From 76c60972692d7ad0492110b5dc5f6b65e75682e0 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 13 Jun 2023 14:16:30 +0200
Subject: [PATCH 001/142] had to comment out pytorch-cuda  otherwise getting
 version error environment.yml

---
 environment.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/environment.yml b/environment.yml
index 77ae5ea0..d47b7e2f 100644
--- a/environment.yml
+++ b/environment.yml
@@ -13,7 +13,7 @@ dependencies:
   - gin-config=0.5.0
   - ignite=0.4.11
   - pytorch=2.0.1
-  - pytorch-cuda=11.8
+  #- pytorch-cuda=11.8
   - lightgbm=3.3.5
   - numpy=1.24.3
   - pandas=2.0.0

From c7657c3e5061fd220dbf3dc2da5f706840282049 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 13 Jun 2023 14:19:47 +0200
Subject: [PATCH 002/142] added how to enviroment variable
 PYTORCH_ENABLE_MPS_FALLBACK=1

---
 README.md | 1 +
 1 file changed, 1 insertion(+)

diff --git a/README.md b/README.md
index 84a7e41b..6f6fffd2 100644
--- a/README.md
+++ b/README.md
@@ -182,6 +182,7 @@ icu-benchmarks train \
 > For a list of available flags, run `icu-benchmarks train -h`.
 
 > Run with `PYTORCH_ENABLE_MPS_FALLBACK=1` on Macs with Metal Performance Shaders.
+> Can set conda enviroment variable by running `conda env config vars set PYTORCH_ENABLE_MPS_FALLBACK=1`
 
 [//]: # (> Please note that, for Windows based systems, paths need to be formatted differently, e.g: ` r"\..\data\mortality_seq\hirid"`.)
 > For Windows based systems, the next line character (\\)  needs to be replaced by (^) (Command Prompt) or (`) (Powershell)

From cf8696ef04d06c86c064b40521a15c49e401564b Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 13 Jun 2023 14:24:40 +0200
Subject: [PATCH 003/142] type casted float64 to float32 to make it work with
 MPS

---
 icu_benchmarks/data/loader.py     | 2 +-
 icu_benchmarks/models/wrappers.py | 8 ++++----
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 9b831aa7..d0e040e4 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -151,7 +151,7 @@ def get_data_and_labels(self) -> Tuple[np.array, np.array]:
 
     def to_tensor(self):
         data, labels = self.get_data_and_labels()
-        return from_numpy(data), from_numpy(labels)
+        return from_numpy(data).to(float32), from_numpy(labels).to(float32)
 
 
 @gin.configurable("ImputationDataset")
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 5fc94a12..867e9874 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -135,7 +135,7 @@ def finalize_step(self, step_prefix=""):
         try:
             self.log_dict(
                 {
-                    f"{step_prefix}/{name}": metric.compute()
+                    f"{step_prefix}/{name}": (np.float32(metric.compute()) if isinstance(metric.compute(), np.float64 ) else metric.compute() )
                     for name, metric in self.metrics[step_prefix].items()
                     if "_Curve" not in name
                 },
@@ -358,9 +358,9 @@ def test_step(self, dataset, _):
         self.set_metrics(test_label)
         test_pred = self.predict(test_rep)
 
-        self.log("test/loss", self.loss(test_label, test_pred), sync_dist=True)
+        self.log("test/loss", np.float32(self.loss(test_label, test_pred)), sync_dist=True)
         logging.debug(f"Test loss: {self.loss(test_label, test_pred)}")
-        self.log_metrics(test_label, test_pred, "test")
+        self.log_metrics(np.float32(test_label), np.float32(test_pred), "test")
 
     def predict(self, features):
         if self.run_mode == RunMode.regression:
@@ -373,7 +373,7 @@ def log_metrics(self, label, pred, metric_type):
 
         self.log_dict(
             {
-                f"{metric_type}/{name}": metric(self.label_transform(label), self.output_transform(pred))
+                f"{metric_type}/{name}": np.float32(metric(self.label_transform(label), self.output_transform(pred)))
                 for name, metric in self.metrics.items()
                 # Filter out metrics that return a tuple (e.g. precision_recall_curve)
                 if not isinstance(metric(self.label_transform(label), self.output_transform(pred)), tuple)

From c08ed2ca108ded995fab682cd96eee9d3c003713 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 20 Jun 2023 16:13:58 +0200
Subject: [PATCH 004/142] added components of tft

---
 icu_benchmarks/models/layers.py | 445 +++++++++++++++++++++++++++++++-
 1 file changed, 444 insertions(+), 1 deletion(-)

diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index c08623bd..969e3ac0 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -1,10 +1,14 @@
+import os
 import gin
 import math
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
 from torch.nn.utils import weight_norm
-
+from torch import Tensor
+from torch.nn.parameter import UninitializedParameter
+from typing import Dict, Tuple, Optional, List
+from torch.nn import LayerNorm
 
 @gin.configurable("masking")
 def parallel_recomb(q_t, kv_t, att_type="all", local_context=3, bin_size=None):
@@ -314,3 +318,442 @@ def forward(self, x):
         out = self.net(x)
         res = x if self.downsample is None else self.downsample(x)
         return self.relu(out + res)
+
+class MaybeLayerNorm(nn.Module):
+    '''
+    Implements layer normalization or identity function depending on output_size
+    '''
+    def __init__(self, output_size, hidden_size, eps):
+        super().__init__()
+        if output_size and output_size == 1:
+            self.ln = nn.Identity()
+        else:
+            self.ln = LayerNorm(output_size if output_size else hidden_size, eps=eps)
+    
+    def forward(self, x):
+        return self.ln(x)
+
+
+class GLU(nn.Module):
+    '''
+    Gated Linear Unit consists of a linear layer followed by a GLU where input is split in half along dim to form a and b
+    GLU(a,b)=a ⊗ σ(b)where σ is signmoid activation and ⊗ is element-wise product 
+    '''
+    def __init__(self, hidden_size, output_size):
+        super().__init__()
+        self.lin = nn.Linear(hidden_size, output_size * 2)
+
+    def forward(self, x: Tensor) -> Tensor:
+        x = self.lin(x)
+        x = F.glu(x)
+        return x
+
+class GRN(nn.Module):
+    '''
+    Gated Residual Network consists of a maybe normalization layer -->linear --> ELU -->linear-->GLU
+     in addition to a residual connection 
+    '''
+    def __init__(self,
+                 input_size,
+                 hidden_size,
+                 output_size=None,
+                 context_hidden_size=None,
+                 dropout=0.0,):
+        super().__init__()
+        self.layer_norm = MaybeLayerNorm(output_size, hidden_size, eps=1e-3)
+        self.lin_a = nn.Linear(input_size, hidden_size)
+        if context_hidden_size is not None:
+            self.lin_c = nn.Linear(context_hidden_size, hidden_size, bias=False)
+        else:
+            self.lin_c = nn.Identity()
+        self.lin_i = nn.Linear(hidden_size, hidden_size)
+        self.glu = GLU(hidden_size, output_size if output_size else hidden_size)
+        self.dropout = nn.Dropout(dropout)
+        self.out_proj = nn.Linear(input_size, output_size) if output_size else None
+
+    def forward(self, a: Tensor, c: Optional[Tensor] = None):
+        x = self.lin_a(a)
+        if c is not None:
+            x = x + self.lin_c(c).unsqueeze(1)
+        x = F.elu(x)
+        x = self.lin_i(x)
+        x = self.dropout(x)
+        x = self.glu(x)
+        y = a if self.out_proj is None else self.out_proj(a)
+        x = x + y
+        return self.layer_norm(x) 
+
+
+# @torch.jit.script #Currently broken with autocast
+def fused_pointwise_linear_v1(x, a, b):
+    out = torch.mul(x.unsqueeze(-1), a)
+    out = out + b
+    return out
+
+@torch.jit.script
+def fused_pointwise_linear_v2(x, a, b):
+    out = x.unsqueeze(3) * a
+    out = out + b
+    return out
+
+
+class TFTEmbedding(nn.Module):
+    def __init__(self,static_categorical_inp_lens,temporal_known_categorical_inp_lens,
+    temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
+    temporal_observed_continuous_inp_size,temporal_target_size,hidden_size,
+     initialize_cont_params=False):
+        # initialize_cont_params=False prevents form initializing parameters inside this class
+        # so they can be lazily initialized in LazyEmbedding module
+        super().__init__()
+        self.s_cat_inp_lens    = static_categorical_inp_lens
+        self.t_cat_k_inp_lens  = temporal_known_categorical_inp_lens
+        self.t_cat_o_inp_lens  = temporal_observed_categorical_inp_lens
+        self.s_cont_inp_size   = static_continuous_inp_size
+        self.t_cont_k_inp_size = temporal_known_continuous_inp_size
+        self.t_cont_o_inp_size = temporal_observed_continuous_inp_size
+        self.t_tgt_size        = temporal_target_size
+
+        self.hidden_size = hidden_size
+
+        # There are 7 types of input:
+        # 1. Static categorical
+        # 2. Static continuous
+        # 3. Temporal known a priori categorical
+        # 4. Temporal known a priori continuous
+        # 5. Temporal observed categorical
+        # 6. Temporal observed continuous
+        # 7. Temporal observed targets (time series obseved so far)
+
+        self.s_cat_embed = nn.ModuleList([
+            nn.Embedding(n, self.hidden_size) for n in self.s_cat_inp_lens]) if self.s_cat_inp_lens else None
+        self.t_cat_k_embed = nn.ModuleList([
+            nn.Embedding(n, self.hidden_size) for n in self.t_cat_k_inp_lens]) if self.t_cat_k_inp_lens else None
+        self.t_cat_o_embed = nn.ModuleList([
+            nn.Embedding(n, self.hidden_size) for n in self.t_cat_o_inp_lens]) if self.t_cat_o_inp_lens else None
+
+        if initialize_cont_params:
+            self.s_cont_embedding_vectors = nn.Parameter(torch.Tensor(self.s_cont_inp_size, self.hidden_size)) if self.s_cont_inp_size else None
+            self.t_cont_k_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_k_inp_size, self.hidden_size)) if self.t_cont_k_inp_size else None
+            self.t_cont_o_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_o_inp_size, self.hidden_size)) if self.t_cont_o_inp_size else None
+            self.t_tgt_embedding_vectors = nn.Parameter(torch.Tensor(self.t_tgt_size, self.hidden_size))
+
+            self.s_cont_embedding_bias = nn.Parameter(torch.zeros(self.s_cont_inp_size, self.hidden_size)) if self.s_cont_inp_size else None
+            self.t_cont_k_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_k_inp_size, self.hidden_size)) if self.t_cont_k_inp_size else None
+            self.t_cont_o_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_o_inp_size, self.hidden_size)) if self.t_cont_o_inp_size else None
+            self.t_tgt_embedding_bias = nn.Parameter(torch.zeros(self.t_tgt_size, self.hidden_size))
+
+            self.reset_parameters()
+
+
+    def reset_parameters(self):
+        ''''
+        embeddings are initilitized using xavier's method and biases are initlitized with zeros
+        '''
+        if self.s_cont_embedding_vectors is not None:
+            torch.nn.init.xavier_normal_(self.s_cont_embedding_vectors)
+            torch.nn.init.zeros_(self.s_cont_embedding_bias)
+        if self.t_cont_k_embedding_vectors is not None:
+            torch.nn.init.xavier_normal_(self.t_cont_k_embedding_vectors)
+            torch.nn.init.zeros_(self.t_cont_k_embedding_bias)
+        if self.t_cont_o_embedding_vectors is not None:
+            torch.nn.init.xavier_normal_(self.t_cont_o_embedding_vectors)
+            torch.nn.init.zeros_(self.t_cont_o_embedding_bias)
+        if self.t_tgt_embedding_vectors is not None:
+            torch.nn.init.xavier_normal_(self.t_tgt_embedding_vectors)
+            torch.nn.init.zeros_(self.t_tgt_embedding_bias)
+        if self.s_cat_embed is not None:
+            for module in self.s_cat_embed:
+                module.reset_parameters()
+        if self.t_cat_k_embed is not None:
+            for module in self.t_cat_k_embed:
+                module.reset_parameters()
+        if self.t_cat_o_embed is not None:
+            for module in self.t_cat_o_embed:
+                module.reset_parameters()
+
+
+    def _apply_embedding(self,
+            cat: Optional[Tensor],
+            cont: Optional[Tensor],
+            cat_emb: Optional[nn.ModuleList], 
+            cont_emb: Tensor,
+            cont_bias: Tensor,
+            ) -> Tuple[Optional[Tensor], Optional[Tensor]]:
+        e_cat = torch.stack([embed(cat[...,i]) for i, embed in enumerate(cat_emb)], dim=-2) if cat is not None else None
+        if cont is not None:
+            #the line below is equivalent to following einsums
+            #e_cont = torch.einsum('btf,fh->bthf', cont, cont_emb)
+            #e_cont = torch.einsum('bf,fh->bhf', cont, cont_emb)
+            
+            e_cont = fused_pointwise_linear_v1(cont, cont_emb, cont_bias)
+        else:
+            e_cont = None
+
+        if e_cat is not None and e_cont is not None:
+            return torch.cat([e_cat, e_cont], dim=-2)
+        elif e_cat is not None:
+            return e_cat
+        elif e_cont is not None:
+            return e_cont
+        else:
+            return None
+
+    def forward(self, x: Dict[str, Tensor]):
+        # temporal/static categorical/continuous known/observed input 
+        s_cat_inp = x.get('s_cat', None)
+        s_cont_inp = x.get('s_cont', None)
+        t_cat_k_inp = x.get('k_cat', None)
+        t_cont_k_inp = x.get('k_cont', None)
+        t_cat_o_inp = x.get('o_cat', None)
+        t_cont_o_inp = x.get('o_cont', None)
+        t_tgt_obs = x['target'] # Has to be present
+
+        # Static inputs are expected to be equal for all timesteps
+        # For memory efficiency there is no assert statement
+        s_cat_inp = s_cat_inp[:,0,:] if s_cat_inp is not None else None
+        s_cont_inp = s_cont_inp[:,0,:] if s_cont_inp is not None else None
+
+        s_inp = self._apply_embedding(s_cat_inp,
+                                      s_cont_inp,
+                                      self.s_cat_embed,
+                                      self.s_cont_embedding_vectors,
+                                      self.s_cont_embedding_bias)
+        t_known_inp = self._apply_embedding(t_cat_k_inp,
+                                            t_cont_k_inp,
+                                            self.t_cat_k_embed,
+                                            self.t_cont_k_embedding_vectors,
+                                            self.t_cont_k_embedding_bias)
+        t_observed_inp = self._apply_embedding(t_cat_o_inp,
+                                               t_cont_o_inp,
+                                               self.t_cat_o_embed,
+                                               self.t_cont_o_embedding_vectors,
+                                               self.t_cont_o_embedding_bias)
+
+        # Temporal observed targets
+        t_observed_tgt = fused_pointwise_linear_v2(t_tgt_obs, self.t_tgt_embedding_vectors, self.t_tgt_embedding_bias)
+
+        return s_inp, t_known_inp, t_observed_inp, t_observed_tgt
+
+class LazyEmbedding(nn.modules.lazy.LazyModuleMixin, TFTEmbedding):
+    cls_to_become = TFTEmbedding
+
+    def __init__(self, static_categorical_inp_lens,temporal_known_categorical_inp_lens,
+    temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
+    temporal_observed_continuous_inp_size,temporal_target_size,hidden_size):
+        super().__init__(static_categorical_inp_lens,temporal_known_categorical_inp_lens,
+    temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
+    temporal_observed_continuous_inp_size,temporal_target_size,hidden_size, initialize_cont_params=False)
+
+        if static_continuous_inp_size:
+            self.s_cont_embedding_vectors = UninitializedParameter()
+            self.s_cont_embedding_bias = UninitializedParameter()
+        else:
+            self.s_cont_embedding_vectors = None
+            self.s_cont_embedding_bias = None
+
+        if temporal_known_continuous_inp_size:
+            self.t_cont_k_embedding_vectors = UninitializedParameter()
+            self.t_cont_k_embedding_bias = UninitializedParameter()
+        else:
+            self.t_cont_k_embedding_vectors = None
+            self.t_cont_k_embedding_bias = None
+
+        if temporal_observed_continuous_inp_size:
+            self.t_cont_o_embedding_vectors = UninitializedParameter()
+            self.t_cont_o_embedding_bias = UninitializedParameter()
+        else:
+            self.t_cont_o_embedding_vectors = None
+            self.t_cont_o_embedding_bias = None
+
+        self.t_tgt_embedding_vectors = UninitializedParameter()
+        self.t_tgt_embedding_bias = UninitializedParameter()
+
+    def initialize_parameters(self, x):
+        if self.has_uninitialized_params():
+            s_cont_inp = x.get('s_cont', None)
+            t_cont_k_inp = x.get('k_cont', None)
+            t_cont_o_inp = x.get('o_cont', None)
+            t_tgt_obs = x['target'] # Has to be present
+
+            if s_cont_inp is not None:
+                self.s_cont_embedding_vectors.materialize((s_cont_inp.shape[-1], self.hidden_size))
+                self.s_cont_embedding_bias.materialize((s_cont_inp.shape[-1], self.hidden_size))
+
+            if t_cont_k_inp is not None:
+                self.t_cont_k_embedding_vectors.materialize((t_cont_k_inp.shape[-1], self.hidden_size))
+                self.t_cont_k_embedding_bias.materialize((t_cont_k_inp.shape[-1], self.hidden_size))
+
+            if t_cont_o_inp is not None:
+                self.t_cont_o_embedding_vectors.materialize((t_cont_o_inp.shape[-1], self.hidden_size))
+                self.t_cont_o_embedding_bias.materialize((t_cont_o_inp.shape[-1], self.hidden_size))
+
+            self.t_tgt_embedding_vectors.materialize((t_tgt_obs.shape[-1], self.hidden_size))
+            self.t_tgt_embedding_bias.materialize((t_tgt_obs.shape[-1], self.hidden_size))
+
+            self.reset_parameters()
+
+class VariableSelectionNetwork(nn.Module):
+    '''
+    Learns to select important netowrks consists of GRNs with one GRN for variable weights 
+    and the others for input embedding
+    '''
+    def __init__(self, hidden_size,dropout, num_inputs):
+        super().__init__()
+        self.joint_grn = GRN(hidden_size*num_inputs, hidden_size, output_size=num_inputs, context_hidden_size=hidden_size)
+        self.var_grns = nn.ModuleList([GRN(hidden_size, hidden_size, dropout=dropout) for _ in range(num_inputs)])
+
+    def forward(self, x: Tensor, context: Optional[Tensor] = None):
+        Xi = torch.flatten(x, start_dim=-2)
+        grn_outputs = self.joint_grn(Xi, c=context)
+        sparse_weights = F.softmax(grn_outputs, dim=-1)
+        transformed_embed_list = [m(x[...,i,:]) for i, m in enumerate(self.var_grns)]
+        transformed_embed = torch.stack(transformed_embed_list, dim=-1)
+        #the line below performs batched matrix vector multiplication
+        #for temporal features it's bthf,btf->bth
+        #for static features it's bhf,bf->bh
+        variable_ctx = torch.matmul(transformed_embed, sparse_weights.unsqueeze(-1)).squeeze(-1)
+
+        return variable_ctx, sparse_weights
+
+class StaticCovariateEncoder(nn.Module):
+    '''
+    Network to produce 4 contexts vectors to enrich static variables 
+    Vriable selection Network --> GRNs
+    '''
+    def __init__(self, num_static_vars,hidden_size,dropout):
+        super().__init__()
+        self.vsn = VariableSelectionNetwork(hidden_size,dropout, num_static_vars)
+        self.context_grns = nn.ModuleList([GRN(hidden_size, hidden_size, dropout=dropout) for _ in range(4)])
+
+    def forward(self, x: Tensor) -> Tuple[Tensor, Tensor, Tensor, Tensor]:
+        variable_ctx, sparse_weights = self.vsn(x)
+
+        # Context vectors:
+        # variable selection context
+        # enrichment context
+        # state_c context
+        # state_h context
+        cs, ce, ch, cc = [m(variable_ctx) for m in self.context_grns]
+
+        return cs, ce, ch, cc
+
+
+class InterpretableMultiHeadAttention(nn.Module):
+    '''
+    Multi-head attention different as it outputs the attention_probability and it combines the attention weights instead of 
+    concating them different from the one implemented already in YAIB
+    '''
+    def __init__(self,n_head,hidden_size,attn_dropout,dropout,example_length,d_head):
+        super().__init__()
+        self.n_head = n_head
+        assert hidden_size % n_head == 0
+        self.d_head = hidden_size // n_head
+        self.qkv_linears = nn.Linear(hidden_size, (2 * n_head + 1) * d_head, bias=False)
+        self.out_proj = nn.Linear(self.d_head, hidden_size, bias=False)
+        self.attn_dropout = nn.Dropout(attn_dropout)
+        self.out_dropout = nn.Dropout(dropout)
+        self.scale = self.d_head**-0.5
+        self.register_buffer("_mask", torch.triu(torch.full((example_length, example_length), float('-inf')), 1).unsqueeze(0))
+
+    def forward(self, x: Tensor) -> Tuple[Tensor, Tensor]:
+        bs, t, h_size = x.shape
+        qkv = self.qkv_linears(x)
+        q, k, v = qkv.split((self.n_head * self.d_head, self.n_head * self.d_head, self.d_head), dim=-1)
+        q = q.view(bs, t, self.n_head, self.d_head)
+        k = k.view(bs, t, self.n_head, self.d_head)
+        v = v.view(bs, t, self.d_head)
+
+        # attn_score = torch.einsum('bind,bjnd->bnij', q, k)
+        attn_score = torch.matmul(q.permute((0, 2, 1, 3)), k.permute((0, 2, 3, 1)))
+        attn_score.mul_(self.scale)
+
+        attn_score = attn_score + self._mask
+
+        attn_prob = F.softmax(attn_score, dim=3)
+        attn_prob = self.attn_dropout(attn_prob)
+
+        # attn_vec = torch.einsum('bnij,bjd->bnid', attn_prob, v)
+        attn_vec = torch.matmul(attn_prob, v.unsqueeze(1))
+        m_attn_vec = torch.mean(attn_vec, dim=1)
+        out = self.out_proj(m_attn_vec)
+        out = self.out_dropout(out)
+
+        return out, attn_prob
+
+class TFTBack(nn.Module):
+    '''
+    Big part of TFT architecture consists of static enrichment followed by mutli-head self-attention then position wise feed forward
+    followed by a gate and a dense layer 
+    GRNs-->multi-head attention-->GRNs-->GLU-->Linear-->output
+    '''
+    def __init__(self, encoder_length,num_historic_vars,hidden_size,dropout,num_future_vars,
+                 n_head,attn_dropout,example_length,d_head,quantiles):
+        super().__init__()
+
+        self.encoder_length = encoder_length
+        self.history_vsn = VariableSelectionNetwork(hidden_size,dropout, num_historic_vars)
+        self.history_encoder = nn.LSTM(hidden_size, hidden_size, batch_first=True)
+        self.future_vsn = VariableSelectionNetwork(hidden_size,dropout,num_future_vars)
+        self.future_encoder = nn.LSTM(hidden_size, hidden_size, batch_first=True)
+
+
+        self.input_gate = GLU(hidden_size, hidden_size)
+        self.input_gate_ln = LayerNorm(hidden_size, eps=1e-3)
+
+        self.enrichment_grn = GRN(hidden_size,
+                                  hidden_size,
+                                  context_hidden_size=hidden_size, 
+                                  dropout=dropout)
+        self.attention = InterpretableMultiHeadAttention(n_head,hidden_size,attn_dropout,dropout,example_length,d_head)
+        self.attention_gate = GLU(hidden_size, hidden_size)
+        self.attention_ln = LayerNorm(hidden_size, eps=1e-3)
+
+        self.positionwise_grn = GRN(hidden_size,
+                                    hidden_size,
+                                    dropout=dropout)
+
+        self.decoder_gate = GLU(hidden_size, hidden_size)
+        self.decoder_ln = LayerNorm(hidden_size, eps=1e-3)
+
+        self.quantile_proj = nn.Linear(hidden_size, len(quantiles))
+        
+    def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
+        historical_features, _ = self.history_vsn(historical_inputs, cs)
+        history, state = self.history_encoder(historical_features, (ch, cc))
+        future_features, _ = self.future_vsn(future_inputs, cs)
+        future, _ = self.future_encoder(future_features, state)
+        torch.cuda.synchronize()
+
+        # skip connection
+        input_embedding = torch.cat([historical_features, future_features], dim=1)
+        temporal_features = torch.cat([history, future], dim=1)
+        temporal_features = self.input_gate(temporal_features)
+        temporal_features = temporal_features + input_embedding
+        temporal_features = self.input_gate_ln(temporal_features)
+
+        # Static enrichment
+        enriched = self.enrichment_grn(temporal_features, c=ce)
+
+        # Temporal self attention
+        x, _ = self.attention(enriched)
+
+        # Don't compute hictorical quantiles
+        x = x[:, self.encoder_length:, :]
+        temporal_features = temporal_features[:, self.encoder_length:, :]
+        enriched = enriched[:, self.encoder_length:, :]
+
+        x = self.attention_gate(x)
+        x = x + enriched
+        x = self.attention_ln(x)
+
+        # Position-wise feed-forward
+        x = self.positionwise_grn(x)
+
+        # Final skip connection
+        x = self.decoder_gate(x)
+        x = x + temporal_features
+        x = self.decoder_ln(x)
+
+        out = self.quantile_proj(x)
+
+        return out
\ No newline at end of file

From 0a5caff5d02e5ffb022b55c1d745be7dc71219cd Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 20 Jun 2023 16:14:16 +0200
Subject: [PATCH 005/142] added  tft

---
 icu_benchmarks/models/dl_models.py | 40 +++++++++++++++++++++++++++++-
 1 file changed, 39 insertions(+), 1 deletion(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 0fb1b0d2..39eac551 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -3,7 +3,8 @@
 import numpy as np
 import torch.nn as nn
 from icu_benchmarks.contants import RunMode
-from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding
+from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack
+
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 
 
@@ -280,3 +281,40 @@ def forward(self, x):
         o = o.permute(0, 2, 1)  # Permute to channel last
         pred = self.logit(o)
         return pred
+@gin.configurable
+class TemporalFusionTransformer(DLPredictionWrapper):
+    """ 
+    Implementation of https://arxiv.org/abs/1912.09363 
+    """
+    def __init__(self, encoder_length,static_categorical_inp_lens,temporal_known_categorical_inp_lens,
+    temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
+    temporal_observed_continuous_inp_size,temporal_target_size,hidden_size,num_static_vars,dropout,num_historic_vars,num_future_vars,
+                 n_head,attn_dropout,example_length,d_head,quantiles):
+        super().__init__()
+
+
+
+        self.encoder_length = encoder_length #this determines from how distant past we want to use data from
+
+        self.embedding = LazyEmbedding(static_categorical_inp_lens,temporal_known_categorical_inp_lens,
+                temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
+                temporal_observed_continuous_inp_size,temporal_target_size,hidden_size)
+        self.static_encoder = StaticCovariateEncoder(num_static_vars,hidden_size,dropout)
+        self.TFTpart2 = torch.jit.script(TFTBack(encoder_length,num_historic_vars,hidden_size,dropout,num_future_vars,
+                n_head,attn_dropout,example_length,d_head,quantiles))
+
+    def forward(self, x: Dict[str, Tensor]) -> Tensor:
+        s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
+
+        # Static context
+        cs, ce, ch, cc = self.static_encoder(s_inp)
+        ch, cc = ch.unsqueeze(0), cc.unsqueeze(0) #lstm initial states
+
+        # Temporal input
+        _historical_inputs = [t_known_inp[:,:self.encoder_length,:], t_observed_tgt[:,:self.encoder_length,:]]
+        if t_observed_inp is not None:
+            _historical_inputs.insert(0,t_observed_inp[:,:self.encoder_length,:])
+
+        historical_inputs = torch.cat(_historical_inputs, dim=-2)
+        future_inputs = t_known_inp[:, self.encoder_length:]
+        return self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs)
\ No newline at end of file

From 52d0daa3fb9b6d67cdc159c3f2250e481bb65637 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Thu, 22 Jun 2023 13:55:47 +0200
Subject: [PATCH 006/142] setting variables for tft and tuning

---
 .../TemporalFusionTransformer.gin             | 26 +++++++++++++++++++
 1 file changed, 26 insertions(+)
 create mode 100644 configs/prediction_models/TemporalFusionTransformer.gin

diff --git a/configs/prediction_models/TemporalFusionTransformer.gin b/configs/prediction_models/TemporalFusionTransformer.gin
new file mode 100644
index 00000000..d430a443
--- /dev/null
+++ b/configs/prediction_models/TemporalFusionTransformer.gin
@@ -0,0 +1,26 @@
+# Settings for Transformer model.
+
+# Common settings for DL models
+include "configs/prediction_models/common/DLCommon.gin"
+
+# Optimizer params
+train_common.model = @TemporalFusionTransformer
+
+optimizer/hyperparameter.class_to_tune = @Adam
+optimizer/hyperparameter.weight_decay = 1e-6
+optimizer/hyperparameter.lr = (1e-5, 3e-4)
+
+# Encoder params
+model/hyperparameter.class_to_tune = @TemporalFusionTransformer
+model/hyperparameter.encoder_length = (1,14,"log-uniform", 2)
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.num_classes = %NUM_CLASSES
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.depth = (1, 3)
+model/hyperparameter.dropout = (0.0, 0.4)
+model/hyperparameter.dropout_att = (0.0, 0.4)
+model/hyperparameter.n_heads = (1, 8, "log-uniform", 2)
+model/hyperparameter.quantiles=[0.1, 0.5, 0.9]
+model/hyperparameter.example_length=(1,14,"log-uniform", 2)
+
+

From 8e812a0010db7987120da6d1cb7cf364799b7704 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Thu, 22 Jun 2023 13:56:27 +0200
Subject: [PATCH 007/142] modifications to fit in the tft and comments

---
 icu_benchmarks/models/dl_models.py |  36 ++++---
 icu_benchmarks/models/layers.py    | 149 +++++++++++++++--------------
 2 files changed, 98 insertions(+), 87 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 39eac551..4646502e 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -4,8 +4,9 @@
 import torch.nn as nn
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack
-
+from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
+from torch import Tensor,cat
 
 
 @gin.configurable
@@ -286,22 +287,31 @@ class TemporalFusionTransformer(DLPredictionWrapper):
     """ 
     Implementation of https://arxiv.org/abs/1912.09363 
     """
-    def __init__(self, encoder_length,static_categorical_inp_lens,temporal_known_categorical_inp_lens,
-    temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
-    temporal_observed_continuous_inp_size,temporal_target_size,hidden_size,num_static_vars,dropout,num_historic_vars,num_future_vars,
-                 n_head,attn_dropout,example_length,d_head,quantiles):
-        super().__init__()
 
 
+    _supported_run_modes = [RunMode.classification, RunMode.regression]
+    def __init__(self, encoder_length,hidden,num_static_vars,dropout,num_historic_vars,num_future_vars,
+                 n_head,dropout_att,example_length,quantiles,static_categorical_inp_size=1,temporal_known_categorical_inp_size=4,
+    temporal_observed_categorical_inp_size=0,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
+    temporal_observed_continuous_inp_size=48,temporal_target_size=1):
+        super().__init__()
 
+        #dervied embeddings size
+        num_static_vars=static_categorical_inp_size+static_continuous_inp_size
+        num_future_vars=temporal_known_categorical_inp_size+temporal_known_continuous_inp_size
+        num_historic_vars=sum([num_future_vars,
+                                      temporal_observed_continuous_inp_size,
+                                      temporal_target_size,
+                                      temporal_observed_categorical_inp_size,
+                                      ])
         self.encoder_length = encoder_length #this determines from how distant past we want to use data from
 
-        self.embedding = LazyEmbedding(static_categorical_inp_lens,temporal_known_categorical_inp_lens,
-                temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
-                temporal_observed_continuous_inp_size,temporal_target_size,hidden_size)
-        self.static_encoder = StaticCovariateEncoder(num_static_vars,hidden_size,dropout)
-        self.TFTpart2 = torch.jit.script(TFTBack(encoder_length,num_historic_vars,hidden_size,dropout,num_future_vars,
-                n_head,attn_dropout,example_length,d_head,quantiles))
+        self.embedding = LazyEmbedding(static_categorical_inp_size,temporal_known_categorical_inp_size,
+                temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
+                temporal_observed_continuous_inp_size,temporal_target_size,hidden)
+        self.static_encoder = StaticCovariateEncoder(num_static_vars,hidden,dropout)
+        self.TFTpart2 = torch.jit.script(TFTBack(encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
+                n_head,dropout_att,example_length,quantiles))
 
     def forward(self, x: Dict[str, Tensor]) -> Tensor:
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
@@ -315,6 +325,6 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         if t_observed_inp is not None:
             _historical_inputs.insert(0,t_observed_inp[:,:self.encoder_length,:])
 
-        historical_inputs = torch.cat(_historical_inputs, dim=-2)
+        historical_inputs = cat(_historical_inputs, dim=-2)
         future_inputs = t_known_inp[:, self.encoder_length:]
         return self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs)
\ No newline at end of file
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 969e3ac0..89201db5 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -323,12 +323,12 @@ class MaybeLayerNorm(nn.Module):
     '''
     Implements layer normalization or identity function depending on output_size
     '''
-    def __init__(self, output_size, hidden_size, eps):
+    def __init__(self, output_size, hidden, eps):
         super().__init__()
         if output_size and output_size == 1:
             self.ln = nn.Identity()
         else:
-            self.ln = LayerNorm(output_size if output_size else hidden_size, eps=eps)
+            self.ln = LayerNorm(output_size if output_size else hidden, eps=eps)
     
     def forward(self, x):
         return self.ln(x)
@@ -339,9 +339,9 @@ class GLU(nn.Module):
     Gated Linear Unit consists of a linear layer followed by a GLU where input is split in half along dim to form a and b
     GLU(a,b)=a ⊗ σ(b)where σ is signmoid activation and ⊗ is element-wise product 
     '''
-    def __init__(self, hidden_size, output_size):
+    def __init__(self, hidden, output_size):
         super().__init__()
-        self.lin = nn.Linear(hidden_size, output_size * 2)
+        self.lin = nn.Linear(hidden, output_size * 2)
 
     def forward(self, x: Tensor) -> Tensor:
         x = self.lin(x)
@@ -355,19 +355,19 @@ class GRN(nn.Module):
     '''
     def __init__(self,
                  input_size,
-                 hidden_size,
+                 hidden,
                  output_size=None,
                  context_hidden_size=None,
                  dropout=0.0,):
         super().__init__()
-        self.layer_norm = MaybeLayerNorm(output_size, hidden_size, eps=1e-3)
-        self.lin_a = nn.Linear(input_size, hidden_size)
+        self.layer_norm = MaybeLayerNorm(output_size, hidden, eps=1e-3)
+        self.lin_a = nn.Linear(input_size, hidden)
         if context_hidden_size is not None:
-            self.lin_c = nn.Linear(context_hidden_size, hidden_size, bias=False)
+            self.lin_c = nn.Linear(context_hidden_size, hidden, bias=False)
         else:
             self.lin_c = nn.Identity()
-        self.lin_i = nn.Linear(hidden_size, hidden_size)
-        self.glu = GLU(hidden_size, output_size if output_size else hidden_size)
+        self.lin_i = nn.Linear(hidden, hidden)
+        self.glu = GLU(hidden, output_size if output_size else hidden)
         self.dropout = nn.Dropout(dropout)
         self.out_proj = nn.Linear(input_size, output_size) if output_size else None
 
@@ -398,22 +398,23 @@ def fused_pointwise_linear_v2(x, a, b):
 
 
 class TFTEmbedding(nn.Module):
-    def __init__(self,static_categorical_inp_lens,temporal_known_categorical_inp_lens,
-    temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
-    temporal_observed_continuous_inp_size,temporal_target_size,hidden_size,
+    def __init__(self,static_categorical_inp_size,temporal_known_categorical_inp_size,
+    temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
+    temporal_observed_continuous_inp_size,temporal_target_size,hidden,
      initialize_cont_params=False):
         # initialize_cont_params=False prevents form initializing parameters inside this class
         # so they can be lazily initialized in LazyEmbedding module
         super().__init__()
-        self.s_cat_inp_lens    = static_categorical_inp_lens
-        self.t_cat_k_inp_lens  = temporal_known_categorical_inp_lens
-        self.t_cat_o_inp_lens  = temporal_observed_categorical_inp_lens
+        #these are basically number of varaibales that falls under each category 
+        self.s_cat_inp_size    = static_categorical_inp_size
+        self.t_cat_k_inp_size  = temporal_known_categorical_inp_size
+        self.t_cat_o_inp_size  = temporal_observed_categorical_inp_size
         self.s_cont_inp_size   = static_continuous_inp_size
         self.t_cont_k_inp_size = temporal_known_continuous_inp_size
         self.t_cont_o_inp_size = temporal_observed_continuous_inp_size
         self.t_tgt_size        = temporal_target_size
 
-        self.hidden_size = hidden_size
+        self.hidden = hidden
 
         # There are 7 types of input:
         # 1. Static categorical
@@ -425,22 +426,22 @@ def __init__(self,static_categorical_inp_lens,temporal_known_categorical_inp_len
         # 7. Temporal observed targets (time series obseved so far)
 
         self.s_cat_embed = nn.ModuleList([
-            nn.Embedding(n, self.hidden_size) for n in self.s_cat_inp_lens]) if self.s_cat_inp_lens else None
+            nn.Embedding(n, self.hidden) for n in self.s_cat_inp_size]) if self.s_cat_inp_size else None
         self.t_cat_k_embed = nn.ModuleList([
-            nn.Embedding(n, self.hidden_size) for n in self.t_cat_k_inp_lens]) if self.t_cat_k_inp_lens else None
+            nn.Embedding(n, self.hidden) for n in self.t_cat_k_inp_size]) if self.t_cat_k_inp_size else None
         self.t_cat_o_embed = nn.ModuleList([
-            nn.Embedding(n, self.hidden_size) for n in self.t_cat_o_inp_lens]) if self.t_cat_o_inp_lens else None
+            nn.Embedding(n, self.hidden) for n in self.t_cat_o_inp_size]) if self.t_cat_o_inp_size else None
 
         if initialize_cont_params:
-            self.s_cont_embedding_vectors = nn.Parameter(torch.Tensor(self.s_cont_inp_size, self.hidden_size)) if self.s_cont_inp_size else None
-            self.t_cont_k_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_k_inp_size, self.hidden_size)) if self.t_cont_k_inp_size else None
-            self.t_cont_o_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_o_inp_size, self.hidden_size)) if self.t_cont_o_inp_size else None
-            self.t_tgt_embedding_vectors = nn.Parameter(torch.Tensor(self.t_tgt_size, self.hidden_size))
+            self.s_cont_embedding_vectors = nn.Parameter(torch.Tensor(self.s_cont_inp_size, self.hidden)) if self.s_cont_inp_size else None
+            self.t_cont_k_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_k_inp_size, self.hidden)) if self.t_cont_k_inp_size else None
+            self.t_cont_o_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_o_inp_size, self.hidden)) if self.t_cont_o_inp_size else None
+            self.t_tgt_embedding_vectors = nn.Parameter(torch.Tensor(self.t_tgt_size, self.hidden))
 
-            self.s_cont_embedding_bias = nn.Parameter(torch.zeros(self.s_cont_inp_size, self.hidden_size)) if self.s_cont_inp_size else None
-            self.t_cont_k_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_k_inp_size, self.hidden_size)) if self.t_cont_k_inp_size else None
-            self.t_cont_o_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_o_inp_size, self.hidden_size)) if self.t_cont_o_inp_size else None
-            self.t_tgt_embedding_bias = nn.Parameter(torch.zeros(self.t_tgt_size, self.hidden_size))
+            self.s_cont_embedding_bias = nn.Parameter(torch.zeros(self.s_cont_inp_size, self.hidden)) if self.s_cont_inp_size else None
+            self.t_cont_k_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_k_inp_size, self.hidden)) if self.t_cont_k_inp_size else None
+            self.t_cont_o_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_o_inp_size, self.hidden)) if self.t_cont_o_inp_size else None
+            self.t_tgt_embedding_bias = nn.Parameter(torch.zeros(self.t_tgt_size, self.hidden))
 
             self.reset_parameters()
 
@@ -537,12 +538,12 @@ def forward(self, x: Dict[str, Tensor]):
 class LazyEmbedding(nn.modules.lazy.LazyModuleMixin, TFTEmbedding):
     cls_to_become = TFTEmbedding
 
-    def __init__(self, static_categorical_inp_lens,temporal_known_categorical_inp_lens,
-    temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
-    temporal_observed_continuous_inp_size,temporal_target_size,hidden_size):
-        super().__init__(static_categorical_inp_lens,temporal_known_categorical_inp_lens,
-    temporal_observed_categorical_inp_lens,static_continuous_inp_size,temporal_known_continuous_inp_size,
-    temporal_observed_continuous_inp_size,temporal_target_size,hidden_size, initialize_cont_params=False)
+    def __init__(self, static_categorical_inp_size,temporal_known_categorical_inp_size,
+    temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
+    temporal_observed_continuous_inp_size,temporal_target_size,hidden):
+        super().__init__(static_categorical_inp_size,temporal_known_categorical_inp_size,
+    temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
+    temporal_observed_continuous_inp_size,temporal_target_size,hidden, initialize_cont_params=False)
 
         if static_continuous_inp_size:
             self.s_cont_embedding_vectors = UninitializedParameter()
@@ -576,19 +577,19 @@ def initialize_parameters(self, x):
             t_tgt_obs = x['target'] # Has to be present
 
             if s_cont_inp is not None:
-                self.s_cont_embedding_vectors.materialize((s_cont_inp.shape[-1], self.hidden_size))
-                self.s_cont_embedding_bias.materialize((s_cont_inp.shape[-1], self.hidden_size))
+                self.s_cont_embedding_vectors.materialize((s_cont_inp.shape[-1], self.hidden))
+                self.s_cont_embedding_bias.materialize((s_cont_inp.shape[-1], self.hidden))
 
             if t_cont_k_inp is not None:
-                self.t_cont_k_embedding_vectors.materialize((t_cont_k_inp.shape[-1], self.hidden_size))
-                self.t_cont_k_embedding_bias.materialize((t_cont_k_inp.shape[-1], self.hidden_size))
+                self.t_cont_k_embedding_vectors.materialize((t_cont_k_inp.shape[-1], self.hidden))
+                self.t_cont_k_embedding_bias.materialize((t_cont_k_inp.shape[-1], self.hidden))
 
             if t_cont_o_inp is not None:
-                self.t_cont_o_embedding_vectors.materialize((t_cont_o_inp.shape[-1], self.hidden_size))
-                self.t_cont_o_embedding_bias.materialize((t_cont_o_inp.shape[-1], self.hidden_size))
+                self.t_cont_o_embedding_vectors.materialize((t_cont_o_inp.shape[-1], self.hidden))
+                self.t_cont_o_embedding_bias.materialize((t_cont_o_inp.shape[-1], self.hidden))
 
-            self.t_tgt_embedding_vectors.materialize((t_tgt_obs.shape[-1], self.hidden_size))
-            self.t_tgt_embedding_bias.materialize((t_tgt_obs.shape[-1], self.hidden_size))
+            self.t_tgt_embedding_vectors.materialize((t_tgt_obs.shape[-1], self.hidden))
+            self.t_tgt_embedding_bias.materialize((t_tgt_obs.shape[-1], self.hidden))
 
             self.reset_parameters()
 
@@ -597,10 +598,10 @@ class VariableSelectionNetwork(nn.Module):
     Learns to select important netowrks consists of GRNs with one GRN for variable weights 
     and the others for input embedding
     '''
-    def __init__(self, hidden_size,dropout, num_inputs):
+    def __init__(self, hidden,dropout, num_inputs):
         super().__init__()
-        self.joint_grn = GRN(hidden_size*num_inputs, hidden_size, output_size=num_inputs, context_hidden_size=hidden_size)
-        self.var_grns = nn.ModuleList([GRN(hidden_size, hidden_size, dropout=dropout) for _ in range(num_inputs)])
+        self.joint_grn = GRN(hidden*num_inputs, hidden, output_size=num_inputs, context_hidden=hidden)
+        self.var_grns = nn.ModuleList([GRN(hidden, hidden, dropout=dropout) for _ in range(num_inputs)])
 
     def forward(self, x: Tensor, context: Optional[Tensor] = None):
         Xi = torch.flatten(x, start_dim=-2)
@@ -620,10 +621,10 @@ class StaticCovariateEncoder(nn.Module):
     Network to produce 4 contexts vectors to enrich static variables 
     Vriable selection Network --> GRNs
     '''
-    def __init__(self, num_static_vars,hidden_size,dropout):
+    def __init__(self, num_static_vars,hidden,dropout):
         super().__init__()
-        self.vsn = VariableSelectionNetwork(hidden_size,dropout, num_static_vars)
-        self.context_grns = nn.ModuleList([GRN(hidden_size, hidden_size, dropout=dropout) for _ in range(4)])
+        self.vsn = VariableSelectionNetwork(hidden,dropout, num_static_vars)
+        self.context_grns = nn.ModuleList([GRN(hidden, hidden, dropout=dropout) for _ in range(4)])
 
     def forward(self, x: Tensor) -> Tuple[Tensor, Tensor, Tensor, Tensor]:
         variable_ctx, sparse_weights = self.vsn(x)
@@ -643,14 +644,14 @@ class InterpretableMultiHeadAttention(nn.Module):
     Multi-head attention different as it outputs the attention_probability and it combines the attention weights instead of 
     concating them different from the one implemented already in YAIB
     '''
-    def __init__(self,n_head,hidden_size,attn_dropout,dropout,example_length,d_head):
+    def __init__(self,n_head,hidden,dropout_att,dropout,example_length):
         super().__init__()
         self.n_head = n_head
-        assert hidden_size % n_head == 0
-        self.d_head = hidden_size // n_head
-        self.qkv_linears = nn.Linear(hidden_size, (2 * n_head + 1) * d_head, bias=False)
-        self.out_proj = nn.Linear(self.d_head, hidden_size, bias=False)
-        self.attn_dropout = nn.Dropout(attn_dropout)
+        assert hidden % n_head == 0
+        self.d_head = hidden // n_head
+        self.qkv_linears = nn.Linear(hidden, (2 * n_head + 1) * self.d_head, bias=False)
+        self.out_proj = nn.Linear(self.d_head, hidden, bias=False)
+        self.dropout_att = nn.Dropout(dropout_att)
         self.out_dropout = nn.Dropout(dropout)
         self.scale = self.d_head**-0.5
         self.register_buffer("_mask", torch.triu(torch.full((example_length, example_length), float('-inf')), 1).unsqueeze(0))
@@ -670,7 +671,7 @@ def forward(self, x: Tensor) -> Tuple[Tensor, Tensor]:
         attn_score = attn_score + self._mask
 
         attn_prob = F.softmax(attn_score, dim=3)
-        attn_prob = self.attn_dropout(attn_prob)
+        attn_prob = self.dropout_att(attn_prob)
 
         # attn_vec = torch.einsum('bnij,bjd->bnid', attn_prob, v)
         attn_vec = torch.matmul(attn_prob, v.unsqueeze(1))
@@ -686,36 +687,36 @@ class TFTBack(nn.Module):
     followed by a gate and a dense layer 
     GRNs-->multi-head attention-->GRNs-->GLU-->Linear-->output
     '''
-    def __init__(self, encoder_length,num_historic_vars,hidden_size,dropout,num_future_vars,
-                 n_head,attn_dropout,example_length,d_head,quantiles):
+    def __init__(self, encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
+                 n_head,dropout_att,example_length,quantiles):
         super().__init__()
 
         self.encoder_length = encoder_length
-        self.history_vsn = VariableSelectionNetwork(hidden_size,dropout, num_historic_vars)
-        self.history_encoder = nn.LSTM(hidden_size, hidden_size, batch_first=True)
-        self.future_vsn = VariableSelectionNetwork(hidden_size,dropout,num_future_vars)
-        self.future_encoder = nn.LSTM(hidden_size, hidden_size, batch_first=True)
+        self.history_vsn = VariableSelectionNetwork(hidden,dropout, num_historic_vars)
+        self.history_encoder = nn.LSTM(hidden, hidden, batch_first=True)
+        self.future_vsn = VariableSelectionNetwork(hidden,dropout,num_future_vars)
+        self.future_encoder = nn.LSTM(hidden, hidden, batch_first=True)
 
 
-        self.input_gate = GLU(hidden_size, hidden_size)
-        self.input_gate_ln = LayerNorm(hidden_size, eps=1e-3)
+        self.input_gate = GLU(hidden, hidden)
+        self.input_gate_ln = LayerNorm(hidden, eps=1e-3)
 
-        self.enrichment_grn = GRN(hidden_size,
-                                  hidden_size,
-                                  context_hidden_size=hidden_size, 
+        self.enrichment_grn = GRN(hidden,
+                                  hidden,
+                                  context_hidden_size=hidden, 
                                   dropout=dropout)
-        self.attention = InterpretableMultiHeadAttention(n_head,hidden_size,attn_dropout,dropout,example_length,d_head)
-        self.attention_gate = GLU(hidden_size, hidden_size)
-        self.attention_ln = LayerNorm(hidden_size, eps=1e-3)
+        self.attention = InterpretableMultiHeadAttention(n_head,hidden,dropout_att,dropout,example_length)
+        self.attention_gate = GLU(hidden, hidden)
+        self.attention_ln = LayerNorm(hidden, eps=1e-3)
 
-        self.positionwise_grn = GRN(hidden_size,
-                                    hidden_size,
+        self.positionwise_grn = GRN(hidden,
+                                    hidden,
                                     dropout=dropout)
 
-        self.decoder_gate = GLU(hidden_size, hidden_size)
-        self.decoder_ln = LayerNorm(hidden_size, eps=1e-3)
+        self.decoder_gate = GLU(hidden, hidden)
+        self.decoder_ln = LayerNorm(hidden, eps=1e-3)
 
-        self.quantile_proj = nn.Linear(hidden_size, len(quantiles))
+        self.quantile_proj = nn.Linear(hidden, len(quantiles))
         
     def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         historical_features, _ = self.history_vsn(historical_inputs, cs)

From 92404d530ece301866dc83c1d6a475289cc1ad1a Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Thu, 22 Jun 2023 14:03:27 +0200
Subject: [PATCH 008/142] cleaned gin file

---
 configs/prediction_models/TemporalFusionTransformer.gin | 2 --
 1 file changed, 2 deletions(-)

diff --git a/configs/prediction_models/TemporalFusionTransformer.gin b/configs/prediction_models/TemporalFusionTransformer.gin
index d430a443..25be512d 100644
--- a/configs/prediction_models/TemporalFusionTransformer.gin
+++ b/configs/prediction_models/TemporalFusionTransformer.gin
@@ -15,8 +15,6 @@ model/hyperparameter.class_to_tune = @TemporalFusionTransformer
 model/hyperparameter.encoder_length = (1,14,"log-uniform", 2)
 model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
 model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
-model/hyperparameter.depth = (1, 3)
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
 model/hyperparameter.n_heads = (1, 8, "log-uniform", 2)

From bed9b87bc9029c64745262d3a20b4ff6fcb1ab7c Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Thu, 22 Jun 2023 14:03:53 +0200
Subject: [PATCH 009/142] changed output to logits to match num of classes

---
 icu_benchmarks/models/dl_models.py | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 4646502e..be1f53bd 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -290,7 +290,7 @@ class TemporalFusionTransformer(DLPredictionWrapper):
 
 
     _supported_run_modes = [RunMode.classification, RunMode.regression]
-    def __init__(self, encoder_length,hidden,num_static_vars,dropout,num_historic_vars,num_future_vars,
+    def __init__(self,num_classes, encoder_length,hidden,num_static_vars,dropout,num_historic_vars,num_future_vars,
                  n_head,dropout_att,example_length,quantiles,static_categorical_inp_size=1,temporal_known_categorical_inp_size=4,
     temporal_observed_categorical_inp_size=0,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
     temporal_observed_continuous_inp_size=48,temporal_target_size=1):
@@ -312,6 +312,7 @@ def __init__(self, encoder_length,hidden,num_static_vars,dropout,num_historic_va
         self.static_encoder = StaticCovariateEncoder(num_static_vars,hidden,dropout)
         self.TFTpart2 = torch.jit.script(TFTBack(encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
                 n_head,dropout_att,example_length,quantiles))
+        self.logit = nn.Linear(len(quantiles), num_classes)
 
     def forward(self, x: Dict[str, Tensor]) -> Tensor:
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
@@ -327,4 +328,6 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
 
         historical_inputs = cat(_historical_inputs, dim=-2)
         future_inputs = t_known_inp[:, self.encoder_length:]
-        return self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs)
\ No newline at end of file
+        o=self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs)
+        pred = self.logit(o)
+        return pred
\ No newline at end of file

From 9bda7693c66b5202c14572fc8702963a7a9919e2 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Thu, 22 Jun 2023 14:38:44 +0200
Subject: [PATCH 010/142] added jit import

---
 icu_benchmarks/models/dl_models.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index be1f53bd..38c94a01 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -6,7 +6,7 @@
 from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor,cat
+from torch import Tensor,cat,jit
 
 
 @gin.configurable
@@ -310,7 +310,7 @@ def __init__(self,num_classes, encoder_length,hidden,num_static_vars,dropout,num
                 temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
                 temporal_observed_continuous_inp_size,temporal_target_size,hidden)
         self.static_encoder = StaticCovariateEncoder(num_static_vars,hidden,dropout)
-        self.TFTpart2 = torch.jit.script(TFTBack(encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
+        self.TFTpart2 = jit.script(TFTBack(encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
                 n_head,dropout_att,example_length,quantiles))
         self.logit = nn.Linear(len(quantiles), num_classes)
 

From 960320b2921680d6cbc6191480d59a26dab299a5 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Thu, 22 Jun 2023 15:27:39 +0200
Subject: [PATCH 011/142] added quantile loss

---
 icu_benchmarks/models/dl_models.py |  4 ++--
 icu_benchmarks/models/layers.py    | 12 +++++++++++-
 2 files changed, 13 insertions(+), 3 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 38c94a01..f78081d4 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -3,7 +3,7 @@
 import numpy as np
 import torch.nn as nn
 from icu_benchmarks.contants import RunMode
-from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack
+from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack,QuantileLoss
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor,cat,jit
@@ -294,7 +294,7 @@ def __init__(self,num_classes, encoder_length,hidden,num_static_vars,dropout,num
                  n_head,dropout_att,example_length,quantiles,static_categorical_inp_size=1,temporal_known_categorical_inp_size=4,
     temporal_observed_categorical_inp_size=0,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
     temporal_observed_continuous_inp_size=48,temporal_target_size=1):
-        super().__init__()
+        super().__init__(loss=QuantileLoss())
 
         #dervied embeddings size
         num_static_vars=static_categorical_inp_size+static_continuous_inp_size
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 89201db5..7fa910bf 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -757,4 +757,14 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
 
         out = self.quantile_proj(x)
 
-        return out
\ No newline at end of file
+        return out
+class QuantileLoss(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.register_buffer('q', torch.tensor(config.quantiles))
+
+    def forward(self, predictions, targets):
+        diff = predictions - targets
+        ql = (1-self.q)*F.relu(diff) + self.q*F.relu(-diff)
+        losses = ql.view(-1, ql.shape[-1]).mean(0)
+        return losses
\ No newline at end of file

From 3cb823dab9fc3542eaffb1fa13d62fa9cbf12a3a Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Thu, 22 Jun 2023 16:42:20 +0200
Subject: [PATCH 012/142] added new loader to work with tft but still needs
 work

---
 icu_benchmarks/data/loader.py | 84 +++++++++++++++++++++++++++++++++--
 1 file changed, 81 insertions(+), 3 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index d0e040e4..cb6df099 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -2,15 +2,15 @@
 from pandas import DataFrame
 import gin
 import numpy as np
-from torch import Tensor, cat, from_numpy, float32
+from torch import Tensor, cat, from_numpy, float32,stack,empty
 from torch.utils.data import Dataset
 import logging
 from typing import Dict, Tuple
-
+from collections import OrderedDic
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
-
+FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target', 'id']
 
 class CommonDataset(Dataset):
     """Common dataset: subclass of Torch Dataset that represents the data to learn on.
@@ -31,6 +31,7 @@ def __init__(
         self.vars = vars
         self.grouping_df = data[split][grouping_segment].set_index(self.vars["GROUP"])
         self.features_df = (
+            #drops time coulmn and sets index to stay_id
             data[split][Segment.features].set_index(self.vars["GROUP"]).drop(labels=self.vars["SEQUENCE"], axis=1)
         )
 
@@ -153,6 +154,82 @@ def to_tensor(self):
         data, labels = self.get_data_and_labels()
         return from_numpy(data).to(float32), from_numpy(labels).to(float32)
 
+@gin.configurable("PredictionDatasetTFT")
+class PredictionDatasetTFT(PredictionDataset):
+    """Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.
+
+    Args:
+        ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.
+    """
+
+    def __init__(self, *args, ram_cache: bool = True, **kwargs):
+        super().__init__(*args,ram_cache = True, grouping_segment=Segment.outcome, **kwargs)
+        
+
+    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
+        """Function to sample from the data split of choice. Used for TFT.
+
+        Args:
+            idx: A specific row index to sample.
+
+        Returns:
+            A sample from the data, consisting of data, labels and padding mask.
+        """
+        if self._cached_dataset is not None:
+            return self._cached_dataset[idx]
+
+        pad_value = 0.0
+        stay_id = self.outcome_df.index.unique()[idx]  # [self.vars["GROUP"]]
+
+        
+        # We need to be sure that tensors are returned in the correct order to be processed correclty by tft
+        tensors = tuple([] for _ in range(8))
+        for var in self.features_df.columns:
+            if  var == 'sex' :
+                tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+            elif var == 'age' or var== 'height' or var== 'weight':
+                tensors[1].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+            else :
+                tensors[5].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+    
+
+        tensors[6].append(self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float))
+        tensors[7].append(stay_id.to_numpy())
+        tensors = [stack(x, dim=-1) if x else empty(0) for x in tensors]
+
+        window = self.features_df.loc[stay_id:stay_id].to_numpy()
+        labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
+
+        if len(tensors[6]) == 1:
+            # only one label per stay, align with window
+            tensors[6] = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, tensors[6]], axis=0)
+
+        length_diff = self.maxlen - window.shape[0]
+
+        pad_mask = np.ones(window.shape[0])
+
+        # Padding the array to fulfill size requirement
+        if length_diff > 0:
+            # window shorter than the longest window in dataset, pad to same length
+
+            window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
+            tensors[6] = np.concatenate([tensors[6], np.ones(length_diff) * pad_value], axis=0)
+            pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
+
+        not_labeled = np.argwhere(np.isnan(tensors[6]))
+        if len(not_labeled) > 0:
+            tensors[6][not_labeled] = -1
+            pad_mask[not_labeled] = 0
+
+        pad_mask = pad_mask.astype(bool)
+        tensors[6] = tensors[6].astype(np.float32)
+        data = window.astype(np.float32)
+
+        return  OrderedDict(zip(FEAT_NAMES, tensors))
+
+    
+
+
 
 @gin.configurable("ImputationDataset")
 class ImputationDataset(CommonDataset):
@@ -285,3 +362,4 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         window = self.dyn_df.loc[stay_id:stay_id, :]
 
         return from_numpy(window.values).to(float32)
+

From f823e29fa4fcbb88ceb57f07868399411a972a53 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 13:54:22 +0200
Subject: [PATCH 013/142] added dataloader for tft with correct order

---
 icu_benchmarks/data/loader.py  | 34 ++++++++++++++++++++++++----------
 icu_benchmarks/models/train.py |  5 +++--
 2 files changed, 27 insertions(+), 12 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index cb6df099..e380a7cc 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -10,6 +10,7 @@
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
+from collections import  OrderedDict
 FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target', 'id']
 
 class CommonDataset(Dataset):
@@ -197,22 +198,28 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         tensors[7].append(stay_id.to_numpy())
         tensors = [stack(x, dim=-1) if x else empty(0) for x in tensors]
 
-        window = self.features_df.loc[stay_id:stay_id].to_numpy()
-        labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
+       # window = self.features_df.loc[stay_id:stay_id].to_numpy()
+       # labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
 
         if len(tensors[6]) == 1:
             # only one label per stay, align with window
-            tensors[6] = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, tensors[6]], axis=0)
+            tensors[6] = np.concatenate([np.empty(tensors[0].shape[0] - 1) * np.nan, tensors[6]], axis=0)
 
-        length_diff = self.maxlen - window.shape[0]
+        length_diff = self.maxlen - tensors[0].shape[0]
 
-        pad_mask = np.ones(window.shape[0])
+        pad_mask = np.ones(tensors[0].shape[0])
 
         # Padding the array to fulfill size requirement
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
-
-            window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
+            tensors[0]= np.concatenate([tensors[0], np.ones((length_diff, tensors[0].shape[1])) * pad_value], axis=0)
+            tensors[1]= np.concatenate([tensors[1], np.ones((length_diff, tensors[1].shape[1])) * pad_value], axis=0)
+            tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, tensors[2].shape[1])) * pad_value], axis=0)
+            tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, tensors[3].shape[1])) * pad_value], axis=0)
+            tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, tensors[4].shape[1])) * pad_value], axis=0)
+            tensors[5]= np.concatenate([tensors[5], np.ones((length_diff, tensors[5].shape[1])) * pad_value], axis=0)
+            tensors[7]= np.concatenate([tensors[7], np.ones((length_diff, tensors[7].shape[1])) * pad_value], axis=0)
+           # window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
             tensors[6] = np.concatenate([tensors[6], np.ones(length_diff) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
 
@@ -223,9 +230,16 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         pad_mask = pad_mask.astype(bool)
         tensors[6] = tensors[6].astype(np.float32)
-        data = window.astype(np.float32)
-
-        return  OrderedDict(zip(FEAT_NAMES, tensors))
+       # data = window.astype(np.float32)
+        tensors[0]= tensors[0].astype(np.float32)
+        tensors[1]= tensors[1].astype(np.float32)
+        tensors[2]=tensors[2].astype(np.float32)
+        tensors[3]= tensors[3].astype(np.float32)
+        tensors[4]= tensors[4].astype(np.float32)
+        tensors[5]=tensors[5].astype(np.float32)
+        tensors[7]= tensors[7].astype(np.float32)
+
+        return  OrderedDict(zip(FEAT_NAMES, tensors)),from_numpy(pad_mask)
 
     
 
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 1c2a91cd..ba57fa6b 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -9,7 +9,7 @@
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint, TQDMProgressBar
 from pathlib import Path
-from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset
+from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset,PredictionDatasetTFT
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
@@ -72,7 +72,8 @@ def train_common(
     """
 
     logging.info(f"Training model: {model.__name__}.")
-    dataset_class = ImputationDataset if mode == RunMode.imputation else PredictionDataset
+    dataset_class = ImputationDataset if mode == RunMode.imputation else model == PredictionDatasetTFT if model =='TemporalFusionTransformer' else PredictionDataset 
+    
 
     logging.info(f"Logging to directory: {log_dir}.")
     save_config_file(log_dir)  # We save the operative config before and also after training

From 742905fac7bf0cda830a7f478dd0523672bcd97a Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 13:54:54 +0200
Subject: [PATCH 014/142] tft

---
 icu_benchmarks/models/dl_models.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index f78081d4..c76a6fda 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -315,6 +315,7 @@ def __init__(self,num_classes, encoder_length,hidden,num_static_vars,dropout,num
         self.logit = nn.Linear(len(quantiles), num_classes)
 
     def forward(self, x: Dict[str, Tensor]) -> Tensor:
+        
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
 
         # Static context

From bb7ecc05379978f7d7f919012ac4baaa1aa06396 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 14:06:44 +0200
Subject: [PATCH 015/142] added quantile loss

---
 configs/tasks/RegressionTFT.gin | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)
 create mode 100644 configs/tasks/RegressionTFT.gin

diff --git a/configs/tasks/RegressionTFT.gin b/configs/tasks/RegressionTFT.gin
new file mode 100644
index 00000000..b31408fb
--- /dev/null
+++ b/configs/tasks/RegressionTFT.gin
@@ -0,0 +1,33 @@
+# COMMON IMPORTS
+include "configs/tasks/common/Imports.gin"
+
+# DATASET CONFIGURATION
+include "configs/tasks/common/PredictionTaskVariables.gin"
+
+# CROSS-VALIDATION
+include "configs/tasks/common/CrossValidation.gin"
+
+# MODE SETTINGS
+Run.mode = "Regression"
+NUM_CLASSES = 1
+HORIZON = 24
+train_common.weight = "balanced"
+train_common.ram_cache = True
+
+# LOSS FUNCTION
+DLPredictionWrapper.loss = @QuantileLoss
+
+
+# SELECTING PREPROCESSOR
+preprocess.preprocessor = @base_regression_preprocessor
+preprocess.vars = %vars
+preprocess.use_static = True
+
+# SPECIFYING REGRESSION OUTCOME SCALING
+base_regression_preprocessor.outcome_min = 0
+base_regression_preprocessor.outcome_max = 15
+
+# SELECTING DATASET
+PredictionDataset.vars = %vars
+PredictionDataset.ram_cache = True
+

From c4085ef1be4cf7bc8260a462050327e214f4003d Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 14:14:33 +0200
Subject: [PATCH 016/142] added gin configurable for quantile loss

---
 icu_benchmarks/models/layers.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 7fa910bf..de8afaf5 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -758,6 +758,7 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         out = self.quantile_proj(x)
 
         return out
+@gin.configurable
 class QuantileLoss(nn.Module):
     def __init__(self, config):
         super().__init__()

From fa7e4b94ba3504e9f7ada59ad2a83c1a0897f547 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 14:14:49 +0200
Subject: [PATCH 017/142] remove gin config for quantile loss

---
 icu_benchmarks/models/wrappers.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 867e9874..70dfbdc9 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -17,14 +17,14 @@
 from icu_benchmarks.models.utils import create_optimizer, create_scheduler
 from joblib import dump
 from pytorch_lightning import LightningModule
-
+import layers
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
 
 gin.config.external_configurable(torch.nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(torch.nn.functional.cross_entropy, module="torch.nn.functional")
+#gin.config.external_configurable(layers.QuantileLoss, module="layers")
 gin.config.external_configurable(torch.nn.functional.mse_loss, module="torch.nn.functional")
-
 gin.config.external_configurable(sklearn.metrics.mean_squared_error, module="sklearn.metrics")
 gin.config.external_configurable(sklearn.metrics.log_loss, module="sklearn.metrics")
 

From da9725bbec1614d475d5dccd235e2fdc57504fc1 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 14:18:09 +0200
Subject: [PATCH 018/142] added quantile loss

---
 icu_benchmarks/models/wrappers.py | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 70dfbdc9..df19c0b9 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -28,7 +28,17 @@
 gin.config.external_configurable(sklearn.metrics.mean_squared_error, module="sklearn.metrics")
 gin.config.external_configurable(sklearn.metrics.log_loss, module="sklearn.metrics")
 
+@gin.configurable
+class QuantileLoss(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.register_buffer('q', torch.tensor(config.quantiles))
 
+    def forward(self, predictions, targets):
+        diff = predictions - targets
+        ql = (1-self.q)*F.relu(diff) + self.q*F.relu(-diff)
+        losses = ql.view(-1, ql.shape[-1]).mean(0)
+        return losses
 @gin.configurable("BaseModule")
 class BaseModule(LightningModule):
     needs_training = False

From 7c3d5111a5af1b5af7300e380df1354a1aa3dfae Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 14:18:19 +0200
Subject: [PATCH 019/142] removed quantile loss

---
 icu_benchmarks/models/layers.py | 11 -----------
 1 file changed, 11 deletions(-)

diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index de8afaf5..8a8a95d3 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -758,14 +758,3 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         out = self.quantile_proj(x)
 
         return out
-@gin.configurable
-class QuantileLoss(nn.Module):
-    def __init__(self, config):
-        super().__init__()
-        self.register_buffer('q', torch.tensor(config.quantiles))
-
-    def forward(self, predictions, targets):
-        diff = predictions - targets
-        ql = (1-self.q)*F.relu(diff) + self.q*F.relu(-diff)
-        losses = ql.view(-1, ql.shape[-1]).mean(0)
-        return losses
\ No newline at end of file

From d410d53e6b048a33a93b8b30c8cb70be8945a80d Mon Sep 17 00:00:00 2001
From: youssefmecky96 <47695820+youssefmecky96@users.noreply.github.com>
Date: Tue, 27 Jun 2023 14:26:21 +0200
Subject: [PATCH 020/142] Update ci.yml

To see reason for failure
---
 .github/workflows/ci.yml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 048e2275..77fffa73 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -30,6 +30,7 @@ jobs:
           auto-activate-base: false
       - name: Lint with flake8
         run: |
+          set -x
           # stop the build if there are Python syntax errors or undefined names
           flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
           # the GitHub editor is 127 chars wide
@@ -40,4 +41,4 @@ jobs:
       # - name: Setup package
       #   run: pip install -e .
       # - name: Test command line tool
-      #   run: python -m icu_benchmarks.run --help
\ No newline at end of file
+      #   run: python -m icu_benchmarks.run --help

From 38f9dee97a104b94175e327cea51e1042882cead Mon Sep 17 00:00:00 2001
From: youssefmecky96 <47695820+youssefmecky96@users.noreply.github.com>
Date: Tue, 27 Jun 2023 14:39:15 +0200
Subject: [PATCH 021/142] Update ci.yml

Removed changes
---
 .github/workflows/ci.yml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 77fffa73..6c857c4b 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -30,7 +30,6 @@ jobs:
           auto-activate-base: false
       - name: Lint with flake8
         run: |
-          set -x
           # stop the build if there are Python syntax errors or undefined names
           flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
           # the GitHub editor is 127 chars wide

From bd0a9a5614a5553ca052434d58291f751ea5aea1 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 16:40:12 +0200
Subject: [PATCH 022/142] fixing bugs to get tft to work

---
 .../TemporalFusionTransformer.gin             |  8 +++--
 configs/prediction_models/common/DLCommon.gin |  1 +
 configs/tasks/common/Imports.gin              |  3 +-
 environment.yml                               |  4 +--
 experiments/demo_benchmark_classification.yml | 23 +++++++------
 icu_benchmarks/data/loader.py                 |  3 +-
 icu_benchmarks/models/dl_models.py            | 34 ++++++++++++-------
 icu_benchmarks/models/layers.py               | 27 +++++++++++----
 icu_benchmarks/models/train.py                |  2 +-
 icu_benchmarks/models/wrappers.py             | 16 +++------
 10 files changed, 69 insertions(+), 52 deletions(-)

diff --git a/configs/prediction_models/TemporalFusionTransformer.gin b/configs/prediction_models/TemporalFusionTransformer.gin
index 25be512d..1d8ec015 100644
--- a/configs/prediction_models/TemporalFusionTransformer.gin
+++ b/configs/prediction_models/TemporalFusionTransformer.gin
@@ -1,4 +1,4 @@
-# Settings for Transformer model.
+# Settings for TFT model.
 
 # Common settings for DL models
 include "configs/prediction_models/common/DLCommon.gin"
@@ -17,8 +17,10 @@ model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads = (1, 8, "log-uniform", 2)
-model/hyperparameter.quantiles=[0.1, 0.5, 0.9]
+model/hyperparameter.n_heads = 1
+
 model/hyperparameter.example_length=(1,14,"log-uniform", 2)
 
 
+
+
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index c220e6ab..4b761726 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -5,6 +5,7 @@ import gin.torch.external_configurables
 import icu_benchmarks.models.wrappers
 import icu_benchmarks.models.dl_models
 import icu_benchmarks.models.utils
+import icu_benchmarks.models.layers
 
 # Do not generate features from dynamic data
 base_classification_preprocessor.generate_features = False
diff --git a/configs/tasks/common/Imports.gin b/configs/tasks/common/Imports.gin
index ea448476..faf3c8d6 100644
--- a/configs/tasks/common/Imports.gin
+++ b/configs/tasks/common/Imports.gin
@@ -3,4 +3,5 @@ import icu_benchmarks.data.split_process_data
 import icu_benchmarks.data.loader
 import icu_benchmarks.models.wrappers
 import icu_benchmarks.models.dl_models
-import icu_benchmarks.models.ml_models
\ No newline at end of file
+import icu_benchmarks.models.ml_models
+import icu_benchmarks.models.layers
\ No newline at end of file
diff --git a/environment.yml b/environment.yml
index d47b7e2f..a1b3f061 100644
--- a/environment.yml
+++ b/environment.yml
@@ -13,7 +13,7 @@ dependencies:
   - gin-config=0.5.0
   - ignite=0.4.11
   - pytorch=2.0.1
-  #- pytorch-cuda=11.8
+ # - pytorch-cuda=11.8
   - lightgbm=3.3.5
   - numpy=1.24.3
   - pandas=2.0.0
@@ -28,7 +28,7 @@ dependencies:
   - einops=0.6.1
   - hydra-core=1.3
   - pip:
-      - recipies==0.1.1
+      - recipies==0.1.2
       # Fixed version because of NumPy incompatibility and stale development status.
       - scikit-optimize-fix==0.9.1
       - hydra-submitit-launcher==1.2.0
diff --git a/experiments/demo_benchmark_classification.yml b/experiments/demo_benchmark_classification.yml
index 32516df2..15802f9b 100644
--- a/experiments/demo_benchmark_classification.yml
+++ b/experiments/demo_benchmark_classification.yml
@@ -18,19 +18,20 @@ parameters:
   data_dir:
     values:
       - demo_data/mortality24/eicu_demo
-      - demo_data/mortality24/mimic_demo
-      - demo_data/aki/eicu_demo
-      - demo_data/aki/mimic_demo
-      - demo_data/sepsis/eicu_demo
-      - demo_data/sepsis/mimic_demo
+      #- demo_data/mortality24/mimic_demo
+     # - demo_data/aki/eicu_demo
+    #  - demo_data/aki/mimic_demo
+   #   - demo_data/sepsis/eicu_demo
+  #    - demo_data/sepsis/mimic_demo
   model:
     values:
-      - LogisticRegression
-      - LGBMClassifier
-      - GRU
-      - LSTM
-      - TCN
-      - Transformer
+   #   - LogisticRegression
+    #  - LGBMClassifier
+    #  - GRU
+    #  - LSTM
+    #  - TCN
+    #  - Transformer
+      - TemporalFusionTransformer
   seed:
     values:
       - 1111
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index e380a7cc..5c67942c 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -6,11 +6,10 @@
 from torch.utils.data import Dataset
 import logging
 from typing import Dict, Tuple
-from collections import OrderedDic
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
-from collections import  OrderedDict
+from collections import OrderedDict
 FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target', 'id']
 
 class CommonDataset(Dataset):
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index c76a6fda..2240e2e1 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -3,7 +3,7 @@
 import numpy as np
 import torch.nn as nn
 from icu_benchmarks.contants import RunMode
-from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack,QuantileLoss
+from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor,cat,jit
@@ -282,6 +282,7 @@ def forward(self, x):
         o = o.permute(0, 2, 1)  # Permute to channel last
         pred = self.logit(o)
         return pred
+    
 @gin.configurable
 class TemporalFusionTransformer(DLPredictionWrapper):
     """ 
@@ -290,28 +291,35 @@ class TemporalFusionTransformer(DLPredictionWrapper):
 
 
     _supported_run_modes = [RunMode.classification, RunMode.regression]
-    def __init__(self,num_classes, encoder_length,hidden,num_static_vars,dropout,num_historic_vars,num_future_vars,
-                 n_head,dropout_att,example_length,quantiles,static_categorical_inp_size=1,temporal_known_categorical_inp_size=4,
-    temporal_observed_categorical_inp_size=0,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
-    temporal_observed_continuous_inp_size=48,temporal_target_size=1):
-        super().__init__(loss=QuantileLoss())
-
-        #dervied embeddings size
-        num_static_vars=static_categorical_inp_size+static_continuous_inp_size
-        num_future_vars=temporal_known_categorical_inp_size+temporal_known_continuous_inp_size
+    def __init__(self,num_classes, encoder_length,hidden,dropout,
+                 n_heads,dropout_att,example_length,*args,quantiles=[0.1, 0.5, 0.9],static_categorical_inp_size=[4],temporal_known_categorical_inp_size=[],
+    temporal_observed_categorical_inp_size=[],static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
+    temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs):
+        #derived variables
+        num_static_vars=len(static_categorical_inp_size)+static_continuous_inp_size
+        num_future_vars=len(temporal_known_categorical_inp_size)+temporal_known_continuous_inp_size
         num_historic_vars=sum([num_future_vars,
                                       temporal_observed_continuous_inp_size,
                                       temporal_target_size,
-                                      temporal_observed_categorical_inp_size,
+                                      len(temporal_observed_categorical_inp_size),
                                       ])
+      
+        super().__init__(num_classes=num_classes, encoder_length=encoder_length,hidden=hidden,
+                 n_heads=n_heads,dropout_att=dropout_att,example_length=example_length,quantiles=quantiles,
+                 num_static_vars=num_static_vars,num_future_vars=num_future_vars,num_historic_vars=num_historic_vars,*args,static_categorical_inp_size=1,temporal_known_categorical_inp_size=4,
+                temporal_observed_categorical_inp_size=0,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
+                temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs)
+
+        
+        
         self.encoder_length = encoder_length #this determines from how distant past we want to use data from
 
         self.embedding = LazyEmbedding(static_categorical_inp_size,temporal_known_categorical_inp_size,
                 temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
                 temporal_observed_continuous_inp_size,temporal_target_size,hidden)
         self.static_encoder = StaticCovariateEncoder(num_static_vars,hidden,dropout)
-        self.TFTpart2 = jit.script(TFTBack(encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
-                n_head,dropout_att,example_length,quantiles))
+        self.TFTpart2 = TFTBack(encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
+                n_heads,dropout_att,example_length,quantiles)
         self.logit = nn.Linear(len(quantiles), num_classes)
 
     def forward(self, x: Dict[str, Tensor]) -> Tensor:
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 8a8a95d3..c4616679 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -1,4 +1,3 @@
-import os
 import gin
 import math
 import torch
@@ -7,9 +6,10 @@
 from torch.nn.utils import weight_norm
 from torch import Tensor
 from torch.nn.parameter import UninitializedParameter
-from typing import Dict, Tuple, Optional, List
+from typing import Dict, Tuple, Optional
 from torch.nn import LayerNorm
 
+
 @gin.configurable("masking")
 def parallel_recomb(q_t, kv_t, att_type="all", local_context=3, bin_size=None):
     """Return mask of attention matrix (ts_q, ts_kv)"""
@@ -357,13 +357,13 @@ def __init__(self,
                  input_size,
                  hidden,
                  output_size=None,
-                 context_hidden_size=None,
+                 context_hidden=None,
                  dropout=0.0,):
         super().__init__()
         self.layer_norm = MaybeLayerNorm(output_size, hidden, eps=1e-3)
         self.lin_a = nn.Linear(input_size, hidden)
-        if context_hidden_size is not None:
-            self.lin_c = nn.Linear(context_hidden_size, hidden, bias=False)
+        if context_hidden is not None:
+            self.lin_c = nn.Linear(context_hidden, hidden, bias=False)
         else:
             self.lin_c = nn.Identity()
         self.lin_i = nn.Linear(hidden, hidden)
@@ -570,6 +570,7 @@ def __init__(self, static_categorical_inp_size,temporal_known_categorical_inp_si
         self.t_tgt_embedding_bias = UninitializedParameter()
 
     def initialize_parameters(self, x):
+        
         if self.has_uninitialized_params():
             s_cont_inp = x.get('s_cont', None)
             t_cont_k_inp = x.get('k_cont', None)
@@ -703,7 +704,7 @@ def __init__(self, encoder_length,num_historic_vars,hidden,dropout,num_future_va
 
         self.enrichment_grn = GRN(hidden,
                                   hidden,
-                                  context_hidden_size=hidden, 
+                                  context_hidden=hidden, 
                                   dropout=dropout)
         self.attention = InterpretableMultiHeadAttention(n_head,hidden,dropout_att,dropout,example_length)
         self.attention_gate = GLU(hidden, hidden)
@@ -715,7 +716,8 @@ def __init__(self, encoder_length,num_historic_vars,hidden,dropout,num_future_va
 
         self.decoder_gate = GLU(hidden, hidden)
         self.decoder_ln = LayerNorm(hidden, eps=1e-3)
-
+        
+        
         self.quantile_proj = nn.Linear(hidden, len(quantiles))
         
     def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
@@ -758,3 +760,14 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         out = self.quantile_proj(x)
 
         return out
+@gin.configurable
+class QuantileLoss(torch.nn.Module):
+    def __init__(self, quantiles):
+        super().__init__()
+        self.register_buffer('q', torch.tensor(quantiles))
+
+    def forward(self, predictions, targets):
+        diff = predictions - targets
+        ql = (1-self.q)*F.relu(diff) + self.q*F.relu(-diff)
+        losses = ql.view(-1, ql.shape[-1]).mean(0)
+        return losses
\ No newline at end of file
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index ba57fa6b..61b9bf5e 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -72,7 +72,7 @@ def train_common(
     """
 
     logging.info(f"Training model: {model.__name__}.")
-    dataset_class = ImputationDataset if mode == RunMode.imputation else model == PredictionDatasetTFT if model =='TemporalFusionTransformer' else PredictionDataset 
+    dataset_class = ImputationDataset if mode == RunMode.imputation else (PredictionDatasetTFT if model.__name__ =='TemporalFusionTransformer' else PredictionDataset )
     
 
     logging.info(f"Logging to directory: {log_dir}.")
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index df19c0b9..fc1bd77e 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -8,6 +8,7 @@
 from torch.nn import MSELoss, CrossEntropyLoss
 from torch.nn.modules.loss import _Loss
 from torch.optim import Optimizer
+
 import inspect
 import gin
 import numpy as np
@@ -17,28 +18,16 @@
 from icu_benchmarks.models.utils import create_optimizer, create_scheduler
 from joblib import dump
 from pytorch_lightning import LightningModule
-import layers
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
 
 gin.config.external_configurable(torch.nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(torch.nn.functional.cross_entropy, module="torch.nn.functional")
-#gin.config.external_configurable(layers.QuantileLoss, module="layers")
 gin.config.external_configurable(torch.nn.functional.mse_loss, module="torch.nn.functional")
 gin.config.external_configurable(sklearn.metrics.mean_squared_error, module="sklearn.metrics")
 gin.config.external_configurable(sklearn.metrics.log_loss, module="sklearn.metrics")
 
-@gin.configurable
-class QuantileLoss(nn.Module):
-    def __init__(self, config):
-        super().__init__()
-        self.register_buffer('q', torch.tensor(config.quantiles))
 
-    def forward(self, predictions, targets):
-        diff = predictions - targets
-        ql = (1-self.q)*F.relu(diff) + self.q*F.relu(-diff)
-        losses = ql.view(-1, ql.shape[-1]).mean(0)
-        return losses
 @gin.configurable("BaseModule")
 class BaseModule(LightningModule):
     needs_training = False
@@ -108,6 +97,7 @@ class DLWrapper(BaseModule, ABC):
 
     def __init__(
         self,
+        
         loss=CrossEntropyLoss(),
         optimizer=torch.optim.Adam,
         run_mode: RunMode = RunMode.classification,
@@ -120,8 +110,10 @@ def __init__(
         epochs: int = 100,
         input_size: torch.Tensor = None,
         initialization_method: str = "normal",
+        
         **kwargs,
     ):
+        
         """Interface for Deep Learning models."""
         super().__init__()
         self.save_hyperparameters(ignore=["loss", "optimizer"])

From f6d1272f4525f3e39c777670eb2a2b4acabde058 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 17:45:38 +0200
Subject: [PATCH 023/142] fixing bugs to run tft for classification

---
 icu_benchmarks/data/loader.py | 53 ++++++++++++++++-------------------
 1 file changed, 24 insertions(+), 29 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 5c67942c..d0814800 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -2,7 +2,7 @@
 from pandas import DataFrame
 import gin
 import numpy as np
-from torch import Tensor, cat, from_numpy, float32,stack,empty
+from torch import Tensor, cat, from_numpy, float32,empty,stack
 from torch.utils.data import Dataset
 import logging
 from typing import Dict, Tuple
@@ -103,7 +103,8 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             labels = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, labels], axis=0)
 
         length_diff = self.maxlen - window.shape[0]
-
+        
+        
         pad_mask = np.ones(window.shape[0])
 
         # Padding the array to fulfill size requirement
@@ -163,7 +164,7 @@ class PredictionDatasetTFT(PredictionDataset):
     """
 
     def __init__(self, *args, ram_cache: bool = True, **kwargs):
-        super().__init__(*args,ram_cache = True, grouping_segment=Segment.outcome, **kwargs)
+        super().__init__(*args,ram_cache = True, **kwargs)
         
 
     def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
@@ -183,7 +184,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         
         # We need to be sure that tensors are returned in the correct order to be processed correclty by tft
-        tensors = tuple([] for _ in range(8))
+        tensors = [[] for _ in range(8)]
         for var in self.features_df.columns:
             if  var == 'sex' :
                 tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
@@ -192,33 +193,34 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             else :
                 tensors[5].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
     
-
-        tensors[6].append(self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float))
-        tensors[7].append(stay_id.to_numpy())
-        tensors = [stack(x, dim=-1) if x else empty(0) for x in tensors]
-
+        
+        tensors[6].extend(self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float))
+        tensors[7].append(np.asarray([stay_id]))
+        
+        window_shape0=np.shape(tensors[0])[1]
        # window = self.features_df.loc[stay_id:stay_id].to_numpy()
        # labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
-
+        
         if len(tensors[6]) == 1:
             # only one label per stay, align with window
-            tensors[6] = np.concatenate([np.empty(tensors[0].shape[0] - 1) * np.nan, tensors[6]], axis=0)
+            tensors[6] = np.concatenate([np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0)
 
-        length_diff = self.maxlen - tensors[0].shape[0]
+        length_diff = self.maxlen - window_shape0
 
-        pad_mask = np.ones(tensors[0].shape[0])
+        pad_mask = np.ones(window_shape0)
 
         # Padding the array to fulfill size requirement
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
-            tensors[0]= np.concatenate([tensors[0], np.ones((length_diff, tensors[0].shape[1])) * pad_value], axis=0)
-            tensors[1]= np.concatenate([tensors[1], np.ones((length_diff, tensors[1].shape[1])) * pad_value], axis=0)
-            tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, tensors[2].shape[1])) * pad_value], axis=0)
-            tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, tensors[3].shape[1])) * pad_value], axis=0)
-            tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, tensors[4].shape[1])) * pad_value], axis=0)
-            tensors[5]= np.concatenate([tensors[5], np.ones((length_diff, tensors[5].shape[1])) * pad_value], axis=0)
-            tensors[7]= np.concatenate([tensors[7], np.ones((length_diff, tensors[7].shape[1])) * pad_value], axis=0)
+            tensors[0]= np.concatenate([tensors[0], np.ones((length_diff, np.shape(tensors[0])[0])) * pad_value], axis=0)
+            tensors[1]= np.concatenate([tensors[1], np.ones((length_diff, np.shape(tensors[1])[0])) * pad_value], axis=0)
+            tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, np.shape(tensors[2])[0])) * pad_value], axis=0)
+            tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, np.shape(tensors[3])[0])) * pad_value], axis=0)
+            tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, np.shape(tensors[4])[0])) * pad_value], axis=0)
+            tensors[5]= np.concatenate([tensors[5], np.ones((length_diff, np.shape(tensors[5])[0])) * pad_value], axis=0)
+            tensors[7]= np.concatenate([tensors[7], np.ones((length_diff, np.shape(tensors[7])[0])) * pad_value], axis=0)
            # window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
+            
             tensors[6] = np.concatenate([tensors[6], np.ones(length_diff) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
 
@@ -228,16 +230,9 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             pad_mask[not_labeled] = 0
 
         pad_mask = pad_mask.astype(bool)
-        tensors[6] = tensors[6].astype(np.float32)
        # data = window.astype(np.float32)
-        tensors[0]= tensors[0].astype(np.float32)
-        tensors[1]= tensors[1].astype(np.float32)
-        tensors[2]=tensors[2].astype(np.float32)
-        tensors[3]= tensors[3].astype(np.float32)
-        tensors[4]= tensors[4].astype(np.float32)
-        tensors[5]=tensors[5].astype(np.float32)
-        tensors[7]= tensors[7].astype(np.float32)
-
+        tensors = [from_numpy(np.array(tensor)).to(float32) for tensor in tensors]
+        tensors = [stack((x,), dim=-1) if x.numel() else empty(0) for x in tensors]
         return  OrderedDict(zip(FEAT_NAMES, tensors)),from_numpy(pad_mask)
 
     

From 0c44d2ffbdbddaa98a1e2e76447ad994f5cfbcc5 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 17:48:06 +0200
Subject: [PATCH 024/142] fixing bugs to run tft for classification

---
 icu_benchmarks/models/train.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 61b9bf5e..30deb6a4 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -102,7 +102,7 @@ def train_common(
         pin_memory=True,
         drop_last=True,
     )
-
+    print(type(next(iter(train_loader))[0]))
     data_shape = next(iter(train_loader))[0].shape
 
     model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)

From 9963d7bc3c0495c07d7ba0615d61a5bf18bdd17a Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 27 Jun 2023 23:03:04 +0200
Subject: [PATCH 025/142] fixing bugs to run tft

---
 icu_benchmarks/data/loader.py      | 11 +++++------
 icu_benchmarks/models/dl_models.py |  2 +-
 icu_benchmarks/models/layers.py    | 13 +++++++------
 icu_benchmarks/models/train.py     | 13 ++++++++-----
 icu_benchmarks/models/wrappers.py  |  9 ++++++---
 5 files changed, 27 insertions(+), 21 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index d0814800..b2b7ad72 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -200,13 +200,11 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         window_shape0=np.shape(tensors[0])[1]
        # window = self.features_df.loc[stay_id:stay_id].to_numpy()
        # labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
-        
         if len(tensors[6]) == 1:
             # only one label per stay, align with window
             tensors[6] = np.concatenate([np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0)
 
         length_diff = self.maxlen - window_shape0
-
         pad_mask = np.ones(window_shape0)
 
         # Padding the array to fulfill size requirement
@@ -214,11 +212,11 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             # window shorter than the longest window in dataset, pad to same length
             tensors[0]= np.concatenate([tensors[0], np.ones((length_diff, np.shape(tensors[0])[0])) * pad_value], axis=0)
             tensors[1]= np.concatenate([tensors[1], np.ones((length_diff, np.shape(tensors[1])[0])) * pad_value], axis=0)
-            tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, np.shape(tensors[2])[0])) * pad_value], axis=0)
-            tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, np.shape(tensors[3])[0])) * pad_value], axis=0)
-            tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, np.shape(tensors[4])[0])) * pad_value], axis=0)
+          #  tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, np.shape(tensors[2])[0])) * pad_value], axis=0)
+          #  tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, np.shape(tensors[3])[0])) * pad_value], axis=0)
+          #  tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, np.shape(tensors[4])[0])) * pad_value], axis=0)
             tensors[5]= np.concatenate([tensors[5], np.ones((length_diff, np.shape(tensors[5])[0])) * pad_value], axis=0)
-            tensors[7]= np.concatenate([tensors[7], np.ones((length_diff, np.shape(tensors[7])[0])) * pad_value], axis=0)
+            tensors[7]= np.concatenate([tensors[7], np.ones((length_diff, np.shape(tensors[7])[0])) * stay_id], axis=0)
            # window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
             
             tensors[6] = np.concatenate([tensors[6], np.ones(length_diff) * pad_value], axis=0)
@@ -233,6 +231,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
        # data = window.astype(np.float32)
         tensors = [from_numpy(np.array(tensor)).to(float32) for tensor in tensors]
         tensors = [stack((x,), dim=-1) if x.numel() else empty(0) for x in tensors]
+        
         return  OrderedDict(zip(FEAT_NAMES, tensors)),from_numpy(pad_mask)
 
     
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 2240e2e1..e8572ca6 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -292,7 +292,7 @@ class TemporalFusionTransformer(DLPredictionWrapper):
 
     _supported_run_modes = [RunMode.classification, RunMode.regression]
     def __init__(self,num_classes, encoder_length,hidden,dropout,
-                 n_heads,dropout_att,example_length,*args,quantiles=[0.1, 0.5, 0.9],static_categorical_inp_size=[4],temporal_known_categorical_inp_size=[],
+                 n_heads,dropout_att,example_length,*args,quantiles=[0.1, 0.5, 0.9],static_categorical_inp_size=[2],temporal_known_categorical_inp_size=[],
     temporal_observed_categorical_inp_size=[],static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
     temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs):
         #derived variables
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index c4616679..e6f2e585 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -431,6 +431,7 @@ def __init__(self,static_categorical_inp_size,temporal_known_categorical_inp_siz
             nn.Embedding(n, self.hidden) for n in self.t_cat_k_inp_size]) if self.t_cat_k_inp_size else None
         self.t_cat_o_embed = nn.ModuleList([
             nn.Embedding(n, self.hidden) for n in self.t_cat_o_inp_size]) if self.t_cat_o_inp_size else None
+        
 
         if initialize_cont_params:
             self.s_cont_embedding_vectors = nn.Parameter(torch.Tensor(self.s_cont_inp_size, self.hidden)) if self.s_cont_inp_size else None
@@ -480,7 +481,8 @@ def _apply_embedding(self,
             cont_emb: Tensor,
             cont_bias: Tensor,
             ) -> Tuple[Optional[Tensor], Optional[Tensor]]:
-        e_cat = torch.stack([embed(cat[...,i]) for i, embed in enumerate(cat_emb)], dim=-2) if cat is not None else None
+        e_cat = torch.stack([embed(cat[...,i].int()) for i, embed in enumerate(cat_emb)], dim=-2) if (cat is not None) and (cat_emb is not None) else None 
+        
         if cont is not None:
             #the line below is equivalent to following einsums
             #e_cont = torch.einsum('btf,fh->bthf', cont, cont_emb)
@@ -508,12 +510,10 @@ def forward(self, x: Dict[str, Tensor]):
         t_cat_o_inp = x.get('o_cat', None)
         t_cont_o_inp = x.get('o_cont', None)
         t_tgt_obs = x['target'] # Has to be present
-
         # Static inputs are expected to be equal for all timesteps
         # For memory efficiency there is no assert statement
         s_cat_inp = s_cat_inp[:,0,:] if s_cat_inp is not None else None
         s_cont_inp = s_cont_inp[:,0,:] if s_cont_inp is not None else None
-
         s_inp = self._apply_embedding(s_cat_inp,
                                       s_cont_inp,
                                       self.s_cat_embed,
@@ -577,15 +577,16 @@ def initialize_parameters(self, x):
             t_cont_o_inp = x.get('o_cont', None)
             t_tgt_obs = x['target'] # Has to be present
 
-            if s_cont_inp is not None:
+
+            if (s_cont_inp is not None) and (s_cont_inp.size()[1]>0):
                 self.s_cont_embedding_vectors.materialize((s_cont_inp.shape[-1], self.hidden))
                 self.s_cont_embedding_bias.materialize((s_cont_inp.shape[-1], self.hidden))
 
-            if t_cont_k_inp is not None:
+            if (t_cont_k_inp is not None)  and (t_cont_k_inp.size()[1]>0):
                 self.t_cont_k_embedding_vectors.materialize((t_cont_k_inp.shape[-1], self.hidden))
                 self.t_cont_k_embedding_bias.materialize((t_cont_k_inp.shape[-1], self.hidden))
 
-            if t_cont_o_inp is not None:
+            if (t_cont_o_inp) is not None  and (t_cont_o_inp.size()[1]>0):
                 self.t_cont_o_embedding_vectors.materialize((t_cont_o_inp.shape[-1], self.hidden))
                 self.t_cont_o_embedding_bias.materialize((t_cont_o_inp.shape[-1], self.hidden))
 
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 30deb6a4..33730134 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -13,7 +13,7 @@
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
-
+from collections import OrderedDict
 cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
 
 
@@ -102,10 +102,13 @@ def train_common(
         pin_memory=True,
         drop_last=True,
     )
-    print(type(next(iter(train_loader))[0]))
-    data_shape = next(iter(train_loader))[0].shape
-
-    model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
+    if(isinstance(next(iter(train_loader))[0] ,OrderedDict)):
+         model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
+    
+    else:
+        data_shape = next(iter(train_loader))[0].shape
+    
+        model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
     model.set_weight(weight, train_dataset)
     if load_weights:
         if source_dir.exists():
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index fc1bd77e..d7ae32c6 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -8,7 +8,7 @@
 from torch.nn import MSELoss, CrossEntropyLoss
 from torch.nn.modules.loss import _Loss
 from torch.optim import Optimizer
-
+from collections import OrderedDict
 import inspect
 import gin
 import numpy as np
@@ -230,8 +230,11 @@ def softmax_multi_output_transform(output):
 
     def step_fn(self, element, step_prefix=""):
         """Perform a step in the training loop."""
-
-        if len(element) == 2:
+        if(isinstance(element[0] ,OrderedDict)):
+            data,mask=element[0], element[1].to(self.device)
+            for key, value in data.items():
+                data[key] = value.float().to(self.device)
+        elif len(element) == 2:
             data, labels = element[0], element[1].to(self.device)
             if isinstance(data, list):
                 for i in range(len(data)):

From 4a462d2d84a6057bb22c077fc1b06f3a9554f0f2 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Thu, 29 Jun 2023 17:13:05 +0200
Subject: [PATCH 026/142] fixed return shape of data

---
 .../TemporalFusionTransformer.gin             | 12 ++---
 icu_benchmarks/data/loader.py                 |  6 +--
 icu_benchmarks/models/dl_models.py            |  7 ++-
 icu_benchmarks/models/layers.py               | 46 +++++++++++++------
 icu_benchmarks/models/train.py                |  2 +-
 icu_benchmarks/models/wrappers.py             | 14 +++++-
 6 files changed, 57 insertions(+), 30 deletions(-)

diff --git a/configs/prediction_models/TemporalFusionTransformer.gin b/configs/prediction_models/TemporalFusionTransformer.gin
index 1d8ec015..48a0fb15 100644
--- a/configs/prediction_models/TemporalFusionTransformer.gin
+++ b/configs/prediction_models/TemporalFusionTransformer.gin
@@ -12,14 +12,14 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 
 # Encoder params
 model/hyperparameter.class_to_tune = @TemporalFusionTransformer
-model/hyperparameter.encoder_length = (1,14,"log-uniform", 2)
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.encoder_length = 24
+model/hyperparameter.hidden = 128
 model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.dropout = (0.0, 0.4)
-model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads = 1
+model/hyperparameter.dropout = 0.1
+model/hyperparameter.dropout_att = 0.0
+model/hyperparameter.n_heads = 4
 
-model/hyperparameter.example_length=(1,14,"log-uniform", 2)
+model/hyperparameter.example_length=48
 
 
 
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index b2b7ad72..bc8022aa 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -226,11 +226,11 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         if len(not_labeled) > 0:
             tensors[6][not_labeled] = -1
             pad_mask[not_labeled] = 0
-
+        tensors[6]=[tensors[6]]
         pad_mask = pad_mask.astype(bool)
        # data = window.astype(np.float32)
-        tensors = [from_numpy(np.array(tensor)).to(float32) for tensor in tensors]
-        tensors = [stack((x,), dim=-1) if x.numel() else empty(0) for x in tensors]
+        tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)
+        tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]
         
         return  OrderedDict(zip(FEAT_NAMES, tensors)),from_numpy(pad_mask)
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index e8572ca6..f61fb40b 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -306,7 +306,7 @@ def __init__(self,num_classes, encoder_length,hidden,dropout,
       
         super().__init__(num_classes=num_classes, encoder_length=encoder_length,hidden=hidden,
                  n_heads=n_heads,dropout_att=dropout_att,example_length=example_length,quantiles=quantiles,
-                 num_static_vars=num_static_vars,num_future_vars=num_future_vars,num_historic_vars=num_historic_vars,*args,static_categorical_inp_size=1,temporal_known_categorical_inp_size=4,
+                 num_static_vars=num_static_vars,num_future_vars=num_future_vars,num_historic_vars=num_historic_vars,*args,static_categorical_inp_size=1,temporal_known_categorical_inp_size=0,
                 temporal_observed_categorical_inp_size=0,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
                 temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs)
 
@@ -317,19 +317,18 @@ def __init__(self,num_classes, encoder_length,hidden,dropout,
         self.embedding = LazyEmbedding(static_categorical_inp_size,temporal_known_categorical_inp_size,
                 temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
                 temporal_observed_continuous_inp_size,temporal_target_size,hidden)
+        
         self.static_encoder = StaticCovariateEncoder(num_static_vars,hidden,dropout)
         self.TFTpart2 = TFTBack(encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
                 n_heads,dropout_att,example_length,quantiles)
         self.logit = nn.Linear(len(quantiles), num_classes)
 
     def forward(self, x: Dict[str, Tensor]) -> Tensor:
-        
+       # print(x.get('s_cont', None))
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
-
         # Static context
         cs, ce, ch, cc = self.static_encoder(s_inp)
         ch, cc = ch.unsqueeze(0), cc.unsqueeze(0) #lstm initial states
-
         # Temporal input
         _historical_inputs = [t_known_inp[:,:self.encoder_length,:], t_observed_tgt[:,:self.encoder_length,:]]
         if t_observed_inp is not None:
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index e6f2e585..ef4f9e74 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -360,8 +360,12 @@ def __init__(self,
                  context_hidden=None,
                  dropout=0.0,):
         super().__init__()
+        
         self.layer_norm = MaybeLayerNorm(output_size, hidden, eps=1e-3)
+        
         self.lin_a = nn.Linear(input_size, hidden)
+        
+        
         if context_hidden is not None:
             self.lin_c = nn.Linear(context_hidden, hidden, bias=False)
         else:
@@ -372,7 +376,9 @@ def __init__(self,
         self.out_proj = nn.Linear(input_size, output_size) if output_size else None
 
     def forward(self, a: Tensor, c: Optional[Tensor] = None):
+        
         x = self.lin_a(a)
+        
         if c is not None:
             x = x + self.lin_c(c).unsqueeze(1)
         x = F.elu(x)
@@ -381,16 +387,18 @@ def forward(self, a: Tensor, c: Optional[Tensor] = None):
         x = self.glu(x)
         y = a if self.out_proj is None else self.out_proj(a)
         x = x + y
+        
         return self.layer_norm(x) 
 
 
 # @torch.jit.script #Currently broken with autocast
 def fused_pointwise_linear_v1(x, a, b):
+    
     out = torch.mul(x.unsqueeze(-1), a)
     out = out + b
     return out
 
-@torch.jit.script
+#@torch.jit.script
 def fused_pointwise_linear_v2(x, a, b):
     out = x.unsqueeze(3) * a
     out = out + b
@@ -438,7 +446,6 @@ def __init__(self,static_categorical_inp_size,temporal_known_categorical_inp_siz
             self.t_cont_k_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_k_inp_size, self.hidden)) if self.t_cont_k_inp_size else None
             self.t_cont_o_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_o_inp_size, self.hidden)) if self.t_cont_o_inp_size else None
             self.t_tgt_embedding_vectors = nn.Parameter(torch.Tensor(self.t_tgt_size, self.hidden))
-
             self.s_cont_embedding_bias = nn.Parameter(torch.zeros(self.s_cont_inp_size, self.hidden)) if self.s_cont_inp_size else None
             self.t_cont_k_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_k_inp_size, self.hidden)) if self.t_cont_k_inp_size else None
             self.t_cont_o_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_o_inp_size, self.hidden)) if self.t_cont_o_inp_size else None
@@ -446,7 +453,6 @@ def __init__(self,static_categorical_inp_size,temporal_known_categorical_inp_siz
 
             self.reset_parameters()
 
-
     def reset_parameters(self):
         ''''
         embeddings are initilitized using xavier's method and biases are initlitized with zeros
@@ -481,18 +487,20 @@ def _apply_embedding(self,
             cont_emb: Tensor,
             cont_bias: Tensor,
             ) -> Tuple[Optional[Tensor], Optional[Tensor]]:
-        e_cat = torch.stack([embed(cat[...,i].int()) for i, embed in enumerate(cat_emb)], dim=-2) if (cat is not None) and (cat_emb is not None) else None 
         
-        if cont is not None:
+        e_cat = torch.stack([embed(cat[...,i].int()) for i, embed in enumerate(cat_emb)], dim=-2) if (cat is not None) and (cat.size()[1]>0) else None 
+        if (cont is not None ) and (cont.size()[1]>0):
             #the line below is equivalent to following einsums
             #e_cont = torch.einsum('btf,fh->bthf', cont, cont_emb)
             #e_cont = torch.einsum('bf,fh->bhf', cont, cont_emb)
-            
-            e_cont = fused_pointwise_linear_v1(cont, cont_emb, cont_bias)
+            e_cont = torch.mul(cont.unsqueeze(-1), cont_emb)
+            e_cont = e_cont + cont_bias
+           # e_cont = fused_pointwise_linear_v1(cont, cont_emb, cont_bias)
         else:
             e_cont = None
-
+        
         if e_cat is not None and e_cont is not None:
+           
             return torch.cat([e_cat, e_cont], dim=-2)
         elif e_cat is not None:
             return e_cat
@@ -512,13 +520,21 @@ def forward(self, x: Dict[str, Tensor]):
         t_tgt_obs = x['target'] # Has to be present
         # Static inputs are expected to be equal for all timesteps
         # For memory efficiency there is no assert statement
+        
+        
+        
         s_cat_inp = s_cat_inp[:,0,:] if s_cat_inp is not None else None
         s_cont_inp = s_cont_inp[:,0,:] if s_cont_inp is not None else None
+       
+        
+        
         s_inp = self._apply_embedding(s_cat_inp,
                                       s_cont_inp,
                                       self.s_cat_embed,
                                       self.s_cont_embedding_vectors,
                                       self.s_cont_embedding_bias)
+                                      
+        
         t_known_inp = self._apply_embedding(t_cat_k_inp,
                                             t_cont_k_inp,
                                             self.t_cat_k_embed,
@@ -531,7 +547,8 @@ def forward(self, x: Dict[str, Tensor]):
                                                self.t_cont_o_embedding_bias)
 
         # Temporal observed targets
-        t_observed_tgt = fused_pointwise_linear_v2(t_tgt_obs, self.t_tgt_embedding_vectors, self.t_tgt_embedding_bias)
+        t_observed_tgt = torch.matmul(t_tgt_obs.unsqueeze(3).unsqueeze(4), self.t_tgt_embedding_vectors.unsqueeze(1)).squeeze(3)
+        t_observed_tgt = t_observed_tgt + self.t_tgt_embedding_bias
 
         return s_inp, t_known_inp, t_observed_inp, t_observed_tgt
 
@@ -544,14 +561,12 @@ def __init__(self, static_categorical_inp_size,temporal_known_categorical_inp_si
         super().__init__(static_categorical_inp_size,temporal_known_categorical_inp_size,
     temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
     temporal_observed_continuous_inp_size,temporal_target_size,hidden, initialize_cont_params=False)
-
         if static_continuous_inp_size:
             self.s_cont_embedding_vectors = UninitializedParameter()
             self.s_cont_embedding_bias = UninitializedParameter()
         else:
             self.s_cont_embedding_vectors = None
             self.s_cont_embedding_bias = None
-
         if temporal_known_continuous_inp_size:
             self.t_cont_k_embedding_vectors = UninitializedParameter()
             self.t_cont_k_embedding_bias = UninitializedParameter()
@@ -565,7 +580,6 @@ def __init__(self, static_categorical_inp_size,temporal_known_categorical_inp_si
         else:
             self.t_cont_o_embedding_vectors = None
             self.t_cont_o_embedding_bias = None
-
         self.t_tgt_embedding_vectors = UninitializedParameter()
         self.t_tgt_embedding_bias = UninitializedParameter()
 
@@ -576,8 +590,7 @@ def initialize_parameters(self, x):
             t_cont_k_inp = x.get('k_cont', None)
             t_cont_o_inp = x.get('o_cont', None)
             t_tgt_obs = x['target'] # Has to be present
-
-
+            print(s_cont_inp.shape)
             if (s_cont_inp is not None) and (s_cont_inp.size()[1]>0):
                 self.s_cont_embedding_vectors.materialize((s_cont_inp.shape[-1], self.hidden))
                 self.s_cont_embedding_bias.materialize((s_cont_inp.shape[-1], self.hidden))
@@ -602,6 +615,7 @@ class VariableSelectionNetwork(nn.Module):
     '''
     def __init__(self, hidden,dropout, num_inputs):
         super().__init__()
+       
         self.joint_grn = GRN(hidden*num_inputs, hidden, output_size=num_inputs, context_hidden=hidden)
         self.var_grns = nn.ModuleList([GRN(hidden, hidden, dropout=dropout) for _ in range(num_inputs)])
 
@@ -614,6 +628,7 @@ def forward(self, x: Tensor, context: Optional[Tensor] = None):
         #the line below performs batched matrix vector multiplication
         #for temporal features it's bthf,btf->bth
         #for static features it's bhf,bf->bh
+        
         variable_ctx = torch.matmul(transformed_embed, sparse_weights.unsqueeze(-1)).squeeze(-1)
 
         return variable_ctx, sparse_weights
@@ -629,8 +644,9 @@ def __init__(self, num_static_vars,hidden,dropout):
         self.context_grns = nn.ModuleList([GRN(hidden, hidden, dropout=dropout) for _ in range(4)])
 
     def forward(self, x: Tensor) -> Tuple[Tensor, Tensor, Tensor, Tensor]:
+        
         variable_ctx, sparse_weights = self.vsn(x)
-
+        
         # Context vectors:
         # variable selection context
         # enrichment context
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 33730134..9a7284a3 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -135,7 +135,7 @@ def train_common(
         max_epochs=epochs if model.needs_training else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="auto" if not cpu else "cpu",
+        accelerator="cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic=reproducible,
         benchmark=not reproducible,
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index d7ae32c6..ee7cc8f1 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -232,8 +232,20 @@ def step_fn(self, element, step_prefix=""):
         """Perform a step in the training loop."""
         if(isinstance(element[0] ,OrderedDict)):
             data,mask=element[0], element[1].to(self.device)
+
             for key, value in data.items():
-                data[key] = value.float().to(self.device)
+                
+                value = value.float().to(self.device)
+
+                
+
+                if value.shape[-1] == 1:
+                    value = value.squeeze(-1)
+                if value.dim() == 3:
+                    value = value.permute(0, 2, 1)
+
+                data[key] = value
+                
         elif len(element) == 2:
             data, labels = element[0], element[1].to(self.device)
             if isinstance(data, list):

From 1114690ecbd71926a10248324c2abc79d51895c1 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Thu, 29 Jun 2023 17:15:09 +0200
Subject: [PATCH 027/142] fixed return shape of data

---
 icu_benchmarks/models/train.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 9a7284a3..33730134 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -135,7 +135,7 @@ def train_common(
         max_epochs=epochs if model.needs_training else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="cpu",
+        accelerator="auto" if not cpu else "cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic=reproducible,
         benchmark=not reproducible,

From 2a89d216e81cf9ce64f76978169a0803d7c3e2ee Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 4 Jul 2023 12:35:37 +0200
Subject: [PATCH 028/142] bug fixes and checks

---
 .../TemporalFusionTransformer.gin             |  9 +++---
 icu_benchmarks/data/constants.py              |  3 ++
 icu_benchmarks/data/loader.py                 | 10 ++++---
 icu_benchmarks/models/dl_models.py            | 30 ++++++++++++++-----
 icu_benchmarks/models/layers.py               | 19 ++++++------
 icu_benchmarks/models/wrappers.py             |  5 ++--
 6 files changed, 48 insertions(+), 28 deletions(-)

diff --git a/configs/prediction_models/TemporalFusionTransformer.gin b/configs/prediction_models/TemporalFusionTransformer.gin
index 48a0fb15..b2a71b4c 100644
--- a/configs/prediction_models/TemporalFusionTransformer.gin
+++ b/configs/prediction_models/TemporalFusionTransformer.gin
@@ -12,14 +12,13 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 
 # Encoder params
 model/hyperparameter.class_to_tune = @TemporalFusionTransformer
-model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = 128
+model/hyperparameter.encoder_length = 20
+model/hyperparameter.hidden = 64
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = 0.1
 model/hyperparameter.dropout_att = 0.0
-model/hyperparameter.n_heads = 4
-
-model/hyperparameter.example_length=48
+model/hyperparameter.n_heads = 1
+model/hyperparameter.example_length=21
 
 
 
diff --git a/icu_benchmarks/data/constants.py b/icu_benchmarks/data/constants.py
index 031f35ca..c7f6c635 100644
--- a/icu_benchmarks/data/constants.py
+++ b/icu_benchmarks/data/constants.py
@@ -15,3 +15,6 @@ class VarType:
     group = "GROUP"
     sequence = "SEQUENCE"
     label = "LABEL"
+
+class FeatType:
+    FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target', 'id']
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index bc8022aa..7289e312 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -9,8 +9,8 @@
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
+from .constants import FeatType as Features
 from collections import OrderedDict
-FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target', 'id']
 
 class CommonDataset(Dataset):
     """Common dataset: subclass of Torch Dataset that represents the data to learn on.
@@ -147,6 +147,7 @@ def get_data_and_labels(self) -> Tuple[np.array, np.array]:
         if len(labels) == self.num_stays:
             # order of groups could be random, we make sure not to change it
             rep = rep.groupby(level=self.vars["GROUP"], sort=False).last()
+     #   print('after')
         rep = rep.to_numpy().astype(float)
 
         return rep, labels
@@ -180,11 +181,12 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             return self._cached_dataset[idx]
 
         pad_value = 0.0
-        stay_id = self.outcome_df.index.unique()[idx]  # [self.vars["GROUP"]]
+        stay_id = self.outcome_df.index.unique()[idx]  
 
         
         # We need to be sure that tensors are returned in the correct order to be processed correclty by tft
         tensors = [[] for _ in range(8)]
+        
         for var in self.features_df.columns:
             if  var == 'sex' :
                 tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
@@ -206,7 +208,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         length_diff = self.maxlen - window_shape0
         pad_mask = np.ones(window_shape0)
-
+        print(np.shape(tensors[0]))
         # Padding the array to fulfill size requirement
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
@@ -232,7 +234,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)
         tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]
         
-        return  OrderedDict(zip(FEAT_NAMES, tensors)),from_numpy(pad_mask)
+        return  OrderedDict(zip(Features.FEAT_NAMES, tensors)),from_numpy(pad_mask)
 
     
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index f61fb40b..a34ad0a4 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -294,7 +294,7 @@ class TemporalFusionTransformer(DLPredictionWrapper):
     def __init__(self,num_classes, encoder_length,hidden,dropout,
                  n_heads,dropout_att,example_length,*args,quantiles=[0.1, 0.5, 0.9],static_categorical_inp_size=[2],temporal_known_categorical_inp_size=[],
     temporal_observed_categorical_inp_size=[],static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
-    temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs):
+    temporal_observed_continuous_inp_size=96,temporal_target_size=1,**kwargs):
         #derived variables
         num_static_vars=len(static_categorical_inp_size)+static_continuous_inp_size
         num_future_vars=len(temporal_known_categorical_inp_size)+temporal_known_continuous_inp_size
@@ -303,7 +303,7 @@ def __init__(self,num_classes, encoder_length,hidden,dropout,
                                       temporal_target_size,
                                       len(temporal_observed_categorical_inp_size),
                                       ])
-      
+        
         super().__init__(num_classes=num_classes, encoder_length=encoder_length,hidden=hidden,
                  n_heads=n_heads,dropout_att=dropout_att,example_length=example_length,quantiles=quantiles,
                  num_static_vars=num_static_vars,num_future_vars=num_future_vars,num_historic_vars=num_historic_vars,*args,static_categorical_inp_size=1,temporal_known_categorical_inp_size=0,
@@ -324,18 +324,32 @@ def __init__(self,num_classes, encoder_length,hidden,dropout,
         self.logit = nn.Linear(len(quantiles), num_classes)
 
     def forward(self, x: Dict[str, Tensor]) -> Tensor:
-       # print(x.get('s_cont', None))
+        
+       
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
         # Static context
         cs, ce, ch, cc = self.static_encoder(s_inp)
         ch, cc = ch.unsqueeze(0), cc.unsqueeze(0) #lstm initial states
         # Temporal input
-        _historical_inputs = [t_known_inp[:,:self.encoder_length,:], t_observed_tgt[:,:self.encoder_length,:]]
+        
+        _historical_inputs = []
+        
+        # Check for t_observed_inp
         if t_observed_inp is not None:
-            _historical_inputs.insert(0,t_observed_inp[:,:self.encoder_length,:])
-
+            _historical_inputs.append(t_observed_inp[:, :self.encoder_length, :])
+
+        # Check for t_known_inp
+        if t_known_inp is not None:
+            _historical_inputs.append(t_known_inp[:, :self.encoder_length, :])
+        # Check for t_observed_tgt
+        if t_observed_tgt is not None:
+            _historical_inputs.append(t_observed_tgt[:, :self.encoder_length, :])
         historical_inputs = cat(_historical_inputs, dim=-2)
-        future_inputs = t_known_inp[:, self.encoder_length:]
-        o=self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs)
+        future_inputs= Tensor()
+        if t_known_inp is not None:
+            future_inputs = t_known_inp[:, self.encoder_length:]
+        
+        
+        o=self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
         pred = self.logit(o)
         return pred
\ No newline at end of file
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index ef4f9e74..642136d2 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -590,7 +590,6 @@ def initialize_parameters(self, x):
             t_cont_k_inp = x.get('k_cont', None)
             t_cont_o_inp = x.get('o_cont', None)
             t_tgt_obs = x['target'] # Has to be present
-            print(s_cont_inp.shape)
             if (s_cont_inp is not None) and (s_cont_inp.size()[1]>0):
                 self.s_cont_embedding_vectors.materialize((s_cont_inp.shape[-1], self.hidden))
                 self.s_cont_embedding_bias.materialize((s_cont_inp.shape[-1], self.hidden))
@@ -615,22 +614,24 @@ class VariableSelectionNetwork(nn.Module):
     '''
     def __init__(self, hidden,dropout, num_inputs):
         super().__init__()
-       
+        self.hidden=hidden
         self.joint_grn = GRN(hidden*num_inputs, hidden, output_size=num_inputs, context_hidden=hidden)
         self.var_grns = nn.ModuleList([GRN(hidden, hidden, dropout=dropout) for _ in range(num_inputs)])
 
     def forward(self, x: Tensor, context: Optional[Tensor] = None):
+        if x.numel() == 0:  # Check if x is an empty tensor
+            batch_size = context.size(0) if context is not None else 1
+            variable_ctx = torch.zeros(batch_size,1, self.hidden,device=x.device)
+            sparse_weights = torch.ones(batch_size,1, self.hidden,device=x.device)
+            return variable_ctx, sparse_weights
+
         Xi = torch.flatten(x, start_dim=-2)
         grn_outputs = self.joint_grn(Xi, c=context)
         sparse_weights = F.softmax(grn_outputs, dim=-1)
         transformed_embed_list = [m(x[...,i,:]) for i, m in enumerate(self.var_grns)]
         transformed_embed = torch.stack(transformed_embed_list, dim=-1)
-        #the line below performs batched matrix vector multiplication
-        #for temporal features it's bthf,btf->bth
-        #for static features it's bhf,bf->bh
-        
         variable_ctx = torch.matmul(transformed_embed, sparse_weights.unsqueeze(-1)).squeeze(-1)
-
+        
         return variable_ctx, sparse_weights
 
 class StaticCovariateEncoder(nn.Module):
@@ -685,7 +686,6 @@ def forward(self, x: Tensor) -> Tuple[Tensor, Tensor]:
         # attn_score = torch.einsum('bind,bjnd->bnij', q, k)
         attn_score = torch.matmul(q.permute((0, 2, 1, 3)), k.permute((0, 2, 3, 1)))
         attn_score.mul_(self.scale)
-
         attn_score = attn_score + self._mask
 
         attn_prob = F.softmax(attn_score, dim=3)
@@ -741,8 +741,9 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         historical_features, _ = self.history_vsn(historical_inputs, cs)
         history, state = self.history_encoder(historical_features, (ch, cc))
         future_features, _ = self.future_vsn(future_inputs, cs)
+
         future, _ = self.future_encoder(future_features, state)
-        torch.cuda.synchronize()
+        
 
         # skip connection
         input_embedding = torch.cat([historical_features, future_features], dim=1)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index ee7cc8f1..759ec7f4 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -234,7 +234,6 @@ def step_fn(self, element, step_prefix=""):
             data,mask=element[0], element[1].to(self.device)
 
             for key, value in data.items():
-                
                 value = value.float().to(self.device)
 
                 
@@ -243,7 +242,8 @@ def step_fn(self, element, step_prefix=""):
                     value = value.squeeze(-1)
                 if value.dim() == 3:
                     value = value.permute(0, 2, 1)
-
+                if(key=='target'):
+                    labels=value.squeeze()
                 data[key] = value
                 
         elif len(element) == 2:
@@ -270,6 +270,7 @@ def step_fn(self, element, step_prefix=""):
         else:
             aux_loss = 0
         prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
+        
         target = torch.masked_select(labels, mask).to(self.device)
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task

From b757c66bd7a93cae2cf184a8d8ce7cb451cf8b84 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 4 Jul 2023 13:06:40 +0200
Subject: [PATCH 029/142] fixed concat to be along the correct axis

---
 icu_benchmarks/data/loader.py | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 7289e312..9e848fc6 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -205,25 +205,26 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         if len(tensors[6]) == 1:
             # only one label per stay, align with window
             tensors[6] = np.concatenate([np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0)
-
+        
         length_diff = self.maxlen - window_shape0
-        pad_mask = np.ones(window_shape0)
         print(np.shape(tensors[0]))
+        print(np.shape(np.ones((length_diff, np.shape(tensors[0])[0]))))
+        pad_mask = np.ones(window_shape0)
         # Padding the array to fulfill size requirement
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
-            tensors[0]= np.concatenate([tensors[0], np.ones((length_diff, np.shape(tensors[0])[0])) * pad_value], axis=0)
-            tensors[1]= np.concatenate([tensors[1], np.ones((length_diff, np.shape(tensors[1])[0])) * pad_value], axis=0)
+            tensors[0]= np.concatenate([tensors[0], np.ones(( np.shape(tensors[0])[0],length_diff)) * pad_value], axis=1)
+            tensors[1]= np.concatenate([tensors[1], np.ones(( np.shape(tensors[1])[0],length_diff)) * pad_value], axis=1)
           #  tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, np.shape(tensors[2])[0])) * pad_value], axis=0)
           #  tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, np.shape(tensors[3])[0])) * pad_value], axis=0)
           #  tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, np.shape(tensors[4])[0])) * pad_value], axis=0)
-            tensors[5]= np.concatenate([tensors[5], np.ones((length_diff, np.shape(tensors[5])[0])) * pad_value], axis=0)
-            tensors[7]= np.concatenate([tensors[7], np.ones((length_diff, np.shape(tensors[7])[0])) * stay_id], axis=0)
+            tensors[5]= np.concatenate([tensors[5], np.ones(( np.shape(tensors[5])[0],length_diff)) * pad_value], axis=1)
+            tensors[7]= np.concatenate([tensors[7], np.ones(( np.shape(tensors[7])[0],length_diff)) * stay_id], axis=1)
            # window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
             
             tensors[6] = np.concatenate([tensors[6], np.ones(length_diff) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
-
+            print(np.shape(tensors[0]))
         not_labeled = np.argwhere(np.isnan(tensors[6]))
         if len(not_labeled) > 0:
             tensors[6][not_labeled] = -1

From 273b9aea3a76360ffd2ce037545a10b0c40848db Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 4 Jul 2023 14:25:11 +0200
Subject: [PATCH 030/142] regression is working now with quantile loss

---
 .../TemporalFusionTransformer.gin             |  2 +-
 configs/tasks/RegressionTFT.gin               |  2 +-
 configs/tasks/common/Imports.gin              |  2 +-
 experiments/demo_benchmark_classification.yml | 10 ++++-----
 icu_benchmarks/data/loader.py                 | 21 +++++++++----------
 icu_benchmarks/models/metrics.py              | 18 ++++++++++++++--
 icu_benchmarks/models/train.py                |  3 ++-
 icu_benchmarks/models/wrappers.py             |  2 +-
 8 files changed, 37 insertions(+), 23 deletions(-)

diff --git a/configs/prediction_models/TemporalFusionTransformer.gin b/configs/prediction_models/TemporalFusionTransformer.gin
index b2a71b4c..74dde441 100644
--- a/configs/prediction_models/TemporalFusionTransformer.gin
+++ b/configs/prediction_models/TemporalFusionTransformer.gin
@@ -13,7 +13,7 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 # Encoder params
 model/hyperparameter.class_to_tune = @TemporalFusionTransformer
 model/hyperparameter.encoder_length = 20
-model/hyperparameter.hidden = 64
+model/hyperparameter.hidden = 12
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = 0.1
 model/hyperparameter.dropout_att = 0.0
diff --git a/configs/tasks/RegressionTFT.gin b/configs/tasks/RegressionTFT.gin
index b31408fb..a1ff4d55 100644
--- a/configs/tasks/RegressionTFT.gin
+++ b/configs/tasks/RegressionTFT.gin
@@ -16,7 +16,7 @@ train_common.ram_cache = True
 
 # LOSS FUNCTION
 DLPredictionWrapper.loss = @QuantileLoss
-
+QuantileLoss.quantiles=[0.1, 0.5, 0.9]
 
 # SELECTING PREPROCESSOR
 preprocess.preprocessor = @base_regression_preprocessor
diff --git a/configs/tasks/common/Imports.gin b/configs/tasks/common/Imports.gin
index faf3c8d6..b3da4a80 100644
--- a/configs/tasks/common/Imports.gin
+++ b/configs/tasks/common/Imports.gin
@@ -4,4 +4,4 @@ import icu_benchmarks.data.loader
 import icu_benchmarks.models.wrappers
 import icu_benchmarks.models.dl_models
 import icu_benchmarks.models.ml_models
-import icu_benchmarks.models.layers
\ No newline at end of file
+import icu_benchmarks.models.metrics
\ No newline at end of file
diff --git a/experiments/demo_benchmark_classification.yml b/experiments/demo_benchmark_classification.yml
index 15802f9b..150bac03 100644
--- a/experiments/demo_benchmark_classification.yml
+++ b/experiments/demo_benchmark_classification.yml
@@ -18,11 +18,11 @@ parameters:
   data_dir:
     values:
       - demo_data/mortality24/eicu_demo
-      #- demo_data/mortality24/mimic_demo
-     # - demo_data/aki/eicu_demo
-    #  - demo_data/aki/mimic_demo
-   #   - demo_data/sepsis/eicu_demo
-  #    - demo_data/sepsis/mimic_demo
+      - demo_data/mortality24/mimic_demo
+      - demo_data/aki/eicu_demo
+      - demo_data/aki/mimic_demo
+      - demo_data/sepsis/eicu_demo
+      - demo_data/sepsis/mimic_demo
   model:
     values:
    #   - LogisticRegression
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 9e848fc6..c576d857 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -147,7 +147,6 @@ def get_data_and_labels(self) -> Tuple[np.array, np.array]:
         if len(labels) == self.num_stays:
             # order of groups could be random, we make sure not to change it
             rep = rep.groupby(level=self.vars["GROUP"], sort=False).last()
-     #   print('after')
         rep = rep.to_numpy().astype(float)
 
         return rep, labels
@@ -198,7 +197,6 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         
         tensors[6].extend(self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float))
         tensors[7].append(np.asarray([stay_id]))
-        
         window_shape0=np.shape(tensors[0])[1]
        # window = self.features_df.loc[stay_id:stay_id].to_numpy()
        # labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
@@ -207,24 +205,24 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             tensors[6] = np.concatenate([np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0)
         
         length_diff = self.maxlen - window_shape0
-        print(np.shape(tensors[0]))
-        print(np.shape(np.ones((length_diff, np.shape(tensors[0])[0]))))
+        length_diff_id=self.maxlen - np.shape(tensors[7])[1]
         pad_mask = np.ones(window_shape0)
         # Padding the array to fulfill size requirement
+        
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
-            tensors[0]= np.concatenate([tensors[0], np.ones(( np.shape(tensors[0])[0],length_diff)) * pad_value], axis=1)
-            tensors[1]= np.concatenate([tensors[1], np.ones(( np.shape(tensors[1])[0],length_diff)) * pad_value], axis=1)
+            tensors[0]= np.concatenate([tensors[0], np.ones(( np.shape(tensors[0])[0],self.maxlen - np.shape(tensors[0])[1])) * pad_value], axis=1)
+            tensors[1]= np.concatenate([tensors[1], np.ones(( np.shape(tensors[1])[0],self.maxlen - np.shape(tensors[1])[1])) * pad_value], axis=1)
           #  tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, np.shape(tensors[2])[0])) * pad_value], axis=0)
           #  tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, np.shape(tensors[3])[0])) * pad_value], axis=0)
           #  tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, np.shape(tensors[4])[0])) * pad_value], axis=0)
-            tensors[5]= np.concatenate([tensors[5], np.ones(( np.shape(tensors[5])[0],length_diff)) * pad_value], axis=1)
-            tensors[7]= np.concatenate([tensors[7], np.ones(( np.shape(tensors[7])[0],length_diff)) * stay_id], axis=1)
+            tensors[5]= np.concatenate([tensors[5], np.ones(( np.shape(tensors[5])[0],self.maxlen - np.shape(tensors[5])[1])) * pad_value], axis=1)
+            
            # window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
             
-            tensors[6] = np.concatenate([tensors[6], np.ones(length_diff) * pad_value], axis=0)
+            tensors[6] = np.concatenate([tensors[6], np.ones(self.maxlen - np.shape(tensors[6])[0]) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
-            print(np.shape(tensors[0]))
+        tensors[7]= np.concatenate([tensors[7], np.ones(( np.shape(tensors[7])[0],self.maxlen - np.shape(tensors[7])[1])) * stay_id], axis=1)#should be done regardless of length_diff
         not_labeled = np.argwhere(np.isnan(tensors[6]))
         if len(not_labeled) > 0:
             tensors[6][not_labeled] = -1
@@ -232,9 +230,10 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         tensors[6]=[tensors[6]]
         pad_mask = pad_mask.astype(bool)
        # data = window.astype(np.float32)
+        
+        
         tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)
         tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]
-        
         return  OrderedDict(zip(Features.FEAT_NAMES, tensors)),from_numpy(pad_mask)
 
     
diff --git a/icu_benchmarks/models/metrics.py b/icu_benchmarks/models/metrics.py
index 280a0653..108f65da 100644
--- a/icu_benchmarks/models/metrics.py
+++ b/icu_benchmarks/models/metrics.py
@@ -5,13 +5,27 @@
 from sklearn.metrics import balanced_accuracy_score, mean_absolute_error
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
-
+import gin 
+import torch.nn.functional as F
 """"
 This file contains metrics that are not available in ignite.metrics. Specifically, it adds transformation capabilities to some
 metrics.
 """
 
-
+class QuantileLossClass(torch.nn.Module):
+    def __init__(self, quantiles):
+        super().__init__()
+        self.register_buffer('q', torch.tensor(quantiles))
+
+    def forward(self, predictions, targets):
+        diff = predictions - targets
+        ql = (1-self.q)*F.relu(diff) + self.q*F.relu(-diff)
+        losses = ql.view(-1, ql.shape[-1]).mean(0)
+        return losses
+@gin.configurable
+def Quantileloss(self,predictions, targets,quantiles=[0.1,0.5,0.9]):
+    loss=QuantileLossClass(quantiles=quantiles)
+    return loss(predictions,targets)
 def accuracy(output, target, topk=(1,)):
     """Computes the accuracy over the k top predictions for the specified values of k"""
     with torch.no_grad():
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 33730134..c5def311 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -102,6 +102,7 @@ def train_common(
         pin_memory=True,
         drop_last=True,
     )
+    
     if(isinstance(next(iter(train_loader))[0] ,OrderedDict)):
          model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
     
@@ -135,7 +136,7 @@ def train_common(
         max_epochs=epochs if model.needs_training else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="auto" if not cpu else "cpu",
+        accelerator= "cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic=reproducible,
         benchmark=not reproducible,
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 759ec7f4..5872acb3 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -20,7 +20,6 @@
 from pytorch_lightning import LightningModule
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
-
 gin.config.external_configurable(torch.nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(torch.nn.functional.cross_entropy, module="torch.nn.functional")
 gin.config.external_configurable(torch.nn.functional.mse_loss, module="torch.nn.functional")
@@ -278,6 +277,7 @@ def step_fn(self, element, step_prefix=""):
             # torch.long because NLL
         elif self.run_mode == RunMode.regression:
             # Regression task
+            
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
             raise ValueError(f"Run mode {self.run_mode} not supported.")

From 5b97aacef67efe35089036edbb693e9eca05c476 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 4 Jul 2023 15:56:27 +0200
Subject: [PATCH 031/142] changed gin file

---
 .../prediction_models/TemporalFusionTransformer.gin  | 12 ++++++------
 experiments/demo_benchmark_classification.yml        |  2 +-
 icu_benchmarks/models/layers.py                      | 11 -----------
 icu_benchmarks/models/train.py                       |  2 +-
 icu_benchmarks/models/utils.py                       |  5 +++++
 5 files changed, 13 insertions(+), 19 deletions(-)

diff --git a/configs/prediction_models/TemporalFusionTransformer.gin b/configs/prediction_models/TemporalFusionTransformer.gin
index 74dde441..870213ab 100644
--- a/configs/prediction_models/TemporalFusionTransformer.gin
+++ b/configs/prediction_models/TemporalFusionTransformer.gin
@@ -12,13 +12,13 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 
 # Encoder params
 model/hyperparameter.class_to_tune = @TemporalFusionTransformer
-model/hyperparameter.encoder_length = 20
-model/hyperparameter.hidden = 12
+model/hyperparameter.encoder_length = 24
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
 model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.dropout = 0.1
-model/hyperparameter.dropout_att = 0.0
-model/hyperparameter.n_heads = 1
-model/hyperparameter.example_length=21
+model/hyperparameter.dropout = (0.0, 0.4)
+model/hyperparameter.dropout_att = (0.0, 0.4)
+model/hyperparameter.n_heads =heads_divisible_by_hidden(model/hyperparameter.hidden)
+model/hyperparameter.example_length=25
 
 
 
diff --git a/experiments/demo_benchmark_classification.yml b/experiments/demo_benchmark_classification.yml
index 150bac03..58b14a45 100644
--- a/experiments/demo_benchmark_classification.yml
+++ b/experiments/demo_benchmark_classification.yml
@@ -11,7 +11,7 @@ command:
   - --tune
   - --wandb-sweep
   - -gc
-  - -lc
+  - -lc 
 method: grid
 name: yaib_demo_classification_benchmark
 parameters:
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 642136d2..016cb366 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -778,14 +778,3 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         out = self.quantile_proj(x)
 
         return out
-@gin.configurable
-class QuantileLoss(torch.nn.Module):
-    def __init__(self, quantiles):
-        super().__init__()
-        self.register_buffer('q', torch.tensor(quantiles))
-
-    def forward(self, predictions, targets):
-        diff = predictions - targets
-        ql = (1-self.q)*F.relu(diff) + self.q*F.relu(-diff)
-        losses = ql.view(-1, ql.shape[-1]).mean(0)
-        return losses
\ No newline at end of file
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index c5def311..f358483f 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -136,7 +136,7 @@ def train_common(
         max_epochs=epochs if model.needs_training else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator= "cpu",
+        accelerator="auto" if not cpu else "cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic=reproducible,
         benchmark=not reproducible,
diff --git a/icu_benchmarks/models/utils.py b/icu_benchmarks/models/utils.py
index 53bfcae2..ddc97028 100644
--- a/icu_benchmarks/models/utils.py
+++ b/icu_benchmarks/models/utils.py
@@ -188,3 +188,8 @@ def version(self):
     @rank_zero_only
     def log_hyperparams(self, params):
         pass
+@gin.configurable
+def heads_divisible_by_hidden(hidden):
+    heads_range = range(1, 9)  # Range for heads from 1 to 8
+    divisible_heads = (heads for heads in heads_range if hidden % heads == 0)
+    return divisible_heads
\ No newline at end of file

From f767499f90e6a76cb9fe6e12e359552ff0398ab8 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 11 Jul 2023 14:57:07 +0200
Subject: [PATCH 032/142] fixing bugs to get tftpytorch to work

---
 .../TemporalFusionTransformer.gin             | 25 -----
 configs/prediction_models/common/DLCommon.gin |  1 +
 experiments/demo_benchmark_regression.yml     | 16 ++--
 icu_benchmarks/data/loader.py                 | 91 ++++++++++++++++---
 icu_benchmarks/models/dl_models.py            | 29 +++++-
 icu_benchmarks/models/train.py                | 61 +++++++------
 6 files changed, 149 insertions(+), 74 deletions(-)
 delete mode 100644 configs/prediction_models/TemporalFusionTransformer.gin

diff --git a/configs/prediction_models/TemporalFusionTransformer.gin b/configs/prediction_models/TemporalFusionTransformer.gin
deleted file mode 100644
index 870213ab..00000000
--- a/configs/prediction_models/TemporalFusionTransformer.gin
+++ /dev/null
@@ -1,25 +0,0 @@
-# Settings for TFT model.
-
-# Common settings for DL models
-include "configs/prediction_models/common/DLCommon.gin"
-
-# Optimizer params
-train_common.model = @TemporalFusionTransformer
-
-optimizer/hyperparameter.class_to_tune = @Adam
-optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = (1e-5, 3e-4)
-
-# Encoder params
-model/hyperparameter.class_to_tune = @TemporalFusionTransformer
-model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
-model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.dropout = (0.0, 0.4)
-model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =heads_divisible_by_hidden(model/hyperparameter.hidden)
-model/hyperparameter.example_length=25
-
-
-
-
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 4b761726..99251ea2 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -6,6 +6,7 @@ import icu_benchmarks.models.wrappers
 import icu_benchmarks.models.dl_models
 import icu_benchmarks.models.utils
 import icu_benchmarks.models.layers
+import icu_benchmarks.data.loader
 
 # Do not generate features from dynamic data
 base_classification_preprocessor.generate_features = False
diff --git a/experiments/demo_benchmark_regression.yml b/experiments/demo_benchmark_regression.yml
index 3b678371..047f0c88 100644
--- a/experiments/demo_benchmark_regression.yml
+++ b/experiments/demo_benchmark_regression.yml
@@ -19,16 +19,16 @@ parameters:
     values:
       - demo_data/los/eicu_demo
       - demo_data/los/mimic_demo
-      - demo_data/kf/eicu_demo
-      - demo_data/kf/mimic_demo
+   #   - demo_data/kf/eicu_demo
+   #   - demo_data/kf/mimic_demo
   model:
     values:
-      - ElasticNet
-      - LGBMRegressor
-      - GRU
-      - LSTM
-      - TCN
-      - Transformer
+  #    - ElasticNet
+  #    - LGBMRegressor
+  #    - GRU
+  #    - LSTM
+  #    - TCN
+      - TemporalFusionTransformer
   seed:
     values:
       - 1111
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index c576d857..2f13814b 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -1,5 +1,6 @@
 from typing import List
 from pandas import DataFrame
+import pandas as pd
 import gin
 import numpy as np
 from torch import Tensor, cat, from_numpy, float32,empty,stack
@@ -11,7 +12,7 @@
 from .constants import DataSplit as Split
 from .constants import FeatType as Features
 from collections import OrderedDict
-
+from pytorch_forecasting import TimeSeriesDataSet
 class CommonDataset(Dataset):
     """Common dataset: subclass of Torch Dataset that represents the data to learn on.
 
@@ -92,7 +93,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             return self._cached_dataset[idx]
 
         pad_value = 0.0
-        stay_id = self.outcome_df.index.unique()[idx]  # [self.vars["GROUP"]]
+        stay_id = self.outcome_df.index.unique()[idx]  
 
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id].to_numpy()
@@ -185,7 +186,6 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         
         # We need to be sure that tensors are returned in the correct order to be processed correclty by tft
         tensors = [[] for _ in range(8)]
-        
         for var in self.features_df.columns:
             if  var == 'sex' :
                 tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
@@ -198,8 +198,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         tensors[6].extend(self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float))
         tensors[7].append(np.asarray([stay_id]))
         window_shape0=np.shape(tensors[0])[1]
-       # window = self.features_df.loc[stay_id:stay_id].to_numpy()
-       # labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
+
         if len(tensors[6]) == 1:
             # only one label per stay, align with window
             tensors[6] = np.concatenate([np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0)
@@ -213,12 +212,10 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             # window shorter than the longest window in dataset, pad to same length
             tensors[0]= np.concatenate([tensors[0], np.ones(( np.shape(tensors[0])[0],self.maxlen - np.shape(tensors[0])[1])) * pad_value], axis=1)
             tensors[1]= np.concatenate([tensors[1], np.ones(( np.shape(tensors[1])[0],self.maxlen - np.shape(tensors[1])[1])) * pad_value], axis=1)
-          #  tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, np.shape(tensors[2])[0])) * pad_value], axis=0)
-          #  tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, np.shape(tensors[3])[0])) * pad_value], axis=0)
-          #  tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, np.shape(tensors[4])[0])) * pad_value], axis=0)
+
             tensors[5]= np.concatenate([tensors[5], np.ones(( np.shape(tensors[5])[0],self.maxlen - np.shape(tensors[5])[1])) * pad_value], axis=1)
             
-           # window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
+         
             
             tensors[6] = np.concatenate([tensors[6], np.ones(self.maxlen - np.shape(tensors[6])[0]) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
@@ -229,7 +226,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             pad_mask[not_labeled] = 0
         tensors[6]=[tensors[6]]
         pad_mask = pad_mask.astype(bool)
-       # data = window.astype(np.float32)
+       
         
         
         tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)
@@ -317,7 +314,7 @@ class ImputationPredictionDataset(Dataset):
         select_columns (List[str], optional): the columns to serve as input for the imputation model. Defaults to None.
         ram_cache (bool, optional): wether the dataset should be stored in ram. Defaults to True.
     """
-
+    
     def __init__(
         self,
         data: DataFrame,
@@ -372,3 +369,75 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         return from_numpy(window.values).to(float32)
 
+@gin.configurable("PredictionDatasetTFTpytorch")
+class PredictionDatasetTFTpytorch(TimeSeriesDataSet):
+    def __init__(self,
+        data: dict,
+        split: str ,
+        max_prediction_length: int,
+        max_encoder_length: int,
+        *args,
+        ram_cache: bool = True,
+        **kwargs):
+        data[split]["FEATURES"]["time_idx"]=((data[split]["FEATURES"]["time"]/ pd.Timedelta(seconds=3600))).astype(int)
+        data=data.get(split)
+        labels = data["OUTCOME"]
+        features=data["FEATURES"]
+        self.data=pd.merge(labels, features, on=['stay_id', 'time'])
+        self.split = split
+        self.args=args
+        self.ram_cache=ram_cache
+        self.kwargs=kwargs
+        
+        super().__init__(data=self.data,
+            time_idx="time_idx",
+            target="label",
+            group_ids=["stay_id"],
+            min_encoder_length=max_encoder_length // 2,  
+            max_encoder_length=max_encoder_length,
+            min_prediction_length=1,
+            max_prediction_length=max_prediction_length,
+            static_categoricals=[],
+            static_reals=["height", "weight","age","sex"],
+            time_varying_known_categoricals=[],
+            time_varying_known_reals=[],
+            time_varying_unknown_categoricals=[],
+            time_varying_unknown_reals=[
+                "alb", "alp", "alt", "ast", "be", "bicar", "bili", "bili_dir", "bnd", "bun", "ca", "cai", "ck", "ckmb", "cl",
+        "crea", "crp", "dbp", "fgn", "fio2", "glu", "hgb", "hr", "inr_pt", "k", "lact", "lymph", "map", "mch", "mchc", "mcv",
+        "methb", "mg", "na", "neut", "o2sat", "pco2", "ph", "phos", "plt", "po2", "ptt", "resp", "sbp", "temp", "tnt", "urine",
+        "wbc",'MissingIndicator_1', 'MissingIndicator_2',
+       'MissingIndicator_3', 'MissingIndicator_4', 'MissingIndicator_5',
+       'MissingIndicator_6', 'MissingIndicator_7', 'MissingIndicator_8',
+       'MissingIndicator_9', 'MissingIndicator_10', 'MissingIndicator_11',
+       'MissingIndicator_12', 'MissingIndicator_13', 'MissingIndicator_14',
+       'MissingIndicator_15', 'MissingIndicator_16', 'MissingIndicator_17',
+       'MissingIndicator_18', 'MissingIndicator_19', 'MissingIndicator_20',
+       'MissingIndicator_21', 'MissingIndicator_22', 'MissingIndicator_23',
+       'MissingIndicator_24', 'MissingIndicator_25', 'MissingIndicator_26',
+       'MissingIndicator_27', 'MissingIndicator_28', 'MissingIndicator_29',
+       'MissingIndicator_30', 'MissingIndicator_31', 'MissingIndicator_32',
+       'MissingIndicator_33', 'MissingIndicator_34', 'MissingIndicator_35',
+       'MissingIndicator_36', 'MissingIndicator_37', 'MissingIndicator_38',
+       'MissingIndicator_39', 'MissingIndicator_40', 'MissingIndicator_41',
+       'MissingIndicator_42', 'MissingIndicator_43', 'MissingIndicator_44',
+       'MissingIndicator_45', 'MissingIndicator_46', 'MissingIndicator_47',
+       'MissingIndicator_48'
+            ],
+            add_relative_time_idx=True,
+            add_target_scales=True,
+            add_encoder_length=True,)
+    def get_balance(self) -> list:
+        """Return the weight balance for the split of interest.
+
+        Returns:
+            Weights for each label.
+        """
+        
+        counts = self.data["target"][0].unique(return_counts=True)
+        
+
+        return list((1/counts[1]) * counts[1].sum() / counts[0].shape[0])
+        
+        
+            
\ No newline at end of file
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index a34ad0a4..e5df2793 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -7,7 +7,9 @@
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor,cat,jit
+from pytorch_forecasting import TemporalFusionTransformer
 
+from pytorch_forecasting.metrics import  QuantileLoss
 
 @gin.configurable
 class RNNet(DLPredictionWrapper):
@@ -282,9 +284,9 @@ def forward(self, x):
         o = o.permute(0, 2, 1)  # Permute to channel last
         pred = self.logit(o)
         return pred
-    
+
 @gin.configurable
-class TemporalFusionTransformer(DLPredictionWrapper):
+class TFT(DLPredictionWrapper):
     """ 
     Implementation of https://arxiv.org/abs/1912.09363 
     """
@@ -352,4 +354,25 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         
         o=self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
         pred = self.logit(o)
-        return pred
\ No newline at end of file
+        return pred
+    
+@gin.configurable
+class TFTpytorch(DLPredictionWrapper):
+    """ 
+    Implementation of https://arxiv.org/abs/1912.09363 
+    """
+    supported_run_modes = [RunMode.classification, RunMode.regression]
+    def __init__(self,dataset,hidden,dropout,
+                 n_heads,dropout_att,lr,optimizer,*args,**kwargs):
+        self.device=device
+        TemporalFusionTransformer.from_dataset(
+        dataset,
+        learning_rate=lr,
+        hidden_size=hidden,
+        attention_head_size=n_heads,
+        dropout=dropout,
+        hidden_continuous_size=hidden,
+        loss=QuantileLoss(),
+        optimizer=optimizer,
+        
+)
\ No newline at end of file
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index f358483f..21334a67 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -8,8 +8,9 @@
 from pytorch_lightning.loggers import TensorBoardLogger
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint, TQDMProgressBar
+from pytorch_forecasting import TimeSeriesDataSet
 from pathlib import Path
-from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset,PredictionDatasetTFT
+from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset,PredictionDatasetTFT,PredictionDatasetTFTpytorch
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
@@ -72,12 +73,13 @@ def train_common(
     """
 
     logging.info(f"Training model: {model.__name__}.")
-    dataset_class = ImputationDataset if mode == RunMode.imputation else (PredictionDatasetTFT if model.__name__ =='TemporalFusionTransformer' else PredictionDataset )
-    
+    dataset_class = (ImputationDataset if mode == RunMode.imputation
+                 else (PredictionDatasetTFT if model.__name__ == 'TFT'
+                       else (PredictionDatasetTFTpytorch if model.__name__ == 'TFTpytorch' else PredictionDataset)))
 
     logging.info(f"Logging to directory: {log_dir}.")
     save_config_file(log_dir)  # We save the operative config before and also after training
-
+    
     train_dataset = dataset_class(data, split=Split.train, ram_cache=ram_cache)
     val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache)
     train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
@@ -85,31 +87,36 @@ def train_common(
 
     logging.debug(f"Training on {len(train_dataset)} samples and validating on {len(val_dataset)} samples.")
     logging.info(f"Using {num_workers} workers for data loading.")
-
-    train_loader = DataLoader(
-        train_dataset,
-        batch_size=batch_size,
-        shuffle=True,
-        num_workers=num_workers,
-        pin_memory=True,
-        drop_last=True,
-    )
-    val_loader = DataLoader(
-        val_dataset,
-        batch_size=batch_size,
-        shuffle=False,
-        num_workers=num_workers,
-        pin_memory=True,
-        drop_last=True,
-    )
-    
-    if(isinstance(next(iter(train_loader))[0] ,OrderedDict)):
-         model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
-    
+    if(model.__name__ =='TFTpytorch'):
+        
+        train_loader = train_dataset.to_dataloader(train=True, batch_size=batch_size, num_workers=num_workers,pin_memory=True,drop_last=True)
+        val_loader = val_dataset.to_dataloader(train=False, batch_size=batch_size ,num_workers=num_workers,pin_memory=True,drop_last=True)
+        model=model(dataset=train_dataset,optimizer=optimizer, epochs=epochs, run_mode=mode)
     else:
-        data_shape = next(iter(train_loader))[0].shape
+        train_loader = DataLoader(
+            train_dataset,
+            batch_size=batch_size,
+            shuffle=True,
+            num_workers=num_workers,
+            pin_memory=True,
+            drop_last=True,
+        )
+        val_loader = DataLoader(
+            val_dataset,
+            batch_size=batch_size,
+            shuffle=False,
+            num_workers=num_workers,
+            pin_memory=True,
+            drop_last=True,
+        )
     
-        model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
+        if(isinstance(next(iter(train_loader))[0] ,OrderedDict)):
+            model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
+        
+        else:
+            data_shape = next(iter(train_loader))[0].shape
+        
+            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
     model.set_weight(weight, train_dataset)
     if load_weights:
         if source_dir.exists():

From e72927af4a60bca9f6588f958ce7f8a87c879222 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 11 Jul 2023 14:57:54 +0200
Subject: [PATCH 033/142] new gin files

---
 configs/prediction_models/TFT.gin        | 25 ++++++++++++++++++++++++
 configs/prediction_models/TFTpytorch.gin | 22 +++++++++++++++++++++
 2 files changed, 47 insertions(+)
 create mode 100644 configs/prediction_models/TFT.gin
 create mode 100644 configs/prediction_models/TFTpytorch.gin

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
new file mode 100644
index 00000000..d9bbe607
--- /dev/null
+++ b/configs/prediction_models/TFT.gin
@@ -0,0 +1,25 @@
+# Settings for TFT model.
+
+# Common settings for DL models
+include "configs/prediction_models/common/DLCommon.gin"
+
+# Optimizer params
+train_common.model = @TFT
+
+optimizer/hyperparameter.class_to_tune = @Adam
+optimizer/hyperparameter.weight_decay = 1e-6
+optimizer/hyperparameter.lr = (1e-5, 3e-4)
+
+# Encoder params
+model/hyperparameter.class_to_tune = @TFT
+model/hyperparameter.encoder_length = 24
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.num_classes = %NUM_CLASSES
+model/hyperparameter.dropout = (0.0, 0.4)
+model/hyperparameter.dropout_att = (0.0, 0.4)
+model/hyperparameter.n_heads =1
+model/hyperparameter.example_length=25
+
+
+
+
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
new file mode 100644
index 00000000..77649f97
--- /dev/null
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -0,0 +1,22 @@
+# Settings for TFT model.
+
+# Common settings for DL models
+include "configs/prediction_models/common/DLCommon.gin"
+
+# Optimizer params
+train_common.model = @TFTpytorch
+
+optimizer/hyperparameter.class_to_tune = @Adam
+optimizer/hyperparameter.weight_decay = 1e-6
+#optimizer/hyperparameter.lr = (1e-5, 3e-4)
+
+# Encoder params
+model/hyperparameter.class_to_tune = @TFTpytorch
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.num_classes = %NUM_CLASSES
+model/hyperparameter.dropout = (0.0, 0.4)
+model/hyperparameter.dropout_att = (0.0, 0.4)
+model/hyperparameter.n_heads =1
+model/hyperparameter.lr = (1e-5, 3e-4)
+PredictionDatasetTFTpytorch.max_encoder_length = 24
+PredictionDatasetTFTpytorch.max_prediction_length = 48

From 760a25dc54ca9ad982b9629d68cbad9517045405 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 11 Jul 2023 15:19:51 +0200
Subject: [PATCH 034/142] created own set_weights for tft pytorch

---
 icu_benchmarks/models/dl_models.py | 17 +++++++++++++----
 icu_benchmarks/models/wrappers.py  |  1 -
 2 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index e5df2793..ebc78ec4 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -6,7 +6,7 @@
 from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor,cat,jit
+from torch import Tensor,cat,jit,FloatTensor
 from pytorch_forecasting import TemporalFusionTransformer
 
 from pytorch_forecasting.metrics import  QuantileLoss
@@ -363,8 +363,9 @@ class TFTpytorch(DLPredictionWrapper):
     """
     supported_run_modes = [RunMode.classification, RunMode.regression]
     def __init__(self,dataset,hidden,dropout,
-                 n_heads,dropout_att,lr,optimizer,*args,**kwargs):
-        self.device=device
+                 n_heads,dropout_att,lr,optimizer,*args,device='cpu',**kwargs):
+        
+        
         TemporalFusionTransformer.from_dataset(
         dataset,
         learning_rate=lr,
@@ -375,4 +376,12 @@ def __init__(self,dataset,hidden,dropout,
         loss=QuantileLoss(),
         optimizer=optimizer,
         
-)
\ No newline at end of file
+)
+    
+    def set_weight(self, weight, dataset):
+        """Set the weight for the loss function."""
+        if isinstance(weight, list):
+            weight = FloatTensor(weight)
+        elif weight == "balanced":
+            weight = FloatTensor(dataset.get_balance())
+        self.loss_weights = weight
\ No newline at end of file
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 5872acb3..bf9452b1 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -187,7 +187,6 @@ class DLPredictionWrapper(DLWrapper):
 
     def set_weight(self, weight, dataset):
         """Set the weight for the loss function."""
-
         if isinstance(weight, list):
             weight = torch.FloatTensor(weight).to(self.device)
         elif weight == "balanced":

From a356df61115357a176938d35d52d0e439c23800f Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 11 Jul 2023 15:24:23 +0200
Subject: [PATCH 035/142] overrode get_features_names for tftpytorch dataloader

---
 icu_benchmarks/data/loader.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 2f13814b..16be3676 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -388,7 +388,7 @@ def __init__(self,
         self.args=args
         self.ram_cache=ram_cache
         self.kwargs=kwargs
-        
+        self.column_names=features.columns
         super().__init__(data=self.data,
             time_idx="time_idx",
             target="label",
@@ -438,6 +438,9 @@ def get_balance(self) -> list:
         
 
         return list((1/counts[1]) * counts[1].sum() / counts[0].shape[0])
+    def get_feature_names(self):
+        
+        return self.column_names
         
         
             
\ No newline at end of file

From 2e125928dce146855bb6b2ef5eb88ebca9fbcf05 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 18 Jul 2023 13:24:11 +0200
Subject: [PATCH 036/142] fixed datatypes for old tft and working on new tft

---
 configs/prediction_models/TFT.gin        |  2 +-
 configs/prediction_models/TFTpytorch.gin |  6 +-
 icu_benchmarks/cross_validation.py       |  1 +
 icu_benchmarks/data/loader.py            |  6 +-
 icu_benchmarks/models/dl_models.py       | 93 +++++++++++++++++++-----
 icu_benchmarks/models/layers.py          |  3 +-
 icu_benchmarks/models/train.py           | 45 +++++++-----
 icu_benchmarks/models/wrappers.py        | 21 ++++--
 8 files changed, 127 insertions(+), 50 deletions(-)

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
index d9bbe607..d03b8c21 100644
--- a/configs/prediction_models/TFT.gin
+++ b/configs/prediction_models/TFT.gin
@@ -13,7 +13,7 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 # Encoder params
 model/hyperparameter.class_to_tune = @TFT
 model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.hidden = 12
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 77649f97..4a618903 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -18,5 +18,7 @@ model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
 model/hyperparameter.n_heads =1
 model/hyperparameter.lr = (1e-5, 3e-4)
-PredictionDatasetTFTpytorch.max_encoder_length = 24
-PredictionDatasetTFTpytorch.max_prediction_length = 48
+PredictionDatasetTFTpytorch.max_encoder_length = 6
+PredictionDatasetTFTpytorch.max_prediction_length = 12
+
+
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 89a98864..43e40bf1 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -101,6 +101,7 @@ def execute_repeated_cv(
                 verbose=verbose,
                 use_wandb=wandb,
             )
+            
             train_time = datetime.now() - start_time
 
             log_full_line(
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 16be3676..82b34c08 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -191,6 +191,8 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
                 tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
             elif var == 'age' or var== 'height' or var== 'weight':
                 tensors[1].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+            elif "MissingIndicator" in var:
+                tensors[4].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
             else :
                 tensors[5].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
     
@@ -212,7 +214,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             # window shorter than the longest window in dataset, pad to same length
             tensors[0]= np.concatenate([tensors[0], np.ones(( np.shape(tensors[0])[0],self.maxlen - np.shape(tensors[0])[1])) * pad_value], axis=1)
             tensors[1]= np.concatenate([tensors[1], np.ones(( np.shape(tensors[1])[0],self.maxlen - np.shape(tensors[1])[1])) * pad_value], axis=1)
-
+            tensors[4]= np.concatenate([tensors[4], np.ones(( np.shape(tensors[4])[0],self.maxlen - np.shape(tensors[4])[1])) * pad_value], axis=1)
             tensors[5]= np.concatenate([tensors[5], np.ones(( np.shape(tensors[5])[0],self.maxlen - np.shape(tensors[5])[1])) * pad_value], axis=1)
             
          
@@ -395,7 +397,7 @@ def __init__(self,
             group_ids=["stay_id"],
             min_encoder_length=max_encoder_length // 2,  
             max_encoder_length=max_encoder_length,
-            min_prediction_length=1,
+            min_prediction_length=2,
             max_prediction_length=max_prediction_length,
             static_categoricals=[],
             static_reals=["height", "weight","age","sex"],
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index ebc78ec4..500b6015 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -7,7 +7,7 @@
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor,cat,jit,FloatTensor
-from pytorch_forecasting import TemporalFusionTransformer
+from pytorch_forecasting import TemporalFusionTransformer,TimeSeriesDataSet
 
 from pytorch_forecasting.metrics import  QuantileLoss
 
@@ -295,8 +295,8 @@ class TFT(DLPredictionWrapper):
     _supported_run_modes = [RunMode.classification, RunMode.regression]
     def __init__(self,num_classes, encoder_length,hidden,dropout,
                  n_heads,dropout_att,example_length,*args,quantiles=[0.1, 0.5, 0.9],static_categorical_inp_size=[2],temporal_known_categorical_inp_size=[],
-    temporal_observed_categorical_inp_size=[],static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
-    temporal_observed_continuous_inp_size=96,temporal_target_size=1,**kwargs):
+    temporal_observed_categorical_inp_size=[48],static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
+    temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs):
         #derived variables
         num_static_vars=len(static_categorical_inp_size)+static_continuous_inp_size
         num_future_vars=len(temporal_known_categorical_inp_size)+temporal_known_continuous_inp_size
@@ -309,7 +309,7 @@ def __init__(self,num_classes, encoder_length,hidden,dropout,
         super().__init__(num_classes=num_classes, encoder_length=encoder_length,hidden=hidden,
                  n_heads=n_heads,dropout_att=dropout_att,example_length=example_length,quantiles=quantiles,
                  num_static_vars=num_static_vars,num_future_vars=num_future_vars,num_historic_vars=num_historic_vars,*args,static_categorical_inp_size=1,temporal_known_categorical_inp_size=0,
-                temporal_observed_categorical_inp_size=0,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
+                temporal_observed_categorical_inp_size=48,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
                 temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs)
 
         
@@ -355,33 +355,86 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         o=self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
         pred = self.logit(o)
         return pred
-    
+
 @gin.configurable
-class TFTpytorch(DLPredictionWrapper):
-    """ 
+class TFTpytorch(TemporalFusionTransformer,DLPredictionWrapper):
+    """
     Implementation of https://arxiv.org/abs/1912.09363 
     """
     supported_run_modes = [RunMode.classification, RunMode.regression]
     def __init__(self,dataset,hidden,dropout,
-                 n_heads,dropout_att,lr,optimizer,*args,device='cpu',**kwargs):
+                 n_heads,dropout_att,lr,optimizer,num_classes,*args,**kwargs):
         
+        DLPredictionWrapper.__init__(self,lr=lr,optimizer=optimizer,*args,**kwargs)        
         
-        TemporalFusionTransformer.from_dataset(
-        dataset,
-        learning_rate=lr,
-        hidden_size=hidden,
-        attention_head_size=n_heads,
-        dropout=dropout,
-        hidden_continuous_size=hidden,
-        loss=QuantileLoss(),
-        optimizer=optimizer,
+        TemporalFusionTransformer.__init__(self)
+
+        self.model=TemporalFusionTransformer.from_dataset(dataset=dataset,hidden_size=hidden,dropout=dropout,
+                 attention_head_size=n_heads,learning_rate=lr,optimizer=optimizer,loss=QuantileLoss(),hidden_continuous_size=hidden)
+        self.hparams.update(self.model.hparams)
+
+        #for key in model.hparams.keys():
+        # self.hparams[key]=model.hparams[key]
+       # del self.hparams["input_shape"]
+       # del self.hparams["input_size"]
+        #print(self.hparams)
+       # print(self.m)
+        self.logit = nn.Linear(7, num_classes)
+    def set_weight(self, weight, dataset):
+        """
+        Set the weight for the loss function
+        """
+        if isinstance(weight, list):
+            weight = FloatTensor(weight)
+        elif weight == "balanced":
+            weight = FloatTensor(dataset.get_balance())
+        self.loss_weights = weight
+    
+    def test_step(self, batch, batch_idx):
+        x, y = batch
+        log, out = self.step(x, y, batch_idx)
+        print("finally here")
+        log.update(self.create_log(x, y, out, batch_idx))
+        self.testing_step_outputs.append(log)
+        self.log(f"test/loss", log, on_step=False, on_epoch=True, sync_dist=True)
+
+        return log
+
+
+"""
+@gin.configurable
+class TFTpytorch(TemporalFusionTransformer,DLPredictionWrapper):
+    
+    Implementation of https://arxiv.org/abs/1912.09363 
+    
+    supported_run_modes = [RunMode.classification, RunMode.regression]
+    def __init__(self,dataset,hidden,dropout,
+                 n_heads,dropout_att,lr,optimizer,num_classes,*args,**kwargs):
+        
+        DLPredictionWrapper.__init__(self,lr=lr,optimizer=optimizer,*args,**kwargs)        
+        self.model=TemporalFusionTransformer.from_dataset(dataset)
         
-)
     
+        
     def set_weight(self, weight, dataset):
-        """Set the weight for the loss function."""
+        
+        Set the weight for the loss function
+        
         if isinstance(weight, list):
             weight = FloatTensor(weight)
         elif weight == "balanced":
             weight = FloatTensor(dataset.get_balance())
-        self.loss_weights = weight
\ No newline at end of file
+        self.loss_weights = weight
+    
+    def test_step(self, batch, batch_idx):
+        x, y = batch
+        log, out = self.model.step(x, y, batch_idx)
+        log.update(self.model.create_log(x, y, out, batch_idx))
+        self.testing_step_outputs.append(log)
+        self.model.log(f"test/loss", log, on_step=False, on_epoch=True, sync_dist=True)
+
+        return log
+    def forward(self, x: Dict[str,Tensor]) -> Dict[str, Tensor]:
+        return self.model.forward(x)
+    
+"""
\ No newline at end of file
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 016cb366..79ee0e61 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -540,6 +540,7 @@ def forward(self, x: Dict[str, Tensor]):
                                             self.t_cat_k_embed,
                                             self.t_cont_k_embedding_vectors,
                                             self.t_cont_k_embedding_bias)
+        
         t_observed_inp = self._apply_embedding(t_cat_o_inp,
                                                t_cont_o_inp,
                                                self.t_cat_o_embed,
@@ -756,7 +757,7 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         enriched = self.enrichment_grn(temporal_features, c=ce)
 
         # Temporal self attention
-        x, _ = self.attention(enriched)
+        x, attn_prob= self.attention(enriched)
 
         # Don't compute hictorical quantiles
         x = x[:, self.encoder_length:, :]
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 21334a67..fdd753a9 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -8,7 +8,7 @@
 from pytorch_lightning.loggers import TensorBoardLogger
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint, TQDMProgressBar
-from pytorch_forecasting import TimeSeriesDataSet
+from pytorch_forecasting import TimeSeriesDataSet,TemporalFusionTransformer
 from pathlib import Path
 from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset,PredictionDatasetTFT,PredictionDatasetTFTpytorch
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
@@ -43,7 +43,7 @@ def train_common(
     test_on: str = Split.test,
     use_wandb: bool = False,
     cpu: bool = False,
-    verbose=False,
+    verbose=True,
     ram_cache=False,
     num_workers: int = min(cpu_core_count, torch.cuda.device_count() * 4 * int(torch.cuda.is_available()), 32),
 ):
@@ -84,14 +84,20 @@ def train_common(
     val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache)
     train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
-
+    test_dataset = dataset_class(data, split=test_on)
     logging.debug(f"Training on {len(train_dataset)} samples and validating on {len(val_dataset)} samples.")
     logging.info(f"Using {num_workers} workers for data loading.")
     if(model.__name__ =='TFTpytorch'):
         
         train_loader = train_dataset.to_dataloader(train=True, batch_size=batch_size, num_workers=num_workers,pin_memory=True,drop_last=True)
         val_loader = val_dataset.to_dataloader(train=False, batch_size=batch_size ,num_workers=num_workers,pin_memory=True,drop_last=True)
-        model=model(dataset=train_dataset,optimizer=optimizer, epochs=epochs, run_mode=mode)
+        test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=num_workers,pin_memory=True,drop_last=True,shuffle=False)
+        model=model(train_dataset,optimizer=optimizer, epochs=epochs, run_mode=mode)
+        model=model.model
+        
+        
+        
+        
     else:
         train_loader = DataLoader(
             train_dataset,
@@ -143,7 +149,7 @@ def train_common(
         max_epochs=epochs if model.needs_training else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="auto" if not cpu else "cpu",
+        accelerator= "cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic=reproducible,
         benchmark=not reproducible,
@@ -151,6 +157,8 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=0,
     )
+  #  print(trainer.test(model, dataloaders=test_loader, verbose=verbose))
+   # test_loss = trainer.test(model.model, dataloaders=test_loader, verbose=verbose)[0]["RMSE"]
 
     if model.needs_fit:
         logging.info("Fitting model to data.")
@@ -160,25 +168,28 @@ def train_common(
 
     if model.needs_training:
         logging.info("Training model.")
+        
+        
         trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
         logging.info("Training complete.")
-
     test_dataset = dataset_class(data, split=test_on)
     test_dataset = assure_minimum_length(test_dataset)
-    test_loader = (
-        DataLoader(
-            test_dataset,
-            batch_size=min(batch_size * 4, len(test_dataset)),
-            shuffle=False,
-            num_workers=num_workers,
-            pin_memory=True,
-            drop_last=True,
-        )
+    if(type(model).__name__ =='TFTpytorch'):
+        test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=num_workers,pin_memory=True,drop_last=True,shuffle=False)
+    else:
+       
+        test_loader = (
+            DataLoader(
+                test_dataset,
+                batch_size=min(batch_size * 4, len(test_dataset)),
+                shuffle=False,
+                num_workers=num_workers,
+                pin_memory=True,
+                drop_last=True,
+            )
         if model.needs_training
         else DataLoader([test_dataset.to_tensor()], batch_size=1)
     )
-
-    model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
     return test_loss
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index bf9452b1..eb2c5db2 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -243,15 +243,19 @@ def step_fn(self, element, step_prefix=""):
                 if(key=='target'):
                     labels=value.squeeze()
                 data[key] = value
-                
+        
+            
         elif len(element) == 2:
-            data, labels = element[0], element[1].to(self.device)
-            if isinstance(data, list):
-                for i in range(len(data)):
-                    data[i] = data[i].float().to(self.device)
+            if(isinstance(element[1] ,tuple)):
+                data, labels = element[0], element[1]
             else:
-                data = data.float().to(self.device)
-            mask = torch.ones_like(labels).bool()
+                data, labels = element[0], element[1].to(self.device)
+                if isinstance(data, list):
+                    for i in range(len(data)):
+                        data[i] = data[i].float().to(self.device)
+                else:
+                    data = data.float().to(self.device)
+                mask = torch.ones_like(labels).bool()
 
         elif len(element) == 3:
             data, labels, mask = element[0], element[1].to(self.device), element[2].to(self.device)
@@ -262,7 +266,10 @@ def step_fn(self, element, step_prefix=""):
                 data = data.float().to(self.device)
         else:
             raise Exception("Loader should return either (data, label) or (data, label, mask)")
+        
         out = self(data)
+        
+        
         if len(out) == 2 and isinstance(out, tuple):
             out, aux_loss = out
         else:

From c0dd05ff70632764b74cd5d92c0cf8acc0af91e0 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@pc-79-109.customer.ask4.lan>
Date: Tue, 18 Jul 2023 13:24:11 +0200
Subject: [PATCH 037/142] fixed datatypes for old tft and working on new tft

---
 configs/prediction_models/TFT.gin        |  2 +-
 configs/prediction_models/TFTpytorch.gin |  6 +-
 icu_benchmarks/cross_validation.py       |  1 +
 icu_benchmarks/data/loader.py            |  6 +-
 icu_benchmarks/models/dl_models.py       | 93 +++++++++++++++++++-----
 icu_benchmarks/models/layers.py          |  3 +-
 icu_benchmarks/models/train.py           | 45 +++++++-----
 icu_benchmarks/models/wrappers.py        | 21 ++++--
 8 files changed, 127 insertions(+), 50 deletions(-)

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
index d9bbe607..d03b8c21 100644
--- a/configs/prediction_models/TFT.gin
+++ b/configs/prediction_models/TFT.gin
@@ -13,7 +13,7 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 # Encoder params
 model/hyperparameter.class_to_tune = @TFT
 model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.hidden = 12
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 77649f97..4a618903 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -18,5 +18,7 @@ model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
 model/hyperparameter.n_heads =1
 model/hyperparameter.lr = (1e-5, 3e-4)
-PredictionDatasetTFTpytorch.max_encoder_length = 24
-PredictionDatasetTFTpytorch.max_prediction_length = 48
+PredictionDatasetTFTpytorch.max_encoder_length = 6
+PredictionDatasetTFTpytorch.max_prediction_length = 12
+
+
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 89a98864..43e40bf1 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -101,6 +101,7 @@ def execute_repeated_cv(
                 verbose=verbose,
                 use_wandb=wandb,
             )
+            
             train_time = datetime.now() - start_time
 
             log_full_line(
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 16be3676..82b34c08 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -191,6 +191,8 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
                 tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
             elif var == 'age' or var== 'height' or var== 'weight':
                 tensors[1].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+            elif "MissingIndicator" in var:
+                tensors[4].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
             else :
                 tensors[5].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
     
@@ -212,7 +214,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             # window shorter than the longest window in dataset, pad to same length
             tensors[0]= np.concatenate([tensors[0], np.ones(( np.shape(tensors[0])[0],self.maxlen - np.shape(tensors[0])[1])) * pad_value], axis=1)
             tensors[1]= np.concatenate([tensors[1], np.ones(( np.shape(tensors[1])[0],self.maxlen - np.shape(tensors[1])[1])) * pad_value], axis=1)
-
+            tensors[4]= np.concatenate([tensors[4], np.ones(( np.shape(tensors[4])[0],self.maxlen - np.shape(tensors[4])[1])) * pad_value], axis=1)
             tensors[5]= np.concatenate([tensors[5], np.ones(( np.shape(tensors[5])[0],self.maxlen - np.shape(tensors[5])[1])) * pad_value], axis=1)
             
          
@@ -395,7 +397,7 @@ def __init__(self,
             group_ids=["stay_id"],
             min_encoder_length=max_encoder_length // 2,  
             max_encoder_length=max_encoder_length,
-            min_prediction_length=1,
+            min_prediction_length=2,
             max_prediction_length=max_prediction_length,
             static_categoricals=[],
             static_reals=["height", "weight","age","sex"],
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index ebc78ec4..500b6015 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -7,7 +7,7 @@
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor,cat,jit,FloatTensor
-from pytorch_forecasting import TemporalFusionTransformer
+from pytorch_forecasting import TemporalFusionTransformer,TimeSeriesDataSet
 
 from pytorch_forecasting.metrics import  QuantileLoss
 
@@ -295,8 +295,8 @@ class TFT(DLPredictionWrapper):
     _supported_run_modes = [RunMode.classification, RunMode.regression]
     def __init__(self,num_classes, encoder_length,hidden,dropout,
                  n_heads,dropout_att,example_length,*args,quantiles=[0.1, 0.5, 0.9],static_categorical_inp_size=[2],temporal_known_categorical_inp_size=[],
-    temporal_observed_categorical_inp_size=[],static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
-    temporal_observed_continuous_inp_size=96,temporal_target_size=1,**kwargs):
+    temporal_observed_categorical_inp_size=[48],static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
+    temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs):
         #derived variables
         num_static_vars=len(static_categorical_inp_size)+static_continuous_inp_size
         num_future_vars=len(temporal_known_categorical_inp_size)+temporal_known_continuous_inp_size
@@ -309,7 +309,7 @@ def __init__(self,num_classes, encoder_length,hidden,dropout,
         super().__init__(num_classes=num_classes, encoder_length=encoder_length,hidden=hidden,
                  n_heads=n_heads,dropout_att=dropout_att,example_length=example_length,quantiles=quantiles,
                  num_static_vars=num_static_vars,num_future_vars=num_future_vars,num_historic_vars=num_historic_vars,*args,static_categorical_inp_size=1,temporal_known_categorical_inp_size=0,
-                temporal_observed_categorical_inp_size=0,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
+                temporal_observed_categorical_inp_size=48,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
                 temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs)
 
         
@@ -355,33 +355,86 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         o=self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
         pred = self.logit(o)
         return pred
-    
+
 @gin.configurable
-class TFTpytorch(DLPredictionWrapper):
-    """ 
+class TFTpytorch(TemporalFusionTransformer,DLPredictionWrapper):
+    """
     Implementation of https://arxiv.org/abs/1912.09363 
     """
     supported_run_modes = [RunMode.classification, RunMode.regression]
     def __init__(self,dataset,hidden,dropout,
-                 n_heads,dropout_att,lr,optimizer,*args,device='cpu',**kwargs):
+                 n_heads,dropout_att,lr,optimizer,num_classes,*args,**kwargs):
         
+        DLPredictionWrapper.__init__(self,lr=lr,optimizer=optimizer,*args,**kwargs)        
         
-        TemporalFusionTransformer.from_dataset(
-        dataset,
-        learning_rate=lr,
-        hidden_size=hidden,
-        attention_head_size=n_heads,
-        dropout=dropout,
-        hidden_continuous_size=hidden,
-        loss=QuantileLoss(),
-        optimizer=optimizer,
+        TemporalFusionTransformer.__init__(self)
+
+        self.model=TemporalFusionTransformer.from_dataset(dataset=dataset,hidden_size=hidden,dropout=dropout,
+                 attention_head_size=n_heads,learning_rate=lr,optimizer=optimizer,loss=QuantileLoss(),hidden_continuous_size=hidden)
+        self.hparams.update(self.model.hparams)
+
+        #for key in model.hparams.keys():
+        # self.hparams[key]=model.hparams[key]
+       # del self.hparams["input_shape"]
+       # del self.hparams["input_size"]
+        #print(self.hparams)
+       # print(self.m)
+        self.logit = nn.Linear(7, num_classes)
+    def set_weight(self, weight, dataset):
+        """
+        Set the weight for the loss function
+        """
+        if isinstance(weight, list):
+            weight = FloatTensor(weight)
+        elif weight == "balanced":
+            weight = FloatTensor(dataset.get_balance())
+        self.loss_weights = weight
+    
+    def test_step(self, batch, batch_idx):
+        x, y = batch
+        log, out = self.step(x, y, batch_idx)
+        print("finally here")
+        log.update(self.create_log(x, y, out, batch_idx))
+        self.testing_step_outputs.append(log)
+        self.log(f"test/loss", log, on_step=False, on_epoch=True, sync_dist=True)
+
+        return log
+
+
+"""
+@gin.configurable
+class TFTpytorch(TemporalFusionTransformer,DLPredictionWrapper):
+    
+    Implementation of https://arxiv.org/abs/1912.09363 
+    
+    supported_run_modes = [RunMode.classification, RunMode.regression]
+    def __init__(self,dataset,hidden,dropout,
+                 n_heads,dropout_att,lr,optimizer,num_classes,*args,**kwargs):
+        
+        DLPredictionWrapper.__init__(self,lr=lr,optimizer=optimizer,*args,**kwargs)        
+        self.model=TemporalFusionTransformer.from_dataset(dataset)
         
-)
     
+        
     def set_weight(self, weight, dataset):
-        """Set the weight for the loss function."""
+        
+        Set the weight for the loss function
+        
         if isinstance(weight, list):
             weight = FloatTensor(weight)
         elif weight == "balanced":
             weight = FloatTensor(dataset.get_balance())
-        self.loss_weights = weight
\ No newline at end of file
+        self.loss_weights = weight
+    
+    def test_step(self, batch, batch_idx):
+        x, y = batch
+        log, out = self.model.step(x, y, batch_idx)
+        log.update(self.model.create_log(x, y, out, batch_idx))
+        self.testing_step_outputs.append(log)
+        self.model.log(f"test/loss", log, on_step=False, on_epoch=True, sync_dist=True)
+
+        return log
+    def forward(self, x: Dict[str,Tensor]) -> Dict[str, Tensor]:
+        return self.model.forward(x)
+    
+"""
\ No newline at end of file
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 016cb366..79ee0e61 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -540,6 +540,7 @@ def forward(self, x: Dict[str, Tensor]):
                                             self.t_cat_k_embed,
                                             self.t_cont_k_embedding_vectors,
                                             self.t_cont_k_embedding_bias)
+        
         t_observed_inp = self._apply_embedding(t_cat_o_inp,
                                                t_cont_o_inp,
                                                self.t_cat_o_embed,
@@ -756,7 +757,7 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         enriched = self.enrichment_grn(temporal_features, c=ce)
 
         # Temporal self attention
-        x, _ = self.attention(enriched)
+        x, attn_prob= self.attention(enriched)
 
         # Don't compute hictorical quantiles
         x = x[:, self.encoder_length:, :]
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 21334a67..fdd753a9 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -8,7 +8,7 @@
 from pytorch_lightning.loggers import TensorBoardLogger
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint, TQDMProgressBar
-from pytorch_forecasting import TimeSeriesDataSet
+from pytorch_forecasting import TimeSeriesDataSet,TemporalFusionTransformer
 from pathlib import Path
 from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset,PredictionDatasetTFT,PredictionDatasetTFTpytorch
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
@@ -43,7 +43,7 @@ def train_common(
     test_on: str = Split.test,
     use_wandb: bool = False,
     cpu: bool = False,
-    verbose=False,
+    verbose=True,
     ram_cache=False,
     num_workers: int = min(cpu_core_count, torch.cuda.device_count() * 4 * int(torch.cuda.is_available()), 32),
 ):
@@ -84,14 +84,20 @@ def train_common(
     val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache)
     train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
-
+    test_dataset = dataset_class(data, split=test_on)
     logging.debug(f"Training on {len(train_dataset)} samples and validating on {len(val_dataset)} samples.")
     logging.info(f"Using {num_workers} workers for data loading.")
     if(model.__name__ =='TFTpytorch'):
         
         train_loader = train_dataset.to_dataloader(train=True, batch_size=batch_size, num_workers=num_workers,pin_memory=True,drop_last=True)
         val_loader = val_dataset.to_dataloader(train=False, batch_size=batch_size ,num_workers=num_workers,pin_memory=True,drop_last=True)
-        model=model(dataset=train_dataset,optimizer=optimizer, epochs=epochs, run_mode=mode)
+        test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=num_workers,pin_memory=True,drop_last=True,shuffle=False)
+        model=model(train_dataset,optimizer=optimizer, epochs=epochs, run_mode=mode)
+        model=model.model
+        
+        
+        
+        
     else:
         train_loader = DataLoader(
             train_dataset,
@@ -143,7 +149,7 @@ def train_common(
         max_epochs=epochs if model.needs_training else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="auto" if not cpu else "cpu",
+        accelerator= "cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic=reproducible,
         benchmark=not reproducible,
@@ -151,6 +157,8 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=0,
     )
+  #  print(trainer.test(model, dataloaders=test_loader, verbose=verbose))
+   # test_loss = trainer.test(model.model, dataloaders=test_loader, verbose=verbose)[0]["RMSE"]
 
     if model.needs_fit:
         logging.info("Fitting model to data.")
@@ -160,25 +168,28 @@ def train_common(
 
     if model.needs_training:
         logging.info("Training model.")
+        
+        
         trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
         logging.info("Training complete.")
-
     test_dataset = dataset_class(data, split=test_on)
     test_dataset = assure_minimum_length(test_dataset)
-    test_loader = (
-        DataLoader(
-            test_dataset,
-            batch_size=min(batch_size * 4, len(test_dataset)),
-            shuffle=False,
-            num_workers=num_workers,
-            pin_memory=True,
-            drop_last=True,
-        )
+    if(type(model).__name__ =='TFTpytorch'):
+        test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=num_workers,pin_memory=True,drop_last=True,shuffle=False)
+    else:
+       
+        test_loader = (
+            DataLoader(
+                test_dataset,
+                batch_size=min(batch_size * 4, len(test_dataset)),
+                shuffle=False,
+                num_workers=num_workers,
+                pin_memory=True,
+                drop_last=True,
+            )
         if model.needs_training
         else DataLoader([test_dataset.to_tensor()], batch_size=1)
     )
-
-    model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
     return test_loss
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index bf9452b1..eb2c5db2 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -243,15 +243,19 @@ def step_fn(self, element, step_prefix=""):
                 if(key=='target'):
                     labels=value.squeeze()
                 data[key] = value
-                
+        
+            
         elif len(element) == 2:
-            data, labels = element[0], element[1].to(self.device)
-            if isinstance(data, list):
-                for i in range(len(data)):
-                    data[i] = data[i].float().to(self.device)
+            if(isinstance(element[1] ,tuple)):
+                data, labels = element[0], element[1]
             else:
-                data = data.float().to(self.device)
-            mask = torch.ones_like(labels).bool()
+                data, labels = element[0], element[1].to(self.device)
+                if isinstance(data, list):
+                    for i in range(len(data)):
+                        data[i] = data[i].float().to(self.device)
+                else:
+                    data = data.float().to(self.device)
+                mask = torch.ones_like(labels).bool()
 
         elif len(element) == 3:
             data, labels, mask = element[0], element[1].to(self.device), element[2].to(self.device)
@@ -262,7 +266,10 @@ def step_fn(self, element, step_prefix=""):
                 data = data.float().to(self.device)
         else:
             raise Exception("Loader should return either (data, label) or (data, label, mask)")
+        
         out = self(data)
+        
+        
         if len(out) == 2 and isinstance(out, tuple):
             out, aux_loss = out
         else:

From 9e52574adfa1980fe53ffa09b7a8c782629b410e Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 1 Aug 2023 12:51:17 +0200
Subject: [PATCH 038/142] latest changes

---
 configs/prediction_models/TFT.gin             |  4 +-
 configs/prediction_models/TFTpytorch.gin      |  6 +-
 experiments/demo_benchmark_classification.yml | 22 +++---
 experiments/demo_benchmark_regression.yml     | 12 ++--
 icu_benchmarks/data/loader.py                 |  6 +-
 icu_benchmarks/models/dl_models.py            | 71 +++----------------
 icu_benchmarks/models/train.py                | 32 +++++----
 icu_benchmarks/models/wrappers.py             | 10 +--
 8 files changed, 58 insertions(+), 105 deletions(-)

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
index d03b8c21..3abcb239 100644
--- a/configs/prediction_models/TFT.gin
+++ b/configs/prediction_models/TFT.gin
@@ -13,11 +13,11 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 # Encoder params
 model/hyperparameter.class_to_tune = @TFT
 model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = 12
+model/hyperparameter.hidden = 256
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =1
+model/hyperparameter.n_heads =4
 model/hyperparameter.example_length=25
 
 
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 4a618903..8c320809 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -12,13 +12,13 @@ optimizer/hyperparameter.weight_decay = 1e-6
 
 # Encoder params
 model/hyperparameter.class_to_tune = @TFTpytorch
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.hidden = 128
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =1
+model/hyperparameter.n_heads =4
 model/hyperparameter.lr = (1e-5, 3e-4)
-PredictionDatasetTFTpytorch.max_encoder_length = 6
+PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 12
 
 
diff --git a/experiments/demo_benchmark_classification.yml b/experiments/demo_benchmark_classification.yml
index 58b14a45..7adac471 100644
--- a/experiments/demo_benchmark_classification.yml
+++ b/experiments/demo_benchmark_classification.yml
@@ -17,21 +17,21 @@ name: yaib_demo_classification_benchmark
 parameters:
   data_dir:
     values:
-      - demo_data/mortality24/eicu_demo
-      - demo_data/mortality24/mimic_demo
+     # - demo_data/mortality24/eicu_demo
+     # - demo_data/mortality24/mimic_demo
       - demo_data/aki/eicu_demo
-      - demo_data/aki/mimic_demo
-      - demo_data/sepsis/eicu_demo
+     # - demo_data/aki/mimic_demo
+     # - demo_data/sepsis/eicu_demo
       - demo_data/sepsis/mimic_demo
   model:
     values:
-   #   - LogisticRegression
-    #  - LGBMClassifier
-    #  - GRU
-    #  - LSTM
-    #  - TCN
-    #  - Transformer
-      - TemporalFusionTransformer
+      - LogisticRegression
+      - LGBMClassifier
+      - GRU
+      - LSTM
+      - TCN
+      - Transformer
+      - TFT
   seed:
     values:
       - 1111
diff --git a/experiments/demo_benchmark_regression.yml b/experiments/demo_benchmark_regression.yml
index 047f0c88..57f380b8 100644
--- a/experiments/demo_benchmark_regression.yml
+++ b/experiments/demo_benchmark_regression.yml
@@ -23,12 +23,12 @@ parameters:
    #   - demo_data/kf/mimic_demo
   model:
     values:
-  #    - ElasticNet
-  #    - LGBMRegressor
-  #    - GRU
-  #    - LSTM
-  #    - TCN
-      - TemporalFusionTransformer
+      - ElasticNet
+      - LGBMRegressor
+      - GRU
+      - LSTM
+      - TCN
+      - TFT
   seed:
     values:
       - 1111
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 82b34c08..a183eb41 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -379,8 +379,9 @@ def __init__(self,
         max_prediction_length: int,
         max_encoder_length: int,
         *args,
-        ram_cache: bool = True,
+        ram_cache: bool = False,
         **kwargs):
+
         data[split]["FEATURES"]["time_idx"]=((data[split]["FEATURES"]["time"]/ pd.Timedelta(seconds=3600))).astype(int)
         data=data.get(split)
         labels = data["OUTCOME"]
@@ -428,7 +429,8 @@ def __init__(self,
             ],
             add_relative_time_idx=True,
             add_target_scales=True,
-            add_encoder_length=True,)
+            add_encoder_length=True,
+            )
     def get_balance(self) -> list:
         """Return the weight balance for the split of interest.
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 500b6015..d6e075af 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -357,7 +357,7 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         return pred
 
 @gin.configurable
-class TFTpytorch(TemporalFusionTransformer,DLPredictionWrapper):
+class TFTpytorch(DLPredictionWrapper):
     """
     Implementation of https://arxiv.org/abs/1912.09363 
     """
@@ -365,76 +365,25 @@ class TFTpytorch(TemporalFusionTransformer,DLPredictionWrapper):
     def __init__(self,dataset,hidden,dropout,
                  n_heads,dropout_att,lr,optimizer,num_classes,*args,**kwargs):
         
-        DLPredictionWrapper.__init__(self,lr=lr,optimizer=optimizer,*args,**kwargs)        
-        
-        TemporalFusionTransformer.__init__(self)
-
+        super().__init__(lr=lr,optimizer=optimizer,*args,**kwargs)      
         self.model=TemporalFusionTransformer.from_dataset(dataset=dataset,hidden_size=hidden,dropout=dropout,
                  attention_head_size=n_heads,learning_rate=lr,optimizer=optimizer,loss=QuantileLoss(),hidden_continuous_size=hidden)
-        self.hparams.update(self.model.hparams)
-
-        #for key in model.hparams.keys():
-        # self.hparams[key]=model.hparams[key]
-       # del self.hparams["input_shape"]
-       # del self.hparams["input_size"]
-        #print(self.hparams)
-       # print(self.m)
+          
         self.logit = nn.Linear(7, num_classes)
-    def set_weight(self, weight, dataset):
-        """
-        Set the weight for the loss function
-        """
-        if isinstance(weight, list):
-            weight = FloatTensor(weight)
-        elif weight == "balanced":
-            weight = FloatTensor(dataset.get_balance())
-        self.loss_weights = weight
-    
-    def test_step(self, batch, batch_idx):
-        x, y = batch
-        log, out = self.step(x, y, batch_idx)
-        print("finally here")
-        log.update(self.create_log(x, y, out, batch_idx))
-        self.testing_step_outputs.append(log)
-        self.log(f"test/loss", log, on_step=False, on_epoch=True, sync_dist=True)
-
-        return log
-
-
-"""
-@gin.configurable
-class TFTpytorch(TemporalFusionTransformer,DLPredictionWrapper):
-    
-    Implementation of https://arxiv.org/abs/1912.09363 
-    
-    supported_run_modes = [RunMode.classification, RunMode.regression]
-    def __init__(self,dataset,hidden,dropout,
-                 n_heads,dropout_att,lr,optimizer,num_classes,*args,**kwargs):
         
-        DLPredictionWrapper.__init__(self,lr=lr,optimizer=optimizer,*args,**kwargs)        
-        self.model=TemporalFusionTransformer.from_dataset(dataset)
-        
-    
         
     def set_weight(self, weight, dataset):
-        
+        """
         Set the weight for the loss function
-        
+        """
         if isinstance(weight, list):
             weight = FloatTensor(weight)
         elif weight == "balanced":
             weight = FloatTensor(dataset.get_balance())
         self.loss_weights = weight
+    def forward(self,x):
+        out= self.model(x)
+        pred = self.logit(out["prediction"])
+        return pred
     
-    def test_step(self, batch, batch_idx):
-        x, y = batch
-        log, out = self.model.step(x, y, batch_idx)
-        log.update(self.model.create_log(x, y, out, batch_idx))
-        self.testing_step_outputs.append(log)
-        self.model.log(f"test/loss", log, on_step=False, on_epoch=True, sync_dist=True)
-
-        return log
-    def forward(self, x: Dict[str,Tensor]) -> Dict[str, Tensor]:
-        return self.model.forward(x)
-    
-"""
\ No newline at end of file
+   
\ No newline at end of file
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index fdd753a9..d833b5e7 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -15,6 +15,7 @@
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
+import pickle 
 cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
 
 
@@ -70,8 +71,12 @@ def train_common(
         verbose: Enable detailed logging.
         ram_cache: Whether to cache the data in RAM.
         num_workers: Number of workers to use for data loading.
+         with open('saved_dictionary.pkl', 'wb') as f:
+        pickle.dump(data, f)
+    data["train"]["OUTCOME"].to_csv('outcome.csv', index=False)
+    data["train"]["FEATURES"].to_csv('features.csv', index=False)
     """
-
+   
     logging.info(f"Training model: {model.__name__}.")
     dataset_class = (ImputationDataset if mode == RunMode.imputation
                  else (PredictionDatasetTFT if model.__name__ == 'TFT'
@@ -89,12 +94,10 @@ def train_common(
     logging.info(f"Using {num_workers} workers for data loading.")
     if(model.__name__ =='TFTpytorch'):
         
-        train_loader = train_dataset.to_dataloader(train=True, batch_size=batch_size, num_workers=num_workers,pin_memory=True,drop_last=True)
-        val_loader = val_dataset.to_dataloader(train=False, batch_size=batch_size ,num_workers=num_workers,pin_memory=True,drop_last=True)
-        test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=num_workers,pin_memory=True,drop_last=True,shuffle=False)
+        train_loader = train_dataset.to_dataloader(train=True, batch_size=batch_size, num_workers=0,pin_memory=False,drop_last=True)
+        val_loader = val_dataset.to_dataloader(train=False, batch_size=batch_size ,num_workers=0,pin_memory=False,drop_last=True)
+        test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=0,pin_memory=False,drop_last=True,shuffle=False)
         model=model(train_dataset,optimizer=optimizer, epochs=epochs, run_mode=mode)
-        model=model.model
-        
         
         
         
@@ -116,6 +119,7 @@ def train_common(
             drop_last=True,
         )
     
+    
         if(isinstance(next(iter(train_loader))[0] ,OrderedDict)):
             model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
         
@@ -123,6 +127,7 @@ def train_common(
             data_shape = next(iter(train_loader))[0].shape
         
             model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
+
     model.set_weight(weight, train_dataset)
     if load_weights:
         if source_dir.exists():
@@ -149,7 +154,7 @@ def train_common(
         max_epochs=epochs if model.needs_training else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator= "cpu",
+        accelerator= "auto" if not cpu else "cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic=reproducible,
         benchmark=not reproducible,
@@ -157,8 +162,7 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=0,
     )
-  #  print(trainer.test(model, dataloaders=test_loader, verbose=verbose))
-   # test_loss = trainer.test(model.model, dataloaders=test_loader, verbose=verbose)[0]["RMSE"]
+    
 
     if model.needs_fit:
         logging.info("Fitting model to data.")
@@ -168,14 +172,16 @@ def train_common(
 
     if model.needs_training:
         logging.info("Training model.")
-        
-        
-        trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
+        if(type(model).__name__ =='TFTpytorch'):
+            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
+        else:
+            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
         logging.info("Training complete.")
     test_dataset = dataset_class(data, split=test_on)
     test_dataset = assure_minimum_length(test_dataset)
     if(type(model).__name__ =='TFTpytorch'):
         test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=num_workers,pin_memory=True,drop_last=True,shuffle=False)
+        test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     else:
        
         test_loader = (
@@ -190,6 +196,6 @@ def train_common(
         if model.needs_training
         else DataLoader([test_dataset.to_tensor()], batch_size=1)
     )
-    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
+        test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
     return test_loss
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index eb2c5db2..d96d3e5a 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -96,7 +96,6 @@ class DLWrapper(BaseModule, ABC):
 
     def __init__(
         self,
-        
         loss=CrossEntropyLoss(),
         optimizer=torch.optim.Adam,
         run_mode: RunMode = RunMode.classification,
@@ -109,7 +108,6 @@ def __init__(
         epochs: int = 100,
         input_size: torch.Tensor = None,
         initialization_method: str = "normal",
-        
         **kwargs,
     ):
         
@@ -247,7 +245,8 @@ def step_fn(self, element, step_prefix=""):
             
         elif len(element) == 2:
             if(isinstance(element[1] ,tuple)):
-                data, labels = element[0], element[1]
+                data, labels = element[0], element[1][0]
+                mask = torch.ones_like(labels).bool()
             else:
                 data, labels = element[0], element[1].to(self.device)
                 if isinstance(data, list):
@@ -266,10 +265,7 @@ def step_fn(self, element, step_prefix=""):
                 data = data.float().to(self.device)
         else:
             raise Exception("Loader should return either (data, label) or (data, label, mask)")
-        
         out = self(data)
-        
-        
         if len(out) == 2 and isinstance(out, tuple):
             out, aux_loss = out
         else:
@@ -290,7 +286,7 @@ def step_fn(self, element, step_prefix=""):
         transformed_output = self.output_transform((prediction, target))
         for metric in self.metrics[step_prefix].values():
             metric.update(transformed_output)
-        self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
+        self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True, prog_bar=True)
         return loss
 
 

From 803d677b1ea6c27efd98af4c5e516db3f6fbad2d Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Tue, 8 Aug 2023 11:37:11 +0200
Subject: [PATCH 039/142] latest

---
 experiments/benchmark_classification.yml      | 38 ++++++++++---------
 experiments/benchmark_regression.yml          | 18 +++++----
 experiments/demo_benchmark_classification.yml | 11 +++---
 icu_benchmarks/models/dl_models.py            |  2 +-
 icu_benchmarks/models/train.py                |  1 +
 icu_benchmarks/models/wrappers.py             |  1 +
 6 files changed, 39 insertions(+), 32 deletions(-)

diff --git a/experiments/benchmark_classification.yml b/experiments/benchmark_classification.yml
index 61763ffa..a9d999ce 100644
--- a/experiments/benchmark_classification.yml
+++ b/experiments/benchmark_classification.yml
@@ -17,26 +17,28 @@ name: yaib_classification_benchmark
 parameters:
   data_dir:
     values:
-      - ../data/mortality24/miiv
-      - ../data/mortality24/hirid
-      - ../data/mortality24/eicu
-      - ../data/mortality24/aumc
-      - ../data/aki/miiv
-      - ../data/aki/hirid
-      - ../data/aki/eicu
-      - ../data/aki/aumc
-      - ../data/sepsis/miiv
-      - ../data/sepsis/hirid
-      - ../data/sepsis/eicu
-      - ../data/sepsis/aumc
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/aki/miiv
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/aki/hirid
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/aki/eicu
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/aki/aumc
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/morality24/miiv
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/morality24/miiv
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/morality24/miiv
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/morality24/miiv
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/sepsis/miiv
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/sepsis/hirid
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/sepsis/eicu
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/sepsis/aumc
   model:
     values:
-      - LogisticRegression
-      - LGBMClassifier
-      - GRU
-      - LSTM
-      - TCN
-      - Transformer
+    #  - LogisticRegression
+    #  - LGBMClassifier
+    #  - GRU
+    #  - LSTM
+    #  - TCN
+    #  - Transformer
+      - TFT
+      
   seed:
     values:
       - 1111
diff --git a/experiments/benchmark_regression.yml b/experiments/benchmark_regression.yml
index 8aa8d13e..ff23ca2e 100644
--- a/experiments/benchmark_regression.yml
+++ b/experiments/benchmark_regression.yml
@@ -17,14 +17,14 @@ name: yaib_regression_benchmark
 parameters:
   data_dir:
     values:
-      - ../data/los/miiv
-      - ../data/los/hirid
-      - ../data/los/eicu
-      - ../data/los/aumc
-      - ../data/kf/miiv
-      - ../data/kf/hirid
-      - ../data/kf/eicu
-      - ../data/kf/aumc
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/los/miiv
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/los/hirid
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/los/eicu
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/los/aumc
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/kf/miiv
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/kf/hirid
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/kf/eicu
+      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/kf/aumc
   model:
     values:
       - ElasticNet
@@ -33,6 +33,8 @@ parameters:
       - LSTM
       - TCN
       - Transformer
+      - TFT
+      
   seed:
     values:
       - 1111
diff --git a/experiments/demo_benchmark_classification.yml b/experiments/demo_benchmark_classification.yml
index 7adac471..341a40dd 100644
--- a/experiments/demo_benchmark_classification.yml
+++ b/experiments/demo_benchmark_classification.yml
@@ -17,11 +17,11 @@ name: yaib_demo_classification_benchmark
 parameters:
   data_dir:
     values:
-     # - demo_data/mortality24/eicu_demo
-     # - demo_data/mortality24/mimic_demo
-      - demo_data/aki/eicu_demo
-     # - demo_data/aki/mimic_demo
-     # - demo_data/sepsis/eicu_demo
+      - demo_data/mortality24/eicu_demo
+      - demo_data/mortality24/mimic_demo
+     #fails for some reason - demo_data/aki/eicu_demo
+      - demo_data/aki/mimic_demo
+      - demo_data/sepsis/eicu_demo
       - demo_data/sepsis/mimic_demo
   model:
     values:
@@ -32,6 +32,7 @@ parameters:
       - TCN
       - Transformer
       - TFT
+     # - TFTpytorch
   seed:
     values:
       - 1111
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index d6e075af..c3a1a7fd 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -368,7 +368,7 @@ def __init__(self,dataset,hidden,dropout,
         super().__init__(lr=lr,optimizer=optimizer,*args,**kwargs)      
         self.model=TemporalFusionTransformer.from_dataset(dataset=dataset,hidden_size=hidden,dropout=dropout,
                  attention_head_size=n_heads,learning_rate=lr,optimizer=optimizer,loss=QuantileLoss(),hidden_continuous_size=hidden)
-          
+        
         self.logit = nn.Linear(7, num_classes)
         
         
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index d833b5e7..868ebe85 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -101,6 +101,7 @@ def train_common(
         
         
         
+        
     else:
         train_loader = DataLoader(
             train_dataset,
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index d96d3e5a..267da7fa 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -244,6 +244,7 @@ def step_fn(self, element, step_prefix=""):
         
             
         elif len(element) == 2:
+           
             if(isinstance(element[1] ,tuple)):
                 data, labels = element[0], element[1][0]
                 mask = torch.ones_like(labels).bool()

From c2fcc74c14c834b9e69b342647d60e4784c79295 Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Mon, 14 Aug 2023 14:27:40 +0200
Subject: [PATCH 040/142] added range for layers and attention heads

---
 configs/prediction_models/TFT.gin | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
index 3abcb239..10428d9a 100644
--- a/configs/prediction_models/TFT.gin
+++ b/configs/prediction_models/TFT.gin
@@ -13,11 +13,11 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 # Encoder params
 model/hyperparameter.class_to_tune = @TFT
 model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = 256
+model/hyperparameter.hidden = (16,64,128,256)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =4
+model/hyperparameter.n_heads =(1,4)
 model/hyperparameter.example_length=25
 
 

From 5f18f72bfb569987216e43ff6a988dddbb3f6812 Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Mon, 14 Aug 2023 15:56:29 +0200
Subject: [PATCH 041/142] formatted and added comments

---
 configs/prediction_models/TFT.gin           |   4 +-
 configs/tasks/Regression.gin                |   2 +-
 experiments/benchmark_classification.yml    |  24 +-
 experiments/benchmark_regression.yml        |  16 +-
 icu_benchmarks/cross_validation.py          |   2 +-
 icu_benchmarks/data/constants.py            |   3 +-
 icu_benchmarks/data/loader.py               | 252 +++++++++-----
 icu_benchmarks/imputation/diffwave.py       |   2 +-
 icu_benchmarks/imputation/layers/s4layer.py | 143 ++++----
 icu_benchmarks/imputation/sssds4.py         |   2 +-
 icu_benchmarks/imputation/sssdsa.py         |   2 +-
 icu_benchmarks/models/dl_models.py          | 195 +++++++----
 icu_benchmarks/models/layers.py             | 368 +++++++++++---------
 icu_benchmarks/models/metrics.py            |  18 +-
 icu_benchmarks/models/train.py              |  98 +++---
 icu_benchmarks/models/utils.py              |   5 -
 icu_benchmarks/models/wrappers.py           |  31 +-
 17 files changed, 687 insertions(+), 480 deletions(-)

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
index 10428d9a..3abcb239 100644
--- a/configs/prediction_models/TFT.gin
+++ b/configs/prediction_models/TFT.gin
@@ -13,11 +13,11 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 # Encoder params
 model/hyperparameter.class_to_tune = @TFT
 model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = (16,64,128,256)
+model/hyperparameter.hidden = 256
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =(1,4)
+model/hyperparameter.n_heads =4
 model/hyperparameter.example_length=25
 
 
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index 5cf3f8d9..ed7f794a 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -8,7 +8,7 @@ include "configs/tasks/common/PredictionTaskVariables.gin"
 include "configs/tasks/common/CrossValidation.gin"
 
 # MODE SETTINGS
-Run.mode = "Regression"
+Run.mode = "RegressionTFT"
 NUM_CLASSES = 1
 HORIZON = 24
 train_common.weight = "balanced"
diff --git a/experiments/benchmark_classification.yml b/experiments/benchmark_classification.yml
index a9d999ce..e9a32b4c 100644
--- a/experiments/benchmark_classification.yml
+++ b/experiments/benchmark_classification.yml
@@ -17,18 +17,18 @@ name: yaib_classification_benchmark
 parameters:
   data_dir:
     values:
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/aki/miiv
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/aki/hirid
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/aki/eicu
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/aki/aumc
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/morality24/miiv
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/morality24/miiv
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/morality24/miiv
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/morality24/miiv
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/sepsis/miiv
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/sepsis/hirid
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/sepsis/eicu
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/sepsis/aumc
+      - data/YAIB_Preprint/aki/miiv
+      #- data/YAIB_Preprint/aki/hirid
+      #- data/YAIB_Preprint/aki/eicu
+      #- data/YAIB_Preprint/aki/aumc
+      - data/YAIB_Preprint/morality24/miiv
+      - data/YAIB_Preprint/morality24/miiv
+      - data/YAIB_Preprint/morality24/miiv
+      - data/YAIB_Preprint/morality24/miiv
+      #- data/YAIB_Preprint/sepsis/miiv
+      #- data/YAIB_Preprint/sepsis/hirid
+      #- data/YAIB_Preprint/sepsis/eicu
+      #- data/YAIB_Preprint/sepsis/aumc
   model:
     values:
     #  - LogisticRegression
diff --git a/experiments/benchmark_regression.yml b/experiments/benchmark_regression.yml
index ff23ca2e..516a4b0f 100644
--- a/experiments/benchmark_regression.yml
+++ b/experiments/benchmark_regression.yml
@@ -17,14 +17,14 @@ name: yaib_regression_benchmark
 parameters:
   data_dir:
     values:
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/los/miiv
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/los/hirid
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/los/eicu
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/los/aumc
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/kf/miiv
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/kf/hirid
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/kf/eicu
-      - /dhc/groups/icu-prediction/datasets/YAIB_Preprint/kf/aumc
+      - data/YAIB_Preprint/los/miiv
+      - data/YAIB_Preprint/los/hirid
+      - data/YAIB_Preprint/los/eicu
+      - data/YAIB_Preprint/los/aumc
+      - data/YAIB_Preprint/kf/miiv
+      - data/YAIB_Preprint/kf/hirid
+      - data/YAIB_Preprint/kf/eicu
+      - data/YAIB_Preprint/kf/aumc
   model:
     values:
       - ElasticNet
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 43e40bf1..81d8c3ae 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -101,7 +101,7 @@ def execute_repeated_cv(
                 verbose=verbose,
                 use_wandb=wandb,
             )
-            
+
             train_time = datetime.now() - start_time
 
             log_full_line(
diff --git a/icu_benchmarks/data/constants.py b/icu_benchmarks/data/constants.py
index c7f6c635..4e4a9c34 100644
--- a/icu_benchmarks/data/constants.py
+++ b/icu_benchmarks/data/constants.py
@@ -16,5 +16,6 @@ class VarType:
     sequence = "SEQUENCE"
     label = "LABEL"
 
+
 class FeatType:
-    FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target', 'id']
+    FEAT_NAMES = ["s_cat", "s_cont", "k_cat", "k_cont", "o_cat", "o_cont", "target", "id"]
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index a183eb41..c7e8bda2 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -3,7 +3,7 @@
 import pandas as pd
 import gin
 import numpy as np
-from torch import Tensor, cat, from_numpy, float32,empty,stack
+from torch import Tensor, cat, from_numpy, float32, empty, stack
 from torch.utils.data import Dataset
 import logging
 from typing import Dict, Tuple
@@ -13,6 +13,8 @@
 from .constants import FeatType as Features
 from collections import OrderedDict
 from pytorch_forecasting import TimeSeriesDataSet
+
+
 class CommonDataset(Dataset):
     """Common dataset: subclass of Torch Dataset that represents the data to learn on.
 
@@ -32,8 +34,10 @@ def __init__(
         self.vars = vars
         self.grouping_df = data[split][grouping_segment].set_index(self.vars["GROUP"])
         self.features_df = (
-            #drops time coulmn and sets index to stay_id
-            data[split][Segment.features].set_index(self.vars["GROUP"]).drop(labels=self.vars["SEQUENCE"], axis=1)
+            # drops time coulmn and sets index to stay_id
+            data[split][Segment.features]
+            .set_index(self.vars["GROUP"])
+            .drop(labels=self.vars["SEQUENCE"], axis=1)
         )
 
         # calculate basic info for the data
@@ -93,7 +97,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             return self._cached_dataset[idx]
 
         pad_value = 0.0
-        stay_id = self.outcome_df.index.unique()[idx]  
+        stay_id = self.outcome_df.index.unique()[idx]
 
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id].to_numpy()
@@ -104,8 +108,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             labels = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, labels], axis=0)
 
         length_diff = self.maxlen - window.shape[0]
-        
-        
+
         pad_mask = np.ones(window.shape[0])
 
         # Padding the array to fulfill size requirement
@@ -156,20 +159,23 @@ def to_tensor(self):
         data, labels = self.get_data_and_labels()
         return from_numpy(data).to(float32), from_numpy(labels).to(float32)
 
+
 @gin.configurable("PredictionDatasetTFT")
 class PredictionDatasetTFT(PredictionDataset):
     """Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.
+    We also need to feed the model the variables in a specific order
 
     Args:
         ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.
     """
 
     def __init__(self, *args, ram_cache: bool = True, **kwargs):
-        super().__init__(*args,ram_cache = True, **kwargs)
-        
+        super().__init__(*args, ram_cache=True, **kwargs)
 
     def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         """Function to sample from the data split of choice. Used for TFT.
+        The data needs to be given to the model in the following order 
+        [static categorical,static contious,known catergorical,known continous, observed categorical, observed continous,target ,id]
 
         Args:
             idx: A specific row index to sample.
@@ -181,62 +187,62 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             return self._cached_dataset[idx]
 
         pad_value = 0.0
-        stay_id = self.outcome_df.index.unique()[idx]  
+        stay_id = self.outcome_df.index.unique()[idx]
 
-        
         # We need to be sure that tensors are returned in the correct order to be processed correclty by tft
         tensors = [[] for _ in range(8)]
         for var in self.features_df.columns:
-            if  var == 'sex' :
+            if var == "sex":
                 tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
-            elif var == 'age' or var== 'height' or var== 'weight':
+            elif var == "age" or var == "height" or var == "weight":
                 tensors[1].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
             elif "MissingIndicator" in var:
                 tensors[4].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
-            else :
+            else:
                 tensors[5].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
-    
-        
+
         tensors[6].extend(self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float))
         tensors[7].append(np.asarray([stay_id]))
-        window_shape0=np.shape(tensors[0])[1]
+        window_shape0 = np.shape(tensors[0])[1]
 
         if len(tensors[6]) == 1:
             # only one label per stay, align with window
             tensors[6] = np.concatenate([np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0)
-        
+
         length_diff = self.maxlen - window_shape0
-        length_diff_id=self.maxlen - np.shape(tensors[7])[1]
         pad_mask = np.ones(window_shape0)
         # Padding the array to fulfill size requirement
-        
+
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
-            tensors[0]= np.concatenate([tensors[0], np.ones(( np.shape(tensors[0])[0],self.maxlen - np.shape(tensors[0])[1])) * pad_value], axis=1)
-            tensors[1]= np.concatenate([tensors[1], np.ones(( np.shape(tensors[1])[0],self.maxlen - np.shape(tensors[1])[1])) * pad_value], axis=1)
-            tensors[4]= np.concatenate([tensors[4], np.ones(( np.shape(tensors[4])[0],self.maxlen - np.shape(tensors[4])[1])) * pad_value], axis=1)
-            tensors[5]= np.concatenate([tensors[5], np.ones(( np.shape(tensors[5])[0],self.maxlen - np.shape(tensors[5])[1])) * pad_value], axis=1)
-            
-         
-            
+            tensors[0] = np.concatenate(
+                [tensors[0], np.ones((np.shape(tensors[0])[0], self.maxlen - np.shape(tensors[0])[1])) * pad_value], axis=1
+            )
+            tensors[1] = np.concatenate(
+                [tensors[1], np.ones((np.shape(tensors[1])[0], self.maxlen - np.shape(tensors[1])[1])) * pad_value], axis=1
+            )
+            tensors[4] = np.concatenate(
+                [tensors[4], np.ones((np.shape(tensors[4])[0], self.maxlen - np.shape(tensors[4])[1])) * pad_value], axis=1
+            )
+            tensors[5] = np.concatenate(
+                [tensors[5], np.ones((np.shape(tensors[5])[0], self.maxlen - np.shape(tensors[5])[1])) * pad_value], axis=1
+            )
+
             tensors[6] = np.concatenate([tensors[6], np.ones(self.maxlen - np.shape(tensors[6])[0]) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
-        tensors[7]= np.concatenate([tensors[7], np.ones(( np.shape(tensors[7])[0],self.maxlen - np.shape(tensors[7])[1])) * stay_id], axis=1)#should be done regardless of length_diff
+        tensors[7] = np.concatenate(
+            [tensors[7], np.ones((np.shape(tensors[7])[0], self.maxlen - np.shape(tensors[7])[1])) * stay_id], axis=1
+        )  # should be done regardless of length_diff
         not_labeled = np.argwhere(np.isnan(tensors[6]))
         if len(not_labeled) > 0:
             tensors[6][not_labeled] = -1
             pad_mask[not_labeled] = 0
-        tensors[6]=[tensors[6]]
+        tensors[6] = [tensors[6]]
         pad_mask = pad_mask.astype(bool)
-       
-        
-        
+
         tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)
         tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]
-        return  OrderedDict(zip(Features.FEAT_NAMES, tensors)),from_numpy(pad_mask)
-
-    
-
+        return OrderedDict(zip(Features.FEAT_NAMES, tensors)), from_numpy(pad_mask)
 
 
 @gin.configurable("ImputationDataset")
@@ -316,7 +322,7 @@ class ImputationPredictionDataset(Dataset):
         select_columns (List[str], optional): the columns to serve as input for the imputation model. Defaults to None.
         ram_cache (bool, optional): wether the dataset should be stored in ram. Defaults to True.
     """
-    
+
     def __init__(
         self,
         data: DataFrame,
@@ -371,80 +377,168 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         return from_numpy(window.values).to(float32)
 
+
 @gin.configurable("PredictionDatasetTFTpytorch")
 class PredictionDatasetTFTpytorch(TimeSeriesDataSet):
-    def __init__(self,
+    """Subclass of timeseries dataset works with pyotrch forecasting library .
+
+    Args:
+        data (DataFrame): dict of the different splits of the data
+        split: Either 'train','val' or 'test'
+        max_prediction_length: maximum number of time steps to predict,
+        max_encoder_length: maximum length of input sequence to give the model,
+        ram_cache (bool, optional): wether the dataset should be stored in ram. Defaults to True.
+    """
+
+    def __init__(
+        self,
         data: dict,
-        split: str ,
+        split: str,
         max_prediction_length: int,
         max_encoder_length: int,
         *args,
         ram_cache: bool = False,
-        **kwargs):
-
-        data[split]["FEATURES"]["time_idx"]=((data[split]["FEATURES"]["time"]/ pd.Timedelta(seconds=3600))).astype(int)
-        data=data.get(split)
+        **kwargs
+    ):
+        data[split]["FEATURES"]["time_idx"] = ((data[split]["FEATURES"]["time"] / pd.Timedelta(seconds=3600))).astype(
+            int
+        )  # create an incremental column indicating the time step(required by constructor)
+        data = data.get(split)  # get split
         labels = data["OUTCOME"]
-        features=data["FEATURES"]
-        self.data=pd.merge(labels, features, on=['stay_id', 'time'])
+        features = data["FEATURES"]
+        self.data = pd.merge(labels, features, on=["stay_id", "time"])  # combine labels and features
         self.split = split
-        self.args=args
-        self.ram_cache=ram_cache
-        self.kwargs=kwargs
-        self.column_names=features.columns
-        super().__init__(data=self.data,
+        self.args = args
+        self.ram_cache = ram_cache
+        self.kwargs = kwargs
+        self.column_names = features.columns
+        super().__init__(
+            data=self.data,
             time_idx="time_idx",
             target="label",
             group_ids=["stay_id"],
-            min_encoder_length=max_encoder_length // 2,  
+            min_encoder_length=max_encoder_length // 2,
             max_encoder_length=max_encoder_length,
             min_prediction_length=2,
             max_prediction_length=max_prediction_length,
             static_categoricals=[],
-            static_reals=["height", "weight","age","sex"],
+            static_reals=["height", "weight", "age", "sex"],
             time_varying_known_categoricals=[],
             time_varying_known_reals=[],
             time_varying_unknown_categoricals=[],
             time_varying_unknown_reals=[
-                "alb", "alp", "alt", "ast", "be", "bicar", "bili", "bili_dir", "bnd", "bun", "ca", "cai", "ck", "ckmb", "cl",
-        "crea", "crp", "dbp", "fgn", "fio2", "glu", "hgb", "hr", "inr_pt", "k", "lact", "lymph", "map", "mch", "mchc", "mcv",
-        "methb", "mg", "na", "neut", "o2sat", "pco2", "ph", "phos", "plt", "po2", "ptt", "resp", "sbp", "temp", "tnt", "urine",
-        "wbc",'MissingIndicator_1', 'MissingIndicator_2',
-       'MissingIndicator_3', 'MissingIndicator_4', 'MissingIndicator_5',
-       'MissingIndicator_6', 'MissingIndicator_7', 'MissingIndicator_8',
-       'MissingIndicator_9', 'MissingIndicator_10', 'MissingIndicator_11',
-       'MissingIndicator_12', 'MissingIndicator_13', 'MissingIndicator_14',
-       'MissingIndicator_15', 'MissingIndicator_16', 'MissingIndicator_17',
-       'MissingIndicator_18', 'MissingIndicator_19', 'MissingIndicator_20',
-       'MissingIndicator_21', 'MissingIndicator_22', 'MissingIndicator_23',
-       'MissingIndicator_24', 'MissingIndicator_25', 'MissingIndicator_26',
-       'MissingIndicator_27', 'MissingIndicator_28', 'MissingIndicator_29',
-       'MissingIndicator_30', 'MissingIndicator_31', 'MissingIndicator_32',
-       'MissingIndicator_33', 'MissingIndicator_34', 'MissingIndicator_35',
-       'MissingIndicator_36', 'MissingIndicator_37', 'MissingIndicator_38',
-       'MissingIndicator_39', 'MissingIndicator_40', 'MissingIndicator_41',
-       'MissingIndicator_42', 'MissingIndicator_43', 'MissingIndicator_44',
-       'MissingIndicator_45', 'MissingIndicator_46', 'MissingIndicator_47',
-       'MissingIndicator_48'
+                "alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
+                "MissingIndicator_1",
+                "MissingIndicator_2",
+                "MissingIndicator_3",
+                "MissingIndicator_4",
+                "MissingIndicator_5",
+                "MissingIndicator_6",
+                "MissingIndicator_7",
+                "MissingIndicator_8",
+                "MissingIndicator_9",
+                "MissingIndicator_10",
+                "MissingIndicator_11",
+                "MissingIndicator_12",
+                "MissingIndicator_13",
+                "MissingIndicator_14",
+                "MissingIndicator_15",
+                "MissingIndicator_16",
+                "MissingIndicator_17",
+                "MissingIndicator_18",
+                "MissingIndicator_19",
+                "MissingIndicator_20",
+                "MissingIndicator_21",
+                "MissingIndicator_22",
+                "MissingIndicator_23",
+                "MissingIndicator_24",
+                "MissingIndicator_25",
+                "MissingIndicator_26",
+                "MissingIndicator_27",
+                "MissingIndicator_28",
+                "MissingIndicator_29",
+                "MissingIndicator_30",
+                "MissingIndicator_31",
+                "MissingIndicator_32",
+                "MissingIndicator_33",
+                "MissingIndicator_34",
+                "MissingIndicator_35",
+                "MissingIndicator_36",
+                "MissingIndicator_37",
+                "MissingIndicator_38",
+                "MissingIndicator_39",
+                "MissingIndicator_40",
+                "MissingIndicator_41",
+                "MissingIndicator_42",
+                "MissingIndicator_43",
+                "MissingIndicator_44",
+                "MissingIndicator_45",
+                "MissingIndicator_46",
+                "MissingIndicator_47",
+                "MissingIndicator_48",
             ],
             add_relative_time_idx=True,
             add_target_scales=True,
             add_encoder_length=True,
-            )
+        )
+
     def get_balance(self) -> list:
         """Return the weight balance for the split of interest.
 
         Returns:
             Weights for each label.
         """
-        
+
         counts = self.data["target"][0].unique(return_counts=True)
-        
 
-        return list((1/counts[1]) * counts[1].sum() / counts[0].shape[0])
+        return list((1 / counts[1]) * counts[1].sum() / counts[0].shape[0])
+
     def get_feature_names(self):
-        
         return self.column_names
-        
-        
-            
\ No newline at end of file
diff --git a/icu_benchmarks/imputation/diffwave.py b/icu_benchmarks/imputation/diffwave.py
index 0945c0e6..a1f04fd8 100644
--- a/icu_benchmarks/imputation/diffwave.py
+++ b/icu_benchmarks/imputation/diffwave.py
@@ -301,7 +301,7 @@ def forward(self, input_data):
         cond = self.cond_conv(cond)
         h += cond
 
-        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels:, :])
+        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels :, :])
 
         res = self.res_conv(out)
         assert x.shape == res.shape
diff --git a/icu_benchmarks/imputation/layers/s4layer.py b/icu_benchmarks/imputation/layers/s4layer.py
index 0691c7ef..1441a834 100644
--- a/icu_benchmarks/imputation/layers/s4layer.py
+++ b/icu_benchmarks/imputation/layers/s4layer.py
@@ -92,6 +92,7 @@ def _resolve_conj(x):
     def _resolve_conj(x):
         return x.conj()
 
+
 """ simple nn.Module components """
 
 
@@ -162,16 +163,16 @@ def forward(self, x):
 
 
 def LinearActivation(
-        d_input,
-        d_output,
-        bias=True,
-        zero_bias_init=False,
-        transposed=False,
-        initializer=None,
-        activation=None,
-        activate=False,  # Apply activation as part of this module
-        weight_norm=False,
-        **kwargs,
+    d_input,
+    d_output,
+    bias=True,
+    zero_bias_init=False,
+    transposed=False,
+    initializer=None,
+    activation=None,
+    activate=False,  # Apply activation as part of this module
+    weight_norm=False,
+    **kwargs,
 ):
     """Returns a linear nn.Module with control over axes order, initialization, and activation"""
 
@@ -385,7 +386,7 @@ def rank_correction(measure, N, rank=1, dtype=torch.float):
         P = torch.stack([P0, P1], dim=0)  # (2 N)
     elif measure == "lagt":
         assert rank >= 1
-        P = 0.5 ** 0.5 * torch.ones(1, N, dtype=dtype)
+        P = 0.5**0.5 * torch.ones(1, N, dtype=dtype)
     elif measure == "fourier":
         P = torch.ones(N, dtype=dtype)  # (N)
         P0 = P.clone()
@@ -509,18 +510,18 @@ def _omega(self, L, dtype, device, cache=True):
         return omega, z
 
     def __init__(
-            self,
-            L,
-            w,
-            P,
-            B,
-            C,
-            log_dt,
-            hurwitz=False,
-            trainable=None,
-            lr=None,
-            tie_state=False,
-            length_correction=True,
+        self,
+        L,
+        w,
+        P,
+        B,
+        C,
+        log_dt,
+        hurwitz=False,
+        trainable=None,
+        lr=None,
+        tie_state=False,
+        length_correction=True,
     ):
         """
         L: Maximum length; this module computes an SSM kernel of length L
@@ -684,23 +685,23 @@ def forward(self, state=None, rate=1.0, L=None):
             k_f = r[:-1, :-1, :, :] - r[:-1, -1:, :, :] * r[-1:, :-1, :, :] / (1 + r[-1:, -1:, :, :])
         elif self.rank == 2:
             r00 = r[: -self.rank, : -self.rank, :, :]
-            r01 = r[: -self.rank, -self.rank:, :, :]
-            r10 = r[-self.rank:, : -self.rank, :, :]
-            r11 = r[-self.rank:, -self.rank:, :, :]
+            r01 = r[: -self.rank, -self.rank :, :, :]
+            r10 = r[-self.rank :, : -self.rank, :, :]
+            r11 = r[-self.rank :, -self.rank :, :, :]
             det = (1 + r11[:1, :1, :, :]) * (1 + r11[1:, 1:, :, :]) - r11[:1, 1:, :, :] * r11[1:, :1, :, :]
             s = (
-                    r01[:, :1, :, :] * (1 + r11[1:, 1:, :, :]) * r10[:1, :, :, :]
-                    + r01[:, 1:, :, :] * (1 + r11[:1, :1, :, :]) * r10[1:, :, :, :]
-                    - r01[:, :1, :, :] * (r11[:1, 1:, :, :]) * r10[1:, :, :, :]
-                    - r01[:, 1:, :, :] * (r11[1:, :1, :, :]) * r10[:1, :, :, :]
+                r01[:, :1, :, :] * (1 + r11[1:, 1:, :, :]) * r10[:1, :, :, :]
+                + r01[:, 1:, :, :] * (1 + r11[:1, :1, :, :]) * r10[1:, :, :, :]
+                - r01[:, :1, :, :] * (r11[:1, 1:, :, :]) * r10[1:, :, :, :]
+                - r01[:, 1:, :, :] * (r11[1:, :1, :, :]) * r10[:1, :, :, :]
             )
             s = s / det
             k_f = r00 - s
         else:
             r00 = r[: -self.rank, : -self.rank, :, :]
-            r01 = r[: -self.rank, -self.rank:, :, :]
-            r10 = r[-self.rank:, : -self.rank, :, :]
-            r11 = r[-self.rank:, -self.rank:, :, :]
+            r01 = r[: -self.rank, -self.rank :, :, :]
+            r10 = r[-self.rank :, : -self.rank, :, :]
+            r11 = r[-self.rank :, -self.rank :, :, :]
             r11 = rearrange(r11, "a b h n -> h n a b")
             r11 = torch.linalg.inv(torch.eye(self.rank, device=r.device) + r11)
             r11 = rearrange(r11, "h n a b -> a b h n")
@@ -737,7 +738,7 @@ def _setup_linear(self):
         dt = torch.exp(self.log_dt)
         D = (2.0 / dt.unsqueeze(-1) - w).reciprocal()  # (H, N)
         R = (
-                torch.eye(self.rank, dtype=w.dtype, device=w.device) + 2 * contract("r h n, h n, s h n -> h r s", Q, D, P).real
+            torch.eye(self.rank, dtype=w.dtype, device=w.device) + 2 * contract("r h n, h n, s h n -> h r s", Q, D, P).real
         )  # (H r r)
         Q_D = rearrange(Q * D, "r h n -> h r n")
         R = torch.linalg.solve(R.to(Q_D), Q_D)  # (H r N)
@@ -778,8 +779,8 @@ def _step_state_linear(self, u=None, state=None):
 
             def contract_fn(p, x, y):
                 return contract("r h n, r h m, ... h m -> ... h n", _conj(p), _conj(x), _conj(y))[
-                       ..., : self.N
-                       ]  # inner outer product
+                    ..., : self.N
+                ]  # inner outer product
 
         else:
             assert state.size(-1) == 2 * self.N
@@ -940,24 +941,24 @@ class HippoSSKernel(nn.Module):
     """
 
     def __init__(
-            self,
-            H,
-            N=64,
-            L=1,
-            measure="legs",
-            rank=1,
-            channels=1,  # 1-dim to C-dim map; can think of C as having separate "heads"
-            dt_min=0.001,
-            dt_max=0.1,
-            trainable=None,  # Dictionary of options to train various HiPPO parameters
-            lr=None,  # Hook to set LR of hippo parameters differently
-            length_correction=True,  # Multiply by I-A|^L after initialization; can be turned off for initialization speed
-            hurwitz=False,
-            tie_state=False,  # Tie parameters of HiPPO ODE across the H features
-            precision=1,  # 1 (single) or 2 (double) for the kernel
-            resample=False,  # If given inputs of different lengths, adjust the sampling rate.
-            # Note that L should always be provided in this case, as it assumes that L is the true underlying
-            # length of the continuous signal
+        self,
+        H,
+        N=64,
+        L=1,
+        measure="legs",
+        rank=1,
+        channels=1,  # 1-dim to C-dim map; can think of C as having separate "heads"
+        dt_min=0.001,
+        dt_max=0.1,
+        trainable=None,  # Dictionary of options to train various HiPPO parameters
+        lr=None,  # Hook to set LR of hippo parameters differently
+        length_correction=True,  # Multiply by I-A|^L after initialization; can be turned off for initialization speed
+        hurwitz=False,
+        tie_state=False,  # Tie parameters of HiPPO ODE across the H features
+        precision=1,  # 1 (single) or 2 (double) for the kernel
+        resample=False,  # If given inputs of different lengths, adjust the sampling rate.
+        # Note that L should always be provided in this case, as it assumes that L is the true underlying
+        # length of the continuous signal
     ):
         super().__init__()
         self.N = N
@@ -1007,23 +1008,23 @@ def get_torch_trans(heads=8, layers=1, channels=64):
 
 class S4(nn.Module):
     def __init__(
-            self,
-            d_model,
-            d_state=64,
-            l_max=1,  # Maximum length of sequence. Fine if not provided: the kernel will keep doubling in length until longer
-            # than sequence. However, this can be marginally slower if the true length is not a power of 2
-            channels=1,  # maps 1-dim to C-dim
-            bidirectional=False,
-            # Arguments for FF
-            activation="gelu",  # activation in between SS and FF
-            postact=None,  # activation after FF
-            initializer=None,  # initializer on FF
-            weight_norm=False,  # weight normalization on FF
-            hyper_act=None,  # Use a "hypernetwork" multiplication
-            dropout=0.0,
-            transposed=True,  # axis ordering (B, L, D) or (B, D, L)
-            # SSM Kernel arguments
-            **kernel_args,
+        self,
+        d_model,
+        d_state=64,
+        l_max=1,  # Maximum length of sequence. Fine if not provided: the kernel will keep doubling in length until longer
+        # than sequence. However, this can be marginally slower if the true length is not a power of 2
+        channels=1,  # maps 1-dim to C-dim
+        bidirectional=False,
+        # Arguments for FF
+        activation="gelu",  # activation in between SS and FF
+        postact=None,  # activation after FF
+        initializer=None,  # initializer on FF
+        weight_norm=False,  # weight normalization on FF
+        hyper_act=None,  # Use a "hypernetwork" multiplication
+        dropout=0.0,
+        transposed=True,  # axis ordering (B, L, D) or (B, D, L)
+        # SSM Kernel arguments
+        **kernel_args,
     ):
         """
         d_state: the dimension of the state, also denoted by N
diff --git a/icu_benchmarks/imputation/sssds4.py b/icu_benchmarks/imputation/sssds4.py
index ce6e8c0d..d359fb09 100644
--- a/icu_benchmarks/imputation/sssds4.py
+++ b/icu_benchmarks/imputation/sssds4.py
@@ -298,7 +298,7 @@ def forward(self, input_data):
 
         h = self.S42(h.permute(2, 0, 1)).permute(1, 2, 0)
 
-        out = torch.tanh(h[:, :self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels:, :])
+        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels :, :])
 
         res = self.res_conv(out)
         assert x.shape == res.shape
diff --git a/icu_benchmarks/imputation/sssdsa.py b/icu_benchmarks/imputation/sssdsa.py
index fb384efe..55764104 100644
--- a/icu_benchmarks/imputation/sssdsa.py
+++ b/icu_benchmarks/imputation/sssdsa.py
@@ -389,7 +389,7 @@ def step(self, x, state, **kwargs):
             if self.unet:
                 for i in range(skipped):
                     next_state.append(state.pop())
-                u_layers = list(self.u_layers)[skipped // 3:]
+                u_layers = list(self.u_layers)[skipped // 3 :]
             else:
                 for i in range(skipped):
                     for _ in range(len(self.u_layers[i])):
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index c3a1a7fd..c5360da3 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -3,13 +3,22 @@
 import numpy as np
 import torch.nn as nn
 from icu_benchmarks.contants import RunMode
-from icu_benchmarks.models.layers import TransformerBlock, LocalBlock, TemporalBlock, PositionalEncoding,LazyEmbedding,StaticCovariateEncoder,TFTBack
+from icu_benchmarks.models.layers import (
+    TransformerBlock,
+    LocalBlock,
+    TemporalBlock,
+    PositionalEncoding,
+    LazyEmbedding,
+    StaticCovariateEncoder,
+    TFTBack,
+)
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor,cat,jit,FloatTensor
-from pytorch_forecasting import TemporalFusionTransformer,TimeSeriesDataSet
+from torch import Tensor, cat, FloatTensor
+from pytorch_forecasting import TemporalFusionTransformer
+
+from pytorch_forecasting.metrics import QuantileLoss
 
-from pytorch_forecasting.metrics import  QuantileLoss
 
 @gin.configurable
 class RNNet(DLPredictionWrapper):
@@ -285,93 +294,150 @@ def forward(self, x):
         pred = self.logit(o)
         return pred
 
+
 @gin.configurable
 class TFT(DLPredictionWrapper):
-    """ 
-    Implementation of https://arxiv.org/abs/1912.09363 
     """
-
+    Implementation of https://arxiv.org/abs/1912.09363 from https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/Forecasting/TFT
+    """
 
     _supported_run_modes = [RunMode.classification, RunMode.regression]
-    def __init__(self,num_classes, encoder_length,hidden,dropout,
-                 n_heads,dropout_att,example_length,*args,quantiles=[0.1, 0.5, 0.9],static_categorical_inp_size=[2],temporal_known_categorical_inp_size=[],
-    temporal_observed_categorical_inp_size=[48],static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
-    temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs):
-        #derived variables
-        num_static_vars=len(static_categorical_inp_size)+static_continuous_inp_size
-        num_future_vars=len(temporal_known_categorical_inp_size)+temporal_known_continuous_inp_size
-        num_historic_vars=sum([num_future_vars,
-                                      temporal_observed_continuous_inp_size,
-                                      temporal_target_size,
-                                      len(temporal_observed_categorical_inp_size),
-                                      ])
-        
-        super().__init__(num_classes=num_classes, encoder_length=encoder_length,hidden=hidden,
-                 n_heads=n_heads,dropout_att=dropout_att,example_length=example_length,quantiles=quantiles,
-                 num_static_vars=num_static_vars,num_future_vars=num_future_vars,num_historic_vars=num_historic_vars,*args,static_categorical_inp_size=1,temporal_known_categorical_inp_size=0,
-                temporal_observed_categorical_inp_size=48,static_continuous_inp_size=3,temporal_known_continuous_inp_size=0,
-                temporal_observed_continuous_inp_size=48,temporal_target_size=1,**kwargs)
-
-        
-        
-        self.encoder_length = encoder_length #this determines from how distant past we want to use data from
-
-        self.embedding = LazyEmbedding(static_categorical_inp_size,temporal_known_categorical_inp_size,
-                temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
-                temporal_observed_continuous_inp_size,temporal_target_size,hidden)
-        
-        self.static_encoder = StaticCovariateEncoder(num_static_vars,hidden,dropout)
-        self.TFTpart2 = TFTBack(encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
-                n_heads,dropout_att,example_length,quantiles)
-        self.logit = nn.Linear(len(quantiles), num_classes)
+
+    def __init__(
+        self,
+        num_classes,
+        encoder_length,  # determines interval to use for prediction
+        hidden,
+        dropout,
+        n_heads,
+        dropout_att,
+        example_length,  # determines interval to predict
+        *args,
+        quantiles=[0.1, 0.5, 0.9],  # quantiles to produce
+        static_categorical_inp_size=[2],  # number of catergories
+        temporal_known_categorical_inp_size=[],
+        temporal_observed_categorical_inp_size=[48],  # number of categorical observed variables
+        static_continuous_inp_size=3,  # number of static coutinous variables
+        temporal_known_continuous_inp_size=0,
+        temporal_observed_continuous_inp_size=48,
+        temporal_target_size=1,  # number of target variables
+        **kwargs,
+    ):
+        # derived variables
+        num_static_vars = len(static_categorical_inp_size) + static_continuous_inp_size
+        num_future_vars = len(temporal_known_categorical_inp_size) + temporal_known_continuous_inp_size
+        num_historic_vars = sum(
+            [
+                num_future_vars,
+                temporal_observed_continuous_inp_size,
+                temporal_target_size,
+                len(temporal_observed_categorical_inp_size),
+            ]
+        )
+
+        super().__init__(
+            num_classes=num_classes,
+            encoder_length=encoder_length,
+            hidden=hidden,
+            n_heads=n_heads,
+            dropout_att=dropout_att,
+            example_length=example_length,
+            quantiles=quantiles,
+            num_static_vars=num_static_vars,
+            num_future_vars=num_future_vars,
+            num_historic_vars=num_historic_vars,
+            *args,
+            static_categorical_inp_size=1,
+            temporal_known_categorical_inp_size=0,
+            temporal_observed_categorical_inp_size=48,
+            static_continuous_inp_size=3,
+            temporal_known_continuous_inp_size=0,
+            temporal_observed_continuous_inp_size=48,
+            temporal_target_size=1,
+            **kwargs,
+        )
+
+        self.encoder_length = encoder_length  # this determines from how distant past we want to use data from
+
+        self.embedding = LazyEmbedding(
+            static_categorical_inp_size,
+            temporal_known_categorical_inp_size,
+            temporal_observed_categorical_inp_size,
+            static_continuous_inp_size,
+            temporal_known_continuous_inp_size,
+            temporal_observed_continuous_inp_size,
+            temporal_target_size,
+            hidden,
+        )  # embeddings for all variables
+
+        self.static_encoder = StaticCovariateEncoder(num_static_vars, hidden, dropout)  # encoding for static variables
+        self.TFTpart2 = TFTBack(
+            encoder_length,
+            num_historic_vars,
+            hidden,
+            dropout,
+            num_future_vars,
+            n_heads,
+            dropout_att,
+            example_length,
+            quantiles,
+        )  # The main part of the TFT
+        self.logit = nn.Linear(
+            len(quantiles), num_classes
+        )  # Linear layer on top to output to the number of classes and allow modification by predictionwrapper
 
     def forward(self, x: Dict[str, Tensor]) -> Tensor:
-        
-       
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
         # Static context
         cs, ce, ch, cc = self.static_encoder(s_inp)
-        ch, cc = ch.unsqueeze(0), cc.unsqueeze(0) #lstm initial states
+        ch, cc = ch.unsqueeze(0), cc.unsqueeze(0)  # lstm initial states
         # Temporal input
-        
+
         _historical_inputs = []
-        
+
         # Check for t_observed_inp
         if t_observed_inp is not None:
-            _historical_inputs.append(t_observed_inp[:, :self.encoder_length, :])
+            _historical_inputs.append(t_observed_inp[:, : self.encoder_length, :])
 
         # Check for t_known_inp
         if t_known_inp is not None:
-            _historical_inputs.append(t_known_inp[:, :self.encoder_length, :])
+            _historical_inputs.append(t_known_inp[:, : self.encoder_length, :])
         # Check for t_observed_tgt
         if t_observed_tgt is not None:
-            _historical_inputs.append(t_observed_tgt[:, :self.encoder_length, :])
+            _historical_inputs.append(t_observed_tgt[:, : self.encoder_length, :])
         historical_inputs = cat(_historical_inputs, dim=-2)
-        future_inputs= Tensor()
+        future_inputs = Tensor()
         if t_known_inp is not None:
-            future_inputs = t_known_inp[:, self.encoder_length:]
-        
-        
-        o=self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
+            future_inputs = t_known_inp[:, self.encoder_length :]
+
+        o = self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
         pred = self.logit(o)
         return pred
 
+
 @gin.configurable
 class TFTpytorch(DLPredictionWrapper):
     """
-    Implementation of https://arxiv.org/abs/1912.09363 
+    Implementation of https://arxiv.org/abs/1912.09363 from pytorch forecasting
     """
+
     supported_run_modes = [RunMode.classification, RunMode.regression]
-    def __init__(self,dataset,hidden,dropout,
-                 n_heads,dropout_att,lr,optimizer,num_classes,*args,**kwargs):
-        
-        super().__init__(lr=lr,optimizer=optimizer,*args,**kwargs)      
-        self.model=TemporalFusionTransformer.from_dataset(dataset=dataset,hidden_size=hidden,dropout=dropout,
-                 attention_head_size=n_heads,learning_rate=lr,optimizer=optimizer,loss=QuantileLoss(),hidden_continuous_size=hidden)
-        
+
+    def __init__(self, dataset, hidden, dropout, n_heads, dropout_att, lr, optimizer, num_classes, *args, **kwargs):
+        super().__init__(lr=lr, optimizer=optimizer, *args, **kwargs)
+        self.model = TemporalFusionTransformer.from_dataset(
+            dataset=dataset,
+            hidden_size=hidden,
+            dropout=dropout,
+            attention_head_size=n_heads,
+            learning_rate=lr,
+            optimizer=optimizer,
+            loss=QuantileLoss(),
+            hidden_continuous_size=hidden,
+        )
+
         self.logit = nn.Linear(7, num_classes)
-        
-        
+
     def set_weight(self, weight, dataset):
         """
         Set the weight for the loss function
@@ -381,9 +447,8 @@ def set_weight(self, weight, dataset):
         elif weight == "balanced":
             weight = FloatTensor(dataset.get_balance())
         self.loss_weights = weight
-    def forward(self,x):
-        out= self.model(x)
+
+    def forward(self, x):
+        out = self.model(x)
         pred = self.logit(out["prediction"])
         return pred
-    
-   
\ No newline at end of file
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 79ee0e61..7307bb52 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -319,26 +319,29 @@ def forward(self, x):
         res = x if self.downsample is None else self.downsample(x)
         return self.relu(out + res)
 
+
 class MaybeLayerNorm(nn.Module):
-    '''
+    """
     Implements layer normalization or identity function depending on output_size
-    '''
+    """
+
     def __init__(self, output_size, hidden, eps):
         super().__init__()
         if output_size and output_size == 1:
             self.ln = nn.Identity()
         else:
             self.ln = LayerNorm(output_size if output_size else hidden, eps=eps)
-    
+
     def forward(self, x):
         return self.ln(x)
 
 
 class GLU(nn.Module):
-    '''
+    """
     Gated Linear Unit consists of a linear layer followed by a GLU where input is split in half along dim to form a and b
-    GLU(a,b)=a ⊗ σ(b)where σ is signmoid activation and ⊗ is element-wise product 
-    '''
+    GLU(a,b)=a ⊗ σ(b)where σ is signmoid activation and ⊗ is element-wise product
+    """
+
     def __init__(self, hidden, output_size):
         super().__init__()
         self.lin = nn.Linear(hidden, output_size * 2)
@@ -348,24 +351,27 @@ def forward(self, x: Tensor) -> Tensor:
         x = F.glu(x)
         return x
 
+
 class GRN(nn.Module):
-    '''
+    """
     Gated Residual Network consists of a maybe normalization layer -->linear --> ELU -->linear-->GLU
-     in addition to a residual connection 
-    '''
-    def __init__(self,
-                 input_size,
-                 hidden,
-                 output_size=None,
-                 context_hidden=None,
-                 dropout=0.0,):
+     in addition to a residual connection
+    """
+
+    def __init__(
+        self,
+        input_size,
+        hidden,
+        output_size=None,
+        context_hidden=None,
+        dropout=0.0,
+    ):
         super().__init__()
-        
+
         self.layer_norm = MaybeLayerNorm(output_size, hidden, eps=1e-3)
-        
+
         self.lin_a = nn.Linear(input_size, hidden)
-        
-        
+
         if context_hidden is not None:
             self.lin_c = nn.Linear(context_hidden, hidden, bias=False)
         else:
@@ -376,9 +382,8 @@ def __init__(self,
         self.out_proj = nn.Linear(input_size, output_size) if output_size else None
 
     def forward(self, a: Tensor, c: Optional[Tensor] = None):
-        
         x = self.lin_a(a)
-        
+
         if c is not None:
             x = x + self.lin_c(c).unsqueeze(1)
         x = F.elu(x)
@@ -387,18 +392,18 @@ def forward(self, a: Tensor, c: Optional[Tensor] = None):
         x = self.glu(x)
         y = a if self.out_proj is None else self.out_proj(a)
         x = x + y
-        
-        return self.layer_norm(x) 
+
+        return self.layer_norm(x)
 
 
 # @torch.jit.script #Currently broken with autocast
 def fused_pointwise_linear_v1(x, a, b):
-    
     out = torch.mul(x.unsqueeze(-1), a)
     out = out + b
     return out
 
-#@torch.jit.script
+
+# @torch.jit.script
 def fused_pointwise_linear_v2(x, a, b):
     out = x.unsqueeze(3) * a
     out = out + b
@@ -406,21 +411,29 @@ def fused_pointwise_linear_v2(x, a, b):
 
 
 class TFTEmbedding(nn.Module):
-    def __init__(self,static_categorical_inp_size,temporal_known_categorical_inp_size,
-    temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
-    temporal_observed_continuous_inp_size,temporal_target_size,hidden,
-     initialize_cont_params=False):
+    def __init__(
+        self,
+        static_categorical_inp_size,
+        temporal_known_categorical_inp_size,
+        temporal_observed_categorical_inp_size,
+        static_continuous_inp_size,
+        temporal_known_continuous_inp_size,
+        temporal_observed_continuous_inp_size,
+        temporal_target_size,
+        hidden,
+        initialize_cont_params=False,
+    ):
         # initialize_cont_params=False prevents form initializing parameters inside this class
         # so they can be lazily initialized in LazyEmbedding module
         super().__init__()
-        #these are basically number of varaibales that falls under each category 
-        self.s_cat_inp_size    = static_categorical_inp_size
-        self.t_cat_k_inp_size  = temporal_known_categorical_inp_size
-        self.t_cat_o_inp_size  = temporal_observed_categorical_inp_size
-        self.s_cont_inp_size   = static_continuous_inp_size
+        # these are basically number of varaibales that falls under each category
+        self.s_cat_inp_size = static_categorical_inp_size
+        self.t_cat_k_inp_size = temporal_known_categorical_inp_size
+        self.t_cat_o_inp_size = temporal_observed_categorical_inp_size
+        self.s_cont_inp_size = static_continuous_inp_size
         self.t_cont_k_inp_size = temporal_known_continuous_inp_size
         self.t_cont_o_inp_size = temporal_observed_continuous_inp_size
-        self.t_tgt_size        = temporal_target_size
+        self.t_tgt_size = temporal_target_size
 
         self.hidden = hidden
 
@@ -433,30 +446,44 @@ def __init__(self,static_categorical_inp_size,temporal_known_categorical_inp_siz
         # 6. Temporal observed continuous
         # 7. Temporal observed targets (time series obseved so far)
 
-        self.s_cat_embed = nn.ModuleList([
-            nn.Embedding(n, self.hidden) for n in self.s_cat_inp_size]) if self.s_cat_inp_size else None
-        self.t_cat_k_embed = nn.ModuleList([
-            nn.Embedding(n, self.hidden) for n in self.t_cat_k_inp_size]) if self.t_cat_k_inp_size else None
-        self.t_cat_o_embed = nn.ModuleList([
-            nn.Embedding(n, self.hidden) for n in self.t_cat_o_inp_size]) if self.t_cat_o_inp_size else None
-        
+        self.s_cat_embed = (
+            nn.ModuleList([nn.Embedding(n, self.hidden) for n in self.s_cat_inp_size]) if self.s_cat_inp_size else None
+        )
+        self.t_cat_k_embed = (
+            nn.ModuleList([nn.Embedding(n, self.hidden) for n in self.t_cat_k_inp_size]) if self.t_cat_k_inp_size else None
+        )
+        self.t_cat_o_embed = (
+            nn.ModuleList([nn.Embedding(n, self.hidden) for n in self.t_cat_o_inp_size]) if self.t_cat_o_inp_size else None
+        )
 
         if initialize_cont_params:
-            self.s_cont_embedding_vectors = nn.Parameter(torch.Tensor(self.s_cont_inp_size, self.hidden)) if self.s_cont_inp_size else None
-            self.t_cont_k_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_k_inp_size, self.hidden)) if self.t_cont_k_inp_size else None
-            self.t_cont_o_embedding_vectors = nn.Parameter(torch.Tensor(self.t_cont_o_inp_size, self.hidden)) if self.t_cont_o_inp_size else None
+            self.s_cont_embedding_vectors = (
+                nn.Parameter(torch.Tensor(self.s_cont_inp_size, self.hidden)) if self.s_cont_inp_size else None
+            )
+            self.t_cont_k_embedding_vectors = (
+                nn.Parameter(torch.Tensor(self.t_cont_k_inp_size, self.hidden)) if self.t_cont_k_inp_size else None
+            )
+            self.t_cont_o_embedding_vectors = (
+                nn.Parameter(torch.Tensor(self.t_cont_o_inp_size, self.hidden)) if self.t_cont_o_inp_size else None
+            )
             self.t_tgt_embedding_vectors = nn.Parameter(torch.Tensor(self.t_tgt_size, self.hidden))
-            self.s_cont_embedding_bias = nn.Parameter(torch.zeros(self.s_cont_inp_size, self.hidden)) if self.s_cont_inp_size else None
-            self.t_cont_k_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_k_inp_size, self.hidden)) if self.t_cont_k_inp_size else None
-            self.t_cont_o_embedding_bias = nn.Parameter(torch.zeros(self.t_cont_o_inp_size, self.hidden)) if self.t_cont_o_inp_size else None
+            self.s_cont_embedding_bias = (
+                nn.Parameter(torch.zeros(self.s_cont_inp_size, self.hidden)) if self.s_cont_inp_size else None
+            )
+            self.t_cont_k_embedding_bias = (
+                nn.Parameter(torch.zeros(self.t_cont_k_inp_size, self.hidden)) if self.t_cont_k_inp_size else None
+            )
+            self.t_cont_o_embedding_bias = (
+                nn.Parameter(torch.zeros(self.t_cont_o_inp_size, self.hidden)) if self.t_cont_o_inp_size else None
+            )
             self.t_tgt_embedding_bias = nn.Parameter(torch.zeros(self.t_tgt_size, self.hidden))
 
             self.reset_parameters()
 
     def reset_parameters(self):
-        ''''
+        """'
         embeddings are initilitized using xavier's method and biases are initlitized with zeros
-        '''
+        """
         if self.s_cont_embedding_vectors is not None:
             torch.nn.init.xavier_normal_(self.s_cont_embedding_vectors)
             torch.nn.init.zeros_(self.s_cont_embedding_bias)
@@ -479,28 +506,30 @@ def reset_parameters(self):
             for module in self.t_cat_o_embed:
                 module.reset_parameters()
 
-
-    def _apply_embedding(self,
-            cat: Optional[Tensor],
-            cont: Optional[Tensor],
-            cat_emb: Optional[nn.ModuleList], 
-            cont_emb: Tensor,
-            cont_bias: Tensor,
-            ) -> Tuple[Optional[Tensor], Optional[Tensor]]:
-        
-        e_cat = torch.stack([embed(cat[...,i].int()) for i, embed in enumerate(cat_emb)], dim=-2) if (cat is not None) and (cat.size()[1]>0) else None 
-        if (cont is not None ) and (cont.size()[1]>0):
-            #the line below is equivalent to following einsums
-            #e_cont = torch.einsum('btf,fh->bthf', cont, cont_emb)
-            #e_cont = torch.einsum('bf,fh->bhf', cont, cont_emb)
+    def _apply_embedding(
+        self,
+        cat: Optional[Tensor],
+        cont: Optional[Tensor],
+        cat_emb: Optional[nn.ModuleList],
+        cont_emb: Tensor,
+        cont_bias: Tensor,
+    ) -> Tuple[Optional[Tensor], Optional[Tensor]]:
+        e_cat = (
+            torch.stack([embed(cat[..., i].int()) for i, embed in enumerate(cat_emb)], dim=-2)
+            if (cat is not None) and (cat.size()[1] > 0)
+            else None
+        )
+        if (cont is not None) and (cont.size()[1] > 0):
+            # the line below is equivalent to following einsums
+            # e_cont = torch.einsum('btf,fh->bthf', cont, cont_emb)
+            # e_cont = torch.einsum('bf,fh->bhf', cont, cont_emb)
             e_cont = torch.mul(cont.unsqueeze(-1), cont_emb)
             e_cont = e_cont + cont_bias
-           # e_cont = fused_pointwise_linear_v1(cont, cont_emb, cont_bias)
+        # e_cont = fused_pointwise_linear_v1(cont, cont_emb, cont_bias)
         else:
             e_cont = None
-        
+
         if e_cat is not None and e_cont is not None:
-           
             return torch.cat([e_cat, e_cont], dim=-2)
         elif e_cat is not None:
             return e_cat
@@ -510,58 +539,66 @@ def _apply_embedding(self,
             return None
 
     def forward(self, x: Dict[str, Tensor]):
-        # temporal/static categorical/continuous known/observed input 
-        s_cat_inp = x.get('s_cat', None)
-        s_cont_inp = x.get('s_cont', None)
-        t_cat_k_inp = x.get('k_cat', None)
-        t_cont_k_inp = x.get('k_cont', None)
-        t_cat_o_inp = x.get('o_cat', None)
-        t_cont_o_inp = x.get('o_cont', None)
-        t_tgt_obs = x['target'] # Has to be present
+        # temporal/static categorical/continuous known/observed input
+        s_cat_inp = x.get("s_cat", None)
+        s_cont_inp = x.get("s_cont", None)
+        t_cat_k_inp = x.get("k_cat", None)
+        t_cont_k_inp = x.get("k_cont", None)
+        t_cat_o_inp = x.get("o_cat", None)
+        t_cont_o_inp = x.get("o_cont", None)
+        t_tgt_obs = x["target"]  # Has to be present
         # Static inputs are expected to be equal for all timesteps
         # For memory efficiency there is no assert statement
-        
-        
-        
-        s_cat_inp = s_cat_inp[:,0,:] if s_cat_inp is not None else None
-        s_cont_inp = s_cont_inp[:,0,:] if s_cont_inp is not None else None
-       
-        
-        
-        s_inp = self._apply_embedding(s_cat_inp,
-                                      s_cont_inp,
-                                      self.s_cat_embed,
-                                      self.s_cont_embedding_vectors,
-                                      self.s_cont_embedding_bias)
-                                      
-        
-        t_known_inp = self._apply_embedding(t_cat_k_inp,
-                                            t_cont_k_inp,
-                                            self.t_cat_k_embed,
-                                            self.t_cont_k_embedding_vectors,
-                                            self.t_cont_k_embedding_bias)
-        
-        t_observed_inp = self._apply_embedding(t_cat_o_inp,
-                                               t_cont_o_inp,
-                                               self.t_cat_o_embed,
-                                               self.t_cont_o_embedding_vectors,
-                                               self.t_cont_o_embedding_bias)
+
+        s_cat_inp = s_cat_inp[:, 0, :] if s_cat_inp is not None else None
+        s_cont_inp = s_cont_inp[:, 0, :] if s_cont_inp is not None else None
+
+        s_inp = self._apply_embedding(
+            s_cat_inp, s_cont_inp, self.s_cat_embed, self.s_cont_embedding_vectors, self.s_cont_embedding_bias
+        )
+
+        t_known_inp = self._apply_embedding(
+            t_cat_k_inp, t_cont_k_inp, self.t_cat_k_embed, self.t_cont_k_embedding_vectors, self.t_cont_k_embedding_bias
+        )
+
+        t_observed_inp = self._apply_embedding(
+            t_cat_o_inp, t_cont_o_inp, self.t_cat_o_embed, self.t_cont_o_embedding_vectors, self.t_cont_o_embedding_bias
+        )
 
         # Temporal observed targets
-        t_observed_tgt = torch.matmul(t_tgt_obs.unsqueeze(3).unsqueeze(4), self.t_tgt_embedding_vectors.unsqueeze(1)).squeeze(3)
+        t_observed_tgt = torch.matmul(t_tgt_obs.unsqueeze(3).unsqueeze(4), self.t_tgt_embedding_vectors.unsqueeze(1)).squeeze(
+            3
+        )
         t_observed_tgt = t_observed_tgt + self.t_tgt_embedding_bias
 
         return s_inp, t_known_inp, t_observed_inp, t_observed_tgt
 
+
 class LazyEmbedding(nn.modules.lazy.LazyModuleMixin, TFTEmbedding):
     cls_to_become = TFTEmbedding
 
-    def __init__(self, static_categorical_inp_size,temporal_known_categorical_inp_size,
-    temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
-    temporal_observed_continuous_inp_size,temporal_target_size,hidden):
-        super().__init__(static_categorical_inp_size,temporal_known_categorical_inp_size,
-    temporal_observed_categorical_inp_size,static_continuous_inp_size,temporal_known_continuous_inp_size,
-    temporal_observed_continuous_inp_size,temporal_target_size,hidden, initialize_cont_params=False)
+    def __init__(
+        self,
+        static_categorical_inp_size,
+        temporal_known_categorical_inp_size,
+        temporal_observed_categorical_inp_size,
+        static_continuous_inp_size,
+        temporal_known_continuous_inp_size,
+        temporal_observed_continuous_inp_size,
+        temporal_target_size,
+        hidden,
+    ):
+        super().__init__(
+            static_categorical_inp_size,
+            temporal_known_categorical_inp_size,
+            temporal_observed_categorical_inp_size,
+            static_continuous_inp_size,
+            temporal_known_continuous_inp_size,
+            temporal_observed_continuous_inp_size,
+            temporal_target_size,
+            hidden,
+            initialize_cont_params=False,
+        )
         if static_continuous_inp_size:
             self.s_cont_embedding_vectors = UninitializedParameter()
             self.s_cont_embedding_bias = UninitializedParameter()
@@ -585,21 +622,20 @@ def __init__(self, static_categorical_inp_size,temporal_known_categorical_inp_si
         self.t_tgt_embedding_bias = UninitializedParameter()
 
     def initialize_parameters(self, x):
-        
         if self.has_uninitialized_params():
-            s_cont_inp = x.get('s_cont', None)
-            t_cont_k_inp = x.get('k_cont', None)
-            t_cont_o_inp = x.get('o_cont', None)
-            t_tgt_obs = x['target'] # Has to be present
-            if (s_cont_inp is not None) and (s_cont_inp.size()[1]>0):
+            s_cont_inp = x.get("s_cont", None)
+            t_cont_k_inp = x.get("k_cont", None)
+            t_cont_o_inp = x.get("o_cont", None)
+            t_tgt_obs = x["target"]  # Has to be present
+            if (s_cont_inp is not None) and (s_cont_inp.size()[1] > 0):
                 self.s_cont_embedding_vectors.materialize((s_cont_inp.shape[-1], self.hidden))
                 self.s_cont_embedding_bias.materialize((s_cont_inp.shape[-1], self.hidden))
 
-            if (t_cont_k_inp is not None)  and (t_cont_k_inp.size()[1]>0):
+            if (t_cont_k_inp is not None) and (t_cont_k_inp.size()[1] > 0):
                 self.t_cont_k_embedding_vectors.materialize((t_cont_k_inp.shape[-1], self.hidden))
                 self.t_cont_k_embedding_bias.materialize((t_cont_k_inp.shape[-1], self.hidden))
 
-            if (t_cont_o_inp) is not None  and (t_cont_o_inp.size()[1]>0):
+            if (t_cont_o_inp) is not None and (t_cont_o_inp.size()[1] > 0):
                 self.t_cont_o_embedding_vectors.materialize((t_cont_o_inp.shape[-1], self.hidden))
                 self.t_cont_o_embedding_bias.materialize((t_cont_o_inp.shape[-1], self.hidden))
 
@@ -608,47 +644,50 @@ def initialize_parameters(self, x):
 
             self.reset_parameters()
 
+
 class VariableSelectionNetwork(nn.Module):
-    '''
-    Learns to select important netowrks consists of GRNs with one GRN for variable weights 
+    """
+    Learns to select important netowrks consists of GRNs with one GRN for variable weights
     and the others for input embedding
-    '''
-    def __init__(self, hidden,dropout, num_inputs):
+    """
+
+    def __init__(self, hidden, dropout, num_inputs):
         super().__init__()
-        self.hidden=hidden
-        self.joint_grn = GRN(hidden*num_inputs, hidden, output_size=num_inputs, context_hidden=hidden)
+        self.hidden = hidden
+        self.joint_grn = GRN(hidden * num_inputs, hidden, output_size=num_inputs, context_hidden=hidden)
         self.var_grns = nn.ModuleList([GRN(hidden, hidden, dropout=dropout) for _ in range(num_inputs)])
 
     def forward(self, x: Tensor, context: Optional[Tensor] = None):
         if x.numel() == 0:  # Check if x is an empty tensor
             batch_size = context.size(0) if context is not None else 1
-            variable_ctx = torch.zeros(batch_size,1, self.hidden,device=x.device)
-            sparse_weights = torch.ones(batch_size,1, self.hidden,device=x.device)
+            variable_ctx = torch.zeros(batch_size, 1, self.hidden, device=x.device)
+            sparse_weights = torch.ones(batch_size, 1, self.hidden, device=x.device)
             return variable_ctx, sparse_weights
 
         Xi = torch.flatten(x, start_dim=-2)
         grn_outputs = self.joint_grn(Xi, c=context)
         sparse_weights = F.softmax(grn_outputs, dim=-1)
-        transformed_embed_list = [m(x[...,i,:]) for i, m in enumerate(self.var_grns)]
+        transformed_embed_list = [m(x[..., i, :]) for i, m in enumerate(self.var_grns)]
         transformed_embed = torch.stack(transformed_embed_list, dim=-1)
         variable_ctx = torch.matmul(transformed_embed, sparse_weights.unsqueeze(-1)).squeeze(-1)
-        
+
         return variable_ctx, sparse_weights
 
+
 class StaticCovariateEncoder(nn.Module):
-    '''
-    Network to produce 4 contexts vectors to enrich static variables 
+    """
+    Network to produce 4 contexts vectors to enrich static variables
     Vriable selection Network --> GRNs
-    '''
-    def __init__(self, num_static_vars,hidden,dropout):
+    """
+
+    def __init__(self, num_static_vars, hidden, dropout):
         super().__init__()
-        self.vsn = VariableSelectionNetwork(hidden,dropout, num_static_vars)
+        self.vsn = VariableSelectionNetwork(hidden, dropout, num_static_vars)
         self.context_grns = nn.ModuleList([GRN(hidden, hidden, dropout=dropout) for _ in range(4)])
 
     def forward(self, x: Tensor) -> Tuple[Tensor, Tensor, Tensor, Tensor]:
-        
         variable_ctx, sparse_weights = self.vsn(x)
-        
+
         # Context vectors:
         # variable selection context
         # enrichment context
@@ -660,11 +699,12 @@ def forward(self, x: Tensor) -> Tuple[Tensor, Tensor, Tensor, Tensor]:
 
 
 class InterpretableMultiHeadAttention(nn.Module):
-    '''
-    Multi-head attention different as it outputs the attention_probability and it combines the attention weights instead of 
+    """
+    Multi-head attention different as it outputs the attention_probability and it combines the attention weights instead of
     concating them different from the one implemented already in YAIB
-    '''
-    def __init__(self,n_head,hidden,dropout_att,dropout,example_length):
+    """
+
+    def __init__(self, n_head, hidden, dropout_att, dropout, example_length):
         super().__init__()
         self.n_head = n_head
         assert hidden % n_head == 0
@@ -674,7 +714,7 @@ def __init__(self,n_head,hidden,dropout_att,dropout,example_length):
         self.dropout_att = nn.Dropout(dropout_att)
         self.out_dropout = nn.Dropout(dropout)
         self.scale = self.d_head**-0.5
-        self.register_buffer("_mask", torch.triu(torch.full((example_length, example_length), float('-inf')), 1).unsqueeze(0))
+        self.register_buffer("_mask", torch.triu(torch.full((example_length, example_length), float("-inf")), 1).unsqueeze(0))
 
     def forward(self, x: Tensor) -> Tuple[Tensor, Tensor]:
         bs, t, h_size = x.shape
@@ -700,51 +740,55 @@ def forward(self, x: Tensor) -> Tuple[Tensor, Tensor]:
 
         return out, attn_prob
 
+
 class TFTBack(nn.Module):
-    '''
+    """
     Big part of TFT architecture consists of static enrichment followed by mutli-head self-attention then position wise feed forward
-    followed by a gate and a dense layer 
+    followed by a gate and a dense layer
     GRNs-->multi-head attention-->GRNs-->GLU-->Linear-->output
-    '''
-    def __init__(self, encoder_length,num_historic_vars,hidden,dropout,num_future_vars,
-                 n_head,dropout_att,example_length,quantiles):
+    """
+
+    def __init__(
+        self,
+        encoder_length,
+        num_historic_vars,
+        hidden,
+        dropout,
+        num_future_vars,
+        n_head,
+        dropout_att,
+        example_length,
+        quantiles,
+    ):
         super().__init__()
 
         self.encoder_length = encoder_length
-        self.history_vsn = VariableSelectionNetwork(hidden,dropout, num_historic_vars)
+        self.history_vsn = VariableSelectionNetwork(hidden, dropout, num_historic_vars)
         self.history_encoder = nn.LSTM(hidden, hidden, batch_first=True)
-        self.future_vsn = VariableSelectionNetwork(hidden,dropout,num_future_vars)
+        self.future_vsn = VariableSelectionNetwork(hidden, dropout, num_future_vars)
         self.future_encoder = nn.LSTM(hidden, hidden, batch_first=True)
 
-
         self.input_gate = GLU(hidden, hidden)
         self.input_gate_ln = LayerNorm(hidden, eps=1e-3)
 
-        self.enrichment_grn = GRN(hidden,
-                                  hidden,
-                                  context_hidden=hidden, 
-                                  dropout=dropout)
-        self.attention = InterpretableMultiHeadAttention(n_head,hidden,dropout_att,dropout,example_length)
+        self.enrichment_grn = GRN(hidden, hidden, context_hidden=hidden, dropout=dropout)
+        self.attention = InterpretableMultiHeadAttention(n_head, hidden, dropout_att, dropout, example_length)
         self.attention_gate = GLU(hidden, hidden)
         self.attention_ln = LayerNorm(hidden, eps=1e-3)
 
-        self.positionwise_grn = GRN(hidden,
-                                    hidden,
-                                    dropout=dropout)
+        self.positionwise_grn = GRN(hidden, hidden, dropout=dropout)
 
         self.decoder_gate = GLU(hidden, hidden)
         self.decoder_ln = LayerNorm(hidden, eps=1e-3)
-        
-        
+
         self.quantile_proj = nn.Linear(hidden, len(quantiles))
-        
+
     def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         historical_features, _ = self.history_vsn(historical_inputs, cs)
         history, state = self.history_encoder(historical_features, (ch, cc))
         future_features, _ = self.future_vsn(future_inputs, cs)
 
         future, _ = self.future_encoder(future_features, state)
-        
 
         # skip connection
         input_embedding = torch.cat([historical_features, future_features], dim=1)
@@ -757,12 +801,12 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         enriched = self.enrichment_grn(temporal_features, c=ce)
 
         # Temporal self attention
-        x, attn_prob= self.attention(enriched)
+        x, attn_prob = self.attention(enriched)
 
         # Don't compute hictorical quantiles
-        x = x[:, self.encoder_length:, :]
-        temporal_features = temporal_features[:, self.encoder_length:, :]
-        enriched = enriched[:, self.encoder_length:, :]
+        x = x[:, self.encoder_length :, :]
+        temporal_features = temporal_features[:, self.encoder_length :, :]
+        enriched = enriched[:, self.encoder_length :, :]
 
         x = self.attention_gate(x)
         x = x + enriched
diff --git a/icu_benchmarks/models/metrics.py b/icu_benchmarks/models/metrics.py
index 108f65da..837bcaae 100644
--- a/icu_benchmarks/models/metrics.py
+++ b/icu_benchmarks/models/metrics.py
@@ -5,27 +5,33 @@
 from sklearn.metrics import balanced_accuracy_score, mean_absolute_error
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
-import gin 
+import gin
 import torch.nn.functional as F
+
 """"
 This file contains metrics that are not available in ignite.metrics. Specifically, it adds transformation capabilities to some
 metrics.
 """
 
+
 class QuantileLossClass(torch.nn.Module):
     def __init__(self, quantiles):
         super().__init__()
-        self.register_buffer('q', torch.tensor(quantiles))
+        self.register_buffer("q", torch.tensor(quantiles))
 
     def forward(self, predictions, targets):
         diff = predictions - targets
-        ql = (1-self.q)*F.relu(diff) + self.q*F.relu(-diff)
+        ql = (1 - self.q) * F.relu(diff) + self.q * F.relu(-diff)
         losses = ql.view(-1, ql.shape[-1]).mean(0)
         return losses
+
+
 @gin.configurable
-def Quantileloss(self,predictions, targets,quantiles=[0.1,0.5,0.9]):
-    loss=QuantileLossClass(quantiles=quantiles)
-    return loss(predictions,targets)
+def Quantileloss(self, predictions, targets, quantiles=[0.1, 0.5, 0.9]):
+    loss = QuantileLossClass(quantiles=quantiles)
+    return loss(predictions, targets)
+
+
 def accuracy(output, target, topk=(1,)):
     """Computes the accuracy over the k top predictions for the specified values of k"""
     with torch.no_grad():
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 868ebe85..9dc6d96d 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -8,14 +8,14 @@
 from pytorch_lightning.loggers import TensorBoardLogger
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint, TQDMProgressBar
-from pytorch_forecasting import TimeSeriesDataSet,TemporalFusionTransformer
 from pathlib import Path
-from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset,PredictionDatasetTFT,PredictionDatasetTFTpytorch
+from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset, PredictionDatasetTFT, PredictionDatasetTFTpytorch
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
-import pickle 
+
+
 cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
 
 
@@ -71,20 +71,22 @@ def train_common(
         verbose: Enable detailed logging.
         ram_cache: Whether to cache the data in RAM.
         num_workers: Number of workers to use for data loading.
-         with open('saved_dictionary.pkl', 'wb') as f:
-        pickle.dump(data, f)
-    data["train"]["OUTCOME"].to_csv('outcome.csv', index=False)
-    data["train"]["FEATURES"].to_csv('features.csv', index=False)
     """
-   
     logging.info(f"Training model: {model.__name__}.")
-    dataset_class = (ImputationDataset if mode == RunMode.imputation
-                 else (PredictionDatasetTFT if model.__name__ == 'TFT'
-                       else (PredictionDatasetTFTpytorch if model.__name__ == 'TFTpytorch' else PredictionDataset)))
+    # choose dataset_class based on the model
+    dataset_class = (
+        ImputationDataset
+        if mode == RunMode.imputation
+        else (
+            PredictionDatasetTFT
+            if model.__name__ == "TFT"
+            else (PredictionDatasetTFTpytorch if model.__name__ == "TFTpytorch" else PredictionDataset)
+        )
+    )
 
     logging.info(f"Logging to directory: {log_dir}.")
     save_config_file(log_dir)  # We save the operative config before and also after training
-    
+
     train_dataset = dataset_class(data, split=Split.train, ram_cache=ram_cache)
     val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache)
     train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
@@ -92,16 +94,23 @@ def train_common(
     test_dataset = dataset_class(data, split=test_on)
     logging.debug(f"Training on {len(train_dataset)} samples and validating on {len(val_dataset)} samples.")
     logging.info(f"Using {num_workers} workers for data loading.")
-    if(model.__name__ =='TFTpytorch'):
-        
-        train_loader = train_dataset.to_dataloader(train=True, batch_size=batch_size, num_workers=0,pin_memory=False,drop_last=True)
-        val_loader = val_dataset.to_dataloader(train=False, batch_size=batch_size ,num_workers=0,pin_memory=False,drop_last=True)
-        test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=0,pin_memory=False,drop_last=True,shuffle=False)
-        model=model(train_dataset,optimizer=optimizer, epochs=epochs, run_mode=mode)
-        
-        
-        
-        
+    if model.__name__ == "TFTpytorch":
+        train_loader = train_dataset.to_dataloader(
+            train=True, batch_size=batch_size, num_workers=0, pin_memory=False, drop_last=True
+        )
+        val_loader = val_dataset.to_dataloader(
+            train=False, batch_size=batch_size, num_workers=0, pin_memory=False, drop_last=True
+        )
+        test_loader = test_dataset.to_dataloader(
+            train=False,
+            batch_size=min(batch_size * 4, len(test_dataset)),
+            num_workers=0,
+            pin_memory=False,
+            drop_last=True,
+            shuffle=False,
+        )
+        model = model(train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode)
+
     else:
         train_loader = DataLoader(
             train_dataset,
@@ -119,14 +128,13 @@ def train_common(
             pin_memory=True,
             drop_last=True,
         )
-    
-    
-        if(isinstance(next(iter(train_loader))[0] ,OrderedDict)):
+
+        if isinstance(next(iter(train_loader))[0], OrderedDict):
             model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
-        
+
         else:
             data_shape = next(iter(train_loader))[0].shape
-        
+
             model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
 
     model.set_weight(weight, train_dataset)
@@ -155,15 +163,14 @@ def train_common(
         max_epochs=epochs if model.needs_training else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator= "auto" if not cpu else "cpu",
+        accelerator="auto" if not cpu else "cpu",
         devices=max(torch.cuda.device_count(), 1),
-        deterministic=reproducible,
+        deterministic=False,
         benchmark=not reproducible,
         enable_progress_bar=verbose,
         logger=loggers,
         num_sanity_val_steps=0,
     )
-    
 
     if model.needs_fit:
         logging.info("Fitting model to data.")
@@ -173,30 +180,23 @@ def train_common(
 
     if model.needs_training:
         logging.info("Training model.")
-        if(type(model).__name__ =='TFTpytorch'):
-            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
-        else:
-            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
+        trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
         logging.info("Training complete.")
     test_dataset = dataset_class(data, split=test_on)
     test_dataset = assure_minimum_length(test_dataset)
-    if(type(model).__name__ =='TFTpytorch'):
-        test_loader = test_dataset.to_dataloader(train=False, batch_size=min(batch_size * 4, len(test_dataset)), num_workers=num_workers,pin_memory=True,drop_last=True,shuffle=False)
-        test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
-    else:
-       
-        test_loader = (
-            DataLoader(
-                test_dataset,
-                batch_size=min(batch_size * 4, len(test_dataset)),
-                shuffle=False,
-                num_workers=num_workers,
-                pin_memory=True,
-                drop_last=True,
-            )
+
+    test_loader = (
+        DataLoader(
+            test_dataset,
+            batch_size=min(batch_size * 4, len(test_dataset)),
+            shuffle=False,
+            num_workers=num_workers,
+            pin_memory=True,
+            drop_last=True,
+        )
         if model.needs_training
         else DataLoader([test_dataset.to_tensor()], batch_size=1)
     )
-        test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
+    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
     return test_loss
diff --git a/icu_benchmarks/models/utils.py b/icu_benchmarks/models/utils.py
index ddc97028..53bfcae2 100644
--- a/icu_benchmarks/models/utils.py
+++ b/icu_benchmarks/models/utils.py
@@ -188,8 +188,3 @@ def version(self):
     @rank_zero_only
     def log_hyperparams(self, params):
         pass
-@gin.configurable
-def heads_divisible_by_hidden(hidden):
-    heads_range = range(1, 9)  # Range for heads from 1 to 8
-    divisible_heads = (heads for heads in heads_range if hidden % heads == 0)
-    return divisible_heads
\ No newline at end of file
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 267da7fa..901e2c8a 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -20,6 +20,7 @@
 from pytorch_lightning import LightningModule
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
+
 gin.config.external_configurable(torch.nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(torch.nn.functional.cross_entropy, module="torch.nn.functional")
 gin.config.external_configurable(torch.nn.functional.mse_loss, module="torch.nn.functional")
@@ -110,7 +111,6 @@ def __init__(
         initialization_method: str = "normal",
         **kwargs,
     ):
-        
         """Interface for Deep Learning models."""
         super().__init__()
         self.save_hyperparameters(ignore=["loss", "optimizer"])
@@ -134,7 +134,9 @@ def finalize_step(self, step_prefix=""):
         try:
             self.log_dict(
                 {
-                    f"{step_prefix}/{name}": (np.float32(metric.compute()) if isinstance(metric.compute(), np.float64 ) else metric.compute() )
+                    f"{step_prefix}/{name}": (
+                        np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
+                    )
                     for name, metric in self.metrics[step_prefix].items()
                     if "_Curve" not in name
                 },
@@ -226,30 +228,28 @@ def softmax_multi_output_transform(output):
 
     def step_fn(self, element, step_prefix=""):
         """Perform a step in the training loop."""
-        if(isinstance(element[0] ,OrderedDict)):
-            data,mask=element[0], element[1].to(self.device)
+        if isinstance(element[0], OrderedDict):  # check if the data loader is the one for the TFT nvidia implementation
+            data, mask = element[0], element[1].to(self.device)
 
             for key, value in data.items():
                 value = value.float().to(self.device)
 
-                
-
                 if value.shape[-1] == 1:
                     value = value.squeeze(-1)
                 if value.dim() == 3:
                     value = value.permute(0, 2, 1)
-                if(key=='target'):
-                    labels=value.squeeze()
+                if key == "target":
+                    labels = value.squeeze()
                 data[key] = value
-        
-            
+
         elif len(element) == 2:
-           
-            if(isinstance(element[1] ,tuple)):
+            if (isinstance(element[1], tuple)) or (
+                isinstance(element[1], list)
+            ):  # check if the data loader is the one for the pytorch forecasring implementation
                 data, labels = element[0], element[1][0]
                 mask = torch.ones_like(labels).bool()
             else:
-                data, labels = element[0], element[1].to(self.device)
+                data, labels = element[0], (element[1]).to(self.device)
                 if isinstance(data, list):
                     for i in range(len(data)):
                         data[i] = data[i].float().to(self.device)
@@ -266,13 +266,14 @@ def step_fn(self, element, step_prefix=""):
                 data = data.float().to(self.device)
         else:
             raise Exception("Loader should return either (data, label) or (data, label, mask)")
+
         out = self(data)
         if len(out) == 2 and isinstance(out, tuple):
             out, aux_loss = out
         else:
             aux_loss = 0
         prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
-        
+
         target = torch.masked_select(labels, mask).to(self.device)
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
@@ -280,7 +281,7 @@ def step_fn(self, element, step_prefix=""):
             # torch.long because NLL
         elif self.run_mode == RunMode.regression:
             # Regression task
-            
+
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
             raise ValueError(f"Run mode {self.run_mode} not supported.")

From 9ef8252d57f0ba400072792ccd9c8f4fb7aaa7d1 Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Thu, 7 Sep 2023 16:04:18 +0200
Subject: [PATCH 042/142] latest

---
 configs/prediction_models/TFT.gin             |  4 +--
 configs/prediction_models/TFTpytorch.gin      |  6 ++--
 configs/prediction_models/common/DLCommon.gin |  2 +-
 configs/tasks/Regression.gin                  |  2 +-
 experiments/benchmark_classification.yml      | 27 +++++++-------
 experiments/benchmark_regression.yml          | 15 ++++----
 icu_benchmarks/cross_validation.py            |  2 +-
 icu_benchmarks/data/loader.py                 |  7 ++--
 icu_benchmarks/models/dl_models.py            |  4 +--
 icu_benchmarks/models/layers.py               |  6 ++--
 icu_benchmarks/models/train.py                | 36 ++++++++++---------
 11 files changed, 58 insertions(+), 53 deletions(-)

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
index 3abcb239..09fcf537 100644
--- a/configs/prediction_models/TFT.gin
+++ b/configs/prediction_models/TFT.gin
@@ -13,11 +13,11 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 # Encoder params
 model/hyperparameter.class_to_tune = @TFT
 model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = 256
+model/hyperparameter.hidden = (32,64,128,256)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =4
+model/hyperparameter.n_heads =(1,2,4)
 model/hyperparameter.example_length=25
 
 
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 8c320809..b10fd4b5 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -12,13 +12,13 @@ optimizer/hyperparameter.weight_decay = 1e-6
 
 # Encoder params
 model/hyperparameter.class_to_tune = @TFTpytorch
-model/hyperparameter.hidden = 128
+model/hyperparameter.hidden = (16,32,64)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =4
+model/hyperparameter.n_heads =(1,2,4)
 model/hyperparameter.lr = (1e-5, 3e-4)
 PredictionDatasetTFTpytorch.max_encoder_length = 24
-PredictionDatasetTFTpytorch.max_prediction_length = 12
+PredictionDatasetTFTpytorch.max_prediction_length = 24
 
 
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 99251ea2..f41f2104 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -14,7 +14,7 @@ base_regression_preprocessor.generate_features = False
 
 # Train params
 train_common.optimizer = @Adam
-train_common.epochs = 1000
+train_common.epochs = 50
 train_common.batch_size = 64
 train_common.patience = 10
 train_common.min_delta = 1e-4
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index ed7f794a..5cf3f8d9 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -8,7 +8,7 @@ include "configs/tasks/common/PredictionTaskVariables.gin"
 include "configs/tasks/common/CrossValidation.gin"
 
 # MODE SETTINGS
-Run.mode = "RegressionTFT"
+Run.mode = "Regression"
 NUM_CLASSES = 1
 HORIZON = 24
 train_common.weight = "balanced"
diff --git a/experiments/benchmark_classification.yml b/experiments/benchmark_classification.yml
index e9a32b4c..bc1c55c6 100644
--- a/experiments/benchmark_classification.yml
+++ b/experiments/benchmark_classification.yml
@@ -17,18 +17,18 @@ name: yaib_classification_benchmark
 parameters:
   data_dir:
     values:
-      - data/YAIB_Preprint/aki/miiv
-      #- data/YAIB_Preprint/aki/hirid
-      #- data/YAIB_Preprint/aki/eicu
-      #- data/YAIB_Preprint/aki/aumc
-      - data/YAIB_Preprint/morality24/miiv
-      - data/YAIB_Preprint/morality24/miiv
-      - data/YAIB_Preprint/morality24/miiv
-      - data/YAIB_Preprint/morality24/miiv
-      #- data/YAIB_Preprint/sepsis/miiv
-      #- data/YAIB_Preprint/sepsis/hirid
-      #- data/YAIB_Preprint/sepsis/eicu
-      #- data/YAIB_Preprint/sepsis/aumc
+    #  - data/YAIB_Preprint/aki/miiv
+    #  - data/YAIB_Preprint/aki/hirid
+    #  - data/YAIB_Preprint/aki/eicu
+    #  - data/YAIB_Preprint/aki/aumc
+    #  - data/YAIB_Preprint/mortality24/miiv
+    #  - data/YAIB_Preprint/mortality24/hirid
+    #  - data/YAIB_Preprint/mortality24/eicu
+    #  - data/YAIB_Preprint/mortality24/aumc
+      - data/YAIB_Preprint/sepsis/miiv
+      - data/YAIB_Preprint/sepsis/hirid
+      - data/YAIB_Preprint/sepsis/eicu
+      - data/YAIB_Preprint/sepsis/aumc
   model:
     values:
     #  - LogisticRegression
@@ -37,7 +37,8 @@ parameters:
     #  - LSTM
     #  - TCN
     #  - Transformer
-      - TFT
+    #  - TFT
+       - TFTpytorch
       
   seed:
     values:
diff --git a/experiments/benchmark_regression.yml b/experiments/benchmark_regression.yml
index 516a4b0f..c13929eb 100644
--- a/experiments/benchmark_regression.yml
+++ b/experiments/benchmark_regression.yml
@@ -27,13 +27,14 @@ parameters:
       - data/YAIB_Preprint/kf/aumc
   model:
     values:
-      - ElasticNet
-      - LGBMRegressor
-      - GRU
-      - LSTM
-      - TCN
-      - Transformer
-      - TFT
+     # - ElasticNet
+     # - LGBMRegressor
+     # - GRU
+     # - LSTM
+     # - TCN
+     # - Transformer
+     # - TFT
+      - TFTpytorch
       
   seed:
     values:
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 81d8c3ae..0b01992b 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -21,7 +21,7 @@ def execute_repeated_cv(
     seed: int,
     load_weights: bool = False,
     source_dir: Path = None,
-    cv_repetitions: int = 5,
+    cv_repetitions: int = 1,
     cv_repetitions_to_train: int = None,
     cv_folds: int = 5,
     cv_folds_to_train: int = None,
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index c7e8bda2..64eb9a45 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -228,7 +228,8 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
                 [tensors[5], np.ones((np.shape(tensors[5])[0], self.maxlen - np.shape(tensors[5])[1])) * pad_value], axis=1
             )
 
-            tensors[6] = np.concatenate([tensors[6], np.ones(self.maxlen - np.shape(tensors[6])[0]) * pad_value], axis=0)
+            tensors[6] = np.concatenate([tensors[6], np.ones(
+                self.maxlen - np.shape(tensors[6])[0]) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
         tensors[7] = np.concatenate(
             [tensors[7], np.ones((np.shape(tensors[7])[0], self.maxlen - np.shape(tensors[7])[1])) * stay_id], axis=1
@@ -417,9 +418,9 @@ def __init__(
             time_idx="time_idx",
             target="label",
             group_ids=["stay_id"],
-            min_encoder_length=max_encoder_length // 2,
+            min_encoder_length=max_encoder_length,
             max_encoder_length=max_encoder_length,
-            min_prediction_length=2,
+            min_prediction_length=max_prediction_length,
             max_prediction_length=max_prediction_length,
             static_categoricals=[],
             static_reals=["height", "weight", "age", "sex"],
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index c5360da3..cfc538c8 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -371,7 +371,7 @@ def __init__(
         )  # embeddings for all variables
 
         self.static_encoder = StaticCovariateEncoder(num_static_vars, hidden, dropout)  # encoding for static variables
-        self.TFTpart2 = TFTBack(
+        self.TFTpart = TFTBack(
             encoder_length,
             num_historic_vars,
             hidden,
@@ -410,7 +410,7 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         if t_known_inp is not None:
             future_inputs = t_known_inp[:, self.encoder_length :]
 
-        o = self.TFTpart2(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
+        o = self.TFTpart(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
         pred = self.logit(o)
         return pred
 
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 7307bb52..bd844d64 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -743,8 +743,8 @@ def forward(self, x: Tensor) -> Tuple[Tensor, Tensor]:
 
 class TFTBack(nn.Module):
     """
-    Big part of TFT architecture consists of static enrichment followed by mutli-head self-attention then position wise feed forward
-    followed by a gate and a dense layer
+    Big part of TFT architecture consists of static enrichment followed by mutli-head self-attention then 
+    position wise feed forward followed by a gate and a dense layer
     GRNs-->multi-head attention-->GRNs-->GLU-->Linear-->output
     """
 
@@ -803,7 +803,7 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         # Temporal self attention
         x, attn_prob = self.attention(enriched)
 
-        # Don't compute hictorical quantiles
+        # Don't compute historical quantiles
         x = x[:, self.encoder_length :, :]
         temporal_features = temporal_features[:, self.encoder_length :, :]
         enriched = enriched[:, self.encoder_length :, :]
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 9dc6d96d..61ac14af 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -38,7 +38,7 @@ def train_common(
     optimizer: type = Adam,
     precision=32,
     batch_size=64,
-    epochs=1000,
+    epochs=100,
     patience=20,
     min_delta=1e-5,
     test_on: str = Split.test,
@@ -96,10 +96,10 @@ def train_common(
     logging.info(f"Using {num_workers} workers for data loading.")
     if model.__name__ == "TFTpytorch":
         train_loader = train_dataset.to_dataloader(
-            train=True, batch_size=batch_size, num_workers=0, pin_memory=False, drop_last=True
+            train=True, batch_size=batch_size, num_workers=0, pin_memory=False, drop_last=True, batch_sampler="synchronized"
         )
         val_loader = val_dataset.to_dataloader(
-            train=False, batch_size=batch_size, num_workers=0, pin_memory=False, drop_last=True
+            train=False, batch_size=batch_size, num_workers=0, pin_memory=False, drop_last=True, batch_sampler="synchronized"
         )
         test_loader = test_dataset.to_dataloader(
             train=False,
@@ -108,6 +108,7 @@ def train_common(
             pin_memory=False,
             drop_last=True,
             shuffle=False,
+            batch_sampler="synchronized"
         )
         model = model(train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode)
 
@@ -128,6 +129,21 @@ def train_common(
             pin_memory=True,
             drop_last=True,
         )
+        test_dataset = dataset_class(data, split=test_on)
+        test_dataset = assure_minimum_length(test_dataset)
+
+        test_loader = (
+            DataLoader(
+                test_dataset,
+                batch_size=min(batch_size * 4, len(test_dataset)),
+                shuffle=False,
+                num_workers=num_workers,
+                pin_memory=True,
+                drop_last=True,
+            )
+            if model.needs_training
+            else DataLoader([test_dataset.to_tensor()], batch_size=1)
+        )
 
         if isinstance(next(iter(train_loader))[0], OrderedDict):
             model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
@@ -182,21 +198,7 @@ def train_common(
         logging.info("Training model.")
         trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
         logging.info("Training complete.")
-    test_dataset = dataset_class(data, split=test_on)
-    test_dataset = assure_minimum_length(test_dataset)
 
-    test_loader = (
-        DataLoader(
-            test_dataset,
-            batch_size=min(batch_size * 4, len(test_dataset)),
-            shuffle=False,
-            num_workers=num_workers,
-            pin_memory=True,
-            drop_last=True,
-        )
-        if model.needs_training
-        else DataLoader([test_dataset.to_tensor()], batch_size=1)
-    )
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
     return test_loss

From 3c9b729a4f559b2ab584728a12217602aefc8e19 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 12 Sep 2023 15:33:41 +0200
Subject: [PATCH 043/142] reproducible forced to be there for evaulate

---
 icu_benchmarks/run.py | 52 ++++++++++++++++++++++++++++++++++---------
 1 file changed, 41 insertions(+), 11 deletions(-)

diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 0f8b7859..f91f69b8 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -9,7 +9,11 @@
 
 import torch.cuda
 
-from icu_benchmarks.wandb_utils import update_wandb_config, apply_wandb_sweep, set_wandb_run_name
+from icu_benchmarks.wandb_utils import (
+    update_wandb_config,
+    apply_wandb_sweep,
+    set_wandb_run_name,
+)
 from icu_benchmarks.tuning.hyperparameters import choose_and_bind_hyperparameters
 from scripts.plotting.utils import plot_aggregated_results
 from icu_benchmarks.cross_validation import execute_repeated_cv
@@ -45,6 +49,7 @@ def main(my_args=tuple(sys.argv[1:])):
     setup_logging(date_format, log_format, verbose)
 
     # Load weights if in evaluation mode
+
     load_weights = args.command == "evaluate"
     data_dir = Path(args.data_dir)
 
@@ -52,7 +57,8 @@ def main(my_args=tuple(sys.argv[1:])):
     name = args.name
     task = args.task
     model = args.model
-    reproducible = args.reproducible
+
+    reproducible = args.reproducible if args.command != "evaluate" else False
 
     # Set experiment name
     if name is None:
@@ -71,7 +77,9 @@ def main(my_args=tuple(sys.argv[1:])):
     logging.info(f"Task mode: {mode}.")
     experiment = args.experiment
 
-    pretrained_imputation_model = load_pretrained_imputation_model(args.pretrained_imputation)
+    pretrained_imputation_model = load_pretrained_imputation_model(
+        args.pretrained_imputation
+    )
 
     # Log imputation model to wandb
     update_wandb_config(
@@ -86,29 +94,43 @@ def main(my_args=tuple(sys.argv[1:])):
     log_dir = (
         (log_dir_name / experiment)
         if experiment
-        else (log_dir_name / (args.task_name if args.task_name is not None else args.task) / model)
+        else (
+            log_dir_name
+            / (args.task_name if args.task_name is not None else args.task)
+            / model
+        )
     )
     if torch.cuda.is_available():
         for name in range(0, torch.cuda.device_count()):
-            log_full_line(f"Available GPU {name}: {torch.cuda.get_device_name(name)}", level=logging.INFO)
+            log_full_line(
+                f"Available GPU {name}: {torch.cuda.get_device_name(name)}",
+                level=logging.INFO,
+            )
     else:
         log_full_line(
-            "No GPUs available: please check your device and Torch,Cuda installation if unintended.", level=logging.WARNING
+            "No GPUs available: please check your device and Torch,Cuda installation if unintended.",
+            level=logging.WARNING,
         )
 
     log_full_line(f"Logging to {log_dir}.", logging.INFO)
 
     if args.preprocessor:
         # Import custom supplied preprocessor
-        log_full_line(f"Importing custom preprocessor from {args.preprocessor}.", logging.INFO)
+        log_full_line(
+            f"Importing custom preprocessor from {args.preprocessor}.", logging.INFO
+        )
         try:
-            spec = importlib.util.spec_from_file_location("CustomPreprocessor", args.preprocessor)
+            spec = importlib.util.spec_from_file_location(
+                "CustomPreprocessor", args.preprocessor
+            )
             module = importlib.util.module_from_spec(spec)
             sys.modules["preprocessor"] = module
             spec.loader.exec_module(module)
             gin.bind_parameter("preprocess.preprocessor", module.CustomPreprocessor)
         except Exception as e:
-            logging.error(f"Could not import custom preprocessor from {args.preprocessor}: {e}")
+            logging.error(
+                f"Could not import custom preprocessor from {args.preprocessor}: {e}"
+            )
 
     if load_weights:
         # Evaluate
@@ -120,14 +142,22 @@ def main(my_args=tuple(sys.argv[1:])):
         # Train
         checkpoint = log_dir / args.checkpoint if args.checkpoint else None
         model_path = (
-            Path("configs") / ("imputation_models" if mode == RunMode.imputation else "prediction_models") / f"{model}.gin"
+            Path("configs")
+            / (
+                "imputation_models"
+                if mode == RunMode.imputation
+                else "prediction_models"
+            )
+            / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
             if args.experiment
             else [model_path, Path(f"configs/tasks/{task}.gin")]
         )
-        gin.parse_config_files_and_bindings(gin_config_files, args.hyperparams, finalize_config=False)
+        gin.parse_config_files_and_bindings(
+            gin_config_files, args.hyperparams, finalize_config=False
+        )
         log_full_line(f"Data directory: {data_dir.resolve()}", level=logging.INFO)
         run_dir = create_run_dir(log_dir)
         choose_and_bind_hyperparameters(

From 03ad5c2e550719c9775ed6f8f5c6c18e72f99129 Mon Sep 17 00:00:00 2001
From: youssef mecky <youssefmecky@youssefs-MacBook-Air.local>
Date: Tue, 12 Sep 2023 15:35:34 +0200
Subject: [PATCH 044/142] added v1 to load model

---
 icu_benchmarks/models/train.py | 65 ++++++++++++++++++++++++++--------
 1 file changed, 51 insertions(+), 14 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 61ac14af..8de301c5 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -9,14 +9,21 @@
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint, TQDMProgressBar
 from pathlib import Path
-from icu_benchmarks.data.loader import PredictionDataset, ImputationDataset, PredictionDatasetTFT, PredictionDatasetTFTpytorch
+from icu_benchmarks.data.loader import (
+    PredictionDataset,
+    ImputationDataset,
+    PredictionDatasetTFT,
+    PredictionDatasetTFTpytorch,
+)
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
 
 
-cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
+cpu_core_count = (
+    len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
+)
 
 
 def assure_minimum_length(dataset):
@@ -46,7 +53,11 @@ def train_common(
     cpu: bool = False,
     verbose=True,
     ram_cache=False,
-    num_workers: int = min(cpu_core_count, torch.cuda.device_count() * 4 * int(torch.cuda.is_available()), 32),
+    num_workers: int = min(
+        cpu_core_count,
+        torch.cuda.device_count() * 4 * int(torch.cuda.is_available()),
+        32,
+    ),
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -80,26 +91,46 @@ def train_common(
         else (
             PredictionDatasetTFT
             if model.__name__ == "TFT"
-            else (PredictionDatasetTFTpytorch if model.__name__ == "TFTpytorch" else PredictionDataset)
+            else (
+                PredictionDatasetTFTpytorch
+                if model.__name__ == "TFTpytorch"
+                else PredictionDataset
+            )
         )
     )
 
     logging.info(f"Logging to directory: {log_dir}.")
-    save_config_file(log_dir)  # We save the operative config before and also after training
+    save_config_file(
+        log_dir
+    )  # We save the operative config before and also after training
 
     train_dataset = dataset_class(data, split=Split.train, ram_cache=ram_cache)
     val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache)
-    train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
+    train_dataset, val_dataset = assure_minimum_length(
+        train_dataset
+    ), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
     test_dataset = dataset_class(data, split=test_on)
-    logging.debug(f"Training on {len(train_dataset)} samples and validating on {len(val_dataset)} samples.")
+    logging.debug(
+        f"Training on {len(train_dataset)} samples and validating on {len(val_dataset)} samples."
+    )
     logging.info(f"Using {num_workers} workers for data loading.")
     if model.__name__ == "TFTpytorch":
         train_loader = train_dataset.to_dataloader(
-            train=True, batch_size=batch_size, num_workers=0, pin_memory=False, drop_last=True, batch_sampler="synchronized"
+            train=True,
+            batch_size=batch_size,
+            num_workers=0,
+            pin_memory=False,
+            drop_last=True,
+            batch_sampler="synchronized",
         )
         val_loader = val_dataset.to_dataloader(
-            train=False, batch_size=batch_size, num_workers=0, pin_memory=False, drop_last=True, batch_sampler="synchronized"
+            train=False,
+            batch_size=batch_size,
+            num_workers=0,
+            pin_memory=False,
+            drop_last=True,
+            batch_sampler="synchronized",
         )
         test_loader = test_dataset.to_dataloader(
             train=False,
@@ -108,7 +139,7 @@ def train_common(
             pin_memory=False,
             drop_last=True,
             shuffle=False,
-            batch_sampler="synchronized"
+            batch_sampler="synchronized",
         )
         model = model(train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode)
 
@@ -151,13 +182,15 @@ def train_common(
         else:
             data_shape = next(iter(train_loader))[0].shape
 
-            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
+            model = model(
+                optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
+            )
 
     model.set_weight(weight, train_dataset)
     if load_weights:
         if source_dir.exists():
             # if not model.needs_training:
-            checkpoint = torch.load(source_dir / "model.ckpt")
+            checkpoint = torch.load(source_dir / "model-v1.ckpt")
             # else:
             model = model.load_state_dict(checkpoint["state_dict"])
         else:
@@ -168,7 +201,9 @@ def train_common(
     loggers = [TensorBoardLogger(log_dir), JSONMetricsLogger(log_dir)]
 
     callbacks = [
-        EarlyStopping(monitor="val/loss", min_delta=min_delta, patience=patience, strict=False),
+        EarlyStopping(
+            monitor="val/loss", min_delta=min_delta, patience=patience, strict=False
+        ),
         ModelCheckpoint(log_dir, filename="model", save_top_k=1, save_last=True),
     ]
     if verbose:
@@ -199,6 +234,8 @@ def train_common(
         trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
         logging.info("Training complete.")
 
-    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
+    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
+        "test/loss"
+    ]
     save_config_file(log_dir)
     return test_loss

From be61df80806026236a86363ad6a68a6ab4a48736 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 12 Sep 2023 17:43:44 +0200
Subject: [PATCH 045/142] merged

---
 icu_benchmarks/data/Checking data-Copy1.ipynb | 582 ++++++++++++++++++
 output.txt                                    | 475 ++++++++++++++
 saved_dictionary.pkl                          | Bin 0 -> 20818976 bytes
 3 files changed, 1057 insertions(+)
 create mode 100644 icu_benchmarks/data/Checking data-Copy1.ipynb
 create mode 100644 output.txt
 create mode 100644 saved_dictionary.pkl

diff --git a/icu_benchmarks/data/Checking data-Copy1.ipynb b/icu_benchmarks/data/Checking data-Copy1.ipynb
new file mode 100644
index 00000000..9abe4d48
--- /dev/null
+++ b/icu_benchmarks/data/Checking data-Copy1.ipynb	
@@ -0,0 +1,582 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "6b471a18",
+   "metadata": {},
+   "outputs": [
+    {
+     "ename": "ModuleNotFoundError",
+     "evalue": "No module named 'icu_benchmarks'",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
+      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
+      "Input \u001b[0;32mIn [4]\u001b[0m, in \u001b[0;36m<cell line: 1>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01micu_benchmarks\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mdata\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01msplit_process_data\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m preprocess_data\n",
+      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'icu_benchmarks'"
+     ]
+    }
+   ],
+   "source": [
+    "from icu_benchmarks.data.split_process_data import preprocess_data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "9786d96d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "0\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(1-1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "b9f28515",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "dyn=pd.read_parquet('../../demo_data/los/eicu_demo/dyn.parquet', engine='pyarrow')\n",
+    "outc=pd.read_parquet('../../demo_data/los/eicu_demo/outc.parquet', engine='pyarrow')\n",
+    "sta=pd.read_parquet('../../demo_data/los/eicu_demo/sta.parquet', engine='pyarrow')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "35342dde",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "dyn.columns"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "c6f3e4ae",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "dyn.columns"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "64c217e9",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "outc.head()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "87724d04",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sta.head()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "fddac54c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sta.columns"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "265af37f",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from typing import List\n",
+    "from pandas import DataFrame\n",
+    "import numpy as np\n",
+    "import gin\n",
+    "from torch import Tensor, cat, from_numpy, float32,empty,stack\n",
+    "from torch.utils.data import Dataset\n",
+    "import logging\n",
+    "from typing import Dict, Tuple\n",
+    "from constants import DataSegment as Segment\n",
+    "from constants import DataSplit as Split\n",
+    "from collections import OrderedDict\n",
+    "\n",
+    "FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target', 'id']\n",
+    "def ampute_data(data, mechanism, p_miss, p_obs=0.3):\n",
+    "    \"\"\"\n",
+    "    Generate missing values for specifics missing-data mechanism and proportion of missing values.\n",
+    "\n",
+    "    Parameters\n",
+    "    ----------\n",
+    "    data : DataFrame\n",
+    "        Data for which missing values will be simulated.\n",
+    "    mechanism : str,\n",
+    "            Indicates the missing-data mechanism to be used. (\"MCAR\", \"MAR\" or \"MNAR\")\n",
+    "    p_miss : float\n",
+    "        Proportion of missing values to generate for variables which will have missing values.\n",
+    "    p_obs : float\n",
+    "            If mecha = \"MAR\" or \"MNAR\", proportion of variables with *no* missing values\n",
+    "            that will be used for the logistic masking model.\n",
+    "\n",
+    "    Returns\n",
+    "    ----------\n",
+    "    imputed_data: DataFrame\n",
+    "        The data with the generated missing values.\n",
+    "    \"\"\"\n",
+    "    logging.info(f\"Applying {mechanism} amputation.\")\n",
+    "    X = torch.tensor(data.values.astype(np.float32))\n",
+    "\n",
+    "    if mechanism == \"MAR\":\n",
+    "        mask = MAR_logistic_mask(X, p_miss, p_obs)\n",
+    "    elif mechanism == \"MNAR\":\n",
+    "        mask = MNAR_logistic_mask(X, p_miss, p_obs)\n",
+    "    elif mechanism == \"MCAR\":\n",
+    "        mask = MCAR_mask(X, p_miss)\n",
+    "    elif mechanism == \"BO\":\n",
+    "        mask = BO_mask(X, p_miss)\n",
+    "    else:\n",
+    "        logging.error(\"Not a valid amputation mechanism. Missing-data mechanisms to be used are MCAR, MAR or MNAR.\")\n",
+    "\n",
+    "    amputed_data = data.mask(mask)\n",
+    "    return amputed_data, mask\n",
+    "class CommonDataset(Dataset):\n",
+    "    \"\"\"Common dataset: subclass of Torch Dataset that represents the data to learn on.\n",
+    "\n",
+    "    Args: data: Dict of the different splits of the data. split: Either 'train','val' or 'test'. vars: Contains the names of\n",
+    "    columns in the data. grouping_segment: str, optional: The segment of the data contains the grouping column with only\n",
+    "    unique values. Defaults to Segment.outcome. Is used to calculate the number of stays in the data.\n",
+    "    \"\"\"\n",
+    "\n",
+    "    def __init__(\n",
+    "        self,\n",
+    "        data: dict,\n",
+    "        split: str = Split.train,\n",
+    "        vars: Dict[str, str] = gin.REQUIRED,\n",
+    "        grouping_segment: str = Segment.outcome,\n",
+    "    ):\n",
+    "        self.split = split\n",
+    "        self.vars = vars\n",
+    "        self.grouping_df = data[split][grouping_segment].set_index(self.vars[\"GROUP\"])\n",
+    "        self.features_df = (\n",
+    "            #drops time coulmn and sets index to stay_id\n",
+    "            data[split][Segment.features].set_index(self.vars[\"GROUP\"]).drop(labels=self.vars[\"SEQUENCE\"], axis=1)\n",
+    "        )\n",
+    "\n",
+    "        # calculate basic info for the data\n",
+    "        self.num_stays = self.grouping_df.index.unique().shape[0]\n",
+    "        self.maxlen = self.features_df.groupby([self.vars[\"GROUP\"]]).size().max()\n",
+    "\n",
+    "    def ram_cache(self, cache: bool = True):\n",
+    "        self._cached_dataset = None\n",
+    "        if cache:\n",
+    "            logging.info(\"Caching dataset in ram.\")\n",
+    "            self._cached_dataset = [self[i] for i in range(len(self))]\n",
+    "\n",
+    "    def __len__(self) -> int:\n",
+    "        \"\"\"Returns number of stays in the data.\n",
+    "\n",
+    "        Returns:\n",
+    "            number of stays in the data\n",
+    "        \"\"\"\n",
+    "        return self.num_stays\n",
+    "\n",
+    "    def get_feature_names(self):\n",
+    "        return self.features_df.columns\n",
+    "\n",
+    "    def to_tensor(self):\n",
+    "        values = []\n",
+    "        for entry in self:\n",
+    "            for i, value in enumerate(entry):\n",
+    "                if len(values) <= i:\n",
+    "                    values.append([])\n",
+    "                values[i].append(value.unsqueeze(0))\n",
+    "        return [cat(value, dim=0) for value in values]\n",
+    "\n",
+    "\n",
+    "@gin.configurable(\"PredictionDataset\")\n",
+    "class PredictionDataset(CommonDataset):\n",
+    "    \"\"\"Subclass of common dataset for prediction tasks.\n",
+    "\n",
+    "    Args:\n",
+    "        ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.\n",
+    "    \"\"\"\n",
+    "\n",
+    "    def __init__(self, *args, ram_cache: bool = True, **kwargs):\n",
+    "        super().__init__(*args, grouping_segment=Segment.outcome, **kwargs)\n",
+    "        self.outcome_df = self.grouping_df\n",
+    "        self.ram_cache(ram_cache)\n",
+    "\n",
+    "    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:\n",
+    "        \"\"\"Function to sample from the data split of choice. Used for deep learning implementations.\n",
+    "\n",
+    "        Args:\n",
+    "            idx: A specific row index to sample.\n",
+    "\n",
+    "        Returns:\n",
+    "            A sample from the data, consisting of data, labels and padding mask.\n",
+    "        \"\"\"\n",
+    "        if self._cached_dataset is not None:\n",
+    "            return self._cached_dataset[idx]\n",
+    "\n",
+    "        pad_value = 0.0\n",
+    "        stay_id = self.outcome_df.index.unique()[idx]  # [self.vars[\"GROUP\"]]\n",
+    "\n",
+    "        # slice to make sure to always return a DF\n",
+    "        window = self.features_df.loc[stay_id:stay_id].to_numpy()\n",
+    "        labels = self.outcome_df.loc[stay_id:stay_id][self.vars[\"LABEL\"]].to_numpy(dtype=float)\n",
+    "\n",
+    "        if len(labels) == 1:\n",
+    "            # only one label per stay, align with window\n",
+    "            labels = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, labels], axis=0)\n",
+    "\n",
+    "        length_diff = self.maxlen - window.shape[0]\n",
+    "        \n",
+    "        \n",
+    "        pad_mask = np.ones(window.shape[0])\n",
+    "\n",
+    "        # Padding the array to fulfill size requirement\n",
+    "        if length_diff > 0:\n",
+    "            # window shorter than the longest window in dataset, pad to same length\n",
+    "            window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)\n",
+    "            labels = np.concatenate([labels, np.ones(length_diff) * pad_value], axis=0)\n",
+    "            pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)\n",
+    "\n",
+    "        not_labeled = np.argwhere(np.isnan(labels))\n",
+    "        if len(not_labeled) > 0:\n",
+    "            labels[not_labeled] = -1\n",
+    "            pad_mask[not_labeled] = 0\n",
+    "\n",
+    "        pad_mask = pad_mask.astype(bool)\n",
+    "        labels = labels.astype(np.float32)\n",
+    "        data = window.astype(np.float32)\n",
+    "\n",
+    "        return from_numpy(data), from_numpy(labels), from_numpy(pad_mask)\n",
+    "\n",
+    "    def get_balance(self) -> list:\n",
+    "        \"\"\"Return the weight balance for the split of interest.\n",
+    "\n",
+    "        Returns:\n",
+    "            Weights for each label.\n",
+    "        \"\"\"\n",
+    "        counts = self.outcome_df[self.vars[\"LABEL\"]].value_counts()\n",
+    "        return list((1 / counts) * np.sum(counts) / counts.shape[0])\n",
+    "\n",
+    "    def get_data_and_labels(self) -> Tuple[np.array, np.array]:\n",
+    "        \"\"\"Function to return all the data and labels aligned at once.\n",
+    "\n",
+    "        We use this function for the ML methods which don't require an iterator.\n",
+    "\n",
+    "        Returns:\n",
+    "            A Tuple containing data points and label for the split.\n",
+    "        \"\"\"\n",
+    "        labels = self.outcome_df[self.vars[\"LABEL\"]].to_numpy().astype(float)\n",
+    "        rep = self.features_df\n",
+    "        if len(labels) == self.num_stays:\n",
+    "            # order of groups could be random, we make sure not to change it\n",
+    "            rep = rep.groupby(level=self.vars[\"GROUP\"], sort=False).last()\n",
+    "        rep = rep.to_numpy().astype(float)\n",
+    "        \n",
+    "        return rep, labels\n",
+    "\n",
+    "    def to_tensor(self):\n",
+    "        data, labels = self.get_data_and_labels()\n",
+    "        return from_numpy(data).to(float32), from_numpy(labels).to(float32)\n",
+    "\n",
+    "@gin.configurable(\"PredictionDatasetTFT\")\n",
+    "class PredictionDatasetTFT(PredictionDataset):\n",
+    "    \"\"\"Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.\n",
+    "\n",
+    "    Args:\n",
+    "        ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.\n",
+    "    \"\"\"\n",
+    "\n",
+    "    def __init__(self, *args, ram_cache: bool = True, **kwargs):\n",
+    "        super().__init__(*args,ram_cache = True, **kwargs)\n",
+    "        \n",
+    "\n",
+    "    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:\n",
+    "        \"\"\"Function to sample from the data split of choice. Used for TFT.\n",
+    "\n",
+    "        Args:\n",
+    "            idx: A specific row index to sample.\n",
+    "\n",
+    "        Returns:\n",
+    "            A sample from the data, consisting of data, labels and padding mask.\n",
+    "        \"\"\"\n",
+    "        if self._cached_dataset is not None:\n",
+    "            return self._cached_dataset[idx]\n",
+    "\n",
+    "        pad_value = 0.0\n",
+    "        stay_id = self.outcome_df.index.unique()[idx]  # [self.vars[\"GROUP\"]]\n",
+    "\n",
+    "        \n",
+    "        # We need to be sure that tensors are returned in the correct order to be processed correclty by tft\n",
+    "        tensors = [[] for _ in range(8)]\n",
+    "        for var in self.features_df.columns:\n",
+    "            if  var == 'sex' :\n",
+    "                tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())\n",
+    "            elif var == 'age' or var== 'height' or var== 'weight':\n",
+    "                tensors[1].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())\n",
+    "            else :\n",
+    "                tensors[5].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())\n",
+    "    \n",
+    "        \n",
+    "        tensors[6].extend(self.outcome_df.loc[stay_id:stay_id][self.vars[\"LABEL\"]].to_numpy(dtype=float))\n",
+    "        tensors[7].append(np.asarray([stay_id]))\n",
+    "        \n",
+    "        window_shape0=np.shape(tensors[0])[1]\n",
+    "       # window = self.features_df.loc[stay_id:stay_id].to_numpy()\n",
+    "       # labels = self.outcome_df.loc[stay_id:stay_id][self.vars[\"LABEL\"]].to_numpy(dtype=float)\n",
+    "        if len(tensors[6]) == 1:\n",
+    "            # only one label per stay, align with window\n",
+    "            tensors[6] = np.concatenate([np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0)\n",
+    "\n",
+    "        length_diff = self.maxlen - window_shape0\n",
+    "        pad_mask = np.ones(window_shape0)\n",
+    "\n",
+    "        # Padding the array to fulfill size requirement\n",
+    "        if length_diff > 0:\n",
+    "            # window shorter than the longest window in dataset, pad to same length\n",
+    "            tensors[0]= np.concatenate([tensors[0], np.ones((length_diff, np.shape(tensors[0])[0])) * pad_value], axis=0)\n",
+    "            tensors[1]= np.concatenate([tensors[1], np.ones((length_diff, np.shape(tensors[1])[0])) * pad_value], axis=0)\n",
+    "          #  tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, np.shape(tensors[2])[0])) * pad_value], axis=0)\n",
+    "          #  tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, np.shape(tensors[3])[0])) * pad_value], axis=0)\n",
+    "          #  tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, np.shape(tensors[4])[0])) * pad_value], axis=0)\n",
+    "            tensors[5]= np.concatenate([tensors[5], np.ones((length_diff, np.shape(tensors[5])[0])) * pad_value], axis=0)\n",
+    "            tensors[7]= np.concatenate([tensors[7], np.ones((length_diff, np.shape(tensors[7])[0])) * stay_id], axis=0)\n",
+    "           # window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)\n",
+    "            \n",
+    "            tensors[6] = np.concatenate([tensors[6], np.ones(length_diff) * pad_value], axis=0)\n",
+    "            pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)\n",
+    "\n",
+    "        not_labeled = np.argwhere(np.isnan(tensors[6]))\n",
+    "        if len(not_labeled) > 0:\n",
+    "            tensors[6][not_labeled] = -1\n",
+    "            pad_mask[not_labeled] = 0\n",
+    "        tensors[6]=[tensors[6]]\n",
+    "        pad_mask = pad_mask.astype(bool)\n",
+    "       # data = window.astype(np.float32)\n",
+    "        tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)\n",
+    "        tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]\n",
+    "        \n",
+    "        return  OrderedDict(zip(FEAT_NAMES, tensors)),from_numpy(pad_mask)\n",
+    "\n",
+    "    \n",
+    "\n",
+    "\n",
+    "\n",
+    "@gin.configurable(\"ImputationDataset\")\n",
+    "class ImputationDataset(CommonDataset):\n",
+    "    \"\"\"Subclass of Common Dataset that contains data for imputation models.\"\"\"\n",
+    "\n",
+    "    def __init__(\n",
+    "        self,\n",
+    "        data: Dict[str, DataFrame],\n",
+    "        split: str = Split.train,\n",
+    "        vars: Dict[str, str] = gin.REQUIRED,\n",
+    "        mask_proportion=0.3,\n",
+    "        mask_method=\"MCAR\",\n",
+    "        mask_observation_proportion=0.3,\n",
+    "        ram_cache: bool = True,\n",
+    "    ):\n",
+    "        \"\"\"\n",
+    "        Args:\n",
+    "            data (Dict[str, DataFrame]): data to use\n",
+    "            split (str, optional): split to apply. Defaults to Split.train.\n",
+    "            vars (Dict[str, str], optional): contains names of columns in the data. Defaults to gin.REQUIRED.\n",
+    "            mask_proportion (float, optional): proportion to artificially mask for amputation. Defaults to 0.3.\n",
+    "            mask_method (str, optional): masking mechanism. Defaults to \"MCAR\".\n",
+    "            mask_observation_proportion (float, optional): poportion of the observed data to be masked. Defaults to 0.3.\n",
+    "            ram_cache (bool, optional): if the dataset should be completely stored in ram and not generated on the fly during\n",
+    "                training. Defaults to True.\n",
+    "        \"\"\"\n",
+    "        super().__init__(data, split, vars, grouping_segment=Segment.static)\n",
+    "        self.amputated_values, self.amputation_mask = ampute_data(\n",
+    "            self.features_df, mask_method, mask_proportion, mask_observation_proportion\n",
+    "        )\n",
+    "        self.amputation_mask = (self.amputation_mask + self.features_df.isna().values).bool()\n",
+    "        self.amputation_mask = DataFrame(self.amputation_mask, columns=self.vars[Segment.dynamic])\n",
+    "        self.amputation_mask[self.vars[\"GROUP\"]] = self.features_df.index\n",
+    "        self.amputation_mask.set_index(self.vars[\"GROUP\"], inplace=True)\n",
+    "\n",
+    "        self.target_missingness_mask = self.features_df.isna()\n",
+    "        self.features_df.fillna(0, inplace=True)\n",
+    "        self.ram_cache(ram_cache)\n",
+    "\n",
+    "    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:\n",
+    "        \"\"\"Function to sample from the data split of choice.\n",
+    "\n",
+    "        Used for deep learning implementations.\n",
+    "\n",
+    "        Args:\n",
+    "            idx: A specific row index to sample.\n",
+    "\n",
+    "        Returns:\n",
+    "            A sample from the data, consisting of data, labels and padding mask.\n",
+    "        \"\"\"\n",
+    "        if self._cached_dataset is not None:\n",
+    "            return self._cached_dataset[idx]\n",
+    "        stay_id = self.grouping_df.iloc[idx].name\n",
+    "\n",
+    "        # slice to make sure to always return a DF\n",
+    "        window = self.features_df.loc[stay_id:stay_id, self.vars[Segment.dynamic]]\n",
+    "        window_missingness_mask = self.target_missingness_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]\n",
+    "        amputated_window = self.amputated_values.loc[stay_id:stay_id, self.vars[Segment.dynamic]]\n",
+    "        amputation_mask = self.amputation_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]\n",
+    "\n",
+    "        return (\n",
+    "            from_numpy(amputated_window.values).to(float32),\n",
+    "            from_numpy(amputation_mask.values).to(float32),\n",
+    "            from_numpy(window.values).to(float32),\n",
+    "            from_numpy(window_missingness_mask.values).to(float32),\n",
+    "        )\n",
+    "\n",
+    "\n",
+    "@gin.configurable(\"ImputationPredictionDataset\")\n",
+    "class ImputationPredictionDataset(Dataset):\n",
+    "    \"\"\"Subclass of torch dataset that represents data with missingness for imputation.\n",
+    "\n",
+    "    Args:\n",
+    "        data (DataFrame): dict of the different splits of the data\n",
+    "        grouping_column (str, optional): column that is used for grouping. Defaults to \"stay_id\".\n",
+    "        select_columns (List[str], optional): the columns to serve as input for the imputation model. Defaults to None.\n",
+    "        ram_cache (bool, optional): wether the dataset should be stored in ram. Defaults to True.\n",
+    "    \"\"\"\n",
+    "\n",
+    "    def __init__(\n",
+    "        self,\n",
+    "        data: DataFrame,\n",
+    "        grouping_column: str = \"stay_id\",\n",
+    "        select_columns: List[str] = None,\n",
+    "        ram_cache: bool = True,\n",
+    "    ):\n",
+    "        self.dyn_df = data\n",
+    "\n",
+    "        if select_columns is not None:\n",
+    "            self.dyn_df = self.dyn_df[list(select_columns) + grouping_column]\n",
+    "\n",
+    "        if grouping_column is not None:\n",
+    "            self.dyn_df = self.dyn_df.set_index(grouping_column)\n",
+    "        else:\n",
+    "            self.dyn_df = data\n",
+    "\n",
+    "        # calculate basic info for the data\n",
+    "        self.group_indices = self.dyn_df.index.unique()\n",
+    "        self.maxlen = self.dyn_df.groupby(grouping_column).size().max()\n",
+    "\n",
+    "        self._cached_dataset = None\n",
+    "        if ram_cache:\n",
+    "            logging.info(\"Caching dataset in ram.\")\n",
+    "            self._cached_dataset = [self[i] for i in range(len(self))]\n",
+    "\n",
+    "    def __len__(self) -> int:\n",
+    "        \"\"\"Returns number of stays in the data.\n",
+    "\n",
+    "        Returns:\n",
+    "            number of stays in the data\n",
+    "        \"\"\"\n",
+    "        return self.group_indices.shape[0]\n",
+    "\n",
+    "    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:\n",
+    "        \"\"\"Function to sample from the data split of choice.\n",
+    "\n",
+    "        Used for deep learning implementations.\n",
+    "\n",
+    "        Args:\n",
+    "            idx: A specific row index to sample.\n",
+    "\n",
+    "        Returns:\n",
+    "            A sample from the data, consisting of data, labels and padding mask.\n",
+    "        \"\"\"\n",
+    "        if self._cached_dataset is not None:\n",
+    "            return self._cached_dataset[idx]\n",
+    "        stay_id = self.group_indices[idx]\n",
+    "\n",
+    "        # slice to make sure to always return a DF\n",
+    "        window = self.dyn_df.loc[stay_id:stay_id, :]\n",
+    "\n",
+    "        return from_numpy(window.values).to(float32)\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "504fd6c4",
+   "metadata": {},
+   "outputs": [
+    {
+     "ename": "KeyError",
+     "evalue": "'train'\n  In call to configurable 'PredictionDataset' (<class '__main__.PredictionDataset'>)",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
+      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
+      "Input \u001b[0;32mIn [9]\u001b[0m, in \u001b[0;36m<cell line: 1>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0m train_dataset \u001b[38;5;241m=\u001b[39m \u001b[43mPredictionDataset\u001b[49m\u001b[43m(\u001b[49m\u001b[43mdyn\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43msplit\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mSplit\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mtrain\u001b[49m\u001b[43m)\u001b[49m\n",
+      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/gin/config.py:1605\u001b[0m, in \u001b[0;36m_make_gin_wrapper.<locals>.gin_wrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m   1603\u001b[0m scope_info \u001b[38;5;241m=\u001b[39m \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m in scope \u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;132;01m{}\u001b[39;00m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;241m.\u001b[39mformat(scope_str) \u001b[38;5;28;01mif\u001b[39;00m scope_str \u001b[38;5;28;01melse\u001b[39;00m \u001b[38;5;124m'\u001b[39m\u001b[38;5;124m'\u001b[39m\n\u001b[1;32m   1604\u001b[0m err_str \u001b[38;5;241m=\u001b[39m err_str\u001b[38;5;241m.\u001b[39mformat(name, fn_or_cls, scope_info)\n\u001b[0;32m-> 1605\u001b[0m \u001b[43mutils\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43maugment_exception_message_and_reraise\u001b[49m\u001b[43m(\u001b[49m\u001b[43me\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43merr_str\u001b[49m\u001b[43m)\u001b[49m\n",
+      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/gin/utils.py:41\u001b[0m, in \u001b[0;36maugment_exception_message_and_reraise\u001b[0;34m(exception, message)\u001b[0m\n\u001b[1;32m     39\u001b[0m proxy \u001b[38;5;241m=\u001b[39m ExceptionProxy()\n\u001b[1;32m     40\u001b[0m ExceptionProxy\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__qualname__\u001b[39m \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mtype\u001b[39m(exception)\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__qualname__\u001b[39m\n\u001b[0;32m---> 41\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m proxy\u001b[38;5;241m.\u001b[39mwith_traceback(exception\u001b[38;5;241m.\u001b[39m__traceback__) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;28mNone\u001b[39m\n",
+      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/gin/config.py:1582\u001b[0m, in \u001b[0;36m_make_gin_wrapper.<locals>.gin_wrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m   1579\u001b[0m new_kwargs\u001b[38;5;241m.\u001b[39mupdate(kwargs)\n\u001b[1;32m   1581\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[0;32m-> 1582\u001b[0m   \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfn\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mnew_args\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mnew_kwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1583\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mException\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m e:  \u001b[38;5;66;03m# pylint: disable=broad-except\u001b[39;00m\n\u001b[1;32m   1584\u001b[0m   err_str \u001b[38;5;241m=\u001b[39m \u001b[38;5;124m'\u001b[39m\u001b[38;5;124m'\u001b[39m\n",
+      "Input \u001b[0;32mIn [3]\u001b[0m, in \u001b[0;36mPredictionDataset.__init__\u001b[0;34m(self, ram_cache, *args, **kwargs)\u001b[0m\n\u001b[1;32m    113\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m__init__\u001b[39m(\u001b[38;5;28mself\u001b[39m, \u001b[38;5;241m*\u001b[39margs, ram_cache: \u001b[38;5;28mbool\u001b[39m \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mTrue\u001b[39;00m, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs):\n\u001b[0;32m--> 114\u001b[0m     \u001b[38;5;28;43msuper\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[38;5;21;43m__init__\u001b[39;49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mgrouping_segment\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mSegment\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43moutcome\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    115\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39moutcome_df \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mgrouping_df\n\u001b[1;32m    116\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mram_cache(ram_cache)\n",
+      "Input \u001b[0;32mIn [3]\u001b[0m, in \u001b[0;36mCommonDataset.__init__\u001b[0;34m(self, data, split, vars, grouping_segment)\u001b[0m\n\u001b[1;32m     66\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39msplit \u001b[38;5;241m=\u001b[39m split\n\u001b[1;32m     67\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mvars \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mvars\u001b[39m\n\u001b[0;32m---> 68\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mgrouping_df \u001b[38;5;241m=\u001b[39m \u001b[43mdata\u001b[49m\u001b[43m[\u001b[49m\u001b[43msplit\u001b[49m\u001b[43m]\u001b[49m[grouping_segment]\u001b[38;5;241m.\u001b[39mset_index(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mvars[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mGROUP\u001b[39m\u001b[38;5;124m\"\u001b[39m])\n\u001b[1;32m     69\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mfeatures_df \u001b[38;5;241m=\u001b[39m (\n\u001b[1;32m     70\u001b[0m     \u001b[38;5;66;03m#drops time coulmn and sets index to stay_id\u001b[39;00m\n\u001b[1;32m     71\u001b[0m     data[split][Segment\u001b[38;5;241m.\u001b[39mfeatures]\u001b[38;5;241m.\u001b[39mset_index(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mvars[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mGROUP\u001b[39m\u001b[38;5;124m\"\u001b[39m])\u001b[38;5;241m.\u001b[39mdrop(labels\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mvars[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mSEQUENCE\u001b[39m\u001b[38;5;124m\"\u001b[39m], axis\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m1\u001b[39m)\n\u001b[1;32m     72\u001b[0m )\n\u001b[1;32m     74\u001b[0m \u001b[38;5;66;03m# calculate basic info for the data\u001b[39;00m\n",
+      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/frame.py:3505\u001b[0m, in \u001b[0;36mDataFrame.__getitem__\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   3503\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mcolumns\u001b[38;5;241m.\u001b[39mnlevels \u001b[38;5;241m>\u001b[39m \u001b[38;5;241m1\u001b[39m:\n\u001b[1;32m   3504\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_getitem_multilevel(key)\n\u001b[0;32m-> 3505\u001b[0m indexer \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mcolumns\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mget_loc\u001b[49m\u001b[43m(\u001b[49m\u001b[43mkey\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   3506\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m is_integer(indexer):\n\u001b[1;32m   3507\u001b[0m     indexer \u001b[38;5;241m=\u001b[39m [indexer]\n",
+      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/indexes/base.py:3631\u001b[0m, in \u001b[0;36mIndex.get_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   3629\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_engine\u001b[38;5;241m.\u001b[39mget_loc(casted_key)\n\u001b[1;32m   3630\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mKeyError\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m err:\n\u001b[0;32m-> 3631\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mKeyError\u001b[39;00m(key) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01merr\u001b[39;00m\n\u001b[1;32m   3632\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mTypeError\u001b[39;00m:\n\u001b[1;32m   3633\u001b[0m     \u001b[38;5;66;03m# If we have a listlike key, _check_indexing_error will raise\u001b[39;00m\n\u001b[1;32m   3634\u001b[0m     \u001b[38;5;66;03m#  InvalidIndexError. Otherwise we fall through and re-raise\u001b[39;00m\n\u001b[1;32m   3635\u001b[0m     \u001b[38;5;66;03m#  the TypeError.\u001b[39;00m\n\u001b[1;32m   3636\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_check_indexing_error(key)\n",
+      "\u001b[0;31mKeyError\u001b[0m: 'train'\n  In call to configurable 'PredictionDataset' (<class '__main__.PredictionDataset'>)"
+     ]
+    }
+   ],
+   "source": [
+    "train_dataset = PredictionDataset(dyn, split=Split.train)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "c7f48f70",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "yaib",
+   "language": "python",
+   "name": "yaib"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/output.txt b/output.txt
new file mode 100644
index 00000000..c197bcbd
--- /dev/null
+++ b/output.txt
@@ -0,0 +1,475 @@
+{'encoder_cat': tensor([], size=(64, 6, 0), dtype=torch.int64), 'encoder_cont': tensor([[[ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        [[-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
+         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
+         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        [[ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        ...,
+
+        [[ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        [[ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205,  0.4882, -1.0575,  ...,  0.0569, -1.9736, -4.7018],
+         [ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        [[-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067, -4.7018],
+         [-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067,  0.2127],
+         [-0.5397,  1.7179,  0.7785,  ...,  0.0569, -1.9736,  0.2127],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]]]), 'encoder_target': tensor([[0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 1, 1, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [1, 1, 1, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0]]), 'encoder_lengths': tensor([3, 3, 3, 3, 3, 6, 6, 6, 6, 6, 3, 6, 6, 6, 6, 3, 3, 3, 3, 5, 6, 6, 3, 3,
+        6, 6, 6, 6, 3, 6, 6, 6, 6, 6, 3, 6, 6, 6, 6, 3, 6, 6, 6, 3, 6, 3, 6, 6,
+        6, 6, 3, 3, 6, 3, 3, 5, 6, 6, 5, 3, 6, 3, 3, 3]), 'decoder_cat': tensor([], size=(64, 12, 0), dtype=torch.int64), 'decoder_cont': tensor([[[ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         ...,
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        [[-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
+         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
+         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
+         ...,
+         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
+         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067, -4.7018],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        [[ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
+         ...,
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        ...,
+
+        [[ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
+         ...,
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        [[ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
+         [ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
+         ...,
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
+
+        [[-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067,  0.2127],
+         [-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067,  0.2127],
+         [-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067,  0.2127],
+         ...,
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
+         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]]]), 'decoder_target': tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]), 'decoder_lengths': tensor([ 3, 11,  8,  8,  9, 12, 12, 12, 12, 12, 10, 12, 12, 12, 12,  7,  6,  3,
+         8, 12, 12, 12,  2, 12, 12, 12, 12, 12,  5, 12, 12, 12, 12, 12, 11, 12,
+        12, 12, 12,  8, 12, 12, 12,  3, 12,  7, 12, 12, 12, 12,  9,  6, 12, 12,
+         4, 12, 12, 12, 12,  7, 12,  7,  5,  4]), 'decoder_time_idx': tensor([[ 10,  11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [ 17,  18,  19,  20,  21,  22,  23,  24,  25,  26,  27,  28],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [ 15,  16,  17,  18,  19,  20,  21,  22,  23,  24,  25,  26],
+        [ 45,  46,  47,  48,  49,  50,  51,  52,  53,  54,  55,  56],
+        [ 48,  49,  50,  51,  52,  53,  54,  55,  56,  57,  58,  59],
+        [ 55,  56,  57,  58,  59,  60,  61,  62,  63,  64,  65,  66],
+        [ 11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21,  22],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [ 36,  37,  38,  39,  40,  41,  42,  43,  44,  45,  46,  47],
+        [128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139],
+        [  6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16,  17],
+        [ 72,  73,  74,  75,  76,  77,  78,  79,  80,  81,  82,  83],
+        [ 86,  87,  88,  89,  90,  91,  92,  93,  94,  95,  96,  97],
+        [106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117],
+        [ 24,  25,  26,  27,  28,  29,  30,  31,  32,  33,  34,  35],
+        [ 67,  68,  69,  70,  71,  72,  73,  74,  75,  76,  77,  78],
+        [  5,   6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16],
+        [ 39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49,  50],
+        [ 38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [  8,   9,  10,  11,  12,  13,  14,  15,  16,  17,  18,  19],
+        [  7,   8,   9,  10,  11,  12,  13,  14,  15,  16,  17,  18],
+        [ 62,  63,  64,  65,  66,  67,  68,  69,  70,  71,  72,  73],
+        [ 34,  35,  36,  37,  38,  39,  40,  41,  42,  43,  44,  45],
+        [ 26,  27,  28,  29,  30,  31,  32,  33,  34,  35,  36,  37],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125],
+        [ 38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49],
+        [125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136],
+        [ 33,  34,  35,  36,  37,  38,  39,  40,  41,  42,  43,  44],
+        [109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120],
+        [  6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16,  17],
+        [ 23,  24,  25,  26,  27,  28,  29,  30,  31,  32,  33,  34],
+        [ 61,  62,  63,  64,  65,  66,  67,  68,  69,  70,  71,  72],
+        [ 38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49],
+        [ 10,  11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21],
+        [ 14,  15,  16,  17,  18,  19,  20,  21,  22,  23,  24,  25],
+        [  6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16,  17],
+        [ 59,  60,  61,  62,  63,  64,  65,  66,  67,  68,  69,  70],
+        [ 35,  36,  37,  38,  39,  40,  41,  42,  43,  44,  45,  46],
+        [ 38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49],
+        [ 25,  26,  27,  28,  29,  30,  31,  32,  33,  34,  35,  36],
+        [ 57,  58,  59,  60,  61,  62,  63,  64,  65,  66,  67,  68],
+        [  8,   9,  10,  11,  12,  13,  14,  15,  16,  17,  18,  19],
+        [ 42,  43,  44,  45,  46,  47,  48,  49,  50,  51,  52,  53],
+        [ 69,  70,  71,  72,  73,  74,  75,  76,  77,  78,  79,  80],
+        [ 28,  29,  30,  31,  32,  33,  34,  35,  36,  37,  38,  39],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [ 46,  47,  48,  49,  50,  51,  52,  53,  54,  55,  56,  57],
+        [ 37,  38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48],
+        [ 30,  31,  32,  33,  34,  35,  36,  37,  38,  39,  40,  41],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [  5,   6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16],
+        [ 57,  58,  59,  60,  61,  62,  63,  64,  65,  66,  67,  68],
+        [ 80,  81,  82,  83,  84,  85,  86,  87,  88,  89,  90,  91],
+        [ 44,  45,  46,  47,  48,  49,  50,  51,  52,  53,  54,  55],
+        [ 10,  11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21],
+        [ 92,  93,  94,  95,  96,  97,  98,  99, 100, 101, 102, 103],
+        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
+        [ 10,  11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21],
+        [113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124]]), 'groups': tensor([[173],
+        [283],
+        [107],
+        [ 34],
+        [388],
+        [429],
+        [126],
+        [376],
+        [265],
+        [213],
+        [314],
+        [431],
+        [185],
+        [360],
+        [247],
+        [203],
+        [245],
+        [222],
+        [299],
+        [ 43],
+        [190],
+        [181],
+        [345],
+        [ 16],
+        [278],
+        [452],
+        [395],
+        [518],
+        [564],
+        [266],
+        [ 76],
+        [187],
+        [464],
+        [266],
+        [527],
+        [ 11],
+        [315],
+        [ 42],
+        [513],
+        [105],
+        [468],
+        [313],
+        [188],
+        [525],
+        [201],
+        [554],
+        [559],
+        [551],
+        [181],
+        [549],
+        [335],
+        [382],
+        [369],
+        [436],
+        [389],
+        [145],
+        [475],
+        [442],
+        [ 43],
+        [527],
+        [207],
+        [554],
+        [331],
+        [239]]), 'target_scale': tensor([[0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.],
+        [0., 0.]])}
+(tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]), None)
diff --git a/saved_dictionary.pkl b/saved_dictionary.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..8a15fe6a723a8fdaa797ab67c82361959fa6c0a4
GIT binary patch
literal 20818976
zcmeF4349Fa8^<4;AQ4#`B65{Qq$oLP-8$=TrOu))N`esg5ofi8)>WZO)KSN(TWQ@b
z#e}-=s546Ai2J^8|98p$TidSPnVsW(XP^CizJ4}4^B%wVop)y5dFOqTJSUdS1}{Ee
zq!dri-n&;^$8IT>|K@Dav`LKy^=hRg`}K(H)-EnFqHXtH@e%EN#dVEON%pQ8*E_EE
zi~mnZDKX^59-ghccIcIoT=1WpckI?XzE`)nE{PFc<GRIli0_q{lAODGm+oyl*L(Rr
zWkO1FK-M0uyL4=m81esIQj+tu?$fPD$F`lj#J6r^*~hYR{%(D`_89OoF<tv~>D@7|
zSFgAMB-GZu;@fubme{*jpSHa%8)kp8VM=mNlK2;!Oh}1=VJS7TwMiL~(m17fOtzS8
z^&WQ^o|2F+IeWX_1A4?;_R@8HJ7v_!mm7E_XRDG@zkdC?&&lV-KQYidrOl}7S7v5|
zXVE%Q<g1_qxp=Rd4rnIrJsn8N0TP}NDmT)B6!iW?2TE|MoepRsp`#9H;ZdRvq{j4x
z4&<a737>T~57L30yfaJ(TBG<#9f+j_*;gqFjnaXXY#FTsskt{v2U_4h$%iux|3L?G
zkaMvP<e>jD9VkK2Y8^PkqP043mLJKsVrcT44v<$pXhuA7g+p6(Ae8Z2bsz^5NVu-D
zDoqE7Ha&Rjpbn%Y^RNzR;O7%Mkei>*=s*mE?&&}dGIM$W?UW1f07O0y{wUx9Ec0oQ
z2Pn-5qX!5jdtnc72NJ?OK+Aj(;Q@}QN49Z>W|1D?h}%&fpmoy9c>oQZs^tM%p>G2Z
zK+pr4>p(wJM-lWuOdYt|*aM`ZZ!^+PAhLK+uQj<AQjxM9JwR&S>f`|kdO%AZ$eH8;
zw9+ft16X49L=T{m^(0-aP~<xga0UOF9w0SO*Lr|G4sG@Tcc$GR9)Qs4L9Xi_pcK6y
zdH{`#d+GsNrJv3d5KlPor7Y|6d4k&Xs^AH7QLv6D$VEasPmqhcT|Gez<39HU<fR8+
zPw)h#h@0pMYV*MiPe7daAZd;#uu9qDUVyyx;9zqva7S89@B*3`lHvt&^5A<fK+pr4
z>wxEUFJKYx<z9fG2d>nCBcu+=!7&o9Gh~wO<m4O)Uo7XzzCuY}=LNJdmZar%{BX|;
z)L?^`H#p)F*#_~3E5rwy`Q7LZ_UKsL8|=}bgEt_rdeD$~;wp(fy#Zm%gMi`QKsyQF
zd4n9}TI>x<60*u0#87*WH?YFeE8d_rx<B>?SBVME1#&aBXf9BSD(~h3SNSm}7a*>C
zaPos(AcPs?aslGJ2cvf60%!P|Z09O-a`=E&nVr`M<X~E)52(%4XdieT8$R~|t?}+i
zA5e<4F+RXL<;VK~qK*fTC-{Jz#4PXuXPLL%2c)L^ZXa+*s$K8_F)VoO1Beeid9FSe
z{Bwg;7#imWt#D~zZm`E?@|zmW8kQTJ;RmvvSV~RL4K%QKMs9G1xE;B{9{cy?2IN%_
zKFQ1tjwo_BH^{||$GHK?4coH$%Fbv{U(h;z{Ct5X@&@>V)JzHS1>~g%r;WZKB@5s7
z1y=a3rZ2EWTpeGK8osqJI7{16zF?1b<Tn=ZPWA;>sJY4)+>syF`hr+)ZubRO$vEu`
zuj4fNT`JCyun^}w$a{`l3)o@SHD5s3_TbSQdJseRB6?7Z8CCV5Cg;2BK}uE((t{kF
zn63w*6gBI?9xzJ}2zo$s9Vov*4|4GHLOrNOo*(s~7N+HTaCg?P)`M7@{GtbhJ`b9b
za*>0de$|7L^xUKeE%EbK(#|620mXG-(l$NF!TjxdK+pr4>p<U~dO*;F*VTdNd-R|M
zGWO}g-I$%G2Q}G9;-`U9$MrxPUuWn+3{y_&K@K*NaOGmaIXx)FBodd}G(N8f<W&!<
zUDSgZ`drZi3pBf?2dz@#rXCPwJ?I-|0C(nRl7E)yTa1*KMi@l$foR8rvTqtdDc&Y=
z$;qCQ29T3-Zy5kFK<iS<vzEk({I>^Dr41k#X=Mz6Xw!q(at1)q1Dfl=)rtl{&;y$5
zz}}h$V2!4A3_uf^?->Ab9}jZH7yzNegOwi`Kq=Ofxa6jNeFJELLnI%sV^0Ip-+`#^
z0oTv~tWvHi>C@5x{XaGU@_G+SwKRa*OlxBRXIb6B0BX_oYXiu|%~=LO*zh3d90MTI
zdQf$q0TAtbFlLnjl%~Tk29N@NjR6pCd2sr710d)D&2_*l(*RPl;k*G@VE8=)AkKTR
z;)wwe`8_D<<p&769z53jfgBvq;|GXx9_;q>0|Y&wxelDo>jz5kQ-B{J=z*g;kSoLw
z#Bes8)MFV)d(#ixl@Znb0D0Ae??dx}SW1WI16G(^J|7^$JgD}5K2QR$rTGAH?7`Rj
z@&PNHxtkBfvRCg9_UQJ8KPb(<vi?9jODg#T@~Q_v5l;~GfaW@IrMf@R#*C)^AQeYq
z{Q;5JgJo^}!4;-<@&_>l^!Ep)nK-~7lw`lzA6Tc>9DhLQ@}S0ge-KJUsy`sgdr<9q
zexRA5H}V4_zXvz(<p)Rn;U54r@pec6XoVYP1AsN=R|x<bc!LW7A^g#R)MJ5s&@2Fi
zQa3IDw8GPG0)R&H&k6vc9NruNw6S$x0MN`;C;(FPPSFA&H3NGU03rBnE&x(v++P5Y
z|MuY9^a9`x{CKqhAkKM^{Z;|c0)IXz0EluPEYt-8BEJWVh|jcOGz5ZBwnhYkSo|sk
zf>>&jeOcgPl|WF6#F~L12M1~ef|R7a9|&^Oyj~zEMT-W3fXM4XdZR#~m8<On0a4F`
z2R#EpN~#SC1mvX$b$<&4p#<&-1UZS{8wdz`Kyw}V>d!zxUiILY%s`NmySD;C3I>!8
z0%wSQCkR-jY{eizwB<q1x<Q~ew@4b~=J6*%AQjW%gFp!yd=>;^>G4GnNWs9dL4c_1
z!O<B(V2?G6f&dZcLFu1@z!e^?3j#U#D>VqDqTG=nP@Ccpg8)GfXs!b@as&hIG|LwZ
za#5*pFbJhViC~b5Vs8b5SpKXK3}U(eVK7L+q*lQ|1J`>6gVwl5@`Q-<;MXC+AQv;n
zlQxq=jxPxY1U(Q^2PUr$21H&Dj_nHuweh|g3_=+CFc@fO!;@f;i)>yYpfs0qg@D{#
z%pC&GFw8du5Z64Irw;+8=pGRQuHaoh1k_?>l@M^2(p5u%CF;H#0`5xo_d@_t-h-Bn
zLO>2;KMVmSXxWye9}(@rFTFxQ3i9?20XcY(6arFF`b%;?KwfyTWoigW#pg3aKn;vb
zLckrE^HT^Q>UrR^Is{mv{n`-F5<%-iKu&^5_;OKlV+hDW<gO4v{_Vl--65bw4jc#p
zcjNbiA%Mu^LEqyc;3^eQg#dycxLOCkJ`(~6dO&j>D0MXiw8XX>As`2t_d<YmibE){
z!Z4CnDH)O@6kK7nPbkPuGKoWr{N)!4T4HrjC`ie&BB7vV!pesNjkK>83Q}>oMkpY3
zc#u{z6r^TJqfkIzdN99nD9A;XmZ5;Smj@%?E(o-;t6o8nf?2_Z06`CEt^-SA3xSp~
zPb>tkkY{rtK%DbHpH>L8f`Q}(QO|=e7YYH*G`?L35cELqI?&^BAyAuioe`vDXOIyP
z^g!x5FsP;x+?kA~Mo^PX62G$;yBk3*GRKf~s>sUyMqq`JIST{J6v$f`5Mdq!63-Cr
zdoVe^Ft`HOu`syGfxU$RvG2j3rwaoMl!+<=&d~A0A|NM2`xXJQ<egLm<RaVQA|MrM
z&x!zo9?)C|PE;%kv=hh`1%w_C@;5CC&NB0pqCgAXdlv<{I5CDCM+o&ERGd;25OzI?
zpI#K0Wd=#RlWd<;6cBYi&^;>(h&mqR$Por=aosx%Xkd>a3|QxBR2WE2^S8qQ`P+jj
zmBWB0{z?b~+G*aO_$LALhJ*p4oCkx(g@KkCO!CJX?Wcx;l-&D13|wW+>@c7SpC7|O
z3U>S&2CR_yXBdcOL}nNu^my?3#V`;<<n=IMfp+?0Ae4Br58}KBt=}jHLWw2&h$Y{<
z#Q>4dgKO1_f!ds?Sqx}oU)y3J1$D+2110bqUku!tk#mZHE6n-17$C}du!%1QQuAbU
zF>rV49xDb~=Ki%}Acns8ivjYt2Nk`;L24?KFcIfHNG5(MMUPzJAT>Qnn22&7q`wgk
zG|?<99B5(jTj3xl-71HJ60B_+4#?jg#5D^CnyAw~9HeA@zi@Dd=0n3lZt{-{2Uj@U
z^-VzB$AgZ8-UKBmd*w}V2Uh8dgIshl76(@`H7*Wvk<_dd$U(PHN&$i%{Id?+A5{wE
zq~G>ZAO}e&$uYt>-<nH<yRpAf1h_M4B>(Qr*5(mF1LX%rfUDeH76Ax)V66kiHbel8
z{FxR3tPpZ20%&K%kqD5Ju&WW^DvusU0OGm_m3$+CX2u6Z0)igUTnF|AMFK6nTRsw;
z<xqo2KwS4=NApNP*zsT|DKEsmJ=irM5)j88d^R`|q+;UgNN|PzzeED<_-u{@gf0)}
z)rbPAxK=X?5cGiNI*<_?1!~f`Qxs^Ozq&+$8h8(k0t7vvxekmO9|g$Y9(?qD6wpG>
z8BxFzNo%4&s~k&<0t7vvsSeb-5e0}k9(1`A1*|jZUKF@9({n_F5*T8l!4>v=91Tj)
zDLxvIe|xaHeKaVExpOqQ8(Th)2Cd=$RW!&&*HO_R2SIzv0L!#LR0dpO?b$LQ2M0;G
z&JdBSEFiD<Aewk2B{d6_1q3~yxef%CEDNkLF)SOr_@wksX_M^LrccK%y*qYGOi9k3
z*rj9J_>>7LF%Xk2CfkUV#wjs6%fAVElJoV5>((wVF`{*sj%^Yn;(GOp8$k95ZQZ)Z
zfc~xfbnDTvZRalWt?SpUX8EOhT-(m^?MOh$Z)EM(wtKJmti2+7ckCM9F1}0eI1+%q
ziS^5i9my_re{I>+xBiP*Lz8t~E2NAXY57t-#xucKzkdC?&&lV-KQZt{O7R*xl5k66
zW3u&58E(l?Lf(Y@F)%D8p}@=C$S?ICcNm_MAS$#`)sJ!>FWs{IYyLH=w&km-XIyd)
zq8TN|AMhrcbb3>uFHz(7w_6#ArhXVUHy_d1+)XbB5QQznqJxQ;zgXSx4Wg!U=hqvF
zW?U=a8Adc|{(E)bBx;NuFs>v~>qn{hQUezqE=?khO-m1rwnYB5PusVNn5$P=Se~fq
zahL0rh-R#q{?5BZle+93SdFN$*z=7wiKd>2^m>n|^}|}aW50{7A7e>;gKtyn6EWx9
zl-`J_DfO1IDbbAnMLT><G%32tk`_dbm%h5)ifHN&t17l38r$OB;CQ0oUtntj5%cak
zxw{ZG4eQ^yJJF1)3n%v?nso2j$v#AlKj$hofM{xm%AG$a8f)ybY%o#lhgSdOPM_T2
z+a70*>OF1garQX0SD&?(AE}Q|^yk0ZFIIeEC7rhC&Ho>-Ss9?U{kM1C|9|?`K7SsY
zc7G`G(ERP5s>6wz>OT8&6wwS_<j!QGNx!u=j3a97J+9dVqN$}ePWz5%Z060=Q;5Qh
zA|<9<H0#4`Ik297jnpG^vk$wQNm-cke6?ye(Tr^?AI~G2l$2R*5mD<$>3XJCysq9~
zIBI8T&lm0Kv$pT7q|@$sRAcRj)a4;{c#irr9NlL*y05|XzP`W6Re!DhH}=|Bsr{Ea
zJZt@vD_pJh)7dL$sqd#9_IF8iz3%8b%UCgB<PuUjQg7DYyOe0`qJI7>EtRG1g66*v
zF^3#8uO(_a=zV@Y(TwCur8g2ys?|AhGg0HS@7HW2n!0-T^PNOvyFab9mnamENKPYS
z&S-VuAW_qlZvu}J&1jU`>NwG)TsP;OBx?OAeVzCJ>#eK=$Q{4x`4rf1O-p)dNr%w>
zer!jSZAXm17ruOoRF>3%jiWP(#=iY!zYCVib9Ke~%S6m`&w5-ZYHF3g?k%Dj0ky~7
zC7QIa@8JhTjU(rWJ|UWFI@;zr(b$Jx3w4&Kfy*mi&ra0bsna{&L`_A%ADEkH#_?Sn
z^AJr+d6F|9QDeQ*^#h2eW^a`eL^O88*z{1M(2qA7iJBvCv@c53biueJoM?u*(d`mM
zlRg<*u{2R*-W7wRh^Fp5v!yK2*f0HamnRBUYBjDz)O@GU<f=qXOXi(45zUA{5>}IF
zQbDiIb%+{|R9IG*XzDi|@7E(5``*;5jfg_FU0*gKYF_hr$45j>y-FEc5X~t0X|q;D
zlg^Bp7Dv=LZNuqyL{pnwFP=a&*4Nm*Gf~*ua8);==0QUq_atg6zr0!>q8ZmukLXV{
zX+gfd1Bn{rYWfc*ni|xn`B0+PkKFC5^|_$X>pa@iXT9H4>OM{@ekh#}rQ;J?&u!}w
zYw2}#|E1P-Ir{kD&%>O4p6@32H`HF2WL02V&(F5)7Om}o=;vu};=fYuq*C#Sz24M(
zea?D5ELDFKmM7!qSBcs;sq2AM;Yk&av*TV}z3$iC`BS3y2Fvl)b_V@;)ephrlx4cF
z!E&LqTx`cJsr@{S+WSp6?FWxNFxNcH(jPwh@VSvh%{6kCP9|!4Qa<q;q8Td_){G~b
z)OE`9iA0UzJ8MlLntJll=xIb_C%tuGCQ)eEGH^Cgv)7nb^N5-@uAj4zXvTnRmlhLE
zDpM$WDN*C)2K`nLO`V;zel^k9mdiZW5{3My-(OGEyeIFtRHCNgH4gtyG~?agq1%Wi
z-JjEDCsE_lLksp0O-;ygeLvCI!sXvNNEFgL3_L>AJbv=VbfTu19XU@D&G2|s?-bFb
zbtO~I5;gW|k$#?NYUyO-Wumd?*0sMz6lPvse3Pj8qk^~Z5H%SZRD3`*WBcI2kBKG?
zS-RyJQDdc3J~|!Q|CV3l>_lUKtUlR`D6~sF=|j{UIwvd-QPZJ=o$?aR7?XWjexgaW
z%iRwoYJA?lYADgvUnYH7h-hq&?K_GPg%S_*gj+Q0gZ&;N-@}pV{D$o)K2Lnz^F-F$
zQCiy_TGQp|eM%f(;ojpb+`is2avkIL<>cyn&v?Cs<G09tk<#P0Sijz4<-Qf$M{)gn
z=$km-{y(3FgYyddE6iW?OPF8(n%~7veN^4otOU^BUntfmvHOXgzX_jX|M$EKQjMDt
z{GJvg^;@ZN@U%Sd#r|-*`@^-`ZfZ4Oa?MkRe#vv)Cv;vSq32I5H#W;nuJ_ZmULUlc
zKWFD3Wz`2vw{!CYOLc##wBJg{CzVh59A4h%@LH`GTFsZx`wa8}uS1=gCrRwMZmeJG
z)Gzd19p7!|D;C->AlLn+b^bG~7w%Oru)JSi-pu=xl>yqH599MQb?3Phd;Ny<)5$YG
z9gc6r_wuAz*NfHuFXqp?8k@)7-Z~igJ~_^-`p<b)wRYZOtsRm&y;6r~o$ue?y!?_s
z_sPnD()T|~`KDAntj#OuvM(ytoBz-IFK|A5SLe@i*{`g1a-|9f=N)rx9Xwk4m9aj&
zzJJl-Ix8aEA+f%T-4EZl$XW;5&U->~j~A8uxe#^h;$)p4vA(+4zPL$!#r|#A)?dNn
zOq%12H6K|^uh9D-rSqqBe3JL4P};AJzDm<RiC!<l$@#8nDJNRe;p)15TFHkw=ce>V
zBs13al4!{nE$MKzp4j>;bR4*OH_pfUx^-fey1$USKFS?_(){=8zDXis?Rf&O8}{nD
z$YSfe)b&g1@Wei!q$M4Ap7^l4nG}Yz<5gHIH`dZC^nP5i_oTCoRC-*1TF-Y=e^^zq
z4aswCi*tkHEqP~T-q);R0<D#kR`bQ$eFc^S8~sK&&Q#iQrp(v-tPDtfUSP!+R?>;@
zbxHGHm(t~}bbRjeoQl<PKv*AMUMJD&1Fh+D_jQuQmW$Z^SRIEcb$+D|&)I%qwc@3<
zUrA~|#D*t!KUR+?R{Q(ZzR#Dco=6pr((l2rnr>F#b5AWr*?J4bBa;6)-<wwIgI4n4
z?&El{9ADpG&x${+q!arurRl#+UA_DHPEzQx-JjOlYbku2^9(EV4`RK2ef#6^eO5c~
zo1w4r_-gHMX2oB*(kV7vvHM}a!GFIGjeg2=oTIPu_-buGlYVP$7)MMt_8!-40#UW^
z+tn&>wc@qy-*!12Spi5re#F&ztz6Dm);v5)_ZwpWo4o6+*v`Kp^!*F5^-Ao1c>VL=
z*FUDZ&%PW*Or`a_;_Q98tMxG>p!T@C#0Q)|+PV3ol|BxYj*ph{U+$Xk|GyWKWdLp0
zMb>ddtd);a=@lD~*!}SQj_3CamdbN=#rn%c%yZ9rTqkO3mA~#SqIiD)v{|cc<of1}
zbLV&SpdpWYlI>09msjgUG~@c|5&bRuFUYrdAW>snO@A%Vvuz$fd1JaI|1mpqp0woO
z<59g+mi(_PnR3>W|2{3!&l628oou{JH1-^>58=xC5abYj@uBrRpVVdVz-lCgQa^vq
z%C_d-Cw27RrMu?)|JSvQyzeac_(!$sfm-p}_8a5;7{cbqP<mcrq5Ya#z0PFqcwsHQ
zLgyvXdj6c9cZJq@;IMq$>-jf6r#Q2vtmQuVOaI*Ei9(fHjVlo~-{~{CDpAvtc_&Ro
zGvbeg)g+o!(5rJDqQ)bx%>!<nw&8R;5>9H9>%|j@#`+q&cP0v38?Nd`6#Fr-A0wDl
z2c^azV7rg){y*PWvW_Fi=hN8kFDB=sq%wt~ml8EzZqRQ9(bU;V>sJ$vZMn>2Em6pS
z`u+7p&3p2WOC@R=UgPlZL^Iy)9lDKZ()~GYb`mu%J+xpC(bR+-*Y^{REnNPcgG3>{
z!@wg%vHyBlw36+<(EB&m?{AWlGOH~j)lKgHL32i{12oKYZ7TZxz})1x$~eAjV;-VO
zDNl0dBWkQC?L26B{ruK}nZ#FU*fMZ7QM1>WR`ZCOHm;wukZ8t$YnOzyqa>skA9s7M
zMC)?~ED!g%K6d+Z&N2YootNu$xP5(a@_Db)J}DiatK(`}ORx35fYo$6dXGk{`EoRV
z#rFCNeQ=NK8soF?*K8w&E%tej()C2?_*`w@t?i|wc1P&(Db;gg^pnlw9LtT(a>Man
zW79(|{bBuhqtVhI7I~w6QA>Z=1>=%%OMjTT(d`mMlRg<*u{2TZ2k$%GbbZLI3V`J=
zPx<5VMdx`0%Z1Kz$=J5?@jOyE_`J})pBHNN_+&m`Ss7sVdV%%yH~2QCKFNc-&!cP(
zIc8o<d@vpKKEIx5MslUn8;K^>>YTWlsI&cUWYZTPX6v()^}${8X}t8+?N%hSxPF9l
z^A$T9XUb~5VKv>f_G9C`UpOyF3rqiDNqOf5!Q&S1%kg~(a($2Q!MQY!7vG0?*^+Ag
zz6P)NT^nD5^Ht#c#Q0tr=`Z+Y(wBvF<h{5a+jkTp3MC%q3Abq02isA3o{w7BB|x8F
zf8CG%yxt$N>o(y$<~WZz4fCAh@r}nf_Iv!h-$Sd%g;w&Vm3(OBdAiHxkre>m@5p<<
zL(4jBV&6-0mwqFck0)Hu?EX+<qWRlBRfiLG_Bk^8PS^D?mIqzsVaoH>s@bGaSQ{^m
z_2^|ilFJ9V!gY0BQ0w|vxGu)?jWtOjS>IcB*YdKR4?(K;t)0ETh|RotdWyyT8AVD=
zCt_}f>k4?2@m*S(w+@dhyk23UKNagK)>9VhDb8E!-1>#?ay-+PPukLBdferDC6Xw)
z?|bk%`f07bmclO{k&Nx=zn`Ddwtfm8r+B@C$0_~CDX!Oy>ovbg8f#<hfN>>>ik+ts
z+k0u-d-PLUKe4~~)%TjQeqsI6()AhEXROa!s?YZ3aW&>$F*u4mzf0Xo*}AG(2Ouls
zNwK}a_Cm|;1=eq@-&(HU)_KjTJs;va<G9|er4gp=y#PEf;CTVh3;0~d)%A0&^KL2K
zo+urk*yka}VlT&kvzA`#cy*kgw4eV<OX0O$(EJx7Lg%5za>w(mGv`^g>WNzM#-^o*
zMq48P+NbT?M9kHzEG$n{E92-aS+jgt=?BI3zV_awq`X*LPwqt}(@;IrT7FEaw~S3m
zrqo)G)ZNEBV11J3`b#Ze#l|akzu4SOF9%q5rmX+YUc2Hh`ILJeX!KE@c0+6PZ(=>c
zdg9Fey}QhxWm+=tq{*^wO#G3snnaTddUdWt)OZBvUB`LXao%-TuUlw6-_V*ad><Fr
zSI%QO52ZM_zOw4|I4c24^&6|6UxoVrwVqF{=PxO`$&wZ%ldhgWjX&oqHh_ea+M#mi
z&xytwyDS@QG4I$kQS+AC&MRgK%JNa__;qz$0zUU)^SO_4<owVlq@Jakj<$JDH1?s_
zLLE8J!SagNvlBIU>huoH^ShChzWB&}oG2BKtM5%)#|h!_h{t0PIeud|j7<-<9MAoD
zqtS9aN8V^(l&GWk!LYu&*dLi?pxo^ewin}?O(0&UmfAS&JEFGNQ8+)VH0N)r+5@S=
zF@LeT-y0+_Q@Qi&jYMgmXWH5M*3cgp+g*w7ziitLEEi?U1$}<iXSw~5JN%^i@71-;
z2W9=peLbX9Jb2wD?79of<<;ZbG^~H)?!=UgstYIgB5J!1tx|p|6%Xsj3;HCi{{E|9
zJj+12+kaQ<87t*1cRJ+`-+Eo(F6D*&vH>HPkV=($la78_R_Zs_8|Uf`mZyv7XPNp<
zSX*b%Jol`}byAs3tzOP=iuH=FdWGddS9!3yo~Z3{o)J)c++9++lJ@mI{D7!h<07zr
zWBu0Bdmvb!|GPdb<%?4B*xtvxTfADGheWM%l-p0%@<n@kX-}WC>xXM4ALg8!(i@S?
zSl{p0YQAtjluD&HlEO`@)j4r9QRB1k*K8x2x_bBXokU~1YipfDlBE|P+MgrQ)^Cfx
z{r7daz31=$dVY!hqw6n|N<?eFiNpEZvkm}S+YLNE@%a3&ejlakg;Md@_7g~UpGW!e
z?d<)B?R7g|-@4cJt+VfEQMX<fu1|vNlUUkdM&>u9{r!L2_fKdWM?w4b0&VG&Iz3W{
zr?vO!Y{%0}bw7{3<9*tH-KS+VN^NzVRF0%vH|Lxrit{w8J5M7Xr!S9Ft@uD|K4?vs
z*!c?8%8y#{+UtL!Ej?OU&qizWc5D4SFZMVPyB}-oBw#-RUE>W@uXkAqp!GSU?Q;m#
z{jxJ}E3N4gdtGH6hb8oR4*G!o4cOm6#+}^hGr6i|-a=d#V$MS1L&kt>mlhLEDpM$W
zDN*C)2K{hd2r};Rmq}k1(vfkGJ+|*CLKI3o%oA?WtPd;aWm@NNWz|Qe(ydfHLdVgl
zl`eOS7yFaGy4{Ktdg>3WDz+hNZ{AC_>V>`X!Tz<cR?Q}bVmr<N%L(s4)VTj(wcfFs
z?%3~Q>&I9U-{9Mn`b3l(hl+k-`=Uns!rD3~xbCHE{TtZMU^}CQb_VP5%X%!g4|0bu
z)%>Tly=QE@4j!vMvYKw8*YR4(r_}k7Iy@X-|Ln_A#8e#rC2XESt=2=;^OcnVcY7Wv
zm+x|gtCe-VTzwx2{dLd#S1fn;EO)KV+vV)M2CUV0*3#=}T#?rDBX-=ad3T-MU5FWu
z*6EV!y<%LqHf!DatO`iWbvbLt1Fh+$HC;-N=f`%X)q#VgGMJ`(6L{2GDTvZ?yj#bC
zS(8Rvy5y>_?v`Ii;}cmw?_+u5eWMnhcRPFE_^<i!vJL=uX}^_j&#--=yM1B4d>B1`
zS@WH>^g8Ol!u}rb^Xo~WW+clqZ!Xpsy4(Gvq|9oINFf<3zFapK%Zsh@61zS&ma}`5
zv)0!4p{3r^k`8nADhtb##8~?w<?8QK`hGF#x7LPnBwxnf<C;w%np$e(wC{-GbFZxD
zUdt<9&ras?Fn8+ojyF+L(eDT5CYo`4*Ty_VlTx1K%tzE%uXOzYqN&+ir34X;-7q#i
zlqmG$jYguP`%9|SYFx>(KhCq#Y98^yv~j(v^Q@2(eeuEmD|`<B&*x=0&-9V7n&kLQ
zD(Ka@4pDqB(zW*@jY|(L*h73wO~`S5KhfC2<=;6-6w*83dM1_@7}qnoKa|*D{&r8*
z;Y3Y!rG1aXVzK4J%6aZ(YS~GzR{Kk_9OUWmpxsy1)0dS1R@y6j$G_@6+L`wk`#V-X
zo<|A?`|)2te;jS&vuHn_XiuN*yyW)UEv4qAqHbJ&Rsphp-chRjXsz$GrVG!z|9##S
z`#e@_&*|iT&cMoX!Ad%5pBG53{v_M$W~Gi_R_g_;=@vTA5w6oj*E&r&|E95fXUlVz
ztqoUoBWfNr<Z(};rt-_H^&y&Z{q%_bM3WZe+dGh`F|MZnV4|r(eVPv?8hc=_c^FX`
zefZo+qUIVoOD7XGJt?2~4bhC132Vj^P3k)3`9z}D58gird!9_2Z?vTc$0_~yy!qDg
zw@O`q;J5%B7l7+Qk@H>ca`ywtdQZmZ?W=|o#r2?^TaSgd{&a`;aakuq)vjmMjvt?&
zm6%S-z})P^?q;H<JlZ%9XH_d^#>2SFa~docdB&UIe89ry16J+*fUE;Rwev0O=R5Ro
zYXb2A=f8S&{wut1$NP2`@7wWu=;ieg?LN?+J{<S;>U<5@UySdyig~_g4BuI63VF_#
zdh*feX+&cuy>(!w<+)$OmVvX0n!UzQ@m{MbW=GDGBt01(kLsNwnzXKD%2}euJ}uJE
z6SaQuK10}ZO51g}m>pk}RRP?#{AJ3&yB=5Wn(x>a=LW}HGf&iQk0*b5`Yo;2qOU5Q
zr?LtF{a4BV8AUYt{4FEk)&6)q&gE%8@wv~d^~^YOe&`cY*-}kM+dL<V<74F+AFKND
zoRt8z&s(bd!20vNW7A}v7scWDr?v~4|3V7U9CFOOmZ<5Vyz~C14s0BqN#cop`^$b8
ztd++S|F2QCJuLC}j7!czG^51$1KvcFPHzhIC2IWsb}IwX)DOev<|7)LyXoZsqByV4
ztM$iXiRB}epD#m`E<DB8ug=PVRDL*{epY>>HQltP3*UFf_nmQGRdSw|=>0OSJby|p
zoou{Jo<qf+Ti5=Y<$2W1tBY?EHGfp__8p=oLxYMBh-PdbJoquuq#;YUJR@qXbjn9(
zdEelcU*qgVV}GnZ*^4N&OFZdA)Eqh|EDurBp@W_B63rNseOZ2@_`X8X4j&VnlA@a|
zX+c!(_jE}VFFs0NSHz~Jhelgs#`_+(=r?4wKj-DqA^m-WT;XP1E8rPM0#BO%Ufnl|
z8e<2HD@impaM9t?L}@?nND^OsSZj}1O*hUP(a(RSrSR15Kf1bN{biC2^W3u@*NK{1
z<*$2-sO|F?R@+gb^OKXTzxc4$4&d>G$CD=3AHm~KmG^V9Y?rLx+uF}7SGcTSZ)Q|o
zIJp-wC+Xg?lYNL9f6i5G08#4ZJH_i0N!D}1;{eC6xZ^tT?pp8NHQ)aq?`$*P7yY24
z{bQvZSV<?&@B4CnYo*_p#QO32_5}TTy+6+0chTk}ZRx>zN3}8E87=+&Sihxdw^V!Y
zII95h`H8&mgVBDR(2_n{(qXP%Wnp=e7~0O4wE06@dT1*zXRljr+kdruRV!ZF#~b1E
zpVU^zN##h&b#u;1qQ>6~Up__jU+*_*e0+-1$A{AKu`>Qo?D43SpGw8U+PF=z<z}k;
z?8{NaRDA#a<@@hS)i0&uv8~5y#f#S=H&0KonC~X@l*l$tHmd@=d%p@E$HMvtY4eM=
z^tgI|ueJP`Qg0cXl1ydvSN45C+x`k>uFJA2fNJfs+VSIil=vPcdGA`T^-x%8Us*{f
zj#rX*yb{&}Vb`bBz3-s#`4sC5<Ml=6_hVS8AFQMk&wuis|I8WktgEHf`o&7VSV<=y
z-)tV=N&EU9en2XjZ2jl`_}9M2iS_&S*J0?-fBVC{FIgE-t^ZQu{zK-=qrtZ+^+_ho
zIX9&@B8vU-($1%i_2a+WG3z)Q+WnC_eNu;Ko%clP`AJmIM^*x~avs5USK4;h-N*R~
zorliUtCXPyd9IdG^3!Imh$fvGGcAs&aoUE{?TDr}xn4YhXsoY1>$8}ilu!JIgrBi8
zVa<4=){j<>Z>{CW-P<cx%afLL(~=IM@5!;2Zr0M9dLq*6J(6ggUlivT)x`XwTN|$G
zMy_YfgN8irNz_z+d9^;4>zwPSNAxF3%Q_s)w-Z?zVC8y|dA~K%Z;0ja`u;hw?|;)a
zo*K*9J<8em?E5v_NMWb0-u-+h(b(=!YwfibK2cWA4|pA1d+$<9`7M&@IykEiVEMnC
zC&+q8sO<+W=@8lvfcK{o9hb!;lHGBh>ew&+bC)Oge^8}X<4Q!$clu1OO4PJu-boYD
zjQAsAHHos`kDt_a%JYdNA7a;u#P&hhbsPF6%r9r(2SHz@@s)WWvoe6!O+IbkCSKrm
z(}&&7#0OIz_nwzG>#)H4>)j&fVa+m7?REu^XFQ(0E%zZszaN;}avySh*Ty`S`;e3;
zIrCYr<Li~KA3!uUd#jWnqOlvsrsH!ua`28md~PK1!(1b0>13j|<C$E&4@xRiD0(Rg
z*Lb->zZFDNXD6*+O*FRUGLN-HA^++3*Aq4G$vZBUsA+hO!@m>Fc(-@xHlj)Q=d{^L
z)VTD}f;~i26LMVNPc*i0`F9Qyh4c;sj}SGFpS&@hs3~Sg&XYtlJRa3MMKoz$$&|B1
zjeT0ApC_7HI@x%cXzaOl?XMAqnO7IzBx?Ss;O#p^O@;;)9}vyhK6vnBqDe!RZh1!3
zSm~6HPDl2?<<~en(bylWPxc}T?GjJ=5H*L+3Cly&bm(BGyhJm`WM7t_Xj1KR_XCL<
zpSQ0XN;LJCNnaKs8rx(0jv_>%#KSz{7R~x#JNoK*NG@OG3Rmgz%=rGpG4ondn5Khl
zkHf|KD9wF3v-4wNf3mFo$#{I>@rBQ8$aSilKd-@d0NVj6+W}hZlj`ZpN&w!^zIs1v
zeSg^5@rkOJqr~$LxcmE8tly{7+P_NcaYAdlGFD7~=UtMxq%M00RwIh@LE(H*_<V%)
z&j<BsK9uzL#~zq#9!3<OkKpr>XiFom_H~dY4djE@)2>}ltKC2RvJ~B?9NjIysXwf$
z*oI^}w#B)@@s`Z{V}Fa3{Vk?e`Rm>yw$ic=d(tOEE0(s556HV>a1_zhooBX`wOl{o
zIQauJE!R&CTL#W1YW5n_Y93M3#`SX+63rNJ?b2eRI6ef&hkQ(`Nm6u^B`t^=FMV~p
z6;W(QZQcWQ@O+q9_Tq!rhp%2AqA&7Zf1tne`kVAyYr{BVva$ELW)p~}mfAS&JEF0f
zuB{h}<;G^YN%mg3gYyZ_L+RDI4!J%x9;vXbF45F)I^M5GH1@rzRT~k7Y`eZ}Le#wG
z@s5v((l!qzj@y=I+%~q8*iO1*{{_~2I_f=^10CgHZl==vby*caEBE!P*Jt!|Za(5c
zZ0@F)11w%_85SK(#QeqTes2&}Yh4C>k4c&Lm@GwZ`Dmp*u^mtNGI!d|ANG@XPTXuQ
zWDVD2t(3F1s90-nSW7o+>DAmkp;-l>T0K`geu3lTq^_@0hlkg}&%PW*Or>o;Y`gO$
zOY}U0_IgBn`tW^%%xa5BAsH*;`vmwt0eNnS?-SsBOQbzW4P11%G|||!^iV3^Cs1Zw
z@<g9bFV9#hy;ItJk&>Lfb^BiNJyYr@j!0>f5FV2)Mi-N9L`vh77@g(cgggm(6Y|Hv
zu#|)XmhZ*M2HEO8?l3$hAv9UnCOT!*$d`ZiNY3_oO8xrv>pmx+7yrZniC|Rq#~ofq
zXZPQ1Ss%L_i21=9=nEASfOhvmYqVRom@{-h0tfzXA|wb5gFpw+fqy#i_u)nU_s<{j
z|L6cZfDWJo=zxk2;Q2{K^D$~#9iZ*7pe;Shm7DwKTse<1CO2`w%?>6(*%h?0o75M~
zCOUu)pabXtI)Dz4X$SuKK||01bO0Sd2hag@03ARF6mj66mtZ)eLJ`ZQh=kr^qc>$z
z_>N&P>%c!NCbOBqgrfs2IPlMd4KFKLC^2_QBi2lJDQV0nI)DzK1Ly!cpg{*%zK~%#
z?Ha5~s$}W)b>Q{Cs*+%g)Hw&R=$!Kjw?YS0Z~zZa6->t{-OvFyIMUo8i*8tPna?r<
z)slfEGG7fa1L%PC4zP9PvXxrtOAAA!?mZiuDrxhT{sLnz!OmMsN!BPyhV#DRmgoRF
zfDX{&z`q*MfBk`$l+u+Bw27|rz|^1v=m0u^4xj_*06O4C4p>$hBFhTtVm$u2tk8&*
zPg9B~2lj~T)-EnFqHXtH@ev)nwTtf`pBT|5E-^kOIalj$@%>uAZ~0rwgp`E1<m}e}
ze?fNiY!}x%F6G}V6M2wjipa7=4J!P-EK$NouP#c&Sr#R-M4mITcie#19owZOd-m?w
zl|-4nOI(}yE-CS3gJjQcapX_+`wX`%fz&oR=)V%$E3R7yOG+DC^dG59=u<yAI|;m3
z@01uw_Dt;Ey+;aJG0F13_#P=S*<J)1ZwWcsvvuzQJyMdrE!k<`%MxX>ck8b4z2jcw
z?^BYdxZb^cC8m&NnsT&m-zBcYi~kmg>(Zrrzr@z<`t<11v29%M_|`8{o|rN+u}_<1
zuiCY$HEG(o)`uwxIZ6y6yB1IIPRL#SH48Zv8dd$u%xv&1S|^Hp6?7mM?^V+Q&7{4j
z11UK`!V^N}Mmms!-k<0|2`;tM0Zk-y)B!C#O4NbWn7+_~oKz#>v(Dy0I*^lhhUq|S
z6d$Prv6LYDDn+4DI*^hrqjexP_a^B;3*0C9aE9SO=s*r~F4ln@^k1d}B?wxr17}#Y
zRtL`VBiU9AO@7k>@~Q{Th$pUaXp0VnGJdNL<X{2`*Hu=f=>XBD2X7tJfs|w()&ULt
zd_o6u^V1m}h+)t@9mqjuP7k1+aseKI$mhWy1w4RdJ`M5!r5R!L0HI_r>;djTLYN0=
znGYg7z!CMxHqOv2(gPfEJIVvJPFgt+pn+4hJU}b-ZQub2dO&j>=tt@(f*y#e16Lb+
zfK>EtM%oEP77yyRCf7nLQnsT9NX=WFJODutXsH7^lRSV{dL?@RORS#g0W`9nq>B}b
zeCGkK;6Kv?q~_^b53t9f%^u*+wEM#Y5IQ}`b=?D$qW41&ppkJ;JwU7U(|H2o3Fp0(
zWnDf`P@7&AJV7oB*6{?nNNDE?a#6RdCx~I(=bnJP^x*3Wo}d(Q6Fos~KA7PNi1QvK
z&G7_QDO=nNke41DZ0-f_NQ((xKodh!yg*JKeD4JadO&j>@SN@iEaJV~3lQ|cl{#>Q
z)FC-IM#6Q5OtPJvoFn0j<viI}D9P)*fELD*w7iZV?s<V4Z1D01M_eM?Al`6=_&_ti
z8@<6E9gBN|JsNcI2IN%_8WK-jC9$VBAZ&RMFx(qxC*eD9kb_){y+KJrR(XRMYVYv|
zR#<w)8?;9E$KK#7F~PY&Zl)H^1xiun-CW=*KgQ$&#B~o&evk`<Fk@UUK%Dnr)Q()>
z3}2J&TxCuUAJ8hZ^ZI}sOpEjZwRsxt1FvJl=RTk{-W}-!N|83k2Uw^4cppI2@!;_U
zACQxn1wP;`^S1kd)O6qN1MWz*3qBx*1&@6I@qs7L)#rkLZjcH?<J_PXE)C2L_P9)b
zQ-fK<a)UGcK(-T0sp+|a2KLU#4bBj^BRAM%|DN1{yz0RxnYqCcMegPXxtQ@dHz2uT
zTQ*<W8SUu{TBnbnFVIBZ0AG-rDIvaqy!7C-(HEp-;oH8z3g6ZA1(t}b;|o&5xAp~R
zX*<dn?9q<=#sc2SzQ77KSNVcF^21tR5X;T&zTheur+wjdoF>0Z#TgP7;+zM0&yi~Z
zJIuP~3kcgDJbFV9V(4B(4{9-^svgwje0M!a$%;XGkb@J`^&pg@W<A&gX6XSz4`{9f
z<rnBd4t`#!2eruaqaM`4v|JDF&id7Q5KEI^^nlRkK~qvLa`4lydQg&{oAjV1e%?yj
zSp+?xxDHI(rUyBgzg-UqdO&j>=(|%72zv0kI`Djt9<)HlK0UY_v(xmTCL2loG*Ife
z9%$q13_XZp$|*g_!6p)}Tnsp;2c?)q;!>N&=k<WR>Or-OdJsdOD|%poX4mwfRZ85{
z1EQ=4eZvgk&iqXB&k}u$k@C_AgGfFQ?RZf3O#>*!+axYI*;CR0a#HRs10V)yT}pY@
zk~oq7_8_XX0pudBi~$gBdJtRA00??Oa~-%^(Etc~Kyw|~Thjon(X@^MXd?4H10e3>
zL9Q4BAar=J@&f}X#d;E#+_bN604;Ec<l}YhX+Zir5Y;{48XACA$~7f@S{k7L#|A)N
z??I`S22h)6Z4BTnt2-D#ExLYf0J*q1%K!))9^{;307P03s?IY2qJ0m>tTKSoboj*p
zQoye<0HQ4qPXBHI1U;a+4tQl6Kx#IeHvkI^zh?l%c@I`RF#saJ2PM7y0AbgI$9g}I
zgX4Mp08!3^-F|+6pa(S9fwOu2KnZ>d@B;)ra8w6!h4_IO&W4kEECXq8`hmMLqPia-
zuX^x(Xg&~2>F|8O3X{v{14Nhy)!xqsO5n9LA0UoB_<CPHV1+Yx^MP3Q>ixkU-QMsA
zrP){3A82PuC4WF(_24Js34$KbTnDaH_XpaT(bONL;z+DNAo6;!tc^do!qiUwAclbc
z{-87y2l#`M>^J)Z>(rX#4+vcz)L8EiLWxNA2Sj-fs$I_yG&A%@en8~+;O4#j;D|r`
z1Ar#p4haCQaHDJhu*Uo<0YC$9Z~-8MKN^sFERYYH1%OcM#sz>@c=}BM&`ADS0U(sa
zn*)G0w(bi6nz;%EKx*D8S^%VGV6Oro1fR_XKx&Np3jp%p9(<c#0NjBeuNDBrIS;bm
zDgavG&j$qnQO<*fx<EkW_h1q6nHG$OKoH8-h(HjFUxh#rOHHyb3p}h62uhJyGZ5t9
zK&?QKlC<{&L2jDY3k0QT(I5~Ic|AyP6bQ6(wOt?}>Ur>>XCO#PwLyV^y!4>%Z-F3`
zz#V}gC-Hj&0YMLFt^;5F83@R$9{iFS2vTzQRv<{hfYL$W46*M70jrd)7zBv6Jm^_B
z2-M~lNrT)x{v-&bVp@C<C_#hIf<P=iz6b&-7&tZv5OqB`IwJ_|v1U;aAmThI{ZkOQ
z!lQLTAP0Y?27y$RI}!wHQ~Y5NAm{<jbznx0V4$65`GP?%Disa}p)@EF3{p|-tzZz#
zpA~{ZEcZVQ1}T`-Di~<sdaq#68uv(^5OE&-IwTn6V#avVW>U!UCBcB82SVz=<ki7|
z$m_weeZinM-Zz6m2tywR1MO^h5)5*Y%_{_y=2ET@keiFSL%<n^`Gx@Eng{drA)pl9
zBSOFxyvv7xTCA)R0?typY6!4I-FHL4UCI7_2q4OP(6Uho$U*FfA)o{;+miGnqCNPf
zR|rT!-rgY~2M>}$Kq^XqNzMnz3lFwT4FRe6d}au!fpJL)xC3*33IRkt4}4aK086xA
z8v<G)Xk7@%NiYdtE=q0;0Xc}=6#~e=J-EF)1hmM310mpU{C+S55P3Z4dprbOrQ)d&
zK+pqM>%iA%LI6PzXs!dLu7-e?*mffX<RJ522(V6Z2nALcM$#%JLvn<ID~$FD1-VHk
zacGgh{6axXtPTnVDOpw|6tqlO`B0#d_SHf`DlXRu1%wU{(rSi+)GTQf3dl<j<~I%n
zxv0`I6cG3FVC35cfmU|aD+p3BE4UCK=mE`jU`cEt&@$$Ug}@c^Y%T<da~|l^3V~KI
zkh~!3dC=uTA)uMYw+jJ+9>`q>dOR)!YLl)rf|TqGG6I4gNL>d8)ii=TlhM=&YLZFf
zcNSxJBdA5@7?MsES-IZ`tS~ZXVPKg8c?$y~%!5GU8KQj;CdU^BSKvAp23I+-w=f{~
zJ^1r<VPJtWQANNRI(}FL<YZ{yA|RH$lZt>`WIJ30q$2HE5kSxbn(M%ciba8T0=c4q
z(Bnb=rbWS7W`0r>XrX)Wq97M1#*pI(q27avQ;Gt@t_ShcivqLEAZd4!?Q@C(qOJ$J
zXGH-~$AcU>!ayyqdxrrH>@kD^>pYDL1F32Lb{HUkdoZPP7|_ID31L7x&HEGoBw*f<
zFhG>^V9>ZQ&@zKb{#c{^)G&~ed*6qFtE`zF1~lRGV;D%mj$gxo6%zjp1F?+A3<HE7
z4?e#b24aZ39tJGXPG1a!5>NI)ocEyh8^u5<v1A{y<a@UmAo6)|ty(cqn-evQ0gdcy
zTMVS2&e&q01b*X-fjcvDPBCzWIX@QzL^%&O@x?%Do@_1#?oQof#X!s4zg7&y(D!~Z
zK>qfiqIWn*O=S`$;=BjR#4n}jkt-aerY8v#QO<+(H^PA?nuUb}Ei8U39OR^1<#14f
zwN1kT`P+lIX5l~+b-IUxl&tR;4$jbgXgJ7C{&C^p3WvMC35ff6&~eb4paf;Fyb12W
zDqV4qiw?%(;3}rZ#X&BTnw0`M==Mn|K+uDK)`9z@N`aj8+g=LfAn7DIMi}Q?b7^ol
z_BV<EcP5SG-<{doJOXH-{GbSMmAlI#06`C|b)eXW2%wQa(;|QsLJma$?Tk1Q0df*{
zH3D4a(ZdKpT=$@oZzRyn_<%@2&;y$5z`meJpoMqKM}o5)Y7hyC>mKZA9tj9L9_%FL
zg}Ao|y9Pu8;@E@F21kNaOk5oauF(INNT3~`&5?l6<-xofQ6Lr9YDNKq9?)C|GGe1Z
zO&WKK0<H5`mncvJ?}1T(pa(S9fidHw0QuX4kG_urTF5yg3Rog(O%!OAV`))<pa(S7
zfm%1B08z(-E_b4Ubq3vw0(WM5j%ZK<LrgTd!k&+#K?yp=M+5S24_3F21|>0fjs|yQ
z%jeOcHT=Jd2D#`uDjMV<Xiphnnf8aufGey$TL$FdAPLtQB65`l<n<mz6OW{%W`VMR
zpa(S9fuNFQfi)&Jg_oa{-YIPo!ebyNTTHeQDUDNNbe4Y;@+9O<$R7j4QW6SSz8AOr
zT<>v*;VB8B$-4FxQbvt@`Dc#=WBvN|>pmx+7yrZn32#*Oqnt-p`0`85?7a$>o3)jb
zET)!~LsF|h;Xom+bG*#@Ln8-sTGKUC7w*lWKg7wcl9jv2fb=aI{?)%pDsN9k220Iu
zkB&A>Em+|Q=d7xo@Kv9Vx|VP5ab7AFnxEEr7x@P1FL11C+4?0*7a9AIb5@=8r~ibv
zKTgkcn^Q__j8B<#BYYRHmOAT;tJ~#RIQ+?-GE+FQ>T3K8)-}z1b3_KOmRi%_?5zpu
zYr;?S?xuqS3auCze1TV6t>r+wmQUtSE%7_W+bZqp#`H<!K-+Jkrg?hJ<Atg;vHv@i
zCFwIJ^qP8)mrK(<QjMv82YGvHO;KR~ur(#?9p%L8{6fALKX`uyC$(DBhbpZD5yM(8
z@LhY6XH6$`th76=;aT31O6!(-p^3eZQ7gqyf0z)oJNytYRHaVu`%~`kA2(<dua-)k
zA6mj|G;&Xyc1wTa#VRfNu%_c$+STEMGXCON)!q4PuS`9#mDieD%g0s?=SKdVZSDnL
zZnce9=yylwC+qJ*!?kXI;`Ws3Ge5b^E3MY}#i~^O1Y+Z7wO>xEbl2m@-r02IGN-m~
zzBeNJ#GKO{>#ElEg;en>-Crj5xW)Zm??Br8Q|m8&c8-@?Umq*}N{a&<nrxjjq2?uC
zth$<y`5&#UkoL_fj#ZuBu+j6V3YR!n)t--1oxha|$6kLS?SH>nd)4Z@-<;>Qs@0G7
z+t~I-CNG!T*54k}?$2u8;|}kvHY*BK7Vf;u)2<s!uU))-|1F+2we?@E?{^9<+)y&~
z1}~SYl`qlpN;Dqq&#@0~yw~Om$Ew;sH&yz(=<*Y}WAFGa;oWPV<<!!&GQnee?K;V^
zrmIb>{M#&E!P~0HV!+)La)8zAcd65D&u*dP258&9`JLF-dmr88gsMG1)cX4t{h7~<
z1NZT)>ZA>;?&chq&bz8=`H%6U15!9ZTl*+BJ!+4?6dJ!=_jO8#d-MCKjJNik;9XTI
z>N1AGjsxS4%-FPG!DgN{EqqWHn(b^R$GX1zOrJUN{e7IftK3J7TJMF(oiDNBiyhyr
z^nTMGA3F5t!5^H~RPOsVd-cNJ`<Y7reRPc=L;YRbIk~jS>RpSf_e<xLQnBYxd*vkc
z-&sFTj(qygx%6z;IMy|I&&<=Y$=5j6bkE45m2=eF#G$I|!}8f(0!MT^$FZv3+jf>J
z*l#oEth#zX>#qAteg1GZyx4v*YrAJZ&ud+s^@FzUlEQ11n|5G7ue2)dgY~@vt*cbN
ze%#|i(!M3vd7&y*{I=ofQajxKqR<{rDzy#Y+3!^gwzzzF&3;a(YUR8WyFTLV>Ah*3
zRy8vHNLr1DfARLzT2F!xl?m&4;StB0sx@9mY`ndi&YWF8`8dbAN<9uksqm!gXR{tR
zs4pFmyT6at@a>g<h2A;d_~Pkfj#gFcd{@f&+vC-nE^%tBy>gT5?^|1xI9YbuDV|lW
z7Zx=uXx}BCRdqCOLauU`E8IVxv^-yE?Vr4}Dpxv$hI^#q7u)g{J<O@4Leop@_G0Jd
zaCX0*^)hOvSDVYZsx8+K4!`oVuSqUlIA@yiey_)z)Jki)ThnDQmO0U*dv=qpK3r4$
zgRa@UOsuNjw`TnB<vmTPFMY6<w^}OKya7V<Bldn<t^J&xr$?v{N^Q?-|K7KCKY#yP
zkyMU#b+-OE+D@tZVjEBC^&Sttg<pBrG^Bcmyhke^=3P}=Kb-yU?DafuC+z-*7QdXW
z2e=(NAjW}p?XJAJ^2q^SZWTK3r@igP<PrDz+kx2HU-(qIw1i_-3-nE|?0;e?&#GoV
zdUBXscA95Z<+{I9I$THB!L<K=vDBMCzRyqPSyfl(Dg3)dQ}=%pXP<d!3dv^Dn)Y43
zL_hJ(XPi})a#`!+(=dO|&C`!Qz22vpt8V=`BJv_ni;iw{WPJ3)E4<vQR6ArF&$bW0
zn^x3xmlvwG;VAuGsrS8{jYq2df4KE#>9&z)IA>L=cv%aFmiB=)-)z%cVeQ-XM^z6p
zDW$gYOZ}a-_fcH!U#DgN-i1y-dHmTiUM+RC9?hxP`1r;q>v*lG&~-pv^^dmw?e(+L
z`ggy|p2y0cJ;|fK>VsPK%C;Xz>U6N$zoqqbt2ItWs(3RewkUaV%^8kWb@e!7b)IQQ
z=@$B2#p<Kb{iNDn=)8i~^wOf`3Ora=uFWw{Dy1b~N~Pn+z(qeVzWE!+nmQUEqm+Nv
zcxuh6RlEP@yBzDP^!jU7|4`|4z}|b<O8?#Jj@Wsrn<eQzZ^R$vT~%v3Xw~7rW^h|R
z%ivhoN=?VK-;{ZSV@(}hA4>f@t@RMo#kK=VJ^z;4CH;`|x*wNZ<Xlx(=ev+PUt+^!
z{XC%bJg8lM_n2Dj>}8%66<aT)-p@FG>*68nbDPvsSMwv+^DwFUY19g*(6ko2KcD^!
zM~k{z&PwecII?b)nCz!GXH}_m*~X(*y>c~PxyE_f`Xl!Hf|Z+Q7kKLqFO`=6s+4*9
zsyjTZO6&7nN9l6Z&&bN}j@G>uThGPrw_-x=iS1XP;G|Nq>5zLrrTWQ)<|7~f`7h<7
zZ}O;g`tZZpb1T;Uc9XZKLd)CLb-9Jc$LjV%=ar>(d#UD)v=@%j>(0;`pR@Jc)&3-*
z`#YO1-0r^}kZauff5+zduMxvqF7RD@k!M9u+@3Oh<|mhVvFg|nb+?$YB`|Wq`83{<
zs+-Dx$KE=)YJ(D;?r>`9zf**Nd7T4~$9Mhi$8J}6+LV=c$TpofYp+^;_nY&)v--NS
zQS&!P{T5RFE_WVTdk;t}=@ol@BX@mp6uwl?GabdF^zW?KE5o^wKWCeJfmd2xEvE(R
znr6N^B7>)0X<yH>y(x(M>exL_tU4-3+RMk;^exKv<LIjTYdou3G0lJE?~||bLRD=0
z=BQrC{d=9`W!4`WIheDj*7${v`*wC6tw%>2rWUMlgrimEn#axRtJIpa$<?_>pKRw@
zS681qO6@C#_o@Sq-lq}z-W6@z4{E79aqY|j&Q*QYHOJrj>}>zXmda;%T2-!bWJ1Gr
z?4)a3IJ+)~QrkJZZar<sk<j$mo9EuO(gX-DQ|+5yzP!o{RafJ+-QUqT6??zihQsQ2
z+jRg~&4*g)u9>}8!E&><a_*wG^ItgoUGDM`x^6LT&o`t_kE7>=wEpgFeWHE4^2z+E
zC4Q%Pt*EQVue0+OTDP;tCwIH*>Nqv2_qXQTR+oF}x15O2<XO>*rz-iJZFqzis<!#Y
z?`%2X?E5Fy?W!dn_V;~wj<ct7k7suEy%4GQ|MGg?-<Hn1!D~gO_DAXP8a^$$uzri%
zyj&WX`SJX3&9`~fmp*jr8?yX4+~Qf)jNN1A9?ZPWE3MYEgwXyS>-KA=F5H_#e~1&R
z<2|Po8u&>j=SZ)e%9m^A<O`ftm7e#9miVmYM61fx&+RC`)^#_urfQ9Q6&vr9(z`D<
z>35ZrOC8mN)uvVcZ5FTK9jVZHnH<Hd)bF(STR2Le((6{q9p8o~Tjxxud5LFD<sQ$Y
z)_QY;0}8De7<_>js!|`nw2jx0I)C=Us}~kED`?*(9#xnR_WYFkzAkvr%+s;S*ErVo
z{OKBb^Gturv8vYl7*<v8es#ISmwWwQrNXz~f3v2J?)Sz1Zm%7+{d>-~Gn&Lsz0Iku
zw&@c3-O=?n{(gV|xIvqEseMqoA6e@7Y~M4K`@6I6FS^TiQkSQ@hbOe%lxq9mu7^FS
zQtvWHi+Y{!&?jMgHj}g3JvpslrC%R&w5YT1nb5jjj)lXY+$l4KV_nldQjMv82YJ?1
z=)5zw+e_6h*oGtbI0U)EmwH^G(C}=1wEZsCIy!QPbMg0`Et(9w#j&Dg>z6EDWb8vu
zsD7KfRNUxQH#wndTdq=lXQlm;D!)pFBlq|srSgICDC<C2V!eA?=Vmi0rE=#^?7Sq7
z_LDo^o>TYli`jaHS6c1mi`MrEgr+NP{;BmBKRd@eQhPo~{dcXetEzPVq?)J7*>I$;
z53Yvi@lIsXyL-=atn0QKJ%T(&KH=n6xyJn}6>j>xWvd4Cy~(?q+B;61T^~Vg`AGe|
z-`Vuiy4~2yyQa1uc9&;W3;Mh>=}Mk^JgsW4o=f%j8hHy}Kd@G3Qc8z3+@5jitfz@J
zJ$$Es+Z<nbnv_<l%GusL548SW?(vV-`NJHILze1yR-gX~eeYhX?WL;s)^P6T`Lc4Z
zcK10)TA@qg@h-EjbF`_W`JZV0-TJ;a>$>^gi0Bh@PIIiN+~XFl#}z7F4oq*@=y_Cy
zOB}08mz%;<BlP(xE!)%jKCZO{%Dq|eosWV~b5biU<3OzWL`4VGbl_l*pQ}Hse~*(}
zZS&~r_;^;v5m?hHbpB1-?OmO(SE~K3_tC7WRQabhoUKDrv$emwlapGl$8FDg88y?Z
z&E?!h-+iXfocR7ejuw@wTwQH<h3;=VA1Q0!)p~DQs(6K#7j4_qI<8!3x`g%zZ#NJ6
z>iweEd9_sT{&?H)#jf+H^nP-Wvy?l0d+R~7^7oh{b=KUze1m6IrJi3@>F}B*={;}6
zALZ53=67OS?|pQW6RM?*zdc^P=@Q49mR|4i;9K~WcUI>Y^1b-M`zttC)q35JN|yst
zT^Fhyj@<9}$Q8b;#}zI6i(L<p_Wgv;BWG{>8sk$Y-3Z^sORYsK&e{Ia@v}UuDs=vD
zTDM;@q4vb~t50yO>YkBBE9a=UiBnp|o(%`>uRO}D(H@>weY!hfb8tm(ldD>9Tj6ql
zlzqUdrDDrR=zhz8eE!L#owqnvRjKh_-rIJTD%fu`Csu`)i@rs}zxp>x<yljy#$Bo%
z4r{*KORw5K+KXT4@nGHF(YO!VuWKEpFRe`Q*j~F%a;|FYyo-E;^cOf=wfd-*$~BKU
ztKq6u4))@u{kfjl^DnLY&8gV<_{Jvdcvm&0(U6?8syydd(@IUpwBM9@gOf__)oWJ&
z-ZWo&{!w|Ka`yDp{HJeuz3pL=OXc!S?)M%(Xq&sjcVBPdSkaQVQ)is@+sU)8D@Im*
zlIsZ1v8F=jqo8g3(@Sf8R;BVOUT&4EJ+OvL>wAsPrc3E@9)6WQkCi`rl9yTo`-iP5
zS??$(wd#kQ*ZsKUBIm4%?ccH1Ct9@5@iOZVjU3F0Ri)=+v5hY&yjHns2ln%<D(mb0
zN_~7Q72fDJN5)4#yu!Pxv>eB_>6lR`_Vn@P7kH`k<Lxc#{5b3$9~x=!oEUIdv_APb
zF}&>Ww|Q69HouO(yZXMc*!G^5{iHr$sukYo8bOBoyS8(zYN2T@c7Hzo7mhU*JKn|7
zap~&snRlle8*jSCYenxg>oq&6`3;^`eU^CVuN%!y@vLjjmD8I1cJUm~ik6<2<Le0@
z-r!l)ZDp>koj&gx&${xPKJLbCyu?eb^~Z1Ov?b;;PrJI>o;%wAFa3@+Avc%ttn2x!
z#<4lZF5?`jBR`b>y~>oYq9@ci#<Qx9=IN#V_py)LlzX~<CnuFUsy`Qcy|c4js~jd)
zweoMt*FMdW$)T#_!(RuD=g(Hl;FVHG_1V$yjYjTi({AZcJS%E%e3jPzZk^|X6;=AZ
zIjQ3l`aZFBdwa)M;pA0EYi+*I)28;)>+0WGsi(H-lzaRGhQH|N#W#QB?D-(}Ih?)y
zVrzflQ|ZzYPAl5MXKiBNn-_UjG<C)o*(R;M%%h?YSP$4e4qUw+5WD}V{MQ%x1l;DO
zR<Y@^?#Funo4xcWAKI07SG!X@?fToz`}2dpJ;}4G-{k1B@8ju5c~*64sf*FWyPV}&
z)#Y=!?d_Hy<E2t-yFja|z0Umo4?oLseB)od(%PzY%BNkv$>eEKSI1k?wttP<xkl&d
zdXBfJQXdbC8YeaVb;}-(b#3BbxIjIx-*|V^5wTqdT>Rl7XRA`z6QSW<ivF}1=x%dT
z>G$TgpRP5H;aS%LrM3?)`0ENzs0#I8sqN?H>2NoHkK4SnT4wj@0oOL&;%QOCnUBK$
zDu07#T}KXI{`}bU%e<>9RlQl_m(YLv(R;kw>S}&wee}k@_FvrOrP4<$3iOZmy3Nb2
z(XARTtzP>UFSW9IoJ6g;ldan5o4l*)YJIZqUv<RjXDM|~aIEV)C2pR5yVW7yRkiwJ
zPe(tTH#o6#1~0YRix2<q9u7pNA4#k6@Go9&)ynfSNAuFK@_UojyB1aNm(J0qQXh4!
z%^z$py{yiAXq#?Ujt_h3WMv$iy>#Yma&_*}C);^e)!sM{*8bkB>CD;nlaF(9snB-R
z-n_c3{9SB6wosqMZZC8kyx4fG`#C$01ufgzj+<lcJFW93JDV=+cvGp>S|7!#tM4V-
zi+5{_5+}<}JH<<__Tp3e?{Xj4O6N!E_?o>nA$?8wY2IlSxz38*<s^6buFhxZXn)VC
z`}f6cJ;ST5tj#ZAFTJ#_7iTX$tX<DLO0V4Oh}gD|5yM(8@LhY6ms)M(QTluNuS%Jx
zue!streiDbn%aKYU7mGi&39{hdrZ4Ot9g$*yiirTeq(&EIuJOrZk3qqr#RZR)~s5!
z`)|I>v8qzd%dB)bom;iJy7$TzPAT2cWb2#>H81h5>Z`Tf`EMWhxRA7O$#tGKow@4P
zk0T;4@=~euj>x;Ez4rj!_3!e!{deeEd0s3wAAdXX`v2K`P9@dfJ>H2ddUx+xPHG+g
zYX-OFvkcBzm3#e5doD}8zO%h}<o-LW>lbP*-HyuH-a1P5{w}tCadv%T>vm$-EtIPN
ze(R9bZ0+yv<m6IE<zoF^>h^%O@T@*qRj%^TYPeS4SXZ^yledkRbsufhpHw>M$?tM6
z;e@JfI8uGDur&JD)RH$i+O*$C-s`_R{fuK(M|Ix*^YB5}IJH%(dLmain+JQw{P{x$
z&#KO=|9i;ixiWaARIYNc4R^j@n!eT6!#u0{e9*6piy9B}tZUB>lMjC|@(3@t+Lpi4
z-xr_QS8!s^>pUxZ;`Ws3Ge5b^%cV;Bzj|uk_-T*taI9&LyFQ~=zHyslRj)Ug_eb6Q
z7dh7T_Z@wYuh#A0gzCLMLw}$9)n(3}maRSMvR|329P2u0OPL#sd;Z0-s>L5yYUtrP
zg=bBrny1IvaO}0$Ij450xccaAj&;3wu>RUL?e}v+)t)a>|GjMelBJ7`eaI=TQs)E1
zLkBDlMBTpDb4|HRysN6Tn@JVk>{0u~ADmb%xVctj-=L?QQ2n)U`3JY}+~wp}rTt_z
zzE3xu`t#U~Y$m0ZRlhLZ=)gZ6FqfQtdePNw9I7xM{^_BZ|2ukLg_Ylz%$kz0u;lL?
zt2*9uN}++DWOAs&e6X$awd<Z-KQm|#Pm9{7&(-fzje}D?oW6J4%;wMXIVe}D-#4;`
z=W0D*W&c}6(%-K)>H;rT)z0UQGO(v>wKSep?N;tV5Z@q!*NUqB-mK90zp2sc&%UQl
z@vQ0Y<Q-Et9z4pkuAk1TF)ruWpLtfb+VV1$6DQo}Tvb=w(WB8jFRpu5z{INlIy%bG
zsFuORs<vD|IQ+`bz9zX<Y&{8^Uv9j|$C<p`YE9qB7>_TjF3e^UtM#5;3NF#>0Vh;1
z6zsmF@y?r^Rjui<YGh2C3KR7=d9i8@N31qVe1B!Gw=;NDVLpiU+jc)w_L?zqpP%Ej
zqQ~<8(7aRV5l$=mLEGFFzWaIuN2^Xgw*02~!{I#Z8uj+0FUD0}!Aq^S#~Z8Phc6n_
zWoVAeyjYbg-=XDa*P7ew6t9*#%7>x#U$rtWo#2&LN9l6*_vd#TC-={{j8|KoO|QM}
z;yb@v&dacY7pnHc)9T;bM<<MXQh7Jen#R7peCL_{zjLf==RZokQG41!o>l#I@ekRu
z=St(1Qn`-56HmTgd*|d8&XMMwTz${)>GL_EdTD85J-^sA&RLzj?sTn^ADrie>ND?~
zpD*rljdNBz-Pl*OXpbwLP@SJ^+Jf;Du5+lb_d#fV{cO&!tGS!|IjPhd|LBzLe}wiv
z#W}0%m%bO5`t2{g(rQhwSPi`SSp)A0S2?LwXnNu!CuS6UaDjKE4_CkE(WK91o)!J!
z)0)*bd@zS+Rju<Vu&UnkCl-o+aGYmNtA9Q;zG4E;v97D1mmTWu{~O1uvihFolIKT`
zx89k~v8sPOX?ecT+CO=#+MZ9i_4jhlYx#f|t10^`HtE{)0q?4QJ8bCvj*sv0N~vdy
z-U~C|y3dPMTOS>LZ&H5M(CDBioZ9LropSvy)pez-;W(SW@hcX$T)cEQCzU!IuT=9V
zTElTRUa{K=onKyTJhbnZ6ke;`v;+Hjp(<6rGAFhud2!7dPHQUnyy?z{FI7279nPbp
z4O0tNIKr{2uFemq_Wt(PFOcf*V(W`s^G-M$uGn<5wqMMVI%{rUzQMby^FLZyA?=$}
zoU`h^ZD*;1{Wf!`?&(8F^$vNDRzA#YP5bO@|HqcfXLwY3e2}_*oz%bF&|4k-OsuKi
zw`TnB<vmSuY3tys4N7#n!%3yRyMCN7*nERiOIzn%<Qt^Fz=>6<(`~KCHO8k*x)HvM
zr(NZ0&#XSMrc$pnDOGrb0}8De7<_@3T3IU}YkHM_zgDjGkR63Pp<|`pX${ZvN+~Pt
zfunWp<o-RaOz_xVyH0XiQC(_>+g}vg!?CJj*BfwkKWlr0Dz5`l^;;<wj@bSWd0lip
z^abmhX1+NhgGYVU2iKi_8NzZe{gxB)nLI0+HviQ6i=Un2Syflp%d1+j#pT0m_H$Bc
z`DFgo62DVCt7<qm^5<-GFYv6Y(C5bPw!KvCoL0jT`kY<p_Ez7Nx*v2lo~J)d2-+Qf
zh-XzJ(~qRpc=#93n#T5vS=&APd0r{C)<2;-V&<V86KfbuQmfE((6)V}t51tX%znhF
ztzvzt7Zx=uXx}AXE86U>3F&LXPxGv+tJk&G{l(S`XZPz_FQaC9wYi*7wHLm#;}q;|
zS1s|dzwg6yoLm~aKH}`@y=k0K-P)qW$+FW<@vNw|{;;an`TgWld-d1R-#w~sYM1@3
zn;fgk`ugOopBHH9w{uj_rT)F*sY*U)8y?|VRjJO8Qu|tHT8rJEPydBuRnt9Ejj4VI
zd8JkAa&Q!$)YlD;;uHG&&DyJ0-~Hx1FSj~MSGfWYmX&LBjH5*zrB~?hN{^e6Dn3@<
zBe161(KrvOeivFFXt^G6HXTCyOKIJH(uP%cbB;^rg_f7t?Nb^J$vLaabB<M=QpT&%
zxv)DNt17nKSlw@Qn<L|+A70^EQ@QT<l?vCdvgfh#XHW7{sjKxbXOpXQk3QMXyQ&`V
zL>9ff_bjKjE}z{ca74Fr9P3)3Z+d0_6GM4cRjT@p;kedbIQx89Yuio#b4lF$n><al
zDJ}P{uBJoZqTyfto22qe>9=)1fB#yMRL)gZeLbdy!=KzKGldhI4ypTzln#&9`#opV
z#rpFRq3aCVUSICcUwdWhfvvo&YOlY1i~h`K#)12IXI1F?TSBiVq<T+U=yNG&$1U2H
zkI?UfTIx<*J9B_jN`<CR==RaA8ZNC~`xcM-dLNX2p5khJV&B(LdOxA{)w;b>?-N+#
zv(@*f+}}TL&?a7PwT&0Qi*rD!_FArZ_Kf@=duJY?Q}zGxS?meP7UIcHh=gq2CtEU+
zE!kp1j4dT1im?-wG?HvnmaJu0iSt;;He<;;q%rp0SR(81Oy>T+d78)F&b{}Xd*Abi
zx954zJ?C>?=X1`z_pF7hrE9QGvPw@(z1Z#VoR=l7nrBq~>S1Y*N@l59^VrOeSMBq%
z+Hm@6!GEt(``oBCTu_^8`Fu6+NM`HxuU>1DeSBK4sG4PT|JtiE{w49I|Me7ez7cg@
zm2CI&w?<!la@L{$jC)SGy?5`eo4-qTt2w0>IM%#-pD*(M>06Sh`u#v(xhbK4OHgY$
zu$ot+HlDy9%;uM0zMUJFF3VoYs47|Wcb(xs4j*%ch^CG17M|O%^;IIet}XNHg3UXx
z5UW+mFSP#7s^3nIC(!evR=f0u7j@mHY<J-*GdI6^=r|Eo)%M5o8qVna4!f@pRR65{
zUT?kS^}42;lFD>=?wDk@D&^noehUBimF$s!XSM$(t9VvjDsr=WgL9ISzSuk6)1TbE
zEjdM7T?=m)cPmK}MMa(e3cJOgzif5+$?c@S)|i)J+|434h+5tLQLnKBf4V`Gv`G0U
zM!quWCNZi79;-O<eAd5+zn-|#Wx{iB{6&;>Vf?hOdVTvB(W_s*{llkguiYS8RV`1o
z;rz?ZpNH9>N2T3T{M=i&B(t=%XVvW$>&{9}(?P#vt5<Q*b;)jhvHJ&yx_@^`QqnAg
zPle21ct)~H{r%eN&!g^EJQ*K+QG!0qK`w{)vQ3t1p3LYKRZEv>SR<yu4M|k({m#Z>
zJ!>DAL|3!M<IRd!_J6bV)fev`DN`z~S2S(E_}Zx>J3_rmTHtn;9NXW#E7_&JPF#q&
z`a)i>zn058?d}Wf^Ld@3M=t$vD0{WsUQsm5jyZi|8b0ZDiq^{h%%$DGhImEQ8?&qK
zAzRXUMb#Rw{86mjr#Zc5tJ?d(;m>bguGjv&B)SHgN2r#A)^HbhT^KPi;ZMn^YE73_
z`1Bp7H4FdtswBEzJo;p&>EE1|MAbm+R7Ck+?>aNyt`qg#6RXa%kkirI{>bzGL{qKn
zxp)mPpSx~U$L*iCZqA$ygcmis?nmqOdi`JNx=O_hRD3gUqGY!6wr_<J7v6qXyHAzc
zaIEsT=e?h5-fQ@mB&sg`s_ukNEB=sJsmxXDe*R{ie`j^PcJIe7^EXR9hbO%Cg!hUh
zsyb!-X8!QPU8}dgEZL=I`Il0>f%cCV<$G59O<Em~+0RF2$1nQ0n9_K>(chKoooK(G
zDP4U2tK?gvRsHj+bv{s<@ijiL*7XKat6Jv^X$>D}{?(G80YhUm{!O&%=he1-+iCc9
zqE@?pm~YAbpesbL4!x9p)uP##iCRtA{@JWOC$18$8b5pS@{c}DB1YBUf0RDU>9u(8
zbmWeOmx!o3FL?f#8d)w9)~dwM|M$0h{@kzK+*>4&`iE!zS?#)bz2S`7dwYD3ujjve
zo@iC0>8icnZvRnhr|99FvM;3GeqXKoc4-Z-wSSYh@M`<f^oHwQF!9dq{s##(bq;p-
zBe6<fZ~atzozZsK>V3pmr^o7CC8HeezFySz$FN(B_2X9ix7p`0>5bp2`5L_JkFRw;
zkdmJ)er`+u={JMCqN-KpZ&o_L-)VEI;ID^?s4D8YuKsqb-3LgiJ}Ldq+k7{p_WV%r
z@iVWzx>OQHtvW}*taMg*m*;4|U?~w@wT^?x;pP1Up`(5Zo^_g7#X+fE<L&pUEslF%
zj`~Y-ntqwN^xI>WCrP4eOs)5x3LSMv5>+=1_62vkzExtaywxAI_N>@-=+!G#u1i*{
zUGt1sr7LRvmF;G=E?upkSryJH_fxIoLw1KZt9)x7=)1PV<r4&|j)Pg{AX>au&u26{
zy&45tpWFA#PDwP?dhV7S-fVxXowiqjO5bggL{+o+h<47izr9+=FIjUm+8@Wto`E~7
zjn2?Gmhh@ppC`#WJ<;cx*fmbf%D?^NF?+s$p!iwekL7++URCWm6Gp>b>hWIpm4!mR
zPS-Q*x3+FJ{EFnS{{F4^bHA@%Xc3?Fs>E9T?a@1L&**jjb|`)R{_~Sde_NGAMAa(0
za*SI4%|)VC1MyHhAC7gn5+S`BbpPp^WR<e^d%NNr_UHM^<1XHi0%@Rf=Iwi<`&RIB
zz1H)wc?%z2_mjt}o?R$G&E=rC{!QDM=eaqjb`h<r^|@Z}ygj#Er?&izK2NAk2jQGe
zK<#s#)^Ju)V|`9@sYxYIK9e(zS4qvvpD6P>wf1}R5?<@RKD>rEdY+qE{;M|L@@*R}
zs#*66v0LSR1uKswM;AVEl!&TUjmN7kU#;QNe7kc;*rsDdt7=Whs_=UIF|^vJc3%lz
z!-+BvT**#W=Z|YGf7ap6>hGAdY~769&t4-=Q@#6O;JP}MrtIEH)N;_<p88+My1tmT
zFBfY&5g!Qrn?U1HvF$r9`jJ4*<?wIjTK<13u~zo1Q}0SzRVfXl<DX_<&)YuJyxSw6
zW3N_~{6p{WM*E}vum7_C$FE*2<ddvYv(EdpI^HGo@3)+|^@=2_?yTv3;;W&jB(qiR
z`fbs|>8-z7uUnEK%f>(6IYIQQ*7Ved$M~|j<I7Ix%WA$&`eqkq9k{cZh^AJJ)2OY_
zZVR{S#>y%U=H8I3)-0p@*It$JFUhEyU0zm?7y0K$l=0a>c9vNkdh^SmpNXh?>p&x4
zffx5nyr{o__<v?KE+A?={yemPp15ZD_4MQ7iCL=E9#-M~-*<WYthQf1P&lRWGNs?&
zs<I;O@xk{ct2DI6x{hg|O_Houqw7IguLrU#KBMax?Yf@zzAjeB$9n#p+W1(tqgvw$
zt@vEw>v6{=)~dGumG^L3_Y37UywP>YTK&o?;aNS0IE9B<`H<%w8dbCGL%bj$0s*(|
z)3ob{qpQ#KKS}heS!HR~KG**7nq|LJ<K6Xq-p_k&J0gj$+jG`k=G(oA=vAfi<@Nd1
z->%;}o-e-a&TgXhJ(sufC#~yC)rKFpf7-e^b2bpSs9G+8!pY}0veH25SoOV9J-6z2
ztJdMLPDfr(X7qi;uIsHT-zUk(D=U>xBD`zl;!n~QEqq3@TdgV=t?96?KX%3U$JI$w
zr?)&uL{+`>$9TC;RPKT7wtIes)_N?5*ZW>9+I4b%jcU_PNo6`bcT5spH~clS{2wzD
zC8MgAr(N$;gK`yIIw0?9!dkViw^j<TmZN_-UdF}z?b7=>eTQky!oR&L>D8A%*q7t+
zzfVeT)j;z~c>g|M&Cb~mRNF_KqVjv8+Io;Tc6ws?*o%^=nrBq~>S1Y*N~~2aKUU!a
zJ>QA)y;<}9tcusFb79O%=l1v>U(bK{JP}m`&3_i<dtUdQQfrr4{l2W)N8aPny050z
z{HT48(Hd{_$Si3tzjJ_Ct*;z<e)Zl_gCx;ZX`kNsP@l*5#2&({D!mV~{#}&)+4^kl
z^!vtY$B5{<G0U<3o&h_EX!`1;8#~7hTt`GzyUuIY+TQaT-mdW&r2AF;FU1l`AP%hW
z2Uf=?%0A>qx$s)Awc52gdf>2E3tu9lY4`P^-xfW7j)<<ZANrRYFY8(T?b7<$=za-W
z)AJ9{>$zQ4+vU%Nx6fsVx0&<!uV)C<S`MPOpR)a|_T^Qo7k0g`D%qv?cV6ZR>9s?Y
z{*FL)u4%Su=2x}P5>Zutei2pW_Cs`SP^8><PwhBML{qcI=K{s6)J}@J?~iP^(tAUo
z{2Se8gZ1@F`QZI`qwDLHt`Ag>ynX-6w{zptW!WnkRaW~dXN=iex<KCzk|_Gpy(}jt
zyt+dYRe7(cYVDh!XWmZ>l2%Egsa5+!vrebrl-Aonp1M*JUH#Xo+Wq;&gx00b{BlfE
ztD=o-D%mSayRX+yqx(JZ_WYn^Pw#?>cW(DTNX%B*4z=}0^l(<+=VZQCX`ucX(Z08P
zzQ57(Vzr-^-SITKTX=56)>nyWIy&d2d5`6~O03fR-QJ#-aO)a@KHy*#ceBzF&EM^^
z*JwK_Uq8Cn-JDGp?<Gd{+SymH=6~ULVwL8|+5ES~lg>+4Y1Ia|iYys@Tw<kTmk-MP
z>9tFe=-RQv*E>(G*h#EX*>COE>a{awcq`2@;#RerKhk#xflaMT9;q<pDAB8G`!D=y
z1{&`(>wEw71AQ)VJ4=r3Z{C%ZRPA{kz22BUqv6;?n)h>1I!_>w{Z9XWNJ!T$SGVmV
zP;DIi&y7~vS5R&LoZ9+tG@R1<OQ-ui@&^JU5m<Y()zM#1?<K7CwHD>SO#Ey&F<Z@^
z*XQ338~rT%M6Er;@9*37gjZA*ZC@byoOzjlAT@g5Q@TD-x^DTN)%@;Zf1a;A?&1xJ
zm3BS5=Jit_ruRBs?H<pP(>J=`8ZY%FOY25|_G{)N{%W>;*9y*EQ!4*uBAQmpeRFZ8
z_6H@=)#~+Stkbi4{sHUsjJ6+kU!VKK_-Z-+7(ztVXZD2T4?ccEGF#1VM-l%60uySb
zkcjFX)jpzFSoyH$!-7XfHHiufN&Yt?Yeboda`k_1^Ty*1|M~oKMD~cBVL`*AB625x
zE)f<S7Tn-==Mhm6FORNyKnndI*XW<i)lL5BegEyT>8K`@9KSmw)yTwkc^;?gnS8Zv
zHmbg+MP@xoH9Yeh=W<aE+Bm#?UMkU_tmykRRd1zJt36Z`F6K^Cglgmu_39U=>Iwg7
zTq&yZpH=Dk^32;;h9(-mD?Weu<itOHtiuabqH9*4TbZi&cK1tFsU|F)TJa^Sk=^44
zyiC<o^#0n~RDFLum!Te2`A;;v!@mx17?y1O>t9ASq!OKeU3_DzUf&hZ8&nhe74F=E
zYGnCl3tLn5oE>(xEmhz5%d51f8s7Tkpm(SS<;cA$f=cxES2A~}>K)#%Nl&T?HRgWP
zi)!Rw2aoij>iH>C(T}M5I#=uVG1YKS_r-&#%747_GuuA1<M*#;N`B9qJ!<P=Ab)mR
zepuH}@JYS>5`~K>cKY8p|G&PbWPsQ9KhS;u|LwQ?{&{f9jUkj_^p~+UMo{(Ezc+L=
z)r64ewvMG5xvE{Zaa29M$2I?os;~6gDPL0!KarF;iE7ZaLM5jrYs#On<siEK4)LSt
z=1qG>)3Wep9kzT1)r6R3w`Wt0j6Cu3JgV}a)qZAIx<=nGl-e2I^ToS;a{DfdodNHo
z4r@PHmj~<kl=?H2-m{e6YhdrA`-_b7mHTf3wXdxCvyM;pKeOU>T0dQ&a%TNLt=Qki
z=y_e~Im=Te*QkZmIebZV;}%g3pVv3XvSerJFsIc@D$)54M*m9HyC>tR)l?J4RxPuZ
zYGj>mA8erNxi|Tj7^=P%+wX6s8s776oj9sNCCZN7MI}0+?e0BPy_3GkbAW0>V_(}t
zR3kGb%{)R?{<Hcy@Bg2-QVcLVecSV?vEP~(dwH=#Ykxo55o_BK&-!Q19i`6V8_=Zu
z398{Q4DEY5*?BH3U44#9^sM8dm#BK%=B$5(YC^8M<E~SU-0|VQn^Zlce#n1^s?U3%
z{e7z8e`lB*5={TIq{^kVRHM6gt(cLjxA5cvnW-im+O{?;)ySwj>7S(PX;7wNE~>t?
zZKIx|8ouW9`219Z`Vx<aYV>oLI~AttJ?&XojA}x3<Etg9Mz$PMr3_V1_N9Z$QT1&-
zwy^@$@S!;}SEd?Ny-t&=RHLu;`KAU{@50$fyi^n3*<YkK)yM)Fy1hcxv;W1#^{M*4
z=yIb0)$n@X)@V#MD0th@W>lkpxxM90s@`6uv$dw0Q0lGbZK*~c`)tbFR6SGHBzC0g
zYj&wb1l91zJw3Zo4cgRb`MXr32M)gdK2>k!B`^1(ns6y`WIw8rbDoSFK-Keh?Hq%s
z`kw03Y6w;N&+K+p9v9U5oX5L;^82P$?{QAyVRb&NPEV_!`}>h-_A0%9+4Wowj;W7_
zsUFXF$@>kr&m}1i%<KKxzun@s9nd|V<`Vu^wUbt*qxX5!@&24LKFq4W3FRsH{gqMs
z#@Y{9#m6chwewz$KKDEB{>iAlK{-C$&Va|mJT%W!HtD?v<-%vV_|IFidpwQX_f41f
zgNN^)6+Jw;KYYx-lcT6c*GgYzELHEF${&0|HDOuAFXO33_Ly{k0##43t#u|*^&Po2
zW(w8tiO=qyPBo}en>;h9MrZh}?QE*vwX0{&rJC^3#j^{jMm}G#{35EJbFcSZO4T<b
za`g(T;cXU&{z^3{XX2}?sYb_UALpa$9Z_rFda4O8_0AtdHS)&H_FJiX7VVuAOVt;V
z?$S=G;m=gAxQA*`eCGlCsYZ|gW^FuGZ`hXfN2n%*-fD1^YUFRFqK;Gb^l2S`imI>7
zSkF1C;U|CVbdhS%^a~4;s7Ak8;OaH1-fXW|xk)u)^PoYusYVW7wDBHQPt~K3g@n-S
zuRPHtE!FV(HNVL~HK^kUM;@aZoquMLtW>>w_jJupHQ}?gi*r(qtXt_u9;%-EooeK#
z>RUN+XhEvs?`_^vh-y&Dzq1xg)|5Z=d+1z;W77Q%?I^|*A2psRdOON%yThwpO6w^x
zzrx+~D_rllf;`8#zMPD%d&ct?=5NvWBCF?biT=DL%6lu?M}7Tz;HHn;|L>=va9;sm
zZG6E?8?XQD?-IZtZReU|0Pp>UdO7L6PVfFr7{^Y1Uj<h4W;CzU62!l)ng`FzcrW_H
z`R)&QYP;#we3`XR9eA1Z+^2P4BCYpNl$)64X4d+2r~8A`^QU(IQBgU-Zgu+uvwFX@
znzz;IvEqbrcyq_$o$41(&6n2q3~<16sG5C}^v>&sf7#<-;BF6h|NV-!_6wNxekt#N
z2ETBZU!c4nUEad;Nyz~3<HHy~vv;3Mz0YshpU#~9=`g<$>+)Ef>m_Rcm+<#phwWn@
z=o}2JC&#|3|K3;CY4<JB+9B5VvW`#Q??2GK{LIIFQZium`;S%JtV&0;edP>uvB_`#
zf8T$N`{5hiKg%$$lyh=f#e;pvjGcqWOTRMw;nDqzis!88Y=`u6*Lxk-Tcn%=9blc1
z+4Dus9v9;7T%45qBmC81`{ELRMgO+3^H)&M*wi!8IErSk*7qQ*^JjH>%==R~?bilZ
zHtmz{^CDF4cg;&V@nVP3bNif<kLdL4;v3V<$mb>Tk}qEDFzP4%e6`L4kKT^`SRZvx
ztX1z9X8WVr@kjnpuYPfwh-l*pI5+I!bCLD>J8S=99iQIuNnY&0eWGd4Xj&L*=c|ZT
zZlc+%^?ls3b<!yyt)3TP*Z8jQ`{h;I(>#Z_J~`-}<h%>A?`uji1v#ypoSHAu-YZZJ
zV)Pqfo+;aTro#JuN(NZp7ewJAik(>3#b#ZX)#Yt<dTtq~5_KLB{NcfK60aP1wae}2
zB<U>|z1N9450iC%S;wcgU)Zj6dF@wX%|mZ|dao0;K8f1jXZL$PD?ecskJam7M73Mg
zb?&~>oSnB&qU_jz@Au{ue{f1Z+&+&7<@o6SdQtd@Vki19+4NufR&2k&l@@w<&%1Tv
zk_+GOJ|oKhgYer&w?B&OSp%#$16Omn%Kgov@HNX$z47Y34*dqHuZIRtbLu&`n!{CY
zKO<MQ%QlWe_Vgat{41(<zqi{}-gc$y-@k3xj+6*k&mS?mua)6^rR>9Fb-y9{-^@K{
z#ee?|t?MuJ`jy`6@cfhd^N+Xwy`iHiR9^2ZYTwh1`eQ<_y5p`>4%k0h-Tu*5*F&q*
z<7NJrTl4+@>q1fh@b+9J&m$78e5|rpZ#sIf!~Gri_tVMFb7AS~b5x>d9S^-k)!Q~_
z{VP;)e}Aia+hF>9b4=a+J$m5a+warx-pWf}?n5=<QsT&d$=A<$GHw7>&)c<gIC-D#
z=<(mIjZe;h*p~E1lJg&WtHIIa{Qp)e>UeVg`?QWfMb%ektmhon@RK+n!pQj$w2J=o
z$LoDQvU}Wsm#KxU$6r&j?YQ-%O6xA&n(zOAt`%gxv)S_>?eYV=()I5*#{L-E_Q$Y#
zUtz8NnofPr6s^99X0O(LNt~WPwfnB{x(^)6$6e#!7^gV4u|o2D@X#EYD^m@sUZ+V_
zs?pc_d{cv}cj4?KUaAT2>@QNAYGi>7-Cm*U*>7wgaL<%Ai5+P?zGjz7L{JTX+|#og
z)u2s{mcL6C{TS%S$V=V9s`&?K_tEbEd%co8j~wIEX!jS;`%&cc1<Nm@>N)p%-=$Q2
zGa^^7pc>w0ap<p9gK{Rmx|(WqZ1!<Js@@T`_N}Ly@KW#mF;pXO%xu4vs%O#OIk8lI
z5$P`Nq#FKA<%)Z#2E}(Cu%9aWuZNenvfbDE{wDhSO=RSWm*-J;GrNB<I-%`u9`?ER
z7M?sHGp$z%hqkTFN;NX-PWmUQdK$3Z2My29&+eX1xdt_AlV=9i=nS8=olVudcJ<7;
zR1-eBcvee0N@M!x&u!yMypA)VJlyeo?D}#}0f2Vr!8skS_Xm~Zy;gIwIz6NFYDKeG
zelHN!Zl!fJPR*Cn{1yM_D{yefbB*WT<X>WFVe1|5vD#0pPS0rjF1MFT?T*&^$!eS!
zJjJZ%C^s?74fDM|kI$dnAJ&(6JjwlG&t2|RIJrOUv}a+l<o>Ye##c*HjchrjN*Sv1
zAKp7%dOl=|1EBoPDSy;2KJOza7e31+A!ga_*|cylUg+-eLZ{YG;q#S}0b!pPM7RI-
zFQXdLJh*)y<>-6|qkp9wyn8a9T1_=!Y}GPrsYcf6_Q3|KYWv-2(0~4f^=Fm-;Ff%P
z&JMfUmSz^`N2uGcSZSWAsD2}=-Msc=W8W|A3(`8d|FD#~`+}g};=LT}A?Wix*1;K?
z$BXq44_Z>Ezt`Y--`M;T>{o&H#8_8G`wLc199l4hu8Vtb^OizXgG&CLwOF#I{GlB+
zXM9vXmjIj}{oD^ekLIKI+$QW}j(yB|*yj}W8}%Fg9{=t4aB979O1_+u52uW$8!nHO
z2=KmR?)wfe=dkHrm*kdyBg6Fx=b7CYLLo+f8CzonRkh>D;Lg|cFv^3k^6+LIwtNOH
z6w&5O!;c>Lky#wfir475pz`@wI2Yso+S;^`<aNt#EieE55Lm6ZR{MMrej+JxQZoE$
zg-TAP65Sl<3S^}7yPUFb9qJXHS48Mfg`dJtMetMXTdMB-Lbuds-tx(tJ>J{hFIA;R
znf=~_=jgk2;*tx$MA@-uNB=v1%G>!VsHb?|K|STap5nY_oY!2OHrAf-kH(dvs&}78
zwD)Y=d+=n-6aB>xuWN>X!M~jJd<K7pKRd~v1MTDL$-Z<@IXZsl+sfIwswovvl=-A+
zFVJ2%xxIjY!@r&E-}1iZ+};my&N$ASO>TraTNi-)0`3dAFJN58==r(wzFSteCswDY
zcRa*X^uhdZ(d?Dyt7CuCzB!g97hZ=styWUex(_wV9rsx^_gTCA#IAJ1cg5!~pPcxo
zk9Bx~N_5TYb1PGI$~?N{tR?@6(hrLEzHZziT3(`^C-;w&ImpkPmLIS0isucQDZ9=i
zb^Ck=_>(!$Uv_cTo37sL!ZW{dE?4r!ob}%YYFFHnPqX)d21j$+4X5qD2|t0KsCmD4
z%l=v3g|m-%lh2KLXMd5}R3i&y==KU#&wlK?j(yj$@4C_F7GCc+yxN8JxHw-qYw~?4
zO5OR&w)=640ao=J+n!&G`v0AtPp9WEvV60Jt!XBW-akD*Wh(j+jmOuyTDOm>hI_g%
z9+V94;1pf^miq52mK;^`pH=nO=)44s`-nO2;~Dis{yWsqeBJ}??^6x`JHy-%dY=ng
zQsq)ws?lA$R^)kqH)`oWe`ePctI{#LZd#rvg!+j3_!O<b;cGsR&!1eM`x1{QxjsL4
zxl>`PO6$Sk?*{uLQvjOXE}^{`*ZeEW+E;q*l&`7!KSyDIRyOx<R_y_+c%na9(f4T@
znYYrZ)gG$6?=!7-zcuhN*zPiV|MG7)P%hS#3phW_*=!zW#~=Aaz52=fLCJq+KMz@z
z4xYQTJ$Ip89<JBk;r*KQq)-xS%>AYpRsVBnt>R@>I-=JXaMI?#|LGS`0cdvnZ`993
zDQC0oG&_Fzxxg*ug?`ywqZU%9@+I-nFDr_F!*A624a(Et{n?~`6Vc8YjGlEo^b&O@
zZ`%j^o5HX7@+*`FU*#d{d7}RHJRw)zao4GHMeg`;-%YA^&5MA4!@r%h4g~(3x<6aR
z#j14t-{akuu2c6Rv8x=-<|$gfc(<2#`_!Hv?v#8)r(YM}m}W*^-|y6XVLz0rW!BQd
zjjYq{gAG(Y_a^@mL)Eup`~9s{!+SdGoI{$WfBtwMN8+vD7Ti*Q4i7Z`{-5zn^pCDS
zN1ce*eiOz0+fyn4uk8lvC+cTv{XSOtg;nYJ_Y*LC&*Qv)t9}3Qe{RR~t-GFY)vnLt
z?z}FXPlEGFlG|WG_BZ5x{lEYE6W-=g@cz8On|-Y9VI7~-*3tRTr)Tv(5AJwROYJ=^
zp|P*+A?h5FnUZE6p^AMP?cJvl_4GkKbqWVw^TDfKdiN`|D?fIn8>s(@H+!6No{iJ?
z?RNToUT;0nd!1<KB%mLGulWYH`(26wypA*ak3-nbE5N?3yxOJrxk{dgrFA?99MIo@
z{subl<XWF^Y9#Mlh;t!k&ZQg@KDu~z0oBOo3zlC*)pPFkzBm_x&U;)racIF1I`8ql
z&07jl4J!F})?&$;@+ZoDnb-YWMdfIf-BzWebsmjfcDXHG^d}9w+Lji&@B8Id+EWd*
z@1<RS5vY97zcy_73|c7u^9)c<c>l2D{X<m06V>kUufrRLB^&?xmr)I=STzq7ywJYb
z(Y}aw&I!)FG}ga?b_VT?6WSU0@dH0Ln}gZ$v)ccZw{^z;=irITQB=FNKF2#HpRDu2
zIzG&=zc+L=g^KxK+V&ZAsvp{(uM`8^HXdjeceCPk%DG-f*CT<iJKtYX?(Qsir|sLN
zc3%V0{9QDAmF5*WEkAna-9~SJC3ANQL+P9@R_ls!Zf(lB^C=F<%X7JC^?_G=d9};x
z`TS^C+V0*%oxwZli#!Kpr=ZG9eV6Bf$)@pUms$Smw)`s1PZa&WkMhKOqZ7uv)xJ0W
zXFt4@3gDLZ+v@fV?F--Si}2+mX#Ew9yJ+?*^<SaCC*!Hrv``bqnzL^%{Dtp!KQi*f
z%kyX<d8#}(Hy7n4R(a_?9~<TDj&gR|`98e(EiZON*Q`FbGBrl-hcv6d&+7NZ$W`sK
zjidSU^d8szE2_TIYo~lo72{qh<6cXuTuMv#@rdr)wPHr9-oldyWTu*MXxrMXR3oGA
zq<@mCr$L#9xv2WmwvBp<YWSMZ<MUGu>PtKxs=E72s@G{!HTin%v(k1p<=|br+SYwm
zXo>#whyE*!!~c7{4Es#)FH)P<&&UE9y1hab>mrS<i}WnoJ13TM^hKn*w3BN1GnFgu
zp&At58RwZKx4<~h<i-$+A^OYM8Y8HB>$6?Qk&HO`Pn7%I18c>x*Qx!bC<k-;J9y`6
zyL~AJh|*pKs{gig46yGn`a720o=pn}{rHdGKaRKgS-jUL-tF_>mpo9rW!1h^+|BDx
zDL~QhJ64q+ul~-fUAXV2e&5wQ9_zGmI<v<aM5z~|*vb38KxXwP`9C*XReweG3sLRX
zy3Y~LY2xdgChWiI>DeuLoMls^<?m9B9ys{+`&7M^m%QAEYQm+&k^QJf&UrF!09DW1
zwQ~%j>U*kBt07dwch8C*PBm!EzLTS<M%PMTW-L|joys44K{a7n#4qEiM)sI=e*#tc
z5AP4!#*=yD#+yBur<D4>`SSc*tDZkFF97oba2^!B-_@;jBM&|A$#cI`jr>$`9+bNC
zSa|DCSA364xe?m-JY#qI7(XjHm6kzt^QJwcsd}?I<360?R>I7OamzRj%Eg@dCfE;H
z+kU{dtq(}40Jhz4MZe#H-=+x40Q+A(y#E#6+wtBm;(I%uhaNl+@y>yF`!Mh8;r$xW
zUyOCFddBxW#kSU&M92AjM{bRoLN$Eiv%9A!kNY)hlV=9i=nS86v98q{wk7=$YEMGw
ztp-P_M*dbR>Nr(TpVslGsLFqM&(Jnb>3{B)u=8tD9Kf~ZZ&Lo<TCd!i@9@?q2fZU_
zo~mo>Qx0?bEoHReYSVoxr2xR+CjQe3Idc3hA=ktG@u=tKw4WIFdDzc9qkhPLhdP_j
zd!YS&s+b>Z&iq*0>vM_$cHg&b=OFrc-oYuR?u(*0|Ea^ARx4>CM&~;i{VP@P9&`8o
z^$loJ{sc`Y{Dq-?Ps`4eoc_<{>V_t#|Nh%!(@{+*IevFWs*#E7@;pw}Gx=)UY*c+s
zi_ChGYIx>1&gG(teRUr8pJXJ-f2?>uh>cx*mY-jpk^xpc)Y>m9H(u@L)h?`e#(HP$
zt4i<FjMkSqW&FukW~}EN9ft}(`CF%p$>UMeFDyu+8vSO0tJkP{v%OyBCe?(^g9hEE
z8aa5;#(PvfRgXRvlDyvF$`eh}QVpM9^P3D*gF1e2<T0wz`DYf%O4Yk}PuJ{J6Fy73
zI44!CS18=M1;rFuzS+XoRLx$eOOyEL&+6xj@Llow%O|Id_a0aD8;aVW^PqOHe{V1=
z-h_*}(-fhRNB&T+esQXv@Q=ooqUy^tZ(kXzyw^Kw;y-_)wMU}bjeR5f=2(_ocy{+6
zU0AyM95o|)*749wRK0C;*1tm4fBZ$%c2w*B<TUI5{E5~Mpgy5KIpX{g)IVF+=cHg~
z*6%I%^O_Z}=+B!8HRgWPi^7Tg>)??-R6RdsD*6#s?)E#y^9hsloKO!if5naGz`ND&
z-J0+JpYQB9-xoaiX#Yeh2T|<A{=N^+x3+q{B>dyi?FslinvdG|F5WouW)Joqb;f>Y
zy!7|OzuB}~wyir(DL{;$n7bZ~_j<yMeZ1HaU9<Y!%G4O%?w7pr;msc2%1iBYtAG1%
z7gxK|<$b;p#(#Wm4^iic%#<|q2vyJeXU-j^`k(bp4xgW5b^Wk9JyGWW>8+1e@w6%(
z(dKRHEjMrddqYQ4s9690VEwyQ{$*7<{(fv%x_AyrN}QAo-zEE$m^M!~#ev=4uY!84
zt$&a=UcA|3^!?sx`SJR$c;29yO6X_pdO-jF3SpkhQXIgxcG>Rqv5pe!DCxRuv(7^i
zrF|8}PRv&_cfJz*K-=>vck3N2zMsNh1osz{*T;zBAEMZa`=7b@zvu*W&ed|Ne~FSW
zQS3zh7PEdw?)Y%uP3mN(^`G}8k6gzI|9<py82F^lM|dtN8L+MYl5zi`@a6IPmr)IA
zCZf}?i*HO7{qbz~(}sVf-j2!hXn5zt+CJ9t$@`vIy+4WV`A9LqDfbbyyKLKCx6kv{
zx(}VVSLtl6>9|@#skfT9r5btcvng*=^-Nil*paHQ*`*Q@RKp)P=X@6Loys44LE}$Y
z7V*n?s`8&x>bKMK<M#GS)bixTZeHxrx=v0syG66t_s4S?>QSSyzbN(>b;SOnn;I>D
zmp;!#4;*~^eX8EdOJ44i{G4+sab!QLyqv=!d^?en0a2b8h3D;%enXVQqx<LduK(t3
zJ~hhO9p&t~H~E(sTG+l7+wX6s8s776ojAGhsfu!cz;kfjxJAk3H;>VCaEcG0{2%NS
zBv%Q$dGKO~)_wrIpE9bKCCZL<<382lLvv)VOyB>4s@G{!m1^|0KHt=!>RmYdh?i=@
zJNt{&rYd?req@hH_b1SN=shPA?Sr=GHt^EMOYM3PaAm_)c#bI<z;n}M9bTX;@Z8k2
zXEf#D&Fb!bc~hzd-mkao+=n#<pxx~X>NDzd#^m>q!jlJNPJRzLv~6wH<oA%MJL#WH
zevWTYreQ9szO-$lo}wDQ=JR-r)6vR1X5YzClt*-}^kv3U^`FmV^gSr@`GVya(Re-Q
zUhlh<s&7W*>J?PO+bj<Km1<DV#8+2SjgHMe&PUZdqSn6kR1;q6oj-<Z<c*o_w^H>i
z+B+wfsxKnlrJYp6pQ&7N57nUf&I9&SjUNBa+IXtour2A2P)!KE)!-=A$lppu9jEH)
z(>neXRbQF0o^w>gPyW{FBGsVj7ZxN@jefJh)oWC}*<P=5lWM}|L4$5njU2pa<2|aL
zsz)CS38B|td7?>Ls^Rl%ev^S}P{$9BJVrG-|I8v;se1SB>6)Et!e?n0=cF20x6+L~
zR6X}Q)yPlPw{qgpf>gub+q|U^)u57pXDybjDSv23AHENn#l@_6t)9<}^$!Q5f2D=#
z-6QrqT=*lK_jF<R$3lOysr|{QU#MRguc6PWE+4N!JAih8WjnyDKiO_qiUD|^efWJ=
zUO%jMexmK=$avoYx39kv{XLb}{#9P<39oh~ES*~MC2Cw`_qYKsQ^kHz*bfThBeZ}1
zsXncS(Ek4L-Ls;HQ^oiQ#z)F0H{y0b2PNA;|L{C*?0MSm{^19u=$>+PTYi1tFR#*`
zW;(p}$wBWVXFdn|TUhqDc-!Wze}!V@<sA0NmP4wPNuD2&ed(ZbRDD~IZLE;|{D67#
zyQe2VKQ(HTX9m^i44<`~P1U=0^~||c6F$0lb^%q)55fG97Sv55%QsutnyTmQu&ZsU
zq8;^H2dXeWOp*Qbhv&nGpAW&s+~*JQHJ5MXs&?7NQOKU&<C=d()mM7$l&`6VpD=b_
zD9TODa$~lxT;YC#eJC?@dxbuqdiKA#xIR_i7hP^Npc-E9+Zv6j1_f^$+Kg)SFSob6
zNtL&KC^2uF&Ae^2lV~U1*na`P=fm$&4t$hDbaR{5*QGdsQ{L-s_h;~&^(18wp81V)
zxsq9I99}*zmFQ1a^nIGDUFR}j9g{Wdn39V;`OhisiT`}U2f1@@{?Jd}?Sl<+Av@TQ
zol?$nQHj>xh-SBF_Bw8#(3AqO&Cl&lU*r5Z*8a*mK0F8C8#<ao<!wLgfcqpf8qeU}
zk9fBa>j_T0JdYNVrwY~+U_AjHH^h1Z?6*YQ17Dtb`^r!a-xZ&qi}eK7%uAk7J>`ZK
z)jO*FCs7gg8-6e{s(nPUu;8$eu;7tVO`^g=lK+j!8j(FBXIRkisEFLjpG(jH!3}PA
z9uXBWeRRzOQfNvv5G4Y&Cxi@}SU5&<nwHBjb>iz8zm?2ZCBA0=-lgc`3T12kEd^4u
z(?|Ru00JOj7l9XURn9l4+g)Op%K0u@<JyX2y>1e%x-eg^F6A%XBBJW511)2>pSn-%
zR=FIYd=T(7%{Q&qcRu1YHA;7S+udot{_8r?sz%cdbbYoF2M@Oly+G_z*)Fr6)AhzL
z+Xn>!5C8!X009sH0T2KIJp^t%)+9&a?8k`cDr$d)Y`0zgMM#&A{yIH%7uSFQ2&7J6
z&*-qH>%=8VdNuV})qiVO{}pfPswc*G$Z_9&9)0WXWdd~<2fn=$)`0*BfB*=900@8p
z2)LDi=<`~BDm-E0ftPm^qw3c3N4`M-1V8`;KmY_l00dGefO)s6bHTSD00JNY0?sDz
z{h`&;<Ie~C^y-PNad}>NFT@802P^%cU*ffiTdLiYpxQV%yEQ}ZKmY_-63{z;kfjZX
z3j{!bErH{8zZjgb=`RTi4p#aBTQ5dzq7!(n>`Ub$bEWY@#o*xYZ%XTFA`fqS{O2Xd
zr9irN!>CU~hVGRD>GPHT`fzQ>UnQ-ob^eplc}7mXUTJ*H-yeAYtaV*8@A-j!+6b7v
zkF3^qCs258xk6|V009vAFM-jMvi52^|BPg}9(^fW(<;6Hlvt^MPA4^Qy6@tQi;~@H
z_WcF%TeBa$zklH~-vk1^KjQmGzh8^Ohi~tLb)pjp)IY`gdsg>v(Rtf#zt(;rgm=ci
zfGEd@%Q4-S3H{b&I!&Nra<ID{mD88^t#q2^j$G8FUjM5^Ryn^=5C8!X@F&nFqQ{Q$
z^{+{0tADyyeGW8UrS!d3_8}b*NR0q5`&y@F8&BWy_WZ^Z8`uH@AmB^_O7q<CySe98
z$4$*2p0&bZVpQ$&PXBogtnJXd4vV+zoYv1M`u^`)&-vsnJnMRf_&`7p0i*XV|LgVG
zAo4Y=#>IJy$LccS?S8M<9;=@dtS%>{r;UJr`)Jo^ZC3v`G^=%D|BV^nfB*=900@8p
z2#8KVw0&84;a#-GwH3#D-6VD?FC1V82!H?xuqGhNdNJ53=DI)D{;Cuo>*oZe^zi$`
z1W-@;uBS@lDG!&X@+YJ9O*sX7L#Uk(Ag?D0q((WJb-mW{f3xh6<F$I6h}ZPgdcD#-
zeb&FLwF}|IAh7B{%h>Iw?i05t@1>%)zVen2z4HQCUB?>_*yAn&TIT_{iw<(@76N+r
z>tnqykAHkh@4c*kcS|`VpCI4_0)hH#@cqO4q&SU-6MQ$dJn-6|rPVI2&(mtdE1hql
z_jgv$4|d0+mSdoBsre+nvyOo1>zV`kmDcaI66c+p*8H0lUMp9GcQ*n5_i(kJ-AxU-
z2LTWO0T2KIcMve^d6n0A1C^@U=ObRi>7DNxDF680*#zWyer{J@_Ag~+hug~&`Pa5i
z(`n;LR^`X&-GSA7JGJo`wHMc$L%^)(M<r|R9^XRxIB!$QNAWudfPgy)`0w*<_h+l9
zYn_jS@DI+Dv#NZwuFFLD-D$r5>pB4kXL0!c&}!-N=YxHs>xr##d0u!g#3!1X#oelQ
zl$Uhuey-4)KJWA5l<ZTxo|e~eypCU5ZI@l;W7odatkUJ}d~CbgW4rt(P`X;nSM5GD
zR)rJ&eNgTAg(%^yZjX(or`7-c!`s!4@S3jOe#$Est#%ppYo}kYm8aV9df!j1zD{j^
z5Do-HCZKm-r{3#CX7woByA)kqp=_<cCB3S5J+jg30_92X{w};-r*{5JAP%g*clvXK
z)%ml!--ES3AU+V#LO`h>+^N3@Vr12PT%_}0pCHVC*EauME3ZI)9_V}C>wBAs9y`X@
zzb1*Mc0I=iDrc?Vi^kh*d*yT4v~omv5C8!X009u-k$~0xAFR{kk*OA&^!E4KeVx&M
zW3#T8*)PnOt4sMyw}?|!&L<QE+(<z0_@R8AXzP0e%@4%)`KHzS&PUAVAez5uwbv|u
zZfu*7XAqDHjGmOWSJU}tBq%r-@dK^nKu&${5j|g8%Mszt9WMo6ckz|)%TRL(sGT?I
z)^O(1N9?>3(E47E@Yd`@$C|TS&Phi;TgGlbb)SHPIu3f<W!~p0>dhCf0|5{K0T2KI
z5C8$U5|HO*Ku3KVGIXyb3J16PEAkBjsS%LZ2|>RYoUrLH2^_$Ig&(9=*YKTv1bCT;
z!RlN(R`J+ZB1jnoKmY`sNr3n9AJ{i_;_Df|mB7JW9L#QQ%-+`n@r${?4$2MXCgQ$O
zf&A9&b`<f0fEog#Kfjr^j#bonmCD?%ed<`HE6O;jQ|+{Rzh)(W#_wVf5dA)5wcqoS
zo*3Q*D?z{=1kBn;4e^FOT_-L{0taw_A2`1r*mbus>pVug@`H3i00cn5Z3M9XTiyC^
zr*U$dpCiBe2wV>R{loxrTe4d9S*v`NQ}%7rdoHvxA$vlII!-Cu%}YPGQ9I(M<`2(W
z;V?0(_IMFebqfKb{%QC1c9oOS_Q9;{S)H$=#MP|vTcvdII|#5Ku;KOm1E2pk*b4=R
zNBIE@U;B3)tlDS9(<DdX?8k^v4t77kAbmFPN#Mvg$A7nhYX0xmc^7Kfun42Iz8FnY
zZ$Fz|*NNW0X%{!%(>3bHW?gUAIV5J;t2JJ;&%e@Yk7(g}uUBrdPc*-Qy&%8?0khV>
z@sKyzz%K!_{LZd;`6cZR%UGWiWLA7y-|u+~KkmEFqi@~4OrVeE!0Nrq9kmy^1OX5L
z0pkP$?c>Did*c=&1`se#K=g4C*7nOp%^shW(@|1Z?Tx7M`13LAeaflv`t!qQ5coF%
zz5N3JjsyP(0jCf!Yn<Myc&+YFv#Q)#^@G_J59{Ytr^jcNUm+b3009tC8t*l0J)hO_
z>V4kl<vO$0C7K<t)%^>BYP8jMn3bMhe+<M0--7@MIFEqQetfg<3r6F2UWp+`AOHfq
z5pdc$?^gAn@m7wo2L$XSz}q|o)-06zkwp33>V7G`>FMnUP`Zw_U+@~A(f#Rp%@^zf
z0T2KI5C8!Xu%1A=Efe~!$#k07t=8upDS`k9fB*=9fN=s=^-~8*$9N79gP8=3j?0*7
z8{!555C8!Xu%5uA=1uoqoN-aIO0CZ!QUn1I009sH0T2KI5coHNeu>v6ZmD)pvRnUc
zla>Du)SqSb_g3avPg+Lj|M*|8#|B&@9)WaHnKxS=2=O{ijrOOqx}H~$uxtMeq?@%@
z)A?s4vpMMXJFEJc0;QwYZmajlVV$1Ryk)h&i!u)#c8>ZqWawTA9Kb<DKTz8)`-juJ
z9)#EHjGkkwbiMz3n9<Kl>GA%%+WVl|aI$}j^1fr%x#uzmy{h#5syAMx>)djFot#f4
z9!7sRtDRL_A4Lym7C)=wHR?zH*ITtus$0@w{r>5e`279SZJ%GbRXN|FZg&Z5bsI<I
z7X*?C@V4)Tl0B@(F+`0=_F7NVeA8-u=Oady_xEKb`|MuF>Gt%^ey*4O9hy1;v**#I
zo~r%d8r?q=*M~h_CoW0S%0cP=%=&k|o)&04*68<o?cwD*t>tZ2cwTtG4r2r^hyH$I
z0J$x(Qe)OAU#)kZrE<*rLYV#hq&NO9MHg2nTkCI06czP-PPUu%ebYZaecl-q<o5R)
zyYrtra#53d{jU;Hm6v{bz5cKFI=%N9UanIb5BL9FFDI?ni9X-V==+D>>$P$=8lKhj
zpd61;d;G6A>%GQ5UjKCcKac;mYV88!E=#<qzkT=&0xSvm*AL!5>uqm&yH37evn<O;
z#U;=85><_ktJ!`1qsk_=-*~TwO7_YA4Yg)HmRX!r^G!Rx6Q$n^c8b+ciSj~u>FHNj
zy8qerJ1_lJM(q&wIn}Ilj1h0hu!)6ZBqa`JJqICPb<eR{+dZqFCH0QKy7fAv<zw_d
zp!IsA_TqXr1VnqTmdE{hSG}cU^!#wO*9WrCDz1Uj$M+!MMgmIfHQe$$tFl*m|Iqt8
zEBln<5xu;px80rQ>%XoO-c*S@tKUVRZzDg?v*ICoJfhAcgxzA#1Jzo7X4f-T{k=+d
zvPuW>WNTWb_n(qh4s!c%wJQ0A-M<IwH?_(x(f3Ib{kc|j`|VzrVKjZa`CC=btxCtL
z`Ibug!F@)F(1YLA?%(9J{m^*pM}R$i_V1uvtSuLHtrkLwLcpr$0XZE}xWG;j009sH
z0T2KI`v}-Qztd>{l_=M19e1)jywUSfaQ(r<EkiF5Rvbj}^FVg8;-D0deJv?c7J-1)
zaa&%)+dXb$G<~IdjNjSbKLYXOWnVO{b_B|gS>KD$w!ud82=I2Uo!sARm6}J^nwSI4
zGqUS@YjW-URMf7|a!WYpm$<ugXx8(MRq@*0ujQ2V?WzTsFWV$X;q1qV(f41sT_31l
zP?Yb@ZXfNc&u&kb*Zw_LcKM&b!0TsL`(|jZr%svIYt}qXt?k3Ye7U-mzjTWPQmyeJ
zJP3$LK<|DCqFu-8K81K5URP@M`W>XV?eU+N9G7GcqVG2$+b`$GEETbBhRq-V0w4ea
zAOHd&00QO_uxow=(hYmMPF#{CItRD+3&`hALLYs;JL`Rx&1(_OiP@>+$)e9U&^jK)
zTX?h14>YG11di1y^L~&IvG*Sh?4kI@2m#T~DKe5*Tnhpq00JOjKLOGESG2Alv!6KM
z7Rcl7tktadFstGfy&uo6^<#W{JFMe{0I&19S=$w;U1i<>%Q`+@{6woAqRmJ3w^!?Z
zM=Kw*!&|*Sn_c%gtJ5{R+*rjg%KA9iIicU0Os5Gr7~`O~-UNCsQ2JhLdZNGYn{B^V
zo(K;DAOHd&00JNY0-_SoJ3mv@{K9S!;E@2UeMJ1@G5fs<@jGL^g4yM0b)C~YA6c!P
ztjC8CUymK*>tB<2;h@$Jw1x|`zYxBE^!`g)`DiT{w}&@tpAE$OV1E|3`xo*L0?7pI
zy1)C+uVDSzs?yauFT_8*RdymB5C8%71X$183lyJvix3h7{zt&-_3Zyk8UF(T?F8(4
z&Sagg-uGwi#N2TOZ|9-;+hcbrSUsQFuJo*)pYER?FZ<qzI`3ES_>SAJv%7s^oqmgR
zgQN0g&FB?fSwA<}6`y~9lh)6y{X>-aSgmIkCBLu}1V8`;KmY`sM8N8PQeM+DyG&V+
z-y*({VG|3-NLo3VUCzA4@86H(_RqYP5A1O!0bcq`mF(b+zS2Gd_<hu;Aw%~{-~bM8
z@&i`wl-~8DqFu*|SD<-WZvCDWSHuGX$pnnfCjdnd;E90ReXd12N5U$5c`8fT0s<fa
z0w4eaN(9g!h5jg>`lFQm6~BW32!H?xfB*=vBOvOyfZ6-WnH9g@^_IL{Cy%fBdm=A(
z`rBpqXRY(z5dOh@LA%Qh>4N|W@JE2xdFgJk%kA@$k^c!354^ma80Fyh_3~=-&tD77
zmqi1^ADi`9n;@^<D(d^RRd(~`HRfBRb$<e<g=e+zfYtHf{afF9J#cddH>>NLQhMgM
z)JOpYKmY_l00fi?7+nW~>!(h9J>$2M+c_BRKg0EG`XRwFb&kp{p}G*`zWY4-*4@hl
z3JxOpfjX~7C=dVv5YSJ+>iKd?`wPjxtG!Qa?eEpP?no(q{p7{I^7})prN^HS_Cd9A
zQ0iwE{dcjw8rFjV2!Mcl2`D}9@cO%ZOUzi#)$YHe6i)5)x>C4>`EqqBf9V#1x|4%3
zZ^qRi00JNY0&XK<*LiTfr#qoSiYG+%j%puKEG#4}G%R>zRFkN%kmP?OvPNW&$Qc$i
zJSrl0^5+s^!C}D-Zg(CL6)}Bu%>z=X+x$p(ewRJBr00Plal~mlW@ekb^T+QZW~*-S
zj4o6vH%7@I8V~>h5a5nLwk;XTjrs7hWS7c7I``=O^^a+9Nm}*LdnC`%xNDNAdM{}5
zCr!>INp`DT4p2S_3}01!Y}x4(h||<4-7#egzU<w0jA&J(=?1!f%AUEW@2ojX>{8jT
z3K4(oi2gp2@TRhTP!IqC5C8!X009sH0T9qbfVbycz4p}364Cdqug?%s^-hP2BTkg~
zgNUxbpMQE#;@mxwsM@Yw=)SyjPDxg&o;r(bKmY_DO~B}V;L))w|N7YZ(3L;84e`2N
zl}U*a!l*ytdUKx7az7ih=JAj-#4HuVJ7FaVfB*=900@8p2!Mdw2>c#2V|(U&TO?66
zq*0Dab2luOjH=tp9r*<T5C8!X009sH0T6gJfn1A1)~=cxLA2_lW5ur^00JNY0wBOQ
zfzSa(uGD^gDbcG+>$~y0GxsmByIj8-d_1Z^ol^v=jRW7U5v&6N5a5YG!(aN9ZM!T%
z(yBb=0k(hu2s}*Swf$XZ&fXtSpx|JoA3W@Z_~dp1qOWgroBdY%tJddE-oo#0-?~ME
z_Y);iHLBOSmi;SSkpijJ{$Mn{3Lm$BdfStyB(-WZJzUQhfuWxj`zfsZHpwil*&}aa
z^V;(yqskXG=UcNgQ?~1+VlNV?GdMWkW5vp0wN>929+dwiVWqt0K9BTT<`+M>AVJ0E
zz^i}3uG9&*?LFLW`Axml@hu3jA+YiZ-``)fJ5J2j&eLigj$3q+7*#g8c5f_&_U-SJ
z#giaG!NJXbP`l8D4I|TEl%QgA;H3pO+TY3R_3MWe^3BY+N|Kcq-mn7%K!862Ge#HR
zobQVmVz%0q^ZrwwI<>IgQAw-XWiQeNfn)-2#vHzw^u$%kZsol+JOjsOKR@6+;YE4p
z1^Yk%1jHv$t$f0rGsk`>qNvm6?Xw!svda@E*FUwk^N0gPR@QcywV$=s@y?vOAZOj6
z!^A1d%5Uw8NAz}tbv;9TAfSi9`z;#8eUNa4uu?rXh<we^OGm1XU3Xk^np#~ZaX$^M
z+v~mbUeWdG!a*G?6in;2O06y@q^FI*@$qwtO`m#5vRk!T{ol}E-FW8R(V6!Wqx#>N
z@eK%o00@8p2!Md-1oZC5XxDYT@E$g<^v>})W)i!U7Y?uk1V8`;SQ9XNeGKBaW*ri*
zoLT#;Qhf5fZdR(49)5qAK!u1uc0_-lNZhLFhgZ+tI7=eoRY}L;4c<L?iUiV!OH=vN
zxmSv%Thrq-(W}ZSAcUI;VE^|A`@gF#bF1gKskIB?#2_%_P|FK%H%=gKQQk|XLb{^0
z!(TfnDJgIHuxq}XoGx!XV2`^9__l31_re#KiCT3R9pu(61n$?llWl6+W0F=qwtiT5
zQtzf@mhOJ~)`pNbk|eX$E#-`Sf`AhUoLk-D&k_~)5z*D`{Rx~FzY}~nwLENIS^vA&
znw}$~t6A&SweEAFHhhsOC5IPUKqS$%W7M-N*CnkadNnnl#CO&axR$MCh4eoiB%<n=
zvISrEZaYSds#fAt_RKwfXU$o{s|Fgcv-*3jToK;g1oWOO?QUwwJqUmR2!H?xxP!p#
zLBCGFd-$m277bLYb=FsH^?A#ML=+vqs`}Wn(<czE8Yutx-q{2yR1W?$^2T+^?fTR1
zL01noKOs3q?cRqK>GQE4uG(_Q`a{v8S^u~xLA7x(dUt3SHgMGCO2>#*YSdm_Zw>*o
z_6bz7)~@yJNVmw9+OIDqR&h}B1N;sGAOHe(6X30%#VQ{1x(VpT5hqIgLDX`vs(e`O
zJBxVO?&Axd*7DT4j)2$jYu^}~eQV4WNi@y2B}2I}A6}MNt5xkNFX`x=Pigmcm=|x&
z`G0}t*W0!49dGI0FL1N+pfjIKR;%6h&uID^7C!&6H|jUSi`vy5e|2n%r`=>9;YFvM
zE?cR1xl58-Rl5$ss&H!U<~7{Zjf)nxSa@BsO8dY0bJGr`FH2^rS?gxaig(Q2nzsvj
zZb*97u66|JI-7vnejw4qrI}Iovq|N*6Rm1i>5H=ekXiXhydc06fnym?-p}~X8ObW;
zDeo+8F}g3W)#rK2=>=l&Lf&~_e!4nV5?$5Kg9yZd_4i|Dw#hqx{4N4z!$I%5L%r9r
z_6NiV0$K?4s~tCV%*+eKZ5oKtoePT=_J8{l5k*V2`Y7w;buJT8Rqeiq2&ZkINSRk4
zKM(YM`!4xz7wf!2;#GMWkI-s|+UG{r;R2Pj*6-D>&u~jPc|L|#jtCC|AOHd&00KM`
zV0HeLUGeb9REtfAmU|XFd+vlJnpSUlcH&E$4okeL7WN1W0w5qJ0i*l>6xz4HPZm#t
zB#Qp^@!!XqZrCM>uFo!QcX{ycy^?4;W_zdQ_j|`nPF17zDbV#|wl1(51lSOmIdwtK
zx<Q8t6da8BLDYmUiS<^TlhkU%U;348yDUL+i;8kz-a4tpvuVohlAz!KKd`<Yu(_)v
z79|45x=tHYf7NaRO`U@hKm6`Y0*6lCNecV=3V}MCL-QJWH@7)>nLxn-e(=9~ptn`G
zYQ4JNboAy6*MR^CfB*=900@A9TM3ld)wt8v>z4>rEDmn<SL7Q6QX|l;RJohO8XqB0
zaDX4Ms|TsoHGF3ufoiM1FFYv!NdkSCgMB4}ltBOlK){&<j$I31^5@m#1PTuD1Agj(
z*{#iwYi`Y2d?JBB!2y1Ndf?P{!0dJu@q>UG0=0(CYP9*S{X|qvKfHSG##s`HsCw>|
zV(HfOI89ioQkjoFpS8~uBXWC1)!_D_MaExw!V9&BL!TBqcjOq5$!nD=`7?e80T2KI
z5C8!;5D@*`RoLI*&k_~)5pVzp_<?=(zzz2b<Piiw00cmQJprrFn@4&#D-SyJxnws7
z_MU_ook*Z??}BGzw``KER<l@+DO>Pm@3vz^bk#e~W)@eY@wN*aIO=kxV?<OHb==wC
z?g|lq?1=t8kr-8@{($R2;86rRPpfq}ZqZ31sy-@jX8!it+hw*@%XWeArt&$EMpcPj
z`mIyvO)hnnuvVpX@jD2xAkZw={CTYswh$;dJjxGP_}ahYcrPhKxrL?95-1-W;=h~R
zHn{&$0tE;7fqC`d-)*3p|6jc`<HX?_8;RRg4I36=w4OI<G|iAkIV#QFuvijB2UpG9
zqV@e~NmLy+uJq3FIc5^*!yJtIv02yG&Jxl0t*_4z)@qi$TH~FvXYT1cYt9n0RI5Fr
zg+Kpe*O|i#ClRQ<97OXQ*b4$Y5HLCp7U+5&@&+6DC1BR`hF$UUOWGZliE?iBSI4G!
z+D-NmVfDr{?~cyAmwY9dj+xme@BHz*h*ipJp96Q)UgQ!4KmY`c6JXWu28zeHg@^$J
zj1!n#|J2&fBMuPg0}e9L3h9d04u9>SWR}Y5DCy>v^}l<q={e$7RpNu+K>!5w6Bxd#
z`q;A5ClIZwKSf*#0{jxF^`0km+WB=vR9!!$kZ)$jRgzU|RY|_Kzw6A|`{RkI8fbhA
z->1Hg&MLn`Iv@Z7AdqH8+0Q1G-%bLl|N381KhK=HAZOj6!$efA*&}aa^V;(y(R9PY
z=RfvF{YI?PKs8#l@zy~0-sy00#EBAr5Um=B3%&;d5O5xWp`R7|DXjZ8Nw4z$e&)Oq
zLykZI1b8E`;PU%jiuC`JI7PKSH|h<~TRFlW5U`T~Z}S*fvk26GB-;0S+dHeTGizTD
z)_%cjeC@mByIrjF3Q4c>nlIP|0w4eaAOHd&U_AlSad?Aw51t}+tMxfYiXZ?2AOHd&
zV4Q&7_i|R(8P5S?Fq6QsuG7ZUU$vX)RWofv+#mn~AOHf^6KMELzp`zYB}i7O^*Ka}
zAOHd&00JNY0w4ea|0YmkSL04wuU{f|>%VQX^8dV?TWe*W^`w>i*_btthnyj-RF4g~
zMmz$Zfn&3uA8?*HP5af38#-p@1!A{~=Mk{j%>?XTPwp1>ddGSEuXEeEYsjxV&)Y+}
z+gt8#X_t{ta|zge&Y_&U&eLigj$3q+xK(8idezr2)8udJpO*ru-guR+vufWRUeZzG
zVf6Q37R5J9x}49;TCG0MrgotV8%CzTC|RwdoRh0$r%^xhzdq1;3HZKx?#5XXiB=Bn
z!Um4IT<I7QRjaM~zVM*@CyCYS@0a)t0w4eadI?+)E_bv2)N4c(Z8z@EeASv>BW_i_
z<?Ph!?0)}td;05I<tf^z_&Uifb*kU8%2%)7bN$*e{Z7d&Wo0Mpctk%JEYSYJynpXM
zUT5}ay?*1i`k<B*FZ(2_wHM*|CQzr(f}BNOyFys0*|}|&YyP}e30sI(HOsEA1|N?q
zQ0ElkO~-8SwETYWcu5pJK7LNI=~E9$qH6zQhaays_;<-H)$6~Bwbq||cR&uWsJi{z
zXS2`Fo!u*{?#vwX(zRco@QSKge%?Fx<HOm!cB@|h*Lz)sg)enDx8^Ewi|Xa1^*Yh#
zb=$S?m{!gRZw-OLRWr9}eLq^_MXkv-QgJH*(fZ4sGEaBeb4z+27!pUU^1H&l3!aVL
zvPlwEXAk;y`rX4vB~evtIg9$<EZZ$=Ii;Vpw&Q?7CnZt!&~nd$XV0CGSgBUOH7mSR
z+I{51-gPG65B`U1Y$QhgfY*7l=Uyq6ZcUHVM0A~5yl3@8FJ2^8>3c~T$}KE)mbhJ6
z?eAe%JXX*DW}P1I=g2DAr}bPfvz`|N#k(_8w(F&0FA}Ia4uLqT{r<K6U1!eTA5W}O
zwRRz#xdg6dD_J4^PX~!8YWKb*YSY(SYFg*l=nWspKC9Z>K<VRq5O5=bA4A(zxf6Va
zxLsM5eeEm}ec$@}3}LOT>{E*8nVw(mf2HdUNmOmuE_7esIj1DvREayQ--mcsMK0)(
zL`2mod*+_Lv*xU2<DRUzn-$Ob*7KX5K5|}yf&<TfV3z;O@v>S6BgYp=<@QH3bxObV
z$|rn(f6?wZv04N9125m-xv*$q|F<s@f3?a!+<)x7|A;!@M7Eo^=W)^Ov8tZCEuAzo
z%6>Mf{B~lMipp2*Ik0A}<F+f_t&>_jo2J|@2?`GI1Agj(sO>21PMyH{6FvHWlKW4|
zDox!Ad<y~~00JNY0w4ea|06KFYRk+^eU~M>l$UcT&9+0deFAeW3R$~qZUk|P_Bu5=
z{TKZ<NTO<>^FT!T-l~4C|FvxR9|-)HK&@f38f|`SKM`Gr9BO&t?Zyd2R2BWcQoZHb
zi7#zBEQzjB1xsf<a`6vIH2wYj(}NP{?vbq0|F$If#!Uofj4r-8-xo227uCAn&}eu!
zl{NAT0w4eaAOHd&00JNY0w4eaAizI?aT87^{E#n=SFL_jac{j-<8Ml$>!*j>U7j)G
zzQkI!miU;OGe7TI{j?NFwZ@0=ARr<E-p|itWuMvS?&*!6mwsB<k$S)Rn3-+z&L6*v
z@Tx}p%Xzup?0H8bw#~2^1V8`;KmY_l00cn5JOaF&CkH$DI^R!idC2F5xm8046uDCS
z^`*ovwQBzmqv`0~w>i*t=Cz3C#C-n8t}}-fP9jcItIk0&E1h$zJN#Lq;yxm(n$>?~
zPAv!=>+?tQ&V8)wP6D-ugH`)q@SaX!55+G=2&5lgJ$K_Qi9{4Nl2=>{0w4eaAYeZM
zQP<zfc2_H(aOcdi--%kapE%zZsNGLgYq&b=tG4>Q<w7E=ivHeJ>pf5CwDaqT)ylWG
z!#Z9Fyx*ch+y@C)h|{%HtB<liUgt9L*B>%|vT1zIJrb|Vi=Sw<L-hXS<~8zeZgcQ5
zQL9?{m>oXrf#7b>jXFWBQu(>i>1uYlv5J3Pt2{*;6<;So!NCoFpts&AozEWVcdhAJ
zb)L7V>1gGN@E`yJAOHd&00JN&DuJ3k@+LN~Jx|iAqUINNg8+{N0=?%M{l4?GT8HBn
zog`3ju+tB$u5-gzRUcb+`UIj?)!G?o{yn~*Kd)8779u(awSJ&A+_NR8ZaF{cm;^PK
zgVu6!dw9>lvDwcLI8UJFaB#bSA^#wdOrUm_h`w)qeTFzqtt#FA#STASYw+)qUbV_j
zqyqvVpq_wL{WIz<LP!w!9|5b*tCG|CU&{C&2xupucmGx1uG72!f_7r=xZ<_9%WSKb
z?E>Ld?JfnY{r}me?Jf`Ay;rhY@7KALZED(Ml3q1>e(`&0KmTO^<cpHsdM{}5Cr!>I
zNup`(LKilSOn*@lRr^nQ>eRw|M<uJ(?)HUs`n>E{ZPX5<<NUl_Kc;NKm%ZDL5xvUV
zKSYV|eu0~n2c7v`GOD8F7j}XG2!H?xfPj+-RHz*MY2=OTlH1iuWs1CjfJ{Kt^{r;v
zEpvf_fSCkD*+<_@272Od(z4TgZ<opFg@Ob80QEp_2lTWhsa<pITKJMbuO25(*VOW^
z=eugFzAro||49OUz(Ef~Tmu4u2pFA587N+S4+3@&I3yk0?JImhQc~=*m-;??c9a>?
z1OX5L0T2KIy#z|^YTRk-^-BZ_4m|jQUO&clAOHd&00JNY0@?}eD^hya_MstO(e%X~
zGka#Z8sZgA(>?#mre$01N}}s;r*cMBIhMie?{U2J6Z+d__h+Ny25!CH?s7x=AOHd&
z00IvaC^Dtw@IniS1O*2peh|_qN2R$N7E91ab9lJ+;gdgsLgC*|nbBgm1U1G%@4k;}
zuk$ZAr+;4ggzxV!+8rlu)diQ|?^2}ypF}jBIdwtKx<Q8t)Lssv?PITZzbD?V8?(LB
z^83BxC0^9&_X*?=1V8`;KtK!vh4$_5lf{!DLBYX@AN2Y?*RLJZ@06h6z@s0qYK<|^
z81syIm}jik4_Wy=;sF5=0D(XRtlpRY$2GTREk2PzlvM3}Gp*-2zL%7t+`>|4iCR@U
zM<Ebbr+@!u=9h+4FMC{qYU7}H-*8c`bGqLne;@z?AmCI2qMlpp)bcPpU##YroIlZ{
z|0lWsl#FVxQ<Kww(QktUbteb2{T=ax00@8p2sn$tr-#~Io-yLS1bvvpgbFF;9MwCj
zeMGUa(6BUN!6T!ZM1_SU{~M7tB6~#6u%O{l5xJ8;mk0|E3vO_`^N6U3>7#2NkV2j1
zB_`&k_#2Chzgc=aaheXg-hRcU_*i1LnpiZdqQ0DGL1e28J4hf^6$t`?00@A9fCQQr
z&D3*7Xe_ZyWn_BqpILT4b$-~w?L<_)-f+mGC1qoY-71#@ln(;;=Jh?arTcHhX=;?N
z-g;s*-9Xoy{d}wxzigLX?YC08vVBkx009sH0T2KI5C8!X&_kg6obTrK8M2*-u77Mh
z+dJW>SR$%EQ}ED@duii{=vwvrUKMr?izTA!H~nrUt!)`gtWrI77T16P2t1mA-u`d-
zx<|*Z{Hy5qQ&#pXlM*9@QokX7w`P8WKWDR_m&EW+SP23k00JNY0w4eaAmBCvGY|hb
z@}t%}iQCm}<&OM<00@8p2!H?xfB*<Qnm}Uk+L<rz+eWnNqhrOdAOHd&00JPuH-V4q
z7pXe%`X-`RFZAe<=1if@1PTsT`hlqHQSG*y@74&`fdB~bMBuUFhk8D_c_Y!PJmmql
zfB*<QOrZIq?+Sfddjo-jgOz^puovQ!+X*xeE7`SuzBL3Y1_!JCb<4n#T{^wFortRP
z`XN#E_OKQSncmn&0;$#hU^KnoW3HS`s<V@*Rio+QdcFvREUi?xYq_0-cjb$k^R3A^
zrPgnyN5&DTGdMWkW5vqh{Ee&bR$CHBSShc$-%zh<l`Q4s2vlqiy!sdHN}WKbUfcRV
z-6oDeoyj5fQpdL-z=nYLK)ntn=WiosE31AiHo10hEc++qs55Hx4gv)SH~T@w?o&cP
zYP*9##pJ+C3;q!qJ7VIp9mHws-yg^eZ`c6>Aiy7i)+hTmN$k3gn5}l@{QV<+<74M+
zC0f-kdyy^(BopA}J&*U&$dm6_muypZ5?+*dUa$`YKtOx~_n#b?W^1J%BvJHA_S+X@
zAOBf$iq^hXqRr6a-w~r~mnT-w{k>=WZ$3#@)^@ZTy(UfTv^$CD%FA;YD?9R^NwazO
zpF4=CDtdn%>w1RxKtK-x*7aYH4I*E|>U<I2;<36+{%k&^M9k$4M071Ouh;w&bvF{L
z)ar6VdfEtF&a)u0RfZkJZq;V>e?zl+-uiFM_yz<(00ck)1VBJ^0%o6M$2xvqc-y`1
ziWd&B0|Y<-1XvSb_5O-@oH<|GXt}cXSEcyI*DSKGONAYTl`5r&-ybG0{MjleUkTn$
z+^W2dSNPlWO}|@7Yg@*WK>BcLDu3E@cJJm~GdB~xs+<BsxQRf)-ctsA=7}M0SG8qs
z^*TSbb|IV?1Xw*Mi4qU*rK0wHU*7Vuee<2HK4~nmOL^k~d)!5!!KDgP*dK93t-6a2
za_bfXbMk&Nc<c5UqE*`tOY=*}-`6Cw^wZCJbsm!~)N8i7rJRvZ5O4wk-uI1GYai?R
z>Q0Z(3BH?J9$2k|vO6BL`V(eneth`UEPqI>wMCtN2ljemB++&K2R{w!6FFGYtEu@U
zzO#-%yl279ceidLqN>(?Xaa@TO5E=DPAPq}+C#1JYUPUX?k3Q*Xr`VsLSu<mbvHHS
z9t1!D1V8`;+(F>u`bDY^yuOLJMFW+p)^QWN!~fR#rO#>?-9of#p#0-|XA=;8zbjVu
zt39t&lyGXtDG*MK^GZ=pe3p~ZyMxw#7^C5h+KcPWArPqF-01g8)*6i;*X!G_-)e43
zet_RW00cn5ZURQvH?zLpDjs(4yC-V;R+W!i*H0tg?p{CdU;csCsf+f#+I2y^h0{7e
zNNsql+EHH8Y1Dt)mGT8P5z%x=<gG6I@@*udYW^*6-OpFeM?}@fiXZCv<mQdU?RvWM
z#69EgY#~;w|NTg`&qn)iSzm8gdz`ym^Vc&}KO*s>A(>7+_3p%HylVBes~dWiC|%HN
zmUex*;gDwUJmpoZe;=Ng;e|~(y`t*P9mD!n%#z7#m9BVe+mHKCg?i1>{5Ph(J>hb&
zR}`%p`*pXsYG0T1s$J~}(secgrF9ri`CV`Sn^~o=_Px>WaAxHn@qz$P1gx%SJmsCG
zEk^gRx2r!e5CiXldL2s6-$q1N|9w#deGbHd_4kuQ$IqJc?FIs6!{OZuRob>$zn)m7
zto;G;fq)hQHxEA5aZ=9>#BCagQRy)w-~6e`Rw9Zf2Ctp@;=XM}R26MJRy2D9`FWu4
zdH4Gr%kF=(=98^NG!3hLecizuF~n*ORL)wzA9TI_ic9ga1p0u3{C=gCBf^6K2!H?x
zfB=sKL|?zjBU3FlU2iyK(UP*UL^SQxYg_-P+r$yxR115A1pyEclYr5Er&;aKfq2%W
zX`OZ_aXJU1{>J<D<)-zYm6&%YVXb1eF0dK|*brd7?ica7^Ze|^ANE>sG}B6=Ruhsk
zzBfH)4RMQ#^8WnCwzItxeu^bfaDX3JUk}*a)e(yl0p8zt?6yydAAWZx0nDdx$Gj3z
z_Tz<}S579?*-5~`8V-6}b-VA^O6lv(7p?;V5C8!X009sH0k;xZJ2l7ZDPb`LDi#N~
z`YZAc0;v(uyI&5jTM*eQ!wv!t;E<Xhq*mAPoqYs&neWM4zpZ^Gf|Nl31VF%<1eC^K
z@Ozrgv;W*dzyTbr^aHb78>j4}i+tp$Givk>0uJB+KlpDwFuNT^{2-u)z?t$VYo-ZV
zDv7GI%J#px?eoz@RPFmk+0h>aFD0y0smxz!{z8qWZ)}r9)$DIe*=m0=M}peJVN~N5
zN7FsEnpmYu{*2#200ck)1VF$I1h9@A>&UT=-1>FosCUk)cW$^>Adesb0w4ea><QFO
zSA5H;5jzRgT^!hZ5@K{B0oMDfs*TSqmX8%b)bq*B8;R(u_xvlfxEhTY_aA57e~kJA
zt_Oif5fHt-cvRlZ{O$6<_2m}t$?WA#y9|0~_?F(mUZbkS&a87WmD0uUAi#n^pmW^u
z{de>F4B1Y=;Q<F0zV`1p#!Wm?=VpV=1j+}8-#WkaS?!`*2oxOP2j<m-f46~Z{+}bm
z<$QnqzLmI5)v#d^W^(BGS#!SKK-8+yG@mJW=*GRYaYPi&viqs?!xnBQqAH%>Qh$y$
z>c?hXZ`Qd^X4$JX-rdr~*JiC*OUzQO_J|g~!KDgP*dK8OYA*-T{08=d01pJzj#q09
z$3xy=1HS~$4{Y{b-WvTRrzpRq-C^1BuP3$|wj@sCU8^i@xie(?Cz7Z-JJY7u%CGsA
zh^D0y>K&@lV?D7-XB?m1xYz70#HhNX_9B-c00JOjoPb&Vb#lDMEkq0;V4Oh6(n@u^
zmfK074>-s~y$9-bC^>%{F-zril$4i!*zK}Ii4T4U0T9qnATfCD%oq1<BU)8|intO4
z_$46v`R!G|?^R*fuvlW1T2+#wmy1*8xf$(;RC~QueuZ>E00cnb!s!M>E4{IV1k!0G
z7e{wqvYd#nHxEA5aZ=9>L{yy+yQ)v4-kXSMdZ9;;G-nEJCRS;n8a@2kDkonF-cCf<
z+rwHUWO`#8(W-&C;Cm1N0p}6eP_Jo~Eal^fUe!8Z(dzKdD>38<1VDf{0;ela+%xXZ
z7UC3b(K=z;?ULJw*~(iv!X6N?lfb8MXU!E|d@JEaS+jWmNZ<I_Ia`V7x}to<Tj%d>
zCA{m@A9p^JH*y<ct*rfm*Z7Le>oxyG-Hk-A@|rK$1p*)d0w4eaAYeU#K1p3m1i!zN
z*sa#*94UeT2!H?xfPirVR;@D+l#cNnAO<rDXkEAAAKpyc5H|>b00@A9^#m^GSrFMO
z!wzDVTAxFt2m&Ag0w4eaAOHd&@NWXV%-i|5O;-M&mwnW&%(I@fW*+`=<VUS{5>~3m
z23#W^fw0=w*B!hOL!72&oqMS@UhzBv7Q2~1#qLu=KWe*!*rje^pK(g9-%5{+BSy7T
zuWkLGZWBkK-~c~BJus^sa7(+4e40xjPrhSavQ61Z>{2;*tk)Td7N5*Px6a!1YK>kS
zrJz8nJK{%QnEy<g&9nd9LENH`reXY7v+ft;YZh78rNRzkmDVVe>D0=N8;I3v*LxD{
zbd7t`!!d}yFBNb0xBYTOT+>Edh^VS{e=k<yANERo0s#;J0sRE_PsmYc)aV^V6m`pU
za-e+bFKK69DT?1Z)lQ@R{Yuw6(`(t~%Ix=1c9yb@iI;gadhM9Man;>wOX7%Vs#ITD
z|E||-<ns(!t6qC}xz2x`on4=K;Q>305fH6@b)WOyygoy=6Rm2DQsApux0kHq`?!9Q
zsspcYBBE%#XTi;Pw{9Y$s@dmW1ops3|Dtt%AaCK#8V|B6UeTTp{q5EJe9Zc~M|(N@
zU#;G6pLKd#&s}CuF!dO%S~qM~I;{O#ZG5TI#J6@6_&w&z$)q|v2`_3l52WvG0<7;-
zX2+-XKFE7`tGvtXazXt4XC>y{Ndj@uTF%}B^*WTCzm15hrN@kX^QR_TiD=5pc@ui=
zV9i^r-<lPkw|)%Rb1P|W%UA*q;NV6-*gmsf$D%z~6W&zR{&s)6t$yyg-f+mGC1qoY
zQGKT1p&R$o#t~7~?D;EZ$1nQ4F4p6>R>v1;o@Ai!&3ax86fZCRdZO47h@;x?wT|l>
z4X@TNgfo{wmfcUCAGUBi5k;F8&D3*7Xe{Af^_H4h<7fWy2C~n89NMhUfzrqKAmBy<
zPrbA$tm3WB#O=ze?5lp?tHQ2fv4pj<vQH_VEdxt->GbAyBB~mFKR4@oCGM<#=Y2d=
z$v*kLhZPUe<7vL=yF#DV-aw$>z_TBiUC%nijJ-Gi!x&;zmFyI?Uv|*-_A4&M#}d(X
z@$E0mL><^ptX3tzu={uUzUZ%3+0}0Jnl!D`?j%rqI0TwEB+B>j8)JTBRXw*V9jp2q
zl=8!B{hiu)l<eX4_YL)$R>@L6jwq?<&mXYg8PAa@M`xEKuk9r4(nsLOMokwNy1J2A
zrTVPIRUiNYAOHd&00M3z@cxm$@v(EZ61y~PW1c^=dd^5r*N@IFT2r|94M`LYJ<;Ts
z1vRfoPSNNg%V)IQeM%Bl%hVpccHNh!BvCbYj``>Qu6|51Tiw)lBd^*C{8g=aXz@RH
zNusOtN>G6nJy%Pj>Xr6`iZ8qTIT2NNJeQC=^P*)$bj{!KMBQso{7OXAdXrYvjmW;1
zSf$!q8eHKd0%on-&>OGS>l;NcOWxLFiO!Q3`&X2Yc<cPVtpxgj125$QJIo;<>O5ZA
zZl}y&kn>}gn&Vx{vF6BdIo}_@ZzZCtsQWfqwa>R%>6pF$iP`b1-QQSmIOUp$5M~e%
zWuCcSI|J=M!`t_)-e*JWZd(89Z$6?{owEM0R6@N&HF~TktW|5BQ<_JuH9mv~0TBr-
z{;=1AqnTC`r|aZtPhE(7e+}VHXZL(5ZOc<WBC3vDv2#+{n(K*EHRkuT=dZoLk?^V)
zdh|$hrqE_0y4J~3@T<YQwh*&a#I_kWg8&GC00@8p2!H?xm`4EnV;I{X!|nX^_Io?!
zI{$r!@p(aHs|-7cKL^(H3e0N}&57A5{_eDV)J<1>%cv1M3G`tO=G20~v7W9xanHCr
zTL{!14m+0J|7Oi6TZ!8>u!rIoBLpgTpAz~}+Z{v{HIi3c3j!bj0w7>N0sr?<(LO89
z*Yf|}e&T#v5M3o=Z_wIxgtaamQSxa2r#28#bwcc_K8<>BBBE;L8x>Zzo4$!yt$ces
ztmB1%UHdm#oo=A@R#u;ztTjDe{6woAiNR}UzPN82af@o@V|I8}>mt;~<Nv(p)X!#@
z8>{%8vOWd*z&T%P&iT??Z|3CvWboGQF+{6sP0#J~F_iMJl_$c300@8p2!H?xfPkn3
zs*PxQsKDUwC9NuIeqlEV@JPVyeSi>uk#$`v>>%J^4+pF3oZfl$YVCB(xPn{rsn!p)
zhI86F0bcW?wOrgDe)zLhPQDVnoj}du;CBB){y`v_K+~d`dd>)qC01+qIp59eGh{n4
zORXy1K=X*yes7hXNCyN!Ks^Dg`X$s`gpeTcKLYzF<ft=h^bTT`{x4<x4+OLm5cRwS
zR(5MA=8h}W&L_4joZY4HPVa$JyC2EvwMzG9TAROf_xxU~_3Mf0H*9K<+pAZ5t(RiD
z70>C_t5q6~XjC^(cCXzU-}ON7n`6>>Mbp;hYc>5l?r%v{9rR-xPiWqAlGSQ=`@%Z?
zj8kg;R(fO{5nc73r^(xOM&A!u`-dp;{n)7K;zCz95~C_geqkpFfB*=900=mVK<P0f
z-~6e`R^oPbQkfzzARrSkYyN<!@ycAFAOHgV5vaQCwb2DGOd(KkfFGb9$n5~DmP%<~
z4X6Chir4jqLl!M58%v-MII!Z1ct9YTfZBP`fx>|z2=GLpW`~4a+0L&cN?M~#rc*07
zZXnQyIq+1LumuD_00ck)1e6H$N$Of6`2C#(3JyH@fs()CcMt#p5C8!X00DLcN+r}g
zRHMgwBAPy5{iSqo%wIx8(?Rdms#xRkpNZ(&v1X4?{+_d#_<N#N`wS?h<8PPUpIe{o
z+a$5;Is(<k!R~TH`XB%TAOHdn6L9Okt;jdqa}Rl2U(BmL57$0?@+WY3#+(*CUiwzD
zTj#t`x?!&lb%>}c&)+ml{maeipGDs{!l?aP&jaQqyw&^1BE8i2?XCKKuL`?{#S&3e
z@3|1XUB}z`#!kOaAb%hL0w4eaVi1V={p|T`?{6efa4_Nrc<xXAxgT8k;KHgkF8;>i
z;%}DTPN3i*f*-K*d&C0*AOHe^2v~g{VxV>D{@<(Z|I^xUxn*F<E}h=oPSom(@)2*H
zzqgeH(m-6D{(V^O>+25Qh#^pI9E|QC&inOF_j}|I1V8`;oJv5{{t>5^huQgJwf??t
zy5d_#jo3+y>hCdEPA1jaNuciJV79*_eh>fw5C8#Z5ttCWs!yZdn+Vj8Lxq%bj_MuN
zKB8Dyny|EC!6T!ZM1_SU{~M7tB6~#6u%O{l5xJ8;mk0|E3vO_`^N6U3>7#2NkV2j1
zB_`&kS?sj?Nd-0&r|G<0Rofg{vYD8zCKipVsPDgjzRc-}%_NYjiUff`00cllKms?y
zW)FV)V?yjw8QH5nnz!yYcRA6j2T#}hz1#d>iKu#`=bmmYe5;AwDwhM44+85)-Z=T{
zP#<xc8m0T*l(F5L{T4&Cs?l@<UEl1_<-N1M97F6<*{;u8cIY^%@@B%D%JxA)00ck)
z1V8`;KmY_lKo5aUH(q+FZLb(2x^B$Zqr{{Y8;PizxGZO-85d#*)Lsr&k2~q9v$zHX
zK;Y2??hOr{k#<ZB(W{S+UHR9eXIo9U-E<>yyDF0sBgALj`X5};GKN4o<6v?>tr*@3
zD?tDRKmY_l00ck)1l&fz?0Lmz$DjE9J>UPackXdtjQRiXd=OFyIfO{$u&tG_%xBYv
zZig-8v{<L@wyh3SqR6QdC2>1Nv=zzOc96n#tM0qgCY@8HBS}dIl2Sh&c7C?`HqBf!
z*EMrp^LhSrzwVj2uJ`r6p4a=j=9+8fGt*<*WD!^Mt=*|F1Rwwb2tWV=5P-n$32cA&
z+;Z2fpD1K?_f|Oy0SG_<0uV5pK%-`RwXV2FiHNGTZ>rLxNApsF!Xe2xIQUNs2Q;rV
zS)g(p%pQ&K4gwG`i9pL^YM$3)$^;>+Ce;C7AOL}#37mN6=@mYj^o~H`kYpU}9EEJ<
zJAwTUF8ZeXK|-K*a7a2{+rKenX@|w7!d6|j;OxT}H!Kyl>c#8-^VRV;lnOshIv(Pd
z=l32;zx{E`BfnV$bdMv;N{@fFxcbk(+m5MbQOmwJ2VcCtdSU%Z0?h`8>_@C!H5`6=
zVY7oamkL#CYVC*DZ!vT0oJj(;n}ezGg<rcSz<pr3`@pvG+I6e57XoSsB=x>CRqqd0
zQ|tV-!u-tc`5En}ru}5!^G^C`@5k50&6E*5?(eis&F>v&OyP|m5P*Pr1Zv-2=f}bS
zED^djt(upvJZ4JnFO!9=rsXf?Lf{tyy5Cn5*1qYjQ965{OX{9DMVO+d^TIy}K)`+i
z=YKZv-`$^nPuQZievi}7?}tasd;R!j6GU81D-s`h{;+LZzIe=WRka@_Ra=edT2v~m
zSIfW4;{Be?sYia^ZE8uWuvPoq)&IgdT}y?ns=A+H4+2pL?6`LBn5oUC2vr)z2b=Tw
zJ%4Ze{-oPP)$;{u+wtCa(Y>8?J5in-fesIK+T3WsWRbSYdHsJ=tBT|QYneR|fB*y_
z009WtPC(c5V!G^_!aM5sJyqp1g#&&-00IzDO(5)ZE~GpCuZ#aZS-?S#gX(w<vgh|+
zo2dJP%H#OX1X`YU><5Ft9xw7$mu~B@$3dMZ3RCss@++=s-lIhL>CUYgye0b{S#SE_
z5=fCx1mZp~l$HEp+x*~@CLDdjil5(`pOA+2i*$An$hY~d>8%p>c@xv>u}{gotG++;
zX_1zi#sh!yi$LEtcYZSZib*1@<`*4mn=b^GuD<e#trwRFSsiiIq78?=veD6{`@Z+*
zH)n6CnA5HK($3Tq0$Cu?_nh<Bb=`l8uw9e7{?AG|S%}?T)!~wHy&jsgZL+Xkbv+l{
z^pb=7A3A)NqgtEa{^LCdFC607t{=?&q5tDOJ`hoLS24-H<OulP*Rf^)C7+M#Sp83v
zL|m06UUI?%14l2IC`{F!Cm%ZIqvk@St+HH6pWg&tZvXm=9j=}zWHrC3QF{nL00Izz
zKz<O&x9@cMwd%$DZ=KzG*hFEAdf%TDcH6K1?9VoVtgIi$t%vRFvtIw>@*sb*&vB;B
z+tlv^nC@5Hv%^oHbzbxIDHB9e8rNU)>ml&M==JYhcIIfIT7$f`^}Ff#E%)8~pg7<-
z1Rwx`v<ZZLUOTCDN%2VgeUP@6pH%zQzNt!&9?eSy3Wt1-1MfPMt=qOfU#x4LD5-JO
zlyW9?TGHY8>pm2=X`^l}ORsJ}Uf8NjntrtSk?qC8R^9$Y?E~w+UMlikWxr3TD!uK$
zkJmMRN41;Qczkbn=Yvb0`phv!-#hx6`llR!bS|tuQho4QM>jYkr%TJ+)%bxa!>i}Q
z>Vq}r*4+E}iaA?#<F=zOzkUDDouu^PPv^ZN&U)L?rGH<2_D93|4iL8Jwte3#SpRIX
zh^lFgBb1wM0*$x7^;o}qCyT6W)bH@<YJFYru~uzYSN*XI0h0(so!>{zXHwl&`V#j&
z^vg!>-)P9%Ng`?WW6-{MY^#1{N`>tj_P&ZA2i5y)_iog#_>qYMRl}j(idDI(O(%+^
zRCPRH4+1g-Cbsx@ul0va5c#Hlj85HWa^=^DOcJ)Js_zJ<W#^LV&+k}&*w62sLA7x3
zi*vvI56`dgbnANKg{f-mJZD(ZqJFRaYrIHW{n}Y}f9dA4&bsE-Ndn!OgDgkVLjVF0
zfB*zcB%tc~m$dA-O#A#xG_GDeQOIgZ)m9_A7L|%1)%9FeQok2aU5=^OZ*DxG5P(2-
z2^{>Vg#(&b`uTfSd7`ClN8EqW(x0!#E!Uv)tf|N2zGv3-{JQ#4c1Iv;27zD#ryluv
zx2Ywi0)<1IIC#Hw*6^akw>e?e8T-2~PtDlj<ct3Kuz3wjn#^@f*Za4xne~r_6CD%|
z#6j}?Aoz?;3Kaxw{r!DXeyd=N4f_aeAJ(vMZbpefv(16`0Lr}w5MKGgE9!_H^>>I#
z<%?P`@<9Ls5P$##AOL}UCD8WxKkuKLUMx_%IOJ=*Qf~<Cih%0pY_WIeC!?>JB;b(W
z9Cp>$*q0uGq~9~+SDvolWkj``UW-sN1Rwx`Y!dJr|Jk3;d*9#=_g~&OsC%@r<-JV!
zY1ezSXfN8!r1!q*9!IecfiMK#U4QfPe++v|*s8XEf6Db+_WdP6>4Vz5-MCd}pStQL
zVXJN~c<J6ly9t4&!y)PSdk4ic#~}a#2tWV=`9L7*eM<82y;Qyz%lC2_2Ywv~^5I-T
z9U%Y#2tYtR0aM?Hj9+@cH>5t2u$V;xQNMQ@H=iz+QNPD-TE5}ucOTd2pi&{Ly12%*
z+y0FqOFJz7`MayO>T`YmdC7Z|gzE2cCvC6q6*t`Y%&C(_T#XwK<cGlS2*i!gwDa$-
zZaR<E*!a&q>VCV~F-`Y=v*whp-+bc4)gX3lYo9pi$X7oUsx_!wjzd5Nf#IhYHalqZ
z&+|SM4!em16|wfeHi$Q~`W^3`%#Zg@($49&^Z$;4Vg6rMdi<-!)hCI3(=cpQq$yc)
z>F8Cb4i{lHZke_{XHatRE%l!nT_S8%+juACZ`?T6m0#ETq%MDD?QU*2r_JTX?+aZj
z>yK^exo$Ayy1}-%!CwfNKp^gYY`^>_)D0iZC2-y?t*&Y_e3ZzFnoBxAJUit2*5|hG
zP~@1dC$`$N_T%@zD{R$vD^}&EHk~MJ)1=oKTeNF?=++j+BCh5~@1>RyfB*#I6KK5s
zt;hP^J6Y(`_#R>b0`UnX_5D0s%5jPI>G$%4^#e<UE_KTZ(tLl9VGtjVLjVHN2@J2_
zV&>L4lZ32BFOi%OFqeR?`F>J%OIIE<CHI%fA}LL(B@bLae9XnC6$@Ln<<*zH{Mkt#
z3fuMJ5%XR@e%S<}Ta$__%7Fj`AW&;+zbl^=4+uY9-ulZ6=brzXuw8#z*Y~>7myH*;
z>h@s``{rhp2;213BR}spwWL%erG7oSWkK%b;=RTT+x6j+wMRCu@bf%GQvJBF9|91_
z9)Y;;FYtRV(Ukq!YcXmB0SK5zVCl9FdmPkxqR5I~Tz<th&3lvx-D+Ap;tvGUB%tf}
zfl=+MW?}2^iM@W;zP-+mga26~Ojlj&K&s=yRC^^=TaD;iR4SsXsrAAy2tWV=5P$##
zk|*#=t<(Ry_}`O7+M0aLDG>q?fB*y_5TC%XsfVwtP^DO;rSWTk1)T)qzBf##Z`g(a
z1Rwx`<O$S%_JS3ky){83rODTj5+MKq2tWV=5P$##{+EC$@7wubpOXCFl<#6ES?4Ha
zEgQVOxW}}~LX}4GfjssR_;$|ea~HfZ$H|&*sylA`-TRLcX{&t^fyenwAgTA;_?43{
z>>ucQN1vv%$BDSw;h|2O8x5E&P&g0=^aEYvK)#I2)Kf13RoACc?YOmT)NHTT7569+
z`Km65s5)f%8y8&o-VDc2quLG17q#7O%@@QY?(x2D?)+r*6_bQ&P5S#P_fMGnx3O1^
z5lO49apm<pZX9{@NBtcl`F8y~PTSwtkktFqyybY~l5Ge;00L17B=tUSRpsbf53t2<
z)OOBlzPRtvBmWIIK6C115tYOC`^yih`0ym5OS2lcs_LbC+*0MIYCEdVCu!L+<vAeJ
z{fHVjy555t)gM#x>H2P*tX)%h;72?HQRfYQ`QmvKoV(qunbTf7vP8&gu;pke4to8@
zE4I!WFHF<6RSsxSvG;gki?%$b=6OA)Oc1uJ?)Hxson^}CdXHgJc9)eN|7vmdNy1jO
z{r*AH_s#suiyHq?^QHZMskr5P@23TAN981&FK)fk&hPh}sP6r;oJpSyfuMes^zmeB
zO)2?Gz?9$r<*Oe_#hvYZb)S!Q+m~(Um)pPoVu!0I3S0H3&pNMp`jiR6I-Pdy?Opes
zD1Ne3Yxi!{uK1COLY2zKt*-Q!d_JmU^*>D#C>)3b`a#;`z~FM#j&8W>17VuBFCN>f
zUzt*2yDl5Kf1@F5e_kg}NDuzg!U4@IO%`$0_IY;F>l>7}^LtfvUx(<nFZ=wCuKA*$
z-Al&xdT7qJ$pQ_>!H;9u{Yl^F3+or@^b**%?|TL7pDh-)Xj1RLjayFCRvTWw#mud9
zCJEi@=b!F!UHg9JvmXNaNZ{rBCf4rPeS*k$Rn_*V?lZaa>q909)vC(BAUmfFy63H@
zM-~fPwd8~c2991ZQJAJd+*KWq`yH3G^KXBm_JMU@FBPhI*ly>%TUuS!X80(9!olP?
z(B02eJtrRI=fM}RuU=SxlCVwF9=BDM@Ao{g?)^b=k@j)F`y5I6mG*PE`CdNvL3X|m
zl5aec>gP$7lhpY&s6MCu;rsPpepw=t(jb3KJ??k@@$Ogs9>28hdEe*7_UGNEmXr!O
zz#+{zFm;^7uV@73ciGzYviC-aq%@kB<bnVMAOHafKp>w8=w5#gvcG=)88aK*_KA~q
zZ8q<Yt1lY6z_CToxNYjTtq*?SWJN!yyXE0`&l>L7su#Ynvi#WG8;-4d!NX1d)Afvd
z9Nn5vV>fk`C$RF^lLl4)?tRC09oOUYqsA^AC2ZA+E&knW{UH;Ct-9p!L7OffJ6_nX
zy5IMw%YM?|ZIK^o$dM%iroEq9_wO0K_HA#=+Ba3{(W7~(Fhy<WX^Q_z|IX6%@=R$L
z{Ln+d_U9AYzcFNKhs8g?r^9kh`}rL`(G_g9bn{tfU32RsVXOMR@89(O-uakq+etka
zsjD2_<5h6aBZUqEao>AndVcS7tfp?uzCXt={iT0xa^hE~y&=Ntf=j0NKcvejVY}XY
z(%<T|sXs!fR@pum^!`v;d!&bejRa14<mNNFKK7x=x-Q+jclSn5PY|Z*#rtoa-FnzW
zVXKxLd`tajMwf`Js_E+(?Vs!O&r9B$By3mR^HUqgW_*SK1Rwwb2tWV=5YR`!wC85<
z=lLD$5BvE&;M48U4i9zO+-SgLk+15$-#0B){k~6K_Ef#6%eA9#MAT!s&t3g5oYS>b
zWKFl<Uw%l%hbIYJHSO>9#4TUf`^oh50)MNfuQz7hzpSof%ht<3LOBwLfNksaaq3E5
z2tWV=5J;au-*e7i*LD9X!d8`CS4p3^*&ehk$emog*La~?A09F9_2ZXK5VoqS?{cSQ
zr_aAc<Lbo|Mbc_^wBwyA1bS6EXYZp=xnE>me>$>n`}q&{6J9m__eiG1iL4)e`n^11
z{lF4otIB>4T$YdS^tyj{<h6fs`4!hR?@=PMuDaVzmHnjW8<dyMJP+=_a*x_?qJHN>
zwmkR#Fk0B`pv|Si%R!bW=^+3C2tWV=5P*QK1jY^DtNM3`jTEwKYklz>0wxmB^&F7t
z?}x&_8*Y5&)X5_29Fp#HQJ+f<>!+^wbz#?x@4bhO1KD)4&#jO?+V7}j+a=%A|8DiY
zS1lX;fk5?e$oKf7{t)<uKw8hclvHgsqH9s9(4|SW?!hNbIQoPYlSNcb%1_FH00hDl
zNb34Oyho&jz;6V0TswEn)MiseQu<rT{09Mf0;RL}xuouiQ-tjrwZF*|%a0to-s7Um
zZrZKTs>yvN?T<J#mz1`A=Fn%)J>{re()#E9THjRZjstR0_0faYmp<Kc-&|DP=a%||
zrcSP$OIvrG`Oue7t@zrpP0Rf0j-_AL9_!eu4ffdh+Pe$h5=m>?<BRI@(^^kbU9PJ0
zN7{Dce&0!TeAr^I*O?!Fc6Zm;ow#aCeen|l5P$##Adn>jQGd^uwR~A>Q|bZ%mw>JB
z^VQ|I%LRpiP68LN|Ib&)-%u({SDg%^*nZ^4L#w>e@pXa1fjFQaxZ^;Sv1C_ymJQxs
z++*5gk#*fw-J|UDzIP<wx1%s54+Q)O^lfwJC!?>JB%-RHUG_sD4FV5s{r!D||4}T0
zH0$4Mk6OPp+Ke(G009U<00L17==y#gc2EE7;(t#TaFF8=HIB&#0SG_<0uX?JJb~?Z
zH+=N<TVEEo>FHnnu;i~jpK@%|3GMg3cFwtjh3)!NY2O>`ZXGJT5x1kmniEzPP8Mma
z*RQm<FW!Ia?AF633RI3m+U-X95P$##Ah0um;q_b0+&X8HK;aN44zm7Sm#X^h-22$_
z5*XLx^P|Qt93|4$QwH7h*3%=4g{}H<$=V~ESNQpR+M3kcZdu>{X<gszMqf5w<f|?@
ze9)$g$Bq}aX`j3LUpS|0sX)`^p!z*1y6x%u9bCTK&H8zQ`au8!5P*Ok1THSW;+p0?
zN(2grIB~FS<o=C@teqrKIG7j*sz&2{zkZ?q<%d*!c#^P%gQ~d44g??o0Y3tld_JmU
z^*>D#Nok{QElaO%KVAfB*!vYhcD%<Q`s(1ldYt>A2&+NA%k|@$_5I6AkAJne`p?%5
zs1%2+|1K-*aZmjq009VOl|b#As`Tj5yj0kxS#1y9^}1wSuZQMrn=In0s^8n_`hGEX
z^?n~(6<^))j(rF~00I!m7J+svR^_HPohZ<qIXrRlf7`jJW6|}6^$IE$R4ORjtEgpB
zLAhW47FI8;S-5{enI1)j2mG>Kzo2YE*$cL|?paj$LhmLEobuU<5{<Q4vb4^QveQ;N
zuI}66gdQFHuXbErT(#qfk((>zqUyKv-oEgKx!*fkRgF!9X$U|70uX=z1Rwwb2tWV=
z5P$##AOHafKmY;|fB*y_009U<00NmN&}HVG-><#0at_@!hfFs<m4g5TAOHafKmY;}
z2=uRi!KtG6C?~E)C_2?FRa?~AGH3W(!cS|>{B*?d#nT<zbbZgNU!2x$fs-{&HCQMd
z0uX>e9uj!1Wz9SKRaokva7Z!^^02E>We7k30uX=z1RxNJK-6`gNQIFJ0uX=z1Rwwb
z2tWV=5ZEn&Z|Yt7PKA?J3R(U8**8qPYs7lTtBam!^wgdA{@?^@r+>Wm@1bipIkxJe
z!7sLJ(|d!HuX@R!J5Bg%+&3bu9x=N0nYVxUl`utR<6Snr>UvhMad#`H>3o6O&0)8L
zF$eR8!1NZ=e)r~O(}k-oSC$lwY&%VO^^CLjtvR>aU=da;&E44Pq6Hs2vKn=rE~$LC
zY~1_4Kiss?Nn7(KK&dDMAn^YQbgo^Yx%2ou$2R@{6|n4|6Z?*Q?vSaQowW3eH{a`7
zH0(!5R+Gl;_q@2f|Lf9M9-OiCL*eSy@sF1tyK1rU>e*v&?)!M{kA<ur{=>|!Bln!@
zpk{GMIu6q+PxiXpwDeK0=e@jDzrVZg-^H97&uw**()Uj&J)mI6PmZe(JwK-Zg2r2%
zr1aH0|9(P~hGlX=+GX)qA1%JVY%c5SjjyD)-Sd}jRdxNU>UUX5bEZ59KmY;|NS1)#
z`NI9qi&y>5Q|@<QT=nMj+xD9;zJJcp$;VgU;Mk^q^|WQb?Z0PJeZMm)d(Fz6QSHmi
z);YGTZT(f1zsZi;lm-C^KmY;|fB*#i2}E7j&w4(8VjM|>fV*yh?)+Wdtn)}C)+iGK
z5XdhAXFvS?g#Y|!ri1><!It%C{H(L1?6j4R=^XNFG@-T-*gb)h4w+wc;7wbEta`1s
zeXe?gn)@tq(0ClYT-i38zyt4|cj4VFD;KAwX447p-QLcTAANM$dM9Z;aX{Tpd(B$!
z=+f{6NC^Q5s3zb(AD^@)^&GxzIeyQPw5tAUwZGrG$w9+$P#ve(g8&2|009U<00OcE
zo}YHheHZnfDNs1%Qyj?l6I0Xg*Lv!|2KK95tV&I-Cw@Ty0uV@xfcu?X+kZb8*MHso
zzg_?2ao7D=tvIZ@^$F|Oj$@B&IJN6Yk(Ac|w*GzM&Edj!9X`Cq$8TS;$dT2s?Ls;T
zKmY;|fB*y_U<!eze_4Lp`EUOuY*+7lNZH=?`&murVoJZj4+ua20uWF`U|XLld%yJj
zCV}eVpr)0~v(ohHmgk@O?q(6B=6Q}s5Ku>;^~rA?HFfF)k(MrfVN~(N5hVgmhJ(7u
z!4d=@009X6F9F%#b>~a^j$<!5c#mew1*(_B*NdEgjjZ#YaP@y<Gl~C$zN>A@@uw@)
z|LTsHS2<y|S?z<~o7{T2<EKfmwW$o#?g!Y!&IVuI=lWEu`&^ps>i7K&+xL6FcNMih
zy4u(8J0xNEyYJOZ>w61s`L=6V-lnBfJ)Q;Gvtg{q$NVMW_nn&q``y~<pOe;#eA6d}
zb{X6Bs<|TH)UOBVem>u?T=qi%0{@r5<NLQ7e{S(|;p(3EU3Jab!&eGZ_5ap@JrIBZ
z1R$_00=>TZ+mB~H_N74KU{W0HDpuK-9RihxU9??H{9d4NAP(pU?l@rD2)AO#jqA3Q
z`AVR0AP(&52d0f9__KQgjqiAJzhR$#E>OEU?7o#b3IPZ}00OoXm~~~>D&sGjC-Plw
zZ%6!x00bZa0SG`KI|RD-|JR*w)?V)f>7230ox1tCuN>30|C!4!8aio)gYKF`b|Tlz
znylD!&ZN~<mJ3<Eds3U<pM2xj0@cI8%xDSk)wrk5Z}?c@5`n@Yi*XRXCy>%a0;WC}
zY9cFqN}s^tKg{epa?hzwSgka7W2cK2eC*hw=@Z6-ED`X&f5EivZ|Yt7PKA?J3f1cV
zo-Hf-cKge2?sfVKfx^LrILOi%NnPwA@O9}c56)Qnp+Mn~g*dq4`RV1a-|Ks^<${a4
zSHF5`WaWL>$-EC6&T!`a^X9#OKWH5@?)XanU(ZkW%oNk{)cbyJG3ByVrw-mCTpiT4
z<e=3jZ4~z&KJnI9M{N*Jjhk1DYPZ@~C!O3g_s$Uu1quh^ApL%DPpe0cJ88$K!W1oN
ze9ddc1HTYi(}(Z6f5T}fd?HXgIHYs7rBn#mNuc8D2Os%EtwkcL+F5J7hJYRd-<3P%
zW^uuo|I?qca{t%Vdgy6QtU>?+5P$##AOHaf<O>1Qz5|FqkCz_1YO&D8A@1+u`kb(+
z_ng^33f<~{9~IsCd$I6<$Nzpm*p_(6m-DIV^*p6)vjw*nRLhyJGnO6qpS~ZI&6%p-
z?SK5@Qw~_|*sgaBuh-;^6IVI5=${rnfBf(3ee3w?qQNh=Ytws!W1G6;#)<d*d18a{
zKRK?}{kxb`<GHO43J2nVe!w`e`*FY>M^Oks00Kb-4te^>3%ZLhoFEOd!*K{e00Izz
zKo$w`9+cg`2j$XhFYY*P+A@LK$>H_Jf0+FGrZtXj+N{hO)xNxJorA)GIG`V-Jq~1X
z?n9j*009U<00Izz00iO_Sl_ei7pHYw;K*wHlCb~*2tWV=5P$##AOL~i2#gwh{oyBc
z{7InUIQ&MA{~!=dK=u9c>NW0e<usiyY*X9cv-rf&E@PWsHCNcGwtvsU_U|^o`^o5b
zGrAW$s2qpKx87c(^;_!%x_b`6XChMkLg0Ymf2nuj{G|eggG?OIdmvyVfmv5}tup?i
zdE$ae_Y8fn@d7j@4$ofHWXF_u-geT~u=7vb=iB(d?4qHQW;nKSuyKsVX9z$50uX=z
z1Rwwb2tWV=5P$##AOHafKmY;|P*0$0i#l8841Y^R)&BJ_I92o><)o!i{Wc}vYb|Tu
z(XYZ%Cv81p{O!jUZ!c49n$F&1_6vXgezVAm_Bp-z^(Bio3Ee6?QtaE|gdQFHuXbcr
z_V<{6>C=jTKYvWyUu)*4BZe=Y?nG5p<FjmBksbnJ2;6)4#9Ln-wLze8_}@5~G41kw
zN=L2`u6A2^{jlLDyzgXH2i<o_m0=A&cT}rs=d*nEC(Ky*OBz+@Yg_Du<q|jD^cK^8
z_vU5OMbhf^`<r@KzEk0(l|r|A%Viq^5P$##AOHdT37E2uZJQs>Y9I98<kribq_tpe
z+ed~^d(|;T!~ULybaKDDHl3IKBOf(@Kz0bUeXe?gn)@tqRH^M`Y@1JH&EFFT)a|s_
zto4qqYTI{tZ1cCujkVeq+_ubdHRyYoe#dX@RQra?%a%L3b;;5?JIYR5>Bwr_akS#<
z2Os%EtwjROHix)zl6L;2--oCAeQIg*qkITJ00Izz00bZa0XYI+ebcB`gRj;)C>%_V
z1G%WypSt{2e}A%Z*RMoaJ+R-co&Gs#t;iQ$S+mz)hgW{aF-=Q{-f-@?o~wjz)gOV;
zN;vJtvfDm<ey(GhMyn0Ek|OYjX75#-S~%W8;XoYN*AM25J?_-a&wb^f>2gSFyrvvG
z2z;`j-7)KCtr4o#j@siT1Rwwb2tWV=5P$##AOHafKmY;|FrPrPeMT=>e^Z%aU1~l7
zJcIxQAOHaf*iYcPV;gSm{`B`wQffajY5)NUKmY;|fB*y_009U<00IzzfS16Y_g!_(
z*~3?gu<xPt=4TrM5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb
z2tXhi0<X2Kc}KqrOC2;Chh$obQcNN6>_ts>OnK*R2Q{06De;0I5P$##AOHafKmY;|
zfPi`eooiQU?mRxv(XHw|$07tE0D)`}IO&l2MF-xrMY#Ihk+*!(Ve!`@Exq*Gi#tx6
zwoItj#&<lq->^?V7rOPpd$)CX_LXwQep=A@n%9a4ej(6oa+opg@_kB2t`NGl*EfIr
z@yy4*6et{a6$jZkJ5dV=KmY;|NQyxJGnZX7bkYpR)ukhwe=>Z%5OKB2jkVeq+_ua?
z;b3AMB-Ott2Lcd)00bbAB!N|zzkkxLEx#2TZauqBpDr8zzXq*CCg~$(K>z{}fB*z+
zBT)D6Vor_cwmLy-TfJ2KJ7?^1r*3}kD@V1SJ?P<2-`?X>v3308rN^#X?5ne`>{?~~
zMe{^f)NP*^W=(9rXOEv8+w`Qe8^*NCRmeqExBe&uAOHaf<OhLCC-=;~bHqY{!htyW
ztsmq^@28d!&`ZF59va>G`S7Sb;<nZqXYE^aZneR})wuS^4*>{300Izz00bZa0SG_<
z0uX=z1Rwwb2tWV=5P$##AOHafKmY=}Cvd@}dxpN(c!AKR*FO4co!7UP&AIx}^JDri
zXuQQSMfdG+LXVF9S361R?puSS5J-oBuKtf*z2j>-(L$*ZNS8p>7In7F8UB_)v&~`9
zfm_d>JEKVCo2J{lC_OHLrvLcxs1rY2FLdi?4JWs5ebgo~?y2(|K32Ho|7-Occeiqy
z&KIusuYbX*qW35#YkKJHv-*`iyt6P><HiyBApn8w5MaKV?fEM08?Ak_)5EC=1Rwwb
zvj}Xw{HC_A6mA!0d4)F+fB*y_009U<00Izz!0rfK^pBfw`1|}FLbp~Lu;l0&?`;#Z
zx;x)F1_1~_00Izz00bZa0SG_<0uX?J-2~=!DgNQz9#e!@Yd5~_u;#bU7Smfy``w$D
zO*dAf(w{Ny@_kB2t`Mea)cTMQ0uX=z1fmhxyLIRL?tP$4u~!!#bw$6<hx{n~^!7?;
z-QU08x5BFnH(apb<7G>PpSs^sOIn-NKIpy4t(QAd^|(5d&K9SaI#KoAPe!+!(Y@G7
zTMO2<ePsBwR~_3lDtA-zb*^2Zx%2ouCuxl~Dv(Pzfyv*V|5lTazY}rwisz@7zkaXp
zMOu34wHJ4sHf@=Rs=7I29|91700bZa0SG_<0uX=z1Rwwb2xOJOi393(+H2N&$26TY
z_PA3wKlhd6)vvy3RI9;PYaP?I?R^dJ{9TzfjxGAnv)@_$%m+U>akb4~>P>j4bc2K1
z$sw!b9QA?#1neWws6xx>eOfFNX{&wY@E8IRfWXcK;=WJznlr9AOdK=Rv0bA+hhR#+
zSyy(gGXA1@B45-L?yes@jy<m7)UG3iZTfZTD-X_C`k_#zJC81GK>z{}fB*#a5a@LE
znF~j(_|eH1)k6`h5P$##AOHcg2&DZTr>HcxTyRnM>Q^rn=&u}>EUmMn?6j4Rsrvpo
zLnj|!eS?F-A<Z}l8gCxodV7u5Z><waYmi^w<BdlSzwwwKHabE2#d{5|?eXR&$4^av
zZs<>oo<IKg^}cm<>7v0ewrkUSgQH8+;=8wV<VPP}w%*Biy{^%q+3g2?>!9gyc)8NC
zHL8r;=mcrdI2QJ}cbsG!0uX=z1Rwwb2xOhW^V4p*@1ou_MOgiMk@K&Sb>0)U>&lwF
z{yMz!GY-0Y4q1<X>IVS`q)*_9p<Tu{y=ty__^$gmoOZ$|s1FDCcQ&f;W4Z5<KI5!?
zYtF4USlFuPeDLvqhUaDoRE~qH_lQ1wQIj20-g(<WchA8calb2f%FW_}F(RtE<7QMX
zf7RcgY~1xLfx;n)IB?@0g#ZK~po&1+^VLlUHlBUap=FE1>V=2aUD9WJg<?P5)@RDz
zFFn6Wgw>bMyZ&F@_FN^xYS4FaZr}3JD}#D{FHksyi37JE-`M7z?!Q}7rueo&o&Qn4
z(K?}8-)?`|&Am=vAyn%<tsXh<q#d6MKXtE<Q8T|grK(-rI7T4=0SM$ffd^-IJ*?X;
z<%(@l)$P@+%o)|bylkBlS8L7ubj0w*)1B8^*1V%%g{7g@&4>QV&sGk8znf~gvZQEa
z+iAj2tG1}KWzO)ogl-MnZ{BP7L$}Y*C|m66x@w&kHQA#~aga9s%ktaKfBPq4oBFkj
z?EdOC?r!BYoiEZ>Tl)`w+dcl}OZ$Z#M@bitz!N9`H-AybqU#Im6;vvyTu`=GQOlx&
za=-j7tX^2NaQ}ibJ&Fns_+`6(LD_<`7i?|av#9We-c1%b<;y&G=ZeN<_ITXUt-pJx
z`n>YzED`al`&;fxAHLRQ?3?qQq_o~`Yw8uuEuYJpPU*7tx;3ZoaCGZ$Etk%JZv6Ra
zI$o(ORkr`CD><4a)%`Y_7xw0AIB;d>e;n|o6ISi5DW0n!AUi&&PVd$sYQKN8${F3R
zTr%0QRaNthYR_#4)AipCAMJ4Gg>!{htE?{DcHyBDMOHLwe=t2?-1zdFrv>egTdrSz
zZyYAQ?fP}_c^l_7c>G%tReN1}*=<w5SSeJi>z^vsHO{N@4?DXj5Vn8p-ikd(qsFT}
zHPqv2RBo#B>8VxvRu^s`_{oslmpV~3{Tl4-!McBrUC_0`ddF66R;^s;CeyxmvZgz?
z(r;Ty%};+TG1`A&uOp&aFH<IHzN+hZRDa!ks@l)BC#zxS52mMUcHmP-SHEGS2vSw&
zGu3vYo<B^>7j~VLbib}E%l6l>=~cyN*z!mR0h0(sUAJ}f#jQ(=CXaSH@x@BHd{;k?
zrk%Hf{4st0<LBSvjnD7XzvtH?tEy{W@7nc?-?00G;)>(DUq1=US$Ddy*RN62>Gt2x
z{;gl0-f_e4zZYq#>iZ)}+w&WDliKgsuDbR|y&snE`TXKiS34zT_s1dUE@)k`Y|c;p
z{IKQv-_~_-*Y8^wzBuvdXSX<U)vq1B`&F%HCS~W~KVEapwOcnjvg)n3-}a_U>izAx
z!P|vyO^TPUa#Wo!{n{a^{i@d8up{>zqPuQmyJqEj(3bPN>*s@KoUmWRch)$ztEzD$
zsqvC>%pl-*ePa54Rr7|l?L=KSGA&=+^RMpS;9eKWs-ItR>zj6dRp$%cc1(+NRsQI@
zo^tK_#jPp(T|afHDfKkP4^?rSmK|MjlC<5luJdHqxxDFR&zq*ExBdCxfy2++w`!Yh
zj;$*D9t8J(ysNtAP40R?Ry3{ij_vCxVf{~wZ`^XzI{(D=TQ+~v=k>J8ljY)<KI(dp
zpS`&IRgd$2{scWA=a(;Nzp3*dS--M!KJ=?c(E5ej&hGpJ{q;R#K^&rvCqZ`O9`~;I
z#<k~{KdtNJxa9`*hp@-pev`HajqkF@qt-|D`OFl1LGj@73QFhKkK*n(g_CW5xb5cZ
zwx_lq{6lf&Vn2Op%jmls9#O8?Pp9;_ug1HxwhBLu`W$l5?>^%m59?Rda^l9fs{BFc
z4L1&=YE(Yne7f2rYQ0qFi#l&ol}~kj(z54`8`ayYc<cV1psx4nrq!-t%k|Fl-R+=p
z&h+DHt>2kiZrFIxl};BIS$|c{1LE2VJI@W<&%7LD?;q((@AtbG)Av96@PO`D+)^&5
zN_C&V{OqebuX*DuscqSHSX0ug8t<ZVjypb)Kiv9lTAZTB1^JR85ca$sHeE7a|5i%S
zI;5)O?)64o{mt`V|J$eLP4k_1%M|;m>g(LN_PqWly&d*>4$|rQ9h52amZ*Nj{r$8l
z`E7sh(RTmi#zE5KreAro_e*U_?>+DP)nD~~RqM#u(evE})4Ar$yhYWzgDN{oaaTRw
zgtZs-dDgJ@b)_@yeN(FHDSIAORjzGz;$Cm7%5OS1*?7{OUN^6(*MDi{OX~cR^>X6I
zRo3|UJqPc%KQ2Gh@`qiQHYHuy`{}l%i#o6L=1VKCyz>LqeyF-WN!yO!cw*~*Z~WP|
z?OorpZ9C|>U)}esb<clY`?BM>?0HJ{cqMC3wtlkqb*EPyXCGX+smAmlsug?njHOlT
z)$Lur*iWObb9nhAeSISvucoB;z7NH_?d7U_JMHtks`8VLKet_4zx7(jN>@}Wmerv1
zP}t+HALz~mOnI-(&R)rHd9M?qZp)Xv`y95-`|wxpeb~Br?LNBn9}W9Ff7ItK!aipb
z<VW1!0R-9iJ04VS+~@wo@^RCJZHJ`Bebo^vdp+ZqK5oBNU9YJ2R9_z@ZO^?9PfDX+
z$EB4oYFwz!m-c-bw?2OFJ8<`V+sTb*UC$M{_HEU;<DlyNZabT<vhAH#eN0);k@X|$
z^;=LrZ#zUi?*x@&`f=Iz2ukla&vN$%<#&&V?N?sDVe{!a52S6^d*1VId*fdBcG&g^
zo6hyu?|n#qzt0NV@7j&3Vf~MrPS!uw=Vk1vT^9;k&zEg))xR6`>t|v2%dYdQO7D07
zPW68GdP9}!{%)nLeOsR!QSEoweyA#b!`cb!m#%cGIVWvT*L~2ayuI^T)3@W!2cmNF
z%QtiKmDBfIU!mAfWyimPqZf~#TW*U8Qa^vv+V5URrKM@r+tzYT?dR@wyKO2P_wKwo
z?)AH__;c^;=~7*OyLRKAC&(Y|x-;s0D(ZS@*nG16GirJ_{!CHZ<2bGRe{tK-)bpz<
zzvA)<n?Gtlbn^w3W7|B#&tKd7dD6eLao6YlzEj}cA2d%3T4(obH^2QsaqD+nRsYVa
zouGEiTD@ial{;=mRlm42ZGYJN3EpwQyX{`bxtex8Z9fnAjfb-PW!ph@zmM8~OwT9F
zRW`lsct!fxx{Q5uzN3qSZ2NiBM~!E%9kwBmCIQ*=kL~G$u4nv?r`blkD%;Ob_M82E
z09)E8>iNyoe7ovHGW%r5J=b>7@7c0;+^x4)b<N{+*_CZ?H@&ysY^x>U_dJU1bA#UZ
z9JpSFeXhk_FLCpwbw4NSeMw#Uf}X>*?YJ&ZN!d*+F7vfqUC)&yW!L*&prp3laWtxW
zf1jg!+xGS7sQyQ7pQQ4+^DxyK<iGBCOUu5g*K>Y;scv_#y`VUVd%wk7p1U1XZc@ji
zjvJ=r({;UwUA^~fWcj-NCTaB>mxA`I?!RfBFZ{|?T|Ym2ri?SH{E&_NsOkN9Mz!ac
zFYfu#l>D}=<NNs;^t>3y^?r}dlsJy-N7Cn4Rpseghx8jqz58|jURKtw*Du*^zw^E6
z``z_>Th(;FLH@-Zce0XS)jUzwj_v&X<~3>UmmN1%r4Rc%!KBj3@;7asDeJ#!>HV&|
zbnVwYj+ho7LH@*z2UX*O*G^LRn}f=ceSde_&qsMVdbi`YcT)LPT{oF-C+axoZC6#>
zN!>q>TaK#wCS^x=90t`RXg(Em{S)@M?(=J0`$3#cJMR9TQMbC|x?laHj#J)zve$iK
z)BE}D?hiW8n|j>!->YG{WF?)q{<^k<?t5~)Y}<v0P84A|c>BNZZC!C;y4|4jd)(t;
z&;N1L+5Q}^uKP5m+x6=oww-@W=N#mp?(sF>?fb>K-+ozMQOCcaa>AakZAq8beBG~|
zW%mcQSH2y0&l9p7qo$9$j*xHpWjP1^E>U)zan~nQYg*$-TIEJPznGRU>~~I0NvE5i
zZ2u&EG6YnQS4rE`MM?Ml5ZAt~^Qg4W4?*R+@!-zaRI4fN<<=90fD8e5+)<@L{ZZ9%
zZyn>V7u%M9#8F#6e0}sP5mkepv+^F-bsjfm{FL=0>(|q|+S~27s?Wno+0(@(s$JQ2
z2{*l}@jhugNwu@Ca+2odmZz&ew%K)g*sikot+Rf<vbBD;^&hXl?)k)4m7VvQmfrSp
zEbV^lz0L`{ownam%TKGH%Eqbe`OWt9?sZXERaIZvb}==*s&%EL?F2m^Wy^8X&Won`
zlh(Mat6Wv<rBT;COv`6lym|f6eLp{{{jlddZ#vW3-|LU6eixOis(fDVy0*j4yYoGr
zF8+RYz1RC}NBh2cRBqmJ+}3SV=dpfqA9ug^e4uJu)p;ezj_&q~%QtR*Rqc@#J8^lc
z&TslSr)qwcH9M-~ASruM*HiuSCFOTg<)l6D)>Tj4-+i&|{1A1XYFa+Oe#!nvJKTBU
zTwyy0UE_gkSJnEa?E7z3y*DT6aYxtrLAD)q`62uLLpQw}Po}6YKG+TSK6~$XF>Tup
zS~tl0aZ~!S*ALa>y*r=t>(^o1*|if_z4d*0$^4q%HQh7kr!}|qXjA^PJ#ta?t{Q)s
zSff+bT-sXiwl(z%=9bUNs_gHGW#0oEwmsusUxlql+;Vi~A2|HHeXF+F=GdlTxqH)j
z`>E<}Z+W(F%YIM9wDd_oFQuzI-S<0G+qZq)SXVn{#ctU1Yuf2lkMCZ4s-AOo?Nof}
zkyjQqS|OsUYe!bYuIqT~pZ2!j{NI-SVXsq6O{e-kXiz+<E-z_&raqsZFMcI`J_?HG
zpyMr?Jlg5R7c1qG)~MHmvfp<h{ebRQ+)^$F2R$78`j_l|o21hF`5SeAQhw_yM|K`3
zyUr}@x2%0{`k?yA&hxzXgSKVs8}_;&Y`UOzhM?=?wAK|WSD1V5U(ep;cs0mh+2gY7
zXF=b6@~#`}UXS#a7qlJZf6#H=_0erVsD8E_myL7N(g%%8rX7zuABdYz*E}`oyp^xV
zWzW}Xr4O3dn0{P#z0|byw)YR!>rZjVIo<iwe!kVT^KV-IsE)g|?FGe^>bOq3zN*Sk
z`aVEX<^5q>=PKvkzS{9qzw>fh=M&ZO>Do(r-CtLEvg<=`dY6M~s*6)nc3r;cuir6M
z6<?;zuT|B@YbUAk-?Vaq<|jeNO{<TrKi)Wqx-EO1Z|e9Iv_2g6co0Wj^O>M>Z9A@-
z(-}*v)T`UOe6gPf{k|ydeWkS0`OTkI?@wwzqr04-@!gi=s{4s5d#cA9ReqUbM|HnU
z+McTQ4(zOGTxO5Q9a#>#$4Bgkdw-qmyjwQC?Dc2T<ACb&qK?n1&o_SdWbeQG*$KNp
zDi88ipI82zB|?^isn@4z`xQ1`{nE*fd*1YZc2$o*u06l<lHM;n-pHm8TOV(G*s?7<
zuZr5wWXmxvz3%(oro@YE|1dSZDLjMx&>dGv+t(cr-tk!1bLQT1lHS(M5&Hu_8FKql
z$B#qMc`m45yT|?P>)IdWx8L!BqZf~#TW*W+s{7tSzx97r>-m1yA?|$BPfeL8#2p`O
zSx;Bh-f7v9y&r2z`lRE-ufNOgkJ>N%#;LUTs~&Hn@=={n*ZIg+yQ=wO4+3Eb=)SMy
z=Y#6=oOhg7J#M=8y!G*xYx}laPutYnF5Ye1>J#K|+&FXj=$_}f<pimF{hF`p<rCEY
z9N+ml4KF|0ZNKY%)%%0|@>`ckdVi3gwjGZ;&(T#+`tQ#DR~3)A=S5Zdb;XS;yRy$Q
znU-Gl^+S+7RosK@*gpTUb^XQbchvKRY`&m)&~-d0-t+aiF5X_dY2A;ATdwzh9oxIU
zkExqW)OHNIo=N)o4PE8wdQKxLyRzT4F(rM{@1^sOtM0a{d3@S-qRt0h4!SjHJkfPL
zXgwrqxk1-Cx_R1aKkD_nY4ugr-;=Ia)Oe4ZFQ^}?I-WF7+saeLIjSAoc;PPuc10j-
z>mEV#8CA!3#ZqQpQrA^(IjYX{UOTeuf2O66dcBeF`Mj~N`?(5l+y!lW{foODb-r&}
zKJU7JzHNKsKv&#)%Sn1$_WK#$^hx{cEzkDx)wF&u>yPc@f^2<BulD&A+v7*o_~RYd
zWVeInIUN7>eY?6i=$apR?aIc5n_g9%2HA1T^ZK3Ew(F0pe(fLRpPSFM6ID&SuC>LV
zsJy-TvUdIsid)-`n;wr@i632@z5UJB?YQl%s@}TVDXd+$9)9QjxSXQqcm0a1ri_n4
zzvuV6{}^<i+Ap8GKZswDebvW<{E2%!>UY}H&gWfsvu#`VcVTtgpVDRRb!$%F;UuMT
z<J>R5dmR&`QOCs~yWZnb?<>o$8-+dJnU>DWE9|yAz9*%A@#eQbY(2uJbIb9Yr}*uc
zjU!d*U4Nshd%s?_%K8(wzeJtqrIjyiy>z9E+i#<e-=^eK)z6Z)qq;w)WzV+xrPtrI
zxWz3u>3CILp0^#m=aabOx!-v_>-)VNgSO+IKUC!p@>ADw)$0;b?S-vhRQrDU!j@}W
zI@$S_ZS@Q5Z`kW1Z@RGdNe2N4KmY;|fB*y_009U<Am0h7S{Ko6N7s5--1ZNucU=2^
z`BkrTs<Ib%z1T0m9|!iEL7?WA9&O5hwnr{0y{pC_Cf4XwHK$rt(~SB&2>G7hr+?3{
zg_lFncx(D`RqtVtwWIq!c-;G7QS+<ruW{|^ny2Wp>vz5l+8>vb?)<X%RZL5-tG!im
zH;q@2Ke~9@VpmnWV+R5dfB*y_0D<TPqF(R#<x@QmiC!}^b9$c}vVB|k^}KHTs`=}R
zI@iyr{Y#f$rrHg<4m0hz-*{@vepUBv{Op(-XZQsH2tWV=5P$##AOHafKmY>C641R)
zg#AZ5+<D<#;pO12OWCTb*MZ_jy{hk7MqMXVozK>|@cQkyE@jJp)p4GdJzLg8{rt2g
zUc7$B9T(z`i{ys@1Rwx`WC?iZ`&rrcp5Ih$C)+yam!^tq*6o<zl6VLKc>+oGTY1ma
z=J2i;*tQ+_`c5{#TKZUn00bZafvgkodoQak`%Ql@fS-Tf_aE4{opgL;y&qCP2*?m<
zIB;d>e;n|oqg!QsA}s_U009#SB=sG{pt>eKuM6_WJuds6Dc$L(^ti9ayR){6Al2<h
z()N>%kGSOpegBo?r+3`&`|pJ=4$TgH>geh>Y!s+o4(>RalqNmy_>~v;y55xM8NBV5
zv|qZP=ajuaWJ`M6?(-#eKh`bB8`nv1>v|7xz1!B*E0|k8XREsHpnE+=w|(8;C)aKN
z$06q~XkD>v&URJ({%hU#-1yg}14l0&KeyZ#VVnBh|MBkkn~x>6f72!P{`TD9?IP=H
zTm0y{j!et0?boeI^+#Rh#9jXhn?LM%Agn!G(}m?Ao6hg|aI*Vl%T<*=?mB}h`E_6C
z>$b0JoJq>AZR-VD^Ec^rwW#Bdt>=%R?{(O6T-EisY3rCt_vfhbAC%8~J*sOvX<k9)
z1#t;F{^haD4mo_UDmhy<sD5FOCw(3;y*!tj>~*%S=PA?tcKPd4-ScwY_H~W#`DRyF
zd<EIHopaE6+0^5*>no<Eckd(Hs=DX@LH*9Q<5A~HN!Qc%@hr%HQ`Rek{D>Ma-h8t6
z5rdxlwdMFeHE){lyj!MNwaWI#p!uV$J=yfS&wH}|c+*Eck9h6q+V=LRxZAq?G~Mpi
z*KXSX-itnSvaY7im!tMGQ}UV0Rh3^s=ZUnB#~t@f$)DBpp(*`Tl^>>`&w~6*yT67l
z-;{nC)sLX-y{sM&ia+o1u;ZRLU)VTGD_z)n#Z70*I-P7iyzQB_ZQZ}a&~0DUcQMkk
z)2v##&P}F$?_^Ea{d4Sst_{{ZUX9y-vU)zEUe)K7KWB;1%|UkjH#L3G`nTV4Z+}dB
zTle`um3`IomAK<x()puaPne!h)%A7O>?HO4t*&`iQ2$8!xh}u)SJ!^o`i1>nLtKAl
z&*!rB@Jk<5ZrbCND*uAwKI!8@^-^_QcfFFH7sV}4mPgort?K?vzS!}$Lr}f_&bzAT
zpI&=jKUHs=dfgP&FI(q(-gXbWtvVjO_JaC_>BqhGQMGMroVk9>wwo$$t{vI(z3E*$
zVbyQ`Y5M-O=fC-0zN&e!tevoNEt@XtdDSnUw_hf;?T%xnYS?)@=`Oo%>K7|T5C^|@
zRy98P#dqBO-u@{&4ya1+^&@Co71y-v$a3@J;TLzh_sjYpH@)feps@Z0oqzn=*|hyZ
z^|0l5-0?H*{I+nm-A}*%7j*uJI_~+E=eIv<Ir);$kDGUY+~+5~_Snu(0=mXKFHPC^
zsf0}*biE^cJnZj3z4cSI9d^As?R0+kaoB&tehuGQ<CxAtRlG*E6Eq)5>UdV>6}qpZ
zUHehnJ1Ae+>r!vJAiGH)PkLUK_H|v{^5gO{C4XEFZvLd>)mz@UE6@7qk7riQsn#i7
z)?T;f^c_ytRMq=Yv9tT{zo?2wZ@kC-uEa}QM&I4=h;qe#D*HYiH@(+SzioHiKvy&_
zv&Z93);RdJd))n5<78@mqPA1i@2x~F&y;++)}u_fo7DBLu67Q)UKci9)cDW$e6s7E
zrln7+pFaBVfbLh^QZARY#{JzxP=0U!R`vT(ubm)2f{uIrRJ9#ct}VxP{ccdVU0eA2
z`KgMJyIyLWCN+;%)!(ybC#iX$u5x7CL$-hD^4rg@w?2O9Y}ubw|JGHGY4g<}f4ujr
zY~7aar*3+0|B>C+JwKQAM^*ZP!_V8dYMX72sTvg5LC51>hktwKQ@zhPx@>Wf_PYDh
z6|<Ue5pi`<(O%t0)LSKz)@mc_UitDzAB(6u=+5r<J-X>p5mj3sJLZlFSC<M|t<<5}
zT{k>3U#QaZHAj8j<HR|_Hm&l|M>jtBw~3Cen%4buRpY3vop0V*|NNtOydqTV1K*b{
zUDI=!uw4&+`GLZ!8#ju$y79^KXD(f~S)`?-@9%T*!?o6ks45#*SxG-)*`X^7JFgRQ
zH7h)*2LvDh0SG_<0uX?JnFP|hj!mmvzx$VA_Xn-t$sSMpeg39IWm@qNw_JC<-d0Vj
z{bl<_TItjNeW+W0QtcvJ&aDeyoOtxJTO8Zf)cZ$a{nCAZPq+QB^|U3OZeCvdvi;UC
zeOlve(&hTqKkELt^M<tZtNQ(RSbnO?leJ^Zb0c9ndefQm{FK+PsM~S-g{tRJZMBp3
z^<>oYz0b+pzU^`d+YgfZ{f4e`;`Up={9*aI>EgzXuIDe4va73old@~d`7Ej*NzZFl
zmlw6(rsPYipXn+`S3h&d$E>St9Ln+tdp#j**IoBZTEpJ=uq~bG>qhSUFez=(<k3zi
zzE~+|y3TpC>vaqEtDZyQ5G4*&z3)iYj_i1EYWl3aXDn;)1-q-pA12o5R5fR-nttAK
z{fo<2cK&H<dR^Bcy6x(^E=$X<_dJpG^?_e`UcY3w{p^O_FI%2Bz3K0nPRc(o7t^+r
zx;|1}j&0Y=S@l=deY~K45%&FTs?!De74^8S&ojDy$Gs1fcK)Qsf6CeQcPFwk9;zDu
zlD3mH|DbU)t@*#|at}Ul<Gcose=EG|*5B{>DtBB>TUGN&+MaHHZadnde(@G|zwY$}
z?3Zo3@X(1uHHW18ov!lyo?EhIzv_0?ZBJJ_CT-WWb>FbhHF(p-eP5Yr`AusVRsO_{
zZ|{7<8waLst9~xd&z_e{R=2~>$5f|Fi>GPvEbEV7ye73jZo8>oCr#U)Y3I2he`Mp$
z)by&yqonMa$}_BANsTkMmlJoKbn_<_m#WM0u0!hH4jaE=)9Lm%tM=WvO<Pr6U)gFW
zZT^&>9|YpAKZMQirJ3|Pthao>?K75EsaLmm`QjiAnx}h@%g%>_#w|bpv%Wv<`QOxZ
zy4T_T?0e%+)%?Y4H?8fg$Guy>pm8DV$6XF-sqW_llD4laUc$}`!lp}#OIUwQN$2(h
zzj5D|{i@e5blY>!b1r{B)$NC0x%skR)pN38f8P=|T~gy!+U2DEJhoeY+&D0$zxc(M
zZvUqAxUa^$v$hJ`H0gB&UFF4%U-Iwz_X1(LB>h}F<&9m?wZVEPX%4E#QC-(hw%dK;
z<o}W`>R5DrVZDON1yu^l_9|*wR8a1hzlGHcYZmTbP^L#w;Q_yF*DokrQ1*hYt$P*~
zzR<hL0;jy{cFL^1rcbr)yXVnMMb`A~qT&w+G@C7SYubVF+|mobxnlA{;iqZ)NBIzd
z00eXqxb*h-yNxd>n@daGdR?<__~{o8EuWLs505_Z(pn?S<!sfbw*Tv{TfX?oNn71^
zK+Pbad)}3n{kU{%R(bM)(|Ru!vKqHszx>yJ-EVM*J>M2-sq0tAPdBvcTy2IhO<n&`
z2tWV=5P$##AOHafL?K|>{5GmT10QW!r`C7F9b0w#nd_HKxTnOiT~~iy>ChvZ%yMki
zr1?kbv*dvQ1a?m#?t0wrtp^{KT~Etu`e0Iaq{!Oy2-GLsb0&5~CtgAT0uX=z1Rwwb
z2;>_9+piy0U615jyHj5XKmY;|fB*y_0D;{Th&oT+y;Y7v00Izz00hh?P;Xr2%kQi_
zUqn^E-*vNp_rH6GPdq+R=WY3(&+O3%?;rpHlL$n8ZqlSW;0pvGurq;tp9fL@XzzK5
z%WdZ<WGmkZJlOBX<KF2#NuYLcNIG5zHktR+=E}vwR*m}mRa5c>y>}<+c!*n`>~m#t
z%Ok&81fs5UC!NnMYIp+y2tXh@fw<R=-u%%$(VH`9zQpl#-{%;XzuqXpx@rR5k9+L$
zZWWgcG@Be$#|!o#@V^A2-shCfr)pj2e|_hFc>=b+j|{)<exICa?RQ<x3rCbGKhHt!
z<RCxt>B-@}KF`Xw_kMvMdRT=31hPOt_Ie;{`m|U@{oO)Z{!%UkejyMwZdK<qy*0e^
zowT=2=Y@X|fPno3Ok0-;@+WKSHLAYTkya$O`s(r)Z{D!NaaFY+Sy@lX+H>AXKNqOF
zyr|corsPxI&#(u9C<IhLcM`=1oAan1*G#dObep7oy>4syNw*W_$q~rPd_z_LkfTRx
z2tWV=5P$##Y#?CD^-EYkOkt|}eyyxMQ#jxU1Rwwb)dW=ieg!)lT6L~AL%?BY4yxld
z$e!%^Jt%!pc^u!FfU5N{TkWX6Z=~8@5Z9esGk7b=UeNL25=fCx1XRbnt@gsUdD8b$
z!umxzI|yXuIkd3#Fuhg6j#sADW9yhR4)0l1A(xh##sh!yi$K_QbvIpp(V@2aLLlgR
z&+mAZArt<+y~1iom(FYa#Cc05eB$WVd}(Lu34tsS2z%XNdOE*xmHln|?s@c5;pLEp
z*xgkfy!EtoJLtYl)Z<~Vuk$tCyYG*xyJqDI5mk2;lk7{5K+@~Ms>@S#|6G>1>Eook
zZjiRhawUC!6NuWM^P3vAhX4d1009W(2Z6NbwMm!n*Q#N^XJbmbsNeDQt3Ug*O~CYZ
z^rZZYdR=5{zNENOPCDOthr1cv<DMOoUgwPKFZuNl&^3Py@;2&yJMvxl%@vau3R^e?
z#R115009W3O~ADAG$|fQ&jYhkUQ+F&`nd}16@NIO*=zv^yEp`WH^G+Ux~{iuwVTv9
zYDzh_uAkX{-nNxHew(r0YI{4F)_>C)k6*v_rF-X|vPaGoz54hX%a590Jr`D+|FLoF
zz3NrVnW7u+8okHNwtMGn)y8{wI_a>cHFHVnqJ4kb`0U^I%jwelF75i%wZjj{*`iH)
zUblAmbNl6@YFgt6<z|~ezW(kjY`t~0e$vn1MJ-QP{jm!HlL#bzU&N%ktMtXz>qkEZ
zw${h%ci8dYkAv#{s=k+totAZKeK*_*+rFyOsg4KiK|qFpt@nZA`t8R^Rlmkg<;(A^
zJYT>eT@HS6?zexhRrj{(-0?fdRK0)iDP>!ovc<7Y$GqO)sfVsAlS^9t+F5r05z9&n
zt8O3Xpuci(=gYDjNe=-CKmY;|Fp+?1&wa@HV<J--pKP69Z9jAUk_q>eIHsu#d(uJx
z0(KH;ci6W*&U#>z6IEY*d*2ps-1xm?i@yJ_HUBBuf2(7=Ms2UO^Q~-p>LG{y`3EPf
zx@>%naWe)laa60FqYGX`Kn(#~es_kSl@2|k$t=fI4zlm_da&P($Gy{glE@d0`hLNr
z^W{r>P|y78f3~!@nzK4qf(V5Doj|^(3*yJ|Y!cwTB6hu3Bx-v-)~(wib(((Rpl~1#
ze(MKOM|Jo4D^oRUy~qav2tWV=5P$##@|A%1ceiZY@m{lhjaTXofn5=><vs|0a(~XI
z=L2@t*VvaHfg^6bY|`vr3q;!b`H{01)_VC9VTz{LB9sgP2tXj41jaNuV)>VU{#c-J
zAP&sw2f9a_q~9+|d3Jq2lJ=s#Z2I1*U%cubN3joqFa&JB?<V{H5Yy7D<{#AN{~1_e
z-@$uU%-O0vW?c3AyDIIKL(}1KXsc6tI`=PglG31f<~Rf(009U<ARh?$T~D%~b&)Kr
zix9`eaT0Ny59bQ%2muH{00Qa>n6{3LKW+Q&dGu1Dn}hmD!eSN)=#E1_`?^@FzE72w
zJwI+i`*m@RYd0;9s>%)GZp(31>lSh2f&38I9f36ub>Hj$a?>4Kb$4~sdCb)BTqZT{
zsD4kpYfopJVH*N2fw<ojCV#)d9rk=%WSxT>FRFBU^+8`ga?~mZHG{+X*M9oBc;GV!
zg#&S*uOG;cpIJYz_|;FAQ`Gdb?@^6fj+<{@<0sBrGT{@)7B%g=ma_i%#aF)XcjM7C
zjryMPpnRs(Q}=gvOsS_XKk~ibno@sV?TcLqWSM}j_YG#58Fh<K;HEwU3I?2WkI0I~
zCxV4I1WcKS1o<I*eQ#QN+3Q_X(zp8R@)mF2u);}7RXukTr%#X<0uX=z1Rwx`Y!R^K
zdE2D?bX!;V^M7tRLE7=t4Xrv?n<4U5gZOY90uYE!An5+0-|^@rk`n^v5=d&Ep{pFf
z{;F$#QZ1?L{!&tQgVu9xIiA!!G^x0v90))F0zu>N5z9&nt8O3X*se*P?<i;I?@tEB
zm+IqwJzCd1&=$LXT-Xl*2xN~y+~+5B-v@|mKYJ}kjUWI4(+G4w?y<|eRa`DmIS!_^
zBmO`jO#-G}C#q(l`*)DopZC+|%EiLvplh8;bzGQgFYbLvQ|pCa5P$##AOHafBu^mj
zx~r=E$=93`ApijgKmY>q3ApPVX=(f#U_mE=xbK74=^M5o009U<AbA2w-=|N$hLi{a
z2tWV=5P$##An?Bg;;sv*%KyJU@xOTluB&<Bh%)8pIjYpW8sd=-0$F+AvbP>-t+RW}
z)zN0ygn%vrs^$w(?dYnPX?DHiMAYrL&;63WZQnhQUMk=K2jalLADA|N;*XsK;{HxP
zZhn`Vs^?N|wc~P#szLLWsCI+$*~&Yr-$6X$9`{~1tJ)5V57o!h`d)3)_vdw$7dMW)
z`J-O{k?+E9u9&<~csay9@8xTLZ(Ons0SG`KDgp1fZ2ET8d1qGgMQ!J-=3|{R+v}WJ
zjayaqQhnX&wWrEY)poKno>Pw}A2_Y|V&TUjYTW3W$3*qVlzh6l$l5i92Y$pOAlvRi
z>En45oIC1zW3c6DDc*g5RNXZzR|wP8Z#*`2e^6YReq8r;Otk1MQ|9DZcmHL<0X1^A
z>g%_@bno0#_Q=_)UsarX{-0~@nX^@|KEB5Cqvlu7rL9rpKWaWx$H}mM$&N=+ImzY=
z>p$sszOItxO!{O9$o4bc>6587rQ|CC+s0+nuN!3jODgWX>&Uik>w15CQg&tAIjQgG
z>MF;y?>B_?N4Ec(mR|L9Sh8`eD}7S^lyY|d9w|KF@lNkaf;iYY4s>7lMYV6+eBRIB
zq|d9{&s@J`!aXHUTn#(VFg2ZDd)Tr+>A0c1oqtatJO9)*U-Yv}d+pp_K|K68hTR|a
zJQg)ySieZ8mq1WlnRYyCyd|A4YO6(k&PFz$pMOcsAN<N^KLqlTz_nlZ8{A>fw?)3I
zs<ziP-mCI2$d0Mkonie->O3M_P7rrh$7TI8ExmicNfi&<?Z{p)lfLrhcUGP+;E*f_
z-Th3}{e>VuZH*h#u9u_wlh*ktC@#`I?l=EQ$}e5>TI`1VJ-F;Ti)ra;=Wy-p$0@0P
zo>V!i&gVgPg6iYPTi5;|e@s2@UY7-_-+9W`{rSEwM*UeIHF15|)NvBOq7m3@)xB*x
zcl^#tN~3v6E(kyX0uX=z1oDYMTI<sJR&Lb$2IP}_Zr6{O?eV5--7YD+e(j=qe?E=f
z)K#8<t@pvbeh1wjF#Wiy@8Mu)K*4}>?h#Qr*n0m$^?4)6p8QBdjw}%fx<8oo@t}BB
zecbl8OlqDRwjA4en&N-b{m%6AOlcSV&_f{VJkOMTw!MGN&)>A?&3d9M*y^FJPU-2~
zzs#{!P3Z^u=7(QDvSq*S=f#419w~GXIAU2zVb$&9oUH1=M_bmZ_1$pCbX9deZT`o`
zt@o-|Ef-c#m~-n1|2U*V&UP(q{p1gyY^#`4t+IVC==T}2_DBx_8wr?x|3#I5s@EB@
z_v`Zxwm+{B<Uj4K=DC)%Iodcj<1+*x009U<00IzzfIb3YpNAk_y!U8im0!?xWY&)d
zt*3JQvCF$vTrN~|&^IFLv8?L;iD`CBf8Hv{Kilu~uX(8ZUiX)q?x<Ehy};k9X>pxz
z{si@td^zqPp&W@rz_xi{oVt=10uX=z1kxwq{avnY+p^cq=@U2GgRsvX%BH*Z^ZQ2M
zbZM2Gt(u!yF#O)>RdTj!#fJ44*H~LAm$aH4?RaMj0o%?yQT_FM-s!tPmS1z@{zHVR
zYD%2Q`r*ghmi@Bp0<wH`r<e6VZu+d=NA%W9ce|;wKj!rYPd#*1nH&m-e29an{U+*s
zLbg2bIKuYM&#lPvBs~Nm009U<00I!OmB9Jee)_q1;4?>7ZLKeULjVF0fIzkh$UYz7
zrVmR+_j|RlpYC@a!^VMZx_o<nI<5N3woAUJx8=U5t$yZvd{KW0{6fH;XC<xgzCWt&
znw2YrE={U+<DO5G@{@8P0D<rXlA70q_lT4b_>F+B^QNEO-%92`2*?vidVZ;^yr|E`
z$`i|v98q~Dl`rj9Q1yOg)9u84Z-Xu0{|NFk?)4=3i$5IDY_^c)kkokMEhp{qMRoa6
zpPwdQ<;(A^JYT5h5cfJ(b$r-j?}!Cge%bfvADy^rOMUSZ0uX=z1R#(l0=m|LY_prC
zHl;2Qa0zTbbN!MD_mnv3uN+)1C<Js8u;usaIvGT<Z3{R2-1&V3@PNmjM=uq+;vmXc
zva3APs}K6}k)u{QS=U|FJ<7hg>*HC=AB7=#AmB&9d%sh5+s`ihA&>@vp+CMm;k?f(
z=Yn+Z{_8(J;ik$tbY~7}v>9bW00Izz00g2E&~@F9UDlPNtSd#0WAZ@&0uX=z1Rx+!
zAnx<vY31KK=8VI87FEdErmw!eZ;Ll>{NAx$XI*m5ph`s>9dE?N{oR|Y{9eD(-d6oP
zn0&F9cDqqN1Rwwb2<%MY&e=;R&F!&5pm2y22Y&D8VgJtW6WqD?vE?Oj#IlmYs@umo
zX{+h)B?<CR_j|#;?UwcJW#em%n=yEaldt+%w{C~jY5Ij@oAUcBGk%9=+d4<qo}V_Q
zK3P9cP(KJj00I!Og8=hGGv<f3x5G_+1{4f9=N^H=!NfRFH5$wQPLA|XK5$y^#R3lL
za8MQZ*nt29AmB$J>Gids_qDsn!`=r9x}W4dF8h36(E6_**R1cC<w*JstvXkmq5OU1
zLGhXN@vO%^^@9KeAdpo8wqEaNwLNs#OV#h?Rb5Be8h`k$W}QrTykj2%5P$##vPEFv
zqb=*y`fj*`?#$tdlmFY!MIDQ-FRWKkrJ!m-*<M90iwes9^0%;hVa>w*3(E8;Dm>tq
z?fL~}3(8)wwRO*;!WVisS>Tk<R+La12tWV=5P$##AOHafKmY;|fB*y_009U<00Izz
z00bZa0SG_<0uX=z1R!7`fu|c(8@T=Dr2@?_4i<(gE<*qU5P$##AOHafKmY;|fB*y_
z009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*!v6X5q1>HfZA
z>wp$(KmKTi@Kfypptu0Pdid>ETs-C=009U<00I!mAc3LJZ2Muv*m(lAn?nXelPW;~
z0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00fi~_-t&u%d5BhSfFrF
z5eLcw6GISy00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz
z00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<
z0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00IzzfHDHFlnxrQWbmFj6b{6J
z{ry1Mz<?nLKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHaf
zKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_
z009U<00Izz00bZa0SG_<0vRRnr{hMyytHVxK;b|fn9~n*4>wOYs5Wr>%S#0c2jYN!
zkkxTO_c)4u2!tUp^qFlxj2Jsl*s3>o>HW;*l|B>buN;Ef{G7|Kym;@ss^(BQ*boOn
z@yu}uKmY;|fIvPFV4ll-pX(j2_ZSD_jRW~`uAq((fB*y_009U<00Izz00bZa0SG_<
z0uX>ewh6FK9B-W%e)+{O+oJ)sg8&2|U><>dJ<pog5_kjw2tWV=5P$##AOHb*0=nk&
z@}7_b0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z
z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=
z5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwbmw>MC3x&1o`hh|K
z0{RF%y!y?tzx&_~k#*Ha5X%sN00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|
zfB*y_0D<fhc)CHgf!kkRDooMrlBQ-5fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb
z2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##
zAOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|
zfB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKp@)$s-0i1-K4%1
zb0{2C#6h-u3$=rQ8UiPdDQP#i*+gNgs;MPbAOHafKmY;|fB*y_009U<00Izz00bZa
z0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z
z1Rwwb2tWV=$r5<FLA8O~UtTKmU6Z9wX%I+;z|CEHKXZAd&jk7_hh(TzO1=}Y<$P*d
zJK4ewKbbc&Z%*#ILDe|*+1PfMS8w;R$hxZH#k_In`Limn*nt29AdoczL#n-V-lpe=
zI6-Ra`$N<AD{JvUeINh<2tWV=5P$##AOHafKmY;|$R2@`tygaw)c6O-HodpY^nc&n
zuzb!o-EqjwtG2CK?_@=@*CX`QDC_qxb=6N#YhV=u5P$##AOHafK)`kaHNJi5FGn6Q
zU8JqOaudf?YcxloyXIhfoXPtC!Xd-jcfR5yfx>|}pdT;}L>mWW$6eAx00Izz00eXp
z_+Z$$TkmfAf<WOACl0Q7>GrbkO!<pI;b3AM=omY(2>}Q|00IzzK)M8mKC|tI5o70x
zq*V4fu^_+FrAO%yfB*y_009U<00Iz@Bd~Qqi?ttrv_iyH-ut%md*9#@jfY(Hk{SXK
zfB*y_0D)Z*c;e*$?k(zAbbVpHf~p036qM~%)Uv3c+%JC%s~6TR+`phqkD|f@e%Y>H
zP`04#1zTJ9EGm4VcasH9`CZv$9|Rx(0SG_<0uX=z1Rwx`L<zjnwM~Qb`+p#0RhN?k
z&O2pR;~UH8bm`t5?^wF*m{pFS>Z%uZApijgKmY;|fB*y_009U<00Izz00bZa0SG_<
z0uX=z1R!7{fwRi=X*uVGl>$}6!NfMkCkQ|Q0uX=z1d=7NcE9ao)*QD=pm0bt4w8*}
zN`n9dAOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1R$WFz;nm8uTydKYLTz1
zpYYBl_`R2f{kwl}O;UVzZY{PT009V?OW?&O*Y_*8d6DDQ)$5-5ZqU@34w?-P=0+f%
zWtl+R*G8?Y(P5~tO&{Oq?l&uUnj&o18y;wLR;`~FiMX2O7}Zg?1NydUv7~T{6IXTk
zhD``S00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fI!v>*#2I*thW#K
zg8&2|009U<00Izz00bZa0SG_<0uX=z1Rwwb2<(c$MR&EF^_{4iQ>CRJ&HQ?Mr3yJ)
z^`8T;eq-OW);hLnzpoEkGj6}-!gdY%UGA=8n0*j{00bZa0SG_<0uX=z1Rwwb2tWV=
z5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0`?O)t4yDkb6!{}@<r`$F=_w-
z2tWV=5P(3^1nzmh&Fvj?LmksKX)cs!Hi0o6o3;C?%LfjcZ4PEfA>Kg%0uX=z1k5DR
zY5CbL4|uFn&UXFJ+HO6D%&(k7;gHX9uw>q_9z%LB5$LZR%p7&`3Ib*k$d~(})H6Ts
zf1B02?D59-@4Dds_AzUYTP5J2hJ!sl7f&Go0SG_<0uX=z1Rwwb2tWV=5YR;+?0vSV
z@87R@Rj&>!znCXfX;eQ<$!GieecRVYt*g;ts4!J^^%Lwu00Izz00bZa0SG_<0uX=z
z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=
z5P$##AOHafKmY>a38?ygc~<QVYdB=+$d6|V-5S2dNC^Q5KmY;|fB*y_009U<00Izz
z00bZa0sRDYuZLkjnRUl};jF($Pyz%X009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=
z5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHaf
zKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_
z009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz
z00bZa0SG_<0uX=z1Rwwb2$)FVjjnAPoZtTg5m!yD8$Lk*0uX=z1Rwwb2tWV=5P$##
zAOHafKmY;|fB*y_009U<00QX}czmC`->lqeicqcT)+&DKZC@L;u11HUB5jRd6D&Xg
z0uX=z1Rwwb2tWV=5P$##AOHafq)A{{!y!XQemqm4>2OH1?I{}q5P$##AOHafKmY;|
zfB*y_009U<00Izz00bZafpiE|ysB4+m0!#gx;5?RjI&yPI#ENZwiD?0^+9XK?YCUm
zs;Ym_fxU3QlR2wQpO$l8SSgZH+xr>*LjVF0fB*y_009U<00Izz00bZa0SG_<0`?PF
zGH+OqA-$If6b^CXVC{a}$E-PSm9Rza@45M11Ks!iu^;XE3eF>DokwgLkJI)uzeiJQ
z4*>{3AUuJz$7@^44^MRGlzVr)W9hPERynBM9CmJ1wp0<YrT?q;Qx!k#Kwwt{RL3j!
z!mW$$sx8>}8-b*r+s=A9zY*m>2tWV=5Qsv6c}=)^O;}D*f6sTr18vT#_0u8|SEIxL
zc_07*2tWV=5P$##AOHafKmY;|$QFSE#$Mawi>_sJw(1*Q+cY@8{|5qvLlSY2t=>;<
zAOHafKmY;|fB*y_009U<00Izz00bZ)N8qOAgU&qixb=>$D*Ih@IXa|<00bZa0SG_<
z0`?NP=lM3bcgziSlGfGhp80Oj)R_({$06wbW%foQo<jfv5P$##AYeNI)$geY`#W9R
zx#2$qAOHafKmY;|fB*y_009W(2Z8gh8P>SPx{5heFNgeyCTa-*y9q>n?;iOMS~G6H
z<-#@&c1J7TLjVF0fB*y_009U<00I!m4uSU{{Q8`&|6c7RrAIWndH(&awmM1cT`zv|
zpR+z0Dx&Ih$F{FiarA0oy9PaPZt8K{xu>;mlh#Oe|Ez9r-t+i}j%uCqm-BLCKbt9R
zSKH&+^*`=&p=rfu+U2_Z;%eIUpnM2G00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P*OU
z1mfP$NGiV#ER*o@hg%PN;pRK5<f3X4K2in*AOHafKmY;|fB*y_009U<00Izz00bZa
z0SG`qHG#OlFHn_Vb-l0$0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<
z00IzzK$$YvUR$~A)aO<@XtEs2HLCl@Gd-6$s2m6SECe6`0SG_<0uX=z1RwwbSpsi#
zZPVcV{vQb2RJNX^hX4d1009U<00Izz00bbA?*yJW`M(5;Iu>1DSg&A@f@%e2dlj`T
zDk%5M-@@vJH4FDIDAS{;@PJ>o>lc(QD0{)y);)^~U+CRrfm6Or-MPJfzohL7p<8=B
zzNqG2|5_$gHJEEJDZPH+>#dJzw@_qF|9bh74n6l+C3LH*c8_W&sD4rHnv!qJGkyAB
zvEVBaR@3(TmSfJn_~ynlgepzDK1rAV?BSJ$P2Xd&2&+l^M|oxum{nuiF)v-ZL8PTF
zjgPMB|Lv?ltrN0(%EyOQnceGKVXIbHzv-#>+O8LAtJ@Byt6zLPv8eIe$G`No$ch%%
z`a{bRfBZsZP2<MbGtX?jVUK0&99fN9u3!H1dv|$s$fqAWX{qa%U%yCuzw4hWRW&d4
z+EL{nc6Lu7>~-?)t=MxkY<<1y?5Uw1PrW>n-qusA^sTOLani=qW-k#@HT@dw?7>4V
zo-bMb=@*WzDm#y{CH>B=^xM|4wJ)uI@9cR(xBgaQwEx1+yQ5hzli2#Ycl|!9ziz&$
z<Ayh%Yfo0cslMUG1FEldY}d4|7gUuyqGG?gZ#19a*se$KnH&7$y;B`qHEI5`{WWZQ
zRqHlk%Of2GOd=5Xd=|GZX|EHRQobL@I=37?aqK6L30u^=zGchy?+;!1@s8Wpima;c
z`&54M8+N~L|FUjBEN9*6+FgA6Id|0EBy_26|NZQLde%8RYIP`^OG}rJ>HWgUE<ZW8
zYTD~Ze(kDj|L>k$*!PbGpF3&kx+k}-T6pSgVVn9r7pJS8lCs-l&zUpM>oUjjQ$Ihp
zU3O8^Q}_GEF<nnR>7Jz@zPQ1$RaLFy`?aHYzh9i2y8pr2UtId}lwu*P-g^6OkNmLj
z&967!FQ;4Qv_HP=6Sr-0Y*kg~f4_D}YQL&=x3uhpeXf~wa?gQ><rVfmnd#}0`n`y%
za>72}ZF)LYago$`NjYW^Nb7t<x!HQoI_!DQFP-k*5Ov+&&oA5K*ZVw+soQ?>V*7sG
z@#7b_rtHu6`X}8EruOf!ewn(CB<q*%xc9Sf%5wo}`ytz2rln8%`F+*p-Mf9<c@yf5
zal-0p%SSwYN&6v=tH;;>#~YQ89wAh#ZTFRe{7w4&;+CgsUB+}fZhdXju;*LV>5}H8
zdfYPAo+_@*ZvU+N(pR=RvZ|_luN~FAz4qPhpmkF>ovfNZeuec<_wS2Lx9_%lP@MQ3
zPkWuh?|LBY{-F7~-*bp*zlX!Gd_R9x@3-~2TUGJu_8at90$UG%YR(~LCyT6U(DQ(9
z`v*PmuZv66^Gne6m9FDa?^CGG=eC=xLBDtOdk!${{;0njbk_yKs_eQ!*m<bm`9EmC
zDd&lx^Sm1eQFXtkM^vBIf149myZ?BK=+$h86Qr`wl|`+W>U=@*qWie&`lMwq?Db9B
z=~T}<z4pS+^JUYe)vjU7^;@@BHNQ>EPT2ENzN8Bq54zInnzzdO>+Sd6ZNKYb)AxHh
zr1g6>Z@K?2fBrR(F07QZP1h|rqsQIv?3q)gZ<ZeS$cOXx$=Ryj_)2O!?RiJq<wxb5
z)_Iq5_5AL^%Qb3TkS`em-g!mb?PR?EtrSz%Tf+MBo9Dm&7j*qEdmX1cz1RPwx6?k)
z#VudgI&)HXlYXz7>hi*_Uz?t8bItoMeDcjjj;%VX&oh7TaNAPHR*n0<gQUkzzw$O5
zH~XKZKkS{$nl9S!s*jKQ{s+fZzxu1*@3&6L{$#$JW;^G6S?7&<A759#q~=em$D6SB
zqCWo{*1oQErp?1t)l>GprmD)d%}(0yWpm3nom<%FWs*v#n^(~DXQmy`%6UC)u5sf^
z^?E|u_F8=O=blafwA=}+arv2+U-mqjc6wXaYyA9Hbv{Weu5RA@l@<#=Sns&1+K&ly
zzOT@~%WB7V4Z2TZ+i~xAcWm2^dOc@KKJR^E?{?7rw4mpkgYIJ{^_-OM??<a1uVn4X
zuE)sQ*PT8nj%1Jfy;m@(-$fliy?l~>zp<+K5+`jZ=s8a>SKZsHfA5no_S|;y>vuu>
zRoyRe{XlmnV9LB|XRqY9qW+%En@_&v-RF=!|B^nL@2uLwdH2z!|7h6rR1i<U^LbeN
ze)(1Hm;Ifpn?9`nQPatOFCp#qUvK%UBQz`LD^*-%<HNM{LGwS=$K8Hwi>iu~xORfZ
zciqQb|AU?b@f#n4uJ5BB4_XH^{kZJ<iSG1nJDaZF_0XX0wCba3ew4HwUE_(j9io;K
zb-eJ)7Zg9LkNf!%b$<}Yu<_%aKl*KZ-;d+Hj`!R4KCc$^J|)@Xe)nP1-tWi5mi=zO
z^;-{&`nxLEZd6U`xglBq{5a*ye%<qwxb}nQg|h7(WH0Erw>-c1Mh)Am=ddxg%H)Fd
zm1_^}SLX6^xgf3CzRG2trj^SD=^Jm4?0oJs6>>q^ZPr7Tet%EJoS&{(*6E{{Hdf5V
z)r0mswO+*wE9X?H?0YQ=Hva9&gIib1>C(M6J&-%2`VUUNtG8eHZEtkWd%$UQa4sqJ
z^CPMKLD%7?AJ?@G7M1`0<6k)I{XHw@Y|%=8Y1QG?J{z5=8kLh@zMyqN+2=RH{*K(w
zpS1ShanHuXR-aWqmzJhgZ(GX^irch~M_mU<JD+Snch@E2UdQQ*Ki$t=>GIpPn=jX!
zrqwg*xZ!u*XWM=^{!G!J^-<H0yW?ci8n^vyeSb+<zvJ==n?LG!;^qq~N7waO(spfm
zuE5XFr2Dt~`%J(4Jl_35<7Civo9z5FC~p0ZKf8AJhu_Tq*0Ei6%|C<IqodyMl%4;Z
zHg9wNiCTX*pI=;>wm<B3S5Uupxw)ElJ)_=_G$o(QF|NwCgWq{6>V8$fe|GI8J)iI^
zPnN4}`l!z<lkeP%Z*Dw8sOBKse%|!5^By<7*ACkdNRxo-er3A7sOP;j+h|v1+wz<l
ze&z=Mc<)rlHV*Fj$~JY^$Gp0$J|wd*==$1w+_mlXBkS94y}jyQN86%7?^E$R?yWc5
zY6<wQN6B92d*jgc(!DMXQg>a~%{Qn0@nxU5ZIhF<I<wz+t400Ijw$NCe>^Mtw^~Dv
zy0O_B$5!?Ju7m9^j}H0tV@DSU+va&*f1_R>#m(oAqfu2?d-~b^px?38zuk9%V~R$7
zZd5j3)b>d#pZC13YCFh(UHA2{`+UjjPrq=&atQlg;Hc?>;=ucRNZIXsw~sq-LcKAL
zSA)t;>bUnhm+h4w@3?KPNQ;B2@hxdPx_S9sC+glGm80tZYudV<>iYTFGyVR9tba+h
zqaRQ2^+DWi+4G^P>1}_%x%>WiTh-R}7}syv>n>B%+xmOv<zsrkFtW=}j%xKAN4@)X
z&qKo6_xdHfo%T5B_9s~ldp&7tI@$N=n3g{5?=iBTF0J<%+0HfX?;3f_mmN1%rT4oJ
zoz{NY_xQ-(x7D3q_xDS5+aL7%f4$T1j~hi&+V0}p&$;9Ov3K6#Ru$LZzn7wb2zFu=
zdy8VP5#v6#SZ?fMei}6fqh73N(AZ*&pvFpau`AKUuGq!CFLp(dCQ_BAs3;<cf`GiQ
z51Hg~WL#6$v~#}y+;yj|S@W5)r|o^#{i=jq%)ymA*SH|-*>8`ze>vu!HkH`Y^g9|I
z);R6dP5)L`i6v!@vRn0i+|>N3_x!3Q-1a+3a=mi7Q@NKz*}+xbP<q6BuD&F7t{-Y$
zxZywc`SAJsJ}y#~q3W9}AJ^{`>ZTXp5medj`~1sQzshdmZ#vJH<Gb2tn_k#2OVe|i
zRO5%*_mlal+|)f<gg<t}AwB<9=yC`>PgB)vwfmp3*%8~{v9PY<OYV958v{nH->TwU
zs_rl5;xFHWsuj1Yf63o(N!?9-XCn1*+xtAJr1QOB>Z)JW&Q<<5-Z=l5c9Xv-bfxY0
zf^_xE@p8RSaYx@4ea}C9P7zwF>e<vk*8ZSvzP^9ATHJ3o<!Bo})%6f%m+O5t+w>~V
zw&RC<4};n-qRxM9@7>wH2Sn9Vs^Q{2P{`{?u3rCo8H$7N_MqyY@Q+Kt^}Zo>4&po9
z7d_i^FVl3nuD@-cOTN>I_K^BI-qk))PDl_iA>jJGOgVk1eHpoos_#zf`iI{4cg@fB
zy(QgqED=wx=UR7%YUi=?5%tb3MZcZ4&9YC<*rrV-H@VyXTfKJ0)cT68<Ua4RwZrtD
zU@Gx_pXd43PyHUKuW+v4=~MkqsV{%-`k;Q_HFWs+o%dX{%4zi#Q)zm>Ozl@zTX-(t
zNAq>RGL+s>_cVRw<9naew|>)cAhr1N`EF>p?fJUx{m5EPXSxrnyKb+Rcw$|oD{FoI
zEtfyneidtd&z-Mn|E|{fs^`70?~=RjBig3ts=rX@*}n6!z0aYWj(T5;3a7qLLETsJ
z6|VcdljEg+ua)b0WII3QN+)yo<>nesV(IZcz6kxk_^b5F_ug`G=X#Gqwp;hTrPSlu
z&NF<cQ?+|ne|dev<>sEZ$I|D@*EPNTeW6-+yYl2tFP?X)%GAC;l8Qat?I4ss?(>pR
zc8AJ`n~$kJU*&YYcS)sBb^KD<RW1AJshxA_+GF~i)zspLS|9ov4}@N>^_0Ax3*|N)
zUtQzt?qiz9Gi~2JA053`<1Swp-14kLm!A2}OWzizQnaU3r0RSk7k}ILL6MH<can2y
zC$gR0Z+-Ci4No0gy2@1gv1|wTd5GYo+JDnM&*WN{%H^f*_pNlpV;u+gT~OG{N5=eq
z^V@#^UExdWe%HYCUZ{v~de7XnoVvfm>-xT@ZaxcsyZ`RnuQa1bP5QdG=v%Mtd9U10
zx!v63Tx@;vb1FBt@Ar2@>33cKgif!!pO9+2Q0rjd`B3|Rw&C3G-TBhTJ-*BO#CMib
zk+$cA)cq?FE|)*HTX&wHYCO~ZrqtqdKUb*s6zl#JSH8K;pQ_DX?)FOQquMhSZu@+8
zO=tT&v;E#?>gn!Yw$I&9tx{LXRhsUfR7-qS|5BYV$M#p*YwOQ-PFSsaOxN$O>ow|m
z;i{iXPvtL%bI&)a=Mq&PvHGWKACBw%z*Y~nu2I=->aS~`IG53tT=VK&_2_a+Jzlk~
z6H+ga?R>yBov+`O^|k*;>7~Zc*x{z*bgcMX?~laN!}UHQ^z86Rm%FByrX0B5M-_Tp
z$2s5WYjs?4{VphXdee5AdVKkJ7G*bgyH?%5%FRFZ`L!?mRQ0s+*-zH@{PN>6wB&A|
zLf?bCzq;SI7fdU2mp{9EukW7R=F`$t3J$4B-F<~r<86J;h^=~D_0J-A`Q7`6UHa`#
z<4ae`)xQ3?pu;Y!T=iKQTXL00)^pB^`|Q|rz=X2a<lp*@oATanPZYNDvVVTM=VeQM
zT85VM_T2IF57+#lbd_>Duyr4^^46Q|d;5c>u6*j$MMv+u$(v<r^1TZNeR1I`lS@}g
zalGt+^@pr?O}~O$Qu*c<F0U6{-PC>#wVo2|d0pvhZ>s5I?d!?;y3WH~*D<Q~impG`
zIT3YE87rK7yotR}#ht%yJ4rQOtainfZ|?88Qs1#mJ^Y&$F5d6DhsG7TOYVCqvOd%H
zp89z)WuNVHk^6b0(q}pztMuslyUtTWr{{i7hSo=&)2ANp>pO+I?Nt?zx(`T&tMpRq
zI@@qnexd!Tb*XE(Zh1_fL#f1<*Nv3hc3l~IoaN5X^tomlpW9wreN_D;eB27l-afhH
z;$T}(wVF;Ik9~2F^9k)veIBK<&vrhMS~}f*Y$~3wd$=mSt-k!~`sYshvNV-cIb8SK
zRrXekzw7uwy>F9Bxb63RQb`wjz08##)jzqzJ0CY>*?q5RTXB`P{eo+L#jS_?Ic3X-
zJDuw~x!Tfm^MyVT&>^6jS9AHRY@oh_X&Nr?v-%nz<nq|=Po<j9bRQ{|_;S6-?o{Tp
zTz0tbLkK>udZ_dwA4+nKH@WJO`Q&~MAKHGIpQ$_4I7ZnO`gi12`B8S}7B0^Zbltwn
z$-RE&(&ua56iTmdz50&ltKD$vjrF}8u6#q!pSieGdbof8CA7X}zq+K-wd33(cd7e3
z2d?p`_TVdA{N4h$?9t)A14qB`W#LQuzDMhex2!kU->XaI9+9mbw)a_6OK1DNiz!cC
z-v9Bl78WXL`ui)X#?SS8tio>7-)l=HzOX-(l*d)!PW`*Vy7rm=j-ZGyIQW*K=iwqh
zY9CcKUQ_#()cLt@p1JC$-Xl?kr&eC7edvaVE*I4wsBqD*Or@%yar=knnOlC`yiN7x
z#tZg>z>)}1&;Q)QeeZJ%{YzqL;%B>_w@v43|HX9NG>y--f3DJF%d^(g$!$H<d6#ZJ
zvVE@ZT<=-Pc5|JZ`qm@HS#Bw}18)67pPWx<x9Pe@jvv}yIltWI9kRW;>*iGBg}NW)
zD<5k8pK7@8^S)5}L*Iwv&QJIE?5TM2xMu36^rs#!e@`ekx9NJtHoj@Q5b=GDpR!(&
z-|*!JZ+KU?S!HY~>=9D+-WauB5c61}$5yKHhyTRW`n3DFRB^E7EvJ*`U%9*G`8eF0
z&;I3ziKQ<N-0!i*(x>WYuKsd=%hm0=-r_2^c<#iKR5`>xMrcXdr5he<okQhAwKsWx
zPPi%V)Vzk`MCs=ar|fYJ&-Hhix43TXRXa6&P~<A9_Mduq==G>;J6EOGy_Y}n+V?*?
z>d}@JQ#s&~$IsmB##R+wdEt8N%s6|{(iKzrw*zl`<<s#W6}nROJ|E>D`+be6lg}7(
z*X-66Q#rlQzi-{{%yyMfQq|s6{=S|Y2M*bz&+fZSDRiZ{7aCfsxaj)(S~uwCFXHLm
z7f{`EOx0h^uVP8p@3F<+cdOO>|GCxe1FzX>cA2|0{cd4u@oTmHrB-h7oDkC0KG*MG
z#!Bxxo(q1uQun#+JD#q+-0?!s=T!OWmP<ZIb9L+H%RP_O&DS>GQU@G*+Wl9~E_9`B
zKB=d(-KQ7ja;<k?=PS^=`20%LQ)<%p^TN0ORK9;~YsZHF*yqFN@B6s$C2iYV?0son
zyLF!%sm7!BD|N%E^CpBpc>IQ^jxE8Vx;ePcgOL8Ge}_kRT$bbMj$^sRduim?t@@tz
zX_1T6Jy#1A&vbsleV#1Zr|r5cbUN<yHK9*%u$51Co3h8~)!J56(pE2(ZbrvhcXsOd
zVUe2r&E;oZ`>*js3vNmAki$*KQPqBztAFV8wpjOnbN^j$VZYptT-~wXQwjAv3zd(k
zr(7l1`}3;z`c?j}=QULNSm(pK{nOVxRn9-wy=L2dsd`m~%jHhZo$5UtUAuhU&*ZA7
z+|D(qd*js}F6w>O#v@L8b#8}>t2Dj8DaZfypC?=}>Av|zEScK6$<!Y1^;KwnkF~yL
z-2wZ&RQQsjy+baUfbV-lOU^aEuj@UO*y(fKpAUWRWZTlXzo#Me$@XyfKkoagvc0j}
zy~tm-OV_<z%b#z);iDgx)gs0Ib*{4W&W$$=K5$yeC3V|@X}k-*-L|~+*NY0<@7$~M
zbNxM*^M?OzsjV(-UGXJXU1|FbTJN`Xg_4i9{LN9<j$NuE<T59H+q6QzkIP)-DmVS|
zjjr!ZE_0P_mc8-Wb^ko7^d(j8m-07#&oZ_6cl2%1_x!`>6uC;>_f=Dkw_dx2<Hs-G
zp%PnC?VLMY%yV*;rr!bM=BLsZ>U_#qKC#XbRXlz9bDx*n{(e^MbQE9H`$N9tQ}&vM
z>wecEwRo=gMPjE9^?g{=_Tc+H3U2+O_q%-MC+-u7`l8<RP<^LDcl;6iWw}b-a;6$j
z)U&R<z3o{acANNNnX9zDHxWBs&+U)A@0l($%3S3i7j)QVm8(80xukA;RmJ1FHyQd~
zv?@QY^2FA|{ag;M&vZSUYJ6Ywi%@!Fw`1G<UDv~^=M?w(qp3db^U7HIOv~*WpX;8Q
zs~xe^tL~-d;vbuzYyMpAFqR(GxTx}1%^Q9B>)sR2EuLz=>gw-W525#aUG1gPH}+Zc
zu7`G?Ttfad2VdVEsdc@geVWc+L)~W*_Q^8TzMPydx7+kSxov!IyG`|-b?DMFzj^7~
zB3CKymxYq1d_%=gb>D-s%XECB;&bu#wI1fOkJ3YNLijcZJ+{+Fr6~u&-S_<>s^6p@
zF7}B9FCjy}C#A|y*b_?ny4M7~KmEHOxvwKapX2-P55<T94<ESEH-)RD>{RVTrS9)*
z-cc)hsdix+&i$Ug>b%a>F5CFIKQD6qy*6QYuHW4ec2oD)QV$RPJeBja?YF6>GyP7S
zZG5Wyq4qsRK9oHoJhl7PEA(vh+cO5Ys8DjnRY&f#>Srx0)gn6|H)Pp;uW4J!MSd~5
zYwz<1wy)$WSK4f!?tlNqa+O@=1=qf{{QmE+P=WNzL5!23r0w&EJDux$D!#tQXR24O
zx6tcr)BRi1_}uHiRP>2?OfE9?cYEY^9op?{9Tn<)Ud~6XA6=!W2k!5=ht?<afjiFc
z*{jZnQuP<@-<Rys;l2Y$zwl*|o7DZ@Ypm~OihQ@<{K83t?wnOryNva{1>1acz5k@L
zJGcHT+iQEDE|qklp69yH0oU`^Q0Xad@_xE->vAxS$F*KF&6g{_>pnv&=|w%~DwqB0
zl1kT(bBkQ1>Ap~^@l$EfrgqrQYi!fmjz84+ooaaQ_acP-TzpOUNqxr;{hp(3`=`>W
z?*DN6+rBUFD;>p!3g`OWr&xNZ^67@>_P(lbdj&Vs-@)J-Cu8f0eZN`o_a#GpPtR9A
z;ysAmrTC6rEa|I#QF_I_&0J-!?Jo8Ao2uPD-~F8Gce#bzo)?BLr|G^!Zt)lVcK_YC
zUui~>y5!<&dcSzeQ<wLD{H%pV?$UN&i%MtOAF1!|nCdf)KW?Sp4PEc+xrHxjYL6UW
zKG)NA%X)mdxz7uI*_pcD(Dy!L<(KMvl*5-Fyy0ElW|fd>a&Y}#pzrjlwrf+ne1E^v
zbsXm2=MSxqn(yd_yK;=3-gkfG(r?;-Qj7o3t!^K9%}%q+T9c;jTlf4!)@!?O5h@+`
z{9L6kmG@vw=e2TuR+~H4c`aAIxt<#cyM6Bi$I`Dl7jf-BvC4yduOGR3{p)4!98!4>
zx!R$-4v^#dzDGvs4}H(fHb1WQfzabR?uAYt>V1UJ`7Cw7p{Lz{<?JF?dGqHZ2F-YH
zX2C7F&dVr0;vOxxr25DBo%dX{%4zi#O5VL}pSz!0rLICr>U^)~qodbq+~w<nlDfa&
zPPJExyBw~|pL%bbTBlz0-LreYw82w_uGGCJCg#s7Nv$tb^HW`aY92)0%jCM>LbdzQ
z;il_r+xWiA<@z0k(CJNiQ1dhq9{P7@Lgy#S$tC4_u>H=6@AvRzJGtGu<7h7NsOL$w
zg>$zDS)cB6KGk?~dAQu6$3O1;OrPgm+=L!#-bIBAJ-SkNUW<5U>ws$K6?%Rs^3!eC
zrtwy~;jxZ``z|QB<QbbCvHcwj>nfpTsNcD=&4*f#bA_wUajVT=-WL&W?&l)6KDmF&
z`~0EYvK^G$c0WFqbYeeWSE_hW`$xXQxz@S0riYr>s_Hp4e`>!rmvG(h5vCfC>p2ZQ
zzvwt~#nFX|gR1}ey4S19PvxJ=b1hXnRNQ2L*WZhj^}$`61VW7qzP{h4d!DQtPxm{z
zzFJEz-}BSunR>kYw%zjR4in}TwI+SFUtPVv#vN0=q4o`Y=i~c5HKAX(eof=q-VZg+
z$9H_!d8Mg+zT1^={nYtXF5z6~wx;#Z9iQu-maNBgT+SuF@BRV(6Hn{Y?&Ff{JVEX+
z;{3x_Qumfr`!tllY^TaCp39JHv>9>f%(B`zsOsI-Uw560cy7P1V_Q$Q=V7+_rIOBd
zUo+SAq0Vi5)id|`sO@^tHl6FbEtT}S&lkDod8z90T|aXF_VqibuHW^S?c{RjvR^>i
zsd~>Vb$@a1GFQp<9M0uE57|ztU2=tU_4jJi6Kfv9HGea$54rtOZm#yHt4A)6@8`E~
z{kq@vN;RHty@iS=w-4D}E#H~5-B*)&n7X-sFH@z*)E=(*)cxAn;kx&|Rq<S(W1-7$
zyKfLXo$u!b^e;YtQgMh~|GDPRHIAC<vE5gv(xp0IGPR33m(&f9y-(rVFHGB2wZwOQ
zZj1EQ!oycNxb|gz*$2HK00JOD5um;c$Q4f6f$$#^sFv@ExW1qJLn=RZQS<s#!sYd+
zYClW%=bC45>G5?g?>gVlExp_xOx?Mam$LWAJv)w%eD6IMw^+C3Dnr$FY(BdC;_`DK
zmOE6vnC6p9zohI8b<T-=_By!74)2sQ2d;ThDtf5rEO)qRzg5N89T#(trz*d#zv}!p
z7k{qysLQWf?sNDmhb~7sUhH|5$X~Y0*6n(qF4y;1xa{_|PXN7_ZZ+|gsmAZq>h^=w
ze(%-Po>ay~>K-f=9(v!>bpMHpFUI3sCD*(pmLAvVSM2n;wi{vh@Z|?@cvrVsWv<fp
zzA%+e_c@e%Jk$PTT3%{jNrao~J#fe#eRkhvN)cLC%Xz0<j@0L?p`TCk{#&YUYJQeW
zxa~boxgEo8_IohYcZGa~n~wLX#!qEG-PI1)_v}+k&$T}kOHb%^wQYW`&qdRHlT_p9
zx_+t_yTy7|eopBA9<%RwvR$TbY9EUVH_cCuzv@ccZ_s+br7N{AWiF|?Q(woa?4#Om
z>fy2OJ#pvDHIJkEMKNN)!v}8kO%YmBcG-rf{`&)zeLWu?y;kEcUl)*ZIOP8QQ>uOC
z5`OdNBL>ZQZ)TC3Ol=&^<-6*XovM8m@thd9((i_@_x0RDmZ9rIm0vFXPGzU>-^CU6
zn2U6spHb(YRCw<9T0@^BapmW0A2Rg!GjF_c{xR()e^KNrbDO8t+V72pDzDgYG?lrJ
zpR)bcvVY874tYM2`nbXMyHb=MnM<|0#l5D`G8Rv29hG``?0Yh<^+xr(+`?nmhiU${
z=jW-V3q4+O<>zaC6iTn_J&IaM?>bJ3^rrWoQ;E-g-bd+Eo!8d7zv;TqHNI%Kv8C(&
zWA5o=^GE)*LBMsq&`mGX3_Y&P`K#Qz`{>m3Ob$1l=Tq+&sO*P-+5}EL_|JQuz3x|K
zXeqD%xYiq~JZICcbw2Clai!MWd?f8O*}gxfO2@S?VXMd2bC~;kOQGJ!P~F4f^5?!^
zELw5s^L6h0bbp5<m3Un?I$^-l>pxZs{wgV-V{*Gy`PPa**Z9TNzlEM!doIT=uj@QC
zmHwg{SAE+TyZ&tRua@%&uJu?fJ!0I;RqF0zr5caxIT%Y%sCygS<0y5HxmLsbZT%n8
zO}(1BH*V0dOhfyI`b(P*YHDcl{olrpjVm;++)&rMsj<`d?u{Di8|wF+e|(>&#{QS@
zI=X0Cx5awTmkWEnQ~H)$ZTtD7BX9eFoXG1ooYwRHaQ4>|zbQgVVL#;JdNk#rTd%(2
z4Q>7EL%*#XU8tn5d_(E&vD=PAE_i)PsVhU-rOM~BLrxfU=&UIPmsI6bt^VWpdt%M&
zCwx)p%4+3+a_JHHcT<}i&Yu3qB9;_Z+RiiXoO0|N_df7m3F*c`)I;i0RnM{7udjSn
z^O4;Bo!eKxK?Co5W_np?DxEW4vI5ijE_Zxk)k&|M@^s4`nhRU$s~rpbxFpxSP}aj`
zAN2e*fnWdW-$PFr^<_!PpVpF3=%y`>9JBx0Woat&DWM-rkNEArH(osJgCdsnQ>yq{
zZRgn{UHlUK&;ofJhnsi*#Fpm|Yw3Tk{TFV!-AU7@6~5*Fa<uzzx5syWwM&Pg1tsmY
z*VS>)19kP;+AGpgJaof_K2xcyKlOOJ^_6No+jT~2>1@ZhRMLg!Y1&_P<8$%U%@6TF
zASD92=QXN$u4VDH|C5_u6~~qD`trVurcN(%k-F!-sl`imzQbjg={}HYd{z6^^-q1?
zoVtCwoPEdhJ?{FpU!}it<5j)ZdS!kYOP(|HjYaL=`lf`m<DjZnU;Zns-nhqw@60J<
zN#k=$({oKv@m1$DzUnEIUTXfPvSa51ZfiAZ+Uz2AdHZ^G<%v`06uC;OessfC^(gzR
z)>Ena+wPmo<yN_OIP=nbt{b(m@GZIadve#Ks)tbix!)&({Xczw-8Y}q*JHZ&`Ccz^
z>(?z;>hT8bxzrVNhR-ftWvKBI`J_Q0_ve1@_NP9s=-OvHpBCwS_XctA%2cZCYJB^p
ztuLA|ugF!Vx-OILQoW~;yFd3lM%Jflx2gF19=B5|XDW6~czKg&`n^7{gpA7}*8aKc
zIg9JM0_l0yXRrRMcbDeJz9@5*rsp{F{L0sR3}PNyEi%?~XYlRo&sb*c{G!@rEWWP!
z`szQFURS$K(}%v#8!JCkE~@yp_N(-{`pfGXSGUTZ%|BbXYrjF$N|jWlqx^lJ`-Rf)
zdz}_rKXng5)U)Znk%(`*Phs01<#eX^9_9Fy{i=Q`+o9`L>ErU3&+&BaGL5Huzaq7G
z)N@=nT=W~r#RSB7P;HW2&;6$X$o)=s>z*^1##24lQ}uV<r{PNfqkc==kEHe^sC_S6
zyYD&qi_PvIw6N5bzV18Q+D~x_6;3^0f7BifJ-Eg_-S=!v-!m~Rj<@Ks$0ctC3#{0*
z{$s`W{e4~eJeA8GxH{RgGo8n$7N1(L)>=4Q!?0B3>%73!9O}72&eMtcoo;&W=d;W~
zSHJIgv7Q&M`BLp9mvGg4Ri^%O`wHcj_t|oBb3G4jIh(Hk5nu27Td8oeZ5K!vh5+|C
z6~^lS<>LDM44wXemM{K~%HQ|A$5g*;e`t3q<C?7<@;-TPZfajK^>D8FPHa6~`##WP
z_Pa2Gm+L-3=y6k(pRe<i*m~vr$*S^m`*WR(Ku?(8*-MRcYWGKc*(1i$ROGO8ciQLB
z>0cJUCHHuvtIzgcp00jh@lu;la+T9`+~LYMH$ArNEberva|?B^L^mJbylls_Sn0Ui
zV=R5K_=$X7+m)~ROfGs|`K6ZLbpJATeD3}B*!rk@qPpR++Lg?Y+YZxtUaIkP+oz`N
zjlF+c>-kgTg6=$16;Iy(i{;k6XJs9ad%!Z(XF6UPs`nn9I)60{SG_kteWyj$jwyc`
zkn6dGX?kj$O*LGsCsb1A>f0S_J;$AI==v7*VtP+d#HaeDZn&@m@`nUcSr`A%Dsvat
ze9b)PpH5*qUL$^(-&xIt^G~-f|Dn3)JJkGG_1w{Yzg;yCi0w~}JFe|amyc<@P~$##
z3zheSY~9p%HmGp!a@pz=^Q2s5=>1;PeiA#rZM)!3XWP%X(uwj=>utH;iszEAc1G=!
zr5^5!i>+Q!&&f;E_ctRIh?<|I8gAQusC05Q*p5%!_s~LpUryw+*We?&{d#oU3YYxg
z>-jf5J8tQUN?vfyogGF^tE*7*k+VmgxY|ec6-utU-wqqxHMwQQR8Br-={Js^+M?o1
z3cjHw^}Bg;yZ7z3?N_RF!d}($)b{*H=(Uxm=PS1HP3=>ibJU8z>A6to_^Ng+{7vio
zrdgZ6zsInpn{B0>pX%PX?tYMNJk`FC?4Q~^#kL=-(&_RN@wm^0WPJ@g^!V%aiLENN
zF6Dlqa_fE%k1L+aj!^#cIy*PFsvo-g+w!-azr;!>_EoH<Eht4gs{L8pNv-o0_Qdjz
zy$>9#+_~nP>i3^Q&xg74Gj0Fk{7|*-)b;n({(PU)`LbK+mFJakhxtB~?YXm9U-*)`
z{mC_+Xn(0lw0paD?^=JZ^ZC@%yYkUZui6jN^{4VlowKfezQ)Z^dgXmL-)?b#mpa!M
z;jZglvA?P-sdlRjr`C;`g!|ferPk}V;imnBYu!QV(H%cTJn9^q3Rkwv!~#TlDM|5_
z*L|^zN!3s4;imma72mWT<a%^<i|0nQNmH(-@nhe6ME?8Uc=4<c3c<m&{^a=7_^Aq)
z^}r1RF$vi2Yt>r1n02(Iyj}P6kp9Rq`>$P=;IL#4x%O|>?~yI351IIxo+}F7rs;Bz
zFUl=T+x>@F>1@vz<Z{E!OF-;z%ChA)x4wGDs8-FoQvP1Iuu^wFROBzxalgNyt1q>2
z+P2@Mny&Hfm$trW!n`6^S!=(i7`wcx`A07Py5kOaJTZ>iN--XXlD^JyZ1sj($Jo|S
zsC2&0S1J2L@eiGk>Kr0ff8Xl`svYQ_gR9~x?^RP72MSl&or}McZzlH#Wc#Ue-c-YV
z_tVhhq{vUMA8j{t|Mkt+R6ld^l<LI&^J<a0_W@nwsep2?%T3!&s_{duGfeZ*eGcpH
zbEO(jdt6QQxo*48OD$b$`;WTz_<BC5#!=ZnmHTS4U8e3-$7jke)%~1Y{Bu7i)O}7>
zd%RfdGpc{29-dl1H61tE;wLu0F@3JN-bdz6pZj<d`nlk{AErJI$@aPOU>g@CMPh#{
zwlwXpvDcpx#veIz8852seGOH4Oxu@;|D%d!@`!yOK{tP<;s#UWI?ht*se+K{ndW7w
z?f1y`P@GZ==ibj&=@aMUs`7K)QxN%ZOO@Ti-}St~*Zv!&SJ*F0({)R#@pb2Sa`}C^
zb?rgCCBKIua|yj)<f~nzuGjZET3$E$a&P<Q#vk4D?y8ks<=IbM{pb~=+Er>zs``g`
ze>oSadmfi+JbAvGn>+P(qH5<Vf75ni`@Ki1o>g{G^UqYnxz;15dQ|mn>aV-ckxD%3
z_eEU82j9N_jAh2oFHAYw#?SS4Fk<a5t2nF16>7aLkMq<zftqjH-mj9^E4psVo>aqa
zIi;2^mvanZr+0p;4hsLM1YDnsvERe=1uxh4lSA1}JwIcGi}v7azloX$gbEksgaiQ-
z0^I$Q`@MCQKDmslbHZx%?|fv7uDdQhvapq=`v|IaldHeHuFA!2yYH1tI$!%}qTimk
z&P@}y_@H$qH|hHOf&<zQ{?(Zima1?`TYjqFJx<-9>-V2y=^20d%#&6b(Y}(ayuW|h
zaoEUZDxoCzx#Wty)_(JM!<Mb2B5gUE?gy*xb5^Ur_x&rT`-kFwY_2l*_G8+QQj0Ht
zXE8Uo*l&&{cf4ZEfk#iMYc`eC`E{z{y5~jQ@l5Z}Qun{BEnKXNbS2mNESL2*WoIty
zYS~V%dW4=61|2$UN)elbYQNI-J2SfRxyNZ;eWv?|s`%BiKdUMqmp?WBQQ_2l#Ps>h
z6<_yxC&!E3&qL)O`#l($yD#^u`|Ys7U6Wf@a+S9Gu%_vJ@mJ}^zC`057rry6jJ3~^
zD}UK;U-vIu^{U#vtG};#O{n`MzVmVA>zdwI{p70Gl_z)lRPQ4TySV1RspxUN$05>l
zw}Vjnxaaes><*O=wJ+d%Kh{?{sdBl7Q|t27!nyV<V(UqbtFAq%JP&N`aP2Rtq?gw(
zvCb!L^A+}odjC#$zuGk3Jtu#$+5LkSmb$Y2t_Oay?F)-aZh7Jv?H;;wTz#{t^!41b
z)vG$6$i-jGGjo;HK78uosf^F2cGy18<aD<E&UZTQ_c+A<kBYO*MdendH}#kKQTr^e
z^H0<9#WX+N^``E;QI1FPk;6^x*PWNhcA2_mJ5_F7o}uHZ>`mR@w10~DazEvEo3@|S
z<IDGfa&vP(cVg+wW&SJd3_Xr<=a*YMSJ^A-gIl_OhapsY?(tpLC(c_^k+$c}y7MKu
z{IT7s&r_=1J~6N2l2m)jooDR(a<!hnY9GLtKh>V8@L1~u+x}dw?aNpD2&LC_|D;;t
ztNNGe{#mH_Dtm4HbFW9!{$V<9iE~I@DfXRhCHFeP)DF{oFsAyc_|*9kwSI~9-n?qw
zY1-bX_@?zk>661v`;&-I)td-6-G^}<pH%gdi@$66k^a?(ep@%XaOGgyZez!{oe#LC
zqsDjJaPIv{U460k^L6ueZP%gF$6BXy<*V91hyR!NT{Lxi;hTeQf7OlWI?nk{&(&^i
z^~9P-o8GTP{7s(e_xik2mjm@4D0jQIZGWky6YG~+lk)mdxOL|z-0`^Er_d+XvAIjz
zej_-fCPjP8Rp#EFg#EVbHaQ(v`{LdwwbjQ}9$8Pdo|94Ka_#4o{?x<ec3|t)J+Bb)
zV$Jude0A>~QSm5_uI)~hp2|13aO(WUG@ROhqUJNE=dP-H^5rj{pT1@6{brHBYdL+@
zgRWk#b0MnT#tP>iZ>ar*+``54Fjr}MKAc*7>N$?^-d&m>`=a#4ftruy67GARFRyE>
z-L3oll;d$fZ&dnB=bb7&y8fw+o0L7a`x{7i>>Kw!@LuW4!L)tq#@B7Hs(7yF87jTr
zo)gLp!v^@CAL^EdYrPU$kM8$yL_GPqqU+B6xlHYkxQ6TAm+&3W_x+Jj`o%s->e6)I
zEcN)l+owDpbDwJqeR4jb-MZsyYVln69aBl~I?oj8sr3MNxGg`G9b)}1$|FnN_D9V_
zReDr*34hmlGgo@o@j;X)b?IxrE0kW*o^z9Ozw_;GJ?{@^e?9S=!d2S#i&W=rt~})Q
zwtTqL?R>y(ttL&IT|h48z|9x>KtP8;==GDZNvwOgq-s9L?Jw*XGW7nL@BCc%m13nA
z^`VlfKHqKaa^);I+2*gCPUJ($#<ySE`l1Q*idvO2pHTaj)sm0Q&(zKJ9HR74c0F?T
zs1sNFsJ;?gQg-GRZoBW|tDLd!0aE$qJ|3j9p7hmjxb%kleHCB%_}-6_^>cBj^ytp3
zQjO<2FHI%A?{${YPqhbM;kM^CR65`7lFHvT+!t?IuROlw;?}*NmTEleTrih#TfWr2
zC^0VOE~$BL>fyQ0!+nqcvVM776>d3Ru5R1sg5cm=#%ib3y)@N&pE?hq#s%LzbJcI#
zzfvnN)jo8?Lzj!%ccj8q{e|)mwO$L&GrRoy4H|ggGt<jx^VjTv!C(;hF#&3v%q^T6
zzq!JHOw!CJ);y9sUuyn{@O^K*c-99+wQ-PJJ6FHr(j(jF?2bL(V9Gz2d)c=808~1z
z_C@J&Zf}kXDEp{z$LyVxQ~S25hr6E3#7=L!zmZBhs(mB;l&4$n&|HASk~vWOt-9f^
z17WE5wtVG7J&#il4>e!1Z6~4D6~ew$p1Z=HR3!Jhx>odY)t~NooJ+h={W?@WsdEp-
z!FN8s&ZoHgDfBERpnI<BD<0}03;|#Lof=Q6(uH29Q}(!q*UEee<qY#Xe7-olwsTc_
z-S0HouHQ`eBW>69ruTd+H(u3itykukwIY4()2Q^S+MCLsd%hy_F@0Xey3Z}_wUwsj
z)QxXz57L1E2-MZtKF?B1$2~rW)|VPjU3;S4Q)Rk0B7Zj#V?rDPy3c#x@#2)eu1;=;
zvD~`l=l&jmZob3zKB>nRqgys}N!@&H<C)$ww#_H?bhh886XlY>f4Re%m)>*TsD*{D
zl+#y>+xK{=t6$VpYSMInC$;#v$F$^oOoEGTdy8Gqx^~Cf@8Qaqi*v1hPY~s<?e9sV
z{=?S4?zk++Q=K#9>Yv;1{+=`QjYaL=`ld`pn(m8nw-eDmechYl*6V9MCOFti*ZU~8
z`T9DiPgSoeC)0A6?!)ND4|R_|c0RuDdxh3(+Rmu>6JFlrnSQU&D`QFSec0IgxSxZe
z^~IWh=<d_m#-pBtzQU>gC&IUR<i$BFztygpTMGR`+VYE)?)LTS$`hx~DNLm(Z>lo)
zb+NDSn{(9{_x#?sKGXSNwZ#|pKKS<aXDl;ze&H%z`NZy@w&#thq%$3_Q;YAr{zZF-
zTrvUI{!67_a<1`xb?>dE8ZY#{L$3U6-@mX;XWP<z)w^sD_jt@bp33(6o}cOJm+jJZ
z$68<c%2#*)G1Yk7?O(SYn8p+L9Bl1X<!Ac7mmFXBy*Jx<w)=eI-c2r&>-iQ-k8jSZ
z_DlKu+V2acSND08YCK=_m|D|IwR7%pYF(9jxM~07=BLsZYkcO)H&(r?c>3};Z9i0e
z?&l*{eNuWTzErtGhg0^NhI8EuhMqPzoIU-Ig)Rr%{T9=7rtKk>eRo|uY}bpar4#q=
zM14{F394~Ncb@EPza*4iv0sp@6!oktW6evr^3@#|5zp*7Lh<6h*QeWFRq?2CO*cH1
ze$PGcg}$GD&rrpO>Ib=mtIC&)KlgYON}ns=6VGV((4FJzn`L>=$zN=C|Dc6sEGeI7
z<mwK+AIkOJk6d{RyQ%$?)WdbxL#f6KwVv{ok7}K&@>k6pefftTS3>3It3A8cgX#X2
zYkXIGsq~@dC%NaNyPuazJk|J=n!jkDrt6zf&uw9!EPcIqA?ua%<#zjeZpwPO?Kahy
zO24$VL+r1IlBRq^#lLIpw?`cC=G-zj$^AYj7jIwVD3^Vdo>22qU-{hn`>yS$k7?1I
zn*1+!rtjmbzGK3*PZ3LxXczzG^`C#_=Xfl)Y?p7h>v|`4dQqRbNwMFRs|>w9^xY30
zIeXNJt9?{oag~&vxwX@3v)A?9#Wp?FE=<F@pHK4oPq<C(vW;)LpO9L7+jfyky3R+o
z=(_9DBa2+6ul?sxdSi39?YF6>)4kW@8jmV}=yeWdj|h+bT{$^_T~5dJSoid82Yy~w
ziwympGUV60OY>u2l(}-qb-kFI-C~>!B}2_KsC;zC;nd^F^%i;^E7niGWGd?hZaZ@O
z{y5d2xWnak9op^seK}Qn-QRQKj%RzmE2k6nklMb4sXd{`N$&h&pGVYc{-XVRlb+8t
zRp*_d{A0J9Sl=-Zoo{aIkX+ZPD!W7VCtvxPK2KAN&;7hr>C=4<sM4qAPmRyY_NEKM
zFdi2l!))22r`9iQ?a6JuT5a~)eg`qNbgJiN>i(hUQ?>(8=y@hrew7=q>b2G@^UK_2
z`&|$GWZM@Om9A3WcZux|y$^2N{;71KzHh<3U+k)n;zEUoem{%*-X&E&-Egk&_Ed`=
z!Oisf=-QrBNl&epe1{AEzNByaee0L+-@v`&kPBX)Qkrt`)xIdbsjXj3?J*q}Q;F}p
zU;5S`dcMk?AJ@K@N)OfVatpWZH=)aEYiBO$Qkj=gcG&KN%IQq+y_i0CQ;na>``J{#
z@ZG<CuZvynr}}{^+%;d_^xXV(^R2dcrt{KV<E!Q~vHe5sEBeZZ8b@;vH+@crj-Tpt
z-qbGD^MuRab=;0UPNb5*E63RBP1|Lx_{#e-ResMi`*IsLaNYN7Wwr8pGBr2%{9L6k
z^z&AgpDBO2KC8{m_1uo7$Fx3Eiyw=d?|fD37}x%z8eekt=e`F4ef!>c@vILD6$fAM
zv4+;GdcINqvG1AXaxX^Jer)|+$Gy<$bKehA+0T8B8(N>QbD2=<XJ7L#s(<M2%jn))
zP|a_Iziv52d{=2YpVPIE;zqScQ@eEI>)P)t9>v8qUy7>;r`mn!@Bw=+b;X?FvrFI7
zcez~qW9amzJXFs|*`I3XxrcK<zvX(MzHh3U5Bpxv%64+Qxw*#H=Q^IpN^d(3akmFq
zpY1v-m2`4>xZJw)pH$<Swl^+rLXT=Z)%6#8bfxZkA+>n1$5Yq*L#<b<Eg#)>Z5)sJ
zygSzYwbb8J5B2<W?Vq`%SG6;0A3b)s+&^Qj!(=-sx7fGHRhrJ@RXlEAudY0C>YO69
z<o-Q4N?&U89$R~;`L?Q_Q}a*#`KD{1yib*zJJxtzYx%1B-#sUPvDy8D7M3cx;}v5L
zJbFT1v#C_&r}8&lFV|Xp6*t-6bv_~MgS$2f*v=2AbTUm}>m%;<q3*pVxqMu1)A@F+
z_|-BlbGM&d^l`OUTRo|8674wmc_Ou5;|jN}Zz>)2JW+*HdP9f1wrf*+Rlm!Ti+}9<
zxm@{k?Q5CVKX-iY-)Ry0bmvuwcfz1UXH6-<;fEYt$5oMD?k}$2wN>q>2!Gq}|H^i%
z++zL$x$~=CIt(qUjf1M*UHz+#Q>x{$t*61Kc3bwNSC?%zmA3h%l8$SglA9ja{aoGj
zzUrBKe=qcUJk~xMcfPr`d)eMl_e`jJ>b}B#w`;k7>)r!3jVI>Qwo<l}%gsFw#nPAR
ze1@`%Yd<BH9@BoHiXY2<*L;1Ad$L}x{$6c*RQEG|`*Yp<HmwiUzP`#|_r4TYJh?o&
z-^EHb9@jo~EIsl#65B1e5817HKIG<ayB{s{Fm?NGXQuX0@pb#X?{c~7SB<~A{;uoV
zmfPI=>KUV2HB)lP_@ldC@lKm&SIPCef2Mkd?R`>@Ek?I&wv`kgD!j4Hq6_OrEY)l(
zw_jt(y3dYm-OMF#{e9Q=)5o-E&Q0e2yR6Z^Yf25Ip2Azt`@`8^PyD9fmb&A9>hZYR
zLvHUU$o6__e(9^_dapJ8l18=Ac0Mhqi&_X|4gw%h(*#1DkJL0|i^A)!(^HS<x{pAi
zlh2=1bw7OcDd%1`euZWsb@#XBd2MXB?mZRLc&WGJ)a>J0=LtQg^~D{Z>gS=tWA`uH
z{JGBmV(AH0zp?Ytt=~}hsHl8m^HSxX+IVbhPpJ8~uY7ddbE@%l=V7_T^Ht9(y{i6c
z>Q9YB)e>&Xg^Hi*dPmnTuJ1-uoJ2VHyj549uY1F~`Rm3LoJ?h`d83@KsokphuJ<sx
z&nI)!C!cfV;?C_}7iF(#$EtQN`@8PfRcm^g6Xka0V7s2MO~?Iw(tX~f8qc-<Vy91a
zy`^iHZ97gSUF`aI&0n5Zak;5+vs%JU->Wk%x9)t(R4?~)Kv$nQhp0A5eQ&zf!cF@{
ztoW|uXDaD!_g8es7sUJFca8)PYF*|l+;qH8HGb;%tZUUiv7eButd?^GsvKe*f?V?W
z?3)W!Ke>d9??1}ViB#q{vR$Tb*Kyo5KRJG?`(L5$lDVYj&h>fc+P;zgi8ZgE@I{dd
z2dZDJGO5ely0L94p(SOPZFnx<!B*MXutSf(PM_GSlB+y#otq|Z@j>fKu5$7*OTTgS
z)E1Ro<*NJbu)$rETUK(HXFqZEqgRY-SAqO#4%9dq`+GTD`K#`;i1MrETdv<l;Y!cN
zA(kE$e=dJfkGV*$a@*?hwI3WxFZEr?S_{8<pHI54Jz`N2N`|UuU-_hRZ==@iNM#>S
z*kOAf&h;IuRP?Ccx5(Y!bzf1QPw2j95G$VVa|WSb=2EThSo4TjJVpKHD!JBMwt8Ia
z&2_v=Eq$#1W1Fv-7vwH&%O5*ktowId`R3kkLfP*+PX0GN*Y^hl>k0jS_kcZ@x?;}o
z*(I0s{rdu;^oRZ)?|<8E<bQHs*SXps89%bG5dyKEr=j;BxU19mcipM?0=dhRi$34n
zed~AqeP-MA^81Qh>xpFEPs^nNJKCAcJ!xNdI^uIqZhM}cYC2!{Ex6|uq4s^`agy7u
z`~I(PJkg+a<EI)g^nQbDe*b;$i3(6iU*n+AtH`!yyt&31*Y+T%=N>mw)tAb7j_W;@
zR6306Ip^Cx)89)=C4Sv&eQy~0cH8FMWh(b<Z0)dJmm=MPpKSZWqO#UG?0ChP1CO3i
z*X&zzkE6EV6RGxe{kHxO>84&y-5WP(Xy34GL;a;q2Q@Xc`2KHW$Ho;JS8k~5-PG9W
zd-p~S^$qp=&Og3SQ)B<jcO6}{<gO>^0|5{K0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!XP!Q<1^?w4IdNp-#+@N9Eh7Jw&mo^>L)X?Jlzl|LmS7=<hp{{pR
zW2f)k8#UB7)bBh0_&!aI{V(5jbkS0w7TzEL0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JP82Z7zs`t-#Q|9MjhnFohF
zS{1AW0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!^0-gWS
zrtg4;@g*cU=(K~MFKD=N+m`Q?xye6n@ZmKNo;k5}m3qAju|Qxkfd{{Py}t3y2MS&J
z-aof()9LU@1tp((?b1=_wfL-nG~<AF2?8Jh0=W`6W7AD`X?5R-GIg0NU)T)-AOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAW(e-9^CPT>(1;xxvVxhvhCT+-M0MHQpoD7Ig|(l
zKmY_l00cmwRtapm{ww#-KjgzQHF?9<4J{_On^Ky}OLpnIaNox!mqMmeua|GK$;hAg
z8dpNb<&a8yfE^$J0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JP8Gl88Sy?T26oEZhT++p<xdd}<kQ4ve7KWfM2ZXf?{k(=DRQ{SD>So`r(
zSLWRIV0-=q<_u`Mw9hT$O5d{URZmPDyz=NWRhfTH`YA!vgO|1F*m`)8tNi-HGmcvB
zkT*(h>4!G7f`A@@`<Lss^T)TWScxU|tVApj009sH0T2KI5CDN%C(x<siH#@yZEBgj
ze5>0DtvdCeT0-h^sP%S_^79h-r1u?boHKAtnY%n@`7U)2o;j(smAq7;4Fo^{1V8`;
zKmY_l00clF1c7I+T4&8m_MBWo))t2lZ3}sT00@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2>2&(_Ki2+vdlH_
zm$}P-E?D8V9)FosQqq*@a&vw^;P9VKElp+a-=C;;`%T*k;)4JPfB*=900@8p2!H?x
zfB*=900@8p2rMShb<?N5zT~xurKvpakJn!I%<m?Zlw9-l?GEj+a8e1GHV5<v5C8!X
z009sH0T2KI5C8!X00BP)&KYn>m-g38C?UZC?Ew8C^>M(@C<v_}00JNY0w4eaAOHd&
z00JNY0wBOk;F4YXF5LIA$t9#62j1F%HV^;-5C8!Xh(q9-7j}Gcjb0;5Q+ZvFQEzQ{
z*!WUc_BpX-=lvg>ScZ~uXd+h-009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0b2t1
z{Nu$}p1-C|MI{gIbLDdzpS4sal)P>5O|y5~U_qghm!AFP+M66Px9}xxIU*ehfB*=9
z00`trpyx>oe(|?Iyk0`)#UaPG2Ae?u1V8`;KmY_l00d|PxBsQ%U0d(`LCGcO-f`~u
zUG|$;+R78RJpYn~_kK`9f&<z?_<lf-7-lh^8`!jS?>#4$Dru$-u|WU?KmY_(1U9_p
ziG81F7+P}4Tb^FF#S_cDT-K^oapv})u;9_b8(cE6)RkR-w&=J^UYk@xRtE>}_5^((
zkRpL=Px$L`hu<)@G?k((>l}T<6)m59uCSFqn|o#bxII2AbY=P7{`arF{g*|xO2Jhn
z2cNS_myU;jQE*FDe((nY5C8!X009sH0T2KI5C8!X009sH0T2Lz5CpdV<IKYs9x}Ov
z1P8PObw3Ev7m)`DfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2-px<Va?%h|L$j#%T%RJnUDqqKmY_x2wdN)GGygXClt2w
z=TBZV=b??;R=8y4sGql6aX?+gR!;uo$%p>>qi+h|(u8q{xZUpU@bPVZCzp`m5TYG~
zXwxa>@r#QezW(RyzgR%#%^{^W1zSJ>1V8`;KmY_l00is_bl-E}o3maXTiQzdoRJa)
zKmY_l00ck)1V8`;KmY_l00ck)1Qrv}{e7GnH(l2Esc}<FT{&^j_1fP5_0*D*s0R=L
z0T2KI5C8!X;3lyC<LABE`pKsXZpm#Y^wkD|+cx;Wv0V-uQPzsAjao%1K>!472=t$J
zSce{enOx>3ZOV!?AOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaY69ICoZj<+Gbfj(QfPeX&NeMtpYcgi?NXE%5(Gd11V8`;KmY_l00ck)
z1V8`;KmY_l00ck)1V8`;Kp;GUFPC}n&?(oxTtI>Y+5!3j#(|$62f~lg$PolU00ck)
z1V8`;KmY{v3C#Vx-0relMwi@D-vY!00T2KI5U6DW^H=-su*3d-P!USjGHaB*dI*f`
z+<U2$hL0^x<=z*re!<@#m{^9E)k7X70Ra#I0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!Xh(Vz3E(hK*;?AihCEx0HLaR>wr<SS97z~jo2!H?xfB*=900`(1IPH(u
zUiQrICY7OO-Gy`RUhS7}m(ipI=alu29rM)hrk1rLrQ*X00w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&00JNY0$vD=A6|K2{Yxg7q2z-*zHr@{-6xm2^6&Q!Sow$>
zCY8RW7r0Og0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHdqf$m52Jbd=cV@gvgseuy&KmY_l00ck)1VEt12`p%F*S)VcPbn$+
z_6h4Yb?QH*G?g`u7sUqw5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5J-T)_-XUHZg9z@(pIi@=@%dNSU9n)b(ug_
z!UPZi0T2KI5C8!X009sH0T75vpvTUC-SLK@&zHH&R$Wfqyx&D*N>>@RGLSh4fB*=9
zK-Cf$+w&j89y+0IMOSV<aM-Rb-)L2V%#*|XvsP=^;H8I4NN_+q;Ohrf`+z{fKmY_l
z00ck)1V8`;KmY_l00fdIaMwQl|9ROxWnnAj!cpJJMfj4xCoFS>8w5ZA1V8`;KmY_l
z00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY`E3A{G=?M{8WOesy}`U{5d
z`o*bJN>h32j%)OI`gc=G$RBdhtuMp_0TAFL@a%D`zq~<*>kCS*Gj`gltE~TKp(`(&
zIcE8{CeA2y<<jk*eE6h(^9z;aD={>I00@9U-UR0Ccm1u!ep5?Ij=g8dmFrzHwS@G`
zA@9}%>p>t-0z0mCVw*$We!cW1uf2A)8}HfV#nM*hNflOu00@8p2!H?xfB*<k1p00L
zpBPQOnz}b`(9ofwV?+I=O$RkKwD|sSW5>o78dq+p>)q7Y>3jD^4fPH6`_4bUPg7(6
z%Xb}Jw50SS90WiB1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;
zKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l
z00ck)1V8`;KmY_l00cl_2?Vw{>xo})cgeH@5*&2e!6E0}weY2*-!5{K7d+B5X3@Xj
zD_!Lhx)8iTKuO?VPxQX_s*64^bmf`TPXFYAe|=q0a`@LDJ=<d4xdo&d2W3Se00ck)
z1QI9k#<K%=yKv8uMd~tfj4%%bKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)
z1V8`;KmY_l00fK(T>1HO7jJRdtEH*@<>RCOcIqy}N+FFhMPv{F0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X00D0V)_HhV
z-Rq;qmcHd?pPh2=+^63tQ<dHTLoo<|00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!KG%5P10Xbq4kvGqQxNEe<u)z)?^T009sH0T2KI5C8!X009sH
z0T2KI5U3FX_ng~p`FrLKDRPlDQY|PZ2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zWJ6%$b<@|Gz1_qzcbN@v7zzR)00JNol)#tUE_l96kIAK}yzz+<uiduoloB#e4nZ3k
zG6Mk+009sH0T2KI5C8!X009sH0T2LzDk9K-yM6E5Yv{z%mwaW;v=@(kXL1RdHis%|
z(kKcDfPjJk&Oh=z|3JOfR=p{@FT6nj1V8`;KmY_l00ck)1geOD={oO{etYlpY~z%&
zR%8`bHHrcPAOHd&00JNY0w4eaAmE9><*U5Z<=B>!N=Q2no~i|E0}=3bzpK{t2C6P(
z0|Fr6k-#xueb#%saYM_{a??jYd}X!vqe@@WBQme5_d4~eRX==peA#*Pv)5Q-rwOEd
zb-)1|UEAjUQdi!5%1)cUamxFpuB4u8M~xb=<!wDCm5|_&ay#(aX`vhhKmY_l00e5B
zz&a1ls(XFZ*ix2*H-EF;j{h20Qc`ywOU0}4YC!Qp00ck)1V8`;KmY_H5jeGesS{Q{
zZct$=SGoM$V&J%LB_$&j3)zAI2!H?xfB*=900@8p2!H?xfB*=900`7Lf#+9P{=|bj
z^(wgJ<dc7S)(g+RQrgPRK0D>!xlg}QLV^R@LHK?k4~H#IANJ-Ur;jRaC3RnY>Dl`{
z_+9&vrLNp>=28o;?m4#PlBVU7`M?bVAOHfX5O{9tW;^_9>4gQC+;7t8Z!d09U(uBp
zesb)9x-nlC)gn`=SJ(jpAOHfjPGE~-r?cj~)wi&fE7kq6?}Y1LE_CG)3)h@@^87c8
zYL!%by1L)|74Ccd^MYGa`63(yKmY_l00ck)1V8`;KmY_l00ck)1V8`;d=jAE=R>&O
zdyn8$-JE>(LFfhn5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!XNRPnftGv|Z*p`z@m85>pwpPN^tGU{=XshdP+U~t~CY7nm+N>3n
z8U#QfB!OM$uRUYyLnaos^44wI518`N7X_ER=iF|~-!pGWVJrLYH22lsTij6imLcm1
zd4T{3#3hjGJP~%8&Ew;?BIFJNAOHd&00JNY0wBOeVD;4>==S6`w->gO%MR!P0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X0D)>DF!q}(54h)mj|*;DEqqXpumnE6
z?$+s(F8;Kr6&aQ;asvSn009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sHfgck1*?>h|d)@v)$t{1F3tS)o0w4eaAOHd&
z00LYDj``}d-rJ2ETAIoQGk^Q>eeaDfLrbo5Lk|dm00<0wc(>1|{Ax_;Tdwo)th(1n
zjV*mkszwK&`G-3@o-nZtB`JFl4gw$m0w4eaAOHd&00JNY0w4eaAOHd&00JNY0<j2u
zx$T1IyY!e`QgZb(mwS44i^*l`G8RYV3j!bj0w4eaAOHdsffI*K|7wk2-B*T|Lw4G7
zgN-K-D`H8ROaCvP==i6$<4aS?eIGxRKDqpGg8&GC00@8p2!H?xfB*=900>kUfn|HW
ze9b*~4=Ziu!>6w^u;-YOrAi()YhLr@JKrsR$?B>Nlm-Mq00ck)1V8`;KmY`4n}F~4
zc3t(au>9m>um1Pw(zmSbR)Erj00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x<V4`@9tY1@<-m7JO7=PF
z`u=@Bf4wx7IkAAPAOHd&00JNY0w4eawMSr+ao6`e;L?eOFIjs;QDP7P0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5U5@Pr*5{<RxRFrzl2=Op?Yf?B?AEv009t)Nuc*u9iF?lc(>G*{l9pk<Dc4&
zFCoDJ?Ew7%<G|u^Am$j3yg>j2KmY_l00ck)1V8`;KmY_lfQvwj(}%rz$myd>SGm`z
zSFQTtyW>k;Ie7Co+wJ(Tab>QOi$C;$00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xq(`7~?!9A1Y_@X6
zm;BYvla71!s!kPKnI17%1Ogxc0w4eaAOHd&z(Jtj*8kDc)T^m`;|2{K8<uORzqIL~
zriK>Z|84BpxI*K~4RyVn8asXO-l(Czp?=@_$M<P!?0@;Lql=bxkAA+<YHQv*qVO%d
z{Ne7AyH6foauoV|9Vaziy!5+!eO!c+!v1QLQ+nMzv~%zArEkeq@3wk=F=Usi$A37j
zR7qRCsiZS)m$B_OwU>$?yF8)uH|?*X?L&TP5t#E>kDae<|8fya3K~>DuGa8a+~s^l
zJ)|yIION&-_vXwjWT|?7=juOm=FfYs_43TJ+N5jy+P-qYEo&asvSKP-^HrrEc+vwq
z9aq1wh$V$x@6YM<(5No6OIzuyp9uT7q_6cTmwnLl(**Xq?Y!Q*{QUERl0U5_pU{0@
z-t)NmD^4j*Wj-bJV`<;NU3PinpQjYDq@Pm7*DBBZOx^KI@Iwns?MM7D>%AY=(*Inc
z)^Y#K(e6Lpc4DXf|ED_asmfJH%`SpVYl9aj<Qk7{%?Q<geCPAlw=2!KvfrGNlBV;9
zQ1M%j?%I9ilb@Bj%iP!3w(C9HbhiDLN_Sn`_V+*Z&VtfaQu!bp1X3jM$xX*S^4|F`
zmbTKhOt$$}TRIiTx31lC=w9Q_DsqwX{y5xAJ~s%><%Igqo#r-7D11p(`_=Z3wJxca
ze6>OQjp(a=3B7);S8ES-uS=Cr=<$L(zZpk9alvk{%_(9@b<O$-5$!dWRJEhj{Z;2P
zerhRwtxvc8>q=J-t!qwQQu|5ChcnHybz79_eYx22ZTH)Bd4$?O%sroH7A?Q&Zr^=V
zx=P)8@fEMy*6+ELC-?R*?2kPj)@uH~&!<A^*F9&Y;`tgMp*Kwex%PXi9p@6B%RV?|
zr)xX%-8*dCgUT-3^+PJ@T<a@TdfR$SJ)Q6QwW{5w;!lkyuHmVaGZ{NFKNm`UpEq^;
zxb7R~uE*DVFVO4fI~T6c0p0Ye&lgPXi}l>eeSFZhKNertd}FOYT<tbZUn}FWDHm0I
z+j|TueXjnx<F;-*l|58{SEZx;bH5KS?DxG+jjf-0u8Ml5+6i|!*F788agbWqsPx<N
z;3~IDk8BrJ|F+>2w^YNq&Z|Vfi7i#_z}DZkJmQ?U+N5|+sHEUU-S6;qpGsv{s{Xd`
zlX0gL^(G{>-%mZqsd-we<E-wy*!F(0>iHSkpQ=C64y+~9{<G>nt?Y07KDc$kcoW8c
zKJb<%o;I#|j7ksHUv$GS>CpDmV|Q3o`jWo(d*aosUe2M%&rtc<&ReTBonCA37fZLF
zo5nM>_?}iBPnvh!x1~z@nqSL$b<3$c-;+7q-QmK6cmHL}=Gaojxz_w$zcXQ*UT$}`
zZn=Gha&vztGSqt3mb0ABmb31-fp}rQPsYvJwp}1y7y|PCxvM*j)&I*ycR$B9-uUj1
zzA$g|xrM9zpXH1Hqvr9(`yNdDGRl1A0t9yVTk$V{|MvFQ&ACcn`vRf#Qv3d?hjZ_%
z$I{0=-wdrU_P(fV{<`a8RlLB%rZPKu9$zhPxqg)Zm;>Z<3%K>ZYY$F~IfdGHMm|5h
z-%g!pbA>YxH`?lK&yBP*@D$PZ9Nt$tu40<5v$^ukO%L~el&!whxurIq`{rd^kFnCJ
z?j7Xj&-EO%)#KW(Z1YVuooT;!<(FD|YX3B~@KoE8tzD+=PZd8_yON&=+;*tSo14GP
zDR*~n{AGKo``f99Q|lk!`$3`J*D;;nnU2HU<CUpCxm{9jQ~kc<bGI{9yH=j3{^z{b
zcE3XA6Z$+V)IBBce5ms-nX7NN={S*UeEHnPw_DVUs^3xmT>W3z0r^7$R6h$9{zI$G
zU8%N{+~?u2KkmIw^PUhV#kBr&Hh=N%WtwBjqt@xQf9LV-n?uW<o&UVt@qcR9tdjF?
znmcouqHVLSY=7PSN4kw#sySE5-5R;ZBWm73g>#q7R-fyA*VyT8*HyOZxaJ9=^w_ow
z?sQat@EtD7L+xM4`$oR)8hFwJI~`ZQu*g-WcK#yjId%Ew@eeJ0=!S0#T^Xx9a(~Fh
z?W<kN^<aDNihKV)biE9Ee%N_0cWzPPlDE7z_}eYMSgN9u16~~R>^Xf}RVb-zkL|q0
z^*tjhz3zCx9gpG>x_xoyC;F|5+df?nKC;Irbrn-7^xDc)=YgU2?NZH$nkU8%SG{*4
z{7u{Aou}P<?&>EjDr}{kpXyvd_xw|Lovy1-H=f{^o762|sCcUSuQh+UU#Q%=?VT&0
z%8pR}x%DsEUSI9lmVc_}#VWg8$01uzs&v$RlbW~7;gmiRuCgbVf9U%qa(=POool|L
z9i=8it!udRF&&r0^VoHt*jM}Wy}t5gx6sR7ztE@mzP9cDsqgQUg!*2eXn(0l+xkc)
zojiXI<#w&V)Sj!{_PFxVP4Bv2M&;xB{7ogj?Y)FrO(*J&I=>L%)bpR}UsQTIoNBkq
zaNpl~3#8w*->cG#dRIy6eGFB&ypE^ba((-9tL*jWuWXoU1?aBVOy@zdOX;f}h0?3)
zx2FE4^}zicq@D++;okbI>3D7m)-b-<7lDj>|6lMotiB&bw;eZrqzE1$5RrgtzF4jP
zrq3r4KVlUvIcwAB8{${4I%;;2%zPv9`mnwClxn&qcOrdX)A3MfHl5$+9$%DOmbsNn
z%o9zeTyD5|38?m;wyzv;%bEwZtmsO)J^t59&2d@Wn-=Lpo!`j$n6^*f@nx=X|M0uZ
z;8EQ?O!rk&iyy0figDChYL44n^MCH|Gv%Vsx_$yBG@YmG#;44W)o*;~%YA;}TOU{d
zR_Y-dg6nv$s!t-@V8OcmCDnLhKd)M3s+=jiT>A-?Ufgd9EtTU2H9ukQ!M5j9zS1!l
zPpyvPoLRWoXYegmK&gIQtKq)u!?%84{XCT3+@Ig7c^>@3><@8&Z${`-jia)^ypJ2p
zE!$=44%Oe0&kx^I^1Y7stsnKdxIU?MjIZ$A`KLDTG#xj&;`=_&Gu5wq|6Ihwe1~_w
z6S`mco`?Gy|9t6neQt7%3$mV6e&^iu{d-^EmC@BJbClh=KS!(0e%tX+u2(L%?e|ba
zr{gMbC_SosU|jyxI8NnbeGVY)Py2qpy6O*A-(2~G-e*zSeO=r3_doQ`g3?vW-;3gM
ztLzs3t{h{{`-T0oG+pPU8eg|w<nsG+i+xeZn}>GpJ-&41pgV8nis!3cq^_519F46<
z^}Wel{B7s+xu;Y0k66!5+kAcPvxL%1tpidI7yGQKN>w{o`9HI0`Av8G?we9q+Rm?a
z^~>>MkN39u$GU&6s%KMwuK7+VJ+AXW-SnaM_i~v}$abdI?p5_F+bz3IpNG`^$o6@k
z>T|}IU8d(5h`-C!<3F5Mrq03i`Gfdj2)O<(r0(yb`HJU@QmE&AuKBpO8(--~dvHC6
zQKe61eJaWc2?8brROcDC^Es71ZhyIquKH6wZ&KOy;A=<jc<e)O7q-%rm&y)Te|epi
zi`$f6Zt+#?WLJODZ*$`z>~(GLw(EXXI@5g#5ue(hNHsjy`LVEDHIMY|FY?b#Qun}2
z!+mo$)o<I+Lf>D^Ex%OmV^em7`h6+n)A6Kv$9-GM9H{z9J>2&EHI;Pjubcl!w^2(q
z$CA|Z&~|)IHJ$7H)i%9q{;Klly5F6R9<eael&Q9_Sa#*IZy?)gyU&JnTMpf8+*w6x
z98~)w{~U1mgAcF$R>>{JzJ#fi+o7pjRZjTx+(VZ+y55_j+DEK@r1`k}Q8d%`8ah7L
z{fk)jBJ%wacuyX-^U74x$^F%rTQy(S^*60P)BP;Ozj*0)_xiZt;t*>-&7H3}cSuc|
z_6ObgT=g&O2{qnP^99#%U5>uuQTsl%7Ve9)tzOsm<vV?D=OL86T=Qew`i`BBtNcPw
z?DBBuPt6D9=M=Y_yF5al@8@Z1^Bu|_-E~Cj@qDeneCy@fH;AQ&dd_f#i~j3Na^34T
zy`RDrpUZw-J-*J_bn_SSeA!{U9`MyZZT0$|7yH)l`+4nKzim58B^@;nOf{UVUTyoG
z?{s3no=b}FE~xH5>7GY%*(39#<~OeMG*i2MwR_unwrP7eEw3(LTl+;iU9Qx+CRVtv
z{nZjr?x)=Dr`!H@rK^Y5HCKyNjjy@*bFJr~r)&3-PkvVFa-g1n2oLie4BdH<X*}KM
zYwUR3<GZXcm35D>gL_|F)+d)gwp+}@a+TFGui`2P)t*epsnp_&^Eb%5NA5m(co~|5
z={nLiKGmM7@L2O~+xb~6dwu6?%OjO^q5D}X`KkJsIFCt9s_eD(Ppy2W_L!cRnd&z^
z&lL0QT%|mpr`)DIsrb|S&)NLNyO(KJ$%}t>&aj8uwP{w#ek=av@890ux;eCz-^&*6
zH7~vDibHOmUkWL|x4gxb_1k>XsztLbRr`Zq?X&B@pFN?zSx8m9OFFdu^w=F1m9k#h
zE#e8O(kJ}Ixr!|1a;e;MI?AomqwDYcx$j%Q?f1G<OBd@q%C(xW?R>yBov(cs-F;>5
zc&W@&Rd&QW7c|Y+^gK7U_^#{x*y(-WYx1q1`+4R%&iPIs`&=Z}z9`o|F_#|GaW3`v
z)OwD)U8|m3@Q<^<ZOUCW4)`ALRr<Nxr_e|3OXL!6yYD18q$Xp}n~?wF?-Z%}qp3ev
z`x@W<(HG`zKDUT1xymE!QH{IR?r++j<oL1X&0ObrayzhgOTQCudQOr`e7SyY-4sV(
z^L8rV+`>)gk5qiBKB@GnKc8cbN3rsC9XIyrdhn4wKB=qN%Izx$+_L6DEi1Ni#eMeq
z`IYOoskqAZ_t<ybyU(<)q$at?8{73ps_D4)jbiEX^&AVOH}&zHvX6S6r5^6<{I1sZ
z#u^W}@}=6nDxCXyqta)(E>`K$^&jx!kY~^7)2fn-Ol=-wYR{dg-FxopCoC#*m8SOy
zb>r){S5-XM=ekO-uYEaHexdzcxl!qL%R`;#r5di=zC}FS^OscD8M<~+`ya01sob-v
z_4y&nztL)I-aDeO<-j%1Rq2t(V_*Agaz3Hmq3_Le<tOf^=Ps%H@Ve!r^l*pU-p^C5
z8%6(8J>OL6WPeq9%AY#VGhJ``itjo;P#jVTm+Lh*cdGMn$}Vc1!X3^vF4^`Aibu7C
z+w$Q~C&qcm#lO47%@_JWK!<?u=c&pDuJK;zk>CFcHSe-rx2Bqo`?(%UpQ@b+|5SO~
z+U3esaB<Z`r5E{7vRdX>w&jrd_#T(5O~1^~)NR`jP18|!t<m?p)BB&?syVi#?946P
zw!Ql*CpB)y3b!4<Qb`w!o1Cw&cEhDNmHM)^Lv>!`%b$xorAMr1a+S8{0kxWrY7f4`
z#r+Cy>8t+<y{_*Wkm-E^m-PC&C*@nO?R}h7(vkI_$@_C}9{<q7hi>?`FqNs_L+7#&
z=L(C@6+&~6_hEgzrD65IPOf$oTaWEMz5jyn{7<ZV0#v=I_GN7M0Yo~_II~r6`+QHW
zz+B^H=yHkqv8}XS2Lzi>g$9~geyZ^<lz(QpU@izOk$~-fWUA?O-#2lMw?vo*p1$sD
zsPy`pxB4D$<#PsZH#M(D_;HWCcm9i|%t3DIw(B<CbYeZCD`op!-E#SJb^96zxb<?a
zOJqHAJJ4-^rt#!_Lc8U1o4P~WE9aN${6g0*+jVIw>2e+4gx$X9CBF2V_S013``!ol
zt)J`nROE5Z)XhEL4Xw}gye_r)p}xan+Ac(VG2Z!>BEQh>N0pzjM@ZHC!s7S7R8pi9
zdTb^4{I^#0+47dtr7~}V9hY6+_~$8w%t3eF94nr$`CMqdGIuVw+~0C_tM*5A{kh5=
zdwdt=P^}+yd2q)QcDYK^{Tg5KZSOTx_k*e5<D~d;t#5SoQ1-Zn%fI)Ui(CADZm!bz
zI^4H@>iaj`;jZmmmEL#zp~f4oaMN<A)^(J>Sog(}w&PQ1dsXdC<<C8@5&4+5FY5W}
zD_q!XD|OE|P2-u`!#%G{MW3&Eo~~ZkdncjSlgMxJ_c~K)zpi$0kI$j?<=P%~$A?to
z$>$65^Tc(Y?0cS{s(zUx<qqB7RQXYM`U*Fl_xOshTR!gmb@Nr7o7wv7=4%_zw13*>
zlX^ONoYi%Ua>?J(xvp*d`yYB|L8&WE*IBv5H*H_O;){ApP3pF@RO4}vv(={0w!Nv&
zheG-5+I{iQ&KdS_yEe_aO0Io9U*i_^cI`g$$<In~$b&<w&#7v$OLttB<K?=4qq3Xp
zevL|xD&K1Hw_T5Nr4#Klm-(Kr=a1lEE2(kKRxg##7Y|px6c5{QiiaF-`kfNj^C?yO
zVBgL7&-cts%gq(v_FUgK-Q687Jb3qCwru8>ayxQ$$9`U^+GVc(q5D&+`I*iGd_U(y
z?Tg6y==RT0_g-AjsY9oidFr}#xyOp9T36)a@9Vk2RUfII*Q@OE#WNSZqTadYdxM@I
zcHYaKTU2tD15bKjr{n4u7P-n?#xa$hzUL#Ny+baUK&|o*J-_14&vySLb~@kfP1e8U
zR!{r7?tiJ$%Q?rc4=R7zUYTz!w``ZLTX$WPOFZszCA2>7^|)?3FpcNh9#r|IK5y3j
z{);?+4&9DwJwMm>Nu^h{U&^1-i*Ucc+d)0wQxA_dzmW5#+BtW)oPTa^)&85xpPQdb
zU+&|L>w8AF=~X;^`J0|=g^vHP-dF7NXopFKZ%OgheXk)@Jj!0v@X+<jm0zys>cVcW
z^8reaX?sY0z9QR~+jCoGuPB$OFV%aI!e7;obiZ$>Ixon@U-Y9~rKo3JsrwzQRO5;E
zotqT>Emx`9C%5(2ZLg|$w)x8GQfcS5-``6uU8wo1uY6MDX={(^_sLR=&wbtyN}p|g
ziFB%Yi0*vMSG-*2bCjK-?|-T4)6}0@pQRqI+E2{IKi7GiusiqtR8bCHshT(X@*m&*
z(HG`zKDVegX*y2n#&@j;)xL!<e^+~{^zym{?j^s!pWFQb-SH}RJkdT)*HyOXRHo^K
zeX^wHdtBjizT9qK<Cd(K+ip{Rx$WD@_J;1KrhG%iH|@Wv#OJO*F5cXB34L7lQF^$>
z*;sl)?+fVWC%Eg*t8L?@zJEa3C!TYHmyoIKF9>@=$yoh}D_{A!n~Pg@E}x5kXwJU-
z1K0T;Wv8!wci(zb86REk5aWMnNwo{paQVB)u5Q)$dQ9!Iji1|ojcjl3`{qZN`)s(!
z{S(VvrCcAmxo3~x;r73+`9&E^+V<O2({X)&n9@U)-*%m4n~t(ag!`T^nChq2Z6f^G
ziCg!-ZT{CqwMbJ=vE$pe%T&^Zp7(L*mkP&fw?mARp`;k!VoTTYg}c4m>XYj&^nQi!
zeH^iGUu}|WA10KZ(ER0g9op^der&9J#a!o&q8`Nk++1boagsYf-{UUyZ#i_Yac31W
z2hsk0$x!1tcRsr9JNEC+g?^49zq?27K6!Xaagh62?ryGg-&lH5{ce16c4<BcY}a?8
z)9K#Bk>hD%FpkD`-ejCHQvzS>2<Ywf(5No6OI;4Ru2W$5@|*7V-8W@y4x#6*w!;tC
zco|wxtbLJG^QGp?p~GGG%T3eUwtp&JsQGm0c^j1<#f1tF-A_#4yQ1P#<<kw%eO=0J
zzu>03|0Cjs9>09$$Ia7JpWyFH`r;N!Z?(KP%)MW0s?S&ZqV&r947s^O%{RF7F`cg=
z{>4kbyVu7B7YE<tBh~M63%7k=B6K;q`)O+WZ0jL(I%>V78!qOvx-zx;vbAT%kxyK(
z+iP<QZfV*dsd_TiXByx4{L<7OIX<;6R)x!Ye7U*qB?&#g>`Yy6*Ee>*^O5sDEUR6n
zI{y#-{Nm2fbl)Mj_}sj>^-<5UT*7^~3)gX2JSTFMT<bztj<M5oaqz82cU_QbJm38}
zRsE*zTels_dbyt?-1<Vzw{m&D$#%;1S#56a_7qxQ=zA>O`RR@;xy4gG?_B#&ZX76k
z<?#mY&Yk8qOelPDP_@h4{B7q|rsu-k@qO>7a_Oh~t!cRHxEDG-_r8BB`h1Nmq4b8D
zulvf!*Emb{kM;N1cig+rw60Kcz>7nkJ*Q8rimRmZ5#ifc4!C8_gIZQ7Iq3Oe=e^vy
zMMWh|`<HHerP={C->1^c;kxIQ)OTB`dm7aDNnG_)<)HXe=|woz?n8%D+_=Jhmy3Iy
z=UbmC57B-o$yHye>Y>VS8ZOrZ_5PY_9_i~jDfG&Aa=TUKg1_GT^07H_w+C5YZr?SP
z?Ul>J<re)jS4qul<Zv!-LXYcnM(EL%q2C+j%1?DKEEj(+4zcub?NgcR(QVhJ@#J-!
ztJ`*e(>0x|-M-UP>sqdG*KstL^z!p47q{F$sda|0aM=#ZoeN(sJ5@Yl&F`uGI__{T
zj<NN~^MBc`s^`@Fx$77DdhC2<`<Dw-4zb&Ft>mxje_Zn<S&u3|mA~tJwc69Gxc$ig
zu%{|l9W}eK{(j((6vCqx2~g*Pw&60%H;;d4;X^lkTX0KT`&8+4#}~PLTyE3zmss(A
zulu2Yqt({DcSPZf16O;s)#IyNrh3IalJVUiePQ0_bBkPMsQo9~`liz9j=OR^N^j_J
z({+&?-_%}tU7nje*Y_4>yG_p}RPjygpF6&<eN*mx)v4(7#S40W_&sc{`u<UWaa>2M
z))B(r_xe27dlRtxxJTYQ|HYEx@T2z1<dNGvAy<2CtIKp9BjVfUmr6Q$eqSqYQGdRr
zuX>i(F|O{=`&!n|X~+2kcfP*ck8k~<`g^SV2)@fF_itbCVYu>>?c{QY{;t2T{Cw3L
z^xizQbMNt`iUZfWn`^xqT2H9^!Cd{l+Vr^2n_{Ootq-wJqATw_?cQ@&KVeZ3O3LL?
zjT^c6yFRyLr{@}<x%5!|E%Z9ZSAMqhR=NL}x_v*7Ozq){Z_6`OI@fuJd_HFCR-KQ@
z{#^TlruK2gcl{m6Ev~HJ=95+}nr-FX9WFe0_g}VbHkG#dsnWS}h?Sn}ITA~cYd^P5
z&oz#k>f!3QvYt@mvhRF!=iyX5?tM*DeXjTYP1_aXgMbME>fBd1+$2xL1pyEM0ZsyI
z^!@Ji{wKF;_ANPWt_5AO))U<M%JXbfH&xfV&)L}V<nLhW-uL&lA84zWS~u8+=d%CK
zWvA;t{J@hQ*y*_Xg$0*1?U&r~seT?R+_wG3PRG?QW9bQ1zrOE3QTpZmZ<X72eCAFU
zs$NX<;U0ft>5J9hxbmgid+On?`;4ijcb!+5rZ?@ks`$F|;oRe?%5UwT*!^YGeSWU`
z)aA#OpQ#@1=dv$-x*X+rsf>fNUAAu5dkL}T6RG6OHNOb0Cl!8Nc8GSY+YiL|G;)=x
zjEkyuovy#>Jj^vdbq})E!hPS<@U1`h`bvG>NZIFW915k^SNqg`4&@%twEvlwSGRsl
z^~&D|4dqT{{lIO9>vP$6dfj%!74MP`Z9hGBhec)TQgzQcH~;q6&3~lZsHK`sWp3q6
z&E8n|*-gj$RO9=8Zzz<0?)x;c_1V5JkV?8x_3oOF>-Zt|eR7wk^XOFK$G&gImA_~|
z@^d2gI3)6y?J{+{u5V29ljEl{F2=S)=8~E_HlJF}-}O0|N_wvE3{m|e)c&BH4`r8a
zIM==d^lUkFuW@G;rW}0lKl#?rH4lOwz3<(sI8o;%sfLUBZzxHPlcDcJgw9WtU-dnU
zRQCCD`Q2%ioxbNqq8@XRuJdcx_djyIZzJqZW#7fr4%7LNZhX~q(3ihyxl@au>hoV^
zmu@>qJs#IRLoPkG<7lqG`=zp*`yN6n`dr69c|H+pUT&KYSHGb2$Xu$`ts0Mg`^Vy$
z`tyUbFZQ_OT5qxK=gvR1c^_qu>+>R&^i)4dH9Qr5wsuhM3*kL>zOwzxh3hye(%aTk
z>~yB@bEX#G^&Ug5rWfrtwoI)(g|;U)f8<{q1VXjzAJGi`orWJ7uw)P2-(lIO>%m9%
z_@u6)D^>eiDu4335lc3|>V3Js?-onXo5w%2@Sz*NEka9qKPXoDedSyA4gWs{jJ+<1
zoqz0l=6+tv{W7%M*FKBxcO7)&xxPQ5o8I=kO*fr<Pt?^d`h%;U*y*|Zm2Z9Go`>MC
zlD_YAs`9NBf7^3NIUV<N&DXe^n%>ys23P*6^cU6h-@ARG*W*n2g}#sLD?fQW%FWF^
zeumbU8*kZO-FbAX@wnaxqx5i(qoLbrF8TG_`ah(bdNp-#+@N8(hUFXTFKs%gsiDR9
ze;YeCuF$x0LtXEt#!lb6H)^PFsNZ+~@qL;a`(M86=%Qual>eKw&6RJzTl$vgkDmU>
zS?9b~ay5+WZa?#{Cp<l*3?(mn<*nt$Pxzp$HrcR8_ieVEHL3J1x$50kPoLRaZhrCQ
z6HAq})tgGXYu0@5)onVzUFyo%c25}9>8wMy99eQn)AICu{;tlgPaad&iZtbe_`C#u
zar7Er9e?vHg{^e8=BVxWnRd(#Bg<G)*#B;;bH?pB`d_7~JY~hN4x9C-m&#n_jorun
zq4%Zll(D6#ht%bt&R=H1nq9}1vV3>@b$4I>%m<2Uk(acaa`ab6JX}_rbZuYHZNBh;
z9b4a3m`c}tRq1WVZLV~}u2knM!ago}@YVxQp8NY@CAZ|V4|;x@z>N0)9J2i0_m!0V
zX)XDL{$cKbSI$4~-O^OXD&c#F?Dy<i{f3vOlH1--#@_ekpdRm)u_SkSQqebL`L)&>
zy7AD`R;E%8VaK?YAG@*ZF(b-crD;FQCBCR1S#m$ORr=(7P2Cq9{=_K*R(YeOq^VvJ
z-}HIF6<_3UE4j`CZ1X?6?%k)ly*IARUFLpXFz5YWow(^8W6RuS?(OKzqUYEh+m0(k
zN!xx)r911;b>{5){wt-cr1C*H2&72h@>{30U3B5#(pI{b>D`ZC-1(Z9rk1rXRUG&J
zblHFRy65d87pa;Tr|zHXJ|>r47hZhZ8CU#iLg`DY+OMwvcWa-t)cNfvl#q5DbUFKu
z*J0GAzkB7%$)zvp+kTb)K3`rmVdjxdC8QmP*yF#dUVZuR^wBvRx9$065ldcwZE=0S
zEuScJlOzA}$@tc<4J&2otDZvX-D2{zhi~2fwL&FTc3jwE%@Yp&>f0i9`S*z{T>ar^
z9~ZgG?j8Ryu}|0L(p9SJQTCsB{dtF;wrFITi+ui;xh+SJeW5g!a=Bx>f3wt#3wG?c
zY9&{>)yz*PKG12kimQxW54q3tW1SDX=F4@C5?hZvKgiYXTA!iQ=YH-Y>~Hz+ZC`rq
zp7%>v8G9T>{$UAl?Z<`IbMbC154mfX_siVnXE%2oI)3}nW$yBn1-lfV-Zr|lmA+ej
zpYi8!GWF3nOH-+`YyQzoZ*<Jg<I7xSF7wsBFIm`amGvf-)h62<^2gh*nK`laB~|S<
z75~@P9sE(7=O>oFWGdxM#SZHJUFzYpI*++>g^^>*+-0oy16}X+9{co)|GsgXQDyFO
z#u2>+eA9YNnX8=J|Bk))>N&2=RhstSt^f6#Yo2QTdLiU*dcQY*nJ@oRa?AP4ZN1lD
zE_$%678&cg6Z>2=RQ|E}y5{@$iOb))|Ma)Z(9+dz)AXk6yj<g(a#6)UX1h`C>#qL1
zFqJBOuKov{e$X`&`VK9*q{<%ZeGpYT%Ky!&o4>N|6_d)`<=?+~`JT@QO)0BYj@a<1
zqkGMnQmUk=XWMn^eoMW!!f%=<m!>lJ`@^Q?MSQz<k}5aRZy*;Fpq?|ihIfA8$jhEw
z{n<h#Mg3FXA@jA5uKFGw<^PvKzr4QZo#V^UGSvEo%16|jkgoSGsrGGqerWq#)=lSo
zd`;y!PnA=&16#S?)XtS%ziv?pB`==7#u=xb(W0U&ZOg@-&bA+LrQ<G7EPd2?;~LJj
zUvT5uS0B6XQ6HAR<j~{hU3A!nQ%Y_bt6p{URgGud975~6X1k-`96oz;3AvbqZhP<*
z&)4^{P3={+_t5^_^D3FcR?mEQZ}ZsUMQpk21z*?iIcZdpo7`^Nj8-R~GpXQ`L$02D
z*>;yqD{_@``wHcDoqxni&o%F{<*YltAfDg*RlfMzwhN>SLx9>hG7S%7_5X6A_93~#
zhYi2vyHDp1EnVgREMNQ|HSYWR9)_)6+5XV(i*7vN+JEfzdYP;2vgc*5e=+ceB3HTI
zW*hGD*;CIIxysH*Tyf5V?k^U(%UIuS;L3M{{!g#G@7pgIxynt~{_F6Id%RlM%FyGc
zD!;?~p1Obi$e&fJO>VsEshf{Kc-e}Os`7LDQ_pjR+kN-FR=9Dk$DzmW{XpN`xW^k^
zeYU?>sH@*sywvt%xyrd$ee(<F{bh8aE4lK`P0ua$>#VeBR$V2wOr4wUdfrz$-@JbP
zi+f)=Z0uV_F7mszPZ@o0$H$7?<;D-stY4?w3q>sH+OAf;d4pRfH4Q4NMVhvMSAMCb
z$M;U{zIV#S?ZktwIrqIgUo2f^tac^eJK(nC+)kH{IP!s4%iN{-E>*QiU-O<&dVSrm
z52cs8pFv+70;cc($FWkKEBAeURsW~%bKCB3$|d{t10NkR`@#>4Q1TCLKYHZg8SfXl
zN%^@iyJfDv-MN4FPT1f2AB)Of>r5;|Nl`CU|MDF!?120sf!?hew;I&(gEF@Kp;hKC
z*YB>2^yWGLbPC_?8v5_*-1_7(Wwmnn>DJ{x^ly(}x9qLIom5it$}{FqAGrSHQYDu<
z_NLp1ocTd1OWogpq`tH4`h88?=OOhyN!Ry5mcRb*znpmZlu}pfzNh5t_myP5+%1&r
zzNqW(W^k9wR-bPFNi|;R@9Md}S06gPZM)!3ryAEo`HS*U??K4xdoeHYHNUleZlsbf
zwcn8z^_;pieSX9$&yeNUT5IUWL(ANx>At0Ie7PQuzH;5YKW~~=#+FA8e&K}SuYOd9
zk{@jH<(LmPomO(m+n;-`uGj0+N|hXU!j<!1{{2TKB@b_R_>}|aeppiSpw~9}^0=#~
zltSu$7eJmz>!#C<cimIJxU|i+Q_Ij&^*ue|@B7}QZ~da*is!j3<>$VxTj;fwzUCFO
zUQ_#0o$rUTE46uzY|pP>y5Z0ByG<-r(zL!^&&lNc-dyLm?eBbaOd(7897g4SqR(^v
ze!t?Rf|4ukvd6+%2Y*sf^0;qyoc{A8W)zfsvC{ADe{3_mP{{`$_}%ob2Y+3-%AJSz
ze)OH@ZwoH@o8ET}xbD%qimu${+1}58x=V|SFS%Cdng7>m%a@8;mvX;QxjQU3_<%qC
zVvR~Dsj?%KzwhV0Z~eCC`BXYv{<l7U@2`(}ylo{HxzBfx{a@DyT2*YNEhkkvYQ9J9
z8_D66J`t|6Czk)%o`-&Z!vED(V#!$L&NW}teWXzFO}{IWN_^Abbrj>d-2YA8zS^Jf
z`<}k+7J3C=$UNSw?lb;^mCkwm!@`$*?GLjy{oB;hg>Na^Un)|}7a(st<BC5`D0Mlw
z)?Z`ajc<MKy+LKQN>@I*>0N(^kjf|Z-^0|kkL&l7V(AI}oHm^|Qu_j?`l$F+yH$l(
z%kLWcIya`;H&xH7aMRz}8~Kk<#<zZLSm|4uewWa+9CCbByG+erKF3qJsriCDZpAL=
zWv{%o-1rF}l(|ak_j+{0P3wXCyNhFn4;;Dm*egmdDc%!^Els(a#-H3~*=~0&cXtVy
z1_#snljF<dqpn-l12+i7Brx#jo9}tY&d-;%A}_ynO4~&j4lYe)%sN_9-mdS#AboZH
zUMY8dnx5CB9)C%F$i(l^uU7lbikl5Db*0cP-&6a4?A>|TUDNqL{t$_T5TV9WR7<Iv
zsiEvcsUcPKSj15C*j7;$^H8d#1W~irSXxuXyrafMLrh6TLPmm$h(yf3*XL?p&*%EB
ze$MX<&pGFLcHI9w?`!S7*1hiazK1pJefBe_myJ8ca{JKEdfo5)YsGZjQEn^)mCh}y
zy`QevX<6&<(X?CXsq5Ar@SUkytMQsQ54iXrKc1d-7Hsc(pq;J#a$8+G_hWZY&x(qN
z-gnKAbys;W3v3M!_r7=T+2=p}QC3tO^3Q$F8!~)G)@qbT>&2;i9s9Rg-CoICg?k-0
zxyRN+U(7oTK78dzT}O?6ByTp>YoAuNYt}xblK!=Rm$oSN$8>pfzvr&GHZKY``+U8U
zAGCKrQ*QlQsoz%JZ`S|$Zk60hzk@B>{jcplYw=INGbt|`R{Q(kdh1pDJhjpF7GH9&
z4;P&^DX$e99lxzw2UXioqsLcHUZcMA-C5+E@%5#)9WrT1-fS$#XP=vgF4%X{psZ1N
z%H=!!dgW~<W<|$wi+{e&4^Ew)6%}t;<)YQP51Nq`70dk|xhU7C>~`dy-+63mR&?BB
ztsh=IV3)VEqG9)ww!PtqZ6{}q#&UTyz25)P4K6$8hB0|ZW2#rJ<x1ZVX!iXZ4m$er
zvmbpos})=Iz3^uB^Z2#D-|pzeMrB3C6<*u<`)hsqSoT%N#a9{d%&Di|pS2n%Y;^C#
zC;jd3Szu#$s5Ngks(!iiqM}@H?|E6H+c)|>nvmDN_pI9c9j2nSZ&Y4L<s7j5V)wm!
z(!!;V#=#4`dg`WE&R41yH~HmS=N@}xw^B!8%41rt^?i|!s^3}8vzo0(z5J$jZB)71
z-??^ly+)0<RdVXJbG7XbSZ_qHE`NA8D~FZdr>XWmu*sM9IB9{I?`Gxj$Y0NM^Y$OT
zmvvTbmAxC)&&1!o_0tD?f0i{Hhy3b|hrV~w=UF+uJNvq@oZhad-&3u%+)=;dq<@MX
zoqrqEf3@FhO2^-<@~ey9efdTQyqE_zhKE-FKiX#bMQ4|L-<HcCd*b-<S+nq`=PW&Q
znH|Sw&BnD?Ja5GvZyT2t6-U4Mlb?O}{;}Ct4a>#H1<T(4<T}d?%Zq}Qz6aCU?;mpF
z38&vP_sG1c_}1uK-+%Pbk$Gpua(-U!daIt3Oh593$IrO!oU9za{QVU+?{nO3S&H+m
zIM<GgZgp=~H2nK_4;%UHzpl?(jpgiI>-F14&V6$Kk;Aj1;IA%uc(Ly+^hDk)Jb$09
zrYyPLzw%mnthau@i{9Pv;@{6zDjM$azz0{YvC#abR%1Dy%l&_cO3(SP`S_;cPd)pm
z9A0_x=cU2-Zl9OKN`9Aq|Ek(@wfoPsUQvG+UDS0!s&}=&Yws-eF5myv{mwe)xx86e
zE<WRW$1<?{lY<uPdghp{Ufk@I59iu){D`chuvzRabKBm(y?5ycvu5M7f0*#w_0D`V
zD;k#jze*i_oj+KwH5yOY^_w#{d+mw5qtQ46gMnHY_~ps-f41yKqw_{#tKPRx$JpNQ
z{vP$&rrFz9kIIXJ?R~!2W9Oe;eZjPGd9~R6z&~C0>~XK>jmC2SPq4FJua>u>)|-FP
zAt&AX$oRa`C~kkX>k6-pdVO40H2m>je{uQ+$32_Xi=~@h@A0iop2-@88-2d&7JasO
zIja@#UTV!HpWXDutZ2CIZw~nFt2e)x6&15b{<y-bi;T!xjjOG{?z`u0@_g1_Z1s0F
z8vXl~M*YsVQF2zi=Ke*u9rk!uEB4#sS4YnJ$$X`1@#nW)cFsKy_9=Z;aP!CVNpG#P
zc&W4FZ!VcSVDkYBl!}UzPTOjwPbV&1Dk}cry*sClxxHJdsQBU%EB#}GUoKQCD(>^j
zKhHS+=0!^#h3DOK=>k{nHa@Qv+q?hjsP>Ki9!s_IzF7W<?kD#en^%j?{y&nPO}>3m
zwOH+Y_Lb^Y`oBg+xj%Nq;wS97=1Y0eu(LexYp)(f@snojdEWWAEwjqUcjhTh+<p8y
z=MQ^1Z!|Xgc}BJJZe8rkVe55&Aa5_0d;hCw*He2px_st*6Zf65*ByDiSn2=!HT%1c
z<>Xg8-=+2T`TjO<9x(8wytCq56NgV4IC^YebWC|~R=Ft8UCQa%s`X^E_%A27sQ$b4
znV;{n*tf>#MaM=zf9WWBLnp2N#0tL~pLaGa=lAWs-mLG#H0n8hB{|jZ)9bC*UVhu#
z{=DvYO?-6KFS5Y@@X%g98{NL#|4Z&H*Bkv^q>h%C{=c4j%Z>JfR>^DhywobaTh&jc
z|5q~qJ$o&8_zN?$W?{3&*{pe*+|K^{%g}GokstMYKJP3%H2eEa&6Z#3c|)V_H!B~I
z+iaf7$!`=tt(H@6Jng++>+eXK^}C4?=j=LRyOAGdwIYA&EZAA{T)leKyDsf4?K*m$
zAN}9BOt&Ln$?N5zv&3<#N2C8QTW5(EdZgp6^gFzVR{wna4ZGc#)r*znxA*?mZ!hqd
zV{aaoH4AIiuU4K8-|>RqjJ)BWd9$(F^PYO=lW6yU?|;Nyt(eAXqsvA6-gG5>_kQM)
z!(MpujXW@T2tSB8kO%5rpSHSPqw{+zui5L1TIDx;y^xN-qsy)Kr|9qhyFc|Dp;1`L
zpIUux*y#S69iJip|K2a)rPc9U$)79TuXcag=z6v8&(*K>JiI9NciepH9>d-oofQ??
z=c|V0;;Yf|(kOZ5^sV=Lqvx+y%d7R=uU7xARsPEReq+YLS3Hr`i?!-kseY~cSCZH4
z`<2d{I_rJ=-V__HZz?~^bN0Dgqx~o3{d1EG9-EZU={(fC9%yg7X8qr)Mx7tbX>yvy
zccbMt`uTpV<u#g*kXPIP{cW_p+neW_-M-f6){U;;Xud<<GjH^GcwFAud1$m>&nfe5
zXSZ9c^=tJxx&9jMpX8BWbgVbt@X*<?((kr9`~6RQtg!WJzvxmbI-dFbzB5*O?t`qo
z*eL&Mb-pXtU!%s8%Gvejuif+ObKc6nYWT*Ff3o3-v7cr2Vzb^yt|a&KsXu#q(lhU8
z?Zwm&>MgJL=G1wPJaK$hEmj&|N8PXW{ibsDEAiN@dB0M>MZ4eVbBjv<H}0j2-x|~H
zsps=@SnKz*TK)gATlIhII{N?j=zrmC|Cg88@}jI>Z13+a5BcLNS3NZH(Y#)4)bnfq
zXR)(kqw~!vd;a67zb!T-pHuLTfd`&<#Ptv6Ma8QY?z!_b!|u(Cim$vd=%uws|1K{o
zZn53p9(?jU*JMS-m*+Y0)eT4eIWIbzkD7%ecmBzP%U$w9UM=o*+~gix4}CE&DtbS*
zR&4ZqQ%+v3>%ekys=eN-?`1ZspN{%of1~>C`j>SdyZh$X@}lDDKkT~mjhnufHyW$;
z*K+mheNR!la&nvX-O_UXwfcIW%P!hr+@2HjW?{AC>2~p#`LW*|mlq9deNIv7d!Oa*
zH|t!dRdO5s|0+{?jec*SUU_dU`J+W{x?)6LbZnJhw6~u|{~w2rmUqaBC!BuI+#~a5
zVVaL?&0p2lPy1<gd8yHZUs~gd$$4PM@G$C@m6o~nkCXFqxZsHI{q%|RC*+lI&-wN|
zf6zyh^H$-jhn@fF<Ne;qYsItXIb^Lzd%vEiX#H63{|{ID|CKBKf3I@?AD~jZ&T_xK
z`QgF^j~m_ZlDqTHj>|1GZTH33d@L&})_OkCS?afXU-4w=drOR*xn!xmSZ`dEd(Wez
zuQ%!(g`9f%X>_~iUpVuNw?7)52mXhLM)Mo;f`MunX!d%hoDQ4qbj-nT_g=2lC`|u9
zP_6$Lx7zyYcrM@n)&0&o=ec}N!Dss{x7_emhUZ1Y)GlecsL$WZ>HXH|Ti<{5(2;q)
zh@VRTU$MRS^-~G|eMRr#W4=Es|Eglj&pB^={A%C%dz1pJ<00j**7DT7j{V!LZm;Cc
z#!7nqb%FQpIrM8wmO2_M=~=Y<^**0y);P;Ox9aHYMVt3iecL-<HL88H{H9TI8}+_x
zN69JozJIf>H;VU4a$Y*{tea*oJtiwE{^X2ve{<K8qqCyp=j$&!-_jE&WX;A({?M%Z
z&H7zWv%eG8-rosSYTqorQ@Pv!V#f={{&9F-G`wKh+n-!#nPGWRar!c^oWJl(ujEC?
zdfz8p<&J4{eRG#bvU0fk$Cp2IL$4RIqGPl3P_yOFy5=umKlYb*W<|rlfA_GF&;INB
ztW}uiRps;d+4}jvZ~taibi8BW!Fvt7;rXmq#6z>O+<O$wzFz5^u-xx2JqIm^A6;|(
z)>|F(YF-YT&1X^N*SinwY~%gS;^(}$_iaz+?ZsA)tJ1o*(*3a`7C&LvHDAhW#c`LN
zwCm5md?{}>wwljo_1CQ5rJcXeR#TST?q7LFVYB!Oxxqjy87TU`Fx9_Nzay+9r@b~R
zw_a=3^=AFvB;~)+<$Cx1)wb*C=TAlLKlT5<TkOBin$ySU&B9vaBGs?G<@C8*QOkS$
z<J47F`|O3hvtzT~OGy2oS@D<l)9Ug?vmU?s-A5<p^<t~_P5W!~_qSRt@6!8LKBvcp
zlk%csx!)rd<@)f0K0a-al_usz!%Dw*E6V+%@t*3NzF%E$In8Ueme>Em(+7Qc+)G(e
zv9tf)u9AM|UNy1jm#03HHydm93-v2K7iiRd{gwN@ZmX};2OfC;B7eDjoe_D8m42^p
zUrbKDnC6ctS2zFG%XiwY()U%1cE8^5dUn=!?R~z~?DnT!zWs$4Z}oH@_<!?|##>R#
z?LD_?)VRz2&Z$+`D~%)lPDQKAmGr9i{;-J;zMY>sI_qex^nQK0_X;cd!BL-Wn!SDX
zsJt9j;wk+;R<p~M^l9(?YQH<w+3NLwbc4%|xnWFRR4nRuD(PMBy{KkgZ?B)1YyaqX
z7Jql?&-#=q;XV(YlV7&^BBfgK+jotB?X<-gDOHQ7?7Z{Wf4a{irCRaNXS_3foeSnJ
zRf{Vh^yX*BUo=mtR@~v#{)_&6tL~+a!o_=i>(8&B(5qA{{`{He|2%Mo-lbMyIe#m6
zz0vCg{XDhb^zWTFX;fA%-o4bCOFp~li&?9YzV+hQcAvHQr{9^BHyUfzr`qoe9+w?6
zdc&Tt=0(Ad{(so*{odde8-DP^O;&v)FDhRB>2BZL<C8b?qTyfqFaP=<E}fDG1`qZ4
z!QRh2a@Y$`zL6IduY2gBF2_FoW)9Q&q`lwY+~v0yK6lz)lk%crt=|Dn>!&<cDtDIO
zPgc^iS@+K#bMkj5-8D4tXsmX8r(C|nuUFn?V%8|Ew%<md=TzHoz4gu<we%X7e9^m9
zE!NvlXKz>U`IK?py4aP&*6aR24r_hxRq5QZsQ2srZdjw+u~TQmKM#0ep}7vbFE1*#
z>i4(J>ZjTNN3C9e+x^Kwi*-G7Ox96Yt=_9{`P--88#FL0Do*Wn?X!KhzBO+zUa;Q2
zLtpq{LLL}Ae3kv6X#2>b>051HPV4Qn)oCxTzuartoPy1MzOdC?gC9Am|ER1{xJBuw
z<M#Xe$gH#AEsI@!&Hg=yXVv1jR$FO{f8X~|)+pqk^?on2)$Q8*|FFun$3wl?ZLNzB
zSa-b{Svjm%Zl!kheutyk?V5EyRf&)Eezng_)m!U<FK!t;`h~2~SnGE-+FSpqMP}?V
z{g;pBt;S~iRjcHue$?J_x%b>!b-lg*6z%K#UQRjQI_fz_NB@7LX7yifd^YQM-)aBN
z8fRydn~uA)mCMCnQLZ;@{%(|=TD|^nZ-0kxG<M(ThL6m@s>trU&iML|7g_CvyuEn-
zuReWe(6=Y#_2T8f{=<PQ>@zv<C|vrAZ{OYb;y3c5Vn^}TsD7KxJAV)SzZn=hY4s;o
z_}%!tS@?hV+VNMtA5c_0aQG9CT-Wo_ylB|!=j-))+RNWxVe>x6-Ig^PQ+?`v-=$o;
zM(+<&ee0d)8{IC|tJ?C+*ET-<d)JT7=M=2<yKtSYextu5T&*9}TJNvNU;m@Y_by!8
z=zv|%x___Qv0Q)4ZFT9~kKH{zs~3NG-BUlf?urkxX5kZaee&18%{?QlglYf7ZW!?A
zP4<|O7X|D6zENjs=l7h=!gBsw>H3e)y8dsM-qod4R6OURCtld>su@{P@wRPy{Bzuw
zT}nm8!!AAbo-dx5r_|9{$<G^o|IGO&?mJ_zJMyC9l^1_r8hr2ed9~PR{>sUFVcg%(
zJo)e4OGU-C_quB6s*m(8wHnLu*{ti0{ylA_@1WIszfs?nt0boqzeE1H&v`?J&&WCp
zR?@fD`^{R<G+S;_zEfIG&sP0TyjFi0)8)<mp1bDSyeK%K#{#?DxX?{`(ebK<d+z+q
zuzPd%nsx6k+rIIm@mZ}{?f0l^{r_j3t^NkzUhKzHuN#sT71_C3Z1lNftK~KNd@|%!
z_gr}EGe6&Dv2Tsf8->;SYq@$mtv%?-kNQ2Iw-?*{o<g+$+Y=`r^m2}eX1!0`D!Jv}
zJCEx(F4OJESMuPYxjdBH?^OGLqMB8&+;hyLUH{Qz1CRf7mvMQkuu<nD=iPJZ0$1%e
zJ})X3{X6qT|DE#LAFR3h>F18ii;A`GGs>+Sn{~b1|D`Va^-BL=vflgcwZj+7AJP5f
zUSsoCVSDwh)V@~#{;Jlm^&G3C>;K@56IbbZ%+Rc}Bmb=RJ&9KRzdHJ*m_72x6;@qj
zMAmFf`)zdjyz_5cW|fcc%$tpk*1J}Dt@>SPv-)ZF_m5Kfwes7l`h$+{dC{Xkcr`B?
z)|%fc)o&EvMgRY<AIzL~)H5fH%ZiGP`ff?1<?yd&|G%d@1|E3c5!XMM7ZnFBwfdIV
zZ}M>4b5Fc=mgkN1X%%k$_5y!7_U2((U~743^!G?wEpO$0zcJ(BE1t+&h0TtKpWgEB
z>L>L2Bx^5L(zCt4&+Vx8&HDfKt(M#9-=|b7uaaNXdf#)@W@D{-)m#6CZomKbQX9?6
zii$TJboAqAKl*M~RQ%+7UEUt@`LwL4`0yjoEVR$S_p_qn`)@At*w>Djl64f;8@HXc
zU8~+Ftu+2tpTAeq$NkBd_Bd&QneS$u1rM6F;X6z1H8rb*+bs9uk-u2zy=+dwO25PG
zZ1>OGX_Y(oK73|YRJ`?_4|^?t*37J^*xvm~w6AWTCr`Et$KHJRW<NUTwXCDDz2~xB
zJ~?B`dUsCDIvf+jmHWTaO}pOY@3o$=>o;d^_SzHqoQAJ1diUiU9q?ja6m0c*Oybrp
zFD|#twA~k9^RcYaxcDjqo;mf@`?FT#khL!ydc(%IWbH+UX!Ux((fu|0|E#rIUQxbJ
zQc2G~mtC~MxIHK4MM3`GEIj-D)5cEQdq7?^toD0u?Ok`ZxBbIce$;i;=tuH79b5H%
z_D1#7-ruc7`^NnK50B0I-gv9zp7HgiwjDBQNZxF$b{<$b+yCVyw!A3o>{w}^Rqnf^
z)$&klz283ZuZ6~peLZh4*4l5Q>OVSQ$cQVi?_Mf8zWKL)?_IKck5XZAv)j8ZcGn`M
zHw%jAkAG>i6|Y;MRCL^G+KAiEyrx&FXn6S6J>J~^oCQj)!g}-1xmQh^KH&5>vZCS8
zZ|>IXhPz(LijK?Nw)b!EUHZYSz1Xbx_3M>uUDztD*Z(WEJMjKR{&M*`Bl52rR^p}7
z{c695*r<BV{(gC-@s_(^^xv;GT7NtpzT*YI8F|A$^J;PWktaNU#%<?hMaM6eKWfx1
zz3<LC3i018oW9H}=P&%yD|u0|)$iRltG{Ob9;#Zo&GN5Ga_5>jeA2+tWAmb7tG=t!
z=zh9g{AGUZH^=2g#d6;bD$4a{`+KFhE86|R-+26@4Ne@9)rzei=hMBG{8yJLle0M$
zPq}=DU$4B)#H_Pp*9WIOxx*=wvi9N<_q})di1|Lws>L23%-VGOd1hus$LF`(^57kI
zoRT#fYmLWB=MlBatF?Z!<G5CR()#s&2PBo#ta9%mGv~cv!TCy!!e;5&+2vOI`@Evp
zJLJR@PQPdFk$GTadANN4SNA*Xoagd66<huOYOVdX>b;1f_tV}y-0a^GH(P$AzAxKe
z9y_{yQQjAA*8eANmE5oD{{yJizpIV=RBn5p6F0kktG>6=sJzx{IjKG8v~sn6-C67H
z^Zjk!JYe8Uc~P;^&ucr2ymH@3Zg2d|Y5$L*()m{0Z|r;Tt)4$RTYqmX`J+W{x?)7$
zQP^4XQzbnb{r~9OoA+AXe(||)eRY9vzMai!c;A-GAA92X@mW!^*5{wq*8lGD!xuc}
znBiHo@kcA)^28pa|CKcwE5(QJrB=e<EPl?5d*AkCUI{N6*XM@q-Wisc!<R08YfQJN
zp3f^`wdeJf-e;|p&(izV=C`xvIb^Lzd%vES!%FX^H0%A6O25Z$@Bi(q^nZuidu~$c
zchlA0ujKcwzF)0hwYOfg$JOj}mqyF4m5;+W8oTdv!$;<oa87>?*XZ{-Q#(}pJx00r
zshj=(A5%RWU7p(O+GqQ0eQO@ru{?ab(Arz|KkfFsR&2CCr1DN4_0F9qocL5;R7}Uy
zsB&jLXRTJRTF>Vi#Tz;8Juf4tT5NBB9o4>Cdv?}(wfb3StKU)Qg4OzUt@R!~Z@b>l
z^qrJ73Onlik*lr0?z`u0@_g1=u~pxd`{6sMojmHUN!gr^<<7?%6`#$@pJ{w{)bICF
zJt~zO{hhy7%d6Ed%JI?v(G4y;=7uqOtFc+<apmN;>i=anyPrzuapmmX*{(M$AGKQU
zUdK)DvGvdw^P=L!-N&zU{;-$xz~=H$F8(S#2P*emr_%Lu{gu1k>iMf^{gvY_y}s&}
zzkT|>K?AdTF_jnPuO<VpyfEmcwMYLhZ#1TCHtRg1-1CHbuUC40;eI;4MwK@^<-@tQ
z96uuK?6}^WQ|CGI#PM0talf8#+_}@#*Rrn~Hp*Y?mDB2dV7dM}>O8v9{kH0T@OtfC
zsa=nqe|Gf+)5hhsVkNn)zJKX;$DMK5{m<t0Vy*g~{==?2-?-^(c~P<1--T(F{4bV2
zqWj6c#^&wCTJ7KH`px32-1~Gzzg}y;iu(2C`S!lwsC^nQ*7NmZrFp#GbHY^5a^<G|
zAGFUl&ECFxR37p>mWOh7Z}0V@e>b5h`qyixdf$hvx4+KPuHHOc^mhCA-f!I58x78z
zjm@%id;gzF+P>U=y;;|rweD@T+?2<nm0PtRXx4djCArPEM;hOieveh}{qy(PYRZz^
z{VT5*Q#~71ezwnY%MD*;cwTg@_xz}%+f~{Rb@uzuzi{RkZ+|pAZ!Z>gJyP#;vs&9V
zD&FTbIkmoT*V*dVYL7GLo4D_cz3#}X#S0faZgjs(?#^3<>3AyrPPVh(Z`S&`RdOr+
z4!YI%Q~yZIoqfKmR=;MSW2U?*H#?s6(`u}j&sMc-@4QlJA5`i7TF*{j|7X9Ql~uxf
z-#hp0^B?{wYcJm3Z{jl7e(lSw)wsr#JMO=IweF>Q@w}V%dElT|dX(zL9Ul1Lsx=mx
zzf>#EJ9y)npC8_<)F^!D#%<Slv)@9cqTx~V9r;B6T^1}A6?gx^!+RcdTdz{9vDNuj
zf3<$cy4w2b_hyP(ZuR=9S^b^;!J4a|e(t!uUgR%dHJpFZAt&AX$oRay_*LVr*6}`k
z<wspdjeaC=FV@;`)DH%NfncCf46J$dHNAVTJ9nws*eHF!Dsq1Q((daou-W5z(Q(My
z7Y@B)<6E+Nv68(nSoZcO*I8y*-q|sIZhX`tGxnJN%g6FcSnd9$RrN;h{F4WlyX1wu
z(b%g0A2O%()9B~peYe<uoi(SA&pQe$#d)jlAD107dc&Tt=0(E=M||(6Pn<s?haF|d
zdi!maeVhH=jY{&X{SI$u`M-CqvP-S~wD<W&XK7#Qd(@r%eo_8^c{x4n%~#RxKkJ2B
z<F;J=nXhep`1h_KozLl5Zv6FLZ`JRHo7GRFzK2vTe||?cr(mP=P_;a!^(ws&(5m~X
zUiFrLdGh?9ExXa^yk6|==Ml};uT}54G`gRT`YvU&`W1K8#=(!CK4bOgFP)qh6*pdc
z`hvZ0e<N=+R?DAy?TxQj7rp!PjShG*ZxuHB|7MIGvG@tQuK7}4RBRRR&FZJ9ywkhU
z{@7}Hjs8C^t(N!J=v&`^^w5!cqfvgG)3Mg?F4W3Lwa<HMmD{TCwWs~m`u|>vQoqs9
zX&NoBQTy*|<y^4dzC&O5U_u_)8Xi0ksue3e$7uHb;Rk(u+8!%S%!`VRK9{LhUak4Q
zv)147sqemj&>tq|ft`zoK9^mz!MHsq=0(M#%scJPuhl+h?(FrN_4|Tm%kAvXZL8JK
z_b}6W;_u%*Y~-{5x<2b@O!ca@+}`I}wf2|RU;X3DpShve3)xp4Q(ih-xz_b_XQ^NG
z_Z^#^r&}do9&ImHii>B4^}qG&_r8&}8Y}5l@BQ{ZN9pMH^?qNgv$X3h?;+N!$AR}R
z@|Vlk8Iczi>wW*Av$bpW^Or{T*WT}l)!M#No~-nL{fctG-0zc(zw3ijp4{P-Nm+`&
z+I5B3M!h~R>uB8l$w7;CJ#$P}RLmav;|i-TG9qghvSU$jyCX*Kd+X-UW%Xiv|4-eo
zyS#kgE+b#f16#{Owa-y&t+)1{e|*7x-JZ@`g<BulccTTjdm}3<e!l*)^DRAbLRJ(!
z>+4@%uiqP!vZCXlS1*}0rT0r&XTeYRo%cHjY%(UV73=*@OlNJE{$6-aDYy6kT&3%Y
z`Fs3v<zq`dOPv+Fee2#qC$Icj)@)39nN!Oh^<A!x;<MKG$U0m7YTpZLR=xJxZH3o%
z{{C8DK9<cX*y#Tg+-!M2K5OKd`QPuK7ai{yc;I<QT>oHRG`w}OD~GMu{eirwSn2yJ
zasNNxp=ocP)V{O7Pf||5TK606t>5V9!{x?RwCk;Y-;>{VG&XDfQxv%!&42X&)nK4i
z|DP;nvsM3JLqBQ%<;wRwvD=`{w-}dI!oBb5wf)fd-^r@QiKF&gdAFZ@kX4J7{!c(T
z{pa-S-7fwzKlYpB@}go#J&!(oqp|xwH+*F7`v7y=cV*^uzc;@2uCt5Y9uFObXTN{i
z*lBwY$Qy;#{-3aB)w^h1pBuJ&XIS27?6LFDuD)Q}xV%x=-rs*z+rHWFzc%`P2651;
z{~y<^e%AZhokza-;?r4C@S)W|-+sexH)i-Rht>YBq~3bv`u(b0A9vYFyZ-#km-6;v
zqyNurt?$PawSGrEe@y#r@B6)-C0+(CwfdIVZ}M=SVzc#Y)^~TBE%)RHcYo-wm;5R#
zI<C6qZ=Zf|(7>#yShVMOslJ`%_a$FdJsSNTmvZtNokyGf|6GeI|B|(5Gd<wg0s9}a
zV&6V}7w$XPIRkba(0A_t`Flw3Ll!<{(Y{^I8gR(s|5N@>-?{qEwe6<|pFQA^YtG$l
zRMxf8b_#jHKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t%yA6-bjtcC
z&Hvf7JTQ2e6a8T79`Ehe_oVmpqT(EvaAH8gKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&
z3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ
z!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`
z1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9
zFc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t
z%;5}Na{qBNj=XV19vD1?9~37J{(Z)l&rM(Dku0z=Jj~%~zUtvUwfCW`EwShOd0_Am
zeo%WHR2^thF&GF2f`MQl7zhS}fnXpQ2nK?IIh=vF2MoSz@=EXIfz|Oahr2@zFBk|0
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nITkfti=C
zdf0}eN9KXSLudAb4h)g#Iv5BBf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl
z7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>
z27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?I
zU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}
zfnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl
z7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nGtuz`;{?`}LbkeV!E+PyBqlX*a(+D+>%BI+Gt1G#!N6
z!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DZk!z?%P9e%kz3zL|Aata>0(F&GF2
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>
z27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?I
zU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhR$
z#=v9qzdGdnKTXO4|NnVt7~vr(7zhS}fnXpQ2nK?IU?3PMFawvYJzMz!#}3&4kQMtb
z+;@?_bDc9_#{qrk{-3{x^gd+aLl*7Z<*WgREdD>`@ARFk?_Aq{dhpo;4!P#s%|>Ni
zyIk|vsoj<y_(cx4`tB-I*8R<A`7-Zp`N}_!__Ph5JnW1I$87g;zT{z_oU+hjGxN=F
z9rK4XzV%^S*>&iCiU+QL!L`p^FgP!V+b(q7Ekj=)pAX*Py2XFibx8ilZ`^e48ViqX
zD?c}U%MmNgKRI7<hh8&!e&@~n_gC$)<;f2Y&o{by!s*-0d@1if?7odI-Q>}%R{Y%|
z#}EJek6+HpA$i-Zd+w(5eLgN*c!MRzTsGwKHodZC*BJ9z-&gZhUf5&XhfaPsKVkbb
z?%R0kgtl=WeRQv(U;OpH{GMT(9>4zyPh=COY&7+fLq}%6`qkmv_n&`4_OF?%A9cWY
zCT0DP@7{I7eN(d=*1O`q>yLRWyJgUnKD#{iUiS8H4qRx`VbikPuk82YTi>0TZT-dL
z<5pYu%WUdrcO81^qaU_i?{ejc%a)lwE&KG|ej|SS`**S<e$Z><jW7M6O%C4iyW3^k
zp8M$~cV@30Jaxdtg`dy9cGM?39<=@)ZR1*Z%3XI2S^dQfsyA!hh5L_KZ*sQ5p<Q~N
zboMK4c+&I3`%K&Xt(=|UpD#S{sXIQKm`}O-@JAL~{gpPJesb@KKRvY5n{B+)hknZs
zf8<ZQJwL9^uI!8-`o3}Kxm#{Ecx<-cvqSG%?XDr&b^qA5-?SwsX5=Z;mz~*@Kcz@t
zXgvI`9Fhy!IqiSrx3@iTwXv_a>8n0{cU}65e{}8rLAK-fzcu5iTi<O%dKfRnON!*N
z-zC?~_;|u0W83_bU*Ma4`LX`>qdsIWe5c56DXukOtpj&@<-iQ4_m%ncg>OB+^8SB*
zwawqyk-p;Ln1}bc^u2$7+!hz~7e}Y>)&KG1mYI?DxccYYoPN#cZSk@79j|<Cxre%!
ze!AV~o89#9E~VA??%w5xyM2`1xy<dWUU~Yb+1VGozQa<V&Q<#EnEP(Lu*>u|e`8mO
zXa2#i?&C*ZQQtVr)yKd4<clMAB~L%>if3^O<!SYH_~X>M`>!`8KjP4zZ@&4MY58Th
z?6lLMx8KjzhuXQX{xNSqa@B`x&1%ad`n_vlzb*rMK9HFw=qK*@llts$+~S-+^E-UV
z2Pu-H9n>G>Pvwvt`oPqV_-79&Zd2q>^<uwotar#E4}F={i~PGBnup3^rGBejW8oct
z_+Za@migBE&uulUZN9?4JgI(qUD^1??tQ^Vz5n=GTU?7D{>abhV_p>>uFEUt<xyic
zov_!2A7=F7m*W2Hu|t<x`Pfg|@`&s5wSJ)M@)X1ayz{w7_x*j(uBE5mTWiXb7j`Xa
z|6e%U5aQH&MLdf`^WDFfy?^l`XT6>qx45*PvF=INL+YofJ)ZTKBK}jfeiCoJpWA8b
zC+`f)%nQalYxeqtowWP-yT^a)$<42Cn{WAp`QoOm=T+05n7fU~wO`w5kD0fBk^S_l
zDGT0v;Kyz2sc$?o;rRzw{363UyQjFt>gPXr<bhpEFul)y<q%&fnwL_fKNJt}kIN2R
zbkHVqm&}tZ&DwPGr&s(t<CiNOH{*9(Z}CQkKmLtZer<j{=IEn_j{D_Yr7J(*=!X42
z{Up;5|AFQQ=sJ{Fp?nNe|6qT4f?o8uUKM}VU)I_D#rjJ>{8_%HpLtduqNn@({L~%4
zarpIDyw|q=7gsOtaLMW;_IoFX+TjE0H%0j~^#gg0pR*VH$xHH^>+*_m;fXwX-aKi)
z!Vkz*pI(!e`Nn}i8}&l&I{V0T(0%y~pY(z3WnYvcyU|O&;0Gz<(K?Gfd4T`&Z#>}D
zxYPMgd-+0Kh)+BkH~(b^>rwoWLoTGZ`3tJAU%cvHe7g_Tv#xbrfBlAUbLAdCoVO%z
z=nslJ>*f^k22<YH1M<^KNWT=xN#`|X^WL$C_nLL{xu557=sUx|cEOEb<|$sW%5}FL
zy6H#xKi^t+|5ZMo-WI>sZ?219>t5$4+C%vPUitU!U)kyOxl8QFPRiy5e5Yu>=ePVr
z+(CW}%|qgb9m&yeipFhSAiufpyhhw9LwQnuvwoKs&C||-)F)ru$QMU$xXMmzY&<N7
z^0WM59uRl(E)>tuJi%X~yaMr*BK?d*Ue>?+@-UQt%OSr`5r37CovTIuAwP*n<ALs5
z&!zI^J$a2i`5QUfvzz{mi=U~FPyS6ly|vd5p2dgz<}Ll|C-o!u*~h%Yk0E`Wn`ozh
z<Hw8f@JIDhKGdTx<WI)KKcRN~n105ANBW2>>n&yau?NHxJ@l&|{X_Nej5mm{YLTAu
z!_9r}UhaUM-p=QJX2cHXPW~j{W#J#}wfgU;{#V|%UttgD9bf$M_)WGx=HLHyerf#j
zdYb>tFY1{W)N>sVX`HZ+`5J%h1Lbdi!tc%F+UbXV_)|IL&*e~@*NWm#KD7Tz?VI+?
zF3^2`&X0^6KX|cTwNF)_pW}P-UDqA`{Y9?H?IXn%{glZwpQJp~kDc)@9>{_0Kri~I
z^2Mok^i>}p+L^zMbJpyA@TQ-wyxP!nrf1II*ulJMzP0}ueq8^j7Cq(dOg;U1o{;iq
z9^@bNG4InGkIKe{2lw?ye|o5gXZeXe*g<|$KeYpXTo*^y;o`@5=&2w5%Uj}7zvQY9
z^$WE(U(g@&6MU#=ov1(hLjAeVpY`iLeq4vf$G`AHAM+@=@)^J2|0%L3KKZwP=>_$R
zPlyNkfuGUGe88XOVfyfUe(1jT_)qI=M=m|Y4|Jb>?ThK<d6@ajc-h0aszvtFpK%!v
zejz(3v#b934Y`$2yY`~>Ex!|2{94&K=}q6%uPW)yzPoPS<>C2unU>p^vCAh5tup!2
zCEsX!&TOAz{lY%_QHJ;>uM)~PDe@2eD3?R_g{fWSIVgV1q5Tm|>&XYuxFI`1`3#C{
zn2ujwf%g5*Eu24Ai|oa3tOv!5c~Ts?Z@!Ul_y>N;<tNsao_9GPus(Dip$ye`ZlE6j
zN&Sz%@qhY=6Z0p&jWgw2-0>Uh0P@oDiyw03bN=Ccl^<A-L3x&b<l_(0$9#~kFFe=f
zH`Xc6x8)7`$S341`p}lk?)K~VGkR<H>ksd{W%d21WO%ZUwO%C0`iFm7Z+UJ&K0WE@
ze8uyDYLTBo{6Kn%H-4_Z`{paWe|pI)eNNl6%V7JQv*ugl+iN^9E9bYXpLh7Y+gvrX
zt?asemi?1Fsw`gYljK|HX5z&DjXeHtp5^DzxcE7LG*9DO-eNy_)3~+g=lqj={)xZ8
z4&L>{>zD3XTDi*|L$-duTWOQ!mfzuw6S}uO=b3fK{@ee1=6t2W<DTk2=FdG#$4+~2
zg^P#I+g5f?>O9oGZ+ibOk8QQY=b3Xp_u0#J`LFFc)a>^Yp4#{u_q^Qmv$pph)|&l(
zh4@STpWJdNKJhAV;8&TSkz>6hPg(cik6od9?3yC|<rncRAGof6NH6wH$AKU9l-ZqK
z+?OAX!+mA#_=|Sr(o5c<&#c+|G<j4#*X1AcwSJ)a3J>h&ITt-W@8AdW5B(s$<Wu93
z-<<Cnhj@bG+B|mJxf7o0f57{>`I6nOyT!M9<~#YrdnWP<)X(ih9v?mJ=hNHPZ|=)0
z@{j)b8^7Rh@)MMYl;tP>Y#i3}>bWn!(nnkvFC>S)DbkOfq4fznsi!?X_^0veH?=pr
zdoIEb=3#Q=O?i?$@}c!=qmkbFrI#}O@R6dpk-yw$rxfKGs6X?CxDy}HyyG0qI>GZr
z{j;CvL7vZBk3j9Lv*}T8-C$hi3%udS`5C$LP9^l-gS=$?vu5Wpc}l*Nzsuzt?d1>k
z@gk3^2g$>;`s`%=B_CEo{Zv9cLV8z=+NH>!szu|3<_~!uAG2oLg&$hqy6=5r`zh;_
z7fyNSuLr!;wZxvDH!AZhe&jws=2tNFe}3Cz`?F8}<%-=(<arLwf7n?ZkYiqR-}sEn
z`p<QGu6^Mlzkg$$?j`4F<Z5Re<d5zBm!rEZF;9s+>oEP=cZ$mt?FW|H_>vQP4xiQb
z9<P0X`;R>|e9sl`p4s+ZH~IEq<jODRVSYzn?Uc2{FM0gLI@5l~b2Pl+i5%_3mvND=
zzC5NKxyCJS$zfM~c<-RiKWD!`w?+Ts&pGL}kJ{eL(l0suhTRq!x5;J;9oDVwy{`ZA
z&)M&H=~sTUe!t|}$wwV}NsrPu-`~7nzj=C<_z}N8<<w)}Tz$bF{~IT>*Wv1sr#`*B
zUorl|_n*G&sQF6t|2!Y|z=#R+w3WRN08=^NIA&J2+i&kyGA?n2Z}uf$|N0ew+KWH?
z3H_2IPe9|;uYO#I?9D&<iFoYw{EH*M-GBblueN^j;raitNa>P~-|I19_pg=yGT(>a
zUh{*cN{{ci=*T~O(!0dIo_B}`^~EE7<tzRyzq$^MTON@A<U`}Kzrzc;#wR}PQ}AM(
z@-bfMD_-&L`a!S%qx<G-%=KUS@$%)@e){YGUa(|8D&E*vJMRO+!}jgE?}um4+qQ3=
za?M_AU-i>@O8oKP%iVhZc6ZI!wx1Pm8(#C`tl`UdDe-6V!C%GsWhYJTd&gg9Wby)g
z<AvYQm*3G35BOvcelJedlfT*3`q=tP-j*l%kN(qh0rM#2XVyjLTl*h*lODz|{^WIj
z%Kr4Ck30rl=V#=Rlk&*k>fy(H1^E@frLXzKb^NfacIxv}aV$Q>lXZr5BKh`{_`|dJ
z{q%=V>p|D?@ld}LzqQJ?T}#eEjGO-Yrx*Vdk9by(J(TG!j?G){v%7iAzS=nSYaHx`
zKk*OAXLt7FKibiUd^{-gbN$hm9^%nDK;FZjdg2e#Lp|qJ?vv|2`QnFP7zcgX3lHqe
zzl_5<DnE3cpW=@_^e11Cr+qp<r1{Et?4#`q?em=@C`0EGP(5fr4atZ0&(QfmiuTp`
z<InWApJyL=lwZqZcywRhlux1hP(OI42ei*LZtckDr|R368>hIZm-rVS{FFT7Fy3kr
zPxgKMhdlmcy!3THMdPF=J>)t38=v;ZLl1Fo+}gPh$<eR8s*Gpnw)UOwLp+FIb``JU
z&-mGeUgE=b_D_-Dx}MGl<jJqf@~ZLb$9!rY=dbwjyaxa7Lw0hWLNC0DL-+ZovVJNd
zp5+}p8lV20Q;-kk7f28BB@UeXsFz}TuBUy9<mun|@Pj}1^$W>UPka6+-=(M>BrnaE
z@&)<ix{gO>cH}SYp}l_arGI7T*`9lOUMt@8hhOdW%f8MjttYj^k2qA1{&=FN=hE&&
z{ZCzaiJxCJ?~JzJ%ZpF*5;^pv2fwn8ric6LL-o~T2le@j@$(n{rJi=FJZ1Xhi5&7>
z7ti>G<h#%A`~dIrA9UU*4_haSTPR<eC**zWWAe-|%Fy~WMe_pGUy8<G3E3gFA3mXc
zNnh>oL?3pSkNF|HXy0h8G*0)~w-Vw5;vq%%&10U!7%voGkiSFErSQi;&HMbe9P-N)
z^<NJ4R}RaytM<BaK=BRD<K-|Nk23$VUMh$Dv>b~6a>#$nVI>~ew;YzUr|ZsJDq*?t
z@mK3A{wiPS$9~m54ZqgA`j_{_J#^jvQa<(mqvxfbN6O>;j-L2b#yh>0&41-k`&!ZY
zFn^Ud_?LL(ud`-9PceS$OXD&gWzU_hzsaYkamu&s0Oc+7i+=5w>4|@NS-y2XVjRlm
zCF@b+BZt18^Rk0DHV*Tb^Evx>^;5(%e@u~omP39-&vJ-A`YPiE|EYd>Vn_MATtDe`
z_No@~MNjKSW$R4uIY4n}9O8oCk>foI@u(~=trwy7AvxAv-rLeoiu!}r>nU1K)5rZ(
zo;<?d<jSk!N_%?JU;e`*e^yT4#}J3?Ew1UsAI#gX>tC5%{m}!mm-V$c;z!~`J^IoM
zidXAVWqxj-EMM7gdN0WMoX_FY`1qIjr?1~>Sof$$Uv|O^KFCSuE#vW?kM|7OlN|mD
zT^AqT!!Zv2%P#Eey8RP>GH!OjyL!f{9=oxZb}-E=?5{p~t~<ZrH`?<%`n!+Ew7*n7
zIZ!|BY5!`SWnEx>#GjnUxUN2ba-Tn%m(p`t?bRa(-{wK%rZ=AO2Jw>efFJT8dy}gj
z|7BnAZ}As?q&@j*p4Sh#`~=_lhwh7C^EWg<>KE!iy-#2BgmVDxeCKcatp^U>Zl6!u
zelO|0Uwn#3=N;tqxp$k(9~j?dFgfh|k7xS*ewSsx%u{4nNWYFo`!{yCUuO?~DSs+&
z{?Su=UUb!hgI621>x;WD_P@B~eUtkxd%*(#b-nArV;`S9?*fD0J-+M4x1HLn4gdN5
z<yYI_7xNDu_S7ryzx<E++uD&c<Lskuy7b8IgD*br{U;V#V4lJH-E)ZzCOq@4-h+=`
z|JFmh-S)pIZt-kfP(1Mi$lswnD$laJ^DpZF^DI<<RL`5<I^fg~a^H>IbLh5<J^a@X
zbJzW@!S8g*{lSN;-u%63a}8d5{`>d+?1NA9tGeHM?j76DGx*^3dJnqxi@67Xx!6jF
zuYG6t!Saf{WSuDP%zK`X%7^NU7iH@`a+TTL`quoZKk*5z&pp@j{4Twp%CnC5{KoS*
z=(#(2*4f6Vp6i~^doJkx4)wj?WZmhxFgfZ?UwzS88(cF_37V%<J@lI*InZ-{NKfdw
zEOZ^>DMk7~&-u!sakCTNy=Sy@>7r2&ec8R_c?VwP2maxl50V2tkGId#ulDLe_4pmc
zyZOjGWc_Tt<+?aFKZ<j5tY^I6OpbNAylnk#-KHP&iacPwCI0klokE^<pm|pu$xrHQ
zpU#iwS^M8KUfI)gTz-uY^OXDikloqAxku{Ho_F(4_GMS`&rigSdB%Kg{cPO?*$3i5
z-o~S{JcrK|@oJrHJp!Hkm@mlHj^9`h@B^5(V?W5x$W=dWFF)&#e(GD_<0bVo?I630
zCwalVF0S~WybiTXk$+jIk;}f2p60!-+noE+h(F9|`+b7=F)nc=4(-duoBg_Z)_e@j
zyXyPhmU(o{c{4U!Y~;tec^9f@9Y-&5&wkbe{3y*s%G&vz2Yuuvev!rjf6!jPshzE#
z)t3*%3AwKGBXX?Y<PYum15{6bQQvw&SwH1aKD~3Bh1dD>(;wyHNWPSpjN5(V7dPVG
zxb2Vm3;$pza>S23j}PPHr?nzKcP=B2=qJv^wQ~aFhQ_CVek70Kk3Siw>&^$r6CZfM
z4^*E&s)sM~`MG?8Z}s%&K0ES1`iV<C8Hc#=dna)xp50frK9)z6?av^+<qvkJk21aS
zAP?b>UDU%1z3{+p#-)G#neWNt$ND8leDY8JBR=^9dD_uWyf|0j-{wc>Tk?<o>8~Gl
z6gS3eT>Kdt4}H|9ue=~W%=`Ee_xck*p1;_4>mN__)h{$}Tj$90?&Bjp|E3SUt%Iyz
z%HaV!44gUtpB5-V`r@S=;sf$?^DBMWNgR<Up4bn6?1u+seuY2zTYgVbzG5%_$p7RG
zamN3Z*~>WSFTUsx<t04GH_F=aFaC#bJn2_Gap%7N_z%RZagr~e*)OvfJ|X?ZBYBV>
z<g=Uc(Mvt|>Fv7y=qJAPYdp|-q5Q_bp>>@8$SH^By;_l9TVLA$+ee6ZdCIw{^@jbT
zdY;$GbF-4)ccpexPn_VBJ=|w!_DE6wP+#20C-n86w|wmVL8zW_*oWCi+fP}4<ITRx
z`1mJ%=wV-@f9>=OwRhh>3_4F^7kSnD#@7G#i}sb|sYfs8K>Fp+#$z0iJ=mYWvo};v
zT*z~-lc!(fXFq71<SMhf>(2S<FW;0y@gxrQFK^JZ)rcqaBmE%%x4x1m%rDxxF0RP6
zU%|V0vtGfAe56di^-9{Fo%Kr|Io5f`VSZvC<8s}+VZKskAJ?s;?N{Yj_vL%{_3ys*
z4t?c!{h2@U=)Q58mz5zqkZZm*|Kkxa*4^|a2S5B8zwALT?aa&cXLtRv6HN23GQTiy
zi8JvaZ-^V$)f2zYd9PjL>b1JR(xde2pZDy)#qtZ5tOMmWc~M@Gcdd)k@s=Bpamfef
zCwdtl|CPU>ex28}o%7EAJ+SjPzt@m&)z_cv>X~oJcMhzc>*f{bf6m$I>vw<7gQ4%f
z(APP;`COSj&RghvfBn}EJxYH6XPz)WIM4Qb!&E=>iFw5QN{&1uew6v6{LfCVn<ve)
z>@3doXWmoR-ul>jh(CxA`+0KhbH$<f7N_FK{@OZQUKF?1$Kp}^tEWt^d{+teFRmef
zz;BB31b<IaUc!q!Xq?93yc_@GBt>y*odo$ayNW~W3;mMEUh=r-9o|3Ij=j~BAL(uV
zBfj{b`32&UJZL@4KaB@p+OsFS$n)ts#klY(-^sJqb@HwJApYrvCwzF%(RjtLJW8&5
z#;YIuNA|-nzm#|7d8od9fjnZ~av#bM>@P0-eh(jb64&_GFMHC@dVrpOXKY-~FT@`n
z@j$NYo|jl(>Ysn{H+mYcd}tiniyQvWF7g+<S;z1%Jh2-&<`HqFUw&^K+8d|m{rtyy
z3X~7%=lA>O5A!>}QBQki<F=2MPxTAsCw7!qQ-1NRK0ClPugUu<@-u!_4viDywGtYC
zCFCD4)e}FEeIfgoL-VhFEuO@ob2IA%^SZpje*B2N<u&WRe{A3Tfo$KNrBqMjlArMI
zc{O{=uf9VE#RdDw+xqeS4Sr?)NN@dF2N)lH<X>?MtydxYLwN^UN5RxT^qZn}kGSO5
z-g|`P$z$RlfASLhxDQ=tNAnE1<}G<b9LWQaKIHKy*NsCy*PnTWeEp~Gv`^zeJNqAb
z!g@#<$~X3R+Nm$Us4vgDZXQwJdd|FJzSmwn@I&h+_sMtNIw!5~K0nYu{mUV~=q<i$
zMRwA^@v;|t^E>`4F4$YY+ClB0{-Acs<narBAikZ;xv!lv{pbztXXFd|=uh5}XWZx4
z%GL+u(o0@&oj%5g5AyYEeA-!$umd?M-}0k;s~!1ieb@D4zp5VOSN0R))%uBE_EF^V
zAO4Fsa>^mSYDIeC4^QH+S*RZ<ZyK-Pu~`2bm-?wc(BF9!9{H8_<&eDsXESi+fBxUJ
zc^u+X{HJK2<T|@5<6pdq7xJNgQe@X!k>Avc^nvCddh<_lB5#Uo<1`LD*vH|+ed`13
zDe|lr<xlgXex1XJANd7O@`-uQeevSlMm=)qp`CS)>()0Zuf}Cw!IyEmZhXq*@*8{@
zH&jnPl|S_Z%@^e0EgfIlpK<Ds9Qy_P2<uY&0Ot($bI#vj`ktZrj$G~4C*OVNC%^e}
z#Xs~PG*`*~760_bzj4wZPp-4Ka~J-^Pw_^sb-yzIhUOFd^jWi?Uqbe_AECGPwspGS
ztyovm%lj(g%X*R@vm<|Yzr9Gl-%(g6dw%43y>+v7t#zmR<cKHZ;&1H6zl}$H@CSbH
z{GWfQ$KLAW9kK_%<p=r`r{u^#u3LA}8`4|8g!~IXelOs862y~wcqAWR__H2z9luaL
z@?Ga2?n8M|-ef=b?K|WV*YT%)islW!n{ZxEA9C5#dxh?kuRnfg-0V^<TK8J-@*DoA
zKmL%S@yP@7Ex%0lh;>zS8IaGN2Z>Mlo4xppc;esg^F#i?p6tsH)OTP1;zr((Z}cbM
zlcRt0CDdO2l2`O2KakTb<lhi4wW9qBdl;Xxcr#Ad^@}I>&BKr#Dj_|f=R)Gu`JTKj
zPpI$zse2x4zpw0kz;{ON^ZBE76J%fLyp<fNANk3;n;on>@Z-Am4!`6-{1MN}_#-DR
z>)(Cr5ZBpDUKLN~XXSEODNC@EIMSc8@uq0K#SWei@WWIN`<Fw!rDz>kD;ig|==?p!
zYTK)4JzXuzXZ0dKu%0T1vyydlwPpK>O33czkpGs$a?fXKwR^SoE5(KNSc?3*5{gIb
zH2GE@G2hwm$|IhOrq?|eNYOkcZnRStr}jVM8c*WU`GMz}=4XCwyz-Rx;y?8#?d1)6
z%8Sl*=)-UH$KUKn(|LnF#z_ty>8VVPIHRBb<XhKM`Nn7c>~|IF(T{w*nE&{__|y+Q
zQ*@s^{$jkT|KiU)h<AMBp&Xi*_^I^+|2Hr3XZs}kDev2u7u8p`P81g@8h<*jjxO`t
za)_t)B3@H0*RSjRuN?B%a>(8(@)Les3Hb>*`m?Tvu9rjgpzlt=lxO?<YEi$n`jt3r
zHJaa>h4O~;0r{m`<cH$dxp6uVk*B`=B=4G^Qhv3wFQG5_@|Jqe{m3!j%BT8)^0EAG
zUt%BQy#;xk9lVzz-`baW&(!lK_32~YOxu~CAUXEm&Og1^MPKu`{fz78bNe9v?!3tS
z<=n{mnfINXZ^`fCQT#Z6a}KJ$_*WLM?Bu$B%p=ZWwKrcVtEWHnhu`HudC<J>eGLBQ
zJQz>xq8<O_XY!u)DgWSC=6%=Ax616p&*{Zq_*shj<)_N@(!c)11K#9s_07ZLRJ_oK
ze)z%%yJ?R{`at=gANl<$JF}Pb#?-Fv>yMr2X&p;1{TXlC-uTmY+T)GAVakv3rtQ2R
zLXLJ&eedhB1HR-V??V~4c)^cx(gQE#yKbE9Xg!?zxBl^}9=_;lKTocGiSg=}-uh(+
zc9D-~&E9|VH{(uua9^I3_r#HPwENmSzcF6dAzoATJjlMv_?+j<SK6@$q>tw~@~icT
zvbb>0PTn<to!V{bfnVhQkD_%BetmC2{)Nu<@lCGhM&!bw?+pLi1vh@#){p+=FXMsy
z53;BJ#f!X<`hocd`d-tPXHPzS=-P7+R)&w<`{GMyf4y7Vb>DHKxBoY+zVAWd*ZAcf
z{1_*F<sZmD_`Nv5zj5&oaff&QBA(ob;*C5g9@Ha;|7$1Dt4~k&wS(f#c-7Mmx(>x}
ziuUQ+rTI<%Fb=%Z!+0V8!IyXy&(L+_){inh)Q8$x*FonA{1J+O{%xPl?~O-XKzZ0X
zF~3QX-uPipda^e^y!q1ox7~KYJcHHKPX6;=tGt5`a+Iwvpmn8njB<+lN!Kss)~odJ
zoW*)IMeAU8fa=jxK62kVs_Wi+_j@n(^zV1Vevjq72lvzV^w-Y!$IRR0DEt3q%6{KP
zAI~j32iG6uXW~H~k!RT3`du7ZZ_9V=E&uXY^2`JBqI%6j`+{l_@8T1R3;Pyvtv>(e
zkL)XcoL}-6e&zdH`mvsY__p73ekJe8ulkoKyss%w;M@Gq&aT^cv6KCkeHj0M<jOnj
zZ2ch5yDq<~Z#|(O_vyv&<SqJ@v!i%S<?&DL%)5Bhue>6E%SZgu^%Ui6X#Dnt5MTBk
z#>Ia0V?XxA7d@1nXNXJk%$txt^<p|+b|p{0t~)2B7dz4in$L~19P)=0#hLgrZ{r8A
z?&FVM<}vM3bYI-lo1IhCkNrtH9_@|8{H!1A8NA4&kRIZkzuJe$L(XCNuYBXae&jFm
zQ*__FTMknml<5hrH=I{F7vVQh`&yBGt$+Bf>-<#Nc@00&zx>8;T~|)!sINbMq#pU=
zn*Xt<cKkwr+VdlR;XZT@gJ=EN7m!O2=L70FS25q<j~|<t_&5Hbcw=Arp5E%=#r)#F
z>#6?wbAIIgi1d9;{Te5^;tfBp8@GPh1ByrX5l8xA2dKXFo_WZ4<S8hw@T#76)@RlS
z)`#@f4qCriZ&=S-KhqD^8dp(&&St>6#Jb76F0YAa-vPB=vaYf&vyQaBht567bDplO
zKkFa$opaGkSv~72?d30VE+2^h^uBph9H}oJo!i;x$S35xE}vSD$phwDdBl8e-j`4G
zo6ZBSi+lIY3)-isKlS8&`9a<y&v@iz`APruCKt*l%GMY1Xu6(|cg+{(uN3W1_*q)!
zXZD-wrPtke-bO#?mg#-%cU<i1)ek;??!o-me!+duBlqch#nDI4{dw;Dv+22?{f+C+
zVIjHDIbtPb7xObe<j?FVZpDwX_)%sL^R#sy`|%%`+RM5KiW711(umFX8hm}Xw(tM@
z9=`LW{=1&K$>_7X4#qq5{~4xrTmC1L7V9!tzF`-B;r$N&4DkxBgQ56<@>n_4A7o$1
z&mcYV?3|r^c4))hdY*V=KJVaFW+Oj^>-V|$je)=GRf6<@><;x0owq}N2-SmlgY<^{
z3W_I4uN19=tnb97yd<9~v!}cwU$Gm1Q;$EB<2rwq-?dYQ;()))FYKzUAAZUIt)JN;
z)zf*8eCWP@joW;uALkYJ*Y;25Q`gOh^wiG$hj;y$AAMKEdfU8XKEn@r&a?0=|9CDZ
zPss!9mgbMO2CQ}9POlu8rD)vZ!*ggn<4-+&i#O{U<8a@49^!!?s&8I1ZuQKE*4Nq@
zmvQrR`9?eZn?JSVZ^lQC`I!GYe}v}swpHqt|M@+ubxkGYH}rt?FNfAK&dJLm|A71e
zns4xK-24Gg>?&^Y0r4!}^@qQ5$gb9-_~7T_S)B4aJkd|w&>uh6h3rc&>sR9;pZ?+%
zFZ$6xz3`+S|8gA~Cx1xwb<V6lJ&lJScn;zFP2@sxX1!$IlP7)0-#Xv?U|r?B*}S4m
zj=W~xw+?ci<9rq3RXfksp!@WePw?!zxMUak+I{wAZ{x?K_WVqK=V$!89Lj(4miV7F
z`~3xeq|D#Nhx_a){^TXs)rayMl=t|f{^U#jibp8U$yJZvKzfP;yoxv1@yUMdrR@BR
zJ@Bp_`-uZ`^&_6lXW|Nyt9{yzeD%bYdfwkKKI?erTl}AY;m0`5i~Lo5k;_l@@4Ei9
zH&5yZ|N2Y)Q9IYwC)fD!ryjlWfJdl3G@sFn9Dd6G_!E8ksrKTJeU<5{fAPSc)-B?b
zUi?aX{^mYC&CB}bcjAYh&g*@r>ZYvcRnwlByW||u`<mP5o1Ze@EnP}}4{u%Oe9-r@
zd<W3;Ugv?%2|X8tzW=q!a?9^<#tGd^=5hT``RJfwi!QrR33|@#_i^S8evqyweV3HJ
z#w&lx58f-aj#Q6+>`jjRp-i6f`aYQPs3+gaTl95aBmVTqKa5YFWB+s>7RSzyyidlz
z^yfPNbq-HH<R|nnKI1Yk@OLOqDxv+IJg02_a?W9%mY3uW`$hHblhS>$`Q5nX0plkZ
zPbu10k_+kQd{X>cH>l?v#yY_I%YF$@>?&W$uk^F-5a-&vk0(fP`58at(#O6Urq{`H
zom~0YINc{lzo~xaLwZ1YOx{&?KW)eUzSnHs1?k5g+S8L=*pI%}hsK$nv++~?;}v@E
zTs)?EpMTj$TVKi}<`a2Revm($%izg4&HMT_9z1H#&-jsiVZI<&oI!SFPj-RgkUsk7
z-;jQgA8W^c<~#EozVVaV6aVb1f9rJlPy8$UUW|1-zQiHa&Uo3?x)}fB+j!_HKSJ+I
znXlQO-Tj|Q_w9c?kM!IQdSA<YD_=w3!8E^_|Hy;(SH8pTeJ^?czw{bAbeWZp{UoEm
zbpgHF@U360)bFD{CHCQ0&XM>de({Na^C0^}eDPcU#Gl!X{_0s5neW7fe%V(){0Q1t
zK>OQ9qyAImN1oTx1ApQa4|uE<^(+3^iGQ;r{;Xf*Iq@M5$dSj)AI{t5UGXeGo5$7H
z-v8;7pU(YV|C1*y^hGWo$=Cj`vU=or?qOeyH{+6@mH8Jr;$3^Z*!Rdo(0r=Ccrd@y
zn||gW=NI_UUVp9|C;p5dU-)%hJ$YCE_*Tz*=lFG9J?Ht_^AlzL7?=0f7ybUg=|`+T
z?_kek*dygj8L|uh@F0(~3qAOs`s|5Ec7^=ZI*(j-$2WP(+Tlrk<5wRq?5mzK{j6V%
z8>(*|<N1pA)-m)_PnkaS<rm73zU-`D^}DZl(aXzR-DR-*{KEM3!|v|$M{&Y_zGsVf
zsJ-|x9{n3PJ?QT`6yL@r?(j@6{TWB9H-A?T>R&rZ@6_MelfAU3KfcXd^iKVsUFAde
z(O&-GC*oN>a_MWl#z`OJ&>ud{o5q7z@g#oqqaHrwAL{{SahvwXZ^(z@oqY1-1Nv&O
zfBb7#E1LJLC!9~%KP%hk+xMz(-C&(#-ZZZ|*D|kKznGt`bF7!m`|M&pW4%BgRKHf_
z5Bw9dhqz{c=UeQQ^2RRs7nkhAj?j9?c_Y4ECx<`b$$k1bm&dE~H~SF#0%haTuYAp(
z>}_4<Ij4OG{mElz`$GF8zw@S#=S1QQ|M;c{-uS!xr#?Q!mv;D4CYK!dQ<OK;^(B86
zXZXdFapFOL{E|J}_Cd4v8_L$fDQX}1$&LrDe@7lX1P}k?2j&lXULKH-p?qQeAl~H#
zc|#s`{%KzG99sV3fAX)g_@|e7N_+XkefdRRl}F5b=0ke9p3Y118F|K?qH`<xA8+J3
z-_ReF&*?3%(F0nK$;;Mn(E8Up%=$@L-i6lN?z5Y99($6*p8B`{!jJditQW0w$y4SJ
z_|snhD~@<@xfzGdHCX+Nk9+@#MHZN+?fpD@k)xe;r~cT%y2AR#IJCpJcKC+I&%fyH
zI(g~5l9r7V(%(F1zUCM7<#*<FdePT7$fKX@?B%|8uH%hA;7fbsbe*5l8-L{E!+q$w
z@$27s)YFf+S05kjiWl*NKl0SK{<3~c*ZJbqILN^_KIK1o$#~SW&Vu-~zQUt*oqp6e
z4)yKBtaI(R^viDi1t07n&ssm?MSVQDuAckWS@h)p>`!0)$;bH7zx?ODLf;kjJlJz!
z&r6**d+rA1YtN@WhjyMUziRKf<gD56pD1hZ9Nc-Yb8qsj8|mfxtULDK{@*j_Yx_Tw
z<Xew>@6dOS+KZkGd)`fN{L#<&l<BLT{ao97QnS~`>f2Y6<GkGSAa-!yzEj!r1p2ZE
z<TvJFym`KXCw4Zk+xK~|Ks)+a_bYp@Ko98m|DIorzwrI1?>cI}lKy)=|KiAR_n*J?
ze4m3i9{jIGOY0vt_|!hzEm6X==Og@uTzQJ#%JL6fYxaAj@{RXjJSPw*;#ht4#G(DC
zc$Js%E5F-s%BP+;k}JNUe))m;mJf_WK9+}|@u&7sW_Ny~EU(Bz`ekSBt>e6h!JqZ}
z&G$F&*KeL)C4MW;;UeQU*=(W1x|JaNd0zywFTaQ6(4T(zg8Y@f#)o(H_$zcD%9D74
z?$b-0L-O%*$+eS@I`onrC2`81*qc9+%YV(Y=0)|~*IqsMmCfgnfAdr8WqxIR^fV67
zot=~D7n)DWqo3#D<cTwJ!H>z&p8fctaY1_O2XE>dAAQXW^i(hHPdn#+>3%1jPt~Uf
zy{t#9cf>cI$rV@p!Tjz2FXM}z`q!`c^Z%5+*X6o6h4S(P&+oO!^m}`^{l65?k<DvR
z-enK(OYw8_v;K^Seo%hLlk4J>Ua6n6H<ah(L-NFTYFF`To)nM#jUL*|f8s~HC|hT#
zM=$G2`#O5c`_9v}qd#Q#S;=><*<C;Q#xwoc2O5vEag(RLe#z0l{AwL%ed2fd=3%^A
z&uAy!`Gq_pudA=@K6D-ONBN(>@K61E@4`6vo&MCPH~q*bH<hD49*mEF(F?i``MEMV
z?808wMR-tFpMK^m^Jj|W(ntTs!H?8)AMzXe>W?1$hCitX<vsgLXg}rrz&=Y{L3zNw
zj~}G_FY)BOLVa;(9M)svSbVB4e(ks27Z>!BZ|JQb`Z^zSZsI(~`H=mnJfT1HGkNlu
zanl1DH~#5IPKxO{NTo76o3GOT^eZpv51;se@~txeF<;8Vcqa!R)(QLwKk0q@aq{r1
z9sASA_^jjci%<ISXYGyG`0GXbk!RfWGj6<9>W7}td&kPwA>^^E`N{sq`UOAqBS)DW
z{$Rgoy=7fyeS=?d%ue(Wuhu>CiFKQN<vKkezeth4LHWK`G@f$EKdME3Rx9GM5}N1D
zS9oMs{s+@M&mQuDa|HjFPksH6&mY*ueRd<)b$QiwDF2m1{sZ+3jT5GRp)8N6Cm$$_
z3vnrqtl#*v_^pKW<Im2Ayzl8bjQ*WBh!eciL!O`?f8wXcXP$GOsekhID^6gg`J%Jl
z?>Bs#EBE-}ylv+V&LyCIjCGdvAb)2kehTFc{wyx|Ej!eU=0$!_Uw$rL<W>1vK9GOt
zBmN=3Vt@7NOTPGk;u-Qo@~j`VS5G{U58b!^wO-}7^vARDThH-3@hTq01*8wY%%h%v
z;?4fr{@gsIUwmm#o_6jV4>{`TkA3A!?a9TD`}D_)dgga}lCQqs3(-qGb~8_>$S&j=
z4<6Ed$p4`HgMapszsNTZ?VJm_5AkFDk@n9|_$j%NpR@|)V{yU%Azs9tak3l!Qh(-0
zcyJ#du9L^FQ@cAC6Ic8MU-JH}+2;|Sr?8JaXuZcz=*LgU;eXokfA!c+yr_o{{d<1J
z-p0*8)raJ`FR#0<U-#*y%s$33w)bC-?y|%@CGo>f&PlX`#x3s}FTb#FHP1lujnnxB
zedGc8l6-oyYl`g459#S#K;D6P<OlS?lX~g-qH_xU;W;hi1M*wrw{LSFl1G2`6c6+w
z&+{4l;??}BJ$d>!e&fMAp4}H`^m5;I>o591_H@2Zp8NcV-qsz?r}bkz+IxQy5A=4f
zBoAvRF4zYj_-1cq_2_B;fIsb1evJn&)?4ITj~Opij~_yQs~!H0C-r~jt$R-T+rU#6
zZhIc1zBr<fvUQC(Pf<Po=X^+<nGZbgx8KI2^*2AkGk(ccPo8w&{K>BPSC*&rXFkQ3
z{#?g9|8n0v!p^SqFM5%qe}3pa3-gWk+L0?B*uj1Ck9PF&oQYqgXdPvLql`cEh<JqT
zSP88wobRRkEayn_5P9+&|L}VPW#^>oi8E*)N1p5aNPphHp%>I%eSWD7>Cb;%SKm31
zdi<XKwZk`mCWn63FXl(}e4k#N%Uk60Fa9jQxKE!H)#K;<j~wwrA910b`}|ZMP*1)y
zpU5-fQ{3oJ|MI4M4ZS}AU6<E92eDry2Y*-oefw8-I(_bvvgaIrKWZQAKEBJLbBiw*
zTj}t%@9aL9Ux`!bJ;n6*ZuW8f-nltB-edIp7QZ9%`x5ct+}$~hI8+uFl~CL{M=sZ1
zJ^ee+_a3J*y<Ml5cKR`1WxN<yDhE%ltDp9(z2^|>t4AM*Z}*kShsMJm<jPZc6L0+5
zI9)d{u?t@11@+_saju^G%8-93(~ti0yL=>`^`n1z4L{I*@{LRU7?1ky;{_UzyvhFL
z>mSNz>|_6+pA^kYDdG(e^uUAb@)Dku>1Vx(2jejB_)aIe%IuB@e7Q~zd+;krKYZZF
z{9%0bhT21X85flAn}zmm*0tipyaCN8^0?=`<}34y`NTY69O*j9KF0c$JoU{--t#lR
z=$}6Ph5hMoUQYF*FMaf9UL=>j$+ym92Y$r=%)jD`-|!E9Vco239@ZbewR4@G^hq&2
z$Dwa3hd<JneDkC6;a|V%8+UqN*?SHATzmJG^~-KhJ?-hE9b`xLhV;}<9OKJ%JR1+x
zueg+_<wN-e$`|ZSK9m=f`3pO^Pd|tU<4w`NhJEE5{)J!bS^h>(dZefw9_eSCc)_c2
zxK4lkkT0&Zv+fWt`f;CJ*P;9AI9#VcJ8EbA_=0#gUhViJq>u6O7yT(i@dwF)`h)a=
z`h)s`{0X`ajW<PkkzT%EW*yH?&KaO~&N<{^?_pVQ$Y%%aFmUGle_EhqKdimDweNQR
zs~>qdeGXt=nKk<ym|XeYzDIj@H4m9*^h1wSuJJ>8Mf@2bIo>mNogU=#OYy)T`I$20
z52=5XC!W|Duly47!!#dQ5AhrGhB7_yN}hWBiClUaAHC!YWqzz3x#Y=b`th7beb<d&
z`*QK3zZCIHu6fTm_?2@W{dyj*9>3uq_!i%uKcpz`oEPzD@}PR2XYfBrFPOH|o<7En
zFJ<-?FZ@+Ke!*W|XIFBJpIsookmu99pWZjm+E@BLsq5OO<#e95p7y;9^&x%I{`r@F
zp!&+@JMvsl%dV@hOrAKQhx@L}1MX`FU5E5{ACe2{3Gt;qB$pn}O<h-y+;p7m0@bHy
z8lTGQ(N8(;mmGeqp6@<ruk1cdc{UFHq-FB7qc1(k)n0w*I(_B8lppoTSJsc~$|<@p
ze#NyiRF59yq~mbC9MTUjcr$MP<vMgfMf-8Qz_gySanOVQ`f*N_%HgltrAUtU&TsT<
zT=aF$mG-Nh-(@?`avrIyp8lY76VC_ab$M9+m8Z>L^0WEC^9;`&%=4b(nMbT+l&!0*
z7p<$H@48xNc@Ase?z-Rgd2iqILFdHQlh%jYr)BMZZ`Zl7b*cN-tN6m7dE5M}9X-kQ
zdsOqkde(n<GJfk>a(zEG)t_GMVf_g4Y&~L~r|kKW{-ApF(~t8V=-dbL17+)9`jCq^
zX#Cpq6J_JahjoT}P=EC0KlJflobO9Ryh8mzdO+(pNFOK;ApM~6Li;AjPawa5<j{|Q
z>)*ObJS(SY9mucM_Z}Ag#l8MgIpzWFmBk79<}3NdI*>f=%_G*O^yDw(Sa)0B@)Py=
zU3y*K;wO0J@BD~;#EX99Q9K!kvT<9t%df^Q4xG2R&wkK+i;rqiKiad0>z=RMzvv&P
z^_AVnEBTNe*(aR`mDz_r?4F`|(LCunf#08)hdpnzJ~MA>XWeNXPY-&MV;=V$*|=Rd
ze(MBsJrCB-JPye*9_^v~)(QCGS7|-<#XUdeSLO19`cOP;FQ1jmbM8C$Op*QgL0ZPQ
z`^v@#U007B<91(JexRTF)-B{f^`Uy&Lp(q{;SI_cDevm1^6+n+pbWJmPyg&IKNzR7
z@j!C4SEdgnSN=0D{cC5O)-U9eqYSm@&-7F#Pkr+YzoM7^$)lfj54|CIP<>_3iBiPB
zddloX4x|q>9;jU<<bNqz&pH>S4}S1vewCN?M=xkxX+1pRn;!f~KgLB5>tp&s&*iMk
zt*foKy!Qvq!|cqT^t;yV?_BUZ<002|>pXUK-S-)e8o&MhH;-MQ<h@(^`klFb2)XpK
z-Zd`uAA4x{o-5ovGsCxbzUxjOeEI%@b)EN8yoaj3-?OSuzHvczqnGRSqK|c@{Rm#I
zcgfcuec3_2<#*o8aUVZvT&Tz2<X?JNhgfIg8$b3Z)?bkQ#XGdmgyx4zXnf`BndhN-
zDn;iS=IvV1x>6iKaa#`AMV!d5%ISUA*~9qhZyjP?2gzl3^N)N1T_@jppZn~tU+>*n
zmpT8j-m-494omNQK50E^{q6kUK7l>hnf&y5hW7L{e)_s@Jp97C(>PqGhce_p`gc7=
z<2DYzJGXAdhkjEu9^=Q8@wlGmdwIgXqg+2JKln91_wk+TXT0fkehB#;z359n`zig=
zL%;N45A~GkVI5B&>wEE$qITrzPg(roIpv@KiWl)|f5zX)qp$J0&(7rF2M^W>%AOz5
z$946r2gJSm#=(E^YTWpP{Fnb2hqyN$a>#cbx{e>u`{WzrQ5G-SLH^@>!MOB;NBpKJ
zPeXbeC;#A=?4e%DH$QWqzuUjcE3WIub!F$g&VS_xa$JY(!GHD7Z|F^5_Qo%JE8|t!
zJS}eVs-5}G@2=RV99pmV{gic!b)J6st95{S)-}rXQ_p#!>-4kkwXWh<#;ZT=oCD#J
zUbAMucVJzUwzK{r$9U;s{X>p*9X_ldwX?3Yz9Ls#lWTm|X?P&dx>$eE__Y_8)}z)d
z;!7Oin}3Nzexd9fpT2l>u7+oE%TM{4di;^U>sNfThkn>sJ<mbpMe?9_-XB-SkNnTR
zc*l=86@U8GpK-gcp7n>gRMs!v?F+2a@l1~K8c*7ve67BGB~IyS-=kml;1}eiNMHV;
zzVlZ6si!|^{^SSh>EAp`Z)N=$hx+oYJWihLc*ckE;x*+@T&Zt7DYB#M%Jd-L^=eTb
z@;-!fQqK#$cWplQ|B(ED6ZY~R`j8);KX8S9-3G6+Th9}3%;z2K_Yw!M*L%>lU(7xD
zw2{LuJAX{i!N+a6=u7Vn>^XSX%ij6zlF#;NyFTAdudg<IM)$$XJ+=M4Jx`i<@Zp2@
zT4<+TdJev9)+Wn7dBA)PL-U+-SM#88_<t|XY1M=3n_nS6;-BVE_7O+?Q+@5lvHB^R
zr{o7H|I6plI>`FRd@2voM;_x};$FU#2bINxdCEFdfBas5@-01}{*6<-v4eFCy~L%s
zqYpjk>HNdEAfCjH_W)<jo`3aEKAzGzWheFMW&HSY|NpUf=fQec<sHDYlLRCQ>l6fB
zAPPc}L@88<`-4FlNGY3W6)}K=L~#(2DHRQ%tte1sF(8a=(>jPnG}00ju=l#4vUI9I
zSsbfi*#t%tR9=y-<IMZHj$h}@``!y8qt8G2&N;txp7ZSIIp_C#Z^RY3^wR^r$%7B)
zTmEQ1y*NMn-3vdW4|za-BDZ<$#IM-FeEzSWT=qxw(C5$m9?BC?Kh$qNlux1k5E=*R
z4U!M?w-%}!<VAibKST4~FR?p5q4D-Fe#T$u-FrxJBL0l?UQ&IejXuATSJV;wU7Mew
zgT8w;aibozUs+GvI`WUYk9@}A4~m->qAQ;mr|y$a#3enmmv|D_;+~xJ%dYs;W`F*}
z&;0I`pIKL2+E3IEzSHqNj(Fwo{L#KDP6wgCL(Xr+rFp3*^}q4%m7R0c(cUNc9_6+v
zJI~zc6>~fH`0kzE&+;qt{J!#$?KkZ0e9??<yAM72k|*}~?cYcC_ptZ)z_%_q<R_bV
z{=ZJY(^QALclUSC(et|x_r~t&eeZ)m>$u-0kM{}w?y}$g_`6BkejjQbZS}jq2e$Q;
zmDlgNW@cwy>xl#7{Xga8rBDA4xc3Fd`G3#p)8DK0_dPwQC;ZVP`}(`O<iHPo`a8Px
zf{y3(p>DVCsq?qH?56!@ch)oC`GJ3sQ@^(N2;TSbANpWde0V=&ymk2lzwv#NxJQ?N
z_}+<r=*4sKA)nJPyIR-ZMe)6m@y4+mdhQ?a<J=706SYu&u^+fMlV{bT_H%jOc}L!o
zAN;?$@|^u!Ji8~-Z~ymPeIb8A`Ayv*FUgbY1>@vB-)rRWYxT>Y&KrB2eD9GrES@&z
z_tk!9>HqEYy$8Nw)=l!g{O+C|U+NtDJUN{w=~KV`johBwzt9;L)f4hLe(hiO8EtxY
zAI-k%4t&|KjptX^v45&3(8rJG`~zS1ZFLO#^ur(NgMO_~uj&)}Q+MN={^;AeRXwfW
zd}uv&s{LIZp&oL-g@5ajkKD$qALv)UgZv&J_)|xiXPkQ|b%M5awfQx@$Uo$gALvj1
zL`R%k2Os#;#<%(jpXiW7{mvioi5@?oPxR1HZ}Jc5p3;2w!XLY{m*>{OpZWOYZ~PLU
z_$D{|@*nb}V;uVE7;k;^<Y)Yn(>&)y<K;{1&@Vf)pZWY$J*-XN)-_K&Kyu+<8=_-<
zZTFYt)lWWhqeIW^Wj*$!PkONL;L~{HJf{cp(Oa*7zPCSFhaBu|K0eW>M}9>v{L^=~
zt?#*Y`Mq@@{Xy%K2fyryuIKvc4d3ELKmOSPrXI<QzB&WH>JRgs+wf%_>&w^Z@o#?3
zuK3jEPwdDK)LGV}S9YX-dNYn6Sl4>S;ZNIm^vOwX`y@V7Z|rD&{w9y(m%P@eAM^Q-
zdr0=im%4!5<YEW?<T4Lk>)9Xp4}P@i-+i#}4)lw6{si4CTgUq-e5ixf^U(c+dsX|L
zeM3Jz(KGwH2XKEte)fXKvmZ2mSR}7FR+otvc42>dGoD|m-?HA2Z}IJXhkt%VpOZ!B
zaOW52F82q{d(N-UdEN`+(>;T8gma+xroOZ9_4BlIJAJ#)M%Ouy9~kd^j$h|f=S$~4
zd>ik*r18-CTs`N!NY10b`sQ<f_V8mP=s5qm7ZB(C9RJ>n@^A8x-?`E`(K!-bay$PT
z2g!p^^UYINLG__|&b#<UhhF&szj9uXN5mnNr{p2~q<uvEi;EWGhhLEkf8vDx><{cf
z@8+qG(W8HIqGz1>;tSnGeg?%e#22(KWG5)Dp}HT6N5~JL`wZy+H-i3uI>`P|K8O4c
z(l^96G!HK7f42+iAJPNlFOdI1dV%shRDVGGU?M-X-`S`5qx>wt%ir?4`Rwd|K|g;N
zC+Z0II{b<r<wbUO4wpy813#5#=*hay+4e#A1L_+4r~TBv<35v~<UxMLf1TUWXJ_NY
zkNhalu)q9locl-jj=mSS&&ZqN(KzdQpJD%&fAFO~u)aK{?LNePkoVZ~7yF^3pPlH1
ze%UGWMEX0s(F1<WLr)#%x%*{#jhyW1IsWvs13u`BKJ35zR-Gb__`iPsOW)!JU3wyq
zd<DfX|1@8HOkVTxZ(k%o`Q#^hWiRse`t=5WAwQP~#J}~$gZcd0K1vSvW$wkqsd@Nf
zFZRuP*ShF>Pe%{(fw~Gk&*cT{ny+mfeynd@^u-rH@|<4PY4R?4j6+v{B0l63@`^Wp
zgdRJ}qx_N|8qbeC$7d_f2CbX&(T{oJ6Ovood6C|resN+vIpuTm7^nV{x5W=V(Vu<5
zda3_xlgIgmKCS2dLh^x6`eqMywNIFbPx~7`urGO!N$=`qa=E|9KfRfcK05NceHtC>
z%QsM5h<kdlPtb>WCl|l8K05LY|HCJ{u@gJfA2gqS&EqHJFrS?An11ukbFZMD!7o36
z_5=GGzUYVj$qm`ZdibF)^Z2JcgJ1e&Uwp{Z>N5Hf57t*#S)X34Pmk8s?_LET=98EF
z`mK*o>yp?0WF7uXo}_CZRFCoxacN!S=)rjUG|qkv?c>(vKkP*h?#tOpd}#BZ98YfJ
z@Qq*oW}flRb?lPQ=@I|*N1yDVU%cukzwfNn;rxkzus{8n$G_NL-pcizlb>y*Cw%fx
z`H4O74e1fz;zK_q4?B|^dXK`d=viKqU+5EG;=(+Bpr60;3wFZ~KO{dt=lA$vXL93*
zp5$x(F0QPH9y?o4JN*KEdNrQDw2i|z`N_+k8Gq)BBXZ#z(jU8uAM28z{_S7J*&peT
zKG>Dr=|Nsb7uwhGrA|VR{>h17{Hqi2iH`l+IQHXT{EfWOc=Pek-`T->^ld(U^S`ve
z`QqO^e36@<TZi1@%slmuah{`R9J}Gqc>Ix%o{VE>`Iz3U!_UxTKlcLc&flHK@ohc&
zz(>+|Pom!7cj$^|@hCpf<B#4?k;6E46kq(nbN$-Zkw3N3w_oVD|7iQWTmCMI_@{Sq
zk5Bh<`1Bl~^g*xe%@6q#d*RQz>`R~YWgY9{hh8DR*^NECuXZj(-#Gj`Pm4P!{>di~
zXj?b^!ub{*-*-WED*qEN&^&pHf8z(=p6eH<=F4yDIO{sc8AmVVV84tPbs|)c;KTaz
zAU%=GI@a~P*T3f=C;J%ZJZv5IMUP#`#b3}D7xY0+`?dJ@9>_h2{O*1NAMQ)!8_%uB
z9_l}SD!-<G<By)4$K`8rZa=^meY1yr>l{lj_@YmKU>)*$?%d=aMw{O7$$!zKUwnys
z<MA&q=$8lZ%fHzd|N6;^KE93TKlIKo@P$uw`JH$+pC6%%U-R%oesXBjulqxBg)i%&
zix22If42|W2i$MSllu8h=1X#T4($v0r(gSydGwv<1a#yB`qPiT{T_-t<LwXpom}=W
z^sJNeTE}?L(W6ItA!n{j@A~Z@iN^60ZGJ|7@|N#E@uMF<5FL4fKX@<9pEDl$m3o<;
z={HfH@I9;i#4hp}`5-;=KmOr+V0vdS^L&TF-{npFrSUs0SasjIr>)bmBR@bFJ$b}D
z`oce?2Z(R>RgcgQJ$TM8__V%wU@v+xFVQ@5@K@`5ZXP?*ul(VAVfhz-=<#cM^Lr0^
z@|_burDye^JZv2Q<G04+!{4b^Z)vj+JCF~b=!g^RTL=GsuSsv_rM^9<59>kug#5v7
z{1#p32W@<4(+j>6Q$F(eZpnP|qH7&;XtTe5^02e_dCtG;6Zt`Xp#JRj>n-((Iws@Y
zdrWk59s8R&wyrqDpMLS99#vl^dN1bt5A}(8_EU8Uc_4b;yFv1MpNO7$kY9)w`GOtA
zD?783_kHp({vo=?%Y%?y^eqnA(f*7M<R9{Z^~tBs)n*6!^F0Z__nywV$T)~UazXDm
z>6ab&nfS9lx!6}eMW5dqpZXxTddoQZN`1zz{1F}c@EnR~b`huOTE}zsAbz0xVdL;)
zJVcK^@CDhAKB4jUaqA31am=1sPq35vgx<)nj?cVq|Ag$~UI4%Byr}=34f)xfKjOnU
z{20%#&^2D$IDSc=?2ixU9QH@oeENp;$8Vu^{f<li#1~{Ae3%c7!-xA9@|&lf^XQBJ
zk=uB3n#V8k59Mw1o$J}(Im*64pZr9<f<C^~HS(8v#_=2T(YN2=N1Umb_z%D2xA;{r
zSeG5hp)DTRfxh^IcuT~;@%Ujc_QXH>@_=#ppfCAcKmDN3j_A`1dXW9`X*_$Y*Tk=V
zihb~*zA_Fy&#h;k{6&AB;|H=A)X#3{lh3;JXS{veeraE(H{<9N9na~VA6t*#&^I~E
zXHRmYgRb%PO)tjN54(D9evZ?I*2kayQ$9i$qATBuch9ZwewN+Yft}Pr^a=4pZu(#^
z@eTPSdFT~C<e@)w=}kYrt*`C5_;a3@&-t@FXZ@519qZ$}h2)0nck}tHe*Ovh9lfy|
ze#yx`=Got^<NQFs=8-><oy8wM=tI9aN%S1z1ONDi`l0Xd)uZGWkMw4~c&8V7Fps}N
z@>mbD6J&4oxpRbl$v)uw9lyU;SFnTqIMH{=z8_VG^LOVVzqjxC9YX*Ak(S@g$ESJD
zXYP%Cuj)QT+_NkB=!0LYf5@p{{f`g7OYr-A-_QCzf_&p#2FWFViBo=wzP32EU*lK2
z(}VME(p8trAME9Kw(PF{@;eUC>A^UDK!5Zl&ybH^oUiBue~_I1t^q&cPvRHizlGvc
zTx&!9FylnL7|&nyyVs!qMDZzJ>0Lkc-j-gSyTvuXK*#+_qI!{^qG!JK?Ju6=o4xcy
ze#777HFh)KI*a<}9MAD%eP}#HH{1H5d`@2W$EQ5SZy`UGXQ207@+beow|wzzp?yz2
zSI<NKmh~yW^W1odo;p^WpXqnrhx`g%c4i0dtPjM4e*8gk#ShS9f2hBO{8E0=hU|!*
zw*A5So<sS@y5a)TAN{(|q&Iqn*{^Lr{TuJ^6rl^f_cBf$gPwkfzPdwO9<eUFprc=&
z(YC%ew4SzcDTjXYcy3+K`Hwa}@DK5AeQnRhskZuuKSS{i<$K7_A-{&s%}}24+&bvV
zQ&8Q&-ynZf=joS6#gRHoenMB9KKVN&4?WpW#R-3h))#Naw-BG``yO1K<L?LA_x(QO
zAK!A*_uumJO*-<6=hiyvH}9*Qqny{=!#khx7j{X%(NBL#mwovYyNWaY{7D<~3wB^H
znB(+AcEg{xeURSh-9GHO@z6N@TZjF$&BvcQ03GYN*HV|dkJnEx`qhi-a&@e>{T$+h
z{^&tJKKQF~&Ku-{#`ClEcXY@rUg<fXtGDrEee>13+W4Ty)HnKx;+j6?iA2xQ)mCp<
zAAR;Ehv)Vic_;NwUMP=P7yUu=#0NyrJn>6T{0@ucH{O2k`y=<HF!vvAdBr|u|C5j8
zAN1{W{$8*5(eje_X~wI+ymzB_NDtQc{?_?}y^Ql-+Wv1Iv@Udhh4LKaKaicF^Pc?5
zp6J8Ox8&n@>@QC_|A{Mp0mYH^vTYpn+<N#oUp**Z&4coXcKSJbp7;9q7Uq!`@=xRO
zCr_#~;GT2t`NFv;uGj6o!^@7kdA%2m|2;_gu7%Fi{!XIx`6;{bL*ENb6y*_mWKa8n
z_Y~G+XUPBg-5@lth3XgU$s>c%Jnwbsm0bQFg*wLjHup8^9Qy`3p4)fOgZ2yli4VW4
zr(boT-x2$JkuALAO=n$s%wubhCHi{^_#&se4Z06X<lil9*=4f*>?qHu|L9A<e9!+p
zch4f;)pz<ce#Nc&gdN=%n9nZm-Jp8X@88wM{6*bFf8L*pfBC?A{K-9p`YHWcy+BTU
z;0t|m={cka`yahpN4}DmAbHpYJ@**wLT}__AM)UvpWws#<WtAWquR#nhx`YU6JPky
z4}Je>o_)}H)BXmnZ$5eCS9#4i`<(9`{aqVvaVbvSXREvHkNnC!-*=;LUC5uH{Q%0x
z<f31GV;`Y6dKM4v<=ERedbBR2FLs_NS|9SS7P4Op=`|5O`JsjC1IXS`T(ywjv`~H<
z7VRhEmmffUL2&{3A58o3M`#_0PyUeTIpnvHU16@zKOs9o^>GW?3HJK+o3^|we>&HC
zE*{v+dno4xafE+)OucSgi$8uQKKYG0pWOV}`2d<X2%Qt~1N~iK^nEABf5-=&cc6Zl
zavImdZyvhg?VtGdyw2bCP%qh6{hb>1m3qbZ0RDcp{)=CE;i6f0&hE~>_}-)5dG_qi
z->uR1_m}*AZhx=b-*>Y=@h|rl-ZRJ>>}H%g0{_lW^1M9G@6}=IPVYJ8S@R%z)`9jh
ze&bxre(E58B<`I9*voq-=V<YoekG5{_xy(6%M18{>Qe82{aq}7A0A%0@4Tx=4w=)j
zBUFE*#~;Kcv>!oyLGxN@{f!U(=TCoqZ0$~6E<gJF<m}}h)P9#pzxd|o@{T;izr62K
zH_@lOq;92us6OM5><`&p{YGE%mON)Ye7EA-b8#)77|$=%N9rrcAL&J#zv3tN18u*n
z<&XT?y$Qr0e%;e}4n5a~)*+Yq*3)l1)Q>;+N#<D(f2l8Y=oK0-UK5Rjp8K7HHpHj(
z*h_t%e9{Mdk&hkuC)CgG(C;g}NAg~YKifaaYd?hcGy9-%<R%yN{Fx%XrCi#O9>o{F
z%*Pk~4vXUXxkB=|7iD*GKrVVs`O>enA$sCn{5y}3)3_GmGm(FauUwa%=->XSU)(we
zqsNcM1G?<$x%)AE>xcXU-{>2!EgsF2@8xrGgARGk#}|I{IXTIVj&&ivJl8fa(Yo|=
z={bk(z4wY4W9q0x^)<BrI_J1IvCrA(-N(4+rFZMwKcW2}UFQOI6~2r^w}t2=qB{u5
zGg(af_#T#?IHNE7g}7D+Iv<F0`B=Z-hvs;B+W1pWTYkci=RY+n?|Yx(?^k=z;{P@D
z|Bv{8*o@bPJG|rAd(Qa$)G_1n<^MUczP9{jzk%eF7a;%U_u^4}7>BO7qDS-T**JR8
zHok@G2=Q!P@s6*wtGE^a=4ogCL(e?)^h5dE`N#h=<bA7qQ1{~6&fVyG4@STAgueP1
zpZIcalTX!&@(wxEuIdkV;kW$6?|sDsdi;c6vj;uWzyC*vUG1CrfcoVv@<aPCeM95%
z&)?9aKj%8*$b~=k6ePbm5!aqWdZ9OU9i$&}nrD5_>65=}(<A=Lhd=R6UgPlNzSFwN
z7drT`K6?1zKm3#5c#bZ+Kz_lF_~9qcrR+;Sb}~;J=DDB0droia0`J4M$s?Zd3yp*5
z(wDd<hv(*tXL0LX>D-B~cs1YtYCip-Lm$@l96C??yV~~gMg8wr-P_<t-mqWNBl+YP
z{6hCa_;v2Lo_id18u|DaJ?ZCf{LB89D8I_1?uY!oKpo`Xj2!M;tjGT3B^Upb=h9!b
z#U1@XamMfY3wy~&{(h6ZXgz-GIX}=w$9~zu*7s)F&u%T`Z-dZ#<6)7$>|^|we`;IT
zeH=e_pQG)$dGy5oIZm92EAMI0b^qkvUwnw07RpOaJpZ;&ZT86O5&r8w&Hb14VcJD}
z4XSIi4sW&PAN!~}PFz6a=mmPtj?S(4BZv4xN8Vv?{K+fqC?3SEyrEzHr_Jv6CHY3Y
zYs)v{&OGr5(f8as{EdA4i~RB)6!+xd*Va!o4&U-l@<Sfa(fvcCxbZs#zoY5(-!Jh~
zaU<UNh3DGh&p6NdKRQsq^|bjBy6gn`oA}i>-+q>K(S!1W_Zr?$c#q=!hVNXA7x(rt
z?*+`)kDo+-hEMeby@)G$4_)tvmTrCEjyKO*zq1bh-S0#1Y3$>J(EEM&6%fCW{vbKf
z#}7L`S14brAAD!PZoVVr=gx2Vqi6TN;vR|@`NBPnJZGOb-o2K*A)eLuP#%*n<Z*e~
zx_*zSjXvZ@>fv?1cgyWtEnmAc&%F>iAwOYX^XY+{kloe$+T_O%{}nIlmhpP;a|ax7
z)q`tB$*Z2Q&uF8ouCq_km;9<tukyV*PMiMR)8dz1z5ge#@$3ripX`rcc0&*U=-6M3
z!-wb2IokAW95jx;<Xgy|_G|v(_mlqL7;(W5$S*#`w{_8R@8LevcQu|{pC93aUgT~5
zqfPI=3$wm`g&#Y&v!Cz1=!M?c$G)pBfW7{H$3OXN$|X;@*VXo(-|wXDOZ4F0o&4-Y
z9{I`smiYF6`<K)I>EOTUyg&ARrMlC8WgqfBmpt|r`v<gtkYBv&m&fIG?|tnH@`ruW
zdi;ofyl3{De)XfTAJQZL(2owpFTb$9eLy@w@<V!s=0kEr`!ggzBp2k@kUvA`d`J#x
zT%!FBs=LJ{y`n>Z{FPqGA%5vezj=_q7?<^pw)xO{&^U-Lls6&&hUy&1Um*VIXR;`N
zK>0?!E3Q3fCwUG(_6OtL?~zYlKu6m?V4VGDQU5;3bNR-4)}e3w;tPFzp@)t(I@;(P
zhhNXNa~wL*{s)~4@ngSsUbDYL`N}+P^sIv()Q>-UP@h@P`p#$0ckDoqiS)oO=-^X7
zKIlu^zJy=$h(Gl3i68wr&$`ffa@&Xa2RitH^bYN_^kbi~zPcBm^yj(wBp<!l|Fp#u
ze$kUp(UV7v>&17H_A~RzMgRCkhd%Jb-|3H?*`Iv&C3JJ1d`Lce<B#MrKaqU&BEOSI
zKfCIuXMSLP@<RGF9{>0vH#yX^@(jB4=|0LlZF!Xb=+%1sO5G3D6_9?cM}Ge0xpM-2
zIEPqY{Hu$}k#y*XU(f@(+U&~jtV2Hf;*Zu5U-XR+{>ZM@hx*Y`zv!obd}v#T{pl6T
ztN0#-_F44QLF&ci2fyM*J!iar`#iezjK5w#-syw>$p`T_EPC%iAN<lixqE&7Bfi)d
zU;HMKeU0b8>QVBLm;TIC=faj;?y=N`{F0ySx$@2pUUcoO&b<kAKdA1KZ_-ZqbG{T`
z__u$tpZeRn_(B(7&RhJ*yxIGne8T${&K&=~mA>%l|6Aq<=&~F7`1Ae=fA}$<UCc|&
zd?~K8Up|uWp|}+P@(DZBFF)gl{KmeGp0>P!zI{sDI>vj?-9r1%ut?rS`eZMN4|Yiu
z$Knc|9GB~9%Qq86{*%bQiS#@O2kFtimPk*D-d_$v{t4~B{7)XkpMAT9{1Oh*FTZP{
z{i%hSM~7|md;TIm#2I8y<M^3)qc8K(l^@-c*f;nseM0$WSi}#sKSKM%Ae5)wTUeie
z<6m3;v5xWfW$|d=mhbH|_5<tMkNjOa_h;6F=0oes%lLun0f-K?Pe6P@a>LB?^aagJ
zbYJS;(LKNK3EW@$F3I~&aqfE_ah%V6zr#<_buP0H*eCGgUQ<34U&itOtn*rR#WU^a
zH{yh!LG-oxz2DV{EB5Cf;uVTpZTU%?KG64lh&%|BPkx{u9r3sR&U^mgW7{q0<Oh5t
z(m#8MEBVxOdU)hxTby~tVe`9uzd?R^)pK;oA)e$fcHlSK_#~(OhMsc&z^^<gPtuP(
z{f6_-KjzinSg-S(e9md&+usN9K8s%InZLS6SGS2H_iWnuv#vZ$4sz+INBN9?(ea#I
z;zk>v>1X(47xw1Q{L_96@deoh9sI}(?3s4dPw)I-SQMAf7Rt}+K!2xH9%o<ghv-**
z<vk+1Tztx<H=S_)1|7Yyi~Wgz(Tn`;`!#m+Udj6~{wF@P{T_~;<tKJF-uoc;{p`K`
zjVJB${f|r^;|EZj!r$hTn>Rl2hz+~<zjw~^*S&b%ZsCpR?|<s%n~x)T_&I&ihkV2j
z_%(j$U7K9;toH@-k@?1nBl@xLK>5J=&if1R{gR%3?=R?+UGOVTt%Ltz{ggiV3w_Bm
z<b&cAzxd);u(iJV#(Dp0UH->j-lvKK{%#&R-s6*BT%*e_@<rA;$*=XHJZ(RqFY6cw
z<tuti6mR%zA$oAoIQ&>2iVN?Fq380aar8t^$nKs)egxSCiVw(6&^U+=w2u2%eB&4Y
z?4#avkIJ8oM~@w}<tzGjZ%eO`K8)v&zLP+Y-{Lp>(X*bq6@7f0Z#+GaJLmHg<E;yw
zYtcuKJm`@JKh{%ML3%>Zc<*P$wS7o_HV$9vd3xZd=;Bk`xYRR#^z#G!!t@97TZbO3
z<9nRslbq-oug<~`dytF2vWNN=zv$3c()WH8(&Hc`7gR5b2l~~fM}B2J{(vv~GoBvF
zZ9UJ?=V$u0=@mcFK4Lt&kbUTt|Dj7i@{4}{NDhduIOGS}W@r19dFX4aZ}CZgkp0oe
zC;spc&7)7_^rK^+^IeVmVdwu7PdfJg?bn^%`99J<>3-SY@ld}y_oy?~t=jIf(TD0g
z<Dfdw-`Rlfzw^2KapOF9&!Vm-5B_`?>3hNC$NL4()q#*6Tc{p&USW6kvG1u%?0fuL
zT=n|zdDIW~U;XH)yYUB|FQ9uCeup3aD!%c<f5-#XW&F>+4%Ny0!+RBa=0}Nshv+*m
z{$gE!cZ6LYxah5mFF0sMM{nwEdZh>F4|WjO>N-fj5M9VmA$<)(_kaA9T=ZjKq&M*n
z$%7Aa;e%fJn{$D>m^?{`|KQ&^^YQ8Zn0!pn=*zFpN#cw=5Pf;v`s$}V-;j^o#zTCV
zXI*+FhkkmaZ*=ho_0zxg#05V`r`L~DZSjgf{Afde&&Yd0{*m{_@|<<dhvHveqYwIa
z-we&emvQ8_4u3(9y`4At19{oeeot@e1>@u!br`*(Ydw0l-#YiAo9pmD{!9*SbriXv
zJcciMk)B$4{`uN(zgK4tLi^1h8|i-#+LxYh$Zwu4<X`N`-`P_?eTkFY-@GqzZz%ub
z(|q*AiMSQl_I-H~KjKvWV;BBpJo=vN7oYmo?f8%<JQuIz@myVP9rNh}zxbj@=QRD!
zE$BkypgP?-F46li<LMdwTu<FD@5sya%5L<9FZ(a~<U#UV&%L5NFaOi4JnWt_k$n8d
zet>WLr}goPKK|?j<YX^(j6820ZS?sYev__o^hZu|+dr&leg0z{{;Y$p=k{IaA@zoO
zSKGP7JarAe)FsYK>RsnA>v%6pFXTvczpNgxPxAx(p<};Q=ZP=t!;F7=VNZ1nd#k&k
zIJEDQOI+|LZFXS?`GUTz&mW-s4EAavI>RD46U7<-q1P6YFOlBRYoWM;;u6XeiQdc0
zNAjO_=vzEFFIk6vJvR@M7ajfN!Vf#PP`rE3<huZWhsb*{^znx;e2~|8@BiEnJNL=s
z#@YAW=lGo+`tF^?6TRz4Uw&q1bjeLm*2g!0q<{N_?=zhjoGZn@eCHe}?(84db?)|e
zRplpo_q!@>c>x{y-ghP9OCI!HvAlsk{^SXHPCVgT-jpZN#kaVZPyC+Ty@q>J{@UxG
z&*%fc_#vl!Y@F{TolEd3f1zVM|6p%^DStUn^K1R&^ZP{f{0>W=#2=Jz_y_;vH~4}2
z-AC%@7m4-(@<8i}7k+Gi;~(tDZ^+5N$*G?`=mkA?q)*6h{NFm-=CKpG`Ga-Tx9T1D
zfA&T4;TvE2@tyH4-`Llz!+-U&C%)~o@*BR$Z63MM%{q`h*-?DRGyIufc;77E<wyI4
z_4$E)R-RLz_+5|soE-d>UcB!x&i-J(z$ZP+v-}TT^zh;Qz(4r~Ims=K(Kp}v&~xkI
zlYimUd}tgbKQs=q{~$y!G50%mLqFFg5Bd0qe)}8w=+AS0D38c%{Epw!bNabB<e%co
zd}v+qn(N@hz5&tK7T5BUe&+yfdBS@G{$d~S93SdydXNu{v#xw%e<Tk-%IE5I^W`6U
z<k#duR~+ZQ;qUf{Z}vgYb9@<Ref(ILe(m$}F+I9(G!8%RGd(B2{T3hc8@;MG>@VVz
zpDybEPBinUIN>M!iC=1yll@w9;>&#eo5vph9}wf^Y5I^y#UDA;<@^qP{>Sg^H}nI|
zqbGbp`New{D9-88czRc#X?u^T4)vZ%8$a?je#CX6@#K;h^h0#)qx5GyeX*bY2>Sbf
z{$8T`-gEW1--S4T_)gY$md5+8dGqV1UcA+N7j*2)5B&ez^rhYrXY!AF$hloysvFIl
zEXp5{A0*1dlST2?!dx%g;w4dD%lyDk@k^ijA^q}WZTgVEjAuu7;7{VqeVV#Uzr27h
zz97Ga{FOi0U&RN0=t&*Kzwim=QGB2)FUxo2M_(MJ-UqeuKM3s;6Ge5XdQ&{A@6?gL
zXU#fCd>h9vogd|4bt812E-%v${jrnz@&Q!WnrB`7ndiBBja<e-`Pn#i9lHAY2fg95
zg;_tN3(;$#bz4XulSO>Aken@~zZTk`TF6d=utk5O{ru8B22}5$hadZ*Ha#us*F*fq
zzR6$YBl%5S$`kHq2BEy9ew8oOTk^GhB#(OTeS!K0-|pY!NqHST>&nCSP5DlJD=(QZ
zPnwrVpZq~x!7lib7x*<h(XaXLW%SEO>?lvjH|igCi1F6*T;1WEA&*#xzd-u89)CfP
zAMg|U$CtbZ?RWAeI`nB@;eYffzv-8M@QHuVQ!aWn4jM-;^YO{Q{ho{*?CE|{zx_rV
zUGb&gI`Sibvo3!c7Rf^&kp9@my5dFLX|uO@&{kJi2cP=M0rith{V$Hy$?`w@;6uN7
zwXXdLUFbXw#p|$WUHMI2hi~<?JjYM**Rqpw=#t+&dWPsj=W})<AOB+?blv~Ti`Jzl
zcJbZj)Ll2-;FO)$?ruE#q+Q>C|D193dop}Ehg(k`CbxAUd+`(BRXR7aH-9uf<Iy<%
z?62QEnDJ$v@!|p<^56sV2gt9*bE3F!VUBOLtq1u@qI(B^3O7HxH}mq#XLju7+{cgA
z|Lh6*k9GK){Gm>w2mSO&9(M3N^{?N(q4DgaZJhnnKE}S<(E9uwU+j$z`K(7C`=)vL
z#;5(tbMxpg_08Yie>vBQ2j|jWyk}BxxJNVIeBZIQ>MP#^_%4v$eYfqN1&T*<xMwqu
zU+^#bgZxX}(_5nRvwY~D)O`7jyzIz+@(ko3{K0&FfIs6~NU!wH57-x~XVgLZ?d#5o
z_BAMOpznJi`S5MugYNazef-aJe0Z+kJbXjz8Hay#pnmhntzR4B%X9L0E{=>7=i*Y{
zWlwyHJAA2YQ(k#jT+0jM3w`zWu*i=Rt-}xHYyIRS4>S&{!?n$W>Nexl*Yr%E>}j6o
z>;R2({!mv#_s{Ox?HBTyeMcVJcXam?=YDQVCx7U7jxrt{_e}DM`r5kkkLU7=dV?L<
z)A@ni&JoUs_ID^x<4b+z93vk%7aE68bs%}wY4R>PjZ1W{V^{AT*p=V%2YIlC^g(|0
zoqQ+Xy0@b@dXR7J_lfj|4ov&7ll{Uv+EBiS<g$)7IjjTm%U<~9AKJsBeZ%_Rldz9_
z+O(Uv7uWoO9mJWq!#8`8(>|H~?2kTq_!;}6Lk@N@U%zqA51#Ws@0lQe=>>ni_}v11
zSr?xWpY$l-lN;u9`5)if=<yHy@yl%MAB60bXdQVA`d-UEFQ2+!vETc?LEAX{m%MBL
zGR}UN?~Tki4;|w@#~1zbNAo<l9y;z-a-8S(73<&|eb2QM(baFfeN*0GZ+u$cdiFi@
z(PKaJQ+{o6s%;*++Tu^V<5S!E`mGDe!!PYi;!?l3(iTtRP~7@XS3XY^ALhCDL|<Hr
z2mUR8$+PxBd|HP-%#+9QEwA`~Pd_^L9s1W#UgM#2PVS>g-@5MO>^IIq_HlCH1OMcO
z{7hW2CqKi#eS)3&H~p{&f8k&J23kkF^Lx+rqsQOGuXq%{)-zxH=!g2zBd2-dz<BbL
z!*k=!rw{rekNSpw(4|jpazJ`T-+OiZTc6$NL)$ofXWV-a$RBbZzsCo?XB@GQeU)E|
zN6-18du4H8pXJBm)%jU`syp5H=~o{>^T?$?+xU?uoQJIgJvUx_T1Q-pKlQNn=u6vt
z_Yl4Z(T@*w#Itea_8fh2j9+bfW)Jp}573bh^+Woj4|X!o`Neqs)_0CV$2fMTH~Mt$
zuntrgk;A^g@A$VkQy+WZhA-<o@41)J|HWNDv(JI=o7b_&>y{t)(GULcS0m6m;zfRt
z&%~X6<1=2}hqTZ<D6hyf;*<WwxqRXt!#PWx>mE-2L(jZi&pLh=Zyq_#Pn6#vyFq^0
zLUzQLJW9{x;qUYdjZ3}Z2Yo0{L2^R#V9HA_D6c_!h5X0w?)gg#@d?>G(LI=c>(L{7
zTHkuci!*5dp;z?v<HvLRBfVJ<vLpTB%lO=Htiz7{&N`6Z@k2g!7CQDx{KDjiAK=e?
z>)_LU3_8&I<iaPq>OFcvAJP;1u@nD;_5=FCujkfhU-a_1{?voEeZxBR>N$FzL;jyA
z?u|1aTF*G`MDr8T!5>5qKiceoK6>UQJ#BiUU-s7?7V%-9vJbMaesxHq{aL+`^OKIY
zeGl6AJV%GW;xorvzlFvp-{evscyFUW(Q|yMTiDY&+UlcTymuO}Pu_FX4R^h7YNyVi
zKYvfhd*dtjop<%fA#=J@=Wln}P5aI6<n8e~u3snUM{oSS(`J`%d;W(XAMv{`b&>Z@
z^v<v3Irl{TlD~;_b`g*AAOAH^oBhZS*&984v8Or@sz<f`|IF+wkL5lr527m{+E?)B
zp3S`zdg?j%=jul1U3I1Rkm^8nwR%Y&r#I`{AMoiM=lmn@LwO#OSKgHme2<B~eD3_f
zuJQ#s_(YF?us@_XZF;r7Jk~<*^IAwBEyUk5Me-)HKl%AH|Mh-L9-#;Nq+jFo(~tLd
z=o^oYd?LOwuIN=gz^|b7T4-JV#=q#3|3T~13%}xL#-YnU)OGkE5BZE2$MS)=5l8Z#
z{A|2&{91gn3pCI7{?NUbaq^<P3dx%&Z{f!}^vRx(z7y%iJruvEH}Xvs=~;fmhkp6W
zbABjas8hr@l)v!Fj{GE1e96Nt^nNkXzH8sGuZc(Z9WcjR&wDM;A-`ik{M*lsv)_Av
z$-l{`O;6&0pVGT=__4nD;}41Uo9x$?r_f;+cC~NtPvf;c7tfGCh<~VV;0N-W{Y1Wl
z)>(YUvX9^Qs#)Fs8(p~l5ywp%Lyv!YPA>bu{FZ1Oe<d%z(8ouP!xz1oM<3*ocdch0
zv<|;Cj~+cohyKL5arh($JE8}T*ESDb&-EKe&)WDmp1oV>cdh(e{_|a=e3Ev`w*8CV
z@kKw*rM_>0-g82J>^lyqp2erQcAp~->DT(6i)-{f@Ad0@$e+;RAI6j0bNN9Vf7<5L
z3%;#`AM=d!JW>3hD<8^l_Bs0?lyAvt|B+{nlQ*3|<VoYv(YCLX7ytZ$o!z&Rmt68O
zKhaM<ZRq}%9Q3AcCx`Qe{B!Nn<G0*q@9E>`wKKOkcf+f8nbtXfs3TsoX7{B_XROmX
z$N9UM{%!~Tdq0T}{$(GKALIk^55<{%#C-lmFZ`OE`qex38~yT?@%9;ZhSr7bf?xSV
z{?QNR8z{eM%Tq0+ANo*Vqr;!X7yjtOb8_+*<Hech`svxap6lmd@*ZR_`u6{3L+5sW
z=X<ZbcVM?fexq*ne!<@@5oi33zdOIW*HDk@m$&Gf9-z4M+<QcEY~Qr6&>KJS9P*d+
zANH^xTi^a;{mmZQ<Ib<Y`|;6ShhNc?=kfx6`H6M-2~?Nk*L;3pzP5NwdH4hW<`@27
z4f?^i_4zOU`MdG-fUb4%g&%e`-#GKhg-`R*H4h)=$rtQJul&S#a`6j(qhFic*41Vo
z^w~8Lee&@y{Od=L9@!V_SJ&E~*jfC`ht5;>E#ufvT|yq1bq;-b&JOJ3dpYsIzUmhI
zu!np^PvnBmC+Y!cUtnMRfblT(fKT%vzvB1!L`NMVpOO>*-Y5EAN4;SmlfTp-_FMgs
zzUY;n*o!>qum}C&*L?Q?^d>GKJ(_1ceW=H@6X^+G<ZU57tjlibiXZ!*@%ZcY>sxlB
zFLtCS&&|i5Hoxbm@_{^rk3?}ve~`bjhxpJo&U#6gz0AWOz9Ih-ui}|p#_>z{#PrQB
z{7pYQCVFl^=8yPfCvv8I?8<IwC-m8i9@rt{m4C7``t+r3oVY~?zmUGj&Cl>f9)4y#
zKFt%4;?Q`{(c|yxC3-cVT<mE+{nEdF>yTeR`RH4lo*_BVg~pTDc;EN?of3VK+xo_9
z<5L@7<n&zIe26dn7;k-f3Vr@bFVJ)An@|7H_mlQ5b}^ryi6?xw5PyGY<X?&G(?WjU
zLh`qe-2Q&FddBZ~)Di56jyg&n!yo<HC;6Lv*6Z&B)|K!0E&mYr?tl0be{c_DJbTNp
z?s4o7_>`aIef5C+oa>2y^W=YR{))c^tM)qY>bs|msrUT;Ngb+=)UWNi_W}ByOVpdr
z4d&~2f6i{|Uvzxej=t}!%yVAzcQ(;g*ZO-(^oB0})U)`K_xO|V#MpsfId6N8zJ1s4
z`|!o@%tK#3BsYKc9O9S#yf4HzI~eEtGwaenKE)mW-7}d-59aen^$0saa-nBFd0J>b
zzUhJf=p8+BIDahu$uTcm`l|(<bFaUf<$G6p@g1FUzU%cn3jW1!$RWR6blFd5zVO_q
zMy-!O<Jp6L#1XynOaAD6;SCRccZ1U}d3tp5+Fv>3iJw0{Y8-vAzjf$`-K}SR@*0Ov
z{6O=qXB~9-3Hd$e5A^3bxy-k|c*DPV5ZCl?9y|E|1>`sR3t!sq6Xj=lRoi|m&)bLX
zqxN^>wB<$nuX*aH^k@5rb*zsbI{335y5<{izp`)Wx8FebTH+bs=o+UEz@PmAAI8y(
z=jK_*{jhQVe^=w7{LEhBTs+(F@a>!{A42^4ZrJZW-QVFa(L88hw+{W-f9aQfwCPhn
ze@uI-6X->qfQ~xW`&{GI5$LH8wCM|<^u|Bv$$ISQIsWzYC+{IVH=iG>OXYoX7-yb!
z(8ag$)>RjY8~x~O+uy8Xymiop`tf6ZXdJrGbMyHjq-T8K-#qf#m+-G&TOL4H-cl#v
zo18HD(vJ>)$U{zbopJQ*xjc-|q-)+pkssp+f9Ug5DE^Eyuh*}$$w6P}wD=PL>}Fp?
zU)^SXc9Wm1r~bx|esvL~SL>-q@M&Go@$ET2pniOsFYnM7JMu&Mg1_JkKjc#f=qIN<
zO<sI>uFap>*?O6moJ-J?_wXeT(`&D9Z+x-0{K!v@mw&BmJbThN{gK0S_fWZ>wsqK9
z{<UvtL-rvz``A~}5%0N<_3)_<Ql~kesGG>4KJ;9@B%iPk|Du0!!Jmwi=gen6^Peg5
zSNjXUBERQ@5P#<L4|Obi@C$ABAP0Rw^{V=r{OF4#{p<<(qkYqQpB<TV!RqU#b@;}Q
zar90v>Pz>C?p5s9zU%WHvwF-p{P9=*XdUlm#Ibq}zx+a6nwR_==ef4^_!WDa58YQ7
zPcNR!m(CmH($6mZK>V>MJIh<vF%Oz=oPK=k=TGSK2lj{AFAt#u@rNG%=#8A>Rh*cI
zAM45s`tbqjQQLT^AAS19pLO(;%RJAG$FF|lpm;Mr(R}iFt{>m{fyP7QJ$DW<-*avD
z=a2jwy%stj%R^8er+?>R=V$i`&^?8Dc~5Y}j`Oa%_^D}Q|Gse31rOdnW9;)Aue{{x
z2j~7CqT_u2gHwO8+eWuPHtPPvdEI&2J(}^*ICY~ov@hAW$VXoHLfX)MpSr|-06O$(
zytqK$zNhUz13!Mhz<%-rJM&+3_=WtBuJz5cpR3a#|CC?NBWDYJALaiSmX8zV2lbSC
z*WVR#KDTe!N9+rJN9$hOeD#oXtMT$5{dwOE?GNfPc|$!(58}xFAis+v`GCFELE=Y0
zbf3n**_+>_AG=>-PkzAO{Kk5T<dUz%fqY_pdBZqy%m2kQKeDd<M_iIq-sh+GYx<!_
z`ZnG^Zhy`1v(>HW(m#8M2mWgxVsC!qUWNau*CD@2KFsH*#*;_g0{K1UZ^rX8=OF9a
zXT=deG>`n`A`iLwt96qOd!tM5`q2|_>OJeDXC68D8^4hk<zx90pOD|Vx0Y|@GyaMm
z`RR#YyPwie4#;2lsrC6cI{aB4#{XJ3-nQQ<_x^emU+%qM`{=fpZnt{sn0tHi?eArX
zOZ3qtuYU2E{ni(E@&P%-kGLfddg4}|VF%|q`o|~z<KH~`gY=AlbkPy7;!xgVPkhrS
zKJbsO{fWGuqvQJpa?1bsgU*@!Ry;euh-beO^!=sh@<q1otKtXBU+#bHKjKc@<A-14
zU;Lp*59$eVC~t^&ZF!3S(~mgCxA-+*zQ(V%_>y1f3qAXRJdu3*Zk$}sgTC+ZJ6d)z
z4*&E{F6exj`#*Wq2ax{p$v-BF&IkNOobpTlub#mV`}$s39>))TvbXyZ`O19pqTjmw
zNt>VX3vp*2WG{Y#zWvobf%l2(8Fsf%XhU+y3)=L-pU6die$QUUYtxtc+Wdtd@iXya
zJUv15*~2(=p>dF3damvJaBcFjgLULb$PVmBUyy#RhfnjMe8NxVZFJDLud+LT@Z335
z8y(|4N7r-lgkO9ahaUgJr+#$heaL^&&2jt|f9B~oUfVe6d7}GS{W+doiODBD*vG|*
zerTWTtvqbshkrS@qbK?$m-Y1XH+-UR9J^fn$_p3Gx^s3%@8)Y;k6$Joa*0#=G~ask
zgbw@S9~zhQ@F9LY=imGR%ERVC<M9VQ=lA$f&)H|Ji+}UTO@G#*7j5!G^$5MPi*pjp
zJWF5B@6J8^#yJB$=Mn3ehrV+qK0VjAK6%VjUs}ihAWq~*^R1J7+9%~Peg&Nq(KBED
zf-bv|)4qo<bqsy5gZ1qX^rt@dTs)DFpWq99`;PWsUvx|FyDy&7nde+dUU?i}_EGf~
zd9qG3-hSt~y39U_KjZO@zWqoWvNt=>gE*i+_OU*_dtb(1^czPH>VN)bJ$7afdSaKf
zx4ehGc;uhPv!n9}eHd??`zqtuf#2baKGYH9!iV+c3-a~iyG-|BlSTPtSj3OK=U&u#
z&;BIN<yU!=9q5l-><;OLpBX2g^DBB{cjK(bKK#x&^E_t{{$w3?=SSv=4@fW8ht7ZO
zh7P|IFZK)c=$HSX$3MiS^9em!M?P_0AdmIrEp2Fh`XVoW=*x4cpS|hLy7CkK(u02M
ziv#+>CwcLM4n&9lh)4Y4-+hPs6n^WuxOEPfkL6+K5&MUI&pP&X@$SA^+>?*|@~v_F
zQ=HJJdlK^7ALUVT$UoT8y&C_fXZ&c(v(_gky3jdJy@W6KSMIyK@6qpf=EttS=B1y!
zVXaOb>3x9j-1xD0aG!ur`qD2h(5GLh&S59{8TziyehBde`5%9yPkxWSyav@>{L6hD
zKT)sCm(K0<Z=Cm5=JQ|oZQiG`3wvpMFTnq`JvZNTD4&x9l3V-zAAI?eFCMvI4Dwqj
zUdRLO<Kh)v`l1JR6j$PnUgR6=sNdutD6Ykccyezj{>8g`OFlARKR*;V;#ho}=RQH6
zGT;3Ze~0!z`#Qh0&#;R&J=+f;zN{mkLh(eO_8al0-@bwm`wRZWk$Wv|f7f4o&!@h5
z*I&N+=@H*?h<9!GUhe1YpY}0-m)X6b@1flX(zD;i`<<w`rGM+XuO$b4(7*goAN)+4
zeAbm$e6Ql3+264;P97xJAVkOiUyUC5tq0kcAF7AsEoh&B_-AkZ{Q47T9zWy6OV{oA
zDZSDQKIIR3h5Vksia&mQ^N9~#_V!z+cl2amU~hc0pY<X8T8EvCmv7|*`Gw#6-r!w_
zOh4q&lh^9RFaGc?U*Sib7;hcpAUW<nW6gK2IBIUU^_#DL``7<tLFYX#z0!-gC5L<`
z{{MTgJ6`pfsW0rvOAh*HfA@;^vz>R?=alU>S+86Fu3dJU|E`%Ge%X&+#F78cQ2Y&h
zUTBSDk3@3XKNIP{h3u5*JPGLu+TUUFm2G;3;%%bnI|%ujzMZF``bIu9PM(yH`47JM
zh4_W`Kl#mgZFRJ~?7fP*ik#k4ndiB95FhGQd5GRU7r*j9diF2*1)uh3`@Md37<-{>
zJ^8_Xhdf|kx4ya(fA#})vUThiz5clz(hvL5r+OTk5An^8`tj}m6Gh*1`BMI2cbI<7
zKE}&85Pe84`ZC{h=W+IMej$%IW>5PH{?S2S9%$u{R(~rmWWRmX{-UnI7du1t=11Ol
zt7G}Kb=0r=wfU2A=0o+g{A)h_8RxnBUHxhu<K%Vy_6==u1?5e4;5YnO{mmcPN&IP(
zQ=44sdwxcr*29PR<7ed8Z(oG;DX!!z{>Xmv0sTSq>5Kl@nH})wx%032&JX(W!#?Do
zfBHaQo;1#U`l5erD4)xV#_?-<7N`7LJd)de0zbePy4GQDc~pM19{saF|AFG&c>9=q
zWq<SB-wV*@f9!9*b>&a}&_1ai761CtHy_%c%|lNc-}0S2!msu7XZakG!#N=HIXj~x
zZ?da?a@fbT@ekFZ<nZ_G$e~WKj`~$xiwl0kkNFqBh2mEHp`&d+zthi8$!VVP;@Ul)
zIFoPC!KZr;=XP-`o{dvyqi20}G{43#zO>~D@k4%mqi217XS}%NFW%>p3t9*N=39^5
zjFZp#rTj-u{w)5`L6=|Si`@9No_XTD7r%3(Kl9B)UmJ=O>zGfU@@ejG`sFL*$zy%;
zh(qhr2Ycz~AMA%Oa$67I(D}o7e3&QSh+lQExOA^2AKNd)3A>6zayehRH{_@86FnDC
z=FuPjgzSk={q#f6kbLZBeb4zJd!S=Hq;Ko-Yy8qHduh|(Ak4ZF-|~nyJ6KnK#%Cga
z^^0F|M=pAy2l-Qc(-S{uH}|OM$z$y2{*=Fw$GYenXTE!6>#1+}7yj@|5BTz3Y_{D$
z<J&#9de;4g?~Hs`0m<Qa9{BYgtbXqwuDo&S#}C}%=}~m)N!$C1$IrfV!H3>5Wo*{M
zv;JqZ3#W{U4|E_t+}G0^`-n^XBss}L|NIx<=;M!lpmEk=H}#@@kiPLrKjz6F{E&Uk
z=ZEOixB5^WC?5Ek{l$EK0j;k-h2jWZex%<x@q>SLHa*KH^uZ4NR(_*T<Iy1>dCWJ@
zc=Y+dw!CA!b)1Lk#X68((WiHIVJC8m7yZzCB>rPQL?6HW4<G2`o82LK>4jd&mvq>j
zpUB6?(HHv0LH1UM@(cPg55M*e&+XIn!>&o69=tE0Cv?eSJ^IzpuJ}U_@_T5!b<ovz
zZsqst2K>k$;!V9Fe%O~E;uF$4{nDHD=?7Y0-e(W>2D>K8tMtGQ{FEQdkL(5QAM&)k
z?spXSCwUWneyLyHm8a!H{!E|zSew4+huy@9d@0}3cOrh!QRm8I{Dyt`DLt6SFUUzA
ze6x>n>J)ZnA8md>PIBR&A3@{fWpeAMH~GeT{0bfGl8gPtfqFUp%sh6mZqmhvJm5L>
z9fQ9|wR+}@|Ms4Zri|_Ij$`jR<MUI;E<NY4z4u-*W9(!1?KJJU>!**s^^!OKeEK!h
z#*Ewa_0tc2(^gZ*{QWKdh@Loy?8T4M{^TbglxM{c{^bRJ!jI)u`x)e?_8G|D?8E=G
zA^P~yk1uv(7k2gBc=<#>JGJ^fXJ>j5FZ@tHe?wnCI^LH$m!e}H{iYnAr=0l0Kfkpe
z{@6+0v95mO@hhJj2hmI9Px!=_Ho4fBeB`k%IgE3!Bwsj>$~VwBcBcOp;t$#nApRge
zAwD7bA^qifuG*8+0DZb=(B`-DG<%AF*zya0g`W6vpX_@CZHTToKwqBA^QHVM{+!#?
zan5<rzTr6(7xp*wtj}NNduV-i06&&b<!y0_E<W^I2Oad?YuXp=JMz5x$NqyK`#iqQ
z)3%N_`5}I+t4?Ge^#?y8hdjcc=+`{^mi<(oguVWKqWXy(zCWZ7@}wU538W`_A(#2;
z9CYbjKl+fKjd!m>PX1#(>p<hsS4W$tE>w5On<)=D_!EBcr~Y<NU>$Y0ws^6w`Qk+#
zP9AjeYhC^3^9yv1!!LUq&+q7oUh!*vbc|=0MEQoE(B+5XG!Y$k%6*Q1;Gcc)g<txm
zfBwGkwrgJWi)B;C@1fM0<~i4B(?9>GuSEVtPulK<<xNOF^3o6cq}`m)*adz0$GXsb
zexgl2{!I>kh#%|7EA|6+cRuo5o{+DMPgK`rK2wJ|_u9YIA^ZmN3;9z%wZ47N{)s<%
zO&-)QpIR3m(Ef{0IJsT_&u5l8nSFd8ub%YYnjfim?ceePzm?aVH@tV{x4x6&_nt$V
zcaE2zoWtZ7>w1n~>+?(g!Vkp<G@rjhaUkCDZ(aKhy7Y%X=Sy(}`6<5mul4vpzsC=M
zHlKd9+0l6><C<O*$)89+!=ikY$Zyz%-1KH$`zig>m-*}<9~wvR@&iARhvaSkN*?(U
zU3RlynU}~O=p}#Xq9fmHTaQ2S6ZF($#^FmoB)4(ar(gcS{``r)`4#`6C;jYUJUZfu
zzO~s&eBy_m^Ii@=__PlB(MOMd60NVUq8Ib&7ysyq56{!S+Wf4A<|oR7_W!gadqU>{
z`A{6NgZ)x|(67yI^eqnLZTZ%BnBMdI?$dX!zGwA22LIo%?{WNo!S4;|+qo2aJ_zxV
z$S(%rp!eCHd;dHLlaFEB{E0ux!~B(=<PCKXJK0aYzYq`NlK-Oby+`f`##!HZc>_Iq
zm52Eg{xW`(AO3_MdFa9ViOw6oyJ8=F^!n#o`ynJ3d7<^miGQd+5j}is(}Vbjo<raL
z$+z+-Ijm0~+KJ9b;#mC4C-Q^5XukUr?L>4Td(s#ClLtCK$*0a&+Vmm6IUjlMe8S)C
zcg}szM>(JV$ZH+^(+8v{a<hYZ;!WPAudI8t(?42m>m~aCZwFz{%Qn6f#a|0s{I>eV
z56n0n)K0!z?UbX%55F81Tl@_>KlLaNw2*(baL~H)*kqA@`M-M7e(yQ`va9v<lN%lV
z#-nE(x{&<xn)&+W5pD4-4#lTDJ6S~E{i)xr`Ob-dkjL*loD1wf_(G2!=#Bs4lYI1{
z-tnAX<*l@nwtP%K5Wn=Q?!#XTjfePw>I>if%7@nV-K9Fucc1$G?#lP&>L~9Ky{B+*
zz%I@O@;kZZ-N_<(#H-(ju?zoXM|R~Gp7RH7_SNs4;<@?i7Dx{DV)~o92VKwU*>m#f
zr+;=CU+?_eKDF5+t4H{Q`R>uk>-%?l;Ft0hdqDKaK_BE656<`a6$j26<m~n9Iedw`
zH=K9=F|Yo{dY!iOhx;G;#y3AxU!jXH{-_`S^uS-pgHG~gJ|v&-!P&)g^x2;tS}5*X
zD1M=Q2^Xzcboc>x9zK$I?b747+-C3To!>Pi@*DDyn;e7CIgXw93FM#j3_T~0^<l;<
zyFvcTzsZ4rex?nbYuSNcq2upnvM2tn2l*BMqj&sk^FwGJ{_t&`MEb{HBL5>Feype6
zLj1~S{I-SY;ESH*Bl#ocC#U>&&^}M@_SuK0bmXBAam}x^6Rq#NB=1Mut9b9@JfPpX
z#d!BA#@QFem3uV%uX*l~_>=wEy|4Qud6B&SewlN+eb)U8`^!iEUc2A7@_+j+y|W{`
zIfv7m-y`_F7`usEf3E?b=&=tJAAXM|PveJw`@0tYZh?67yE1;e@96F)&i&k!&iAGC
zsekJ!E3e;k&CJg4>)k)m3;9o-zujdw?KitS`Sc&{x5?vEJ9Okbzu%`nbpbuIhdPBn
z(gS}s-hB*u<fCVLknf=V0P;6+W50Dj<oxWtBR_PGHjbY;ziT^ZyGKyJ^H=S>M<Aa(
zA|K;V-gZxAJ~^$Q{Lv$M=?Oh{5EuAkN8|aAxZ%I@20!5+<fR{WGhaPw|HYs6p*+D4
zjC1ef-orR`mHMAwu{-_Azw|@T_?AD}McX*>F8<Xq@{n=%UG}sd`;pUl<H*6DIo>$>
zb)K;vzm-?X5BaCIariJF8VA|Wcy=@%-`>B7A9Bg(?CH7upKbi%+dAm+e`p+j2ci9H
z5aOG>*5?Q8PcF!RjbkVFa~>4W>I-q_eE`2j*F1Fsl<yMNU+||-T=C9xzV_(IGe!BN
zh4P5})<XLRRR2N0d);&8og2L9+F6~vqhH>#zxn$L@{j%3-z~~|onGvR#@P?8<L{;H
zv|!bJ=bpAs*Sc@?duIE;d?`<QuhT;EjLU!U>93i0<`X0Iqdpkt^XT5p%P*hVJ#f)m
z7hiDDjPd=Mz7o|R{7gNh-9mLuBL9`2_&xpeclQ|D@}lvOKZ_Uf&Cldv-z}*p)L~Gb
zHIM&^S9ICe-y?$bpw7wnZ}ubm1hl_;uP9#eEw0Dw^a~zXch%vKjqn%xqz@?Xy5BL*
zerFx)(}Q;IYvvn=-?VSer$_sreSyB+LqPP@0rcd#{3ySXPd>z#dF*Yzy4XFK_0)^>
zhc5o<jo<PMc}tzy>))gB1NADo$z>is@XPMxz%P32hHv@E`u-k*cuJHP_>cNu9Fd#f
z@dJpSyhCs1<C{ECe3F+x(f6SJH=pOeRqfx^0Da4!>_gw;k$=l?>SFtec=LUjIN-++
zeQ^loQ^?=MBYONHzmHJYI)Cs(^wf{m$G1AjIawUi5;@hyzWel^K|QJ7RL9YiJfL1s
z57=*=!^tb&?Ca`8b+9<HKKcCpStxIk&-q3h-}v?1y^8uZ$Ela;o8P3L*jMPmc?iGu
zY4isnd&!UN%Rksp-jiSH%f2X|svG1-`%#W}AHq-Z&EK_s4`tuuuh!+a_%cpCU|n^c
zarElmhMe-bI?H<e*LdUbZ$AIRAAfQF<(H8D(X+2w&pdLV?|z4z{G0x?(ZwIsFK&%b
z<Y)30zZSpZ5R%t8er^AdN3_}5c>JR$4j{YXhrRiM{qC6}IR+v5=*jyYcBCirq9=~+
z6UN&Q=!KlnIBouc4`?0v7UGlN$Q$lcwXF~NGkN4|>${ipU5oYE+jlMQo3-i1eS*9!
zZ<$XndXuj_w-2GCU!5qg<Cp&U8$<_R5M6wclicz;x!jB6Lpzb*laHUL-sK5)Q~$|p
z_C4s{*ZEO9(Q|arRnJ1}LwP9Cc-Z3CI?%X8c^Ex;(>U`z*Uw*#hsl4he_k-&bEqF0
z5A{QN$n!*W)$#UCZRcC}@#Ldtau`o;_Y2zcy!;NWhfn*PeT*F9!M<a@eTiJgJ6G60
z>@Qh=;cHm5K0m;({Em)(d}%}D=+n6l|K=^~zc;2IsLo71sK3#n2YJdqChp}2>xg&t
zqx@>TwmhP(uFST4C7;Om_D$>Bchrx@tJm-iom=g<_(tD2_qv{2&%UL;HO}`g>O6Y0
zf0&03ztnbKL62O{(dt=t<TvbSAF&?3&9h$W4}Ig=&-)AeIJxX=^vq7s`uN8uy;`3f
z;#EEoZ;9ejUK9uRL;gdb<kfF~!iV}Gk==5h`K^BILhFlLNKSQ33)xjXLiUhP=usUo
zQN$Oz{D7Xs1-a-;zj5Yy&L8yC8-B%~bFqAb5BF{S7aeGQ_R_X4zV!1~ZS|8n2;Yz%
z`Ll75Jl4ZEq%U!x?K}$UUmS=tbfEo8e9CwHO#YGg#2vIAy^H%?kAC%|pLq!%;@SQ|
zUgu^gUY)<)@8Q>dC4I2Fe*Du1`Pm;mcEP`X_nz)Gt;62rq-X0HM-KPz=*pAiur59F
zW5^Ha%{ulI^Vx}>?H`am*b(3S!20wjpYUsPiAR3S->u7@^v=%o;huzD$%TJ*rzh*l
zU*^#(dhS)^efxoR)sgH8jZ?3w6VWkF{i)9Me$V?n^SswH4%&x|(~qv_{=XIP=hS~K
zq<{N4eX+aWQTYA3bB}t=IZoYW-_kbD--Gh^oy3j2h(3D8d+r?Rdo0M0oR8&Ob)Iu6
zIsDGfKJM=T`2K5~H_bWzlF_xsKKb`+u3P^FGsgd3)eqnC<Q50tH*;*sT8ppy;Hk66
z4t?2Y{$|yJIb*&HAOE|S|Mlh9oU_Z<M*Q6Yd@X<DNxOXiBh$yk2R-4Jzl$gP2l?sK
z?`PcqvY-1CavEoS{6O>B2RfhAAN^>L%h~_?o$O{lyZT-T-`3N%ui?{sPV1PDuD{#B
zZsd6FquXA(-Rh}ho-bIn*LhdpJ!Om>eh)+rXdQI;9X`<Y_YsWKf68ggPx$ftr$*^}
z!@oNBk$3O3))@V0cgGz+^{&^?9^+T;JIIZ$dHxQXzklZMr18g}JbLB-J+}3%aliY|
zRkz%F@}@J#*Y*1Z{;<c%_a1q};%Q^_$RF@69@qsx$)9*+H)wrvK>tv@vk$xA&${}3
zf1*xfKi{qCSHIv>9GDN)yX1$~h3G;1CUlO3_9ZA^LVgOxI}|7IC2MwHx^%`m9pooa
z{<FUPAg-+IyHUu`&0{x+4z!OZ%Cm{;3H2|#@+<3zXZw7we}AN|Vi)`x2leBNUesgW
z6S=2QXVEwR(B>cXq8}al@I2)n)K(9vAIL`!>}g$f6*=u&{D+^26Yt;bC*;Mi`&)98
z%YK3%^sVQ-jgEQ@9pl-R9PFTez@L8i-t=ZZxzObg>Qi+mddBf@`=x&L?59wDj~;!C
zUvbV}^hZAWuwTm)>RR<V|J2U<l)a(z5)>EEIl*)HNzSj{Z>w*i_3^15<zM0leQ`^F
z^tsOWZn=G{<!g8Di{u%0%=dcijgI=5f4YA)j$g4m{gK0W+w|dkZFZzz$R5x=22_u$
z|9$_aen!XnMZN7D<J`l~(WMvT$i@Homw1BgBY*OHdLRe=Y11FHK0m>Sb)6IF760VZ
zFK?NT9y`(xdytPj(7BRa{E6Kle;^;a=+LkDWfy$WA9?U+Udrb_%yay+Ytk1t)~9EB
zB_9;m^eDbOXFv0egY>LzJ#x_xe(_Ha)}tSOM4#xeFMXH~t%II*<W1*ndW6>X+&V2Z
z9<rD3@LPy4h`xQBKcFig$>-)l=U?%}KJ;%seR{9qUVwd}`xkzX9{by8=#9SFhrjv%
zB%!>L$Zyz_-2BA8fp7N*?ghjldCX@=bhM3^H?-L=`H_FD=e=nQQ_u9KpFI4UezozV
z9`t@!+rGt5)I;<Ft;;U#lqhZ>J3xFVdhQ+w(ko;S$R051ng^}ZLU91;585Xa-7~em
z6W6ca7GL&z^`^Ly_rx)KibHjWJgY9#uWn-ph`xO$QC`EJc~HF3kGRF3=kg@GKy;j!
z)Klb?ztorLc`l!j+kE5X8AvbcJ@vM8zjLp;&w0&!NU!Pd<VF|&=->;#;)^`=Y#rxV
z=M(&Su3jaNdkpsyo)1Fuv=Co|P@T}i<b$4uMR__=e6&!Uv@r9;WZUe^PVCI@=-0lf
zjc;xDLj1;f$RGF*|8?)kA3Wzr^v~bi8+mSB_8<qe4*wG;o<rwqazpFl-+X$4^e->d
zkNC0<d&(pBBkSXnUfIWfEq?9q)<YK`=E-CB8~mDQU%-!j-2UoZYClUq_Wzta-=Vkb
z;s;-I*0(qByqA$b`GbCZ(33WM(!V?}@6xaDSe#SHjSuyg^O$|beUyDb-jkoz3+g-L
za(`h5b&=m)c|XOkAw7`Kdi>4%F8iGO5r|Lvq6ck!=Kez-`Zv%1$8Q!j{%!<%{Kq|x
zdocW~!_))rb@;P&-BXcA99V~b*i+oQ*CwZV_~!5Y2A}xgC+aDF!p@K#l5c*4Kl<RG
z^k_f9AH9$Vo#cx?+-vZ6dBM8;S^Z_7qZj(+@AM+y@?&~pKl;Z%J&7~#tMNxZa;RtI
zaeVPRe6c5a__zBTh;R1AkA8VeUi6$@`M2lvqu+S<11&WE4~5>xLi&eq=+3(1+NWoY
z|6bBKb-TYK;dh$ioxj;1_@Vv8zGgqww(qi=I*nbS_i)~Od5^=M-n)r=@x)J@C-@zI
z<(K@$IOluxjpt|95r^UlJ^JA9{14I-zb3!@Y9Chzc+W_0#>?N<)s`<HxwZK<^c-I;
zBoBEZy7-`fdV=%<`2);##SyfxLHP<Fke#%Rlh5Q)bi_TsRmb3mAJLQX<kD|l^XP-!
z>4hHYK^z!|4*xKYzgeGN_zmO-kRH$_XX+vS%6#(SAHV!G<*^?6=^yl~4wE<OjU41B
zH$I^J0@ah&hsHy4LHdBkLHt8}LGvK~Aw8%Qq4<OD!T2Tqtdsg-e|+*2erdcsP0!wo
zxG$Ce)tUAy@=O%1kAM3oKKbwF{om=+lXdZnFMdg0awVd}A6n?%G12-h9HfWjC)@Ng
zERr{oURtOwX(9b4^0OA*Vf)2(3;Dw!bnludA3*UUep<-hDR-;g(%(e;Q*W*H2K9@l
zi6T8gd0?`b@?=}wv77f1ei!AupK<2-BinD-+xemyo!{FWyZV}!e)5L3x(__P@Dt0g
zn%d#py6z3s)qb~W{@rJ+`OXzb&F$<vaPb+-K7QY;W_9X4e971HGJR{azq%sj<&XSd
zn_uD|zx-1EMPI&g9w0A2k?+tW7rk1~eD?SKsrv+RkB)VGF9PKu>o4k`<MEF^dcz{U
z(igeu71FnJ0aS-Tbr)1O%g5-bThuksJ_^0xf#^W|+t-~N=%<D31PAHMILO}iNA)9l
zA-xy}-5069>4SdAVZ8UD=o&}QkX_VE=4)FQx<A!6AG#;OC$t{<)EDIN92y7d3-UWC
z4<w2YD2@{OZ6dpg^F)8QNIXIL0(w8%!lWyY!rXu5QHZYf<VPsKdvB+%aj#;W`&{?F
z<k#Oq^WpFP-KPHE39??Ib=XgvJn|!CKl{(w7vFo-JI|io`Mqe`@x+sky?^_4XLpI#
z^WFX+bZ$v(>4kr`F!eybM0#jp%B#(9?2DN<wZ)x0YQM6NxSw=S=(&47`?C2^f1>Yq
z6TL5Tp8|bnD9=FoLECv(eu6p9c=^e9;>LS!p7$~6o98(`-A6<BU-Blp+R*pV=(t~_
z5A@tGd2XI{%!lTuT;{p2w@*O(fHrx_pJ*N5a~scY+W4S1^sQ$<;4l2v{Qy7Yf8y7E
z?z!=hpPHxN{lD+l%y*A&efLW02>vZ@#F6pjNpycnfAq#4>_#uft6Su4e9#vppZyhI
z#+wg~rw{tHkEw^~MLsZI9i-oUZF;vJK8zzLJsIzL&eKNU?}g}1eW2b!pZ`I6X9s@i
zxw!CLeDP=BYg-4h1OFf|dq8&4HXo|r$VCougU`&Le{OB(8~YN!wBPa*`yqSqYwugd
zt$5(q&V%9sns1-Bo;bCx@%qi<$LgGX&i~2h+(tff7-zoq@UI^o^Yo*qZr6`bbet<a
zH_!U=HMt==+P-gRFR1?YeLKDT{~pM%Up?;}>Ymhk&v)c!^hU4v<{#bzcz^C(X5WF%
zf5zFb<QH`8bJpVz(7Nn}AMe%l%Wv$5FL}*#?{SPHxApb&Gv9%TD}31>z4zk>_9gLa
zKlh$o{fmz8FvyYXySI|x`L}&Q+j{2Hll6@gH{OqvTR!q0-916C|NANO`MtUN20i-$
zy^r1X?d3oC*EyYgdT1SV=~upT-pM@3e(w2<ckksKV!yRt+5gy?{q@7lALLMX$z$pt
z^z760&93IFuh9`-;z6B(4}9}m^Q=qH&e`~|Pg`Fc;!oW1Bk$*()1AZJcj7DS67gz1
z-xJB_`0@L4>+m=Chx{GC^zYm%|FgS%t8F}gVt;Y#x&4ej{M}Z5BX6^>b;O1HPjRbX
zUedPCb6xKo{na<0^RtH^8(~L!@cSe2O#k9ezx@h|NA%E7^xQrtZry{}KeeIvhV(D5
z$}9M$Pbg2K>psYQ^sK|K@)+bF>^)I*kIPRX{Xu?|$dB+LFZ<nqe)71Fu`YSZBhJh>
z4jN~Ffc9m11|Pmh6KD1}<K<U*MI6f);>G@q?$WIf-0|jF>v!^oekgBPS3Z@u<O%w5
zE|WLWF`t~|BbRmLVR3AKlQ;N<zw2+Fb<88b{H(6QFaG6m^XS!nO@H<U^X+rDP1$+o
zMz5ILiErcZLtpl{MDnOZ<N^AUZ}AVULoe2W>I-t}mlx@W-N>V^;kTb1nRCJF>!x+u
z;#(Z!*Eo4V{h^NMf8HBHafgn)B45xqeefIc#-E_?Fwq?r$u}(W7kXvymOa?bey7d8
z>Na#Cdi>EouOI5?7ux(e^=N(L@h_f?Px;gn?8Of<pUQ9cH@~wM&+MX3$bM~n@hAQZ
z#ZR6GwehQNrGI$}og61F%)^gyo<nqu(~o~B-}sJ|oz$i5A#V8rdg2y6>+u72GhcqS
zFSwUN*Z!ftRTn!ixzDCI`-i-WK05pgeRk(B>@MHoAEL*u;)CAUjbG9mJ=%B3BX7tL
z_DOl&{T;pNcW>srOrLrGZ$0~#an=>L#^INqjFZ>s*>ie@=sRz)13uBqxaSx2Xn$ZQ
zbm-l4dczO;=0SQ*M2EkkC(odZKYmL8`uQdOkwe>f_nz#@f6;};JFjb-&kxae-pcqF
zkH(2pdgSl=p*W`x$o`(=Lwu*+=+At5!Kb*hpW@eZ{K#MG0Q|GB=i-AOh{Hs52cdmp
zqDZfU(C>s3^Sx}gogaop=K@F{{8!%6reAT-j)|V5$FJoBc2K_?$DjFy_1GQiPeiAM
z{9zCdvMawHgh^-6x}HDR=l)vnkD~!~Xrg?{zvR_Kb&Yruht6x_NZlq6+K1(7^TaKb
z*X`f(fc;UP)o=es7kZ9=ZG8Kkl{_z>yWgP?=P`Ac`0||rJv!Gy>-75f&+dQdH|rq!
zv!1*I{oPG`y8n{bjf47~2b>qR(a}#1?W{ZHJ@%LH$YVY<4x$UyKm5h{2GT#o?;w=V
z`4i+%Ei?|Q4<I>V(oxqW8khdg{~&$ye|~Nrxzt_qHh%dH{i>t+A-d>6af!~b$Zjq4
zyEQ2O)VJCYfAWR;Gf^E1&9kqQ8y)h|8+!C<p7`dM=s^8Y_I6Ln@A-#1T>gQ^CDKc)
z4rK2?7Sc}(-KP&ie%iv+%d_2PuVJyb^00j${^i`xJ+}8u?r~CH=XiRuAE>MSzZ>dY
z`C8sHkABq2>TCLNujjtrxy8MrwtG738}I#_-?zH2B%k@@^Lqk$%J;9{XIjVmGVe=`
zqepu8|DE|=t@-%(9crTQQRxGc3w=mm_Cx!aebBwDeszYn_d4cl(}R7S9kuP}?!}B}
zUntM#e(Ijn{>NX?ckg6>kw^8*6XbyM47%nIi}I@Z+VW`bf7u?izu1>r*y2;(O>EK0
ze)||b4vVdP-RkEjiR{xt{MvV*xK0#zEgX;c3*Wofhc9|^WZ1TS)4s~S_DSfueb+us
z4s{j(f%a$nEWdJ3>APLu>q7nN9rM-S^r~MRK=CH;^6x})xDUmz{T`Cv{R27tJ#2iF
z+j!smx<^9~T2K9h5B^KP>_T4OO|YN&+Ux{9ch1qq7rM~jy<i7&(U<$({JxaDz6Yfj
z>uRG*kC6RZ@h1Pvv+T(q<O}n)oiq58=jO?;&X4l9{KOvUnV*>a)01<i{Sf`hVy=tt
z7Lwn-V?X0x?Ak(lkPrAnB7YMn=<##@ZJszW4~h@p_iFPW{=|Ph=eOwOeDe~~F&^?~
z=N^AQ%irnPf1?YxKjOG)<NsfRzhnQlC7*uJd!{`(>N)gx>HYmSe=i>TJNy2wJbM0~
zyuTZgbo@UK{tiC=ytg68tp5LB;9iAY`GNQlU;LHd>*wcCeDFW}E&ujhKYHR%{baoT
z)cIGv#*fv@@&P`f_u=kUjK6Z<c~_4dGN%)d==J*dI@wPi`-t`MiEnkQHo3$p{+$b~
zXMOA1w~R-}IC40zWZtsAe)G`5H@Tb(tVcfd*+D&lPyNPgTUWn2UOnSnfKChX+d^_$
zhkf`9xyhAiUGM1=*#}zB_j=@Tj)Lm5Jm<99&TY=4<L9%@F5mY24?o_l<JiY}lpX0?
zKYHl%Kl%|b#%q&<UVU$89lv9QzPqFs=QYSq__i-3;_K|i{fc<Su`3Q+ws6V%OEy?C
z^`k5HU$JD`Z~k32Z`pdwHe530BP*6|{G0a6mrPwUb??WIIC;ggi%!|?j*;o=>B7=L
zX`nPv8u$~{z@kI9+~(U`Ts>0wyL$K&HFt?s8Ym5v21)~^fzm)}pfpe#C=HYbN&}^V
z(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v21)~^fzm)}pfpe#
zC=HYbN&}^V(m-jTG*B8S4U`5-1C!Li>Mw2e(3)kBj6PR**A@S8`ss&1Fgj73q<A&0
zG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v21)~^fzm)}pfpe#C=HYb
zN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v21)~^fzm)}
zpfpe#C=HYbN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v
z21)~^fzm)}pfpe#C=HYbN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5w
zX`nPv8Ym5v21)~^fzm)}pfpe#C=HYbN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=
zlm<!zrGe5wX`nPv8Ym5v21)~^fzrT~CyzYvD_@%XtI=Wcpe;AK_`BymG5UPLGGS?;
zG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v21)~^fzm)}pfpe#C=HYb
zN&}^V(m-jTG*B9ts0Qx%(fnJ#{H<F@2H_rioOkYe3s;R4KDT<9s6A`^#58cm7xsDK
z%3W8F7Cyi9Ffl9E=+Z!Gpfpe#C=HYbN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=
zlm<!zrGe5wX`nPv8Ym5v21)~iHSq3loY(u%lr^J;rHAr^-^Rh<$f*&ffzkjCocZ)G
z-gdxx_m55%(J3qqlm>>^z~BAvmZ$9a|LlEvyie8o|2dotWuC868M+!SB}qxvz8c68
z;g&L7Ly1U8qe_JaTqOx5Ls3M^P$IHTNkxStGvS<L9uAIq`2Fsy-sgMoqvx)*r_X2a
zbDn?B`#fu{&)#dl_Zpt{JkMJD<PlrYA0Et^my*JOFdz&F1HynXAPfit!hkR!3<v|l
zfG{8o2m``^Fdz&F1HynXAPfit!hkR!3<v|lfG{8o2m``^Fdz)Xhk>58JLSLfmRWXo
z#rUu)cNGSN0bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&
zKo}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV
z2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801841Bt*+HbF2yjlqg4}P=*Cv8v4
z2G78mF89u!cHwbbP<Y6W?I8FbBzF)7gn^vGK*L^D?|-lD78PsEDg4TF3InlVpn1)C
z*LVD9h6*djf&;mSFdz)%YzF=|ZTxFRroOENo$-*fTa7%uF!0|Dymju#KbsZXVTT3(
zyLKW81HynXAPfit!hkR!3<v|lfG{8o2m``^Fdz&F1HynXAPfit!hkR!3<v|lfG{8o
z2m``^Fdz&F1H!<6GSH&pKR5j2<Vsu68xQ}<sr)Yt2m``^Fdz&F1HynXAPfit!hkR!
z3<v|^WZ>lz6I<l&IY$XblZSA&Dyfw)APfit!hkR!3<v|lfG{8o#DamZpP$yJ&cHpk
zDgN~)HLQN=gLY6@d|bEAW4G?M{lXbz-W~bRuXoxx6=Ttc<Q~F+Fdz&F1HynXAPfY-
zK<bpSCk8B=rGSm6R;+aGu@qZ}-P+t!p?>Eyn?rhj+lY+??ySCjmn|3#9)f6Va!p}C
z7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4P
zgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&
zKo}4PgaKhd7!U@80bxKG5C((+VL%uV2801&AnXie950*eUwMD9DgL^;?ZnTAercOx
z>%V;VdGgX#$}QGP`RCTX6?Q78*md!}bI*+4tvEc>y-MSm<8~-hys=L0q7x48Ri-%X
z@5O%GS^9u-ic8e$uLjOPsyH0JHFZPH=T9i7cticYwWn4%txU1#rw^{E^5j9qVfVHL
zb}Y(wPPxUml9nH7KKZCJ#p-9)|6@@9v&s~wEV}jCldqjqg3;t5?Bk=<PZ$segaKhd
z7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4P
zgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&
zKo}4PgaKhd7!U@80bxKG$W9Er^!bPL@2-<>`;FO&BY75KKo}4PgaKhd7!U@80bxKG
z5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@8
z0bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd
z7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4P
zgaKh7UJTr^_qMAauK0@*j5ZJPYCrM5?@fJYkG`wv3R_Tkkai&bK*j-Q<3PN}Z@IrP
zAPfit!hkR!3<v|lfG{8o2m``^Fdz&F1HynXAPfit!ay__DAx3c%4dokv9lXSgMF!`
zFdz&F1HynXAPfit!hkR!3<v|lfG{8o2m{%Nfp@C^@hn517}~zmUz1BDmrPE4ZD`Y>
z$+`ahw^QLx#X6NtPIzT#r_%qHuS`x%PQ3e6$5)4T8q>e-T9rHdFcI*Rbi6)!z^vW}
zl_@SC@nY4r9d;-V-+6b%;NI_VP^Osnx8j9===-f5RSfv`$mMf515Z``YE`L>O*X}`
z!;?mCE}miMbj;z+Lmpok5C((+VL%vg#lYxWc0S*&Z+<IY=&I84BVj-o5C((+VZb2+
zoSo^`2M;x_G<c_-Q;_39(to!{I&4-_To@1rgaKhd7!U@sF#|VMYgY7)$1bv*Vm78k
z9xX%+JaV>O_a>!ID&Vr~pO|yed)pL{`Vk>-ys4o&e!<4`{EWy4b{cWtnSKp2Rp*Dg
zZZEg(EE_jq>i?(Zbn6dpeRlKuU74lcJ(y$GZyXO%PY+R7lBx&;!hkR!3}hn)pl?8b
zzp+E@YchXLv`ph4Q7)y=`b8J-K4SZYK~oI<O@~p}AF@mRgKiIUePKWt5C(*S|7M`e
z%yQqIeda<-P<V)CJNWNbD3UNB3<v|lfH3gq48VRN=*}M-6mNh2wh|KZKXXat9kWyI
z?2bDBU0CwzdkRavq#n`U4=(Rd-d~i@GeG;L<D<yJfG{8o2m``^69zc{c#lR)pK15-
zUYo;;x$F0Tcl~@t5&m}$aV`=4UDRZbmz<K0_*Lj5f}&Sv2faGzc8sw(>f;2RH}G6R
zW30~83hMoUH*S8U>4=9jMP2+D%}0Bu>0y(@5;unSjrS<%Z=7Y+_(XdM3ggX3P|p>$
z>pu0sP2K9WnyyUo+*fsaJU@P`3L5L)wsq3Fvg>UQ!^X4eyvjXK(CL^D-D0fy*69WG
zIk<TpdKH(?rJKG_({i!)zna!7YUQ|g81MPS>;0o<C(_lI_D@}ZMn0WaZkX~?O1&pC
zjL-GPbvWeudr;|Azo)!}MW*$HjQ4TT?bQpu)MUSHiuip{=yu$q;|Jn#ZY;C#p$)Z@
ztf2Az3eA6++a#A2G-7{QSkUeJlHdJre?`>!q5aWRj{0-w)cSkXld}8oquwqspT-(Z
z=Q;TIfUrw*`l2pfrw8Lcej(}$gx?vctB=$Cp!8x~Hz@s3uA8&YmCbJN+f)yda{e$7
z^?6t7pS{;f@_uv1`{hQnSo;r5@5j~GZTgz~Tc_52Znqs(bo<;<`5n&R6#V@hQt;sQ
z{RQp+4cqm-@`c|AZLgQ>$Hp(>7tgV_J7u@~p<g+jfAKp(VIT+w!qU%R$A$gXFVxwY
zx%{xi4{-I9bZ6hC7ruVyE4%yoV)cAcYfrJ_C8yjch`Cs<DGWq~fw1)RZhfW=S^Cp9
zJL(t~ebeuHgrz<E{akWA*}v{(|9a5<Lf1aCTU>|f^Kg2*)A+5NUf%Ea3xAKU-3FcS
z_jAehgaKiIXTa+^!Zbcv%KiRbta$Ojw}5EpAl_Q=t5u~kHfg^%8x&7r@ncCn!gYRo
zc7Gp!`AOq-9k260_xq^C^>X!>^z3;~RIJA%X-C<!9l3q4v2K51y`NkDOn>L}xpCci
z&BJov<N>pKA5<I<Zda*M(L?;Ir(2%5-=SZ`8oByNx-cLNM27*d<3>>Ss-UkMoi`*k
zjRgZ<`$1TKAGH5ht{<DYo>;W`DBmM2{Y_Voaa~`VbHPXHwxICvpY1@`ZsMIE^?2d1
z*55BXf?Vg%&t-|)oaT2wcl-RJ*ZHg0`TbtydcXUroSqLL<)eMB(fUVgJkaK`m3FK#
zEc&|FdEV`O5cKb)Uq-uMhV}gg-EOnX^_}|PX?bp+kYDLR+yCYI@r_3Y|Fm7jx_{Al
z23jAla=g}6y8U<HIj62G{2Hf%x(?n?R4`WU$SohS>W8NLY-u@sFEA`f>lam-_D|io
z$mNIi`LQoJUP$?ksvN8FA?o+x#^12McdouP-K(A#w%^m|%PBI|=cyd^{zvN{)_E!3
z_2SyQS9_}N_uPFYT^JAs&SwC61ID$meJ_C4=lmy-6k#C#44@w+O=w!@zK&;=pzsie
zcA(RjB>&Ij*o2p0d5QmsEj0*|0qH-%_8;L__-*&_UOQ^-AjmQ2Kd&7tf2iC?&ab(6
z_Yqs#L2TQ>fA)3xKmH8Be&`nC?XQsgA8K4_@J?IDL;PD>v}%yu=Y?p!jri5p^t<YQ
zm5bk#=8QX|U;F*EZs$+Gf9LmeMfG}q^+&nRC6#x~PPMZu4^eI3Qg2~E7!U@8foL&+
zcxS|&L>>LsU$fiwqSZR3w!(lgAPfit!hkRkW(IWro_OUya<*OfCZ$d)pj-VmUiFvz
zn)wq=c?xqMlDY{4!hkylg8JUC+xa4_zsqj+40#@Z&%2E^pYD3Q{7M)Q2801&Ko}4P
zgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VIb-Zlxu%Q
z+vo4eXRVk&{rcYq<+C7n&OKsKnd14ZuwvACm--6>QD6Y)stCsRJUx8B4Dqj>(tcSq
zI`xpEh<vYz?~g_WqtJJxj>3R2APoE`1F?<^*WC|w^E!V7^?P0YB|qBx7GnMWf@-Hh
zUl-@g0J-}uAMs+<wH<aSid=b`j_>{Ay*$x=?7kNsNU?Po7QHbn-;0)dB#qo$Ji`_q
z!p#HvH5~`2jPW4o_gVW)`o`vqx7$F!?+1NHFlR9Udx}@=G~&K9{TgH{ApM@QTN(Wg
z{Y)@gdTp$FleCLywTq~}FTdI|#<f@>)-O7Y>U)uT=ZyZ3bGBlQZqI+=^eVsly|nwB
z(e7jQo}IQ!*xh1PzwB}!@SI(q$Ln*zPHKA2o|N<V+`^p7C)ynl_O)2^;n%u?yN=iX
zDZl^UdIs8$;8&|kWo)vA2jRibcA#5-arR}b6S;h(LmvZ1ZM@^geMuJvgn{TVfbj>2
z^|e=w4g*qCVL%uV2801&AbJc$ZQT&o`bO`~OO1sAVL%x0mjT$PfLIs$g<&24gRU1B
z2Nm>vas62Pm*RCl_yJ7YEBF!o-~J}y#~DBVj>>X1VL%uV2801&AZIWD{Y~@^XL^U<
zeBnJ%^!q-3U9aYk^+qrFFpniDJjAjc#A;ObtDnT{Iz1XKeWu;Rd+o3x#|P{Q*$u<S
zKUbev^((oLFdz&F1HynXAPfit!hkR!3?w9ExAVs3ojPNH+Wmr(oz*Jmmgejfl5S=v
z&$%r}o;*wpM7_TvKF})$HJ|9N+wQYd9sj;$jfyv-P3r0!>wLe@dyV(?F1o#X!Izrs
zw}VE!Pg9InoMgQ38}xhd>-t{(K7NfartL|}2?I_Th`Jsk^*3AhIOXZ9=7W#Ccguwq
z(Cc{sVgJ2TeoV{P?7e)=gnaoe!T9pvWK5K@{xcA-{!gF8>pH|#Z>IMZSr`xogaKhd
z7!U@c&wyWcot$=^SiO19UT(T>i&r_Fz0p+vlyc!buRK=mU+yCe2m`Taz^gwz*1wPH
zJQ=U~Io9{(-cPL4@!a@Mtxivo{AiuWkM|oE#5ww0JLk^RLV8%9L+TNXe-gaLeuj5;
zN9y;W^E;j+R_LtuQd$@g2801&Ko}4PgaKhd7!U@8f#@(m*IObd517^apfcrQ`G^;*
zuI;cxadsF<Uop7%`y0;VC*D0oXHN*5e>mQJ$^HK~1Na_l)cQN@!5r%PiR1@szZHJ}
zoZqPRv#`{c>NT`p@osPbdz6*j?9KrE;z8?QvEtF``B1T{S6J@jl%Jr>=YQss$~$JK
z+F?cR9N?hR(N2QqU9P`=(Z#!u*lu~id-4l|YENOiZdm)dX}#jjr(f;O^g4b$FYZV6
z-=Jg~=MpO!b_uTtzBu&jz0m8p9S`Mq-sThGPx$kqkGO3=rt@CB%XuBYgSN9^9UXQI
zng_qGPv<3_Y^q0iwR5=bXR-R-QxDwKtxl`y$}4hwbNYzW^q|HQzj=0B&$_p5owTm(
zdYePP>lv%-c^!wn{@(5TaQeI4N0udZPdu&Sg?^2rezyy5T#)o|en$@P#qV|y@9S~>
zIo|d1dVe}TQvYGlGRJFB>0bHr>-W%K{N@+ejTidGPgt)LFJ5)`4J%*SrC#9QFFi@F
z^XG9tVd2H=y502mVYT0c<vp0zBPe~r?RB%uxF^q%GyMdZ?_z~mH*y%YcIdZWkL#I^
zhg><l&#3V!^$o{3#Pv^I`p@}}ZMgoL<5~NJxb{iYqx#;_U!5ZQrzz@qk7|BY-#e{$
zRQZ>Bhugj`?<d;thmI3&<*>B_&i|ZK)5G?DboHU_&{PkJS2<eWsLE9D6*=7Y;h^4+
zX<V6GyJWn7ryGCbm2Wz3#GAjM?jJASgTAlnd-bav=0U%%C)Wvw-7CAl2eey?v3_oP
zf4}9R=Ix-a8*BdMelp&~V_#5M+Z~OMgsj_dN$vhYjSq6&-`9zFFR?=Ueq$}mdv|6Z
z@Y*hfZ)bc{zQQ`bOTB*AccNyW=j;Y<-)B(I@3uZs`5o=&*-ehhd&p_^<jymVRk~m8
zJl1++yzdvQ_k;V4e0Khx>Xr8?zmS`EBt0DQ7+&@Kpns2c=@tDNf1F;&>p2fze;<_p
z%k;WByN|B@)cmpD=mj6<vD_jze#38!`$P7MxSm@KO0PD(E;lZlzHhEvPSdyKd2@ze
z0RA?&{T;v4jgzMN+<2GW+9Nkkg(Y3bcUbC)_U9J8>Sb<!&$aXHn(kE()cv0F=2Xu$
zEsyU*g%x!?A>Xh1yVW;M`R2;Gy>G04hxt{|83XjaJ3TkS^gIUBa?WZWp3=H`k?Li1
z9Oe28eLsH5Z!iCPQd-hVh3oxvcEAVYtL@GDejjAN#xL$VkDP7Sy-BH)ip!5xI{4u9
z4De)%c_;j>^uqB&Z0>qAePrj6&lWzAp=f%nWZE8Mb>HlEAJg^`bh+@1j8YR}Ko}4P
zgn=+Kpqr1Jt_#9e-gNyT<qkEjG<c^SYaZg=Uc%frrEWX}x^YYLy~P&`f5P7neZRHC
z+74hp(%Ii=K7Pj`KjK6)2lgGx17yU_z)p$sVSifGKVJCN_zUNJu{y6s@y0R>AKFkm
z$?E)2*X`xDoy|I*kF)b4owJ99wf_ZOFSJ7r!5fD`*H5l5^P`jbG3?{1)XyCQrg4+*
zt}egQF~H3aNQWOWyN%a$-gEkV@wU5oy}zzqNWL&23<v|lfG{8o2m``^Fdz&F1HynX
z5Kji2uG9T0AJ1p@a&ND8f}noy<(cHS!hkR!3<v`W3EX)Ql730$9kWwyQy!!(2?N4_
zFdz(gV}SZqIT`U<eyz{sIucLhB%UZL?ZsPfjLxqSZ^FeDBb^&ha%wtw6HFd3tM|dI
za~S;MBRV}uYHGp&#)(+()06wfXMbc=dG9i_+;?Z6xzGv<vA@C;G5-Q}<16|ZMd<y&
zsOVu_J)_#ba%#P~_ozz`ny<0LlSXbXo?!=tbe|XcCHf%{=|J!TIbMi!rYNG`PSHeH
zkTSvm82G;5gxPZ*-=cu=>IZV)6@z=fzrlE4Xsq}`I}sEHauNg34{|C76(1$n{c2UI
zj7@gDdB{n9IXgaUrxEv^>DM4rYZr!|;C3A2&aFkd+v`T<ciFM+$TJB8u$$2HePD0U
z^-nsk;CDG4<JGcY$Aey>dw-JeEv^K73r55L6|a6N_Z0?&0bw902Do_~dVmhO`K$Js
z^o`9IZ?}QC4tGDK2lbqCT{8wo-?H=hZhiAxg6?=QV>$bm;>HC@|8sp8EduLlr-=9?
ztgF2u>LZvv8Nl}>ql%{c7f~*IzD0Q|VL%uV2801&Kp4ml449t#E#<aOT32?x9Tpz4
zL)(*Q5C(*SKW88+cB?<Hy`%~Q!7zaDV8jZ8c?!8|c4xq=Jv`pOcl&-)`mA4c@$Msb
z*bx5Scq9DM9ENq?56ikIsCu|<@A5lgAj}MK^9bUe>3H}=`?338cp$~M{XSvlM(QRE
z2m``^Fdz&F1HynXAPfit!hkR!41fXbJJuo2Z*&?*jq><C;;P-E?mS2?AN6pHy83eW
z<<hC$0`Egm7!U@80bxKG5C-DM0G*%1B2#~c$llHk5&ncflkfS&ZzPgC3j@M{Fdz&F
z1HynXAPmHl0mPLF*6;uB`uWO~hghw%oZ9!{_i*~VsFjBwg4!{V9u(^KvC#TZ8SjUt
zLxx`ovQxzMX+6F25zjG7?kx<sWx#3wk@A23ognb<bsUl32lKsM;Wb!ZgB};qzXXK=
zVL%uV284m^%s^1z>ovV@cIGyg&(^hAt+cNB57~aB>3a<8{(Gs{pPyqLFJ9t(zC|qi
zq?3Ea%8s2~?}PKBog(@{Sa=EQoDS}~VR>$;hs4X7#LMa0iIXumCzXx&yqi<+pL8Nu
zg_>vcTNGFHyy?*;hcC1^%t>wY-_Od~&0+7=^?S*e`NNO-L)wwFqbRo{jF&phZtXm$
z>K9I*JJ#==TX%*n9r0SR<|n7zFQ@6V+2wgTz1%cCr?hW*PSG2qtT#&g_11p5bIHA?
z!#^ST=YCA5`mdDt_d5gGMbGj3Ibz+;{I0)TFWL;my8Zcmzj*5-um=EZ{#bAHf)Ddp
z*q?`U`yitnWYhIgtjc-SL*(~gttyqV$+z!2gq=@}e%QYl@BZlY{@r)Zz2KeqcG%qS
z@O*+fg8{$iC%K-iBf{8obnP;Xb(T8Cj)AcG-E{SVUhNm^?q`~hIAcz~kaTbM)Ufl1
zbsoi7t><+2@j9aD;zwwH*xy%ntzVBuOP^`?@LpR`c#wABe?Ra#j>_-N7<i}pALWKV
zF|>WBzb2PVE|r}4+R&y$lXLz1Z>PeYighZPobbxfPNn}XUzwbkoOt)Cj;{{wG^T&u
zwJLYQhD{YtSNVFq9W)*vzi#S)9~RiU$Etg9%fe;Hx7a!?|Lpw<!?!Q8V}%6@zq_``
znuWFwU$1dR>yoekYMbJkTGut|Tj3?!6p!EYb-y3a&a^pfa4bEoPy1E2DUQla**~Z1
z4>qvY_73TXDt@GNXj)&C)4lhg^G~k$KG%+OyKN4GzCU;UHtjwxn7><!&EY>@{%P@-
zBe&T))b$sd&($YZ>Amm0yTrmTQf<G`uX-Q2qiA~5<BOCI3(TJQ*umPr*c_Ix*lE~P
zcdf94!jC4-O`E=Ky3HZ_Gte)7a>{k;?l1I8m#demSmAY#FG{U@*+H8_ue_&qyl_L?
z!@KNQ<Lehzx$=Tr(z53Ly5H^?G<WiT+bND3liF|O&^2~caoXm|PZ#~i4m;lHwH?Lk
z_ermPQtIkg_t+H6bbo2gi<8%??1G~{_;TUt_O_kfkhT-LzR>BHb#Fhnf8gTv3iy|@
zgKM-Yx<u)a-q$brk#|P*xwz0)6)(hd%<S5r{N2rxEVr0haODR*o1as_ghkW7oVn?o
z;;_~C6%PKG_pG8g^O?%i-y4^yfT$nEWj8l#+Of_V<rIJ1mcH)tsYeyCVWl-Sw+!E_
zbcpo^hqQiFuJY>TBk$b1+2-)sTDxlH>wQR>;@%G%cYEflT$Wel)-6c?$2FT<yp#Bw
z(qaDcS2UTi`h?P9)xq^{$e*0ca*IQU6{ys&)iR|+%Kw3S+bVvsWP`246syIP<t83f
z6zTP1CHE|U-<Z!!F14K^S3gd#r|EvpgT2qSn)E^0ZAyn+ech(lERwI+vsdi5<ArYP
z9roWX>G@#FgemK6w-~nPo0fn4r{zCbYsVU)f9SBm-D?_;JN>QAA=P*0<^AT9zd!b}
z(qYZ7OBZ=y;2vd)-2HXwbX?KN*Gy^r<R^W8%Uajb_M-c}U-|UDPG!uWI&^wnlyCf2
z`^yVV&#-l<TNgGs@m0mt>*m;|i2GOkxPH^aW%k)lafn@@cKrouwpSc=IQO&`ZC5JC
zNbJ(EV^Ksq<nWO_gZADsal37bPRAFzKBe++4L?qtF!z|E*mG>xgcqx(+NP-UCnQb(
zAi376jke!N`F1PA|D!{W|1y`f>p6eRCfhCQ?B}NWy7xfs?0(6kp4i)XV7+u3Sfl;5
z{rBeEYMY{|o&A?H4Q8xQ-^q^a7+2h4m6lgNeB+pXwpVm}|D5tWzt#n~&P#o6efZMd
zn-s+=>(zG`%vz^(XxeY%{eE=oAFug3JyMtNm7kk$sr<$h|FTt1M~=r0)35%nT-RN8
zRB*yo%O3n`dAc1nJUk#H?dWTV?d*zoZ0u9u`)e}npm0h3E@Otb-D$f;-FS`sdTkzD
zddI2L$`sRzmhE9ZutWKc4VLYn`gqB0Hiw-5q)_^Z`3)c6rJUkN3B3p1P-BYCAvb=|
zbWX24aLMld<DT7Zo8sKJ2M+wl`W-gK@5Zbty?OtBo1$sE(CJ}#o&!rx^y|B2hti?W
z{><gmcqPB=18%P~b6%l}<JXO_og(#f#Y^_OZuk0ow3}F=Q@gw0<#qO(u;$-VdED_=
zK00Vqr0-3qmvZ^^I$mW{y|rJ;x?)W)eoJx4)uT(9-gQQ7+Ni>gZtePE`JG?=fqRcM
zook<AO`q3!V9&u%{-m5DH{YyYIIqGJeUE3wUD5u|$=~cgmV0EU{Wh@Jq`~Fqm(8@7
zeBZ3j(t6A6$EGH>T7COryVu;icmDH%lUdhCd{;`Brjx2<X4pF9?BiU$X!@58XFT>$
z`-3(`E}qUbo!$p>!(QKgaIV?`TZg+=7hO3c@v!~($DS+sRJolt;#crqseZZT^6Tey
zf9_D0KNRn?-5bN&bQr%+Rk`4%(<Lw1qm~@%RpaR;X<6|xXvfq(3%|xSYwJy#{qu_R
z;#CX2`M3IcQnx7})sydAH#51yrTbJCem1XEu@xKCeFPJx_Gz4Z#dy2tj~!~f`P}OB
z@Tx0T-_!TScNK?-*T8jvn5XIQpx4p({|(jZUb*j;)e5*Y<!IW}_i|f6s(&7Cb7E?G
ztwRd9q{on9cWzmrvXHE^sQnu65r}#Ni*%e=r&G^w6p-qZcs`1FK48zwUTN2;-Y-gr
z)IMh_!_LPc<;_%1%Cq3<6QkECr%3&x)PL$$9(C#Iw9ZquYsb}TtrxBR?a@`rDRx;n
zDCOzfeo#8xQoUWJO0Q2*I>b7>Zr!TWu0A?lnW9rWCEBl3G<^?UN=-?qS@=uk6luSw
z^Sa2{n*pc&FD&I9`@7xx$>a39p}ViS=e8@qQXE2G;gH%B@630s@wrEjD&Ww)rw5Hc
zx;ZPJ5#u1@U|#>C_Po@w+Y}J<Ek)>G(1S3ZVE-KEgNpauzjI~p-HJn9FKhg8=RKz$
zI;h+t)<aJBH*obD^2jACD_7pGD4Nd0PWQd&;)|T-<NdhB8iOj&JALm;J1C^>$Z0*P
zzDH$h57o(>z0Pes2xlj-N*5k=?VDR|ugJCA39B}om{L7e0iQg6?`wOe?pMISEy_Rq
z_KJsX=;0fx7k{|XKbG4%q;WZja{!jf=u~Iwki9mL#-Y%Cd^E1(Jih)|A7h;iJsya4
zB;GsP58|bv7ebH3{!sWoXnZ>Luj9TztRsO)2cGx)KjKHZ_vCjvo;#}OwLgUA_p}|v
zO6KDCOw*mtF9$}g9F)@bC)+L3^(iO+FtAs<_2myLQ{=A0rBi*!RK_@@L(ZPVrDI&8
zb{SI{`$ct#`{=OEji0SOUVgi6ifKb~73pDZuuYM+J37z!C1YO7E{OTVG%nY)oNm10
z_UqI1uYP{}>aEpID;=7~RR^{2nwu}5yZ*9*n-1l&DC(Z4#l;QI)!B1Unc~sr*Y=(B
zc`l2?$6m=_`P*X|$`oriD_whTdZyyA<I*;1_oXLV9M10Ex?i_4c`T=h_82SV^!4nP
z&he=mw`u;6-YwGJUzK4~{9)$f!*)NGVROjc-)p+*d}dmXtB-CT<bJ11*R3a`l26-7
zc9V7bFW$51J^&ggr<+%~d5iAP#k$lKp?91150s;Extv~yeG*(;EM0GL<!HLDKkD*P
zFAnK@Y&)oI6E$_0twZ?RO_7UB<kDfEG)1?1FZVu?UZ{5etGgE1YXj#ezTIcS&V#m7
zM86UIodLRz$u6?4z2g1fFs$pzyH{Uead^?f{geJaH@8KxaF2<{)vN^T(?#`0H!qi|
z9{S~~;(dB#s9lr(GHTox8Ch}Azf5Rx*Vhfxv(Bx7-vRddo(*0Y_RpbPv-}UR`$4~_
z>vY7CaQ;{96UFaf&qtgE#$VVA5QhT0D*VjU4^w93GbwK#-=%=Mcm(9ra~izL-LJgh
zwWP%bt+x(0nDWV>3$1cDJd<#^*^#Vu3;f@BZagRS8t?`?Mv=#gr#=4UU>kN8*!On~
zx#~d67IW0Rd6UL8{`>4Ke;51;;1}}=<_nCY*msBV9`;t)lTp7mzkT!A+Sh(lXcwp-
z&Y^>S8TMQ38=X`8vf_mrZp}JJ4|ZhKAMqcE7dkL9@t&STepNuUGt?jcC&Z;!XuNUV
zrvuV$#Ph=s2R?zYW9ty@*(rh_AoXLR9|OUwDWbn}m=*u@M?4gGML$G4!E>QqpuZuW
z2=52=LOyV4@`yg|P9<cmr?0dY|Lgfqrxe;%ywR;c$*<qx{YD+B|Hy9{<AYa3{|*W<
z-uR7CZ66p1qKeQTvI}~(C&arQ!f)(1I(;taLmWas3JPIAfPSY#=s!B7Jm~bFSmo=^
zsR(O6)-8S^^zB$9^meaE+ZSZh{R=3k+owSDz3yA4`!FDz`pb0Xa6O8M^Lyx2`T9fV
z9kSgbcy@|(T_Uo(`2Vo-=CnOX`PedGy6@ez9PBNzLK=4%l<b#%H~Finqt0|XoOLb`
z;-_IB0~SoXWcpuwr=7R|jP5tW?>W47$Ba)Kughoog>>CX*UP5*sN3^PIlnd1b@pr1
zcFdK7AB97m{)_8U{Q<J>d<p!XL&V9NB3EC|F9tuRj%V(=u<vr{blv4tPc`+kaeiE{
z>GV7jD(mc8Scf>}iI(TaN!VL;NaMda+3kJkI7emfen{7?SH0$Q&jr5&hlsa}?JUpV
z&9py6tsI?Scn18<F^%Vta{v2$s+`J?*ZCzXzt_cS#X28$kf71+XoI*1x2QXxhn}+l
znbw!a!#S1V$JHV96`dcAu16pvZozNV^)F5@M>_Ok?4vV9`X2m}sogMa=S%9&!HHGA
zUwj9pM{(<^c%}Qr7vfi`eSJi~mnLs*ooSbA)4fLS5B?8Ri;Qkxw!`-G&JoR*Ikn&m
zLyy?l-`?WRx;Hew>5vUXKAm@{9SiZl;FYe|Xgn(XSBM`&+!#fAj?k9OoBOw|nxTNY
z_%H6<0HnWhs>FtrGAET+<m^mHce|e-zk~gYL)xzJd^$8;H=Fw55f8>8;`k5;M$zl~
z6ThePFVZ=r{6$?3YCH|<y0BAb7u4ASP4nXwr$^%}FkZVw>R;vTs8LJj?A>0|ssA8W
z@_pA%__FS-19rR-dVy2)%6q(je`NCOBepz~*IM3pYW2^DC1fGuiEn&x^6+A3QdLs5
z*`Ma$yHkyrdSrj;F(<R)y>TuY;{FhKhPYYiY1m%`eFFXql*{+SthJBUJFI|RZhNZd
z#H;r!4o%0Yes|XR?yVsw9P3-U^(D=Z_x!Cp-<;-SUs6zrICE3PIiMVx&R-}ua^lqs
z_7=a$!g)Nv4XtXmKC?1C>l{*?yLPt3y2V#5IF!}S(cX19;(^znT2bVPn)mJIJ37xO
zX&pTF%XK{$7qx~rnw-{mey%LU`CoK=S=g^<ol3J#D`2BM&32Y~^unxt;se^3{Jh6E
z2NdcJzXcHc<lk#Dul~ocBv};4=Na8<*F_gtz;j*hxV3xZ6N*FZ1JGgdUz0i<ZFy30
zXgaQSs61@o(sRcYhq`_>qjSx9SHF?V;_&Xz)4Q$Ql-uGE=cfW!4_~tH!5fb$9iDt{
z-H0U}b}9}{&l%<Fg><aPbV$EbjP^tk=kRif^%94)eL)^DY}R9!KAB)8FYj!R$p3%)
z#n6t>ud&|=h<zg6Ha<D#qF0Y+^*{8>*ZbezKc(3UrNfqw&wTooXOCs=L&G|ldyZJ8
zQ@aEwQ@<+qSphL$a;R%>)XqWWX|K;Im@zw5aftC#huk_G>AL=`%cpi7x3ca&PV@bW
z`<3fB`(9{v;mc`zb6TFpw?M`^nLEcG>9CW)Zp$He&J*X)gIyU2{S8Qe2mcMwwEv^L
z>AC^suny+Vx4}BV?fMn*W}Kg$yRK<m8TWgn<J^7@vA>o>*yU+^MEs8qVZWz{btC+U
z^!or*#y(bz2hg(%6n=MYku?jmzWac9p)9+S=6rhkT=A08AEJsh4k&1ujvHZ@P1}{}
zcZ5viq;Xy=km{jS#y)n6+&ai<`mht-d+sk?$hy5r&y>|^`K|jOTGF@VLrGZ({}Nqa
z_AR@hZ_6e}vc5Bn_-Bj<7>D3j!npGB$~JpmQ@@?}U0=)>KyLiw(m6dnsC1kM!6D{X
z9b(+|in@6j?H&F=Al@5?cz>}%?woiooxa~#?FZuYuCS*=|M45qzJT=lT>GZ&4sk+2
zdL1h3_6_LXhpyky`%xMCC=m8_Ao`gOIXeZ;)#31h*RQzZo?ZtOhd3vO&L5DeT_Wo|
znLo}w=I%@F7B5YyF|umS0}80K=jrSx$Va>3#(m^dq~ijW@tiv3o*!|11^&{q;_D4A
zv~Iqma{Y!Yav7uBeMucJTh@Bi6ZtG1&bzGi%C04HTRMb(gLN0a(*~S5?y?<|4j#2R
zI~45rK*a0N_W?hZDdM~#9pXLc5YI;u@l|KliW4^%I$|Rp3j2XE@4#=3xGBWnVVnY@
z91uJLb@~AEaeW<f>tx62<lI!mnNfVU!`5zPemtBNKL$Q@sEe<Jy~8Vhcyn93Qfj8X
z;Qn(b_dRti3$fmIx-Nk|#w}iPeaeUe&GK0*`~Q04@a0L?qymd8R2`PjT65FjdWSAa
zw9s#0Z>Mn?h?7OTpyM^{_CTDwh4ZG+FQ6A;JR6hn%&_q<Ct3q)mh087`(_0_ALq|e
zKXCc&)f#0q&QLgi4&MO;KJme>LLXJwndLXY_rY+!73u-MHR7zXPQbbW>4@*jvKy4Y
zZ_MW<m)clo;diiOz-|wF1D$WNFA4Ji^mf>RFkT@J5`J3PTQHxa{-|G;UF3r=7oKi!
z+jvg!L=o?SBI*UyjguH}qKeQHb%^#)k<*uuPWd&JQ6JbNP0{W6Kz~QiCxzXRL&^&$
zLqDa6{^S>;9|7rnOJ&5T=n(HkhxdH*>w|R<A5}o~U!du}2;F*DxBr$qm!3<<c!_b)
zEmD5A?(5dF#<)aFhtP8<B0eZ<96z{a;j-gfZ1gkCvly?|zIocdX>wkR!wKJ%Ut8?$
z3`G&=q5|o2=wzM#;Wgi>-r#q6`usSrheO@GK(7b6Rq>Zr6mOeoQQSJNSg(Bz&MNq^
zIHc=&$lSPublo^h$1N(;{XBI3_bQ{FUXj`lq9RY4_IUB8b&FZ1i1jJ`o(H}MjdM<+
zZvmTqS>VRfo$^{0Qo8-MP^}^s+9ii>_XR>v4hp&Y`jxJ;Yf1ipf2YqcKDhB#mwxN&
z=XMoacqZ#RVmNmih;;>ry7S?1ZXt)5pD98wGDW}IuV2@p>*c7)-1`bT9dWL1k<-_8
zdMoT56fq9Zy>(>y8>}>?LtTGxn~(TuzuKqnI?$utqVD>-b&5_ef}Ki-yAxk6xx4Oh
z#UZ}GY>IRp0{aaR_8C*;zEei+GddY|G96N0IsF!PJsrXyqQfQM^-1XXY>HY^y~@8T
zd~rf;*wAF&vONnG>>scbBEA{@AN*{nR`<$%udG(M4(x%5Ck4V@2`n^c?7-SrtW`RE
zw$`p%`FbBxbiUE~;gAn|pAKOkq=-0vw}|KUin{fg+x%&xPF()*!-s9)r~B2aj-8L%
zK<>He`=BycA1*Eq?I>vE>cyqQFXI+<<84^-59a-JL)GP{6wWh)-3$ApF)tM9II&Kr
zp5G`s|G<7h*V**>I6DAM&$5$ertF_n^#>c{JN%D8*zr020&#}Gp!FoL*Qet;)(P-O
zxkbt+;;T5s_h0-%_(S|cw8wbkdA$D7Uj0gk9W`DEzhKmn`UTy}bUbq^yY-tmy-!f(
zb^RCmHivHO1-o!iNY|lU`=Ik4CsVtLU$QQ~oX)>g*4ee_Jj}_6yLF1X^Fp|F57M35
z)8QZYijL!B>;kpxFG#c9w)dcZhj^T*BJ~GTSr^|K_585gE8X^!pnj*DuaJ*8Ac}}f
zpon-N9n$ORWb}J(-Hr5W-(TCh|CFP4p5irH-!?ywMNt>0iglCUh;tM;<jx~N{3M+p
zXnYIZR}wavwi8p?bUfC@MS9J@bmqn>cU^nh#<~-BAhb)wfdifTqY*zxQD>iXn!oeF
z)B&j-OIowLx9-=iOdhN1;CeUYPtIk{e5UgB_r_&rofinZKlUeb<ItV?jx|2_=uvg-
zvyrC~>*Tk9ush)UyuIGKcSfTEClt;*KsvC;3+lwXkLI$#C+>&1Vh#}x%pr{n<K#g<
zj~zaG%@Lc%&6>6w#5;4Si<?2bEuDwC`XU|i)*QnA#G&c)ar`4)*PmfW<mPGJzREkR
zALv*3>ti;M_7m>;kdFQ0I@FDuG=5OGZqSVbI{vwJhTrMB^$Lxbrn2t$$j5qxBGp$R
z({&n^b@pf7x`_5SPDXtw>e`Ff_M!VdcYlnd9KsL8p{_sro$qxW>i6$){!h@zy`Qk8
zJJl~t%j5aFl-v4olb4HH6xXG9y6MH63t2kk;v1k>aELfz4!L=Xrqgi@GG~8>zr`(b
z=U>ru8h7ee)~!3R4&mYzXgc&G=+T_r4&$uXaR&1%)$7py@O+4WpmtQ$1Bi7dMd}}i
zT@!voAlfU|yO<Y&&}(q60eIpN{yyjfs3#EZQHLDQ;05|P5c6>7x+fp%bK`ECBI*Og
zxKOv-v5M~(-K$_<!1`8))c<#$ys5(JDqpX+slVN+jPpKqXzX`N=soC$8dGelFF{7T
z@QUD<BB#$FUAHe1`DjnEM*2LaGWs_~s=u4cx^;stpX*;VosJ`%OrHxf#u2{|{9yg2
z#d_N+ez9bO&FN!x>sFn1_0j1{hurVorh^ZRTW%5KpAPXpDC)))(|p9AU3JCkd-}fk
zt^%6+qv`oLkg>0TBI2ucsEg;~`Y+;YOp#uP(<5j)7l)#oC$O)AL*4v}d|m(1*->yE
zirl#lG#z$5ityK(>ZROwm7uriQ0E^&KKiXwG_~JTesprye&s**DeGkDd8U{)Bv+9h
z)&@H$<n~4A(sljXYg`ZZF;S%BCo50w(>V8v@ix~kW0w9$1-WrXm#&*vVy%Cf-Vgd@
zWbxI@N8Y)2vyF9I)R3E(a$361-WycD&JG>+eAD?dR^`}nKHv;a|K|LQG~MmGg8Lop
zYZS3ArwF|e_7xq{_&&(+`vP@(4wp~$Z^*DK>5yChA|3Xxpb+(q73$U@Uh`>tr83t0
z6sg@iR<hf;e*BJp-w|@LNrTJJFPmutah?;9wr6@S0{psMzePH?o-$phbLCKPr}~!8
zpJ+P1k2;)tT8p+T6^A%~ilT1#)}2#F^B;L<RG*6rZB;ttzF&M^JHBRfi+2)#Q{4VQ
z@UF98arxMv%OS>FzYzNl-6H+|l1|3Cm8N<YXAj|i|8}Wozj&kZUfUEApF)xIr|8mk
zaXiS^wezs$I~_-ymS482()3&3D`~x2r^~w!=e^Ln>~Ei~{BCKIb?Ly=XKvnhQl+Jr
zuYFyCUlsO8fJfM4peG>i3HEmQfuIjyyoFzvelO>~`+xd0<Hqk5_21C(-l@!;>*+Kd
z{xYYi8`rt~X&o!I%=dkQ^_Ry!Z+GdYsmkd(W#!sxnV(cEVo~gKL)+T_`l^%#d~3vs
zQOj>GVNs;tdE#W)C%JwDd!bjP<Ah%_S8t@BHxIr1)8a2jZnI6(xp5TnZ$TmI4}|^N
zwB9IZYWL@!GpKZ~-_UeCzbU$%-?86U@S1yWyYegLl?STV2Q8bn55LM`9@HT{XM@UA
z?`CD@`=`ZT9I^NF<%boAeGe>cl$>-}xkak~({p#J+_2J`np=kNRi>D*Xxf)EH=R>n
z$5E<}=w!MMr?S)e8hR6ly6Zvz(jj+!n(lUf<>KABbgn<q`h;EP&IyT1I_xG<#h~X0
zuh)nD))YB?n@j(3Tl%`oryf;0?6|Z|+I{JXS>FTRbN``1*R43KID9_k)?pvdm}*nR
zIV)}v=a8AA&d+U{&#m7OZyFSG;~dhVZ|e~2ZHn+O&~z$eJ+4E#4(DWD{0GN3O{epx
zPR9NUltX+jknW$L^9SOI;a9>smLm1D)8~bqgW6qm?TB8-RMy4saQQSY%T(s<eMqN#
zQ5o%<;!nTUEZMeAX{$q-ln2vxmCC}_g-6!v{a#TE2s<_6_P5t6+dpaEd>e@G5dcd+
z_IjV{i+-^&E+Kvn`Se^nocE?f8pjWR1jn~kH}6WdaivYy&+A*xYqVzbHk%^iuQ_}?
zWpa<-PUf?8*z=k)yZ_q1u*Ko7Nq-qN?u!h?A?L4q@zdG=%6HoyTZha0PObj=ump=k
z*r7SRZ{5t~3YYFv93sAwqUkwo7&kcNzH`L+9ca2Pt^@W}Al5JVemO;)CyDb7b;#+P
zoS%XAOOeCDfK$IV{Hj12zs||<yE^q>>*BqGViEq^pivj^7uI}g{|t+a_CUu=zhwBS
z{6d{wj?2e$3JL>G7{L09>WlE30<k|0i2VZ4cY)BGf%rZ3!(qK-iZo7|KBI2^8uff?
z7cKMN>2DuNO1Ir2*6pT<b&%iaRNn|od8hFJrsZinkC$xvy>py9Zi?Wc=kgnS9=z)t
z`@H9?*lE~Pcdf8#+!?ipK-T#uXur|fM_``;QhNopGf<h^_k#U2+`8Rudc}M0-?_5)
zZiRE_DRS%Gw4!BuSP$$_6iw}!^mmYP?mZCxJsRHz`xoUGGPkbAyhD-do2KIw_DSlH
z+H+vnq5fuSUp#MqyL(OJai_n{`ranp2ZrZDoHm`e=s5$_u2170FkkmhdFaWvmj9wC
zQad>1ou0!xCGRU!caI)x1F??^?VCf62gGv$X`E!mE85OjSZ$pG()vP%o<Wh`m$nng
zRR4j@*`-XskA-+P4ym8|$>aCFwrA>o1?0{vnO>vJ!RFic8OH<Z#ywp=c!~<@o+s#h
zQ@gs?deHHKlQCa9MQ*)`bj*`cMH<&)D&ssuQ^a`AAsxp}W$t}*>AL;;Tt2r>pzAuP
zGS^Q`?QDn-c8mMF4JvSJldY=m?dSFnT)aN(JKNBMu#Roh?&E^_yQSEeuTS)9Jz#8)
zRGT9Fa&+E-tUGTO`Sf|X?@ypUnCB@{{rZ<HZ`--M`g$7(eVjw^;udlKDR)jhmrnJ7
z>@NQ>uvfeF<qs<8^YmQcq-wK2&A<2me+T}H+VfJ!Zd2%|&`W{PV`0~UjQBjX59l|D
z&qaU3e1vrP>!IgC@5ObYAECcMuLn|lJ*OWc?hVg@bsNT4v}5Q?_>LCt54{=Rm&SeR
zc#HWD<>@+s`tRvFl+JVD8HoA<u|B1GIrzoCa_GN6)Pwr<sK12nC%}6~zoK!)7{^SJ
zdtWpi??;E=g`#PFP!9a)5bfM8q8*tc9k(IVbsN$tQeLQxb^}Cv2GV#`*vqi42GV{(
zWxDPI??B9pD2IMW5&1y0OCW8Z=tn@O=WSwKaEjD!!?k0KgLIwQ^`~XC(&pTtfRs;8
zrtN~$vuNCgPR2L_q~jQsX+02ULy^WCL56;#L+E?pgWB;;Wy}v@McS_2%FyrPg|KHq
z&xO4U`WWIn;7`SQJ;(=QT%hw6WNN2`9t6zy!>qNB);p|#3##N@vgxX$3P|^%Zppm4
zf7_}VS?AY7hFy{(^eSM7%EJaOJ$Fo*BI>~*jjy3{$)|ss*RO7pr9+JS(64}0kKx)2
z^eY`=e5U#bmAQ2$(oNTcRIlM==%YG>U7q6s>Dv00zi-UvC70UZgG0=VI;8VGCu6;b
zd7VT0KB&yiCp6tyKRWSM#nkKO*rxUgl!M+u5$%^ls&DIL?m1}xfDC<!@(X>1;`$4I
zN}YN8ZwiR_gnB_AeWcW@<r`+Eoae{Fyo>W);Wr0j9nR?=^!;)&orkfH9rHEp6sRx8
zc{~T=x6zL=?x5X3rtgimXR5zKrur5qqy2D*b(~|oWcU7Y&+fK$_H3+k-6Gv*Mr9fw
z?o?j!kFTyN|KtH%hxA+*Dx2zEu_~t>mw)U~<IU$*+wm%g`6as`9WUI<XdkdAc|~sj
zliU7Ce}{b>7#BID^Mg)CJUz~tfnC=tf?rd-=bK+2tb6#V0uCBKvHcqZ4lCfpy>lfM
ze&ehH)^1k1_T2PL1%y5fe<wZP2mX?*d2mc>zmY@N*nZ1+Ufq5Pr};WN0p??hruJ~m
zcTqu|{u-5htT&^EuzN)np+^Rd7!QI%Q+uzj9H*ai>C~?1R_65Spwc=0+-o}gO{Pfg
zNPfwWy^_E3x5qM+Qw*woy1g!^Pemm?s^<it-eI*@o7M~Gutg2AKg}z;?RP={&g(jh
z{vO|hU3A5*Bc_keV*!UIkLc6xRDy+l$@Dx@e76*Kl@a}3n!L4jrp@89o0~Q5Sm%ra
z;`_5e#6eR2bmJAZOF5O{H#0@*C)3H?{7=(ir*n$B`_p`GpC*@Hr0b$~lMWTJbm)|K
zS|0HeI;77N7MZ@!sK~GjJ4M=0{gzGjz^IgieO=Jm`Idh31^-J_5O&L`qSN^-tmWYk
z%c<!1_)f3q)(^q`&gpzc%Oh^UDVo~D>GQ{1rq3TWncD56CeyfKud=Q^IrgL9Q8c~j
z@kJ^uzvKG7r0<{lc*$*c&^#dCI98}@&$@iX<2XgcznCH&H+6P+8g~R4`yD7c^%uc!
z8x(SJSX}zFHbYC*z2!&S6uIkg>4@7lMO}O$&FA8^xOAub18qM#8S!nV=(IlQFI@jZ
zd>fGS1I@j4WceGcH02kbZhh}v3AOTA9CGz_o6g0HL?u1w_$jCDiuNm~GS~l+?)1Gl
zwKM4YCC&GHKGyl&={^rCJLSi(@>n0y{XU4>2h#Xb#AD<A0_k&M92~McsmJ|=*4RLL
zJ;;dH(Z%=C^E`ESYVLiTp6iF_Fh!^Fq_lik?Jha3Ub=oA_I#(~XspU}`k7mO!0YdV
z(#Pq!KCzO6dLF;78<luQx^E_GGB@u<CEawLAJ%eI@AN8r9p~_ST^v39<tPv2;u4Tf
z^_W=8h%=(+qGP`q@S#)X>kpZC$maBD`aN$>9`y6r;iK0av1uGvzRalwUl@ACrikw~
z?OI)Q<&4C`HsU3y-45r00IB_iTi4?piiRgP=lZ$921OC)N6_ohb5ByN7EhL&cu)a*
zHh5v!KZkBr9OAqgQ>5o5>g3c@Z5n?x>xkkI?UO^NdJTOaoQ(I$A^H(T@ah!lJ{p~@
zvrp>AP1un*%sPLj|Ly%#nyonR94F*cyQyC?XGg&~8Fb%Iy!3UhU3yKA$~>;CCyhTd
zmA&#8)bFwHH>wD`P}Gse*SeLtc7phFQ>6M3)f=d+vp4C^i{|!WanIv59r{pCL%Ofq
zsSG_Y*67zd+-tkR@3D?BMNS`$cRIco<`uyUhe5?5a`%gBI?liH3lE*!Iwg1A0#@tJ
zd+%zoKX2APjjj8-wX88N(JJ=DoR;lh$YuT3@Z-b@bC0Q3#a~)cyltXYX+y`$mn{Cj
z?{Ov+EIj;djoj9IP3G1A_?0AUY0A;Gsqf{sI$pM{^{6NES+||c)9TIUxvihpU-D^Y
zrMy;?61Q!9{pB<2*ZtQ&JZbS6^<kk-14;}(!lM0D-v(cNnBM|6%F}FTnMW_QIDF^B
zq*D|6=e3^d+ri#^^a2)(|C-d{Xv>p|!x^1x&b#`JT$T=X`#58j@6=yP%X8}x({!v?
z+#)?6z^!b$e)p>!=Lewa^xQC0nbTi!KAKy+@3VQOimlk7bcl2EgGR)Km}1cTG)(J(
za<EG|MVyDuq0{T3yi@xGcb<hV?iJ_uxkb*tqq7Uqe0olzuDw#e-TJjSJEkt3@{My+
zfI7X4jxRcydygFdNax-w_gzoj`G8*YF>YhrgPx3YRWNUW|0kEXZo1*YO^UV;Zhh0R
z(wdrEhVNCTxcQw5uf9;@B1?xj&&MfN>EF0klles~idSzPlzw633oH()p3B95JGCp*
z^3)#5`6G}{>jjysA5Ax1Uvl<rE*<sr8%^u0EjRsx<XWpX+BpA~BG<pYrgP_MMkO8R
z5&DhrgF-*Xyo+%i=iSor6XO>2WW<T-_KVZ`3jKy+c3)4sJ=ezcca%>~#(M+m;>xK1
z%&Uy^|DED_{os%9lTd!pK2yJ*a{R;(zbWY3u;(Ja7VAjx0KZ4tkX%K2SQ~6y7jc#F
z*U@z`^{?p07u;Wmh|6@0-#4qXwB9ls&kMx!c}2vD0uj%E{h;6zNZSkL3+JzzBJPL#
zQ{?#9JvXk)A$<-iyPbc*o9=uMr~L)xY5PIEuv65<>*~f4-8^BMPkDDMqdx<wK1gL<
z`$0b9qv`mI_v|;)`<Tjz$MK3<{qMC;N?rZx9@}d<zx}zUe#5Yrd$z;YZe@NvY=636
zt?JnM=y~`;iILxw`tG2u!*Xr9*U0@rrX4Hf>Z?oVz7GjI2=rKrrt!0|^K*#rK2SW@
z<&ImsH$I^_bUNM!Ro--bV9QPI(>V8v@ivWP3!CiJju^J`Ue~|z`aSk1#R~nJ7ufc$
z$-sOcHuO5L)8A`oh*`Ey56x-$F(Y1XBn<oq1Jpj~xBMTs`mcYfUEQk;zY7rSLBH|8
z9^(HC#a<k-_w(h4RdzvLJWN>g>3gAacAHPUuEW94mg?;)ReF7j3JP%_9fnQsigkS`
zKiG#Qc>TQY&mFRLJW&3e%CUOix_b33yP$8&CP(aeqpqG&%cuToQ#q*T<o1KbI~{R1
zQAK=TKHeBqyYcI~oPR7T>GKoc?lWQMK|59me>8_)`z_87!hTOvM4Sgj8lR|>b^0H@
z{(17S0+sr;TBh8dLyb7rSfS1?@0Z<G=dbXqKRdlnQ2GwedkhMLiZ8--sh<$`ZyM);
z_JRG;Xt&tsj_-@$d%8fByJl^@Nwa@mVI!^u@m$zni8u<n-v)6jhzCIXLjAaTo1Sxv
z_*AEe=LWic&xqRz3b}Jix#vY3oeuH-oTA@;HEv(3E`4~T$!UG(=dw6F(W~`<u{~04
z4t3A*<F@p5mrp&aIPAEzP1=3w|F?e&GWJW2ZJt_aPvLwPaNmhJJLY|r&!SjlX^9@M
z-kI0p@R#S?mzZ=eMFG)HX`Bg-M}yyd-n>a;8vlK^LY(=iOHZeDp0Zs{SheBAl<KKj
zi0`H1yF$Pp2KH*VzWhN2nIh(Q^jnHp|Izx<{W2R`)oOiaWx4|5yg?xS?i<~gaNaz=
z`<i=hyYefAaRukA0I5HT`dcxc;r{pz5|H}+Q6A`YKP~1x?B@fT#v#%1p2~E;IhA$$
z9xCr@UgE~}M-+!2CG;M2Lyajmhe`Q{eA%Mw&o+l#oE&%w8cp{p_+4&9gC&jjpDbqe
zd2?EFr;7?&-(U0KKaX8`p*8sZ%{M%FF1PjZ^xNxwTQ|X?y!X%CUt;hZN7aHVd6#Uu
z>Zr<J{)#3uR-aJM*4kAoU++U|$)R2~o?eosXuLDzp!tmI^`Q?qMX!DzzkiSKzj?)2
zzehSQJC)t~{jl#KXpA+VLEUfK>vIZb%uZDltBtOC<+E)w74WN{-@bZlwbRNJyWI9v
z&xu#>R}^*Ug6aI0X+_KSupZc<bcpY?>k#uZho3I0H@bPbOvNGAJsjTsd3v{%n{rzm
z7Va_exSEw<afmp34mo?%+BZ+zH%-oKnIhGjAan7GNXPy|4j*}ERG*6rZB-oNyabB)
zKBf+J=Zb8qGqYgL2}u@*RDY#1=3gD+ya!;j7IXf3ONpYESJbWlqL$C;hkmDf9WTAw
zPq}qHO~?A}&V0ujpL_JE(xGnrr1|^14JvSJldZ}W(eF*s>Hg*2iLaL2UH7;$#RDS~
z@98<@SJm!~VQo5$->1c%m%Y-iQN3SOPDiS5J)d&xu#aa<wSlnjpND^}x4CsVm(JO@
zxOA`mll~rd$f%=NJwH~z*Tp}$&8P2G=O?21rq56B6KmP)_!R5<()(c_U{sOL!%>r?
z;uq)Ysq+hjHQ#SNHD1@#`Pajq|L9MzOdi?&0?RAXam}ghb)A6U)BWb03_plp2s?6A
zk?yAsN_N^VP(D_1-dX2LWOS-Cb;w@ZR1XTO9G)M@#dqlTt;7F@>$pXn8=*t^>2&)v
zIC~$`5jUYj#8vo>y757`zY%sdw|vvzg?&D8T0h-;jaq(C<ECHNrEv$QGG~8?SGr$(
z(Cg6t7FC(rt(?k!J-2TB4?4fsTldarRNzEbTma%!aQ-yT#Y9{=>>#kuQNOn?u8p3b
z#I5W7;x{bcT?3CeUljIce0PcSL&Lv}xLw4{AZ`l&I>h5ZMm(mD$E0erKh3{)r-~Ko
z>f<#(R(>ke^)zRfpy@O|Pvpi=e$wYRoAQ9TW4}?Sult?vwViXnr{81aWSsZq6rIkK
z+&SvHbltjw<|B^XDf+!0a(X@PTwGnc={`Qc%5m+DOXvK$^js?{<2-4KxKGohPSut_
zQ{1{FrR|x*EemDgT?thxq-@P&oo;>aT?w`FSPT30tW#;$X*M34o9{5+a!BJgI2rYI
zi$VDpX?>jfBk22r%&lvk+WV%pUbObNM^`CR<m^32=hknC>y8(~-sBeD?l%j|@96j)
z71>np3wt^E1EPu*8gE?p>40=QYN*qD!k$mpS9Ctnt^alTbl#7O%=ytw(<zTS8T%&V
zjk<dJm0$AdpXT+en`8k8?yI!pq1<^bAij_06pzg8a8;G>k1LA0_;BQ#KA+!qVAK0L
zUEk@-)9;c~dDo=Bj2ibvhH{JaJEk}vm_t(@qE?RXe>RmTO?$j})4Ij1k$10~R%CvO
zENqhh;pAJY70CKtV81(SeD~In6AE@o=-XHi;5<3lS@8V~*e!L~^q*tD`DI;#rNb+(
zPZ?35Sw3s)xMIEbH8`tGQRi3W@=f<IzkJgjtIEHS$I>C};7-wLdvtp~zti8jolkV@
zDBbq~{EEXuKcVYBq;tr9mnzoj)GmxT0S-C;371aWovA$!b~RnRt!W$z{hk$->3gE}
z=49%Rf=s_dMrAth(e_IHppa=?iQDrDac=w*$JRab?cCkADRTZTtRqa{FUke&Z<Ol~
z+I!2y?Ka1Qu763s*|~_|DX6$Jx$f`re;9{xj)f^!>DRb*TDk9(4)I+Kr-*Zibm&&^
zIBz}o!Iuk9x3_I--=N>w(8-9eqzFBnL#O)Z&mV8DHT3>%who=zr_bXhGiCprsz2CV
zI{1g32XQ7kJh0?MzrI^`D9*o$bi`Ze5bfBhz8tUerh1msa^Qht$Hl!GEx5dx#bN8h
zBWv}3uc&qZLreOWd?<-UdR@O{zt6+)>v~j=^-Ffj8_t0;MflJBM#PVKMVza|A<o^w
z`B0{a^R7&huGb(VZiVwRBOU%zuLyrJ=ZB@~x_J@#V-lVjHvZ*Ai$i+OmQKdGVevwo
z&*c?i|J7l|d+y)4viELfiu9Z^oeaC3DVpjDZvDhg_nRVK(G<CJR=9IbkY4xO9fRgh
z-fsiZKZ8Qk=kdE7ofkNJ9<`78C1V@{zWzn+d8uQ!si>f-eVCS`=Nxh}{hdyx@6)Y}
zcwxT~^B0GhKRBf4HR<Fk>(zG`%vz_sBAo}_@=1T^x1JRA^>p_7?4FPJhxI3iIImQP
zQ(sf(e)#q`+Z3stfQuJ$J3pAl9iSY(8*Yji-*t%VJ4N^*bqF3gJ2;mPzkn&yxEGzw
zje9g5b`j1V%%z*o*ID%(+cn|Es;M^CCxZXY2mB>LA@wJ5GS^;dy3_p#@K;f!aTPzz
ze0<pM$1-dV={dnvp3%AHysO{HWtk!uKW&=+(5dqEhs--<n<Dl}afox$IIP{QbnUt6
znaV339*~iC^tHn_?4CgCm!z_89?I(PKX$0`=5woU$cxM0pIB~PDeJE0@4i>IdqL~w
zJ1W<2xFVND&zq#{s}!rnljSBJR1`-{J+i;_n3GC}^xSl;w=n-^q0irC_G{k*u9H>I
zKmO6Z2W~%~>H6C=E|=EJR8HyAbW)Yf3_H6ZH=af%-7kAHu4CE`oR&iz*f*<ZAG~7j
zImhQ8a(XTo-(s5XbUs9RZl4cL$NpRBYaG)0n%d>UQqHOUE^Ot4Ubk}l+H~pAw@v-y
zD5sk*Ozk;d_t$XWMd8v-?KE6JqUqc^$TS`CO_<MhXu7_letKiM-nNQgEZJb|*3ZaC
zyoqCbK+gwLCQMmp<Gdga>A4J$ssDIfhm>3`uQ;WE%ll5P{`s&3Yt2oA>m9lzF$*&b
zuKb{9^K*)#X?!o@Zs%QAdS%xVxh);ub}~<^H=F0SbU5OH*PmKZ<cQKC7jNR!FJgKg
zGRirfkKLBvwYuoa8HtDOs3Ki=#!7a}x7*)w=R3KLJK^Hu-KNJ%zsC5@;VqTN9e?Gc
zgLYU@_uUBYI|eizdH|gdb#l)JFAV$V(5=c8IeQ0}e(R*ZU*vl7w9=u@f6e7zX)XTO
z^PNs9Q>_2V{7(ioOR!9F_3$P89=!3G(jn*H(xqSax6f97w=~HzMW^F~PA?2wKGvmC
zN9bY2e@*IewB<>y9q??4b&IcBa7a;v{-i@{cY}Pm&55b$wGJsA>b@sp8i(UHKB-Xp
zi1`g4-=*9lo>zzTduM*hPWvU@4-#wHX}m5iZ@Ql1%F%r^PG#z!H<hV>Bx<tXb_=xI
zayLAaaJbnK<rK?|d?w}1<GYksg#Hq5G+vjFS~)1C?N4@C%j3M<s3NDoM=hOOmqjH#
zYJ8)<Ue}vkoDb&@M!I9W+xN}tEUmZ9INkz&5%^bhi1MZg`?FIt)z=YM%f$`QbgzC#
z+)rosF}0Hv9N)R+bAvN%rwBj34&Am}E`FD$rxqFAzHEo>wkcA-yiV5nyOGcJ6J2|T
zf6Y|iGi~=M$JxDg>D)egn(kE3r*?Intn*iJ`KEr1tbIv_{sQQiu|o9s?1ng}*>B|f
z8|E_|()Jy+9CSPKdwrE%uZQs)@evfcxF(wZPQ4j>eymZ{0{&dBf2G<p3R-^SvYVSV
z?O5lG8Z>@l`!@z0X7Q6Rn@`%2H<zWujEPUy_-4ih7Kbe#pZWAF&mL1w5$(It<ivHg
z%apJdTzbdolf_F~Uw`-g=`TMnU_E#JWd%1K%4OlZs@UJW{?3<fX@C6@wd3$R<$G1R
zz&iQdx)Dn{>{PJp(fxM!t(%!#;nIByh`2%yKgu=x+h%u`v`{~a@B^D7&a>AcjhBEu
zH7J}puTaJD>qgja<9n%JhLb6OR5sm*>el|i`D^@6Kh~_l-Ya$&vnX==L_1U-HgM^=
zV@ii~yzyJ6_oweCEb<Qnd$n6%{-6R5XkYU49^V{L6lwgfPNw@0{FY7kQ<#?X>wW0<
z*}2X4tA4uc(DSkUmfenvroRh|U!K+jywm;J&}XQh7v~~EA4a?<>?+XXu^z)Z-IOO?
zInEEq%@^1o)wBG4V?Hmr)JFUg_65L?L+vP7j{`BE)BOmDTLm9FG}R-%Y&hewhuR;s
zO%d<WEt<v|pd8Nefu3cGRPW(r-F{fF`Ezd_S^fqqO_`!wKM>j(MeNIWieCM?S$2;4
zUB(P=yVIt!>A4VY<14uHn@rPn{(j_F8(s6tXWM2fr{@vk`(LK`;_&C5eS6qxMRCvl
zhX!4@;;f>`y>DH=rTf;X?9~27%fn7+iipeAq0YWY^PTRK<;owpqiA~5<BOCjV!q-K
z^Mq409q(y5ya!#oaGJ0Cu8z}u!~;>J^};-=LtR`S%`bCFyPos6Y&!4UIn4VU((CAC
zoqZbXYo|!{C+JZ+)a?tW`CLDE<egD{E-tiHQPlZEkdOWH*w<)^NqH7LePZ+)r9<u<
z5t<HrkyC`erhET1pW~BD*V)THEYxX0iQz|-Q~X=f+<ga%?685l@rfHRX}a$GLFChZ
z&dH{JEtKnfU}>Y|q{GS-@m)xY35%wEIdjuFWr`SgImG(L6uEue98XB+=Aod{abF#B
z@yhW^r{e%8d)@yPw%_Z<&9LRuei^h(^;=G+@t>RwdkB!zE4Xx>zmewC@mMG8?8LCc
z0I9y|w~YJvg)_%pwqw%4quPCZbUgtXdZH=1<wG}q`kl|&y<(lt&6lR>I{i9U`SkrD
zZprQbF#OKco)GkYa=NeIscfo$P<vNUGU7b;Kc1ZH)k`k7bbb|6KZo1;V!YuH_jk&>
z*Yecv)NAL0Yu<V9oZ=96exUtS-v(cNnBQ`X)xN*Bb^j?xZ4PyDmiVqM#j<~Wy+GeO
zi576$>vIZb%uZDlb^DB*=5LsO^>^jE?y^ne(CPXbGRB+Mo%i0=Vt-!CDeB@rXg=0g
zrf7P;0{knc=k}nS&VI<{bLU=j={h@qyElfl=`enuGClVi<(%#-;o?xx&nTMqd+aZx
zei)s9TW6O*ewQ-6>x|g6QE^zN`%7b9oV-@)5b-@g*y|}e<&`T>+XKFfphMI1dfzxz
zVna%qlZr$1bEn9yW9hjJruH_j+@PPw4j;Ydh>h>I!{6BA&bl`=zUh!XqTfrCx3<pA
z@)IE)`uOuHw+{Pw##DP+o1vxZ-tyyls9Wdh_KQG&;*eW!A>B02UUz=p+<ATHKU8_0
zO%eWN9dhFs(m6d@cU?};;NAyK$NS`vTQ?&e?M8>#ze3UUoZsm+${cLIZJ$k%`t2d>
z+7<G3dWX(`sEdPzpN{Uw)Y<ouuUkLkIXIl3_;#NOI}h3vX&eYA<GQACdnjkR?}V1a
ze8VB)KDqIkiw_w(tU#rHt(GZM<m`Q>c3=1lsGpq1*-(G`OH*o$tXlJcf?nu1!r$Q)
z{jQhk^)OyW9nrr9|N9)`Sn)x}<EY9`&ld<wd8!8lEpy`w_0LiJ8@F$t`+k$@dYRfM
zAe*ieu-@hn`&h8f20}095OIAv#Puk0*X7n9x^&L&%gtY0{Dsah9_xIa-spEe*Z#ex
zL%%^BEY^F7KY>j9%SRLErcGZq-3G$Wf%Oj3fz&@n=YMXW=e*7Xdk%i`Cq?~Zi01{u
zE{pbv_?EjrPw%#JQ*Numi9Ox^b#S{X_~zg0=Sltl-}9+)MeCBU|7z=yTNi1^VZ-a=
z{`fvMhi>OL>}#fob~EIWOIB8{yj{7)UnaD;>+6Q;iXz$*hloEhMSKsB!`~WyoH$|b
zG36K1dBs!)ufrQnPU|~Amt~5kalxkVNoO~t`P9w~S=-Jpta9ZAx1`x9$F)=4z5trf
z%>%l0tS>pF?Uc$Gzc|#L3ykrULtQ)4U7zbObYBA}Cok`8kI0{{IK+6%q4vDB{e@oX
za`jTh>0L<Y-Vfq4IK+My`rQWI`D(~#?|Z0Mji;BSW$jDl?6*iqKj%=Vf6;u-E>F{`
ze?cd6?SM<?c-N)t&e23Z_Mey{ea||1<};P2zc(&ZQCv_Z?~+Yd9aTEy&Q+%VS59vA
zeT9QR<~^%`^m`;!-ktbr$=!92D;?_M4|LD<myK6nb4&UeMKOQ*E1Jw$eL|Tc{InGN
z-C5(iw}zZh98y0HWYc{)M`m`os>=7r6-Ca@NYk;OghQ?$$0}X-UE`%GN7JUhm)kN$
z_(OqooyW<sdS2Xz+K;oF%*~@=OUHW8D>_~GB0ekX$gSJFrlX&jqV9fJ&vU3dPg<AH
z#*zQw7lD6-BDYSY>83nz<#hf)E}y%e({x?Dk<<Qz@|>N*Y5ihVKIn6>bbdZ=of@lj
z#GjcW=jZ0^?z(Y+=5yEK(sh1+dVSsdp!sOmrpU!#aOtoUa!Bo+uy2^6Q+u?IPrvi&
z`w5%Oy_cw^`~BYN^}LSPx_Ilb=fh9t6}|Ep@89d<-@=kl{WwvRb@5!;HQ#CbHq~qB
zbxqF^iB~yJ&-OcAXAeTY(|w{;uhGd)_o31Encd{5jaTQ@_x5x92QFT(xP1(vPd~V2
z;j-gf?DL*0;^~NApy;*##3~*yR`;RnM3E;yUGyJ2?4WosUGIjyoLjpN_S3`*!{#U8
z>cf4P#BI9QagX~w^$Sy(?myt-IFOG0a@_Yav7f>%(l`N~O!t>Trsp*3u7muBCpPE$
zxxt3>_TwVnl|vd|WGXvd$8+acBi;1ATsclZMjVS*<l=vkj`sjWx+!w@8?JvK9s6T-
zi1<q#>edllJ!v|3KG%8aBNOlGIpkM$fI;e4(#fXZp~ZN>A$L9v(w)xZ*q^3DE)E;%
zy7^IOho|QUQkkApN@d;pnO={|*w3g#wC}LsvmLf}EA!)FJ7^60d84)-bIONld;<MW
zSay?j-wAV?f8c``Zn@y964qb)TSc2~Eo2@0Y~-oLI{B?i8#-RTWO1T}?`dG)CB|tw
zpWJu<PoHMo_`QO@gMC1BUV?tt@v>#DM?I0xqB!xhHdC*Cp@>D1wimbOSJQZ7$TY5Y
z=-$(V#vk2$-gyjke-WO?6lwfnQ1XD9<$AU2zF7fvanJZp1cx{uBWUFA&*}F_NBe+G
zk?P-0W%xav+F5os$o1a*ITu+vM8Dz?`lnk2KYk<CBcdjo`bkg@{OQg=;o`e#I@UK>
z&pO5QPOt8Ib;mwsiWrwok?W6KIz3ka>p%{5=MBT&p<Cx+pBu*?_8)M_truuIcm9Dc
zK40f2(*C~M>`(LW-Kntuf+EK+(y>2~qG>z8Jjo$_?~v*K2~LLIKoR{7$i<KA_F>R`
z-99FoubV&V`AbyBegqxr&abBY=*}TSztEwseNw%L%ADUl_xgr6ENZ<?VP7}qUG!@p
z+9BpYd~Xc)Eu24w?^n@%3pd`K@#NF34k^U7!rlje8tfA|zZCISh!aDc8uo(%ss4)V
z!2gYX)R-TDSg+7^G8fN>?*`EFs4o!jk=j*sc01@V9KxQ3>(l+hI$39@LjMJV?^q-5
zZ;JHXpQy>)xnErRWq<o@<#$Vytlw@;UAO4tQdX_^-=06DO@1ri53|-jTJNyJItlYA
z_VcZ|X>h$mmn2%)S9X2oxUu&BQwsYj@!d(ZFKQ=3`=b7eU6cMYYTOqY3W$AHIM*K6
z#XODc<2s1b0K&e9bD4m!pWu8i^nd)0&d&u3zq_``nuRt+r|lW@B+zsniE?ybIb@7a
z6ro3%w*R2ZVf^Qi*57YgHy<N^VZWYrD$P2rbja<GFZuLO^ZM0IvMyS<f70LQ=4SD$
zpWnWEYqisg!ymV$ue*HeQAM%G3+lwXkLI#;c){GIsi&(YSR8WSe}p}RL*2fA?tV1=
z;oiBD3cqnyQN(z{A*UCbrmyV(>xsjcCt2}EuD{W^V=BAdx5wR&OUL<ug5P~`uG#^c
z^1$&%(=p!w9iNwfEIqAH`&G8?dr-*d{5#xvLtHxaTMjw>mZn4hHbtx-IMn&iO!GOt
zh^BMT&!w9_r_=AU(sO*U9`TB_z3%>7?*=E|JYlyS_1UNIEIDrD90rQ{GN%@NVdxQ?
z!}IK?ZRUKk$~tKG`~%Tm-6Hfhidg3X>GSy|Pkia7jU`&`vpGb6*CF&;in?<MO!K+(
zN_FY<oET2##t)=({n%+b9e+6)-#O5sE{;xT_w}03)t}lm{c5jnuXCQABBN8CsYCYK
zu}&Y-yG7djt1@hyzs$u6?pj@R<&4C`whl2q=#ct3bmevVLGQ=K_3^%S=ba!Q>r5Tu
zypIl*hYehM?wI0`p0AJkgax-;e*L`e&mFSeqSNP~c7>?P)E*sG*;GH&m2>+3{VorC
zmfr~baCSq^9xmyPpZuiHZ?<1NI9=Cp<>~xQ=V7NZ^`lUk>fffaX}^ZvBKZ6FQnDKl
zrtQFKIqIJlxlqON>qgjq^WgNopnPhP(e2B2*ls(;`{!JLy5*q@Ew@PHJ2-jI{f7o!
zx8kgFi`@61Yd0%hdv1EBGDXwz)HL3c#wBnv;#qZQs?YxNeESlU&ZQ_0d%bn<j79}c
zC{yJ2JD`2}jfk6|$gM}<hvg9GOH$;{J@cB5^Ts)Z-H9Un#iq#VCAv5q(|r2fOvHWY
z5dL!xx$j&dojY&YX*z4y{C3Bnxs&(XutQMf{1aTd-|HLG>v88lo2G+r4$-f{k5hft
zw7lteXt{Ezr{CE1ibEB*oxRA4I_l0D3~N5;zlut_Zl9cKJ~!@!Jst6hLE|e!7JoM2
zmqL~fP4ykR4_qhf^hO#V?N+Ax3nxEYYges&y$`7|A8h<!$Mahh;$)ydB90W_wSzu@
z@fC4tK=>~)@8i5^*x%?n2YNo;mqqvQQW^1iK&Rsp>Sc=bd{@ND>d<Y!qw|wf|6;7l
zo1Xhj<KJ|8EB2Q-MY>M~GF>0|Ez{?N%zYPtrqlLKW%`|C+CHcpm3dEBPw10Qk@DeG
z*7-}k<~yyAuDos^KjN0@I#3tC#jcAq=VsCIl9TE9h`41S7pK9cV}5drnEyQE?rZM3
z?aHqdjenpr#wUuLJpk#LXWgPM-kYxH-O5(!!lSNzbE{1e=SOhJeb)`?bbO>T^bH+y
z?MOF&a{GpmPVF^em389=@@br|X}bd-u}03Gq#Z|Z9a;VcD@|eF7Kc<{jhC!@o}8Ay
zVQK2lSp`qqnD03JbiZ2FvGY;eFVw{!n&xA@<rMLqEFktd`Gwdo#bHiqXPo_$OV{aF
z$d46&^qf$)a;)kr_Ynq!0bzhL0Q+OmNbf81|7O7Ld=j?b(S1`+<^Ro6LIM|`5W4jG
zUB(P=yVHh$maeloneOk3w@mF!zR4eax$ty*+va#cI_$wZg#NBWoOh=~zt26xIVW74
z57O}+aEh=4n8qET9Cv<?({`ov_uzgygdLB5|Cag<>A7drj)rqy*zX_qx%J^ocW+Yo
z9xIUUTfctylEROVzR<$Ham1Zb{~@*8;Ji7UkAZV;DAG6+_yd7>4r+I$b`kUkAojJv
zkB93O>iBY-7arJT!wyW5v+L3C{6fZkIK;UdurC6sy&N*0OXvUdE8lb<uGezit8Xja
zw8S|De=*(%?10qIO5Y3J2dMMgn0{Z1?!$)6#YNEhms_86>3GjhQKye`_cKkeF{tvq
z)Az2lbqKqHQ$#=3A=(EJ`{U?7UYu`({sP3g)j%%p1LvPnesDf6ZC~tsa7WSfrpFg4
z+-F1e;tx0a$8sBp?_|OsLGOq0p3V!f(_y@zeh&C+xN~BVPW@xhw<zMgt+Q&yiJJ=@
zv4NC_sL72V?!4#JLkCr?5dMtEUddni+hZBZDT3cv?;G}dKey{FI-Yu!!CzF7n|GY1
zL+{d|E>0q-eBC*xZu6<0=v8*x9=-M_zkctw{`7Zl`!{}fwE4At=X{>a0{*)H`iCbi
zKBIuxACLBm`45Qo9fx#&rt1+JPe$V+sD8!Whf7DiA&{<XPMls)^H{@;Hju6txpt~M
zX8`%!`UB~1<4)*0lauMWH&oWyC(b(uXKdGm7ptb)G#%$Ix<&AX_KkQ_4(a_N)A!5C
zv|VyC^h1hX?MYPsgG|p$qcZnANC$sF-Fd5sYvcNlE*<rzNcnOqbL(it9Ru;)Zqe&|
z)cv07^E%n7-I?+Zxo_D8eOop;VtYmIoEk3OseXrb6h)n0Woq}pxDXir*rCRo&#ktp
zolh74MeQh5rq^>TQ~L~+(O%r5X}{C)LG!Wh)gkOcPW55@jzf%}*&VrYmP>ca2hKm%
zA=SHdGN%^=o$hoVi%NN?{U_GtsU8%x%=w?fmhN=lb7;$V9@um6lRqigM@><;ehE4s
z>*%N=_Wi{goz5d+Est~TqmHnTWEb>$PDi|d&wXb&Z0Ycmxy7LMwOC#Mj*Wc^e1A=b
z9W~_kOGGUl_5eCh&~<fGWv_Z0<pr|oxrx}%!mSIrbkpl|`|`q;4nGKxp2Nt=Iz3bu
z|7V&{<1ujF1Bc@}q~vOO#VN%hzHg?(<$b4C|9n`2<reWBG^ePGcSJmiDb`88yU<%L
z^H@%i`<{+29p8=9p^ks#Bd*OYB0kU*x&7i8UpO2$cv;OZkDs<<ja+@aro$f*RXowF
z^?<QGQthZ>*yahR^)a=hhqW9%=O?N%^+QBO*7+fkuk+`c=Ih4GsE&iQ-dsE|o}bg3
zOwX(R*X?h$nfz8k%PEe}GrHBTi!QKq*y7>_=j!Y^sGK4<ultog@3PV>yOzjp>5%)5
z9+!S+zGIEgJ$h7`BHm+EkZX^o>6{-N`!jTib&To#!d4FPE>T0j`dz%Q6IJ^Q>P_c|
zu**UFNn>?=?AHhjA}*y|o9;Doe~@W&`_pJT_OUp{={3q6Y`$%uZHlJ;d9UAJ)Wx$Q
zANr@?=(oKc^~3%*pwo31)hDoTC#;D5F+n5c#jD(<Oz%1)Hf>aXBNrEjbod)Mq;_vz
z91z_{rjxn2Uax$*{XO);sA9bR-fnrv?{Z&v>*(#rQf;ru`Eh8v)A<YKO~-4e@u*xp
zGSabc$Em(zI_^8&cT3Boex_&|XN~9AA>xARdv+>AAL7t!JUaJ#`h3ME4K6>wY^LoL
zx$i;Kbf>)N+CArQ*ZmIpPWubWbMqHXce*bD<x{#eom3?=!}bd?-g!m(KBFR^8+PmC
zFTH=#jurZCzoF+zL{%Qq@1@CGTW8v_La*)Juir=Yefd>yzj&1E2m}9%0jeK5l{tNa
zrqepOwTrurm!<covZ+53<q#jM``!qbZ{=Fi|Lbjs6uuXN^N$gKH{l<PM{F*;*5>9#
zTK}9PN5xOh)iYjp<pINHJ$C7n3D$GhUsiC_p<I@!pH}xh2WmfpUDFgz*O$8WYEEhY
zgWunL!-MB?TR9cu{hmi%dOEH1l<mqd#5@u&yzjGlrHZZCp!`PNzIvzmSmy_g^c)zc
zGW<{+((?zh<aQqy%-=1=#<_JIQa>8{L%jX%Uhn_No<VzWnYi7K3gZ1z`xTYp#|aAY
z{-c82cW-p*I3JjzQ#(6X-fw>-#tocHNs*pU<y6MGPC)LvQr!27bn#cXE{E_7a{flA
z`;@V+;xKERyz#TO$IEZGsZ8xIPUB_pzBuIK*|_uu$I{dKv|nXAMZ|kh47z{o+Cy0L
zo$liaYx$_gv7_ENr!T6=*;%5J&W)Q<NvHFpu}sci#yQMDdfv3(<0#?{1Ro!^`>_lg
zJiJr=52lAcF|>WBzb2PTE}fkC+R&y$lXLz1Z>PeYighZPobbxfPNn}XUzwbkoOt)C
zj;{{wG^T&uwJNvY?}@$MvAOos{YLFnYypQEMH5F~IN;gCDpvTfEic^JaOnY+-SEMO
zK04F&xjiaqbohFGFRpuUzTYs$7yK?|@bca}k0?{zGPv_6U$)J(z2e!vb#D3gt^>AH
ztUopRnP1lLvvoLR<NaT5=(5+=p;Nkb?IG;>rtcZ$O!-+ld)w1%_MT9&!s~}u`lRr}
zLn>Zq%8S=>cP_s+*L9g+*lyA5{jX_r{p#YUKD6V7sQ>WNSN89>eUfsDJGbns_-*%J
zZ9#85;5~#Lf4*p7-{ueOR}{luZ>gU@41D>?)vY@1nx&kg>D^x*w7S-^mZy{{2EEQ`
z_g>;Bb06QTIK*?++O_P9H>>VYZZRl5AN3RTmw|ZcMPBbaW$^S`&lcFNbm%Wz@w=K^
zy((Tcb~~P-zCC{Wb$O-Nj@e%EmS0DI`SOKZ?4a<u8rutWxpBVDA)Y6@p;P^umJb^r
z-20RC|J}ZW=7(!1BM!X!#jZC~RM5!X-!%QFzwfB}@ja<ZhxB@;_93_9tk>Vs`Z|@H
zSD9SzoeK{sujur?D4(lxM!j1;Ibu7-2F=d)yR+6|+burQ?xjc1T)NX1{GA7#eKsoj
zPVLiSEAO^F)8DP1T;^P>RtIezq8|$S%RuSJj&zx}b(``V{jH^3Ehq-u+HLSVul1t#
z)Tqgq*8iYt!t@o&D{?#rt*`oh{f+q^K9qiXw;eT{U2;g<HN`XR?219zGjLtnesy*>
znm_Q+<U3mI{Z54)XN>#p=GU%UsKSPn|9Hu*jx~D!&f=$4P#CZ0nlWh5C;6tP+fMQG
zt-t=f@PikXSB(1lkLvxgmSf#+l7>H8YgN}ZDqhI>?ZTeky-w5lbCcIAim`rwazAGb
z^eeeO<2^e~{Xh2ZJ^r$>{2%{}8B9nTXC;+NR1QVNH+xG-Db##BC_WBRluA<aaY{tx
zkYoBvlxiFbh4G<izdy(^A#yIljMIo2%ot<F@7Lq=e(t{Ct5)l<*V=onnf=fE^}bKn
zecji6-Pd*9_qx~G3p*xB?X>0JzUHNSjB7U)C*H8xj+=aAUMnvq$zxFYO~0>|H~vVf
zKPdSR`yEQ43<>mlAEYkrouf4c%RVO_6un;0>$ERJmxnZ(BY^|YI`r5te{5E(X?Vbu
z>rGo_(u`Kq@yFX;a@dc4@KkGI$EsHh`<550=4(*&y)RmKeDAa~S6}Uq4|Wzb{LaKF
zJFWGp2Rc<@l3smkvwc7Gma}eb4H}ESr<Ju`+I_RK%WvECiXU$9%>A9bm^4pi=NG*{
zU6*!A`~IfoH|0K8<g@B~u0#6S?ip`+>wd3$`FS^Vnu>YPEB8D2{O`Q&HRIbu%qNGk
z{YO@M@gBs2n$9YZy44%h`BcB<Q1^Q`N&F3Jf139_VD+n-*IwVxi0zB=S@iy9eBR&G
zblLYQ>qhV1sT*!FdCfn!l91k1^<H<?bmM!6bx$|_UKsw<-v6j;`H2TFdBulrn9=DK
zt(zhKj|A$z?yBN1Y5t$sW{(wi*kDd;i2tt~Ca=EzT_5`C^Q}Q+(!5obU$q~snttB6
zy>9i>;<Bpdy(+pV?YhIB=iD@_lNGDtBQM=S?-TO!TXy}Tb}#Z#^?io2p4V6X{55S|
z7cGzSu3OsqM8vA%oPN<wIyYKa^9QYWRlPTtcU)P&mr|E{S?{anl|#Skdr!1!IOzDI
ztp0m_|4{pW=}YX#ccYtzWyL*F9$C}Fdji?+2{ctsS@j@my}ZwF^1knv^&TC2Qxwax
zo4wL0x?f0YmvtV?s@Jc3By)~Be1pYLAKUH~%NqB>N~bRIZc#pZ{XXEJ+n0UMw%_!U
zo(nfEf8~w7KeEeX(_2l&vf|D`(dqRbTwUz<8kg@oZDhvNodpf!cToE!@2uw_!+L}L
z9iyTA2P^*#5BTb&-LIbAstZp4?23DiU+<Y#)z~jP6#L1#uZ{I0oqng?zv>ms`p%W{
z_xhY5@AIXy`c30oKXt*nd{??I@+itj(s@l)`9=B6+U}=UFFoy=ard{Hf=&5;N*bSo
zK7YunA4&C^{(Xtaf8O)(B)K#-KfV`I_w>YRPhRxEfzP#u`2WgbQ2H15fo1a%-@6{B
zv;B_sP)`3|PQC7j2Hn2bb+({=KgM|WI_$iaHl6vqPF5txem_UiU*7!#e(Qo+zdw_;
zUR}o3FFv-PIB&)iOV4gK4b$$E_e*~IY3lbsBH#V~z7Y8hv7QpB+xcDE`9@NCS?5gk
zoTYzo)igVv#BWpg*G1o#D%$QJGe7)qD1q`M(69Sc<#GH!l#=+)d;dD`cOzr{vd2}F
zUe<kRUF&7tx0O{dNq_2^UsNy3ZrAU9L+r2D@g)5&W3TU%HSK$-k>5`paqHRVZGBg#
zF4*fl_p5!BV_Ldje8qn1YS+^ED$CwSI#E7F(>JYs={nzk^Szz2Fz<I->sG(tzsFcs
zfA<~x*|VqY^lYc;m?n>+dS0|$+4t3p@=<lWem$qlJC3sMQx54=^E<3*&*k#Y_q<C_
znYQd^Q#(a5Z+}g#pC<RFlvn*bybJ61ZQ{7I>iMsqe{hL)SDMx77iQ&aSnpp;px5^U
zliL65-qLfkemAc!^?H47m@bjQ*2wza(_p#$=icM}<9?_AnWJIOzmq_G54JApFFtzM
z>c9WS%udrV?{~ES-P{baEUW$LwO-X@%=`Ysu>O&Uer&tHwR)99+Wont@~ZVAjec3*
zjZW*Q-{p1t{&Ze`s@`uV(eHJh>iWA6k*}ir#$M%<wf%-WUpsH)`WIS7F|D5uY_#n<
zr*HaPr&s*aA8x<#bw|wZ3<~f2%4L(5-tGBLQ?Y2g7xg)7I!!OCE>GIW%h&%h$Gvu?
znHSyM$%<*uW0dEOSCi_W{J!&77<b#;PT3gmVdPCGeSeW$%f_<q0S}r^UOC75Wy^J8
zp%?F$C%vzn6hF%!)OXLy>bI=te(YRU9KY1&Gfy~oPNytPy7!fqzuE?0UT5afk9Hy!
z^^aw@t9m{5`W@A>+aGfDpEmyeC+D`(FzJ1RrsT)>-KBe-@8j?N_>_A-H@%Y=d+j&2
zkA5m^I?`!sdi&X1pSs1eFSKA&$1y0rXRQC5kty$(-&$CZp5(=<=jXEfHkk6QSI+HZ
z#RopI^!!in@nWZHT;bx2_ulQ`u_OJ)ov&H`6W1-b=t$FW&9TpKcGl`+N1Be?9Pr2Q
z9KP40BMS=lx_<Ni?!5l)o$%?I(_VP2Qx)co>(@Nvx?@+K^<1YaH13GKK7Z+V`=Y;V
z7sowJH%kJqTK%QB9rf^-c2-Qgx6|wSds_RnbAY7t&8qU#_-R^sUEbp=%17F{Ot0l-
zzn9Q-dR48fe$gG&`s5|j>}gVY(dWHc?|;VknbO`*Tu{r+Q(e*dUfGy7-n{oeno>XM
zd1GDktNt!SnjU2R{k{0R18M2Be*0CP_C5W&muEd6Z%Vzo%R7m`yuSyLw|?I9nzHH_
z)tj{S>wBqLv2OM`ZG2^w$L|Ua(^vd(gJ<sVq{$&k{$<B02W&Lqw0n-fwbOJQ)IO=K
zeyckF?HAp$@8uPxSJd9e@5nYao!2kdrT&z&*V=yfM=xoWjY)Pi=DQC&XpDU2O*d6<
zo6dJx-$@!2o$pMXveR0hdZ5!U%=<md1-*XL<F9^wCn#$_dF^Id_1%xI3+6p{E&80j
zs&;<oV~7v0J$c)QT7!~9zwdjL)nC^0-=unF*@Z|a-n(x~x~h9qS?NYO#&p@|BURC>
z+xIQg_$m4vr|3P2IIm@;2W3C=_G`Rp`{DTg*O=~CT$!dny_S>55Em>7RApb2=vFPg
zUiCbwebs!I^}H=FouqLM`gfko>ObpwVv-!n&QFSShxvEUIHfbSl~)e&ed?;yW$Qz)
z^opKW=8a48{*Fc3^VMF<^N!m+fwX%?buW+c<-(FqS|=NS;GHjj{F*T%X_)mr!ePCC
z|K8<VN4#Q_y`P`gDGT$yFWzhYvhE91MJMX{UYmY+<g%^qZpH6L$8=J>9rOGB94FeR
z$WL9<gL*C>=f7w=uYVtO{k;A>Z++j@`pw?=P1^aQ$2)m3>w6LEEk5JKQ>XuGey3@;
z&8lbIHTzxjI#px*UQJOwPimKS|5(@Wwp8^zuWGt!=cP?4&ud=>UBBvkBT4IQm_Jwv
z#P{%$(sh0RZm`npz4x-7llPm>${T%uWS7ULx9WywWXl`o+_>-Uk94wP*5{nPzTaI|
zd*dA9Py%@*khCu?dhVU~xp>j~d#u0bZ=O0}d>i)q{l?#)w*G6^oiV>PZL`fkebw`0
z+jYllZaenIgC2RNm4??|`oO+-OrO(=nDseU)%A+b|Df9q`g_Y&_aD#M%1X0MRh@U%
zm2T1RXH?a$tlt|*+8<PvUzG2pdl+f?Wy?S7?+nznUQ_P1Ch^(p^Pal2_uaCr-zQA^
zj%mNX2esudxBA-6cR$hT7uIE-()e(`-&8EyKK7elU4Fl?C?8qx+1IsRzusq0>gTrA
zm)r7~=caaGe4c&e==T&NCfzqmd+&A7@oC=sC}qWQ;{GB2YYBYv;7OB~-*;LI#&cBn
zHj>V3Vt&-)q;%4Ao+SOQYkqtlQCB`AUrBLl-usbtt3U73Q>HDu+0;(cG3onSdGq6U
zOv?J*(4>1dY0t&ueXl{K%YM$C^t+r@=ck<`^je-Y&t>IjeGk9vdjHxRrstkH;k-l7
zIqSt%U9qfvZqa=~S?$tiqMRCK>s7z$l{MdKbT0Vm+({3%o@y1vyt7sGJ>fyOE6Xkp
znof)to0?AAXAC-D{2kBf((LE%2dsU?Pgi-clZ403UTn!jKl5;>sp$QtsxYq|D!ab>
z+e7@H--#|;zUURb-xTk2SCuXrmk-J>WWAS`w_aKMu%h!l=yqkF%hLY^1=Hf`s-G`Z
zCC93te@_4GihGV<@0nIpu&i^bLDNb49rU{AfB(tDFM9O5x3!vz{d(U$$sQE-=dqoA
z0c_g7D?SgYI-U01vuJr$`jnTSqWhpEJsnj3eg~ibowvPae7h*#HtWUj9rDpd+JnNn
z-^)(pZ^=8a8~^^z$F`e_8y@i0NxNS?yVW#2W78$pnREKHtzL0Z>o%?5@h9)O@vWb~
zqg57qzIpG|4Y!!Q=AT<dG2Tlknob*kEbsMu%};Ez#|k@aFsId29CZ9$R{vGM=TJ5M
zyx-5OOZ}pDB(B4<(?#vjuw8F<JV}04->c{s{if}=>-wFetb7gXzERozR^>m6zJJ^A
zc4gc5Nbe(8?zit>=KrZx77prOLe}RodCw=Seh=0@ATJI|&avOJ-di0Mop^6FDV^00
z#q|}_@!omU(shZS7gj!!-ZQ8wKk9GZ^nG8sY|_%ZJ>TgSd%X{k^`3lP-wQ6Aud1KR
z4(abZZDhvN9dbDF;3cp4&<!&>c`<3d{cYAK|8V-4r#e|N-oL9Voy2di`DO1b`bDqk
zInS_NIp^(V`^BoBmsU+zTpOPcylJVqJKa3?PaPPa(<jB{@%gUz<Klheq<e2g?}ett
z0cFqkkluq!eeUfeouM3tawt;|dW{E)wy)}*PEmfeOR|$q&rf^LBWrnC-$_WKllT0s
zZuOJ=bW`%hCr!l|SM{1^7wU@rdTyB1PmD9u(nY^(7uzNEll6Ql)=P_n)9%G)eeO|q
zy)?dxmZ!z{Wv>VOmZoFV<AkF8Cq0iXTTk-REBk!u{J-9Q(-l8HxYcwl8aL&Yf4}NG
z-_43;pX2vRFYS8>O(~y$=%$<e?tM>odd0HzChzZ}R#m^(@s$1EX3}>8%g(RLpEq@X
z63=UkrhC;V?cX-(j1Ru?*hSk-#~*KZ$zebG!BeeXkzA6ns{50%zeA4x)5gF5<lI&o
z#&&f{m;D?i?e`y=Uf!?gjCIkEUio<UX;WYMx%C%qHyu}-u;W2fA78XxH|%wOV*93i
z7a)z#ro9iC#P7suPhRxEfzP#?idlcBw668~{kwH#_m_0u)3p4f?mKXc4_y0Xt166k
zFfHBG^@#7H?;WSfrL6L#=NL)(N&RNkm$KJg)%1GZKMlHl)9&#{eyj3VNqU?$KkxfG
zv3^oNReet@slTl6YscUFPCFM&dJiJ&drL|7CyB27q;balr1x6up6~Y{vSQKkl>MBj
z-}H)(J8QeNeNn&4oBExK$oKXW=goLx>DjHiWBfhUrl+@`z4fVEEc-&MDvWlws^=83
zpR{z*?|LP*tLnK<)pXPD`81`xEWN5)-jROM`%h{5JuI(Q0`~WH$E@*vUF$XVUQzrW
zLDlK9o-Y<%HD&c*HXZxIvaqOp%5InT9DHFbFI&Dz^s?>`7u0%5a*6p>pUYOYzs~Ci
zH@ozNy(g|ard<_Axk{(qN6%Z1Zojc#_dJIE{QLLT)8;+txp`84(fdd7TqrGF^t>pZ
z$LCFxS5@aG=;|j8)6U};^zta5qVH`itnH%wnwl;v&hM2@)$+^w{%kyNDw<9@_e{$#
zDu*;a>$*<!en+Wp_3LupTU2lQ)vnj~n+DxJ>G^t9`9<d?@Ats`ZckG&@4l_B^^3m8
zleb+_y(+8S;xkS>b^5R7chaz`-?7X~x7YU&)7oc^OVaAadtfnrz?JJwTV>LWR@1QP
zd&YHXmo{$;dU@J?kGek(%*tO?=d1mqn{+N!_xz&w#f!E}@`GniIPcJN&U&%cbWA%x
zOe#;3OWFAoZ`f?dO};U&)o(0&97B3T3H)6Gd7t|w`J;YIGwbiyB-JbW9nzxrzw);0
zw>+!rFX=mHW#td*T&5~{4e1Ug&`$~Ey+1Lm|Fe0Y-~aJeuXD7Z$T92tVZ(Y4uRVF&
zhgu7Y9HRbLb)TvzUHl!i>(~vBdi_h<z2dKS+V8o!FCEwJ7gn{8@7H-wuX<k9cNpl!
zf`;*Znki?mwf*joUeYRxgO*R)bBnr{m;JrNLDS3nef_%BJ89P)_B`jNS)Ha~*6+X8
zz1|)>e|^t+uX?1@RE)o0R5abI9i4pP{nIYJ?~YDWF@C4MY3aQ88;aI1sweIpSB?1V
z^*x!Q@`>k#_<?En8q><rDT+n!wJ+@Lipo2->s2r7*1oF!&!Fh$ea_VH`h${7+5Psq
z?@nqT*PC=z-}|bHZqa)e+8KARSoB=0tafSq59;1c8edt<>+=3!R(_M}rQKgkDo>)f
zpypTgUHhu(_WL_)!~TB$o=;4_rbP}#&qs^4t7`paeXk(CFO>BjjPsAI<1V_-EBbqH
zdD~?@hp6gzUVEk6uY22h^)>0fRndI?=e5Iw`VJ9(i{jFkzvBEe-n&@4Y|P6~*56TT
z+W9)W-&Bmx`I?rl>wCayd{zBDuzu0!$D4|2@0r)NylLw+@*CgJjsB!p{VmI1N4{e^
zYy6tEUa#j2QI2Wps-N>@r62jM>V3!9PyDX^pwf}Aq;yt!W~~?dNqVn9T$c9SD{Fb|
zx9oJ@?*^3Je_iMm{SIVV{9kqK(tDqL@#5_T4eRnddu7d2*>sxrUVh}a>EBDM`h9?+
z`V!-(ey5A_-L!UP-J8sNFYwakkKgHzWoLATxNyoLZ`?Jk|M4mJd~SMYVV6VM_O<Na
zQQKzKGwz!Gu6doRG3kByvhoi+>(FDr{IOZBvT*Y1+u!w}pFZCj6mGclwev=<f1y<r
z%ew!94m~c)Pv6@e@YPAXUp>23_Wc}kXc}fcr|WmUUe805+UK304{WsUI;U^?Tqi5W
zbKb1!wC|!8eg0dtU0LVJdFkZ+zF%GI=hd_OzH-^5rFVP2lZ1ngmy*^~+4;SGhdOKf
zUiq%-ygn=4cn_&x>8kIQ*Zp}!-gRBHe*FGt-P1+=Oi}stt6kY~;(|i2EcqwBFHzV0
zrk|7dD%Z4oK6%TVer}rgytr=l2lf58vij}y9&lRwvh3V~N~h@a_N?v5t>4eh()ug=
zcRTw{uixi6P3@2V<;AM*hradX*PZ>cD`vK;#-r{#aElLI`(&$EjB=~$yIgtc7X6M_
z+Wz;)+g)<lkACn}D`LEFQI~Yuc$!jPmOP`rmYq&J=S?b)e0aa4*Y9|ZUuyH2C!9N{
z(<>&uKUXx)>$gAbwZE!<N2*_Rt3G#&`kOVKv>t}}OJ4Dz8)kH>mP6h4t9kj^@|RnE
z?dH3m=nNX8zE^z?SoM9#s_Eyodr`jM+~L&cw%%d<NKs6B4!6&RZ~WSUN50T0iYLGC
z{1wLCHn)=(zx0ROZ+zVmvpZ>+_MF;#4Mj2U`onLWpQ_VIzjISGzifW1UjOK)@tt<g
z(eLs!e)E<WU8jR?ciDX#O!?L;=XRQkN%phs{Ji>**RKq^e%+t@X5~-3kaxa{)-U_J
zaQ&v2_j?_6t6y~9n%b`DdE%hkReg?JHT|-`bJ(l?U31&9Hy-rJGp$~+-}WY{zovfY
zCGws39%|j{C%vaW%)j;Q^R~XL(=R#n>z-xZ`boN%ns$#i>A8UXtlv*iFMgk<U-Mpd
zKQrz*_iN9Mf3}kqllFy4?`y?#8}jTGkDI;Nl81ig;Z9T0?`uT)G$kF^L(|e}>x!MI
zJ0`8i$ajn{%X)s3^n4>P{Uka``FWo|m-QY?*7vH^%lf_2I6ryQ{oaS?<u7mjs-Ful
zDD<<QgVm+ppwBh4_McZ@)$dpQtiOm^$JcAU*zYi1OajqQH!YnON0nV(G|v3>^A9eu
z?n<*dX}I`|6HlG~tNES0=y`Kdyzki0o;_u!XFIUh_fxVyXD|ErDT}Rwe{QwcrXL=;
zY^%FlWh42fVXyb7>voQpm!G=6-)=t;?Le>R39<cct1q|ZG0#oyz_k8ic}z$8F>Sw4
z7Lrq3S#|l{y2w}6>Ab%)QMdZ>cNB-|DLbw8sRuey4rRw1Wv$1obgEtlS<fTV>ZS3S
zc0Qj}Ui9}-^7^-=`q6I3bk^U;NqTOPlwbCHYA3Jw#%~|?;5n_TF@E<XO)v76Pupzs
zPha)?*mhYs@rKQI+~gbcS`qW|RkXfy{~-?McgEA~THVW=t~W96?RUE9x~*%wyx+Yj
zT0d?4F%GLboyL#yqT{WqUD5NQvfB0fzGcz&O?j@3&$8q_q?7Iap`n}xBPVt;;#x<%
zVw1g}pV!HXS-%7AT&h=0dY)hOz3;s3ik^?=?I*2%oVT>^3Z<3jJ@=~m9=Z277Boyc
z|LAvq*548C*K=+CM7{2Jx@x^{Dt+_ts?}e5+ffgXX*U(?@*JcnA7#IPGAMdYv+MM~
zDvaN6YDzk*{#1P)okl;&za~8oXlj1H&((VEuj%KSk^i)~z3$~n`>UqpXWh4|mv$d0
zsXVJ6$a^1NedkemF=?Nd_Iq^~es;%s*W7z&D+%NCP|r)U${}yPqThFq?eg|pR{ehY
z?|Y|ixW(i(|J<qyv+fgntylK_4fJ|_9+9-3ist9Fqj}$V>Gl3-RqfO4PhHEC?0;4H
zMeR?~dCRL`W!KNU9_m`Z*K=*{>vpbE)^~2R(n->P-=EEj<O%=qrV~$F<K}BSRby3t
zxG3EudDb<*=s1(wl{L<$e3v?j&$8cpB-gxgZW4b@%O8L8o*Uo#`8!&JLh`Q)v)*^D
zYrVYBGwWL4K6_ADm42nsO*_x6dwEmOkNfpKfwJ$tqvtttQ!(i|MOpc2>#k|#MbFdY
zI5+;lJ750zHDgAy;-a7X@=soS&)AWynD@LO?Yr8syy*SRq;`3~N9^3XDCRxq&bkkb
z^~!qhIqFYZ+H;~_-!G2s`|SsdzNeMeE=_*1Jnen{SikJ_RmU#9_sJJ8-mVIJ<(IzY
z#a{c(+CDE|Y4x-AyRg;EdhS;DdRgz8#Cq{P#j?`}o^|N4U;fyvR#UO)J+!RtW~~34
zkty$(-x}h-mcztpPhRxEfzP#?j=g?=I<0-Lc01{OxGDKb?~mur?^TY2Zr|(ggAUu5
z^E;~xsvP^R-*vS=y?!S!Yx{GLy<p@0mU^r;XgqBByJwuzncB*VX?k+)r4Q_T$MiX^
zrs5wqdwA{1+dkB4D&~DIS-1LrPp#<~zqgW<PFqiD<z;^_aM1Mf#&v_P-|utCvis}x
z_n4dB{-;+jJ?)xt_qT@FUpb`NwO-5fu9LKT5Lv%xm-L=|(R->@x6Au{u5R_KmQU66
z)1K!PE$?;x4640*tNq5h>2+Ryn*KeRvfhIjl-?%w+pm3OQvYS;_xe6jzuT8J?noys
zP8gP7@xu+CxxYgWQGN?+y6O9*$bbCZ&!(i)?&BntSFN{A`z}c2w<`JM-N)6fe%|}I
zWy`VG_cY4lr|$P`>iRrtVc}~~?+uit-;s{^r{7ri`!ZR-b2+GbM=$-ADLc%5s#P_{
z?`tfm>7@59qJ9@mSM|Ix@>AA5jg24u{YBqh`{7nrjC`b}7Z&?&oJsN+)IG_%#fweh
zC+&OjN%LH{{6RlIF022feQQ(l-?Zaf|FrDfGde@8Rt`niU)nw8r1HFe_~m0yTjD$W
zJ>44Ozm`MR`RjGxk+q*;y#-qWRsBA5zob_5J?W%&S-<;H)%dEKk0IR&r`>b>t(^r$
z4oUm?toc>xZ&tcRzcX2OyCv_uZv6W<AKQNG$*(*6Wmn8>H5L06Pi39oy!F!dn@Qz)
z`{`Hx4R^kF-pKVYwDMwJ{`z(Qq^f@U^<AE-`zh;Qect&myZ#xQF0sy>)1PgXjalD6
zYFfRr&Uc1%?tjM->t4}Hl0#X~W6EAHNq(u<{IdBR(i=+P|CT`3_gIJZ#-4rI-hXWs
zmBarn^F03--KVDYQTOLhz5d?9!qz_T`zvYsQM9})|BOz*_0)GZn~qJ_|6cDiW^Mob
z)7F3Ox-;gt7Bnone)G1=dQP5HFK>TY>kpqVSNnW<L6>h?-_hHC_SUCvvFr=2A^vMQ
zWL?h-TfISj7bWZc$FzD?&sV?WvpAlj`@p)itNI+gYWh_@hbT%n>PJjxjYk*MdQl#E
z(?#bcYrDMslvO|J{z+Z(dwrj~*Y^P8@3K~%?)5%h(e`Ek?qE~t_4|7u!~VAV+Rb-A
z(JAY@9C_)KWk>VUsp|Q1+0V-dMXxIP^qX$4@pfLlT2Shj^}X5!MIKd+!~1f*VpaSN
z=?*2(1PKi4z0!X7JLjmwH(31ivF)N*)m(n+h+EG-Z|l1{Rby5BFQ{~T-M`nReb(=?
zCe=%u-(J6KQTF#ZlIX2D_W8}uT7B$D-rp0<JHEWXbC~zMp{(C2L`VP93k&x8J<zg$
zm+7+mHkk6QSI+GW3e(nkzslq9mo_E6;{#Wozt=KjMw*KKn%}hNpLHqk_x-t~{`xg9
zdHY#V?y(H&yJ}_iTh@D1gQDZU>VkstJFr!!^RC;Z`eohQsym&e@#oDy=KOb$zv;RO
z?WW^xvtIn(As=0&-7995-=O?P(mmQ<^Rs?0AZxv>a*Xw=zBiPWe&nlYy4QPrN%7Re
zl3(;b#h}~e{T_7P>pR!Yx(@RGu1VeMS9R{0ReojFYkEA|>vPdTw{QCI%?&EfDXZVI
z_NRlQQ+L0R#$VC*m($vnWgqKK=c;3u-uvW>7jI|9yyN-UTYmoDla`p-O2fQ#i=O`{
zJ@?9*?>SJvG4H*{SU-*5vhRZ@Jts}ePkOGF_jzHgpZ8vOulLIP)xPfMNLl$yvk!GG
zZ|Zma@SXI%<*M`Z+WWfJPx_8j)A9$sj~VvA;TDtE{Bw&OvYsOjx?ZpQ&$_PbynJQ#
zb9Jk?pu7*tpT&7g+duZJJo1${-Rpgiy#2-c@pl>*)^wCh-O_clFIoBN_j_Aq_gB{Y
zOoOJAb-vQ-rLBv+-;*n=ei~mz%ahJ6nv(zP=O0{R-IZo_nu;gwy2GC5+%&7xRQ%$j
zhpqnmZ_Mm86`Lye$ahoLYvi-vdXr`+%l5DRrk57qCXK(>IKAxlD{u7ukzF2}-fAjN
z|Llr;j$iMYR#liZk9Et>`}=oE_3e{;#mz1~Veg6Sj%haq^Zq_c-ReKF%^oZ4u)&;G
z)3IN_A5?ZfgZfTNul<%She6Rh;L7!;tukpwt1g)L-dkDsjp{-t?K%F!US7BOA&sAS
zU!<xuJ_cn+`rU88&-2K)Us&~fTm7P+^}Um->h(HbRoy4=H{G&-ce7vg`hEUVm3@xq
zN=->O{X3gw-@jZ?>Gf(SwSVulGgn{jj}LZw#eT2*UhmuFZD00#fK9s>m&ETDH?FhD
zy3bzJ=@-)5L1Vwq;p^7lpyfa8f8mfo()oSX{H*6>QJVe!owC?}QabDXkF@t(VtL*;
zDeXPtB>orl{H*gctatoQcPu-jLk`~ieB{0-R{z~zw|Am`#&p(rqhI%0(|!kXLD&1?
z_=XZ_k_5^=&r4hXWtGpqth4W7KYXNB72dSuk%xY4yT7%{LOPN4_o3=quV3fyW%c8C
z`^Y~D*ZIP>Ke_41G3}}_so#Yqe~+ENzURDGJ<{nF)8g&C<z?Hyy3(u5d3dkqqjgy?
zNq($q{_6VPdm3N8elM^t?d{W=ih0jB>Q=vB&u^>hC-3hf)UAHM?r~LZM~C$PM*=6k
zZB!B`?sMWUyT4(p6}MVxtFcF%_~8?`TIBD4cYoRL6L(*Jt1;g^arYJfo_^z2W49Xn
z!MUF~{KVZ)|Mt7?Y>gZ9$cziV`G&8J8(C$&%`dtAo#RF>{M*ypJbcU|BL_Xc{winv
zZ0yMU9{I{cyI(e@i|2j%d+)pMy<<im+~Xq$Pgrkk7uA<9y~5d3e|5>MF(cZYa?zTv
zIQx;?I$v7lkdb$vJH4~x^Glrh`(q#M;@cKq{GjVLc)0V&W!}Bo5i=g`?7h+sr@naj
zjLyBUx@Yl&wtKqs{&O$hZ<R04?)>sicl_l)=g;bFv%zv--{{wKy7@1E?J4iP{9Vs<
z{@03Ed}q<C9`D>gd84II{n4|XpZ{dPJGcL@XS(S-zIWi*ohHucEWX^hk6eBCW1Y2T
zZ}p|4&wQYB=QTgs;dMLR{ZIA2@%R&4|9Qd%ozCO0+x7gP-P!uc)XC4Sd+Y73dGk-3
zyw<9JX<c~WZV&Hs?KQ0nfAaWd*Btn>&V@_fddGoh9n<>c$vd3%g^4${)|`39O>g<y
z-K|MSZMD|%7e3Hh<Nv&O>~}vtt#$eF*PXP?;SaUeJ$sR*Z<zL2>*_NOJL3b(J=!|%
z^xy6I(l<TQdeyg&dgjB||FxU{j!Q3JW$~{(+4|Y^l~3tx`Aip2IsfbXz3<ZLE&LsF
z=Q*!gf9+el{Yv8te&M_au6p5uOS|Ja;kj3D@{9N1+k%gN^TyvU``7zh(0Gkww_p7C
z=1biCcz2xoLHFttA6sGE9v8HhI^}<Dc>3HK9eCZCZ%^FnQ&YQoAPwoq4`2GtN8dZ)
z(GFA|(<|Jw>W@yp<G#+bPan0(*1!F07vJ@V*IakTTkq{m-QZJq|HnIT=txII7k~67
z>WO~ojWmAcOG9!aNAjZw<{NH*>S7!I|Az=&{m>2_h|bKDzIDSkU!2<28~G5uZ+&-(
zoo+ex!Itrpmvt#0T94!-|DCV=+?#iM`pT|c*Iabk3%7sx(oV!!&pbrD`pmby`xl$v
z(!mE*Pr3f_ElnQQpLP&E_=W#|{mh>oeP@?{^sS4Caa_u+?<IHr{5$qPXr#N2<eOjf
zV11~MuKw{y-qO%`jf>vLe024ve|C#KlEx1-K51xN_%#po{EQQRd&8^Oe5&)+AN~4_
zRsZmG7ujvN{5i8XzvtVtx^_0!x30(mUH!->7ko(5JA6ZOQBICfd-5b-<Bj$7t3H2*
zKjXkR`yf9eI{J~G`?5v%-|h=z+i=}wulnWpKekAF{U5yMzb-y)d|UdY;~rXm&I4oH
zS1+~2dsg}8m~MUP*DSmAY8P+#V(Y(_I`+a1cbnIej^z)&c*4CKkIZT9H)r`TUG~K6
z);C^q%Yo1B|3Yi}%eHy+t6R@+eQ>dfTV8tn+?IOMJ00`-`!Ag{rX4YBz=qy=FWfeH
z?(NTY`@^qt`0iOB-Q>XWv%BLlj+Mqun)sh*Ki}G9+7H&=?5i`oe$see_Vme%Zu*K@
z-SHWhe&8cV%y{7Ai@(sNZ(QW|rXxRn&VMd5rXAap2Ris6NBql|KH<hG6F=Me$GoAV
zoLtbAR`11=cR6I=DdXC^t#IhHJKE#A^G|Q+6|^4ceMI`lt}55gx?&gPTVLAA*KbzT
z9)4?N-OEmT?1j#&etp5VE3P!R^XSD-uX5+w&v#Kg2xC3=M?84TeXCx1#ZR7UiDPcM
zpmqMY&Um65Z?T(Dyar=Fdq4MQo8Nilyhpq7-uR2>KXlEJf9uA5;yQjE%14J^gZfnt
z<%=KGXWx(b&d$I6(&ukz@dxHje5RaVFrVVuZhADX72ly_e)tXRj9yxor>y+0SFX6$
z_^y6RvtRU-{<3Gv<x49!kI*{PuXfgleyrmp#D7HV(KxMR>k@tavl}<Pa_omcy6mFu
zcdUN!<+n{;tZS#yM-P37E_)Z(y?W?Kt7pGJPt2S7rziA=UehmWdP>ghtGGELzYWFH
zv0ks~7(XP@<A=n5kbR=Z<RPEE5(khU`B+!hkM`CLd&qu9WM{<<@Saz{_@*ll{#$pS
zM2`5;zjcQX^C~_u4)yS3{7||1mX7V|0evA4Y2(HxKH|8^Q#<`>XFjwSM?v)TXMW|Q
zBW)bUuUx;7{NlLqNj~H+ADVwiFVKVL`NVnG-Mi#}KGZ$;5EmNv{Bc`u`JE3;>z<Q{
zGtJvFulw!Ji|p}a_nc#uw|?aPb8ddF1Nl98=_bcNe(_OrJAd2sL$8_jl^43W{~K32
zdz0m7b>nvDE}Q*u?w)^nXl4f=;$r(l=QYr{pmP;z_0SU+qbIGN^zKVsec^=F9_gs(
z+y<&geo1L^r7!pGwZ}4FTm7omrSIDDSD$>~4;_BbIf(hQzT~r~)`RmexW@IL-(k|^
z7dw|<`Q(h-Zk^Y;=)(1<e*Zmly7T0mfqbJ}@d=HG{<8<s-Wk7g`jQp#0nHz@P9Fcu
z8;`zV%%8gFU+<jt&@B@;o&1mUtkHcv`=&iP(+7GiZJio#QKT>Q41Md39!2`(ryR<M
zv0ble`kq8zxp5`+r`&v7A8{Pg5tHb~eDj9>y664;{IyoNwR?_(F1aU-E7HxHw$7X{
zTW=9R{+^qT+Gor8otamkx5jdp&+jOgK6=YzuDEc$7d!SB?52H!ew=HoXWd%I;t+Hv
zUH+xRKRWh>E?w(iJ@Sb1AvfzrJ@G)S&mQ1|J&>=xcJkQ=^&vhZz1Tl`Nys1acaVQ&
zN1$?Pc0f8J|7*Wx|0bWG=7-s9@vD3&j)3xwLmFNAP(FV2BW-_%5Bd0x_2W471GN|5
zYwtOMa&fwH`(6B^7g0Gx$38_GqQBUK_dYlF=ou|(<I~RmL^~+o`HXQzv=2g0{k<Oj
z#I9f8@#$79w?9)~KlV%Zi}KZr{o)fEpLDD*tsW$w*p8g&2f5QbDBhqS`aut!B-D<6
z;0qr8-%p&p&6!7b_ITMBKeGGQBi-j>);W6**%$Ufyv_eW>-ptterDV*-@m3K{uh5)
zckF;TUA({^vQG#9@+;?WxBYV+=gIoV5B~TQc2~SC{*Ue0$9FD&+NVc;_|!i=hh!(k
z$LK=z`G+%K@yb7Ly4TE(_!@eSsUEucMc;E!^`y-Qf5Q(-N94C+zO?q@)#xYH;}0Wh
zZ=aDDoe%U2#kb@r&Q8i#4%MTVP&thG)`NAT9P)GYQ`)*^ALLsv_AUHtH-GfpPMSSr
z|C}$~{p6iD&0p`??)koac9ETAAK6{!Ow#x5`}W&kx!>&W`MP}eiQVCE*fVL!9zgcc
z_}MpfrP){G(Z6#*{iCNp`#g3w31d6+qP=-x@8Ud4n}2EcE~#Jn<Y7MXFRh;aDt*A0
z^2i@P(5Fx68#npM*RS&$boHZO*sVXRm-3<Z`o|afNUJZcKl32oGCvWO(<kGB=#dA$
z^k@E{aaE1{pXVR;8J>5r^AXkOFWG1Gq?7DC`yTB+f1#av`j7e2P<<HbC#BU-LiAze
zQ+{kOAOG4x`sEzRy&KO_tP{_3q}yMZv;K~kKhu3K$v&~G?5sGJeHPEFZ~blfADg~$
z^>v<VLGjz3JI~u;mz8F;KC|em6VHC~zV3Mp{dZr4ev2RR$sXVO)BpJRKA%{;orV+6
zTJ@L<H=oe{!W%YydfA&6Z=d$#(_g#6mlkWYyKmod+h4!uibdO=e|+ZdUp@NiUB|Z}
zzM<!-<YL_8e{<&HYy9T+4msmrySQ%M+hRA^i?}|!_Hp$7mh_5$eDlRqK6y(^8hv(E
z{}I^-ajyRHFYaYGoSU$p;$8RH#D(Hg_uQOMC>J-XCq6L_@r8ItTD&~+)QyY2>`(W1
z#J%_z$Nb{9)7O8^qJQb$Z=;v|?vHo9?1UeWo!Py2W}QI#NM7X0U$e8$6_vZUB%fb^
z>@qa(`iuR`H?P*Ic`<J1TjXe8L=NWD{A%x<6F<t)(_Ve3oq50qJ&E=7kFL1V{My$@
zn^*ecoX~kF{bL8L6Zu^n_1omv&3FHC6rHb^h4hu2=%uuIk}pj@kp9s(dSt&F<!wE8
z$2mHF^GZ(kvC`f8<T83sR5|%U_6a{p`mUVcvA(6_dZGVu-1MG5QP2E4C%XMJ`(Jne
zV>7$=OYs@yVqD5S4<MKBI7j6W$H%U^2diKAWBD)qiTl}qb_n0n%8gSRJtz*8CJ*`$
z(fW=^F7y?~_350~`iy$ZF4JRlp!(4HLqzMHJ#g;JPw2<`Fdpl}ype}_R&Ktb`5|xf
zV7~Ybafkh+dg2fBfF7jB_>oqh-@so)d}?Pr((0?HoLpjBy(FYJ?6>)&C-PyOZ|$`+
zZ|0NVH_y)P&9n0qbj1zgef8;0OzW4PIfqgYY7g-f`-%GvY5N!T^^0HqsIQ)U`m29(
z>+0|5J!SN)yU3^Y0Ld49a@9XRrD5D>;S-ui{m?Ibvt#s)ekqUkLwjhx@JB9?Jwh+m
zi+mavy@ljyA8P*u>7Rbphv<^8@j~M<kK}LO&{a==<~R1EKk_%f__yyfe(Od(?aZ6{
z^itaR^@mUO>^r?@<GfWqwEsm{d-`R(^2sG4z9Sk3dC?2{W4)utPl|u&q47cdM81vF
zd4zTEJfFNEIhY50(Oc_6TqG^-pr?>tL3E8Dl1oH(mpqc>Nl%rd!_L!Ja>So?s9b&d
zkw4?rUO(!Q7erS(X-JOzV3f0UsJ?#5Gv+Ht2VW4s_>)f$jT@q;A9e$M`bh5RnP>3<
zL^t-Y9sZzsweIBWkA9#RF^MiZ)(=Eix%SXL!@NZK7$5o2Cuwp<AEFP{gW97H<9N|C
zU-*@#|M+*V#?H`3`l_CBCiQE+@vokC<bXf*<U?_%eHO%*xKTW6AEUgB&b#dgN7261
zy=wb7=X%cT$ODoyzO7esGEQmZk?xiUe&mxE{>`^?>m3@uw07n>=A(x%bfNL77g4>K
zuU%HO?=X+(vMcD2Gqf%tyF;J&2WUM*`RoZg(&))oKcf2dU0S`EmM^}+PfYW_5FLm=
z?WOT+UE_;BIS(R7=UwEk9lenc(J?R5&coD$`gaddTL0*<YtCbx!#c0i54z47*)!)I
z?342qe#ZK+54Qh>^6ii9pV<@rLhIPNvi@OGK0S~1<inS)yZ?!|d}oRFoo~MEKaT(L
zOWOFc&xYzpOrwuK@*^kXH9mR(t#@*eZ+;@0U-J#EQ|XA}ef8+I^<y0<hw4e|UmB_h
z<>MpvuUx;{No%K`@|Z?f{m7s3#rEi^FU~;+DmPE&(Ym(I&`Uymh_j@Pm)(i>De{dU
z{6+nsNBI58n)BW`>5>U;&$YbA<M|eJKVtl=M~*n@g+<$*SM71y{nMX4V$rteR^IpV
zoI{%4>ld=4ah}M}x`EaiyTiU|4~?I^$&>uze$4(2Dwnpev(HUJ^!XQPo{ZN#nIHM)
z8)|19(orAui(mW9q<rO`^BRwFYo{N6%y_1*{k`RW{qXGWeKh?>w9j?V*tvs#r4QTa
z)+zh`<oWLRE!+pw-aRt>;6pxs$<4jAxSpINI43a<e52#NO8NNE-grEpWFMR#K=W+=
z=q3FZ53&>VKpI`+g4O~3p%+j*r<~ug4%q?a_`<LD^7V@!`Qi|Cj8lE}<V#0B)t8SC
z^~nR@P~1(9_+b}}13mFhx83M>JM4S-edm#oy@}6f?B6_>(;t3~hukKw+Madj_nzq9
zGo=SmKll+xiO=w_Kl5t*;x~vN=sA?<4(NMt7mCZec6{`EEbb8*w>XGAp>asN$H0!;
zSNdKDxr&QoyyZET{UEyNk(Y7#PK@V++8K}agMW1J9phbT_3T^mE8l$3PxEShK>8Td
z>MOTy%pdx(o%-~Rzhme4FLs~ZWrx{)=a2k>bP}Q)>pRCqU%$89zs^UdzJ0N7oW0j$
zd(Jvz=7jD&8uw}5`SW|PocZRLv?2dPU&%eryM2hb+B}(O>&SXBPx6!G8uO(g`LjFL
zzkQf=j8EA6cz>I{FrU$%@h|+O_3oa7e)LOk=odR_{N!p~&}Z_rK0T*qC(VQXAUU#6
z^w<3!<8nWPJ#^m$ALa|9hrV+p<MQ6qUw`(71NM9Am=W;-|IT0Y@7@cw4(KO;uO2$m
z{J%H^{aC)m>$m&z3vYV1`#mrI4SoITU;HEPa9`f~r3d_>@v}?jfxTlt*=zI2KB})=
zzWYMT*=_dT`5b!uBYVrfy4Ocv_!IucIaIv=puO_}bk5ml-yi+?yR%yE6?*RlzVD!m
zHv7R=bGqg3<MA8r3F5>3t(eBIa|%dK>dV&;zLxyl#xJgR+tb~9kH$w%_#iLy#6Ot_
z^GNQ}=EXWO589&#*$L~9{#f_ymo)ppK3bRjb?hI%__SW-tIr<uuj~x{VsDJg`EcaN
z`nNx1_u~7a#>K9(v+NQ3D4#uo)+;%%Gw8ZUh8}xreD1TcTgt`X=qSerzSv3jQNH=p
zfAlZf$ybm6RF1AR`s5|8yl!aT$v-RdkLr^<l%EyJnLb8j-^D@VB+rGtkE@*jTIuu4
zO?mGsi;S={key)<y^rEOmx#XWwA}vBF1h_K&$S@C>D)$Id-t<rT6=bq-Bn*XdeYvL
zf$S#!rL{9o`A~i3U-{A{pFM5zv)%7r@f-Ma-%9&ve?52N542<F`6>244aG&;v(M~3
zWFO;xOq?ZO{Hi|t0of(%*8WhrcvSqru89lTC3Y3sZ-{HzEA*^C`P!k!PI=xTPLa>P
zNTbWnuv>9GvB%oQb&4PB+&J*1Uvh9hL_XxGe|7=NCl}+>PQS*%&dL`z^2^ST#CiC#
zF0C)?m_EliiG2`%h@0rI_=A3m7sO-Y3-w|<Y4xLi(@%az9Kg@>i;%w;r%0PW@+A*`
z*?f_oeXe~_l%sNbf<AqdW=F`+ILTQYNq+dYUfB<Fg6xI*5y{ax2tL?%{IF~4L*szr
z3FD=Q<VgRGI}NQ1bj&X#7k14!>?5N65)aZhdPq+6&-mD9Xg=6;_Bzf(oDb#vk99<@
z=+a|$+kE4jK8t7Z&0af~@*K#$EcNL*xyV<~x*<RPiLdcPPV~$?&@c70V}JPvc99*I
z&z?bXtnnqGeX;#AI|r>>=NR@Q^b686dJ2_8`Qm=<Abl5~X~&Kzm(~tlNY9-!*(X_V
z{1Ja^zrud;Q~U?CKGETip>=7V)YA@K<CBj*xj_CI|MZLATSwM|cuqV<e*76e$&uY<
zm*q#)9zAj<|Ex&9^p>2=hjBV55hu}i^0eNp7wgh`QD1*ah@SfR!3Vyectkz<<_Xd#
z`(g9Oznd?}U$77C89QKq%@4CL<b)r7#CY+GANE0gboED`>=VACp5a?R^cj7q9OBP9
zq5tZ$v*hR;11cv6c9=Zaarw?S>8p7*zw}GJh|z!W589(=eAcmX8K-{ei?~yLbd;m7
zeMEd&2kcmsFZ%c)2Pj`XDDJgB@QW{e;>-N8FWS*Ne90#-s2zIh#kBhP(T-iEC-xus
zq@V1nb7gkUc#YpU?Hl+T{*WCMrzT-+r#*W~U(KI>txw~_m+=^HM1C(Xikq!#>tCE?
zpTe)i{W!ZJo`I1LJ0so^M?(56uCmUpAAW-ztXF;>zv5JWLAi5q>xbQ7XT`71Da6P4
z(+)l3bY3h!&MUn%9(Kfd=(RMY|MUVKNWS7;^5t)gj~}9E{G@iqrM<LsUiOrn)H7c4
z)~|AMGcU^d9sQ_BFQ9hx3o3{DLl1W4J^DVUaX1$@K7NH=;WybeezxDpUWm_}JD6AT
zo^@*;tRLf2-}-{)6Oy-mmv{tPAL38?5AE-qgIGuQb>awmYu?_n%}M9JWc{&i<@Tfa
zG4Im&G46==o%BH5Wj*UZj+0){PjZTB`zhtt0sV&Tjd9X*<@C|K(P!o8#_^*=PV5gm
zE^cGDjoWz9)t`Lr%mX_wO>WwuD{X!tJ;A?z$;13Wa%R`rQF=_T&6Dv)`t(6Qd6BDm
zgZjsp`s8AK^bT4N`Y{j2k8k~;gFbo0{^^Z;XujEP^q~2{5BcEBc<^Wb<VQNj=N_qZ
zP3uX0^sPJfpWEv8eXoE2nC|a?kXsx-xyNzQU*D%^SLhY~q<ue&y`jJO)Gv&3a?a#D
z&v@83@{MR*_|dO?^`(>0c$4t@6MwSRM!#RA?RPj<J@S_;egCN!S_d8S<$dp7|M~9s
z%*1KdGx?;Uc_n9Z^!!FSWZ(4%-Ai^a2)^}>nHR3H|G2i_Iq)4Fdg6BuJil^Zncg}7
zGe78&gY$OZ@gaBmNv`P1XBS|cANph6u-Dr2ug=}XSK<Zx5$)K0a^=VMLw@8*-sIz)
zIoc2Lj(E>JqGSBzs6TdsUz0DbAMx~MCq8lMM}F~Qi+){r*jvwC`BigU?1_0mhdk^D
z(KSElA938;hplks{1(2=<Fz;Mv*p=mjcv0l?z__i_R{>|!+K^v*g0{E_N^~2dCei)
zF3~>q&wH&l`GbqLmpx$d(++y}rR_}~K6m6(FPYH&y&&?ZxAcg7*$edDqxX9t_#iL#
z&V2E+?2WjL9DQ%gdbcj5$8LYzVT;{0zWvp=&v|j19hPcKlNXfl_f7mx)5#xMcI9s^
zHm>b{AO6+%oh<dOTl#4p?N1>7trz|fKYl00IPSe@<=uYy$dX;ZK`!(rt}E*+V%!Jo
zm;L8g*+u#&?yyg@4~z7r?cdaMeq|r;JOlsaWPYK1`wsM!YtLU=@6r+d{uBEsZsPao
znRUT$u@mOcIGp21vv=gguGv?SH$LTy<Je7cpY=nZtY`LE+Pp#Irf2MdahY%S3tE5d
zGI>Dd^p9O(_vi~hNq^-V2foa&{@68kN<HILU))Dd${{)6iyY}SIpB-E*N%RX8+ntH
z`NOw)<44>Vk!DxL;rN8)2=xc!x}~q;D`=eL56RK|86Q27hU7qB$(`OPXTQivJ>}{v
z*Pn5czjYSp#W>BIafnOw6Op{gS3JRAh^NxfxnxA=oz6?0A4gQ~+#(H~n>r`;oGH?k
zc7CfL&m*0aI_Jf|^KI?bcb*!@C9ND?_Sm_Q^GETFG`q$>v5(I0AUf=*_VW1)e#N-d
z$DeaT=T`EKn;bm9H9qoC&p7l8_0JE<M+bdr=h66(ufLea7u4Q!N%yhoiTS55zFSTo
zv^QV)faGLejmtT|@k;BDoQwlKbf9+Xi*NW}aijek|16&!^&W=b^YA?0eVKS3=pH*e
zu&^V2hSrUBCT<g_ITwfGB=MGb(0b+%#qHKTdd{!-GkyXcepFig1o4M|$bR61y}&oR
z`W3%<exkj7r~L~*LI3P?>{Fn99eq|$njYJyiHF2F_7~1E#EblHS01D1_2NZ-O}TN9
zYm8IKgTLjs%?o*uk2pb`g&sf4kMqw^oG8u{e~5edckR*BpEyQ5CVtp^m&Z2x>L$yy
zxA@J;H?RGhm$vVE*BPsw_VH!9`#SG)vmf@&?8>}b-*(U=$1T$Ko;Q2M{<5p=iM01z
zq}_vNpV{ST|JX(Ln0;oa*zf4y@#Fa`I<dWc_0hu@`;R_9K<|9V*?w0(d6A=Y6>+8c
z5QmB5#JBd}^j+L#9pHoC5q}s5e$5a6VO?A2>_Jh)7k=?iZy~)Ee_A)zUqtJLoW#A>
z2eeM4(P3Y$5B9)1XJ5s?_CL@*gdNb1eHMqY|Lg`opuKj^>)By+p?2m0J;+X*Z}&RH
zP2x4E9ec!on16gj`)_<e<@)E3=q-DuAO4NqWoPLdI`}bOe47{ZC12bX`InDR<B%5L
z%GWP`^l!ePa|-)T`zq&e&Jpd4={^0XxAc&F^cQhqkE^O0zkQ;6sNSO#*ZN)rzvTU8
z{>tx!?!U%jE3W*FF(Z)wil{yG9<}#+yyxS+PJS)km*s!hbK~ImjK_0(alCxrr!daQ
zr}tF7cj|pp<BsLzgTCKk!ngj67ae{bpZu$FL-rf8_v|sdBhAi;Z^U2XGj^IC%8KO8
zZ;B7Z0m}K?I8WjW``74)<wJJL{*b+6C)siJ>?_%4c7;7QZgz<O6@Q4&*)M)sd@jy(
zp2Lngr_|m)fIak_LOuT5Jdr<ph%Wyto;413HzGTME<ZsI<R#uASN@Xx$deq&8y!fW
z)Yo1+{v49O=TZ^z8}rlB<d+r29TCMFX=wjq+-Ye4VSGs_9!o<07m{~G^3IFwM;g-4
zBxG-rFlqeq#e+%6KENb-$+s>diZke!G^EFAX#KEH#!Zg&()gfq>)N@vakJ~zb5~zS
zzZXgmyYoEyy-w>BJ^KLrC3bC>5C886w_J6R_J@!8;?{TEHLu04v47$==M>J>*g5sp
z<DcA5a898==zKuhzK=b|FS_=L_EYR0yA18?BYo}IWAQ7yY`pqK$9SCYm@n&@UDluW
z_F3kYe94<UjRPNZU$*G}+kIhdo18tjj(W@9h}YR+_Lcsy6A_&wboG4nyH4ov6HtHr
z1!Q;Vy?Sx~fR1%!e;CI#=(PAvJSKh<r-?JgY2r<3_DEb5<0NsPILJ8#v>$^p9=D(6
zzr~Z{SbX4DT#8R|DL#v${=~uJKj&KdS1$gA<ftAwtM5Gl`vmX%i_^(bT)z6dmigEQ
zJI?FK$Dj0|^{OuYvpde8*>U@HaWMPME<)>=AL7^eC;keuLy&#pAK9zep5NvtwbxF*
z{@G>cDbiiP81A8K57~2eT$<f<?t%~4r9bMAyZNJjn7sI7<0n6MQv3nMpZJ5?kuP~@
zCr#h9gXE2Ga^|n}ORo6kCyWQO=f<Hv`I!&)p8awz<@p`^=eZyM!mk(~`#i*rmOk}I
z&vphS2l^BBN!q%hNA!=r(KC7{&JuUU_>6sW&M3~Z&lW$~U!!9^i~Gcr(&(~p&IRoA
zrNwFD-595;haS7?Ih^+T!I$$H=l#x$odZ~B&S$M3aX-7nE?8IQ$2!v9`V_Z`Pn?@u
zf7Ycqk-rw#&^z<ZKG+X-?eD1F)vtQa6Fh$shu{a_&XMrLPKb}B$pu|{h7a;&pUt0g
zaXPyN>8bX{r(8eU(_^Tf_WF|#trLFGda;gT+I+)UkH0a`^uoHKchdGT#%rAHcjSwH
z(LeqMCVl@>94D?5zlqbJ^Izw$;zRMGb7}eX(K#;^XN!y31>-hP#%ui6q4D#p^hcW9
zA$gcjs2n|dpkL3W^{XAdkxyUfoqF^@TtHv6*MCIq@xw0Z$Gl6kxA;L{8hz)M+S3>I
zA_>V)xpuIdKN??2Ysb#$mp(${Fka&)Kgf=Fo@U;xb8_M*p!*f>ox?xve)K2D9W<f)
z{)Kw>`_MUu=h)6`#Ov;X@FV)A*UnL-={J6%a~JC%`c>_Xmwxh7=qi`*T$^6=@9c|w
z&-0Z_^ZWd%aYn!G-Y$R4zsiU5p*Vs6hx~#5@Sz=g%F$77o}70<zmMVmgYz8wFzwA_
z+;=(G5?A2YeFFLRZN_2#$PFDRt~O5cray6h&09q4pB+v@<HEoA$oS>smz?nj>8bIN
zn|o5`1$}ZR_jvC_+#t>n|3H2N@<05GeCV7d)>A(z-?<6&98J6hJuefliHn>E#kBFU
zW8y0H#CPH@<Bt7_&z$pkt|!gU#(LsA^q~HvBeK_!UlI30@uYpIxJQ~li2g%dtv-7n
z{Ruy>AAUps{6)8)@OxP?pWLlK@tJuskLIhpuO2;Lz@L8UE&Y-vU;Sl8{!sh?^~Vp8
zgZlU-cX~xH=n4H5ztb1<98o;RZ;LbV1@WaGz92hAkF<m0CP>cwK0QE(Jn-W=i*kN3
z3E9^q<nO(o&wjC!?7!z4_L=NB`zRgdD($@9^Q(yLkaJ_}*?v%3JN4-S{bnE84ejXz
zJ&1aP9>lNu?4W+oL&yG*zDZjT?4$K!T|)b2Xq?WMozJsh_#_u}Jx_3-gZ(icd>9Wu
zfG&QdttWO%J9Y`5#>p>{4<sk@VTZKCmwxaetv-Dc&sn$Dmvv~p@YCW#dK%*x@veB-
zeP?lk^<$lhXT`bbi6`8f5&wz@yiX}!7w^dzPq^<ZPKU-n#GwTKE&=NuvJ2uX>y8~`
zC*-qJ;sN%J9kTxTA$|*5cj6xLi*?8k@^8*H(BuF3dHZSQ);D_g`{Hx!S-y40KCpY@
z8z|p+*%$k~$Or!QWB-Cb&kgKP@rQ4A*!a<7SNL=JkxzD394QVm4$tS=W%Ht6^u#ah
zBfhocPn~n%mpx{$(WM`d9+4+K!l!!l1R9q(7C-bV@<YG*C;m!5=;7Zy&?`tz%IS~u
z1@_*4)Og67{EQ=_@j(0-mwf!vNAb0|R@^PV;aBa0J@4j!`C0VHM_fQ|>;Obp+!4=5
zoufK0jp(_#b64fgQT2~+=ddvLtAEew)e{dXH-6~8LgWWu?*GQ~W4}M6+&MCO_SO0^
z4r%p_TfT8=Z(ijapK+lB@#EgJ_XnZhFEAeT=nX!l^-o{S3q7F+>Z>OXv2S26oqvdf
z#4+{}&X1g<h>M)Jh^L&-vA^s9v|f~3x7HK?N1xdv{!g6AKCx?_8>4TZ#ZEd0muAQK
zFZ)4ugFTne?y@KRi1<<5$ZxO{#?4Maegj?f@Xep<&v_kryJyCZ89)2Qk7@6mQC#dC
zSzJpW=z;N=XY(7M2l3zhwej$);s;1B_R0Juxj^~>wKHGrKRV`JK6}jXpbORGSILk6
z<e#O<mEN$+kiJ@X^p%~qkF{S@U;GByeQ}<&b%7o~!p}td);)U#oufKuHLvv8Jkr1J
z`t*Eu^nA%Yn}6$ye96oF(Zh(oALN{Zy`evm5A^wM_2>tElBRd;1U(UdusifYIsUC*
z<6sZeXICOx-}o1AdM<{4{i!eCxbUx>+{r;&e8ZmGH=Aec+qu5=9qqh!?2Y>DzWp>i
zWZs-FX^*b^toU;N#Ln2SOGDqUVz=<G93A~Y{ou#BfOCKPZye-+Z|x$IFSIV`8NE<0
z&5x*OKG`wzZ9bv>89mb<dh}5~dh{8cZvN={rC0P)J@nW;{uSCcuv6?4`y<~zhuw0Y
zfZt~KpnUe4y|oX(k2L$Mo^tK^1NHE)K04MnyMQ0#)J}cl!>@gWe$-Qru6>sN`3rJ^
z{J8luZ`z6LoUcZIz`lz|`I%Uc{}4~|SH>&;FdlYNT+6RPakBfk{F2ih^I@LI9UtUq
z-1H{Kx5g9iDI1@0xZgn!$<6%nZ}c8oSK>eG*tw%~a`6v3);T)#kY1R7a-bjd4AKMo
zK`!jOe#B$!vpA4njeN;B4s_HLC-N`kK>x%~?3Q@R`0!8O>g$Kx@C)_NFY=${uAIEl
zBS-NXePSogGx?c+NU!Lh^<|%dK7F(<#T9Y<<V4>gxuhXE(i3TVU|s50fB1HuPyW^q
zJ@b6Qc&%%3hxi74d|0panVh@*kG=<@op~oWbon>(vrhP7@ddpWkJ1DCB67hG{-J!x
zpLO{f^?&pjeR@UzV9ZZiXW}_}Y#icZag%fsioc<CLmt)xv>%W0i~Xm4cx>-{%(`M9
zjNAN}XY(25Opf9j^G*KZ3-T~7{NUg76Y&>)5PymD#C6gz#)s-b_k798cSXd-;y`u~
z-}IfGF)sGN`jbx{)|YjJj`axTTc6ggG`pd`^~fGXWcT<(@r>uF);~L_KXDvCue~(;
zq8-1lTz&m0*FV0DgMEz1-r*~@k97D;^FTkzO?~{!N1wekE^!Mv8J9FV&e4qDK3m*j
z{L<R%-@2kV=2bebH~Rp7S$rlgV3*i4_K3f?-sD5;(SE@Cm*%I@MNgc-u35j{L$p7$
zzN$ugWL)IL{)-DB`)1!{y^GhAkUe6joKuhkWPjOddPx7-2mXxSThB2qj!7z~m&#*(
z`T93L{qWz`5B{tx?TsU1ERXA!J(CZuJL@bhom5Z%#%I0oYxI^M;&&kX&rg^?`vB|0
zI)Ki*=#eyi5>K+P#vRk-8OOoC8Mk>+k3Eq#etJa@%r}1%{aK86tq<$Sx?#7_VTbrV
zaey?rlc%(KAy@r~5A_GplU7gv;xzTxQS+i-e2A;ew|FC>ctM=%{7Brv&OvsNy){01
zL4M>UZ9c>c_%|=~$@(OJ^~EFTNULu?@xi{+Z+?JXXGiI|{>f9EN8aWIinomepX`b8
z!$?OO-}+^LA^U4x;~#1V@uy$%>H6Ezb8TsSvya+Ae9}MtK>mMWL-R&H<j#K43wkVG
z;>XxScHBA;x3PQD>}ouR<WIz3;ydx1{+tIWH^215{*^tY-_{3xF+O}brxPDKm-9T>
zb7kMf@|@c9dC!BP=f|E8L(h|yqvyG~^SO_|=cc3f*>ZkIJLiS!>1U<SFE{1At1L1C
zwa4EIXHWgrC9}qi{Bo_|ec<Dt8#^NHJGjvEDE^6j%saW6Z}UW6&L^EaN<;a^^}a{G
z^3d*=jTs>a_}8Dk;eh>K`j33$SML5nyf;7|vA+J4N2E{ABk3i-PtT0gb9?K6-jW|W
z^3jo|uj1f{_!;79pC0+)Q=Oq4hI0739<WRHLCyo@^E2XV<<iR8d+XdfPC|Ch`eqNL
zm0Rb|8Q4jDi1WoM;w$F~>an}pL*q0a^u&++G!zG+r$6%R`p5r%{mh>oeP>7i`ay@>
z_-*`raIae*JAB-@5q6v(f$T0n<Q^IO>${NB>^c9YoqFz(q3gRa>LvMAew!cXuie)|
zmtWzpJRjpvwU71OvlB1y)9wcwH$L=-fBHkOw5MO_;|JgJ(T)8?zY^2Vqhj1A{*%vN
ziL2;^{gC?h8R}~<zIMJN-h|4f?R&)O_GR`#o-a5@voCT^Za?Ha$9_rLep1^0!}+>%
zH2L<U^5gRWzw5OBZ$G-&7k;{Q`{kG1dCHTwEY*JJ&+ol*=9^#A_FeIKf6Kne{Vw-2
z>`(2x$x$3jzT`<B<YQi)kCTh}63?+C<RR@GM1RgB#5L>yy8zh@_4UI}(I4j|>=1jX
zf93k87wngDI9H$_<YND!oqdV=(&Xd0gYyIE9Mrj!b42sv9E6^_&q6=Yr&sjNd_wzm
z^MD`vU_SMW4n5Z{)}s&f7hld*ilTn)<H?8KTF>NTe()WiyWzw6fc1}m>k$9sX}pl0
zJJ(gu{aEXb9Py_;JL0_{^5;M4H~%hvi1yol)jq-f2WkGren?t9Y4#nuPvLw~yb_<|
z@Q3`WeUp6=Kgn-H<@~Mo{4F}@M^vAG<1hJhbkyVT`7QPM+b9Qq(fN!tRG%NVF6a%t
zu&&v27|)fR3!)=Uuh5I@!MURINB7z26?<=;D`&UNvwaV}p;z{y;)7UE+#xO?7x4pq
zLYE%Ka{1_(Pxa76hyKR-Q4c-y1J$R`#&4YTTinhMiMOo>?Zi>i^g}<!gCFsf@zFzi
zOh3J+Ykr+SvoFq_>33W&=;@DK%m;bs58_LE>&CdqiGG-mI9}scE^WN%if^q$<A`Ye
zqhnmgp?`8lPuxpR&U@G)=UDa;(0=2>zdgOp!^bQ#V!g8K*0puXZ}2<TDSsm_<R6u5
zALAPF479%z*Ek;&pFn;-o==LK#Sh{VaZ2omzY_N;kNt`##6J;@m)&zu*Zp03?tNLm
zFXY~a=f?4US#}LQ{#-mIPVgK@xj2HI=6Bg!s2#r9VeMkR`9;q<HNWh7ymv!C=np#L
z7JP_Llp6;<!k>0fef9L?d|G@Wo)JIE7q57ph7bDy^NqfJl5$9o#69Xo{gI}BU40(C
zpGD8b>-a;DervB^aiVxnx$$VPp7mlK$v0l*@~tEM*&j%I{-&OE+~49$JN#Ss=vu$#
z(Kx#6bW~pWi*iNRIPe9HGs-{Kw+|;L^MXElaXjpZ`r4C=@o5+H#V_`A&X?@xoG;m5
zil^-Z?Hld$j5m(Yy0X6_XK0+}jsC%S4w9C}kNv551OMbgAMj1i%FR3ZlB0b3z#d4G
zD>*qYv@aoV{5hZFzvzQ`P)~dMYrdT)nqPj-x<C(~<j%gatJbgeO%B!vJvJ`-VZ8Xj
zuX<5W*j;=;agjJgd<4ab;z@ClxDvV#W8Wtph4P(?sPB1?eEfTkX<XfLzw&c$-tFlt
zJL;J~<*`1!gw9!DjLYJ=r?^raDLxaIi3_F0pV~ojpfoz#L;F8*r2VG255{t7{h%+c
z96kKXhxCFUrH9h=)OujQ>`S0|wf@bEd9$vpOY=bwp!tpKm3~C}`avIm?1BFH8)@|E
zGrIbhPjAI*?#+njd<W0_$bRqO#glhAWZx;{y5B$a`}R8?`@ZonuC~<3t=o0hyl3LF
ziAYcFckD0aTfg?Z<YykFqkPzjXurq}eft)6!T#C0&_6k*`HQ-i(|>*ivcv2>KR`d&
zH}Yp!*-i4X--hOsUDb}=wVv$j%_sj!-sZtP^FPwotM~=~#$!Fu1AZm0SN>Rghz>i;
z9`pC=(JOSV%Q!yqgYiP+vHp!yJi|_j*UhK-qj%0D&5Lse`|^n9$$E%(Tsga|J^ISE
zmsXyJ^a84{AM~VQ9G7}Yh(7((5B*Xe(R?)x=@t7<zvxN7F#6e;7RN*96XHnwaVXxi
z-<OX2eCG?0Uvr*-zWu*_IkbNk=R)fbzs{HBORJxR*0cKdHL*T<$rtB{m&9ex(VX9j
zyR2L1`r;<&dnV|j=RD9lfzB(fBYsnw{Sj~KA3sq4-}v(JufBTvxDof@#Ba_C)iW;i
z@TtCh{HU+oy*F|Yk1BURZmVDX^x@g}jU93Bh%UZix80lW_{)FJpVe{BDSpw8J>u6Q
z@?Z9+_O1L56wmNO;z0gLT*u#t|Fjo}@~iBEb9?#xCI8RP^8fsta&aTO>>j1(WBj9d
zf!|jjU3z0Y-F@<zZ+Z7GHov8Vf932py%Wcw%ii&)<f0#XBJDX!jCZU9dMLipUfe-0
z_%Q$YG;Z;Sagdk(#h>^fXY<3Zt8bi8{2`4_tnb{^`AVdVU;AA9fAq=K{6qR}zb{Ur
zx7yR!m=?!Fdai$ZKrZM?(`#s6p!QI`yl6cdukqjm-}*Hl^2LYzm~#*EpjYD3y!`cA
z-~8fNJer2$3iB7`!7iCk@hv-O{Opi)JU1jydTo6`eCtnpdd?oPKh~rEj9*%N{X%j^
zSDYxVUDVSk7y76_e3O$nK-?hi@%+p4N9$UB`#gxQ{>&5oh4>>+`S{abenj=OCl~UP
zubuiZtzP60J^AQB{TMGfK;wb(BbrC`p!+YL`+MGGp2TmS>zEJczw9u(f^YrfpFTOy
zl`qZS@H70c{V#n=L(iR{_yjsHvk#?jN#jFDJ$4g4&yCQNPd?&gsGjFp>dR+OjGsNV
zU*@0WtH%zrOX}m#IP_yY^39|E^`kvLjbFS+p7JB&BcgsFf1$nS5YomcjbC)}3(+Mf
z^yEY3(&$0@PR{htdStKZebi_6**cH*fxcUh{43;_p!qV7#zj8(Bv0`uzM*>j6}dw5
zN>Ayt=eOpceDI|nBu8{1yUGvRr}1mzaB;Eyi~UyI2Z?jli|2RFN5sF<MUnsV{3!0b
zod>vo;2fHL<?rm<q5ZUT1N(mWM(qFPyBDB6d+lC>{Ky}9FX(7~?N96*vyNN)SU)L0
z)=Nv<XC>9cuX@tz%a=~V*e=#%2iYO^%0AV;iv1F=Sl`gPkH~+B10nmN9Li6tN8har
z`YvA@^26+(xQHDWm$2{R7UPBLDQEZ5V^^X6$eDfSuh~W8cP=NN-6a?DRUYZ`XZ)mo
z%#(KPIDaAz5+6i4vP<q0v4853uX%&=^@G0W?dn1HN`3XDmFri&dBzvL&_DXlPtnm%
z8eM7Yz`EhL%@==ZUc@2v$o#9<^~0m@gQJH}@w7M^YH$9e&A)oqujd==1av;;{EvO$
zAC1?#M~~mOPn6G(YiFN{u6*&3cqyJ^=}&$9M#R78KF)je=ls&SmUsZ)`en!Q8<D-!
z4#smidO`2ZCp+PsNt#|j`RJ09e0ByM@-biJ@4l(~sLs*I)p*RC@v?tONMG3%`W^GF
zL-xu!8nmv|qwmmqi*{2#=vYtqz?XQK-Q&-!56J#nzwDrP<Zd45D|+N){-Ao&<V^06
z9$8P&e6dsPBzciv+@F)1c2Jz?Ig@;L5k2+~I^WaI^C)qz_llG|M{u5_K2$H3ODCaz
z(9sX{{*tu!6JVr^4t}Jed`QmZC=ETI(~owMp0xUrp8Wtc4=}dZ4^$rcz-JPge~3T&
z6zS=Y9+0!N`udCc=1V>IbktLBJmf)M>iM2CebygDS3P{|7aeKsp#GqI{E(OVf%t*e
zt+;@_5r@TnGyB7yX{Wxla^u3E^<q8PcS85z-G_2t-tWKew!)#)?r4u|pLE<q%g=dW
zY!|%;s=e{*U%%uEm6IEJn_u<ti$27s{v$@d?YE7~JdryjM{<Sa%1`h+{EdA6Lw)gB
zjL)U%0mKKSFA>Q(3F)Wu*q&ciPER75C-V}~{E??|NaF(?cEr3wdMHgF=z;e5Ll+u{
zaTzZ@V*6NMJ^nA!MGvZvj`~o3MD58bVj5lJhjBin^=mxRNk~2s^<#f$z9U`Z!ar1B
zx$%kvtvBWDp7=(4c7%Sg<8fVE-|U2P_CB_YdZ%2z_k!4$dAGjpphu2dq|F|_^^Tbr
zuCf2Pw%?K1<FxyyKYPTY-E#H(uCg?guf4SLr12ORzF@3RUJ<P?`U1_rwDqTb?AQ6T
zb?1D7{bJ{oJO7OHt)6`=`w2a#l}<x?MxXHGyg>i>La%Gj{XTB^E*m?~uA>{Dhv>&T
zVHfB<{)`tL>w!F=`#bh=-W#-^Q$FQ_*G~HC1>?Hkp|WqZzp(Fcug8AaeJSUm_QCd%
z_Jf`yd*1B62fF&PA9Y@DUnFhc8uRUsl-p<N-*W)_YRIq9+a&##57`xZ9}zvMpI9#`
zt)IA#*ca<cxi~krQx0Q0@wWVk>>7;a=tAZAOhWab_Ru&WIg-12Sux57y@=+6|B{ZV
z+`d*CTDM7l(|(A*i^$GEe$9H1X#GR$Up;B|BO*Hn*(?0xSNs}LeW={|VLaz>9%B7V
zJMT~*q7&)HeC3clAUQ}!R9+O><w!TCi_-6RJAB7}<-=HBmo&fZxwYro_FGA)9({)D
zi!0<iM|8i$d98DIdhEFlKgEwJho1YO<NOlRZ~Bi8jQz@o;zZwN5=Y91F)n1!*<Joa
zTKudYzafqmZ>ncME*<-gY5dryM-*4GZ_)nI%eX(6HlO4h<=rixdB#m|`P$trb{_H*
z>^45wRsFE{%Gp_HUY*}aJCBy8CsA*t^{2jjQQ8@&aXGgpH|^=GI4;_4_e;>TPVg_^
zx)BG9i`>sf*K<*6{L*XVu&&W1XY+=xw0hz~aiVpBzBKf_1A6`=j`Tc992$`vO2^TC
z20qhJyh2aut^EdcZ$!Dc#(rMDdh}VGLa(8G>w<om4|du-v!nF3YmdYC7UZKxkEQ7^
zz3$52{%mx=*5zX~UwifSgO2A!)(L&qPTF{5JMHCbhhKI|JRzTcLwvCdp3C#g?pvWJ
zZCvIhw#UEuGfrqds|WEL`^P_Ir=fn(F+Ox4zR4@89zM}E5A>9sbxvemK>Dq|G?WkZ
zLth~|MWm<ljUz4JxYg56zV#pDX>qmZyb(PwWN+=`<wMUI?bjo+hY%fzjx@wiOtY)#
zdA^>6N%iEr|0nIazkKH!v0wGolZN<F4wdW2IH7%&^Css!^qqcLr|bp&Rv%hF{2l)o
z{UE)A%GqD>13OHwlxs)t;<}WtfAz)5`0Lu;Q9CT&)sJ;VfAN7o`TVVP^y}gibR(h%
z^@C4*NaGWdllTEY_>jh*^$EpQFxHD{?I3@`zS0x=!~VIa#1F97=F2$I@qVX!D866i
zJVm?^`{9?+V;|`meRuAQKlYfuYbR}-`n7(b`4dlw^Vl77G{5Snp?!<}j{f|Pu=gF_
zvdu~7zGVHeZT-`G@*rpHoBf8?uXADN%+7_$ja;C6cJ9&fQ_juJ4|>YU+q^novtLWn
zck(087_XDN{VqRY9_YFCVt;|2b;ZwFN6w|uwXPzfkDl{p_nM^L^Agv)@5Nqw&)4}e
zyDbg1=Qq%G4vwC4a{2g)bjZnhIlsvNDu?Lc-}v}*@}P&-H@%~m<i$Q(7vxXQ)-}6l
zKc>EQPX8f0tiE}~Cp|H4a&`W%TswTyM{?zNlh8at?Y!4&p3Dn*;)C4f^Y`Kh_ax<;
zfBm5^Zjd&A5dY}7U+EsAe$ZDh=4%(}CAIV3ZbbV8{+|C}m&76R)uU(Zw=|@W^pL)Y
zANVVFmmL$w%eNn87wv~4eSD$IenNgpoMk-H=<~DI6%^MPAGxvX{Hgxwh53=D2NA6c
z^&!9LT*dcUJ?HSg2s^J_8hY;H`$wMNI;V3U;2cDM_<;1=`hwP(amt7Cq4A@ehU_If
z#cuISo-^_%>?0JXYR|5jZ~j}FUeil@pkH<h9q4@<{*K)f2e51Iak9hwCjV$${DX0^
zv*JYg#;KhDme0QGPd~=R?i;st53Mif4&n`RXUEA0J@JP1&mNGMxXS!km)5y_^0WT%
zVSPh(L3__5&9C_pf8i&#Lr*`}we>EJW>?q;<A(euIgqb;qL1`RxpCr;J=70=*f0FE
zOHe=Tu6p{RkLsbXzH#8An;!js1^+Lf9vhGG(I0f26Gfy)^d4F-QSX&QdM{1?tta&P
zd-|)sa`sqz_4FU>p^uJrDZUg}uw&>;vy<Wz{7JLB><NCPjT^uC!Kbu!#14on#2M~`
zh%eZ6{Yx7s{*5ojALwR9@fi8&C;AQY(LWU58wV7}8He$ShrA~u4pXijIiY7f;ym-G
zp6>)n^KZss-i+6Hr0I?J_=EECr$6$C`i11AANqxV_kVnch<v1>?-+TXmfWO$$I17J
zd{@YOx!T9~b@k^vN4{%hK8)YE$w@pxzR-M|C-WD-e}ixQL_XAy_0Yo~{WCw(5FgU^
zdE_r`AJ5O)@AF^IwZ)y{WalpYlQe%P&W-uvedYWby5jO!Pe1DONBkx~!~gP2%K0&V
zpMT;n*eiYmDu>#kD{e6!s6XQ%7yYBh&x-?`R~a89M|_Gqj90#K<6l0+mo$0eGtx6&
z^`-f3ajf<b-{helzQkkZ4Sn(<U-a~gKlO}<e8n^5svdgy7DtFnyL^p4$3ura^)KCx
zw`M(b%fwA5x0IXL*e`neGj95WKlRXu;v4afcopB|r@ngPVSI_l$sgb7DAyl)P<+|N
z(ea~?5A!6RG#~VsUeI^*OuqOtPJG4u$d`COo=1^`af(O9C*n(Ssqv8mInf*Y56G_a
z1N;KNC(dxbXaCDiv&YKCBicuF?j=6spYS10g5ooAFgtI4q`Ujj(eG*7Uy{4}=8s&=
zyR`Xa*VqGkVSSLV`EZ__hT>yr9?&s=^hR1aH1APg=#hH%8?iol7De-D-p!+VlaBgC
zAGCwwJ$j{{eT{KJ<EBUAdj0EHTqhnC7m1U^D^UFAdnggbY4W8bDi<$`_k6EYoTnZX
z*NKnt3+0oOI7xhGeOgcAS#mKi<SVV7eUSZ<{7Bcl(0B62r}H%LeR#g&-kj%Y($4AF
zRrV7<)>{%f$8)~t98bTV_lOtS1Lue0H|KofN9TO_^!&^BVf3q>ar-{H-_`N^QOf0`
zE8n@L@6|Y;`S3Ac-1?5Y=C%A@kMS9==XCfp4t%-)B#j^Q;Ji~m?nmm+y+-#ZeeWIO
z;|p)t^yy`9TD%Rt7i4@B&RX@D3pbz8_S}yD<>$m1?6&!qhWs7-OmEDmdzRuXc3wQk
zj*Caxe|F!zK>C1ha&taxKIkzyniunFezaGvALZ_o8i#(QJ)gby=6$w2`>e5T=)NfZ
zB7gJfdF=@|PMP@G&X~4&)gQj0^@BdT`tdta-p3ld{c(pacGvj!Pu86G#z~h<XrKA3
zPptE$7nf+$lOOH&hu2?q;^N)kN%cFbo4sx0>zCPQ$@W*@KIg@4c37(0-tVbuPyhb3
z`_Z2qchH1x{eAYn{)LfuEZJ^-ami~A*>;KcyJzoy<LM_a)xP_w$98_r@0Muad(q0f
z{qm6|yL8d}+%7MB>ad?L_D}Djd4J9RQJg_P!~-!d6fZ*c#L4s!itD4EDyNsy){A!1
z^xAsRzu!k=|L8G2vwu}zTyK47$6k8AuD<<%_@CXzxAkkC;hX=qp2er#es}%+JN7?l
zq=hc~!hV|<{5TJw*VY&N$-dEJ^P7b1B)f$kd6;+OH%{XcSMj^flf<*)YVk9FEPm#9
z`9JZI{Fs(6E)!phi|{9&6nC<t;!Sqg{=_`eGxI0Dw(k*F(_{TPH<v~~@}WQZ=xe8(
zo<zRXH;y<ye0Rq?djAT2@}j5E`nCTu-<}sa$ByTd()MZQ(fpV%`!46k^dO@8F|HGr
ziNDm7PyZr4{LnZ07*YHjf6s+ntXJ}3r`RF(#Cmm~nxAG@tSfN~d#k;BT+Y*!+xJ<I
z^64W!r1b~!CGN8B*%@&mzQud&DEq3tI4h1vyd@4`uk=G6>{%R_diaUu@!V5deSC>u
z^pB3Ta`T{H<5my9((1|AFZ$+>-=HV_mibd(n!m*ly5c?aNH5H@ex&gQ$r*k9@zeC1
zopCMz)iVy~c(I-J5!<P!Ur1lch213=<3=y)yEJ~(N1r^{H}s8D|M*sJ9qJcd^q}_a
zE&t9wpvP~J139YC&N~kjSD-5{;7|Am`Ru%W<5O-t_%SYgsK*~gI_Tq9JL8u&KKf!_
z^e0Xz3;7*!1^aKG&tKW!J5P!J6TjpW$H~61gX-}=`Za%0KhCH0Prl~Ud^^92^0toY
z5sdR}9<`^3Ft#&)))Bj6{jhWPsrrNFU7SEJ)|>IGXZ+$8e4!_;oqQPe*gP2*KGfGQ
ze(0yTO&VS68Xv|@PV5AG0NGb|i~Kydf#eL;hwQt3ihBI3eEjf7{EhL@XL4X)`3d~Q
ze$Bt}@E6)gd!(G+nGf>Ahj!#lj`)Mh=~2>rs&9Vv1C_Ho=(0!J;e$M(cG8d>^dq0#
zwTt~kdDB1hVII2c{J6CbTj9$2E$1N49h`$Wk8|GPe!6{|a~b{Ohdj-X_?o>GpNoT~
z)f2aiw_~|_@;x^fXS3JR=&{q{b9U8vx&4asBjaO_#c}MFxQSe?E9=tyTi@&@I}F)p
z$gW3Bs;^yleT=@}NI&TxyA$<Ez1Uuw9@0DGG=6CP@jJ$ij<n|{;t~6Mb~~oU3)a1L
zi!OcBKfP0K9`J3x=q>r!&(RO(0iK80pGI^Z!2i=D{o5Bh-{*hr<L#@_b)MllfHXez
zLr(lYeW4HNE4MD7^<muBlXU^@gN;+#dUhX0d}LjFj)Fe9*smKGv_9zvzR8CiBmel+
zPekqM4}E5@orCc+{H=8X%_liW{WU)Q(Pwdn`Ld4bDHN~JV~CD*#=cn3)+MyQA^xo!
z>w!Os7~5$d+mWO8((36Cil4+u>;ZjcAEhIb4|$P~c~LHnZ}W^#a>H-jC(s}3-@3Iv
zts8zLo<q`8ek$rGzXs8fMxUNrANXJgtRs47{ix55;SWFd5fSyLALBP3_DDbMkhFe0
zH)RLd6-d6u3DLtpWDnRM`4Q1E4&yftX&CFtN0)r{Yh6S7;JlMvfY!VE(0JHq{YKQU
zak01Njhx-PgzT=k2Orv*C-Y{UQLl`PKG0M8Y`yVk`Y{gktUh^1Bya7^m;U$_<Kchk
za~!wl_@0}4zVE)MbtZ0%_ZHk&aR1W1(TMI9$k)!j1!<@sd_nw4n?L#1k@=T4e);r-
zoZQbaPWLhJZG7^{2cPm|T05v7IT(lg7y5_z$DjJ*vn2n)uJb4SgZ0Kf(QkT3AB>Nl
z(a)%7^0nuG*+KpBXVwAackmlgJ#n&mjPpY7^7YSex$j~<@TdHodgP5BeKbGPYh3^N
z9VShFvEzPCq))%d2R(W~@7XQ$WIp+8`%!U>dFPisul8J?{P=xoC|-f|g8h;vcWH=@
za{a{eZoRO-B5(6<9QcW-+;dRqzCwKN;W>lzdKl};w?Bn3ZY1}J=0QF_p?v2>(#nlf
zT0Mx5q_lI1SWo>Vq+jH3zi1vIKWAUTUhv2CmY<~W?1gsxzV_^c=aMmvE<_Lg7{9SQ
z5I_7TJHRjV<LK%~oJ}s|AkL=m<OkU`=TP(;+ApiGo__2@*h&3C{NWcGukk_lT-rQJ
zv)|?eJ@W*`8RAU!pz{BZy*rP#tE%z^oCE?$|DcShAYucClA=ut6mpJD2?|s+fJS+O
zN+Z}|1B#;HKs1Pe;)}|&0VPiqc@?A)p(N}JK2ak`D@1yTKtu!;4Kf7t2E+4h2NQPA
zJ?EbG-uX{vRyWsNzcstH_CEXC5B=Z_f9>Qjukn#uuX=pR&o9y+^ROr5$syj$leF{m
z;;MKjzN!~*#0~h$N6imc`GooPw_o_lN6lk?RJ_oxzjJ}S3_tbC`dg<wN&e^eFz|sl
zImIpf@G(B}Grx9t;m>~UBfQKbZ$*taFC3KRd#HBnCWrm8F75cB#v7*``KzZV@(uZ#
z-+5`bkLvZq$9<4_jMI)kKE~^(J?^LTiF1bgZDn?)%>Rhn{2@Q4onF&Fc*z^-f&Hde
zsD0pv`3F=yFrU1HpJm77a85MN{U(1PualoRfAO#Q@@wQ`-{gX)`B456UviM&{i!%l
zF83_*0(qW1S6s$dzUny;eki{9;m=>nFU2wOM|^VMr#;@oxxW+FQR|O>oxg{l^@ywT
zC)7UEH<VoX@E`mJ`=-zOnP0te^ba+jeD;sNnjc=`I2v(<pFAnz2)9V-t?^MFdV%VP
z!j(MqT3J8y&`)J{kK(5tjedb&p}*`+o<lE{;YdF7*?;RcKU%IMoG0+p9{C#2-tni;
z%I1%L0#30Vp7<MQzxAh2{DpnzU-TmfKFav<1N4gBq3n;IM@rAtvlDW$7w!BvJK@*G
zW%i*TI}`8Chr-`_*p+qiL;RKJP4tQ%^&Xt(tJ?WN)ca!2dHkk)8$bQk^E>p4Jp6Cf
zzH7Ii_MQINS7qzBuk->f`_I3B+A4>hSBrYB{eh3Jf8wi)sU!Ad*eHFYKX9v}@Qsun
z!IhrT>!|PY9#lKN@TCV)Kj;n0&*^Wxbws}O2qnLMsCxXAQT6(x^h|s7Q?Z{id5u@!
zZM~ugotnqKqod_Zj#kHwmcM@HQKo-Y<=9`pNXgBfVtcc)@y+_h_9%a=?XiEa%ltsI
zdErw<<ssr7JM|o!f0qyO8~h8(FUXJdL)j%t9{!KLYiHM}{^B#L9v=J$$}d=#vU&JF
za--%&t&f~=(;t8RBGu1$@}cC>PuaTgw;$Gl!hv3^CqKU4f6$-ZnMXhJlMi3xB7g1p
z*dJwinQ_(^`CA{kQT+8Mr!tB!KFZ{<E;NoapZjX-aNg$E?4$gPo}<<&Uxl;r?vvaP
z$V*Z8Jnl*Gp$GWMqva3m);i>CaDb!xNcH4}hkp1g!^1re8t;4J{Sf)c)r`UszVa6N
zv*!i&N#3li-TSECbM^c|K8>IG_%rkA=iDzZ_ue4B-kbFMIlrrm&vBIX*KgUpKi)le
z^~Y+S6M0Y3b29I7s#o@VJijmX_fWom+xs^?bI+;uE9YMKpFe$QmGbwles6m1om=m7
z&&<{8%6^aP?+l`T*X;M7`o(eYy!qVU9(>@c<=<~LPJeQF57qBy{XW+3485P~@4*`9
z_u+mY?{}@J`QQK-zhhLdY+g8;U%h?M-|u>z<Lt9_=<jz++MVn0#pfM!_FlT%CR6J6
z-E+YB`zXJE!pHrkdw2J$?9csdypMH{Tt)Sd<K1ubTl@}x$Zw#YAL#G=@BV;)<45p`
zeny<($Bj3?c;cQD<=35q**8DpJ!bbn?A`C!+((M<@)LfN-y|QuNUoTVx%ahB>(|e{
z4nNH=!be`_9!<MA0C)bCy!h%b{<$wEFaFje&Y<qm^h4nbr&tyrQE`l((qqpVyl3Kh
z0jl2n4C>>14ZgGH{RZ<W(_{R+4?(VIN4^uM-g^-E(m#CJkGLl;qWD;^_bIGPnS7}A
z8t*=fUb$CB=|kLi_Ct@^6{?*cvRitfp8nH&?evPC=||7$Jv)UTKAv;(AM_DF_M;#E
z%I+2U9pgL?<oDpn&uRBO$o{FPr*K!VpL`m156G{`D^TYR{pA<#p;7*czcL>^qaV=@
z=#lk$?nQ6u2|VqyeYpPr>^JxDx2#@A{Z2%j6VH{!J#le|uWk0j3r}0MF5b#ljMLxu
z6qLOuBYuj9@;bj?5+C(9-h1BW^BoM|Px;`&uRXf$v^DGUEx&hS5A4M`*Yl{jpUUz|
z_q*~z`K35gr3dm;`z)_iFCL-nhyK||_uMEwr6<l2`qOjv%8uv*dqvrk{%Ev2{zF`$
zPs;l97vdX#z+TDWT!b3WFOUm0zkd9n{S^<LZ^SWil--F}?)Sx0{lr;*lYfqWJldsr
zq8=3w#EEF<?AX40USWUW<2)Jl0G`Hq-@|)H`fKO!_({|{`5p4AN9n2kK;`G`24x@Y
z1?BhHhu?4TquSXYyJ6SPWzJiE-^HIeCpy>R=iDb>bk9J4=ruj2XY|VXTN#xn$xrds
zUg~4xJB#Y+uir7inV#4O@k~BLujm;nPr}DJg5QXgKG;|BnVr!u)Va;y-Q;hb>t^o%
z`v3lq%cs`kd6ECNFYscIaDx-Ogp>a4RK9FHd-hx^o-4^Gu8A-7%XwW~BB$TedLNS?
zp+Dxg-_9fQW&VsmbPnPV<u!1(etKaa_yd04IDP?t{#HFb=GVkQW%IxpwI9}ndfu*I
zTtE4Z*KYqJ)$SZ(|MY{i_ZyV?5&lKJekgxJKkzlL=cLNkiOT2bBR^tZc^=BYXqRWe
z4PWcTA0;0=V%hlEpM2t$xNrZf_MPA3*X)~n3HfUk-sFQL3UBKqpLy($@zz!9Pvd(;
z^bExZg+r7N?(9{+7)O-lb>fFSQy%NPIOdTj+IRbH-Qt=!C|>a=_?VyH5wGA*F8%S7
zFY`<Agd=%Sd9-!$fBc@Z{9L|EF7v6k9^buCh99}$X5H#h?aKTye#$6(Q2qF8`2jiQ
z-EbtIdC3*mfgftTvUySR$Fh3!7>~L)be?i=s2+8H8Sjh5HTR+7v+o+ZSH)M{HlH}@
zT!+Sf?n&Ky%KMx@<SD2;(s@K4A}`S{Z}EE-W$nHX>pn*wF3)jqqh3BMPdaLg)t>e2
zV<(OH-mEgY^pk(fyHNR&GP#_$tyft)xy|RE%K04*>h-T`*U!0+T;@Zamm-yipw4Z|
z;sokGNIY?mAaBvmFZ28IW${?t<A?Y~aXFUx&oXW_zT@tEz`ofN`|`a9{*0dTckbcU
zyU$S8Pdz^TPV|5L2s`GN$tgbY+x$7ofALTF@we86I^S5g{_6EZ$<1$>$Nb`>xF~Ox
zzt|W3?5B8&+E?p0zc?hXlCRSf{KP-=il_9zeT}??{>g{wG3xw=(r^9c+3r_ky>|8N
z$az}+jk+Ipev*fqpFLGkdA$CpbAt0fImks$a`0dLn%_f^uS$OR1<r-a@!mjLKYkxo
zZytD9H$Nf|RfZR8z0RBXXh*d>=fr;M^>?pDAKZtaaKs0NGd{82zM%BbJkD?YCp?^E
z?X&Ur*ZNTP^p;%YF%Lc{eS$mx?EGgRB6WUH&)>&&Xpj8O19xR|%HN}(cOK<0#253h
zf7HCFdC0FG-o~NMALL?>&LMCl2m6S52K{#*AN9(7@)&W0-xDX?8>r`hW0_yX7mnf&
z{_<e?JwDD0)=3}WRfV_twA&~CnZB_D`%jPQuXTzKF<z)=FZhwqInw!+-q?To1t;-V
ze1fa|L>v?!<=tz(XOB}}`Oat5m%Q|n{l56Er&4)Wr1C;}qx?|b=RS`fxzCd)8z(=d
zU-CZV@HJjNKJrEN`ZuHQBkec**@=4l!)~G;spseTMf~)qPwZA1Wxw(Vc8k(4agZJI
z>*64P$$!#E{+*x2kG$k3w{^pTob*pUzlo3WC_bohaG}TS!Ft&p{bv`kA3NtC*r|EQ
zjaolSZWMp^Tk;LRgCVDW##xVgb}z1=>=`cB>D<YWMQY#dC!G0l?bbuzyieshhWl&h
zZGSJo{jtAq;M{=5zaO9<>KyGoA^lPFILD*T|GwvkpZA8G8~A(Y7td4O{}}JNFaD@=
zzvsE`58U&+SJ1A%dhc7A&v}Jk6`$-Ie=J_PcXXcce8)bx=Wu`F{)3#>;U43!yWjYc
z#jl%Gw{O+~Pk+||AN}EkA9?Aod_Y;8frogDuW~#u(YMpSb@~ftd|<V@b%;OCan4E3
zMf8oF*2iye^xl8@!LHAlGIGyW#~u5Xy{C-CdxfQk?sLo<+fE+Q@B9ZJo_*iRQ%3fD
z|MmBOW|dV&+&35}Z$Z`LBmRg};-Pl&M*MScBhHCa`Wf$DM_C+G7H5su&i?reRJ`Y(
z_$BYF$lI(#oF=b(C+i@eb*T3}9ryU|k=+l<N8}yuHQXDxuW)X6U&5c$KXFk$WIfu&
zBY7Tw%fE_G>W#x!Tu~N(jTbji`e{Dn-6v=#C%)_g9_4x(e>a%@xi>Igybw?1b>!ja
z<t^mkPxwvwo%c=oG5f}^dj6vfKld!;B%k&14>8Zz&QIe*53CpN{JnnU<Ja6@@ee56
zwY%@;SDfRW^VyF)fPKiL<XO%`>;~T2QF)hqOa7%Em9ItWeP!=0@t59PM%|0SQ(Wce
z#a;dqKH?2r`Ca#laQ7Tr{KelrB8sp55s%?79vDv_$tO?X-|<EDGY%i^C^^h;9jJa0
zAN=JB`ibkt!<jw6UtCxAyoJ8G=VSNs1NiHwo!#<-sCb8x%X;mXJOU+`dF(eiw98BI
zv+r=FNAfoLgMM&@pY`z*sCmRCaZ3D<x1;={_`r|RD|r{b@EI$WPm8DggL^yvKA!ud
zKaq#a@AY?oz#pLI!H55F-|4)^5Ai$hBV&Hx9Iak|?_=?Y@(J^ho1fI*eXG0yKks9Z
ziyu`cAActf^P}*9k2oT}!5f7qy!3P4Q;$D<$;HmdfgegA@v}bXD}2R&<CN*IdjXVQ
z>8IYh;R844Hut>r(K%fnAP&pZOWJs@WIlP1yvF%jf9=+3K6ui9l-%&ApZLRD{FFz?
z@8oUeIvRP9=PvRjc_jZuZvKG(6z};1eEDg9gkR+cJr`1MUjB)^<U*~RpXcX{vp#x+
z(hqr2#9tZZ2jzeKHa(X2kRM<BZXe~5^p0NhGx8Su1ZR37KXXq3U-DU(dmT8^2YBH}
ze)t>5F2xD!GY`J}zxW1!d5-b+M_i-_@=tnTJbiLr!5?4xsDC`~(--??AFRvz<-PU|
z9^@jobyz=s#_^Zx`6GCnS3CWn7xu@#(^K-%XX8-kJkMw8CHeIue@PqfJ;BL&K_1Co
z<zaAkE};+fMj0M(8*Lt7etyHc<i+w0@3DJM&fm!U<YmUmpHTPQ;<9)v4x`%j7r&j;
z^{ZMRJG3wK7q0R%=Wpw>f9wN}_8+xx^oBnW$Hh5uMEnvr=%;b^jel2Wm-d<dyN|VR
z<kXMfMD-U>`5FGL)YC@1B$syQYE(V@(jOmwpPgH;xGldDSNUJ@GUfqr)DNDhcH`tb
z@)UmFeC8)ltT)bnMY)twaRXIvym-T}&>Q|I&aZ47y&{*geT;b_+{HWVgdd#bEArS_
zw!ik%Jj&!JC;gPa^S9<zPp*gq{e%-7?4Nu~nO?yefB2J&Uc!ak#=}QH@q!(bgM9e#
z=Wro6iZ8pNrzpPmll=DCKH<-=k&nE_M@kO$aJMe&C!cy{?Qmx|<YH&WnMXhC(cig=
zKX;$(p4a<<?psms3%Z{bH_StCozKJ%W%nHNC4N($!{0`#o&V(b?YsPspLGt2=O_4;
z`xyQXt9o|DpW0{C^APtG;;Q=(@Ab;_oDbv=)-7HrTc<oqJ|WLn#z&k%?E|~B-_F(U
z``w>MsvmuD?iT0W|3^Ob#&Zn$0Y3D}^9=Ro!_V)!JnyK=E5=$cp34vA*QkAF*Zi;h
z2Y9ffD$2gBkNsG`=NItOFYb%`anwDeGV1<Pp6I^Jb1~nq&@PX4|H(e(+x!l^<Zb*p
zzvDiV{n0=A$p0EIuS4az?124yJ}J)N>l`m$@-ybKKk(E}PVpK)a*_is?h)Cy=acS#
z;i{hf!hxUX|LKvq$6xbH`k|hS%jfasr|7Rd-@5RVAKIrF|KX1>zoIPuh$r}qi|&Wa
zCyv6E9XZeOQ}TZE!9AW|<l*R8?LLkdKlpo;K3IqM_EG!dIf&=qo|~Yan|O|*|2Gf%
z<%|E{{MGC4+kDo;8(#fS`ttk!Yvh1utWp2fAt&v-%F%1o{ob0M&~N)6sr})%%KWHt
zo)8z^Td;HS6cuO09s5S#oa0dU5GX&YY=7B5eRi%EN7)t1f4T4XoZj;{{P+pyVCQFk
zOPuvQh<wIJ8s*|I@uk1)o!??l=8gJRRpwvJYd-!*y>m1_MsD{2o_}~QQZ=9aLLN(x
z_#JxY9L3%vWheT}r}=C3_@ewD|A>;89nxcd%sB<s&cE=l`02+Fq3k|VcEJv`TetP`
zb0{463w}{vAdgbdAM3B(^9l9xcJ1n|lYOK5lUG@sG@f5~k1Ec4-mD)!{6D>B*X)Wt
z@DFfA;V7=_=blHN?EXi-%umF!c*ib`!;d|fho2Hh*fIIZ#h$IxeJcN_?EDhP$FhFz
zYf<;B&L?m(-u&d@hf#i;pLV~?p0zs{@k{)pIBOpL@iQ+<K4tyzx4-V?_!sN6F8jb=
ziVN!P7k_SF?Z5rA4tj{6eT0+qqI&XJw{iBB9H@1}ReWWC;xeipbx%=6`A5|G+4$%O
zOFP~8-7h`p;U5?5(>a`+`mulWlaqgTZ^wV3@@wZ1{hU+cd5_<6KC%vePJT!);7$MS
zw|ii}pOi1jujEtm0{SD5@*O_kMQ~21C*~JN#RGm&-Y@^9-}IUu(GT*=tHdK^af!cm
z4=2v?5A@u74B|h#G!HrLvp8iP^hKO?PNNU@ojmkT**In6)zdd|&v`<gK;QIJrdRk@
z>9>1J_b%d^I4%!B;aEjW`wjOz?7=!v_5m+)z@J{&2mXM3_8YZc`4wv4;40qA55#+M
z8@}Wt7kTide<;69U+ACy@I~nledj;5qv8#JO;5GQ{7(L$o&QDoVf^G1><kr0Q0GVE
z@zc)EQTH73YUd95wDIiRx+0G3o85^w<cQRM&=XX>{ZzKE^c=N+=0~-|*FBJY2w&$`
z^4eePz>lAVBV72=xIgTLUg3i;KMEK5lrp^K7tS^I7arD6pPy3S_+12hm8Y0b9%Nqa
z<TM{S^h4>Pex-gi&ex?}jk5hf;bK2xe|`a9a-sAv@{9OskJNnTw_p4asvhO9mFX{e
zOZmb*gLdm7KRIHVzlIa}^{?{t>Z|4<2fxODmGd|BrmSzo3plY)c1G{WsjMG)=^_0`
z`B{FFUXvdVD0_r6{KX&qP;%=BNAZ=s<VW?x7a!x)(+BI+Ph7-L88t7t%KC=hzz-h!
zm;4)fZ`4Em$*Dg%_%VJ@JNb>*U%fJ1$c?IpJNrT5L|*H`-*}YYp$GaIr=NOd^2B-h
zE$esg<-hrJevyB3PUZJe{P;g*_C>!v2NQ4IkNZ85`#tA(@mRgGdolU2`v7I}+kJpI
zF77GIuRWKG^LXyT-uP4d4?p?}cln+BCVF5W`4@54c}4!|97HeqC-c*1a>&Q6OI#xl
z|3#naoqH>JxxAko@-yx93x9e{FZnI<!iS%tulN}UKYD20@Uc#~X-7*w4L_jX`tjkX
z=q;R`|Lqq!$t`c^=jGel&F>z<{QSIqr$_t@e4T@x!`!#gBj;%6CG+A#ZhB^Zb|*gQ
zCy%mjc1j-WF-|*N-~eCeIqN{3pPW<dJ3Ur!ymK}^(vLpD9X|BJy5Y@V8Bd?BAAj<|
z-Ms|4oMWxi?<=E!q(}5v{>Y!&FZ<69oonc$eA)c&SKt%vRoVLRC5QET-j9!cv2N!~
z)V#5u{&2(3IL}v|Pn}oY`_M0P*-!Rm{nn-3`n*S^tlhrZSAL1U@@wQKr*k!Y*q3<8
zufhRUPw)6s>#(2fj34B`*cbhzFY3if_thwU((bucj9cQMe%@OZcikt6o8qnh{06zj
z8~SS<;tD&|j`BnF-#Ao1>!QE*O}v38ySG0m{OK3J$&TzRy^8v2|M@NP!usVso=32E
z_s!xVdsa3dy^Q{vzSCFwgz`J^zy8g|^jkLAa0E4u9hi@vS~vS(KkQXMG}<?NF%O)r
zgZ%6PwO)3vzx5bLZe^6;wV&j(-*Ftd)WZQ)4+s3{i8x0e#XI}YPUN2`eefJY{v!TF
zO3&q?#<N#^_zn4-{@U5G_1Pc(hksNLANbi9=L7a8Zm>gkjk0HPmVFv$f9<z+dL&-h
z2l{EB?Ng+1uzq<xoaE2)FZ&=)vtRq6-uOt3cRt{M-527=53qauoqzZ()VU@4TYi_G
zq2{$N>okwwOTS|Kmn~lZ@U`k^9r4-iCcSX&^6w3O=iIk!{ecy0*3nxpc=kcxU$$0V
zf9EgygrE1w{O;7c&F_6N{?GbRzelw$dJi{c&uhdd?{B!LG>`B9h+CcuIp^6g{OB|N
zr<d%4f1-ExU%PsCLf_@f{JVPkB!89f+b8E@f0v1WLD?I76o<qw_9rf|V`X;FAMoSk
z7pGCrdHFm1;H4fO?3BM$j&{tiiUZ^jw>>{}p7FfUd!4cD`wi;7XX^b)&o}uCagBWZ
zjdhXN{>np)llS7sukmN}NIiYv7sV0#30FAUFXt!w%|DZ?idq+(?FYW}M!sv^_*tKQ
zfrsa_@N`a=CwosjzL#r%yszs0RsG>=|Ge+3AAKbUDju@yXh-Zxy?*SNe~EFEUcv{C
zai8ESZjqDTqVh)TB)@s+k$&_)=AZF>AZ7i`=RAdv_g7K-XP?-K=ksx&^h5Qln$PdO
z*rE3Ttebu+!$sby9)IUA)Vk@Tb*V?y!v`+>B|D(^_L(1w^UF8wvvt@v{G(jtU?=v4
zoT&3J`QQK_<McBhIZ^Ai&&DZh?=^L9<xfi*o_Fzg;*z+gKfl8tqT(BW#$NU3Px)tg
zCu$z^^OvZ4<578pcGSGeo=+%iNAW||$NAOkM{aT3dSbhHq}@322)?mi{^5QNAN;Kc
zmG9_>ADqmuEN)sSiof=_f8@sB`pJQh@#fXvc=hB~Zbr-Tjpu*jyF8vh5l8rY=U4uR
z-(%1GtluB-$NVX~)t~?2m*lPVpFN0E#`7zvJV+ToeE3Q7i(AH7FaG>BN?v{tKjYA-
z7x<Hl-z6XWkNd^1nvWb&KJD<K@ASs{%|}j@pC%_h5g+qw*WbF($WOibP~+jK9Q&J3
zKXR$JE@k-QXPkNUS0=B0P;Wf^%xk=Tpcl$0yztSlirPo{hO+#IK3FHd>?g`?{wPnZ
zCy#m*e-uv2k(!@;v1}i-E2HY+%ucntr*i(5_d1u#JH#oJJ<<z&<wNE(UjAg?*^_zQ
zXY<qeI&U~vYNuE70rSAmy4@SNS8?y+9?p4IKfiZyKj3#F?#tclxcBg!*Sg6a_lZ2@
zL-nTz`0)SuXgAJzEK>QNe2^aN=X}E6@I%RCUQ|7~l(nPc7yioPo&7W)e*BYlDw9{e
z^|2@UtN!8|`Ot`m@y4-Nl)mbx3{Ub~CqCpMCw&z++{eo|<TK9wWxmnKAJ~~VW}JHb
zl+D+S+AsEqzw@zkuyZl{kq_cee&ZvR4?QK%Z@dqs9Upx86Mn)y0Do;?>=(bTT|G)K
z@a6wdc_BR1(_8+=c>axD$98%_|L~zV^jW(yJtB9rb<s!j%dg9QXy_5Wfj6AtP7aiQ
zSO@>1j9P!B>{h*dEdJWLj9tj%<?G^s`%?F1^w4_fKR&4SnMc_<sl==C-iLbQ@vRy!
zpN50=kqgewNAwtf_R5~h{c6~SeGzZThtg;3fS-1hURyuD<|mJS^xwMUJf)nC^SO13
z&(>S6tD%4T!N)!;lgqr;MQ)VbsB!iyu7~{WmL0NR_10~Cxz2|C_OFW4PkI~ur+Y^F
z>pUnQaZdF7Ts$B@eIu{$-1?5U?@r5G$bqkNxi8`O3-Cp)AHK?{edE`a>5=_I;mfaD
zhkE_27a#JV`l0G0B`0eAp0C8RdX)d;ulZ$Ey*PyO$FX1Jr+=jQ^83#D;+k`ZcJ=0q
z^P3mNpPyGYUKz!YA6M3&9a#r|Xgqn9^;4z?%2m|;hw(@MZinZ5_uG@p@0}}8JL{fv
zFIqln#COX*XZXQ+FWBqD=dCj0dl|kzgUT=TSN2^R{G21*XS>&xA6l>baDTV%`&<5G
zkGb!jJYqlXpZ9n8J^P})U(|WZKHFDz$KJi~!oTyM?2&yt2iR}<jy%~saE$qn`*P2R
z;0%A1J!towhu>{R`GwI^`7+w>xF}bYo1aC^BknrK8^@2?7yXp&pK=)&8u2dn*H8Qt
zzr;E6h&Sd}Z+xWqnNOMDMztGX%G=08w8JM-a@Z&BaKMkB<nQ=5_2lEH#53_qImR*V
zsCBUq<JgD!<9@=MzTn3$y%$7od{FPfIM>-nI9MNj@_r!v$c-O=MSlIrhsq<!s~?=m
zi;w=+1#f=Bevl)U^(P;k`6YbKL$6T!pxwIRL*G#AF^>Mj*FMSv>7ViR5+CjO>8C#S
zXLssR^~T{JsdnR(%^PunhkCfGhZ`y`*k}7?pOoz<J)j@%>);1h`LgvWJ7<~~bxy*U
zUeJ>$kFw`9&b8!o?qj#+!8e{a#6j`Yx+66&eHSmyi!VOpLiHmz`K(*Jdh$AV*iZdY
z{fuMJ>`6QRalQDE%R1q2J;podu><GJsMp>v^PZgdK)es<Jv4rezEtTu{g>bHWBRja
zzqe=ak?K!w_Zd-se#QM?+z0s~+}IJ^^><%jeb%AB{jm@B+4<7F0Sb3~<g1aApZ$}E
z{qqy{Q@kOs{_3rZ9b3Qg#yR(!M?Zdxep-)y`~W_29(a(0T^n!zytfx|BA4}8zk3gU
z3U2g^p4dP7O0UePOb&eDW8ElR%KO2_`Jde6vVU-a8@%XCiNn-;AG>kR#fAE~Zu?+8
z<R!oP$c+!YQT@n6f9Z4Cudyz6^py2B+Uc8dao#A8`Agd1LqGUK^5}=Zekl3jNFMva
z-sq40Ag6KUG#~t9x!m7y9|s@vM|`XN2K%sXlsx31H}ufD<(2YL{#u;(KCb6%%G$jL
zt*qVixHzt=tlsx7QT6zu_$!;=b3yCz996#Vd8)Gb!7juJ^~$JxM>$gVPjB#5ugo6R
z>re0K4gG3Htv~LEvVD!X$m8IFS|@)P*BQ&!74cR!4_xszAAS)BI9Y$RTXrr_!H1sf
zubtk*mmSOd<Uy#s&pin0e#3o;denW9douTR+TDvNqmeH@D7mZ)Uhw6otP?))kk`xe
z)a#EgTu|-i6F>M>W#j3Cc#9gZ9i<N_Ijz_J*$?xY&;H3<wc~@*Bk`R6nKx2;!Y-{7
zf9>|oeWdqVl(o|rRK5E*dO~kJmnT1JU&RII5dPo!L%VwW?mQ52({3L4T4y|GX_vRz
zC)EBqH`%{P-8ZUNCJ#B?huUZDsQK`LqkZKcl<6PJ-=OrCpH`+1u^&EB&tpG&C;vj#
z<L~!5?vwa$zsup@wfp|5^Ah{DZ}x@0z!AR6>|K9)gc?t;^rx5hNj-a!hmcD?WgP1L
z8}?vd*@^kR$3jkd6TbGremjTC6C%~_{E9C<CpSMrAM7tmUzGU`@}S!BLD`#mQ1fdi
zSET0G4_{Os%KzgV_eUP7-8}Z4pRhlX(kJ?*zvnmJqxJqQe<<EnQT@C>uU{;CPGLS}
z{XMVXkNNEqr^{aar*EIMq@V}Zfm$ER?^vfYKP1ja|I2^jOAd0O)=hr(p3A{g*}CyV
z;b8sx`Fr=Cx2dn9<iOWHdcJ1eaDY>!?3n(rd+Wo`b3^Myt(W}dC6}^w<Lf;!`IxeH
z`=y>-_R0Cne$qd<;$z>G$wPj>r}DfLg&Um7sh&LICH%Z6B<`wjM*qh}OTPNhwQJXR
zJLZb(7u~*Y{q`ewJEXSuI_3AueUGz>idXV}Wqwwi@_S(Y+#AYcozuLRDqnN&;vC`o
zM*KUf-8n|w7azr2_h-Iupx(U&KM?)2bGo?3Kl8WpaMXBy*?TJ9ca7^4FWgVTC;DOY
z<4X>46y?A8d*jSwyto8^cxtyE<H!v+d5e1o@{rqj@dv*073;;%{L1ca=><v-{gw5%
zF1Xuo>%iZBSik3U@-n~U_I{6Zx3c$>-A~BV{0<x+dQM*R(P!=YYbPJQR;KUrLw1G_
zT&*|S5q|8^eE~ko`0EEB^RqMhVt?sB`#`m$)@i@w({M76yq7%W<Y!Rrv8>)Y=nMJL
zIInT!u&*dPMLqAZukeDmeW1tohyP(O@L^x<*SU+G%eUh=_Ql_s7d0>YU=Q$z4|_uO
zGaj{n=0(kC9KAG;^~Jo?y}7)ef06G;9GrLHVmt~TIKoYSi}G9W_jh5u=jV3`;tHI6
z&sw|hAQ*r3#*^o~al@%~=M3M&@H+_e^4H|HUVhd3<eTy;WzVVL=APMl-4meBcW^<S
z3;0d=*(df)F7?K%k7es358NY#L#(&&RpX6AJtyR6*rRh_tJHS~t&g6wGq~W(zO?hJ
z{3Cw$U!G}SmF+KT-<-?DdsIJs<d3l)e^h^E`J?fuJUQm4`fHcRR+Z()@_GH`&&~nz
zXH-A)lUM#5&msEDpUGoh)Oo|azH@<U_x_@K{k1!Xm=|v3wQlVwxyeWW`9FH1o*g?+
z+CTe@!bd;Ue!<N-g1q`^C!c#6a*`jVhvXqIyXD8C{_<b3o_~+^)+N63$KtuTql}76
zvA;MM+x1hAkMoXvP+lSr^jw*r5byX?e7#SOM*ku|pa=Yn^@&&J6KBn5yt4k{w!D*{
zD*HE{w~9lN8ZXY!Pw|7lra$}^$`4vEN<Q&Ne>Ca={a_FD+I>2`rPs>rQQ0}(Jw7|7
z=g#}qL%-dBDwC5wvqO5%4p8?B?nT^(l9%5iC%MUmvOm=QhyKd^6F-YDyM`b8XJ^)_
zoxZVm>qgz%Dx>s6S-;2!ZgABPRj(g@aE$fJ?r%`*Q^q$^{PiQRa_mnZwHxnzz>lya
z{)fMiKN&|(e!zO@3%li4oL`&^`3voFe)bssgnH*Z=OldiMP>d(-lac39nVY7WzJdp
zIVZu(IP3EJQhBs-))gs!=85vj*YL3({oqKBSa#2a8b^L*{@K36$NsP@=PTzW`(PXz
z`{B=@Yqx%SW__r2lZU(!Fa7XShPVFO@iUKp){DZQd{yJ{iS6cv7aYt-Pa_`2S%)(D
z>9cn8lN)ueqZh{6e|(j-tG8eFl^+lfwOb#&+zXKxB`5yoMddBN=jUD{+9Nrw6Lk(|
zfAmFvc*Xhkhm&#E;rDs=%Y3N)M6DAgm+uq$JDvV6mG_{W&#gDgjX%7tM?F2VKl*96
zf9mxU|Ku&=KY#9iU)+`l$Y-?6SHyqRdB!;P@(6s@i}U!4W8#W@M?Y~-euG+<xF=7O
z_c)h}-_|R>iIe6fhdjx6WpbeS;ZF|d5AoD}g?tsI2l0ED^7(kLNPpccp!OT(&y?v+
zr0xgY6VX%k?A3j!`)K{VALl;WcR}g7zq^Y1`=Y+zsh@i+e|PO?*BrjvS>K&pcaP=o
z;2Lk-<o|ifajX4$)w;ji=I_1vdqwWWd{@(X$KT^6zxnVduXWl7aTqS%AGB`v&JQ@B
z$sdf<-#W~%pK<oVcTL^j_`a$B^ngC`cg~CaEWgKJ@XPc{9$|j|ksq>ta>2*C@RPTZ
zgMTECdB`Jv#C7OzUHb9U)*(*GNBKQ*P25r@r+YK}%`bkq&(mMM^_fSWL_RdGkG$-V
zyyVB%^FQZq{hZ@b{KO%A*gf1)cC=vn<1g9kqBZKu?8NteQFiUQ6F)%D`2+h;FZ5HU
zPwa?3^E>)0>qo!i{+QP|_CR0M(=+<*c?rJuSG)22f_0N$zD0h1n_aLoRD7pD^uhev
z*%$pXA9?7d{?>~R`OOpYG7j~ekv!(7N6PG9-b+8}FL}iq=LmS1&v^XRn}@yFcl%--
z{`jiLkG{*-$i+Ui`+HyFxjaHX;2f_$#&z}LJ}U1})=&N*A8<dVzdS-e6d!ql=NS6S
zZ<<l_$ZMSEt=D++;qSghJIe0Zwf&;+?2Vn#L;FTw?LVqrypWHh^w7T2Gy6?H_;>ng
z-_`Te>cs_o;yCs6h`!N3_QO7{7Y_V`b&EIbTR-c^hyKHdT`QYc8BXM7PsZWLPoUP%
zuHk}@^|Dj<qwo-i@qwT5_`+SgxanR=o-EGEm&HG2=NnZ1&u%=26d&=Y|LjVhPOnjR
ztH1LHKfrIW7v~p#!+FPjsQH{PJnwMsz}Gzb8HW$be)S`#`PnVGjB}pS-};sDg#-V9
z@>}lX;SC2k!@b<M#_y2quX}s?K%eLfeWstj(<h#z?8-SwJQlazFW^hANcl(9`ceB`
z%HQ~olrkJ5rI+;9y!0MEo<Hb^s+R}C%{YAE$WFZ%z>mU7J3r{Ym0ad!$Kp<vJ<wNr
zhSF<#W4wB}@gMwy^;A*kLgzvGCMqseQU0=u(mRwtric78%Fpxb;+=8)x_<n=cJ;<7
z>*t&&zTy|>$4@<K-^qvKV?J>~ybwpk4gBCH4)FVE>?c0bJLB{tw|UHO9{UJ?{pG{t
zSC3k!vi|Djz2dKW`HTF<IfP$v&PDA{%!9~foyu^xKJDyOS$}e&<{=k3{QVV_yw-v8
zo8%@BeS-&mgA;j_$!UL4>tH|fHSNmu01ocUoloOD=2K5!=m-7cNBA{<Onl*2^y3HR
z5vcs6^ox!775y$g{I7Y`!-2fyl&6}9f5!*44*Zm3yR!9>lb;nwtVf*IUdFBY_fK2p
z(DQ2gi+j!;;=b|v;e*QG%qNcMFV1KeC*>FBl~<S_AJjPg<bB%9b$w^6SO4EvU0djf
zpZWB2A8vm4<njXjOZgjo%?BT4=P38y;-kDy9w`44C(R>|lb@mTIQQ)0u>4Iw&(-~|
zz<W&O@jRX!s65g<<cj%_{+<Vr2Y&8Z-Nzn!+MN&lW!}{KyWg?iAC5bFwesG{{geA!
z-#b)xkBbj{QTpzEZ+gbxh;RIecq=aOZ~PCx!f%O(sC<(B(qnqTPMxF3VO+a3`VaSf
zaImlRfWFgTW%@<$jb}e7d)ALW(;M$`&;!qhl<7~@d-^FqWOwGFN3owc2*1+bHQpB@
zH$K*B{np8!x}O(E#Up;4U#3^|%s%7K|JomM1?4ZzPd;+N$-12%;p=&sarB#>$)n|A
z^q4#mCvv0og59!rdKCA;zNts~HB`PI{fYIW{2a=kDeH&ADVE{py`n10P9pVQN35s6
z*5iF7`=USnCx4VrnH{0>GI^c*cKYeL760kEmGjWz`3rX6@yaQ6_x1X#kI%83r{#U-
zb)PTK<Tst8<(Jxx_Z&|dfBo@|d8l)nybfRaC%eQ~ey%Kk#h)DVF8rMzls$K|ZujWg
z-5;38{=mUF?Z%PceDZ1hv>T^v9s0vVo-BWqU(*ZtkdGYx{<L$l=Zo&W$tTY77yKXm
z#7kwAUKp>QyyB*I@+)gc#Ygc{T;<oq-B>oC`0c#s+zUtRty+)%<Z%wgAJxvUMtPO_
zVR6d&M%-~O5jW(^#)})qlN(O<N&Z4F<Tbu)Lm%9a(J%VPp5UfFj(48)9E`oOZ+6Ge
zc#iLRqCAj4;<xCfapF6DqSw}MpUKCL*%A4z3oh)7{MKun=Xva!pR!I=yLsS(4}6`k
z)x)1(vkv<e`?DAIaFhQki@V~nJW3o_ub=y3aaBGdua}RAOU^&q@kRNQRw?~v2lSl&
z%a`PHW!@G3PMo}3{vtoJZhZI?_Tv0a4phG992fH==QQyjuJT=RUw%Rk=P={+lP972
z%QNhQcKeIJyvn}HFXV6P<&miKp!c$!3sK**R`&aG@1@IkQ1yO~0vCA`z4E+4o~GTp
zjk6B<75=Ds{0`smVf@|DEf&A`t$UrhqEK&t;N`rEUyKv{1v{jd^w2ttmuIpo`8oXQ
zG5us$?8rIG^GxRtddmNilON$P#TVxRa<Xsw1ZQ&5bNWRt_apG4Pu8W~dX(8A|Hl8u
zxG9eF!`1<JewCl)SLMb0yYc3Qr}d!piQTau`lX$HDx>z5{o~6ovePOm{}8t$9{e9l
z-%#}^zp0&na~~kjU<bZukNQ6S!h^s3+Mn+_X(WC}U%mdQdf&NMf7iD+|M*AOnKZ)Q
z%zM&qCtrU4wkrzXnb7XL``Z1k4V5QW<z4ih-$tFw*@1dE(L4E_ybEP_>`)$p(ns@~
zhrOAfov;u3$q&*~c13TL=^KB@Ppaq7=qEqKuY2ChztekmkB{-r`6xf39i{gudx-qZ
zhaW$JuX^j#j^fYH@=N%dPd~UtdGy1Vzr%<9+830ZDF4l$lFK^qC#U^1Pm~j1^1?ws
zc#D(P!H?TV{;L@!pK<u(OI~_O4)t)N&-}gq`dL4mjk7-TE0cpi*6#Oi^qJiDho81C
z_2PoOn!j~!BA2|*dW^Rpa1qz-11j%PM#%{ecE~>MlXmma2kXa&T>J<5<$?T<_3%Ud
z8Gj^Cb`NMg8uMQJ#*fKQ$nCicJs>xIATK}e{$r;X9<%6%dseGgofqUa_E8=o{}D&s
zKZyVK8|AN!kM|q&B2xJZzVZ@sDWmvDxvX29kJSBv_!cSo@u{NnB5?^$?iEn+O8$UG
zImA8V#7}WiUP15ZQ+Yq$xQC~A^fBrmzV;E0@;~k5M9Hh{o<)E09iI9tyGODg@+Rvd
z5B%*bs$Twpk38Kx<|QXPVrOuHlXjH-!p(a>%EpOD%1`OnI4^6D^D5hK`$-Pt*t`DP
z&8Ht4<t)qOX;66ZFZkmVsrBlQFTE}0YP_H2-1^|GH*B{2h{p=wN$~wR=i6BCe2MDk
z_f5Xb@{T!sFWqgEDfQ@=`E7BEe-!t`EmWQ$4v4?{qw<XyH=XOn8PDg$Pw`5<@9v>+
zvQBxU{`fjK<0qb>;<&hG{njO}h`aWe|DcEL5#@i_t9+i_tJj}iup@kpEA{11XI#D2
zkvG@)eR6mY1V2>0b=YU?5FhQg{S{~RhlhQZPa03p<(=ZGd5xE^7>}QN{IrutKkeja
zfAVSl@QZq=9NVp5JH17#>dmL$Xla};@>kXm4)|-YqE-FKYn*=OQI1r9{glzF{_3Or
z>Wz;ywi~ZsRlE8~<M@b!{7F2Omx%N7CscgZPn?zaM1IQdbB#kiA9H{19$sE)UeEL7
zr=BaI{JiHAei!I|3P1OR@{dU63+m;m)+Y~DmN)QU{1!arN%(qxX`QJ2M4p5%f5;E=
zZ|;xyRb}I?(|mA*JO1(?>ym%-v-VfndOWZ4oW}mh+xU5Sn1_7uBM%%R^*qOO8Tte-
ze8^>;v1~m3@qCDWc&><_{18s|&%Uw)^1xYsPv1QErQhVnKjv}vQ@$m>$Xh%o5pU#s
z;;6jJ`OY|g-}zIVGftc*4}H^)uXg3wPCo78w0NaIy+-jZ<6Gl=q8>kS6%}vEZ@-$+
zc&>$)I4l11yY3C#GsvqVr5E&p9;l~Z?33MTSEe5*z2TRX@ncu)EK+u@zj@fV`T0Bc
z!_L_WeMa5$D|;TR%rCGDam9S>Up;$Lk2<%J+w&E8z)5=*6^E=Zmere=oT&Lw{=oe3
zM6Jhu$xGP<{f_%WZ{Thp;Ocq0=lR<8=im5K?aKB;8Re&x@e`*#=STTf^PyGq#d($W
zkCZ-%^KfV1;!CV|k7!?%%_pwVBle-c{W8w}*jM)Ly&d_7_blj<^ALY3PSa2P@Mllv
zSC5jD-iR0a!-YMe<WWY&OYsjL?$PY8eSjZ+C_UHTcNV<o=KT@z*EsylPw&Y`AHDZs
zJUcL6yZi%A<f50>Vcqso9>bpKH%bosz<=`#<Q0dU`%v*kS)T4b!FvbJ<>E{{*Zc0A
z^S}E5_adI#q3V54UfFZO^B;V8_I)Q$89_a7bRRCikVm-BiIg1X@m!G{o)>z)c+?iF
zJ?q)WP8#uCQ9Vjd_a3p{bIHixJnn-cAGpC4p6+GDdCxUa?Z(+3_bI6NQspafqZjeK
zP7j<fmE~)mCz8wgO&+IgKiv1qi=6-5^WZ~H_fG2LeU<#i`ceAfp2_+nKYUT|-733x
zvJUqj@PIS<*#}(I<D-m<<NR#&lj5<wRQ@UtwU6>~d91u#+4+zEiGD$zKpyj==EJX)
zqw#wt@@YrOLC@?ve<P1omN(1)^`}49sf?OOfBJ{7{i9d(9py)?3$FModrlSSH4nS6
zUi_^~KhHPS>xY_0JG+DnKF(S27k8uF<YB-3Iz99r0DH18=2teaIDwKI{_s(6J^DMh
zC|6PYhC2V?M?Ux{TbFejr(HR&3!d~Bo^c)4NiX2fzm?+}zk`eCxh%JbfODjL8kHBz
zhxt+YxBDLVQ>gutr|O4BKPfLm=_R}JzN5Gk<GJTPsCHER((ZYQ{Kb2U>cu%!e8eYG
z<Kzjb=QP?;`A2LwPX57fsOM+Zi<9ILC*_smF8Q^KqvAKd_{BKQ?(GLY@@#pq{2ulE
z+rHXAczezl$BUQF>Bh^$>7)2gpPVP$gV=BQq4Y_ao{%e!vv0=5{FELUhssCgu_*lD
zipF~TKp*Tgy^J*OqjvgYzs*BW#3}XkgIw}Sb}6pN*Z6()sQ%i;C-Dwn^N~Zm*G_-s
ziQ<O1A}))M%G%Y7Z{m!&kDq?V>n}eNm*}1MHpDgU^18@RyF3czC!10C3i_F!Uu{O+
zL&SdWxlwk8vZv_xBcEnvxT4Ma^8@Zp*gZMO=Una{MEr@=e#l2rc|2UjF?gwu{Gz|C
zvLAXcZ)!&QS$1SRdy?m(?M3~GGs-!Pe%N2<6MDt2<vsEs`_FH%TYBR?8~TO6`uMy9
z^5$g+@?Cn#U#OS=Mru5}@cucuwWIPq_HJDCC(71iJZc>3ew&|?FOtW7K7844ytlIt
zDEk-Z-B<DtsCyyzKd5+&x~KBqk^3v}&!hU`BYyiGi?|%;)z5q}Uyv7QN97;#31xYc
zdr#|iPbrU-@4-p@r)T27d`CG_ehoi)5`N;axGj&6FYsI9zB2pa$53(BIg-A!Cw3ri
zuqXN^&d0c`EIy<14`ueG?A*(~`3d~!KRL;xj9;`rW%hy(zTyj<#2<cMo&pE;aC7cP
z$!(l6yy3>*kYBtPS3Spwd60c!=k`NA>Kx#_z^+mHsvTwL$|(I;)(=1P+6TDWC;ME|
z@Lnu@wc|%0;mHo<k@TJZ&~NdZ-q|<%09XCx)$TV@d4c!O^^-UAN9x(Fa{)i%IRL*S
zZ{Zi*v-3mFA^d`Kh_ZR{SI>X&FZ_e~@O6I-C+n0aS&uS4@T0f%+r3<Q&)T?8Wmnqm
zhy60I^9DJshrG@uaJNqKSqDDG@t5}B{Ny4Be`(#3k8+d?&d#g+GCeR3esO*FQ9tV-
z4}6TbUjC-^V~zKR*@JP`gP*eb@WUTvUvOf7=4FTYDzh8wgadp~=VSQe&)(T5e;^-<
zeq7$O(M9Jkz4PtU>gQ~D)uYolduIL2RW|tj@4vWy-QQI{d#8(LeEv1-)csD}`)i(W
zitC<>dOw4IlRxph%J8C>*2k{dyS##Z%SYr9?3+C++b?+1FJ;tzvlIJjJUQtv{EerF
z=9PD#_L2Oa|0!E%TsJ#pujVm7{wVzTPvgl&4*l^{PcHc*zw39I>gAL2M)zXYLk{zz
z=0){$Pel&x)@h%$qxKu6C*+jpqVi+%D_gg9n4h0lM(w+Gq3|THb>U+@^r*^@;IDt=
zuim=xqu<K>4L?Jl`3-)D|KR8OLGfMx=+BJf$H@sNc50v1!^M33u)JSco=8r3S-*A4
zFWkfF$FJy5KT-10V>H$~4>@<SNA_iXqouxI;=3i@KZ)<5xkq+C?S5GvDX-MeeX{=i
zu=iy2Gv52M&iCz7af_d1pYZVf0ItobeL&eIKI+lP&pE}pK->`r#YcIwI4Ey+zHt6=
z58z(F^K<tSo(G`L56-{d=keZ-_k(=bM*Q?%nSC*zdEC#)2iY~fC%=C35a$*0dH;z%
zh|}(!+;f=My$Zdecg|bh!^1~^=PTzKW%kOy$TP)Z{)Zit+r0{^-a5#|&+v!rRa`f}
zdj6I_q0jsWKKkL${?*Gr;lS=tc@Mo{ug-(^3H9C=oUIR)Z_yj`dLPm}sQl0V+HdXd
zwd|+&(NO*b?x=O4?2Y}AU%h^Do#rKve*6{r*^BY`D$`SxoN%jJ2R+gbr%27`yym{a
ze(>Xw!qGb6&0q01_*sv89d_kj80DXwE78&)G=9IRY(DG2hd#iGA5w2UaJ4V=6n^ww
zS%2$M4`+NL&dT<WoNy*T9NiDQ2Zy)!F5v`UehI$L<Id^ym*0XbzstWmH}F^d0so?&
zUs0C#%TM?deu!T*pK<2XZXWU1Ii26)_xM-*JQq-g8$H2?KaS@~_GzEQSMd{G?zQkI
zC%LQ}l}9SetDS4eAwPt-GQRMNc?~)ECHl@E^QZKmURh_P;t@OWy-N9=_lo&t@9ENe
z`Wx*+J}eI)54#~he%i@vy!D9>-XrE`<9e;ncayif=}+fxanvg1-_28&$D-Ehy?=gy
z{jx`tU*M0}A^k(yp><im@%%!$uM5vUXUl2V7wlM`N1v6QGs<yiO<DIB_ugNN_ssN{
zAE?rA`i?JZ9DftzhJ23xI#;3A!9SUw9pFz+_wdU69r;jl=%=3C@FxeVK2o^4heDkj
zt<OG)ciOEBAM>H=BZZHC^n$)Szw)>I3;%5#xlrflNa-*A7LUX?`s7@}Kd^Ul(QEeU
ze23!a?>NSJ@pB*NJFWW3Gsuam*H1p>_bb|+W1L@%H?MVw1N;rYq+Xmh4?lq~e}wwo
zi~jr+`TWkuc<p|dVqN^ObC-VBr$4^@H{9h5<Tsvt<TcK@jz6Ur@Z#U$NKWz@FV7-}
zxMUptc5aZLq2y)P^xwHqJmHVnz4M;uPVxiok&5^D@=xw{l-alO#>oT3Z+7S$C4Uje
z`AO?>ew5e9uf%`+^~2XanDz5J`bEE{%-`}q{F}0MqWp&W<XPrX<`?Bz{13d$10VhQ
zJ^p3Ts5~O+tCoE3@!!_EjRU(C?<1w}^hz1!FX*{@GWkV0uJQbt-lH*ZD($AxU%Pp#
zsJI#P_AEac0dYtiLB**^V_cF)%g@CddMZDORK6|Fi1X@EdA)lJ_cy40F7>MM9M(BN
z9;zKb`7M2q6iV*z!~_1H-xW{zYyLFqGdZHZvJ;fP>nFbzU!t8^m-vZ~cqdNq|7ARG
z{GA|W^0O~g91$<XBXQFCTKv@y)gNx+nE1{=G^6w(?kim3pxymCy>jlOpW+<-qo2w#
zFO|pfhxWyM=8yYiUr>664?U63ME%rHz47cpzNL)H&)5h1gopEm{>t<nrBC$7IfH%E
z8+zuROx`L#=0}_d<e~hB`wROc50#&y@>qE=%3kO*`*aT`52UB;*>geqZalqpzGIj4
znjO+V{pmGa%p;F<?`A&Fd7UdgM~=^*y`SkhwC@|Eo|lo2{iF1Z-qJUE;v7$pt&jd#
zC!Ce(1--=AJo>W}RK0QhElOYM89C`kr0%WQt^MbB;BG(oFZ}7hbAabND1P#5<#>Ne
zzwCFM4}bPTK4p}Ao-?ELWVF<Ca`zSVlwTLGl<AN4IX~MElplaM9QjXp^4s*Gii+n|
zl>akcKl8Oqqu=Kr@w1Ld&5vr2<H$??*iU~{T#ESO6Y*DeKY$O44{D$ABfoj!p&vf%
zn*6bBpVhN(euqC{@2L7n@xzB-;)nSil)u%^PU*8aM$g$Dd(@A-_?Xu`+SS|FC_j1h
zGoBpQPac%q_5~mH?AQL;AN%Kd0lSRz*<bC}Z(aNyKgQ3K!~Epo=iDpAb@G4s<0~#1
zk1u)5Cw`;e6VlIkev>^|cf^%E#*ruTH;?D^k&3sd_eZ=>qQCc7^y9ac`E}#f!=0Xb
zzUTcu`9aJJ@P{v4?3aAq`r&B(#z%Rq2QDc7D0$(|uC$vkQtkFVmeteyXm9+4c6zSN
z4(TC&a3N2`MZNykgAYFusXV5N;)`0ZxXYf*!_KXTJlgpa{+B%RWB0wv;(>U_-@r{7
zuJ|dluW0Z3X-AE-PWs{eY#k{5_TC(QjP}JY=r26wJK7^PKO8;(HjaLy_&UE@hx37W
zB@UzTbS{!lxo1J$+k5Wmp2hd2yl>?k<-KY9==p6tcNy;-<~aqb-g{)?hxwx3IKO#r
zq3k&dY8}pb__=RV_B>mE&vl&loZIF1?h)l7&cAS`&&G+1+T9DtZ}juL$$H3*KR)7S
zoS(d&Uy;{2mVU4+avEnI?d}nbbANz(e!<?H-`KzVE$d^4&ZqR*Im7wLdw}8(Im|0w
z@|(^N;tl_avOD&IKRMWs_|KoPOa8(6UHoG2%HHcRAG`J52fKD&W5?#<cb)$|cXBTU
z5B<bB_*q9Y$`6WP{JAo|_J`l6fArSAlLx-2@%X8aW#iq0*cbb&UO)RRAJC4!e)35D
z_;2?c@*e&ib-$sX-w#CUJ!<D$@Au-XoxSpl?wQyJ+)LWHKW1O>f+t+;bHvLy)Hr$~
z|736Y>Cc|nfjA}akQdNfR6R<6wex@0@BW0Jf)hVT?@)e#JmjVi+SOY(N<YZQ4%j!G
z>63Z*9r+sypGesSdo`bXC;0_@%6^USE0GK3SB)>{U;dkGH<)#1t(?D6#t$CRe?-3c
zk<)%DYv;fD_n7aA8|vNLsW;9#_#NvbuleaUJ+}_=$bB3>_{$gQnK&m7$v?$cWqFbO
zL0P^dACl*ohd&fQ$>&_|Js9W0=ts$KUr^(%Q{G^m{D<|iA9<fVfStLwwhsBAd;`@_
zer5gEiyu7gFaKqK)x(7y!^gVfI`FqH>yT%naI|iDj`gAT%RRsSzz1IH?UVY^QaGXb
z#d+u(oXG`W@8KvLPcCspSsqJoQTeRAT0i@29`6ys32x+rGs>>uf)Bptr$6k>I_xk0
z@UUL}?I%4!&4=O(4|y@a&2R8u{3E~bIS4<>--rj|2`W!imdA;w;*EI9-}7hqiZkMB
z>?aTO{LQ&zy+5DzmA5TitM0sJU*Kk6*r$A1J}f^H53Pe7@Zp#EA@<0Q#dH3M-%|Fx
z((^t1#0UODf9upA<zMv|kLV5m68#N5HJ;r3sBz@tm#iNZm*V`TJ~i?+`eU3t3WdA4
zO7G~Kd;^a3MLeW;_}D*sLa*o_zUHyd^p0H^XFlWUBR=FaAANEz!k^yDH}Qj?{bo1T
zrN1&h=EdK-=?A{-L;hvm@Q)O(=0WK#J!2Q<hf}n7d9~-(&VR-^cUm7l?2G;?TUWFf
z`UOY**pK*&k9zwl-qLsahT31a+kg6vKm8}iSgCWe{q<ZL{?1u`hwQxV{0(>a)6RX)
zfzE&W8*d)^Okep;^5EnAPG9vio;{F9yK{<llF$7-{`k@x_}llW$K+$z=3&?P(pUVW
z{h5zl&{OlWhghaJ{E7CGHvA*IbS}`Yo}YCd(C(Z>AN6D3^pans=i2dyn{`@0yyOG)
z8Xxo0!_sc2-uu{%b1p8ltB;iZl9L}GuX^*tk$oz|nS5|S#WD72oOR=GU#(9(A_v@+
zW4wf`e&mXj9$6<n<tOQ<cKDLd{?K1`2Vc0e59=|ndh63}oH9K5J^h{k+;clG$#3P$
z@(Sk=`6wz6_xlEUs(eabYv1_u$FJZ1=w+u&sf!Eh*$2F%UD9{^$9{~Dei|R`#+!%!
zng_N2`WdI)I28YqHvEG5jYILbKkD&A#W(RzUdkR&{ZaXX`Ahs8a+$ZJ;d@o3d=sqR
z`H($0kFp2nIC}0}Lf_?ADE)ya{iL_3{bHZ`^K<kHKYAfgG+uvxjQ>>Tm&6HWek1x9
zenkGrzwu-ElbfEAhkZN8(O+@WdDeQ|kGeN?e}LK_&)58&9Op&9C-U4)*>f@e&F^<S
zAH>)5GX3yH@#Xi(;kn)pU)$`57oN6i{hc?T``d#LT(y4fom=m7&&<{8hn{xl1Amz}
zweI;FKKRlj=UO<~ck#_WD#M-sV9)f5zM%9EjsAi^F|Yk0k9Z(%y2sE@{_R}q9>je1
z-G0+I>q4WyWe?hor(foWgLOD3>&ITK11|IprHAyEJ;Q}O{3|<PC)OqIlgGZezcCNo
z;0ZtLiF&~g@l*E4e$oT-*)RB!UmRuUu^iVSpW#2uV?XS#e1ZSuFWqanCxf^9M0m&#
z?HB!lA6(^|{Fw8Nyw|zG`9R)8zu;y*dP07A4te2DFX*>+alkt8;}_`(T=`x2%l~{g
z%KI6<gXKG4)6Tl*+>4e^8u47ud&<7E=s91+1Apf>=Q{fdPkPM1kdxmbzkLuF@O2K0
z`K@!I`vdEg|H_L|eDQnrBZpjd`d8L2{i?s0_VwG|zv-EKPOZ<}|Mma<AD2(9FPQ%L
zOZK{Gjk<lcPI$!ily=?tt|Pla>7Db2@$f;_!w0UY{ilEIP`orBD(|C5{1AUY&-e{`
z>AqS&eE38Dh2NyNu^vBmDh{(_)IGa%g?{d9$?18Ad8~_lqx^+9jUP(hNX=uN+S#r3
zqVN<4_%VEp=fC+^@@w~=i8wAV;1|Rj`)Zx^6Ta|*3;Dzg^BYHA`-?xy@0gFCi!1bx
zUZeb^JWrYb=O5TV`=H<OG(Y_&2fuC|)@vQg)@z)y^`X|okD|`K?2&!)W7cII#>u1I
zql*)0#L@c1zbL=>&hDc;(cg(@{4u%tL;j3k<Y)CGzxYgl$qhGlU?1oyY8~{@{upn+
z@ug4l2>N8b@HGx@<fKRBw;p`(hok$2W|Z9Yh`%NmdFiG8_M0BapU9`5_1kBDfuC@`
zK>0Uie%3hi!d<(3gg;Y{+Bg0LKXTen_NgDgDqrK*J@0jXa$n^9qMiP*L)5;)hu%ax
zR)!aUD!=tT3G>3ie%d#9DVrBA@=W~2P4+<#=mkANwWHQ)y!F8wF6@jSH_m?OM=o*F
zK9zp#oKJrAwHH57Be(waUpycOJA(t9l~MAS`{lV`<GmYtO>TVI0XenfV;|I8C;1{C
zaOEf9&R@epnLQfEpWsiQ$f;d_ah^WuFK&xp<l%ShFTZ3R+Qk#`6VCjX{`?BNqj&6?
zT;?age%2w5$unbo79Zrl;tDy%5pl(N3Dw^{H>y3xD}0>8oD1UfDdzystGsvWeNy{v
z{nkgn;7AY1VVx-4?Wgh0sC8OTGb%rFo~e3H<~>??ct0MV#_0z??WprDJ+trRP|wex
z?2x}u=BMbl{-}C5vMcussQ4s~*cWoy7kXKRYix%T+~h-%KRr<|zfw;Q`2~B5R69P#
z<3~TO2c`GqQ??%U^ou<9%Q>Ci^E3Q6zrnAPo4=KZ+9&&Kznu&1t9Ik?H7|d!pLk$C
za*6}|6hAG#@UQeNmgPODxUHXfr`|eHe2u5C{3QPz?VcX<C;CPF@WUS-;x9h~KX}Lf
z5qEiv{_rA)eV}h}l`oUqx|GF1_>jZ=@(k^8HoyJVPrQV){8E4Ok_!&jp<P^8kJ2OK
zBmU$xzkJU=unYGe@jW_m)jHvUAHBjyKkL+P92~8mecDg;Xw~o4#6fXS8TDR>`!jLD
z`xSAV_`$D=AIkiXIKnR(CmzcW_&@z)zJR~@Z9F;ES5f{PU;dSxD1S*F{#My|_2fmx
zk0=M6qrA!}oKgFMs<&UtsP|*wOkV2~2gDiiz`9ZEg_HH-V?V81KY1Lg9bfhMNBQuN
zRDW`qU)-a2%Jjv)+YjyPo9!1pvw!ZB?T6p{!e2ks`%LB)@8Ll1xL#$rMQS|U@v$!J
zB`5tNf7Bmk{qP|-YCpvZ_KT9!K3PW<#n1j)2l-KQ*$?u?c4hldZu`tG%!8`ezsiqJ
zlzP8Ajns4L_*~q*Nfq^+-23$SxW_h+`yS7+wWIE7{<`~(A6fjmNp<f9=^v@TtD#;Q
zRWE*uuc-V${vc14UpN<Pmj}u-Q2kN$Wj@#Vo|k%j<TI#zfPC^wd62wVKl;u7=r?`m
zpV$pMV=w%LJYdi%{E{AU^M1ShqKG)iZ{?Y>J(k5;@y0pO^Dp{F|K+jH8T6C>mg6s&
z{@k;c-B8F!#0~ifIoONyi1GAYUhe+Od4&EuU$AfIE9X$>NqN5c<R#X{9<_@<;;VSi
zj@*;r!wymF6sOpyb<59;XOGV9aOHoT%i!dmg#UtzeRm!~;R!eWl*uRGFu%C%K1#p1
zPW|jNocJGn`6qnj2gcFE$S>-V{`kTZ?r=nnw{P*@i~i6X{#RK$>i$C6d5yk_BlJtV
zeM0H4{cyf?Z>!!uDtqq2PCfT^-ZUS(vH$G3*LjZTRDNgYeMskQ_k5nOSeJW1{jJmU
z3;jJ0W9Rk_zV=CddEe4_PQfnio45@><Fxa4_L1K~=@0v4XZ)G_0sC!V$&0V~_(l3n
zpXs-D{+xdEA94Nkql%K3p2God^a*bKrgiX7`oYU{5#K|XXC5@|Q}^CfU#0Fjp*&MR
z&kf~?@=5jH^Po@OC!sf<1A0y(KlT2mdmjDhBYi{J7aHrWTf2DazQ_Knr?1A*3;NB@
z@x#YH7)Otci*_OJwl3@Sz6t7mG`PUa`)vFW`BC$`--C<$R`Qsi{E^1*Lx>aN4a(n(
zC*rldH^v|F+xN4y%XgiR<9T0M9uTQ~8h>$ITsKeb=bY;tEuI<2&gBori>J!?TR&<X
zyp*+fn!*Y7J2Pce9ziZ;c-U|E4ffNztykWl3=jL@zQ%sx0|)nL`a1{dk6*-ze<i<s
z#eE5X=l)q<shyv;K6)%K5qF|r^?cTNc_9Cfk9-3kc_TT*1MTuG_|QjGoWY;m;*0V2
z4X*HGpKw=?+9!7D??3uJjCy{--+!eq=3~G1#lG2J`e%QY@m0?skWW8ypyVTu@7vIO
z{bJegDamX7)?@wn8V^VP$Y;E9<VCFq#g82Nkry7yRn&Ox*6(+y?9zNF`RyC|^|x>8
z$%nF2c-UY2hgz>Pe)?-i;R#3U(%*QL{Mu3LfJda{VGrzv-m*KC{jyhiAu8XLugSl}
zYk7lu`IqyBa_lFb8z-NYf5-#mGs^no7b!XAqw-JutxGvld1zHX>#=_40yNg^ALX=8
z)VSDBIj&28c@RFR_3MY~udF|OtwTTU_JKUc+eiD2S~tGVXZXw0l;zXvmEF(khnfea
z&*Fu$yodg?3-&GUh<m7d6d(44KfcQNDx>DrAH}z-T|H`Clq;6;C9iR4<QL`SXZTBH
z{80SRUZ4AOzTP0fpRf<lA)O=Sh063%nLf(zw5ylDI>*aTjo0oTK)bkQe)XvM<(@<y
zY+mc&@A)~@cL?Pz;<N8r$Q#6Sb|J2qpWozH`Az3r?WlTw(D~YX(D*q2Dmw@3&!0N?
zJ3pZA6Rlr;#MODjxxjwdPy1w?bAa=Mc7I>cb2Gn7MfrR8Cw_nGzQB4=_c5q?@yGq1
z`v-AI+=$dTRJ_3#KjXwf{p<sIwA)u@aglz|w;1Q;XY5aYBCk;&_sw~Xz1k;sp^WNh
zoPA_h+KorGlan5y<g;G+n{}$Ue*IDRv?yGatKz--=x4~Op8TFiShw@1b1VJje^BRF
z_ulw8PaE%kL7aqVyl-R|S8qId&KozJT3<YW!R|X=Ii>FJ<Er;}b$5E<F^g`vXSKTW
z2N!<r(QT)#SwDNHi)MWOHS5$rJoC{Nv){UQUHx~?eaqG#Sg~e($xAQU?~C7hM)^BL
z<JqD4{C-ir`%r$#@9EeF%FeXM{?=_?W%dq-_`7ud$fN9eyx+yc4Ig~n>#EnE-0YJ*
zt7rG-gNu2TcRBmkAOGp4|HRerJn1>Vjz2q*f6+7XMBXLNIOjSK(pP#YuF*638@pz=
z=F^`&;KP3OGY`4>Eq2Nt_$BKS2bIZTT`><eFUn5fV!V2B#=5NAy7Y^b9}<V{vvKf1
z$qAn*zx_b9D?68?>g5ymSG#)m2lT->_X+6v4?aBmzLTen_}#VdCA;TwAEdu~WO&J2
zlp}=;sy}MIqovN3sOJ>ofV{-}2JDAERlV=4UVO06RdJL4dQRYZ2RX>69v^W;{Gk8x
zBk@?f_#m#(?|2_izde_u*Z7;CJ;BL1{lp*hdOusdcWSVo_S^okH~FzVjves_C2jmJ
z-Sb7yY2Zjcc_e$XKjMXP>`{L>dS1mJ8z)~=_MA#to-7Yz*X*qsC9m_nxNiUKCkj9N
z4M+Rq`J+5t91>sUyPi9Gj;Wt`>U%KydmibWfQoC@Pv7a2_idb$JdakdpM9{O^prl*
zH|IG0*#-T92mPU+^cGI2aVWjeP9F2IPn3N8hd3r4!$<zEpLKgqjSs%o%igWqIQ~dn
z<Zr}J_JNXvKV(nZ`9YLkiVMo@+50Ofdl%RBqu1JF+4C^xXZ*xNls)hl^pu{*vU6D#
z<*&?Z9`ds{R6IhRQ;oMSeh!73cDPtKIZ%3t(r^1-(#H46*a<z2l>X3PwAAm$?_BYL
zFaJ*u#S8X@>Q8U+mp`#5`s*A+PIhXXc9ecwhxLn_{5=1zo&1r%dHElFQTp#3X?}i<
zov~Z<@oU<RW7qh@i9giOI^3ry%TrK(*8149d;;Zv#5Hk@-{#+)d-*4Rf?t4x`S~yF
z)DAazSf8@}r3Y|PwjTKC4>z>z*T@UVL0*0h4%Tg5aPb}qf9>7{z7c==25<A>ubv(l
z4<Gp9uRlLdUinAd5A(=3Q2p#He&%-$F)z8{zz^_G56*hSX3LLwtneKr)OVOr-&=CN
z@O(7hM|e+yKjdHai~do*q}}sMR6OE``BU-1dk*G9`EmW7e>|5(Js;K{sd=r-y5)WR
zU8$#y_w>HBeIqlTfB5`&9J2ZD8}0t=-6zkRKWF~#Q=a(uko6AP;E-qTKIzl*4|&!T
z<(KR}dH2bC{_UNgnSaRXpMB+$+A8wuq)iZb&zjHQZq9v6i=^%0&;;IWo?r+pm_Gmb
zD<&_lb&5_f_*!1>8;?I`=G<4^RU~~f4lM)O=74Qh-Fe;(3$H8MrEN}I4x7HomD~N|
z3pdsJMG5lvjXwACn|8adHqmImaF6rahop(XAR{n$jZYl?v*kAxt<phWQ_41a1ddz#
z(jzXJbz3o38ue%N5@gRrAQ4Ce5`iZWnE!7FUj67VC)N8!-~6ksUUBaM&!`V8{lar<
zGjDo)z51}ypKti#^VYrOFSSV1Iue0IAQ2dJ1oF8}YtL;4eLsg)zHZ&W4=WxMP42PY
zdrvgHa=nQ_BG4RxR-Y@zd(lBHuRdw{rk7o{xEM6rd`GfJKM?3vewBQm{N6+}Jo>Ts
zIcFk}2qXfDKq8O`Bm#**B9I6q0*OE(kO(9KgN#7C&!-1@O=FR*T|9DJYtQ-nh1Xbg
zqZebljZ3rp(@wXyyPjTP{lpxD`ur*7YVUVAt>QM;-+36{X}d{&tv&Ziyb`a`{LZ{r
zaV6(T1QLNnAQ4Ce#xnxLns>E(F3oYnanCcT`!t??L9VTJ1O`=4Q?94?bFJ_2#2zu;
z^=R03x7Ga^w)LgF!;OIZ^x<AbN}mWM0*OHP5m>arcfRoE_3x@x(b=1=vt_a5)>^M<
zl^la!Kdkpy-IBZenwv8u0*OE(kO&L{0<GQ)^t=BcVAe?)x)sNhZ@w2bn)jkoPf|~Y
zQBOMQgOgVxkO(9Ki9jNd2s}Lkp3jYyK0Wu|)wcROp;r6%ZY86e>HZt<J$veTkJj3y
zo%;I=qm#COy~e87ogUvmBy#3NAQ6}-2;8vuYrenMx{uU`mBzm}Ic(+Y=kI#rUKcE_
zjg>aj*WXTCHgC_l_trX1<MaJya%6iVkO(9Ki9jNd2qXfXLZH=qnnCTq?11{ho98X4
z^@?^1|Ku|<5a{$hSF8Lc2Fkh0L?97J1cncRLCLFz54pjVc>Q%h{`$YYcv8JrbTCm!
zsk)6otA9^0`|mvOhK1J^y~d&2dXW4Rf$@w$ukLl?Jm);}p>2P?%Se%QFmY(6^PSGC
zTfKK_*1wtj+5Th%I=!FCuQ%U~hx3znvw6w5cBvW0yZ>(HXR`g$BQBYBTR|U^CIX2-
zA~4Yq7;PMwXt?BhJBvWO`C7NfjaL7My<cZ{Gr1)Ki9jOoR0Qt$(zJzZyzQ^GVWrQV
zy~lt4^&^w(y{1n^vbCCFiBIi*C#N+uvUeho2(*JhyY@L=$Bp$neXZ`nX#IGv-&g3?
ze4|}&Rlip8&Hn#DV7%IEzt)v{l?ZeKfqp%g?3RCTmG4+FZ1+3;W1Tn0CjyB;BCz_T
z<(po1)#74MX}|KLoHxIZ{B++(?w7}?I!}(*`hMY@^NueB<k!7kRjcu1wDN86-08=>
z;+wy#b($`iKL7YDCNHiH3SDvP@t@lDv1PUKLOcEawPE3R<f5Oh^^+f-Su~^lx<?q)
zytmbRi0uE5-$jnrk7mC_U}Xr5^*LJA_?6B1L|ec7V9wjy-v>6+gYn*8Wgl7p{O#u4
zx3owc1_OujW{0ib7j^r6_E!1#YCqEc+Oxl^{b_ZdCrbZ3xAo_HHsb~NR___J|8P9l
zs`8`bwLbN7IP`L&?^EW5?d64u+vsrXmw$iBH-7%Hm%sYK+Mv)AzVW|qe%D^>*GEe~
zcIs&_yY0>E)VoE0aLTVP{FjX$u8o!s3Xbi{@s(G;Xop2hFRe`!I%v-U$9n$lmmkS_
zdy7D?zRzr)_<hMvmxn#S?AGoN3ohMSfAUQP1_^;)pJxVX?I}wlkO(9K!;HX0IoEb-
zox{A>PsrZtdk^EK|FG{{%AW`f0s^=Ec;Wl6*<)J08GYbiw|?)bhi+6KE#3O!Z$G-;
z()H`(h3+_YtKHtX<<xqwX{U7LkZqnf>zTW(RUdEk`uV$_xYq?sYwgm5-gw(l`<?jf
zBI#h^&?&udmCu~_?f>taUvg)WG;!#p9vr;%u<hq<ytL>Qee;*roB7Aj+*NdowrW?|
zKM_a-5`jb@5l93Qfr!8_-nRP54}JUYTDx@mCRc9vi!a<%8y33vK66fe<tImK?b3r6
zT{nNtZ(LsMH$8041+xx1>7v@O(Jij|+wt%J=C#FmrThH;iVbEQeMZq~%5G=hu=6Jm
zTy|rTbhvSN-wB(1?DDVPTqI2#QV)7(2mEN#M4+Dt?0wWjGm8~>7D*F_Vbp`S@4V>a
zFP-}9B58MVXw^r3<kGhux%pYQ6cdHEichEdA9&*(+dsJE?jmV_aCrOAXYa7jm+mf-
zj&B?;+yBp7ed{fk7fBO`)PsN8LA!D6l(#(ZLvwDpy%<#5?z(bZBCs+9KD+tthd=m}
z+lqGS%6NCOZT8-^9sj(;okh~b;R!w1_l(nLoUr(=fup0<`TP!aXTJkITHLc=B9I6q
z0*S!b5ZM2)_inMz`);pwi;mne<AU$~*Pm<MqGMZq`=cLz(?y?t-S_URwM&QXcOc{C
z_q}g;`_=z`(vf!-!$ybwJ(@2cKI?<W?R!VjZQ8y?=g35$bp$?n<;xzJbIhSdtF-l*
zx;tXn;!%|Mr25}K^!z<<FNU3V+dp=T|FC>-FUlG9sNHgRt#fd=zxAB0XS{ITyNf}g
zgM&;;k_aRMi9jNd2qXfDKq8O`Bm#**B9I7-9f59rZ>d*&_rIjr;gUDpUG$0`wd4~Y
znY+!c#YCaITzukIkNoK7+IXWgZvWp$?)c@0YQsubJ^#DMe)Lnntqm(}7x%GyyyTDw
zBm$j6;9ZY@e!GhoEG?2I4ygxe2k8ge^8=mo$F1_&;JMR}dBrz>SL-xwRgUbR2qXfD
zKqAl{0$X0Qe4{t4esz&_V&Tx9w&bWpAQ4Ce5`jb@5l93QfkYq?NCXmrULnw{_f>M9
z-o78yX}o)-k2z0&5O~3bw|sWiv#%|ZCJw0w-Lr%KSa;6aE&_kw=yNZ>X}9ZYXuEuJ
zTq2MNBm#**A}}5h7%$&T8<cen%XfXgdh&iBxbW^fit$3bz5mhHyY*eZI8T-nfkYq?
z=oSLq@?*(&IKSW8?|ti*o{yIA2VPX)_PF=`rWo&Zw0LB{@rl46ryX?E%g#Hu7<M{)
zvvsyCmfTuvm5$F{%~d4=iNMMb==A#(!_IGISY=xxkO(9Ki9jNd2qXfDz&{b__VcFX
zf9VmI%(|_ZC^*Ew_mSmeX3l-pT}8ieSbfs+O)tA@agj7}7>{}|ca2XR{j=pa6-g6^
z@t_C)%*>wrzlpBjPe!AArPJ4&@uTM){)bwxY4=u?oD+eGhrpow%Zaxmx#C115$H7n
z$DTN{)gGU_w@BI=4!z!|oG%d=WCRB7{(6wHO4$;DL?97J1QLNnAQ4y@0)Kwi56_yi
z>fdU^N>4v*hpU!9e@cB=>Gxhg_2Oq;@yA-X>B@9qWo^B>$0WlxtL{AShK1J^?b1%2
z6UdV^5qJWDe!V}_t$BN8*S(r&*#Y&1H_uy8OFEo5q?egk2z1M@HRC(6P|npR0*OE(
z&>VqYzi&3)=UXv%-gj2)^H^=@=&8qlYS+h>)mHR~HrLGTkq8VQ0?qQ!Z13;yk`n*K
zzn{O0OZ{89{&ngef1>kgHFCaZ&F617=f0&yziF#-Xa7WCWeD8yrD+S-c-voVqoph3
zm2HVYB9I6q0*OE(uyO?2eUB%O>(zS%<87V=)8`+5#pK1cr2WAm>hZ9Z`*lyTa<7uD
ziNJ6n&@Ve0lzE41w^G_fAQ9*X0`2;%e#n}04lV+n`aR7Jd%x!UYpwf8ZM1Z75lP7s
zfkdF+2<&^t=`&7Pd{;3j^uQbM*#5yKcNd+a2QNKr`*|BLEryluebhrUixqbk?b3eh
zSI(aZBm#**B9I6q0*Sz&BXGq|Z~FX;Zn>*AUg*Ex|J>ux*y6fkywYy#Q@ii$c1x~q
z|J{;7TW_~;9Cp6#o(D%eZrJfj`G*sMcKuJM$3^toeGZi4p8j_=5#Mf?b0405AF6Qa
z_I~u6f7QDCUEgn9a{ff17YK|tUiaG`do_Q*f8SU&e=l?)=NLW&p7-@1ov`*7uPKI=
ze);fOA3Sc~JBneWz4AxhnrF2452NijR?DgKd#%=wRgN5=2qXfDK&KFB)qiFG!T8Qd
z?nCawAbjs8_3$5h*eQRUd=h~~AQ4Ce1`~nN-batM-(apjrAh=kgFv_L4aYm*&Y+)M
z5`q3AFkb95qBtnu;fiwgt2|!zIoHwJdtrl$SGVop#r2K%IpCWc*2hYl$vLdw#XIMb
z4{iJFT}Fywr~T6JW_pqBi9jNd2qXfXLEy+mKV9o5KRmPOHf<*DSnm^C9p6lzY)=Fd
zfkYq?NCXmrL?97(dIb7)Pta`Mr<XQs6M;k^5l95aD+1mAE>QAczUgIGEiMv=#3A)y
z>~=6(KfL19<3F|QW6NrTPTT!ETBF4w`y~R2Kq8O`Bm#**H3A>n_L^-rIqx^cM5Wcp
z{G;V7uYA!Ci<VwmYn8T}-?Tce+5Kj(+neDzXzla=?ZB%a{pF;3r|6)qZ&>7MW?#dy
zj+C=K1bXeKa=yWQkG(zZ8T+U~`MpZJa-=@@N1yY%A1nX<%-Ah6IV1v!Kq8O`Bm#**
zB9I6q0*OGo2poIj$X0uN?%pEl%5i8H-yD|+Bm#**B9I6q0*OE(kO(9Ki9jNd2uuV7
zTK&FCr~0@1+&ss9;^?0(zo|$ZI)lSRu<=}DB9I6q0*OE(kO(9Ki9jNd2qXfDKq8O`
zObi5;9Z+9*^SlMMq}|0~VrWyYG7(4w5`pd`(CYUUvj5KWZdiC-kvNPEhhG0axBHfq
zGb93uz_26Gt>^y3j!Vj)2qXgiLm<DavGVU~M11<UQ#o@YkO(9Ki9jOIeFS=aj}qtW
z{t9!3L?97J1QLNnAQ4Ce5`jb@5l93QfkdEx2*h*ncq=!14%$EDbLK=K5l93QfnFdm
z=y9!Coa%P__49W<ajy%O*7{9nZ??{s#gbcVNfU=b)Pu49-SckSaTWhw_}H8y5l93Q
zfkYq?NCXmrM4<Bs?6=;?m+v`agL=F4^z~-^=sAb~q1KFcUh?Fa2uzyPsqeyeo6myj
z^N+t`^5R;nbkM$=(W?BT?Vq}w2qXfDKqBxD1P0~#LheT*kO(9Ki9jNd2qXfDKq8O`
zBm#**B9I6q0*OE(kO(9Ki9l-z#P4!VlyYk*W$$hyFeuL%QjTM0&VAKg#aMCZw!S3)
zL?97J1p19Yx8K)F{+Awc$*kK7IP`nJbN)mi5l93QfkYq?NCXmrL?97J1p1A@r$2hw
z=70bDW3^$USM2nr&%fxFyK3E{{oa>e&%f^(r_VTH@m)pI#3A({?I8U?dwyWY9Y6l*
z9T#0+v`eeLSCsXiIQnPHZz=`_ha+aaZJP`K@|&Vn+UvNU^Cbd_Kq8O`Bm$j4VBx!7
ze9>$7{8KGy;xI^h&<Wj69*ICAkO(9Ki9jOIUj%+|%C9c`|LmQ4{FhVu|68>Zl8hxa
z49XJumL=wX%UC~zA|_js$`(e5M7AtrFD64-QVe6unx#9HZOYc7MY~odm9$CyejYqO
zGmegPFX!CL`+fb>t8?GyT<1Ekb6wZD&bjaR{n=)_?>rMW-f3*JG51dLOc?e%<+jh6
z(q&V6&l>~0w~8I^Tjh{jCp<sS*CF-tf_a+{(N`D{2801&Ko}4P(r2LC9Xnk!@}E<E
zTX@2RLwa{w_=PW+4<6FDK+!-L5C((+VL%wjKLZc!|H~Nqb??`)>uzO#DBHfQ;;sGK
z^ee0M)4yHobltjZy|N0o^y^yxr}7?U70W6f|3jBP{klGSd#lM_<q9uOIIMSJzydER
zv_*N(TPqKL>iBQGr0|PIy<1ILGTSRQ9MSiwmFEsx=G#V<CtLF#RnCipw;#7f+flv_
zqsq-nJKeZ+*VEQ}+Zor^nA>BrXA9FFPtyJC=2dq+Y5k<XY%%cuvK2e*x$l3xv=BC)
z?|#>_9Ufflhs9~{j=6a4ym`JY?DfXIb5E%`(GQ9>>(t!${^qm%sF)V-YQM}w*0J%$
zxApx;&+mK{<E|G;<%>c4i94RGwvTIHVdDe!<Myj-KW49UcE7Cq!(Lj5Ub^)?=yf$7
z*eau+Ep)XXwWXiP`F#e0uG7CSB|mqA#(TA1en#j<)3o9#{kj=d`cl>Nm$2ouM!>wN
znAJQds9s_D&r5xso!H7NOmcMn2fc2R&;6kC)Oxf}8dW{QmV@S#rg~xfrGC<Ux#};d
zA4fe`aNTaqxLS2OPut*!#e0rCzuB0!t9)A+_xXg$PJ`ZOpP6?w@rfVW#-RRSs%Kj-
zktYl!!GN0AhAqd{(bc(1-sI|dbiJP1mam?Vhb`OIC#bxj&t39bFK9jywVuv?gI*7M
zK9_X8p!Sn?y`Vh1u4gO1*2!_LzfS(%(~j(4<Bg48Slng6oL<$gU+%@lsD7E1cBSSY
zVb?wC+uFyO_EUMguk*{iT>rZI8{52c$7UmRs2fMvc#~HBNA5ah$}MX)E7-=Q@_1Y4
zZMu0>uiNS`>vGg_>&kEY@AO}NX}yAL%sO95>8G2Aq^@UW{aLf~v7c|)bFCj1lir6&
zf4_d;GOzpayz3n`kGGZA+LEK&UxvM&RJ@davcW)B)~&F5B$ZdV?x$GR!MJhEwKsHZ
zd`tT$s;%?FqE4^4<F(yBtR3Zj-OjuIw&!1!r?cCj@6)-i=lUF1H~zTi*1GmV*9rCf
zN2jOl>-SzcYR8tRuJrPXLHW&ky`<un8V|zeS89D%db^GTs{FX)2>U*X)c^JGy2ARW
zI)ARpy11U#>$caApz=)b6&B0=pnPPlURv=eX&!j~T<mCzS2bS<TDH}{D#v!5P~}rL
zosUv4EULdl%xl^8J}_<hrhI~VkE&<d^B(qGNUd+mmv#I_U1xOiY+YCNGL0|l{h;x)
zSnJ_8X=7Bs)wPQ{9-X|rUsvOWT81yvxx3oVR1TWgMSZ@jo_{1&ww?DxmB+k+rgbcC
zd=7gqK>e`uYqG1T_k-G_S}$sUw%e(Csb%aSDt0T>um8={wF@d9ari5(23)yiAt<Wt
zb>-eiRk`rMS1J{PV$^ZjZkJS?(e-0{JZZ_Z?H{VVuyIFMKd;YY(wf(x{i=C>v6NZ&
z1b_bCW>9-@RWIuP(9|xhzNY-k)Ss>UCOx<x)E`wIO#6tW?&p2qmUsPaotvokl~mdG
z@0vcW)Oqd1nd`l<Xxi7Q_ig1XQF)HKpOtp``1=Y`?Lyim!+>tz81;HGTK~6~pm|5!
zdg^{LYWcsK|NK9boUH8&RQ;1K+sYG=qua;3x}NuS%M~w4$+fjFN?T6e?I><MPP={F
zxSVv{)b(d8|IeD7wBF0-JVfyK|9+V8vuS?v`hId)y|cD{splfD%DOll^m^31BWivX
zR4-TcvKA+7jWg^#+UhrNa=t&|$+x;SUG1ffS?7)YY*x@UUvcyAq}!X$dGc~U?RXs5
zuF}@ucK>1RCN6(jYj2xR*YfiIoL1FeeXlTTIW4<V&yk|)p<8#u+LKO>>+7c9O@+-@
zb@f%g(<+CJo5fx~>3qm!&nCMt`ITurY8-P_4tvfKcAp+KUfF7Ang{j1t$uCEG1*Dj
z`?~klO!j5+2i<*@2b15}zOSCM=cR0#SCv=S<-G8gH9eV^sOz&We?SkxUo&8edr95@
zwN~Lvrg#zbe)yh$zeU#kQS>>icVWN+FE2d&zIFZfu5MqayZ&1VzrGSQKUVWdwXBOH
zLH$zKp6hX+KLj18T2Gfxay_nIVV`pbea@TI`yjuzruZctAJ{=q)X52YJ^o1SNYMFp
zUaqTg!gV>QePpGc9V>}1CG5F}?fQ(s^m`lC?%8*!*D0I)U5)o@IV*M&RS$T`3Np`G
z$FP6zC9Uz8?7P(NtMg!buA;kd>s$gkQGd@9wViHUQLmfg592gp(Dh%<bEDd^?R{WW
zJ>s@A?W0WVrtSN-_ZRrBF0V-IxwdWl(aB@{K~cw3sa!YtL)!OE@ly9ZNq3z*Hz=xc
zR<*0R{g~oB_k-%I(<kipsBuDV7q(wh{jm9@s*mk*(0M8Lda7Tj`#;Lqb5f}4m2^33
z-kFtluGYb*^N^I?*y<;)-$&(1*Uq$_hmDV>^_lxRebTy)UTI-k<IZcpw*1kyoUEJ&
zhSejgpT}JfT-hJ_g)4NWH*#I+EpmVV`%+QQlU(zv?mJzTvA?XKDL={jebfFm?ERqg
z9(Vp++jH9T<MI=<z3yBx=yg*+S-1b7`lXe3+v+#yIBnHK@1TBUIww14ZIerzRax)r
zFlgMNp2==?_Ze5-5gwApp!T5FQ=bc(%C_v-Bq!|t%p}irUuTzDyKWi>_f6}JdL9*a
zoU~V;f9w1gy>u9MzOz<8Y(A%}A9fy8e~w$u8n3qW3CnBP`Z~Lo>p%Y;1Xn!R@*{P=
zb#cseU6oU8Wn2CjRZdhsS+7bB<DS#TZJ%@=9Cd!9+LfzzN%JQCZTkC_#mdX0SL`wO
z<5ks0<PFD+t5v7-v<<$Bas4ptb2Ivjx{l?w9phH<<B^?*bUw1uh@{bUzNNb#^}QN(
z+^)-E<Bw~7T+7#uH>_Q$^NZf5a}n<2pXKcar?Wmp&1YP-)9tsCx~{X2sMk&Vxvbxh
zo1aVjn)}}0e3l=Shp2Xzm3C>_GjTL3hRx67)(>hAN!3d#zDAv&q}sXiBa{9&G@W!s
z@1d1PxI$O!wMninKhElPSM!ftlO2#BMn&>U6;1snbw923!B)R<<9^oKyLxWoT5jCu
zkf!#w>?N$6sJz+s->C1ax@u>8d^&ka<v~gHlT`kw<CAsI7Mk*nyxdpU=e(8k!n01#
zr2NlSKkE5oUdyidaV<CO@0x<v*Tqrwder@a?RKhOVavAGg{0*qwXVAACu{qVu=+&J
zlZ&lg*!_jnzqZEQ9+SPS*+JO2X{)|2KT7L**mc`ge2gkLuXz}H2hCG#)uUfm7<C@g
zY8N+;2^tr=73$aj=IPpnpvd>L)5fs#;>xaq$~Dc0?eFrrl9#ph(v?0@*FDobN4+05
zFHWm|)bXX&&UGA9-yaQI4!XXY&JE+PC$8J4HUCljb@hBhmCL;9^t3fE^kW-!dUHK$
zoTXjZeS)p}=&z#dc#@R-pyvbX_+6FLTF;8LUsrjPPOdAv$hzF1`>Lq*Y=0kICr>x8
za^3Ckj^pyH^N*nW)~wVEng?gKUf$P{tn!Lps-5q5?twR)a`HT1hqUXt#rH$LTeQMA
zVN^df<(s<v-Sl2s(ECx}BM!SyrGDIa5j5{q>#6e}wH){F0w&#F-7mT-FKRfi#zl8l
zErdm#ee159z6+Jr`?~SCzOJ+LNA5ah$}MX)E2wC)Pt|`-=j^6+E$VwFrhawzUD>~D
zxlz}bthLk4OIEMjKJQZH>HJf!r^^8Op9znuHSqZ9FMr~zsLpfJWmoKx$C)s$Kbv^Z
z>V3|gvxeA1)b-ZX&eikeuyPrvpfC`I0bQOHbRVtObDh^CzgnZeOnYdnVICNFol*1q
zte1nv*&(NmpVxi%B40(*`HOnL#Vg$pZqu+z!8XQSFKMr{V^`NzyDzq~Y2Ax@pV*|@
zS6<3-^Dp9;EsT2Z<H|2w%eBQ@T5{s9Q&HR7iceAdkJ?T>&yHKp+PWTfJY{>nHRjq6
z$9X~hKkR-&>W|v7<*6$@*E~e^i>$PZYQI@&NB)>sbiLlXmcRG3Bm38QW22WAuH1M=
z`-&gE;OkK5|E_r{Hu+)mt)S~#vD6D1&)N6e+Arlb?x}ibWjzV2hdPeDmQByQOXa>U
z&ZV`!>H5{Z&l>c)Ze0v}-S+zfruR5p-4E;cN!QP6KEZilSj_5p(92f5vvqD0RZiG>
ziCW*4UnV6t>UntBc5&lI*!c;Xmu9UVewcUE<r8_kZhM`#C69bt^~1RK9M?~E?QN|q
zCOPag<0Aa37&Z>su5WAqTr6^I^BY%Q*5jLM&v`GK;z7{+apyg)_HpB!+CHsuxa!yR
z_n%Df#m2qwYFw_r8<6!lrs@|r@5xI0yypEu<7`&zCB?g|cz|46?}>!v!`8eeotH*E
zU$WiK*1k}c6Lp=yp0a}KxbjxEZ7)g7%gXu~Rgb*xb8N-KtoX62N7!}7RG)nRo+Hn1
zHm2<=U&R)W9CgG$8&n-(LR<Etj#u4psby8)u;p{sHo3G}mGyp53>!y-@}iC_XgO^D
zqV5As<)C~99jC5d(D6#WF&D3$H_x}lL)3Vw`l;=5P=CxzJ)Pa><+>?O<mG<Ud?>AU
zN$0J$`^$@6nDo>=r;B@CeIA+BvhDb;%8R-V>Dt9zA9Zp~*G+cBeRce4my>#bM*4a3
zt!_<Mdnylc<A~bcWKX7jVb<^4IzJUT_s%_~=0ra%55%#g&}7f1`?~j|U0;v-9fh#(
zSLda^?ff!nd3o8d+169%Uuo@&O#SNgje0$)IOnP#Q{JlH4|}e-X6Iu+->~OeKPbkv
zPm_LO^Z&5=q`h96o>Qv%pju8kZqgrlrOMwnt-BrDuiH`|3_5@6ykW<xd|erTLoZvX
z)87=gxUY`ORJNseUgfCvo!2scl2^=|{Y1?N;<hubUs3Oy;zX(5kNZ8Zq}oTFcU`-r
z{gL&{73MXr2i04bHw3*-K4S~x#yQvR(~f6B=f!qCSN;`MuI=%qEzcA;ZQFIReBY*2
z$D#8ZlO4IbZ`(eR=SttC<wo6K!9&(iXMcIQ9yC8Hm3p@9%XChd)O}myQRSqa|E1M`
z+<6OnPH(%Ot-MaPgQ(@V{dfAWzO-H;C}uUk(e3}kUN`X=R$kEhrhfBwKW<!5^C8oF
zfogrV9CqD|tAAG8^L!|6jLT2lyiWH#Ea>&P`e(hp?Rb;6z1doKkz<NSS-&53|0nHw
zx0<qKwilNN>?p4oH~wa&eb9IxwO-gb5H?TATK%Z|GTZBoPF~n`S0^Xv^`xGg+Uh6l
zI-OMf3Oc^9`e&_v(EVcAdO__~S1)e7*2#?;S5577{Uv?fq-WgwLC2%+hx1Yn8(*TX
zhjH_{wA;Hrmy2tsw%Uiyr!rfA@b=@jXgkWc^`43=Ibr=T>-BBfp-zq(k6rPh_LFou
zZa$D!`>^?cR_n+8&M56&#SNWYo!^+Q>->oPUAI37`uxt6Pt%WzCi_Us&eeY7#s?k$
zS-Y<Ot~_t$wDZWQ{U>b)w&pdeU#M{)D`ngJx~TlvZfA<KrFP#ozamc<2*Q9a&$4x0
zr-SYD4O6~n>%Kakp!sNC>e=EcX*ue=sO7wTPb8?Gaq|NBQTG+L%k*PH-TtyvuB-e+
z&D%@0oyuR>a#H&plU`YwKa-w$yKg#=(%pA8&Y<6+PP(3shrC=*JI^BTab-8Q<mP={
zwY5J9D#sS@COMDXb;^`m)@)XYi&4MR7q^}3?++H6e3L!6y1%I5yc!qXS+$Ti{IF8z
zwG(Hq_gtZ^_X|w>sHpF?q}?tn`Ds)=(&jJi{!RXF>p4u?a^jwkq}@KR&p&LR>zd^0
z*2k#vA#3fT+M$jY+t<_DZ<*q#Zv8d!X_|lA{1uD5sQFmd#&2t1ob<Tj>TQa%y8B`O
zeq7Z0N%NVv{-W|~t6kps5`Be%O)*gK!S5TtefbYwv0zf?_M74|%(bNSq@SSYzG^*F
z{;114ZCy9@ANRg)otNu#dR4oAxu@enwe`69n5n&K|D)bl^^03h>Up{Aet2FN6kY41
zlOI$+T|M1-l4=JgyNh~X?I-DS-1wWczG{D}Kj*F7`=}}x9{5V7LS8ZL^X$0&o6fbe
zdOxn;W~F`J{8iOAE9V4R`CU%b_p)^BdtR<b9e2=n@Sv`LLCbOXKSA50Pgde4`hYsV
z;a7*`cgTr~S&6%O(<3V1YCFz3iUqODV!^E2r7k|??Yb-bSN$!oWi`(WTedaswDUT~
zq04jA@>leVi<ADa=GBHb)G3%S>h~Ulwll>K-T9U2y2*Y_{3X4vnDmKzUw2NO_3L?m
zzgrhS@^U@yIaAp7aqV2!zRkM#o_@4b^@0iaKjed@CtbLG!8Kn0X}g^d+_+sKEQZ}*
z2L1i8TF=&d?@{H1%{!<cwZEwCboQpZZd&ic-lreN9~8s(W2-*;M#Z4#F=6Y4)hDPQ
zCY?vb?Qi>btG+mHYK0MLq56JB*s>|_%lds&UYphXY3+w>_v`9>ASt=F;&jq-&RN^!
z(q>iG`*}t6d|h2n)caw}VdJ2#e%$pTZ2P$JO6Ad1R`b}j$|gNs-4A;1;<{eY@!4LF
zgZ8U?|0(MAp#4Uz7j}OVwV$lF3!B$dpK+<bcMW^~6*TY9TRq!(R8)D)Q`Ru7|7ERy
z-t1A;*LD6J)!x!R#|^4?*6cB~J~{b&8e#Kesh^L(lTvK+6gAJ4b~X3CzxgcR#6!@y
ztk#Q~H-&8%_WU7g{jhq4t#8{elRoz{^`|?>(Or)^Pr7#4r){2H$us$7)cf!d6+fQZ
z_u&RdS1jl-s$5e$-MprK-PQ9Y6CdjRwDWiRH^l+l_ru0nTlK@f4-&V2*z=vB^;J8H
zTeih-R5`ZoD{Xmk<BRQc>9F!_JqHW>J+8RVYfSCKt~(}qLGPR1bLGCP_YKwfkhXt>
z?LVyjQePc+*2+8|$Sb<aqixMo)O<Uwc6qe}TY59^<fU<uI1m-%_8+%>(Dx~`TF+H{
zNJ?(hd^Tu1>|Jov3`F(YpzSu@Yx-ARKW}s`KlJmp35WGA3|Qc$-A;Y(>$)7((!%D?
zLF=pf1dU@s>)Ga0)i>z*iR*f*UO~%At!qj3<2uf`mT#M<pmq?pUfcQ8){g4Gb)j3K
ze*JHruKgP{@uc=gyokFG3K|zo^>ojh(9?wKIgPFTN>Xy-?jzD}@0usoeuI|d+PAb1
z_CA7+r?lRWx7BY}z8@IW-fh*(>Ul#@y>#}dyKaj&lN{4Jo%Q>+^ZTIkOuX8@Z+ktn
zJ+Jc`^=>s~$!yOxns_zkbxGeh=^OTb()!!#PtAw(R@TL7+t<U|gKa-{EiZ5LuG5!!
zXTD6&(RF{%HfUa$_I@gE|F+k!q~$TsK~ZNndAV*n-;n!bE?zrto^P86+x8LEUUl_s
zok!~aK3?3o61Kgqd?sx<Nyk6i{khIFl9C^E|DV-*uAfV}k{_2(-8q^mo{;zHFs$9F
z-=#=9uGs2d_xwb6UHATkj%RgUvK60>zOc<7-+6PYpI3x0+j&V+^6+bQe$_JL%o?hC
zs^z43jM@);>f)F>9$nt(>N@%r8`_Frs+_p<U|J7t-;b-WslBUwq}b%Dc9ivU(C0R3
z*TX);`i&ZQU6s*K#iZt+eqHsa%QJ0XXB=svtvo(0Ibnat7q@;|^JMB*w_hr?>n5Ip
z-q+=|ajysMKW@FW-&c>@f7E@6>vpWisvWQ&1$FDL+D=#2<sZ~n*CG1TokzG@Cv|e+
zNws6!&n=?LQ}uFP*3D;LuB+#dc`57cGcVVZ-Y>fLgP`v*nChAI%<H+28rSn$PRqYc
zb`kczZU58p5>~IM_4Bgc==991U!!+k{TaQZ=37~5$2ii$sD7K(c1g{vY5eRrl0uU{
z+)qlMp#7NgD((lBmzR3FbvLWm^TMl6PuqD@((-igA<Fg6YbVZJ@7dykIGGgM(mSr4
zw4RsP?pK{R^4z3Q7k_MB=UlbeFs=MIYQL_}^;P+*T_;`UIf8Ap<riV)*or&M1J6MO
z@11)}&53@}JQT}w<)nFw+aLDt3X@v5ZTFLx^*Qc$+w;;sFV6?jGwgbn_4>B+hNR`?
z#h-P0CUssBbREb`+%Vah?!BU*JcO;M>o4r}u>IPqulu_?dApvJKe_6MbtbQ<dmby-
zfBv44YaVp#jH~NW<4;zfr<nALdM=f<c47aHjIMs%eQMs@N40-bJI=vOej9fk%G&<K
z#Ea>^$$xF%XP#`KX?#KN+tR04<iJ~tSGphEreT#rP>lQhE-UR_=N%^b@b3!qzK)pU
zd|vLS9WPwjmq~8kpL6K)yR2Wg{d^#4d3o_`ou01auF3E6c0X$Vne;eK=c{?YpS69D
zP9Ix5tM9Ft%C`DVTTWDeGqnrrkEV4rth})GT|d_etB+bgs-38ET$L9!oLA$bJF6CK
zp({PJF8AD81CO8n@+W@UsLN+;U5|S%=c>Iau4Vmx-26Q4{i&&c(>Y4o_f7ie<-Th>
zj~j>6p69sz+p>eWJf__~soy=u?Z@>#-nD#N>vrDcgthOW<5T6D_KRuXkDC_+Z6DXq
zY_$)&UsC6@*vdiYFRS&=S=;2&W>wbvc|}`xoi#Z+dx`5muGhn)<p*6?q~7-9wrD%b
zkIO?|<AkcW&c968!>)g}>+}7KtYh4LoZ3Ea+|#vpl|P#Hp-IO{wZEwEg(4?z97(G^
z{%HFgCaheOedzAH;!WotTxWbHw8blSW(!SmJ?#CUal`a}jw$cd*@5nQ&~?EyPbNDv
z@nw>4YHxZ^Quo}N>n8oVk9}tagW6TxdalRo>N%1txncRheoSccgS_82tp{Q6tMj0X
zb6nTO5nJnqPL8Ul?Q+s~VR}zQy>D8#isin#4+&bfZ4auvyz{G$$5uIPyvbVqu<N=l
z-jG9Fjf%(#il+55tM^TLeb(=XwX>|%k9vL)w_Q-X4qMN5UK?~DXuDq4;;E|7m~pl0
zbe^`sca401L(Su|8qcEYrL$XG`AJ-V)wNgs)Af7GI{B&|rgGfB50`d(bsu3W>-dP8
zpTup)`BG9C)h{#GZo*-`3j-E-#6=Z#^PIKouJbxwys~{=7q{j5oL<$gU+xJHKl7mT
zKU3bQ<~>2ny8d+6b?0+({pauQ>UdMH$6YT}c~UM-2GsqQt+L9dDgH#gA2dz{t*6WT
z)$8i`Y?b5sbyDqh&vEj0y;Rp7TlN}NPSEEvuIpv(oHVLFarenV+ee*mQ@g11Y?7Pw
z{h&PR>V?%eDsNe7S1j?@Ro)cV9^%$l&v|T>qvm&6Zztyt$(+B~vP0EB;+_lZ;yBl%
z=HsfKx^h(c($3EBYh3fFoA<ca<J!HnuRQ#z<G=AtJVc$RsCiM;{<6}JxG8wf?w56c
z*t5-pt9dp(7Y=*hRz4k6PFD6qarKBAH%hHtT6ufi`8Bo2zsXB%#}DL1J*T1_dfLXk
zoj>dJP0K#x_M7$nx=BA<e^1Hw`yWZkv-KXJNsev5FBW;G{e-LguFh9X^AgssS$~p3
zRe#g^oAvuapMz($o++**ecxoirIuH#<5m49Y}r;^(#Z*{kFEMi+iBST;;u8Q9fU0h
z{d-7zPdl=IjW;%WaWUvQf$e&6*D>}}QBk-5FkO$@pRQe4`;huSzb_h>hqUq)TX8h3
zoVamK*FLTvXQh4Eb6Q>fwDU)uUO}&iwX0&OZz~=Y3m>k=VcItYz3)mtkt+-c1Hyob
z0n>NiY~45MA@_gHfbP5_Z`b4Qw|~vbZ%cXM<KvN?hjc!&(g-kUo|ct;1$wyJ7nsf~
zitT<}zpy<I%x_xwThC7CO5Ax#tG#+I7j*xTm3m3dc2>q~(!=zgLsIvXnwPNs=+5!-
zcHMRyPg<Vq@3Xj;ug1;1mtFIv$`9+ey86YI-|5Dq8-LX6VV?_Ut$y6+U$hU(pK0B7
zbw91VI&8l|c~tAU;w`LPRj;UJ-oK8DVR={UN9{K&?R4=k>({fwpDjJ$JE$KJPxFed
z*0-SiQIB)JsF?J3Lv8bh-3iv*_x|Ry{G@rXJ)cR-OPVj+{i*A@lz;v`Eb+0}G3@#s
zmXDz4%wgpwb*`xP6PB;M)DQZ;T3Ypz&NI|`&uTg8b5vLTspoZhDd+th)^<J-R9@J4
zpVj)V{2^(%>UU@JUJm-+N>=N|{e65=?SuY)N>?xJxoKAG2aS7KtCv+i!|Fwzt*#SE
z$5CB>s$OBsao-mR+n-wh<EecgZg6zPLaC$5OV-L+UC&H<1wALoYQ3b^Ti5;AdQaKp
z7q;Hl4=cx{SJeA)<8#{WUE7sw`FTIL*5wU(xt@3a(cf15OIl9S=aZ&588&~<YJKcN
zMaHlDj$2mFPuLd;Zu&di%wt|KuAjtxJ|4IKwA*Lx{3UIDlFmC_=Yg)}XLbLn(<|xu
zb=9Bg{2=N3I=)TUUH^Ugq~)jeK7i|f<F1oxJ`whuBI)``?N?M@inZ+8K6Ucb;#1up
zNjco#)rdQ9#kRjoJ1;i%uiCe1-OtPYy#Agd@Aj|Slax0dKjX$p+9&-!sjdE!_D@qB
z40}IrUY(Wpw&pXa9PBo2OnRT4wf@rPU-~Z%29k>NDw|2?QELCXa?rT1dmg8|KIso@
zUTt_ooq~$G`ljc4I{l0NdW%PnI^v%Vs*Z4lLD$Ks_0swshx&epE&t2P{-N0Ip7FVQ
z?vWR{@Q@bT_TQlEjjMXL;(#j0mA<a!s^hR-R^xEoGJcTMa}`(pWMv%*st0*9{Oi!Q
zz3TccmFsc+hxWG5&0YCP=Hw3E^!skM<EBYoTtCZ-zv$wRYyHBWuV$rwQorAiYhR}J
zy65b+ubcGXKI>ahjQYN%$*+^TpOhWB8jtOLNz(GpS=;2&W>wbvVbNrFs{PuIqfyU+
zbnTMn-Bo`N?EeeX{kr$-*mbwE?aOv3t9Wa_HvP&f{q%3wI$gK!TCc3aE&aOI|Eau3
zS;ex7$N$i!Prt5@-rj1mSK0PFm8`s;{U@E*Yvl}IaML`LtYOH&g#lqe7!U@80bxKG
z5C(*S7z{jjZj&=!KjKX<YZzlli8k#}<?Fe1n~(8|6>BZ}Y|fyA$NPD~L?>T569$9<
zVL%uV24XVsM!l8w?_1o<H(^@qaLnOJb74Rj5C((+VL%uV2801&Ko}4PgaKhd7!U?B
z$3UkmY7A}g$TVNj4iA}|;Di*C_cu5F{$^HrNyscprkj79w(E_<zFg=F+U22S3`hnl
z3<v|lfG{8o2m``^Fdz&F1HynXAPfit!hkR!3<v|lfG{8o2m``^Fz~w!{O!V-n-3W@
z!PBAo`;r5iUVQ91izoSlrNINnlh^lNW>$E7#MljUd{=nE*0*iHzR6r)P<Z%XJHT%Q
zg@K<Kxc%<u-mS9!3okGD_7Uf=8e8{czf`exo0c2=!>9U!!oyE?AW=*h5C-fs@Zt1v
zwYv?O<>w9UGZ9SU78B0C_p7rucsd-jy8iTrRcClQ)a~<i*SBfE?1kqaf5jKfCJ(`^
zMrsKI!hkR!45$ozzsn`v_dIo;pERm^NLd&V284lPVL*Lu7`0q1e8@P20bxKG5C((+
zVL%uV2801&Ko}?$29kb{qEvX5aTW&yw%#8t4mM;I!a%kd2>bWYvZcLfBMb-w!hkR!
z3<v{xW+1P>!_(=V^zTUXJolozFpv}ju6{3z+@!`K{RjiXfG{8o2m``^Fdz&F1HynX
zAPl%+Ag#aSbn>^uIu95(+YgFvMlVu@0bxKG5C((+VL%vg&p^`e)k=Tqd=G8qdzG3V
zG3PVC)KT@Q-p$*6(*CCTe%h$&C1qitlo{YTZ_-%GoXgmS0bxKG5C)2efp*tU`()J(
zYrL=+^n0CAzsnQ0UGdC_j6@g^284m)WZ;sAFL}31?;(Cxu{c>ze^kfcaouL4<}CAq
zVyF3gT;I6zaNmTb`u9WBai>4SZZrsbzT>8Mktz%b14%H@Vc5RAw;4Lh7yOxrBsiBo
zgaKhd7!U@80bw8=2Ka74I=%n4SN={((0y0l>iu@arG2Gg?MEN1)OVtv6)f%9kkJbR
z!hkR!4CIZ0pubBD`uoAK^~fW^yzwXc3IoD`Fdz&F1HynXkTe72O@cLsHh5&3pEVwm
zwg>6&X9j}qYoy+!^LnkE;R_GKLzeBJ)Z$v+o=*kUm*;eXmvrCr)Oo(}AUud2==~tA
zxUc^G1^K>9I^TB@-i3Gn@Tq=WJ4h>zNWak-xO?N+yX&0$p)Z(k9-_^@v=Rn{0bxKG
z5C((+VL%vg&4B!VHQeubUC)Qe7Y2j@VIVF8|5$m%W-a%uQg~#1z5A<-tE9!7m+$k}
z-7o&u`|}gKJox3E-x%@LE2n?;<pB%4q_F7$1IixQvO+;H7!PsnM%w>A1M_BHdfbD9
zXZvCCw#Sbi*=x!Vo}e2Zet*`A=kDJwoVodsK@&U^R@!Ucs?BN*^zw#T`CYH*VfS~|
z#p8jcIg;<%G&}vdE64j;!_u4$8MTRljs3b0+ivA2ezBry{BmCy2*ZH--donnao+>d
zwSVZP2Y&3`V76Z@ICy=h!>+9`+ZPlbvS0^cVvy7o2801&Kp6Nf12w+f_ut3dv(9_&
z+$Lwde#D!R@$Sy=54+`%*}kCgAa)>rKpYr-)b9OPU)jemcAU_9@=2#3{vTh5)86ho
z{<WdweZhD<{B{MCEC0*D?VXQ$^Uw7w7QnuLKC;OZcU3Nc|7tLA;>MRN6}F$d_t;0T
zt6Z2hbNO*6^sbm19{u9GV@BQhzAq>|h#i>yVC8$2njSIdGe0YM;Vq{v?fc+tzu2(L
zC)eyff5rq~Fdsbp?;0ro7Y2lZ-)G>si~iZ}-iv1XQStYs%Uxlhv>14!-pcy-E$-#F
z8vA*ZMlVkC!1{ffk2-hbI8TSBbCH>sU)HbF1+#p?P4iG%Rwknr2801&z&!&|-%~fe
z-!*1+{pk&>&hXMgSIoKoPP8k%MXoR)3}l~yw~sh~)!4cp`+~{v5cYkT#w`k64*L8%
z&xGDzx7eY7<5gZ*%)W(+hQfd_APoGPfvo*b_}AJ=Nf;0YgaKhd7!U@80bxKG5C((+
zVL%uV2801&Ko}4PgaKhd7$_|UqQ3tl?d*QvQuq`;EA^e|iyh?C4oYj)kkJYQ!hkRk
zoq>UE>+IWW=3Fl-e)aa%We+V}?CCJNX3|m^5C((+Hw=6@eO&EsLuUD5(Tz4DRTvNk
zN|AxEzsu8o|2*mIr8rkI=Hh1{E8qFDrAP74SUyMS`uEGFB0rzAFIuO~K-|BV9ydOl
z^0m7+j=j6ixgYvEwB_I4ZKDsZ)oGz;Lfzk~+@}4q7oLCo6<@Hld5G&z(th9F+YBA$
zXPpPval3c(cAvDrX}+I#bj=g;d-k7nUayri{Gdp?sAy~55ji!6Hh5&3FFZu!A*%gK
zJ7GW=5C((+VL%uV2801&Ko}?;2I8LEO8ad6J)~lpkJ9l#8L2QJ3<v|lfG{8o2m``^
zFdz&F1HynXkRJvv*!s5Z*EgB#3vQZ+{LG|iTFeXteJ@Jtg?o?87GK3|dong*Kp4mt
z1M>IWlKFdX!gDe_>%6Pfzh9^0MXn12!hm}QYAyO~&Y*+G`+32r`<bxc4~kmf{ftQu
z!hkR!3<v|oz`&W4mfv~E+!sAT;UNh-C<dPiGlpV$@3_>*5oXS%t}q}B2m``^Fdz&F
z1HynXAPfit!hkR!3<v{p7}({4TGJL*``S+n<LEAJ-7^sN`84g^YbZSk1HwR77zq0Q
zWz_EthkdUn>idXczmFRAJLJU4yy46WkB=C;VUC{^X2rgv&U4c3vci|>v1ta9@*C+V
z**GP93*Skui=}RFn~tvOUyb^{bz1GxXDnYDWaa&(e6c3l3IoD`Fdz&F1I5X}#sjuG
zZ0JLOEaVmYp7HjLZI^CSh>OKJyD}<aKo}4PgaKhd7{~$x1KZZwx7W<MUQo<}&Z2`b
zAPfit!hkR!3<v`*8Mtio%Rlt@ndL=Am%52;VL%uV2801&Ko}4PgaKhd7!U@80bw90
z17Uxs6;!U&69$9<VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4P
zgaKhd7!U@80bxKG5C((+VL%uV2801&Any#M{his9za7?jz_{6dv0>irLi85~gaKhd
z7!U@80bxKG5C((+VL%ut4F-aKcQJ2&-%j+MbY8EOGkoDec*v6-lt$E*Q5HJ`t=qKR
z;2%EK7c34Qirrpie8PY*APkfm151bRzs*TqR(e74;!ipZ{&1UbJh0gw1Djra=?`95
zyyv@xjb<)h?Hzx|b(@WvvrI5!Jaibg@9u4ej`9VChdkLqQt>;jzjGY4-=ui3)lb9P
zk3Lwb??gW?9?<mSW6xPU$=6}r{b<<l^-(|QIVAN>s2g9<>p}G~y{{J39@TnjzmF&V
zKR#mYhB^KPTi>?*`X+O8hu`mVN%uWZo#!Wow&I+~5e9^T|1$8m3ukUVWY7e!)bPIy
z$p8MAfiov9zw?l}FM2_-)BHWIZ(Mn}FZeSLaH-~7QXW^k+mKnlYaZZVP#6#fgaKhd
z7!U?3R4|<*abLI+284m)VL&~fE1r=>8_C_r_BrnJd7pY=F<NiZN*E9ZgaKhd7!U@8
z0bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV21=0u)9()RazC%X
zYm2IPDcY*dF-PUS)Y{qPJPlcUp111uf7hDt2?`H+u!A&wMEVp4ij9Gw-@9;Cuh{sM
z@dyLLKy(I<Iq~sMg^%v{1ce8&ga7(LbdQpj!hkR!47g#Sck_0iw7+S-ZyR5qa7^Qo
z|Nhc*jpNT~d*}}<H+Y~MmPD#BAPfit!a%7qAm3H}`Mav6%0sEzy^LEJNP~f_zK@dz
z7t*INAPl%+;F5<gdACdNA--#LqrFHi76v+0`Fd{M=3~69V^-H?(W}PL29Heh^Ug!D
z__K^d7!U@80bxKG5C*c(K-|BdC+)NK_u#R(fo<#T+iT`ruUIka@3i1YP#6#fgaKhd
z7!U@80bxKG5C((+VL%uV2801&Ko}?{1`he_4l{>rF~=87hKFLZ9T|%-APfitaTwt5
z0jG_+G1OZ0*_=TKkM{+I2eE_F^#fgekn6%g78$Vp{fc6fm&G{~orD2lKo}4PgaKhd
z7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKiov>8bHyEZ1alh!A%{h4@_`@(=QAPfit
z!hkR!3<v|lfG{8oWQBphT{v^|A%iA(#foYF4ng`K)_K6V*?v+yT-|@tdA(N7@CAj3
zB<&z8@m%x}27Y27FXyRYzbl&5I#MkCh&PlL0}X3G`e3EL6MaG9LF_>MpxEO;X+?J#
ztuP=A2m``^Fi@NfRGm}t!x#KjUa??t&XSDEH3L63X#dA=+t2no&EMnt#+8Trc|p^8
zo@?GjzAzvR2m``^Fdz(Mhk<6pkMGs)=J&m<p($UH``P@vhuN|0WHmXU>BYyMvv`s(
zC_IQAh#wSd97r~*mTqqk?ElNy`*rWvvFmPSJCyBMR`J$;ZTgi}`sv@Ub-HfdwO(0;
zTl#ga|5JI7vWjIDkN=@dpMG5*y}i|BuX19{4S1>fzD=vl_HA{Dyrl5#?FM-}5B!f`
zY&ieT@4GBNsJG`DZOyyL*)#+DH*Vkj`U58TL2=VklDqoS+q*RG{m)^(Elf`HZ1nVa
z>*4PWT(ZCqib1u0`_soSI&<(?uT(H7FL|%mWBrdOcI&p#OAFy|)pG~+X|ZmuS8A9R
z?`pryL)P)#MPqK8(RY=nV%GH*{p>IhHLthR`){Sx9Z<FIpk0^yuF$lOqmiwA$+jHy
z6ST{~kZSer7}}v)A#XUk&&ds1U;CqHLc46GCoSr^Kw27axP0*BJI-6~g+<%?9)94x
zZMUty&Pxkzk0&egF5I)xP3xb2$BT+t8HebRPX?mqZ(-ZTWoKTaj<fIDWtwNgsQpyh
zYM*ZFj$h*?jZyWbovr;j^1|XD2i$n?7abP+rHZch8DZrX3;$u`O4jN}%`c<23+t~^
zKNtf+pU;G?7p!-wl{N!wj~%(gkU5LJs2ErOu<<`@^-bqwLGSB$%-ebwHcn)veyKcP
zbTv-ZetVBRf6&M&E4`o?HU1>sPJRB9m9olzT4mRKq%|(-R~QJxfa(0k_5G;ltZBE?
zo%4j@Ug`=1Sz{n;`<Af!oIJYAn&C56_+c??^N_dwl&Wvk^E0(w*5*a@5eC9AkhSyh
zsQNs)_{h_@n7Gu>3##!bujQR9tXtQ)$6P-yhOt?xn|}tP`gz*zf}TTXy<S{@EtdAK
zJ*nqIedbSZ_pKKeUF(&W{IvZyX#aU#k4<{JK9|zjRk6ei(>QeFFV*XL$;U;{Z0?`J
zo|8%aXy=P2-ol=9=Dq&h`<6~Qt68OjU^pJqTIaf-eZ^ZFZkg>BJEp~N+<xC1TXE=Z
zA5|$NjdArZmG*-!xN~Kz9u*2f(H1YJd|G$kR{yFT_4!WRvMN7ld5hMU&wBErm3~;P
z+VQmQ4j8-47tA*g*h$Ssoi=ZNa~D5v_{f&Mo^5r{94{~Ue>Y$DZoL*Wyu9O_o2QgD
zXgSd<Rm5)ue`cV6r<PY-vh`<vsiAKE*R)<ybC*lL^0SHuyg%nZOLzX#FICjdo2%<V
zp9f^M-u5qV^q%T7-p?!U^Tbz`rl0kyr$bls?n>@<UvK=l<F+fjpy*1EuySpGHz=&U
zu=-}T{)TDqZro?oC%!9Oc*7BU?bBj}7Z<b2U*7e4VDU%Or}kgz7aQL8;*MW+E3EVd
zf94_Yc4|xiyydAU9&+f2_dom6H=!+lY|F{(bG2f5zHW<u+jbFDURHi59##)qzqby`
zlhhLigaKhd7!U@80bxKG5C((+VW3zTxOwoSV=t;xxgaP!h#lm|52`%>TICB`Z1h}X
zvBWtUhcJ)?16euWPGYRmM{owRVi%>RhwgU+NnH<aA)Bt5x96(1_1!cp`Mv7;?=ahz
zU#jnqo5r1%1**S8vaLth=dn^h+~4Jn%9rc!b-0!v_C3(7)wk_ur7q7DU$b(*)aOaY
zFAN9+!hkR!3<v|lfG{8o2m``^Fdz&F1HynXAPnS*0o~t2%ldWoxsR=~t+_LOZ#?Y%
zq@K@9KfnIFEw+BwnzkI%cNc8gy~z1B14(^%_t(8jNf`Jw1C_Sgr`x*Y*LYF!*V3dU
z3<v|lfG{8olqLf=oxS3~W+Q*_Ojw%i$f$(@VL%uV2801&Ko}4PgaKhd7!U?vAnShz
zJ*a-Jo)^NCpfFI33}o&1d|~ym_4~-Ma~t*dqDi$28aG|n3(Kq2FKq^*{=T%dS(DKV
z1HynXAPfit`C=gK-)&@cruRsT?LOmHvFAg5=1*_;t(O;^*Qn#{yLOr8g++DTQWgeW
zGLZIr@@oHKf2Se!Z@l-54vT%^p*VQ3^}DK4lM^>aR_%D&b_a}I=7&W!e(B14*X>Zf
z=fvec*mQE!zqcB;*e^AVJ5R3Lulrk#8XKl>@NA<=?_$g28P7pip1kbOjpzA!!9IJe
zc;?Dx(|uQHvV)+1uYISFo3E~YT!lhVH1(U*{rjt&*5uQTWBk10vFpaX_RahmUa{b#
zx3(M8ztsvaZ`gQb^RxbZ>RQhgKC)%6XItGf$4eSN`KHJBcb_rY4~w?_cb_M|sx<wq
zSG}yD?eXd4&D>%3_q%tjP!J5qgN{$z*NbJJoA<wi@!ahfj;MLWLO-wAcFFVeww*f9
z&nqrCcl_6#H=p9?9mBo{m6iIg-fs#TFPFaf@!;JyF7ngHp!(|S^{czpm}dPJ_=3su
z(DUNe<Gn_{pI5Z?K4efiZ-4suMQ08k>k0nM!|S`>`=@HJ&i3<)uHOd@D_<8MU0<Jl
zMZ;&`9J0vI3U1N*@>x$_w9*fXS<9otzUPs(`eRO9cJ|KQKlk&Fj}AQUvDRNr_w$Bv
z^Lf|tB5(4qx?siU6<)7YNDEDTEqvs{WgX7_&KJxE4?))*wVr97quyWK;FEU;H2A>F
zIz}B|TJ2g68}Qi<A6F{4!nDRE{kml!>2nIVqp48gtUiNIfAF0Jo^5ok_rZVLwc^M}
zzVVVq-MGfxQSt8^{xZ|gJ7(>Df@hY@Y4Y$HvwcB3JS4pz()o*AkH<h#=N@@~&&BrN
zgU^dRoplBM`|MHArK(Kbd)A+aPw{PGUijAOS>^fHDqqlIqnB6A%JWv89y&X>b^YZR
zU#t33#TC9_HhHi$&mu<{5C((+VL%uV2801&zzqXe|ND|FhhH_$%NvIMjzd!Q-OQ3m
z6$XR>VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaMljg#CU%
z()DeQRb&YR!hkR!3<v|lfG|)B4BYkOwgdNR_Pv)?to!bcTV2<8iB~MRWYKxgHaume
z7Z;QI-KFh*Z0`?>%}3MzuCOaUb-zorb=}o>P+iIGe)bh_ZMbE&mv;>MJKn6-OX|I7
zSN$YqH?HCVa@*a0K+irSm-u<bq`y1bVfQm;-PmltZyV!&?@!l$`<FL*PxTq^2gT7H
z4{dn)=5ze8h&^Qm7uP*v%h$L4%8!bDCZGG+`z;svY2l2?|EO0se2$-2?BA*76_;%N
znV(m@r^~1P&aVHSpEvw)i|gARdB+4VE@s7#RXuF~-QTsxj@)6$oW)*`^*^52t=q!L
zIOpamWer+R^x~pAUQ_vvZJv5yw|+DHsCe&{w>2O2>?}X8xcZ3~C!DZigr9fJ`}gAV
zx=u|z`0^K9w_fCjMU!8~y&pG@WUalceOOX*i{<a=X6+onq>uM-wc$f{{?5xQCUs7b
z*M2r!K6vsS=PmcLj&b9;seRP%>}9Q8(DQQ>A4%W0bq+3a=KN>r&R_arc`&U5Vegyl
zNB8&H^Kv~bzfwP%U1o(hlRca6pFFzDn&C56_=3Vi96NYm@ki6A_Fw6j8tVLO&;@s{
zY}KPeA#149H}3VIe<v|(_40ncmUsQ<zJKf~t9Kpe+d|v#F^Rn4?+skCzz@oUtM_t5
zZu{ofA27ie9)yP^>>zGF5Vife;|FZp?uE&oZH%f9?TY1`UmXX}xwDROpTno!e${ga
z^=Ywgu9sI#o4>gI+kPHMT*xY>?a#Wu$Kv|B?RkzXFKheVxca>P>EjojIe4t6L)-ok
z^tn;i>)HOjPG02gTw&e1&OPS(D%!RaktYlY1HynXAPfit!hkR!3={(c+gx^ci%%Z;
z!gGzqz*RAgA*;WSv(0h!-S>ES!wX;gdB-0XU-Q`gr+)3{9kce{a9Dk8{ax|2<eXgf
zPwNls`;?zoJn@i2N4)>pm%iZd^AI+U_Pl)3rK20I@$!nh4m;7mW=Z8jSPaWk)cRr1
z#iRB^yLa2(e0`m(-}m!^d+)XM<Vz0u$j>Y8^Tbz`rl0kymsgDY97@+7dyI-rCpZ0j
zt6_`%sHp3=`~LHv?r`EuzHQX?C)e}EK-m3JUVgu0Th9istr=2r`{#UDxLMoRpL=B9
zn_k-Z*fqybKID&YczMTHuWC@?^=BvfS;xHQ0Xn_!uX0+GPdAS7U7;FX|J?N1SK17y
zQm~C`KS`Gt*QwX&;5!z3dBY>NJf`#T+6%nAW7_Y1MeYBrK7&qw@SO!-SWMc!r9WXH
z4F<CQew$9epuZckRnOM%veTGj=~EaG28xS;-XqT+G;+#H&lT$4!;X7B?0foY&6@4s
zL#8zjTkDp{iT6F1wB}#>6$XR>VL%wjHUpP#yzkPpF8{_4i>^kM)$hM-{oZjEe%N`g
zpI6Mvch6nfjV-x({rfO!>1~U5lbp{7e7*h5^XB+@#YYDo_gL$%ru%uvq~f#59$eqI
z{r5Lq%R6A=!xy%BWxnSMUGpvSg#lr}Jp*rz-tMg{2Q2jSieZ00k(K&4xBT+pQ`^q+
zlS22FBt85d1J5j()8yeZX8TFw@98A>gn?3HV5bvro4D;IQ~jv8ZO4fZTu^<n9~Rp#
zd4AruQ|I}vvDD^M#w!d61IaV+RoP)j?mBOsCm4>0<Yy|IJyz{_+I9zwUFHjx77y99
zMt54(eRs#LuIszR%R9c%f7f%mO&RZH9bNJ4&O$`8O$M_5yF!zGVc$#5di}89ok+WW
zUf(ZGn>Snihy8n(S*vgB?+nI`iK{-SxZ8IXw=LuqpWL=X-HQ*XU&t%I`oZ=q>U8)+
z!8OLUlVWdwa{Cu5-o19D2WD-&qnFjLcl|jZ`-0(k$l5%LKEi-7P`nIO>oDT6HC4a&
zqGItf^Xri<ICuQloj0H2Cyl?>B6_LGJ?Hrw+m7_p#^@uEmcoEAAPkfi18aACcBe6y
zFY|-qT^paTck!BKzMvZ(O3V6Wv@j6#JDhp_{(D~aw*B4Jq~$%caA?a0pDpt8hI#!S
zx2kuaJytw(WwYsiUU6}QPu?BS-~%sjIQxo*&%QZik)KyozgNsUVGHy6`}??hhvhG+
z-|3qA39GlQ`WvRbyK$dUpZIBG(DA&!`@Mgv_Udfk76$G2?Cl15I}iMiZwrI=8?|1K
z^*^52t=mE`C`Rok>+M!OcTk@e>*ji4(bl+h-yP5U^%Kihb!c+l0xxZ}HBYwXeDY0?
z@9#civY%H>`<y>+|JA%FYV>H**K>tY?d86IKDz(F?bmy0VN|{I)^5(tQ_32&oan_x
z-j4~2qdOkj@bb;)_=2U$!(InmwOPBreB%`x{&lO4FJJfBcb=dd9&E?gmv&ou$~Aj`
z>*o!FKBtOWZ|4f@)^+YN*SCeHeZA}ZamQghKgEvYeuq4%_PW218236n<PEd-y8-lx
zJ0H6Cd7B^fjqCSe+eiH_NZfXJoP5$rH`ac_PYPZAJwxVAFq;g7^@pt0xBd4pg38P0
zY!;tZOJDqW@NOFy`FX(;4>@$i`=5R3hsCVVT;Be^Wm5Xunzt{<wjR~^naYK{;OEs|
z`1i_2RSS8=eRsI}!ky2oQHY8UF50j2BjeV1w(yMSzdm~D8|%Eh;<#D+|L|6g3WdDl
zmG@nI(?OqBE94bNv|IV{Noy+=T%oNvSFCbef43jrvKAlK{jElg4bwMxdByh+JnOA#
zW#4&uLsx#TlZ*X@#U~dZdHNO;m-@CbtR7jbZ|l1tapgqi<)*V&9N28+4_>k0q>%#_
z9(l$xFYma^<%9MtJi6YCi&5<&Xgl4xwCTD|{*1~0s8=?8jvp1T8SvV~_Fv8SU7_yp
zg`2KNecqQ=yRhHE&RTs}-wQ;p&VJ>(Fdz)rXW*za4{vnchDlytanhFk%HFthofj1C
zGq`Dqw%?yfTHZ$=UA=d&hlhGe<EGhkcXdtc6*YIc<SXA6+SW9wzdLHuJMQy(+wD#I
zhQ0see#ceq+4vbRZ@8e=iPw#&`--1eeE6|r&R%`MK+iRXwTEpVxNF1AGvD^}g0`Ns
zq%Eh;uKjv$@$5V=?0iM7ANPDH?e@0M4U?7^b>8yYu2TQEzc_EgG%s)XR9TZD&o7wc
z<rO!q+yAgtizj-vFs}WD{oSdqeq24I{jV9w+WUK=kKNy82%m2GY&yB=-&+k^>}L({
zuX0+GPdAS7^N!nXQGe{Aw@mkv!sF%+Yuj+rYkpkpb7ZF_hn_y#%PUq{G{5%~ZRUA-
z#iV{W<*J{!zxz+Sec1DjUuPv!ips#k8;;m(pB5uL+Za_J+9mz`nSNd2on@O>e5~U}
zPw@A7*#Cwe?caLhNiQ#$mwY*@T|fF<)x(Bfwa9agI{iBCve)veLza53QP;m*FD(X=
z{(F5<zcb<beepN8dFp}P`pxk3hTFfq(R-@Tc;6P<vi_jI3k_Q@srAS7o?+JSXZ81=
zw(QF!$JO`Tl9J1FW_6r5{{8%V)sFqbPa9ReT$i)%ZzlbcUT3AhIsaL@^Ox4X!yEK_
zzOMbi*53s)>3PnN*Y;d;>1r=;c+~|fKCke4r9$4Z!r;$44H;9TkT{wQBjo+C`JJhL
z_p`5fYr`$GJrzT$6}h_lcOCLp&g=6E^d9x@%@6nK|CN_l-22!L54Je(GcT`r*Thd>
z{G!DiFD;DhxEXEu(%ZW<?)}eUzH2lMKk2_mWYQ<A_k-ruN%JZFRjoT{*X2GuJh1qq
z=~Mf!^aX{7VzvWaoE&t)ohw`Qs89%sLC;kOywrT(rd4M9aq-<nV{V(#ca^8&;RoK^
zcH8RfJlCjuf7W!p(pLL)TX*~#&lc+DFRAOMbE~-bUFS6-KUn@4^!&<Jy`=0{*H7H|
zRIKgcEA8JYm3sfhl9%Rf9<hH{7&MQ{TD`pNYjwXz%$pq+o4)M3iv@LdmY3^!pNGd=
z4}Wjqk_EnN47v`u)*~tTI{OXEL)dz8|E?bG!}e>dzUg^nUhZGGXQP|eKmCrESIql9
z-`2i5Z8<^n#H`hG^*c13+^F$~zhfo%eV642_4bnHK{uaauWz_~@Z>wrTkd5At33Z&
z<qKMD^pe7;_HSyJcAlimpIl!z@shRsw(|^I<BR(|BklIK&!yot@B78P>JR^}(6+rK
zEiY^5q*3)rYTeTH<I28Fa&`L1b+_kRX|Gdp`%gN~*zV7@eVhC)Z}+3}uZ}b9{$}m5
zBX<}wXR#L)!@l>U{!YPE*1fOgJzQ=0ke$Et!eZF{bJX>Xc45!yqV5yo#uwVF_G2pZ
zzM5{H@^ZcR$nyt{oU+ohjVAuo`=)b3wZ83gkM%#E*sa?_&lSeiH|lwB+<t?$H^sBC
z_k;E;^@IUoKo}4Pgn>LU5OtrUw#(C)MOR^<*co6ynKg{Me+b(yEAK&t)!Ss3S)0$W
z`sDSw7<#MgUf#-K?J=wMgTDV_@+Z^%u;-OouOF1RtkmOu+q`1f^XIJAk80PPPYZ7P
z_vNm5F~v2xFAN9+rN=;8zw>kbKB~w+^7JhxF7;jWkTs79`+h~#`mV>l_g+g+zT}XP
z{JfwmKEl3FQtI_x%|qVgX63xkmL6F>cQNS|_1wU<y&Zntp$kqtrD7qk=!&1Lo;L*5
zEA08RZ63qEPo&njrB7HMlCJOi{bSeiljh5If3Eq_$+u<SB1aey2FwiT{4nlybI;kV
zvForC{cDy~F60&88(VScZ68%B1V!B}M16h}x1Fv1P^rmL^F+0*n`gN$3~Y)4-d_>?
z^>_Jg^RX%JgItUHyR)#r{~O)s<OZ#;{m~QLG!H@M>ZdwE*JEA1p#4a_;%30r?=zEP
zE$Hu*gVqb0Z@L;^P(9PG=lbtaxRPHi>rUL?DZ8@YpmMYF9k`(SN<Fs>L_KeIt5tOA
zQQxZ!+oS1w0a?HA`hB6K<VSt(thNg~|9Pn&G_T5AJ-+iIxNcqN9&>$HJlOi3lWl)-
zHD23#iacT9e;J6HZ%Mlw@BN~~Vm~Mk|I2xp{|A+?)(g{z)D;GVfu9*r{Yc8;;#}VN
zGW}iKyxtdo76ycY^cm>+P@nnJ+kNZh4d*rLIQy<$rg^S0{Ta%p2CmK%Omc0#hvJ%5
z-TMr#uB+q8`gyuZzr5vXs=l`D)Rr98KGbrtJa3LV4wa9vWn1UIIyqtWu~pyI_qA=w
zjr+Z$Qhnb)tbLT)d*w3T_RX(9V1i%jJcRWVTlJq=GN;MIXUz6(qxu|Q%8y@k=HRhj
zUU{%(Zz4w+5C((+VZaRoarY^u(%udG6)LFbGp@?6^e<MqarfU@X&*O#EY|szaR~#$
zfG{8o2m``^Fdz&F1HynXAPf{I1FrVtiz}Sb`-|^q`nE7C`=<BPlDZ#Nf3;oQ_ls0{
zQWge;0bxKG5C((+VL%utZU&Cp;r(kz99q2)6vMI+^!sXI<!+ev?#6vaec}rW4=(J$
z)%yju<+|Fhi`?Pw4P3IoH}PP5K1)Sj*m$4S`nuoy$o1y?Hmx$-7apSVkkmP3+WlnJ
zuSKuoV8GVz6ywS%4mM>J!hkR!3<v|lK(-jzw&TPHE~viP7fgqTrjwigz16VAzF=wa
zVDkP4{WcXh+iI$pS8VXwnjsapf6mvTY5ZBazx~S_y{Gz&_w$NP?m5rj*mk7v3bQg#
zqK8`s8a-XB)jM53_40~m^xXcL7hWCZ3knZn2kHAk*!K>u{`VzU4!>%gS1h==!6)wy
zXz+nota$GD0o%5FVX_w$-L7ULJueJ&-u1%~_wPO1OA7PCi|A>Wfg@k4xcHeH=J|r@
z@bHfpAL{VLwM%_L;US70nBFs!`#-;rXDe#j`YvkT{*GkY`o{geVyU*bW%nXS7!U@s
z&p_{y=MNe=Wu@mDRVI@DPNmwvu53H*NBw)<ruXx8`pb1;AV~&n<+pL=Bspg3OBfIa
zexHFy|9aF_x2zoF<qdy-RN1}zz~YalPwl_b7fhCi>{>)=Xl6TK7kO3d4%&6OUury*
zhPTKlg#lqe7!U@80k;f1u>UXe`*rWvvFmPSJC^NKR`J$;ZTgi}`sv@Ub-HfdwO(0;
zTl#ga|5JI7vWjIDkN=@dpMG5*y}i|BuX2TVPWz_gRmZIJS5$Aj#U*F2_51x{TAg-}
zuPVp0&VQ`aS$|*cfAmrPw~l}Pd;h|z|9SiW)~@iIH@Ro4L%;gYf9#IUy6#nHMLBZ+
zq%~L6zHO)V{>H1@c0S_cmF4aBdvE+-du%h>yMF%NyL9>2Fki))2kmp~;}vK5wO`n2
zzxF$Q>hIn4qF%fHb*SIsk1bn_{dBO`eadV54xKpL-|vkR_nv#w0{{Nyiw;<|>uA4y
zm3jYu=<N@@i|Z8L-Q&Qap>el?-*5i#TXQ@e&b{`Ynjasq&>MZo1$#EW=sWLeZ^)-F
ze({C3&0{<5_E@Efo_|-zE*IW5tX#kIhfbb5;o1@Yp)>cn`rT(Hdz~8;#?EOo&Fit_
zcCSpY^|jagPj6MKeb6%R?mPGCy<fYP-l0`a*?8!y-+S*hd-=p3JFNF^*rCcEr!QRN
zb$@2cr#D~oV>y1>{fe#H-dCZp^s>=|9vZdLJ9nE6Z=QQzrNTp_59)k+&x!?a-GE*1
z>a@WtyYRzn&mFhcd-t{>yG<NZp>WTQR}Jo6^+)gX>t6rkxa+2R*H?Jp+I#jH@0~Dx
zvvb})cfR-d>5Z#xeEA#i{tXv(x?{nt^6_2x>^G-Z8@SX1H-G!FGndX>>@^)Q`-biM
zFE7XQj^AnTEuUZJUHtt)T`Rmi*=HWXKR51q%bWh3ay!9pm@o8!AN1gU`TYJOpZlu*
z)%%Rx?9_`!l#hpbLqFR6`^IW*r`7({qyD!Ke%*N6PnP&9Vqf52UfZMI%$gH@^bLy3
zz8ThW(BQAi?SOdz*S+)M72oYP!(*JtZSliPFV?y3TkowyuY3L3d*+tUSGgXFA0dbK
zApKB&ZG7M7M{PCaH~u&Il$VBoJL3x<JAd&1%HC>p`ZS;UKI*~APwnu|o#lGK|C{^%
za7X=*KPdM@#sOj%Dxw$2I8|i+j=kdgQ@7b=xrZHs&$O@C^V)mYd%OOz!U^|H`LF%s
z2lx?wbK0~O6AmhT@4fQKq?V^{R<XeSQEU6o7&C73!Xaat-*w+b)e2AT`P^5%`&TU-
zeMpCSU8Ytl{PWEl|NcyqnuWS|?t1X0qbnESiMTPV=1IGCAM!(a9AQ4t55Hy}ut(Zo
zx8&h_&t6us!1!qo&(tSwa38;CUXjOq-PGdp@w+{-)Q5lOwffcG)RF)C%73Ew33v4z
zzRvH`_0Gv(cA4+DUwuyRs_VY?X-8bgk3n@Ec>mNXU#$GoqVn|vd!k?Lfpr0W(TDYb
zdiW=C0bYo|#0SO;Pb$I#e1Z5iNIyEHo+=;v#2!KT0KdAV_KX9s7~`vmJ?fBg>kxa0
zihtbe^~SB5tnzdS-#Ww(bQm-ZV*ZIY%+n56)gSpz$Ir{-5cWZQgm2oh?w}8L#5_@t
z^`U${_{ATHN6fcc<~sV(KX!wC;IGt2U*Zz&UmxR*Z?j>whkhV(wtef)&o6v?od*&p
zK>Qy$Iy+$=RKH!m^{j<Ezqri*;^0<wcm30Hf7txzuWRteH|6U!aiZb_Be&Y>)~Wt8
zx3Bo}UyEn?>#NnfaY@UCzM7Y@z7t38eQLm-Ki2u$CvM>H^rs?zts?Wt{F8UVKl8fH
zdx!sf$IVxG%rE&Dc_i1#pT0h>?t<1kRw#73=ZWKMy|Y>2vkQi|c%o3Fa9M?q2hVO&
zsX$!se0b+JXP>#&yL0jdJ*U4~vGDZ=uV20D@+t+^5fxcCK<p7@-O(ZARgrv8#aHid
zc+QYlYZO520c1Qs%vt)wCHwv0UH;0w&om#i(nHR_I*mPL@7dpZAmdh%^^b84d9nIQ
zQwkN!*F)kxSbK4=;XAZk=7abrNInBD?ofODb?ra*K>P&6kKpl}o;|NScAM26*O}i*
zway#5>Z2vz+9iX(Xn5m_a{s#Ju2!{Qxq4oC9!@{R8T24dQ)V3*u-%6@jqf$j$37Vc
zc3pm7lMmAm{=t3*zEs3-i6bC!iTI_)ANXPX_zn37?eP!fQXhXhqxF8n?*Dvt`TD>*
z1t0LvK7sqRM-TKQE|Omn=kQ~Y{Z7Lk-|f2e_R0knvFp7%jjVIg36%@1f7qjn@W}Yk
zn|A1XS<lN3y#2<_3dCWqGmep!%Z6RMOV#qc@y@n&wzy^U%?dS+?RfPAtv7nyM}PWb
z{KRK?pnvYqxoYVRn;%xUK)wi)=cxY3dP0Bb!91|8a~)*e!2fAae}^17b?1Xd*D8;1
z@D0LGqi%;Bd+0&63o0^?AbBvze1Q132^p`7$iXi0Py86aBM!0d=*~U3kDua~<SF<q
z>mzv{`Lep7q<`?(X8)@B`t7yK_Z!o{TzGen=c^PNefN*oADCUOeBXh6skruEpWZ%d
zVeN9egAexeI^J3Ln19wk>;iuxzJrXPILEk&f5b`bjrA73Xn)KZEpJ~t`Y!*b!&hy5
zv(p%le1h{K=8JahBk&LEasC2gH`HU@WS@aufY?3jIey4G4-zLp>|aIlW5$I(Dl#u1
z?R1D;s|ZgfWL!Z}wIB2(UV+3z%ETMu6zzz2*a`Pphgg@GC+=e}v{UOdkMN8A(=Txs
z{g^+>@P<E<x026d&(xzG_d)ys{nhoKe1!YRVZQMv%GfRO1^vqP_~qP_dho^i4^NDj
z{*lkR5AT%GgZM?ij1PYV(T8~;-qH{988>{g{~!(%C*dt9qVM<*D>og}cW(JU<b~x|
z&pT<aRptAcKOME(#Led}_M1!?->F&C8RhlRhxkGqVq740$hkIU_DRHj_(C7z5cY^4
zsOvZJmwbqImHi&b`7rq|>-9D_UVOHHdF=x6i1iygWBzqW{-s0We^?~mw7O_o_4(g_
zU4FideX5AQ?CY=_&e5@N{2d<QwS3+A<=m9Km3G7fcp~qG7wndG>z*@D+PTUPGt2i0
zoZFKRGOzd__D`99kjr&=R>#lzCV4dS*{?8e`bTf}ap()*#Anud_Cv@o@Bf!`WXAo6
zy~n&d`uGJN$bOLZlJ$Xo)68wESNzlJ6(0L;<^%gjE_^c{;um_u59j5~7yZB^`xMp%
zc)?!qTjZCI^Oxry*co;~zvM;KM=tSAjeE>DdJ>NqH}{c4JY*ciAL6Up4|2GUeEc6C
zkc-}w>7TenzuX7WAB0Ej3xrSX61^B7^^mI~{)2qvfXGpi`2ykXg>4_Z@RJtf%bx=x
zmw3)|IId$y=tX=WAB8{S4|yEpWd4zh9_qOm^|3?hGk+?gAM=Uau!tQo@01xAi2pKQ
z%q#YRy`Wc8NPdf7lE0uIafS7kbr_z|oB3tFbjUt}=N{xG?Dxp;m`Cz4@)q=<f7*fQ
zO<rk2)-x5cC-|X19b!kIS`Qv|NInRv^;G{PPJ!qP5~o3U1<@14PC@1qWSwWf%DNL2
z(E}v!<9Qc)k;majAbBM7kKMsDb~N|#SMNQh)qmfc;GBW{n)3x+nfmC3z99V)Pv9RO
zu|wXYAg*EW@JT*FUO-&H4~bvo<?6VJH>~5xK_B7@egxvb=*zh9uM;cnz2LxZ-}&SX
z=!;#U2YCwml8+ESxKAF5-y`Swv2D+u*l1DtdjseX542;QfJgMf?#Qpmd&do${&efx
zS9|1jtlQ|tIXusCf+G0}NPY*>KlvZ)EAfbYj(K4{cyYnVnvZ|E!JoA6-w(We!hhcf
z!k&p=*e!@%;1|pz^UV7J_y_fgr`RWU#`{0akIFap!uW_2*fDVn9@r0&cXCdFU9iqG
z@9e+ehr9*3r#9&E{r{a%tMF?5Yd_g=X4OL7Y9EcQzhm_R`6K6gGuqwNsNrTc3dg)&
zrAM7+)e27!IqAwhmenYKub}<7mg{c&_D2tWi2ILx_xO+PE?-mrybt}j&Un=K8Q5=N
zXRLeZjlE(ACM2#ff5bohmbgj2MLp&f`v8eY_&10j(;x8{gdgUc=Vi<{*YPWm`Nob|
zw}?yl5Bo97*c1G6A70^&eu#gxQ`bx6QXe}bpJhEGeo>F><mEhHq#w>bv19z2`6e#$
zypD5s_+md#9*mywgq^V;!#;^?jDz?B61V#d?Rx*dH?Qz{{)@d3Cn)3Ztb^OPS>E!n
z(^i(hk3yOFgWah0d5#UT&aqD*E@7Y8Gj@ku?zir9z#dKS{I>jk7~&Ssy=l*LZPsbV
zg<Zn~{jgsE>5p?N_(EU&fH+A$PP}AaLw)o_Kl&y9QV+jC9@wJc<Q8|{zrOtaAI3pG
zL_O?}IIkl4s%^x+i9`4!>jL8=KT`Fg46lq2-c%%xsz|^1J8=cSW*wnSJ^YaN#23}S
z@P810#Q(8Z&JR^&o`{pIXRHInGtSM?(}d`W-YOEuSbvEp#4-BE{-{s<p^RL3Q}yD#
zS<Xk{3FJ8%{g@D5O$e_b;|q$|iHeLjD3Vu#<hAfl9?bf%U*9HAbvbHX`TNG?i#)Gx
zH)wpvyH;;j*ldS&vma_#y+EErUUPky!Y<X0-mE~L%KId=JFn(ZvxaU{t8n>cZ=N}M
zzdD5(pYQj|2WM_x{@x_#)jVJ4c_#D3`!~FY$vVjTkDa0qda^E3kNIak#9p6nw5s}N
zM{Q9cKjV2T`$+tU>+plVoU^WZxlgC=2URIhpFEKKOGWa05Wk~6@nBkqzpVePU4=r`
z$4)t6#fHk|^6?w&ow&mIup^#hGC%Omdd)mhCZ3Q_p*MQ)J}vTyBk03*_*Z$sFOd&2
zkMO~K!UOlQC-mn15Bg(%IF}}$W*p=<cOP@w{uAEbqWrz6QP+K2wrkHC1>`e-_`$#q
z-d~q(zGb1-A*Z*x^}ekOy|zC1?&gneSva9>%?qC#U9UWU$3Ln)II#O+&(|)XKl0cY
zV5ii>Kj`Q5(f>E<nr2%S;P2+!9=~bjjawG5JDv-HJnup7`7c&}e$DDF3-}GZ!y|gZ
z2k#B=o(cT$-ZT3Yc>VO-KRmf^;eX$AYudh6^S3tCDu9byol$$>nCj(m13i%kGCq*&
zyjKk`@bTu$H@w?-S>3`_^{?8qLA`o~-L_d<|E3+cF8`eae#AI;yJx^NO}cMeAg(aK
z*fDm)eiuInsgK{F5AVScr&>=wZ}R$!wkm&5kbMGqHt`1gEZ66j#eZvmQJn(*cm15X
zuRi~1&BB^}_FTQgebo#9^k%H7G`?zKpP8MXsQXm4@_i%k)%<N$y)pYvty%bdUdNA)
z>A!j5lX`DF{#pOJ<-D{!@xFD7wy9jcAEh1hjooawN0Ys)p1O4bd!*iJ5AAr{v>|oM
z<2Cb&pJNBa_1<URessIlTa@o>iI4p5M17A0J7hl$&%Cdw+70hl5^u0~e$T?Z(=YiA
z=QZSmJU1n8WPN0v;Q18mKKV3pmHiKSiJFh_ybV2=U)~=f9ucqcPwa&8VMoM6uH%RB
zM|=E-b(ncl<2QB>599&p#X7*g5WB^1Sce%Gah&(-h*RiId*m|?cqjkE51AM0F)#3g
zUaarv%X-K7;R!v75A0KsM}5|P@+antbqReK2kS5M%(-FplmBq?k&mt_|9uzhGxpEC
zVXySVeGvJ~FMb5y_zUyLIPh=Q9r9#&f;Z*``(Qq>8~UL>=S|E9>nl96p3o1-Jir_M
zFdq12Jd}yAtPji&^NC*Qfn8wN%s2X~?XYL$l7Ar=eUS^YkLLV=a|h0~xDRp;0J6Uf
zitJn2U$B2yk$e{383*=5JY?SCh52B=!npBs_RpLbfb3U^tN1T@A^n4#Z?MlMzT*Gv
zbEwb08{UYwAo{XTL|^vv=)pNuP(&Z(qbFtL&>!)ixCPP<{qRHPgZ(GtK@Rn?H}ruY
z^dKH_pX<nBzCpDf=NGg?Z;*3Ll}F}>b8z~@ZfK8R;MdHjikt_bAJ2{8nfS^672df{
z89j;btk*m*Vw{|xb6$kKfXIO#>>GXXL-;`s^GQD@L~qt(<`cVt5By(6^*uTK5`D_`
z{N;BCoGXI(73W5jIhR-C6Y-3A!1&;ad50Hx1Mz$8h5Fbh`ZEt8*BL+Z)p0W4_!;`a
zBk>A;(Hs4c&pgu)*FmnsBYbh6{<scbltJd3I1Vr9LmWU~t}{O7gZ{`%(GSE<_}xDJ
z;fKsS{HQ!)2k6Uu@qCkc#II<NJp76Aai4KfkMU6t{oxt^!XDt2IEftEv##M^TqkZ2
zPsz8C!}>^j@)X8_UhFH#_le{99rBSwKEwJ<Kjcm1IUwg7Dw1D=j7vr4#f0dG9icz=
zhaZtou@57Eg*TA;@Xm8{<iVqDgl`kFewk3^4Zl+nduAOYZt+|TB>w3Td&V!ZbL<WK
z#{X2L-f!@i_)ol4k$#dw;txn1(IIxHB7RKXu0!lwhvY3fWPWvM;_Wx{p?sG_^;@;f
ze#V44dnPWER}n|a*U3Xz_uvVB*(c%8+*kd8e2_RuKiE5Y2JZ*(oD{o<N8*rb@2n53
z8?3iFq<vWAJc{Qh%scN75a-l+f=`}<@_dx%+4wcjgRy_!XJP#KCp@Tr!FA?=`6C{q
z2X@E25^p&#V}7wK=9hC6=8^L-^y0k^)(PSrdf`Xx`#A@qo{HEB^Qt2K>kzwA<F#65
z-jL6^4f9Bz!8oxi)-&vwddM+hSo>qWL=W{`f$wFozXI_i?27jW(3kfY-s<?|p?myu
zdHL_Jk<af}IM?7hW%MDRL>}_V6V)=;d7pv&S4I4T_V@|oQvHbWVTZieL4Vi<`ZJIC
zGxNeaN|`vr`xV$L@ss$4f6+hZ8~8ozKK6}&5Jy0+<8RDAeoQ{Y`-a4G>=PvJ@O&4<
z|9P*6eu)RfamJ57pg-ef93bc7>^o@Bee`7>nP1*hVIIf})qZ(z2zzH9u|MpSdB$$A
zEBpYvfhWd8nYhb5GXLBsju3ws7xj@(98lK{c!C$&F^?epQ6Ia2FZxAa)=%_AKjbri
z*d_JwCy?>7AI9$KAHDDs<_%t$f94V7Je>0w=9hJc`}h@hj(qfiH^#yIu?~UMXJ1U2
z_V7X(gm2~#zkvtZ5nqUdtOxMNevonC$HX1@f=Bf_^NilyhcA$MAufUB0q~E0%rkMH
zxQ+erJH}g|e(Ck=_y3{%eP6x@h+lDjfIspcu-YCz_ZZe`Sm!I&_$tC5i2SsX{Tud9
z{z6<Pui-fY^MSo$Pn59-=8N|PS)a)JIG@DMh?BhM2+!CbevIAW56mw}d}ln!2Z<-d
zAMBF(V?80B5qGdF&W%{_m_NqH`<BEZz8}eWcrO!PK<t_K&f()PC-mypWqHLBFTSvH
zbi17@j2M1j?TNeGUws6Ag8sCJC&thI5qYfN_#5-5`YCdegFUcr!Y5_)p`MDw8R7)@
zS?|%4{T$E3cwVFC`8-eHxuRO;{2k=^9`(7-xfK0!A7p<3GXLyD(2H{|71?)y>V370
z9-z9vM_-V7*gc59>JYie;dv&{KY9O(b3UFI5^vd8FhA&_BJl*i=!Y`>gS2Pe1hH4*
zC;1TmhX3P-<X`v$`3mh>=g6aIk6iqP_(R+w4$ux{ohH5!H|P(&Xb+OVARj$h&+sdd
z`~kbG{>yWB&LNmTo?Bzb#Alwn!#{q(^Kj+|`OF`PU*bQ!|A2jPUXFd>m#j0y4dN^7
z1o1)5W8g#OlX+!a#6OVxAaR<0K+YMc$Gk9a#7C}E#xAI*+9&eRo4f-%Ay4N%<D?z@
z&@bx;c0_;VIVy5a&N&wE)v)gJz6a|$eg`uDDsrwtJ>C~&enIxV{C<!9I_oC;SjL0D
zF<$s#TwF&kaTq!93opz!`zidJJdkymyzcP(D&5nu!Wy6Z%oFyCe&~x{=t(|;J)kf9
zGRo+Of3qHvmtY6PC)R81j=Z1#$YZd3;t%UG{zZG%b?#$ttpBXb*ahn{ai06+`S_6z
z;S0X$4`e)yAD&rvh>z4`K8U;QAK;yO<Z<l(Sr6Ffkay_xHeJWg86We7-4GwJ5As3m
zmvsg^<s1t8V_hR3X5UYlc);&#Y0rKIIn=}duy^ibpX8zHbyW}i7rn7p#)rK!-^f#W
zKo0(kJoG^xey$?(&-n<5zW4|03I4}?;V<|x`obgYF@A$y%p3Mg{6=5;0cnrj;{V8p
zFX91y$$HQFh96U(_Q=KWv1{TANFD_*@Q+_02mhhWyul;mg+EZO&$u~<!|zysm~Z5<
z&p;0A%QXkwdiKBG+vu~8V!micyg(lMa=yZTp6l=ca-I3bo|zBgpALDB4YKb?E~wfs
z@rv<oKKuQzF8F$l&wQZ|<HGJ}$9y24I7u0VC+vuE6W@vBD$);jMqI;g@N4`K{U{T!
zh(E+7ko)jLo&a8b*q;w>zwPSsdFMLy(Tj5!t`i49{2l%1SLF}8LasV~)&uxt-mriA
zqa8fZFGxS=fxVzVdSTbdp^SVGJ!pr1AnoV}K0wwj^y9h?S@*&s`vT$s`xyM3^LcgM
zCSPG4C;y-w_lb|#H@pysh%dxt{0cuIE>oZRP?7$qN1Q+};y3Y!yo&Q+5Ix!Vl4rBO
zCf|ZLt`pxG59<MC;s<)sKmDOUNPGOBb1;x~j&%XK%p-XkaULG1hg^6;ALQacAa+d|
zp3xWoWW4B4ef&$ESL!kU$OExE`eQv+?Fcy_enx(!L-H&9kba1d=$|yAKY0&-2Z!~R
zxOUNJb?4ml^bbDsMmzW>&cHkQJnIttGH>`B^GP0$ow44~5AEQMxJAFzLqFmx{K7wR
zj&|ffl!*%<`8w+!<0W52U)F2x6F2Ce^^$!m`qB@25qG)Ib?k!i<0q84&-l@seF*bF
z|K#2D!+hW$*ggCpmwM!p@JfHo6ZO##L@x7=-o)?n`@j4h9{iB_L0n-yCQk-APvzW;
z-=mQypa=1gxXO7I@d_T`8NSe0Mf68c;sf?TnekC3UXqV-e#iM5=Xk6GoZE2DOT6bh
z7yQ0Tt*>6EKgLgfTm6^wM8?CpH|IXQH_CY+=S%R8{**aiq#ftZ^b6mdM{@oMFXRb~
zm-^(h$RTg%I}^+safSNW4fcmU!H??a*a_<$>pJ}*pL~cog8vgID6=0SzH&ZAd|-c1
zJf=N<#(tmYMED)`uv7ZuK6#`GiC6Rk;t!NnWZffv>5%>z5Bd=Y*+&!qi8sH&U%q2W
z-hq8Ezr;J{hj_%gz;*Iw&JD;1h!ey|;voJ^JI2BL0N5?}nFo;d5&HzOL*gg?46n>D
zymF3%y~7XdBJm0NtiQ}7>ofY}XPk$#Zm_<gANd>g$=mS@;u<`{AN}GVjFWlbd=}nl
z&p9}9>4)bE)aM)&ga`Vk9_{HLyI@?jC;ky{@C){@*e!9BaS=b5Z|sKgGY<IWJOq81
z7uHwWfy_Va9dV5LXPnqSagXs4$A~l75Be~E{EzX&8|xeS4*a1%cDM4dqkHf0bhQ!a
zh5dl=#Xf_55dO@#@E81%cubl8$*YMkv<Kmhe3m%C{NN`bc^CErGH#H#uOhse5Z*xe
zGa>$ALgFE)>WMvq%rD4(Q-?fv<UaF79OOM))^*|r_Jmzx-`FqfBKE@jx$s24j30T#
zBk~FSi1CoGfaFi`L);+lun)%`h=<gpJ$8nF(H=gqKaltWs`HJVgZKxC|A52;6;(fB
z9O%z@;2nABjX#iwtLrmy0zaZ1NIYbHM-KHt_9f(1th2-=_@RIF#V;8L^FVvz19Cyt
z52&Zc2l!Hv{RH_O>oj&t8F}P)#3%A6)^X&}59=XvI1eQcV11zs(w_B+e#r~TE4hw-
zCWJ5aC9a3YU-CBW2K&NJ@o%o<-}oo~f}PWj_{IK;>$C&$C*ls*LGI(%_&@hS`cunX
zC+=hK*d_OgSMY}&<2Tql@sE0(gR)K{4?Vcf_{+!1-%0(2fARw2BJo${8#!D@9(o~%
z`9L4$i8A|Z^aQah)>Gz>ctBoG{J?)HWB1g@{&k2xVUc(a59o;>x<dQ`#4m_foCk0p
zy|5?r#r~KVo~PjV*cW!EL*`5HX9loicm?qrcvX>kV*kUs%RIq9c1nBv4dlLxI{P<W
zC;ougnF$#;NPbN_@(ui)@!-GgJE+e-2p-7;Xh*z49()m>;FCB(JvA@DPgsxeWBdXh
zST9*W$>W$uknymuMsMyDH;I?56RbDn&E(TuXJ5|x#Qfvm$RnOpW<Ka2ez;G4c!fvo
zgglG93%-~q<^j2kk8?TXGfvKFSf>~d?I^<o<3<kk*her9{E#woXb+$8hFu|#GJZro
zc!LM*3HxGwhd+3s%>ILVoWro5l;;7ztjpXdo)aI43)mC$fqvwB*dO+a9g_cH*R;n^
z@sFs8pCFI?n0ygrpUyc7*U^h}QO;Yqj$ZgL{IVX>Kl6tj;HRv2%s2iHuh<Lz0>Urn
zJml~A0sY`N*cpC{eDnmVkKKXX$KJ3r{EYn)c@gz!hu-8#@I`;<4PWqpol?e*u{Xv6
z60eDu?0cA3<_%QmnfRh2cC5blgIxAQ+~@DMV=weWefCx80Y5>J`K3PmVHe1SXBDwO
z<_EilH|!OZyvi*D_yOw$ah`aK-Q#!Er_6o)k@Xk5U>!sr{=&RtH^?O(5r0%U#0Bh#
z_bou~vp%99a)?*->k8=?Wc*Q4^&{dP{tX{oCl2C|tkd`%$U0&|&hd!n@D0!8C&YQ`
z(T?*+-cw?q#`zuJIpn)CJg>zb*)P)%`4q_cC(nmL@+#sLNFIcK*gtt0_Q$$_y}~PH
z?1y<KZs89g^Ui(j9lp5Ec(7M^Q|DhTo1z4M0iXDxsuz0T$CSy>7&q4$H*%OC9n!Cg
z@E;b@GbnO?rDE8A)%K<`>t<Bscc=K34vC*(k@p`=$UN&1d)6WTs3Lx68%=&e-k~CX
zYeMpS;xF-*eG2&;ahd%G@s~J2p2_n8)xXJKsZSnCndh7MDRB-zWL}Beymwh14}N*S
z2*1Y9nGft1yW%==68|f2|I7Z3=TF>6FXW*g`(ThbNdADHv?HE@#1s4wf2JK}<RJ(C
zXit6CL-J<$!0#A8<6-{|AMi~-tRpJ2eo_y+Ag-$R%XQ`fy;(1?8^)y~_pwLz7vu-*
zr|5_E4*SM_=?^=kzB)d&%)Uw;SK4KG(IGs<MdnLI>^Lf7uR7G}jeX#sCS*Rz|B=f+
z6jb9WaS$X<>X5h!s`5-_+J&_%)A(K8Cogt|%rnnT@z<cpyyG7{U!sg0;ulE2<O`f*
z6E|6}h$ozj^ZZ)v7yn?rBc3Cl_{}<meAYYSEqOn2ntX)z#CMSQ**Kr%xg+}+>d_zl
ztLp;oRYX4D3*kK=#>F_vtB?<}j$-eeH*&tnc^~gLseZ_L74_IBbDg-4UvVGgJXMVc
z+*i+Gk%J!Gr#|Zva@ilUzO(-E{WiYqz`SxEM}6`M+F^h234fexs=q(s_xxO^J@4It
z<S*=3d0&|KYgD9P?2|HjQYL=^@fYlsage7Fr?7Y83i|Qg1Ny<Q*v}%5{z2x8^K_p7
zkiW9NlTQ#IxK3P9;{f>&{zCuc7vv?#BOk{8@khqZa|7xzE|7Nk7xM|f@JRe(y{0}$
zef~}f^8xRShxkHVgGcO~I7d6?gLpuEz`i-ZfG6ySaq}J}`Vep6oq8bm;eqiIcbGT&
zC;l_f=!tz17ttF%iIddhd;>q_Jcu$#-pzZ!w5K29IPr&l6!8Onu_x?;xTfZ}tY`Qc
z^Fn?@oMIlyV~A_;!ak7uv<Hb($iogm@+|C@{V4Zohdj;$IsXUo3-Uv?J#vV{tQ+_n
z<DxzO!FmJoe1Z4z`JUJgM<0C7z14p7`F<1k`MZ+rv-s`=a>(<TA9#gN#)o|BF>fIL
z#JCv;_QLqo^$}jVk9-h)m?z}H8-7GR?xPp;Nt}UK>Vfb8(w{1q>mYi7#A$c{@o)5D
ze3WTVf7l25aUbOGMx%#1KIVh#*f)B>1A1_ucHAev5w}>+utyL(A-{n~p06?<t~2lW
zH}L?zRAfD&9%bgAGV4a0mdE{d-fI;{FfQsb9^zoR-&UFT?}y(0z(Wr8IBy`XvQAT0
z5q?;2iSLw&v&aEyN8BZ!VqF7iPd(x?_tA&-fO?#7(jR#kJdjTzpL~Yjg&~LM25NhF
zWIZ8Y<M&zoehcJxS>%1>ZS;#C=*M-CIE+82@fAP9jyaE^-TZ%Sc(woX{~s@3+oRsh
zniI>v(~Er)|9I|;{o*gKkbZ+A^NgKhN6ZI&QIB<&de{x~hMjVbspePJfBC&0c1`>u
z&amFn5B|-3P$r%*f9S_LK-^%y=;wvySI;|XuT?(rhIvOW_0{<yPSPK7m34~vKtISs
z4!={zK3Gq=k3PsJzEPiXQN|Cc2U3r5&7V4<(yBWukKj7-l{iiR@X0tChl<2&ka03U
z<ZFzF_rl?e@zIX!#AV_(_0SLdfM@D)9Yl|)h(Ckqg}&%To=TZ|l+h1FKkn<0^;Jdu
zh4IrKKWE-puh18Mkw;lY#=|(!n|W4|^AGZ8;w1gE4uYJ!(2o7c|6}jogZwJXJAm(Q
zHk-}OXaN}-XhT6os3=e_0_`3sjImaOT0>=|;Sx!!DH>pEB(Wk|AcTtnQ^jGF5^iB=
zv|ymNmHly(mMB3Y<)To6A~z*QluKfg{&POZnXmKC`Ryikn95(@Ip@6Z^FH_IdEayP
zxAtf9hJBH1c_!<UeU3a2)id#dj{JaLB6{{|B}_fImam|7(Q!>b(EO1ieT!TBMEcf$
z7{~tXz@PXzbYFf*I>xCF<TVaI`Y3ufj=jydp84#=j-EsMWMAt*{(}#Cb{=WIaq5kI
zuzIJinGeOE=k%(6(l>jHr+n@_#CMjS)5~C8C)H8<bFLsByuYIleIGr_pW+$6@(2Fd
z8yZg@_vyhp`a*W57yOY6Kjd>x=bX!U{hD>mgZ?hhy7b~XKY}BTt5?SV|JH!MDC?s>
zN1yNS-Hg|FskizYsBd&WA-=>P`p&(q<D6FC<h;>5<IGp*-AB)J=LhIHmlePAqVdif
zJ=YhAe{xti`IV2n7lQJ!dLhrCM?UxEC4HN|8s<28*gD3cE1&4Q&?Sd{SUn=IIwSvB
z*Lv>LBYE71{EOdMH<5nyU--Z``t;+u`!MN}gFfsJ_$mI}H{W>raz0BA^NizX^p6g|
zgziK8f;MyxOh5GHJY0N=L;D+Y+GmMda^P2;U{C&ELh~Ryc@KB?GiUcd{H>mjb5-YT
z&H<dOIuGzWiufZBdy7+Y+rNre$c~WzLw1Ag2UA{t2F)uWy~-!{1?Y+k==~=oH}v~E
zkUv8Dfb0SJ0~GI&+z|gz+(YX^aR=pxM13GXAwRy^*>n0r7m6F>@j+gEk_SC#p7o*m
z=(C4&f4{roJtF-XpZuve_96Zsusq@Yn)&!(cl<YW>i>%<&$Exb!r$4MzgkCp$cOex
z`c&&f^~`$s)Mv^U>}s4onZEVC)`iwdv<^M!`{_6BO<wcpf&A#$7wgaEIemEJhxPwE
zB%g8QK!+UWk=wbhd_zBp=HXL*qYw1;^Uf6$b*c6t-J&?*pU`=937uEiAC{2)A^9Nv
zvK#xl7N_iMUgj10kuR*rUM1uQ;>>%0C|>A^z1W96`Hywj-T4B&T93VrS0}`eb0PY%
zpCuRhJ%{|#I_`T9?RWL?jV?R$H}}cmy|?dHypJFk)bFe3(D`%EM=xFcxrLjW|F5J!
z@jle|UdHQN)Omi;^ao$rWuIeqSv@Ro%g6dOb`}4w`Io$EpTWP_U!HMa9yMPc=cnxI
zeF%G*Z~fdil==ww#g{xS{=G+HZ+##;@+Zh|_%A!NAAd0(8VBV=^Z2p4zz?DGByoa{
z{RV%8=;Bu#%CF+l{TyfC<vF|P!`Ov<;?H>Ycdbs+Ptr$6UPNE~vyXE&d5%BPoA1W#
zx7Bxfhd=NieTaTre}aE~fw<&<>Z7<w|Hq%_5PkF60bTR>J2}J)JFu_$;tZ;1_!r-q
z7wMf{_>cLn`3L`VP4DdMT$KLE;qR@TCx{>NvJd{nDgUvaItA5f|G&rYZTj6)|Mx}S
zKIYy-Hk`9=(-1##?&p24ya445?}wc;%HQUDkL<bs-~Q6_)r)R>a#D5U<U5{w=j2ya
z?hn@IvCdna8;fV}znxFwm%R9s2b^z<4|z-<#lG^1cv82}b6!e+?yFzcGhe*&19?op
zX`DDFpL1$?Lq3$B&@+#}@LzswJ^nyG=ssj0e=q3o2>CnYmu;v{IZtry{j7bva|0+Z
z8E2oQZu=fey|>So@5GV#H_vsVK4R7IagSbl(&k~fbjs`Yzw)Rp!`73>^b6#<{I=IU
z^uW<Q9dz^$`U`m}<6b=DPrpKb{YKK0AMr(x=;&jt@7i<g;5WzFk6Isn>*jd;Sw|g_
zm*rP_&`0y194FtC&vz~K0`aH*SciPhgM3G)Uhpe&k`J=0ar_HE@{v9e|LP?>i&uK_
z9?ChJ{<q2RcdU8;{OxZa;!ou99U{G;Pafms1$t%|`9L2@f9z>o`h&U*>BBWY;Kxwk
ziw}I7ub-edb&%fGGyJiO|8vQYt*>8_C&{bcs+Z33*?}KgUp*y<x}e_Bzwz`=4*3zP
z7y4)@|9YNv%QaMo)Khg=y+Kc$n8&}=W9av{)qQ-Zljx%h>46;T1-$1&Hy?0$HNKJ`
zoj21Hz0x~>q*r|{JGdqXI?z0J<^S&UKXGV1^ZA+jsct~_hw3SGe``_vC-S4*Z;4lN
zKwfb<Qq(VtH~CwBl`q&s{gj8q9lhaC9v1)dr+g?6@Dta@laD{igYuMl#<`Dg`N8kN
z@H=^%-;mdH{kVOQI%Pl3kI^wse?VXM7wDUp`oSlCd2T+w)H&}l=uuv!2lxG74t<sH
z{?W0n`RXBkse9~^{Nan=pl6)^>E_q|_2CN-?(e8;-p|kn{m>_Wb|2C|d&tB1P<QBG
zUiO^)#)%W>Lg=~A4z9_`PuYR}>D_q#X*_za$%S9@<R|?EeX<j}p7VF>ySARVF%Q44
z@nJl=o~Ix1Gk#=Ueuhuu&?A@e;*}narw4L7ze8U>KnJo<%16KSU|q;>@$Ej;KjO#w
z)}sgX`LX;$9(G}WNdNpa$Ma9}<J0=qvv2WSyxBixyt%LMqDOWjr#Mn4JV!6<sye3r
zs^|JN@uz;sx9sG(IF}F5v#$M~`SPzg#us|->%T{e=;^zB-=_ZQ1DpfdpF1bgS9z{q
zK{x5pqyEKn{Vk*y=LgOe^bO{z-y=nS#y{oJjB9-8tMmi*OU^U+lllCg|G6)I_>J}T
zYu0nVpf4jo{g6M;f9wmKADid<ROfcii^yYN;haQY?VM|{KEJX)`O&qW^C5h&AAg}|
zdXd-YB~c!eXV{q^xi1d*opJQ(exmah{@7ss`zGRvo$MFfr#H`?o5>IMEs6TzMEU3y
ztZv!kpzE8zhk8XHzuo$fAH9!~U-i%GseZ<L7x_W|A-|}f;#ZxMr=as^&lA-nb<eeP
zbaC%o-unvUtmC}ewR%Htd5j$DLX)rj-hGYEx$$Lv^qrT;&(1&cyhPloH}bzeQa*Q{
zrQe16E&Zgv%y@Lz4XUT|KYHw}uc8m*<l&|7+4!lxr~5nhW%dE)sXx}ym$~LQ_V=#!
zp%5K*LLVJS&+LUybxi+9&-zj466zLuuF=s)<1_0@UemKUp;vt9ztP7hzr=_2^i%kD
z-!()Z>QBV8arQ^9J!fa@@kf5H56Jb|CDAzYSclyD2=;Z&?(8WJ=+QX+J2~`$_!JlN
znEUjv&k+CU&>z0+Yu%?G@dx!=_~&2ps`DB4=2ym>PmlCQZuv!Ds*goqU!?D(C;J}n
z3E5Ttm8Z$;IlH@W9r0$qeT#YfT)gMibKbUL$ag-@i=9XNyI|MyIFu)y<Gy&qS+h@j
zYk%eMjs0GT{gH9{YWZ6~sa~A--YKv8hsgsKw7&PkuAQF`Ox>{dl9_|m!{2-Ig83`@
zD|uc&j=tZi*;>TM%w=m%IRAx#>eeyu`1-=##y0Ddiyznr(674VIemHVTwa}JM{?1>
z`sulK@k4*;(gSSbss6irblAsxdi!GZU9*=ui*NGclb_HV^nL+<_F1mI_t>xJKl^$g
zdrbv>&&&_#kKDdL<sbGV^eLZ^C-;N)bN2J%M}Ok)vGoP=M_#k9`XV2Qdv>s2v>%mE
zttVfww{_5$f6RmGBRSNq%op_2)ZN-2<WXoneuyqRqstH2LtmBX|HMJ`<Ppd(FFf;O
zSDrC;(-35L`6c5Dzxd+U^onot8xMVV4gJ4Z<Dod<&;D;Nl$ZFo?_>BkyVJisiBEp(
zntjEgdZ~{PfAlR5*bVB#`Jr*vp<jH{3zT2g6R3WPPkO)VtAp=;u<EVYPyAVz-&kLs
z@SMK&sn9z7#QRU{k{4ayHJRr*KO(pL`0_hU?4_Q{H}o!F7_T1lH~z@q<r9C$>H96;
zBl&)cKZ$qeZuE`*#>bxB_>p}!Hg#Pd=f{xUpE>*klfJZSuwoDM`3GEi(X`{QS~<R{
zuj(ItvKKz&ef|sSU4NPRj33*t@NXy|>09{$dh(n&(U;OU6mR-O^-w>CzCO>o^1XeR
z`T9kDqJ5qIQQxkwC69ia9Q;`P@)vPOF8;=D^ynJWr~c5jc%={P(nH3JYkcr4c4l{R
z%D?dMx%<Y6JJ+5=eQ2V*AYRdNEnfKD^h1w(e9l9C6@K|EJzAf?)4zU^9P+(9jb6%!
zfBJ&vS=Y7a*1@0rLoRgjo9l`%&)Gfw!8pkN)`Rq?-k^g|eU16<L+9k^=wrmUeYp2V
z@|!po7y23V?E}SwICNf-^ZB)P#lJYVu50n`zW8<@+Ru_hz7X&D^4`w5iRbbI{^T+8
ziBtB(kMXX_i+^@c%yW>oYk5V!!XNuW{OVWP%RDH*dTzdW;79CaUrmqlC4J-1ee%;Q
zdhCTRzrh#%@i%g~re}WRIpn{tM~e8zuXX8*9`I{i37zk<3qNrF>;0<p4tld+aZPW|
zXYfHj&z<w=i|n(+zw-@#OaIO_^)dLhFG5G$L-ELe@CC(9qI?a#?;9z~XC)LT-6Fe=
z7_WVB^=D4sG^9S1ki6=o{LU`Pr+lvu7N4%wQ~%FKf51NKnR-Dk^$I=x2s`pq&)r9_
zglT8rgF=1*`A?#}AwT>6!*?C}LUhE1yuzQ2YeW6Be#U-SzhfUk5Aq#7n9mRG3&`!7
zz4fp7Rj>GmxRY1d2e#FVm+N!>I<M#3SFYW&uX0`h^+C>`^kw{-e%y!pH2osG@nil!
zTGa39L-8qp%8&M$^1FEEr~HE)kpGGwbqJ~h{05>ce#Eu=_>}Lh@40#k^#kf1eW}al
z^Lyj<d;HG+3epEe&wP2wef)YZe&~muU8}R=QXe3G)I;a@=&7H^IfryUNYDBJaf!a~
zXYdR0=e+}Z&iS2Bm~VahPGm3R)N|(`__5D6o?qz;=!rkEhk59@uP(6<eeid3>0>;{
zpZMTUke~dfP#uwHoe$6heTi4{upfQUCw-Bhe#EPO4825t&cFFB|Kiv9pf}^37w|Ln
zUcTe!=*th#IP}B;d-G%R^HcNXSNTv~6leHJRKNJQK1khFAB{`Ix4iGXSAFnYzLuvT
zdY)TXpGD93fZ`4vdDe6DohPHOPje2E$d2qyF7)vW*$Z9u)cWGjdqwg={=lx{lit`*
zp4T_w$9Wom!M|&IVt4EDCvm1O7|(y48{6NhkMyh`)nDVEzUdKN{@UoT{v8(l7-xO+
z*@3?4gZ+)C7k_tZ9($0(`7}L?fAiUmocb8+<CDJWm%qx}^kM(zoD5z1BqzDd!#Dq;
zXZnNW5_iVy<HRNYotw~`@eS60uS`Dkz4wqWtcR~={*SNx?`dcLY{<2B_1DfRoL~6v
z+qLf>_^bSFzrkPB3;hz5AJtd&z;k(B{ZNPW73f2GSpG+!T>OXK^fT#S>NNilf9jci
z693{)(0cfX^rWse<LmfR|Jjqj!sOF6z2h(Sv*7e&uN#;(w$q0C3V(N#=Y8tG?^XB*
zedrf_$8Wqo*WXo`XTQpyTzl@mb4uqu&UuWpj=T%`3v{17-ACW|G3Xl4AC1QczoZZO
zIiIulE$3YI`ukS(H1DC<PkhKf?8jf+w?1^g85iF}@N?@lbnD-H5)Yos6Y2uL%lxdK
zWM1R<{8#=o9$E*=pYoNutM22+d(n9d<{h-?%z?`L4)xFbMtRP2{Ss6spnSio@4z>0
z*fOTWf9V(c9?^6D?(cErNA;e+`@3)7#fb;w{XLuK=v&u1;-d}ulRWNPp5XudnxEsF
z|L3~m3_bio`I~>Di=I4z4*ujB@m`XT9`iYV>?5pa9qW3IUv$_3AI5u*o;b$0>y(>b
zp}140AimQs?4!QYC%)JTeen;~75o`TzcA|se#9v~@dHS&^g@1m!Jl>TZ#=)?=e{dO
zH&I<mv>tu2FL}w2KjYYgeO=?%eR1jk!LXBS>!;tipYqZRR6p=-{k&G6+~)^S-oT%}
znEgs9?~fMM1<3E9`rvyT^&tCc`yi;V5+~{kzw{l5b?pb((f-AGsJf!gsB83vU+d`e
z_3Q4dr|On<^yl~yM@gUk_$5EnFR&-S$#Kr}`3b(n7d!HQb|;_viOyr>4|L4K2YU2C
zKJ$z-5AsL)a^E$2_B-rop6B`%{G($WyYU-z%roBpt%NBLKCP4Vj7Ohd<V*I!4>Zm-
zdqVNh_+h<INIvYZ$s<nqeWG(z=UvXj?3bYZ6#i1L<i>CAU+`}}|6ot^-8Y_nJ;y)4
zq&Mh3ySTOvq&IT7hQ{H8J=G8VSclx~V_kM|ADx_+*XZ$Mh+YZNf%t^vg7|^thwPCk
zzxi%P|156hVd#9v`Ix-x`zxp~_k87!3-8{1?AVID_7my?e{kRan%%sQht@Zq{~K=|
z^x4rm8Bgqr9u$Y_9(&Sr)(QI)?<2@>o_ZnQ>tFc2yuvQ*ZM;5<-Q8E8_={_C3em?0
zdftoZAMKlsr$6;V|DvB&$MCBTqGLa0zICB><yY~EFZ*$F*_RuSuJ7fIZ$o(lpW=gl
z`Gsra`2#=XZ<$}kpZ6q*^o1^eMGt*;<#)!Lk1l@Y8RO8!5B>5Bbm-AMercRMPEY8V
z=NcLh%|n-exew7}Z~F)S0?ii(;>mc=jYrot{}B)PLWkVwKy+N2hmPm?x1MYBpz(?3
zxlfPyb05+l`?)sWdia3Wac!LW<a5nm=tn<=Km0)S=-YbcLw+vbh&$`Evvu$T%@+s!
z9ewN42fF-~T;dE}`XrBW<TBrS_%F?;XLRsuU2=Mku5skF4x}&FQ2mhy@>;$zKI!Ab
zeTc6Tnn!Oi_0JBFK3%h)dGu<WYluI7>Av~=09psVl!qMnV0Y&N=0odw?wVcg2c3J1
z3-|HGPt0@Qxrz50{L}hSKJxwqAKrWN1Nu_G+}CgEWAs=2$bEg1zRK^`IQR8;C7%2K
zL!aq)*?qs{?>_Wv`a%6E`ub9Tr{Z^d(e-yyzFX6uvM+l=`zPlC&Ns|={y-k@GxdA)
zrk_@a?Z3oV>Q5hOJiR+-p=W)rJi(6k`{+V`L67X@yoR2PXCHQ^Z+wvtU2#D!dgV|2
zlYIEq*N8`X3(9l&p>O@2_W|A`Kzf4Kh2((NgZPE?4bg$tf&3BTAF?NucgYLoA^Agp
zDxUQ>`fm2}-c_EJr^v_8$*+%x`ek;pzH6wzm!GUBKe=z5`{vUxImstq%UAr`dtK;0
zx#-__YxKxp<uz!&JZfF~@f?abc@?@(&+H>EAUfpl7R|?B#+`E>aYdi{Bl3txbkL<I
z<IJNs&&iu;9RK52iGE+iduZ={mfk;e-7y#SG`|bt_ayy3r1N^`>->oStBd^4`#Ik$
z@I!iICv}kDxpsap9?%g7=Htga_OuQ@=}%rX-+0J>`7?W&@3}ngd_Z22zahJr?^>Q_
zFY8#>bNoW%_z$_o6Tjd`d7X6lA3D(b<gl*!;?1@3(0KJ%o`&@38mf!*Ykv>%LyxZY
z2kNA{>KZ+0eTW{vp<ntTFFwt4Z9Zgg{$ZSX5I^d_I)sk%FLI*8Z|L8(e#!a`-|L?`
z@cG?uA4V5F`bO7!=-?k;5dUz8GmiY1&wgXqDv{mriN1C54fzkW9z?GV*)dVy=sR?B
zs|(^r99Wk==rxfZ<N@dP<lsmA%6)Y8DZZER|2cfus(*K$tDp9LhVR_`{|@KF^v&-4
zfjz{Bar_X9Bl`09=+>RN$D8kY{L6in?@ZnIy(j-Bm%h<=alWg=w|#?k2kYNgqz}*W
zed^afd+q+Kem-O$;@sDI_BZx3=Cc#}klfZwz0(8w{?5j}!~O*Sklnl|<qz!Q`x}12
zPsxdn?{>tc@%ZGw^lm;q$jAH@lA9f&b)a}?L;l6zApX=PXnbP!LG;xv>MO-LBzFlj
zzRByp{OP*kr+$y9pU`LMlk`DQpU95p$v=?(<PZIleX=~KU)P7}gP}ZUedrwD{@8lt
zl+W<PPTBXGhYx)Ty5{A5`ePsbWM9c2tb-qZrGKJVh#z`Fhu(c3;CmDHqG$Sq<ksil
zhd-mMuF0>EUpQxS?gZJ(dgOD>PspJ!hWcLD-VdP<>5E>-VgE@U{OHTw_g)Bke}q5l
z>)+{>-`O|kHU5o9haBuqAMC_mVDjz0bzM9EBe(U<w-2$dYj$Bz@r5t<>6ibDW9!hH
zYx;ov(>~01j`lz7>3chJLGeU?<aeJRd4I)z;>LG0<i{`cJvV)ko89Px-V)_$h#&Un
zZ%};Ex44o=*oD6FjW7JTk01P!pI#yRpyNJ0+i!dBd$uF*+O>De+Xp&VeEsAtZx}be
zbMpB8{^Bj`$9MXc|NExuj)@)pu=)5?zx7K{|3Xgs(4RP8Rkup$?~3(1&hvaHpf9kV
zeU@>4cg^2L=&$rg&V%)h&Xe`4eh*Clq|eYtndjV(o{ZDC>2L6-|MYud`cnP0Ykil$
z8}s``?jO2p|MQp39as7LVE^}qoaoWJ-)HiF_|UN)edu@T4W^ysXY#5$&h6w~_r;(4
z{L;CU-{rG^^7r8CyK($Ve7SEO`AeTF&g41g=I)zE4s|B$r}6p#exje%k3fAfL|>e$
z|MCgG#fv^%o;6PV$`A4Z{o1$5JI15KZsOTE@>vIe=;{~fMf~MF_K_zN`H^#I_LUd(
zt@J9c<WF@TU*ZoR^e8{m6S?tEzs{fKSADnNUo{>bdf`v_qaXHG56F{<ZtlCRLynQ6
zx~YFthtGcI?EZ(p)zjH|@vB$-<g&hwzGT)LrXF$P!rqSm+d&WN^>g1`_LEbO8q-<!
zCwI*~<1780A3gNlzUR*!(>d_O6`y`;ufERBOD5g7+eL$&1;6MWT5;pA>c?@r%>J9}
z_nOeul|^gsdE*~09N#(e$ZuYJ<7E>%5B>06e|+a@leQU>&$+R4Wa!+OoW4JI?Yne+
zw)1FuWOsE{Jh2yl(ht$Q_>}Jwov-k3ap^sU_~2*uRs2o95C`@z@|yADQr<*Y9O|R^
zseX(9@pt*cdlr0qUuRu;6o2|`{jBxLfq!(J2gn!BXIvZSy)S;f2UKtIL4Q!3vY&DC
zw>&_9_$0qLVNc_YBM*JhyFQ+u@JoIt?)?51f1`hSPkxiH<#~RHPkBgw)SnxVULv{F
zt9&jVAU>hIPfw6PLHQSoD@Yz_J(zsCrWgGIzlQWePVph`^fCN`ALw7ii++VX`WStJ
z^^M0j`r<<z@eg&*IQ;`Z<6q{v&rj49`Gnm3lbre?e26P?r2nM{dWGy#Li5xcazW=W
z=8I!UU-U0u(X0F_zW6zNkW0Ol=f#0Iw7xhLfBe>d$=}`ZbACe~>MuLAp>tt=Bi_U}
ze<qJU$v#AX@4oK_oa_3#8NYj_FSefVYWS1=mi?1`!`pA_pFd^aan0Xf`g_cuK63n)
zegA65=Dzb;`w-{bf41-P2{$d8Sp1!I`#pIX|Mp4F(Vc7CuUXIE**b@|&+)qt_BZx3
z=-3CbgYQJ_+w5!6wXXTz6ZoBA`%mi|XJ191&^Yf4=$+pD&YShoBbWKc<JbPrb9!QT
zeq*1DKj-z{)9^F)aDT9V|G_S<$zgxYuHMJczwy?!FZR8Z{G=Y^Ltaw%$<43zQ~C$<
z>Bat^o#>N%`U`$&oO-ISu%7%6ofDPNIhlH(4$IH_M(1GqIdwrkl5fZ_|LL>jJNe3e
z{e`?N|H)(WmA=<H>ZN(c<41n?|GDH>{k1;a-&4QwWB>Do51-W6>Am5U2iET}rh`v;
z)_v!K@@>kYe#`Ud<CmV*HGU)Ct3ULto~o1jL;b3HYFxKyo_fm9tb<Q>WIy$vUDQK$
zj$h(Ke5wc5@m!ydA9=r7ck|rY=RWc1ka^-q{iQE`13LJ?7k%o3=v)8IUi?Y_M<32H
z@NFK{@31qy$lvmScsCE?mw)3E|M*iEpmq4a@#M3vxTZ&PSdZS&abF%FpKJV~BY(>m
z_EqA<IQ;|v(MQlb{pq8~hkt&EKZt+x^{3{`x5hz!i9SBX89gSy^kiRPK9nb1%b)aN
z9y;XEXBlTcy~#J`8}B(iLUiqa{9O~e{G8tSk9pS72hzXs)=52?Cm)+{UGuF&zv$76
zb24#nerX)K{0Ck9iD%ED`hs8fr#E@swQ=+=KIq4K_!l?yU>^C6hveX=?xW+{IPaCj
z6FphSy8MNIna>aKLvQSWj%)9MjHj<0M_<;ZNBmibKE(|>`T_H;hY$NNa>zgClbc=e
zi7)H36Mb8c-&xOn{JYN{>K3}z#gFlME$;A}@q<1@-+1z)Yh8M!H*tf1^TmPZ;sIUb
zoHtkpozyG+=5za5=bfJG*IeV<d9ZylKcxqJLH!o`@;QF>!Sa40d+YD*C&V}X@DqKu
zeoQ~5UZZEedM^)&=aC}$$)^wE7yMBDllR33J@R|}k;{HXK2hKC?LK{u7X6<s`&;`P
z`x^a=`}#@euk@|n>mTha(P2mZkN33v&AF()Mn7ghZQp7?Z2$YoIn{3CrVlj#zs$LY
z|JO72viHsS+2%2oeWrQnplg3>|7pDUdWRgfWW!V2kEveVd#@L!E*Y%Q@m`X@I$v?#
zk$P?VpnAV--{`$7y7ue#>G-f8H_!WE^U$#mgtITa{GAiO@vHw!XP)y8dUoIczvEZx
zjXb4}$TOLz#DO}ej<_yi#vy;;FXCUk@F)3J+{+8%9UbHOrS<q5BoDfY=0o`!dS40g
z1I>ft9^wn~8|XX&djAUfHRMmwz6i=E(0MVm|AO=h=@q8FjDzHY@-!p|#24gWQ2l}G
zcOt*lN65?knEjkr_`4Ey(|L{j?LPk&Cs~i3v*3sSs(<Q-@09GX=_~i?{DeIEFm*}&
z5ij~1^+P>&{^6WjK9~2^RdI!$a|rR`JjHpR`RpY>lG`}>kzW~SJ^m!`sXNXCy*E&2
z@kM_A!M^%*`qJ<E&WRq>b^2mY_tiVk<r#dlkGw^0`hv#gHM;r(^~iTt=!#GI8y)wZ
zW6B%oLUF1;<)4i{pZUbiQxEy+u>17D5BL!}_{E3(2gN%(=riS8^D|G%Q{-X?*Yry+
z^56^K^v&P+1wO=yYksW$`o2nBvb#KvE`2+n#y`JdKYEb|Jog?(en(F~;2aBo-b3gU
z2cy5xk3szy{?S3-b9SUR<M7K)#_12qMIZVk@1tCs?^^z6Px-_7A^XTT{E0t^dwQg2
za#{~s&wTVe#|Jx*Oa3w6IC)y$qBn>>xm@#m{wV*;C+PBTc^RMf3)ZndyOdCV<nP|I
zibwkubxnSBzUaQbkbT&}e%1H6-qXq-uIaDQL;X9p^yK^weQ3P8M{e@yGwu7yi*I?&
zI`+ZzAdfhYS6|qPKFGn3tiw<EFFKyni+qbu`a@s6qG$Y?rytS(m}h@Q4t<U9nb-k6
zXdQ8bFZcCN@+<xGSJ&z@d(oS9_`mV=W4!vpPWm0=@niqU|JfIx)}=4{w;nz{S1-(?
zKXG9`KKUI!#VJ1d6?*t~&0ocv`^mraXML)^lN|VVU*Bmw{`o68@I!C>24DEl@6rdk
z$PM|S`R37war}$k(e<1?@lP+-<wy7zH{`;%?>F^(;suhMzsVc?LY(5q`t)s|yseS`
z@X3DSk)82@fAZ5SKcz48UDGT7BR@OH<NN|W>xoPDgPx;jUH<C(O!<nQ)faYmp6KuO
z*%cjp@jL6{1AY4|^X&i3Cx`jovqN>rbL+8#Yj(4LWIy>A|M)}C`kq@a`NltedCpG!
zmR+H7;=nk3uq%1-ML*)i`sxxtlm9(uZ*^1N;jjEuJfVXR*XZ$od5ymDXFYz6F8TQ(
zJ<+%4ko@M01OCWAt&cDI#Rq!iWlwbFLwdyzdytF#uF-+)WPS9}V{d-wK0mX*K1JVx
zE;;l!>?p4210D3~%Xh8(!#Mtz_7Erf3x389@;!afw{`gsd!y$XU-F&!<yY#3@#d>z
zN6h%d$#d5CcH}SX(wA}Mv_JG*e$ao>i|;7-1^tqb9rOYCwT^hP4mtU!`}_-<XPou;
zxBi@7jb|VGTz!vzh+NQqiQMAC{=|HJ0XpbIeyC5UFPQrRa^VBt<Z1fQ`u}36^VY#1
zy6BJtIv0@d?0?a-UxWMrUG{e#h)?UerYHQ<lYGjq__trQ4*SZJ`1YJ$@gbkk6Mt|m
z9_{~~BjS@DydR=Zbet!U2VLg{jeh+;Nd5Qg@)y6RCw?aHiEnv?T=eZ+k=*Xvhs%rf
z$)47sKl9LYUf?`MUS%isyW#ss7wk4?`PYZ}4}H*YV}}=xKV$5Mr-#W+K6J&W_;O!-
znCCwIiW7O(eRAU8c>{eI=Ugw>CkHv{l^)S!AO1s+&NtD;r~L0ezn3RH7iYdJlo#cH
z`I;WpN#oE#m){wuPPorr>IOT|Q|=Sck<aO0e|rC079Vu#cV8Oj&z|d3)B|+v@4dIz
zhmyxS<aggSeUQg}^+3I&SNs^qzvzXX6V0P{blFkgPyhM|@~Myd59?W9U%?*yi9gDl
z;(}l4x71a9C5m5i^I!bRukOn);tv1(!+Q9j2Y%<d{3xH2N8jQ-I6Fdi<}d8$zH4&w
zH|y#*ji+z+Gavu_2R-$RACb#*{(!Fe5Z|tIoOsX&)3<X0hz|SV16};%i+yt(`{p?I
zMUOuuzv9(6>#!p~<OlK`|F$l@qfd|c6-VqzK79cH5GUkn{OYxrefaJJ)(p$v^d@es
zLr!(VeSN-l-A_L-A9@bSn{v`0x#`Dx;*0*sjgEfAdj;~7PyN$J8K=K-{%9Wh@||($
z(4+C_xh4;J-A6|pslW1p_4Hf%OzY|I^w;<?529z^0=+NR*PttpLUzEve4!rFyKCdD
zL+|=3$gcc@J|MZsfet>c%l{JP4f&d1(1*T}e!Z9Wp2>GP-oyEBid^0ULhtdQJ}2Mv
zdY?rO{E(Y|#HDNR>-5?BQ+^}9(MQj7er&$@@!WWJhw_oQ(|7S>^F2p5(R(-Z@FO2W
z_MjJbgZOt1#RGq_P8-sf{iyN!TJecra`G?N>Z-iVuc3ZXexP6U@h^_a>Akq;>`Y&&
zCw3%9`WZX0n{iOw@jLo;pMAXN6%U?UA78#xvM;nQePmpdn|{#4k2p}5>4hC~9=Yg|
ze^?ja>YVY$LH!whn9mOOmFkCUeW7?DH@le6PpnTq{$f75{Di!o>znwqKEOKkl>M~+
z3Vr<we<LqDu#<gaqIvkENA{;jej*OY!LF`7$G>=VpB>RdpZxp?l8?U0L$CJd`W5{g
zJ<z9mmwD22c4l|$TZf*=Nv^DCp7Rse)-jLY(3Ac{UoU=hUGv;$Cw4&3IO~#wob+iP
zx!DO_$Uo@C`uZLG8jr4Z%y*4H{CJL@?`qM7=&IZLLwxaHenGGF!cXka^{WtH<l{%s
zeg;~PTq%z{j=x4fJxdo}xc8c0`#V4K(T6y3tq#hQ{8|6Re)v%z*&lszSVDeEZhag-
zv5#_}pFw&SALxh+^!Ype<I{6-i!S@f8;RDlK9paqUqXE}#J6?%6IAD*zZdsiJO0R(
z{P0WkjhE--598R+_h}G)$Pd|9AD~Y0NAaew=4bi<-!uAt6uR#_e)8+f(ew8X`VZf`
z`<>#!`nx&%n>@Ewubd<C1M=y=@om50d=sjN;wS%ZKtAz(zWn2RS<lI9zk#mx#T$O<
zfqwWkzY?#|Ijp#6Cw60hazgepkA3WO=pSFkS=YJ0{t7+$mE87G=<y$@pJ8WuVqf_K
z^79f}pPu<Q{-OEuEB|n99{V@r>-QSI(@d0q>=)Qg97Fx1ysHn9AH}!4Deon+v$~RU
z@CW?6HqJbL0@>L(*XFZFUeh1Cd5s=K$F+6v0qwWdb$-bY`Hk;#+0E~K$a|2#i9>eu
zcjNqLTcP@)PN*+1>yYv8%VYcpsw@1*H9vB$<2gj1fASCcnE#6x_JQupPwEYN;ul?d
zbd7!q&2K~gg|0lzuKZA4lXu9WenRofF3!K`k$#PHpFPAYyRn1u=!s+Jy6Bk4PuzFS
zZu%d3$1i%;(+{A7ALH@seT{kKv7T%G<Gqz@bge^9^xS7}*Ze2t;Yai#FBtFKj(qxW
z`hffc$|q2Lg7|~#KFmBMf63SU4azh6Z*|t+O^Ivy$~=B5uklyU`7^orJydV?C;0Gp
zD%O_|{r!afhjY>r(ziUI4#~^r$tU!-a=#PT@Ai$ID|uX;k|#0y3;hngTi5#{el2e}
zxAA?x_2|cZ>yjVe{EM9EvKRfczj|mL`jubYhv?u3(wF)2r{6!)UqXFb>YW_Wc={)o
z`}pEF<fR{eNKg7L^YZ-ze`hz}pIBeLHVz;7r4N56>zcg$9lz{>Z+dL<#kCKv{><r{
zhOFng^Gy9Wy5bMNe#eX5=o^0!UE}a453nnFp}fbB<v;IL66FtQygDUciWhlN9<iSB
z)-}&_{_UE*n{oAbSvk&lD9_{9e0-@J>W=v97WLEe6FT(b9LBo%@pq%@96kG8LGNch
z_dSF26m{2o6u+P4IX=`|=d|_-&SRkZ2z?LYdlUSsyUvG=lMf;JA^(Kz4pZO!1&S|r
zlNZc~>>$sm<F5IUI1v}}jrCyWSL4tV|MD&L9HJ*4)qC?H{-E_-lc%|^e~-&J{SWlr
znDyz+dp~swKJuM?zINxalPkaHs84A_dCKoBkdOcBm(@RZ8Y$X;$#41!dbY0fB<n){
zuHR`x`Yxe7jSl<5l6?Fm=iy)7k?++fb|5!8>@U91QO6;^<Sp~8<2gF&nEF)0pC2~;
zjH%a*>m=eAI%g7h(DOuoTSD=<txz74Pkh&ozVBAVk$yxx@h^VvzH>tN#T&bbXMMkO
zDt2_vr{Cod`c3|#-_fr)e{;?#?^zFj)<d6P8>f!(8}BjgYw#gH*-L#UKeVp*XY#sr
z=-YWE6o=N6-_#lMvI{-CW;b-ikGhOsaf>hg4FAwi7zgzo4c2{%bs>7t_n-2%I%}V-
zj@!rUpYVr&<MaXQIlGH*c4l|}Bkski{Sdtx?;KqJ&F`EWK>CFG8mN!Kw?4;rg}#Rn
zSM1Dg;)z}8-8lXUjiaxTA~_(sB~<@P$R85<CwWRp9>^~sJ(N(ME1`Jr7VRtenLGef
zPJJD;KZE=dX8lnIq5Y8S`xgKH=Pz8aT?Os8;I7Yo{-&jC$5jv?Q2vGD9?IWPJeN>D
zg8at!67q=mM$J0)-#Ms1;+|fei|Nz(weO42WoPzm=zM(Wp<9>#U|8I<lfKD#_o4f+
z<fnNbzx)*1KZqCkOT5V|CA80i_=56C3H^SU^Lh15z2R5#p}K5;@BCjM<9y%mTYLYi
zPo6gax!s>#G^R3-fBRi*b&&qpQNPO$?$eX}sK4Z&;wbBZ^A&vSFU)goKDzt^s$=+q
z{ET1ASMGa1W#4Ll$&c|N9w56y<I$xL*HFLfe9Jkd-#dYRhdoi>ieGWa?@EYo$c`m+
zj>=E;hvL*eNWPW-)Wt-9H=;k_Z|I@JPuW}l=v>Xd!FyQwjKAs&#E<)sUyvI;e&ibR
z7xr|WNU!|BbN&J4XK^JD<Ii<zU#>5P_!5Wg&);0j-}+vB8i!wUxE43idO6NIiTY~&
zGNcFjO&y{KeTezVr*-fNlTX*a4>ZnvXuk#V2aUrAee)-Fb4@O&9`b+a9816LTo9k?
zlE1_7y936FD|JnM($8B5>UXym^~w0;PxR!z{?D~~i(mEDz61Ylh+d-g+7O?RT}y~>
zNWaj1eFOWjpXcOE{nG<I+JBIPKG@OtMB~&4*XCK*e$I3Jx$iqc<H<qao};U7lFxn$
zAI6i%HAL5Qd_i)e=bF9f&3UtV<Z_+-q2nB!Jm$F%(Z{F$!E^pVUl2X|{Ks#9dHZV*
z8{g4iB--af_15_abe>_p-+NUr<xl7D@|5xR8~RKAnELFwdaHjcq5dWj|83|!)M!z^
z2=S@Dx1sU!xP6Oy=b9g~J3nGy`gWgR$PePo_qOgkmykcam+}8coL{IH@~GdtHs1do
zbnShOdLs_d!6(0#f83Yv^%3-;PU{QwH5r%uP@Jo?=FzJ-L0?>Z?i!l!d=GtfTb<Td
zpzAq)^+oy<=YII)50KpY7Kk75>AO(x1@v|N&^r1<{PUyi8$2g()-U@n&&`+r(B%*M
zin-f8z4zL~dOQ5ue%^R}hrG@&Jm<gsL|%mE%lDAI#fSOq4be&eBv-~^%3&S(SiQoR
zzvuSeQGDV{T#}#O$n8G*u8l(%ALyHJoOyX4AL7?K=wuv|&${#>ub}U_JmNWitjEsu
zjURgBFZ$86hq~%MIrF~!!oTn>Zs`x&_Z!bY@M|1D%KA-iagPqfuX*^TC-G)Hd*H*m
z@`8LMKN{znUganA#ku%*pMKZ{n(w?o-a<z{ly~fh<tO>oJ{Vp4fapT+JDk(VgXq$?
zy62kT$V*TiOZ2`#UqVjr3FKSnTh5o9i}8Q{El;X@<TXw_s1xE(+_4M$`fiZE`3rrk
zJMOb5KQWF!IS0=1{Jezp)P~LzO2`k|&^dg!C_eQ|{EqzMgFWbnzWKlWU>-W+lK#Yn
zaq2ia;?R8O2hQ2ViFxQG%1b5GZ$sZD^4qk7c!K)$HhUP~hV-h=(x-D1`H}vd@7YiA
zXMKlp^d{ek5BY)~^+V)we&@M)`bgK-p?`ehLp>o6dy04Q?7lvZ+}2Bet;?_2i=6nD
zPayqT-#YwU9!NAFk_R36nBR(*|7#R4<hLKCFLudmd66DnlPm4p_*uNyre}Ib&wBhC
zA5i{)?xW{g-GbJ^pZf2-+Wyi0SKiHdwXb&W?HoMslRJ^#@k@W~fM4|&s)PRTExxVq
z{Ujuh`E7`vyawe5$S#oG&_{<nzbTZjtxqrd4A=4$em!SDekCsLYwUaUlj2qU>n~lK
zr_azg+mGm*^ac7EeWZM0J@o~B{K+@+ojmCO4e6`o>8vNd`{9rLk6q<s_UF&?m%M>5
zexZ+Wp2IHAhxCv1qMt~)=~+Isj%$6MYx^hqF`k~(8Sit+<-A{gB){j0o~u9T(+hjx
z-?ey`FFjY!ApY3jICW7T;jj7|=ZDTG*u%Q=1HW)BK8)uF>LmTrC;G;D-^V}j<62&p
zhxwsv{z<R=9bM1mW%~?vQ8)0Ze`Ig_W$_1nH$o5I%Uef$7$+}!k1vnNEA%2R^a<>S
zKjYOA*NK~EoOks2Z~abXUEf!FZ|hvveeb8yVF&vxa#&Y>_M9J(-#GK=M?ErLy~%l=
z(~~@FzWefne4$Q5&z*mphn{)(GoF7!`3JHuRPUkp+K?YY`hnzy^b7er#0MlF)Tbn>
zcTk>|4~+Mo-S?OBy!<0R)eZ3n$te#&&&iSKIh223j%Obze?#>YvOAOqpnO0-qeb}+
z;)nj+hvb#7$*VsQ|FFqZ_5Vpi-@ev!eSvy|uDZkS@;CnQ<Gy*u8z;_;qetWA?-D)F
zb38iMQ@7{`l9wL!Td5D@^ljwl2kIs{Jcsfze%-eYyI4nG0`W`EMDo#V+TDEg=>>Yf
zgD?6Jcld+&<{$2}6F#ix8rtWx6MrKIJ)#TgnO^Xh{Gf+Fa*2QXDBTyw=%SZ-!usTr
zZ^@@VnNKeEa}I|t|DZ?lfe-p(H}mC7bghFg`IdgwXX}`czH!ic=J9)U)s05}`a4bX
zupfEI%O2LjuW{^PefpM<$;n^!_x3gVV&|{?G40>1JN@9}KRa&IFu#H789Dh4x#*ET
zJl8Mkr}Yc?W<UKVJCobB_%x5+tivwquRelY>`Jfb(kH#3OCIykcb}Z>sJ`=ebj;_M
z##<Ni8-A#-!EYO?v*yz?eyods@)*yb^>6xE{V)F5-#*5D@|uS}#9z1Qe3w1hS^wI5
z!zmA}-(gILAG06-kbn4RB74v)KFLom_OY)Jr}jy(B$xdGf7I{m*X<kZAEEK&rB{eg
z=sOm4*@OMbr{DH}hkd8Y-~67E^JL@o$-cjJt?#!Va{ho1Xy4@jgtCvnM-lJzjbHcq
zb0WWBPx(YXWUq`v<Dq%z$s6>@zwBSpQ-`hRT*NiMLmwUV%{Q-v`05tPohWbd8;GuW
zNR&UIJi<PZUJ~WUk)r&RNWY2n*@p6e3Dv!BQQd7r{*b7?>67?Lj?3$mU%yd8epkZM
zJooW6Qsjr|@GpH$^5Z_g;y3&h9jJb(SLvro-}7!!9)Ri~6jyDi9^g|xm)G2v|J5ON
zP+pXmbG~?1H;j|#y=U=vE6&s8e~1rAzcAyG9w5FTz94xad7*Wn`%qm=ly~em#kc*S
zb82y_FBVt6%Q2qcqUYS4AF2ngooA!x+(7@#f1JbWU-T{NA1w71TYH~h@qc_l^z{|=
z?(b{po&DIA|G3W|ApceW<pD^a^aAApdZ9o3(;vQOE?aZL`7aDq<gqUQfc(HZ&Nt;%
z&+(5gKWV-<ssH~EI^I9=3wgcKOZ|6(jUDRW55y<GqgVZdeuaPXZ-}0CGjFcEXxi~t
ztsGx@uI`{Oe@*$_AH8qK{l<0ld+xJ?{ER>SAilg`Q$O)XKTw>ozdX%8{EnTCryt|V
zi7tBV&)@h5Kf=HH<kUZ?bNnmiHV?nBxnIA3XLoY&8{@l0ak;Hfym}AgyxloEJ$a8J
zUy4(H1OM`pozMCHRZ}W{%FpPZ{Orcxm)<{f-7y#Sbohz+>K?zg52G*k<(Gc1NqzG7
zpS~kef31JyUAy*9dHX;|+(Ui>{oV1B(>{9gDbrtF`90d5mfbyk-HO*%P+sCU@(6w*
z|G+mr@JD)AKk<P-=)QfZ{iFSxarW!_0{brWj6;tc>aY3y*tPv1y5`}#TR){2cGuTs
zJkuBb*@w^{e}JX+jprZog=_10zauYr&aUp8ul^fH4s^WdcFo=}`IYCq$F@I|pX4R$
zh&Or4H9iyR8~-J=4m;61JCGY6?&HHcuEn4C?)oI_xF$Ec^x;~*fcPt6%7+fbC&Uj_
z$D!{q*#r7c-rt$X_v)8%_?9m`m#^5*b9#gP5FPrIuf5Mf58voHm$eQ$^1Sn7bj)|(
zH59M<0DW`Px4!<veDagacPZAxukrdYh)?vb%Wv=_zV#=@nTL-3f^qcYx$(xKgFoZR
z<GX15@}C^fUg(n>AN0mf?vpd`TMr$2U?+T;=ly|o(1-Yj^pSqYekH^Y)K}0O`CX%r
zf6u)SArJkr2fyM+?nCRlhSrl$tOL~<es7-j>4$#3zZ0kYf;{*W->%Vx^eK<<AM1He
zALc>pswc)lbl9K%<wg9t@0y-nTMu2R&gsXTAKOQq_rQl1JiM~EQWxQ1{X74@i_yR8
zH}o<37SHdW_`&O+|F0i6zwfDE$?@hn2Y3FVKXzWP?{VG@opbv;wv<c%<b9X^OTUI+
zeUyIAee)namXMs}(0{9U*$=3X{8~QIzpLx6_2v30^61;yg}$ITPyZIr>Mefwnf?@t
zOMDs+$*Z4cU%wl}KG6Qb|0!@@#{T@%zS#Q=|EGf<oEOlKb-n+QFUYT7fa()WJJBo5
z=cQ|YN*?Ef=&(CGi(_%0eP8O8UF;vpFMj#6IG{iNO;7Ig7k`(;9@asJJnZVbEBh4V
zAUWyZz7o=h{jfa2?~Q}__j_&5EyxXx<7f7%P~7la>(RUG#=q-xBlp<@9pm^Vy^#mr
za?Vw+zi(Af<@;rR&R_Wrd-%IGe(rZ<=*@k6@MH4R2Y$`txAqh4WWU5O<p<xHs;B&v
z->P%+g!*K@`se(@e#3Wp_?9otr#JWQYw%~E=RQ5S)>ow6@hjh{pY%}5^Do!?^v4g|
zkerZyM~mz+Qsh5vXrKCWL;BlR$iMidcvP?Le-ib(&cmJai97yI&-NGg6Z%E{u{_A0
z_z<_`m4Ecd{6~I~C**-d^;!Hu@}noOB;Vx2xBMgz%Ln*H4<F(LpZJkC=v|*rkNC46
zu&;4V4t!e|TF3hIh+p+wza`Jfi{wLJezzYYpM3AVn045lUep!qtFQD;Kh{NGT~?>a
zLq6lkqu-`i->pFFsE7FDhxA8od0kz$j{1jRbzXnYKK#e;y6T(MCD-UX&-ME^>Z|W{
z^;P;L^Ym$+vx{-|5BetkgTF_|zrVj=Z~CEs`x|!QulhCqL=JXlC;b!@$L#7F>gTPa
zUPE=5oxC4UKW2A!;iu{vd%ETi(77M~Dk1%Mi|mxhzr}~&V<{njPoy_?DItAB`h)C~
z=sCLlT0e&#`NTi{(YxpBf_20jyLiqn{I*TshpyWH{3Ub8Rel##e9Nc&nZNQQab-U&
zPs@Yk*LT{7$RqN&xY7@s#}CO*9_z>(zE@y>^y!oQ^n?%B_9^1pefseGoAQOY$G820
zeVcO&`R|jD{cz=qFHflaKLhXA{qDqVFJ1h|s(U9_{%?Wb*|I+ozuw1feC*kcAK7Q)
z5WajzYG1Q*$Ax!qK6Y#+ufJ*btyAXC7*oA?!&$RWduxBiZmZYrbj5}z`z!OE`}mzt
z-$C#*{-qD(zx>+!ApT1a{6?Kq2jnUF(D@a6v!CmdKj4Gks|)g@`SLn{@ZASKoQKel
z^D!tM<ZFHgt&bn~#VbAWuSDbIVJP3}6U7a?n8!cK!yoylI>*2GmAYphJ&8By+>ieF
zm+M4&moM;Pobye7u1=BzU+hj!^`4*USDfRj$DXq%ew_#L!?cS!&TsV*<e_)#$OHUU
zKbrZ7e&r8!9$)OOKDckZI?ccMh4uKG=k_K1MW0TO?7$D$-8KKryvPr&Pfoup$FAn3
z{_&^Ih#TlRKKY+|NFU}yc7pU-LUG)N)=#7ddBHmLVxMT=Mjm;R-;tM{jAL*6KYDhJ
z4t?-T`xXArZ}HDh*k7N>Ui?}<)^Dq~^0n_O<r(ie_^rGtZ^%2|8;C!CC;pv_@<Z~7
zFLFS9=~K+tmxwd@5C8Ir{3B1P8{~IB%8$u;%)N(fIA`6aA<x~HU#v$T@}K<X+PaXQ
zoJZ2*VEtZbpNsdrdd}N64B^{3z4K%1(y#u;_ptb-7jnwC^vrJJPM+f@P`sJPF8Fg_
zp0+MI=|Mk4KM#NJ$qVMM=&y`Z*U>Rg-{gGv`^O#e%X?<_H2-Ijf7*A8D}9W95Z~fS
z|7QPgy!g=%@l$!oJg9#8z6v_O7oX%~FLFbF&!X?suer7!{>XuU$X-w$GLIf0Iq@k!
zk(Yht33(8%8b0pPOHbN7JnIcpk2rB*Z^wIg-%omP@A|7x+&bZ(zPEY!z!O(|`l-G8
zI_CTRw53yCxBrz#Z5e*%@DEJ-(yGCV{`oKe(btMw`H=s)R`1nk{)L|TBSm(H^qHv6
zjuzFM5~?pHOnzMRCvh&%sx$nDp3LKS<o2At`H^eum~S4G|K%Tjf_2D+Zqh>+zvdao
zPxS%Dp=&=%4)Q~BYaQ`GKJmg&&C@^7FL})`A^zGBzavHab^VC<*v?yg7vlVm-|08?
zqxR|Y3OmZ9&Uu}S>u=bZ{@I&-(9?H8&&84agucA*+=pGs>3oPh#_5mHr+4(shwNQK
z>zbcv-4gQG5~5#1`YIv4m5`hzblx>uEUlaO*>!7?9@(FNLC>wD9;To0Q+_Mn@h#5f
z3;m00^}G$$FLh3TDSzk>^?%Nb^o7=y|Kurszk2ALN`GO$pbnz1ztT_3Tk5ZUf^`$=
zNnWIPe5l9z7pUKtPsArW^ltyfU+sg{0rcc2d5IkIk$gt)o|6YX$gcQA$9noYb;~$)
z1V8pG@;LhR!w&XQ*44N3TY8ljtcxDI(j)(1cm4}g&+;xmq!;f+(c!<&7s<<B#^FzX
z6kqs~U+mZLZ+%D~C8U>z?pe1E96h%)MBn;C_D0wKhMw8MzRkYOJaNW<_;n+H{dY9t
z$ok@&e)SFN1N+jGJSGp(tNZ)}icdIL?=wqiUkLdR`PiMD{G5HnJH3#L9oX48`gRT3
z1*UziXMWn(wR39rMBhB|Dn92en0L^kGY6{M{(jv%c3&{Q^8UoW&flMj4|Mc5P+wsj
z)HjGTdbU4Sf9Tz{dHjI=ohQ<-`{vWXzQ(zd@$x4Bkw5sYI;c*HOX&Ted5OMXC?S7u
zL;8gDlE|;sC;RH{&-~k2!>3NF)H8MXi~n}!;%oPuP>E;rpg0z<uGz0Smp<UsHLt(?
z*<pHTC;1ki^zWRHU5(Rci+g%!2lK4M4%SoujAsw=h@SaS-o_`taDIZ0e8YdNV;=hA
z%zJ+42l^EGNj(sM{1ji}R~{B`)+3j^DxW4gH<Z8S1Lu0;5W0^~<K;i{$wT6dzo{4U
z5WOWj7xey&obnz$%bWa!eb^m(KPyhm7hn9%{SrD4ASb)p|M5@q;LE!18)topj=sw{
z{>#4PN<^O?jQ5?Ae5Q`#kG%3UeK;pG-*|jM>zZ#I{>bk>`sP_*-zs0DCl2Hnbwj_%
z599&s>1*++Zx)~Oj=V08(8ou&nDyATxaIfk;=c9C2lW?Ff9P6XfZ|1dz_)#`b=kqX
zuGs<7oB8}l-SYm}ck=Rpb9egGXQ?mphwr$Z3(I@r-TM)Uu66VwgLVH#Z{#6AKY;Rq
z{A3(-U!2pc{6jA5$>&i1Q&-8WpV23JuD?nYfBG?bU4O&w_F4R^g#1Um@FVh~PrvAU
zE^a+f)DL8xQ76<n>(dLq_#;gEIZmAO7xfylkLTi%pYf}15nuFV9zWq9Str;d`JrF_
zD(?6XeeySStm8iU(5HX(n%<0~ckA#2dO??*=*o-ab$;Tz0sRL0`U~&#=u5v?vak1;
z*5!B5di2bn$mg0I^u;gup=*6fx5!S3^ymLe!=8^`y7+SoHw}BA<M$K4`jwyGK696z
zPX2zq_d@>9opWdB()Jh58Pt1p-M1ck`f~Kpk>Ay0eXVhRFU9)kpl=*{>V!VSbLZ*&
zU479<k;D4>2mQR~@{xE%4}J8|hw46h=BbPFKmRe$bNoQh$%!w1ioR=lmbcKsk2v+c
zfOAcL?Y_8HKg5~1Qg6lG)qDN(56<|)*eX%n+t)d_aE-n^A)f39)GOEWj``wO{E2tp
z3#f<A5v-4%_rmg?c!laH^j>^m>V~zK%p9zseXM!%n)ThM*JtP5zU-3QwhYN*@&!J9
zUx+{c#?I`>&iXw5D}S*&`tnca75;}le-nS=%{=jn5BJ4~=ll^}{>rcUhkP$TxS#b)
zA7@?9#gqC$FZiN=cCjDQ|I4f5oj>85T;j?5Jbokp;6t1mpZcR0{6ci(WpY~2bA6_K
z2-%VR(D?>`AQ%4Vh27|lJ;Xgf6R-3J<q!HckN<hD?=VjOF-{)9m*@NhvcEiFK0lWS
z%olHv-t}#nU*rkjwc;1b3(hy(rw{zOPaY^9*n_?Bg^qaO5BSG7f7J)OPkw$%@8}r^
z@dfFbK8<q?(L>*UNuA&?=F9u`Tl7btzEko(!*|vnoba|suit)RMW4n)a`N+x2mbDU
zv+u8ehmGI2Va7Rou_Hd%OWoH`THkZ;o#==DJ(rinqx^yo{gP{a2!5;&jcai8+}Y<o
z@#v7a<6k8dC*sOE2Y<JpWq1Ao@j*`LzHyKqVA{vL5~2ge6>RiS|Gpmoajr<8CBzrx
z4~fp*&=>#Kw+|<uI;9@Vb5OqE7yN{OSYQ7je;ZF9(0%jK7dOsj)GvJC*L-?%pPyJq
zUXic(sd?x?{P0ifx;D-=dgLd!Yx;EGy7G|w@(jNs2frZ?dh`$FIs7<h)&E%6eAoQL
zIM@7zypVq674gKcjB}o8ygVgt@$I>J)`#dC2i;F3zx(LseRS{z&GQ_Z&(B<=k6(IC
zJ-gqA;@Wo~_7UoizC*rC^xXL*zc3#f=l#1p3+;QX59Mibp$>bFZle8?@z#N!<3~Mq
zKhbl1dY*iVfANZc>(Zb8fE@Nu{*KK4%Q@{Q=Ty6mn?6wazGm9|=XQT~(U?jfVn2m0
zw61)lPh~IPK|3$i&)YxnGy5<8#2@7kc~oA_cs5=?2k|E!_>cSY7PL>*zxrJy`%!VA
zzTn?J$Nmw2zVnt}#H0P8{fP59c6YwuoIro$y(D=&_gxNs>Ce2k@_ShL*Y`m3_&X7G
z%6X4@>?FUM=ehg%r3c7<;)dPiCHjH#E4f^IUu0Yf>9>T|FJX!9Nbif|M0$~T#f5SF
zOT6P#Jro!8j}E(&+kNBdFYmi%NA@yaeTVv^5|Yn8oWIEf_T>;AerG@L|Eu#ic|!ey
z{1e~q8!vx~AAODfL>}Y6#_J2jr~R9KI@I@g&d<^J93A-?k~>k}(m%=b^vWJE=ey>&
z>_G4AA<vH#=~MpWXZ+Utx5lsQ@7nm8{K#IAUyB3&kA5P(>I+J!P9@T_`t|gkH+}s1
zQ~N7)=tVtM-_=d_RhQKj^~Q7b^hf#&^!d$T{r)uT89iEu{*99-A%E1Tic5JH%CpUT
zo_c?eK01)!TNkoBJJJ*X6j$g%apLd5AidEG`NgYvfb{y(^3{uOdva2R56|cIeEZ6^
zd-gToQL&SLh<s3dvKu<`2m18KPw5}JFQ2%Ucg3eTH;#Ucr)T5Xl^w+seWFYM$*=L`
zag83-f4CNx=$R+q;@^1un`e9peGkB2$cr9++t4|#@5A`NI*NaM$vepxJwo%m=Xc*V
z)Q_p}=J6NuS&u)kGxQye=j@4ZekQM4r=c7FpPpaQ2Y%2oAAi<Oeapl0ruD^%cyul#
z&qDhieYbdY4lB;(0r^%w@Lc{x&wTRf3w+PyzWoXL+0l6%{qZ~cpnoXuIp6TTk2v)G
zoH+8Gjygm>c}9OP&-y(KdZRCL(gVKf2fy+e`N&0XNDlMZk6ifSm&WUx@I!9<NOcfB
z{CQ6R(S!VyKk8Tb8UKUw4-_A+`GN8Lz%@O{tG=g@KRg!)#^G0<bB%v>%XuaJ@*nvU
zzve;fIQN!U_z(Md4{G1)dk1yQ{vFc0xDiMGuAIG%V_$WhU+Wv#-FGGYj{N+9-1<}c
zfbtFb?Mu<&$Mi)%@|SV^#&dQy&ihg0#hHDNb)4hM1L*Qc`WIK^;}7ghF4xfhg<ta@
zae}_QD*o|@FaH<By6Ajh=)w2>!MQtB#?yy!;sAP1PH{s%b&|dDg%9J66F20L$IZtF
z{qO_!=WqN`zR~v^FD|V|pX^OO{j>Y{XAktnLn1!;AHASw9sJ9e_OblKxr^U}_r3!?
z<Jir8{eb%A`$PSX{z=}F|M(fZxNhPi{-16mXT5LJ*Q9^&GkT+UDDTrd<lpQkKhPsY
zpMNDic1OpxxaKENo#ZF>ulB3_K;MF2c{IP1l^>ye$o`ES-j~(-6vyl=zmiY<`A&#j
z{0q{%c#{9=)4o-ok!XGN^LbvA2Yu)K;*(tDX9sqK>_xx&eB<Q{`Y_KpeDFtcLNDUQ
zbC}QRSzPdY{sGY=muu^~Zyo%5PQTu3)4O@(k}vp?b6)k7p81<|1pDR^>M!Y?{;Wq%
z&+%uTxbU2v*w1tNBNw^(5&s~s{L8-L)Hr#<dibSZ>${I%bpv1G#5#Fjf32UUXY1fY
zoO*Ado>?D1>|&n#{24#i5x<arJa<iB^hq!Bn*9ef9$mk`<M(>#OMb(z_)b1u^GAqJ
za+?qNv-!q*4z15W@lAi`^Aq-BKWM(+brkpN2)|?xaSM0()ce;Qdep?u=cgZW^}8>b
z+%Yck`e~0XeCO_yJCI+E7TGzG{3X<Pl~7!kkX{a$_P_z>UG&T5|I_PR^t;Ya=+k>J
zeY5;PPxdRW?W>^vMV{2(;75NbZ;D@j&(GCw@9nKCzw4Kr$H;fCJvU$e!w0$Wsjs33
ze#7r}Ui|76Ke?>0LqGaG@8z81L+A3oEA)RVoZA~`Kjb?`c7gdm#XigLlv>B{9_a)9
zU72-_^Ls1k;9tFTA3gk$+j}Ye8P5;1Z{=^s(X(rEkQ0CS;s3^4$8+-f|2zCP*CoI8
zyjQcX`tLqJK;L<j_3>rCb@j>i6YQXmhxFii3CU$&!SC#2#V<bobV=WZC-l5Hj1RwS
zHFvwG_g;HgZ|6OS{Q3JIzp=ODcjf(lmcOI$z7Ed4XS>Hfea*(^?<@ELJ$euB{hq&L
z@SI<#J=|wU&;31x-|0fv-(BzndSM@a$}jD|=!M?Iqv!15JwN}mj%)Ps$$t11&+J0K
zp8Go#`b&E3i#~e%)_vFhzkqy7PWq%5b+uH#^h4@!)+u>8^R&E(u6lrueW*UleAn`?
zzEU1`jjn6-<!^G!>rfnfuKy6X=;;^G!<W8=9;~ar80Wrm`ZxWGy6*jzagcxUSAHgM
z@GtSLuCg;ENA7#+w}kBByA^cQS@Y<}`5L+Pr|6hZzwD<D`%YHhXq@%<6XbX7YQFx@
zHNEHy<Y|2K1M;w=b@0vq)O~h#jbD73rw*WNJ@oM98eilxUcPn>(dXB$%@=3ZN&40$
z2Xr5uM9-o1Aw82DAGtn1b|0VU&?9+`m%q?A4;nA7-N#2_iH>pfB7f5d`L`CuL9<@{
zf2imI|K#Cs?n7}!e)q|T??m#+FZxjVM?T_T>Y_a3nxE>E^n=Es2h}4epP7#zc8AV8
zp?UPD&y){Re)|CWf!4#fzCixfhw10=LvHp#$F+Tq{RaEEPp{~b-+LtY<u!84kLZXy
zal<ds=V#WZujHTK8E3xzm;RT0=+JxeLvH-Ai##h&^J{g(ee09YwRQ2qPLQAJH{}ie
zl)g^Cgdh0`zt%%9$I-X^&)>}7S`?qnI`wxu_`**cCST~1OC4l?^;{oF4$q-|g8p9o
zh<p7k{p+tY?%5xo{_a!0!Y@9_tDkaSr_XeL@7jAzeVq4~`VRcb58}o8{90VFH@@?|
zN%|4Fj1w2+qEG&89q~f{N!K;{;?sU2^O14ri!1SDe#1}wT?su{4?o7apYpIXf8}rV
z?YaBv4Sw)x9sHq(4|35TzmuoYlkcqWIpnv{IM?RWt9kgdKE6G7-?*gXIX=y!AM2A#
z-Lej(ADH*ix1Q&&%|jomv!0vhIl0(_zo2UzG@qRG4EaIsZ~fkaeK2$mKtJ-4^9tv5
z3r;`wx`A0^J5SuR%N|D_Jn7dEefx9s{a(0p59epj4Xp3n1Imy7&lhyQrXK0LoL}gx
z#D#k0nw-u{=!xF=jk@7{fPSs#JkN7-d7mKvu&?;mm#{DX<bQr<p7@0Fg*>jm;OF|S
z68aqk-=(QnP~Flep)cRt->MJL_cr$J-g~GU>WaGKy@m4xeYo}2D`=ljZhaj+st@+n
ziQ<rbp?bv*;z~TS7r&ur{iJ@EKico8*Xp7^PyV%DqWI<y)|EH;xB7_>e(c(O>+x@X
zDE?9});CU_lb7T%{O}v^ADo-0^YqA0{8Bz65B{vr54_*yKd$+azL)<P=X)^U6*_<5
zzxeiE1fnPY>4D#g3;HLgxa6n!NjmNuN1y!MI?frbBd+8Z@`yij(~IZy=)HHw9scP<
zoLh%I&BLdBL_U3@bF8FieeX40ql<6$f&7=h@mKyWUey_SS^ZEiTszM<p5Lht_Ce^1
zE92y6@hz_L<-YoYfB8l{8L!_|UyRQ@BTnQ!<IzP=eG+HliJ#dYlPmK@)*b!3`_{#u
zYv?||bKiV^@4Y&C*<YT4_L=q_-p|{&@FVZ%eb=eJ8JF+D=|kM}b9Kw_>SWy6e~1(L
zfV_Uc$9m)tkK|Vu$t^C-^Sw9!U?=As>XN*w&d9gU0bGk${PPR>MxOIr{-j_0(GR`)
z&W4`Ub11%D%U9$_7k~1dJd@}-e;O&u8|0H0#Jl+Roc&<tWqIFx`sVlin|;*#%qQ%i
zE?SRYiEr_4J$(;*%46~+Kar2f?>T*<=e}$C#5MWkcYebU@xh+ZIQp_KIWs=^llo-5
zdIQnRaqgpUJUv^FzRAPB`T%s2zWf0jIqP$1`bb0%AN0kK(c?et&5zi@e#<z0qUX6!
zMhBY5-{l#6SdV_vFXVZMj&=DVG>$%uQ-|=6Kk?x{`tF+#@tLSTTbF+58Qny6I`trq
z=*fDn@$36H^XU)2?C!q!FkYM)$B+Fzf^p{4v-|uJU*vPm5BUK)5TE!Ur{|D8Tw8~}
z@d4@GbL-eQq2s=B=&~nytxs<I8g&&tbga+MT$7I-@GDNCegr-G<S*=ne|ca1ww^wb
zU7Z)Gzw({@E|0TEB0kL5x1oa%dXo=*XQq$Qztbl_v2P)tdHNapp*QwYzpZCJxygZV
z<Ir>dAV1?%AFEE|hyCcwwe{GUobtLp#kKPc^YLMyOCQGL1OLty=->M6OdfpO?~u!S
zuK5qX-FMDGFZ9SC>=)SIef~#2{Ncwq_t7VxdO=Qhr62sUKY!r=`cr*0zU60p^JD#w
zxME*^YJK`J&VA$fv-o8n{$^dz%|ox5U!U*u|Iwm+(k;q&-p@H_b^d9;Yags0LHgAP
zvIo1u#=qnD?Dz|P@Dp(*KYI@8Q+{9<c^DeU@8}<2@&&uW>@V;mj>M_?{1QLnpMBLu
z{FukD(X*eSFLlR#*ZO#J8BZ?jy6@V$`XqY7H^eW*H~r{$(bM1JPd+kE|81W}UU5sm
z5MB9=J@^~H^na**kLWu-eU(0dp2%U}>%9g4RnO!H&#f=6`5Sup=D+l=UyvX1<J?Fd
zbMDJ7zE6~w(087MfAo#RH$O+89P|XekMVz2>CJf-K2G2AuRF~e*V{S$z_Vv9zhld=
zzpHjG;{WjKC;5T=?^-@(SNb$g{&1cnKI92{lo!MubZ)8s@jt&u0gV%<;+mi6lf;d>
zgg<t7E=K?C!B1WD7xVd<=j<*YtA|iNcFm8V`TPiaF29R+dDQy+i@#jE-Ma7Ga_(z7
z&^)L=l^?9*n*X8iIX;aOch)B-yF2I6Z}@u`$ZqllJ2;1-PjZP*^!Xim_@_Ll|KP92
z<4eBfpYovm2hrn?{FPs*r}7N{F%I%O>sTMsgZcDto@;uLr_9H%`}$DV>JB}5e`1{a
zA-=@DdGx6cc%S0^0KKYr`XKks<6rbjZtFUSrQeKW_U0e*9lcst9v~-v{k@#?M*jzf
zU$!AR{BDEqlAVV-Cx!G+PJHth@#OCbAi3#@pVE)NclDfqh;w@P9$&tM?2Qk8#{c}E
zHRGL&>j!e4IEDCEC*%qI;?sS3MxOP5#r=J;zQelohF|wx(-*q#L-u!0?|Uil@%g=Y
z6KD8yzF_=}_fGim6}RtDsYB%ApRS$j^F#l?)!$*F=UiRA;dg!y)9;^jyI&}c;|GcO
zrRQyh>;&ZlNZ&B|qF<QL*-75*7TJS;57xgg#_u8fc@K5(vWKsE*P<7O#J{*xFT_7S
z?Azrj@!+{UYacAH%d6y*$K982(2)<tA$`f8`cQh-&)eTS_YimJ7;hc?sITtpyT~se
zqlZ7&<fd2eljKe7(}(@J^ALH)`p|gS^veIBe1TtfVJC9pJNsqNy*Dr)|4{yrZ>?{>
z{0Z^pdno>nj(*TM?}yQqKg`pwh&O!Tk3EupDSwpit4D3<djWAH@A4;du^WD&dJ5%t
z`AVIkC)e^@-baW2T+;{4=khlFTHkzjFi%}_&E61y^uUkV$8-M4-}tZb{K|d&izmph
z%@a@Nd2W4q0DWk`YQH1y$!Q;K|DZpye>LBCbnbiqNN@B?@4maTE=14!R&|@5j6+ZS
z@DtbarTj^c@(`pK&)Jh6b3Xl;k6-&O`)2E-E1$EI{08l#`G<Xwc#`MUHK^ZXPjwAF
zevtbE@oT<yq5j)`LA}z4sk`bX`slh=@6gc~;Zq#Yi@Zi({Jn|${>QJo<Uj5lGEe>?
zhw}{d#GQPKu6#tl=HrhX5dVqx6X<%*uGX=ymRI<%ICEa^yv6wvKa-dFE4lcWbv!2*
zee!E`jdy+_Ud55|;>~k<HBNksd-34D_4Qr&Wlwy#Pfqm7qyFVJe|4YS_M?8sME-!*
zk@wXhd_Z*gS;}v|e1RWwlhZu;7+wBtym9=|y6kP7`R<!XPww+W{%JnG$xSc((tUp5
znmqLFxpnY`e{pBLb@)B`*^M0JmT&aCp0l6z>0du8ZuH&GtISi+#3jBxN7r}pzJphn
z_`7p(d7u8p6~16&hx-5RH1)gwoih1>e&`#2;=*(KG!C*qWN&)LC%dqdbzIA{ZJ2!}
ze;~K(CO&HUUGtYj`IO!Hr+k$9kWc8rbMw?``N@5KgL&xiJ4i2{)1!IpB5zvXc>1SL
z{*O-PPxI|3@UOpCFZe6J<KO&}yzGr1eG0$!z62fTyWRuYztI=}!l(FgzN~-r-IMnk
z=%a@YIm9J8_+$rtnY!uR9X;dKZ~4n}e98|gzxd{7^yfT}-K>jmd4@kg^yMdfv!D0r
z_~H)`efisb`NTPez7d+|99}%qKmTAq{Q)`fho1bve>|se{%>7&bj{A}r_RcQ{L^^x
zLSFKtV?KV&V{h+`!~^~k*~fkPlAqIu_2~h>uJJ=|=PK$^-Zzgto|9YOPe17KKjZMF
z?^Zuuvpf0Br!Rhlo;<_<_@#Y@{|AeX@3`ePeGa+E$<Mq`vYt4_2b2$?`RX3Kx#qw8
zQT*|Da?lt1(K|n3ck?0pnrA(HhzEN0od4QCTHpOd`N#VZ`Vcqj7JWc`sz3C?p7OZ+
z_T%R9bNNmju%CD$hx#oqy0-7+x6Vb#K~M6#^F-t1W&T0m{EGhgA3oWSe$rn2AJQkj
z)OXkV4*Ab|)<>WI)ff54eEwj(`Si;E<gkuw^XS7hWIyp~zIe0_KEx&Z_>o8X2l`MS
z0`q%?-WyJNVEqnbI!Da-#L08k_crLgO8$<O^CZ8=RjNk~pZ-r}{kyZ|AwSd~L3ZW0
z<Rmx$;+O0wkH`o5bL*Oio;*U|NuS=?$Mf_zeuIww*?o3(pI^xH;)Fk_2hMB6rTgjx
zI^q{S=PKTFp@VPw<tO+<H|4~a_0YpN{^(Ua$uIIa{>YK%Qt0ZBp!#FKVO@4{t^UcU
z`cHN@9v|%Gy%RlpU#<^;<g^aGK>h-)5BWQk*P#25T_F2F`>ov1y>dM|4e&>P;CzvP
zIUf~2{EB_}leiZz{1!duko*pcTPP0stvtp*oXe~K`Udq>pX1s*_xZ8s@}>AyC;2D8
zhUh`{Tt1ec#lL>VwLEP+`Q&kQ)i-^X`RWgT@ozl7#h-kSu4{a<D?Z(4XLeAR(mwhE
z^U>3{kc*u7*Vp0KzCoWwUV2iW)phxfAK;5Wu(!JGoPd2jH(&p5y!-ew&+kgGi+TLX
z`sBkWRFBjRc}+hl9_f)j)B|y8p7$v1?|UQT*@b?c2k2|Wsqc^ML-ZxS_eKZ*_#-#{
z;zOKS&$$Y_s(a2)ojdryVE9t6`47MJeo5X+l(+D~p7Iku64goa`922XgFeL}Kj5E!
z&)K<>?`nOg?3zD$k4TTM*~5KuunRwv7v%%~=6uUO#(jB<AJP+f*p>ar2R)b1#ECp3
zUXm{U$R*zNiI85<lb6u7K7OrZ9KV6&v_CYS{pELdmA}xl4$OJ_CH^HY#j}1%{PI`%
z)xOd^eWYvk)_pj-s{h+FOMk*|)CqQSe#Nf(botIa{%D?lMIM&-<PZLd9;A7GDIUa;
zdHAtDy7C`%pPcCOcYaN7c};#7x7HIEuB{^;)MNS)FV?ZHeh?Z*fB2D4`9He&6o2ZD
zICI_s={3<hiPr5F*(ni!@)morgMP>Q@{fK>UhrIg=a=+He&<;FIQDVg$$sq0pRA`p
z(U<W%h@QTK-N++f<+&k#(1r3O|HKzP@+<ucI_%1?tPAx^-W#wZzwv)I$ictZN8jQ9
z^!Qy2^zBa|`Q>4B<zaTgx4ghV=!+fLm7eUATsyC3PxeKJ9QN(j^&PK%Kp&Jy&n0B1
zM0rqOqmT3+Dv^Jy^YRKgq5Dvt;XnM-zJ|Zhqr4)|(vR<g^po_cF4Oy9{e7<Qe#n6?
zM5hh)w~73@4f$CclCur<9V12lrVo?f^^x)k`^sbD(Rlkrb&_A9%g^bBU$VdWlSj<c
zk6NGqSQnqx)pzE7^U;@|(7^{j6XhNL?wUT>k$ezc>p*p%p5$>z&P3y_i%;^|x0*+f
zP&}cdUg6jJ>`PCH_9gZi@`iljJ-$3;J#oig5PkNBnLo%$uQ2yp-V5Nze#_rO;)5Kn
z)pPuKU*$U-=M?0nSA5!!Ie!p8@;5!;TR)WkYQBA4_C2oItqs{d(ch7@q54`vago@T
zx762eG4ri`R=1e^m9F(ku#{Kz6WyY|A+f|q-dCrf_%C7FJFn3xVM*?8@8h4p^Jn=*
z-)o#cTb-9z<vV#z9_RPyyY^hZB_F;^dQgYNqx>Z=i1*PV`s9&k<Wc;Q!}}Tco&U)r
z_~57XM<4ux9_gXEu77t5ALMoo<!N?PCp?GrEw8y&?-JEf_F`YiFYqfb%Gdfd=bQRq
z{sVpABu~gU`h0xY578ev=vm$!En1)7%e(A{56DkE=NIO?PtWva-=*%@m#{x1hy9Yg
zBtGbg|H;4lXwUhne%pQvU3M@|9N?F{ke|EeZ}h<);s8DMhF;L8A9V2HIsf2i=+Kkz
zgO2{a`M>z1851jhMn2>D4}GCWpW@Sf`lEOI7SH($KJjBeYJL2a=+US7=&?WkN+|wI
zC@(<$8>Ih4?+x*nNRL@x>8B0(qw_p|0{uN1KZE4R^9Oz-ukthRtsy&z7yd!-sb|k~
zJUaZqc?CU@%X28+ptzwg*Zcum-!*>eF%jSVn!TX!HLOd2=;N=1^0WOVeV35@P&~1V
z?=RSyUy;*1a=MmpjhDB{kw`Dz7ulD2Z)D%0ZrYa^?|qOu=Y5d<k8Ayjyykt5enno9
zCyhtPyo3Ml%`Xn${+ep_x}C1r@MM3*PWlr0hCcc47l$5sv2V@T$~l|Av+=&q??o67
z#g%-(&#mjbL;fhw`M>ex@^{$kk#jZn;eY%KU-}00_=WQ}_V@R>^1pTO*!PB`=1=|Q
zklzC^4t?iz=J{O$b<z2zIQIS#zv!vo^g?gq&3*C3KgBmWtwT@NVNa+YK<CxzFZ_%@
z`>xLUi+wNuGGE=+$Jv+jSNEL<y5_&`i+_FN|FU=H@m~%5AOC*)R@x&=h%Avpc4a>w
zUBoZsx^!i^QK{_2B~%n$*|#D$%f*%KCDB6XqjGIgRNB)*rS;qXt)$=MK6rMoagTFm
z&dfP8pYQqS`})qAnfLoW?|ILA_W7`$$heVl8{;LulMlI&lXeI__|cB(FYp8Cg>fSG
ziJcLqUKn3c5A@gU|DiYZB(@vY7nnaGpE#aIpXiBnl$-Ss<iZc3584fSXMKouZtM}d
z>|@YgXvef8_H$?#=!f<VJ?xh86Le{p>}TP(vF~kuS9m`cdgLP=exCZKpTbV4PwW81
zZ=pxhArIpL{8<o^4tvF}NKgA<U6yeO<9X6iAJj8p@>4(Lqdm|Lw(0K|y-`o-B@CG-
zh9P!KIj~piop|J={s~i$=!5#BKVp19KTZF`J^=d(=o!1@y$BG0CnA0vd|=v@x6d9{
zoQsRJzaXSv!7m4)C<pHcgS;10x!$IEPqnU+gYLvDM>l;TS2fq9!<gDvw<^i;`!4kV
zj5B*I8}LZm(&8NXIluGH(^HTCbLHY(+0egrxaru6xij9Y*tPZ%+vR4@T>MbiH!J4g
z`@o-S*Xmuf^0x0cKmB$8n?Kk-H|(*!FB>?l>TmFl`QHtD^SNr<^7C#6{D#%-_T^8P
zRLTV*<zZaRxMWO8%MslgRN6+D_j`~7eIPgQDv5YUr(A>WPOrReyhFc^zJri{oAp!t
z8}>~*rrl!Kv={Q>r?F@JGyVua1wGcIp-VpcYuXup4ulW>1p16eSr_MiGxY<~{ut-b
zF6jsG$JjsV;E#VIp7ival#_mh@8}bM1LDunE8|MWbI`%BAV2$%jI#*Sk0CeZVm`_H
z1<1*GjdC!qzz@<7F-{=NI0E_TU!jLxQEtMFJE2Fpkr(=upYwK<3;odFL6>^L4`V;{
zpNuydAHxs5Qf~S&<YK&oe(~p&lkcqaF@B|<_)eI40e+kGBJ#c^{(^dB{7rrFz7cYO
z=!yD94!$$*MBmgOblJbAoTNiPQBmx-vhc4pfZxNvVCUE^c1=9~l5zgF`uXKuW9$RJ
z$hZZ>Z!tcEK6F9a1%3c|$j5wu`4;nW{4oBF@g(bQv`-K@k%M-Pmf%Bv{5$g_P+b3G
zT!x)8pJ6^gd&GY-o~PXOgN#e*x9IO!ufWgXPeIaCKK5nlH^p@neh-lMf_Mj;c@gEO
z-YGBrAoD-;&OC|!l70rgux^DMVThe!2h3mazr^D=(I@^0|3ZI(o#RjOzoZxa6@0K0
z<^|Xbi2kv2`d|3b|6qTtv%sHm9r9xDj8E~O%$Hd=p&Za>{6RlPJod)88G7gs`=Ect
z|3e3U&_Qne1AdzI6vmg(Lq6m|KgdJ-ARhVXKZB5VN_(SSVz11nK+=o$NqfV-(Fb-%
zKE@T;JARaOv<JRpH|U4@7ZHBCk#d5Z+n~JY3;#>I1}O(&?2_`Mcl-@>2{RvNK8`<t
zF8&r|Jpj9)eevG*HhcS}-$5S675EwY1<FPFsYmD_C+khDOQRR!p-VaFUkIZQ{2Ts2
z{7$(^$GI=)pg-Crev5G{@{k|>BM<dNKY%}Gos4#Yo%5ak4WvIs9_V7vjMr$_=%4jw
z{2567fy@h;A2F{E3PT5EpN#c6{33||Vq5}}4;0f84ny+8pY-B)!pMajgoziw6OUel
z;>C1geUKl-pOX&1OBi|}<A2sa=wGM@@`K1p7&_Pme2@cV9zwh4JARUU(4&2jj(Q+W
zKZG32r;wX|iF!kB@<&DHImk~xg}%`n{Tclz<pPnH^%>R&un**?|3MER<!64!xPx*q
z?qoblf5dnNKSeu7zof$tksd#Uf5s0%2R}}~2wnO|#!>KLJb>Kj1;pR6UdcEFME}e;
z;D`UiAJV_#*YLaa3#6ytz^~%}MZZaU%FDPMJ&5BD_@fsPy)v#Mj2t2#*0141zk-~E
zkstZ^jy_p0V_v~{pMHV<n|>R;a()IopkKf~>HjDP^-6s+&ZFLGkK}{jHu}H3GXWjs
zKtGHF=$D{F82d#3)IZ2~_<_g?G9RGbfsE4`#{?neg&v5U*gJY+oFa?(6Z{bE7`;$W
z%s1didnO*ek{|g<hurYPpJETROZ+<H82mE+0E8c5=z!Q0>Cg-EP;b~da)ac@9;kov
z(VpOsey}_I4fRGlKo87A36qZbCHe!QM|#>d>j0Dsy2wL+!hDx?Mb-(RN0@mtdPHvM
z<B!=NqWtI$yP-U+kAn11q(?5olovVibI?OS$b(-b9sZkohd$#1`ZekoI_MESV}G<`
z_)`w#!~T&AebB$4FTSH^kov+tNiWt5<A3Z9KMh^vgfDgr9m+*{Nl&@y*XS4E2Vxhr
zf69;Eu{ZdkKk5^?un)>j{m^cp!+3;sUhzBiKz(B8<YRnA7&@e5yh?w9pQ-Z8eJAoU
z9%9_b`aJ0w@30?=--a&ZIOZ$(y&%M%s2B8yebDc6?t}3h^DxF8^vCqy;yjggQGUOQ
zdZ#}nJ>vrSVkhunJOH9E>KXe$?|g?J`)rJ_C>Qj(p8|j8rQEaQ{1@+5Avg1Q?45kf
z+X++8*fsqD{(^BN`p4hEhw&El$WK2(IhZeVZUj9dCvp)F9sDfg81^5r3+Ph+=oPwz
z=`R?+FrGvIl!J9%<iSoD=W$L3IaoKQe%c;T@09l4iu34)bz;f`Kkfs8=t(@shdk6b
z@A>j>2KCRnG;-H^?&Pu8&8(0|&U4=_UHU++%K3ihZthuONyYr7FZkb7n_MZ6KF}BE
z1EE8CLHO}r#nn45Ib`0+#d*T4n^BLfTSK37U&QmB@)HL6j$GtJ9>%qdPpA*%Cm;2J
zol+0vBOZOw&ur^ohM&{qyFV`Skq7&w-O|peXWBje0myh4IYG(;(m#XS`m<lo-B50j
zaSBL#0kKaIKL<hwWSt3Qp95qa5PhIO<cAJ^5u}}xp8O#4faq03^e-a&VdTdiLF@`U
zpuX^L)I0IWOFe`53;K2RialXp^xODN{1xLg%1wIo2Y=|okM!7SRAfAhT|yVXM!y3e
z<fY!wCvsB%_($p+{o`*bC-Zy!F8QI4|3WVKus)4^(52kyiT<8^)C1!@=<uC-rCj7k
zAJC=$KyLUW5B#tz^hOx_#I6`8GyX$gj0@=3kemHZ_@YnxSNccB=cFeLUHW(Y339Lw
z12P@}u@}atjPJk?`wVP2<@2H(<3QFu8UKNd>zL;-&qdGF2lE#E40^<0aBhlm9sUhO
zZpw-Nsdwf9@FC26f_g$9&|y3ef5NO=;y=(g>yyZX{u!53e%2wG|Fe$CxS4$D5xvu%
z8CSxGdM6&eW0%w?<)U9A9ph>8Q!esh2lR{R4?hZ`U*v-id|7XyJf!10VfsDzz#si#
zPmEUyBR~BH{UYs!_D=a&mjTfSdL<q8NBe{?eqKc6Vf_n)F7qhLjh>-HIZ2P5Ko_~t
zEA>o0pg;5hQa{)i^w1alM7gMU^gujzNBvM=AoM67`bBQq4R%iXkqdi(KXjNs6AwQS
zd59-J?F__!ksic8kt+z{2O>9qh40ACcl<bhAH-g%Pmua1AAGPQ^h`NH_~PH713%Ug
zs7L66_&d(&g2amm9rTDC*dO+TenISsFm`~QP!15iQf};-`i4L8e1{K+{(=yDB|Txv
z#rg>Cj&?-7QC`|V>ETBhKS{qxdk671)Engy5xO9L4HVN8Mm`bY3sN4C`T(gn5PJaG
z-xiT_q7RUE3bKzNBIn8(hXf(x3dRSF^BBj{PVk?MA80q!Gv67HLXY+-(t#dyf)KkR
zKXwg5pZ3Ie#zB+=JH_9F(8qs@<5=WkzCk(B8{;qRn{h09K|b=KAI970iFFO^opB!h
zGyFl;<7g+yNqZw5d>BU}2lESY-35IS&peE<xb8rhc-C#8OMDoj&mcq(VMx6OA>(Th
z{YORotccWq5MmEOi2cSz`cv$T`X(&43*<p=+CTM<p0Fe4d(;!_OQaXqnaGblqi5tH
zjJ;x~q=!BTJ<`)&K<tcpHTDd`4?7|s@zA9`;m0X2=~&N1ANVtna)7iC=upqJSH=zW
zXVfEdpfBo)b`M>~1<Yr$f5s2=YmD0%PxGDrh<;=9{1uz8JAAXpxM9>QpDmpK+#Wf`
z`;?P<hcEt@bq@B2Sa-rcn7=c>L@&@s&(st0G5@E2@z>ZjdLxYgg%5h;JL$1E%E35~
zaRT|MPs)S;M-TA94p}Fo{jk1+JtGJ5VxP2I`aSdmqJQ{cAJ`G`&=nE8$9`y6{2m_P
zu~XLTSU(~jzN}xNPxQcd%1!x@n{|2QWq*-&f&F6#)H{4wkD{Go&)6O7OZYk3H~K)o
z&<9D+JehXMJ}Bb`%1?hxy%DB<*~ep@6#K!RSZ{<La-bLVkNoJDeF)kG@z^tA^uYWQ
z{ZKyI9py%^v{UMd`omt3llnwX_B)s-Fz=vV(I53j`x3w7r|?_YH{)&U1wY5QTpYKH
zVbYsI&JBY2QIPRI<9Os_oX@=i<f9$nN4EKa+PM!K9rO0K-_ynZXy2sA4hYjO2@}sa
z6hBS9igd99+9BiKZT-_P`!m=neh@pQJhTVe3-%6O>Ki+P4|YlzddNllCLMkZ`@#+>
zKkN6ryNF*v4(MQ~l#_JuB|Z8gJqTa=jUdEt(2s-Uhc99D1|lDbK0wODyoml4q#p+<
zA9@BwI`|#x9ebw!SYJB)@dID()MArIJmtY3aQ+2<&N&$|p7xJC$cbF&hcN8|Ik5-E
z^YDWY<0||O^LP-x$U!?skBlG4j~=K`(!mcq#g1r~*c17X3x6W!r(Doy9iH-$4?lu@
zAbtnL4}+wm+#qs;*dvJCAoUJnuORjZqIZyVAoYx0gXj(YQZBxu4-mgYyG4G!Q}5^(
z`OpvYVGq<hVb-Hqmq8v`q~0k9{Ve)GUiur-BR65}0{x%|^d}<a=U!S6Qa%xpGYHX7
z5Mtj!i2PBJ`VtYp5QNlc5Q=&ujDP1mSP%x$<vZ(yK}h|Fp}2lX7<&e><DmNH`>ab(
z`Eg?Fq8xS>hLk5RV$WcZ{e*qTKglBdpkPqFMExEnC+(K^uXrzn`%(i3SNY@gTyYM&
z#!nH?zBTnmz3;bgQL9I%|FmuX4<FJ~PcI+z#)3QhSITid=Fyw3zqZTi+vh}kLoUX9
zoO{H-vi^#nrM<z2dcZ&Oop)ETPxvA)?Vt3d1L>bh2hzVGAN?ErNymJHb^#ya`Of?W
zI`}=#6*9h|o|va$$M`w;^BsQ@72%Ivp=bDl^rOu4K<0ZO>og#Kjd?8NP5O8G4Ulqy
zqMx8XMMS^wBR@z#28#L7FNj`1^a5f}%!imCv97`SHIQ`7gPG?u?<bz`%$uRhe1vr(
zkn!t3M_jx0fbFX0S$Cp7h$l=wkac<ZvX04k7=1z)B%XXCAL<(<Ul3wvAnjX3`~(=(
zPvUPyBp>ZkMBabGPC(9`fV^uIgv?_={30m&d;BcOIsxzBaQ_j6F6r@4v}4|D0U2+x
zz5pE&k&E;o>%bs-2tw{PitP-3$PZEvoFh8^oJTkPuv1Bncch31yR6^&^TTFT%ZZ3z
z<sF(Z#D9v2orsA44nq1}P_)0OVag#Qex81lb(A1~Px~Ps_iX6*2@}scAYuF=NWUZ^
z=XXV9zKS1UK8|1EJNpBy`!K&{|B3#GeuH&L_FJHXKFH6y1L@$){D}UDbnv79VtpB8
zeTa2$)<xkjmIM0mhc0xOZ=)CJlAip?g?^}i^Z{Mg(-=1q7S}z{C;EU7VdzmF=#dZP
zJ95AWzsY$W?1lAw#!=WO>(L<nE9W-pcd<{=uRx!1BVop4jH~Eh7%vbHKjgvwi6=}v
zauUX#C_l*fgLJev#uey^^yr;>K!4~PJ(G`i4<GbP`-U&&K#w5fE5e+267^0z`ltRV
zH~s;8rW~Y49`s20(L4T^a#DWsQ9kUR@gjPo{n3um8~Yio3uCYN2mA%?ig=Lxl%H`r
z^)K2XVNpKhr(BeS^0Gc7`p@zfMjxyfVn^tQc=U-Mr9Sa@*dczEb^~3)Aat=y>Ye(+
zzQ~Wigg)`u4f*h&#Pgl@Lis2MVf;Pv!54d_obV?exj^VZmwebK@?ig@LvHv(kMz)`
z9%y$U^J&^4{^IW=d)@!)fz`IHKcH{?6#MM>1^k=sh(8whgOQtf?1TDZ{6H8zW54Jb
ze?flu@E!j|eV}jrFa06@9y`E~vF}7W(O$6^v7JyZ&PgB#VfHN<x6`gTS3tR`Z}?L#
z{3-Wx**AwT_QJXi{XgYmoX-3gdt<&$dh|yBhn}EMza`GI&@cTL@#um63%=}kVHfnD
z;{2KQH|8JA8z?u3eyAV%C+r40!cQ=tq#mg!<Y#`)`VRAa%E3Myc8b3so_35srhJUI
zNYD8K<_C;7_>MiI7yKo9#IL{yI|6AR(4+rgT+KX}dZ#@2JN)4`y&ZeVJ<k`<^;tJz
zJVm*mUi18x);E@U)EoYs^%3@w@!yn-cY5%v$c^37ZYV$FN9-GU(F<~+AKEeYi@vZI
z@?o#&gK-t@0DD6&#?i!6@5E!@q{AO!57-}aP`;$cJG3hhKZ3tu9fNiSQcl)e=<m4a
z&bkBZBh)i|K)z#-AoG6epLWZ4_=3oZT+DZP4}p4My@mB9+9m!M|Ak#pZy@zWdy+-;
z2R{&gAbw3m><s&YKl;SZ39}EuIvMx4s4v!OxVORn2#9>>kM<6K{Gq5H+A)3%#4pnh
z$%p@;zL>|s=g<qQA3f@)>N)752at5|!LQ?2s5idD7ym{WebP>$$9REuihm&={U~zd
zFX%t%xADjL8RSPF$cg{KPY{M5<IHXS#xM8K;g5Xu=d5F*U*y1l(OwvTh)BDpKgQp~
zhjxT~ln*+@V?XEx`I(<$|JX6|<DZd}bzJ1(_q*{sygMTD#oy3Ri1Oh-SO*7@Yny!(
z_3QTVA(K4X1@*@Gl=`IokPl=Wgg&Ao{xK>t9>AWF6MMo=8E+yt=ZINnWZa7V)9*r`
zb^;&TIp5I>-x*&p9-ti(k3RtUj{N8my7(dDX;;`6@7yrX1F1j86{HvaE`FQu*e7;@
z{ZfzEKkb9@3h~%8egk^w7k!{V{2TGKbCD17QGU{a)DQ8rSM-KH@tcf;sVByD#5117
zPf-up74-0b@S#4*&$yKRKKB2pXVSr+`4e`@{DyScC1LEC@gVZkpU^Mh52+WvQ$P3%
z<fUFHH+m53jd=w8828i8kQaS%{uw^xXWUFV_>R3$Pn>6k5AB+EkDTb2{s6k@3%RH-
z+9L=Z_%ffTzeGOLBPV?EBiIG{Ax!(ho{$SWCqAfL<O473U**R(ch$;^NWIgZ(HHfL
z9ib1-`%^yT<-Hs9PI<6f_Pfc)d=opMy)n-sJ@Y8S@MpgS{c>JeED!x4^@Sgzy!b)-
zMZTj~>I*+YJyY-44Ro=8^1&BGpTr{{<8%BJ{tbUcI{H=Y06#`N^M3Ntu0+JIhaqx-
zj6cGVekKfq_R)j(_xa9wpD-jn@`JQL>VtJp+9~yqK1IZ?@xS<0{4@Rv`q(k{iXGD)
zXh$N7^+p(eL5RJDA@&x8_=7OSUcoTF#3w!P&-kUI0p=GX(*LraihVP#z%ChQVqdfq
z;`vVdqQ7Ds!FSp{{W@Xd$%h}uztZ0lreCF>1IbT0;Da2LgZ$8Ay@!5*eh51Fd+t?}
zkMT4}yP_Wi(G%kn+B5PoFM}@qKlvFCG7m*Q#<z^m`3{ng@gwD+zl1*X1nhwR3w`02
zkqi2i8>C+b=@&utgI<~UfY<>DzcA#yANfG~0g(L%Q1su7FF@#kVn2kvfuv)eNjW)3
zNIk4Q`Sd&M4cabGn0^es(!Vk80%>QoKgJ_`r<~9O85bfS<2~e#iqu~aeqNzjlTBMo
z@*w<>mwKfA(LW=nh|DLD7sPKc{v(XOX<w9^dSiSGAKE2h5W46S#CzjMX;0J}<6!I`
zWF99XdLh4v#HWS$qab9xHw<b2L5M#xg<<1hG5*<0y*8yax8L@AqC54vrmcTtjU4AX
z|L=%bs=aq_^&H50AnlOf7iD~Xcc16)TH9;;ZTs)o5BpZ!3uAwP{d&%AGLD8G_lWpC
zQ^w(h`8`n1X|f(qI@YyGU;TuOn&0zioo(fXf7u}iZK(HXl^pAS7d*8`pIHNIZ~Glp
z_zc=-eU*tVYUR+=*o}|R`TXmexzhF@Tz!A#9k<aX%)J!$Y1k*=K7@$KgPuU~yej8b
zsSoxYSU2W7>$U7VF%M%N$~$;fetCbI-%VtF7dyr7>A&bV**Ag@<pl8?tpCw2STCgC
zh7aR3{2cy`F#ZAl&<{fTm#7Fo#z%zluVOzchVjcmi2Okq)ZZ|E5s`id{~m<+HO5a-
zk?~Gcq`ixXy#*n1)9--Tk%;)GAf!FWBI}~`kJLLz|4Ki}`V#%9Xb;d~{!MzmGd|<}
zQ|?R9zcLQz-URd+r?St8UXhRX1TyZ%4~YFUVeFprz=!b(<z}4_JEooRPB!Zfq@#Su
z2ST5D3--@=3Hr2m`Y+bg7)P*fhWz*o=1YW`mw@nPTu3_T@Ey6(3v%(D_61$?K@a_~
z{zDl4=mYv7?T-2t5r2*z@PniyKl|>4nFrF36Hh+qqG$Xz<5~O%{)qe{Vt;&Rf0TVf
z`bV*UfL~OEKFE9_2=PzM|CygMAEJJO5c|MRgU;{M4zL5-Fa7|;j)<rJ>9<Hn`(u8<
zd<j3ocaZ%a<RBe%2!o^}e^5I76aI{JeCPKacrTyzcjmA7HNwnSxW5Kn{0YCufFI-@
z9P6D$*FSxK-)>uc{3`S*5B_W0{IuzflN!FT!Y4oLFxWT!5q^dC#kwx-i0{}rNc+J4
z@L$+7@qC9K{tdrQdh8YcoYTRdGrp$3V;l`KKI3;pNY6c4#*5Hn9L#t81^$5km~_aA
z+{i~5ze4*Yp7nj`z=!cB`AAQgbmT{G_$~C#`6a$nKGGvM>6wSX4}CHYrye-(ft-9t
zFNEO>KjuBmPoRUIgOG9uA>~9~#zoi*<)&OBQV!+;B2qsf{Gf~8n6EYIFsAm^tx9s}
zjeC7zh+RS#IVlhN6_N1^c7c4@8%R9mARqlE_JG|&2Y%Ei?FjiU{QCG!eO7Mqz!f))
zd-18UTepSLFNnQu%fI2WJ1cEk;^X&0{6fDYe;E<?y{T`z&JA1c*|L7i;(Pj@+qY$f
zfB)@VyK|k+JGU(A-M4eye}@~jEN)qR*5*!s>)ZLcdz(-5N*GTY#~Njq-Eq$JQcoA7
z(_d-UpzF62y`-?$wAN#G>9E)<G@RCRQuW;qn&X*9UG=U?r~9k{e(wY3`$4Em*Oq(>
zw*GbK(!;*-qGH-|>u$HQ<%y~%Thn!K{mviS_L}KM#i;t}-0`Gkl|TH%PYT<%+qmM%
zEpz;|FsfeSrn_dZM&mxMzrt6=xctAq`+`Xiy;nJx79#(NS6*CWV2$l^vUqo!W2;Z8
zRWau(zWLDMPYvu)l1m$@2itK%$<6motg+M+(RR5r`SfT&bv&wv`){f2|I+IHtM2i!
zX9~l{8Avpv=dMS*SZ9H!ipMUia`tJX=Xtsj`7+e20o8g=JKqDXH(v3>6?45X)T^!Z
zsD1Rw4@b8CVzDoaX-T|sT;*FXcxs3*i@N3g=h^S{>-X7*URtQTJhsr=du^2;Yu~)o
zS4CUOfgEePZS8tR*Bbc@?a~AFBVtnZB=VIFM;&L%(k1;-Z#3I<@u=^;q!IqA*k^L<
zrHA*K<i$l*{%Q3G8Q)*`$Fk!CThov1J!kTWXJ7YR#iZ6_;iKAavwXj=0aN|J?{m#!
zZPCDihxR-7zG2HfRg7DX`E46r_|fzMetE|rJ$C4F*OT4}@B6Oc+qH`>e*C(zzH8Wc
zShp_+UpUz>RJ?NTfG4gjnd+ND(S8r?^4g2@Z(ra?#T);!vF^})miV?~TKmFbd`11I
z6+Uc5pAp9#yw)>?X_ZTrude+W*XMz0_Xn!=r<yPBIz`<4u8bE%IfLfws`ZfNOGX2a
zz1FwYo9!R;$~y*W`@HYoTC?{Z8$4IB`)-4JuAaQ!lf@u?r1hQkBv(<_JUK1>XM8ga
z=+HpF9Va~Imu&T2$E)UiQ*p%`TYXn>MR)(rc5iL*UB_q7yYBc}2N&hzq7GZmXt<++
zs>j`U)BZh+@~&ZAJJ0g#@fW)EXu8Ghdu_*oHxB>7lSR{VrA43nqPFAFO)jjt@1dn$
zq2Y)JZ)pEUlhvLn46>`P&wO^()B7#-)5h7SJutJ;q`AH+G_@}jM0diymsH*`<SSnm
zgX&55_koYzaNJ{6H~6k%w~KFmJwJ4V?>bKS*A<_%Z@b=i4KI1A!M{FiU+QO==^>6C
z$ex!kw0=LS^62hAqS6!hzX~-RRUfvclf6HsD}B;>_+!Hl7q4r$UEVdUbMvqh2Om&5
z9~E`!HE5qY?7J@hA|2iHB9VU7aM=CIp!fy{^}GG0>C-&dQTIJ&+vtVe<Hb(G?vL9V
zZ=1fu<cWG;%=UCae!=$dy8IRGCc~dKpi8f=(Fv-5+t-~x828Vni>j{llE!XN^}e>r
zWy`#@F{s{5eeZR_ty5-BUG2GwFQ5I7+T#vf@41R$>k_8w-4xyYy}Qo6`m<$zR5T@r
zEM3|C1KsIm<+U}v>33wq#<5Z3P0t&mzF#Su&enQIt|0x}KK>4q&$aQNX?dydv{2W5
zD3OkAc}(kBH~pyhzhw20RQlH@Hh<{$8f$#lFzs_6y7<=|)@)G28&~>qF)4cy`M7#7
zd#`3Q8qPUymLC^IeMAjM9nWX!<mNxN=qsunWa%;usAxd<d^w2To*#7Uc3Yh<{kW)7
z#w^c&G~oJq!8PwcQsK<EH~OyP^?MGwVCvw_eunAl;j?@C-*nr>n|xQ%^>%#b0X^%l
zyrj%m#s7G;bpMOnpV^vUoX;es<+S(tGX8e#w`TR=Rz0Y`FDaWZDL*9gQQgND=_DN%
z{dn48UFDa3Z%h|o*XSH|{Xxf+)?DSMjjriUlsD;dn#f;wIO+1l<rjAD-j(<{E0<O~
zW%Y0`EsSeNuD+`kB)6>H#MPf|`9DA9<1=3GzR`09&*)h5%(HLZ>Sg$AJ)~t<tzMe*
zT(fSQJy5qCX^&Sk{{MS_JM6t`+vB6YTW)K*xb3V^-z|_mPlF!Q8jnRi_eDD7wFPzm
zzL72EN=l9yn-6=p+co1o*D&myc2fF-&rN&X^2Q0@6y;sVSru<z_@{%4^GPB4%g{~@
zsP2#I&WCow?uQ4(Pbj(hzKJ!KdT|jsY{j_cvn9Xj`$DGZPj22~&WPHJyrj@pJIv&A
zTLZes)j{-hYboeGNZIdk-@(*fE>(KE?+GPM@1DAib}8!gotG5a{*H+#pQ>Hx&Zp~l
zLsaR+Esv>ro9n-0VB7CT#I5I~@(1;ILEn>-OXTC~{(`CddZK<*!%6p(w(~pvm1YgP
zeml`~4aNS})o|GFJ*wKdYQCU*TB_ew<uIL3_I!tGdf9S3Uwgu+83%0gGPF|<Vdr6F
z>pSkdX|nuWO&|B1V0p>E-Nt=}ow0p|TzSW+cCPBzRr5uaD@*skH4xWd2Ic?X+`_*W
z>bw0x<qFqGx+%iW9}u6el)vRJzH60q__q>eUw%&m)}AwdrOBla*DlGO;TYcC=Gf{}
zYE{fd#ow!&G~?Pd=-HxLuQi@&G+jnl*Hc`VTlRfX(wnx2F#4+dLt?%#K1s)`j`MBF
zCwm`OSNgd1Wc$1K$Txq?6{~vOwaSylxcZ5jKkj*|Fg~*9>TON$`ufoRN6u)qvFjpV
z7FB<@EX*!U$A_JVg#TeH`iwZ{;I&><oPAgSXZqc_&I`k^-vLNFUiSTrsOizGZj?O-
zW@~!YeKPXt)@w%Z*BS`>y*$(NHdpBXT5I8<p!5B<e7EHsMz|*9rLb-O8?@doerF!*
zDhAnkyxJ@(QvIE$xO~O@s^b}vj;mqu9;U11NK0-}e_8l{-|sbDKeF!@m^#mC3td;|
z{kG13WaZ-hf<nQ#?N@f33csNJdR^beb^5f!lv~7oukASS#^FDBu3*?YR#3cH?kwz<
z24vrj4cnJ7eP3kScBAg`CVuDJg575g@OvLH-!Bv#SAEo9_IrK0?<%r>nc*dsHw^j8
zH?0TP?LO{3Zq@ws5~}x>gVuS%z902QvrQL|`rZpeQ7%{gL0q}1CtESB9#!MR%A5AM
z+Z2CW&dWrV!}R&wFgu7kuL~PDit%B5lZrp@ySLWtea8mRRV=jod);>%)N}RZ^`2{3
zdBh_Jb^2zlX9~m4iI|F4-AC5ler;LDMGoEN3!(?V47+_f_`=D4R6Qi6Pt)rmNzsq0
zS6kC1r6*H-Y=5u3yvP@KKABd2+4Hzj(+Bxc@%xQ`*;sezK1=+xQItzI?CQB7*X0gd
z*Rno7;dj&@d#!J)H`_nxW%#omT;JEXtsZ6lrmpn+U%zmNjtzhGY{9NGzG+hR>rI}k
z$nPS#iXT;4zi!NJTRhj%7JHVJBkubHy8MDjNA~=ZE$P!fH!tc@46EMLOFEw|=R=~(
zam`+h#(i3Ug)fWwdv~3C^=Hd`SrqG4SJ;(xd0pjAicZk)gX;R8R5>y}$!ehD@khQ|
zfA}gtDKuTW2CMr#vS*czzHT%vcg<nV1~t5KrJpngl`CmE(&iVW2i@P}zH^b~Z#`z0
z4vW3C^q|`w;?fKHJpk9fhrKVZD}H9B<M%kL-D=-eJnH&`jw!9V$}evi)J~3FR^{x|
zM$hxYP<8!ZJm(rWta{$kmV8m)U$Hga1Fbh+@xv8!JzFuz&UJnFUVG#7p4}_ulfnxx
zf9!;B8<*r=!=&^h^3he#X^*F6`R~29%8#{gUh11d&J~zCe<RYh_5HXoxsqOog`Y0F
zhfdV<465S-<M|Rk51{Lv3*Iev9Zlbx&`m#V{2Mo(c9o%B8ZgygCPmjSZOTg%wV%n-
zz4M2*y=Ho0dY}fnU3}~7`Jo&9LPg4%VWxpJHSp|t*BxK$;G(=~6wAoCUlm<@SKzgY
z%^$kG#u{H0dk*-&qF3Kt<7cR*hor_cY4MRg7Zf!;?e@k2m1gWU;U}-Wpl$b}@9kf$
z?dTnfa&htLiCa3Y+V3YXDRlMRGjglmU$!Nm>btzQ<a7011ySy(_vJ{JRzH#!UtQ-E
zP1A{bUf0%ivgd$pOP`i~>f$T=uCZ<D)1G(9zS~QB)pyv^&L_Tm6c*O??iJ(y4DHf@
z>-%c1%YXVS%^Gz5cB1Dx>N-zp8y#Ksa^?X&>#w||%-4;&?<FKnuf^iGKC5)mkG|{p
zk8}UAa8TvVzN<Ly^pE#`w9QX`(x}=Ss;=`Vm9O5gzdicYffe#`aqxs*7xbD^oKFi?
z>qj<U)b(Rq)3MHCJIdBe*!;lO_^9vM+MX`zd$+cyGrf+cN?*1;kUs3)L0$9awCJen
zQ_L6CzR+V@m{xykE8n#0A&9Sd{@u24kX^g_U045M>b+^xban5q2HCZ%-^1P!HNCEF
zN^horuO_MarY=6*YjPcR=_6@6uD@%X)c&q1e*oPTcRW<`$>a)oS25}Pv8ME7svJ!Y
z`}><UCzs@-;+X$?;;-v_l;mX*d$I-N?zgAypW@1&<u9z?kBnNMy5^6zv|~{YUHg>K
zL9cPgfx6bi#qx%Q<DRdQeQ&_F^t$w(Hl3u#lcxBjtv^-%s`H+(b{RKbwLG@w)73Af
zMJH?@#Fh9xKj_x&wmM(>uAplBv?ZVEb+@GGhpmgM#v6YRrsu9lyjW*}7ne_5e;$@U
z?mHv4<d6D3T$nzRj%PmLD$ZHCwAv}FhkF_RtcNiBvF-e=tUR{-u10y4L$<!+rVq2{
zuz1&gr_0sz1+sFdHJ?%C>&m<#NKWzGR@87%`%=}T>3l)*#{F*k{DUd_w%#i)FLLR&
z_d*-zx;F37{SIeFZ}t9F_xRY$^pNQxNjr!;9#GA1>UU^V<sskodk(r_>fp_Op<&&j
z@11Z|r7gbgxaR#wDxCTDM&DI*{oPgF_qJW9hhMN2d8Z{UOlur&+kMQS^#)hJKlWPR
zR&Ta{&<jG{?IG-aYggi<?tc|px}e|J)@>)U^mOlw29>kCe3#XKQt6Lwa$(JV4=weg
zqHVvkHlgI^`zF>{>bZ`(^s38l)^)Gde@e%qT-q47Jo{eTao~-^fADl+QooB>Ui@Oz
za_V}=Ev`M0-}HUsr07TW`(f$OOAvlA?w?H;RbA=n#vp##_wu2Exc4}s=3jZ-io5zw
z{LynA)9%l8@xQcvov)USD$3c0x?9~xmDaBtbK4frbv&bE%`?xwb*tw(HaleZKh*13
zoO2aj|2>uNyAA5Odh&YDRs8&vkI#6$`$o@oH0_UM{iti_Kty>}_XA}0q?%rK{-w%?
ze75}_QkXogUYhe<vu>L`S23)7w#S>&JNBrGhpp%{;+TWida4-2SJ&@ni*$ncCH=kh
zq!UkB*1lrS6eeAsr1{D2-^luR(z||-#q|EDE&7RjkH(h#uI&HHwm;j`M_s2^O(*)t
zuFrgS)zkYe^mU`iC+YBj9qV^HVpMU?HMD*F7gfHvzh|0Oe%1MQ()OM4%QTQ_Ak#pm
zflLFL1~Ls~8pt$|X&}=;rh!ZYnFcZqWEv=J4XB=vH>H)dejhK4FYl+jiecqTD&Cac
z=%PEa_ngTio_*bO6_dU%tV{1<bX=_u<QC6uxEfCC-JUReFdc9DoPloo*qJFbt#9b-
z`rXQ;=_J*En&RWyyQx9)s@8kO<Bxo`{_s_P5N_RkS?i;Rt@B;S3p$?LW!j}{eOJ-d
zcZVkY>xxg>w_Wcih4@nu8y$Y(ca^KJ^IgZ+CN_WQ_8Mz^SrqkQDxCB<9Db(ClhG~T
z8qjqvHfb%&zEduyxBdKT()x6jUsCTgh<setS0;C+f&bP3<BJT7E`I#Fv3^uN{I?EO
z{ulcrSHn?%f5O&uQR_QP_v`i{+PrPQ+bPNyHyro(_iV{;>fLhPbnmI#XqTc+-+8(*
zt{u3N|C-smOjvt!t-LCxT~5{W3%2CbZMSLDv;AJAD4#8MAS;J0_j}{Yp?e=RX?nWa
zTU<XXzE9&?SoI#VE%}ntx2l~a^*%rH#r<8A2CMr#vS*czzATD-kq1=${ciF>M-_Fg
zqn9^2s`?i5slMxDYrf*!{Hfh1m3f7N;`+I5;jnj=U5W4ZRPSq>T(->9jbeLqH5~W%
z33Qcva`P5*M$}&9rG;VTNNc?_Os=T$^y_J3RC$BasoHB0y+Zw-mc7R1JK^3-DsLF_
zmG3GhHUBWhC+XjvGu8fr&drM7RnL(J{f=7DcU3;3yr#lQ-FI-EkF34Qu6Kr|hrjIk
z__X?c<W)se`{c6nL_J45v(oW<oYii%?+U8E?`=yyTh>KY<#6?WoGtYqR}NS8kK9SE
z$Lr>k^!}A@eqwtU`;)l8XQKM9rkF46cWt7+PZ~8{7{9Q1QIBC^u^!S2hskLwK4`v?
z^!KRc(|yk%Dqr3A0Hg9%O(*MbRr%PSuWh@HE1ujk$CpK2<<kATkQu#S|K6sqdeTkD
z)%S0Z`@k-*y*U5&1-`2ocO5w_zpA|?-LH!LRONOppRW5KHHS4D)bPfYepC#47a{xp
z`|m~+ihgeV%f`Ax_gUf>8oJsKxFWZ1|JrlF{}sLZ?i$}UG+kf1_c3JW6}r>M-EU>Q
zrwg%TSLP`P9@_8R`-Uy|bR&9j4fnma<G>q-|KQn%rp|S_RzJuq-p?u2u&MJ3L3CBm
zS%l@I-gKd^`!uR_(po1s#n+U7Ns4aRxertEy6icl^ZOczdw($L{Ho`)eqRge$LhY9
zUnul+)lAsBt8MYJ=jd!p&+lEUqVDq8@~(%f9J=?XlcuLyZd>x1)|W0nSEzJU{Ycb&
z=q;&nbyPmO*N>8>XWM#DRC%J-kFIpspDZ>W*6qu|7f$wdqpTdZrPqC~Qk9;nUR38f
zw&oMh3%DBArC(LMh?-ATzd`hE8HWYQ5wz|a_Feb=`;4BMcMYQIHOPL$zJK(|4@b8C
zVzKWk?$vBY!#U^8@?Aw&^^V-O=snENb;s*or%9Tg?0G)j`U|3``a4Rt<coX%O*Maz
zK23cO8(+xA>z0%Ic&0H-Zd30<sL~CSS2q5Vmm2)*!}g`VYv{^-@E|#pI)^B052o9j
zZu+|BwQ18qFS7Nenm+0AO2)6zg&$2H;G+lCdUG`&_U#(F{$8r?bC7ZEZ`yXF?(rsm
z=f}mU_vcNgdwSxQPOJ9&$#WgWcMe<&cb)M~ld4~D@-j?U5BK)3)^_v`MY-~Zvh(Gn
z($7Baftiga&GlW!q~EWMdyk0xrtfc=rcXIsMb&i{Tk@HDuP-UOvil3-_)^#Xz@T#`
zvVKw456PB4NKV=Bvh^GF9;jG9vf-r9cgxBrn?CBftn!}j%maGXUwKKH?+X6o+<z<_
zRC%*+JIdNC>Eq77qvnU7h)KN%A@a?_Y8p7|`h$)st+~o~6_b8X#kSw8MV_eb*0yxI
z$Kz_P;J3MS%LP5%?<)q;+k0)5A8X&d)OQt4&wo|v|JG8>eCc-at*_^YZtz`2-hmgj
zlfQS@xmSO-%ugFdxlD&;=WVv7PikE=<MU3xexH5lrKJbc?O^bPUKjM5Qk;*AVRE{*
zAC0`a-ls~N4(B0~#xT2(jaR+55|l6Ke6Orrs_9L=E0fWkRDHLD=6GRxh<g9m)^uUz
zvpwF_`#@3Ys<vBO^PRD1W#1b6RnMEou=2O9IH2X!hquosg<<j#FS}1<I(^iAF4=u4
z+tbVL2d14~b$ues=k|O4<*okklOoE_e}l%`;&)y50fOG&iTZmq%=c7L^_{z@@<mN2
z*4u~&Z)pEUlhs~Q*z1B@r_7wX+B1c!zoTt?zNq)tO{beb=89E4?po!Ap=mip`l4MY
z9oFTi3XKl+VC(!is$8mm-<EuZVpp=`jI`5Z|E5s&U8%J5MZM2;bdw8f?t5sdmlm2D
zr`SffO@p2-s`Xmqxq_8PJaSN{Z`OK+f=P|PP4NjE5857IqwKOf&Y527nZ~I7fUD<#
zM7e{`uSL~^Y&u=@I9a~B(ue(CM;1S4@`z_&_e|>{?QvsV{)IB$S6w$tJD)4(D1!X6
z?cdYh@6p9S>i%9>x~|WBcGc7SE%a?caU5VetgHVO=_GaU$hP}fq8@Ay>pnlHtGuSi
zDW>Uj4?-8B*P!><g1(#fA4$=FZDR9>Zm+S%PYcn%DU7<FZfm+i88@e04pa6OM%Pxm
ziYiyoIR)FltNvc6E%~DEN0ireVeQVfc-41w#B$1>qm;cTCi}aBVe>>WUe~w|I<oo*
zx?d6XyX-oPto}(aTMt3$Mf(X057@DOw<AUsJO7T`?6Hjx=sL1OPB(_>PmB+H?^#T*
zJ8XJf7FS-?b%?lpqvqH3duh;#%D23w!|%C{;(6kz;i&zOYPzU$Wa;|0db9n5UZ#g%
z>p}JRO>E0&`uAqy`fb<kPS^TyTs@|pzr4MJoRv4(_pg(-Z`pZvQt6|<Um7<Zznhmf
z%D%%FW+zGQH=E)!?(~oMezeU`zN>iEoNp?wcw?*YDqg?mpbMrB-t4=Mb%(xp!cmpB
z_+c2<p2Fhep0l$pzpn9KQgmD&U#ES}P#1sG>yjD$)%#c7<6}>k9>V;ZZSkhxNy+HD
z{rlav)t_zcC#pRCcAW5-U$WIt8eJVPBX`x~ZoFy#o<(_Ak#fYvsCJfAx~T88#Z70r
zJV~v;=;C8aAGW_YXq$Yd>cKU-uJ1#O^1D9om6hMLy}5oa)0BR)_vzwkR`q-b`QmZ^
zPwsWyYxSSfu_$LNy1w6d#w$%OeYkc>&Q-j$eVwnCjVj6|jsLMwz5j8ZIKyt;ZmaX9
zZ%Pk(Yb}@>{;<$7XkR;6GjcJe-jfrhNNWE_<Ri)v7FNB-VM{*Q`FfB(Oug$AMmOo-
z^PBLmD?Vx8cD?U9es)j)n{K;!lOKdZ^=#XBSI0}D+`8|jBu%f`A-n&fUdQ5GhGzAk
z+Wu7g?Lv9q(3WyUt-rYMiA3d-r7LU=n0iMmDXr;VPq=8mle?FUsF({wQ{@YqpV{_Z
z*ZX1_od;$%nl#s!)kE8M8&^EJWsa|krtCK=U0eEF-Ww42zg!C^<sYKfuWY(~ukASS
z#^FDBvM9?Zsq{hbD};TQmCyEftHb0o{hoDF^bb6=-?{e<Tkge0+4|R=-qi0d*z%5&
zZaHT3-1UeT>n!jxG^+>R!-$Kz+sm5|9sbn74kbC=s9VmAUZ#Qn)_|+$h+UKWzjZG2
z|FO%eoPFBpd7dnW@vwbgYT%<c9QRn&4St5v^ssgFWv!1Mw$6`=VfLXKzyFalT5asQ
z$k&BQy>|?s8~?Jg?$CXf_^zYt=P8rYzwPhUs>-L@ezSa;1~Luk(}1b-kf!Of9&Q@-
zX}Yi|g{=>pir01C&~$Btttab_7t0g(9%tJ5<F22jonQQ2^|ZsHeTZRQ?aEa;qCRB9
zvg;qR>1E~7l|Jm9@3iCNzQZb>o0km-ojcCH|Ngn`Oh1`^e(&7AE%u?ykIK@Ky%!!e
zeNyLdGCrgJvft~|{V+X5t+$|bvUXJ{>ATMOrb*SWH+gZfQ1oG{9PB&9MN{%+bTbXO
zpn;Xgt+=c2#2>vv$1^(CJoD^Zw|c5*sxChm_s^z_s;=}*qbd0dm2S|u#<lNp&*6vF
zuj>7Yr1P;a9yab3<86DlwCgjUUG?;S3w_<_`nd;L`D+eqHmKo^D}7ar`g^}o)9J2H
zS$bjZJgNA&<N37n>pH)cG@YdCcS6a{_f4#^)UzFB+mEUAN!5!iAJgO2r0B=JZ<OVa
zcW*20{czp<RnHx&=96u=LFrB1TeFRBQ2AxQ>wY&OEqc$_o-k_00h_$I7*&6zC!Ki8
zvi22oNu%!jWzc(K@|lg!7+xu-3!~~$On1;*?_J)iahaDkigKxjRsA9PTz@Y*uKc~G
zwH~uehsB<2DEoea?ddtM7_}Xo_uX4-_P%3-=PDlE<ieW!9$M<@#;E$!l`ii4U}@)1
zdVN(lKiT(uY)fzY{J&}XvhUW~mfp5?dt2nu<sWUQBl}K?Y<Z&k8(F%t`xv^?yLvu8
zOm0)_vbyN%?l;5eF+Z{$lX`c`G@rQhxU}-S{vMO;Jr-T*W&ORa>7(vj+n%n__Ge_r
zHMXTs%b&^eZPTD<i)y{rc&cb?yOx#fuoZnq9CPqm&owNRcF%omvEFR?9aB***>K$P
zmn=Uqy)E<dwB-=>Vp}+Do@Z;k>2*5O^le+$Q<X=Tp2FJ2QP&@IOli$ke$p5wmuh_e
z-d*Qj{n;|#RlMY-2LJl7eW~v%rhT3@E`QbY)wbkIx*eP17q=eEOa3rF8TVa}wDZqz
z+vviNrVsFgFsS^x&q>Cm=i0dzQC`#Y)1Z2^?Ypk`_;mTFw9jqG@;CMSV@c6X`ksp^
zex~|u({x=O*S|Kg`9rtYSmV2nrtDFbuI%pvJbT`C$JaWzD4(Gldg!;~gvb1nt-dPS
zHhu_`=Yo#scA0kRTHiDxN7QpcuBX$ro)(o3`b`RLKOYlSzO?6?aqCO<ceQQH7xx{G
zwDNBpSNWC;o*LpOg&n%w^`v*g`@U{;?cEmS4O*vh<-4i()(eepTt7+vu=Af$<CETx
zPWn8<>8~_v(DmDio-W+~rPcda-Q!~~LpeQQw`pV4@l>IvOI!aL|4aj!1~LuUrhz+l
z@BH4<>Q!<X%IU$jdd%d>G>~baa5bQNJs|2lIc_>vYlQc7Gfeh(kHhRG?(a^7<qzXu
z=<%7oCu{GbzSI6)H~5Qk$%budCqZ(XmQQ?VJqt(sU8YP=g`p?*_gqES?J&si>Hcoo
zk7o3L)<C}_f6+wWoBFov+^}W+mV31<zNhcGeOp%e_utO7JJ;#FbIYRMeLL6vceqi@
z;+Dl{ZSM59zMY@DxA`=$Br!cc(5PFT^{1EQbTPbl^uF6)KKN}vDJ&T|dd8l`oBTq<
z!AF1DZ}P}3zG>7|@2Yfa`cp4|u*NVi2vzCYl5bA;BNiOD+Z-<{rY-mD5AOY^H7_ss
zb)jr|s;#JW$7A<R^IgSPe|mWT>spNUqheJ3ocd6;ml|KQ-b)I*&KY{@PmL#gX<<~o
z#7$S{q{}LvKX|dPigEdu&Hi-F#~n6%X(95f^6y_~^ABByuJ&BT>mRxxe|*L7ytI*e
zupNW+WV_s%e0nsX+Ar(j{#z<zch8+TZ--Lf6o&N+NOVNIk;m-t+6GS*UpuK~>-+z@
z%F~U=m!Vz_+`aJC?ia3F<GF@YnoPZ<>zd_W80ytldelBz@x+`z)+@@(Vp<Y^vaqcG
z)qUo9vZ!0$1{DXr(R^MPFD=wv9$V-QKdt#gi%yv5tD-ICK#qx*)vvI;+mF7hIBeRr
zgH9b;lrJ=-enfn6#mb4Fv|8nxMv<><IO;e`mM-b<-}8s%eupo;q!IqA_~&O&J8bP^
zll-`-%0H=feOviNUB@+@E@-{nweM#%9logAoeMozF>X7Q&9B;SvwXj=fd+Hx)}FC>
zv8Nk<pKBItiw15OdEpMt4_)V}V%&0=TEB9QZjc@u*J(TS?A{yv49)7H(9Tha<=JIn
ztrktkmieYov|ri&<e)=t?ECaS(>>cUZ2d5;cv1gpg-wtD(<+xL-_chN81rJYQZFth
zb^gH=pJhF+Jm;>zj`Pw&)%sJ-m-f8QwfQje2CciR)<c#r84b8L-Zmv~khX0f*M!N}
zcE$6<-W|Tcm&G7`r1c#?<0|S}w@*v|8Q)9;Iy7K<o{<#&w9Xl;^3{D#D=m7NAJx$c
z%O8!X_Zw_Wcg8cd#=h3{3*U82YyH4BzNTB=BV(THF!;gMzAT!SD=qr6>)C0gcXeJ9
zCb#MJv>>~(#gFS6uLRLCweG8%Zcshx{@&-#FDLBOv?T8;9<gxPdHzG&<ugo256@KX
z(fIwcpZpBd)kD`S2R>D0N=e>TJipN)8x{<yl#hzTdpEB4<lHS@hH`q)-G4--_x_0=
zeQ{W?*?ys8RDIZ%E@-@D`g_uP(Dm*`(sa_=k5WD7W2#<tucNBUsd^vE^u3&<=&Rln
zcO_rgz6f?>+j?14dBXO=ZHo^&FXP(xu=SrpkEh*a__GFVYiCKx6IB1>hunJ4TQB_R
z*^1AdH@-)$>J@TH<4bL(zuBYyRxfP~sy9>L>o>h}eA(hM&s9`?cgvQ1VdH;O^=^u;
zZ2xLX4q3XY=U8;*la<%j^rro%uJ?jW--DCg?~a<@)_O;-pmm#zuiSe79aY!)u3}R2
zepmQV-)W(#b#B@6RIj%CrE?o@@?62F>wB_#NGko(@AC)kckoi*HIy9}*_vM0`eIV{
zBJz=a$F9)Qi~5KfcCEc-^2+(isCJO0%QT>(0n_tB=<6DXsgyCx^B)b^`u<#0xm4d5
zwI!eIc?R3kM_muGJzcjaT9rNd`5NC<bbTBZcb)G)9xdJff_^u`_3z?(VNzO-`aW)!
zF5d4lWcqVkf2!{}x}Gm7KP2)|J@+8eNjfb1dkd!1>ngwO?{ewl>l&Tc#+=sW*fz!a
zw9z%aiSnAh4<OPP`Rfj!anf6Rb!xHRcLi;qPezq5?e`ht+L5Vu)zYFXJCBa5Kil$$
z{SKUM@vfbFPfFgj>`Jr`-ExTiVir#4_kBdYhusshJ>K*@$29%8?M$rCLJ9xa`<AA!
z+_Kb*is&(F`n1y_uPx~6`!BZWFRgc^;_?l<my)!8kA0$ecJHGW`mUq6?_xTfbUkJK
z<TSAB$jcWG9A71$HX^m^?_89Z^FOw{Cl^-^Th7baE(h%*!%PD@H4t|HR9C!Cng2~o
z*SQzj-wzej>n@inJyZ9glA^1tKBLl!J1<E(zp7p6&X<%u1o4Sm9@+i;@{-=Rb=0`^
zoK$|<eb%Jp68X5g?rQ42Zc#s~VgJrOmNxvND5ncue{a<`KT_W03432k)y`G(1@+^q
z-&N%>oi8c>DqD_kX79W3q^D}+GPF|<hdtE($$vb*T`mY^>pSi_5Lte%rq{JjoHiZq
zy_&+f_EaeOquP1Y_f>SIiz-)^?tg2*w)us4&dv6)>Nt*k|GQS~`5JcqfOvZ({#lYX
z6K5X#(YQ*vw9qzv*e;K)^Hf*YKV{!{wbed!y>}T#$8`A%WgJ*u%3<0b!szQ7mxb|3
zIzDWD&i44U`l-0}l=ObBD!-`j6N%~K>L+UcxaY9L_zZZx{s~RWiu10ctMC0Fx9aaN
zspbo_3)At;<AM<WvhT7+eOHV5u&o%k-3R4Iueverxl~*7%i5cmUbkK|dcW2{kX`dV
zY~E#id|Kx}#dDUng~jn#TH#+?qyCUBeh@jL?wcgFA8(3J(EK-U4XdtS#N{jISDhb;
zbX*OK>yxgQBQ3e3`hh}8r~00r>H3kqH>&IRBW$H3s}Eb!r}Zw9ZoX-a!%Xpw+kT_|
zUNPz5cYdQoHY^xYDQ^l@*NIH!que5zeqTj5{qjDp66>$Lg!gE-<3WACt&)q1VeiqI
ze%~VMx@c1AY#mRi%H_&?n?Z7#9v21iEwA6ZUODioDpN}GuHv=l^=SX&jTQ4rBmIl&
zI*_<tA2l3x9Y-9OM$K1f>8K}LQPqCJ@`aT*ss1%-KDO-_gvld&k61Q+m>ooYH$gUC
z7~iDgRoBmK$>-`h+#tDi?JFitN7nvj)2q(k<K~l<lk~dVuWjQF-Q^3S7qs7(eILH4
z+MNr%AUy=>Q}=srNsnJl@e3P=*dA}|`M9WZ>9)(H>EZWX$F$$qR^`t=jcyF`qvH3l
zb>pPsMY&|damPzp{<^hiY+mfi>LKazx^8}9>n*zDIro;~!~0*?VyrK&XSx=)tsZ6l
zrmpn1jUQz1wb^!0Pq&{-nqHxt)6;#PBPn`u`z>95L8RlFe@EW9`iz@j)T0=d-9Jbx
zy>07(Ve+V+*HB$|5$jb~Sap7ybUt0>O^S}`@n2Hq$oM3y0aNS2L3B-*&eXWZwA?}Z
zGW|WMTwUi3J@u!?lReil>Ubwe54ykGHa~S;o}}lKy6qt@y`=YPOz~rWV;aN8*;)Lc
zH=57u;@PH$pmq}VT&wClNcXtfw0)bVAGJTVHC<GDNh@8Do$LCZ)IKVFO!XJ0=|(+w
zp*x+fdQQuJRr%`nvm!n0Mf`3mZ2F#?Df+bAq*0gMLx*>g!qC?JA=!PyEd3D+j@xaH
zr%MmEjeo=BF}-h0yUNfm4VZdIAt}0cX;WSrSJ%%>t&_*4OAVNMkK>uDJsQ7X_LHAB
zQqBxB4Wy|7)B8)%7t6?ZRdj7%>X9+ebr}5MYF`y?c|Sr`{<QO%UKcG?`m|eH>@O)f
zZ2O&n#&y~bJ-hb?KQ5Mx96e*t;!S>1nD#g;F8_~KJTd2w^@{SYqN#CoQgkoAa_jwf
zR9)x0j<V~urr)Uwqn~ttt-5cJbiNl?tep5st5v?M7`OjOJHP9GAnAP--TYL~QRvDS
zwhkT?Z|ix4Fu7#c7j>P7W89yiT^i6;ziHEnyH1=|e%<dQq(!e#*DG}OGp2qY-F3a|
zKJSzkJ>9<#kTkut-qTg>4NaZ1(nYt=onKDascA_*E~dS1s9Ha=`O=;T#^tZN?<AWq
zuD_S97gyIMO!c=J-ODPTKX|b(tB0`l!L;IK+XLyt&IO9Uo90?r*L-gDl>^4S*sRnO
z5k01bX}uF+E8n#0A&9T+du>UlxAmNN+H%DmpQM#v_r6b%UAy`{Y~A1VK8h*5EqL(S
zYb$x1eN!mA-kx@P-R&SPdSUw@w#KJj|8e;zRo|xk0d!U0$Fe1#u666EbY#~t(oQcs
zf0aFNLwfAV7K}UpNNaosUtR07Y0;q^u3}nxi_2G+zHMp8q8z&SuUm~AS~8-2<=heN
zMjo@nYa6_@Q7ms*IPSj5uxZy0I(1}G-c@vU-CmSC?!HXg`IEX|pqr2CdEKyf88=>b
z+-7_Fw8z15`KNW>qP+7>+poptAJ^_|$*<}s!}5vyWwwUn`tz{-y2rI?(TjS&I7}Z&
z$ES6kR2Sc<dbc&5?se)g`>|#HQdN$)>jYW;pMSqfRS&B5rkXFzp2Omkx_7FJ5AO~Z
zI(B=aRoRoDukl?$aoxhTu<Cj9p!Sv4I4g{Aki2ofhn)wsH9qcsZ<hap{P7jP^HlYq
z+b-^2_-gkHSFQ011(QD47nfg{|H<OZ`d{5=o|ow%(}T@+5O+MFnqPGs7M4#{9`fnl
zU$TYXE(>e5Xgapcw;gTKm#Q3L=Z9^NH+{a@G=1IsCHMtfQTMo7mR?%(e_iLCbkWgm
zcWKiL+jrL89?~waF8-?P$*$)MvKQTUB1=#AecQ0{rz`QQ>&>$IPb$4^`?c-7l&t)=
zrq|`iblHt9_tE2)NB4cBr1sTxjju#HaqoAel|QJTl>Hvpp2%;jACZ+yR)0b1(Mu5O
zdM`qx6T~n3UOqHn`gh2nU%lGym(Fdt$#WfT>DS}R5%xQaX~pYqbwBpLrRgiTEcIMR
z(QmsNzG38rJ2XFZo#!g1{r*c_{<_b*CQUEy`2kz<o3?*hKN{Dr($24X-a%GRs_A9h
zi7FrRG465&RqM}|e75)%RXI$JC$L9Vtl#v?@nwt4JXH+h8};24*>pktlK!r02T}7S
zU7n=*^{=z}hb}``dr>i}-o$igG#$RE+MNqMSJCvnE^+-e3-7!A<%8e$RrTP?ctG|%
zi0$ch-M0*)Bl^dnb>_I=MLtP~(^{8`%UAXOf^GTY>LIQCs_TDA+jqwA*BbEe++%6O
zFN$)u<M<)B-t*QAKYFfWuR7<=`e9K?&NY1H%C+z8esys!Db#g-=hs#qJtVu1lvH}%
z^ML3YF`IID@5c3>oV&%#@Mk@k)%Jg=i}U327XFVWe*Zt+=d(n5y3W;!bVNR3VbkXZ
zg0###!@4ktkLmBO%=-&n&airxwKLiCYNS_PKa$O-`tA(*&{y1bi?ICg2}4)cX?4jh
z>POZd%6ocIf3AjA?{VnP=c@jK<W@Zw5=1ZVcU#BZVgB6qc-8et-TC6$hw1#P?ZK9O
zNv{W*t`F$j-oC=*V;t-{>b_?hM2~t;3xnjc^}Ff!36iP@Th7<X%3<rcC8}JuzDF)A
zm+Jll^Zy`}J(p)|de`T}y7t48rjzu0NJ0L(LB&CDG@sYSa~;L?{ixv}z3BdKdjBAZ
zzV32l^fC?nTLZf0w@GVL^?f_?sh)?6dX778x}^2#D!;Ve!EjYyncSHM{#ye<=fJb?
zSDtg%U&nbtdiZZ0s{9}JyJD`yn|=?~G=0Vc8NRhwrxxq|sCp1>-nMzCC|}%gTK5ff
z^G*Brp`-FQ^$tK>J8&ict|KpBJaBxKyeg($PSbwhG=16gcebYwT0gh-yRGZ=VRD(?
z?@Nk)TJv&Ue0A;T$MvI0y>n`ckFEY8s$5C=WmP-TeGW2+9`Z%iTTnXHbysoU9(h37
z{bVscbW~B-x#;pnM^)cqzNq_zuA~dQ$K+bP?e}q9v5PP{!_F(J#*6LE)o|SRU38V(
z)H`WW>4ud<_qtP5dSP-!jSp)lNySH%Hz-}y`V`Xz(JR#NNuT@D#V=^Q?aKF{alR|x
zZ8^Uew_QfHN7eakRQjY-t$$g+th(>y`gz%;<d?Ne)$wIeKKRSlbJ#kosd(g7McR*U
zl$9r{zHOh^iSn6R=S+&O?)Q<Brl)ItK506-_5;$UlT^Q9ijQmSUdSu@A=kp<{VTCQ
zxqIQO-7j3V##6<p_u0bIiSuAF9CV&f^?TfV^r9Za!m9fxy7Ps}X)0da2TCg(wS2nI
zF-7GYb^I1JT~xlR>16v0RX(=o3-fP5@w&<-?u!;m*p_`&UG=1!PEzw1Q+$$M$J5Qv
z^mtI!UZVC7rqZd(?OHxv?IP;=ebP0q({||Dy*Kz7mQOt-bsjWnKDyUWGkTX*Jb&<F
zA3f;yue$rwr0ALI_n@n~E)kSZcAlj>ecbamvipN!>9OON+Dw15NBymy?P%)v1JHwO
zD6UJ|5@y|3H@a3o$eVP%=+=8wdO`cxroOA5vk1#az3IZV)@k7zv~K3=ch%pqwI!df
zcZX!@=(6XG&hKl$^m>+V`i!?_@%Ocmer(d`BwXRAt7gLbQQP99j$>?17k5AEnW{Y+
zzhCx~uL~z$R=>jXZa@02;w}qowP-rF%y$)8kIL}Z>vQNsm!B(CI=auri1g5#?sFh<
z>De+5P?aO@`<iLxmz}S~O^^M_V%&3WLHT9nur0mp{J>OtRlNl53%L4S_x^<I^i=I4
zYChF>DTC<Su5Z>43mxO`%cY&)^!`GS{f2!XeDs(7CXd|WyNb*sTt(N<HJREslBFAF
z=epx<?-#=43p;lp#_QH!5WO(DZI9Qz?jMz2kUmX)*WI7#mNTjGP?+3N-<3%!U6{PG
z@nd(-oj7lYQr|T+{l2bg`U@Vs_S#C`X5Tc*+JmX}1XFa=o)_ri&p8p<`ch5r%06n4
zoTlE_NQ$m%y}6ptmiuw#O%CyXdfaea`wME%vfsshu(ZN)^=V6f*V{!}^`G&*Ab)(t
z?>zL7*1l)jd{y5iF`du0^I@v;B<-Jctv{%i*Oq)qzu%>sU)=M2Y3EPc?sfB%-S-#A
zm)h>FKT~V$YfZoK#c`(Wd!^#KdRSOi{xEt`<7L-dqP`m^){ksB>Uy}Ybh7fvrf1xf
z;jhPmqCS$kFRhD@E9aGE?UnSV$DPoJpNOwNxc8sdyu92ujUwMHtfm21&u1Wa)P7S}
zI@xg#=}q5@5&g0&VbycoLHX2L!EbY!sxQ;~g3vd;&uN;zuKr#(o!?rDnJ=#G+l%s|
zcTqdC^D<NEMY&9eb@e}K(Q&0eDKt5Q?!%b+Zn_=lI?o&?r|stvgXDAdKCmdaDZd>?
zH_R?%<HPvL#s{6BmX%92ebVnK>gH#9UBNVc-TFwH9`kNjG41^~)pl!ZzLs?=e>nA=
z%6ZcmR{qz<oYv*oHpTg*FialegZlZjzDK>+B<`oV9%dcR6_njy%+fc1XweDt{7esS
z=>b0}qU`y}sOR-{o%;xSPxpv+BahkPwGCbv%9c-8dR^s*j#zJT-yM<7uS+jMbadaR
zvxS~5{UG;a!ch15u`qg~T_+vZbxyI+=ui)7t%K|08?+AO%J-!1Q<>(&K9wto{hPw-
z)poyhZo^HUYZ&#OcGPjAsdPd9L)W=TTj)fs4_nfG`ot;w9C=7lu27KixvOZ)I9XMW
zu<@*}_^ACr+Uwq?`A5}*Y`P$SE&E-Tudei_&Ix36Z)y6<Ela(qdXT-hVOx4v`w>y@
zpmU?Pf7g9~D~z71|8aF+*wlI7r0Is$tE=%z`$JXx56jo>iB@G#e!j*xjkdMJFum9o
zulqcMuJV~$mq?1P>U=6FA9{^?&y{px>;JCAr`^uczbQ1m{%e|kp`1rZdtbtoeTC6=
z#Sa9@ne@0x7e8C}r&Z-}z1^Gki(&1~wRqL>yjV`z_pW5uErQOu1bvrXrw&>Vlchs?
z+5IeWy+<}2B!AF%S^0FQ7wso341d}6YunNX%?ILs57VC*A9ij)Os_i}WS6?W$CX#L
zABoF1YW}F_;iIOD%D23wb9LQOlw0*YU0lBumY@F8RaDh$)O<;;Z@J1R>3s@Q{8Z;_
zQS-(1+i9<(#O1H+er#Mlrk&rm@o*;3!~0*?Vys^%dPv&7ZC$TQYJVfHU(}Ufb{#3q
zPT22q6~pG&w#SR_V!0MB6n`4kzlF7@u=up@^_$}Bs^4%`?yz~jEAguHJK6V?vh>BX
zdmpvXkE@3;zh+y!YxmQl%Io_3g}TPUg+|A=_7hc}^BWzqVZo3}`J}PWonKDascA{x
zReYvukH+tp{p7of!=_z3=+u!#`M4O>&XP(O_AZOAc+=$(-`li3Y)T(zob=XSom#B-
zGmNGOQ}y5)-KgsUuB6j-?lNgQs_Pb}?Ja3N=;miiKf3n+(xM~#yA`_isJbpfK3(!W
zane6eukdPxoUQmkqi%K9pI(x46<duQS~8-2<(#WHe#ou&y!FD5Uec(`zC}7ozmr{F
z`H6Zr71njW)pj~T@22R|tLpeGET1S}TKkBi9AV)geY^5q_PlkFK4kAr=t?i1w>2Hs
zJ?>76Uf6zTP(9oBUG_ect?8dTZ+wqh)hpy&#h~}Cv+s@TJvn!ar%Mm2?N8O;%J#!)
z`7>AeMy<cN^MR;*vUG*50aN=NNog(WeVwT3qMj2lRlY~YJlA3HgR6bpv1H`v8G9CQ
z@?AsQ_cf!+$L}($qAB~0N>}z>0#oT3ce(yfvnl<Ebfea<Y&zL{3Z$3ilT`Yk_b<Y}
z%gV<-j;koUzGQ29UF&(W^{+d<DgDaM6J1H~YP)ky?y&uVsPVeni>cqS)GcR5FVnz(
zYryvTWueIT-#VB1KkEAtV!ALMyDY5LqUqQ&-&Tw||IO0Xuduw^kJ9%6!t6seo_FAM
zVbbs3xN;9NOitJDdj-8atNXhv^9xxyRoictFVjG#0eu?Kb?+-_I%TszUGs5=&7Nu0
zr|H6?6xP3+ir2M{VY)U<txpEg70VNp&Xw=F)~VB?W81o=Xdhx&oS&Dsu&58&aNKiF
zVfkg{(UrdKNv)fd)~S@Y72}?Rjru-GP`aeor(8K7_!~Lr)U7>Z^J34{cbc7-yW*fX
zn$PRv{ia^B@4u@TU4B%Sj`Q|rxV<O!KB{RxaoeM;T@_0DxOyXhq3FX@IbJ)dW$XL@
zy2^`-rsT`$W*Tro13~kLLi_$wo9S=%sK3=yMN@Sd_1sn5bf)AlRJu{`yV#zN@q{Xd
z)vxUSYFg<z?=Jf;Z`eKVq~lG!udiw!uI39{*9?o-y$+q$egJ&K+Idp(Nw+6m{Mf&W
z+fS#JKdJgn`dpeSKV5bamrhdkBFo3NeMD7xOq~ZvimtBnqG{35-H)bCkNz|%lx?>`
z>206K=#np}{IcK0d%doPb?xUw)t{=Lj+@WayV1JGYf<$mrnBYUL{To)aNKo*q|ZOt
z#?RIJU#8YK(xw}=9mL(wN;|))_aUR|PggqK_hpi%7q`F6@=v||!5YInRXyk$uck#u
zwcNJkV?EAQG^Iyf{?T?ivi?@KJf}WX?WM-otoOoDmTp?>Flmn)RQZR!2W6@q>7uK;
zKCPP1ws}gBJi6{vrB27u-$RSLZ&6<J%ldm;)2qIVX<NRu=l#0*yY^nJD6cEOzZ!Sm
zU~9XUl}q(q`?T|=HICNJmvNq0Z-uhnpLRP->vs=r<15;iZQ<0%@z2y6`&!d4d|5eM
zJs*kOaoeXY`Qz><l0U3nptqzkOfJ><wAP<>@s0X>6t<?r?p#G%+P|tCw%UcPTuGfr
z)6GZs`I<0#VSZ9}U%|HYVSXg2{9*Z$j!)`*ttmdb=ZPY{pnA0Jd*eE7ho0SggC7@l
zjkj!>S4WjYtasPKy7tl1rsL{<hpBV1rs;B@({)VyzJkbK_I|ACdQO_YtL-?d+_LLe
zVtU)g4`K4C&iiEZA%|=}SJQ{hqjblk-=xsgIp(D4rfpww>&x}uV?ch<kGc}pHLpsW
zj;VRPX}Y$pAIZv-^tzC%{Nnma^6T0+!EYsnN&ilR?(->O=a5~AkGk%pD;;*5Hb$L4
z6>7S)^`G%C9~#j0-ar@~QJYVk^v}~Pyjmfb6pH+V!f}5mAgTO8@)!E|wCB&R@(*jD
zs_{v+->`O7=yso#FVlca8qodTNm9R$tm+5T%7;w^-P?5KyXoK0Fhw8z2cc=Z2;y5_
zzPn=Ig(~M`Ppq%B&yyAMuA?pOB}k8^<x6`%DC3{~?pmt9J7@TJ$V`s|TyH<7#%Cg3
z(|$Cg|FZ`A9r=qU`rg#HUFU`^_iDL!%i??bp4+!&g@6C;T)T6f&O5g(>fN_<-G7H0
zwJdH~eAebpf9u=%xqF*W^GXua<In-;e>-5$F}^N_Tkh}Hwy4b%FDdN!Y*D`tp5EdW
z8h-s&uhZVXdXr}wb=A8nojKK;%qu<edp`(O>DrR-`Wyb#?D6|XdQmZLxz8`V^`LXU
zU+n2Z+44-;x%J3%?p)xxinI3rvT5~)=Xy~ws(#M+=IJM=RaogKh4qSOAG6nii~O`O
zs$SxzyR+8GEiZrfJ6{#!^1tnr!}fh}Tv0A9ME?E$&~W*=CB->eoP6qWfBv|4QO;Go
z;e@pd4n46#E^VY9Y{w?`dOUP&@A;mHw#%K#r$+<rir&5AjjI=VvZ#mqZ>dz<{9wOv
zofdnhutn*K%U55r#0x?s+OOTxwmsUt@2g_7$DW&ZOi3S4HzHq#dNnZM?{ybma_{rL
zYxw!ok58LE<ZUku^=d0UY6H4{`byhspL(*Gmc+HY9{5HsKh2j#-SW1*c<92FH;?wx
zLfz%Dg<hA+m1<TwbfK?`wv+=ouAO*e&+4aC%e#tuPkrgWL#kEH7aCGOBEIucgXzZ}
zS0is4MZU7(!Ot|m?b-QX`C%wam-M&ndgXo}RaoF9jqq2+&aId2y!e1`y|}2#zrkI*
zoqPCBAA98$?;H5iJ=YwN_tQqzd4_HIra!j&tMSt&`L5zux7<H?pPPsKNug@H&GP-e
z2I@Y&cIL=VUwgXo_qk@VwrF6|6XUo0y4@sC72}rUxCSG~KT%_zSKcv5k9De7u5-uD
z-+Qj$$2E8Fwx;Ym&ozv?Z|Q2fBW`%-xSMh_JX0vz@B5eanKt>DRen_LaPg%TmmRs<
zw;jh`KJb>ix|RC2p{W0~!tczP`?r^;uJTM_TIEvZdt3K&dp7=Psvj5su-l_+n&#*G
zu434FhUs|K`cut!+x1gVsoU-qKP{Z~NTp_<?KsYN4TI+Cs`ZfNOGX0|_I#}DjqT=p
z<sE~xy=3gJ?dtjKJy-Ff#|FMP<g2ZoEC%T#t?z4Jxbp0K7ESbB#qPIXzOZ4@Cw^Q^
zOaB?)OanSJ&|*WsyIwq?B=0(wZrJ_VT$Pf%E9k9kbLwgTsF-&h|Gs|vD=N1r$;U+<
zww%#$M+2L-eEh(_mKWz;!&fR?by}S|#d+7U<;MLc-S%mPe0fLHEoRDlqwlW0w#<`7
z({iOne`KXOCq8k*LeF*_HUFpmTi09V6&kKwvh!}$+AQ@<VUS&=bq`GBJ9hIs9h&aH
z*f)jSx4r4wOW$4PCxu-)Jl*_^QKh~t2Gx`9@89>j?fg2QedD`|NuA3v&F9AAJMMkE
z$2GodIRAlzia)yd9Y4c#^>9M_#%<nTKi+c{W!KZg?zd#||N9=0?*1bxy{PXaxSlSm
zK5R>u);W))^)PM0-Myv`tDJWYdw*B;nVX)doR5mS^g5vah>!RBXr3>Ny7-H9uGs#N
zg`IBy&eMe={itEpcYA7lSLLEdW|exb<8$jS|8QozAG|`vXKtQV>ylHadqIeusGcWI
zJKtG74}R~ER#QD!u)Lp}PP_a;?LhWB?Iy#YH6VL$y-?Bz)xYiKOY2^sE$b^m^=9h(
zimM*zc>nP;Jy%h7UBk8Xrs~}k-Kh68NM}k8S-NW)%o|YUv$=j+C@Zh6>DOJq^GWBg
z|I{~y55MsFA3HxY*pG{wCp|J}$`@-qQ&?rz_&tt1f3>F@ZLN3YYBT$rKVGxTYEQ(c
zKl{I{PU^kRa}_V0x3;w2g{!;_f7S!_ofe*S*$JaZ96sCAjk4t#-0r@MpLu<j?+VuI
z{Ptmg8#2=mLs>l}mA-85GVj$I%YE1IFDuSDx3vEf-*wcb_oVDa<m2jkYTE1Z5AM?a
z-Z3XG@G~q-J&5fj3uhY8se!Xj|7>xudFy=BShxLSHEVAw_2Z(>vX@6%y_!@QxBin&
zzN@(CxQ^ePaDJKZD!Qt7<TmXmlA{0i`G;)&<eD{phUw^`Q2K#F9e0cG^(WO=+P)8(
z$*1=2e5N<Odb|9B>4TRn`_i)olRDRxlphlLbbj=oNB(fhWKR}FI!T93-6vC}tE>F7
z{B`kljgHulrxkWhZ=$@Pcd0P_&!eXMNukJJcX-l8J1v=b@Ep$-+-t|?=giu&z;hip
zJ$S{Y3iUto(!#iQB-{TLTKc&9vn_w;qlUlq^{|qhD_HONC2y{~y+SU-U+W<)yZZ9v
zJ%;bvyf_EymZMPDS6nZb?D*TZ^t#SVq(vugJImI+e_r=SkEizEec-U!U;DNr^4fyB
z+qEs_N=lBQ1J3_;z@B4#*D$P|B&9$246k$Cb5{=D>SrjY2lSVrof=4My-wtdc2w6b
zM0%=W<ggVB<sO2m`!}ZPCar&4?J$$eZ4D%KZ$-D3bj=6imS1<dROy{xcI!dse81Rp
z6$@?sUUzv^?Lv1x)p41wd~wTj@Vx7HZGYr;dENNvipdAQ;#bVOil?lreAM{fC3#yg
zZapWJf7~^<7C%(<xfd0al1t>{>OO$3eqN*_>PIy!(v2H-wVse$b-tZ;KHdFwTJ*yD
zSyek%&8NFwRpl_9FKnG!SG;UFf_?`x`@Xbok9P0-Y3pHP{}C@tK47dDgtGM=cU?-B
zpR4KP-V3rNzv=yTSI%phmNTlI#~uI4=8q~@mhOLR;QA`7U#_<Hdrvq1Z*JjV(;7Fb
z@(mkrhQ)_#B;6EI*N4-U^0(Yg_X}b4e=AY;<@Yqu@!6t&A3VLqvkmV&{p=5quTqqA
z4L^Ua!^<U&H+xa>_v$9ixTNl(m@cDg|B~fXyAO)qRPSAxwudnKQTyGnbYXmwj@Pvg
zlop*ZeVdApTTia+i>sbnimRWf`QxtVhVfDLS6M#2_iV%FZMx&b?80<>Sb5;z?_URx
zfB59}UQ~=ao{vf|t#phhbR&AzjcN6dy7|WKPj%}xqxWkKg!MZ~&EHJ%(KT-RwbsHz
zy7-!+6ISm@#fNJ$UJ6rorAjxdUUm0Vw$O`Ln?*&o%rEYnJ*4uu@yk6~j9Y$N@^fAa
zv_&pg<j_6NvYnplIM8%`x^|z~w7g;S$SgkHcd1;{f86#fJ8psBVIw{namKJR&lJk`
z`=--VZV^?-scGelIzATDiRI41ZfQWaUkmdKvh-xrtBz-~e9?YyF06iSk2l><i~g^?
zhfS@MB}Lct_#lWry@cxTN~`{^Tf3rnuXy9?MP3+I+x%d^ah(=>V9^D~z4Q4CbNq6L
zRFtJSW#`r-&$)AfryIkHTleX;Ge>s%+AB2N;o+Cp<lcVCa|KQJ$6<DG$|ddkyz}yG
zKMZAmA3cn3Qt>Y>Z+usenM*uZvC`ZnqvrKr<rNCf{Q9!LtRAz{a}8^JSLLEdW|exT
zu+HL!n@-xh)Jqyw>p^$>y{-GXJsW>C)wcz8moJFktv`M8<*P3*@-vjvLy$hdy5;`C
z``kR-*M;Z(ZP-I)RsZF?j_a=9`K0sLf9ku6YhSqX?0XhX^j$^&oc&LoJ8gpRDpvlX
zQIGBVF7;i<a~AHj)8jiV^IgN|XO6q-!|hi1x-rO){^q+a>j%FPOzy+G?K_>j>Ap9-
z43pJE+lz-TTzT_o&sE&>o;CM0efC{1!*ulE`tN(%W;b$PQ(x-4-+iamlp{8Ju402B
z58S@2$tKShT>kla|LA(!I?q*1>t2g)zCGT4VsOKi8$H+Xz!s-GP~4&@r;7SK!>=iw
zblC}`M;t!ea}BF@tGM9Q|C{Eeg}>I~Z$nqDf8dd`_Wr@Q9cym<$EFWoSnRuks`Fyi
zdFXFzR?nAF^MBgEb-h)d?Wm`;$x$)AewbWIGEPZqVCl5M#b@<i;U|Tr%XYzbwTk}I
zv?#9|P0PLGst)gWsJFyV8iUI9+8?*Sc=rR>`mSTmKOb{g$)RifFbvXz?(ehq|FUWI
zhv#}3>e0ibC&q90b-PJkQmES=;?gsHz9%XAXZ1Yzy+c|}^-SZR&Rl%Y;rp-iUB$m#
z(ERcfOIQ2l4TIWAlX^WKI=1(GFAP^M*?G5WZI*hn_}se7Kb+a_2TvEizUBF?$1GXo
zxr)avE`F=uH%mQNv2*KXJ1;)qThCSuvU6SE7cR}`k2<kZJ}G>1ZO`2feW@hx8cthq
zcdx0#D(90%UG+So&6$6{WyosZ6;xfPikh$K4(AV>e$01%z0PkR_O~H3jbqq(XItYx
z?^0p<pGQsilSW;351ql!G{5cH`Cs{AczKrzjcX4b;Ms=D?tiP+8J9ioyN0&yL#xWu
z;I7@yJ$$E+Jy-CE8{RqYrrZqA6w<CTv`YiJ?{y_j&n|7sOLP4Ve`@yleIq^B(DXfM
zRr=I`?C;e?P2Z^MZntzB^1YunQqBxB4Wy}oOLjQy%EMM}^-ZH##y{-#=$fYa`MxS9
zy-yaGU)=jwY5Obq+kUPmNWQYU%e+@>EcXi)>vtOb`UfYL`nr*J`|NI|cb@i4m0Wqj
z6Kd5PJ@xHsIa~0l9slrH{=KTXxOnnTpS*ra|Ejs9@bq05wyZI$O3qb$`=AYHPCKho
z&Q%<B<dOFuJ*6b)Dw_WN&7|nN_AXA+_vCGVCl~p;-+uYRhDD$FuA-~=gONLIJ>T~D
z`W-*&ch|DdeOEE<eXF+D%&2(XyEA-QjJiH#Te=hX`Qq@tE#~=2Bjf%I?b5(~17Eu5
zngjB_>-hQ8k58LE<ZaJ2R9)A$E#EUY&#HCFsnfke!@D|mc<hCyv%I)y>U`*djb^N@
zIc~G>D*ko<fAxOq<R5+4@$V=1|F&(t4Zf>5Xz2-yZ{KC9pERoWhH=-=P3M39)OUN-
zYP;Td9aZZ`HlOYDtT6fdom*$_{Aa%PO=DcUm8}<3=YLGo-Ex1ownc5Gc&=ko_hn4;
zx$F5o2ETaqSDtMs+a5?iZO1G2I(yUtKM4OYXWEpb?_2J>iuH<TAG6nii~P8FXRVW4
zUjFWPzKG~CEiB!z`?I+!CHV}0)<atL5XAS*1x2sce|VX1D{_9JP_RqoN;Rt-y3mh{
zCw*A4Y~TqSJl9b7dsIPo?dta%i|@Gi?H<?oL1_A&L{oZG{aur&`CPkKtNXpywCK&5
zc+=RGTUPq6VbuF)wx+Aw{;`_1H<kLP(Ud=c?us=({`Ba$lDw<vt!#7ZY5%C0PYO+q
zqg!m~ch`#tl;m~e-`8({MdcPH`KY*R-Th}@(Xk{ii`bJb7<b=S_IGu3r{{ObT}9RN
z=~=#b=livtm$RiI3v%3c{nS(HwtL0bg&lT2yXv<`Z1sCJsW5K+C!3UU(QzHWIpO>=
zKQ0Qz4Gn%g@RYkBxM8BFiniU~mX)W)MSopBWd0X^hUw}-*YElyO-I#kLK|kx1WCQO
zkrW@(@2r}p@A|nh<PTccvi1AJFMR&T&d&_?T}4-Z?=(!#HZN9fIq>igJ>97K&PdYv
zV!HFt?1l1szr+7_-u-pY`^F2yP=zF%U}FCfFHAmQtmi7uf8e0vkM4cPcNKN}_b~H0
zuEEIhPt=&_xr*64$a>yEc14e2_8b;}?)2erkN$eK=L)+1yJAu0x9xsIm^`N6nYNXF
zQ2UB|PSsTYAbI0{HzluWx??XNc*|YgO8pGe(Sz>ai_*Q{mNvbi1J3_;z@B4#S8%^}
zOWXEn_r9NDriaW9+-(PO#{;VQ7x<gTTz&fnPdBQ{L%zy?+Wpb<mag#&4gY!8>i4f&
zxYoBFw{LsXwU@rT%6AoA|J^Uw$H(iMzp&HaXRY^jA%4MDO#5EO$VzigeBy?MUed^Z
zrY#t@f96_z*!@M_?ct^6jqmC)bBX6VM!hdlXz8}B{^OrJ&Hcd>QMaAQ($l@(ZTs&p
zN0smVDSzK_`-j(fvM8(nq|&=~4^Nc0)mPnax~%0|&vmRccgd)E{a1N$QJ3BP<Lke@
zGN8s<FKvul9^L(7jqj>l^vJAI&oth$>y`U`RAGUq8~56=`8l(;Ebv@IUHyJsdm?|G
z=XctA_+DE)*RkKf4j%vT$?Ltih+cxw)Omp*e%bf(p@F3CPn-7lNzr$GoP_+%K3I4`
zS=k29Hq_nfY%8CtJPn3CaQm_*n>^RC$J<W~Zn$!z=PLFYUHae=BR6=i;*znuwyWo_
z_guvnzwBIX`6p{V*AYK#3T6FhzXw0tvS89=-&I_9{mv(yzy4ES7G?FMn*N0LjoZAx
ze!LeIRr!!_%6p^luD!O*a|I6@@yUoYhLw4);>n-pnpPaQ&T|!Yzt06d>`@gvw_djM
z;sd_*R56IJsrONX_$B>)*JCegTYpA{oGDDYJW2DDedl4o-|H^C<lg6fS#)iGR+Lxu
z{fDUeCSA1Cl8Fb;@zTbs-6}3P_5Y@Mu4CB!NL%7v*Rv@9Ntc~4dc@(gJyR(9$GGoM
z1?3m{Bpq%%?A6@|+_2Ge4Ids;tI51A>pj=8^y8+*pH*1zxrX<T{CdS*H<x*`n6!Oo
z{C=&0px=kK?fabSP3Dyz`MvKd#(kGMt^B6%?}_xsU2|*kLq(r^aq-ty9z8T==hh?7
zxpRSM8|r?~Ai74(rj(rr7J7QK+Wrsqw12lWZT|n!#P9!S`u^ykhU+SIKXb0H8)e^r
zyXw*&FaG_ORlX?<((-YOi{I+^%~DSn2Jtcdy-B?u4;|ZkzL#OTdI+oMPT$vg<g;f=
zy{K4a*7!Y+Jb$$(i&4*OlkUFRLn@COzuc3>{@*V@WuNnwd8&xM>USFa`UfYL`lb**
zVaV^wmsb?^V`?9zQ0a>Lb2TjMZ%n0kRewQp%f3TR`XG97zo$J84&wiLmkQJWJZie{
zDw_Ts)2Q_0+K1`<uCAvc_xqRinKt>DRen-vx;~(9%RZs19Cz<Mu-(4*&GTHx*S9>s
z^_V4#Jkv<Mr-gMEH{5j6-ld*1SiN$cJ8u5ot78b0ss~s5*S6QpsCeDGGkjOkRlSRH
zhyAX5P<&W_x%#yH(9`!`@4Jq=-it|#&epSkT=mk+Ykk+S+C{A=4C%7kPYQ$l_2Vbr
zmK!p4gYP=3zCRVD7v0}Yts92X)m@H^UZ#P6YapoKl>MHxHsik674`m%m@cW`r$}0#
zuJXJ4-FI3|Ibx%iG`gy<Ozunr|E+=PkFEY{{Ip4ahH`rNZyl=qUnu7%RPWQ-l27-$
zPUVH3X!C=Xp0N1#U6%T`peSG5ur2Sj#+5_7cV}z3?ZrbEuDp4)ryJwifh+kZ&7S)B
zA17{?SH-l;IcwCc2BVvmd9LE6^VXKuyKt3PDA=O(#O13mS>m~ZvgZqJNq@w&dykm;
z@e0p%bbTBY_1tV&y10JS^?5?YlFugXvvQ&5D#l%BjGAB7PF7s?K*#%!pXr%K<ePEC
zf&IVzVxzB$7i?Fn=r2u+^0N3w#ZCQw7+RDEkq3NyQlBOr?p*1sB6L*o-S(Z%-E`j@
zUWRgdP}R4X&((MIg5+-a#<jCfT(Ze46fB<hX7{^ySmR}ARu8hjhY&Tr*xp<X$2~`<
ztK5^G7{A@u?Iw9?VOTl-FlXA7qwib(|LmQ2xE96P#t$IaP(<uX)YuCuL9s;LqgX)^
zQIlxIf`x-6mMGX`#i&FSF#3rG#n=_GSKI>%f<#4%bVZs7NIf7$@VmUsorIBLc4lX0
zXWsYzhns27JoB5fv$Ol|NqwuzudM0U-Prb)JqC?Qi`K0E%JMzhY2wE1jvkljs?1)g
z#$WJ<as6h^`80`5SKf2O*%hacPGVJG??bq%XV@b>#*TaWmBg32ntxe;e_;CkJKf(=
zZuv;fgm*fONOU#p_xzA=_(N~^8MWI7iLRnN^{CssUNJn8anK#Fs`D`2bkyIi&#;rL
z9_^QA)uMh^4S!8_r`xt}+jO&!65Fcl&yj~-zG3GrmZmD5d-&~F&wl9pG#d4EUgqjN
zuBiJ3TlJoJ%ieG9bJCb3Hofh_^Sk$HJ~4@1b+5;v<(FNjLXXe;owuUPi}O{yp5*<z
zWxl?{YWqAj@86#hIjX|8_XVc*RDQZjFKS;W)|<6`bMAR*<Kxfmk=oL%^(K|~KDWym
z->0@TYrUrVs_y@1vv1ptO0ueHJ)xIdI(bg=T>bA-UA5INSAQR}XuBV6w9O5RmQG8G
zR$KjK+fJ;1uL*wZ{$4G#{L^o}?zBxm{vvgyrsL($wf=I~D>q%57Nw?s$5yV_HAA00
zdxN1#UTtvAo-bBbk4>{`#UBoNZCIZ#Q)+o~DEj%RtNKG*-%HIO+v}_E_Y7>+=lbsz
z$o_tgX}s<4Z|mCIef_OF-|+JE)Kpt+egC57)l*ViigA2k_ugX<s`J;RTooK*)6<q-
zyH}g(Gm~=B*!l<ly83%%GP_0nZbCdKcOAa%@$LFOamnY2ZAHJ?Qq%WNY?WL7w~ote
z|7uWDD*9@#)4uA}cwic<7ImGqHJ)i_?F25r;*<k>-2F}xi|W2VhWu-%P<-ly5$B$@
zaNxXzx;h72J%jJ&$D*HizNg=go!&}g)f=zBV8*T$LsC;cebN^vU%JJ@#FlRVY@5Tn
zJ+&f<Rb{_FizOa^S149}YsPki);W8A;!16;bERs>_4jYU&GbDRmRs}t(4oKYC-blI
zor~^l{*N~jUyAXktK#3QwH4O&BeeXk-s{%2Ys$s;_q2+(SJXLz?RstfZqamHWG%OS
zqa)i)Tsbw-Rhj*~?_*uQ-omFhsF|L)Qq%A4<^7$x@A9#Jp9g+F`Q0A3-85lV;!0)X
zR~3(QqF6Q7a{~CC_4k6I=Bw#*QR_Ru8J-rcs&Q^R-PZf^vDmTAPi7DEB39kG{p_=E
zYX4s9TV*_5#oInlw(X~{^-|@x>G$4Tl~ehvn~uN#8Jjk5c-rgZ4w{z6s;2MP>6X(S
z58~gW2|es<op4<bo;T_CmuY@q=Yy<g&F8VE{T6GviTBk#@Z&}wrm?B(=QpnH+de<d
zJO2K*#f^6jJZx&3SMz=kz*hPF-X8My@vkpUsKs)y)t|Jp_rJOwu_Q6&P_FYp_jfA3
z@^4%9%6?s5o{+|_!|N>Q)8mC1sV}wlyQ|oL_nP-QU3B@^7hSXKxJSQDW7V;ZI&R<k
z-KlA;s@wnK{_@yAUj5z6Q$9&u>Eg>bKm4>ML(@>y*5Aqe)rEr|xNXj`G^=L)orl;z
zy5Z5A|L&bOH&0D`sjh#z`Lli(x7_pn^TdVkT|A>EjYZRT7ahFo;st5wsykkh|HR`y
zo_Xuni&NK%`a~UeW$kgle|KE!TE%)9HY4!Cir3C>zRS!sR(<8PeO3&)bW!R{`%bNR
z?pNLBq`Ip6T{hkPTb|c_qXx_8r{%6@v*4Gd_WWy~Jyv`=I<>93&-qOAr)_pQdh*!M
zQmS1JzZ@xUV$S-*krx~`e_<M{esIB4oo~Bgek#&2{eFw>a-v_Z!&&dcm@n?{6WH!Q
z@%zE9!>;GS0sHU&$=2PLrFm85r|s{W__Fs_ms9WA^3*R9su>5>xX?|{`d#U)`DEXb
z(e=wT|BK@XezUa4!X#F$KW)7~-Zyc6605rUy|ypAvHyMf(c>R|<fN8!lcLmg-iBK4
zf%)fMck{XHRJ&GH`(H{s&3g6JjjDaA${(cnYk&8dkG?Q54Mh*x>79T5z5dsUN{jy8
zknQ?yJ@@xz$MyL+?fZF7LC$OEjCE2}-M?p$^_)}Jt}MT7J<s}HP*#6s`S7~}U6t80
zjTif^`|eX7`%A;gi7$O&<>NmdH0zthm2R-&Zx4LfVP;ZlI&aqITW@gs?4)ST%GFi>
z%Iu2c+IQIZ?^I){5B*^4`@Til@wGm>s>k*BseRe+xXC^<I$S?3iA6E~UFq4Ej$h}p
zH^!&2Y1ZfTuD<h<)voOK7qOma%KR?+@Am1|Z|m<=hhoR~cc5(7*J1jm_kH~6{3KR&
z^<6R1Zr1C-M#~;sKA_LcG<KD}50#aJZa!PzKi93-R(_s8t^NGT7oS+a;|9O+_flh5
zU%$_^U61emxas;MJ1-4A{^pI3Xm!CG)6&@0*7u%5dyicAefYfjO~=2g+$kr_ytC;p
zi&NM7^zow__wTbf4MkO4LQB7K^wmQyJbGSYOKr6;l;5+O`eFNf1>k!87mqwRw)VU<
ztG<2u2@iLEd0`s6W^J#leBkd&WB+}vtMhl?c8hvWf_hzzFO}V1*9?93><xw{v8w7l
zuTayA`a7()>TO@~M#mQ~`6h`~J8yN-&Vy=CO<bvJ-o=(K`@21=c(I+c$M&mr;vI9-
ztcu^+WmVnx7q_i7>7ecR{3gw+*RC4!>2ptilZtA3Upjr;C)+)L<dih4=KXH)18p~*
zJmt}^(yS_b4@?!`;F>*OtgIfJ=2db0sKOr|wrZ=3R@JK9_P4Y8F225YwEF%51IO3z
zT&L2us`!cNs&fW0zwNNA?<c`-UVlx0=a;wMtmQlRy{6$c&o57GY1VqnEnlx+z5UVl
zgT78<SJU6$X8m55X}-MIk)q4n`aYV>j_u!*l$t%)->ddz-}GEEYq`AZwynR{SCk!F
z_k(#k7F|B}{WJX!`~JuG_@2^ruUdoddMyo2dp*46jCOZDp1Rghp8L1h`u;`DtEZ${
zRW(jb@4rGRx99HT=WX=iiZoWedY8d>{NE8jq_OJ2U-i79)AcLU*!7wnpP4ah<qxS!
zRr5)ezSC`M{=4(B#{cGu*ABU=a(ohtX8jINsQF~y2Z}BJ^e5JD^;FXViLS=}ygyd|
zRqI-4>8{_OE6RSO1?Rta{<gEyShe4yM_%`jDW9j*)j5PVf3o^h)ZbUR^881RJpSGJ
zX{`GA(Qn+e@!3n#Sk>0|)U%E=U-_=Pe9Tc>wtqHtt*+a%eWN4WOk6oNiA7!Uc;w@j
z7axAd;)I%wgR6ds&2Fjv{bJYWqps{<aOG!vG<daFqN}dQ+ie$~-@Q-siAk*Licf5R
z=UMbu_c^p}j;D5c{obYP4^B<hmOoRbj2r*L=W~)!HS70OLe017%w3lszQg<^R$Vdj
zzAs0AvMBMbw#K(;XUj)wCcM*OL_&4TLFDc#ob`S1*z?KWGqc6j)!(%c{nLF8U@O1q
zhbpZ5J=NIr%Xqv~_shfne#MN$x88Ta;k8d~vn=(c<A1Z~LtlNeER9`%dr;d;hKyL6
z%2d|Rp~Vk-{`{xw?6D$=U28tOGTH6!nj}``d0Mvfw7i^5=b!2GwTbUe-0A%TmnK<t
z)(2a)-!@&6P}j^MYyWlYGwkH5NBgCwnzbJ0D{TUAoY$f4^1EM5Y^yJ7FZAd)XLk1=
z6I*)3W<y?R_t5&4Ds|PqSl;W4_1wePcfeiO^Ju4u8@D@pTw<!O#&1@;r!Bp9uQt<X
zCPnM>CvS4*Hhb21`+M)U-h&YB{HjC07y2JIKXt8H`&T#Lh?j0X=9Np9C%URzkFWf?
z-^D<FUH{sSd+U|=M=VTY)x7sRp_D)A^Bb-@w^dD&S9SZ}c79iXx1#sr3kFPR|7}ug
zYU}r_w6pf_L)m&RYkR(&`s<g+Rn}d%(ze?6llDm9e+lfo^LuU1zWIlg>XyU*GS~UP
z?%$h1epMY8FX?*Ui939ohN{EsEa=nYg&8R|8wcIL3!oYwrs=kSS5fbjZSFsH_M+66
zzIX0l$BsH^K^nXE82Q&G=ifXpja4r`=cXqf?z|v%tsiZ)%?*o|PD@HnP3JGmkpKzU
zBarp`xuNAd>kmg>aM=8XscW@Iy0k21{kx;K^O=5!QY`1XH)MbB)K$FLo=|_^kkzkP
ze?I`fM~{E>k&{}^O+ry|e2C#iJAK`2y^klS<*FhN-LUK5i|G1eE5GjV#(d|;^GH+8
zd%h$--*z3owMp#{mke2+`qsBLJ*4+V{eDc#U7xLe@iCijTB|xXE%$yiZNIH`RM*a{
zy-xe8SL1=HZT(=yYv(uLWo8<?+U63gJ>79tD)}$`N5upEA6c2?)l%W%svX}Pa^V+;
zp0+g2tFHW`T@r|az&T&n%-{Bm1qpTS99;FKFYd1VFIBtZxooWAy!%I0|Em73UMT74
zzq~w(E{}HD)~tU|vQ+Zr{kvzXzay%f{<ke|yldcLQ`4eUHP3yeoBmzWqRZj?&_#Wp
zH*dYZ`q$O(BW&B<w(6Dry1YCgEn0o`i*7w{&Dd_xI%m&MV%I}v)_UoIFJ>pP>ap_=
z`)*;oSxKzA>4x<-{Y$6mNvWvs-)HI0TUq|t{+@}czq0o0#^3bH-}~)4Z&?z%+IsIN
zE58?aX?a%Ls(O{awbeFP_nLa+dX=GTR*qu6(<gm#@}*lWOo~>~FVk?|>p>{LYc9%<
z+uwNYq|v`wmc*vE)}v7DJT{`sv%TuqBw6P{?EC3;>U=zM$9Xdn+nSZ1t$etjk5$Ju
z>bQODcc-SY>h9}r-T8)>r>C)M?B6*ac4h5xzkhdJ8mnG+W7}Ky7&Io0RS(<olcqh7
z{W`U+w${hc?Nt44&gL_B`zpz*y6yR~^$!c4e7Yu8sjl5NGdA0-N5eU(Z=F51U#%1G
zn489?^}gNxrgeMHPGi-)=k~hgvwml`=zLFnTD8G<Lu!)P75o3#H1_+DyC=O|@56t6
zm-trQ^*+@6hYwh1=~IV(pTw?3uVc3B*Zsa@X!&<~x$E_Fnk-L3*S!4}>b%zVYvsa|
zf8Sy9w+Xcv4!C}a{bu`n<EHjRzjVWS&&PG^6Z5<JJDRcC6}f~Kw*9^%?QQea>yt*L
zDh`V;-~8~?mJCf}*JJ;9^>;5%`6SJ&s(JDD=_fqg`Q?SFZ&mqan*Prd7ruA#jG8o7
zZSm#U(Fff9T^g(UJ_od2U##EJgWs{8hsyljXxU@S2lSbl#-_gKt!kZ|(si#|gYJ4Q
zja3iq-h1ppb^e;jRM!5`-jDVDcQ9?&=lXtH){mjZ<Gagw)z;r_+48*Z8#P!yKaE9w
zuaCCt>pQjLxnFgglg6&k%szDH^;^tNW7Al_w_^HTt<cIXojfObuKssvDB7;!ei!XG
z@5eN&_N#Z-^y<?ps!K&pzsD3yxzp<YrC*nx71gF1+WtZ1&nYL&ytC;pi&NJMKaOhG
z{;#h+Jt-}9HT`?Xrscrbm-f15=(A^UFf@r>i~c*KdHbvA?=9G>|MSlG^xLu1TWM^%
zbNkt6-_-uS)R)@gE!w&JsVDEebhqj>Hofus3uf$EF(i#uW8+<vzj@b5l#lg(PW2p;
z>Divw#>NTbwrItBHsz{C^Jo3#K_Kt<;Ie)nOQh+0T@>q=g{$uWXR~kHjY?E1dtN8T
zo3>NjFN<NB|D_(^rS2IQOln!blB&mH`$k8$nYeOl0u{%T7|yybWS#$c^QrtP^>NSk
zkw8oYM*eZLS;P1KGKo!9>rCEs)BS*~9L7XmzgYh6Gar3nVyepl{AF6IfA34yUf=QJ
ze%mxGn_s5!;(m%@H-D!HTuu8O<$RA@S$(qaD9ZBNiZ>nSMVHfEuS3nB_x`Bp@}Y9g
z>Q~X@IB)$~+ga}UY^|5HvvvXxH2ZgY&#QX2uiCZM4y|@<RqM{4ZF{z={nNiy4XQS+
z+Pqc89X+eI{3*P9t6HsUow%Yym!4I<?>cg9QfKYfnMwjAKmsH{0wh2JBtQZrKmsH{
z0wh2JBtQZrKmsH{0wh2JBtQZrKmsH{0wh2JBtQZrKmsH{0wh2JBtQZrKmsH{0wh2J
zBtQZrKmsH{0wh2JBtQZrKmuhz;QUdSU)y2MyfkzzgRWo`Nq_`MfCNZ@1W14cNPq-L
zAPNGZ-#-@Z-^Igc?pQN^acWEFt-HxfGk!fSrN+%6ipfnUNq_`MfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@
z1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W3Rmf#sdf8anKi;i;&-KKl5Q&-VQ>iB%`=@#pKdt1~;%
z)sgS)e9nP?{xT^yl?}PK@9f%S%+2p6u2fbp<4J%7en~)Fmvq-R-}!&p0>+R436KB@
ztck!*E1R@l@3QqOspJsSI9StIWnAeHXn*d*>o#b$ETNJE#{uU7)`8GgI9KOTwCnbq
zk9^CUZ)ocX{aHJKS$|(S^Vqv<S5o8Vu=ZYNDhZGP36MbS1kN9I`L!M9%uCB%WA8`$
zPXZ)B0wh2JBv3j8#P@4u;aJW|W&Sh%=l7{gXXJ+2l6oIMu<bK9)UDLjdFyWS(u`kE
zOQ~)+ggIKWd$rq&j(u8RxG15LLn)4f>@$HmLnI)+{~t;?L{{{vcmkz<KTtejdQeIP
zLccF@b-jt@+#{=>rM~X5eQW;Re5hPXX^muC;vrD(?<27NrS)FQ#lvUrSTlZc?(<`g
zi_#hw>`(S5-@oJghoP+l@$3|f@pSv=maN$8$yZaT*nha6!?}!@E{2icRI{EhgqqK^
z-7IJ4JV6|f<sQbo4MlPPP-<H4<1DuJ7jv~`tt1d9fxPoO&Q{ZF5^zHR^GV$I=iR6A
zyLoH>Zl0Uoq*W3i0TLhq5+DH*Ac1lr5bARX`c>ZF8x-#=8MgC$i5$ssO@A-U^|&aP
z>uKoibluOk?B_j4bX~qw_9w16*|x75U$OQ}=<5dkBLNa10TPIV0RH|*xu~n(HJjSy
z_lM&6{!rHbjbql+OA;Uf5+H$C352?D!25nw{!X?Oe<vI3dO^QQfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@
z1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LpcDz%{=3Je$cSws0TLjAm<W{1-_tGk
zcE&U&=_?74011!)36KB@kN^o3O`zQGAK3nf*KN>hSrR&jqUQ<gCjk;50TLhq5+DH*
zAOR8}fxHApzO(Z=2mblX#J1-3i{(jx1W14cNPq<XOJI4YvxW|PWq2C8{?}vvj|517
z1W14cNPq-LfCS2qfbH`&U-rtcr`Tc=AOR8}0TLhq5+DH*AOR8}0TLhq67WkPl;^yD
zTg}=?fCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@
z1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@
z1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-L
zfCNaO^a(8Qbk@*euMAJiMR!}#u}|v@7bWGQr9Xn%LJ}YW5+DH*AOR8}0TLhq5+DH*
zAOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}
z0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TTEnfd`uXJFe$dJ=<69+G@vEziw6Q&Yo?1
zwyOQpzf}#YHm%ycRmB}WtG4_pynCx!t!kaPqC=OSRlV;za%@uPmlh+5IkN3PyVUfV
zl={+h4;X&L%_I6JR5K2d3@f_ni9qMQFFUqR-}$L6_2f3IApsJ|LBRBS&Fik&bywt;
zqw9hwy4PV9XI@k>KB0!qA@5+>@QAHCp1tMFR9ExX%kp6mSaQ^Uolf0naB5q_XdS&E
z0TLhq5)cV+Ul;biPV67C{}8ryuA{Pp`luv80wf?KVEaCz)a}VQGM)rTz?{G-ZMLeo
z?!i^5shZdPb4hXkR<7Z^978E@`aFf@{JuY8n|CPVkA4(Opz;3my1z8HPPHp7s&%^0
zeQf8CeSiO#=7Ue!=9n2tQCifvVZ9_!4g|JJx=xw4;j*OMRo2%#HGB4?g;h04xv7jt
zQSs$IKGFW7__N+(3B<mSr2lT-Ln2o*uEow~)=mN>KmsH{0whq%1aRNxN`15I&|=;m
zAOC5CN?rBsC+jOu0+$@rc6`s17pJCLD(^Amy)U%=98A|<yVhN={qdf8Nx7)6*;*>&
z8}0DLjrEW~hy*YXN?k)_UrIie`uJh{*8Y1ijAh+@8uH(=*B<Lk-@R6)Ol9|r&&}Gg
zPOq0YsZ{CdCyjXTy@m}cU8&4J#9!L~#>s6OuU9Fmti7su*Y8u=eos_p5B*=Rig`o*
znLytA&REN%|BLcJw)v;}J+7~G^cyuB0oT_p*Y;KWz-$zliv&o31l$qGd;OE`*BKw@
z@Ace`G+K5^AnWrhmp+uE?TuPpcJ-LPD--I^9LjMFu+1bu0wh2J?g-?)PeFOnL&VEe
z{2qDHVbOnvNq_`MfCNZ@1S%?udM?U(&3+#n8aLO^1!<q(dEz=<JOSNxr+AO)0SS-*
z36KB@ghQa{^U!jyUv_S-8z0Vyq9<!2Q10VAwD#M64_@`Yu<dly?-rn3-tUy!ZckD6
z)-+dr#&Mlrown(`G<{!n_1YOCf$}7vyAIjTU!K&<r`4wGvskWth_R(4KmsHX7lFL@
zl_;<Kd{4|D7d3hd0=D;+uI<6^qE#IKzQbADFW%o{e|OEx51)9}b#*FjYwYjqgm!M4
zwV#X5Cyp~!xF~;P9iOhoWl{a0s@L~=s`@UNZaVf=)MEZlTWIT#?eVOdKTIcqQX=5{
zd8NoswvK20jzrdczT3(A*8crf@DaxY!zA#(1XTNt(9=V^zliNVE${yBf5(!I|Kq(Q
zTbi|fHhi~f1M7EDYO^i0RU84==i099>(2k;2xqG~>pEnbFPj(4MFMsR_}b^#@s?Id
zfCPRbkas_+qUdW~a9t0cH&cuGzOn6d2i3S`x*Gzz`*`Hny{|CcH|v(uT@SL3W7)dw
zrdRA+)$OmW`E@-t9X~REPHD4M#dQy^N@LZX``&nb^G9n{=T+T)2t7Z>iK*Ip|I9U~
ztk3V5&+gxcmOYoq+uyS5gepF7y(~`xBtQZrKmsla$lg~9J>J#z-jucOeQNCaWt?1}
z6WX?KdOsp-KjZBXkoAl1yb;$4c$f7(eXN5hM-793?EJ_yUK}T}g>}zWnZGo?s~%fF
z=3PB}w>Pi9EboRu*5^FB`BdKO+KVN>?!Kz%{I0Lhs&P>2>8|;@wqNx4Wc@$C$Ktx5
zFYSNh<Tj1htMsj=`w-jfhHJjM_HDOMX0P%7^SZw@w@!7bsOfmpE$548sjPppbEu-@
zRnMPGC0%!XhnBxo=0Pm?F`}QY#)E7fE!B8e9BG&5kf-i5IORCtIEY~!6t%t=z26k;
z*L}_=mM=OS`@YOKKi&S$n%{RjY}KcGE+Xc4wT^6+be%G7!(~a)s(b%g^!49WfB3dr
zbiezmAMfSps;mBp<$B_({j|Gw0=C~<^lfkLtqU*J_w~dz0axpb%5PWoy0$B8m$(m$
zHJp{7j059IAY=l**YDWti}m;N@cZ1O4!HErnq_IZsca5rJwFL0pRIG1qU?mi(bSL7
z=Z$Qf%i?vvpP_pWfc!<T`@YUgP>=4poUQz_e%6iWcQaQ1ZU*{Cm!oNZ)EAo;HEv{_
zH)?g+)noduOhm=HBvajUA6NIcrq?Ocav~?+;iATm?Rs^`Pto~ZuQRgw$@pRk6umFY
zTYs_iSbL}hie67$*B>f(`c+;8O!t|(<wUAD&otG%*B#&GeV;pJeeR{3&(^u4tNSgP
zUET2-TYgvL3-eZYoG^be-=`;cyWC~#Ni6ZM&jnoDFO~7E%PE%eRc?O9J|1K5uh{v=
z?tjta2m4W7wRPX(+fGruU7vr&ZeMr&#Xj%p|Ihao;&&4m-m2r-Th2^Njl-W0tpDJc
zVRb8QtF6D!qiaX@o<zCE<2|ofwW#Z-?sX6OLth_s@6ToVi`uWDUhsCUMU6jO^~(Cs
zcD(C;5bejhPs8|)T|-~zvicY5^Dl9{+nTSg#+@zyRCa9nFMBR7a`zpU`Bkp*zWK90
z5(tTa?dwxWZKE%80;=m^-gJ4*Od$ah2$evo?E`K9o=LQ4>wY1xoxJO{?ed|G@7Tw6
ztmjeiJL`E#sQGN~lVh_N+I)sTw&%6&buh0zw3A8#Bv8HtLcK1U`jvOT%ksmIxOqhX
zq}({<y`Hjs#hDjXj8F1%a5e7AcZFZQ<)-VZt8#hAi>mzUtuH<l>-%i*TRf+VEvy<Z
zy6MG_D0)BwBoN!Vmur7<J{qgqdXF?VJF=M^YCp>SWjqO#0s&X|Jw@4deciEb-}ZBu
zqV2gl-=N*{JhzOc-;25y=-$s~&7ZeDq3-AM`W4G~%-hbQ%g4f5=4Vm+AJm(7A7H!x
z^R6eZ%iC_Ru07Z7qy6$AkaymPQeL*MmTP>ejowhkSzbR(&tF6BPwXFJRXoRyU3K>l
zvi!EjnasZLcw6nE9TKofpj7rHw$^<W+q~CZS$S2xv8CsI-XM-kl|9vZp7s2&)boWh
zj>P_DIIQO+p^OvyK>{Q|0<jX%-A}}R9aQbZWa+W?ugY)PexLFCbvkvQ!D&_wvT-BE
z>&~~(@{8@U6;_Q?ReD}OWcF>{w~KaU^)bG*37G!w$@F<hDCM%Qqo(=t&VQD-`}g&H
z@iZMT<yuZ$U0wB)SP#P_U_!vv__S@;q-`u?kAUs|q`f%z1EKVry^(EZDHIN_{V=1$
zQY1hEBtQZpfpd>K;L<y5mZj8UIf(rthRgN*Le_4PPtoDnpOfbGzi9rfp9DyN1Y#%P
z`u;hq{n$Cu|6&Mi)atUU$Mjv9P|JfuF{7-aLUvzJ?(uo|@hGo)4-NTZ)7ZzyPR*V@
zX<=1O61%F#mn>a1-m>m*LeD3@KNwp$biOkGd~=pvKUL#AmUL5|zTRV$@pm16$w6($
z_dI!V>RVm6i}p#tgn;aNpEcfe9?7l)w&P8hlw+Bq_O<26nr$vU0>1W5rB+Ygkzo6J
zp|Yp?{)lP1>-E?4_fRa?_1YirnU}cYkazsYRzCJ~K=>caIRN~%t>U;dJ#SEz6YKFE
z&fCwv{bhY5KmsH{0;NSjbzc&Cy6ydk%wB2rdik{`>$x-Yt^W7i%WtHx#Uwxi<xZgJ
z>wt>CY@gyfUgcNTbk}}nwV%~r<|6?TAORVHyyv+nZ~7b-<+G~k^)2-C09*c<^6_0x
zHXdd1s^=oQ>9TrM@wWO&*N)1sSkkk$JL^6ulzc_=-?{IN$2WhpR<)^y!m-@^FnwQ?
z<%)T4o!sqm7q8pmI+~Zat@o6&?oVyy%X*!Z<r6tFOadf80!0xpJs$|YoUQdfubra$
zRaLL+_qb_)?Z3l!{dcof{m*m~AOR8}0TOUYKs-l?J?#3sWw!TCw$52>+0i{mwmn{4
zj`eb~Ej!;rd^xdTTZ<=9)c5I&Z#g|6f$$0BJx4PwA3hl~53_#1!PWaiW?r)t3B*mn
z)qHYg*L2-1m2wzue?GAOgJXu(t<0;g&+Tcy>y&95E=$P4j046yb<G59ony)DthrXk
zl7JrqrtjtHp37#<@2AzQrCbPv{@#(wKiPG|G`?K=-B&wJ@56lg$9lph5c_>Ip5Mi)
zrt>$Ha;Db-+vUQZW#M>dy5A0`1@t5Y0$J~;vgWh>o|5hLMP@IAz9`a%Sg*UW`<-=O
z<;_>**z!?#UOu6f_ra_b>d5*&La6ylp$~N05X$|It{<xNwb0XTpAYEt`I_?DdcS&2
zwi!nPBtQa15HLNLaeW^X+WxlO{BhmCuI<~pUKeF2)ct=}zg+jPYx}zU&mz`t*6ELc
z@Acnb%UDw>5s*DE2rVAWU9aP&_TitXruKBpp<dtXj_vx;4k`)cA)tHy8A^WJ=d5`c
zvMdQ?C7^r%Q*QZjoy^L!)boY5zob8I|IU4^d|l5&Ti1Kne$u{M0$I;Vb@SQw+SfkW
zw%;;)MV;5%s#kYBgqA<{eW0)T;Hn<%ON&z3yo@#87spungywgw`%?PN`;{N>nU`=J
zl-4*1ZGEIaaS<?mKab_?{++i{;~Cd1ptmGI0wh2Jp%V!0Jkpgvw$6Kf+3|Ib<*T09
z&zWQQU-x}R)BIW2J7422Z#^mwvh^S<2j(LI5+DH*Ac1luAiI7teyfgWZ#gps2Ulam
z*86WNJ9+nEw#&QXNxLNQ69L=fH5PkC-zU2seX;aIsr#wAFL#~pdYsAZV?Pk9y6zX+
z|7Vw)K9f>%(BmMRZ?VRkt`D)cmwuA~2^3F2bq<g>J#Pz5_r-bZWqH2@T>V~0*KX1G
z!$sAv%L(~)*N0p7+GCyRyVt6;t?<WGb=$%GgHPG!m>Ef_aWEY}S<CrdRasl81Y&vL
z-SoaRlyc>It`h3Fr(YyM0wh2JYbH=?`)Rhv?|o9a?sw4cHP0%4u|=KV6kV@B!mP<9
z0aw50%4#>1`{JFNJ$ur^s+vSsU0v^7_IWAV{_}zL9~?8RZly|X|9#?evFGc0Q;K6V
zgf{u=KU?+ao+n~ISn8_FC2xLl?B=!Ss=Vl@?eMur9dPNLHOo?8s{0;<?mi^z^Q+ME
z>2g7S+v9=uvJn9Pa?xy5n9DYS*suGx_Dx0EDe5|5dmPc8Ujprip7ZAx)%7c>ZaLWM
zRbTseSM{ju+TuhzehBzJf789siZ#EV-V47LU;7&0<CyhbcJ-LPE0fqbn9dl}`y<`@
zRM*#XPuJzi{3JjE_6THsFTq}ov{WR4t&*-&rfs+^DK{;WIqN0?k$~xQQ_)V*;k?hW
zbp6j;AItC8>C}A&r=f7r?SJMkmOv=uhJIw5$Hk5o`?Z(*c}c0a-~Nc9#UcsRJ>m7Y
zdhWkbwMtFb^CH{Ey3Gmr`kjlphifls`#IR!YqOtP)O%5^clZ%EkLaKH=3svWg=x_h
z*Sz~KSG3CQ9GC62dw#F``%YPYSL2&@OPfG!@0*DJmUi#3bwv}feV>unUeP>Qe^>-$
z`+?Zvi>8|OeZ;)?At-Np-d1$Etelun@9&9Z<>l)B+LztD*FV$pMXxWcf6W9;`!lq1
zS?8}T-<s)^eyog#ZhYyJVhc%t1W14cNPq-LfCP#p;Ol;w^{n}QjiS%NiX40YTX)eM
z%C-Lg?SoSOKhCRF8p?T0sC`i?>v-(*J(O{3>PM-JBeX;JoG)+wP}hfY@hi6H(xSf@
zw|O;df46Jh_1YirnV0z1tly>3&1bsq$;xHrDCRq*%~loHJ-8|@T1CH1!+C$tq1%t8
zmOs{gP2THaXytLAVLA_DE9Yt+>2d&fQ;l{0!0(;=-gtcTM{8Bbs;2!GO1Zq(ucFJF
zeorQpa<0}PUv^#X<8;?iS$?cDrJ}NS=8X?+KO*y|RQi$aSo?Qe@{VVg_d`H-KEU{j
zGcT$bpM=W6&*)$+BtQZo6W~5%?fVe&$;Jo!^^nJU%)Z6?_hW4P9eN*zy5B2xzw$m8
zF17Nic~aE(gj9Z+ru%+QZ@WJ6eFui0J!xT8P2!7#?f&#_&-Qh;D0{xS#af@~I$3Jv
zWb1?Yy{e+Z*S!4jiDz9`r!rRk^MUmr95bwLr7OjFjAj3ocORX1y@`dB%+I|1iYi~!
z_0V>`v93e(yV&bUhZghp`1ns7REDCVygya6AHLr6iLD;&E1<G{Vkq&p*Lj&e@Su`F
zNCb+0&X$)&QTtL?^~%<d(BfUSzbLzT?+=SAAKJK}KZ8%%=9n3YZ4P<YNtP!85+DH*
zAc2qx__`knnZf`1mX(L=b0XDw&Hwrw!~b3Ve$KYtto?7APxidXbv({*ZU2rH@52_Y
zG0Xw_NCG4fHUX@Yu_$P|-jBGp9}92ziR(Z3%hdICQC4rc#5dl5UiX*g)~OCnZJ$Td
zo||)Ca1CX?U|kW_wSTt16LZU6d#p2k_ga<IY#fSu4ri-g*Vj4M_HB=UnLTk{+X}np
z<J!LIxr8o1+v})nF0tBwpxM7k_q?iS`>I`A{kqjot!mxbvu)2-wSW4zszKGJRhze}
zxT9y)mOq7eZ&j;RtrJ&t=+d*Q_gzPhP3lx!@>a+5b{sq~ajkd!-^2qNyf7!VH837)
ze*3#uKmJwXOS=qtw*7bWW+kPjvlkq&!>Q*?Ph6|5emAv)@h4MLyP=f3Xs_Cnx-1`)
zX4RtoUe=&Vv&I8^rM9%__V}*<;=3Mc-s09#NmljkAL|Q?z(qgo-|v|>hNeX+XcSdH
z>;1X^(GQ`kY(Mh+#*f@|>5EI!QqfHg+WpW`jTfY)rg_KL@3-$hsmX*|mAaa@URn7@
zRsF6Rxa+E<D1~23myO>3kgJ!ars`^(!9QE->VDLgf3&l90$K0N*KQ@A)L8?o7G1sF
zH>s}1(?U0%%C7Ui;@z~Wc&mBW&*EFKx(B|-gR6S(`|RGfA0+dWQq$FY>A#_@=Ms7I
z{kKN5|G7RdyXO8woA2@9(!{r#(fDsETj$*WwW#+$SNl}ob`Kl6qR)1H>r|^$b}n6X
ze7l|6PkL<nqLf-*9Q0;u_S_GSJgaT<mbI(%YW5nLlLW#cQ1rfh(mO*Ibz65)Qtm1n
z5qbN^cX?a;G~afzo)5*E&-FNvjo+;Cdp%YAfipi_oKVd;WaaESpR4l<*mdo{%)YJn
zLQL)CJ(noDyzli~tWVajuHs#<w<`OwUZ*h6MRnCrzU=<xf%?rKd3=Lvnacc#<(#AF
zbD_NT+q(ad^`jWCn!ll?o4&6kwp$i%GpDBC#I9ebuC?gvVBY%kj&qck^@Fc?(|u4V
z<xJ;8(dAsvFJJbH=4h*a-S;z!&Y$)9Wa#;9ts}nHOV$$x0blEz?Rr9=k6Hb*HE)Zy
zled3e&j!<ds?4vf{mXoMznc?znf6=g<y_BK*|-fQzTDgIyC1~!$JoN5ogc;KPps$X
z^n3O1FvK=*MSpesQ@oD5x^C$D6?%T(*B8`h>$yYdcF^9~6zAo>)i)>cd3mXZZE-c7
zZ?@_e?btexh8@v9!bq1X^69F09~)~p^y^$+|5W=4^t<VPUu8e^baDSBhGksDe4&KJ
z@#4y_Sbv9TiwEXS(Q12snYJgE_nvh9RDCA~yj0HzTwNbk^@f)2`}dQw`p|Ds)wqz^
zLA>gFDynfLONafes%nRro^^Z_{T@T!`c?A}<G@s9`)l3!to@;D*R)->%lW=97wfa#
zo}%pKoe!$=S;wI&pYAx-&2MWzC)z2hU$fRL`irpaekGRpQa#5qZCBRu;3}W^e8W~)
z^jG|y#n|4LO@IBc^{Zui&ua%9_3hR5s-aDv8`1N#aqCt?#qVXzpERIWn;8|=GVL+^
z`1T{$`7wnaJ-xwzdWZdxLdAWPxF1ukr(!zRO_76Wx2Q1IlX6#E<6YN|>3SiS6UUdY
zuxh`pn=bC-T!n4TH&f2#KHo~c{ifrB<@^xPz5h_<_v7_{YcYM!SyZ|Idj9kORM&;j
z)(6plRe0L_4=t*_Y29jDD%;P-60bTp4L#k~exWEks`<wBeDC*X<rmBOjLc8n`x`O8
zYTfjeE_)6Wd%QRfeesly7smVfyLqv57T1YV4cmT?Bd<O2Jt9|OU+a)qkLmf5XvcS0
zJl`uiY^(jc<I7gPvDpdb{x@`PMcq%y>UGUacm5QeKlXhI{4aVw<{ei>`)|8`-Th%+
zexa3Dokug>?cXuYyYEK%qQ;dtK5YA8d%hQCujuQi?fQ%2XscelPZO(R-pf=xzZ1i%
zcB}4p#rK<Bg>|3j$oA8+c-!lhu02&hi0QiaUFVmLKht=zy$ECfEK~71lr=2!$y&}<
zKC~}3RkdH_>N>3Qhv{qH$D&`v>x1tw{GhH*K(^lbj$hp?{g5qx^=n>x3ETT(+FSkm
zOr_3a?W4<ms`xtySs1^2km)Y>Jlo~gy7j9~RaP(JWxvbF8ZYx(6)*m-Lk#DAuTJ;x
zO5*n-D%G`Xn&0*awRL~4doC;Hx7{vNd!k>lg>7B8v)U>8`Lu5RrsKkPxuU<Xg7&D+
zjl}y3-1pjAH$*$S{TW*R(C%B&&!MaC-<d3`Jz3Yq(DJz&mtsHYp8q1h?el2g_c8z6
z;9oZ_@3T&&FYUL-;;zSbuBcS$HTNIde2)j0CMq?(9*O%S+uTE2-{FsV-Etkay`Es+
z%KR4lHES4lP1W={gl;)q|L~oFa#dH?4{LV!TKn&Kbo)E&c$ei9>y!OnFY9>1byB6O
za|)zqtyktZ;$bhZX7wj)KHK|=y!K@4Y`MjY^FkKJeWov!`Qa<x_VwMjJy-o?%HP^`
z<%cythqY}#Z-1F`l9f~Sr|y1B{63N_Ec2(Rc-8rs>-3`99c#Vic0G07UfFwivUsdx
zy6c0O-_`gN&$DfZUHJ{Wrsp&)cW>JdlKF`#2b}xpD(;h9sp)fM-EtUzp{TfC#1?kF
zZp-ZF?LXOhN2$c;<&(9%tbek`tLnjZICj-tZ$r!P`@3uC@2vHz#+zzg)t&dMaVtyD
z+Mdw!**-UbJ=JqJ-F`xT-T8#^=UR3B%9~#{E<=mYdLALWe?vUR71Xqy7)MakelXp)
zhf>bge9-L&aJQwZ>xFK*>3FbRPM520eqAoi|48!|w~k6&anS8QF~4eGBuf|VFiZkP
z5>V|MOD)}WJt?w})>OA?J6O)`b){7Js@@k3HN8<)ziS5Wx+)QMO>;;mPIvzf+qU<q
zq1!{dMU8D<m_8R2+s&|D0<w9id+sN`-vTdj-l0<Li%?GHm+N@=K`pP}v1je?(61Mz
z=3mx+2{oUsb3$D^SVv7&x4yjj#p_4WVcB_;?ReAv@m0>%yi)m}cRYn!KGgfXqWp3_
z4rJq3Rj;pfnSYEgX8#%6eYG!tLwW9-)eqbAH?-@C?e@v+W#t`9KHYN-)BG}?w(rkE
z8<(d3=zgyxwEU*+4y9b|`wd^$LDsY8=Y6iuWo7HA7$3^`75&l;$2z`T^}jE>vFu-D
zeui=#DcTR)`>U+>Wc?CLd?@QD{Lo!DZRMB!j=rdP)pOHdq$|GLWBMH4v>f^~RxNrx
zvt7Th>#41J^0wdBx**!g>zA$ay4M%e^9)<%M2=!O>-CoTo^Ai#yjh8B4yNN&>{nZ1
z+s|KJ*|XJNUv_lY5#9W%>w#*0$$H+W8n3#3`_AvHzirj?i~Xmo?|c1wUn}$5*7rc3
z-}sT6E`4!HYFlM~BR=ms<oY@wvk(78jrAN$^jo!VAzf_0t8h{Gp{%!R(be01ljh~%
zYFvbFH}CgsRpnLR@wJui?Rw|vI;z?qm1??de#CYzrP?Qlo}P6dA8I~ve-K(&9OvCW
z*yM}@T7RC(6#LHEC$4PweY5eYN=?^4)T67|U&W@f{j%$LvHwlOru|&1<uGpR*6y_Z
z4tGyVbv5gEgpkklc}vmd@Vz77DjVmrbwjmo+ujGM>XW70_9JwAMgQ){R{gSZ7izq%
zcFXL@<|pHeCGh+0yH9E|p;o1-uH0<jeKt$#R2HSK=!JeCRkVNN_dr4kV?5M5XOn9u
zHXNVWR^QKqWc8t)R1(l3fcqF*D%-c(j=yl|lQ(odWMK-$eNtXM?6FI)tUGmD8oD0T
zw@=R_r+u5MblvXL9+^30VJfQbbBnxOW#wi2L&n$io7na1v?va;{>mGVwJEl9cKDn3
zyiRug_8lMUdAiE4qW8IZ>o4lN3M#IlrrYM_d!D(f57&jP*Ky=87tMQ}@>M?X-@Prm
zyzTemXix8bL)CjQp{Lt=Pt(+n>baflxmD56)lBQpdcCln5Buk=YWut(bbDfdiqC~a
z|Fd#b<qQ4!UsnG_Ua^G5evLJ3>-;cwJF@xqW|Ki(-??RBYFka$Y1wrxYrJW_VmWc$
z%N=&~o~-gcCda?CmW`{d@v8N(sC5VDfx4P|oH*_;Tld*T^$Y6N?HAMhzV5eO)#K|v
zJn!{Uw|`XY2Gcv9x8vY}i7O76hq0?GK6&lR_B-M}MK_$+ucFH<_d7ZFS-R_x?fkLN
zC)@tp=9|}^t^JcLJG$#Z==p8U_q_dHRC(L?6GgSdRlTbF%&g~hrukg8%l0}d+ROXC
zi0|^MevT#G)_UgKj;rgHFT1YpuSL7A)&tlrYX5DzJ{Dch_c|r3uV@Zp{TNT7sIPT^
z^{n~*-Mr^_MVBvXUCdjrt^Iu{^H}u5_WdQkk8||&1_SCH_CqSFcs-VdZLI@hdz%g1
z;?-l$+PG4t<KBF=PLqKRDt&3!e_h-DoaOZ^vueG=E|_reCJic8I{k}@4X^wAhLyV7
zs-oj(lk3&1%&NM-*ULH|O@BwKTdv<8i@P4%xuVjQ4!WV!n8jzUNOU#p?_hP?tJ`0?
zTul8D=c(<m?RBnbd$!icqU`v-KXz4LQT$BnwY8rq%8qHfLo4TuU(V-iuI7no*Y^4@
z+LPVqo5qXx=f1+?xv(rOw!i3b(f6Bq>(9DR4K<(b@u?b5p{=j<XU*qGp>mPEPva_H
zHJ(-Jrq_|Y<xGD^9Lw`JT|Z5Egi<cl-%q>tOE$mqUbkJ9m-%ZN@5;|m?dtB6WcyrM
zysmvQzj%KlhQ<7<ux>x7?1|~R_Eq_Jx#!s~x7Mv+Ez`d|P`~*jk8e;d(|cY!;HYn}
zu2*fU;&TF782bUdH;=IR`|sZKuDpDQam!Pv_<QCDN1oNTdCS_>qKdx<K5XcUKHK%J
zQ~kjoHrcbqL$&^+Vm^dnPv$@3MSo4hXunEj&sk*g;A^U(?lZIcRrL8T>;L)hq{zl+
zsPU%zfxP8(pD$(2uUc1JrHj{JhWGFH%o{^fT@G2-yHNARz7B@UHLG8y*MC{Lyz89n
z^3I;iWc=N*YM<^ZU3OnpRJ^Np7G+n((e}6&?-v<%`+Hm4d}Y`9qQ<Z7@d<mjpHszV
z4;(_%qQ;r6dS&w`y7-Ly2GzdB*7(Z%{K{5&TkR3;WZf5+T0YhIQKiSuLmUUDVbwlQ
zOt&>3LbapfXnWrTd#ZY4OSgRvV*0!R<z)BwvUrs9rT86+Yqi}kruOoFFFfz}(sh4V
zAm-2ddk0m%tmoRk^Vwc+qT9zR-b=>19jiv)AC~SjRQp6%>9+10i?S2_`1$wzcFoV#
z_zBgn>3+cW^+sk-mz&H^)_C3DBNm-swO<K6J(TxkP5p3nUz~Nk%JPZl0J>qBU03Jd
zvif|-yWSsWZI5n$g_2+Q{d(8DL$z;u-OXAql>J_){B`{bWqy^LAFi%9uq&>|uI?|y
zdVGgX-<wCdrq7M&`PsO2t9>ipcaL51UVdJ6ecmkEmz}q(;$_z#T)$+`^RmuA*paEs
zzUsYhSvlYFdDjQs`{ttZtNJ7KblEr$JzjU8o;SbleoL0$bY9?kADWt;JIKnJ^08em
z>pEU8`E2v0Jrc-5!1a13^TGChN!;hT+OLT1DYf;<)p@mRyJEkHVb}Hhwx9QX8}P}y
zZ&Q^=Jt~#$SIQ+`<fG!s^xNOP`th$4a1i<FhFxC=VPEA}DEG4}KVwUm-7l)*UA0qH
zFVoF_FNgJ@XmvGiY}t+dIg#)FwZ&a!M?7CQ4d;E1U|L>v?jK9Kuipu%#)GSL>_=>?
z>v@^?zDHHw6>rgQtm|*+&n=YxMgE>S&b3VUhpKUBYB#iUuGb;g_VfN8s;Kg|uO}*d
z;<}1(Xyeh;AJu%7rGtZO#d_ph#q)kKEIUV0?LS33V%Rm$SnZph(}vbwvA<2jzT1`e
z`Gag7QN^3CYr5rp^Ng)N+s{*V?Zw86{*u6&2-G`glWQk79G{ems@88?>1$%C6X*JV
zO0;jQ-J%`U{Yu_+u|C^j<@t%?I|}7mUL5P8y%!|<mpAPDe$QEbWj>z8ao}p)iuL#o
zo8HHvoNs@{`c%&?LQl_oe_C|;yw~lb%NNbhw0_(Bo1*NwT33D9)jh9~<rmjA-Eh(K
zC~y6tuhT03T>o7@*w-Bw$S)grw&PKscwbpm82&)<yKGUXz5mdn+MCv`mMQYXj;R*K
zA@+Jrd5h)pu2Z3vH{Cz5+_UY!n>Q=b<q-Qkg8#bh&YEAGZ?T5+?jLQn+u8X}#`nF@
z4%t2-FDKLT@GGzCKF@QNU-kZl?7dsl_n=Wu#ZUIWh^u&&KY7zl-`_5}oUH$2iFY+#
zeA#t1?`3xLj&oW0tn01m^D5bOLKdIZpS<~G^{e9Z)|<DyX+PwxSGT-uyvgEC+a)W9
zdSvrS7B92sIv)O-s;l!_U4OFvonc?=w5@t@pH!6Q?O)k=k0oBUE{B?4^!=Kx`gPa;
z(DD~GZ}ZlhcitCWUYwWW^#so`Of~B{lq#R^bKJcBB63uP^NvGX<yC&VN_V|~%W6Ms
z`)sY-u#>f3Tl*#0$y%>zKC~~Zn${C~IdPrM8b-Us??TU?G@w?S85Px8wa4(|+mBr5
z$F$Tml=C$7Q)r6!{$kZw?tf%{W?lD8^O=sftmgo(@@4fqwEGlYe{Ai`WaEqRHT@=b
z{W>Lw^5Bql-^6@o-<`|4E{pk0uV1?5eD$lYU%L5i@5f?a|8)7O@?(73y8gPhqiT<7
zIyjiBYQJM@S5+?Sd5!OUuFfG<Jj8U}b<h<jT|e^XchwH1-Hh+wna)$OobBra?1|$j
zZy4u@MX73iiY?vsb;h-QTkE6lenvTebN)UW^T)Slou978xmdq$zZ9KcTvxM(#de1l
z&f1Q0%V(>f%8gs>?M1&gs_J*mz+G1*dG+_(cc0W`Laj<u&0Bxo@}}==gi_A+KE<_t
zTl*lH9gKJCnhChx53ae^;$!o^w^Vd_*?12<-t~GS+BY3(uI|%Ce?))thHdx1=&$el
zI$3?9Us=Pk@83lKUER<6vTOUkH|sdi%?CeCxrp`Qd^hiP#MXS5eFtXbt5fT|^X7Mn
zN<}|h?<;KENB_pErsFbeIdS}E4QI6<YChZJ%(uNzxyI&K*70sDAFdCvs_s747B`h0
z-}A3j>QiyGoo;)55ckb_!@lOFimNX_veu*WS2sOtz07C!9;1v~-usfQ<#oqH-g)c2
ze9q_T!CXK7{az3HOEzv0FXJrUC(6RG6RTz&|DonXKW9}}=cT^v+TL&YvKP$yV7SiN
zK5xl7US;`g`GfD!6|IL2UD0Q|zICc&SMhhfv4vyj5C3E1F8W*a_YO_#&-(jySw7i5
zs^dQ%{@t8?`lm8gUB`6Kc|<>Azkb30aeJ?K-MhW|CRtU+P4#^jr2nJd4>#TU(3B*v
z%KV5W9{fYyN2~mj-6us8uh{pA^F6dM#*b{BLpoHpZ*&zO%RJQe)73c0+YhpF;3__^
zzpC=K?q^EfPBiQ7zt7=(or~nH2jf$=FO|i+`ksnt7ygOrYQBqh#d>XrW#_B5<8Aw`
zYfrV$3O!x-`k6PsE?-~y*V+2VCU+jcIE_tZ@2Q%`+v1&foXE=C@=s+aw(GL!ujzS@
z>vEIc8M3I`x{H4CUZ?HfHHqU#_MWL2k8=)D#e8B|%%3+5|HW|{%JoF_%XJt$VpVZ|
z#u7HYj)YRqH|JR9fyz(c+-%iXD(3~_{hF_^YFwz&#qa&f?z2tfW#30u`DH8J);KN7
zPAK~aQ$HTR;N+R7O#dW_Rb8Dw`?4#0Uh6B~);tcyj<4&Ls=c<)<y7`m{vaLqi?QjT
z8#;|yeAbGjR5b7Xtnc#T{dsI*)9Yg><zkz!vHFX3(wF+$XSl8h?^P77V!x@bo3_)P
zbN2jQR$up3uFuJB+xK;TfO^Dn?K|vhzfshBVe7h%e(-hQ;i?|n>!5FYwx4$uWiRh{
z5Pg@&_;;nI=bA;A6W>pgg^PZ!BlC;#yB|8L@q)A{4n>a(+x5HN7rVCaYX9TQuCM#?
ztk;29-aAnBk8B^Am4j|RTjN}}UR(KH?O%P{^>ttDsvh6GY}aS|`c#y?qV^Yg>$TmV
zDto5ukIHVS>3M&z5lVSm=R30NY`Mgnj(=4-oG-fGkA>DxDlV$@Ywka^`5q50O?;_p
zAF7+KYG2lLj2~)~W{n5-O2I9Z_1Cu_uJ)^0?Sj86wRJr!%1*nT+E03H`l6JYkAti6
zrLr5^byU|M-_M0?)fe0QPNn9r>-~jl-o=(~dQRxOov!<}X!}K7=XLAVUB_klb<fd4
zJ%7>l%hfp3osTL%bnWTp*X5?0U)LWozv;QBXvbAp)qlF_uKWz$?(kFZefYbJMx>>#
zy5qrDe$)L&DCM%AH^iDxcKr?|K9q5g*ALg@Bkww_d;Kld{CPPRUA|QNh3%^8H?iy2
zY0(^X$9L9sQ?zUAyv~&!-REqk`CUJcb!A`d&r%D!8dtvThO)mi^~2Zwi|=}5{hxRK
z=$-?Hp5Jubu-qffTiiM-(dCf!Iv#pH({WvNyG;G|UC#D?BJa8<D=%AT^Tw-wUtzir
z45eJG&y#JBOV~rdAnZ!>j!Rj2+5ABKL$mu_{^F<WRr=E8>tsNW-)vB+(glm_?(oq?
z>sG3?RYk|oCfBQ1DO26|Hna9`-s^?#^*L*P**M4=ubTgu4>DENqiTnWzsf$+F}_1l
z_~lw%w^!Ehp~ma-$U3iG_hZraRr6FgUF-+dIj5{2RO>_DbkR@SVc+|8*Y){6Ke1Jx
zDHq%0LA2-lJ$~EuiS4l!maPY|#H;Qb#dKTTU?=aqft{?X`+i_3`Ay#&3#D8r98LXD
z?F(F`XC2qN`E0LGdFQcddENb<@BHHW={hWaH%zQgOph%ryZ^8?p7ZY0W#z^BDgIpr
zUt!S?RanM5mU!L!e;E&3<4f0$@AXPHeq)PQU5{n`99q2V_cv+3_M|S$$E4&C1_#;v
zQ>}xh>yxZLS-k1_Ou3YkaTDXkeW`9(w8wCH5>P!CG))(2nyxRZa=QCdRsO8|a<P53
z!n*UgsQjkmg5~aQ`#~~4DVl?=@oH+v^c>N4IrKwp=QyIjS=X(+`Aqv;RZg|e%hFYL
zeW%O(%o?xjuWDb<bhG!Sb-C%DOUd%<_P_1?s{N5^x-QRH@|!+)VY#2bx0sbf*7Z}B
zPn<7V>o?73dM+mVDGSTi5yp2sZ^yv{lh`@P`aN&F>U^K+c7C^Q+E1p}Khyd`DQCLQ
z#$L|X`L?TmHoeZt%K7fkqR!E5*K4{z52ak*^NG;Qo386w%VB<J&2O4dwa-?i!%y29
z%KMsm{fKqn>U-U`U7xk<aHN0d+1I|`_Pj3@d#3AgDCKnf0ncBu>Yy7sjahuwio}<S
z?FogmsUNEAa_H&4)~BrXgudUj<)7(wR-8Y&Vb|kP*B@K?b-!QWJO9zs8w{v-*bk|x
z=H35h{r#e?_W@+*5U%61t~0WHrSf-xWbF$*-u3t>Rr|KqQC&N(cxK(dmP@{(&l64e
zD=4Qsu9%+$bO_kGuIbddw!E&+8%_7`;`@zR!=~$NR{Of=4vnh%T{CdkRf%oY^`qY&
zi@P4%xuP<ws=m*bH(mB!H(7jEzmd<>zpVXadOwzRoY~H|wz;1t?ZJ^}wQb(AcC~BG
z<6}9NwRK-yw4J=yNmbvwy6)MwD_%#lp8sM$8;XkUuocdG-V|H;y!WxT$3xzEnzy`a
zp8HC-J@0+l6UTqn^Vd-F`RZShlkafW^LX3y+ti+~e$lPR*SvCFkFD!xQFd(K=g?m5
z2hRL#abn8B_qna^yp`oQU7unrCu_fHyy-XzrJSwxH7h@B?+5;{f4^tm7@B6a=W3s2
z`~E&FM=@X4`&rldM88bKdAY{6F45ngf6totzC5(@m~UC<f$#l;?fPuJFDKf`$`ASS
zuJ?J*Q*_JQ9=E#nGryhRtCC|F9Am$px*A7VA4^?z`(Kv_uInn*wOiCVQmNK!>;Bf2
z9o6`;ovu4CipuYbw{N?q^UAlMrB<Kqd6U(i?s^|;epmYs-*#=^R~2n9>o^QGU)J;5
z!t>?){dwN=hfvDLdR;KR?uJ$l^FX{#6ul0ie$g-8aNc!Fw>~j{DD$jnKSVB}g)tAP
zde1%c@+x{<hE~67UgUjlX<A<8muY%Y_y4x)jdlM5zj2+7Rnd-EHShi;l=7kNYqR=e
zx*nRAQ>~M>)}^BD$j)KVe%bHQdJI3l{m6BGOp8|E?~U88&-K0r{TPepy}sprpVIXA
zsG*e0x}JuXPkhhERygl?&@Hd~d-Blo+y0KSYkRieTPw<5DEmYBVLJbc;^(T~SnpR<
ze&^kHiu;MY_FR`&@vxOHaw*lYtNEs~TNKZ{`_iJz+Zx}dcJlU{>-E96{i4T%ZvC=z
z0>=A!&sn#>Ld)-ae#Kf}-gOe?ZJ$&5vgdm}F+Cq<xvE81Z}&}Vivz}O(HiTzVaxBL
z`Lq7=AmHo#S){3Zo|`wnSidZs^_*1qTuyaAEvBpdkojNg@wV32qU`vZZ(==SdsX8r
zYkJ;!X{&sxk9)K)@B4$Xmj?$^HT{k<%4PkYa@Ksd#=of@({n^sIotQCzWy$ZcpbJK
zR&mo!$2<_%cZPrdevNAm;4f3#^FXvGtG8U@ZS}A1>!K~cLboICGh+>hGGAotzia<O
zA7?WET)(#o`_FIu$W52NxFmI@w$5*hvg7-C0qa}+d8fE9h%M}Ee{XvowY_gB6?+df
z`#0&HSM_XPwQH-LTJ7Ab)}1}u_H0%Ar+=#&RBc+dd8>*$dRA@uQ+W4QwOZ9WaYctN
zJ*#@(b>!Hjj_rPu+8gtmi)Np9!<<B#K{Mf^d*?5n(Q``TOIz>K{jGJkot%`KJ~8W^
zJ3jp3iNv*Pb}zGQ64>{t=dSPG`ims1YMQs6y?_6uqqcAHMQTf}I~cgZKC`d5e16kK
zNmk9;oW0*4-sH?r`X}Y0S-HfXugevK1~=MwQCgIO-)*B`Ib+xl-=@0ykHZ^2^jxPo
zX{>tDdL8cC|MmH4(TaWuU1j4Vbie7J9RgYRF=9SDwg1w}nfH#qsP_}Ks$J<lPaOZd
z9vg32EmL@ucc1>)Z&tm3^qWIc+Y0}vb_sO;zS%Q3HD0GWHf?!L)5!zQUY4rVF0JCM
zneLa1^4N4AW!v8izV6oe@o_8CqSSVKLa}#yo%Z$j`gBNQs-d)le#A*&^EU^6Ua@#V
z;#%|4Y4r47ZT5ZqiBzSg_3YbVWS=S1m!!Vc)L)b{y*`NLU{6;U_8PK#izRatDmmB~
z2TMnOS2600&y!fy);W1rJG%3g`Lhu?f4lX2+~2!@YFo3_&Rj(k$b0^q*Z=)?*s;On
zKh-3$>SO1;c=?lm{U*t)GLBiFr-hO)*7p&*%>L~e=brR=N;Tskj$hs9dL1q~weg%=
zzez*aMo(RJL7hu#($G}oUsQPP3zKhp`H|(RD=n&Btd|6I2o!xCvHd*Fmpwjb*BNEZ
z8w-JP{r8#OcbBC}tor_`!(aUD`NfH@hSoo^-k-p4Ti2D)`i1_GfDQrtej|3h^VUYq
z8oo6@iA7`Q4gYOD$I}_J%o{%e*Y6LwvLDL%a9%%L@rli@>zQPF4(r;lqU_tcUbwas
z+c?hZujzY(S<A`BRo?X~mh!Upmur01amjps-m8p_bFA+(l$zhMjhnpw7WKP@*y}Co
zIw!uLT&m%s_~xzG)q4yoyLtUCm-4#r8QJ1;$M>z94c>TqqN}#*7wzc!p~^4X&l=uz
z*zw0L_;PuotDjul=8r4uE={O$bHF&+?!t!kI;Ec`v1#L%I)3!^$JJ>p`hDkak8bwX
zt7+`I$KH3`vSi`YNx3TK4fSUN7hcn0Nt+`lB;}^E{(rpTkS&{zo1BEIdA|b^dwE&E
zxr#S^J`+m0j*rgXx5bEelUQ|n%eQuJxBq>KOkMT6E4vH!JofvM`+t{aRabuK+V%Cl
zRarf{{)U>r{U+5tPWfs;;z}F-@%Ra&P9C1-)lm7x?w4$zB^&p##2?>!$BTZu)3Sv6
zGl$s6sV)CQ-ABv(s(<px*PeKPVxm%8d~Dmv`}^pk%FEWp*y8ctYhJZIF0$H-?Hm#Q
zs=havH9adQ<|6?TAOR8}0TLhq5+DH*Ac0aL5c_+k^uO5etCN2*{8jI}+Db2#b&l;I
z0UrcHd5-6!t*j$Ef!MxF9^3O6nZM%S5Az+)K0?-<Gq!PK+ut=)i8t0ZAJ_JLeWx<E
zdWsr>Vt?EAqp0@+Snt}ud$0QYMNH53cb-f+*`8<RZckjtLJgODKe7EJKmsH{0wh2J
zBtQZrKmsH{0wh2JBtQZr5EFr=hi<X^&I=c$v8k))W48M4%Hd}oI{SzQ)wcB8fw!;U
zZF24EShc3x>nDBLd_o$V>i#{lQ1h$){*SG3Pdlp<=-+?osO?*Pk-FB^Yh{Q8RwwZG
z)0d6DYyRBSR9Cmg5DAa~36KB@lqZ45$M3YxqGuMTDlJbsY&8jx011!)36KB@kN^pg
z011!)2*`e)8(X}q-}i$Pl>|zWfa&k<LMdnKIlgY+=Ka33>+-(8ry#3Omn-v^H-Rp*
ze|yHcCw-pkYI##*>q&qFNPq-LAT9!}cj^Aty4y}pvMO5l-p~#E-d=lFQZ9;ii+Vzx
zC12HFe|j2=Zr=3Z3tDeJIn`CM-3*g}O9G*tf5!fIgX#Y}cYOH66A3w(aky>tD`yP*
z;oFp|#{p}k=|1V@+6x>1s_WuJSF`^9=&^o{?!Dxa#R+ta1-<7r?J_edH$^{%s<*FL
zuXg=^exKS_ReNRU%CX1K`(oD(KYDjg5{quq;mp?#Za6J*rK)i-=E^$f-n`+oRHeOp
z58Cmvso$qrRaLLA^wW>td%boa{ws-9-)#2!|Lyka=V__vxGr~(xqR`QG&bF#f6b;%
z2QNuo>E7=TZ*t}*{S)7Mz<zVzxch_giLTn7zrNe8s?T=&WcF@pIrog4Pg#{x^*G3Q
z+KxYTRqp|Nx89&ScI|#n_v!5(tXCbYe!AVoweR}ff+SX5((sP&THZ1%iB+ffIQ5=4
zhYd|)*HGTC()GjD_a(Bfmz!Qcb<aIl%t?yYtp3XKop@E}hBtgXBcb}`(0-HZ9;bXY
zAc<9dpN~HC<)je@9X>Xt{><UTrcd5~$@mYFSatI^2Yz0$ctH}1%GO8M@h5&bX`dE-
zYm!j(zM0n?{OW*ZNmlJM`-;owH(iv3qL0j)|NNsXXD6|0tmkr~-@kie;5NO64^Ly$
zQ*PPy$j9Cvl$xsR^&&R=zd!Yq=P#~VnHHrgUVA@xe#@~_mnNa>OCR<4?Zbn=PfA_?
zeAIhex4Cs$qN}QYHcdag&xVz6zx`1fx|+7HsB-DL#~r%rn2KswT2#AOuUi7X?-$&*
zrlR66KMovz^1<U%+v?iyCOc1<w97i*q`p<wuA}c5bmO(Bj7nnHL;C&wp7mZ|nW|LA
z!FK#DUk_@w-+Pl1svQpXPagT&6Yoz<e5vj6KzlzEaP=Js(e817J9yi=H+`GNs$zS}
zB`ljKS$}Wb^x+5lH0d@a)zwxF)_?noA!8F;8rts@bp7da#h}5B_Fa_5s-gehN;b}I
z#b<rLoB4*n_}TM|6W<(c^)u~|011!)36KB@kN^p|A+Y{s!%rUc+i_`Zy7;j=?@dT&
zrlF~uUZPbJAOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOV{MOuu96
ztDMcY(iRDj011!)36KB@lmh|Z-!IdxZ>uHaZ~geBDQT(bKMrsB&~u&Uq<J-z?@a0X
zVf*j!mf9TMZ-*TlO#V|%B2!)7y7^u4bbUV<+ux_?aLK8S=iK^D8oTO#uQrtYq5qy#
z<=@*+UpD%#`Eyg(>S{fJ-4hOZcu~uP=O?l1wNGCD#AfrRCBAfe%eQuJxBq>KZN24|
z?tgm!wr>;LdgHcDs<yjsa*|amXTE#h|7|@#(N&D6P;|F<nl8Ncpsx~Btxkqq+4XPJ
zlcLntcN2=TbI#nI-u=TTUnH?=tlz`7b)Oc>JQn?Uuf?rLjQI7^)RyMG4@7z0`)XN!
zvAwGBU7sA@KN&G9G1W_IKi~BDH@`|^)ps_3>xa50k4|FO#~TjWvgx?VNv!%{_2dJ#
z955k?P4_*f^BX&CF(uJe)x6A`UeoRMlfG;|AuSbkeV^poeyRK&<JiX21z&e-{P?&P
zX{;LS_r65Gmma#s?mI7BklNOL8;tBTW%`mdbj`b-tICIRE|=Gj+v~Kizt^Wj5|yfW
z_)fR|T*SA%JHBt-Z1Bd@lUQ}>$nPpfo$+~MOI711R!;D{^)B7tT6f#Y3AH>psK&D@
z{jqaiy!^?(ev`zay3b8x&F}j<k+;4(Z*A19;al^Qa#Pv-n%?WhM^~(~adl`a^EYq&
z+`3n8He<jSX>2;=o~s6IG;w%BEr!DnORqR&@T`{;TdMp0?a=dMUl)t+{r>PKXMWN@
zp^^i~LA>)I?|#6vyy|?&G~Lu6lnebjVd`Hj?>Va8y9+%%w&yFNzq0odVu|nHf9a_0
zTYQn)QmhN1D&9XSO7lK1FfA{R+oHppH$C`*)|*dGOGQ=l$M$?!*lWn{EtbqpV%Ksz
z7ejkx=gPLn3GI;p36KB@kN^pg011#lDG<2hu6p%b_Wd$-t);-Kl-gi=&h5IKZN~dN
zaAt!C2F*-j*Oo7hUi4$PX-Vuly~nBdyg6)W603fG%swywz2k#vtUBh(I_KWJ;j}bX
z?fB^IeOrupH=(YbgYG)ovHNo$?!D*YG*&hJosKRi)BFooZnoZUTildpRnvM<?)Nv;
z`^}zX-cDlCz5^cIWZ9SxlUTLkACI3f>g3^RtlE3tUP<$hrl+<Q<Iz;}K9`Zz>*{_-
zRv+WzBB1+wpxM>C{d&#=3zBlvCLcb%)oa5>q^|ULPYm3q*YM$K(Ry5ydcW&&^yoBp
zZE?xnEe758QX0A*^uUX6-S^C=X{l)AmpXp*^~cqzD-~DQx4-<uRl6KmQEgkrdVGgJ
zp5N?)Mgu-fW7Aq6%s*_zYeUo6)z!YqwcUFL>@{utreCMJ>U(^%J`yN`KrFxi5&blM
zFWt19?tOtQe-VAmI!S;8N{c|#hac?Iq}!C#mAd+!rEFc4#TV6Uw%cD+J4}D?z;b)F
z+4u1$Qd=B~>VMWt0wh2JBoH<M-M@c;{H|K{e^YMQc+-)ClhAZkyQYW!>$A~GtU9%3
z#Hu-`jZBJGSL4Q(-5#SB3>-LSenQR0!4`LwomjsY55Hr5|5Wtb^zSd|mQ#%f*Xes6
z)c({h=FUrF(c}Jh@V0et`Zjf?u6ffw36Ow$0!`0YwRgjc*-5PW&o4S2@M5E>No;!b
z9fNMX_LNbHFLghXSi{-~wC~gTk})^UPJHXyd}102lpBHGFPxt4I%Y&-s^8u*G^rZC
zDACp1kJ-QLA@9#lTx+@YC)-N`B;cRGiRVAp@$;+ar&K)-{(CCS8V`T5-7TA+x-g-Z
z7l$xMqdTv@>i6)8-#49<#;&pa9Z*-C-Hi}hwn^a9Q@S@g@3U!1ta{PC^B2$PIVC9-
zwe{YjFFX4__1yK{TYr(nrm?KgDnEVov#WX@>(}VsOD<WQ<kh@uVv`pJG`)VG4XR_+
zO+I;Q-=2HauZ~s!{^)u;?{<HKYS)@~oRoU`yHA<=-dV4FoI*ou?`<ts+;VKY;R#ib
zLuma-e@K7?N|(T|m$qH}+C4R?sg^F8)mv8o<dLsE@&3fbx32Dmd1#-}i*JAV)X8bl
zYTg2tBmoj2f$}0yZ_h4g4{K19WYzU9n0CZTqZcGpHyp}q^s%)du-%0X>vc*$O;j4&
zdB-Ljv|skl!L!p?^}V4R_PxFKtRz<5bicPATQPZQ8k@%YcMe3qpIlPsj@tb{OJdaq
z%NqTyf4@m-Z2JAc-aWtgd}<P_j(nuin0m*JN};$;*wSr=4SZ(#X}!~OQ(b;v{c7q}
zf9^gtDOz>?ww3?*Rqvc|RWds%TC=uin~%C(`~2}^6I+_KUfcVFto54aYk5u6$pg+_
zmS$DcdO|PPW%h5+IQOK_Q(X<cT}AuX@ZB38JLc|bY0+9V2i8vlBoH2f&6^&4LF>&Y
zr+GE&-vwhny}zSIUgV|oclSztyx3krtGn%U-zirtN@COfcG$7O<UiFU)Hpd{9IM`E
z4<)^(+v_KN*?dB3TfreVb^Uk5P3`yZzjW00Ext%?Yi#{fZvKkjRdgNxVd)iz44(CJ
zVq43t-(qX8?eksV_KNylWNiH$tG|tIIDg&^`_?3}XwAx%oz8gh>qJ*W@1K7h-teL4
zI?YL`t8=i`Z@09Zd&bSDtV(0i>3^-=@{dh_Ok>r~-#2^arpD`3o2spThaJ=3k+598
z-{Fv7HhvhVR1zS8;t8bd9(U-fV=AhBscDN{-JhBI6G}N(Ty*W4^0;Wd4tMSU`uwz1
z^oa{*);YZG*J<opck6=|ebQw~npaK7Mb>h<@1-GsR{N!rukT@PD*BwfD2YYKf4|}S
zli!$|#HvSadE@U^F8nTarKK`1*$xsQf$}9#-FwA>s|JlrsCpdAcf_#eBtQZrKmtAq
zjQPz)v(LL>P7<p=cFv2JKl#^hl33MeUx!fF@~xNM{l)WNrLk$B*FHJ(m0B~?Shd%+
zmtD8dp|et3ns+RnP-n?k_1B-C=GDA@u>9%-V*8Fg{f+bQ50q-Ztv;I=A^{R00TLjA
zVhEh~_w##PnT|<qYcZ_}x7Lk^-8-`N;)-hO+BwwvVE$nvUK^UmqOttmI^3B-&qyE~
z0(X4hy4m23rzgHNoR-s*7zyaUuWOor>B#RYMxF6_Lbc1m*O*)I$qCyH*x{ihR{hh?
z$8K=c(2o;eyh5+X_wVFw*LVJQ>-V_7cmI^?mxJ$f_0Z=L{UZUN1Wdo<BrDh9l2aSc
zx%Hb=RG)2PT_ivPBw&NU-tP}@a^@%f6W6NpbzJ{_X7}A?Y2r&&{)C!-;WZtWv^jD@
zqSA%EhV0&A$=sw=)b{<5YkPOz+NfE>x8^5Qawz3-AX_K*Z7{OWl<7;-teW+Gaa-$|
z>^lgJp1SISI+xU>uJy6sta|_GH;1IU`u5Y8jlOIC+|-uJ`p;K<)_yn57s~gW>Bl>F
zeE7o?i7yVm#;>d%)8{^=UJkta%)a9C`Arukv1%;up~(Dv|J30xe)jz0BzA3e*I!OZ
zXD&%%QH+188G%sy+1L9kzUm2;bGi6s>pq8e%+A4l?ZdLJf~NOBp_P-3$E+N5^XcAS
z%kt~iYb(F&b9vg&_WYyK(|fhq_wgsvSQX!6q?-Mm<-FWv=Z&u7ZLL4C*zt9)<hmYP
ze>W`e-;=dn{_*iUt+VKv#i=iKHEw-9&kJqe=8C5)yRv<pX}m2zeT^Si^{C!E&`r0+
zm3G1;;Oo4~RXwHh`x{s5K~}p(jZW9^`M9<p>+`X!<0902zRv$*t*5B@?5bYV-|2|{
zbh%>C;70o{O0%l&-}BaeFKOwaTkO8`!UbtomAzMFdS9x$AI$sRQ{8=l>~|ijcw6^p
z;`mX87j2%rci$E()6Hvc?6q6nTGiCmIjH!F>EiEr#ITs(cG&hgi)b&ezqgHk<&0rJ
ze4FZO-g>i^Z~E|qeVTNelG@U&^)ep`kN^pg013oIKzBbb%O6vl=_?78I)PB%<Mwra
zZtH$LuN~d<yiy;RvA0k5zG5u#y60h`=Fhs0<Q;FN_I)hzJS+5Y-oHOnbospR35HVM
z^gWsPTHJcXh+i*FseU=+UH4d?1W2It2o!Z+DEr-UEb+0vcLTrkK1WCS&}M}0`wyo1
zUAO!8WA^WQ$osRCSkx5{-RH{Xp5IkJ#Aes@{v_0Lv7S4K@0(-|yB=>BeBG__<KtGO
zv8ro6v~Nbh^gU6OGpqHNrLaE>Rbx3Hm-)H*n*%?uSiB&KRb}mR9skRb$`}$L0apa_
zKChOQcjd2ZyRq)~;P)*p=bmx%DXY?~>Y69*lK=^j014P7koA2QF`w@H8?L@DR4R6}
z_UaS6H#ua>BWqWiYSwy~ue1qd-Ph^nD{X4-T4(xv(04iA@eoS>{n||&TkkJ*s$<tr
zPv7-lFI`wsZL01@a5l?XzenIYUp8-<%QXSKC!(va{W7&L`)-9QUbntbzgLshuUOu{
zQu%4?JDj@xA4~pFf5(y6FW1knXy5F6`7&;{=ZmZFt=aaI_DFyP;vpbgzf|$@w2+>X
z011#lF$7G%9~pYNVra5f5+H&9C7}CUJe2&te((RktoZ-`5^!~%pt5W0z4HII#LfTC
zym$0Py`QL6?Mua0g!=o2tbVD^u|rQU>i6Tadd2Y}hE>mdN<CfVp&ORH4{j@7*B@K?
zi~cT<s(#h?=2hvc>y0Yi7H8Ta0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TLhq5+DH*
zAOR8}0TLhq5+H$)2>k2o`a9IUe^Fvv$A7=!`jg+7o7hrQW}|P~;*jYdk4cJFQ~$+s
zCk#Ghiy2GnR-3A5k6{uZ0TLhq5+DH*Ac4{*p!;1fT^j#5yx~L7b()i^RM+mh!&?tp
zK4@S{B?lMdK==D#Mdi=>dmQGwX!dzG%t>r>u=RVkqU;nkZgtoDQ1h$4lfrbn-_<G?
zp0?L1SN8J$J}%UGO}|QmKq%i)DGi!z1qqM<36KB@ghk-N>d6OeIbcFUEr!GQ1AF)U
z;`6BqwLCbeW`9k$*H8Mg`Gho9omw+u)tu8tCNfpEKa_OU_ftYoANfe5G4+lcmAcYU
z`iXwHC2&)6<}rU7G9ZmrxA~~swa*_vHldOO$3gM)pw`WwZ1&OD)k#(@wf7YMZ^{iD
zZ#r^tQmT61-_P%HWjZD`Rkv3&T91XmI}csetI4acr@k~6F7(qbf$y()>y`K3{xYE!
z!{OH7ytsMGQBx8sIhc$CTYuNF)a}?^HEE3mNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq<5CGbGAfBW9^s-Equc5StDt6f^vy0d57o~>&C^lw#zs!gjl
zZ&h(e&#End3h&;kR;yYkuISLEXI1aJjvSlRsW`ps{0VKxjZZ!~_^Lq<)t>qvMZBo<
zzMpY#z1ow~hc27<kA+)I`;Q)c-o;D4e&O2`igYoIc$xmG$^7m9`q0ee#ZBgYwc8h;
zr{jP3&555MIqyG;a!5!1-+y0w>1p*|Nko5bg;DQIPfq+`=V#wZVHfox9)4w2*oC4T
z?4ll|Z`}K`Rlh%Q@qc)Td{JJduqP^bqaIOV7v+#2@yL&OF(2YZ{qmtE&z)3zX^Qfq
z?*756zw6bsqPlPQ*{2=*$@2fSA9j%cuqR(`d(I^b|C27OAMxM^yD}B+qP-_SefNaL
z1M63d3V)#cwp?)f%ridw&$vc;v7gc3Dn&W?CGtVNXb(6+QC?R?zq6KyA1cK>5*78L
zztBFY$P0df4;1Z(isJxb)Q2$2BfRvzD{A-np>{R6LuKtleehRQ*hfAn^1<$3x;|Sg
zd1z&du$Yc8<`e9Q3XW(G6z$HcVmVwN#PaA5v7f{+@`?V5_GI<IA5(pQ(0=>v{@iBO
zG8O$m7=9ycTVY@1HRO$b``-Og-D=m0`elmiuuOHgSm-~D<4vl&@4e-NwW=}i#QB0S
z%7H(4VSXYV?Z7&NFv^MXNS7(rcPQ$W<wrc~Ls(SN4#HwRh(~$YMZBnDKIBKdEI-m^
z+F-{^uIjwc+yp8c52o?xKi&S3`5SsX*7MjD{e*QK{^L4?uuR4I40}*<{YQR`V{v>T
z9{z}S5QabSQ&i-GJ=r)E=MVD39_&EHaVWM=OozQtROE(sA-?WsJ^H?O?IzWx3T{}(
zWs3T+KN1z!A#j7A2<s~PALU?ARQM17#B_v_4`ER;eo>z|E->zJKQE@s!lFGfES7_x
z@JI9$=}?SkF(2yDRnZ@mNB{m|gPIGPO<9m2AGo_x>`O$IjZ?&<zo76}RJ2bPkNp7B
zF)qOy_OmL+kC-1^p}5W?3`IQbU_FMPqFq`0&>o~iF`h(4Iuz~4I26Y##tZy|Ut$>X
z@JFU7hxUs+5Qf4pj7#_{D*QB6(Vi-dcA)>n_QHQC(q)`r59umJe)J>84f+q|;UDrN
z9$`^2FJK4$`%?I$st@BqmJfC?&Jc!zgQ&18UKbIM^95PG=x3yh_EA3+`QRrM=}^&5
z)-cA0YCIzy^9^B9v7U(a!%wtFRHQ@2ba8(Q-cYeVgkcYgcu~P2G(~@lJj5{OCE~?#
zhIpBR2ih;ohj{n}#W+B^s91kckC>h{EcPSX2^H%>ewD&6RX#Bt<xsy&(LdmXc_QY=
zdW?AFN56>mBP{bnj7L~h)CYd>7x7S8Ii!mUzhMvMMZd+cSPtU@;|=-YFYH2*59LHf
zd0j>SpdQgK>J|G9?GaVR3;D!!)GJfuN4gkBJlZYRgE0J8t@lXB^$_bZ$`ASX`6|Ll
zmyLh;i~I<~FDT;0d>Cg)M;Q6geo>Jh@nSzC9^4RyisKODKpY>kFzU~%@CSCl3o7zQ
zJoq4tenmeaEVdK%#iD2j_@F-!7CC?i+J`XC)jK}-eBJBXZCEW+%!^PJ<3U%!0sO!R
ze8C6p23JurzW%S@cc<UleZ6Xgaesk%h<S%F>Xj+xJ^X}!P}o5j{SU={6YUZG1~1r0
zIfPLTiguwK?4g`Ykso%T-~&5Q)DJ~F5Qc&e%AtOwqaLJZRoH=j*h6`6KsxwCQC=)3
zhG7qW>MF()6!nYaBI|kwzfoS~j&!sG^@;V1@yIVK+7CrK!r+T|)Gt%iE2_*tF`sBx
zOh*{`vf4#CQAIllV?BU<DEbBIS+CD1k8)7NLlG}3{Dyz%4=DOeOvn7jd_jIF%0b0?
z#W3=r9{4S)j0^Iiyh_n-DDoqmb$nv}p}ZJIJQViDFyf)$C{yr~r6V5oh+)J-W$B1V
zeNdDW72FVy_F(-(JQV5RhVqDqA|3vqJXFkw`fz<hIh@Nwas5X=jAt<(;~jRv0r4mY
ze^4L7P{c!F2VvMj80n}Neu?`R#EXh@um?_<e^_@gf1pT5Sf-}u)v|K%SL6l1p-2~b
zp`EaUenmft=?H@}6!B2lfg&I3M;P^r^}!C};Xk;dT?mVc{7_uK5r(23F%16@kABCz
zLOkk)KTz0(B3@LKhr%ut`4I+BDENtrb8A&Tv{RK&Y&ZN-Dg0L{>K9dPmntmQ1OLER
zOc%q*hxsfj(xDg^xXvP8RMZ21ux=wC>VZAv6BYSH744ut--`ahxR<HuziAll1~-I7
zd*C3(>xO0a5f2sXgMX+GVR76b9SVEMkFc1Icqr^5pKHxpAN+$q2+PXjdhzc5*B^KC
zAGWIoSCoSyUZz;T#B|Irtb@phc;pwyG18&PCo0lqD%wRD{vfO>kNi+^9%5aEJ*10@
zde9E!M?WBpe4>iwkdAWj5AnzkMZ8MUPbde)cz|M^MLJZRPY5F&D)U3t4)}-i2!jLS
z#W3PgKNRs$ltViahRWIldnhNWSPu43FY?0<+J!LEVFzK%cPQdzs>&y>7w``%mKXCQ
z9V+^v8!wiZg~17Y5SA(IAU_oGG8O$tKY<U*A)iW-9}1qZ2metI>>(cc5D!H<>>|D>
zg&*)2{$U<qJwZC+kq=y82k}r<yTtOaqbeunLl}zsp@_$LlG#T-F&*v3ItUfx#rtbq
zhoHFrARo>d5D!Ius4O0SK;aMaLtzK`ptw&&7>fRe!Y=%VUE~)P=}<8p_GI-VKh80b
zAMv7M9-=<v7yD5a#(0DxKgNwXp2RTnp?=ICguzo((GJ3*!VXllBZiR=Tx8=w<cToK
z!!F`s5AnK+^BviEf<4ra{zW{vAPhx*nTr17yhY5f+W(>b;E6EWqf(43tT!q}ewm^?
zxQGfqP^5z^!mtAs$20iB4pfYXJ*0yV<^%GJiun#je%OP8I~4UG9bv35$Ok{r-^eG{
z1HVLj=vQ!q9nmhz!#?aFKjNXVi*gu$Q2hQ4>1Yr72mOU~*hRevqkfEg*g^Y|4~p_A
zC(8#OVtyGH#3R309`UGGY^Uf4@<UM$?SLX3VdMj6=*zEee0#kn>sEsgxFWx(NQYff
zkzb6Lh0&i-_zjhn%NjpKc&`a|W%Z$dMEfWw_K&zup*$3J#NSKkhQ)ehVfd+1_y?8Q
zgP*vLfg2R<%d4;lRr!zfy!GZSkMS;3a0^wz(NtCZuwKa2v|f~hU97XX4k8}?f-n^E
zP^^bCg?%X2ON4P<!}A&3?;u`Ol!w1aN4zV={KC8kN5n%B50%AZJw!ez@<S01MZBo6
zC&p(Di|d$JPE5}mhJB19DEbxS4f_zJi;8?0PslIEV_ae0!4Ad?>P2~|Xb17A59v_Y
zwXLWR{z4HC73)3ixXSTAR;^!+bklyrxCcKd#wAp&A6!%_mQ#h%5AXy2!4H{Y+(E^9
zzzK?U-0wrTIpW9p=MLQBKi```K2<wa>8Mwh59=(_kzY0*O#7SVe*S&1YYrHPS;s5#
zp`1)H9!38#UNC;3$PYz5m|sxT2fK*Je1v`6e<2^{C+v$V=0`aw>P0!!3qQqt$S*4L
z%i0b5D2M!r7t0}xc4HnQUMwf(Lw;Sw^$LpmU>{+zez9F*80`gLlt&nfd{D$gQ4WfA
z3t=eYaejgQ9MZ*h;M_-SCpaKpRM?Tl!*3}3`t1A8&mVo}ivN6f4&`AF`BjQ?$S1Z>
z?0=-o6!n7#xWNwU0~e^6j(o6(`p|yFLy;eL#Cla>_yNDMFM~ZO>PLN0q{E+IKYrYu
z-~FvY^|>wEUpDvtO{)(&vFU<N+wNSA^&ob|{1n@Pc<e`H>Bxt93KjDsjQmi<%M|uZ
zRc06U!*1658}LARD8>uM5ftMc^Ah<GM*X7VdIdWem*Tj_eJ$pNN?{KQKSYI}$Or$x
z3HcF^{Ae$@i@cByyO^&C!w&3=ige_Mf*0%~AHq=BL%%`cC(@y4ANCjM2PpXCJWj<Q
z<<M?~W$P5$2SvM}s2Al?KlV%L2b^<(gSh^|A8<tbVGs6^PgMAa_8<&D5Qg8P!XDz`
zC&I7`g*_<ZMTP%Rq@x@Z^&<>L{RpEyh?gn&p&d|FdGw1)VGnxN28Ta$){2EG6zxTQ
z$OlC_+9j?lqW@?o_@O)$;{)Rn`7oXk4@G^LKgbVzlizMLcH{Qz{O9|o@Dqys2#fy0
zKH3Zaq3}ag)Q|Syca`u5{vsdB!yn|s^P3_6e$IjX|Igl?NBee`g&jvGO+g3(h7uV>
zhENA0k)_h=i8YiVrDdpAvm7cF3{qDm1*;(9fPhk2U5W!02P()EC=rK7k$YsBdMHyt
zK@lWMF$)JoQV`oz=yLUb(#2Pw?)^K%o$~vK?|SCvecpG!d!KW2FY@Xe^-0#Jb1AO%
z>6fTK`N*aE)lagz?A9sIT8hV|IQN<R?M$Kf5!x^3!}H9(MEirzD>|QKbpNP}&ZR!+
zaoRWg>U^N*R9=32$DbSh^yfGJrF(#$>^)BI-01nCb?QqW`a=Dhv-;xDeB|X<@@t$q
zqxtcdhn*}>sekL2&w3iKj`^6Qdyn4t64I-$Ui)<J%#r`phx^cb#`DtiB@cb)Z(ZuN
zFM8`?NA}9}Xx-$A=4HO<p7R|}kL=<rTfaG7`pAFtfTup}(|6B#<%7?9);Hhzdj8Dk
zUUKI*{_34}C;E3Zoge!kzj-4)(yyican4HT1>gVr%bs@r)wj9+_viGbe$aZz`V+@a
zuHE;--tqg3-F(FJa}Mpt6212}rE^z0C-gm!&NDlCzf{k2ByY*jKB&i!jLwJtJ6CpO
zC!_sBdU^H3PM+vKZ65jDcjSrIY5n@IfBRzZd99y(_4B^tFQ2~kh4NT0nuF(-UcM=g
zy%g_$L+1v~3+YSt6V09*b8t`SbGqNn9ov87<(d5W*G^Vv;^gQ2*f;Uk$!?$d(R;ts
zyC>z*cc~v-ve!>N=W8unXRU1gs2{Z6{Zd}*l82r=ksrDT$Wxqi&QBbF>tolSI5Pjn
z(~DnX{mJ@3`icBDrSsf4diB<>W67T#o7cT#&g}e4_GurwuFCS!`yQajJ~#Z?antXi
z{rbC)mgs&mSMyq;_pkRQ+LxueC(7<UGVPCjTB3M<_S1Rrb58v#^Y8OhKXpoerGBt+
zmBml<VQ)Tqti5%}{L3l6e(lrwl)n@|<+l%$pZrVgdG{R1i~5`T;n#CP7LV;qAJtFK
zwLbWJ9_2wl4|>1x6UUD%FZ%N;_R_lf%Qvy{Wb4$Yxyg_8=2*(3zWi$`P9NBN+<Hs(
z+t(z^uP#4%$!HFx=K<w2FaA?r`e{z=@;OK9kjt*G{+>7I)O%@(lfUnL@s;&6^*QPF
zUGle%NfyVxM14+dJ^Q}&bAx-BU(XfUIUqYf=sd8?i}Ekg{Zh)K9$7we$shaN+6Qx8
zyAS4p@{mjO;pbkc-z1AK`I|dCeW^aOlS}iG*L>AmOYzR7IC|8N?|U-(J&xzy?=$y~
z|K5`xeGl7T>+A2A=<UDfNgT4bU;VM8yrug3bT9euzN}MS{~egQnWJ+pzxC=*A1I&u
zke_+W<Ij=Z@5o;t)yuEmcfI@#_xb2Y{@;J6FTXf)`)(b6){FK@KjQR(^|xOB=GlE@
z7mxH^hd$-kSM!>`{>+P?f0x=^=+&)UJv*}NN8i@7MEAfAo$kNV`SEijdOxA_(fzP5
z6V(&v{Xs_gyYJ4IKGhM&j<q`<&aFK3s88#y?7dE}PxU-E)wAnYJU#!~^@ZX}`$gaW
z>GjW!lbzn&Ja_c^QeXepV}1O{{N$M^u6BK^JNY>erT)aBI_5;4Xgz3MNH2fS9hu)m
z@%pfSGJDzi$gjUrzKQB${aU|v>(y87ji2Pk?N@dV+&jK!oO92!_q2JLqj-M$rAKkz
zht4B^R2Qv_z4xYh%73!hKFQ9>VX@EUl(&BRF~`dE6YaNold*C1(|+_^On&>-n~QZW
zQNL4P?C5-AzpJi#;w>NX=wp9=Y|Deq!G5?e+)t(NaPMidIQN@;?lV89@k8<AN_Ic@
ziCdz&`Y7F_^!%+${=MTrFQa+z^K-R6#aV~^?o)E@^vEtB_W7S=&n4dXqPw1Z%~#!i
z_gh~5#w&jOckggux-Zlxmz}fqO|<Tb=1{wQ=3GDfD~`YZOZN}D*Z8@g$t8b&WaM9&
zzBE_&VC|LNlU=X*Siij1S(zS}?EIGKd6}rrQPX_Nsjuqwx3pe%b)3(M{3hB*)JN&Q
z+LT@AB)bPEdhRB=*C$SS>EE{Z@JlYg=(f8!&5Pc1QQ8OVY`;@~)w^#!*X}=@ejefP
zoFISl#6JII{fMLY{5wD9WBolx;_1oqPqZ#^{OIK&>(9FD&ra|A+dbQS{E)vqQyu+?
zL-|Yj=*g&#yk+Cu^Ue)>&#nCO@uOd&bz%MG!`dtJ>wZ{I{nhQdPPXj%K<Af??hVgR
z_lqApy1&TSeAVyezgy?udx*aA)wA!H-Oo*Bb9HXWCA)bdKQj9L2btf=K3At|eREn%
z&+SpsIkIogt^Vjs=R`cN?bAIfu5^#f#~<rYu3g{O<KFPy;rzNM__KRn#mQs+?qBCk
z9&;2=_TEQ&e)^#=`IY>py!tM!7un^z{|g>_#pBPr)#`ue!~DB{(?0TN$F+KXxD;Qz
zICPGdX#dU2evr*`V)NI}dmN{E+9zxt{phcBUP|YXzx?u{IjBo6tydh1L;8|kKb7hA
z>3e(1BOcptW&JxR6ZxS$<fS<K%&&1>k9AM}jVCWr-{w>5!<_wo&i9df5bcXR`o`M%
zd0tCDU(j1WPI~L+hxDC4d-djH4*Df~U+4qnE7eE&#CgxS54>-XU)etO=lNC#*~QV5
zt&3mn?UUU;PjT{Fzx_2Qdh$eh)J5_76HiY*YU*#QYu)y7>c=|d!H2)-```KD*WPls
zaq`mpd&_10__rT&{hdeDS1G>F6+QO6*iSUi&V#;eT<z_XERMWH>q2v!;_7Gp{Pn4i
z%Jk;to}lM%PUPmNFZq@HTBovp>R&%H@<)34$f!TEd64Y`*?UKv|IU!!`^%rJ$>$!F
z$NSHJFZr&wJnH-}y2;&kFT3SSu6f%N?!Nn!tKR*L&wB2?cJKO-vw!2-@4Uw@`vZUR
zSzrA_FFbeGceejN(SL7;p6iDmd)@_i|N6hT`;H4={-DQPaqp4lRhnD%Yh~*;FZs-+
zdEEQhc<0_6Q6BrH9<n37eB#LL^qvFt*sX^wKY#B{=fk>DzKOn<*!`SYJv*A4e#ouc
zIkO{szxUnOs4h-*>CMZ%g_GVsAb<JHgKX~IANePWuU*~R=}YH`KkCPvCVIY0b>u_)
z)qLIeB|rH}`&@QTr>g9{pgET8OY9uTeNM>!oX`6K*~M>4^LF3ZH+rNmm*N`Fzj}G;
z&6SLko*m^w`X%yT+oyb<SF$`u#kF(LPbnVR$)$7U`7hlY?x)(l2b$NMJ16~APgakN
z;@I^==7;p+^zGhbxBe}+6Zq>1ILFSB^Fu~@^j>i8>G_w=AJ#5@x_{VFJlT0ib$tK$
z?)3cnPU4TAGtZ59ao$tjTS#9&b>&fqpFZ_dvRg0mlV3h{#G$%m>!TM>M*d`eSbg(~
zNBxVVN9SsZ&h3&PJAbrqsBiP^cS_eg<rRNaG$%Ai{kKnc<Yyn`!Rp0f?PT=xygB#x
zT=eMqoyf0w_)m8J_6hymsNd#s|Cl41PbnVFiJv<1>>dB_BitA4{P~fcL;jv`&p&_c
z{>rZ}<fjh5we<TH@#gB!G4!oZb?sN{n4kBFy!0izyrul|k>wFzvZJ{6?L0YW>^(o$
z@4su-r+9xK*!=Xvj{33B{JgK|-8=01QrA4><EKyY_L1zKB+Dm{I{Iaom)(5Km7TwH
z$zF<Am%ZzvSGUw>^Oy&{`O2pbx&G|ptgG}Ki?4t4^V1hQo-SXr`Y)}+Iwx72bAbC@
zzjJ`j3Aya^t{-vwA=hqw>}2b8e#Fu9b53fnAAk2k&!2c?uPkrRtNf*TnS*)J9~R9I
zo0lHh+mAlkQC!JBk^f|G{rZV7-MiisWcy3@=P<q(Cc4j2eD{;=d|>bG`c)REUv}rK
zbPkcd=bv3ZtUrCpPk*&fa{a9fYbQ_C2k!SiJ80b0pLLlx>X%$D-DlqSt$WaQ>^HuB
zcrK9LJSy{ZZaNS8#+xJZBbV$`eSNe}_3Z8yvb<9re*XO%dA%ptr}+9g2mHI<%AM<y
z?4ECa@_uPfXbxm~%>(JJ+s{klklzw}-aU6q^q%Q^$h@q_pX0ygs`Jl3@1?ifl~<qp
zr8?rRQ(v{yBmdIhx9Yih@B_c(0pIwM+wS`BcjWK-_bazAdFXNJyiRuOk@v9q#xH;5
z>z;J>`*-ARzx3wqJUb6$bLHoIK|cP@fA?P;J<9Jp2c6S?ABbaD*L=<MXJ32ko4?@J
zpSoMR7saFRCf`@`=$juqvYR8i55?(wqB*dW#j7{b{;-ReZ}0f;seG@pqxG7Dxm%As
z{H<FYe|oHc_43+B_0ajk>U&R;)sdg9-}Xg+qG&w_rSoAfevjwe|MKl``RGSobl1hd
zx4_?jUxl8>k6nJ=)i;0pU3UFGn7?~@FaK@<*?t_9_V1{v&x7vEQoQr$yjhp$h`n^4
z_e*^_FXV~t)wQ&q(tE)=k={D#t&g3I{7QA#_RCM4d}Q=J=XonVC+^?U?_aylsb2l$
z<6m}8`qjU6ikt3R=R%xwp}zBJ{q{q@-Y@dYgZflQKl0F*`sb&w`t4Vq@~Wdx>%gwl
z`%}KoiN2ijt5bjTrdLP5Wc#+1UmbqtjrLvLvhkJmsXux7AwANU`lKgMRG+^-%Ke_7
zDX#gKWb5XqU-{XsliqhNiu1hrUN;B#k#i!S{OZ$-_j99j<>xMW*-@SS()#vG_mO#c
zUY%!p>$48D{!J<Ge!1WE9#x$4a8&t^DzBf5?7Q<#-~G{t^&&mem#yPmIX|8o=g;$r
z{7}8M)Q@wI;;ow>nu~e(4ld={@<Wfl;5{F_{`V@|d2HvgpY!1PJLvh5$Ma5}sGfVr
zeMFB_ynDjEfz^}Ui~Vl&douL>qOU10{Y3p^=R%L{WTaPzzV|3SKkGI>cH~E{UuAwv
zl-IoFL4CJBcKONt<gK2JtxHDv`(CLmU+dDB@=oM0pLLZ}eD&riulKd@Y0s1INc8ie
zbJaYpvz1S+_lkY9?<Z@2oa2f5TT6K+fA*5UbBvQ;^@lC%XQF!gocyX6hqWJ6mKWvI
zC%t>reR8tS-+t>_2fBZ)x9ocPWAjumpFI509LXpT(xdMadF>OKA5MAL#p}1^*F5!`
zWbx=dh5r8iOYZ-;4}JBwf7ajl^CESv&v%@kXW9L{s}6nR^vRE%JhA!s$;a+}jOI&L
zujiOw^Vn~CR3GJMM|$L6d-eL`$By(k>E-cV(muq|qr9bgh^w6}52`N?`FkFnBYKp_
z`Qwl5SiL;@;HRE*p)YjKoipczU0(I*`IYimhk1%;mq&gw>YtrFtwWsr6XjdWcR%^z
zxBA{U{MxZojpk{;eSe$Nv_5k}eq`T)$ZqcDO0PayefzIZWOpvfOO$`2{wF)VI`yY_
zUhCg=inlJ5M?6+9pFHMT)=q9-`RK{~^=BXXohfwQ{QErSINeX?!fzr!b0eFZxvC=%
zfBR04&Ka4X{HP8)nZ2~{^43n4haVYR&pPR?Z!P<KkjhVa(+ki2!7u*99n>f46X~&i
z>r>yTU-PFIkMbUr`YnC8(34B{lAr#_$iL^3y>#DKU%7eJNAo4OK0V5(K0UTRS={8u
zUW!N03)1hGtvAW?*G`Y@WTY?6gWg<E7W<rgzC54p*ZK~N&f|XbxhI?teY-cvXnlU~
z@8?i?eD{bai)$Tm{Q0Se-ha>8``8t)yx^X@>P|Fw^C6p)yr}+KiYps$J!l@*V;((6
zr;BW!ryAW4rE}n1u+x{;Re$$ByZeVM4?lJ6I~m21d#=dN5z_ZLTB`5oV*h=Qf49-|
z<QzMP@|F5PcIzP5PJh^J9rKyUf3i>d#*sIrbA|TNIYIY^ebE=WWZy5XL%e&6TskKw
zik_D~cl7(E=heCFzL*QWx$~dcIDYg<-{(g?e(Lfo<>@|&o9z6ZJERv+Mtzm+D1M^z
zU%Pm}2XasGXRk~@<*R;LuX+>3$;Xf0^Go0Rm&|UR^zxATS)aL+ae8j`$xr^$bJX)9
zUR{)ry|U*^UeB-m*f?^jjyU=2XAbOSah@N2EwOVV?+t&SX05(`Scmgtz4S=04-`i(
zoulgIE2rnn{-e2)aniHPZ-21+VNUYZKFQ+H^JM+jYpzqCqtfeZZT|XqUhL*%ee|X0
zac%u2Kl$vhdiwF-p)b*Qh5nrr`Q-PWbgq$IU44ruPjqiM7xIzitt_88kk?XP)UWzo
zZ}al6zB0dw^*=1xy6mg-Ykf#>zWmAT*f@IrD35c3?gMm>vZMQ>6o=}N{oF-%p7n+H
z!<@+GPiAi&eiOyBJLk?TvX_moeUj}*?dEF#tV<v2nGZWQPJZ^1zr5_`Kwp|O|Jvn6
z{*~=Ndu4j_QqOs0m)BhAQCw+0rF`|Lw;$%jPab{nCnLQ$q(}Z}J?zNdeQ-bO3+qR=
zF8@9A61|tu??sp5mh8U&OY0PG9{&Eg=Vv+}{{4RE%so_^Q>orW{mJiMG#7FFoNKh7
z_HDmx-&^_r{`a`f`I#~g&+DnW9{oAr&cAcs=YZYwL+-x0FX^#!py#hH((}XG`JsEC
zKRwdRgVsAy+_VmMRImBXTb}w)dULP8{Pbk^Q`aMpI;Hu^BQNqNI|uaWKKJi>xNp55
zya&`l_03P7z2m>X@ZPa5_2u{PjL4@S{i1xRPx-yC%xk|?|5T&7nWK5ppKARcRzLQ?
z`J6Z8M|Li&m&bj|56x8`r1zYX<wO4DiTaxChox7?ddW*{{rX!!TDSS|r^o8a*t%rr
z-gE8TtKXk{=}~+s&qVc*A9-Ty=mXiU50~a(ANiL{@%7)zr`F#u+wRA9KmPjfk-IOC
z`aGzE?vKj7clhDdr}rxIM|$@nx@X;!<leWw``rKHQ622M$fdd^KlN(Yr~3TK{HJ(!
z{y6EU{QTL`9OT8N{quVv`}GA6{oD_J%ei;oZ5{UJQ`h%F?UniKn_c|Ww|+|g^2&$x
zr$;~kkk`^aul1k&{JGX-r?<YY*YnG6uH=$Gzq0k%?Z0`_%TF(UqByi3q?dQfQ@wT3
z>sKCh-tT+SUC+Jdt8V`{{ywDm=A%C>`uWzmKWJX-(>K!l{qx#7eva?G>>kA0)mvL<
zQ$P9buX9uO9Fd(<{iA)5M}Be0k1QYR&$`Gc-rU61Uw#x{S^w-4)e*1W66F)`zN8l~
zuleyOBmYy4)`NaOsPCoqS#N2d?Gx6IepAYi;^pr;tfzYZ@|06vaqQlU^y-<Py0!D0
zDBk{%vCk{NiQ?GF>bF17Eq@e;{K)3gd}LJDe8~2}^P(=f_dI=RKJqpXd*jIUrx%Cq
z#~jqPzUsx(qd4rm=u3WU*>k%#ZpqL7_gr*+=F)uDMNj6xl$XE#Qx~h}&puI{b3o=#
z)+gCKP`xGEAAe4)F3Mk-9`%RLUG4H9JAb6FA3dsPo%Rjsv2o&2eex3JHFxu*M|s$n
z=(|$Chs{mbssHvtKhZtwIdg8!%X~`T$^JaecfH?Vu%rKu!v7x+8TFwqSw3~x>E-9I
zPwSyacI)(A*Sz+hKfC#oJ(ud0=C02rs%L%r#_2ty&h$LeucdsY^_TiZ{Z947PxeVK
zZi)6m|NP~h<h6Qvu=_^8mg=H7vipFH{MJ&ObL6@8T$<NJ>#SW~wEz5_3$pv)b3`w`
zR8JhzPqfbe^pQ{6d)hDl(c-^v%ip}!(I2WucHYX`$<}R7=0h(JnV)!^`VfcBM=uZi
zME<plFTEePJlqL*F6=wmd8O}tVBfs=x}V*@%HD^ib15J8_v!TdKze!U$@<j~ySnuJ
z*yUwMcJFO@_<Ihk_nknG>grb=_lA1(WN~Es);RW3eSPVl9_dT@^i{H#^6=NUdi?oy
zo$QTwj!|9|C!ctF@)G&Wi{i}R^FvQA&0k*r){oBB5}nJ|Io0K_x$^R7oO9tFci;7v
zN1gvgH@Vx!G#B$TPv;8x`FX(nJ%9Yrd9tteLmm6?{*<3x{6zk~>*ZJ9I`qe0^0yu`
z`aV;KUVXB>)w3ge>3fjh#6E}SHTid4>gu;~)mP>}<?$TxYkhk0s1AR6bLsr$LGdV_
z9jkYq{r}9ctLOj!ip(w_((4!X!w;QLdHGpCKdfDR&$m9X@s;H*ovTva`qQ@``pLif
zo$o159&!9id62y_Jx+6@@A~vZFAx2G*?Z4>sCxJ4ME?3h_p&&Cm6!DF_QUy7hhARm
z;OF@>-@TQ8UtJ#52ew}QymuzOJf*zI?)Pccv#W#j{`+(KlApY_ll}b6pT07E{rQ#Z
zBD*|fc6$5HK2aPy8GDZ8o!E0NkNl`l=U}SS`=WK_qj!$wQ^$FCAN0Q9FHh@|*-P>K
z)FJcdN9ITN{-NiO^!&-rllXoISid~>1*f0Z&6gkAXMNBkyLkQC4|=jV`K*g<j{L;Q
zgSD&6F7B`?zj{mg`8kKyiM~thn{!QO_kF}}Zsu<u@@`7;C_h;~GMYy@`HNdic}ns0
zWal2OgS=nLZ(rn*pWgYk-|U`OacF<Vk;U;Vt-It$uP^t6yvUB$X+P!RFF*3vKk_Hb
zGf_VF%IfHg%zvUh`XNvL^7He2tH(}%qIk0E?st9q_1<qD_m+K_{OKECJsIo2R(4-E
zkNsbpr}5$@J^Ms)`r}7tr!V<ce^7a;uKfHanlC$9{K=+sj8i}QN9*W0pW-LIy7Eqb
zhox5so7cMZRr^|5{<87x?&VTH{G9j7-3Q;p^xjv^Pv%GVUh#exPfr%--qSC#lTrWJ
zxsdfE4_O}ik{w&e{?^V<KC*f4m)0X6(wFQLorA{dyZb|Ly?(z)R%epwCyFom(~~Fi
zvo3zl-QMxv@A!SIpC6nH?0Kx7-CX=0$y|J|`njZfb$x&O&Xlix%cF00l;86vpLLVV
z+R4^CQ9r)t`KixNu3eqR)AO%PkLKmPc70^~;m=ddi=J%0exFwITTAh+LpG<QqB`iD
znv40C`Y!b;e!o;tzVg|}{_LH1dEz<O|GhlVbJ^!dKh|lVtRI~l_c*!dij4Z^@A-4@
z$k+2I9-Vvn^o{1>-yy%^RkwQD%|H7Sum5-c)i2epUZ2+Qxm2G&J%9b0bKe`*?LK_&
zE8qW>-~EH1wEJxj{hsGu_F1>t)lc_LUUkXshaL5$u6fhTQ;OFIe`F_@?CLpJ_Q}4f
zE57?oZ=T{kXX5pFP?~?|wUpOB=->LQXQxN=@Sbz-tdDGt{CdvKhks?yx$i3L#GY5{
zwvYVm%UYU;^@yV<Z_1;Jmv^E*&BL6K9X;>l&Y^mC_1WzMxpsPFNAdbb{*~!BW%H2H
zx$HXGOZCOQ=^6j_MYsNed+y%v)lWX_r(gG(yXU;}!Dl_|o9}!*#alPZ<DBzDerP|-
zC42pkTGl^WC$>Jn(mAdFsV2AY_P1oO|B~#v;phCAFRDBFvDd$Pc6m+~t+VT2>StPy
z=f?UDo6WC}C0ZXpeUWi#-RzV9q&MfKczx<moV@h%VC|LVt-tSt!=nDwSC_2*61_L%
zL-P<%&;LHJ`o;@xeTzHo{?gOF{3*}6_FmV&FZdsnues{{^Ur(f?Z$sMqWRme^XvP~
z_nP;cd13QeukT8<es(l>Kgam4^qtC%@^}yXE~oeJK&tEa1Lk6V-ZS!u7pK1b)~iq7
z>GJ81pZxy+PWiv}^PhLw+pfLiZYfS4bCn;>P5$b;4zm1Y)R%SpIl#KjhaLHuQ}gM+
zbl&9Azj%6m(IdU}ljSiVGK$ytq<8N7a}K?EVCO_%e}3vCKkL>fvNxV=z4q6-=$$9B
zI`sC39qCULoeO=H`mjE-bvOQGtwUcYntcw<$sEWhy05kM^yj(OH?m{(WORO<4`i3e
zbLJjlFZt8UPoBuH_02_`K0KdfY<+R^SFT<>J?f*h57m>c2i4ba*Fkpgq5IN(hV;Ha
z`sV<C|HkgU@4j>&crN5a{T|kL<)(SeQJ+iu!GEHCkY|ZY``bG9%iO&m&A0nIodbFJ
z<CLfN##;w=AG!{*xs$PTGgtP~z2W)teaKHf=S~0oOMdi?w@!ZQmtCK_@|i1&mtTGR
z$WD*;jUT=K%wbdNbBePs{7UQ4mwxE=$xbfWty7(zv(sIE*ux%i?sdO#x9fjywBC1I
z`0@um=8Ajoe)I>fyzZRWe#Z6oX<v>?-}9`FJ|{cBB|E>hbPh`K&f9cO+4-L+>UXN=
z`CzZVIqKiN(l~bahCIlQ;yr)j%n$jk%_G0O`t_Z_?z@1VjJ`9()AL7m_bSqhBctDY
z@b`Y@kKGq@K>I`HPj8Ow*1tsl*!5Zmw%^M7D&4o{K(>EKUy9@B9O+ws{7@co=zd{0
zCw6-IYp*O0+0pqX%Zv0~mp+=8T*`y&Q=a-&Zr|*io;UlkMEk;zY=6<u6a1{tdF*q|
zPdvZM@_KGP5B&Ahdh+;Ate^U&yyBP0zg&uAx4x1c^+{IGT-eb($l}P(HCev0&x7;p
z+^IufuJvo2yx9F~|5JSRmGxKlIcuEy{Mhv?AHDv~n|>lc`S__zF2!}fJqPS4Ke==s
z<iq+^ZeI1pqqx?gU!wcK{UQ&2W%=~ied;_bi|fA2gZ$9Gk@X?JxRPBQiYJ$~ld<!s
z$0?t<?x(n}b19EF`y@X-(wDW%GwGcZ=cDwTwocc}uFeuK|Ak{W`M{muxMM%3eRkhs
zzuV|h9(kQh`cfQD`Q=;7wRM*K%+<WjmA>(0>mlo#jPxbD^Qta7C;X0z)A_T0&zHW;
zN#DpW4*Azk-jwpHi}JJ{Jz6hWKJr9<p7%|^pZJUSb0EFC?DW=0W_Lc2zIn*{NBP*v
z?DQo+^~_VBOKiR+x&7E*eKg*EtG_;XWaoO4J#XyT=bj((o9Mo*on9XFa~64tQ-9`S
zo@joRCq2KZkJg{o)%Z=<+4xOm=L5y-zwfC@_WPH~PA~5g>rXZx&zpU)Po6vLL3ZcO
zI+0#H(j$GzZe9A+H@$jf<k!6H^kj6dtP{=0y3u{h&%LRC_lmjFyJzV2gZ!*dUR0ky
zy?mwq)D_Q<jMcA|<@e`)<o2gd=c7;c#8od&9{Xcn=ErVM;@n&8>P=LiKk~QE>ebbs
z_XE8?QUA`NIC;hMM{%ALeTi2OCp|lQF61Tahn_4>UpVR6%@@tnI#3?6y4F*fKfV0U
zm3g8(;!wPOsa`+s1OD{tA-#UYv0FdAdi3Het81<kotN76qaOR@PwyO5c3#-g{cN7<
z(2J9g9_f)@9MYpY`Vh~*oc!bwkNi77{&>(o|FxbIMRPe(^_J?Gqq&oj-hSwx9r=0g
z>3c5e(S2c`OV1&_=anAWE9-||Jkp~)6ZNbA`Yq`vyZY0;u~bLia@sF`*3<m-`YOeX
z<B#lR{hb?hUi6L5ar4<<dHLx>J~UtZ-+SNpjDA{=%&tE_XDpFF>USc)iBlZ?6n9X)
zIiPyhPe%2~*n5Q^@>?SR?hCzmaw(1<xqj?u{bl1T%S(^uL7x2BC+f#@;Q63$9DOOi
z`o@#_O)|Z8b-nC|Mg4gG_nTK9G*9;;y`PJE&-ngyzqoh!xySsx=kIs-o)YK#*59)s
z>qlO)y7cN<hyK);PyHzm`$50Mmf}5c&I`J4$j%l2iOzNZynJnZ^NPdTD_ie=&*R!S
zb8$Y=zLVvnHy_W9{daDVzdT4^Hm>$5UYtI}Rqs6V!|FXprMxJ<ln>=;KjNB?j8mL=
z`@r9G>|E(%dT!<MoKN?ZJpB0~y*`{{{Mvta|2u!jwSTg6*mOQoef^Tzo%<!(dGp-z
zr|-Vdqj@xr{Y2AwI4YXcQRP2qUi*d4h3CgUq5W~+Imh<3{qPs>oT-ESuz4z*tGwdm
zr$^^rpQt`sm;CZg{^Ha}dVR~IAM@o`>chRo-u(3T&pLg-@Na#6AV2F?pUjTxlI77a
zeJLMKbCrkRMEUj8e3kWy;>hSbNq*Fq?>_Ucz4yO9>^pz;E9--PT7Ui2kr(+n7yWx`
z-h<XhMtbY@{wCw3*N6VPF8x_2JBlYe&t&AMj=tIH^^5fUP(Cz|(p>qK^6AH%$l^=)
z2l{u=&~q=py4H16G_QklsXyyN&kxe0^W!}AoY3>vzqytC+=r9Bdj8^<Xg%We-St$K
zSABW-A-}G_vOMM@PG9WyjbF)+ULNuk&re_ctpn?4-Qw7Nzx%oWZF>*D<noJdyUV^r
z`OvyNr&GN3c|JYw-Yd8iKiR9--+uYv2Y$%|zVRcs-7VFz?uqsZ`MKYgXg>VZ?YT2g
z`CE@a){npZ&JEf5Li1#omn{E8=fL}xpFSp8U*gr%k3P%`<>lv`k*y2q)g!kr_1Xu2
zdi!HP^<V1O?~^8b^H^W$eJqdlp!N2<jK98XHwSt6(TlepY##rASiT$mI{@w_aq^od
znzMbde%}T9px0mH_;s%I`ap9fFR^*Z_E}%%*?j7uKE>@F|97^%AI-<Q-OJ6(Uh<P)
zU(Sp9l99jq^2o1$?;q>uXD-f3{l&@GdOZjH&^Z!^>{z`xWY@1a>n`<0&tE<=T0a@>
zqvvHW|NnXP=EiQm?D|0aiT1nuLzZ7&lwVzBH?K`8zH}bwJ=fN8s#Cw7=cRr<XZ)O>
zm!JKCtMB#pw;Z#d$iK39?0)DAonQA0dXLI;qG&zNDO&G-={d7rviX{a{X+Bee2PQy
z{E(fDp0l3sK6mQ#a~}Ai-$za~Pkol`rMa)Axh_#1&$au|dd&&VQQzjEj(p_$O?kx0
zht3x}x%8YkHzm6~`lFwoCv`l3Q(k^kp2q2;da`{${%EdbacF+*hfU|edaZvVKdj%R
zuU}>Bt-iAShfQ@3E3bUzQhsxAKe>;@IluDK9~GUy{pP>fn}6sPKmToKANs$e<Ntr7
z+;5+2<IH`c{bMJ)pWIjUI6c?$=#M=4tHZw(-@0%7@<+bzNoT+RZ~VPj{!^aXE9+Ce
z<{^6@l22D!?-I>}pL5ao0Xu!!ICgRB*cW>DzV|r2d}M51?=$i8O%y+|as1RTotG(Y
ziuZdG>olL{7e7&5_R7^ybLt#@zm@W#dMLkj@T2Ef@|!3gCqH%3e972&`nCS{i@$vG
zVApA${H>>cWPX+H7d<wgc=}Qt^4BMQ??LOs{(M4jKA!uD{5?-({h&VBy=UYl^XF&3
z$o!GL-#O~Zht@?lC$c(a=d@p0UniU9cB1=G7wu#FsJ`E=?4^Bcetz^ze*96Kb2{nQ
zo~N#t>>SFk9{ZLbdh`YF`QXkCbq<~b=h6EFoon=Spx@&-pXj|Z&C~gl&$*oZPZ#}B
zt;74}sOE53ecMlTp4>krzqNF3<RzQS(mK>}PSH8{zIT4vQM`5dJp@@EdU5pRiA(Q)
z>(hsG;X6qms1Gt$uOE5QTuXhF?h*d%<Y|5NBQH^(*!5M<Zf@31=8w+VM020~olo)n
zCi26{kKTGFncjI`qI~9p)ssv9(|NC6eCJ=;I{0J#)aT#*;g9m5xIQ23)$>Pw`ap9h
zn}eULyhpuP-9xpD!znMly!x8f%`Oi0i}jPAfBPk)buIa`qxG7zb?B?IJ|~J}CtHtx
z)t@Mios7+I9qi>&-TKRq>||7z%&&cu%kCSyJZOK=J%;j_v$(SRX)fYh-=Ej=6K}oz
z$?AEYksjr-9`RT`8TGTYUiC`rRj>A`Z|g;UpnQi-{gmpofAW&uI_fWO`mQ*rUVe4E
zF6Y`g;g8Ni$=-bQ>S5PUU#g=Y&%Hj8pYJQ*XIMRXswc1WrLH>YzlUJAUhiqo3G$Oy
zef@NP>Y%<Rxq9Dg=B+<@)Fq2IH{ZYP;>D5Cy7^C3hn<Yo>j#~O`in#U{E*%{P#iz3
zojm2CUrTf5FFzjixo^BWheh}LVddLz9`iaZdVZXX()qxp^UqHneTrj8_k-t)9_g33
z<j-EpTRqu*rcdm7n`Cn+m+JC2Kj(!!)oWfd%4>ZkKl;k_@~owP-P7jky+Q9eBct~T
zfAZe({|?vRKO@V7Q=Z<l)`!#j*q7M(sYj2_71FCu_VXq`?^AaA_GSJ4-iJKsTw%`}
zx#W-foyc#A;+Oo``Iqd-Ub%X9G+*=T941*jJ6XNz-N*KC@@pRQ#MUAAy-bg-Klv}!
zwVuh3UOlA8>h+22>S3P~`zdcJAJ)Dki$nF?N8;)EP2^`iUFQ^cs_V_cddQwL_nE)n
zz|Z$Y*C(DI8SB5PERVjrAKr7`Z{AZA`MIC@$<I#iIjG+IQ9kixc6I3GVV`96k)M3_
z4e7mi_^Ycwde5mjP85&wRJI<!Z}R-9$Dbe8&d)kpkA2hgw5h+jqw|N(ePw#*SG;xl
zE)lmW<uB#&ysA%+`ayp3nXf*}P3L%6{`xY<k{u_z=fwGR4*mXxUc7v09_Hmc0_o9y
zknKY$-n_j3#n~sMUrTZ7sz*<jud+CFeyp#0^Fs4PdU@!Pojm#34}LiLv-cd>clxsT
z8oT?~ePs^qi>xo??>?qScKhO-PIP`IPWL>$xubcZ`<SdxGV+`3{94C)+0p!vAF5Nb
zm-SPp{m{E7>^na)KkFu!`d7d9N#FCSAA08ytM}hgvm-y>ebuwy?fDOS#`}NZ*1P+q
z^NaSw9LvsyT|Z=g=5|oZYagfQsCw%`eYiKA6SR)K<G)u?w|P%D+5UR3tnFW%_lbC9
zpU4lrzsNYfSDV*9(u>FH)uYEruMhc9oVn3s_2g3j)mPRBee1|WKT+T8&Bu=F*1!7N
z<=rpkE1Qqq`t?oEzxtkE>v11=&-dPvM?LhuP?sOl^QWK4f1>j$U;U=M`eb*X`A+G2
z%|jhBKl+LKGGF^Kkw2<qKlF{QtH0^rkz$v3sSkBZb>%~P{$%8*ez|1lZ{Nr$9_h`G
zT-H9t(_{OPhaUNx15Wzd`C<E4XJY-QKE<o=o|w+be)Y}=I`1XBe#jH2c>eAUe)QzE
zv~JIJxi(Mz?639OH*)VU{iA&R>Fp!E?>papzSH@)j=p>kIwvR}KXpp^^;3Vc`qu9|
zS3dF9Tj~q#7xLGSyc7AM{q}PR`ggSOHCLT~{&_FG-Sz*TpY^DZ)=k!je$kxep`R#^
zzR`2Qu0GN)QNGr<4rC`!b*zuQ_Y}Rp<ejLVKIL<M==V$cv3<}xFYNSWociFuMDgZR
zS~vaVPmkhIe93=_;_PSXe3$&_$@Z1(xkvR%_QuPP{K))KU8LtPzI5+a@4lm#2kH5#
zi}d_aU;OA%-f90lpZw8W-P`q(w-k^1lCSdQ$6wz)r}Un)CGwYtycDMo6i;sa^gQzu
zr#_ksn!EX!GrK(G(zy`NPR9Dt^H-<r{bzlpc~n2i-ecw>{;=shpQ^OqhfQ;}zxLn$
zcK`XYqdfM7T(UbaWSq`H_2Q@e@;N_bdC2VaHx!zi^^?uT94gcAm+GvIKkWKbUFWHG
z^^tw@XYX_2xv_raU&`OQWbyWoULNu83-vwEe$FHNPT|-7`OzbP??HX3PeywEqC92$
zcR&07;qU#Z4|?(Km)#uX<&W0w=V&sT8~-U!?ffRTU$S+&r_p-!MXxS9@*_`F_o%79
zzFLpHJS;zX%#A$Jd3JvJ(UT`yhyAo}vOefb=bzp_lTjV{$P?A~98LZ6lc)JxpC5W|
z<Tp2Y51W2}W51^S`Y}(kzQtJ|KY1tK_oBO=d(Btfe%J4f<dcUyQ5|#ESJ`-TRR`@a
z`&xE=?sM-QRLA|QfBli21OCo6KYHslf7G|WD$}Dp{QCa19(Hlo#a^=WpD3>R`L`dk
zdFY>v^0Alf@~KB(nZA_Q`LPf7Q$NU09`WQ-oPJ9EBfE2BADlb+)pedq@$Lz-eymSk
zf1W76KGY{8zuM{b*XKc9&wcx5*B9#F{@90}KXuW1^?6t{hf-hWz)sc|eQ7TIkzeDT
zo6ga?Q5^Y1QQcC%^h^ElE5%KEcIQW3dh>HG%z5IHpZzNJ=e)|Zmim;pdE}R`{flcJ
ze&UdRQ<|%O$nv7R)}cRfWPYBj+UrN2C?C7LWcDSRBR}yI^^N@Gp+|P};^+QjM}4fN
zeCmmluVmLx$!<T%%~P4*MEOwN`qQIxH?j8s8K*e<u7hlzIQe%Ut%tw(CGwZg`o&E@
z=UeAQ-wo#Oy{w*o)j@OMM=u`P{l0)7J$nDkYkm8rIq5^*iTqKYNRRYo=j=UCZ@m-w
z>)ZRR``|v4r~c}d^(XTqPt+g(%Jll_eDrM}*yU+m^JbSH>ra-KADNvV+4p<iZs>9L
z(>-!SnTK_IK5^6adrx~l%KqG@u5&xlbBN+locno+?j!MJ&yjoFbB6rUoYcqq`?-?6
z@$%3&o?X0sp`T>yc>n7!d)oO|-)5J;I^xcBI+yO@wd)mUZsMl<;jriG96LYGCp~&D
z%Gy1T`sCm9-#YC4<#oQu>_<)YN_~5N)>6Oq_r9i=zZ6eje|F@D^w#4$nBM-g)0@A)
zcW9rQmyGh0ty5lPCztH{LVb}Nr#`#)f;jWjm*>{~hy1L&b)55wp6iKIT=nLl4{~We
z`bO(P>*X(x{MIQR`H{uTN57WxEm3~=i*<BP^w>U@>eOx?)gM$gU(_$QZ*wlaFXTCD
z>Tkd7y2##xohv)4%g=i6_v$B~_0zBW%w77oe*W_=d)u{lyk4dk?>r%W+4EL^vVGBK
zDR1>;_USzHW9Ltg?wRTP!FgGtI%}zZ{mqYF-)MehdB~-F)i=*nPn<mb&^a_mb0RNM
zynLm2>n!W9Uvcc@CF&oS{A=%=^rMf`dgQ5{%)UhXDsH0ZTYc0oif>)d3w>pJocdx%
z^-6Pff9c=%3cYopb)b94drG`~;@Ig+e)6{8NuJI-y>*K3+*(&({PZIaJH5Q}Ao~*e
z%TtQyhwX!(zWE`4w7*D?jpL^d@*|`3=ebAw!f(H{&Lwu9M<qKS@>>^qQ<`5X552jP
z(fy$>{n*DPdXD|P;#|n1kIM9=dh%lL1$y<w>4RMIm#=yJ{KYOG`tucW^nNatADw%$
zIOp5%qv+9oq4R6s(EQ97yPx#>L3;g>%}rf(rn$(6@*q9Zn*&)ra;XnAFa0+^**wIR
z=DC*kWv&0@Cl6Y$z7Lz`+q~-P(|o!=-sAk6hipCap}gYB*m{kd<kr)NI!KS|%8Twr
zd6B<-<j&Q3>%2Ta&MWE%8|QiCPxkK=siz<7Ad4gGU*DB`&*?}1WO?<C@-4CP`qTef
z+MhFpU3b^jIg!m<yyuA>*~{9UtJbe<AD8^qYdm?$&QBfdr9V;BXQ{rq*FMSO*t-v8
z&lg(1=XiQ9*&AoykbP6?&%I<F=D@y`SG<1ZnP_gCQa<yue%}-59@LkAH;x_o`5vS%
z`N^lQx$vjQ_DPS{gXTf@K0|ht-#&?BC)-alJJQQX7AHUQ>-s(4)`$E{@#fFp^I+cm
zoG&szvc5}wW9?-2Qs4CakskF!M)MR`@>7q!GCy{6NA;0EJBoJ>(9cu-PSJ;aNRRf>
zpX<@1`u3MU+GqPK4)wi6bvg&n>-1fuuhM$WX^H%Csb7BW+Z@n(%uAo*>CMZ}x$4uS
zIV@4#qo(txAM0^XSf~Eo&-_aBL3v95zaUFAPn_oA9M^6i^=TbTG#7vVEZ&?-b3*>@
zr*-+^lD~N8OTCHwv3}EYRlmL$)o(ogS~lLfX}q~qUs*ouMdzc>JALzzt=s#gb5YOv
zRL}Xaug(+MI_1TlSN`(pOWnhw`J#K<dezm3zN~|+AL}p|_Q_w|erccdsebRB_Q7s{
z$nH6ExoMy6PoJf8V9xeK-_93(=Sh#wJHMN~`G;Qd^WS#%^?#mjU0-wM<<B_h!aMG6
z`SEkube|9E?>xvu#_1f`Z~KhSvFEw8|MX<fgLwM-se|lfbZ*@zOSB$w?rH01m%r3U
z`?n6{hx~m{tB0*a7LS|i^@HNwV<`U;#kX&9?Cyhp=ZKrw`I}qm=bDN9^=17?zn0!N
z)`#M)hn|e|Nbh-&-?>EdaIWdaTR%T~`|5c|dHCJ`1&_Vr@#o!Y*E%MOo9G@<2kDXi
zMA16U-TL%ne$I>Y!C$=jqJ1zA>+%0zvE}JG0sHQp^gOWBPxO7{eE9C-?_8D6*F^6t
zcIU8U-<0|hM_!`!p*c0Kc5~*Bo=bAsyzJ&nu3u&Mu0HmT|Nmeh|MuND@zv|6?0q91
zy*I5xJh^@8kDqvcrSn3sF4;NbhklNz?B@`6=UyJ?iQe-;FHU{y<d5oBRv+E({K)L;
z)lOztPrQ4cAN^WcK5RaEdC^?-MQ?uQCBD?(ME>md2iZ|w`|JGalfC|Ab@)w`x6hS2
z*6sZAubrRggCD)`C2{SCo`2)%^+CqQ)0ft%Px;9F<y%X6#Wydx<d5vl*FMazdNRAO
z1Tykdm#i)s#dROa_6zw>bWSIG>+)Y><H+qxo~b^)`(wXU*K;hdpG)+C?jO&u`@wyL
z`|Zd1Qb*mTeEjP-^)vZ5zOsDCE}!>H?fj749Lek_o1Q!K$Ei<xJlXy4xBltYzxRTB
z(er5iU5|5$_Vu9W(7$W=vCGf9`sQ!H%PxNCx4!3+U;Bx>?yjZ2OY6|5zTH>k(mbjs
zi}#)&v*T2kzkA$s;~pc6w~nr_GXK`6SJ%%;exC6AMKU|R^Fr32e))@^$luQy@*;a>
zb3*Hx=0d-g@|l}Csb{_Dx#8Dy;JqjxfAQk<$4?)nxb{Ov=V8jLj(qyU&Y%B8>yWSY
z+Xp}O#4qKq-FoFCH(%xEF=u^x?&(pSe8|2j<wNJC>!_apVNu`KJ>_Me{Mq%(uld-~
zdeK}<c5(9ZQ@?ut^{c-)di{%McfKd`6YqVge|@0m&HbUiJkBM`%T7l9`aV~Oy>VpZ
zCl9?mWaP(g9o4fVKQt%nNBz`ped4`O$o8jkoumAyKk@cWT))%l(Of3V@80644tvR7
zngf6N$o$Fd)=Rct>tLt1ZZbcty?*Tc)wR!4fBHAC%JiG^u<~qL*SfKN>>d9-vhN??
zRpwLj^Ihh<i(dZeoY-gmSdTblM}4)P|K8g3=zeQI)*(N6iTa(WzfI|$nBvr{UB64~
za39K3dJgE-kzYKyG-rKa{U(|J-tqINzPjvF9d`MhBV=D2&kvnndC<Hj>ZktnrS)3B
zy4Il|`LKD&@}T_erF`{QpB>xJl#gEC%Jj&t4w?rU+XuOM=yAFi)M>o_kiGuw`s0Uv
z57GO6aIcxW^Tlt<ORv5-&$s=j=cg{JgZl7$Z}W4{s>^R8KXv6V*-;!nGSV-R{}Qdw
zzFLpCwRP%ezVcxGoDZD7XXuf?{U)o6=8DdRIC^Y<lb#>DbEXfZM|z}3dYts^-lzQZ
z!%k+WFWF0elU{!0w-m?EK9I|<Pu`vb^|5*CFHfnSJnZs0Kh855#q&q^$)*&K&Ry&4
zQy-pV>!3G>Qaq|h#_kurzSz}cFU_fXeXZ5|d8cvBZw}2%uW#$d=50K`QeKo_zs*B#
z9<n}~kIXKQdgM~x>L<B%)RRX&_io=S&V%=ieB!%b@|WgUJz4+FM~~_|w`6@<54}E@
zDBt8KAL<+RN5&~%_2%yRuUx(FLiFd#WbAjEyzU`+`IFiCm;BIr%*Q-Sc}w4E`c;pg
zeN_k5BiFBXekH$&*3olDZw_SiT(w^1_U$>KcfL@(e#q*f^{|`kMC(BFM|z}ROZ~V{
z%&BZXamb%+eWm&E!}`f*F4o6SoW987$o%b_JnY!M>^Hl-^3jv+FTdLPv6uYq1DPLL
z|6RBI$RF!34?DSj?B>J1L~;6GqI~?w^5C>S{fnRIUaP-zR(iheul;F%&X+mqS08@g
zNv_>}@4l5E{rRIl_{*bi`xDQv>n2+_e_ZNUJ$~X!{pw#oWd8D?{Pnkw;?Vi9AM#8;
zFYwm~|4nKB`ji)~pWOWVNB(|q?f2C7$$5~cb?qyC^U5zD@|(yHomcr#U;03C^4mvx
ze%8k>FFo0JmN;~7)Zs_A-{hr!<<qA)^(ymYNBYvb==qy3y*`khKiPWpRXP_~yL{?2
zA3HX_^R1mMuRLURtFO!-r@G?QE!Cs1f9?F#MehqgZ;+85>6a+K{c?{Wed#<{k38*v
zt?YSJ$Gzuyln2%G^QLv`d*b9rkMiqBKK0DGtX;lRK2!(kvGHVm@*{gs(fjTsn`3ET
zQ9k=&zp?SXZ|tA_?Y$$vJo2J^WPN#V=;dc`U4E!u^RSotq!*9+B+Dy~9_j6e`MPJE
z1N+R6Y(M4GSILh0G$%4Uee;+nJBrgc+4_*)`;y)|Jx9*1{QB5C{^zmgfaXz}4?pu&
z5A`9QpE&iAzj*r2k-jvCiRP$|b2QcC$1bma)HiqEeS62}`CTu6!+k#bk+XLFee=$p
z+~<p49_K@TcKwuZ+k5yWmtS<--5Y=LyZ_7Y-Rln5|NZIK=|0Gd)^Bd+;2gPc__34q
ztv~0${OHY*UZ3iS)35VK?|fIL*H`Ui?-jB-`m-K#`|W({N0vuF^v<EV%g>)5ySb9h
zt^UnZyY;HyJp9$+r#?Cd;`KXGU!EH>s)O>#uP=7<bie3VK6ZKd$y2-i<d4pYb7g+=
zsH-o}tNHkMzxs2nxFuT8<fpG2I_;->;&hvb`^PyXJ8$Jd*XLZzyOfWgz9!1U-*Y$l
z^Yc8}M|sHX6Xh3ovT0q`PsUCAJJrwie%ECVjqiSv>*wBOcRrjOv=3x-zmnxqr{5#=
z{Ok)^9^`MG?L)s<yFBvCM^Ba?Cq4UGs*Cbgre9*~@~2-~7r#;-oW4`c+51^N-x=ci
z9w+Y||IR1N<IfGMpYn<0kNoxHd^+c3l&AAJD%rfu`AnnlKIhwd-2-IjXrle`-DY3y
zqxZae^3bb8mQP>up?LZDlb7~MUU|sk$R+zh&zp6sZ{B2VeRIP0p+EV|!#<dk`t*Jt
zWtZQ6%HzInU+RftC+n;4kEwt0;>hi@vVQriCy#o*2g&XgbFp6R`sl5%RG*%V^4Vu}
z4~i@K(aUF@_D_H2VSS}MOYF~cTX}yIupjO*v@h&r`-z^{%Hq6N#nU?vrG0mfDtm6&
zk-qdiRbQE3>AB=jt{*$rPG(2?(z@&i|B@eeUe&Xg{K}<z^{;Gy-4pH^_37Pn?sIne
z$o!EW`A>8&v3Gry`QehC-&!`G`QXyJ#PdV-*`2S>jg0n79+ZzCx%oXGlb!!$*C)Ty
zTpDj*ksX_d9&4Xu^O@q@gY3?=?*eoWF44TzCCiWQVef1H_OF!3^M>^D%EL~N=G1*7
zPi(w>lg}K@hpa!;AJVU-xKh39$=0P`@?p_D-2eJ958U5d|ArG~znllpfqAl<r+L@k
zb1dHZbsyE=x!^}O`PS7>sV`&~Us)fhZ+<AB^&&gULoVf^CpVvdvHpz{kL(l8buIO=
zmiBRpo`3sYF2!%^-}5Mbsx$eE)BmYX{aR<AL+>l=b&km;|K_C^kLs4};yi!Ot#j|(
ziFe<(zIEwWy!X94>dGe%@)J)k*?mX!=QlscA-mtdS|{oot*3pGk-v3Kl+XOc^KTw@
z`jTB<vj4urziYQCJx}uKSABXiyFR)eekk64tLJ;1%n#Kei%0p9AGy@0c>Rk*ab*5T
z&#!v+wUmcH8JG0xo8uDARetg6knInDb`(z*SAFNqzSQShyZV#AzWAw+&IK;%-AnSS
zYfk1ykNnlq57MLi4&7Volijyu@n}8jlf}tLubz7r{d_1+es#T1=&ieX>Gfm%=I6Y4
zPt{L;eIFIoX@B}Dt#hJ0=smTT=CLW&(=XZFtW)3i**QS_BM#*iKT+Pbbe<+ob(Zwj
z<J>OMc|BF>JXn``*;lk*rRUomrk`u3^-uR5e|G1@`D0(p##xvBu1t^OC$_)(v9G0g
zeVMy>c5=yI{90OX{n!5dIr-_Y`$>=XbCMfhSsb#<i{|INdk)00qj{K*b1sg3qWuu>
zx$)e|hwgQ9{XK8uvGvuf-n!}K^M3a8kN>WgexiK(c2DceI@F=($8JvglCN`Ck3T>8
z<<qbCkNtEG_DkmkH_ao?{vf?|p!HZUe`M!xzLl#Nhvs8n#82~OM|sH34|$2sIY0T3
z-oBa#z5eOtBlFW2I&YniIw%i{V@Lbae&qL^I?;OdZ~l|sx{x2)e(Kxv<J@{a^sP_p
zBcpivtdm|IWJh}I7bkz&KGem({IN$|{D;qe-!Xal%TJH(gI=7xWO+(?+)v{4Pe%F3
z>Y{u|kM!zQPcHSXzC86~m!BWHpO8N`A6Y*8OpoT}yqXX4L-nj{(l=h5dhGpf@w}Dp
zi#}iS$cyd+a-V1K+5Vo9zr08<uX$RR`5`}jt1lnQ%kJMBV7FefynDz0yH)p#^{6LK
zKjQUUvX}L<-{|kf`TzO!_vNe${eF|IFXvM}_2|vnx_wvj*RSWo^TLkqTa;fP^3bb`
z^6HN)KN<Cd^w@i!o<HihcKULyU*neK_Os-Fvh4N;yPx!E|C}%9!Sl$D;>;Q8`O#PI
zzO$paa>=f~{?YU1e9J3e$*xa%k-vOote<@JC|-Z=uf5~<8avt_{$%T{A3cAhcR#92
z_TKScAoCNC^!m0w^q%zmn1_D(xo7D0TdIrBX(`T}<nKDwE5&=icrTd;>Z5cX)K|BZ
zhhANCm6x83`mSAl^JSm><*T1~acJJ&xB5l*j=bg|US8C1>AX~5nIE!Orbm5}`J?lT
z^r(N7hip##`y9}t_>x^-_R9S9i<2LJ`J6}l#E$&T&3U139eRH9lF|A1oYI#)PnFdd
zCoj_LLp+L~$gkh+TlwQB(C6LU>B;uRb7H<^=Rq93_)<K7vi-*D$rC*n-Us?ded<S@
zlE3p+JH2}7UM0JC)zdc_J1>67Z|cLkJjd=Yq(}F}^!;RgW#jvEj=ZRTW#8TOOXR<n
z`daIM*nS7CZyx67e3(Cd{hSZ`=KSzG=sMMxzxnjVujf)f-gmy&e8*ZRd*eL^d&j@O
zQC+m&iTn<mQ@x|oukGJ`X1?wx^K@U@AO7~a=Zan)dF5*!@u;5okAExO=tlDJM|$}u
z_FPV~zMWTeFQWSF`k+VqOh)Iq@lzgtQ{UCQZ+af|gT4plNBdc}Kk@45gN*8t>$g^}
zpM3U3KkAj%SG{=oYHwXKnm4(5$mLYGe)^$ThaRmD+n0Lm&Y^s$j=Uv5dimt1mxn*<
ziyxUk_Pp`O#?gDu<Y`^~I^X2dd7`&I`P8c)|B1eH#p{EfjMhVLUhl=ut8wa;?)T>9
zSG{vnyZ+Rj`eEmf)<Hff?bDVYdh`YF`QXk`&!OkR^Wr&VFWGVGv**eAaGpy3$WET>
ztktXUyTSK|I@s@-%4_TKlb5`foxA?&-DCCv*XCdAr`}qBeUy!t*SgUD_|EKo<~*S9
zQ0)Dvj(u=%H7{A8?O%WLpg8;CTzNlO55M+9@A*^5`sAZW`T3!_iK9pTu%o$|WBu68
zPaozWk2)oL<HVyl{?^GZ9_h`odUpEyRp!@ylBfRdmn;w2dXOERA9av__m$qhpnRn{
z(pw)HoqOvn`7N>i>u*|z^-PpcU4439>Gew%=iIWR?<4spnwK~-fBVO-J{kGdE)PF_
z^T*cZzbWN+J|}iRoFnl_FAu$Sl968D_JNGnO}5`S&7*qrE#15P)j{>y8*e@COS1KK
ze(dtAkMgTmvZFlaCyzLN^!#^S{Ouo_GpdX9rM{{^QL^>u1E+QKw=Qz$NsmXh9{Kfc
zAKMQ*|FZqDTc@APr#$rPm;RijdUN(YQM>c1A9YH3N`B&zzGTPtLl#Hw^D7>$1L@^i
zOY!w@K7H}4oxbckDmxE7U-C~BcTnngsza|oRF~Zx$xGy~{@(FFpXj+}pYqukbY8`g
z+1;<+2lVfM{bf%(|LWW9qIh(V^VbiVAF})Y@;tIzf9EKUzGO%F$ffw|$?Ow*kJZmT
z#;<ng#hf~K`(qvE%+L4DV_*8Rn?32PZ?S9tJV$;1ci#Sd*?-4Mmbc%<>RL}}JzXb1
zcI&Zj`!CMA&5z#_<z4DiKd3IbWS_{Ny)yko{y6z9>CLNt=3f0IJO9XUNuK=lh5RO}
zqi=Jm-hR`QogcDu<6L`==$#ib@{`B&#h;!(yXQ|mvN#mSE*|;GL$5F7H&GqrXKv<;
z)`j+g|3v=sl<eL2`c>vX)n!NV<Z{ZRKlVwc_nxXh`&!yZdF+pMpmj{Le$;JV`Pp&O
zv#+J|#lJHBT5cL|9p)t79L&{ov}EUBwtsWql-|3gzSL=b{d+&NyZ6ZEAs^X&Lsn;^
zdg9ziWOmdSyYCM4ol$>%>z9A?O?}eKD<Az5%>(rz4_X(}n+v)1`1_7;pW^r-Kl4EL
z%JkwUwqJ4j@Z9*0^xswaUfVnVbBX<ACtD|f_A6d>tEb)kvp;dS@&12?Cd!NA^)3Gr
z8~5_FKXCQE-u{+j<|H3lKiPcvvs>@OU-bR&{P1gUx$C=JJ*0QeoD+GipTECfMz3G~
zIO*B-kNi;|&IQ?b8`=AZ?7p=gbRV&o*5lqZPxaLiN1j-}%8j$`vVLUiLizPiFW=I4
zlQ?rkd7LNjhp7+#<}FXDFM4t*554n4ULt?{sK544Hb?TQ%C4h*nwxvm{#&29$-}>S
z*v(PAdtM&%^?pQg{5<c-?jBW7KV;{VtdGj=Utj##D?5i+KXar%(X?JPH*;y<{7UN<
zZ_eV;e(@{i;m@v~`%@nN&LMwxar|m`o}4@Hapy=Lb>*>7?D{2_`tY97uRJCHiPoVn
z_4I-C6XiK7PW{uHC+fRoN9W7)!{1yzpX!j!3&oiaz5OR+{mhNObL3p{*DqO|yyC@?
z`FS4sJE!W=%dejG*G~3bhw6GSt7GowtuOV^JWKV;_Fubyhrsu){>(`q-QV_Oee$!L
zlRU_;?7HMB<)Lp}W&Y+n#p$>4WbtJCHj%&j>Xzn?_RU<Kmy#c<BM;g?{g^xdQeA$?
zPhNG&6UDQW#aowi$4|fJHPvSq=ee>^;_NSf=bpd6|D&(H<3FFuTYqsVt`uh<YBvvk
z>W?gr%-_#HWZz@@(SO&~`b%=}Idio?Wc}&CeY5lDM^Em3ERS`H7uP=6QC@Osf011r
znuEF1>kFM@c74igebc?cujEfJzxj~Wt1O;g9`AMENt0jw`H3?h=iNH##p_pm_m94P
zibv<1zdFdzJnWCW^!(_Nzx&X7P+e5lJp6e#JF183Sq~ZciLYKi>LY)0DKDA_KYIDC
zi(P)CUrYHWPI;VHb*K1)>is-%&^*&R4x0CLkGFr$8Pa3VJ3o4KewXN;pXhxkujf+U
z)^~rQy!_oG@=cT%t>1a~9Mbc1zNWnN{K@WN>(#&fD2_axJARXWDZX)@Kdjwc-OuLk
z-etE=c6zi=CHq>6SI3;ivt#wv;roHTv|s92pYIa!?BeAo+i!i6?IXQ@QGVZ>>ga=9
z$~RHnCF)Zi&xiAco)^zYDNa2<A6TdLn3sC$)BD~byD!X#f63oG)l~=S?VrB<+$0Y_
zb;MgA|FX~3rn3CFY5e5hzo$l?{7*H#bLgBof6f^h`P(=9?R=Q8{YLTbNqL+Ha;YzR
zb=7gsI#>SgA$i<Co`=RARJKk$Xuc`_sPn7aI-APwH~S{<QBfUzI&bvO3mM%<`gOjX
zBj=cZ`&X~}O=Z6a@jN-d?!&#~zwdE>`*U;s=@<1ymdA5v-JAAVe{-JZr@lD;WcTI`
zmF`>T4(*?FM>c1&d6MmK^VH5yKKo(6#PcKTrxaIz`Po~CUVhJM_3YKl&u*SmfAsn!
zi<jSf!`vpC{}eADdhXPvZ#?~}O8dCw-~R7MOy_0Oc{yJ=#jD3&|JwPjrFdK$H~Goq
zd{2IdrB|={*4B3)G@i`vxu!3tKKRw%eXL*o*>Q?ryKedUljXN=cBJQr?T3ubFAi(3
z%&(Ncdhy7wY}`R*^-lD6&)UAdANku?viW##^}XgfU`KlNT`KQH^AcD34)Ffx-@0UM
z9`W?%fa2swaegi~4{`h^PU}9=dUHIiy4E}8V=wtvPi8+=s4w@P=R<$&o+IQ>cJ7?B
z()sdppPv`yFXi?7PI0KeQ<a?$d8yydfxHyAWM8VU4ld<8)$P`S?RV1GZ@==CpT4v|
z&OLd*^|x_<`_DJ#V*cdQm1oL2%-?#v&rg^8UjJME|DU_VnvZib<zq+ps^8P~`-JTK
z_dkBhpWOf5FZuLc^nHW;$>_V;x%J)SJEr6(4&~F&TKcXkTX(;*KYyL5Z!~xH{P_TT
z*IU`oC+39aN|v9@j`Z@AQ9o#ZOLq5D?PPvu8c+1k{o8scX9CV~>G^T4-Am+Byg1}1
zo*pMXJGxKY>+&K$q{pTA$7FY})$V;W>8H5*ua(_<?zzLxzqXGhKliw~+Yfqu`?-g`
z_pJX8g8iUW_cd3YfBtzdz1{8-?JsuS^yv2~WVFt%Pe1&cf33U}zhqykBfgyCx-M~(
z|B2S?$N5C}5ZOKBUbBz-<>y>>-OhtJb>-t<){p;0&y{|hGxXil_0jA9TVDOfD}MZU
z@34#3$?m`FP)|Ox`ef_cbUj_C_`{;PSl6bsF8gavo6>qVjoZ{;zve1_lKFQ|egCpM
z-{K}ZAMDPb_nLnAk(W5_pZS*4`sq(qT8H^{?zQt<qUY7~Jk4{`tE(UThwAeym-csS
zzwuvw+0B1$XPU#>pBn7LVe@T!55MH{i*CCsujktD4ZKI~yL;o=$Nub{cX{GDyC<6F
z=zLD}T=1W0ee#GSdu}{m-W%+tIhiMlH(z;C9{$+A>u*2Q^&E?99{%j&=#gDKKeB%L
zq5Axn?qlx{`OrGmLv~cxzLC2Qbvpmb>iWGL8Pz4@+Vw8Y<EZTB()D&d)`8}3U$K4G
zUmU8_ebN`Zd9J1BiofTXJkj@%bHUGi^n>gZtpmj&Juc0GU)g--i?w?W=*fN$VSf6t
zK6B`tIxq8Xetowu`LX%r5wBk|dubi~uzqBA>n2ZZpXP&&Yn^Ew=2K437k!_%>eXvr
zvi`b%)?t5;pY_r&(R}%-+y2y@Xn*A?YgcDt<0n6Q>o7lL$Ekn*W&PLoB~CsvR!=@`
z+DB}D{gg}d`r!+Xe;GgL3D0@drT4t}9v9#9;!k+Sa~|=Wi*Nkb|9k1JFa4BDZ-4QP
zzV|top7Yo8c^7}e#h>ud>%QUX&$;xczV9FY&dyENephe(7AA1kr7yqZPrc$M7w(sj
z_>o_F!S8(OU+$c#To~Whyq&;y0^12}C$OEsb^_Z8Y$vdtz;*)L32Y~@oxpYi+X-wZ
zu${nm0^12}C$OEsb^_Z8Y$vdtz;*)L32Y}YOyHg8f8CY0_*egHXUo5R4jtPiUvak=
z+~FU6U}uSY*M0K8f8n`*u~WYP(Lev5$G`fcJLO;f#7$pz!KJ^sbJ)Dav2Xep@BP_p
zj-4rd^tV6rFYo^Q*B&eX^#{KBZWn*f?;o3Zrw{FZ-$Op~$H(wYna9#PdVaU^b^<4J
z0-nG9@?`eDt-YPVb^;$ifoDDL-Y@=|t8Q`OC%^7dcRY5VkL=)O|MSuBy2mg5`OaE?
zeAZK6TKZmo&}Y2)Nk8}DKi)arc=;pW`Vao@Q{KCCrt#E2QQHo;6WC5*JApH60?+x%
zbDnt4o8NbA%k3Oa<T*IAo}%q)wiDP+U^{{B1deI~+wYT8^ZR7l?<>CH)vtWhv#;7Y
z)A`{`-uY=?`A?pIto+G`zUH%k=kuPwbJ)DYb)Wg_o4)JHolQCI(^kKoz;*)L32Y~@
zoxpYi+X-wZu${nm0{fr9eeQj`SH1FqA2_xtr+;tg3%~VS@BWHk{oS1<{^X0Fe9i41
z@LM~lD}Vi&pZa-^f5lsNZm7KZ%f9DNKj(+9I<}U-{Di-M?dyK!w~w7Fe8WqgeeU_U
zeAltlg}?q?zkI<}@4NEYmZ$d|_J31vBhRD>?0(0mKj}yA@xdMZ$@@O`8JB+XukDn7
z`XgU>&#Qm*eaF`FPWO4pl|TKYcOBc5&wTzXzi#&le{}3r<ySrRXRdqRxBbDflg%^f
z3~tx5oxpYi>n8B^FM9gzKlRoBb?0>By49U}UJv}q-@4NSuJ}6_o-E$@+#CImhdt#3
zJDc)EPx87Sy7dRY;_7R6CicG<zx3%(`MHl>`rmiTPrKW7KYiJkzHet!UUK#wFZhig
z{*|3APpvte=yQ0o>V5a!fBoVszT>^ePFME#57XZ_t-kaA?6*Ghn;!R1Z+YSGJ?0%3
ze(nFh(SdotKYQ}S|M3g2+1Zq9|9(#W`g@K3edeQ<*RJzzZ+X;Be*8Xf+L?IlnfG|i
zQ{VW$os-4G`upBJUng3A$&Y>QSA4~b|7z!``GYt9=l6N*+i!N^-9GPgUjMCU-}rid
z!zX;tm;K<~Z*pPT=j);OyYUPE^J{N<;dDM<_NL#y`ek2m(+k(~M9+C!Z##kQ1hx~{
zPT<5&;8$<*nP+{<_kGfZ<$W(W>zkhS-#)fe-sPje@TK?o&g*u{{@nK;?)=o-zUs4n
ze`kr)pToU)=Y4zs@^Qa;>~!O2TzTGk*Zi~h?3}Lbe~){*$^G|pP8WIUIXPYIr|Y@b
zcki*v(>|QoTYFn=JAv&4wiDP+U^{{B1hx~HCouhf_=c{ZC$_bmj0qgppBHTN9L>MC
zdsK5h8E0)<b31|U1a4Rpc=BI7<Q*@+(|dNd{J(Q}@lC(>xu5*LKRCAKb`FPe4sKYt
z$#%6jlnH$QHJ@|w2fy>WW1I4@{vG(~&qufV&Hf(e4duL^DeJu8y-)wq|HIyS$JucG
z|6eaFYD5>&qpu){aIQr{c2|oYLiDI1h!!=1L^j0gBI*(?L|@{pzUpFGZOK+!SleyG
z{=T0_KJ)p-B-iwrbMAfp<7w`k@;=WwGxzSynRCWfcuK9Jqj<Dx9x<ct7YOA#%JD%v
zV1P#h<g>xk#wRBevnibatWg+$EvDIic&|_Ib0K44xu!FEgi$fY@4Fa1&Q0%jp*$FX
z0pkn^*9-CcZo=nkDK1agmRj#n>wezGQOk|jxE_F#VIW5iAFmPXzOEEUL3nVce&8tQ
zeol{h?;h`!LG_}ectJa000v+{F9VG~pPlkxKt3Dq_n)-Jua{48Il}js+UH^7^6YJk
z=V>x}+fmlzReL|$-f_U^FaQHE00S@p12Dj<fp%r@`rR6pNqOhi1DQ38jXz64zCEyd
zo**t5fB_hQ0d)rO{MFd!uj-Bp`?=(fn^D^m@i-uR>#fUOpUm0QsXY$J_i6*;@vdfX
z*LkBh$!c<ig64R5b#K647=Qs7fC04z0&7jay1HK?1z+}%oiD%V4<1nzG}gndd3E}n
zbxEWG)%Ihx`8K`3^Ge;U=Y(}C{hiyRZ8O(FjUP~ZRO`!8^NpM`z2?y&k0>j2RL@qm
z%c${Rysncbj@wb^VVQL~cfL{I=Cldy)2qtnG9SbfrBpjEPAZ?)`D=xBO{^M!D0iJV
z*3<cC6<)lh{8Pdk51#cMN-ic6vuM@z8}-6~90S3j55f}<B~UT9cep>NRZ24Dm6d;M
zG^Ws{M5>gI`n@jIE_;1Nye_KV+xF_k=P&>R|1=<ePHgo2HCoI6XAAHN48Q;kz`(y5
zkUMS(=|&HT@Oen#!5k0&)>rYF;}~dAtK_~tAI4J<9?%bP9thU~)~;|$pX(#NS9#ah
zf+;+}gO&Y&wd)Atd9?wn){j8F$M#(DIz#mG@am(Cy)Xa+FaQI*8z>ss^Tw*fw}`Wo
z_v488VE_hT00v+H1{}x0`X!UbMy*PrQVebR)98vr5-97ueEIYTWA`Ld@ZUW+PS0ht
zCF|b#eE#rOCx}vF9ui~cAz`z($d3xo$?%-a%I9Q8_doe(0#dSQU_$OsmglftBdn1{
zSBS}a19I2%6w+yZzuWBm)*D6yj$~l<n=@bES2u*PO4n1Kg}ee2sNS5}zTmGfS|?IR
zVd=C>efEz_q98o5&<`Bx8i}^>V_==<@p`u#Zy+E%I6^-V-iO!v-l2FtF>If9-#a6M
z!UH^@A2|MbAY4c7(dhf&&Qd-<zZ9*BRNN25>$;4M|DuX}{Sevn3<&oz!D#<MBLgWs
zz=M_jKs;X-w&y;szw8Xv+e4YGI>+(^A0Z$g9<2GcMXfNvrva&X$))!5*=j@$1I8E-
z)^!!%N0pkV82ei!oj-bus`X&Z(L|{*00S`K_y+X;yP5U!9RG--g)jgEY7Gd#<G~T@
zzOEEU9o<8Xal3x$9T!j4GVa>Gsx_a5QKQ+q=Yw{(i3ceN59kN{p9gBcpO8;q12vX~
zUs$;$m9Wkx=N_Fq()BW7mD8&QeocSAM|tP?j3L3*4<!&@_%!X$<b;eoHhXNcIV-Wo
znH)CWD4gZK&3`}nWwVD=V9G?-Q~qsFP!Jx_4{#pfI`Hc2fN&jkc1~QpaoN_WM5-6x
zY|vu<PoG3mR=KEA+XuZ4<*>=+t}-18pUSvSj3Vw^@NwUw!$<w|v}=EZO3~S`Culzm
zzyJ)ufUyQbH}!hMcIOF|%ekA6y;D0rff~&_pFMmw`NIeb{=0`YO-@A~eVjr;K0REj
zv|(y+i2%wQKUunV!+`BE1cV3l1Dpp|Uk8l+b;NJ!m&fFNd}l!{1>phxz<JLDey?DN
z90p(j2J|uTWvhrYJx+xYrPQZ(lmY`V00S@p126ysFaQI88JN8HUghsy(<ms<1NsdN
zC^4}0e8ieFbEBwU?6l})>~r5J${WY*e_Ac~g-FUP7v%WjSogIFlvmz<l$bMSMm*(}
zZEt*X%V$6|<(26j*EfGRDT;#1JdB8+*{!$dEegKcgOcwgXHHu?YQyT70}&L22POT$
zna>%tonHgOJin~Z+ve!;?HzV(&lRsT#A>#$8Zh<u)L0@#%!|**y!eiGocO)cB61jj
z0T_S*7=Qs7fB_hQ0T_S*7=Qs7c-cT+I{l}lGxw?9e7P2E*)zcEJiENaWzQ=<hXFGU
zXdMUFvgg<nD@KG<tEktwz<smbf2;7|AvK$Gs-1S<J0pTxO<^6A_)_l_?o}X_vce(L
z7o=AUNurKIv%D(3hK%-)zi>7)jwodnw}dNSAKg!RWsyAhOXv4aBw`+`GA3cf<V3;>
z114u~*<U=12#jt!t%ui_8ALA6wST93`Ej{yTB)>OAw3MpG0>n^$$fi1jHe(x{JkGc
zJ$m)tkRwq<V4XK=ldL9JsH0NtI?q!19BrHnukDd@E&NX!M<358Zl7E)TGNTwpEqk;
z9QQKsCFA=y3a4s0a?13YM~6J3QWV#VZ5V(77=Qs7fC2su+`16G?s^@63M%s;_4krd
z`>Pb#@jSagFkyv``gb(i_3Ha1{9pOd0vK={1E**EHaZi1gz81!%{;&T!`+3CA_!=W
zhl&k8I<kIhFy)o6Wk&U@TPT!*#&{5~do-Ze>*HN&-X=;pZSSS{>BSR>(Nw$6iPr0=
zFs=q1vF__ianw;gXzeFf=Wp*fZct*!bJQ&A{hZ%wf2bb@U;qYS00v+H24Fy$fuC~q
zY;Zj&uMLC;mivLSUTu78@i;oE?Z|pfznQo125}Z{-Z-gXSov*)H734Q^V~a2B8ky7
zz6YwSA!A{=rZakkQP!xgHOOU_fh(2fmh!uCm4fhqe!$;(P-#o6{jP1UQP5fsW?ipQ
z4<81Mnty<g)+0(7fB_hQ0T|%J0589v1)EF0?t6XkPpQ-%-#FH_->AarR4b=G?p`U-
zJCRDU;V%Vt+loa{y*N9jL&wj{#Zs*le1R|k12EvM27F&MJyYk#BMQFkf&O+NzW7_$
zDJyKVD)i2$&*G?6bk_4j+-|*o$Ji@JTpza8894jy3YxVpjq*lytwk;vfB_hQ0i_1K
zhL*|E_j(91i%R1|dKiEK7=Qs7fB_hQ0T_S*;|yFLllSqR1+mm>8rMQ=OO|?XM(_RB
zwidNJyMcEa6&(Hj)oha5=U#>09XC)ZihUi7%2mB#qZgDF-u!59-x)qnshF;(JPUaR
zCQzjmw-?(m00S@p126ysFaQHE00S@p126ysFaQHE00S@p126ys`WyH>EXV7gzIBF(
zc_h}e%cPSRs69rM?d$fy?=H1RrR#jztB3c`mCncBrf6io=izs#nCTtYH-9%Nii(+?
zFTdvx9#K?a{wCvonB#SwYNg)KWvwc&?1<9`qaOrOUg_Pt^u*Eu(bOJ0eAGWryY@FI
zZyZ&-%)vk611RfEURdk%MHBB(F?;&ft&{xhIu)q@l^G?%01U8dpw*2L1rN;plY;6z
zuzFk(moWyE{yl*7Et1Y3y+zITV9e1+sm@}cPwMGKJ%71Ktzzq*58Bx#9;Cdo^5oU!
z`)%7ytfK9Q1wU09=tCw2-)S}dwSfOo{P+8a_?g{$d)}gI={neJPx_7+D$wVA^5wvd
z59kJu^mzw=2&4Oto!L3qE%bl%>Rn@N^(TMYT{pGN?;Xd@9TGzAF|=C{sZsqV)tk<8
zUZb5b;0y-5k4;!~d*UUcm1VN(9Lp1Ygs7!6xJDbTH(=Jf!Pd9*uLv@GElVyoEuA<T
zqpP}I*j_t@fbd{tKVaoN=PeHLdS!cu9ouuo>kP4qMtv_-Oz~UzQCciDn$LgEv!l`Q
zKw^)D=;#3vJ`XAAm>#xwxId>=N-{N?@ul7=+^aw;1>xb}`T=j>Lx>j!U;qYSz*+<T
zS%nvGDgTrRJb2c3D7lzOKzMNEeqilchuUEP24DaNV89pyRgb*=_QP&r<k}yTe~hdc
z_dklq*NC@`i3X*@01UtY48VY67`S<DVD2>yt`I3YMr+l#A~^Ixc;cZ1sy9;>&-=Vt
zi@l_1V9y(?4&Sm%wd3Mg%_kg*m^|5GU0M{f@y4#p4{g5PCY`9I(0&jGU;qYSz&Q*I
znZ6*sT1XNF;Q{^N|DOlWarUE?Fkq~Kc4hDS-5QliL47=EJzC;<sI-=c{4f9mFaQHE
z00S@p126ysFaQHE00S@p126ysFaQHEz@mXVZ`3APO|DS0Y1I5IQswG3AMrL4?LTN_
zAmx?f_cNS*o5gbzF~NYd8pv2!uIY>(Vbmzj?)qZjoeQBv;QBQqYn`kTN%Z1ZEpDeg
zZl6WeGT^5crOVb!AZF27`x)BJwgK^Y4Q#)9JvCPRu-&T=9Sj&_fR*zPaTt4@G<yD5
zR?9c$tU;+}88B-;X5sy1@%ygg`qz0LuXnrg2I6RBeO_j>`Wxzj0T^K4fZp>cpgd#d
zr+`;;y|RDiB0?B|0T|%Zz}WaNs<_t=p?pS<s9^vGU;qYS00v+H24KKB4Vd-3x$<v~
z#uU1gNR`q#j|y51126ysFaQHE00S_f)_`~%6KmUQ+l72E00S`Ks0LQQIrH^>bwh}l
z!Z_3_1$I2oE)YyurP|{%r`l=vy)z;xYgF4F<bweifB_g#XCQfDt<M)tyhFtty+3cy
z;xWP09>><s?6P6+H7cgmd?Ax>_P@KYz<Fvmwa#O+IH26d0|#$Vwftn}@AY=9yiKj<
z^@svxCw~z{dE>1M(d(|)@u$49r*GXl$<MA+v#IW?fLul!_;}=z`BB6Cs8;Sjc4p^b
zw@_*o->;B*s#Crgs+LB3M)@!R126ysFaQHE00S@p126ysFaQHE;HU-;?wpc(_qS(+
zHIArxyyuA4&xn}6lr9;X{^4HAI!~=FdbjbP=Ls(y>mRc6@Vy|Sm3@kpYjkeeQ3CSm
z!BMYsXcr8?0DlJTjVmzwbN<XBVi<q{7<k#h?3fN6KQ9+c)v`z1X0C%8KcKv`<okUl
zC9i!%dF7oNqq1GYW2m#R^SSq<+f@pqtng{a-$u4A{wL*)+w$~ZRX947>cy9@F4%$r
z7=Qs7Fw4N;i!N7Mc|Rl0!X{6qb@ABpkT?s?>PM&t24DaNU;qZJG$5?|WEIu6y~C#+
zCTI4JB;a2jc7A;Q{jy(25LS8ctnW~AF_D1qU}Zm$y59IsXpwaO=q+kCrQ#Crm;2uI
zHWKYWXk;L@$J9ReD)jESfwD@YuNOwoBXjMK$v;L`j3Z{#S^i#^YNy%WQQyn`y?xby
zslTVj5-IBSyG4AR#5N4T01UtY48Q;kIJ$w1h2@&g=n+Qc(#LDWy00t6QQpbgyqT=V
z<LLb#?Slaru--tUh;LVntMHWj?-irJ7lrk7_5QvPzW*zx_$~Y>Etc}iLUi<i2%m=(
zw8n$c{ohf_7p}PRrQRvrt3WDM%gKB1RsP;JjjCmE=!5XYLkSdw2Q&JC(C<MQfB_iL
z%YeP{3TAyS{4STv9UHrqXw>ilvB!0u$Lrm0yn)E2-s3Tj&2ZhhxotWH;XzJ65YA&^
z-8|v%Nxdoe{8E}uYP}1k>ro@Ux^=v*_9paW5C&iX2Ati1+WSgY<<t9kz486!ve^xv
z5Uq3z?K|+%pb&C;L8->$c3+cCx}f^dnU|vp$hU{#QLQ8D6)a*CbLFQlN1o5kYqQ7Q
z)Aw#}T{MqPFa8)>anq9@a@fR_d*0ldSEtWemqaRXeeoF!#?*?YQrvnzV$GSkQIt34
zZ!+$OIbPSPR{Bk->vwc*2(_9`zJK$*?_ZpxAUvQS=zku_y^bQCHUrBV{Wrh&1n+^v
zYq-~Nuj!t1zIRt|_gw$^cX-j^C5FG@o@1W(@Y4U;e%n2#drr@+Vc&WW|7D@uZIWAC
z?M{@trAyB4ZI|Ts6L@^_@UgRMWmAD2&XiocefdeEmM#~6>tD3nR^q61qOO9b!T=1w
z01UtY48Q;kzyJ)u01UtY48Q;kzyJ)u01UtY48Q;kzyJ)ufUXAK&QpJ6mtmm<{C5vx
z+0AdCo?9eG3YDVpykAUVT!5HA-yhIsQmaQ)Z(eB9sKm3^f{58HIx~6C_rLF;db96F
zj~QO+u|&+n#Lcz&u25=^;yB!QezNxB${_0K6!%|j!+@0r?46%-KesY=)C~hL@Unr&
z@zbUk&B*DOyKq3`$^SR8jzF&xBUAny^MuOf%cZivCH4Ey>y0m7tjh?dyfJ^958q2~
zdX+jF+3!(^5C&iX1{}-4qJkG{AK8*lKzJ~-A2`-ogx0_S48Q;k7-ir^P>%-bZ$BX*
zTORBkZ%4a7VXr-ff7-bA@oHZxQ2ZQA?`^fu3GE#xd=3LJ00S@p126ysFaQHEpx(gI
z(QcQk`9~3@+&S#46AL@UQZdI=@(NxsHl9fF$va1H|28j<@Jf#tduo2rJ(f5t`%Lf%
z-FQEO$|Y;#B8{H!YT>>idphhsML~G@dp}TrrAAIiGO+mCsI56>Bv65!TJ}8Gzg7Sh
z^Z0<#y9zaqBXU`x?x1HIa>o*-bXmOL_e^{$VU-EC>J4kwOe9v*k@_Io0s}B0&%h7)
z1M<3FyhM1Tyw<Uo%0HL;h(U3Q)GC%4@Kecxr$dNRnr(9Cf^L&84T`7s*!RDGrwg<<
zetuE&Gmq=7iKR;UUH7?9FIGK6K(;)X-4Cs*&*;D7jV?c6MauzQdIb=Ha-WxreRr$b
zt?(}YblLY~r0&}&y&ou~SNq;~_f|LVRSC)>0?mHUMeaBqsd2}0%P}fli}<!(zbTwX
zv{Kx^%-$CIv(X$gDSWy6lIQ<>&rEpVY;)^7k)Z=JshGlkfvKN-dc0Id7X2=%L*F(l
zQ~yC>zfg}9&GwA?U;qYS00zu8Abc-~sdT;;@$)6&ccc`B-&tZx{k~y!zuNV0Sjl&x
zS+fd9S40qADV{%vm+iv8d^jOsbTHwaYX9zHC7;x|b9-@`d(}p5FaQHE00S@p12ACU
zfZ6x2d8yC7G3>FzfZXRzNN3Hww?=ulqOYM&7=QuiGT=FQFWuGsA_f2Qkn8x+A1kj*
zB_KTL(GQ&KYJ%1}i-Gc&JjeV0kl#-!v$pm4>9r!!R4K(?jT)cjY~_l*Vw-IP!u;!2
zk!?HjqqFKfH~P7|d;>@c12CYs0jtKTv|oR~KeBBSVU702%NhONhNG3QcR#f^UPtUr
zJbqE{ZN1xt@?Zc4j58oLZdvVqd)F25^+>#}_4`&_o}+9V*8@;84CKghb)IX&o}CE<
zga>Ep2aa;?FSz-1)7=3%{Pd!uctJa000v+{F9XLH4<9?LR<_-B>gBG3&3ih%e2U9?
z)G2D`)o+U`ehz*7jn?mHm(FVwmuGKVYCaad_wzDdTq;hr<74d|2Ye0#FaQHE00S@p
z1FRZQ`@4s=eCJ2h&wXUrNec4qfz|T_alrr#zyJ)WGhlBWVSN7Kd+ANDQmr1;9ToO-
z$$kGuZA--CQ1evll>14`Od_V%@jK3vU+r-~zE>Nty8oyhFNb_D&&w~@$E$k-_QC)R
zzyJ)WHE@A0Sedl@1_fXC@VG~@bFFILBcQP!zFZoz?b4n!B2aBVR-3O*%L`8*SHDA9
z=Y_Hto88GwA#yn-yWN1z1Cxn8s`cfl`KG+Nug$mFVT2Vrs%NX(rT2I*p}g1|bV{d*
zsf2e5>y0@Y#p{6x$4x2a{=I5;I$`}uv+EV=bCmb*?7jDn&j+-b)anrxJXm%8M!hf~
z$AG>2!@RBoB*d+D{Yh5VV=_Aqdws>|IPBGn&tU)t{%OE{=O=3~t_-3AS$}WppDn;A
zFaQHE00aMK!06vwD8J;|?aNORcu?ly-})*(a~uQ0?@93G-`DT~AAkRQ%h~z^Ygag}
z<K2;8pZ5mf5gwJif5O^z1o6Du0PFV^5uclD^If5m^KQNRC}S@SzyJ)u0PhCyefZb+
zA@9c#@xuTNzyJ)u01P;e0rx)&%<ecnzfFn-Lf`#<SyWCN>-6c>dw<U{&k6YN9vr9V
zve}Y?ea;ma(=MDUrPO@h*#E_gbs51_tp_%Hi~Oh@<^3$Q<G21ryKN=#@UjQ_X97~P
zXh3+ckTtUC3Ncx4KzJWdE_oZ5YrSDa;7A74{_Zv_pWO3UX*x%nN2BjXj~QO+u>^z%
z7W#oBT_e#Jehd_ynY`!w-*-?D9vq<`G{1d%Zjl@*)ZV-)azA&p-#ciJ+vRHhQ3M{~
z0sTPwJdhf%$lHA&R^#R8m!dV1#{5Bc+!>Zc1+LuWdwp|20u?CCdnvr{#_DlKT*~G%
zfG_9pwQ%2%JsozRqO8!_-XBMM*_=PYnr~aw3Ilu^kotQ?YCoT?M$|B1jDc0Y?`<e|
zKbgFpr~b$;!$SW@JSQ;rIe{@p6Q#lc48VZn8z{UhdQ_{mm#DMQ@sB842m>&n)&QP&
zJJ<8>$GzXU_Iz&)QA<D9=dI|PL~1k(7V+rN;{GWD!UOsN|L1|)?<eHb*MP9@H0$i%
z>c+h)L0N=Vx-8!BdnP`W@Xo?NZCv|!wJ+s`GZ%E5bZJmLwMVJvIIQj$&hljoUIawA
z=dlTFGWXBbJ3q-HAUvQS;5@)};MLaw;X3N<{N%jPq~K~%L@&C;WepyA_6A{<TkbY1
zI3!nYKe>F(Gby^;-e_tRFSNPgGP1>43bO4X!B)Lt&6<fsiq3vLLHl6<24DaNj5V<E
zWXIK4Hii?qobu+rHs5B45u>@NeytZ@r}-1`-#v`0m)<x>@niz>>7mQ`+$jr}ClTJ5
zzll%lg$4Zx2oLB7I1jA84jB9Eh~LtiZ99Krl}i`_;Q{@?dCvoWuV9EA24DaN^f6H6
zz|1ZU4p6F;`qYk6U;qYS00v+H24DaNVBjwUn7>5X{3Yl&FrdW1g(i(kJbNvO=*DKv
zDjZ!AL6ojnV|9A*yh*%mw_kt2KeBBS;g!PtRp4Lqc$w{i*M4Jl+^oL8##&rTzLT6e
zZNQ3_1G@AIARs&_=?Bhy&Y<o58W6sxSeZ}WS>s^kK8vtEh!pYti0?-w-;a9r7k;m_
zh#UrB00v+H24DaNU;qYS00v+H24DaNUN&IVd#f+kf-QRnjG70{S<1EN6`#X^nFcx~
z-pgO_L_AT;D|rg$>2@oQSjASJw!IM*<B3)tA251Xp~i8<Z00&X^vBBUQi;{fT{xie
zWY;wM^ZfyBCbfFR3|V>Zr&k<$J?D`sN9i@h|6$l-+r6t)DLXYdcqJk80pXPwg9;2A
zek_WLDfN8U?ETGepPpMJM+z0#XX%mC-%L%Ra@lJ0vXNWKbE=g}`xVl|fE)vQU0+d-
zviEUV_to5ePU)0J$y8wBUp|}=FglnxDqjTkXpsK)6QVbDdw)jn_i>!8JUQ3G|D<uY
z<0aRN)^t|A$0d&Yhx`F~T`yiDYWexp(s3>Oh7l=>>%}$<zyJ)u01UtY{{}7}U(xHA
zZ!Qv0nTM<MTod-}Odw`6f13~AOK*CWvO-7wI~whJ^?efluY70$3^<Mfz3)r#Zbo?T
z7IbrMzAKcP?O{>D3$>4INhcsY{98ZhyU}BYS9&ZFsP=xfS-*qTKIhVU9X&eQ?Q%8$
zC;~dR2d(|Y>ilZY54HL9o)^n%f2bb@U;qYS00v+H24Fy$ftWwYjyuDWC<qTM_XB0U
z+W6E;?+ZHH`Q>u)xBf-DZ6&Nxc#p$q8s7ue)nN8>ce(YcYYlRlWq`MJ4-jwtkzIy`
z5_o`z|MUaB@B5-Wb?d5_b-hMCd>AllUU5EJk0@aP24DaNV1N$;9nO?oyM6gd!a4<;
zQuiH|-p90;eq*(7>rJVZ*H4P6pL}|}R7Muni^4cKtrUELFaQHE;H(Dl{?@<0zm>JE
z$4{>niKeX3>gQ_CdVYx8y}9+B$j||q)F_JU!?rpDtp8rO7oWP;A{Pw601Uu@QUiLw
z4=h!l()f@b24DaNU;qYS00v+H24KKA1FXy+VO$HXEm`gSC2I3o+gjA_><0D>Y+fw2
zZcaa`b%ccdVqaE2_Yv1`Zypl0&&kB?#WoDU01UtY48Q;kzyJ)u01UtY48Q;kzyJ)u
z01UtY48VZ?2F6SZU+%u-ITh1==O=3~t_-5~Xz#gz@VW4OLoOZly(Fpeq2m4a-V;#!
zJXCEy@pB+vw^@0<uQm>!UcLAC9P^ymqu%S8n^hh!^Gn%#?_7M|>VIWMi7)^ItQz3u
z`o(ITtR7dyWsCvw_a(NQRXDmLg0SYnn4^zUoyCAz>kithr_ThB(2e&aD6jna{(v@<
zT0NpxQMlhD9^WbUt@VC@__`<DCz85fA*5r4&T?L(oiN}G2E5+*;>Eg*V5*hEbyO{#
z!8O`wy@A*pbV{d*sl;en-_pM#u<Cb}(e<jm&LH2bpL4>GGC!=G=e)%syf-M&tHj8Z
zKgT?wR+0C0g_2tp%D4MOK0kZ3_naM{^D+Oq)NwNFIX3F)(IEZpCj=h;-2-pmLx>j!
zU;qYSz*+<1=P!EQKfvF!%Kn~(FLQlad)A?L7=Qs7fB_gV#=v(;9s0Ienfj0St)!ke
z8xsvmg#j3V0T_S*$1osw9E)SLR(&f%K2CXaTKSwdz4<s~h-Z`YcO=(QbhP^_!jaJH
zKAqX+sC_?K$S1TPgaH_U0T^%&1FXJZjJOW?N48BO@L-Jx=Q#V(N*FNK0Po|FS&Lum
z(GrjI)LI_$!vGAx01UtY48Q;kzyJ)u01UtY48Q;kzyJ)u01UtY48Q;kzyJ)u01Uj^
zfYdz7TKB(qtaF)lf20tB!Z={PSmeOWE)5P)s+Jep+;AD$;w&|buRa>s3j@{|5YMxY
zZ8q245FZz7dJAfW0c#DU&vQ-KvonE!j_u*<e~+`X(b}^MwQDsH^9R{+XIK(7n!-FG
zQuBgnU5AI&{Ni!OtZi%UN5~HYFyJ@_<jy;Rbmpu#1J7)DcAPUDEr9_TfC07)Y;Jug
zGIT&DWh*K~0|PJs126ysFaQHE00S`aDg(mv9;11fxVbjp6-t%zRk33q48Q;kzyJ)u
z01UtY48Q;kzyJ)u0N)1sOz;TZct3)QDU3@Mf6m&rTKgT&lw7-g`ANbW)wT!uU;qYS
z00z_<__Y6bJ1hD=r($-9uQ=xYTglWO?+&OGIqlp-DrWvRAHJ8~^eVN-h5Lr=>9G40
zHJg1mdd%=jk0olk%40~Guw7w9EsuDwJu^NfoLJ4@0=raizV1)L8+T>ys5t+^J;E!$
zUz4(7)}$z6Hq~7fkjrQTwHnu|ceZab)ynD1YAugUi=kHW(bpsBqK|J8wKUo@%7+0M
zfB_hQ0T_S*7=Qs7fB_hQ0T_S*M>SBm*3X$|ir%BFvBT9p-}QMff{K}=*NpPrUnCIL
z`N?^oNx{{k2rn%5tnB!oHzX3ROt4jNShHp#0r~XcsMk5P3kG0-KLf1H&&Z!SL<|Ek
z00S=@(E9s6)8*R_`^Q8O-uZT(`Xjpx3njdgm%l6ZigV%g`%lVtK0#O^ul;7qoBP^)
zn;k~<;>%YTY{38wzyJ)GWx(0yc~(0vpVxe5^&`{+126ysFaQHq8Zi5PKXH9(+rI8#
zKKHjv#8dDu4@GAt@A>}s9h6myk2|)ldEeIAy-Qthl+IJCS9+<q#QR&sx9$2(;WQ$}
zMFlU^KC&g9*kk+k2mB-3CJ|O?^z}@<4zk|c0V`S#=+Y~In9U1K8kKnVS`cw|vNoTN
zv&AE~|LOI7M|_>cHVnW348Q;kzyJ(5x`D#Kd^jOsbTE<2tMgnF_Uud`ypz}Gjb_K~
z==~q<g8>+@-hgnw@#tu`%hmj&2>9<Fg!gUqzCS3G%lr4b-uGjjrMwqGJsPCH{e)=c
zD&O}ul)Ilyz*l<^uDIP>-MCjJD2rG{z3!u;9KPRIG3)pV{T_q?7=Qu249I<MXjHl}
zlfsv~FL_Soa>$2MdXs8}{OnQbI=f2OwUHVpVb=En>T!(U4Z?Z+#@ENAYn{pE=f3lk
zwHH?gQN1bEZ`JO%x6Xy|IqpaN`~3(<^CI+P5C&iX2AthM!9M2-jA<86d7<9Nt3a3N
zcZLn!L$&g9YNc`kVV7yg&iVHF6~98{a`O3JyFCtvP>^pAty-pf)eg<;C+6&TfBwBf
z|6G3dxaj!LM|LjE;ing?*WVfA`_VHhrrh)9%?ph_?DTdV6)1PzquXUScUy8gk@CiZ
zMLasRxPOXh<){%YPYr2#hgeNs*Ih&0e6O45EXPmobrk8e8Ccfnzxll<cn=(2!@Z{a
z+wM8%dw2DA&-I^whZh}QV)z^GIp%o}Fa4kG|J&)F(=%(>x8B2lS?G3~<knU@a>D=&
zzyJ)u01Pl~pz_ka*JHC2sa0fJ4uXRL7+}gkjoSG-T=R*iO373s5DW~!01UtY48Q;k
zzyJ)u01Uu@=>`(=el{)G?JNc5c`*F|p+Xpd0T_S*7=Qs7fB_h=&_Go8{Wa@O+eATi
z9xOa`s2T<w-oVw1gIoEgBoSw)!yjnQ8yJ3|aKIv$BZODJU;Ufrk$WE!t>pYLB77Kt
z0T_S*7=Qs7fB_hQ0T_S*7=Qs7&}^W5x79~~jY*(tS$5~Ga-F8eQP5fsbzYQf9GP;3
zItw*-d=vl!FaQHE00SxvxNhxUXwbX}0y?J$6<vm@EO;M7-uoEvq^~EYP67mDxq&C=
zPOLuenMTYe#+T|dzV!1itNk`}UNE&skHRhcjhmNF)iQKn#epCAM^L3ydOu3({9N|h
zEB;*RJUT*pR=#(n&J)WoAPNImKC?rUJm2TA$)&>LksJnKzybqr)~b0wq*oYGOQqj0
ztffD7$S<$k*boZ-<-vk3$Yho7U!ChZt?q5g3*|nC6w4-kIy4?JFQ3h7MlRSs_v@xl
z32zK?nch9iC57n43yVB!PI-_>%w}f01b^Gq45F3d{)KHAfB_hQ0T|F?K(D_W#4#wn
z-(_`ras1eZ0T_S*7=Qs7fB_hQ0T_S*7=Qs7fB_hQ0T_S*7=Qs7;MYK{>;5ryN+nZ}
z4G+AG3$hm{zh?p>|GR;#-=-e_sB#RI%fIJ!rvD4?2k1?)!?=bXi~Um`Qc#-*u~%%v
z01P;<0ein=lz#5K2GMdDfB_g_+d!jfBc4tQ45U_*ZA*v_2J|(+`}wAH9s(i#i1Hoh
zZLNHig64S8x7VUH7~sQz(!Wz#ORv}aHz-HlygX9g%svZZ`Ze9M-{%(5i)P1<`e49H
z15)!t>AhcWTgCHmnYB%REYcvp*Hg+0n|4ksz1uC7T1DaS6Qd}%KOmiR8jw4lz}eEN
z?F~bIDSCQj(>UTRRO<=(U;qYSz*!6!ec#<#JffZQ4E%h2fJeIxIc-K!ulqq#<;ZI-
zQo#TW$T6Vx`KMGJj`DX;-9X!0otmB@pga#o-M`YST&aGG{l*wj`+fxG>E>&mZn^8K
zN<H5c?^oL^!w(b=SmbhqXeIC8XL65=R64cCfz^Cke+TK6-)KMi`*`O(l6{YWY<MvG
zc%yt6fB_hQ0T_S*7%<9!)$`+v>o=;MW|ymW-U_qhM13#-1I8IBaQe-(lRjAlga<kO
zfcN{9R>iON43@fYs`q}a@gl!t8L;ZT1*7ZL>;Af<l;c>v7OjB+7=QuCF<|ugZbvQO
zae4q+;tU3^H;Hd_dQBMRm1e!?ZFD_)|6OfXc}D#`;tc2B-`n_A#a2migW?GY59kLt
z51jQn@b_z?ng4gG|LfyRo{y+ns(mlOUOw@70I6-Y&)JR2*X~O74_wElQC?~E{a|tV
zQpZ>AeqMjqYn?~RsB6dX(|nc`h{$8(jjmg}7aBA#f>=ePj)&RhYVDW2<X1aCl2OMa
zA@66?g5Azi)@fAxQ7#OyWZ=oU6RVGVrV+EqQrq>6<^Ag4G>_c-kf<fE<Barlh_YY+
z29y~np8MS*dH#4tKzSZY>`34JY26S4^5wzaS;6|>5#soGy^kZr&HMZ(MnC`6TE4x0
zj?c|CVAZ%Lb6bhpV1R7{K`zs~XSt*hF@Jw!*rd2s>xeyypT|jkZwcRT_Nex{E>$n~
z|23Z&Jh1J7?W=_IM#uZ#vCezs(DGLqxc1rI_Io@7iQbf&*W0Z9uWCE?!2k?6mI3kn
zvAk}xwk|B<IoEYs-P?w(A1<7QW_>TB9(~ufg->lhdH?%L@88d6muJ*_X=ay8`gCYK
zVqQL*T#kGF@y&*ni`blnc@BRX_+45t8|&Qp{*P17UCwRemBVk`IG*5A*apIbntpIH
z;|Be%Sbm#T6n>973(b1}TU?K`_%*AY_MR&`>*sjh$M*9+XFopTaeI1gTeaTJpB<Kt
zdcSfi0pUSQKQQb2((HOlyLnb`7x<iLWlFv~d$R|n6RT*pH`E6MFaQI*8!-F6w%+v_
zJs*M9$JeUwQ&#J>`govz7_ibn-9UNkzspT$Wm`};48Q;kzyJ)u01UtY48VY68_@gy
zPx)@EkNz5yK(*4fcu0=XbyKKTI`+{+>tFx|<QPyp4?C;*)XoFJN<ORRwb8p?R(#5h
zLp&b|Z`*j!?yvXk?0JxLEp&!7_WU^N=gx3U(MA}60T^&*12z1neS9uEn}C0L;C&qV
zg+-n<r#wg`&PHcGhtYNzV9kJ|{as|#I)p}*>uB$}$-N)NYC1<bAIzR##M$e!>bkDi
zd!gcTw6-VwK;eKzE=P!?QSJJvX6IA8?jkGsv_Aj0I=|d;->jz7`+HF9^~C7>_U7?7
z>vN;qhw`1nfcSfy*KJn+KCmh-N9(<g^7n$;aTKf`Pu}Mn<^8#U5Wil>fz^J7xM0AU
z4X`>dwqAe#OC4Li|GpH;ld2c{Uu{6@zDcd?{xNk*B~xeTt1X(hSLpZPtH;M$<wx(H
zpx5;o<xD;PQRNs45Ac9~!2WsQy0v?uLGvPrT<Z1rvsLA&z0auk_YEuMiRXuBecP(v
z8+z4i_Va39>SN`6KpZf@qJd!-GjI2dxlDMa(eHh;c+3zJ48Q;kzyJ)ufUyR8R}Nhk
z<MN1r@W4VpFt)d&bQpjE7=Qs*4CuYSH7o5x9Bdk3{W*=<@v(N4tXdD0m3qa;W9R!n
zPCa)yw+)o%LF^CPFaQJ2XF#uU2U^Q<KJ#cX48Q;kzyJ)u01UtY48Q;k7-PWbb>8ij
zZ%k`YDhya_z^eC)<koAg9n=m3FaQHE;3x)oeJ{wYc|cLmut{;N))BcL9Ob-lwsurr
zy7zi)b|Ph+*PFyQI=v>0>P_+SHfkOcp&qsK324oyb)HWleh>y=00v+H2GkfZ>igBK
za*bMlN+?%Yhf0lG<YC8v*70-BnxCB`h)7@n24KJ(1H$*dRg`L_Ro_Qaamn2;cfLxa
z)5(n&>0kf`V8EV%vJ*F#>G)nr8|WAw-r9e(!H=Eau(3{i{n6g%q5CQh{J=kg>P4ZR
zAeZUgvs_Y$TwYk@S#!#RL}E6DczLJX@dB)_qiQcMtB;FW_bpM6zUwP!eV&T^9p`PW
ze3ZfipB}{XwTQ2yTE{CmYksjuY{LKyzyJ)ufSCpwO&jraQeYtEm2$@|vzpH6b!pA)
zOIFlXuF_4vYNc}8oSjy<N1ZSL127=RfW7;&_}unZr>19!*u%BY?zZ3K8A$YE_<_O!
zi(HNnR;cv%rr!67<n&pjkv4F`^W9?2{1b>hN|%5XFrc3Sy{{8;%hRs~C<z7}*8uP1
zOIbTIa@RGJJFY`%I@Ww39vFaue;P3QJ~Qj(3r6%B7s$$bDu}~7=aKAt1Rh@YAoO<-
z24DaNU;qXj$$(zZMXW04&000@hx7^~deM=NE!qMDyc@9UKBr#wCglBWTCm$$$}7LB
z*eWS*P(0z46=x5A|J#@s#A@<>9wL4i;K_hq?@tTm=-qa+?z2h7&D#5fQt@Ix48Q;k
zzyJ)u01UtY48Q;kz<{v^l)h(TR{C1k{bTBsN~XM%_4S18#V6b!;g!NV0=!c1^_|Qv
zPw6;m*3v7Tf7;lqG)jj77=Qt31H8;{DYS&O=M+LbLVO?$z<~7zc>Nu(_w`t>Bk_Lq
zZ<<H$eMt1CUh$zE7=Qs7fC0xgVAgmkp&p}czWsCVcptaiHmx)|PLvM=j%c8U-?Wd<
zWoHv!>AC6mVZ$Co6VMtDR=w}0SG{IEpS4$yBVJR`CVmY3e0+dMyA3&QAUvQS;5@MU
zI>67B6OqCI48Q;kzyJ)u01Uu@asztJ*Ti}`yp4<3dmM_@<(1uet6Zn4ag<l`)_<&u
zSNRo)?HqE~%MhnyyH$t|24DaNU;qYS00v+H24Fx>1FWvkZggDYbzjZi*7F$tTUo5E
zH|y+iM0MX^v+lG_)GGd4-@s?m23(7W<QQEyh4MzL`k7w!>NStLqn0DxbNNbPZ=N8j
z&-rR4qJ;q%fB_hQftL-$HS}2QpYo78I$2%MMz8mytSZOZ|Gq~13kNK6IYL<X@bXm;
zTQC3vFaQHE00S@p1I}&$^O9IIFA2OmhIgy4AFTiVX;gexdq@5H7*M({125_8yeQW=
zGUW>8mHHfAlmY|#7?3)T^cr`fPqZk-*$lA$elxG}iTjb-?|-9yhYID2<HR-$z<~7z
z^!i=EOF7mXM+7hc126ysFaQHE00S@p126ys&TK&Kd#cXt3~h%27;r2Da-YLER;$q(
z7=Qs7fB_hQ0T_S*7=Qs7fB_hQ0V@os9nWM%tf&(P^j;UJ)^-1wI;E1SS+r`NBGh}g
z{T|OiB6u)s-6Nc_FaQHE00S@p126ysFaQHEpw&Q3zouLE``jYDQ>zE9`AfTbR&N*h
zoJjG{4ogS9UpbX%<%LC_HK#mCB)svfimj632E`NJSz<@}?oaE65Z-w8;^0=kDM^G^
zTIE~o^#S>fF`#xFIBWT|-bXaXF8igr)c^JICC^7xE7@-;B7^}LfB_hQ0T_S*7~t7J
z)^Agfe^fb!It$g#vnS-!`uCAIztQWGOO?+`-xd7wM)A6`HEQSUaLp&4@<LXQztM3t
z?VMJ6w_7TeqSSo6X6-lHhuZQ-l<zoiYvrSqcb4zA`slAQ36xjz{(a5r`kiXy66VV_
zi)znb<kQE1*899x=hy4^iCmjfzeA1MuU9+8<>(!+T#w>(YTtL!JD)f%wcFR5#5X#<
zCXA}3*6|Fy=jZM3Cm~+7$D7rB!u3JzdB9sfqy8>Z>Ot%8;zrX(Je?F6NO`5u?|7y7
z{upoDtc~YT8jo;2=9O0euGU(=QT<)%brk8f8Q|r2<gkmGw|mB1CXPn4-hbw$9&NoD
zxnTeXU;qYS00v%Vz^MBmD3|Ybr<ESR_p5)?JaX?t@~Xaued-MqIQ?eYNuMm@>{M^i
zoSdGUejhgMK{Ns3!ODIh)!VJQe<Br^+Wkh|hcv5PUY~ymar658$x-6gdwhgd<*6Od
zW_3QZ`;pq~B=VVSz^w6=yw)Qe4PjnTM<@BQNQ3xZPpMk+^8ChLoV@;Bl-Vx9-!?Ua
z@XlQG@`ub`oJl}<P|^>C`+KZV?|I01Esxdlq*liz{H_ql`gPl^dEEpbAPlf%fVJy|
zQa4J+t+Jk;_xEKHzhk`jjN{1Gaa4L;M0yy20T_S*7~si(S<gRtaUyp-pwho<l>WYx
zdOpO9PsGWFfx3aVw>mXFLmZWCSVj~u00TT5FnWBG+}{^+)A8*3@3EVGUtF(obboJy
zGyFfYafe1%XTD}rOT8T_{T+_<`)k&nwu!>S%O3RZm%NpS{_t{tFx$uHUsn5V=Dc9a
zE6tv-SMT3xym}Ynmiv2=)pU9vPkZI*eO<Qd{W8|;UDoKoroAV44;)^@{cZPJ?m6ds
zclCD9^`C!-7ad+=_#5sy=6Me<{h#f(-E+F<^voLet@rR>7P{Rgx%ECycFP-5e$Lhk
zzy3+>CQ!KJ-i(mKJx9$WX0hwG{!_*u>P#Jt`%bKntFY_>HJS<!ukOjWEFCs{<IStY
z9(AwK<N_yV2CRs#aGuJg$>IH@x|SuE#{Mw=7SW6Us6gU_dSB$rzmw#!@yhUylXJ&y
z&uf!Q2_6629h-O~hj+awY8GeZ=u&z@zXWO(|JQJX|I_29|IRC$o)NXw8eZj16}=BU
z387X|h(G_=$(Ov_CsL&h+C406gHIymm2G!4sOR4GF}0e)abTS_)?Yi+`-^ZQChKt{
zK79<tkNtUb*sUO{l={^E&r-zwQ19)vD}vt2JNpE&M<F7u*N<VdChdGyF_!90AwE!F
z1I61#RQ29;gYw4CuN|1;Up|ORQD0lul@|Eyh3A1>4~bG*6|q;>ah<x3j3r8`cf5_u
z)!Ec;>~(4t^==O<<;6c)P_5Z-F+?p{X@?L;y{2O_OLe|YdF6l+Phw(*hEYeOaD2qv
zex|6;OxrzbG{yBQ-Od;B>8(<CW2qFCmMf%R`uUaUH@`_BW>ctNEqy!mKHs}e646Sv
z^>@ASje9%)bmFY+y{mP~mr+T?YCbDJW>)E$X@qx9-d?ostg)GdS60v8^5ngo&xu)7
zd)^}7s|^%BSAJHzj0mbXUtKHqvSOgY+JV_mPlr*p)Y^`aS`oi`Ur#2yGPT*>ThaZW
z6Rou8@w)3Ttv-I1NO<AIbe|PVFFqi=(RX9x<xdCRC63A-WBP6MC=pJKqPYK_kB(d1
zcKCBDmr3gr4_Es*m9oyB?LFg{j!&noQS9HU?aO7|wtO4@h!{ny+NHK$y`F1G-49@W
zzt-nOTJx(terofbnCsu_s;e)tikDZ`s~+roh44nH=Z$KQ2l5$f;B?bIfyHYi5ND@7
z%gdV7PoMQ&1mTtWTxYgiQ1&WON_#%6`aI9<rPsed=u3I!mF$md7oQzYwbCm8s22wG
zFrf805G(n4`+a8C`z=P-bNAD&lbUCoBwDFQUq%`H7|?s1rCH_uTHN)u#F-(4cXpUq
zrfH$~Ly5D~=rd;GFBdO7UHX(LrO|O&RsM#7uXh{JCyB7m^2B3gu+JUhXk50x?Bt-&
z&Jm+%uU}1@v*63>`x1!NeD`+dgp=h{iBar5zhUd6A4XHNxMXCzCw}YWsZ!cIPI`a7
zZ|K^$>b#atc;%-0Uk*As;346ic`km}Y}@8Y!W%1BDOj@5*+2qX>w(w%twP*Nf3LgW
z4sWr%|6T$r^Pu<lhurc4r^Q}&x$93HopOD!md+#U+{RMBz944PY!9uT%xQY_`ZK~C
zUEbLDs8{z4BA0r3js87h_o!ZJR4MhUUtCVVt9dV8E#OD=qPTpy+iIVyZJN~j{KDrx
zgm?CcA6F&qy{ptwDfQm3&`%n#DphsO2a$v~^7cLKGPv_J-<$g>@ASI4JD^MCUdkKw
z`kgJDH{f3eW^Ao+`Rng<*_@5`j=$Hnz5!d^E)!O1)qBLO%ztR_xEb|%r+3cOoqaTc
z@=B$@n^{Xgx6rc5&zq%Cdo=2}8&$4P$i1}Pd6KAH8Wo4qa!0*r)!E(u4z-F(<7GAd
z$1MwbxvWT`MzLPM8M^}Flc-kS|7zMB+v?vYMp3WdbK-JXJ?=tW_P$5wCneUXdiD_I
zl|AFySJ@bIhJb&05RSK1Jbte7=xsL>soqrDp46^=Jd3AYrMz%_Xjb}^s6Z-3r5?;m
zUua;>X+yrbOL?PJ^V8~GZ$yzYy@nKtqgrWJe-YQi>fgOmarqCO7xkjkexj9PA9A-@
z8Rw2T^qId(>``t%Ksp#uV<7PB*RS-i6G)Aw)N>`d`_;4&dH%lv-k)0<H9pv^a)ot?
zc%{_eC%o(*GpJGbIxQn8uRMEb)%09(4=Ascnx|W8|Np<DS^K{(5jWbDJd@AnXv|xU
z&RI7pw@si~miKJC)NyzJn-qix^aK0nfoo^CRrBUurmWDc^VO{Dkhq?zW53>1IyHqV
zrMMikx6?n*`?&VKdxRD0)&5JXdsJ`u+8rvFdezHYIli0wEqC94np#cXyouv=wD$uJ
zxlH`(ogQI?7k2)BeMZIPXu>;N{I)J~*Yn53DvJBH((O@S*S@;`m&e3Wc(26#V*W?s
zh+Hc5#d`V<wbHMydUlWU!staIS^EwLQ}AUER`oOSanL&stLJ4v{r^5c#+?sp<{Y{&
zgYe41R|=OYHuVMJm0HiUwAxck_gkOS=4e#9PJ`glsCo0H$}QRV$CJC}|4HOhh?f<X
z_Z#r&+kl5uFS61uv*O_G{YfF-m6zj6wObKP?9nWLLOt_G4a>PdxwwsYcG$9f$INNP
zY-UmL2eN5E>N%)c`vp7uPOOisu<Qa=O0DA~h5SMstWvl>^3Jwtot7S3a*gmtv*$x&
zwI3obz73?umK}97cMLU}dRzKz%A_V&zFJPSa#qnMGg5s1plYdiyVRCv)O(C(m3v}d
zVyo(NZV<C5ey@`CZMFSE?|dKSEVCwZ%WF2PsI@(f%hlP`ZtQicH|LNhpRai*gz`$W
ze&5TjN9%DmD}Tj8dDB;XnA;|oX2m70r)IakU7}qdP~JE{!N10p3HPbq6#G%T?a{|0
z^4?dOR4?Y)GId^jo-E2MT^2O0|7NvR${WWvSUUKJHtCdi?s_`N_r%c*YL9CBx!Qc?
z9)7Woj7=nZQEePX=WAK&SZvs-6VzxbZO6pSfECdd&Qp+057`|bdGB`jqxPuucx(N8
zO=*3+rq_F31gpw3dLE2fz3WC4TNX{6opSs6v+`qRm7bYK%%a@5knZ0Ntmv3)MLdn6
zdh_3Flm1le`s8Nk*Sapfv=M7l$el;n+9>~7`_9^VcJx0GNv+~PBf=-IGO+*7TZ0nr
zCR5g!l)v98pNJ=vH!ke{N1fd<QB*Epb!@EM=kE3R=9F;{snImLjoBV2y?bZgB%rw-
zd^a{;{&e77qLfDW2dVN)-rwZbYtLh<l~VPXy}u*1#U6c}LU`pL-F~hep6@9!n_7=k
z{Mes2husRItaD81*W-^oze1H#s~@@f*RBY9EAQ+RL@lN2k^6oe>*=BkU7A*KaXjIb
zYR9*!eQ#E(Ul_e#DsG|vweAOfm$rUP<g!eS$(KftJ58i`_s^qG!z+gnR%z6{eS%lL
zY1RE$z3bJwj<VifQQpf2q}H_`=RN=P(T9@=uPkvoU{lQ0D58}wTa(`U<@uYPW<{n^
zR=KR`o9+B#GpN;+HmNOz*8979=a+jwr>XnQL-S)3D6ib|;fJ{zE`LPD)OIv;<Y>|3
zRGH#uQ>apEtyi3XQi;xmj}1(sN+~Xf*X_V(FFX(AdPoHF+77GYZdr0^><{B_5oh6m
zm^+`7RyT=K8hw1^&Wk3b6VKz}Xxk0e4$OXfI*jth;PfAtb?q5Rd8bwDM;cYH*7H~C
zctoK-qsJBH|2p}Scl$)j3WauydC70+?U)_Olo!ffcTt>9+-_|1WkBxzNFiMnuX63L
z$Lyl^X!LyVti6ZA%X$KG<5c^6hLwC*vOlU_e0DfBigNEGI7>RE^K7t|-l*pjQsoNY
z5=wt(3F+0&hnamjsAQh1;Z%w(mu(%<Y)u3eSZ3=Yk20%6iL+5SPOQ?X>#kILmmPGQ
zTxZ`k>S)yaeNm&!n|6PZt@Qh+s8N*a2Xg0aPaj$T@yxnssa#6cYu5g&7q>V6a()`+
zm0HKav6f%$_ny`Hj9&lSs{4j&-`AEpf0e$ESSYkamhUHVgcVA)Pp|#k&lL5UX}d>3
zWghJLzu6-Em*88eL@x^WO?ao)`Le9e?-e*N;Mvk!lvhscJK1m4>klcf<aHfOaonR`
zwCe2ce~0qMJhPWx|Nfva)tmONqvFp4deunVbvTPyO>taGw`b+(QhGwa1PYq#p-NEn
zu2tVYp}aDpNSR(kio{W~Xw?00Udt_fuKcWa84;9M>a~7>UgapgZt9huwfhm`ah6is
zL2m>;uG;JoWrcc=vo*UsVSE_xoS&3fqw3j1ls9(W)_=<QL!GHw>UCWZm!tIl5i9Af
z{(U3%DBd1$>iofP*JTpcdC@a$S+)CVgcs(1=c!xmce9C7s$Jhqd|dU~ejauC$0_q3
z5!R_!d(A3G?|G-i<(btE)MKuJcW-A-I9WcGm_?&om)f<DXYsVFRBsv`_tZ&aTZBz~
zLd>SUcJ0}AspIbcH!1JzI_}pQ+ZKmVDcbX(_vdowb4NOP^WTqbdaG~9p`13esCR$R
zTAtkTHLRzzH(p%t&-J>$WmP$~j%?}Sza^SF8||GZQuoue?$>+XQ{InT*Ub}FE_(Y3
z1&#4wub=DnxiDT^icNoyJyCFaEY(V*-d{AjT&u=W=yjYYru(c|dhr3_g+{N>Cam*p
z^muEb{Kd_t-Uu!bMtNgw;*A{M^`fZR)T`eM<+u#)Jk9s!ek#Qt@#Ctby?2$eMyswr
zTI=O)K4*LJYMmd=?EFTLUl-0RkSzlnrVsLH&@+dPSF&YOc{DxSd&VyvpH6w>(;ZF9
z?XI6eS)*_S#t(>^9B}+0(VM;VRXkCw(mi4|g?554V5Nb+hv#*Fe)1_Xn&LKwrnWiq
z^mY_c%cs}cT^@QRobbX*f0Qg*r%Mb`%QLxs!@jH=L98aP?<qB2RjTTm4<d=9(yQyZ
zPF+XF61^#$w?^NeVXb_l-j`Qfu3qyfnEgDTwfS6>#>4u$9#Zk~_8yuz-g-^PWR~iD
zoAOSj@84TJU#8S}8NK)O^4?Ish)-{ox*JP*W&E;TJ=c7dNO>i%-i5fO=G9_-|E!`-
zW~BK1L9L>_^~|L{cfTFpVtN0)gf}W3f3mJ2C2#&2PPJ0_-4C*5U~SUfxTghEDDS*`
zWJjAib5bd9{J3R7FP9Z5ly~yFeu-4vdX2XrpXHyrdP6$Zn@Q^v4_Es*mGa6CoeF%@
zb?sBiJEi&&>-(*mH(u?x;m)~#e)a6}bHXa?6d(VmtwKqgR#rW5wbbq4qBg6j_V_5x
zCpCXD>-!sY^Ip~9?TyrEYJFa#^tc?37<_ul!8FPnt-7A+UGIROkL14cBA)Wj;_V`;
zdT+WxS)<(ZK|JrMZ`BT^HhdpN?NRGEcdPU3HSdkR`B~(8H!9t)lh>5`vUnx||MFmz
zXSL<Bvfh%tI3`XxTDn!rBkCyBYdz>6zAbJ`82ps*#@1)EW=t)WO!T7Ke$Q&Yj&~M^
zeHW9<W{*|o<b9spHkZvP8s+71OWTrCPvVJDoI{#?zUG|}$_veYPsgbG^tx|nb~%YR
zmi&6sBb4&WA?4?6t?=ug)My%YJrK&})xTA78+G5~kjuoc-surW^k&<%PD_t1xkltt
z?R<|ye`1CGuI9aXwSXT{%T1G7pI`Xghgijg;mvk7TD_3+O1;KiT2+qP^T6tS;(7Je
zZnJV8i{lWFbJV*1*}EMx+($p9(Z46`9@Q&NHT4=-Aa1YJw$|~TtUqs%iqEWh-;J(^
z_jM0=y+4#5TXxjV+%Z%srOr#O``;dSD<Mb0ZPZby_Wcc3@=48y!OMQ$`V(*CPQ~%I
zO|7wXnC%eZokqQ9#`^q<LVQ}U=TiCa%?K&nbJRS-JLS%&BIU#E{q+v^o;s@LBf=|t
zjOn-0qeM92mA)N%pYL5KiReYC{-gJOHnYq7dFeNAeez%@0j=@iyRq@|rvvX2vnbV{
zrS`M(J6diWJ|XwgcIQc=j?R@)%NLEVpFnvb@5kdr^J};Fxy4hZw0C~#J<pO}<=Kl@
z>*u_!qb0<P?@4RECk5|%uTw2Ge*<s({VRCQ4*M*I@<P_;eMLNM&c}=6!uD~|`g=fa
zexsi2tBps<r*@t>*7J3CSv2L#3l|CNT)QIZt-P~O5MC+Fm(MHBzK^AMeVxBwpHVS6
zn&?I0y1*(E7mRDurSK(c6(f8bZ>5)V*vw}1?HmKEr-u?&$m@M~dvVWQS?t|rBhraI
z>U}<Z+<8gV!-qL-ymQR4B31k+JR{D=l&(3i9_m}jCZ^u~L}__jM)y1r)8r}9i{&1E
zv5t&QB)l`v#qXMJ+Z;)hQmKEl(kq>p&))lvR(-B@y*)?EuUEhE=;IN2@2gB|HMO=!
z>G(vc^m@JbE>%vIpy*wzzI{UVrqSbZjVgaqiOz+O4NRh1skJ`|`MYlGKV|%(&Xjj5
zo$t|Jdcli5${pXNG@ZTrc>R3Bk@Ba`EY4%oi^6j}N2S*HIOOKvvUJ$+jW@3n-pSiI
zW_$6nI^I+7>rSf@BVwCX{)6z&6LbArU3K*(yffS5q<8Pkn}k<Zu2QgMp|gR6SI#;9
zR!l;{K*B4hwAozw#r+V%J4>7n*c3B0iWo(u*HPZTm(~70Q|d`=`Wox69qRo>IFU<x
z^|Whyy2^RKBq~L(jAe5UG)khpQ0aX%*3$Dnf8^ZpIp)s15J&Z<y?!kB^CctOJ@H!~
zPu0?1J%!JepVclSg6d6s^_l&-(f5<(w#Vvqk<`|&^n4y(w$R3@ixQ|(@-qJ3UYxa#
zZ0X^@C7K$=di`eX3W!glR#WZuhqZij3oV=cyjcomog=Gt%AIF;3Uw6jzw_3hguBVq
zD2lJgQtK<J-7l_3YTMp*!`VKsG`P8&SJ_0$J0rI4^bDV#OtrG><_Rkoz5RrmMYX?c
zrTRDa8*4!Bdz)6LvwHqDt@Rr{pQusgZ$DGiXQu5QHH-dk>l&}gnL@Nu?ODUx`VK-o
zdXEEE>qqZ=TGxYdwETMewO3w;ZQ-5c1D_CQr9Cg=&yAiprch{!EZ<M!h~8}R+q%eI
z&mR-Hy!+?Tr{R@Dh*1=K6}B6ftFx)y*y~g;3gw#3F{NLRKl1zv1+DcUb=<}CxJhl_
z{dDW3<{2l6QVQ?K%ca`);<UbhFZfFB+Q+ka+Er>4g?gl@b)9D3^Nao1Tb~P`|NHu7
zVt>4DzuPFg_5RK0sMX}vpS`%%-p8<4p4QKwRH^kw(6&^<8cXKhpIWSQ8sU|rzOH?B
z{V$J+)zsQQjLsi8E%vg@U4OzW<*t9CS30AQhfu!K-<zzZpH;NUj1-?gDDTwz`&w>(
z;ka8x*7}cBJSF@7cyibLKZ#K^>wdRh>sFdwj?(cvtflAuIk6Cb&>Mk|t2TQ?m9jkX
z7#Zwyhu9;pzX!x|OWogLeLt`33yd7I@tb2_E2&wu_q|zYV9jYmzPU?zC+p+yb7j}A
z^})JCVm9p^cYF>5S`6qlu2f45YVxj_P~xL){Zoiq+M5?j>htYwv#Q@L{e&1zHBOu{
z&$Pai{YJh1keW?rbd5GUo&nz91Gg%Etvz9T?-%0^SJ*&!XLz4x$Adht6JDAAdEUph
z@7*JyG7sYZ&)WP;;`p>~U&;QccJbNaly~;4<FfMPk|fGHje4&~Yd_#6|I8mQ-PrAw
zLe$dgc9PHXPhGtso$^ZF*O?N>Z`6CaMwMG;>mrXbt3wHI)N9^%tIMhPMZWwyNe&yY
zY&H8(rw8#RZCa`II+{MR{^OZ-&r)8g^?fp{^XvUSlD+cO_LK3US?N=v0;$my;*%P$
zAhloZ^QoP+^X%w<Ad(6c;t`l<_R{O$AM~YaDU_p@R^5l-y<WAx#rb&sp0OAA{>R&n
zU8<ea<|yoO(|_lcP0t9cj3T2(ZC#i}c;~*MYu~E#S~}62;`5EyZLe#61Gc(dCTgiy
zyVZ_!&?}!*JFJ=)Qz|aG`^_4UBUO*wc<rT|m7`1P3H=hNQraueQ9n2OdAw2O|61Jj
zwZxergjY_TG`2<9#3zJT>is^tRCz|7fA>nvFXn$Fj!04M{V=8TPaJM(TT<#tJW)!e
z?=h*J_n5x*ym`CAu>^dz2c`YhcXPkx?)y(uy(rYLbp5l~(vkZ-E+-PRDa5OmM$N0B
zG#<I{G4Qr-*IM_3zDrv_CcJXmoCRN2-<Lpm<)`074(@+Gmhi^1cY6+f)-ay%PO0Aq
ztna_`LCu^)_hk^?xMpUL9SvKg6JFW%!Z+^i{L_iRc>l@cW){t+0>7_Zu|=_jEUK15
zIa--{aLTl%F3+hw&TdqpVXxHZR4>jib?45vCDW-|D*awAQ)BX_(c?}NtEn_@z0$W_
zwsl0aH4)Tm%5BfGgKm@S?7K$wqTG6u^7lLC6Y+%VMY;8=O(*r<g?PW(dRWh=bUkAs
zy<Y7K@;*LlO3??z(a6gD40;_Wz00{=)@{qT;g1NfES5gjd;WxE!Yk#@U#T{o+WsPU
zT~4FYsg0Yrd=(b?Wk%o3q*__OsMqO%83k-|*<s+~`}H5^wt>#+;j7dAt92Sxz~*SY
zJ>aX1#(VPHKw~`UeSPJ19Z|XSUa3u|b{<E2`P7b&)GJ@Vt9dV8E#ODgQt$Oeg?_Bp
zyxc~YGe5z<#+3>8sooSk@J6e~DM{6<_Bp#$KHiRx5U)||w%9Az?0MpZ`pkZQqIUd)
zR6gOj>BXe=iHECwoJy@C>+2LMjnC@)-&*T8>VAP#xq9_yl=Er>ysftr|71b6X1~P{
zrF^vo>-VbtyTj4*>2+lM^5vre+e0ZYoI{#?zUG|}$}6>cR=fU;y?k1~2d&P(qGPTV
z@id0=%HH!Ewm$k{H06~>&s(S0b<R=C5uTqLMZw#Kfv<NP&?kwoLbdb4u#)d|(>{U4
zYa|ffIDSCX<bdN3iCGl-pHj|$KV}ixA4c@1(l}U4&&oV^YQ5NdF2L*OuASXh&6{_b
z^3M3NKW`4Z6-2dCZNHG4FS^jBX$2R@6Q#6QzSj9GjGp&IxUZ{}Bdc}Foo9Fo<%J)&
zEa>I3B8Bowx$i5fO=qwFN`21j-`PUkTL0d*I=|HYVAl8R_1s-+Ia2*x@BLEGL0RA5
zcvY#YYd(l1a;djJd*w|`_gS&@;se4f&3?Wr)MwA9QJ?F*UX$K&wk&llHtf_1YBZ(d
zRyx0c(dnh)Rl2|FlIykGUoT|ijX^2C-3NNTX5*CyzIUBEH=>};Xe#XwM$JQGRJs0c
z>l&}gnL>DHxrbk@BV!YZ(&MEzy;1Wbn^kV%bLD5X%ZQ-7Qtff$HQ)X_Zw*Sgn@o8l
z>+2Xu#i#dkORfDcdQnK$zQe&(D_NNj%<ZR7yn-u45#DLk`p;&StJk{pD5ulXV@s|P
zwH^i!+gR8uI)XSGYj)e)CEE1?WsQ2j$6-}@v57Zwc-M=fyfOQ7P{}-1!zr&U-Y%l5
z_of?^S4z#F%liJ6m*YycTM<lnrP6uW#J?|3%u8%lea;P{mG+La(skwD9(OAtN5O4W
zE)#|~+u3OKLMoS?T^3FG^1?-;l_lC7t@?kHX%V?R<TCNAcY1^oDT<Ge(rvTHeJhPm
zY5KE=R!z?p_ke=0_F(TlI#xfo`nj^)`jz%qA^r4wgKQ^bZxf>_)F-A<^U#XR71xVx
zbq0*?Cq|cV_IzW){d}bieD=cgK(2?xD7GpwBDPuOKL~Hsd;Mf}XTd*n=^d9)Uagfs
z(0tDd+KgtMJRc3ZQ!%fNS2ntQ@b%`lM8X>X*;9=9WXZ^OPyE)$Q(jpmD0<hbZ=X=H
zo$Y6e`pmT5qgGQKm(km<sVy%peU?dh<&q`cH%zf*63|=^zrNV|dr8;aHnV8-dGPk1
z2MW$^mrS%$Do$SiZnt{fTD|Hw`uz%f<xBMorTf?I*-pPa7)119w#P~D-kCRvm`dYP
zn_lkyN><aUU4N04e7xT87vgT(FsOC^kM9#+S-DEVl7-F&604~8yk#|C=_2VpR@ccQ
zMpLT&A+;iY^}e1=%%W60!hY7y6RCC9h5f9>tu-FA?-OdRkJoj^5I5iJ#F>5kgzKc3
zTVLN=eCgV3Dwlb-Oq~~>CyPokFzCtZX9KgSR(|>3tuv)6J)>gQo|kuM?VZo46ick5
z39k2^QMK$mWbp?>K7K)!@_cmM;<m$|Q@toYZf~u~(|7#d7t}1e4DLM5_vU_T6a&A0
z{Yw8jfz;8st|29F{uxfKrrdFrX1{MO?$>5-XLo$$z1!W7f-idzjsq+6>8OoE?l?16
z)A2fPR2+BWjU~UH^a!QAQ0V_g@#OU(R~qh3rM&S(^J};Fxy4f{UdjHbcJbNa)GFFr
zS8;gRLK~+pN}!IyU!K3&X;x$!<%P?NzS+({HiLp}cxX8MWRcz%pHg1wU%_j3*k>`+
zXv+N_FnYcXv&y$OpR!yJO4G^x`$Q_8(t7nuZ}s1IdevWPaC0}WvWb*e+WX$e=lZTU
zE#;Hf@pL8cZ*uFk=P@;lYOg>0POOisu<QcmjjWD$P#V{}jj~(s-+Yd0rC#HJ^eV^b
z`R>K#tL^_%`OLchH2ppHM8WB?lod*OVQs(O_XqT9pYP^=%iZ^%ro6IeT>B~;W6lu0
zD0q#&onv72^iU#~<Gkm8KKgJH;hprY=gr#{jwM!6@NX15Qd{iN$0>wYrZ(GqE4u%4
z!aIANTk_Z~B8SZ?8r8p~$~F4_jnU<s{oGWq`j#DZn_Oq#HOecu_kJ<%aD@$oR|?~O
zjiS_f$J_ol6ZdEPf0Rm;Qrymm8D%maZph&$mwR(=EAPJIIh9Mf<EeynliHQ9xjr+S
z%B8)yJ^FY=-uo((D&?jEgZ30WkwxWF>+^!>-xGF^>Xk<2^345T8ysDfL6x$|(rG_#
z%}%FUDegZ~+roMzygcu=7pGqLmBr;q`4jJ-)w^y)v1QT39>wYP-d1{lNNIYl@s^oA
zeO3R=r&KAm)+;x^-s@M&t=HMo89fh(S>-QlQa^pxcM(J>eK$5<{&e77ViZSxUHj_#
zUmg=)StztbmhUHV#3;5tn>Ax<sbm8FyNA%!Hb<V`jv~C%==r0yUbhn$jBC@S@FmJC
z3k|F}ZOAuwsa9(BXm$Q!vnK6)Rxy@>@L)tgFuQ-JcJ1R?JnbrFjat{AlbgRmH}6#q
z-rh*9rqXeAQvJkz|FZ?}6i+0)Qfl5>*7vir?z!4HR!k`I(YF36gm;#`+jHo%hVev-
zQs<M@{z`w8ELx{a3~>~iwa$lL^_X=Zz^Hn7U0+ii_uiU^ujFnTO_Y-N=j6P`FLhmG
zZGWeC&eWZKG=VxAy)u@~InXGH@=9LULlNQ@o{#BG*3KWf@wj$&TQzUqWokC3-CtxY
z{r)M+E4AJqvpT<C$6IbcGb^3m=e51^jBd}Q5}gYl8<<3Sp;12Uy$^`bgWk$J`veht
zFzPt)R_?tL^NaZ(i6f3eUY~2G7pq;#Yf~~&N~8PR<n2Z4&KjFZc%@N(8keiHsomJ?
zlvhs73|J9e;XIW~y*+BZKcH7U>9J);-OL?BS>=`wKg`u|`6J3JKW<sj%Vk9h<&`U=
zmM<DzKY^M}z52JfoEAM!l_`EUg_=#hkN+jVp|@jpBvV%D?)CWQlyMKK(G=E~adbX7
z9lv>hzYHp-xPPnNHfr5xaXebzM^~F)9Ea4l)cg^w?+<z-@Nv~<kEmSQ^U*tB#S_IU
z-6LjG>%7@Y^Vd7rd+Ml~j|lISnzz~Bac1rFr1go1t9_hGxo3WJ=e5!Mx)ii~Sl9-i
zL<$<~LGAgsvv!^x{SQP^r7W{`kw=-;p+qj_&Zpq4>Es?it@DS=tq18G+kjW#ynts*
zZ&AHzmNoJFmRh&v&W~+W`|8GKjBu%(Mp$S2;JK}TxSUCNV|1ZQ(+Vz*C%kj~fT+m<
z#~%{4G^+o~E%*Lc)85!t|2DCj!usYt;>T4<d+#dcm1ezXA@}$yP1pJR^%)hDqlr=~
zt;ekNYUi0z8sFuzZd<+$e?)kt)O`kC_W$-tFS@}sg=nSH<F9vmqu$Fn-Ly|&@fr!l
z(YU>BR`r{upAgVk56`16|2SpdBchdhpD)*4e`)pcvqYjd^^Oze!NA`QRC$&@plgi`
z0vhY#@6MI{f8(+TquVBB6Qz`@gZ1^_j9PcmQOj+)Z0m?-Ya*yzO7#!5`}b_S)NyzJ
zn^Z5F{ocIzeM4Ti=f5Abi0ls|ywdHbPrQOFL=mI;^U`nL`sBe*;%HQR{wmG4+-soQ
zSABEY)KY1Ea?{`J-e~2H8@X*DJgDghwO0N>^F1qQ)5?)!Hhy!=YbE84O6QAFn_lku
z>}=^qy=c|h-Tw|Xn+?_u%zk=0jB2IP$E!oF^sB3$-J@nv+#ap>pRCT$`~6p?^PwA^
zUfe&#+xs7HJ9epdPMfn*>_h4H-9L{$4X+$RjH1$b^h)n_t#80qx66c8iq|z)x~=s&
zy4LR@rTO)qw_9#`TIXe%TWHzj=gm?m@8o`bHmBpvmCmEDS9-m!qe{yeJZxiOujmM(
zHy!1@|H~`uRS)*PLU^av`=3htm7}Ep?A?wT?xUYlt#lM0Mztef#HY7P-HoMMX;geD
z7Y2ARVDx&2td#GUFCPuq9!k~HsAKtSao5)pXNC}?X;l1<TCUdj!&u3G_Ry;7x#AvB
zwRF_ui*~(ipk}wdU7}qdP-mgm@k|3^?tD&K-6VR^>UC*f?hV#$sr`PoI-l^~WN`YA
z%ewZAq`c9rBR-_{PknB-&!ntV?tCpqr!(ugD6OYy!=TpvKfX_RXST;l@7|d=39p=U
z{H>UTf`NorcHP#0%J@T_siRQsyjW`UDLrrPrT6MOu2a{MvBc47ul-6tH*5Ti(e=n3
z_b!!Au0N~hMSk~o=7f{wQ;FWZ*&_Ux;9IFgFUs{OPPg)MT&Z>|f{E1>$E9{#>-tl?
zzJH)s+`_yJypi|krsDWpmRuV9!}wc7?s>rLdf`Ic<sN>qj*LwtdQ+}Hz0y@!<d+$J
zGn4YlXI@DsgFTZeuZ$>Crq_@nag<lqI<lpQ|CVUVE7d+PU?tzIqD^L``20b6rC#@y
z#pURAePq2HrSH`!ZIAEfe#_nWpQch&TCQ*(*=$PveaGtl;bjYLoVqB1+T-zal}B&8
znMhe>#MYgj;nS0;S@d^X*LY3N6v8@tjOn-0qeM92oxIL#Wc2&vMwf5Z{P9Y!_pGL`
zT%};iLT3XB?^L?}0PE?OHL0II>$?cTD@&XX*c3B0im=Yct4dW}^Fbt0OIFWorEzUP
zQ`Bds?H=WgQoqZ2**|sC*cM?EpZw$gVc~P-XSK_SpjK09e-fwfI_}pQ+ZKmVwG@}9
zbX)1)iH@3HsW-FJ`wyKL^`g^$!W-MB)>t~sc8Ksw-ku-Xi&wnvvf6E_ey?`kYOVRC
z;!?Yxm2vE9<KXRgi#Xmw18Ytj^37e!E7gu`VkO_c6YJwDEW1E?r~B>j7R&qZCA`t>
zae-3T$%WG{KNz|&iSSCPaZ2{uZ}xmzQuSH=caPTk$A@O6Pl*boM$z7JWbN}axqZXF
ztQ$eJQt7&T_R=@~J@!Pw>9Ld-^12?o5H~NLrQ$SteQl%4*L&SMv&s{XYv6U;s`1eF
z>Q#FD#DA~wzRlZvdrIS-Lz;ZP=A96#maP4Lkc#J!%fzqV=@CYZrVz)JHk&KIxF13s
zotq}LKELp}4>5~^U$e-{{RgdaBrX`&rc2>V)GBH{UcBD-634A{K228AtNor(o6o3m
zhE~rHptk-zvzK1~{-7`AmAuX0Z&kcnuaiRlN`I6rTBl14(MrAELz7!hdEzlL*yj#0
zi}RBbYg9dZi0Vb7=GQj6T(f@93H1p5)@n}73|J9e;XHL#TJ0b8JBNXOSMy%HTELH}
zrQCU1#5T>Ihf=72eyKZmzAc$f?Xh0J8M^}Flc*GPtbVuJ&{}C!ikJL`-j3OkOedA-
zT=>|)B-uP1G5GY9gK5+rh4wfbYplO^sP`A)1XSihm`8AAwNAP73{R2lxADt%^<48+
zB9&rO_nC+0$0kstIpF6bxv#v4r}pS<=K<Oa11uZxh&s2i)UPjyUQ91myOP(YWMVY!
zy|=2@=Slhdo$`rzLiM812lm!Hd?k0&Xu>P)jT2_|^ShsJozy(z<Uig65srtzwMlp5
zo)%1@Mp3B8Sy{g9`-iLM&n8;AW@e8a4O^rWz4*tNxGKNr&SP_Q&RtpT-DV@wiPfAo
zXTg`%_x(Ti?mXV^>FOVN5+a5WV`C^qiJBU!=HYuAQ;C?XC5D(o(Hg2%sUf6>swzP#
zMO!hHlFDH0H!)ACAt90wnPnoAgorf1UtgYgo|pYNz2}@eopbN{$H&@x?KOP%+UvW|
za8G7u<zl6rpF+R0*3j{7z2jM*eeK|P@@m7Ye(=<mf1mb2UhP<4_lu^oGj!)S-g%_U
zth`!rufKKPy6d4c^J>NQhV+>6>({2_)r#eQ9$V_VhVG+b{|g3g-0pvv=zCt@6MA;)
zx_Q?vy0*Te@1cFWw)xM0dv@r#de61Gw(8TjXUG4@-|E`BYwK={PW^S?o_Ac;<(+Jq
z;IT*sdR%z(2IF^_oWtSozINq|Z71aL!PTC6{QB)C=H=ok{TFw<q5J!JbHyTQ4z*w)
z7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZvfqCQZ
z?6dNqNqOK`^N`Th>hkxF`uoprnV;2$!`8ZX@PNUS^AwL7@PlsmKRGo|@xaqo``!_K
zr)Kqy2b_1-%zF+Ql{Ypnyl27@du=m5uW$6diWJwFe0J~7yL^xr!e8%m|ID4YpOnLv
zQtxT~h8zquB?G;tAAiB!58a-Xi%l7O_znhwfiE-Ax!WxrdptcLU$oNw!_Qp*qYT#k
z{P~UYe@w`W#V=c}d};DAzr1$NMX%<C@X&vEocPEc*XPxS=Vq5a-s=DV;YoAD%CD%1
zFc=61f`R76z`n=cbkI7duT~QnJTzt<G_O+>*#rZ@Krj#tG!z5**8QIO!>Dy@0xOG$
zLi@eW1CJbb%7BS^bH@QE_Py`k;}&M{^gr%7bk?|G+0(rqJZ*m4ck3O?{hYASKE-%2
z5DWwZ!9Xw&3<LwgKrj#tl+VD*8*J16(|mefD*kNib7!>Qaax|@vpcUp@T4t%lhuan
zd~@}h{qG!_S1Uf;_RnA2;nEp-OU0}IdfUYhtT8t$7U%r@kbU=CcR^MuIHujGx4Ygl
zBWoyJ<@lfMyZSW~v--j{ztw;3{gxY-1y(l?<#*~>*_;^oc-Qf3{iF4y94>s~`H$B9
z-TQfpllQ)$?<q^(%8SK2_kQc$m#-X`mx^zHc-_vU)*P2tDn2)Fr6WeonU*&-Hm4el
zT!MjMpg0EhSaRQPufOqOUTs)hc9o~K<KBOJ^s4pm&zcHP`QQ)7oqO_xyi{y#CXYCI
zo8wRT=J>o2p1aD>VefaJl*5{VTO6~_X*2T_k3a0#b#{4ec3vuOu;`ua@7-qPftAHW
zW7lC*`78AI64f%E{@vQPAK$UXUoLz(uMJbaPa1Vq|HIarSN7k*rE&ar<(W_Y^r3Ni
zsp#)vOU3ki7k6H^^V*wc)APo{PWN@+;Gn}s=e6OKL)V{p$IuyhA#Ccu1HSa>d-{L<
zig8(8F|Du0mP@Ula`pY)_cgs%oigLdu{nHX-)|1s>qqb9Df-=gXurR+<NHIN+;@1M
zqPkw`wsj9^Ke=C48#Xr2;V&2n27-ZLpm7;UpDW}-pTl}CrDz>o^4|9kUU2xByf!TK
zJ1W)sce?3wx7uX)kZQ$3bzAA<<=z`BZQUgo_grt8CqBrl6$|wv<Awfxs>bcw;Vl>l
z27-ZLAQ%V+f`P&eRO{=ghK{>1V=>bd40s<q@XP<MZgc0ODQYcz1_QxB%VJ=MjW#`f
zonb?>z%O_>?55e>F6;PK9vD2-M;)}RZW7tH7zVz7#1+p?o}IU9EKG=f{=fsK{A%}U
zd0^dmNN5eQ!Wn3P#;bEr+;2qQR9NBk6dhhueg7c+9&dfiMH3K;!9Xxj?+mPW*Ae%w
z`N^AEW8*1ZM$PKje|%nPxc1;xmkd7p**wL5`(HS%`(sn`z~XqQ??Nr~`!DJD694eR
z;d8Uz|H>N*>pQQo7Yqag4adOe^V-~V%WF^Od%gAOfNg(2JA?Dz+H>?1pAOH;#X|pX
zRJTrV-oM+6v+_b%*XKvt@0yF;a1{`igMn79`u(BfmM?U9Jr4{XT2dV}m$N_R&c`-c
zsq0&LePMIqCGrRcf`MS5z8LWLQ5`ysIkMOE_BAaP>x<9SPM0%IJO9l2Z)AtI9{u8B
z-&n4uSoGXW?G=(6`|k;x>hr1k_WR36JMI5!&5XQMEVK_X9t;En!9Xw&3<LwgK&3Kp
z(bdmAfA}w6%&Q#_`o{Ig-?8NfSzvYZP^sr6;tB?WfnXpQD35_rtMAbJ_9NcS0~>~i
z@^+>2R~7ra(dB0H;_rXc>*f_EWbl;@XPmIy&z{dy>_4&lj_dt+NFMm#c_?=uV%>jd
zpwRcu8f&~-e^0Vn<Noi}5VOHRFc1s`1LZRC(`_ftm_59GP2iV#*yR3oIv(}?@mXNQ
z^6=Tn9VcCM)1nNPTgT<D>)h>@jy;|pkXJgs`gpg^Ry|`%UM%i=;)A<Be%6@05SGhV
zb6t1mAII&r@R}#GO2OvJb5rHjP=6=1%fpu+eC1A4@@mJ!K3t>Mxm{-FaM)Vc4jwRg
za$fD|`<(EW75iR!<bfmeQnA$U!!?$EyB~h^{?NTY%7^yZbmQf=IH}RG(x02f##d@R
zREmCM|E^3^_3!W-3>3q_)qlP1;s@54o258u+OWTl*>rJMtvGU*r`J1n)!A98IC$IB
zj=Af-SF)Cdg=)p$!#5Nb@*Cs9Krrxs8A$(K=HVZoKQO=ek*rej|7H*)!9Xw&3<LvT
z#XzBdXE4SWcf6td`+4xtG(3D&t;SgMVc?{0JN)sqv!2cagNLYtI0yPbrFFP!{r9CK
z?&|*c!DIjGwttn@M<Sm8&OrKiEFpjO-}h8HFaLd)V>TEF27-ZLpxPN|sL$uC&2F{t
zW5gc}1Ovf9Fc1s`1I>qlGbVoPq+8!>UsD^Nf7|jOU32HVS*7EjdoBFe6@MO|1%5RT
z&8KoJGn4k;S+Mw>+m@{<6;FTm{C&=BvwTfp!|+g<)lz&^lb^^Rzw+6RHG#oHbF72n
zPeN!`A_F&HbI^vr`($8VX;_I|MNCc4z@<;$)Bo#NjLV9}19thtCttsNg_>%?re`=J
zXif}N`oA~&%kRB-%+$l*%jz31Ty*ld$?Yd)%?0}mx@P4eBc^77!9yj~L38SnkxTV3
zu*#LYPkrIw*;!!l&>ZSunHR5Jc<AVtvVQwtIIa6*Q}S~0^gr%7bk?|GSzz!Gbr9#E
zrS*Z!TQ7IckJ_(NQwV#Xx6)7hFMc}<tQ!y2b5m@ZNJpH0<z=^T->L>4yZZOXef^IM
zGU)fizV)YnpR@SJX|?{I*z&h-eg2r$KF9)thp2<5KL>}l9{u8B-&n4uQn1*6m#W(I
z(%&Tu`Hn6Aa^cH)rSi};PuuVs3{)orH*R>~`=7t~XkIQ>XVwvQFwigz?6~jtt3P|&
zC)u}8uG#Cmt4>*p#s2<$=#}^H@HEVBhSgvo7zhTMKLhjM+H>?1pAOGj3O4^8jSPc<
zU?3O>28v>!-Aaqj8vI(`s!aUnH#+Tb+Q5-{L*t^``uuI~Ij?7d!9&zR^`C>HdQ&Jh
zI0NTg)#}#4kB-VK9k=}1O*ig(^5c1>;<G!iKk%e2ev?%@Uhwm$&bj!yfqAvysMUAq
zeftsb=7sQqTjrhh;$wfystq0d`}gRv_mbsi<?y%-e|YU~GbZPO!9&zRoP+2C|Gf|B
zqb(iJ_(eADrMJf9^@aD}duFeH-TZM@skq!vZft+k4S7~9Hgp~{9`11d>US>80xO$`
z{YF0U&yT-7HLnd@dOwN$gMnZm7zhR$mVsSXdT!iSmru@%#m4$OgX>r5_4Lh~tyB~E
z)ja(6JHKCU`#JCCfz`=F>kUp`VfWWM)C2|(Q3r7jq7RhU2O9QvRK3=3+-d$wZ+vZf
z78pE49kkqYP`w>2A`b?FfncD17`XMxPxkrf*$?wlv3|UVm0%zk2nK?IU?3O>27-ZL
z;0p#C>+{s+;wS1R7$}8-tvhY7>T4&xmDd+8*Xou2Z>&2quQu$p&7Pf~*l1i{t@x+)
zez5bt+rE`oE4F&@$`jw}@j+g#IAzVp?%8{#33;{Rvj5y~&1d$UoClW1!(nr_8N1nu
zBl5ui&O<5Bq?W9E#a+{TjlE=I9vD27QU@*h%tY?h%Yg6S{A|h<!}eTZY!=uwJhVUK
z)ww6`HzIFp+-_Qzix2tdw7fPv_}BqAE<5stEHHRz%sQyvP8*R21HnKr5DWwZ!9Xw&
z3<LwgKrj#t1Ovf9Fz{stHhb>XH6A%?K~~@RWiKHWVjzF3-x2>f;O#uHJRS=18sou0
zV=^#ot!oDl7(6*I7yIph;k53LP05=I*O+{E@6Nk?kk=J!`#h(i=7(Id;#&JYGc#{$
zT;|1V7altLrR<b0qh@vNKR&M%9K7vm$K3VaD_L{F2UmOQ@$0vnm^U<*+C#px<NHIN
z+;@0hDjsm&Su^iBWK>?QSm@vDe({1y2R-w|!+E9P^sAQ)JN5NvGI;T@Wsf;)uQ&2y
zaq7N9c06#|F<D)))Vd1&V4xTVs<nSa+zs3RJB~VH?3-82%HfVTzHwORPIL0+ioblc
z)Bdm4%*g8-oBDnk{=c~A7t;&>qx0*1?w`5y_LK6Kg2nPuSG^6-oxjQL-yWAYHl}fZ
zcdOoO{rKt$S-Dv3Jw5dsaxf4K1Ovf9Fc1t>KLZ1Idw%(Ox6aH0OXFeNOXk)5cH>1^
zW8<3N>c94W%Z<w_1-<Vd_{HZREro~b?|hL#FwlG$c>cfxru=I6X?cBNwKH@3$D_6#
zH2k?NuxWT$^4|9kUU2xByjrojd|#ka^XPHm%^Qs0VR8<)A3J&Yl2)U#Qn9h>tK9D=
z)wPc{_1~+9|Hk><fV$R6Q?Ea1)K&ctTWem{SlF%8oA>Yb;;g*Bu&MbA|G_{o5DWwZ
z!9Xw&3<LwEF|hIm+w}i5pPmN>50zX8rB!vqYxh2HrJwd+{B~9>KD+b!15euGH(5);
zc0c^+{h@n*lvf&_)NO}9o_5yLc|&8vS73P_D)n~}%gZKKH5LPhT>k9t5A|D|1qKgM
z2i12D8teV#@bll_&u^@L9e%2Vfrk2?MRjB!Q3eCSKrj#t1OwH<z#k4@qvxuF-^nW-
zv02;iB=q}3$1Pvz@_JrZytPx?i=O(yysS2Sy4QoJ&2Rf|R$sW+TaOOd_V=^1x+1;;
zgMnZm7-*>sY<%IJ!!|s!O-<mJd02MN!GC>ax3O8J;5V)q_2jv~9h)~5w$$^H=Dp>b
zbC-NJZ*kU8nEDI3JO=i8aI5vs9{FxwZCGBmu__n{27-ZLpi~APf9=NUyIi{<Yb-1^
zzR(W_f`MQl7zhS}fnXpQ2nK?IhGU@18K<3p=KMFZrpAWLuxTwXf9uxgk6G=5tg*0Z
zvkmVpoq^8XZt2+L=>d7C`?_y%&|#x%#gwmw_e?lquWiQXm4fre-Pvd5L6h<n|F~p_
z_uBn_N>(bS`Gy<}1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1NF~9
zpF!8GJY>YwEXD7&IXe5(NuOthu=Vg6fBoyt)3ZW&O8><jZ|MGhUMx2DzZ=_F{)+wg
zWa&M!?>7hR^`rOl6br?h{?1)}{rhLruU<0j)YqTMsvW1?`Pe2ab$u(ZR{Z{-U;D{7
z)}5IZ!aFb9dF@TJ>3Ox`gR4FD`1RXO%qty-zx&#iGq#<Or?}1oj~sT&fQdP*|IQp1
zgMnb6QW==@^F#LCZ`}o1V0k=L>b^!?4Z}d8`)iEP-{kghkIO5Chlbhxu-Z}>=&;7{
zw)eCdoi!B}`+h^E>TNfz%f*NMb6Vb1IP=s+^S}4p8%zKFN&gOI`ueBX@3$7xOaJ{u
z?{};*`Rv}EcljXy<)fYUf3;@DQk-(=`V;RMI-?*j<_lp%z0PbY=Q{EX23i6Gr~h%!
zp|i#f%j$}U{=4JENA9>jFBe+^*O6n>Gf?PxS>NMLpXnD74DGY&#>;JSQr_I~l8bw;
zx6BhC<blCMQ`SMHobzgnW5S?wFZ}y+gL3%M(SO>1cDwOeQ{ji}95>+YU8m;tjs5n&
za9a1rrsNHcwY@L@-K~1B_2a82WP#0-hx)$16tY*T?;}NArG5XXq4=t{=TO8O3<Lwg
zKrqm>46JhePxf8?nu!^#>-&+`AK7N^wQKG3W*!(kRAL=8?Wqgz!9Xw&3<LwgK*KOF
zY^`et4;VZ-?|t4%KkdKx?F@dh#|>*=(syoF2pcAvuo?^m1HnKr5DYXA28w+jzj<U^
zeiq(Ws&CA$Jni?d?*3_Z(bdmAfA}w6%nRYD)pzK9`w{Qv%?0g5`uEt?BD<HCi}DW)
z27-ZLAQ)&F3{>0iPDH%b`FkPNwr?%t^hcJ#K*KWdbgu_bo8R``EU<ofsOxU|J6%{;
zd$B$k2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3Q%L<YV$=C^;EddIA+
zvGMkgM{PT3_;Xof<0)N6&Fa{Hd|p?)rq`-dW*j*-uU7Q$wbVCO%I`pz8b>9b&4?)&
zXsHa;{QUWi@_$Uo8w#iW^6is{Zm>=bT)*uGo9;2`wXD8y<qfv!|7kuwFBcE{aE)H)
zcA1$s7Pizni@YnFfg3kG@cqwUd^8WNE*>hosv^2zpkWx;X8tkX{QFPG=Yh?ehb8ZQ
z|KJ6OkI7paHq5CBtBu9L!h0qhvDY@^^M*#Rdv<yF@`JD3X-W>$*E8wAlha(}9&eBQ
zd!OEu@=C?Vs<-eH3<Lv}&%jXwe$egyC#U9t!9&!+|DJ=&Ke-WMFc1s`1J%jE!mqWv
zI3GPWug>Tr>R=!k2nK?IU?3O>27-ZLpk*>}&Q+~$9sKC1yrJ>->3t9AwEOtHRBV~M
z5?KcW!9Xw&3<LwgKrj#tl*>SU|2y*aZLi$CVqGv03<Lwsm4Qnx?z!GFPkfN4_~2?!
zJ%0Um6Z3T*c;v8C22?JVyD#m}cy;cH`;Eve4a?0Z)&&E>Krj#tl*ho;Km76So2@+|
zPjRK&);*y8<bGKpe0Jyc2cERWZ?Y7RzVFk!ezas{RtR@~am!yF{p-P5V`E+ay|TK-
z+jGZWdp$qz^{lCJ`>~UUFKIO@t2SJ-<{zt%=rbj+R=mIaZny4t+oZg)vAnJjs~Vbt
zyPp5W-}hK)Vpdn2e)W=Jr@sD7)>OFe8$Y^jiv{oJ<zhqg9QK2OU?3O>27-ZLAQ%V+
zf`MQl7zhS}fnXpQXs!&ry`W>KV+VbhRT^G0`rg0Yv)upx+&#g+{@|Uy51#)*Ug@~=
ztcSMSdbz$?wctOmKXu3-E*X~B6>nZ`#2X8yzm^A9ClAfF&qZFrKrm2!3_SkcU;l8_
z*vWZQWA!l?5eEaoKrrxS2GZZpEmi*P?0q|&)M;#9?Re<FJ5GG$j_dPk#e?5H^p!rx
zeU!Hpyl~OU<0iMClvN7O>(gU?&(kO8)rK_#w>V~<(`M%Ng<tM2Aq4}$Krj#tG!_H>
z_P=mi_s6E>Ed@)xuU5N$W7Sdk2?m0JU?3O>2AYzAdE@Tvv+|%xc~fDzxj+8w<h!3e
z?VUXE3m*JFLZ#v<T}I97*nfN;7(6s(9n{uu`cLe><9a_Hk~cQi7FRmI-d#uByXGfv
zX0>7O^H%z4|HW@-h49Wlj@xVDHBV%fivC@~hQfw^z1r!%?i(C**yy~maoAed4jwRg
za^BLh*Yx8rxci~ov%sd|VY_KvE<WU+)AHKz*ZbT*bLZ_R<#olhPlg-}1Ovf9Fc1s`
z1HnLZXJC(2em?QDuYH^qi-lenR@?ZUm+idvrrGp7@W1oW-0MH`3kHILrf1;evyQwZ
zf2>W-MkAlR>-qIQ%mTlfhy1O6NBrY}xAX3w^qu(3mE*Je!u~&bs9W3B=VbTaduFeH
z-TZOYVzu2vDka`OF4^I|cE6vJ)fGQF`cM1MZZ|#){O>&IxOM&A$Yox<cHyC;U&_kG
zrux0f@Y}S%_t{kZtNXxUAQ%V+>WhJWJMMC5P0zV`Q{ng89G(5?q|dWr@z@y`WT(D8
zEh~hj-k%P<?)~vA9XUTM7KirPbmQf=I4LiLE#>bOBG3By`xNI`{a<_bBfH#s+lN_w
z<J23EzTmOnPRbe^(|=zo<gfmBx2l~NbsQKB1Ovf9OJ|_c|NX=?zWUy;`uaYse-C(<
zhc7?)%AKa<#bSN=4?9i!{pQ>H-}=Yf+P#;hxb4SxZ1I;1U(Oo~|M0@$bF<$6%2O=%
z+@!BdZ~u7Iwu6R0m(_;rx7}dVJtn=DRU7uW@a7H1?=U&9D<0Wq?zL;}^Jd=EIQ-q$
zuAH&$ggh{Kh&rhMb5Lv_4ZXTCaN~CWv%c?neNX7wsp}S9x9r;bioS>T?b_x)|LxhK
z=juJz>e{ML-<}=+BY&%F>#nW4EjsnreS6+<RhM_NW$Na=1+AX*^GEJI`r(iA!18!#
zL8TI@1_Kqxz=BEh$E<YG?7Ud)FmUmRksT)FrQ#3I-SM7&V;;(j#kMa@JL<MYx8_ZS
z6}M3lP%sb-1OqLcfpYKJEu6MUKNtuGf`MQl7zhS}fnXpQ2nK?I7SBMruP0mTx)xtB
zK|nAN3<Lue&cNrxyIehG?df@7%i<v|%To8hTJALqHofqcyH1*!R~n}Lha3z911*<<
zQu}YE>$lvUjLd_9U?3O>27-ZLAQ%V+f`MS5sTlb9z>XVTeD#NUVDQkCb<kAx-%|a~
zTlxAEr~PG49@sp2C{)+yY}{kBcXpVYH+TF=*U$ex@aTDYrDLIeit%8eJ{ef}`PkK-
z-D5^xSFE(p2_v4${QS1Q_NBQ$ry451`s^KHtL_ZU-r?Ibo_}d>UM?=Vzh9db=1tA3
z73<DmSO^A!fnXpQD20K3H(dA5;oUpb1O^X{SqG)mXy`V72F_df%pIStIwvm{zju4*
zU-#Ye-&tL8?4S18@{p}x%u2=8e${hcx0heeTPimHii`}KGXulg-8Fy2&*$XDVsqvq
zatj86fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+
zYGYvAgN_+<!|NYsm4eHR8G6dT%Pv#X)L0vDi%*_*$9uyDWQDM{^)Vj|1Ovf9Fc1s`
z1HnKr5DWwZ!9Xw&3<LwgKrj#t1Os(vVA&(jn%C`~DS4^b_sM+gzMo9WipBj$@3hUx
zp>O0V);)u;5DWwZ!9Xz35DZkxzcbSixrNPOAQ)&E209E}JYr;r33*^;^UyH6U-?%5
zyv<H~zrEq2tTw!G>)q!bbHJ>uSUhanlj9Hi{m3lv3mz)J5+cH2pz#^_=+1+-+2!q7
zd3|ABfA>0F-}sD#$6z2Bs1ydO<=^*iE^$`MzDFFvKrj#tG&csW*|G1lTb(vFt8d)%
z{gXRfdfwQqRJ>#DgF0-`V_a4$UiOck-JcshBZJLtpBg%!`g*+;cFKGG+R*zO_Je_7
zAQ%V+f`MQl7zhTclY#Gc8hY5E`{!kWU+}PC()=+iT{Jr{7fY>zhWdB-*1LI+wJ%w+
zI4cz&K5@%CdY?EyD;96M;hX&iy)iK>7K?rTes{aup8E8&g;|QF@*DcWKrj#tV4%<Q
z^Zxv!v)j}Zi#smf<kfdBpPLnn8|}T^3eP>VC@U7zztd<expnUqFWhq22YI<TWRr)c
zytV1jyja}thMzvROq<p<DcYC7U?3O>27-ZLAQ%V+f`MQl7zhUHn}O`2`wr}O<lA|D
z<G>fc(e{9+p35r*ue|+->s{Ejbxm`@4<3B^-IqRpDyvp3_I=NC^~!zUVA^MIwI6fi
zGg-NK?A`}FGiu?RS#8+%g=t6Kw&>QpT)e8sHV3TuyJ=amSnT`PrRo*yzmQ&Cf8V&&
z`tP*cs?Sf@{hq8?tnYps_JV<6AQ%V+f`MQl7zhSh1_MW2`I}!%-E>wSSQ-y4qdFqX
zU?3O>27-ZLAQ%V+f`MS5x*6F1>IEmvX*V^m6dd%kZ{5-6XCLI1f~#(E_HUQ%Ju@#B
z)8EOeHhIwKRess`yR-Aars3h76DLnPbN!F<a<RI*QAA%X182T7Z0~=)|4ttG)jSMs
z+wa0}o;D{h6@RtnzuUImVfmU;vDkeJy<i|12nK?IU?3O>27-acWuV$#k2iI^jk`Pb
z@9nDVe%|BnJ!a)C4eOuDuow)qJO;*I_Op|}`_Q<&HoW1%=ZAkVa%>*hxIDDHifr7B
zk6n52_=5(_$O9{jhsLdf@D>aN1HnKr5DWwZ!9Xw&3<LwgKrj#t1Os2iz$G`mv&XtU
zrsjdcL)1Z>gXjaL^?|SIsWBD|1Ovf9Fc1t>I|GyZJ-kZ4x2EPT1*?6(BK}|?7zhS}
zfnXpQ2nK?IhGF2BQ&&Ckt-npr%f*I?Dy#+r!9Xw&3^W%8&i(a`JFL<7?YyqoXRkXu
zAGhX|yi}aC%#wZnxyz8eQt^&eTX)%M=D+ghf@AjD{*#GMjmYbY&87Y$k6@ra7-;Ih
zPcGGex&Ll(rLAi}_LiduZa*=vHtaBP@raQfCgjzMjm39;oXxNi3>3@2_S>%U!cSHm
zpOuQm+7G>8AQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z2nJFH7HoRqEq9$XGY<?Nq7LF5L?0-x52SrH<X|8e2nK?IV4xTV8tZ$o;is~`=Uq%~
zgibIJ3<LwgKrj#t1Ovf9Fc1s`1HnMkF)-q`BX=A5=&U?2c=&I1&~z0Vo*RRKRkt|%
zx6AgPnFj_BQ3nlo4jSVuhmT+&7zhS}fnXpQ2nHI0fun}L|8T3G6Y{{|A?l#;IcSKp
z7dC@|U?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>2AZCMN2eaK{Pa0<vSRVx
zu}ijJ@2yX>YQ>vw_-4OBZ%oXp6_4HffM-T6d^0N-51aPn_=A2wGAkGPZTiZM2pWQc
z&xd!pddk|<^T6PtKI)($YBX$?&%pkpciLv;&^PkH;GsFzLHQ>mRt5vr!$7CqR(*cL
z?)PLZ6&q`(n(E(?Zmj&mPcYCD7`U<Zudd6#{#ahA*b??2atsE7fu?0(_72~k@%&43
z^T5)0Xxd#1@4-N2G0<nPJ3Ak@=9IitocL0Y9$gQ3Bd>3)EOsKAU?3O>27-ZLAQ%V+
zf`MQl7-$Rzjy`MZ{)4|UBC8D>BbM+{I|IL(*tXM(ZC0r%759E<gGKND`NgbMyms^o
zT^=3$X;v&Aci*LNuKB}HvWCWv$3A}hA?q$vQ!4)M{1M;z)lnm|V)49br!0GZ>vlE8
z;%)CP`{ZfQf0~tw1D?O^JFWk;LQT0?>i2X1^vNSH-?MA0nqu*q-`zTC&h8y*O2sSJ
zTlU^dKbn+R3cmcu4%hZ-KPjtkto`JMMss4|J8NI?T#pau<c*EZiNnYx7zhS}fnXpQ
z2nHILfmMI?$`6)&eYKiWan!(hyH6SO@2oc5b-&L)-|US=S&EIDWq1n)f`MQl7zhR$
zmw{*AUE$-s_ne&v){Te8-L>!*473aeR&2M!r0cs+%3BJyj9rZ^gMnZm7zhS}fnXpQ
z2nK?IU?3Q%I|K7pzW&5%f0>iFRIEFL<yrXnz>XVTeD#NUxp@B>*M0Np!(YxT4gD@=
zePekn$EskUei&Hx$g}2kyJt#XD%Owpuo4Ue1HnKr5DYX7104n~9x<}RguGbHF1qi)
zZb!bIHxzb0@6cWs?e<<?2<!WPS?TspnzzP)4gUUmRxUP7)r8exAQ%V+f`MQl7zhS}
zfl?V5_~JL(9`Mw2d8xS1lkLAZezP_;#bTxY-fw-yRU7Z?-1qR1?T&vvFBUIcY3Lol
z+<15nSHI@?Zl6sXmzRpQ#Z~V7KK*ytcJoUv%Nhzx&69q)`NX>N7-;Cv?aGTPRs{of
zV_@gk#~-u&-S6a0jcFF=Y}{kBcXpVY2i6Y{DZdRZe{|<T+wAi8tURzd9vXT-!hSFi
z476MZE*-P%#6P$CFslu}yYaebJ~MlCUMiOQx|;riO)tFVu9Ifw<zl&?(^TrZmRq5b
zc`y(RR0{*=EqvyV&sLq2mx^<K{>Z&YKm1W%SF9GEBF<nS7-;?sESNNZ%t{x{&I5yo
zsDn5M(Fe-w1I_<Q5g7&p!9Xz3vKeUT?^CefvRRJotBrxppTGJyJ74*JUMv>+yPX*S
z?|;`+ZFLdx1_NKkz{1bRuJ-I6GxBQ1CHMDhv%<Wo`B&vu&10qhenD;eg+4F&e0Z0u
zr>s3aFBSJ6z0)=$hrW@Qi=}>FaL6VPO?hk6p?SGjYMiD1o#~M`U;bv_z2C}8Mc-p+
zYAp40xVq|B^Vy8pgMnZm7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z_$mf|u*TX$o_J(Z-q?8Eot;;k)@yuT8=kcB*4vD1_jq0|uC_^^AHMqL@Vr>u>h_P8
zIr@aLd8zo*hqnFa?8ipt<>Fzdbm-Lf^QZH2@zUvwcROgqXR>1Pl*>C_wcyr&Wz7Y*
zc<Zvslkb_FRV()V_Jh0jnmjKn7JIMI?%vk#&CN>1LoXdSW20YA&x*xm4t?g7D;AB<
zYQu9rIOLOk9vYi9G=5bNi?QlwVA^MIwI6fiGg*pfbzEh@25(Hti^c7)UU0&kc2o0m
z@sgX~*<;-vQ}a@>$4~CLYwxd*%M0NKv(Nw5_d7q37mHi&-eb%f`I}j>`29tTu5a^f
z|GZTE%i#6<4f=RwUI;%r@wY!&ZtN3zsrc%io4>pJK{w~cVr~B}(a2WYUjEY|4`k)y
z;qA}dZhpIQS+RKUv_tM2vCEXKSX}k9%`fQj@RY2va9H12KkB{eYguiWo~L5v2l53k
z_gwp>tXM3zPOtdGb6fWvF(j)EulnqvUq5o{-C41C>)+bHd)D>;&d$H~tnSYZexYQn
z{tg-$)P;fFFTe5fKFiO_TPoJZ-L_o^eSGKF-pxzJKleTJ@YODwl{Yj#yZ-O`kJ)@?
zUMW}`pD`Z{R1yQNZ}|Qu>rYsmm5T!=&VIS)t{-N_;%(oXanqtc<Fit6>jyqKX8GQe
zvZlgH+Ovoy7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>
z27-b5WMJIf&p-IqIdA5bigzB<{+#{#49JVclYhA1Ge@p8KPwm49lhz9&)olXRxJK)
z$iR(0S#xq$E}nDZ<Vk0)|508pKGJQ%u;>2TrlzUz-LB&<yZ^M8@^bMP-#uotAANsB
zUM_BU<%h>Cf83C)p>X4af7I)ub*5#7us+XX*lNBEl>2@5SXZ~d2MGSk<FEOi#K<fd
z2nK?IU?3O>27-ZLAQ%V+f`MQl7^r>*>iYMu2W~NV-*sM`m(>?u)nl6jR{Y(xtWxo)
zq3=K3s^^3}uwi+q{*x6M1Ovf9Fi;r`Y}NLag;!lLBM%H78nX^Y+;-${BOjfWmx{Ih
zUQnU=$~bKiMKBNyR5Amrbld3t`#v3!*Ed#j&JkNM5DWwZ!9Xw&3<LwgKrj#t1Ovf9
zFc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr&{Pccow;0>onD%rHx+(9
zyvx;7)}Ed>H7?ln!dvb-X=dI~xa9tRZC02!H7^%C3|u^7WQPfPvDo>%LwjAc+k1Ik
zaqOS=+47LBU(AZdn{N1Kzd>(I%u2;O);_4i20g}Q#p0gtpWNZn^TuYy;<g7JGv<cZ
zKhBE9QvW@z;avvr_rkNUX7z=|=2hu>P1V=JZ!i!H1Ovf9Fc1s`1HnKr5Dc_j27b27
ze#cz*$;`ac@#j;IS^dNHZ_7%>?4tV)>~`ead8OcCr*!Dl_VcIn`o@-9!I60|5DWwZ
zUuNKikLPaw?`3A>O^siUDx_c_7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}
zf!Y}8ztz519dyauyrp4nyu^ItGEnZniyZ6f_U}qJ6n~9dJ>e}F2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?I=Fh;O(X0Hj?{{bCfx$!6L7ao=1LgIB)vr0e+h^0p
z<>lg$H(&l{-@V_;3gOImCaw1DdLy$+$2CV@u-OwkkIe$V%tQ0PNkoRhKrm2g44i-M
zS>2x-{6f~;@uL%e`;+CyK9ScId+oNv0V|&ITviD8>A%Can_qHSRvSLP?Z{Jq-uCZ#
zbHN|1vG$NB9+{K}1`m}~2bI=uBc5O&7zhS}fnXpQXgCJ8nX%ccKRWmItTr68*Y=-G
zd}>5qEFSZ(^Uq!K^!s_S*lOKx_gOq}bY2h}j)Jh<j2XD@&d#e%>oq=aZrF^sEK&;w
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLpduN#al8KstM7SzPw3gH>y}-&>e~8>zK8bh
z+U7t1?b)H{>OI%$+Nw|Ao*n-qf2(WjuC2Q*I`!9md){$Xmv^#dTAkZ*^*#6Oy<yGx
z0T14E;__dwd2g2;@2P3uvF6xOYo54Y*R^Xdet%lGHAk;g^Sy3^PQB~V^_J?9|9s+2
zTb^*#hBb#B_~~m`oxM@bn-8tK=B3}+u;!%e_IdeN&wZn&%@e&Z{r1i4*ZgJUUp_p0
z%K9~b`r6t1-nZ4W`Br-`|Hkt@2jp+;wP@tiFTa<;3B%g|q{FB4@)Ren_~Mie&tH%~
zanDvu-udLie1&~)e&yL!j?EuC{p>4VxL`*9!;A0ked{VC^4^bi`R-W*p2_MP*@urj
zl5Kdyl8H;>IOFT*b-#T4#H?7f9`5$sjJ^v_pOc-p>hZh0JnF-Y{6wc!Mt!*D?4{#-
zw7IAK^3&#Jdp-8fxx1aXDC^Q|{m-|U{OM9OpRPM+_}x!G+I>d$aJ!W*S$mTXH4CpA
zy~W54SF173-%}^Pbnue<<}US@-+B3qtA1;K2I=jy;D#C9#xKsE`{AI@d;j_KrT+hV
z%RT)c>fEZvJQSz6*_+h!rVihItyRs`3GJ60@k-m8XVzUbc)LwL%T8{0-YdJU(x&Ex
zzGvTa#D+_<vySXM{Kn&^Wa6bKPkGm5CtSYHsSC2repWO2rJ2jrZ2pf$M~^vuxtjYA
z9<}zQ``Rsy%X)mv<EJ-G-~XQXrY_w_^Hc6V<cht!EweN}dCCj&&z^kM_pp~QpL)l?
z-dSqjI6tX>^+fOL2RpAZ;DKS;O9!vAu-BWzGxn@67M-KPTMz0mXwlTX+Oe+PSwGTO
zoO80i_3_nzyZ_){JhHSN(>`OL?2r9Q$L*W_W!E_W{OS|^Ry?-TtWEy4w10MAxZXKW
ztvD(_?%eB-AGh4trTy>p_aB{j&WYnQ^;B1WVtvyxpbpcz(vQ`zdQ?BwsUvj)&8yFW
zvp>7}ufLj-t4DnywMSn((0Gc)vN!I!Lq1u2!JHhbd;QA2z&yGC`Ip~&>80ayamd5G
zb0Q!4vyTUQ?2!4#1KIh;Z)%?m`6Jsu_Ut2+7k#pL({;|hdHn*)8>ZvplaGAyAz$`A
zcRh#Yqpx{xTTgb+D1@G4p3{(BeT_YG;1~CorslzYNIgLJp|o!FEpZQhFRR(O-zS;+
zpigGU{m?y9-MELTOUS-@<kvZH?wkX4=N_xx_;((hCw-G2n8w8(Ufd`7bAQC6x_9pM
z0e#`o-(0@aONTEr#Ot9S_ZYdtv7MLX@^G)P-uZ^+<s%RLK)lJzxYuXLckcYia*xej
z`npVi5=ScYXCC5H{~_!9P(JJ$FNB^$o(tkm<>Tg_yX*X2W-ommN)LLzh|fNFUXY>q
z;E88U-(u|@=l>u3_~R$l6Q}XokU#sZZtx}#yo(>-#?=*j5YJ@#WbsYe_p)QZ^YN$o
zsZ)BayX$tJEya`XyZg_Z5BwyjCvTYYVVr&%zx}Ni)EB(2B|B%%z4PhZL4D1B>SOBC
zzS~!_eS&zj-u+D<AeYOle%H`=?;0E2_2!msm%i@zKESDab>8#9YnH9iPrWZtZa%Rt
z7zhShwQ`=_WAvjl7xlQm_j2uP+?Vt>_c;AaUm??Xuc!_6InOJoufcSjtRL#1`mTOU
z_WbiacRwKOpX?du-*ek|`W(gsfB5t~@Vd@@i7amUq<rvW9er`g1L~LNp**ZF7UPUm
zCIe}`s9WdI>yQ-PN7SqHsUEz3Qg7znBh;-rH=iOq)}{00LUxQp@sQK?=H&(Ho3{?q
zgU0DwpXP0xekvQ6zkby3?GrX1`e2(HeZe~Wm!f@v?9nsMAJh-@J@;*W%so6sdFqGk
zL;Xo#(VxV{zr3u2cwTY#Z^le+vqjBkXKZxJCA)53(|+rH5Bubjt(Nk_KD+pH54Jz*
zU4JuPZRmBF{gEdgjC);1-##1ndW%f2u~5AFlRVT#ePd;I=fY8`SLi<Fyf`2FwDagb
zVI8FJeCWIEvD4hJFvDLxGxhR4-rn<^MN2;?P0x+j&HAN&qTfOFte>TQ*m#Qe#dx}|
zHrcvj(fc%o^yQKIEu>%UxV{7R9Wulx)GzdpLVEUvU1;Cf@%%Ece|o>l`&rO)kbQ`!
z`o^!W(*GUhUmkezyj8FGqpvU6SExVe3wUDJ`qUmj^zclkZ@(bkoiqN(o>TG@m;5~U
z^)2Vey@_A@!fzUvx}`5qy!alA`+_*!n?0}S>Bssm8Sm`yo9fvI`H}fC&JNzh3+;>7
z<*7gN|5ntO{|}G&fc%OJzhwKMpTg-w7k2x>?8QsJ-z9GO`@F?Iup>_P<!2lp^v&ao
zULmB<Zz>xP{Nnp}cyKTAIi7v6uQ06-{Mw%s?UQq(KCELGI{)g``%BI#d+LX5eIXRL
zeAD`32RawdIW#U$`tre>aqE1(i3k4Ec%4gqEk*O}vuFSG2RxRFU+x-pmwZktPWy&G
z_29lMFZ&L?Z^@qT8|sVr$Mb2IPJjHk76149gY?bAO?T>Z=A&n|u3^{b)xJMSrtk9#
za{BoMJNBLIT=^a+KJ1U@k2=Q(J9r{{J~=1y@Hw<N`LT{Zp80bwjHmmgUd@X~{7@ZE
z|Hg~w_uOy!8vNl&eR|(nT*jR*>x^5Ecl?MWo%j6${M-LZL-$noC;Np5yz5_hmbd+|
zU*??$deA)n>_47ozrFCi+n(uAbM$RT&e(C-S~c|f5f^*L*&&OY?7Zub`dX@Io*%OH
z?5S7t{NpLr!-sxr-gt`YSiLznX`Q*}IFHbJ-*=&BJ-+4TKFbf@+0h5>8$IVlob1wb
z&iJL5%K8s`{IWxKUinYwjq{t@V}~Al>WK{HV}DZ~_@|Gz)Sh$2uQ-13<c*{E8rWut
zdNR%q8M2$Aee`^GPu_CYpb;~tF2D49?BZpQAAEUEh|_+lAAGPQZhDY?e6b6iJL{bX
zJoArNe#OauTJP4Gw_d%;0~+^yhw3rShkfJh%TwN|f9uo%ed|-ypVEECkNqaYRNwxJ
z2an=6jz_XQ@Z%ie1y2y4Wa}V%5Fg@T-#DINnirWpd8O;|DsK6?ucSy%Jm%eJ<mX)N
za^y=_b|11@&EH=>X!&;i+SlOWqg(bk<u`k;y!3k|>3zgKN&fP*ub!v!p$F6b5;s3S
z_i+#HeA?E}c9^_!jeH*XZQtwWZ?<BMJjm9`*XO(LGxB$@k+0A1J#XA|{GATF=B*cx
zxb3UIzY!NXjf-7z!qksC#2;B4{A|6|IS2gvl$Dn5qxj@+-|%1@&uPBw{rI1Ef9LEg
zSE%8KU3TOrE^$KjsQ($qKbbvx&Ii<g<)Lnz6YG87$v&#LLj0%u;Pnf8c#<F9@c^Aq
z{^Vty_#r!Fh%foEFCOuVlT43Y`s}f99RK))b${jBFc$EuueiUa_fOB|LibhAYwNuZ
zrVmp)#_97*57JAKKELie^t_)7>ATOf@1CyTmWtN%&mZ)@H9h0>J>SXZ-NU_aO)o|J
z=lL#w_Mm%#=L<XZ-9zw+FYD=(y?>4`^X$o&K0Wm!Ugw$&^?x!f_B7CEM}GF-b3uP`
zew}~k9$(_dqkPDazP#u`{R8r=uhoXd=AY{I+M)f+r}l3*WYB>x4H^7)t0DHq>jS*V
zW5*Bs^%}BPn<3|4Jo<#~c5F4o`vYF5*k^h9Jj=S&f5?@=KzbfL$L+hmW54x*6rCsM
z-g=*B8i)G4*L&*5=c4+abL-ygzCzaT)UQ6SPm=W??`N5ppL^EKqc8jDf)`e-NqKQT
z#RbLTyyze91MV5_U0#pKOFr&dp2PUVvwlO?Z~1rr$l{kD-uVms;`a|Kg$Lb0UD*G$
zPVHyl7r)<ux4?#BKtD?Rk2-a~(Z}5v^&!s%=gd8b4Eb^2G48pb@9{&1`aHQ1%7dS@
zPwUt2VP5CD2a1Os_d=-e;z7K|@#1-59FOAGuTymHJ;%(G^$X+tv+KEmFZ29+j_^aa
zp1$=d;>CJ#6q3blo<GlX;|){Y<y*x^ntxNv>eTx*-gi=;-j8zMftc{zsV`IyUiaEp
zbwMtK>H*SI$Lz9~BK_LXI(q8PJ<qv!Z?(_r0OHL&ednAV{ei5XS`YOJc2XYrN#m!7
ze|hOY>_hYJ$=1uOv9XYU`7~BMb@gX|ijCJk)3>kq5D&d{zw9GS*E=uiJU!=%KlRU_
zb;ijldQKRp&n{k^3+IhKe^B0dVNbmD@B!)bm&#tRiI<)@_z}1DP~5&pY8~104W`d$
zyr=s@AMyjO6Cas<<EdRd%2!_aBGZF-k+*SZJVpI2U2ngv*N4=DzGuDhRL|>X_e}aQ
z)i<vHnioeJXPQqMmwS`vr{}ECHN2k&`Ge{uMScs7n=gbuUs6A%qI*PRqy15z_Qg1~
ze`K#0ypQX182jLT7oWp;-^l9@vgfdR^1hANE%el#IK*i`)RA+g-r0A~oKy4af_-t5
z#Ul=J@<R{e#rd%gzvAIvJo<@#V_bZY9yHD$9{6)!_{E?2txu6Ye%W^(ofC0*E|}Mk
z@lIcSg=F^d!@qTfWcjA^^0Xd`i$5qXeLO{Z(og-dt3K_2it0r@;UCZVw~jufpCbO*
zfraj&*3(PTI^*gojhh|gslM0$&b2)B1ARn4H|{**0qQsGd0)Z(f_>xsd%wXs6|enL
zM|kv{Sv+D|pP$ZNen_>TKCM30t@AHm^JID3Pj<*D+GpeH1OMjj8?-NE$c}idcdndY
z=hS}jR~uR{Z~F^VKlJb-&a^JYt#65+ztnE3XFdDSxP61{lOaCHsU8`!PZqDd$xz(<
z*av-<tUp>Oe}4E)QQYi6e$Cqt>(z<+qh}pE;*eh<bbs?4v)=PSexB3pr1=(-ji)F+
z=apZ`pLJyGjHj}3bxx+oUz&&U^j?zYFOF0nU-*Q6mmt+=$2xgK@!D_uZ@<kO*T3`y
z`H`XLfcSj=D(_-ZeV2>!Dt0|ypzr_ro}up*dM?s~__m+WJj5%xHdJ5G_Y+S0=GU)3
zp#SH&bLn$lpMyi+A9Np35BA4?q~qeUf3>0WP0#Cj^`oDt6ZNG3*(c-bNZ#sAed=50
zjq_(7KkTRT?q$YP**JUR6({@lwU8h6$UZrZo6H}3^5+lIV^@555f{7St_{_LI#ws@
z!nnQzjl)zw9Ty+HG;Z_8q32|wd3ty&g!I@WL-rwl*`t@{MaHB2@W@{}Zr=Gx&ndg+
z>C<x#$?RK4hVqh!yy&OsxmPN(&o5rYsh;`Ak8$gveK1eoyn19G@?QwkeA!_aU)J+W
zw%#~>vi!uKo^So#_mA`m&w2LJ=S?~<UZ^jp=RWl>K8QD{zS!fJ4C&)Vyy8jo*T=+1
z_PoG@ehl?JNDtC0^ge)j$bW5!ANlZaf9$vX3!!>(9vU0<5Alg#Uh;(Mk{)CST8~%z
zZCstGpHklsz_)$2UOlt#{C=|jqZ{mYe(R<0JEV2&_ZP%NmJfTzt(T8=Wbr`zN{`Gg
zJ*W>jNA&TnZv4(QJ?o);6&KlhGJR;D#eqNZrr)P!4?jLP@H;xj?T2_$<kz@($Pgdt
zynQj=(CA#KZ~FuB>p7(lrsL`skB}aecRG(Z{L&YPe&oD}$2?x;LpBf7_{r*oAO6i-
z&puha`h@lJQ)f@MU9{QEgO?wo&gJVKBwpvixvg#g=|T0A+GXE8S|0X|tPYG9dXC69
zwZ{&BWOnUaI#16yeP~`>g;0H%7azN1dC@b^4?W}d)wt(7JM`2AWZ&}}sxRy8pL$|f
z9kD0AbX=VpXAj@x)GvMG^w~3>qPU^>?3eY>xmW-EipzM4*3mc5kNMQUJn#kOA#QqP
z`^+BuX`Jkc$9#(B>6_;dKjIK4R0s5pL-hs4%^u$5K^BL&tTQfdNRMCl7qa;5OFEv$
zN8dQ!#ckes5Es2P4>CUG1I5FiIH7syeA72>Ka9gdcI2UNh*LgLoawxA{L3#z^ZXTy
z>FZ#8dyd$D_a@IR&rA0tayn0TZzQ`{rS?-<JP=?0UWL6>rU%(g(dUakKlDDI=bP7W
z`nG-axewVs;a$C`BXv^<*-7Wq`tp1e4>?8WL|w2WKI8m}A74-&;=>DMmkik@LvgaJ
zkC=zndtcb+wARZ%eV^6e#reJl{^|Mr*XL^Nr*+_SNpYwPdg4suF>f6|>9}~*m;PZL
zee)@%=T<+^Kb%)}${yqg+DB-Aibd<#Q!nx|59vYU_RF|^#<O_XSI>~06!GPp`~2VM
z3cjCI8`=l`$hqO4%zqlMxbb4%`V`GO&+HrLC$&qTp69A@e%LckHm{yi^!mcRB3;L?
zasI^%#Z9*F#@R7XPWdxl2=zPhq$ppQ+DZAbK8=gaPP&flyo!rHWZ!%tv@fZ?acF%i
zTbKGJr)VExI-lC(SH9`I^NlzAp33PsyY@+5>XE-RK7F92qyB|A`+`?kXdJKf(tTjJ
zTG4aT^IM&I4tf5oGxh4Z>z?UeTpQ{O?9j(=L!o^vw9ove@!4<Z5-;vao|7rMkI=`f
z`%t>xz0W%O?sx8Q(7y3&9sBH=cOP^wNbRr(t)~zD-69#%=Z9R#kNFhYXUF$A=((@5
zYaUPZ#0A-P?{P1pk3ak6`3#){NYA<9M_laTlWd*(Pmw?9o(TCb7x`1y&P9r~%{wRQ
zI_Jy0I_6j0@-faI*?DtL@WvlI(CaI_u}7aQuJm3dPUnYT`S1hv3$GvbK{9lIB&X>4
z#1Hg*=STlFFAw<{hvu!LXPjU2sUPE>YwTJFtus$wo@AJg%hNc%=(7v?lMl4Ng=BeK
z4^z9TOiw(prQfR>Du3rtKZp9L^R3@H_xiB@>-;*e`Um^aIAqT{{`4#Jo~zdB8{#z1
z52TmoW!`gwe<&Yz*)b03vq!d0KJu`REHC!0rw1Eq|4X&&eAb4}wZ2A9QJ>I93ZZ`E
zIpVoxyb$8w^Ns9#p#C0C9YOUC?XP`PpY|)YZ{NJ06}R)k9zAx|qkEEjlX3C6m!)Vw
z`NIp|@tcl&zCilWI><hmJ+itMci`PmKiYjp7CbZ!56(r3>Q+Bd@5YNobzi9{ze4@S
zdG#F9kMt>hQ2!!B&pUka<KCd}n%6I_(?|K&U&&r~LUDNA%8$OpUm?T?KjKQ!`t&(g
zXrI|l?cm8hM1Pkbp6!D^57WB!evr>o-IJks^>g?AV$t)RK7U>>c>R?keSXDb9O5s{
ztEpvisZaZ;4%Df-HBN^1)BZY7>^UFwAbs=h59*44_ROpM)W1B~<%eFmXkXO{-rVQp
z3H4R;cr!14GPI7L6w`V3>4~$c>#Di<*9U4teW|wZ8Tvh0-{YX4_8I$ZKgsmepLz8o
zPJM?gKKl9<yT1RCuCty${@J7NeHM1;sZ0IYI`+ivxnw*&r|Mk1X&&m$`G;x$DAW)1
zAM@6w=o~}y`V%xxPd_VGmcQqw^`5K7p?iWnJV&j2^1CNI`LkcGRiiI@?z*3g*FLZV
z*)h+)_4sDjbDv&{;^3Ekd7C%x_v`(hp?z`>^7_{NV~@{x_0UUKt>HJ-KXtFpdmebr
zvNiPmI}P~tI@9m3h|B)kM|q1Anupelqc+4#itHKZ&pC2_jMs+ju<N;$u4A7bnVxYn
z^t=|A^Gk-#w{u<So~55a{YQTxvqR53|0(iM&pODSdHoY#@}O^?e(H~GUCI-FjpGyA
zH>jSRSF-cy9FS9VUZHcqj{cIObLxGY)UMYn=F{I9aekpVz24DZ$o%6+y!6B^E_|hE
zy>a_!ytaMwyn*Syt3%_SkH(>S{?wUqb*WB_(^F^0>626B*K^%DBd6d0G@hd8yyrPP
z<U)SLW8OOG?=~6=ojdW9)40gceMa0V+7J6oUtXzxI*vDpZ+gbbFx69!P(4HaL_Mj0
zc^ap$E}?nodw<a9;Lg2sZQSn?LC;yAn|scu^X@tNgLBW1zCechjs9U=T>7VR&oSuv
zuD{?#U-VqZqx*q4^%e1w*<qI-{mXbN^Q*s!lV9`3$({@Pvv}FT1AowcFI{h5p3pww
zn;t*-f@z-OGtRF4N@e=cb4DEMlB^ExH#Bdb#7*WG;<pgmU+8?<e|^V3I8V?#wBGsU
z*S@gBpLJyO_8I?Ve%ZCo-zDg0^wfp*>{(AXZybt;J@rrDyuPD<%8S0bPkAt}j`Rn0
zWPjBuR7a^@`J{C$&J^)yTs-NxdHuRnv|e4|LEg?kJ@LCYnAeBpNnii8Z`M06jg6)9
zQ*HWbeYj7kf9Tvm`w6WhyPsIEpFsBsdhQj@pZikk&v_Jgiu7wk{+&PPQC+J){o4IN
zo$E)|m%H!sfzAQ{{6P7V({pZ|9jMQl$B%h-=%qa9PxPF7`Rj9d5<mNp9dV?x{vc0r
zii;gGKlH`zclsL|`4<;C<;OVG*TkEyrzbD<rB0yr_JzLn=E?R8e`!B8?|EfD9k<^7
z#JSMV?PI#0J~WSy)DKzyfx*Ck7)alDaDMa)^PVgEJ=t?9Mb9ti%XwngeVFX|Wu5Wb
zkX?Q}7uoY1G)|^3Kl9Exq*ohS@7&TCKN;d#zai@f?C9_E)PJD<NQUC)AM!(n_Cq}U
znJ2SLHgCOgC>}C>aTtg8O+C2}8HedU;+6dr)A-C=2aStIoc0CUAE-}T2koDA^sGz!
zJw0eYp!NE>=L<XZ?2~-b{L{LpXC43ckBpCW+`Msm5D#Q}&W(NcyO_J4y8amxS8F#!
z9oTPv*ijGip{I|eeM%hGd0rY%dGP)nJN7kwT`eAY;9Fe$xu-$?%!>mr&bNJG*Xw)t
zdwKIOKJ(DL^*e8I!pCpiu<X*`<q-!=;}@^^og4M69>r}v#FPCe<5iyK8yn@5)}7}b
z{_vyDj9Z5%dD%DP*4t<N@J|o={VjUN*~cS4kUm~heKLLh1Mk+cZ=ch7cGC6q+)wo>
zuLr$Pm&#rTd4GeRai4cMmt^Nze-xLvjeB27JnXwqJBMU-Ex%M|$GH+WlplM}2Y%HN
zzxo_~{OSM3@zK(6v(6D7okQo+xk}MFbe^08_N<4V2lVN&LpBeM^RJ&cSM=0_x)qOe
zWId#3Jw0}ve|5#4?=g8USf9$`Hcy7e`8AGzdD<5|rhKI`J5V0{$)60<@syuJdhGKj
zAM4o_r#$3ooPGR5`-*?|p!MVw)6YZnIjFzs58~%v-y(|>(ogHeynaMa|ANNd3&f*O
z8K*C9C~oW6!<)E`%LkwK!Mt_+@}uv|$GCn@_B;>|ee3Cap3oz+<GEm79^xZI`PhH;
zs6Nwibwkg*x;0)%kG_4j4!T#e<2;v(X&z*GL;ToJ`z<c<Li}0hc}O;&t|!yupIwM2
z@sd+yk6vmo)weIk`Ge_t<Ip<$qtB)?JNA=bnD#SvQ{LqV#Zd_PEi|8w<4yebjci^X
z;-r__v#;WX^vsjlk#{;z52p6SCoXZ@7knFM$9%fpJWS{DtKK31#UlUWO0kfgR4=t7
zU;UXr-cpngfAUUc<NT4?r_aB6`^hgo@zvJX=%tvh6Bqx5;!4*QE2sXQU$6Vqd9rw*
z@!I^l_tM9o^8)e7KHk&1rEkAee#q?LQ=a@o`^q2t&M_H!u6VAK-4~3Lt#eO-p7UgO
ztoPi9^r85~t*^1eul2_92HA11$BTLPp?N$T7k9B}U8P`caqEvM;sL62=SdynKh4)T
zUhM~2eYpR+XWLgYJL--d<7wTfdv)TRIET)ix{{aoO<;Qd=@}>UCti8+Paj|SBtvnT
zPx)jo9ml7-hWN3L+|a0w<e@*|OTKu4wawFm_)hD=e#l$>lhr}WV|~f;NKxK}#_gN^
z6en4H#_=mJ`&vkk|GFZ-DT>ScZ{DBtx{&PkulE7zdmU*Wrt7?(OwqYZQ9n@!&RNQX
zexR=8gJ0|MEHC|sAO6Yu78&v*5Bp|)iu9~k_w;;URGd(}@?b|?=Gi4feypdL+Q$d`
z^vPcL``#+_y*%HWPRD)khaODz$m&V`hzsBH5}$tIoH>W;JFSCM-@X+ZXO};`LHmH$
z6wQmDJ$_T!xVYt`?&SsXAs&8GG!OaVpB+d~eagf0UY_iTj||zDhrIA<9<rOp;ki^B
z`dp$A>J$8{Z+SxXhiCT?d0MXzk=ZwHJ@h`D_doG0{}erMyk7DA^jzi#;@3UTdj92+
z`oXvS%*)65kSCcQzsBu9WZ%4a$oc~Q%@;!VJm~Ku@nD_0qn9Fn(s4X_{egeza|P?{
z3w=nBoT7Q?b5Vbf?ejPLfFFGwfA+<>cTVZ)-_94kRAvWn_7$J@6K`a7V4eNuU;mI-
zA!G;EcHZl2-si3UeV_WW8+L+$V4(6C(7*Im{YF32|LJ+IK)>g#|9jr*BgQ>1JRhw0
z9AHO(C0oa?d2#8l?1~feZ@uRWzt)MDKYGT=DY8>&oIS`Nbble^B}Mxq|8zc;`7<s)
zeAr(+n@>@Gsb8{v@O)&KEN<h_`r1x}=O}%4)v5jW{AVB1W5;-E&wVeA*Zsgc`Km8=
z@NC?ELgSDf>mfU2`p~@l3H#7JLmt+NU!MM*BKxZjQdAG(;Ya-R*(0a*hky5c@j&~^
zK0Ex2+qk&xuXxlqJI+1c=;6nD=TJO&GcG>k_@GZVZ`?X4E_(bMhvwBcd;CCk&%XR&
zYR^94j~+YnHEuor=<x@|2gPTd{gbc!__Hs@?OVCXKYz{(KE(~iEkClj+510no_qcA
z<CYtno2M@ihzE7YJ`|txAx`ml-C*A98t)6)C+Bd`fiDdi{C2A$?(wPY-;Fg-w%#~2
z?{%5+6rE4cP4>KgvtE8F;-OfyzL0*Qd3AwLGGxzs?~mf$>qX=C0Ty~c#X5fNtJkxT
zA84Fie%K>>U&Z@2#tX^Tr+M%P`Gxe&L;2A2dYgWlH~n;+zWCC4GUTri;sLUcuR`n6
z_~mI|*u$^)ZP{br`~U2jXD7{5UdGiSG@j}e;!{0~m!7!955*^6{vf|lT>L=eDY{qD
zbM8E!Qq(75y3Tsz>_hpce9JFgPfn5l6zR#648_G?%A0(R(-*&cU25O+#q-?r(sPFF
z?{A<!P1cw7Y4<Mn)r~mWch95ec?!*YKGE}BOYNo4QF-G9e^6ZNRekdBejy+B!~;E_
z#g)p|;m16F&5Hv+;x_L7Aa0oU1^a1S-O5{@#iDba)-72a`jyW$^#?LLolo2P*$$Ie
zt|=AOFJAOv{Kz-W%e?d8e58EHBeiQhMfsG9@(`DEZr}MAr}`kfCz9<0Ju*9Zf&9wH
zzCq(Ds#|`<&7X54KK`6b`8hYvmvfz>^G}aVj~(YoeDululdU7OCqCos(5EL3<Ed;M
zT9?ZB;*Ts2=iNAz55zaD?R}ASo*!|D7dkigovi-I&LuRTo=4}7eTW~3SMy$1rfC1I
z`#<vkyw1XpJX6F&TG!<C+~7l=_9K<W>p9~5i&viFR(ImTZ;E)vuk$WW^Ww+5=cam+
z2YvQD&%|rJ{P>}7-nl2UoBH!yVn@HR-Z+#WbT7jT{_U%C1nJ=e5BeAXWJn(m^5LHz
zUf7i{WXC!(d-TMKfBf=eTs%-csW<mD^_9-6M`&Gb=zhVz^@XsIzWb1MsUG_&(zDJy
zd-62z{RFbtS-vj=ecyqs9(=FJ=i`3Yu+ZnP>A3jBk;W+>yz6)BSU<3j=E?eoyxFsk
zzBtT7>+q94AF4%mj+{g2JUL&^5q|Uy{T82)9v+S3%{b&IMduxgBh}N7(|Prf#;XpD
z(-RN${G(S0(|FQ;LzYKvnC3xd$G++7`aNV{9eQ1%?^$P`JwM#bt@pgaul4M?|N6Z$
zeL$S%)s=Ph`6HXxZ^XqvJ5U~;51yam#~b~0-0LiT%X5RBG%oSsO`iPVOWk;`nJ*NF
zx>9%6LC<4)@=yo-;!Pgnq^B>-S3c^bu}~h+`6EMl<4?bV&K-UH={Lsd^OGXI6jQy_
zK0o>z85&QKo%H!-oIX2vu#f8AKGN5B#mycW;*G3cQ=a+7i_aJEna-1)3w3I}=RBEz
z@rx^sTfb&koGHpf9QaE6hj}O;`{B7RKI7tWF7-p_8uDx2Idm?~<3V2XXWw|5SDLr{
z__L0EcGLLuIdNG}PhT<4k9{=`({ZwRtY;t6gY21y`aC|#^zkAu>)4ei{>5cId)7H;
zcreaCyX@IF>%<{0NDuluhV*$UuGEilh*$ZD)4HZc^`{;oKAb<}&K<N~9?q>kp?{Fs
z(?5;VOZ~9VpLu9LMdwmn@}ZZmH!r@_pLOnW>b1~*vS*&Y`cxmaq54&?^vU9(m!1dy
z3!!@jKJ0^i!H4G;9@MjYihsY=zhC3;Je(i-i$fmHhx6s$>HP7BPxU2F>r*|ifBhYb
z`;~azU-@&Nbnm2R-1pJN$-g|*v-s>!%BSZQ^xPuj-G1Po9s7U}{_%<jagdz{<Lt9T
z4~oM%Fm7Mc{OPf)&ee&y*=5f<d5DJ}diuXQ=g0ZB9`C8_J_g;#?W=mR@6MGxApiV|
zLmp|K?kmRav;D+JiuBp74O6@6`y=T*|MbLZ-1{ls=YirBkNUJ;-1g1>TgSe5*daSd
zWOh^8IJ+>lZybt`?EJ}>e|pXf*?v1uWcK(^Wp>34<(2CD92$yG9M-X8y?D&0{Ms-3
z4#kZR<AsnuzhvwAhvI<hgDfxTxdfd%sPCtJPyb*S?@+wZ{^Cu&>L2=?=a+H*$>zn!
z4jIZD@+)p=K1F$nAEtWjq$uBVQM~pEug0NxQvdoUeymp~?8=M2c-0v>^{2k+u>;jh
zZ79Cl^eSZ@k9fg9e(>t~=y|Gs<*Sb6rEcv5ef6K})8mI7<79TAadC)WpH)ZntYg<a
z*?hX*I6a6Tb)p{iUC%$`;$%;sD1`XN7c4ZM&g0X#*Rk#??lolhTIh2gdU$lt!UIhC
zH0~Z`UL4}$&%K%KUZrk~`y9&Oze4X<L-#Xiy*h#7QuqA1@5>tx?t$#O-y4?){+$PP
zr9RY!=P|^Ox^hm%m-4Tk^f9RK^Dke0PCcq8d3%2Qej9&y!i&7+sgC%A`0%>Q>pW<`
z#9=?(v)T80(LOxz+rHP$-)zMiaqwe)^M5Qldd%s|)v#~9*K79AI(*};T(nQ+uD4I@
z6hiy%oI(4Q`j?-6Plm?R`Y{jLR}U$Q-#8h{N4(AjyU;u|u5OLv!8m*Tr94<C4*5bn
zST7G~T_LoNAM40aUokI_bUw|?I>?`MEq?n&rsten&p*V2{5<ExWt_fo`$LuwKgRiY
zuIcF)`he$yaWZ{+;*lRY#gqs7wIMsj>hWV;Ayk+6CBs7NN*zz_*@rZrQtcMfug$Ky
zRIgtDi8Dp|Dbh2}9vRY0<26oC9OOdic?><rJ=dXl@Zr3$%RdyKdFVW<EB?$w{vf;j
zv(KJ*)B_n>4^unrLG#v=+1CdOA-xp!73bc1e$4Cpc(M=Trw{pIk8Iw!{sYCKkGbFK
zNBWd|oOSx0`@Z{<`wE#JH17WDzGIzndg-`%_Z|13sT0~SIpUSJHQs->PCY^QQgx+Y
z6+5q9)r0$1$_xI|yycZ*%A-8-W1b)L_Ju$DY#;3(e|Qm>c}S0}e(>Xb7+0TU@rYAA
z_@ZZAUdHvIrSkuL-&dWo>mDI*aa(7ftb^i{AD)~S=Sy68;s@g2diL3&r$6D1zH<(p
zKfJKRuDqfAtrNF;wlCtq4}J9y`NJQ3_+r;MOznz~9Vl<S+8_0Z2lejtve!+}=V^FV
zSMu_Dh#lwCez7YrpNo0Dt1igadEKS&sEc&J<g0(<RXpP5M_r_dFZ<69{)|KXL2;1n
z3p@5NU9YbA_x`iE*~K?KJg7JG{E0)}^dLUH9<L46gV*6+XWJijlIn?HT;hiEH_slV
zr;ga?-}_9SOL$jD__7}IXFt>rf6lXUeA*YX{2;rjEKkTU8Csvp{HAv4L-X|cF%MI{
z^gO6n=y^nj{8*1S^Z0}O;<->f?5b<}@=?c7eD<Fo`eb&={HqiDs{itz+O^*EQ@({z
z95A)3E}cvJ&z`vPs~_v@;-&}jBVP9KNe_RJUC*l&*@OIuA7A>XdHskU9@K|*@?jru
z&^mP|KC-;hJmiNzh;O{82eSCd>P8%7@v<jgdC_A>oZ=TRRA2Vpdh5hl$i6u60`X8A
z`krQOdgacmNBb==h)-x-9%Q`W8^7tg+Tvu_x^hum@(Uc)VbG$fIUdxddz$CEdx(DE
zeAx&4tX}N5dG`?Kntk_6_ey@;OQC)R*@yZUlm`?aeRY`XJGc6Z^Q$jfN46ds7bj#V
zMfXZPrE#X?_)3wU{n3}~hxp_#Px<&<)B6IxZ=&Cc3!m0Q^Xz-y&3O|CJ@u9%`=0Y@
zUyzqN5<lb*AN2VVhk0ln{>V_=Y2BxG?Jqz4vm;LP{HG`meCU71*(I|N#R2V0DjQEx
z9rG_f_VHm{Ug@~k#a`d4L;7TTULQc~#7&Q1af?$u8n<6$=UQLE>o1<Xar9mT+YC|f
z_SJJ%oaw$<&%S-J9$(_6kEe9rc&cxm^FU7HvEDqM`N5NUapIT0d3AtC{EE*yaqz>g
z&z+rDe2Ozg{Sx2eVUJ(-%tQ0^q5a{Ho_L)j{1}J&G-S7-(LO-&kmXSb?TdN&v*Z1A
zb!*)HmLF(*sowv*FPX~xrQ`hL85Yua-sQobx-_pYz0c)!TRJZf$ew)^M;a%6Jdwr4
zk8$?YrTSvu{l|L##G}6TE9ahndi;>tx8Ax`Pn`6`X&lN6vd<r+Zyw59e(5+lU2h$J
z#3_$7FS2;(r}qf!AwBDAL-B|=-4FSsc8pt3hWIgWz3~+38Bb+)oiF>VkEi48q)5+r
z>c>9uM^F65Qxu1|)OmVteV<7GwvIi1%zK>y-5=bS)v^A;zxsjn$nvtzb4tJA2fAP2
zhrJZtH`I~3QlIQW@6*?Yp1XdRfeiW8&&km1fE3;9>3M#T&3oSj@`F!t`5jI8=x;9H
z>7~P$8M5Pt{dx^qtId#KE<E_qBerWhWZ_Zg-qh{mWtPguQ}leUHpGj4hx!v#_fUP2
zod+^>Zc|j((0O3jxiap)$B+JC-ncq8&u{974Cz@<77zRM$@IiwJv5IG>r-CY;}2R#
zrpHbpzw}bH&bavSZJeHQviatQ_N%_oJxibVy3)Akg3n7mH;j88k)d&RpgzqHyWU6h
zK9_Nyr|IWp_WAR^7@0kvx24~oX2-fTF7e41@@t+fUUAAp9MI=}DXLTbLp^&v=zMwJ
zn1|MR?izRQ%zHk|N1XSrvB6z$ZrQd5dVQ(Sdj48xKlzh4{_)~-SI<@F)AP;qu{M+k
zO!L!6@nqiX9&wW0r;N+T{X;$A%Q)FOdJxa%@hCp=(}V79=EW}`<L*sqyy{V$WctR_
z_|1#oI{knj<JMb8HeU!+`($XHIGsyKk3V{3C@yFp@Jug7_Usq_t;0VwPqu%?$?DB`
zx}WrnJ0H+Ge#mJ((f3>+L+6(t==C3&9%Rq`)bm|mU>{mXPrUjYdxh?w=G{A>@e~Wq
zi!aqT?%wTQ;eLq^@zOWXp79jrZG9nq_UNT(Khklsygd*2!9P7{UVXC<?Vma;gz8S-
zcud!)av^(##+_ew<l#Bt^%P9U)vI}Nl8Z(ARGU8Dy}onq_0jZs?>^x9=(+2AsqQD_
zDR&=o#opbPsj+{auli<s&tV^WKGXBuC+mmqG2#-p?<w)ee`?2mZuYa9$uG@Zrp7w=
zFYi}a$BsVXxvajNPkDPDcrJ-spEr+pNKakyD}VXA-{4K0>AeOIFr6nu{^;9Z&mD2n
zcP^p$#fKkt4%K;z&cAi))wv?$n_seXklM9wY2TZB?ymE9nVs8D`p&U9pmq8Rd-j_h
zaf<`8<GiT@d6{QVUY@&n<sXlp`{GmAo)7HVS9z*G<JR$;#zjt%9lVK;U$VTR{P}fG
zoGW~)Lwy17c$Po^*5eb}2kShS)Qx!LCr<Iphd=vb-g@WOdgs#q$;Udfb7frq;t@Zz
z-_ECfG>#|vvBSQ2<iVf&t9-;~TwUR@5WnK2PqyE9)K~DNujq67y!z1Zj9bU9xcGy$
z#pygj=Y&1{*(Y}CLF0um-5>TnkD>dLdG^TZbKX2Gq-PyHnBGf_L-yQ<?GrxKle)3L
z&W-vqPe0XjZmfgW(@&8-_G?4+!k_iV`Gd|0eSL;Z53*ATt;c7%c>lqp)}C}<yQP0;
z?LOl^Muz_G+PwSR$?eX2Ww%w@EWHQ0Z@3p(ZynTk%(F*!pK@-USEzqMbp_oc_|XTT
zz68}H8S2aG33`6<m&PR@@r%Rrh(7-GEqUY7eUcseWOl@1J-sbw4H_|X>hd-4v`eQy
ze%y-fYWSOZ^kpAi@WP5U{%#Ea;?-x>tLLTXJ=EtsHwx8nq4CrXnLYN?arX2_^WwCg
z42`Gy?8!Gp=b4?fpI8s+JKxYenIHDdL(gCO;!o|<X9wz2kiFEeaqHAOyY$I;Ad3$g
zXU}@`;!lyDb^0Dz9%ONn`Gw{7$ygT*1Os(vz<t1T$#c!~+q~zd`-bP%^l!X)e$V}u
zuleym@BYr&SFW)1a~%39TIcze&Wpoy6nZYo*nQglI=!#E_ow$=uN%CcBG-ocRk5i5
zLj5~kZ{9fB^Uw3udPqN&jkDwR8a;WkYd+N{TgUGIWADDBuBxuJfzyj1Ma70KwiqHY
ziQPoczDkS+C9%ZB5=A761$&FJ2Q_Lm79@5o(Rl3zJMp}hXt0Bdhy?))0Sh2V6S(7!
z&)>=To;&t9FJLf6{^1#Ot+{4jbFID4c@I$h@HzL;(0ckjcgTx8@z*+f(Bv2=&v+s}
zSSH7KqIKlb=RO0iGtWQc(76hI4(W4NC_m(ro9Cx^JD<gmz0PU(VRFcoKkFdBVHqEO
z{;<P3_X76;=sw`SkI#DXA_wB5kDtcLwVwa@*vo(M;Geid^YoMF{w=QbqdiBkhkon4
zfAN0XJwEkWXI}orP5$wzJN85Ndw!M&`pEY@<@t)8^x@+#d#v;RhX3Mi-a7d+FJEZL
zKK}7XzW9^&T2F4)6*>6WCBOLjhmYNf^r$QRkp282AHRHf&Q1C3lW+RSBVV4`O%9sh
z)`=r~tcUolGfqEy=rJ!Z&TY@ZsfT`gtS2AMesQ8tzSO1tW}SP&e+uUQ_P!gcC;4hD
z$^(9NE5Gt1FHj!j2QB`_)vx@>qr5?VLSD&%^g?#3D|F7I$-xhevlp@#8s}#j&3=5D
z7x@xz_aONdck@u5_`_fOpFYSw_NQIRCmzNjd(e;_;+IICaqHNHW}k8O2KmolaWW3Y
z8Pdys@q^^yM?>S}us5-upZu~84fzl4518ZVTu(m4m&o45-4}gMk?6TCG53Y}T<`hM
zxO~f_Iz>a{(75Lzc~>u<H`&2Hb&7^Qe?eQvPUGs?JWRR9t<y)KarTqTKl_9n$eu*=
z^^o3NZ=7B<B-cE9&F8pzaj`$;$GCOkVjRjd8j^>WfBh92w~umMyrF%JhSuxLX!}|n
zL+c<r@H=Pie|n8W@qzqJEQ?dwxcra{Q!cyNDSrGnj!*yMr+BfW48@6D@$j7H`Nron
z?58*78Yg#g_fPuLF81Vj)-(U?3;g!a!M$H?yY~k3tM(;-<p-Vlv|gR!BVQl%p2z=#
zM-D#qf@Y__>-~uSpx@*=zn4Hx#)qBig1zDk>4EGQM{z1c>!5Y)<zLPxAATs_;s-O|
z=(0G7OCo;jp*Z0eM|2|oGPIvPFWR4;=k5FK2lU8aqIqaNOndN^jhjzQetxhU@{@k`
z^`C<3kAG-&tX|zq)Q$N>bxw}BlVjhgOZ+JZjUSDl{G`dX5Aea{$CvBaYhBX#(?9a@
z<I6s5opJt&Gnzbo29nEueDtzA<410y_4v_gPs+!a^1dpxbI3X8?~RN@ad%$uFP~qM
zAD{K&CvWnERu_rp$zca{zM&!g@<a|)NAga-d3M<M(0KMyv^cUud?7t$$ZmeQN0h}|
z+><Z)%_p+gxVq2v^uRptoM+Ax@vzUu4SIj(ym5}Y=c1iM;%}Y4ZQlES&kJdf`!|$l
z@m42Um-zV0ethg?7yI4Mtt&(R!}N=P;>bUI*5O0*6D=Rsr#$O(9X*i$P@JK7L41&W
zcCZVIr_Y_#6aA2!vbd@Tc}cX+`&)YPC2d?hAwKa#n>UU&55<8!&~qMCr~Frc>Q?;d
zlMm~lar-Irlyu5N;}d6c^?C8ar{2^7Iq466{;^xW^=;$gNFRCRi@Uh8lYI8jYaLX7
z<j|u&<wqUKr#?Whb@<3j`|**V>+wT!%IvhBosb^$>?N1~iS!v4Z>SE@_{dHF@rfIL
za`<JO{zUa*9l22bnV0uOa>)}HpA(Ch`0)#E-oD_kxRaCTL$1dU)dN0u($5e3I^!jd
z(0cI^f8&sRdeQte4vo{perO(A2l;6pZJu5BF+Kd{kNmOMI{V!Ihx#bgU&`)j#;s4(
zr|6R(d}xSIee%P+b&#Eqy^TfoyYG<;=@)nPEbi=NpZpq6#0SmGpK&Pv^rw9G$vb2p
zIrQ*Loa@ESdT3u+pNNk>{f3?O5I;GPJT(9Cqw&Eqed2`Adi?Z2<M^y22ijj~>!AIJ
zhUD^xeDh`U^sD5<PcC^7e;J+Q<mEc!;smWntGmSHL(`+)V6IQvIw(Hking!xC;LD>
z(}#xSl*N_(Wqzv<^6=@;Xy+wA@I!p!&mS~B<nWsw=ec=ylY<XBfAD!<Ab#RuJ^RE_
ze2pjVzN~MX=ZC(>4t9x;{wY3-ir!oBi#_(Q`F|HY_LR|mHr%vHgWua-T<E<C{k|7#
z9LlSB$t$$4)R#D`Cvx)rjB!YQqIzJb-{JQAc>0LA(PQ3wAitLf-5<Qyg5F!DKKq7!
z?68hJdacKgPn_@THLT(Okxd$s-@Lex<NvAUpY=YUGjDyy-#+DsyxJe$i{Udazxu&3
zyFT~Ou1(e|<Xe8|^SwjAU!hOg@9snT81(mv>gtp+|JZT4(JK}At~GSU*9NauwAg%)
zqwo6tn#HG6Z$9={m9>lB%fC2!k6|m<{?2ljUNc`f`-pE8r>%0s-`4r%D#e^9FS%~Q
z)yo$C|6G43vFs^rk8C+&jpF(>Pq=gPd#e^>j+?sL&gZODj2&{@p#9pcQX3~{n{$pB
zbN42z6%DtKy5-W}tX$JOzi59^yDL^Kt~z+T?S~$+e6iP+yHDEmxh^&AzW$Lr%-g?R
zp?`}Pv`_e3hWaiPZzwP3$ro>Z5I@8Z<&PcYrv3ICv>u&EpL)vrQ(w*neCm@Q<f>0~
zrJmT$f8*q;GjthR2dyJN$N49p&a*tv#j#8td#n>D=QKTu;)hS%#7kc?4()gH(E1gc
z9r~zw>&Q2*ZulcE?2$itl3#s|UUJanv%~Wo`^@8K7yZ8H4gXU<2mZ4Buun6u`i**&
zXLTx1*2`bkrE}Q4=X`ZxJ^kuYKQfP2XZ9C6>BXON?0bCX*+CvV$f1w_kUV{Y9Q^i;
zdGgI$Zyf4}>fJa$*(VQ){Gz9^$X<5QkA`LP#D%@`WjvAH?9KTcCr_Nw(E2ZHAAQ~Q
zszdR&zm3bEb3z}<`oWLSzOzrEexsil*H4@i_#k=ehF$C-$2@!ZV|`-!O%A>6^nE?Q
z`|Wwa`R99s5B<Gz=jahF8|d*Iz)p5$oQ%V-TOBU(cS!`kiV@JS^qoB4?R)2>ev_#0
z=xh3`et@P&|3F(uAKLn)jq6kF!$)rFv5ubfgB<hZTSq>$9&H`|r16n!-Z=a!*3#F(
zX3v%0zd2u>|IS%_o*&WBdgnAg`w^N?TKyQe|BSOI*W))1=|{7Vz2@B)>Y;nQe&SxN
zzG3bk<51s#?)B!$wQtQQde6cx@~nrRFRV9?HotkdKGTL@+@itXDabpy^m#wva|<+i
zXrE`G?Irn;w?y{ggUN@kH(z$&)Ti_{$j+}jW?bz*^&(#O6SU9t5pvX(_t?%c{N~l2
z{V31&tG-}7_YK-UEJJmzZlU*(<RqVdqz~AC)<f~Jk5f+4)`_q8750PqMD`lT&u?}b
z_x{2Dg!Z?6?fs$O&G5SkXy0%0drUshKqq>CPOkU7^qZ$2vdjA({z7q5S7`DfdGf`6
z>--L)-%<2CC}{rh-{&~i@q=7w92zh4m!0g=KgHiXyUfeC-@)>E4mAGJWn1lk?vIu(
z{LYnezvD;`d-+YD-`(=L5hOR|`5i6tp>^!%*PX4~ox0-M9c#Zg>36)uS=_`QT2CH6
zG~|~$$of#9)}wu|j~?SmTW1_yriWeRsSo*QFFE>yb@sP$eq}#MT0hXAay>chqMsjV
z`!o5i69;_OL+j9KhjpnBzxgt>jy?G3(MNJzpR}*U3mO+c{oJ^`n0KB)=L6cgLcVnn
zA3f}Ij*<uQvj=*PbpJu)N8_g-4e`+j&BHSL_~AU~KfA=idZ-`yU0S~nY}~$)cYf*H
zhi$gXnga)PDD<J*pZ;l^);G2<^euhP@A?_n7xgoGPFU}O_r~nAO!4~M-)^wM`O6i1
z?sN4~>;81v;>2GcwfB@amn-xqzaI#V>uc;uG|$h}FAw%n+OMBLeGJVH`@z0)j%MG`
zcf`Z~w$3?Co;;9azZs{WKI2(;{8CrsLF?^1deo6|cBwyTKjA0WJbBiuS9OWD5A+jp
zv5p<`&tH1gF@AorgI#H_^<Q=5Clq(}UFM&<fcVsrJ^<xOzmdN(lxOSdAs^b0_{j6T
zqTkV<&!^;CPo8nK_2e6e`1$AgjUU#foE#@-VNu@ng{)h0?F)4JA-?9J{b0Y)XW!X}
z`W8O*p}xprU!ry7>!ajYkES>89riQ-@F(&Qnzt{YxbcJC>|w9GLUo`X)raS7sBYDX
zx<o^AecvwaWtVaGLGhJ0{gFQLc3$Y4&L8(=ezAvtkUucrGk9MC@g<svId5OnE56Bx
zCa1BO>(G$@;=vB%X#D6hB!^%8k$?3F<yHOCLqFPkB4}LwWq#-*M}Jc%jYaSGd>_>3
z$KFGEp9STGea6|DwDF9maq%<HpY+c-zw`|>6nFb8(f*Sc`wb2C6Y}usFZhh>bL?QB
zeoe3S`jI|pJv6Tm&<C9(_I=u6J$d9<ho7J9AYc6XqkfIczk0xDf5<CZeagFh%a3@W
z&41A-4&vhe?t8huPyFzV8+-Jfzf948PVbvfZ`Hooea-Q6mK)xw_B-mHm%LZ<9ESG1
z<i75C1)ukX<a-a~J&)%z?}3sw?l~&+Api1|b*vucRi4x<e)~@T)Q9uj`6HgrEB<*N
z9J*1Tp~sxmrSSa5K6KVGTE0?`_3Sk7T!j4eoJX$bNxye*Kd^g)1G_)b@!h3s|No2U
z5q{F|`B)yV_|AF5Z~kc6V*HeT9gaD3wPM!Uqr1JjM(4u*l7H`2{eMDpoA0#p>{XU8
zu6_5B&BpZWTJ+iF+?@~k@u~%R_O1Bp+xDY8iHCd|N84xW!u~?rw@Kr3?%D6DKWX|Q
zIpjlf&68su;zN_GJ|RAG$aO!GKmA5s+YjO-Z^qH~i8!cJ_iJ|P_w@Ta4d-q8Vcpji
z#Z^AUH{($@j?cL!Z@veoPe6R+sTZ_-LFbqHQg6`yVF%<_%BRPA`&7Lc=LdOEoswhR
zJ<Q)9u|Ms$j(l{=MY9Xi2aPB4tIU6T<sIs``nP>%U+M4q9va%;P#*Ye9P*Pr$!{DV
zIcWUsAs6O2+B$sXnpdahp?T}k5TAG&Pg=Zk+<N-?&pz{L^H5wMJI$*H^@BDKop%r)
zdE~2e`vxERiR#k0I@2HABb*z?^?COo=zf9ryl$Ow>&c}TCZBrG{eaK?MSK<&osad<
zy`&yr&RYk~^8<=o8R`el<ubk{<NI#$^lKkh8XE_3)epXEs4xAeqyDE)E^dGATlL`_
zP-pg~zNX*#9MbbQ+VhR)D)Tw+IjZcr8b7)0f$V~5zw;o^2jj5b`GT(u%htPRmaVHd
zUM5$a<a+#&-}sHgtUKo~%=?16BroNYXPtYBaqIG2&w2Kd@4m|}a`3?%N9Q{FA;0uX
z`vIDFo<aV(ccGp0>X96L)|<y~-2DJH_8wq8eQ0|4Xa7KY+-K}__aJ;wJoE|qGcFGN
zh3uh^on`*XpL$9?Nz>1+q{)+a`$v75&-LV}qpY`_N3%~npnQmj{HLGvu?wbtei(<w
zt>Y(Tk9?>Pb>Lo#U!RpX_dfRXhac=0Pj;ZKV@JjdU&=8q&eo&pA&<S{4%wOO%%@!P
z*h3E*Ke^(8Pu-#M(=Q&mU&J~2<;D9a&+GaontkHN4)g3)H~82ue(YoiKI7~bKkNBr
z9TW%lC7t!lK790B&ko~|9XW5Eys{gg{UHA0gkKz;^X#LS9D2mtIDO)QPQ(xCMT?WX
zvqPNHZtKiLcAA$@_K`zh8S>M7e(vf08a}9=(bn@zKGdW0Nq+S)G`snOmLL8>e2LbH
zlX-a}4-LhOJ?e;m_HBM%obySGllX|Q&yUdJtUk%{cM9GwLUogMpx*h(fAPqAG>+fA
zJen6*@=~vH{N^FM%+m+U?iGtVE}u|dp*TZ%VlSE>;x0aD`OJLFtNJmIR!`75ZofeL
z$-c3#jHCIfkE`!8ls9=m%eQgyAP1j$d4=?|N4$-@pO}Z_X1wJA(kuSP>1QWCbxm&S
z;ji`VWS96FCl^0E=t+Cjm%OV3NItvqsWbNKf8?t>{AI{q>(J~@`;tyQXvlB%PA<FY
z6EFP6@xw*k_qksE@*lDfs)GeZc~c)~^^+*y#_Q=LFZqnC3-S{2C8j**pm=;$QGV?Q
z^_+-L-N799`yD>t^ZBOpI_s^BmUsORt?#1~v+p|>Jm;6`<BvX){+l;$|LRNP2l2z)
z&-$x<Z5|ETOAe$rQNPH%SO@tDts{>;{7qyRB-c8~4s!Xe&&Z2?$PV)G8)t9sKk>px
zzWvS~`<Om_(0m!czJQN^iyOs3oY)JE!+({1<ja#h+Xvz*Kd@~7u+#cP=csXgr3~?v
zp}HoQ95m#Y`=@b!Sm!y#d?LOwO!+CVY+czn{pQi)!7p)#@~WQj@z;9t(r)AQny-iC
zL;UQrzjObwmp*8m{l>|cfBuL!KKjwHY}~wie99%Ce*44yfuHsX`_WKe(~sPz$k89!
zfzLQT`xdfST=W^^>@hAMX!FpxI^wVTpvO64o_+GnetnxC<g!D2tVf$?mv#J-5BXuA
zd_v>Nm-XQNgLTk)Xq`ADvacSWx)H}R6c2LiAv>Y<P+ZaK2rUoxpL#G4jn}isyg2b&
z-MCM<H>wk#?^?$m^5jb$kSp%|R5#*5UixjG{4$zc-_!N?`#xv(eO&ig-`h6MUvb9g
zd%-@Z_W76lGgO!UuEO^}-P_&cjf<0U>->F2&KviB4zh>6Xm;he_cuPrf%y4r9h7(c
z#;r51zd-#|evGRN^{O7!pL$jo<Usvcy^yCK$<_DJ>exDV(N8Z_5BeSX>f8B24tvq$
z;Nu@8A0Ims`Ns~Zp77ge=8Y%PCl2iM9GLU+%rAC}FU)a%($6lKezC`XQ|IDtUY+8j
zAF4<7qF&QKewFq4CF4sXup|Oan&@}VU+0W{WxqKmAU@x-Mf?6%qW$cA_WdsV-TwA|
zihke2B|nip_PcwH^GM&yK9zL#D}4;=oBrO*_t)IFom-GU^svKxu5(X;;x4}Y;=lO0
z7a7ku>x<%<_afus2<exP3FF$&-|d-}wZDs6<Huu99Qm!43;$1s`<TDa8hlgpN1vOx
zVqqP<{!Yt1&i4WRJ=uc~O?r0c(>fLDpLOD(ZhvsVs)vm4+`7TvWAR6Pp*UENc5j0G
z5J%@OIhjB5#0l-(MML(<GrhT<K6T@KcHdE#&Q0}3F8|cEdHY13(bn;gUF?ATARoGa
z*(dDHdjgu>Xy<wAw|~Tk9pYl2(<d(GA^-71`q_`qIC*8r9`WKoIrNavKjZAi55)rw
z<<q?h4a@GS`e^3MK9Udd<`0^FeStsf9C}{z{8NUW`yl(p**N<=Cqd6u;_bPL|Ku4L
zFY=*t58_Y#<UxAar%xESo?K{NJ+q5m?@g>{Ct4o$M}0r%^<jP1x;#(h*?G+_Xk6bi
z4%L&oupgkhQ1|MC-|9;qvku9zU!ZYz7-tWCIgc;VepA=_2K&(vKYjYAc#1Qc97s+X
zii1AHE_U0GXun$m#Z|p#{Ta{m#JWUvsD7<um;CXU-R$EhyRC!CC$8>`_{5z)bwED5
zjMK+oeEd@<@<K0UmpXF(*$?)Wd#rQSet_zMUh!cMWUn|v{<EhZvX@`-M?V^hr~HZ+
ze)gqa>%?2#T9>r-^m-4VFX->a^)-2A7dz!aoZW}z!Tm;l-0#$taq)uc0?j}9e`??n
zf7xw~`4zu+=YG1dn0`7>_$`m(?cA}i)eSl7m%aQDfBuM%d47v)j<dr)RLATxpZQlm
z(7NP9vx7YSl|AVnK4_jk_NqI2V2-<&i4!@Q4|a+_ntt|@BmVrcztu0=Jb%QOUF76=
z`eD6z^CM~U+$Y#UKL6AO`}m1pz0qr*vdjL3>{(P~$Kpc!45|zDBu?&6`jPX(Iw(%!
z%nx-dKKK*;?l|=R*gaWa)o1Brul(x=DIblGJnQv&_gAQoq4g{GKjV4-RUhO+eOo-$
z9lNZvpY(Tm(Ep+L683}mLvrnF`&eG^*(dCjFXQBsXI?yVo?Lw9(a^mh^_Vw~mM>_Y
z9R0<9O|-AsV?DpokY4*ke$*cn2j{*zHLgC?vw7o5t1orSFZ$IbKIbVk&VO=@<0FqB
zkX`m8yM1nGp4}NAdf1uz*p>Rx@|%3-7Z<8yb#K4wgU&~Gsuyt-2Q*~Ayg+fvdQ3ai
zJG-1)XlUQFlN_{u0Qr%~Zg!E6Rv!<XJ7BY6yEbpoU+8C_bHRFc=-c|9{)pdxRTuUr
z`Dpt@9*pzHdhr+U%&Yefo_ioZ_Y8I#*O&Ece(U4TRe2Bx`-h#7{p`jkZsf58vcr28
z^~!Gcu+RJA{7gwctdl2wRG!6~o$`P-53T3F__0Slx~HMbP#y4}-Q+@XQ&(tp%wF+_
z{GmVBi>LUQmml+)2lhgC*l+f^xaGWkp`O^uZuL!`aeX__EBiu!hxRwxxd5GW`kj83
zdeQW-&pP(Y1Ao!bI!M2J9X-bR>-+0y@yl~cJ=uTkvj5m`z4uo5*(u&I?~C+`Ya)K~
z_{R=@vhTiL!y4`%*`$F!>(JuNPWG!aae--1)=S3MI^+1!?Bo~!t%K?+`_aEU?LMym
z+wb~-{-CevtNQQ13gunh=KeGeb6)=B3$6ap*2%AYLi35_sM9jE&c0`ladwc0HgDWI
z^JVtp$CrBTLve%F(JTJO6H~r*^obwD4|AVdpL&e**E!`JbpB?&I-i_l>f1T1j+>nP
zRNGZo>RR|+QQtGrhssbK^(B5mek7lBPajh^$#0!;`M{rcIdAbp>(KH;e&#{H%W-)$
zp8C<^0r`o>2l<o84w&=S^TWJ)PUJsSS7>puj=kdQyfIEcG!EIF_TgtQG@eKvJ0N@1
z6LddKG;bX_#_bdL2=9G7N0_&7oCn^=``u^1_w2nqe&{^0KkZxc=(kUOzTo{oIglLt
z+j(Z39P(42dp6{+bCA9Ijd{;0^oS3A_~n6q&pGTz^V2vx+2NeU&klN^_@UXw4|yfu
zy!gr|zsO^kerp~3%v&c8^gwo_q4?|X`Txy{x9=_K)BF*aw1-{dFCXe2EnZ0*7YEOe
z`0OM1e)~1|wZ5Vs=RDeaXk1^=SI9FDjl<-p-#pY0p}uHdAIo`q&7%`D4_P1TO1zig
z#Gm}-%?Xvo6$kqd%8&O4>Rug|)u+6xAN7addlYi*L-X`IryxH2-2T%qtY>GUdG*Xb
z_3VCwF1yF^k6mbeKz!UM$fKYC{@;VNhdgr7$;W>2O+EBk&mVrGtxJ8z@mW{)JgZLi
zS$Q;$Hg6otqjSo*zF^*Z=Zy0MKU$qQH=Q5&Ao*#ZaePqSvInY9n0(}?f6hbSA7ih6
zX&<7UZ~TPhix0buqwR}4&&Xk)_>$*6jdAtQUUqqZlJBk5kLOs=5$wr%^{o!ovARxF
zzt-6={8Y#K2AUk}=!5>>3-a(m_fF3_<g-UTh`)OiesSa<J!tWP>M-lmJ;*x9PWHH`
z80RMx5Bzzqo7eC7F=N#SR{P<xO&i2ppTmbHmmGCxzu+?u$)OJ)Bv*W(br~=7_J=yR
zAI-B%UZHvMq~Cnj7eDB;&Nx5V>7J&)^FzJR1KGn5d0@XfR7chshxpC2(|jVkAb*m!
zj{KCHcHy(mJU=0SlQs^`!yLCi>9r4`aVS3QQ^)d)rUwo6b@u4H_|f={L-T0zj9bS(
z$bR*q-pH{Y;!~gOh2lnjS$(PpaZ@MqY95V`T*xoykaN-g^EoJf<IsMhAL^I#M=l!j
zTfg)<ll^WS?H&l}(O2;&iibYHANz(JeOzBuuja|&HyS@XtP^*9=8dZZ^32N{x#Zzv
zk9ez-)Wc7Ht51GgZymkRyn1%85Px<-a*|f}{Ij0jX%D^Zx6jlIv|c^phxVO#i8uMy
zJE!ai$gZ-u^Vh!RAA86*Z=ca;-Z($mZ63;tdy>3ZFTd$0|M{)I@k`zF2fAO02lQTx
z9qLq`@v+CaIOqrbMAOGla&o==x*zjj{NzpC*g?PkV;>r~-u+zM#KY&j{=XM}#<;)d
zL%WyzJ0hRonlD592eJdLzKqMCK3OJrVaGS`)@R!Ai(52U|Iptnca9#>vLXF9Z-46-
z>@*M6BfpG)QAn=*(FfUu5AsWWiMRO$MeFT1{i{qr`^1qw^4wVc;>iE9b?maQ*ah`9
z@*sO)nSAFDOg{4L5Au8uAfJEDt9S81n`gIiXkXF?jgyD=c`5tN(<5K}GB1ACL45iY
zyX1jfXx#HYqzB^9x-nliZXG|=YZ<biJob^pPyQQ+*2@nXKY5TH>LmHa(fTs9&OMV{
z_Od7U3wihx#SI^t9Q<hW>^9$6w4UAKYTS8DzBsZ^9Li|*f-mjLdG!l(o?ZNBk9f1+
zy(RskpFR8|kDvM%q_+&yfA$zh)60%T_Rxc7AA4YqyN6hBz6|;6p2XgipX2nn7m;rr
zWCuOT@BZrk=l<os=6$jEm_ENY@1BLveT+Zi%^s*u)K{W)Ij%0f*K_~%IjwsqbgxG9
z(>m{I`DLDbc9tRg*oDtHG|n&eD6jGa<w?Bc+rAQaXk6UggWXTuyXB1@_0JCWLHifY
z59`tCkA0)==z-R&Tm0-d&hJEi^OJx0j6>s|AJjb>A0$8P$GEuK7vgIF<Z}?Y?36F^
z*+H&xs9xzo+i%9p_>I%gfAOXtA3xYlPg!2rt!~LNPEOMHnR$94JJI}PFS;JeoA2TH
z|GE9{i|1g!-%^J1s-F01pUNv5AA9v7<M`!)ANr7a_T%&ayrc2^JkK~i_P6}oZ_vK9
z&!PQoy>WI~XFlHp$S-^NFCNhMR@~p+N9ogFJ%>W}LHB9%>~l{+vxnXK7k%c{9XrkI
z>u5+0xyGS!s6V3FsXvyXb^O7HhU(D%l5hLiygEVKr{X{!J&+tUOn>-Aj{3n*kN#oY
z{xYu*xo_By>}3x*{2{OGbNaGz>lYUJX+4@h{Il<K|L_Z{NBh`0kaF<ToAHoucIYqU
zuv7oFUi{@(T-6mhkbZR9C12KOUDK0je|xS+>jUZy+OOG1jN7034!(MbpB!=1hv--L
zX)oIQDeL%4F8^Sz^Iik;2cN#`c^M7mk)8a)?;gk=@~z`H+B`qJ|6vb5p!WgC54~h`
zk9(Wd-becTaYzok`O9u~M6UO{n|^rXi49$vHTXQ<dUEmW-|7hR8=rjfhriYv_r91N
z=tOqlgZPrpI9Ts{pxxI$a)){Qw=38qFY+N?XlUF%RKMo=3B`r{MC-_vAAP}kab_?5
zFvr!KzM_xmC-muO>_Y2j>~lU^uOFlN%TCCjCD>;8ll_KN7c~y{S3T4x)lZ`Q>ks55
zZ5*Z?{-}>KRG<2aeyNYrn|;H)e!(y6^(+06ed^KYH|CA2Pkn+sc6tv#YmcJ_?KrJ@
zgLTIB8*=I8hj_8qJ~U5`dHo5}r@zrtTX)n&`@Y@kt*SiQkM?c$56^G>7FYh0@BU;x
zKjm4Uzz^B&cN64U-o(eadxQQZKKhA!kT{Bm`f|^5zty*VKi0haAAS5!fAD+H<^IPG
ze(3k^m*n6_`@0tJ-SEo~Kl!t`QJmyCQQsFYeycbB${&Bs%l8sofBvW$A6Dzd!M=jx
z2s8fj1NANSr_P}KpzSwx=YHVcpkMgADBm;o9>u=XFQ9!$E;OE)&vTwz>G3{<z8v>{
z+V{7_OW%O@zdj;f;w?V855z0S(e?p9Gp=Z89ojl}L-xS5v)(v83yS=K>`dA?`}H^D
z_@Qx_<Lb|NqCT$9)P;H|L+{t<Wv4pS-}&Jjh3XSsw$6G;uKHxZ_mAxNeq4Q<r-!_Z
z2Y(Xr)9<|;I_K$u?119Ue{#fyJa(8TkA5_C-r_Gq@*zFqp>B+$&Fg1rhk6lj_Tb|$
zWOqH}hrT0D_8~u@arUyGKTzG`%k}2Tvkt#;=Yu}u{BVxzGxbpaf%J*DzR6B@;Db4i
zwhkY;5FcbOd-&<x<exa;vp&boJ1-KQAMQu>&UfbzKF=H0qZ8e$-P1iE8F!!d9!lLo
z&l~h6dJbR*6tDDKT*XZtS(lh`GS3b)Onqo|3h8C1acEo~javuH^x{`P=;SvJ*~{-7
zH*Y*K=jrDUyCFSj`v6V8akRL|qy1X8zw8faJ-KL@_Tr--;)CqQ&pvXYd3KYhJ~NK^
zA${bb$w`z4a`B_nZhF`+uGY~jo{8k?gNf$Z$qtxxYW%B$`ryC2z24vG?}_?cj_b?%
zwD}x2zp%(Yc0h4eCpn(;?2{L?x<)Uoo!0rB(Ydeg-OG%-@1XHP_bB@hpM7Q=ok&jV
zA&1}Q@taS1Wi)>FLiU+=ucU`Od@%3p?1SV%_gT*)P+vv62g?I};*fUhZ|sFWXY$^#
z4C%LSVUd5HXZQm>f5Ch%uufbP&C3scW#~CWJkjJDx4yA{%S&VV7nDzbSkQG}7QcFT
zJ~}VmmyFv->ec<hdiCu*$1mRcBy_&JFF@~)?Hl^6W0!cTC$xIV`Y^7Z%TQgbbI-lr
zkB|%1HF@~GpCiY6H}3}%y+8Cl3BOlFk36WaL~+s&GLPabKjQ0L^L%3+eQ4|GkzaWy
z4<;XeXdb^lt6t<4pLNE|P@O=2K;z<J9Bn>P-QgFPw2z%-Xy2$ubqnR&docY@9jOoF
z-kY$CJ><~ixmO=`p7@@Ga|zm?;=q6MjjJE`8qeGADc0eKJ}<M*INEc(d)cIQkL<Vp
z)NV!3>Is`4bX~V%?MHuc$gFi%FRWWol)qdz=TG11u+NdL8uIVk#LK=D&-~ofy!}MJ
zadduONxr``)CY}=CmN<7?w!WfMPkmE(Z-?uPG0r}wEDNd)su1iT|Mff=E<>MU$D+T
z!pAQA*gE}C+@XA^U-3lKXFYy?_&nXdrC(oShk5&oe>pA=?8fhWWe@q*+u!VDhyH<&
zT=7HO&+?gi#5wCqJnYv*^^cDp<My|@w%&U>s88rCxj*b9>(wdw<al1RZ`onJ?+dWg
z`HUaZqp#Vg@=Bigs)MX6_K*khdw(fD@~R*53*tl5hi0es>Po!S8@rq@xnI<Qx-$>S
zbAK_e9@VLFc0%>(K1H8-C{E_>OXnUv=JE40`P9ES+h^j%FNn|nkZ<#lT=l_k`9deM
zmwa_ejyUOmiPo_L=6KTdB+^Sx>T4`b4&;Y=k{76M<UyR}QNMBzk~iz{(Px}KsQ<cO
z*&pQL6BqMozkQRao>LxLy^^PIvtND47klxE3;W0=Pd%9z4|${Cc%pgo)IYuI{L4c2
z@*6FF^2J``X#8k>M!fVFeD;C7n1|vc9^z-cJfi6#*SbXg%zjXp<T#h;=ZE~0FYngj
zV@KwhT=UjJ{Niu_y4Trvo@4OYxAvWRwDq~4_)QNv<Qh-fc*?Us6MZiMdcMcUKK9s`
z^q9xb9{lXa&u{aPT(o%d!#FffE@Vf}vx~iGXuWl5kNu?%>@W4APqU95^zoZ~b$}0g
z@2!p@Jy2cR&q=cr;)BK$%|mkW8BaU#L45iGI_Jr;U-gZo$+1t(m!Y`nSNe=TVVpjQ
z4^0j<k52vU%le>?e(|(k9AMeF_&FEFMSRtvdG)By(dt`WtADih)<OF4vC}x3ooMmW
zx5XD9G@tR|pSa>@2Y!4><2Q~bk6!B`yZL7vT1PH>#LIeo*6|;WPn~$K%yIPr$&pw5
zP(N8vBws!8Uq00_J!tWCj@u97!(M(H7cctsN#p!em(CS5Bo8egX!glJzu04(|7i2%
zqscRlU%tsfCm()@A6jQV5nnx&H}!?Kj$HcqmvLi1lplG^b^0Ye^g{JxJ`rEqC9lp^
z^`Vd9x1Jt;@{7IvQ4j1ulV=?A8;S?{X!FJs(++mh3)P$T_>4pIP`vmlF8r`Bjh7*P
z)`_dQyYE1Gai3!cIgnjtD1PR}E0MqKvkuA!JD|_M+`|&xi($_DT+QcV=tS$p$Nun~
z<vGQ;zG~h$KJhm%Z#j<Em#t6v#=SpGe)IC6F6G~Ign9W<m+l?biD%Y_{;odkL$vkA
zA-(wMFGF(F3$z|Bu24M14-NHM>*YuO%sU@F@51CKhkX0hIIL$kK6=?9ujE7gkpJH2
zimQ3Fe9*%m^XzA@aq^9mlW{gL--+atE6?P!H<3Q$;=o^e$W0_4pM2Lta^wT8pOMEt
zean8**R12OI3>ylKOsJT8&4XadHV3NQ#|nV1HbWlNDq0|(*w;XvP)f<S67f+{7_u&
zPyFmMpMDr8m;dZFu0QH~_^gNQr^mQ+78+Mi>JQCsa>T=U+4BN9xv$j|)KALL^9_5%
zSzPH!dGwmM4jNA+2bS4o9A6^2;#CjrQ_nMgZ@}*=L45w+&hPyAf2{oeQto%OKH=|7
zeLj`)l6L+&x7|zpT`0Z%=k&h$^j7T)|DT25Pgrr6X;)qFy-o%HKDunH{m=c;(uL0#
z@sqE<eIIDTxc2jRd!}U}4*Z4U<$gvkKdtxuBY(Fme)!o}hWe*_4Evyc1;tT(Qa?U^
z`yDEF8+RX+2k!;hNe)_HaIX|k^WvnR7|%K-NB!^{O}_Q!)hRx8=l*6Kzy7Ar=zIE%
zKHz-SUz0YD-@ZiKxArGJ`U@J;57WNn(_irAI^+B`&mQ(B%8PaErH>r)(ar&UkUWSF
ziff`cvX4CaoG0o=-eHcbTl)&)SD(3#-SqPtil=$;vMy=kW%+d8@Kc<~w?5~^gIxCE
zV+T3pk_WAW)<fgsM6Y@F(*rY)#v%Q%%wGJS1IhJ#h=%r|d03{0{G2B*^^&g+ApS(_
ztxIH|`ZsTX7-uJakUsNhc|k+=STFAMnuqi!l4soaU|_CiAHC{FJ(0(5eAdAnXTN!Q
z#78a~KQzu>_dD_u$=5IK59oZ-XT0~}uk|T6{T3JYk#8TflYjbvb@bAMwvQ9ZX9u}Z
zeL(Ru&rURZ$YV#wjX&a&c33BVxgLMoV_(TD6gP6w?1%KA@w1mb;w_&1HV@SUzC3@(
zp)Zjfc9<7$G$c=4jib$rA3F189W3ih>|&3&LUQGS9a&G}OCEjn<44Q0adD#0JUQ&s
zxA2jN4~@_F1?k}jn!WWfpKIMy-Cx~%jl2K4XO|(r8jI|Qja|2(eB$r>cu<_tP`uGF
z?G$&&E_pL<9Xk2cCp51P^%Z>`pZbOP=*LIDzMv1l><8AD#liPS>7zeU9i}{SC*OX-
zuOH}7>^5Jf$2fgx^2@NET<gSByzC3luh72n{$cXSS#MtXc*nwXVV;Bbo#zGbAIN*~
zp-Io~d|IdC*7F8lF#VfL{mW22Kz4d=VV~!Ac6wg-o&k*?pWn$w^N;`HDSkfJ#Alp6
z<anRpx!>nc=IOm`%JAJUJG^t@y^8lT;ydf?(cNBMqjLehSMgqgKk~>9`p6}ZKJxqy
zn)$;vTV>6G13DDYdmiZh3j5{XJoKK5z23+8{c`KYBkh-G{$>8fU7qCs;74xw{t>@#
zThMPmHx~6J{PwH;Xk6d4|J;l81^d&!ht{PYa-n(qTK^$mpJP`dyY+2y7Zsgb<dDxV
zXdb_DexbAP=Dma8;%uJ(;>urg*_Cl2&w8}_vyL7#yXiwi_eAsR)p;Qv_9wmai_dv#
z9If81SNH7pK9n8e0r`W5;^SOr7g}A@tFG-&_VJh9?64lnx4IA)<K8Di<M^EO#@&OG
zUmeJgJVN@#1wX_mpJlZ1FB&t>8DHm*d$0O%Pj>D&hm5Nm`B8WBMvuBOZyc%zcA&|D
z*6Z_V=sdL^I<M82arzVaQ-<QHf1=q@w$Aw@ZsKg;=u7;yZ^+B~QEz4SqA%q-@~xwv
zeaSB`Ul!)`+!A`>BA_4J=ju@X%fGtO*Yz>|53TR%r)d3Co!M{pEmYT#ed=*xQT;CL
z`UPD_pZF~3`Z9iTbS`;6f_84v4~;|X{Qac$(0HQt&Pn{xxV+;xkJe`r{e5eh|BHJ(
z`{|<Uw}13?^x{T-q+a!6o}I?kv$)`sM|~jqQ;+j8$I;}XVd_ESr=Q)B-TIPs&Qtcf
zFL*Au4u3tw&kjgFWG~EkqOE6l?yJ;?<~MtD+&-j7eCe}}y{R|*0@^<Jd6a%&9J(iY
zuJHVR)xq0sKlG603+IIX?>ta9#?j6v{uqbGp}3PnewkkLP+VZjbAEb`A=kO*{_EUz
z?_-bqsde&04_vrY|Le0#{@wHCPdv?|^#inh<aadeL-S=QkB}abCnxufar?yoKd}V=
z-~Yo=76<(bilcs35A9?5)-TC3kGB8pbNkoxu;(K8XSDk@+CF#QkcW?4bw;mxGVc9{
zdeb+ocfV#2J&Ep@;+K8g_W`|!^*-5qTJK5e_xuaRS3bm3ymKBOyY%nO2O5%-G(F_v
z6G!sV&MD`MIA{FP^^iP#&I9qHpWV>7{=^>jfDf(Di95R?K7O(Xnopz$AAiW@55&iQ
z^2k9arhhq)CLfY3&-}v=)g8JF*&*)qq+I^tCyzbJPd~Z#E!ukPQf`i;t)sV|yj+K$
zpY};T6mNPL6vcy|`j2t<Rd&!P?#XW)KRM1@_L)bMYuvgrw9Y)eIqp2yZ}4ZnjN@Ym
zG|yk-XnOoVOy;diyNpBrqw%}{=REt>3ls-_keA3lXdX>JIrO3N^B>Jl>+LUn(Yfb5
z%zL==*}3ffw$DYq|2FP?&GXRr!+kF9`}6JzK1U_T`AI(-vX@<GakqawcjbE?@lhXs
z?@c`Q1Njq2^7VUkqI-76+xZ6N0ph1$p811@@>+)cG*3>-XMgIqPCv9Bl7kknv`<{w
zV}FuwT)m68b;jvSJ*kf$&^jo85T7`tJoZ8MK=wj*Lw?c^#g%?_KtFxt=Kdy!9yEQ4
z_{fE2<L24vK7&@r_Mv;0{fLiW`1zN9<5M5dxmS;NKgK6N`nLE%e&b^ozuAMvXFX&e
z%yIh7%O@nCJ?O@wzNr7|>+%BeLH$-=_PvwFA~}hk?-v&NEgl(nG(KotywT?M4QO7z
z<ik7^KhKfQN%f)b)pJ?@$7dhpIqDpAE_y#kzvpD<slWH~J)9*tXRp@}YE}DQ(iatv
z?4!;*{Z`*V=Q)D54u7J4iO>0@@9L{iU8`I5Y@FTZ*@sS4zv`46<B8_U<tNN>`vvmX
zKC_P8GJnX;_!;*cP98ntYTkPG7$=9n;sNbn@~uZhebK(wH`oK&K`%Y@lb1*@KKa(q
z$Q4ie%tLj64|?zAd69kWkT-hJ*0U?uu|vGccW#%V`#if6t;3fzyPPY|WA=#$KIe#W
zenR%6)t5Ms3zOge7Ef}lht}gmL+jA&faYPD9Q8yFK4?9}#}E609H@U~U9-=5>3ns6
zf!0HE&5NsgOMBUej~wIBIQi^F)0g)e``UR8eeT6S_K91@(Kvgo;}84ve`uV4^w4i!
ze#~1h?)>n5ZhebC&~rKT9>C|A?kCo}=j3|<@5{@^&HG%`=hFPqU)^Kff6dDmzoC3W
z<Kmb7Nxam9ycj3PIK621Ip4>2zxH?f?%%#IZam{_y?c%CVf$Rpy!e~fzx*A(c$0(n
zJ$9dK_#SyZ#1G{S`h5d>64^;VyC8ouZsLRw4Sg?OeX0{Q6n`{6^-I5f0olvXw1d6Y
zmtoeceGAp6I<s%pnZBW)s9$}=I`wTmxllc$$!DK&a-ebcrX2RM%e?j8!?7P)hmRh}
zE_On8z|?C!ezdrVmvOZH@7!`eqsil+zHR@xZ@6#i59IO(4aspY5f}Q=;zob^ZCpIv
z7a{qOUV5Ap;(<>6;z$p>=uwx6?)&29{^h++&Wk%FAF@ll-3Q5M7vwjjA8j4^5Fa_l
z*{d#~yo-<c%D4E6k38^){4&4g*}9ZNpLzPw*3*}I#KS$BJ;u@WpvA{Hw9n{Qw`g%<
z7k}`f@v|EZ`2qP&E+h|az5Ey_hy76flTV(!8@HYx^kiJc%{>1g`R3V2POg^^>(noM
z60L{yvRmH8$2>c%OPU?{(d6NmfAvBhJNaFP<mA153H|>^Ks|_~yofVe{N+nt(77Hz
zIcSI<4cUVxAF7j_*LR`*Z=G>8`yqKyl-LRF8|%ok?=oKG(T~r$;(T)6xi8>P`Dp%;
zpJ<+)?6-d*d+ir|&^)A{T=DSzb@TYy#UFl)13T-Xd$s35-xKrvYTSEs@5%FXXU}EO
z{oHedao_jyoZw#RJV)#Q`n>hf{f=GMxkr)54$ps{Lp<Ml?)4m!wC8Z+J~zNmKD!b<
z2dM+~A<ocq5<UEJ{_%?+^jas6^5eP9IKRoaK52P{;;4T(Kg#M@eexHdJX)XgnQ#6l
zA9>I``{=iip#75Q{ic0m-hM+vdLe!6;2$*JSR|J|ajUmpT{#!<>$i}-Q2(%w9QxTK
zzL_WX;d5`nFAn75gX#;P_2$`$<`<eh@<pC;e&~bxBy{ef6UE2+GJfZab0^Vzd1#(L
z`hDs*?i{mD-5R%!ec}P}TMyX{?F0Vu-#weX@=G6-FY&kk$S>1tK9PSpo^serAA7_P
zt*_{BD_?L&m+S6Wrhwl2LGr{`Jk+y#)5qx}U;oADz1D)FJj#!Fu*W<<&5Miwm(hBB
z=Fy4#wvTc>esb{po){Y9C&#!v8&^mC(x31{{AhiFJn@0*7_E-b^r&<4=)vz^CeHMW
zllzUl8y6@0ho9o|?~3jV{7k>mu(5o}FK@=t@*pqbX&=%D<x!lSE5^myJ(s*PREOEW
z)q`>R7Zll}PwP|ajy(MZT@S@keOu2Tdc~#8f8)++aaF(O^-J>AJN?G#wI8jA{FgWA
z9#e+$0mVmL>4)ww{IBs%J?aP3x0qb@oWV}#nz%yx<V&3Bhy13;I`xCb|3zc!kr&7x
zbz;2C4)ajH#g{z(qRZsvxOk}xc_Lr^=?n5yhWO-%Jo5D!ddXMU)~9^?!8|$8IK9U8
zFM5na>r$V2^$1gr_4beXj30l+p`Kmz<A>ymm;FGVb?jgtI@hDgmoMY=pdo&FA>VVq
zdGgGwWAo(EgHJs9BYx_e9OsvD@)GTP`$yc|qkKPyT>F^bl!s<FI{ji-?z_g))<gEu
zBfj*b#nCwAkM+jup*%qHA$e%^h=%OLUxxByKJ6pNxIEy8=IsNhZnIzTk9_?6F-|@-
zPM_~N@Y{HqK6=cPE8pb7<A+`{y2riE8cvw?vz>R_uw?^2b*`Qv`@HXRFLVByM|<z*
ze(m|odi~Hm<OliA|Lm{UL+hYC$dhq$%GTk7_~e0Hh%ft%dHYm95KrUcUe;IaTl4bc
zyhe*Zd5|7_#wR{HYsa5Yo?9g!pZyKlt<QO`BF}orF8uh^Fa7Q>o_opl{RsNSiyZc#
z_3sG-wi>z3k#nk$9`pF^3;R)A`ENe;um@kFeB+}ZZC)MGuRhf!J~Tf4sVvX*@yB}l
z*kNA2pg8eMe#jL+`6W-EwH_b6?6*%6`Co5bT<Bw0qWsYhv;M@vJo$<2%5k*y;tj16
zCw3XvSHwr$<)g7kzBpKiHc!9!85alq^2i>1;s?o7muT^T_#uAt;t16@{zU8Xs~_i{
zaqHO6Pw}RoKI`y{2iknfC5Qj=s}7)Z%=zdXgx2X_^zfJ6?4S?g$44)INRN5v78;-P
zn0@%vEu>!@tXF63!UxO7*-s98<rgjQN#o;BzK6gsA86nAl^>t0ia+#uUK#rQ(dU3L
z`Q3-im&HT867^yI)_z6jygJs`@mXiycs;D;_w)BK*>}v7%Rli>y=Z*akt6=}K>IY&
zdUc?l^Z|OrL*3~I^ocuu_gbI#dR~=Zb%alT)f2R@pwFe%CB5~~zDu-U)iZm2Ki%ie
zKL3X7b^lhU{DIjY^aFaK{aA*XAN66}I`13#L;s?pK0>ZKR6kICsRRD08}sT(eOiyU
zFVTtgJBQ2W>7iF$qZ7%^aqG<6Z|s%_b!va8OYwj?ZXe~m{V2ckNRNE0KjUTlOWmqZ
z`GnS^6YUS<>KU>RZJ(-F<KhL)^Anx>RXm`1a*V4t^YTGHKK80pddW-LI3!0Nj9X{l
zq~GF#AF=~YK6#1ckk3DQlOL_V)tmZLmx<yiuGT^EwT|7^nOB$i>?3lZb>!eTuKx62
z{;OZK_2Meu=Jj*=6CY?jy^x-?%eZy)l20#WkMC#MpZ2r;MvwT&7c>vqi57qBjc1)%
zm;Upo4An)}9li9S*};A^{p6?*e2|=!Yy9hqo{PLM@?J>2<$X2JO|)~(xu;)<t9+HA
z^^hF2b@FGt4Dr`P_S461=XPT1SKsQ;K1j50a$G;Q@9`zt_t_VW>u*q>(-)xr;GXK9
z4Sg@tyzlY2NBe#d^mlmvZV&o9L4S8?zh}I}^WP2CLz!LrOwRK&$I<Euy`ZSh7Ia<m
zJNM-y=NFY$=P5^>8%L*K_JzK#&egT`iyPU=F7e|BJB>rn9r__Ti`#yAPrc4beMkQI
zh2OcrFXM~qPrgrDLjP+c;5=~ulvjP*{&U`_8|RBUht3uA-q)ewovquQy5ia$3-4>3
z58m_m{=Rbv|LBH>`&aF>OyU1U%X_cBz#jgfJx8F`H^0#4q4=Q56`!29f5Z)K9Xj>e
zXVAR3(U+KZ*@yZQea4~jMEY}_{W&jg&OP$!&p7BC?AAa2wbi~v`l3c4>p}kHSH0S2
zP#ws#^UuBAJ<2^@{irwhaqs2a%dA&-<U?}&?u5RgFV*`VI{k}^&Q*50|D_-NPe0i`
zcF1jm_G`0Bp<mWReL_6y<zJtXXZ@Ak@=LD#i;w<qT%3*TlMo-hX!Rj~;)j-Z=p0dp
z`jI}R{`h5{9nKB*lIMMAj`Kg|u}2;uzr;^H<0lWFb6MQP$9_t^*71itXq-Ij%|msk
zJ|TP2;wj$p!fx{Dh4|$KAHS?uZ{j0A=J7j6)VFw}#asO7%W><}8M)%bUVfXWhyVEL
zQ)m2^C-Ttd@r$><Xn)aXUjJYZzcN0?$xm8*z1O4%l5bxo$`^j|6c6jgMI6~<9a^5m
z72-p)L*9(jhqeylhslSA<as~l|IHu|@*ApOXuJ&N)%OVL<!8ptJUb!z@|^3$xeUqW
zS3RtEu6$YZ;=q2id>Gf)py%|m_ny9gM~;0e&h~Z2UtH^<yx@~pST>*h_4v$}p?VZI
zG=A~&{uGJ}I?ok;*ni~Xmlx~si>G)S=Xav-J?fwM^%wmKk}r?!;SW3QM|MKv(Dy*u
zC2saNy*aO+LUubh*hd~Y5FfeZK;!I&^2>go3*d+54_s@Ni*~r{v#Ry%#P`6t12!AB
zYx9P*)4cqeCzl*@*_GqTkB0JW9ldCA#ILTzNuJR7*bDUw^^A}GklsXfl}JCi>~~*K
zukKgsG?83(v6~&_<KwUSME)eE9P-Ngf39yVO>SfP@nu|+cArug)`^?=iw`<ceC<c5
z50%l@xxdM4%D2De-#X_`>T&Plk9_irzbS_v^310_#v%FQL@zWC$xA!<Ctm!puf@qY
zWOvTnH$LYQU+DRW9DdVZMzhB{deH9I<R_+n^Va9QJVSi!%W-jFr#$h?I6c0{=6OrJ
zJP*@HE`6Sxts|En_R$ZG<Adr)U8q<4AD=qGCw`E9?DAfNKjz89XT5sjw>YueJUhe%
zAG!QRL;4}P^ue+|`gIu>A9>bK#9h4@N6U{ml7mm3$`5|^=3Fx_Kk`Sv_eA98`GRJb
z`-VPf9a^2^XCK4|*{h#I^WugN(gX3M)iL{W9X;e*hab&uw0Z0JDZcWQ`4RW83VKd*
zzwlh?UMrsBEFS8OpXy9J#7Uph&%D?6KF9BI=+8c9vma8fey<O*OFoRtr|*UMT`>KR
zed>e1?#t#OKiTK=Kj)FYoiuyt<rh22rN=xPiZ>b`zs>U-pZMXEKY6lF-QyD%NDe>9
zQ8%7vJ(qe9K`uGI2d&>gcC(8f<MhcVJMo)Wuk3~NL;D%hLmreLevnU&J^{(&m;S78
zu@fJ8`1D=!a$daILl0EP5TEbOkt<K;6Zywp>)45(ef+`CU-M{u&^Y8@&Hm5-&V+yB
zgHN65TlO#etV2U`%<~VjM}LIuvJT(kzUQ!0ywxu>u21B>Uw^QVlFt1{F1|$j1s}{h
zLPPt@K842VL9^F$3;*z$XFu9JyX4b*9(klE`Tcz|yCHe%+`2?{LoX~F&w2G`oZae~
z9gyGR3C$<+M;sFICm))>^r|;#T%ADvL;I4S&^#;~H=oEKeTUu9xH#)`@@l;}@TUyb
zrFr!z9_~f%FKFxVi>rC}2DE#Xb&34o7k~7RtRMAn9$Jq-<r-I)&^jnDX!6Auk_Yjj
ztw%$Akez6JiR#Jy#QSl2*v&pj{uhnxSybewI&v<^v$~fr=cxMuI+0z*#Rr-v2c7fw
zJ3GZuo!US87kkMg-?@*5{I$>J&3bi_^Y%HtXz1Py*(DzTDX1>oxA?oT{)w}FwXkx&
zs_XQZv|s(^_!mXn4>0AjE9d3kez8xDryO+Iejvwsb%mxMpSbwEDUrS}3h9URqal50
zeL+9bZ}kuTHW8nGXT3fJ&7<+zPxO<Qe5qGol_&QK>&bIJG45XC{_Q^F9zw2h{u`&y
zJ%yd*(+A1L&kudw=e%fr#rsb0WvoMcFN4qfr~7&hYq)=8lLqfIe6L2H*(>huvHBup
zKNL4Kzt|x@=F?7m-ed8XKK{uYI#HaVdHI(wenEQpZykHBr{8-T^7v=I%zkpeD!90x
z`<LkZ_lkghr0>`t>bzc`)(`CmeFEBl<k=VY6Vz|@Nwj@Qp8bjs4V`O^Mdzb_=zakC
z1;q#Q2Q7|`#m3tCW%2p{6aC$(zZdd%snFl6LVwrl^9K2GP8s(ez`39v)McJ;{%$ru
zcfm)l^9HIT^3{)ffc5ON4zgbz;&aYH{Ai!2IDgcU&yVnXu7o-6J_+%uJNwi<*zW+U
zPbjYTrFfXP-o9{eAP;TdyVsQ=JO5S49{bhvIP^T8^XwHreBvn1_OWsGDvt7`U+Hu5
zEw17W)gR3Ilppsj^Jw<7i#~Pg-okIlpG5xiOP$ioPy0pO#DiXO;3s?S8~)mFIc^<3
z^3d!OXMLMJ;(_K5JIJ9|KQRyGjUIBuIcf2wmmK=o<z9}?JfO+ppZw6HzR}h}cFUjr
zsxDyqiOzZB>N4l$O&y@|ix)X%ei=_Bhg|U$NB%+mMtsoX=()XK9K{nqdHMsyrytP=
zjYFSj>Z_j9j6?I#Jpc`j>)-T1?<d&FuEwG`C+b7uYJbX~I+p*e8+lcy>PMYerw<rc
z-{e8_>Ll%8zdXu6KI7_$9`_1;%Q(G|UdTUw7`L8YcClZc%zJ9cb>1ggM<09SCy{({
z)!+3)dgW7}CttqYbLiuTI2mW3dLV}$^4Q5P^NHd>E>!30-o8jwuk6vM*(+cCM$?=1
zfo2D}_?;W-TU^asN1yNKhztM8MO)8qeM%hhL-kG{JLH2r>*<kK{AeiNP`vf0%!_f~
zf3S}I^g{7xr#Q-+eG9FlM?ckP(>{FSgvO`OL;i>d`_Pb`(7w}m$%E$6;*!WNe##p?
z;v^pGQ@+%{c<XQK4|+~<-imh_k}p2wJC{6HuwVYgo1gqQ?(b#LkR1H(SKg~a`4@M3
zj6?F+<#SZX4yc~UV~0GF!(M!7`sjz`k|Un-DewFd_j*XLK7eM2akTugBlCw3TF+kb
z&`;RMZgS~?#`z1ahdFNl+JE}4{fTxC8`lq=gU%o0_7mEE(|5_SZ>+aZ%v%TLk=@Wd
zKkQ%giQ<uQHV(_;Wu9Div4h|IfEh=A!<;8qTs(*2gXD|5ycvi1$<fEWPx8F#y_EcV
zZ)H8?r+Cw292&Qdeb6{Q@fCmjq+XqiKcrvX$rn9n`mD=w^Eocw?j7>PANjLRy<0~v
zWG_1FQ#{-wbDm$$ANh37U9#clH{*dv5A4%#-Ja|B?B28KSp#<-*t6OH{@br(zh(M$
z>DlCrf&EtaUwXZsO?x)oan1o}4(vDh?5#&tnm3s^YU1rr?EX%*_Zk~L^T@+vtC#M5
zUhDnmkE+hu=)k>pU1xZWUUP*WlMcS>^=j8Ae)-zNhrC_g`o<0Wz4Xq+>a8!#e*TCb
zyj9KlW7;-9ec;nA8s={Q%VS@><R7*99eZxR$tOb|s;#pglH2~QQQzFQ>ATfK?&|u=
zb+^7y!)N!N*z>j1KderE|CY0l+-FL)!;K@?>b>vC>U~cfc4L<rZ&t`zP@FNY*ER2)
z@oc3Gk3Q?7JJ)aXQRToM?^GvzZ${;?U2eF1(5CNHuGwnYzn%A!w=2i3zVyJ;uYa(@
z{^!q{@vRF-JXP)9{=vsO?*4k^-OE3^?}E$5R(|)yGJUQ)>K~QKeg1su_3uxsRFB)_
z&|6=gS;O=H@b_)*ee|=+iTCbw*0lbgRYpv>e&Yu^&8tiwbm8rrKQOQIXy@*~nK5*3
z<)HuC@tV_4Y+9WE!~3pY@6r~<)lc5_%Hzj1E6)3Pi<O$M-K_Z5DxZ%3?LG4=hn)ZS
zb^bEBd9lKePWk4;`^>L&n=$OBizke$+5OZSAMJ9(V-qWvZ?N;kF+D!1%xbsS$ydJe
zUd_K-S`EALnoB>az%l1`e`DOm(<<#x8up``I)6}Ux<%KY_x-{2%5kUlJ!0>xrdHIC
zx`HphG^FbfubE!Gve(bvoAu{u)nx`<)@)j<snrKQJaNEt+fJzdaooTa=AHaf<(d=v
zE;Vwu+bd^$xO%b2<1f|fO+K@p<o}`N&*^i<pvP);weEp?|K;@`-dVl=kYByF<8v2P
z$RBdYseica%y(<~l5crSlt*}0vyOX=-0*`c`^s>Gji0>cshi%d!gDV;VbAl&JW^5r
z<ZUtfnM2<{{@&`inSG~R+2OhB4JUT|>!VAJuldbB{>|yyp--n}|5V$b<U{*5?Gg`q
zT6TMD!<lctSgFbTT&{SkZ*>h<x}xWt=BtkUKl_9n@)OygsQw_miRyOF8a*#vaoaJq
z{bXOM$30s2?QqnSb1LUf9lqH?ug$2~2kO&48h^ytQKPnhsxqSI^6gq3)x22mlztBm
z>(a8={_WAXuCsmf;+^WsQ;+XEuhQz!HV4mcH@~*;>@)l7+TJ(4-Jy5u!oILCU+i^T
zxBlzRs_nNfxBPyU!;hO?+t>WP{hI1slTMvktH+^lR`%QXgHLMqlZ$_+*Z#ceExkXi
z;Im)#A@waD=;t<nq|=o9C)euHK0WHfYcG0w#(UMCKb^GWvm-vMZg=mrtxuacqq^aN
z_uTiZ-+Wv(?|e{au=?4S`}}r?>D6^xw|;v5)-!7Rdarf+EOp+xbE{)IZT#F_?WWY~
zS$+Lt>e16a{rM-gb3tFf<=hwlbXcd!wZ1M->g)PlnvH4u@TiJ9QrG5<C)!u)7wWT6
zJ(Z#T`$ChWJ3qJdV-=`x$paKOsDCB0rwr4dq-**=@1F}wm-%a4zROUa%FqkEdZ@q3
zCwlD7-yZeZC(~>7sK4nm>e+d-!qaE2c+RVHs`5pS`f)DkgYwg2_my@!`HD|#_a5hy
z`;9tvep$y}eciaeDQ@Cve>&InJM{$R(Y*mq{p|AVKYC$S1$H{_idLg<oLoCMog0bb
zzs2!KU)b)H&#HG;|9V@?v9qds{NdpF|FzZ3+CIK*ou{sQXy&x)i2k>%d)Dt~*Y@?$
zK0n^|yL~5DZ@%-h5x;o-gW7qnzle)<;wgU{_c-gf$J{=(mdEK`FB{zEyK}0E@|7sx
z;w|su>K-Ug#^qTaoyU1zEWhWm$Bo~ySut%^^8shA+@e_b^!cN{`^@Z0kI&}+V3q5p
zRKCB?*{AH;Yf3Hdx!$=0x4U?sZ#H?QNdegh@s*)EPjt>C-o5GSSG{mg>jFCW;KtQc
z&be`uPiyzN2gmonZi8QZTp7GyrOB~f=GMmD2i-Tf9rWq82b}Y1<rn??H(&3qPiy|G
zKj*Bt>$hcSUqI&@bbdnmp?=NIt@>?x_~5lCR5tkb{f~U>!4GQ49(8a;zkx%Rx#^9H
za}mAo&>QE@n_R7(d+Lt;*$3=r{Y71=NBdTv&iv{B_5(d-Xy2wBeIoC-{B=(-Z{M(s
z-Q?<@_&vY$y?yYr#~v`G_WU9b`kg$ekE?#O$<MBC-lQ0`@u}as{f>_+?o*zh)RjIc
z&#zy&XV<&teO!T_TgI<>-cwIL*}O<Ram6S0+WN(&h5CKG!=T+SXx*l^f9+!^j`aTI
z%m>$*v%!d3{oCKIhM!PucW8@3{kVtmQ~jxb``dnWucTKW7Jv6Pbt^9T$YT$5uW)aK
zKY!`inQu&=Qo9!#x1Z4VDWvy~tJb?{z%$K?wI{TB{@ro^e4cyk(Uo65=;D^O=STcd
z9~-=E_dWkHsYQ{f4xxSt<pZj_dg#1MB*$|pf8@=65NCSi#d+*KhkJrL_Z*_G?LU0>
zy}poWzrlSsd#J^AV>=c<oczkEuk5r`(dOVIdS1L++hWJ3cGz;z4NDa(+%)C0HjlL^
zW?Ve`{1MIC{vUjL?~bb;d1$BFzEFSmr~KO=_MPV#^zH|Z9N(~B)7rkV&)zw#&5j>z
z(W-#%Kb{NpFXQ^+{K0GgWTjmuSKRmYo1Mnrv(olId#@_qzq$16-#+rz^_9*K|82`R
zhJRGmXXK%{ut%?796h3X`t#rDQS3La>fGM_!EVz>-10>2xyd;qpYlt;zUKKD+UNGG
zeSPJy?WT9@KC=StKbUogX7>R{t^VerAG}dzSKd4AW2oPip}J1Y`YNOC)5fCr67Efq
zKT!X&zx6fehI_4f?^m2R>H@moq16R6Za+FN)syFA`S(0!U!v88b67olj`g03z3N(h
zx#zm~s(<$c<LqV+JJg?iS<fH+!M&M1_}GPy9QPXc5a*NUWJo_05Axl|`E8%m2idD%
z(n}9May;LA{_vd8<+~RwUHSOa>Iv8M>3{s(Pa2KnTDQi`AAP*jmJ_SqAOCvSZ;ajb
z50h%|A-s2xclPTK?tT1#>g$x3fBL%)&(E!rYrT97+kVdXeznXi)h&<QcIPdZ8(Vcw
z*cbML^AnDHrSsO;?%cG{kDS-;P0;&c@00L9di3C4KU%$c?LDw_&pD{Cm7#N>44u~t
ziuQ~9iTZSoLh;cjFKl}8vMY}H=lgZz&RJ+3Ii6=m{_4<{UBB~QMIT~Ei<dk1zGJP~
zl{>e%Z|Eim&aIH^y#+q!jB`xiq=!GwSL^H}eMbKI&2F^5sGpww%h$WFK4og<v@2Q;
zn7GypmB&A5vQ@{aZ&&nR&nxsgcld2w{=d;{na3}mJ-IS`-sNY!J+Dc@Uv;Ri`**u^
z=v(g(tL$*Y{om+v!T1XM#gQF8uOOfOJ|DqH4!^|3y?gH&>wehkkXI|tf9O0BfAZFu
zJhnQx>5Wx=U*B@x>VHt574I7T`TIlusULBZXY<z4tFJk~%&Q0eLcXlSZ#?DbWBAk~
z8ozVOyn1wx$a6+rdLJd<DHk6U-`f55^SFtxI+EAg_~)OupcCzD>(mcaFY?CkMEx4-
zx6T_hJ?<6kp=aXruRpu)L6d6lljWz!3&pnQZuGa>J@uEx?MwgZ@TY5^L%NqbX9iXN
z*lvfx?^m6lz5n{ynoF&)Z|(CO@pSH3C+~^!)mU^6xpz1R*+ag#ibK{XK6N5~^-vxb
z72R*;0ov#4)_X<wFz0sqEAC||5BWKUdz;TM#Mij<-hI#g0}a{D&U&biB#OWIq#XAu
znDg#EF!RJe$Zn{w!K_Pl1Njf>g=wF-LUus=2I>d;w&%gdqJ83f1XEADZs1>k{Atza
z29FNf=%6Pznq3_{VcRJ^&YM+xPp<BoR@dBi!+WOG#@&xSrw-WmcN;FXNt1@9?|$UM
ziTgBZP+!)o7wg0Y4bSWR;zk2DpH)TkUtGmoJ*pG@YxnO|9kR#N+I`qM_j~vD33t5t
zV*8~=S6<%g@U~5#f4TN~x%-B`DqcR%7f1ErxzYK_9^d0&FSKvvnLM9E@?XAlAB!8k
z&MWs8_Z;V%ey5+SZ|Ask0lz-z{=pw{vya@L^eguk_dn;Jdylw!UiDt8vFQB{^c(@5
zr|vWEhwg3qi~FWHU9#%zkDorfNrUGy`(oR!^ETb2*(bH{-KYa7|4((ByYV)6kFTA3
z&LMTHj^uyJw?{S{ea@KL_mAu!eMY=@X?Ay){qB6V_Pq;vRA<fw_Y&`Q-GALXy<c&!
zcTbgX_fPXaHxWOy_nf{T;C}6M9Pg9Ndv5djlIJ(}$XC~gwx6-;N$*$eN6!hhI{p0n
zM(Rtw)b4$s&%4*ySMDp~Ef4ijUs3Ps;N$kAp4g+?N44i_{m^--Kj?Sfs~o)FxQClv
zG_#s$AG_bu3%$RD-cO=^4&d`T`@s7;_Z{^C<y#!(*S*^N6X%Tg=K6&Gpnj}VKlVQ)
zZ`0k6y=2O2b86pX;fMX{^9Ju_J(u&_eP18h|K<;eT($kwiaN9J$@O`l_qg6Kkl%ax
zb^fu&v;X}34DG#xdQ#Uuml5a0P2W1{%!WZTs?d4we&QVR9>x1oeaHEvzZiEus0-(Z
zd&BEj95lDrM-yt_D^uU{<ozLja?m~p?>B7Ho%jFuC)E?T*!}p&-<?_Yc{u&fSN5rY
zeD()9>YYA)(f!lC)p<=G|IF)~o^R}L=aA3k*(u)S;1_>=Tc3rioxO3dTkdJn;JqWe
z>~H<nearc-Z$tOhFB+Zm^6Pn+UVfU_KcR7bho9n+s6L?kh&Z@E&?}C{^=EebTv5NU
zU-c31shnGo9PhQ{SD#DN_aHs&a9`0+_*D<RH(`(X^Ot;mlYZmwakcn;er}0}`v_!T
zqI20j%KJ_I+4-uk*F*a2p*+KS@ydDqD6yX2oELZKocBFR=Y#vFy3^0S&;8`q-_P6k
z#-_!4Lw`4-^Brx9ZY%Bc#B%f6);<r^@7;U6ul70O=DUqK=9~jt7N_lS*O8;HX;&=O
zvfplRZPlUp>Au@+H}<@a#jB?@|HhTuEM2&-`drg`=eK**$#*|=z<Cd}D&*BV=cjYh
zxmY&td&_w*V~=&>;C$b7=!AnRr!^_a=ePH*K4)bo`+Uxvc<@`JpWblg&uZWM_B|D!
zFY?>>CdhT(TJL`I{E+)rU$dB25eIqoy@8W{T^V;kj~0b{hWiA6o&W9&_{b%PKJ_6U
z<oUk2`tUshadf}npM3KNpZxPz9LP8R&F_D>;h(Q<R{K7c&p+I2+_&9(e4c*KfXmmM
z{>`OopRc@h>4g`3`~8;1^|P;@wEK4LYu{T~arz;re`k}{wa?jKI=szRzaG)KIQ_Eq
zXU<)&L-F8Qm-fDE<xYkCxL0j;&a>yPu|>yPz3g%SP6xhsW%EKj><f0-U-ro=Z@jh4
z;EOsGFTdM<`JMJ`Q(Sl3TKjMGa@%6UZYR9C*PoWI?L*%unbm#DWvfnFx@LdwGj^$C
zwC_2K+pa4-+2r{4%hvX5-)kS;blDF&6+Tb$_XUR?v&DTY?AWgMy|Vm1g6}Kref8tB
zcAK?St&fy_uR&h^cGDmFzW#Ch+V>yC<)XFL+u*Wk?P_v-PvWF4@B8Dm8+0nv<M!YE
zaHqSEoKqRQ&f4#HKXF!N-uR<-9&%vkS{%hyow?822Y)>Kdu#V<-@1VEBwq5~V!-@|
z)|}n67AJClwbOP-fA5Soh0oc1zubM_`P6!~Pyg0`T&rTANn4$AV*jNJ-*?hiJQset
z?I~?<JHJ)2!wJLhKm2=bYTrk{aOAE3eblLK3w^f#>)YJ&Y~NP3I<(Gy=8x~Et2^~|
z#AEmVV9Q7USwHGm9(}K5rL8Bl-u33Tg>w}ibNpu0UaE91R{rCD=ly22_QlM%k6CW!
zvF!?ZuwU(K`$2!PukF*Hbl>g!J^$LWwh#4*b>BVW_LI9eDLmh(OXK>J_ZZ^u+}HO{
z=(^{L|9!@ciazZgpbvV^_WbRAj(dvFx7=r(OYS53ulFkKQlIS8H+>INJj6kNl}EHb
zZ$GOaaoenI-^qQaH!INkq;uN7x8DBu9)jJz=Za4sR^RH`-zBkc$nY*Vt$bzsf?jz?
zyU#iY+)u<+-2Gh?|EzPKvO_!|JNYBd%XGe|^E!W=Ua`N-i<5H?AN}-_1HDher!Li(
z`-XY<3U#P1JTL1<kbUY#9XLn%tIoX-(1)Bi?ib{`XYkWLW{2}k9oiS{fa0Zp+Q<C;
zANKA&*zT&j1NeO<Ave4(Ad#Rl1Y?58L`0)d%lQdWArK5cA{c6dla7F(v>2FRs}{oG
zqeKM?)S6gmeG-C2p@@WYh!G5r0n&uQ_#hY+i>NS!5n>+wqxW-|`OfV5os$PTEq|@~
z{eJtg_IiBR+H3D~uYUZKeAqkO_WK#_tlN0v9R5>o?9ROOFApI%>(t(OzO!?BB3F3?
zs(zVuEOmT!E7X3nMD06K`$`n<*w5OpC$+C;U+fefsVk~;!8!Xj`x5&KbuoOX+o1SS
zkHDvWpZz4NE}`yWKKnlVWaqQ$7V0X-#jpLKx-a~-uh-6gvs3E)z<kz=Z}kBDs>9HS
zcKB9jvQGPT?eOWl_TF2Mez{Ln|Dg~4jpuyBd-_nXF`ju)df*RmS3HF)=JlRlt)Km|
zqonMXJsOAoiqqmF+*1e9PyArt?q9h-!%nQry)*SY)Ohqu9>(|H`8oMnH@($Sa<Lxs
zI$vV1=0nxr)qCKPI+pMBjeqv0{zpI7?OYO0@oUt5H}l#5@)Q0CH^_yY@F8!4NBXf>
z{m4^)@*pR59plh1{U<)CcfmL9ohv%;m9Lvmol1Z6sn6k89nAN&^Vd9j-^0Bv@^Mb(
z9E3ls8ySy%n%_FLV^{R@(_2q}?(`GaHt}g(R2&kY;Fs~})BN;@>aTuee&=fDp)dGC
z9^|PVIlv9;q<?nnJ-kRe@Lv9CJoPX5V1DbvKR%uJuowPFkL-p1#W8hm_teQp97OrK
zeAPV0mq*FBP<d8T^%&GRN%37r?SIGx&X6}d(9XV&e9h~boaK}Jj9+@cqbdIDD1PfG
zzeY>#`JFfhXXOJZyhh~}sPhzfEdSxR{G+Xx;rn%-`Ds%AU83||qVk(<rR0IaEtLF9
zv}C{RUmjngaHmAsb&0ay5{08BS|>ODQlg#uNy$Ftd8pqv!ZY;?_YCAYzN^o=CoMk{
zw^8Q^?jyLT!QS2LcP~W#!4B9hJK)FG1$TWf*{k>BQP*_1&Lj@$FCK|&>O%0}eJS~y
z^|;649N2lU^J#gi`xxv+Jd?ko)<-Vvl3v6Y?c^Wg9X-Ky=lk~E>K*LS{sRAgw}n5r
zLO$Z1IBHyYmee@-OKScSbw7-rN>tv7PxUQwBriB6PcmNGDZcBde56xKp5}!s;)w6~
zL*0uIN1W5Uci{I!@?HKY9~EcZdvM>&y$$y=__Mg>*}R_RXP(_-AV=TXL*AR_2mIdt
z#{NZ|q&Io6ajjRKhJW%G_wV2r{F86PpSB+Uw&%wO?f-a#--}D~Pq<|Ng<tPW{rIPQ
zitxq$+B5vdzdRZxKlTW}-D{ISsBe&mXa42BIz5rMb&<1s@o-7LYkv62pNu0ez)kCc
zH*k@C@^AX2_q0EDf|3il;nO@QJ>x?=zdK31WH<0rdvb;+^u|uK&-us!E|I_A!HC!L
zQ1_(x6MN?`?7;mW^7oxTiod=a7rv06c+Gyu7r*LD+QT{Pg6H-}+OrSi8js)L!~FQP
z4sldG(vMwuW{>I*`lF+c`^n_J-tzMXJ=@>d7s`*^r%}gNhwxn;T)j%YSp2ik^Sdee
zs`l~)ex(k;F5!|mKprR@B1iTj?<IG7ft&0h{axJAPMjf6xCIwb{%c>7ls%x<`KPA#
zZ{+M8L_OVkqx&B6@-HpB_~awT4Yej5e8<iEO&w}^{+nszrape*P-{xho6cQ()llp9
z51sqqt#>^CeEk)h_Sty$P%EkS4;=mYr5}6AUM+O#@^{~N)|Gp-PX1iq0pECfxAyy!
zmri(MR%5T$SMLA#{MUT^c`6@-H{y_IehRPHzkE=><eZ3o(WiWsU9l5>=NZnr2Mo95
zlW>lmJ6B_m{L|khbAR$Z&y3mU{bNSA;46DHKRoi@I*rR-;g5dWTPHt)S8&{W?a5a^
ze33IMpD>R6L!C;U2>$pxD(WlFL*?1d^WX~{Qn!ag_RG!>ozHm>r=0&9$GGlIpzeR;
zS3BouN%4)g+YjG=HXo{9gu+V{zo`AYa|$%;XZh?r+`9NTxv0n47sCPj8}dcT)pyk2
zPif=H@Zaxrzrj2xIXL$~Gv32*`L}aKaZr2^PuV@3cJEmJul~<Z<!$1vJkh-TQGSLG
z<BC_t!Jl^WIq?O*{=OK$w2$V8?g=>0SD)}+oJQG+e2PEGlj+U5FMPL8bbc&e%Lmw_
ze$K7z!^L0uf;a<*$dg{=De^P=M#)S5LSFQ1T<yu(dwMbtN-o;52m3SgunY6a8`&Ga
zKYRX<4nE?f4NW+2KT5C0_1@o^#IHP-9oeU{H_!OBUUEVC6Ff!XwY-$Ryhp{2r0~Xl
z9(YE-D0%xm06E!5rQYq2G7nTI;t%p(d82r&KfC5H@;USIH}d7r^h-bVEZ&)y-aWG;
z>yjrLpPu-!yn$TlK^|y5@`L1Ge!)(h$KXSpmG6kt`uXm?@vT!H*zP}kFWs~L;<$b#
zYW?&N7u8izb}PQh53El+`)ztqccg#zgR%qiQKwb+VDH^hc8)LexqmJXW4H8*Z@5Ap
z_O0wuJ9RAhK|b)qvpAJ@%>VK|`z9Crr<9}hkt2OrmwF0&;(yj<AIomkCG7W6cE^s4
zYd(5M@tJ;25A@03;Jo#k*Ld{74*8$`8u^On)`MUACMVz78G9xl@zJw9SiWT+^OfQH
zoA@rjh8xDWPZ8hM)$p%=1E<t8tP5_D5Bbq6`yfyK)c3SUtxvpAFW`Ud2QJw!@n`vm
zI;MLXsD5x0|K_I;cn-hh$M6fKFZ(X>#dsxZJXCzKuS3bNMAe_I-@1~juaOu1+Apv_
z<In^C<ooW$+Q%d%U-rT;=#3x1De=gAc{Y5K56H8{1MSs2eK()HO5WqW{U-Y`j`p7U
zHGRsL*gyHgA9^N7&*HUy+S^CNJNgy(%#UhE9{Bg39UBMV_z|z^f!)F}{Gi4sU;7Xg
zK9dW7DN%8+TMEzAi`k`m3;Dt~xW<1x!$Wup@7M#J)z34^uGt5<^ILMGucYb^aNGRw
z%{sleF8f;4y6uzY)B5vAeyct~KGw(o)mh+xyq=udiT)`3BTw>0$%h@f|73o+ikdIa
z2gDKI*(-gJKf7d)+QB>R;5&bzH~zs-`M3G)@5sw}hj`D=<Z;F|4k|C=hxjg0`I~b{
zc4c2{T`x3+r{tZuj6e9n&*TB{!+Pv{$jLdI{DZvNqrA+x-jgrtJzSLM@nd-)D(;9U
z@@jdr{Tus{_oz$QufcI~kUhxD*$er}^F5oNT*-;t@Gl?dXU6p`p7THQl?U=a@*y|x
z;SE1v-~3-Z<5%gYzH2WovVZv?eUlTva{eM;=GXimp5*-efu7)<aqxv-c&R-c+&uC7
z?|a85`&!0_f8@gc>`%y1zC~Z=Ns1r(FrIaa7v_gQ<O#RcN!78%0rfmooKRo0Zgn{E
zL>}!~J8?%{$nQvuk6-6@epll6OzPZzSK{|1e!t@PQR=+j`@M^JDc{DIxaqs!TjAH=
z6(K+SC3X$3_z(LhUwUJ|_9x`b|Jkp7l(=s^<MT`T5S)>R!W(r#_^+PlykCCn{)zl6
zaRVOtyBz9=;*R~kc!O$JM>8+Y{1N|2@ej|El7si~73E*>o8QB^r2LJ)%2(ht{DTAX
z0{a#7$Zznczj%NTe5=#w?-`DXZ}zF|35A#VLXF3+*eUtLSMgO`vF}3hOD^iCp5di<
zDo&Y~KGav>Ih+#5)lcmI;gEQ)ek0C`d&Y-@<RPAVhQIt!e}0c2`HOYpo4n+E#>Kz!
zQF*%c!V_|l|F9GOCNCAI*#jK%Y<=udo?ssGV2||4f911qh27}Ke#nFUiu2?{@B9@m
z@q2c{-^@b};+psD-go1Z7e4TXvQzEZB|XtMxw3ct$yeUQPw-=2)O^NKZ-KMsMd`u#
zC>%iHep3Ex-&v&b|ElCAb(Fm8D1SkdPw!FwhRU}})ZbMRzv;(*((i(uGs?&KDL<x9
zc4uAkUFVMSF?kDr7N5my^-|;G(|6Q9l^x1M<t6!@-NSh}$8PusJ7m9b5@jE#_NZ~t
zlKj1g14;Qk+-1k`o4omn_<|4Ay2WXEEo%Lq^~*TyJAc&AvwfZS<cEL!q2{qp?e#aG
za|AT~!1~(q8UBB;;)8im|KHF3V|f}omS=e{?+_2<d2kE<@*nX9{=;ebFTRNj+L;%A
znvcBTmh)!M_%2cM!WVwKrS1LJ@ZZyBzw&c=ggBfyC{CmB#54cp_i#Zw&+y4U!~CA@
zH}I<;{>)>aBoE-v;)l2<U-ey{Wc}<E4#8tMsGr}T;?w;2wEu&X+TjDA)<G`(g<Q;M
zzhk}P7<@Iqd5o_eO1_>^b|vqIx9UssCEw*U{E#2O2RKrq<lQZ0hvb05o1I0?%bwYT
z`Nco`S^9!E`dcr&xBrx<!993VqV)A5{a*?2SNRf3&g?a*IHx^*v1^pR*|YgjxRO*I
zqFc(}P<cTeg<t5|n;*V&`URWYe=pg3jDvsisf`=M`LMbQ`JiyXzLnjIoAy)uQ5<w0
zDUKS)K1}{#T>C~ir0!zBAkNCqtVi7${@ef3KR@KBa1ai|arrbnR_F4a|HFCj#VPoW
zlCL-kw|%EKd5}2Ie#qIn$k(&+jca}S!)f{;XZZ}dqU3E}@3oij(3AbF{6*eH{^Y1`
zO5Wmp%G<Mj2OJeQ*_r&G+{A79HM?M6&N1CL_j@Pt5nt@zc>LSE@&|d6`VzTXx4HoS
z<nQ(g@<RKuZKZjygPz0@_=P`qL(lZ<JG+5H_~&o%Kz>fnsQL7_Ui`2Ra%Hdl0fk%q
z7S710*cJO%M|R(vztJ12J$pp)LBIAZ^enDezjYaxow74_#7^Z`&Sm5c_NDleZ@5<^
z?_{rVhThySQ-7vMI6~gu(|c05%8%`n)NSmSzP90<C7=CiPs{ke!vXbI?Y*bBPO13I
zuEa0>@z2ka%B$h4@yG#wqc7%FJ68hm+rHO%lXD<`PQP#f?ufgd#Rq-^$Ke7z=P%+k
zKHSrGPaodvuOGWncS7L{`@|o=hHvtWokj7DT3@%6zsN68I0Ik!H~;r6|B_GZkIE<N
zsNXv{k9GfszJ2$*E$62G&Y$n}>O9vusdH8D-J@rRo{i&o8L0PupMl~Bzv6^^#&`C?
z9@vfiOuU8L;wBsuhu9r?(v$Y=R{Sy_J@Y?t+_U*hv<@Z2QT`{M@fUsycf~jU=A7Ag
z@+3FovWF6dA4&1oEyYi#)Vh<((>tZ`s8dRh+e+Ol;6Lm)DZO<|x0SE#CvPiJ=M+i#
z?~bN*ej)Ep%J1u_yak@BBg<FiL-uF(-S$KBQ+5weJZmRUb^gPC?KAi(KTv;k&gA!u
z?8x8U<@fA}-LPx8D=y39*&&=Yk9w=$LHeB~y|N4Qpzgz%$GsTi>+koN)`4&9w@&y&
z4&ttLiDTlkJV)LF$Kf)0vs?X*Z{4UkhVK%tkehXA2UpX6?Q7&??85rsaN?_H`y=%i
zc5C0O4y4XwpDu2Ck3Vt~FTKYnJI1H=<+}dV&-O7TO8?tS*=5qi(R^lSCCXk)lwBv~
z|LnSs(l>jOzhqvbPJr?=lwYC7MRPoM)hVUNZtV&lzt9wJ>?~@3CNGt*c1q>3@?ZN5
z@zFj%>rC_@uE7!ctvIP(=NusQ&yM&X`%}k}PaD^Kp8d{3UB@|*c&R>w4}8Jd<V(GX
zy}HlqJ^XY3RXgz;E~-b-gZoD0WPbhNto7kbJNB-=<DQk@U#n-zE8si3pl9>@4p;aq
zeUq>CYe!$&Iag3`wI5<1+PNQLe({K3*zdWI0<YvH@&Iv0{PlNDd{+nXY(E2+<u~#b
zd53xELA%TcjOXuc_<dPY`3Sssjtr;m&)^<ClN&r`$Ls=r85f^^ZzIq0-T3qi*NuZZ
zUyv{GcYcL$`&|AY{+mY~RXp=|E#yt)Y&`2?7yR8g^h6HsH}MbtMeq3Fm*fav;H7!^
zGrjAN`uh*YBR}@VuGq8s40*y)<MS(WW6#!ST=~EIzHo)yJfq~uj?~Sxw@x_WyYmzB
zQTMmMfy3$q^ud1RrSepEufD^-{T&<jATM^_M~)~zQ^&zKyiwP5-sb#+|F~CbU#6XL
z#36pk4{|-^2LF8bcT?qA?45nM&u=~O3UwX__uw5rAwTwETy}`UW9u(bae{qn=lnsO
zAQ$llWf$(B=+91^U%-F!;m>}A9{4MJWykaj51e<vFXv7uJi<S}vo81dM?SsBsVgt~
zW7Gc|b)U`sHgVN`5zi>OIq!3BXCCq}zj#X?>{WZ;jf={|Jb$aX?D`RpKG*afm8YS`
z)eprF9MWHXLqF}T-~DgvMdyv~zwnB?2U_}hZ+-M^KF`{FrdRFQ7kOz%uJq3yQF`_N
z9_Tyu>^*(TpV+&7b5d~tj-vR3qwJR*!b9<fotG#(7AN2_ToTXWuklfNGhEgmzs_Ue
zm+|11xC3|j7o64~PO0zcXFm8ZZ-R6B8&7}F`oly0jgKF=2v_wpj&^?k?{`^n)&Hv!
z*X2vb6A#GQcqlt#_x#WAl=zqU;~7r+KNRerAF_Lt-@_U9zq2U+>y)xP@@H4<RooXJ
z*%f;ezvzu!(G&Y{e;pNv$<;na{v)o)Yvcp`od5Al^<g+8KJjDvl}C_2y@^xymBz&f
zf8?KV0>ARpjPJhFhk4-6H5dNnrq9jW(uBL#X}<+ee7An<)Gn#~bj52AnK$(EtxfV^
z@77H&_6f!@Pf~b|vVVGn$L!vH2J@I-oDxs04`1RcoVQ+b<4@Y7#@|-Tzftz%-ne<$
z5k8&Y!VCRT_G!ElWhY7HU+fmXkR$!EFY}`0OHWDhV?FxOb5i>@_G(^n*LVKIpTuGN
zCeJ9pLe+iv760=dPQfj4#y(74#y(X4q~bL^EYXyMXZVPUGtOtkCGi^7zgud*40p&w
zJ&U~UkJ*R)zWoUOu}^Y?AMjm&_D5cD(Rkj|pYiZXU-<LhIP@p(h%=~tr#OT^_M~no
zJ{Vu!j(pS`$xS_i{^%cmqt0jK73|JB=|!Fdzl{sW=mFp4r(VE5jbk5bUi)cwMqc)%
zp7}lhgd6k?PvJb=<o|HT{u$rwS>A;Y`jrpJ|E%A4=Z0DL^WHk-iKx6n-Y4IH5Bv)4
z)~{Y{|9@Ko;vV}~&zARy-{PQl;(O*9+PTMNe7Gv_f}i3#9EAtoi^J--&O7+0_V)Ge
zvC5;|Z*cES{l@PW)c56M@Ibu=|M<pF-g7lCyvHYb!a?KX+j!=Ir`q|h?jgS<KX?Vt
z<QK++Tjr++dV`1bChx!>d=j6?-8i0or#JBd#SebXo76tmy$JUv<PC{;-mAa+4rk#Z
z{DjNmahw0)-!t0hix>9!;;a1|Do*;p$$l?wKJ{UIq2dH;zB=mMO8lWO@tYp)6YMwq
z9eZ^!c{|F^_@O+4-ca!nB|lVrTr=w0!&aTMwW$u~y?Ic1vHVJYpuU63Kb)h{lmG82
z4?x)?Dt|!91(k20=0oKjsQ8VtCluf8!r$kSA35Jb;RY&yLgj;~b)soE;yo%KC{gWD
z_Kxy1lpmqifyzgc!Z-LL&cG>l!4BC^;s%_7gW7q9pX}6kxaV1WlpVuU^$36GN1X=$
z>NNIw>ZaDouI1tOUE&LSU}x+HzKVPHZSrAxyLs^8-X6bp&Z*AiS-#3X*}Z)QJHVH9
zpw__;-JfQU;*{^+^FLHvN=iTe&WAdbc%lA)s*8)0{F?s#PS(8ii_$+kWN+Tn2mQlG
zd}lv;q#tsF+x%QU1-FgEe%T`&<Nwydeq^cQUZ+&t;OD6Na*6u?9{jyT#edW~QS#@P
zs5*{z^ej(fA1GWSPja)K5+zS^*U#^^?3>{cy@*@%OdtG3z1@8o_NyKIkYDj%_9f1H
z7H8NY%C7KZJ*YZ1e|LYAzl&$`7IIW~@H<ES>CwF8gF1iYxAJWJ0sogzUQXZg75J6)
zM0h5jhM)X`zU1BZ<@^)w@jrQnI*5A?{92x&A3f8T@r~y`f&802@qcwEa+42v&ksse
zzKK7653lK&y%|TGWH0ax#gF#h^CSKRue2Ap#cTTa-81SQE<aL-LfIv1eZE_t^^-Gu
zm2c6T_xu1~{Exo)mGv2iUFLl31z+alN9u9rfd}>>{D@t#BlzU+E{dP_r}#nP2g)z(
zZ;h)zf0F<6Oa9{i1DrKKyhGu#I7^@GB%k@a{_KYS__hA1ct%hB6;8n?eo4;kMt}LK
zd`6s-k2sh0e{lTYFyqU!<X`N>{?5Hb^Rst$<M(9tk@9qW+5fE{f9YL!-q~NR{^hRw
zo-}KhYTBVQUw8bn5pC-4cRK&&C*m;vUv=TuhhKZo#)fsd&m)ez7wGSQy6@-z2Yz|)
zLudW>Zw9Jg-FDHcp&NFo21lGe^JhoxUXA?5Q@{O(Q}(P@%o=*+rjHF)H!PZT*@Aoe
zDtuWdxftL7ow7df?endJT==7W)%xU-<jCLbN65waXaD$^>+YR9viikGcA5Ufkt3^t
z1zT?&zp=McCm<jC_Io~Z^#8T!$Nmi;{8`?szKehD*%AA2FHD`CJ(3%Fq28+-(4+6_
z3+x@0zt&OrOyGh0ByfN~%Dd&$;+J?07t{&RFaP>&Pffk_xu*IUoPrb1N9F5qI;r{n
zT@n<(?rFgr^*;HCIsyLi2j}E{a8uqO-!YGM!ZY&q4Bzy}xBC^^Td(+u;t$TNW730m
zaNNAwc^1D>aY8;I&g0WFT$XnxB`5174|Xozpz;p+n)U1F|0A*^c_RB$x1v|};=O12
zq_{+0@=(vem^yy_+kdjLi9hT1ERUo&-|0nu<sJb$b}xfm*b%<?mwFEWF|K{R_u?{n
z(IdOC9{NPtfqvFOE~vV`aoH7rvCm>R##`{Viw+;0($lg|{zD$O-1FG!Cmhz>lFzeK
z@sK~+4>?DWpID#y<k9$)Ps<zRQR0F1(<i$kFZhV6pVA9lmmhd9pYz^&^cPRWb#X#I
zV!y>t#69_7*2#?H?|z#{9wQ%?kD%hP{l9ro>w|yf!q3UqvvKXSe1{Y8fS-7VEB5c^
z<1fase=?tb?90B4y{NOH=0nK^<xks7Gah+nN9KpW@SXj`DfZkgC2#p9dmwi>OJDd&
zTw=%S3-V{wJoZn{<KdJ3_^=PMkL54&Vs%g7;iqx!x8M(d*3bS;JL{ug_M+};eD=kz
z*o$`dan@;E>tpZY0o-Db-W!h`$Q`c1v5W)o8cxDJ<B6Br!7=y^=ZtGUcyGT%pZG%k
zUckAg@$~1n<YE6KZzV4{$R5}w{m>ix%Ke=68V??@5Bg*0?23NrBkLF93CjPiPhH)(
z<jjBhjc4<)Q+CHr#2a`je&ZAFnGe2s#y`EVH~KO^KR3U+Aw02OeB;AB-s6)TjEfKQ
zkhg1Z9P2VZdoe%C{!!z~Kb)f**L!v?PxQ=B<Wtto{^<=>*C2oS3^}Vy(FaQI{E*$q
zL(ESu>;gaTLyH6ApK}U%mOR(^<Rq`+@8n?p)@R-9LtJ7X<O%QK7Jr4a{8E0x&-okw
zClB^XUhD_X^ILw+PT7rqp7{~`!%w$VUI>3sa>XAx(K9~819-u1^e0#E*)zSfFYVa{
z{cBHN{1M;UcT45#>;r}C?#HkPe#`Fg!=8=DK8+)vLH#{W<FZ%d@^A5r+}R(y6`#d#
z`3e2WpVfujyAbD$OOEO->ap^3^(1v9`9(W!o%L5gnttD94R%N0>VV>%_Tr~?qx{k{
z`TISEeJT944<uLmhl}i#U3yOHoRfW`<PGmod5d+CC!F(LUgvvKd2rHf_u}(sK5+4s
z%bPFC;s3WB#0T|I@d5twe|Z3Wfa~x@yb)i;dAMoc1`pvHzvlPZPP~_ot9z(Nt0Snt
zdoSLrQ@|(jMBd=NyiweTw@KA^jE^7jT3y{f!@A@%@QYmROUU7lD_?)+ov+{66wf@X
zkE=h*=is!uGAdsB?sq@pr}%H*N3WjYx40o+%=)wO*o*v-UCC$2nLOExc%M{0B5oK@
zK4)I>*0|<lFXn-B;)(do4$Pw;J0L&yo^d6g$zPqF9>iVa*w5InS-&`Jz1ov6zWJs6
zhW%=1J=*bWc0-QF7vITGfAa92yvfyfRD5Uu-pi}`P3gON<mK$ty2*t+oinpv>-CIJ
z?d+HATjZbGnMXdD?c^)$h@Z+k;fp*`p2VKiG0f-ufZojSyY-U0{_<4$Dctnzeu(zI
z`@M$WcfcR_uhe<n15h`H=kiy1sBz)jw~oB|wkc<9Zpe4tn|3~^zw=N1P<ZXxJp?%4
z{`+<>U2*Es5A;@W6Q9mK-A94X=2H(>mv&yN{*Ai#z2Uk31(%E)sEnshB!2lj*4s;s
z@0`~;wf~dJFT@S;MnC%q@x^=Vv`^5Fy!2zQ?8P|VyZ7UFK+f--1EcJmpPPq0v18}e
z_EF}AOZHXl!8rCw@Bv@qEPF5xoFPB!aSnm94}L&?@(JhY&cE3SdAf({J^7IXKG~=B
zSU<bsPwLOcB{y}DHf}FlK7MfQX-)U;-1p-@{GA`L2l~J-xuS3Zo}=P~XX~RU{oofp
zd4`*wGcOQ#)PuAqPq;<T>dL6P3Ths5K*g2j#8<97;JTg)bq|G|!!L3|`6v5eN6yRG
zojeA0&l#l$xQM!sVLs2O`z3G+PPh-_eiFUl%d@{{#n0HK`!DQ8zV17IjYpm+e1)t0
zia#0OIitKl9F{MrW1-@%^`ir;hvwXH(8h-RAFkk&{P1Jl^hLk)jpEP!7V`Fd+`6?(
zCLGsa8OPsYT{`3XU1z_yr-I}3126GIPT~f6!$JEDIKdyZ=Qrf)?^wYP<2i?QzX)IK
z!@Vqdq<mQZLqGU%pTK+kqHqdd@DYFhel9A$vWs?nzv-k)Hk`ew0msZQj+q}7ckDmp
zQSe7zXdkN`DsPiF>Tmx8Z{=h5WAZlgW54EMzs9#8uphJEGOzFU5$+RdXI%a9V;^q4
z_+=;L1h46fe(ay%mOMj0xQ@bGan`u<UH#y+eUA401|RH(ec>aicI?5t;)D16$NJb2
zDjuV74sP%txWR9=7eDwhy=yN%%ERFmevD5(){#_y>q5yN)sMV9(<8qT-{}EAsQf7T
z!iVwr6FnQ>IP}4;@PqPK{m2Vn>`2~bohUn$kI4U=v)NCpM;Tv!;XD7}Hz>Zf<Dd98
z4*&Mudw5_S?1dl07yEY4?k}n5^Ka|aP94TQC-)n(ULpUHf8gIUyH@Y99`nE(a<QJI
zSqG_mwl4T7&tspcc!ZLp_9%U^3--!i__w+dxzn5ViU;gTJN?Lszmu2u@JYRioXLy7
zuygt_AAit}eX`$fsrBi{9@sUxlArlYRNcz`NO6Z8;DUM!J0%}+P+Ucg>zvNLS-*Gm
z-ni^QJ;(g&HO{xxNz%^bH@@31uygt$FSH~V@do8r-BNlgQF`i>(sQ>|T&ttjr#@pJ
zAWpzTH1`qwQap4YA9emu9{faH;9th!|NK)t^vth}$6iYm->CS5s;i>xCn?;A@9M*T
z*9*7VyY}vHzyW?#N98f};~YgDj-LGe8uAmj?Gxx9j(S#S^ml9R&)}r`Ecwgp=mEZ?
zb@AfQ^}Btr|JPxEjmkIWyW#+OYDW$zf1*cq2K}~|%46-fQ2MeTmxsV#ILq(EkHkfJ
z7r$W_;?dgqYaYGt;og?-sBz^x{1hIG>+(~$0T0Df6z<D6yoZPU7&RY%W_SG2e$01N
zJR%?QOgth-?ac4~m-AP05ZBm?akcl}IL4(9bwca0e&fSCe(M=-nGYZKZ|>7s7khOt
z8b9zEUMH0Ys9Wmq-i`IhquHPPJMhJO`2stW&yl-(d)C7)tk=GQJm`bGt%qO8J5hG>
zLea#3=X&y2`vLkQ2l68~`7M2GPrmG4zGpvxlDB^J#a`U=lP8)VH9vcD{wd$`%s!2)
zzM(&Rfgk!gm!og*=?%`jY4qGD=RVU{sY9v1(>wc3oNMPBNA3NCOQ)@CsC%Js7FDO5
zx@qfQyzJVi8pf3ui3j+A^Ui5X^NSPw5nu2Ce)?{F@d>~3HaIKps7u3JaWAPnLY&8!
zx`K6}`s2s5x~}|1d?iOzUc_JNpI^g4&+MFD<z4K6-5Hm^!!@{Kf9E~?5tpnV{*pU+
zi$nH{*5keL>Dzd1z3#vJ2lm|is;2w9{9asu^Tt8N0rsFhN>1uD-s`U)oF)(J(B3-a
z+v2P7eP^GZ=|LXsy?Q4(kh{3gpHX_$AAj;2dP3O^xoO9q=>dQAD*m%4R6ZrH=+Dpa
zgMa=;PD!nYp54b`zmu-nyn5Opo0{$uh_CDz4!Tc7PU_9{2*>H0T;Q&L@DZ+)gE+&^
zj7MMk%MZ;%PoCui?2SCgi=E*M{;SKg6Yb;~^vb?b>lDBA^WOO^|8pLrevL1b-HU7Z
zRA;b0`@lLXuXxKF54!6$vwB<q^zx0rzx&fYE$d|u@Zj^;-PySRz%5PZ2*#g0XUX7M
zqk3C@e`=gfYu~=+;9YxK+N0XBuiUS)H~Fr1`r}K!B2Qs=?#~z}pS9E9{KjGD@^bkA
zd(_Xk@)iBW1N$@a7fz{P@jHIVUipoDNIlCu?2^B;XE?*Iy=R~Bg1y;yi^uW;@x*sj
z9ZK9q?eF=A`~!twsCqs-_e@^+CpUlpK)xW(FPcAP#^U~-%K3=*@(mPUsJw_?#R2+5
z=^5^bPxK{@!(;vE%lP~iPU$Cqln;vw#@D`%vKw(kdpHeu$lrR{v;DJv_(IvO{_KvP
zj1Q;SCqIC9)<rI!@ogS_lc#nie)QK*UJv(C>l3%kM;_+Uj{f<PxXymuU*j+QfWOF7
z$%*~(3;u)3E43$Ae6eTk`H%U$=Qr{R)b9?}f#}gZ_+n4w4OhsCJjltuNqhXFzQZ;B
z%%4<TO3JV3m)+uvov{b=vPXVye)dn#>Y~Oqk31Y+iy!croo4*-9koBjFL|gVSReoJ
z-Fy6syXHm7KdJe)mx>SUj9(>Xm;3>)$&17xc!J8?b3FdV?>uXdy8nbv`7eIti{hbn
z_EE+cANh^Aq96YeH|Ytk%hT8ooMC7Df*r9NekT7`FJTY(r(bcC{EUlFevlLnkPk{8
z`onYk3-icZ;S_n{2adoi-%)l!zS`+WU-%+V_-h=v$DZJ;@#t0FV&9L_w{^im_GEqV
zfd7*Vy;?`+`S`USxQb75fY;=#Zejh_VZH8?u|M@6|K~fYeWUv#?$5b*D(`ZC3qR~$
z9YGxh-FMb`=dYjMSGm7tUip}I<|B9KMEJKZekE?o_l!r6#&hq~{X5UrWjy_T_ue?<
zZ9aNQxs!|W@Q-?L{9Kp!){p9k;=e@M32GkHI#GH+=>g>jN#QKNR;P15M^5es!w>$X
zE-vn<gTo*GMh^DV?k9^g&Kd1*;jR4;d_>g|=+ECVgd^e*oG_2|lcV*BU*y4`$%S2d
z&+l_Q?TzO<KIAWOL_C1|@)ndl!~^f)cgo2);teWqR>zhf%iq-*;GMh*F4DL8^z*E}
zdlvp)jC{!N&%`NqBX2UE_=!LD0p||#BlEEj@<H)YN3GZV^y8eGU9vO!fZOcY{aJag
z^B?g=-jI}k*guMo_8s&ro_H3&=+!u=al|9kIP8SK$j8V-{PVs<jYF^aLFFq+<?HOn
zyyQ<GzGwbR|Jtzw<BH?#K|B7VJ$r*I<}*IN*`0RsKI59#d?^3Hzj?j)Y(C?o@<;8=
zhfg?y!eM;l!+6H=Y#i(JY&_#wAN|lT{9=#r#k?rK`MY`8tMT{`zwu1o^ok$+u|NIV
zelmRC!T!nJc&PsPVgKe~pZK<3>(k#n^2OxOIPB0o;tHIl5A*O_IAnb9`LXqTPY&{F
zaZo-jE{T`=q52!wv+wkSYM1<w18SY{(0kOn=|g*VVtjm%BR=rynVlM+-q<A^Gmdqr
zgV3w_*#$r07sjV&^O&FCv19hg|LNWM?9+Et-GslXFTxq~zz@&jv2`0CpW>x>WPJIS
zb9wR2eMtI{?}(dzr^f!oO}`&v&+<2Ulf0LH{oNk<CHoMk{r=S7dl!$5Yd;Gg$xD8u
zUhf==-O;<aZ#?J5zI#^ZAwTk_H+c`cAz$`J59|>i;;(-4BIh6SE&QO?XFPVu@8!S7
z)sJ29BYeR>@?)=X-Fwf@SB#He_UQk7vs;v&=%1e99y@_!D8DOFxQoJVl$=oKjVQZD
z^+WNCI!{3De^B^D-s}V}{OV^f{oQ3Nwl>}WRKIlo<~=>I1Mk%}>4SW|SBFBMm~q15
zg|F#rndfU8&RO!=pZ2u;-Bfi+?eO7z4Rucs|N8s8$MDZQ&K>b-zrcRUhuy1}%4f)5
zJ2=LUj7vW5dBaiu>RcKQ7>}RAQ`ETR!k&{l&nEv)DZgRQ{Et7nhYo-E6$<CYE&eY~
z@f&t$z2YGM7vGZd6Z;i7pA^12XZGyxC$m?7SBiaFCn_$Ii}&&x@5u|E@kjV!UGfV3
z<v|(u;XgT&6Zw&sXY;@*@?fXrhQcFuk#>zw;~P)>gd_MCAH`+g#S_n{_+xzcuJu>f
z5I3w(yf+^?8izdbt9|m#F4dcSHxBAOezf<Ds`F(%P94WR4C`SJp7HHH`ynUqtw%jc
z{Y<^feCkU0!#BCW7xq9N^rN5ki`%}l56|}h_{Bdt`kr}~cKFiHdf7kxpm*)f&;Qt)
z`~^SwP)B1=_CYAWD)ncdN#Q>HBUgO!UleYj<d0fM9TnG-%1h~;-1xh^NL;}ue})_0
z(*t?a6aL^dT=8r@?1^2nU*`ns?al?*iNB+u?(Dv$e(;Vy*(JW&jeUT2{6&4-@74W1
z7<F>?@9)SsC&|BysQ&DC{ruUu+WWgR&IkP7Tpita?fJKJNP0bM;SVQ#?aBU1JM%b~
zu`YU2Uq{VH4$dL`T@imD%(%`q=+Qn4uGz2AFMnb${E0l-1v_A;@C}~A0ps{iul~M`
zIADD7gx#X_YTYO~qvCiSWoOod;uoHv@`<F0C*3~FE7?EX!w0{Hw<x}}Lw)Bj_~GyJ
zAp3^IpL~Xk>Ys2(e^fk#@Ag6Np|dOgMbG%P?`9{t@6pfy?_p=Fj^B68$aDLvr<Z(c
z%V}Tjt=xaaxA>=yz>nB3yLWB{Z}^$N*W&N795M5ed7n9BMCE>tcwyc6cU}q)@$2tl
z@K^p0C;c4_e(hXNT|+z|cX}pY6kp;2T(|#Y@9a)|McILSGV*cMI4HlipRoTBU%eN%
z?1${nJgW<!)+;`_U&kK(Zp1j|LG1(8TiCySjD3W4xld|-<C2Gap4LG=_MhT2e&y%<
zMmzlD(|h;Q&1*dST>M(Mdm;9#&cn>BZmPb-9_R;O-s78`)Mbo^U;WJQ-Z%M>qkX>b
z-sd{mv;OphZ}x#d@2!u0TZer)3NP`CfA;I#kX(|&X%sGt2k;M0xwjw>f+KK6{v)4&
z8{(m7<G>d<57+oFe78>iEq^vH|JEM={1<K*H}OS0HV(cr50JmelkAJdH+dKQFVW?9
zj9Y!?<s&M26CD1-<r^2S>Konu`$+OTd6au7rw-iryD>u}t83<beckSJMpgdr89erP
zt^6L`?~?t0IRB?keh8;M``r*cS5LN|^Y@VsJaMn@e{tiJ4gJjTo{YZ}<^GOw{oOEe
z!{5Pk-)7;PZ(cm<>VeAo{XIkZw*QkSZt8D*e~0SlYle=#=AFHj_(mS)LFr#!D1T%x
z><3QRCpi!B9=^g|_G$h6#r}fdh;RH&KlbPM{HT2if8mex!H%`(*Zf#MV0={E#-DrX
z@=*M?_50B;pZL}T`YY?o{8yZmx8ei7!DD%iXSm1j?NjLk4#01?N-pjz!5O#&f3%ZF
z%4bmf4RVrK^K<+qg=28ndhOTLE9se@;1=8#7vziZj2ze(yMuH1)J|Pi9JYV)o*c<f
zyfPpD$lZQM9OpMl*(G}-XY!H%l<=|io_xrQU#hdSGjbCT)x+5ryA+qjU3^);`o4Pq
zywUv^UUBz8YxBhKzwaHN>}y?o`IcKIf3vT3)tU3=J-PXLx@6M(mmj!vu(hM8`Y^rG
zzw-mXH*&t9uDsigw}0XB**z`!iTuns@Jc=@A7oeLK(6}B6Zx%tF{!!;e4$tN!N1fw
z{9O(E8hB;j0O!OZ@yWQTI)VL{^CvjwJ$pvs0{@2x;-CKb@!h_OAK+7-0GI89#7WQo
z{-HQQ4)Q4Sv;UG0z&H0^;j{RMKX?e&=|y`~JO1k2kzCX#tj9i%p2(lS@mF}k&fqrt
zLfIKVgjf7ooRo*bCFk(wk>|2!IAL7R@Efk9>;W|%DlWlWR9=a~DHPu*xuNn-6dorf
z*EavJKjYMQPk69N4)zoL*txrVj;Q@N{*BA8?K`wn-(VN!lMgvRq-XjOXXwNJ#d~sh
zj;7urUa~9e5|`M8dGI4IBlq?^!*Y_}wd2w7Jt=mG>Q|!t-M*E)QSpZS#aUFIz+cGS
zvwfX-<(!rsS}%VUr~EF}IU_rP&*GZ6;diIr<DdT3C-@Kf)KPfl-iWwm9{Xo_U_T{3
zz+3lT=o9|17w24X6W;S*`H_7Nyo00aZ17ip>;F*0P3vKQ>fQD^_9OBh_>3=kJ$u)W
zed{N#$yedBzo%^;aYQ`CkNueU>h<K|-md+D^@+RUhBzmlsVAnq@h3ir3+j;O5zp8I
zJa>+ut}5Q>Z(jL=@AN?~;yOIVr+mP90r{&Vh%@v+4(y0Mis$$yU;lsH{?t79hv#t9
z{)d0z8}7kp<Io#>XU}lke!{x7=Rf9$Bi2bz_6yp{H`qC<PE?}!hQIbd;w8Kgf8iV)
zN6DMN$gji|cqZP<<IV3rh<GC&vuF4TAMtDb_KWZi&Wc<13;M%<`SF@j*B-X&oUKiD
zLh;)^%zj1u!LR&Be1Lc4=kJz^E940;)c?tiJ@9vS;k&pYzxAG7bV~V&{^Y10Abzkv
zc>;Zluk@gO@&z~KA-=;G>ol)=96x8*`mt;N@BEP5+;j8Z_||D2ewzMn9C~E8@)bB_
zew07sn>^vGIN{urpW_2gnqNQT@-O&^vIqVmzrjDg6Mx_dJ<8MAGs@mQ8`pfk%g^ky
z?Pukg<}t5vyvHa0QQze|aF0LncX}arI8T0XSG^F0&+<6^@RRGt54jr`zvL`0XFvGm
z|KwtP{o$kh3a)4mw=&<uk9`Vzw@%~hFOK0)J9;sm_xR_3@R;1mnf;QVxFN1shxzG?
zeX|eqvNP|k11{o|ALjbW$vpBiIL3a-8Nc3>hk4<;XY!L@B%kcbcYbF5+Oq@eF%MjT
zEATMewe>Lk|1^_7^RYkcWY3=UGrslHhxYhI@y8#MAAZ4K#T9nJug#B=7ya087)Soa
zPPAtS^hp24HLrD&GdpId)?@u}!n*l`^=n6e-X|aAqdoZ>*Ltl-JR^Vn(yQ^T4_~=H
z_K!dPQ2fvrzOp^LFpu{5fq&l1r_?Xq-&W@|KKYV6+;>iE9C;!>P~&Q6J>((}miMFn
ze~I6lJC}0q<@}02+sFBTDW3hm5BHty)0~Idzsl3>!}Qm$j>?PeSMC2$_AS5Jed`fp
z_gvlAQV(Bw?fWO}c5!cuzSt)`BTxA+obo#w@rgXQl{zn!kK!L*!gc#`e26djmWQ$%
z>u1;ULi-K)!Cv5^xM2SxF44RFtNprjRd%FaVSh`H-m7D<Gw;<$?5E{->HyANoqxKA
zCa;Ef@=N)$XZvIFRWFy9$~W1o_xQ7Z_M|Srezh}BQhHXWF%LWD59;FDvn#k@zvdi8
z+>oEJSNREifg}2tD13z%@PmBKBi^u2_JryOZ^aeu*{8T<A0V%>F7qUXt0;d&Q~u<P
z!VOejf#!VbUa0d3)H+f78&sS?#ch-yqx>A@4=6dJ^n=PnQSwF2gR*CopQCU9Rp&*;
zJ5>Bh$}ZKBwU=)gj~%=B#9rOMa!%@8P@Ncl%OC99-Ph8O9_$m~uYEi`g~Rd<c?db<
zn;yjndXs;M@4kyK@*UsR+w@mIH=p`|yvR9+db7GDJi)*7e9!6+__ls=MSiNzNAGYH
zAMUw{JK~7()QNnDE7l87zIf7OBTjqYV2eNFlRwJC_@(#c0!PF<cmtQ&J^YrJ!U?zl
zf2>zK_jR;i_>Vt4=B7I~HQ=;3Ag-v(^C#z<@PHqSN9+WC@=NQ5%W#!^w1Ye9X{fk`
zvTJ&x=hP2eAy4urZ+6QstOK9o6+iTz-q`_sa6WIp>X}}x&+mrm$NbhuF3#t@m%qwe
zd_Vl=ckTAFAMI1|7x6%xhJW-cPnLhkXXLf+t;wIo3;Be6$^7gMU)p)L9(G{;{M`E0
zQ>}wNJ9khY=D+Ha-s_K&FFmm{_`v_+C%b@e_%lBL<G1oM{sr%>6QBH;A2}ECUVP&}
z@G5Z{fB0sP?A-k9);Rc+PqIhP)<d4Y<BPqQD8KT1GJYiv$<O(PdcNOVI;R!)-~-D4
zoCjEsctPIeMIQ8wFY_AL{y<#uUR;CY@-ffWYkl$oalrYCcGd+K%_}bAPd>??@Q*+G
z6feam@5NpA;hFtgAN`OU+;qO;nSHS*{p36JN-oxk@}u;B@7V+YV_)JM|7G{^6D3FE
zvkUt_>$47&9N3BVra!Y6`)m5ZCpma_Z^FFvP5%0O#;^6!zj&s-{ER)J#_>CE`INc}
z`Km{=PkQE8`k`<{o@t%F!!7n@Ty|oAPTt<5{K-1_k#(~t@~0o`aZX`;^JiVrcero8
z{F|N913k*C<qh_6;yk_LhkWp9eEA7I@k{Z9AK=qIc}G+0CU^G4pXrso(S!ACCq9vv
zXL93L?2W$32`-x7eOv8N{A%YtieJy_2<%>ezhm*;xv=wNe~*Fx*w5kL{PuG>KKr#k
z_933(lia;02friAcIL;Y`bYYseW7|Rc^KFF*{^+${S1EK8~YWP$i=hv#$hM?-8kfi
z4|sz=d}+^r{mu`Lii_|;yyxfcbKqMXHZH&O9sk<NKggS$_<{c9Y`xY;KJ>}o&99$%
zwKE@kOZy;iIA|X8;=?+;XZOZqSK85&_U3nwgC6jSe>m*feGvBUJW_xB;?sNbB`5gH
zF8n@?-S9VlKyL8ad4l|ioxmgax8apKE56_^KHxfe$q%&Wf8=Lf+Iz+q`C2#r?F-#k
zkw>u`_NtxzF1*Dj{_!balE*j~=P%?g9<ncT7LVj1@=WWH*XWNQ)OY?PPY_4=ll)EI
zgKu%se$YPFeGl*L3%$pWdV&3w{^}L#@Th%{by_znURxJ=+h5Al@uhB}j$nKg-|}62
zv1je<)9uIXr^&}SD7*K2WaFapV|4`Y@nIhNpf`D-e%hPYINGZdXlH*(zwq0+iv2J>
z@l)q2>Qwgs>J;9)$8S9KDss|J9fDr$FYM>x6+2=d<PMMdxBV7BV|Vt=<mUb)duK1~
zgj}5$S%>fB!XB(oJN&|RI6#m1K;<cLMxF|vtWP_5A+M9~@NeV5TYdt+;0wPYH|^x{
z#)l8ui~sP~{vZCqHT7Tq3NP6`dmu;iiI2uNFFwcx-ixalkJRnuVe0hoTYJysD{jDF
z^2Qf_<jLgD-*TTIj)<ei!=L}NXnt{5To-qJH=exN|8Mf%c;cG#4&UXi@)G$1IkGqR
z)A4J5axkv=o76bsz5LRA{+=AZoX3!p^^=Eujs3DGIHoSfulP56gJW<Ve$Zo5ahLsy
zYwTP-Lp>3m!C|<so(qTJ#j%&qp4hXdr&6!?yQfK?IQ_dTUoxU{j*|8W2idcIE_;>t
z$q(4K_w-C3<ZC}5|JIIP=#AXy+k1XtJod-`*n{`RN6pXP$e*6!qWcE&OL0<OCGSy(
zgKO*pKI7Lunti|__$g1)p8adb9zDYY`z6oT2|w{GK8Zu@(>m}ij%X*p5jTw^{)>O^
zNz0$yn{Xe%c>z3@r|}=};T!qkL%W<e`O#iJMXvl8<;Tt&=+FG*lk3KZaoj5xH^d7#
z#&772z1errH+yG)?meq(%k%8t#Vz}B{oxsWcJ9nC$b)_1oBqvD9w_^QOYDZ<7?<9(
zCufv>!7uIQEvP(#pOY84@(1hlot%?z{8+E=&c{&iwP(+$@#zhPKlE+itRFuo2Y75=
z`4zu_^X`?&7s)OCfFH_p`GL5^ZpjxOdA2V8gv#gJ{yBU<)q3QC{Kh)qHF>iy?aA3V
z>;ffMb|8MUTl#0$>;nFw)&cLe^Ng}*a$uLJ@8pJBpZ$%xR2?Nx=lkM<{9pXRk9vFh
zCp>`z<b~qXcyN(jkPkh>JNEB8|0aL^#RvL?589`_vlHLR*Szr9z7KUz$vr#I_J!n$
z`nxUmOMXws-mQzD@F%!VPWETo(+7KI@9dUb$PYi{;JfkVb;iRVY9Gn3$(dfs&wh#>
z@elR^hpa=rRUEM{_C-(n%Ol+5Bro&vYxfkyPy0USb^O|V=|6B>-m9JcHhHq25+zUb
zk*9q>Iq1({JmbSNxzH=Qvln_bzW(Hcnn!<jK)?Fgf5Izr(BHnwcm3!cUg6*T=CdC7
zo8udgozbsn{n<aiCvW=+`yDu|KWbk5!lBd;yVqVE^qxG8!ykR;=lqb~*f}|=`>4OL
z8{=B1aqxqBZ+>yteJkUm#&^HUK7(A@z5EaU(zCq7-~VP8@+SUG|NM=-$Opy0@9>j-
zidTsz<jcOu4-W8KxFK%v7x#|vZ(l<`>_L0;&>#Lvl;6-3`LIL#O#4{&isDOtM~?JF
z&h+7#9QBv4_>RJrq}uB*ui{_&!;7T$BPjdvY&=xm0QL81{r=s(Z~C=<`eaA^g<Y^0
z@?>Z9!9Gy+J(L{zm3S_1;3w{F@o#ZaUhQ5SJ4WqS)uq*s?OQMX{gM0izp}6T$h;$O
z9y@(ZCExb@5qXJue78>VNS#OBQk?T%oD?_h1L2)@`dz&Jfw)U=`11F}{H}!_-S6{v
z@cceO{>87zmtVm#`+DnezQ*6_%Q}oF4;6RV3;$#v+F3ujs^hBvlQX%nGknXh#1-c^
z^i5vYC$B`!t38~CW9%GdmnG`{qk5e<!0tTLEB`^+LFUEyr9XckAu0K)6TlB~8x>#R
z9bAEv>VV=qoJz|7;5YvjkI0RF;61z09)*vd`Iq;`<!{=tOMd3r{HXS*eV6>ed;L*%
zArFLy@@4mCQNK&`oxV}_0LPrC@oVpQ6s2e90P5hVcq2}r;wbyD?`9|J_xA6;vmfzN
z9Y;Iyjvo0Z`B*nxLD>&~fkW~(_Q8)(`GkJ#HL36BNA*X|BTqoxw`WiC58v^b`L}lT
zz)#h=$r~kK`HX#n_v!`oo_g_KfBJ^o^lm&j0$=cFUG@+3Yd-r3>o5--@t%FiU*MB@
z@T0$e<fWf+&7b|m4d3YzzR(XGFhBhAUVJdGxI{jv@lkR@;TtO6p>P5fceFPT|MVWk
zCqJ-XktdNex$rORl)uT#`60jXUi}jFonMPv)<K@;)ee97E>Zp>j@UoRxA5;f`TxiA
zaX(o9>Y<7s^MCv%H4a=*ulL^g?pe7%C9c@#@E`J%x5$6xS;iv=>!eS7z)g1Ny=U=G
zzD=+Cq2ieS^hdAw)t?^a&z|*D@0M4~W5|_0)yK1+^G*8?`60QW{6PLD55X__q2z$-
zZ+~FlBmT-mQU2)JI^>Pk14pb|-bOC^nNL4;PI5A?`HZK%b;&1v=SSp6PsZ_`{nH=H
zPT?qg6R%OYtG)OGpT!OMglexp925`D=Y8%QytgmqAMD8Za6hT|X@2`;@xZ+X&*txx
z@+W$fFTy+L@o>cc$vHdxa1QA20Jx`-IAeYCF#XzbdD-&ugJVx?!YlVha{p!Dgp!Z=
zP0r-+{KNe^^=7zdpNrBDN**Y@LfJPeZlL^tKe225QTUMlAWxHTiiiBwJn#YKm;99<
z<HLHiH;?_GcIsii<IlLL{RF;HeDH72_=aPCe~iKb{LzQHivDQ+-xPe+QR`EOaE}EJ
z*eAndb<u9A^&hux?UD({J^y#p)eFpH-S&s5{T(}~qw?kw<?r;Q{(!=Vr23)kUfu$K
z*ggD)|KuQ_5(mXC<H~pRr+;$LkN(-E_v9m<l_;D|%0I*_{NfYES5kQNLQyzkAHZ(h
z8^Q;?cfKpXaIWUNJe(ZTPSx+!``qVJ=i_(&9|b>OG=IvB#r-`MeZe#Tui@|iy7BW5
zP93P!73tGG1one(bwGBj9_aqAe3V_OL(x~_96J#o%#V^AJG6iHOkVQTq~wc=SMuz%
zD|H#|<-5k^cjBh}H~j~H){%a#KML>gV}A3Y?tj`(tE0H@D8GV3_QRRac(1?w20!uu
z&nS75i@bpT<hk;5=h)-~$K_EdK2UiLia+%L_niD)4f&BgPoB!p?6=&HlPAJY`3uT_
z<)7VB`a<nbO4Ph1DsM{)hveTSD!(hy(!AcwBTKYf`Aj=>=If4MHllq$Rop?@848C`
zeuApQqW<m$>i_tl=J$7*k~$wjFFR<`<bU2Urb6KsDqf+fCw!y)r$pTcM(uy()$aeN
zzSXJZQ~W{xug>fDUf!#dWPZVK`86uu$UERVoU?DFA9#iuw~oRyc8Ky@`%C$R_izP;
z2Pk|-#VgdgR*5=?f{XAEzR;U`D?Ehr@=x|I@0K5mm-2k^U%pD-@-cA{ZqchcoqCe}
z7&+1>Tp|x~+_U(`A5r?t{@TMo?ND({yS5)(Fn92wnM)h|M;#x(eh<PA+<#Rk(_XzA
z#jm)gE^42wJ|Uk&$(LT}i5^h-R#N^1AJx~a54Ap&9#Q%&QTe_+oZrbm?1SkQU+SVs
z)f3>0`*89xeh&ZSP4ugdAupCEs6(<B^*Qwq_bE~9_sm}SDay|I6UyK68UOmjC32FF
zlcRCu(d>fUJ^Q}^>?Nso<R-6Hk8qzrzA1kqKk-Gsw%;7J_YW?ewyt4a{LcBIeTnz_
znJ48#4){{H_4g9(<J~V158xI)^=BW}XP-xY;)VB_pKFI|XI_3tZ|vLr_@@tgG%tTg
z;X&q0{_oR*w_SAj;FO-0zfZ_d{GX3W*KA%r?T}5)&tG?E<NgD;G_})TKI<IaJn~!l
zy>{D6#X;**#}hB@54@K@$j8-*Q2aP2tfO!?srA=Ue4;r{K9i^4;aP{@ANhTfeS~#s
zXFc?hR6F|)zbli6$dmEm8P(rB_G6y4^LH%V^Yh)`v9LbBzw&=xj887UTfb*~`yG~i
zik;~1S^ntvv+Ulu^k`gm=e=ino!`xqFS{ZS^0Qv!*x#CmzvyQkRD1eLJ^8&GK2Y=T
zefY#fzxwaHw&X)e#WNKC$qU6n@yzeX#7X&!{83&b&o{381-@ncl8=hRzLzK*-Bv1o
zY%7J=sCCMd>ZrKwJVBl%ud*M48}fg43-92c@AggFv44KUPuP)uaD>0gGvTRm`K{l<
z`o9a_^Izvb`tuj#IB!wU;FqX;U!G4-@&M;d@=A8iUgeAW$-nd`XZ`5QckfYlnE431
zm9Lu5_}a4%RGpMR$QwM%bIdExkx$tl$h-0F{sa5)OyBO&$+z^k|DZ4OHxGI8W9zif
zWT)E0TfaAChyJe@`?fEVzrYXA<j5}Czuz1FKW)~nUdCQL!!h`!Kfj}Q`%L_6$3EGs
z{hMd}v1@+ApZO=blE1nIoFsqpu-|~I^rQ}KU!fm)8P9t0>AUselV0!%H~6vNALsh)
zTa3%!?Z@DXb<>}D#5H)NKfASF^2LX7-Saa(`!f$Y7>_=zn_jgick{|0Q2Ty-lastg
zd{^&6%_q*|*SM(m<Y9gMmfTSJ0e(?-V}06_n|AoekM{hBo$&+x$PtAva8g~@clgX6
zwKKl{+IePID12pi_G`|G?T747Z#?BYx6D|%rD@;d{~h|>_vATC2G1JR+wyl9#IK~z
z!{NBPq50J<{GD3o-|8v!>F=_^J^n;a@=(-w`7J)#34Pelk(+Y|_`)ybm-1Kr`4@W<
zC;2~osH5UZ9Tmq))c&@Pio2asd9iizJAT0*_zUVgxuu^QM|=LxPU$!0443q0=i2i_
z`7qp03YSV${RdSSqG$OCYCcpQhrO4me4viXAK;Yu3m42M&oQ27`H=dE_MX`lJCsk!
zx7A^sw>S^hUVrV4tNx>Y?HL}}H<6Ea_%}ZJ`<<Hh>a)(N)h*SD>~lP;bE(TYPel1Q
zT!$y{!0%+)CHrH?@(%XsImgw0XH)#)1BH9=2|n`Y^fUe@j+2KtNFMymdW}n-<l@;l
z_(5}i{ZQkYkKeNw&v1^Nq3p@CIO^Ph{joFt;5}R*U-=}Oa<X3hz&UtfT>OiV_!Otv
zEs9_MS4a6hzd-2$g~xCl|Kw#JIP^kMc<o+{d%|$TeoLJN&N#=g?_tm4n!kG`Z%~hu
zKg$E`-{l+f9{(R&Jx={sT|ge9-lZ;u56|}D@_KoYcJe{?E#9m9voHB13b(}za#9a4
zp7=(7?orUUeVqO4!S}8|>i3@>Rrxzo?gz>L<PYTH*?RahKlW@M&+;tioa|Aa&HnhY
zIAp!#qCTLG!2iTo?Yw6%#xXCv<HzcG&Vk9tcX_+`A)h68cqZQ1huL4#BiwPm;l3uj
zVIO{X#=rRsKjGi>3un~H$(!G>3-Q;z9yr8~=s`brz)snT{`m3UJv8^+_>uL~55G{K
zW6$z<`$kkfgMG-O)MfEc-*BAX#5wcvf6wYQ?q&J^s`Lq0={e~QizZ#R;GVupo+Ynw
zUl@*w3-(RcEx)60?e*8r`1Ij?j2_)T5O3fHyK|q#IQEhDkH!@@@nL-YvkUQyUc6Ty
zHV?b9KDci@@e{=tDqltAgXl9?eqremw`^&k=0V91Ww$82LGg={AF3TnUP;9}l;4U=
z{1;xbGx#cw$*bV0{EU6VBX)0{?B9F%g6faTA5ra4c!<g`Q1hVpM6D0y$K<oERJ{n5
zza{lu-bc>XYo9AG5I4QoU;Rt|Ew9iI9*HyN%lJdD<VDW(DUPv2I7<HFPvV~TzN7ky
z*PhMKUVQhAvOj#PAC}tVgPuKS9ol%PdMf_BXE)^OJ^LbG{E{R7P`E`u)<qxY@vI+y
z%%{I+`oW*~#>bENVZQ)JjAvceA@4E2@$m(Jtsiw?$2!zI@Z-JpdXJB6&(Gnse)22h
z8An{h4}W5h@)EcTKgrGf)@NU-U;F#;xrk@{vp4eMul!1T_KA`|zUj$2<y-uYo>1p7
z)^A<(EUt>j*$%&O+<N6t=4B7o;~AcrR~{wrQrC0-sh%ny(%w0~d=g*mckTQ&kKXri
zZ%h3JmG`oflsh?*ukZeUuKBEs-@sYxWRLV_pC|7WZ{ZvJF)q2v6SU{Y?7+IS9?!mv
zgD>OrPxIr`GaS-Sp345^C+ZjUPag6tc@Tc;s5q-XJK=}c!#}JCzwA#x@8x^!nOyDH
ztOE}C?is~jr<9+Nk8$YH?=R^SZjmEAVwXwn5A`=6{aOe6fD`Nn4x^=Y`Mr<4U*2tB
ze#)5-t#3{0t?a+hFReWGnO7hBYy&Qt-}=ejzRbRX{`qO@pMK@_@^*M=pJN>Px_yN9
z-rJYB|7YJ}-JXqW9P_D%iwE>UU;IeCOiJ(K1DuinqQ;>=IAtC<1AmMw{^3i02`{x*
zN0Q&eHTz$6W%rrb89t0tqWI{P!keVlpYQn#eo%UWA8?M`*d@x3l9tw0_a43`P5wRC
zjpw_#P@>u;6>pQWhdRn0QTq>Zf_~*&byRzleoK^`N|atplwWj9$;<u?fBeFC&+;97
zvPb?$FV@Ze*r{>oLqC+gcS`Y#!eO*dPri@+`cs>BSw2v~Yj`N$c!tmLPF@Mu;i-JW
zcX_D(;*2_;c7B)TcQoP&YJ3#l%LlYW<rS#>0;QJ{6=&OY`0rhlZ~Ohb`W_kmy@wn0
zoiyW$x;Z>jS7u-IsSYeoSO>kh@9(@udvyf&6x224Q|42LcHT)Zp0!sGLrZzZ_P)1H
zaZ}!e8h`aKcis1-S-Vs>zVU!X_kVj-HTJEanE1WfTO00$sQ39l#{Tbx?|!G?K8d_i
zKFL1hdHh^_MCJMVrQfqdcADp6^eB#_@&)6=8~EmYME)tCVL!gZBX}duyFU*1_^W-R
ze2hL&<6BqqX@7-Z6yM}#U+6u2A?L(h<G@dOAU~vUb^~WdK6pWO?1u*`{me^G;t9DM
z&p4>{%QNVce~8!OrFQI!{jw|l<fqm{4sgLZ;sJf)mwn2Ij0>0Gn(?fIy?PJNt;4*X
zQhvFksQVvq-uVJMw@-5K%kMRutH^tus~$1)k$ImvV?;&o^d!$=ck*L$KIXywR-J#$
z$QC<i|8SFE(T975?vIKC^dP@N)#t@8{<FskH+|_1$Bt;Rdo=gk?B(xXy5iKMALy<A
zaQViCtNKQ_aS=XRpLKg?muWZhT6h2-%xAyI@0=U*i&ek=w>R9f=avS1Mg9HY_Wp3j
z^}Eh~Z%?)G%{MQeboD^x?@u_FbbjUU?ZR<?=fe99&-E|3WZXbyeDd&j|IH5<I<-^&
z&u;hw+)95TSNR;d!fCWLKO9ikGavg&f447pzRNG<7x2b8fjofy$rsqOesBv-e#r$c
zrGM}T_QlW0QGfPKPAHu6T%zomACf<Rkw?jM<wNvF4xUltIbXLP_r0w{d-{eC&TUY9
zq3{L82Z~?R@5NE=Q1w;uU0u~aU!BrD-K)-=H}A>KeJ#I3lrK0}5+~Ik{0<7<qVi0*
zC7;mG?^4`bac@Aq*F4&}SL40&b8>|1aMQgU-|=Pr>K^#hPCj6KeEaSm3VGs7-5S3+
zpZDUCJPRMjF)ulppWWf#cm4eS+V9)wMSuDrKmEy_yv&c%hq^j{Mfo>9TDSiAB{%lS
zp78IPy_wH=?Abc?BUh9>^(P18d!|ounS98vM9B%2N05Vc<BxsG*NyADar9?L+R-09
zq56BKS9*bS+L4Dm8!l<j9_ZV;=oNnXJ0Sc7?!Y_$&&sp!?u{DH{PgSD@0ZL&Zsdqh
z`m%oOFdlyKfs&_Z^YSm>jZd!PoPDnI4u6-<{#cy0Pg1{-C)pp#v*l6hB=P`pSN^B2
zF7C?DQTLzB;~tJY%)KjlvwF9Eum18gd9`}C{K7p!`v`f4-)VTZPU}F)u|(}R$icI?
zFW!om@&)+dyZ7=8?>&>h`!B|qPnaKNAN-2lm={0B5ufe9P~YW=_L(TY<xlcEfA^5x
zsmIBKoxiX@`K)?|XZ9c;r$_e64(0RsNq>T)D7m24gR)!HJayFi**Q7N-{1p1z<>UW
zZ|&d~`J(boc^7+@Z<CYynD+8w@BQ5#_OJdXzqWpLH2JyuUD_jik<Y4sz*Bj;d=%ao
z2hOoi)c(PK4$b$uPi43KfqYSN_Ka`13diw<S{J<45ALXwqJF30?;`noFxs;#{F4uR
zCO7woyth7fz<%jZo=^VTxo>U#{6ZW@#d-I+*{itgy}ZEh=;SNTk^P?6c;Y2Fii`3h
z{-V9SjJ)wJzQQ;2%2)OGT|e<x{^orN=l@*a%a6p1I;uSi_qUbu&u%IIs-yhv&yDi4
z9Yxs}dGJs6Z6AipkNJ&y;Q<eQ>%ShD*xzzLNnGJ){MY_M{x8p!|DyJ%evi(+;FaIc
z!Xt5wzoFK}A3ej@d=F3Im3k)r*NnRMuvO=5ZHkZVLtMx_%<m}ZOFZEB_<;-LPJZ5#
zH@qZ2xMe>io|spi#dy}IPEW4dk&8MHT;sR+bq^XJ^d;}$*T$iDaY6or`W=V$;ln!E
zCHWbbp5Zfj!3B9Nyp^}2)}zix@9fri+KKz*Wu4-uXZ9!`!k@V1_cZvk4s~zi$ZN<)
zo<(o+GI=__<X!BJpV&u<|MY5{SA6;nCtmmMo)$fd-{zw)dU1cyckx)glpf8ezG}Vr
zME!jYdZ2H9CjVxa=12Lnb;&d7pB#+OFTA%t_CXKkN97gdUZQXzDfuQPr=;wWAF;0z
zg(FGHr_|4T>qg-pN=`}DQRtoh$*Vle@9kUcAD!pPU-^f5wC5M}%>MYH@ze#9-u29=
z$9v}P(f+%$?8^Sa{?R_){?YHB-MbL4$W6SG$HGbWt&Zk?AskbWaUVgP)6RF#>{)(p
zeE!}3yW_+6EzCn+&U4_NeWUY!=Y;g6KYO4b^5H-5(RtD{4_<KC9V7Ovu9@@ob-T|Q
zRr&kCSHJ32XTRsMq3YH9&0ew3meG~-Eay}9ea_?jzXA7K+_!ml-MoFCJ#S34V%E?j
zH+^idQrB^Q_0a1+^)Kh$KUmrC8&7}dntm5%eE!OA$<6QB#83N2cE^6$AN}xW`q5r|
zfH&lqlz!ke+(`Y|CyN*Q@gsF7btZC#U-(A(JH3ks@(6xM4&o1cu}|dpN##*+7=?H2
zT>NEc@+W#A7k=!1Ec^9ddvYg7`r${YesITlsJe#V$<UkMWy%NHlX>ZZ{``F{^T0Q8
z*7)Kae4q#VWuNMi?sL1(tDiiBUE|xj&4*v(iKpz#d;RfEPu52csP(cR^Qg1>?)SF*
z#ytEIAL>f%P~PD0gsDq7cXyu-&WjK9O+WO^PT{b+*9Y&t@PSv%9^Dcb_>sRW4QI)P
zAJG@;yLG_>c;*?tp!^%<Cnd_C>uB3g;=Z2qDSj!hl_wil+>`I~`&_R)!8q*MegY1N
ztM)nKt-L@SW>4lb9$ez@@>G5)FJ*`9id^{vduEsH8D8=?dXpz;PmkiWam97~@Jse9
zj(Bf9a1U<i4=2nA@4e?Qxo&<ZF0cdl?EF6xaR$y>4|!P+KQXR4uR0_=F`n;m3LnOS
zW9s_OYw%}X@Phx+6Th)8_DHY%5Dth7_$Oa_U{CZw9>$S(SdaC|pV)~yxbr4>WF7c-
z|4}>ZR_7&m<H^U>J^2y863<b73y1U*Z|$4lFkFOdsQl*aA0Knwy>mxaDElA}a(2!l
z?}XFN5uL}O^lu*glDjyF(l`Cn7k<S(-{A&4XD8b8H~jNcbhjIC|H9+5ds@F;bk{-O
z`PYFK+*ChO$5t=1?>>6L;uGF^e}BvG_1pvScTd<8KG}tQ!+8|@BA@m;_TT*jd+vQz
zQ-AA#6WLzBPAUID=_9Fq^tMv=TcYAZiPB?I@q=H(XV1QqpExAXwQkS)u{UvvJ+WKk
zlD}v9pL~knu}^lL^J+)0@Jc`R7kQF(kOx19H~ORO!TjFyBl@;p{JBp>F7N<umZ)}h
z)Ox$6{8L>CpU!{m=lLT%gn#ba=ughp>F;>TbM%uR!71{hZ}vey#z*yMkMx?9pUB(P
zh17%OVZLjRPjQf4$S>u6_%2cVx{|;9UZUhyqWCCL{FW#`DA7_s@70@2R9(JA$+eF1
z&+Vn^?BWodaWBcb*g3u7pWWb>y~?B5zxQxKKE__f{W>ZiQXiD}xi{_EeokFV{wP0^
zcly5>>eTQ;9mwxt<)!is_on4#_P_E@^W)2Xp``LjxJgdp4LQo^#07q+pL2P9$gALv
zxMMtgt23%!z;k({`%CuC{{M=0@*evoRDL2~mrokU_~Z`Hjq7~F_}0T->=)>pU04_U
zk{6L1d-N<n65q_@nZ3%7l8SrymUrQUUl`YSRGcF><KvHi)0=qiy}X~_;2Xt9iHgtW
zmv7Jqzad9@weR4E?oD}?x7(-L-?%4d9C||Om0!4rA|AmTaVyVd-NV!lZi(a8h1##8
z{HH{%3zb*019Boy{zZTMRr|~j#dmoUdlygi^B#o<DF3m~7mw)|j>8#s3HFu!#0B{0
z?;Wr^_vPTFang?EDb|T!l$^-X|3h*fz>nyU-|0^tz9(g8#zD2yAOGw}yi0w_6XZ$i
zF{pDo`BT!&&+@sH$K-qbq2^DjF2yg^CH?+~zuR9s2l8xRF22DpcrKo*Gs)BB5#-Ds
z#U*&gpWw83=zf87VRbw8L3z2lhIlBCBro?T)I;R;a7O+s-xi<PqxageW6$)*&gq*y
z`OePB!+mAH*Mrw^lYfdw>I>q9I1NAfIsWNGJWWb({F}bhzxg?w7iZMb-KT_qo~;*k
zuU0#LpVau`lzf7G*bBd8@8r&|P<ALTd#|3)?#x%B^k6>wK6Ygue6UOUha3ErJ-f%F
zKfA`CaqtbFtOvCYanN~)Iv%?>zIY%{W7pQlp7Wj-zWffsy6|To;ogq*;!j-B9<@&T
z<@fxBoyfQOgL(>kFc1HP7xv-!hBNBmsC-pC?UcgDq;Nod$1gmTuk#aBehaVVPx3!F
zVE@9e;4?Y<9hUXN3G$R@pxTLV?9};={7XI{zjM#nxd$A>pZ9It8m>pumvzfS;Ft3_
z_+wu-=@X}acjZe)RPc(N;XS_-2iU*$uv2lsITCwRmmx>;ay}&gfKT)Rr<@DHUB7>k
zM|$u6LsIkR`479NM}Cgq5@mn-@hkBd|MCR(DPNLD**_&^cj6{W4#qdH`~lvOCz|8a
zFaGTZ#0z;R%AfJCo%~Ur-YJ#G>)$+o%7>S%ezuY0^Fug*AN{O@y~8_shIq^WwPSzu
z;63?zh7;nj_QsXZ+5h-X-o_z6_9QOKFW3z{woZEI2l(;#wB-ThgW`jIzzz0+vQP8k
zi@e}#&QJgHTk`PSDTR+ot#e?()?3GK?5$Au1>OJj{L$Mk+4SL=eJ%S}=Rn%|9fP{F
z_u4x@k~etvce3q&J;QCkQ!uXga2Q_ty|C~0<M7is-h0+x9%3AMIb24?Y4JwC%yYDd
zOZ<^tzytY)@!+|6<SW*zKmUU-`muBRHox}bru;B*R$hWH?d4m>W4C_CCSIz)d3G);
ze(-bltlsFnUY#bXdZ&G$zyDx7@yb42-12u}+}9ER#b^5o^5D<*7yJyr#`E_k%m=67
z5}f7dzN7eEb^N|#MxNVWp~i*Fa7JDwkI`S9SG~)<K=%suBTsfjkMsiP?epL~{aOck
z!$EaIcm>C_hp+4kPRb+YX>gIhvN!c4e!!2sXaC;Ab^aiZkt4Z^+uGxk|KOLL*t_w>
zH|Ja0i%0y=_@3bn{*0@g@yN%1$v%x-;DdGJ3$<Um`0_2cO#WtH%RF!g4w9?9+qmY%
zFFep5j^oFBe8Va0#1FlZD}Crk58}PNoL#aj`MW&MchA}xUp{Ai_J-0sJQTOp4UFfR
zpR+5tp+3d0<ol@Kaf`pk<5#GC6aLAKpRr%|MqlJiU)E!s<kS4}GI=gO@nQY=#)oy<
zSCB9NAa{5OH|UETQSuW<oS(pB{k&f^f69!-{XG?&cAjNj>b?HnkbBhr?hwD!52aUe
zlbygD&)VUOyx5QYu|)YVT%<4l<Gc9enOw=)d7*PS_Ca5s>BsnRf&Zz8!W(uZZjcA+
zJ<85d_Ew_ohu-ZY#S!ZwANoT11^K{Hc?&*J>qp@h3hz+3fZ_`!e-tjF@<4H$JWCWV
zqWm(cyc3>VpLOeRUE-{H;RXt)#eM6M7m%0rkhl5h0aYKvN7@TJ6L07fzi<qNZ*YP>
z;0ixC5C1ejsvrOH%)a=SXMUo-LoVzLzx;@N;Ul@4M|=Fa563^n)ASd4qxIQuke_FE
z;XQqkAHOyazqL;IMz8F{`n02Gcx9i>UbIK`!v|_T=0&}iSE1hLxcYhDro(kR`poxm
z&wG5D4<#q^Bya2RTu1TY@7~&f*srh~_=56(c+TGS6L-WR<A^(+^|vpuKhsV;<KO&?
z{pI+^vk%Z8zo_}eW&0-YlZxl~6HoDF9`fWraFE>O!_JSy8C1Sye}vj!YA6434@I6N
zPgUn-|JEaZ8JC>lj5-6n6n9Ym%<lNF_`=V<*N(kfKir~Ec29r&*t)#8esMbO4IknQ
zTvtbdbNs^Z@}2XDH~v1Ib5wk*ciM-#_u#wpVs&NuU?0Z!-T4stiZ9mTJ-IrM(GGsY
zM|fnN<OnzE8LqHj_KHux+Z31RDdT`~)f4H3-HS^l>OMt@nzuy7-<?IBYmhU!IS=!m
zeWUD)p6HXF*rhrfKhfSfn|-l5g!b$~dwhx;Z8{up_*YVTmnVoHxen{WHyq$s_Ur5$
zU!LV1<d^ek&raY2J7!Px4A0mfyA*feKMLp2wtR;FuabG-jksWch>}-Q`$2JmUC4*|
z9csNv>Dzgm`!H~se7dFNDGtGB@rr-RtKooo$cr9P>txrN*UB?de9%jY+MguVUj1|5
zS?8U<etKU;@AN>=^eAq_Y3mYityf$XXXx3palF@_{h|EKxYpg4^96GUADX$eVH|Ro
z$MO^T9!kE(V^8{tlkC|1;y(W9o&NL}FZda~q56{_DvpR3;+t`NN6k;4>{5I3mXFe-
zx;wq0sej|if7mBEqSi%@@^<~;pzr$8FS&WPZ}hAmedW095e|so{L8a>@ry6t(Y8Ge
z+n2aP&-n9RKmJWl#wU*w)em3R=YD|tiTh<GD&F~>DtU;T?isijqb_aV=y!P9;U6E?
zO|I?*c<;UrIg6*pwJvhRpZJ5CPdoAzukj_X6VJs-?fE(TF(3Z<k9pV|dH9Zsx8jp^
z!Xx(S9Ev~r4(IqQe>5Jdy?quuP&Zc(=hx=34snSb{XGPD$bQI&JlLnd4<w$uZ$duw
zM4xa`{;IwEQ_k_(hxhpKo}P@)AMnAi=$T*ee>ft4W(WMsckA}uI?Rh2m%Qm0KKSl_
z5K3?Kh}!3%@-fst%((0vf9w#o9u!~V1AfIte26#X#IBq_>M#EJ?peGL7u6BOP0!|I
z@A?^!|F8%97kYR8pibxS_~HY<{LQ$N5BbmA5B}~G4f!p7*w6Dj{HP;nrygK^#$n&=
zkzMmo?Ztn8m%+IBh6AX513QJ+@&xvm?X>4NaF`#nZ_oHf*&RLb3;1Dt{oybFW1rS*
zKK2S%*e6O({Fyzo7jj}Z&iCbw<RXuOOZ?_nw_UVq=!RV?axp%>%wv7lZ(ZasUU()4
zeujT?v9B`@KVp~GWn6eoALP%T;X1#vf5Jb0@L^tZ_Fa4YpyVRYlP@M^SM+T@@6E%W
z=)ro)%Rcw`cl>7hy{k4gM?SsBsVgt~W7EEvT-dpMhWyo+ytkiIe{!!Fwf}R!*|U0)
z{_;3_v_5v>@73VXeKc~D7ohaW?@@XXU)ZtvQ1R0<J7wpd`5pNiS3OsI_GLWr8m@}N
z;(>LsTlnMuZNod%IP6ZIA&*MG!-sjufqmg$KkGyJ1wVnK_O0sZ{3R(|FuwUqR2)V5
zBRhvv{Li@7NniBinIFPUe!{PegBqWF;Rm}DH^~Xr9zX0x{>=W(tDk4%>Tf)Ju`lwG
zkFzWD@gs3lJY|p0o7~SPXSl*X*(JFrWtaR8&f$YSTaSDQ|Llt#__OuWqxa<DJ-;)L
zJi|QVpJ(f|K7Qmo+_0}_cjTguqt284@I?;f!Y|0ry7>V)+LwIjXKU|1eB7pnIBZ<&
z<d4?DkJ+_$)@wc9TbJ+hR^y@Ol`py<VSVg_+}zW_xBQRX$RqhDALEge{^TWIdXLIq
z;hb@-m;TJpZseQn#JbIoPwPSXsrUqc**)B0ALc7j_*6$r{X2b64oS(OM8)G06;DbO
zzWMvo;=l7Dd4l~3dCEWFE`KE_IKeKxhc9r^K3knnKlb20oO4EY?0%?c@z1`C->^61
zIiFV7b*|>`LfQB5d-~vy-m_=(v1flLh+O3h&Uu`N`a7iZI_DtbhP+E1ihSTTI~GUC
zS3Q7S)t}V6#94j<FXWr-#C&ke?_J18oU*@>Pv8$<en$kS=v~~?4_}_O7gxo7zYAlZ
z?9O-gLq67rkM?-WmX9ACds<Wcf*009j^Y|zrgw7DPrUH^dG!r;==W>Jfot%eJn#>n
zj8A_hYF~hF`r^O*n!l=x(HETe`v&`yUrZgp{_Q{6*fhR#Z2JcGiZA;aa)T3jj?Uij
zWxvhtjmJ;CXSe*td7<+z{rCfWBTsrVj=Y_}vOjsHx*@w~NBoQZ@N@YR`)8NVUCF^d
z%()!-!g1r_OB~^s>dy3Mo%o?=|6f3R_Dw(dW2fv}-oifNsC+^Em0yVe@&R}$Pq8ma
zytH0ad^Zl<#s|H^WpNQM^Dll*9=_v8KA@d-zzy<XH{?PN>dyA}zH5(P)H>uve#hWF
zxsW?5-okNy!an(dxZ<2aJJh=140~ela3?AIgU9TMT;VZ)67R{w`o%~72M_6yAMj`R
zAYXu|^rfHg?1Fy9C)9lOj$i%733(<vvJUT!PjB?E|NpRe-{H1a)!Fb59N^GIG!z9D
z^%WHfcA}zYZLCPML{UHm1Qjt#P%Ob1q9lq$G_fm2B%mf@1BoJ*wE+VH_L5LU5s|7$
zIdp;Vd#{(@{;v1t8sB)<+It@``p>?vG3%IPjyC7>tY?c0#>E#q<sa;ae&lX_=Fy+r
z*sFCI$GFB(ZXM#fe&(gG^nsmNhjGyC6Y_KTH7*o~=o!CbK6XYv__2=U*Er$<{l+)>
zLGmLXDBqyR)~P*tTZeM|vP*HCexs*FawZ@0$a4IUyY;g-{tUnB*$sc4NS^Kw{XS9L
z^1U*06Hn|PzXZiK-z$LPwfj(cg?mi>J;xQ-jpx2bej^{0Pmmvfg+Jd};1`mAdd1)R
zzJxrTpEWK$k{5{2<imezM}P6h53v*eTRU=)=jJ^4^_>!anxAL?%H@;RubiIY3k{9q
zUY33Eb7<p<1N4sF;fp`#U+JYh%l((<f%I8_a*&6S4?f77AF*HLOMl2u-Yj3TZfG6s
zPJBu~V;$^{UtquUMeCziXnHi%@6a#n!>@7hP44nJd5LoE;5xTHZo_qdyz;Dbu>aTW
zchP<i?f292V83r>@17&cV||ywb8>Nz{i5lC@A1e_?7#b0`2c&6Z?j+j&yHV8zvF#8
z_s95<$MU22r*EF~pxL+f`a4${pZ}qEkUui6an$R_&l*pA^W(?7{3zPE<Sd_n)~TKN
znDx#<@>X$0IZVCK9^cld+<QFwJD;NEm1z5?o&NL=swcOk$(Nk)WBuey4&+K+^v1d%
zJ+Lq4!3WxU%qPEs<~JTbvRwaafAMKt_Xqm%qvEW2?34NRH;;b&ne~zbxmvGt0)3WG
z7#GSH%;O#bKM&jeGi(0kvXy3Ex5m=9zVGx^W}kliKfdJR!|!rCG#`EOonk1D;U~0{
zKk;kwCiV2q{>fk1kM+xM<Rjz?-7D%xPTG^NeUa~(Up%Cb<joJ$XU}iMKmODGiE{vd
zYJPgdud0_<8%JCe&%`l$YF{h#{RF?O5ceUz^~aw)0)JKhBK~{7Ngg6@7+-#5oyz4~
z<OJ25-#xp1VHfx`pZrE#qA&K}y%m3>UcMv`R!{%f2YuoP={G%+-^wG{89PsOzB0af
z=`Y&;q#t0<_{0}KB<>rB+>9f?mfzYR`J49Et(^YIljKM85PpT-;D@}}H@P{_8qf1F
z`L25fwDTQ%p<i=(^KJhhMmu^ckA~(`Zyox{kL1DlwZHOO^Wum8${*>4a{;-c*>#TV
zeBqqn9LwJHhx`t{JU>VCE858ieZNJ%>ODF4JN$=o-$8ZnXI$^mImdX8WL)Pla>56s
zuk4XMvm5r}dp`7D{6LF8;-L2QjNf2q=H+kf7k<f~oR!Nn#RL7x!~T;qJCbknfAVqm
zNY6Ys;YZjPz2w)7XP@b>xJG}?Gq?VK-gV<+Zu#wqU9k(lv!j>BV;}UJKXC35r?htu
z?VRLZklvu7b0#?}=MV5_UGCfSUYtJDTXCG9BX{F@&Pgux(?0Nv_JRNN{dVs`$ScHc
z{5u~iw@>VXJmeqZ7ro~{<qa7R#7F$nU-6E=w_o(xIQ9$w<ffe8AwPORj>^dm&93ML
ze@1>uE9ZCU75_gevI~Bd9e94s{^cF?ULNC~g<g;&yQ7EvvGv#o^P}wxe<1##y%&rR
z^5aj~qxX{4+gJIU^`hk$<V(-w1?rvS<d^&{JEjN5Q9p;b{T`D2;>-Tif8)`A_CX%v
zF@3Ur>s1fUgD>N<Gxp_NW}oP(bs0~8`aie6+kVeDSAN@i#TD~wk6-q}?#xRc$eaC%
zd;FH?&h`oN5A+MNYv&>QN-yNS_%uKH=&#&3&^q`-d}qBpM?HPBF7p}3eB{O7krU(>
z64|kQj6BGjJ|sHdSQq`pr}pZNhtE6*nxDPWuha)|lwZqufe&_pCTIFaUp+515Bbn*
z`%i!IPk-5|dGHVIhxX>>ul3^>ts6~_>=hs6ZytG*Jd*y<7ktxS{c?Tg=SR$sCU5xz
zKLq6?_>=cJZ^+N+o%~8Wd5ic#Ui`cD;X~ZzpU9Ow=n4NRzb0Sx#^*P^-{twF@6gZ}
z`)(Zaux{^>c^^jJr9C|pFO5$=+VR8IjUVgqoiE=5^Zj}J_%4KY@-BI|@3@+eyzws&
zl^>dy{*bGA$X$DJ*SyvveycYgKY<^5<@>z&Ay;zY|K+9ZggnJj_C{aq1HS2j^OSn!
z>8H^6qL1<>{Fsj%tjj*}KkR{h=;xfsf3XvOPTnLxw|;u6zx<P)I`8re<f9)u^j;9Z
z0-ewJ8GcrLvY*o;e=-!EYkkMxcT?PVyH9n_RqpwLd{rJOuFG$seRu!myrbMblXxW#
zL;l!)rXRK+;-`Bm<Jot5Lr-(x?W6q?NA#md8Sl{ImbhSiam_l^vp4r1?9_WsP`Nz7
z`}X{RdH4bR@*nJiUCYnx7kNN=AN@3*dgJnc{5Aggeg6;I_h@$B=f8eA`{0%4zW2ah
z$PMyO&X?{%=%449-kbEi)Aydyo@cQ;c`BsGo;SEZbie5RY<808<h*}$e{5ZzBjDTf
zFwZ&Jk8<}*dp+`<|M{cWuQH3aAHMJJ{Z98nXzkJLz<G>Z{GZl6f3@z+=T=^MmfSqw
z;RlZ0`MBesb=oas=y@{#?RT)^CI2i=(L4Ii-uX9n!;iY35?}Za`sW^A{v%$%%n#%V
zzP|;<aqAGT%!eO&9J>*3?60`TuK6LfcI?x-=GJ-JcOLi$=LvR6pVae9{4M)X4&CFq
zw_(TWPsKHPt9AN6ypX?u{53tJpX`C(BnRg-_oVFE`qZ-r`-?CBo*#tbv3N^<?9DmW
z`uPF;iWlT&zr}y!%Kz|5Pw26E`L9I$(_?XrKgEZ<fM3sX@NFD&hWK+1&h^_LaxlL2
z(+~a^UugD6KgmTua>l>+k=Q}5+c}3H;4j2$@mpMGNAe|dH81|~P5+_%LwnDO$cbF-
zJ3EpG+IQy)@s?k6KI)Jku>bVUIP@RVOZvf1#3k{={Tg)N!EZSSlCOA8FZdbrkcau~
z52P3Dz`E><yuiNmx6UEPrML7Es>g?YLu*H$?c<z1E?ai`txvjcEDzz2=r29u*Z2+b
zGRGw!==^M5@*?BYUw+kk?2CEr<Hb82aPA8~c$XP^KyKa##UFbyE`1aS?Z0!Sb>h?a
zxaFNg&kI#MaW9d6(5FOxqe5~@v@eicq5Xl@4|6{H0__{S8;bt#gx`T*^MQZ<+~bbF
zW#m0NNN@a}?&{ARe9Io6TxrJtm5|5K1Lt7*mAqSiV_)5Cuxos4$NsIK9s16N^EkUT
zo_h@U9G>T~C(l>;Gx@*Y?K;29v+yB~%Om~Y7;<#4B%f4HzVzF@igOXU`YtRx;ji({
zpWAnSUVCUC$eI1g{~*6e{^FMYP`!9&{o;n-NwOz;4!tMDZ#c)WH+~SN{kV5_elU*w
z!nuhbf${`#)A_=A%KZGl`HffkkE)#fJLJdc75hT-i}X`oY2Vlb{*|NU75oCf?3^I}
ziAVa$%lIApYd!jjfA$GK_Md(lPq}zVpYa9L-sqM4M}Ecc#K?_ZDt8Y7#T&HWSNXj#
zefGOG<M_Wa{2;rfCs4fe{+@HK^SSo`wRbLce)ay}fj@rcLF;_?=CS(%=XvLSD1V~Y
zeh(oJ@V*DT6NlBa19_C^zRox720FK~6Zs5(P0!@l@@M7V&k*07gVeKY=QHO4?dcmm
z<iGiKcEWBTeU;bqcl<NG5--??^R{;El^>KhiBtTnc*NgoCl1j&@y2-itEboUQ~I0j
z{I1=8nAds2{;*GZ9=nA6y8Xs4JBIuOJ+LqQ2>x>2*2zxoL$1?!;-dUPd;9D@oZq8&
z&S&;fd~|PVedOd^A<oHD?4x{xeDot<H2o%D{@Ohf`S5e}(m3K3yA<#ESMBLB|7joi
zY5cG&@?<yUV%_8k`8WG-9mcgjNPn!4oRTJAd7F8x%Q*ZaeWOp>$%p8@dPtx6Z}CDu
z_5krqKGw^g$b}xNr;qw8w;%Qm&HtkHmuD&$CzTu5y4+Kc2Ya+X`tcj$oA;jiUG|87
z`amDa(|w!g2<{g=|B`>{=RG3NU!0%hd*Yk8C~m122i$+q2m1xtF+HT$^j+SGW)I}7
zzk2zOa&omUs6NNX7d>L<_5pt#W?pH(>5KQO<caFdqdf6%k3He#haSJ$jC~bX#aZj;
zca*aO>!EjOcFxb4mp>BE#4pc<$sOPHQ8~M%|MbuEPkM+S{+C|KkLZ(gHb0B@J^*BY
z@@Gij`6c$kZ$bCu+R1nM1A6T~**=I{+RL{*zp+2!gZz>GxhE7C+@G;`<5>q9pXBE`
zl6=s;y>YCQJ-Aor-}Sd%>$I=tV+ZVs+~g_Fq1NMG(LJaB?Ap5M2YX~U(76wv?A*BW
z2y*4m=#71l-};V;^CEvH@0QQXv+cKW_(%Ha`J3}C`9bo=zjKEDVkhKmT=HNy^o%~D
z`BUcr<@m*i{pAPEZ{6ng-mdoMCvW>k&h&xasi&Xplm5~h?eJ?~=?^<*w@^Hw&rm)6
z;^)xTN3YF`f8*Lm`%GTO5r3^qIU4`uZ(jL?b+H@y3Oj|q>j~*MeG{+gsr?rZ?1%eb
z`)$42+h6OXU-SxE7yihdo>;Gax1ZX}>-D4WiR2~klppa2?43OupMSxpd7ydH9>^*6
z!Fh#Wv+wdc_Z-eg@>lnoo>#kP^`1R@Cui|i{Dk5@|Hls{zx*71&|V&<y*SI>`8j+T
zpI(p$f1q4@dT3nl_lV!d!v{SyF1@ke^p>8x_n?>dPoAS6G>`W5k9}yzuHCno7k}<+
z%meYwkK@ny;t4spKO!&tME|VQJrcd<x7?Gs=b&fuFZLndgY=(WK=~hg<_Fma|3H7l
zDer^uL*#6Kt<Qcc=Ld2={CS_wds^(vI`GB*l*=FGnc^fr%1_8|oCDdJ`vT`#_9}kx
zBgQd)_UA{*RXhB8&SD+-HxD`JhktoFe@ZX;1^Q<FxeoU@&^Z&zGsyvpU-+?)_D_6-
z?92H{Jma^VbKDQR_kq^OuIQ=#O_Xmy<3RQW@dw!zv=7iZD9>|$q^FYr{c{iNeu{n3
zbN0cW_yPJvALUu@h4@2u=Kn)_50HO^`~hTd_MaWoC;mwuKp)wyag@_{G`qkzd$oS&
zYyH?KbRSM0&PnW-e{v2M7qXpv+dZT>$X^@JI^EOrWA@2C27l@v-Fqec6My1<fPW${
z>(2a0o~NDtuuk!W{nHzK+c$oU-1#f_$>N$k-18rCSX`Dz$Xn&f&h7k_JjnS*9G9=~
zZ~UFOORwpZyh%R79{4N$#9w&~R4<;Q$;UpE8~?7I`N-Ki=%an%PuY+AcKSi@-5bf5
zJ;$RT^nzb7zV8Anr`Pn?`q_>BU4FyTzkmBitIV($>#+{cC&-I_@%!|WU!muT<YE2R
zN8c;tXRJHNWpC_Pd-=C>seDXcD(|P)+RMxA7rQ{yEB*nkoqUj<&}Z^vhxFRL7yEGk
zV?1_cU)Yy?T3*AB=q0;$-m~BQEc;>4_LUvD52qK_&wkPFBk<?G!@T$~KNP3*GajUW
z+FPIT@ej#?pSGX!Fze($*ef}ci+y4L%C$2N{=KIHr;GZ3IK9LdanyaV^F95ikM7sm
zx&4&i^UKZ|{D^Y{zhoczOGxtkf%B;E9C}{kd<U(A{?Z5jO+Edz9(>S4{)+$SFWh_a
z5A@kQ<R|{|6a0vA*aN?y9`a-A#Zh^jdsypq|AA&N>=Tk}qWwvf_YOtzEs?*nzw$@&
z<*(R3|6)J*H+m&rkf(DR|HMzoC+(;Fj=!?++S7A#m#^U$$`Ac+P&}vK?1G(GhrA6R
z;w^N)=DiPc7th&~`#0^~-#P!=N6)GCbFPrT$RoUW=sPm<SO2fV`-9etKjr@4oB53I
zeZ!BQw#_vU+2Z<<_Yw1c5r4|%)82!zUU`)L_I(EWMIP2~|Gi&D?$&D`&7UZKR7f8Z
z`9b#2&(M!V{*xbbpU2<0=ad)GfAN(+;E(tLdd>d%7x|IABKeosy0>#qweR!;pX?Bd
z%N@#B6WL{lDaVTL#!=7DPKxqB`|h0N{?vPN;)4AVN9;3w5%1jp*yrR+oRHT#*VzYr
zITxDWxkLVIKIncFU*Z+oxY`>}IsHr&AC;r&3B6R#F3gXH?7?$s_t+4>6_OJ<kfZk^
zA%3C!**OM(#!+q^_E9<X`!2s5KK_i0_TK&xE6qZ`*Yw_sd3=Aw{XR6V_dy~3vrlMx
z1-1A8H2j{8JjvPp0)F@(>y;0sewzPN%m3;#|9I|_*<6=%2EXOp$1dnMe(1A(+~u9G
zymjk8zkcjIM~|&1^B#2izls*u62+Ge#or1~-2d<$pSsD?nMChTRhaEN+Pxxt+)h7#
z-65}7b!I5~zYrbDM_|=nelrw5cFHk7Ir$AM&m?+Zx<h#l6yGZJUUr4@o(ijRhnD*t
zI6HL?VsFkf;-9>FPLFT<{-QYNe868hPl;QQpVdwtV;p%oduPA=0{`hfa(<58PK*2j
ze@!mpfpY@+kc&7ocRso8_r%tTf9=rXnsV*Qojl1+oW>WwCw`#i3;Z^kyp4+w>rQlD
zvJdhLh=2Y3Uj07r*mB3Czqn-9dB^>d`xJ8I$K{XYAYMDSk+*v=afUvapWmGp$=&+s
zBYERH<%U1<AZL1HU-k38f%gUU*A7|-Ju|=Y@heWT8|OlHrJZtlxbuzu#h<)_T*-r;
zId7+((NA%d{%J=K>8bgQ>-|RM)`frTAqRZ%N6vTdV?39!FY;pj=#PCO596CZ^%8&N
z%}$iN7jVvZZr2`dU(HM2^uhjHe}&F*6<RNJ9)$csqUVxT{PcXmy4bDz{|@;}_YuZ{
z?*GjL=_`5gkC5LXSNDR@Ip6s1Q=Av=5B?xN@xcy^Cm(U2;6A`TA#^X}IR`nJhyJir
z=pI3P`sLi5h)?aw1LkvIa=?fCjtZ@ty&1Pc@^U}oUd25V`LJ6^PUOJe(D;y__`Rup
zOq3tV-{lj|Dei~)FXvI`Zg%5-lwb4Sl78X_e~)&b!rsJ1{z<;UZ+U)i|Gl?t9C6Bi
zyRYOA>51p%{H5~{KJ*iR<b&kuIgkCepWeT6e)0TXJ$-kMA%FU6f87Jo5BF;F1kX$H
zlm5oN6TR^Jdh2Hg#$jjd8rlzf<#&?&1^Xs<_vhx3w|GuzUh)>l?5F!l_nyX2yAux}
zy>%XyZ_{h<(|_e2dtScZCzj6AS9(al<W=IJ_$<G6J{Mn%Bd&{E#vwm`THJx+s(p2j
zXg~RJ<BQYsNcj)D7I*2Vdl}>LEBqRLvw!Z->5cnI<FSA8b#JRbd8lWX<N@V3?#<GV
zc+cMbr~eluAD}n#M$ez;?Dd}x*!5jM|HjC=J#TZ*%^uk&{`r0W!}`TnNbc<4IP_ZF
zh3tZVanE8rabCPIzy0T@^d|@Fh3wvV^o5?$bNOi^eYbD)6d%UX&-na4yU`z?>;`SW
z<h$%9QG6p8=OcM_hxD*R{E?6MCh*IS@XcQN0ngX?6LQ19cp}f`59y72ICj9V`yMFd
z?-J8L9=+OWpZxfTt{%-7*+qrofcQ|M=U32kG05MG?|#?lKFf14_cZ#MU;LEs$mhgi
zaq9=SeB>X_yJg9Y{MYwTKL0Nde$U^ZvGR=XbiC@ECm;Roy;hr1p19$^UHQA!Pg-T>
zhudwi)@rX^I^$d*kHcq$zWatRaW3yIt>eK5?){b1)?al7|KkTe;IvJSUNS>YiPq)5
zjGW{h74lQ!HvcNFuxsaHevo}=&)(Ppzb}sQW8R08U$Hymxd+#ue}VLp-(tr<S@y4w
z_~BpOI^Oy*|LfF0-F2l|-&=AYN$%cD6i@6QdGVv-oP3kL^8@^o=hX5`@j#v>-_wtL
zAU-@lmzRm}&~sP(kq27-g>Uo8zr2ShzKI*|Ma(ZBf%eVw8T{&JesPWevrp+i^)o&>
z@{{<I?|9!-x$*JqzQX?SyZpHLWuKg@jE{eL1AWA|`;W8}>oX4d%cJ;n?>*WF--(cS
z$x9M_ufVz5cZlQ#*2xblhxoBxdW>&+BA=mO<SHMbkN7w4P|UdXNBaMkfc=pduy=W+
z{i3JxK6w<ovro>Oo?pt_AivIT*av?K={<ihPcjbsaQ;H`$G1Dj|Lo0goBhCbW8*@3
zCbVu+5<i>^*e5>aU+f)<XU4N`@`9dgLh($VOYhl>^PTq2N#qaZb>f)wsB@0rf7vI`
z5B&bvb6nr)Rqp&G-qHhhOYY*SxNTp^hhH~dhxW(1$Y1_0p3xV23g_bDZU0Zi{PKI}
zWpRr?vVZi>yzZOO^uYYqDc-Oj{?2;rFa5~&<m=olPO(>hP`UWTZ<CAh*|YxI^V`<J
zKE)9<{;fyc!iRbA<J@mu^j_RGKiYiw#;5&Jo=BeN=P$@lJ(O3XA-_Ox<T?1p2R})k
zP&qrKKk^oFjXdUy`0kLM&`<VAPRj9Xzwu43P`P%A#$&JMXGf48@wYJZio7SGZ}J`J
z{2?Evf7;7KoQK_?8b9+q`)HqxD{jiK<k@~t==qIvmU8b;I=70K?pu{xkLM!Zvm_V!
z7dykZePCDO1^+BgiVO5xym4O+#Xs$gORoG<o=2T~V7_PJoU9%#KXoor?m30>L~_Ot
zBu_L{4#|}rL;lYGL*@Lpdro<@d6a9%Zj7THq=&}QUpYRj_S!+?n4kYqZ#?^_J^f=h
z{F?SqJN{c7F<<smP7eC91J57a&wDO_FXv?KjISL#q@T*|GduHq!n)}byVZUux-Sx!
z=rMljIeyH`zM*>_??JndvR?NG+Ov;D_bukLF86lEp-1+Ue%MFttw%Yz&=1I;i<k7A
zA7xke(|)OUPk=A$mY12wxb`2qUl89tH<ahff93Pgy=vAwUl_;xFW&R?-ex|p@m#`v
zlX&ZX$@2*D5sI(w6PyR|<++poo_FHQ`NVfGjH?_U@;v#vJYAkI|MxsA`8K}irs9Nr
zSiBcEl+z34=*%P4%fsYl{1*L@udq9Mz(4Z?^hUXU;w8Duqs#~CJCp~~U+7*`9#bK{
z#8YTo>n9iSc~W%0?fpsj(EJa(N0Tc*z+XGhkiUJCuS0&xJp3&`<hcUB$j*%m?U(o>
z-x3e)gL3<>KiW9jS)ct<FCKYrXZ_@ZfBwiitk?RD=e`fRZ<B8tALcmnJ9#R<X?*R;
zLp$|;=O^EV*3S-{ujG-=z4VbkHy;1*Uex^FAFyBiF!Vbt`eYpZl7s!T5AHYc&o1z3
zJpHW4@8adf%FP3_+&J#f$-#W~0bgkH!#5h=`0zV-{p3IH!JL~spW~<Kr+aYs7vxC)
zjDtV>Y+UCq@r8dCU&UeaV>kAhKeYd7{*zy@e&gAHXgu;Uo_&$8xc9P8_FH>;uiW_f
zrtkF7d5v7;5&Q|{ADs6*$EUCOcP?|^B9E+4e|Ba6?Q6zOb|qfhKflkgPiXQN$HZmv
z(SD1^^vm;f_2$73^m_v5IdQ{tNBr7H`$Asyf<1_5_~7@<2jw~9A${i$te2jPqx`n_
z0gWd=_FR`9;G4WXcX5A+e|*q;_e{pOzxcqvap^PtX5Zd-5oheD=T+ose5jmU$;bVs
zJV&1Eeir}uQtrMO|Kk6HkGT2A+iw25F}~#0o)?)<ehTH|Xz%&aPkAZChxaWqf0d7V
zPsjgrleeDsuP@&8F8{F7EdOnt?#<~r{qcT``!IaTtMO%h@^a(xyX?@o<it<9hfz-N
z@k7q^h+f!Fc5FQJh%5XE|HdEk_v*zdd4ziW(0BSTUdh+km3u8H-z7Ilzo7L%dJNea
z#5c5V$e%&+5sE{QKZ4Fb(7A>`f%Jpla*qb>8-Ji5<fowWMDk9gKjiBCX5I9VUb0`$
zYv>dGqj&BJ6U|32>79KdU-`D*1Gzth;t4(=IeDH!e#RY&@=o#8`t@fw?3X{FKdJxx
zt^VZYxt{jqz;5JG{E7MO5C32s_3Vv&$&38(VZZou`oZ4uX`k)Ce&k|Y{lyJ=E;}=i
zal|X@hUUdTyEUJ2>;w5}Z{IUMu@mj-r+Mioy|<s_X+DTgXy5ob=QVzUe{}wFK1d{A
za<^Xljz9T2KY}lOJ8zORx!Wi6up4@8T=Vm5@(lS6`<SEOve$ku{+Fvp@>}mi`%Z)R
z#^lrF0e^Y_Lr*>Bl%;e3&sJUtSNp*MGuym+mAUuG-u%+h`7eLUDl;!!<vW+Jb+^@K
z<jsBu>z-5|CO>okFW(f0+#}$}J*eja@+flEUw$ucxv#K(?cIBlBXnP_UVHPpzxF)J
zxb9(`59|ZJ@hd;YulMrIXFte^Je+sL6Xzaie)?-3=MwTJSA5w|?XAnYjqCTjX#KQL
zxwtPgkN4!9!|;h8aYnwYp8w|um6NA=l|%dfz+dh6xxd|F$qf0Tp?2PvG#<X`r*isb
z9r7h|q386-`1IJmv%}OY^FaDUP8He*Xn)AHLir`+NBB{Gi9en5XQOwowf*52&Ha8q
z<AHOh_E5YNANW7^$8Ml~=TG&=pSYpiJlZ=akRz0zCbBzv$WQb0_)^b5^UL;`-WuPy
z%H0ziM>#u(%IO`XcgC&I`1}Ap5C{2b`%d2K=`p#nJATP?B>RYO{sxkh_)cDl>=%+J
zwEhZH4mmzPAbz0rL-`J5&xz!eeAy?+J`?rhA1ai0$Rp)RkX_hMc>_JP4)#pH<u&wM
zd*@{+K53Uot}y+TyoSD*M_hCMQce#cKQR~I$M=8QqrQ0Mn7^@J=X(Cj^D_Cr{o;r5
zOYY(yIkFq$%k!*<9E`(XKyo&Y^RhTX{_aJb%k-BQ`hS1+pWmYw<N%FhKJ6g6xR+7h
zq46v1-uqK;eHGe2NX|pCT1Wbcr1{MX>F<1zeu)RxWn6lX5Ai@d_3T<a6!+*W|3;3`
zd6xX?A>^0XqjMzxXk2#Syk~rI9zST~vwO64K<%9u(d342&sB`$cgOVJ_mO;e$h|Fo
z-G6!x?|V+3&(kA*MZff;?2%pI8^7#Ze{v%q{@cEZ+sfTfqs2vfL64qv^^d;s&rex*
z#{G-;!>o&b`JI9H8t8-XT6qt``x4#{@cw}Hspp^U1HVP??uXoO*&prP(~&oSt)2HS
z+)sP1Ab)ip@|?_mLgUa6dgHwh_u2TcPWRIG)&AK}epX!ISLvC!hHrXA&y9z_MDMeZ
zulsiQ<lgVF&(>pK+@G7*J)?Q?ZJpYaw|%ES^wqrzeX{T5;(p(~x_#!)_+9<^7v~9b
zq-XBctQ)`PrN8{3`z7~s_M3dITl{n1>z)kX<Zk_bXC^Pghw<qfIk<QJ^|QAB*t`C8
z<yrI4H*zr^f8qWL|JLU|(!C=-a~|W{Kl@LA_&4u4*%$lEkMnEJRs1kNME>*{|MDJw
zj=vPQ<U{hN3Y|L^3!P^meT1G%TPL(G`L}(s?|whXp6H$ZhU}ZaV4ud}hqPx`p66+&
zKf9*a{FA(qUD#LX{y6uY-r0ZpOW)}Od!`5UmVC^sy|{y4_K08lK_9K3U!%|D%1`Le
zp6H!={+(a89`^^(IQ&k9<Ru@0`a}96jzImP`60gPopTVBhe2{7S7<!rkhguO|Ljm+
z;k{b+$G*i~`(xidpOmlgr|il8+XsGLIlsXU)bls|hw<dY_{NWVs2{y@&mg~*Z|Tn;
zk{ABa_K#fC-pnhX!=H7a<!$a+@*WYt-Xn0Yp+C7p=UZrf_<-~h+AoNIC@+KffO+1h
zpB0i9ejquMf9AR5r=9iLH+H0)f3yGOf^T~6K9;{n(|h+g^k!1TANkTZ>xATufBI&h
z>^nb0Zq8+i=F9y_n*JryzY58xLVQ-}-f$?g_eA<oq5H}T#m_|N7WYCGy3eUF@BNat
z&mHFdBU=20_^kL*PS4%%bx8l_i|iWmgOg(AU%K*K=cMyydw!)tcItVQ`*+{ncmK$b
zi-Vp=`F%&)8UMxanTMW>$MnIyj_(SmU3#ub55;@W-^G1-m3sb!pXFcam-p8B5q4^R
z`3t+ow|$(;Pj358uJ-($^&1yo{Aso~u6W|S#Si0)-?lFMBCm0erM<jLKm2Jo6wQMl
zdSQQ|d{2G=o%5hP3W~q@kw>dHE|f1r`44nHfZ`YZR8DUyv|rFVp?!qp4(X3`xc8Tw
zqud`r=U=~{cb?bZy^VV#@^x<WJ0kK^E`NpM3b{L<pw&b74(4~S1NAdLlvktigFku<
z?KiZ3=-dM7bs{~5@})%2w-for%71!)+<6Fke*t=)Ut#)f{t)5=W<G>J@AL4hP`+i|
z;<J2D+%yh9&A(+_N;>_&_Y|OYRLCy*7ygp}7LTEF@-#nrXWr($>_mA<#aEV-*H9$?
z3dL6_E+txDh2lmceRf{<d!o$qyf5NAWBw0-_wfB*^q5PJ-0;(*RcBWD`&&=A|BWln
z_+FQ|?frc35&7;|qW5<b$$@^;L;qjsfp5L_%!i+G<4AsD9o{by&#(T>!ME)3$(3fj
zS7!ajV@G~Z>pQBx*XH;0-WM<)KCIL4XYpa*%p*^7ex7^(>}wbAGkVLC8FKR-HowEQ
zPWwWh`0@S{WC!Q}<Ze$o;yEkL{U3-8c76RDe!0WSGbf$6(U13e@2WGqeEHe0xbUM(
zXV{6niu}k0?R!jU^3v}4uX@$e2mkEak$v<0+WUHsU1P5+_PS=}8S<iE)??lDgMQfu
z@-{CyZ}`*$AN1ATZXH=aeI!3}H4fC?zM<h(NA7ad=g(a>bI&JTyzIH3USsCMuO0Nm
zyM2E38Re_La=-1~yVDvoN9_J@=dH8*vKjj3|88o3>|-|hm+yUX+1&ew+FLI@U}xmD
z?G`&da=WjuIs?D>_aC|cH4j>I?mK+^vEK`cm;9K0;Q#n(znf6cFL)o5{@CX^{_eKL
zUtjmNn@8-N9JRNv|MJ&=_N<RBT{2^T$x}P}#y-sB_W|-F_Ch}(`y>CiY=6va&iwdV
zGx)^E8pr?o`^TTY){O5Q(kI^;g!u6uq5V1hPqx_glFL?^F`sdLcaT5v|5>e*T-Yr;
z^Zu!MyqC(adT*6oTd)4)hadIqCH?0g8ZAF`Z|}K-eA7L%=M3^>`K#xN@>1yhfp)%t
z@>cg>o_m?s`0{4?y!_p`_>niu+uiTTlan9y#_{~md+naD;S1m7r5(BXofJMi57*xP
zDtT!~-p-fiF|OzR#<3311tEL%{(Jf<d5`>+zi^J@2l>4_JG%e%eB3$FdB!=^{eg1N
z^E~$<U(er+=U&8pfPC8dh@9O=c+Tw{;9To@y8UHO_E|gotKRc9&*QYSZuc$nRrC9u
zt^Fex&num;oagAb`x1H~PP!-X{2hPhb?$PH;XZ&o<@w$h5ohU}^N2W$Z@)K&?ltX;
z`Nco`>-`ez^gaxG<WIZ@^Zt9^_}kOpv1B&K(T{)Om)LRo73XTtZ}sy&i*rAHa}TGS
z-|{@vd51pPN9(gc<m-8^{UcZYfqw8i@*n-(lksE9%}?I+(LQ_rYn|G&Gx|f{$W=Vy
zXWcis*AW-I&p=-8SJ<KZB>V6E2>V1|*qiek{@E|&AD#Ez^U2fAD}Q&6hU}a@S%-e?
zn0}g<eC@mQoAa}KSN36l<URW1-~Jn)eUcaZThaNB-<`YnaZbGT@CR&r&%d~0M6Z3H
zoLt2-$RCiGbCdgb^3$JO$kY7tY2Pz~zAM4c^K<0yJ0ad%){orL&auwh<jxNom%h+T
z=sC2wKPl2<dIyaQl_%npe>5IBL;3$R_jvL)zc}=+GkYCz;$!~g*Xzz4dE>_}+j*O{
z=f0PI%P(L1oa0_`x0!#x`J<a{@c6sW{Ke_pzUet1T6Yee@0?TK{f<XGV8^wW&7D)6
z^L)?2dBJlO{J1xxSL*47_X_EU?|{=U=LY-cd;-}Uze#WTA#$Q0;uv}R9)$O4wD+DP
ze=44PehuXn;<9)ouU0Qlpm+2PAN+y&_$hqQbAFJ%dY{j@)?;4JljIltm;G>lmVYL)
zXZJqNYxc+g3$UO3h5fgm{Eg@OzH>yc$=AIvdzbH7KYek}s-68H2XYjrJ<kw_=!^Nq
zBYFtMQRg7z+b{l{f9Ds_X~+DY{Tqtzi-w}OW`21%%=>Bir1O_NK>lmJ-iNeK`YX>6
zXXX3!Q@$lHz@L23xXPjN<sFHBhs|E-g?)9uk^3vp^?e(0T;9gd@>lc#(l_?SuIPhu
z{ze=Vx9G8c^q5@d&O5h#f7!a^kNh4x$3Nr;tOH;4*E-n=e`z1&!D#zrf8|&5U3z6a
zcE``rBkQ9N@@(gS?;qkvJLvZ<{I&DBeAs!?`O*7#@?3g}UveTx@=brOopIO?KVp6K
zlN_wedgS%)E!ds)pv@zWkdt+hzx6qnurGNw`w^eaqrG@%U&&cN`^X;YgY$-Q$ldt(
zHIDNSxswO~PG00kU-4}`_Kt6U-8l43y>pK9OCtL-4!JwGi096op7*;i(*AX8EPd<y
zPG4o#d7EB3C+X+h>m2S}=043ic<!EH_s^{Pm&;a~Rqj1`{E&-sa!aJI;sJk2&msS2
zKj<6#qF?ld9<wX`+>g*3`bwYak^SNq?IZsAD}0NW@_YN|{NVn^{KiYvuS37Lhn_Qa
zXgwYJy?mnk_zJUqMU#ts2GU>tmOjx_e#?Hjr-k&>`x4&Y@PF;;qjBi3{@xF9Ut{0o
z9nd{VqWLStcZcS!kUTnMhmhTOxT4Sff1K}*k$`+OQG64p#B+X>-$ILP;;QqE_{GmV
ze~72r@waGsy?#(W&ku_W@@4t6^EMRM-7|_4@*eZbr{t^d5sb@!vtRkIcr31v15|E&
z@q--jBOXC<0RP&FYx?6~9xJbrx0xTxkL)Wx-7Cq5)En1$@(|}Ta<RYKLw*z5Psra}
zxBLZ)3y{1!<mdH+{C$Pi2farK)9+gs%zAN5KJWKOzHdmbkX~N+vwK|ro^RedmN)y|
zvU2gu{-vJC<NUr?T$Jy54uNma8_c&*_=wLw=%4=iO-pAX`>|gBgWjpH;*|W3Tp>Bo
zBYMdXp>tjKOF6wX4n36bp&{=rU*%7=mq*(V$nPb}gZZn(Y`>yudRC$P^bXw@RLGw!
z7Roz^;uc?g-uHI9aGkmP=Knr$>HR+X<h#tsr~O|~=>F09V24w;f8qx|vF6<SM-RE@
zZr@z*mNjO)2Zj&NZQO6l>%Dj5Js$64c@NEdKkmuhyF>3$dEdo*T<Y;>T<=+VFGjm}
zt+&PYFF9!4x%b?(%Xz&I<b5Ld^)LR|$v5nD!<uvd2lWY8{N(R`^6WKdysu;3?(@9|
zmWUtypmllA3tETw)0{&*zk}Xm^&T(zyWe+@Oiu2T-7~w_bFVCpd){OJ>=%@GI7fP~
zOdf(C_uTI5yf-3_dQK|-x^I&wK=a6x#FL@ud?HRlaaEkFP##htKU^VyooGFY?ia;_
z3bm7m4n=-@DAK<~dQzddBJRLk?@&6QuMDM~bM1GYN9m*cQF)8}BjddO@;dp5JVpMZ
z-uD%JH`@CZ54-JtO1>-a6OX)yuUy=Qp8t}A{6CR>$sgox_#+qbKJOuY$HO@CZ~41(
z7Gwu#Xr1y)>+oKe=QC*jNM4cWIQaxUa~{N}JWJjtKI7l>GUK?fb$*feu`~DEp0_!7
zdd}!RTYkvCw1fPz_b(Fh$<NBu_)~G*^Duhi{W<)4K50Jh8R)Mc`_s>R32E=f^Im}W
z9>fvv$$Bpvf5tc8P!wOG^;IZNIB$Ax%YNN2RY+ghf&WL!KG-39*UtFr<<a<X4yIS)
z4}0Wi%;R1U?cCsg+`P)so-^UcJkAI78ETjP@QH8FXSDNtO}u8;>_?nr_u>~j_58zl
z_;8NLxBbH}f8%)xJJS!J_7VT?WAG!da4$n|_)U4fyoi6*jy|(XcH+K=zhY;Z4=HDt
z-k0~j2mSWD0JQhqp?mAO_wx5Ye8;D5^8fv=L*8dxe#Ut6IerB{?A$uppL*+cZjdL+
zJN-^bd*gW@#eCu%zKkoMus_Zl{1V#y)`R9>j6?qFwXdE_IM?9Ia|CiUF8;jN|Kne+
zy}`40U46#5#;K58DkN9u3+EQU2O@8Ie<Ho|yM;vi3h9IA?)05JjW3RnM~C`*?(aPu
z`^oMS*}Z<^6MrJUDi<%^pYRXjz5JMc%P-^y^wPf4e@O4FpI?B+A-C#Y2;c92)NL8@
z`}Tj|OZI-?c3W)s(Ct=w`}>~rzU@~2-+%YM>)v<Yd;RT}{KNb9-td3Z58iI2?N-|1
zrkB6%eS07NjwhWtTDs)LTV1{56IZ!#{N6);@a#3d_s!ekhjuyioPD>rV4Qg4Ub`K;
z!EQerclaNh9&+~UU-EBbD1m?8&EGj}pZ(4s!MXbN_Fw%|7o9L(QsBo<`qEQ=d*-jl
zD+<Xa=X?0k=RI=mi_RWZc*-F+{_eLgzGYOQ^;Wp+ORrny(tBMts&MVi|LqeOfBmbY
z3M;;pKYop`Jm7`<Ts^MvecN96itV;pWp-M8=$424@)eg|F@_bN&s+Jj7d?0LOU4zZ
z-YphQFB9o!VzqC|lh0}&?MH>rU1PT!zW?EsXDU2xm$UwU=XYK;s_?6C`}C7vewQmo
zi8t=D=|i7<+4o0@xz41IJNbgmcX{a-#*<>sQ_(qJMW?({-t4x*lh-@yvpekh<1wuC
zFUza(D*c|cyc)kN&-Evbk6cgE-8@}=&YN`3m-I8vz38PoUUJR2!fc<k{TzzPU+UFR
zdQP7Q#>I@A{6>Xy;}>>2l+UVNHLvn+-{}{eG+vsg!paXQ@A&KL@!esy4_&`X?pa>x
z?LzUN^Hp@}XT{fwDyJ72pOcnvB=Qs8K6UloK4g84M_*wlzbbEd&@mf)YG&{2$BET`
ze(IsCpR(Or*Nj#ae(dWXe(S26TsW$*<Kx@wy!vtXJpbBJB6(Ef+V}M9NvD2SG`*|P
zK2})qsl4i!<?ng%d&gHTTXm+wXFq+@U7zrR)n+QJ&LPTEZ<DU_OXVrwiq3HtiZ1sj
z&KH$^bmeKsNq6h$#;^ESUg>97o_drtzRZ(Y#k(vw&ZL<04W(1w_<6(g-f-$Me|5na
zR(d$JJm;B|X6F^A9(DAj@?GfmPh9Jb|L}uLemsU<{CeB>cYo+k&;H&xai|<S{hqcS
z-{Q%}@BaE%J~X~~L@#n5$Wb{uao6R~JAIW8FCQlkU2o>i=*)A_-MB;TDcj|^yFd7z
zvkzPM^ttmg`9X3^#D9g^zM`xC$>+4R{IEiKMTO)r6svKRPs@k?iK(Ya^PfYJoO55Q
zbNfSHdi1VKzVW;9P&9sr>;P7Bn{Rn_PrCWfo;-T(Lzd1y_Hk!_^r|P^I^O6@fBMy1
z9&r8Kz3Q&J9r@cWF8$@4zg+p?_n+{Z<%f^Sd8=hNo_yh@zZiAM?<7|1A`e)#8(N<G
zJe2P2p|badF8{*wpL^Qpj`;bg!YV#j<@T?`w6n#cGp|cpoJ-7h9i94z?&P3dg?Y}a
z=-iLI*G_t<yi+cd(#h9C(cONiujHz{^P}qRBRVnjgh}ZxU&C*Q=^r{jLrxtI)t7~C
zH&l*$9e(PW&);RG8TTjjl%LQ2nlC+co!!3T<ACEY+j#$@&Kiq@iB+6ktn#7pZ>Suy
zT_=~@)fe~s>@vqsJ3!~XVA7p^Wxf4_?tK;t@jWdLwI6v)t}p4KdQs)OIbPCTd|=lV
zy0@DaJAaw=sdq`!PuQ(*(*3e;F!@89ze9T5p>Zqh`cGOvRPXRJslBAXG0tLPxBhP4
zZhiCZ-;H0@SNfH9k#v`TOj@6Q1)cLwOLuuiCx7*o9F=P)ze!YIy%(mu+86m!hn2sr
z^q3!We@OlnR^uz5bpNw`o`2BUZd%%S6_QgW=PoYfz9&szD(w7cH^2R?kRDc;{x#`J
zFO+xulW&E2@0aw@JbGIF$bS4sh1EK$a&mw91*cwc-qQabCz2!7K2d#z)&485_*1T3
zg`FN&d@0X-KIzInD!a_`%D$D$!z%3bxboA=Q;(7+=Q|swy;XFj7pbQmo#ULd=kFeJ
z%6-?Ih1L8+%jYZKj{hz%$@L}89xCkYjJ{Qv^G!-0b>5E8_~GhHXD98aeJB5nFG*K^
zKzSuM<=y=Ftk66acKT8EQ(n!xqRKlz*R8Ls@5aylPMVyi#i4RM?Ij<7_7Ok--5BO`
zFnPzfA9>(eYu@F`aXuGgf8HxlZ$09Sa_#Wv9#OgZr^U`Mb@lkIaN7OQk6bFO=2gzV
zr^QY$7ptB>cV4RSnLpq1gKyt@?U_73RJ3u2B7L0}<xv&V>k2a;O1it3?bbQ1UJUJ5
zt!rpGJE*XWTU~wUm#cY|SNh-0tG?<#w7lB~d{)R$RM^R<+ehoHu<~olyYV~zCatgR
zrSd<@kN?scU%mO~H<iE7s>W5GcAa!3zfL}#9#!Ki@5b-yjZ<OQzw^7*xC>R@@z>Qa
z6d%<%oqkvSlz0B4TVGXgeD^_>|LMxR_3;xGcKy5gt8tZg<L7-%MW?+_O7kO=qWt#i
zXP&nEeXl*WeDBjbJAJY56&fdT=siXDo0Ptz{Yf`}RbT08+FvF2t~{S_e`)8d*1K_?
zPmf@=@AL<DdRvXt_3P^At3SE^q&xXl`=flQo@!rVr5{zfeSxKaXuh{FRKDg}Xzcc1
zeU-;%xqAwjc7!I^>_1c<+KCs5+yC?5Y_!hk#&L(0{un>8(w{8v^t!9>-t(^PU3sS`
z+D(g<+?3nr3bm`SlTTM)&70-b`srt4Cs*~(FEGccX!f>PSo*2kes4PMyw*Jwt9h#Z
z$nt9cE4eDqagx@q!fc;(xBm1~NvD2hyi7X%Cp!I6MR$HC{a?4f<Ui?Z-?BXA($UUC
z-sijDU%qC?_Z+;|%ol%g%`Yy0^aeAZ`{8bHc+Nh5ay!(o!|L9%Du3I`k6HHThg>(R
zu(QKX&fh(5;}bW!@5!SI`JD<od3N=eZ#BC9fqO0=Re0z*U%TmzPdjr|;dw7x_sfsI
z;L1^j`8>U%(;m>NcNKl)8+ZGMdtbHs>=E}_|L7;4y4oDR=ZeGL{mOf<Jp1A8-*w7e
z_giar+ckIp`Wrv6!NmCV=dHKJ*REZEt{?iDAOF@zZ{2DAS+w!dFx#P_enU~eXFYfJ
zmYrX6_t_)X{i_rAK4+a-a`3!W-o&4<PkxdAuJl-awpR|j{@TH`iyS{`{h;~KuqszS
zEgE+ycJoyB{PFkQ{l%vpcIv3Z5B=q`&(B=(?Qw@koxahfCqD9oQHAGkwDU=8o^#Q-
zLce$IFym*^mE5j=?TgmB<njZ?3ypXG^fu=`cC*XJ6;}RM`T6TVVafgO_lptC{jBm`
z<@9%2JmobvjUIUJ`QsIZ;^w5^x!ZAk^bTA8*@m-;&%f+%)_(Gr*PBh;@wG>O{@q)z
zH(O#cp7C?MispBZ+4-(F{q5!J&mO$V4M(1G<$ANn-euW~&bs05vyXYxw;u3-f4|%8
zz3>0R8(w?DI<u#L_F>yCJ#p>XyB_@d(d^%>KKqKl`01)Yc+4uZ_c`;i_qg@fH;=bE
z;<VjA^%pnH{hoC$uHN=L70)l9xBTE^H<-C$^xyyT;ER6s?j<vD<G)_#KQB1&mXUgR
z@*|%1?+^R<wWBYr_3)jx+~JzJ@6#P~@b~s#a?Yi5wDIA6p7ERQcl_*CBl`y5xz6&}
zjz9bPapK!w`28op=$*eAz40fjfBsDyTrv9b*S>x8Gp@gGbm<14x#D-9x_0!z`)_;N
z`d_|k4z-7${hQ6c_N|v)IfDAZXPmUdT{qkA;?cDD?bG)8ua_)aK1#$FKG698kA45~
zm!E(BH6w^0_{DYKz1hi^T{?pJBL`?*i-qPNiq<(4#S3}>#rqD2`kA|bV)OevbKi?b
ziCui)M=KoaCr`R?-SJkJESZ5r<x{nL+AaHz&UnWyqYBfGlHT=2TkrnOFTZb8VW&^m
z9kk9pc6io*jfNsVU$^DMkN??5OJ)+QJif{|vOM#$q`Q5p=2uQHC&fJXcKW8i)BB<I
z6~CG1BrTul@YCmi^nkVYxM7s&x%p7cb1J?oyHmbU`Iq@%w{H1Tg&kkj_h5&X=X>)>
zcluGCN0rmVMDuihSGx+$(_tmYYTPV$eoV}DO-iS|OiFkA)9GL5&$@c{GZgb44qcs7
zs`Bc*miv?R?w8!>Lz``V&A7s$>&SNOeJ(GK@4u%1uIM~xCB4NC@A&wmUiIq{%=)CY
zgXv!@`Z=FpZ<8xuy=11sjJsRk^E0D;KDW}$o{u?my(6Ezti=7lGqZfNRaTpUuiEq9
zfAocKt}>Ij^IFSxdjCml%p^YNh<&%b*Fme#Bxbvc_C9EZc@LWFOgi(x)Qg3pSNq6T
zKfdSNR+&kZ_d5^#^ze7geDcthXW)a5dFZB(xa6i0%sfy&^~$GxdWWmeT{;7IxMA-{
z?fk@*W^VfR>ATE4bk&)^`|jw8Z(aSaC0@P8I`_EXLpP2RjR$j_Yu>)YyZ7IB%^7&<
z+2^i**oJG(Jm8`?UH-)fE}Ow0Tw|}_togxhZWux3_&np#kKVN84Xe*QY5N<#a^SC4
zn#uDNIXNGZUxoR8TotF4S9U#VIltcdU-e!5GhT=D9cRA%@mKlLq2-mlhxX5Q_=5cX
zq*(E#Jm*ci%Ma*7i4m7_{CU&OcS}0`=+-y1zFdF9pY^k6s60`9h4dvc{Yph=zoa{P
z4z15|htkGRxp;2|(r=h~*&R&}U2oSe?+cR7{4D9}If3#VFKO*6%=wbe`;ero@3ATG
z&gnzXrB%D+C+S>I(w+ZEc_v-SRe8?0(Dbx^4S$vXvf~P?=hj)??LR%KFn_O<^6S=L
z@vl6`OS<A~((>wD)$yPFB%S&=DNPS5?Chhf&wMEDGilFr62*xM?dP;u#p^8Z_C5K|
z^(CGAnRF-LtWSH+I9$=4+^T*#Uqx5_l}|gr{war~E55Sbv~=!!($)ORyLqj%!%mLV
z)_43_$F$h3YiNBnf5(3}el_mUa`&1QR^xPhtMB+%U&V*?b7=i0#q|5=YW_RA9Dk5p
z6Z!8dZi`R;&qRftJiGHxrANxU{$0IwRM_!Pe<nrpUuaA|Qs1Wai~rO5rQB0a9j$+L
ze^J@Zq~(>r?8+;BR^IhbJ+fZ;b0wcnPmBY*_Ub!+r>!TC3adOg{pF-IIab(>*XdE$
zUwyUiEYEy>T6t%`JVz&;dN?WF`EUBQSeWan=#&e(<DVTR>Yqq&vi-C)J(?C%&yseJ
zTA|+y48{581IZsc;{bii`LbQNzg>ML7v=J#p*X31Rqg%$phCaTsL=nDsjwP1%PYN3
zf5t8<%=SrF`d^K=(B;+p3Pa2HIQP)4PQJ&gGZj{TYoW^75v=0wwB=R4qP)7dRK8H_
zT&!`r`{%r0PMW`(7M&Lp$+yC^qoi}Zq&t2q|D?RKTjlc|zw_TYe#Q&*P=8U$Im=Vd
zNoV_trq8*Kq=)j~waa{FQW{_Ow?lChR^O@V$~!$!|HYmD<6ED<@Ujsk_o3^|{GwYQ
zJAj@1=UYE?9rJA`j=(ApsOGUB6_VF{v6^2weXUTtg~o1Om0XmQ`=of|mD|pK;)Lbn
z3cEa(omMzhPMLS6T_xSEGuwA`jwcROSdB;T^EpjL^EVZCdQr(;IsNF+da84sb#++T
zjq=JrR_o95z3#iy6SsWl2giw>oN|9Vx*M;m@8tYHt-g}OwB?n2y78)hi(TH?Lp6V8
z2U+g_*G-C*zfqp~=cF`yt}xdzDLpi9RP8Fhly~}6jjO!t->nxP6;^tq93K^G*J0{!
zbw8fx*NU#<PBpIbPF|gU8VBaO?r6H(hpxWUH}%yxd5%xI>NmK&c;C6w?<}XsD{7y*
z_}|r&_oVz*c4NH7Lh`N9yxo3lmzd}Aych21!;X9NA>Vxas&oI>-2Yv8;IVJH>W1S^
z8dv&~>)}V_jr;+9cW;+*5PiaZzIx!X*IqM%`d4V43e5|<^L(Wr^DTF8)M3Y8>U$*@
z<r}U0cN<>*=#^)DXDm^>3hhgWRsYnRr0rXlr(U5`?~~SUzL@+Z-Px)7`SLMse|A_Q
zxhLX#p)u!4I`?r>y5qNtC-!%>O}_t=(|-D+abjne)jfssZvM_cclGK2(Our=-l4)y
zuJX<KA~{Wp72nD`zUZ0%Ckk7+-Ta-y_Sx_JQPr-~mj$n%^7nq$Um-cds-5yqU()WA
z?&MPWvy4wwyza)&`f7itEhncA`J<sYX`JZxX=r`t*A}b3yBDwYWZH7_9l9S~JN8sz
zryuS96z9Q7^*ikoO&*Eey46q1e`k02OU!j2u<XO%{M9FZGODo3XO(9j{)bB2|0-^`
z``pB>&JL>jd=Dt~p`t5(>1~IVe<7z1yYUyQzT+ePQs-}n)=z6Ex$a8N%C(<w{4rjI
zmEJ3#cK%9EmA)%K@@=c^d)2v1W)sN+raw&Dc!`yLD9`rQI;!$Y&dQU|q&vM%{wAfX
z^Mw8Guv&K&*Q~q3YP_y|+V=|WbA{q#h3?@htk$VK_dDr42i)1{D()!n_^tFrdDp+Y
zf6RJ%RQa*W9+dmevj5Y&q7fezcJ@lH6?XontMB|zCHJA_)%fB-hn1Zx&paXNp?r4j
z*neWzudA=*qdfh~#`n3`fnR+5Rbza}d*~6m!cN|m-&Ef9$5)5sVjn9chYCCWCYK7U
zag}%Lt@s{VZk`Ie`Mdf`J`2TPCD*Pz?KJ5wj?l-%2OYZVvS&W=mQkL+lCHk{)0O*f
zV|Blm<&~VP{Z?M}tIF9^V#=qYJA3WwJ3W+NOpBEsFI0Kz-<c0T^?@fIa^dZsJ6G$>
z{&yCdK6U%2KINJJ?<(U_(%t@6@>7oAp*XFa(@xO%ug(XHRi1Xa{DpfyZ0kL~HG)}R
z<%7yIk4`%MRMMHR(vJ=+zmKm9$-P4RQ(@&7mD8^dtNE2z`(LfADtE4#6s;%akajzi
zPWwe$FFTyp&Z>Tsu3P(j4()xO3WvsXaXqoxKjqcDL(6l%6-`%qlKO?FA1fNkyF$Mk
zh3=Qq@1eW=a=!InIO@G;yy=*;$BFE>!pdHhSN>IbwVw-BZk}n89^Fyc$zfW)s<@!M
zo41Qg{?A|{zX&_K%zf+Dn|fNU<LYOgw)=grJ#}1R)lWJ86Fa$d`N(|h@z-G|ck#5s
z`Sv$%Col25!fKsWd*$<;*ErTuVJD}mzT<DP>r<|&=abS$|NN?bKYQvm;||4*3MaMy
z^au8Bp)k)K9e>@tL-#f3MW?^)X#HTvU!||gyZ)(HNq6h-#;@`%<@l_yo4>Mi<<+>#
z^W2IScNYrfjgY;5`M9-rJ$U!CM-X3$*3)6`AA043-+#humLEQ@Fz*MFUMM+CI}Ur8
z6g{s?zk?>9X)*UH=}v#U`p&*OzAL|-<>WCfc6_p{M0SvJ7@BuwzoB$De(odM?~I0`
z_t+CVJ-%`I^3#_-_xe%Z13crrfBvh_y!D0=JbR57Y`W{F*Nr~%sQa9~#<!M_GTsf9
z$F%ME@e0Xpv9Qzs#jfw<*3CcP{yBd2+;-aX9RJQrS8=hshsijXbl1Pr>&h;acjL<!
z?yT79?|kdKxYCW^?N3!d?LDgdLzw!MwEwqT;n4rfm;VQ}qdPm!{9sbL^EcK1%T(U=
zS1*4^94fbLSM5WVA9~K$ZhGU>&Ky<P>E(RuQ$H*Jr@YgLF7NC9U#xki#V#M~_|Lp2
z>G{gL>)-Lu-&fd;U)867WEb$&U*F-Zjc2bLcbLz`Iy&vGqtnikuKw?FSDx!nI{)t}
zy4r`XJmYm`|H`X<Q{LIteD%tDD!asIh28vFpX;gOYv*r<*5`al&-ed!cX2Y;U(s(m
z{xSdgqNUf5A9C2skG%CU=ghqqRPBrW;oJ{A;$dfe>E@B|9-yIi;#1~{^ws}W&ij}1
zUbgO+AAP}<qdb@MAFx_ir8mkm->K;Qf5l0szeu|C6Iq}Bs=Jr%>bv|X^>U%<YJZgH
z`jgf!(f%(srk&;fpqIY-SDWvB-(QZ&yUPnIIp=&8o$(~;s-N<dXVRsAyzTd|**@vY
ze<;uPNv9qpZJav_JNa~cb^W_IBTiRX>6h}VpYkQIJ9F#*wa(8*iQ+cgYT1n^UwG*+
zMv3XSlTJUGw0x+-)R#LeE#6Fu-FiCx@t%-4TcLL1ZH3yUpP82K{9{*7{}v1LUi`bq
zZG7TJ_dR(uDW=~_I`uy3Y2~Y5r$=d*X!rY>7o{GHCl%tO!jwnSmELB#JTfucC2hSG
z=6aLv<lEV4S6}g^yz8&N(p&Wv=KfDg=Y0pdns3tbJU1*fUCB*(>O<10&q;Uk%<(3r
zD|sXz6)hj=P~MxE`zc@U@|&){@;8%~SAI!(weBh}QQqxS7vIuPC4KR->sPt{zrQ|$
zogS&rcqy(|nDMBi`K3hjL-*;aFY;!5vNM?ez2nC|PK&9(Nmu%zT)v&y>9>0NG%a@V
zX}<ONtdJZktmaQSCoPV^&YrsZ^EcZ0q&3gEXk1~}zw<kt{#E{WX!(5oO~+rietx~e
z)c>UUg9?Yrp=#HyFZoHj>rW38{oeAplP}nOmzREFT#eJ|Pqi-Pm3=9ncHZRc#x2j?
z>cFR8IEK}J4K2rCg_VC$UfF9`p8MO;=F9K@pnZ=6rXDO7-QEAH@9cNl`XwiA_>?Q|
zebyLu>+I&~_N|IDclPqkbI_f<rmd&X@~sNxTNP&9os?!b9sbGA`){`C%WfW5nEosE
zsPZ?{%CGCkuBv#K<(*znTA%Ar8h;h0Ur##o>J2yhi(}sP-Ydt6Id9V4z7MUJ|5w=M
zKb8JFXH@9<Ql$^6mr18QlCJhgd8ZfrNQGVhYF)~^{_3B4_5b;QYbzQ%zn}Wr(Ulyl
za{Hb3zS#5$Px|&IN8Rs=iS4m_zdQXv((=s;D>*3_mnx)p6_WQtW5tj1ZoZUL(pCTb
zeh)g|mslve@{@VLl=QQozUi({c)@BjFz<`1ai=Z!9!cIKCG9&RFyqO~AN9IJpR?+Z
zMzGS`q2)P#(nH6o+EwQu<yF5GRnCqRtGv+foaC2PyM-#R;{4EZd{)RGRA`^3#cEvd
zQw_x`pBY+SJvUH3bezR%mv&dt=`YaMQDOSONomhd7Yftf<fRq1_SAeIyo$fdyS%H4
zcgj2a$n{Q2<8vqu^$%UUh1zf9(APT(yY+VUliE$T@A7Epq=mvxuNS+1sJ;$u*ZP5h
z|IeZur}CqVRo?NlSoN9r4y9AC7mFU6A9U@8?pyk$&0loEyN=)Isu9d}erWe!T(SA?
zmyHrn_~NT}dB-O%9361(qwf0HZ7!Jm?tt<{{quLmNguGyTOPf~hp(RF$M2D~Oa9Q6
z->LG#J6n0azt+*!?`o?2VrcneA9waguX@6*<4Li)2kz$W@``EwLX}@;d7k^x-8r~>
zzF*yYDj$0O%X{6VtGFU>t8i#stoZEqjU4^oRJ4BR4m-am@2W8K9p@Y6{3gH1pP)N5
z-XA(vdiclkL(kcFiwnk+%3-K|R_!{w>gua|Xa2a;ckQOdE<Vrq{J_pTtmLb_+t*5t
z%DecezFP03=j@^FtNE07`m@mW;@D9B7rUJ}HZ68>rPE7st3vS&%8x61%>VzF_NTt;
zuRPZ^^t`2A_CII8r(bf{CtNw|<dOB{V*SwgnOAh*XU=-#L*v7gU(#tG^eD@xrRBNi
zbxwmRFMLmm)}QeJ-{zkbH-7imzw)8+#iNR!?)e(Npna#OklfJ_|LPNyFY|PW5B$^n
zMD;NDtD8qXIp+SL?F%}Q9+5|)`oxTjH?O<JN8Yu^#bfa!>yz&Mw{c*$&+(GBpUN|y
zi#N*A{OF`;o`pvIEfy{`zs6$+zi*gvW4?6S+nt3@JMU=wdS^rO^xX@%$GL}Yb@Dw{
zovBc}3cGzzf1dPnj@Wn0dmXg;Od`LX@g?c%zC(G%ca|q#D~dkg{Ac~(ek)%#-gUPl
zf4jw{zZ_q^%K!O&l;5Sh_Z9!p?t!6l=pLJ2$^4NV>4kVNem(XD-#y^Q&t5e$ukT|f
z{^NDG{`e)2ynb}TeZG3&vDaQRA~*faFK(j~^;;~&&wR17|HZ1eKZ}K@z2xK1KH}%U
z8@o^c_9G8GYt6e{ITlZsthfK_pStLTvH9go^e9n1{gQ9%PtTw{1S<C(XXC$lpS#@i
z-_E*pE*~~e^&XXR67iM%7*BmgE5|Ri|Hjo0|Hg&Mm;4GE*Ld<g^FZx7G;WtC+IRW@
zvmaV}XnocXtyjBA(R|%=A@%N;pnD|ah|_5NkSkg{@>7q7+Er+t3afs`?~r~ZdXBZr
zJ72civ)+2+m^`eHetBNW?!+VYzIVQ4&-)#7k4sJ&!z@oa+vT|CV<-6Vyc8;j;v_Uq
zqWTWaBVLK4Xs8_OhcEJi)=MssUtMjJ@BieqpZ;jh-^vHzR%>tgw(lQx@fe<P{2}*x
z?QXvr!>=Fwilct|&R@*^zD+qlj!vB0|K`0a?JE7V&I*UxJ^g1N=+3XH$LCO-^!(bj
zf6o<%z5A8-UU{}ceyPIz|Bgws+Yb5P>pu7Vr=4-}g(LET_K9Ccm)PiseX6h=w|S3;
zy~%Usncj~$<Csl0d)5w1X3l@WFZX-uYAemW;f??Lnd@G8>)iL;wfBE$<*&wt@@C_j
zPq}f_4@L5x7QI)0XTf_u_XSTo>&UN+GQQA9cFP~7zo_g-zZHe^JqJ|%7wh~|jVqs-
zFN(V>3i+Ld!fE}>wEf&q3`O~3V%m{?6!+x`na{{4pmk(^BOj5U$XEEaME;N-<h*F}
zLGwWK84oJ2kUX>AxQXnrx`#4eB7Ze4%4ZVEV_GEl3a1@+MfDph&$NS-W4_;ome=I{
z(PGiwQ>##(nctg4SN)V*SB3Pu!fszXx#aJ9lg@og+Bz!C_DTEBa)sS@*0X-vJS+7t
z=}yksKIx2e^aiF}DmuqOCx3pwh7bNE{otgu_%sx&btoS?&a~|oDxaiR)%j+j%I#-`
z<W}K!PyO%BmM*zs40C_0akHGhROtCiV%2_VxjbQ7tn%c}530}gB&}VADaWM6uL`R?
zW@vfE*U)l$Ug6Mjrfpa4dlmn(JoP>4O76<5`LjI7MR$5$$u-OIlbHH6Dc$W$%JZ2&
z-|~ZR-+JvCnD&q~Ijks5d3S!e<16E6w$J*cb03qQ?>){?KcHRSza(A75#`xF=`Jom
z<J^m0y5l9+j4PC%OpD!q%daZz?58`=bnBlle`)_or~W0K?UVLgufn{aNV;49eCu<*
zq$|A|THf*3?O&%KUH+NtNjm4R?oXAcKBqpUKT7|Zbhb}A$Itbq-Q{|!`Yf;L&Tk}t
zNq7CNmwaK@uZy!)o~k_gOgi->>Gab{SMpSzdXRLEmvm<ro&0jVq{XKSt)s#mFX^hE
za(Yo=#h3DwXVThLnEEg&o%)h=+Hum=x|LUaEARYnSD*Z#Qx5r_U*=z#zp_7=`Cr<#
zb}-xLzGnPQI@>3m>mzSC-+KE~Avsr={ypimv-A_$KIvRvj@Qx2cj{T%LDJ+5b3Xn7
zCjaU8w41}`dANCDJARCp^3L@qo$JqYaO!8~AF1ET@9pFl{N7GpQJ?EcdQPsly}yv_
z%kgu)sV_Nx@}G3tQO?)tbIzCY>(-Ow=l*0p{>UxgQ%QLwo%|(jT-eDc`=>mdb1USR
zD@=Vhj{gUYhUx!Pzmrb>E4n(bRD3H>e&$Q3{wB@uRG9XYw0*BI^&#nMUgg<-Qo7oo
zq2(#BN$E;1%DeS+^L72ZdUCF?n}65WUVG1N&ssW@d0E;`($=4t{s;Yy$GrbH-#YDz
zxqGMNEBVj*q|;9(&2LtiemCh(?wwup2Nh=joIlr_^+WYvyUYtZI{k3c$xlb;e4Tt#
zACk_z9_{}qK>Amq_pmF>_%><$&3vGvQ!gsIlb`yOdqr3BQC`_m?n~14rNWAzEKmPa
z(K&z88P_u&CY}2~EuDIkbf*{Hyj}lXSJL)5G38g${LD>TZg#=mpIm;s_nWg{`s<|W
zZ(@#9(Yd~)t<&%QD$IUKCx4UD`Ft<wYJEe?Q?5yO{H7fy-T8fadWDsLQ=aoBUHJj!
zot(P$r@WHR{YttUzw)cfGY^=QKJERFUApz=Ys|p(!&QEwJmsDDbL0$DKB*sBpLEWb
z`)3~5<s~^@*FN*$lvj?I_4tChKIJg=CEMEvnDR_I_b2IGU(T27NqOb^lm8q)Y3qR5
zpWVZbpX{Ifb^K?0`vTb^On$V3xgROtq{#tVH%$Jsf7a_)VJEMYN76Z8=Z{hz=?}Ai
z)~Eg7?!0;PcMjWUzw<}rpX*CH?IG7w#VOAjR}^NRnza07DAF5w+OEr=cls(HUOuic
z^PHq}pB9U*^f>h(>9o_NyL_3wPm8J7<WEn3^YP=>{HIr)GpD!8)7~mt`#TCNzJ``}
ze0Opu&kDQ#U480H(jEVuJ)~bux*NZ%Pkxh5znU~ZQDOSk>))~K@2)uV#xZoSxbs@e
zc6$FwYs~Eax{tknot4*`+4C`Hu6N{<m(8Ji?=x!W`#1bG{_S`2iyo?fRlCK~i*8)?
z6<^8^IR3JY_dn{aF-*T)?U!<LO6>fhdizje>T%NPr;|><nKVA8MRKSx`AwSsRhV|4
z{c}A@+vf^%{G_XWQJ(vrblP>&nJ*=s{3K1!D@=V(+Bzz%##LVJoATs$S~}lrL8m;?
z{{LZzxnD`=@5PeNd&Cu$&;D_qKT6<_5?FW%RR8Dg_qBZCCG^J;|0sb!O5hJo0-1ML
zd7Qj3v1-3i<;mZqbn+{I&3q8;e!0TxyfL)Ax~E&H@+#j~KJC7cPs)F&z2rV5ZT$J7
zdyvFw*PZ?7&q8CKZ<EgRV@2n4By^`=IbYJ<zT|$Qo!ihb?=u#fc8;vDyBAcS_i#h$
zw3DRCA+eJ)zB=s2sr=i}^2*=wGaXj_v%DIIURBu1C--AodZ@gzUG@A-dCDQ_ZhzBn
zCEfL}>Qi5nuKG<{p8hB4&i^J~=%IXc<IlH#T74TDZ@YHnk(he2Sai1!-Mn4@YJaM@
zsl4i^yxX^wXVT(a$A5M2uDt8t)u%m8OXH`)?zwx_uPZ0V3OoPb$!EUx6<^A`^>uPH
zPK6!+oqcusP|f@O|GwLgK6>#fqY7_X^Vs*k?3yo+h9bLzCqC<?uYBUAKN@%IOS?<D
zv*VP{+rGd1Lw9=i_r?`=^2+g&?)d5IGe1bWvj=%pC)cXJ)8~9X)zO_i)u$Xg`uMYd
zy6bP(ymn51=#}qt=Jz|r1*jhFyQIFW?7O7Ojho*$MXQJSsjypLSKs-Stk3;c-|1sj
z&kpl<B*yKqTi<-^^Zl`fraS(~XIiYzOUkQ#QO>R^Onq2ry7ME`$|d_%@oPnwS8`Tv
zUplPh*6Ex2%HL#pcaNmL;wSeZ>CXP=Ti@wxt~cqfzkQn)?Z>n@?Yz}~O<LZ~Kkd5o
z>*U|n+fVB<ZiSscFiwS?Jk(e7r+p+{^;14y{wsNkCtY5tp4=zJO0LRN9!Yn8b0|O7
zkvOzJc_q5<tB^mfF!TILX?bjgx!$BZ`Kr%;=xSYAUg_ab`Ll-#^S-Q$GuD-u<0M_p
zKWTX-59O1NH)(tJS0O$UJAb4;?KtU_JDNYOuv=%f9_6{7icWr$#&3lc|H`{{RQ1(-
z^wc^NJA26XNl)sRvVAvySD)i0-S$7}OHcXjnZKU<|JUv>x_(*j|E5)#dXjXdZ_3$E
zh4iPwq5gEC+Tk0f-prT2=T*mk@N1tuVpQq9@|5$WH2Z<`<-6*Suc}^owZB8l%~PRy
zCdHxi4{cY?&ps=x{G;-!pYpV8`dguKE6nkduErf&K3{tsI*)etrQ63!9@Ra&a`SZP
z_di21<&LiSoV2{+x04e-77M#|RDNLEa(;IxcJs_vzvtVZ--ipePwc3|e6OUVox3U=
zDu+&AI(_c?S9a3%Q(yHb$0dLD)dL^*+*3y|e<w%Z$jP{g-8$8q7gqV1@```Y=_;)5
z5tLWweC5@+L(4O-?)qi>PJg@lZa&|~tnfKU?7QW?4qAPt!fyQJFWb|nX)(u7+PW$`
z&ig0&S?L>n?U3Hi7c(AI>&SLb+~I@UY_ikUqvx)%+YR6U@X9kV*K54$eo1|Y=?Bos
z7dqcV^}B7T9nAd@R}wpU<Ez3hPIcq+la;)zr@~xk()?0|=C81tS9zz`^R3VIOiJUk
z!fw2(pYm#b2Q2&WH-GhspNwE7-$~2)8+u1?{2v#;yUln!l-577)8nq5ygD4pXVvb)
zogcFG0bg8y_M(gK{o}Vh<!-Zk>~ZCNU;d;uXaDp)k3Zr&TP~SB^WmpH@Wewd9Q!|s
zOZL3qG55ISlyUw)U2^SEKmV^cY2`yve|XE`&)Mmmvp+FH8wU;X>HqY4KPFMVeLz>^
zsL%108{hjk-tS4|f4tWNwKI-!&4(}RwqNMwN3VP4u`fC9cJfiLf1>fti+{9wxO|Vl
zyYe9$|9qsr;?w^>|MnvfJZsInTshw8OMm*+TOM%z_~@TsweM$7y=MIQ=bd=cO1oV<
zN1t%~A@_RiZoe6Cwf2T@`~FcE-wusOUn{*=E>Ep6<7m?KqQZ>(=*$y3nm$xm^{@C+
zUd8E2%PYS=w48rV?B0j6j}^K<>u@Na*)I7_x|=uKC!O)G8n0Tv@~Yp^a&ff6<h!CX
zP9}}t#4cV=s;|?wPrfFlE4g&#nTID`_2-W(%-=00o%0OUm!WbQy8p)O_O0WulGCK+
z=?9Xo`0)Mn3Oj#S*`M;{V^Z2U6;|VQ@~g&Eo_;RR8CgFqZQM?e=v#%=dMo)TH-CrK
z^V%%Wb#&u)a_joruL{$its7QyQ(ld$yo&>s-6=OuBL0RVIaT*+>N~9ZE3fi8<+=Y!
z=Y3$(^1uo^{)g7*y(_xYNBdr3w_jDig(|m>3Ufd1Z1l&!wCP0$e_`!e=c&Vvd-EaR
zeEX`iu=0P()83Qr_NTMwt{xv1cKxTV|Hys5eeb{f<OSmj@4x)U7r$oHi^diDU0LF`
zqh)XZuT8HVPm4!>_ep>A*EiohuCU^F((-e+zv~Z=IrBT?#NGD!^nJJA@8ofXcmM1E
z+U8rAo;0p-hgA>y!h<gP$!J>q;Ti9E%S8{kcr+>g$BAqI?0`$pFMmhq_l2K2VWo#Z
za*wmd6>jnOH~w(b?_4pSFFx^ukKg1~ulUtyTC^T`+FGx8|G#|ts!@l_KXTL|GjF|d
z)ZuEEzWJ-GUiOReP)t2vZ2DPOZoSue+g~{viobm7pB;S5xyOtvtnOVtdeb+S?eO+%
zNAQ%#A99yRzxkFCeB{2HyzXARpEIg(mm`ip>8v9!9d%gsS6=b0{EfeQ$y;9ekfk%#
zI$r$xC12X$(yK=mc5>|Ws;d9=-@SR~gC20hs6y-O@XUXD&_zFb`Zc2poqwHYr}Z!V
zK!sOdbM~{3y8PXx{~Bs9=~vMwzw;Y!TJo^-MimaVtIB_whny;`@|Z53R(a;6<>mbG
ze^2c@oOGUSU(K84*$+K5U#{BaJd@I0zFmE9$aDVco;=I*J%Oa1UlM&ks>8~UReAWN
z<$2GKCWi_=Ur%(N>u^#&tM)1Hq;tQ}*)D0%FDk6~QC{`Sa`uwgo!ctAQJ(siG(I~t
z9<0te_x#s`e!KpDSC11P@u*Yw+Ii#Oj)!99*Q-1_%lY*R#jQ!P`>savku<;DVK-03
zcNM3VckxPnbv}|`O^f7_akg4dRbJ^wRc_u6t&5(+w3~|V>?rL%=^U5;#CN9`>a8PD
zUYYW*=)6CxXmYBsi`!K{<((hw>MMVx+&mSsi+N*{ukj0&-YeIx!z%Ap-u3V5>D#o}
zolEHJwAk_0%|B^<r62rnqWcOsssEx66?W_AS1PQ=Ro=-*yJ@lFOL@vOY3(LOd0^+y
zht_9aH!0oOq58^yb^cNRMCXp~ezCL9YX37nEi`T4AA0E0E7pAVS4KlI`R&%%)#C$J
zenR;|$)UP8!AFNf?{Boj|Lw4edztqn-Pu`{zf4-*<tb?&=)7m><XY8>%hO_&w=1un
z&vfNqz4|j}?Ya3`;|g!L-+S$L>;}92Y@Fp)d{h2{Z~o_*YoC1mxWb1VxAAsgJN$R!
zX|a1w(aAm6lQcW%u$t%3Z+XMYKi}=w<HT-%)K}xqx4g?Es``}AQ2N8$zw4B{?zh%#
z;+uZ`sEbD{-EDT;HFy8|8$YnY9PT(idWS9lY{S{Ne((DCes|jq=1@7*@9EE5Z;P*8
zyZ&5%w03BypL#URd7t_5Z+-OEoz|a)m;LrV_g!iK^=1z`=!$C|ywjh|9{Am(Uiz0`
zy8G-iA9UE=F5UiabL%{3!{zV1@1HK4J?gw2pYg-hm(HR0_bcq=TKP}s%L=RWo^p9o
zhkh?qVb{OnUwOs9a_uUVk7c~@94XQNl}wb!cUa}k)pPx-ynBDH>R087la`ZL=U0c;
zAGYj0Kf1@0KR2qd^Q-bEe8DQ-R(`9SM}2j_DBgCMdFD{M8&|#cS15m}&~x4n^Ev7c
z_uKQQuixb6QHOb6OZvj~FMRBpzq)QzVfCC<dG(xCdGeifr60<>yuYjO&Q)D~=V!Wl
z_A)KzekI+FEAN{YbG}LGq4To$M0r<*-TFIw@7{;$)=N)e<_TS1Soychf3n)9&)w}e
zqYkV06O?!Uq0|3-e?MvW7ZrB<qu#l2`3v`a*w%Y|YgCP=Ts~1@o)<j_xO&^qeC*Gj
z@PqLazwy|ERyq5uQH52UuKc)iaW(ZIX?at{zws)p_{(!WIhe0P?K-UZv40hI<I;mf
zc2!}<lcdvr{f@7~+^3|g^<{Y_C(lEmd&4t6yX1FI*kIWV+-LWr9=7G7OJ-o^2k9sI
znQGoFmmgPo%VL+OygEAnzi~xZ`FK~J`@Gn6#s8t_eC?(;KJCoWeDRdm+%$UNx#y3k
z#f94ct6%$~wJy2*fbnACcaPin#EtHI@@TQJ<9o5@HEw4Y3$;(bZ{y_eZT?7K_M+!*
ze#v-6$zi_wve^H(y0Y8OuXcILoxPslSSZZ+A#a%!J3n@3t-t%Hw>j^zn_WKsea9-_
zSg7*q9HV@-e|YM}kG$x{afOS$|M@$zie4yr80SxR-hZ=AUv~3&u`uoPj-spQCd#Yx
zs`AQT|Gt#p+2g_=3V+r8#)>K*niu6gP15t_H_!Vkn)Z9&NpWaiIBEWIXKi0SFI`dP
zL-loNyM@X-yKxrler8fXx!86^?%lcN&R(DXZN7AuPx(H_Vqx{YovvKIv)Cx_s&My9
z?(?C|w!UUuVU-^%f6y@-d}?Oz>&JH#R_BO?D(~)#)lZvmFLpfY;z#GV_?<-WB`r1%
z%>x#yT^DcEFZRB@_rG4c&BNEYadiIrPgrul`~70Uu(I=oE}yUc&DWoF{+`_~G<uGg
zxY+j^^Bvzk#k6?-MmwLh<~bLQ|4=dgW76IGgO&Yt`HlKckHzZ>=WB0W{99~)+40}W
zXTJ3->Um2wPbc4L?Yinmzkgru<$T}ye+n+txwnfO-T7p`c~|E*7Q4RVzgvG--|>G(
z%fDNHRnPv1V&+3hSMv-lul&@|{$0DFey-b3^$R7pl;eEq&mQ&gr)>V()n@PPdEADZ
z{lziwdheCvX|c1*N}rTZdOw)$=X+0wj}?WLzB&(A*x40(m=?QpUe?by&%2}hr#z?C
zfBmZSz(SR;DE+MFsqQ_McjvN)y!7Z@mwe;@{T=KgA-l?bot9qc`=Z~s@w)rY#jZc^
zW$S+V(HC4fs&I=P-tqBAz3SJa3f=o9cKQB7#p|3m>C3koUH`y6myf1JzuSOS+!$J3
z-J5jf^R@f#zTwVNKdm2`v|sleVZM3kV)vgm|Cldd-T5)=J3p|Z>&4OeVt2k;DF0?0
zaW}Ek<2$N8^Vp>M(?q{_n-=rja!1kgmHW9Lc*Mib_|naz#4aDm_*c>0I9+}BUD`?O
zyM3z8p9_7ynbbdaeqd<*wEp~#>c<aN@~X}^Szh&DQRUV9JXOBC(B<8EWoTYmwHvB`
zRlBr1&+jVi&Yg>`uc==PO?P>6cTe~GUO(S^md;<xcjk-y6YS#4V$~1LZx`Dxy74DH
zKdtEY-Tn-f^B;P<>izum^OMqkFEuSz@oS;VtN8YZR?Z*J7rXnk>OL{c{XdESX_)K$
zBmIBhGgp2n%kw^IDBYbO7CMh#(c=yE&z1ii%5U|)bT@xD{?PlKydOg6ebS`#-9NGU
zeV)1RMWYJ4`;STeXE$Efcjv;ce!lll-S}<&9sGXnr2C`&LZ26hj`yA`4tw`2@4fQu
zofU_!<M+K?7dICB{N9~&x^q`|&g=5ss(&|cSI@pC#V*b)RDJT*J#Ve*r~TfB_gfOH
zdp70Wb7S>6?qbm?$D}K{DZjJ(k513Kdi!=~LHjeQ9r06l)P7CM?_#&_;zH+lt9tu0
zEq3e9ymL|-U!C6HQT0R5i~m#GvC|5x-{&Z=o{uS?RA0LG)7~Fe{W|^c>T}#X%HQR@
zxu5KC=(#WNS<v0RUUgm`THf6Y;A63HzR#7qdp!R?cCm22ermpQ?d+wiPrY4inq4IF
zUyFs69aZwj-yIbv=R0rb*XLb-(wCm{+cSSXp7fq?MYW%=U#a>}Y9G~o@Bg&&?*1|3
z<(-Yrykep0O7A;6Us2D;<(GFheD&9NIBVnC>&A)Idn=XQE_QkM`}wZkean1N+^R6|
zQ>*($<wN~lj+?acD(v#8953mvf3*(dRQi<r(e2}W>+zFv>eg4)5A|b1&nb6QyJ~%V
z9e(PW&);RGnGQWSJ!ikCUvk$cTsg}8r^*9|mUnW_{hXA(9ltC8|9_{Q9>FU9DDU*P
zlK(=LXPm3zQ1Ur+esbtA{XoS>mg`5JiQRje#>@2$rOgksU-Gk1^rU)RwNHMkc2&7`
zCyIl{Pr0Prwfooiy#L!TSnU^M`lVcZnEWV*`u#uby?4A-)wTYuh++d&Kt&UY1!IdH
zjCE~E6cjNQ1bYLc(L^9dV-y9WqOnHBf+cn_lGq!zg^H1=K~O=OO0S9nN>S0@dp~c!
zC-*+wamQL~?|nE2#y>oaG258W9COUF)?NohPrZuPI)z@_Q`zbt?P_{nS?8(BdF!+C
z!%NHgPu}{{L-!{9J+Hj$5BCZ12U5R<#9P)23%$7S*^50EdiPeWYaUvEx0UyIEQALO
z@he5o9@4FP(R=;3>_$&Jbp`Q9;*y0ruT{4E1%1|ii<M8?%PV7tV%B;ryU=UBDtq(J
zn)g;bQhnBb6wQbKT4>pi9-b`p_)$JprX35txahra$UL#o<JH^edG=Z9R)1A*mG|r>
zo$|y%#jNp+dMwm>G#+WckgfXi>ea7Q_QsKPcvJNJT=QG9?sID7V_Wa|vpFLY+BtmB
zzP(QWb(4~X)_g>7`8)cWcDtYS<c24GoaU8vf2^{`IhA#tS2^$a=grS}Td4MwE938=
z@}aV27kbOT(0jkv^wPceRlV}wcNo0iTPhzatG}tN_iZYB_Ic-QZBJ#dyw`uVS7on$
z&tJXy&iHS9=)Q{%YVuvW*SZ_^8}Y!*!l?G+#~^-dp^m@WPdy;%$kfaHLl2?{(Sztg
z>I2oE&?~-t<*y!?aLuG;;;r`2l676PWa1(AW<~mIq32g#y0s5Q&-!H{>z9SzIS*be
zWc{_!;u}5p$NXJ#WRQMZ$nWK>e&{|+W$S#7-s(4cZ$GW=sqB^4afKHU-V_-RYuwT6
z{#IrBX(8=c*m<p!mW&wCprlCq9y0D~7qarBvd6EE2Qu^7>Nk3gS1RM@7J6~U;|Dt|
zq&#+6cAzI82oIq0TiY^r6h$xI!h^M*czVy@JiZt&3%&Ne^i`VwZq0!+UZ}ZWXWge9
z{2+6GZt?B;udegDkJEjQ?$hwYqKN(UhjaD|PyOcDKRq}-VZ3$zpa*q4h(93vzvI3<
z>aB(YzO3Pw^6-RAytMp*dO>ZUxTyMS%WALsg^m;bw$QUz$Hl4#z1j!QAo(n0K7iUE
zm02$=wA!)sj5qff+JBYR9+e;O+WDW4-up>nq2^Z^9xT-SDiaUcuPEwx(LbI)<wfR~
zg~T^S;)91459oD1ShBXSvhs+m_G$gt0aCB#^JH&cdf&ApK3S;yL1b&(RIh%MSGLXz
z=)L}X?J<vwBJ;*V&ps<2sb1^x`enr})m!6E{bTMu<CKS*EKOLCv!dEhya1JNwab&S
zuP7>?Dy#h}<9|ibvfG>Qq_h80yHwUVj;!+(S$VKz@>}SQpH&V$anwSue(nV<^vc7t
zg<g5fZuIK^DwEGb%YV>Y^BO%oTj=$#XgYjiCrDfXSr<Va$0P3T(dGx2u2}-Hhw>gC
z-oEv5`y9PS341{JU_7v|2Iaj%#^;naPhEG|g>NnM-XwNeNO@56VGpS3d1d+u(yoPG
z{aTLk0%-@-{2E`hJ<bCl^?+KRme+L3ftsFGW`BqM)aRj|qgCb{Y$3c_X!*To7wKOA
z*{53QmG{!wU*PW+GR{_ipIh~i6OQURDy2R}`kNOm-qCygE1GWckX4VLX2o}Q?K5`m
z#`97|crO>T+MP8Y{bGLuTI1{0TXdYPay9G6(tq7?X0ORfR=Hd|iBlHhw-#FE&|7gT
zt6rZkt8DR^Rd2;9^p=01S9>hkvZvhfv&IFz)h>G0O$*uQSm^mF?O5pb--_Sp>1R=7
z94++fC9YV=Jh0I67xbLtEVSxHZ}khk*T1sU<^3-8Gs`c~d+pN?(8^DGT|=E$oWDT)
z+(JFiTlw<pc^=RF%*z|&XQ8z}AfBk7sBG1%dgWJT>wX41EY$O+%2v6&dL3Vtv#z6A
z+qL3YUVUDFv-0bBs%-gDxq8drv6J|m71>XC$a7r_wf`#XJW*Nmt8B&3to^olL$B>+
zl_{@Ce6rB0U*mwv#NVu_aRhsm|GaY6`OSK-Xw3u94$DvJm!ggfvS+{AYspqUmYtQ-
zV-NjNwD?7j-Rg&0k0&eN$h2dj-XHPa2Jdrdy2|W3v&K2{VP95MevoM&eyo04aSOdw
z4!vh5`HCWbWTDr-w|;B<s%_b`55KUG^<{aXH?H3J;GcQX@(0y3A1w6R_4=1L-QvTm
z$0~>3>K}S9u2YYNMg5C7UMVClWkvcATKz*$J`0sc^*8(p`$%Vek+pvEgV+PI@6b4a
zJ>&yv2h{ek6SU$G=O!(Wp7R^;X*_$({N`Ubd8FodY4jV!9@+)bgXr@je1Ph=+7D!S
zLsrCo)l*K>kxAEnkPd3S=t1_=@D7p=di7{J>z(rC$@nSHf8g6fZ~j~5&?|qnEn|m;
zmOaXMQQ4ZG=!p*&GR_`SpQ6=2^!PdN#Ss@k?;ganD{DIQ0QCIWOSjs!{2;3y-irD|
z-g3m%a*=tjNPmi=)~9mOdMSrLSjhTiq2&jj|7g0^K6>m`wB~hQz1maPWUD`(zgzcy
z=!s7jdOR>b7FzY9$9}6k>pA&7q&^F+b~LUdtG`*Y$CG6bdY#u<WpDi;-SVFrz4QG7
z?6lB}?_PYh#sxj&q3H3h`7_J%-ksM!&koWp|5dv@zBK)@?=S2ATA$`63)yebzBP{M
z;XNzTKM!f&L*+wd&n|18qNksVI`5Qcm9@OenqOt?v9R2JYJXH#d-BTV+UNO!m+rOi
zmG}IWbZ`D!>7IS+FDhGpfnM{goOL|495Vg4Q2j||uYG>U#W>ZhN6vd5+~b4ZIa<>x
z2Wq<BCn4)NV>bwI7JBVy{hFV2{0cw7KJ0);WYWpczFp%g`9P~5=zA`HrOlWpW~F0x
zesk+rubQ5&(rxo6HlFp5^nykMhE98MRPx-@>mGOI$>VB}{mAwwOug>;Gv_C)OSiqf
z<Az5)@pba!Ygcddi+#UK-hQIhwug0Er8MWxmxeFC@IT)ZCY|32wYl>1`G4;FUBde%
z<R>5XfavwR2tPe~_yu48eSU&nx9_&uxV46TnRv+k(*+;x*7T<xrq=XxLBquj7k~J2
z^4;A>bQwSH!DQcc&cA%*tPhs?UI}`5V_ZSThjuJvpKPHue^rm)WJUao`z{c@q7@JD
zClBF8k#bi3mY(@gE~-72tbCU%dv>cG8c$S)N04~Kb6!Q-CtfI0-ts5><+zjX|NJMj
zrz|r+m=_JMy!nOSetCDwywdoB%<uQX+fJVJ#r7x9NtidxZ;<cEB4<UbKdOgEkhp81
zb<dJlZ~43W1G3uf%Eo!o^Cx)7it0})vwmBs_T-f<9`owqIV<kf=<C;~&z_WktS|Vp
zh4?-Bt#M>s?z!OJbLK9ZT+<HaLE0xjh#sW9zQei?>U-qanstzL=C7i~2j%D|GKd|l
zk0;JJX^*F0ow7{3<bwy5(Nix-|7jnYbddT$);r=B>pPft9((!OPgw{*7Fuyg_4F$*
z!mou{*FVdC)oWbQ^N7lM=P~tb9AcdVHO`>7{2e{(DfumAJ+sg%ho1JzMQdEWdh_~`
zm7nnd>0efaZwskcF>C$mpLzXF^S^vz!_RlzX+)Y8wLCKWt^<Cz=%aa~HX3r^HlHVl
zOlrOi-n(AUSI-#VWXS#JpLF>T=dU@W|H0FL)v-gvA@}!d*kk@js|?}$rOXHXp7#NX
zBNo3|_1JA8?SodmS>ui8zh1iKub$qsPvci!*|S^IJ$v$|=e5_`*XcRLlGP8&<=^r@
z^!Q&<B+g`xFFBu6P_GrY(PM|AH{Rv)tMy+n`LL0jH<**cC;q^D#mI~^{>!}*_d+1+
zC3cb@S=%RmS*Ycy$3o4ItnFGd`4!1eTvx<Di=wuN?A51s)D_vY+e*iNUDqv{^-vLi
zwvc&R6tz8-89xiXcA0k;YTU47{MJI|byl?cW9hAVs`kPk<L}vH`44)nA3aEWie9^#
zZuu8_&98Fa^*V2U`fDNUQC3tx(*CGy#dTM2{Jmuz*Sxajryj2sf9SQ}S!K%}^cGKE
z{8%Ra-`}agUeFtNtNpBcYyP9Rc*;Aj^vA-i^?80^rCa^Ts<-S`z10uRr!xLj6g@vF
zTCY_O`&ke0XZ)rpDt}fzUVk-R>qTZA0Cl~gUlvjisQ!drk$!1@OSZ;A^_HFJ88=Yv
z@?>pKW!kaOtDkh8$5wrn(p$Wvx9*oow~+pLs5~NPjpMrRtL(L}e5;H-7JB1g@uPa{
z9HDwGr?Ry#SG~pwWag)Z#63leXNx!0YkMBAq-+0?wSG(1xTiApS!ng!(yM<hBRlUu
zS$5>r69*OH)kCj6<uR*l@rGXKL#1WfQDlBtsBuSS=8MIf$G6&}vgP0Cm+>>>IX-sQ
zcl7+t<GZe;>-<E{YM;e}##>L$>u;8yqu2SNvS&|Lf5vXwU0!~T9eI)YTU)4pWt|W3
zD=+?f{k5Lo&~8!0zN~2Riyy)ZvWM!=DtqO<bdOg}$Nv?HCyMk}`PY4<rdzT#4mInk
zu^w6T+!`;{dpv4>mA&#_x{gC$*{Wa1OJ$31)oVGG)n6(tYkSB#4wkI_LFT<s3-Rye
zh1R)R^;Uee^jZ5;bR0eZ)i{?|#^3P+4=sC$L-f;nKi*ouEPu1~RzEHK&|CGXp8T6`
z_Jcl8?L9wT^wjUxeCEE-QuOp=k3BZs=c1QCtNDF}eoxMldEXHK<9Rtq{Z>Eo>S?E1
zh`rT99iRFp<2Tj9`o>3X+gmN(srTe<S`OK1y|>c5SgrGs7sty@r{9W`TkNyS=9^DS
zDut{k)k15(TPZ#JqiSKL?5yp2*mqP5t2M8*UXQm89_cXgr`yg=tA$?sMbmj75VYo>
z_uQ!FylLFCd*kn|6T~r%L*D+#+o$O7R#m253$1>!zq8QVFXq)NAIRQ#R4U!t|D*T(
z#v7k%r+;~Tx1IJsWmsZix%+6Z|6cu;f2qFc{O5dPA@BWK=-Kb3Yd=)B<`3V0vryxL
z7hg!X;vahQS;+jekl*!N$UTGdfQ&zR^TOgy^;SKW{Z<@&VB(<%y|?CSHNR`Q;mW_g
z{p=cR*2w(cW&4kRmYlQQYC{w~|E2$V(XzMFddn`0-@N*)@mlvS-hQ;wJZOE^JW;*H
zA9~9ltF5=<B6@Eet#Kosc*uR5BI67{S<!3H<I`)$>aVGHo=<4J^6W1+-CM7`@y0JK
z)cKKDW?l2_tX8^?M^@S6tz3Ie|N8fNq}=+w^4`4AbA`$tkGeiyJN~6_&OdNWGW_XH
zU)|(KuP=jn@252Xf^$w-XV%;cQc&ad=~LgFI&rlRQwy~|m9gJK`g_yp(>m<D*4K%m
z`ja<4n$A9|C|c(<=BJ|CtFmW@=U?z;p^l%*Uj3TC(sH@&E8i-6@qu)!Us|8aRyp+L
z_AhHaUOQGg^;Zh5c04`fpy=&Os-0dgFP!U6p0(L++wT8WYN2<3NIbSs*8?jbdiCSH
zGWCGo{!8PYCsUs{zPdlQWa}Qo^UHG6z4%VLu2aY!Po$GyQTZ#^pDE9{gXoK*+G*`C
z&=+loa=QP>JI>T&p;w>IH!pwQbmotR%C{w3{qyuzeX7@ZuQKt|LgtU6&V!<|w?8Lc
z+eaqdLh@N?ohQ*3?U$BQ-j|V`^SZVFELX4ZN2;86UTA*p=gpVjaKjG{oRwHeytU98
zNAx=HJz3|Om)}d*`c(GHTk!<FSDtho2bI-7RkrR=v_I97)n1i5uXWOr5d#{OEIfSA
zzP(QWb(4~!<yY0#6Hh>%<5_4ue?f1>JM>xmLw*oH@XF(77UHiKTKz`Pe6~>IiptdE
zp=H0<f6FfPx(?@+z4q}V3-un%llktDV)wh>9MkpKRZBYmJo~)<dgG#YEF<6Bqs<R4
zU9)t=zftdv@MjDE@!M~btv{Na=EZW?+pO_Y@Ab-+CvS4hYip06k$PyI@2vB4*YB_S
z_LDCbq>6m^3bgu-Uuime3-vs%<@3tS+pOrlpQrYz%sC6x^bH=|ezRZnn2~_YH_gYn
z+lphH!#GER=q<!93$1afmOk%z!kdLW=d;j@UtW6AdB*(n%3I^&*`?{$KF^DD*y*9x
zQ*Io~>j$2FR=UN%#XEYdoa&VqOE%Bt#&?}~FTv`URUdk_v#4y%EA;TCsQWmTf70ns
z_w4qYPZJM`mlj%a552aFtoEwx`KM)%rMLL^=8?Cr@$9tL6ZBqr`0~)RQ~T-FhyPnx
zt@X*PSNp+yvrzp?W&G4aJ$GsQ<;rT0%J_#B*Hmx$IeO{?)qeIh7Q%<NYstFLQd!%l
zeo)t0?5-_T`&8EPQd#paZ@FCFbUak{>c<Z))cTMue?(6{3(038{$|A=(k-<1e_8cb
z{6cTFTibeTJy1RC8>s7z##@Vb^g3@mKg*j=I~IENY5#Owyz=FyYx^oIe=1w^0=?$1
zw5)vPm33WFS>tk1+3L4fpH*(jsRwSoVbgU>r+jqnUaxPwdda$HLvP(vqStX(+467n
z#4FAhdB;KX^Y<bw)ch*bFAJ^xK6;HyDs!)9A@SEjYrR3Q{ZZMQ_g1@E^^B{9%7@D8
zKPp@GqSyYaOg;-eezM*(7R{&qRXOW8SoxIiyfX8(C^Anh^!T^#M`}y2&)HS>`1I`b
z`fr_k&=(ySEvNHFW#Wy6nqOs)C#zocS|7Zzj)3<(+j_>8kFPt#!aJV6=&GSttUkm-
z>-#t8tvHU}T8Gfz_{AO9pY_8ALoB2n3#rFKs~mc(9D4p9lZDh{Av{=Um9xg#8h6IO
zD0<H|IagSycB$;mdn;a}x7I84Ry;-T@dDpP(Od7Z(?VV6RJO{Y_xeXZ3%&9>o+^91
zdG&kctDWxI=f!K*9}AT)m5DbNTJ@sW`c$^Y9lgh|#}D;b=#}@{FE`z5U+0_3<(Aj|
zh05jDPkHLq@mJa7Q~R&7_g)U~V_C@iU=|X$E%f58RSvzij-j{gu9O}-iXwLAMfQat
z>#K#HpJYuZPAM*3XT32G44jl$n74m<^K;JuAD|V-&}%<b_IT0!DtqO<e(HQxnSNO4
z%`f7Oh1%Y|o3$9U@0)83QT*wi<L;cl&-bZ?{GC$^Ex$!y?z-)bgVlfZ#A6FR`!sIn
zl|B2tbox;)T7HJUTsx@8LY-gB$i{c7@MAAdd;C~-q4(AmZ$5hIUcB)5_x#Q)@A2c!
z1FyW-KgQ8S=XWgD{6$Z^v5<V#!e#tf`eWIZRqxsDjX!>Fp;zAXKhHj|pI-YOzn1@c
zew#Jj^Kb9|hVK=FjDyFEIbZ%AFTKB$?!~pD=^ysKc+)vIeUW(fTm15z)hqA$MN#`H
zr|<u(?5#VVUy#l|w<xL|_?tKHym?DH&vTKB;=b#gfBDE+AEb)bID7NjDyMp{o~-E}
zpH}*!N4p;Q_RQ%C==p8A>G+i*@0qCHlC$!n`G{YYM&4ua;vVVDV@1*}^zOs6rqlnH
z^Pjx+rHAfKJS;k|DChZ&6>qcZb)3C&)^in4Pk$5{N1jJn^9{YnhsI5nz5ZGLgI?`N
ze)-B@Juu;#Nhzp*p9A|XwB{>%s~mcbkH}g5+8aM>zl9z<EcEQK+PCf@&|A;*RIks?
zk>SZh`l*N=9(wf;IOMv2H0`y{5D*@*6aKz#^U+>?+D}jWjW~J7E{DIHE@-&8;o=Wp
zPM&*u-Q%u2d0etrqpx3|K6_FPzVOs<j{Vbv)05jyp7h1`C(o%t-nT+;<s;of=C6g!
zw{sS?YBP4>s1!u6yr5@Zv2Ro)y(rR-g=&||Ryoys@k;$vW&A)<^W~MbKb4YKY5u!4
z2hMmQwNUvWzFDaKwq)-<rf9m_>&ZI)dHqv)$Xj0Xt3AXM)_qM^+3T;B&bq7isBDcV
zdiGnC+iTs8`i*#CX6m8xs512yMfL+G{`)@2tKW*7=)LEDUi!peuJ_#TQ|BcX_G_?5
zqy4U0lvp@-uVnGo-M&mL^!kUN6-BFE^xnKy|Kz-3p*f!azJF0a@?@TG<wcJ#?U%|{
zd8^;(2YoWX$KT)lDg~{2vg*(JW^mJQ|GswVxN{b^Jb3JC%iuEyO?&YEj_Z{y<le(V
z&;OJcm5D1Bdi~@c#X_rI^j`hed-d86OV)jp${v6Cg@yE^C|Y)_-ukXYRz2&9h2A)L
z{quPA(!DrlrF-WNZ(nG=|CCkFxx_-<PpGWptg^=&>!gKVd{%o@_SOT<pI7$$ihWd3
zw0J?!InY8Y?xOeX^7!}0*W=$Sul}TR-tp4>TA#{Z|M@OUr4hgK;u?Nnp=FoyV9DP2
zkxqGUJT=`~chOt%9z8r+NIe#M-zBhq2aMh;Z?#*l{^s5Oy(8`ayZ)!1v318CJMPqR
z)hqfR(Z6GZ|N8fgre|z$MvIQCT;BhTjs8pCrsJv|SMBu8>HYejard9vjY(F^yPu8x
z!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`
z1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9F!28|(C@GZe|l@f$C7gKudTk>WwU2LO#)YJ
z9?p5_2Y=e7_2)@_!(ZN2db7=*3zBj%f<`d#-x=uLZ|z^cJ$O`NVbW*Qt9tF$sFW2y
zXn6Z$uROaX@o@MXQy0Cs;hZEe9}oY1?Bf5yKrj#t)F%eIPW;h5KRM>Zq*^%iy^rRt
zap8MOxmce#4O@eOU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>
z27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?I
zU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}
zfnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl
z7zhS}fnXpQ2nK?IU?3O>27-ZLV8vvh%VmFfe9g1|l;p+H*I)X}wmbYQ$%;Sf{KT&w
zSoB^}TUg&m>(>)5+T^*WAEh3)y?f~Tr~PYU8W=o;AK3GtzQ=>GKNtuGf`MQl7zhS}
zfnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`Qs%
z;MyN<cwA|fkCRH_*KK~Y!w1*AofO4|&mMl`ewSR81_lq|2em&BYAfzUJHbFO5De5;
z2A=Ew+D;wky`3u7*Zzdv!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<Lwg
zKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&
z3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1Jz+*<TZn5
zcOO3~4GbP?!w;(C#ZhZ85DWwZ!9Xw&3<LwgKrj#t1Oq+;ZRa)^zsZ&#q(#whA#wx*
z!9Xw&3<LwgKrm3Z8F=%A!*A?-%}1$)-?x6`+2=kSlY02)Kb}6ZbI<W<xmdT^3hRP_
zy3N4BL*E!R>+I*!a`DhpUwCKqlqb``;3112)RkE=`|6_)*#B=ICRwrN_iNtu^#==*
z+Qz!VSJ)8@1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ
z!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<Lwgz>3bmoTrCOxNz5x({k}2O+MW4<B^}FbrlbP`n+xKd46JQVO_Pm
zT)S3u|Be0y1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<Lwg
zKrj#t)DH%}-eSUE_IPP%QrEG5uo+ec1HnKr5DWwZwadWz(;KgQ<{l%{YU5|gRtLT_
z`n|N;Si9^+Yr#M;5DWwZ!9Xw&3<LwgKrm337}%l7wWl^XdU6^VJgiK9P?zRleYEAx
z0ekh`@w^Yx<%Rtko;zy9lF5mOt<RmmwCO%AN_p|*Qkw(5?a`tX7(9d@toZZb*nt=A
zcvR=HX<l6KxJT}5dDDc{!#1n-KjXOt<I>v3yuZV)k84qF*u2Zz-yJ`E&&!hfjvbbs
z(f+k_-bm{!R{Qx#wd-GA`=v^^+o#j%U(SE}!?eDk7pE(oUK_EgHrw#llXAbuvZ9r*
z&9PaT+NyS(w(4)%zf<%1w~kAe7wQ-njq7>$xpgJK_CMrcAQ%V+f`MQl7zhS}fnXpQ
z2nK?I6^jAy-l?vquUMl|`@N|v?<duM&!fd)AQ%V+f`MQl7zhS}fx5^*UD=;nb|jN7
zIpFXsC#BWKx;QprQ!o$=1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrpb}8F=F67yCUq
zEM2A2xZFDy1_T4aKrj#t1Ovf9Fc1s`1HnL12A0?F<BQrG<${4=AQ%V+f`MQl7zhS}
zft87Yyzkp){T<`V#9{O`7zhS}f#u0S-oO90Y0r12o&V(ZsfW*Xe{H9Z^WIML;_~Do
z+71SSfnXpQsJ{%{zu>SJ{@h}8S}mM9yY<?S&v+osi}jb$usj$D1}p~J^>}Q%b1r%{
z@$l;{Cj4cOmxd;J(P}sJ!9X<`sPyyrtoMJB@52p09{EYSqVRC|)8}n-&+`-0YGXC$
zLew1$1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&
z3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1H~Ay{yubl
z)%WRi`j_*c{xG#rzi(mXQ++XZqf{^u3<LwgKz(JPzOI8|_qC@sIC^p#JOmHr_(6S*
zTVZ$oU?A`Jg4R3kk^5TSG$E}t)(^JA%3vTE2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}
zfnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl
z7zhS}fnXpQ2nK?IU?3O>27-ZLAQ-4k25#Q%->vokUH?<h*t+9R9e3`y>J|Nu=-;uy
zfBkz#(=#?WqeaJ6F7JQFM*k&m({a^~t9JV4^nU%%xcg7-#w4rNrmfLdFi<fDc6jdL
zqo%YRohpudcAYDy+%Yk=@S^Sm&pW=`lvJ@|h8Y!E_AD>GVdhGwUb@41m$aYIury)J
zflpoE=O2lMi?-OH|DY2_C!pqg=;x2m>-ze$3HX=!$KKv_{>h1j4Tt^yvcZQ;PihNw
zU^{I2*eQd?&q+Y~1)@JNIp^*}pZOvIhy1!U@XePeCKl%H&u{O)sKcP!ze_;I!@?I{
zKYsPgCVrWK?|wDum|frhCIL^~WWvp*%cmqD{jpGaQCa7wHJ<2u-@NwXwogBkdZ_k1
z`R2H12M-^SDAK;#Ro~@S8~p6FQQJ*VLHMJ;I=@t=oTBnt+j6fBChj`w;ql9?FRVA}
z7rK5S!$YOf@*Czm_>&F?o;7%#C5eaRw@}M7&Y+GTGV@u{vcFP#E3OsQ=Z%M1^Iv`Y
zBfU#MUzpTYtkif!J^woc$MqR^<M8GUN{Z?avp#Kd-J87^Bp%KgaqwYV-9I-`-238#
zdfoNu_(W0XgIcNUSYDYpNjz0-zv}PWe!u9AG%u>1AqNA&K;;-9&SagJ#F1{lxuD&f
zgXg5>BKs%Bs~)~)(+!?jrKDK7p08LnS^E&(hi4sU&8OpI$y!cjs~mdWe_FEkOJ(XS
z7j+%e`Yhhj(|;}J$&@P>XRS4C@zu?yCZOf@J9OD*?W@mUtpqa9E%fZS(yjI@rH8Ld
z)fe>y1HnKrP}dpYeg@2XezW3xrS(PmQ#-QuOY><xDszt3a(Yju`;4M;Q9dZA`(l-?
zeN(0N*aNcu;)i*W@)pugwNS@f<-GUP*xByNzn{40S!<U-y$AB-tmDT1O;PLDepO4h
zo(t(cF|yT9cnk~%>H-6LZ_D#mQ0J3nvlZt&ebIG?a>w^tyx`k`i;}#^JugUoVAgoA
z>ujZD>psfjU-edb^m?vE4>E4ZApBGtwLc*T1Jz&vKlf1g*?DE+Y2g2SuSfY{j2Vv|
z=54b5uiMW^XCBw<{CQW6$c)re*YMT!<0B89KQj%i-#p--%L}dbx3=uj_~*?t-EUZO
zm+S89z4KkO)7nDb`&m)Z8y72`=RQSI<9c5Cp>vy^vupENNl~=s2l}k`aG#bJi}FUf
ztoJkYM^WczZOeLYwB*{hD{Fh|ANu?><p2MD)0Kq>onOdV_oZ3y6RdpJ{IT>_{P6r=
z-}}}5gjKKR%PU*`&Z}o!@?u_l>wEsZ{N&BAeq#9@>s?+fx-Zaj)gDJ{erDBM`?Jso
z1HnKrpmAB_imso*STGO_1Ovf9Wf`#UGd+FL{ij;b?~b#gwI5&6^~&3dD&u#$Kd82>
z?`NrOy%)vr@D;81u8Qh)f8@#5e9(K6qVkI3e;4lCaoFK^%}Fh^{7%QaQu2zz>+<S{
z_a0NF?BYGrqFAoK==mtGtna7Rw)~U3dktD+#JHp=YJAQr>-h8ine}%&>+b!UOiMw&
zKf&`=^b3FYVzc!}O->(Lb*EuJ8vLK{age^(i1}YM=svS1o%cYVnz+Vw&yAX4zCY37
zvA3Qb(05h}(k@7Q)T8xcZ-;Y_p0wVPV^Y#V>{489lg%5p8v0&(@kQ@<I`+)B(rZRs
z|4M_&A1(7eDD)T4dv5VrZ%nB9UI+EFj=%#vv7VED+ZJ6LTy^KPRPCUjH8^I(^@l9|
zpyqogHG26D3VQ0V>$t-S1I8|RbiGnm)O|MNtO)Olj4#OeDKah=>byaIZMWCIJE_y0
zq$vJ=r-%D&H)%;y6xaO6^Mn3!)S9KDNI$@v_kZ_*Uv1a0R1{zO_K;R<9Qt-z6n|2B
zdyV@qxi2k>eCMfJ2v4<*{63*5di!NJ{oiqw_?j1W9_N*HzUGx%{p#!oxA=Tf0{-Zr
zn<uQh`W?x<llq?2ZSc$#Y}D?zpB&tNdV2mZe*M~`?PsMk&c5RA9`k3V8#Uc_&}D;X
zrGNdr@$lE1%u4TAukD&YnLi^XKk^axynf!b?PsRn9`=V;UHkrLy&HS|V`ImRnpv~H
zvCegVvsLTc8c(c=^Q>>W-syXX#6OMyw?B6JWdpwZxMn>g9(LPp&EM>I&A9aGCA*&X
z%!J`7&+XP<viP6ROng7(I}F_?o%q?9=5uSlE2ift{jLl3D0X<I*>0~6UccmF+lAv-
z>$pwxk|OqaSakoHb^f0I+R?}Ec+2{wqIluK_pSdy-*rnx@v<MBI`zggS1;v7`lrux
zk@N1C$<OaDJXAl)E7P8Z*11&m@K7%Hdg7!Ry&J7nDvCNkJh@zZy>p<Kp7s40&6m~g
zJw88k_sR3tC>2F7F62$mYIkkr%ey{m{?l4rcg=M}Hz+MH)c5f6%I<zdzDKJ0i{3L<
zs$5?GBY&m*tK9n)>%O<B-s>0V=oK9;9$InUFNhbt_};UB{qa{xQM_U6QBS2;%}a{n
z)^~QD_soeOCq;468jtR`_kA;xqIhK2E+=ffTZ_`lj+TFYw^{$m=l1wIDT>y-C$4X{
zc;8hzb!b{rB(8xPr;%T3xAeA!Q(Kf2`%Rg;_x4{mEm1xzK5|ay+c)~*IwkHmLEd}j
zo{aMm_dX9EKjV(w8^4x-?3agMepR~}jowczw0O^|_x6KD)3fqbZoa8)4(jyeu8m7z
zQG1tHxkmrm_QFQjv?vut?s*lnj=$!!c+mG?t$C@xd!fGZ+kLU?R|_X69&!(m6^R3i
z8*TXNhU>JPU-<Vx^!MI(b{aXi?b4*usOQD3vfhgz>o{hW*>~35-~9VIL*FC%_5S|(
zX(_m8kF9>wqvt2-D<i+}H0aLh>7*O~dDFJ-r_~_ODZt;J^XRi@+&MK>WS{lU;~zF_
zH+V+Py%f(SNC%0JnvP65_g>V?{S<MQe%4paI-Z)(^Q#^czW!o^b-qlxG`Ql&?G}H!
z3=Utr&9Gme*Q}%nkJh{>8qc&`vy;9V_0{loN|i$Weo0>0x)-wa%umqDmshXxuBfcv
z&+=sUO+^tNiemo`r55j<ynd<Nb0+&TMelx*^{i5O_oP22+m35qsuVIW%f)i{Wt7+V
z1C;M_<-GG(<6~WsS+A;voI8g8{pDlkwV#uMfBWdu_kY=aPCDVD>pr<~`#Ckgqvtsk
z>E!>a(@CeT(R5DQzwr+aozQnq&GRdsV<}qvcG+jiw{NxnII+-*57xYUy7Qc${&VNq
ziK5oSzV4tCHs53SwLVI8->lCI^nTOxSMIm1c&>3;`(u?OANFxR1g&!D=_mGqq{9nH
zdlr%pq#g^s_l%WiOQwFtLH9Rmua3XUdS20VWOxQ?Us3Hrrk<>*exfqx8%4&|LhP}y
z`My1V)a2nb@zCps@`~*FLD6*lg>^^KibJ0Mo^k8<9CCl7Qf;HPzp}>7(_8%{E?CI(
ze%6oWg}i@)J)o|amA22~Evw!e7t*uV7x^m70Pnl-Ud+3HdA8-WCd+<Li2DZiMJL?w
zQrk<K{O3Iz$}6gWa<8H9)o@QidEUp-e570W%(}M@J^Pj>B@gjC3t8vbzw5e6InH6@
z8}Bh}=ac8x{QX7BYaBah<m!_T=>Bd3k`H`h|1sS<zHm=U{pBL{QV%lc25<eQe<1r0
zkab7Vi$hkt#Xb+|Pi-OnEs9ou(Lc5RnN4rp`h{iwP8NC(d5;>@@lZK0Z#te@pC?x~
zYWcM>zby3p%S-qAo0nhmd;3Z+-Lu<Dr+v`tAL-=Ji~9ajQMu@Rq@3?b<>k<MgRJ*#
zw8HzYOHRGL<<r|uOXpnr+L1qOGPMTLA2fRF(;hf*QrhHzYX;o${N%LV`|A!ma@5qC
z=M(z7vQPi<vx_=!{B&~0AD>zHyHV59(f5vczfIriDbEXTdjGHIA9>`sbhXYmjU6~|
zN;=|=4TgRG!KC#0ul8?u+Pq0=^UV)_a8kD!HNT(WJ>)gt{o^Ue{`Hge?$*6t8rOKT
ziM*%F`j7oQPhlUze&G*ye4F&&@!{muPS1Yd?#OSG8*bWs!NL9JB=gVzTia8HyqhY%
zIORtj)@(mD1$iI%($Z;vysyd3RB`t&cc1ps^Hb8so9#U3&-4GY-{E~>3!fc6?WSY5
zpI!6Z<`+*+?Kykiw3_<4|LHXQuiK7VqicF#%Rv*z@Bc)4#ZmXPJ#4$lHSZ_$yk_55
zMlO7C@a**3T~2Iw%c$uoc5~0fd(SuTe*U-951f*o)wBP%8*M)`-SWp5x4iVgiRl)r
zY;o54eP`B;3v%nZtMuA*^z54aYwvvcw9oFGS;G(g!w=vCp1$Aj=+<A4nw8$rYTpif
zZa*uX+V|lB@7y_~=J$$Qw>Y8I)^|=$w_R<Y*8iM0wFb36$+Vw5_;{bm=>f0Kd3@n^
zGt!+uS+aA}yJpq+6ZdPh%RckuK6mW1Rrguu@0N65m{n$;)fVbGJgcnx#=LUgc<ik!
zn$GV4n9s~xMQgn|`^0k^yx4bkO}x75XMOfr&}VAO`D1Rcdyg5?ZF)+4As!ui_=cZ#
z`EEk`+3N?7Z9V?Abl#t*Ox$<v7t^J8u65Twr$1B^A4s2j%cyPcxa{vq-tlIA$cuVT
zuw>RxkaI>+%-cWq^F@*N)GtCV4Se(EiAh!-be)D*Q0Kk+leVMzRkq$+$*Z^GnDyLE
z^Lut_{POHm9<s{HXI7c<$%@+lO37CH>@z*g8{afP=ht$Pc|bh$p2KHNx8kX_E}{4Q
z$9le?dd5fnS!Lx_Wq2uy*rDeROD=jI(|*;Ktnma{+xPhM;!NIj?%6=bRqZM&d;P(m
zvtn&M=fNMdBD{fKzq8hB<;xn+uqP|xmqpQwOPZc_+&7r?YOArgj7h4E`n`|5^C>IO
zdGjf6p8WTQ2mP_vc2jD;+rhofTfbQA&yT+Gahewo*m$FPKl^-f!h0r1b>4a4z#d~$
zekZ*5qC@X)yxqi<a}hG{r*e-%`i%z;SopJf6KlQ)()F3|w%fbg<dpjg?jIDb`A}5v
z&C|T;S?7&4epY+LU#py@FLxhj)r&n@vFP|LuX5bGF>kmZ<UOn<Zw_4i-f^pzxOe3K
zi2EVX;>Q|i)n|?KMe|wqq5s(hZ>Oh!_kB_nTh3|S_}mtaO0R$a%duV3H`7&q`Skk@
z@0^t0`{X72&F(QiJ#z5povuFg-IRNi?)^u9F|yB;n)Ek5Xu0OMji=Pa1@4iFEBtPU
zc=PcCv#;$pZ*t8&G55s_`b}tjcAtqg-|ykQbnYwfZ+6&g7j>IjbDzjPH1_a4hR1F=
zrr{;mOh`fQw;6BlS3jS2dBgW!nVEdK_Lk#by|zJ#c>nnPA79zL@3fTX$J|G8-^qP9
z_qg~Mc<w3JzO>H#>1pdDn;f=Rqei6{yRN>V<$cqWmwN8<)<%E*sOBD?{5%Kuka@;;
zAh3`4-lX+uA9U_=ecI#lcOQS_z{xfDt=P-GF6TM!-|=VkJV&SB$kfX{DfhnU$qzE_
z)_W%8qhIA>(fQ?-gC9lg&x*`fMJs+6)n{GbG@telS@(bF6}^3!_xnZf`wZ5(0>1^-
z4&HlG-jRnq^knbD=8R0TBEM_Qi`Ko6=ReFhMdBG~)x-Q%BtN`?%v(^;_r#5Ap^jgr
zW!-NPuZgFuf2?2dM7(3Y;&~J66zd+p*W&kGJl|qnQKTH}Szgq*iNBYNTCd7F|5eue
zA5B-eXuQKNeZLo(d0rGXPO7Z?F48^zt>-3`w^04plCAnxZ?#h`y=8A!J>#7f)z2*1
ziqEt|e5^L|yItZB=Xv52&pkQk5dV2z&$)=_^}KJ%{*`rveK7Hh^Ar1Rp4W4}=iJYG
z2E13ozMlOo_ZmFk=X`L;(ACdsf9I5x_e9QYyTy;MUphLyZ|p~(e%a>zw84U*M>YR$
zT>8!DkKFP4H6N#+t@r+@9eRG0QXhKu`@F}(bKOTaJhS)k4<@F(hrs&|<Og}rg!e@#
zPk%kc9|Hex2CV0JMfF+Z5mphG_4g3`-d5j-oPF&-xBl$5S8DF#crO9}$1m6qFi)6A
z%tPiE^ND%NJY*fxeKqgVvJR4ueHiB`WQ{}pz8v}FOKr!dAoB*huoJ|7koysk{f^pM
zRAyWPgMlmz@c!D0f>xi4&ZkE%c;i=nkKDMVsK5Vb9}HS{czWxeg!i~WjY}$1kLro5
zAp10A;@--F@U*<p^9Rzs?@g1hC>EU`<(4aIZ<JfH8Q}L-oWs~(f$aZyF2!?I?%%nG
zBd&4pkA9EI&E}lcWO@p+5957izFSRx?whzrLI&>`{ravC3>}?<-sqG&Z}OHe`W@)J
z<yil!g<0<}$j3h4L+@U~8&5C&-PMLJUaho7skTw?p^D197p0ij53T&(xs!8RQS`<q
zYkIkUt@r$voVTBzf0J(UUnxE9)KyeJ%R25>z6Ce!*7b{1S1nZwwO#(+0;uOU?nC<R
z`130d=svsV?=i5Bfvl_i4v2e8?k#wpJ?yo`jqmO@t>*7JaF4<I$oki^?d8W@-)B<I
zclKC!dEbEbnfCd;+O9YK@|11g7+3TAkxzE{rs*w5PDo$5_1g2d=rJng@6zZwrd%0b
z)gL_B^J|^AJjb#y>$!vXCOqUk2U_cz>i^Yo<KLV(Zc+kLUQx?gaRYuqD<67o2ifYc
z^6bgfr-*%(M()S+;-9vBamm>&ze@HWaQmpY2EL!3(DmR`7Ec_N&e*-hThoV*Nk<;B
z<-y;#9iPs;c}k}Xp8wD9K6&4V=L{h8l;@-DkD0eTXW$-_`vsmi@SOIPhZfD=al1(=
z_x%@~H>~I2#*<Ro(fQ5uSL)^cl`%U`I_iMQpQK+l`|SFQCXHPNxqoeY#FVwaKl;xJ
z&oMywT)5Xk`&~J3Wcq`xx}UV`s3|5=pQ1JIytrbO%d7X|4C(ka&uMrbbllE!pZsi-
zk5iui@E$8XzcziZ2FE-<G2La2E}yjQHmTxh&6m79Xnv3XYNzA>7IH4liq`uZR$TM+
z8ow-Ad9`H96NfF-@|085e8}WOt`=JRY4jGqo?lu1j-LKmNIq}>P;R=e&&XE2*kz&R
z4@LFFv)aa@_EIjdz2*9M-uBAPkDqu5A0Ar%lvVH9QD4*Z{=Fe<-dg&i_d8k3W&J&8
z(R_z&bIQ?6HrcdPU$NZ#A+NmtZndavy;sY7Bg|9gndNWjH6G}DAD*oJ*7cab&uF2Z
zueoQa7FzK(Z+qNRgVyh$Jw4-QA@wWjxFK`?D2kRHs>jbNjo$jJ{)NoF6^LEdzER)5
ztu5J#*VwB(S6XJD%e>R~1xV-I4S(1d7z_jh!9Xxji~-^+>lVnm1rmQj^t^WqQXV8<
zrI7m<knhM;3b9*}bqT(Hd)ws8fA;dA1hn>(mD1C0rP11dR!g6EoH!S-?t|<v6**^E
zsQL5Cx-VlsWJUEWZ+}@dz36_su9o9m$9VDmKnvMVDlYupi_O*_H96()zw!6mEF`M(
zz8Ki;Hy5;fbMTxLBtC#VSF^BE^B7(f^>>&$pRpU%{K)(cSp9->9$M!MORv1>IaB+=
zJXgKSyl<rXqO$Vg*{%IRCcb4w`de-Ec(KNhzt;mF8t-%;K-}cMLuK?@PW9|RK;|j&
z6B*?9WX$inimZ1Q!n1|iFK@m^`u~jU-hM^>JFCoi>i8g2PkkQ`eotp%m+S89z4KlF
z`5w0?^S$7r$oE)4{rkMu-w(*D&-(tJ#&5<O)Od#;B;AwAS83$$pn=#?E>_x3>SI4s
zTUhCP(%P<%zb~$+`yOQP`w66%%fH0~akE@x{|>4@qX*fqc&P6IVW(A&@9-;fo?-u@
z$i4`~|3UPiwog8ge6&N{<^5j1Tfp~K6!{JU`z~bi>A8#jT2{oq+Cpoe&b|}G9{LTR
zpz^LV=T!377BbHCLlM8Hf1sXo(U*(tlRVUW8TiWTN1Bh{_j{=AYJVBWtf=vZ_b@^4
zIladd>1wCSI_}7t5B~=7ACPj2>hH+bdyeS!K9x8c7z|W`0qfp_eU75F9$0!UkIelY
z_nVYgd&m#!IYMRjTjU3+mv$^9zt)Rh%V|Dj@`IGaj%p+ATJ7LFtQ%QT<25qt39=&N
zjGp|8<g*YSJ@mfUsOkFsM)>p4+MiqYu^uTh9+g7x9;d#iTmH(r3bKA{e)OC#NLM{F
z&s~to&prpgW?#U*Uoq?W>b#{r57n<KEfe30V%GjgzW<#8>)v}s*Z=QqEUT^5{qeFj
zS5`y3D2lpnTGlBatgGnp8)OfOCz?;=6*Bc($T~@WMeHq#T3=TAh1ZW?{j!N)CW_eW
zp^k^QkH~75m(Q}xDp%D1w4C*PRC!Pte=3T+7nQX<_c{0@<=_oj=NB^NK)nw^#x4u_
z?xT(i^?IoMTluj&D_ZrSx9WknqR2js{Zq9NKUF_Ork|kNq4zq-*w6lqeImc})_N)>
z(~gDMRc%y%RGE1PlFmLIL=S3y^dQeq;gjdC=<!$XYiS20AISHyK*kefyp?b4QKTN)
z<DQUn0kR_PSO_l`@^?2tYhGe6{%)a`xAJAx6ZaS&kn-5&A^9m!y^6$l@-ZHu*AD5l
zhpb4tg=&|T554ABS@%VjKhST`%1689BD`4j;7?XMc7obZWUIXDiSHJ&ZU^={f8JFi
z(iM#dYuqcPx9$P!i$3dnJXXF+@mRDT*8Td5dY-P7T$E3*T(gtD8THlhbxQRW^*x{Z
zCfAp^va;C8J|!=D^DNTq76UwAuIp&AX3a;wlVGLi{ho1M&tL8D30K=c_&rTlwEUsk
z`k`lDGW>7<Xi%yxtag8*{$OBbU|@r*Zu#Wu_ZBAg73Vy4!?csu{VoX%9_peWcroOj
z$G189)+5$01uh>R+74^d@5AkzmI8x^@PoQL54<>9H2s)Ajd}a1(Tz*>6)Sar9QFLq
z-^T|Z!N-694$g|=4@Ki-lnVxefnXpQ2nK?IU_cpY-+rsdKfi9hQsDp1gYuSF&iWj%
zw)5rfN74LCMqfPijmu{w)xx57t?cDaUAo2gz30wPDvfo8*AC6c-2D6#tC#94Zo5&#
z{hnT@S*gBa(Z82dSITvGrP*$;4qm^sym5KWr<n&|w(7};HY%;GxV(5@(c0zxwQ?~l
zFGat95aoV<+2BK_Co3usy!V*ZkCu17oA-bBfM0Fbuv9Iie|L<2eb)zuj!r9ut>>=N
zYtzxQ)0GY36SRJ(#qaMF`8xpoewE+Nc6jWqX9x71l_K-IQ+}t4tmvhaAA1$`eR-=t
z*6-+x>aBLr>+c6WnQ|WTJ7N!Y+=|McJytruTV7sh{XSLoe1E@E$nS(fZ``eP^=p;+
z{VGU%AbJmVK3TG^x1K-g?~GMm{hlqmUOVQ##KMn;G+pbdHZLX?T7HLK+f&*5E~w^L
zncp8<sQEov{m+{>mfxV~cj*>-<EQ?nvKN2&9zvy1{j%JBI_1kf$9wb3OSk+dtDf;L
zidoyS@_GI9(ye;4>b?5CbmC=EEcZT~@4Od9{SIYbneV6;Men;*R=Un-m9<>f@$urU
z=MQ<)J-f7DS?i}>&5zzgiwErS;=h+(G~Se3PS>Hlvb7%8mfjkdta|l_+Lp`x-J^aF
zFRz?;yfy#NE_gdV{k!jzqR9TTC}urxvOlR5R@zUp_QT7Ub)6@lw{Q2-bzQM!+VhaV
z@1fZCh$(A-fApV|thnLbBX8Ps?iY!L{5=B;`MW|E>i9D+_^z40hpO?A@4t~x*9Y!}
zvZA+sm7C7_#2Vjn^~z_Z<#PM)mDlmBmdtl5tBuw?Eq8xTdA<Lrw9LHf-=Wmvy_453
zSy=P;RaV_;*pCKJPq`n(-}w$8_p^*E>55vOc~3p$=U&%CEzft)sfTox(SxL0{qXw5
zK3CE5lh7Y~d(-(RC&5ED9@MWaxzcfh`dGiSqPHH9?(Nq!J?pupQu!*q&gQKTp0Z-z
z`n~*KyL_M6;@diZTmD-qz4cu^*1bw0>jr48SFAINtb44B$ULXyUY>hx>L)))K9KSt
z>zX3#nuR*w^?4TKiyf6h?7?1mgJ%oLr${{@=L`5$ghxxS>F})i(JRtF3;8=OpyuNo
zrF_#q{Lo*;qVr7cs4dyEcb!Maynn!t=O=lw+~+Y~e5`i5ch1@W`ptH{@7|3{7G8Mp
zed~YFcimD^^wtkAz3Be0Qr|1{?DNuV&PP9gd|ubrpG}I&*7{jj^}PSf_;C+IJmTJn
zxI;W9J|UA%IhBc<q*I>n#8VG(i1ia0#6D!~23g0}yS#a;2QGg#B_Hb_GI5o4P-XIg
zS{{3~JbJAUJqSP4j~@GJkNaZM6_pR#rQZ62JN@FHhdnyurbKbii#q?j_f8WNMQ=XA
z8>sUcy~mHX!}x$!+@3Sy;KR1Me{Q14_<;I%R#n#e^&F+|f8aNu_FHA<4Y=ASn>TDV
z^u5$VZ6DdXKgN!{=<O4Vrn3*@eu+3kJW?cHQLZTBFZi3zW76>p5PPsg-<KpE)O4ON
zXgcQ*kaUoEUEh#(9)9T$=S0%E$LF~Z^PIR$TxI>E9@6n25Iy-AFZ>I8$fx70<&Z(f
zfqEHN#+z{jHJyCaM?Ul_qX&r(^h1$2pmx$8Xw4sZ;XDLCApJ(BKH9@hkaQ3|_h9$|
z^G@d_c2J)7ufF|}-ld-}toa=Q{boJEU+@c%^4P(-6Qq6YQ6wFtANZ~2M^AqE#SVCc
zKlJo>#@Scg-DCc&6h9(<!wdOA>Z3e*5WV^h?K4l1$<N=hLVxkR=N6y!#)K54e;_<y
zZ+%DRU0%%EZ+NsYZ@bGoKl_3mx@@!d)#tBP!Y|+hSrNa)4?*S)^O$@feviy~fcnvc
z=<#dHbB_go<iig3ZPY_M$mAyuQ7=e-ctEenyaSoP<bw}I+NT|4?7>dTVF&q0rycaj
zq@$-EO{abArXPy<$%=|ccI|S)#=EsBtt_~rj$`y|#bCht`xaUC?B^_OyKww!9k*#-
zvhbd9>-QXTf1{G(iotI5Vr60AnRRa+diE_%N-G<dJHG5I^CInX?zND66p(TjvX2AV
z57riP9|Ur4RJ8aaAN}Kfzr1Mueun$0teE%tq*rg=epvb8$wPPy3<iRMU?3P+xftL+
zUhi{M=6(mgBKJgr1J66Y+mzJeL7y9!D_i>{^p+jgdA6?Ty?x8_O0TWxP+R-)Xy@Uh
zrnDTLu53K;98>$vb4JdGAn~93f8q-1)YI&!hu)e!@2v#c!rGcgD|<VP56CzwlFoYq
z$X1*sy<F6NgeR};<5}tcXI=00o}toqM18?PU1GqCYux)0A2~-UYM*#6ugHEqFVddQ
zV=InozjzN^k#$@1t4#lE8}ssH<%hp&q1CSHmDjv7<5g+QJFfhm(L?VZoOI@yBKuc(
z)O3|~e$<vsKNLNFil%2BFO5sc%9|%=^*1XX<<;k%r}&437GGKCpO#nqls~Ut%g(I&
zynd~IX+0Oy-$PjTC?B4z?W?T&PE98c;J>wnYDZQXdqH^QJwuRj>^bbPao-;HWKvg9
z`L+DT`dv!so1L?3^I1uydC+kQxw;HwUH3Gf7Z2-OdUaW@z1oZN&hJn><okuheG9X$
zBjjV90M*~}%G6U7^<FWrtotMEP(AxBOHVrGJycI)9R|JfR=V0}`5E@I9{|Zu{fsC2
z<%P-@vifzoe`kbx>l^C6zuIzr<0tI>zZtOR_5WRJ<w-rT<)De<_kSYwkaz^2#MQh=
zJ$Z5E8IkB|Fkmylezj6q?e(<M_1nxuN-$791J>_D^6I_k4qBge|7ZQ304<Mf&9_Q@
zr;K_U{cGC`8(q_)lo$0L+LFs3$*3V12nK?IU?3P+{tS5MWbP$B<ovBj54qQ@Eu`JL
zj{HuxzM=YqwZF6U>L-?*b^JU(w(8HSxAwuRxB9F4to^D~K5yUerB}+1O4q}<SGwM6
z)o0x=;z!kLw_5dPeZQ+(?MJ=AKz(F@^+4BAWS$piMdAlY{gp!NJ|wHYQg&!P`kp;9
zf5)8f3-W!?c_;NfsoUV0Dc`-+?_#$aGvfL~mVS`(zK5n4m7myuOt+3N+><KuU247)
ztBC#0H$V8nN!@0or0@I6$b}CMo}Hf6v;VgnZ9lUH`99yJrPKa+Uz3@shfP|a_Ce<^
z*QZ(0y5HyftS#GKe$4fKCZ&pek9BUZdyg5?ZF<^o$Ddz$K=;`-{rTHRpT7Ug?sL+q
zeIFk1&Yd%A#)on8aLs=_Kj<$<ty#*7RzE7G*YA(jml$Q;>lW1$k1Wg@-y+`z{Rf>e
zI;rnG=)6E?erS9zSEl}~XzkBEJ@!`$FWk4|u*2_~lX^(ptS{KV@edB2(05L%==IY|
z*S`Z%X}P}7x4L3~&*w)xJ8j<$OSO&qe&&LkcI*1ZsjHUs-iPlZD0Y6cRqNXtPfS6r
zk9`lw@9P!W=Yf1*0i<5tukrU1D7Ry?E?qX>{$B~m_pQOBI`2GiV2`n>BK2fFuV>B2
zc?DEn^UC@=Bz?bFdDZcPPf*iUW}mM}c~I}I)h_ZW!XtL`ee!EYT>nag$sg71b5D6_
z(d-?!o0O8S-_^Ec<(F~aVwEk<I=}DCRP99O9HV@2e!>nA{{-nL{Ue{^qFWa3_ww_@
z62;%%Hu>_Oy*wzv9!2F_W$yiW?w>c#S?f9VfxN$v71?K~z1qLpl9`w6ORRfdJ;$qG
z@%>Sd`sym`d|pvx{L4cf&*de1`}cd_*=gk5wo8+8@%Ubg7koQ#QDUK<KlQt8`g=Xi
zk4#*!FmD{nn*ZIeCLOcu``;v4(XubE-inh&_1^u;cboN}d~T1glf0Of$D;Z2&Y!v(
z7wei`y3SNv)^(YEjE8lNH_IOQW8JsVng>??ta|ok7HZs8S^ITdpK&)1Z{DDk74yzh
z;tcItcEhWO$`kwLyofzu-ttBBbI#y*k;ow52lVdaIWJR=`+vonzr(cZPQ!jQczVit
zLi5p{?whd(<oO5f>bxf1L+nt*4(y{Io-cv4k4!%LOTC=i$q!Oak$NoD{=$o*o<HfA
z=EGhPyFs2$DUxm>?ZBgj_<!Ka#Xx;~9}au7j*_l3UcA=0XUSRHU)l2M`OcF)emK{1
z?+fzxo!MV8uB5Yn;GRzJ-*jAw=elpDUPap7;e-KW7d*ON3AFA%)DMyApT=+Q2SM(o
zLA3{&zZ>u2zdCOGn-j-PN-Tsw?4o}7K@XB{A^iZ!$2d?gdPTjj*8Lwc<**-qIS-RB
zFEXx-yCUaw@-e;^>iq)cK<+`X6P}RiS6)=Rv&vUJe9fjCJh4g%%v!E!zNNd(IPCEa
zHYkCNXHit1^nR9oxPHGAxx=|fPg?KDF)7GC4P<;ky@#iMMdgS0KUA+W`ItAL*8lX9
zT~B*v!tfL%ABcbAcg!D<bZy^~>A#263tIV!%bKqClCJk|mcLl`t6ur6YqI*a%BQuu
z?wad{Zcy^@k#joVzR?fYDS;ZFknx+kitHz>dwa%@a~S@`I%Xkp7sPMt3UVJq|A`Nv
z#yd;a`IA+q-N0ZV3j>c_@W!wD9=UO;zGA;AQ}^Ef>!zg@1+$D<<Ov3Xf%?h7i?_Y{
z&1VzVDlIRppR9$|!9Xw&3{;<i?wgK$?UZ((B=r@$y?4dt(`Pp-1y+WK>U&(Y5De5l
z1LeMdTIupd8F=%)-QV71=~qd8!J_?)aw|6jM|SOU!p6I`C<RuRhn3sQqW8f-Fc1s`
z12AAczYhI*m$aYIuoOJhZyxaTz+fO42nK?IV4yxRaL#w99npGHvr=F-9;~r#^{cZV
z+~V^^$?`(WuFwYq!9Xw&3<LwgKrj#t1OqD)1Lgib@#S6qiNC&Z_T6V~P^z!kzeB0T
zdnd17s;gMve7)1=UtIa;sf|l3D;6EEtiSK8tNE<oF|MfkqT?Lpf`MQl7|6=N-VGm`
zd41zmOZ62$8q##Fr`o)j)K|Rt)`xrdK6ay0;EK&d*7>-y=HvT6f&cS+NWn+&@!!9X
zwX*v|wd3PG<JRvv<o-sbz+ya9dwin)V4xBVSl^3V(e!XN_Rcl7JZ|t>rMiaL7g)C$
z7<>Qk=fD2>hNZyD@KCMM=KHtx6{og2sMC|XHZCo1)bV+Gk6nj1ynRVhS5WQqWWF02
z*yOr5doM_qHxJ%;Y5G?U=Xbo}>y1h)D=Hr=n*8o+Ll>`BTBB57aroM8hW+}yW~KUy
zwf%Q=SM>H)wD}qR2?ka~2I}iM0lWG88TA$Q?-0~CxxT-DRaf74^X#(z9sz%^fxqiv
zq26;^^%d3Y^ArAVjv{|&Aun3|P>*Fde{Wk+|Lzd7+N-=FlixypkD75}O~Gytt$tbK
zTxmUjN5jHOw^!-<EW6QLe-|XHp1)gXq1PY&UL}a1C|dS(x$eH+JKr@sEsECPjjWX3
z>%V1prP`++%MNSY&|CIp)l*Ma)W27%vR4m&w4$N<KXDY)`C^TCR=vhut)IV31+p$!
zNPSj1dTl4GoV8z8KJ2p)`znQu+w#J!`);eh{Jld9b-%&icU0u>Rw>TD_McmScH1kd
zg|w^ViJtWhM6alJB5S{uucy{Ov+0dnzmQsJ`H|LxZ29?~7j^!5@0}(l+E0D|_lf<-
zbnE!SJt_F$@iXq&z42=a$lnVEIZyI;04s%A{TZH%qPK6%o34Lf4w-rwr>w}|@ym<W
zK0mA8dd|bRFdx9%w&>d6synBppw3TE=I?NW)QhZ0x`hjW_hPg4M@>#aEw6tkob?CH
zo1ed{3-Wg`!HefTxA?3#CZr(kfI4q=o~qnx%!um`S^7Z=YP!xtcmS~jR36C(DjypE
zkUc;B$=$sMtubO;0@8m)s~(M~$U`1_viD(gMkb)nFa3MJw?B6JWdpwZI0dzRmFXXG
zhPbTbX34|>Me4DTxB=2%=BHIIt6q6hd7}+q-Ef_j^Aiibzh8}=An~KwN#Bh6YWO-O
zMfe5P53U(;{VNS7f0SCN{OdUCJkk0!UFB0AS~Ppd?Ixw5?x$7e@84Ru!@Nzl|8@Hr
zDM)#pZ&_uX$J7hzc}DYFcB1F+f`S@H)y~yA-!yjMyeTP2|14xaf?A)Bho;j{P}8-%
z`iK4<boD3VEl6Ad`Mc{NdQj_A{?%T}fts%UVcvk)QG+rbtKYL_*K5bzmw=2D*nHm}
zKWg%Dnt<9q@fOrLtLu>RJ7Dii$6j&&J1MyL#Rv7e>(lWG_|pLojhww?gA(}IphmY|
ze&lxv*zU@|pSb5)YnMR%`xnZucka)dPCSEeFl+oJo-1k`Ryk|^>L2)#h3cQ$E`KK*
zWL;ui)Ao>=@7hmf?w35&^BuBve~g}SvXFQNzWbMFTTW{-I|V(zXuR3ZbN)Pkc52~Q
zolZJ!jiz%_aKc5`eRARUb5gMN+*NvQI(l|$;mLjO*k`Nmvr=&9PnPW5^sZSc*l)+5
zUwJ_H+38=tYu;o;`(<(Moe!V(*_|^}Q2S%Gi=KE2(jKVyYbxuwvYvn%H)sdcxWhPs
z<O6k`!~e3Pwuc{sv;%5>m9<^USxEc^b)Ds00m37QonYrTTeZHe@x&C=bx&o=6-AA^
zDq}aO=^BUB4o%nc%D<*_?g5owZI5#usC=nC^b7RzV+ZGR?_S)B-=3a$tEl`~a@KW+
zbs;ZupHwZ>=aR^}&gpn6Uz(1bx4%|?<(YFhh`)e3Uh2P^uKq~9AbtyKdHM_1>@PRj
zykV=M@1@G8%KAPy_miNtPNVnEoun%dSKt0f@6yi~Cd?D;A|LUS_<>&eLQg*SwdmoI
zxR2gL?D3FuEBsrCAHDn4q+@n{|C<C{xYt4ZT{&=M3Ua>-a=#0zJ=zc5ckBGs^E$kN
z>R;$B)ceHNU)?t0-Zl#oQ0>8Qj_<X2!M6h!nfDpiJ;zlKU$f~3Ppnd6p9@;^7oHX8
zj5zqPt?r+jP~O6<?`3hX;^7OgAHVu#6TeJAwHtp@<Q&F$5-&lGi=M3QW|h_7%9XA8
zQf|G>H$~!&SHH!dWv`VVz0O;e)y}Lk^?S&9*utHD@z28^opDoQq1vmm=Rf$Vg;syF
z&OfVO)iZw-Ilo${^{A}<S3mLkVfmx-Li~h(Eyp?mvaWy{-?hBPb7adeG)}3^e6vt_
zQd!p><wM6y&#%bvrS+@5JpU!XBKK>crfYrUyPvw|#an%lfcpIVppmOjKA`)%iK5nv
zyxCs2-ni=Be@#K1_u9VpSNBiU3v%BJ(hi6}fZP{?Jg)}nH>mcg|6vEnz8BQ~F}|S3
zhqkADsGq98Y5Uq9_i>=Mhkt?0=h{MzQ!48|Nc*R<zCR6LjI-v~`HF1ypM85?)OJ<I
zP8}a)Z=9)5{Ri1ghaV7okjq8dr(VymH6Q%|;l)DQ%Zl2+tmVjuT_EG7>l-rXLQwlb
zJ)owmOueAjAM)vXhOFgP=Kc}HPLO_qYOl)7I}5G(NL&WFC$o@!7W*IkOp$R_w9Zo+
zr{EzgVt-cT-kf_!5Wc}K`z-nPt=1nWAn#Xz+CP=Gzd8;&j+UG?UaFr}N@m<G<h~5F
z?A3jCUYUC)Q2nk_vNaCW&$wEAV4sD@^%-~L@a7Gcx%bd{pfc?-|15-8?DUX%L;cz=
z>57_9?`62>z)qfzVUM0)*?*89d$`{LX@~ue&O7W<Ju>ZsjH@E)Ank$VM+RBHD5q%o
z1^UxkU3blOLpLZ{2oDz0PYdCvC~ALHX8bJF-(#t)ex@?_Z5Hm(Wt+9HK7X~6h4@vq
zvHPSGKO570ZUS;n@Nj(NJ%;Uk^8BPITH}s?DlWQZ;eIbaKP<_MI_~Al#4SbUS+&se
zqr7@u-|LF(@kZRqin>268b`cx*8Eie@MQIOOJ?3!>6YHIGxU$0GHCpq1Rkn2AD&tF
z)}d$L(xk+>4CMVgQ1^AZKjxkbtT_ibzhj(n_7!*cm_MuLy{2b}PrK>Z?PsTFpLkA#
z7yHgmd2fev#}W6ue%`h1XQtSRo_xG#gWf{&S;)L+{bXOlxr%-ISv~uIyV3SDQ`+JE
zERgpmXa_raPY6BvIS11}+F}0CK6Y_#MeoIhN~P=gV=t)wqVji}^`CrhkFS%WNdNO9
z&rL1ld{J%8x?U{reAe^lta@)>Q`8?>Z!FYxjro%oi354jva7cA#H&i9@|soV-c|dr
z^A~%0&T;LcIlpfF@Tm0d^G@4u@YSPJo^SA+B`+@B;k--QPiR=uc9Au%Av6BOXA7x^
z=SJM?a8Jhjj=YaU{q&dolqYVI&hsY9Wkq<`{Hz-w{F5I&<w#fiH16|$4gIDc@JPKN
z<w5wh_C0y$G2`x`?$43oqiDRc%2B_f##h?6kbPoN)Oz8)(((J<2M@jcth47Q7w+3}
z*x`50N%1G%6W@CL?1Sz)dt`dSdBb`RZagX7d(okHH{Nby`r7op8XWWd#I)-(-)*;d
zx5+iXce>)Jd)gkh-Q;waHM)G#vfHF|LB9!&&+aoZ9ePg3%g)$jT*~`%#2?-ZykgTG
zPkHm(5oyb|mmhO|pGoOEkAK*#-QXGN{sV3w_13`m({s;maO!WrJty7ureB`2?Hl82
zekV!%x%`z+dwzY}yrkm;`)~7ln=g~0PapNh?*}eOHoSY}O?%G$BDu8v($BUz`-^1h
zoon5-&*={(*B?3HH|++$ll1!2)V+V&_4f1+cYK@l-|^w(zOf&D`emE<(~U+wJL;RO
z$E5uJnfDCZy}$0DBS%e5FTUvgPRE}4R=VV-rMF*q;dAMcgFo+d^`Y;k=Up)PrnQ#*
zD`otK{{7`+=Cz+w^ZRZ3$$Jk6-TCzOo3<R0KHLACS4J)!livH}CHu|pF+M%?@C`rf
z^4)}#_(H$=T^aN8`LFhGciOy3Df|1mz3x3`NVn<fmOs9@<)sHsOu47rs`SVgeLKCG
zj=p!q`)&G8PuE)bansX!e3-uRLCZD2Z9FCY!B*W*+I7^Fn)Vnc?BC@#Z%iHf)OiW<
z;^=Ni4H`9PRPy83pSpX<R-=>7dp8{M*E1$3+n)I4;t2-~Np`<*r{@m2@RQ`Mspk*e
zw%_bzcDL`=8!+(88hx8HTK;)>v*}5*H8&eR<@gcFlg+z5HKg0yHRJQuH_uGF=hlyt
z7B}yB*n(GPB_I5F^!X25KD`D9+_LxI?^rY=IsAiGciuAU?V9?(Sa<T16E2&W@cS$L
z?v-1wJ%5WHqiX!}{Xg_syw=k9Qjqy&A>&CLEEkEFS`X`$g~a_zqt%|L#~&=z{j!zM
z)5BM#u(rN~&F{{#qIJ(~^_z9PC|dsR>AmsFn$CKo$T_=Qw0P70vCm_F#65r__XQyL
z1jwM4Q<?h-?jtA%65m0M6O1?a7Vt!V?m@I&;=1aUk4ni!_di9;@!YMhBI}4X&y{y;
zpK9qfUh6uc^HS51@h=d&LA8@|dC?kY{CRnycP`QAe4cELcct`JoU4?c`!~klL-h}p
zE&p9!di=b$Q2DTAkLTsZm)EYo58<6#i~1?N={_^7tag?wXWi$KuiSOYil3IA=Svpq
z`cZAU=(v03*as=*U7x)Co*jA9dCw>-a$i>zz4)T(%(qIT?n^A0{jP^<H!|Z%eBiwZ
z?qfjmE9&=3ROVjGL*^Ct_11h<J@#9u_G-SbuDo=IZ|AO8x^|Zn+ubs1db;G)+gm=p
z-L$mt5&e39|NOL???rH~;rZ@HO}8C%+2C0<-;db$m5~b{96USa`wlm>+PA}=+s~@`
ze#F@89~(Pn)XbFU$Bo+k_LGC#Pfw|j^9Sc3+Q%-QL!a@-XBPf$)U=v*kiR|b53RcP
zosoXO-_fnV9yKfF_xXG;W6q_o9r?p1Q`6>~AN=5?ZZm4`6S?Q)UJT^BbhNw2<YseD
zYVsd@u?ys0jQ+eb^7~GM?wp?X*|_<@^RE6NWqvTeJQoJN`KtR6&!6(9W1ph8uIEix
z|5aJr_hc(Bs2)3+XNunb-b>f}imY<p{e||cT5{HPO8pE!<$eOWQt0g)tCi0C2$kvw
z^;q_LeqZf$-6v(0d4IXmNZhO~%<BJEK7Lo17j-`6mGP^hsLwg`%9cOp)$2YbuS~qJ
zG?qK>+AbWwTE}ggm$IVn-#pnmCsb>lSl;z|d}z9kpLgEY{fWwX{kq!ui~76rKt0-D
zWbFrfMK4}!{M7NvDyuyz(;r3cw<l+fhuV(H+HaNBewEdai^^)3%0=t-%JJOELdsk5
z%+q`K?^e3^9N7B1q3ZXR?D?0b4?Xje;eY!_gHpL@-S6eq=e6UP8~^a-*4M3GDvH*4
zcyXmt>AjvfX-4lxYn43Yc?gLAW<{Ofx_{IA0ZrHQtnQbx#?f-~@x3f?YTt(kymRM_
zl=~p=f4C3gzUAB-zu0$ryTLWzJALeiV;Wv^&4l#Q(rJIZugT2x+DjX~_RS%$r`)%k
zaKlS&FKIF(rGAip^FG>)-CMjhedw5)_n%(sxyxG{{q>`o_n>%BX}y2GciQLwXjtO+
zYrOZhb&C^PZGGqDn(v2mzs37eycah5<*yDsb?7UJhu*nJ{Qz0V8`&E#Z+~XpKm6ph
z??+v9->f97|LVF?Et&b{)vs|BS^MMN>r^Ve-;}9)Z~t}EQc=YI<%MJK|NZ>eKi{zA
zVYTej@BSB+b^MFU?AN^crRhc6^X7TpbhQiFtA}{xp*KIRbk=Rpj=bsWm#mkT9Yyu{
zMWxZ&Pg;6-%!|~AKjcNvzw_Ederuh|t7rYiPS#P-^DFX!R({SW>~D%9erh3pX89}6
zm$G8f^^Wi6*A~)m4?W(!@_EyX@>f^O<?TQD`*$d{c<<!(OXZ>$f0)Nbk?$jB#jNdc
z4ykR_bxvg~Uae?);&W|dR{tO$&lMHD@nD|iMc#9veMM^>sg$1cTeVT|jo3$kUVqr%
zl3$T@u=}PXUpuAUCrMVc;=Q(mT<LX{eO_Mlcq3i?%9Hc@vF2wSilP@Mh~wV<zqL<8
zukje!oA0E1cG7>3@1p6Inzvb>e`-D*_q;MZmWw@~AMxz8eK#!SMdn9QBz~8RFST2G
z+rp_WN_o-q|E%d*-zTy174_e&<%*6^T`6bH3(r1lUU+e*w$eTOz4W%P{o>*A0~aRz
z&YJHa@V?sJt$V#RuJPoW=M4P*kH5Rc^Ge=V<vS7lE{XR>$q&MhKF9H7)*aCEU)C8#
z)&&dooJD(C(b`8?{)E0#>q*{vyzd$@ULMkq+QuD!IOvou*KAU<ko%{snDuvO>_-(X
zzf!+w^sj9%Y;;YFlA^_*H=i}V+;y>NdG>L&jb1#+o1VA-8fR3_+rF1y{mGJb-<Vg<
zI&bpk<GZ$er-b!{^+VScWc_^$dVU|y?=$%Q1nJ0%9?y{;3<LwgKrj#t1Ovf9Fi;)?
zS<hE_^JP7sTlYbh-aAk0xhtz|ofEU_IiGkuWxa=_J`25mS^Qag%Wge4A$#p>x;{tr
zWb65S*730N>HR=f*&8>#r$V;u@buQ-(a5VWFFGv$()v3?@Au2zy_lvmPFc~q4?@5E
z84C-7ft8N|UDt>k++UHd^S`L9&;PU@{@yF7>3qKh)cSb-1M)lv)cnZwn|PEJ^?8q$
zQ~NdDi=X;Ftlp2ZuK=xdtp`uX9_|yle?$iL{*mWvd69A8d&ZP!9Qgh*`AFycFUZ`p
zQjWh*iA;X{m~!aJ2Qp6NQzRcqenl;Z4A0H??eU`~52uOR!FQlR>{ir!W5%6x1MR7u
zq=Sq<<<*a?E$e%{@B<QeK+RvStnX{F@1{MF`NTax^A+9{c^{p87LreqaS03tf`MSb
zVu0_Qc}N@sHD0Mq90YZJD_16-7Df7Dq4vk(FZAUxF!0VEo9{brV#0c@sN;al{)6wa
zm+MdLb3m<6?aC`-Z&6fxR4#86qBIPg|BGM0_GtTADevpC59WP6^z5I}gXHJD!#Rlj
zAo<ysa}FULWPeY+_$TKm^qgDJgV;g+An71_>_x^-51EIG`rT8NbsiNxFJ%24Bl~^Y
zCH`~I&w52Z_TA)1k8GjN&!RGZ85j%%1HnKr5De5u23Ti{qOR|(!xk1j7gCOUJ<c2K
z3lzPv(C6FiquBR>?BhUfN9C+MTE7ovUt^)Rr?OQJJ-mU|KB-cAeV>8-9BA2veWY8c
zcEZ2D=Z#(wzXFv{m9@RRvbIloea}SwQ_CUa=e++zeIVyr_V4U>@oz=u0p}0SJ1V0G
zN!R*RCLez%6g%k;h#q@E`a?Y+=Pyv(rCe6j^2n5@9Sik&Tvi!Bq&=-49&|j>D{6V_
z^^kE;r2kn_=Pm6rZuk#~zETLE;Oh5m+4b5n_a#}8`K4&x%Q7yy{xL49$ImTP{xu$H
zy7FzwS`Xs_s+}tHzLDy!erP=NWX4NT{SLk@RDO`P9%RxjRNmA+l{G*7C~wGCKhTrU
zvQPE&D=TXGO3BuIsV{nK9w|S%@6i1*?+38IQN%w$Ew8d=mo?t_r#_dqWd1Ha$iA&8
zT6Uq=drH3V3StMS{8_T@ud>QkeMR+*tHy<_GI4-)o%d{5UqE=|+`{`d<R{L<JIMRI
zMf>N)b@dzk5!7`y<nkD>;zQ_<y}jxDlat_K<>sNh+4Y~&Pw%no@P@Z9Nh*c^sri5Z
z-{iCZ@m0(F8<eVryIio>RolNZC8-wvvgi9pyf%K7Qnm2dRfZk<<5ioMs)hQ!x=PEB
z?L7ARZRV_1s;`*$`we;XpVhj}<4cD$E>#N`+_YQQFHT*xR4sh-@eR-X^jE8ts)fCV
zw0)|_E}tjW!oRHk_dj$QI6tWtp1O34?R(FipOlM4PyKzE?r(Mf&HV=ZWZw5Us`JhR
z2lg17vd>mzzpco=8)Uz&SgHL8`)vBpzCQ54vj(rTB&lyaFpgPK{kdFO?>VY1FMNCY
zktYuLvgV$a_meo+a-L;=^4tPh?~_#~KS;eimt+3WPv$A-cji4Z_Hj>vo^$O%Cv3jQ
z?rVLN&<^i~@p};T*bmZA>g{mu(UaCYa!gJ9A)g}cGVa(zeb`SqeXoo-#d~Vh!*`HK
z*L!;9nfl2`T!eQJ`#}9WoYVu-56XkIqewbPf2mLP+7HsH*Xjr9)Q=46dwk>r$%hPL
z5AQi*H|f|1Vjt=F7s&XNA4ISGQV&S`AnhT8^oRP9LG%_{^-{kg_It=Y0`WW0Dv#fT
z<cC-6fj^M>j!e6X+Aj58e9`-zjy?0O6eJ$dKJkbA)O%X1>#n(O=mw=~;gpl7cInV|
zaZ)XO<>M~5{-kfCQnipcUoFgguF0F9@h^%zPpWIkJn^v0b@%n&`L5Y1NPM-Be2N->
zn8(B)klztAPtlW~xCT-$`N;=jC;8DMQy!W8tOuG-K1J#Uv0IUJo-dG2+^1gRnf8nP
z*hRV`@d+e;DYE}!pF}<QJNvQ&TMn8ye*Y&@_6?LH9&+BJ9_+`z$PdyEGVM_x>B!WB
z4AMV%tMAD9w_12=y5(y->@qB^7S8(FUc;O3KPssf4!C9Szu&QFMp9|~`m@8I>Di-s
zsoKc<mDNJdb=5+C$6YPF><6b#z46S|OZ62m8gupLXTLHnsTR&!YuMten@vrsg?nDq
z`RBcNnwV4z+aJHf+RYA`l$49y(=d;+#(maJ^0S^IgS`I$@;zZi)@?=h4<P#?+C#7Q
zus=YL>>>HNN2Px57u0Uj6<N2j7o;Be0cjUGFTw}qkh3D^w4%7vFaCMhqcd(wilV-E
zioaw<_Dxw4zX!E_PiCJ5W^I>z`aO5<trV%>L-zaskG;1E)ppsogZ8#zG++++L$ISA
zIV1>jO-$T3fA{9*287Ss5wvL_T+IesqP9$IB{6oib57$l(iKf&!&b24X~ecD1`$-!
zup!qg*b<Z=ShK2T)%+XvRK4%>ynn0K&U$OkF~=Bl)_mTzzO}!7a@V_^y#IZE%?<Fs
z7on5?{blw<KC}32AH?H7Uu-}7^=s=#|2>WW{SKWx55J3e4nA!^=JRR$+wUW8{y+2I
z{Eh$qzy3SF<o}%~|KG#ppJ)HmfA1grcmB{n{yV?M|DMcq^WO_{&j0&%{`XJYm+y1y
zY41Io+c|E{XZu+m@8{Oj{yB1M&i{RR^|bG&Tl3Yv=zU^+>|XhO%%^=G_WASwzLihQ
z=RfE9wDnta`*XulZvXegJkRQB|M%iZZ~y<%t>$O_{Wf~P+y8xq&F%L&&iQ{Y%>UgL
zo&0^!o@gF~qxbPFZr`uZ&UufsCx2b7p7!hdtQ>uvq37`%f9=2c*MIfj`n|u7&VT#A
z{B!^NU;I7)@UPFF|J48UC%^lxf9WT`j!ye^&cBBao&4uM|2;FG_W9WF+xGjY&H4Y|
z2s-)a13vlBeLngBo+_XJ^Kbw8f9OB>Xa4T5p*<hrcK%1@w|;+rRo&5^QT=cIum9)&
z@NfS1@BZ@M|Nm3$&+kwExt9Mvh5vgK=X~<tf1tl#v43yFIsZKm`|#i2@cF;|rT_Tv
z{rmsy@BR{<Jny4U{_kM+<UR4<3-KNp=bTUezQEsid4H^D|EB-j|JlF#zx|W{<1hJr
z%X-e4&nN5nWSmdN?TNmRbH95}+wWO9zrU|KAN79ZKIpk2oZkohep@~HeaG)xd-B&a
zpTF|={@-8u$zT7SU-HRc-~ab-{kebRAN<Mh{*u2B{QjT(!N2*J|Eu5m<zM~he)F&W
z+28&@8qRh8{=@ly{(t==|Lx!TeZTWdKDo{(-^=EV^T~BSEzb9tJz0-Ve&6$Ru)h3#
z0-g5jkp1}N?<;&-{=GT(VIQ88alRM-Jk0++ocWyde-~%I{lDXGf9_%ZKm9-bcmK)1
z{Qv#k_Wh85&o9^Qz1g4tcUoNM`S|4bHUIkp=X~0qYnW$$ZsmWE;gfl2KAwxeAMxLd
zb6-AL&nMptoqR7o*&m(u-=}PjbUYtF|65P=@4JliKGA9CMLGZPVfnN_*Yo=loy_O&
zzv$%ezr6q1)9%my_d3q)_ZR*?VsVr+pTF<%*RkdEbLZ#GUzhy#VC%d$`#$9yo#=i1
z^S}PX_j$)Z`*(l#=YRa^w_o}8&wu;=Km6Ih^0ROM0KNYBlOO-lAODGO-~E-J{rLMS
z{{!E?|MvZF{r$i2SAX{7zxE&f#^3)7-+lM}%*TK=D1HC^usGBBVvd`y0uHZpAberW
z2@6kghyxZT*wlj!z~!O>!r}WL3JxZ_yKEcVe7?aA${Qr}^K2&mcDomU;=AM1m&y0z
z^Q6_CY<JRBz~+g=KCyMYj9UjHo&2mRf0xR()w>?OzXMtcR~ls)NWF}QqK6gTL=up3
zp2z{$&iUYW@G@!+l|YFoEDMBoDm<;hz-UmKn(-XcXo^woj_y;oQ=~hFxKkRB7o*e)
zRU|HBH6AQX*($5E6Fi#kG|PiG!Uf%Ekii*XVPN)grO72s(q3Xc*Uh@>e8i9H9qW-_
zYp&5tBekGHJr^WzBoL_(@N&5B0~7#W9ta7p_UDHq`^Xs0bukV{Km~-mIU!LL;!s-&
zjxHQzNM~@9X442e=r&Wn$abdn>u?U25EEi_^;(B6n1CxQLoB*1()}dHz@(o4cF4hv
z4$9$mZ_$KC2!stl0l?lK5CP18A~_$#Z$v{dDWWyJ1zuPXgg0ReG*f0WZghC%&fqB4
zb5)!}qAz+xcwG`scVr@;eHxF{nlE^y+vsSXsdaPNv3tyj(b*rH<!q5tge47tA>P{{
zWoNu?w0bqbkKhP<)SkFEpSo|1CJrb7IGc`Yy6fH9X|0OU*e<Mw;NAB+$>0IQ-PPhQ
z@zRbS*J)|2w;;dl?p}5<daQ0Z2fW3@uR}h0_$Rl2@RLHfox|r?$5%o=$4bROHV9YZ
zM<y>f(B$Rcb#DJ1a`M7w((v+N1s33ipaK@|x#Fa0KvR%P?wQAEPxf86TUVGHtfwn2
zHCJccIYxt@Ian|}ksFCdHEBA_lbT1wC(m`BUFZ~o_$g?*&K|McyLwzY?u5CcIKW-+
znBgQh0@rs!9T1)_f=UR7f*Y4|ri^A@o*M`p>;Rq|7MSpelZW8*5j}6}fd|Bun;IR7
zM(toNEQDatewG(VKMndat2pDLXLvmM4fFw5c<Kg^_IFgz!%vZpuQ{+8l$tN%EZ`wE
z1XI_Bx1f#k`~c4+D?RQJ&t4FLi;y{D2@ncpoh>546k#Ih%%;Mjg+|eEA0Q^ey+cM8
z>A+<MjyuJu@N0w?h$4juH?)?X%tv+r58)7O1Lt6Zq6ILEojmFhu<C$_MqZ#Iiy>Yw
zz0hWXTFuiDi5%vDu0J7)-DA{EjA#fi^P(!i0UR!`E(A$V*w>8~!9)QR?WfUNU`DtS
zN8V<Hny#^(M=X!gBDOV~yB6mHpj`k5P#$F!=lO5d#ouz+^^{YMrjl+C;Hl}X*5=&h
zYMlG@({*$}=k_R*FcVk0#mhu*HSB}7W_Mz9HAY;FQ4*rcDz1iDNl>HnCIcM+yfVAs
zdN+><dJE@RjV^;bq|ye&$Q|MEc~yU;12}+T7sLm*lki9y)*4uXmP%8KUBLlXac@3s
z5EKI_kAQG0Mzsh#G|Ghiva|uX=p~o<AxaC|$L5(Eg2H7`M<s!aa4@f8iQXt&<s%=$
z&-5Dgc~A{q2jPA!!4#q5t>Q|&n4n(SCKZQu-3w9YaKzF*WVWu5>d9C&QOTsirl}1)
ze1)Q+2-Qp{CDLBTci5j~BPY1PQrfS@W*C!pFIec5tD-AfU*S3URz2TthN(F^YkyoY
z%x8N#4k#4>H@~Jw0W!c2P@!8)(zNIkvZaeJk&m}?vfWX`$##>pqi(0v=VqB>oiY&q
zLc7~O-{O&1xqISM4ZiNpPg)~b;2NC&d<7CY=LAL4pBS%By2^)FyUqwXBixU8!~}H!
zdw)h^L5wC1EU@1PoTuIKG^1##mzqZnACKuP=B&Tex1#2RGfq%QH%SXyyefF#otoMQ
zuX^Ff%+>aR+tptg^oPFv;r%Ov@Rm%)?X?09uR4Sq+fh1U;VBMrz^LV}t!)4<7Znf=
zzr{nr+Mw%f+t}vwjyEW8kj&4snfND)z4#N~9iP5Tz8{|_t?p#IlcoYTPaO7%t>a~!
zF%a@t#qqbecy^D<jxzR4h}<ZHext&$*RVKpy{-1(c5p~Fhf1JC6qW@-I~AVRU|=*T
zP0e@?X*9*Cc1QOq+bPl=L)<Bi$BR*Fg(?!4u^JB+rfikf*$Ey^cbetF8{vX(G|1o#
zurM(DxYFbjCTTCRp6g~^bw1)p^^W!E-nGs-&9RPWS@!`703S~<$++5|zj?_%GM*gs
zY#JN^7G@`a2BHv$x`W{8!a;^~1~+Lojj)4mGv$kHXIj4w=Wq!zAx2lPb?AZ#xUw?D
zqRS%PPht#A>iKVn9Ng%j9A5VpO=yHb*Z>p&?Ck*&!2BnY^FjPZGz60(TEko5g#|%)
z6ShDzWhUcBhga^5<tT1cZG6!q%N-X5(;b<}XYb>Y6Yd91d*3^vWA~U5qxj$6@raRA
zge3<6L%cUd%FcM(X!UA<58$V*h3DZ@4-%t^0}24nrlXqfdUtkOt70^^3#%b`_q|Rs
zc))OXwYW>Xw4=v$S{myu$S=FQmmQ2As~gS%Z}IT!kWU`|$?YHfq|j~W@Hy7;m5|S|
z@;w<}?Y>WcO0TeRJOix2g8!CZR}2-faL==uG!1A9Qpr8@812cv>vro3bA$DCrKRTT
zj626@@G}Psh9`0((WoX(XL(Zdi1_5W&a(@hLJ&U%P1o5YmU~x^OUIoscN7QQb-nw<
z*Uy4FAUxd%l@JaE_wD8sHO;&{HxM}30X#A+FyRp=55eandfwCn4~Q!_H98WF+QC{_
z2*IHJEH99L8uVpWamGc@@ObhY=mW0s)D0f(@2H-KpCTP!b6_(l9sABRmfAFO3+5ZW
z0n^>e0pJpbXHp;{STIGX3x-E&^r7_;gE7QJIP0Ev#)HM^%8;uLC-adVz(Y6$^$sQ|
zS^%@y$)g^TZ**ZF5OX!a0|-%X!s}EEZ5F82eD*-jv-&YFc8}5IqzgLhjSyZKs0wfZ
zn^AgmH>D~)DG4qSKxH^xXju!)6|TgQw;7?PYi#Ec%VV^NZO!Jc#ks(IJH%YYdH!2<
z_6wN(PCV6UD(OuCJT;xw+Prs(Bg98|>@tN;S!NPu;>y$rk0rg;un$^|-HFZB7;!O1
zNr)<|xEf+5L5<Fv40HhS%It#c-8>@bJN$$Vlt&C7v@1B^>?I5v1jPW#BO#oMQ7ys_
zjWS`sENuWTddVezh|<FLv3cf(pl}(~QE3$q-;H7f7orA^I02mLHPP;B3c3!${XBvx
zLd9Ffm3T2hy|PUz4(qxXqR!!nrF+P1T_M$zv1+1{NrO#O8+iB%MMDv)nNCWiy^Qa$
zKgmW;aDk<?UyIE!ChcCZ&?#3%SG2yubMUQtzTFH{b9C1JxL}yi_H-OjDgbVNO^*U(
zfE}Phx0s}9(I;fjy!eBAPqv$+J>=of%`(S2XW;94W#{Yc&$<3}E?@WNC+!q0a1G9X
zz5<DybAlr2Zxo+)v+InIGs6A$kC>niVDHaJEQryhfd%&afb+CFo@Nv+^-}Yw;o~uV
z#hmq*`c~AOaK;G==_YAmi&q8jyHiv9;8idDn7P_MaJ%{|gZ}Whe_{X1AiO10aeJ+R
z!>bPA#&(oWSa^y<958CRYik>T%S8o*!*B6Wur}yA+cviOyyFeZ8zl4dY$pDRVlV#0
zcgLqMlkdmpNvk{A?xd-J%@c=xV(WMrXAFcqR&o3-E}q?^vZIVW6CyXtpx>x4>@_To
zTyLvAxE&l)&7l$~5rt)e&`yP?H5eETN>ekQLmEvns@>6j%65u$#}Ic)<MCpYTA_->
zWvs@7g(+KQb#{VB)178{@J6_x8x1lz11t>8KCU#mgh|><tmnE}SDlaeQN3e5x_7N}
zPIIi|S=N1k0>H-;Ofs(a=WkxJkBleBJevkbfQ8u!pn)jFq3$3!x^R#ooxx3-O(X1}
z+f4Z)+nLs{!#P|+Oo-9dYaO~^0<NqKvFNf$_mda{lY0K!AqO`)D2La*MH3n!5H<h>
z0DF5t1Tg=J<a`jn5e>nlh}Q5Hcws>#?vW->fn@4TS6CCHT)pWWTKb|#ZtA)qn(oL%
zK6?+3)S54Nq}%9do~d<n*|B@fh*ACCx_tuR4q)^4j@lw+#C<bt^=g0*;HNEv=iyTi
z5~GO&3INWgqnhq|cXnE<Vl=i3t08#zy-qTCz;Ji9xJ$gWqsMhx8tW~{FT1;!9gH5U
z8_of5@$l=APagis?H~N4&~4}NIo9!&kk7I5JsDr^zE6Kjudr}D1FXP;|CV1@3>C0&
z&$F2{4QL8d$vyKJ?a99DcIygrgY|T!rRM64JI847GY1QXCvqdvs3uKkc~bL;_~g0H
zvkRR<5I+S?*V!YMdsmN3$DJ^D6bIaOz5B%1&w@H2JlzMC5Do?R?dB9U&AdD}5IEQY
zJTfdW;Snbf!RI4--qZsRh$}ZWIuecA!CF`d!Jz#tFOYs3^kr6Y#zoKYc=8+Q1FrDY
z4Ib_9sGf(PA{}3IU^6Hk`_40#+B9+t<{P~M)7{De;1Y&sQXnE&Fh!^fhDT}iq4g1i
zF~meT>z;PTgT?5|kgE<S^N}6ELpTKW4kjpC0JGT1qaKlObYUM5b2Y#N2vKjs>r@ME
z7O2&H_CU_F`Y|tdkJ03$3p(qK5MCIl3UC0MQF?PXr7Ars2`&*pWjI}ESqsb+uEdeI
z8KI_YZ08ZnW3-5E&E~Gfxxjoo#9YOB{#$kS3z+>*Jk@9_=}iDUHJ#PkymyHs#7B7S
z5~0q)0Z@gRxDuH?8GEZ?AG9mG6Pv3s;$n=F5LH%jHN;AS8l5*8=m6lA*#+0Tc|_1#
zILB&q8QdY2HXugs2#3$B`Xe2{0Svn!KDeEPN7As?z!J1nno{fv4zP-Q^I?Oa7(jUh
zgi|r9McAQHChV7`4ZuY&xx^1qTG&1|&)g6cE=P4i<#1WDDp=+7V9|M|SLFX-3|$A|
zemubxq2jILO1zk$UfCuUhjrZxQRi^P(miChu8``<ST#|}q`{`C4Lp2>qM-=YOeZDM
zUdDIWpJXE^xWH1{uf=8<lXfpy=#;CXD_URSIrvsR-)@GfIXY{9TrkXMdpZs%6#zHC
zrbhuXzz$HMTTIfl=o7MMUi_K8@w0oSspe*xW8oXGRIcllov*WBx#TmrU-#xGZ5u3b
z4bFeQ0*Rb+f+Fc}6rXmp>x_^y!u|G-n4k_|@6Sjqh|#2h1@`-Z^RzpjW)v;;QuC<c
z<1u~3ob{LbR@9tu#t91PCTU@dR|W68Q&aokRWJOQx!OK(JNlJDKmWxK-+xCRKmIiR
z=Ab{ae|HdGm#Me~SHR&_hj3#%N+&El#UTzDwcNF}4Z!810>a_<dMH>Mbe(M*+kD>f
z2IUQs`FS=I|7funf8x92)0fHj<MX7|oosi~RKVtm!#=Tfyo@sjLLRF){uUR{?oruM
z#-0h08)eXMR2cRe7DukP)gIgq4yoo)36zM!vOs93!qXZIj0UBt8P6e&rWn=k=ssmT
zMY>~%JEie>F-omaMdC76<H5p|t+F~h!K3L;vpjesT+odM8JqzY24){unq0yp?IqT8
z-K?w5NBpSXu^!#K);Xs+*6}RsK0pEB;|V4iSNrofFWE=NlVhGugCoGg>;%w26yi{K
z5FA}N$dJz9Ce5Z1cF=96e39);>(}8NE+Hnw=<2l&T`&PxR)$z~S)}_(jDbl#|Lu^2
z8y%Fx>)xUXjSvVMfC7NMJs<*@|3q>=h~J2YU{XYDcniF+AQJaTlczv3b*3w<iBYcJ
zbPg?j(IYo?T@X!oWFnuvhevA77d+B!bTrS@y1DGwJ!ZtHesA4A0dNPfd3#4~kuu`G
z83wOb3;7Tv5z3naSW+Gu^E`a&<}sQ$pa9@(I;!cecW0-yDn?_wuo{AQ-|HlU2Ml*t
zi@U^2J9=EFrLo?E{Ia`y*}>?sy5St~77xD;`Q+iB-2TB&3f*=NpJN?g3Hcl=-;?px
z?)&to^a=~dGr$Tg_;2}j#ZUnY_dJ_P(}1QRmE1Fr(VpzPZnv&5H&{<sT57J&xO0pK
zKXb5Pcp^6vjcU?#mM1ljh)<sDJiE{-1o2bQbe%n7xp(!rbleGZM{&Si*Sk-A{Vb>h
z!qa_F3E@z1-)>G()6C0r1A&7bz$3!~6CQE$5PUwO=S@BEfVgr~qa)F%9jt|g5DePS
z@&f6nL0@JSXI%6Qk0-x@KHv&Z-Qdywj_P^%Dbn#Z2R4J!vF|)%sZArdV7}2CFx{;j
z04`y8CIupb1yh8&V0e^9A6g$V7(+~iv+ikUJXnmb47uuXG9TFiJcL6~?_h$W1u%=9
zJn9koMi=%0F;@dTfDrX2yiT>yW`SDGXAk5&s~_`X_ZUr1x}dY(2;qf+ssIPD8KpOO
zQ>xOFlHd{nREE=qmbJiK;Yu8Nn-OZd#&#aDJVuMy)@<%toD0miL(Em2=f72Fzku2A
z#8Zu?lHLTsQ`1?k&3l(PLVSeBE)nV+8~|0Ai7S!Wld-oN_CdR{JF&SMBQC}$2~lMg
zS3|5MsL^?oferv(nO$(bn@0q_g>$S%m%$xUX#-;9j&S(Asz1^J9Kf&(;)B~scq9#L
z4J<)Rr76X(-~g+*Hy<_#iUE{IKsXhnT7(@MWx{@0+5lYil1uy$rG@Qd^UMuF;WDVB
z(kdRl8^s7NL=7Bq0yxuaqTSULbRC5Ic?460inod@@nV8{Wt&tS)^#sLox>4J_mJ7T
zLaHZY)kGze2Aiff@bDFih9Xomos>v>8Q)=ll8v0;0!wMX7Mo#A+Pz?*Q?81xXnlp}
z;9K>4yBVhD=&b#5!7!ig={TTN0NnhV9tFq%J3xhQF-g;+PspBm@dx*wY&S`J$ittT
zWsY^uz}NN4&ez$WbN%aFzV6LW+9_Dz8l3-p1rj;u1Vz%{C_e3G*BK#ag!}CuF+m-`
z-k*_J5Ti*03+(p+=V^C5%_v&xrRGt?$7A}6IqNU=t*ANSj1v^nP13>^uL|CGr>6G7
zt6umqbG3cocJ#Z0(yt8qBj5h$CBHK0yY_aSm<3O{D|-Wb123|H{gr&qgE-hDl>r67
z=Zy7Gm(RJ$L)Ug^v%U2@7kGob?mQhiLXQC9+_BU>OCmk)5$hjy<{$h@`#MB4ON2dY
z!l=*MJeF<q+<`!`4D>BMnq~?Mn{rQb_^vT((2iEfjdDgW8uoc+(~BaeHYrap^r|Y2
z|4u=^OTAIOE4UPZ^>)P70Uy9-%f<(a1n=vGmu%${(<eQf<GUO)o-}&Uw$0jtfOG;h
zZEuJsJz<i03)9=A<ddKPo$v64na|qWv-;ljsF!{gC^wU@+!Njz7FhYUPq-df_~3n=
zbmx<7@S<%4n`7m>RVHpWLFu=)WSeJZ0mD*vs?i+F$wVI;YHt3wCJk}cN@4~M?RQni
zS=&xf0C1DtAEXWkO%BRwgk#42b83gSa}B>m_V{fL*YaZHlv!VLD74168Wvq*8ULEM
zUS5+D<ei0!3qhq_ORC=)wr6;(4S&qOW~SnLjd#CO<S#H0c1@0O#SovBUN>O-IqJF=
z=PZrpJ9S3-N$0zK{ZnnHZtFXGq!|rLjq|MFakHf-^|)rwTIVbFy>IKR<}=?=C9inr
zK2tbq1V-54vP4Ic{Z8-uKaS8AkCku4n7QN(C%Q7w0XTBl*!yf?Ki1Qq)gxEhd-X_N
z2*86sY?Qnut3FLw0wg!IiS#_DiDt``_U})#N_U+MICR!-^Y#Wlbp!lG!6)Cd<t-=`
zlcpcA9Rj}%hChv0d2?w&>t~RJ^AGR7!sFk5p5V#j#Y2b&B8<5qx}b7Y2KWpcd<AfF
z0fuJMv?!0f`1XzmK4$1EB0j>kx@7H3Mu}8%(XMq5okf#wY`lR7ZLqW%Hw&r(q^@&R
zS7e-=WP=-k$`krNp(VKdhTmxeU-y0Ub#MMo@6X4~8vNVSV;b)+djop|@4bP?{6X@V
zF7N#$UTpfedt~ij|Jx$P6+~G64nP4-F&8MUO6K&R$`VgtZ)NQvf2IG`)YN-Jkn+#B
zBDb3bB>-FXtLGnbjlAadyR0i-{dw;B_69zD1M-Vj{7lgPJD03qvWm+a+%)xFQ{5%T
z<0{+yLx@7Oa8T03hi0!i)k9Yxo9%VTZ9Fhmi&x15huE!NX#?q9F#5~+_kWYn1=GU&
zA-B)hvuERQp#dcTaQpGW&==C?+=42$`GV)Be$5R8Z<z!IC;&S()V$?V-{&SgDW7iL
zsDSWd;Hw_VOaQwv-VONFd(AutaCZO;$A}SOLULOenr4t*qN>DD45VsAsZ|A#Wd5oM
zczr%x76sP@3opBw(NE27Xh9NTE9YSH>W$U{ZDTTK+0)S8KjjPZO-fF$Ju0zfvSRU+
z4^mk6j*EWpLl?IVhi>({qLzp30holO?Rt*tU)6GIEE?PbCOCi#&pzv~4Em$r{>8Ol
z8Kf@MlOEB<bf&I4C=GSMb7&_T&+>#M2%Sp>9A32u*Vc)JiMY~CJjEdl@Y%E!grH)?
z$yhh8HpfXC@i7%u5MH{6RSC+t4eaLtv`P%B0W_FsbaUt=#=R%B{x*-iV@`|?o~hq6
zH3K%$92y%4ZOP_rgz!?gRhw5A08_P=LPiL7&CpzGWy}q$xF3qt;WnX!ats6j#y()8
zbssn35CHxFUL3!>^hi$=;vracgHk3O)hD_q3Uo0APWjaC10y6TxcqP9coy3Xd0L*#
zhniurxp<aBL?Xhb&H*f>YOYm+KKf%KJbl`7Mk`2tL`W%r063Lh#R2S5xl|V%5q6>^
zFv>bUn@w@ZuO;;Aif5}9O9eOKvym7;rzEzx1xY32z$BFs1L1*7g`9!_1?Hq8sz447
zD7GE$QLdP@u$7Nus6^44^3uTe=Vl5Q356nMx;YW!EF6xo+5j5)Ya(lN^}}@N<BJ~Y
z9Y77hq#KkDZ9C!EDee=agR!9C@FKA?Mn~6?3K^?qzycf-DG0ZptyK}l(ioKwJTZvW
zh%LzeIx{U>D)kbi*f`YGpjM=E$PEesq`YoTvM;@)PZ?*w3>3VE_129Uv}LAUP^UIV
zhh)2oBGyopilj?F7Yexul^EfQRM%oKbU9^NlaBR>?0QB~&gzRia^no-C`wsf2#I@0
zw~TIGBjb?Kj;Sao!Am2xpmy!Ko7Bxw0IBd4RQ0$=$jzgK6x7#pe*otiehE|oMEU~D
z_b|HxL2wCHglDypZZZxus0?rbdt*tIv0|cv#>>74<IiTIi4EpWXF!MwU5Nrn*fay*
zx|b|J<*J2Dd!a|i`hbC;p)C^4;A3N0g@u`(h6B6=Ttk5vi<-uS7>ymeQBkGW9H_s#
zAd)xr2)95LgtJbchz)QJ>x_&&qT5-*fFGAE$~m91U4)H?1PTeuIJ&T@%X!v%ccUH|
zH&i$aw@-`PTYsA~Z&uHI<b!b5**=JmaO9KlBJo0x6cdfhOvbIDotz4eSO5n{)KE_#
ze|Q&<FxQ=yN_oaw;3N^CY?M@0`X&s_T@Sr8!HXjuVe?EwRzj7CRtWEum=U9RTEu5X
zt<#Y{0UN-8F4;0(vnOTw6R|{jG1X4Hwv_=G^@a3iI1E_MXFygZm2`3h3rvJvb3)X8
zdz5WC<x}@!w49tJjtHd!z^W4^fX6VOjo1O%4cS<{sas$M(zqC~6nt4u5mrMmsjkAS
zFR(XoNew8k)Z)fzK`Ye&6ab7D(0r3ph($<@mbRejY9I?7)5)>sM`*)Zpfx2!MTEl<
zz+{+@?!r@w2}C$nDIyykoRKENiVd#J)rmzOz(>L|F6J!mnTTRKY6S6KD2C0p1-Yll
z3=iH77lfI(&@vy{kY}&aU#{e=LpkTPoZ__>1fuo1P)Tz90Me5*GGMB)8#xD15f@@q
zZbjRRb;;0>YTC^?swbyL24L7o^L5s(5#AX(iuneusX?U?Dj~dUhYVCKxPL8=(&OES
zBL3_{5TrTGs%Sn+%sf7YOAZk(8fXa=NSYvtkQgS-q6iC#$lw+{cXnOOC?-uBj}#+Y
zfjn5?XE@!08!t@NC2$O1@W_Nsp|F(CDl}<A3*vFTjG*2};i!`+;!kejS(U_ve6j~R
z^;k+8*);<DHJ}Dp<7(d0Y^9~<>Wr@vqjv4g4sl^2!Y)C}0^|bMBH#pz5&e)zsTv=$
z4IT3g@z|NHvYqy<Svu6uO_Z~sd5fUR1TtWUWcyex$>sz9aY@g3R%XWQsrk<(y<j`z
z3VOEko=)um+i&3NsIn6mj~g&E=oE+pC>tQWxl@2h2){)#@RI!{O~y0xB0j>oFV&AY
zsy`(_;9v)^Q&cb*rRlRCEF7K2pi@DC;d*k~bqCS1;J^=-aig%JL6M12Jwl1YA*djU
zGPX5kVu`t^K$R%1>L@EnauSYg=I22UE4b$BjJHT=tF%C_Y7k@oS_QIUK@8<M5#X?D
z;BW%cXFW2@aawaw0dU`QGbjP#7eTF-11H^6aXL&<E%i_UlIh?KI5bNv$~m9Xw>P$^
z98K3w@*x|x;!`(j<pKjz)5P@{ZK64AGl@~}^wl;HKJ)-IkV)xQ0km>iE!SD~v7pU+
zJ5*_S30gWcz9b)a8aPHq5q9@V`;;FO^*!|nUk-CnCB&ouWv*`;O5s;U%=WGud5;aC
z=Lv8C2XB|gYymIC0Y`<i1ki^Ec<pmA{-EUVv_$suyOCjkR7CjQ(>&7vT?*g;4!-!V
zy;Ps0N7r^-cXyLHAo?*SlLBV#A6H^DvM)A>KXFMXWq?oNp&^;cSO_SLOEJn)Q5B(f
zGX<7{erOIjVDrO_Jb@j<=F$huh9#&nkB3YGbwOu+W`rC0v0&9G<cFZWN}r_<>4hGt
zI#p1KFt$AdUH~bBloSdi!a|c;L$>sIUqw(`_d}zt;(C}+9fDOC)*4uXmg-)?sw~sy
zWH5`_W;u_<jTjUI=+Gz9L>M$eWQ30{5bui=hi-xtaTTT#A*2v|I7#KL;z=eXA~bz*
z#Sn7}*|lKnHXobRj-ja3P)Xn-R4X8ZTx%~a$av8;i~<00NuFgq^;*;lDhVv4a;_Z`
zswwf(C!!-9Dgr?O)#mL$6pQOzXUk?R05+-ID?Q38lX@&7a6h$gMN|pXLogK^-h!gg
zsm|C_$_19gWd#&i=b95=X;cD48J;GNOsJ-eMAE|^6)&4-Gz6W0F&Mo%ANV(sj#+IH
zu5W6xKGLMT3#1bin>`D47z!T_>ExI+G7dfFz*}q%G&h#JtVhF(YFfXp!c}6oCW<0t
zrvNG?Pyld$Qh)$p9>bl6AEpNEGZ;6^<FdyI@o43nf=Stw&ZxwgIytcl#?8rT+yOL3
z0})<6x~-jq8KR*e=c>)SL`9P{rP;~h2F7dUDxcypBIiz34D&zSRJE2^ZO$#&*dl-&
zR9MB~#gUO1jzUD!xXz@=RIp*MILX{dSXL^i?r;Y9EbV*NBR$x&a6LueBjI8u5t_=#
zfNNT<v&soU)~D0fZO&ALbDdAfg9SIj(28yNWO6$}0l;|yqgwN&01_3;BtQgU+(QXe
zq8g&g)i@g5_w`7<02R`wt2;pO)!kjfN_$!&*9|UU5hjw(x+=zFG_<KysDPc7TO^OL
zKDdm=OjKRP(Grmlx;f$_T;8<DfUV)4^{89q2o?(Oplr<UOieSBl6i&-E?<Xz1Dnh@
zxcCXpREF;8U4cd2y+?&D@*egmkz^ub5vsm1p3%6Z)GcY;7N|iQkAqIDhywr<@BpmW
z!IpF)!AUr>5hKvA4Ep&men|iS-qOd9KTW?m=!do69VEPPI2dG5!u!FU9$Rt4SHR(Q
zEQBviJz?P~4spQ31e<!W0k~XLK=|_ahl9cHFWX6NKHuR6<qeYgc{US&yWNXF@!j$1
z%jEm<dD7}mwmWGmVDrRbpV&HH#u)=4k5wFhi;HLXsO%_X&wL4FIO4E){`e1W2QQ%J
zPzjWX!m>bUr^3@342%Y)sTt29jiwmY?&v;cJ4L!<h&!e6cri+?P(|W0R^!3Kl&!Kl
zJHeyrPP06CBV5pp1{s_I76xV?SDIYHB<&^EbKR_~&PV*H-mxCtyVg0UIo9zk>pnmM
z;NuA<8CU!BH!s;o#*<^7O@kxA!t4alKosIocMu$1ILMIB;3mzc5q8jRrhJj@OzYR-
z94;Xy#OUg^4qY$-S5}5tbXlbPNsNI>J^$^HgBu-`!|UFn35^g48-N0Uy*(fTnEym_
zK8W9lhG0@eYj_L1upkI;!WL+z%w*i?@XDRB9L0^QjW2p+x#OZ>x+4?$?0r0P!u_CW
z?|Vmd>>e{>6#v^h9x-x?u;c(>i1(&Q*%@yetzHfA0sOSJ@H~9#L1Hv<KmownbX3z_
z@6JwZRgA`VVKoHrzSl_x4;b#Q7I%r4cJ#PTOJltS`DJ(avV+lMb;CK}EgpUy^2x(L
zx&4Em6uRvkKF2z~67o4#z9-|W-S_EF=@k}^XMh!0@Za+5ilG7)?s+zorU6YsD!FGK
zqdnPo-ELiBZm^!NwA5UkapxEfe&%4o@I-DT8r7ufEKh145uZHQd3K>w2;!%p={kGF
za$mQ{#4a4g0Z+W%^|PQ32*>Y2{5@je^~Bxg)-)O%4;29JU<dGRVSx#cIC%&@AJOxs
z9(X`pxv9~SXw(kY!a@iJ?Pqy`^wXd(vx+k=dWOf7-#{O5g{N-tl>MDzGWZjASUPik
zOBZS67MwZC)4klS8~{c%Jd*+u!GbA5T`)XKqYtf*7>pq%!ddsUGaf8PSB6}5IGK;^
z03O02sCO_y(E^yoP9F7$e4`8dfS9WR9zckC6JDoUXtO}A=CcQKp4E?ev3rarCtc84
zZ-nr|KvjSP*o@MfyD3%aNl9>t04l@jLd#lUu5cxeyv+zTU1K|sSRSKAY-=`mEzSkz
z+acyE&hy`@vtPjMcjBo=Q%P?E;Hl}X*5<uS93ei!W0weZ4i11S%*2(*?8(?$4f~*7
z*`3&2jS&}Pl!U0VimM@364dCt$v_7HugosE-pwO|-oiOnqs!n9sk8wxaz{9PUezDz
z01jZ-1@Xb{Bs`LawFZ`;rP7pQS8#w;+?x*@1jPW#BOsiLQ7ys_jWS`sENuWTddVez
zh|<FLvAN{JcpyThRXlt*!huo&H5h47?MyG!<AFH34#NFtf+<49Tg8=lF+shuO)3uS
zx)-9(;fSSs$ZTC9)swMmqLN92O;a2A;1$ZN$ynWgEjh_XHod@7+ONfC7?XA{Sm>0i
zqAOZo;W_wLJ>PDIsX01pe_SxkXL~vhC=~!Vzoth4GQbW{p<7JSwCEGEXI}ily(imE
z(jM~g=VqB>oip%ty|VLl_UByxI+w3|^OJT87PtoIKVN}F&N)Gm^f!u6yV-R{$Qj{&
z`$tSr2e9{NBo@SI(!c`yeZYCz9ZxfgmU^jq)bR0`zGBY$OMNS9PB`NPg>;j&u*Iu_
z_uZ+feekLme#~5LAGjU;?x6H5gMRq-s~7yrAl!CW_6GI_zTgHXz9oIZt?VoB4eSl<
z4eSl54WK_PR`JjLmnOBUg#>MH!D3VQ2KEL%cmw-;#0NjzeYL%Ty@9=fy@9=fy@9=f
zy@9=fudsprFEwA`DepV(4eSl<4eSl<4eSl<4eSlvv4JQ5$o%B?cP#D}DZ1?(j##%$
z@|84uZ(r$Y@4H^Jfy1w&L&7z4J=OIq-ubBuZ_W4S+bVXjH?TLbH?TLbH?TLbH?TLb
zH?TMGg*Ncie;0ac^Dp#}_f_`>UVH=lTfvJzpgrE+z}~>#z}~>#z}~>#z}~>#z}~>#
zz}~>#z&mZ=&Hkd{&BlGFcfy`^Z(wg=Z(wg=Z{P(t@XSB?AKZJg-6ZWH&vI^-Io3l4
zzN1OL#b;dQjxB8Its4Muoi%v&|4b`zy><4j8*ISs+5rP`T)N*O_TA{=22NZ3X-0=5
ze+@#~r&wtT1@JYN@W5NP-L~L?&ELFL{gpw#`t6VHUm5h~kJ7*UB_iWDzvv!+Z{TG&
zu)ogcAI;v?{H`{0Z{Ynmu=W+d#(w4Pq?^3VWxyZ+ULG9!yz=wev+WJ+4cxPV{bhg8
z_O|w&H_+evJ<|69IDnV`%+<|ywl}ahus5(bu)mScy$9f*LvwP(y2uu_drurL2^P+}
zCujiRv${H;NB4WZ0o+6z_`D6|w|RZ9_qTkH^o>_>t=`$*z}~>#z}~>#z}~>c8$f?l
z%eFHd$=~tbz~Tn>TWoP*o4z-&H?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH}FL^
zu>a=li#+Ik%{SS=#8>15<xMWT$9(z*(3=V2Pak*>us87J4XCdTl{{oH)Dhues6{$(
z+3?=LWg9r{Eqa<!zUmEb>ymZFWsl)r<CUwdaBrRK5k2Z`7qq&-^43M3q?^A{WzGvu
znsD2^?gr4C0pYKElReqj-T->XBK&J_%iP{9$)twd(n(t5e~F+krAJ@l;$Q2Mum5mg
zf5NZzj=BD}`ibkT{oj!@aGwek0Nnh#fsBMogpx7%zp(%nlX^|t+1|jXY+&*$!(8_^
zl`+}r-oOiQV86#-_~Go)_6DxsK>HTZf?>}&*3e=XAGm@2ydJph3tG{y4Ep&me)#@7
z`uOpu={E=c@wML_q}~8M=@AXBGj)wYX{ZC9Lp#xUmM0`Z=v*q`@Tx_)woWWe#Fb{^
zDGp(P&!(jy1QjDr#=3E}IZo1ukEy7F@X|f3N>IjaU_S?-c_MpUFW^r1o_123NB%J<
zMhDL{ULUZD=Fr$cXiGL{BZQZ_t=hc00GO(^6f#1vYlh}hD`Re0#r;sE4z~#<lw%+O
zF!lixt^2qMhXC*g@Z$L0rAK<25D&qk8<aBPs6Np>QJ{+{aLT849~dD)!M~FOAnJ54
zM-;%&d{}HIo~00xh_I=101K&_Yn7mn{+I|)pSGOQ3Q`{tQpz6y=CV8o)Lfz$!KtA@
z$jqb+5zL>BRk-=JgkD{7vgt*s$QAJ-1p@Y=B(}H(i7#U>r!rz7JaDOyQxKrQoK!>=
z$iV@{w!=Nj6_XbBjF0tg%1eW(aRrG&k%HL}<18GGu-X6``D-F;umT7VeSFa)-5JyX
zOu9ko(6$qfo#H+*Iv5KI4lfcbV{~*KsgSW+1}wlak%Dmh*;*md-26$+&e>7B>1@yv
z5Zzm=@*usn=mF#gg#hYCcSPBjUec$Gvt<VK8Vl5zLFX!R3+mLy=#XqzQN$XGQjv7&
z=RzSDp%Nopk-$BGR>3vtSdXYh&nU`SeUV2VCIdNYsfI+BLXi>QqyWm!{1QlnO|PaX
zC&5c2wV*omJRFTj5T{fKxQ3|eagUINM+qsYujQry&NciJr~-)e1(xq&b_Igq608W%
zY9rlb9B5D(-~jf<k|<-vL<NnPeG$f=%|sI$%$v@D5EZ%-1(2|52EKJKS$@h@3z_yp
zkB;>L13^PuB$~m;#;^(tGd&FlcnP?M0x=dfjR`RtJ9ML>O0PLke|145Z|V_lfhq`R
zojws8;2PE$8GA&xvxEUZE?JawK4rTI8xIK-5|(jvVN;j$to80jJu+^na29T#7Pq(l
zHfNsGS0duC;;7GAKFPmWc|(tsVKgo?8MndRhFZ%`fiv)Qf=XfBsFHlqBgU_%l1edu
z_MmeIk+jiLRp~b|aP1s+SfDz+*E&iEBw5CA{fg9mL=-b(v@b^(<Ph;$QR{T1PrwE+
zpi8!l7wnUal;uyv63N?W*Ic0&0HXq=H^X7Tay|pHDygKCBUoS}?3xpz?%Shm%PF6_
z7o+9mEOA6A6#!P9C;>c%`E0}v$Zp8S;!WKGGmyr`fTiHea*D7Tf=P81UVVYRflF#Y
zd8HONP77M82A}|7ynyDLltL^*VzjgcO;-b1;FwO1H9tZd)&i|587d+ijsPaZd~_F{
zQcNJiu}Try=-`Yr5msz)Wv)&v@&G;(mT@s>anD2)(@`Uc_d+pjwk^m#MP_*LZnz-K
z#D$jm$c8+7js9{aXC2Bpr{xr{wIC3!&xJ~o;|GwQq>%wrjorvOfQq;fqjD?SUaU)o
zhE!8;UJ+jy&JO5ifH_t?sYGKRgVxlb$2Nn_Hu1m|%Ztdr7KO)q6N>n=&5$9@VOB-+
zQDWxtDO_@haPt7W3<^4>OjC2iHz}SWA`<&7yo_hvOm5Oadpnp{mWFWtUWQVvU6(+G
zFL<QRI$THQu$0e!$!;NthDH?ysl+(6w38^}Pvdmp%O2>|^47Wrtg^VBq1Tb(EL_c7
zOjoQmS7&^Q7+t!uun-|PXs4?=pim23i+~d>ruD(ID^=s87S;#zVqdFo4ND>0Y0sLa
zL;c)Dc@o#Eu&H5K2tn+k={gHrTZ$NKP!g;YL({OQ=0BJ8f^Ci~c<sI7)DE!yy4F=9
z!1*~KQWeZoJp#uGsJ;^-Av`&t62hT?rV%0$N?6Y<NumPaLm$#1Ei3?mgB`#FM+JjX
znm+5n!qI6AIu#Tct|zBmcMvTL4*XylHwr5n6qyLsqm?)uf(nu-V_Q=umY9nQREg56
zj<SLzC*jCuejenof@`kMc#DL#N(<zw1~KNZRUjJ{#88go0C`0rq|k8*NT2n{Y{$tc
zy?_dU5gp>CP~FOblQtCFMN<e&`E_xL1Wy1C%{7a1&gY@G=wP{&pX5U}Y{jQ(r(vBL
zsxZ?PqcUmb#;7o20{;ldxu9@3C^4noy8$&o+5iurzE9{Isx14S{DuZtA}c-a5f54r
zfs2qi;c^Vmq$G7v0R${U&mR`cXqpmqEHv-!@nR(nWz2kqse~E+)~y0dkx_(=E?B!X
z83%mw6gIyVXPrHff90Vvz(ZJpJ?RnNljP{hQ6<F1|1y{Be3EO9`P-dtvoxCYls0GK
z)pNmaf!YV|q9{5Tg34XiO)(tNV}}^c&QwaDnSp2sFGHd#zyTawrZHPk8p1u{f^#@>
z0s@B9$>EWcAvReH*r1+Q;>bI-4oU%ZFr&IkEFYn^X7gqb3&cQ0gde#Ex&X>9E4bv!
zxQ$b;f=ZQ!$^Zv2+?m-XkOSC=&Bee4kk{3?oUX*KqXVJ~W>UZe*9EPH?5J%O93O5p
z2KW>f8j_ieg@D4i6r(H^RS{}8Q(!6Rhvt9-Hb2bB6WB3qE`7jkSb{3^c*rDB7j)KV
zM!1n53s#LnehAvD^jZ3lUg(jkQw5a>W7{*}1&}gGNufX@EHtS#WJ{0tRRpzlKQ!7Z
zu7?TLAy{=`t$`(IsqPi5$}(+E2D6xLmh(v5h(R%c4t*j`gh3-jM)>Fg@xDlL=q5-J
zS79m<LJGl$lT_X+o@7EILem#l3^AvWT?@8u^RY?o7>Y^_l>{zAwE{B8wf542j2B(Q
zC;$+b<XOg3uSKn(lE6YL=h`8mni4O4B09pMA`k>nZQc$<vAE84wrs`%V3W$d(xa?0
zsmCG$_fz{;M3pc-1XHo$Ehq||>Wn?5Twpm|RzQJut~v3QMkPR$;c4Q?glfu2Bt7g=
z@v?bFL(usbgVC$=5q}e@b&grf2-i0?S+|~)cY$<*VzXzV4nyI?A)OqPM#iDX9C(Y(
zf#$|?m-T3PQBCXDRk%v*)<jW+>=Zzy1PTD|PYMtK%wxFI@Wa%AeFo!Zd0dY17)=^e
zFe#hT8I>4QCnr|HxH&nEJAlS$Ai~Q>x3zOHLo^iRT(x<ZsA!T_G)rXF=Hldr!JtGI
zWZbS$|I?GN2}FcV*8{HdDIQ}d>8Mi`j{Y;4A*r;ynSh~ULJlgd;_%|gNR0>8=F7#H
ziByEe*#WH~SNLoN6M#B^lekn+-Qf)IS=#rkM|!Yl;d%<buASOEF`G)0=pq`^S=ELh
z)oQ`kZO$Vgoa=l-9xU)A?9k3o!62WF%>gI?IPYOpW7~W}qGFi@hyV<?p$kaeo~s$+
zdR))f+xtm9QrT2UpRVoz!S9Wth?VxVT&^2jz#>c}opn`=$7pC%sZaqsE4N4<VSR8J
zjhU#rilZeWA9QoXN4UIcj{#f5J?l}o$Pp|Q-a*-z-I<zZCMEL>6<ogF)YsUHpU_O@
zz<rG^-$hm_@*ehx?_cK;1re%YV?3jANvT`XxGg~M?IP*4YT*F3HU*Oc7Q=Ft9is^N
zoP{HMF$DeYp!6$){`j{)v43TdSj#wcd#!-Os|~_+m&C$ETxlkr;t&S-Y+4FJP%+|U
ztQ%LG<0Osvn2IV0FWtkc1ZCU?_HzK5C$h)&0`7G0X(zRL<R5clbnr~$^#Pk`4vh_j
zwq$cQLU^g$s?DnlfT>zbAtMC4W@s+8GUkR=+z&<SaGOv<IR*j%V;?Zlx{sT12mpTo
zFOJ_`dZecb@enM!K`9fC>J!})1-h65r+jMnfe{iET>P{<n8h|jo|Y%`p=MZYE}o?j
zk%+LVa{vpenroGykG>kh)2A(Gw1U(}gp~3JfVnKs0X3KCMQ}tYTmqDJd^VfH&95bt
z_QkW}Icc#}h}N293pQ_VsnFsUNYE-<F4+y7Qc5<(&WQj8=A<I35XuRZ21a4h!d8L?
zQDEsdSL6#4)KB3ep-`k?BE&cgha;>ufJXkB$QrBw!b2Zl^hkFGH2{-tP&%~jgkz_;
zPmB)6f`Y?~#L5^QT}LWptd;=_a7?5i+<vxJMHEY8R6g*;AW|c?Ap7ggv}~!=OORsY
zP*a0ik;)-AC<Ku5x;4qZ^pZYhoc%IT@EX=zH)hb5nRY>)+87;@?J9~`Ls2S{F8y36
z<RVmJgey{Ai^0(4lx0mi)+4g(8AUm(FY?HZGmxVwWpyDW?jhYWx^<0=Lq<EMqMQUT
zjnsnLwdZb9H%9@a!c$Py;~pV5j}lT)U(5XgoNM?cPz4a_3oPHm><R?IC0G%j)keC>
zIMARnzya)yB~iwTi3%Do`yz}#n~5eim^YmPAu4nw3Ls(A41DWeviy{*7BcOH9v$lg
z27-pRNHl|wjbRlQW_lV9@Dgwh1!62}8WUnPcIZY$m0ok8{_28A-qa)90#y*sI(;HG
zz%{HhGWLjWX9)v-T(T(Ve9CqaHXafvBrN0T!lo|gS?k@6dSu*!LWTl7D_3#*w79+X
zu{rZ*^~^^;2xpz`gZKzXJ_#=pFZ4(;(YVZH+#1@+so;nOaBxHo^#t;Vcku{w-D#<m
zXRHNI5&_CaNmZq9!ob}1&^r^nIMNX|&opEuREcPX@J@*tF^Z=}d{)#t9qALW0SxGp
zE#n3IBqL?{6R|{jG1aafeODy{FzO5G);G>s&S$_<C6#n?1Pe@rU2{UzeS4H`IptIL
zVziu`C5{NC0>G*hC4k2;pN-f7*$vrPys2AY2GY0~uoQe*P7zi^FsZJ>t1qxOa7hg)
zuhinkX+bO102Ba>7tnl@Qiw%JjFz^b>1rSg9Mj3M=0|A5TA(#0Lq&wc5x``akM6=#
ziU~wGRw*JI9h{LS!io*9%+-lS9>7P!GA`yU?wN>UI%=eGscrv0SHM=%e~HuJbd?-S
zKuo1V!MOrQ_SCD_XxyB2DCeA(Q@qxKK(sy=DoI`|fb=Ad447){M$Q3L#Dy4@ThaDn
zT{1MJntJn!_`+~@KsN)-vEoT38v7WurUpH>8Em$R2c}qFME<oXJl>m7#Gh@33~3It
zDw>ZHGmlT<l0$@>2iRp$&?#k_nj5}J@eC1>*k|EoJnLq1lMdS3!Mw6Gg!A_@lw$3=
z1S)*NBX!o{Ix>f)eD+Ir3qdqAswhY$#-XL1L=k@)rvqR1K&O_s)-_<2#qA8ejudC%
zYTjbHVy(G4<4eTo(v^jU2)RK!UCjZ7THsm)oM17n51w7A8XvW=KA0E#T77F+3fWG3
z)+`<B=O)UNxK@Qt4Z}hRVi!%<S=icA#8`upV4WD6hCOXRUYBOQV4LF#UVE=NwF7Lw
zu630NaDEPmR0Z=?kHB#Ps_%qI2u}{Egm5UJX@p3G64o<IlBfXq(1&zL3kyKtU<dHP
zQNdu8rq6n?aC91jP6Y*q>&a=?9Yo8513y^CjlxO@k~u8E&ZaOjfVHq7KoOdDgOQQw
zBplhm6*_?f>fKgw&DEJ-)nHp8dC14jqDqA2p0I!>BK#&18<IGE)+3|X4_DakzBBvR
zCgq3{(4$gSRba|bky0XfwschvTcQHM0la=CWW)=I>jpJx)y60rr;`c%BOK>~!r`FA
zlxEj}8X#?e2T&gq`i3gYz9+w-0hY*0k9)*}7DV78WKOso!!s#K9aI1Ti_r6j#WI?v
z1RV>_dwaZCNkbVkUtua?hQD>Iz*1xsVWSJyE=|S(pFD-lZ^c<>Pvl>Ds0{EBR$xzh
zg!d#ldU8|=aq%B@xXvfJ=9s_T={8HFNl$5W7G6CU>=vkf;4X@ygCVHgW!)6R5j}Q@
z(d<m6^qComhVU{ZssbFq!DSk=1*IX}6D~N1BPSqWIGr3GIT>P;wSW!kc_og#Q|q7<
zKnF9btHkmVYHK!c_OL(<R7CiZYoH6D?6QJOu8i9_<tnIDX{Zcv0K=V`Z2~!fjo4fa
zTmX4pjmzmu>^eFix?m;+OmJP$YRHb-R>AS%Mq_|aVWA<J$yf*|j7u@fQc)G5b~6Q*
zf_`WYIAHU`j68uI!{*Wl%!VbXGLMH$0(C)WeP)Ck`LSTtDCCEry-J^@59x&-sXA3q
zi7>W316}|rgOn5sB*H?IT0^$<cwa?OTlYhwt>Su^P#uC*7uFhBf|lxD!Ky6N=43F7
z*=9M9#Elpf1L)8v(nJ_ELS%%GE)egF6o+nt6mb=%5+S4zd^kzvt>Q^0BqB6@am5gG
z3fZ+_>oy;o)Q+L3)KE#_B2+6NgIsGbEy#G$HH-oPaY>$KJoQ@C3MvUKq;jqu5~?Zj
z(kG%L94Z1q0M+L0KopDXTxZK>EC4pC+$%lGDwBFFB5*&oZ$(rI(?c*78{UGV(5cSY
zQ_2OF!({~&Sm&A(UujeVL>ZnYj!dYgj6~AI9u+T}XEX$ze=!)nIv?>jky_`NwTy6m
zQ<HV;NqHAYCnz?17V0n*J{;1?F==ERddz{h*c@nXEO%Lth8NYeeqDvD#BNO#MaWJ8
zR7#)#;Qpil0l++lI}JZf4cKQeZkEU87?07UF$I%W(~L$X#?;9P^5%LyId!v)Zv^Rb
z<`VY~Y-u@`VRNJ7bdpx+OJwfwi1UWQphOmA-0o5T)03|WM1)P(1FrIU^hQX^PVF`s
z1_i}3NtVp$AO{sz@xi^!%yBx`&#*))!qUJ2tuI&jYy}g5I)Ib7R8ZaF4Dean_pC>H
zuxH_V3cjwL+P=Thv`IH8)m|YZ1W8Z}wr=z4&2&!TiSz>lBnfc_`E-MT0)X=#Mm6OR
ziHc<sAOf%vqZF%A<!T%~VBgmx^_EmfpRVoz!AA=!I4kXGRa`f?fJK-{I_s(!kI~Sk
zQlSENR&J3z!usGc8Z%LK6-P@%KIrC%k8pX@9s{<9d)A|Fkt0|ryo0hayE8S-OxnyP
zF5^bf0iqaO{125(MccV-jV<3rRw?ox_9%6hiHJq0#*6Wc#wDe0N#nLa4bpfVbXr9m
z0GNOWV7(5uq!S5F!jX*_q5jICKk@BPuKmg&VX?!#fxUsffxUsffxUsffxUsffxUsf
zfxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsf
zfxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsf
zfxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsf
zfxUsffxUsffxUsffxUsffxUsffxUsffxUs_4bZO)`uQ(@`2IWk`0=OdHwXQx{kwyX
z*R~mZ1A7B|1CQ9i#4mn(#2CBB-oW0#>ulhhZwtT$z_;GM*yY!goo}7!R^6MAf9tik
z?)^3}c&iRd?|09$Ck~cVAzV3ohV8dEus5(baKZ+}H!TrnW1H7*0KRh;<*Jf-V<bcP
zyfb88XbpRQfdX7BEL^k{*bf#i8f4p=4eSSC7P8a5fxUsffv>Xx_(y;!)KDN_XFvO<
zkKBNKr|q}fBRBt)Ui%S0W$j<I>3+A#Z?E6)b$rpUgHO9Qy^-~s*(0BJXLTQac~#oU
zMH}cl=ZMdX8V+oq*Lg4$EcI!71j)Tn$V-wgzB=Nu^vN0OepE-gjqeRiY+%2=CYH1*
zdjop|djpqj;N`yq^MHKzf061)xS4};sJRt<WUpIi{50L(`KLY6`|p$A6XlJu-yUY2
z=P%@ZM>F!>>FZB;_Z7Fn-oW0#)f?b%Bz#`oZ~N~JyvGLiSHycfrz@uXxM%;M3)MCc
z>UPDG-=2E|uf2i%#kIe}W*?H-lu46|M`o=zZ*O34V0i=j3$A1uTejSSy@9=fy@6NW
z!2T9@+SAFa-1|IjpWA3}U~gb=;N%VPcW6FOw%>LS+yH!ru<*d)PH$B%b?en`)t!Eg
zZTNm0=x?4Lz27bN#CrpK1JBq%{jQ;pXAA;*=E2Hz@)fXO1{Pz#%uH&KvdjX>Sdn%z
zzXAH)LFrcp{i$!iwtr>N{MI&eZ(wg=Z(wg=Z(wg=Z{UF&IOnHc-~!-VuUG8y>&ecy
zPIRm8&Bwp>+FQ3b@P#)}zv}lF{}+A_?5pn$><!$$0rkyOCAWKS<=(*Fz=v*Ne~<Xk
z2m21!+HbyhcuIT5sSSMeo4H=@qwhHNlsD<(4aj%e`@P-XZ{mv|=NGUq&wM}mH+a-n
zKcuVu?_aT*(4>+7cbQX&{k2DtW@E=2NE^&cESo&$Z^r5d>Sp(XR~NVWGaKmdD~|Y_
zQJ=%%j&zoo5zf!)bFPb>vFGRZd4?O`rJS6hNsj7BxADD!i4E+x*Tj-GWp7|_;LC4-
zzE__A{qlUrlOGfh$Y)<dM^0|JZudf?tPlUDioNqUeQ<m1r)=Q-ub1;3pR$<UbZ=m9
zU~gb=U~gb=U~gb=V0Huh2f*1)X<psi%dSrirV_e`1N7x~wm0ws8_4g{{Z^fQ5@u5-
zO)eGjh1m|9_qrS4r%ETsuRHD6m`p#1l4<N~>|x(>Z(wg=Z(wg=bp!cNi~UE&>e?3g
zd$|kC+SI*)y@9=fPu{@S`U9cd_G{hwli$5>dAIiV+waD=e13cAy@9=fy@AJUp#5o)
z+R6Aa-9NBP{gpw#_U+f#er1qaeouNtyV9B3-k>zp0neeGXgtdkk|1<06>xafB3xT1
z7AE3KGw~FMFu-TiQV@cQ5hr8axY`^iX~f4=R6%&@9#$nN<2JCL1JEils0PqrqS4Kv
zlNk4&(E8gv@{Tz%I(Vjj&(sXqM003tAhac$vk}5e-BxX0T>#8I(it*B(2--7L}eUz
za;8V4VS#EUY+z$hSoe|5j$_~tVDN;fO{C~pkMu|(9)d+TC}qM?eWH7!Ko?WsluzwG
zFhYWYe<ue()ahQ1D1f2)u-HsIOCcf=VN>S-7E(3WDnTFpF%h0VZ8@VAq&^~~ls^ER
z%C6!7_NZK{3yugoQ4$zs9iPpnIONw7dUeIKRg0y9oAB9444_jITik-Al5t>?%7}sR
zz@<V?L4X2tQV~@k2L}|}4)-WmOj_8=M=?~QXia%(VEc13g^Pqjkuu$!h;bGUM_6qD
zjr=u{wYmCXI`r{HkMs_p24K<+N{6<caO@QKiP6DWP;hvWSQ(?E>qv!+)iPiKj)@e6
z+t1djh+=7s$_JhpL~6tqWPhERmMxWf2~unvYHCm`QaR)Xg#c1scO}-G^Lo;up`0;f
zS~s`_hS0pmIFxIdqY;pPE)=*3l@{TO)UU2!pjB{9I@Tld>KR2jt1t4%T{DoQmTE|3
zDHIv;O$wmw%rAjN*z{_OauU2WQVXg>&%@Dp1aV4*fNO}V9`^`Yc$AQW`dV%Z;9SEm
zfhvGVUtswjW>+8xF2RcMtTxh3#(@Tv0S;hqEQvB!OjOW#|34N?xT>i{T8n}Z6;#yN
zPC?R#;#{N`&+<qS8@#7cR3z%P(-b_B+^#$@9~;9eEF83p%D_v&H55qX6Fo`d9m>2}
zk`xsO>aQ+hQ``ns5bo!RI~E>r(Xh@&cCERQn6%MS5SJ{<IiJT4$eEOpi-8Trq|k!r
za(phh-ksT_MlJy=oQ1oYYJcl*^EM_Z02sA{IO?<Nk9-ndth}K|iiyT$CgavHx|$;v
zz`<?0?v|s7*eJf}kw&(Gq*BbEJ+O)}mo{3eD!lxDb~i&D6(F^XYBbMG0!fy!D##X!
zBnvH$88NB{Xcj3ze6G|w9qALW0SxGpt$$#Z6)~iT6GCA56R|{@Uzzm~7P4w#bS>cG
z<uSr?J_D92sic!5SYRUTniHbrTdhq8PWjZm7%eAfi6cU(0C1+rnv8{>VLn@*4hVLu
zDvT0CZ)pp%qGvr~tBRDvmzhRqSPel$J^`nG=@Yaya7hiA4E#sqw7_ye0l;v=Zkv=s
zTti0q6qE)P3z>3EC&#@JMG@Mt7HCb$P*LG337df#au2N*L%@QSDWXb*8-=5qp|&*A
zL|75v%3PgT<N<snEaPI%;+}~prlUp>?}cL6Y+I0fip=of-Ecvei3=_BkqvqF8vWUx
zvkv8)({hT}S`diV=RzgP@dHRt(#U|R#%|;sKt)`LQMnauFV-bPL#nAauZS-UX9sjM
zz#J={RHCtuL2GKzW1GQdn|NS~<wfLQi^Ai*2}S(bW=IhIu$q5hO)upvtjeXMK7kyz
zOnB6g>#l+d2v-JzBtoK?G^-*kBqD=b@Z8yTF{79?X*^PlYz6XQp$&u?umTHkUa0T|
zkMNc4r8{0Dmhx%f&MyQ*qq2fj(#1OJB#QXcI34)12RgM}_qUy7mHi|}@zuC$P@1i@
z)LfnMRbq7MYQjQ<+@PJV=72&ia4iB(u$a~d&#qLBk6KtC%!_@kzBMd`Y^ObImJaoE
z6Xi)<tHP#+VIc&ui>B)=Y;7rGtU*bzP7F=Mp5{Med!t_HScz~2uf12C+5xs-*Sbms
zI6ntOs)Bi{N8mUC)ptT9geM17LO2xAG(sdo3G109NmKxQ=tDZBg#{pRumgDDs9-Qk
z(`P+cI693%r-A~*_2jhc4x(kjfgdd6Mqx#RA`_u{v=WCyP(c!9Y-`HI5_3_3Dp6Y1
zQC5)TBplhy&x0IRaLv^jZ;{YeX@OkTAjbT)3S`5A7|L-Xz+u(E;RK}5dSsU4wC11!
z;J)W(Py)m+f?6#HPP(V!beN)A>Y)H6)4>^VXqH%%b3UbSZ){OHny#JXLpE&1r*72B
z1qP(1iR%V6X%)w4bC!b^>Oe3kHABQ%z(Z;XrmhWdK^x`d;D!cRA}c-a5zk%_fs2qi
zVhIomWDY8TfJK;$b!JjiEOS`U=5X3pXo?cF)Q^)TvTg=sOGRO@oTwDx5$0B;?I~+A
z4)|oM&2Pn7XHVo`d8iEV5LRGMdL-6h#*-XfLX{Awf2pq@9PvqTj?KTZGQU5ehLk`W
z)KUQD%31h&o7K){+%E6Y!4On-<e~v^M31R4nj2CneP#xtA$)l{5+E~=H>3rnA@*1>
zuU5`_j59r*94^k(WG!HWac9?$Wh*|2o4kTj03FQ8Gvzauk5F5)d9#NFVxS_zk6Z&?
z0A-gITykaH#wk}prAk9(fCCur%xn|L0c^zPV&DSE>uOw1S7O)E0nr6BDPV%@f>uLz
z)V2za4>uYEd<qK<$xOyVKw(^pQI?9T2(_CjuoUz|bHD+cA7<nU>=-teK43O1L6vzt
zWD=+gI_ond+{ljwt41L|1npJ&EPY5X^hnjIf=Yz3?HTX_NExK0P#_T&n$#MyrN{d!
zg4((t8f_KV!-VP(th%t)z!J1n_X<{JnKmbbS<E)ec_ePcpcp`hK9MHEpb;V?d~|_$
zU!*v66Qqc%FqH@)h2X<UDsL4}G9eM6>5D6dm{Z8E1zWfI*raw0MWu#H0vDlL0U6|4
zduc(&i>_f50EkQSEaR!yqE=8zU?G)r?T}DSiI+YR9pO+B2m+`!ZwI1ST<1DlHe&&>
zN#$PYQC6AMV-bP-seLP=N|+vkso3xq6opQ8#-36xupBNcpujrUocKzk5+KU(G;w4?
zHDx4{9`>ks**v2m==_Vp=+*g%zlqd3$E;<9>zkUaTTjZnKsrIO*|SiGq4435PL4?<
z<IrOcyv61~b7Q&7dNjPKruFM8TqSmEq9{Uk3ZPN~1pxOa1qcA<G2ChRVQRoWgK@Jw
zE_<91k5;}Zn3PTFj7p5DlM}08+?<@o9YA9=5aH#c+uAvpAsPyDuG+jyZjY|;ynPy>
zykRgXkp&sIcJ)6!`I<mP*mOPMDxcypc9M=dRpIDAgBg-a%bN)pDkkKh!YU3gj*QfJ
zP;I_kjG0J9SezZu8ghltRxkmm12~CG1=StS0H39O&w8W>dls&z;Op9{%@eb!G>I;v
zF`ZRy2vV&UY~AKO0>Zh@C*;8bPr?rE3>6IW+1MO_0)X=#Mm4t0CnPGCNq`8za2vXS
z)a|*NA+E>ue7(J&)FYKmh4ks_4iNm_D2iBVPs`=H!38YBMABJT#dwT{HkAq$u(NWD
z<Pp{fm(iGss;f9!BJx2uM|_0KoAwy6HQcivb&DLqLg5{hjoF>4X=YL~&rreT>#%QN
zllcZ0KcSh*&>g)iu&BHDsIW!e!yYA)Ohha~)i=g78kdy1C5_txHAv%e&}kKM0AK<h
zfb}}ql1?Nz2}d?!1p1XhKmWxK>Hptb`uOpu={E=c>5ux|L1+aC7pn<EtFk#iP={A9
zgzHv`g^9S*OgzOQ4Di{s6ojB+#K~9}R~_nrxEG<AiYf>%J;JI4Wh}<Bu+1A46%Y<@
zxGZQR+!@%$HjnIMPK?4ElW`=8uxJQ}4TQF2bM``bsoSc}s|$dsT1z1#1iNNvj<npc
ziu-YvVA%&uwC>aIG7|xR0JjI_ts_OpdZd*h9)d+TC}qM?eWH7!Ko?WsluzwGFhYWY
ziywOjv)E?H)AD3K)C`Nw#j_M55)n3a4qzcwbFC8e(N{xw`n2VYR*?FLkW&5tFqh>y
zpym?22u=+JLS`mqh+zI~tisK&CG_fwlT9y5MXrb!DG;y^C9%aVNPHQ4Ih7Fu;ektq
zoPq!a=A<I3Kn@NlwjJ(Ku9&p2XMC)0Q(hWOjVnkLiWJO-7-!*dgw+Pn$X^p#gB3t{
z=;MnX>CT`AVA2grhqj$?>=gHj(ZN_yaCnhe8Ka}?NQI2mGGGCYi51FtK0$^ZPZ1>z
zXhCkGv)0$c9;swgsHs7laTGvqPza#Nav96M^pZYhoQ*P|*I1y&4Ej}(*t{h*hAC)V
zXNKBSM<XEpTqtl6c3ozv6eXcN1kIf*r&D`Gy?RDb&gzRi@>ChfQA;%>vJ{Gp_$CEV
zcIKBrB5Zm!ML7vx8mR@<q37XfJc2l-LclddRgZgwEIdj`L47SZ1#qt6mp~Ojq%W|1
z53?%}1eah%cvc(fCgVVZ$^ZwjH<m;hD<&#vyzGlG{%j_i*kImt285{4l_-FOO*8PV
zd&%-su3E^n7kYH84;Tm<+9J^mJ~oC`SeWT)IKWH5H57=ksA)`y(b%CI6;*o8f%>Zp
zB6(Afa0^sHIP3I@*Z|kC&dAs!x}7Bq_;JaiobxH$Mc8;qppdYPqYIn5oM)|fH|mjb
zLxr<&`?R>d^|v|moW2qfe-%f4&hknA#mXCcqzt2RnaQ{f?l#m~b_$$<rxR2P<3^R_
ziykq4J(X07`LhR|JBXx>ma0m>iGge9u)_k?>AlubIv~k1hU-_P?jxd@5u<%M!XSr;
z&x%^7BYgrkfB{{yWxQaYWTY&AB9=(rM!V(;y#N>$AiWt51D5j{kX1<~ogBde6Jgh!
z5Ov=kWm``9)V&xjCufNxLa6|->O=|PG0bNpc0hJRHWqK{7MOuFE(R<GUzSsZ)euao
ztMKXz><wH}1IjD4xN%z0N;Lol0OJKT-=q{`5fY=NEoizL$O6Z7a;*6g+OQUAP03IZ
z;cx^n8RnzA@RVW#5sp=g$VLZeq=~R%gDZ1&Vvz^%k+6)5Ig5KHqL_{vLA)1=VY6*P
z?kO_EgLlIPVJ0rL%ttol*=zKdD>>^>&N(foc&!D2XnihJk{myP^dyZ8m}=}s&H+@!
zg&37v(e`3pGBl)`dh?3-!f<v#Hv`PE;z=bM`xvyQ20gYJY_^F9rdVD?{<SDP-kVUw
zpKXQ=X%4e0nvW7Qk5A!}Lxh_L*kw@ADP@|P8@@^L3=xsoXW?Z$>t=G34%*wnys|We
z^Y=28V(q#FDty5sb=KiJGKZyn_DgmPK{Pa~C`cv7p{1Qf5q}z|17G$)r<S+YHDHy+
z?F_w+6ldXT-eS69t+_hmOT_5Xm4$@}xj{Q!%>jj4;93NnU@@%^o?WRLAGNSPm>2t6
zeQQ_>*-m@bEFJ3SCd!kzR)tLs!$JsR7fsh$*xFLWSc8&aofw*iJvINiq!(;+T)}Ja
z6{mK9?bo%g5&_Q70g<X;p6U@ePC)gY5DDSQ0hJIA1vHHiiBQ6NW=Rqi03Z604ryTl
z2psGH9ylr(jMDU34;GG2W6-Ifz;Hb|?YbxuED=V`^Dx{ftY}bVB2<r7;=`#WQN|>)
zc{2(N3RFqc<p7l~#*sa|r4u-y-fIQdT%GZD2yK-X)MiHq=C4&C8y3W5j^hA%MIofn
zaS2GD^~h|;$tb;m3V;zE;-ygC%7K$M6x&5p2u%5Paft*^01nMHi*nBAp||K@xs)TH
zh&!N@4O{VPEHtb$LltJaVpJxr+!z%`OyD2kI2RNS2PLModpDp4NE_e*)b|N}LzQLU
zli$z)OJt?TJ>o$NB5)BhCtQx<nUtgsDu94R==sB98BJ4yj)msEJzlJ&p^TZYFqJUF
z-?~*`DKd(%(FJRlCgXrlp2Ft0;;geL@~=Ep26zZ7uqQpjdy*VIIjV%X_+RF7olkPj
zF@L+$ZI(upp3>$lym~I!El~TwT@*zJLr}TPx+#Vudh8IR*_le|Gcynk;bllv1vr3%
z%QR*SN<+9OTyPFYPC&qLIypRYGQ=ip0UOlwN*sBo)<G$N4rWwWiRB~I)@<JFVSyN^
zi0~uVKo>ySWd)a98MkrDRZywYP#NF=hC4Ic1abfyvAGzy0P?yTm(!Klb#y><!AuI6
z;JTpIkR7$Hg5$%D#sHtfLPIi>u@F!gmtvHqqAEh|W(q6?{m>k6z~+Y;c>+6z&7}{R
z4NFjE9uJuW>VnSt%m_E~W5KFX$PYn#l|D-!(hEIOb*i8eVQhN_yZ}-LDJc|4goP%x
zhHUBazKWo>?uSNO#q}_uIs~gOtTnI%E!Dk(RavIZ$zT?<&2k=z8!;#b(4kMHi7;q{
z$Os=@Al?@#4&4MP;wnrfLP#O_aFWVf#gj}(L}>criXr9{vTMQCZ9X=s9Yay6p_0Hw
zs8&D*xz=7<kny5x7zF_0l03_J>b0m9R1#Q7<y<=?R8!)mPeeyJR0M(ms?FPhC>Gbb
z&X&zs0BlmZS9+9HCiPfE;C^b~il`E%hhQo;yah#}Q=PG=lnX3}%L*v4&NU~#(x?Q8
zGCWNjnNUp`iKK@;Dqc3vXb3w0VlaAjKH_g8wazhX8R7b-ChOLd@-C20P;B-r)L|%m
zIHZ$f(#SaUm;-OIIndl#?y?>YFRE$%x(Zi`-I^$hkevdklt2N%{Ye1=fO!me8h)4>
zu+L!JERV}E9-~QP3MOS!I-?R}>g2>K7&j-UaR<;C4MceP=(ctaW{8G@oU1nP5*1C-
zie`zd+FYF6Fc_4`f{fc0>VJCjHGzn*>3YCbKE-3~Bpr3C!qI;QGbELkHxn>aOvpin
zRUBR%8L9E0+I+bfGm(n0I6I&<<O-jyU;<DFa1xgasymzkK1=(a^+*r)EL=~)*R@ld
zCuUP=5?w@NI;+|cq*^W5y3Kh6gmaxw$b$u*gdN%$Dj4Lmu{i(*0Ovi7YHXWNNK`D7
z01<%UHgo~0+jBKTT#xJddV4>qM=F~N>C@F6Ao#se6tU8tmdka63s{7Sq_eJy@fZzl
zDitbVXXO^jBdiZDqcIazS8=pN<b!UG_z0Ib?J;0$xMw};7CC~2!aFD%vpZAM%%o(V
zp@PfTVc)<e^9`PU#|zuEfPvw;#+L6Qs}y+;dz7Y-iHJq0Mvd`|#wDe0N#nLa4bpfV
zbXr9m0GNOWV7(5uq!S5F!jX*_fqr*T`jtU{`rB`;{mLL=vBSNAy@9=fy@9=f*WJLx
zpM^IMf2H@$Lv7*Sz}~>#z}~>#z}~>#z}~>#z}~>#z}~>#z}~>#z}~>#z}~>#!0j7&
z(qDU?*!D!L+t;|2FR_9BUU-Seu!q<i*c;dz*c;dzSlYn;^LuGso4hx$H?TLbH?TLb
zH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLb
zH?TLbH?TMGW*b2ND=~z>+4c6gdjop|djop|@3VpZN6-5_t3B!7z}~>#z}~>#z}~>#
zz}~>#z}~>#z}~>#z}~>#z}~=h8`ytOcHM4Qoa?U)`i*aYX6;u7rC(g7KGLs0;*Zjv
z9xx;5tb~|HzfdXvb}Tyj2Vs;|kW%{1ODR9@*IhWOE}Z6*e2QNVYMk#wcdFa9dE}&Q
zX=1Xafdg`O)`F5zf&M9knkix>RCiUXN$!^&qG~54_w*6YmkV@*{J|GUXPGEZS6&K3
zzS4k2#=4qHQjH5Ezm{t=HBCi>wU<f>EEXU+Pw^LGRV2Itb&xzvn&~MYy=|YMW6?<d
zg7qG`AhZPcj}#N3KIr@|%TFeZ3RlXz(IA=-<RNoU=`$ge^cLqEwX8ha+q`a5={Hlp
zn;0p3(#pkzDP*{yg5PLWkW1yIyhroHt3V@92I|c)sR1-|pTzP=mBIqBz8bz^o+9aA
zE9tg+mrO9T({nm(CkX+)m!VmwnEvDRW?6ne(~BRMpz=|w6w{YFz0gSqePlIJ_>+k>
zFB=wh=J83wVpU#Mk{e`ea_}D!mRZEEGv!8ITIS2Ih1p9tMR0_h+@NTw4c(lstv`_W
zk~K1^c=jso>$@Z&c@`-w9dkv(3vqFf0*|yRmuQXz=9gAn57JgcQQdY}dlX2srRnI!
z4GG4c7%gut@%sA-h<Xd(o%BFvU|w)<N0A|2mEPYdoV45^mB_#M=#jn;ku-90rrk@H
zo7O?oVCf%PJHJp)2^hrSnOr2|x<H#?Ha0d&kSEFWz$p}em`Dq1PUZzGQA>r3sNgU^
zx$aKRY+i!2mFIlanQ?P5SPDfJN%|FYRqm@O@rr^<)S>v6ijb7IVFHAvZdM>Q9N(k5
zY8g3;XGWZsmnT`DCk>#9RH&cKA$j^>qHc9-mi$WZY)~@gRhr6I9xY>a8pbH4xVZ9H
zE$=y3Qcb$)G|y7yYK)mAH=Q90PMK5yCA`JTnM?z%TME*4>Ie<(J^@sVxstV=(<3Gc
zB)+&H^sXp!Wbs_fGe*i4oJ<|Ry@z;P`I`(eB{|2Dy!ug{>`6vYiKT;U8~}d0fFz*8
z=U#W5!t?Hsh8ka-c0|LPRv9)JG0i{3u$aoNZxLAIQJ(A~lR!eue3KmH=bneF)a>GK
z8W>3dF89icX;%#+b>+4Ig%0&=#p0XVl>Ip|4RA~EHBDw8N!l*NkX1tmmZ!_nKlOux
z;!ptL_KD%aB@fDT8!>N^?Ai8RI$ZHOGaNtb8E*U@qR9``NA&}FD<dCXUL%?S=?pDQ
zBQ1To)c@CZ!WiN9LqH--ES*_d%L}cMIpDML!T_=Iz<~neg^c_Ht176Ehq5dlI}~<x
zn|W`gw_o}vf8_67Eg~@`*(++A2_#Om+K^^sNry^Ah3JDcm64CiDNsRVnc~)tf|e9k
z{3<6j3^S9d9;UXv&3QSD^Lc1|MCe8>8r<^8E!FcPCU~<HdS}Rwnrn~<)L0T%Vkwj{
z*^<xbQj(wF#F<=&>BSx$40e2rk(E@BcxnjRsrd(J0bJZ#u?4D2PiR9uqe$kdWU4OL
z;#bP-<hRTQJSuD+RC{o?6L4XgET=fXX}Q+V{9NS2=_M0MKT|q4t<%&s&^u{rzM~-y
z=BE|WKlz<o5_&Bk-aEVH7hfA7tl+az9ndO^K{YD(nAPS>&yp+eaN<KkF*<y>OvsNS
z9<n?obPG0!wYQ_?9ooJ=%iCwgF2CK51a!9}&Smwy*xTLL<mQ?tg;UJ(puy7$d2)K&
zxR3Cm=WV7J$6f$f%x97U$MWu8eig<GW8F@Dc)$DhUoQHfq4^C1p78BUf37m#9-(0+
z1Rr>VgEtH+BOFZr=v&%;lvm7_826JmW)A2jedfcwWpmx@50NQdxale7u)lZkNHyrU
z)fApAom6mtRA@kJhqIw#u}HSunTaLm4z|jyJY{*wP_;{L`6FTM&=g0(vY;8#a`+zt
zjZ;t4Y)zDCG7<PmVRnyOw9dS&uw~cs<>*WwvLKZfr<^)ePHxtcS}Nsa%mrykF%e2J
zpCv6;NMbE=UdfS$uNBpbS7iv}#4SV6lg}gO@k*>PswT4fhi;_)$+@u4A<sk3WK!~`
z%f40<<(gPs^TDRovlDy7hE>LCKRvbAaKxPjgDy!Q0xB0zP{#8itM|ra>^*3N){+_*
zT0TAd^^r7w`w(HnQlMa!&?XK)@`maQKQc<u2d%O!HGdV5q{xsUz>Rx$tS<pOK}OI>
z3Qq}{5WC{icyOg=H|Qie_rz-AVl+7!k?CgPKl$Efu7Fq<!uv6TFf9Y~LPJUb2G<N&
z4LYJ9)teCq`J$ZklkeXUv3sf&e5M<7o0AC0D0fP)`Mp~V2`qSxazZbqP}6Lb!CK0q
zTqHa5Q60;>f>gmr;v|$_)juIO%k%K{g_Mpd01*bQZ-HcYa3gJ7?w2pA3ogsn9i_Bq
z9BD8SG{7{y%!XbY&?v#k3rz!c0Od=E%i&fe3WWsMf576{g3fNyBUzUWjgPa`I`~L#
z>4U~AAN8zg1^dCYgp9jAzRCmfnI}jIjoSWSOADOu9gO;z+lfi3p-9U}4hSZlhN@Y8
zALi#sYk7f>G=!N$r#h1Zm-l6Pi1j^Mo@;sDk_uw!UEU~JN@lSlfHSwqMECp^W%KrB
z7TW}9nF38iRVYHWGFdV1ui7TRP3TUYP@(|hf~9hvJo<KuF_&&W%|d8Q0oAGDL!53)
z?}Ick`%UXYBHQFQMPm@!{7X<@bMMAG46>>^WpW>0qK7_HgmxFtPR-8m=i3Nmxt$XT
zfF5}Kk>9%F;T2v*)1Avd^hV54w~3<Gq8A%P0e}_9CyQ?xZ*U$`E|7OY`zW6@A0?no
z;ZYGWxYSH~P`sOScA!qfj!d`xNSAB9&?6QNPHI~89C&hI7(iY(g`~Yq@gQj|A6J1X
zUWF1Q?`$9U&}hxFbKALcqeN>=d2%YF8l}1&XXSmEx}|_V=*yrS<s3O)R-*y#{|G%o
zQ1xnz5}Y<l^R3-nPV2YbY_29fC6Qy432IM0jHI2?_g)ho=#~wYHM4`5B!F6VxzWh1
zxvF4Yq2%<nd*27?%bVanAE{oQDSrzhgSp#`?dLq?xaY_GRh3U^ACDfRU1#ey@64qr
z+st{TCuFT1K!gUiz{-X>?JE<iT)Lo`4P~9Na`My2Z?DvVn$Z6f{}5JMV;gO%v#e6m
zl28x*ksDF}(oPp`X2rD*V5?rvPx3v<hQ~K{Ev=g@w3%(P;caK@dz<`76}(e~<cjNs
z7Yac!$Z_Z|lD`HBR{xa8cl_!)#v{h46CVI8F7=t@V?`%M=&~Nw^6_F4Ls*qCfE-6O
z$lnI369wo4SCFFP*_qvkJ$Q<G79lA!&>c}E%*((Pp!Zz;XaPYv=?HVgnHraG)BVgh
zVVs=Hba_OsWp)+PAwM%wv@l_IONxV!%m`Rj?kmTZM;-em@}>DFwR)e_k(%;n4YOw*
zCLdz%BqWv_%mLZt{ukeA&a;j`LuM+UMw3^SUrd40Qp^YQ%68~I7JSq-C2lX4SV6rQ
z+vK{i5d4*}O`)vlB2yo#Rklz*RCYkoAUL&ErZ*+k!#nY%%2SixmCN0|RFY*p&y=c#
zIM2*{?#3tM_B^Uv{xDBw<=bU_nq`ZzWZZaMjU7)y@f+);q3*It?>qEQzCW4gPkIn?
zBPuCWdOVVHYfJ;xj}|bXPbv>6pafuHjTnT}CDU1tEb8<gB@Wz1u&3foqAbg2$$DP9
ziaBW<8l{}NpfjRGANd+|qZZ^aPthPMrjJN(uo;e#tfkjBpiAbfmY4hh*5)(pG;Us8
z9=kkLdSnC}rcH<S>QD^(D=@8>ksat^(qQGjHKf#m%*(}0$f5#so^^Q+b|^;0ur`tX
zYdpN9<rnLO((L&kVw5NwaY(28=unSXoQ3HnRq`j{G<5wZcdUl8PEavpEs5&?%VLUi
zZpUeNv4z*m^QeOtJ+WA>F1SoEBX`GtRX`6vHD!LEk*>ad#VV3sXVowS<7Q+#{OI)*
z<?-w5Q1ueU!@G^A%wLT3k}u3Jbe6d7zON`5=C9z!=|R@y{oVSK9N@sVVsk6Zw<^=N
z%j8*U{`?Oqyi1;)6hV!9)G-q!>4~X7iaJ~9Vj*T1-<ZwanVUqTT8upNB&9>?U580W
zE;6I=8imSrU^?JOX#vZqkbC=LZ&b(0GZ1e$gg2T-ddF!=ou*Ax7TMk%G<o(4)<%W=
z`TckFD}#RiiysR0$DgL(9Q2!OzdMNfyzK`)&;7ACus5(bus5(b@cA3if8y!v^KZXz
zxHqsjus5(bus5(bus5(bus5(bus5(bus5(bus5(bus3kDf%11|`)|&U_O$DJ1A7Bc
z+d%#Xb>VlZ3$35Fv)y=aU~k}o8~B)SA`iaEgWEmu9kE;O4eSkkkqzu`V_)P!?`!T2
z><#P<><#P<><zrr2A=sN+B17U?ioGpmQQOuyW5u;@MQ*>-Q4Ew4eSkk$OiUzpAUJE
zS6ySjL9W{D)xPD)$>-PO9$h`fhxw=f(*|@Sda2=uU2tD#Z(wg=Z(wg=Z(wg=Z(wiW
z{tcv`7RwLZ<(0GgQzZGTG9{JS@d_v=WA{(E9rgzH2KEN_2KEN_2KEN_2Cm$|{<-bS
z?QZYAfxUsffxUsffxUsffxUsffxUs3-N04<GIW*i%Pw*cwm0z94eYPDr>_6An$zzN
zO20DbH^2Sa3w~u#{XIzOKc;eY{qm?+{_nc<W%9+pR6hIq<uTXaft8*$eb6Isw=Tcx
zE9VWn{tftc9rBt^o|>KlJ&b&oCsUtB`SMJ7DdoO7Y=x6Yd0D%<7W7?;R+K(Uxo)Vu
zm-FiAT2$`;yzE5>gYB=&U^%xfWvHSz(`lUoqYWKxAg!OC03XrQAPa=EeSVDj+d@UM
zpnMZxb8@LLIUZTEwK|c=;3^eK{>D@<Q{ZrnR1zW@z@%!X<z(5c45?%)X%l4BluTN=
zgyzRl7tmQfBC5zlSWRS=p?t~*tqP@QZr$u^&<5eW2y>QRTABT74@zsTR;s>asYK#L
zq0bUm-ra|4H19SeLmawGxeDF6hIBEy8C`SmPr3hqk@wT<+0S6TstU4eo`t4a_*kSb
zINwek&NM87o(q~uYNAat!ASG+JNpFvonwAS@f$e_=)Gn>9AN76>9kW!jM375RGS}g
zmg%}VfSq803F53=I?Q8GE-Df7gD`&wOU_ANKg;(OA)}H=XF~dJxT=})o#b$9J||^|
zv}M0#!Qlvsfb0b>!Y&QLUOtGlT4x8+uzdJVku6GmC<z;7P?}-Z`dfbFbx)1w<?|k|
zQ2YP0_ihQ2EX#2iuo<oa8<0rxl9bk<FDyVyAwIBMg})!+mX%SD?w+|_GnL_PW^V2g
znN>aaoC6FrhQw%d2#)3>cxF6VOc-!IkluA@U)rFI;DhD|(xG9iZt3M9n|82&C$v84
z+>$Qq-DA)1_T0WX?Dmuf(&>)5l|Sy!ZcVnTsTG;X4`Ry*P|l-PyHcX~mAozN8#R1$
z83}g|IsZ%_ey(#p<D#&cq7p=IXaR3|C2|(DfXaX$D?<C8i01o}*9ZE%ee%}~U#ouq
zP2hQc+E8$t`l*SJn@sMTpkTf2F>)Ll?gZED@O|WYgH%Zyz2B0Qhw?M?KU0m5@b|Ee
z?&xq_{KInLbjB3dUqAe5O#4Y;I}*$^;cxnT0>3Bldjel|0_T4=c>iRFUyl5z8LE+h
ztJhBs`Tev1?82vAx67N*pM3JukAB|KPe9tBEEHODbo{lImy%h25kXCFzgSKaofO@6
z51~)|T)>Y+xJ{94T()t)2{_mL2BUJC&$0Rb(_JGkTKy&FI-@SBVbOXKq5MnTJf^#>
zzg5Gjqb&4H9$87vB8?X9Qj}*;u$oY}M-I8_l{lS}Y)c+-8K4Z2l|EhtKa^eX??X-R
z+NR;|mi!eudPen?LWBOKRqh~9%%5fc!9$-x{%x!#hSN7k`FNz~fJ0%$q;o6(lMd*s
zXFWb0zBk*HpFCsomb<83(!yc<WLc?hhzQfslg2Ohk<hQ(R(^qxo`Bwzr+cg$$2T_J
z;0IoE_x*8rpnfB5Tea1!J@-JEmM-uPvbh@I<?iBv_iG5Z6l`b0_u1$To^tci0I5L=
zk@_`!YhZw_ui5{>)z2J&e;@m&TRyf8w^JdKe<-Y8)J_-v_J5|m`MztFvGvCeCm_x+
z=`6QZjQVx4yyaBTB{kQ5rK8bs8QB5yswWx7i%wPd&xiZpm;Wli`2_0oQZkuLZm1CU
z$k3Bfq%P{aSp+FN{6x+sxhc({4{?HRL|vUREw6%Vu`RqQ^W@K&XsHjq_$oXl28$O1
z>hQ$oa*6iJkj8iXNh2?dY&D$LhEZCeNF<6}bqT1I+$FIoEW(U)O}PZSqIi55&S~hf
zc|wz`qaGM3zy=%L&gL4HCnGc(nYpb5bDRMOxhOFb7QM12DqKDbjVo^gLd_w9+Ff~<
zkuDO}MjpF2iLRE4es^eimoIuj1FE<(bx}_Sg~`0e)>zg6BO<0|UL$ZNbCzTA2*e^=
z-87mdDurztVG8e7J4%SXO*X4^j6_P9uWb}Fj1H+<NCv7!?APm)c7>bCwuR(sKW+R;
ztko+EAfnbei)x&zy(Yep9I~LaGmzHYI>$y#nAHont{fm`!J}b=)f1wLh;ZAmQy$o?
zU`Ez;vF07N(`s~<SuCq%JVQz37!u@R4|;9~=(yb|+h`Dm*jgA*JgBk`Yu;fSCswne
zmDR0erG(JOLYUwq(qabaCi8DAkh&!H&;xC4GZ602iiu>6IY_K^i}Ge>Wo05u*G%Nt
zCAtC16Q9BaOWVa|+MTKeOR~;l-KwRVR}d0K4mu5ei!vu}7T|&gG%zJiW2+ggyg-!L
zn>57x-F2pvsEX9<RVmAC6^&jK=%EIl7wJNopo<0RS#m9NZ4x29QnkE@oGdLWuqSa2
z3(C?`m<cX5F%e9v!i2V(_+-`pT~a0+>qAIdx~1z^dhth?po|xVi3QA8V|?WXo>9=9
zRz8mT$ANs9YiimKzv%!eCc?076_N2Klo~p#g=jr_J@FC~HR3)JJF?v!x<jiG<J`U}
z2%~Cw8)>MGXj>95+Qh|6O(w-ElukSlAEl_Jp^lnxA8i$oW|{J;xI<I3ufVQPVxK_f
zh9^V>ZPBhO<a9UtMgEc#@Y^f_*p;aO{iEU2&&iM83=g+<L|>cVJ*a+D(04`}CC4-;
z{r-a2nnBNm`?bEBx@HYzG7h*pVLC8tN=y#AD-Sd$TrmnL<~P+_uC+<V#R9XCr4$k2
z^%7@OiD?=$lcE=|-eLvZR2#f#`_6JcOZ!5ibsbYotI=vmlPediC?oe>lE?EeE~X1N
zm9EdbJzq)3AtRjb#`Z6JunEeckfpx)l%-R`DAWP?T+QVVSl3>7ue9<WPS1}|X58qh
z6YC!r@j>&{a(P@aPA4Gh`I6~HzrTR`1CNHE@XsFoM^0s#UX4EVBuGBb6x4T?XDS~|
z)@1xmF@1LB^(Mn#n;&@sc9-+&&vFJe{*jyb>`=oejeiE)$CrK{%g<K&JX2=*nv06*
z&4dJ;{a(KE1o+{V`#|@mzU&6^>i0|TrQc)2b>(mFX5{ypo(>jpFR{>v-~*ecCiiS_
z%I=x%Dzg7@i@oW$AZwTWAwO=Oj>o?x5~hB@p5ft$cY|%N`oL4RB)ViV!@;Dotjf<l
z`52evju|(ubOy%<$=3J$sp(S6hjpAl!8|!x>svm^P7=O;5gMx>rOPZZO><#dI8(C6
z)qKuTMV`PNL~-j_yhZ}|PC@A+yS3dU)k|TUBJHa6V_|;e2;gb>esrIkc{d!&*7o!~
z?U8u2;Uw=;q1IJ3okzLXofZPaeFE(u2jEr1d8p!W1Mb)YZ(w)A*CsqodA>Idx%GaB
z@9eQqC7i$iwXn^}fLW3cnPxveemnAXFP4x$igMu!V|~E6+jpY+R|fqb|M)-8{>mUf
zY8sxCxcGh&Bzn49lt?eoKqLkA%7Js!t%9573b!fJImvZ4wkRt@w|&Ijw&@+O!BSx%
zW|oNz;A6P$R%Cz`NJB-nJ~TFuA*|*kF5I0P7R*OTTdD%iCRt!0YL!XV7;jj1rXpGB
zszCB-APbUG3%oApArn;dl2|7Jt=Y;9ur&@`O6(3Z?;<lW+ed2@u4uFJ>X5}?MrO{i
zzRsE%^__PU%5#auyh>t_Dp#tA6Anqhc)L&$LumlbJG|*qNt>3Ktz6^`Z(K$<^1`ik
z8B0yG$n|qc%=BlOIxA>GS_LAR{lr@kXS5PcR{Ca<;=)!X$<<O8&J&`!7>Zh1sT4k6
zH%sd9?91C%0|4`6#L1hFg-(E`r<a=eY5^;V6kuJIF@deDNt}f18W=^It(8X*XHEn%
zLZ`|MY3*e;mfMQ(Wm{gB%387vDZz{48>Bef)42Xhp<@~ZMG<H1BEkjAZ}lY+3;7zE
z+S(!#%l6eRny{oPNntgE#F8<m7h2Zajz=FiPAOKL#adY)A@Y^o&92TgW@bH*Ownb%
z<*Bo)ep6T~9~aiCO)%Gy`|{Xwv0ck#(m!0dq7N(5jt+SidhEd}$nj8rp}>`h*>D0f
zC{KeTnc&GbK;{iNqD?PUzH(=;Sb;u@V!E!*Ir7;c`b9(?OXdkZAT+W(`3`f=nL@PK
z+uS%`7sF}gj4%a_6O~>?XlkjankcV{nWQ=r8_Y(2$sz|OXjDs^_<nlnvy(fbyImcL
zW8OrPC~RO7_NR@?lT0_UBT8f~7V!fjkBL7m^2|{C{P4I31(wLlR~}a)60Ar_R(oi}
zm33xklnR~kK<wgtA}Wnaaq_Y9CRY2_#YJ>Z^~IC1oKIuxprvA%Z3EFrt+^n5kS$z|
zNXX(6P=Nt@2$nj&f`u7OA-`l{1FfW3U!>-`ROS{O6s0Aya&d)T&UozQVvil={>yNo
z?OD{7r=d%F8@3LDYxP7{++VEh;L0mrvm$3oFSM*TXC$Kb_M1jnjBu`CE6NygSTiSB
zMi)_-jly2$G7bvNB{qQh#;T4BaExObdO<=H#W#pCl~PP1qWCHc$Gl|{@F|2wNG@Cw
zR%we=ERIyoa%Lo=R%|H}_N5`{MJxB?%mK?3NU4Q1B^8pnIL=pAg+@}ZNCqm@>6$Um
zNEB7tdqr*}Ly(8=iDgFHY7Vr<i&#~xg*|CiR93pm%Oekr$-GV>1F8}fSnq0sGjpRB
zOO@WI4VvE4wNgk6Lwu2WQmCy3_X@7Ga+b)HV#i|jT#H5|ORohBVJlUcT0%m!+(H&p
zohaEsI1?7gC|jeySWpq)Lk>)pR;DGFIm3R6v@$50Dl9V>=!!OC1Co2<Qb@Z|ql8kJ
z-`1&M1`c!Bl*!T{Ud^X~DVgeajFEMb-#!Mx^nCIfrke7HgW~>Dj2Guuz~39XS4zFk
zo(%5B$s+lKK^oR^GffzbCn7Y)FGRo)Ibf9fy1&(~m9%aV`U#^pQos!)GbFMZseCce
z_PwR%d^D??yXy#)c@+svBzKU{k9mspO(HoWE+v9(2vvIyBw=occ?#E)R9{}oiUk@_
z#0J{Hlox>!hNqS0HSJn0q=4gfHH=n!d$0iKJy=3VDKf%f0rW~@mCM5FfQpQ@hLF(2
zLhyBfp2FR-8MYM39QR-qa%~t0;*W}+6w6#0>b~2~h?}=tfhm|F_w|dgq?G2YWVO=d
z!ns_zfGGh(PeN7Nn)VYO>|hEm>KAopn7_f&#UInu%x18OVkdl#4AaGZJ&?|puGK;%
zB;{*kL_k+1Dj)P7T`8HPiv=Dt7MbqR+9yJb&A=2kr{`@j);EmO6I`_E%I8lO=8`xy
zH^y3XzAn5$OT-|u#quM}>oQ-zyj2>S3sNn+l2$5Y<vmh2v7j1nq)E?7kg*i2k(e+}
z^BldTO7=rT(W^O?v?S4Kf)f^DL{a>F#HUDSyk_PLSy2mXV{2}qn{`#vq_9?awka48
zRB!Q0D3>-onY7Z=B`XU3w}v#oV>%=G=<e8$B;sgd&8Uze$;CXmkqy6z=RN|kKV7yL
z$@7bI19L*Ly>kJ@s+T2+0uYhvc(zn0yQMAj$gJ`W=uWMNZpu@#*<XBoc3Eq@Rw1{^
zNSrmX^jj9rSid^xR1$HLAma(*0^SRUY%VDI3ML}2@)~U>7K>Il<tZabx@$&AdKo9u
zOz4ieoEq3KQgXv3rfPfhNeD@(!L+hQFNZkST8$-gMr#OqaW7t^xy~&d59%Xuz<PF~
z$#^Qo)0j-<i%A|~!iU<U6+}#tWW0fDss~M*^a@f$qU8((T3vFM;#r=&lXitbo(NI6
z&I9&IS<@wl+7T1Utdtz$NKT%?@#*su#;`v}rjRvVl}L!VDPEH_zwrs8aDWF$ix7kG
z$$jq04er=LjgFXFnp9UyO{gn?lTR(E`kI9j;i0I}tXEVPel%z-`Mv6uW+&57@=WXx
zo>XURGvq$yf*MH$W`nPN?5mwV6cWo<Q{K8wW0kg+!Tb<9dO-|EJCu+VQkRx;;|Uvu
z$lwsCe(S9Zk6;%&QfbfGI$J@xhU)m(Q#l#M&Qhiw1QAOwQ!eVnLRzxneo><XLYp2`
z<Q_7kKUO|_*kyH~9vN+`5-nY*AUj^?A9!!&L5TRz9Oc5DNU)AsnzA}d)Wra_*0Cg0
z?OIBeCG0Foh6+N7l8z^gIKu^&ENbFbu~I})CcCo}37l=hOQLmEXstmi>KZ*UQwVK^
zu^*H9nk77ue9nr!i%XCy1<$)CL<I|`-|3|rVctb7S7>QjBP2Z}^4m2+#N)%=4Aj<|
z`WzGx#`wB8zc6_@;sN22M5G0_=!99DDpy$%P+ok^M%S^;2x6BiuUQ|H+jdEs>4HkC
zH-@cYq;dnjQxFbmfSGhVVbNB-0loCXb(1b&^V@)_O0ZqJBYOk&8>D+DB&s-NhS1|J
zsRzs{?mAWpBVkxrE;g1~1IlZ>tmPZUzFK4@841_hAC_7uRQLs}5iRXa%?#w2Thpg{
z{kAu0>=#p%iJGeEMF1~mwy&(yi&hy__G{qf>TSEKIU8*-tTSO=mzcIqFAif*f)WMZ
zW0jh0H46~E`ccFE<&sPp$SdChwGv{uiZXjlF-^3TQg4e3753bNktm=f1+wZ9T;zvw
zWPiV}J8<CHeS^M9xsi4*vMyV4c6ScShc38~x;&HB_>fV@FgCXxF(%ec(*gM=h%cAS
z=a?;7Oy5NEB5c!m@4|2r7Z*ja%Y`d=$hEVwvfw1@iY8!+7DX(-*N1Vv3g%)G_)986
zDfRADw5!N+Tir(kfCD)EUi2Y<>tSoNtZ>bmbXoD28No{!+K9L!Qi?3|Qq~L?6|UQ(
z&mrF_HPBD8X{elIb7#LPA<W()l0k}=hzc|9-wN4bG2+JsQbpxrNUTTP@9ssa?x^o8
zX)bo;tJc+cP3sa0JSidF5;EdzlCXKY2xS0edKw0)zo!(`A_o%eVR95~UDHfd%eY@$
zMUDI==t^B@QzX5L5<+rH?a0JWb#87d1**8HDB<d65h+FXR;?FNkT6O08V&`H#7#xX
zLu!#BB<48ddOLSbLi#P&u4*%6%)~_wkaBy+Hd*DkWF(YRRy}=xEX~ST5E1wGqG+Hp
zrZ-KX6y@s7jJSSke3Gb_j&zmDvz5f7VHmJTHR+~^ueu2xkjndK#7&crjJ#%x!xOf+
zr3Tuvbgmd>Ho?vaseH&H(&_@R4Izx?@wJoSEX#=MGrq@y&~@*roZR0<TOV5;$mLFI
zwJeM##B#72oGLg>xQ32+U^N@?h<I0l`K*;fO?gRA7sOYEmz7$XUwk+g)&N?ef7k0+
z`Y?&JUd6RQmLzJ0oF&Lqk9|T(1E^1Oss~^Mec|bA)evg$$wSXVDeA~RYNm^0E6;Rt
zj#)X1Zwe_;+WeAwfYc#GS(9T^$XAnXTG1t>-~3w9IxM-3Y*1Y!6I$lYys~1u66H)n
zE5`$Frz_OD%5SnFNMKtz6UYcvP^wT}%!pQ%us~FKL%Cnams+Z{ta>dn;l(EsQFx*F
z(KSjVvQbGMudpzTksU6yCB9aDSnw0MRKmyAa50E#-SnpSMrJht;o_v(iC-32%&R3_
zw;tWKI=%wV83<zmMW%9=ky=YcEF&mKkQk+_4k<{Y3%*$Sy;eCq*3(4;qos$lOmn6V
z8ZR4ri3Q>LZ3*j@NQ=?D(#3Qs42|Aztq(aDNvU?1a>u!XcZ?nx+Hbt1FmjI&af;ip
zZmhxPL-Q_BewVf1L6ev$^3-I3^ULqz3b-Yeb0iE^siSz%RU%$Agc?;x6&J+C3cbtb
zZl!H?i;SC_KIA;1TzqP;vqlVPBC%ght1B%Aaj+83=}Lj+X<cZ2scwOMgDYu%9hIC)
zx0izwc-n;<L4dtUyyBd4JpuWZ1f3FDMw8e*gkM+ry>LfX6!9b645h*{StpE{iI^}F
zBV1Y_0jff{yHTK`Nn0>Z1j*Op)sj-5xKli)BokD9#iZsQLLVhM-*YyCpgdOV8^Pxl
zDFYl5w_bA9g_86@t~5jm_c~^!$gDh*QcGF6Lz%Xi+}lg48%QZ6jC&|x)BN*b?Q<1z
zofdP-U#<cu%#}+>sl5dgfMx5#D?t(%?6}v2G(jAy1uasZ0IL!xjl=47;gxJsY#6V4
zXBDGGyL9X7wM1s)!2;+k`I1%+jUlm|W*Csb`?4FS`67ahNUVj+AztRs2qzSAV=j`$
zyd4^5GbJRW9NUFQRfS<@t*|+}3F}%L*BU;Fj4Nvzf8nc4D&4T9RLfA{ULg>>(+M$7
z_J%9O0`p-m6Dl^`izV1$nMbSK+P0Aa8_Ark##YgK2fEdR&4$(TXo&FGQD`Mf6)C(>
zb!yMur`&}yf=l-56XY2it+7&GQzBuYLzE;w?84-?Ey|Eqi4-tM6d#IeD0K~!%nEE=
zto)=XDe9yY;{sSRXo&)F;6PAdPU7Ob<KBYvH=1AMdycw`k6n|;Jevw=ks4Yw;?+{Z
z1UHt4$(Kg7RF#CTBj-gR))68Fa^+?TSRrMJ^q2zUZ0#6zvRdxSiMA|+>r}z&lS}*N
zadQo9n`G%6@)yA_+Y#jO#rPA0!IDhm*tay_zC|ZBhtq{!>|W`2z|u)6U8RV6@GpT@
zNIxBUAqfS#5WYhFt<Sg`>&R@WF&k;EceAdnQ8g{3qZeNjPK5j>%0fcPUI1Ryf-pAP
zFBM-xe%adeRlLXz&;rLf@J!g0rf}ob$`+H-R`gy_lzz!(1zGTV(M9V@?um$07U+sN
zk#eQYPZ`xHNnPa?w3!I9t`u<MhyagpzZ)(bj8r7nGRk<(0*(d9u(^-*zR}Fq>YAx;
z1c_6k^F6nA2shqVsDp$}Vh+BJ6$aYsY6}1XU_OV9Gr-p6&0)UTiL=(E5}g$gBN*Y%
z2~1bU#7eFr%38&hs4BLo7n4_14Li8<cj=K4`<Esq$rBN=-ps3Li&dX+Rc0fPjaw`K
z_K&nHBrqL@tD<I6HRCh7s&;~ZtT`b(Urkz)VzKsCL@^ad#$lKg^0vqaNkzcNdnYM}
zuIE6r6jL~Tu0$C^zja(7T_hBx3?Q*eIN8~0R8_&@>sBU*n5dhi9JEP_H-^?MpFp{z
zLfFBsM~_~Q20(&_s}rI2T(Y9c_HGXatt9jji6rF_M#PhaauLkg+S*WBYMtjJn<>{s
zakh^l^017LOLM}4<BYB3Hd*eq1f*9>jRCN`n9f&J@FY%UxirV#vj+562K~o>{u})X
z^k4qv-~Y?{n}h!A+20*h?4n^rB&ZoEDX=6Lb?WH~4R~=7cZQNg>;rYO=eE%eS1W|q
zPWM!`CTF=`;test)?EHL22}u7znL}ERy#p1a&R6A(}+R1_>ZiDyo`eV8OtThd8>L1
z*nT&q5{`Rd6%W=^eYflNJvV*Eb8P%@DniZTi4*0hAHitN(S=4SS9oyWVkvu7&(#dT
zt<{#^DMK2BZzEe*SYU!zSJ+@CW^@e!iL(KL4S)~ZLKmNLRE(gMG^(X_yh_$=bi1C4
z*vSOL?zaWNNBt4I*%aWJs>>QcHn^p0qzzGE!)<&>m>Cev@Qj+3^^6tjLt@cm>7-~_
zOU>I>M4#freUKK&Xd564J@#P9b8OXVT{Kc4JXN=FtJnZ8m{B+glAjLVeR|$Db2Al}
zbTwOB-Dd-oDR@b|^ddk7@4`<{8R{{y-FY<o#W9>ydy^drSHL;!fp1fhAi^x7Yc_Q5
znjj5liy2)dh(I{5uu<m98MT^e%@sMGv;&4p5N;M1yi!fkQuBc)>jDix{SNQQwL@=o
zk?+=u>_l>bBp?8s9G?t<@q&c`ywQ8tT2QlHtq77W+$dYS2f~|fl#O7t%oZVyDD!jG
zQXu(2Kq)mE$N{P&4O!GgG{_8sqD<u23itJ%s@LsRJ3=)xDsF%SaG#*%L)8FL5rCf|
z+6<b41^~Vq36zZR1ZK=*53a|xbL#S3B^+N}=0O~2f>w3S)Q%QNBuxtKUoOBp_mjX=
zD_B5a$=Jf3WMqe>AmJ(8ICC%SR#4bTu5|+|P$5bJkD#gUjzBss%?=W0By?fTQ4=-x
zVH%yX5yd(0S+=dZU^1TLP~-?HI<9$M>|ZiwyN7Y}ZD}r_sg82qLA3#JJK@JYOF=VD
zse}p0&K5tQO}5rhMwDm@5pf$=A=0A;UEgyb6B4c{0g_j|X2n1gq(HcP>;gd$?u8!g
z%Ild1{Q!~21W(u~^zyx8N30t$g)IrGo_k>%!0hmj=qNj}BWUVJO66sNSLA~5n1&T9
z3w1=Hq$u!+DxUT6pG<Uh&9!C(OVeVW68Y~WY&_so;GA%rf$Mn&r4-D>jHcnQLoPq?
z0MZE-jIR@YJ=Z;U`Kop45I@E3&Q&`WM{SVB<sr<1T$1CcwrFP^@gQ3&7+geVDeEb&
z|DoI}E0rawl=Z6;Lz_>)v&rJ``p}JBTKT<Ii3{>dV?@Y?<$Gf;k0&T<W$q(A6?;MQ
zTo}Plk&a#gMAV2Zsxb>UwV`h4fiZbSWWh_4X3j<kMkBA<YRhKpP7!MF(m)fh@*q?L
z6l9N7Txuq)4C-nRsx5$vh%`W#;T%W(jNWq$mwf>DO;2ssw~1X|oCHl=ad3++=($xL
z;z&+*!TC>!fViKfw%QQ3B0O1}R8~+hzyqXKh(S0Q2FVegZJ@?>8-j?K<5aNBXq=&t
zn`v5E=xPs=f<lBhnWwYfQ44sr#EWw<lQWvf6vbm*?)AbhK_mXc<&XTj=*k^Q4{FLp
zpH7Au2FRvx%7cq-WQ`fndWuVJuaex+6=+P0y;D176C%D=rde82MM6Cs=zt<}y#$cZ
zId*<W)N&_wWRn(2N>Z37-r*hDdMu!*N|M|Ik0{ZT?@m8bU0K&Lvp1|l#Jlieq!Vk`
za^P9s+R#DhS>5jCCaqQ4mX`I_EIKk!hT%kDC4nYP#JjqnqEh80P~GI4E`W&Kl>z`K
zqAF@utge!=Yy|O4RbI0+Z;a6;!QgEtO$slZLX|}+mM{TS(bCx@&<9azF(gw%IDz+i
ztZ?ZRT$VHG3J=(avWhr-+{X$AKWr-EqY^Sh1<`0zWMIXT(lrxYFBbqKx%*sLx+b-`
zY_`p?0p>Ij6K49?R&n0^4baun>t0!v<Lj!|*uy?hjtGQrQx8IiJm9L7r7)hT0&aZ_
zK>vUtd`xL8_MMp-efX)dcD`;t8=%q?ylPj*fTFZ$h^KI74tjm4&KRh!M9r+(hHk)8
zP8b;=923^|q8gc=;xdFj#PSc5)Iye=4BppJvB;YmH^%D0aEWL17)Lgz{>X=L<7{QX
z6Q|`g0+&41GcfQ?vK0Q00NsOvxim?-pywCTM0tqH4{#*)9yxHa1fFKiL3~kvL0vS6
zs)&|6ClN;^j+Wcn3cYRO9jn2h*dPHErll%sc1=octEeS~w8T~$zU4X1vB}Fi3ucrH
zFQ{Tn$*i5xPzN7TwSE^~T_Y?0C1K5uNGZTaIDsjQLs*Ulrw;amN4Sk1DyWKHgEoQw
zo#?U^XzY*}>2*fEC~FU-Ty~(((i~6ijbCdDNy#Z)DsGvYwUp5f+W;-51#keK<g;`}
zw}wcD@HF`lf+_$Hn{lNF)@1`85wO4t`}&|@(WJ09MX*AMb>-hK^Psi%588weMFb5n
z{~@MuD-qp(bqk@ib4kiHm`So(G^0}1xt=HUix5>+Tr*yByo}>sW;S=!0)mPp;aNW2
zpLLsaANv1mM0cJX!bO4NRg=)(!;314BPxQG6h@P}q@jg~l8Bm95D`h|p4&X1#;s4R
zTgMG(r^BTzO3D-x-%{;1qe{ME5r4jVs}&SL3Fp(Wf=$dYTb41+z=oE}YgVL4bBcpo
zkmmuG40mqj9SW<NnNg}j%r%}_jboum-iq;bIc7jF1<y&2^O%O)YI~48P;W9o#Er0m
zo$LATy(>0EfcOe#lTGE;H5haQ)Ge)L_3{ikkN|NAhXvXV)d9S~DM_Jk<7(U@iW$Xb
zylRKc*=ZEYozowmRG?JN7TAHFQ{7!md+f_(z$zs&pX2P5@f>GOj-h5ECXCDw?u{M_
zTzMJ{&IKw7nFkP2ZTc=ewJLpUw>(j1PN%SLs4jbhSW4;2(Ygpy<O9U~DITV+vZby&
z#aAIJ!Gv{#l_t2iVjI8=@BsDZNYe41kLR$7wA(4Z9e)@^6gw$cNkSu0CUYFS$!LzV
zlj9!j?%Ni^pjEv<K&SBBBCV0-w!v-T)~^A`DDa4+wC9YD;1g%Q`J|ugb@ENAG{W03
zNQUruBI_8%78Vy(H`w#RXMpy)bL1J@RFgvQJz-9Iu&Dxwh}1RndQ6mplBD7u4!QyA
zRkw2Y@t~FgxQbHN^eq)(2YYUIXE;)uo2?hRi88FO<5p&obzZ9p5=0?@Aui>qDuvbh
zoVZzEW!pd#{XPutidqch06iK8mSn<WTiH~9C<wM+dBjq%jbf~=<DP0&NDB^sKxNff
zJA%1t#ZZHYQyjKHMs+}_szlB5t{MZ!NoTQ@)RD4#W^kD>h(|c;^W*S*2o&cHrC38l
zY}IN@X|>p+1sEo)Wm+uD2oePd03&(nbzuwp;YsF}Xg(hV2ms7;U2*Pvxw)ch32+DS
zyz8wws>bJxD&B>o!FT^1krPvVr+1_jwDV4>7w$n)ouAqLy>c7xUYf3CCj{AXtHjLY
zW>vJ-%9>@}CaokaL<GXyK;EV5Q<g#(UknT6xirPD_JB#0RebVESPEfX(!4FSX1ZFM
z{27vqu!APaIlLhu9sN6YedtO>N!ai!F3~k28&CTuB86qR8E15+O0_4h^#FYqq$Ff&
zu>oe?h?WAC!Tl@Z>+tA=BKo_7)?XR)U;pvn&i=|E`m2V@i=NBV*E2PzJwC73qdjOe
z?0cE4&-H=FXQVO#f5{v*NRy{9EDW?Ym>#L(TF6J<D__pZ_0;5<n$sS$K8>2fS$mE0
zX};)j-Qh)62K?wzf8+LJcRqv%FCN@ber|pKp%~gC=zJXH^z-mxZv(W?%@4#5Um>@A
zuMB=FcW-0&RI<mSElEBDNMzA@{OV`vLu2)gyxs;nK%&oZH-Gg=Z}d5u*~Rkn>eca<
z3!d<b!&P6xykRu|!FWC}o!P~?O@`i$gGuoRVc@AjN3#BB@^@?wowN`0hKFd}mmQBc
znp7&HAt(9>d(we=U%qsRC+&~6a4OchaakR@X1FStT5)+yjSP?bQR&8h_iIhL$}Eg^
zK?IAXT{#tyuaxtD<t;k-$65jsVQ*u}ovfnE>yf+duGFf&uG^`07FDAm)3SYx*{`C{
zRQb-U)ep!%Fj=vD=CtauJaAaIx9XqFkPkgR&61!2YbsAkTM9VY99(}C^f`Jljr!WF
zJHDJh8krj0(Y!p!HyCXDdedW2xAGCw+JdufH-qMRXzhZYTws83Ab%Eq)CLcDr(^kK
zdckX+Zu8TFM@ReW{N109wH)|Il2@m`S5KqgQ+`gR$M?Agzs@h5z)N@M(X;i4f93}`
zdZ2Y9eR<(%4*qm;vC%rvE8~27?Zqx1&&KEQSoVSfj7N=MaJ??Q@NP|4%17|QVJ=5)
z<6LULjl4L{Z=31IH-7J(&2IY4!V}0^B)X$PwXfn|&>OeF&oZt-8@38tUAF<w90q2D
zrydQhY$QkUykfLjU#2r2s0rw6H_-dhb5-~Kv+{Jn$Fn`V>pa!-@smfc4Qck%6M25t
zKeKW?pYh!IG~a)q?{yrv^T9il`=y;^_>JA^Y5tJ)8>ciYdtm;u&!F6ZKjt<(IDJGT
z+-~mQy{3=yD;xb|9+F#|e6X8$J~0ox9!%19W8Xs!{wZmD*3R~Jx`|uvbYQ2(FX%72
zJu${n$&vVlEe@r3THn!h>)SO&s5iV|-T5Sco6)VN@9lQzb|}BKp|2Hv;SS&z3w`H|
zbBy4@`OK~d1`nmLPWZ6Fvx0YQ9o&^R65l;^2b=bfPsjVsCy=+6RQJx%ic_IMA79_`
zq{NUHJMagEJMO|A3EQ_sb%3{6l1ab)<1K&hDZg!YHll%FJN(<#bXNQD1GI|*Ih04?
ztI<6DtFC`({g>aRFK_sxZrtVWFUfaqgFYegdZ+VEFT7LC+_e6n>?hCm!EGN~Pd_{^
zP1dt`$$umHqb55(rK^G;wX@SZOy6FoHNW}~oPgg44?fW3qn6jt(KQPOd&{%=Oy_56
zz#e|#_nm;>775<+csxIyPxHLRbtP_4zP7RK+Wup#4fM?Yk*_~<xI5yh{;uy2)6s2z
z=1|c#PLJ(WkNOJ!UbpLBm#iG?I}i4<-E-u<(;8F*^xOog>b;Ml8Q*)hRrTDqP!+;$
zv<K^p<qzEda9TcIx{1$!%07QDT8EOB<UKcTv>d_r^s6p%_ikKk*~&gS;+wtMuKa6(
z+90xsbiX%FW_so~pY8YzxUIY`G-tTJ&k45p3y8<3#LvdQQ+s4K=CR{e@(2h^i7dCZ
z)iWHI8~;ZG&EVB^`mEgW1lze$2_$fisndi3Ux#4>SG+v{W3I>1{x9w9$sE_m^BnGp
zx;!?0sJP=XYKHEk?%@P~n}BcsjZIyj)-`*>{%OgZ9;}Q0-?z6q)ZIH(tNyTzH@Pm~
zHTFk4+vd>vU7P;4zHhnB@4aPDj6}Ed+m7&kjb$5u)n?vzIG=N(1HXE`q4Q+_2HUT(
zsKY<g^UkO9-~)WeE{z%UC-A3+hkN_{HsbWoqwDDo?)bIQO}6|ar<*mmWak@^im9$*
zl<~3Onj?wa#)CwDFvRWFHa^(*8?5!O4Ek^X_`lBn%Ai{xqg(%s;Fi-P(OD;b8nfx8
zK_`3u=pTk``rP?#@vB|D-h;P|_93G_XwR+fDQ<P&wW4{Sp66A=SL>f1d@qzy#(_p~
zcX)uj$K^qjU;3Zz1Rg(Dj}>QUmi=<P{s=P|P1qoVkk@+vr6<ID-A+3HP%StR0Ds>m
z=7DDPbjov=^NPnaeZuo`*HZ;eStqTT`vKVbKRr0z->LpjKc7Ehet1(L!P_#{)^V?w
z<XG{gAR6x$o<u49()oXEGk747cwsB*`3dV6raCKqe)#7rhY8<mHu%2;wc%QN+dAqZ
zx>tOvudZJ0LFGBS;s*K8gx<P|!EAtj=0UVSCB~3o2O~U!VfB&danGvYtU_rnNF+@P
z59O)y>9EZYCxFD$onI@Q=9#e9y0q$#H;{g<O-kobr%dF{xwV3qR^3uE9H83gp`gki
zWzDXokE+l3kDBkJNV^A3AJED5hVq6a8RGC=K3j*67=Of6M&GRk$@q@sZD(c1ul;xB
z5mKAxEMkoUuns$7%%|3`@A#gbyf*i1)$iE>x$qn^3?2zF2shufQm^M;7fntLt5wZT
zz5Z6cor<k>pN87%>;?z}U?ww{WB0rhJY9Il{0lUnn*Rykqmm=>I>%FeIr^81i_@dz
z!<|O^5qvag;gJKvOYt9=<bCD1ZN(^CD?Ou20|)>kd2RMwhFZd&rv*R{;I@f-s#f%y
z{;4N`Zg<5kv$gJaMp=WXQ@l3(K!yhTK&!V}{;6}rFT1y%&ac@IJArS1T)%yQ$2a5;
z81eBYt`+yUI4VbwJ65J<*CTx?`}x5>+5wtF;x9YhFWt`fHGALfJa;;;xPMNc^gjE8
z%||D>I*@N`d^J?&e$0SBW{`Y3KQl&0b%Ix|WwyRy-`MJmfoe#c4*xcz)5cHh?@tVr
z0DgKPSomd)-~{)G!#NM-_ht~)2dz)W_$sY$*MsY0Xfhx3UAP;*D?o$2%l|DtZyVB$
zJb24|ub9Xrd7g+PsvQtsRlk{>06s0u)U(@l8Od?2yU3m@y6{WXU_C&o0Ef@kh2!fE
z;dLkCewR%z$vZ7Z4RQrIfQSFth?L_LUkm==(Z2Q|KHBZePWj6kf#X8`7mlcKdaMZT
zc~sm>5?%nlGklo*xM<te2K=~@f8;=a-?9DR9gU-n;E#M(-Za#m4XdU|OL8&GU7cGT
z7kH%vs2<^uer%vRfCCtILHf3<-}ECV@O%0nIo3bYRGtO8a<ndj{E?ga9s18Yfxq-o
zd-Jy5Z1;{w<aeZZoZ`DZe&Ahs_ZENs;sma4_Z1foKRB8;(6OF><c}`a#MU2tW`D$J
zc8<2bYv|EW!JXTYcg^tupFe8Y5A6Tf^3Oeiw?DgYH~blA`<HI(XB_T#OaX2vfQJe1
zD~IB*glte-k9t2<bkyNn#eezszV>YSFyo?s+<)H@ufJDjZ74PP{fD934UW@}XcPSc
z>jdbp4Em4%{5ScFkpBI@tiL(vzn}fxLFIrWiHMH5X64`$l|q&5NYu{Uj7F73cXg(p
z*Ud$Nz$LQW*4fR_r81$dMg<ed7TqW^iLJ$B1dC!$uIsK@cf2VhJ&-OKFc6pSV^s!o
z>;v!JfJ}u0t57f&+1dwp92052H-^#p;>cRRUQ;WbmB7<xLg~Z}*Z|!$Q<)mLj|a^L
z;Uc@~a$(XPa{E-4!hS}D^74%4csyXLDk8U;2vvdt0G~AA?yw7NkcWO80~-Jzjvp=&
z*+_!y_~J)U>Vo1ZQnhVBN+*4yZ7g09UMKq(Pm?}{%gQoatxL}<9>&SSBydHZu;7vT
zJ;Pm9kd&mb&IAdYPz80Vr@OdLbG3rhFol%b3&4xsm<=!w)e<inzN%ao)3%ggg7RC_
z8)haU7mmf9uM!HAt^iI74YSZ=50)s$dpXfaf$+g)7=|ITC^noZQV5c70DC0330f-Z
z;-d;#y1eIJ|Dgc&D|kvMOevTu6cLfMW>I{Kf8J;x7R?yDxEEde?BtHbXd{PDISQ{Q
z1-7;5S~RmWdNCF(fDeETIakgo)gZ?46-exew4&4@4}?pM0lJ1TBBChkZ!IyiT43ai
zS`Dd|!koFNR@BT~xx$m7EOE_2(}pJf9zyLv4A5?P8wod#=(659W-A8jF$(0$q2uC`
zXb=pCEGj9pikca^t(%O1_2<I1NMz$|m2i%x+&U0Hf(=$pNF%Bigfpj?ypXbYb3Kzy
zut=_VuBS~Ox)e1BW}>UGnsYK-M6@bB&?xfErRFtU@?23#$`HHbJ2KakPPB6cY)N+b
zHJA+$)ob&C9UyLqa4?O^b5)LqDnth!k?&ye=}@l9anT{7;{5MG`0NNTTuHKZB2tl{
zi8H$bWgm)}q>HK6Kr771SQ4NWhDDA-=TQ_mbQL6d3Q+}+1RPPa=wQujc2Sn7;}A4S
zu!aeln>0Z#Yn0z~Xe0&e%0c-@7gW>zb|e%L)ouu{D$9@>;1#YJS?-86(zOY4nqz3y
zMBcFD)NaDAa9ELzay<2DJoU>jOm`;oh_cg)5%De@HR%35_b~whVAKxcsL!cC^6?|y
z4R=ILDxR|(d&9alk!(T$`*h*P(N>@No!pVEFL9PqZ0`t)93|PNmKG+S?*?n<(f&ZX
z)$g^NB!Dc=alNCVhj>Vejl(k<hvPyGW<8<tXCq4F?P?9`vJLHWZ6-@brHPG~nN+gF
z3n5T`B1*)1RqB!~GTUM6-f`g4Whm$Jb-A!H30nZ&1Sc$oXppDy<*U}EL;M-U?p$?e
zvDc<j4?Vcr^v2aH?`g2D^UyB0Ed`)a)HC3z2NfQCR9QYWTN?qYsftE$ai7A;cfEmq
zgC%XC4sHuQ8S>whF#^Sa05F`;+V)+jkJ;IpHtLXA<kVtXEsjPMO{ilY)EdbUS@4pK
z&A<%YqgJOO24X#^hT7V#9rqwC2PjAoO*IoH<(Rq0lZrfmPlj_W=2je8ie@=#1o0rm
z&}}{7o+)cOxHn8Bv$UaOJ+dLs;OM)Mvw%3aRMvD=A|bV&2^Hn>2dGZsVxTNzyO;x5
zF%vVIPo>(Y<H}(ZX&UAh>1t^!12Ak<^)(A?gomO=v%aC4HZ&U{3gJ<EsX(OQ`&Dk`
zj)xCL_KqRwkJIvt7XI?3&j(|yX9K=)HC0NMJTyf3c|rD0t|=r%OSx#mlCaV_Lve1D
zHQcRSI@|J=uF@Sbzsl6iG?xY}T6M4#Gv_6+VkdSawsXa$%{Ey}UAW_Q%L5;ySY#sp
zTyVW0hKsIbB_ZAbw7S5Wmer~Zj65RaWc@T==9OkAEvvR>wvoY=p;;YAZ1rnR(iA3v
zgd4X3S*Ik><envF3skf#Ne<PG!MY}tZCp8%728!$&DK)>+(fw*xOWP)EFl-_QQQxq
zi+X<He@)dH-<7rTB9-qFZxe(=8ej(9+g0dA9+>S07~$(<g-chksW%|2YrIQ$WN)Az
zUD|V&jo@Mso~fp3sZNm#0`e#orLeSjoB~hWfdc>|x%*sLx~6tIm;ntnPSwOZ*yVh6
zux1b>APm;k(x`h79Tz<CgEVe5G8NQ}mQ0`+4nf3KQRa$?5RJB!S4&EnB#8Pjj_k_`
z<Z-tG1Z=pvWZYs~;ZZCBX<X+)9_=E=`ZETqA(4jKG!w94YQT6VDPC_~Se5D2SA-Ri
z3>BQ>a$~eg7M#Q&MH4VZN=2A}tlVG*Hk3;`YPqf1HyU>|k5;)i`H+oS+3J9FbePkc
z)tM-p`G&SsU1l`3FoFFM&hjL&@PaZ47!_?{#K$%xxVk#-fsgX$;E4vM1SUa5WU~vB
z@D!RR9Rh@gvJDa-!6{tEHEU@jWsVN^e4n-}JeR;zAIA|6JR|3lc&iy@8c`EIVW_3|
zP6Dg9)igbS7Pq>-B7fB(8sH;bz(^uW$I!;hno1#N#QlHN^Z&eOBiF5fWBa<RZH`8B
z&*t$iTx}2R4x}-26iv~CA&8E$FvajBpB-kDorzj+Rv;O|Wk|#V9Kge7T)6^|xo6E}
zj7#EO1?IC>3hhmiQE3_pg;-bq?J^Hq0ss4zhEb~j-eqIRXYz`VZj?+qWd~}2usE7n
zG?!nYovV4uQfS>`MI^$#5(ox(fY{{#U$=*!87^57<twH%TomvL_aL;D12}+F%W9K{
zSqvx#Fw|$F*Y@1z`7~~QV%<7!K;DPdv1m{#KQl8r*^3R*+YlG)R%(E)skTBn%W*28
zaHeL|N)wyH>edJ=1^v(*a6tFN3Z9^k(evyBX2T_jxsQ)Z0xMCom>J=%{Bgml(WoE6
zzLl6IhTIExMC<5F*9)P_6;d_m1}II&p-7+*5^2EMvZEITDn-2j4&ZhwF$U6yf}&A1
zf?79n4ocu@?KO1k%ATVUCtYvG`6kRjj<tG#iqFK~#fp*kgBTH$aju&Ms0!he<D@NE
zNp8(L4|}o8twju)s1ydz5oVh3TCFzSt>@SzDJe{8uHo-M2I?i@5^8<3;nXi4a5s@U
zMgf3~%4;2mGpRlytR<{8ftu<7NM3Qt@^u@n7BD?Py$ne@Uh|r{CxEx-!3vPxZXW5H
zrlssuu=ND-|KPD6WnD@adq9ekSjh-xaTC<EX6Bj86UJe#$niYgl~Z@toJ?;SZ}p{W
zV-J90l#&P|65F0LI)cPNLd^Pglc(>Zs@;@IBfL$5WC(BNX_i!x!X)7Eg6h(8XjVt7
z#Z-{v-QYIzjLn0ZC+R4Q=yIb~7CSC@N!qF9N2}?cX*^i)P>y>LvtIV8GS8aLm^KhT
zEnR0#cAStMRed8`YNoVKlwoDHq*oZ;R!ip&pfMVVaCwDopTiWXVG=Xj^C&6NO_|gj
ziA$RgCr=CpOEQsT--7;uV&@W)La&R!SX<L=3{j0jT5$Y_n<VSCoOdwkJ%k!WoZ@if
z$jC>N0Ldy|v;0|V3>;^5GihJjM6PSNl{MIn5x?OE*eZsRMO0gT7hb1`b0j>?l)}bb
z85mcqYo?kJOszWK^SPCw`-<xGrdt&%8F2$`ZC>SYGk2xG;<CrZ9*V3cz#YK2M=2;t
z>4%t4<7n``?~dpOh*<B9-+-hqd*2eP>}wYq-|z&daEWRbRWm-Lqm4>o1?*hhrt%4k
z!E<!2MC>U(AL7LicF+L8R|l0ht1)n?aAXlRFGz^)0o}}AOk1;-k~u?yudDoi)Kui6
zLNQxA>A@pw=FimmDe9Fek1!%0GDb>73WIO*AXtq`LKcy0J-B{YL#7@ZVBU>r>8Q$@
z9JgnKY74yqiA|yP<W+H6vLxb&N}c@OLF=y!`tSev$Jt*Q)NF7HJL8gH=2Itd_hww_
zuDDOl{cXOy*?+1Xf7?<2C5_+qu)NI(9PFq4LEB&W_XPe)Pk=sg7ymQt-l>3(ZH9Av
z&-4STd59T}_K)~b{($J-u%xng{p0x$*uq;!xOX0JZTYO(A2jN-Tl~xGzTyskMUQ@)
zzN2TK;Ya28Krr*M4SYk(kL~53*tp%smi`kT%((TJj^{@44aqpy3*a}jd|}!bN_C^W
z{P~-HaR*ZV@>#q%#S3?Hm40jU*GB(}>iu}~e;0W^!R~kN@kdnfM$>Ln_(#nDT|<7s
zIekHkr%vx;t_$j^R&ObK`JcVLWs3VIKk^XokN<}$|KQW+A93<?uI{2{@3<lA_%>rU
z0&W2Q<e!0H%a_jXGp>E4sY}9A*eWs8gnwV^e>nUBXK*<Jk(YwgdqY1A(_a~+<!@~P
zN7}pWh!<Z_uKBLu>t<UIDmAnBcWw9sef10<>`YZ3=znf`HjAkFdRBNo*lQI(d%D;5
z_FMgX0^kIGAA4{x&%t(_wm#?bv8o?8mJIZjW4*Q4Tg`6Vb?cW`iY=F5We(gh9VIUt
znCm^Gs;Zee{jPud3H;vIKm88=komvmLHd@aZ$0pj?%<>5PaMNMkD@dAM3<ja@|9cq
z%3eR`gnuV_{|UU~NqUFJ>pSBqbQD=|vUOAGhAZ%ZMrAwxHlyA8<H#$szf$}0jeSi6
zbOZ?hnyG$jm)}!)d}n^o=<yl-sfyow`={Qy-?@KJ;P(W+^aT2sWzWBKnBQjKaRPkH
z|2gjaKgTZHLnNE_|AyNC%ef}IO26XV?)Pvd1(xW_OI%DqNLCD+ReEOt0pPv<-N@^W
z#o>m}4avKA_rv|Z`9vPKf2{c7C;U6!?+N^#z*|oMekVZUsQpVWFBKn+`4REY8~^7`
z^pS&arfJtH^gr}9^&gs+o&ESd{HST)KFnx1f<OL7{SgDcaI{yYSKJ?QF26(m0VnXc
zJ!R8=ChQNmHIGL8OSkoC+rRwp3A}s)`)6lIUT*Sh@Kq;pb@Q&c$Md~b{n-Cy#W<jV
zBlU8_YVZ|3YRBK~&Ix?{zJPyO_wl}WHvhdEKLQWzU+}$ouJC`e@t52E@Bi<A|G(?6
z4Eo1E{>KY{Wl$VPM6{EdmAze1MYkn$;f!v@d7`@|yGeEtR=toV!mIT?RjtX{UY`+;
z4X`zrKaN2afYoniH*^E=I0sT}l9<sigE_va12V7J_zf^AA`otMaDBQ^9eaJxt#QV4
zY&`r*n8i|7Ims#pri`u;jfFwK9`y!2*A4)Fy+1tS{aeeLp;-`abuX{5!A#8PngSAM
z69O9mAGU=qKI5nuK`CieOWpQn@M5`lIrh0OxMsa;3!bTZL(OinAqs4G+W<~M<qXg0
zGKU0i+?dA>lVZY}aY~W;6o>moMrj5KvoOb-5CE$Nl_4BBP4H|nqi_%;KOG$UOq2g&
z-*^Jwbf><d?HgLo7Bjl;DIyS#E9?Q8D`(VdrZrdOc+w6SDna;Q0if3%fblQ<djjVt
z03OE)4lJVD5a0lQ{AEzx;kQkb-B00Q_bvVq<v-6(|BxH<<e2b2LgL9rpbWKXcuU!A
z!a$c3_(QfQBYdM)cmMDZX2FcUxSic}M;^5}t2k5Jj}fl&|J%Sv6J@^jYw3kjsa~Ly
z<U)_T)(C{dAF8<^$IGy#17KZc-Sm|nGgyO&cj5c)=1rh}sD9k#0+CO}Yc_cy1;X89
z7YKrIFZ5tnDg_7tyI%y8a>C##vejO(Bhj-(2hvtqQfvd59o`Y`uoF8{hDphNWh$bu
z6qY0j5grm&s4UbGg_5FE+^SKX@xSZ4P5_)VfZx^i*XQ>Heox@`3H*K<Zy(hroX&2-
zq|g};sv&@jh%`W#;T%W(jNWq$7dQYWYU7k`VFGf;u<nAMTh$?s<YX6||97PUaX(9K
zwIOUpc(OLBte{|k2S}|DgK#hmk|R9ZK#lDd3IGA%+|#Uit@@gU5aFS?6U{C&s(v#$
zqj`jD1z-W3eM=nrh>Bn(Gc%*T_Vf8L#{k(JBQ=6><LE#zF3o$FmbdhxeQ&f^>TV%1
z52RJU%bQQ&!EJcd?Xnz_#{?=k#fR;jSX6Dz%z8%q$HLSjo-9%rC5MrCJaEc%j=9$c
z*aRanf`bp%fp2vLkB3MNFFM!`US2_u3fJuAo^LS76QwwX(PRXPhD6l<ebK(@9f2n^
zNcR)EC8}kp1mSQ%ww<eHw9fJI3=IKr0N)(*wd$KqvdkLlYH61IjQCxXy~}4j6=*&Z
zBRKlbAgV_=`uYTV03(+HM-Hd>y#0kvMiT}NehP7GXe2`TPuU!pL&pBX;knh^Z${?_
zjJ5;x0EYTgpX{GwaKgB+70f0LZK>AIs2Q)d3wnMbO_YbI`~XK%?~wx+OW<kN9K;v(
z7t}=qP?h0`BN9i;ZCz*zqw5F;J(rypHGNHMR$AXlP&vY%_q=etptgRqc1Gi-1s_qh
zeivR{BP;$TVa<+6DZoc~1Y-h<kObklVzvq&i)>BZGZf>^=*0k0`l0Axq629RI~8r%
z21NaQ3-E~zRs*#BXLKWw4B=_=Ap}(b9%e+!p_f5VlvwG}I`kPJv8;q=9<;`v##=i}
zhqeQxYRs73T4&LWN?GT6p3I*JtD9yS%yIG`-~^=*Z5iX%rI;AF)`cTVRftA-GS4BF
z7p5dp58#X8qaq^l>$!FLG;V!j-8ybSXd_|~R?f_fhUu%3`193Ut)Ku(IKTDGOpZ6<
z1Y8$fv*L#|r}(2!)anK-8SdQ5I~4n$nHi-j#N5Y`vRNY;!dv;H9parJqp|S_-tpcG
zcSQG4h?2x6w`kA}5SJx9g`4980F@#f9kaDTbpQu2?1J>xURGC0FkV%YIeP#K<<9<}
zZ#klY$^-5|&q;X~3)jCTns)w0dmCC8s3bH~01?$$--V}ErEl$)C(6v}6xI#ZW$&LA
zw~1Tov!3Fs5S3uUy1_~l+*`2?U<P=AdUGV{c+bak*hJdx6z@ZRBu7I<Go*0b1F30g
z7}e;etQL*gGfr9~%WZ?(!g0<7BBQ_~lCqvNI)YDpYJpE0_)&-*3R9BgiSY6BmFhTz
z#TO^hJt!Z125720#h$T!rsj06(7)F0wd$DS9TAuXz_4S}w^YhH*mE;K!;#wDY`xG;
zlwr1-&A8qYjg`-Usap!4ar)D0F1I_71N3Oiumlp19Dn&g;{?(TPNB@c0lp>7xa2pu
zaRT`qQ=GyZCfAbDYXr#$5CBH<Z6Z{EVSq|Td<3^C&VWyNo=xzL>eJ1i>i4Z72ZX~%
z2Nvu-uMuQ~@N2hMSm}uSVg`Pq&&RoHa_n|$cB)4Mo%y)h1LVv2J=LYf=gBr?OR0ZA
z`cLu$c;k2<F+H2_Y=krQ*(CYYKiubdu%BB%e`U~r{O7;@>%Y)z{oO(T;~)R??C%cZ
z!gN{BzP)7EbT14+s|scveVX`FL`%E~qf9|c>0`I@$Nj>^B6YEpSMm`*4m9Td(4O?B
z_Iz?Gwlp!>(!hh5ooj()M9?o0XjX(OQ1?P<!u{A`Vnejt*K3@Q3EBa_Yy)eRiFi7>
zh$3HUKq<32O|wX2QSeo^ZPGOr4dxe>3>F7ao=5ysaUltJKnLMr)=b~>incL9VADwZ
z26Z1yq*{XGwPF(JjpnC}@5~s9HkXIdMRcQphvc5w=SE?cd(1oXOm6LSUe8t4hl%%=
zMq*Do@-*QJ4Hh)%U8V(0$wi(~zIh6`_-0TZhMP4&*7mbhJEEnK0M?&|pO{BV;<K`L
zTU|`hW~cj9g-t+d77X&Bp;<@<dkj5Y#?Lcd_HhZ4m#`GC9rA@%+bvg<#NH;<oKD!e
zDJC*+tt51+QclbMnlM(8S_;J>N@YI2iZZCDA~?cRZm{U64egu`uD+1>ku@?=oV}#I
zJ|zp0`=2VMBQps-3B@5-?1-~Q>7z!^#w)LNs$;pH_Q-t|NVBc!_|kiVvAb-@X}|pT
z2Wk)z6fL$RG7nIhVLM>&#z9tmFpe!e+hp518POK_5P5M4>L%J}C%94P6R=h?ZmNT>
z!RX^ci(dGlqIiBaoXe#oJr#2)p=R7{N&!#8^S~*J{xZoHx<j3T+UTn}7<{?<U6x`x
zK%3bPv)O)CI&Ug<Plvz4lqHRa`2ZKl=1NNC<tcU2ZF`_oB)vYErN5qRFP3|;qlT!f
zrLrzC-!sy9uG@?{=2RjvNnEeOKx(UJv*ov5a*Z{)d0!87EXj2~!`O)!cC}BfoinM3
zlf|w%<ylDb%<Gvd(CQhhXGWG9@5X^<(_}8(dTHfaLKBh7vt^FMZ7T_Ck2bO0QM!Q~
z?=|NL8EsYm48?YrNcx7D3#W3G5Ks0rE<q9x?ZR@1_UQ&uf<$k<URaao%OMSQy>U7t
z8n(4!xM8Ghzev%sl6PMz=*Cz1vYSkX5>oO`IN<xv!>N>A_MHw!xd79_oR~({WmLB8
z2T;(<`ZH2|R~tFMt(XRQruSJ5s~?s4S)?Jgh8A4DHb=koMS(a7klg;p@WCb@(DzQH
zJtXYu=PeyRalJE~-RnEr`5B_gA5=f$ACx<deC+UC*aB47WMewj(#J)A`<WO<cz-1*
zg-cblUgo*TTx@}@t{VfS;(><>jEflg4OUv9w}&z=zIG6`y4`uU*5kK+@r%E^JfdPs
z*lTLa0xDiIZAdd(Nk9@2F}<ip6}+fMAVD%svA3hZ6H(3AJYiF`nHu{r?d^MRhhw;{
z4~-uQ!pNhGT|RPKJRigj-CYSi8TqAo1xY}Sv+yL&g*i5lt4&os&A{rXhk0CBJKBu0
zKCZ5M&iK=}&N=-8Jpep-UP1e<dNB5T+Fp(HF76$Z0uP1LP0BZH+W>GeU8YwIo^ZY|
z4{nY)W<TrkkyWSBV_0*q+27FUfd?HC{i1$r<xo#;YR_(bIU3C>Z1v&|aON;jBiYB@
zTHf?svFhG6@pQUn6cf;vKvSk5jjBqQClYx=<*~6V7;tvy$8Dwph1&|r=ORAX=k3qR
z+u5}gPH|TU4Ze-w$>}NDUgJ&Y(_EKLeE?Xot*aC`%F}&(O~wafJum(8ey^RsZ1jtU
zwg(2>;pY$idn@PDBXnd&@C$eF;DJFj!olQUz02N@a>Y7j_&&KSIiSmW^U@w!=J)!C
z$gR5Zq;FMT_TL?LL=9rMRbrE^m5A*}MFnc@<!p!)i)Cw_O(?l_=&Y%DO1UtkcFT=l
z3o{3=coZBLxFH>m{UWeAMY?A5q9w~o!rv6K``p4i+hL6@qt-S@>w0A(S9Y9o>Yy5K
z=7~~5FJmLJA;lyRl2=K^3YByV&Q(0s*tRB4yi}2lmsW<LJI^EXcoG+k*d;E%dd?2X
z-MZRJt*C}x?j)xolFMrqii?<;sd|PmB6HR<OJm)o!Qsd{Ck1ViUJ2qeUKd^YbTj>8
zOvbwfjo>YLB9`^h_QD^EtiAXvg^sPDz$KxRc=(bNlQYv>b$MEGsQfhZZM`fQCTL@c
zr5X?Nw#TF>Mk|p){K?41cg|ut7su4;UE$z=b26HMb~gT6JKZK6gnA$+yA)q_-<xb*
z1-LD|CqJ~42Vp3({rlAE*oeC~4XI4)PAMQG?!>BPD!v2?t`R465kgZoVo)hAVsdDz
zj`CDc)$m%Hh0?YB8nxD)=cfx*Zps9taA^Gu<gF!mY$5yOO+FB_V`wOi%%Tb|K?O`!
z57W`l1~ker@<G#~dVqM-<Kb8<DjEw3)?eV@%!1aih$!<}QSs|+wH{u}ZM|qbdFiUC
zg0YxvA=ll$p0$B&<q1|oqxx@81?GJZqkhfq(xlWdr83e643pl5)J&g4zK?V)Kj5{7
zXmixGW_e(FF6AcbGb;C0?st}gID5;3!lh(3R}3&~r7U&lPgKwSV>a6is7!&TAr=-P
zP9_uMxK<l}OlVIbkZ6F6=u*p<N1Tor8};-$97uw6YIqZGx25M`O>BO*>qep3@I&DW
zLfu~iy>f5YJq)-?omd>hXY`@ZErIW4SFdL8@3*%}$Z@-^AOUp4=NCV^>E?+}(Y0s!
zMGreiVKYVEB1nfQ0C3Ul$>Afj8@vyZ4e(jedB~gQqXf1gnqF8TrEU^|cnfU~<o2MW
z`FSA+RA+%Qk47p6wh!)Jm`Dc00Sf(5w9e&7zlpEQui4E^fQ$Q{E0#Cg&pj%t89V!#
zXETm#O`~s4)vB)2+RidsS1)>C`NatV`o41f7)Q-te?95pWfHt?6!+DgZMoinvhlZc
zFOEIQW5@*BladgpFu?DOY@j^k-FakZ=8&oe$l2pYS7z>7g7b<}PM_V^UQ{m+!D~Lb
z`gA7x?&11-yScW%=aI*Ke=G;8*K9wJKBH0B`JRXJC}Nv?Ui9Oly*7Y&2R&$2*D0MC
zE3`bjpx7KUZ<wrhH~1-A4X6qIPQM6~*4U<+){Lu4sU!vB5^lsZ>|FI+8IS;YD?bI^
z<ohNYo4(n3swXKjxp`LdiPigl!klh(Z;<m(Y%lx!o<`8wJHfsY-dX;%aV)p*@pbDt
zk2s@S@dMz*MehkOCv|CrD2qtx#|N_%$x6Zj+BB&Fe+^P63eXEPs3LH7HuuXOHifQI
zNR<qPBZ`u_4D5i<?*ToKoCGr1uEyNn4v7+t8DEL5%qZs~<w;}j6<)O@dcB5O3dD**
zCDlXy02RZBiEn_mO#9?FU2PT{$5yk{oEOtBPV2V8Ho$ydT&3T+N#D+egGp^|Qa&oa
zG=fqgc`==w5gqsN(qmfMASEtPkh2Z*$s*}bLZ6~o(ME2)N-MS~K16ST)F61ZEv8Q;
z>aja5g7+nNZSc14d+8u%=)Mv;JiujO6<If1QSBoh)s6TvlXKI%DUZe|MYmhUZyjqM
z0HOH7decxZ+0cUte({FLoZoaK@)k=XQo21^@@z~4>B|EHj@H+zhgNL3F7%sEPBhUc
zB0AKR6|>=-NXm{<Mn~Yk7RNEIlP6=Smo`Q;Z-W|pz##$LOktvtMc&#SP1YaBpKXwL
z%ket+mtL!!4~}j4CN4W#9~xgaRtGVRD==LX@DOPybTbL7cbT9LnvS&mmLh?<K&qJ4
z&=;d<P2D{lL1$aAJj8`&=U*g{h>a|y_1;S>cv%-RHuDwOFKS}1%o8Sz)l!lE=3<ID
zw{ly3F9!VFl}Ej})XC$dD!yk)Il1ozKNP9M-x_jlU8?B$*`AMdFyS_WLFb4b(aTrx
zE8;=)dRg@&$cKBoo{~R|bm0x#g;t66<7^*mHRC(vnh3at&+pmGa)1MOrsutoceOxX
zX-J9EwO2^av&#|80`ipRd=Qkh{Yo|U)VJ{>WZOsMr`|HV-;2aFj6AuObcp^7rQbxF
zLUb+>h3zD_sIj!D*o}<a?mBXlmyRP7E{G>J@{MbBJTawdd?GnRKYE&+y>!XGMDY7x
z|AqeUp!HV<{m*~=ud}~0i0u7EL&z&4%4?>QNm!`MI^fwm_1qdHJcWy`naOqSWY4X7
z!$EzsKT^%Lgznul!m$CiVzY?>RRC6pncdJ0m<A9GOA&+c3%^{F<1|)>d!D2mZu|yV
z1tJh`b#Q&UP#t@H&#iIBb8P(dkTN%mrL1z2RZQ(iFj_xs)Y>6Cgtu~S4AcPDBDmZs
zLmGq+UmWvR;8f2NhZ7yjkEa6u0EW-<+!wmId{m5}lr*ZPb-YT}Y;?PxirC2n!|t~Q
zz(@TNyV(@rnX1bgKsLCgYorZPV8d;ENSGNA%<znwmGz7j>O*4DW9g)5SWC^@Rz#oT
z!+nqz$jAopH48oVV99dqRS_&2DG(mTJ=!WZfGcJc4ua&TgLj{v_uYI|Ugjvtoe5CE
zyYSOfhI&kFcOK1t=R9s}_9i<LW`F}Q)m=~zZ9`xYcAJ?Qy%-A?Jlx1!Iit3nY0VWm
zp0opoN)T=q7ratU(Ngnycupn+^*g*HY5@sr@o?Q86xoU71xY{vI5|EU0%M0g=>CYP
zc+I+p=SY_D6nYs22rnWa1GN(zSVXl2zyUl>1f^h~m-;%KXHgT8lY0O~naJ^OVaPp|
zscqe>c7(JuDlUKnuyqFOP&GhQ4B%&oc^1$Bz*i%Ik`bQ3jODn@jR?g7&9HDOEXLrZ
zf#JpZSJ#q}88U2mOW9_^+<zw-*<q<jcnUYp?Q^gdWHQCqx`7pFAlwvi0LG{7hIT))
zLylcIf>!OS!$3^wn(cEehC5=75R>s7TU}j|qT`z9W&I1H>UAe;B`wVkGu2VfJE%4Q
zz8D)+WH`sGPfG_N@980(#NixKqA5hgyKvN?`}f?(goG<ffaDdgSuxNADG=@+yFd_x
zd!Yxr@_MF0KS1O$!4ozLy?n3O5#LM?4y3&_=hy}?JG>*>U?+AY&ywRuO2`+YB*`A(
zAz_8eLLE^kDLTciy4KaqG>7KdME^4lb;^vJ<JqO}P&7bf2Vg`%c^d5^JS_+M2~JoF
zFJ^g#Y8@=KwOSi$tFs#*41k%;T#nt-gPnE6gU!;bS5Q^KoZ@QK%W7UTZ8m~$oUHq6
z-r+J0Amey0B3KIu0OMPe&n6Y8uuO;W!&^VuIOaj=tO+6uUJ^7~6{TM11h+tCj(ZTj
z1}I1nO*IoH<<59eZ2??Fqyf4N=Q!$T^qyn5zyUB(8>egw6OcQGbr<y9st$1^C%fSM
zXL~^0&r(}$2wM@JtW7E_C>Y=YQY*wD91Mfx2+uZ9W4na{Kma)RG;3a~zGfjrcqnQ#
z>rYViJ1YrwM0GI901M#Bfw9m>)WtJ1qrLX?`7p--*&HJ^f~(?PxHRu>EpO>X`##uS
zsk?>5JWy!~YayY>wevdy=er>-l9Z%S?RnYZ9ht^~s%(BQE><PU_2&dtpI4!Zcj3cG
zC)QP4Gdr#s#fLHI6O=-0;OiCVq7{fv)u<@rIcB2`F#C}h!NCXXhTAmM+YoC7Rsc?g
zzv1r*9G<}bQ0)kMAR%6682p$=_#zR7a9D8B=+q185pFXP|474Q<_jp;;0-X7iI_0c
zzqSHrgLSp^ey{V2``%f3*aylHf$%&7$5%qgjl>9U{qUoS>Jg5dfF8ieWx$cc30}39
z=}L=1c5pp69IaS?#y~YB(nOnP1D1Hg$OGY+NS%=gnV#a}uT`?(B+muoG(qv}1xY6q
z0~^ZRjN+pOdH}!8hy1OFtqgeLblS3n<2C9s-O!e*az>L)7xer>nkWxZ`2mik-XjMt
zmcY}jIfyUnFQ|(Kpen-=M<kAx+q%#cM%NJxdM-OHYWkYiOj~!b=P@B-5FV8{nrmm2
zT`s7t%dDNzxM{&hRIT5ISJ%jje@R%gBT@?R5gx&qfFdM8IIftj!p9<8Q}+zTcr!W<
za3E#>Q@sv#P<9}_-tlI%0Z~8Sg7L)AG0g5Wx;!dTvzQCvp_n321>iOc8M@K~YqJ55
z2v}f+eSJ_c&;ul<mGI1i*81JJ`>jCY{klfTpiU%E8;?`WB=szsQ7P+O&y)E@h^i{C
zSq5{Q{0BHeDYQdOwHg-DPREnpxK;O{6wwH84J^g-!j!~nFJ=V#T+t;xw=SQ?txv35
z#|_B)usRkEO66x}M$Pi|Rs8wttyWL~C7fSfb6m5^lmR!2fepQ`DXb*O;S_JX8ag4x
zxRC{-2@rFtM?Mn{y(>|(I0J;kJ)#1HBcoA2g43>b1FrYLceW$FhH|KK9|;EC0I@dV
z5+<O!LSo4fhj3V+Em0l70Svn!J=n0w$E!MJ&Q7CH?(F~hmLq7(TJO4^6LA-FlO9Y4
ztWhHKInF*A&vDk|7-}YB!pIEa-sqvgm8ZerT%eMWc>odBrtiX2tJ1f2%M)ehbPDT+
z>azF4_`(EH2!Ghi1~(xxI>nbCD#65~A&kGd!U50(L%{&`?nu({o{v*(BJFmHUmSm*
zf6GxX7G^TX<8Eef`FN!VGDS^mmgamP0i9B}9%_v&w+(I!d%K<^qrfAQ+MY8yg6Ynj
zx7CXKKi93GxAMHqSQ3{!F>yS77sZ!U>aO81+yL!$=g2d*sV0Tqd%~RbU{eJU5vgnD
z^_VCHB}v8K5rSC&3@2>*mP%O%dv4}uI8vLNtrxn9GOVxTR%VfPUaJWbL?M77F6F5z
zh1L3;xLIFi+dvcjJ`C=PS`6d>JsJj<WFp6Is6P~TE+HxOx(JN5b=*_PDGF)9;SZ>+
z8f!-|SK+2`8Z;%uq0^n|U~wm!SxQqVTi*ab5o2t1q6L5l@G2Gt@d$@UiVD69Z^u*S
zPBW#DDq3nL1W#1OYNoX#NE{#ljO5xS7F2&>fJ#Pu1f%`&zGyD!*@>R#zAy$w0KPd4
zs>Yi^74O2);Jg2h$k{2r(>qcM;K|2G?%+Wwno)d~VO(qu;mKbnLGw%%F#)b3O=W8q
z9voefkV6E*+dw`-)u$|lE?xnEA)ad^b{7wrLs`Yc=22Hzmn?pG_(yIbhK>lIGt$gB
zqcb(PaIFW|k4?zbZv(7PBU%bn2A_YFfIr*~(#HfS(!Vn3fBobC{{1V1&>3yVz8$v$
z8?KHBuMO|1YE91e`iyXFfUUXwaSW;etbQ}Qp&Nk5!BWH^TzZUb26LRo>Tu7Kl*5hR
z0INU*!mSRjPZz3VukX1v&UlWEhu?&=Sjs9VS;fGV(KVv6Fz6Mz!h=F5LBl2ORXx`t
z0FRalyi<lW2#+17Mijz=rh7Vt4Q66S-!~}m2k_zfKo>u~Q89v2(x{f!@hVxf(d~LF
zVkZ*}yWbW7AN5D<W>bJ?sxE5)+2EG0kv2qu4Y%<jVP-%u!!v4D)-zV94~a#OrIVsz
zEj4dj5q*je_d!}9BOAciEcDoeCCjl_MX+e3KzJ1QXsg%&u9#6c2$G);-hF!Bck@+w
znWH3kCO`%6!cR{b>M^n1c{KZ-^SG_qo9sxK0S>@acR@Y04S_}2ZDwZlVk}tja3gc&
zjM{dlHCN<#(he9ZLAY65@JcmBOU>uuIhhdD@9>VO1thG+!*zF1WG9jrBmn{7<oIL=
zj2-r%`y-;_HR~RpBU!>z=w%cjyoi7d)J||<5!Ds|2k<lzl!AR;>g#ZxMNLFb?g12K
zBFDRhA@@|Kwso)C5z@}6xBw2o))}lr)c{d3fS)1eSwI5-UyTGxMtA}<mg6!vA`}NS
z!@{Mo7=x1rh8O2wT}wu0$gtrpWt$0e|D9xHhovInDcm@>&%suZ$rNAf23DYfa8tkm
z7@xKq+Wp85Id<U)TD7YV12Ls*w$HH`?ua!)OvZC;b#+CGj%%Km-@h_w9c4eK9k_Si
zRYh4iN7Ph^co&{#D4n~1&s_`&SCjzBD_*l=pb1hS+&y-IAPDzD4|e7COoM)a$YX*h
zY!rI=Ua=#-nI0TSduh(G4PbV7N3_9C>`0y^$B&edFGNX_J;Fo63YCRAqEJ$Fid%K9
ztC?vI&9h1U&x14L=6H7LI}{BN*#Q_4P@YD+2v5s_eu5L0!i!m6p;`w^ZLQXZ+Uo2E
z2m@dyGnZrc^k8Ql@nEy`>J?N~FsHa0^|G4ROq-418z<}jns>NN1IRd@iwM>N0>Jpz
z<g-b|DJ;_={P5OKHja5vI%|T+f|mr1Rz<1TIl(PZnd2TruK@}YL{rU#Nx3r~R9gTS
z5ov%f!#R%n8NKHiE^q)$)W#{>!UW`wVci8ix2i)N$;mD_|Jfc8_p{Vi8^TtECu@_+
z3JL~zfYb^x2nWL;Il{9I)YxvJ01yDqJ<Xces;^lH5gv*f&H58m{mx249Z?-jGQa|O
za$qd<5q0s*%xJIud_K%EKsLunjo_+y7cR}aTgzK|(Y_D1SL$vdF%MK)!dgh^aqaw$
z!1-=SizFo}RC``_ct@sjpemc+i;GoBa{W0$)#p{H;$8SK(usA|*36D;M)6?``UIuW
z8u)s}xo8EVQ#C5ec#hd<1I&ITMsV=Ky5Tkr^)|#>ffayL;TVo(BY3xUFYY<^p#cXq
z!RU2Mo<%8^Fagx0rO_7>A=#tSVo0Wjj~hK!?2{zqpnR9^2t2Yunx@bYi4eZZ5QT79
zz$466Gm4udJ;H4!;vZ>v%zOa_8@vH#G7%GI`qx(AY_P7D-tTo@ao;;D5BoqlA`qTu
z;P^@ixse#btsj0gQ9Z(u6VL+~xePdRIKivdGF@pg$PTXOhNBhh&lsqNM4D*RY`_vv
z7<nKZ6R9&2A=6V_{IyCJoaDJ+oF*uKy&&m?Vqim=n^AnUKo8)z`H;W$u$2K%oK9Po
zaJ)ucrW@K)RnBO#>4KhLNE78DDnG!H)O+N>#S(a$H3#uU{RMT=090i-;)ul2a$6Ug
z!st4JLC<BUMNMDRnrZ6}_B<v;48o%lM|16rvdaavb(ys@8aFNYh^qCw@ah^_@h=H$
zc0@`6KEfjy6HtUC2*(w(RrpwBYwDh%7;i?$0S=_>f2!A^4$2Os*E`;fHX!QfTQHs&
zI)>SOMwdq=Y8G=LJQPy|ssP+ZAwySsU~M+w5djOVu&)ma26}+Rv=W|q&|1G6cfS=#
zykFM{8PtgcYU6Q=nWUaYGb&}B>v=N22vJqVHOpX*lm7rGD1~;2saC@x+Ua=G8@K8n
zlp-48t%0RjUYL?t?Zu2hpDVhg=ho%Zxb=y3>$m}VA6CbrL8<)A%&1wuzKTCzz10c|
zpoH_QYmRGHnKIx;F|eW6HHDP~Ih^8cS3@VH7&o#&Gy!5x^~h(!p?4)}7H5EPxJOig
zaAY*<M{wG;Zou^(_|A5u*H8{s?jymV8z9yuT*3rYS4b=w;t&oCv?ZzoIDlanqz4-o
z`FK^Q%-Lxa%ANf`-*N<PS?gWbb0Y3yZqkFvfHg{FKF8T7<2lZn97D}SOc<FV+#5X<
zxbid@oC{PEG7li4+Vov`YE}ByZh4~2oK9igP+j(Z7+;tm3gHiX+2AHbMyL4lLnWAa
zG=%XtS2zHgU?>=%-W^Fg-t%#aO{CpU@r&c{^KUuo#llSHc-+klE+4P-K&Gf^&C;Cj
zBcN02)<dn4<+j0XVQ<%SWE6NrQrmMzM=;%)^R`-X|L3|D^j4mi8B5}lCnk=k@1ppU
zO5HUah8v*0?i_i>Hr1ridrz2?9&D-rA|iFoydD#!pd_jIJ3=rEfZ>Eq-%=^-V9(9`
z3`c5nv-LtZQHJ$(+{!Gn&TBP6f+z$q#HBn{rLbC`6F2LtY#V5z--p3nQHy~bphv^N
zl1${-4fTh@&Lt#;UKfF}wvKxWIYl8YIQ#*XRb%Z4<|^D2PJ^a|ICQ!*9W3rdGfQa-
zW$PQ@Ct{4PPP72<0A9tyARgiHNKwIe;q7?J+-as1QbkM6gy4y)Sk1I{1c?I#fRS9g
z#DeNC3{c64k6^Sv-WSaUJv-6!+!w~62*5XoLDhIOsN!8X8hrQP5ji`>cX~%k0X+E_
z$sIf>MKg-eGK`DOAw2oZBxs(gA|}99q^WGp!h@qL5^{(@cpJz^sQQ$p(8VhNFvN3h
z#O~q&b118L*gWbA>ypLWLTje0rOBTmxd=OGlAOaE64J^4FE8=V7B|at$s6M68W9f}
z-ib(I@J${Bt5HeFB66(<*N;ue)MEq8yAdq~Dud5IO28lP2I*r06wzN9^dJBEZ~yu)
z^!oSzvi|0v|L6Db4nl{uo&R?33T(JKBD^-dr>ZqM+v_vJu>rQ`^2af#0<ikc?1pXt
z9tTSigK+6FvKh>A8mq%SPf`vyegmum5eT<BxISH|j=jF;);QxiHXeQ@&SEL6oMaUP
zQ%2W_#=@Xi<O&Z8odgY+v{&_9ivT=YCh$%f(jYu`oElLG3!3if6gHTN8GYZNz#qVe
z=L22*^hU)9N=c(yTF0wo%|^HDsfe9SFzkL?0DRORv71c+o~gR50c3+)x<=X%1vcEq
zhlH5{!3@u+Sy|6mp*|!QJ(f<2hPBkZZAJ7cKHLXsfsAYbU$f9-50)&)UKPQjkpkgS
z+@q~x1Gr*F;UGwUI(YZ#dEd=f<z<eN+?fCsybC`)WvIu*cIVOTch2LsW^b}1VFow=
zQ{4si&^81XVYiu?(TlNQ!NZNrl{0GFnbus9<4HSUs086=altFq6fHHMhv#HMP`|@F
zq85;_77y3mL6MzEUXTO?fRp2sAux8>gYJ)rir1`rc#dQVPobAlfbb#$GEh6gfkjkX
z035*6L{JL$d8x0%c@{MhIk^W=l!+Yg7KYqYncCL9YDY*rqv8TM09$9U4pjq0#Q=VW
zm}dbE0DLtPC>h}i%vg@g+=x&d&<qQg!eR_g8W>)je|0SxnIXf5x0G!r%>8$gksX$b
zgr{)h+&%|eK_*jtts7W@2Et7N2Vi{KZfN%-JLK4fBWTsGIt;{=uGv1vVz?vL2r(JY
zvDMWTDLSrsUVi`Xpmmh}oOa;ec~=!>;T%y@A>v(lnxS;={yldwBwSGfB(HeQih(9b
zfpGWO1%e>l3q9DC*E0?J0V0nHp0H8q<$J}B_-1-=Anm0&$2Nf3;T_QiJFz2qmK;A)
zLcS0sN%ja22`f|<>WD%~(J5}#wXSBSIW*5E^}iF&jGN=xrSDKQKx7ACL_m2O?IJua
z2l@$4SPCy@d4*~nEVZ>-8)~bw8z2mTnao^{-P41eb;N_s(yLccRl%I%YShbWUNdbr
zf^VFx`)l6eG7TW(crGGX3kU$?Ta(Ww6{oOFhw#H&KiN3uLFudsA`4y;G+Gs<Ugre2
zKxK}55WNN{NDxgm6DH-(cu;KtTtuV+x(w$y>Sy$xW4OQpFi{()Yzq^RJBD=^^xUcr
zaU>_Z;QVKMK-|w#TWtti5uU6~Dk~@$-~m!A#2_3DgX9R$Hc(@`g#thTIQKMbUaP)l
zAw+m6YBcLlQ1v@233WtuFv$Q5;K_lp&_~q8Gc%*T_Vf8L#{k(JBQ=7n;$65j?`|z`
z=|%fK*j}l-g~U8iX$fl~p~toJI|ApsAuW=Wq)_d7+2I|T#(}DAelIRoCCT;Y1XZ6`
zp^A6m!$>FARa-MVt{KILG3XPNLTljb73ZQAh)&h0DC0S1qYW_okr=_j2kVC0G}PM=
zYXw#SPK9GQmW|-u*1fpr*oOuj)C8l~EqNBDSi%HQla@wbNQ7jMN{b<x8a{6HSg}u%
zkc0AFx+Cz&25Fi?LnK1@Dnk^)VF8aYSIsDHj`Rq(nTUU+;W6_C6m0MYn8`#;nCV|z
zfwRH7T6(|NdBuJ2tUT-k<%mFdo`K^lA>>A41h;<p(M0tKM@~QwVB|93$l(O9TFZ2$
z#UMMlo*Rx<tUqI*8WL%uO|t<@JYnR4a7?7mNQ6vJaq-tGS#Xl)f^nLl`1OLM6N-Ti
zWo}0C(E>ey-{wR9*27i?JaIa0S;Fxeb(wBxOI10e$)*c>ej!bihp7AjM^f*R0~br+
zY1SOX7xfp^MFUWk;fNy=N6T$pXbPk22nIcuofb8HO>3sDJJ|D>5HSdkN*v9#Gs-R(
z)YfIz&S>1U;3KNm@4~BVWW~QEtl1GM1^5V$U`#*}k{}#c%vRxJk*%qFhGM)K9S1m&
zvj3@GhdL-bkY4Y2GunWtpKrl<V(1uV_ZeLtm8e<Fh44^J5vT%i8-)yA>4CM`fJX!@
zu)@ARC>ZDg64Od}=0R)yZruG=An|@(BV<q~5~z*GDQ1#-7R{)Xb*|^h{31kE71u0-
zIZplqoS+oiA*Nali)g3gNpIY$dr*pKgtrEkVtHXoVzn1D0)4LNlAc?aPvh1n)~(|P
z<b7Bjiw33gGc%)R`T8pUeDzi<D1Z{qudX?+S!K$A8^yqeUe^>>66A1-w_OdLkYe1(
z0?`DBIn^Vd35VX5s9Br=!r>lK0m6~ds2{;;*SZ1Md*D0UkzPYNRJo4?gKmIWn{Wve
zP+cLhWQapJEYOyy4&VTWU63AZSmfhXoib;qQ7Cuz|9s04v}LV#UC)WQi@8Y;CIi+e
zk@*~FpN!`?YjO-V6ER_AhH!86P~ghbU~n!_Nyt2ah-%Y!;i*;WTf60nGIKhGbwhR8
z`(b=xf+&PP>}7+S5E-4~%MX=c;?WSs-(2AUXo8_&fO>Z%>3Gk_DK?RIJH;=K|Nm$2
z-4Y}Ta^pbqGCTwCFh8r6R$6JL)qe6q{NQcZ@pQl)7U3QlnU&QY*<Im)!C-hqWLEXe
zkTv!{`KT5XV=&s~G&R_CJkt&7BD>iOecl&<N}<|lYK|Ps0?Wj)+~^~rj(1CPN1xFW
z40mR~tY)l#U$=y@l>4d2<T&L*$NlL$NIndO>S`9v70{fwk6e)rIl=VW73M)VCRPW!
zMeOW(J`SXW<b>p(FhMT>h8+(4!IiWT9DNht!x7um>^$IAA;ZSA9Hk~%<#rk&Kx6_K
z;*uV+f|;fF1B>~YHZ@d*@5A6!RH6tL&{;DuB?hAHg8EHijs+x`p4SbF+F8!j>11_k
z!r?cl$SQM7FlOPRup87R#G%qkcQAPj3}V3zri?xTz7dPq>4CujaR9euiJ&;b;Yv}$
zPvOIQQs1GcU{c25nh3#FC}Y{v+!CZ12mlz-wMsuw{(%5WDe)5Q<+taezM$q982wlm
z+8_zQi_M^HJPgXX3rCI5`ZFTCQ+y83NJ@a4?=8848!4$q@h-b?lG%hSedz@CQ)ToH
za29cLG<#ixt)n1}p#Z{%hP;K$4_+{x+zbFiJmz-WRovhh@-p6K&Z@$^AaR>%#q=~c
z=rbf2i58qdXLEr<yytfZjf->NH|WOkW7BCWF##q$1O^wd4Bme$0ev_P;`a`aMZY_!
z{mP*K;~)R&g<lzj#2@_Gz_Wp81J4GY4Llq8$qiip0pTaRerit}kQ+JQ&WLooTa(ZO
zJR5j6@ND4Oz_Wp81J4GY4LlooHt=lV*}$`bX9Ld$Ufn?Zj~>u}`k*)p_sZvt&b`OF
z^ox@2DgCm_=eKp(@|U$Ey?$j=XpO(BSv15p{&!Bjy{44FJ1ZAgf+=^;QN254@6gz5
z{+#`uo<#LM+I^-<KVZS^Ic(v`K>R89eGU5m-2Hte?pYf3cWaOIlg>Z!_$F)T*zjF=
z*PRt-#MuWS=lqoMx!&5HqGI;-Qv?5fbzi&1mTf2RFFD`6(X!PFtgwS&i4Ie~dnN1L
z9S<nc$cgrHl}ky>>OVo87qj%J{CSJ`iRC{vpA9@4csB5C;Mu^lfoB8H2A&N(8~Dry
ze(}FxesSxcxwqd?<+(}Ua2Gz;!{2$o|IVpD_f&yO&)UIa9(ZR1&ttT+t%v(;;Mu^l
zfoB8H2A&N(8+bPGY~b0zvw>#=&jy|iJR5j6@ND4Oz_Wp81J4GY4LlooHt=lV*}$`b
zX9M51f#d()`#Ab-vpoGg8+bPGY~b0zvw>#=f7}N2R|fs3fBc`f{K}yGvrYEW?hZnU
zx9Hp~w^-F_TK?rOOu$Yruwq<F3c|1E2dSShFSma=@oiInT}NX>qrbk1>zV5{j+h7a
zHRKF820eriw70k`wO2T)Ih26#l)T{Fr2|X7y#TEI%>ld>9JxM^Y1%m{S-?&sjK1N;
zxj`+mUx99`{IZY=(9yowQ=6TL=l`!v>Hixp=e@F!)Bi6%LxsWB(lvV87fM6;so*;a
z)|8#^td+Dk3qm>!q-wv*2{Vxy+}}0FOWocz^<6rbhWXLrmpZ<fv^(f7Kee@=+%vjG
zlID$PDpxs{P{jmO#yJs7>_qOYbEnqM47YH-c{i@fE~b|>`u}&^s|K@!%|du00UF2!
zuqx(g_qsKd2JXzC5s>;KJl1J`hr0D*mA!OScE)FfD$Qs|x|MnYaN;aENvJ#B8dY#)
zWbXR!Q$Fk@mW_5wf&su<<ELbCG*OPE5}<?!5yoS%e+Wf1WE9U{jIwR1b|yjzCPs1s
zJ(!8|CpciX8J>}0{F+dsLuz(gie#B_O+!UdGQx8&7rmFo2Z6zjUDcr)jd|ThT!h>0
z=?ZGaglM<$;1)nuoA@3cBi-+olf@_k1=N#J5aIK5ux>9BEhrpGUu4hy3WD&t^~?-j
zFk@B$Pmll;9A%<?NSuQ+62;c+Jq^qrBIerJo0$=f!9dTwY^`Y{7Q%ZwHKRuH2Myv|
zhQ{+@L2CwyAl&j6b!M4{5IFtXi_Fa4SgbbzKDfs8kg#IDn0{0uNC)#M<AvBz!CX7D
zaz=yarW>4?J*;)a%UPE^B)^O6lZ4U^dP6C1#)13*5AKCAdJ512aL+rkLUJ%E!(BH<
z9TceX3U$nlu}11-2p%MX5P-w5Ly5kH7;qUk1_j66=;;O2gCDV8?BK=pJ4VDU#7MNc
z>I4+MAkIxVlA%s|%q=&HPCpCbPEZ8lVa_-Lq9Yv0M9FT?B8tobP69BjiTJ&Pw^`?<
z5d-19)pJJY-eBQ!b)-hA5fnlAenK3#P%R+KC_^Q<F^J{cX~pt+9L?M<4+_J;f;`NH
zf&JS!OU=katD6+5QB+D$a!>^$P5+=n2?cH%Md<fKG8`Ha;eGC!QJ@SJI7=0HfP&cz
zU$RsQ2}V5H7iz7hukv<U$eP((*14FrtdGA7ZAg3r_0poQBwfk*h6PPMylW7oi}0zj
z2rMKyg$#G12!s*c6AmPulPhiX0D5&{7{@iHUakXHT~Bk@1@|-7wIg%0PNoMhnE1Jf
zg?H>Ip7P2-0+W++?{729f8Im0B7CdETST@hzJpv%QDIaY;qT~cPm?<1?zBZYk^=;l
zQ$?rH1XNi=CxBwGN!n@n$+<~CQJcf0Z#C?6K<Ez7**}aNbaoK(rV(cscfsFmprQNY
zLDO|sw`^)bX58ZNypRI5ByhjtNx>HOnr@&RCY-WR_BLKus?i|s1UEwP6XrYXs`gt-
zAW?)L?ea=arH6Fx!mFAcLsdFs(dt5#wII^d5*fh#J)PKvg9Hj{1zk%ASpY7ELCmVf
z4`bR|>@eD?8xR0+u?JDRi=TQsbWp{^q_bcxx`Ni6vw9nGPyoOi!X-7zMmv%b-Z|%t
z&b?&4T%Z&vLGBMl-i^=8K^13aM6P8;g>SY0+{s?q_~BOSV2x4}D6)>1OajZmD}#S&
zF3-6!Y@#OT_5n}SkHSZ>vd@zQ6h?Si<JE_1vl!^%Hn?i1!j=n8$rF&JG%QEpnrUaw
znmx#afm!1mDaA^#X(9Z)%r`Ucnrc^Pm%qBhQFrw1Hct3i0Zv%VBHAJ~m929G<!eot
ziz#o7F06P>O_76yCSn4vWhII3AhBJ%D=t|`{x~^hCFeK~N3%%wa8?~j*~zK-%V^R{
z&a%XNiLdqaTK0v>V8I;7-i11wDR<YqTW@U~uKHSOl)a0?hLO;m5SYO<UI9r61}rn5
zoAjJqcz5d+B`$vNs+2mm+&2@Q?M7E33&1dd_z2I{jy$LU16gg6mXjXuYy-z@wWyu&
zx#?v(wPVdn|Boe8<iKGqAeho12L*QVg~3xpqqA0TU<EgrB0T|K(IR$wU@$-&z%3a^
zNO_7HVCPVNPEf*6@KMdwb4e;eNjnH7-A(6BvYTD(({Jm0M`(xZBW)(F(J>1}-i@Po
zAgb&8$ws43855u>2~Un`8aSR7x6|V0>j~hZ#UXLgo%LsAkUAw0zamxGp#74I^buZn
zS+*{H0U-b%2)#7sFG@z6=_4>hXz|(!-73W*810tU|G(6I@tuf02RoR30W?$c%we<+
z86}?uw}5ianm5RFw_8U-h_!?VQZQ3F#mDgIryOzUR|ft2fBb8Es@m({{-yoqp#Qx0
zyMt1V?sSXlWG}UG5IW2Qnul_s@G38)6y~I&z=Ufq!n1T^VIq#W$V{0K1MKu%GD<Kp
zYUeobIO>XvG~#0_iXdF7m(d7D+Z&$y05l~Q(E>W?=xB1N6xw~o9Q8-P(ylo*dT>wY
zJyS6tQ}vOtgTyg3`X+>MiI36IR}}zLR6`}B1bfb*c+8Eq71ZQ@BvOaH!w9hr5&(>K
zKu7yJQ-_BD&<F5l`(345xtkI%LCHG^G2tw}P`y#0$`CNu&a8eQgc1Z#|G^}pF4vlf
z0XP&t47$k7N=R^m>Dg?+K#Our3Hj)^iSYI{h8b0m>I9QqdjYtGr^p1DhH@z$IKk}1
zNHEm$>~tAX;H#K(w&E=1cFeFMb~?!bbVj1vZnUgu7a}c;7zi&^6iONN2ATe}e=ORc
zxo%4DRXW@)jt;J8bi7+Fn4H4*{5Kh6eJhjj;E)+yqx0GEZY9D(<{JO~LamQsh{4T`
zZg__}b8tp_Ef51R<sF1hWjo<qsjgF_2W`QG(?ROY8a)~(W<)zJ0|sE5D1-3v%S%a8
zom!*uhBpS0oahGYXKyI!Zp@D$*@J?*13D%pA99O80VS=g4(rmTeMOt~7!X`do!Xg2
zS!0e1>QrZRD3((cu?~}3kn-BMw~!vpl#K91(zOg28ZJqy)3I)mT=&RgPcQM{!VIwG
zZia%_%-|7!Z~?^1b}3SVOpi8;o#NsoH&mhfWoy<#IAcbD)kM)5pAnLnQ9=ZrV_X!#
z9K)9)tAHq9;P?!iDo7}Xu#EClnWz@+A_s*54&Yfh5~H1%D3Ed47h(40BAMv0ZJL3Q
zB=nS2KtaX@a%uW8J0n9*DiOQsfo=`C<Q@hKeuu7Sc8$_H7KTw6D2p6}Kr6s%2BcQ3
z2{l?PDB4ma%M5-rV#+K!BgRJ&gj-%d5gXtc&K?QRNG>-L4)}FwM9sG|maDMiCP7Ao
zqg@R=#A82uwYyNa1~nA83m@;6l^^v--*~I}#z#5`x4f|q;v*dCG+a{V#2FEhj7Klp
zUU2N30uxby39hJRK7sa{Rc6Gv(;2B`+gCTHlnRJ4IW<N4CI(n5SJLzH3ZzUt60H(q
zPtk_;5~)%InJ8}$CwBkYNQZ48PysNYN4EXIDp-t!nP<2m1j=_rv6!z+J%U7xjxZVv
zsLbg#!s+dDAd^DL>5>&3uweF_8?2_Q(y2oK8N}{fuEk`}4XMG5831NVj72--8*FFY
zbWiBKj*&Is?z+MB?sdy_6iIIyiw)<T+z2K3Mx5%kFHm-X#T~$Ds`hO3hXRs80Kl-r
zUcF_8WHk-3Gf*57iRhY5r`f#_SrzJ-8#zZj6ij$>%3`1gu934t6GfoiDMRfXr2`(c
z#5{^kWCg063R|myGkbmF0A7Tnos+v4teTDrLA(b<&}H4=n!y`*a5fwWVqu_dK9V8L
z9-|Kf_BzDAxpIc91_ILf%~M>?9PKNb@rdZOrlf8-Pz2%LRM2Q9q9Z&eK#rB@1he!s
z*NA!oIMz%VnV-E*M0jUptL6`=x&s=8Pzd2Y_mF{tg3njE#WUV@$g(eLLLOoI)+m4I
zP|TvlX92#5sVbx}zJxWR%|`)*Cy77`CQ$~Lvns44B7rM7SDva2tqg9<nGq2hBajCJ
z-VkQM2n@h}Va6Po5gU0Ny5nkDs+|Me`9c_I6i!fbdUCcpi7fjXlumR_1D!cN*B>jz
zsOd$F?04g-fE3#)nYr1sQEGH+v>*{oE8yo8n*gB-tU<sj7UOkr?<vdqsKls)d9FK9
z2G=VxpM#m9UOUpyP2{&+jdm5Y%0LNXl}yiGS$d>cGq^-x+-wfLbeF%!_F_HIF|lw2
zuf0d6><Jj__1q~^0h8VYR9S+})ER;86x6vAA|bpfc*qUZtd<*isfiE^X0MUR&55KC
z12Cey4&|X-7yt<qJOS2l6&Nsb+m{}ecytZ|oeBa5>uGM!JF&JCPV`_H7b+te3|=tP
zt&KQrf&!5k?J<@_Mj{6jn4FZGD#i#(xd}%y>-`{w9XxZhXImuHBW}c19oU$^7lCXT
z2u)3#3Yahzm~aBxmmbWTI9?n?0NnTN0%9P(BxqC)>~zh<=`dLu>Sh2UbAlO|P)-?9
z^X-hfz3_;})^hA79g;CAI~|aU4zruR+!J|qd_V_RRcbVpFoOLEXMUiVc@Q!RXcdi6
z;$@i<Y*ih1!&|x8xKJRaK&J<~WxWd$@L(DzECE7A+5`y@;9xf4><#V^ls-mq^kKIi
z@C-5Bo!jA*7#|um$wV<wP86!}3Y%7x<;n9jGwd|#=wHPxZ#<E{@=zGyC7i(SbStc3
zXfJYd2t`Vq{!uS4w>qNL?c>gvb~@1jJqx^mn0yzGE2A>Il?R=k-mQ}+C_IrR2f!7*
zrq(DIq)_`}2BIN+yE{riVl(cD8&VK!Y>2DWBz=q<y`5y*7g-jv9{rR+yYgSBx#2`H
z{Eb%OMG`>;-2d~qpi}WQT!ab;lS7q5QY|Xz7#n@B0RzN9L4=>F7I_84DjRsa&Hb5a
z7Wplm6uOE6-r;O?ZjJ#Qz@cQ8iOsYKNC)r%zh`tyDv!SP^KRVg^mD2>0o8y^DwwKU
zM{2Z^7a7Fw%|mNSHNei0nW5-KJ0y@<hicSPLo}FKZ3tK@sE7K14Z0p?a0g=>qaSrZ
zZ#V@-*71@_U=C!j_l)o%{XAh;sN|R6I7;uOH#sNHh~}klw~*DPIj!IoQ0OdVK_U#a
zWE4X<OwgWC9>4(%s~~>fNXaU}qpm$c40m&_Vud4f^hsb98;h}DB|EVQ7SO3ql~Z9*
z2$2zfG(f#C$~aXEq>7uc<b*^r!D~uU-Y(wKLP{`)x;Qcr*@bKkjPcROB7sR}aAP0o
zGcW`K9?VIGRsG@ySCfj4kpUnJ#WVC`revEJP8et$j-3)zjCkz}(Gd<aii83Rldy>>
zBcA>2xh8;LnsLw^+<Auf1oeOLcxP&~Y!`5YV#%?r5{%@5klpN=W=szlo7)*MVN_P1
z_$enjAc^5+;*ki&l2CBE*)7o;eMU=g(iefF>!g!DfMVvv)F8t1q=VOm7t?z}RD#LF
zo<nwN1|2i$G@EjwU23d>S7c6dE{1zrw+;u*v_5XalQV2BK3W6sO5s9+n{wQZ-s@qV
zOmlCLDcu2tPs_-&A~R0Nj4FQ!46cdNS%nNsr#WMW;lpWeT>%tE4G}J{E<5&Nj8HI;
zIXe11<+?R%3c^<n7RN*ifKG-G31`ise<c(V2*H#Xh_)-~4-|7OAi?y!ZdlaLa2q=*
zTb-J4{KHM4jk)ZPVBot6IViA;!@-e|w<ZRnWjuTFy);B%JFS{g?HoFB7sLVV!YDrB
z2G}WuyVosmiH)Daa~FM&gomDjIYf>GES9sgr<@WDrN;Q^*QE@dPm~{r9<oqMi5qC=
zpp}*zV@Ld>SkyR}A<?n})B%iZq=e+8dMNU495p`Ioe>>?0_|&xPe9<O*(b*`$8#iF
ze82-7%tp#y&#LhnEoGDnGhpSUR;5>1Z#-JZOcdS4(GrmknjG;F9w+TKU}v~{-Rcr-
z!GQ2i!h_h0?oe!SNt_|U<Mps^A;bIu4?m#|kwbO#RLn`HYUiBUBjMQzV9p4EV5Yh?
z&S)J<+3q;kjnJR7y3uKhO@MKSz~Gjw$ceU*>3lPe<U<SeyMx-V4EoRi_%D0EGAKyI
zF>=N1fs_QFJGEww&T4@d%$awLOsNV?xXK_rS6O4_oaoK@5n<Z^J7fB}4YB~NdUbR{
zCjhszLB%2=QtCx8+6Q@n<{49b0*s0R2)8_VUQSfSF@N-}aKxi+@%T~Ev1Zj|q2!o&
z8B?heOl6J>4b*Gy*u>~>^m?rUxJ%Fp-U&>&*w~Z?3)F!cZK^OZRXhZM41hPwLKSbR
zWR#$!457hw*+t{Ubg#0lbDVG-^{Rt$CF_QqSz<yInDAi$?1Ic1uF)nA1Gw-IIajC{
z6*dS57qNG7I4=?kJ%|~J(VmC^XgMei;XtW^Yl#|#jUf7N<4~vS{K0;313z@9exU3P
zC2NTqZTA!f5RNnK0jV=<)N-mbW<<N<78puFcp(9x*A;;913w$s-vD?VE4X9b$^`)q
z;PVfItPVeI2HLs{8(Ww7CCtBP(qD2zF188pBMe+D1kzBM0Uulzt1wVy2Y!?6P76QC
z)#*={P!nqO#c^iQ896Jlm$AxLUOd8o&3M*Rj}P?*xAV|9xG{nK5mXrfAABza84-?l
zt24L-5NA3KXY$D$bMcdeso4`gDlqDdzE1%I&PWELXFPjJKpn(DxNFP=K@jeVZp_T%
zo*>^9&8V0ahHjeZ%o#Ic6{mB9I^_lc%YeuX&xq>Jff)&{By*ja(uOxRQbBmvVusQ>
zyIY+})-G<@D9_^0dDjNOrU86c)ko*qz_Wqn4LtA0<yB3@YIhYTi5_ty3j#Qihy!>L
zj&{$l(KW@c0vo`ATsU}>8Gu-^OOJ%nx2#PZ(Mhhy{QrQkLtM{XJFN*j5gw$CDkn%N
z-~v)96hSy>2GJ27Wgy3<g#-`)aI9(4d}e<3IuYTWv0~LKHL7}butsAG#|#h!aMVq9
z=XWa+no~z=bj*FfU-VHx)<+pCK{z;C2*#!OSY_}B_n`b(XwKAWVW4hEv!3JT4V>JD
zvucOwFt{B+3A=bP&W=RpX3wN+bbMPFyTpS;Fnh|TB^oD6F&u5KH38N^87RR*2lK$E
zI)ck3Qo(}?Ho?m$sFC69y<GDKf!I;7gV~FeAcbL|=l*%tJoJpf%@ovmLrbKbhEfm?
z8)Vr!t44DlZ}(6W00;15o3CXr7Kt<~sHeG+@*VbFo!zBVZUu^$ff6irYY^ol9DRHO
zK7f%*fg^=oydJ-yl~9F2fo};cHH}CJ|CZrEAJX;*o99w>y&A0_Fd7H&0Sx)OI#EB`
zV8ghsCDbAVI=Gr!qbA(WQ$pu%P<zswE$_*VHy7NL2W<iPT)a#y!K76%3Mi$VI5|Cg
zC#iE7r6YV+n2|cSMmhT+G^y9z8m-=Ah&dJdYM`8=1oIe(g79jCs#vC!U~U0PqEv*B
zsz2)giYD*g5E>fc$Mr!LfQ#>*h<Q-<Ldt>K-xS)~+!DDa)#xWS{hrE6H8~qIp*)0l
zX~aNy&rH@PJ>_0;dcAI4(RpXi)Y$}7)tQ8ub)-fs`D>Wy^VLf=BLN!l_*5xIqCF8S
z;5gwpDtd^six=N$s)i9(lq(?3#*te7bv{(cf$a5W5Z)Psg%S`xq`SMrhg=W1aju`J
z$?@FTD0l_Lc?h3Hj}WAS@Uw>;<pCVPunOYm5uGJQc~wq~*}peZ$d&znzU7D-O7HLp
zj6N|}v99{3Sk>eonyo0EpyW_b6R_MU%0zfB1zDNq-5@@~O$NqY`P+=4<H7ZAckz~p
zQgFbk!Av8pomd7$3b=sQiqEV`Y64nLg^|=QK9_$3f|}~35Ch>o^NTHQUXJBSmWgMR
z?+cX~C5RuiH$H)&leUKxNlsW2!gF>=h$c)PipQMFE_4d0%KQZOMXj8U`%S%qUH}Y7
z8~B4OX(KrLCccLwwyD{9z^g)r*=aJ*`E20Xz@NB*@J<h=OgaI+03313V=QbS?lIWG
zTo5^?lv*W7Iv@aGM4viB`3C|hrNm2c>S7Oghvr@fKgix4ewROwh7=GEA03!5*E~j$
z5W<huUWrOYTo*^+J9@vJqaxdCyJEY1)X-fzNdLi~4Llq8&<6BZ2K|?R{9k*&GU%Vi
z*FQVZJ9h{ep3^>J+kfLlEQjd70b`(jlcF=d>>HQ<*JS*OB>b715|<nrK>k1Gu_k*w
z6lJra7jil<YolfyCml%>#j${(cI~AM;d8EgfTT=vn5!sHr<Vhb&gg?^Z!L?kP^H=A
zs+^rzT^R+ji;6zhxEP@VC^ZFoCu~(YjgQ5ugjcZC+_6xm(!e&%kY{i@11F{ij>-k%
zT_zW~OKj4~dzr!&peYwJvs&DqMqbtw9a$M&>20ZAF5W$RT$$#V7PmmBu38J3H%+?Q
zQSrTGBbA!f&8%b<vlURGL^TSSp)s`3&zKUWVVy|2WLL`mX)`D`uFXLwJN(4@cN0zU
zsM0InLQ#@zRSkU+m|m7T1rlq+AwlUwl7@O4O84Y17QEc%dEz-^wMX<YRr#+~-O6^d
z-1}EPacGi*N0AFu)83Aa?~{cBQwra%o+!;LYa$bm81m{_s}|h^zY%LmR!mE7W?PMx
z&6V3hf8PH}ZOccTtWMiVan#UfS}m5r>l$!th8Bq$9#STE&Z8URsFihK5@>*uXn|FD
z3%jI1l}}#U#3lKy0rtt-IZhl#Lqwuk%q`Bf=bWR&VsznRa(;@gmvKB88f=FZ2*JTb
zLe%5*VntOI?RI6u=A$(JvVg@s5C{#EB5q6-Hv!kvym<yHm0yP?j9QZ{dT{b~<~C}*
zj2>s{s5ynU{-v*-&cZ(DX}VCi8fzYWMrN^yobX#y%$DY2;+}0fV;1RVjDQiN=P9$g
zM-ft?w1&z%le~==r^_X0ovDPxi)OBOgNIuV9o|qUNHpQ#eaH;cbhFnstRu(P#t0kK
zvZl4B=lCWb!Y{9u^>QOpg5NeCXj2iae(@f3=8|DT^yb6~7P50>_J;I5Zk>~A2(~<x
z5CUIQihnLp$jT_$+~d=;0<FSA7??0?2#_1P1!@zFrV`w?GNo+-L$BPZR(|8L*CaKv
zmm%z{c~@$o99<#`-;48;cCX~V8gE_44#&O_%m}K{gnSUNT>H!cGzgzPlU3Ym`U%y`
zKuare`!sM_HEQx23Mv)Jka=f2Mco%6{@!DoJJ*6!o~QEXKe}}*=?k2yy@t^(K-qPm
zMzaQU`0naxw9815-L~oqwo@O5$oiP=eO#$1Gf4Ya@|SZCZdF^~AT#<y_^)yA-9uDs
zHr;f%NgU}ZE1b^0WWf?JaH&k0gi7$6rOcQgWI!XBm)~!}RBZwdv<^ciWalBd2YJlJ
zi)3vU=%ohh;IUMuRigHW<nBuE-4E75R$$#@mJJG6gbpXthE_O7y;Iqv#U14ye`*|N
za*%JzIjv?)`#J||q3=egsi@uZKkeXcUcQxZ3tNiT10`|i+9`SDfAYP?6P4?u1-|X|
zM(LY-?l)^$*oLzcC4T%qheo6~*dYuAiG8#0lC5<CdZ!${P`Jj2HiCOp;|;NCC~BqG
zNlu5CtylB4Q0Qp6^bZghr#+b_?*G*xPP9%s8LGGO_%*=a;<nymE6It0TWnoLE-A_$
z`t-8$0@(9jHixFXos<Q#w{j>=KZn?Bsez?FTYHUXx+tZ`*;*>!NNIM}PS~1Iol>J~
zibHF%18Fym+5qz8SQTF3v1%u_NoW`@MIpW+>htm1sJ+k!t)b<?wMGEUVI1>_Y>GG5
zFn^3qeLk7$FnMx^9lByjebRc<>wU^EH+)BHYnyxfmS5h*jRNzdq~oU=fwi_n%MOVK
zz5Tr7N$E`<OyCa+SDcA6(q(V4as#(Wk_gKNweptVyX8>r29p|oZTZXDw5z$}0-8x3
z*`)Wx%NB2ldat@y)#a}0Z-=6Opntjfw<o@-=}p-kdb(FvAAY@K?_O`}6&E|A9j(Tm
zyMD?C8vD3hkMzXJ>ZH$D)Bvo6BQT{3^dVccYr@J09d-Gx)3MK4eWaykKiAk?sV~Y~
z<?cr&@pP;)^z%M7M(t52Y^u?Ri%-^@`ql05aiLu7b?ne@%@_c_XwfFZp9ZED$)TMo
zS*7MS2+15!6~SHu5k8FT7I+(BC8!JD6ADiFyuXQ30oD`6)NX6L_tkV={$AxzsU2_R
zI0|}P+%mmW=sE5$*CV(El;AM=<jA$dW?kymlJ7bDFvz;6&aEo6R$Z;>ma9Y~o#9y&
z-92Ny`79^9vnDP;d-6)%a_C15YE9m87dn%{b`y8`$NRZg?YP!fRNfw2DTn0UL3jCw
z#r5ZwW6t}dZ6;Z3ct^dz>VAc_-A}CeL9t&nWF%$UScu(Zu8At^T78s-aDFz<$AjtT
zshf<?RoZOq`0ebEiyOQ0%l!9mXs8ff9xnn*ZPVMScxi4aSAt8$Hc3TaH8knIv&DC6
zPlM2pGxqA1BK0x`)Ux|I>GaTGsg!(|YS$s$T^#A0HS0)8+F8^2Jd3*1@8?>a0bd?P
zLsw*50&(<9ek2II(u_i)Gt#J=fkhOCt4P5eCAa*hH%gjs=b*=MzwkZ=%7P2TdnP7S
z`|5en!h$g6H~{;oDplCAJU@0!frZWl_T!*UdJ^rCJUY({f&wU!ilNvBE*fzL9;aPo
zEb$7#qWT^E6qWX%<+a$|+LI4#FfQu`nwgn?)@Ej##z)SgHSVUsr%%YG1{Y>4o~|Y7
z4HDm{YNid1Uq$U}r#JW1I`3mbmjOwi?pwX!8eOVIsnk5xTS_=2iN*~=hgI$mTHm}4
zT7|r5E3Xxk)xM^YvoAEH!(^S;z2)8Xmf2d)%Ps0Xp~D06UG??b<bvl$D^4B6F3ys?
z;qW<;#yAuB(KvN_^tz$BO8y4f&q(CvuhjJC;~=>|^9Ljet;hwD-hTk!X0~dzHVh;>
z%1)%<OvQNG_U@%Ou-OwS2&if7n@?&4y>J9&485*?Q`a8Chto-c7@oT9o28JfI3}#v
zEmQtZ<Xmp35bi3tPw~cwMqM?hP1RopywI$45Sq!q0HI$Q^zZ-iuQJxZ{Y(4JLH~8{
zcL(Krap2io+0Fy%6M8oAY~brRu=Y-7R=dbw8*0c6uJYQq7UO>A{BXH26nz^F?r2gB
z!~r}i`~v^andHW0Ut7vy7XQTNEgn*x^%!?<KzH5_w$Lp=xV0kjnH7<4cgylWb@FGM
zl-DMzUMlQU9aeSr`We}y&lLy^%RnFLRX0mmS=4Hhg8nm)A@Z$*@w*QTNw-Y%3#wp3
zLX_`j@;TE>y2x)c+R*lw=NG4%g2P2%6G!yC<xqzbU{c`yedb%X@)pq_x;N*uoHKr?
z^;lM$^j_VFr2Fh;`GGjN_cgSjG}Ink5@=9?9!hh{ZqlLRrC|8?lMZ1$*?CEJJ?TSb
zriOm1?$pQ%u=BHka%H|eW%fd07zmud*o5nWfj@XzH_aSn;*a*XZ7vod(b2!H(2ol(
zOtwJjV~pg{XQqH<s4J<p#+uSXozcwP@n2yf>Y>nsnd7Ib;UQg25CHI$ejR8PP!YvZ
zxYs;qbnZQt@BALQkE(orTZiL)Sv%6}S2l&#_^X;lLu})J=hWM4N(sEPa&aY?a`zn7
zyF>O4jlJg2+3)E|RNtfBXR7oA7R;W*7LE+WpK{;Vp#L28z7qE=jrzN_NBT*p{STGy
zmR*{59%M#b=pf{rpE5qzTf0-AEB3i{e*a9bb!gdk@@A9sxm{nWRJK}y6?QNz(NW4D
z^Q89yyVY-kU@;Qy&O5O#C7o#gY<OM_d{kniR~G^D030iRpZzPF{^%N?4c)%(Pp-7r
z>Q)*E!0;2~E`DkPe{UOT3odGKA@u_k|M$~st!W$vCUjNq;XNDJ+`#U$w#(ma2h**v
zoyK(E4T%hH`w*HC_%T?1VrlxU;`Qc`{qR0Vto`ff4Vu<o-9#u5%$i%G8(9<806WV9
ze*?I<fPpr+ZBaAw<=Z=N_?(ePM0|wjZ1MaplaiA2rd{V6*{eEzu;UGUXoFJbxL6S_
zpft`|Jdv?;lMF5ZB0o^~2TFqBKMB`U6Y3E*-0S!!+q_h(_YBi#%8bBd32Y}&+NGqE
zMm`??BZm0+*6u90a|d>~`#svfp5`hInZf-6PsX!>U$+5$_UZYT_rSqB!3oXqe??ec
zF*%YV9K=evfhAs)d>8bY`k$!+b~F|reM=j0{$DZ}*d7b=n@fnHmii;7H{r&eZ5)dp
zL=QrD_BUyTH?GD)C&7IE4o+fIFMlTx?m$}6-?@#gsrj>JE93W)m!6G4q7U7pPuxI_
zO961C{093E8G~27Kg)i`Z@8yVkIx2d1H%)J+yS@j__>HE&i#}T8wo!Xc&{=3mY&{T
z;8NEwb@fs*Za2yD4<QN3i3g!f{HXVxU43)~vDmYYxQ!3mY4OSVU=x36sb9Lz);9m+
z1>BWiI#rm425fFWLTvhWKaD#9k_}+ji#7IaS9U}5M}JTDVgH32C?1&x6(|5ZwaodI
z8~r_#Sf_SQ>qY^Di@+S65oQ8-YUBHW<Iazx-#5*j9f8>a94c3g3JWRM%u<v=x>V7K
zAseKo7GrJ-0EhXzNkHq*hh<T4T`=&qn>qTa*_#feV0xI1#nGMSMwS^oV~yuP`~9vB
zXp@Lf*KUbDGA(2Cj5jh!_KKT&zlR>CEt~eZuE_n8eE_yZ%I*1P>3>vf?jYf?8%CG_
z3;gs+zdNY?%Ao)HkN@_DUl~N(vdQiI>4Bg6&jy|iyt09{cQUiuMgCg7Avd_n>mL}F
z^5!$N@a9UReETMKRGeZU4&brCFYy1INp4*BwWS<pwMQ+)gLz}ykJQlybm#403*7>Q
zo3zAdRz$kpEzAGZ$)9afUYn?TsjyFVSk>9<XJn5)S0FGf1AU}d-7H~cQL9M`{f)0w
zKXNL2bjk4SlPJFuQNEkW=S(kY)L(bD+3YWmDNZ#8hl{`_j_A|vuu_x&lZ^CNo^RR8
zTSR~8_MFdh&iJ9$V_9v|dvzm{?z5NW2jbw~*U*B}P<wDmpg{$CD9tImNr#G;g7O`Q
zQ*{XI$<9l%>q#FfGd1*6b*DyFfSsQOlq>V)DYF*}!$9Et#U@-24E({%x@qPp6MwY7
zZF8{ziH`nlg??OUVX_5EA7dnsJ~IU@LtRO&HP)0C>WpUQ^7E4+^E?!KFmwD=HQaLj
z1OWhh+UrKEfQl%N!oBu6qjT@EeCPMbeN^T1+d3TY%i57%zp^Q`#$VMe8e$v&JEz`W
zQ%c~Sm5VFEl)LAs-W{@cXzVqA&VEl%qWT`~K2xP1uweEaws2%1{*?Q^2L0!#_m#M3
zY1H4XJ<?A)&#w%6@#frmM-F>w59;TN9d>eS@cU<atwYPUlb4sATi1EJSlMa?R@lL?
zL`Nxq%#+>+>{h=+g2hO*JMYB0lysu`v*CF$@KK44UR?yp18}VPefF<x`lD-nHgx;C
zKe^Igt6OOx0K-p^yZEUI{Jm|UEx4$`h13sF{NGQjwWe_xn9x<Zhxcq?a|652+Ae>y
z9Za{vb{f-tHzYE+?L%ln;KyM3iKXeYir1S%_QU%evG%W@H)vXWbrYdLFl%m!Ze&eV
z1MDmd{0-pZ0tVXPwnfdzmv8U9;d4eF5%Ce8v&HkbOiD`5n|7URWUuP<!HzfZp$$rz
z<6=d$fYLZ;@kGYXO)|Iui2Oj^A1Dci|0G;bO{hoMaIfQ^Z1YmB-ZM;}DKi3-C9s`9
zX_t~t8u@tmj~L?PTf4L1&K=m{?)PZ_dYY>=WCr&OJQ>dhe%%K2*{A1U-UA2k1Sd4Z
z{}o|*#pFnea1blu29|hH@?Fqp>VKvR*wI*g^et_~`G3h^V0$dcZ!RH*TI!FS-h>->
zws9<a5IqRp+25oU-nbeIodomsJ2;6=z5Ja(xC3cLf9E!~rsmI@t&HDGUV1hHi9U3X
zK5+vzE(O4m@*C_wWDH*Q{w(_$zu}%fJw6-wgEtW0wBct0{|t}ux2)mq1uk{{Qdch}
z<93rQ{}7UpoOlq*#E*L4+0{o^5Q{zQh}-y}ofe;*4>s|X`lb!Er-1+Q0($!W4+#y}
zhU-V1)W#c6=UBi2F#zm8J`D5?*OlGS<k8>Ie5ilt28u_fK?MrHPAzkO<x+poB-W{&
z)4EXr;UX|cXM~vmp4#|6;JEYS==V)?XGdUm0EfyIqryVUHM10DkS<j;V#o%msl}L^
z0>EMZZW7S?^I=&OTo(*{?PiXCYWAiBDVQE+V{vq+xshcC&sgI*(0;#b1KK3w)3sY-
zk4($hJmZZFlD*=l-tVD@Y0IWPt}AlCWFLSnk#c*!S^6K<nmb51?1m90zyd#g>aPs?
zZ~yqe-|#DgetNif)&I`B`JG+8>mK~cy3d{bhP(TXci}BPAyI^X!_9em__7Vy!(!)h
zd)o=i%HK$Q*R=1_y}ZMrp~1znL^QIVa)fKbgOuU{F%w6O@wU@S+c_3O_{cl(YY9cC
z-~@<rhr7j*!9D2X_k_Ys!Tde_Kb<}sc;^Q8-xT{9C;cx8CQ@=v)PFeqi&~vG_2K4G
zImJ<U?)w7>Y6ju!ExoR=R;#EpK<v`RaW@p|{#wm5*1Fz0^Q`8N#qS*<szv*Q<HfSk
zPA#B7#<Q0MRvneNro>q()g3p6Mi3rohm^x$5P8hg4Q#A(gL+!_65*Z(HdQw1r-L2n
z)u9?at@Aa6AMGC%M|l7Ko%aX%nHIkPkAW}sZ7@4+g6>=6Ga`|U2QQeWL1zIM5MV~3
zZi9Rw%)uGyQtz>rD;m+x_q4ODy4J=Tt14P&)@b?UuqfhJ6(M$u2vjcutsvEu|Mi4e
z%=&mEMQ;YZv6mnnOe_48Cu|1>aJMlf+BoBK1j+B>`T>p7q`xv+f{l;x-BI^+_Sm79
z^TcyTkXtl^W-uvva2NIVXkNpOz5JhMqqY%F{T{_WGtbX=@R>P(Ta|ua+;Zc`XThDI
z2*SZT0MQW+Wa>rDB8tp`THt%p^9c|G;XgJjZm5pbD7B9w2;WbLODa?g$TG@M30@ck
z=}7V8s5N-ugu*bePz2_}!2WH-QVR{fNGWP3C^>7jC}lrwKO=VYT6R==i!JS;CK29S
zsTnnTW%<3*jn<d9cjKhjrO=M9EB@}xd-rig?L6n4@fG*rm$dm6yZX8&K6mpzSLr8~
z45kC{8#*j2B``YCZp{qt8`}G@BpN8<(cWtBa^hB%cay6rDvbI-_`7>N*T3G`L1qYv
z8Gyr<n72k=V@=gArcF8sApLjOK(&7wM|1Du;~Xe|<b9Qon~srgcgw~QWX3HH&kHF~
zO9J;Ro)m21^_t9Fr=OR;!Z@w(1UIgf`3rN<z7YP26`|U9;h&hH)u`H?unG83b5EHY
zH&EQf0+NCt0O0Z=BB8tZH*VpIR6RzGUeSLc2Q*Q@1@uDQFG$4$;mCX1M)lr}-!%>D
z^Dg~I=ess=c)#mljnY^svWpKZLg@qTao@XvliP3<K8lr1ZzqTfNCPl@Udxsy5UEbY
zUkV(~R|E@)Cr9`%UCW=-_D|iIchx=6JqX==KdI8))wcAMyp?nF4&JbW`#XFTZ{5HT
zKX2!=e0_uGBRq^dH}IYZqN{d`y3XFh9NP&Cnh)jm4J_XmCOX@V+yQ{$^Md#Y&()4R
zr~m_5ZIPCf9`9@e$7{8yo$$HoWjeKE&5FOiG6jwt)&hbl4{}go7hf1WH8eVF^#)dO
zgDKJz;1w-mrw0ZD!~xurafFnor~!5k<>v$?{1kueZ}Q8=N>Ea(P=!r}{!HMu(RYM)
zxW4vFNuCW1EQ%uU#?d<v)dho9@RGzK$%1x{RR9j);^S!)T=QrB85xvL3B<2R6*g$U
z<RX29*ImBU!jqgffH!Y6W(X}_E&R$EiU;!*#W&=}&Y}%UWlHuyO7)a4CPKyKz-}SL
z&v^H`m0Q}x5j}{~uMGP4|M=JTYkS-4-~Ofj=Ai%n!tW011-nJXvKPu9hK_@^=7^5I
z)ed+tk6WGl$*Dbi!nG{Itt&a#SUD$pbACkFHo(r9er|&-0IOadozMxu?O-a3AY5vc
z(FjI6v^Bz`AE<0j@d>aD6hOG;!SiyWDvtT1Z-pZsZHv!uB~YLzj42{pikDziYW6wD
z8E$Mzj9H_vVgMdWP3}q>;vgLE1SU*vT?Al(I#8pnHVo|b1!MrcSr)2zk0ql7C1nT=
zZf-9JvsbDecSX@k1jDX389+z%QFM|?z?H1S91tAf!EG%~hyoLC?ZbeH0YVMes7YDP
zenNhym~~FA7y~xArfnw*@8ZRMkP=A98t^j`-FBnVqU~ATXko-acuyX!DPsaSqefvP
zh`!tSs5AQGG@tTQAIVwi043ao?`|3LabUXhsP~n9EOX`}Gt$LC48W9k5IU9Z1hcSO
z9jVcSwqU~PAa!Pq+Bnr2GooE_3k;<o+$0X-8X6dbYdWWUG9t*I;Tcg07_btj=d7T}
z90*!42m$~OiWf~F?64c7J_=+!d!vT?h!*f*dKwuBPb5GCxfR^8ZsigH2XNI9q=GpQ
z`MEjwdQP`II2(`@1JRx(3|V6t+s-*_Mu=OZq5@(7w#wivvI>Zj0emG;_W}w4_^2h2
zQo<D&v1q5hQ9yAJW>{juOu}H(K=5GySv7eih6D$Ea9O6pTz?KSGQ*M)@L*0cmygDj
z5QD)!<`q$aihzrP7=ZD%UC{1FREV|{OVF~NwP}bbU88)oMQ}!}5Q;=R+Lo73#29h)
z<8l3e5!PE6%k)UvONYk<_D4|Z00zpG1TrEV?VKTzA+D|XaT0c3>;>kaz+L#tKB!|<
z4+p}W!9+S4&t6D(6EP5ekSHm0;*5w0!qF-UegTno1y@)}^z=DnMtmKku|XYEqmN}k
zWQJ!%3(SETiM>Soj0xi@kenco@J?Zd(mK0aok`X%ZrPmY$!=I!P5^_Vk^N860el4O
zP)XsipMrVBhx9x=@^;e!zk&l6Ow`DW_)x)IJF~i>c6xLIBm!U###ppn(~UXmh#Qln
zXE!6WdSn+@p)oD{+0$Yrcw=XyzWO7ax&gtyo&yW!f&hT=t%<t{$-zwBA$;@JS2B*d
zkt(Z$f(cI!YAusOt?fYn-D^R`O_y`SL#VY*D6B%}K*!l`WJv%g5^(@8!qM*eHM*wQ
zRbT@+kP8QIG6N7RcIlBY`j)kcBRa|Tn165Y5Z5!;PHVzWga>J($_Ww*xPVj&MGy{}
zL3D&i8OX6|Apryc9BZ01pP8S%PDFTTWUJ=ypsKf*(%Ie0jfn@M0In1m>-=tw<T_HL
zWA6L?qK^WyKFUxDwv4-QDL$$V{@@;z9~+x9by^sx8%iZ%E)3|p=J<@j{uZc-5}aVF
zJRdVWBUL+)Rf`|dQ!F{PRz#Z(Gl&O++X0lYix(~JNH~`b_tH*k6fdLT2Pl}SO1WM@
z7FdITlNK{^Cct_r10~2AIg-6bm;ftdCnR{ljiR2ZoSs=k=_S}=0MBst%A}DJx;zL4
zmYYlTND8)?0U*`E^+Gt%DYKPI6i!tTcHr{o45wDcM*2*h5xALxI=6#J2yY4~gm9R^
z-PBnTjW*IsoG>NAO^4!VoFFGANSNRWFr_L`Va9*$1j+{MY3_Qh{fu?($lR<0=_r8k
z*aO=~L>MarB{<c?k0#1TI8p+903($GM+!T5=9;G`9R}IrM&EFBV*XwPvSA=}G;t=N
zkt>We5RL<(G9n?-U0n1vOAhSBzMvgCNP1&J;7N#q3FTOg;?)8^fN#^G|ESk$*vWu9
z9u6aqiwS<u3*ewJy2`9mgUm=6{R3(Uxrx#@a76VQ(QvXD?ncgreUN_;Iw%0LBHZn6
ziK6*-HZ+*MY6<pmowl1a@UxpeEj@yx-#bJRg!fEr)wwmwDhHvdO3kg&e$#rqTUq)p
zJgasu=_iLdGa{(~FX0mG9gu}62*(+<RM#=f&QLu<y}cSO8`z+-{w`me+DO}=9#1?8
zO+e4DZ$Y^+w1!%}Mw><sWUu#y@Xn|#kOknO6%ur&8&+ll&H^yO636o-p$Hd{h~|KM
zZVaXG#$B%&V$avsLIkxVfrj?j#gV|SNi`~IV{G(;_<@PCGOk_(qaE}MI6%R)C5Buz
z6KM|jCw+*uypvQEM)**`Tue`q<XG-OkANRDM#|`0m3QM-r=L^B35au89+L{D#_LFp
zn&j)J==0S}H6sBU@OY~l?d;`9YH$iMFrmjeDJl+9*u{sbib{x44$?r>0gCMMkxqp}
z?HtHn-vPqmj3@!&NT}qOU^TT4;CwfHvKbk3$cHS~5u@N05OV`AW&pA(3^W>wAsi-X
zNt6e00K+PX4+a)`c~zwtvu&i1EBpU^%MmoP<~winDX@yU$Y@jo%+WxMM?307Jlc^H
zZOEwt6-HtR_d=%xXYLw>a)OdW;vMK#E_xRpN~OBFTI?uBcDt}@C@*TC+6N*iw2Sv1
zhRnfr-Y(t}Q3?)NHJE9HwG+#LNC6knSR7GCeDupbOeC$gi%;wC0^RD9BrG|dBUTJX
zyPT#5n~-O^F{=mGQ4=XP$n_WsAiQVJmgdN@EU-+R3uXifb-Y`WIr@y2;Dwz}I_uxp
zb*@Rl=S`91IORge{pmYMJ`9EGcm~w(oV<lxkqtS)^s*J^K{qB=2f9V<?0G&8q=e*z
z<k2j61vF+Y`L5$eP6Kcz1#jRFt`J9X^i6ILM{HBG^MF@{3>(jKl$vCf+i8RVkqKak
zOM1u(W|rO$Eaqq0)KC?^4}()ti6U4)XU)JA7&udm@naE-+F8yN<sbomgNoU`Oh8L8
z#L0vl6xhXK2_%#ogsdFMUYu%(z;?nerj*=LRF4syCJ_`zII8n?cw7S6aiCxab7Tbt
zVhv1=W-NP}UV;<@0RSVqR_O=IKM+7EC0>HP{PtYb7t|aBqaW)+8zcdEu^E(&hd~*4
z;i&Ohe@0|?iqGL0NeOWCy(L$0BPG=+-eng~GMjLvFP)%%s*K(N&LU2ZX0L0obrgg#
z6hQdUkhhTe!3(C7n*m^m$J}naiW?k5UdFr3Syh-9ByKaUn4adUKDi9ZMWS*&>|lXH
z`R#YSqynb7m~Ql&&R(@!)}EFKfncWkHO^=qO4;r>*Nx-HrqkwDU6hT1Ik?6J@cvr~
z3GL$67(VHD2en@r^xyySKVJBiK}f#*-#sJpgFhR1Ht-8JfIftB>&$8w`D?KuH@M2{
z|A!Km^5!$N@a9UReEU9rRGeZU4&brCFYy1INp4*BwWS<pwMQ+)gLz}ykJQlybm#40
z3*7>Qo3zAdRz$kpEzAGZ$)9afUYn?TsjyFVSk>9<XJn5)S0FGf1AU}d-7H~cQL9M`
z{f)0wKXNL2bjk4SlPJFuQNEkW=S(kY)L(bD+3YWmDNZ#8hl{`_j_A|vuu_x&lZ^CN
zo^RR8TSR~8_MFdh&iJ9$W7)CE(U064>}C0ZIJh@-vMw>y9$eCCP=Ov+;FR5@L&Zx$
z`HsV>I)wFP=Ox+oqz{#u8v3caQzI+D&d&nMmHG0N*$ahXAaMR-6Rrma{@`WZG;@@R
zKic26xmbWiNB_1$KQ6Q|*#f1HF_K50nF5xfuB6r)Yf1}sMl*By`N@!Z9tu5}Iew}d
zZn=Jf0DwL1b)!{4MHENjUi+NUx%XJU^Lykzs`B}59gg>9?MSa**%VsiuWA+zv5o(o
zQ*W;+CGgJ5#g$;n-E&m$4%s_2_L@Iuzo#creUEmZsnQQvFnbPLI5H4_%6(sh{&Up(
zO5C$F>hIPb=_j4=KP9`T%SDxWkQs5QgOGE6%J^Jw?LqxqvCpmZPn;=k&*rm}W66o^
z@3YHRE3m>2h9x>m`D32+K47={Efy?BqTP8X)}^Eq&7Td=i-C_yZ1n0PuI9zs-(vqs
zv%F`g&xUT_>nB&*YjrCP1Yr0Hau+`}fxov6v;`M6xRCk*ivRm*wbnEa0~5L`_wb$#
zY;IupS=;4rwu9+b*iK`*?}kJMw|xjr2>ci<Ke04@R`Ggs$bNXABi8=)^9D_8uWlj~
z2xiSK(T%K$YJi<(fxiJ<T);pZ+_tD0`SR_ZH+;^>BO*S+bGCT?mPtv;dDE_QjqFvO
zKG^XFKD0q8b6l*57El`JES|{Nxk&~W0FfW4`vWDx@SlY1sR{K68}4=dlWktA)q95N
zGi63#vIMphDD6_xNh2Q*{}Drcd~0_W+_?ig-2EQyUr%$DhRoo8fhXhHz^~hYKKu0i
z%X{G9o#2FK_`f17ub3Q35e{M{+`tkqO1=yFO#RPP0XrItkG`diIR7sh3~Y}D`OPK7
zP)q%h)0=SP&Nhxk526R5JNuio!W&m(p_5>~eg`M9sh7VK2zMZ@=<nRd*3|r2vz76C
z$xF{hAkl~J(I;-8#-#u_QhtN|hm66i-k)VZ<2T&Xr^jalwt?XZNA7@IcKlpK6z6`*
zh>e7w3B1=Be@jnqFL0^rm%4f>8Mm8c`G=5%<ivweCVtfW&aOVXf>`WXN8H8-?X>vh
ze6Wc>wA3$MXKS1P@dED3FP$n(LjyLqA0ak<yPw9L0Lcch>%|&-wkx}#`J=xl`>_AQ
z4HS<|g9;RYom%Gn%8mY>Nvu;lr*)$M!bM<?&ImIBJhkzCz;Wlt(eIn)&W^zB01lNa
zMumlxYi23RAYH0x#E=bAQ;RV-1%Si+-6Wv(=fkonxGos@+RYsO)a*?MQZPNt#^UHs
zb0f<Pp0UPrp#6T=2DC}Ur)#&w9+{T0dBz(#NDl6%@YmDN#`CzY$o<{PS|a84e6#dF
zsx^1eb=VCfOn?P``qW<;^gsUb|Gec_2L1GSb$z)TdQXy5^M42Wwa?h^Sl<_Q`3K+0
zmq$BzyS}Xbjz{1tn|jBfzhCo*Z`F(C57Q5Wx7<0e;KseOx2OdXYtzmU`i<t-jUf$r
zTWYxf;T``}-4BoXlzTStY+!z~&q&)+2|n9<Ht^03=tC#RFMR^Pr0sXK!I~oc4*BgT
zt2fcDZIPQ3zNeF$2aDekbjW<1*86i;b2|FZ5zd#M8NpLij&Kc>pPNf6hpzPb9GWvX
z5`Vhsr@n8%ZchV(so@#1JQRtz+6M8NoLDgjtQkYC>~p@ffoGjBt?$jr^W8znj7w_>
zM~(0DOPv)F<*ZQ48V?vhD=%#0TQ<*qys?=#N?+Lelk%k-cpj6VzcJ4}c>6v0qgL$0
zsNsHCV}(f=M?(5NW?iaATSX86FrwelIx2aG_M`D^;Mu^14Lonk3rl+?<yrVE+xxA#
zKYrJLyn)|(=Wgx$xeK?h^AY<KHz02k^j0~;EpF$P4ZV_krC!YW$=b1xZF{TnE%J6N
z2MIq`_@Jj0`kDr{k+10|ZRziHuHB1#BgIQfiICz(i(q2M`(WgVU9xWzx>Wsplj7$8
z^c{WUeBap9_dc!P+S%q^G`gJT@o>qxcicBh-`nT~wRw|-%xIMUllKwle1om4JKnF=
zHP*cEh(y~mPtQw9r>aaSqzh<C>uTPT_*+Qav9_I$cm9`o?-*t1{v9P=)YJ}NRC&^m
zH}HWiONkE@I-C^MM|czKp=POLn4Fb25WOkmI`A0J2A&Oo4GixVb`6tLy$ss>)X-TS
zMFiJqa1ei1K1P~OP;FclrXvO>I7BrT5X-gN%m@-14{-ed>1J71PV40;H#MTfmAoU(
z)tJ?hv+873>ES&acsB4u8?d)=ET7)e>BF#p9FBSBk0<FG7HeFxZxgyy{Ym<xHee5v
zo%M6tqkrV^Hw~p*`le-lQ!h7d`H^`x@ND3#H!yzcj6VM?WMDqc%Rg`limtQ^&-?1d
zJk5bSrMGXsi!<`YJN#A6e(@fk);_<1+n?Rr1wX%&r{c4L^9}r(2i+Q_)9f@}2`0w3
zIfos4kPLPlndZQeiTVIb8wd-R<=(jB8>N<Jc}Rco2IyA?{ri9XYx}jm?e%Z}(tdN$
z|J?iCL7W&bB1K4FTZ~tB&#9sSWBFL6*NYa&W;pV8>{?S5>R^c$b2pJT6R~$1Vs2UD
zB4{1JbgEbc16KF1Ie|xra_wYSn6=9NsP@GNsYidozS_`&^cf|z8qAPY_ZA4ge`Zpk
zPU&b$-lh-P>uzs=to6JTX4FEexpe8BR5=X?QRAX+EWUmQaW&U-m3iaF=i2v8I&1$l
zGVgwEm7KTN{8371Ba4Q>s+k40z11>-M>T?bgBiPqj~@)gY_;c}ik5}D_5c($tP~F2
z8QGb>u1gACg*Ie68SZACzJb2F(uGoekR7rOy^wQ2;mx7=2WCg5&+r>sqXqKotkxXQ
zkKY_G5Gxi+Zr%4n5<#I9qP&Q=3Y9_$He&lviYofMG2TN_#G9$7C^9x7TaOp2@|8qQ
zR^ct|q5{=Ed}$LbkBD21-bdDUuwdRv_Hb5@<P0=@g?Ej0Yt~uMHD=Q=v!Ao@%Lt2;
zO0V*|<!ZC)-0*QZ<=3QH3?Oy-2|71_lF-6F`8oBY4BE?j!LY@nq$If^4>V5{wk7~2
z2)5xyis9$ym}x%kBXy}#7GF7I^~68fj*k@u_Uu&d4(F_8s<Fa(p_jaY7le*dbCKEO
z-1KHSQrn5WLsa{xHvHzevD4f#w@pDs0UzOjeic3|gBx6_s7sZ@6JFDFnV&a!uKd{-
znmSu;pk6Z5a$B+Nf7ZJ+z(Epqe-QpiNH_SATEDdj?}#2f9YA(APD5d3j-ku`D|g4+
zMb&K_6c67cFlC6KqB_WU4B1=mdHW`Jz|0L<!tXe9?`ZRDv#{10^&$)01S%XAlM@41
z>_Iz940W2zE6B~ng!kG*JMS%aRsCR{S0(@ayy?zPz|Ib0a<l1~h3PobQ&u>geXq4q
zo2JzF=fN?zhfDb4rUl}hSzi##u+w=H;E_$kguLsud`nwn6Kg-C)}FRH(X0~P&nXxH
zir23|dc~~83GI*ErvC9nB*yZas`t@pw(BqI{gwRdm48oWxu0@xyv_R;dat52`LQ$H
z=`cyT)j+*!Hr;ngUu_BJ;d_8{J{`$DE>)Ks@ZsB#``jsh@(GB-H9oWv+@;n18d4kV
zj$UsOxucO=G)|_uXJyjqiPlLc!?gC4GHi~Y!_VtOuH;nUZP@Uv?aj8^g^zt^><Sy7
z?i_0yFr8<WQ+&!k1ycIQoYS@@!j;x0kqKz369#}p)U@-afY=Bp@iBP2e{Z!-wN3i9
z;n9t`pr%qGGfuGn`~-n@6UuW}l}>D~il1P8aVNy<;TJo+Uid5%F8eGo#M}OYhfLox
zBK%2swGdfAoHDT1c4)~V(V(}VOE-DB3p{~8C$f^6I3vqd*NCBmwS+Ym%Lb~<7XUKn
zEmPV+`XQGZd99V@?0V?sbXm|Q{l2YI_wRn)*X<80ef%!HRFH1#=EC$z*v#6d+XL7C
zgZ{GD!|r>qTicDSBv`X@u4G*2R5z@`w8EPwrz&}R^K3%%+i8%U0u%uwJST6!HWya8
zVBWI4+i#A|a!YrwicC-Bncz{vS50;{$QY;RX~Z7?2R7jM!Hv&1K@Y?7<1jro6OaZt
zjwWh%mYeA*YRM=I*C~^MOWoE|7|q?Ke>CL~@UZf4jnJS+y?aoMWAc&EEiv!r7PUeI
z7f6f3v_kU}kTrDW{>T$x3aj1j@>g}gj7PWq%2N9nqTtsaShDV2^Sx?I1Jop2M;-k!
zRBpB1@3n0etO9y@B-+@i$O?x<Txl&qS)O&;2iM?!@qh5Qr#6**F@4heqkwk9^q#e`
z7M`0IhfnuHYR3j%RHyV*VQ`1o4b{eSBKd|o+G|+ZmN||3z7~g_PnB;6`Y7GnSyyUE
zKZ+Pz*>FleaHiG{xW}Pam1L^p8ebEghMfS=>N$`1FlXCMpCYxQ(HG81v442z2d|Z|
zbI4%0UvB!i2E~Tj>|Xr-uvlST9Wtv*6;=%wF<2my?zKV5_FRGgEI&VS=goiuD3Rn*
zYy)4>P)@E`RA15BXsuE6&z!Q+J%xcS66rcHUUpf$PzhT+7{qpHXCd@vh?+YraUAtj
zOv4dR{)mC@)a}K?LeG}_hT1&lkL^6ne*0{vpS-9d-33}pj3*}szR8w*xq7|WdEQGB
zS*;v-Ngz9o=Jzl}D7$_5^EJS~&Fo{4HN4(6%)1_AYX4n5->W+;dU9{5HfjEv<+R~*
z^H;X9@+53LEGu%M&E9_izt!BF?cOo^rf#%giyL<wYU~8^KMr;>Pq_PoNJf>HG4l9U
zoOCdSP3F^&lhvIXqA_OhHzV}BgW9hQ`k(*!U$6bjps*Kt*R8BdcUiBrDYgJDJjDs4
zg)2xd{WR75cG-SaDWoHMY^W`ra4mV3xwGz9O1T}CrG^VKE9C_7D8SS4gz6=t5}1|(
z?r^<nl^zy@`g7tC>kPNpkR0lazr>hwsX?LZsf_F?MTm5zH0g-#RwGo=Rwe23HFhBS
zK`I3acR(AVvB~_AS1;Q;=vdX#)GRm8C|jP>JhypjCCVK+yf%8)3PJ~FtOsbK>yMJM
z77|Onu}RXorgbeVRO7+)5~pf6VbEsR@69*@<!S<l<f8Y3#1zmP%ZC?S&Va;Wp2nLI
zO@;v|{@%caek6tNqWo#o=3N?L=v;2p0FG|#XOJao5A$6gvoocNLaZx8N)@2zkM=-+
z^A;MLAbClYY$eku^oL)ZO%(ewqWZGo5I3Hklo-29jx<74QkY*A#w=oMPh6;{vQp54
z5<TV6dNu!0ya%_+(B!%_o>VSfEL&UBId$?_OUmn0(hzwz0#_<>B%nK{D8!5z@u-FL
zQKM($l~>zlnXaoYW1j-4x7Hn9rY1q#3$2k9f9Z%&sR!ojs<Bc|WxN_23yF77`UqFm
zrqM}=M&Kp#;S^L&xK}$^$ny@ED;^isMq6X;4>fHUexwMfUjxqRlF}6Fl!%(}5tTdP
zepVEtdSOQiDo)~piREs{4H0bmU6w>rHgPfdtrS#0W@^v4URZuJgIA@*1-5wHmt^7<
z6_Kg}W7<VXp{I&pAxAeG5H*kIP}Q8)z<AGy<GzmB^TN#)9(Uq=C3b~e>Z;fD+Doqe
z3~FB24eDIF7sGr#<p#2<y=!gHn&rtN8C~v$WUsugWPwswq^`6qIbN+9ZM6zI+Jy^9
zgB;=9&gqts0#Xki2)!z@CbHa5zRFESnL9YiO?-a<ZCkWBs@c9ene8fFDHJDpN+U?@
zsL-{r6$)ZfhzQ?my>JH2mrZKwcyQVgH9M4Ic)*A`{2|7OC4bbFgi-j&Z*rALkwQ$K
zgblv$+?-NbWuCMUiU~N}BRir!>(ELg`vMYrn7<bk-Bw2SHziX8S8DI&F#9aUcM+P@
z96E4%8;<_yiwsc^AgKL`=E;yJ;kAtn4+(4fzNNxDt}D&iy<Xwg&k%Kfl6@9GN%t_)
zvE?~&Fi_q=E8QxW-Y)vrcVZaf<10YHY^>~!Y3>Wn$!6H;ywX4{Zn((6xR8-vA*Bg=
zd&uMDV+CQW+RC%l9>4a7U;NqS7A2FzT2)adP;#SbLF&;`IwTeasu$%bgAe5hBuEUC
z?d8aDM_BbWb~rGGo`&eDJC+}PTaMv&o(i84x{zB3Q=X}{I4@#>w@RTWBfiwnAQ7l=
zB<|QTG1??cI!8lAK0m}v)@FLJTPKa3PjMs_vyXVG3EJuK2jl`+)I8`$s?r_WQ1{4+
zehUrJ$<N8(<c*zQ=?!QoJlrV1abqQ*Vzd};Hb1l$>nGorcyoGau%g}&y4TfibPe<*
zE%kTS#7Y0OBKpHmt~hj+*R0KIe3@ehh-KL6)Cus&rlCSI$7(JQy-HTzX~(Ao5jwrO
zM&MgfHz{`sO<@9c?sZh2P=9~Q<Fg~jkK0)S&349lTbvhrU41Wo7;|u87ppud@M8h)
zoF1F$Ro-;oX1h%64ZxD^Y^gv|p6=~y5Z-9(I`!lAUfaJ6`k|)nfdO~;`qF<kv)&$|
zWuyc@IKjySgTe?0oqu}Du^#z~jUmJ5&Yj5yUD_8P!z0V_IsX({$}2a$lzCWxR+teL
z=+kD3b=FP_Oy4RpkaG`vLqU;fvR2uM;#-9=YAo(jE)1z$Q{z{|Ou;kGjO_#$q~+Kj
z0tcn;w%$Cc(V`RZn@m=(OL%2lR#~#=+HiEXR}93=5htH2D2I!=qag`5VGBe-vWXx;
zKE)L&lrl=NujDGn#;RhXrHmllxY7jOd2W&06FXrPZRGTaCQ5(U5AtkUKcq||;(H$J
zjyaW|iK@zniK=ENcFP1c&D?&@)E>i@RSpaql3oFl!5hRlA2Ir>O~SJVwcsVGu+aFN
z+24<}_Wc!1$C{Dg<j^KgU%Vl`*exR$y=axPRDKswilm{00w2^_Q7;9%K|_L*65bpV
zA-3XC_+TWLJE#;nuh^+$5nAkw$TV5`hvy#pDu}ue+s}|ty3(*Mw4@YZF=oMX=!||A
zcOf%)P;>fe^E(oDPcgw)su|anf<Z!Dsa^Tm6+r?8SBo9GNI+c{Vo*bz#6gl9UaBZh
z1!Wbl#%U;B)1Q*d>D;_tD05;4AecqzD^RQsCd#rh9}k%WlgH8>CAWKKWv~%4V6?nU
zN8f8uD@Ds2MT_hL;-SmsSSd;>6A9KoU~wivXP4+!j7LVskG)p9cr~^5qW0var=kk>
z!)OgTuJ-w;HDo7u&>U*j|GFzM&t0_ondw5OR4};G(gYNpE<-A&&mrGOTFV!_S`$Mb
z+09;TSe{F{iTaGneU<wislblf@}O`kiPaee%v?ze)%hJY`u;YnWeQZfK;2LjCZU*g
zCc^!iW%x0nnL0tN0<yqJHNR=}?TE02E}z3fsLcRr*YGATm!;>Rb!>Xec_q<g_@QtH
zp{_56Ub(jOP7TgdB_{XgBYLW{MDST=b}M#&y}3+4w#!Wk0iYXRzxdG&H&1wqwwdJ*
zJ&0^|nac7KJ?Ib_08X6Q*?eSXf%^~{fOkRrkT2?w98gAh6hsUQxkxue+m%^?Djild
z`dE+hIMxH*vZRBPib}p2Gr15fpgC^|X?t1XLDF&haTJK+RiuQb&GL1PjH<`VzVoC;
zPR<#^i(L&_M{X`>o;-)GS~KWHZv#2xJW_mHj2d|TRce%As^ttZZfnJLwc6y<`p0Uc
zpE|w8iET&+S~E8jX}8pURm2*)vLRWsDTGc6$fMhZj>KFw8T%P2o!+~zy(nKEg4c9n
z_HIx75kvxO(;UmMX~cH#x9wfUXOyp7uhE`2#z((1w=9-f(`t8Ul%4=0IOxWhcIeZ-
zn4sxV1=%JT<A%v;cY`0VR)C7o-|&YpDUD^aX!bZOrHVr}@{0>meJPiNqc<jN9Kd7t
zwttiEO)}R0Am^@bQfMpNG{I}9*S(g0mQ_3{A>|$8l>?PvXyn|S2a)$c8zKFX+joAQ
zD%LI5=#+c{?5OBH;bSK+w9wPKrQ!3%G)9mTvw$W}D8L_sREZ4q!V#3waaK0<!y49w
zo{~tJ40J_g1#xLu0ea^2<px3Qbi_E}Y=z}xx}Q9W;iN3xaf=pXwkFaMznB<G3?{3!
zjFT5L3YMnNk#otjihT$k+I~{gd0Is(YF}!yrZ&Sz=({BnYieu)SmgD~Txm^H+g~9Y
z)y^S@SH+hhK&}M&FrI9Op0VPkYiisc6gxpZSj%ucNCf_l>0Kl#8f58}O0h)Jq3{Gq
z4uV_RBznnFjakWtlDihQm21`K)JTr@Ix}W%*xa-6t(u*NkLOu@<RAJ?Og=94Zps$x
zl;gt3(OU5)6hBxuHT9BAd+yL5o}a||i*7{RNOFWqw<l7r+SHJ~+%O=Y6h5GU5P*X;
zVh~P4wo?yI>hx|k4%|kuCsC#(PUBO&?rYW1C$&SZ$e9D#BTDt+F=!$;*yJu8qF{Sf
z`2)7X8B+G#_ZFZbc~rUZ4cO~%V7GR8aoo1tRJvsZ56sa$jH`kO_A3z8VPXZknberf
zt0uV<Xgm%wfs+I_pYym6QxG8$)FQFI+J{4qe35P__0E5YAr=dnN$2$HP`50Z6WgUl
z_=|9ATK{lG6^wC$M9gw2N&j*(+3b()_$*`K`+RQo;F3EI%guqw3(QJ=;(sfUn_udZ
zpEJtS$2Y8^<!4U?ORz3RvWKr8Zx*+&*Tbx@2%gS%+$Fym>B0lsg-)@3_kBf8kl(@1
z=?3TU{;pn{0&IAUjQ*I&Q%Scz4EIX?^B>BXRk(M`C}LJa+iaAUPt5$1Y?jDDA+pL`
z*u<xki&U+S7-@1RQbB5KGwtA@F@)z73Twl3z?YPOVHDunbFdfEcDM)P9X2rwEhD|+
za*B529myj7>7dS8t5_Qa`0JnlNq=R~|N6)OyZbAHnlAs<Q+7jD{Y6EZbM!@AVp%qE
zDIuP$GKermHXsR_#e=D4RuV~3wrY$G%F6TBjT>oAIixz(F^dymd817ciup^q%(H-;
zL?lZXa`m)cp^>OrYd8kPHj(LMpk2HO8SNDcTTEc7!mL1*43SZuQ(T+^{&JKJm7-{y
zL|)vvtFXtd5Qj%?F$x=nPgYP+fOh5GP$VP}NFhfx$UfuN2UCnHe9Mc<U0b-8=3Hx5
zB!pxbkqEaWJE@z9E1^M#+S!0$>J)Mi-Z~XRgcKH02@pgUU+fU!b`?Y7%;*=(Nzb71
z?d-~eINl8Tt0g@^8QKwJItAI5VO%>Y)Pl{Wi95=G1U*;66j2@>1Tv(UYX~k998;j+
z#hl4zz&cUK`klywIJq^)Bi^7Pe72;%4MBiVETBX>cD|?u<44;tI_Z&|UW7}koiW91
zi_xhsxq*!!L#}C1u4&piVD?~4JmZh2^wUrd*eQiZZso`NX4_ftdE5dI{kG^pXls{-
zEI=;Xb`Eicb+>f1KXPzhq?1nc$##IO5NDj??d3eBFi~&|X{Kip9=}W`VrCOj{%Q|e
zk#gG0X$E($$BffYwdhb*k10fl?Au62olj$oBBDRNSuO+ItC#og)1c;Xlp1q>i^#pP
zke$NfL?$<!LiSa@;e7_lAeYs91yV6sqFP6#rGTVUgv!yy%Bs(YhGR%@dPB(HA(6xg
zVw!Lb5~in_5lCdlFOf=EdMBF1ct#DAm$O}yL5=`?vME@<GNDc6NHgfK_+)!TM%6Ik
zd?)BG2*zBwDIB3VhfIqF7*kP6bccA+VP6nQ&0(f06w{dNK^{Omhy}!MSB~*GmLmdu
zWfob?dhr}|XcH#uhBVRx%3;9B)#(`)Bt^1S*<gkm7DCBZy4Q=o+mb=zCZA7==YsBg
zUFERvLTtnf3F%i-q_pa*=h`Bf*5PhZx`@bOa@tsw_ZCE~F(ui1U{_(X?aTNaw{q;X
zi;?0<Ao=apEpZI=yspr!Hm*&a^i)m7W=bHF>EZ!b39jF4rWMUd+^#X7u;xby+e(x&
zPUDXf^7v&2-51hT-#j}lreW!T1t>#Ve3EUx0fr6d;97ZuC*J~g^7G71C_)F-bb@p;
zU?I*4O46jB%gT0XFq8Sx6^wo5?7hkZmsPE>om$wt7%FnuIFch?yrnTtHsADS7>B>t
z6c;9N0wu{89ny>-ETf3otgC$CnnRwkOep$ZM#Qxg_1pfYFU_7{1rn}vHjLo@Qct*Y
zO+|c;j_R`~nWC#IouhyWN<G_orZH8C?HFA0hZ?>4*|IfJgq7OCWkpBOSVANS(G3Z}
zSFLMwA-OK6u;=|5gR;#YPIBqhPBX<bWZ*}ZB{Ej<R8!_<a+yJ(WvflDW9)>0aod>m
z8kmw=p$C{GiPAvx0u(i#akxnXsYP#^3=S|#sh+He9~SW2)<sN$#7NB)i7I~tlNd+a
zMZA0($Q&!(G&P5^QU#_+xRNu>mAD#<8D#ZIHwKz**>+lr`SGNf`N4EZ3|U6A9q*<=
zKA!_oJLFP)6e5UTmfJmt#*q0Y$7f?le8IMKuFcto#ANBGchzDvFH=L?F@<9zp!cf5
zA;L@o7^6qIy<D4$8hNyhWN|xtXWgQhznWv*>U5qOF*ovCSVMX54IOBuZ|TOhvN=0O
z&9<{=^nqBlNg$3^ZDC@tXViF7)O@nkb<ARR(mR|ueDCP~@mEhIUP?CKP9~7kUM5O{
zg2<H}E<w1DNbu2@NgQ!B=7Wx0)8La?p(d04q7=y>)Lxnv5;da5Xl2tNJZa-8$4^Q%
zEOHe0x^4u%<yowqY}&<qP4PO46Rm%d)m#HO)Li;YsNut@S(K~A#f*GOVWyG>5l%V7
zh%|7hvk6l%qS7m-sM|o!Gm7(=%pT>%%tTd&sD%irdKQs5CLqg2F2#l5R!?u1Ib)RZ
z5I5d<2JIwaMTs9?e2;t8Q7y2CP;+xo;<7#|pnD<6s6x_kI}EpunooZ@Ghl35uH^_<
zMaCdy7{-$78%`%NjDt>L^(!ZlOD{@_FZ1oBSRW;#9P!i?NzPb8gNfq~8*L=;$~C>X
zjnT4AiB56V;)oy?F-VenhWiXjE210#GKJ_z=*LWEnFCLjZPDS>htJzd@#I)Db_T6p
zwBJU<m>#%Vza9L*(=+UU+@ZU}#g)bRv?PA!#n3cnvM48wx0P7>GA$QrfJTYKJiYM2
zX*z~Yp@S%@Ewa}c-Ae2Pi-15?pmn3k;cPw94DuCglbTJ5&cufzq-6R?iSV-A3^LL#
z^jD*Lba9`EC|bQouUcG@YLVt3l2#!_sjz)X7Nt5wA!5ug!TBn5H4Tvr@>IhDdg1`$
zS9rtb)*3=p=K!@rD*8nSy$mv&LB(58H)*J`#x<D=I=RjqWOa>E$67|tKxe?BZ8qz#
zrJAj+`am5Q(dHB6^A_<bru-?7h?Z}5*Os9;N$U~Y_||9lwXwkH>LY8{?_H6Fb)`e;
zO=_4eH)W_hxVN-^)LyIS{hV5H%Jr<<m~COzAM1*qa{41a8E(@C+%xp0(+bf|%!O-N
zYn<93Mr_Bl7>ka)Z#!Kw+1Dz1Oaa3=V9s!o%e{Lqf-o8yWLoeCnN}k9>|$kU(UOdK
zrL2`d*LZc-gig;Fb?Yx1yQR@h_wgbj;@xh1dT9>v$)=ZK*Uz@J_$np6v}~9;+`(E|
z$|`kFahtK_PxUzV@e8-{@v2(|dEK@#&7`*Fi|(a)6f}D{`Id(;QHg6bHbR{67*}4C
z;Rr?B3uGtQoOY7wmw-deIE`SYYbap8#YKw>>8k~dlSDMZ^89q8<wrX@5ht5oBbI%;
ziArnk7(Y<Zn`aqepk~gmsTE4_2S8RiaY&qAy-2qs7CD?_iJD6tNhV2khFNs7O+xKu
zzMYn3emlwJSHODSu*gQ#lz}Y^P^}^6Ib$yJ4X0BoHj8a7;Db3xM*E7sLX1V+fDjL+
zMMLGT?7=aC*_y(&#Evv=PME${asw;mCcJS1i*34OOrF(_3N7-YopU9Nt`V1J?Me~m
z8D%JCa66ewC+W4m9HD@KQz51joy$lD)48Zc1J0^J+w&|yrRgsVkZu>@DirZA4hKKR
zj%`vkhTa2JWn4p4TzK<m%6*9S(NF0iG}O$5k*8q^48Fty7!xcr3b0KVdf6$a(G=&P
ztj{6}c}o$um|{JY@4%F6WAy~1%(WYbXzgm&oW&GnnNmMFn*_pFbA->rWmV@>eq)4m
zOEriXL^nUiDKg?rlpLn^i$^Fbf-Xy4nvxr~6%AP26(vQbj7-_aG~3P*H=oXt6Jk80
ziAHY38g8Lv1Rci<Oww6_Xl&3$5I)>hBl=47)MoUs?#Y1W`b3h)*B+LtrO#3p@e&Hq
z;aIR4=SZ_mL35mMe#qk(NqWo-Iw`8Y6BJ5`T9PsvPE^X`m)UZvM^XCnmjc{4ffjUJ
zAchqiIqVyuon{)&zAICg)2mS(voty>)xO*?K+kr_UpWaisxrb@7SWWLXn7U8(W7ou
zaoh5HQR(9hCjraJrxevN;83^W32~d1?4q)?WSg!TW+yX%%-|eqM3Xs-MzR+rh?a!{
zA`t<MMRa?4qyeKn`j!LENC%V>j(N={H$pL|<FpQbb*K{?R>2dWNcLsfoY=FwHB&t|
zl1ztAghx)Y%LD=~V5eoN&&L=ujwd_dMoI~em;vY=P;Lx#o2>4+9U+bzF9oVxB}AU(
z4)j!4E2p&XbSr2WNT=Ci2Co6~GR#iHtCKkGi>YdR1@kSh@-0T^5GQwfl_0Y;B5#~1
zHKvR)^GBYmTA-a~_JO5C$0aGdnL;u9qKI+r6m}q4Ku+_VriOoDxF7{aHu?6YiGeh-
zouS4UH>Gx($;{|XeOZuBf0<(^2aM?~HI!srIkQJhLvn8^qJm0p36^806D+1D$BksC
zZy^I+)#Y7QjK=4bo|kd*C}v;g%W0^cF={r&(92>mzi|s!x(qs&8lW+cL&meyK&xT6
zofMBY^Wnu*j#5_Cn4?73EUFqWp3ZNE*>;Kqd3`GJW?#5|HJSp0%P3(HkG8jBf<_K!
z;%BL@tfY2WW~x~>5^It@6zUNF%xLVA=tVm#K~TE$N6Yf<lz101tABpELJM(J&(6#<
zuHd8@d0t}-P%ZjO1a~4FZ6gFs%m4@{0%Ql@O*bSy&{9q)+-NZ|XeYRT-)6drhg0T{
zNexk#593;KIvXZ2`q!V<&Zl@869rb+z9`<5<Y-ssvnQ~c!m^q|H!gM1Ud`n{CZf75
z$hL~J6!FC{l^%^56b(Kn1>a*VCW<WEE=C*tN}|F%-*iq+(~z@W4Y22%{n!jK2sEPH
z1w(dD(%9iIbM-EXzTRJ|74J9K=KF85@{eV?8aAB`aEu?G=0_~tLE-U;+>qx9hukr@
z;z>P?$il3JK~V<?9cnjHiN|j`+LyQ#YA=F0BpcC~9NehTh&lZ+OisGW7b%)OoeiLp
zBCApM)OH!I*I@bU^6IirGM1<(9SLiWQOM}Er2=C^ifIs?<>)Go_yu1noUme(gwEJ<
zq$>ghj3EjeKHGwXE9t6SZnJE?C2}KjN`ke|n6zdPwc=EXw;UbC1H@)pPwQDLSgf+d
zmvX6rDo@)Qc#TP9N5$l=v&liSc(1~y4eL$nnM90V?sWae5RS(vL0_Ut4uerjZN9M`
z=Prb9AmuVE%Cw*fu;t_)AUY><%<i+tX>^&NXwD4{ScUSW1bVS}I2jDS#T80?;@TZZ
zficO0RL_Fcyu29Xam1plP8U%R%2s52Qt*6Zv=aQ+OIi2SvA#4$2frd$vSm2Il_uR2
zzX6JIH=+?yvn5~HgDnCIiX+OZmPChKuFhDyv=wL~IocUv_NAu#hKF-7yNJ%rHE7uy
zVXb~Kyq%H|fTOZyj6jq>#ry%_&g5Zeq~Q-eC-xF;^OnxIOX`gKlP=jM({Ya~XbkR&
zGKp$m%@|)Q&rH!Ggk1e=HVLCI4>SsXi6V@%2$jxttPQIzG}^fEsAO|wP@dF7MB!I6
zI=vy1?I80+gy@pS2_)?=ma?Q4)--`7DU^vdq@YlGBKh%)gy>HBOM{u??R0}TAT9~Y
zUzR2*bzQX^Bup<pV^**!W?yD#Sfyo9;jm3+Px9n|Q8eJqlpIy;$vsC1e5(_iGBu`T
z(<0Qqj3vx~Gp99U#5@3b{g2)p)(}3ky@&?*Xu9+gnjP#b%VQq&fFoou8WW9Xr?Hhw
zFo(2!qUDZ=VI_=a>G5RVbTLL1h-y^Axbe)7I;n7$AFA7u8y*swr4R?&gKyb^CVIs4
z8z_!(qvmLvPQ(c#Kb~w8t}lx<H<+$MEN_<EZY<GIL$dliTP>AWu6UC78M~~~8|ape
zQzi)5u7Mv+53@mm7z{Q)PZsxcdQd2xqjaH{`Yy{s!3|^Th?^fthJ{m1x-ya7cJY%9
zjS4PzH_*i-4!<<p<r?TBonjz`_+=bD8J9N!g@>pkOym5S!#R<lS)MF;QJRqGkGF?!
zVrMj}wHUG(9d)ul*O9_Og{#noFnZbPmc|I5zA|7<ij2nE!EbAFy$(C7M+A>)fS~wC
zDXc)arm9cb^fH_c^&M<M7LO#e<LL=O(Be|WJUdC)?D32NibdaS#w}KfHKMO@EYN(e
zUf`xLRr#ZSte+aUKKL|8GDyn)QZtq(r*ga=q3sX<<s-~ZbIMC9SRFcxFDF@yrAdHF
z$Z2_j!tn`behQ_UmOtv6j-P5|zKOSB3|N~KX1raDL1+6zu*AeC*I!vncoLL6%k(oP
za4}5^HgWvLj<(W3vg}2cd_~l-d{T~OS{~~(M}2y3+;QO%`5B*E)yv$a^kj1sXCJZ%
z$}I+}z7tzkCGevdH3pjRxQ}tlR5eqzlk!}S?S_~b`Sc2qh48sWmRei`EB1!&;^<Im
z&_!*c@vHOs4c6TjZ4Uia%uX(+F}5i#AfKH_KZUgbzx_o^Xq+HPnj0mEIrt{;IK`R6
zD$X|8Pwz$e!)HjNDN~y?mrjaA-)zm`m~rfE9GW@5Oon0<E>#oAa+)T6XmJX7t!C3p
z$TWa>auUw96sOu+gDr*~qEK{1N%@vXV0(35w<I!zl}Rjmp~0hy+MH70G7CpZaQNaP
zbeTgVwfSk@B0kBla-@T#PpR5Fn`hcW`saV9Um5i8|M9O)tG)j1U)paD`rmuMJ4kzZ
zj%NeU2A&N(8+bPGY~b0zvw?SSz<z?0pZcy=bCoOr;7nTcNJ`}RdZY4M{d-xl{b=7?
z<xzV!@ND4Oz_Wp81J4GY4LlooHt=lV*}$`bX9Ld$o(()3csB5C;Mu^lfoB8H2A&N(
z8+bPG>o=hPb($P&|J@q1)~LUJJ<kBo2A&N(8(7&u|41~;&Xw%HhVv}tuesIFpw9-L
z4LlooHt=lV*}xyZf#>JCKl=th_xJ4n(uR0TA77OhbmOeoy|~D}Nqn!Bhy85e*}$`b
zX9Ld$o(=rk4fxLzZiG<?tMEX1%x4472L7xK{IzfSG<b6UtQ-6rr=i>N1-E2ufqezX
z*k{)qn;1G7T-+BVqQEYWZXwbcTP5Bqcg!vQL2Ph)wK-7AJ>aj}fZbI(&E6T{O3se{
z-^k576)$6d<5r%TKfQtHnfvs{i+ayZD3<be=+}PehQ|M}Cnt^*?YBP*WdZKsJ#BQ=
zZk0WpfiI??$v^RD1J4GY4LlooHt_u$z~2lhV!zzr`{y|7P-n!W9wmQSOosR)M)|VM
zKJ7dkcsB5C;Mu^lfmb#_zdNY?%Ao)KkN@v0eq~VmQGtG17=B_H+?9Ft_cqawgoPlZ
zjnYGATLRS(n~QV!pp6!{-xM{n2rtqZ>NS{@2J1TA;)WAz)6rg}np|B|8s)h+RbQgZ
zZQ$*(j&^mfsk89Y(ri9?nB=IYtXr!v!7}9L=oUQUQO^I$i#&0D`(ls!aOq0OOM}^P
z18DbPfH_w&6Jobr%-7<K%(Y@`S8X4{I}s&!wH5U#XeQ0oqO>pid$oh(JeD<WJ5N$<
z;WKSbV9uif``}_GEphmo@RhY#LUl&P6mO3C)~<6sAja$Wrs=piRWn^SNL~nuqa-^l
zO;Sz5qXe6bQR!sDki+;-Uv%o9ZezW^F45l`;@fK1bvo0f1R|;SC4ZGA@(Jqs+3Jr`
z#?L(|qC9iZX>wznKgZ~^tiNU`Bku7-YQpi6<(XQg>dBrOaFsD)ey`w`h0f1*v-)Js
z%ASQnr%jS?ut`%*vx`$9KH9}jPmz|DwSd-4ADoSmv+JAMn3sDs)_9ItXa4`s-jx7q
zF}45c-Zm5>BGDpAs8skYIaji#B(h6Nb<-jhz7Qc2SxU=SvV6A?f0FnT#krwHl1jEh
zw|(Dk`=b7H=FIxOGw*$unR(C5{BF(s&Uwyrp7WgFdCqy~op+Yl`sTc;E+buw7i9pH
zaZm<`4~SMzt0U)4Nu{zSAFFQ^$|<!a77v;kmS8&Crq*X!`A-FwQlm(<p@oo&in2vt
zlQ2|l%RBa;121KAYl=%0_pAk~Ct3|1^@Q4rW>D2<S<R*)GTAqTdQCN<SyoKQHMBoo
z%u3}=`!5$V4aFMoqLSufS&fzQLUpgY#p`fm%oeY!=x`0*fL=*XaKpTL)i>^vYEsK6
zS!S94sn!hv6Dv+J5qFMj#@W%t2~l=k`(xeE{uJBlPL`=fyc<ei_)aM@q=xS?{;DpR
zJHuVqI$_o4>bxv2t*FX7wF<Xm<cy}&lmWbw)<wNnszBB=6|qh+efiH)V5lTW8BJPJ
zy2o6~v+7=H%<O^MGHcwVuJ%?ptCs0XdVy1dRbWOGLrUj(O0l37a3#fr-T=#4GCZe=
zI897PNojR8OI*#AGX$DdMZ(2uTZ(6<X5_qT!BnDES6WneyjYW%3~S~&nhIhnYI?F*
zRn@!2#>MofHNsG}Cxe=jQC3w-6bZ{~7R?onv}Re?<Mp@#x1^>u&3IITRu^kIrJ61^
zCAfm_f~i5>S3SYBR0U>1uBv32NZpt~F|HWYbme0GPf=AW^2&;$VOOoEtLq6nJ+)O?
zM@caixk<HEz0i;<CB{xOsnt*`>I!O}tLcs;nb3`wlrUOhDMlavZ`+Pi&U&jTm1H`l
za>EasYHFbjtZDU6&+*}u_lf>Y6%93W6?59q$Rwo%hQ^~7gR1}5)+{-5hNqQ%yjb^5
zFOm~1t>pEDuBMe}sHo}6Jg1eJD_LZ^nxerYLF;2Y8fJ{f_(&*ub!UJ&N=zVc{J*7n
zW>8U=A>p4=uIOlXRHMv-YC|``W$IovuQ#Kq8};QB|1X1@V3USPSytY&jHbj3Sc&qd
zpq610Ijv?5sgW}3a8-S;*HSdpoTZ*6qiUIz)n?6VPHA&(TOC=bc3a^<)@R66wLCYW
z=rOrgsOoT$uEP`Ro?B5h)QqMOlVt<0Y+AYKsD~i;n`b#WtRu4f@}_jk)DtFF|4d0)
zqLQkBT369i<yK~_RW*YQO-`vUFau_TidTwEmNM6sYKjf6%ImR0lTfL^HMv+VQ?x9N
z>KR2NX2<FpW@8F%vWC7=O}8y4n6|vrt?6aljMi;cjVbCHyfW9b$W@UH+^m*2<&319
z(*GM3jB=$(X5LU^`br%x)m&0NQ+3rEx_i8o7n-fBwip$<SgFg*^4bb;do0OQ43&(M
z&`V_@*Wp#&*mILgYs@9aKl6c!72rE2Q^iK3Oo>%c3K%LGmXrUnB1P56Fa<@V?kpK|
zft==ku)X<@D;XB`N{LMAN~T0I`n+LNC6;6rwM}tSbB{@t0u?o7uFp!8lunMKZPHaL
zD=MlPMN{)Z>9T5Gv84(Wkx@%&Lm3v9lK*(Qtjzu^e_YN=6cbuj7Rb6>iziieMNdwv
zWvZ5(<0cf#iYZNpHN*-PO;(TRm^!b*Olg%_Qpt1std^2ew3(WcP%zUSw-#^39l1MJ
z;=MB@PAwQ32~tj#Ww|m=j#bSSozi+`vw|-4KH=1?lRQ(8Qnj0^ey0A#)^}<@Re7vk
z<XvXKX+Uh7DM`7cavP(GsyLOoCD+_>vZs|NjaEa}T+&9&iuNWfQX^xe%qC1}S%LjG
z{qaKe(@@i7cDOBNZ|s-?FHr1g`np&xQ?v{{SyyR5Rp)6|T++rgO4iX`G7@UrN@d+S
zwLv%Yib<t0%}w4UZ=V;aUDp*%ZZKEmKdx;cml}DkTvOEYNuj1>m@}n#p{Amswlj)Y
zif`(1l5(_Uc3esG3X1Qhq~>#++)}_YdL=y}*HMq-OaaS$PztzA4`wD~1+M*>tnnmo
zSFz8MMo619ZA!UcHU~U%HD*Kpr>5kj(H&l-X&7>@qR5r1W@>bUs=CfatcQyGy00vu
zX_%3v&3TiaCXOrD1x(raGkck8+OT3u$qw-UGH?cYLKQRrID>4@(^?IdxBh7!sZu$s
z^Q{#zneCrR!J(d6L$j-)xy(LotEmaKuG@RYrHbF21ADb3+_rbkklGqng$}$Ws*+)u
z)>$)VXwtrVl3ml(G=H>kG>j@iDi`x$mE}sBmGiphnySdC9LSU!J6Z=z=mTegDXC28
zM42ku$xOwr@@9m-oH93(Oq!22H=*gLYDCdbwUX9Ya>%T!y_1t_TWo1m#Y)Fjv1U=v
z>Ir6qm1+VFb&*jjr+9lN9i}EXX_9D3j>=x&88SJgN}0nf$;%k>oV-_(dQQn|T5<*U
zr)FhE<v$mi_lgy>z}>Xoe#&OeHBBSMy3C7I1+9xr&i`u?O_7zzOt3=55O+nZ&Px-y
zr)wrsN}?<oJBki-llz>6RX0=&&y18CcI|Ai7L<)A<h}X7T7#9ze=6`y`1blJvwkMg
zR|gd(ROI>`8BJz;N+p|`HX@B-`yyFtU*IIx%(;bZ@^TN1o@x>UdH2~==QEz?m2$;^
zDrZh<#@O1REDo%Sr9jCU5<ROWxVolb<h1gV(Wt{x)|@%7T2d4YEwg=Dsuq~b&1tiC
zTslTkrdVN_?Tmz+QpB>dN)@xClq&|gF5fHer~*d1`p5NIZDvUqsur0SvP?CmTV|??
zOy4O2Ze5p|5;5AQ5^hK@*Afi1w36VqWs#*sQ?jUX>n1(D)NIu3gMHduO<~7gjTu!N
zlR>GZW?6|+N88J&Bu(`3BGoddUoohtDnlYyVMCzQVNxZ-HCemN059V@s-6t2B*VW%
zCX|Mh=96i}u%vWK@t^t4-Q|1MIi)YMvZBv+YEKnimezaAGu#H#V!E1^l9n^<zao`E
z{gZW=P?7Sq)~2GaG$8NPe=Mcg;Dt=Ah!hPurx;>;>z`7EiI^F#W>rvxMh{{Om3pdm
z1>zLa5;&Q1lBd;(Yn8<GDslIme0Gv2w&rrFqZ5t=ZaCnnaW@=v*G9y;>k4^EbC$HC
zD{e*Jh!fqZ%HxT2vJj)`q~;`#sV+j+$<<WluJy&KVb>Z+)lgO)O5Awj(&RgumK~=`
zHgX%0;+&JbB+t{W$Lu=QO;ui!k3N;B>b6VzvKjux!{F2OokRWlz=7?Vy28DZ)a07b
zKt`IXtMJDw8ReQ0P03uY*42uX2DOY;&pxL#$lKQ%WM8Ma2?k0vwXWHOmR1xrD>|}L
zZi9iG<Np=QMqah9l`73Lnfy;NpqpdWH7z5hDygbk4Tju^in2g8ngS{<vTFK&&4w)D
zUu@M98Pq&?k%_g<w&@sm%)Mr|JpIQ#QyJoZs_yc3b>A4b$|ftZ-Y_lp7w?`bQ*`1>
z(~;P<;%G++0@XXsPbF)}xaW%Vy8n8IbV+;*+z;Ms-gP$e@}Cj`47p8Nk0qHr;XkgX
zDl!LnHN%QhTb432x>QzE{;L9B!i6l&3)PfTk4t!+7`bkW$+e84qqfQGT6$uraksQS
zs{K`JXfn0BTqV7K%q!zhv%x&IA%zK*BHk08miNqIdB=NX?U&(}T2pI6!5Eoph=F32
znNn0#fmW_GVHi^NSrw%oL-|iJ!S+glBG6L2REZ}=i_4X^*iNp(o8V3=8Fgo7K=Dqg
z#O<jqnj6%-h(p5@)`nVNMLDk)o3%6<Q?jHLfoaDo@rvBE(SRmp3UXTcugiGCWXCQk
zv%q?-@0Fx!r&u=SO-W`>YlFGO(ng+@T12|QnozZv&}4uoRci(qDN{8gsg5!8k=c_i
z%E<n+tjTk&+@!CSGBu{mYV*BPUo)gwQ{-Gj(NQa?ikc<&&~Tl5r8Uj^rM9gUXdZLN
zd6^-zr1T7PTJ5MTQ~!(xOht;Yq)urTWYmo)GHZ^jDNL(+Jb+{^OP7?q$);+S8#9}<
zr1gxR(7iE9G_y=(D^?4XGF8FQ<0&m~<TMGlZpz6?BNv;}@)~Ns@hXZ`-ZAo-gyp%+
zU{K+&q|q^#8^NQP(poe!F)7xRCQ^|D+iL~7Tih|ls9wgKa4uoxs<Nq$DQT3)>S;=`
zsOjm*eNgsXSKn(@O&VHKD`Kc_iipcKsZz*Ul0{s~H7u35D_XItz+19pR4z;w99vgv
z>y>#z%gU%F4K+=oO*)R;302D+DspkmA5&sRxOFw9=F}plqWmcXuC%D^bg@<(BZ$?s
zR5A=|wM@N}HJP&6fIb{d92<GuvUzV-(G2r)#V(^sF(mIau_BRoJg@FKXf7%_wll;=
zN>kDb)e>DnPgv?}+FEU<W!BW>I`YC86*<c!s#Hs=SxwAMXc1x}lqvB4stvWGR^PC!
zn_@)@b4o&OSxxhHOqH}+nmrAZm1}ysVV2O1SfqMK<l;CDO%O8$)Cl0EYCRqxW^3kV
zn5v@A>*z(gTrZU+iYWz|o=pwCx;er1<hrH`N>a;f$ZT5w%w~`-nE{<ib<la9?bQ-Z
zK@WJPz)+A={J&mB&suC4QaP3QkGmo_sMjg6N@9cQl=86BYBFD}|FqyVq^!H@Kjygl
z$1P|%d*sT>UaiOKYi%p2nbw4Ip8u(utXzaY#S89?*(J?6hFUr)cUct37399Mnk=XP
zW0ttR6fG*mY5Z{(^*tAJFD=HEHgq*Pp|qV;9|EhS@v8C6JEHep7HetF4j)@in>spj
zhFf7a3~f%IW|!AcjVT$HPV{FTJw9B_J4+o!U}|40WOC+>f=W`Zr&LwtY8eB~8^eqy
zR3xhByq;o27AuCibzZJCBqvq7+^8Z{Dk^fhjB6-+rlQ%Algb}M?vf%=#agMVFK4xG
z%Bb34^>|%wDVe=tHyP~=b*zn<bD5;gPGw@pzOFhHjBSNqmRERInpJXU2U1KMZ5k<N
z!Zf(b@LKXl#;T{u?WJ0&8JVWM*;dfp;K69FWl5VlZ0<rTX}x{LmQqvQF_1S$A1$8n
zOo`eihb>8^NmY_W+N@&41NT?AYgRI+b%iALjk-pUm2uJ7F~mwr5$m+7g-WBEB?EcN
z$TKy$s;VuclGMr-v0BR08SR-_Zdz+aMn%Y?Qv0XK^c<6#{`B@03z|9Iu40YpsKt6B
zHqQ!T|EadQDl1g}l!R7YLsrPr3^Y`$YqhM%)wR85o9VFXd~Z%EM$HZ~75-1rH(Su^
z^8$U(&1qT1O(mnojEPjMrmQ3j%*E!kQpR`OYpnrp+mbV9cpI7*Y`ocj<Bw-7BBPL}
z91G0_EG@gI?zD=`VJ21ftaC~gBX5%HLPaIkhnPICr@k_%eT#cDcx4bLNH(ZAz>xnL
za!sORf}&U}dL1QUUolo1o7ZfZ<ff!tYRzftD#jMZ=C#^!a)<gc=2OzLugj$It(zrE
zS`%pKY|O62p_YSKwP0TwCyf)>IlyJks!k<N@&xwGRm^FVTe0HUyh&9_#+Anjv>HwY
zPI9fDBc@$ZwHRAJHm@~omYdUBxm}PdW%kOvnzfAQtTNTSt<Y$?#D-#;b28^7kEyOU
z?4a!+Rx3F|GS!@Poa8nOhkaKZy-K$rMUPTcPuVL6`wn8qShjE1(Tc8at6?kTRbq-_
zvR3!3h1Q&{wZvYrQadBgV%=U`!oiMe->Edtj75-6t%P+P3S-1ES=psHi7JSbYwkN1
zXvQ2A?8Wv{&6L)rfsXoyik2!_)lMTu7L(O%Yv@#*s@!HAcoR;BGzWAvE9z-IA*-3w
zdMzU%7dvS<$&Jdfskj<(0;Pt%)X@<q^;8{?wG*3<?Tue~iyxLuG9EZpa7&&l%Z~ca
z1#u(q<a;uFB&`$QNW8=_=1}G!X7v&jm1x#b&{EFU>`UyW3EXn5P+~{yYuZb-?%1KZ
zhB<AgYa?^2;w0BhBq&#lVl-m1nw=7P(i9!)$BW}7stLyet%(vb-hwKLFEH71EOaPv
z5SuKfnRGC$i}ggTeWwDe+?tEEq34~{Omb6Fuf`Hia*L9KnnPVXu^}^3)+QX}4q`ix
z9913qnNZ*AoSrl47%9ClKH;cjE-<I#O=yzjv@~rE<7>rBn6AB)TX86J5Nj<vpiM0s
z5v%MVPN%@$7l#HM#ED#TsA4O&6Pjh_w8d|y0#)u%tcZD;DQPk6z_f$G1pHgAT5`;o
zlff8yOja>tC)4f67b=o?F|T2go04&5x_NUVu7bJHluoQ?Q_4j)1-em(G6%6U+qyw*
zJj@o%X<gA)sMXfBQcQ58Y;QJiPIH6t60ISp60_W#*6LZ&Nh4OS=FPx=ak82#F&ASu
zCTp&4Ulc1<^07kIjDy(1xVc5^SeQ!8hRkVMB_?YlFxj!ww-%dnhGipVDrQNi0=MK&
z4aC}y&FgMD7U&w5gp;~eZp~Q?s2L{>C%JATUV%%T<a$Lt5wDrpCmVem5wl?@wv*W^
z>eb_G#7p$f#3oWujjd|VJDYJRi5JI9)CS@U;w5HZ%;}Oe9m_2RmW;`~liW~IG*mT1
z+ejIjES0KUt8J}k&1pI*8aEq}uRbvSSu%#)L9BMlrqo8ntT~9C3LMo9&l06lt?r<2
zCpO7U$wba%P!`jImetT|S(9q0m&OTriGx@vG?$svx=lS1U#d&=M4U;zww$nP86~O8
zRYS4Wta)orZOm5aqGz%8)qJej;*CS0K~I8|V_k;=2eI?en7J`kHYF|IImuHTdOI_^
zOix%^Rx?h87P%$k<eF8<)K=;RN)t}ibh(}|+OiQDg$^mZI(9N%T`$oSR<jObixSgu
z&@+q8X`{ZqG&zfg6-Ni03#<*Ba&bwMz}%u?I#wYzuhmLX?xfDl8D;usyCG8A)3O?B
zPH5;PH<{K{IXYCZ5!;AVQ%)sLatCLe)ScvJGmbi1ZygmJ3LM1dHZ{`<TJ^0tL(xfY
zXjoJB`l(1&d#P11w2hS2F`jcMbr9QGP-KR_k&0EeFR_=#*7ZJb<A;q%^TW2tRw&nS
zOU@gQwHA|)$tGw%*0K1SYHuA2>>aW+=~&mXK&30bJYM4HUlOVo6;sBPRLv_8FOHYk
zx#Os1HlAGCV#<;+*H68G)lR*ZQw_J|HZG-N+Nx{KX>{33oyoA#H;GJ1tpObqmFY@y
zg6pZ;PL<4Zb6PPJC(z8>7T5}H>^SS$43kZL%?TMDOgV^66>XEY)og{yHXU0zHt%eQ
zIcTHJ#7=)nshgEdX`VC*SR*l6qg}@WLnS_CX~HIB*I>yFm8=o(iAj<|^?22Ii6dn;
zswOYgq?3}9T(0I+<Ro{jptxXP%U<f-xRbt<T=R*Qm!zY(Y2G<hh!e*NG$VEeb}~&r
zGjy=5iB;6H$+wkyjpS{cHB)TIQQfJ)NpAK*NheZtEVC`J6~?va9GXt`o#alI9P}N;
zPF8H>P8L#;TXnUZiVQT7MqF%OlRKfaerhF-+KvS#kK%OV1hTT7jM;FKI~SN*c2aUO
znyQ@JS8H)_&Y1mjOq)Eg*o&=T(&I&nAWo{3*-M?Nb<lSZyJm;k*V@of5$GAc*wkg7
zwA~x^%*8R%n5?R4FEyEoRfx?eP&F0lNvG;5$emR6Hp~fKnJ4tDSto^bg8B)H5;WjY
z%Pcn1Mv7b1@*2ugQ_^VFR%jI4S4*{0jJ`E%&8bbt3S;wzhfXC<a-*hm%D%3>RI!qv
zD6SQIU3;l9Xhyk_vYK{FZnbLUEcPuKMJHaut2)SS#kNAj1D0}9agsaNus3Dc(NY?k
zS~rUf_w1#%4y2%}*iThobt;8wDX1pyah#zzfl=Rn2y7X1qQTy}y;QbgEwJVczntWT
zhF!{`mV-FP6=qP^)Dv1Wae|Tx94lxQW3hw^HpMm~Zq!<!=k!EO({9Ukdvb!O<M*-)
zW})g47sa@zWSvV)+V)EJQeG{#EH-a4o*?N|%PLoMW}!K4vTq|ciA+h2L{nWgX;F$V
zHX=QyOs-c#Ri|Pnc}xRZ)>++BV##QA6Zq($Z>|xm5}Vgr@mj96V_;m1aRPHcw0eq?
zjiy>|Ut%xSOlFSUzOvVwk`oqzTA-K9t(w!;oVCQ_U|f+Rh?6F+V=3hsCkv^{V|r!N
zl9PsQM`WRs2Tmoa>O0x7lRMhB7Fcrz2TpRwJ~~x!k~<l))wdPe^;+$KuCFKbp2bUI
z>#)3|mTre9Z1ikIimpk{lO}=Hs$+qy=2Ye+cdVc`=~T}_?jUw-#Yx|x!(82VuUayf
zX>3@Ebj$kXTe`%l%W8dHFE-B#l9lO3ViR@-6dkp|Ty9RA?U?1}v{}cTcGR&eu#+)k
zPIA4W5*2jKR1-FqEh0;X88-w*%G#(~a#K^5cCD?MJu;W1>V?*tQ$4fo64MEcI2ENT
zH++bx8I#rZQkBaUovJt&XstQc)D)6KXEP2ZHewr5Y~xNUvGUlwEe~-@YHO*=<LEaG
z7%8JC7MUeu=<t-e(44mIjlEDdq}S9F@zzaRiJY*M#h5e7oC=)eM$Pz?T8Bf@29j2c
zZ_p}nGGxe&6t5YZSERPWG&qm5ll<qp+OIeZdVeh?mW-~FhQ~(qC8(UR$ke)XADsM+
zubh%39kZr7rlj7iolGy(6LDIGz(_G|Yfi84TwtkS$(Sm~CT;X=M6tD!%IkLYgteN3
z*s5g8#puUmxnis^Hm{jWP41+vS<^6qo?Ve9i$&Y6*ehu*%cv=sL@enbc9iq#rdp<?
z)pY6wN!GWj#W`eG?4HcASF8z}LK{&@&Nx;v7nsw!LG?o+zRV;sCG~nnLMzqKwoisS
zOBt#bx#I_vS<AMrt<YS{xlu=NoV;|DJ6cM{JIBhg1u@1<SyPgCJx-uAqLpZ<7sMnS
z${oaVWnSV`?j(1l)Tu(Y<T3PE{5JVyO0pU;S<MI&+Z9=5)|^_`R>;a63LL~v>S`+n
zIMq*Jz^RI@+*W8bpK`yLUye3o3u5!;b{*3Wl^n!LI^$H^xid~R+>$#v9&bNh!nsrB
z@x}2HO9e;f4VQEadcso4Uh2%HVc23OMX{ZI?unB>b~OwSoNF;1YpFG7*icf53LVRw
z3Y_FI4(Tpf^-VcbQdgD}4q`{Sxvr7ct>_7*%qeG;8)iys)v=z*glkEyF%`9}$-W}u
zJ6>v|V^UX=NvJSA<Lxn4PLix1+iHe+qg5lNIBD256_}Dnd5pSIplCZ)G1QEduB#`k
z6d9U4WvHnoaRRMIOhHW6K}!>-K9wBoJNCv&-|)aG#a<aSqJjE0q5L}Xn}a$e`R<^%
z1~>bi<VvFllpl55b!&-VIP{#ScAk3OOd2$t++gCY`!~*`!1Q>SUFN1~y$|mtQN@eq
zHTb7&Q85(;M~}^`blL=ZQAn`Curjyw?Z1Ws?Rlub<wfbY@p~xH%0s(;Z`XM6sij18
ztgx|N{TFh#Q6PB8Bo2nXyl{*(dM^zs2JN^ps{W3t{q&|?wRRJrJ3NT~z2n23cdhP!
z%mN}hRy*yJayLx<ivq*RL;1to&z;<FB?$^HB>g*->9Cpv1;e(UyZ!w)ET8vb8)x*#
zo#PPEc)WMUYd`#7qxHmVjOuyIO#6L#hoV*oyKW;u@PIf7&O8wPJn@051JBQ?@+$>~
zi-$ws&3>+Wl|95STu(bqsQ2(>5){Pwe|OIRFKaWQdBKg-C=fgZF%IhGRF@7{-AI7h
z;i2&OYP4#r4K(a{<hX9n+|Z_&c#VkT?8LEGyyN$kVE?`%>T%I|Ki*yA`LTmZrYLIv
z;8$wLqoV&V?zexXDo#7^jyXFiP%IuEd~y7oPxox2K=6P#z&yY@fOP=t0M-Gl1Ho7a
zUK~5-%?WG&prT>Zhx@duUGy^nf`=f*f#}zXd+pz^KXsh?=yyd_boBbXKkWPS`wDml
z&wl)l7qvL^`}}TL+Qaq0-d+!U|L7j`?!0Od1t!PC8{Hoo)o}3+3Urr;Y=1A^yXe&y
zW_MagMaOi%2jKVO)P64x8((3+{tW$##&w2%kMsOJgFl};={FKK6z%>@(f#r2`bXIH
z_IlPEM>MKC<FqXl2p*yk2kCx42wT0NKHo8@p9ewxzI5T5A$uqGpF{k@usrt#{c!jB
z)IoU;RjPg#3_fyWzcM+6QO7TzIlt>W8@CW3c<>?)ZrDC`(dkzmra*srC~o#alQy}V
z2@pJ(<G`=uNwnkd@w*e>qFwJRnm?DV|E_iGN~H>eif<jeb<>X@uAo5hfH=TBz&eoL
zI)MAl{k`8j)8}8HzB|tm0gvF(jpxw>HI7bPc~6}Ve-@Ic;yXEY)}PUCI{}Kx!|BJa
z%YCowK?-!2hp64><JSG4LA?$DetUa88~9C+-zC>PIi=32Qemoi$4__tbkkG&X{s2u
z>xI#ua9#(ljVsP)rFT9Hyn#3T9^n7)0eN3{-q&vEauO5_`}J=XPA+)nl2KBjUx@3R
zyt&RPs^{l|)^ES&SxN1T9#DSNZP%?OLC2t;S1-x$RIk_eO5d&XVz3j8_os_KPF|mv
z_G%u%Zm{0l?$vmO_5LedM<2}V=!16rC;I1;U;dSJd|y&I&VL5u{AbX<-zMvCrth6c
z{hOxsKD?J?%0o%tLs5?FAH#V4qtpFhL6LiPKUT8qceeHLo^*HKlO7ZvJAWk2duIAB
z>KEdE7k}?}5uNW$kF&y3U(|70v~u_71sCrmK(Tnp_HjZ`+V|>s&TqS3oi`7vUAM3M
zPWSx<qCck>{yn{ESFPP7+dPzX-Z<Ivp#83(`HT0dB=<g*sPGh)-!(;ht`Pi5?|rii
zN&gOII;<u@dmh^Ld%MPiPc0>)V^GdB!%u&oQ<v=VRkrlp{yqbH{JvN`zrfw+7sT?}
zu~WsG*B-T(h>l*(>#)5iBB=erbHu%Qj(AwUr=cIoT}PX(_l|1!-DewdT<-62Irt7X
z-w(bxe$J<Rw$W5k^y5YN7mVY2j8kUg<oD-&RL@IXG_S!wWs8bwRPoSvv!AP8We*V*
zgZBNvtMeJLcGJDiCH6cK^4^|@1Ha%G&+84}^NNFZUNbD$W2f4WsGi4IPdiPh_wZx_
zj5ZIc&KI}zalKP|*E@kX@HS!XA5`=>2>bg%yngN$?00+ShBn0{T^=qNcxjt?^>+~<
zct9Lr9$+0vZygBQbFoDIodSM|{W}uI2je61@0@<eXO|lvtyy!#DhdP-hy%=nbk~8P
zKBpqAey4jr1@$sJALTcHZv9RJJMKJJHrqVjIHFPA8K-Tbz{EWCyyBBn`(3qz_=VZN
zZzk3Dx9*=ZxJsL4Bxu-RSeaY;_FqH8hPZx^Uq2YM>#g9gH}}1k#H-);om@0}Y+j|)
zCQzU|JY@TMaA-fh=knb<<Fz0DuhDt}1P@`3gT0GhePMQ|brcv)9-?|ZxnKP*JiZ#O
z+G+y@ip2x&BZ&5W1Q>s@$3NBce9#~K`P@mr5wCgh>vPkqcEb8QDf*S(`>D0(#)ln)
z_S^~ho7{8syyi*t`?j3?!}*M8p3lIz2YcKnH<&o<{*Ch}5Ii6bFb}W}q_+;>e6v{R
zn=u|35Bv@kyx+k-+<DjP{>LmJK=6>wI0)+dQs8&TsgHhFM8N}iNE8PLst!Cqr^>Gs
z2p$jz#yk-HJ)r6O7nbMqJN^Cwd3K)rnCa(7RNf1Ye&RjxcV4xKmJ|nJzb-`7{^5G2
z^sZ;}iZ|R3=+6CsQRi{!IZy38^}3lf>R9cxPs-ge^)Ko+M)kdCLA_7Y>+APmUySQ0
zF%JHYgQ&-~@T>f&+pb$nzys$YsGnQ#8_z`u^K%i>jqj)&f2Ye2>S5i$x)I&y3Pojo
zL_e!bhpTQR-~l`!4x&2`aNf(^^Ilj_v7UOlo?^bbJ72+PcKOWqdkoNiINnEraZGO<
zgSP(RcPacn7v|q5asD0W-*Nsuy64|9|1tkD|1tlu4qzSdavkvdJay3G*>8T(9)36Q
z@^=Hom%H(`yMC4P>%Fv(`i<xPcz2EG#||bycX<ffc~JP9-QTU!<v-o$k%(F^T^>*`
znDOiNxZyc$-aLmbTfdK{`#$lg+=t)g#z$+`9I=W5GvmSQbqV(IdDSL&GeI064iE>J
z2T@-K3I-p!v0s^-LSSMZyn5bOrgnQ?@yV(EuG&F>;33FyknMHY@9)*aUS2px8oif>
z9XocaSo7MW_L8t-NzXM!xj)apeJ7{R`ZL;XCqVFkIB=c^M~>_E%nfad2@pIW4&0ju
zH*BA}==7@&Q($^LWcocA^~HLQdDO?F*B5I~?et=)!c6hRmG{)?@Mj?b+Vg<-y^8t1
zSG-R;z4u9{%3I#oo%gjHx|{$x4_+UC`Sm-A*Zt7!2km~2sQ8QDb5i>~$L;rfIN6Iz
z{AJqD#kZe3_2nmX3W4eIkSYIu)z9?%jbHV>{@f1A@z;9VX+ph+ClereNF@$3ecvhS
zi}k*39ABn(d<nji;|u30?axzYix<DIYeIYJJ--pWrN^6J-*eCop5x=?b9@kAiQ~(!
z>-w_wy*OLv5AeB@-sg@JZ`nG|^Sa$-Z6-7?xN#Z{3TA5_C);jVp8GSV_S2ho)!I$M
zhV{3+DE&5m4+VmUAjbjDyQX*E_3{oytqykGMu6ZUigDm{{evj-3LD$ie<61p1%d~2
z9L#NbRp*Z;?WSI1=Z~a$&rIJ%gNhHnIDXEjd$v&^ct9Lr9$+0vZyi{)a`)#27w;rM
z@Q^4DhMx1(&Qq_ONkv7`o<p*b^zTrn!)oFdhVA@hQ2O)Mv0FF&_~8l)1P_P<ap%E>
zYliHd*nbX*I)>$ZlHcz$Nw&SPJqPT@5sm83IBg5{8?*hqyx8{fULtqjO9VcH$!EIj
z8{Qj`*?R-PUwZsy`#mim?!0St|6>-AprEMNZ3X4JJg@!q`#22kW#{-8yo2{_eD8`%
z+%*_h=9a$w*HB<)Jf!>Hv~1TaTd(!A<M!ngdX1u7PZsM(w$1})>vy=O5BF(RyXa>k
z3gWqr*5^9n_p@O9einOP6nVU7&X4!Z1?Bunte;W;y)o7PKY#OkJ9=EOi2}g`;sEo&
zt92kK*O$T1aNd9I*0|n3cSQ9cH|G{c9i5(6;g%fVbKLo!6MK9dl)PK>RdhRUU5}n^
zJM;Gp{(SDF--ug?`{Z$-ya@NnUo?7bUZv9}(6C{)-y<89-#ugd>DBc<sh;<Neroq8
z#QvU#{3qpZnEDqDI}eyAsm&Ac6>PqKIeSBgzcy~CLBX&+=T?;KCE$lx*RS|B9%1{Q
z<9C0ueg6a6?>P0*?}{jRNR9`;)|FJ-@%s7GqS?p&X6fB;CMs`XJ1#|k62DIZ{8;(f
zyXe&yW_Magf!^?d^DJVYX9>#h>|yh>{Ne5APVTpo1O>hNJqUL1uJQcX!6a-veEH1z
zUEkTbg#d%W1D@j=jpw+E`aKkWh3mY6+vAz+`^CK0o7`aHtot|4qd@S0IKVu}_BxR4
z^GBgq<@|at?W5oUJRlA*4`SDW(F4kly6w8P1SlpC+5VhOwS7E4EVJi_xy2uTfAZ${
zC((Z|b?g4VbnS?KKPUVP=ko)+e$FuLPiS6n<1`vp9`N4u^xm5;3UArIzdmT~i~3%|
zcKzP2@!(TS2@pIW4vcvaw(G#qpJ2TQCDZW=>is@m|E>=EohH<KcrpPG;31oF;P!Pp
zQL*>0{k_{v?UP$53gSJA!F-P*#tq|kxavlNI1nQaaQ-FO=U+-yoOa$Fb9T~fp;yl-
zhuvtu2O;dwvqS%)@!Ud;bGXJi-Rosgk3W|j*Q<GRy&8B0uM_*vA&3Ki;~;3)gLvhy
z==T*2nsn})3-11h0>J~~0P{ffb)e@JpPbt7svQIvO&+5DUhugsuj>5Kq}|kQ49fG!
z-1al-&&8~;v0eQaa<@^SyF9oZr?8jWa}#l1(cSkQf=}@2=KC`dPlzWG<EimcHE*u*
zYnj4q;RORPZ8NX_E|P7;^-saR{wdqz7uNfvv(>L*FE1P;jowS$!l1vu`i<@njcT}f
z2lX3MeNT`m`l3FMyVn<MPwn(#sX`!lKpbEmU>!(r9Vk4$8m-!D0|jQr!;53byg6a*
z9~1~45C`$|z^nVz4}CZLx$0H+5TKO@(XStK@-J-fH9~*F`M#>4jB9bT51O>e-AsVs
z0de5}JP67<5jH=)y6z>_=ek25?|lm6d!L++-ys+CeHa)QcgE%Zxg)CoxH-2lY8dtF
zon!ki>T{r@=2uYGwT_)C*1Yzpy#xpz?BXD5^ETD~igG<n(Eaf0dvUsUy}ExMcKdfI
z(_u9M5834*-Fbm}sol2)zJkq{=<hjrC#TN(Gumw@*~S}3G^#t}v@H~9&x2^skH>X&
znO#@s#9y|K!*M*7-Qy|nFE;;o{B+k(H$AnV`h}f8lIA@#eHR6~!$Z`clY{<m-Sp#!
zD`?Pp5cR$_xBS9$N&S5;Dfo7mZ_&T+`{iHMt|Q2_zh2!(lCIrM`y*QYba_C%aJ+}o
z@8=@Kws+C!v3ZqFn?Ql!A&ha5_jTue?S?KVq9g97Pw#&Eu<;hQ`$D4DAHRRE#C?Xi
z&oI$_hIoD_p5H0r^E>^H%dmdG@ta4qhxG*O$!{cf9-O%Ho;n@=EF?gC9vTcQb4%a;
zYbemlL(sn;1^svHSpRWeJsjuNy&A`CJ#RiJ?PmLV<|Vbib^nyXRoW~gZedC71*P1r
z>txck<Mw;fV9%TP&w*F)n$CL-!}h&scl|2o*L!In1%`u%px)0Y%JZ*-dc9`0?rTc7
z-PrG|$cN+ld5j~*(arlnf_`1VqLsTpFSvLo2?`!LuG=#=v?(UQ?DF9E^*3lQJI`qd
z%6T>TiT86p_|#H@IB+)(Vt;>+dOqFnlWx}w`tRBBAMcNOe(Yd^IPfM8mR$4XlsczM
zg}~%^@awrQQER8#X`ht4Vd`HLm>Ca2|6Na%>%xQj`xE>Q_H#g;#t+Z;%Ix`G%i2t6
zUU1_y3Iq>`1Iz=g1L>^;^nrq|{U16&fmR;8x?Uja=TXsrvF>ja{kbjhFBs3)!8l=@
zg76$D#3|x58*y5H%Zt))<M&V?cyJ>QM1AjOg^lg%zmU6)W(&vEetOfcTDu7_Gahh$
zI<xcB;4d2diE@7!{224*gtdQA@BkhV2bc$LtpibiKbv3g4M95{r#|{!5%r3Ppx!^8
zuHTazOq_N9#(5M79uNnZ2UrKvTL<twDm;%W(etPdR2_JJPL*FN(8>euTle<9^`PGu
z5mo<1xgP|6;C*aid><R)IvjC5dO-P6w_UfE0Ko&|0Q11_bzpXxo2K<Xyq5r@$wRu&
znHRO*{iA!#yYs3=G+X%m&F}5#als}U7Q}I8IF2*@8pq2!6tz0obsGsQ;yigY&XWiI
zxxZ2I|AK*+wwYId7YQ1=_1+}darZq`qVidId^K9N)drd^49fGvH9y_H|J<+RC$xj(
z?QH&Df%~a4yPpdDfj``ri2E|Lb(|Kpc1N_8_}}op!#m{HY0;rY#}=gq4R15NMGo8N
zACrGv{s}FlXNTukVfnf(O0_7}@<5m8hUZrslGF8H!v{al#ku)4TD59*-C_Ap*|m^l
zdxM+(PI9GEY70P4gcA=cO#m>E76VRk1v#fPQy$V9ln7PIlwZ_$Cr;DlF+!n=u7Uwa
z2GX5BsK+VwiwDO6L_t~%_%&{5$KQDZzTM?Jt;vN-Niz_(cul(F5mOA?IHNy2fWbq0
zJVb2zqq|@L3`Cd#d`|Ge2M^$ZF(67D@X^2?41fVJ00zK7I2aJ+^A&#Zc*>NAa71Nv
zdy(n*dsW}95%J1D*mY<9NR~&p;sJKR02lxRU;qq&0WbgtzyKHk17H9QfB`TNRR)qh
z4+A|OA>aW#Fa{6@hy%od^E}|ufjt-i17N^A27>ZAoGt@F=^y;$BkVR0>BbG}fq^J9
z;PyNe#a@)hG@ZVSGH&pLj~;jc4~zlC0pb90fO&v*0P6tO0jvX92e1xc9l$z(bpY!C
z)&Z;oSO>5UU>(3ZfOP=t0M-Gl16T*J4qzR?I)HTm>wur@0B0L}Q5nE`z(>Yy9z-2)
zul#bGKiKmoF2O5!Wek)Q2b`tU_a)gyIp-8Sh`~eZ5sn&Q01SYE>@X0t&yDP`<X@{n
z8xQ!)V*@;Z2gX3UalqLN{yyFQqaGjol6VMygrEmuVxXjWMR{`b8a#m~#sK0V+Ht^p
zkA2V>$aLNYt-tU$TnuCyzo^gS0X%>Q#sK0VwK(9~!QFc`-oa%Ao`;zMui_ncdDMUh
z@W2=dY8>#<!(KE7f*Mcon@19O$czWK=VQ9#;kF;J7fuG!jjyoO3+Je!S5apmEb|ln
z;E@a-g3UwJ*CX`bo&nM4wY}GnX3Bu*;|~A$2!jXkz!*q34mf+*yUjql<AHj7B)|iB
zU<@D*5C@0@V;=Bmz#a^MfowA1_vcbJ`<l(R{pK0%@zDkk;DIrKI6xdA4loabwhr)F
zi#-?s17H9QfB`T72EYIq00UqE41fVJ00zJS7ytuc01SWuFaQR?02l}t13~-UF6{n>
zE4t7dFaQR?02lxR(PSX($1CU`k8AMY4iA2>5*Rmk#tl4zN5%l+0C6C49PrV>9t?m1
zFBri3&si0X2QS7NHo*WG00Uq^Tn0q@9D+YQR=|U|Jcv6Z;TaeJ17H9QfB`T72EYIq
z00WsafbUbBLGS<`7z2m{#6dc7z*)l{3}lx9tY4ge@Bkhd17V8;KC0M*0WbgtzyKHk
z17H9QfB`T72EYIq00UqE41|q=u${+3e|Us~2k^icKpY?r5C@nCSO=o94)9)L4+g*h
z7ytuc01SWuFaQRk$pF6h@NowZ;DIrKI6xdA4*Z)3Jle1a17IK<4B+#Qvko4>17jfD
zall6bdoTb7zyKHk17H9QfB`TNYzDG@T%BtB!DbuYg8?uA2EahF45a$`0DaCZcmNNK
z0mK2~0C6DhJmAsk{~qfC)&;B!;;svrpO~MRpO~Lm2e1xc9dLgg;87*s9`haZ9rHc&
z`R@13#=OP6#k|dY-eMkM9$_9OKaY5{c)!QI#Jt43OnzQQeZF|VBB2E^00zRwK-AYw
zQTrdZI75Ge!GPEE!7moP?g#AikpvImfiaM59B{U<2LoUL3>XX~yM92Aa|RxQ#e)&)
zNQHp`%)emGKa3Z~i!qRD9B|gK2Ls_`Al3N*eI5tk0X#4U5C>w!0rvv?pfeEG`4qHq
z3F|nZUpx}M<{@bD4u8P_7ytuc01UXnfY+a|u+K*cJh;n)8?g@?U;qq&0WbgtN@4(?
z-<%=vkeG*(;t1td1`tn);|ctLAI1RU0C9jgz&yY@5Uq89_ZE9F00u;2Al-FXq&~ox
z=rE9Oyt-8{I!t@pQ@6$gcKA4f2k^icKpY?r5C@nCQCkQ2toME&HC{{VulJ1Gw@{K_
zDCfL^2k^icKpY?r5C@nCSO>5UU>(3Z5Vm!I&uZ+!K(HAI+q^`7csz!chhWdP@OvND
zxI@3fy{<%Me8ai|MP>ZZPpm_X0mK2~0C9kMfOP=t0M-Gl16T)wu?}!AvCotN#CtH}
z9pi*?VhkV-f*l9kd+fnLG#SA9#m5~yfCt6^;s9}gIKVu}bRFQc2>YNgkm-1%J|AiD
z03H|vhy%m{;sEmi>j2gPtOHmFunu4yz&arAI>4h7dr=v{`pHKIJb(wr0O9~~fH=TB
zz&e0+z|D1lvx$9B8NhnLM-DuI2gU&60C9jgz&yY@fOP=t0M-Gl16T*VSO++(*n<Hu
z00zK7SQx<Z0v}2603H|vhy%m{;sEmi>j2gPtOHmFunu4yz&e0+0P6tO0jvX92e1xc
z9q?lv;4EVwRR*xm@R0`(;DIrKI6xdA4loa}4qzR?I)HTm>j2gPtOHmFunu4yz&e0+
z0P6tO0jvX92e1x=dmZ4t!yXKP0WbgtzyKHk17H9QfB`T72EYIq00UqE41fVJ00zJS
z7ytuc01SWuFaQR?02lxRU;qq&0WbgtzyKHk17H9QfPpA7;P>_7QREnX1_NLK41fVJ
z00x51fZxYUL2pmA(GEJFvHpw3FZ=-mU;qq&0e=~Y{aiqv^B*-H{Ea%a1_r<Y7ytuc
z01SWuFaQR?02lxRU;qq&0WbgtzyKHk17H9QfPwTG2;1*C=`(?vA~F!R_(Xqrlz|8E
zz!*RrAP%Aw2Yj@HyARrY4KCC091MT~FaQR?02lxRU;qq&0WbgtL|`Cj--CnVuLxrs
z1|Nbl-cEjoVGPg*FaQR`V!&y=ft-&McmNNK0mK2~0C9kMfOP=tfZaO4S-~C*fB`T7
z2EYIq@PYvxe{oj9gF8HU5u<L{#Q3^1zS-u{jaYyUFc5YIvOQ1HK94Z)03H|vhy%od
zn{mL|jP4%uDID`Dx)F??g8?uA2EYIq00V9_fa62XXe<wI^9*}nAPfw|&J*PMsDcOZ
zz!*RrAPzVK*n@%SF@W`tk2!b%4~zlC0pdW!IN+m#Js3!i0jy)3C-4w#9+Hc8Xn}$B
z7{L0#c>@pNfid8A9B@{#2LoUL41fVJ00zK-e+;;NJOg`t>|%Ks(O4bR;eCg9$gk6)
zV~hJ*lo~X=&F~gEY@dHj{&D#yw2+=1o?nIK>$WJ>BB$knF3%0me`|2FNl%uQa_cld
zc2c$Ps9$(~-@dD+)L%z}h7TS;v*pO5VhT);hlke|G~GUR3ymuFt5<l$fmw?Q&>J2O
z&)#*(j$Lah(4L2j|NN@!_idy+pp}Ql6GxqVF!w7WItK0Y68?&Lob%fANB{l#<{}zY
zyn5e~T935bO@Lza(5Pj>FMl_kMuG0|AbPw9<+wJg-#?<#?+c!}xA?b%yJ%4GoN=$W
zI-%Ka8Wf~A&t5UL+j0^Xbo-QOSk4dObK9NIZSd$VkA9uEe!Xt{?+>3^M1iUC5Y+FP
z@H^P&2{3+{ji2bpcb_)yv+!WeEd<DUIP{<S3x}`RM}m%iA1}Fm{O`3rzxfGkdtv#W
zhJK`XJwniVi^};tQTvJG%gl~1WBCip?+5TRx$AdCzn%jACBE)O)Z;$4KDR~fAN)$~
zd~r~IH}lI+#7}DR;}>6Uty{1Y&F^~{cZ@skKf(PsSO>5UU>(3ZpsfRbeXl?}!MaXU
z^zn*4|BJl4=hNNd^QnomF08cVO9BKBhy%<6tOM3{VEOU+uPvUukpNTULG<I3p!+BK
z^_!yk=hpEO>{yRu{rY{yshuHROV`?au$YLB7cKf`YM-`?X{w0pQIfkJC9FI-#XaO=
z{tk?B5o26Fe68l1&$e$NVM9@`mlf5oR6lQ_pWbnb+q^yUSh?pKkC;w?;DN<~+wmou
zJ<-3v_~l<vKd-&MFVHLdH(xlu{zoeolb~U*x62&WrQ%lN7l!rwCHj@w`Gv4P2XpA{
zYx~^3q?kk<!#cj`ml)4)?Q`(>4?A|+O@RLLfX}_mKKH`P-=|GpdhW40%Ly<VJdFF}
zP|kfbw-cZ}4+|#lYFlU0R1#Fg`w*<}LjWJ(1MdmJds0$e2k;y|f1gL#`isu@AN+Y2
z0fGm_0p>xr)`6&>CtTU=m46z|+DCz6@{p~05%zY^IPSs+w=7#sqk^<$yH<@yPNtb+
z)PFBS|M9!2o4*^Udp}~W&(@7S?~8Q=2p+;52T}QaMn9u<JSN(B&nb7wueW`$fdVt*
zA!z3lx|jQFbp1nHXjswj-*toj{u)vJzrF9coOz|!5TMvR_&wjk-rf}h?>=o_m*+{?
zP?XO}zuv!zcA6hMsoHlGJS65J)AO?>)t|ldr1{d=A_4>tUc^CI&qw}#|FA=E{j-1o
zgT=$Om9Aeh@RMaUD5yObIqI0{`>hs!^GMISOZQWt*gS-FJq`Q!(xCKjP`{cBMpoZJ
zfZ!oYaS)W_9Vb7%I^Wgvnr`LlzO;ghj$yt36#cTlFWK+UQ+yuc^U(Zx=y&{y+McNQ
zf#JEUVmx;>b{s^FW6}JYQ*Y~;$G)+M1Re1l1?zJZ+<sp`y61yLyPgjIh<QDyTk9ge
z$Jl?50WZPgWmv~&7k$@pI}JMGzN~QEm*v*_99BCg)jn&(^SL>BqNCG!UC4uVJ%Zcg
zCHlM!`g`+4^&i(8dV9TLQ0|NHYW^Rdz3Y-4yVg=*c6ku(y(3}q2j8oM^}RaN@ygcy
zG3lOvb-UgxC*5AQeT%~c2p$jzCC`JVm1q5QsNGtUDh{4-&BePKEvG>6fH=tBJn;MT
z9`ik#^Ii1crGxf<M6b?g#M({w{7%@P<B9&^`QdneIG!Kw^*Ye|r*j^fS>qoX6pZTq
z!?D-rMkT*;bLG}^XKkUtX!DS*c^{Q|hJK299|p$7+i^Lj;!T%)wWyee6)UcNYGTXw
zUs9l$Jm9z{+Q&5*e~drA-(>#%ZrF@x&Z@d@Ed_!H!~y03)`4iO1DRfri2BjG{uJYl
z@jftXF^P?Xpk1F4c7L;Vz8~$*Y&o*1nEJ)T!)ptgZlAh^MioW9uQ6MGx&7Rou-fzc
zdW}aOEB9RE5z|RfFsS#(h4ptDzpn!qwY{4!9AE#V6^jWFJS2((zt=Csqr35l`%r>;
zA4=?b0pxMM$nJbm(BmSi-?!0ke!d0IE%NrcML}O5!s>sv&TFCF*IS*?Y&QiD;2}{Q
z;Cgp&uXj(EUwZTG6;rz{CqczS|CzsV_=<f52p(7*ymiDa74yz54n6wyJ1*Ms_jhxz
z`3~xF+PgPh*|VWkxey2*5C_4T2VuQV1N}Ps@6R_EQSbmB5C`7R15vMgfnR-BO{u?*
zfCun^I1qmx;C&nZyl*4j^U+bMSMceb&5AGBN`T-2aS*<F(0JmglMm*8MS$P|abV1Y
zu%0gp%k_tT-=Bc?zM0ym?PBW0L#D^WPIe(r%sco2KZ)ig;sfy!<oLjSY{9&bO|)^#
z^!Ow!^-sHH%$&Z%<`K6ss^<@adY#&!el-`2tiFQ)!2{yJy?O9y<30-y*4#pX;6da#
z7<%=9USGG{Oo8A5aS;4@5S8;hqODgwujy8<?n^5u5Ii6b#GMCW{XG}`!hH?6ufhBK
z8ZgiJJUipK3m@FFY%vubgZi8ZQT-0f?;q%gn9p~?^WeOF9vt}fm*1fNE(pKze6&Q*
zM+-`PESR{fZJkL|Nl@_BChhV&w_Ztr;3150fbY*ZPlfYTI8TLjAhUJA>*wzIZ9gi%
zTce-24<pF;VFV?9{k}e=)@SR+p7+H%5)|~CAG9aN^@VRP>U8LnGykDL@Q^4Dvi05-
zd@soCdw~;wQN1q<{r2~HHk&8iy6E4}<Q9enah-se*9io5T!wTlU2E^bVghuRhipA3
zGu`8)4_~Xf=CkcvNZ1kYIdbPcM|fT{p4W`$HH-heX3SH}Q_R!g&(p9Um$}uy*RN>t
z@SxMSQ_(P}-w)w8&NuwsbQ;y-;Oc!#YCY0!H^~-myyltTI)9#9=od;6`NR1ichCQT
zPkTPoJwJtdg-0BiwU~g1Xz<|I`QvPTj-%b|9!Eu$f81}H+5M(|txIUfpZCl~ooCvz
zU8}|;C(}&v;oJ#t?fI&h0_}MS%5}(5^Yf!`JCqrA#5M}_mIu+FOXRnIULS{Lx}GOn
z-|K#8_LKDodhUFlC(C2hJ})xuuV2r-*i@_Se|HVoM}diXxVF;uO9p<ji~`f+At>>c
z?0Ob<J*WFoM3am9yhZpmd&Sgl%ZaEwoVj)PQ570iEd+*x2kp8GAg;qPUx$Nu2v<D#
z{d^m=H?!r)qGAdj{NX{=<0bed_VFynCs^a-*L9ECx(+T|$6Y~rzHw0e4C{Ts*^19}
z+wHq}!1<#;SVcs|uzdcwb$^v;c7n3rz)!rtAkq5^MEm_ADE{EOC2y}=0>5JOJE&jH
z1tY8PAfjT}e@{mL(!1X3&T(_Q*E?w^^%{fnd}t>>7k=|d&$>(ZQy_Ri9AF+`9Y}8-
z@awu$v=glN+XZdBM7vH`yFbA##PjO1^SnCI;tKwxcRf0I3pa0~p6Bt~uiQG#kDXNQ
zI~sJ%^nNO@>bL%)^Zf^Z-bKQK**eY>?fg03vxN67C3??NRN~|Q4b>WV>-;N86-ArR
z@JH<5=iC|}zmKnC+Y8!rZAASZl&$MlMQb-AkJUvt!2lQl1D-M9)%X9X*`4+K@!j{2
zJV=6uo&{EnVNv5DTj&3>^&Zy`N5630`ir*{Q822{5q9c#)Ssssl>Xy)=Is2=9ChBq
zaz7dRk@&vsu-;#ae!27hT<{1Uhp*U2GL3_W*A_J0K6MKPf`=f-LC)Y$PTx)Uk*MR)
z+t>EFeMvF#3Pt-JD5(Bq>-<lq?MCHuBHibah*mGC^8<cI<2{Axj`L-8b6)!8!BsSD
z=yYD#FL}D-5%hXqomb4(b;D6<*Q;@ZU9r#akNe|L&V4ht6CikqLL3DB^EK%H2jzEZ
z_!-Xk-um@i<Ye2q?wU*2ANTnx8Wa?DUf?)W?BmRAjgP3GmzkbF4qE+ejX&D;W?hKN
z?-%GNe$Ti2JwMayMNmKSbw|k_KSQr{t-S|}iQhaReiO&9SNw?fyLd_d;P}g%<FByt
z>h%0&$iJD|r|n`Y8V|A0r$Aop_W-0jKBAsSfnT`3>A<YT#3>H`@81_(bnf<lNVf5*
ziL)-OwB$<y1P_P<%!6#N10{WLLwUH~Ymf27c>djV8butq8wcIX{WZG&p)C|BCJ&;$
zC$}VjoW93E?(g?p(ay7l#UIf>f3nT9=={JxcaI-L;WH@jnU1QTqF?VPN*u$F^wuXl
zM=bGk#A5jobw1*Ig1_Grz;`hC4$Ad1vFi--sjbUl<twPiHDMhO^b7NHYPaPCaez3;
zBo0Jd*Ry^86YYz6oiD}(<09hq=c2|Xjthc)Tu{>Q^j??ayizdED+O(wL_Kc}zeb&W
zF!w709>4?Q0P`T#bpXek!9L!^_+k7qIgc2&_=)OuUa|f5`?)Tuw&&OL^wG|eT935b
zO~3<qKpdn$4<0;zX3LR9#S|D#9z^}Uz%Rd2{d`YWKWOJc-SRhV$K9#+2gf7c9FGL$
zdeyM{ne4g;z3e=H6ug7?Y&`cCafi4|6nEJ^9`gFSlw|FvdK?LT{QhD7{lly0XGhhp
z+w&{c-={^>&(`O-XgnnAPqshT(7re48Ne%e#qUY!|DF_<`4fBH0rE%x{rTo1nko<f
zyZYmoe{Qmo0>J~~AbuWr^*%knem@ObJE^W;qUekIo?-YE&i8_ceO$d-cVQRrN40xD
zD&oQ#7s;MygPyzh|9Q2pBfhQi4L+jDhiK0)f<Ng!{};Ta##>a^yP%!_@Y-Lut|Nz?
zXgu#ll=ExwBfZ~uyyi{9UqS}3PP(&BdL759+Q;!{YR8}8D_DGabsh?K@q5&cU2ADr
z;z0CuE!EF8uk=OdDck;uc7Dd|^AJ(BkM9HSd>;Uh;PJq$#l)>R@caInsIGrr_1mlC
z7T6W@ykl6$<(zS^w>qKOZW>k;?Kl_yMDscjjC*F|?)LrJVYeq*yoHVLupSSfU&$SR
zdiA{`S-ZFn+TH7*!KYY!2Icv~VY$B@{mAS&7h(PW9F>0g{rhmXfB!@KVqYI4+V?m3
zliB$T@RuHcVLdKietiCGizja+K=6P#$lg4_`L4w0yYRVX|GDM&_XW4xL)<5hd+-x3
zez1<&uVYc=B`oth+v9h}aTh+gW!Yj11P_RV@XdqSJ5QP~jV&T!LAQ^iVK2M)7mmBB
z>Bp7st63N{to7NtvFClUj$|8yI<If<J1%El=`|!M*!qjk_aFRu7Xg9?!~y1k(>f6J
z<B3lj_gQ$b<`xnZ%=YIm+ULhXcppeK-Ukx&^$q@K_xm3B4>teVK3^?b`^nz#;k8~^
zeh)!E!u7i?##4;(#OH2mpS$>8=gs+T(LTR}>W|;gA9A}rQ9c)fa-SmnoY`_@Q85iW
z5BQu;?Q=SouS5Ttzi{}9eFO*|5C@nCSO?Nu2g3HeAGiAB_Hi)mm9Dk-U@-v?V)3x>
zn@4)qUAmtF!2{v|^8o7r)&Z;o?ym!)oCk#;V*E~*={Sh`dq%dOPv_QqAEIt2?C0On
zzi9l<;`KO-dOjm4*Y(5CVBLp-@xpio>As;%<2CBfUk&Pg$MD<wJcv!T+WvRffPFM7
zxM1S0wsj^=B|z|iIKVu>I*{Hv;Pw4Xu>Z}}K5ZA%pz@IE@pV}1<2ieH&K{n#7yq2S
zLH%kj7+HM>0fGm_0p<bL0sD1eU8CMN3?H|T0>ML+;^6$geOFDXzm5R2%R|(!e~Ehk
zMZK;eUB7bcG(UDyweM)QFkQbzt#|8);+*@BSWbc9A(=S%qV7A3j&J@C2`jF?dgWJT
zTQ4WTaPtuR90KGsdya!w{JC`>u-|rqcHD{YY1#drmhI;rh<beL<d@s)O;CTgh2O#c
z-QDl;b9>&vo|wl?7#EDoj$Lah;=r3Y$oBDCRNB9A%|Cs%)%$_^h2?L5zw+w?e<MM|
zk8&QL^2U-P0!)vG^<R#nmEYJ(MMVjJfo9;#j+a$iaPckzwCACC<mvl=d;G{epp}Q|
z1zXx*@KP}m9Roc%W#{|;a&d0+daDUAyF3Ivj{a#|sZyzvwi2M&JY*}*%AG!a+ygZ>
z(6C|a=I2(d_V6|$D$dSnU*qSh?-SAS&U^Bku5G`H0>$7VTkC`<?LPQ!?PpG2zkvn?
z$9FojbK|BnXjstgIPhwHcH7UA?Dc;0lv8RyUQB`DAyFKZ6h~nx|GC^_m+u+3lLiGv
z;r)f@mpwH3j$#S~4~T={%!3APes6pAj$eqVII%gMePpvADUkDU=ltIXo>J~G2|D`y
z`=BVFlkme|23pR)yJ*BiMHC1gf*1$GFB{tKwLLp2FgrXf?6b7ripuM0*wNor8Lfqd
zf$p7~&HnoOy#yFc9#;K##=KWwIx;U)bUU7ptFr0#Z)WZ!nPQo(TdP0&^s#xsVDsR1
zd<EU!@M~VqJ$Xnk3KWZnpvM*b2LoUL41fVJ00zJS7!ZSjOUAV5xvud_3Iq>9ii4r$
zM;1K&=?^5T_^tHc&#O;eKt)F}RylY92EYIq5RZZS4X=Okn3v~JU}`-4+N#NxHXS!m
zpf^0+T>ZJHhJ3f30F&e4&>%W|=FAxc=q?Z0Ua>1Zc169Z-)|$L<0JIIcL&H$k}9@+
z=)K1pwAe*~(d40iUi*Dhc5f#@@PIf-ejYUYVdkF)%KuKohRY8A+4Hj3c2Zz+Jlxph
z#Omwc-9UiB;vwgnjtvekTt|W6fyKer@5+Am@i`kwSn-4(9)Gg*xKeph$LoiE^})Yy
zTTg<9?=JiJtRXGFAwk2>|1MjQ(`OG28s1Ur)8FSd{D%02r`9{I)_<<pMuFhL90#v|
zef5hSThAw9!`gk<{&nfUHj${~2TQxXdVZ;m1PC6yh=b)d&g(Ygqbhkoe|e~X`LJI9
zm$#Mz!Gk#tin|>B{ZY@XBEV?y&~DaG)#{8{LxA=?yfx~&yB2KNMuUo%KQVUX-uo)%
z0l@>}0P_IrKzi%I1AkW9*}UcN1n3VBTOX)$Ua4&d2@pIW4lob0wGOnWD`uCw_9GG$
zbo+YLvYY2OYCe4)1&YbTxT*EZPMBCsfbR0})m>fA{;5hajS7zMJ#InXMa48zJaOfp
z12$g2jsU>};vn;R(B$H2PkwOfVwx)cXZRKO{?YU!nku?|`~Z95WMJsg*LLaKeKpNC
z&e=pNo%2Nz0fGm_fp_yDD#yWI_w%G%=bvzDhqW{)_{3}duj$sUn1&UvZ2iwI-!E8A
z{lY#Y{(htX1@kD-T^=Tlnkn7#_*McGn};7ZKlpr&+}$+W_{5Z}E`9gBLK<|G_MKGc
z?w;#N*zkgC8&{q3$~u~DBy(TuRE`!?Ab1cV4!k~_(5mdm<bxjTC=fh&5eIdvY%TuJ
z%d1J)(68?gXvf{}YhJIXZrdL*qr>&n#xA2#$6*EQ>JL7!ji!q^mk+t@(8OOTP;4H$
zHT>!DAu@x64e#nTvDP!k|4e|{<>98~<vy!AeLV$&hbYBC(9e@rI8^zhqMjS6+vxT0
zKW^E_eY?@RZ}-rocPdUl_Xpw^zBKfd@%wKpod*ODEDj#7{cMFbFRi9g#}zNXdeh0b
zeoI8df9<ZcV$^Fp2~aE^{<r$CVYwss6JTN<QoX)aRQ(-q-BY*5W19$&^WgUR&kJes
zzDHIsr>SDgJKy}~whwkwAb3a=2QPQKyv<1`ZlalD_ntqWUS|6W8Z^A`kAB-eDB4PZ
z;DN=#H!BuY8+z7i8gy)Z*Mf^$?p#WN!RDdguy6l7W5W^xwC7>zZR1vKE?!MU$B*yn
z``^n4t|dV5V2*<mKAR=YIBE+O6`#Mh!Fku0-$jA$^6=qrg<p(6=6wqEhKJAh_nNfv
z@1LmXnC|sFCDpt8l}Sy?ZrVYC+2ujh^L9b^t74O!!AJf4I{`+6hhqlSzOCkuld0eM
zY3ZWz&-GkNvyDqH{`u;A?ps3AMYqpOyJauk{-9oF@ep?Y(v8pS&c5Y`i>j}sK=6P#
z$lg4-XxNvP3SQYu{K9`aHJ<xY`^7YBcxKU0NA@c!rcuR)Z@zHn4I4I7QSpk+H@?@n
z=S&h5^y@hXE$FUcttJ-JRPpN5XZGCj=s^Mm4~PTI1FQq-tpgidT{!5ha}Lm;;QK8P
zHveSM4jMH)c*T2%+a8!gqlP#A>$WOS4%$M2(dMD*E7c!5?~bk1Z|uF|>LHi(_>Ka>
z1LC0MdGM|D-_NU0T|k3^AKY;F8C_1?K*NfAcf3_@+2Y>_P)r^s-1$+HHH$V=Ab3C=
zU>;x{NN*k3w&M0tZ?&68gMzbbK6US&$5v1vct9Lr9$+0vZyhLi`t)%R)Yw3Q-~n-f
zd64cpa7U?6f1lg%8v+!Ahv#n_^~7sGt)Reg^RS`q$nK9lw~a;>qy9YX*#1A=_VA<o
zIxL~U#5`16(|6#rUo0Sg;RVP1eP`|Kzb8S#J~u3`{=ZY#(4gTZXTQ>a$k+oUY*=UM
zEt~dVRZOCeO%{#(WMs{62@pIaii33TdrG%n^WTPzJGs`c1n3SAv!DF%xGht1@_@nO
z;XgHZT)c1c777FpVUB|*FKkeB|Kcqa7)>6IIe+@j);DjXKyP?hR(j2^jit>LC>9UT
z^!#=D(l2&UAb3C=h&K;XeLi|n^mn{`@)ZN;{6&L`L4W>myIDU~t21T|0fGk>2OFC=
zyLA7KnW4ucXMH>Mhlf{EAb3b64qh2^>!N12t|39gs9r~aen<N{lAw?ODKjgqx$Emc
z2@pIW4loa}4y3mZ;P;$p{+@$z$GGG7T=#y@-udxU%Rl^fAq9HNL(tz};kEyxs=spm
zvDFt7Ab3C=gl`@^+q!J6mS^uJK=6P#Fy=x2Yma_BykI#E8-8_Puh+gCxPb;8^DaE|
zzAI-)c|b9E$n<?$SGNA=mhTs=ra<t3I9N?Lj63DF?{EB?0R82m_w>iVZacSt0>$9r
zvQnS6JhR7Y0t644!~ve;!k^>vZQYKg%T1X?-Nr)&d(QrH?P3x&Y&o*ul9~;sk)Yv*
z^&b)P_j&>Z4~PTz=E1s8?kcLcW;p?Z2a)5TbIWee{<&)l1%d~}LGb6n?40&Bey;jH
z2|G@G>gdngeZP&0f?FS`a$c!z2MG{7AP&Tx2MhZw?YE-xdI|&&hy%<6tOMz-1L;2B
z(eHYfjA_wxUE`G$=nfC#JDu6Nanl(T2p+tMgR7oA{<j^KiU<%qgdq+lwSV`d3+5lB
zK=6P#@P8h>{`;M~uPHx=1_ht~?C#=b4R#VBcnD)0T>E^hDSyq`Oo8A5ae#S%bs)WU
z!0+>DUbold+i%|e^xW+f2p$jzm<O4z1Fgz_Og`wbj{1d-R*inP+^8ZN6g>6YR$u<-
z?7;*WEgs7K`~T{kcK2Qi1P_Vg;JwRQ_n$HTAn_Z6_MGRS_&f2eK@Cr=v4jG_gO_p8
zq4#l>y1%!Xh6QhW;q%s0wymN-@Zd%q+|%!Hrx*6VOM&k4FtKjiFKYg}llqPKm-^z#
z#fN?+sbcRpPdTOb<Ha=Wc-5e4Wc?F+2{1Drt{y%7^?r~2Nr2!XQ5<|*YUaxNwdN8a
zct9Lr9$+0vZyi`T^#3Y9IBqcoCg$OdEAJ@%@x{}LXn51|a-UV5zMcZXLy+TO@1s-4
zzWV)Uk}W*(h$+uJaLqyL7Y-lt{<geIMKtXA@jZS2d-=e%B&>+{M2PX8h!Hb7Tt98>
zG71C_hy&+&;MeEP^<Pyvu<x506bK%o5C<31;(d>-UQU7O@qqj6ai6^x_t|GFt~N;T
zt)F-83YsbU^?4oCcG}JQsal;eYY0#*9#)^x>Y=+16j7kJJe+>n{_d3)|4o6MhZ!R_
zl=}VlWz;L|H?{WS{nfTpukne3)gvZ1DxzNFhd+0^pwH+v6et!CXS_eD*6ocJP+)RA
zJaXswon<cFOT5O{t{OAz`TI6gU}7G&&l`2@?ghIjFg+fszEb_6^X}M6MaR2enbf50
zrX5stoSZxH++nqMQ?Kxy?gy?Zb>}uJIv#aThh4>2e?fuZA(c2-KH`%lb>H4Yf#Klc
zj5@dG6}8<#fZ)N+IPm)WD(t6rUZC4q7ydP6{C)}q4`Gah`X~OYM!jF>6Cij9Lmc2f
z2K)OMp8U3Hi;L)1;x%41^r)VbKU_`x#(9@FDyVp11@#*TG}?d3h+FnhpgTO&%sJ)n
zQQZ!au;Jrx9=~b)yML3Q;4y=0-&XU-$y8MQvc>TST0T3Q1Ran6d}+g5Dy<@6!=jeY
zUe|HJ9ts2xL5_o0rmsJ@Y3^1EOpS+G$DLex(q}s<5IjUN4sJg7v?f>nvyzC48(%-Q
zV8wxL1PC6?aZs^I&fued{+)P@!wS~bAADdN4Jz*IF!qEtmq>*`@PIhLJit1T-a4?X
z^qOBAOPeVWJS2*P+p7-w<Ebl;$`cjG?q6FtpvNyn6m<Ld+&k~dZ@RYqDhl+MhuoEa
ztgSq74*`M)#DTc;0MF^>&*{G9ng?FHeb+`3be!9G-&ceFy_<-J<6o}0?5ua@(y-vZ
z!r8N~dANx9jRmKE)^gavl~fcwzDLEIJ6*q*0>ML+<6!R#OWK}#+|M*sy#A?2X8!rk
z5()$le#SxP<_+I1zHB+k76vWeTOE4mx6h<)R5UC);hgC|O<qc(jvLyJ?Ectu+b9q`
zAPz7OunwfR4jd}jbM}{O7ZYG+Jk-x?zi-O!?IdU@`Skb3*&P}TZ?=JmhW-C2{luc8
zV&XTh|8f+q{Ki)5H@^PW>G#~cW-%27D~x_|!LAzn36S$J>G|L8dbGk+0u-AEd|qbu
zc?tf~<L~LVhacV7VF?WyHd;0M*>a<bC@?u5-fvsxqjz_1q(Jb1IKVvcY8_a$?6MZ)
zTWz60@DSuUc)Z1{&$at`K8-qh{eFb2>(##b%3rn<U}ikjm^A44JKC<KLBsXGce!KS
z*-L5Iao;(04>w-BlLQ?fe`DT)hYAKzx3JcZL1(QjJCy>#LpI|etmhFf8PlTYy2dMM
zwy;&%kI4r;*3qcp?61xmTI%0xsN2})%YzpcJiLGaz2V{Fn+`s8%iWX!qs7Cc>sI|b
zr`Hw&1P@V&gIfn4^Zmoet)xN0u-#|Y`mO~RwcNRs0>MKz;$Xx2j|lmDJrM<i_I`#B
ze=GcA{4wuS(QtKf>HqDx`C|$M59T;%f7WC3rX8F`MZ@|HuYd8Fm*>!^V3TvtoBR3Y
zzY$<Kc(}atr6Vd$+f9MsA&PPE#+x-;F08nd0KMTMDxZr%?dPI`?fWX+vXTT9Prr2J
z6Ls@ek)UC=e$Ps`-Moj!uej~$RTKywl8J+$oezn+ze{f$w_<bgY61*54;z~|yLA7K
znFI(Pf)oc=zkAw^-EaMaW*ZL;qQhs-oI!x#0datNknMG#cHgysUHY$01PC4w2bc#~
z2hv*yUjO~h-Pe?#LxE!QFyx$*b8-(Xph3aAdQGhL%<(@HV0L-<v{sMGKh(|11G>w@
zO@AC}amEeXsi=72nt%FitM>!V7S?`rOhNtaGilKAg+F#T+AMuefWhG5%GUqf^8JF<
z6euPSqTdJb<hM;*Ttv4LU@&<YGwPczPkwrc0F&ck#)u82e!qPg6%9Y1dhT0KJhGAi
z!2{yJm<PRA6|{bFdod9m$9FojbK|BnC=fiD<KX)Lp0Sr!-$R0i-@m$j!r8U{Awcke
zI0)}Nm^5mpbj#yg2{1Jt(!H+Qt9t9Ddhfk()-(zP4~PTI1FQq-tpgWSUh#OHTH7cv
zm^@tZ=TYB`nzn{y3!j*B)ur#AS4e^2!Ob`j^>=spCH8d@7$1y}m)Gq)F}Bgd!hW*|
zFg+fY)Ev0~$L2dI5Ii6bFb{&h4$RJJU*qSh?~|zGnd9G_H#PS?A}Zo~uV7y9g>l2U
zMe%w~JbyAA&!0?p97Vs5+H3!=@4NY!4d4Go+{RbN+`6dQt!qfsF#pB%ZCV}LO`?i8
zj_~F<;`Ohuez9Zg`2?699?EW>->CWYc@zj95C@nCSO>5UB(o0GKk;8R>is&OWD5@!
z>^b|(wTnr%v2)9A&;Ge<3k8CQbmAat-y^;5Z*JZlvybSsjk<+{Umm#i>&6R6Q1PyU
zH|~C<;<0&t<Da>o-9x7AAwk2T<wq7g{pk-xG`#eKf^tiGmC6Hx2gCv90oH-^)`6YZ
z^?K*Q%MKA>W;|T`e5)yc&Dl(W-~n+EKM!i{7<AUkvQw$sxajU__h0hq+Z1T!A*k2S
z{jvR$+>gIlNrB)2agg~uSlDN2zZI3&Qy_Ri9QZ#E8jgHw$L;-hQJ}v(c=h|!*HwPH
z{J4{H@_^tW%5gC3fl7yOYIlGF!9$edU`?ODkNe`UeKcx__XJtr6ZG{-U%xc{r9}jo
zT^{N;y#B>wUY<im$CtZZ-sYqeH&GyXu#1C51>5&kxMd{)+Vg<-jJxxm@saEQyL8o;
zYp5vr+>VMDzW>yA0t63XjDvlJvu9oNa1j9plZW2ZAOE`T+yV;pmWMaqtl4s5#hpY{
z6y<)*W;gt>Z1Hp32+*E~qGtDvf2aH|A_@+vw&$9MCvPM`cX{a0`?yNo-`h<6#<{hJ
z_R9NY9RY#|#6kQ#@cMqHSo<wU_FGc3!88JNhll=;zFT&1<v%D8JOnuonjO*ngC3Q(
zQLpifbE-F)G-L(!8?P_Zf6o6--$8)jA&7A>v0=NKRj$}fqmGSMjefS=s3HnXj)(Ct
zS6p`1J9DWhIILh@{lN#e(Xilit5^TuC96K8z-aPN>OUXfbbpP}BwKj?gG;*1{b4r+
zf`>51!8Z->{c`hBhp6a?`{~oWpFV87z1i%}KWa`~L4m>K;gyd6+EDBL9n>#;y6xdd
z_jOo8f#AU|4o(?#V#P~0@1{WT5al=s>;0g(&MlbNxnbPGHE!n?U)|-1zt#|-Jr5f%
zoV9S%o}C0}<)Oytm%Gkezkvi5y?&35-}Wb-@<*%6OTMK*@L(4Q_#Tz`d(<)KPv6=4
z=516|9MX7OaqelWs9Wgw`s7u6v%flPXsLg%q1nRkcV5%4W}STm2p$jz;hP7a@9#Bf
z<KI8guppj09IodMk8XX$w9|L4BSAw^o?p;(Tlf4Izx<U34O<uea%r93J4sORhVy6A
z#S>Q$Ab3C=U>-QF1IJ7{a$%M8r_-R~<zqj-YvJr4X;ARNW)D7B@yO8>m|Y%D`KA6_
z>&vd7z;N>*;V;q*yu0k<vxc<zh5)0<!x`^Ss&#v#1r*46$hoFtgTo8g(QM<hM?Se}
z_5b!zAb3C=U>?M-1LaPiKJI}U8z?Y29-3S{?a2>LT}*-Q@^JSnlbV#>w1WaO<00tR
zNnLj4kk<c~zmH}czni}2<cjsT6CijHArA07g3O*r(5>N5hYyh%1PC4w2bc#~2hv*y
z{xhfBh2uYZl>)6ic=g=w=U$o8`P2Tp2oOAo5C?~5>}%Sza2ExF2gCv90oDPm16T*#
zUk7#!=yvbZPtT@6@Zd%qh<cyT#Wx*%>Xy4H$re_9{+fXeFZ`1NGvnb<!Jf0fT)UV6
z!2{yJc^)hr_EX+~it~wR_(ipU9zDJJUjzsq5C@nCSO?Nu2QC@YqUXBCD``}*?X>!z
zj4a$mfWhYBv&+Yq{dCt-5*1wW@~bzUeCxLa2p$jzm<Lz~(pv}SY$BD;`J#vb!Gm2K
ztpDwb`om{@LV;EuJ{Wk)xeL!<LxA4!(Bh8nE$(Z-oB)HxL)hLYg#O^YQcHT3%0nC=
z4&01`T4f#?S$^1d3Iq>PiUY5nzYV*GJH4>?U79Qpcz;f~-k+1|c#3vCL#F-&^*N2A
z`n|u_^pob@^&<s>hh*a5>vI2|P-E#z8dhvNzs*T`zilMIaP!dn%~MXP{dh42ip|6L
zPG@#*+;j#F8WxnDHm>y0jU*^|-j8oD?RWM%0?ZB%+5TPofj=wlY~J#B5)^d%xjVxO
z+F$&y?+y_lcnEVG;J#qo7mWLY-M=pw^D0{NYH0b91y6ta0|9!=!-!8VDW$x?VR+x+
z9rEk6xUWU$7NrIaZ!^3_4%_z~e)sUY`5p5+^_QVb{&D#yw2+=1o?nHP)ooF#MQ1tF
zbFi#4q+FlA_x8J|>wUwAl<wQLN2lRKa_{ZlLD~0G_I=sDkG#+6AQzVD(6!^ey=C1F
zUAuQ3KBPQ@d+zHhOL9B(lx^g8c&MkW+VNgF+3{XkBd23GS-WGm9<t4xj@?<hcPFK^
zH?x0V2e$9rMXpr3bJqtN$(b(QACmX^U08*D*>KseJ$v8NOLm|Xv()|Gj&ib0_kKNk
z<;!Y4?q&6Rbd=Mj<$Xsb^RTSYqf_5}W}!zHdDl~JuyoH(53!mLH0pD&jHP>ZWUb{e
zE2VqoKhQ_6*{i!;klRb~qE}z0)Vot3X1@>fvUJ~0Jy@o1PiE$!-d%e#vL5Lmw|QL4
zu6_D+?b+qpp7+W0^nIZBJ?Gg8<@Vw=8`}w+*a^?K6JB5^ywFbgFFWBy!-pIn?|egB
zsoW<UUAa>>Qn^<)Qn_0;Qn_C?Qn_O`Qn_a~Qn_n3Qro^Yvh7<V+rBlj?OP+;zBRJ#
zTO-@PHL~qnBip_;vh7<V+rBln?OS8pzBRV(TVvb4HMZ?rW81zpw(VPE+rBln?OS8p
zzBRG!TNB&9HL>km6WhKuvF%$E+rBli?OPMuzBQ5imV0j(wnX&l^ly0;%kR{+OMYKD
z@d!_JV(a!ptsZ(#UXkU6`cbw}Kh!G!@fK2k|CR^3<Ui4(lv377QFtbQVE(iD&*cy5
zKRkc1yuJ8*pNBd;G$i-od%F*3HFNXJp7oTHIwHS9{xR_jeyNt<o&4PJ{OW_79ot2c
zre2-nxXpSrN5U3v4t{(#M@qDv_*#x6`ZoFF94RbY%`&-CShkC+=SpsFI~(RoqHTMw
z%$5A!uD>T&@?-l!ZpZCy?c!VsTRaE5&nhip;~ZRlLuo15?YHMkOCoPA2A7tywLMX<
zyyV7q$0g;ZOt&{%mY2NT%Dhot!X_>Uy*?~2`LXRj`UuId?Zz%gNU3g<UphkadMo<x
z5t3iqFY}L(GTD0nb%f-`_VK18By8Y7<QVuwj?q-N&NYveMBbWScBJITc5Ig;C8w?P
z%Q8m++1_~dNGa8A&Hs*+oVKeQ93>^YjgfV+fdi3aVCAJpN!Y-FTQShB%~29Ia3FFF
zta$t=$;+*7*$NUia3FFFoF&I-7`JolRFFj2Dz&K~Wn-Iia|KD1?Z-DONM3L4zpo&P
zww<`Sg5<_ldshVs8#u@$21XpLAi24HR;8j8wC(b06{U2x!h0%8iMAHwDoWVI;$ZoT
zic(m%$xj?DVe^iI!*3rgMSbi1zoR7)w&_!jmay4#@X*|&r9|63i;tF~xV^mgXelh)
z!|RWhGTjbdeyoH|EDp}T>sSe!I2`<S_py?f+gjOvGTV-J$4c13%fWRWj+KJG<#jw(
z3j4NY#<7wc+cw!ws@s@rD@kr_Z(LSc@_K7^S7j;PZR~TEB`>!>rd5_i+uo>gqLj&Y
z<KrhvQQNNi;6y1%+x-8XD7m?PIJ%02EtniU@>vxrnXT!URV2T*{@+%Syxxx3P(@00
zTX@Myk{jDUFPtP{;~YHt=1G!Y+s}JWl0?{gN+(NRZSAU^EJbM>@~^5=wzk1Ns!F15
zyLwlZ{MfF2tg3_!9EcnPzr0md!WK*pYEGys`LVsXzpCW*_UMVHNN#WMjXy>5VmslT
zQzUHQK!g~W_`xX>HgPyO`^!_LRJRv?J4Fi1)+V=_gw1^p{yDpvl<jT9Rn;U>wuv`Z
zll<N$w5cXx0|z3<K!u}Dlfu63C#OltZe2b(O%iSExvsjD>h?gTnv$2>iO1EH!oDr*
zT~qR6`)fr_$;<8H%C)3ax1;KuE_t<8kZp%?+jq|CQc$<s`kpR{vUMw4Tk>M-UcI)I
z>9*<nGbFdR6K<>{1$C?b!<mwsTaVk%k+8YXLEA&;NI~4z-%wu?X*)2Wz9i!I@`U=5
zAKN4U)R(Y<1Gi(KZvO^Swzli#w!Pe>0S%-~w(Wx(NTP3>n>UoOWtW2<w4s!(?e~QZ
zr7&#|9(A6S&8_pAMv~WCgCiSDLD<xns7dmh(M&H<li#CrZhph-rry=IVk`FIwEXiO
z->Y-^;AS=EEAM0DySJw{yed~t<~?}tkg_rsoi%LH5i$-~@y$sUWvuZ!yX{Jn3U8|Y
z&hav8f2r1MU!m>B6Itb%&z@IRhPV4qdE`_X^DbEXRSg-7wp1xyTgCzJU31-;GS+C{
z@6~!T7M}FMujk6xe&Ld8=gBC&a8$=8GQ8dF()TZvG4DY4tryE!^ywStUnb*#?(-kJ
zT*ew_9G=l!#=^fEl)pyC_9NT3Y9XU^!_eWaWO%#WH%o4mF|V+=Mq3$+{(X9v+hiPY
z-pwD~DPxTd19!BOv9REO7j=-a{he!`>?EU9xyr138Qz|Eb%pLS<~{rGHV?>H)b#!T
z^p<hJ?mzy1NX8l;A9+SU84J5y((Q2>+t=v+$&)fle~jJpv<z<#{juqDGUi=(=$XMX
z7UeXUJ4D6-)9yL;RT*paef5^tWGp;q#@LZEwqLz{#b_C&H*1~sh751ta?=BE%b0h}
z!2gYtv1s;Z2i}!&z<|}4e;{Lxi%xw0BN+>~Up0S{jO{;or1EDnN*yQM@wp6dpYq4s
zU&@%b_=vS%$yhYx;&W%nIN-`|eZG^i#-Z20m@Q-BlzE5e%Gmyaf38&Grf&Tq3+4RV
zMeP<XmN9SiD<}OfW6_QO&)%89`%wM=f9|z(tyedD;$w{>R7j=ykR+cd5~b4ZR{BI6
z5mC7%TZrnTkmXinDP;*E&PT4DYu}^mTC-pK?tgB~@BcL!pIOeCnKPf4$7Alx%$((Y
zp6AT@oH^TsTW_VBoI7*$cB;Plr4Q_&8t&Gt`fjR;_x9+#kE(QN>YM{qQpW7cc7&?G
zb)iPbsHPWgFz6)J<dyGjI78LfXKaZJRKxyFNtdZ6o{bwHoAn{VDYcK}pqkRQZH-)1
z{bj%HmX~V!)@AeJsV0Xm+;j_7U*jrG3R4Z|XdNm>HF56eTS`!sIuoCdYD(3kZOT&h
z@AFNpKs7z3>G4Wblb`*hb``3=gsI)DQ4KHMx!@kEi9HJDtwmLO@R4TsQ%yP1X-FNa
z{)s<s_ft)Obz`~uRFg}_wR@DRZ{vNF8&M4peEoD|s)-GU)oDsqidoj9In|Vz=a;mk
z>hDmwz;jg7?|7j_YpTgR2Mm9Ss&Dw*j8~|Jn;)qdq?(w|_jWs~(!wXFy+t+UqmR$G
zr|PdYrEVvx=|?hpccGd*?v}J}RDCbiFW8-GxLBuFpHNL)GbW`cRjJ>GJ$<tD>Oa%Z
zu~q-52mY-GEaw9_a;Z;|FYf%Eb9=tH-9ERpkD`T>*U~9UdPK1k?<wwhPeFNc=kLAK
z@3p+j1+Q?UrXS@hK-2H9g1$Got=zc99=F(Gx?eI%%QN~_(XYy+U)5vJkMQ-G#yE>z
z>9Q*x-s4nWNr!s8&GB?yHNY+HN5^?!TovQ0bX*YQs@BF;Qy#vlN`K1Hf1%d91F5F}
z7M%Gd)#NvaULH)<S7GTRL#c+hpX)cAYT`F{t{FvDdh+?AKTu7H8_@bks{VO%MvteO
z{{G?pKT}P<yHxc_RDB1Y=scBb_=n^<)2SvtKRMehs#2kh$L3H?S&=X(Ox54(;SKYt
zra#oN#3HK6r$;9(rRtlse%uPG;b6`qtEeWHu2o|lRcXs>-8NE9`Etm-EmZx1B{yxS
znx5@k;~i9!|F|Qxi>j~Fb6fUO4Oi*!J3uvY&mV0LQ<X*?`Z<$oO3RYRPf+z2c%t?h
zs_BcncRx=x`Qu3oE>ZQ}zavj<EcJiv=4LslCjL}!NE}t^m3O!2p_)=+bh&t{{`KqH
zCQwZukYjQos>uy%o-Rt&cezcS5>&%8zUfhlYT`SKmz1F@RXQ7AAxp3R<6W=Y6%X(B
zh5C2dws|l%<&$0f%$5hU{xF~F#`CG{;%YV?v;HFGJFjq~re7KJ^DsY;wo{y*pNDp2
zbhIPwYM)@9Gv+x*<2+}t)}v_WVZI91VbFG(;Oj8FT0fz`(5CflSAPh2_jjV5iuHI{
zk4MKHIIYLS?*QKSv96i6=X9=X#{18;??2JTBihodcOL?;aPdkyc#p5Ujg#B#!TuoF
z9|Ze@(0Nn__XiREeZ=kcklXEZ%RT{C_ffK&v*q}%27vxN?5~CWwdg!xjs3MOKWBoj
z|8@I%N#x75{W91eP_X?0@%xMYEtN`)D%xRJ+F=S3@IbrNNW0W>xw#qu%3XAnJImz-
z?Rrn|hl=un_Y2H>jXdua&k3FrJSW=EiSW<I)da9=7YolrW4|fe_A|h`VXPZ=a@{cF
z`T*bMG51{_)T7>LceyPe?Xt(Nc=WC(a+^J7=My@W4yWQ}_Pw^N31HRE_3H1M<N1ty
zii&Yl$IF?=_bGKdU$@)mm3gc9{wKcw`M=-)Wc}Q+4$m#^DzBu=u5{QHPqe-7@m}|5
zbzOvLzb|epr)K*B)aAmf=`uU+A}SA2?N0nUu}L7yc#NO5nn5w7+`2hs7FGYcYxj%H
zLX`C{+WMicq&oW2ZFHTD*Eo4CJ=p*D+WIW{k#_H<(oFVU`gZ0bs^RI&FE6E<_%^;r
z$K-o-G0S>1cXB>JzYTl(WW|r~(L}}fX!O2cvwCka+rGzoNWs=ab|yYwRy|VnXq&QB
z{rh|qMc7B&YwZ`Wq(_u=h+?O&{N;J|6=SJJ+xw}}_q1s1H%)IEZoQS3i{#vyqqkG_
z%`bgm2i0)5X4Q97O}w{9=Y3SALsKz-H|uw9=J_Qpvwr6~R4(uw)$}`FXwjN#^3DOn
zU!v+8J~!ik$K?|9`QC0vuS*M`oc0#gl#lTJNH={yGOK83{X@MN9qWhDkFj&XJ+vN5
z>`^eUo_>rgPY3_!!>)Mn9&GG81;`g~tXI$)xAIy)FUsqc@(K>ue{Q3NOSGiZxBt`Q
zt!ZNQ?th<NXZ(;3)P3@)&D%Rs#rr<q_Z!jnaCqSBryFOriyIEB)0C=&@wvz5P!1_8
z5(b5-`g=XRVLsLLhdP#6L^b*J=%l4oeUsLYTR}A(%z0!L)x^@ZYOJFwZF#NRMye@a
z4w>hs_x<z%IVKm1rS><ddAcZ7-{m%SN>B~Y_@+lGs)_F`UQ&jtROxJdg)F`L5B&pT
z_Yd&OM_%EkAGYcHfHuwRvx}2m@o?)$xS9aeH>hul)3T8A;q=b8QRTgU2G32;>UwyK
z%1Knaqjml{^6UEjH@x59!TtK&MdRWYd);D3O1%fi*P_NK^E=&YSNhZuH6EhId3Bw>
z+qt{lKChOG#NC-0L$lzIEK_L&6}{g%4rew<n?&6wPUu|lw=DlJkGtVIHH`O_8*@t*
zue?tmD4fL$`@tL94?YW7)<5)5T)%(f3LWrCS$=&!>Tx%(@$y=Fur3noB1KrwgL&5G
z#$BVez7js&R}<h^JIv|$J+jZE@a52^@ng)>^7cHfXnhW%HQl1D%Z&Q`xeCCsb_S#0
z`KZswg4Z9{GdZ3b>KIBIT;a*AiJp@5VJO(B$-BKnNx=$%m_TeGrgx}WC=i?VHy9sG
z2o?%RJww4FS?3k0gP6vt#d?K;E&A3=Ev94(l+HN*-+urC0w7?G!2do^S86J2;lVcu
zfPgy*;8}NPzQ8IFu$uttf{4U{C?X;_009utK>&4$4oKHH0kLav;KWV>-YtCIO_`nL
z!<%uTY;eritkEDk_7fY(_P=94y+bdADh6K;zWO1R*Zw;Wl#LDwz126%P5WSn;JdG8
z1=6WWr(RjZMlT1uX8G#WB=|u<3VujOjDjBpV*b7N>c0+p1V0V-4E754rh`v?>Cvyx
zyPcAP{ew9xeH_a2nJbvL;td9$O7*Q*;5AA5{n1#>V`0}=35T~Ctmqjl-RLO%RjlOp
zW6H0wl2?w>IkHJ!Iqol&O=5YRsFqD~+j07lY?9r_<1c2D>^PRucvv5M*JYD%{FlM0
zO4%hGR~S6{M0QE<v9V8f$<0TgZ+6M4<D)y`Bo;@j2jV2F$H>NU63-*Yh&Ty{dl|g*
zbDU(yaVmc<$*!a6Yq=z&$CLrNB;H5*cey0Hj`eA|Br8XUZMh^C$FB!-NjSj3&1GOR
zEu%({Hl=e*Za(Tflv}dnDDYZtN$Y6yIb|fsF=9||$>_0XT5d_}c(iIBN$)X?y2Swo
zZY~2mAIu}+00UNKpk1>(5)LqMa~a6^AdkfJsF*XagaZuRTm~x9GU~;nN`<_V8;+Zs
z<&~T`rZ>+kx#d_fGOxt@NL-Rva@$e(P+p0}QSM}32?rQhDFcHq<ds++zZJ<RiFVX2
zkxw#ngkQ`j-RKDXkWa$lE(SX?@=0DfrhJrN!oiQh<&^wVv>$Dz<(J%W{4qVhgyZ@Q
zIxWpF-RO96U4AJVkI(n!m%MVk_fLMw>T#iN0SSk@7~J`M0SSkD7;Jl?fW-6IOT*W5
z9B*Af!r@&8jgksTq96X33rJo+4$mtfu{e&>aEu<q9=lm$ag2DVpv3!V@_a$b>@mDo
zL5b(Fd3Hg`ZN~^-VadwT^n=1uv>gp56qZCf(xw%bSROwQEh6C%lR@X-ib#5nI<tyM
zb{!wiFCy_ia_ui789i1%P*h@ZT>Px4gyRZ>t|N;|b{%U^7nR&_yd@Qrcpa@v6qBOq
z=zm{v$*H67o5dx!9Va^!m+Uwm>sDOC0S0a^1M5c>mvD&5p!B%nk{!p`bHydzN7ur)
zO018uqi>aX9EHZ*D&YVFH<W>2Cfq9Fa1Vn!XWc3pJwDrbtK^lVS+)`q4)zQ#-dRF&
z`q<x~gyfdvm!>5oyN_|rN=P`sz|CbKZ=RBp*N<~@NlEYV+T@aw+m5&Pm6D7eZ{J*6
z;&~KKC@p#Y_`5@CiN~=mqqM~HxW8Z-$>@=%LRpE|k(Y+;#p7({vXZFBGabuHZaLcJ
zEGO|e-Y8X0vU(g`QeI+p6lz*Q67?vx>~@Le@#eFYB^>M-JbkIMB;xVU6L(2&I?gBG
zCAsnVeB51<9Y^PjcS$(FfOQ$D_+eGaspD}PH_t;#t}0nMGW%AQ+<qLYUroZ{EQ2>m
zHOZ-C)9PxHmyY&%s!L8DZT8$P@jj~NzDE+_xcVU<DG>9&AM&}ucXUSet>;Vo@0W#k
zpS!R@o}57<>3@B)y;7&-a|breTqTR9(_ZL%_?9tSiKu#4&*#VGoAnP7RqJi)Jn!YB
zyNT%f)$8{!U;X4R!duxswbnx$et!M>b7$7dqG;1TE0SKB^b4`7k;vtL?#aKqG!GN2
zTC_{KnRhhaBt!Lah?Fl{_wQ+PN0l<4pOr<`zdk>6v|-Xg*{!N@kJL_4-~Xrb?LR$6
zh7n%+V(~gpRBC%dR_t?1r~9WBO!|2$al7*3Q?tmK$u*NU%WhF#=~Tk~Xw?IK<C?A@
zcBvAt+qK1=Uu~ILA%obhwRhZ~XV;S(iKyByY2%mG&mJVA>V%k|`qjyQm>AXKD@J7`
z_CG9(s$*OJcHioOJ7l-3Xnf7qb5r8-FWC0#Zla}T!&Uq>E?4clVk`F((Y5C7C2O=S
zo*^66)pr;Fyu-5XveWv{{H4z|?|DqNOO^aLtE_(y_1d9skF!44YX{ufLm=(-$8)tl
zeMuHY^Y`sqe|D}@vM9Q$(b{G|{C!*&O<zoykgr(&eKKq1RjMmjY*J@Y$<wmk$}2sQ
z!i~0m=SV!%_v`2AP_pKC3uU`h9lzVo&Gy$@O-HT%WXHstcM!8`HGVDsi48u?bN~K{
zvQgCv&*=4bt&<L=9Xu#|O<SEvd@=1@rYxHFxmaV*mYBn`DEejgp{2S#yIb~(HYT5+
z_;cd{SroOqoxwW3?lTKCtkM0X?35a<FZH@Vy!ueDkJA5^71irc?QZtz8MPaaKP`Jj
z^MAZI_NR&e$VPSCA5ZVTy>~h>svkWcyY2AKHL}}OozHGNZ{Ik*{+7C9WYM%ma{lZ`
zU)@BEs@wRx-Jb9E&7WE2@rr)8YtbKLziyJNoZl<@LE|2|3KuQy7e(J0cFAAxr80ie
zw8W{#bI&X;<L9NTA1r&z&7T(a^VU~-J^Id|k_G*u>Zr-m{&NKi`Q5HJKahQMtzFv*
zFFpU|n_vI*)<Gh=maRQ{am%f{h$uR@`NGj()!$E`&T{BjD&xZWuA7M{>b80PyxNmJ
zjwH;U^y5*o?fB3Uqn<qgwTXi$^`}?t<n_CbaBtjha@+4NuX<D~+)hi|tX=THPQpsH
z;!&;-+BovhapO{nS9R@$=l7PH^%oIMMQQiwwbN)nuG8*UkKHlnpRRj|sG8V0Fzf9c
zdx_}kmiB~FUpl33;b7JdyVu?7&8O1qVoSvA+AWK!dgqbo<)L=RdLFJ_;i=QX`aJ!Q
z*UJ1><-9DKvgW`#yiwDR{nWeaKO~Z><4vycNX_AERuNIuuJ)*!6Yq0n{)3kCivQs+
zL*E>&u#9*`?V3NzJD#T&W>$H*>P}e{?eO%dADZ;vDvPRaU)Qa+FV~`_l}mO`mF-qu
z_0Wk@<LBN{;wZ7IX6H?t^_P9vs+EC-JBg@jc3lFma^O_BMRz|l_KTF`gqOZE{PcIN
z-Z?=;)9S6CoK&yDG2#_H`$*4UdiFa;L{nbnQXQ^S&!5$On7r=aYQAJPm^S_7z`X?O
zP7Y4@+w+cJU4BKMpT)|4tL<H<{F~h$LtPGVek}pL`x}|<56=9yTKb(Uw-LQ+v|bpw
zplH=`dsh>^T6N3D)eoQjn~18s_PvV~?zN>R`p!ge$Cw=tH0#f)5~9_gEz0#rx86Nx
zeV^_!FLf$koQ^lrIF(!Pd*wazk5>-f{@aB$L=<()d?KUm6t|XRy><mQK05RGfulrJ
zEq-_1firp>BBCj)d8~GYqr4B=^}E7r-w$5)pWF0gxa55sWwYwF{pGFh@9!4>Imh?9
zaN_QvGArdhpDs51+VKxdt&qL0toA9_TOUMPpJH^M1*7B1-3ynR+O7CL!b{DL6SMNC
zd<Q@@UD2oP134T2DT}IKH2UDN!)3xUE7kiv>2=q4d-;XaHXb8pRc}0syT|k2Jo2$S
zGKo=D{5LGKu=naNr)9Ou!EW2%zU`RIO7q|HbnVO4=>9<(s@MZ{hQR$#4`}n}?xV6(
zs+DTH=acD;Z$bW@U47kF61!E4lluBI@9axnIq?W#t?GD;o=0kz;Qo!*)!N0!Xud{r
zfARaTiY>3Oo`|XopR2U}p5Z$PD~-e}(*07ypIiRHh#4|3)ry~0``M{*Mm)32UpI70
z7FCDd9oKYExf8Nzx_iXoc4rG6mA$5Q-`!9!-`PDfD_v5nN&9|-GYBhHp5tb5{c_!1
zC&ABKn~ocFXaD$Ie%`92>;9E>H|=h|Rko^Qa}OW)<yS{!yH!aC6a+v31QY_nPYdLJ
zy4K&stE!}vm42NuC~4Byn+b2VYHz3Jbr<J)`4q8R-#Y#6@n@zTCtg!l^PZJ(ocex%
z*>PA_{*-dP_3M6jKRfq~EQ&rjEF<rxH!jI`slrLEZW+HEe4@kfR^R<4>s56+Sf9^6
zQtr%yjStA8>P;_?Y@RsmxGbtJdib4U+4@|NMb$nzmfU{eg>y3084lA|Z++}o?US;o
zYS(w9Sl63g3D4<u5s|)cuIz&hy|N#)Qykj9ck7hP(lKIGz4Dy6H62~Py7!qa@yBJO
zYBoI!KMu#VIk{A}OU>HpR)4*IPnQ>TO_!nIaJ_QCYF@BY;S_mq%L4_seL|qlaOn4N
zu>wt&EtcJ;M&}v3Wxl@J4!ib~<Q<QCe>l<3XAT>mmNWl)SyWYjH`8o;#ppUmr{4GU
z&fAPucSiI77V^Q2n$<IjsA_cH>y~U`U-;&A#BHi~-x9aEGrC?#?cQj4<?Y_7d8@qB
z#VUMdy{%o!%5GNS7_D#gx@Toyq;S;jP*KmFYJaebcS6ih{p#dDOhi|``#HGHo%(l3
zv~#EC)$@v%)%Q3f)qlM2tKY|ro_qDXZbs9^JD=<tci<Hd@AYBXjvt%aX<QyZZ{?lO
zX2aLpAH=(VjaT?~wcmNgv*@X>hnMxAB&^hIIWg<c>N+B`@f!8FIOU^HA1ix=uvVjX
zxXpe0#_9F9)Ey)9Qn!6qS5$je-(CFk4$HR7yp?zP<kgN4JzOpBynZ*0hU?V#T(!9H
zz8(l1XtDLBoPMLq+b)EEeFA3pbB`8(*QbW7Y67hGy>nYQV_W`q-|B%o<Y-ma-{(}9
z8|B>SK80|fKl{;FH<3sj6uzRU-t}y(+=;@)teuh8MHs!W*N%@?J<vC<=?da@tyA*3
z0~=<plHIPNe2>?to%uiB8~f8lx?dy|98eA-J_ma1U$wjD)0a)C*LjO9s;bi?%DFPX
zL(5)P`x{3JM=O5ix~SjTP}_aE-Ag0W&z&H=RN-Y+w;dlkV$`z-h^T6`-RM>KtmcCm
z4acePZz}OxE&q1=SL$I!&5ld+vRkz8f-CmA?fdM!@(1BwA@JnIvg4aPcbxFny%`TD
z#E-Zrv(oM}3pA|J{iH0aev!BGOZ}#0%4RjxF_biT$?%web1_&UP(5&WAf|VyStt;j
z^*0zFtP-r&==bNJ&e!DM^SZ%=V4;B2GZZY6bzU(L6NqV?TC7(nShsIIU)ujX4k^D^
zYmkke``f?NKPT1nN?)$YMKw9&&!YLL`o2Bhx&YPiQ{~3oLRI-^)L&xWrw<g)^1PsD
z_2N`gKAhh9Hmd%bd*}G5rXMbny&To#u?-vDPSuzA{-8UkhKo+vP=#vZ>MbRzQ<Xl;
z^YXn^QtCZ8z7|#g`8ST-Pc?n&h#C)3O@1S-TV1NY@|Wk;ryAZ?HLfAm#IF;Z1gJ_+
zd=Y9wCFQ0+w=|{d4<GYAO*Oqs+1H+-nq0m4#OJ8`_J4Z3HP!I<(`qMCO?+-o_gAS(
z1&b^UQb}3<Xx=xd`g?Y1_BPe@I^&0QpqhMY^Y%_ueZS@||3205YY(*hfNG-ejmh1q
zN}GqD{)9@(7c1)YqUvvSsYhR`>9JLp_NSUW`^5r-sQNk%YVj4-aOHW!zoweFJ2PV_
zRcU0IN+YPGw0P?66srFCPpADrHGR=<=YOP{oV>g41ggH;h5JmT8qRExHi>HDgw6$j
zqbj{TuGI`GDYtG;nMKvVF4x{URMY$4Uu7QE<VV`Qn@ZJp>D!r$sD`I6zr2)c;@cM=
z`M2CuysQ6edY+Qrdd)hj{-FblZlaprG~9YC)#TinqqkG_%`bgm2i0)5X4Q97O}w{9
z=Y3SALsRD*ppr6XSGFTm{jCc%Iz~0UaDzc7sV1*{Z^Id?zCL41T%a2EZ%VpMHSuiR
z_}HxSKc)7O98^=<wylwis=w^F-SSdR-@0sGJk{jTg_~}n>T6u3Nnxtt9IZpes3y++
zd`k(cQfK1xQBA3Ov`ty6{(ZiQ6{x1CG(BF4YVxz6)UHC+moT+^HLBsII~UwTHL*v*
zytSxG4?fcDeyS-aIt{5q)j#pa?S881uWl??pK5Z+xOR_H^=-UwawDqYfv=x#Of|9L
zusThtN-@iNG^d&}^Zb&QRQ(+)7kG|p`W-K{XiYVF=YZiaQS}X<oAC<OaPuP-gH#jq
z`QC0vRa*Guw700HeDv}8_Ei0~rqu03HT_6N?=Do6$K8_FjjHdZ`USgF4HxUw>JzGo
zYsRGXq$>5>u%}O!Uj1kKIkxH_^}xUNfaQDuM=td#^2MFMb8gQUx7+8I_EEHO@>)7Y
zNslOY;yuM3?<puR?)<%X`n{G{x!@IU)byiV1!(&HRnYe)x0M^W*y9#EO!rGhX?aG!
zD*9EK^s9R8`4PT8(->#5D_wTQ!+V^{E9p>=w>h4!s|L8G{pdIkjH_Z?m5vKyT-DmR
zYRbbmRq0PT`Y+UacOcdD--0v0q?-KZ(946V`YJ4aWGL0}_H+HR=5r)|bLW~-RHY}M
zFZu)3l(+${f28W4H)r&Cs_E|^-v2Yz<hx5%pG4Jn;EB#tsfK?@o->_l;`5WU&7vw5
z%6M!J)sz(pgThq(y&m2$pKAI;9ZM{tntXb6(o(9vN$bb0pc)S5JhF;vV(D5n)=`zV
zyw+_a)s!!X%-cfMA6RnJcB<*w&NbdaHTjP_Lc6H?Iz6{#FV%3B{=NfL6Zibl<}g)h
z)S;g<siw3ndHe)be}N}zpP`z*xO?~WRFglRwBQm|-~BuC#Kuzp$8K(xgKFYW^@hYz
zm0o#wdmgGOB}SKvr|Mt7u5AL<^Z_|07owWnpyuhKRDGA*)G0wVJmZ@lrKl#pvv^4v
zs#2x1@fEW4>ObD~x?S<`ZeOT>mu;H|V^coa#m{VcFzXNVsct-<$}X;E<1y<mQoi#F
zH){HoF+UIU^JqK8>G^qRM@C0G(ysOi<~d`Yb2QF#_G&$fb{^)dU>yc+rwP6e!>jcZ
z`U`DZ&vx~PfOmf<+NoHNhxK@L+=0`2Jp2yeeIM(ZX?srRx@NrpT>JhLZ9Jkay?XZ{
z@Cp~Nq=WbPy4yIp%^vIzg8f0TKM0*iWpIBG(ceehUJtq5KDX=>V09lQyE$8q?`i<(
z&%^#&*k6mz1J>AI%kpz3==xu`ua`u=T-z^${Q(8r9}vI4*xypA#HgYjcBLJrAOR1w
zOO3QkEti|C0ifJPN4c|HUeK=h^nR!)A9%mOyw}L{Uh$mZIl*(H{hSE@d|XWct9G&Q
zJT&&3vTZ*DtQ*F<VJFuOGp-NtT^@7a<v~5_jdqvY^3g7P?21S4dLp;kV|G5FQ|WLj
zUS{8GyP5!2?Od<^t~s91$fu|nH+8(6d3>KzxAS$oeO{Tjitm5o`=9^&{ZH1<9qaJi
z(ysDKy6j4aUGYTQ`yTIge^%E;i1z#9wsLB=A3$9$yqYev<1V7|5Y_I)uM?XDvW&<0
zS*sZoL&~k2Q)W^1ue)}?$Sg!z|DvrQ>Po7kFWpAh*?5hU*V2RiZ?CP-k{@aJZYs@W
z-=%M7E}|NqzWnl1s)=vodvr{`M;EiKM{_6V1N7Ulr%zV=_#RDEe2+%&`!%cg7PIYp
ztcMhAJ!EI%^JUc|RgbnQOVz*6H&KLr#J$#j@k)9`NrxzQ`pRFPS6?xfYP7wd8huZT
zwtmy}rs39GX}L(wojH0tRp0#52X;^mcWYLCH`T;@dvxAMRXQ{k^LMj;=VqQ?(lYCJ
zu0!Pl&rwak<AoNjsV46nF#IK|zTtB-{&!q1F`w`4cJ#Wm@X2X!QBC;>-;Z?D_an24
zcGf@Ci_x)u82uPK7u-YZp~M~q^Xloxxbk%He?IJr2k*hgzEgmF@y2=ut#K=__4A^<
zUMa8OaQ){tTDU|@I(_>;J>HroR`34z>2<~r=|J5lpW3{=6IHzL<9)voZ4ZYBzJ9uK
zR=c?2usThtN*JGeY!2m+vLaznn5w_m!yD#PO@F9kiA7YCPmfMoO4T=M{kRoW!@-<K
zR#8nXU8}}As?wI%x^1MI^5u|uZhGHOACO~mp;&5vgPNy{QuSSKQ>O&g@QiPIl%ksW
z&f+Cys7jU2##hMFtN+kHAa?%%uYBYcZu()Hz7J^Aygs`)*%c4BeuS$DKz)PyrZ_DN
zDIZSnd>d8X>u2!X^sKIjx2T*%wL4nppCiAn-+#mV{T<w|&s{VwZn4)bcBIsMaC|Ll
zj55E|t#+kP9Z};UYMfWs>ARh~+wJpexk%idnK3jA{>U<wMo`iFo#Sw3gS1K1ed2`9
z1%J!(|MIvSu2aK!U%4^2Wbw-T^nt=zys#g<q5a^qkY)Wt|HSqCC$7)|pOod-=c68X
z^BOO&r3dRGu`W`C^*oqoZEoB(TI(y}(|t7oj<v&_j^88uJPKb9Z5ltuJS}g}(~8#T
zAX?Kc%DT*`&!4LR9BXGV`kjyZd@OkVaXpjcsiBUcqz^;EMor%B9ZCvT2*d<p12Mfr
z%|e0LtiQqdU_!7^K<XI^7RfrVNFBs9PA%3e6dcvJo-a+w4JZhJ00@8p2!H?xNK(!v
zgS*Vly^pw6kw-TskT#-3V*EXS5x47&Q-Ql6ASQtw4;6T-cE`VEs8}3w%2irEyD8Q$
zsy-ilbLE$fPRMT8b$tW3J(89wi=t+0p-9ia*>DwqA0_4)ljnKKAFVnswa;g<J=V)k
zYw};q^8G$#mu#0R`EOQtH5mAD`odE(E7fa9{lT%Hep7al%uDs!0e8j-6iJ@ctYO#V
z1nMph<?Ec7zO%y_0=>e)SgD3<5C8!XU{AnoJCKznqgt=Y{`IMo#A`b7*21q>KXQ&h
z`Et1TT&-KXx4THZqO&(WyJGp?%S03vy*=mjdjhL^R2@!%UYq}VHrpYhsQNz6?IGZM
z>ei7De-|chQ+rZxJ{Dg6F1*7%ou^sBvI#qhQ8iC0I(-iM?(>v$7mpIRtJm6pUTJ5!
zJsrH;e|G&IIo-aBmiT$sgI;+*kCyU)bh>N&Bii$WcPmwS&xKz5-p`AN(f1p-yLbEi
zQ`qOsJYlz$Ti*4k)8X6oUV?bs*)NHFLB6=Lzuax*(&>5{@$(_z)^%)p?b7?cVU`~6
z@f%*@n@z7?f4uvPd4;bxeNMUCy)>=pz3;?^BdJsNv5MEO-#1qF>J8^;wza#vk@K>a
z>J7)JJEzxqAbxkOw?RILl@Gs`9sJFvx@(BtYP5dQ>z-A;t~Wlo0|7e;xUGLSTJ6#M
ze&{xLdf!*Ay0i0{KwKaI0?|p}`>k{2e0Rr$&1$7te`mJ(C0iIO7Ki98?d(%6%04f!
zlh3|9>`RnbA(~we>=u80hsJk!>ZgBXyVR-NHT(PS7Jt0Plaz3s$``~70w7?VfZhG=
zh|k(Mhv@76MYG@d6J^c^uXSYRtcmOk^WvSE7Z2{>Ze`#0$WJz0*g*jONPPN{?B;4m
zSrIXgNc&nFz5jao$@kp#@L5?)MP%e9o2$i*_~waR!(^vai{+}ki5V-t%;|TVYL#Q&
z*R?pQuk-pofI1$Wg8&HFPk`5Y8zo%(Q-l~n00ejuuxnh`u6PtqtoqZOnqRA=0}28l
zpq~J%^*MHhqd#SE2?8KsmVniDL}uglia%E0i?b^nj2B#6@`ko}YQy0l4$FG=THox?
zoa#?z6|d4CBC4`_KXNLZeq(3ck+=7LBC7J57ld&2eaFDA{e#)poCpyFKmY`sC1AE4
zW!7rm<LBz|&D!Z!e`puzYZpW-C%f81i09hxXgD>#6fN-@t&iQZ?pAGw+232Y`BSG$
zv~#`vXKr<;9@o-(K3K2k(uxPyLBP!fjLv6=`vOnZ?)bONi-ViX3TzXHfL-G@yw=6r
zMcS_MTEsK(<Mf56WGxPMk7pykYu7Pi-*K__9oL&*cGaU=@z^ylnpHf!+I3dL)#8im
zAmAnfyw+dx4%h1Q={5g$mnp0Lq;}h@#K&s9M+pyleF9qFW5V?tf3Hf-$F6>X>*pe_
zIz@n2KaD8iI>r6obG2^m-tHnXs!rL9c%y^B=eIqPYi#ozexn*4bdXBc1l%@H!>m25
z+fS@HBRmiQ0c`>mYNoWhDaRR^m1@U?J0Ji8AOHd&00JNY0$wIy^?Scw@$b}r2~Mx~
zG25Tult1-43A5+Co@rL|dF%>@b-xg+@Ob49!o5O3Yd$OM>)_3nK&WFVY4D`sG5>}d
ztPqF|WDCUf4mAq}Vzd4R<AVvoLIJ5~C|D%xyka0G5YsrdSg%lURNs2OG`7YdG!Os*
z5C8!X00FTHJWxIT!aqCL61S?@bopEF`>21y!EOhM+m$~F!~g;y00R05eBzs({PUZc
zL{z<L>4*v6-f^6`UE4p?IPKl^V?-3yS9<JswdK2STgP<WL84Xr-Bqftf7wprwARlb
z?EJ#l{}8*>YDrs_yTHjiRvad*RIfdSC&kX2Jw8Zysa`wa&KQBEL!Y@bd$rXv)Lk5S
zJx9F5J$~Vb-G9|tKv=7>QVrK200JPuo`BhYH&&KB_Dw>Wy2&wquW7+wTISfCC$}HU
zm&5j?p{d(P7w~&UYn6EN#kBtMeo^%0K}om$btsQtRBiS1ovHHJynd&YRXwT>=VZTQ
zCC{usB`d1Fk8^tnWdGr=0Yj@VCvH=FQg1$%p0gk9f7htNL^O@x6w|J1pWVc$nkN;V
zK4bU%(=6qqJbt%p!BzJcczR5X-z)mSXA8HrZhux5MU~$_(e$2LF`p%$J}L9khd&+j
z<l+}L65i_8c{zIR%GRyivHDL;C2rRPb6)<tVvP-imwt7+^jm%Nt|!9l&f&fnzYUYG
zpx}UVVB2%RyOmnx&P8+c#r{K}_HZ!ze&crcF9ts9bF}780tJWvEeG#p##Nhm$A01#
z<z0_D9sX-0AO0(C(jEc@2b2Rm2dD?GUk^B4Pa}Rl1js8r8^5)AFA-Jswx7+?+j$~!
z%HPL#5z*9adiDDIbwJ6)FYZ4fi>7+h=ajpC%iB!5+;NK>Nu9EfRlK8bfBV6$_Z=p@
zm6g4E!->=%t~VT`?!NkXSE%G8dkGXAP!7a-4o(!Pbk9w{ZYFlC(fUQN`?Uve>z*-w
zoorS0#s_yGU?+hY<sW%;)~bCn)L9M%mc&)-_uf%iGz|@YJ)`0D12S*5^O-<gAOHf<
zNnk>$aXX|Ix&3DKl_`%@e5Q9czv$Ym+hg6fEtLGxqS0B}*{3@8$$~Y<r%slk(aC{*
zi4rTs->Ob|d($UrL{xnxbmxpeGiMOH)T!LPBz^l~vwt#WQMGH!-=BK9@==+WI+ZVo
z7X(1SIDxg&56ko3x<rP81IodFp9AAhRCM?dwVevP#eUB~{#nbv=%}6CpUOX56)ped
z?SINpu{pTEq`^WE;7_1Mo#Kn1-+YufrCKSpdp?=o_-^X?V3DTzGl<=)#Yuhr!l6kM
zyS{XUuvT?EI0pd`u%7_$^?mlI2r+^H2=F5C^U?ONm+SgB@ro*(I_&-Srh#2jWvg1Y
zW2yZsmMoO*RwW%!5C8%F1h($EkQw;;7!gJFrwlGZ00hhuF#BB&#h=-Dz2Z-_{WCt<
z`s|^Xnx+%G)GPUHS32Cj&v+#J{@7b6RC{NfUsP><W^S$K$1cgDYQKy1&X@9?mPOST
zW6tjE)aAS^suq%$<o;$;4!@{6D0`=^|1^p7^H#g+Q^W%TAOHe(5zxCYo!XsQ`*?p>
zLmj?ZJKgHf?)eCa&zX53X1RN1{0Hgq)_9)P{B&!547)(U%>-C~m&(m)g>4|9Md1CP
z63R4db&5dU!9gpZa2*7U5U^`J^Te<N&ko;vfUs60X|>|EOZ~JS{YD=mP-i)Wmn}JP
z@4%x33Jxd-?4AQFbt4PE56u3)Z1)m-h$uRI-S~YM=I$qMQx<uFFhBqVK){Uz%#Q0Q
z{+vov{cZiK*Ew)OwyI9qFG{>_`JOK9(D!{~QOZ@c*!lRzwxfUCxP>^aPQ3>>mCuDI
z#m<{OK1htJQ}!a>=pdl9n~Z97&_OC$6L6~DXB{u=ej(PJ5grJDfHr|n_fK!#efM0#
zO0{Fc9S{Hk5C8!X009sH0WTA1Uiha8t<skeXmoN|ll0s(joW9)(WbnfY0<t@w{@ED
z=Q|JVmffbj@(1BwA<*xxQg!{yb`stS-fRhkI);*h6$04;*#j}XL(M{g*sQ<7_+Ub?
zP(bP#3Kq#auNa64#57JV)+-bo)wiB6jjas`4Fo^{1V8`;KtOB)sSTg1oxge-ajS|=
zm%sJ-*AzSW>BQy4?aH47VgLaU00Dgjs*UVACZqT&BC5ttt=XV$wN=FJTH>GVi+}uk
zB@spSl^*+DZCkd#WBRWvNVKY3p40b9|6bqxwvyPTR!ds<lS$K$Y*|5Asa|{9bXeB)
zw&&9bFV$-Y+!-UV<fBTjw`sYYctr<&v-^=VjTaNqwdb9+_dFW2oQS4RmCyb557|}_
z)@rO&!!-zi00^)rV75MHWy#DxE*DrU|4F>2y{E4ldRM*q1j?5~i%H*=`K*2_@ru@|
zl6&ur1*t?7ee0gut)HJipNOj28Z4au*dJlyl(MQv)#3c|<WrN&9A7{bRo}<CJp^jJ
zF+AJ*tydDasXeJTA4~h~owuwQx0HycPPeDb6Bo;8=~7!yU&@h2+^$~fhf>mU`_q$e
zZ&`Uc5k-e|IhQ%_*%d@I{b|AMiw{gmBfPZr7t_<8dU6Tjtp($b-nwn=QX;DQH#K~@
z(of5X+m-kDWP_ZyFX_{36@h{S$^o7OqxAsq)}>qLe|ps}(b`jy?0sY0cN02&vYbG{
z;eX44+vY3ru1B2?-|g@7uuqJB7vvN2Nre7Ev*pz3dK&TbArPsZ>h}A3=jocI$9f)~
zUE!HcuU>x*hfZ%0OqfSRQ@!bP%3a0>-<9nZH=RUMr|e@DZ{s8P$boHXgtxM?S8q6x
z=I`qb$EdpzzpN@<Jb4*`f&<EdIM0Do<CI407rpMktXJ;O*Y8<Ltg7Dl;0^@rBp~X%
zQ?J>bGc&t?_IHbjXj<%{g@GF978BlT=QDx0KmY`ylfcHo1t00tcO@~ayyubG75<BT
z=VpH{$0`yn8l9z`eX8BIP6PJXyMBg!i4rSBtNjE{`#)Or=j2v#D~Vm|RPK%|{^7?<
zmoFlsD(`h&PUQ>Y1pyE+PQa`4q>(On@6Rks{uqCvqQi&V=R?6hG1uQB-}vTRbkt7n
zPo>#@V)(Px&(4~k`%4-u1Ofg8?3&N56^~X5&93v{?Jw8xhySSDJB`?_TAb9^jjl7-
z>s}oX&OrbK>?gpxezQMCh!F%pfENK?`=L1%uEL4cIw7<BPg?b-qyq{9AfTTB@Aa0f
z!`GiOxC8+ZFiXJd_u0(G>lJ@awdc&n%X&N@Qg~j;XS>po5i@V}eH)e$QI+@lHs0Ys
zlE2hfAFp0Qpw4jMy-zsr@I{>`t=Ro-wlBww8<t9Vt6lXe;sF5=00FxQm>u6XYoFcU
zM=@)sTm6mx>zCf|KeviN!Qp!4K<_*cxWo5KMEYKdTkB)k1p;m+z<M5$o6`#0KtPK?
zwUJ%NWE5XTpzh$Hl~1@10!9ej|9yvhR)4yJ@YYECWt+WkB&}B5#!ju-pl!8P1PTr)
z2Y3#=Ru5RI8(H|R+^=`bUz;r@qUbZvrH?#cX&G^wvd9aB0Rp@UjJd1pnPs2%C8Fso
z_1^sO?6}E9RIN7o$7XM}olRIN?{cCJ-|TyW;!ho(UFSEyFnLtF!}Ez%waZ>n<B67i
znR%zvu6{mI%b8d0KL6=6>A9a?Mx0is_Q`T8U(L=(amrrA8yy79)&tQ&2dQLD;Ei$L
zP3ZK=aw592o~Oy0Gr|J_5YQ&@WbRTeJ0D#^SgCeQxB~(p00JNY0w4eaAmC*Jc73N3
z@nm28<KHWZ+c<Q}Y+Es={VEb|%Ild<oqx~o&%OLR5k=o0wR=w90%77d<&{4O_X+{K
zzF!I6Yzc%qhLQ$P8Xog+sKE+>?13DCnBJjgp+Icb-(Y+&Ay_CN^$Z1zWSv(G!~|j*
zrxxoK3XbYq&zHv57=#7_AOHd&00JN&Hi3n^Z(rhnaxrnMicOcl^_K=0FIsluLgIGi
zPXaN300@A9J^~rP6{`8ep+!Vgz0@Py4>|fRB5v1=O%MCFS6@g(QGKPyepj^$EWT;t
z4~s~&YV#E%FK3oqOq|wthWCG?`5%jjU23(YU2cD{%D&)Y!b<hpb8Fo#!y4sWOn9kY
zJK)Y3f#RJ@o%{Zc1;i`5BY)rSHHXY6qO03~=e6t`(5~y|>CY}AtkqblhHDT20T5tM
zAY)<WpThC<J<_1EvgDoctQG^Mg~V&xf8e<JnIjeuC|?e5pSnD4Z=3nVGrDH=vY*Q=
zAa4tzPi5QHx>3&sL{#0ezt!OLPc0x$DXV%^9nOvm?Q<7<FO?{&zK?Tz2=MAZuo|vC
zsW%_Xtxsjf2P-ZjqG|KLr*(|4yNDQ7^Q5BF=YlGqj+nkFmAGANwtb+fzsY>!6-|}D
zO;~qsJ`qLTUOu$!EB<O+&z_AI5?=aN<wpwnSA+>~y)?67Z0pQaBC3k=UV3ZQ`Q@wZ
zqT?0N8*;es(xfj3e!Q4K!2#vKzUP2<D|NuLFTXOh)?xy+hl8l|j<oDH`o0GD=H7Ef
z<Hl;<39E2;*P~8{zwq=!54G;Fh(N&s<p9qC>VfOm15Ve|h@TGut@f!X*Y)<Rn5Fkt
zT#-9-^!<~Fre@Qt*Wa|xo11-l=Z8cz)tf%2++E!E-oYU?=95V3lzpt?z3<@A+d90s
zgz#2Y_Ua9%`Rv}Gt-C<iU1+G@aE!VWZQL66iaFl^`NlTi`hWOElXcw|61&xC{i4@B
zt9=pl#s_yGU?+iE_nyk!yLSPBI?F-Sewy(Wo31{;ZV}<Fc0Ln`3j{zQItgTae<|E^
z*<xZ=wce9OyB=-xK%%p>vrl#V&pMs1Iczb3MkfdMB}%Li^^V`T)t%kXty8)CRH5^0
zUmKfBL{;AJi%#VW;spT^FiwEiIwyq7cO4n{+QV1tiHbD7WAr_k*S>&S_FKI!>b?`O
zTkL&9kbl<l&+2mz|7-}Te?N@=UbwX%fR|ln`_<L{CKSH^`R!8{6IRNmG$0fZ5Ssw&
zeOir%r$x`T^Yl3HbFH;cl@=%U^#PAQexzHWC4{xA<H0!yfPno3c<)zZe~J(z2!H@D
z0vDIP*ty#+sl+R)aH`oL-)lK?FCbRcYx~$Le4rrUMFPCWTX==*#XLZoK)`JTtgb(F
zTYBua$LKhQ)9zXA52iO9yYoZz_-dpiPa5)Dn24%>x7hvE&%Q7bRhzxga@@A6^N6S_
zDtE>1w~F7DQs$?5L{vTXRcwo|vj0Iu)eXZdwfRz-LU^m_^)&1U0T2KIH375x=c&Wd
zd*0}=itVetI)m7ydhHU$9q(~LgfI5@LRilSL3n)T*?_AtuA<b>uoDE_L|{k$zTImM
znNL`$o6-hbKtL}6yZ1jpd^g@FMQ^^r9SGcr0Pp<?mGHIZ>s!5kqqMQS$7|ge!Zr52
zNLK6G5DwdQg5W7OPnLD1nXpEb`QW{{+>?nYdW-z$rdcn{A#PJM?BNFl>?W`$N6hqf
zlZY&u9&cA<?THbSiKzPeD^&w;OeBPr+Ffqc`$`x+x7%JNzE=}FOo{6^l~`3JJm~ca
zG+#0Da%M^TT_&g{4tBNYub+#!>J)*uPhFn2x6ORwv^w?O6{onHU03Cly@)qD2rS6e
zVrsY0LSj^-gAP*3n!u&8)d%=jE+E3|%!aY8GgHYWM#`Eq!UF*i&?aDYTvF{{J0{!#
z0T2KI5C8!X009v2G6BqMz45$Ouk#V@d#9}I<@HS8{L|y#w7cVXB8sNIyyQZD`A_0D
z<&{4O_X+{K_JIO#wgf^QLrK93fgFLHftcQ*W}!fA*56=!Fd<kdAoUCdi)5Ww48#Ot
z8mAWP6$*~(ThEuq&W2n~OPyUeeiGrOg;)PIb^P`fG87z84%j^h>}mjn1Ogxc0w4ea
zAOHd&00QC?h`;%x7IOWKM0Dl#KF&JaX1f=JetT;fajWtw$ra!IWO>P7FUYL*jhYP;
z#!tT_dtD>ti(Sv}7Zn@VB7LXHc4?$^<30$000=}IflCjT{?Y%)9@#4zZTW}vf&d7B
z00@8p2!McH1ZIEpc=a!X>BOkoWgp@J0T2KI5C8!X5Qo6w3aw^juezH+#pIxrv`6Qz
zT0Y*tl0d-$<p9rtTj~L&eu9F48woThUhI#>ukDnf;Na$R;Kpaij7>X!YqWWMCLyeJ
z(LKHVGfHnEqUw%O)557!7ZB0(r9$oQT@zYGMAr>n!e?goTtY<E%wD<n)c#<TESla<
ziWK_nP=*XO#=%UT3qK$L0wCZ<0yocFe`xdmGei{ad^T-DO7BxdQ8zM$P0>T3`xoDJ
z{k3?e9M|Z+Ij_Hb7zz$32Le3@Zm+2(3=ZVV-r^*I^5vkEgXWKPn*Y!2!-Tgg@j*cV
z1V8`;KmY{9A;5bdil3V$ygxAe4jFna2iE(FSPf5{5(`T~00ck)1VF(11lF|b7Wd~9
z8%TxNreiKWbo&4F%{7}d20wd_R4uVFRP@!e|EJ1$lA?k8TK^m#v4<#{_mk&ZKRkXn
zQ8dT?KA+!Pcq>tBq;}zorpv1R+HwU^>jMwnlCVC{_e8JOekJC@ypii=QB@f?QMAO6
zGL5^FJu=i@4n}K&_ljq4B!9I*hJr&hlmnyBhu7UFEiX2!QmySW)EN%HPMJ40wP~vC
z7R@i0NlR;)DeG0`_a1s>-iX~Cyk0*ceINh={0Z2#kFnl(_@~(0F)SSYa>LoZ_7PFl
z+qr|(fdB}A00@8p2!Mb#f$pE){r%j8(=sd7jtO@_00ck)1VkqA!Iq@4<EzY(MN^TP
z!*=fy_^MEgKO2U3l4w)!<|tAI0w4eaAOHd&00JNY0w7?NK>ChSmt&Ugmi4O9@ZlZ=
zKmY_l00ck)1VBJc0&ZE?0z0l<XC!86g4G}Z0w4eayb0`&E3_!@*&VXi)OTiex8L$E
zm7(C^<#O=t6ZI}HnK4-wRWD4uG_n4Kbuus2TYuj7{?`S^{`EJ3x|4(6e1tm?009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009t)HUhV_uhO&Q7yD(XH4aL-
zY5l;nCCTlhWwZL`@%*3M`|1{1RK06h*-rV^WXg)pd35yR%JIi!Ev-1F_ABGI9hRZq
z;GjGQ(WbLMd@Fg<GY1JPRlZZES9#A@nw$5FE<YdNMikZBXZ)i@?e4zkl><btKK<fT
z9hWW3AiOo&%A;4)d$>ZY8QH7uCRR26=8sy)^*54e*HFh$(%?zMWBx5K!3u$#ftvy`
zy+h4Hf!M6S!T4Z8uuwqi844E3I<FXr3B)u`E!Ha(9M!j;FYUj@EN*|yTU%~RC6V;M
zKG|NWK4SNKZHw%cwe;7x2R>+(zF2mfYS}Bw^*#?&DOi2RL0MG2^}*J+zj*$*EV>?f
zW7ei_cV^1GmF-h&JhTpb@9%i*?o&imjZ{zSy>FZxe1HHy0&e?VRokPayg|`#%E!Na
zoaoiwPq%z`>7Lz$w<`8Ir491LzE^MADcSAn6d$j}yXDe|%|CSBA={<A`LRmJlAfR3
zJaf%?Vz+KTKC-|Uy$%qlSRAHQAJo2Ull3H$>TTEA?XKF4_zl?}O($;GNPL~Xug*8K
z^<AWVFdMGoZ*QfH625$Ke$n;Z4-GfHmLtY*RCCVzxy!6yH_A?{x_!#~T*-g4I=t(y
z`=s1o$*feb9lZOM^x6^0-K~!u-*D$wM~U4U$u6hvd#xSAs@yvrPNd(LXuHqr`QY6T
zs}(-4^h64m_wQ+>_}zYA-9F@f?sgt#+v`@-Wp!N5YW!OM&9?uv!ZUi^Z2WHb$NM=$
z_*cgNoo?5#j$gEVb^CJ|$v)QkXEZ#c^`&0-YCH7$Q@b<!d+D@4b^5KItGVbsXEi>+
zIvl<BJLS%5d&w#Pk?I?*`;j;g>Nl+Wvv0=wt!kurwC?ljU($*%(sifn7rpWG9>2Fc
ze9`)8L~BpkWv|}yYt@}-<wb2T>)%~o;i>Irea^d|&93ly@#ht;U0e{)jmKw2o2Sit
z{L*Onk=k=s?;EuT?!{g|5pDS=TK(e{d$q<pow{zdo;2#8b-N}~{AT;<L}|a+^?t`X
z9#Qyu)y`=9y+e9$+&+s`kMh1R`g3OXcgCzgz4;;f{6)RLJ9>Gj-I?X@v_Ew^Sf4wc
zKC|(&=D<3<5!1K-xjyzRiKOay^*#q~ai=vd$@;pQ6Yuj#_VbE=#Q4W<$<^+H40RWW
z4qL{gmjCvwEQ%UkXXG~bOS&9->EQjBWl^-ky9X--)@SqEt-R_XR^^=acw(gRtd2X2
z>YrCR@M^eD&GWJvFYA6)glBGF39I?2*S}NGU!;0A+V5M<7q`5>!H(<i2N~)8Pn7#w
z{o(5JYjs>w^nP8%eo>zXv*l4;4xBnS+YiXvpV{=Ywgcg@-M@o3PqXF5EB-{SHx;|>
ze(%?t?|R=8^tu!MeNS&X&AQW@FOl3?eLsp6KI{9e%Rle>RtaD9@~qgev_C~rv+YgR
z{*?M#uPWt6R28LsXxYiSo>9V6o+DnWl^(ZVSKj+s|4wLy7ipe@+51}YS-ozxoUp3L
z^wwLv-I;AK!QYL);}VIF)qXd#{_S38VfA@a{9AqQ75|a)#qIYM9!@FRPd8ro$*Wz!
z`gcHi9$2Z>{XUG^VfH<H`K+Sv9DM(~h~EKY#an5oYN^q5!~L}ctnO!`lv}IeDE^I>
zn`=|^|If^pZ{GflrjM2T|4-(||Ks)hbmLef-{oZ&!eu*N6gel2xlrE6p;y+YMd}ZT
z^u4@D>z$N(*sN;h?mZ)`wO!Qq9MMv(bQMT=b)!5}3PY`NIMKX9;_&Y;5$LrX?Ee0f
z5+Cb)(0cByUXN5?u)44H{IkBU@Nnw=pH;j{I-no`0{RKCdY`f@9Q`S~;?nEyGe&tt
za?ff!DpEL+?1*$<>1VN9S^xfeH9WiO19d!F{i1eVk5q1P9|S-E1V8`;KmY_l00dYP
zFx$>$X#>LgF9Eaj!2TPOzAI+Sfmwh0tP;l+tA1&%aLkqmqyBh}6G!Ubv3g&Nn_I5)
znwMu+xT5xp+Qo-=JVu{myWERYL&H)K@Gb#X^9YQFqwY_~`L+A{7%g$4{?00P8=d#Y
zd%m#Q@Xfy0i0;p5d4+or00Hv^c)f4i9WL+pAB3-O-JjKd6sP>VrJu&E9bPM+tk!Ak
ztr@l2jSY%^Q$GIX<HW3Lh3nSqyq-U+<x_7yD(>|D{wVI1_Xp_p*ONtBm&EG6k{@nU
zz3*|N_CM%NpIhCro>yyk`LkMXjM~fUd1X}|jN0!u_q@MbWcT+~yc!>?c4ee+c$ZIB
z&mHS<^!7jUYG-P3XVu=eT2Hg)tv5U^F1)YnwNLGiw_OPT`UK4OKe)vo?|BE;r-rK_
z00JNY0;~z>?Z;!~j<x-63y*dGrrz|a+jnj~cS}1ak{!I;IZlOdc3)rko3*LiotZK?
zL>C9W^{?6;>)!#p!&9e6lyilXmc5bs6<OWaieI_TyFc5m@L7G&TPqyB?M@}%q1PW@
z_6irP^Bm0jkJMh|eP5|J)aurDJ}bLek2@>j8Qo_|YkWh!o<a#flB(xBsoksZyY*aM
z9&yf=0ITOHQaEg_mL&CaA=<gsamY}|P*Si$;HE%aAf|VyStt;j^*0zFOb8YVNIgTr
zB3b7Z12KV^#;L`6g@U8{*7K#Y&IN=A0w4eaAOHd&AToi;q42z64b~F3s>p2Zu)X^o
zUq16ej-ABqY6lC%0RkWZ0(uAx-_hjovx~P7QPr)#*J`^u{&lRydwEt8QB+Uaal(zL
z<<x6-m(Exa%yntDY?oRsYpnU}wMShKus+vo2izGWQ2(2^2IMWjgFxNIAtCPTPiyzx
zPDE2)<;Ym6hHDT20T5tMU_`wQ7f;_5CPtN&C9K<bZV6Ac-#Jn3b$h=t>~nYjKCkwj
z-f&sfqv~+f&$IeI&g~(fw;b5*&YskpkHsy&<3{aJKX-c1&6A2wpQ7}~8!bO>cdxfR
zyTzT|^(w3Qc=NlhMU7r^j|`c&y1hI{vX6H@unIp?I&lBWczK1{H$Bt!ri}y&4k!oq
zJqNs7sjTWzgrjeqj#WFyE#V;F_2oM_Im?N6J?eD$Xs3!juM_!){1bUzvMA-(>3SOR
z^C94t_OMwy^p1C%rN??6u3h1oO|M>mX7kHwe|ppBl)D!X4E(-*-pM4AI%OZLc<uV#
zV`Z=2a3a00>J7)JJG<%?#N+Pu-^d@a^2e!p1xD)^z3z?9>(z3vH$J!n0Xqpe-F`*<
zSJs)?)qZ3hPo(xCtNV676Nn20Kp;8^Sgo(T@_WYmcf#%AMQ3SepK7i5A6z%LPE<5*
z>`RnbA?h8!fx9c~8lB2rtH1M3`{&)xa4KIAF9?8uaRR*S1BCCc{h*K^#`44X6BQjk
zFup3*_$u-X`IWU_e{|GN?oXxL`!!*ozV!s|FKMt41o#tpDWS<!Dcc6wDb-4$TiZQO
z_0MZ@QeWp)|EuG{IS7D&{RC#tnO64T?p=hJ+MgoC2m&C$i-1$@6<+ZwoJ8*rSJD9m
z0T9qnK(yyxf6Cwz1VF$n0ju-t%*N{#f0ycHrq54aAd8}QuRHZhKHHU!P~5IZ_We}J
zFRDKG)dOw6eW{{fRQ+tw>E6#jb-Q0wy)9$-uW<*;`?YlX+xJaB_e*)dTXpvvC)UnR
zDd!heo7Z0Y?cJj)_(j!*KO`OQF{!Mdx7t;oA|4O`0T8f@fZ6_Av-a7&zQ(MbZuMt3
zKg8$kzSc(hTkRi#e|LYEF<SG-t@SbN0s%J@_<O*%&u7lb<!7aCPAhB!0WAV{k0T(y
zV=dmxvyy0W(8?!V2LU4lM4Km~w%15nt+>T^9o_L=N3;EH@E0BXY$2b`<+GK#k%eDg
z>#?FOTo!qOFhGDeffey7dp|F?PZmv=kH~(@_jz;pMb*-qTJ0TB{+!H8d6yG)_*Uz4
zb$E82*L?iN11CG?^jlTC>=iX0_4^^tjqL|!H!o4k8SDlDmIRE>Q&8Nqv;kp(fSQ2W
zeou8cI0pee1eAJBbT#|jX!+B_4sJjI1lSN@UH`DL3ZZ}i2!H?xfB*=900@XiAjiaE
z)la3S6KHgD=<#vAf6DaSL8480J*RyJzH-NmzKaOd8V6qagK)19(CXJ?eI2~n5(sq+
zB@Lc5Jm%j}gB1dCfn0%@-l1lpKy23EV0<tkSSTR%3<Zm1omULR1Y#Pe7V8xXj_O;_
zm&Vo@ga!g200JNY0w4eaAmAK<*Squ``b^Gk#O>-FA;b&<AOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w8cb0z>Z1Gvd;?
z9RvyvZY~GcE4#P`0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&00K@Bz`R?`yLI!tTc^vX6VD{#1OX5L0T2KI5C8!X00FlX
z7(MjDnW?ij5T{g&kJIlDZi`=w6Rv{*2!H?xfB*=900@8p2!McB3ApXOERsE5%?G3l
z1V8`;KmY`;M_}XYDI<%ozr!!8*1P!jO~rqz=oeM1F6+{CU3?|KsQN+4oozb3SjMlV
zy=H8$TeM0Uzgu<ClUqv6K3Cc=s)}0f6}xBrwzyA&>ZSd>^?G$5u7LmufB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zuq9x2zZ<qzAT$sVi-6hZM$4aAsfV>7AU1)HzU+fve`W_^rDD^7^&kKOAOHd&00JNY
z0w4eaAmBWK>I;h&i@Ugyh@xpN)-0~EaUF4+I-irU00i7eAk;CG6s!=)704Zk=^bho
z3dCmp4aNr(f`tN7&rq;P)_KK1OdzIlYO!9S;HbX!d};1W2P^^s5C8!X009sHfk*^8
zls^4VyJcr&y;|^0!sz>_9hZ6O8zWx$<(7vI$&u{FeGmWv5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5a31NP{B*TAG~89
zfkr0>UgZkmf&d7B00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?x+?YW6hOs?==$lSN(SZkw9h_ZoC$UR!Ob>U}1b*q(?#{=*
zm`g;}vDJSV{oAV5L{xqG({>LPzPOb@o#7x#J*s7==<7KilKMAWH)=QGt?GIY=O6$A
zARsz{-#a#V`(l^5vQZU1-LM}7KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)
z1fqq&{JcGLKjl9q^VVqLiFASh2zZA8@A*>RA?0jJwC209zV2))5jzNg00@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@9Uv=9h&3?&VoG(6_t^ad*gatHDR
zVtR*~g#xize}nPCgkYh7)H4(;l677&5EF=LoLa0`C^)KbJzv^?G4=jxYeK&7GDsx-
zuTQpDzCE;d-r<c$WzqEZuNv*1Ghv@Bs!r=av`zbWcgUjZ(4psF&Yf{Y7F}yqP2MqY
z%rTj_vVCfehnCj*i=%7m9wDM?%H7}182I3EVpWZkgAWkkN1#&irz?KEb^~#%Dk-n_
z^_Bq-2L6`y>UrNc>q^8dCcIU#&nYeSrZ08Y<}Zoc)hRw+i#J^O-eyxD+(YbA-uwn{
zy>D{m9}f_#T0XD;`*R;JmF?Eb72E!DQ~9YxRP9rI&(o`xgh{k%_RJIC{P9*MiKKVm
zwtLm(cMlP_>GRK|x9^pEAAz#vV79*dtm()F`~TV_YpL0A6@O=k_W$D9P5%(l)oeYv
zziHLuwb!PTXj3Ks&1%{8z3)zcYZqaqdhOud9@T3{BzH+8Ccbw2+gpg;8p$rF?oZzQ
zV%3o^UXs0{qW2Ff_7@y7HGbI_7i7`&_52f>Zkn4ZvsPZuhf}{VywVdXTv6)jNa^5p
zAHNINk~RB&eQhYl&%X8L9J?a-c4AhI$EeSTQa>A2eW6EoWmi9t-SI>kmv-ttt9G<o
z!m&GFdB<mWeQh<qH*atF)a&=HBv5l4BG-`ST&T;bxlkm@?C+J+{#e&fX8D~?_@v3#
zziyS?re@P)mp`|T=SH&2Y=3qnu4db<YJZX9(Ynv;y;3W_NY_P~_jNe_*q9>myJfHF
z)EooL{?cf^EQ&^YUfp_M{rk=O+-yB%)SpxDIjrJ!x*el8{z&7%R_|M#H|Q4sYi?e4
z`;Si_BT#TKQV!fUe<G4SHCC03Tig36af?Q3&sn|imUKpo9iq?UP}?8rd2;%`+FqR7
zGoOf6JCap6R_jTl{#o~1MT+0-eD@<?U)Z;$)>`6rW!*1nS9nI>``qqcwEENSJR{l1
zy8ptd@FJztuKSVdQQr6MN++*)?0)|B<_E9z+7+(ebufB)sNI=uA2{t#oetLLyywC3
z3ZFFx*5OUb_IT?-TXOm%sXAU!#zz!8wbp&GzOLrP`#e%T&ntd;?Jtv>Ry-|xMc2Jk
zIcD&?CuGr;_dXS@!xz2$h*n;@w7BW$>G|hmyOmcx#A<w-*ZdEyaP8{Xa9cdQ%7ItI
zoxko>_Ak2Zmqk;(`+$hz?(x1|2L15YDp@qGy>sf2=I{I<^H!&xzmG4?er$e)-(~3a
zIatjX%zHOC@7*qLZXXZQYoAs-NnL)ewzuDzvdI^$drB5vMP2`G_We;^4xBolP@{Tc
z{%vPvr`2rw&F=33e`4<=rR1O4a^n?$yxYON!&marF4ddwtkbRZ+l#8a_iNLeFOl3?
zZKp;GpY?s#{aLK$twjn)t9>ilb-n8~;O^S}Fht!~+pOK9&lBghenxaItnF9oP1e5?
zTH*D~yL#ngUDnC0RVzNL*WL2_V09l2R{nXntLzG&6%V_@v5Sj39=-jCtlY60PqzBp
zEB;xf%c*db`rRoNwH@k~df%)athp=C11s%*t_JDd;IeF1joNXb*XR*7-^=Ycs!BZS
zdQmxNJ?~cuS4$Oppymh|-G@-^UYW1TYQGhu^BJ^K?ezLlt@!O)7od(ui<9~~uX#P{
zc#NJKEf;X_%={pu`5MW++24Dw`cpsG>hnnJ^wjRv=aJ^`N4l?-pOM=8TJJfL{MmJ%
z*Lobg!WFeX<h}2l(tfd96AS-P?v(}yWmc;7+*!SD*Y|2Kw0xp|t%8ST-m36m^*pfN
zuR(8kN;;q*00R06X!Rqw^}7C)U2&=P-szR=MtMYXuhs5keLa#Lk?ymeXQDT}NbeI~
zxo>qmfOY?}-thG1BQJMKxivd4BU=0^oZPBL`%8GIL$vRfsqGbgJ%`zGD7W}i=Z9$L
z`yT7D;<XM1{i5pSVZBG>x#;tYs=wb-zUj&L3;TI1?|KyBgMgO_L>iCKy6<HsZb+Ni
zeq*z?e_kth#_+OxiCt>ePPh8wHSTYBxU9xg)Zv&d&tCCo_xKp=_*jki*d2~t<HNk;
zv1^>(Xgo&u9m{v(Q~v|;XJogj*?kP$=5Ox6O7TZ;*(8gqtk)ge9iDo9vsQbB_4lRp
z?qkf_o!)--XmzKR|614E+`A&?!TuYGsG2gPeBF%Ow-UYjdxso7%FfzB+@|9u{`lMO
zr2Qn?RI6M%b)9v6r8hjM(rYx{tp%T3RqoUo*)G*<pIz+_*7FRM@bsqBsyn;Zd-9IQ
z==i7G+<RpnhI&4_DCeT{7lpH4J6W~Y_12HP>L;Z>7e$Tszu4toZ$2vS)cwC&=Zbr+
zaVAl&^PY#b^Y^}^e(Q2k=B-M8xJ~u$YbMHbuC~+a{7I+%v+ifJyZl+T->AK;*6FYs
zA2Vvd+uZYhpAfx$>K6O$s;79DPrKS(M&r>tPn}nJ(BjT&+}CP7t+yS)+MN~`-q+3U
zE5h3@gnxYktjnF<;aPn?Bl*8RHF2&Qt<TN6kHj(B?z8&+aXOqx-w7F;uh-b$3rl{z
zs`ccs-WL|(m5f{4`zX=k?bhq8`hoR^<5mu=?9$u+uGgJjyLi7JvkKp??_gMshxdF>
zr^1i6b^+3B&-Wy~%2{vyt9EC0yvQx>D0O;7Iaj!A*=zQF%cwuC_?7FTjC)41lXrjb
z2Wvjdd3SOSe<UsR=ctF~<jvvts;c)Dj@EOh)?<ACMeG+JwK^#~rAob_R;PGGir20C
zg(-F!t<Tlt`EEHknhw_P)#VZAYzZjkmAA6BT9SB;KiU<p)%l8{j-jMrg+QJ_-at(6
zP_s}VHtTONK9~?J6p(s`f<>~<D+Xc$F^yA;^$G<?^{wYi`)@A1{_y>2-52j9k@UYl
z*<N|RW#gfr9Q#|=(p5PY{@PO7B)d%?{_UZKmGiHXMb$pd+s<ooY^N-$KJ?_s>^t@z
zmqphcXJ2~%RQhR|x3Ya|jfYkx_Jz$)lqN(}ePh8}Gqc?nmaVFBa_|8H{0NK=#SMA?
zz$>y_RZ01{H;>%))4YpBum19A&VDD?94EY0vCk>}tI4!d-)uV~yIq~)<F$C7xLBp;
z8wd8vb}4Uu)ymBw7dx&bR@JF?MWpv9x8HYby|33U-tUEa?XcTj)q@x74W74`xLtYw
z{#p&c!MmeZwcU_O6g6AlMal=W;VS-?eOLZ)o-OBP(beqxy4CkCR{rgN&XxQ(tE|$k
zgrnCE-u+5??TF;gYP-X%|44Q@bzi9my`s|Kc{3)=IzXTv<G|?wQR+Fnl~Zye2Fv^X
zB%%Q>ci<m7prdcWd19Bk9EW7K-@$6StiI=2jbF>Z*>;Xrct)=~)$hSO-n(lYnzC->
z-^6Wdw!P<8e{SPz)gIpEl6ClFcD}k~{LdSSsA~87usS}Y^`&0->ip2_Pwmd`b|kC#
zSl3r-E_%;d+rcV4z4klhPW_ySa&ET%kHnePa}g;Xt^3<j@0>Mw@ma!Jwc?9(-D<zN
zUH-csU2-}1^3%jCy0A{q%HgR;2$U}ev-ODE{E6Og$?Er(cR1Yp!93Cr&6^YJ@YU^j
zoQwUwX;=QSYA>=1CsKRP>V4kzE5c_p&&{nIS@(;v4v*Dyh;WRpgK%4UwOUUa_0M`;
zPsI35wlhU}A06@H19=Z@jtz^ZqQA$9Zok{!>y6sO`uEVO@Z7RqHex->*@4mgcbj{=
zf7kV9hh6h@c*Uc4-lbk1YIklcM|$m1r-S*q$##?Dzl+R4unjQMx(MaII#yBIql%rZ
z`z^J?Q*+{dzPj6kjanSYl(m#s{BB>*8m;!3UBBa$zd}D|PWX7&0a;YFI`7qP|7Pb0
zTlL4Q9%9vhWA}THQ}1_9$IGi6cr{$7^UG}fc0W&u@A~_lY3<v_`nprkpQ!ByyX>}_
zFLv39c*Jf8Tg^YM`bb@V&Hf%K{_L8c!aE+V{s*^TH_Js`4xBpwuAD^L?K&wtt!C42
zmOuP?V_u%wa^n?$%I|=vs`Wd`>$={2XI($oJ$}hLKHl@t^yW(>cV^eeEB+#d&-%XA
zen_={yZXJYwvXN7-|qfutMR#IU8UZ61ByG**VEnKZ*$YcZBkfNz4&aews$VbA7-t*
zd8~fyjn}Jf$`%$ym36ADe<!rU<Mmu=#b@=p)pi7{b}sAok=gL9{vNo+ztVmZMOlrr
z8V$$pb`PuaVfDFJ{Ijyhsc@{;UsnB#@_ec7w3=?O`B$C?W@>dlsX;@`{theAX!SR%
z-)F6!N3v6~2WpOh(%v<z2_=G$B(FTlY`+y{ey?8DO10H;=QaOYoYdEYE9T$x@wT%v
zYgNZ%^c?ePu0Nl~ZCs4zYb5vIHu@*2^xDfZ^a_Va^<t#;F_FgOmGuc`RVzPBjIYsY
z$kRE(TAJ<N5eqwQ$R1W}t&(YFXRpf<c3RbTiFPhZ|C`!Q-s`sW)@%Atjk%YJ=&JSH
zS-q~cj)(PiUgO2;a1|br>PffXSJD9m0T2KIg@9B2s&>Vzq?47}H6G8(UWDUL0%qTr
zME9rUhS~8`r~KJfUn%h@oZPBL$7gt_!>)RnS3F9-i>kc1*cGnZ-`^tH=TyDODqg$t
znO8iv?bE}%B1QNh;9&wr+bxQF4<m6!($xK~de6-oZ})f3&D!Z!f2`L_sl$sj?#SwX
zq;{3r`@Gu&tkTab9Ny{Y9e$*7bGP1)H2+PM`(~dz-u^`47cF+WZ9bDxdqnN;QS4UQ
z$BJsTS6I)>kQRM0WYvmXVY`)g{UUn!TKTVa-RSe;HurkpQxtb@Z8xaf-_dfeRW6;n
z&Z~biQn*f~*J!+UmrK3&@m?3LgwOi>zDjs{(`nV6UHet=j)(XD+@~ro@3!OcZkd;Q
zZM}+U>9KoVFzfhumw&s%XSIGwZ~dseFR9gPf1Elzx1aOsr?C2-pf?{CclBF0yZyVB
z`TdbpanJkriFf$C>szzyOO5(d^22Q!slBe;=k@&X4%hB>JnQ&a^Rc`9*<C+cjZg8<
zYQB-(yjjPm@ZhDa+C$MEj(T4dUgxaKmpVLF&mHS<Sl8=D!&7ro&Xw^It7^6VZZ&?h
z{uK^(>$o?M-1XDEi^QnfZ71TpLO^R@AW^QHeLnR16TO`1O{Z0N%KH#BZpixwUeWxX
zChK+hyu(w*O|{f$dqV4Zvids<|8LFAc3)UlbI{`L*6YgtWOk`rIk2)z>$x?1U9Vld
z-m_ST%ex<yRrp^0UFFq(f^fyye-8O%ET8n&ziM~)z1Mr(*BdU%MwRzCmfrBy=@I2z
z`JL6W*RFb9D}LoVtKTW7!%=wi(wvQR%!o_B#UDv;=~#N{bF1_D6`ghHfuB;668v7#
z$?tDyck7-Qf3)idWrn3!`#Xo<Zv7?ivO4c=%jtJYm3l+1tlQ_T!eiB+$*aCLT3>0k
zo7DEH`vbJb&DG;B>hRV6)c2Khb^LaptIH$K*%DClnYXgFT9QOx*Q0(ujh=@(hLQ$P
z8Xog6*I<P}-ax)UOz%*$P#`wzZ!kWX5G)jsdWM2Uvd$|8VgfOZQ;YQq1xNL*=SyRq
z9S9EuKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;
zKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l
z00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)
z1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;
zKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l
z00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY{H69{z-
zB?T)4@&)n-VtR*~g#xize}nPCgkYh7)H4(;l677&5EF=LoLa0`C^)KbJztvn_}~)+
zKmY_l00ck)1VF$Z0`uxtzcbIQ9mFeY4>iOB0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
z;u6@tHs7=DCrf_OR9xb)90WiB1V8`;KmY_l00ck)1V8`;KmY_lz)b`$#Qr#bVR$Ej
zMi&P+l|k470w4eaAOHd&00JNY0w4eaAi$o$eVf<L_+-&8qNV2#9W1smyo=~n5}Wgb
zT(fo&*2+HR2oVH800cmQC4qm+kGc1r*E5J-9db_~HY4{=qE|ml-}C;u@;iysIwiGL
zj|=N|5~CXI75~EB|Li1IRUVKrKiB=c2rFe-1`rkqI7Q%(8KH-IKAKL<>dMgIsYOTa
zAWo@M+z>AafB*=900@8p2zZsisLu*MpZ3ZeqF37%JNDX{)K$c&K01BuW1G%xCcIV2
zr(AWLe!q6iE@D*&we8WS)sS6;mmVJ0=ZoLB?jmYc(g6hl5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009u-N8qOdk9Y2rw2Sc4
z<xjWGUS!lR0tE+@19r~=zZwA{g8&GC00@8p2!H?xfB*=%l|ZrOTbj(h=bUU+r@!_>
zt}_Go$)c%S^96Q+00@8p2!O!<5*WEA<)h$-I|-C6hyP`a|APPsfB*=900@8p2!H?x
zfB*=900@8p2!H?xc#}Y=V<>6xq~S6DrY=|^kUtO~i0K__77D~>{SC$k6M}^TQqNGZ
zNY;79KujQ}acZ$%q2Q>#^?YgGgo0Fo00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H@<0;6g+p3rbrI)Q=%%E5o11Fh$6!O6XUzPd7lu+n3PUn$%2`)!1^YT1wL
zAOHd&00QnHV6@&BwY|pb_tGulxTEyI3J|bHAn~K2e-`<Av20dH?izS>@6_qUDYce*
z_yz$G009sH0T2KI5Kt4CwLU(7`zD8Ez1s5DpW78Foh}>Imwqqv<B5m=l6k9=zSbLl
zsXp@c9mJ}Zc(1_B<FOfpmu@Z8!T;Fs45C&g9Z(Pe0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0q+qA
zbqpm1D+J;L1p+a>L(M{g*sQ<7_+Ub?P(bP#3Kq#auNa64#57JV)+-bo)wiB6?Y|tp
z_^eOgVCS83B>k^XwpW%9s=lh>)-JM^{xhu2<c%NwN!+GW4pqFh)cCcssOmPJM(x?z
z@!M6m_56!K^>JYP)EW=1;m8kq)-8TWj&?Op4n9DDAA!m*ja^qiV-0buDk+cDPEp@K
z{KB;E4=*|?^H#+^r&O!m<hJWh@$p)`cDHwU^J7&`+s52h<V;|jY`2Qup5a|T@(%y6
zq$zXHRy`vd)lc*1+W+c(mq{dDey-r*{)0~vx9RqiMSk5-`8a{H<zTkHi&W0dhO7Ad
zI#=GGeqDKih^{#{J&?3+`axn;PfmUPq5gSOWv8`LTHRf*?cPi5QYHV*s=8fl^<1wV
zjZ+_ObN-f{gqP~IBa%C(`tKsy<<xzn&n4WOtIwi^8?W~jUe5<Fdn1L*D?O3I<^BE=
zDSp@AfA)9(fY!%l$u4y{5VQS+R+D8mUd5l)__h2SeeYD<YlUa@y3uhnQQSM#E<?Ox
zk830UqAUN5)|YzStNW$(`cu0TwLHEN`sAkd4fn~cRLw>2Iji_thojeir`(D1JFm9W
zbUQl&=ue7vT|FsxT-d5ch(_}^uW>2O2qWB3+Qn{HUj2toh3j<xwbA$!n$M2ec=D($
znp*AWj^uw#`|;gMw7o3z)){a7{9v1yi!zii2j2a$k^28cxzC#u>+tOwr$an99v5}$
z_ttE@ZhIahH0LbeHp>bA%#Cxqt$bOv$L-|+^Q73$lhVuc|FQSxfi_kD|Np&aQl?Cm
zkTUg-Qc@}V&_FH`%@HYeMN*U|Dq~1V*HETV8A{1eNm-JJBJ(VnhlG0#*X;Z2U61qm
zy<ff8d+&YD-shZst$#eN>9t<#xnAqF#<Q>2^Le!LqIQ%n1;}&6Xv22Z+r2~6r3Yny
z=k+;WtmN74Cr!Gu-=jv#Q?&D2yvry0IbG27v-6{Vr@K9eJKOU+?T=XbbHD0w+Rsk;
zkolrrf5fZ2Y=`IJSiSG^x*pGo@K#{`noXOx*_{-!Ls@I2^L)I@#cEtgp8O8m`6IvT
z&uZtXe%=4ZT5i0?SETjEYFzqNkKOo7wDd^(DcW*=`5~|NHGbKLPOeva2ThJ&c92!&
z2CaY5_rQS*I+U4vKxGw7wBMII)hAZ`$7}pWJ1=GR^3nE7P~S%zy|c#e45a<a_sDL;
zy7gGl;uHNoU#Fkj{vyA$gLs7##ZDq^e^#IGO#A-#J#D0RlU2(3{d@YXCf{kld+irv
zeQxS?eO~h+<vHyayL{2M+fKgM>HI$5D~$N}Y%cE!qvbhK$_H&<l_=-$Zr??||LFsh
zCYSwL1r>Gc|DeSCwT|R`)Gv(Hd7oGHFfP2FCv@`UJnB_gS6`&@yl3=Ezi8zo%KQ^6
ze;zGgUYGZ3yhSPpdyF;4`kjeaemT-{FUuv$xQkSec&`^@<uAOh$FKEAQ02;T5p6iy
za~ikxM`|C`qx-&qSNmY6a?r<MWDG<*f8tLgwI^8)c}`K+>O}P`k<MGumecVh<Nf+x
zysXD<*r`33eDS?!ciXRa`Mp~IW)(ShzoWL>pGbe}b=a;y-KIak^oQq1%swX+Z~S!5
z;+Ze@$`J}G#_C=nX!ST<XL_A);p3m3`fTWN%PVv{uAJtN)jCS|odK`oX?L{iDaw28
zKb^{>9R>{ph5_dcFm8i_&Rh0Z&S>}ZiC*3<A9nd;b^mLphh4hY^Pp3CLHW*4w0cJS
zeTHc3kF>nk_j+#SVYi9yH>7?}*I8cWgQrAA-nBbLzuOV%y>OzGb32X_Eq|=$MN`jz
z{~oCHS3wIWnmxPKCsuyMZ$E0}8wRp615WobGGA8aBoW#c>AW7P?y<h#i&W1<EyufO
zzc5mJa9hsq{xi~aw|cnE?-%c0$&Hk^L@DRCKN2Os)BSj&=Zm#}c3PiX-V*hFQLOH1
z{jSHW^EM-=O|3Wb&$Zn0Am1VD)+=6*1H13N`jsA(d%$?<sq24TytJR7g(JO(Nwjj&
z-e(21yv{D;6(5vw8!vuhb<RpXPTLtP`}fN~I&EjP`Ffm}cTel<&ZwZH)9+VCn(yF-
z>qhr2y<ho-iLzgfR*ytEFL0}eSLZD8l7oLV_@CbicIqE>^Ssk@2wgt6b{Oyck@`ow
za^rIpXm{+f<0?{g+K=S3`<)Hx<k$6IP~zR*kGstuX}`HGM}H(5Iz5MRdmfo6`GaET
zIz7F9UlVKli&lTP`g*lLlK1kFhTZC)X!+gxQ@8n<uM9@_ca3A^Z~eAU-Fh|N@ow$f
z>-;+2ozCBO`MlOQR`PVb_#GeVe#URTjQs!3K&<T1uX=RrL$BihJ8J=ob@`p)NWTYA
zq<;Mi-YAzRWYF;NzwAIaa-!W2nexMyl<K)bNgi~3Cu+P`e%q;>MCHLNy}aHZIMvJR
z@2~hJ&u;ynDCu7JW1`4Q6n>2!k-a}Oyv3F`r~X&wv%9ZL^mN&tpoBSp)ae_m_XE26
ziI?j}6aMe(X{WqIdJbmF)!z2__MIvyJVg6`D$???@*BE(@Ea0!`QsgrAC4CNI>+>C
z9VOTGQm)hAfkax)FMgcLmG)@D-ZJ3Sexl9it=36NUcG;amz-$dou_q5Yc*){@Z^8<
znp!@kU`nBs<lbpb(o(Yh`EP20)WWGnQ<8e7r567)Trnj%CAs0;lD*PWNA;;wX7N9}
zV}7}xcI5f?H)V(aEa`2`6=l0G>(*wTrNhv;O?fV@^^=uoxFyfIcYjoHoC+$QD0p?f
zF0V~eLB*9hUR-~}nz1VA_^-OJ|G59lp_bq1?WlDxt!>SXEq__ES_Ku~ZMFX4p1prj
z(V~0iOo>EhVDRYYx<4^vu?i|~Tvhz}d#+lc5*6wDQoXObbo};o%PGccy?}o4!u3U`
zzFYH=<touJUi@UWdSC4}^4WeJH>y~n-~4&ijxR5$T6avXZ&j=@_3g?dzFGa93MzKL
z@PdX5wk=RW#rvnfTDi)}Wy&qu?R!c0%0AClIq(=1jMU#X=(&2sIq6-L4kMLIxiu3i
zl$!YVb`^A7xob|Xx4z$|+~UJWKmDQWV=Gm>vFX)cK6$joQWYzt|0BhWQfuERykf2L
z3Z3e)qvUNx#tmDg{6eRC*yW3tU$)aLUg@7+_;brLWebI}3SM<dg*s0j$rlPbKK;p%
z-qU6l3Iz?<4&T3^W07-1UZY>*qr<l!mCE_#S`};bt37sd{hr6{>QD6ay~7GD*-&GR
ziWMe0A8nhw_0sAO=P0)rZM&YcI;^`P?|Ky}Mq9tGe6I~%FW)w0p>m5ld2Zv6jv8}&
zy$<VDR>LmeT|cD5hZ|I)VZ7Fhe$|`R?55kz+3)(O_OB`^=vTkSN^ZRE$F09~nqTtg
zw4BV>vFkY}OV?bdf`+YU6#3xYYgQ?*Q1aq5oH^=)3p;iE+)6Zbs(-xlN&C&}VLP6^
zYQJ4Qx^%DVmvVIV*~MRfS+NHDMlY}u9Sd8nR+gwc(8?;<?78iEZeO@u1qEOK;pc)I
zHf~Zu!<1WFU2|2&W)*boJ8kf}BktX-f{F{bRsEq{zs)MqQTkOAj_o@W!*|x^JJ#pv
zS*<?1^<1>+iSD1Xntr)jzc+gEbDLCF!P~wc+WExYzo|sUL_0@5Tl7h^cVE%^Nw?*_
zzF)Pgf5-r9$e^pHE4SDy{k4q;ny0CR#|ZD++}_W5o!{+!X`<)%njgRAB|4vHi+*nX
zl3)4l{HWjQe%rg>^1Ytd?BqG^kD%}SoW5^%%7@J7*Ld<<u55?bVW+$&T0XCN@LJxq
zAMX9~mRo<b>`>O*<d%K+FWR}z3JT6zu;<FjUA9_&qvXl&@S+dX#@4!ji_)QA^{2l0
ze79-YR;#RnUj6<e-g1(+F4;D8-9{A@+`VpM`gd(stDxfPvh8wp&A(9v1>?2;^2-l-
zbuSg`bwj-RA>Q@+Wd~VRZcy)4g7zN8PTwVu9eeHm3R_jsP`57hIzC?G@1^3smpA`p
zo-+8~d5E@O?8b#jkMH}t7m`+fb@%G^D(I;Dor1LA9S!cdVb#WyR-{;UcG3QmUOHn1
z9c$0*`R$(C*+N0X%WE}Q(s|P<%PmSfh*!AOC(U2~;-Z;WyfM=Dd$nIOa)S9DFjBk8
zD&_pX2k~2etnZ<n_Pg8fd)eh*@y_lIIp;4`LB)q^)Zh8Y&=txrblNX=`J(+Us-1kV
z)88sRY2G!Z7O9}&(q%nP?>V|e1s#KSPM;{>&AlVn{K5@pZ&6P1?$-zW@>;WPDyTU4
zt4h7|4A`cEio<R=cyq0P?NDCh@(K^N>~wgm@(Y9Z{Xy*8>v=*a?~!X)<xhI^56f%R
z)fa928_%Bj{maX?s7NuW_Xj$?dmX&0YPUYWsi0%fe)k%${r=YbHed6`qLnIEc%o(Q
zijUOVprXZS$35l!+CRifZoJ<I#mjD^tvAtrmlJP&7!O`ywCA^O^|*b{xgT^IzsPcn
z)T4VZZuc(RZMxn068ad7jDg|roC?kUqOpn;d*s{N@=&wM$}1dj-gE0(44SH(qOR4^
z&X<0bkJb-IT3*MKjQ^nho~!RVzi7y7l=Zj`k2*GP$^6CdDT9g4gWG<!%is8!CU0$g
zdx<jm7Y}ytTkO_F=M-J=^2Ax2l@9I7+okLJvq|oDZGXNaTS$lR)W}(`X}4@4iMsr;
zj$gEWMX3);CWnHG6E~K8_=)FFSwX`WU%h)$nfs1ee&OUNXD_&?Y^D`!e5KKl8X46x
zEU!>E?xKxf(Yoij!xwF^{Kn&*zrH-V=oZUwgpYE+-c_U8neVM=v3#|!XWg5wtXL!M
zFlZPs3^-@NuYH|ga-FyAubhc;zwDPDZuzjwAC$k-(Z0x``!hOzuVRgTMr_S{=GC86
zR>!_?d{V0NvCmbc80oo8(91Etg7&@o_9tswdq%ENu|jxCRD62Heb=2^XRdOJez)Vc
zPj)mr`092msJQflG7Hw-vEB+Q+I@c`(bI#<J9_sX_`|X~BMw<X#Xg5t_0CozDHK#Z
zweQg%xBZlE`Hg=2Q6t|lkd+yT^g9_@nUh3lTcqczk?J1n-vLCbXQGz7a?8%iBWldI
z{K5twKhy24=N4Oj<K&@_HyirlQp;;>)NXaliMRY>`GudX&epZd>Lpg9;?1eI?5vgl
zXDg_<;OYSjKfGyz6*P?WJSx%3Wjy)(-0q+Bwh|55Z#l&Y#YT>5m3N7iXxO0V>J8_l
zcTqvdu4{h({P`E>tDs`Ezh8}a`FC5bf4FDwUlgcYue2NR#hm*V?NMG~yuJgcd&g|z
z|N1-Sy8hS22c0*!@6#jQfBId{Z9B3`ew|&$D}MD|cg!p^Xqk#N#;d(<>zz<}V3k~7
z9<pMEPW8+2L5)G1KAvXzjedQ9!D&0A&F8ltGx9(Eq3dHSl^qX3e@75>{q6cE(bI#@
zzhC-0^^g1VzrVQ9?Gu&6Hy*0Dt7M~}RixPc!V4NM*tS3g1z&r&@ql;g{iwXgXwL(j
z_9OZFRcTmo&ZZn8J0yQlzYB)me*33yo41&D@h2-Szmfh(G<52poW8$Gw0!p*y0Af!
zD$7(*vDtIm^W460x$+8QZGX4-ICXciOE;@X(XGCr%B9}_cH34fT70P9Ys;@I^Q#pt
zy463?^6S1Y<W+pn9I4g5E3io!jE)Dp-@kg5?zep=>U(Xc-{<k``%60By*k(Sd!6N%
zd>t=-$6tPX+XD;c?zh~c-+CGO|DAz^|9jvzf4AK3dqdHt|994c6C11FflVITx^KHN
zGnK)=c<^dp5^p(obno-hl#=T#9o9>pJ?rHzYb=R%UOAFnck?PsqHg2_{qN5ts{T5@
z6E)uHxr<l%5|sz9^y)ip@VO)I-K@Mqr+Rs{5B6J5qK;dsSGSw9-}O=LUsYDaF5g{0
zq{D|BltIITu><>Y;MD)he36bXujbik<;eB~CCvGvPG7&yf23Sp{lo|Dd-6Jcy;|q!
z<T$-2fBv3J4%hkh2L&eD{a>{D`1KAhXmS@9DpO`j<s~ZC=+*glq;l{Z5__z<=*mH-
z3xpC4_rLY>%JTKjSka=}@2@+}@Ae#aY}&Gd$xnQxBz`;mX!bShW+_mf#}O~F^^zZE
zjn4a>0_E?~<oCT#6ui1#m)9mKiCt&j_sz$Hp0(nI(jHCNTLvP17w5K|w^}D9?Q7QQ
z)6AS}RZx+4Uy)*3r?gh7<x>i!6i!L*oz^5RCEK6>rWQyooLV#`sb^Yh@jt^AQ<77X
z8_q4+D=l?YpE_k0d)*60o?*Z+U>GnA7zPXjh5^HXVZbn87%&VN1`Gp+0mFb{z%XDK
zFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+
z0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn8
z7%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXj
zh5^HXVZbn87%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKe
zU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs
z1BL;^fMLKeU>GnA7zPXjS($-bU%uwDE{*3XgTdw@E8D1Pn_<8(U>GnA7zPXjh5^HX
zVZbn87%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^K=2tz>y*}N
z(B$FC|87ZY`IN#bMN*P`r!`4S$@b^JsRdFCrxr~~>Y0{W{LgU3l;o7;hI32yN=qHp
zr%sv0NlBxpH*9gB<W_6Ys6)rLq;6EdEUCMy+&SCT?CVDte>D37b$<KfO&+^$@;|U&
zYUd@NHCSmqc|+x<SKhQpwZ87H_kVh8b{Gr%(sbYP-xpb6yEaeXa?6GTVL8;>q{gL(
z+E4t$N*cauNU4kiR{s(A*X=rUi8}E`i&d}p|J};0UbxGuYqwiF+VrVgZ}vv3=EyDG
z>K|UA$}XAp-n~nIQ|LWl_uNXY#;&)@J$I<z?YHi*x^&K!ealz-!tL5MvhA#5Ro5!?
z{_xsiUybRo+aezQtEE+I-E?hhYx$jns;+wYcU8Z7pW69O{bqsa^I)S>b;j*Y3SGY8
ztB(dGWmxc)YtdIV2iN+}LjKQ(tG0in;x_Bo$M>aeExa_`4(danr=H5c@bt%DsfGnd
zPguBVvqBFLxziTa?%eA6A1uP#r;L88Mv+m9c2H0CT~nqksJc8{@7~fgiZAH7NrB`?
z@9Wzn<$t-?+OU2PFJ3sM%9I5b{Q%M~iO)Cg{?$$UwkU9Xp{ZpaX}U+zU$g^zqW?kK
z1Hb4)JpGRx+DZH!U#xoXo2EO%yudH=u`?ag4-&CU;_n^Ou+ND>pNIRK_JHUCq6gyu
zz9$WDv3&RChr@g$7dhAscIOvH>jFDG)qde`9g{<hyPNhjY+YsJCX0Te|E<zjZaA5;
zK0IErbH+RECrmnagkM4qc1C^V-&X44){m}Prry4I-1}D+Uacm4xvS9K^Y@3xIr?B{
z<eM_IMN(ScuPyut_RlyXo_@fthP=ElW%37y!u^O{(ti4#{Ma>!ebYYL#W+P@<iV3f
z>=Su}sfY24J;MtKALIw=4;{jX9ZG#=m~rA1@&9fSUYsI)!aw5|ehAY~Abes+(k_ro
zywnG~!5`qS$=BlIYg1FdIb$KeTeB{6pDT4JJbyATfj5j=p1GuKQi%G{pYkB>q#v+L
z`s=|FyUzQf^)ZVu?V$hh$J9eU^agX3x-|WldYKmE0>nR~ANdKx>(~LU=TAPlH9X(I
zPx(9ZAIe*CXV}j$kH6aF;<jUk?GMjKr{C^e>-G+N73uUN>xsJy-`K2b|9uL70^$c4
zPt1ex%(^35zk0Tmue|L3hTnAkLme+zdHdj!I~4tb-+y`Ol0pY=*&XgD((inGU$!pS
z?o#9<p8f<WPk)l1da-XQk9h0}#P0Ar_-p1L!V;bKC+h-<j4K^7{&a|a>rf|m%dVRH
z2DMpjNrXRmki1C#na?sNwqDriq2p@K?&_^7AInhm4}A5^cjaYQex0cvSlF=Gw;vx-
z>@VOS|B8PC3twMi!;l4Y!}C7#I=s`rrGBrwc22t!s@~<5cU6Dsq{{u%syq8^Kdx?>
zxZ%z2mmX2WDv$o8P3@!V#Gti>UV9*EwnX%!9U$_M4<euRhf;<;^;EV~YE9m8yB~S=
zpsL!pZSMLb4y%2YN<2Lx_n|N^?3X4jn7i)4(-{iBnTPR%?QYC}(@ld9S|y)2{z}1O
zhr|0e_+<W|oa}%2!%q`t{R-dA3ycf=HGU9%sfT*04_=99zM);%5$Smv_v+fQe9jPd
zi9WQe*T#jVDs?(;p$~o-dqQ98XC9$mkaDyKWE?TSvChPfX(#(H(pk5&u44TNPprRa
z7xNnBSVxl|{TP3&3&_tr3LnTP9zRmGd9I|J?>%Y}X1xHf6*6mYKQU~th2G3#vfP>t
zwYN08;YfJDzVGCeHiN%AVqu5OgWrF%<biTi4_l;DFZGcRp5c%7(QhF8W9E7IV;s<K
z+QB}V{)ERpFSRbX?TG{7a{}~Q@?4ksMPJ!tu@9hMsh@QL_QiSvKgarjd4Mqc6~-6#
z3(|h<K<Z07$wxcTx9B%@OW*rKx`jQke!_m}KX^h8Vfur04g6rY@XokoJRG~c^QH4{
z8DkNT-69{o>5tD#P5RgHZYRR~bH*e2D2LwgKz*dsevtJj_DVhwp0FzreHq8h5A+lM
zgMP$6$$1X@zz=1w-{<J6HxGr^rS$6yN4}a+wbd>QyGI`NAqT&NU&Kz(5Bp<Y1L2eQ
z(x3NSG;DhI0tdt6`huGVJ)d_;rbRq{V)&_=AGSJi++uv++NE>vmw!EK;a56*d;j45
zdyj?3GrWWF#C#1Ov<sv?gvm!d^9b`j{9v!_D=0_!^IyJe**|ag5PV3)j_eS<>=1pt
zM(oiJ@jG_N_=YFeMa*OHK|Fk8uW}qR-tF}BI-T_*;|65Cg1=(CfwUXH!uXPizmNzY
z2LC$uj1>>eV-m4P9bymIHORUK{jm$iIerhszfc~;Z?b-3eq|oRpW-*)eEY>5uZ%x0
zL_GO77Wm+)q!Kwoz24e%=LIK{L&(KWLE`b}(to1|_07NO;%O`Mo(%hU);Gw*zVTb|
zLOSaxJEXis#vLf-;1}ozc$4Fi`JHrKztS%{WIcv_>Sw=L?eF^@<{$QV*aiGzFWmc}
zKl>xX$R`XBl&3v9#I9*CDCaZg#~~A%HvMu>dicDHcF<n@B<ofDDEUD2mPoop_(4Aq
zy+GE@I%GY?IupL=f7WUE0oFI{7qDma#vYMF82-_Nc?>zMXIUSzuAzTvpB=(G>oK<|
z=X3Tu*fHl=tdr<}_W!IKSuf(p=r8Pn_G71nCCYv$j6T>CdcrUKk)L*9hv-Q>Jfko6
zLzs0h`qST}7rFlS4X<BOD15$6zoQrSLs;VUpIY?4yY!jxJt}&m2mMa_;0?bCzr^EL
zh^IV$3j4-hNT)pZOZ^~x!6SAcd7&Sv5Bo<S5Z)O_gyDzs=tcXnYvP$l@N1N(pO8a8
zvLC#)>KikkY@HN>2kHlD7ktt#+C%-M!y94hM{oGU{*Vun4nMSqd0Vy%J>ZWp?UebE
z3m@=EKEkvMc_4NIKjfo4`qOUgm~{A~y)vG3&OzXb_D~MFG97u?D`EIYZ;*WKi{XcM
zQjRd|DEb*Ww3qsrFR`n8Zcq9)-`xko`*Ow)@@W@Hdub;KKlmN$$^84B3jNBwMSGZ6
zNN2tyOnX=_(H<#x^MLdpr?k%(V!wl5BR~BHat?-F;dihb{5SJ3?Su!$5$#4#;*ksD
z4;e44r-_$H|DYFk&bb-wqW?K(fIs4C4@fzXe3ZvdSO?NR<kByaN5(lwKFT##x4e>X
zLQ*KBYSGN=N972koTr(uLHNVp(C^qcJaH~byU0g>Vb2n2AL~ox(I1qDKhEE&hw|vd
zxTXE%1Bs_U;hTOyZ;<qAf3L?GpAz8*d}VFd*Tyd`5<(915XksMKJoC)_@rI%fPC6R
zJ3#6KsYfDmIlq_4c(X(F(ILEoj8pss{VvZ*IXC8<2!04-57;~P(@*R#u_q8eMn2?V
z7vz(OUBMguf&CFi9_8sr;weu*!!PORMLOj{$s_x3>;XNnE6RcJgdTD{qBneyPJYrs
z;z8EmApGbMzX!rINc|vufy{Rx?F1PIAmwyOJM55tf^Qw7mk#N7kb03ve?^MSqu2*X
zJ=hcWLcQ=qI{iUE5)aSV3-PoEp0F3jKkZ>WGmaVegc+}l2mAo%AIPJA(m~n_q7V6*
zXQ+>JE$Zbx6Zu#Vl1@D99n$e@#N&@Z?z2G3;m6^H^0bFA`ol9wzk;+El>I|F^fma`
zyK}>X;o;BU9~e7`%nq0rxK99?UqJRBP7!|vGB1Fv6C~nSCF0+8h#%D<?X*MYKZ(pY
zcF6oB%ggr2Fudr{&c5@Ee7b(OQKu}4?2DPlb;v&3DN21<Z#YHP10Z~H55>72_gfpA
zzg4rq+b6AyXKh(Jv;67syJhyt+><fibI&sTv&L(icFPuG{zne;IQ7tO^dX&d2zX<@
z2jLgJneU0u{QF#o^#JxqJ=hcJl2;jKKg&9m@yR(Z>qE}fs85H;ljBf^S!dEdiRh!l
zg}*O3|E-VCgx}$E&S;0|A?uT2)?FZe2c#cC`bVOaOPGGrq1?9<hBt}2{vaNEb&HH6
z9r8|8BIAbf%RT}-MIQRWEB=b}O5|`}NjdJvcqhVn7wPy-_$D1YX5Y!V9`8>Gvu{9e
z<|*`LJ&nC%cf>Pp*oR@q)X#Y?;~u*to$|~-GK_xQ7t&th(UbDjPdtddvQHqMenU^>
zq6hrJH{*o*Xb*Nu|KeXrmwLfB^^lGn^r0T&d2dCSeCSO){e)fMPia5%DE+|uF{v--
zq_jsO>Bym7<fDD06EDN`1NxH>UhvcCkNr^}zb}9f<Wr7x;<0D!fO7E2IG{X;e)J3Z
z&=)!Ih<>s>{eoT5F6@VR{0QwOO#dN=d60696A-)2=#un(y%MLwzpG__W_%zI|3p31
zi+<FHeA+F`aqdbv?16fz55KQN%IgsQ{X*<BQq<{9Jp4wA^pg(h7ad~HPLXi{V(-`i
z;}-kpcS4Li{1eDMA@@Rz7m#@uxzxkE48}G2`TYjxwAdZvi20Op$2`mUp&w*^=2zxX
zkaI59!;%;DhbP+2?^W=l$j8pgubNrrP-^mQF!S$qG3{hs#qU~3*WvWW1-`ag<_&@H
z0`fk<4jHHT0s4)31UrOJi8?zZ9>1VNc+?^OPKTVYN~GT-MSf3*-ss1>Akt+R`+u#^
zmY=#jlCJ1q#tZX|j3=FV$|GNn3+loCK*kaC82yVK(a+MqvaXZsZTg$>Q|<5jMR<W<
z^uu20f671mY}>|N&e>wsK3S>C!rmF-?^3W{Mh^1{<>mbwa>&oTh2Gc!=hxD1*iVuV
zgeN;>e9%trx6y}n2<vj}hH;6%f^YgmBKv9C4#FV)$bOx8<|7dOkPE^G_96Kqp7km`
z(|*#CPx)P=nq58O!5ksN>^nK1;e3L7L4F^BJc;b5Bw|0jYixK`h4#w|pBqA-t9L(B
zp~cZcArSpRerE+%``f-j_$2@PSB|WfYi*toIDdb&UGoNH4}sVd{lUJUeHG)9`I&cq
z*ggD!v;%qA5%$ty&#Vzgs%0wP^)PQS?=U{3JjOHg1LK)^><XlQ#xr_=oGWpjgneUQ
z@JG9u{|TcfJW>w(r5_kq_yK+&g<gcQH+aIH$OqDH>O)TuIUw;M{RYAxi2fkFfYc8%
z59koSS&y+kfk*6%e3_G5HQswTUx@m#1M;)}VjM}tj^KxV74k`kAI3X+T>Wd)`wve#
zZNUfamdN}CuZ%zJk8@6m=tF&s7x-bl#=epEN@U)R7Fi#I^dtSmx=<qeGcG{p8;SIv
zTZC5~%6=hCe)y(d<}>soA4osY|BO2j`yrjM-8?MQGyh&sgZN?Wk@~O)kl)GEPV5}y
zcU&Oxj1$V~kZ}#77l_{g@q_qr!jxxSgkS4)U}m{@7UmAmv$R)^N5(&XnsLRx1HW+f
zfYkmwuE`Z5pOnLVg`Y<+9ip!t(te59ndFys6^Nc7_6H&dWIY4I8)&DmT{?2?5IfKz
z?U9IG=up=mGM;gS-vn6~fW%A0j_gp@%X|)Ehmj(3B(g5FLmiL0cxjI^On)+;f$*fW
z3x4m<`WpMiKQfPi?7Q&G<bwzN9(Ir2k{>_9Jt6xS=6B{v@{`U!7UcUJ_<PFZckt_s
zSLQ$b40^HOWq*L))KB~97s|6fg<sAu$OmFy=!2ew(HH*t?gl*2e(FIkh`zKFyJtR^
z^~wGvztn?q2T$yq=s(iY1G(ty6q(m_h`rGs9a2C224XL?liv%%FY@O-(S7EIM#t2K
zzK7}*TYXIN`=UZ$)PMKu%}KMH6dPIf@y^+1v%ebk@xGiF4Lz-Ll^?jW)X1dSS3TBg
z+kj7!!$`gxkNj)=6$g@MUwB*g=J#HfJe%*&fSuar&m7<9lp+jb*Q^`Rm-PetJM`u}
z5WNV~Uh;vo6FsSy@8?hs<a;};8{viW@Ql9bK|0^fp`NSHnbGp(!4u){=x~n7{*rYp
z{X==`VZ2FXelchmFbsId0P73pDdr2#KO}lrvyl`823Vi4F5sM+^K{lbtVdWk5N3VJ
z{KEWAm~{?vSYI%&vrfY<S)Z{kpgc%E>=}K~A7q`)x|VZK!mN*wLwj_H-sGn|?EsNa
zKB+hBFv{yNQEUf3`8^!{LO-*9mGSf|@gVW=1;Q8g(XX7NkWQHTk;}a-^5{qMvHqlg
zup`nX(k|qnH}!$k$9e`msSmk?LDq?cIoD*}1ai*EcmmOn^(FC?LtpBr-Q=e{?V~&>
z$2IZj3!)!k@*x+bUe<xE8)-M?$VV8Ye+g3`NPB26NdFLK9K#R#(O&!%@lFwYf@j{Z
zF>a`z-~Vzyf!@4Zz%I~-c-oJhp+D!U#7ljV1Hu<be)@%bVR)h(dJ>PG=s`Z@!Xx?U
zSH=x~k^GE%>Y+UK!5{q(-_%cj+C#q+FT?1IT;w1hWITcRAM94{cVw9MA&>g76V6F9
z|2~(N?Ij+6$hidPLd+lVC;6rx^hG{l^aDw!9rPFagY*;qK>xxk^5|#wq3DO-KtB6o
z+JzkQ^SwXz-N<FVfIRYJ&&UVKhaBF~&@On9$b2mOpZOg70_hKsc>$E?_MFq;H|THL
z$M^p^f9GDF^DovrtmpXsG3B|Jz#q}iAmfH{0KfDD{Y$^XGxt}rJbWXcb`uZZlw(}b
zZuEdZ&P6$&#ZKU#Fv#BvV0<xd*bl+K{2n9u`Hnm9wehd$DZ|7&McHo3!52vT8RzK3
zJsR<hNBV_%f$vh#UWw>IdFC7X6+P&8`h|GrG5i4Wa=s#r9|f^5koI!F!+2v{;m^s3
z-7_8tV_%$;GcG{REwLBu!zog~MBcxmC-#mX2GI|MUl9L8zrYXug<R~0a_}$1<fA@0
zzKN$FsTX<FOFVW)Ipk9>av1m6DdQX7h|m0c{$rdI&-e!655!KK_8azueCkCWh~LES
zXczUJQj5M`wIxUR_foVIdC12euz&g$J!lv2RIpF<z)taJgylV-4C5!IzvFj-j2G_h
zxCdf>VI0vO`i1$1_8}kppd9(RFS0}IjClq>#CjQfr5^kc$UPB$55z9{T?hH`-^_=|
z1+j0=-SJl-`xo9{66PG8FzXuDBa|ne@l@^a_hgJ4`XB#II`7&TckqM0)X%$5;;D!A
z59@s1yAaQJFZsR>=Yz}}^gBEfrvHe?P8sLu%Xo5%*q20j(jh#9^auTbKLvR|&bpUz
zN`8=j<Gw>8JTi{4C)$r(^ujOTU&u#)aj!@{q$3ymhIf$jF8F60gZLkbtaByujzc2-
zd`bE(En4KrAHqM}_3rRzs()WN1mfpF-lKrD2ZSGx@lStY_wbB;pbz7Rcl!7@5dVZf
z#~)t2Ao=~vp8mh@8{*d`@;llVOX?55|K{AGJQD_#*t@%6DF4#st1s(vP8jjGI{bQR
zvmzN^6bR{X*U`ro6)jybB$57MeB&?S2R}+Wc86VIN4!s$c8Fg<4)=J-V|_^f;7>s6
zV>}Vhy*+ZMpZuiL4%U~1>38%aOuHBl$Y)$3AH+}6KE^e?P!9cRrwrqlsUN#zTw%AA
z$6w(mK-$N61+iD`9K_Bq?vR;t(xjXr>XUwz@s6J+A3RYn?LdEc0I3gT+<=U09r7I<
zkog+KZ-Mv$9Wvj5tSdoTj&;67)&=lEIqU@g%X%Jr!9K_jk`JUmkPpAoKCuJ%h6nWE
zoC13%o^eQd@-dDWC)gkLvOXXm@@OBtVi)Mcx{CEG_JsYTAM0N9V_nEPm3G1#cEi0B
z<DUB{<gv~to_ZNq=!tyx$;7iR#%|?$3O~<2n{h*bGA>y^q9^vic!EFlfPeDQZv2A|
zsSlnR&oZ9&VQ*4z{08+B#vVcBg2+KGNPXCcEYCPZUs(=5(Fb|>JHqftJ@CnRLOymt
zzfvD|L3zs2F7&26$oxt>=pWAK+#>R*pK-$YWL#1|c0xYpNq9kje#c0=(U<nKe;}Um
zgCCUVj>yH&lb`cM)>-TmS?}??7|vUe&$^2JA9hDQtc&PJ+Cje%W}gZopM5%V(3g1D
z!RSFe^F8Z8cn9Gb`@>!dgYXYxZ}<U;*nvd+4)wZ4^mB{YHRF(W;MW*$_%FiPEAs~R
zvQJ_@!VV=e4rn)sJP`dPqK6&Q9}@AWI+XP}4PzJ4qSRN0;VV*<{UO8jBkKX|7=Oh2
zfqrJ(FfZb7S>F+lT@z;hg<s@gkJ1j&gLMn*5$uWjsE7F<`HVZll%v1Ui}LtS+Jl`_
z9{+?NWnBcHtZQ0Sy8GH?xr&6pzmET7z9T>HEYSxZ;F<c#Pd|}PIs7N<F2)o3V)wKY
z`=XuTPaR&Z{ld<?A&`1?m??f=qYLZCE&XMOoQLqP7XQRK59fC5^GFA=Bi8rG!`|_4
zgy9c)Ap1e)6Y2-aN0@x{7l<4kvL28KkLU-=aYTQDvR|+h5c>pKKY+~RP7%2v>kE*1
z8N@C*pG9x%f^h&cPO*E&0dg4Uj9ZZL12PWamGK86m;A_wC-`8Tk&d1crT@mCfz*dR
z5)UuP$1XU>L4WDb;1NVVc1Sz1BkYlMc%+@|rx+)UE82;C^g&<Z@k6Ae7ySt0Z{V5s
zbM8fdg49p@S=V4cAo1i!KjeV43w{`1=nenmBTPP!_L2@Vu8@yk!e7BJ{R<D+1?Axt
zJ!lW{*c*HjhG%${`#;h_%7ct+#xM0TzFDthH{>HsI`T;8-2?j$;z8E?+}ooEJQAiK
z;fr+ogZ_s%>=6Dy`~&kn{lNJIywMK)0P6zyLq7Wt`j7ndiwsj9eOZsuF6I~RrPw!8
z4j!;8>=(a6y|kBf%E24;(@y#UK9NKJ5(bHf2jY>3eC&e#8+JuH{s?^;Pw0vMAm!*M
zS)Ovpqa5Q`hOr0u#on+B!n7~*@AVPqSIk@35B9)082!(A6zgTew1e{aGwQ*v&=0;r
z*74X6{meQVzGx@>Ql5Btq#k&LH{^iSBjvz5h&=LRAMnZk0K1U(M?R2t(hmG7{fa!!
zh0!0sKzZ3d@=2r~$}wNT7xf~Kc-qT6%sCNZ=4IlUAE*bpGK_r2JN3XH{SIPJ$OFmm
z6tNQ>(jIumufYq5zO+l$k3C^0Amb544#+rjipaG?>I2aaBwnJlBN>J##*-al|2ow5
z3x3)Tndjx-GiIJ<{$rkId^4{jpZxeW<kEinn{mav1j@7DCX7D?NtejI1p3H+g*WtK
zULs8YQ7`#3|MvIjPdxeYZ}5*i>_g^fK9T*69TU$u2boXskH}#i!+L;zp?;8cA?eh|
zJjQyOc=9v9(Qly6|G*3TEA}1uH~4^8+5?ZoGw(4@kV|>^MlNCe407o=<b#|eksrT^
zUxEkLU&PaX^3xB5sgHJ34x~QLztEp_<_Fp>5&6iWybjS<hsei%;2%AxkMT;Fa>S#*
zv@h%#ePw?7jq;%66S=fQ=9Br+6J#7wP9o!kc9Ea*$fthnf_TcICw2xRhje%$9{%A2
zp2!D()K5O-ksqZ0n3pM!zSN7|Fkhnw?PnZNAMKIjll2yU7d?@Sz2jHVgM8FSJo?cu
zAnlV#e-e)z;%Nu#L)LNhBmBS{NWXFqF69!Bze7*z#ShRA#G?=GB#a&)^}`QHe)@xc
zq#Q`U(N6Nobu;ay|LHf$7wt!H>Y-imFA@3Vr#$kA$DZMl_QC`4w4d@IdeCm_VZFfo
zOg+dYA9|w?h&&K^Aa>6^EAfmIzIVrb!FopO!~Dj4N4jixrr3w_eH6;U2RzX(#vAR&
ze$b0^0Q4rFbszV)+{?1wBFuN-Xg}r1$L}>rr#{LvZ?g^~4DaX<FRc4O^kv;Ck@XIC
z1>(Qp4a6?61L`N8^BUgAabG|=_F?Sv7zZHoIOn9Dl*c|n>c!6KSH>+!I{C35^r79@
z0Vvbak9#icj^87(-j~R{<rdLHhqOb7@Fx*F2eAY8$1=Z8K6VT;UUg`<A7vakMai?Y
zZ>M429fFJ>#v^hVe~d@B$UcF42F5@8KKu;*fxUs)5B3UT*NjWX3-S0p;z800gUAJ`
zU&^7M(F^^tSMpOHJ0c$W*c<W)Gmnsu@-iR!*`J{Y^?<B%@CVHQ^b3go_!G**7y3{y
z_KzQC-hmg8{K#iKU~hz(5Aol$oAFBf;YY^Hdgxc$VTb7J7WrMQ4t4di9@HWI<u@`9
zF>kP5C7yL4>qU@phknFkPxJ?N!~BImffwXpXYfxs!pOtkun)>J-Xvmo(w|X3_3*uZ
z<iP{u0Xt=!!58yCdg7n)lhlLWl&3w|DLi85@Wl7eSa%_p@j^c_Zr~C9;DdHDPN|P{
z>LnlglFt1N>ond&5avFEcZGAmF8y@wSr>-Lhn@2~aDJx&KiCcPVCLWJTll2?+$Zuo
z7XEG;-}ij!O!+sve3=}olr(w4?ddr~%<u4w9by;UyU@SPQ}h$RKPmc6-O~5IkZ$3(
z_`59B%lk-vZ^e3_b-Bzh^(Q~SN8x;ubr-+uB8)!R1NFc^zkde#dxwv0nx5;HE0aU;
z339&6Jk0t|j(>QAFL>g7621A(JN-dD*cEc<XZYdo-of+yn=YQVGVe)?cB3clK~H{P
zNjmKy%(*sl(1US^9=r>NSL}g(4RU!WOPFzkJ+p7de$kKfTHf38dkW?=`U`u-&ruKK
zl<`LU(3kxd>Bz;78DIEA>_xu6Wgo!4i2lakv7h1mgLMUdka31TB`nV&829*F^5HM=
zXCUeD1CoxvCLdw?hkTsZP!E2TFzM(8axYJQ{46}M59IfV$l-k&>x9qi7rtifk4F{t
z!5{vT`apgc0-xjush57CpQs;w*`JaRen9Mnbmj}}kb44nm&ka*4<HY_WBjqtU>snN
z_(l8$<v3Si-AQ@I4|4H4)X)5c{ZdZqg<R?frQao<`ImZ`9~n=`(;@uQFW3?5DD0H}
zW*%ld;y;OJo}-^Y?m41G#t-MWAby|v7|)D9>;pfGz98)*AM)^%*fI9ax`puuPmBYQ
z`r(T({4%exKO&5L^g%xQg6Pfo1Q~CPYvkcy`MoUTobSa_5Bf)ow1@W7PUa=%Dfpnh
z@CxtnhTkS1ejdMp{SwB{fY=A)n)bmLd{7U5hjb8L(GSG0(hq$9miNWH$Kzcw-@!#L
z<p?ugm~WY9@h6Ob(h0*e>v7rvkF1wie=v^Whkm1furvCRFn*8vIR`)w_A`7Z2|k*A
zbkD);GqcT}TK>X9dsgR|O_+QV;g#=x^ZjA^pLG@e$~uK~+QGOX9zL)q+J_&*U+`Xv
zb~3-<ckvgjvsee=??L2(ltT~th5o>fv1{f%>S5i9-)CN7{m*>HIT!tfUqcV<gLLFG
ze~}M+VZ5R*ewlWoH|>CTc;uWJdtsfZvpf7b>GTuws0TigiyhK$65)w?m~lgYNQ94e
zH|D?Trojip`!M_n{lU7GeI2|ZAAaz+tgEDc*dODPenDUOr~l~h%)j@o;M=8#7X7kT
zwh(`BgY^L8it&s-$Y=e?cwrvl{EPDNPdwv|am6^n?_tN(i=6R;TQzK$mSZ-EAB0cr
z3_GAa_JIEM8}@)+q%)rI$JjG`BcJ(zeCUb)r+o)<mc3(Jv834#r3`!OscfedJfR2c
zEcC@5Sa&f#kcXXM7t&s62mT$s8CTdf{f2#_H}fm=4f@gFyypa2_mB_%xKF3OglRYJ
zq2IA9`V0GIeU1MBv3t&;7|-O#ui^g~AC!l0<^}fi%&Uot%wJzjD0$ac<%)(_w_(rt
zZ^i@jQJ%)Vx^^s|GyM0rurtcx5AYML1G#@;oZ`1Y>>oc&KF*1l9~h_DFM6Q|c7&Z`
zC)g+Bkbb0{%n$f2;<111i1Q8X27gL9;%Ps~_yX~#@QWR@-@rexPJ|cQLpt(k7vq+6
z#sTF(#u4q(*(v2|Cw>=yz<9&|;ivHn@IiU<Q=W2!$uH#)2JvTvS^pppBp*nA^dlaB
zgTG>4z<**__$Ta%{P;cOkj_2?KS6(^2lB~J|In}WC+XCKe(XD#H}HeGe_C~CpY6vL
z$nP~l{(c7g3gk)s@w?PRKKLL%cFMS<y|jmP<Piq3SI(ViKmCK<VUO?*FX#cnJLgg8
z0YC7CeEbjXqa6EI>VdaD{knABQt-5*KK>31?IWIXNxw19;fHo&2aE&wCO_%qr#+m5
zGLF#`e!N1)6M7(*^$UE#D+mvygXm9tke~5->9?n}Dj1?2$VVQ0P>%jWALb{@kxu`j
z56C$<$hw~XK`-WekaYrxpN|yrkH`nv59pA72br%x)-i4oJG4XUwL|<q?Zp4$2eAkG
z9p0HI>2Lalaf>|el~@PJ`dHt=E69EpemVC<F7+dqb|R1d$8Km3?L$v^q+ievd!}8q
zll2Soh^Kz$5#*yE^-xYCew2EdU&u!~<B59kU)0a|A)WTpZ}10C%<Je&z34$a`Iyh)
z3HyKt@)J)O{ooV%=u5lU_mGah=*9Yo_Q5OZ)QdjI#sA|cX%BV;PxyDr;ir&?|6+cI
zf6_tPg<jMPf8?WoX%BpX#MA%8qc`%=gLa}f>muqSAMGO^IoK=pqBp#<-bD}k3I1pg
zaw*4nhIh)7PbUYzLppXu`=lKBg@4Mk{-Hk7?GWBU<bm);JCRSA`3-x)p0Oj^0dk)Y
za<9+*I`0iwU)Uk~=`i!}^CINXPwcBnCyX8-{m#A+L=N_ZKc@eYi@&E|C9=<VitHb0
zk3{AZ9lD*5$&X(HnNL9c6i7Mj7kxOd0P)ixyn*<2<Y8Zg8ISM|zmz9GdeWb)m*_|O
zk9hn)a?ykG^gDXf&(w$i!p|^XC=V~>BTRn$0saZzS!YVUXdk?w7d(KhgBdUA!+wo@
z2J2qtEBpxiFXj>UO~m6L@E_#k{Ed5g{4;(5In;w3{1Sc{9`SdCnKzjqs2^mWpd9@P
zzwpla6#c=u5B)(r{ZBujCwxk~<a`DH&G{4cvcF{gMLoztF8VOeI4{C(B~QpjKk6eN
zh+Oh>j;-T^{WH7}Mt}MN-brVF%>I;m=?BWupWF*_uFHO%c2GZZv0K^!fA9tB&Yj?$
zeAo%&1$(1CAUvWs`e4WCqw5Fig+J<(dY}(>hh1<^iG7g|en8o7>V;3<hw%=OcUa6b
z%#)nMb1#H{!T&I?Ghg9f@wdd|FG1{>cF<0EpuO}1NI$?g{*w05eh|5o$4;<6`U|=A
zC;ft5Qy#>w$&VhS!!z{|4>BJ}|AM~g2_M8$4t_xRpkDYOo^>1TLN5N9d4zUwPfZv;
zbx42ekaYowe)vb)!TJch)*=21ghvoN19ftujfY3p$M_}oC-@1F@k@K*fqucR@E^=$
z<j0=yAM^+D*faMQ^au9`j5p*HMn3Z>@~}5}q8#=q<#IoZy>efLKgVC_Fs)NstJLx-
zMN-aBN$#E2BrPS|pZ}&7NG+UNG$pBLT59n>!xdAKQ<59bE!itAbyS}^Wfn_sXTpX7
z!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+0mFb{z%XDK
zFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+
z0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn8
z7%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXj
zh5^HXVZbn87%&W+O$=n5$`~@>{oPj3@NDWM(^kWPVZbn87%&VN1`Gp+0mFb{z%XDK
zFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+
z0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn8
z7%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!@$|bKw77?R)Z!FPyTniQ_H8EpHeg>xp!KV
zw3KXr{+n7LwQy?Dl%$?%sm1>cS4>GxNp3i|WUsW;QGM!^SsXoG*Im`L(3|bg2_-5%
zUM;2XS6vTSZn4ejhdbO^?6?J9+xo`K53foN)%dCI$nVDFN;pJ6+PVE-MG7qJ{)37X
zkq07o(7Q_qbRE9lHJ^)5_Z|JgReLQRE*knul{Kw*S|Ih<A^8&ppR4la@?z;56?oGl
z8I3L-uuCN>M(+WP1jE4D#{mA&U}W>FTi)b65EP3`RKK4ndIhB)jh?||;ORSleE#9b
z_9}yh2PbwA>3+$Si*CQ^6@PVZ-aK*b1?!a8=$H5Ns@+mfx$O1&yt{6;<u`iGvs-zx
z-7;*q&XoE64!doiF8_*d>%PBm&mm>-KY8$LoEo`?0mDGzGa%0if*#JQ{ykQfwEDj0
z+v>cuM+F^Y#aFEBiTC)7b$jjV<G#({U-zws2gAdk_wL3Hj2*bK1GoEl-g&r1JAZD{
z|2G3!ZGHQ{TdPkE`ua3c_180=wNswz_W8gc7jIDp4G+c+VmA)_zHc$|V{@+s&slqx
zFs(BT7zTpOfZhGJNsrC@RKv63Ig;NQ8M{dgy9x3f@Tcy&-{0BUm)mnvr}@o$Tl3!5
z$$ML4C&o?^$xebk51eiK=Y9U73Oi?PQwIOyLHAxbtHz&g_U^R3tOw<Nw%vOxr|Ubv
z^6|TQ@VxuZ={Ee9Rm$N1=E135n0%2jfIS-g&+l6pJ`5lK|L>h7nmt4|+D)mf#(>>9
znAhpi@|V@##hLc}=kF8^AKvg0?Km^#{b9h^w>S1}c#IE^@|=Y8lte+l&R>k&|NM@L
zKi<(uH4GRA3<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjhJko9kZ9-OvEoDb
zy)F4pB3Asws@<j@!+>GHFkl!k3>XIV48&^QH}(AIdj!MB+0BR7@4@sV+@u%=3<HJ%
z!$2$;knc_1h7;xQd3vQsEI2oH7zW%j;C5U_nm^ILXW^Eopyl^!UG4Y&jPKIM3h76`
z+hydR-TPY8|EB*JcK<<TRXgx%eu=l7pnR_{-aMK5<MX>s$vf+pXi>_ISJ>;{pVs)P
z?#S=P<O)TLy7wYFeWUFUSw3F<k!bZwUhKk&roZ3%*zK!>`kll?|2t1kc@4_?0zK{Q
z$)qR7_gQuGkS;!}>_eQ6yGZjH`%Vn|rXM1;Th<AF&%cqLhuf9IAA5y<@!*wyUd!>T
zy+&?g?1OdvfZuS6Iyte9_uBpwMV`}sF!}!Xyc@6njCH?6I^Uaef8Dou^<E`XIkCFO
zwW~+BzA|<a-MyTSSK?#E@1mivR9Vw{r=`PK>5-`Qu>Ub=7%&VN1`Gp+0VfP_Z_52A
z^Q-*3OS&-Y039-4M!LSNwcym80}uY8-XD@<*sn#>m6JA^Jca?ofMLKeU>GnA1e*cY
z4+f(<$8pPFP_0I{4z-JS%a6$)TYI$QTi3txdtdPi2X#Cdy$u6~fwPAJ-96jLn!8?p
zs(ap$#3vqoa%-tAr!9lQ;DPhKpkTb_Yg2DZ-><qJuo8ubv&RFQmKp{Uj{$yXkSPAZ
zD?N1kMz`^H`%}NtqkXp;uXlz)tyhjrX(xX5bG+p;{tY%Q^k(~WLUuej^-sugnqQYs
z+NCZWt-nar^1Aj0EgpN2e#L~B4(K|3z2%mNv`%TQQp=|lO(~X=+&isFT1vJ*|4l8B
zS~#_6N>b0X)Z%}JE2bo;BsZK}vR7K_s6KVdEdHm}Y=1o8`j!WGD?9vWNpEASx2oD_
zZ=TIchaZmrX8V+%r>I23dG*?k8q#u?3M!UJPpvh1;9eC}TtB7q{8CTuRzb&cm41F?
z@uD5dZ}fK5x|fzC*LR(dja+R76<^PHMa7&ox2tH;J#(f+EExE_`S)vnYq#8r7LPXm
zy2nqazfexGPMs>BS3WYu(&1gNEuK4l*cN4n^PV0!__j^oS#Gg@=8o;%S}j*#_5KxR
zoxFRQa*B52`RE60C-+_WopOqmE?e2S`z!BSAbr2^i)RM!X`il~;_;;mvh6&xNBNED
z7ca~)Wpv67%QvV*$9VA*uX^{_-BNDay!A@r&ViNoKls2*3*3Kep&d<*es6&#I=njS
zm5<k0;D}M%dJIcjt-#CI)SWovm!%4n{odl?kIv2e)Cwh$^)qsuqTl>Uej*J=J1<9C
zenh9ejr!Hyp@ND-d%XJb=H`bLxa3s*sq1#^QbEBVa?hx@_nAY=Emq%Jyke;Xa}~(^
zCy{ZZLz!QOM~t36=VU9T+~Ta<-%hCA{xwU7WnZ{2X~@>^RifhhjhCFf;_G#)<4>)t
zzxC2eTa^CFE`0f<M$b;@_M5Up=^tfy-h!sJeysbm0weWzIeJehR=L;jN{5lkrJU3|
zsNp&XHqH6@=qBYB?d%~|>9d~b*KxqJi>+88{SUew?_QrPyk7mX=0#s$rNEA5?-})3
z(H|8!;_!-nGpj6AA0~C}e@*pCVJurNb?|eyFID9}s@LRTi9b|{iyrPcd(%d>WZVO{
zf8Ah*HE#L(W0S7lX329vo&SK3{Uet3U(s^00$=ZQOP>`@ms%1Bwz+6og$i3OiLEc1
zas0*4a)qMByYj#AY4aoJhIA;`2k4(@_{}^0`rWl|r^+gby#*a3T^Gl@T*vpHOnRlt
zHsv+mmM62(nRZ7k9k#vsxsKy2@3ZW1#KkAKKQU>m1?ISF#wU6A?y~SpcIfoHh5eom
z<K+h$yf}aKnXC3IzcAMA^4iV@s{A+B`)<366~>w`xAoQkdEew`if&PEQP$ITPX861
zZ(eA{3T6FXg(F>u_qeV4{I>@lv%oKZd8hK`t4>-@v19!%gC9!UZ0(<StVj1vJ8kjV
zqj&fIeeym_hiBT}dQ*o+$1EKd?Ywwak2*O*I^169f~Q0G@3ibtw~v<b_*LDxxQy?5
zrrE^N7i_U~Xt!RM>rP#`(c7)c<e!qRoT6O!$#9!NBf4F9-eyaOIz7CKce>8|{MU~=
z7rJYwlGw3y&xh}*yTsDr;d7Ub?s)GR%MM?vR{GqH8QDV7;-<sb_i0%r-I93K$Ql)2
zXnEMu;hY{FYhC))NlS-Kr{}-6cxvvD9bVYwP>%&a=ML$x{|Dn+z1jDm1u~B9u*TQL
z3*O)FcT0z>uNzP|<D%q{4)d3Mu-xnQ^M&m2=`Pl>ciUwP+2QH)*Uqi<U553-4ZCm1
z)72jB{DRZ;-*$9<xJ!wGA#g?LfoDsNKVUh<od*Y%eesH%Av>%Xn)}AfsVA%=t8%nH
z^5AhR()ogO{h(o+-}~>aIex413M1t&()rcza*?*jslTl^VfKXnO_M@8)Xn3>%kvMy
zb@rXtvu))cEvJ|?xm%;`OFvL<G1C16=Qvu_+^gC3hdhspG;CLoE<Jtxv(>+ynlogF
zYu-AkuKPH5NFwKD5_eB5Ib!UzbjvNu^;4wb<1eipI<NIE%P+j=(^<s|FI{ixaR0F6
z#vKQ)vYaC4mf(nk**|}v<uXf$b#6JcuixBtmJZ#%7fBnO|I!}Kezqj)<_EbhmSN6Y
z?Qql84_|ue>ys>r^X7L;di;eImJaRC={fh+;rgrURQ#joN=t`wo#r;I^JARn_=N{s
z9Gkr3hKzrlC(H9?!gCfcDc!DMQb>oJ|9I-b>SL2b5*J+l$U~VG^M!PH-=p)o7kxBG
zNQaBMPrmxI!AT)IeBi?RXKMX^z_P=I{a$={UD*Sc9lm+u{PinJ9JdaA(Yn$V3y#|2
zjZKT+_|1;Pmc(+6@7uMk>mEyo?R(~}@Iz+0rNh^!-8gI8`&%qKe6`z6k6pZHqvaQd
ziWeAB^{ve+Ug-AzS?1@Ru<kyFc<yoR(C*zi>GLvPtloLvb}Lf6qF0$=x9$B^fp+)B
ztds0e`e7NCb{uWk&JP!UW$50|79F&lqTPJqb$b1Vs~e3y`Gc~<62qQf@z&AZmJVa}
z-d>&~JFVC4I!t$NBI9-IO0VKq-*Dv}-7k4n8H^1N_#?kiw=VK4-s%2O?q6aRp4ah_
z72_tYRaqUcuUX;EPH!m74xRQ_wE0S2Fra#iLi4OdL7l&}Tc_&I=Vg4}vX|XCb>%S&
zY+Ye+zaP(JS`y_xOorDtyP^51Wjid1I{k^)o|}kwGBTf6_X&Q>nK1p{i92&=3k40m
zI!8?uIe!1H!fW~WR2?_H&R6?YP_ahAJTJa***@hK<$Ybe!->K_{xV*8-i>`KKbe28
z3M$S{ey7{GT?bUGkbcs=uO>d|zXz`K!_NNSe7sNDq0{%y^8KIO*UE6$yBnTAXWBNU
zL#OLOyWcC?ooC2&`rj$a^9#Fu&lj`KEBC?Lp-P9k`vKzZ?mz6_5xPxhJuUgrz4vzN
z_nrD@r@zCLc2Kjz*rPo^JfI}zS@_kOc6ASi&(q!VL4LdQCD!FSZ1~xZr<QHZ9&(F~
zzg_i6or6a#r)YN`txK2v?RVJjcffwtZ}%QXrr&k>uMZ4L{>{=M??rUDc-=*x9KJMn
z$SGFcIrWo9Yx9I8%6%ST_C>n&yPbd1rME2n>4^7R6bb3De4kLErkl?TfsdE|x%r4r
zc|wumwt^#DUC?@~k|^&BWZ3P!p3J|zXL_rL9y~v!!_8w0zxYS})0V_%7d3CP{*jHA
z4t4j+GX9bkPnAFKhJqm-s^_}bpEf3M2<)}BWZ%A1i-hd(=)^A`ZTEc1kPdS{F?wa6
zJI)Q|DpI}q&9iga;(0rU&TjC`ZYxs!<Jiodv%bs|lK8;mYj@PBlq;k|?ss)|>h$*)
zk)DUzy<gya0y=D4qHC?88#h=I`?f0j+0)<dw?Mmhq`LHG3obeEQ|{9iIBb07@Aeis
zCuE0q?+tb76?Wfy{<Z54SrVCN?NIvHNW=V|!fQ<1d$RurM}D;=`ZXWv<hnhl;2ss!
z$&FR~6%ST??AlR(sGy<K-|M$-^5*O{4UVZtPn4}Ua8|}YP~O|xg?0B8y5DINFXs~(
zmh-6$>--k+SFE$<=lp7|rNb82e!Avp*{w>4=&3{9e8PSMdzQ$%06Vn%J`Cf*PL9{<
zLHQ0sy!5m?r{JBP4zVAnDEE^xY<K?S^!=^fz9-UjyZW8}&dh23(Vj0l?H8HPZXd4m
zqcYyEKb@xQ=8J)Qay)(Cd8-x3yXDi?(qq@3cSz~5{p93kD{ncdBtCFkoogChm!WRm
z&@KOh%hThIo61z1)#0TK6;!nI=hE(U`)j*+<`?juwO^!Ma`7LQ9qP_$h}X@>x_G<q
zJIM8}E?nxG=aUXLJ!ILT?%gpwxJB&WE7a{9<2}Bd`c>{FbSTf~;}z!n9MPg%dvTlJ
z?*230cA?YH?w-u4Jvm(;M4C^o-~A5T^*`TB(EYwZ#@n3(k<L3{xAx9=qU=!T$K`w#
zX&C*zMt;{X`$cyjXgA(i_se!ZIO5Y!2F*LF?2zviesj~Xve$&RTd&-+;F4YAcUd_m
zFFAK@jvuU-SKKuCqp6d_zfWV`nYQQh#y3~|(vt7}<oa9MH}{UZ`_<3$e(*`z(QE$i
zd#Qhaznp%cy`X$&CFO?d<X&cNSf(V(^}U^)^1U?>eL(rU1Lhf>pOCx}mg7zC+o=z9
zd!DQN-KOr`LdHw|yb9~i`QXJ4b>l}DPe1BV_NSeF$nht~0b$9f4D0TFb?0kt<IzKh
z{4PW{&&KK;)2_em(w)xtGN0XdrSI-mfBE{ic36HP>lk@|?=<Z6cNFB4eo)%649juB
zeqPRJeud?`6R*Nv>&raJJYa{@8%%BgXsZKCqMXlkVcoqL@$8qKqP*9#3+sMQB;!AS
zedf98Gqzcv{GF~Y%<n-Z>c%U6D*ue}?GN0zMCnlX9?34Ab&5p#T_U{m-5njm^XJbb
z@4dTZh9!~Txgigf@8o4zp3mv#8F@a)@0M@a+ImMuxeqP3xFfaK?p{0puyiQ>OQd1e
zgK|BhJAbi@m*?IxeC0FWZ>v-KfU?8=7wy?M_Juu4BJ->r>iS6+uUpR$@AlkCr=RZJ
z$Syw5;mPNANjs!KT|0=^twRz$UU#k?lz6A#-*TEyu5+D+b^9lsKiAo-jMvTgy8AXc
zew~Jwyxi*iiDy=XzkeyemoRSp*102gEVCr?&QAI>`Thrg$+?2OUzYQ!uH8Y4m+SUe
zg?VQqzYD1Qeu<24^w<->pLq0u^~LiQro1=ya2R?2WQPqdU0q|-ki8aI?eN(5)xHx}
zou@L(4KK9!A1L>&x_)LomS`y7&&MjP<4MM||M41SJ#NEx@8wOpzu#4I%CFmgwaf2x
zUOij#ZQ0#-?x>B0LOQhjovYn@HK)JV(dN^1DCh1v)O{E7`dcg1Yp^t1DBc+9_hb^K
z9QPtYL(T*ILe`nFMtLp~ZP@PnNp8~_#+*CZq3&LSb6Godd+!rze%dA1k%Z+rmtC0q
zKd(@J*FhJi9R~k519G3N3+wh9#OvNO6L0tXh)C0Q=LN*;zH_IGm-{>!9y<BZzT%^f
zTVA2=TsR!>@H;#5oIr-_ws_>nTTh&{?9gr<%Xzg9U!7EaSd|+4EjyI@$*|pdJn6C=
zVcmDP!o1r2eOKN3O`fmFu-$j=oTkgad*?LFXxDvrebAU3C&u;39de2n_uKaT^;?cx
zI^=uac37<A?h5-~K0l;G^o|teeMii~;_r8HFOk*I>A9~iAAhF-Ubxrfo<hFs;T;U`
zju?;J<1mgH=iCe97kE!r|Jax1wp}?h{C5qJ$2lwdus&hk#CrpIzDihk9xdasXWn1v
z&a<4ZZ=LSr(KAx)G<x@WA6~M{0(I~2h}Zp{5M8|7Prws);TG-QwPQb^?p>5F9{Y5<
zKcqd_r5&=qc8bN@zuxWY@0Td2h`l>S#*bTM{MjLV@;iPz#2+|Cr~O1edEaXn#y_}4
zJ3A%a?fj+NU+UtW+L_E}*RQ18{XRsdV-HS|dBf?t2)*1Q_1K|&7h)IQJ}_Isr$g(N
zQ*`YQBm6soy5G_0;^qCOE^N22km>jx-p3h??f1b9-&uq2NPox7`sHjz`Q01Bb6=iN
z@Yui34Y@`3DWLpbfz$J^Nc)5JmDBl>e7f&?>EbzmiB!)-E$8&yCenOP*FDkZi}dd_
z*_D&)#Ir5TIYLko{}40`D&JCXr{_Ac&L_V+C&O}|!@L37?W1Kn?;E{Fuig*q?i-1>
z``t&>>9-cz(d6j&7U!T&k?$JWq3-<4t9X8&;ucv~M2l|gwac&TPrjp~`<<hVm%p=;
zVVR$EOuO;N@4Z33a|_DfcW|ByHo5QBeHUh*pg=jV%X3ZnI~1qq7VxD*ewPw0I{j`h
zziYQcez)Khy?!?uYkBhbYIb2cUvU2tZ`A#cEUU-!Zk~HCuTl417G1p4-&K&$t9e{^
zFO_s|fgz`>XAjw--F<hY-}jW|+@4oE&9D0|xo-c>dM=FO@6U0M;u>xKUM$}|h!pvr
zyhL3)y^go@k3mWI`*&=1-#v*oJ*%8&yZt?9qU2xEw#mdR4yUW2qI|Cq?{KvHF}rmz
z>jXP=>MynDmYTmmuiID2c=jnlLEb?{i}GGaes_tm)AbVlX!l-7zIT*i`S;tAE5E-i
z!@76J#M_NGyZ2LapDV-Cu4I_^VLG%sm(#t&AiiQ~?i(+sp0Ff7{$q>n_v}ax+2L;u
zK1v?<O{Qgs^hcthY_~2f&pTvTX9x0}BGzGf4x;<rxs2DHcj($F|GuRR>)yTF#dqKT
z;~gnE4q71JkF!JFJ)zroQ=DH(l;;b&aE+QtpY?fWlhUDX-^9I_+wsEh528iRm7_(s
z`bV3e`Nt{BaV*2~+>x-|@2*I1JE#AO&NnZ#bVxqGaZhruqI>Ebwd_!y)9dCP;@jT*
zT*vX1_gPNS&YtAoJ095PqGc5-Yz_ZCF<qGVrPANog>~ak7ay&C+LhOxSH(M?cjR^$
zt>4j=*Zp1>JB$`}zjJmP?{;4CD!-jQ`IYYU-a+Q`OP{PFSH92kDy%ziw>#go+h4>g
zUH1-J-d7TKyD!)Id%JjkXX+H?zSU{i&c4~t`HfEXJzMhm_5Fib%a!{-8Mgc0o=n%-
zt32N@;VtE+&0DXc<w3`{jNe}9f~Q0G@3cU>eRS*Z9{+vhg9ZQb_YL{3BFNv3bc*u3
zWKP4pKe0pk9z^nx)xuuCTeCZ-C*7^xQVx{ATa{tm_etEoKRI{%MN`Um-LBlCT)*0d
zb@$s&<K2E|Jt+B0wdhbi`{={Upy9#Tf&Dm$bbgZeb56rv&mWP>i?rSGF2}x0r-$x+
zsn_v#`_-*CT+k=Stl3J3%*#4te{F|O-wEh`Z*F_<nx0x~^1!`Hcg`Q}c{=wm{2o|`
z@^>q|7tAUs_jPA$SpIG%R$<+Fc&y@e{><(AJ68Ew4@diMZ$yhvZ!9(}({hWtdkW&^
z{g&=MfZcoWHpP0CZE$#n0>kfPGk0w7)@r%pdma+;%bfcX2IW19?%rDW4!~)=Zhq0l
z%X=rCU7C1r7>Lz)^+wzOlVZ32N%Zt+?^Dt`rL`I~d3f@_8A&alQY@u-N^<YCCTS_z
z{`@z!Kx*OCqA5u|(^8B78LpU;oRZvdZpmJ0siXSTDYN+hh~$62oYSxSz4!04?2!C^
z;SIZgp7vIiomS9rNb!n&dTgC&1r=YazBPZ_YiFyVVo-UPdUvSRWcD{H>ny*~ukjnH
z+`Uf}Z<ph$Lsn4m#~E8XuiksiiWcqoYF7AE`XA?ITG8T|TKTFT8+F(+=!^%ue$k}|
zZQhdl_9(h8{XMnaN>qgRU7P+W_d|zO$|?3OJiTS*N46-x5IxW@UU+f6(N&VBEVU9H
zB|rT#KECyVz4NVTvF56g(_hQCNhJ!(c6uFtq;}^!uNc3`iWPdTzhCm=wT?{`yNgs0
z{8&)2-treNE%fP6$}N_>Zs_b*d-hs(SiJqAwx4g_VnvD{%sX$#{OrG3UgOlUzg^$^
z$~jh|qu=`)ujNPTpMI$_qw3IqZBROlR4(NrwZB9wmt#n~8<s!0+H#A>-sn|5_o(EM
zQ@rN7Lz%Cp<qLU@D~_BQ{MtR)La{>nKT@nV<kATR=44o2p;JBl-cLBy!{mz%15Ynr
zwdm5`nJOsQsMMIXnRQMnzc5Gsn={^-bWC}Lo1Xoq>3OY>tDxb)4fjo7-*&I^8vU|r
zx9do+^83{uJGoWcHy-;-zod{~Xjgx%(kp!1dB^PZgG%D20Z&Yr-tw^0VYUkCb!$#I
zq`bx%MF+NAT_jyej1?b|){}nC%-PegKW(|iNb93qwCh;pfL^Un9)2%*!Z*)sw%j6m
zM2n@XcdVMxf49nNSa-*`a;2ATRY60)&YS&`JGFoR33(>(P+0|6)S5G*`k~)cP%vor
zE%l8z&(Zo%U3od)bouP!|7g^vSo7+~tgM1Be}C1E3!0`|Zc*MlIt@Sa=!d7CeCc;9
z(a^5n<DGunqN>@hJUmSW4I|wr#+&b;)&JS%Rlm#`{fgxk7JU1X5*gE$D6dh@H$e~A
zZ+g1NZ8scLexcX<)IpCnZ#VImBPwX<_c>YSH>z`w5xXqE(e8dS(bMI7$3zWBI`23w
zx9;N-XPPxTpn{HaU7YA)r~PB{`EyPWuhE|0#k#!Te%-Bnzt)%BU%Y%>o{{SZE>{L)
z!$Z*aWkJ_B(cejBRsFo)4@N7`FF&f2i+}YBb@dSM*S<1Q<odm@^Gp7z8Sm{VQhU3~
zDj2kLn4sTpJ<@ymgir5StAdV~x2U$V$cZs3-iW`97bfaCh*Q0$PpI51b@x|Rtk7vc
zO&T!eh8Obhv4V<De<$FS51DVn_8PZ+RB59XG?eY|IvnqN>`3ePng_4tIX&-`_2y1b
z8@~7R(<)MIvE}>EIuuNr9V;Ai`)9?nuQ;LnM#+=kVY}ZA_*H+G@!bb>y!VKeRj|g(
z&oxb*muUqZqx~*Htjh=G`w}`m6YbrLojz}D>T+9yANO0aM!)=!SNmM+&?Q^HSv}H{
z=+}9>d{-W{aHQ{2{IY|1$?eeN(-O5Zeo%g)SN0uiIZpSfLCF`?bDc={EpE%jYy6q-
zC;jj5Cq>&Yg_1|+>-)k%%V1*ikX7ET>&CIPU*4-kin`wsI*s@G{br={oa&os`69jN
zk#=Ch|H**U?+Zn`PIdeJA-C%#xAzo7_xGB%`>hpLP|@k{YrM|)pH_4Bg(BUzo~`BL
zWg$-QB@#8C{EmUw;du3nte1C(@_iidz~sCn*CX<KT$BUlI|RNXX!m_Z-S?~9-m7}`
zt~p-*Lw>(A+Vd9p4+^?nxA>L6&BL#@J9Y6c<rM2pPI-RGx<6D<F(}`e(&_os?jGfS
z?v|l~hLP@z<5dp(mh*~VVYk2Q@cX@wuARr9UbCsi*jypIdCo3f&dWg$`*j~D<wm+s
zc3MuZd+fq?`&`ybexYCUVxq})+JAQWbmLma+qK{AI*awVQ<Up-ufvg^_jy%Lu786P
zwj1w>p6+&^WtU%$n`pys&nIPmS-;)>KGJpM$rbf(oO$DK+P?!O=Pkng?z_x`^V*I5
zcAWxQZ-IQ5j`iQ^?zL+EaQl9xL)snmcYh78Yg~0yt)nW|2!D}ctlw?P`gD3X%e-Uw
zZADI6iH7mIAJy?B<NZ1plJ&R^+pTX*y1(zfobv0oU+waHb>0~-If?T3Cg9m`|3f_z
zgZ7?O>MQRFf)>`DD<*n;r7;Tzlss}&Wi|Ag_j|ASxbC2Q8J1Vb_;q_e8EO7l?H_5E
zSLcp)=be6~(+-1%0mFcE2Asan;QYZUa=zgf?e1Tlx9qQ+PVdvbUYEv7o?AZb^2h4@
z)lLt)^vb<2toh9GQ_5?!TUWWQ&+gnYQPRC$zsKwM2&3=+s>l1o4p{L*?AGnyGl_Tp
z#kF@<`Ka+WD^>_k65W0m->>|3zuTATvK?NBBmIs;qLfP%KNqPUk^T-OQOd>YJ05=3
zlTrH06`QWjHan;o>HT7&lykf8i8Mcc)9&8I?RNm(?nB6*_R2YP?))iT=`dE`b+mhL
z!F{n)bh?itpWWXtu>0<#o!`~@NvG|R`RwW=J<@fr-9D9cyZdYO@CqZfN5-8&e;D}X
zoPOQ!y?>wO79-Ur*565rRL?{$=k#36<jWd=m&0v8>ed56iFeyxliyoAPn31N)WhxH
zC64tx<Fvj=?_CqET)fUvy{fmy!#|#yz3`+JRFvn7k*-(Ve%~uz^N8EuGn{St?an1+
zy4&CT@IKrQ?S7Y{OSd~mkm+vi*6w#w@k(bu9yI*=_||+^mN=p!#Yp{?U*&ofU6=lz
z+HS=f{c4Zba%1gJz1BZkc}{=N>o#AsdO0mWvgoyyX0*v33ObI<^J({q<Ih+@#o9C8
zd8Y7{Ng;#r;o+ll&DP%e!Er07IDf@}MO}9tva$+#_1$=<e$cOcZtwkMe(tlJqAYJ0
zmiJh0-?iK6<#hdOC&#b;L9Xt5p+SlF`~6s~-h0s>iH5rUsf?HV>7ayVJAxK=`uj$w
z`C@H<(VjEJN?*Hko=DSm`sZxhaqK1QlV?lpG_~UMUHY9=Zc$fHwDEqA^PtQ_iLR&L
z{e*mZPh}T&`wm8~+X&mOqwLc8euv+PzLBEtepfe+BlSyOm(%I(cf5`lzvJ!pAG&nE
z^)mARI|E*SA3WCb{Qf(z|D81xD;aS6E+Z)Ub$_?-Y>D^!_wdqg2uHggGUaDf9r~{g
zO7Z}|LBX^Ae$nfG^7{9ZPW6lR{duJI1?AnB-`|0Xm;6LOKQQ|J@4pXd_=zn)PW`XU
z7xZ@qvOPfw^PMc6z9p|4I=j`Ly_VmotDktczxztm{C=N5z*n^B_T6ow<(J>HK3l`O
zzZ2zmJboo!7_I%sy1cwkY&hiA->ztvG}|rurMKI^UzsTRW4&LutIx0gF>?Q6Kz>)q
zukiW{)6x%>S#EiSc7Got;nVXxma*gH9<?DtOzV`^Dz$t{@stuN$-UE>q@`s0^WW40
zsfAOErX=-DOD+CqxME6jN^--wC3~f%j_OmV%;Kb^6Cd8Zrr)XER?|09Di55x#d`0v
zS6{e&#$GF9(Ca@ndTP5hwM*-roqn7jPA4Du?Z5?t3ijF-POtE8-Q>6K+hMi-<hz^m
zE&1Jgd(;o@&b@q>b@%p<j$XNZmG#=1iDj>;vfO%bNb^DM3;$xB^XQY;ztna1Kf>fA
zo$}}_%TGOL`RXp?-?YG+j%-_6xYkG7`q2NEDb{DNkNLx*9n^#TQ>RCMRJqiz)`xpb
zzNhLPvIvt;qSO~Tl1JJ{I&wkslTLop(HFi!^dTSR(MuwHg>mffwdZ&FP?2w5+MZ(B
z+e``by1VD1Sw%MgW>FvYP%nIvk9Np><VOzuL74nff6AdB7}o!9JE)X%$H`SqW+>vx
zhdnOcRO<UVV-JM;_4*m#UXc6Q{b4(%e&T7DMEU`{thr-YmD<l`gyo<wSSe%7FIPOj
zE^P0#4@6JGAiT)>qYXzp?_K=$l*YwNoKQWkTRy$>*yHMz$<Kavf3u^i{-iHnxS(UE
z`t76Z&Kv&zVYT7n^|`y=e=v;s-@CEKs1xbx=?h+)eaoc1YRo&cekxPxfLdSX%;n#|
zvQI55cuVQt>AThH!CjJD7EV_mk9}cHvo8;+XG%05ex_G?IA84%C))kGEkoUQ+5H_)
zUVb>74|yZb?YX;dp55x|+_}3hsJ=6dbrzRhFs9lDHKR_8&mY{fSiLu`(uEUttq9|q
z`D2@3+H##beDIB$cg)(X$WOj`<F2^nlbe^SLX{rR-fHOla61+x^(mb4$o6o#%9$xo
z_aC=K<(N6HSB06Y!gA0D{aPH)H}2;FTh)!l=k(1v_ji^1{OotnTel~SpR5>lP1&5k
zhIxSx_}yM*@cCcw%nb8CZR$1Khu(Tj?f&B61#7AtS4Agg-n_ogk#K)d?z#8B9#d-Z
zL3LBnk&n$DaX@`kDs-g#;6K7VQVu@JNB@Dyp?_!>d=M^p>8yvdwb`%ew{aI<w5(Q#
zjBr0w5B*L)<kQdSK{|GWJ;5*O=!G33k96W`5Apav#xLWM`Gx+!>7us(YJKqgu>D{k
zAnC~K-S^>HPe1vS1>z@*FZ<UQ(;r-F%^TBWOoQ`xTNn3v<4}`@o5S-b`SAy&<A*4B
z{!53?Y1ZnfHMK{zNhL2nVcl}%o|;{9WLmp=r*FA0_sMV=|3-QI6Y~vvQ7=5yc&Jvz
za#asls~&%8!EbvqEabfYR=$=?hi?t*P5AMgr3z2jzbV`f<l|>S=2`qG?dh=lfk$6$
zaWK4|D3QHs;TI<yx4>?Bmv%Y&(n;%`MbisEp^`#ZG#WH1?={Jx69sa<bbtHg(2%=}
zlzcr`_E4^E`}$s5CU@xC7LC8YbxzLEdnM~!RJ?MIFt%+~;6RU$bA=YBjZL}k=WOBf
z^YY(z>jgcNLO(9)RByug<Pba#dV5>ZV_T9#)mCRrt?*Lz5c#6@t7l93Tw6CBP41s=
zF@Iu5tV2Q4neVQgknOymcI*kSSFw|3!*kX6|L#P?>l4D1XI%>tpW}<u+vfM%7oNAV
z$9ImFsrgC$4dL}H>snckcA&q+-+!&xe@fEIuzbcFau`Rn55(VqtY^CY{>L2~^Gs3r
zlW_Xq>viH8w~RN&H~t7DKka4xhMx2Th<{|hB8;DXsbYx<HH#hz`*q@(cUUj5PGFv5
z-fC3+{8z90`k2B`PyT=G-Fdv7)4f0NbCQS<L=7S4c~(Qss?Wx4sZ&9WRV^aosOhS>
zsF{j+>`_`X*L)lEkP6R1D59tcK@1@=#S}wGNb-BR&3pUZrH?(Fz4zJsWc}k~t!EA2
z^;zF(t>-yAtH1E-?`KTO$9et4r^VLYd7G)ve3{Be#7B12{_#KN=riW^eEe^l{`k1w
zzn_qbOX5ZA+>hSTYw34VaZ4To#bt3)+))p$SDX^p@Fh;^AC`?{9{uqwZmL(t7oL^z
zghyrl!gy={?8ApY*<sl!hku>!c>ke8cWM4QwNJ)ZHm*ESy#B#*f7$H*Ek4WVC&txp
z_@)<3y8EoJ^5ftC@JExM8$CUR!v_E9=Z|mI)NX%w`{~nX4cU4|3d#5LcMhM>I-;q)
z?~N00yLU=c`<wgke0;BUn)Ch6-uj$}?pSALKEJ^`6yL;G=-j|w`2lfIJNDu{^U_CS
zpPsSK=lOWcPw+SV*nxXb|MkE{Cnuf{IOmZ5FT6NC>3zT^Gnanu)5JXHWrxP$FT_dn
z<_?~^-k$e=mK=WQy}!Nuy3Z5Ompt><MKAvN>!i=87tQgL#iu0tTc7oc-~6F`gC9vJ
ztuW}8YsV(?7XAYd%6R3sRzGv_O~0SNDL10+gW)^>&&<R+w3lD;Py8ppYW;X0_Vshu
zef9j8`E_VV4*l^hAF*EY7-!=8`@Hm<%|@lr`uRockslpA?2VCM-#$Fg1AqM5f9^m3
zk;{)vtVer(+<6bGhsME+^;kC^=!gBaukyaD295aRRn4E|`Jrb=O<nlV5g&Y!$e-*-
zm^W&NAMNj%=e#>M+;@5+?|Hw^cJKaI{~1XZ>Mzf;AJ!r7qrb*kcES0UzI(%&$(Jq9
zAMwOCO*trUUwhcix2)E8bn4tsFU3`MWL|vjb<T>f51;z29?B>2M;^S}fBIlw=mWH$
z>{@$yx%&tD$6t_}JsL;+F}}D4`C)MniUa&If2=-?2U+8YJH}Op{3kiZJ?A~~#rWbk
z|1bWFXZl0+)?qyL5Fd?-^ewbw`WW$F@hb8VdH5!83H5+~7H9cO82WMkTmB&~ikI>y
zaYKB7VO$gs#Vh64j@aP&C64+cVc-0}`NRQvi*cP#_*dr==L_eD8Qr(u`ix(^l{(jG
zr=NEGwY;A{R?m;>XIy0{FFX6e4=&p7`ZH6=kIU1Y^Ner4wHF+D>&3fv%gwxK;E6}v
z)jhXw-;>{ZA=f=$R=?gQ%ieSM<vnuyy>#2Vg9r7<DHn_OP2TK2Ak1Uwhx4+0?v7dS
zL-y};$j+_(+w=LL^Rj#cAKJ<LJeT(>)6WpeK~5;1;ZNQZ?w5TI(Y>a)>0IgDD{j*p
z_ulTW`@X#7l~=s*TIxR2bLhGIP-XYpSs2bs-{n2{#~Xf?q5bfjKFXt^=k7zUz3TS|
ztkgVzE;*$4-%s3r?ws|xFR%_gIhT_YU-)JpP<!iizhJx&t%DvwdP2`aImtr~`$j(R
zE#%3eKH|%{4nO9#59A|<_Zs%WIL<Bk{=xfM<JlMIBm3ih2fe{xcz=RldT8IA+vubB
zJl0Q6>mWZ~4}NgPPp&<CZ2rE;`1T90)=y8g(_cL_zWJ;luj<K1ZttV)GyO7;_fyuV
zY(DKmy}&QN`Frd1oL<=negqHJjSutdZ#?k{AL{ud@{z-Q<~OhVDd!RA8s`J(e8c}a
z_qfLj;~+VlJItq@`Jj6(^OIA2HlOy&;-vU^TKn1~E`0Cn{JoU>OmZ4u{U?tOTX~`T
zXQX&AuKCGt-Qury<W_bbGoJS3#v^&$$1T6b{JXBd=k#>tCl31cj@Qjh)w_?A2dH;H
z=Uy&6|KUMBJs0=LktL_~h?nk3oUg-sD)(H@-R_C_L3a5aKQZgO2=mA*%_o192gpC1
z>&c_N`w#J6KI*-Oe8oM2yfyD<XZcrtk3WUtsrjtax{br{=x09b7mw8AN!~4g5<l>*
zUO(gTXW{|>#ozHC{EYU>{6|*3_#ke$kB~QImDv%bhvtL)w|FYQci$?Hci-wf;CukB
zM}82-vCxn6*VbkI)=NG-%4hjyeiqVmXumz@-}!Cxi%;UVvV6mQ^wRzs2fy}NJ8|2*
z;sAcMH;?DW!-sR4dq-vWe(pKl2Zs8Y&tGP}S7S%?7xDuk+7Igpk)L4C{DX6#{qVk$
zU6GUjV!xi#t5eVa;UD@wJR@=M=swu}s`HU^G`&^EoA=#6Kkm-WFKL;Y(Cck`F0l3R
z3%<(VAG;TJZ%g0RYw!Jrdv|gwYfn!1sN^9h{@j;o@7|Xmb<TF~_FO+auwU(sZ(e2R
z`8Q|&aoi^DQxo?D&hg>_Ir(?{=y@ogeYalY(^GufN8{f#<cKR@Id@9_dk=afp6IWi
zeSr4UzT@9|#4G%|PZ!t375(T5y)nN2<Ot{UJ%)WSj{f+7{Hl4>(?4>1--B<w;E7(4
zt5o!!%slMVeuw&H-RkjT|M4#Vlat=^SIYRZFZg2j^wqrD^E>vDe&YjQ+KczvvlITt
zI`osb@Jr_N-p4rBBVO4z{tdr)Wk2#2aag`#K6!_DCSDlFe#wK3%dh)h$o<p60Xx09
z^^)E4-}&%c{Dbw|SNnq}`)Gd)VW_9#m;JF{=8gET_wA8~$ip|^U37Ilke@ow@z2hk
z@<Q!G<R9g=;yC}yFFHT)kNl=G|E%mh82VxJkk`2KdH&x$k#iAru5mx)e4<`{t-o=d
zL!3X9okO(K-}%eD&^b_9du8$)-*ZS_agqO)hl&sUd#JzSuzXK^H;(7}s~2a*Q+bB6
z_!PDy4>Z1ghWd+x<Pv9{@5O27JZ1U==>>fikMX9Rd_tZ@f8_z8p3+NkReSB^CB~O;
z+aJ8)35Ifr)ABL)L{9Uu1ITWzS3f)|vvc+A96yl#dLH&;kNC=Jr=C5BNDt^8Iq^tN
z{97L$v@;H#@T#4B81MAgbL)fZ=@B{XyYa$vyZdtfB1HZs3;B~$(S38NIChN@zq#P5
zxpMM+{votSdg<KfoTELvbw2k!eWCZY&3l~xhn_F?$UV8+v;F`6QTP0NTH~^JywV@<
zJKXp12lUc=VdIipKCiv~F^>DFu-?4%hkohD{`enyBu<*&_ip}9fE@1Wp!0_JBHnBJ
z`vm6)yo(e1LHBL=qPOlH>6thrKB^bb#5?1L^Jy<mX|JBXyVtTV>$ShmJ^JGj@9y95
zkMHpQ(D}|f>5=_XhV(+&I>}+Zo-0G^%tAcs@19&44|u_&@$HW?x$Lj{aDM%u{iWaJ
z(oZ{O_o41jwWm+~BR%Ku#U*}}|K>;4i#Pf~<FaStu^;y}^xr)W|Ksm>_#yWU?g`nC
ze)P+_?1%d~@f*+h$43|kvdZ)<L~*5BNYAQ;_A3kVnuXr?Wnrb|X5ajd`*MCB(mTjs
zvtR3xw}*ORe0Jv?z@C)Z8UM)sjK_b5_GujT&^XFaz446Ck6Mqso1U@<`=Bgds^?cc
z*PlNQdB>0TkexXfX~%yX$9$gK2W9?LJ!GHmo#pH9jpak~ASjOWAN&gBU-&71=jZSL
z#Bc45BY$<D>AWN!y04dqc`x8z*1fH~EIj`mcj943ytLl5Bt+}CF7efRjpMz7^Bm+C
zv&#IceA_z1cp%P0@j<@tJi;#d8Rrl2N4@=p^e@ya`b3ZDD|?`S;<5Vy`G$C<y}XtF
z;1BQi&po$%Ok8At<lzUz8FF}k%dW}oUKmf-$8X`6p82lA_ZhzD@ON0gs~CUnO2-^E
zZ&SPWzQcHD^6jUrbY%DTJ3jn$ySLBj)_(NLJs0VBQd7I{JA7{feUG86o$p5cJ*>Yc
zac}<g63qj4S@7GxyLG=V53uk2I6mDk&`)|qAMHno=D%ju>8tE`Z?D{@Z~wJ@pSgPF
z*ta}`oy#{azv_zR)<3jo?#g|iT5)9CoVi{%EP2iyTlUK1%^Tml{<%M!H|Jbr9Ck!s
z*uVP&_c8Jj_UYciy@dBy`1;+jOQx>VckcY(Da)74V|?}e0RQAY89w+g=PKXR(qndO
zKcM(&U-&af9`f4{a@aR^&VPjc_;LH;+{Dk)ul+B7{r%5Rog?S(w!P=%SLp{{=nwnS
z&U13pM}KETU-&_G$8U=__~n1C-+J(%9r^7u`Jj3D8D)A#Z^R4ngS^hCVf?b+J(g)d
z^}?^`$jRUFLT~v$JQ&A3>ct=ak$%Xp_}4HFXy<*Ib(`Nj)?ps!bbgkf43QnMGkQi}
z#by4K9G;WMy_R}$RK0aOhtLQ2f8u3`><a%-K1MG5DTjy;any5u)cIK)k>}xyy(`Nb
z#aqZ;*p+_H@%ERW2+u**A<pIV)>+@%DvOs;y!TxEmj8<5^etQm{h`0?lRZLqPw(iB
zGCPBwL-vk;yo5XyD)W!*oxk93_$zuwK7PXZp6kcH*&X@pH@$%JZu%oXqF3UX_~Raz
zoq29v^3fmm2H6$6V?TJ{f7ywB#+P-erw{Cj+~N?s<xkyPi%a;!FMq4OefQitjANbF
zBadMx?B8?qvNz`+=PCAsXZPgzHJ)|ipFH__XMHbfJ@f!y_McpkzjbaS4}RIFd>gV~
z_Ut~0Uh#|cP@ZeQ%}Xxmdq3zLZa#T-#DDz`RjqiS_xy$X3g6v2_s|#n%P#3HJy#Zg
z*bzU&Z}Y$E`2~9CoF#6!M<AE?F7%6EW0&F#q^JBn|0REsUl@nqaK2Q}Z-`6sSoPM+
zU&%Z0MLzb(uJOem$`c@d_y=~*pXl$oarhDQo6oxKmwxt<oqKM4cJBN^4&#St9(>ww
z`I)#uE^z@*@|JLZa>!5kIewTwQxCPL&-^vy@9Ymh=eaV!PcQi~WvE^rM6cDmhlKQ0
zJKsyQLw=K8_%5D3IPZt(??b|Rf1l_+z~7rd{qQDV5!d+-dSZW-*_HjLFX4XDH+slU
z=>t8WcjlF+xOa*e8Hfy&&wzL#euzK(pE$(-iLbEynPbmJWMH0?)*aA)srhraKDh9Y
z|9W<x+`_pHu9~-BpPV=;Zn)17-`szQ%i<jrAC<*Vaa(*B@72>oc`DRjTz0>&pK(I;
z9#I@7kNJ(~y`cP5JQY91UG>_D=jM?I;6<D_PM8PC*F&rqU2<RV+%2D$FT3xMH_H#?
z8S*fBe#npf8ehete*9gfk>1hcETlgn(lh-EtqVVpow;97CYN@`HNW}H57pyeJ*3b6
zj#-&r>Idny`N;#xqkniVB#-sT+rxVm_DxRXDw79>^N>$J<t!vWzR6Vx`Acz?-!u+%
zU+cc#`AR-4Up8Op4?}%$kHqhhACk*@wS#!X8^2C|?R+<7KiqF<=YGk1*icS#$Ul^!
zap(h-$LNO_eg&U+H?MQ7@%0P+1bveyLGtAJnDy@*7#|u}dvVfp`=e}Mv^SsiD3i;&
z=#P5nxqCr8Ie%$q-S{#O`N^j}-aU7}WMA-V9C{n}!=K-YfWBMu-W>Yf1;0mn(qR`Y
z^XU8C+n4<O(EIk?yGQ=_M*R-E-!t_)2J!;GU*LBN#BcH1bM=15f#1^(ig(V<?l-)j
zaIfqBh5k7g;|1Cu{hW`*P3L6iaPzwdcc10noE+Xidw=VFi}#}Zje95eHr|KxL(cK!
zqM!7^{?k8t?zw%ZcgCY9){hT*r@eOiLGt0py@d66j(7F$8MKG?$@)Th>`z!v@7X2$
zf!0q?tj9W`c|Et^;)D0F(0f+zrRarywNCkwdU+Xn@XW658~gS?%DGWJ9{3UbS+8|z
zXCCd9VYpB9k)0U_PsYI)q#uxe(QA4D>5u2^u2_^O@jLvMvfrKYoe+QKyCA<u<M)~Q
z9q(88L+y;m@9|&!UpP-F2l<TayCvhP=MTw6ZomI*9PNxlK63b79%bhr{!|$nM}86V
z!9Tk%h4MajXa6Dnh4Mi8r~HRp@>;09JV@R{PIlt_Cy!F^T<P5B{D?olcck7uuzt|_
zSN>o<^12YMi@fBwZv0uN{lbI19lxGi2mM#pkN%1a_Jx1RLi~}_`uJad%f2eJ|FA4>
zd7g#*2|q)=&_9HJA}o79%%6qmIXTG7uT~qaOWdZ{^hTN9*niK(IdU3D9*aNi_;>qH
zFW5EI-~9B*bK}u_>o6{Trl;gJPOZ^>h`4N@#U1*qEZ(RWx2)Sb)zhD_Jzm0b_4i!a
z`0^n6LDs#Ke3L!WXZf~uvJ3Ob3&^h>`%tePe<Xf6FZdlYaYlS{uO}XAFJ90O=Tdna
zy)z$wqdoo3Lhp}#mj~tDP+l%ybC0GhFPGPOukQOc@2A5&pT4uJVllJ_c5Xa;sK>A8
zA@bW{y>TGF0@a88vdY?<PuY0lqy1A4wGaD;_1YT;hUXjcogP8;?w7@1<3M)9@9|sm
zUw(wWhZx$Cd|95%z94@gUl08?|De5g?i1K2d)3}Jkl#0s_MVeZ**N@>_<~Q!|KkbI
z@-TMlzRYv&*>{+i*l)i_>GvPyrScyBNS-VYhWwIz*Zl^+Bu|&`u~+$?dE`~(;-BpY
z-terRKZf|HC-g)AkavC&k9en_^jR6wZ}PI=EI$+0;~gKNzhR%o5A$pJr#zW`8_#p&
zLhZ?mfB7e*KgMCd_<`aJ{etwF{_-P`-rEoRAbuK0JNBpy^`{T^ncedr#&=JkUVHu3
z(;I%teS-W3`ffxXLI0GY@3!c*{>F8GMn3t5e&q1|2|p=65pR{vYdrTN-dC{~>%@aR
z=s({(aLU!6^vcaY_j$V=xbFfvf8SCn9=y^|K3(wXIdgFIfB$B;A-m6++wA6X4}Y@8
zTzT|&7~$W;n9uJI=y&Z0tFQ6K%X8(_lh-^OFMs^~&rR%^|9c$cK<yy8^i%fS-{<Jh
zpVKRP>wN${r=R=_yV9ON;kVg2`{I9;_1E70H+x`rkR65ZG0f|pnBVtZ6@Sf-^I!Y}
z|IJVFE9^Z)dgXk>-#h<{XXN4^&0~J~4mt28epyc;6xZCpSigMRc;cS?8XwB!v>x;N
z4oy38(Y(<3`n#7BSHp5xZ`?5cY6qPg@JUYcLGeJm;Q#e=e`4RnLGi@?8V9e&x8ATU
zPk`1%eth909G~9ek)Dv7yq?>)5c$Ut`B8r-;P31BVgA>93f~KRu06l(yz0J1+=2YE
zxZu1juT*yL&+nT@e6bGpXMX+6@B9j_SG>~CbNBVWgLS?$uX7$=jALGL4L{^`|F55R
z)<s_9h>POAI3DJS{I$3$PAlWpxZ<+>-}?f2CBJT5?cA%`U;dZ>CKvy&tiOH5t9^y+
ziTy(L?pM`^7`D@%zfy+$pnmL*-N-lDJ-;Cz)!uXCkt-~#50QNAnLOH&mwhW2JHByX
zIKFyF9_2zvu0rkDLud!=Lmo`;=#hA@EZ=Z04D$x@R$ic;xGk@T#uuNV{3M*$_~LYE
zXU?VM<S*R^@gu&MXHPJ!XV3Cj_DOE@K=qJ)l1n}Lt;gTJ-n8o2L;qvz1#<F#^0QO^
zkiU?hg>pGp$z#M(_fXD%>b+;o^F8bDtmuXD`73@!KS+PVe){84d-KU_tp`u}JapFY
znBy6*A&PVKnxChCc!%_j9vff2ZGQ5|ON{6H6i9yK@Y8tV$3ygfMVzq?af$x0C-#G9
z`XF98&l!gvi?{L({D=#_M`SPVO~py}VLtL&C;9M+XNX7V8}0E;U)|@~U*q9D<d;72
z2l%xg^h`az<wN*kXZ~J^UCJBqDnAfs>96w>J9B<vZ_3a)$9nCnajci!;oCZ_+d9dC
z=Y0JV2duI0rw^xiW54WDKl^|$dQJcABQ(DD^xXL5r7!XjeA_Q_7|;0Tr6<bx(vH95
z&*TgIE2LlAL-EFV^eyxc>Wv5KgFIGViVr9*>SrC!74$^ey0xb-<kH``>e&tbv5)ji
zJwE9%dyuEf$LKTppgc_;;vA>F^HG>D$+zTh#?d}R_XgTIx7!c?hn>(PdWfHJzwzum
zN*?dK%%`1o(X+5$_?*7spPt!I$nNY*sE5|6-g@*WAN`OglM`RoLr(I0u04795xm)V
z>k93RT%K!B9`i!;kzafMk^YgJob=oJ#S{K5jML`Fk8vO#%tJ2o(^sgS=gP)|_)&)J
zko^3gb;;-10r~BNe(LdVec`<L_B_->{NtS+WUb4(@rJMPK9m2Fmx#OW0hGl9`60cN
z-_S>P%}(&hF7OkU`ALW`{?d5tR9rQ$JW!moU;GU}%1+6F7yE^0{Mje%_)+If82Uf^
zsUN>fU+58jtkeAV)jsog%Fwv<HbnN|TnpX9InP>$dj065?`r+tmUE7KFTYbIE{Uh|
zQTe|7Rs7Xn{^9R$<<;_9_fGO+dMU2U`@(aX{Lt@Gxu23h8lS$)e~f2dfB)@#E-!a)
z<UNaa<iu0f_pH{hzr5bxA^Ck4-&@!Raw|Ir=!Z}1F%P+|3xD_~N64e^F??_BoMOH7
zKs#k}S(ooW=!^9k*Y9un{j*D7`Ol@V+jfqe?@#EFbAY^G|8O4jS-*X@zwTX(Yrn`%
zKELlnpU7ig?eT|S`U2_S^wB>$?w#|y<*e85koo;Azn`X^f4>NC^w)ZneNW>ah2G%_
z5Au5JcfRC@_)GrPIYItRzs-*~`>$-h<WeRF`S41g!}ahx^n%=eH`Dh={0pSV<O$J!
zu@`#4PTcR1-}g!MjGq%%#3^}>IK^M_d;A7FqK|mtH_1WYL;2XRah1jEFs_GX_uty%
zf!~1oZGHTsqwjxpzTEY<?Y`EbGw06veYsFSz1PJX#HW2{H|p_9&*{ZQJ1^Mpw4UAb
z?{)wA`dT|r+@NR9?{E1XXZCJBeuy6Mo9qM6;rf-WKSXwbchBuByRcvMmHzP?%H*^E
z`gvb3U!Yg?PF^kFQ6C~bl&`UWa`XSnP<wfldkFH#AKiyQ_8@<dZ$ak|`H?(JekA|$
z-q7!idM_vslczYB$m8Uv*6E%N;s<Z?M*ZZI*2%7|OWvjo<+b`5$8+`Jc%IAqjicT;
z`aynLJN-O|=7st}`$)f_d6k{xjSKOX*N<6$KPXR?mpZ>d`)>cB^P%$qq+ixSKk2vo
z1p7+=<l~Uuc@D`H=HKM;T%05aeZVI^oQK#?Aq@S5yn!F%ANUP%j^E)g*t7R0;d6eT
zJ%|HPym#;9+@~y_@q^-y`NUOu1%GQD{3HMAeYbHTdDt<z$zk7>LpilKj&|1PK8?OO
z*Yhj<k8_dd&P%>e@ZEy%G3Xh+g8Umjr-$r>-tliCnwNZ@JJ<QHB}C83?>YHH<WJZ)
zR39S0YdysxzW}WtALOzR%FsHjt6X%B#;<XdA-$HjLGco**B*K<pQ4XYToRA?5B&2l
z{0|gQ?ThxIA5zv1(#NnYUgL)yK=%^9kMi%c({uL=?8D#J`FGpxD`Y>$vtH*AdgXf|
z-y`$O^qjozwaM%63*FbSAMNB9<ggy~ewW-j{64vVryh!H;xT>o_mJ`#^ZQ;xy?LO0
zfOv4fh(Eu(4)JUpWvCqt_sQ?MFM7~IcWt#;w|473Zqj;--EeJlJ5;YfyTm(v@Vyp)
z!LIC^_WD~FdsJ_KeBVY7>=VB2w|aV`-Z=i=Q(3%`U$bBLE8>;Bh<@^0)@42NI(Y)S
zu`b^q%5#jzzU9Nl)807bkjI7kPcHeAJS5x)^6)485Phbv#uL}%N80gkVceBZT9^L(
zKfWRRq4#*ihyM7cXHYx$s_Y(r;xzlCU-Z~|<$3tvr>#faqOZmgr}-EDgr6V>-t{vM
zeWmC8seey_ocs!V43T}aXJzgB-Z=5Ld#5zz|9u922%6t~?2A6KUwq+%z1t6Zsl9qU
z*$3_Dw{h*C{l=?uh~fzUfN%4P8}xwO-kZ=%`$2B<YEMr3MlSnhfAxp-n|_)vM9=BF
zdCAR=*&)8db&}Ki#1Z53KgKl<UdgRY5Aeg!D%&sel9PWSFW#(AJ}&;NCztk6J;a~q
z)~TMKGA{JJY)6qlmT&PN{GsndLO<zwe*JCl_3gg@?FoNv9K7&9=5wDV|5SFL#cw$$
zLA=mw_Q%iTkG*(~H+iFW^uu{wJQaWBr}9qv$o`!xtpi`;r+)Guc2q8smp_B@0eK#O
z$luDVJg5KsrE#na&+Gt_+xp}i&imdc;Eny@$9+8B=#~2b=X`o%-<_ZBpZyY_$wlAv
zr`Pn4p2>@hPagH|2b9HA`a%Ec9epE@_RxJ)sCU)_@#Z`gp11t_EbgV<yStxoKTmGw
zIQJgVJ+}K9=<kW_1ADYy^4LfBKJHDzdm;DO?l;_H7!RNLcc1K@nSRl$@V?x89PROK
zeD^2z-FnF5IiBp3{h;UaIq`uVvJ3gF{S`0hqwfUB<^GTU`(6OQ_Md(#^NZ|38LHRM
zc;+)dKc$~M7Jtgd;}7v?|JgnJ_Wp-ny5GaE^+N64zqu!*PxMFG`$2kU-}U#LKGP3n
z?@7JC^WFxAde0Ap>m#T4INk@TCkLdz#<za=hmd{phvsz;iEr!iKFNDU<KSC=^VvUo
zDh~;f-{x2Ie(9P?oB#dTYg2Kao`mDm5BIt3kDcQcAN&A4W1sw(GJT}q`td8)OYfBF
zkA3p}hUff__jKOhhVQ%gd+D}!2M_9z|98mz`^?_E(<?|nLch&_lgIl-$j@nK9OFWH
zE0piZ8~L$tAMGdPU&KlKi*J6zKJY8{ucLWrwfoy&@>&PK$ZqJRdU~ZFq{sB0zUj~Z
zi7)I9%1iLePS`#9*o}JQ(|_?vzO7zdlJ~MxaZJ9Vo%dn%#l4QWrrvzy7w71W_#yw4
zcPh(MmBZ)ii$&);`D}PE;W_j?OW*LtF09i&;ZZ$3VGs7rJ{t$`cyun{hn*{&W8Fu{
zGyGktb;#G9U!nE6M^M(@x!8H%Ia|Ij4mgj>chraS@;BNUUtVln=QZbd=M44Mscb$x
zxmOF3oaUjA?g#Xz=i-_?U71|k=}#XZxgh@R8y?7|opSixx+kvrhm$VasayME>uh=N
z^P6>Vckc!9P7df?show@g&(|VubhSW$$Fkuk6+KjaXb&}wJ#O%QEI=iJ%7q>=^g(E
z#Z7rMf52bxhv7X**iU=y%+G(=KZrN`Y2BfoD3cGGhrHsLa{)gm4^S3gtgE9auYlyT
ze*V&W*)9I%HMK_jQY_LR=R9S4=X~hAXkY1{cI*UC_|Q&&_apXCymBu>|HOIy`73-w
z{3<(N$V=o;^vJp0eT4Iw{6&BHyZp!b#&h{U`yj9NK<D=S`evO|l<`8JLZsL7y3h{f
zed4>kPTZ%zp6d_UiG8ELc*n1OlaHyVFZib~_TN6>n|$oZxOlK$`hj0%yj!2=kbLx+
z9pF{oNIvVg9`Y+|Z-0%;@0iy<8xN|NcUhnD$VDFf<HvmT*t#J-rBB9JZ{NvDF3-tj
zpNwNa$j?5UgXxQPXm34uBaijS$D#JtFOJYF<68%P<|o7nd|0P-;FCUEr*((tBlne_
z=l6M*-Qt5@S-16Bw{?+2KkL>{zUlqHdvX8HyZ4;V9o|1WpLp-C{_L&KdFYOHW~Sbg
zx<_{of#&soOI|J?m(R=Z<%8bKd9ND2&vcG5PdJXeA5VC}pM6zN&*&|EP){G(3FLo0
zr&s)x{jneHUVL*;ArF<WC>x(%$;*wao%!U~;-s>;=3GJ^`oX`;kN81xiJsDb&*>dK
z<<Hm|`RO-1#k=S1&c3LJ<}t5*QlBr+`n?79_S-nh*3JIyn>@pLRGA*&%{a!jF7xsS
z_Q^iDr*h6yAHH|CPxg&o*k|(62m9?gx#<Tqj_1nqXZ?+DKb7q(#Gf*~Ha;}JdFZ|Q
zq4}*_du8KjXWkIWftPR`ayd6U$K&0;D2L})@{^0cXb+7O?zi^PeH0$_Gmies{3IlY
z{>Fu#hjQaZKXO<v{~gMsj6Xcn7yDx$VAj2p`vdcm$N0+T#e;dYvrg^wgP!Blyvpt~
z=$X8ce`D|LT%2%EDsSXR`MGf2zUvcT@y`E|%lm3^*?oxbirklo!}QvBLcRwykNEx0
z<l9eK>B#Qw#?jt)M*7{l+9SW%?SSU|eT@F{g)lz*UQa!_<sIZCkGQVAxUHUFRPUZo
zzMu@{1IqFZ`Gk8v`3JOLA=+<wsW`>2izCYPhQ2`iLjU+pNT2Dk{Wq?D@+j@d$3L?N
z`pzEdl{jWxh%db2(LT7><ZrYO+i5Q!RHhg9KkU!X_zui-@v>ONC*H+js2<{j{W||T
zf7^HH{iOG4-m7@e>3ysBTHaInyI}QM=)ITc-hUa_`!Vl*y+8Iom>k}ZLhbbPUfKIv
z?`ORagl$9G23~mUrOp5AzgkynS+u2JOY^C11KL`;ee>VeKCKJ1F5J>|N?Yq9-;`Hr
zX>MuWc4WVyZLL?FzQr<6g>J8L%DeSC_PEEUypYx#{$ZX64j=K8SJFnsdgHUR<e9T+
z%F+-1<B7Dh@RlzQTj#~i+S6)d<5+x@G%^qwhzvvqA_I|ukbwi|US{0^Z@-z=3d8wh
zIWkaB4E%bz%Z{CP?AWwaJn5|VHGgsDTWQ3|!~d@b<?8l;gZJEUr4g^ESy=A4vERQK
z*yXh2H`r~<@#*Y_9o_#*^VLy#V*l@Fpi%D=quh;sAKYmB_x)BBpGF2E1CfF1Gf=MI
ztiEKk=M3fUN9-5#mqvb{KfCJh?CIqiZ`I}M=h{7feCl(<o=xiw8*c}q#F2q-87S8t
zW51XuHBz2bt@q@e^}S-8w>sYgf9o0J|3wBO1CfEqKx7~?5E+OJL<Z`QfolJ~P#nKD
zf0q?`jyxCh_kZ!;t+MxSo$b-%ewMbMoxNXOW&TFhyK3`a>F=a#HD0xOiQ^Z?K-}-*
zcq+D4wcgvraclecqml2(cP;*2xmdqaYP(A1OQp^m+eZd!!9euS(LdM1KX+8mYGDg;
zj@mQORqwyw*(W#brBkm-8XYg%VukhYfAfJP#A{Y<JMGctCz3|N8H=8|=?hywn}ir@
zd`Eu=u+jIM{Ecdt<^G*Os0XoJ9s`v+FU4_6dp}Yye3hp)u~%duG7uSv3`7PZ1LZKV
z!`@GCc=nO6rx7C$Q4hXr2j$pQIeq%i`N3*u?=UW@G?pVvF@0M8w#13+?)`ikG4k+z
zda&J2$wr^8_j($!K6og`wqu*fKt~xkY@KJ;Uh?({X;;UN@)Y);+;`ujH@W?`v|J3w
zD^%`w{KO-=J$*-7DjN5Leb!&`!mIzDRvWh%I_{*W28>CwkUWi!Ge+ID?61?u(^_Lk
zd0X-?XKcP&-?p@~u%q(F{{NSOJx8SLjo9Ihw9@#0XNjfAKyeK0yUQy_|K_Y0(?-Qg
z$H(GWi>)F9k%7oSWS|fO_3|A=sd5xr*X;g0{_b6Me)q0kd7a($J+4185E<w^25R;9
z-1Rn3=P}-~nM=*LI@-Qt(#GDAf!Z=~<o$o|{m4DnrHzUmwb#z>-_gH+jQtzq?{sTx
z)p54SKx7~?P<;lj=(GNPiw*fWDHY3=?1+|+*V=C5iD|8{TsaH%Tm8MoPrG{MXVXe!
zp?M!)@V{=KZ^Y+Gqv2iWkGXrHML$g%9Y?oqb=Hn|zM0f3uDaxt>yA3_g``ojQGRzI
zlr#MOfkrJau=uLSZ+_*Zq~5TQUyRQq1C78yweE#$wf2twj!Nu5W$6e1@kCl_9%{A!
zah}LPWFRsyyD@O{UOk@M`{xVgDvh%nk8!<`fyh9|8R+cyp7l0=#~F<yL<S-QwPN7d
z-(C5`#d^GzHYzUo^2*mtojN0_7QWl}om02n?X9HJ=yzaC#Y*2-lsjIzzXO}qkG|Ji
zJj;@Yyt50I`*#eD%Fjt>wXgY$Gv7)hHU<xks%KH|$UtOZ_GO@2?;*R&xU;Wj#Fa+|
znwlzY$C1;>Kx7~?5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgcs4N3*L)r$m
zuGG@6WwDm#Q`-i#wRHRDzpZ^*7ie9$rRkKm)<wQ4uhP=o(!A}+enZ<@uQ+{+WuE%3
z4C5Xgb<os_6VpN%_U!xWF>_5ga%@s5e16#zws`UV@ktiW-DmE#FIfNewBGR6YZrTV
zz8xP<8x@BQIs4|l{xm6VRQz)3ZT>l9^{Ht`v9tH@(T_h_eD$@TNUDvY9oM=%aMe?G
zzk1GTNu%PZHTGWl&>@#39mPVtS6dI_`2XJC3hhf*eO_;IHIxVMo$aTy<|kigVU}F=
zR^M6tbhUZQ-PfZ}?f$^^+kTLgiq+;XwD)q$S?7mp*H`Ks)LG+Ho1c0ezgWI+eQxs;
za+kfG)Egdq?|Bn;o_kDM@3`tgr~UDwqd!ec#X|QIm3khYAItT3h4Sih$JMW1?Wk0_
z%C)2AFMH$IIgg*5bT*bNf4Tkaf4Nw0zL3?gSUIZgmx`66P`g5QTdn7X<m#;FpKtfy
z_Z_^}7fDya$GQ!kI_9;nlSaqR_Iu^dU+Q_I*!I=3k8^*v|GJ}goRU@wi;a)fT2G;I
zv;2Ie>I=<Z@6RihXBIn7z4K9O9<O~z*LB{1J-o!GyMC6GiyfUmtAG0+um5bHYd%g(
z#jNp4tuJ-nLhUPM_Z=Ol+<TyE_uHrU<=eh;$5Uxn!D_$%E%u(JQthghug7u!{d==Y
z?OVCyl<F_FFXSh}_q_U*isk0bo!zg}=cQ74)4tq$lWO-PU%g_XIA84ZTKm0n^}f(P
zR7xKT&6D-K*!;iLc3JZksxNh(&T3z&bGUKJjWd<%S8IEkUFQqqccsg<x*u)S@8!td
z(Q{y}`n#v77FPRxNF4v)zYi~z=f~%@`M$o<^s`d=Vx`6@l+ShadA09>X4gC<Yu#PV
z?_{mBUg~R=M|5<aa^G*%YrkUO=hmuSNBv}}{VTmMh2}5UuT<)MTkUa1J=?p_;Bhn3
za<Q}Wkj}b4srPw1>Mx7!U+jHnvF)nGYu5b5en+BRXX$yNd5hhzYPSphXtn)TIDWPC
zr_shO_FgXBudH&d@`!59Q|vh`YyE}l8%;mRTWDP!eZJT(Q=b3DOHJ*y!tn17$}M+v
z-PP`2u0EIh9<ZzTtM-0{;}?p<Vf#vzD|K&HFXM#qzfsGTKF?Nryhis2-lH`Np0HJq
z12(#|TdvkvDZdkrQ>i@9xvW-LY+l{f+J*ZU%iqsHt@N?d^E7Jw%#yp({M2f^O8Lh^
z<78=Zq5QO?&&$0}pI!Q8?Mt!s9o?VUzq-FenB}h;tv<^ihI(2r=f%#JqtW)4yjlBP
zs6Nz#&MH^y`@>SlEqDF1OWcd={P*7tma8{irC+0;rwirbm406AzVq)9|MmOZ)y@x!
z-Iqe`vc6|1w%+}4xbMC*2;*8fZunj;d><A1apwu+XJI&wvcEHep*+<po4?i=^4rn%
zh5K3Ua=8DcmTS#-)_kS%)abuoP^<U!)tcwTwO%;<gxkBfR~oAw7krP=C@3E*6sIcn
zyjtJ8S4uwP(CdFZH2Au~KkL?BE{6WU+;Uc&&8n}K{)zw9#xS3$bh%VL%;H&n$cOq$
zVX5(}Qtd0vd#%P}#~sB&_hR+_T%O%oSS>$P?QuJrr+0S$O7C|^?I`Tu(ftd@D^xCb
zKb$ABe$SxN`D7t~QfQs=`S&x>bGx}tZhm_1_HxnRRml%KilxTstksp8&vf<nSv-aH
zmCDb;al&~@Ef?|w@%i`vyNI>or__Bd)IQXs*}YsSz81QVER<)In-}PpwV#FVt=zkW
z*y!J5*NgliuZ>bJq&FRXE`O;tb~H~4`+M(IDP+gBx}T_(K2~~Nh5Tu?o?A!6$UtPE
zcm~3I>3T00zp{U|td$?_ta(c1qfq-wf1i{!ULiROJ@2f)=iA`UA)5?Xrbn);<A7<8
z+;GL9Ub%9y()Y{7j#sPvENdQmTj<<bZ_jIe--oAq#jO3!s;~As>gC*At{ho$pi$~;
zZ6B4MuhH)b>pd?mBwwZfJ&;<DS1bLgwjYh-M+Ukw1KvB7id~tLdRSYz`F^>w`!1zc
zSgxG)+OOL0hijGhR%;zu=b%c}muoMv-@pI6(vX)z`D)!he{S;=a+kfG6ngJp@6T&x
z|E11TYP=}--n&}w0pqx%4w^b~Vp=aeR4e|}YTWQ1CM#bl*ABA!6?@-Z>2{Th-(6*#
z&>v>)e^z}Mce{GIR`*14o`3)EwS~N7>1Aitcb5Kl)_O~wx7KmHSovxdZ$tg9RJqu7
zcC=j>FB-L6>iuD<?JKpfS>t5oRgG3(x99L{zyEg=!@RlC%AFObsx@z=;!3gpG3&db
zcbAzs*MVovkt-K#6^AM{Pqpr$I%-eF?xS{d?6<*@!=ITeR|vIlG(THwc`Kz~g~qWT
z^@hd%T}G+*m-W)VqxRX^{VRQcQ)~Up$_tCFFIC=3<!O~VFO({Oy|u5j{)OY!+TVB8
z`D*2FtF51vvfoONQ>;IzRDK_hlf`?XyrZi>&*G)n`eOGrY**?%i}uCJ728#ofoh%q
zYBg@P?jwv_S+;9GTDia1SZY6Z+|{wz-yJCT_crUTU!nK+#qyzDmIZY7d9tg{pT&2*
z*4Ns9)GH5_lB?MJgkswjD_5AO*K4`f|Bh+7^VRAcTkCmhl_$n|y2f|Ov->_4>wm*`
z<=%JH%e`Q@o<=E)ds*^^zYA7uIlOl%RL+`TeX;Mm>b)JltA)NBxb*CuUp(c@p1CZ1
zq2HHxEpuX@9MrF?U{^gaHOjg=+Yb4S5VL+~ytC`ejelXkYWbC7-^&zxPSegkW~r!u
z7G~uIoozp%KE`ro8K_nMU0JD{nzHn`QR++mo&B(V+mN<_7v6el^Z)YNx>CzxEsM7_
zpV~H{t)<&H|84Elx<Ko~ElsDiwJ!2ad6kytmga3o_8Z#Pdd2BmEb~;Qw<C@h8Hfx-
z1|kEIfyh8)p!y7)GxN!F`t0*qT5nkWmB$$(1CfEqKxCkj44i(>Umxu8_47%@$ix4y
z2bI*ZI96mJG7uSv3`7PZ1C?c9nR%}GQ@1~kNxKSGmYq0S$iNBf@AA~{C%=`3_}c|e
zTkPxACZzR(q1>??8Hfx-1|kEIfyhAd41D%Tx3zxy#xrT9u=tr{%g8`vATm(j4BWf$
zXN#Oa>DaVV`0E!YpY`Jjw<ZxI4^a=I9W<IB7<Bktw=Z?+xHMwt@bFsGRqapCcq5G%
zd5C)OT|21nY%)qWyE8D)R-3JI#w~BB5hD-vtOs4)PLIC$g?*3x?A0`4<RR)ow1eme
z{=FaQ>Tw{hFES7rn0**%^zR_Uom%&SZI9jM5054h|1S?APq7>shzvvqA_I|u$UxU)
zV4>YVzGc`|ccz8#uluaF$2~v&TT&~m)OVF}oXUQu+Ew}NdX^nm6d8yN)Efig-v^dj
zu9v?fjB=Frcb)a7hf$u$Kx7~?5E+OJ%+3tlwBf|9yS?*rI`z0C4j%a2Z3!GR<HaBK
zyyU}V$?JFdUB3t3$>V=_U-i$UTP7!Pi3jg`;@!7CNZ{RDEjIm^i@un^(Hs6~=DyD!
zl!iF?S2z8#|6wnrS-AespS*gVg<eTRJm8ewKK$dK#;3*N1y}v-!QLy3OB)rR>pymf
zUv2qBn#WbH-hThnUQeHkSW_NK^${Np{kV7k$6iSr4NqG9i^I>@eS8`*@=%H%lxiPc
zz5TJr9Xan`N4=PixS;#vzZ>~_#2WLk-A3cC{b;c<X}#b%gC8A!%8$pV#p0aj?YZB$
zFJ4L`b`B3+-9N?kMFt`R-!d@br~6NNa>0M3Sy=A(pWn_9|F0`B(9z#7=<L52=?Z%g
z*AN+q3`7RHG6RRsd)Q;?4_{A;#pz4!zT8K<k539=a@xqvdJlRhi5PkKMh_b8ccdyM
z@8vu8S^L=6CM1<YzUYuk4ta3D6Ru0U3R+jOSa1J+AbE<FKemeuL<YWPp#OkPXQW%c
zlr{=}D|!6C$UtNuGEm<PbX#NMl%BmWPhgnehQA*@=&e<bU$y%KX{p%z-DlpoW|Prr
zxwzNxwO1Z+`@?CeDE}%I>wDLtbdiC`Kx7~?5E+OJL<S-Qk%5jeu+vAw2VeH=__S2)
z*sQU4WFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtPEYcddjzpb?2Z;O0IzP?^<
zLRv@<x~8{?tBMRn2I`T4a=&Mb{W|tNV5Rw^TaUf+>^uiw_j+3INZzi7)!Nra<Ehm5
z@UeY)-!D{(w?@;it}gFIyRC52@R#37XIJd%`#Za>uhhR^i0z*|r_Vl*r5)vAcGI)C
z-pD{?ATkgchz!gQ4E$>AAMCzW%d<(TxZzp{ZL{~J7t)B4ho}e94r=8GW(S`f*BBXy
z4AeIR^K7-*DremCb{cW^<e|Q`vwG5%dQK?RzN7b(vH$<Q=dIq($B`ogjmtpWkhX!X
zE43`%vP4Vssci$=TDpDn-_|~@3$!lW(sW8&>muKjS7~W(Y2J2ZzoBidSDe1ZGEdd-
zdY5jRHf`{c6VgV*4Ih2ui4T{4Ic-$jbJ1&(r7wRXZB%Sj-aAX~dh-DPJ?Uq^ns3a9
zX`^7J_O0IbA<8p*GjQ2$>#x05kMU`}qTShP<bGd1@aiXNu~;d4%^C+^<m)WFZN&1c
zp7599X}x2$_^}W5g4J4Q9Jf9hSZ3_)^Y%G*avHI*c-Zm%VH5klG%;;dtQUV*NRE3K
zy8gnI+Fnf?1?$txG&SwD@*c0BwC-E!?21e6x%q$nzWZBAXXELIzrXJWD}0#N3hPxL
z3&~aOe6-N~@p)t*G7uSv417NWM}9qh`-2aDC#e>G|LpN;WFRsS8Hfx-1|kEIfyh8)
zATkgchzvvqA_I|u$UtNuG7uSv3`7PZ1Jz`p*zYQ1yVAZxjeJGEO8IU%>PvO?rJ6o8
zjvX0@3`7PZ1CfEqKx7~?5E+OJL<S-Qk%7oS=P<C~WpgZe%?gv#M#Xp^RNQ?~=V({`
z&N^|~f#c6!eq1uUVg2q%eM;7-zyB2FZtUOhDs}(r(|V#*k%7oSWZ+u{8s+!wx~iPz
zp7Xxl!T5jwj)C8exxM#ex6McjVRFfw4_vqOr%9>!=Z!{Q{FgI6Pm0A!8}2dR1*gnN
zip6*TGWy8nF8d-W7XLEzZ(sGkZ)Vb{xadYx)_(41UnjN3f2TXMQ{8^Q=rioSRX$4^
z73bWm{mL)r_&li<c9w25%I{HS$$3GK-Bx>g=BTt<*jf7$=Zy?R21;k(N2~6>{_z_=
zmllhKezz|^ukYVQj{8&E{*-P}<@H$ZvXjnxey>SsrLjC|Vz0<RWFRsS8K@Qm9}oR7
zciv`SC5?txulMRli*ERXT%%&!v19hwdi%L^mBLEvM@N7Ewp4kt{+)_K^`*|!(e_vM
zf8deJ-+w#J!j6s~`*)0iiwA9b_#q2ClXeznNk3`9pU!_z-#62U|KcHwr&u2uhzvvq
zA_HBUfeR-+{>n=if0i~Hj#*%hho=mfl1BV~9xDCsi*#*kXlk0&w8sh0fBst9sMyu_
zC$29t5E+OJL<S-Qk%9VWpwjQJt36)1-+h<cFKY+u<=+$C|MrXSy8Og(X|b5Ko=Vl1
z%X7W-YxKXDX!O4$U%F}9w82MCNE;Qa?f<GhexveE?po=8rTk@A8)wo|Pd#|qQh!f6
z3t!prwex=S^9e~7F8lT4z2BSkeiE^BdAQ&gH$Jn*dy|t=v8&ljy{(5IEfvdM$MALg
zz3{<^N0M4$xpKySk%2~JpwZqNhVs_Sxo~!sqh8O$%ihp?i~C!jPKw2w-hX4YOZFI_
z)*J4%!NMoL_2@s7M#nLm9CmB}{_m!ZipOn|9Pq@b57I`(zn*?T``kZ$Cxzw4s)K&>
z;-o`oe3%rA<<1lP)fWSI4qNJ+-)%Z2X;jQQ?`PGIKYW>+|90KGX}w|A_ww~#zvT9(
zjp=*HOKG`SU)IpnwAdXplk10$PHT<z#zU4o;!LeDOODQ}@3TVN(evFrF{w1xYCT&X
zx7pj<@7$bgbewShhcEtS-EO&voy$XyVRv``?FKWGM#H23JoK$scKIl2RGjkHjfZY9
zZ9>vjaLAsA-#6ow@kvLqR(>>Vo{s(=_ZxqBv1O@E$0V~mX7O98`jJC-TB`ps)00NU
zQ@2=U;J7d5%2f-i{k?(lE481UHBP<WlNOSz)c1V#*1poek5p^@>8Sj*lJ|s%_P^$#
z7e7lXjkS`$TJu!<z0k>fezMWJ4}FzH>=+N?OtrCEdKSkmkAYHupSoJ@J1f5|kG0rq
zc4MHc+#ghH9kZK0l)K(~J0Da^o<hH`)zzO@I{z$Jzhb}0KxCkF1|Hl0jIF-Bt#__c
zSh@uD-J{pat6VbUmiN+#k%y=U(GKd}50vXor*8Am;@7<~DQQ$Jcio-cZ}7>dfB2(q
ze~`<PW6O=s+p%x2X-Ty)YyMat87PN=FZ+Ko_wt9lna=L`yVG}C=EAK$PG>hPM`!A(
z&s|3>)pXCzFQlbnwboGV@3_Qv-~YQe)!LspZe*Y{8R+QuB5gz323~mUrOp4lhSrr@
zmS|bBrTNsh0c|bazWHx!pVkFh7j9`frLA?5Z_2B*G`BQwJF?%<w$>|7-(s1knwn1D
za>~rV4H}g^y6cP2f41Vg>1B&PcJk<BC+6|BjemIRRn4!bA+Gbt=wBYW(8y$i<u={m
z$)~SP?jHTeVGsY{x)iS2df>4u{POk`w)Gp=XP2v{q`i(xUV7-pH`C*rE<EPkAC69@
z-T&C(pRV+3veM<B&wI&zZzP`IkX-llt<OK7A9wd26aVMWBmSQ3{>9!W%rWh~^vW}O
zwyx3Vla$=%(a!k8-x_w@Qult44teqJwQnCYJ};knFPQx9hErA;nf}M!t1SH5++*_i
z(VyO1weM5!r^bWE?R(!@|2$yCv}E29SN{6L*Cr)@TCe54Wq&m;HQog?4jgpG+pp)>
zcl=xDEWh84!&B=z=<|(Ro9-B!Z~yh23!n7<rtjp}g9mc|ZSAklSmYmnPkRl0sOPQ+
zwkOJ->u(+8Sp0}Z9=d<vKl9__i#(y+)~Vk=kGp5NXODX+`QYfIkK1OUFB0nt(SAU@
z+pn$uH1_(%?>{S{U*<WePwv50Ho85bU+T5DK5}UH<{|UEy3ygI@_O_?C(XIzxOtvV
ztlv2FQ8^3o^&OnGkKulqU%hpa^ZwmF7<k0br=^}7SNWW?et5z<yL^^dzwzulyCi?L
zu%piJ&P!j|bl0BEIreKk^W3uJI=e6VbV|RhkN(-$Tl!vkz*8HHPWh3sJKy%eN3TAg
zw?pmMxp(_DC+zTc+I-<dKls6^FQwY?L*}vn^w)mVCuRD=o|Mo1^$tt-`0BOf$R!rN
z^OfH3<?VZm1IP7x>+|=LeYah5&H0CYoNROW@w+ax{N$w9PW^WJ*%ecg18)A{S94ze
zVZOa_wtJ=L-Xr#TD{nXa4m3agvEI>-bzf`LA19<Yo&L$oH%@vzwH|sDqV;KKo%AML
z&wZO6chAk|PDmg*_&12xEYz+b&hmf!PKe=p3YGCxDu(ms%d_H9xQ@KPp5>=J$EPyA
z$NN>kT6NBucYKhKYxDr$+C%n3FN|wn_zm?_-@o?xWls7e@ti-Q$Mh7QHTJhhzj9Y|
zjvV~A_T*NE_FZ|oHSSq;i`Awi#^LY8LG@6-aDB@9!G#{#@_&|?_mjl9_#p2FU)^?V
zpVK}{R@`c(G4oECl!|-&!+F<Vw)*rw@1^rhI`)CZH~lz|&u_BDf-9^zIn_>CJmH7<
z-~K~4TxG4ZKFZr4JH$8n$!ERvqN9FwcIp3<RgXU9Z|NI(|F&Q2Sz}+`@2x!V^bhij
z_Dz5Bgdeq!_KlyQAKDq$c<PO}+H<1^t-IE%`8cjUeG%6R`Kx00bFIt9pSbA7lamnH
zM;6k*EX@143_mQsg~*SEDBfh*0l&s>`7L>&_)RYD@hxv0_T-Lxj#>8IL_Wm6$YcDE
zKNvpnnrq)o_UOOP@C*JqK96`e&wT^`d$U#kGC3W3d+P;j|KXGLfR~;avEI~A(vvqh
zZ1hT3f0&Zbc?JGv|0@Um=FBhBus=Uyob^{b_XnH3`DuO}endOZ=_P$rwtoIz8Q<2=
z@5n>=Z@l2!{-_r>j6=`FCGn1*^_)L;4ltg0FyWYU=l;ir@8t6)<Jlj2tG)5*t2lMf
zsATkj8DAtrZhG<QKm2}r^4dbZnpgbdlZ4#X9gagDanSk9xcc9;`@%;bbKchpG!G;v
zx$%lm@y33rm+z6oyz~-}P(3_h#rwL=v+>7?{`i6R3(qg~-uLz?kH4R^9PrpN`#d`}
zk0&nPW0haM^JRX0&tCJ&4gF95B1xuQw!+ziW+a~5N9)~f+$Ya%oclaE?9TIUT7T4M
ziSp}n@37L4>%L0x2|b@{n}J7N*nejJx&1RQp3O^NtrLHf_L^t(al8LBk<at5{5XFs
zzn53g2Yw4b^!}8yx0vV7qbBBgm0wu*$RGc9^;Hl2H0ihIsJljP`}TM4Gx!bm$uGGd
z;5VFiLVM=t#0mL0WN&!j=Qexohethl)7WI8E0<j9%5Gzl^z4EA9dO$4yk8r<RPM<)
z28>JSf%6l)gVt$$`AaxHJ;#Ije!j=1r%xSucK+Oj5B|wM*$4CR7vx}n;w(Pc5&zF0
zkW*YWzWs&#fPFQubr{e3@T)(0Avy4CJaXX0c=U@tT9@(o5%RES`e`2LTK(l2c&1<M
zKwN|5%g;OOUW@$nOgyzO%KQ($$f=&+QuZ8@*E-}|?A3GjLl4-Iyxjef{_F(b&bQjr
zAN~j*A(EpIlCKcryVA&i6vCND?Kt?S|M)B|gzSjk;Cu1{ufMbOtskfI3;SdL@EY2?
z_VVajp>@imAfD_id7*WM`a_QjAw7%uulu)J@xb4Q$j@eBm@m+Kevp5qAMAl2lb0Ar
zdv-1l(~dtRr+h*l^yM8x2MyV{M@~Ly9?zkA@dEPO<dx@mE{^Cg4mt<%=lp<m@T=_H
zbMd1PY9Asw*%kRAd*%PxtvJVC$qDU;{VW#U6WK?8-TeF+y&jeB_4=iQXXNjR@IX#H
zTc3X9w?6I3Z+_)0q}TexFrK;Ja_+Pq`W*5ju0nFcu)VV9AzHt6L;D2rlZEbY=#6|A
z@6H8bzAE3P$L5#6(Mx4~%PXu0zwVvL!G7qYJh%{&n_ZTQ<b(FXy_bErkM1#UnS9j5
z9oCqVitG3`j{Tx1c)=4ML$r?}+Bf<~PW#{-uAUsuOY{_K7tW&`+O77+-Q}9m1JC}V
zXO8^r&AOd~?01NZZ@tOh=gm1i|DM~v8{fG4u~XyH3%trVl%Y68&*+QswX;u<zT$x%
z;FDbX+gEai^WX=HC-{Q=w)L3DI*ji;B`(r~taXWpklgGP|HjGlHS7M;yw>SCIgMi-
z<i;btGk&ONAurzVh2xr^Joqx6{UZlCt=GPgPk+yiubpvV{`o9k*gg4;i%;=h*>n43
zJoWV7I`mVfkLJZE`!ufho*Umf=pTN_>A7*7L+PJ&YHwWe(tf$uC4VTddKj)td-K>|
z^VlEXWsr}bRPVk-fAZtq{(Ejd?YwW_|MsyHo;#*J^_+ahhx*|G+J_y!Jba$hmi#(@
zU(KJe4}2Jhec>nEFL^M30Q3EP$1@A*x%&v~7w_m1dEKYc3-==A$ot{o`x?*bq4oG~
z2+AwSWnZm}o<h&rzx>=i>U(cqe(B<)7tc+&>aO#CzQyOM_X#1|NAuz#T)%y`FL<?Y
z@?B-fF34#fyzq<0b+1Mb$w4pVANGNtfb@^PsW-m$LhU^Fo+L!_;(eQSZ=Cy{(|hOG
zk9~sd)41|6JdmG$Ss#BxUyN^F`U=exqW<n3Lv)W(2=nXCcz<XAp!L#o_m=YMu$<S2
zS>HieuYI>a_DelJ=$HBErFQr=j&>n>zlJyZXP+K_e3ARMz321%cXN0#KKZn_E^>IT
zURitc85a-q!8*yMtUtM}pInd~d}p3n-$fcP<i&h2ymzzTSs1qGN1$<wMeBst3;Azo
z-Ozp%!u&qW;wO~P{NXsxa}eJU&yc?;7x_beJJcKgM!j{}XS~@zdda?w%ikE!eUy71
z=MnE)p!a>s?x~EYo*d$~dVW)0BX5UzW(VZAUUC}8xb9Pwtpk$7y0n)ElHa+*djP1P
z=j62xaa_A_Ui!lR?U#Lmq5jd!5c!>m|N8DX@(_9W=KJrc2j%Gjf6Kqa(C>tKbfL2R
z9g0^_-VmbaA(AT#!?+%n`Qbt+UWUl87eaA1>@SXGVc0$_i#J&)o`p!BLiQoP7K`FD
z4EZipmM1{@t?z#9v+t9{UGJyFQ|->$vCn*`%-b_(z3TB{9(jViKt3Q3AcwN{_D$Xt
zBHr|u7noOGDBiO><62K4^c|sldwHXJ=TG+o>{xw>?AU#97Mh3rAr?A+`g<1l{UJJc
zhR8pJ^Mz&m4&~+0e#6kdl;vk37P9;B+^%e&Lo8&E^t@E0PgzJWLZqMW<(!kf-}GIF
z{h?R*rMJGn3-?EU>wd|3()}jCY(DMCslD@``&jklqTiuC&?EKqUVGzy(reZE4%u^R
z{_g|Luf6jJJ29T`I<3=uzWcK-_k{KbPx@JheZVv1PqZUPSmuYw5uUHuf%QZ8C-jvZ
z##hFd`OU)*Xz%_|Jka0!J!SqyJN@_x_9)&MpWe_{>m(m}ZhiZqxvzS8M*g1KbMy00
zz7sOPcF=SE$ft~VWr!zoS_iwLU&c2sq-WcATl4LW4*oO^+nZNCeKDSSjbnZEZk6GK
zh7Ugc%apw4wXfu75BRVS`Ydnb4?LH**e~%@KkfNh&-rDj-u;7q;kf*gdU9xQ-OAeG
zg<Sl!=hjD0J*S`ieimw<h3vapXdUHZmfY&ak#bS|&O-a2h3ueMWH0QBp8FnRu>q4e
z-F&ytQuf16*uhbo|74Th`+kwCpLhJDZ~peZFH`=MU-8`gGWnbN*`;~Kp%>OZbECta
zo0|VF!{395o95y7p!ZzPU+TqG_Un5O=sOwvN6*R0pW=z!S3kVsOJjF!YF8Hb_~~0$
z`}3|dhV*EM;)`+UJHA8x<UgRlM}X`d%Et>KJLO-<i!c6}eTPWzvoKsoSjIyZiqlz`
zwLW~7jkA6yfgX_$KjHrJr}k4G5$X#)lfThJc^rNA-2K1$a6J2D9Px|&=l6HkdoO-n
zJb>ar$d~*~9AppTD}7)e^iQ0?o4n8WY38GM`ayQ9zww}Uz6<a^PW+UoS`U92&TD==
z6+-Lw-W<>LI@~Y&L2vl2EIG*I+)cmTzdM(|`p<Rl{QIxx&GA#t+w25y5D)GJpnl?w
zdwk<~?(dMS6Y58=$ot@F*R3>WR8x+;?jig=l)o3_&%7`9_aMfF_Af+#zhmBIKfUOl
z3&+oyGj7ktr;Y#fF1>OP572zjzCeFJhnM5tzvR3<KW@%Jyg=)~$AC?L@ym&KH@AoR
zJ%3@J?Ys9b(E8oeus?dv?yTQ<{4pMUCmq_mam`OI>xKUQhW+si(7NTN^wPLcduTrE
zeD4<vzTJO`ru_cdZ}(s9-n!_i`RJK_bIziN#)J0BI`Iho-5n$+q$gR(P9QlTULb#w
zh4>1QALgI%w9i5_{<hD)-E#Jg9(c}w;W0$dt&jfkXXeEh{c`S7FMe2Oi1gULnn#(x
z2vJ;gZXvJpo4ixLsa=Tj*+MA4%R>1p)J|Ste?qk1c!KUtLJa#?svPc%vi!n2<RM``
zBOihIl;2pF{iR37$3vkw96qOaP~0Lvf8!hj{k;PJ!A_uj1+oL<^B-Br4@33<$qUWv
zIln=#m95M7jQkmUVaMwE3w~6+b;vKoXK}$e{2l+w&$It5G*2PS>Q5i=1L+~;PhmKo
z{fFWt4CM>^hh=$H7K+b>(BH3TAw39@9%f;dTwy&sqMr~iFiXGbMIofOkbc2%ePNk?
zmWuwqtPmEGKYT7<$U^qPuR!gsJ4AVu^94UiZ`lJsO8-4)=j_^kYfoRaXD{-k&<~5_
z;-z>iu0r?A;yAgC$IlzjIf;FcQ`{tv^_rL5?2(@EhxFGykM+_!`ygJD56{NckG$5e
zo%4hC5YO5{^?0yP{F!yh%gCpm{LpvW^wRuzV6Vn07R8S&bdIqfS!h1|Lf>)YL4IVP
zjql!X%k#UPJFw}?)P11)MEZ913EQ20^aIoK&)qLx*k{G{uibuTD!%LY`wdUs^TtPJ
zrcM8O>6vZEOi$fkLha}wv>y6xU#&+zB(HD}N*;FNeotOxeEs|#6<(C_65cDh|2XiO
z1>ZRNhObh*$+z%I9z427C6D=xZ$9hwevcl9_AS2-QGAIQ8Hfy2jsbB`yc0*n^AIaH
za~!D=7zpz!_XqN<@H{EsQ7=A=AL5_9#W>EZ%Jfj4k}uEt9boM}hw91UJZ2s8e)*d+
zJCN^qu0E94Iw1M<m-l#o0LdZmvtIIt_qFn}5bH&E?6W)uPvX104lnjce(8LHZ+^&g
zsGWNR=-i{7b0%JmXP@vWKc*+-H&537XwMFmwb$Qz_)mXt=G=xi{m4gd@@Yptyh3(G
zf5@Y3T=m-956{Udzcw$_E|kN5%j=9!KKY-1p6eIZYwx_|IXU!G#*=f4dh<i_;{}HH
zNRJ@ijB7sQ6^rKKPw20|1Ei<kSDMFrKIb?4<GnFGgyf}H%I+Qgy)rZpKaFpQSNq|d
zf)D+SV}G1u$f2G;;FrzgIX{UH_Mp9eqrb+7{0qMMVW^*Z?6b0Yt;cxKx*_=?IjxgE
zD?|4H_@k$xeEbD{As-Cm7dz$``627HF7uI>Ug4XZ+L70Nl6rA3v;*^)*Lc2zux@(o
zJV|eybM>Qt_SyW-yUy>*_Q$?EKYQ-nY#gY5RJRS+oHnz2yL$uklPg5=n7ujgiZ}Gu
zbM?yh6WVw8<nA54r>5WTx1Jty_z!kGzbWT^y!VWM9=Y-0Ume|?^ZTH_^ZDwH55GEO
zov)G=mTTJX{1;~=mwtNH^j!}7GXMMd>;Z~v=5^0(Kj;lRQMM1(Wj`SOGp@4!c+;Q%
zR3;amtY03$4!o~`cymwc?<4T${ib~|uknn}zU+61^rl=ip8otkz41OB59IV5pX||n
zj_<KN$6tu{0UC!t@I6i#e|$&dx%}rl`Z4SG{P_ib2$~P#8<L+NG7o*SPv@R=#I+~P
zF+H)b*0atZSG#_{t)_kFcfP&9q7U?=SoC|I>{Xfm*)M*;dZ7J-<P~S!n~9tF!2`dG
zA9ko-Kl5v6K4@NL@g<x`o?w5)17+ir!#MPn9Q+ME^xU}!>TjQ*dMM9PE`;Q@9_z-h
zby*MpBu<In^aS79;m!Esse4rUf_d?x?C(PK&yVk1JnMHO@R)`CHa^&k?_KZ?wWGh}
z;)jjL-mI5C(_8h{!w=inEY#ok7|t7b^W6I&`$W(AJ9!7xPM*!bD)Vdn5j3y7g`bA*
z!F>;{4CN2}y1YaFqn>>HDgCkU{2e{l&U$@6A-^%M??Tm^hyIBZ_;7CXU6sER^gHD2
zPoBz-=!baEKIALx$oTX^|3YZq5Y3x~;XcR<*)e&b_a62Mf8O)hcl#>e43YleiQeL$
zUh?nk#lEvMd9!)7=U?&6PT3*!UfKD<?<_+7LX_u*D6WRcugF8$_p%57c$>{Oo+k%G
zKNapLejq;p@oN9*ulJk$7d^zQ@ASzDt(!i|`zCF%$2^yh>Xp+!3y*(e!d=6zm^=5%
z%kP|X<ggw&<wK7?bfeQx>Y0P=Ckt;_c1o{jwwf=Og;Q^Q`t*n2nI{*b{R#Dhe^Ku}
z9Y2LX=({X-pzM90ap*b!VP8Dgp8qi)yK}FgpLyuDdh&X1-<7RTdu92m_@G`t?a9xt
zE3<q3jORJNtVg~0!>;)=dV^Q(wKI-+*fo8G{78uOT^_=3$mh(%KN=63*Sh3C`jZ>7
z8)!Z#US**;0NFX@@1T9pLU9io7m9mOJ{}@{F^~1q2k+JSar!}T@d(Wa?St|0Ods?U
zSJdlIZ+&-Qz4S-D_!Y)0_Guh(jh&ef8kc{i-+01<?+NHRJu<#^vOjt#4j32Tp0fk<
z%g>x2=^MR~r)j5N{%9U~j<_jqn9sV@n<vb-#A$q($2#q&@vINOc%l!5kiFuE{9!%2
zC12==#5?lH6O03`+kD1z9<i_N6o2M34m%{T{`Lny>_a{KG`{`w9DnM`Yn*UC^ODp4
zs5dUXh2kqY!hZS}i^ikZ^o4!;9>zTU485V}<iRuh!-MtP4|c;ZyHA3~vCn>&)4#K%
zo%=xc>l`TlIOp9s=*gYe9RF$l-|v(U`nwnRZN|}GJfk=AIOv|n{hz!Rn#cK7y*!wF
z@;vchJNgLm4C#G{_=TSH-yzb25Un3Cr6Qh6#ZW$WCC;!b`bvL&H>sW-$@9b!e1>Rz
zdJ6S}_zaOf7ean9MEahE)?F%w^ORe*4_R1f-G!dpuPl_u%Ny)7za&3`;+T6C_CS6-
z(gS62TbW*oQ;^^1pZP`lC>|S6yodCh-I<R(<W|;?f3%PMfp+{Vze}IAXODMXu*}}q
z-8EPKck_)apQiurEzCo1`(R%E@q%aZh#cnE&phO&C;9r|cW*b{{`_h0F4R2-v)ZYb
zXB0x~3sGKK2z{UH`(637eErl(r}lW|$)@%VmKnRjSr>er|J?!YoKO8dzVC$bFD@(7
zbLU`aJmnC{^VUT(?z(vL^we{CIy6oe+JDGC*#SABdiHAnq5XpN35NcbK12CNu_&*A
z@=qxKL3+uLD3gnxyBB4D@>b8~NBW2QhEMygKc38|pY`Jxf999JvM>F~5h6Q<;-Wl|
zUt{O|A3tY&>mV1u>7JOM<oEb%>!BwQAI_EZ3NP|Qc_BY#T=S@x-||!BvrpQ2Ze94&
zj(+fK-j}#1w-4cS|Nfu$&_3y}o*md<`lF0D_2MYGJy*64{jCR*S9|`*e$od#X^%&|
zklQ%MSI^$8TR-*mkRF>ySwD6HjUx{?pL`s;$MXJ9UN1j4zV_~ky~htxf8RYgACOCX
z>!QE*hdlH~J^A?;d>WU&h;!t2-|V|D{Q2DgJVNVsE|Cw&|M9B5dFT;;;XX=S5U=P<
zzW=QFu06f=U7GL6ytib>^c+8U;Q#HHd7yJ5{jx6g`sddb{4PFya!zDN<`2vK2YLBz
z>$G3w3;W}hKG=75WPIZq&-(1AeRKZw+&)_`zWGz{SJWHdcLvJ&e<x?wdtvW&oS!}S
ze%JiwwI26?{D}2x&kpR9_2b8Ucq5<jLw&Jto-2n)PWEG;%*)QiLF2I>{X(SA{DAy|
ze$X59ksmrISg(0Q#G`)VFh135XFlWNSv+?B5|_nO`GEb?PyTFO<_qsP&1WC&AAN-6
zhW3Y^7DD4^p>>kq`HS2T&v>H`<RO2!@B9v=&(J*(bWfFq^sErFcW6FH-Vm)9S}&w8
zS$u}|c*;Wd9-{bJ2;J}b&X6DEf5cz>vOoOUFMgB!;xm6sKjg#4(~tbd%lr9Reln}<
z_hUnOjYsdH{Dpk<L>bTSPuUqBtdHMQ_S`<xSN*lO4*rT<^hmvYTzn^&JlA;gd;R!D
zh=2J8f1|zkAkG=~*K=hU-t!orUOUgY2b9OKf9tg#d?{O(et0!Lzi1w1>oSl2%FsQM
zeIdX2!fwSU>!!ctr8m}NJpNEWXncNOJMRbVH{L@uZ#bTDL&QrKS`Xxh@d5P@=f@M?
zjYDpH>#uBFe$+l$m$G)y{KkXUtv|W6XP?R;8lU{e)sEcsk9^vj-*fZfQ-6BGFVSoK
z(VuXB>(O32JX<&YA&+yXexbbPW&g%8uJyBna2?`2|3of+$2#=a&bXc{+XwPnH`I<C
z<P}e>N4;~B{EfW!50V3a{2+hHKDE<N9w!gCPIm1)XFd8^r}3;$J4hb;W?!rm55}=>
zsGo7{v-Y0j$$f?A&XMLdA3Z0x^@L~~>k0cA*EsZ4{AWk{iKpTNdxFN(pS=9EJcAz5
zS7q{>k6hw1ls`azOg^Jte&qMe<R{M0+KGGmlT&~Bf_XeAk8>)1tV?@5;1fT_F+RD}
zo8P(C@65>;<(0l4v`+Qzchraf?uB;7lV`h+zz2WHpOVkL0`%`mk~c*893G(iAbNy9
z`e=XXGkcPEIoJ7}5BjdX^Pl$?>@Gz2#q3v}75X=R-E;nipW+|LVLj}bzAMvnWzV7g
zaGrD@ranaNLd0(tvg<6Q$63g3Lk#D0zek_w7i2dfhTnnG&n#r`S;%gS#lc$+9<b-H
zrzX(-S{8=-uYCRLe|__Wi|5Eezc1$=*1fHBAUiG<?UVa^|L!ZE*|GBsdt;aO$3Cz_
zdKvDkeb>(OuxuV^9&+)c{Gxh#&Mzv{S9WGS<~NUZS+{wNuU<Rz(=YQvc1VBl1jRFX
zOz02DYhLn-4|u_!cJxh|eBzO^_-0?&0l)9LxI;hjLJsRSu6S-fa^f-M&v;pA{8G_(
zNLfh#L$nWBXkFzZy(xsw=e|Sr{#xED50<xzD^T1K59yn@CXR_$<WX-N`KNirP5OyH
zeAsXO&8r_&PhaVeJkj_|oH_rn53cOiZXS8C{_+xb!2ejce(aDwlY?H#gUCT2=>h%M
zAHV!?*pEEy3cvW_cjQ;v<DcJRhwM~;e#Gx!liz-l*Wac2JLW%h|5@KlUs@<fpYW-l
ze~;RJI~S6locJ-1zi$$!#I3D=yw}A2Uh0-RYwT~2e&w#_+@N!>-nRGsJ#+FI=N|LZ
zPyWt(LV2M1`8WILp3yz&mv;;uG-Tr*Ip}?Qm?!%?Zhx0$9ojwAy=C|@%g>Q}=C==Q
zaCL63{O_sLJNqx*$Yc3C`YjLs@X(Rh{Q2RT$?nf?I_<QNrYCC-dZ^EV+jY<RyDoh4
z!}ggy7}q@&eQ}PaSM)?bJjoNB3;aDhzs#=mmv87Nzwvj${yx|D-1eKk&|l-pXRQlA
z_*1W)_aExzt@0Lp(I@)OUg#5^_yzfn_kZk7nZA+3{?S){N<M*S`Kx_UCYNzTyTXrq
zZtvUi&yUz=d4qZU?x4RHV%OHoj`%a<;Kg{_TOa>VFPwMy&k*&4`q>BVm9--mzvMl;
zc;_4!#>McvSnZ(uALAGo$`4Mx;kJ8@|It^e@7nO^e5k!~q5gPt&NDCh!uHl7ju@Zb
zhWBCgpZ}vL`~iPVk3-a+-uqsx*G~O*`q>pz63Fj}m-0D&N}Tb#8phS1p0nG{dh9%N
zv%5Y|!tdAl5qdAKTA#Sc&$4r9K7NhgArJc@uXULx3+<QvVNdiv%)8tR&|mkA^jtm?
zqW2sfMf$>j<5PU(AH-Ab{l0{}jXr1(_2b9P<9>ymxbNUE_!W5|#M42aZ`|5+$Jo?+
zfBwjGW&RhR{D$$#2etG539>)(8iyY6v)cKdlsx3`DDtz`Zyn?o$LJCNDIeo6JP-Tx
z^VX-m{`?@M5A25D@*E%T$M_Zg(s#D(#rJjU@yM_Hewg2c##1J@JcnHD#d{3vCBJy=
zIX#wN@-M#QVBevA^5^yo5A+*q=Q+EgzxI<H*Zky(15dc6NBe;bzW2lNn>V*R-&{E0
zs}sAu*P|V-dc<oNA27IEJO1zneLw2^=MDE6cJjNccWbAg+Z=xUt_v+cIsf}C<OubN
zUgY&}(y<RLzUjxQb$M<*(78(9!;i>Q_#tu2_c8Py;up$`ti!&DtMrUM@jK?X|MWo~
z>b}A`AdG|7!yl24Uo^gP=$C%vu>W|oe)<>2b?X&JjLXjCAI6h^@=NT){Px3slE?o0
zzKg$v{H*WY=n1)@I0)sl@^O4?Z$3P+SMwQ%zS(E_kp0)6zr&;a$9@@?z0pH_Lhn2L
zj*<PPU-%`TdF6+pfAyT-vmX1vU*LsYd#=3rf9!KibGv!0-#Gjs6nBlwZsaB8rpM$k
zF8=vH{oFG-N9ac%$*rB=jbn%61;6d@Kl~jDJ;sB$Os~nM9r-=S4}VP_`wsbc`+*Pp
zuB^ZDydMhJLm%V~P&^FF#>F4Gq4kIRsGohIkMb@Uu9F^V?;cA2?p(kw>5cmmdAzt<
zuV~%--`aiHmtV}E|9uegh=0R7`*#k|pMTJg|HI=W>)x^bk&Azw$Orj*{h@q?{fK++
zzvv&ot}GtdXL`q<@#FkDz2RT2i$1##asQ%Tng1oX_0tz}Ij=$bqU@eVKk=OZ<)7t~
z+SwP+>7R9ouf_}Gk9Pcn^IVpm+E@L|M=tRZPhot~jvvuZzYw*@t93&4_=ogBf9N^X
z-+1cz3;tGqtezjGPy8r<#Sh7|=xMlL<iZF0694205TC|1pZLJPif`fH@$GeW`uyC_
z&C%{0A-|O;>n{!&*Y`g7kRQt<v=eu%A3x%P{EFR$XkF}H959Y^nepf&y<r#h6S~L1
zW9WCB=jj2xW`F#cdVb${P(6L%x5y=bw%?w+$5qzey{+|WXFc+8_c-p6$!lH4s}$mi
zzS0Zlv@E^A3q7G%^c}zC^nQRJHNN?bryu<xm-{UGsD0R9y?N~;|7^Yd3jU$I0NTfL
zk$#5#-4|vd{{W5CQ8Zp5<j)JC{bC3BrRVkovSa(h{_VH^#<TD84*j70Wk2@KJtlh+
zmqY%H%RksB_DWwN{Z@w7i&y@T{_^kQ4&Lamb!!j#BlAM@K<&(HT)Z1kT)?;d8sGAG
z&&5&mD2un`v5(dV&95JSrayW419FnzxZ<n!ko^3VIH^CmmH8|4h&#s9jy(9{chuW=
z>%xP3DDpt_dX8u3RpW+89{tTn9{gHQi1yL?w5NaS$tnJmQ<?sfJIou5!w$@AJ>=Jp
zob&{e&%FA}v+=H7$b;wR(Vsl@&N|3pee}%y_?IVYhp$3tUC_AXGQKjjF8VAFv_AVo
z|DpHe-dlTr<hwuLH}H>zkbGGvKJdT%vG@Jzy+4rGK=#e=%5#ikzxZwXFMgX}IYfTF
zSR@B|L!^gUC@z(X)(Pb!P(3t1JGMS~CFE}*{f7J_d5w=p<Es~^@UFkIami!fjOV?W
zedIUIL%)p!<qe@6^oPCcZ(Qxwhkk=z>o4CBPa*#B#Lro`=j;?u@(X?*Iu}_t9^L2h
zSJuH#X>VQ1&i&S>Kfk4n2lBfAaIfGV)BZT8xqr7V<KV~rs&@2K+5L*|VDLgu^7212
z`j<y8G&12g^mE?Q&px=<;V=B%IlFNmDF2t|7zY{;@<Z<Z_yIiOfn4<4_;_`1Lk{gd
zhwc;Tm3?&I<=z3G^n$*~W9S9>%+F4YXP>RtIQCON^0<d{&I{jX;2&n)JJDBmjZf<%
z2mK<qb!&$&`bnS3iwEP-XYI^y{p9Aa=r2_7T%bL@h2gr{o%rt@;@;GF&MEvhKS`hP
zZC>wV=&5xm(|>vZ-5cT^FZi`T@(*(Jlh(uD_2-`;eR1ETo$<`WKPlT^{@8rRF+RP}
zo<C(5=1~v%8~%l!&<99Q*fl?)Y#+=+PU|F}=gQ{8r+wB>9%p{-vyflPLg!&fkCo}S
zJP<EgNUtFMg}w`fS>trHUOr22=&${eFR&l`Cl26~ALJMLKl)7{=#_Y6T=U2S*cbi4
zlljT3zwzl0WY6@#{N$6TYKISb1b)ru{Z1CP4QU(Lx>C!MElagDpV~H{t)<&H|84El
zx<Ko~ElsDiwJ!2ad6kytmga3o_8Z#Pdd2BmEb~<8?V6g#o;bR9ztvw&3gLHqR<`E)
z{#W)Hx!RPZ(Qxe#PaD3`8lNSNim#k}@nsjD@Ojdxcx=BLPW<G#DM_Pa&*A5-zTfko
zCDq2tT5ZQiTmJvCcjob4P5b{prwJJ%B^gUJNMx$6i+yCiACb&mM5ZG{hD?cwQzCIW
zAw?8rR+6Exjuc5Mng?l~htoXQ@7HyozK>S6J$&}spS?ck{ZG%<TJQB<@AX>49zJ{T
zQ@7j0Z%>)*lr(xNH@ijc=EJA-Q0a2xuyPMv*?oW#`qRN{T-?f!I&tbn8y{StY~|$L
zhIDeq&Q(bzy3bg#r2JW{zHt09-Sy*D|HgZsc-cAiD@A;n;?&LcKWlf#C@1MOrBBk<
zovD7}H|}*xyGiE}ci!Ch-Y*t8akInZ%ZDCt<3h()j&FSKj4vxJaBSt&#}7QAPsKS-
zP>v~m|I=glTH@HscHIl^tatrVC+W29`Z(-9chuXfj@sjp9~?J^^-H}+F8}n*79FNJ
zwsX$*`vxwlKhX)wq@UlIijVOR%VB#Cx%S8prl_dsH;y`k+Kukte7rJ6zj26NJPO!$
zpMmQ)TbO_I=}XIfJ3`sUPb%bG{bK9yRMa#zs~?%(LlxATt>UJs{&@K{J^#JhPkQ;u
z^_uT}#Bs$=TuizivxnZg^3f+&IzgFkztX*~Pq%%y`X_zZaMRU8<~c#><}Yae(0{#j
z-?~Rj7YUQ0!+DDzJ^Z%Ea*J%?^7dUzj~HL7C|hWY?!<ZUppO6PesrlKTiNl|r$_d0
zU#cim47y+XweLTs!<8K-<Q74{^>jP0-;r!94}1L)_Wa4~`J_yh>*JiX^u5<^tTM3r
zVr3g!&Ft|~tJ0HHP$o_Pq{riSpE_yzet++`#m=_-WTx(GoAw*^{U+U?>G<(~-n$KJ
zJox)q$5tk7KXHpE=<iWM?HNzo8MNP-wl8VVmnOBpT`u2v>!Mp{I<_!r>(tcw<e^`;
zYuC2oXm#{G@AsMg@El=GIz;t*y!eUs`3BK5T0Mha$D6h{wU4^_jrzW>TRY_6ad_=v
zKOVMR1?9A||2(8g%T+4r+`jqx%TFG(R@ugr?yB3Uc<+@eTUh3wC%$*s7b}%89v!Z{
zx!FCFuUn&p(dfW_nJGTr<6k-NeYRTJ%DDGmzvGm&_kn`O<M;g@zk29)du_PBebXt$
zj&0QY5p_Cg*ZY3siCTxK{e|sEo5LUY@YIPpMYiyQT8|&L;+hT0G@~9LFMnRo!FtsX
zTYgoI$ImU_VYQPj4EnreQteHDry^?q*)o4@=|`r%Clhv`@v8Smvp43yQ*N_Lmtmiq
z3R}<gbN#0M^SeY)dX1-T<)*)Hl&SS9?e6^jt=zV1(j((GIJPmaam>`Zb$eb!^8a|>
zJ>7Bn?VK*tYLA<#gvr#w^l?5e|8Bo4>hss@de^J`!PiFK*7L~u%68hi9$U2Bxcud<
zGgMIO^FY!&E8zD$JnZ#~U;o$z-8ZNw{Ns5qQ#$NCYf4@!=yMlIvoGp=;LO`cE>&?c
zQ!}JfmD&g7+%ng(m41I$32T?BamjSMVfD&%JzL)o(c`x5y85Z{moGW~sc}xG8T31Q
zUp{;H4IkE9<=D!&?mL-k=j&l_HCV963dc5fxbmS1(<ZHQY~j(zb!+zYpk+>6#}!lW
zJDG}qOy_ILtsn4<3d^9^*-5uIbzKy<->}!=ZuQ*mpGkhO?-jV|8#LeVe&99l1{cg2
z|K;l+IDQ#4PW`><Y)fbA`y=VLYwL9{ai`mVy}sY|<k#-43qS4FZREF3So&=TyRg?y
z)OY(X#o-O7)|_$abQKl7zIQ}C!ptb(H9n@x-FoddKQo<g>U`hJubWPiFA9hPqJStM
z3Wx&sDp33Q=5x#KxmHC*l!)tn)41MKre9H=mxjHLO1B>U5Q+k#fGB_hZ&nyHWNNKV
zPPUSMM#Z4d1%|axH{QJmHGHS$y(?7QjO+LxcRNw*=ym_aRC%WVz8~}2?{_GYVi))C
zkm#f{DS8Cm*Ytiwoi_D%0ORuO_xI7H+GYAadAjXq`uWOq+kfHx!izq-;aSI4CUsu>
zjVrF_Go#)&v&El$Us%4c?C1N^VV^_HmUWWUeuVw|O=0^J_xq<|<4kJ3py&T$|HRAA
zsb8s*q65#-*~+B;or{tZS5mKIqW*qdru^CRy*IBoZGFFtxNUjQ$t_O(y}+>ND_-vr
z=;iIY7u;Fz`lU|XwC#PmxZ@2v{;yx!`k(t8lT&0X_c-O=Hb*_4Q{<O-9s16dllpH~
zN#n>)ue51<^(JK-)BSE{(D`cGUVm>SuC(cXMdd&4`59H*QTZ9)_}m#^R#@Of&8Xt@
z@{{iGG)bu+_Pwj5)QjqQ#!UOM<?k}<ai;q`XQt~XWgSKpM^y9JYyUG{e(BtgS66Gl
z-U-WWeGV{azf6x)Qtxm1jVr44;^wEMpZEG52V3v2$2GpD+KKw_Oq<?sbpPo-&#3z|
zmA>wVioYx!l9LxTO~o6R-+wIK{las;ndO*bN%@`0bU#h)Ke}Jj^Um+_D5-X)p5t}X
zJ?`gAlGdJ^E~fLt>R>u9@&DHfB;|fs*f^p-57hnrwQ7>9D=@t5y0t^wZgsMaLGN2+
zn|;3v->5L`=ygt1)aMQ9YHN1iQsMq3N*JvU=xYl*Oul^R0XHsm;-W3OgpD)mzccGM
zj*~uYxasO4^PHqpf38W=X8(SSu5VPoABmrG4_w)OfD#?DTL-`UuWlFjbBIy5r}xK}
zwAbG!(4XHmmG;sr-S?l9Qs4CV$Gvp)TORg4SyJlB`@Vm?@0+Q=?{@FC8><YgzF1|7
ze&=~o?5=sc#`oXeP^{pzvj04!Ny}9#C{6Xp?{^N8W~ZOeXF6?ae8le`3OqP<V!b_n
z-Q<|&usw%dd*laGlrS0{3fmO6?AE_$?te!yzh>&N)ZwYzPKC`1Q>AwNn_nTnVt(aR
zPC;RQl^yBYsoYd<lVMdm7UuWr+_>84U8nzpGX|c$t<HGIi@Vm0Hs_0T8h2fA>MUg&
zpC8)m#{SQ)QnvDgXRBSZy5BNoD=VIUeVNPpFI2YkyMyvho3&rDGR<gLZTj9O)SGn8
zgR7T0w($H0Uo0s9*nB5ldMUT`(N;~LxNn$~F0Xj~oce2C7^j5(bnqG%xAOAK%eH^v
z>B-7gzPIhUqgtQ)ok}Xv{p_yg->!4VI>#@+Y1LzBpN7SbDb|=#rNY$B3uF7krRbou
z^x$u%9-6yJC7pWQQKi2+ZTQdiJN%}S!l?S?=BGxPcRp-9>mDaAy2X*S{B)0(Ve8wv
z?-Tcuc|~=PJZ6cql~=dg{!IHl)~cZVqM**fFVEcS*ve4@9_sXQ?=4Q!`QuAxRX_3A
zoIG0?c3*q{^wj0g+_%zkV_3h`OZxHMZ+!oFDfhR6ljf_SeCE(a%hZjdolNtDmV+DZ
z_x^e(E;9aM8TEOB-#AQPclwP(?BY@2{MYvR?)?ToD_dym`HpRI2i@oW+MAlyrtibL
z`8U-cFTdQ!Fh#HSlU_bz{)Sg>e`=N!7n80>*m;s}ztYt|-S*w;XG#y^Ny>e%sN?AT
z%@O<F)c+G@8%>}0M(v;bK-t2m^)c<Q#qu`?efGs1Wh*oN+y?t#)YQi#tp6<q4>Wl9
zfhA71FzkKu!y2?2wPE1~Cn&?VXS!b4`4Cq<+pZIdS6_F2({bsVE$zHgar2^k2Cj0F
z&Y=56((TiIekktxrk<ZL?KkTC-nzdh=fC$z^OdU|)4Y3d^Zgnhx7M+hanIMV?b-eu
zsl=Zh&jV-cIEKAXmZ^G48wXS8o%FNaOBZc;`KQ%N7>^EpmaiB);-ZyG7>^EquOr1S
z+WV1izjtOT-_$<p=J)*5KR*1HbJwY;=++MTexHxm?Mz>fMdd%!&wqt&-}H52-2U(C
zdS%f+%gl9b;fzO4m~%?i8IG;g??2d@PRe^m?3bD1-fGfK&+awIv6XS3C;1(xq|C3d
zad^FtPq|+`bUR!2YducAA5o`m-A|*|A!>hq?_2BbzVOnzvejQMRb&gfk7Am-PNvgd
z&vlt<KdS46r1%N@yF}9J*>=8S>-j!$+jd;?`}}^?cBb~B?ax+j*R_24qI7BcxS;#D
zZQeu`uc>*EDZhTtpTgSN`dp$d-wpB`=bq&cx^8C8O^$7Jd*3<g{A^u+Vb^U^>ZLm_
zb&tKbuxIrtjxd=zB<1=k-8kH?qxAV(;jPm?dG)qEio)`#$3~B-)Sz^c?ab8gv+1_$
z_d682ojwmFy|V)P{aClO-}4>aF6eV!LH7f<c-`{d_6y1X`|n`xJZnl`Dyr+)O!+bW
z-jwb?Q!|9$(QKvZ?>1*@Tr%CR+wmaj`L^E2(c`w|`$0*IQ`gZeo%H$a6<^TvzogmQ
z_B$H!+IIfS_e0`3u9%u<rs7ZcK4Q@8wxrvq`<<br)aUo(sA&2;IqY-Qw$+Pje)!!F
zyq?nxs$ZtR_oBBS_BqR_>Lq<YbBizO-+l0kPrrYdZrbm9@@tpw^?~1ZunXIN>bva&
z>Gngf@3roTN0=D}+~z@C`EI?|?;9p19rbe#e*L-Wr01K`NpB};n(sXcll}KMf{v?~
zeN@*Yner3$b1GhXd6k>~`z3De_dj`bxAB9wDqH#59*@@fYQr{VE7N^WB2)FbAD3;6
zdS8h8d%Ub;{rcHd+Uq<&s&e|}b)Um;Irq8aCjAhK0-}H@fC6!U*CX!s=x0<k{k?Yl
zryFln$G@O{{2m8wvy1w9Ia9ymdif1|A2+Ic%xl}u&*Q!yfKED1olgh#8}xTZy&qAh
zqrN_h+h5$*1Jd5k-$(L#9xUnQanIMN+w=N8HtFU1_YO^^quw{O#b2h^r`PjKQLh*A
z7w&!5u+Nuf%Q{JFKf?Zw9JW7kUw?#+GpY5iZngcH_Is>VLeb%`=0RM~^O$NU=zTA-
zf9|N(=YFT6(jlpTw;<?!k4&GJy0!22+^%i;UgzCOFE_P+#pPF@Ut!Nz_}#-Vz1}~R
z^342Q*r4Md^>Zk3`wM#CJgN3k|GpKMKmGatpy{CV)wF%w^OyEa_bV#@2h88++y>um
zP;oJ;xTEr8>bMqFd|rOC<#~MDo?p=8HT`{;xco=`o#(jxvG3Z-xO9rE9sT>(Vbfmw
zpXu^+pRY6ZyQOV@O^;Jt*XO3%33{F#)PBQnKd%1a#?6jtnjYt{{{7wu(eJ<5k`DWO
zUsCD?y$|eVA9kLEtrzzAM!m4@2Ce6%TiEiTaU{*&^nE1#dO_0R{%&#5<6c;M)AQ5z
z>vmJWC%WmLl<PXLzprMh+)WqL`AJ<rru%KHe^c`>F2APFM{OO~xZ_UuJCjMRZ_D+Z
zsW|;!2m7^io0nYI2(K@5S^tG9DjnSPPHKMK&kH40hp6I;`ujnC;|lwnf86zKc@Hn_
zd3Dn3#r<Bdw0FXsQ>xBzL<c){@Vo!&cB2M7)am2iTO3=d_s5nr&kMQ5OSx&fwNJh+
z$3d?+O+T-IUg^@*^;%qh!~X7*ZS_pwe{%Dm>GzM*o&Ryy*Yyhf{;8?Bz4o=FmM86Z
zNmJt^e*aJ)uJ_ZTy8k79{<xo5*ru>$e$CY3scNa*PKC`1Q>AwNn_nTnVt(aRPC;RQ
zl^yBYsoYd<lVMdm7UuWr+_>84U1$BcmqzXOYs+zt7k8~0ZBC0<XRqDos%?&Ke7NzO
zzYaPgH_ukqI(yKmN57quXDi2_eaaO-6y)UD&bDXOKjPV8IeDfT?W#@R+jG5tx;X!}
z{>m2KbkCSN6$bWj(xsPjN48wj;?b60sdTyDbNyE5ee$*=^rwT@xVV+i*t6cO;uGg7
zp&1>}y}^E$4VuzuuJX$>tNv$ogQD5W6nh-@<qwAspQVT|Q@m`#tNZW!+-fK3G^Nkv
z{ck<|!Xa~1QfR85_}zQs8=JoEy->wP+D$soEk148s%sWIar3%aYl;tUzS6OkeMaY(
zUp8WmV=JE<x=+1XTh}>3IlD{U+~J)+Rkrfp-1eP6dGrpIbpFuytS7tNJ=U>>VfVRt
z9ok-2apXkh#;|^=H%Dz6di%rc9oxBe)x}rbGINm=lr63vIDf_c-#D4(>Ke0p>^^d)
zii?bYSoUxE^wqC@FkMAOzi|xgaoO)Z4jrmY(Qh1L7morBU+6WtQR$}~TbTc!w@zGi
z&3MNaZhH3CO{ESS?g;<T!PKlawO@JpHPs(4zfm0*-28a8pY-y$uP2kP$7LHHs@rtl
zLMJHG?bqS^H*Pw8>l`O6({10a{<wcvT0Hic8BS2T`3su=pXK9^epi(#5+*~3^_v@=
zwjx!k$QGWpw%kdh3Q8Ac3vJOoTjI?0ya~Er`nA``$8SB|&h5O&*8H<yIjHZT-+om=
z8TL9W=y5eJdwrafmbP_%d0i*Dl_yRAq{m~+{P=G3r#Ch@YN5&&+M+vg2HoeAV(+$3
zM3s*|!oUB0Au7F1`w4qr!Ee2w_e+v$|MYJI$_*Q}%&~=j$4BhayT9h8U)b~iOx5%1
zPu%6E&L{aj$(G;kvL#M*7sm6urs?ws@&DJ)dn9dso7zX+{C-jKlwPg+Z&y*#tsU|^
z)GIY?-H1&pC=adn`hfnYZdFNV(0(PwKIys!jc>szH%~nEvJFZoIs};qP0ky+u*0mi
zD$``Y%oH=-|8(2`<g#NokGNyCii>{7$@cri8)hzb-rH-L@}gfobi1V9SMrKW??=?>
zq@AbwjVEdyqW0(a{<z+5y8BbM*4K41o%R|(Q|){02T~sHIm~o*>-POIqn<lsons4a
zeJ_)^ZPA%HZ+<<e+t)p}I&ss~J`~k`J-^5EsO+MOJ5zqbK1X0`9%RaIy8E?PefAmK
zXv^=<e&h5Sr;=WtsrBb|{#(+^(;XMv@7GBDc0Zq~&)?1Co}V${=haSF#`XNU>2_>=
z&LZsdLi#+A^v()Ad%z=|hn_Y^dGVfdW%u6wxcSO7gMM!>==Z_=erHbXcK$Bk&a<ZE
zrK;TU!rm1&6)RhL+jWojdG&#nDlTSfhMZDk@Tiyen&sHamN}o@`r+*h98=8HxMaHB
zbiccssru8OFWYeB{=<~*^!grs;>tsQ{~_x0_gCgEJFEJoOP!L!*|Voi{ot3Sj;;J+
z{vpH5v?z9LW8Tc+cQv|bnPUrYnOE`ok*^gyaUEA4dHAOJHFHL&sA($x8Dragw_@uG
z6_!gT_s?xouv}#e=N<g^HSaf`t!!u1_oahgKL@q9U2nhfzf*H<*+(vzsnVt2{ou13
zZfUmZ=?@(@2L1l=p`BKDYVi7Jj$ht&_azNqe0s2Bicj}_yVD)Did3ffQ_mMCPZ{uu
zV;kM(gDv^#{v8&*zTfra*RD#>HPZ&)@~ab;e%rw=-RC7#)1%IwIeCoI>GnPAu=%>~
zdYb+SGo!#^HxKXj#L}Nsrs&q|sL!XP@|QH7bp7%#Us`!#t>2xfskaj}{ppOE9s5mM
z=?IgpL(p;cvOhTgllmP8OjEY9!h=VDm{&AeMMW>YyvpCdc;BlJI(m&VMYs0%uXNV<
zx9;AqY~|{W?=GwIUa2BmY3uiX$Ne_x#f!)8R+MebANl<Y1=sFb6cx8$xMk(r{nn_c
z$U5G)MVmXX`)s8$%~j_V4Lq;fI%S&lYwI)ZTI5_c!AX}__1PoU_xNv}xJf^RqJStM
z3ZTGVb9evu^V5r+Y$g4Siu(N|Q|WZ$z30ZBjn5xatm3Av&!u_At*&YHkD6;&s;HQ$
z_uiI1^YK3CU;m=Y6q(nipWltz|6}`{k-F)#xhgKA(=NR7luv)^JwxSWm_hHa?9z{j
zGyMKOnp8Ve$C0@FX6pCRsM^icbCprGyY{Zir(d@2M`bJdd-t|7>gQqH{6)NNW{tbB
z&zD72@0}&5E`0caCCZIw&X_#m<foS@VP<uxy+_pt+6|bef--9sN<upl_WKZFyA$_u
zCTyHZtrz$A>7_lhe?LL=-C5tL-=7hG!TyeIQs-mP=ZV4|*G$**``w_V+YOn$?WoHS
zUg_A%52p@Yu;Im3j;+++FAICV67)VurrxWbQ2Fw~O&^=%2>)7#pyPj8+4lWA7UmR{
z6xKRsL)Wi%+pLntyAOT(+*1c{RkrcSBfo0e;;}6%EQ8Kh)Apvmx8UX1bibnV|JC(R
zO&;H9rHYGD#T}KObe}(Ss~=T-UVh{{DA;w7=p}lMJ!XcJw0RKrJ9yb%&ujlPU7qgu
zSx%VT`GjddEO$yeO^;Jl&n;xikLmZ(Tih|VOzY34D$_JQ&SCxg{Z9WE9WQ;R%B2$>
z+qnF;E;m2(&uNaWeB#YR`uDm07bhx)ohM=I-J9FK^Cyqqp~5n3yFu&y`-Q%r5BO!f
z6E%azku-bL-^c2;_NH&AwOr%~)6=2ip5-dGzki-%nx^OH>DQn7&6mgj=Ge-cuAI_;
z@zT*w)O6E5s`t0k^%M3xD(Tj9)5Uatw`LV*&VKE86_lpq690d#z!Nv$_(Jan(;ZuB
z`uz2;Rg+v@fv|s1BkA?(e>vjI^9OBkN-BSP|Hgi$>uh$SqCRg(d;Q%BDfjpHIdq-S
z*A~Y8+?g%9gpISt6IGiXU+D)YD*BDX?RU<k<xhEG)i0wak5fr!RNu2pnxAy9bEJOT
z%PNkXs6>a7p@ZN3SGU{mxqhqjK6%@*m3n_{Nsl@7g$*C>Hrw^@6P&Uz_v3C8mpG>B
zrB_tv2buO0_5DTNpO=n)%WZ!T^~G^d)w}Vr6{@5&s_)SR^>gFZ*Ogg%;wqIX`km)V
zvAg7gg`JK+YOR9Jo0h9JGk=W=N>lx@_3yF9^?eh+?U))L@%x7Y-M&9&)N^O7b4)Yp
z^^mFWeMQx7x_T70DQwxTf6v_ij$(ezRJBy~RBorjW`(IzJO0hDkY6#saw?~wFu%%<
zbnR4bD!0k7svQgSdv$JHZFKt6zt-kkdtZ3-YDXwK$ULxb9i+blhzC(X6c7bO0Z~8{
z5Cud5QNVr$R_wO#$McVvtAwJ%j(K4Jj4qiyXnFmsyH!YSbWC%4oqO9JQhJFK7fa@7
zO0Pu$Q9u+B1w;W+Kok%KL;+Di6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KL;+Di6c7bO
z0Z~8{5Cud5Q9u+B1w;W+Kok%KL;+Di6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KL;+Di
z6c7bO0Z~8{5Cud5Q9u+B1w;W+Ad?F4UA7y?-}TTfOIIj(+t`KG*GyfjHr;>ik3W4f
z)|J>nFMG<dcO!N>iO(<f`jY<p@f3ajTA!e6RKDXc-|;U=-|-K+KC<1O<Ek(--#wM%
z*k2#VL>GU$Wc&Uk{SyU50rLv%KC;`v@3bqM7nC)&*1K@`*1PBF)W3_o_@%82&waX7
zUXLF}-ah8I?T#1C_gGpG1w;W+Kok%KL;)`aPD^dRr&6DN^Ms;<%!ByWLDcu>z1F#u
zivps6C?E=m0-}H@APR^AqJStM3Wx%tfG8jehytR3C?E=m0-}H@APR^AqJStM3Wx%t
zfG8jehytR3C?E=m0-}H@APR^AqJStM3Wx%tfG8jehytR3C?E=m0-}H@APR^AqJStM
z3Wx%tfG8jehytR3C?E=m0-}H@APR^AqJStM3Wx%tfG8jehytR3C?E=$QJ}C*VaxoQ
zsp_dCQn{T9n-!)??f5spLVm^k%Bh@!!u%>b(zR2$soW;Rs&*{Q@71|+wb8rA(zw}M
z9oO{!O?gqTdHu;_&v@+DuN|T2kPY*oeS_~WIq01EPPS8D$C3`Fzyrs$KDYNr#VTC}
zYgMd80Z~8{5CuemKNN`jyxSJtueq?xL))KO;Rr<sKl32n^SDgcx8-_e?^NsiCVjNk
zNh<aI!j}C`;>?T!aj*N#^io<C1w;W+Kok%KO11)Wyf4|~FTKx<0&(BxlJ?~H>}>cw
zyQIzI%<NA7TfMzev#Es_uT`dK>b}~6PM6tp_uS}2MgNMGEK9HFufmo`HEwbHse4+^
z)va=-IzgHAS#wF#19RHGyiG;Tu>Fquy2jM+5$n#XfB&V&ey^gUe!oBNv~9;{Q^zgS
ze#7pYe(PO-*2#BtTC-7^X1eP%)AKFtx~*~IyBDp>U*p8ZDub_^UOTT`ktv4NLFz@j
z|BIfYXO!m?am}Z&=XrkX9r@8C=QnsQx5yOtJmbJyTU1%7{4&$WEoncy?gyrO9;Mg!
zd!887F0S*PlG={n{m@psxQ<7eYDZs(l8)xRq-@tQ=>5c`*~fjKD%;z$b$^$*MFCMD
zdIe11$JPBu^}Aa1N{GiW3Ix5*3u_-Hwsh)vo#&_H`|s*z+x?epKcABRnR)Nh)bHx1
z{ATLD(%<bRQYH$-t$^SCFx|LJKS$x_|M9wY%FL^sn<o?<;+O|+>m=LG6K(5{+qg)+
zC{Pj=@cVp&*adqY#I#-|aju%^bJ+V@QqSM_c$2P^i8&=cMS-OK9-rxWyza}VTb^w*
z&hL2HcKt=Xrq7S!@}KU!jk`YjXPQCJ3o><|UfTWRy(Uxd8JX7C?>?FoJJv;9$1_v=
zfT{S?-PfafZZxQFroMY%>UYt&{l<M9Ot-zb^`X6V;|^M%cGG24`$1IeH&cGt-!sLi
z#zFjKmjXfe1>HWV8g9SSd*#!P+w%%hJ*SEvw|2-U9$`2I^yiENr~mi&-58H_8T2~9
zt=IbcDCuc8og`lr5Cuw(0_k4Ad)0SS^3ao3yzuez(nYp1=y54&_L;igNw;0o*NH*z
zBSmHJwT^<8yREx)^G)j?w}15Ty3R>@IsFj+wE{`oe<hx9-)Z%$+h5ND$<0;;(mk(;
zyMDHggY@eU1^n()QQKwuzLt!Cb|?_^IS|vmj_Nu~{Dk{1W6<kQv5)rs6x-)*y1xTu
zx_(^eho<*Y@h=L90-`|N3WUAR2^(kJ$IGyBO1+Y$K;Y+TV){EVLF2Oh{66mAnWQ~C
zzB3*6I2E*By3cQW)z9?xlC*E;dw!YL)$M2<_1`6o`}~deyw+7oE7#|n+xSbqC?E=m
z0_iFc*L_yniTb#eu0MUhjVhh)xRkW|rjPSMuNQ*an~py&|CxTCp`^BN`~4j~{;0+=
z?tPH!5?iT{LsV&w-(eZ|I7)k>Ko|vVeLqU#-ueBuFk>%uL;+I@1ihXxZEx%OO<ZxC
zng>bu8}|Gzs(PmN3HrOcZl9_564TY?$d4X5zrky{MM3HJz60Njf$7F8^+f?uU{?j2
zz5d$QU!J+n5sD6B=0U#?r}n#I>VA1qan~6tIikS-DG=3pp<h4$*RuRCN(KDtpJ_YO
z*UeG+kLx-%?slTmNBo%mUWe$L41J^Aqs5nb1^8Xtd`s!I;P<{w((TaO6dxS$YLjc*
zmd*>xZ2LRPQ9bu2e(Zb?A?Wu<GHoB#d9A5+oN0Z6wx87FKzc_(sqLjeT-WPfaY%VE
z1q#~~w(QowXYPM%<=0Fdk*bl(?Nr#TFjZ>DzxfsNE9O^D<rEa=SJ{!Soytw+HW^m6
zV_|-;&W)>$-nFrgBi?PfY{g>7i@Vm0Hs`j^JsX_aVZ37-m#jRw+NIacb8My8xF@~b
zR^9cu-)QjUko(^J-7(E*S8Z@_t?HKgt>dOi%2w**7B(GRIjJEEm{VY5g?c9*I<$vM
zD(QKvEhqJV=d&M`Up_ykYqRg1V#hRzFH=nVahmutrBAlC>v#S~WoKHysLls&|Kipg
zs((M!iJP`vx9IV*K5gZ1gGa1>ZnxD=P;S^b;qX)L+o-%)(0%&J6+T$1lE!OlwEXRj
zd8<{@`TjmVAHCzlbt<V`JT!mxeTOViFzkNkbv+Q)FZJ|wThh`=IX`W>bWQ*1U1q7c
z$oPk4)W<`=aYS|g;pWG09A0*;`?%>9SEkC7emn^p-@c7%pE~}~mCAM|_4w%(m#On6
zQ|rU@eONdDruyUM*Dao?^1a$mdU;ej`1O-?J;GkEr`xZv?NTq@_TB35pWpw%p*6QT
zLFwi%Xns=8i=&Ri?YvL&gWU)B(ktrggG~F2dLEkgXZv-$-{YDoyP*4}UweJr{npd%
zO!eEeU$6UFLCbYr{HBw-KK#a?DZ9AtPet91zV7^{y>ve9$IZPD`KWYJy7X$ttvqU-
z@aNW^S3dQV&baT(nQo8k5nGt)^SQ9?``tIf+L?Z?1ph&gvq`fb-tPWMYp>nxB%ML;
zD+jet`abJ7KGXZ7_}}^W=xou$>pn!%%1vKi1@-TBK9_E}E$1Uy6KBBZS(COtc8Mq9
z&-eR1zKLD5&$lG)_?7NH8MJ=b>le3rUhVg<v$o5*59H(pW!=7|x?MeHt&?<yy<U@g
z$$rnn^!T`~CtLDO>l2s%r0DInkD4w|>hBtU;|hBHlCHhq@krYFM$mZtUf=lDL$~wm
zKWTP)KcY@Q_@7f7)OmS>V;iH^A!>hqpNrDlP0De?Zya&07hNaQX|H~nYCo>?k+|9k
z`hJ?&hkLJ2*NxxNY-Q4Z&-02WQ^!wJ`;g!BVK2K~?w4lU{ojQ2$J9K?R=-}y_oSAa
zdao!`e!bqKN@{sr@6}~${rNrrFR6Cvj!RtU-_lOF`~P0&HMW$yoqy@`H{Eq?iymHa
zy6Khl{IKg>p9h&rr#t_msvoxhwth!Q;`aAFU@x7*&YNth=cP~5%f0UJL|yK8+@spp
z@Dp`E;`T>BOwsSSc-eWKXLyzWdH+GXzeK<F`ugw8`nY(lKkgsve)Y85{p6tS>GomM
zdcQN3PWSoAq@Q2z5??}|dkuOY=GP|=m^AI)70OlyeXb&D_F=D6yy~U9{=DjE>i%fD
z?Rp)jqAEXoUWISVowh|q&8XtT597>nG%UU5qiyB8?C;unM`Y!Oj~n{=idUNX;FsyH
zN4>t+d6}v5$i~y>Jc(*Ou#VGZrp^<iYBwr9!}<xkZ)Uol*M5-n^04FORxjOkq}NY(
zU(x+}rNg$z^{(4{Ka|HsuYOAT-z$*ra{_*k|9<WMyno~Ke8;VKN!^b!wU4{$r04q`
zr=WIvJ3-S)9Z!jC=fB4obX>jc-HyK{H9zQgWORG4`&vQQwQlb<o^ItY{&N4XTh7`q
zPv?n44;XmPwfp2{ns1dK_2{PFW%6ugT<2p^w`2PG@SyL#M`iE#Jtx0*rtTkx{e3BD
zJyZ8<bieFBQO#4g-}OxA(+{C2APV?bz_#;7|2_M|GVFaFTj~YX!^>Vje>Ii%io>fs
zQ`ebZ*9%hq$M+Y5p6^6uZ)(2!^~=1qZC{Q1J{&qFl~LVyjjDgP%_lux)93%D{b%ZU
z6IHwVb)Ri%)8lK}zv=rnrv1ltJ{;A4DSkwOlBht~=YmS2R&n+@?D=)r7~|TnX(w!)
zNv#+3eE_kK_U}aN&$mi?`pFeOSgNAZ!SDSHv5WS5?&z7+`-DM{Q(@0HOxKI*exmp}
zx!R@I&2z%)V0xV6_V0GS!g)|sjOzFJpnl@k)3)~9#@{X9w*Kky>i2PNP3!$CDQUNP
znW_9t&7Vy5H)y;1^QorOe&Y-JdstZesIG5xKVJQe>UqI*-ycazeO<q((~UmrUD$Z;
zB4vtU=U>w5MRlFXb!S|RD(+1AiAyKj+VR>)DSvV7`7Otl*{*b^JD#SGhf(`CwV!X;
zIN|V9?%Sxsl5z4&zvs*Ni|V<BO!@KK->`NskGr;6gMXDOGEK%KX$*Rw(Kh?2?gM4Y
zPt?byxaVKG*O@`%O}9Pj2fbhCW*;>Eq}iK(J}vD1Y^j&t?*U=;Gd(|(dVKM_|FeHY
zO*h@6Ixq9vzogl@>0&xR?DdoBxWxZoD-hK>$&{a{_Y>XUU#n(M?$%1j9b3sMooAZ!
zCQje4?Mr*+2}Os$n+I+yCsX&e{oWr5d%o{>{o|HDyxski)?T~WiHdG|CpACr<496<
zh$^nA_4FH8*!2>3y`)}O>2c}jC6W&J_s&E&(e2QoJ(R8U!0-O6+lBp|FDdVh=>4%J
z%{+69H?H$uxAw_T+Wo7f#>0B>`W-&(db6co)W5&k>M!j5&UEW#S}*nuopJ4FLEDMy
zI*9!|Y6ji!+{S+wKl*pM-Tsb+W~ck#MQ;~BdcC;Q`h1jhv<h(iGtFqD&dJfQr)^97
zy-(?8SJ<YoWq!?6jnt8;+)jnf3R9(a{F`4PzhZvnR8B!*ew7{R+Ns=BZj)hEI~L~m
z>fE^6=%_~_I)4T0FFf)4edal~@yXIH_o_R3gkvlFEFN4iY)c=<Rvz=s*m_t0{Iz2{
z_wQG?>2YuNQKlK)ijBrorw;p{xZfAWjxB7}wRr9P!r4x`j7B~2A__#Kz=cP>@WpNi
z&vcSXdbxk!HaC^NeU0+VR=eNQ`J_R`$`qfN@JaKZwk%S_mnq(O$R02Lr`Oji=`^KJ
z(&nwHev02@DzN)abI-VF!b(SImkzJ>*z=M)7tV8R<(xIYZd-cUI49}6^T>0{pY`KN
z#}+1Ym&wUlQdH}RrQKFKC6y;#@m{aZUky;1=Gjke>va34E1bBP)cNQY*P7=BT{NS?
zG-Zlj?Mt~RAPR^AqQKuPU}}HH@87pBrJ{f+APR^AqJStM3Wx%tfG8jehytR3C?E=m
z0-}H@APR^AqJSu1Qi0w}?!K$|_|=Xu9vxQo7*YGob;}*0=#b6xpz`NKf5}<9(n$(U
z?xEsa6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KGNVAMbNdU_!gY=?9v!B9ee9Btj@sbF
z&CJZM<hGjh?*W*KFS#R<%*afQgZQ=MyBxNRQ!-bNL?sG{0;0hGDPYTfnyqo({NOF?
znl_&11m*vYp#1Nz6nJxLjWVm>S*yIb;nV|aAGvC!3d%Doj+%1#<;BV`t9`wu;_=_F
zQ+|2;zm6}s`TaG@FRvO|;jVkzZ&bGO%CdJhTYuJi6*vEC?v$jw7oIx1%^987DqC5$
zYNORZcU-69Vy5QC!&h|J(D2+Tj;-8w&GDy>8Z*H$#Y~Nhv?~gT0>KqHuJ3yfzi{D1
z$1nLli_q+M%0%B}=o@@Rg{xuecgAqhNG(x76c7bO0aFUpJ964@N0nKiY~xw4r7qli
zQQ0C}dHgG<40-yj+@h!$_54Wpyf0`xVgC+Q(00<b_xk%e>H3%Y=_#;sL7NALoi)|T
z6y2gff7{>-+Bi!b;U7A<=_L80fG8jehyvNIfT{aTrv1kKT)3_6JzW3fJ}q7;cA{c-
zuZEnQ=~ta`#ri8|Ikqy}$4B}n3Wx%tfG8jehyux1AZdSB*DId5zb}^j5lmKu`}eEc
z^td<sC|lWQ@!*1CTlzS*a?%N{mfX;8sbeb()?awy_xsFqY^B%lUP&+CFm74S_~ySj
zw(^bZKD_JgrxrM-nXFwmInj*!xavdAfB)FY6qlF1r_%gyCn`5qdhD=oh7OyrgqhXh
zi3y)H|7pu26_m-D>Dd`c+|QTfKK65^A7&LfaWUIQ&DP&NB<^^=pJeNG0C9WW50LT~
zzS!;HnNCU5A=~zG>5nKN3Wx%kQee%Bg_WANUZHH|$l?>f+yC2P$}jW#Y^Z<qz*SDt
zSa0qb`Fq^A!m*7VPWZ6$zsApY!ZK6yUfK<(z}IbSUbgQgbDX#sE=sB8roc_Jo;&o-
zDVvq;>{GkmDODPmDiUT^hstXn=~1g}Zjo*L*MrB;{I<nLWh*=0J^cIqk6W#>g`=;k
zIqKt^bBfZX+rAf9e#a&MY<~Skzo@Xx*6%h3?U(8CI<CUF-m~YeQ?~Mf^(Vgb=%Hnb
zOwqLd+2TL$=hC8T&(`N`i2JJ__x$L|Qj49a825Y(`u$qnKCXCIS9ZR7{(!A2YQ`16
zt?lgJx6MtZZ(pN?cIaU1Jj>R&gFatm%HHdH+NR30bzG%iqJStM3Wx%tfG8jehys7F
zz|~zRwCdF-w<s$9eQWlW7G8Vu|NrvbT1O~4$UKmBko0{Z><%0D_sFEw3)}BZ*Q?g;
zj3=9n->t|kjxYPQp3u2?xiU?+_9b5w2&2GBSG?D2^H&2@N$0fPo8R}FGu<g^3^Qkv
zROi5TQ#*{jXtoMUQ+*gV=*fnK2hMVY>FHo<e8jIPV6y^+Z3<g<>)$i?zkSHBnL09c
zR4TVqVY9+isU833SIDoJUpbXiP?%q3N4j<@H<jCDSk;b&`Mo+ft~NR+=gxY?hdlMl
zLg%V3#nYNinYs%pKki>EuUgoAj00!9Qt!QsN^MXd?z`gmS`&X!ue4tA`l<uw{+Vvy
z<=tH8m91*U`uSbUmHJKX)2sEiD;lm<BR)Isqo;o;Tcor0cLOi)aePiu_x{VS>hQwu
zMIXJe<g#;r-lYC<!)@zkeXvMPy?jBFUmCCYlla&DV7K9`yGy;erPQD89P#FPot8aY
zp=jTt7mlg&a&8gv;D`1G4BY3*a`krmv;CQ0Hl4gr%QAoZIpwWe2i#qIpFhiK2S1Rw
z^ms1r_w=;2L-#5|$1B#C+HhI9PaNXGo^td+Z#NPb<A5L1#HqJKKI23@AO3G#@k2iL
zj2HbxSIW^7JN)36xQGWkh#!bQ#)C9^V|V_!XPn(?{r*MheCvDfmu-8^o`23O;vjDN
zO+Lgwe%+|sQ;+`X<>c%0ntbY?_V$2g*WJGA&-qS%)J%8tWBiDp@rC3=@^#W*?4WK(
znsVl)PR13TNwd%4$BX!NvtxYG0sD{Np8nmD?+*TRKcs!#5B<2Z{F<9ioW0V4?7Lp1
zKiHAh=~b@VOZt!FA>*pAJN7BcnFq`_(#&hdpM1#p>-8w79)7T+UFuU#J>tcVa*iL^
zQBE3vI%yX>>gl9isF&+$;?eV2N09beSFDquWF6?#=R4(Y`IPG<o^<!sl2RXi8E4Xr
zGv`61DaRg?5A}R~-zAM5=M~I*%2|J;u_qq#A@y}Kp5F|wU%&Qym5SUr!sYzfE1&a3
zovcIFE&BlHGvs5ZlXBK6q<`q4j}ztSf*<@dPWt*G?N*<BzeIP&3lbO9<Dp-~^YcI2
z41cB6oL#Puu)_~&FQNzg9{nL7QqH&$C;sUV_4PQ|H?gNaq(1FI{810Tj5GDIgT%}6
zk?|%?JlNx3_ltk@$BsBi6F>E_XCH?6!;XI92V#dG(zHW8NdNS>@JBs;AJKJiD|hpY
zf4!V|nXlNPH|>yy*y+R{^jogm>v|HOZm0X#)AW~mka30di}DK9Z*TL!vCAEFhpa2c
zk$sW*3W<a9B(0P7$tPaYkhuEo`}S*FzMA>xbprN~evqa<`H*sW`wfE{?fw1KKkvVg
zuiNYK;vZtqd}n==hUB}ogCD(IZwEWb@rHKDhxo$}^Ncjq?Z}6e;}??e*1ldJd&ZY?
z$oj+1jm$H>9Gy9?LHxVXEgtMCXFkv_<vOv4dbwA9U0>|o{E_cP^w3FrUex<xdzyJG
z-1+{o=#iazu>b319W(Ci1MIV$50DQz{;=<2r~6_5g6vbIIp5)Yl;Z$tot%&8`&hQ5
z84q1wbf^F5jb4=drLL=<=Dbj+ZpVBgpYdfLFh4k-WZa3H{*vz&Kjk{p9e3h|#N$TY
zPLGFrtRK!B$k%@l;=EcXx<cAPPxM7E$|3qv4k;%d{6YMtt83KtiI3|J(ySBW9`V2X
zM7kaJ<m<nWv99okole#r?Lp?Z-adZy`%k2?XZ&f8dc;XPj3a5rQz!ND3(2RS=z>4)
z_d((#4(2cQ-TV;`ej)kHYrQ|%v7fk+xX_FKk=DzJ6aQZEqYK0@^&#!D-pJQU`^+Eu
z&3;0C^kM%Xt&{bOALbu+_@f+BPJF~gKd|HXe)><`)FX{O)cxymQI0*N9sHAq#K-(2
zf5iXRJLS~Jjx=Q4na}v4oHWFq@y9RiLCPWdtRwPC6F2QJ4%neT_PS2k6E`|Ty`1{&
z%k+!(X_vUDPq|L~F%KD6;-MYv^!bRL8>#0;w|aWMUY~iZ+tGiwIPs(Fp!>m&c1WWS
zY3h-N<YN!n4;UB58GAi0($v?<xUv60&R0l7?4h2o`)7a0p7{`z=z@RZBdwRC6YUZw
z<amM~@*#fRh+jA2M=!@8`EGXP>(uKr?vVLLztCS#ldt<HjXi1NrF}1bsR!N08-H$e
ztCwy*{fEp4(vbMk10DQQ?-%pbjqE#6w<DkSi4VHv>*e$hQtp+;&W+UPdXejTH~Xma
zY1fO4vrg*i`FecVLA{)`S2=c!m)AHmkFdi(^(lws>+Mhu@kbiE<>NOfbzSu|{c+2u
z+>7)#Ea{h9ef9y$-QuAil%uDce=onZrxX7=u_K=}BtExwN}TNT)Z=`DH2d(3hcDiz
zVAQutC_2bIkaduheSrOn>pHI6xt`|woa-sb`80OqbAHWr7spG89qo`$J>r41136A`
zo<~0Rl;e*y?LfDBl<OpJ>fsksPF&R2?a`V0DjY{RzLBOr#6SLYa-RbKl%orz9DT@#
z)Psy8Y3w+@k;X4^q62Z!U+l|{dGVu9AKJHwdgMcNpdEA|4v7AGntVuo{L&8bQXe}#
zUi@&p$By_${O|r6c9hdU^k$r?M?B<XM}5k%huD*boX=1X5;vqB>bco5j>JR%uqTZl
zbj2P!;-#E==uSM;!%ipm)W@DQ<4L=E{LDw{VTUf*(++mThkxb=<(%jFCGENqzd^}3
z1ZAJ&n;-V`*M}By-pf3K_#^F?^iQwvmDc@p9Af-*zsxW4{gQcu&WtnTe_omDYnL>e
zukc4ccGP43&@b%qPd@XRe7!zth#$!BVQ%HvLFO~|kbRvvAmzkK+}x*NT<}jhY3#_y
zFXi}WyjV}VuB5RiO<bfYrycqQ=|APz(+@~I#7RFO`8w$*cBCQpq-lq=9vAUZk9nr|
zn{~?dDgA)d$Bum3C9PBUr`Ol*DaSACfc6<D_DkNce7eHOeVbRPQMCPo+P{~Zcz6+Z
z)JF%1f9%jh*N^dbt4CbC*NOc@O`hm}&oTQIF^=?$`AIv(LAkD%-~AtdI@#w;k^Y+^
z<Ksr&Gj=2Un;RMTsK|U~Jed!q_3@!y`o}zj%nwL?^kzM?573SqsYiQ|_E=Y>DTlOA
z8h;0!vAy{Ep$8P{Wc^`>9@K~UbtC7!lv6L$<a!mdUKv+(gy;ZueaLqkKk}Im%r~9r
z1X=&+$Ufvo>~wvJ&#iy-Q;(0hy@($8@gni-agwJ0kbFo#bn1GNW_+Prf0*Zbf5>Ou
z>Tz+NK|GXGPuBxK^v8|(r5wKyzl<;A$MK8)(;xO5;=&I^7wUPD_VEvChqxg2q`mxO
zN59;3(Cx4%ACgahi3ef_v142r56Za?Vcn6&-ff=amw0sQx}Ym|dbwK~doNNSJNiey
zKA!X!-LPZ+=%n0@x_!_zaWn6!PkY3V9clVWKBOMy^pkOb%twele%#XJL;SNY$=69-
z^iL<}q2v<>`Cj$)a?<$0FQgn|M;bCNltadWG<NuRquw9d!Jk__T~B@7^m09of1P^V
zdOdx<Fy8t;KtCY5&>m@>_<<Z>XqWcLr{DTMMVfM`=hKc(+R=#}^Bcc9*)Q2&NYg&&
zxsY;*o$gmp6A$^MA^AFW-SxCSZp5RL>(2YX`guT)sRtJ2Uvlr+?VDEL<$F?`uY@K2
zL3h%O563gEvvl35hrM2oK6?N3eA<Ed8KFIYPCoS^dNI!QLvL3v*XKL_=#Ng_AMH?&
zH1+VWlQ?vLq|pI?ZloRjaee{O3%}SyeIL^2D{0y%%{(Tbc|bq0gV>RV_+frQ;-VgQ
z<P*P6uYHjH4nGim7)RZ2*!4v{+DVt_CfxZug6OJ`n_HTB#dwp(UeDL(4QcH4`g%Or
z>-l<`c3DUG)k(YhJRl!?{A0&D!j5%IyLvt9V-KlEn*Nd2iC;*4(%92);=ztMpxz#S
z^nC0o$1m-XpDD6mF)nWBrKa*}7jmA;d?7w`^dh=K{OHvE(!Op-K5>zT*t_XNIdM}D
zdybp>{?9yt%uDP!K2uJ8<|{-8<`?smcp%3Kp(r2<gjaz3;jB}hAJWObpfh~SQqx8S
z*q8Nvoc$b9jy?OjPWE%i{L+ti`f-4??vM8HN4Z;?eA*@LM%_PlUiCS?LHy7UH{Em{
zsi!k3bLr>HFZr?UzRUjnUM}YiI(5AmCrJO$1+q^;uGdLJ@^wFy(+<Qx_UH~N#~w1C
zlv5w-^*Im1o;3F#$fq8}4!^WZIr-Elt(TLA`uwF|j1Os@%v<vD!}%g<&I7R{4T+O-
zo%kVMFMI5$hfca*w|qS={Glr(4$|B|CGAChJ(x~&J>b<2`>vZE^8$U{?7Z@sKg<{6
zBMn&>_#>Y*aX`u;cBHXm{PcN2S|2~^>E-033++I?U)bsGlTSVLp*`B6KW@Yi^BX(r
zyZPhyAL?UI9MG+teEiUF+M_=Ckb3l+G;x!!r^$zm7xR~P(E+_#&pIim9z+MyBmQ@P
zo%^ml@6uH*pLY0tRf-_}pgw7c9{7dCp_6imANot0_Rs}C<m>IC1II;lgydrfDaSu?
zLE0lO;)LY;rCv`@>+R8h{Oje+KkRf;Pp4i_PveL7&<mZ&hk8F~k8#2d(jIo`t@|eq
zozx?rbrqD<hi>&bZ_&%C$9@5^qg*HRj^i}sIKg~HC&)PH{>Vp1bVeuY>-C5;Eb$k%
zKJmHvanqgtxV4K8x?kEuAIiN*JzZz=A%5_$6Fc%r>tx;N`J}1OdeO=J;`)L2>DcGV
zXP!a&M?2_<f6~lX?5U4m{eBE-h}{U6-+3vgJ^D*rq`gSokhr0~-q4Ny==~#~e&827
zw|$Vfb-K;pOy#3D{c_V6JH4EINI7Zxk6x6!^_TTTIce%a_5=1`+R^LLFXG|6inJT`
zdb*unPMUtYk@=>V6BmB`lKF>Dtb24pFUrvglF$A_9K^?Z;qPBTboL_tbvxRppXf_I
z{US{{{UIO!<U`8c=yu<U`qzH>L8mHt2Nm&L&bgmGbIpryA6ztH#Kp}&IBWkR+QmO<
z<|S!2Vn==G_8tU&s7G2Sey|(yzx|ec>`8mkO{bvwv>TLqe2fqIr1kkhKBSy^PJQM(
z^ALN=$!9*|2a=Cpy*_D(KR4>}V8?nSjX&Zb4T+!lNK+2U*GWD6L)s?|$wvqB=?`hj
zX`ky+_C=kX7eJ1~UiYmyZc~r*N_5mo{KQFo5PQ<}gL;tj4APK#Iw_|f?dWOZ!X9Er
znsH?R*Vh~6ZloQ(oc%=iN1FEZeV%gsk%r_$J>RQc<|(@3hj<|6_;ZU>&(HQWdc$n*
zuk>%{zYEDei7w28!Ik&AspVPKih`1Th<!}o&q;G$Ng5K5PW?E-KE*!F`i7K;CH>Ir
z<KHi-j~@8d)8u0hX-7}%`IO@ak`LJ**$1)nBJKGl=ZE-ZoQT7(AKi|55|qquFEanV
zh>lRV=ePsi+NB<T-0aB5o-`yMKTVr9J9zV;BmTTz;k?N<(jQY~J~B_;h#l(|a$d=N
z^s29)uQ9*zPksFBY4T}LC*`a^=0{L6pY(d<>*K+GqT8EFd#x+IUE(I6_{oRl>!h6e
zq@f-U`MMu{ycu`&)9q=`jkKdvx2IkF>*ds^AK2;C{SZI%l;bpZ5If58OTQ`Cj}!X(
zVt&&e^+-e7B~3k@`ufrBiC3?O9i*Hze)W9fC7(FShve&eP!4IwjqEp+>*rOZiHG<}
zL-KXvS0~rq_@|tF{Ll~5=uQ8jZiip&Dc95Fdy(@|{6g+C=sI$rm44|&H}Xl-PpIe9
z4*9fCnsVs2KCp+h7nJBpJ#-`u-P$2O${Bxs|0(Hd_Cu)ahJKW@F1Y`}{!MvUJ=y=m
z*7sYFc>UUAM|{hEy64IIzwh_w_vt7X?tFbOdSs^_?7MDcAMi`oBj;mLk$sBe25Ft_
zV|qUElTVs>NxKm{{1b<8#qYHy{^I&~qD7y-)`$57^?AqnI_nZ0^)&gA^Eu9sxKBfx
z^Ha_zIWHs)IZuGpKjExu2hYE{a*<Bf6>(6HH0QyjIi9eds1NC<8#$lE-ix%0KXf8Z
zzlkg9)aMIn=95mZa@}4}>v~Ym{!ZGBdVM|3c`D<ik0<dnUZk-jpZ0bCZfWfC!@PB~
zC!h02($ps%l*B>5pza4dh#hI2dVM`jeLdfAn)bZx+0R(-)YH?tuK2~?jk+CaFS@ly
zJ?1w$F+Yi$ai^SqQa@XXZs<eWZ-2e)@$2S~dEizab-q_TwCC3jd*%<vGv*Qb%rlNx
z9H-c)baI@<4q``|{g(3q%BfG9eU*8yr^zRN%DqT`XovdbbG|`2`Cfi0*X@bZjns3a
zTRrmidbCfuPV5;E()1TS7$3$3U8sjWq+IuhJ@X2(UlAYeQeP*2XotAm&Lf%UVM(0C
zfv&Vm9P}SM<~Q+D4za_(TYT&@P+y<aM@PzaVn;sfntb$!^ouwl@!LZ5*W;((=mD{V
z{QfBPrvUQ;-68qt%6es;Fz<9?Pd@XOe8v$nj*xunxuwZx{~<2cxf|&p@26m=_m_C_
zPk+dV<U{=Eq#RPtK0^Q9&Ii~}u!rR17hUiN$%o`a^u`~goO+Oah&?2qezFgf5AlZ`
z;{z#oBYxb7o_e{iA9m!sk^Xs|kGj<-pLj^S5nY&Xkor*92RpsKZimjCH{g%@I+<rs
zw`U$fxB9Frh#z!7XFVU?$%oVnO1(bqa(u?0Zb!Zs@#kenIqg8|Gd~$W^0Cv)$;U6G
z9`@+3`z5XKkIWZ{9qmE#buw=t{vq>_ddwf}$=B0*KNvskDTm~{5&w{WF|Wzj>yr<u
zN1FQNGafoA$3HqypR`Weq5q`ugJ0TX9+1zt>gSV`Q;#(Mq27*Lnz*TlKkVrzX`Q+~
z{pC1=J?%i+gZQC6_Djl{FXS^%Ip4qz;zwU^q#^A>bY*|j)6^rM_UI>mu*){;^Aw%v
zzh18AGmgvy#uGBGq?t$POnuVm4bhvpNb6+$pk7Xza!9^T)~g;r{iGbC3&f6jOFsP~
zO}|M)^3jKU{83Im>l9KB(O*x~KjOl_KJKKc4{6^mO}-n^iE=$n|0##M9r@7TRhqXI
zdZLd`z5lciDMwe*5Puw>87KNjnt28pH}*?Ny!gkD-ahf^#GZWoK+0*4_{gUo?U9c?
z@l%g@v8R4m5~o*v>~!k&h!?%-4|b&K57+Z<$9e1^^_d6g!2YD?lcpTf5Bw4z`J|~2
zX-~Jqo_;e<5PQh@QV-oIXC0I8McN@w`iXyt9qo+x-}}+*SNgtAKEyBeA^m1u>fa&7
zj`&Cu563y?0sSS-xT6dGp<Tui>iOu6U+hTZAJPuTQTmNO#1AR=OZ@2e=mW8X#7+P8
z_{rBvzwk>r{wOD(@gNP+g>j?*l#@n({E|<5Uex2HKK|TD9FX-&ntoBAd`LdT-YZ`0
zA@M@yAN|4~?Lu8w>NCzd@q-=yNi$xoOZ?Lw<Ah)0#4r6JjbGB#V}6<@<4GL!gE;jx
z<3~Ah>uLO<pH9}PZl|ZQgS10jkb0Cu^08z5=m&98PCdqrv`*@I5kIs?J9=E?>+R@i
z>M`%=55yjS(ECg?vlgry^n%PU)&<0#v>Q1e!Vmh8rXI&dH@dZlAL1SHzvI9DJ_mlu
z#~<W(3+(Vqd_2FWlls_^rr$a#XFWmo8Rj$dOg~;@4~dihlZN;WO1Ig_e#N{dUfn<O
zGv2yBdVTz`u1Mntd!6{lj(YT;cCpj*84r#p>=VSH`^7(G9Q=~_i3k7W>ttUc9!Nbm
z;)nXAA?@o#A4oZAh+n;ZxA@Qp-E`{Z_;rhy`oyEpKhpTs*AeB+XZ+D$($se&b`U$#
zkpAf9v`@PbdoQ97q&~X1jfa<CuX5Ufv=3QVI$4i8i5GvAlMg8;4QUr*r&I42_V~e`
zeD+mzBp))q`nXb_E{O-Ci+-FUe)ca&e%wSCzhqo=o#{8!_0;Y01G&y;y+g(e>iK#a
zyPzZv$iB%qGA<l{y!674Zl~`f=)^o=9@8Eq4*cPta)|!elaGJWtar|P^nCo#9;98;
zkbRr+AWi@D@xTsJPTYE$c68z&GS9H1AC%)4>g_WRsju77FY4*k?I~y6iI4BE5f|e_
zy!^dQ$|3djaiCq|=KPuQMOWgXAJij0%5~DNPCX9%a{f)Z?vH$(#G~7j_A18?>i&rz
zy&?KSy&k%<pVL3xo_52MabW*}wBvTZA2y$U`K7K)Txqxd1#RDNKTX*YFYA+inBxlL
zO#94N@~KbSFGpzC+j>3x(k^7*V4ZNBa63*>&OQcNxAcehAo<us;vkKGeO>7NA&r0P
z>%<@V_$5t!;$vOYAJUNH2Wern?}4*FqCew>?#vf-$BuDiyxGUGgOuZkd8DWD2WbaV
zk2G=P2NE}S_|r*!{8Aq~<|pOQt(@}<<~jDHby5!9_G5IkjqGpe#klDG@ai}7ir*m_
zC&)Ov`KLY3+nA4#_}G_qQlGzP??vkKz3sk3E0lYy!U4X&@A1g+8J&*lUaRQcQ*K*x
zO}k@!*CWoT$^3-q#(YLU%E{+Aj6LPpW5+rpjU8!-9p#Ka>lnYx1Lj4#Ojl2@_TA!5
zx811fqbsC8j4R{AxNzP@KJ_5;LMQf6_fNhXIc^b$+x(zhC-G4ay2Vl0rm$sx&D2q;
znyK7Qh0O|6rFQ(AUm?F@e&tk7L1BKC9qHPs+*EFpVO2X8=J)E{xY}s9-7a{3ha-oj
z20S`fnPTe|cO8*BcBT@F4l)m-TL-~cy3`N_L;+F2tOA8sRk^Rqn1#xXW?L8UqJStM
z3Wx$FRe`~6{<-SMA(NEtbUS_~Ex*#K&F^2`bBRhSO|645*YDS~-txWjqUL!u#<cxp
z?P4d}=`}8OD<AD_K6$h<&3BxIBW}2DoQjKH{g!f3Kok%KN}2+l3tCnEZe~tiwy>m)
zpY&H05Cud5Q9u+B1w?_k6sZ33iFfWfYNZOwxZ;y`L;+Di6c7bO0a3sn1#JKQu%yS&
zOzYdV)$5NOzDx;42bl-54wAADus;b!fg~z$W514-5BO@NBNQEyJP(q%cEZFocw58a
zb;mDJQE}e2KW)tUXsxo9U9YZj&g177E8BR;C6CSh{jgQab{@K>sPn(pFIKj4UezZ1
zj(u~EvW@ju?)g^F9)*rDm=0m~T&W`phytQO5*3&}YsQ?FpBJm7(ChbgJ<cR*DlwHD
z1q#1v(`E8~qnrlk3~S%FLl2?oAoIYYb&&L(%GCFg#P81EqhcP!_1-J(2}J=>Kok%K
zL;+F29t8$|(Ra;-Z&b=N#d7aIvGmw(!<BGn9S+=F_p36~zjTs9duFUeDhh}KqJStM
z3S_?moHq@)<*WXqyAF5Y*4N6se`5YPXKnLohg@~xo63#){W1N%(8kUOHGK5)*-Gbe
zbt(<)f9?tul)4|UwC(R*=)dFbsqVOLZr}Mz7_SaNcLJ~9XQe#cd*Y&7GIa~OKC<0D
z=y6T#!<`S<qT7UDT9)o{+-S!yO@C)o{Qv#;OtXD|lKzPTqJVh?-YL_f``!y@sh~Xk
z?bAQ4f95Z0+g)$mQFq<z4qWqm!L5z{IlzGz-*e*9k8k_l@%vtd`5tF?3o8ydVecz<
zU#o0oc8`wqP81LYL;?F0_%!#dm%3fJS_wr5nFr?A!Kin;9$bH~=_)GPx8q2}qJStM
z3Wx%J6j;^ny;}b{?+3>=`sth4hyrmd@OqPP>P<fBO{b(WZk?sQWGQgr#pf*>bonwR
z6dhz9$U2aHAlN>TtleCq5d}m6Q6M@6g8tp>=vo#}qJStM3Wx%tfV%<@6ntKO#<y=N
zzdW=1tSQ%SS*C=+bg<>UhHQ<q`PyM!uITlX3QG5t6Rf1D%`rtodd^gCwEex-Y>hwI
zD2ufyAPR^AqCiPeAgbpLljeu-hzNK74x8waU3%F5yo%rV&%@gJ-N%!*e@Q$$KM(JB
z9K}u)5Cud5Q9u+B1w;W+Kok%KL;+Di6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KM1hi~
zz&#spnts&h#f~sN9hf(-9@c*Ufyc~r!m{;>yN*a5J5$+8{qLaYe|Jew%kvF8Ki^=x
z9;}0s#;BjCdvL(C>d(|)uiU8roeb0S=B9Q}Z8&n~Dg_T+Q`Gri>ldqZ*|%$}*B?22
znKI3iHXpOKzrCtdnziBF0V-WqZ8hw(g)=6qlFq_5g)O`F@0t7G`I28VRWns9mD{PX
zSz)Twj(_tj<X6nEoXROE%&)Q|T|1SV%55^NYRAI-UY#3P8y(#=ySn!Uk9OU#*fGT+
zcQpR_$Po)1q39s<Ai8xB-3}n0L;+Di6o^Ivuj70)t%w&<Kok%KL;+D?X9dn5Ua9AM
zFAa5U=h%Y!#r;1hc9PEJCk&bValtYtsoZ&XM3>t4z^=dTJ-1jz%{$InIDTS}B~G@}
zYb475^l^&{m-cf^vv!$dZ@c)GSx#K^>bI1O0-}H@APR^AqJStM3Wx%tfG8jehytR(
zt_qy@#-I`N&RXRJ<*uG2M-&hRL;+Di6c7bUx&js2ukQcMC+{nvojNdow@tkKxDB<J
zJ3`Sx=7Fq(r0fIiPeM^3i3)W2q|IIHo|x(gMTaEMgCwq<FfkS99d&%IS}UEXIQ^x)
zULJS%M#omZ+2yHG-yFNfv5h0E{?us7rZtZ3+~?s{1$m>EJGS!fdK*s4*>AOD8&7$>
z^^d6;^BiF?9m4FnQb!aJ1w?@)Dv<i%-#veNWui(7cRTakQ#LdltNb#Fno3M1M}Z5r
z^{;i#Ws4L0UQ$xumy)=W{XSbs8K0!@RQXQh`v3TFmJ*uPfq7v1JwW`^o=_AJ1w;W+
zKok%K>``EB&XqMz_@;c3DSCb1s-%`D?Kw$%W~@Xi3Wx%tfG8jeWWNI3XSuZ2rF(xk
zVuOO;JfGkCu{zuTOq2igs`IZt^Ody<>fd!6dGd)Bx9m01fqVBlx6PP4raL-=z8{Kx
z(DS*ZJYO6(j&4tF8#6amCJ$bA`uJ%_Ox)}U<JBSPPO$Et8C|-R+wKTOhmtT4g07Ej
zx0n4n-u+qEyUQLUkN)!Q4Ng)R^zXSW9<Ls{=)Sd%Ff%%2`~D>T69q&8^9uayl@S9U
z8Z^%d%DvWXzOQu6XI0nkL++~5arB?uyUuHyhF6%T-~|P_Z#{DRROR=*K=VD$?iRQY
zVJowHbfkBpfG8je*r&jr3syE=*8c}bC_2bIFuxA^UU}80SKPVPiHi2^I1;faAPR^A
zqJSR-cuvAL`sth4hyrmd@Wub$*DGm^TW4u6SqemTy&-<0y?zrtMb9Yj2grOfGoO;R
zn@cpJfG8jeM5jQ|zb_wM%i>8C5Cud5Q9u-MSD<U<ySBZ)ak}zL-UAfw{9dG&9{T%|
zlK$)WNKNVDcYf=3?kgu)N$oPn-gfaVvm7^$Y}DwGAurE$gvrt&*eHv&C?E=m0-``k
zQ6T939b4>8orj6vc+Y!9@4wgEw&%a%j@R!#9(P>Q-p4H}T-wj^tApQh6gyEs6c7bO
z0Z~8{5Cud5Q9u+B1w;W+Kok%KL;+Di6c7bO0Z~8{5Cud5Q9u+B1xlI%yPbLNDH|G&
zRl@XiVBVxY_;=5rUYV%E(w66{+~VZ#0l_c6AMnX_7v<!+@t-TsYWC#b>m6Zs=)gKC
zX>8u+f?2(Po2#N?_ra&UdC+gG9XIOxi{IyG!rJNUeay?JU3c-7V;r6O-(B-cyS+ye
zb$&^kkJ;MaeM+5K>cKfPRJ!as`M26D3%+(rIt$wrw#=`Ys+Fpp%I#FxtT0t-$G`a%
z@+;<7PURF7=2zK~{{Mxk+$O`Sb}Y>A)wyxC(b+n?q+g<dC?E=m0-}H@APR^AqJStM
z3Wx%tz~3v->fj3t`c)sIY~|*U+kW$}>Wfs;S^n5}myT+(MA^c>pEd5K%Z|U``4Ma8
zD>u5=loFd1_~rVk=k;$g&WVdQX(O?U0-}H@APR^A*`mOApBI&F+<K;pi`k-^^g|R7
z1w;W+Kok%K>{B4>^BwyXkcdTr=oCmj@I$kQkD08ZX4nWHIP~;Y%WhcVB%N;e({0VK
z^XrI3bq`prY+=o#%bovA-xbO>{(AC?6NmMftxWN-I<<bAdF5=!G>gk#|M2w2-z!@g
z*E&u2_r<XLsMHe$L;+Di6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KM1kZhP<PuwC(T<k
zTG__rD=QI-0-`|F&r^uMc%RG7l-^Z8m~_xZZI>xqm?@oPhY<ymq`<6!hqgRt#9ZaY
zB*iSThytR3C?E=m0-``T1?u;^Y|k4C<~l;VbO<-vQcG9h)_afqxP8HNB`ldb=<`9+
zqCgZ1d{X(A`<Hcp-HD2G2JJnjOZhJyTe(~F!rU*nPjgH$iZM%)kLJ(Z_RQ)5n^d;a
z*7rjs?r7fc3HqMX!I!=|<lRNZN|+fPOuZL%;=`@h-9K)eGDX{-kKp*2q+?N%VmW*C
zQRmcey<UZ7(0wQA_DTPqs^9pI8JAb>qK3nrsAziKFSzW}TE{%xN5##!kAG2pZ#b$w
zB&v8a<tOg>qPG{-IEbI@QXt*)ub^s#-JiYM@p^tCX!)XR8ovE$-${-ez2cGbzgHmL
z?{x58>x%Dh_~_bWW;*FI=yi@;@6ugQ*;e08C&?EDL;+F2v;sl*^`zOmoqyPpZ~D7a
zru;|s`<<H~zxPL?wu}1teADY!{EGrMDR9eAn<ihe=x0Y*GIX$MZc418fG8jehytQO
zQWOY!o@6@iu)j|vrC!kAIg)0d?)Q;1UElWi2}X8lQtP1^D;-<;!Ku%+{o&`*dA9QR
z^IyFE)aq*;(=@%0ihof+6c7bO0Z~8{5Cud5Q9u;<KLwWm@a?PLG+FQXWzhR1aoO8`
zy-WO2|Gm4o{n`H9pnf0f|5nP*|BL!P%$>bUN)!+UL;+DC+Z8zVs3-q1Z@_#fE@r!)
z(!X#D#5Mm+wUh39H%X}<^?7An{%rj&5OGKK-gmb73AdX_Em1%e5CwuOpx=Lyw7>7-
zCrzi|BQG^X0doq(^*e;$c5J;*&hHeaX>KNn=S(UP*Y#-B?WC)j{v9OK>2%|j`l5g+
zAPR^A(JJ8f9fq*^(MBiUvP}Wg=jpotY|}3({oy-vNvAD;$0_W7oAi1~pC1z6)bsi`
z8Rv)&$<QGw`@6)ET?+WU51e$nsPCic{-S!n+xGobKaVb{X`aW4o0)!YXZK-;tsmNV
zv!ip<*~KIB+HQ1`#;DGNoAmkQmRn{lbfTud&v>Py>Q7idanE<Z?Pb?dIeEQ_YoAYE
z&+U$=uuWmhZvA`a{<jzTHB+@ybyB&V3Y!(CO6~YJze0Y+{K~1Eg2MbNJJPjNxvAVH
z!>V>H%<t8?akbIAcK*GyHXk*i%5ukxyVi_0r|Y<aF<V}oq-^7g)~~%ad(uQ@D;KXE
zpHo<GfwGlF<;tls@6A%SvtGkHHyzM)vNFwRS8e*<o_J;728~*6aBSh%m%euNoO0`&
zbm^ts`E@po?Y?E2N|&~->rBP%H7;)D?WWz3H+1o2$5tNfTyWd;zEhl}65Vg<+p)!<
zuh%$!x%`Mal_$Ql)-lCn7H@1-zWREH_%g*CZ|S{v*M)PPq|=l>M|3^tyk8f6sFFfc
z{lssj(x0x~x9NH(F4At&dH8!>23?rDS6<xw$JVZ+4_Z?$&sHAMq3qVxXO+scm8DNu
z*ZAlL+nu0n`pg>}4_&>`v6YWiIQh{veZN*o=S>H9=y%Awzbji9cAxunO0^q`r><6R
z4C|MABW^BOu<fEx9oyOG{xg2;ytS_rl*RdtJ~;8!&MMRFI^m-+7tb4@;v(Z8mS=pw
z`O&rqtahTJ-#BWtexy&8wQH0q`i(>E;!&X8pi7?JBel)3g$K>Q>vgB~N=Il`hYy-o
zy7s`!ik+xwYF4km=IrCoXjCCDZkp<km*3mBo!&GxVu~_FulAE(Ui<Dz#W{PtqvB%H
z^|)u&<_!-m7^i|V-F`j%TbH~lr+%lxGTrvw>Nosk@wa=lTdac8&0o;`YBlbj@Kp5!
z^0JK;*EBis+QA3p*~Ya$&0DduUFAGmc;oer>sNjGfV^yB(N_(db-A-ro-MR(e(Uj`
z`;V!+eRx&nyi76Ze(BeKa-T<jC~zv}LBI8MyM<k5ZXWo>bd@dKu(;~&6G|^qK{@sJ
zA?xN;pXmhUdsl6{`-?-~bdpAWoRgL=?Ag6k=|i?Tw(<3+U)^iS0b87)Oq%{lkEg-p
zO>KrAyw*uNhc#?IZoitvPPVXI`PNU5d1$R;3uo4u`)=W{s~y{D>bVd6#`QbI=jHEv
z;DpO(JGSzM9*wuvon7oCoeTFI_3$3guW(Fp&sh~W5B~Qu$9C5Keb63{pS{|#g{H<+
z{03VuZSQaW&)N4acWmK3*DZK(%s2C$q|ns)<mm1HI(omy=cv=V^=s9r$6R4dIy8Hy
z#Y6Xe-c$+W(V>3R#@!1FepbSGb%^Uc*>5|h_E9&#uOF|R!+xKmqM}<n<X^Dhhu#Og
zyUhv8yFQw9{i%a;@{-QC@3>;nqLbG-wsFciwQ_GBxY5ZL4!pT%wYqz4aBShY8(Ll2
zuyDI$E9VYAsh=8^lV>~Ie0lJ9!=Epcmua$JW{N@YZzs+E;Hsy7QDe95PF(amPU*g%
znd$n6e_g3U-HR7FUi7PnZWr`-yQJ9b{fIifvfdr{w|H`+V;iH^A!>i!4{F)>;_tUA
zFY4`jeU~C><v&+#{C3AVOB~bGbuyjar*G*+Ph3<k&yA+q5BfY^((R+V{>!wV5r=er
zeD?K=lr6mPp&?aHU$aEn%5Q$Ce#bi#Rw`S#rR%Dp)0?bTansa3)br(u)yh8pt&=X}
zesAx%o-Mbp8#Y_n$_~!SxAr(;rV2_^^B_}xpSUOIiS{FxDO0@ewJ%GL-nK&7#-tzb
zgT{B~4JTe!_q4gnHpX>6lBxBV)bHXXC9ZVG<@NsWse3=0=m?Xk!<^=gzM0tMSI2f%
zJab45r`bqHr#^qv{r=eW4bOFJcEvnpJ8gX)Wc$w(E0w7+M+K!m4<x;_0{3ki^YX$&
zepFr@{MI!ur9K^|g7WeecOS9-l(EV$gFeTx{lX=MAEk<vU;2HXAgOlUKCSxT#Or@>
z;^xk?rsSpGe({$4-V??;w(_(}FD`!eo=Hwz%+w62IBZOxMQs;2wzAUI&kg;s)9;Qc
zW@=nA-R`>gPd@0QV>c?>_{H2ChduP!dSyFvUTtvX&2`o*TUh?7hQ}Uu<XV+!Ui;*{
z8(aLi*zwD5Cm*}vv8U%cwlXRAKf=ag`gdYe{_A~qRomNZEpcq&)&)O5_fWx-*uK+z
zV81RG9Qf)iCu*8HF1|OsLf`gz(^OdYe`nN%AGMgEvV~{8e_Xv(v!%**_S$;bMUTx}
zucG4p6`St;Rm)Y%6f57~sL{m3R;z3$?^~ryzx%<4+m2n@{+RET8ylRs{ftlF`a${S
zomKW<@#WUh$}fX{CoHJ_z;EAJc1X98DpO4Qd(Cd~jcEM(Nj-AcIJWZP4o}^C)WBjV
zD*9bde(loTcm1}5-EBu+RJu(58Woh(Kkm}P?#I>ct#n3x|8VNH{qFzA#-WZIiKo-5
zl{IU8|GBc2^uspZ)^O6*Ge6JG%M{)8a@)r;o!{xyMO~avXFGAzO(#9S+h>g*tUK@*
z6*cvCf~J%H_i7}*KmM*z&~f#$zhc>WNAzAa+_9Cn-P!xeue!dXqN0~xUgcrGhwHbV
zTl-1*drM*CIO(9eea>$EtFo21Ja+#FTMk{UY~z1!J9~N28!J?{F~91bz1wyvR#7qP
z=iyk#Vej(=tryqxPxR}}miu3NUGr5=y3Ew|F#QmU0-}H@fC6*(Iq=Gx_F1N~mGm<z
zUNg4rPh)p4a!fPbc+<T<7`Fa_kGJdIV$(X6ZG2_HOI_|bwpc~Q>mR!R<k1hj<it(C
zzr&caV_sj^YWbo?H_ukKFz)LhbkdoW@4Nn5zt`!PY+32}rQVOI(-kTfb-DcR+&o*k
z?whagefrUIdA9QRGf(P$!<gJWTgl&Bx0SYh_tq^=+u#2R`W|^$`>6NLZ1MN~=iS?t
zS~f#vigmtx`|=TQFH~+UH}~2}2TWb8gqhW0zq9AoDC{#{1!cCZlce?|?DLRe`!nv5
z{YST`J5I&LuyH1}Ue&2r+`Q%Z6^>AJ_^WyF@xQNBZ`NPq*vi`<IeP3Hy_Pz{%;*sG
z@A$=KKQOP$9rxV6O!;M0_Y0%;GqLixr8OEYQMU8^Ivd7z-!e_v%DDdSx2bmY@6Lri
zUwJI&;wo*fDOP?ts_5p)i{4+Y;%2t|omEr)2s-}L9-X=O6}J~SwzACQFaG+>XM>er
zy1gfp>3qN6MM;|7^Nre$pT7HS6_!Egt7-d8-3K+@uc-XT{XLDS;*QGCO$T@AcgVZH
ztEd@Od|rN{e(o?+{%n1|oVe5d-f&XtC-wVSUU7MS=Pc>v;~rl8bNQ(woNQs(?`-Jx
zy!JoS<>}6wu)j-YTRqd`ROY9r8@=<^JjYgc{pIR0{Tl!5nBvqg-#@c%hqcNyO^<U}
z|AWuJ>c^|<u5!G1#KeUkcJH`c*+$#H!*NaB@~;g2&ngua!_Jei_0m0m3)^ncdZz9-
z1dStU_79!;R`qSa&2&uh#al-&JnYG0N0^=taUI`H&(EJ<sdv=$!J{2pIlT1X+4mkZ
z*NK{Lx_|J*?+<KQy4A@vuln$eHTN!9<-|=lT}<b{_SWJuPw!toFDOmNCI0_ffr%g8
zGW*TGE0wL>`-g)cJA30?Wr}~Tn&j#V^tu0xA3JaD>tq|_{_aoG+ACMT{b^5C9qU9z
zecq7%plPLR54^0{5k{*6`r5*F({9Kcx_Gh^7j4lcY@DZ8Y_+&i^{<_%=r@k2zei@u
zU#5QV58AG+&*yl>{o0F9R{Qsi4;-QBAoIX`9r)dUbvylcQrpsce{4x#w`j_(&gm0e
z|31Nr)~~%ad(uQ@nqGSO{kv&Nw=?}c4=)}4mdEw|G*j)Ey5C}(-%P#loay&1{Lb^F
z*y(?FAnf&oss8wVem!Y+NxR=?YJ9}+9}09T9#M02o8Odap5ErUardt*n->+|Y<I&2
z$6b+=XDjb6>~!lLU#xP%van5I%lw+DI;o>mxt$7|6{bq<_&2{oe#QLCshons{3<)r
zwNtsN+$O`Sb}Y>A)wyxC(b3KN{pa1?;kZ*4C{z5oa@Ws4{d}GhiViXlqFV>i?EvCQ
z6c7bOfoK%C*Li$!PSeTCjnT9sUPJ*=Kok%KM1h?Z81|1t2i<t|Gmh<?cTL-_D<2)D
zlFshs3!l2`!bvKr+<A6Hm-^<;)DZ(N{9Q%Oanr{<wdDBOD%<Hb5>?K;v2yPN*E*(I
zsrG;a&o4LAiHlzSmU2-*6c7bO0Z~8{5Cud5Q9u+B1w;W+Kor<jfmgp;f6nq@%T!SA
z>Pd1$0Z~8{5Cud5QJ|zNaP_+z-n?@ABu8kc4$R+Y_g?z-T_4X>LeW9yfvkh1>;vpi
zLQx=z3Jhrdag(wYS2{w`A<6R~iEAfJOe>DKx#~|}@0%ADD^w}I<=-`CJGS!PeZKwq
z_kYh-wy~b`TdA9-%vQFu^zJ{de&~bw%2xjVXv6Q0E105e<KSWQ-X3=AWF-uyLzq2R
z>WBiOfGChe1!}(g@53&-WU5LEgMPmuiJD4GB}ai{A35W6_1seRZNX7Tlso!Hq39s<
zz@l}K^quP5DyKBd%U`90W_4g5#C0A-dqPn_6c7bO0Z~8{ut$N%+UFl!ddLE0ibsDn
z{(?H2O63W6*5Q=?Pd0t3(R`H@+B0J%Qc*w@5Cud5Q6T#jxaH?N_g>#}vvWej|Hs~a
z2kbfC{o}tg2tll>5yY-h<*_NwtF}fI9V(O<eMHn&yOhSJEsa^NnNZZIJ<l5{V#JO}
ztPBxJMD~vSzP`M!{+^HT@toIn-*+N7|M7OL&+*yE=RD5yy7OM+qdUKr!zsNd|9<IT
zU7s)az!GyFIHqqE<KMx$?hjiow`Mjfhwb(r``!~fj?H1}J=KW4RvLEg6QlDKF1GT9
zcTdYl<q^A`dd12^Hv1s270=n})-hXdIz0cTVAOxvo;QD=t=>$q_|mT}aC7THS&5|{
zz72j*Z+v{y=l}P1Z+)`t-9z#czZyL}d&z~TU9s@cY<|Idf9ItBD)I9=bYAP>o1ULa
z`AQ9x8kqYUc;?NIr(d<_;H+M}d)U64cRai?-}yV|{p8p;UaR71qYgZ|_nm*s;EsEZ
zx!{Eh-^n{1c+}2IT{J8!!nrT;H+zSU*Z<;{?@WC)`?}&c+d8Fmr3Ok3lp6T@Y2cIx
z*ZT1bmC<>Lr5;K@nEP=M-zzn@{q@T@CC8-(N)416C^b-v2I{@PpI_$JB3n79)WE#f
zz;4GMIp*d)-^#yjIIs0w?l*sFpz+-H%k2Ern7qVN52YWJaZu)g`trd1mCZ|TN)416
zC^gV>8rY!ARm*mJ^4`2wY`L9FN~H!$4U`%vHBf3Gt$`Ceopt()+YZfZ#o?>{_R-3u
zd-4+N)5FwZd!09Umx)=4UyUCAx#VLHU%K(Iyk1O4PJJUqzi+LNS~<JaK&gRJ1EmH^
z4U`)ADmC!dQk$Liw_m)Sf77u1I~4Qg?@*L_{%ZC7b&uOsj+?y3-+u6ZKDV&e+*<xS
zhV%EoV_2&l%Xy^+N)416C^b-OpwvLAfl>pd21*T-8YneTYM|6Wsew`hr3Ok3lo}{C
zP->vmK&gRJ1EmH^4U`)AwrOC_JG<;L<!6KQ5}T(7e)C4<iUpSG^hwrK-0nArop8kD
zSF*1wwt8*!O*dUJDucWK`Mn{F>@zt_;iwDdym;u#Ph};3GxQ)1zHJ!)J+kJu_t^j5
z%i6yAOqRm<@76Z=??e`@U-i=qwma?p=^5Os)u3m8+G)W`#7$0|d2*X4+f{1Clzc7S
z|84WfZ>qd|_UyaOoqc9!&Ba+0Ti<-uOONK?cI<V0uY)eT>zY>o*I&A?)p^~{>ve8*
zZm+$1b#DFHf4eW<eW~utbZ$7iSNCN<Yp>n8Rp(Y+`YwM?ukP2LzuhYRzo~9k%2jHh
z)Ih0$QUj$1N)416C^b-OpwvLAfl>pd2L5*qOkC{JgFm`)d|qOyhyTAH{O?{<j($}d
znDyFr7mQeLRQ_$luS(zLzF&<7)_MAZ3%5M)y}ZP)Mi1>aeD(07&-@^-6@NTviw)m=
z;+4E!{O3i_tkv!ODOs)f(Tq119emkSS-m*t);_Hlo;WgZE}nD#3$J~6@2IR^Jacf{
zVK+2<oV66*df&l!Y;)(-?5l~+a@SG=r3Ok3lo}{CP->vmK&gRJ1EmIP(Lhu0ZL-t<
zv3r+&2IMV`t<GwH>UuAJm^T$`k-3~xYG8iVK-+zeoY>{&5qXKF9=^$bu*qsG9P!k+
zL3vAKt@nfFy!!4_zbbvruX(!UtJFZLfl>pd21*T-8YneTYM|6Wsew`hr3Ok3lo}{C
zP->vmK&gRJ1EmH^4U`%vHBf4x)Ih0$QUj$1N)6Paf#at?b?uBL#^)uLdYGU6pcY>)
z=ad>KHBf4xIT~p0@2)pEey`I`-};@rUc6_kKP}sPn@L%{c*BNU>^NYRnORHY*-uZr
z;g`p>uGEW7<)dD?Q}a7svF&odpEEnF7n|cV<=X#f;G`x0`NqFq8J(B-?bO46JfZx*
zQUj$1N)416C^b-OU~X&Ril=%!-1Wwf@)Aoulz#B#IGEe<_jRTJ_l19P-!YXzd5NVS
z{x?79efTRcZ2Zfyd9Aqjahu=VXYieQt=RLEbyuAJYM!NVzavNf_<>HZXSL$*hOV*P
z-)6m%Hx=*ef68e?K72eY!t-a`dConD_Rea>ImcxW^;oli)>QoYv(rys<4+HifB)|5
znsC2nd|$fEjs2GSWOV*b$CfrrbAQKPFF#G)&py*<y}M-JZAa&G8>g&!>Q&btGA2(U
zeo8DgP->vmK&gRJ1EmH^4U`%vHBf4x`5LJAyCn1L{k#A3YJSwn*z10>;Z_4a$-izG
z<tW?#Qv(~_bLpu5>wlD&`0dcc|FnMDDm74Qpwz&;)4)eVXB^n|sB!tX4d<PPTXv`B
z{`*$-u5VfHOKPPC=B@^MJ({oIYwGZ<#8MAU@q_!G?X<vdkNrE3Snt1k_)VWbccZwZ
zUTUD!K&gRJ1M^w~?f$jqdxyQ-s!}gr^1;6M9kbGitfg^Y%U|wSYM|6Wsew`hr3Ok3
z%zX_^+I`to&$!^@yrpsObN=}bTW#_9Pd9t?gS^D~OAqtP{^dTU21*T-8YneTYM|6W
zsew`hUyTM@`uXAOPVcL+Lb+?Hfl>oa(ZD}`{lVokZ+RpCx?)pgF4vSAC^b-OpwvLA
zfl>pd21*T-8YneTYM|6Wse$=T0}~g!^x%(f9G|xo&TsOS{FE9fHBf4x)Ih0$QUj$1
zN)416C^b-OpwvLAfl>pd2BHR*|IQy?nf=oEya*qk@zaHRv>BeyFSz;aHC`Clb3#@t
z9(T@Vn=RdIK%T;e6@GB`?E41gwc?u%|JC#8#@n-6@rCWzd3&3mT#?s{(OzY{)Ih0$
zhK63p_d2NiTAkPH{Qb_Y&h52Vug<MM`)~KfyD!y!na&Mo_v*gvXYI8+x9Z%gOW)<s
z>DB$(^S4{2f73Phn-!1Wx95;y*^&Ef@Y-bujL2TuVdk4pKKD*`^yQ;|eDgaav&RnZ
zx5u1+Ovrk?_u#N&4<1tOkMqzu>!Iy7Ir848vOewiU+a@K`sc^K*89l^_Z^+>eMaxI
zuI})9Zl9a3y7}%qH)yDszwpjiJm0s&{<-!A|J-2ddvAZSx^KTdI!r(9)z>oPlP<XJ
z);&(TC);VtUq*KsGCSM)-=lj?=`p^_-<n&TeZ+%DwyH!r=Ama_h^~FerT_2WJ!a>B
ze7|pwZi%1Usno+)tB0m%fd9Gjfa5Mb@cjkf&H1^u_`Umtp0}KN=8#<fzSBD%`&j$8
za^u#CCwU_FzxZ6wk=OPgn8^>?;!YciTlmOY9p3tQy>ZoeKYob=w|efak2CQqf9a2H
z{Se*Q)(^LBee<FVeKfI}Z_G#ke`>RA%`X~YpMURt{{}nWFspjbIS1q~e$0R6eSR3v
zul%~6&n~g{<ZM)@XHV$9>p~Ut__cG6U-F|{mR|Y#SJzvh(r3Uy>pbz#))nnXruNz4
zkU2|Lj+oi)*0UB|v~tFKA9eZ8pyexLuiEy%FWhml3V$;npZF&a6o1;*i=StD9dq+9
zR+*E1@3)l?`i*H*F%Mc#fA&4>%ux>?(7s*8x-t9xVeHAzw6EOp+)oy2*JshnI{%uv
z-{2k#R;;JDi1b4baEFHm9=G+uGc)7p(Dz;K7Ww^hYb;(FIdsuEzj%Iu3jNaC?HBj@
z^W^Wfuj+L~?>*1mu}hoEdY|0?`k0XmRoB_4k{#4@`9r5<zixN@^ILalUD@umm1eE`
z!L;(<_p330<?Tmq{{EbSQ$Mbr56+8+dK`1^GbfME-~aQTQ)iz%J~!`y_3vN&{+T1I
z?E@wp_ru9sjjOhYzPRX7i%*-FqvPB`=bn4ARsJ$(Sk7PLKITF3Ev}*SNgLgWHL9bW
zHkUNOANY;>Uta(Ijtjl=boYC!^@}|G>womVVV(P?=5WN$f7pMcNz-!qQ=5Ni-#2c?
z6*v8EM9vQy?-+R7l_#%I&2#+H^G@8KUq10pR#~mj)Xcbj#f^UTh375f^1HV6{1zX^
z<x6y}kL#?rE{<ETA00?;`-@L>)raQei(b&Bcl*&hzU(7D&6B5x-22bV4_;`+3beoT
zIMUaTA8mTqcCI+LjO&N`p?*kS`;m)0sdf6TgXkL9whz6UZy)348`qyg^w@!ZA^kQN
z?Sn6R;z!~hAKKQD6Y?X-Kb=1i9e%7W?)9T<Kk}oiAK&yz-~0$a>OuD`<U&u}&?`QS
z<Ii~9PrrE(9ej}2e&(?QzR)*LfB13E#(sK&=s@f7Yo50K?TdeTAkH@rU2XKlxAVh#
zdQeALM-TXAFLk5*?%wH?wI3Yu-T_Nh#toSM&NWXhUfE~0VKcw;$`Td!w=N!&_QMA~
zrqFzK6O4AX9$ly(I$zmwc>8BZta8GX{J;%1xM#tKCgcbIX4JaNZ9chL$C+=v{k7vb
zy2iE9Cr{MN*R3sImH7GnP^pJf51-u=*6IiHw)`OfJAa*1`sHuy<ay{^HZQKjSDYX7
zzWvR&A9~IuXkYo(xdQb=`46I_okIDZzT{W?%q{es<eZ54-oErszvjvNMaT}&IUL*W
zsi3;cJQ&xjqhY=C%!lN)4|I;<&wltcPko|}<nM95{oLoF6VZOqJ)pMd<k*jnIyHsS
z@A#j7b%8e25A~-|oER5>>M75IkX-mCpEkW|n-8PDtv9Y6(Y>O3NcWDn?(x=g!~fB)
zGQ9ici*NE$Lq+}Ro>Tj&6=puL@ftI#?>F^Z=bqHPrh7~Gv*gily?%1JzqP-%`xN!8
zadgO`-#+erq4khl>Sf4Y=$lVIa=GuN2Yjsm{bP6g^uHIW9J<2>+wXo++v>eKdJw<v
zz3CAh^4m|_JZK!B?2RtOC;iwTKhQkm+V)`&{q{9q8-MoGZyv<2{Uh2BJ@I87yXeOk
zJDSHH_C-&;@I!6>3H3w$@)nH!{7S#|-XHKod>fZ<<yCsH4qeZG@*_SVdh&#RZO?&m
zy>`q?_Om}a=<26`=Lo*ZfiAsUPfw9A{Ftu|t*0;ZAimABPvjrp^hO@%B|7|+|A`Ow
z<ah2-%}1ZV(u;oksIQz~+UgVam~$wi`Ox_k>qY%gT@>jXkMz_R)}gO$oqqElem1*n
z(5nkHs!O4|R@;2{E3w}=qz8OL=Q{nR>OSk8>(F}gsB_JS#_daw<TsArh~!hp^F#hB
zZnTX@v=2TZIq<8Wz3>Z-)4zD(*KwZt<i`j4`l0psvJXFq^U#OZu@kwfbieqHM89}4
zA05vJ_>KEoPtWp`d?L@Nm)Vuvk*@v}$^#MY$Nmw~!+%R7J<u0@j_kSZfbEZ+nU8v6
zdY7M%o1L=@{fakwalUEeoBfT82mR)YCw^kSHuSs#Js)~rL>E2z!9M8XgWTwwr)?g(
zP&^waZ>r5-=oO+z4*iF%zT&03-90O3AL}7I-~%1|L-Is)-)cYe$sOmX+WZ5*{8PMo
zFG@c9phs^hBo|bFm>=okKcyG@LVg4JIsM2h<Wi5+i}&x|uknEa4UKU2>mQwc^tgo@
z4;|5Mg&%G_qxv40JcoCCbK(6rn3Lm+JmL<DSAIf%`nOIzpr1nVA->F`Z~C_nf27aY
zw%$5?@dxv@$wlAx)fUhC`GtMRi9hq~V}JZ=v$r-nke*{(9q1f!o~kQje%E&H^CNXH
zJr!YUUv%-K4t2hl`1yC1r5;K>eD=NVH{B29EvVjcUO@R({!7g>Zk;*=eP}(p^0xC7
zqT}3^{~<cssd?y_ANh#u<Gi@uKIUmhIqXADa%np+=*c-0_d|z%Bbtwna}9lnj`{eo
zueNdXp?>qAIwJ0iZbavo`Q$+_wyn3nb>{2GC%ZuN=qL6^{lxvyhv>1p{<uFm*gg7(
zb@;I!ef|7Xy{8>fon#;NwRWU$J^qaAkL%2X=BtnJZ5$nK`XDz%2kIw3y878M_D8?K
zKfSOQ#Fu%IU*oagzW9yL?O(UHx>J2BK2xYJbuZ#R1Rdk*R;VAUL#@+xUmEFKXTEks
z_cHdypEgv7TW^1D^5esOulF42;71!q`sS&_-3Oa*Uu}Hj!~HaV%rg%4+mAl*<9&$z
zpywpdKc17&HLmS>5I^yG%RF?@bzb1pJoMr|?!EX0ey#I73gdeF>Mv?XIqZi&^2j&n
zLFb2Y^2RoP$YH%aXTEXglspRUXPx|{AL@tbYeRIb(>87$x#$c35y?q!+7anDB6{Z2
z2Q;7k_46<G(}wzEoLP@9`NR)?$OFG_dH>tauk#(M3)CIa?^E$0F4P0s;#1u68{_C1
z*S3z|@%uQgjb23jlLwk-AAC8NoHO#Rb0o%%w)lb0QR~zd=!-k+wXM?+oiFHVr^e9}
z$L6CC`M>9U`NMqsLGO8eSKvDbdBQsK;)C4Q`@X|>I_e7hIIki4_|vZr)^ELgd-XrR
zb}vY8);VAG;~Uz~cvBI*rmjz|<9E&_$iC2XKt%cx7f@V^8+ihXcXo*DtdHj;dx|?K
zp2cm%xQ}sh4DB1~YDYu|eSDC|e0s&dykwl5*2lJfbW<ok*<1WLfB7rq_vVXx{%t)!
zF%I?HS6h8<ALHUY_RE*@i8gfK4#lH!^t8!CzEu3^XAkv=xWzBI$dN+(;fMb0W88Y<
z+W4kle6R=qK}Q?^=;(*mqX)?+@A4CRWDoX?<Lsedp2i0{<cIVb_tkH``Sgho`|^YM
z{hob|qemX|(PIbvJ6Ei?Kf3g9KDy@XXHSSOe$4Y-8#=!K!l&nE^C5oFqc8fgFaMB-
zAUU)l|FaI?^u@1ISZltDc~|_&tH$L+ZFyQA#J9T7xV#(@eSA8PO8orq0ko7J_$R-F
z{EPqaGxl>XvMW2qJjTEIv3jMcNbXpt#D4dh@%}T_AL}OfzUD*wXrphvb<n*weyvl#
z`VPW4`r7DfqpKf%bYs6ZKCOqwtyeFaZ#*J?tiwmD-+t=%NKczy&_iFpdB*i?+s8co
z&?~vDi}cO69-0U32lZQTpSUmnj9X_P;}E@6+q#J2Rh)`X`CD9<`1#-0n^$^}FQXl_
zqutDRj<TckQ$OTy&QJa5IH!#}ucM#H^Z2pe{>JfbJ-*DdPejk1k*;~<z`yz+MOQyW
zHzN7jn|$~;Upvy%hLJyQ{EJ_5p-=BoU&f(*@I&ADkH{|MwB9}tpZJ2-X`2W2%Qx~@
ziJ$*nrLRg4o)gtso;TEAp4+t5cl=-72i0+Lo%Kzv-^z8R21*V5M+4$X9v8>r(EW>f
z?uFogD3$;BZP0+a0LtG`zjI!_pngz~IH$DLW1j!?t7D8i|DpQAzWkOyJ4cM`cmBlr
z_#}^dMw>k3i+ti2f9e+dqi=uo#Fu__@TcFt5$6}*BPY8=y=gnw<s0>a`_1MeeWnn9
z^Bb}|J!(VG`;b48Q(ktjMvwBidw6vkKT>z2%P-wixHp07Z)lx(aKCWE8G{bmc=LuY
z{XW%xk-qyVaZ5hWE95lJpJJS#Z=9X|ZYqW1-Fk6{5Br)YUunm6)<={ViV)xS=NH=S
zXC8Xy>4&L)<LIJOgzRtL<{$Stt#O&wjoL*R`C%`7usft@bn%T2v>ryf&L7A=;tbk{
zUA2wVKmMS3_%R=fBmVDvbI)L%II<31h%fn+f8ql=_uS(_>&QW0*5Swg^bP6NdTse5
zwcb4T6i3cYNKS~JHof5wTF<`bX`6>{?TGf#j}AVx$pfu}=F=<vLH16e_lCZY@xIS{
zK=1p!4}`v>@m)-<XrGAQ8~TpPzS`dZnOEZHe|Nmp!`E96>Zn*(=`TX(i};tH<Zo@~
zsk|glL-|@;-4)Tf67xCw>NM-L<tceh-ov-^$b9FGeeh*}>+KuS{>}m8@}}nkdD6bd
zoip@@Pi^}{{pQ&h#&P+WeC&cBd_eP!n-BG)=e*F5FLbogH{W?~opETr@gl?zf7G^4
zo1V<Gk8%53hhOp=M@QTGTG77N+Xtd!o_@%0j7P+O?AJyg@>9=Mo>QQ4&neKj{Gi`?
zfltqM{(b@U_aDTo_!CF!DtwzK9>g;}s!Qbo@xXrcVjTaFU-}*@=0*L!r;<O&%|7&s
zUc@-heEKpjU#g$z)%Qy38}gHve~LeH)0=tpK)>=6y_*NkgX&T1%wwl0SLBoZ=+kpP
zI{1!!>L(95*_j`+7yH^@-cH>+k<UKb&_3)0$qTLLC;UL$I`hyakA7|Y>o;FMBcE}8
z5c$!r6;pbeSN;5fd=dEp<bT%jXMUugKDDin^sQqD^X!vqN4n;T7pUKS{V?{68}=uk
zd60bg=NHD&=ZE~kI`qwl_M<0s^lPV(y!PS0aew3DMjId65I@j7XkX)yUFpd@>&&A+
z827c_xOMbsAN^22OzBO(b<Rus;@3QKKzu>_S!W(Z2g>)gV&uy{`q9PDyvir#KBWdq
z4SZPxu?}<pRxirq^13{!-+2S|L;dP%^;>MKkM+mr5d9Dx^CFrD&9{%bGLEa;ol}q;
z^kAMm=v<J`<zxL(ALyWqo_Z4Eo8Iw5Ui+m`{)h53zM=ZZctjp%K6;)9@kd_c&Ijx4
zV_xh}okz|`h<*z3AIJ6E$NuUt<JKA153P@NlKUondOm^Hc`k<LYeW6^)ke=eh+l2|
zXp;*c_}4biKF~Ofbo58WH;n7Z0nNh~d9?9^PDIbu-UDby^!~x$dG>eZVf_2zp3gn6
zyHAiupuFpO6?#8m+&ut#-jnDzABsPEpr@$s*iOxppTwv9p?(w3@{M@)94c?gpYnjb
z<(^5tHEx~!1?5Bg%9k+Jk6-*%{pE|l6CC9yuXv|Veq=tn#<lT*j&>Z6e4A$<aTe``
ze|ge;{q_|v=$L2ReEn59zW8@g+28*3VV!<xT$`TkUn@pB?hWNB@ogU%*V!M^hxstY
zmv!bvG;UvdN$JykdBM3IzweNj#0`1T6>s85Tl~0Jv)(@52gxJ!!S2SzpLOV~KVp3A
z_kEOg=+G}Z_%@E7crb3haqT#6fAX6b^NMxi%sg%R#eVX&e)iWk9{HB9=pP^UvrgN%
zb^4tH^rSzwy~n0^a+)8}di!aU1JX-G^kTpK2FW?7vGIYGmRPFd-$}tIJ<7|Gp7qX0
zaiMKKJ@b2dgU(CmGmO7erbqD^_xD_6Uvffv5eL>g=d}4LetqXA{_yL&34f;xM!xBT
z9@te}MEmj&^YNvPe{lfG$8Orjz2|}KOipq`^3sp}BN`{Se~;e3b4h-3lb4<3Wqk8*
z$p6TJKYY*^dG*tq{rL}m>Crm;#dhpR-@f=E7e4V1>6<^<AD<9^_|_IL>H~D?$N3xi
zGfy5PAO53!{1W0D-N+~T(Xo#<q!(@b`};cljh*<1bx=G<f58_!7}t+~a?t~R^wXF9
z<2-zrFMq1<Q^=krj_x(3$M~$&L#c<B@B?+9`Wt#ak}sgRj&UBx%@gPHihQ9h&uGgJ
z+HoIkc|~4=sr*sY?|kqbr#uYhY5UmU`6AD&ui|`i8+Wha{zJdGCO<vW3w_Z8G>_dN
zyFl~eeTQ>UKSUp*1JP55K<BJBzlzWGo*Pr>`3*n()&A&Pk3Pgl<dZxQe-YW0o!QAe
ze!#xk{2rgio!jhef5^V(;hR60M{e}t^sdKU-eppoMsk~P9O_5kzWBx;zYrJd--zhK
zIL;s05t4&^_<`u*Pn=qB9zJ87um?KU**C67$2@30ztT3&AFPMQ?N1N(HLlHW)<ffv
z{i3|~p|9w_aX!DKe{}7mpI^W@ZeR21ogeFu`^C1p-a7omdHU_^IgGsK<5xb8{pdh^
z;-9{ZM>G#wPyhS`nn!>7tz&O~U_a~3gT~S0U(PB0Fw)UakLW@BK<6Q(SI-Ib4?VBR
zn@~RW+`=yS)~0{^<I6aCpn2?$p8VxKi1!@+-P`@H*=O|TFD_GAz4h@Q_PA=*igEW6
z&S`wRXW8bp?4;8l{cdHIorjz@V4W2z+dqGC*EMckrFtLa-@irghnM$kIAYuHR6OTz
zwNJnE_HA6ULND>(-_-WLMts<RRHtW8=)UVhm4_Z4{lZ>nE>Us+dDIpwuGXt(hYI=0
zP2cv5^Zk3u<YZs#$>F|;p4lVX%l_h%9o%=I=ROE}j`e(N9r@Yg@W*>^bj*1RSJ?5e
zGe<pqK>K#pfB)Hilk-TtL4Gf8Vm!#t<crAfQfNNEq@U===BaP+7xRGi=F1oKYaea#
z$6x3}Tis+o_x<|GZy);kwjuv0vF9yko;f5h(u26-Kk6oa1-++%-lr6y=K}Nit$y_R
zwK#(O7k}bW9FtGK^UQpCPrO0<kykw9%X<7l@4?8UpIqY4zW8(wJAcu~HzYUy%#%;-
zhhOz9xy-W;zw~GydNUtg^t2&<=*PU+Zy){mq#x_hg+<T(O&wRCI!~Q*?BV?6H_mZ!
z0iDC-lh^6JUc?7KL61GqV@H0<ZuAS$Ge3p+hy2yP+T@OyqN_iyXK!`3dG;qKza%Gr
zVHc=A@IIb@p|2g0pV1dTwjaLiqmDu+&d0xTbhIPdKlalPd5!DWj)+fa|I|L#7tu8j
z%J)S`e~>-cm){!aCq;B?9hZ;e{Ah3EQ9f;Y#4oz!Ko6bfqW4+qW%jY3ynsLYvo5v2
zc$WWD^Vq?B^72RWLiqq6Df;v$PPEOBeoF7=M}F~T9n|l6g#PfC(vyDjkPkijhWIhA
z9nt=Aoq5)a5B<>jAM*`+n3qEPLwYBlc9fgm^+SBpr+H9+mG3XUXX97)Gf&%meBswT
z{o3?lU;G(ImmfP%ApbJodVH9NZ{y|{A%DatdB_1H9emjj@*i~Ldi0H>gFZgAp}eB*
zh;+>}&i?jOFL|zK59`o(E|EvSHhrV7jSjlxus_rf+08n1-A~}#y>#S*Jmgll^C$f4
z*Tzq+sP2SKt!v{t`GDWhi@H_6{G$!|k33@@xzH0Y+Vt%n!+E9drZ4j%8fSlc)z6RE
z&p5l-&;6I@EA?=kZy#tJeRS+^9)ICi{IscvUR=)(^qv}z{n}Oge36IoOYh{>rblu@
z^3$)j_15u|_&a!c$v%g7dvoFaH<**pyy6EP4mftfM*R@~)<xu}>IZh_AL2<J<#~@^
zvafjX9iRLlFPM+6b>6d?ryn2a#eLle`%d4z9{+^&&R+D)PxOl`Xg_jkdwxNe{n!nC
z{m}e~>|~v`yo*o$F^^hTE8-{mU#d+X;@3QUKz^oP&~^^;J8eil>*$BP`XPJK4>XP*
zI_7KB6FnFg-}sg%pnj-dUcm>x(X|i0?2n(g4t;37dFa|lKf2oXgHhl5Bcc!Si4T6x
z?)stm)|-!C{rDs&`)Q*?zsB)vo;K8PKXkOAerP?c>MQ(Sf*<o2^YKS6d~2Hr#gqKX
ze)LEVbhWLAv0odK10DPrhxCFjJ6VSgKA}8|Z~D}ZXxw`I*dH2??UdiiOZailkO!TJ
z<RT}#>WB0c>ES=h759mC<^0xWC;48RUimA%>UXYN2id{Ai0TRR!~xU~?TZhHu661N
z>!5k)Xye!WSaiv6-1lng6Lkyv>I8D(2R;4fJ3r(TsQ#Bf_?z(}L>Kbs*pB|J&2Llw
z8RJ14{fPF5_JMJJz3tTg_M<;YK5+~6i(_>0sm(4C%}eobzIjl;b&+rLjC+pM{;HAw
z<WKQ}Z_gLxD&m)a7tM>~>;vr&@#$Q1j_?Z@&mD3ZhtXfOAwPoj;yaFdQQp8Gy82-%
zKgoA3h4etb^n@>ZR{!A}pY##=F)m-myegmChrS~}_=;$rb<lSr@;W^_Pw=5X&T|gx
zPvy_D|NpH4cBe;r)(>O9`ObCd{y^J$$p84Ie4q{ar+OfwdHe|KPoeq%J?%IjJ^qeQ
zek|_fy_grBFXBKRhWf3Of8}lGP2|J=5P$f`r@RZzw~u|@cR}~A5xrl+pZU;yZS(LS
zksj!ezVNAEo4(M|CWrbBk`o^gUHa9wzjY9O{F)c}isR_w6W{p54}H<E`#@+OI_Tmn
z_OrMBVB}YSM06vfYyYYqzIaZ@C!|07qGugC`mKvdK749hhd%v7_N8AaUd+=*KgvlC
ze4(q2Zbb2bPjq5i+{#Dtp7T~b>o*S?Cx>%ie$sCrNdLyw4eUd&^kQ7wI_)?=_Txvt
zeInvZzc#zG8#{>;Xg;(ryU{-+7yI}gN838~)powogZ)xS@Afkff9R0QJa(pE{rm%6
z>y4vhJ>*Bwx$NAA>OgYF`LQ2A`r~^2_R~fe>NgKv=d^kJo?PaehYoq5d5}Mr*!!%j
zJG`E!^dMe}P@Jad@KeZs?8={E9A|g)*&AA?&SdB4561b0_4@5&fAy|5y5{S*j(qrm
z=4qoVZ+nh058uYo*ESCw>(Phe2;cMv|KCps4Gl&1esqDE-y5^XLXGmKzq=z(%X9n!
zJ^iV+{uDYN(3S5X|BH5zALY|1H+qnM=}Eu5tsT+6*7>~)zU&{{=#qn;Ai1^8i*(Q@
zFaPjd&9C@F^nYz~@FQ(>wfUua^oB3~stv6(A760lN)N8Q^$D#Sy>EB^S|9zIJoqv%
zh4cjJ)qM7V>U8p(XC1!r1;wTN4EIphqpwXL^v2HQSD%P8@*4M?A%4t<<mI>gN?nT%
zxzMxD`CcpH13mj{L-SIIPwS!e{7FB(h(q=f5BmA7`RqVmsGmLTqo029k1uGxcF{a~
zXBXJ(_+AHHcGoqn{%`*3zE<b&cV549t8;tp)vI&s&;Hwe@$O4?U#4@z*}b|i`&oPK
z&aFDP>e6@lb9!~Z_WbQu>EF=M`?SR$d!qfY{F>eO7%}~(N%^=BZ#)0k1IAa|drvri
z(v+W1%p+d8|HbY8d(Oao@Lzk(-hJ%Ie91@tz0KRNjIH7e`;Y4U)>$9taB|11u3c!u
znN_-Hf843k^{Ll#c>7B;`XBSdck`|n9X{tr+fK_n^ck>vw*v>{FK+qx65}5jmm6QJ
zUAyPzY&SJW@2>o@e=aa&dVX%}#k&sJcuW=Xwaa~zwmWmw)O^z&&OP<FcTC8;t+L7U
zk3TXbfB3r>9CY^OPv+#Xk8$+z0asgghtY@sZE!w)k^N4;=8X~Auw#Y{9JuQ<8T$0l
zrSq1XO}zKvymHA7yYGGctNAC34(!`?-Ld&`mt248b5lp;J9YeY^UYdMtm>!jKFjZU
z`d_Ey7ccqpCOx;DUga0vR{3f>ZG6x8{MSEQb;!>y8eP>3yS)6`y&bo{ZgQT_+U9^$
zcbl9)deF{4+whQaRir0~?%@XyeWdlJ)AHLceEEu_mKc|h8aV2XzWWTXw#f}=@A}IV
zUcT(3JfeN!Ki7QbhKHw3%Grw_7~f#ir>}eFmJf2c)a6&Tp0wbE>i+D(Kk;S#gCC#P
z^TnM;R@bv5|K%^+_1U=lYr{uX{fQhA&5L;5DTlNl@QXXXWZ%d)IoTcJ(>OWt%b&!7
z@%NV4^u@au8J|afTDlzzpL*MyKRU8Uc4kAR|G87wx#ZPns(yk0jSsu${$HK^N#3#V
zFW-9PH}6;Ti+*wX%(}BT-Tq(0s`=!s{;Td#Ie2Jp9=+Up*Q{6f9y}($Jp0okyIuQX
zzRj^a?X}H!hE)C4IQ#IQnVTH8-_O>3CI8dVUdw)Z`fC|~I{C`;PksE|8#CkV_J_Vp
z9dyHSA7$Hi-~8xn*BF`I-}k6K%S?Tz+Sb3siGN*iP@i$xy?6fpqN~pNpgL}!W4iac
zcd1+6&Tjn8uionN;-%TljvWqKV(BY#$Pf6txLo6iCGvZ_PWVzjl|LeiM|Aj=y!cYX
z@k_qA*pnF~FO>K21I1Gc*|iASA%*y@6;HYDpd(J2Ju$Br=_8`}i+EzE#d@7}<3kyH
zF0k&q8&7-x<*GkuuXWh|HxJ(WuAF}P%leBx^yK3E49<@FY458q-+XxX!PR5$|MOKt
zGWm(#rVicxy5VQPkjXdfGIZJ@<FD!PVl~gnTl(3F|Ji5xzjmJ4W~Bic+%!Mqf<Bu~
z$?$`3am(+Wi~Q3*#&7&->-QFZWMKBpDr0}z`^iyNy^z;_&^q$br+x7QNA(~4!dg8>
zSI<>?r&s$rFXb8a3;%b1JKrXpc|)(qcAio_SM@uW7Cm?1_jYRaK{ekxXV8P}#y<Ro
z-nu>Z<ZA7gJut5~kA7R{?>X$A<#*j<MAg6PS-s@EGtYUzZsv*SR9ihV<ks~D&YC#6
zdJduQJWvOr@0_u}wmilke%kt9%N%^yo7H-jU%*=J%g-aq8z~fDDa2<AckBJYs>}R&
zSQfGOu?>S)?>H)p=zKi?mxI^uG=4(X|K9sMtyY<oozijN)BbVx)NF(G9(a76hsIR<
zyM1@fk5+kMWOm3M*S>JVq4#DpSLuA^ayz|O%_EEcuHTNgEIB57YD~kni%)nz+kfg>
zA20ZuH?y{{FWK$R)n{b(nK0|>8^^vhJ$rfgf3ADP+S9Y2zjng3w<nL!_CDf>Hfz5-
zrMf>p=pEg69_~K9`|gAC@#_s}JpO|Bs?P)J2>DLk#@_r(f3&}PMcpCq$#b3q%%g`3
z|9r~6e|+!p?8oOkwC2nWUag}1`TDj;FWjpC>(zY7{^|*OlF#K)=v)|l$-W)${`8aV
zy$jcWd+1*#WtVUI`*rTPb8O~0MqWop|A}WFa?JjJn@~Ns9(^Y}_`op}t967rh#lBb
z8^3$Lb<^gzb^Ew#H}j0EqpVX`t}$|<{vQmTT;<m}gAa81BmY?Dw)0jz`otL-{>e{n
zc<sSi!wJjGuIfd+dahF!LC;z0Oy}<oFI9HBaFc)M;*b5rF}>2W^OYUVKVsc0AAR@3
zewlgb@dNd$b5Wc*7wHjhwC4F|?7RE;jJ^l_>c|B;{$NBk?)V2gigV|%_4I6?>$Y9`
z-xvS*{ml5BYuEVk@;@C{-A8+eS#y54;tk`g_G`E6+Z!zPqoGy5JM_3t?;JVi?OeQw
z6ZzNks`^uYRwp`l4}7%G>`%w#)%spssW<4Cy!4=6+UKElC%<}I-%NcaABcDLIDXWf
zp6fiHJ10D+J6Gfj{6hKIyj7;{_{pBzjH*5d%NObs`9wdxIXBTG2meH0-Xovq2>Nu6
znNQ#3)picUtp+`R<cBBSms`)@)wA*uyUM4|U;K(c<DP5CL2r=#_#=IYU-_I}=#8Gl
zr*-t9F1L?<{wdz&ar>E%E{t(3eyw+Is4L|KdO_E7hI0+FA35YJdee`Ok$1lH@}lhr
zWoQ0%fu5sQf2n%zhy&*(yIs=iFH0}?-tf#h?ztBJvU0~_R~-FZPA~SyFFELyoyC`a
zdLR!!B0qmJP9FNFkBIg!LjGBV{IaEyK8nyiV-fNvbuqf)PCU~)yNM5Z>o<3IeB*}O
zU#i|;qBpM)AD;6d`Nbu9A^XS!@@GW(CZgxe5<kD6o>zL1&m+p4DU_%AH<Tx0UgiJl
zA9apABY&ya)Sc={^^H16{dwAbd!9S#h*_C=>LK*iBkHuPH`r^`dpk_b@CofJ{?z;O
zp}e>5)PA@8W#rHd>L;)BoPNZ&JQw?oL-`rnKO#Ex>fC|G#hJXrPU6)(`^#JC){5xL
zYv`(j<Q4n4H&SoPFZAc!QCFhtJVQ@@p)co-b_)3gdN7`6^hX|kst%MN`42slQ$J*%
zh~$VUKH_}mHB8B=Uwz@;!E>4CJkQ1I6mj7>NL-L#-ls?YPmk=yj_P{v7d$txBl*-r
zMJPXb&aM~v6(pB?#eTca{I6dv^+xWwNgeOGL4C_U<fadL^_)%Lo{J*NBN5q`9<{CS
z+U}4J$2~nW)2|Ljhu_H`=8@NY_GSlppIqq4qn?}jfxJFxdYhi-uF$p;&l7FWG5pbU
z1%Ai{>Fti|^1Da>ep+UK_YLkX*uT|Q9e;kv4<}deKgeesAMP=nv-sf;^gw?7^1gkY
zJM8Ryat~lX&-2bD_ddq?Gr8=8o^y*n)>(1Nuv71uQ++>To^kRx$JAx)YF~0`^8@z9
z2YvBxc0*4;<S+aVKk^&Du^+p7&ap51i6`^PiBEpczuk9PNB_pjV;(u;JnQ)*KIJ>~
z%+rsqaeSj=A9;lxQtQMAIuPIdPd|O*i(lD?e(__Tesc0x{y}f-?7Xy|z1;htM=taD
z2fq|&{DOb6t9{68ALoR8EPu0udO|!q&-e|0bY9^DpUwgGl5u=NbjaiVhkf^3>!9`*
zuQj{+T&G_09@_rsv7h@Obq&3<GkfZHPDB(}>Sg}IuiWd2S9JNW^V)p|{jnqavXA=>
z`$j+DNBm1XU(;`w$?L8^Et3zS{D<#&@5j#ZJ*D~#l81c!j9uBmcSGz)Prd_Eud0vr
zvmZT>n;rQFe(kFca*w3mRA;ik{UE;+H~2>v|KxS<s87|m5&4g}X9x8TyF>A5ANkGs
zNDuVwIfEb53%_7z>!E!j(t~^!ksgbX-csl}4)PQC#_S}X#DVxgm!8BqeakcAP`@_+
zATN94haUMu3OyI_SNmwA$A8#~y~%@r{w}`w4LitBP=4hX@(I16%iqv7kN?`oJbcKn
z_Ct@}oDb{(=@+sWv_6I68rlc44|I+}`i5~I{V7BTrsl<d_N3>C^h2MJ-eFv4p12ps
zO+|T1e(~JF548CgyVE~@qJ8ND|M*46Ii=2)pWH)TIN_Bodb~9?8=kK);iN-8%{;Go
zf4<Y@Q+7J^f+?Bzpy+!}Fn-t$!}F2<H8q3ziJKjH`|Hy(=s8^6I5)M`dCp68J+Ful
zaySpz!}F7M_|Qg29Rc0D-L_!AKG$72F|)sY(ATeR-1#cc$afK)i%{MYZ%}-d`1yBj
zr5;K>eD-}~=?6{o1Luo81)YmfT~aGLmtd^dJQqZab%D0?FNN}D3Z0KdD1S#(M--v+
zG@|n<h0d#{qWU$WdZ{QroKLl4w5PheR#fjo&lhW*HD}<4&(Fx5*X}pfCG<~Db&PrH
z9rd~A2!5-c;P>cxKH%^8QirIY(1-42toNP-UHZqb`pol?_e9$CkV5a5p16FI9#3yF
zBlkU#?}+?v$2@diJ>s^t&OL8N9?|m<K2qo&!1G5$_dO|8zZ5;E#&Q0{ekqLgRP=Xt
zfb0=5`n9%tJB9JNUE6(QM1GaRwLUoJj?>p~sML$<?-a(ortP`ow5|3z>8TH<Rewk3
zzR$gs?={%peTU!Q@PF?^+*|p+&Ao!}<Gc48eb)p3n3$tup5O5;``kIpUGV12+<P|f
zsrV_rtmkL=_Io(@Wa7p95#RB9{t`F*-F)#yPyF=yv$mP_r{8{B{T-h8rU%d0`khbw
zRoi^|%5#ADh4PGhx4s*$_uC~`np6E<0=sx_MMppS{E;2VLofEDmv~OokNhEjiGS<K
z#V+`eKaG2TieGK04sd?jmwwp^zx)C}Bi>zZ?4^C@RKMG|kNwf-H~P&(m)`ZeuQZ>X
z`6I*!eLIId_u}9F<YpJ+{1_kXpl<VfRr1i2dHfyU{Li_-KK3D(dtv;sCwq{G-^sJ;
z32|e-`bfQ_E|U-4m&&)+vA22rlD~-$a`AV3IX|p(zOZMMiyYo(sRQh%UT{vx`zchH
zq>y}b3+bV$s4hyOdyuB0^C*R>b4@>c)QZlR;XAB0Z0EOURDW+QuP!)o@%{S!VP<Zg
z_g(6*BW4eLy2~lEbMu`C&XH?J?lgXfi>FtAZ*HF7lR5u3fA@FiG+sWnTF-5D((#up
za>lG``@$37T4&2W|MPoj`Br|ASNy(!f4jdI@6KO&K%7J0L5d&o`$oGP{_EAF=Tv{6
zB>%`u&PV<7hP>Bvr$21mezS%~C|;a{_!FP<ACx~Jdqed$tku88sde%-{^bMbRYdtA
zh4NPlQ_r*dJ@=$Al}F^yrlRviU8l{jd_Uy<wBJRsqk2u9#*X~exkTU4`QyB02kZD3
z^j*H+fm<g(voC+cFS|M)w9SLg1#*zjKJ-C8{_7m_Jv)B*r8>blxvf{{(!bx4(TDn4
zJx;FpJi~72ifjBsdMd)0AL!5THt@|}>M-Z5`0@Ru?+_w-FUx;@52nt5;#PdABh)`o
zp4jCt2Yj#LjfM*Jo<baX&j6i6-g~I8_=h?MzuqHEzWkCqwtVo@>hHpg`@U(P9s`Co
zuHCBgYUl4Rvfy!TDu=&2=(&wom{t9~qxTIk75}~~@E&F2?dN@ZV4qeM=P&*teL>$*
zL3)LE{%!3`dp_T~;{A^ILeP65`1hype&eYVTUW&ApKp6Qdui2AtG|<!C)vfl4gTEw
zKy{w?F`nzaUttI5ly&Ut{SN)93)q{#u#fx7&G$Lsvhgd=%%JBb@ynm=%U{&9?w9bP
z&Q~|7XZfxD>6;w-+1Yn0&Ov^|KHuH<gcFB#X<b=r$xD}9^G}nrwa@7OP`_mssQ5jj
z_g;`*`6W6KKT!P(#WQ3_=p2Lm2J(L>Z>Es{&>wrzGy537zS}MDcj(rx;+~K{h!1w<
zhY{s%d0xKbckC>l=)rvFhvz2zMU*eehdw#vnTXbjH~XsR<SX;=6VdZw5xO5pp*j_+
zGd(}XJg@EiWlt!+BGPlcSoB;)-w{0@tN&wtug<4;sNQ9N_fhI~=dZd#y<Bt-$8q%p
z6hHLC-s%U~`6r{hyzuI*Tzv97=(~AHe|~4Io`Cj;@&ptwP<;Zelb6*E>O=eU4|!KV
zdF5Gk5C79oKh6Vjr+&h(xOVP1Hy}DGbdDAwxg(NW-hu21V?Ln|NMDe>VB}vsrcho?
zA%7}D=U)nA+-vif6tZIxiX(Q0>IJ9{jaXz?<Iz6O9T@4xwmKw*>Vb%VJ9)D`{@$jc
zQiRUyT2VcmLiL0?20pRm`X^6)batlhi1^B<SD*d<rwtWxFK)FxCy7V#uFYThC4bj1
zuc<%ThhI5IotN^we)k;loRrU<cb*sVFW*{cT)#MV&xjv<I8VfDlt-S^kG_0?4|=h$
z_>o8Gg`W7WebJMrw4r(EIM>XF=8JpsyVr6qLVV-L_mu2!J$m8*AGM-$Acg#oKc>(;
z@@sp)<-WjkJp0M_p8MnP;>piGP#vZIqeuMFyZZv;?rqou(l3;coD+VZV1MUmq~~6S
zUx-s}{*F)jz=!)wdED>${JjYA>pqDe<CA~Tx43Zs5aY!?3wzNQzUjmB969`5gNV+T
z5=#w~8fZBUcs>)S@`rOGV$1DZQu;b*z&Yri!hHmEFD<`22b{m^E$I9(?i^7^IbZo<
zY&$27$9lwhXPvs0pF8)VadiXw&OP;`bKbb~3OX;*@jlLZi;i<N@{PWFiQZE<zxa-N
z-o1?XRo=(4yZcu9@H`0poilcF&q#0X8$4IKH*jCcUg*hN?Cd?8`)2Qp@WY?bWe4?&
z`W`>(WcMrhb<gg85kLN(o8N)(1N`$t`#^repZNp7)5Zt=;1eDF<ixM_>U?p?&+ShS
zkbdb&+^Zj*C!Ra-Z`{7(M%#ROLLAwLz1W+4^oK9|+mGDnLHE@3;U2;J0(zrQbm^y7
zL|>iFpUJ@=<S&2MP#sSW`O^1*^kqK3@I9^P8GpBsejvTLFZG-Zjr)$my{+e9^VrXG
zx#xFsv8Qu{-Q+#=ypLiZaYinF#xL0&+RytValwC#^CRonm!FawvWs&IdM<_L%X{vr
z<wr<Q^dgUv3*Y#`KYj2U^E|JcPab+^Z}Db-^WD4fZ*6|Tp5BkMyYr45P#&Rg`k){B
zvY&nU13mfP(0uFpo9}?uzJB)WKVENE_3yc}FF)!qXUk=mc)fKa|3#nPAwK6k@zAfY
z?%uA^J(c$l)<<+6(=&eU@422`oI}nh=N`1qdBmT+KM>#I+WoM7<QwOU_k{A=s=F`s
z)YUV_RsU|Uyyy4G{E8m6<t@M4=ZE5weW3dnek<Of^Ta;lT)yK+{EYs@Bmbf|=Po|&
zBYx<IJkWiub5K2EUw)52sNVVQ!fnp#^<3M=i08h4^0q%;d4a}yal4*>-)*m#+B8Dz
z{rzly7V9eCAK;hV?1NwD2K~`XME0WxZQtFwr;=C1rN5)0&T`Kq-=N1n=!!$g&QLy!
zNIr4HuHu5fJMY;~zjGdXzwN!4xTzI8KD_JHm45eO_3!J7NB3IV^u+(6ya2_Myz3lQ
z574K0qzCq7AO67p;>14Y(~tObE{PLyOV8}iPV!%*i=KQ&5A4Yw)Xm;+steTD_U-z>
zo`;RNrcL$t9pr)H1l=OUAAOTco1f{YXXh6E`@8DoS0}nJ6*uU)&t_lGEA$TO9Y5Om
zkZ0K;h2oK(@GXDPlm6(J*4ZCF-uL2{A3^%%f9Uhe6p}-op#F0IfIj`wpZ$DSC(mfJ
zqjm1l#gl#1r_lG3<WToGx1hKa58_q*rT#&O9`PkEyuV;K_7Jb~xbxB9<I(SXL;H}Q
zU-DD@=(o?B<A&tdw(66=I`YO%A6(+oY{4TJI()_=bE<!bNBwDka@*g&{M0$EO@I7`
z9*Xd^<@)Zo-OH^i5!o+=m#sA8ljqKFXngSb54-R7U#%L)?6KS5|M>mZjowpwzJk6-
zSNEOq`cMAQq5tfB(8+HMd2wK?M)Yo;boA{PzSGcn{G;d2I`S8-zl8MbToFh7SKNtX
z`ePsVc8>CI@nKv&=ld-6CZvDwo$&+l5mB8RF}{D+R=@HO`F!XbOKx}lUacw+Uy!^h
z<WEqZrw92}UXF6mlln=W0M(QH-?^iHAWsUPIR4t*ckI-r;@?m3oof6YB0Hq;h{4<6
z_QD|xR#Ld&E2mU;IkIgfBKvrLabFA75Au-b9{ElE<#!IwC2h}#==uGIIF_f)H&4I&
zL3I4y!nohx;z!<u=(%^cPP}?wD(^z?C7rkU#fQ92etFpQB7Mnkkl*2t-$8WbRs1;r
z=@*iRUyCE>EmS|ny3~CB3-Kv0I&a09{7YW)*-sna(7hH^2SEI!P~8pDh4LYk_fp9I
zke(pA(76@SdvkePJwQM1MdYFAulxw|pXhh&O&;U?B;F(OSNDGC(vR;b<9VgM%!Yqn
zEc;|i?m5o>es@4`@`C+*zv(%Y+~(5{z2aZIi3f7>Cv^_~q4^Q5gZ5XKp&#W^|Ir)!
zdR`W9_BUVL;L|+`e^GBbuho(0%V+dKPx$9I+WbvENFhGOr*jUU;@WxVTvK=9Umb!E
zajajSaNc?Tl2_F6@|%AAXyX$f#__{$(GL8{e0j<|<NOgH^uR9ik-9R{^E*8C7<$eb
z=zJCb;@kXs5xsb>@fYs__>FUnziWF=AgAwq{9SK$BR6|^US)UB`TWN{4u7eB$MnU&
zALhOTeb4vaclmoH{@$6k@rdqU)Te$Aqz=<|PvCb=zU%b64BtyyXCL)e+((<=m>=Vh
zUD#ROzz?BuD88H@5&1zx{)?}9h3XdZf`4)6IfNg|U*?Nv`PKa+{n1-Q=a7A%^9Ran
z5#^U6q}Pb#PNDOlUKDrDMR{OuVUb+XUhZM|hk8Vv?|y~9$S3^7`RUwr?(!>k;1}Y_
zx$nH>f8teMRPV}<?opuo5pk!EN5^x8Iz~QrPB`cBM<3qPppQ@blb_`==O}-6Zt9Qc
z9=$|AP-lCeAs<)s`WOFhoV@6{+c_+5)XVsB{zLf@;x~oPmm+lTd<nnc+dV=G-NVfO
z-cwzF^841+`x*B=od(?X+RNW-)7bBZ{nuRV>{gBDL+=y4_j5nweWd4A_df1}j6?Od
zdFVm>;M4m|d_eRex==oZ`qeY|f$q7W=M5O^ZRZ(`<KiBwOKU~>0mk!9Jp<(@=L`SG
zpZiVU8LMBtKjl~Y)r;a8eeq8(^0s_po%2aPqX+S-&a$64lLzn}QG7%Eu&esP^P%T7
z&l~cV?>GET4IT5)b1%e?wEh0kJjg#iXW3t#BR4zIBYOC9f5ac%x4OSlU(+l4(GS@{
zevyadHz>b3_x%o!U7+vd`H_Bd@fUWpzc#t?tDpZk580RB#kk=|#(mE~kM!=`6+i4E
zZqVhYReF5}tlsUw0lD`E&Kvok9?30!oNwxW`_cn6-?@YiJJ`qbh38J^qUR0vGS4~W
zoO7PJ|8_6tdBJmn^Gv-@KK5b{`_Mmm=nt~5@2%WF^B?sN{?X%4-n&8Pn(sHrAN9np
z@(TL=((@|+aG&FR5dZYeKg=hWdW?PfjXFsj>PIKqS^lJd{=o0Zr(U5a^t91sZ_jh&
zmY2<g;!mB(-rDreudHJy`)lJ%Jj+k=I=hft{l%}?SG-k!57e>4K}#%sMfKnDH12s;
z9g9EqV;}#%ulV98=#Ued2jwOAeC{Ra$-hG?9_ZURJw<=SpK<<Tebg)Y@NGSL#jW-D
zk+0c7eZcR~i^z|ioBUorFfP8;{hnLcU%k&SB6{ATXZx_Lyn?Rv=9A0$>^|4u*Yfu#
zotw@9>pU0nf6wRq*}sRYP0##Rp5j-=q4PkyxkwImi{~rx4Ed8hz|Zi_{^Alw{+w4(
zJpsi{b5TBk{2SUoB7cPBh2lv36(RphA-zR(J{O_iE1%MF-_!na_SDRKR{kTN*nz#-
z5C7(qPrjlD{-MtFKFK_Kh5DiXh|VMLi|sG3%Rlm;^BTX-6YIp4{3FlFU+9pZU5(2_
z^x_<LUh8+y#%}gg7g#3`kVF1L$KTac$9TTN55JKQ*+E{jk9q@=gS^mubm)tn$?5q3
zpXSMv^eUf<Bla~<9w0yZ_NT|##=m`_dnEd?-uWO;S#Ljn>wZf-;J1qH+C4XCyQ#VP
zP@Z>gJ0HBKQ2()aMEsDG9h!>H8~GWs6T7f0d)S}fd(Mq@DLv9J{qRTehL5Oc{p>=1
z{1`{yJbEJ!|G{rW>)3}r(I>Zk$**pK_TyJ^U-HI$h97!~^vHt`<K*^S3CY1O;);Ip
zZyq$xzuAvpT4!H*il3-k)MxHn<r(+E?1*1_wx2k)5Bty)KcWwDqz<;u^S3ts$t%yH
z?_484bf0J)I{rSY`xtr>&+0w$(Qo93-1Nc!p!L>~R~!HG3?vUEulsTD+4+fe&P{cq
z`xob>dW~N>zxca(@}qg4C;310yrAxQXxI@qY_Rty)qfY)`JtY(4>T@*^ov7$;7i?s
zZ}Q;BIQi9o&~q?!Zu{MZeyF}sr;yM6hV}M^_`x@R$m9F;$UnKPr+=tU#*cFqee~jd
z-<A3f(LIj8gX2EOcX)o+;_oK<ou&I8_XFM|d)`Key~#@+e&oAzXr1@W5syFj(!0LD
z(AbPWlh5-Kz43p3g@5rOp1comPwL)Ap74F0c=g`Gdierf{~mz+XuWyz0s8WgaVRf%
z4s&ifM?AMiR9DL9@~iJu>7CtD&lT?bB8txxsykAMuN10JBRV&rbJ%x6><pbh5u?0O
zzS!o!DHMOT;<$ruy7I;kCuNWwQYbznx@U#H2ln?f<t6z}f4wN4yf>1M)i3S^<s;_+
zKcg4<kRF}?@|1P*g!%Grq#yG#yUA<5gOR_gdF0WNJ+d<!D)c2T<WKU+E86@Gf8@Zo
zyo+!7UcO@o=OgsKMBcL=sz=pD>T~=;`)T7xTU<c%@lRg%CMW*I9eeOc@kBmySZ}^I
zyG6T;Q}eXRk%|xf=+=w)Pa%CpOyw{A{H3YL{zd40v-;J0aOarw#5trsRo`3ZJ5}eJ
z=YHo9zteVpsjt<a_S3KKbsmU&^|Eu*zScz)C*lSl;)9;;gJ0(*KXxvok3abg9f&{o
zMC>6>@Z%hpPt-x`LjEr9(C6pk&^Y?!Lyvs;$Co_pT&8Dje2PzYk{7L4*KIxSjo-hq
zXZs32P`9v?Jc~~1Innz%^%*&0zP4}wd++bGT4hq!_Vp#Z-MRXV%)b27^P1;5df<<q
zk31K-_oQdf73zHZ`o4xf)xXX!`f+}$L*+f+L;C%RI?V67)G6NQ(u?;3{KI=CbbVL4
zX@159eKwntZLr=0kFWF4n2i49KXsnxcjuP46!*Rl^c-uRHonB0-|s?ofcV3|`Mz^?
z{`+o~Kk^fC;=WVf=dbAVfAPzH-tUNCa?k@ivcI^M4}2Hq`!{~WkKFIEmpG<hb|pVQ
zRTsFoKu><bulp4GfSzmc<=j^f^DFYn7vkG{Id=BkVqd>o5Krzoe0N74>-hzHxOX5I
zeef^(l@Hw);a~oh|6zQ-(Y6j<=N7pje#zmyGcJ#_v-5%7=o=mLj6?L;m%oyOeDnpq
zKco-$`1}HY^a$Mx(X+O?-#MY~k;nWV&fnGde8221<7X}OWZTNcpKP^!n>AX03Hg<L
zL@sq4fAgM3zwZn1B@Xzbc=6s7eevqMG3f7Z4EWWN3v~R!h)n#Emw(6`>NE8sL<c|g
zBah%;9<`7AgA|G*@grY|5B6eT=y{#o@~QVd^bft~XfBe=eUp3$)g#8~lRcb6&M|i2
z5Bk+7&Rg+HFY>lLV;+4#{PGueU^o5F8~0MqA@iL(&LhY_#H+ZHXPi6cL3Hsc4?2H5
zZ#5U?d2!`jc0Q1s9`F_IiEnY}JX4RMM{a&6?jgN<4itatFZ1L%`?^;XC-R!_gZ;ZJ
z?%~|GdoB@I>?5DrpTCiVoZ`j#A>Z;h_L9H(fpdd>=|vrke|A^@Iycm(&Ijj=_bk4<
z^qh!3Iq1=M%k=W2%~yZs`}?+Pq%Y_`(mwct@{8|D<t^WVs$<=k_@0iP@uxl!x9HQW
zdWK%1ydqEg`z)RV)YamYKamIe-Ug!2Zt5BS=DFDSvz`~&pWXSXe5THn@6_AuC9kUw
z@GGy=xAR!rzUGT(b%*;G_r3J$Ifh@@-+J+iALHzVZ*=TW-}sXk<Xz`8yG8UJoj4Jn
z;?2Ggzv7dB!FZ0yf9%BG;$K^yHqY~fadn&gC~sItUV1bSdd^bspo<Ut(yO>w`TR|H
zzwMreMt0=K>O6euSD(up_+$5v|Mtx4_g>ew(YQKPp7gykH0~VY5A?(C?4!<SzvvI-
z^m`-r)=w||jGXi#Z@NElPIx|r@~`#k272Ic@{+v6zvNN*mOSP|e#maV3ltaDv5)*m
zuk1h`_7l(S9{bV3FTHtAwNAYV`I|gOAN&;G>?W?_xVAb;-Ru5|zpMY$OP(*CtMe=R
zz1i~HjcoU`TNkeQcj-KLiGOwx5AtvHXY=Tnz2rmZhx}z9$Unq|=MDEt{FhzOWe<5>
zeW8AKo_KDM_wnN#6aUUP&r{+A-~3h{<Uis?T|j@(J&yY_=ezS$UFJCgU3$hp|L51v
zb@F;Hk+-z@E4kUx`{<OPl82tiO;64Rb`THz!1-uj_Ctr>Bk~{mk0{>h-8g;GAN|t@
ze^75(C%>CdzmcB&M^1FvM?by0KlNTty`!Fs`t$q^`62q^!}%h=dTvm!`h68tC;Pph
zzvrk9v`*fJ;?jE(|E{uqp?vH7a<AolagK{q=-iS&$mMxh+&gFJ!M(comfFz2*Xe%P
zeTMHR{JWOUN%<zuQ-6zFaqK+sUZC0z?@gTt`1HQn-(?!M+QrX2{dAj(ef>M8-WzKB
zZrQ!Le(2wKbbsYLj+R34E^cCcv5)wnSI@)Z-nlEE(7W$#=*xJdBYurL2ke7SZE++I
znGfC9L+?4<5BQx6`$POf`!yBi8F>@Za|+cvP~L**){E-uB4qa>L`R-=u8S}7%7@Mi
zd6M7p2Y$m3%wr#R6Yt`Sy?oc;{#U=}d~u2ny7G&-@A0E9*PVG<tBN{?9-Pbc?m56X
zJ;;;nCx5yx;qUUWw(}2N81?L&h4koo16}!9{L1U*MSk#UoIK8L{E<r=T4#UHRrbd(
zKQ`|7#h!ccZ6E9Ps}G!So}Zk9_GcI4<g<@+h@R;~8-0Ab2g3({P;WSQ?S~KNxce`9
za^Ghh;?q3)(FgtMCl9`@)3y%sPjY$AgZ>_YbDSLX<DL)y)i~eeu>F3v<}0~#A0OmF
z&%XRI+MgcOE9zS7$i)xnpS`VPhv=X52FZuMwsG$v$zLnFmxJVj=+L8kO8WBm7$ARi
zzsDb;{HreY{>SsKanB7!sQyo(yx@FhC&(|L{14ehUT_|=C;Q5O{G6TTVdqCg^lC->
z;xD2+okGt6^`d(iDDOh?3iXR;eyje0=E2A}eX$pP(gXW?j$;qw?4qqs^1Ngnw4b~w
zPl#uT4|3TD8s`W2;!ostZlFVd@&bDF$dC222fF-}UZ8yJIY+$tJ7n&)tUv#kgV*mg
zenOU?xY?1nzdkKf$GETYeZ9IBJ^yZi=L7%V`;lws8%`LwQ1$P$s*Bz8`QAfa=DEW6
zJou3h)Jy7R|4zN{d_2GUZqV<w)lvQ(TkHHC-bvHj^gMTkwiVAaZF|mnc(oa=D&*XC
z!;km;*ReA*^Sw`BcE>Zm_s9WrvWd5!_vwLsT2=h}vrj&;!s|W$vOs0Owl9udWXT09
zTmIs`-(GOo0u}SNSon}}hfHo=`R=|aoH(pY>x%b0_E!(OFQE^AhnjvppZPw^bA|7c
zJO`4)Jo@mvRCcz`^QC)6-wS!J636Od_YU4a&^x=aFMIml%{=*k?<0<Av-Z1FGWE57
z?|aB`O}||xue<)V>fhl;hn;qPf6(n~?)sm<L+tMnr|zBjzk3w-W9}in?<Kdk_Ydad
zLw@#qSm^nX|D!K{Jdb+r$G7^Qz2P4(dUpH!o|u*KC(jr345QrY5Ot9E^ycGV-XXX9
zK>5x6raZ%5-V^DUkD&U+cR}v8yk}GAsLRa9w|dY$5IN*adC~oqd~P4~%~RKq$9<6Z
z0_LFu<zeT8^Fv<OZytG!L+6sb=A7_+>U{M4$v*P0I>tQuz$gCck^I{DQ1?2woQuvY
z@ozkZ{5OT<gubto-}&RKoxi)tg2%O~c&<*N`~u|(DBnQ&t))@kx3BZe^S?O9r#!;0
z{EeS_zW~(-^llzG)YtAI-Rp@%^#lK6SAMGBJn?2<@h@IHAMq#klsK0^_@nyBdr$tI
z!d}PsI;i_vo!9UDgU+qa?X_30&aFTDZ}-K!FV%gS&JAbx>b~q}?X^3%>fEYJ-{sHg
z)&1J@w_BxuL&M*nynlyJ29C+=MS1edt``qJYW?AP#Gkhw{i~m?_fa142b*3n?6A?}
z@_I4S*Dm7wsqftM&n^eQo~Q6H7re6P!Y}pBBc|rZ{uzfH@UQP4@NS;MI6u`UPgC)}
zH7^)+!a=v>5r01W#lQAAW@wheqWvPhRC~kQ-rs1^4+mutQ}XIxbiG@8k9_QrY;I$b
zoJ}1k$GpaR`Azw!IG)&cTH9%FK9)i8K(5sKA1?gR8vFH_n5VGFu8*8^>-L8(HZH3d
zqo38=?$`CLp8H%hA&Z!blT?4vzO~wEn?+7K`<0Jg&r>*~>zY5<e)P*(Q?ZE8B0JSO
z-c&sm@mK43t$Y^E<Nx2g`@1(+x=+ht(Rov3?|2@@c9ER;N@0;)&E>y{e$ly3k4;7S
zD9WG8Px@1Ssei9Kztd^_kqwn1EQ&w;$*U1l=Vz_`_-!f{*&qKYjCzmlBL9u!vzJ)^
z>OUShGEZTgAKNj1YuBoude0s6U+8e;mdg*FlBMv$`=5Vg_Iookn2MX|Pv}RCdj6`~
z{4}-yZkL?4LWgWdUW6$>jQKFOQ~Sn#`i}9}ZSCb7H(27iJcSnzestzZI}OZIICPn<
z8&+L)NS4BA&)AOs8`~+qYVGI$DNOOLzc6kR&vnsGDf{xD6c))@biPJEi0!{WcGX!6
zbQ+PR@QsU(y5xp?@63m{zxBm!TTiLtu1|JZe4WlCvlPBEVfklX*tAV$yWOTgbjAm*
zlK8LQ18?8xvW7~;8~!r-(2GxBu(IDTf3nf(Th7i>cxvxI@3!2aIa#eZy>W8a?LTU$
zK>XpiUVQ%IYi2f_xKJe`x{y5R-*?o7!|vRBa`r;M9s>?)H>rwOtoGx6%RDe4gIm6R
zeq-0e-p^8)(*MGTeBAzuU8ZL#oVaj@<%VB5BTM0Ry_VX)-(R1}Qn-GX^RL_Zh%s3T
zQ~q4kAM=xTQ5+PF7wy~s-i}YKu*y?e3RCC5{#1UP*Zo6R*yqXigFnnu7|+qzPT5m`
z%pa+CO75rs&%3a--Wi&=G#1%6#ixFCaSGYFUPM2Isr_22U;IQ&@e}*wIhJat?mL>>
zU!?EIZ@ulJ^D}iW>M!!Uc<$G$k5~_B7x7)>KSg>qu3oPf@mYjL_QFRBV?C)&KlSQ6
z&M&fWiZA^|@)=j3*Na8`(L)OHnZhFaMdL+&82KuipUOYA_NU}b^%wDDe2XmyT{B_p
zIe7|qedohp@BY!`JcUKihjIMIbJ`v?;s5-*+4W+>!K>cA+(%F6;$&`NN}sVm<v)*|
zci2sDUo_-P@1ev?JkMg=ydo^hPjNiOum0xhGj(4W&yl9K`D+U4G1{rAZTd^?SJaOm
zxYzYZZMxzg7OJF>9g48tJZpY){b}D{Pi;GD#nD+29{A41TRpYh0+mrC`~K$dN4KpM
z-9r@VvDR^VP5Ci<7U9i5Sm^1_i%iZUrq-#aQdp$-)V*WuPsyb}#h?BnK2qcHoNI17
zwNH_qo4RkU=WBE473G=c#&xay)vKo>zEk#!{q^o!>-^b^EHUt=n}=jY$j|G=BKeE>
zEV66Sc+_KT7x7n=kBjne(fV4)qnt&0D;h77JMP!gZRd7TUUI&|)chiOjW<{BxwSqe
zSL`p^zp3NSxfI5@X=<C@*`rpxblhM6boffG8)2#*)?bv*J#W?O*D1Xgt*=!+{++_;
z2eF;vuh#xZuikc3`7YY0R{b^AAB)x%*|)jlMe?WgT5o^RIhBh0RDZ0W=XN{3kBIGB
z&(V74#eGuk_&lSXdcUkc#ZR<Xs_u&OVmtNRq`%hqO7U6kKW@tMKYpuCW#2E5d{JL9
zUQ+of&a2hmQ+g`uuhoAeU)rfSDT;67waRB++%J{KQu<BxH<eCmU8<jdz!X1C?T`Bw
z(TV!3w_Q|k6wxbMU&McuH`R{sF|>>B4U70Q&YvR|<(W9Y)^<wH*e_4Td=%S7dM+9-
zvUAaRksXT0i};({<MLs|TKA3XVw<0(uvWh*npZ2oabEPhRJ+IzqF-tkt&94K?IOA7
z_wlIjrnXc0s3^XQ##4T&KeexZ^HS*ij~MNoY8UA%<*)iZzeh~%SJaPgQ?bZjQ}Vhe
zfJOTk#hrS3Zeerxw?4&RQGJ}!qyAIg*?93Sb{U>Ul(*t@Nvd7sFXV_=WIueSuxNcs
zZ~9a6=#Tuyw&%r`M*KAwW4?)Pa;8xIRWI_(lzk!}u^siCY8UxkN-q5={!;!H`%~-n
z7x7m#o|3!P{>X34|FIqI5Zle=w`hIbuikc)H@1uRm-kXwsP}^3OY@6b@!<zASn;?6
zKK|10yZJASa%vaJn;I`VpN-dQAN-`SRysxb(7i;&qII>7*P93Hoqx!+&kWqVYpV*Z
zH$GE(<R>YVuTz++Yn$6&YaGOR_T`5uztW%b^Oow5{Ka<4F0tS9L<-&eG!-LX^~Ooc
zfAp)%n~L7=MT~Kf(vN<AkwWvDiufsduH=6yZ0`CLUHwJ$#ZgnSNPk6s8vQ@Ei_Tke
zrZ8oH{Y7%ec!}-QzWR&w)70^(r&K$BXB^x3Okq>`vyPrqnA*42{-XRH?G@Wa{*}_}
zyy`EKPadN0h(+;V>v&Ndr^eM2^?uhL=SR7k+K%yFtDiaFBF1|Y?Rx9JqWScg!pN8Q
zy1%|{--gF$<tdEu5ZkHv;CC?UHMVQjbDWpT8~SVY-_*R6-|A1T)8AA-v<{u7B0f`?
zvR~{^<(W9I-ZnW?$c`zD{t(+KzH99-l&9c%KDAFuo_hPmSCl)pi_W7sKekhRkuPOG
z{i*f(<9^!4=M~2MR%;u-FvV|?9q2QKDZ2XY-%^;8KgFN^A~{oZi~9Lf3hAK;Q~TgQ
zh4I{~x1HLzh~GG#;xA?Q)ckt;>y<Y(Kb2Qg^^5)@ep7T?sz0UQmhLZ-vuHf}OH=uC
zzQB|{Q-A*<_7}YeQ<v6@hYo*c`1@_%%v%ame_v$Tb$)vHAOA8wFT$dFZK+##d4G#_
z#^w=If5$_AD(~qpvUBS1oz&Xj)O)*H*LC}7<tx|u(d0aZ&9!f;4o>yocftc}@6)eU
zV{=jb6k+PzEb=G*k-}R4{z}SDDZ2Ic&n>&A{GwL5n>(*4?uy3a{xSa)<(rh>=uhcg
zf2uCgAJ3E6PWeH~pY%sLTdF<rxL+;u(b!XRNY3V>_>JdUY!}(vcuH?Y{aY=%>^U#r
zG&1-19Zq?0bjRawdaK%QecojcA3N#V98TKr;0=Cs;IrAo_Z-pn{N5w7E4vLDw_dl=
z)xSGDXyIS3x$l}&tADT2-+#G!<CO<3{OqLay!C!H{ErXzm{4uU_u%N77uTB~`LGUJ
zuf6%^9k>47V;@)d)sN0r?PfeRV$_hTTyfkyh>sIby5aVzhtJ8rcku%4ezMJ{+4FaI
zT4#?g4V8$auX(ipv{PGERzGi_<FCK9b@lfUeiw1-QWvh;W$D%x^U#Cn(JOp-<wwr=
z*ObZG7T30K*y!~STO5o2p6lEiFRJrW<Mh%}==UG-J!Wk4hZL%>QkbfXYwb_zLw|}d
z{YCe~MfG&idiyjLqrU5Hr}W-Z{qa1DZRcQ9(erBxYyG~Ysq<3$cAm8q9{E_8w>maX
zO+FvRbF#VZR9&p@Nnum(1B&8Ayr!^deab)Nv52YpwaOjm73rlYj})ypUhBSbo^~qk
zQ~sde`V_{vueDw4d6kMo{q&PUbz=&P@>9{hQEEKp&-&xODLO^{=tTKywF5iDqVw0h
z6sG*F*8ccCMQleoW4qRUQu89c*iOl<-*Z~M=sYR1xABy}>o>2d80C&_ewV^jp4A`o
za%`KI!sb5TvsX)F^w*+x>K?6LzM?<-`yi=2P((k{D{A|$8hYN2?R^ip@sssO4a<$k
zw)rn?bjDll7aN~zn-|-j<DqjZCAa?Q&+}{BJwz(b#5+8t-Bo=*=>0~P!dmO$1xNkv
zkp)H%$s;ax@_=2Q9ra|M!eKjZcHXD&+?~%YTy4^f{f}Mq!#sr(ClA{5rm=s@V&3By
zEsd%Cq`$fQrR=VsKc_J2A-0`6MbD{4a^NS0ao?hLQNAgP*OWf=r})(0T>k8n#DagH
zt7w0GHW!QZm>RElU-_x%c}U!(u*eTn{MOo^wAUBkE!8V`t@CT;FEuZf?^FFn@m2I3
zP~@M{PF?qGcV72lZ5jvN*|y)|%PiKo>_M%6*z3wA8}~kQ?S?t`E#CO%uP)eT*lM%$
zLst34Hk)2_MgF@Ui!Fc4TcfJ~4kdczSHGyA{Jw$x@W<agFM9rp_2#WZuQ_Fr1E(kd
zeJ1OQ&_431`*Zj0M^C)|xUo}*R{uR2-?!q^JpTW{y^HR9UbjKje`j&|pZ#*Dw{CnW
zqX%-<iu4$fpQ2kU#(mijqQfpJjQ$wsp@Yx3ueN#e4Lw47PNDlI^XVzlv%aXFG7m<&
z=<AR28Mi*}hhP52|MV9<f7Cm!UV#1%c|`G9gzT2W=ohg~Un%4dDLin~d;Yk`sn1p8
zDK)RCKl0Vm?V|mS^A}hzALzi;`hEVk_*xwXF4*Yr`I{HxHnvlCYHmM&tQEy`5qh48
zShQ}?Gsi8t^wLjeDZGE}U7r~9{zrKc-u!T@4L3UQtvq7XYiy^U&uZ<D&wb5pKe5NJ
z58r0gYdMVa#I||O#ajKLrRK%w-dg#O{*`KPbX2F`uGnpGmdvXK|6ZMQA%#WtP0@IY
zu6$I4sXUN6A5#4>9%5U*N@1L@oz(l2_wVq@z%jY;qWfQb$9s$z2eEx)|LsmYdG_qQ
zrSQ~2s~#}6)1r+jte3x^pR&u)H+FjVOZP_d$JHypu+s1Q49`>e&TfmpdiTH=^Ar~4
z%Z4@I|M~LIPRg5#+kSdt>-P>Alc#X$WB-2OI$O4FtQCEy#(z@iy#1!&*sZc@YaKi(
zs}+mR)8>wgx2B@+;#wN(t-t5@`SBd>x5i4_bY1H8ti)0e^RpkM>W;Gil|2`JsqdHm
zow1_tZCiRgRi}MZ`<q$^s5erW`tE;z=}$c$&_{E>fBUM}H?{s~>N@v|^9rN=np?Lu
zb$yh#*19d$FR6B<Kd;-V=ZM&!;$Q#Vj)&+EO>N`1saW(p-csXDjni7!wY0tB?<rfV
zojQM7s^2|J3hS*uykCLwJxjgqmhz_*ou=kR>xzCiR`h*c9G}<ErA79g+jXJu_P;7@
z?(gzaf4@k-_v%f>qI}leJkVVEi`Li5-`svaEcy;(ZrAPV?yoD>`aAu(HLt1l&iuY^
z{VwNUxA75UvRbj;^V59wUg{idsea#0)QWRE9{xY}-UQsztf&&Lq5u&;3Mi@=wA;B6
zWt35w?zNb`%*Y_~q^4*KC?bu3v=2(972c<XpNKOmK6xrmK-to_v_a6PPztoj)KH9q
zil7403~$HYJNAx<9qXUYf9^dubDzvS5wT*$irDd=bMA0%W{!QH$NNF{ykEQFm)_@z
zAKcrPwVyvseH}s{7skGR8aoGnUV|*^%lNim$Co}{_~+U-wdj3W#_jc7D}ML4U*nax
zeBa0R#w^dV)Q&y=_;m!?YkliezV<(T`)3cA+Of|^x0SbF?~lb}uPZh`_4_7cAFsyd
z8q2Tsbt%$8kBhO_h5Oy@KRxXB54rOGQQNV{JIb&1yhb{x?^^iS_1H8&_4EF1@y5Px
z9s4*kR=-x>zD`Vy`Qt6Jn?6p)_A~Z6vuS?p^Wm}2bGDrydwz`V@3`}qUvT@gzTzw1
zv$xcaoqtT1KOWleJ5K%ljK9Bv?AQG6wO_h_>QjfiV#mJ!Gq%sM=M(?jIP1Lf?0Y`q
zflqqNA+pRv)_y%Swny}F-{MnWH~DeFfA5CRf6--+zR82`_r^nx+4A3O<@50CBFFrE
z<L^uJ_k$3Rb!|PzY;8Tqd`{Nk^Kc#aXPo(*=Xu#~`gtMiBFl3k%XKIZar5`{_;HFX
z?*m!h2eQ0>YcKxh4}9mr-LH5lpQAhuKKoZMe!`P4FMmI3DgMdde(+uYzYl$2&sy|#
zG~+xEvi2Mt^ZD)b)W&=st3USj;g`JN)tCIt;fMCD<-XSP<9FMZUx(1+!{WRzWchiB
ztks(v^WKrZ#mAoKV|ngp?O2_?o;K#^D_>tbw~gO=v)5h!PoDRI!_Qvh#y|V~$9$;#
zKXGTB|CFn|^n-tNc(wPx>RI=`^BWHNb(<}Jyx`B@#`ewgjM*_Alt+&n#(B^1_0Rk}
z{Zqd$x;Ag@`tv#FTHgDZKJ(b}{@B{S%|6Y?5q|w-JJ#R6F7m%m$(H9}Yv)0+eV(8g
zJ-!e=_CBzEzHw|nAD;U8G+!5Fc}|pPoO#yr^|bRbeq0t^SA^Tg89!c@T7I2-`@8P;
z_%HbQ;U2PR9>zC)|7L7{o|7$KZ?=qEYxiZJhc@Q(Aj_Y(@aHx5@r~*jAA4RgZ&N$A
zPn)-Uv$5Tyjjf)I(d!c9R?o&``>_3NZ1rqBra$%dfxoXX*N%OCzG<HK!Jpq)%m0p_
z{kU(xAI|^(DqH^d7x>>@c+49=bl-ov(be`}db^i==XYLk&Hej*<FlXq9p8QK{^M@*
z*7IL-r!Uz5mM?tNzhCyEEA0Q-^FIBVk3RBq<=<WAzV`XWA796;eY|qa&r@sp@7Y_+
zbFdxDTU{ITdH8k4=50K7PS!z|&y6gqW1Qzj7U?o>pI0_!9<r8i_k&_vH#Z)8{-b>`
zZXYK$o;wfo`SU8YueEW0ofzB4Sbo#F`20)lSbzSwgDiTSq4&$iaDILvi~2E+-lrP7
z4r}Mf-v73H;Lp#H<;NScV;^_X^<#Xj-ab!l%+GJOd_630Zj8>$xc&NNZjAQL_}Isp
zP4i2iw^Qpj{d|b`ur15i5!tn$mq)F~|6UBTn_eGVKlVJHo45LF$D2NX)~dIk=W{&u
zd1iU&;}GLxA0Ni@W6#I2dd~CXVeIn&&hz}pZtLfYkuHCnnQHmGb1nLL%*l^G^ZjM+
zk6!mIKGtXK<Enjp+IZ~a=Gf!Set*Ztwx5mpeLri_>n7uNPc|NVT#WtuKCH)|Ct17c
z#}lh(WB$4ZS<AQa*nIXn@Z*f@HvN9(SUo=vkR7YHk24#O)sH={(cT#!(;xdjru{xU
zf81fq@1t5f)^}=t?0L^0@A>gyEr0*VTK@VCJ^%Q5#rBuK^n#b&;dNKO)LP!pOD?+N
z6F>gWvo5{Q*Sz?eAHL@`FTL9X&%f~>f8v5mzxk?<e)ZFS`r4P?`|xG=y49_(1MOX&
z^2v98>@C;1)LO1*oOM|L-rx6<-+93|oqy?le(Ci;_xT51aOqud@X@_{zUTZ)5B|r4
zzwAxFe636G{_1l+_N1#_<I<nJ{1d<S(_eqpOJDp4&wSo%e&kB!-wEW$<Jk4!JtK>r
zAB>|tGtOU^Ew%jb*W1Sn&&%gz%h#2yeVrML`T2z`>mqCCw6T4iw6W!**9ptBv3<VS
z7@g0~vsP?>-jqMDvUBmr0c6L1{&np7PW`#Vx$h4!pU;IX^N>Z4D~n(Jl#3tx!k7Qt
z+41%I)b$T;a&Xgwvw!5``(J!;1-d_V^;74ay5Qif=U#m3IuzgF;Ov95@AaDNU3BrO
zU;MEPzx)mSGT&|AK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;OWbxPg!S!KLr`
z!JEJ9u-j>Sc-@O``|O)M_05Ofo~%8biO=nR=DvZxfxdyhfxdyhfxdyhfxdyhfxdyE
z4Lo}PR(JW#6EE1GYR~_L|MH?o{rp$#kJ)Ga#3i4<?0dg_f2zIdh5z`4FL=|J?4NG-
z{SQ94?pJ@?dHYN4(6#HizJb1hzJb1hzJb1hzJb1hz5(9At<Jvq)GPnxZHH&N<$ZM9
zH_$iGH_$iGH_$iGH_$iGH_$iGH_$iGH_$h5=4{|b=l;ZFuket|4^OH+bFNfBY2QHK
zK;J;$K;J;$K;J;$K;OXT8@Tu>7eDsY^$%`(aI=H6f8^r(Uwm){x<7UGQ|Fz!;NYz1
zUVQ306yM<B?1QuK^_uHlbn&TQ{ILta{0+G8JKlQN-+#dm{n;L}xUy&Z2Kok0`UcSR
z-r{5DeB6J2`h~B%<RgdO4)4M4bu7N?-uM0Dz5no^_qyHg9>yNuQ}cKKs^59c%ijKx
zy|wJ$|HHY@f88_wWpA#%^t<<8`np%Yd+(&%seA9dw1Kh5$=%L=%B?^D$=4npwLPlq
zbKD<)&^giT!6(1+yo+D<i2t?6cI>*Kj~7;dD$my07@cR=@GJe&S3USA-uf$hOD)>h
zNx{b+H)Hut&-v?je)-29{6Al@x75Dv6<_qD?|t#R50_eg{!Fzude9B7`rn>*&OY1U
zfA^K2|H`*~@NlZV?PVYT=PNw^Lx;#-^A_Lri+4Qt{Y`$V{yE?BQ-Al;kG=O0*$=(&
zz4v<htKI|iyvHv)?@r(LmaqH3;h05_57uG(``3KJ)qnE(e{{%}=dku?f9E@%^P<na
z>yRziv;BjAzxlWQ;XB@Nc%zSf@eRNGZXcX(U-=skeDC%C*Wrz_z39E~yWI^xef520
z?c;&#klps{ntgqp8*lpg+>P&k*YCRIi~fD@xGZ{{jp6pXJmZB|{nV}h;13U%+86!w
z?SAXUpZwt7Qv3F=`{$eg+b1sHTWTNh;0Hb8y0^Gsf6Ur*+jwkmzwq&g{_NF1{^&hq
z$Li<i?eppy&prRke)YM3a>({sAA9AezU#H`IAm)-?$~<N-|pdbkI^1Femq~#*uC4w
z@mldwecxy7>&B(|wf10j=04x;>z}=THn#n2e(J|<TZdxXZ!EUwLGjY-Hn;BgUiY?F
zT=Is?_N+y7t&Q{J;c>UP%74D&kH75@S^M#2EdIo8?{UW;ed%8xvK`}d{+W+|=~Ey6
zxc45;wS10lx3hsa{@Z)L_2zH-KYMen{W^kUzTR`~*z?=wH~qZd)-QejS)HSf=e{m&
zi*K*bw&STEZ%#Uozdx{T%U}Oq@%2COoJ;QcfxV^HK9A7j;IlV+>VLi4JFd7tX6L>x
zF@Nd&w(h_E>~B6}|IVM-n`$5SLvQ=ZH=KWkePnshY{wp-b6-#RqXgRfQrymo`iwnJ
zpM2iCZ~m6wc>AHXV;?W<<BDTchprcYoLb8+{d(kkU-HbmzwBo}*F4YoTuZI}eA-?&
z`}ruxbN6w(XWiq;KmQMJ-($;<ho#oeJr?u#TdhTpZ*;yNyu-ok@AvZ`Ioy`D*OBK#
zk6Sb+<6J+spRva!zy2U==SK0;>wVmHW8eRunzyfGHXeID9LrCA{pZKWTGqa9OpQ^0
ze%y{(`|)?v_~B3f)W3Y@rJvl}mi@Lbd;X6<|260BZ)*ATmbEQ<eltGyI<je=UpFqg
z$(?`nE|354;s5*2e|XckKmH?!uXw}zf9jz({P5vruRZVeAAR$C4!`G--*<)o{_GDI
zd$ljU*+>5Gk5~R5Kk4iL;yVA&I}gvl!xO*!+}}9dd&e`M|G$6mh95cnwTFN3#kaiL
zpYHwdum3++|LF(3<B;t+U;GtU{Ib{l&Eaoe`w7o@!<*lI_>ybg_fMbsxc432@?UTH
z{y%=zdk?R5_C=3<#AiRb_x8s>_A1Z%>Nhv`3;ybBfA9-_{DS>o_}uyDUUs(+AF_Sp
zFTCP<??3m$hu44X`@iJIU;mGL@A}mb|HMTfyZrDY|JR>B;6I-GzQcRI=1Sjnx1W3O
z;nSb``X^m-wfB_wZ(n%Jr+wxFcYW{S!*2FZzxR({{H`+BU%vcruXW)iA3S{Y554&I
zU;X&MZ|t-0f6+5O|El*NzWzV{?t6akm6z_Z<&RhV{{B7w`bwAI<?(Oad+nRAca_gv
z?Y)N|d(^WZ^|(tuRP?U<ne(1=<@;ZD_)oXF>2q##>$e;}@8vi7-2M;z@ZSFCzV#Wu
zebJvE-t<0ye&65z;7{%G`>W{V(__E*9WUH}?tAyx^7Z}D!=C)Kk3H|QL$>^R1zY|+
zfh~HSSPLI}UFE!eeB0Ms=CS4T+WpwqeVgZ)b+%=H{rj)=eK$Y<@Aug9`FL)&jL)@8
z=jQXXMLPE9CHQeYbxwX>upRq2&imlGxR0G<?Bmv09-Wu_&b9V&y6Z7I|5BV^f7sgn
z^L4d*;p;oqvJP82uhp^F(e_90&)Yg1+de3^b(_Yv&s;rwU8cr!`yjsU$FB=x*4EkB
zJ`Omx`(oVA$2|VL+Uj%6^=z#UzrI<V`D|^T`&-L=`+2whxXjN7Wcl+XYyaf%T@SeT
z$3Jw)7X7>!dcTV2v(}!&#{7ASwf5t-&2Jj>_j%az&v_xspD!Ji9s4>QeIIiz{Ij?C
z;C)|y-g*0M?c>Q_N1Gp;e=L9T-S;o}n@3-Bf7^ELuj{Oy{W!CBJobJ1-+cOOzw_A-
zxb8l(WA!#~uk)rcf4#e@Mem>4$0>^Kb+hrUKk@c+4sLLMd0zzeW8Bs+jrsKnS@ivM
zd)=P*8$bRf=l$+k`_}Tj+p=iBO>zEt=(Vif3&-2q3+s;UVeEAhJubF|+s9Rp_4}pv
zdhvB0v-b1AP2+1l>i@ju-`?;=``fbo`LVV3<AL4BnHJm6$JY9M=6$!i$Gv{<=2zQ4
zD!b|PeCc(!Ur($Zqt6ErzO?_lzVCh?eaYRvW}oeoUUk_k-hYQ{?~hsjyyK1MyyjEq
z>|eQj{_%v5e#eh}&wu!;{eQmQkA3ME|J_$Z%ik|&%je?0zxbXX-uv3Ooxi`fwXX{(
z9{YTF)6WaJe(B?HTXmbhzMr&u`@FGV-yc<M=NOC8_oK$@P#)=>Zur>ewQJ><eq6Gj
zr>qs*@5kAA>AYigXrK1=)qelb#{7E6-{(Zuo@;6Rk;}gMzF+;4f7+XC?YV3`bsj#S
ztsnb*eQF-f!Fxupd;D>FExYOK3(t$}T60d-v5#9DFa18PeH|G4d9ATLe;;8j%b(A5
zd%S-?80~-V^8<aJU|)|=Y#;YFUi<q%R^L8fP>fy&$8hwzh4#VkJ6p@w1<kh>Zu@dP
z^>uA+p8dKD?c4Hf%s(GwExHc&ae`w0_`9h^=R)(XjoZg5Ixpks=N@nTZJ&S3Gk@jp
z_pIf|Z?|Xm2JFWR{``?~`|nEe^UC5=WAwb*6@2QWVeIQC^m@SmE)23`uh(`@w$Ic&
zUq`m~<CWFrf7get{do?1{cZlZV>=(mJQv5-+UEhsY<VB(^EZU^-q}ul+;!fv4U9cs
zH_fB(moU!PneE(D*vI8s@ne4dlOMjx*S)9wJ6Zg5;cWTq%<J6dnQwjhjeqOVTKl^3
zWpBOg1^4{yn-8tE|4tTvznOoojh{~(PqkcUE$g%8y(7z?pYrRk)w41B`A)`JXUy_^
z*3Nx=tj)jnOMm!R-}bs|?04Juu*;8wrH_02x^>dxwclUudfR#5`ndPH^!YKjj{lzC
z+IH?7h|gcop!~M*wSGS5xO8sw(I?#U@{e74zuUftU3eU9>v_t$==-QA1z+pum&bH?
zU)!?i`zCAQWAC$WnzuivxmL`7uYN6Szs?@}cl(xpJ^RGxzVYR+z3b(p|4$CTpJBh=
zwlRObY;ish<NW=(UqAb%umAE(>c0ohk6Y%kwdc40PB-iD_YFDEG3V{yd$TduvE{#q
zJhorX^Y<C-ykjw+b6b}8x+~WH+y?*q%WU~^#Fqa(3AXn8^jx>L<$c+Hc7DF@Y{#x6
zKVI1KIoGnAe%;J-O|`tIU9nU5Xy2FNc<Q+=4}Bk#Uq6vW^Rk}pGZyprP1(+UJdWvG
zJv+ax=jR2o{B<B(tH<%w#}_~T`FbL2=j8WUt+nT!8e4rEvyR1=#{76;%lcET-7jAc
z)|+d`&cWA}t<|^J&Cb{3zJb1hzJaj~*w5ck{JY=u*2f+E(JviNwf5gH8=Ir&PsRr9
z$1VPRf-V2Nt>tmdUpHDi_2V$}?B^jz74z5YY<Zs3)!OT1<D+`iqR)>RxBngve}0eb
zC!X@(uXWA0Tz<&*?%)5A_gv!vZ!P~1+EJZrtv>wxW6RICO)cx8Ij8<Tprt&1J(+6l
z*O6#$#?jxq;*WFa-$R;;^Zu4v{`?<V`+QtG-t>J0`*q6H*nYi(V!n>u_6_t6)D57|
z<2$bZzte5)W9)s6o?rI=Zk;K6v+tvvwD_pL9$VU{zEXB?mo~80&mA0>4*Ght<G=Hm
zm%aTXd%N^lI|;K~@|o}W>?^+My8GL<Ct(+T#`QMvv@2fisTaTV-2Ibkf8u%v*ZBRr
zUuXZM+8aIMMSu4%zkQAUlWI5pdBfwbf8nRE^fiyX-rpQ{dpvts@0F}8$qoGFcRb<K
zfBCED@1InAo%{Udxp#c}1^Z{Z9s7Cqwes?@)BWAuz%T#hP5=6zZvEN4ZjWaVC-w8p
zqndl|XAXK_cGvN>_ki^5&pYgTjLy;V(`y4i^Iv}W!oPj=C-+XOefAGO@Z1Yu`T4z*
zYTt0xTfP3fpL(tRlWLFa?~V7qNAv%UP2IEod9kTKUx|KBh;j6DR8#S(-?y?qhsLoz
z2mc%?vSV{y^WQ)9=YRO_Hyuv3Jja;jzu&Ud^3Qo4*GD(L-ed2Z^Z5{u|NUBI?fZIs
zZnpgY9Uryjf5(mIU~B&#C_49CeC~C&kL%n1^s~R@>9_vdL$>yPeV(5!|J*Ly=iTXt
zZvQVo^u9y3_VZ!%b3WY1&Tp^l*w4}O<H%b3?*(%_wqJhyPPO*m%N?8d9*_8n+uZUC
z-?_IdcIp2QXzT2CMDb0o^`Hk`_lAG9hb&*mu|4w7_44D0E&uz0{P*nH^5385_^5xL
z$-WNRk4HAP&r=)oe#iF2dEVEUwLizqeeCB2Q{zqly-co0KZnlevzF&xYWZ<E)sFpt
zclP;Wb*H}GY>UVL&lj?G-nC=<dC^>b{`fRzZJ)Jbo|m7;9FINUmgc8E|5)EXKR8Cu
ztEu>=|IW^)KQA_Q4|92AuN&y=2F5r2cX#jpJAd(t7d`v$50T|QY{#zG*z<z(n_iEx
zy`7Z&g+KaRkNn2(y6XN^`$zxt;a7UjyZ?1>sh#`yT${giZvOwh`E}3QlXAVs=Cq%G
z*_gjSVZZ)aJGPI<zx%l>f7Qb-`NI9F7JYnY+<yIq;wS&^qdxeIFZjZJYsbES%b(A6
ztN;Jzsrwjvoc4U*z)9MG{r`)W#`fb5$4m9LRcF8N!2j+bf8XNAFMGlNxWeV{EdS2}
z=h@niJN)s5Ki^zyZq}P>?OZ74>%x{l??iTLf5hXjf7l`(ns+MB=Rp?fpn2xvbH6Wu
z_KkGr&ck0{j#;bA{Vl$&7(FiR^J>$%(Bon%K6ZZIGoObaZ)0^R&w6WF{`W$xMg6yh
zk9}-h>-BR}9X>yL-0}Mv%(Is3*p8js<~N;x?Rq;O$M*ZG9FIM2`TKKIE%%#h`TYpC
zW8a_QJU@StwO=Rj_W|tnAB*|>p3LK~D>+6M&BZwSdXe#|`$N26f5=0x@;%S|)ZSEk
zmp5Ga>8G80mHp$goBrJ7b*}v<*Z!jGeSGh@>@D8>q+j`l`~U0SQv1g9Ui_L*eCI#x
z?W+9`Z~w$yUVZ5`_L2SbA9&H9Jm=r9y1%yN|NjI%&vp%8YhLuc{@$Ot{JD2}^~?6Q
zWzV_fmtOT}UwPL4wk-O5dM*4}&wSIp9{Pm8INY{H*XOu#{x~;gKlh~<U;h11{Pdo+
zV|D!fZ~lB~%<|9qAUoD?+j;(f1o-<6{63Mj{Cx+uHqSpd!Pe&a=P4fa{9k#-&3^vF
zht^(nlRN+DT^|48L*{YJmglsV=Vr_Ok!5|J5AiI{KcDn7U-{rGJn!e;e~9eGPr3N9
zFMR7cXUFHksp}uy?BMGU&i;{$?|<>Z73lud)lZ#w>Vku_o_q1B>ri}ygR>9LzSnE6
zchSYC9`s`uzUF#ooppA<E#1M=x}_Um*bkMtxC`>YGAVt|OWWidAhHc=LMb@p<1&~5
z2>=f@KeC*i&BkJofs|ln7nA*rkDh%t!b6*ZR67qL6AJ(hz%yE51wZ4X>KKZm0x_41
zOHlfvZt})d%}N2B$<@bfoN<LbaGu0Gw%2dBF?2qOOCOwO0);CFsRK!qG9jyP+v2u3
zv-wQ3wLt-ZHy?0g1)hv;BM1`>a}Zcb2KGSu2#-Q2iE!9`^~ACXmQ@X`Vgtd|S285H
z%{i-f^@e8AIj3eCq&`8>n%j?@5x3v8=`b)wNfoE5-X?H7UDb)Ej^OirR5oZCEt2@V
zhRajzg0J_ZBRt<!9#)wBg~$AsPK5)d5m90gN6Dd6&&qAR0iH&+R)MHNxS(ivZoyo>
z?xrGN%12e<Dgo#NVEEw`@B!>kKAAk3II+*L-f-4$F@mlu3Dsa?WV}t&$JKY!xp%OB
zbd`o+LVCVhD4f!PT+C<ASd-UiLOHz&6yE!}`N1vvdp{G_B`FIx4lch`-U%Na&&6HP
zJH{BX$u~e$HK+-_#vX>tU<M=r>@_`d%+6+GG0;Ftu(FHEe#S=!U-(c-JIv@zss!i)
zSe2H*CA(lO@iRWEULq7l1!WYlLrIptr~{RXv9$os<aWetoN<LbaGu0Gw%2dBF?2qO
zOCOwO0);CFsRK!qG9jyP+v2u3v-wQ3wLt-ZHy?0g1)hv;BM1`>a}Zcb2KGSu2#-Q2
ziE!9`^~CW{Fu0)A&%2@Ej;zZ$g)`YSNPU9h2yYc~M%-#YUwmMSk}8fbu_g3)x~juW
z9l_`MsA|vzsAkvjX7FPM^nP@NN1MvS3dj0|$Iw_aa%*g1u-(=h;O?rm3PcUU1x35F
z3v~G=k>A=!H|ew0S}-nv{ZS^9Cle?38P*%l`YlG#btR!1OpJ`TY5KVOZaVi4)^EDX
z>GeL-M<*|8VR?S<=N1RI?C<?d-gP1_-1xD6-8AEf=y)#ff?l&RVv}!xsA^CXdW}5{
zm%$830N87K<d~h!#$up>lwf5Sll_d34t^1UO4?yYXHq3V7r?5t1TNVHV~L;fQS}m`
zC@LtUfE`M*^hF)0RE(_!a3;4SX5)-2<bm@f=CQqgyN#jqNnHBiG!rOXIY=Eynv@Ay
zecKkd#hJ}#nyn2A0KEBt8!PZ+Y#TwCXqbb*N;0qq(noj{LP><f?yDz`e}cgUt$yAO
z1$Sg!&MBP9ra|fx6i0ZgkTc>|`}yJnQ<PM3e2Fcg$J13EX6gt&&qq~*CO|d2hBt#B
zGobgQBRtwv9#%NkFFc0Envq*$3xn;p-T-%3tyLgu5H2X%on4^IH;Me#KDtRCzZUcX
zF#PZe_yC?YN=u%lI-z`6Z#dh^Pe9g{glaG`VZ==+$JKX}nRm2y(^XEd_nAIAd07j~
z^LszHJh)YV?`QI^6LI0j;S2lB_S>eT<GHvCdJ~Qjn|uRARfC$)YwTgT3}!$Az+TfM
z$Lwr276T2W1S`9k>}Py*@QVOc(hf5^lPUqa09K_XaLFzhOZ<$Fs+R~wQ9&66>`;=W
zFX}+0Vr(sdGr1ix8)sZ051c14kL~r_Z48}H;?f7FnLy#nLFz!#q)f=_+qSqZ&TKx@
zY;8~g;LQizSb-;F+X%u$!yE)wl7T&tKEk6AN+KL~Up;X|6HuGf%5UhStDkmb!L!sU
znyE8T@ZxOTcvQe?H4Rdqpp3Rznc$4La)&8$15=b#ai$2a#yOs@YW%4q_&gs~4VnPe
z>>Az-e$0U0kB;zYQ+Zf{U)qNlyPjKP3xj9edIQ{DwN`<sLAao3cXoj;-z4%|`{*Wp
z{94cl!0^K>-~)KpC@p!G>V)!Pz2R&tKLJ@+5~{((gb_EP99Q2>X5P`-O;<U+-e>yg
z<Yg@^&+q-*>fqK3-}{k$r9XM_PTbKAz^mydT(F)@q=j4a(o+N#NY<|UCvM=~Tx4Hk
zxh{)WvAD3bJ`v8*9<z3<o?a8c0gS&21_c1tGwgEI9)OS*-SiB*^uH3$JKa3?rhD4a
zdbtGYgK;NNK;>x!hlf1<&cEVwD(oe#1}%u+o0z!<BsWuLF@0|TmY<LoUp&&$7A_xr
zdF9cbaOb+69*aA=Qyg8VoqMR9l!5vsTH#8ap-UV%_CRX$eIxM(Z-C8vcZI%zzJZ<C
z!1QM~Guhhl)0aRfYuU5oDomdKdMieWP-(L}h=QV09FbIRe(&ej2e+B~-cKV<Km)o^
zoWP+D;+}Rr%51e-y3})9CeTfBpTQ)PpvBcCQ=BAX!lFEa$F#$7Apas4VsS!i>6b~!
z$%!W40I|xTCQQOEm{R#3GhhZJeAq{MgyHPMGdQ9avN6vStdnw}Trf9-_z0Kt08Rrp
zfH<6R0ZqWdjdeD5cLhKoKyWMDkQGG*k(%yYT!PXUb!Y-eSG6MW3w)HzhO^9Ki`ibr
z!&u3&V|1O*kp{D>3!*hsjbGrShz(`(P`Hn)IHft_Oz0>m>}EAz)JJ2fNtuw<gM}GC
z;iE_fu?1gKbR@={;{p)?;~n4vG5>_YhICLltceNEvTamA0>JP=fcQ<ng#glb@CsE=
zcXe?W6z5fW6-aSE(Vz)v;H)$OmlWyOSv0C6W--cE9Cu`tmKh6*P%77<j)>7BULE93
zEpJ=pN_uO%ic~L6F?7$|`sO$zLa+#wTtQ^t0(!KP9Cyaci6Q@F%?vg54Iuc=IFc^?
z+_b4s92L|ZYPgZu5-`+tC|AdESq3y=EhdkXY;)Fxx~4}PZ(zZGb%~DfJd!8<a^XX%
z#HG)w9!5EU0~qqT&5>9$oCya;K+J?YSqB=B-*O5LBNND1vq_Rj02n<+LZz+<5d&<6
zx~6MFiZoVBB`*ESs9wwi5`fFlRV8_vU!vkzKF5+Eea^cU5}ibZG+rWh*t*T&k$pC!
zNE6{)x4U-w<c4X&bi(Q20DoJ{%EP%cIz5Tka2tRNWg`0eEf=OGE6F;OBSSDDk%l0V
zOK!NQaf*qK+jtU~L>nrA%c%LipW7UK!`%0NnpS5TfL1%UhoyC6x+~|=m@e*u-auo-
zCf@*2)u1Ny8haQngBg$ju-Ej+F*}=$#Xtio!OAWs`xzf~Up2WID&c%%Cl&x2fSWE6
zaLF!+>E&hEw2(5=M|>ZbmC{$uU_!CB63lCj**G(K?0FLN*pr9Y86rBL#H9}gpFrWt
zLFz!#q)f=_+qSqZ&TKx@Y;8~g;LQizSb-;F+X%u$!yE)wl7T&tKEk6AN+KL~Up;aB
z6AUhB_495hxFhRwPT@>84e~$Va5`)Qaz@-nn-R@{DN3q1BU<aGkEg3T!_*Oco{y>q
zO@L~a;9iY~(GYX-X7FPM^!z%)qfO;u1-!YTn%&B+v4z1CZM^~Ru3D==)F50?v^%>%
zmv0jJt$lQpK7K9e1K_!@DoDniWfs|0*b2CXvDTG@YA~^tJyp@;>RTr`ZM;oaIlbOz
z`sn2KE-cUQ{d~j0H}>~_ChvMP;=(JnS<_zKH_$iGH*h*`pueD=j#ssxsc)cfU>7&g
zUvGEusw|kvxsVRDK)9D2;|9*iONb7=1ovt@jIQxq{1_h>y>;I}-$38M8M6WO+Pv}t
zj;KvD^>OuW2I;``4fq$$N2jYUW9IjMzVYC;{k@-M8`M?lD#~jqea-Dr-$36$-$36$
z-$36$-$38MF>S#46wQH-X}I3DZ=i3WZ{T#?fd4A%qr>M%HFI&)Wv&k5BmLvTPs*If
zwV&R%Z(w)>C*c{iXvRfF!@KI)zJb1hUD-f>@8`A$x10OkPxI11Km)o^JiJ34#69i0
z#o21LbgAdIOrV?MK7&anL5r(PrZ`E&ghhD-k7<YFK>kH8#Nve3(l3*clM_w80b-Rw
zO_+pTFs1T6X21+c_^^-i2*cThXK+L<WMiHuSSRH`xnOPv@ewZP0h|VI0C70s0-At@
z8|!TB?h1fFfZ$fPAuEasA~oH)xCEsy>d*v`u4+Z#7x*Zb4QH9f7PGyKhq01l$LKnr
zBMoL%7es5O8o$6t5gW?np>Q8paY}Q<nb1*C*v)FbsE@`{lQJQz2MaTP!bgz|Vhg^e
z=tzt?#|0t)#yh|TV*Uw(4e6kASQ8VRW!tEL1c2d#0P&lC3jw6>;1#N#?&{($D9)?$
zDv;uSqCpeTz*%VmE-BKlvuIRD%wm+SIPS<OEi)Drp;WFz9TB5NygJC6THdze%3#@*
zUYcU)o~!!*_fxmTJ89P~mR#u?l*QswXjl|))g9_Ol&eFzECZUb7L&(GwmECU%JZu=
zDx*fIbd)UeTV`S$7jEZUHAaUTXrfDWgv%sP`f}Y8mp-d{807#CV3(h`PK5)daZq9q
zugU5~ITNN9?mc2A+{rqSD_p=S1rA3+Dk#eC>Qh!0AYbS}C0Q%^2)9Ck$5ggbre*`G
z2`SQ8EtR<RE2DZb4@dwuvly)smg?e^lyM|M`i7=MNOTer(s+s3VXM1vG@NCYwWtc=
zvR-jiAX&wS3`&$_aZ24xXS#r4nlQDU4i4~Vo?tRP59iM4h!C&gHUJmOMD+Xr_hVN>
z@BO^+t>>J57Tr!=KfDEW`~D^nFKMEc53^E|KsUvl!6cKQ#nmNKoFrnxqCA4fw8L>A
z=gAqm6k>5CSjNasCP{A0G=PV?Bo|a|F#~2m!iRm7M;Oi~);Py7Oh_sgvN6x&P<<#D
zU>!f>qwRd&JfOCL8$gH&7tjPO+*oI0Fm_Ahu71Wx)g@CD6|i)$E-pdoi#ky0>!~^u
z`1(G|rNdcfvBhjJ<6*4i*fF}!=SYKD)dkU-sm3qxQN)Hac_`e+Rh-fsaVB&W6n3+k
zFY2SQ)TB(v>cPT{pYTy6gV=(vDLN8k&T)YVfbkA+ftY{7U_&~n9M;4HXW2F?AOT?b
zAVB=4-$DTCJ9vevr@Oki3ySlqyb7eapJ>noG;mg$fJ=(>>ns}85wjR&D~>xdO3RD|
zMJSc)P)EdQ5w8yNrk1y@xH4FFrI)4{y639?CQ#iH@1$L~SaPLnP!@|zp<z*cZ0@v)
zrm-d*!U&mAY)WZ0G?k^GWwb~t+BLj6%9;Q=>_87)q9a@;dD52`M&i<ERS%;azya*?
z6W6J5pfnCj4B|Cey(nkG)WW?-%!E5x2XciAIHkbhC`biG*<F3g$^zsI9jGL0B_H8d
zDDardR?5_DKs6yn8mpxemwshbFXjOWz-AVsRl-tToRTt*BuL-TbO?z~B0?H35j$*k
z7mkLr%(50$AzaogjtV5J_>e(~k}OWCo9Rp!FiaDsw$s4@{>&3hrsv_@866SgHQWZ^
zLYatue-p^ACjSMX+aKIv?t4G1)NS8D-$36$-$36$-$36$-$36$-$36$-$36$-$36$
z-$36$-$36$-$36$-$36$-@vYHAiwu>hl4xz_kOm2P2`)A+wayp^bPb4^bPb4^bM@O
zf&QUv^_6x(vw{8?GYdJRLc03t5pdEU2l5onSu#-Y;%x6?a|^cH>vZ}mr`_cZ>P42Q
zIOUOi`tGBjtZ$%ipl_hrKz|WzR&quF<+>N=OgQ!#b=WiEf}X}_%<ui&@nEmN_j4Lw
z=6<xkfxdyhfxdyhfxdyhfxdyhfyEm*qOWgjo9Zo|`B)3n{PJFmlMh$USRBDw+2s~&
zSC&5`<kIf)2K6FKRGhF8>7Nk`>WAzb=o{!8=o{!8=o{!8=o{Fz4dnNJ_73*vzW397
zGD|=Mx=>t#Lmk9D?Yhm`YPEE!=eA6so8mr$NhU#yt4pRhNyLOjc?6GXhvPv0MJ~kR
zgx1n8laP}WO}+tQl|fCIgk3PD@;zq23`qE}kMan^*@b6tL@i`vo+nr*<v_V$ZU*rY
zF6RN925taxIN<`CfQ1|DZ0zm|fIxuYR<<E4iV7k%-MP2~r7!Bx1dy(3Mc^0sD3=Xq
znZ*{fy^M#kl4Hl{I-esAW>ptNYo;2%z()}q%H*MNA6IcobHtg@QBc^;YQCtC#!{0q
zA*%-qGk(HHkqlxBzNY9%j5)^zA^^razy)Id34;ykpmJCf6P#t+sDK23;e!D2n|=!c
zr0?Jrs-EuZ;w~u8tMV$4;(nq*6VSj}X#y@O(yy~<R7c#0QMTf^OsRtUp`1$OJu!<=
zD$}8k9C|8hWvndoTEUd@aE);kEW6U<2H0}dY4QL4<YV%g%BdRw5fd)y(hpe^wtT}V
z{Nm)+M`MP%4&~}}F3W%>ti|MUl5NhKu=4zBjmoGIDjg+@{Fa#*$A#PZR*li22Ab#+
z9pN&`lfGQH#HG)w9!5EU1K8y!u2bPaX&jUo#A~v8QO<;^g?o>f33sv%<O&yXN`b>s
zkP3>jyZV%s1;`gVP)XKGKEka~;4zi0l&RT(YC?)MR!b!={mQ6b%mWgD%`8T%gr&MT
zC1o5*kiMbm5E7k4gfw0vcG&7J91UlgWi6^gxU5$k6-d@he`B5L0)}bAH1~9HfIo8u
zlj(UlcSc8qcn!A!xKJjd-~YcKyPEvo&;G$T_4j^wH^$0GkxaRG0^JnX4knodEv_z^
z;v^9h7UdB<rX7w0InT+^r4WlF!7@g6GD&h{rU5+ECApw#iy1Hj5<cvsJi>4`vBo)u
zVM0=|kd1jBhw4MQ0PFY}A8qIJ<^i=0+yFvMxPT^L;l?@}gRxs0cl9$qsxFzLsDPz|
zb#VzwU(|t0Ur*JUz}NRtE*;J?i!Ek*84qJ6$Bxl;K1UkNsxFAuOf`Oik0Lgd$wT2j
zuHuyDh%=$1ps<_Od{G~br6y%URu2|t{DhAp8N?QRP0^7UbB+r{0E~Bl3&i{r1{=~r
z<*+6uILo$C0SN%Z2La+Y{T2d9-@z+XJ>AvCT~M4?<y9cX{X~N%pn<c}1YA<2UuV&%
zj+n(LTXEcxQCem!C_<@RhdLrgi+FXAH?_QN#g)ObE4?(u&^=e37VrJgVbmt_$@!3W
z5dfF4^pm<3G5gj9B`Qu^zC~oH>rk$a<gyHC!dgroC)wt#2`kU9)~Jjcq0&*Z$Zwg6
zaa_2aZ`BwbYM_ZO(Gf0_Jn74IOI-S_>S2@vIDlP#;yM)$l*U1cLA)lb7v)TtTDbR!
znQ$lTK(25BrxZ9G1*xDYyQ@!GS%7?@1C?a0<Rjb)1s+q`N|~Ass3xRHW3^P`(yxr_
z#XKMZ*vw+IN?59kQ&Pr}1nC=^4k6J=L`dT$Vu!8n!qIS+S=OQ|gv)xxQGsL?A2KKr
z{-iu)6`c`hx`1JtFtwQu4)8}NUYyRuxidPV#A~<>z=bjq{r=vMT}^)P=bH{LocrDn
zD|Op9&^ORG&^ORG&^ORG&^ORG&^ORG&^ORG&^ORG&^ORG&^ORGa7J%Hekb~jUPM29
z-$38MZfqdG_jBRFo#wswa~9pg`*7hd`~tVJM0&e47HewQO%kVEO*tzAoh{57apEi*
z-$cxB3GzFN`8ArF7pm0~1Bn!O5KtyMQ@4;TGkzwgnaa?3L_D|=9tgeD4A4hYQ}xDp
z&Dq+5u&%Hhd6yO_KrwF0<@@wlBgt4ED->ZcNJXK#ohdf7F!i7sSr8kD@z@PG4OJy#
zWdx&0O>jlDBp2(*8V(@_s<k*%IotBoP%20rJ;=c#IpX1N;5@Z7lN6UXNXw^%HEYaU
zP(@%fRLdIKPe?-Zm?^_otxo5QI(Y?=T9&619Q!Q}6PKAc5Dfh;e>mo`MK127*3wz%
zbly$EM9VkwVwn47vDAas6uk61iBk~&WJ9?Hw?#z@`lyn80v?~j;)`RIBN8^@c<L-E
z2oiaBERZ^JmuypmffiX>bx=nNK^#&tl3ql@oy?R{I${iXP~uf)^>Rf66s1_jD#hv}
zsfZi1WQgOIX;Fkxo}h>(6EW&oYf=egxsvMQFT&ef?2zu#ahV=$X;d7>2%uk)#C{Aa
zmZk}mKqS<?0l0EK0x73vah^)no>!_F6h*P4x?7`RdjvX#MrG)oO|#dJV$W8V#*(A8
zjD*l5YS}H-h4x%t90oSDRXGcpS*f2qy-DNlp^)kZ2yW6V84z*gW7nX~N?hHxqfF#Q
zHc8Rx(k~WH)mbZ&qcDpzZJrIHJXI>pU1^T=OL#?EmlSxmGkf;9xll#7+1)}xETV()
zBPmN|BPFRpTtT>_Zt-{WlN)RJ5@gEIP(!)WICWGlxsGi-L<}OigdC+}j*4>Q&?y>l
z7J$kTK1+OMayV_hsi4Bi3q*M}a~<z9Jrv^ZXwmT@QuzcB!?{JagG0zw32Y<J^O4QN
zLCKUb0we5sX*9H>Lc5!J3k9==59IheUX{F^{Cpw2?2E2hjq_X)oHHVEBY(Ikwu<6q
zqBJC71PzrB<z_B(x9Hx%1~yUXW?i(tbaRha%J@<-3s%VrCfj0xvkS`T0RmH8t8J!(
zs(G+d2BnHa0k1~dmee2>a2o}+%9!U37Y=8sQ#lB3E3Q&z$16wAPx=*a>8VOC*6}m5
z3yRFbN}5rYW<Sne=NAm($80d<rHDl6SGX0YjCQO}niY;fCzyLf6?q$J4QV%N8#F__
zP3#o7`MsYz9o%`|dq25RH+=(r1APO11APO11APO11APO11APO11APN$?gsMrD$d*m
z^pp1u^bPb4^bPb4^bPb4^bPFd2J(A9cRskwviE-IeN6f-Kiwri!t@3mCyNd5OT~BZ
z(mPwFm_Y*6jS}HzVa3^6klsX_Bu`YISn)upflpom1!Ev(Ku?z`>^<jc_8<{W^)rxj
zgcGT~Ra6ryo+UylVFizYinMr1Y54deJP^+4X{YjJ)p*S~L+Tx)p9Y;d?kIhfF!{vm
zDNSmLUsooT?-=Qbm$&}u<I7@sjiTekBZWJ*<VZgTgcY2!Imj-D(sG@(MQ8oP<-rGq
zY2uA;af9*0O7Y?c|9H-glx^|Gqgufuis=fsHy3PaM~-Z3E`Ct7Hat?|Lx(Lf)_?RI
z`67G6z5%%b_yt8G(I*D+Mju+X^(2(@?4X|xPWN#W=~W@V6wpEv#wOV4l1(ocm6Kjj
z;(g7<oiaBbJ|fXOad&w?Y>sArwXYgf506ne#R**LQ&{5SB*7ArQ-YM!KulVOjdO{B
zg{$>)56&mc5f4yGazf`z<Yh~;Xs{zP+@}$!oZ<zi_*5r?><Wa16|I8BqY}5wha=VJ
zd#Hz%*3>x>!oAOwsPv^q>oYUFNW^4~RUh{;NnGY<<q#S~sCt=PSIdLf@EV|2rPKg{
zsFcnWX0lC6B_f9dYegzmRVv3+>y*i~@|jCudHv3ek}FOrJ@HQd5EjlXZzuj(8yz5Q
zB0P5JgLVFswo@KY)5Oz0>A_StOA_%xM0XXQ<W1{(^ej4GeXh;xA9vA|_-I4h9l*PO
z>GIN}DE*3@e9gn;d`8;clpAlE=%roUJOK$5jffr_^F3MWm1}eVSjg>bKpwm$mD&pu
z<KO#v;aks1Jx^UfyajaExo-kBlhh$q--eg7bMd?vrCo#BYPA%4MKd{cAH9szMWkE7
z$*WMzd3TB8Wgr%X<AWwNh<)i&n_EGOlL#g(RXl$IvXl9#O~5EFSaDTVnkJFry{OR@
zSx8mKASar<1yWqmxQ}obZ+X^`O}|qcBG&@Z+6Cn_f}34XB4Bc3NJCLoy~`TeL|j8n
z26MQuFp^V3&X=Rd>lx5G6mj`ao-&p*R)R{4P?qL2(~|@x1HvGa0wCO!&q%?eZW%KQ
z#avR1S#!tf0ylupP&)rf%*+=?xTQ8%QEB>3nM@Q;22&!PgM5y;HpbzObJAwvNFhJm
z)b&!_$VzwYw)rGdr!{d@=W3pnEGtzfRu<SO7K!NOCRrv^RfZT1z-6wLZ0Tnm4Gn8Y
zt}0Lgw<j=DgDFuNJm#3p)&)xmvxM~ve6;Aa=@t<^$VCK$Q)r+f6!Cn=yn;n1Q|rS=
z(PB0r>38lE8L=RK7DOCVL?IX|rDDEvba{yAcQQ;$1LlbNHCZOe4l$BjJXy=yF`0>}
zp!CU3WhkMjO^V@`DJjT6z#7;=Voee*71<#a{3)xUyitu%l*M;aRyUDi%t{#R2Fr;E
zF-~Y$S<XeP2H=DOLy&N~kGJS#ZUBl`a6#jOm`gZAqEaNHoQw;-NMt7%@?w;s!X>9-
z>7<F6(BunFB~xAcO`A%@{6z$XX@C5AQ6yxjCP^Uu%5x-@I$hQeq)dqPR2D%dl(YmG
z6qbaN7=gjhK{9d5+#=;8r@GZEtBRF$#}cI^sS!@UD@7qGm6GJEi3pB2UU2FX;Ybe2
z#G(7*C5N;M8dc<7=13^|6)!3z!*F%!K=hls&`T#X8pT?iQ%o*4o;lV#A`ayB@@UQ|
z2pTUZhWx{aNpMr&0D_-gX(V0xNxiL&oN&ihXfZ0b6j_3T-J(a1D63)hNTo!LLx4&L
zfU$<0xiPPbC=qWd#Zgi!k#U%tUwrsQsaQT)cmS4>7$ZydNrMR~<_9^|F-5qwre<S$
zgj2<`LNcmK=r~!Skh6lHr;VJg-lc&@TLfBW(mrL8+^b3$O;savA|YH?O6GyY1h^c<
za_iAVqvDcNg6c;I;nab1acyZ^1cfm5v-i1#%Tj%e5>yf-E^eG{@(m!0f-4_ttU-)Y
z7M$qQodk(Qw}ho*6o83rW1^5b5>^tCFX-G>K@`YZnQU=J$g&?j4%kW=L8TLy!v`-#
zR5alaQAAg#MHY!fk&azsX~<+w(u&LsGNDjnGDT4=$3j@Zlw_goS7NM8Q$#lpk>HBd
z0`d~BuvAc?5Em9ZY%5E8!&%9!%H?29aRAcqP?(d+<bd>qYXa4c14U<$EG0^^IFd2o
z=iQS@rCA`FVX>MO<Z>-jHgi;<2iCJ?z6Q4yso}<(5M9m30Na!e5d&oD7GD*4!3om$
zOw_1|48ep%WQZ+6>}N|RPLW<+vP&jpG?yb%9=`!LfJ>|V7l7`1aJN<7`*G$Ro4}SJ
zei_*o<k+uMXbFk~v=ofUE}lS{l77@}fYhzP#ZgmmFr4)opo$SZrkTSw=nI!;0wAYl
zh4eWeZB8U3Si%*2gPKqfr?`H#M5|vtrR>8#%7ug3kfeO%Bx1Vf;$jV4!{QYu*Vwp7
z!OQ4Z_$aCQ86QRLI8`Yrju<w%IipR(G60YRuqn=AC07reoGJuK_!1RuBrJVVDcDB6
zWMrlE3xOOz;iIZh2Y!K%=F0M+Q7fTnP^&4L#|B%ds#-2b8q8!DL~EuRzraTk8_MLN
za35E3N^``S&{0s>&1x*lHl=mMCDEFc2?Z5q{DhCX%L+{pTVw)7u!>upG^z>3{MnTR
z=?f-+0|R6L93_WxqUUTb*Fyc6;H<S=z2rotHaW6SYSZ4Ek`VGhZN^0q9pO;~B@v#A
z!q*Y$i-~wo-0&xrl)~bqVdIw`11$}`kxMG&zM<y~z<pSwDbiT;l~oy_I;uhjRj?kO
z&sEEnC(9GHis(|!)Hxw(X`J%#ijkM2^c7^j-syvB<*{0_1uss-bu<Iw5o8xfOs2<t
zRNmqSF=6R56G!x*S5ws!pkh$_!YU4-M|O){Q|mlHokQdX=Ba!^5+J!L6kw%RuzE3T
zLKdl&$s4CNjJk6p)-pklT(=!QsV3#DUw|CPtYk1D+f0kb`Z+Ew@=2zdT1<Q$VTGwY
ztblvX)pqy^f(d6ThZ|X=xZ4U<$;PrU`njmVWf9z3F4o+^6G$qqPArs1ID8z<EkJyP
z!~N7mD&>2&Id64s%uAEF^f@nW%mWgD=j~66gh&H4i$tU^=WKjMOhLKl9NSpTL1Af0
zPHU8-;ViSPMO6s5^(7uC7)Zr(l~+MgpW#dwFiaDsmeauj{*38oQQhIN0@0h22H<)(
z;ewtltFSeNCS;be@@orDfl2^#)reKqHK?AhXFR|6bGL)L&wKAD{RmI^!sCc<`V`^8
zu=_VVvKtQ^W8jef@qKilgO^%CWP!w3NwbT#z+G<u!y!>>Roq+V0A>zAqj+pW{Y*NS
zS0gc4D@nWItMjX1{(PmpI5RcU^*EO8rAOt=c@e!ntdi_;-J|&?;Y%;#BymH^PdM6J
z+7%na+u1NqO|Kc^Qd^&5{*zC+*?DMdi*FkhFYdZ_p?3-oo6|R4%Q+Pbx#{wWN6Z9w
zVJoLhtqDJpa7JrRn0*~eZo8qF9)3dP^qWw%Atf|z6$o<RmbfWWn^bR8O;GPNhzkdF
z2$PPZmH08?Z8W|rL(R<+cp7&S+04EAk}%tc!^~DXf&v;gJ?l1X=In9vMGRiOrSePb
zHeAr@RxZ~#c%4qSgP&&Sk=JpPqfOAE7n)64%&J}GRc%(sog%x7cEQ21(-z@{!-BbR
zP<%!bH<}?LsjCZ0@j<_?w0cd!?D2wDBeF3tRdUY7=0lfHAsI+EeKs<0@~LCaQ$3T}
zbS;L?Kb&<d2ld>F0LI4zN;Y`>$KS%g`F8h%d(3<9Cq49<`}~=)#2qNjC`q1g3vKMx
zoc`GMw~#xm*JzH?MD9eIc{t3in#-qlOKQxRD~IzZk`>a|gU!t#KEidkktmC-qBYVg
zr?=WlhzobxV4bzq4IJ<FfZbt&mw^Zf=Q>~~lue)jaScz`na<Q5wia!3plxa|S;)XV
zFclJhE~1=b+e`%#0B*_|n9Jw7Gt?gWy8MPdI`f*WA~Lh5pigYCX&kYQdMXedhLVFd
zRzt<hE0v50kRrlG>CU+u%)CMA&aDBxwvP>Z6gxmY*22fmxGQ;mULA=UO~Qms1hD@&
zKxBXe*w^NP*0yK}gs<KInDln<c)K@Ix-@7mi3>lw0%>HL-1uW)xMVmxuD(-@T{z4x
z%&=1zr4Ms5Hjw|`&pi&ldG32Z`5|8)`Q+889N&RrZWOl!8<!5`prpc8pvsp7u37@!
z6!#cRG6~vT?T8epI7!5WMR^2|X@}!LW&J8wg5reM(l3*clM_w80b-RwO_+pTFs1T6
zX21+c_;5w5R|oL0noZs~$1qGtDi*47^>W_^Xsa{R=;}fpafOe<4n!{{#bL2z8@K^9
zpQA!6PRxi%xHgqVDx-RJVBsacAY}nEKB{0vQNfswBuih^p$Q;e)r!C`@KG+4S?P0L
zTgi~Ve+13QN-h|4IFm|sK1UifJxtPsnM|f+ilfk)kJLmGRCd{V$`Nx28n1-v3ANx#
zT>6?-B2CdzAtfg#eU6XDLX$F~XaX~S!bkN6sybo|zNYA?3eEC>2!Mk$$XHm$(jOac
z4wtoHO-yi>Emsu_5&%XEfOzS%p)+HiO0Aym?&1u_zhV;3lYbE${!k<?Ka;a=%2J>c
z9I?%GrZsSpD@5f8zHm)Xu8<6F$hmC!Z(vk~R3u7;t3@ia9$qs-6IE>@nnCCDR1OY!
z0Cr_HH%GL9ny6l#1bNIYE37)45~U7Q`IYC<YIf8cFDHillO;3M)Hi_OJJU$I^mEfx
zAriusmw|XCRh(3PME65ohjMj1mt{Z`)?)HF$u?(ASb2W6MrG6pm5!1{e#=aZ<HGHH
ztH$V115I>^j&PaeNnfs8;?ieT52GBw0qpV<H&@6)>GK@XqQoFxlhqqkt1Y>JZOtSv
z;ZD|pY^{!`6gV6OsX)jhTwOy+lD^P!FCyg;AK|tl@R-V0wJq7ESEBnxRdi@ZZWJ|u
z5P;=aEEBEfm#8?3ECNZ8zM!-&Bsz%*X}m=2ur-}B46=r7LsW%uS+6)MkSy|J5;7=J
zlEtwaQ{(9ZhH1joZ#p=@pE~+nULMY!(a|DaQv!=+B7UALKhwABl5BG^HUtwAWe5_v
z=%&xy<3Mzm9k=l$Q0PlOfXk@)y`OJBIGFq1kL=QY-T>M+=TDA3eNJ+;J6~sSL1i7G
zqdoje-A=;M>x|V~>G&(j%$;j4f2A=umO1jdbHgL=yYZ%WLjdm$XLe)qW1k`}NN-6;
z$3Eej1Mtz<uBm-`YK5;V=y(1VKZOX(OZ!7xj9+@pUk8u_@O=HTiueWH7)j*Y!4lHq
zqG;z_ns(!fS%B@%opR1V=@9XL^Ye@-h_rCsX~*(z(JJ8ab)G+dd8MLK9z0tP=d1RL
z$s6cPp1i6~+;{_L?4xb@!F8pMyz4Q_7$tbB3$X3=*`D%(WA1iIURB@N@MYTaeyV)+
zh6g&5Y<}<O;NYGczxM;q1>o|To;@bw!-6c}*ZHXJxhQs}6DU*Ck9``zbRWwb`e?XN
z50$yN3wo1{5u1DiEK&7pjvbqHvSXjJ570NTr47t`0?cF29C@qaGiT}a<KF;!Jovbe
z4xD0kfv2^Kh$p;r*xZxO;74laC)3`y`^W`t!bcDt;ZX%85$?C%2AU=~!a0tx)fV;z
zrU$V5Wsc_RC{OHyBd2j{9B4%3q@}?L0jy6qiR4~Et<$5l0e_zjIu5{{H>G*?$AzOS
zZ|9-dVyL#@vkPpG1p=k{TnF_+IMR3c34#en$>B!UD2@nD4-Z1+33sv%M7jVT%+z^s
zb>g5r!r?e`3lJaS1B>7iw)OM_9wJLz`cMjx06cGhNJI=HDbg2|)}2gjO9^Sp8^Eud
zg|4PKAT*GQvySCw9cTu5O5Z7*87<eP&K!>DGfSq<J1RR$@H@3-E#g;h6XhB5dq4L)
zxYxY*e&YWe6s|@3DZ}Rf5lR634~t5IuBKBQP96qq63#*zjt{U_5e-0mAjLh<U0O)G
zkIf6UQDiZd@W6yQQ@4=P(m^JN(Rj{m32|;d>JSGYS9pLwmYS+J#%s>j7Vx#CCFQFc
zH&|miY)tVTJtp#>{1D*G6ib&Q@RES+k$pQV(Wyoj#6Dha)Q#kn&GS($xmKH7rg@U8
z;e;Ag7E}${8EI>fYOMl3dea(?%jJ-ummh@{msS_o5giIxQV*eubgD%0QpJmvy>#*T
z5~e;Gqn`0rs(r_qMnQ4jaJMTr+IJo&d-PZf)A1j2$fs>EmTlZg%!N#`?q1{$qE=!>
z)pQp%!yWyOe$>nW^-rl}a+Jt!!$D9^Ge}%BDP|@&-sqroO9$3|%8gE-&XR(>IU!Ix
zaHn8XK3H_i8B7R;dFCo1^g^LqTP_<9O1#RfURG@Y(Tzs1zr+ZXmgPApSa9n8GnpVh
zSGqQ-AM5aCB9t)dFQ*{9y(OinEXh=yQ(}^W5kE3=Qn5=xnm0}pT=xdx%Jm5Gq;b4-
zSJ}$*N);gnX>CVzPaQXH)gFOFH7Y}|pJuNg#h$G!jU`8G8401s*0Niw3+=hQI1FrP
zt8x}Hvr<2KdXvW8Lm|}-5Zt6!G9co}$F4z}mAJZXN14crY?7kWrC%(Zs<T!kM`0Fc
z+B_RXd8$;HyV4x#m+*?TE-CPAXZGxIbD@fEv%7_YSVRZoM^cu`MoLnHxPov;-Qw@$
zCpXscCCHSap@wp$aq6gAavj@vh!{k22{}r|92Mopp;I*AEC7`we3tmi<Z#-0Q$dB3
z7l`s|<~rVIdML!*(W2u+r1A+MhI5N-2ZxZW64*wb=OdekgOVv>1V-5N(r9Q$g?2ad
z77At!AIR}{yefG+`T0V4*%w{28t1tpIA=uSM*eV7Y!$`JL}^IE2pTFM%FSHnZqdDi
z4Q!&&&AMoP>E<4<l<}ov7OavLOt!@WXBU*w0|chHR@+PmRr6q_3`!M;0$z=@EvZ2&
z;5G_sm9eHbE?W+Q+X|7)j#rMJpY$u<(o>aOtm9{97ZjO=l{BL)&3>G{&Mz3mkJ(_z
zOA(3CuW&0)8SPk|G%Fl|PB8a|D)Khc8q#jkHfV-;o7gFE^Lsz{I=J_|_kPL+$sL2q
zSEvHX12CdP94ChQsO}{wjvdyE5!uBPC{xmp@(o~`4(ErnUISDyf{&#2(1_g31Bp4&
z<Qt%o41iD4BjU)fo`Zq`d<h<8IGb4G9K$dnsaVLyJdZ=oDZj!;sl?CtC}PLfrKC7{
zqH1zRo0|4LAP3+X8d$;4_^4W-qNqU3<>C^QzNkanhjdjd0>8jVbD95nozT?opq8eF
zXdZi%uH|y1!90Qs4$b2i_~?KRDkz&l1P6DZC?j#{mmb=gK;eQ+^(%a|nJ5652{V4e
zM^OXB7JN<75idc|E2KSHD1E^KaA1H8fTQG4PV}73<yxp86P#tsRmFk?fJ>QhA~o}q
zY3~I+ySboE_z0pSJgT51!kyND3SjgAY3K=P;5@)g+^1_)M|@C5*)m=&%+y5n(DB`4
zQCTXL_lYVc39)fvA=t&W1V_QLE4{2g!PM$Po(2{A8Rm%X2-hyZq>8gMYnnEZ%QBz|
z<q2!5dNFT8yEoPtr!@@mL|~O<TjufNdb%3fIVrYSWdm<^f$cF>U@8yH8DXS>aHQq%
z69f~ElEa-gq}vLqwT%(em8Eh}0AO68M#0sIhVlqUQ$idiPy>X+{nP|5RV$f(@G-%d
z2P6Qab@(_E5$BN<=?hBhLLy3NTt`8^0aVN3tfGnlRUurIYImSuAQd%KUIn_vv}n43
zVVW>?n+^`}rwu-z<qn5sh~AVm0N1+-7xZLVg{>(xA+wC7H)ZUtMIr;hwi?8$x(3zL
zb&Tite(rs6pZ?yD-cX5nRRze7iziT~q#r9Azz7fLhqGP-R55~&r1j8<+{^=sInm@B
zppguKPtqge$giG*f&qL99%MM1SmPYSFd?Z}$i_U6L(M6_!bhpZ&-f@}$JV8!IC-LK
zaz>k)_B|j6;29cN!O!@pTA-q+K+NUh5|qBEL)(XRRVxC&z(;dgfA1$Q!rzgPqRSP8
zP(isyh~VH3)QpNarQd{+DmSb7qCVP86adVG89(8pr~zUNzNYA?SwjV^Ckv%7cmNIz
zkO6R%9LkBFv$<Rg^<#pwY`LmfkN|Kg6HcUNelqR7pl3H1v<V+Ubc9D0ltj4G8c+d@
z9v}@p0S%l7n2Gyzjp~RG$|zgLtA&}Gs2)1Ldn_tTrSd*er6eIXPAmkwxR&53SazkC
z^(UBGUC7fQ(RZk_<R&;y<5<<=1~Fl2H4{hlpjT7X6QE*H`@$*?p*wM#Hj&FRpb6#a
zYN~oMZ$hytSzh<Y=!Ck4l52Mj=ObBb=cL$Twx)?^7uX(C1*Y<v2?c;O5H3oU>+ll<
z6RJD~4mT!7akmw!Y8xY_D@)~|0Km9Fje@Ha4doGzri3_3pauws`>6?Bs#Y@n;A4U@
z4@dw;>+o?TBF-Zz(ifE0g+!FlxQ>E+1E`k6Sw$5AszSIZ)$TySKq_jeyb5%SY0-26
z!!%*)HXR(`PaAwb%N-8O5WOjB0IqivF6ha!3R_cXLS`9z65jiH;aks1f5qt3^}}00
z_w8>2nQO=)y5cULK$(($lxzTQl<jbSIO{b)6(jgaS`UrL%{-8p6HUGW8p#0oBt0UI
z{OUO<7{Hg{L58!5HO?^%6OxLBY|QgG)SU7we3VN3jE^FAY+Xu<lP9VsXSAtl-ve?0
zo}qyi{EUyP1uBXP#9S^eLFtP+w0%fdwIc8fd^DH!H-X|J{2loyx?DjB6_jg)2oCN*
z&8Ub|`b`+Aa<iH*>Z8p>0l-X{@e@9Z8X&gdYl@DVHB`WQvQYYh2jIW}830Ggp`7SB
zo6EIOKPEWKmaB>d2>_Qe;Y4cYC)3^wdUkU`oA41tM|f00NrXGC0TsaL0n*SD(7<_s
znYd5asE+ucjIw3CT9~Pc>Y?Mi$D*=SD(@3jN)lq@#6qx(YYC2mWmkGxe}bvig**)s
zeTOPbZi3@9j#Vvg5EGVGGjT)@dNox&0V)QyFRbDax)Zl)6S*t{noypurm7e7CKQ{J
z<#m6IPN-`rxpvoZK9aR|PKqsNYnpg=f$cF>U@EVfPyk2+;i6Qz4nILKp~_R>aART=
zcUz&VwlQM5vQ!QV0E`RND7ZS&P#)oEN{FKbYJhOKpPIm>Y9-STJ|-CRfCON)4j)G%
z;yjWfeL-nmNJI&Z>nO-KfND9MRa6n6Dujzt?G6+Sq@sq(t3bDy7EKp0OcSPV)4>7$
zw87`I+~Kec(VLP6;CeUVf}Sj^ur-AyWR|ff;Z30M-p_px?l<?npLEb^TQE!M;Yrv4
zytFW}(y5fb3RY^+W%>sC2Kom22Kom22Kok0pAFEziA`T}q+jt%rSJ(<`Wy;<CPmx@
zw?#z@`k<1z{ab!QS{!xUF+2~RuJ?&O#b!%A;$t6iyUljftNRA}2Kom22Kom22Kom2
z2Koln2J%1tcE5xB&wcNQzw}UYYX+-)qP>d9ny1=SYr=shP$o=)R9xHxJ#0c>Ls>_~
zh@P16h)=(`-QvQDS3>FwU%2B!f-GEF2VkPKsftuj&;%G;*f{m|Q6XbGGoNizDI^Wc
zNcvoK+*Zz%e2Hu}T~(@{TI(kALb8z=$~9yRAQ^<Ob~ZfH?2MjOM{~pmDox6Sq6y5v
zgeb1A&-T$;wXvrfsGvOcH||jcjDEq7&)N1w+t#14xjIVqbUDHu7N7zcJ(L=H0vcEk
z&_seST=2Rj8Dn!5ZgNV9>cV7|Q$2~s;bp2xDiM8FU8i#kBQ42Mu<S~2ya!j+zj@1u
z5FD2|CPtYx^*3P|Y9=n}(hqtj9BzyFWv*ez?KNDoT6X>3M$v@J2XBR1B&qKjzWmJF
zsnq+?(JCFGBuW4r!0wERn=53Y^i?HF4B|Cey{KwJ)hPwz45w&oke)_I)U-uLMv(=`
zR}zw>5`KW`AOm1|GDPtwn2ByxeBLbWC@m5d0qk}g;&P6y5y)^ffMP0E3fF43mQFg>
zVl}sLI=Er?m9EK<%)GNmcJ}cPxk>WS2q@>*7qmG1QYwHbX=Bs};0w$1dq4L-c)-H<
ze(E?hfir~z#hSUer>&dUsCuYNWi0kw=>)ndZZw!=612FwWQvnSOjwjh@R)Wu4&+|<
zdc9<Y^vfjV<V2HifLLWv6DDC7OsRa2888D9KJ24B!f<xs85~gy*_h`E)=4=~E|{A^
ze1yw+0H=W)KpalEfF@w!#yT6jy8<8(Ah?xn$cmzZNKJPxE<x#wIy3>Kt6CBG1wP7U
z!&zpr#cVI*VXWlXF}lv@NP}6`1<{(R#xL+u#D+3?DBQ<YoYEX|CUg`OcC(r<>Z7sL
zq)f=_!NQE6@KGd#*n+PqIuc{fae)Yc@eXi-n18}xLprD&*2Dy7*)}R50buwbK>ViP
zLICMIc!jE`ySg}o`tSY3@9l?F`l5$9?HV}4cBEmV`DFZ%5q#mc!^Gy20VK;C7lm6K
zo?ihnNWQ$M63J*nlWg)NkZzV~V?aaYglmZ$(E@6sdMt!3bIS^=4yQ!1166+Id9<1x
z^~TGIA^!-;05|mwkof?Yu=I1&R4fw0warVaIAYZ8yF*=va&<D7Wk3_wV)8i2HfK#(
zd49D<Wz-0jj*>-w%S?>p!tH#k#^_K3O>~KlaGB&uU#?r?(q~l<qa45i?D7-Wsc@h)
z4oVE-HCeqVXTsFNy+_Q1J6Q*Eg$p>Pz~LxJ1!Cg{FFn;YlqBg_cJNA~I5DlGT7hsO
z#Pv*NTSdqdsBlX{W3^P`(yxr_#XKMZSdLYNf~WZ<DvlzHKoX=cD6I>LP9j1YFA+Oz
z+fIZGXZ;#nHAPKvRDiloz>Q7}GnR}YG?HWaT40zaOuXsf0Kc|e*UQ7XGdi7!*Kiwv
z3uPjI9V8*9C5tneX9y-F(hwwa$qo1AUuGyyZneM&eOV3QGHQPB=K%-bGWWe7*(Lg*
z6@beGRtNE6K^E}qeAM<_6uZ(1lqu=QJ`G^HkL3-0G+d~M%3Rz9y~)OiO}+t^sCqTW
zj!in*u}|3t=o{G52If5h=CNmvyjAg;vvm6LZvZ_WeB4I|PBFW{(^^Hu6W%#&?n!6x
zBQ^7rY46*8<bpQgBZ!XhsDhFR_gilRO%oj997ot{3;P1o1K9mCM{{+QCw9S+(>OH_
zG$L}+(%^&u)~B09a<8D)=~3E%zfT_>-WjUdH5{G7&PlPwP;J3y7uX(C1*Y=KWR8P0
z5bj7sXf7Y=91D(yPq>qHAQ~LN6VJ4?8rGo%!qNPQkMMy-a0%N<t3hOhLs3uwV6+Y&
zM<QUDnRK8s4mW)oU|gd!1~xEdpHnykEw-DQIULbvmQ0<0RCbi$cWTR8#IHOGG{5)r
zEe8*r_ukK0@fQAZ@p-p!_0TO`SCyiFxZKj)ji{0tfl>k&WQN=jRkcFFfIujTo2Vv|
zX*XhHc~wz=K<p?}V!Ekq$4c@ktDG*#+l|-LUep>_4_%E-@H(6I22=}`I3*BHyxJ@v
z@aQ@<Os&=y?QE>huY&m_hw{7uRXbgks*6UFQc7D4Htx2qkCbV!W%nEHi!R&rOSe(0
z{IDS-T#8Q0Enc56dmf*|Xz7lR#x^C!PNx@8#Ii%}Cfuld9E<Mx=U$>eT3qytGg=}E
z%c`Wpm><t(NpLMWp~kA@1oM)uXnHQF(n?_wwke|{=efka>Z3+KJW5_Twdtmw5DhL=
z!x>eg)wUUhwqe{&JMi!WCud5uUOz?HIPex7y0*FmIdDtd6sb+Bx2YzmcN)Zv13H9B
z$I(jsnD90lpLS4lR-(CLU>g5)dbr5RcliN7YnGNQhta+0IkkhgbozyZQ8As1Ot@vO
zY{oPb&E98`q6t)%?y{|G5zzAJ;kwzQ?>*EpT(fkU_>vRg-0Zbo7CXWb@_Qn7$Gr`>
zm(3#*=3Tzb^->Oolvl@7?>KU!J?qR(^@#x7Bf8+=aq1OxX;?59x>aK!)f$eCO?rN~
zy5Qsy>PoBE6we$lDm5Zo1?q(57?b)jyL<{+0rL2x$|S8!s-w)QT9eCXNb`rIRUmu$
zTpAZ6#Ds{C*QIHOCTP>duT@Sp)T1i8HT@FoBpPN0&d%@sJn-N_^WOX6PU^a~*Rase
z0Oxc&Jv@(_aKU=jMOwHuFFi$2fn@Eff8qw-%|-S#mg};36^jc?>l5J|?J;Y&>ghEB
z9KhkDwJt9azyVy!cHU_ZKuC*jdJzAM?@d*g)z3TKJod7QcD_n3LHc0u2^3Iy8o}Wq
zPrvi8_?!xRNvlB%^1k+(uo5IUp>!3>-2N>;AuYbjc-t4rgQx2~>lCxd+ds%A9gg-W
zN84?ay$sbJ-6^)*Y0L6MJ3MW2r{fOOnI@iWBCT9-%ct04g?xHD1IHdnZN3vE-rx<e
zdGD^!H_$h*6C0TR>}DogJAV2S2xTpMc3g$Y(_e4JC=n`cb_Y>VREi^#%FXZnJm}!T
z{k<QxnEAozuIEc0YXv-yn{dH8zeo$W=B1|yDv+%1pVk}b*S%cc#j99cSX!S5=V*^v
zyH!uG3E%+6UpayT0P7icIcg6;NQ-WIhW_5qrW-ugdij8mJ{W2O1yr6!aDVRyUd;CK
zk&qUL61$bigZKA-j_$IY*vE=m(b1h^%bm6?ua<VY2B+f=)0rloY$B~(f6J%XVugHq
zI|IiaNNv6oB;Mc+uzByU&^ORGuoD}Y{_JKZTRZ+4{oc=m4<54cy&t>A9-k%~z|Rxs
zfQTHagGonBK^qk-rIQ$(*{2`9Vn?^0wo{9Qaa}MT!_P3vHlJ9A(uew!NTciQgua0l
zH=rJrE9x$hg0sMc+r?-5s4K7x+<Jp$V;xo1wvWGU{TX|zqf}3qBYZk*nu6AoQL3M-
zBKX1uAJHU3ww69~*vee;VWuXk*L?7nsU{g3=zQ&t@XDC&X-N@Pw|e91%&l*ZGXf7a
zE_z&yGP)UHCNAmH52_|Kebx<+ve&wkAEjEmr4~u^>>7TQTRLNVrHk%ddI#FzVFJ}o
z+TvF~2ogj%dWl7RgzxO)cWB^tlopA51n}$~kLZk~XSX}@f^`qDSj{b*4yLua{3GA!
zjD~PvnJ9j7Ob3J)njx5wUcM|8N()pR#sm;s*tqna<}57FfA8lZ2j4pPy`QFG1Ej$L
zTsLFRhxo7{*ZXw}EkSWiZ#P6_7f+x}Nk8g0fayA%AI^FWP{jy7lGZ~bax)Jk=0uZk
zfJQO^K1q*=BfokM3I^~cc#z?2VvTbQ!-S+_Ash2N4mGFz3Lm8sKjWi_9b1=@;^c{{
z$r){G+V_AQfM;l61wZ4XYJrNP0x_41OHlfv4s9RORjmm80w2v~^*c{O2fKq>P1!uQ
zT6m)s04<jz4dyZzv;wk)U*MwyI;fzuLIekQpeQ48>6aebnLy!!O!X^#w3#RXm<cm}
z!bed9#1?!_(Gf2}&?}@pStxzM0&rk}41lBLP)_um&E;CC9}}Eq%T>jK1b|DKa3VGH
zlWFe-J-fM}P520+BRs00B*LB6fC^yr0BPt6Xy81+Ox&kyR7ZSJM%gl6EzHzJ_0aL%
zV^LWumG_A%B?+-{Vj<YYwFF1OvYWjc<|f0&p-+QE-_pUUEV&5|DsORvn6R{(i6eT@
ztEuV<P%)@|VHJnaow!Y#$YmMOgz|JXRlS%uq1coxulr+kLR~}2wY!G%k*u|IQfx6>
z)5Nn2Y>%k|Q+dsV0zeuF7p2N|_z8juRh|Ne8xy0r+X_{+jS<t8rE*XJU|gU^!PSX|
z@(4#$LL4Pf1BAo<)C4Y7E17=qF~OJzBmkpz_&5>~=aCfY3rg!kB1&jnM?t;;RLkM4
zqKW`jAzYMdcc5S(6*W{|1-iwwXu5!5nlN>n4i50A4L+ac4u@rk-jp-|*SiT9^ki9u
zttm7ivy9dMxLZnsN&p){6sqbPR8JQ&PVfD^@Ed+1{ntUKt{>h4dg$CYfwC|wQYd}h
z_e_XRx6{M(xCs}mmtCZVTl3OW1QkeD_fP8$^y^+O@8VS~E-bB2gmbjVtlg@o*933?
z;}4sl0Kj^NU5?rV5YnQXo?+&@jGML_te<zfdF;W)CpRgVAboJ62^3Iy8o~50fzj6<
z=~sMCg}tQJpal_k!EI5|f<CBZZvU2_kQPTBcMQ*ir|UiI6tl=Xc4{-u(H`Y!yP2b|
z79QOxw%lpU@~bU;mxWHp9i}r)JlRBAxrUZ+<3c?Boq?kdq&D9R5^wMZ*t~aF=o{!8
z*oh5He|9sItsOso352qiJv*+#<ms=sVw4D#HoJo;C@RGfN#*9h0QAs<hb?^XN2&GT
z^x44ZLuV^T-D=ml;)$QDoFHsKC21oW*vfLcDBM7E9kzZUwtAt3xNza=d!lckZ{Q5u
z!0G?Oz$eUB8@k#WVLHQ;9}lH)a>W@<+>v#yGDZ@q_tm3I93ux>?eZt>Q8`WHh;1xa
z!Kh#-Y`THGZ+Go3IIZCZrPCT5DQnqjmQ~F=*jhb;&RSCQ4sb;5e7Mv4jMG<%)`0No
z%AGlD1G~ELU7c&zayxm}2KqIhwXQRe93LU=!z3YeUp@AZVoUkGpNAbhe9?P9XHC3q
z?<f-oXg5DT(p2ff3m4A0jW-Gmr;edX^@_xmo7ka=Ch9Jb7%OQO)fT8th+xShK$K5*
zCoU(@gk&iqtOUX#)XyX#L3LeN>#~m(ROeU0{IN=Tab{|y>yfsfuh&e4#q`u}ak(z6
z5378NV>02#9G2fUsh!&u?PspevGLIXhVc!@-RNazO0j(igwt#|^4jIzKoJ{Wyv*v6
zIHS?+L}{7aahhFk616%SW!eChrAMjwmc86|!!COG36axpLe++p(6m*E*Q$lrM`~~3
z(H=@GXDRDIp+giZohCYiyNPp)>8qwZQ$<(t>avX7cu!}QMq_RWhrvSPGC1>W6$(Qo
zeLP*tj2tV7ScL|#gweRw6eUaLm)42aXS_=fnb{x9(&KE&J>GX#!-Pp0I*Ute&<1eJ
zP7*lhM1$9E+T&sM)rgstcES}Tf{+!_Y3RDgb3t(UxIb}CT^bh5m&L7G0HQ|JrpgJo
zg1$F{>I*DJ^JbK3#Ac7D_Rz>J1i%z&^rBf#Qjh_skO!Rs0Vf5<beP8l$p@*PA*?|I
zI}y>v85(3b>sAixxfKD74-%AY@U!!KKMy~6#Ju-@qEb9l_)MUY&A?b9eiD&Lno7GY
z3#+MkHhqU$8=N(yD7A=xs*}Rhyiko)SaE{{zylMe2L_QRnXz0!3&{g#ONevxQHMAH
zxxxedvD8$(F<x`Fwjl4QUfH_8v_VyplgXs3NcqU5HP5)EqEM|G)Xi9!dZbHOGqGJQ
zDPF*7C_zl!s98K|VJe5@VjbC9965v-sMg|4<!sASL#ZHj^dJX|<cNp6f%DYTOj2Cl
zAT2*Fy7J`$RFO;wpDqDwWIrJZ&10qvUp2p56?O6oBDE|}Cph+7940O^Zy*@@UH)*)
zV~bqeNv)-`(CNIJgo&1K<i#-e%VMbqttoivcM_){{>g@N3vP>w7W7di`2;*Zg~b=g
zDn}%2!tvBuQV=Ba?pPpo;x5^y1_Ld!wCbRa6oNRUWF)<aggcojrF6s?@Sw!2%<AQe
z1}I9gidBl$M^X_tX2}r8Ez_b1qdY+oO(tU0vDTy##&RXq#b1QCx7Z=wrQ<R^*wUyt
zj1fS;B8mMNR4h#sD1k_*djoLgdIVBV&Eh<jtUa$(GboB;M|HPG!}bVt3XRIpJDX;&
zAH|-nER7{cYZ(clN7S-gstfJ8yf_SOXsdD-GP6=Yd3uw^-9sVO4G`R<S27^t$j7ch
zo0Yh_ZAY2Ni)@mj)1_Z5oT{@{Bu8NuXWBd)M0u)In7h&(>6h?|v@R*|Y-jfDadV-H
zZnL|Ef>=Zc<401K%0^03gSdilN8RG@<R>@Q@FmEUp`nIyrE%)0T5=uRc!(H8atS#~
z#T*so#-UR*;4A=@BYc+l%H(j`dQ(A#lNX5cYUVoLXL=~a-O-}sL!|NvAck{`YzK#s
zs}k5op64T*hl7$SVFX6l^U`Q&M}>Ab^A-wb4Ijwycf2ZjJNfxSc-a?Svl{2QA~<J6
z;zs^(QEU~(%S35N!U!5FAIi;K=5EowgAHt=(9ODNed*>NuaxnnViv5D6->6p0%sSL
z(E|jgxK`Ut2UYW6r3^|HhXP)Wv@NMYD&RH>YLzk18!jBqQm1kd+*Vwr%#K%%o}ct9
z-qKT*T&&|~W)~Ehg_Sg;EX{tLz0NNf#E;ov$V(B4(64YSP8sc3oir;Pfle^@hAQ$l
z(i+lk(l%&@c$?TMaPxaVk2rYb%=dn{Q@k?e;)IJ*E>{FOjtQy>0AozxIAyXx)uAe`
zi6uD*hb(x6Dvd%()tOvh8AL7g<Pbtj)g6Z7K(!^IzG}og-G!UgQAivh-Z314sAI|k
z7Jnx}PLeqL*1EkBKaWo7^d|vv7D9L?daxxQwWhXdD7LA*Vv*n`sddSwNkUlb^CXp$
zv#BbT#(ORyCi}_8G+`1>^|`r7#9deoTP*llG$UlPa&ei=iFn8w<cfGsE{?TXnevgX
z1alM0R6KYbkd=P|jl+b8hyc3FE`<z5Cj%7^9tYW#vGPM8sIUibTXL;WQyMv<M18ig
zTS1LN!HFGiGR-n#T&6Qk@e7!AIOH1Ek8uo4J1*e9nZrF;N7~@0<3%SI5+R_J;Czv!
zTtexd;Y_T#lnnPtD<!I*>NsXgWY^Da>Z6PsivlhrOShaVf}H0><Yka^p0#|B6EK2H
zui8GsB@<U#X*wD3tV1}tG0#B}(e;Is`I)Xp2^ho0$ykYleDR5_qf!p=Tu~k|MIdSx
zVRjNdYejC=<yOY#68d;iA5?;4!88gJs7F*_AO?(DkYg{;IdiRuO1G$6I2quW2^@Qt
z6c|rxAtr!}b2dV06`U=FT%S0hFfFI($~^Lk9P49wk@m1iD|h7NKXGpMY5sC61BI+8
zD1JmmLbxhLf?0wcqOu=dCa^|M$6kpb_<69*iCSD2Y>Z<K2RWHdBru6>)+)sbm)SBd
zfiV&83PJh;X4z!Z+*D)!a7eUp6Xz>T2IF&U*<BJ(4XPje#gL{*H~E-|6A;WpT1yFu
zWetVNJp0Tmuc!&AYD^S1CF8wbX6PnzY9K0ihbeHXq&YH|&QGrQ$EtDcl05;!@~xa9
zVId4ot|8CdVw&?xA{XY;sKGv2y2ZFOEC^?WEgiB%FxeZmBugab2ajFP3U+o=d8wz3
zEV*R?UCvr;U16Fd-|{)cF!^BeD8wysS}y@qOShC{#b}8g^U1_&8Rs;aa7&W0CIRw3
z_e+Yo7X{fhHcXR~G7;u1aMDgwM`{)lWs$;h=v3XPLd^-vx8w?dNyP@DOmiXIp&WB&
z+i9Yjr6b!iXK9nk)$|a_t*GjTDYC3qAPML~u|}DYepFF_460u}b!iArsXB$A7Ovz`
zOXgHO4pQ(GB_4CU8pevI1#Ha1(n?Mdl7#9fDi^S=MHZo0V=)SMoROCMM2;wAPR_@4
z<`i`s=f(*Z_vAWUl9nlo!!0wpKtK&!a+udB3O(jqj*X=j`SNBp!V@Y&o<==i%$66a
z`oPV#?uP&YmXmXr{4`EN1q}Vf^E<#Q6HbwaC5}kO)1^AWHz<XoM$ASi*J(<wiAwpd
z5c4c_fb=WMC%?`@df}F<;%Vbu=@_U&vXH4HmBv0rp^zyZ2(_9P$4aSEr|T4orb<#>
z%*RQ%CN+~QBcHBSYMS#jO#r-MEjC7rd7hyxt4x*$l_wFi=Mk8cWD=NDvM~E&6Gau4
zWElb(V_(LxDBp2S%2$@eD_Scv#x&I?tx2X*U8)K9IThm^GKOp}sp(wT;GtD|43svt
zCgvbcQt=iwlYV+Akq>X7Bo<gp1vHAEB~+A=n{T;JrI{^D%6mVLJb2W+_kOrjw|xVB
z1APO11APO11APO11APO11APO11APO11APO11APO>wE^)Bv4Ih$u?7TsYG*f)u5G**
zu0L*n9Qb=#i1u|ZT-=#cmLDShk#bCzJpa9)M;$zR_Ip3aea%XLa!O$KSn%YO(9nj%
zT9v&D3>Shdl@Cccxd4@S0Hfn@xErU~GR~IqVH>*8?|O>^M3mrog#6BA$|SisT%MI1
zk{xbV^P|K_$nCQ(_2gf&r{nV&P6|SzMz~^R<tCZBhFFsC$kdn{280taSt7z?Cc(^~
zWOGR{)r;L(OTmT3?An9*36YWXV^#t2AwqKT3^A4S5u<_>(><@t$V6t!M{<nG%UIEf
zkJVV5v=lMnR1ynIj)j7yVTq6**Ext4{#{Z_g0Yfl5iA*o1K=m3a-kQeZJwTk^sJ*>
zwt>cO!s98jn@nPa?)0nP@-z|7IY|h942hP@a!lJHH)i4_BpShlm4ezVdWn+Ap}c5B
ztsF(sL`l`mDxyUtN+dX9X%PX4H!`sY3lJxbo@Y&l8rDvw3xmN{2Kg+<HIP^nB*1^9
zToDdV^dpI8Od%=cfF~y?-Xh<+0Tg)gCFV0o$YE8edXXr@G=#mUHRW4zVyvUe3dxkp
zge;J*CFSA@lqNBa9Ouanq+-_OLzsW3M7&RKD^_sPNMPjZkdGr1nNi5}faQF4a#>Id
zKzw9~hUA7@sUepFy0MjB%1pecez};mq+awB5XVZ{%8-?e#>prMf70(%6%ht#w8TmG
zqAR7A-QpvtJa|ahy@`%I0?iz+37LA6<rNKHJQ!>s%Zi`a^`5KQ24WPZK2csxLTH?g
zX!%ccihgp0n#pV)B5i~(@8~HS%!H*E!V6&|2dN`mT_Q;mes$w@=gS)|cYH*|D_<^u
zjl|<2DdYmxvL2SS>o3L`p$C<mBjqc?e1V5#&S#GdAxSeU(t|cWu5%m(_-KU1Ke3ej
zc&npOb*2<qW*w`Iv!;Q|*z*TUyU82ia|#(=gx{h;0}$6uf^ls^irpxgD(Rj(P`aYg
zFA51ydG;%x>vB0IwtJH!OzDL92uGTTk8q>|a3q6pATt23rv$GxYOxkRaJs3fxuX<a
zR9Fb3&Rm6g{3+}yoY7-w0+L({*WCu9{8C1J<sS-CiqQ~Px4e86<&=z7sglUz7V`C1
z#w(CK(UbuC9seXj_JwH9G7|auE`7x3Oteuml@O_TF%aq@F0bMYN7Yb~G%6wCwN1&F
z1}R75NLKI(U~I%vTL0mwC{%IdCFyiu8!{>5ayU*E8+)+;q4JOs4|kDNN{-7ZQ>KYx
zSg^41cuO`NBn+Oxa;~ro!OI6Dk)oNLiYSGqUCeT~Y*R(tdRESArE{2pxVdR`(;b&`
zAGYZdWMTjbuQ~E#l;CnkQgf&zBa3A`73Uhk&$IK+%TZ{EbYsiK8CJtc4(Fwf`EW-z
z2}h42L5n@<?D1C`g$QLR;>Pm)-p`{C9y9m7AK5RqZh$lpfYrQ-In0Lzx!CWc&=M3!
zHby=oyLbX+O8T)y1DM9c`QfbB09B0OBWXP}A~*9uVoo&q252M$;FI)-IP$CKpkM%B
zf(IGSCe}E|Fic1)7P2wV<4|+TukcYS@iRV(*s*mfDNdfKnw-(5rhN~{0eFT6R`4@E
zsurjyDiCwIxCEsy>d^KfUDb-fFYwV^Rz5BYnE`iDtErpERts;m0-)t`q`_R~f>uDb
z@C$r&KnE3+R*2x>4)p)CcOGC-9BmvQz=932qKJA5*4PkBY~P-}Tri3yu?Df7-B@B*
zVkcrWdUjM|i;dWgea_yo#0nO~h7|#g1<N-VxF~q%-R*6e*?W1OcYoaOPW#Whvu$Sf
z-)KHv|LOn#rIGd@D|RskNZr+R2-9mn%Mn>&eowPoOm8r~)?5PS_Wor5w7aDJWfb9)
z-9N5M``@PlKmGrcz09m8dzjf`X7#^Oerek$Y1+K6F|676|J`8RYNRpLe>Ahf%<BJw
zee%!0yQc1+g3p(q?4J5>YJkr-m|kOUcXR9ijm6wXbL&kcZDv<9tIavcXS=7Y_-u7*
z25ek!diS(z)BZ9R)VS72<NA*@yXE7CkN=q~zR6-4f~5rgbeXd2r%@W#r2n0=m$}s@
z*Qbrb_>0i^@2CHz#%~2ahW>A7>sEb|?w3OTx6O)wOeFkosL$7@Ta)hRe@WT6-rVja
zzK@$e{`bksG<z75|8KAV{xPhj;?KGqncOSg%5*<JE#1dWAOHJzd%B;WDD`RZkNO$c
zS`qHET~fo+w%%Bz^h}>m!|(L#(ueu%x6J|4r(tab*6#ON4nJ98E)H|+KM9)VUlaSM
ztTMYHP4JW-CieYzgT?hGlQI|T^MuXqX);<va5DkStTrDYO;*f@PT9@ec5~~^<<jhW
za~sShU~auxeLik5A2f|5))(ATg#T_gvy++C=HxKF-drS~l;_i3KK=ibT|WQU?2hI(
zq+Or(7m5C}6={<E^42uFe)7XyD0AyS37YobF9i6)rhnr|vpvm^v`JX^%lZh??V0Xp
zN@~`x{KEd$W!(0?OlHn}G=|+j3-Z|tV|bGrOs+H)tuYoO(*;cT)3{eUro4PgyV6WZ
zGpo~vHTx?~=#M`>{+BkSaeeB$pDvmpE8qPr4XLM!M9AAc&gG9@Z?Ub+y1dVXZvzrJ
zAm{*o!0vJI<FpmWw!D0X`Evd9j(5l8Jj(&?)4{9Arp({&xQ>Bz=}`S%=&L`+9OHm=
z>5#qPvI({3Aw8|U8L+y+vX;9s@XI<RdrqwC>~RSzok}}D!gW7f_bc(be?hS8yziUc
z=YXKYr+!d3D5m+_Ue7q8`CGZ4pBHX+7z6p~aHe<k&VO!Y&^sDyEC^HQDzX~WLR#8O
zTJHy}&6iV--M;OQJ;y-M0sO%B<3MTm+qAUrLgyp$WyoJAHeTk$X72_4PIYK>9<$D)
z8C&fh^x+k!G$z(tTP48d2`3k6(=-3=A4&#Y({n)30sH`t184{6ZU@Rb7eyaDA8|l_
zI=FsFD8G5cZB8lVd!4Mb`xaWtBb8tAJ^ugq;82Sext#PuQ)%xTp*;5WF9Q8QKj^1t
zdq2g_Bf?cb{v6E#K?m>yI1XrQ2jqHg3hB!G+#2);{pEabqBQ?w{r+gBbIAuKIwx>I
z&;k4ajss{1pdE08b^yjprX4SN{`NnAEvxyQa}-MTeuuSm_`Z*#E!^V1L+4exkc5F>
z*5Of$!C${=`5x0kTE+uFK42Wzr{lQ59~{dcd{1{$O$`g-Xl0$B1~>PK`I`fZtAo_f
z_pBZ7Wo|#{_epoZPpBU{>IdfKp=(|qV|^&?{ScJLvEH}ZIv*zN&u<=4+I6FCd?4gU
znd4xUUavh{jx1Je&K=B`@IKSA-e*dEzYghHH$Fn>`LFoKXS~k<(s};}^+HFzz`Sz&
z=9L5897DIR-E$Ocn(Y<GD)o81uxIm!SSZ4EkagEVpcCjs<9J%J=Tn3Dt(&(l`=ibk
zesOt*fo$v0W&iAx>ncCs9F4TS?^4S1QMvlpveth&eSW`wGbmukU>7~`%Q`Fz$@*wS
zMrXZTbhLKTQOir)JP1n5PgIj}+e;QkdRnR8*W-2P-)?b0&;k5Fy5m6T@e9=^)($vT
z;RUBO_OHLG@Q6l99P1Q%UPyEC(RP0Y=M^}wm^!a0b-cGw?{Ab653BtLD$dH-cF34}
z7^qAg_|EUmdN@{(N2&8o+jDb!^OOELEc{m1_X?msp+1%L{S8|ES?co;q(?{pTc`(l
z>p^M#c!hrcD_4B1+#g>$({;zlLot{Y!aOqK&Lac*g1#_c6MOSjvHl(q@~@2dxwO{f
z)Yg9_^{e(6i%q4zzhk{zaGxsgeJbd0oBob6zqzGyg5Q|-8%G(R$;xvH$R}IRk)Xby
zzF2zhX|KP+I9;~K=^E=#ta;v*(yvnNa!55l0PEwuq|yn}v2MNwzB;k?KImq^>ITbN
z?#8S#>+(Jmz70s^#3t+gbs_&@>Rd%uW7>6qaZd7#bBcenJ;Wp6CI?Kb!{&;A9_jt)
z6lR65&d>biX!kfQH{TyRI=QIFOO7_e^I_A^hsD<A0q(xK>`$9HTDjCc{_*Mcu^0$C
zu;2%@_Ul(_zQlTtrj&Rue)TZu`lAF4v`z<m$05=d9_|~`-Zy|Apa<Mf)B8LjY~hZ_
zw|ZvKI~rL(KM&k_zi{_yPI|fMlw-GV`(w{B&^{feg)LmN`u<f61RWgN50X76R(1Bc
zgdL5vzlU=4{5jfnQTp;K<@=hB?#DuxUpGV-H0PH{^zWaxq$WtyRpT^87ogKP8<yIf
z+T7ZF8Z=3#^)UP{t#Q&g`5!7cS=W11282TDsZxWcL}1&}DRp;XiV&a!=wML5(flB#
z)?sO{0*>Z4td!T@dWG<gsQ@efEq6X0(>vHNwcLJVZTjph4JD9L0c-sV_DiW1&;fKX
zC;)x{egJ*|#{skh&<@Di4y35BoFyDja6Bo;n?Em!-2VM}j^te<w;qswO8tWlpo2jH
z@B{FJFZn@=I<Vxc032U@m4k5BC?J)8S(6u2`-qi4Q)$6EPyiGF1waAo6%ea^hImuz
z0CWHy3<`iBfFFP#z;OWW0JH;gv;!%sv%7@j296s!jvMeh@H_B3xNd`XfR1(`B^R&+
z1(cxxv@0p~Xh{d!YTHtNVY_wp3Oa#K1_i(mzz@I=;5dMG0NMd)2cR8*b^zJ|Xa}Gj
zfObIKb|58oSb_qe04M+ofC8WZC;$q80-yjW01AKtpa3WU3V;Hj04M+oC`|!)zMfK}
zpaa`Fuzp;E|5@jMa@RA|lXdk3I)P3G1;7u$55N!LIFPFyNICjo2@24qfLzA|q?_Ug
zpabY&PyqY@`~ds_jss{1pdEmA0NMd)2cR8*b^zJ|Xa}GjfOY`d0cZ!H9e{QK+5u<>
zpdEmA0NMd)2iR-}QgSW+68wWr|A2afdNU{hegJ*|egMY-v;)u%Ksx~K0JH<p4nR8q
z?Etg`&<;R50PTQM+JTg#8J3^`C;$q80-yjW018;IfKsl5pd2Z64m!xIgY`!q><J2h
z0-yjW01AKtpa3WU3V;HjfV>L8{bP#$pabY&PyqY@`~ds_jsv0Ffs~^PmO?3@8=?!+
zmeh3BbkjIZ(FN!<&W5Enr#81Xp9W3RX*~?TOKY4o&i;oAPS$CchSXDsIiVWo59}D+
z<^r~r>t`3cw>9A@2Lv6!57<2p>}`KO9DPwM>BW5vv`>fk^CQpcVqbGWx^##*7?g2V
zmS-49mkzV1y50EeNerfymHn4pyisR227(R-KcLl5`Q{6^J{r5D&n=F2mRr84dA@r8
zU?AvVjUU*09}N5Rdw#&0Zi{Ah#JexN=7gr-pYsa7b=k`S`RPz~$dEhRE5vb*#-q;*
zM53@5juswTo9QplspmMkdG^m2ZAxz3!LiCMeuK}wE%uZHf)3yZwjT%Q4h-vGI{FP(
z3X5-z3-jF>#{v22(D-#!@nN0+;aF#f<tO@<E$|#GjdI@y{Fvvk|J;~FObgo$S?+u)
z!#xfNI)ER*aRBWA-R(e$+|_@4dh-GX^3y?S&ux;DmhGQ+*hOtCmz?kZep9J;7$~d`
zQavBzdmSs*b1GV{XCWVS-Csa`Kz-13ze~G6LjPvy-wgem+3nw~^y9bs)Q!JSuX`5*
zK?m>yI1Zp4;MWeY{`_ucuC@6$bc)76&;k4ajss{1pdE08cEH~8t4a&+XyeUTzityN
z|9r=F(H8DX?^*AczB|hSzpTS=?d}h57nR7-!VzWlS(1-j;DDe5_yHUTa<v2XCw3}w
zs^?UWR>C+RX~y}?tkPNAqs47bD%#tx2gZBw8}9+SIhJlS3be}<sQJJF>C!>z@7<w%
z%I+7`VrP%uZ~uP50UcWh+WRL${-8fE3;lWRz5nMsz5>FrIX@H3hs19_B(Zh->He}h
z2|J!*xoK~E0^#Akgd@C{aJ2paP+pjC*Oqyge{xy9^yS_}%q#^yAJf+V1M)-nJm*k9
zbk&dC*X59Y%Ked8{rVtYe)G)m)lF&h(^y-c+n#N?3tjqhv`{J6_4(FUuX`o0vs!BQ
z&3t`9h4)V}t7LutPO0*b{W~}Jo4XkUK?gDYV9JPMXXh48;(+4lAl1BOeA8)~>zk;F
zn{Hwt=m35o-EkoF_qLS!-A<)Emy+sv3T=A1lx*O>(d!Tff)1bhfzrln8_NUV{Z0Gb
zUslIO!Tpdp_d~SkD%Eow+R}l3l-Bj5wD$QrEBiq|Ip`<n2>s;1&%w{d^m7<5sO<5A
zP``B6uhQCerCf)z^4yt~>vUs2LwjW?{_1Ta2Lv6!58ya}b|9r4P}+3>ln3rtg}h&t
z>Uzw0odO(kXM2S>jusu@{*})ASEv`$^&(Zj7~0bT|1j+zLidxSH9n<`i-B^$xP4p3
z?SsF7zX<6sw6^b%ucNU$`rP8I)j_Q5AzJR!XwQe(^N$(hm+E<iSn0rXS!KMRp{<{$
zrTnno&s?hOK4Uurq;nohrPPa5{tW5B{gO5JORUW&PD_6%T6KftUY_Hg@4O09g@f-}
zIQDlfSgZfv8V3*D*ZBzsN}~g9^8wJ7A6n1%wB}2!=V)@pTW<NH=K1RVgMpv}_yHUT
zw6_Da+$T$&4`%1}PyH@|+|M~_y`Eq#U#!k&1NoNb`{a(IGcEJ@Kt7Z`Pm`_nD^|ZD
zR^nC4crvN4FO{C2aehKu=`-a1_1a%Q=769B_yN1efvv}V`u*>B?AK}Znj{WrpAL^2
z)T{j_=`aVRONWh9FLxZ7eFq1mONT}K8c*xmB9fz(>Dx*8ZCZg}B73YWcs7w!I!%`z
z){3n_kt0W1F5{kKAn0I?9|YFizv;UTcQFuj_|y-EyXI*SHT4=NG;_|{dh_|ae=v}r
z4(tE!UTyG#d)U#Kw@a}hPkh&LwD8V5w|kYZU&V5B$Iotmbw2(Kv&sS2ay@V0e}w~r
z4&Vp29|ujY*QpVj@hw&gPx$;cP`xRRQyRs-?=!|<=y3LX9rs=1Sm(%&m-@V!^aLx7
zHOuDD68K^pCl~8wX|T3mu4kAQZk%@*wH+PL0YL}w12_(#9iY1%=+U6ks>J`@;(+{g
zsOI(PPN5lbSSft<OH`Y}x&Owt@_e_!<reuTW1z4)G|#kwTiY)Vi_ISXeI{Ld@f6$2
zrhhljdhqg7EHvlr5Bg?z*i8%s9l#IZIDmG5Upo*SRqe%*vT+y)I)ER*aX?!;a7lOM
z;<5N`7zjFmAHZ<{?Et@a;HMm+LEjgD!m-NC-g}l0P(8#z&;k4ajss{1pdE08cA)IU
zEc1>BpTocu9UN^OuCbpNFs%a4bwmI6{?Ch;bymlTuZP{eg@Iqzp+(2DHT<7PV_Mj^
z|KHvj6JjwCbO1kq<3O%<;O?`DmDEnxIa+xva==fE%U<Mw;_7g2aP)}SybrKcoV6^j
zLP+v6%$E<FHtkdKOfUxo9l#IF9|vjgdf*o=3e<nn@(;iMhdChVz@i^C_CDM{Y2;fD
z=-4`3JMuXGW|`X<n4*K{jG{w*+s9)d=wOQ<*n3^FtngO<S#NJ+pkwH8{(jRVn<vL|
zK)Q5rEwrW9$Snz2ZvN&vciH$YCpjSK0DfTpIGFnVys@*-#9|=m0DkcKaqwrEDV6f8
zqA;u6T6aaJEia-up?Gv}nctj--@`!A0sMfy<6zL9(fit*ScZY}>QLxV+)n@22QiRs
z9kkKq)>lo`akTKy2KZ>!dZ)42r0shrH#6M3<cFSM);ayRLMz^UpFu4(=O61+eMQ7f
z%qpQBwXPjKHGh1@K}YUmxmaRa%#;($?_i*^bvVEFO&8~b$sAA|9e(nel>OfHyO>rw
z+V>WqypH*u4ch9xZZ7}aUB|{@p}3-ZsUCiH_G6_n@kspC?TJa4b&i?aGW)dlN!Zai
zUsu1&)jAOv2s(V~2OEAE@_XHh_c^7p;I`LC@+{7*2Bzqc^~k{;Cx5z)S);A<|HA(K
z=1rDMw?<RExCdijagN3xMttc0%d1rw_+=g5FMIjbuKO1_r7)_@^q_;A9$_Hp0Db_+
z0ki}B+JWfrm&ZOyyvqSW2k-;wj)T>+CyXrcE{RhLX`6qXZ+@hjhr`}<u3Wj%`|DkL
zjur0ec(dBrm=7EfbO1kq;{e)$ly*QV&tsq*%6h&=OMOZ;o|l&LLO#+SN1czH@hZ9S
z9tQ*+Snz{d3pM@Y{BC1M=Qp$Vt{wjU0}hy02c`9AGgqEpb9g)(SNs_V1RcN+;5dMG
zfL}W>zRib`uY;m6kZm2rx*qJ2-~X}a45S8rS%<M<mCxx%9O1-fr6s>~`1auw%qm~?
z8uZum{5LU>Z5>#9j?|z=hKO3-pKz>lL;mwcZ@!<3#pa}};~wAL_7Vd@2QmF%@seWx
z=W;yffYR!4=%;~qR_=I)foXN<=ltlOV@=+2tZ>!M4P`<#2RONzw@a}hPkh&Lw6Q{^
zZ>Ht^^%w?9t3&amDP?YF+JR|hYxg2m8os;D0YL{A{DAfE*s_-Ym-QEH9atzC15N8t
zv(?+5+jhbn&^jHO1~+UT*Wm?L3bVJ$u;|xyuQ3pG06&1^fcAEP?>HUS!<D)}f1}QO
z7so359XzmVOZfyW7pL}p<ufJn00!EpLtxIHT`n~^%>h9NNA?5So{vhEpW17eWsVx?
zqz2Nd!*-uvFZYUj$vG;O@;%ol?(fI%n~c;zVRV?YU|!0nTXjQpLE4g<Zkiq%rzyGs
zoyOU)9HQ%_E3NIS?LLN}r#81Xp9W3RX*~?vN^6`nJ;;y#A!MK4b21>*d#+(I&#;(p
zSS&Cs78(|d42v+sVzFVd#IRUuScDrEzZ(|I42$K4#R|h>rD3tku=vBUSZ!FWF)Y>^
z7V8X)^@c@+VezM7vB9v|Xjp7AEH)b!TMUb>hQ&6+V!L7SmtnEPu-Iu>>@qBN8y0&E
zi@k=$KEnbV7J9>CzhQB}usCQ~95O5p8x}{%A~d&uufc<R_3zoVf45#;y9^mH=m%e&
ziItU2t@JasvWlsdRZXp|W@=@1Q!BqVwX%jTG|%UmuViv9DU*qPNuf-vC8aX4mK4jx
zT2d|(Ye~UOtR*Egv6d9g#9EVOt8B7tl}(ndvdOYlHd(gHCd*dYWZ5d4EL&xhWvgtm
zY?V!x&Cg`n{7jb3&t%#BOqR{hWZC>omd(#(+5AkF&Cg`n{7jauipjE7F<G`MCd*dE
zWZ9~iEL#<mWvgPcY*kE_t;*NB-Y*Re#4E$%wPEpwEV>)o*rCmbPA07<X;eoV8r7lA
zwWBpiJI4P|LG4(L)4$ug|GUEk?L_S)?PTqgF*<Dsv1wBW4-Fa`nqf%y!9#S0pf1`>
zWq$tmzbxA9+MJ)a&QAWFd?xF(MMLTh`3<4NWt^$Bk!_rjP^JrVMnc%_Va`YhD>&L2
ziEW-F4QuV}G-o8XRh{dMShx74&WJS|`@1s|%lfW!MnYMp_0CAiHjR|eTDyMH87aZ;
zWOhMHv14DmATVh)aH-{jXtw&_xFDfzTUQq(mo?V9AR6q$G#3OWt_G!6yC6Px{jduX
z)2^O%K}xlWo*9t6R&H<xM5}cq&(x%_9@SitquRnIuE-u+>hFrIwf3`JkvR*Eb4B*r
zbP|Wu_Q*90qQQn%$%0t7pGX<NKm(e6U>NaHF!41Q+Jv-c3bn}OEQqh=>X{W;YrDr}
zL-yMF>DiDat2!$ivemNe%ZAKZ*%R53P&W5%b|j`9bIO6hQZ$%fC<hYDPA<uTgtX<m
zav-JI!Am(1AM0D(4asT!Nct>Uw}Ea*2y<TPhQzecwQh(;3t#7kz^v2YccK*-Xh5?M
zbl>KN<g)0UZV1di4Py4VAuxG0*u2*bDaFDMk#n0Jw)CVM0!ynwyGS>rB#S-khQMss
zpyx$51ZG}?Z?3o@A<dt}M~h7&c@fjTxlOcV$c8>}LsHr1XXN}0X0-->h24=<HdEz}
zXt2O??g$JtpxFm9R&hsQ!fJ54i8}(bUxWSu?#Nne-_9L@(XYXYAKVcw_A=NVfq@28
z`@rv{e!xHjihTfWaz|ia)*y77J91=;-|vp3G%nH|fq@1z`@q{mc@T~Ea&sQ!sCG9p
z4-(4mk+{s+S7-7fFwlTzA84B?Kcdx=D&$9eZQSMj2rNHpl`M{JEa8FVwsnm?5STO?
z931C?z&_Vt;Smo61{%=p14kcwAbV`t3l9Y5feOE7$D&IVKwzMO*gi1fs{%-Fn^Cy{
zqSbzwQUHO01~mJ?s*r*ROg;^kpDT!1wKLZXBB`uHHBTglWo+n)<gjUNJ&_m|(%BP%
zfd(}DK%>5%NGfYF!V_6C-(XJ!W?qArOFfa&EMz5VBfzNAV8M1zB&Yp-$P<Ze^PLJI
zmhEMmLWm|i7FY;@iK{`OJ%x}EHu`)aM61RBT?ol(IT8yYIjlu)FC?TTczGeQY-lMj
zM1yVo#tVU^Xizq|2m+H{gP4s)5SVEVX5%6V?8_R|I8+3I6+t_ki!x*D1B)U#4eN>`
zC0NM1qDU&sajqx=b4(2qx2llT_I9rd*=x^+7ehi>iJ8R^7-&GV4-5<|hQzR?4T>XH
zY=8UWhz0YTS{zwxRb~`NwAq^K-pG>esOybbv|mPfBPng+6>r3{ZCm7nXtO69OCVb<
z^Q{s{D4Uk0Bmx5sX!e0tvr8f|tkBL<ND8YutQ-O}uR-#k<&hLtxpM`?s!d;A0V&12
zA}Sy-+coHax&orfB1nF~Km(e6U`c_BNN6iJwIUMB0_ImlVA5!?B!c(`m?#=NZcqt<
z$*;ka`;`!wJQ^%~SP98twVzZ%O0Y~Nd=VIEK(h}tPO6N=uu56|kiFK>!w)IJc9L?)
zX}Tf0AZ<wv(l}|Hrsx868fU{&n^Rj}TdCQOj%{)@|Mz!QZEkHo4Vt9WdKi9}Htg=-
z$!D@oTQ{U$TyZD#-(UESN{w8|FZGXILNgJ@l$mut3t@2VfdV-QRg023lj%>9zLn3)
zJcNPS+dRrohz?Kk^&|)%vwc`$LUpzKr27F0|I{Oc4`J{hO`4S=R0WQhS(Z>=VC|*y
zgn{RxynG4KFWI_QAqcNmdrfshb#mYMnuIZ17FYcrVQ}BbQFRGbB|hwLK&Zc7(X|O-
z;M~CG8bb8#Z@T6L;h7IawIo#Q6IE>pV}^V8{GKq_w{=7ZLe;~GNu3Gx%eK`BA`I+s
zZ(Mgml-J{kmLU98<LrG2)suz?3?PiDvu6Gv!r*_d-WW=#+K{!x2twl@ZF<p`FUXHL
zelC`-V#O<UJD`+uNY$R{d)DeRnur#t>bq&2LA<MBFUJ#v|8~00WI}bb_Y*=0W1K6V
z2qg^O{e$~VLe-F&ZD$ke%k2-FOBi@JA$9>FT2j2sVgpnEXlo~FJ+JVs2Uc~aHQ%)6
zOKE;<sr?9V+iE~KsSI__iQASF#vI+6yoxY5_-@^`gsK|(r$!L!6B<QsBn(_TEbmrA
z<Dby&j#%+ZeST#9ykhhVs1MujTlk(|Y&}k)Ug)eBtJ}}6tK0uYDlxp!)$m<}>I+%!
z?<I^0tyz9QVQ|AfgAWm^-Y?pDlu*C@)Q1y<fdk$(j3h**D}<gS2#@J}{sN(T!Hfb|
z2xD66J4X`+XH8gfgHUy_*rQv7`cVPCcL@WlOc)kNh@NcO`-mWX<sFxJLUreS%@PS?
z@;93Kf-v~ZPnTX1s;2(o^_Ebtz7q6-Fz~hO8fU}vjLkLTGZKdP?p-wtq1t=VsO*F>
z(I@xkBn;NQ&76l&<zK#eenNf5&boqxf%|@q@*+gTuu4T3Uh!EEZ$fpPDxxG|OnA$r
zGK9hH#@8rMX#5lUJmWCugH!=&*UwSw*-^`F?ztf70p~R(oY$oGW2w^<{QW{|0cbx@
z(ei#os{0H}*R!<v9WD7_<^7t`f0U-1(#j+Ddsd?#2>o8d`25J`=ZZY<i)F9()C>rH
z{^DCMzTt%KZ$L{v#BP@zrJUCG?*KjMxQ~N+pran7J|8;Tb7EEQE#oSY=f(OHw-5Uf
z22RMEy}IG~aqWfyH3`F?51n6!P#v-AhMF*@`(>X7guz8z`!pt0UH*DgGeZ50UN8L#
z1Dh<Y(~=N5ot)5`Fnnk7@$U%LgUY#gAdD&7scmP%;M>!~x)7?u_QiH1)VGc=ttAZ1
zp&HPK5FKf;tv_M-*q@VsBve=5Tz4p8OnmH=;e^4f^F)p!RCQ^PcO0R<;L!Ht2?Nis
z44*`Zrd_%>l`y=1=JKJ0>bKPg&mfH1s@*w@Fu31>4|51rB~LV5K&Zd*W?C3w;JmNS
zFC|1RIu=+?816c~^D091{=F;K5XOvn`fxpAaOI-D8wpj9z8$uOP`^BQ?{>n#j+<O|
z5u$vtP4^OppUyo~PpF<;|I$IinEws&I!YM)az)SyLe<8Lt4|Z^wVC425(XBlUiAVY
zit0J)GGX|v`TL^?)tcj(ZxF_~yz#$97`&&f?hc`9Xosl#g!=NKsz-!@_xAL7N{E&|
zS)V`{{#}uz=Y(qaZ)?0Fj5#)LTry$s&l?ZFCsftEmCf0igip*BkdZKOUA_6Pgs9u#
z8`%iMy;k_-BvfC#&^tF_%=C<#@(~6%s`jz~q3T1AI$nhOzvfLSN*Fls*zw|osLbn}
zB@Imd6Z-njVctUs)ZYo(QDNS<fj_f(T_9JziIuKc@!Hz13Y8C`;<41vpKm&R!wLOd
zApQL~t@*OLKVWJB;QGM2@f}Ke?r7;gi&ej-J>RtFkGAntavg6{*OS!g3H~13-f?!R
z0>F7lobwQE{zY4UY#vt%{;G&^u~uIv+uOd`IF4AZ^PbspGd$mKTe{A)x8FDO?Fjf4
zj3*%P0W6ItfcHf3o(SF(k@sB!&+`H24>sQ?{5PJL((4ji$D@HCf*&ej{D{5RF>oBv
zc^t@IFR2-T=Y!%rAGCJf9@_h{$nK|%%~vP&b0(-CI@@<k<K<Yr-UfdIf3xeln)2%o
z@Nd)p&Acv@`yK23M40yvGw+R*S}w8XYnJ-@*W&zylt0f@s(&f9Tzrok(8Kg~g}Hg(
zp&em!o<4Z54)4|By}I!4)y>V5XX*PH(2mN}j>7pK&i8gbClh#HUa|a9taxo5x0!w%
zA@oDGa{hkG?~iGje_E>RS~#9epEp1k)4IU@v$E#7m#UqC`HNuwBKZC!nFq?k_b1{0
zS)Ti6d%qU~?I=8lgy)d(9Fq3&w{TzM;P*AD4xrS2_~73T?%xjbeLA?_gzHVX-lXSx
z6V5x1dETM5-udQ>Z#c2%X@&Qz@P1Xu@B6~_6I?&R^%FhUPsaBia-DzVN;mLal$Wm|
zaTs@AW-*b0miY--^9R=QtG`~+wF${|;M~CG8bjvIjEA#z{0r#dnD=2~&r@pad|=_d
zh525NI!+Z)lFTn@{8Q?4`(WSJ5gkY&t^Lj;^rM(j;0n>p-2A)XFVGIsG#)_hy=YW+
z(vHPMpWF}k@#On|@cl*j{vv6lga2sKtd!xr7dT>OS;Ei0>yg@!!+9U(&AX853kHBc
z2<s0<X+c+Eb&YYZcp{XDtCaaERR@bbx<$&P9|irC;d|BO`4y$#tA_S0<^B(J;CCMk
zy7AL3P}O(SID>fb-SNGThz#K?@3_PhsypXvmPjb~bJUpdmPuvEbBN$}<7<>Bgnj_f
z4*>cB(AN(D+R+Ka;z)am1fG{g^5?LIGLS#Gua@?{Iz@Hm>CRfYSj(@y{W4f9FKhX=
zw|?}?w$%tCg@*b4-ZhLgR6dLs3#du%gTkK=onMDg9kJ?$nlPsOWuFFw!9`sAG$vGC
z{(4h0Lj8<hFX6c#z0dvB?%&pUWvIVn<HjW$>hI@`hu<6OujZ|6aGzuQeU4D)2chB#
zAG3W}VG^0T+WoyMLYTMELEbN=IuP7PNqZj^csC(-fkFEv#jWa}o?1<Czk7=AdZK-v
z7NtKoclhHZRRGr8bBC|TyfVKhREQsiXFd?sl9WZQPgJ!bv~}E1xS8_}_@~VNY2|ej
zoR7;l&u@6XlCiU{AYtIXU!%MT(J-u18JPM<%lU$~{HT-r#@8g7qJ5qShu4iVKNj0>
zXs>VD^JnfmrL5*#=<&|R>mJYpu5(O{_XEEGzcA$&=Ee=PKCS`$9sHe-zk}a^-|+F9
zz$$6`X@VbtAARaa;1A;XgTw2ZDgZ6d1I2!hLVLeOrTU{%%O%wNrQm<A-WW<mQf<gu
zVg#Xa9!nU%lJ>et>HQg$)<4!!j(b|lNlQM`f7hDUe9``%588id|63Z~+dG|{(3&uO
zXY%py2;n=p@Eu$hzJm+xK0LQ*K%9W~@rGjgsaWwUZM=?9ItdjIjE_29XEG5}-R%8@
z5JGeF?<rkxrI(NT`O;G>0G>~Oc^s5fekxTuX6G?B^L-my^99cr;Cpj0Pd%AG0OqNO
zdFn|cZE>D@Gwr*%c2sHYzS7Dg^f)|tPnxje25}Oz^M|K8kW#Lf;dl_|`JuV{d40M4
zVRMb56Dq}zl+q3vs|M=j%lE_3k4Bj9OMt(9x%~xwzN`;iuUWWWV|`xc+6`geM?=31
z*Xf;C8Tw`H-@9TBVa$l957!e0S1#(ikx=#M+hJP>^~;0zZYQMmdoM8WW9>E7NuzGA
zA3Ce;J*)W^Yn(fL_xH=+{bgnRF8DQEe|<VmPQ8E)zXrbrzh%*H)6XXsUh!EEZ*o1W
zj#EXHB;@;DEI1zEcyPqy!O{HQ(aH<sjV*ka-p2R+wD%*o)L+*4e$CeT;9*|VFVAZ#
z)O84aw+5bH=5{96vC}hd%10R7sM^Z{gsKlc>Ua_A|C%?UC}H5hW5<gVqB5^@mNYQ+
zkCpebwB(nTe54<*Y_7kV($7B*ULSMbZ%<J`H$)dSMyGAoeDD-qkhY}8N#m?>nxYHP
zX`Bs9ZBA`&Z9WZ}q|<sBewQXYIQe(-nXJ>c4XM}3$0(z9bo%t~-yol^Aa~VY`m<rz
zFGT~}Xj1?xmbMy!{D1;hDFBDARWiX&pa3WU3V;Hj04M+ofC8WZC;$rBr~uq;*(ez7
z4hnz*pny;cKm#n4l3$1iME-&}u;~*8z*9aELx_L>6V$HA{)@%%k6kf+64+9tmbQNq
z75obdfC8WZC_udehTch}Q<AgB<<m|{Q*@nlrL|qP-Nz90{Jevb3+b)YKg6)lkJ>@n
z!QBm!3~fGivY|gySM5kcsG-fZqcuo7hID1pj@3B*yRG}bJ510{)K1b))=nY4qC$vG
zn>u)CkTz7Csm#whL&z-J?4`fZE2?Nny&=CLbhwN&l{T`CGZM;lLC#1ByFJVq31J0C
zJ0r2plcZs-ot@^4#I~w)oe}F6ztkDAW@CSMMq*jtb<RjA%e3AZDcPox@>y%wFFGS7
z*qzKSNGW#gYZnA2tp+Z&ToBDx{~H%1v~BC^g5<KsS{FoveVFEgz{J&{)M^*R$F3iC
zL1Nm~vo1)fHqkQ!ve(KD&VXpOj$1PzDXd2|SLCR+u!$?O$CmoLB5SSvY*%E?LgQSK
zy*8c1A+<en&4OsKp;fXV*6k-!1~AZoW*-<vd=yN44Td%$?U_O?GC2$4Yq@%6Mb_Hx
zG1-v4c7A#`WXY<|%7$#UEc>z{b5{05b|jR|eVZMLX~&#$Ag~k-<`>F=#Ilo1av&jX
z`K}yDX?E~Z4#da$7I#B(T0fFLOV(|m8xq2tSGplFEp)9LqS3<Fxgjv?H29rp1qK?>
z>;v7mxgohMdZ!x#vrmJVJ#Gk0UJW+ybwf(A@I&O>W``|3>4w14YS1py4JpZD&$=Nn
z+coHU(G7u_*WjBgZb(S;C-Ko@lSp2~v~O+`?HID558RMcw)q)3KZ9AVfnQ;FB$drn
zxg#1Zu$(&r0}W{Qfs9q$5ty(V+-~BI!0gwce}Fr(*4npoM_}}8aN-AdM2o!)c1K{K
z0o6Y6JE<Qq(12neK%3kV*q1d3-R6!Q+2Z%RBPorGbVp#I0nI+}woo2KqrKdm2RW+U
zjm(3DvU?;hbN1Dlya)_5pxFo7X3CFfwWJF95nmg3IX?o+k6I;*V;f6&Ah~T_V-Ex-
zjRptDc_6UQHCTAW1A&1CH2c8O#~#QYTlT^Ofq9_9ui3Ha5(N+#Xdt!^O!%q*lG|of
zE`VsYAEp#QV4wlbKCmjJAOe$5gXQN6B3A9p^@2z$>rl-TiD4NVdLlV&T3b&fhJ|$Y
zL|~u+%|6hmuP2hqT8!{SmdrQU6M>o6;N?<Jq%;dzN!kc7sx(-z-4n@ae;@KhV%vPD
zLWpI1*`^Sp$&LjULSW)*P-ssfB!rDVUkK4^@qZUWa$1hWLP!p4k=qLiX$f9lNGuy#
z$_vq88^7^FU@01u4K9Mfq}L#3V-W;qT7%iR2m<@E1~m>9L10DD4(FoG*!sYtNKV7L
zqDToAvaTqS%5t15iohIGgT$>WB(=TWt3vkLv*E>%P*!4QF$4x0(Ch;P!-^p>Y-xky
zh!xx4zBppR{H7L1)>@Sr#Sv|`rn)z>WIO75BNpwKQQk;OTX@AAv25EG`5@Zt$;J}M
zR?B>=1QN=oWhsfkKm(e6pw;Y>NDM2qvlNoTstzlMz|3oq{AYP2g;nlc0kLY+S64tv
zF|UXU2+Vd3`k$_VXtD^BA285>W*=Bmpdu35%1y0^#Ik_-6%m*;8Z3z*z5ynR29Fz5
zLSXW1@Z^3a1SXFL3m;ZOa#-yrm5>rFQwd)L1{%=p1C5g^BQdN}RzGB~HT3X9O0b=z
z9CDg&h%QK5QiC*38mB3`0G-C!u+-+%=GNxZph-Hdhv9c=!)E_ZK9hCYr6Kj=iaQ~p
z|F)~Q<qcZD1uK>I>h)W*(M_Kg*K<PC-h9y>e&~rFe;=-O8}lV?<;%07ZI%2ZUU9b4
zTsbb+7;`kY_a%-m&BZJ9`e{LzXZgN<jK!kR`85~cwE6|zo1Z+!0Zr>*PB&ZEyWg9-
z!D(e2W}Sh<G`j|5ypM&Zt$Lc<fAyX=cS}u)!BWv&K45){0)^`J%zdTSCCn;^6pekG
zJn|}*ngxdY>@4elg|oyrS6b+ce0`a-mSX#5+hDg0&$?g1tTCa{w(T!w+{aSW-ujq%
zdC7s*s}EtIusXDV9{5A#n*`2Y3VnYhSA73_?{;@i(=!~a>^h@TSO(Wsm{rc(x9vsd
znNe71+S~qE3cq^mf#n`wJ;$um->1^=1<yRh>1nC__HGX@*|1vJ`Lozw+M5rd!h4=x
z8XFk;l#_}=<y+`@a@ERkwfdbKSZWGgj`Z;z3m-eNsdqdE%A*6{=Zdt18+*1^h-=H!
z*b>wG9pB$IZq>|or^RJ#iG1U?H{5U8%XOKyErGL__T~e^%d0@+=*oL9P94X|MSDws
zDqo|m`twIHYdq|Bd!%aA8O)dV>Lyfpp{_H8%8$8tl4tdsyRQFZ45Uj3bM=*eeTIc+
z&uEU^#zNDWPJyY-YHKpy;)J3xULYu7jRJhn144(hCa<s$e+BA}>eaSzn<UN>`KMrA
z$ky9WuB|sLiW8f=<9%M$_J70)#T7LIqW8DngIVQ)!n!i63%uc2<%NM8>b-CN563F+
z&goG0!Oq(pt9<%km1e{9=N#*7SR!Z7;6xX-*tD)SMCkPHKa<~GL~6c#aPUNj){_!B
zxjCf!;gSOy+~I`cfY5VAYVJtpgyPhU$4fox^oCO!SA}J{c`tuFCpNA1XM6XX^L$WP
z``@o|jzX8J6}?|X-r>Yz`uZq#!P(2{&RtGy_URn-B=YfNPH2{0Q088-d>PboGrg|x
zn+*l3hvHiixo%;(DU@2OYjZl~QYUh(Qm+0#LZvJ8^M9jX1U4Fzt>(iBjujg7?WnA?
z)^+A9pO=_bo*FSIbnJ$gSZ-SDKOy<z*JjI~gzcrZ>4~+U$4^nU@?Ia0<zo7BHC^!X
z?98zTFkiCTFW#8%EwdZV>9PGfClu4`YwhnrZI&)?9vaOFO>5H^YCqQV&sT>c$!>X1
zefN%&n|$+cZ#bolzq7ZWs$AiPy1q3RkG=kpKD@Qh30~x!P%CS<mz>ayI5FHkAod++
zq}cBno}PO$d2Gy0PARlBu8a2VthQHKj-Du2WY{5&Ri2*eT`QCS0nSmGz8zG`@0L?f
z2X)){CpJgc%2{Dj$JIG@J;1C{DeqCldT%baA0GN=itqImu^g)u>iw(K@nn6|=1D%}
zq6S*01K;C7?Dx)6#ecN^z=AGQ-*U8(Z@mZ|uKq0Dyk{j(Vr#jl+v8GO-=4>|(pld#
zX-v`6oVB#IUA6Umxcyvo)5IDtIM!%syqVD3h4#1f@xgo$_RR;8UQe-q&)a_P$EN|#
zYF631$BPO#o4n>&<&mHDPCZ_n;N+s%{#CQh^d1N9KI2&B<Q*~G;gK;M>-@U!;OM@;
z$8$ncEdSY)`Np1pU7lfkDOUdMjdxuBqFY9J#$mb0H~&#CdX+xc1uQrD=70F?D&I%t
zOyXGONuQTNO$H`#Vw11FjuOt+`x#pI7rI{@E%8Y;4`=#x())k<?Na|1YW~Rd<&x|7
zW>>ev*Sp`vax-C3rI@eIT*rJlwxjd)r?<~@d|5QC!>Q4W|Kcps+W9&8?kChZB3j}x
z&L2R_yhN7r(e|s^Q9Db<VteUbW5u!WqVHg-Xw;ci3e~PUdOUKSM}$sS=;yQ$pLO%G
zNu`s}{vvgJtUkZudwoPpII-@Br1qcm=|!#Dv~9#s37B=JpH~-to1Sl2EL~XZr)6t?
zjmOQE%sr#GY|W^)#JV3U4=XXjNo_6L4=)>hsWeiXW2wot`VYD8thSb;%RYWKW2m#*
z5{G-dA0AuBNzIqdtF-m0IpuH6mw5|6c(?z>D~>hxes;#&d*EY^Res1^<aA`YlN_rQ
zdp;|%<6rG}^<?RgSdLY?IWGxW-eD$Yl@H35TGtdW#I&&P;7fUPyuQa-OH0q6t^H1z
zrRP%Sw%0h9ck03UPqDS!*ZRnc*$p0IN1;&j<;oQg>-97tKKJg8tDa$2skz*EXVRl*
zm{rQ9yYV_*Z0=~?^Q+`c?rK(;xnB08J#sjyS?BtW_ZqG0{*DuyM*sOI`E<+i`4X_b
z9O5_c-h|`vSSifqSu9J{pPpe>*tGKFtcjJLVQXouf2p|U(X^@kdOgSXvc-}1hhBLl
zVM~;%T$a*lb?ne}_25gGHEJiiXKhpcCT5Lsmx}eCxAry`n-%})*8R%pi(GmZtzL6O
z&PfRzUq0Af-LpZNdmQV$sP%Gsn)rZYm46;>JgMT)yBw?Bv9jNy8kZk%_A-6@rfwPe
zUA=12m{mr-=<7dbaU>Rt>Gh)JcY%DTLoVI+RI|!(JpUpJ%&w-DV%4)#scYYEZXb{N
zQmlLl74N*^)y60G%B5zNgS<ojiZ7f^%_{%+ZtK_QX58XfrQGw>+nfGRn<L!jlzPvx
zO26i1%NPIkHODF|C;K!Wxz1fJ7Hw@OeLJ_<Sg%ncc68dRn^4d5ZB5tQ{;YmS-I(vd
zy%p}nMxNtXWt}1&9$ngbma~`U>eslwz}@fv_$@pM(?Vmu`W|#ySmMqj?5H&AAQamM
zyJdLR{R(E4*OL}3UfS*vRtnGkacl2`k@v8p(ULz~+b_bnUqYj8+h5GMkAYv-VcD<u
zw(QxRfLWs@e=)Z`cO`Z9jD;RzYuRIQskJADJiwON{`NK0tKc1MD+lH-9Ff22Ei5;4
z9f+vcZpj18Iy1U<UGTEr9n2bCPYw*d`aTM?&Wl%m9yx5wb!>^WUT^H0dQ{iZ<veDc
zos-u-SW|WvwnTI7B&-Jol%PQLq|6tRu06!861SXF@W;yu*j`Fimn(9-i#}a2hGU)f
zj_0Q}e375qFIycKi)rJ8#QKf?&hr2(g_GA#?>j!zBg{HS6-W$f*ZCx-jZ(E|w4}5D
ziMH;RH+r0&wB(QWe00K2I(_cD*hqWxJ3McZO2;?d!Aj$k+xK>kz7~yH;e<X5uWf2~
z0|P+^G5mnm<9pqZrP(fzh{bZ#TszwHc=H!~n#{vOaZT|y{SN*?`q^7ak3VM|s1=li
zX`$HTlEsey>#pPXd3Zg=tnu>fgLO{*?->@0<F>Te?%eYlww9$5o;=MIc@MM7ECHVX
zyxDOV%T06bj<xH1aRH~-97w>l(c1j;-Ou>_UZKc0A3}wbx__ro`KCRdGT%M<Y1z{^
zm^Iq_opZUvSD#oeeDSvDm{lgv>Nj^?|Hqhh3jIB9sp303(`n)cpImCz`DmM`*MTw~
zYFcP-dp#?|f}*3^-Q{Sbv0V20r`+>_|61$h!XA@eVAfe@t?SHJJ})tA9D3mI-?rWS
z2lJ)T4}p%YfVJOEg#B#$&ZMn+U9K_aXm0OI9FU(5j(rphUwXdZ$FMh?@Fhtb<BA#q
z(feEP!AjxZZ#&*Ey6YMiis|POxAi-@O5Z=NH?jZg?whxlxJ|K`R?79gK&jJBf82@v
zo3H1JHNMB%{LEQ?`?b1|lbUr3_5FBjbu+i0Q12zu$0OBue#HLHpuOXMg$}>1Q^?ai
zE2FVctQd9qT>aNKu~0PUXJV~?w`_L7Qzb8Adns1_#E#dkLyMckTkEmV6e{0RpGSpC
zZ+5Smr_QywgIVQ~4rOlmh26qJ)828?tiGp`>N|mAr!({Nk^`$(AHuBB(sw__+AdT(
zXl*>S%>Q9;KGr&|n^wo|DQ1n|4~z>w6A_Q?rM2~K?|wTMu8GX#c9CO^4KfZYQtkI6
z9IKS-`-0Z`hppe~qHTYnuG{2}XX&=YbyF%nz+%yoKP;&1+VY;yb51OlQ|(UPe&8X;
zI@2G&mUQL2-N!fAl*Z#;VcN;Je71%YyWhN6@n>2y`R(({3plB0tFBV*Z=COmFYAu#
z)wXb(B#u=|-Tzsr^mfPlysGX0h+~!ZzE_qz{OPIQwKDl1;8>?W{`L8R%a1r)nZA9J
zs^6g0{rb|EQ>x$TpMQv*F5h{M#m<+|@&8dY$AfR0ZsAzvk)QQWJzkvPfYR#FbivEB
zGshmlKzVe?a`Z&GBEt@GtTJ=!Co8YKJ%-Iu=yp!*__m*mZkkx*1t%3*{f>Y!-(L^O
zRCvt0WX=+Wt~Yz<b++`qW#jf5w`yj))8aDb%l+kdtv_})5%Z<F?;;rWY#d#A@5QO(
zIH72a*H%6+x@KQluV+{(JQhB7VpH#UY>u|-BlZ4L^=o1+9jWVqZ+d*cW572(Yr|bW
zT()nOU*BS&X&tQ9G5vn5_w%>)zB7IK&Hc8wpT4<x?Ooq$;<kOaU9+nB(%$qTJR1rK
z-LDPe6Q{o*=*p(9Y_tL&bGfd-+70Uyc5^~gu6~jqOD6AsMY2_$_c-T2n0B&$e=O8@
z2aWMq>opC3Z`f$~p%<L3R9&0XDVI8tlZw`s$JYIR^p9x}Rd*%FD#gCPu~pA``?kHv
zJTnTjN}<L<i5-uv^<wY-)kE>Eh+Ma@Qkkn(hO5=@+`z2T-f=mW!rSXVtc1^lNB>pL
zHv!v9WBF{I&(PlS)YF14&+>i!7_&xu$5+T5-dW!>X-v`69P4EDyUgkHE%f+Tsp4bx
zI!UPU=jr1ywxeRv-uBiQUa0&@{hm*#e4C3Wu+f-oH6KQBKzVg2zt`o}Z)gW*mFFsZ
z{yOO74bEDc^GD-)=!YjyKRl!E=K8TY`ug9Rjc)q1xSlgdqkgQD*6%Y~n=e}DN3%9x
z{yvp{FL>r5rk%$2S}OMbsnF&RC=uIBTkrdY-v4N;xnbVw7noJ9-g;oU$5+p>qf%&p
zm+QTywdo37UZut-)%XQ#^*fcX(N_KWBN!-;4#6cGRtr0S7PHRu;}zj|hf|{$|HaXw
zLydEJryiXD6a#J3!P<7z(tc8n*PK?rpnLO^$2hqu)O@&n<9S>D)PvT;o^Zev9W41L
zt=kW~-5#kLbq3qZIrURZs2idS(w5ZtYAS1-rsx868fU{&n^Rj}TdCQOj%{)@|Mz!Y
zZEkHo4Vt9WdKi9});MXL{5$zf)@kd8)Qc<b^ii+y9hDlnke})wyM$&Uj43ngd=|ps
z*aHP}5ULg>b#^DzxAIw;hcGaEn@9Nx(cwwHo&@1zwht>zsIGQ@uZl3{sYeDM!r(ue
zG%H1@3LG)BETO)@+DqjL1J6Zy`4XaEvUROO5MHnLn(Bn=<i7DW31hY_uKGX1;J%Tg
z>Jq9-eAwTBP=CFmYZJo2xq;0!gy`Gfbj=CEGara*NvPH*s@f374EOH&Jz=nK>xd47
zs)rMkIuq)bZL1MP7}(+7xbB1~ug4KBLHMc0+4~ZzCk+o6Kp0bJ&HO=x!T(&nF_ciX
zA!~^dgvLMG^r9_akRNgUTr6G1idX7(Kq=*rsy);9tkq{U5iL;Fchfk7cvr(-jwcBJ
z?R1^Vgz9GRCxj5jI9EIoN*KKR2lttTsv$Gm&L-5C+aETUFz{|d>;gixq<ERd2B!Ye
z)=tuTUg290tm;f_zG=;u()`v^`w`x@)qrqP8S0!9w=E}(Il47@6=87j-MVWDRW<TY
zjUdz~G>Y6v7`S#=-mQejKcU+lvEr5b{K)!w#poAMAGY1M@IAlSdYnSN&{;25x1U{C
zxBrV&VtApe;kyXc7qZ;nOBfScv;2O-;D&t$A0kw}U$pZmp?>?R4<`r%2fS+-Nr*~U
z2t7v-9@F{!1w!?L83nEo#<bLTjwTGwny}&qq3U3<N4E&|qXK;I5(ZY8Ff5J`J=wDN
z5kdIMJ1+5r>dyI^B@)KuZ#44-VepxsF1;dDP5r~`Eumh0CFlcT;A_`4&W7h1n`^{p
zBn<D}yJ{9fwfCY?*$HE!Pwvl27_57nIS--AzkKujg!+t~bp;6n_x&2>MTmxBm5MOD
z;<Fy!gz7j|L`lM!@Rmts2!q>=uTh@R_$Tyv#$nC}sRGijpQF~Zqn6v;b3xDp&TC3I
zuSxC4Ql}^Q`-Rj3(0-nx<^70M_ZgP1XKC>}TJpil`!%EgC`~t|l}GINtVTZ&`n`nl
z`H{`f6?xtl%U<uP84&vX#kX91!wKErfR=oS-7Y&yIj!yA0eaAJ9|!e7M?FY=K6JF_
z#H!p|##JKEi}fdNANC^*oRBwrb;I-H+6@D05{5q?I=>E~I%3rgHDOHm%RUVVgNwNK
zX-ufP{Pm`0g!&o1UiuRTHd$DwB_VP;IiWRS_|D|x-w~<@m2>Yv7*n=W+s=f+x2K17
zAykF!i|t0JZyjG+OBk3#HJ}e6I?`fWf5PywKPUf4sII=b?oh&*_}D4K34>SXi5x|!
z>e3+ZI6{5Fq3y>L2A*FTK8X-byL4|VVR-$_<wFV8Z>tZUK^U`DyK@#{aK8l~<`Ak%
zo@ls$P=Djiv@pWJd0(AhN{Cu?EU=s~+;w{ARfOvOdsnO>j2ZFt;d;X0%0+!Q5~?13
zJ8TP~etGcT?Sz3HH@WO0MEPQy?j;OAoqML9P(8W+rGtbq{~O|UlrZ?^il7sOs*M*{
zpC;66GsT}J3@lc?>IFg+)pOKk!th!1_eT+`HODjGAdGQ&<9~}Vcu!f~9YWR64pH|B
z_2olVj|c<r?dkE95G{SOK7la&yCO-?3Dxf3)_6r2b8OtWWWwN|Hy(aZsH%A@o3k?s
zpO`BkBVpjWdh=ZgQMbW2vJr-Rt?<c7sJ?iicW%O%=@~cWBMfd-?PUQ%)rTH+ya@Gw
z&6`k^FmT|p<HZS4nb$c>8kqVg^!1&?yoV5|zZ0~h!n|(-e`fQ#K(2ZdD_ybTwY6Op
zDj!0{W2v7%-*otf6Z*M8`ulNO^JR5^z|;c3^?`NcJCyR=(b9butA0&;zG=@NZR4rr
zI^LwNC#lmD{5`n6<Lpuefb)<z=ONnsi?;mOJgyY{RT1N2t-emSw|%p59I;&IJ+tFx
zc)s7Zbe(B$zi;N-5%4P*Pe9%SSQ<|N?}^|&5xgfN@4Ez^=L60kY`#zUZ#*xh*Cn=&
zM*}|uKUBi_5qqy=;5eZ3IFP+wQZoS02gP|lXzjc`wD)6?-A@^tuTJXcOi({`w(pk4
z%dvXB4gLoHX4iE!<<}kH-=_VWd0i;?JJ$P&Fz+8`-Ww^kTw=}FEcNxT#rX*-f1asS
z|59qX_#QW)hw1AIbMwAKJHqBXeehl#-mAlVb>ZKuo0})k()Tl<9hIjYh4Vd}@9la{
zCh)wxV)>(3@!C3WGyOP1=!b0O{QZ>QAJa1bv{cu%a6FklZ-6eQb%Fb5WzBOhRXYRo
z7s32R@cl_L50r)PPs07PJonG`elG;tQFsms&mrMCB<<sG;l9Sf?`u*WK&k!k!M`2c
zza8ZJba1^1*PC#?Nze5roOc}ayhCff^UW9EaAMEX3h!6p{i=}P_l4^xxPF4`Cwi`*
zjPE_<I{(O(Zs55nFJD9AFz&p}Vj=@A^AoV<53J=^f4!n>6O!q`xq;0!hRmB84`=K6
z7tp~m@598Nr_|Q@z`}bA^SvB(oGPLunP1ZQr_|^6!M?2{I*>wI`<+MVM=_(o6{459
z`FFuzpdF-XJb>DJ(WvaC9gB%RxgYN1$@l-@`-|}XMbbzI|IwsbDZ_a$aKy~Agr9xa
zBefxi^FGX*cOlgm3;=%+)*p=0g08~q8sl8?L?{tgDf3gR4i<ZKi<Cz{3i>I-_o~VB
zD@wmt4eeRV{U7MS?>-oG<ELAos_&+82Jzs#<9i<w8NyfIafv5Xch1)=kx=gEs4?L!
zlgg0i5W(%n*C<a2{Q#gJ0Q3W(uO9%kqZ5Y3k@gY^JTHsn&tVN^Ab)UQE$w}Eit5VK
zowahYmS21OWw2IW*79p_{pgo%s}V#B4fFfGYZz&$d>AhlP?OvTg+CuUzYd`~V$}^b
zVNCbSJ`D(ii@5e_OsKm2^`>To`Wd}m!gD`*pZlrZzpe4gP=Ck9jY~Gv-_IKlzc<uh
z&0E>vKF9X^9HGt+Ld6q4X8W+hBr<ii`+HS{FmIuQykAOnAh?f`_C6}`ZbIw=gZ4{`
zTh%{3wVL35_Y~drMEg7~N`G$d@W)B20Iav?4quOXWqwbn5I+jfd?2bNDT`X4sA@xK
z>$snAGv^!dPnrGG%IhXLAD3^Q-|&1TV`p7K!oYpMMtKpUVOXUyF!hg?^961BQ789}
zuSqgR`#cd2uN!54EVkd!Uf;Cm&)j!PS<Sc5<DHGyJ)j3%=a?Gr2Yvy5VahMejT>ft
zTm$$!_&Xne2fqQo;o~=fRnqp;1U~{l`qYoWAH?wohu1Y#09u|0iv1jg_I`^>^+%<a
zOQ`ot!T(&nF_eg;+K{!x2twmLmN0%L?RAmT`!gu5f2^Y%_q3FgmVBiDt~IUsqWwJ|
zwExilw=}%BcRD$tHDUP9<m2BF!gp}tJGd--2N&9Xcy7^vI05bB4aM?PvEo(Qcpag1
z5-J`TA9cFUWFn@z+4~70gy!boQ@Y+tFCX>urKeT^JfHsZI4G(7RH}5$&SPxm`!=-Z
z3!X2)_vT=pdNO|i%u^5Z)RRWq;ym?c+IMs9sM6YfrIknMad_~aG-1UJ;v{D04^MR<
zrCcw=@gUCgLv#1@`f~fj<{CvOREi%dr5!R>4b;n*?}woujWFMr0Dt>(`wRMfSs%Dw
zvv9q}`n=4w8^XMghJG2Y(>t#+^vl@4cf}gQm=RAOt|ttxT-0|Xq3Y4M!?qCWmk00N
zPDtzbUSQtG+H0zlM%`RLbXMDYR`V^^ICuE&@0Y*(%gXp&@N2mK`gEL}dI1}L4SoxL
z%c9?=pHD2j;<Fy!<a$;er-~>^$oIQga6G{A;E2bAqxrp~l^4bvTlg-$jqm$u??-N_
zzpU~7nyvG}!@Q<np4U{U>k#;E4LrZh?M$v?r)S)hk1)7VwU-45RUdlP@gmg!HE%*u
z!oY#Yju$6HWnSkjX<+IfEAM4#$uBMWNIzcLTz@mApMM;@KIYU;6+kyc7c@qvZPtA7
z6kU+Eq{d0(tZ|y63(#qt4NGlKZEkHo4Vt9WdKi9}CObI!ck-F6(=H9E7gwCz3B#YH
z3cSn~kk>o+Z7ej+6(ZB|Im7p5jl)vYTz>ej@2Kr}W>&N39AENHSFCXEldoQD=kf&G
zN^|*9wJNlv{_lFsm*(OXdi|~cxwon@<R&K;h0ZU;C$9ok-c&C%uFpHnIwK0@@8ui+
z1`AE0YKL#T$~T-)`G$DtRA52dR+pYGf67TskKl~~O-3eROBCw7OQ-sfD@zL4TK}-W
zb=RwwbcgMr04QKtfd+G&C(iRe%CSy!&p9kB0DFJ}j-$W^?>Y0X)IE<`CF|`J-{)qm
zg%j$Cvi5pdsPrHnP=JpD7eX|J8%8E@a*<D-#SF9m(9~a?CtT!cr<fW*-0UeJ_4`Yq
z(z8}+v96!2O&|7?Mgggxhf7l`NCgz2O95-|Phdamp6j!s)6&l8GtB+x1qOl+j^GDU
zwYtIAPUhIL`3@&HX@B0v>hoJm`4;Ma$WlB~`Lm;?Blh`QE<G*zE#K{OJugh3u3Wl8
zy7nm$x!5Z(r{6U!H|;CM$Ix=0_w7wTWI@hBt=2u@Sf$X<>y0&Q>3mGvc4NMQpnyFJ
z^eVB*uR{IToK&<YuSyI{Te~84euTa*5Xwg?xh|gfZI*?cy%Z`x*2W|BdlhTbh5h7J
zfbaMt+QJF#ze<TuUhn#ty4cT8#EO6B@4tq>dG`z}g;JlV`KD)S{26Q8`RV}S<e7gV
zz3v}dGSmOjc0Lp;Kc-7d?^=7Gb3u6f-czQ3{w~ySu|1CUdbjs}Q?Br=K2Me_-+Ye)
zrCh(#S`NPTB6m2Y+~*2)o{>5psoPz-{SDFw1waAEQ9$VTG!P%ZelcQQuR4w=LrImO
zfLQZpN)@kE*Xv5CYbXOKAh!Zi-%pB_p4>XpmcCfmQ$oip_IXL@ezkJPXKubVTGxx!
z??LQ%S$lu4l>CZaueQcdTe}YV0R=z-PyiGF1wa8%Kqv)leQ!bg_gd`z4vSE_LOh@V
zC;$pLsse|;^%`4wp_5uD3jO?1>iBFu&%*xv`V)X|pc@VIvk2vfQk_5e&Zi_-IHk-3
zS)$JK?Y9TL!eZ0tM^ep;0qICP4~Zqcjr9)%1wa8%02BZPKmkyIRt3cFA7*d-wCV==
z5?2AS`mstCuT=f?#Y%_u`4CvoKkf7M%bh=QTOWv>Rt1E5-V5=_`~G5h?DKiYYrW@K
zA+5&)<O>uKQUTW5MWM#!2&thVF4nI<rOLm(&j;iR&(~jQ8%HG8d|FC*Zw%#N>pO^I
zzuy--ezA^6spCyQf3CUTeA`h7CyoMAo%d-?M;ukzja8`U%d9G3uHAz5bj-^R^#Jv-
zFKZl@+7DQ5&-mus-1h|d#s}d*0Z;%G00lq+PyiGF1wa976cFlpguU@tlULa1a|M1a
z{B4#$T4z)X#m`fac8yTu?yZeS+S19Fnq1>u?M+we_bfuEXK(rC4$oTu6?6HOdmdcA
z>C4?OSg7=c>Q^LGJo2@~(xhx}KOSjvsr*!=8XqHcI_92FTf1JWnw5G!B&DS%c0JMB
zFTv6B#rOIH!cG0fdBR0b?Bm4V_=Uc2NgtoR`E!);)}9|(+0WAaJhpC6-|jY;<2-Sm
z_fd{@3iaJtbH|CX-?Z5MWo?Z=eLKP0Z`#HWC?!AU{Mgd^JB?=i9CPF!PH3i2*W5Ts
zOYK?ucH7?f9>(zY-mj+*FVy{DdfnhRC}4{MmX0S|x7(67>}$UQV!eN`KW7kwoC*ki
zem2G@XPF>vPyiGF1wa8%02BZPKmkwy6aWQ40Z;%G00lq+PyiGF1wa8%02BZPKmo^9
zKxzHJ&6P*!8fK+`ll$bW*V?%}!9r7LJt00&02BZPEGZy#J0^8}VxKoG=?B~8QGmAo
z<Wj4^`nU+lzcR+RD5V~S8m9yCES~pmmW3Q&9i(nY?M*NJ_zKp4+p7nJ2L(U@PyiIL
zO#xQU8|F%te*7c+?ssNZv*#S>V4n`={0-KF0-yjW01AKtLMR~hdkenl2~k#v0~7!S
zKmky|aTT!i{?s_H7*O8tGTS=u8tngk`|Uxmu&p{cZc6|q2L(U@PyiGF1wa8%02BZP
zKmkwy6p%xKIew|B>W1iov?Vpp8W)Yz6kUK$<7`-Jb82&I^J&l|oz}zfyR^nh<K*AT
zXR=PaG^Ac!aVI(Q?9gcj^(vHHvgdYAXqpo>!}1E#7x<pSQqx?1_^yAx@KL+4`;RbR
z@=aH)aJr)9vfOxjowJqZ^7EnLTld8oZ*zQUE?%M654yi7^Nit#II$>nejz@26_^lz
zqh{!VI~?o$;RlyXo~!S3LQ|;Pd9=6d&C*pbVW}uoz9Aku6<E~l{{EhmuVAV9sBy_m
z`})OUOYFI%{<X-B_pr63Q)S2%Uj_EJ3C(@tXd(v`RtLWI0pUOaP~f8i!y80Sn6}~x
zW}O2QTq{MCeTdWZV@csZPyiIzaDH4;blbZetDLZQ%pY}EC2_3s#k9mCuP*+>DV;(c
zY12;i*!E#a6laM-r3dkV0(=x`oqyfh_A$q?T;!8y2m=a$0%;W(`u*86c}Hbcv&OWg
zF#pRhQ;7zF-(2N{ruh_MiyR6(cQ50YdBasK73C-+qzMYht$?NXKd{}p_hhW;RBQJ#
z?{QxDFc5Tb1V50fovnPsb4|A1XR+K|v3PyHM(7%*mD4H|t*btH8!L@cwWm_~v!kUm
zKDynLE-hoQSd>f8ze+T0yz6WnCo~ry-Wc(H#0ySt%B3r$Yo7uWM*QBTSn?%KZrWFh
zkD;Rn;sUoKpL4A8asFod#Rrl(R=KKXyX;%^&p4rItl1AmUR59W;8%_n8uJYV1?*9v
z{y>$>lE(+IRJ13rN(>toSD{*|O7R?DKJKutRosoo9IIS4Zr9Rx*KTn_Q79id6j6D!
zTDJ$=OQG^(Z9JJL)%IxV7K??Vwduls@+u(o`%<~$%fHci|L!$f3{0zoyx#RObxZx+
zsNG&~#jm|<$8pmHrr+qG!YA|G=cFdz>jkO9oe%2pz5kD~91wJ%!4E#RWTyX<dYq<I
z>6tDqy=#BajGg+d+HK4lXa9UhSES*6%sRz-&n33U34MN-I=(xzR+n75IGSUPwc9<M
z_rD|8IG}Yp@I4M#9T&w)zWLUR+~HX3pJpz<xZ5QE{#Wl~)>yXvh@3eZKEqN`ZhwRH
zK><*}aTGYvzCei<r4Dec(kHCUq~hx_$2zNcuKn$oy{9>$={TMYB~^j~wDx<pHeXvS
zZXR%Dd?Xf|N~min11KQ30_o>%V)eJ&I?|T@c-8LU_5Bhs?aX{)@!CaYldxE%z5KT3
z?@yPGHQqWUVpjO}$;OD0UE(ooJe$@1McLC&F{@-vZ)1KNxM_!Vn)?9LLfYDO$PXw0
z3V;Hj04M+ofC54(uyRU<x5L}N<CMa%<)@cj%k!S&OQCdyct8PA02FXk1-{zy!?T~y
zU*v@1&o#4u-{C_z$11ITj|ThEJr5Mr57dv8^OQ+-{*Y?kTA|WuUAEGzi7jtn*7>&U
z)5&*BU&mt8=tqz5_8U3I<0c1mEFFyX4+I530Z;%G00lq+P=HniyoVHhc>4Gej#kpD
z8{|t|1z5k|{N{4+6{{{sVOGidcqi8L5Az;<dEO&&TOWv>Rt2KD+voM(S2!TsI<PiB
z6RpPs<O>uKQUO};1I5mlkQzc<pny3AX5S7|{jf-nt);c!3!Ap0qvyI==di7`^j+_6
zy088^knk6_mDZL|Y5TEy-#O_1qRcafAL3Z&q}{bcD=eLZS>>R=^Y7}G`7Fmeiw7<Y
zTmJod&JxYFldxVM1sW88a%f8Cr<|iw9@Xhh_0N58R&Kf*gY6}~CH!*ERAWPPKOXfM
z13?Gy12_(h?ZB5?4drb$H*b;hN(v<c1wa8%02BZPKmkwy6rf1~d!IuH6`m$dSjvmg
z?{VcWyRGlV?A>4P@=BHdgy|(WX!@SySYz0QHF0nEJ;bc?g;S+hL5rVbpu9R9FEWG6
zljS*PjVnqGs2%<FQ_MQW`i>#rc%^E0Asxq7V8puI#RJ;^gIQy-yJt6U_I-i*QmR_#
zd;CMVW!~ew?qRX(NvQinh{wKu1LLFHJ?YXi1_Sx&Aog(r@q+@O04QKdfegzlOkd!8
z3Ijn0n*G3%AH#M~02ENZ0#Uy&>g+V~78Z)iSB$O_iZ!mBu2Mp7Y*nD)(DnIzzDdMF
z)7E^z{-6LT01AKtpa3WU3V;Hj04M+ofC8WZC;$q80-yjW01AKtpa3WU3V;Hj04N}z
z0>uW*zTCL?OOAGCUa-I0sB!l=p(&s8kTNI$3OK3)wR~gVKDhliW|c0Zd=eXcy9J9y
zNA(NZ%e~?JxTNT|cR5-pwO&1!)V~(F@g4?(4&Vp&9|v!rY>XJ$B_0Fi)j{fZ)ZX+K
zHM_sR=j1C`DYRD)2oDN?0-yjW;8+UaZj=1`U%iikpo5Tp;8;fylok{K1wa8%Kxqo(
zd^~wayVnmmd#N;Cpgf=eC;$q80(2?xbIt7EclZ#_0p-=<Ly=e2$36I!1B$ByT}L0}
z1{44VKmkwy6aWQ40Z;%G00lq+P(bMl%<)Stp>BvSNLy0lqRF6fnxYHPX`Bs9ZBA`&
zZ9WZ}q|<sBewWrbX`K8!`ApVnmxk1fEAFIpeW)wnvjXdb+q<5@LepG9`>y_d?a=Y3
zu+%h{AHM5fG#|flbA{8GFZre`R=B?L8$$;ACvvvZTz(3^t`nCj^cBaK=HeB4eM~}@
zflH6>!(vhB{6c*4D&QaQ$7!xdV%9kVhfbci^$Zr8Le);uAz`DYtBzu+C{(^79y%2`
z)TC7nH{VDsH7hL{xiZ%CEVe|i2N{m7x_Jg$OFC7CT=7+4+>4;?@lmHSP*@%K)(3<G
z1wet13LGC>rdN;ePGQ!0{p7<TF&j?f^!!*-_zx5S1r8nb+tH!x4$LZ742_?A_tqbn
zRd%b_Z_Mk}o3PR;)RAWG^J}5fgLpszJ_^*>((bJD(lMM|<dbIz0}6lwX%)CUCvU^4
zA!jgaOj`=`zr=b!Yd%HTB8LJ(zyGlmpB!a`G(iEm6*#l?!~A_?PGF#8>5#2d^ng6a
z4r3tb;0S&oRXcm0oA=GiefzQ8e08;8w*><ZVOlx5nNQ8JFOFcP5$&$v-gVS*EIOrf
zXa`HD+_Wj*Z3sAl#iCqtuF87k8{d7qu+XICx=}7&Azk|vkh<TSx%BKS#mCT5`0J<V
zTD;;|Wd~efQgHKlj#WCXe}A;qxmZpp8f*54kbN0CWIT&mp)ub;P{1Aq)*ascuGZ#A
zEEVm^s}jT7+pld?@^9sJS59G8S##MSzjG5$W1%RNj|er+MyUK)8;{WAR;*1I_LEnE
zsV%?1m8sxf%sR`)G>NX$?;r-I)j?kG`k4CrwZo#1_uYe8<Ea%*x|JBP7Yj|f`l&`O
z3LUoR?M6&HA9dP(=G(lxu~O)Dywit5we%RsrVbxlGSmNM39G-S+>}VHG@33gy=$%Q
zS54b~LftpW9gkSg1H|^Yyspm*UH|(8W|jT@YIN>+@E~T5eZ9w*Ik<H@23n^B-{Zh}
zOSMM5E1ktaadhBYFLH-7_uh-O^{ifZ&8p{fpqJknEEVPUH%K2800kUJf%Nkhu=e}A
zy<YzdS#X78ou>}$njCmJh7+2O<H=A`B`8p}Z<x!7&Sx;I6zlm)`goO4*H8vfKyC%x
z&KG<<G2#?vopS3)TlzoV7#4MU^$AQn#U5|+OP8Ga!%LmON@1%KSqCh4IgME%Yxh%1
z$#44WC02gZ_TH1Wb{+Bq3V;Hj04M+ofC8X^PztP5hm8)ovjZ!Iv**okcHm|%HD3y)
zE5ri|fC8X^qbksB%la`r_Uz<@;`eWc9KT&|B4(9d-R?Geo@*Bdf)3yZa2#0M4oG$W
zSaI#oDI+?Z#Xx21VDuyCXD?1ad)jq1);|yw00lq+PyiGF1wa8>6`*~bAFaAUzQk34
zcE7ZgKUT-18uQIce>TXcwDWM%qPMuM55!KZ0&>r91L^y5n(L9+UL9C_{!Q!g0Qmw1
zgj8VXL-pWqUbv_ojY4V&ae)Hn6zFhoSU_y=1K3(_nOx@9$byHkCHkx#wC--B!`N2#
zdKXk7!{Wo(9RF<5YE$u~!`N2NIlX&miy=p_wak=|LB0F;qnLGGczE$x{uReCYuxUu
z?ezHl3Cubd|9Q5UXYfhPD!=a#vn099Nz6K})!$q@3G3xiV0qHb*ct7UI7g*Cs?(cl
zk3lC#7Va2{?Ipb>{Bq7%{XP!lo9*u#LH*HPe_w7jl(*H~yb;PPDU=8l00lq+P=I9x
zq?%71(vf!ldeA$q-Yg%Zkax#ZVA5AL?lpEgg@vNr3g*vWrpS&NFF2_vcX_2sfB)jV
zd&4wGF>Bnp=Y#v-+yTrg?}mQ=eNm5tm{sQLv}tLdrw1|XwD!F@ZTs2U|A5y0h5jD8
zz41xa?m{|_t-uc<`!aOMcowt9N=rtrjP*Q=`BJJ{k9FF=;_FK%F_51QV%HNqUr76W
zK`uQ;;LyqQww}R2emaPKoIw1b04M+oSW<wN@uH9q#~deYZM|FaW7rM~fC9=_pjgr9
zm+v!1Vxg#f#po(w?C9UUC%bOPLX)miLT+qTz|wp@e7D=057-|R00lq+PyiGF1wa8%
z02BZPKmkwy6aWQ40Z;%G00lq+PyiGF1wa8%02B~HfhpV1E~rrNAXYks(CX7TUNj%S
zadU;!n0E4ge`zV)r&_@Opa3WU3P_<q&1Hl9&P_axS*4|R)!yw=)B%lY*!ulnsmt*9
z(ZqWR4bNg)D0TVkb&bjI{`dd}f)3yZ_8$j(dj4;EgA&IvP+lFRZb$7+&))BI(H7ob
zJs><N01AKtpnzj3K+Am?<b$sJVtebuv5q1rEhqpAfC8X^(i9jU{HE8XLWi-vRGKbO
z9#8-j00lq+x)dl-=yosnu(KE_uMW1phll;EHR@gIEVfk#x{f}`4JZH#fC8WZC;$q8
z0-yjW01AKtpn%d9nB$jPLfsHukhY{IgC?WKX^JjDr*SqcwK=u9wfQt?l1}Sk_+47#
zq;c}^<TF{PT^dp^uDFxZ^`WkO1NFPw&Ok@7&@@-j>h8xfM=U>zrKY+3@LfN>UDs|4
zsvpCA$v0iG!d+=Rwc>aE4|BHCTz)Pz8r?jn=M|1G&BZJ9`tnP(PhGQb$6`_F{6c*4
zDsZUl@waZ|!>z1yz&{_h-S2S_3r(SFr`qU%Bb^2wz*14Dd_z2RD)8aQ$K~U+$FS5~
zo4;nq8=H?|p-HF8kSo3l#9pbdy5V~S1BKOrZ+$>GPyiJ8sKE3-Bd>06cNDYEuitN+
zHRI=FI6Xg>6#fGRK!NApyWfw^w+*w(>Geil^M1Dmv&snvF9z=FumdZdLLF(bTk}<0
z{^TgOM4{4yct8O@3f%Mi`!FsM#>quKd4@2c04R`F0lx43<PMj%6y|?bFMf0MiNi;*
z&@`VSY>`6&sr^xoGD4c5fZPfcYL$>vTly#lI+hNr2S%*?vC|<81RWf~52R{mwd<!9
zjjg%|%gvYukvD#QugA2qU_`fe$>$GXrSaCo_H&Y39mZl)Dt~sgbl(5rJ6(O|Fcyn)
z>3O8~n~VF3&EkaS`w1?~Gfq2-<)&P^Lb~=TVClVxRNL(<#mCSew!SYH(&{nCD)FWN
zvv=n4S`PpJKc`I_ry|)ZD#;q6D7tRR)-AMHqC~_Up+1W2iAqA2yGRm|lt_glHK%=(
zPWxV`b548GzR>df_Tb&;*L;j?o0)6oKCgey>&#p;?|IL2mTRuLmV4&soBPjpY}Fed
z7<u{FPVX~WN-eAV`OL88k>5}d2t%Oh?2iv`uP~kI(lEZJV&1UlztgW-_F;n})0wS0
zb?rGbVz*6avQ$<d8FKSWV{4u|m4#JV{&=;Mv$AZBnhm~pWU1G3kuF*SHLpMZ;ljnI
zFx&OCrz`$3u<aBEJ(feXz3acFH$JD#O<iY9Vz%kzMkBj#KVuS;t&=)tx4+>22~4*h
zIPvHB_I{bnbWOGW-zEj~?rqR@3WM6o;lDH4KmXVM^^SXMemI?_n*MWb&5dpM(Hm`>
z{ydFYr{8?Grv0k_`>u)Ys))bp9(P*nqFuW-n96L`!O7*lhJ8Ab*`_@g7y97gsuLO1
z8;79dK=*T#`K`R5?Gkmq)2mN9`@+2c+Xu8<!`>H(D!=XD6^pw6qI?hl0don+{!XKA
zdqYks-*H`Xqhq^%|H{xx-72qhWUIM751xU*{}PaOzc?y8y7q_ummFe100jIJNR|6N
ze(Rm;>1BV1R+T-~_jw0@v7^w;NfTJAsH*onRM{O;J8}J~d;jlwGt1QW@ABEYza93u
zQQ7%L=ZUcE>!|i~QEoKftyaxl)K3u)0w4eaAOHgYM&RLH>v!Hcehf<$J#_2fxif~G
z9Sf@e#tO$k00ck)1l$R{`uQ_G7FVC)$kIY%AG~*1?E#LhdU@+MbvIo8IfH@&`T@p)
z*LlEwrj?W8aLa{bW?nyo*`l`YbMt;*v88YM|FoRJR4eroUEjq(IeNcC;>MfW9|{5>
z00JNY0w4earV`No{+ZwMOyvf@)Dlqjyqqq(Pt2P;YgnZz%vQyBw2u9bmUbSNe~&`-
zcQn+_KCo^o0qef&qv(4$wsRPIa_#1;Mowq8sHx)teu02Y0_`tdJN2u%zd89$Wio^<
z5D*bKtzW~umETTcp0!8KCo<of%uaLBH+p?JX?y=E%&+Dwx%SLemrP|MJ=FWOQ!0Nw
zmHAb_@1srKTet4L|Cz>Y*CEv!&HZ!1GzN`^gYEst?>I9xea_SizRJv)#@tjiPa-}V
z0&^c6UZ_ZJo|9iS8tUejnltcFX4hHMSy(l<hX3xFsq-uR3brl<H*gDbJ;V2#f1fo{
z@2Kdx$JDDSY6Jox00JOjGXY)C`JkLj-i*ze!oVRK4mOX`AihtXU#e;4i)OH>HHh!X
zla{9I`*FHh=T693xVimSN0;iVSJdSshV`tNbNSEAHf{R(#|2tW9Lj9f_}PDS+kNJb
z%vQa4)64hd{4k!`uFF4JcW8H|@ys^u`TC-7`*!_;*{)YC{I>UjDqk^EH0t>n<>wCq
zr#IQ2k*GF}*`}iXX|M6RTG#D)2ps?S=SEbwQ`Fa$DF3hP`!{d6`S<6_&S21B98`}J
z*ara+00B1w{Bsis<M|sI&)>-IM{fNX$3XxDQlG%7bKkwTX1NJWmZm;2vue1m#{4Pl
z?2jDTYE~`bjb8$)zL((F52ObH5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI83ev;_w}ImEhe*6R~fASe;cRd4jI$r%1mauHoo%Q@dGzcWU}@DXMz7g00cll
z2Lb-~du<CJdZplJg?mn8s#HfCm{`W)_FFF8keJTAs_r_N_IoC}>tFMcI*nqtOkz-Q
zKtBjS4*dRJCelwo?(5H{G1(k+&!b_>6aAitt?^-bAU_Cz00@9U{vzP@IuYsoU03IK
zUioVjp|&6Z0w4easYamGltuUKx%vYZR#S}&>H`8G00JOj76E)8(~RH0X#aZ0y){3a
z&Y)^Jm^J#~4G4e$2!H?xfB*=900@8p2!KGU6X;#z?;0jMBwHn_#tX-b#M8Sc8z<uh
zj{HqzB+4Yp#?!hc6XlMCtHsmf=?z;{>6T0k=uvM&rSw$igZhC02!H?xfB*=900@8p
z2!H?xXd-Z8`wP2vcxRK7y{6*f^jT}6;E)piK-2I+F(3c}AOHd&00JNY0+tcTDE;Cc
zPW{C!sGdA`#i${xW;^*!Eo)}D3j+U2U|8ca&-5y^#&J{HFOj7^e(e9)hb>OX%%0is
z+S(I}O4EO>EDnMI2!H?xfB*=900<}}P<q0`-<EFL$D-Eyx8**0^N52iYJGbDuR}lS
zzK7Yajn_^|e)sq+mfuv_K!IVc1a$S=ZQJ{tcmE$HGI{;IM91azil=1;RjsfhEUf(e
zy1mlRe2U4^2&<yC==;x@SgF_B2OQH?D-;w40-6Xsw))1g-R|GTOjAwOM=_QY_@U*w
zZ;z?8(J@sY?{dkr7pHA;Wa%?Qs@AJ>(FW&}E@u=SeEw=_dJZdB_VwR#oUppS<w+H)
z-?GAyrM*^Ode#%qZgF&}<--B)g8&GC00@8p2!H?xfB*=900?*_(0|Lov$iI;JF4{2
z+p11_ed-QJl~#W8qOS{lzS{|_FTHx{!0~1FI#4q?c<jzd1_B@e0w4eaAOHd&5RyQj
z`<?T$ifw19s;$1tJFVyWhnQati51yF00cllJAvnywHiF+!f_7iwLSMt`sJ6oj+;KV
z^Mo#!wO;19>Ej<|oL{C?wqvWFeR8?B!|JVcc(t`Yy8PDOku??A;ismj7r#E2k-k^<
zj)^yB=Cb^%(`s~!jXrA`Q>7)l+;i`>L)J2pjxYI6r(ru+GOt>A(`#>CcJgxORYm2_
znRQpGso&)>uUc-%ZB=hsyoT}W)sLvsdFARg-L;?eTrr~f3TCQ~ec|!W-`=%_@!Eas
zeScj2?s^ur#;>?{SZ4oq%vK%I?#?#dYv(cOu^cp=zEBJZfB*=900@8p2xuelMAxgI
zXi%hJrkh^)ebGj9_7utF^_(xSYc?^raHdFKoBv#e=N~JODbiD3KG443&;prmdfRzd
zO)v3AflSg0Z?QICjw_fc(!pn)@Y{^lMKj&>j`QALf6n{GGI`b304N9qKmY_l00ck)
z1V8`;KmY_l00ck)1VG>z0<ylF?`G$<^t+z;?w%iyK7U2eVUdpC`)A8(-_Li33@M#8
zqS^A$dd3Hp8otEVIZ$vgp&uMG=>8$l`I`Nu8%&tzK>y<K4~P&20x3_R{v#8rTr^=j
z^QxabH|5k#)ps(l+Pi61qt)qqnO9v^rCG11&iI4zTKd=eQ)lkj!NTe(z5dg&?X=$*
zY2M)nf8U?Bk%iR@S2i!)rSdwaT4g`q)#kmHO%9h^&va?Qdmid@#d8~3s;KPtShl0#
zlB}F;2K{#qYrbo+bjWtbRB6gj{3%w!GxyDX?z@sR9Vj>)=?5v+y-^(y009sH0T7Tv
z;EfBH_gFCh59U{|y!MyomEYUVylSg4O?Ez4V;>W#S39C~A60s6Y5P^1nOEK4@Ya_y
zTC8EVYS?`W+w#l0KO@VItoOXDUtRLD(e0KxP_-PSOqMVL0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0x3%1-a5@Huya0lNa?wHk$!r9`4O)!%5#FMuHR!dy_^*7PS#cHm8}2!
zENSyRC%>q54Frcl00ck)1VBI`0scGDD_(Aw^~($M7%Bgq=FWG#bzb$uD;N|Uj`RbC
zT@$8300ck)1V8`;KmY_l00ck)1V8`;KmY_l00b;0u)SNcbyq&W#7Q-M>a;=)oc9(v
zsjjl#JIpV3EFGqB8w5ZA1V8`;KmY_l00g8E2>X2yuTG3QF;?x8rGGk5Zyegzys^xX
z9}YQCa4@kScpdjh2Ld1f0w4eaAdor)KIw8s(ZT1hcD(9=uMd4)f9L{7mcCQ1VJ+6<
zdneV@lpZ3houam53Qza}0w5rtz{GA9?wWMlTF0--`pz!wWSn27RJP-ngZ!QVizXBB
z`ulHuI+N>^3clUEyUH0i-@L;S=^Ytgm3ZhsyBu9Q@40UW-*?X*$4$2_8ga{jg?pT^
z`qDd}K0Exh?T#!x_2w(r?(MkWfr5ic{U8<QD%1f4KmY_l00ck)1V8`;R1?TvQ}J;6
zthJ6vn_ky((#Ek{S>v@+lHWZ(3knYC2N(yK2mYQ1RL{?{?A!YOtE~FN4hVn%2!KF1
z0&dTvtu52N`H<uKI=1Wc%UTT{a^W}!3J(9)55kQm|GA$TQng;4i#9lJ>i2meq!0F-
zRet5>zvm){Ndzj~J9^)jFTLW}swT07FVPdA--)PMsPXJK8s#w3+gmp2Fl$y0gMtJ4
z!T*kf=tr(U560c_`Sc4ib~sV1Kd+G*1k@7fUE^>2$qva@iK_7;@#Es@-II-z@d8Kw
zCNdIb5@q9QU6YA&N5a+O>GAZ2Evj@&CI<AVx1mzHbnm*~e&>@j*R5r?>hPj<p1(VL
zx+BuCz>024e|%Z_$xOD2{JMB_#-f5ZWHa3=@+W9~kqN^({WN^B6I6qit1931r@Vi5
z#sfQ<UlsYYYUjtL-s!fP1yzw<+40K`^<Vng(`y`6D%)R?y`cRtzvL5f^BXUkhi!}h
z=-Iq0(u%+5$X35RMd?Rg@^<@YZ&}E6smKq+|3yG{KhbT^^zrGppR4Ydve$#U>|N4-
z?Ng7GTkq)B|129c`RsoCouE3S({J~0seRai%H<$yoV-@&!jmo;y~VLzb&m_#<*8by
z>$3ZM+n<&^cm8%qw#w?~B0IX~Z`I`lT?gpiui(q0>u6E?=#CG{F=#&B_48Vu-*ua;
z^s3hh{fFFiVzEaLIJT<n`Bk?)-PcD^mFIPwxux?uuejM!b)28igN9Z4=hbdh{lxZi
ziEq&BLs|LAkDw}hzgYJ5fyiFi^?guYvh#VJ=iJhre$~C_wET0Yqe^A<S5xh%UVo{w
z@3lXwTJKQ#@4j@~q>a^zWs0=_5343FEU=%6^z@r{E;|3cZH$!cLj4iYy`SaJJ9Vi=
z=Ly~M>T2lEa@g~u?fL!D@}FAuz82Y+y-p^(wyN7Y$n@j3k1O5tsqFGpt%Fk4uGg87
z)@`<{-{<Ar($hY_EDgHPWO_bb=OeHE8R`5HkUdZH_IPh}_2;?M@1A|q_ClE=-PN?}
zeIE@x#Qf@k_Ya-EFY~xek<RGarSy;U{$#w0uAjZe^YakTHNS@ItSMTxa*u+UZYpZ$
z|E#awVaKI=m|qRb+mwB@p#Jmubj|;&%8|W~uG^mMeIsgL+1IbC?TO}5ukRaqUH|g)
zu-7ngRi&!NM}D*G^|=sP{aICi@M@P|ujvl+eumv=(LMjlE>HG-K%Q^Vu&n*Lpydb6
zC(CcL6Vwmg`NEDPS^0I>o63{rf7EuU-`S?J&W|EH)Xq_>s`Zhm-m>CV^=sYh2P#kY
zd03S_@;6mf_VuT3d#c*m7Q0c;pQ_4tn<rJ3tIDpb^Neb{ZtW2ExUO-gYW=0EzmVNj
zy}xaGy=;A7Cn~?Gzxq{G;}UlN`reh__An*A*L``U3-^5}Q~0Uwcc$F8wbc*R?Zev-
zyRJn3aNni!dOuZFI=(;14(oc3NmV(f%-do8@OppWZ#voih3v_?-X?px)a`j$I_{UV
z?DRo-n6lmv>c6UaLYG}#`#3ybS^Y2UJQg-TUoSrF^?r}(`zn6@OO^X9R1d%PmzAFH
zPrBv>Q8}ibM}zt$YaR+~$LsTNqI5y)?>C>Qd|B~f=VMWRQGZq)AGTlU&M%9Hs2x<r
z>*gf8JXKs&+m$u$W!d4+m$rs&-3RkKf0%Op$NM2`e2dyKXutG(zDD|Ef8Rya9=62u
z{Zv*sXn&NI&+oipYI?8N`=+JSwGX7a99iQbYCEF!5REs{`_-cOhrb=xp-8hLnQrR#
zKCi0dqW8MJ#z);=QI+ra{b|$Ei{2j%8?WmACU2MSzd(=u{WPAFs<7(oTh;buU1x~w
z_+2OY9d~a1UgVdoc(?My9uM38rp>pu`4g6Z)cHm0INkC2bsr$Abs4SGRjaIdRF@sU
z=NC#}s!Z#}&W~xCrl_j*3~!hEkF0*I`uba!eO>)kb~%3MeQJMQ>UX^uHodNTiOTW1
z52veNyOlG%QtRbAGv+z2>hinkvgQ-ABU)dKDn7T>N!@>Ucy+6%=s1<9T1DlDji={{
zysF>#vc1}+bbjmOcf9lWRov!3zwPSRf8HNaSk#~0;#K)WeiqAGe8M$bmoZ-X`OPm!
z)6!FYRVmNQl(66NWm<Z-@e%cLRqgCHuelvp<(DkGVfz_>U86f3)_>&BSXA(aY!;M*
z?CU7q_JY>ilzeXUkLkxn>i}7Pd+kr;hbiqE)DKzfJ-_D{;ulu^mMcmh^?D?1`E(tZ
z-_@-@$?EU2>`;FUdw<gQ{KP?)+ID}F_gDA2Sao?(-+%R6zO40xto}@PRJW6C`?BUq
zzjnOt=bM&}|J@p|VZZl7MZd?w$J2dRzsk>Nd^l)7^E-}HCB56a#?<4cjjy2o_?>@{
z{;%)X$>!j-pVRr$ugdb*Z@E#W_qyMKbay^EbKP1Nl!NT{4eEy}`Bbg1W!X(tJ7M>k
zOwI52e!H&pskR>%bp2#nK3V5^*>+@~Crq~|nrC(SE$jPCQQ6^tuPiJ#zt5T3l3v$*
zpWn+#wey{A^|8JGnZjE%j|Sbh6y@`~eh7M>UsXP@`_r=0i8uz0=l7)%{_FSH{mvt%
zt?&K%W6C;4H1BwgzrI9)r|Qkx<Cvyi{YN?w009sH0T2KI^9h)GKR^~azt?B7(|f(&
zZF)LW`-z)hey=0=^s@RP?0^6WfB*=900@8p2!H?xfB*>OZvwjR^W<;J;cG8@Unsms
z{w&CPe!yR;NbQvXzwd-_|D~Tjy~a_^LH4*Ndv5!ZZpWkE52kwLS9NP2Q;(}2FCzP1
z=Q%!|+xtu+&TjeWxaj$GS@E*Yt8RAujzhQkL&Q~fyxVhVe1A|CmR&E{JNEbZqUNFN
z`#FA}FQs|NtGaP?`yPTU4sP?K+kEAA+$#^abRs`h%_Cmz`kk*$P0yc4yw=}ux}eX0
z>(1x5UP06I+!4;ZBCYs)4$mPdU(@q7s`qxsiv|^m*{bW$%{k-D<BP_i;NYhpyioeG
z$)!6Mi`lM0`#nD|`CV_SPVe_SVxshXJ#~fo^|R`7`SQGmz4|5EkI^0PcAanfakuL(
zQ;(a%m0y>64U4YdY>9WfK6N|pcAc#HxM+TM^V^i;ZhXAXqo$<O-S17^ckt?$*LhiX
z`+A+<qE4sk{2-gB$e!1JpsJox+2y|r#D_(^Oo{h8F7sPDQ`*PPkD%Xi5!sD;9yFzV
zuk%^ZbRzz4@usZjbot@NgN_Gn2g*lwqMqmYbu#&-OaHTM(B!lG?RTJhIoNvt0Dk}b
z{hzmce)v+e#|p%t(Q?o=f66N7m!D2)x8<2nnQV3a9dob0ckVNtEH1DE0{=!J=(?0&
zcc~8lTOGn23wk|hO1>GLe!G85?Zb{L^~=_j>(NyAL)0#+_Su8Z6H(>!J0JP=Gbm@j
zewmV9)ZbDqKIroW#35+E2%0bI{g<HSdwpKS)O4wSezC27v@dC!2AywGH6P8Rs#I40
zkmaB5aVE=eksVX6i}Jf4y1u_*N;#tbNb#nuql5Yp)$<y%%9S<mrkWk<pCa{pU21B2
z-RC*k<;fZcursXFPs0~GA`YT)pekPEhi&oseZ6CP{oS4qk=?(8?kDNWC%e8Pd$x^Z
z-e1$&Bk20omVC0lGoi|ksr#XR{ZhpTcHcic<AI$lEC;{siS!?rdZ*iFCOQxKjn_5r
z^Y33o73R--y2G~ar>oiz@!MZ*PtV`S>|NvU8YMd<TP3Q-kBb+Lr*}^_PR0uy`J2c{
zlu4A0r*%yx${h(;i>Jra8@8y@Etwe5quz!}>9+H*&i^t+`@DK(k5x>zia^wV|M53V
zp1p?YR*^qJ<HPzNv|LsBTEFpO{KFeMGrubGr~liHE1tZ2h!a#rc4f!=t*_tovi*g<
zXbIf0@vcV0AINiT*V!4LtbeP@21mBaYMl-%cHHq)i8)M{%JLg_%p{<zAMxe<IPUw(
zTh`_=H}$*zGLtgA3QFLm^WMGd`NFvjDwjjh_CP)m0D=D!_-M^7TboT^#%x#q{7m8h
z))fDN00{K2x9H&B<1(48T6OGkcRhRDVrHw_KF@mfU)D(T+Rn1dgB=hELcnW3Rh2G?
z&&UG;AaE=JzxRPmNq=lD#38@?kzUh_OT>{V2)x$o!ShD<n8tKz6m>+IAP_YHuh&CJ
zmm2$w)Q*?;{N<-#KilCz!685N16}iM*yrz%Kbq$-ZR4)1Kj*h{RNq&Kis$`hZW)zU
zaFHWhX+06OM#UB7h9eMlzoGKO)#AUoH*GFb{;YcQ99uPKV(GDS24*|9s;cLPsh`-Q
z|Lu|Z4+Me}(7i4W?rCb~3_IV9NM=s>@cQ}8Rt;N!KEJG9BJ2Dv%O9_H&dQmRD0XP9
zBTK!Oi*(Tv;Ku>NskNUUZSVST>9+PORp&=t_pkYKRMkheU0eH$$ZvXXAFB5~Sy{aQ
zJCptMf2y8;jN0x$*Vf!vug`TMotp2csOqn(dz>o2b=lRuA0w+gj6;9pFz7hY)erM)
zIYHYc>U{k50m6EpFP61F^=c>T{)X~F00i<60o%_fEjRZX*uLZlX1mI|ufyBPKRp>W
zO$7q@UX{1+Ri#2(qYfYtH39y7X=^xYj;59G^?Ez%bh7Wu^7c&mPJk>wY+vWw;=k?J
zg|_-1^!xUz@|iZT!yga;0T2KI5C8!XkVPQ-aHj{pSh&SW6|K^|+Y_rMrN@G*EUvHv
z0w4ea`IUg|{T@}<`MT^zz2A!R?R>uj?UujWP1pIuw&$Zo@3SHPpWhRveuVcL{JqzZ
zD!i!up&$SPAOHd&00JOjDgjyV@v5?8DmVD0mVmDPLS^Mx{(F?No~MMJ591%+(3$z=
zVC(aRy8Ko<`@p)X1pJO8qz^X#sODzdKDepl0e*piOal4!zK%?Wumu7l0;=wFsoH<h
zWmgp!-F9{Hl3k8&PO8eY{re88xahX)b$u_IClMbFf#Wy4{OD61c02i1qoHnYsj{An
zG`EKT?wRg+6XoU4=f=?PcD4Jz&l;(BRMbCHucoLG2!H?xfB*=900@A9Nd!deP~Gt+
zF|o-P+4q;DuKP{T4XU}W>0U>dMqOWB<yX0<VDptzPL0{7Q}63}<~=#5$86Q-SDdu@
z^<OK;Y}Io+6>2ehX|<T`TDw83jbH7$AZDAMIJn%0O{P?j*{;o7AAimtbuNjSqPpf?
zl#{;+$X=i6vgh@A15r9%tuFhyo&07`bvyCjw?{a0!iU$-XSzA)dXFeo%TYZ}U>^iP
z00i6!{5bCW%3IdvF(^2g+z;IPF^+=(2&6s%+1~?6ePU+SP}P2uS+#^WehHvo_$7w)
zAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaCKK55`b)>ZQed`|>MFYp
z#(aJL`>)@V7BgKhT=1U@A338?Ot#AA340&_0wCZ<z_$0kd4GrOo3d{7l9OVp)QumG
zM?*l?_floq(M<u*`Qtf%lb`cP|3v=`(m!?2qhZTSwf9=X@<4tN009sHf&4{4)q5YX
zd&|~l)0eUQ${~M^BGeWHKmY_lAk_%8ZQXZIujem_h1FE!g8F~}2!H?xm_^{(`D>;v
z?3%}*(Q?3jG(GpxOc{N$zl#KWjVqqKdx!%L|IERZeh)uD00ck)1V8`;KmY_lAU_i5
zUE}XsCp#otC91}Y#*4+%yC)kb;{}fVO=Kj>B+ACqx+W9lj)bel)8pw4TU6<mObqBz
zZ$qW@{KyP-1_2NN0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T3{dz><N}IyB#Yz=488e&`1V4nepC0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JOj0D(50di1@k$V!%KY5-ce00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&ppn4meGlv#_0u8-1qbv4j03;(fX1l`MS%bafB*=900@A9
zRsyx|DO;}TTfaHbzc^@xkHSCz1V8`;KmY_l00ck)1VBI`fe+3v(f`nwYZx?taZuQ$
zU>XEK00ck)1V8`;KmY_%mVn>qh)hdw`*%QA{`ACM8Bb&}C^(qd4^npGL&ZP<1V8`;
zKp+AFwy#@*`X2#m6bS+#00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w7>0fzP_kotM_<tQZs=&<`*UWX%JHPIqt%1V8`;KmY_l00ck)1V8`;juOE4
zukzRTuWW7aqa7CifB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfPh8<9iI55)c9X6h=o;+E~6+A009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KINd$V=_}fsjL$Xz(YP?vycs#v(vT-tA;K<)ZMxsojY&@-NGEweGxLQ0tp5CxU
zm2SzzfFAWWR7#iR2F!o}2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfPg;&znyew%$D;uu&A`x=2N$fJt>#z(r=!7^!%IO8PA~LfPRpF$AQ0D5~)D|1V8`;
zK)?(F<$h|}b=`C6F{m02X0$TA009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5I92M
z+XlT)8d*BmfvVwvUIqdn00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00OcI)J=P%@qz(snJ$&(H|(e(@Sn5xwtIBe
zKF3zo^}DL7%IR{}hh6WhaYD>?z3JY;S2X&)aLhKv?~CF0#njBKW?M~_esAl0Z10tN
zarki^XRx4ZcJG3BAOHd&00JNY0w4eastAmGZSxI(j@;y!s%;D0ym;y#yPQ<l)60#&
zdUe;mjxJTzpI{dRKmY_l00ck)1V8`;KmY_l00ck)1V8`;@)H4F&q;<YCqK0b>IwoN
z00JNY0w55KK->N;zB#e(Z;oFLRvL1F00@8p2!Mb$0;=A(GR>|x8b}2KAOHd&00JNY
z0w4eaAOHd&00OB&pv3WK%sF-T*G#s4bZ)zfg@4$@bZcs~3TgoYAOHd&00JNY0w4ea
zAOHd&00JNY0wAD@zyq&7TKD9G%bBg}_c<F?9AOs(KmY_l00ck)1V8`;bP^bR$!DK`
zmo=BER-JW1Ss(xcAOHd&00JOjB7xpD{`M-_A=xTXHC{YkBA(to**F<5aO7_yBT*(%
zHlEfsnJ9N8TrHj+PjA?wO1ETUK#zJGDy2)W)7d3s1OIq=remx2JG02(*~M2ek%k4<
zd-)Y3F8E?IldU4Zwr1=YR4!w!qgzG(1dZR{_l<X&E!oI|YS403<r~_ve6O^-3dH=X
z$e+#KKJQ=isp7GqDzYm({?hVmvkpJ`8&jpS{T0~@+7I(fJ`p#+@t<7!>F8cp?PRv=
z<yjSq&hNgB$yUESMd{By?e=Of&tJrJsmKq+|3zSLnJ%~V8NS@HRr~!mf5O(q+Z^3m
z@v$=}|Fra7=B9t1`Tl}4(x)@8s=8nPdha{i=3VnUvrT2MW4+p|ys-PF*^lS3pxXSd
z?DpM?Z(vZl9Au4?%#$x`JnE`-%yt!BN7@p<sZsTvSI*C7wyLW2wyb_GvZKmx-F8LS
zdB68LuwhxP`OL4j>i_k_=RC2D=~7Yq=#JNYp3+@EujNH`KIs15@-oMNn!{A7*L9Dn
z>11Em!k*dt1x0SVqy1NJ>~!4J>$sMkF6eaupHH@5e(gmy-`QR+agMqk<VR4w@8r*_
zJb3*UN0#c^k8zt9+>Qt3B|BfEU$5&{|C(GUtg7C}>soT|oSQl=a8gyT%pLVX({`UZ
zwyD?tctN#CzB%sfQB1bdb-UZT)$RI%pGUpc3+av}@Y-qXZqJ&L$#m(lrAHUKxX-dT
zuZpjBqSojL99yz&^O$UZkEJ0$q<f!BR(ZDF|Ka@=?f2Od&tL!97JlrU8=JJbVh*!a
zZMiSv=BF<HUT03X^`F~uT1Q5$vgTQ`<F@Ygd)$p9j{o~Uy4U`Ubp8nV?Z<x8^WNxs
z{=41CC1)nC+2e?`(;dfc+dJtG$FI7rcSX<1g&p7g{T-fbehu?nL}7kCO5sA=cfJ4f
z<J*~E4a!^B{b{#yg8I+r8~9FvcV4`Ew_~fy+CPuV&Zxq@Du2^pk|Rs0echguH0`)(
z{ps~O&28OI$NfH+Lg|UCDz)W0+H^nZyr4@}^#`wZW#3ouYmfIc?7X6D|6El$FCQ9v
z)1<0n99hcq4H|Y^Pk22ZwEUp?ygtY8HC<3Ybmz+*(__GQoenxttL}PJdCvTAzHL-(
zuVcHGX;<{lj<X8HY}3Z?_iFvs#^NzmO6?rAR;l>puG?=tEhbxcIyb%Q)I2?=N`JUL
z_qrvg7mLZ(ju$nrHSl0sOt<=dUplJvVb?dN<saYc+>C7}E^%ztNelKAdZ*JSC%<ac
z^QWrvkKgrbtWL&0N2IFCRb|)jb(JaUUvKpBluawvIkxM8J8!LD>!huYY*p1?$gXMU
z3%~x{+@sUmpRLGpY}4=G==)InC)YSoZycz<`c=RE3F*Ck2Z1>GZ4Xn@%eo%|J96(Y
zsp6@+-_dx0`kOz9m+JQ6?Vt3)sEk?f9B^z^zt5$ZmVU(2meoIQdB}-M`Tig~{LP5(
zp1QGbp_r{&cH)4#<4P8e>C#hAs#Ue{Eyu-V>zQL(?@nfwh{;yj{e|qwdVihl=~8+w
z)^rWZ!|QV*s?*u-e^5UBdxWaOJYQM&@##6AsFbf4ANKp)ZmOhL)sLdG8`b>F_b0dK
zpj982HO}~Yg$)Po7oz7Eb;r|ljzLxQ{JYn9ukS~xzP};s`77RzuJ<4Ka=bp@WJ<cA
z>j*xds`*khF1<bnXKFfC?JsMcC9>mn-q4jU?0LkL{6WtHQRnk}9#%C^gte=B9Uiow
zb*@rr>oXIMi$$%fo-b5wH|pnd!<Mh=etgjWC@Y`p>lR)1Rj-d#*^m0XA*=kb@B5mP
zU$jo8@#gpYbzXmODyrvZ!ak=Kc3(oo+w^!@<3_d}uh%o8>x!`Ps^&@FF5PDjs{B3$
zg;hPT!Q1sak9bY5YaR*v9ne(Iueu%j_WkpL;yJ%NBK7)wzt{5<pU&@knBSjE^)Nr)
zZ42}5ZCg0(x<*!h(fD+Wm;HTexAJ5km+ik-d$dkhtwS%r<h~PD?ssg_pzj5n_MMvr
z?Jh{~{p2o3wNn3)_5BWA-`7xGj_7++rpNo;AE)-$rO(fMy~w4V3&(V+u6l{eQQc47
z?ECGn(|&4o*0pRCb5p<fKm7W^r}ulm#P2u=n%=Kpl-{kpp!<Y=_mlkIH}$)3;&<P|
z?>-y9U*hJU*L@>X)A?;Lzv)H&ho9eW|8B&<J>7Get$J~(Mg3kqVI5PYJTFtil@EPe
z=(JhenXP*NDd(<Q@X8Kms;X+|u<LVGe#x>M_WgU+*GX>nZG9h&{0{rv0P=_XE*rIj
zty-&k=QR@s?Qm>W(LO!hFAJ-p^*sMvtJ`&fEyu&2uPDFY^NZK#(cRMdEmxG@@AZsn
z>HS_mq-uJ%@1ULYUW2vchHh|7QR<JTy+5woAL1ZOW!*1}%FgvquY76E&_A72QPb|*
zxcTGt`4+cyvidXGx#80p-^@R|P|OyM>N$g`#%<VggN}Dq^Y!wR?#ZgMXP+ZmgWmu3
z>qpRXOv`8M^J%)Cw^v<`>gVTF+gHs;mwmVMEgh%xrC*hGUFWylsM71&2g>jJj<VZV
z^!~Fg@uGb;zwta*x^FK_ZQ<$FPruKD*pgn>cfDYzLecr%w=oe1Tl#~WpZs%6riA(V
zEY-uJ-~H58Z_)iETjKfqBVoDOa^K3YpZxth)5Egg!_;L@^jw-P@wzyvD#z<O(A0FY
zuD`?D5zV7rOU|8hQ>O)vEOonI$sd<ht}HuYpSSRv-|utEZt3~wl5GjAe(u@r`E34p
z*ztz^2R1CLHJ^#*9k21K`$bUuUi}Z6&hNgFTl%1W$jY~(?hS`eZl500rLz1s-Ht8u
zJMX7$@4NH<+WNdK`K|l+R8^Jt<KSm6JhMfKm~Hy)6MJ56y|;ACc0KjnudjZw=ZP^>
zH0*qRai3*xUKL;Mn5I-O)AxDY{4@1B-LGH#KBX;T{yN8RnEIh=wPnBBR6kX>&wVF<
zR^`F#w>Z#fIfQ+GF~8?mJx;>ff2Q-!3Y+Q{jH%MF{_*+U_90Xq=lyUC^Vf53VOyV9
z75OdueIZrPYe#L@?fq-7$3^|IS)K8h)oGg%Gfh>0uU@r%TYpEB{2qLKx4~mhPK%kQ
zqJHca?{-}n^f+HGe}9p}{Chk6^WFS?HNV$IqV%$_3;jMPB1+G5GbMcdu2*AqGWI#D
zRm4SByxa47x{u3lU+3oJA*aa~jx3e!pPN0J7s9GrIi?<`_ex}|tojc6<o%Z(Ihez2
z)1O{CY0SqJH!@qb&3hLuKX2$pCR_bJ7sIFL&)>Skrri(F?GJzd)|RknU8Xud=sYNT
zPr)r-^xV1Mc(-!gzWYkw&oNC^zYig+zCrtq-+a2)pD6E*cbYBP$aHbI;Rk2x%@34{
zx#_t5rOPHhK0X#y{q9Tq{mzHmb0hRTi%7j*N4mW??DqQzZue2#elNyt|Jmzt(YnCz
z{2)8MtoK)=w!@!CWOE{WqW)}JJo&3y{jOv3YkJk~5S9H@{l1^+^|Jjv237Z)b=eJi
zU&OS0{QVWLVSc{xyWUitUe<TD_<HIJ^ZPie%XM4Ni;jEsOV<4x*tx9nsH@g7+c@aD
zzD7C!yq^*F`Yh`H7S?~$^Lw2~O-UE_`=fsIyY1tce%$N#L}j<HE%TwO?^MX<DYEDF
zx+Cg&RCf7p_nkz%Oo{jVeo=l&Z%X^P`4M!VMr1d?w69w|Rn5m_SH#~fo`3$tw(!k8
zI=%hbiY&*gy78dnUhRp}ksZ;xOxAaN-0bk<nf%hF{Cy;ZuNZN`7n_-C4yNsk!k_M!
zW<Q?C!g8?vK1HhduWSC4RgNm|{P{&!SQeMCcHGu|rW{xO9H?shvY&(Y+U~Ni?^W6J
zS|7jZRIgK1+gCmQ{PLvhHCxr|zFgGlMC~HFUgP8a&NuvfmTqCy^QD`8zu$xO`+jH8
z{%y<mrg=X_{VmnvgU+YK!SD4?*z={|@=Qr@`u$Zm|5ROX=(fv0H)LzrwtZdO{dGJ4
z2R%;ns4A7UZj<Go>gxmEzAV2*c7mQik+1sBE9Yl3Ssc7RmuqS|zt<<C{z&nn=TQ;g
zyw<>jX))CtMl5Yv{o|I099vb@`&!gLMQZAIN&WiecRZVxKJ52pqRuaRpVYK?(Kt{Q
zFYA08wH;aO+NkW<^4w`qKlybyAC}#}{qAS#PA|K@B73^)CA++!`^<jxiTLsHQOys!
z%GGs$KWI6o?HdKHhui%d9A9^P){IOR6$h{1DM7j&gUV&Bb%+B$Uj_}ky(b~-xGDF8
zg3cqd^7*}AY)X1n_iKXoTU+z-_xpO+_`4>_4#`%Ds__!>lJWHJ$;Qcefg^tt8HqB9
zvhlR8$wav$;cD^pczVMYRk|e;1A5fkP$^wHhtHEmKX_*IgO07*cFN;F7Q7<U5ouUp
zQ?KrNLFTC6nQRsLC8~$&c#%Ir<6mp~$b>s~<T^n$Xt}ELefik$7kxG1AoHstf2wae
zSg-el^(?50?8=VMDm-=gooP!PRVv$Gk-b|FURbm3=JgKrpE-!Q`Hdfx+j_>3VT&2`
zpE>yDDN3*Ue8Afm`GNSq2+V!%ss?v#-tO3{QO%1b>owYV`S`=kP5rLFRQF3;=eIW3
z*8cUS>A6g|UR-4yJJ?|k3#u0uXtDIxvlxTQ<sfUE1YO6Pme1C4ohp9w>ke6AS^ZpO
z$JYH5uYRZXo-lgoPn(%Yzqz4E&)xHOvZ!?B<nw>(J$*kDwU6%j#Oz5mUS7H1(WScU
z=eazWek1BnVdGWpSCHMH>ljn=HNUfA-!3~=J5aMZcphgi3BAtCE@qS+mUZ2%%t4Qa
zqq>f^y<EZ@^nQV?eB?(^l|8@fTGxm5J18&N`JUZ2<8bj`_Oh@lt3QeCr20O_rI-IU
za>2)|nQiK|Kgzl;<?USm<mXwBEZWY>oRPS$^PJs*mESM*S}&wKmO$OxJ}$Io{Z2=h
z9$R{Jp|W3WANq3MW+!Tmj=-@cr}}*DR=;Cu$Pek-2lo5?hFkiu^K{htz3%@Zoto$5
zbgl1Im6Iy{LKPq0uGg6}?7YCQ|Dp=JomZkhuKPZ?uJUwUcX{p4Nav40*!>u{{Jb~(
zb*QSan_ofut=oRjqaBM6AGAC@lh?U7wQY0pl7g8*mF_!|@?7(4*sVS2xM+PH^!yi;
zchuLpRDMwZ`Fy(0GpfpQn<q>^E_?o@_BFlUynmwp!LN%&VXx~$zFe<i;;Kr$`j2$~
z{#>Ki`<tTsEq>!w^#`wZb?t|V%HjPKjbpFzLDzMunvdrjG%RXAzwtrK51NnGJF+yW
zAG-5-Juj(Fr@P)%Ug7=sKeH!$uVcH?bAYy~>it@3=crZH_)*Qz>-p4gx>P-n%Bq*>
z`oq?ESzMyB<F|jOTKcHxPgUjjYp}7^S@RD&B2`tcD!WmQkFe#cIzQ^RD|?@f?5XN6
zWLMU8PgHhnTlbQ`e)oa$dwS}xewBDb)qFq4Z+n=MK2`R)={#YZ((~oIRCT}8^;~wU
zmZQ3Tc>A`$Z)*GbEb8lIzCXwgU;AC%sTZ7(<v_i0ux))SyT6b<S?dzAr%P4uN6WGw
zl!xy91W|cG{kL`hnEci?Kk$61{ID8!pDb*CzFvIT>-dWLyp8hz{kwTltvC4o6t=&H
z%`dB;@bwBCp7rV3TT3ka)A6gHR!L)z=IwT(*0A$n*y}E@`^>7(7pnGKqPngN+pfCv
z%NpOR+R^Lz)V6f8u1|P7s@lO+yJ62GrsR)mz6x8e*YgOaQ++;(+P-KW3fj-2de4_H
zH`e#f4rfmJli8+xJ!}h$_AkQ52knot^6~F&+8Xxy9*b$|Xg;=GWqoHdYCEFqQ5tW4
z&ud=yFX+8p-5U0~BW%A7yY7i9zpnfIe#?payc)JYdCed8yql`|O}#$!>zCho*0l6~
z-^22I-gE1}{C<Qe9CZHlI!-B_-~BYoXRG>sPt>&Ze$P9mr8kYIUw>rx`>^+){N{I?
zpHv;Eb-HQ|d*4TQe!ufCrB~HXLG6n6^<>9W|M7bs7p3=mem6C}?tUw)Jiq%s)c(4Z
z#+xqHRWDIFUe}#|uN%Ck_qs1b>0TT6QORA;e8b$-Yx%O$dF2>39dYogqU(*Y@j<zT
zE!Qo7P`|w9^Sdw4&zo-NN3VWRIyZa#aj)w~UF!nB<Duo>zwgQ`Kfn34o8Qv&yi5tp
zzTVPp&(!rI`K79z$$n5(<(DkGe)kRd^kIMRF>L;*??;3!U-vkcU7p*z#q{I+I>NTF
z>3rS%6J3vbUC+6t6Sb$CU$!3iyAL5s@Av!?bY0Eo^INVcz3uDXu+Ke2oj>S$iO(l`
zUWVdLd%vIhqwW1KtpCJ8mh$UP+rn<=Dcg?Qc7CLNrLZciKa-uP=V`y?i|#+Ej`!=Q
zDd|P)Teo<xzxz(<ygnZh^!HiZu5*LV$6o!WbhgctB7f<3hip^Z-U|`=8+06)l27(N
zPE_{X=0Q46=S#mT>-xZNxlyI}djB+RI{vynzlCMDZ`kXtsPl{V+5E=yT>a*EOCNUq
zFUlX){7LJgR8v#Vv%DX=-=9)l9$!CO!>0Csem+a}u;{v5SH0b?Q%yh4UoVE`X4^iU
zSATg<rib0uC8i#adR>{{%QyAB=G8CJ`9ODk*m)HBK>!5oArN+dOIH4{#7sT^g{_xc
z{^u)wn0@i;!ZBIuR-WqPIS(8+<G#ECF<q*<KBm~G`vbNq-M6(>3trZLYR5%|GNI9O
zn0r&(HWx1`m}#3PiXQHqHor)w?RwdWiMz}GQasZX4LvXSo-lgoPn%iT_cEmC|GQ$M
z`!8POWv%OA$J_7Jn8qpWzC-Bo<<Wqt>wa7O@`yQbGSm9CUw^#Lm!_xlTW_R4cHRy=
z05oL``dq2ke6p(%)qN<x_2A#LaSMk%Z<>~0_PoZQ7iHUz%ATzGQq(_1?;l1Tuljuk
z)%K(27q)y;?}v-}v0FU<y&PF#zFg6Gb(=p_J-_Sqc@D4X-0I==IM2<LaM*F<H9vp8
zk{vc}{K{@$+3OkE{<+yp%$`)^<(2!Lu<BNhsmE14uPv*-s@Dgq?9+SswyLc6g?Kw*
z=U35rT~~b2`CRrq8I?U-`*T!&2b~Aq)-R%d=ye>p^&hwL-HyAh>qY)kyjy-c?)Ud7
z+~yy*b+cQ3ugA%Mkp}HIe)D;~-axt*OK&}kG21wJUElZ}M|^s}?@RId-NIql=PPdc
zzT*2Ue|OxJU$6KLyUnX^<@al_vDI1g4?AI1)b4)g2ifUW?XT#z%bzD?b0T}T%s1q(
zYV~@4&F|?{w?kC+y{>PO?%4a=VR_KH+*TE>vu%k_mHEu=x>0nTpKqdizCd-kd_8rA
z`TY^q<$8TCg-_?zFWu`wQF&qK-Kg{1GC%TuM$L`N=l${;ru?>Q*y|Qk&v&-82Y-Fa
zhrQ0DrljNjR~;6uLu`o`?f>xcvfKB|$9}))s|g30>FRgC(rbFzJVo}r?uYu_zcf8P
z-+%eAh?gnxs^=|R?VHj*ZhpA!-;0im?3S$8XyfJM53{gp+xL@sfBpXMqTlo){%-N6
z&xdaQxxLT8A9v$%^my3!WB)cJ+D}pST$-ERqxIADk7(Zj@#EOR4s)0)4yImT!mp{r
z?@U|bfP>u}G)<q<ib=KpCJjtTLvGK*m~!0fzHEL;=e5~I*E??UZqIwkKJK+Xe$(mt
zj+v@*;#ahJWpj3+m@4(l)Ao7Tmiq&ucJb=J-+d;(-=*{VU0qSSu-6TK`=Q_YD(G`G
zs_sAd_1|xLQGZLdc)#O<(h~>2{Z6!h;x#_(IEgBM*!#;-=g%+eNL%XZcifqh-tT$U
zl=Q0R4Vp*ui>g`=i0rE7O6No9<3AR>BGZYAgR1eV%dTH;Zv9#`9&Cws+mFF<Z|?^N
z<-|W<Ei3HSZ-e@4T0UFGnaEF5_PK-lVOoC>`QtVoWFNQncMITmk)FHf?PRuaaN94G
z-LIv8w?6FiLVSMd75-;~UiZUIP3LzV95nqu)BE38(fKh|;$^MtRM|0YKQ8Ecgnw=k
z;j0?lv3a{=3WuQcNPf@vzcXgA7`OF;tmCG%SM8ol8{M}uEi)<&`}}@X`D1<G>~Q9k
zKbdR|)@$v}#gC`nHU6$kvO}^}qH4Tk{P=i!_hjQ_yugvaiHt;<MA>*+*JPsHk#Mzm
zdOW>hiz?lci2*(8ZK#xPF9~=K0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&knafeuJL!3lO2++5>?~J$4`i-
zcTYA>#tR(zo5)C%NtBJJbxkJ99SK*9r^nM9wy4rAnHbQc-iAu)`3@K=3<4kk0w4ea
zsYu|{6R-Sy(x`2YU;U+O%d7@3u6A^3DzZnNKmY_l00ck)1V8`;KmY_l00ck)1V8`;
z^b=V2<*#p4c{PtkrTS}(5<vh2KmY_lz#IaPOsQS+vW?j+YBh%lJOKd^009sH0T2KI
z5CDPHA&}ec)-l8OtYKmG#z!Cc{heWJm`GEH7-|9nAOHe-36vZ1_)lwI&t<l#UOFfZ
z1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;
zKmY_l00ck)1V8`;Kp@o#JW;zsn;-AVWkGfJBbAfo2IMj*ILPP+sXpFNKM()`5C8!+
z1n!@g*66FJ@)-0l4r(aFDhPl82!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*>S
zC(v{EfMN^J&SB8h<`DJow1h2x>xZwsGiG@ngMx#J{XqZBjS@is1V8`;KmY_l00cll
zJ%O7#pYhVW!?Ky1*8Ao7Cri{`!~AOB8w<UgQ#^-3!2$gM<3RU3Kr`s4rGL5bkr%R=
zSMA$k+*!HB*RZfk<wHRL1oROoa>vSB`n<fD$<p)IwLSIci&rq&I_`>huD`YDD&|#_
z&Vyf9ZMKTl%xKwVdfDJQW4BZ5-3!(*C^(=Wm^Tje&6Fq!1V8`;KmY_lASwa{Z#lQ)
z#b5rwyy{CE=gs}#`l$>G4l4UWY+%18#cSs>uR3z_-LD*2kjtRqI7BrMp<ECE0domF
z*?7merYHUE*s6otwpdqr&I-p?z5BGQI$ZVgcE@&oZ0}VY=PcRbK*8ZiKQMR3gJ&QB
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0=5z;m;C;`HLvHf{I0gPAN&RZ5C8!X009sH
zf&5P3(pziII?yziK~sZ6e(%qyKL~&T2!H?xfB*=900@8p2!H?xfPlFKey+Uahg*uS
zX1es=o%4Ebo4kspx|-YK@C*b%00ck)1V8`;KmY_l00ck)1VF$qfr>3&n3_14!=T{c
zr62h1Uq}xEAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w7=~fibu2eWJ?^YgyE4Cp~zbsswu1_`7Dw4#`%Ds__%z8S(V)$;Qcefg^tt8HqB9
zvhlR8$wav$;cD^pczVMYRk|e;1A5fkP$@lCxu9Mk00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd>N?`93s|u%Iy_`j*DT)=<0s#;J0T2KI5C8!X009sH0T2KI
z5C8!X$mayUzhm(^7xd3&VfB=64%g}M>1rm@e2y8_2LTWO0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sPKw!%5^%Gmq$zeftK+T3jZ_LVOP;ijZ4+3;yWB~yX009sH0T2KI5C8!X0D+Vs
z@cr&@UKsFY?U-Nvs#>8nPi54MdDRg^UYK;|&>AtXI{LZX371@ZNel`O=m!`FUgv?7
zn3Pco5C8!X009sH0T2KI5C8!X00CnNtk}YOpLEuA22E`auUG!`?qB8|cA(&ZejtAw
z7&D>65fA_Y5C8!X009sH0T7T#V8hmtcPH+c#N70a-%544__e9buU_`phMohPPGeAT
zKtI4Z&^-@GoGM`o1VF$N0<!0s<2y|5_2uBXOt!8VS#xBqCl@fU`pm_r@4dbFGFI)9
zwz=)wFA1#KeKuFxSu>kK!2$iiym4U3Gz&LD00ck)1V8`;LJ+v{P{ZC6OU-b+YM0^<
zPVTeeTL%gbD*Hi*E{&`p00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0{=qbz_vR#3|+j0
zx#=z2iX3>W-7@A?r+#x=wYQ6|V$fqb)ZcnytvYp9GblKqAEeGW_?Icx$V1;X?Q{OL
zQNKI7bjH}x*LJ>Pfs^WLWMjiA5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!w1h!v(VB7_5
za+qH&a!0p1rE2CdSt_M6FaiRS2=uP;x1nT*WUEBgct*TbJiU9eaWY=u$lpXpqD-P}
zJgsXoQSL~%T0A|T-mpcLZpp-e9`!a<N|)ruf6eS3)u!$nB_}(!>N~H$)+6zBt|QX_
z)->3Es<?=}c_h7jxqb1Kj&2qC6Er@o|3S-Dm2Yc{?_auh+HS|Miu}3Y{=p0C=geb4
zRb*Fo{4Z}`l6B+qPL3*-?XSpQZijwZ<-4w6(0}G2;^sI0#*d%xdd+FSIkxM@4<8z_
z_{E7#w)*8MN*{FI2%1mi2jc%C&^FoSxP>=vaBS6}{UU5Wue|)uC#w6U?0G?#Jyqv%
zH@jiigCl>dc6!Aj^Bt&M4zk9H-|=Z`dbfUT`th*yan$)$&4aS~xyX(xzjfOc?RU8O
zA<CyKp3X0#_R$@$d!Eo;Kd<F^UEj%0=heTU=`MaF{j!aP3S`=<t9F+gzUYFT%vRmD
z?X-4hPTs~+U4!;dKA+qCO~?6sy27%~6T18rl_RSj`OOaTq<$Eck{?0!%R(7#_doOp
zlchfvd*!6JveGhDYoB5zzZyR001K)?dCATf_4SL_@>6|$M>T%Kmg}`Ys%k&Fk6ote
z*!0D_w>h%ZYrT-}SOQ`Dli&QumL6S5_w}o&yyyrVTXKHCZ+t8b`5|5V!D-FL#NNDj
zrvv?qgR1kGF1w<61MzB~+abQX+F5luy5@JUGpFC@%=q*xT8$jPzxqBW)s(Mi*l?<z
z2V`A0xY_aApOMZV0oD83yf<|JOtrervvgkZt8VK@(Krt~fAHgj=bB%`Uhn6L&QC$x
zB`9yd*XgR#2lb!NH~56kP3m7+Fw<5Ix({tizMl@4zVO;Zdzmby_VqfhOi4%knL*X-
z^L|0kTf|kBx}BGB+}ryCUe^O&`-@vTRsF%M-KhHo?`P2SG37I5J~G7*o^Q~wtaV7x
z@`L6JdOi^43+jjNd<`ex@X((p=CY_&cfF}R)%$q5?8_Q=y6w1~_re~hc8*$gbCp$|
zs{R|cU4!OVRUg@QgI+(FnlEU3nUYU-``TtN>iJVu`EKo?s$5leMe9-1<K5aJ>~UG=
zLpM8K_XYf>Q`KL{?zz8SllSo2^~_d{`uZhk`GuF%X|>>oEez_71NB$GYU;WuRs8bX
z9;T${@0TJh_c?Djo}&BU%HkDwJiq64q(An#K*i1TKjy=(E0I6k_Zq#vpQI|?|8xk+
z5pLHDrXSb6|6<#ClH6U6nv-?E^{7Ej|IqzaH?8^EGn4mr`Q3?1gW_;o&*(ZH)PK|W
zflTr5fA~szPe=EB%9r;)_2MGD&cCU)ZuMGkThc|<K3sQ7<NKfPu&L)!S<hicWhd-<
zJnT3!<+&+U=X+WG)~}tJ4-Y>${DVInTXb~WVOKqre$WZ44eB&$(Pm>>OtwxKH~0C;
z<qO1ktrffNql%3S#%xvDoOIa}z26a4Jb%8lHSG2tm8r-5u18F*hph9V=)4@XUq-bK
z<;(Tk&ZeaId;ZQZ>4Wx1S@~40`*hpo#|gs6-j7nvBUSGIyUmwhPis-T_3+t_NJaa8
zZs#?(_VM~$4*%WRU;a30>l?jyGm(0|?iPKIS9G5B8ZY8ydi<dFL!LY1k;9H@>h<~~
z{ni_r)qA6OOtyYA@z=4>b}JN9t!H1?aQVWY3de$~=z3MP-(=rwJNZq=U!SWAo7V6B
z`V)4a#Vx<zeJ8*3o!fcJ?R-MV{gzAV{jTra?xXU@!><3`&bK)J@4u@>?QE->(!Orv
zRM&Cc?HJYdfUN6cH#=U}x2B{E`Wz}hFGn5LwZBd6FKZq#{rNbr{+YUO?B<u(`cXP+
ze_g8UJV52>s+Xu7ulL=lJ!GrjdD-jv)9<>~t6fn#zvTrzFH^tamF7p@50Sra@owiS
zzt<hQ(!14*KhE=V3;Vsk<?Xp~<<q-``TDqpy^c3GJF<>@^@q}l<|}@Fb2|^2eq0t`
zksY3wDPdjfN>$|qo!52cQ`OF1?~91isq#yf-LUOL^H)?Vng?u&_q&g4T6(|h2Gi23
zx__a|u4z2|`XgF5%Gw9w?^jcp|BeKG&qFki+Y(RrUBas0^9x<4g;l@hiqhNCULrqb
z??>pi7q%Zzez)Jx6g@W*RXp`aetyYsVd5Z5MdQSlc+vH}t?@KIZBe)VCa=e3^=Go9
z`uAdF+o$!DZWTRWXnMTYe#NKryZ()O90uJ-;P?C8!eQ67rsUVX9+OobT_@SDrtg!x
z`DaQ$4(o^4eo5y`zbb2e=(pUc(tG6@HeJ|uw>`h?_O)fdTI46cehC`pxyl;9WXEk@
z4SL-6{%o56)Ly2k?CUAmJFVTBlee+3986u0^7C1$heh*<u6l=E*F>G4=NFcnt^12^
ze!J}xn07p>=Oe<FYufqOuRr{FvMn68{i4qAHjZ^27tW)SzE?@lG5CG{LG+$*RPlcI
zox`U0%gZgjs^=e7+g1Hsm)~&~^g7s-e17YzD!p*t@p!=0=LS5wA2^xpb$DP$BxQW{
zgkqaoUvZe}Qb}$bZ$`8qX?ncu{;k`d>G$2;{OeP!<X6Mz9AHpz2-6P+pU}BU{VNM*
znxbJRY*Vj~kJc;X^;*==$)rm8s@BV@?S{QiaI_yP{wL~v8{5lQ+&pQfuP@#FOO<)Z
z_I?_*|EkAHSo?1MIPCGT{_**BeGh^!hxfxR?DzQ`zC7E*vgbek`DfGaKL+(j*8C*u
zkE-|8gW9M2Iku|b@oifAp#4>KK2bk*i|6MDSz*3hkzK$2SXO_KJ<rP8=j6GW68^GK
zM%(=l{lQeLtoF6Vj_mf8)ep&zZ2#Qs$-1xTR*tF1Wvy#?JF@EQ^>-6YOBZ&3E$aNL
z-fN|Cp5N5e`w4#ivbEi!dLBpCJZPJpcIBTg_3)F$Geugm&z~c1UsNPhr29YIx2=Rz
zEK|3N{P7wu>;9*%_Kx~IgYvJ*dwA`7riuf<E)#{r?icud{#A5+E{b>aPgF0z@nOqx
zJD-Z`Cp%u`hpu@=RE`^8TaNSXC8}@O`24!BAmSGGK37!v{mw&S(~J1o9&g%pzF&W2
z?KkPRqiR1umtDW@WLkQ+=Xg{7I6rThcHdprdrN#hb%pu&NmZ8{^!Y}wewngQ6||nF
z<cn(GENr=ae~u~~^twJOe`S>$w*5@aFN=?AJ8tub*Ll>GbgFsr_D$`7Uj6dAu9n@t
zvY&HNWlz;UfNY*3dtT2gUi(X`q!aNnC7$0mM7Vspeeso!DedFtN6>L1vMa0oRn3#S
z?5diN$*zdMTYS*=HYK0meL^=Lbey;bRk9=Er0YHqzmG(I=~7+e3*~e<ZsCm^9O7Wg
z{t*0lXvE?dCo)+aysnFpPVI9iy5>(=<%Iq2t|)(0_hn>p32Vn~AIg;DqUU97iI+8h
zk{z$@9<=?a99`<QK7P}Q#=R}^UiS%urt`~F*7a~ycKGWFQ^TTm@%mjpzt2bbU2pOG
z@qWX8{o~UI?eAXm`MoZu^j_`zO-J@ssi?oDT71y+8FBD?-tarFQYF1;-^iBusLoF)
z_pkd6JcqE)+aiCPx^I-6>_p9h=22BD>-;FoKh@ty;O)!uTV%(U`8(D9w0)i6cO50_
zj}-6Mo@wc$+J}@?E`NPyTiEaQnAh~wKSiqRKC0?+=)Rq;%C9$V3rD@K^IE=W9H@%7
z<^4=IKW!aX`NeOy{x7?Kn?66uuCK_Rtn0m~?D*X`51L-YkB@ij$F?7Ln`fgw?)N(#
zrlt@2JzL}(IevfjeU54lquUO<>Y?<54%8n9e!dJE*0nAQ`uvCNd|t1YRi%sS{2sR4
zpyzwHd_nuIt@-%z*1N{vHA!|zwn|iumx`B;r*}^_PR0uy`J2c{lu4A0r*%yx${h(;
zi>Jra8@8y@Etwe5quz!}>C!p$ZMFRVuHEM_TXo~Iq1z|4S<FNl7Fg#lA9w8e<Vq%6
zMSi{iw)1}POTK5iRpd|5`1|*LGv&pH7P6okv|LsBuKcn|Zuc8=9KR~^=hn81?;Bcw
zjuTWxc4f!EFyYGg+YMU5RH<x#MfPS7N`6+iSq_8#GY1hjzww>!sI<G@swvENy|-3_
z^$&cofXP<BJVoidt$6YLypc1RE*1HK_`e8ToU{Jh{<Wq#w(7Ns`s?eK{Ke6&s`{_$
zez~PzM%Id2OPFoi?EB8=U36d`)2%ILjelg`)O9SVF5ED&Zm&ss3@Vp{tZ~wA+!y5w
zKem+FuClN5blaO&dDgn3<FXm_SPruKxyVl8_YSiTm9m-bTJZDBCyn}Y72|cl{=r$(
ze*KV%^z9R#TGq1nD8}n&$$6#H9~r|`YyUFEn$DX20~586?s(nfUU&Vxme>5LwMB25
zIiKm)n_fM5`^i<-FjZ>%I4|1t!H;`w%VxG}x1E<%@6sci*{V-D?UvOpww$@C*Ky{S
z?%nLtgKphCo4Ki*9aYB@&vxqaU4g|ctg7<Qt6kgIzft*5e4}2sk{>~}as92$X0Ko7
z$kLlO9shZUFK0Wd_1ya^?0UH06ep+#<t01cqm6q#w|LM*7FKhL9)J56r3N!qy0zpd
zzm~pfC`(oC*n8Z`)k;lawyD?t`1%{e%AdD?1e2|F-QYG4QaX|H_g%f#3+av}p!>Oq
zV@r=NbpF#>C$wvq<wUK~5jeJF+xGco`+F=6`5|4`q3O9<c^@oW$)JC6n3-37O68tw
zn60X--=Lg--Y>GX-|&8m)={$Kz0RCr=clfP5|>WMSk0o=_;sx=yJW{2CQJGI5@EwT
z$}fGR%z~vX)%3L~zm!_Bay3(}Ui&lB`6D1}yol`Z-u!vran&BEyMn3Kn=ktI)N{+P
zcSL&quyxIMc3$iF)h0DYKen&sTqaV{b4L7ivfnVjU&?dMuVJqod43f3nx9V>l(*`B
z@6~=#|M`5w78iK&g~eHpt-7#Q<@%+c8s^xlOXrPVP-9t7CQGS(Mf01f@nP4`G#|LB
zXnuE#C$6eAZ2gh{ulKL^4H$USjqR2*H{II2!#8^iFJ)m>Re$hmSJnGAUhVRJir%mD
z8vo>rB@R@qvy9oQvab($zCpvX=KY}M2hI0P+QKonFJ8`MX;43O=R0)P<4cBQjB=t@
z-SwvOw)|RdZ;59QIksz&<A0dgcWA+wZCdxqOM7&9tZ+<~QaeYjU)}w1m%>d-#boQ%
z$8{dCYig;ODy=<o_RwYn3dLmWMc>rEWa~reG2QxH&O@!T=M{>Xt^@K09oOjWEl#Rv
zu`gcg@ZFSyj;-3F#NoVJ-3!O^t42M4sw&^@IYw3Gs<QiNot5{^$y&^8(;oeEiyiJf
zi>cB#?!9rtski>ZRO_q7bMM>L{97hlRrMFL8}`0s)cH%l|L&F9cjq$Ol)vwf@V`I*
zL;cmSs(Nl3cK>=V-EVuCl0H@1kAKhAmay*c@~G~2{QX~q{e5>$b^Gx4+g*K5nXKRT
zIJW9NLw>8?spKDyty<uLMoS;K=T9dp<@<x|@Z!0ps&wDF(XmzU-Z{Aa@XDJVUHWcz
znYY$%+2F|5>~aSiT-AApBU@$n7qWNgthPm7EV+_}RkEi`U;nPp@%`^y!`yUm&qJ?Y
z_gOZJN`vw+<^BGk{(F7y%hYr$7nSRD$0ZAxYUTMZFFUGlod@Q#u-dEQ*HbSmKbu9R
ze7*Rv*YA@={T#pS@{6?o@Y7$PTFp{L-{|<~=Fw;8Fx|@cr>L)g+@90(dYrFU*l_Et
zDdRsaHi7w7-QVFBmG{!!-#`A%$un3`-8<m((>pvl!U?K3cU`=q#P%1Mo95K{VchfI
z40lxN<<C9v+VU^wJF?Xo|4!u#PF&>ZR$KOs$WK}4(Zx+~8GB~$c}%vx-1oM(-@9=-
z<5l;*m+bPcyY-u~Q>zVdWa|Ytyf*#!(H}UfRP?>TsN%z(M@-2d^?5#M`Kj_ed(isx
z`9$lXp#ALN6%~u0k(=X0t>1V1xp&JG*EqK6sr{;cv?e*hu~k*wUklnFW#yB#?ugpX
zE2p2<a7Nw?W}E8%4w9<!bbt3=Re7R*OylkTV(UgU*}BT{s$So>o?i0QU(z<Lb#&{N
z)8C#vc+74msPgAszhPPL*}%?U@9F#HZ)<wd`p#>-t@X|?eyhHo@!C&xr5o3x<#!c^
zEp)tU%?C$3Hh=zfC#+Vz^r!K6ZfA~P{ph`evCJQCV_x;ev3LFaRqu|DSM_Sw)O23k
z(bROZzOx*aooV;)xbXG7B~H}(-IO~=*S{{?@vHyoJ^t{x(%FtI-MM?)<Fh(tJGyk`
zm!s-k-+h&%TWOuHTFVb>@!Ah(jCE|$pQ|6dt#;ur9NSeFZ|Xm8-;2iaV}B1SRd`VQ
z>(ZdV4-hn;u6l{eiMpTMQe;}0n=^BmS5@`<h+fN=l}=SXU{~(%NAf)V{=O%l-<Gi7
zdPbd|pWjsPTd20LdS2#vnG#lAZ`Jly_lvOhRkd^2{W(>B$+COb<>?joKflchs?~Z<
z?tfy1-A+_`(cy2udH3r>j;*?)*qH;mKKX}Zt3ES(_On0t+2+`)<q8a4ete}(j;*Td
z`9EED@2%Bf{R1B?V7$t@|1SET=(<y{KXGusMU0f@dAHy7H2wJF2d;58eY}K)RlnyK
zS>FfsTdpX5iLx`EA3A;wvsGsgN`6+iSq`&Ry}lpimaagjO_|-_&xncip()P1YU@kI
zOi}8O<(rQ<)U;-WSbkOFAWN_PblO#)_S@;$s`ICAFL-&eeNL)q+8?iMs5v6nu|<cR
z{PBQR#j>2R>eint<_>83?Nz76yz1!ET`pR9^BFN)bmNAFPv3s#@iE<6x60TD7Myot
z%r@P0`7PINzqe4#uYOVgw&Az7C>Zmqtxp=gclx;nV{UrQ*skL?T~i?DS9czMq-Os)
z1!5xYy{}FB<F^%z`PJM}uf8^RVu6@fb*rB($A@m-Q}V2?yB*uKbm!~}OS*1wY*$<N
zE6DGg9*fsG;nsDI?K*H!@pX^x+~T;Y+j&=X{L-He4ar<K&k3s{`=-Uun0W8&GqQF$
zs`dQ3H)lP)`k<3vRd&1E{@nF~_cpF5JZ&xut321+UweDb%gytcEbTMq-0nr+&SSDQ
z>!JNGH#;Yn=~mmm<LlMma;G-mQ+972gR12qYaH~ez4O|-7p!AZsjT@ywjF*xOZBkm
zyHL97ZTowA{5l{kx0CMLTkg+lc`Vgb_xHNOmZy7NAgesD=WA2b$>I=|9oyGGrt}NZ
zJX*YJdgHP;{@}>c(qGOzA)Ytd@v6F>v+!EZ=+mm-FtF}8rb|WN%?ldee!*Q=9A9rT
zvrQju-0Qi;gC;Un>i4}SQ{Q{_>eqSSZcDFp`*<g670o+d;~zh9SF;^YPGY91SO0^i
z^V?tD(g*cJR=yk4Gsi#w!V*W9%JSQEJ4@${UQlCMPiC98p4s5-$A4b!*s3MIYkuaV
z^L9J7YSVM7JXx>ZKF4<bD8152hj*mKY||P;%6>5YK#`d3I_X!ZW%EZ$#7xn!^R=n(
z?^C_Pz9-ae#f#_XjhxA%(xBHhrsY$0oub-q*mkFWs9J5iel^Ws)$LR0^IE-EfAyVX
ztCl-2IjeK$A069O*1lbSwWE5Rgteb|=e+(uFP^|uX;}aG{LP<QTlA)x^O<ht{csEW
zT~D;ixIepFolKT0D*JaH_<hDI7oBt6mN7p&wyUgt6;Xf8Zga|6_uszSF-^by;Nj*U
z44UiMstapXu3!48VUDe8>O3Rr$8PbW_mIQJ^W}=plRf(97CYQ|79+jwKaG2>d?$;M
z^1s(7>-T*&E*rXiLYu`*mWKVEPs-18GbOz4oc4WdH!c}dt)lPH%Zh)lT=)4chD~Lr
zsO<KY{hXO>|J>~P{a&mny<0h^9+&-`C)tx#U;cX-wuD>%c;e@MFI&WHRr(%?Yz?~(
zi~2eu=zC^Go^Lp)ZNs&W?JE2C`*hh0`u*0h`GU@a*H8N3=|6AJW+IJW*XpuMcC2B%
zy8Vti9WVUeVb-BiHWR7a?<IG+wqVmQu35!IDzZ!QqUYl&UQ};Bp4PFtR8(Jz-#1|3
zO*gh%&fL`P_tJy*8^8I&ejf?><$PyK6~_bF-Osp)?RVTX?DH!VUcWVCRy@z~s=X?H
zGkj)+4USiJ`#ng}x}4%~FWlsc4c{+!+|=!NLNY5%e)q~}7dawz`(5Yi`}@usRC1mp
z((|9rI-y;=EGMkq+II1ML+j6Rc=bC!$WBl5r0q)QQBv8Q$ljobo|xQi!f%f0O8%-=
z{=G0;!t`B5U8=erqO#xjnYUV3zIQFNRp*~`?|mJ1=Q7*0@wD@|KlD~EvsHW5yl>{b
zojJ@_jq1HS+2y`kJomm`&A(-Sm7j0c^__a$f6mzDn4;&sR;0}EJzJfy%GXm@xQx?k
z;VCz~>yWB0x5alQZ=1P&v*TC2`en-dmTup5HT`(EotISa(j%MMrmBAj!mE9LzshTv
z_Q7mb+UE<ZUOA~sC;Hwrzu#(V_=H6l-ZSNm941n)^QbB5hM)h);$MFm=;+dpcT`@r
zWOcS<t9sopklntrp9fH7Pu0GGY@Q-}eLj8al}CR3mAUDQOE!<_^z1a2D*A4A=|Q({
zp3Oum;$=#_*L6dFOJ_>^xcO1=^UEiV`f?TXt0KF)_5<^KIU@dU@voeITEiK6GnlP9
z^yH6b*DUfA^Qvw<=y=%s?PRBKtL681?LLR`I(Ocyhc@lYVycz@9V7Bfm)d$C%=YtB
z(Cbpu@)d3R;K#kTWi#7URsW&;vAT3-UiB%Jd#+)&sp{VY(=~s}Do57u?D2NOeg`~g
zepy_?+Og$%?d<oTzoY%(1<Y2J{hqDYcHh!3BWp#iCCqg7S|7jZ`1PZ0;gyTZb-Lq{
z1x%It<+*#+)sJ2{a2c~zRi9U7*%!4-PMsgdJ^#&c$4$L{*Ok)c6g~d-FG>w&UiHPX
zcm4cT?~aZw>i2nuaUbm~xMbOQW{S%G4ghaY)ZbDqKInRfIOw`=p>k|}9%|u+iFJET
z%44>wt<TZ%e!shS@|V9&$Yr)Gy|-$c^6g<uSk<_tc~q6kx;~WUpRDzuS39!&7TMvy
zzk%?mzI7g$&vbK;{oc23dw#zsBkGS7fAX>k<EBo{bZpg&bJl;`zt%LzR`vS39j2yJ
zbsbCnQ>5Ltzr1o=izKsM-#+1~Wi5M;Vz%nbeQ$gFy&I=9Ta~{*Z)#XH4pharm^J>9
zc~jRh+m-eMY*kaA!<OZj+xlE~|5kNBK$czE^%dE3`woif$8|jyBI3u#Klx&b0~PBm
zW2vgL@59Qr7j%DGRlcC#M^p7)0PK!_Y+uW{EG!4v{WibY<LAquVO{SHn({oPs`q)k
z+6_97MV(L9_lS8rLHn((`5vva^1eA)iy3L}8h_U$*&*2~Q8ivVUM8O2J=r)JFL2~<
zA|p{IQ8u2|HJK=PBwQ_?9#3!BqDr@9VnC028!DyS$^m|Z00@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*;>M&Oc3kA70-*#a@YYFIVk4hVn%2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!Md$1bWx_+reaqWUEBgc$xT#@$~M=#>serBYzVai86_@@wBeVM7bm3
zYVq`Vdczh~x+N0>deqxcDLrk;ifeM3*WTl-dU$Ixx6EAU&i5Pk__Ozj|Is3|$JFUl
zcY@RUie8`0>Xpep{k~z#ZKviQO}F>*GA}Huxz_2KcsgtN%?q6kA3r+&<myXVosV<6
zG}yg>mAUei74OfO%<lhn=E#di&TuF{=}C|O(fFY&zGbB6uNd>i+i5!-QnE|hrTGh!
zpDMr0A*Fo04yxI2%)oXV9XI{+hn9<8dvmjM-3w25e}2_Ir`yVz<<IT#hjYshi$46h
z$@-)E%$_krs<+y3G<@}@hOIAJx!QT*=^_Pl#%w+6Pur_XUVU|+olc?hRkqK$aqm&R
zvQ1ja=esR+cx_y(%C5(U{^pD*|MdkQb@|gd?dsm0ie>F}rd?g(yM@IL9QA9`&sld@
z9<|${cBA^c_0!IcD?B=|-|vi+I5oJu$2G+dW*=?;-UD{+&q=I2+HO@FmE2XV+Ise8
z|7JJ5I5vx&`p6$Il{$I%QKjRg?+<CxWb~%(?9$|=t(z=u?POm0X}Jdr&S0dJ|Gat4
z8vgv^Z;VgB_1sh1_djkIJ1cAahj-4+W^=M`SX8;hTK4?htkT=+Z)9!Tw69-(#kQlz
zsa)z;)PJcSyi)%mf2bX(ew}W6Wz5xEa~Sm(YQL{PYB_)7`MaGv&gl1j!Hk{Gj_PIi
zUOAvpjE)me>Mt~&C_RNqX*^N?r}C*DG|ouLU&_C(O^+LH%v|YIZ&4)e+B^3-)UI@#
z_*4I&aY#yBXq-}iBCWBiNweKwA9AQ)ZhEeDsk}$F9j!lcB>OZ@%l>>r#XFzbesnx^
z?D*(yFO<k}e%W39@dv8qIayD=Fs<xMn~u&CWS7Pjai(#Y_R*sE%kS8Bbo|ozpmgL1
zDb<IRIFM5RAkNedbewoon9>t}3g?xt(E7n8^Vnmx&uen=)w3CuNBM~_g~=Y}+fwnz
zax=QFW)!A$ly82c-qjb~znOjZcI#>19>_g9{%CyAc%b$aDfvrUutAS~i{9GBNXZT<
zjXN5jG~UP`-am@xmF82L7pY%Sd(*h1@lEppjb|DU)IJn`;)#+o4}CR;5idHv|NN(`
zOszlAA<m>Uzwt`rhw_u&(|^MWKeXQOkW#yol3h~j_oOubNNIeKQokdm`GJ(|@JjQ<
z_qS#I<lI**MoQ_Z-S~DOzI^{6p46Vill&%o6i@w!w?kp_pT;FAjZ0o>ywiM1{ek9F
z>OZ7(J{Bq8zM?SA(;{^{kK(x51ZW&n`_p-c`q|#U*Qw|1y=Lv7{kAxy6~^bx9b5I#
z(e)RlA2NDF>D#CM?oh2O%t-&_?DnN&)DP)=M)L{HPt?CC+_2QALuRgC!y3J_wZIEC
zHymAG(YWZ=Yjva3{!EX}duj7m4=gAcd%e_!U;WakXpH**tahERd1igF7}+JId4$Fd
zDful@THlD2{PL?bf6#m%RO!4gQkwUuJZf)JzTPx{lJdv-FtsNs&xzW_t5SawDb0VR
ze0k(IDb<sd;z{}X=nB($gU;i8f2Va1oi}LRK}zc#%1`H~cPb42<l-`m8J*wgyhdRv
zm(~&dI*HQp<<fee??==QPQ{B?9EdMJdL2T3{ZZ<?wS7)q$H+hOpUR<nP&<<Sksq{7
zOBVa-==CwJBPbosPo#AI6)ByUMN0mY5+_pfpOn@Gq@r>up6rm)I*C`Buc*GHR8LZh
zC#CU1{g<~(>uyr&|D;q8Qp!*5PWePi=LcTtx`<br56EviKTtf?llm9!AH2PB^0}=(
zTkg<&LrUp#Hq_bLyUk*U=0hs?{pYvFzx{ldL+dS?7peTM2S&E(GhvTI*JZ?!*5Ra-
zp8TcrCiP1S6Ib$^`T;4Oe~CA7rS_-sH)76&``#>8Fh<vNR1fl#<~33}4-pS4hd7b{
zR6iOIlwQ<7DIK*7aV0;=f4aV&xbgB<=k4Edbl-ydQ~jYU-g>3Vo}=-!-$fi~-lOZO
zFH4U)^MNLt*vma~X1(yh_M_K<BN~*s?6c|n7%ADM^#fgR(z=5DBObhxe}1+4>yIAS
zp#N@mdzq`6SATQs(ft?VK=Ulkd&HNN=0TdLNhv>FUy78tk<xf2rTI#hQvVexU0+hW
z(ses2`AK22vv%aCPt-bbz|s9M;zIq4<~eE~+8-m1w7w(WRDV+H7u3(m4qcB?JdG2Y
zKj?f!<D13}%^ws`<BH}#3e$W=VH!tdhvo@hX*`qCyh!{=$qp&`P3vw_n)gVl-lTLL
zPfF#I^6S$|+v~S%S*T=;l;$Ve@8gyDl0EX5l>8*6{W?-wuaolQm+B=_suwAhOG@RE
z_B-#>VQsG2$4IF^Qvc%TH=2K_Us67rALuxZYf48-eo*^ebKu;IW_I6kbbp)tBF?n$
zP4TqfP3JM<MEt1Ul#be+_RVSEh0ZTzpZ4d;j!3D0()g$I5bbNy_$0l2!^n{{s?R@q
zUxemqnh$8))4WCeX}r^XPD=4Ke^NY+6Y4K?oX#^e4^TNYUr>HJPV)fu<BT~?w*Iks
zwnIws)V{=n<_GHc6#jqg-FLiQRkc3+1VWYGMFA;FlP;j3tZh(GA_TFbaKQi)kq$~x
zkX|ne7$Ay(2vSAqN|6;oiWH?v?~nutkU|K7K;EC9H{aZQhfdbo`|MNp$^0izW*_sJ
zV~#myS^MC}yf+^BvCb+dT{-QEkBE1?@GBznYi@si-LbcwJbQzD+SeYx*t*Kk;@_}`
z#!vc+KYZiFFMfV>0i9=Q2ihmZd1ZfM-tGPJO=rEi?v!%>@6R{oqT}B!mw9FLM{nN!
zotfctE`4o|5089rre7exd8Qug=@(XCZS49#nl|&kgx^8mp#2PV&Jgv5-e?E^<b?VS
z<%j&MD%75RpRjR?NFI7dAIuL(-;JYjh1O&91~2$f551sI=A(5sqWa`h&O8uiA0WBN
ztsZjXk>8>J+M(z8Q%-&Aw|+zQG>hy?3bmg?e5X*qQiwN557{IAgY0H^Aw9L;7zchE
zIpxF4)%Se%fMEyTS{QHTq?6w`$`?_(`G600mY(wCwZqQCSg-R?Xn&|)<s!;oh5QQs
z2K&iwv3u;I{B<#<U+mN)^IZ7SPKQmJx&JZ$=?VQdZscTtZ`<w28=e?CwPdI9YQBBz
z&MlYt^QCXjOxGXf$N}|>+|izsi`|#cJdsX+*fI7;*mwy`r$6*gJrVK29!sZZ^qpK0
z`Qy&R*&ljLE@&LA<K(k1iCCp~J(X@e>9z6UKTer(PXXyUy)+Muv-<cO<}1INzR*YI
z@M>Hjc_6!CJl#L?3+XRCH(%%h`vTPijid7Vr+xL2Te|wi(*x%g_&s{wwLiM{^;2f<
zd+>udewX~>-#g`iO}@DIxH5%b|Kfl@&-=CUC4WwN$S&}YAijk0Xq?y)`P3i7?1S>`
zi*=u$gJ(P%XY<-T7f&yYhk2+Se4B6bsZZE=+m{IA-+7AuN;mH81N)&L+9i*ElZ$;|
zkI185dZT^wMtu-p@++@C@%l|Z^O)TFOD_EKOZWryMmcDF_z~LEU-Hrid^mqHu6Uy#
z_DR-la_X0SQ)bvB@~fX-=m&Z2huB|!s(dikuRJ-mr~l3cgvn*UMqlU=UgJHK@xzby
zl&3f1*&AW}(+B#?z8Xj8yW-_z@8}nQ!2Eaq%if#+@~QvH?SHoMf~Wky_knBu_YHgO
zGN9oc!@C>f>->XXX@9RD_7C)l9+3y1P`^|8***4(U!y<b)i11F?ciOQU6v0@$G3XP
zp?!G#iH98e+(PfoJeM`@^1pN1qYEGYtpTI>Q<weiuaAx2XX4ELw0y>uTzJuraU+lM
zC$IeUK!5Q{Z(~0BQ}O08r0?dnu<?iVQkXrX_rmOj`HDYw!MtT3)GIxORsI(_Qpiu?
zN2ZY6>~KVOtE(966~+g|XS<Le-&07>=_914`f1$pAk4oqkKB{9&(=479lJh#iIu;<
z-W79>GGELy>p4BwUX+7=8s|ly*<|oSyUo!sPo<j=P<mZ7KOp_9LUQ9td3tW1LHYC>
zT7TUm@C)@vKk1)!*}AHJ^+EZRtD3*YhrNff968CwE<)o-zwn}d^0HUjp_g^h`0zLQ
zb<UxUk9M?c|H@8D$CrLdS3iBH|N5z3>B_rb;Kw>IbsvIP^MM~o4*gTVdm8c93+O(D
zyyS!GsX}r<^+k;J%)Rw7hwk^WIU3EPdCGpUi}WWVxgft{n<dU(Yu5v28)e+-U0+B2
zhZ8T^<<6o1_4dqrhV?#h-Hy{Hy;Jgcg!wV@3-h1sU!{NSpErEujo|}FEqv`3zqtO{
z0i*a)aQqz0Z}q|w17@Z>*N|WM&R6bT>}MBDpLzdbf6mVHYs?>E>Gl`qIlIVy(Elkj
z_BHM)yNc#p3ajcTX9`pPjCg#cklje3d0d6;bF*mPRUvy>h0$Kr4X5TSJICMWf7@q3
z`xwX{QI4H)9%Y@yD}6Q2>?-|)_6gQC<zhYTzxwDs`z3$M&ar3uPe0U$FZ)K{kEqXj
z0r3Rwr}+)mBkNZjXX~8$$&XL#oc`e#50GDJ-Bu63hF_%r)?NM3U-p8Z$iA>2&O6)(
z%cp$or}=4pldd0l7v^7?H{Qvu_tL&A@9@^b4f-QpKVx}v%CG-;HU9WAF7zSZ^WZ<C
zdgviNrVql#QGV}cofjH^dLy5B=N96%r#v3XZ~j31K<lG-C42vQ+tXhkH?{D6z}|c9
zdFM0x4QzNv;(XmXjQ1(-hn?4n-{IaZezonUbIm-bQC>Pb4!wU+AODQqHGbrEKBivp
zEY&A{?z8vV;)rwJo%x-U_O+*eVRGxgFtnd=?xY;K@jy=Rd3@*h#rOXD%$n;@FE%>o
zfhqScKcE5SCqFsyk9Vlw@~wCB@RLS=V$O#C^2_x@{oaW{_4y9c_ZYu=cjMm<d*|(9
z>^a-5ch<KDHNM(7^^qH=&)LAUcGRPN?P=dR9iD#mhpiv|$`P~8?6+~!pB<LnVEFJG
zrx(gUx#0PG4OnVW126K$c4E124}Ev=h2sV^v?q)Q^%#G4f?cuiq<8F~F#VKY|J18I
zyNL&OhMi>}*>Ure9f<RnorUZ<J7)g}ou9Ht&PCY?$e!EpkxM*(L%+-ecAdRtuk~Ab
z>x6XgET!WQzxu-t$9`(hxR8h3?4k9|I;g)fUjM8=?5zG-FUcXlern&i&<o=nc_)wl
z;Rkw$XMW>DIbrQ6hgbVze6m;CBR?MKoqk(i$bnbuH@^iR<iV4ElS`PMJLlqe(sO#R
ze)h(E!?$rHr}0*g{OU8V`ay2}7O!4m`UTYowFmhlc#X(^FmBd&XkVva+95yxRR7o$
z@%kmdd4(VPpr7j3zA#>lxBU25FMYG$(+(ck9peu1MSgmM7y4pd)h_<bOYImR@jpG~
z;6XpRXo1Ei2d)0fI~y<35U+pEajk#!9$J^IGxjOQ2hVTa`_9C}j#{X(<vJ_xdH*8w
zH`Eu=y%j$Ah0uI}&WG`8d?5Y6hxH5UCw`3!y^wC4^+$d9H%|B?uP}L(qld~FpV$t0
ztS9QDpX!NnX~#HehklToT*i&u@*79>&~JP}a*|(IJM4t=kRH32WOw7bWgfA|#)sY4
zfAhdR(GEHFOFfVrcx5-#$398tN0CE4P<vCtyB%TQ``F*u4;xQ>(_i~z`x|`5ai*{M
zS57<nZ`|pf_Ts+JeoVW{kzZIl>aov;=4;IFy^MZB{+M>juRi^-F49}&t&7^Fm-chw
z)k7|*J?A*a-*+zhNx%6m;*Gy@#!LIgjoj>={un=ctzPm;=XdCbc=dYc<eU}X<P%no
ze2`r9Vy?UX`Q_tZe0%11YTA!>LYRG~*X)?~Q|F>NzHczV-m#DDTX!Qrg<nNq%wu8m
zNW60FvUT44u+EuJ>;yeGZ{&~jk9}2M`}CV!!qy-2!TLRA#=2mgVu#HS{)~OMc>OaT
z5o7y!vA*Dwf2n`$TpU++K|k?=2XcvL=kUXh2+OCu{CE-1&KZ{|mvG;hC0UAgF<||6
zUQ6%j54{l2KcH9iNq+h*9ZEM(`FrB&z4E>vjL6QIm*z9Sfd9x2Kyt{(&eCi3IS*5h
zb2;aH=D&WJ|K9g0j~Dg+dhN5<o%7MT8{`$wUm=hB#j8g=p2V|b`e%P<T-{sAZyd>u
zPkxqm*+X)suu2d4ecHzxx!D`%ZO(((Z}v<-tQX$>$;WS2zkbqd_89-pgXkN*aL$NN
zVR|aRbq$ZszxZd;$<3cr5C2tp{=NR%m$A>}bB@PfR-bnLeuQ}W=?}dyuhnZjoEMrG
z<mXq~CwPycAI1@{`s1DE>?8O8>x*;EKPqC>pQvZ{eZ7TIK4E%>H+->c;_0Drq(A%w
z`sm(5`EM@$@?B#-_`#XyL*kwP2(vfPd^UgiGyF2^zWV^}(~lI=6KKBCCwl803jNdX
zi1bBy^TquDJ4fCrVPDUlNM{%DEkAw27nC287m^F=Ukb@DUA<7f<W;Y9<?uwmAU@b5
z{ZpTG`X!xST3@W++9#KH6y~>g6z0Ei@s5c7U>72qC+ui{LH+vf)7Q-TxwQv2=%IN>
z-^3di<H#P;7wcmjfA(SB8NcIeT^3KzBU%sHAMG1&>GaI_lUKTO<_mk`Tri@1^p3q@
zm-Zij!-_w@bFPMaQ}?>^vqSi!uRDEt<m3<ha{h*}`&V{Id3sOZ)hA!%$#~F9^%zI}
zXQzbqOMi`r^LX_{bRUgpe5ePi51Nn8jjX%oCH)che$ae1FTE3l;=Ly{|J^^}Oa0=>
z6YUoLB}YX1h;MxIH~4@2Zu*LU<4Z2>N3<^S5A{p?&gYD?ewmN-B_bZkp+5al-ngkB
znpcp1u;1n*TzAI2#}lNtkR8SgIgP(@P%r({5Az?7#xbIH@ZuaoKh$TxrQht9{@{oF
z+K)(H@A&8)x%3O~;-UVKi~Pn*I(;#o;>kk~p>*w#1L~LlY7g?;wP#%MBA+lG)k|O0
zAH#SfuYCHW9w<MQk6iMRBZb<D^3yk{9r@_3{z7sXH~mpwSiZ<7J&gTOUOVJakM{7Y
zzrs*|p?G{ja%oq4(xG&6L|*hmJKBYDp2-LCANyl|kV8J@@TeUq9m?l^aLNpS-Fi%Z
z_Cq=C3)|m7>tYJoIpc|M>l;0hZr(!nIMyS73fWn9$^2&@jHmSlzw8w`t#8KLIg|JJ
zzC$Iadi4vh($$MUeh_<2KJp9WLzw@sf93`GAvwqo=@Ge%uXgEydhwuMdSpJ}k9_h$
z>HJ6i(VlsSKXQ?uy!fQ&_Or^X&pt|f^jtsHC+zzPdLqnUv(C{&-+u^ech?OT{Nu0R
zn{AYI-)S83^}{xK?C9A?`922fpM0Nw^Wzt-F@E4EVREQfI{ouKiTvc{UrF~p3ch?t
zGUvxmeQ24p2aGbl%JVyYPaV-dN&oq+P&&D^E1jO;O}ci4jfZmhP`~o*3H^!rr0b{h
z^b=q7Qo8<$S6=>z<i#WVqCCAej`C{<>PNF^o~6+KnBJKO><B)|seSXu{9|{`FZJoC
z`RM)9#N(Db<;r0THLQaX*+2Es4@gdEorL6s>{ncuwCnuVc|5zUpD|1>`)c_n9x~vT
zFD&(Lv32>mpUivlfQI$bepvgfZhQSsH($MQ!#>J>oqe?q7&q_1oe%omIr}g9m)(2H
z9G6`-$IN!g>0V^ir|w_!Yr8Jq5D%?m-k*7|?K}Wa_-1dlryl)PAGz$q@dUkVkNK6;
zUx){MLF*?ZFMUWMULso0BBtsQALkRj@opu~C-)5Y8RnV$XX`U`KcZgq&V7P=6ZO(3
z?~lw&`LxUKk<Yw!j=?TD&w%#p>@<62-+tDdmmGij)`Ld*F2s3|^CY|*SKl{ycWi#q
zAK{oz57Z;S@l!8-hV;rj5>Fq=OV53upq%?ocH>9WKlal@{y1>v`^}gxpLpl`<kmm(
zvm5&7zKA{ZUYfppui;)mee}ot)L;EHpXE27@$9~lT<S4T-G``8zoaXNH|=ZBeXM#`
zT==PhZ=Cx7?Bfi?1G_;U^~1;`yRJR=e)7Bj(tqzf@XH^R5C8a4uX{CmPVev(?I!(~
zF28lp_fy_|`tBskZ+)Owc)^$TkDPu_SpR*eV;;&U-MrU7>#uPTZ(Q+44)%$h?4<oC
zz11$e!#^`F;@KxW;75AwFZ*Y_d?#f*_@ngG{AXX)N1xeQeETkFrXJ7udw%4hKX}8B
z@r-tN%8Y%uc{Edw8Q)LQ4|4J+ya&b$zYKqP=ZBdW+Gpp<ZJofMFgs#=_{H`^{6_t?
zK3KP`L%xT#Uc~x@tt;x`Z`<$EPky(2_6zFgcgwH8>Ji3=a_+_H3qI+y{Vv|rN6r-D
zNq+n~-^3GpBwoA5K|DS2e$Dt>->ujB@At^bC!Sn*#H(>O|M2O(F&^j_{`5=zc#8VL
z?}64a{gj{H7#I1quYU6mALKNC{3HF>AMY>pLppx(3gt5{{5D~7@&Ck=OL_Cdcss8!
zuk;JbuRMA1Wt`|UJ)u|ll&&51L2}Z6yc!?v!W2)|0r~i!`pM3yUwi7qoASnyJtmL!
zNIv!AMc8^kt`u75lw*I@M_%<=cVl|YM?UfTYrTo-{5j)7o@m$kgVtYqV%>@TB&Yi5
z3B4tUcMR5B>mB`o+BGiPn=)fvQ=fH8IXoLj--AK?@*C{)^xr!)VRC3+{p7}ncz&4q
z&o9w$_nz`Wys>lEVeQ9$;@kYAPk3j~q|*oe=D+b<*lY7cf7HY7$?v>@eQ=IOuk?#Q
z%8tut-LO8e8_vD|Ysi^<?KRuGrTriKX&m$ivg7O?{Z%jf<h;?iIj>~Tlvf`<^%I}Y
z;nb&J*0*c!+To<HEjPVXKR<%}>cxZpSwF0o<e?u>zwr&tH)#CWIpyh(e31U7>Q_#_
zy2viE)Am91L^?U~WFANRrW`$G-|=c)hU|lW>K}f!CtiKzF^}0n`OO#eN;~Y9_MmZs
z>WSlmXZ<6Gde}ep;LmzUUi~vp_Ql%MU;e*#r_8Xk?4b6@8~aTT`C|R}R9-vskss=B
z6_P_gjE{QgnXqw!c)~L|t^dO6=a<EP$fsY%3Gd?BA^QaV=hw++-rAqZC!QVR$MK7`
zix+l9IF`3xRUe+zLw@tfI9qSnQR8lXWyi$ZH;|WJ*`Kkm)*<^={il!dJVAZ(S#QW^
zT-j50nq2DTk6736i5K$H7krxU#*N&5S3;QIWgN+;KIbOn!J~5ei6?TA6B;k$BtN}l
zAJ|EHXMC)e;_dtB4L?9X@Sr|&v47-Nzw%MP*eUIhpWa!|t<&;Zr^!iw@uA;%AYYW<
zyrvh?Zm5@Cu>SDp=mEPxUj7@slHa-}pK*7dV!fkxc&5+x2lR)Ym>=YgeyH~Nm*y|I
z^jkjl@jso5%C9{6h4q^r>kr<<L;cW>{3+DF`m~21e!O&Y;Zr}Q)1!4~yuULJ<dWa`
z;vbUJ`MP?v3#BWk|HhB}5!n%P;hSI0zL{_QX7WmRe{KHoOVwk&FrWCR>@@$M-|JmC
ze^`AH)h{1EnSU(I{=+9ucw_Jq4^N%>_an$*eKF6?Z}pJNIgk3Z2dxMAQNR4wX?8<<
zc+#%(#h6Y`^+D|!Ct-3!?<kBL9-a58j~vFIeB^=F2lkJhh4eyy*=@WSKmFA|`N<Q<
zpB*MY{V`tp5A|0VsvnX=x^~G)UyZYVK)jHPzWDxLIsEE}_Vpj{>Q$exe99Rg@%U4}
z{^%dX3;obv>B`YtVf{4D^nl&dKlP|rJh}DL{K2zw+LJ%>pg#S?gZ#qs;YFB#qCM}I
zgynOtpdEV6-a_jnWXB<SAiECPS9sCs!?xUV#Gp}7y-<Ik@qpTe_=D^|G>%X?$e)4s
zF)*$(sj&L#t9=%^=mmYZZWsrCpM3l}<B&q{AW~@m!k^}Mi+7)B9p{JfGsMepAL0I0
zy!x9(zZb^8_dTQYLw<mL3$)&o-~L>F^@#VrfFHsC(I3CtE&bYyKls%@t+;r@{V=~m
zedOT(k>~s?UODkkYb`kQduI9EW9p}Nq5i5**m;C7zM=D3_{GZ~UF;(_&e_lp_2I=l
z<FA=l#@RV6|Cb)v|LH$H5?0Q<a-PheGau!rC%$94{=V%Nop|knjoo&feaAPCnXBQu
z@Mh7tL*G?JbT16$kBH}p#v2-^i1>(T+#x$0?VkOV^BCh1=PAFPzb?$LHty`4dX2mM
z_BHA;uJp*bkwd=NPHay;`(FLA&$GX@k8=)cpDo@wg7qZwD&Bf#KTrSGow0w>zImb^
z>lpcsi~87kc3e5%xml;Bdxy!tP+mL6!}*wa^HP7U3+l5T$VWcopdR{ZU9x_uPdWYd
z&O*ENk>A1}Rjvv@x$w24N1ixeV}S{~AAJ1>7H;_7kG+N7$2@!TggdU>bKZt`tL%cm
z>&G5=C$`B)mjCjR2hQ2Ri}x+;4!woq^-DZImYp=N_!7pGdf0dNL4NXBpCG$RAK4${
z#y+z5^om^CH*d919{o3d#@~KM{qk9tyg&0kjJ)J@PC||p(qrfx3z7q}3y{BLUu!<`
z*ZJf0mtLD!+JWkeXnxpV^5@kD&0Ash2$N5Jkli3>M1Ht@Rpq3^C<l8Jk(|mw`N(g7
z0Pz9Me<+`MW8Wo=*VvBs%|HGxex&n<@nk%mqsp&;^h18*Cd^Jb7m%)=xW3p&Y1e+h
zIV}I2{grM!$xkooyZqWy4j;yY-10}>jlX#PiF_MB^GP~8?)<{NvUuOo7-#iJ*H7!H
zad0j`KK*iD#%?*MHs0)-diCG9@elX~?3DFW{rKX4vUkp(+1uy`C`Z4HAA5se`?{!a
z{0{9ZXFS<s{No32coCLPf3%M$ew6;uQ>cHC-v-%R?UGM^<3MkHH$#5oN<RHnj(q$p
z>jiz8GUNVKynUAOwBC}(dSc#Kx8*l}(E61^JV15;AM{QA<i@Y@!3#O*fw1wX$9Th|
zbrt{YA%8%7=9T`D3m@Xy6Y^<ay!z;o@gtw{&|h*HkBH=`Li<zw^1d72>NU>fqJMbR
zj(kvle30L`ipP`vj&k-1>gRv4AM%?w+U4)zO+U0le&bFq_2_?`58AIn{+sg*exUi`
z9!S5)OU^2!$M$LZW!<4S^vb?;;#!MO`P4as8c_eqfp2<fpC=vS8DH#g%r753p||RZ
z@$yTTUq7Tn^M@YEXMBb6qhI({4xiHHR}aK1jPdkVdGbZnE|g!t)hk{;az=fnKjyJ<
zu^%;0=rj8&pK^HBo_GJ^$tT@BX4izRH~7a39^}(c`Gu7euRMKaKlvlpVfD%v$4NbS
zB@h2anE#_5ey8&Ig!~-k$OZY~zB{&_@VEFi>M@U%kNtJu!`~Gqr*ix&-x-=u><~S|
zGybg4=4aJ8Hoe53b4hy1{?R*Oeu#ZK{j`6E^cb?Q^8231zFazVexg3-&d#~@*Ez_A
z#XM(?zjd~m-?3RI-PcFdf9H(OaqUa^ujFTE@Wmb(ALYz5^AnOMBE4=FjdQz@oq%{Z
z{?0X=>(T@B$+*%B<>;&anU9eE;uT-)mw92m!n5_!JaDhzJl6OaH)y^>`HUkz*;%L`
z^0QO=&z|9rercB<FJ0Jt*FWv?SDbU<g&uoXjc56#laoKiemHNozcSyot37&(SN@y!
zjE`}n56ZjOA_pF{V_uoZ&e6m}@;Oh}A3SQ`_>qf0=(}+KTa@2?RbGAKA^!EpI-?xD
z(JthlS+C^NuQ)HXWB*E@$&EjLs(FAP^3Yd$2<aRC*meCPpZTdD#$UbWEi@nSqTl#b
zkAA41y!3;d;?<8QeujGKwR-7?demzi@yBn`Z{_E^?}3LuddLT6{w}I?^NoK+U)f3X
zNj>B+F6zO9{yR6q8#`eAalb9U-%WAOWIwAsdkFER-|UrfreExreE8+B;D5@D^9D$d
zh1nN+tl#nptDn6fhw}pSHI5s7HLmnbf0U=E_@wvr0-yHJ%Bzn*#10upVe^?Cryurr
z!p<egB}~8BYvV#rJVhSpyZLWD#JBMyKi<8kR36{-)!$pxu5rN+eP(Z>T*e8HkRQs<
zJ8$GCK=XpU<g%_9Px2W@<@5`W^ig^B;zv348c%3^*dO-IyozD*^h&z<XMXZ$<aaJ?
z|6rd%t`r*g6w*WUk>2S)`)?e@%TEv8Tbmc=o&4&>JG*6k@F~ph%12K1so%Vy*N{Hh
zC(v8_1?cxP{Jsr;guQm3$8JOVLf`pS=7D+Ty#aebzap9!_KoHRd%zAF4=CNf#CXta
zJVEn9d-M*{6Z=ib50*|(?Qh*bx$jZV`9VbT&<FO~dma4ZEh2uPc@516dIZhqX3>1r
zfAdnjuyL^t(%Wbs_@8*P&v#BO-h47IjGK1Vt9|Em?2`3be)06lIEoi$m)tYpPyfwV
zdP6^iwWl7a9_{F-`yGA`z0yDDFXV~&^dBlGteh}8@kK7<#~(0`<{5o8F6^T9UA-w(
zFMc3i$r1Tv7vzVr9d;Jqc!qdUjvnZzc>Y|f9r8pu>9u~ss4wO%`;I?hJdlTccoK&C
z0o6mVV|?tF_MrYj{lrVmC!KuC$%l9CiigT8Ctn=Lm@c1q<4oVBn@^D5t6x6$g8u={
z8z|m*>$iPR3hh_%VP313{j*+37tjB(-_c+4L-F)SKiN@o;9ov+M8vOtJGX`GJbj~o
z<iWczy9e1b^~G@@r~36ze*H<Id{9383TszBVe=>YU;H2Qncu-LlRx@9{37<$`ma9q
z2=fP=Hz{v_tvr8?-^U-|Z>Yz<fn59x^713?JEZd~jDz}&4?mDx+E=gg{4nwI=?A%-
zr|^sT-SnA10L??=6;U~;eHis#*t{{{p?V^kzj3~r2lSWT;Lmuo>(Km2Aw4FCa>mEH
zj^~K@i<q)w;_;9|ddwa}{l+U4Prvksy@c$D@nye+_5WAzZv5L}@4Q{;mv->YuHp}x
z_rmni{z5(KC9iWa^G!N_jD!60nb+<O@Ss2Pxeub}>eY^N_=&uzN4?6^2lL*1Fi+Wc
zcEh+t#Jm1!Up&OWeB_eu97#Gk$-@prWIxO!<Ag8$#<wsz*;VaDjQusg=%sxXIg}?4
zd!nCJ$j?^4_O(MF=qbCRUVfEw`b(~ec%Vo0KtAInpRjz^3t{ExC;Knn_{h(Hht7HU
z9qg%i{sVhUKa4m3o4z?mvfr}r6plR6KmL#Pl|OHM%r|`CAKE`b{IV<jCH*!J@k9>e
zX}_Qz`XH<x=OW}V&)G5kkj_5oukj%V)L;Eqz6zC3p?MFjm(Vy?A$@~UFYv^#!ME{n
zo~NJmUAy=;-pXr-Unia38%J_M<BC7?H0BqNcRaa&)UOzCpC09-2ljQwAIeXk^wT=P
zt~qC?U*esgYhQW(F}d*x)uTNA$f3ORMs}Fp6?VU%UHvrQ^xwD|Fa2hhv`-%WvHs{U
zUihteAqTl2{Zn3;9-4oUT<XWulo|I;+Q$PulrBtv%tQUQepsj2Pxe$k@+fbf$R|vn
z`48gpN*;(`>Dr^u`VZAFOfK_^{N@+CNlx}ydG(n`^pibMzk1kL{L)`&{<0s|7hyaL
z<Hvd@-MG+G^^jLseqsF-R===zjidU^fBYLC<;j63{iToekDiOCx8|dCegh<zd1oBV
zduUwbgZ7`$e#QROJs&y6d%q*ye6;_hm*m$k{v`h~h5DaDc7XnBkA6bq0mT~^;{@ps
zdj+*)U4i^f`ygTUMZ^R91I-Kk>bHJE^6L*fumAMZe4=-dzC!u~^$Ti`{#BuRV|(PK
zC;I9Bj6Rx2P`}6l<)^petvAkfq-z(?5%t%2Y2W%M-FhIr?u_qN*=<N)*kj{NKJ!s|
z?W#A1^$(AD(J%GjMLy#o4COOF)i2)ouvglPywL~s(o^4Ak`K~<<0DMpA$tSa2|SY1
zxrK7#^-DZ^$-dd2sMmV}??Txr{n2mj(_4rK`o<0$mpBgmDS9H^{4tNA`AiP$lJfW$
z#zREqBI1kvOQHR<b6@*Ozl-C%T|VnGy9lkf<a4g9ob}oK!<T)n_MA&W^7yXKcaO?R
zkNL&R@19ye?T6Lldpq&+LH$;5YJIQ{#eDL!2gZy2!3Q4Y<F~Wt><Av|5&wnXCm;Wh
z9<hVwDZREX(i`@HpQwJQf4*156LhW+_20a}C%)(ho`m&VKh0mrZ)V?=BOkrc&dm29
zGtNiZ4fEXl4gPKX{S9{2K7~CpudNUKWcETi@1yu}>S52+r#~rVH;f;-p?b_m<7%C<
zKa<YhS?~0pUK&UFv@2{qh^Spi-`Tf_c!KzY<c0JR8fR!+pm_izzw|z$cSRA^%f33F
zvyRYH{<U{=^qL(opN*UGrBC>S-XqEH9h3a@+B~Os#$8yt`ivhx6vlC)-_r45J{m9e
z887;$UHvg1)JNagLG$0ZvTO32N09w<-;HnUmGe`2tbgibC+WYiezIfurq}wdK4Ivc
z6uabouJ^juA?emVzeBD)c9|Ub^1TZG2VdS*(hEF5<?*b%-wz`{^m}XkNqnnshkLj9
z)wY|?)!+x1AM&yP^w+w}KV-*!Ulr}Sbn;lo&0FgbzmFW=M@#qjBaJV<jDvRY64T|=
zZ^#}&<swSg4``gVFP>ebZ&PNRf3auIuj6`7kNLaig?`a{?U-NmoqaOztq;QFwSM4J
zd3L}!8h`QZAU>3*m)05e$)_HuKI=ApV1MbWc>To#{iHAWW`~5y!EUi{>h-%h#)my%
zpX6f)r5hh|LgTA{))o4sJbNIYb7FRvJko{PLGxaiJmx9AvW{6t<hQ=ahj-(_&w}ic
z@rB~q1O3&meICEbI*wocHIDAP@M#>a3-nk!cvGKt@PiL_)%vMCKTJQ>Zyu75pCCW}
zw6DL;(dirgv3{Dr#?}0wr{<^mq+jZ@Zqp<3&`<iWUWj+&Wt{oh+NTG8M}u7!uYL01
z)p;bo>6I`$#$H($?VE&^Gye1pfAp9<##8<FCG4wy(Ia}ALOjto_S1NgPkDObJVd;D
z^xybEdP?t%W7JFfBwl~DqkeiwFX)YQJkneBiPukjiI?B_I;WxU()Gi9H-5$o#(qX#
z%{%V~%s>6MZn1~-5U=ce<Uzm8M|x}@Za>dXtB0MA^+~6f)+O^+{JM$X*HfSMlU(wX
zS36LCewy_rg?OUR#zi~o6K{OjLGx1mcr#AM!@Qsu?3nQ(zj2~pzULsPd?_TSae?%I
zrk*}|=Q6AA|8^mszOw`7hj9|dhkWFW^GLjT2-o@U{x6*TyMd$lul}wN|CK-P_oSpd
zpOMe+Pz&34@U#7{f%4m3@byWv-!R)KzrXXL&1T>J8y_7w%DEbSq3`@S@7?$hc%Xmu
z+PJdc?5*|1?@Q~Sb0z+}F#q3qC3LQc2jgO&?tIexV(0k_-W`%lz0f*iU&nu>-{iy-
z{bd*F9l7```eT2g9w;B4?K`4>O@939FZ-xp<^_GW-rHvwSMO2uUwQj!>!bYSg6su(
z$ge&74C8@U`hZ92#!LP5%D#yl>@VK%z+S0We)Sqp^^>1G($#N#lqUyNUOD4HZb*(O
zmw4rkxAX0opPcNVa^#_B`lo&Q$Y&i^zx?)H>SH&xi)a0&@A5(Q(gR_<s^7Sq7vzNa
zCztYY%H_kh9yfX3Mqfky)<1UBegmK8i}@4LxTw#%Yuyn~kMKlaAwJ2YUh_qH`Hhcp
zwyx7B>x1+-zT~10!qyA&K;wxQ^(rS!4)y4d^#qcW{ZZb!O)mB6FFDk2{l=59^8B$B
zlAj&*PLX`%$ESUt_Q*p%<IC@IU#)%fQ9OT`KV^O9hd_EIpL+C{-5|I6`EB%DfB5nI
za`oX`*#5>o)4tQb&3hF{KJr*+`I*)|@#Mz?Ka^h$`EC4D{Nq{Jyc1S0x%A&S;)mYx
zGv(t)*$=5#I=v;I{#n<JyL9!~kK2dQL;Nal|71PDvv}_Sp!o{fFZ#(|v*-NEi2B72
zIUi>q%@_I(>7RC@-SV!19W#%W$AfY7dsN0px_Y%oUg_eM(~fyS&*g{mvzOY5^PGL5
z59*`$(&?{y>4W}5?HDKR<3<0Bt8o!;KC#bGy8PM^FD$=)k{@rv`k}wlp?MLJ|8M_b
zeE2K)*I)ftuYQ}ycsC!6t9_P!>5ul*8}-w;Xh;3pr;p@QFT|&Q7{}N@@#HXW`r$rl
z%8Yvkywf9g*?E=mfW}3Cp>r%evFFBJIsJ4`roZeFJEouZWBTKF@9`-g`P5@Srak?|
zzj*seyqRClFO0MGMLzu^m+{0WyU32Q+xB1D$BTIVBB%9%zQ^@QKKnvwKHA^01LVQG
zeV#BmloxMbsJ!{6Uh|IsYFy|oImBy+zTv^Ti68yNAD;Oi)@%L?eV3pAK|lE$&P%Ox
z)@l07PjYTePWnzy<l{#<SJIAoN3ZFjF#kk(<)HOEo}c=?NbQjyALP;wKSfx(>eX-Z
z$<H1d7yO!Mc!&H8>#%zReB(p^w9DS;7x~Bw^;bUjY@>4?m~!v(17`l+A?NJ+ryPIC
z?@E$Wys+PybnnMc_|f!_{q&GO4s3Xz!EeF`{bUE+qe&-+->VVUAN^KNJ?20CGmrSQ
z=8Jhn4t9)QsGmNYU&^yT>>D(nj2k~qebU!EdH6}AKQU*+-yIQ8ukZk^JJ38Z-t@va
z^J|TV{QAQlvr~~Tb_Q?uV|XU7bYb$D2jcZh`{X9CczQ2C-XhYc6sq6v9yu2u`sGmz
z?RxGa4R-GK&n$cOL-R~4$SuEprg@4-{yn*&c`Zy{?V4ZomwpJ7hd&h2`d}Ti-Wo^i
zt#AshSL!i8#Pg$!KfTtTa^lHFf0R=%ec=!O{D%{6U3C6IqrC6o-`Lmi)6}nB`RS{4
z^;=KPe|`>m%wK*|lv{oJrylKFw~dSS-aUeR(($i9c(opCPdn@iw0@We<_A5YujUuK
zC7*iCL-paCU&3EAzx9j$yT`+qcs$~r9~t|r-}HmL#>KglcTmo~`TOjrd2XF&@9@FC
zv6pz`7n+aw#|L@LNAn$;@8qx!<JtL*cIYE{)DL~{5Bcf-J{8n1ls|=dfvJ4#E7UJM
zLiP^g6|!@!A~~Byd{klVKYfGr(mZAV%uD&qYiNAsGcV*9Rxe~%>;uG`7wotBr#(nM
z>FkWKd8i%pN<ZViM7z@MhsdE_>FTA&>;QYrzUr5Hgz=^y(0<2#j{Zm&ww~xW|I@s)
ze(0Bc?s?pUI`?+pt6g?WzvXB5$sueUl@rF3cJUw{(i7z&yRE$W$j>4#e&`)EpYgzs
z(pPp~`_du57P7niBJ<AQ)3Gn~JM-2d_1S;2m-Hd7OYAZGOwWwF^@SZb9`>#Di@jpc
z<!2A9w<#o_`k{G{LVQByB9b#AeTDe1i{wrry{e1G-Mu|IgpDV?(r>&v*NI3z_K7~y
z7j{HD+J*F*f2Tc29{Jau@m`1?nSc6c-azx;yfxn9p>(KU`ayrm1Ic0js*nAM@yh8p
zz0r>R_*cJvir0?(QEv6iXTHbzD4+gmN7y*f5A#}>ol}lH?1%oEH{=w?vw9+L>Qjy%
zreEZvAJz|cf<Dq)<)VIA2gI9K#z#Md>5+arFXCTYM}@7|_*M`5D&Bn2ALB2qzslps
zxCrABZ}^oC<N6cdl~}iwS8qge;YqspL&h5~=COLoBOkpme&Urgj>;Ja^Fvtu;*Fbl
zNM7SEY#vyzj01ZG)eFf7=@rBmH2<M~rqDhQs+Ybv593GTk6$I8-zOdK_LIW&1oAJT
z`izHlPnchbZ}H@2m+;R&(jSOlNZ;@%pYd0}e9(MT4$thGeUWoF`$zUgIr>2_)NfxY
zpZeG(`+fGz-zgThk8<wCz6j%2JU?;DjQcHoLhU;5B@a2RpZ2f%iw~$A`60c9@<Tj9
z?Ly-Y**Ef6A^EEiUn#_Uv&c_{>No!E1bxH{{vf@^i*pf3zaalznB3Z@4|u1i<T75`
zWv|Tp7|$+RC*)&4%)5x%iKyQZ&BrR#&xquXh@UDnPR0!yx4NiY6-Hj1pEZltxrpqx
zb&Q>fNY40ua1Lh%G6Mq!I7hNR*;jZ+?!7(#hQFnpc?gx0&${khU;W;dMgK`V%JH+j
zqt*^PC_npVe_)@&9`FnK|H`{3@P2}wCztlc+xO@ne_s9eJM6#nAM&|}<FDa^A7<Ur
z56BO&PO*pTvCff`{L-zP>@xq!J_{e}{pMeuUT3eb&)-m9IqR%^5v|MmE8TdKlYCG<
zej3ytNKYa8pnj&1T>26Fr5$=EOwWu*^b?Gmc=eLYI+Q}=-zxGmBl4%5H+l!-9S6Sw
zAN&sbU>xMbD}O;)K6=0IjB_pL;M(W!(?5PNe_#Lk<?1Cjf8V*Sb6V%n{8e%oZ}O;5
zKI5g|5!sIv>Q7zNPx3+jF1;qNaWSs)n>WS@Z}e2Zq4A2S-W1w@rZCEhUueJSJ}>g*
z{Q=ZpNKYcNGxR5gv0Mz(gA^LCi2IMfVa1={Iaecv^tZdvdSHB=kI*0YzV1Q%9dhIB
zotgJ#-WQpt-i`g_u>HrZwZPmB`mew4Rq$rK?em<^vKvtS&^{1q2Y&bIYv%mi+5;QV
zyCkSR$ll4X-iY=M;{9G7{jeTQnQ{JL9bmulZvABM@oQgcy@TetFg-VJ(0o&`eXer+
z4e}^w9<USS*DpM=1Jd<hJF!0X%8xhW?ERZ@GtcR%d7vJ4fnLZ5wL@-pS^eHC$%hyH
z*1vUUtV8tM`LuCUzju205mpag$l*NCyrVD1D<ZkQr^bWvQ;%}O<kK&99(o@pAJo5y
zanIvC`-|`W^_ex-pI+!E#IJR{3h7N1((j1ob;Q_@X2a%X3XMYwjaNi|Nfp*@m%g$S
z_7@S|KN=r;3h7I1mwu?vII(B+So`ATgLpE2))W2+q+iO>Z}OeC-VBW#vHyr+!#_4;
z(IJZs8F=i7ZAJ{4?O*>6UvT&$!xtYi;OG&<m;6_Fl_3L%4BT@2m5v!P{G8(lk6mG4
z)aM+|3}gl}1DS!$KxQB_kQvAfWCk(=nSsneEd~b8cHr?-K0T?-n0e^WdQfXJ<V=}?
z%s^%!Gmsg`3}gl}1DS!$KxQB_kQvAfWCk(=nSsneW*{?=8ORJ|1~LPgfy_W=ATy8|
z$P8o#G6R``%s^%!Gmsg`3}gl}1DS!$KxQB_kQvAfWCk(=nSsneW*{?=8ORJ|1~LPg
zfevTjp<<IqzBlaUvMx?}`nkn^Fz>jsSzPVjM_)hsnAgf?@f*+o>G^Bdd9AF%ozL8R
zm7knGzO0M)U;LrDS6ch^GUNO6(BbNx#m@|61~LPgfy_W=ATy8|$P8o#G6R``%s^%!
zGmsg`3}gl}1DS!$KxQB_kQvAfWCk(={g#2J7P;W(je+l#t>QfkAO8E%qsEub;{E3x
zd)aw!O)gu-emfw!W0`@>KxQB_kQvAfWCk(=nSsneW*{?=8ORJ|1~LPgfy_W=ATy8|
z$P8o#G6R``4rU<#o$CB|s(a(_RCn-{$)aZlG6R``%s^%!Gmsg`3}gl}1DS!$KxQB_
zkQvAfWCk(=nSr)2@X^;^+V|WKk1H}}9<m;^ZyvN|a^^}h1DS!$KxQB_kQvAfWCk(=
znSsneW*{?=8ORJ|1~LPgfy_W=ATy8|$P8o#G6R``%s^%!Gmsg`3}gl}1DS!$KxQB_
zkQvAfWCmtg2A(@_{5^;N^3k%paNLJx|IqI?m{Ml^zdX#c3YR;c8ORJ|1~LPgfy_W=
zATy8|$P8o#G6R``%s^%!Gmsg`3}gl}1DS!$KxQB_kQvAfWCk(=nSsnezhU5r0biJX
z!U9hgeFZPy{>MxI?dtJGf5(2KK)EBCfy_W=ATy8|$P8o#G6R``%s^%!Gmsg`3}gl}
z1DS!$KxQB_kQvAfWCk(=nSsneW*{?=8ORJ|1~LPgfy_W=ATy8|$P8o#G6R``%s^%!
zGmshR2Mp|X;BgD?u<N9<xAC;~W{^2z{}IE6e{9HNLlz%0@YoUCj2JT8zy2M*;P6F;
zFFs_z(IbX0`LFOQLk12RxaIaM9W!G1ImZtkyTZVJFciJpg<aPja>Abu9bdGHr~Tr-
zlMnsX<3+?5KfdUKcOH3bSr>bkh1_~(pl>iR&w(FadCYY$lo@-MhrXc=S&Ga+W*{@r
zbp}4U&Q})x>7E~GWb7L}7||cD`Ns#AnDAto!e5^;=**`UosdjVmCNz(&ufYw`<Q!1
z{$$NZb{SJZ@tuNu9dX5}cl`aivcKRyTitQscfLEW%-9|t*q@A<fy_W=ATy8|$PD~D
z0}meY*exG=`HiwuQI7_1`s^i-F8x@U!cOg9f0aAlo2)qd7mokMW21_m#&+*_V!!%Z
zm_C|!{RQ_}WTO}Nm}g>{an|C&dExio`_A59UgGT{b#BsK@#a&zvD4>ee>vBw>PM%K
zQ<lGfGSFT2uBYqWV5ysTe*Dve8WGw36b>Kx^>>fjVseo}=jJKwZa;RzyEZ@jFTa_y
zQH9pEh*j^Q+U*yo>g#X*aqdrZ@8bG9U+hbF&9}Pz?LX|-uO71PgJl)=w0+L)wRVr&
zTfFs8ZzuPvyBJ{q+KuhrFSlA>yLe0SQ{~@nu*;`cpZw2(jk?$_dHZTTy=50$ZMU!T
znB~e0w1a_G?P_<GYe%1Q9o^4Bca49l-tO;F-)_PFG@cXH-CyQ>?|<%*(!1V@@7vm#
zrOOOt1~LPgfy_W=ATy8|=&uYo_jJx#_Z&0lYwp}P^Y#9GeP*YLLsvR^Y?(3h@W1td
zed_DjuJ1;>d!6ZRI~3m`?Y8qC&oA@bdqrw|txGV<@7xTg>gj5HcaKX|{q54PPOHZ|
zzNj}j{GWdptxAt_dS)OqkQta&8EEyLNKcn*_xjeV|1+!VN$z)MAT!W;48->nDLdNk
z_Y&3z^$KU~toQX*<JEaHB1@bZ$P8o#G6R``%s^%!Gmsg`49uzwRPAs2dU|iY*U#;C
z_U}}7*Z9q<cBHD`?i;+<@IE7j@tn>%+N_HA?f1#uZ<{o4qrde(zwgWM`~LU)zN`=L
zuMg2K3img=ko)ug->+tVx`!X{3w!I{u&d+B9z^VFz5T8HeWR{i_nDLCZftnyJr7?v
z_dGLkx7kOYwA7+=H>3+g`K80NHy-=ry^j2!dIoKF$k?q%&eb^UxM7<fzsy_>;bt-Q
zdzLx=|NgzK-r}Qb-;>ib1D(J?f3xqoKkx4!IgjDzWWL^?ugp*8=guQ<EnC%tsx_e1
zdi!g+D!y`hW}p)ou)apL{#K#=X6k!a_bJl<4QK3ctJ<kb*G>vk_c(neKAwjNr}`7e
zt17I%6xyHm6+Xz3lDEIaJI{;jr2a{Vb6xoHC2v0e)Rx!z*iN%ya`ZQBmWP<X6Ks8!
z@qZZT>+goLd_DUe&i`me{$D3CV4Z}~F1I_J=Vi~%%UPB0c@F&W%44p3q0E?hsG|qn
zZL_QP->GzVxh}TbUg!E|t#k4I`abX1XAC;?sYNH08M~K<?zW@3{>(sTATy8|$P8o#
zG6R``%s^jYz<q1Py8HJ2o-dyx%v$FNdA##@xALB{FWB-dLuR127;q2Z-CDc+f^3i9
z-yZkYc=xq-yX$vT>ek!W_~>qV>hjW8^07~?V!P~_dRmpEr_24@yY0I+Kl?AgnX{4O
z-=7ETV5eiV>u;y`H_QM2fB$gsrq5pT=+cjseI55Y;)+x6`1^BZU2NBS$-culfBHIa
zDf#kywEQ0JKfOomw7=6w-mdzk(f{e~b)%>K!k%ic>if1<)8qK{m2g*iZC5|KDu?pz
zT34&;F|Yd?R>|E{>0PyV#uL`HchUdp>2S)Q>9lyfcNII;kE<H5S@nC7-0%MNT~b%=
zbS|HN|NTQRzftzK9`qGItf$typ7P^bZKv*d^tRr__Up>M$08fOxW_yb%PNfZcQxEo
z`lTJ?-&MJD`Tn&|r1k|HEOqnFkAHelquFzkyVu?Sh7bJmxx#ru>O8yI{ajD`tG)~D
z>i2W{+0*jp_WrYf&kknH3}gl}1DS!$KxQB_kQvAf%!&+D-H$jwNTK}=v~TKe)Lz8y
zYB#nc++FSEdf)%=#$|qbm!DbDR^~qcI|HeC$Ue3kQ|)y%zS-~CyV{=eT|Hk+m9IMg
z)8E~8-sAaYo_nu|IDF*S-#u!J$wg0N)EB>B?)SrEdXzh+=kWV~2S^VhI=6@J34~K<
zoa$n`$H%<tH2inV+P7oq1HCccc(1ZQT}^Md-t?9F`u3=2=`sWDWnkvt0}6k)HpaKN
z)m&SrF_5y))}Lmv)B2y~%?$J;11bN$ZoIz_)>Z7uKF^A6b=42=uJZO1buqO*#Q0dB
z^zQ0MuJ=E`!^*t$EHAU7ZOeVm3}gl}1DS!$KxQB_kQvAf)Mdaqhy7s{HhccqY<}`|
zDpv8A)8GG|zNh%1Up>`cZl`aIdv}dPU3=Tx`RIM8Vt0*SuD5^B&%Qdoy=@P2`<a1O
z80f3>Da)1Zcq@OOsH=LD@~2|G-f_qHP6>Mt+f(vt55M_7wP){ByPNM$`<*54zEha`
zezb0Us@_>WKBXtk`Yx$@a(w^zy{~4^n|msM*027dU;T9)o3*38ou6OW+t}MY<n}WI
z|C@o+)|(L;vHyr+!#_4;@gYkL8F=i7ZAJ{4?O*>6UvT&$!xtYi;OG&<m;6_Fl_3L%
z4BT@2m5v!P{G8(lk6mHl|E{Yo|G)OU<;Gkx{>>78>EPSXoxaZNWvAe`HXHnjp`V;s
zrttAASG#?gJzgqX#Q~QeF?9C(&M&KQ@ZdGCSnbU#i@Ip=PTTydSMFMW;RgK8^xqC!
z=kP@vDSY|NZ~S1?*$XwQaIW3XedzeD=5N$R^0W&F9eeQ-s~z)7*($b6j&{|vz^9L2
zZIAiJmCa(R{Z@HVE>&;4<2PRUytj`2?fA00uwC-LxbV{ZG`1a6r0|4ci~nJtKfhjd
z71x;KlhYQT<CP+X?=AM|UswIh#G(qnfAdKX9<b@1MOX2r-A>%^<RARKNa3U%rd~dJ
zqlrZowrd>D+JD^86E`1Sb~+w6?;5MjvGI##yK&@Ehh6&Y_LGzARcalJ@f$8Q_rV8!
z^tH05uwDAm?s`sqb)_*o&pub9E}r_m58uD?%Fh>7$j;Wq-(Ga%i&x%#Wf}3T$q&AE
z!<u)NRru=E8^+x9$GIC-Sl7R}?e-%E?0V3HMP2;V#w-7H|NkzFh;`dr{D8mTbkyKM
zjVk=oN)K=N^cmv{Sa*Ki`12D_{OV(4%DUKYJ&X0-G}oUm{_LKI7oCO|u6p-Qf4^sR
z+0%H%Q<omF@K?u_t>WAF{cW}}(_bz#wwH$uHy-eZ<M(~Kykeg7f4BU(kIdXBt#Qmp
z&fIL=Q|0<6t$fq8uiak`y7Js18$IyEOk981TE~x@``HrS_P4KWa>wR(mvHd-r*FRd
zt*1+PYB}GhzwnnAOSr=ZzklSd<)1I%i@VPI_>PlaDj|P!sR_e3x%|{|C6r!;TWtC8
z(DSC<TEgSj|LFFMANS8vzu*>kzjpb)i$7Pw%a1Ll-+AjZWyGj=LstCus@I)2wv1TU
zKH)p!8jlsX&+(^6i&pWi^B=wZgggIIMqGZy;iv!bi}#f&th-Lgx9Ue9+WN?!{Gv$V
z>z`kE^fqtZQZ$Qao%6y`XO4Kdg!H2d_xtr81AemR3uOwU{cd+S^4+On^Eu+@-dz6H
zsi!R3c<YjDci#QL4>k_}<mVq){*XBvBZiDT;ydpQXuP=1-M7E8**gUseEG!(Z$5o`
zF>>4Eet6$PZ_GsNe!KDeoBnXWi&lHB&#cGM5A5!6yY;zU^|9MMje~C;bJC5MzE*Y@
zTAw-<dwM+D-CkUe`)jyr9^Uqa@1JwVj$_LztgEM0?R8iBes}G%{?C5&P}x;%R=$)!
z)N1_08?C+7kt-Zlwu-NB{@jXJ-}ZP}h5Xt6g2t`t-f6z;F4}Yar|&K6;(p5={hh5g
zzqhE1<Y*P!eeN0ku<i=i-M6IjwaQn!%RRl`cUD^IxXX&J;=S)~an)m^$CllNDIWVv
ze6#mteI@^Pe_LVvU30xxRAJS6+iLpFf4_IR-4`0rIP8eee0SNGpDeDuX@M2be{|%`
z^O=pF-(~c~6@OdIx8N@xS?J-L%ij(-^rTO$f7#4)7yHf^Pn~U{eHxFKu-)@)R;e$Z
z&vbV<^18t2k)yt}?@L7$R{3|g?YGl|7hd*gSr^yd_m&q1FE_qSVU=B^hgEpo@t-^X
z!LQs`q%iW)Tj8ZX{GI(qJ^f^v!VfNe^7=b0d|T1o*y{OtSIf2P&sLq=Nw4cSrS6NI
zD_7xz-{0nvH6~9eQ@F*{Kd$(~+BeL!w>y02g8Ap!>%lUGUp@Tqf4%0+=gMZ$xpceH
zIam_IyMy?yDc=9Z@D@`>zj(y32g+8l>b~stv+h6X!5#lzMs!{u(S1V|R_&9<oVx1k
z-~QGOMR#N5v+h2or}CxjTk5{>d#{`{ZsUudFZ&9*$7vQ9`N_y}f4FB-k;1zET-AJj
zdEhaNzdZP@BI0cm4_^O(+g>bS>OD{0_{Ne8p0(D`o-M1eYCp-}Oktd-F?`m!*W5g9
zSXoA7XH%G(&u?vg{ZV(lcWqgPk3YG>v^B1OtgONv{&@NUtIl(KnL_?|#N`kA=72Mv
zzol#yqaAB^xK;a{n*XWzDu1j>|LUgK-51w=uhUoaIsfXedkgi(_2j0TE;(oDXRj+-
z#Y4Zg<zhEq{X$uV`=7YhtMeQ&xk%yO>pi#RHj7O!VBK@O9iI8~o{t>+OaY_)Fy8z)
z_tIxgn*EAz?RbB|58~&#uah5N?!RGN&#UG`tLateBiDTCsy%leHolBFYT(b0xNz+F
zGKEo|@11tm5g%IM*`f+#IbrVscHD5aG0QG7ra12EPv7*yub-QVaXi_TxDLekdEV{D
z{iOZY*B`xh$O_B8THLnJsi$xD^k0&g+8?ZV!56lfwA8C*cjKYcx4Y!rrJpNOh)0<6
zM`C=uH%a+x;#>V5Lb;b7TKN8}22Ls?KEBT2RbD;%p`xoeXr1eRH_y1SMa19TviaHH
zSY~ojg_GyM|J?n)F}+CPwFj<z<uA{hT%_=kpG_-2dFfn@D*XEIZ(3rJ`yMYMruI2|
z9zD;>w@#T*RN)od-}1;ayFXAs`IGy)8Sfur`ry|VIQsjm-BF~lSwA85z9tp#-77!S
z{+&Jg{1Zpa_qWw2l&fDh^amFo@l5I56E1b;fe-%dt(S_&x7%p@ckdrvym9R=tFH04
zxBd+`yne}rwmtEc;>k^}oqxj%2Q{j2+V)@j;((3jY?x==h23pWW|jK?alum?jrits
zWffXap!bK(VypV_^hegcY40CREK^w5-%af!`6X3oezl6R|K5RG-%~hZ+_aO<fAgs#
z;)+{;bL{<xzFgdS;=$i}Z29p?TzB$$zb;>Ur-*p)DPOy4tNErDRakY-!@o;ms$bq?
zet)S$HyQcl6GatPt$%U<np!WG+WW9`esbv}MGC((^{LtZvBl#>3SU2S;MV({`p=>Y
z<9Ih4u6nN!)2sFgsdV>U5u24ab&q46N@3k|-qe1+>fS`Ubuxu1e#G<BQW*V}RJdv%
z&}_PSH{#{hx88J>KNcyB@6Cl%{(CAuz6;}TbQRb5_A1}q;G9>>6b?OQ*hl8OazYs~
zb)S}scOH<!_lo<berka?%804@&c5rLcii{TxUvdU{SyEBhhKi}yn|jYBbtZt98lQ1
z?i5y?^YABAXuhP*f2F68ADY6C54!Qy*B2U7rZDbLn+>1(@}Pabxxm9^3h@|`zmr1e
z8?efMNckVl#z!9A_eHcXO<~>p!B+Ft^~>Tpws7ivJjSQ&o%lg>{^fwFXFps-jQjgk
zSpQSs$5h?ZI1j7Bc)lj=JSK&yby2+Y$<(@)@{`1O^?PdTCyahttKq8q_qY#dN2@UE
zNetT;*Twi<yYq*3VbqK0Kj`nB*Y11z%X>~JBUbS!J!J>QZ#MUoq0@JNvzTZ3kAC^8
zho-iS@m{DZ9Qz-`(O#s&ah#hC+s`$Nb@$Qca|+GZu3}sV+ZCpV-G%pF_^mHq|E1AI
zr=tCEr(xau^LFJM{JH6`o%q+siWFAu%cZBz8{#=jDqM9>o${ya>VMn^RNYrayPgWC
z_Te$U>U~DaAL?p+)w|1>-mdf7y5rK<@=e`kxdT?eb#f8WdYr;<?|bb5M;|n$Na4ya
zePgi=9($vx!o8mO!8so|=B=Wu80T5Lzh_O|Z!NgbMQ;^<8qjDJ2k$jy#a}Hvu1KMC
z+7!Ni+tj0ue0oe#h3;=#)xVUzjPcF^qaRfjKJxWDwi+^MQc)LM^?$4ULFw_Fs&3eO
zhgR|MCnw$V`BfKcq;S{=FF50yFAr+83*-HK)qBg-{cMbn>yPkn78riQVrRTjv<u_;
zuk+gIr#lyjxaA*4?{)u^vo~PX@8w8e?b(w?&pULEMipMQ+lDu-_S4xKDcp6rySG1U
z|G69R`rH3D?V1bbYou`D<qm)8?g!^@r0}rI)<1L24PQ^5<Hr4XtM;I7Ip+|airy*p
z6n54BNtLfUXKFP)>cI)8|7p)<e=wi{zr5I{PaJpnyG0fL<(t#re)We_lJEMx`-&Lt
zP7J^Oqen;Hw#(QeBL6#{JF|zWeem3qFQ5OA5pR?*zLWA^IEA~ceENxZOrMf`*Bsvw
zrNTdW>ublpa`|gT#FX8L@y0E6u9u2u$9o&e-z-M^UpHJeKT`J_U5$@C_m~|D-z(ZD
zrPl2q9=S{T)SGiOQh3^L2EB9P@pCnvI&Fnx4q5*LNsRL}hCgxo9&c~?&>W2_O!;e7
z`r^F43Y}YaH-3Gs3$Gdcr`OB6Shep?rKje5SL0Lnz^V9nKNZ7u^}pS}Z&J6us9!5D
zw)Gb`Tlvkhw{V@mt+oB}m%Ukb6}``{i|&n5*y?#+l^t$(`hHs+a{7Obc(#Zb-@V0i
zrK<b7kB?k<*eCuuq1gG}RTn*Cp9zKcmChI43-ePVu6SSL*xA4PR2h-~=bafo9$#<K
z?{4+gdrIhjv#S{2+o!_(D(`&a`=e+_-S<U%<vz*2&pVWe&EBKPzv4%Cyz1t49xbAu
zDeQND$p@Vqz^eOv>8bmh{g(gwIe%FEjuNKsIpco4>YO)~o|@P396IWuaD3-yJly{n
zuZZ3q#P7l)FZhSiuZr&l;`(iTBSv|wOY{uJ_tY2Mzr!oX7SEJ0u2WI|_?>_19xmQH
zr}BwcZ=65Eaa>a2R)1fqtL3&^^sxID-Em-}E{;2Uw@04aX^}=17Qeafn1^4QuhA@4
z`IV{k#%%W;fBkb4iz@WpKnkDv(gg!Aym3lVg#+ea@QnA?pS_X7n~&OWla>B7tw^Ew
ztL#+@xBl}#cN#qD%>u4-_J*q*wBzLDcO1U&P=5;JdA)GeIf3;uVmx1tc1!q*<8GLL
z!gVhcDZKyqe@>nBkwFdpfYGmu;gin4XxL9rm{dfxe@~t7i;wRvVmR*eh3l^0zL!fO
zIZ_z+Va<j|oIc-5M{Rd^k;2qFZ}$mxG3tSE^xsn9s`v2m`=_dKv-k0-dmiz1^)c#a
z-S9aB4m|z*!v{91(C;Euq3`Oe?E6j2EqVVQi_P9h;Y*i%@T~K$o=~KaU4p(-N+G|}
z{x+`H!gb$^?)AgXr$77bLX8xr?hnPM&KqMqKd0{gsp=h)?|Q4S>iY@l(GRK$d%v7Q
z{#Oc9;~V2s@0(+M^bhNXQ|~_yS?tU|FLC?$a>;GC*z2YxCeFn8zT(q|F1XL#>rI*Y
zyP@<l=8NB{@}E-}{b1oZF7yg+aNU^$*E?fOnZlGG8ROaQc>iPH)GGRJq0^9E>1!DE
zAr-ctYd5kV-Hj<fw{Cpu-9fx>?yj);)NXX%)ZLi!Kid@_&k@=kj{Ebrhr{2$i02$J
z9KSEG3jg=#PTAqoQ|51^aHqW&S?HVtKhUVcR{zdbYMo5Qr_QBP@y^vc6+3M`i}Lbk
zI~C)5fZhtb2WS@K`<||bo87;5HGk?|b$7?d{qZ0F@$=Q5U1wSWtM<8d(+Avg#WmMH
zIlc0H@4?2mez43%<BKYczvq(**F7hS`BHvSmH+3yrwYAGg{gD7RD50izIedhL+1V7
zJw;t~9$6Rt{uDh*y+3F+-Z@35V|>TbDPiwP>!R^bq5HDBnA(Sl@2dY9_vKaLRKI#E
z-aDi!jQ8_(!&U3O``8pZUxD%bBsKq=-KTWb9;ED``^1RxzP@fa<=0ym;_uJ#A5!-O
zsrUtUJ$#2lwtccpp?8|*O<ipE`}(@)vhrCs>f#DRpLq0lCyXxZqI=@H`1aP%?sV-^
z4;Qd%oa1?4RoFdTU5xvSR5*2iR23i3f2+cE&s}_<vC`wae&n{l4s1k>_OfoheKlVd
zZ(U7~eD#&^&hIS$g;Q^uSZ3@V9=dCszjDod8!S0|u12faU2;8i#%-_ew84}jh4DTx
z72f8XhrRZp*)J|zMe9?<)Os$y%8!Zr#?<~Ybq*c<k{GV*r+epBg?0V(nD1+kZ1kCF
ztG-mG(EVPsnEE{)@m24Xy$7wrlw4K!byfSi)bC)4PrVlwUnMs`$NR(H!qmJ>z2g+`
z{3YdgnrH38ude>ow@w@QOaW8p>M{Pp2Ogh#*?KP&Dg4%lwwS)jE8~l&Cx7br9oL*N
z6XW|R=UdO7x5&Y-Zt-4G7w>v*^rIiS{LLbwcWO}mb$@TCujaew$jjIJ!|4MWUBzbC
zt*ZG}H@(?*Q~CYwaHnF`dz`+W-mY`_s&lKpw$GuD@%PndwJ;vqg>&t8?nB3KHGiX1
z(f0`b1*`nWzLuVPH`r`^yYyK-@q4ho9`@Z>Pos0$PDA&O@tuAu>>fY$cOO#msqd|0
zyuY_n7wt3R{yv7iV~g=s;paD6`uIUNjVkM6T|YABt2%F%9{r7~bM2UJysOZ=WBZO~
zvF<uTpPEJcqKK*Ak7{;56wg6Y;YmA8y?pdW6N|bS&(Blg_+8bqC%$mtr=EMesKUDU
zLGrcw-COGS(!|H_hU@;0)Qz7xbNNp+UM#AxZhi9A-H*%XyPCM~Plc_Q%_6(hRcyB1
z-pcQMsJk)R$yqJjtlwCbKYm}_Y2muRJ5rS|&YQa5Rn^UxT4%Z%&))Ym`hL5=Vcq+g
zp3c|P`=0J@&wWs*W9s)RdOJRThmi`W-g{NWSKYJqbb6|P(XMw_IDYroYB+UYmHM3m
z@m1e>SEa{y*;PEH{$AZfH{Z0v{ztx5#_wSd+U423mbz<FnZmd}H+zm(_1-e3$G>}9
z752M2RrklztNgB*p1RMjicfu4D?W}x-Ecg|s2i?&Us3m6X!K`VU2o&Kr^3yycizvY
z@WJnIbIBT$CzR0pfx1YK>SD9|psMo$-&w;dJxrz7Juhl@-x1%{#4x#A#b)J+^I16B
zQNRD{caT%)T(B;BuN;4GD;4(l5USpj$Mn>Ds2Jbs`d4*6Bt8DVPF2|VJFxC|Gj+fB
zc0TFvj@L!!QFSr(`*o@Ks{NVYyMuN0E!AF(kN$nteRN&>n966Ju0p?$oWj&Q{;K#Y
zd!6!|#K-rCRblfuVzfKWhW$QcU&Gjsy8cYve38elhDRQC*rm^IKe_BJJov^f$Nty0
zZ<jFs9;fir>wRaXm5#ft=qkqXNrnCHSi7+594)mT#lMrvZlvBvnh#ZWER}Aa*Tu_^
zEvDaj>oX;+>p!RR#eG%X-$_XM%`x6P*j6$6{V^Qx;rTsP7}uq`;RQyI9QCDrUn<&#
zsdwX5@qO+0BBEc|)8W+o@;lb4??n0K@Y?g18*|C{H%l1x!h7x1zBYbm>G%3mzsnW(
zDXoU9&gWu!^p_U7<mBZ?Ej_u6@3`3q=~e!WbiZTRD#m%;SHkRA>U~uF-Q869_H!Oy
z@~PKeEAeZ88P{8X-_d@8ziEGOT<y1=^9cK1M7(uCIVgWb=ho!3f7gBzr~Yxp7uLR^
zOpROYXRG0=cRevZB~L0oHE&b#_Dxk-_nt83b1s|mqr}(sZ>#dvwd1Mt_84Eae`(k6
zYQ=hzdOBtFi$@H5piKFLG2Z!I3Zou-mr{lCyYjl>Du3DUnMSPo-B{^W`~TE=Qr&p-
zlU~=w)bFas_`2`^<9Suxu<!p<_cSrS)$b)z_NMBbt8TjA>#d7*`Kg<a-HMp{dz`6v
ztKz*Et-`8(qIBb27gOK;r{df7eVpGZ=;}E|tL3A<SB2}&Px<&imH6`NTW`9`ACrGq
zFr^o%zaQIO@%9H*$WB(_SATcg)LX|sP{Iqp`qz<9E&WJY7gM}c#i#t?R6M`4E;?sP
z;iMBLy!hs^lgfzE4~zCi*xzZ2=NDCB{&8Jwc3#Ev^19*ZccsGdysIi)*B_LxE-&&`
z-A}~y&p&a*e1BVgLfI_Z&s5=Sw;uAt?O%DmY!;inXS6@7i}bcxw0~_k?(*U5-ahyb
zbI$xb6n=-$y_feU{{Axm+<BgHpr`!Kg?7DYli~aQxj1N}mo}RF(1|nAecjA`+m;Ux
zJ#X5rGtV!~6ZiOaf4{ow@0dt;9vt^)sj&NX{e{hrUr*()%U4(P&3)&e9@%~KNyQp-
zd~(|2bG$MW?LX?G_xusvr^oLsgsc9}LrkyhPs{h})Ema!^vAgyuxh{EY<gXLQ8!;t
z>6dm|ea92aRh{=sue%?SFZvI4!*zc@%6W3t`9w^wsvo~yMdyW`hEYE{Eu8XmyBZ(w
zEv&b7agC#fZvWykw-*s-uEQ1QzRsf~I{&W1)ckWUox-Z`kgC4RvfiiAIelG>-_No)
zDa5b8yZLYU@hzvlJM^yNumo0}J2m_FvHablh-WQ3?Zw?Ld%lRc!A_T+vB<kG6e;v>
zxn1b*$hB(!;@`1};rKpTINqzJ!d3ovyhn`Ts{KyYd``W)7Vkc^3RC?{#i!=4cz$~d
z?Gsa2wSSdvzmdXL^`_Z!RrW5Y|NXDmocUbYsXWlPs&_RheHL#Yn?mO;RcPOtS`VWA
z<u^yaA%;`$-o;nFYv2bv=S^X}KXg71lYa63<A$EN`RKCh{0%=<$X=yzz6F2r$U+a_
zTt;lxKD0Z3)i`uYdbHcE@>F$x(d;>%{8ir%wL87l{x@5$+4G4$lYgG`f4BU(kCdtT
z+V1#fdF_<^V>TMT`C}7)UDQRtt63LQ`=4gx<2okX?C&sEy<dy(15@F;?;+%i_x+Er
zx9E4b`szJp#H#x#>2d!M!>N6<c<-Pi4n1YqN9Mb7LYcy4cG!FU+0UI^w)*{V%HK)F
zSB+OpPyL;q82?^z-_%bn@J0!%_FpmmGnX8B^=v<wR#ahJ*J3#S4pOt>=pR;{dq|J=
z$$47U{xFpu`&AW=`^mT;_wKZ-82i~%;i#A0-R?>M`1KiQKm117EJnFfcG>-%_b63Z
z_x(%V_X6_8@42ELrNXKAeW`f+oGN6G>*5Yy8vgdrZ=PJjs(s0gKR@xruRb=Wtc$*X
ztcz9eRHUcgJy*rYdycx{l%E-SuCg=IozKN{>Z-7NlB)A&>CK+!JD;mU>lBRdVp8E|
z_a#;58ZkY-FHeP?2Q-V(uc-<*`yJ1s{~WZ-o8NxBsKR{?DW?8n_t_dLjCQoIhy5O&
z-?{dF1IGQ6a~j{>2uHuxd!Bf{XFZ7L5pn+Ac+-)ic3%0>68gIU(EG!2zZtsMfve8e
z*yg`}{pWih9niR8<!7glJZECz_b=@ap!0M7?~4mBy-#D?F@^U8%6Sj!yBYmq=lDt1
z4e#Nc6SRBW>)OlS&S!mYH@2J4RDDN1_~;_reCU*-3ZuWB3U593?<2Qqyft%w!EV!^
z6sGDEU*(T9n;!3FqCb-2OZ>&xeB+WabIsPMLiyW`RqMg5n%-`nyQ@BSwONev?0NaX
z?H>Bvq_VF6l{()LZ@$8~??{EMpLMb7yWm#S<NDj~aI1e`eVI=jIAGFO-YDR~WB;=H
zX}_FaRH5J1slrq}mo0tu5@%ofPSI7|e21Ox`20K53g}%z3ZtH;!twXMn+>biekWq;
z+@{^}@w;tuMXdU}Sp3Uo(K!XIvMc<yx@aG6J&suSchu_UOYNt`^M~sqKRAWHw`vws
z<5{)etILCYRrh-_z1evb^V=^}Vb$Mxm7e<U<_i~(y7;883~00)-S?+3b>1gFzJIS;
zk9-%<E{xyb#_-_1rmXm@g~t_j(S9z4QGdD`w(m`0>`y8j?dPl#ZuOq2s$7*G|66(=
zA-t>l*i+?`?{?04;ixl5JY4qm^se4J#qxF4=SwR+yy4Slj4PVOp8oxyy8UT3pLbuK
zirqa9{Js>%bEW8?royRnyVO05__)uAVecKQFwPfY`;%C249DMFa~@Fl{=<20MDM_>
zFz&DUP4*!$rPuD6QsWcjoja#6{;sL>2<QDU-c!eZ(x=pUM{GaEw|c#wt-^TEB%Jc2
zydN;`RY>mCeR0&U7>?(p!uq-MJOBCb(N~>MRm~IW&DzyY%^%<0)D83R<9Rs#Qs~|%
zV%2*G>8n4o`6bu=<GwP9mks^F#Ya3-(wB(rUc{8VF`ivcVZ476PQA|-AAg5uz%5r?
zbM2GU3x5agsLj^D{ghLu7SQ>i?~qb|x6bdUI9Gv@7h(78(0=LEmj~_h%>^DVBYGFl
z{=#T~yqk&lA*uI&t2F-l-#;GyP63@ex<~eR<=iVm_SSds5!nw&58PKm@3>$*SBURF
z@D<VTkNB=X;@5wF(-Mo^_jm!_r$Xx-G;iWLighpED_dWbi+uY|E}p-5m!dzZb3*4X
zFn;eB@2{<2)&=t@>V>d+p>c@*RO)=u{dd)VHTos=0`doAyTayGv#5M`VcqZ8tMa}6
z`GrSs^VTgz6-GUZ;U7QvnbR-0>hdy$>`4l}A4s9!%YfazF30aVb9lp54%%^Y(bGJn
z^rchcty^`mO0U`VPQ}!HN2kQcb-2^Q{DpR7mEQ70J00Dpbt=a5mQD-z)H!@t+lhW>
z4D*+p#nkzXbD37L>K%0|eY5pmKWf42r<Bf_sxW>x6~p#5seY#7<2h#Ca9w+p+K0q=
zdNbSWCv5)7c_);0vDtMbzUQwSw*RV&_S3McovL-9*>wBkzKYHEr`7y-f9fmS?7GRj
zMaK8%p{`$?@;BqYwkq82d#m{S<jsbiU)BA)Y4SOzN@3N0OFebbc~8W6j#U+Q-qcn6
zx8IRE;xpe}_T?vws(x>H{gMl9d*UkvjKB9(8J_&$Yd5TUXXzZg3afs%!*|m?h0YV}
z;%&DdF<{q&9xPIL@F`!rYODFC7VSdktts4lz2}zPX0ho-3a{AymPel1{ec1=`T8AO
z4H-14i0EBz#Ad&<XK&)~%2b8@{l_Z&%>u(uSnP~9iYlyHZ|myUQoo#b>-PqYE@73Q
z<M&~!FusG?@170z8~?>WmpzUDc3%I{Pd)hXlGl~X2Xudu!m9Jvy6Nt#Bc}9Ue9Hcb
z|F?Htmme{7_WRB+V|=_Ps2h&*3ZFG;_A9=%<NZYyx^Ih^^0(seD5t`4KKuJa?MCB}
z!n!;*2G4bW@x=WlJZ-%h42;-+#IWHX8?wZZC5H?=cEmO#hRpV_e}^wPe39Xc4;gUu
zh~Z2AE4<2(fkOsvx&2DVj2M2-@q@>%Ffhem%kfigEDt#9tg%IZ#g>VgQScvYUboiB
zk3Uhi8@no3UoYQPed=rd+w7jTW?Osm)n&7o>c_HQICsqfSKe2&im7_r9siFHoOt9n
zwtcPWE^N2FsrtUL#-*E|Ht)nD<A3HM#apxSKU)9!`M$rws50Y!=Al_WQ~A%oZGq9Z
z&;Cr&Q<#z?$N&Ef{P>o2Mlb!92}P%3ciH!<dh6=X(<kqC@ijY6E>rmKb-x_F=x1Ln
zVcmLL&A0bazh7#TiMN)WhGU<8b;Vov{%6_K*ecJ;Rq0VqZ_Q6z_l%V;Uhw9LMOU#?
z?Ov4ovXige=%D#tE;<#z^UHx3-?;CyMaKWj13S`PxcTS{S6uG?=gUsVt4=!U;<+y$
zT}JFK-qicwPam?`m+pAI=qjqOxAE_v9kb>ITfb2B6!vz1Q|<TE{v@`Ol1Kb%$9`zz
zecyh!Y!$!s$*tEqXOp+eW^u$%K5+Aw@0w7yihI3u)QWc>F|q6_Zt<1(j@oRaNoA+v
z7cL)i;JuH%T6QYd^}p$B#Jc6>J7JM)c3f?z=gKOK?etXm!nuEc<RuqPFS?37C2y*o
zCHH*e`&T^sT#>Ok9`M=Q{u4P`#ZEg9An%VqIrrI*pZY-ADz?g3yUVppf7&fqtZ(m4
zwi$KfkjILQeT#=UPO0#No7{ZX`fof_wi~r~_t~FVVeJX;741UxX3PxKVW6*^f7OvS
z=V_gRc+Qgwx86d%l0CH!M*pBye@D5E|9SQ`htK<T*{R6?>r_nXYp29_>iGq^`^!6z
z?|tX6-QQhyY}wsd&l(OGF#5Bbe)q3$4{WrGpS|wq`!2HffJSfQhTpm4-BEW=E82za
z@`u&4@W8X@J>j59MSsJccK+?Vd;I8?qPwxKK6f==tM4Q`wOn_<Q;g%$Q}+NHy>$20
zKU;H5QH9sP{rtRZFZE_w7nfae{s;ei%z#F3W8^i5JA;AlUia(P-!8VQ=2u@&@9Osh
zy<NWD`&oMN#N6MVIC|R4Wv3vyXH`7)iEm!>`1i+^8N0?qUHxg-ysz6{cjfCTe=OBb
z?ALXRpSt72i%uw;#a7?B^tEy+o?4BM{MHRGa@S{f*!YNdi*{q(_S=>3f)%%W^vmbJ
zU9=0^C0BRXbI3(Up1j{T-Yj|>yStyM`ZxUX55F?}wkbuY;|D)Ea@4Sw#}%E1@ttJ5
z!>Ms^HNNh8<2<@9#(SVt*!M)RyYCeqyL74P+if?#>@Qe%{Q7FXX5aVrmHbct;Kr{k
zu*=(J(^&DHUF!S0cE{H}_v^3uTD6yby<F9OL|>Ob%GXohceJ~m{_=g;x!eBfs9&Es
zzGyebac+0GQ_k;Fa`v_Td0lyyTYtZ`CZ6<CnXw)Zt?GSOzvJsG<=b_h*zS5-9hY{O
zoA8MnMqTr%_lmlB$92yg@bI{)MY9;c)2<3v`57_2+3$5aZQoDccE4Azs;}Mp7VGnU
zTgKLY*CX=T-Qh1haL`H%|NgGBQ?b+a%)8~@!tU0uP9L96eMcVUUg(q+mYR6=y=A+x
zyYwge1FeR;OTMnw`|8l|EwTOjQ;XihR^`9u;?=er{?ea{-o{qt@APy0o;sIrmET#V
zT&wn|ua)c6_t;f(AGP>n6Mk8~SoAbT{qOB}omD*5O^@Hb_+F!@vF^D>cjv3iTPok`
zJFNTAqVN2sOkvgUo9}e&0~h~(t4GT!bdTQb{^DDcR(Sr#htDn|#`A@$bKBH;b-V8;
z+g0D7`8J*S^UJ1|hkfFXi3_Yev29qzQ>*E%zITh|etYLur;ONZLK(5U{H57ep0N4G
z7r#-~#j5$x)9KB=D`@ri!P?!gIDdOP9M{WE4X5mUj_=yvRp=}HwW^o<U$x4ChnzUA
zY!*-3dHdZj`Nx}OyRcb1-t;@AiuX;ees7WTADWG?`wp?)`MPVr-l{%Um8<&>y6Sfe
zdn>)!-^pq99nb9#o_x=tW8Wxx3*C=)`tPid-EGO;2d(o)(O+?q(XU_q+#zoj&%AWN
zgndtYZzjgy)uc!9_dBZev+5l~-S_0?&B)h}x%#D_Jy`koh3c-$_MZ`({ku&uf8F-`
zTE5;s4@$M4I){kyCr-Td^WPeGd(o+Q)p^f+>e#h!E;<!Em3MNVvE(YpANBktMW<qK
z-+RUOtuJ-4>YZ`bdLq59pBVG~;n{!w<K}aJQe^zkJhc0MzNhQ!?(ea>yZ*ZGmiubH
zcJE(%y1rJ=3A$RYYG2&0^sfG{V^7~7`_8h{apgZBFnZL4No98<J+YtaDpsx6ot9p;
zpY3buPrSI{W*4mbWKkE5e+ui)qq=rC?yFl3_w?^!rTD2EU)SI5Yx$h>^i@p#?q<8c
z=X1^!&u5%B?DW|KzP!do_sslz48pTIHtWx~>;9qYok`vFc#hX;;dbp?qn+#S-xuxf
z`n&4qRF#i>^_6h5_mF)pe>_j>YvJx1=dS+VQFqDJu6n!soTR(!zwznyuQ~3+@06X2
z&pbG8&{?O>Hmcp&Q|C_Y@*8qJv*zzKw)@>;tgl%++0*A{eYL%&d7H5h>*~I%*>+R;
zySi@ocKLW8)@k8(@s}E3@qO)FlU#Mr<@#$rJhThr`A?^Xd;9%-yW4N|JCsf@SEbMK
zdzq?mfBhZAcK569_dk6tU#s`Ht$qhoRj%3h{qpyfdE4FR2|ZPRSKlG@R{8$ouhg{@
z?fR~!-Szaf^Ve48se4aSweIHh%s?kIaQI0NZFKjy-zhp3Q||$qjqj;3jOT=%8XmgL
zZyud8_}#LrSl15ASNHu;tNy}!M{ct5tzUe-tc%UwAGCVET34Q`e9i91Ta~w}Tz9{7
ztg4?rbrtLCV^8PnDSOse+DW~mBcI>1Z5DmUT7|wJY8Ly-cV6**6y93JzV=SGUGmg@
zXWLiu^%ecfa&-20ovHEbtG~<K>iK88%cbmRyT323^0#7oU;EwTR?l(zO1Vzm$9d1*
zUHJHiMof6)BM%g?%HQrTIeMzzX8phJ%Ac}3b>r)vQ}=Yfy7IT0uid}L+NvDwD%bAs
zVyfzEm)w25p6wQScfzX+Oes1I`|5XHb>&L^K1`>?*WHIU+n;9NKlT;Bi5$MStBX~B
zL`?S`S6%;J{&r!z_et&cH=F(a<(R*#_PVFuKQ!Bpel`2|$723=@AsOmukQELeLY{Z
z<JVX6w|XC?96ii9YctTU^(fb~)@*A}zPfCchwj?9&noq{YZkYAzg|^eyT1FV>VLE6
zv(4V?w<~{DKf0Q3f6!mB+26;g${+pkzT)q6RgQM;Z@Q{~eXV?7c@LkGqw3vFe@U-;
zXVvcXR`syEzTd1XPrKI3-ma%ze_y^XZ>fCo9yb-XZ>)>$`rU@AzjM{y=jBzu(-YIH
z{JpNGw|o9o{T=&e(^GoZUGdGnzf0w>E8hz1toQ%1ci-`L6;=QEb5rRhROuZ;uK|g$
z4<O~?Qj}r=kq$utK?34~bO>D#0w}!*s35(Flzr(XQUjq#4G9E7O9-JSzb`NE`}kuX
zx2NsXZss38%<P#pYd*8qnl-ccxjAm>B|c4B1*<)8mg-k7AEnyW+du0iKg&+qrM_O@
z3)cF3^#j)5rt8km7mI)AuwBZP-jBCx`+C_$z2){jci(kK?e|>LXe<@q@_t_LzgyTS
z`FZ8Hjh0jH`}9)n^0tdCd8NLeY1QZLvX5rzS!(_BdY-lJVSTyZA+&0{QumVW^1R;r
zez@*ox!&=z((i~`wSAbcggoX|zH|M(_L^_=K3%1-)ZaS^pI7>wY^i&cyyja<PF{ac
zt<iGc?f!Qg8-=BKY}WJeJGe^A;lBgfZsl^nD{ZxQtDn2npJ%^!Qqr!N^?TfQssG1C
zYc8?s#HmTU<9l20Gw)7oeU?;;VZ566)?0tM=aC&<A(nc-Q0aRGfA=d3EB$*5rP_z{
zY*cyX=O5|2`9>WDSnqMGy|=5kUU|z8zlUhIa^ByGM|r<qYR+p%rloiY|6S{LD_8ox
zSgW;f_wTz(-FM_2=jtu5(SNU|)$FlSeaiiw(z;ce7iG22%9FzSL6<#p#<!+?nub^^
zKk~fN`-`xBt=~h1^HIiU-uLF@R*J3q?<!QQ*Ap|Ja_kR({C<*!mGV<*`yrz*y5;DZ
zUQMgTto!r4^TLw{-m=T`yN^lB#ahSD`g<(p)>pa?&1%ojE5$5*%l&;@slOwh_1#@o
zeOCO=s;@L&hWE{-%9Zk2sXkf!m#Q!K_rI;$uGHU&3!j(b!}GqgY&7YtzkZfhi{#{u
zrOs8Qp69Jkl-HUJwA%f8)aQ)Hju<~Vi9GZS9$M4uv$d+%``|{s?+wSvo@ryBl$+l+
zyIpuc&~D|tf5%p?m27bPS-%{(YOg}2*y#7J<<6(lc9qV%)!Gjnxxvmuj`=F77gow=
zt-o(-cK=pA2Znl8x-PZ0Z};~pJzf8LzhA7?e%hsfwRU2DmA>by^xiCNU+UjwY}My^
z=R1v-m-oMe)+~9I-ov(Pd+!z74NKjtx9an(_Y$?%i=*|z@ZTk=ww#xKjsAT^UV3NA
zsrC24vHztj_qy?^q*fj({ksjV*1po;sb<-u`bvMN5ZjM_;@(S-O!DHP*?(`RRmVT@
z@zIYhb;2iU7KYzfwQITE^0H9hX8k+Wd5zmH_u=AoqcE@hy43Gh>ixV`&-vx*_0{d?
zy*bwg<I?FB^S1X!|DBXd$IE-ZJyl-#UbtT6a`~$?FOTi7-t_F5CM9|CQ0uv{-2T3^
ztM+$(rTR7dJzpIE#fj&o<B}{M%KiOGY<KU4YfXMF#Y4Gqv{buB{Z7Bq-zC+neY4Le
zrM}Of-kxXKXSw-#rQd^P{rzwC<({7^egCRG|Al9-(ape!VIy`Mx=iPMo%45goG@aW
z5uLri{_oJ)hYlP%PiL>=M+}|s>+*7)9i1Ip?zGhK5koINX~XeLbo9FE@(CL}bodqN
zxV3hF@u%yJO!vQM=}A|d_I&!_RV#h>nvI@HH++2YM=o3ar6PX((WyhWy8HEF*>nAs
zA^9`!vER23JEAMu@!Xxx-hbc|iSmOZ-kG=0J~t*uKmV7-*Iap2GHB9ePpow4i-~^g
zE%fsh_x$L-)bqEGf9&qLUVklF_79tOoc^OX6aDWx^O3XX7(O=H_SDYh&%9|&^2HS!
zJ%7qqZztQ2yYkajM~yDxO*4;r?(jeTw)mW!uclsf<?;)^Qf&9ss|(+E@8usR<UVm@
za@~oSy_X#D_6~a-wev^GkT=g?d4oeHB`<#P{1a1G`#9-)>(sFio$`J{Kl<Ku_?~YK
zd-$OQo^bN6cOSUl$Yi}KW1kwh{*}e~(lec~)E?Je@q7Z$`gE}&*KGb$GU1iwPP?(!
zvq`akH!s%7yvYeM)W>u44tXHYbN#KGdh29eq4kIC=8;|}4eZ=~TsmOxKHER|;ydYO
zXWcz=u|XfE&z*C4$1VdWrmtPO&TJQ4@^N~|C4awc;m1EJmjAifv=#0<^8I4FS<d^;
z+82K@F@@~byyy=nJ$UW1bKf;JUG~xacG~vJX{quef4}j2^UeQ7dh<<dj{E(_pB0fF
z<QxCM1&;3e?qA<cLo_e*gw}_D4C?phm!r42J!RMI@w}S{E%@j0&lKzNXg>OBuRp{y
z9^lTO&vey4&UmA!FFheW**E;|lphQ^yzeK;w{O2|fdL19P_*m$o?h|RTXuXZdG&yO
z<~Vf8`-}ce-#G{E-DkJoKU9nl^Zs+;OD=xy)wJuw$NJ6w>}%=rpL8C0+EruHH3keg
z;*qu9Dk44bwg0ta?*H<}+mltVz52T4=N*-j!*1}e9-0?aj~|Ff>jK%g`S4ry?4G|s
zyonqA-+1ZMj?*R<<Ai$gLs`E)ueoQVnVy_bjBEU@sBaLb=}%9{PL$aX)GtKy)6P7_
zGdw7Vs6E8X?=F5|-&t1}pFs8ek{<Lfh1wUfo0np_JHCfy`-FKE%iZnGKjbYen;)LC
zP(Mh%@yUVsAx}GqZ|&ht$F2U?Z!Pjgai6kZ*q^i`ckg}+5B&27uO{r1T>Bhetj~$3
z&$P=^7rc?MPx~|bHJ*I{?z`((JMA|A#N^kX&2!+L?|o9-AL$|9KRw5H7d>I(+i7vV
zy7x<ZK>m;wcZ}Qjj;n?*d&lTv9wY9E17X||C$4$s;x!lV_ez?TP{#W9U_d?rH@NVN
zeP0<nshCfMd4+u*t~bv`dv#s-VY>NWZ~gTG2YpgRv$x+q|Ip<-Ew#&w#r=z28P`6>
zeiu3E)XCdC`g$?G^S6Cw|Hbx0x9Up82k}oodEkSW9e3FlbH828BiZ?dvmXEUrZc>q
zDi`yJ?);Kp*|+7z^3U5h-1&wb$2^|ex5IsxykjQ3x7Eut+?c{*`|k6C{nLI;AM+81
z#1;N*e}l&3kMtJ@p?y;vvagUU?rL9@(;e@{vDrU5x8M2;PD$RFcGgSlc72x2-)Gx_
z|NH09i*ZZ6cqt#kqw%kJ_gm+sn@uk6_wuLPW;x@gwfFq0C~wRD12?<<%rA=Lkgvb<
z1ij4D{P0MB^E7Y!2%Z~_;xIq4{_5GaIL<E3L)m%9eEEy{DC5PxM!xY*yXU=gx4!C>
zlsw2T_?<Ey?Tf~zU(t@b^A7%NzWjh+6`%9R?tb*dLpVPFq@R7$y3-5Vrzf3#;>fLk
zF|L?z;YmAkvXCD1GJbLX-TN~hL!R(QPx*oI)f->gy3+qA7xZ8I|NU|D;ykP~-a@;g
zr*-$-`kNR1$fJk((8KuJnJ1JtSa0iW-JQQe6yMZ~m&)X^GxIWUXg#2L*e~!+UuF9c
zdx7Ga`9kYKJ|5}CU+G~SNFU>hpXAFA<*VW>eXIw0p&t0cpLKCQv>&U7>g5AhU;gXo
ze{$R8)HziDyYBm7;xnVZO5GD|d)I-(f6(z|vAp$Kb04+N-#<%F>{|8Cv)}o5{gKxn
zeBF<K@@et@V$LPbTkoz7C#U09K4<Q~zx7Fa{R+>$^1oB3r1Fz17cNXW_MG>M`9^pj
z5$d_g(my`v%N0H@-hZ$o<9H5@N6u9XF7)*J2TUs73*h5B^WA%LzxyYo__sdxA@$bJ
z{%79_>pj;$Tu=7I-b*37gZlI95ZMbq5O3HM4D}K>oqOmdu7u;Wf9Sa~G+tPyw|esm
z<tRh?h<o~52lEcsP1(9AhxZfKFGT%}gTJthf2f_Zc|-m2jxQ*lm?t^<LF-3PXrAP0
zhi`I2#5X%PF4R7hi+}SrURIf&=7AS>t86@WslEE5UC4L4?;Xv<J|cdG817T<5yj6C
zokK!&?sjjDAL~dTdgIG_IFC4YI9G_z?q$U@@l6~Q*OkRNahIGdbbnx6=Q{h1bBlf7
zbLHYXbmx5|zIx=Png4feDlZ|&JqZ1rL+NFn%G%>wz57w?XME$?-|QFGoxbeSe6722
zruNxvvy-=Yv$${I@ufxnIpByNe^fjVTL;gLqij9cq5Mlb>n=a>oSy7P-W{TR!h0Eh
zDxcxUP=3OX)QfxKhP-g}-iJN%+B=hq_hRA|`TW&Am-_`c&tE>-=$|M3d;e#@3#`8V
z=AS+JX)3R{eaM*;ZawvjV!m<QI-_p+)#ab3@(}&(^LS?;?8tqF@u7PQ>l^yDvilD9
z3fTwyGcLK1e0JqM0Qn(056c(vAC~cM-LzAN*3<eaYj544IK&^sBjxk1JL=c}bNjmq
z`+)4uJ_^kbU*d-Nh`(RXztaWNhX4D$7W=`Y_+XvI1%7WFs69I&Uwq)t<hyq<F8%nO
zajc{GXgupou6)w`p!VV+J7rh&6Q|UR56;KVsq!Ds!#F4Y*@vvBIAR~o8&^1GufP27
z9K8y!{%Y{(g?{}>;=E-2`IY%XyyFR<><7>I<ac2ly!OjyuKZ&3r-||ZGVE8EynOnU
z;(NW&pWLI^r`a<<TlLSqXB+wEJBj;H`!*D3AvsXLQpis27wlcxej|>lhwPF)hln?F
z#m^--oa2XU?fg~xqm7?9b=YQ8i*eX~#V_Vq>h4{8@AE;*zMc2zE${ID*f`4c7RSYX
z=QL&Q?FYu8FAVilk7s@>UO;vgK4*7$WUu_nbI2Z{coNDNkDzle`z4<~p5p<J<XaE&
z%!?fR48%vX5pU+9eCZx{o^@%*q@sV${PT6EE&9wyiRb*^{k#0B7zexW=fpKAUlI4j
zOZ~-n_gLb)c%+?v;x9eKXL*|T;sCwj+qZtS!1F7AUd(IUv+xJceXnuYvi;}#)-k=h
z{%8EbV~5_^tLysZAK2nYKkw+8f4{kJ*>r*4U2p8O{gE@Q*3os-`yXxn)^B_NADs8k
z<8JudlwMur@elVb<X^kjIsfy>roFn9#e*yq5AE}FANKLgKiTcmM4S~*@nxNy<K*Ys
z;m>pHhhO`f^N>9A#Sh=vb@2ygDoi_fx1)b`OP|7d-`a4Y`Ihfnke7HLD}P!0r|U1b
z<SHK)?QFMgwtHi<g?kqkKIQuPfB)f^$u__J#a7R~@Ky5Yh|4?I`A6@<C3`-;^!+0S
z6y!xv9wZ)$v-GF0GJA*iIrG!se(nB=e)KUf^T)6L<4>44=0^j*Ojg?Z?7z;qW=CPw
zx$eLI_do1ius=Gd$uGn=dWmEFPW)iM;^NSm9=vIWMaLJvV-m0A8TdA?GQaSi2v6dd
zJi-3!z0}#i8#~WuV>$}*6#e88^kg6QWp?L&A(SUiF`jtA4px6|(oM@A)~E39>|4Km
z$W~Jm==%fwvLDaoAJ)NnP9DxK3;Tbz=FWFbE`DDp?&0Tuj<|V)^E*FF^gH;#zy0i@
zD?d)WC%bdRd$(_Y&9sC*?vcGu)BcOu`yKY<{W}WZUh>Gho*CD_z@GXp)^)<!U-m0_
zKS$2F+t2l_6Z`imJoVZt+m9REzi`o|7ca5af&Gj2bivC9EPCaZ{R+-2_8s<PA7d}_
zUG`@E?0@`-9jMn|y}ZLYhJQIXi@#xfX5aQx=R)_=;-5Uk`4h@Z<VQzcxBJ3tKl1N*
zyX66+-WlGfPnUQ;e&UVCFLP+0uG<dz{tx<%?%Vb5uje@RyJ^3!#Seeto@4*iziZLk
zFS&d2as9eZ_~L_Q#;(!7_*_{#<%K^!@Qyuq{kI%)jT7Q6H!iX5AvgBv3hR~S4dRA8
z%lCoMddlnUPxdQly_8Se?6>#cwA!@PesaWwH5d5P6&+oj*ZlFVPhS1^{$(H2UfF)8
zoYl{B?V<JXe!@PfUO)Bvm5bt^eAfMhJRD!_&Amq!mdeY-ZTkg3QWl@(okQOGYuD~G
z&QRc2gLnSk-#5OuUqQaZU*BK<`-@LG_{-vZ9QJM<t+#q*eq-N0Y`OXWJY=3J#qV3q
zJ3Oz*WAPJ~<@v@5(Rm*K=IdNzT=Q1u|9PYMD(;Ghi_bONqPMO+EfL3$IQpPpy}0J*
ziF)x?9G6cVGviZ-e|p%LiTy<Uw_n&F_^)}|f6b3yi?jAY=TY}R#+6stzvxR(`pEap
zlm6}-$+OSPQ^aZY@&S3Cd=T=hT5-+u7r*V+i~1J!x#+cbKR%&f;qpOWEw}C8W+?1A
z@3C(_SLjpdKlc5Tms)H<aX*x2$h*9^pg;cHuel${@?Y^2%Bz&ae)hGHPkE>NTKdRy
z<c;`qo{=x9*S-|GACWiFM?HJ7-uQs*M?NZ_V&C#AdBmrkqkr}2=DiA@%YS`G#Lk>E
z<To(PqwwW?VI3j+qpz~{)ef@T;yE(-eM7AO=I=dPl@I%gea3zq#u@uki0--6+h53W
zt~QRc{7Sw3Gt`5e5VP#h`O5i6oEK;8i_pG6fA(D~x*rMSjrcAvcV4Bp{6#(@uaWo5
zYdn{~xSuf}a_o1Z9?H&#zR$v!=TN+qcZsjXd365{$vGb1@?i1ZxkcIYFmEwW<2k?J
z!#L_89z2Kog-AZsKh#HAo*}<*u7ma&=XUF3{qQ0$fzE^MRy{j49{xgf{|V_CwlfdB
zJ3neKkL3UEljW`QO@1zKRgVX8L4Qa;&z%p9Yu(web%EAL8M04)1npz|S=?af@&@-2
z;d6c=KhRztXMa@oJp*~(Bl1IfhUmP`-^?>?=ealn<p<VRd|-F(x5(vx>`^^EmBlB1
zgKz#uKYWMioD1E%I)^$}yKjKLv$G#4+XvXEdh(pp)Z1^^DLdz1crah~hd=jscn(qg
z6Yu#yp75a_Z+P}xe;CHs;yB&k@vsB+^iq~@uvd01zhkd3v}5@q{|TRy=N=A*c4D3(
z(!Yq^`6GSIr_}Q-J<Z>9d9iy)=M<<NKQ}+~$D4NeQ1)DVsGa*%^C`|x-fUgFzYo%n
zK6o*Xdh+<8GJ9r+#?vlb5B2(Mk4NLFhxAlt&*ag=c-oO`Ja(_2_Uy?y_4yU={OOsO
z&RD#Eb${zV*8OdW+PQylZethb=UnBS6{7nJzt?zTp^j~SFjv2VbG5SZ-LI4D{s6KI
z_a)(Y&gI7SKF7U;=j1sr?(*BEU%UOYUIp_tet2JDeDa)Q@#Fb}vuv{Q0>}3%m=8Uj
zhspQ;Ks)pD+&KE%x6H$RgZp^z6UfuveAVmcUY{QHhImxg-}^FrxgTL4{EuC-7yRJM
zJn_g*$Ps_&0oe`x#UJDFd-qi286Tf`Ay<3rB;L^zy1yb9Kjc^kJb6y<{g+w)qq&|K
zP=M^rxLJDOC)D4%g#GYp-t^U;eHfoyeqlb4KbfEPF+bzrU%YoOML+8g**$v~f7mtN
z@uQ#lLG!adFyzhrv}Zr^Ui*>!%Dun$AKq(wUjv=b<y9eiKO`S>@9n+1^Nf7EnAddw
zU6F9VuwOy>l6}Ovz&Y3ZPWhUA*0{zC(K$sw<AwW?e2O2+=h>Ha_Pb1WqMh@AJd!-)
zD2Mx=bGiFE_l^9U{kZ>QFYKRPS%3Pm5Bj@5b<SZY`a$ziCSTsiPsB;>$(4sfc?16m
zzn}13g87)I_~pH{arh;>=b!YnF6Qk$1wQD{zdYAofA^~7tH%RB##4ySA0awVI}h_G
z=(%$XJ9mCj&yJmAJcsfr{>1*Zr<Z#N<7fx9R}ST}XZ7x3^m8wv-Z=E6pZ=l08ke3>
zd(Z9H<a@5HU5KGx+T)p?%AV6xy>)Z%fnV?2-8-r`5A~2;gmSbu4n53A8HV-d>3nP)
z<C-V==0^|D>7lGY9@J|WzUPVM)?z^16xZwv;;T3=-a~O%Tol*6xAh)TTn{m{6ZTRm
zmXcHMbL}8K!m|2Os2@L+&xm*8m9lsR`D=*$vefvcp7U4nqE!1zpPPRv4Cg~1*jnBa
zwW*Z>-wQc!xVH`OakHK~hd6hH`<Q!V<Jf=o^IW~W%f1+*->G=70m&oBcXIBhowLGo
zl(P2DQQ^4ZdsE|?xBmDemmGZheo0wBdgEC;`pbXKKfI6Ej~se9Px@ZaJfLytLyofb
zQ4X<`{rR4L)>RK1@zlvP72I1mPuSm_D|}xh&v#C-zq!A3uJoS7dCmT5A9Vimy^`-i
z+)w!a!*^-EGj-q6@sr;j*Z+mS1^0UHJDg9P2d$6qQ0VU*<{a)G*1XLtd~c;bMDufA
zat?M5rH}7j+=G~}di%e7H|H_(tcQB%GG+60Ui4hPp`CHO_Z;=(bkIKs&ea9kkA2lX
z=)J4`%D$vsqmX|5&ODUqWq)*UqMdx*eAt2agWgZMkAUo2JLh)aIl33oPF~}FM%<Cl
zc;Adq<9JRl_UWDk(nFr)eWExgzKSQ}2o#UPvik5l6mds9^B!C~{X7rbi_7AqxDK6@
zJh$)IzvNfW%g$lYx_WN?@vFZw-q^Q#@<Ki6M-Kh0YlzxIa+P6Lz31jdzH#+)UZy|3
z^oQo@K9F6uJBkP4d5k|gpNaeKAtAqm;rlo7!2K-L&i$l&9_YD!5%OpKv=cY@H+{?t
znuoXp)obUT)&5K#`OyC4y^;HL`@Fos`PaT^zqIezf1$j={^<U%s7LqvUEh1)nO=A<
z^4a}<SiN@iAjf?DUJvi~ZTmq{Zuk3U&)JRqMck5Kif`hb_$6NOH-2ef@;$b=FOG;;
zz8lh>9LV3pGJiLoaeV(`oGi32(nDOQ54rYLaYj5*HXamzwWF7Lh->r@=jA!Qd?!T?
z`SK2SA+EdsRkrT@k6l21!0xpR?L<7`XCbm5_2k&s>8sv7x%|=nyL=h4b7g)hkMtgo
z|B^$lJlXp_?f4O-7vxV+Udx}&7ti|9&vQJQ2h>hJ#Se`mZ<hy)C)&}&bK}v+yo?9w
z>AAQB$rGRHkAM0qldDWm$gZrnd67eJdO`Xqvn&0wkl%*L-{cceULn5mQ+Dg#EiB_D
z^tWPt_rGU=AM><+>e&JQ+@o3_{n(*-lN0iTC;ixoa;UHV_+~%QeCQ<((3f0#saK}A
zcE%&ebM^FRH_Gf$yu(j8A7yc0d-BbPKIUaT=%s8P%GQg0L-tLcd;$OXWk<dXmJfyd
zhxcUSruM!Q4)+1=!u#Yf&$TZZ4^Q?p<Epps@F#x6Pvq6ked^^6VSYe=a`>Tkp+A!2
zyypHcEc0LIFGz3uC^SF&J^kf9^szrXH=Cb&`kDtmAh~$Aznj1L$*auE{T4mQ_Z*Vr
zx%nv5lU?u|c_+JM_mE%uj-4OKyXDzofA#Wu&!O@8r8vVc_^&d(Lll?E3FU`n^EEH|
zk9Oii7RsyXVSMYWtQ~)32mH|e1wZ45ke&D*Uj7OBDf`tg%oo`yIr_=N<-g>T2R+x1
zzWf&t>aCacz&CpkzkG)dtv7oq+F$oMN&d=C#Bb}S?74BoY2UXv$BE<W#cTEAwLHcC
z<Nnt9&V8fvlJ8L5GehS>d8zwI_rLOQ&&jo~kjww*izog`59eX|uX)NV-GjOxmG9ZN
z=}iy)=wqMrJ&yT#P96;PcfaS{%HQyB9>xvVgMP*nkBmcadV2rt`ziYRyDaoH4*sma
zJS*he{*Hh8%Ujrqdh+Q_zVYy)eK?*nIrP@wbMZ|(b|dcc7x@qV_$j}^7rE}6d>`rW
zVmM#1C%=pK{iN?5AAWM!W`F)uN5Of`I$0lb-N($f$*$YKyuqA>U+(tKI)@CIqrhK$
z=csHw-TQlwz|P5~r~c4${n;HJtfzUg8~!5hXy^S7zvl1!#k|>vGJCbI?1Mk>d+SO+
z{Ao`Q&*>@tDT_DW%UBoZWBTI7Jn${Pgzsmx!@qgai(S!=-pc%#f9a>*`jDgEyx1W>
za!+SHte<(Q5ABA1v2)*(vm0{hZQbz4uAuSo4gH=$oT0b<O`am|u^aFCtb_6C#oxsz
z`xyI~;ZF<y_NI+yC>S69^4jx1TW#K%3g7)?=wlb3J4@k}LpK|m%ra~7_ZFl6_|qQ^
z+hIWQ@8<Y>P&4hj#qT!Ud&YwP-m}6D)>!w^Ifl(#IAri=)3)AemcpEWd;Yu+o|>gt
z-sFyRpIrHlSqs1ZeAzy)9yM#>iuYDnam+twEa*>Pf5*dj^8PN5@3_AjwZ;+Sf7iR<
zeW`mQ_HG^U#Xh0+_MI(z#yk7tm)3<Jl5c%NJI24bs?09Qk@v7`NN@Tm(@Q)1KD+k5
z)j0e|KX$|K`4Rcz2fH*c^4Ym@_^tNrJ6s3->@U_q{u!Q+oQHg`;GE~YB)@h3QSbW$
z=RfZW)r&*Ui_VkUdB5m;1;1OAFFR-H=kLgP4@nR24a0M+?<>sDxzc$+9B>|&H#=v@
zw>>uxJUAc7=fws4B7cJX$Ug4xRcJ51hv@H%vVZntUt~w(oP3DgvlI5s9@vd|M;`wX
z-<+@Pm+X{%%a@!JoXgb9kDO=JlcS$9KclyL@mHRupL|Te;=NY)zc=MP>3n3JoU`;d
zj&rBHPkVOZ{3pNk{?ECOUh+Tlz$e6mbu$jW<nzWcFX*02**qaV%|~9v?v?pB`xYng
z1J!GX59e0%w$9}HJ&AVKfxkih%*(k~J-c#GM=$pQ=Bdoj@jzd4=s~{z^fbPGOCG`>
z<!LbVFZ&p~WM}Mwy@`{~!=BsEpmQC&fc(O9cBG$s0P)N^>8Bm!U;2@2|5UaQv8S*<
z`?jyk>!5v7ytD7wC+&Oc<qc4tX+6X-_0WDujyPxEHlB5srz(3cZdzCU*#)%UD~t2&
zG2A!ociz|9r|etM`(gW!dhNY$<)8Lt`@MedH=%a!Y3;Y3L;dM5-r4``liJZs{}8o<
z?$_`kE-CBpzL#H{FTFfh#ydU9(@xnvoOb$a2d$Gb{@iElk01Kk5BQgTfZv2=^&z@n
z&>r&p&_DU9{Y-mu>7yNgR@N>H%@<$z7N?<hcoB!hUvXJnckk=Ddh*42<C4p6>{r%T
z{HG5&FqCiKvM<;V=!HjokxyUyi2lZ<pL=HOroDaBKBz4Jr-yyR`WZ)jv)=T<w|qi;
z6wl<5>d6sT<vI2p`vX16wSVD7K4Sj(D;Mn-@>+SP{X-m<&x+H~{$u~ell@TsgjYOi
zXI<C_yYoDp7k#1prf47C|GtHLDthD9deaZT)+f}*y3j|yj(>Jw-?Z;rFY@tfKPFE<
z_chK3^b~L82inVH@y5U91>wB_yE8xWf}dD_`>V3~^K<b45A1|q&QJ6b=d1_6GB0tP
zeDdUbcyn*Y-t?oN_~|~u`OG}T6aGeD_8~vi-#$RD{`fVn^98?D7Eky|n8%tYeY9iO
z(EPQhpZO`X8{^~4{N0~e2jh~fopHq{^JeFGf%s8o|Ll{$vpeyUeQT$mIN`Z|{2THg
zC=RkaaYX#^T)nb$3VSv_y|p73+SmAzcFIuR2JKJE+Sy;_Ir19gnWu5_Ag}VAJpN>#
zwtnK4xQP$r$t&f*P@d~OzC0GX-*RuHte^bZy_5ONJE1&P-ii<X%wOD+FN#b0X%GFq
z8^2HS_e#ZM{>l&d6F=df{7bwLhuF1v?_A2C$g@u3qJ2qRG`@aNy?7;liR;SVcW7^1
zam)JiH~EeDCGX&W_=fT}_73@3i012l4WG`7ejnuC+kKt(^v5%P#Rc-|VI4v=pRm9A
zki##ae$aaHC;J$5kH=2ffj9}Z!#la+k@2B={wgkrTgv949Y5hm;tzju&f-U5+4(E<
zOMb?$eaAzN=gz(Skl%}U^kDzy8=gPuZyZQ3&&|&~#1&=jmCa9?{^Y4AH{=!X)`uO6
z&+JfsrtG=(;a}#hEPjjUc(cy@nH|};<lWji58#*o(3c<3k6nd&lc$|H#y{+z@@r-K
z;E7!NXh)vBM*ibIFZ4U{%5!;-{M&f)4E~QN>uEmnZ1XW+?d-4CUHlZEvd}uRFTcOF
zzj`lZKUHu4WViO`upK)J(R->al&{<W>~qTaRAy(+d;Hcqh4J0{Mf;NRj3fV4wr?pz
z_0axnT=rtW5P!v4?afDCD~}haojdHi&V}@}&)SFSBOh0mZ<D9Let7Vle$ENn)5AQ;
z#S8wmli!oCK9oma{rQjm$$n;E;y3()--;vrgkSLkJd?-IJr@_qC6E8g3w+PwI|Sc<
zxPO+%xv$V(JNbro%FfAnVz1T--|W}=D_aNs@yS2TTbz)0dG3Bn9CwZ~f9>%TBKhK=
zGJeTt$MQgB`kN>FF@OBJ7w|p+kDlWV|Kb39R?jZ=Cl{}t^9%lrS9U56iA(&N{_Ndz
za;y_SHGkuow|hHf>rFm=$Q6IY9X!*^dea~O`m+z~CvMqC*+CJze`hCtu^0YkpHwy;
z|Ay=%)WdWBYF~u>AKGWxXE-1AP(H$b*n{&7dtyiIgB@uf+K=|uH4E9L_Usz!$Bx8b
z>t{W!H~!@r@|fa!b^m^mpKA~KE4#L??1w#QN3QkN&$?+Jmgx)i*WS9(E3_l=Nj)Cu
zOAel#<M@|-&OIJKqE9$q<3am_yxuuOo+K~w-HYGbL;s$Ye~;1Mm-Jj2`uC;$dsF@{
zxPNDA+)BqCbMAJ%yZn7h^~%uStvu$SGZ%aC-T(XVdWG$@59`UN5B=!x`yk){xUZ34
z$}91PN9QtmCUib>Kjb{*d=%D`qn;e7zvs?1p*-h8sJ(Nfvggi4?jO`^@0@CUNRE3n
z=Pr3Zl;4P7&Tr6pfE;>2?Z~IEd|f>`#*<I+<1k+m2gG6hiuK*U>!UX`Z#-Ea_0Ac1
zVn6K3Js&^w{h05<_+w~q+LPzGdjfW8-RNT-Ld2u<6&|evKAqnnIr`&)J;Lxj?jArt
z<Edx=cw^Vbv)=lX&rU)lUwiWCr(Qe#&D;4N($jP8VJJs^7Mc&tY8Te~cQ57V?g62{
zy9a&W=X*c*x7z!A-0pXMU#p%RWq&W*_qZqgZsp_Vy>DvbJzRLspyyeS-7@fJX|IC3
zMLXYFJ9oIpl%M$C#yIr%`vL#X=s$k+^i~gj(5K)#M!!d(zwtdM2QTW)!*gZ4)7x`;
zY3KJ3c*mRHr5lfac=TL*^P`u3A@V=|uFS9bljr=5J@W&oo}92hf7DL=6Nj{S&*=Lg
zW$o>s+KZc>i(^oJ;{Heb@b{FyyA>xr7cccU4msi{bl!pTH2aBs&pAMz<{T#94pDwC
ze{;TYj&$xaj&qH?kiUc|f24;zoqTfCJEzIp*)@OVht3Q1cAj&-GoJefe!*_}1-n)j
z2YlD&yEJl)&u{sadhx-zM}ObP(bso!*1@^Ods_35Z--^`#hdkYzR@1f*44Sty0Rna
zoa8+1oaVW6lQMf?ht4hfhiH9`Ltp0vda*ZhJtyCD=W^#4^M%ec>=3{Dn+H9~*H77f
zBL3**xqA9Ihe3X69UwmOKtFb59j&)^?k(`;yhU&4F?y*t4|?l|AMGJt@ej2R<tT^a
zS#S3$;zoFn?VMe_C+YqjF8(}skLg^i-n!7+{H#MbzIx|TaWm|1KI)C9Ob*_(&sqm@
z50B>Oe#ZTuam<4~m`6Asd5~V_r`~ws{eyFc`(a35dO5dy&*J?L`=T#<AXk6+tN7u*
z$+?Pr`LsM%KA?;T>u!Ghn;zD~^Kf1GBR=V^EWVPfz4*Z|$!AB-aqOMlvTyUZ4tTb1
z?pNIJ^FMKho{+xE<j_Zb$cr)_*$Xsp^Duwwt&CUW;3F)PLq2;9`QYF9A&;J+T>j`@
zP8s6gc>D`b<kG|V)<t{rp#J1)5A|dJ=5HM9ragagKLCqzy8pe)@P5+$f_nq^1I8(~
z@Ba5sl-VPDCC9xkp2Kwr=dWEjKkLSh*$cms|JdjFH9NOYS|8&;c0@nxhDZLdJ+yAc
z_QAi0X1&~#i_iGdpWWk`JoW&^HStgUq&K^@F8YZx_%7PRWyf8%#oTWv@&I{^{MUKf
z`Ai-R<?HfaWq+U8cUH!=Zt`R6=RLUn!nwvd#<|8l1v?POt)KZjcgz3z2cG=BV&_r6
z8*#3d*LyBc_gp=c&pX$fm$>iTU>x@9y`%NP3m%<Q<ni);=NtB|zr2pUi~r8^@+j*-
zZ|jFI>+8HhUv|J=!v5;njq#mt%mYv2KVG!MxAByXM?YwOc((q|3+6{J^Ra%AeDd^z
z&Sly`c_@3W7xI%V`Of+7A>0eNCsD7z`wi%R#Pdqg_|W?T<EmG75APm@p2}f6?V-Ga
zJ<CJHQ}I=NP!<pDE1vTw{>z_?V}EfjXGhM_`te74Df16V9=n3Y_}snE7Weh;^NfAj
z{-kW568FtlImF^P-R}kL8{|Rl>|g4Y`JH<X_Z9qEJQH8E6W_8BAD$~`p?1(ZDMS6O
zpLMhUL3spZ2ka8M&*N{({K|gj-p+du??=M2dKljOd7tOIYR}2>KF06#>FJ)-`vT}a
zf_r)SoAs3sI#1~b#SLZmspQ$W<t@<u4YhZUGp;;G-o;PlRrY<zKAj7-lfUzO`GCAo
z{w4nuCycA!zQTT>^R<0n9wiRg57`s>Q2gS@^pQV_1M?IHUpnizW-n~`)`>H2|A#pX
z8yr9JgwqDhS)eC=$YIaPFJ^w|=<oge-;eNjM%>qj>*riv>H8dZ654_Fb&t<3yoa;i
z>;tj~?eUKnd4>7oiT?6t`@i>r@+tSW;stqlq#rx=J*<6(Jm+NR9`Tvp^q`mdhG@QF
zz5c$>HI8!FURgh=9vUx1^R#a4z`8*5!;><^k8&0okKM3G@x^-;?`h-<>cjXe-igQF
z+lY_eQ+toydCec+`sCG);=ilm-`9ZrR$kzKo4xR3c4NO~r|c545BsZj`Z*6dM~Z9u
zD?3k;6Yk^UpL+4s`kNQ~rI)<Ox!yj3PwPsacq6Wgx8#_YxTb7eW&Ol+@*)3WN9MsE
z`31X?H?t@Afch)5JAU9fzc(&D_?@zKfZ_gQJ;QavuYT4kYkl#8M|M+`-~D}=`yO)W
zQ*76rPijYwyipz?UgJl6H*fmT8yZ(0!48xmy{tcE=j;)ds`nh~N3Rg+$$m;99>hgu
z_QHP1QLpT|cq|`f7oooN!#8_ThW0Uet~|%{aNhg|(#QDnV*Ra`GW##uQ}_Lsb#{+s
z{`^n;#+x!ejRWz*&WtY}g?_GVzcO#~@h>it59w>Yl<jZ&^B3_>nO}<A;u)UQn+L?d
z^%ifSd9Z6`^U#idkR7lO@t&X4*Sg7vmE~3THTCQSZ^oxTyHch%9>o3dy@axRKKw%G
zF!?h5yyth`l!rTSlP4edoc}q;tEY#&-tPm9^ZA^EF1~2QEL~Y>-w5{?cF(@#FY*<B
zq90_xcoql3eM3C;oF7>SaouxvO^$l@V_bd}>aV|e0ObkpecUHG56j=&?;3}EJV4{)
zLx29Go&M%+K1F-&-XHmodVT`=hw)kk-T#X7{JK^2%UciUZgEEZ6i2KR^!<jo%dW*a
zd8#-je#qZFw||N2o*M^>Yhip6XS6d86lcUU^Hg?UuPoltgPz8R@;P}h<R|=}zlqc8
z&BK1AOfJ9TkD*-qk$6+|v+jJ8znPEy$b4bQkNrY^2(7E<c!B&mM0&F)_F$e6ug0-I
zl5ZaL5pT>p?9V^MSA4J&?{D05_|A);drxFvavo)G?2EqSvM=rQcMt9RA@)UY^Md>W
z-^RgL(NDYIGl%mdhrRJ*-#LbVcho(obGzR)_&cWlT~=k!-GAD5+?%?Gb?*1veXae<
zJ+=F7yjur(gZn@BYW?u+J(v3;_RfB+Pl)W*x>;Ae<5B(p$<Mm;aq9@JlYF|kUQ@<C
zHE{ha6ZReIqu%-(-}=*6KkdxhJtTX@7yE<Sd2b>fgy^~W$ZnPK3_XYRvJU*sx~a#P
zIH8O`@sPiZE6VtQcvFU9y)rZ|xsZQ(t_;I=VO;i{eVDhn7MAe|osZ<Z@-HYabB`cz
zgW{ocpgh?A;=X~LV*VSvx03hTx7;U+$Ks9s39sycAK3>y#|wVt75IhfwNtM@KFPHX
zS%_cl$+2GgnO|sk+S%WoH~2gKLL^5&WoSI(tB0XK&?5`O@y!#OKfR6XoMB$<k3Bm7
zE3-3psLZa|n{tTGDc+~CU*m@P6n*h!{^B=#<6q*Zcq;EC$3BT)_N+{Pi1^jsx@w0{
za`=b%MlbQ7{_Na3Wnpn1-TRSz(mcu4j$F^_LB9~~=g@ml`@6hB9w84<ubuwvLtfx_
z8qTfmlktaFaR<N3@;vqI6?(2LZt*8}<(zFi{m7N)D~EQ2PwU6-#5?QYd?_E4&xH0$
zo__92?62Zi_})`H`*gV9lMC6OJj%ZBe$xGpJWPJ({fs<Kz5(?YmyJ`5H{JG14}6)Q
zvi|Cop?0BM_9L(KeVOm0%*(v#=f24Or#Po9uCY_!)A(*Cv}?c5WdHm}SsqP}dmYcU
z7a!%%{Dyz|PC-BZC*Sv+p5%$Q%HlMA=x;p!ZCvL8<Am=4@g(n{KRwM)*}tn`JmcsG
zLp`)tFCKXh>37@CG1`%@-tWBWV?KBgr^4?;e5V#7|8UOW&*F!8&0fQ}BW}sx?91Yp
z{X|?M&wZ@C+wUHsdsgwt_ma+K?s<KOr{4FQ`a{q8m;T=S``rva-1n02_m93~@?E0e
znfm=B{pn#`@yYjz#y74qefuufb;8+S_AB^~);PXj_4`WSN&1eFe#WJ@?<~y^>Q8^a
zhc$2CLqhhQh2o1iq@B0{y|3rj>{frMp8q(fhjEbqvNQH)KV`S}PtTnv=u0lW#2@mV
zYr^*CB`zuJFRqyvUgVL+h2i&;=B;0d?oY%6<C1TDaRQ(Ch4NYZ2_!E>yh3@3xZwPz
ztevv_9D1%S{|V*L6Iz$BpZu0x7*~EKzjS^Rub_NXekxy*@5-<B7vI#2qssDl@e~>_
z)YE+ELB2fB{ek-yc?SKAhX?0md9-@x9s8?1n4Mb(?di#nl*z}J_3_+!NIPg<v}5n~
zXZk_wO+Vw%Q<**4m)H@MKbbE*eBWXo<_*OWW%Gj86+hNb9xY$AZrY1G=E48$ujZr7
zzwu(;^fzxj>u=u1G4Bwqhj|v~)BSxiJ7RbI2|xVBc|d;Qyuv@&E1p8Tr8lIf`FhV2
z&X>Qqcf>z*UU1%VZjf)dmsBrb@O`s<!uvS)l-k?(opanb`99P+!8wCJvk%Xmdz`1_
zYuY;>$fwA6FU*duhxuC<^Mu}q;{|{8q#rrf)%&;NxWV^G^d(0eBj0(+x|tXKJ*PK*
zv}1?nX&mz0<7&tLigw+7uc)5=vsZRWAIN?oy{wz_ZfGCg@5smGgY1cY`hHcO=l3u2
zN$bY${C<bsSr_-)?8o}yQ<)t3xBL(K&Y!;Y56AVMi9b1aTW@H5eBqD2%Iq#Ihy1cz
zJcejIc4PmA;+OT8FNibZq5RT*B%az&#AW-cyiWcrPn6Gz=j7-w-jf@m=f;6q`KJBB
zc-qMW=|^AtkUZG_fgff03V(s-V}Imd)-k;2ur5$M5%<Y)pF@swnSDVVv|f0^A0Ei1
zuYApZVxP9(nU8sh1LBN0z@G4AJo+gcABNBIYX4J~$B+-Lmvwb6F@N%vA^G@{@5@V+
z*)4nUoIOEy7}~MCjo)~#-ukdZ=NIFP-^SrT;<)v}k8_pw^fF#Je%M|=^Yxs6kcUU@
z$%S}__$F67`dep+7ya<3-hAX?{8BwR`mwjL-hAk3UgXiwbMlnQGat{{JAKJD9$w}B
z#<BkFOndeK@y`CWw@#ju=bXWQ*fn|jk)vKa`jV@?=jLHN{wU5Gmw&2vKKDEvSNm2$
z_qpyz&BOhv`$+eh>hXau^VHtGhWk+Cki)*%1KzZQ__to-8@ntQ**CjX4*O@-Yv(yR
z{LuXgKhsVe<5&DY>_<Mi+Uw8nLb=*Qaw<i6p8H36q5D*MnEOokj_xVln|ki|Y3@5c
z-}}WI-rj4Ez6JNK-(K>_yPg@>zaUSR2axaY66o)_-*>wQ^&ZA^_pQ+Vto+j7(R05l
zAC}iTFUr&9^Xlc<5P$M$`Gs?FF%RkfUPoEJA?`!@hx-L+U*$jI41WxfKl4+5B)*A5
z;;wjTU*n&~V}I;iJI~2yxBA;J*d>3lzp1y+@DukR@_2cYvUnhWz?1hh-Xr;Y!2WLV
z^~*o7#gBg8(RI{yyDz-<BfY!)z1~G{zvS-8$Mx%i&b{&$=NRWxd9HTyNb4`Z49nJ?
zeaKtYD|-%|gUwUD^No7_LzD-TqknPS?(aRxV~5s5+_DbV$9YP<JYQMAupK?{r=M~8
ziFufx=j!nrwr3yEIOeNN4m)z55Qph+K9C(cchVov&OiLxJ%IQpPKtBh<M@6NilfTT
z^W=E`^WA3u@VKk{7xrIf{g38)VnE^9p{MQG=ZG1K<%=%8c!{+R>|eOx<pUPIa?5@N
z<BN~pXNj-&1MjDd=lvGFp*Srr58nBEf8Y4teg*S-e#JX~dgi4w7A_z3)pFbZZH5AU
z{H__&@2S^T*?!#Mf6r&gTYv4^ea0Dz^1N@;j~>vt&^-lz=CAzOJ(c?l_Y)8=;=Owj
z_l54?{JzxwfLG`}s`IaV4fjFebN4>#-4D5!qObcR=j#yZ<9>(!_BHE_C-;)}HE6z&
ze&*vIPkrcb<_qa>T;H#_H+BE&p36L~oA)^4iLy8r#tm_TzWh@h5U1>K_BHW9zU7{Q
zzuUjm+b`{J&~tH2d=Y=_!^X8g+i&QPXK~1V414w64ql;sOnd$yj^fe&Ab#S>`1UvJ
zz%Ka_-nBCxyQ8Q3WaG0J<Kx$TH9s(a{j|fQxU7sP_J&XEPG54&S3A7mL%wJp><Pbk
zun*7^njbzxeT@?$zRAHaIiB+`b|9~p&x#MeTe2?v%e>hQ-te!zxJ9mU<lA^67cb(2
zb+sSk&-mJlui_s4>0uq%q50F>b8?((#I<mq^tG-~9HJLHq%XvaI7wf8;72?6p7aw3
ztsi~NM|>vFyz!?WJ?S6H(H=kMM_!2d&O+@&yA18!bM|Un@reJa*S~1@-TA6`EZ&RL
z{F-0!U;9446qoqFe#($ukUruS)E{c6A7r2SvOd-;l&2gbdHO-;Am>2)nS9>7tOLHq
zOW)bJM{^HhAH%zS4?5RrZy!|eJV*~{-9q%-K1na}oW9oCzA7JQFL-lqQ7>*P<3W4R
zq53dx8J}MA2xa$N&c)_we%gicp1%0SpY^m3&f(Sv@1EP&_=)>9`G$K?W%hwzaSjja
z$+eG*JK~Z!#m?;?;u!x3{e>NSe`_54STFe=eaLaoAm3u=`~`;bhd$Of3+ZRR{8v07
zNBq(rfArK&Je8l5hj;qh7qlneeAp3x5%;VozQjNK0lUVB@$C=vhxR#TdAfdhpg*}#
zzwo(v;N5e4(NjGnmtEkCJnfXBe&iPWckeI!n*Z@9_Ql`i>HLyE@+bZU#Y6JM1##0o
z8vo}n##PVn*im>t!C&YDL%$ODpy%R6F|KxhZ=uZZ#9Myqx&HJ}W{>#RkH4CSGI_=^
zfA-GK$tlVY{(b;E<u_q_a@h-B>EV6|PvW=rp&xzOk9PJi>u#Tb;=3}tBFDUx@oFCI
z*f`o7&$_ZB_2xyc`Ff5Qd~2_aSMfxg;pf_kOQAgc;LrSxuOEKAPx8LVJ+JTcy|2>F
zceLJDxfgeDt_<C;E4w$>pI<wtL;kNlKQ<q~n~<ORJ+yO^`z60K@!gQ`yWO+>^4jx1
zTW#K%3hu$&%Xxp|zR`W0_bl!UwR5khz4s{EyB~BP2<byV_ebvE4tQtZDVI%}q2PSu
zyyHCNJ0<5Y=cA$@b^pC<`2w^bX-5uy>46XTNYFXcIY8Ndqilb$&&gkmYyY&*ktZ(j
zNA@h9v1b^@Kk;1NWE}HWc7Aq_P?nF#5BMLy=7;7<Z}Cbz*N)!ommi7`;*I@GTrh9?
zinq?0>e&PS@g^SNQ`xxcjc>ig75TM&n|_{?Q;eV8@8|4K@&|c`xC!wmzRK&=TQ72q
zLq49hSC+TqTR%u|>tKDz)y_OUx1RK651xCkC;yNq&<EmEd+TDq(VtxNupabQ5BY<2
z)!(`q*S<ynaG&7^_91r1&b2oVJ|O#Jzs3puRD0{r4wbDdyEMLfe#=gj#h0+`-YdMX
z6PLtSank&q!$QBKKl^4^=8Y$I#g6gGU**r_use3huGtGaVh8p|Wp*I%v7hSCj>xrM
z?2tXN6ZPa8$2>zh?2a7$)ti@b%@<G31CSpnn?Jei3;)LP9IwW+ZcyH9y|gob{NdC0
zQ~XMPBcG8ESr_%@0mVW3;*;LyrJla#DPGXuI^a`(D6W`~^)xT@@SNVxRqC^lUC^Js
zLHA(tI^Rz~d|GGv;m7>M3I3#P|HZ3$vnzHF%@5K~9x6}u-3UF^JBQ+py_<*i!3R4u
z5BrJrG*A8^4<HA>?8-RG@_Oruck5^UmFZ7!_Ql@#0i?g@^fV8sy?I)PR>ko5!Lb|}
zhz#`V<vu_hwjYY0;u60WZ^R|%FmXeC6Q|`p?uWFO|EZV%$>+p<d7km@tKytIL4Igl
z<M?jT`JDdpRQpI67sXrqhJ2D-dA5B=S^nw!7Uy>7H}@CLozAc1>+e0Wet0oY`KtEv
zTjyH&r+a4kw7g%v`zU!Yy?h7bzCpclv^O5ToTJ?%7zg6n{-Hgjk9AkZyLqt>WqRl*
zuQy-%vuk#0KVt{%n0)r>90b{q{hI&SH~EJ=gB{to?faoV<k`RZBcz}4_@{G|`9R-O
zu{-E@o}Rm}Vt4rRo<*4*T7QUl>&)J?XHVp?Blmdj+kC%AFM7J4^ZPvCB`Amb8_#^{
z>7I&yklrEEmmJ@L8J~apt{aLs+MAbnC{D@a<&RKa#ZHu=xF_zg8};lDviA_pKMVOq
z7V=}?WB6X&eWUwQ_nq!ToxjCr>!rQ^>_;5(Jw5)7Cw{O)-$nZOP{?!7D&G2@g#Fl;
zoZGeM2fhb&pX<3e#V<p@wl3DIRy3~sT-@RJ*0ad_r;81_X7iU4eEEGh`*knE|NVZK
z9kYM)_gyOa{FR;i{Xi&}zx&;QeNun-?Y?hiFTQKNWY5Q!zJJ7k!l*y~^hd*X7*Np9
zJlKhSm3(@+w<p&<xcEX2J=ux)VP9c4#<L&uZ|7Ti1beVP{9OFDpNLQV%5&#8c4I%V
z?(z`#8}#58_CfY7j)*_v0zJcV)YIF3%<knO@&Iv<-{48yXTSKyt9Z)}d~YV7V3+nm
z^P!(O<-S0^!JpZqb)}cMp)7ARu6c+%zAvy&__Ge;G4x%uaqwk47}_nr<A>sovfm$t
zeCwy)`z!Agvho|eSx4(-AEqCD#S8P%-uh}M-t$NNtGCY9M}K~1T<?*>xXwR~$BwKY
z`TFw<D3A1984uP?dnjMv2ke-?spr>WKk=I#@hg0qSBUf@4~qNV%PT|r@EiT)kL-ef
zLHa>)1BUTL{IKtdgUbBn>@~Vi;Uk8P*lp-Co%45otFz;T5!;OD?EUqBht589;Lv$G
zdmTSw=zL$7m+S24?AUUrrG}3fdhtmcj$fi9p5X%q=9uj_-}_?eUWF_Sn#W>fpehEY
z&VR~~=QevfsT8XkEOv?vv?&9F7XHWTBhPp)X?NUu@mZHopMNRI8{2eQquP;yx-jsA
z6OSDHqm5oq^TxU;8-+#&dKd$F`Q5%VEw=hLZ@rPWJ03U7-3Q;f{>y2^p2b5CbK5wb
z$UtPE1_s)FpNjhTjC`ZiI@Vy~v1??YZ5ViDu0w8|?aq&rW@FnB78Q*QL<Xv6VEpcd
zYhOI!@+8En1;&n%fyh8)ATm%D1EtPEc|DK!x;^V&H?C(~&vtrWQ*Lm~E8ayp)5kqw
zv(~v>dyefQ1CfEqK&vv)DtTn9@)Pxn3`7PhXJFO7-#Ou#YoAWr6~p`SW|b>vKK7`E
zfxNy8?J08d&JV23k7hXhh=-><mo^*e6EQLn8HfzD3Io;dAI-WCi{r-o)}HV?m}tk%
zwc}Qazfq6>&Or2&=K4v!`2X*#9-l@AA_M=8fok*i*sr|5hw$IHZ{=rkeal<lp2AZr
z+hWu$G7uSv3`7PZ1CfEqKs^{}_dKp%_bZ{?dT<|QMh2!g26la9>ovDOdraEyIK8b~
zoNr?pm}AYsi!A-p$TVVQJnXc@g@=EB`15JI;jZ7g>o+U*dm(LCTyv414DUa9RN8LX
z?C%=G@f#au;%Je9#xl@u@geHl9&x>~R@`$(J8IH}hn#%dOKHT&Lp|((=j?LQY0I4O
z@heH*`1+WuPI~))#wCr$1-IVk@GGu;H>no!QZ25(@14h9F!GtS-SMQ;MxA!v=m(Q_
z#b*8fFt2f!-ulf0&Ya<IX}jP*zqon1$Bus>ZC6}po!=b4&RwI@cEwi9kE@NJS^e_<
zU1gLP{~lO7{XMXDx7T)mzuheTv*KZ_j|^19!1s53V(rroeJ!mOo8=F2+<$!+{<|F(
zIcS?_UP;TvYHX#JK0_9LYo8@HyfbYS)}m|d+jAM%`Ob9BJHJ0BjTm`|c2L?6dTu0$
zGm8vF1|kE^WnlMpPhDg4lP*Xj)|ZF723~c>kLP$ji5Pikq#ZOjx=cT#|6;!DSDUc>
z*tAw`b;jEDJLGoNyVdO>>K7S^3`7PZ1CfDdGEnZnQxN{HrrdJ5^6TBMTzzA^*7*DM
z?Y<7}{+&-x(Z89oHI5w_hzvvqdNKp?JFe;L?|a7eZp?c3<e2fF<`({aFy;T$Hr7Q3
zA_I|u$UtPESqxk}`&xG_Fzmx5V&tJG+d;eE!?vsb&5CPr-2cqLb)Oux+6(J-r9B=0
z^Xy|S2F~4ZsWtC<<L)G4d-JgM;<GNFKL1kEC>*r#KUN=k#&b!l;N^oxocLhJqe;Zb
z!`F5ozV%eB_1?PP-=~z4U#tGD`aXX^uPKY(^Vp+ltK*?}Jv8$Jw_TZ5i#^q^;=CdQ
zk%8%#fm-*acI`j?u6SiLpF003L!R60>7-q;-oHDojLq1(-5F@rJiJ`JN}W^U^Y5JS
z%(YLaArI|tw^9FZGEnU~uwMNd_3v7JlZ9A{3`7Q6iGj}^`(&1@_dO-46t~-co8NDE
zR9Dh0Z1?x72mNuqUr%2ApK0FMO12esi3~&rA_I|u$UtNuG7uSv3`7PZ1CfEqKx7~?
z5E+OJL<S-Qk%7oSWFRsS8E75@dHws<<;rRG@Avc4FYgsCm*1>*<-Rwqb^hg!Q)#=r
z`OK1c%&A=~ZE?z*X~b{xki}Q4)cZS75f@v1o44Lb@z5&kQ*S-)e)@`==lt2^Bw~Ge
zxMkuq4}W-n-vX@nx<>htfyh8)ATkgchzvvqA_M<oAf8XZc|L6yKBE0b2HKGU_X!cd
zdA}0!(T-~yb?*5LgmyZ;lzaZl#u-KiA_I|u`Z7?jzekX_-1@Q*rAG!L1CfEqKx7~?
z5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIf&3X5`r)+I7u)5Hw9%OVOrpekF!1BEXCFLq
z`0HuC@cX+yvG!?)zLqu{&t9XOy%EDk>^5|n&Tn-t(Ajaqh;2r6_Wt_6LuVg4aOgap
zy^bF-biS|4%XM~ic5J!RQo~0Kz4)XJ$1l;5JNFNMIBCHZR(&*SSDb0~nN~e>t=G~l
ztd+2r?isn?XXD1Dd1IDdgKr+T?lz0{Eo9+gC-k}dsx9A3^2RKEs;wWn;;zdcIPCnS
zTC8^5di6VflYLenx$Ts+Qq0ol@V}igcKey0O{>K$xp~)ry6EOlO<m!$v|d<B-z>kb
zv_5bDZM3{W`<-~!xrcp}R*L1?YpHc8_xZf9{AR|nhd!UQJ6=2Nw5xYHXF{4chCEgq
z9~*6dq296lO$M&NVu7b;+xEV+U9r-5npgYLiyyV&3-7+3l)_5$oo}vA{-rYxxo@E%
zKb)Mj8(#m+21gYxd@E@<*61JYzMr(J{^jO%rS`#gjeDWqdBrtyO7Yfi&%4gLYVz)v
zf0&lSYR@&*=0o~r<wg1L0N<3@uJ<RQ-l0Dzms-y(-b4BID$5VPxhl;ovBySX^Je<r
zM`k~b#&>4;+uS#w+`lUek6z`ugO9qgPgkRH@CBQ0eBL7cx~j!;_iOF4crW)Jp;h;%
zQt_zN^DJJ<tuJ+6$?N%l_pb7vma9+s67y}DSALwYR+URz^1?BP4>)pS60tHKFx_n4
z=us_}8$Zfz*Qoegu0C1qs^u&1ex>4mxzF>8M_F<je*e<wd!HKbd$R1T-u30$W4U%z
zX*>2(>YSuq7KZ2Za_1G>eUpK#b8Wrqzd3sIN{!lgN{!Pz2EI9RrGDJg+JBRc{7d!n
zr%-Mf7qiYqA8-A~JN94YlcW@uiqoZ@XN}*e`cnF}`tx?rqq6qzcBv2jt6j^Dir3{v
z&RW;MRNQLU=erF$euv8je4gZm-W&8Ztkr(X?O(1v$99!5P%1Aj_qq4bA%@yCYM<Zc
z^9TNP@g6f2vanjeF7+Kw*1bWk^A7XLkk{$8oW*xopJm5kyK>9b@?Gz|rQZ67^KMi=
zQ?GFv_1?T*<K%5m*0<fxJK?^gTrNN5o`b7>Z-m!+q4`Gq@4wS3#ZS5WbYAD0df7v{
za?71>z1oF()*2^Et#7T*>$MM&TW%aGw_V6@wdJh*snYst$FH_u-gaB99*yo->YUQ3
z=Z&7X@ftP1dXLl7?u+u8k9BS~mV1BIYVB&>XS4d3d;ZM3o&B@bG4K7T+45@5r<&cr
z+59zo{AS0`dW~P|yYqUDTQ9kl=Gm>%zSKI@`aHZBuU9#1zpJ&rUiLz6t@<^)e_r`w
zqvf=E9@1+1vDZp5?>sWf`{v)5sB}GArG2&c^sUzK;of_F_Ux-)Chdle=CR&(*KYi`
zn!fd3pHO~Y?`N7Nr`-GfR&CcT-i#aCL*C_;UfSgMGyZK_(yo~GcOToWzTe)1554`#
z84G!1-u6OXUgvW18inQN#mzofRLe)JeD6}V{%!g@NOskzd(e>IX5T;OHGVj+uw3pv
zTeJN)@A1o>cWn3H|L$vOPtAUxnOD3B=T&RD+V27KiUYOEnLO<$J9Ip9U0N+>eQz%>
zFLhoi*RI?zRLgg%e)P@4YVXDEca28h!R8(3$juw`dXKaF%olC>@=cGYd1Kc4Rl0xG
z>aSV-tL3j!{mX5i#aCYSdE0fZ@@m~jTBUzh{A>04w-*~X<G!c%E0l{zZQA$e>)hPC
z&}w+cs4F)e^_`guwPIGBsMgN1<Tbj!(l}RcysCG*yzYT|nw)CwF|U3L{9wxHEne)^
zRW9ar4#`_irGC*W?eoe9nk}c)Ij3EoXPtArUp(d)qaXgDb83=>+Ch24V=ta{{vw-o
z6td9yE(^;&Cx+ks&?5`acxS~Wb{^fgPztm11b+{=T5J_RtW=LywdZo@QLntEQM}Zu
ze^$KA^8a?J&wF3Zl2<GLrTUi}@2YK=HLtwtv+OIYzS4P=%GV#ecdsk=+N5_^DJ(Z$
zgzZX=A3m=%Z?E+}qLh8r`+2478QbUk@7sm;*Q|22c2=psv|9VT@5jrPm)E(tl$=K8
zIeCo}o(sw?=QVy_a;o*GO8Z&=YX5Fl*sol^TfJRY{?h98mF9KUzfo9gzEf%cM$Id0
zoO1oKUf-#eJHB?WpRxX?or7LUN@1mbP-|Y2HD1<!X+D)=*7?`?S@~<$eOFdJd0AMg
zA8Vhbe^z~^{M5RRrTSOu-{E~-*8QLQto6&P&uc#q<yfyuF>4*O>a*sRRiEWgVSKAs
zylRwO_to{nkgrxLSL$D_);`NVTcy5U>s)D^2;0|7ue{`D#jkSf!}!|ha;x1-RH{#2
zejeVRHmZF3Ci|>Da@#3s#OCm@>u1X!uzKGw(^6QeT{fzHmfm^UQ&wIR#%be?e0t{D
z_j+k++OAmXxgx85rTa@>c|fJ(X0@;Odx6+*^QWe+@L5_b4|(}d);V2$*7q#Us%MX-
zeh<;ppW`7ftk-$5QF0rd7uS2dtb2@BtuJ-|Rqpe=?3tWv7drd(C%2lb&@NbSyULQ^
ztax7QxLNsU$@&U^e_n3BUhR3al>Ab1v+~SRa(n7?`AfTDxp6YKJ7etjGd-I|9{xKI
z<>m|a%SzGvr&_Vnahuh?*59?o{^k8XFz@wkw|yYgH?KGs@>=eDh;rN2n$KnRZ`OBJ
z;XP(gDL3l<MWy4oC+;bjm6x|$eHf2gr5whyo?gy+574N2H|jpGUgL!K0=4p1?(eQE
zZI{=6U#*;0@$Vk5$1@uoRk-l2B(%$RDYtt*RO@@Vy!vP5y`}1#l^2KOX2sFG^PEcK
zS*zudmFB~>wl8&WT<P<?*RxUb>TL%hkEPzv|6=?*k3TT$8)>yzsUPIszTEqOdbi7p
zJLT4g>sD*IR)5vMT>YzUSMI%Vx$UaeBd>nt{@t{^+U2EptINrHU)<{T<=SnlwySsC
zthfAj{dXopy-Vfat@^yDeE(Ljd!TCN)*7!{^}K9etr`!@)vMM#x7=~+-LBR5*DU>N
z-B;LQy)nGMh~++aU$y0XNx3}KdwuHdU-g#X<L|92{C;wJ<)hwyQmg!W`$4Vxh5hTT
zZ<hRe=lNOtX}i{ke?O_#a=rO)mi)ZpV(2HzrT%_ov!2)cK0eC-@4uJM`i`X8_3XA%
ztk=0Z)FUfD4*fVh2ZjBry?+S%h2z#*4)w0K9P-vG<?w#5UgdiGQC6PPtoppyx7K@{
zuz%kAm6DfNJSsJg=XrgHl9wFsG4n?AtrhcHpGM1Rm+!(d^{msqL9OScOnn>meO{e-
zsWl&HRR2bO2bA|X_1@>R<X0R2O3lC4?_WywFO^T0^5atWSZaQy#`nC`IN|eBc~{sz
zESKtE?(<M?SgzJ?D(#2Ih`TO(;IQ+Pp2oxH*DW>hwlQ-SN@1<PcMZ?GrOJbE9=7f_
zi}fv(ix-@>{R_u`+P{#6rM?#{cbrnck1ll%&H8)G>PzX7*Y_aC3$fDpkazp6eX`x^
z!}C(!JdxM*e7DTJn!feMMxW#I8n0BollSv_uS2usH|zb+^geE>e81J7w|ia}o)6l!
z9NzbqTMpw*`0p+>yPQ{i2<O?Ta+qh-d)!iaaMts3^(xn{V!QwTcd7EuOByAw)bIKl
z^}O0R9iA`Rtz0X=VLWP;a=U!5QtJI+xpA~r+I6mQ$H?D2{B%+-)@m=|cOZG~?~Rhv
zZ|}i}-u~o_g;JRJ`sQ_BuD6_8zo*NJZ+Y{XC9l-|s^_hKE~`|(&@aPs7GJGW-&6mN
zZ>{rcmGvp5N4f9!!gg75FFts=v5!p|&{Zpz8_xz`u<6F<Ez+;6-kA08*)*#@i~myf
z;X1Zkxzc;DQhaz`>-!pVvhuWAc`i492*;sUqp(r=U>s-6;RBAGnBbw>b9wByz|*sB
zdtchBJha<ANl(?cQSmdcaq{}^B9!AC)@aO&ms;i2`g_7g_0P&f+NC}`2iICIciru$
zm13iLEq6YZwyQM1ZngIHejigQpQY}tTm5;~y<olS!}CkI<*aon_uN#e9as9EwOl>2
z+Ese57`88$N9`)*L%&L~(l}dg`%3kww0)Mn^|bnG<0&30wWC_kVU_x2tF&+Tb5o`M
z)2i)jy$5WS{<Zp1v-;Qi-lbXn!*g<8WpT7#80HDpmb3DfO6wbKclF|x+*XZq)0<wc
zvd=e45B8UZrG76PpPzWvxrcp}){BQ$Klhi?uh#pxto~W{s=iV8r}ZAE)bBD|{duGB
z|MMO%%l=xep4~SJ-Rp(l1y>peE8XWR%@a$tuT<YgwJ-NRq*DIct$nHYC3)L>qt6wU
zj+gg*v*fi}e5+KSMx9%-_zUYx#i4e8-YC9%$~e`ocTbsrPn}1*?XRt(Zx}DDEm!(}
zq*d<W+f|QNv7buyC^x^Zv|Y7!+phh>_r7PZ(LIw9!$#~jbeYZtIv4EhIAO##BRYG3
z{okRp4;?skp3Yv!j~F`N*X89pJ32eI+-a%dBZgjl(uU)g=xA3SqTXdPu=XnZZNBS-
z=h8-Fnd&vX+01W0Ip*?pN2SfiS9@LB^^d7z(n_)6gv9|P1CfEqKx7~?5E+OJL<S-Q
zk%7oSWFRsS8Hfx-1|kEIfyh8)ATrQG3_Seu)3fY<<!foPu@ohh-v3W8?Mv|%pGO8F
z1Em>QY?h19>3!C-X(_BUPL>`ywulTw1|kEIfyh8)ATkgchz!)9flBk(*uFLHNvuQt
zEhLW6lNjjy;SF!>f55;(h*NKPzw5P)XDC#Q)pE4eHbbVS8@-t3h1Kd&ZofU=T<*w0
zeeO>4!g9xr?IHt_fyh8)ATrPl2IQNK!e-2|mB-90KZ<hxm2WnCt;6?A&DJN59~p=Y
zL<S-QJ&S>#4LosyFP9&iMvOc}J1E}|dR7#Ovxy8u1|kEIf$4>TTMl{ojeoBFRuZv3
zJWMZk6z3TkhzvvqA_I|uau}HWK<`yH9`j@xvAH~~e)pD#|KPX>lf1DU8;ETp1CfEq
zKszw-+iCy!?d-choJNd1RAUG2VB=B8+8J13&nYKeGxM8CUO4nyTl{j{LnbG1&bM#f
zVfX<Pk}Ujga@o|sE%<CwE^c=HJtOZs=E1blxaTtCuV3Pf;Yqt;-rvPt|IF?;yn4tR
zX{+KNZaDj^qZWBRZB@Ks(>0I()tJ}PcEw*kx$NqvANqP4@tZu<9*Lub$UtNuG7uSv
z3`7PZ1Fglttuy}en8%iwt#IW^mrvezyIC`FmBn{_{_$xu6&C3~Y{RYAo~wvM4%lF!
ztG_j0Aq(HS=A1p(*lNB)yJ2gsQq(3g5E+OJv>yXEZMy6|L%*7w)C=d{aJkW2-t<P&
zE_mn0`%L-AT#u#giral~`(775{dyX)IXvuf@QgPua=}Y!#OaHNMn(R^e*fH_hfjVb
zjo34JXp|kraUuhefyh8)ATkgchz#^}270eJZc6`|FG^YsXW4YU<xalttu$hNc<5<%
zec-GIJ)ExmS`zWU^RUb9t1i3EjrXMyBM;FIq925Dpr`E%aXyiO>4||x{oUZa#+jbh
zCeAf75E+OJL<aI_;NvBKF>Hk&j!VnMqt^Pt<2xMtW?BlvzaL(1dF@s9+kDpv&!t&d
z?eE|!{eGpV{JY$x)~QvW=anZ#Ij!~Y!L|B&<R4+9#K=HoATkgchzvvqA_I|u$UtPE
zTn2t|{x9y?`=~3^h>?eA2hk5&Jq~2WwXFNkYVR{+zkL5)_Q+>Ge1`YzrT$&FR=*Ey
z7yYv0POOg%L<Xi`1|I#{$HN}l{*xqP<e?TjSZB}|b3Jn1fI`H`!`F7u(<5rud9B>~
zp3YC4Ut}OM5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)pdJjg>)%n%Tkm>s8)ZfY
zA_I|u$UtNuG7uSv3`7PZ1CfEqKx7~?5E+OJL<S-Qk%7oSWFRsS8HfzDF9W51ml>a*
zxWJdok4>xPp?xhQY91Mg3`7PZ1CfEqKx7~?kS7D--<he`-xtYK=l{Q~tiQ`%ZvFpX
zR{a0SKx7~?5E+OJL<S-Qk%7oSWFRsS8R*q3JSUbam;3K`#dekbduQ#!SM-_k7?}C(
zC&ygA?x?g}>?!{aT%5<HUH_OmCXGBq9-<vo?+23~=)KCuW1dV~9n0GdVyi3$_IPu-
zBM0@lJB=85$ioh@>@3zt1|kEIfyh8)ATkgchzvvqA_I|u$UtNuG7uS<-WjO&_h)(c
zYxX+C_cq^rZ`17cX!iJLuhFf+h+!jk8@f#Af}IO>cAPL`n-QJ8zy9yg*@q4sI!|Y>
z<3|jg@9Xk%ogJMWTkf>f@DW2VK54`8OLXL}|CGOed-&kb$0j`$b0;hcEQ^8ulXpM&
zkH_{eREuQ|(!fSbTzJeGqj#B*R*G3F{(ObSkNxsbcO}(gmfXDSryg|0^C#~)F{u~k
zt#6dqTn6rWd*q3)oHZqBckDgq%KLx5<)=yBnAhrTw9XmZ&O5{BNuw|?y?d&hV|Sjk
z&g%DmkwmPFhedmR{`p>qj!jG94R<a%=&18<Ps_!gYG>2yyh_<ed|sX(g*;8K_7vwG
z8OW1?VLKmu$JFCrPud-`&ZBv9H9g4s<AhCiobB27(?;R+z-cSZaib&Fd+U4KcN8Kv
zhljq`-rDD9gQq1u9b3uf>#0lM7gu_1{kvaCYsGr%7-iR=ft#+swciaLlhS5ktuy@K
zjvxGR)&Y+v?S?zPy}*OZJo{QwE!N6g>>n9uCkD=*@W{(A{(f?rH?|W8QP-Z!z@=N=
z{lG=<zLho`doIIqW|4u&Kx7~?5E+OJL<S-Qk%8th5Pla>uX6Kd7)Oo_L<S-Qk%7oS
z;}|&f%dft-|Ni5XcE!BEf3H?vxhs~}_gCePH@&o*>%x9>UB1-Bbb7_@?%(^<yI+}<
z=7rVTO;5?oFTQAtOSXFB<uos>md{4_J8jyv^X&KOhiS91(epOm{_EcR-Z`g#l(svz
zYaG01!tT8m-eFQ&FSO2i;e|J>wdyK;Ur$S6v)@zZHGaF=RhHg&+&}*vhg>*5ZC9LO
z+2zikdfhu|yJ5ZZght8z+uT>3z08PL(&-iRil40}XTyWKp4@7w*ODx3HGNCzGojZ{
zjvD>(t7*GoR=ke&|2pTETK8I?kC<!Dk*kk*D{WVNXxam}jQMzc+HQF6^?%sqvfsR)
zMl6qqYVEw--|4q{yZd)rZ?S`B9h0^iR=Y0E?svu^C$4qh)vu-X#!Al{rOub(^Vbhv
z?CQT?`F7fF_?LIbEPwt_UQZi^^|rS^toxHWUijX`G-7o;)O&qG`FY)wG)vBQ{g(c6
zgR`zm+Z7x2_oA}K3GE@4BLl6#z~Wy%GV?nV-c4%7R+!6w(xKUZ-}OJ~R8HNdJ8iW1
z5mTQ_nvLb?`k&gg+PR@peeS#Gxbuhq{*9#G_@B6ob^m7I+lOrT;rfHdrR|O<zwz$T
zf8A(g($lco-_6(B@81^kumAgMwJrJ`Hs~)&v+&F<{`};M?@UZ;#c*D+92tlVL<S-Q
zk%7oSuU;!1@!_-A4eMQq*fV*EHXa#>3`7PZ1CfEqK<y0F`n#}Zy>F~_+&BOJ@!k`^
zeN@t}xW=Fz);e<SXVP}X@Vk+AE4TXhS*7&bdH&0jMK5|ZZCA|td;HC+uQVQ&`g_iY
z?f1#3WbrgDh3jr{@F{=VdrDF%KK|uu=S=$Wjig+hY4JT~UF*2L({i!*@9!MF=QA%Q
z5hD-P*umzXkG%E0H(yC2PG3CKp3#;kq0)HW?DmcNeOXT#hZSUfZ&zvkXPe)<;FV)0
zBs~S&?cX7-^zWls|3+b>&tZ+4SE+ICeB#o5ce>&qNyNtR(5UlTrT?Dy^wz%F`9Q7X
zx7&S7sBf+F*S}PJi_hDGfkwr}dW}<QKHBK^?O_E`%g8`vATkgchzvvqA_I|u$UtNu
zG7uSv3`7PZ1OH}V!dx5AaofCaq~+o!H*`$+%hBV~QdsTZZ#sSNCr0c%-vep0aj$u1
zUu=~>Je#&Vp1j@nzkkdum!<87<B#5W#gzwlCGCzYU-m}tmw$Ro(ymzQ`^k2BzG>Ir
zPnm7rr;<|m@5mVcFEY^c7-;u+I`(((zwp!5dlzcOMy>2ho9}YY;60yA+8wXk`P3bU
z9s60*)9~th`#jk9{7Ffx;p*3ZZ<%X8`7-G#`1$XKOqe#XqtNa+!|-3sJ8;OmNvmO_
z;#sSVGwPt3Z{6&VA0;8Sihk3Z9^-#`>M4(2IyQ+Id5Crp{a|{H16gr@g&|jc{PdG!
z)4cG1{`%AxD}DT4QZ7C;X3^`{z5co+V&tKoc93QNu|6`;iVWm^j%)TD8~yu}X3w`#
z<3IAquU9(!{2P;Y!$#Xf-s82)`=u;B>-9T{M$656T^l8@(Z5HXUdF36?zZaRsW-0+
zI_{t8)z3$#)#6e2?sNIQFHBCFg|+@YvsUR}>wCUt_21|4A?x-(eU?JC$POb;-wfRH
zt3`U<cJ*^<#OCs_&yDBaa?#N*r4g&+Vfyys{AWM)rF*8WeBa|~qcQ)Pd{g2kC;k1a
zZH7-udkXHd>`z`hX7xAIyzrLc&t0-W;f(}taq-Q||8V>>X}e>o?=s7M{!MFY|5B^}
zUEV{#{OWuA?>|22>DcJ|mTE2cm{azdXTav8)4Z`<eX{Ps+O0mX_X6eemeuZ|X%E~o
z=Hu~c-uT%vS1&yBpbyh}<88~HvGjmjUQg?Vt@itzTK|36u>UTD=Q;Y#2cJsv!Yn(h
zSN+-#j(PO`MgN)BiuKa3-g4jCebhU9&H6%`H|C}H2^%iA+e@F!SZFu=@tnKebi|V%
zCGCp0cV4yRm?ghT^2YG*!&O>tbe@yfcs=F4%Ul=so9ptWCZ_F%&HDES@`{7y%ISOU
zt$ltrcv{l#c>nP27TM}IUnT8^{U`5!?jMisU#Jz!UGG}k<rUu=EvMc8zIUVketWI<
z5&KVH4AgoKD%HQ$?+eS-FU%+EWhL$M_y6lP|4?qD_nExLE7z{_Y8Ud>D&;Z9z0kSH
zkY|%>al8BX{`BrwCZ+9)f1I$%j<Y@ce%fx>Zs)yF-@N>|S#nD4tL^f<()%9m!}!&x
zb77;;(cwHBRSw^O)qC7(zpt*hUU|!(^4D(<AN={)q+PLD@ost_ciiH;y?)w~FD31Y
z;rXcD%I*H|WR>DO?>JX)dHx-sM&YvqSG;e^HdE4e#d6O_d9PcTM^svFmpnF0PkiPb
zx5=#(+x?zA)W6;1+Vq~+WX-=)yKYwdQudeiyjl9jacg6s(f@v1-s2tihv9F&{DTjY
za<N>tYJGoM?l`esWS|}lxL1t0)~c)YeLamlL>}_AgL?REJ!R(g`?pq+Q%|lck)8KB
zxL)#l%J;c>&7<5kuJ%24qwevHQ|^2!ZP)I<D?E1RN$aeB?-xnEuu?u_`)@Mvy`ke?
z{A%P9g}m{bqtw4NX`N4BUS!sZNyO=shx%Js&lusb{XQ9$ES{zjBM;FIq94RKP+1)4
z8N0yrJBzIMgR%bdWYLQrO&i5SyZr7XOV3JwUr@^9%6)$p+g0|yG4fR#U#09eK93AU
z2I|K^rQbKk_Dc>r>b%?2$V0w7)X&1B<aTBtulIqa<g_zKQTO&^ApCuAEVtI*<*N3(
zfa$$o`*~v2I5H3!2pKqgjqaKe!$#~jbeYbDIv4KjIAO##BRYG3{okRp4;?skp3Yv!
zj~F`N*X89pJ32eI+-a%dBZgjl(uU)g=;-yMmnJ>)?ZU^&p$pA_^UE{8lWhHR|6T9g
z{mo+enqvlR`_e2=BuBpa+a>q@^X=)G6W?BY%2M~I+CO!~ZA(0T#EZ$LI}RE!?WXsN
z{r8)_aL;n<+?dQT{NDaQ+_x+F$yTe~e&$oJCudIGcaM|bdM)|+)$i>%;G*}7NIqQV
zqK{`i|Bf*Uykos1ZoB&Qu?bvhk9%j|ck4a{IQWYjuAlwH50Vf+d-sabiybp5DTQNy
zbNWn=tvx<D;@$Jl*!jbb0$xISb6>LXGMDszHaUH#?=RZ-i&u;D!^f+0&(N{-CzFz+
zmcFC+EZ_bxnfrl_u32c-NkznGi25(F;%$RATz*PX-x;?1)^?j*JSADL@AlKyyXE8L
z=mq;Ox8wLv6Z&m_@R2{9d&!AOXc%F6wZDCDrLo_CD}{ebu6*M9(W4XV2X7s9*0uM|
zd1Z>{(d%wF*HTMQOwU=f^Nz)Ld@0@GlU<MM_xU?TWDoGcAHFbp^3rdo<g4HGoae7v
z@UBynpTGF%oojwJDINC0t*hSn?Ca?)6Ar(7!66@}?8Wm}zrFGamv+3CLV8~~|HH?Q
zI(%Gl9!0(Sj=Sa$cRrMa;~Cey%#(h`E#k-D9@^`s=hGv5o&B5BR~}Wg*RVYA9j89^
z^E*FJX1(*0o!)<SLbB(U3$HZCwI3$Q=NB$@+8$Gjeqdd#8(!HJIrJol-NPNa=cgU}
zS>?jlkDmYc6N|i>r}@2i*kMO(KKCa@Uht_M|GfV@KU-nYz8@v*@8%1h8}h^c9R>Or
zhrQ89z5eEf7xO*-v<+vu`PbtTXk2kXJN?+VxZ*j!=p|l3@ngYX&3ErzyFHcC(|oNX
zUd5x&?|S&)kCu5Q!K-%W<2hdSo3P6)qmJ0&ss9rPzI*F(^SnCKv&HerO;6ix@1LJ|
zdtx0|8oKY{Oa1u9VmvMO>-Mw1oHXg5f1dDKI?HtnueQU&Po(_u*ue*lTIRC%(>H!G
z%PZ>~JTfI8;ujxqtKn-ex5}v>q%ibn{<L4W9~g)K@h^O4p*Sh-TOV=OIG3E<f9Q&{
zPf9;H=bb_GKK^bo9vIL3@F2d7dbrOjBhMO}nit+e#6yVIKh*!Xr>=a|YCo8qKz0O;
z2iXTK#*yy*kNrY+1dRjPRTkcI=3={EdCQCiD4yZPdWAT{uag(={qq|sq(3{PH@>aE
zcKlEL4UxZw`z*O3@^kS+oXo;ue|h1F<No)nBn$Cd3i(GChWlnLw-*EKm0gS1{GT22
zFL8j~zO%(#Z*8?uufnk#O_}!hJw_JyaV(s=?d-E1KWqPj{Z78eZuMhdgO?aG_#aPP
zk%(X79Dfy8>}TZ37sz>SrtdE^?8+~aFus|O_yfh4EW}6NC?4_aYSDg?h1N~H+ULgS
z$3F9$uK)KB6aEXWJ7gEo`onM>`y7-HK<iQp@hcC>LV7{=n}zg(VLN&@8tv=s-F{<T
zAiLdr=&8^DbMH4(`^od$-*EQ_uZ&KYe`eGktFQ7(DnH^!#dAY<J|v%I2lAt}mOp*Y
z_1^fXcut^){oej|-Mp9l{PDq~Qh6yq<j;83-uybwzIV<!Pk6C-9uDXG|FL)9@pe>I
z|M-*8LT@4}y$T|VB1I7AidgVI2u9%%1cA^3f+9t#h#(ylG1vf65R@WS5U|V@g3<&O
zkbn?@P(lqM0TM#_{dhfjUtjL8<C$~L%$%8f_doe$&+NVS+Mm7F+H3EbGnfASPd=o7
z<8t5nsej0yAbntHw_zF2kp5Z7{-F4jg~kK<TZr`Mx6rtu>(F)akUv580QvJE&N;8(
zC-f3W@tB47-68J(>r*dW@75WE`+EBrd5wHYzI(}SpWX4Em7W~jPue#)M}YKZ7rQUt
zciSUl9vIxGlLzlzaJMU7y7`#|ntx?TAM)@E@hP6!r?}4l%riduAH*;FVmIWltNmVD
z>ehAcU!d<dw<dRudHv<&!tIti?C?`xP5Qrd)8V`PZPws9Psl5MjfcME@LO>fpZr&S
z@q%9Xo_Xc5yB_etx%=p?KlQEK?1vumZ9JF<a^(f$!sgp=e8Q{=rwqn7_8~vG_VG)O
zSpJ%4(^uzP>ZoTno|10z&-Wbi#|tMW#>HOPQC8k-KH9}grycUxXWso%YJbMR>=(sr
zJcNiR<ALNTLvkVhv(S9Nu>W?;>>vxp>+9B;ed5x8`Sakpl=&3@>HFa|4?E_%^pRAV
zKGq?~kCgG@T#Ubj*ZJ49B_5ix!e!5={L*!Ht9=-M*%5sE+P7XcVT&gQ^Ih%vQ)tiZ
zmtBdU`XSeS^4y2`#)EP48|Yjrv~&4x7Ru|Pbr4!#pnWxz=i7HWe}ecmUi!i?|HUU1
z$D!+aF_e#27~bd4ZyfU3-9Pr18H498uD@f<w2eRV`5EcWH!uC$Eq?RjAbxSrpYHm_
zlU`2Y8rSZ6?d}K8Ovk_dfkn6Z*-PoMSByJtoo~G`crMJ3pMUn+vFDyQIoWjY2Sz{s
z;!}xnI!Cmw;?q1^@5O)T=-NT^qaHi3KC^52)aW_q>B`pYETp&iO@HV6{K7h*Kk=FU
z%fs2Zcpx5T#WB~#5q{2o`H%HY{;ZvN;Ji#+QeQkH&pQ9U8}Hio;I~Xj=xsiY*L8dt
zC;i-KfBc_cunT@^U6v=CpAhv!Uv?g%_41F$pTGJ&w~pvDKk@<h%>%y7oBe>eVBVZ(
zvJ2zluVH+`d+2ZCySxGNFK8e0`_Df+_MpvQNy)L#F)!9*>#pZ0%1}MX4&eNs8hgm;
zpME2e#~3$sF3TS22dx88zYt%okUs30pRnf;>8oFMXq}SJLVjc3*@1Q5Is~on+KDsl
zO8=p~DjUCXv0v*4#IO413E%8Sf7TuRkT0JwZ^q%i>)MeA_17-yHw($nLVl2i_{c)M
zL3V&|<A?Yx3!QI>H^wDzGY<B}9`U}~Csw~_;YXiNjVI(yee)Y4yVegLT-Oi(#REHs
zp<L&okYBN5aXLixTcNxH(oejTpB(qg%YXBYb!VsIU0w`%G;VQMeq+Am`Qo`aBVN2P
zdd)??vg545d0OU^8|{4Re6OYS)}MCrCF7&FviR;gG=6%i$B*&BZp<(HVK4Mx58{aV
z;y2oxSM6O7(KzX&zU%BzIYjHHJWzY_p4~$8W?sZKc}9q>?5?W&pP6~!LdP%nYWmUd
zj{45n5i?Ws={z#zb<D>9v+ZM7j~qDX$SqGf><|BM9`N`NUcc&x(_Tw|JND<BtnuS%
zgU|WoYoWjB&-|Mg=steggScrw<eWf$!XCqT!LFg_o$@PrBKu>X@<Q{-PU%fA=gRU<
z_R3DgVdIm38mGL9J*ualEM({5c=;*6#lP{pFV5)K`fvTjKRqFTR!=|9rRDK>!wb8@
z8-C@l{7s&0zKtJG^uV)r@^Jm&QQ5fVb@DUgF@E#L@9EFK^%ssCUtxXoq#wMILofSa
zc1#aQj`6t8ZXw>o{E%LFvhTG1<JZ25|C%3h5$~bi)*JWvk$L1l`nS)ar}5xjo=zY6
z0N&V-_Uen{#uw(x>eGXK*Y!hR{whv{yx^PO?9=(FdB+2P<1hHPKZfe_Z|$r%<nm*2
zoxihB_GEtO$zH{E@j?Eu)u^?f*zn8K(?dS;;f=O?>&xl-D^6Vh<kMaqJa@5wcK+@B
z!Fh$e)4tX?_0K=$3+#^{7&p5&{?Ja0!#K5v#`WmWt~zz|O<x?$*T~0{`t-uDGWq6L
zzx3h{?&I6IAimkJ_(spmp16L2pZ(+YR9+`unD<a$;}D<pOCS2-P5gl3F=YSz1Bx$M
zs2+cSu8Tk7JOAW={EGkbH+H5h4tZ`9%H!|WdGhs_g{|^0<H$n17)KW38UG=&i*7NL
ztIYqPcwH2lkE)O#vUkXD?VpUF9oj#MOZ?mXTED~v`7L?!UU5Rd&M%GM`LK1wb^W~W
zd+++i9-nw689nEGm>&FA{Db^IFVYWMKU*O`5y#o1^KpI!jYm9#+F763zx(oM>n+~p
zk<N+bXL!)w{=b#~vXhXvu$?>rKjJntFXYin{I+hhKXKf==0(p}@q!O=mHpC_eT#el
zkAKg3ubiLLlN>xl`3br9ht?T>fDd^KJHRWy;UDB#kLYhcUAK-wcH}<(<ypql3a!r&
zPa)D1dY`>rly`W(g+Kk{2U^de>+%BQ$9vt;J{|I3aaUY|r~cxswO(8Fsljsue#GAR
zuln}a{A}Te|1^28r)H-5gXY0G8-M1v{L(zJFQ{MhCNEQVuEw6#7a#P;|HMId-7WIB
zEX>MBTyNz!;!YN_%PeH~S%}A0D8FkLv-A$H^Y>OLF1L#z&-_;$lW(v`=k>0$Pi4QS
zVAt%O-LWfqz4cX`eqg7~UcGah5q;tx|AgW;yC%mx+VAPd{IhTT7&l(&!Tz+f{>szj
z)$$ShQu=rfPo6jc@oGHyme<Hj<Y)G)<Qtzn#kmo5UmoH;H|rt!)>+SQ?PHABe1!R;
z_p`i*;Q7COk@pSY1-st1_1X_lAN<{;dBLN8*u6Z%{g8M1Sogdy@%D3XPcOdYt$p@o
zo;S)L^y|5`^B(o_f+u!?KhNXYnR6(7vBMD65A997)(?Ak9a?X!zwE_za;-<+mlL=6
zoqUb`>lf0K-COs(Z@>FTcOSFk4zm(`JNITc@;vqJoB4xt8TtB;CVX)BFB~?q5AUA0
zhw|{No_S<9>`vZqK0o=PogUlc{!xAO@OwmdqaXSx@A8({uiIs}QGMEzN6sAw9Ju&r
zzdNGO`29`-;+LO@6YSRS8qS>lv)8uSf9BvmYxJD`){XDKa;t-vc{SPYZyUdI{M6S5
z-v@o}s}s)s(GCCq`v`M>SL=BJ-n}nsJkWi==QTdRYoxcb-w{LC@$vMg+b{H!TSxb8
zbM9SZPkDLXzE8dXf(3qa!h(HoIclFB)BP73ysqAzd;MYkbNc4*gYVsL!S#Oo`htBA
z{$t9w558pXKKhWq()1tv=Jbc>>(ej0U_Zw3)x*BJ;R#30+vh!1^Q#?nE@R%bXQ%pQ
zkL-Xv_RfykpLk{6XQv^uTlQld+KX@Omp|Hn@OSZnfAW8ELOstDp!jD!a8BTSHbm=;
z^~*j>T%#ws_;BAk1g&%8Exz1$J$yggxrBAf`UvH%<j|Ww{7`*-;uoK;lcQ|?5jVvf
zyvReuP50Huk2ruI^64ipF<<m@-FWC@JjP+2Gr#2OkDWOOl2<_T%!7PGJM)5H<F?;n
zckI%71hq3?t{WfT@MyiVKayYBS6P?*4w#?wJN~3!`r=Ff><iz<OAqz%X?*l1-+bab
zL_Ar)2lbfqyIAd^_UugF4ebx@1L=uh`m-1IqhI{EE<WLjf6~YN+CLfxKJ-hj_#nQr
zFL{)BrM>a!2hvB`I&Yqg$NDTEBOed=gyQfo*80`tBY!wg-@MDb^6-zg8Qq6R`M+_o
zEB)ffx?-H_(}#XPzV4`#E_~MlgX<mnA<|1+*B;tW(Z@LyIrxI?+dQfV@nPP~2YX=W
z?4P~6ZeBvf1Nqt+2fNlEJERA_*)Lv{2j_Xt_f)hOH}pqu`WP4ftV`PA9kK`Ufn4K<
z_|=bbuov~{59wolpqK0B$@q;^e|XXkFY**|)%rj`yg>F0`LBM>7x|E$?$f6g+JC8M
z|0z#)-FWe(oqeZq8z;W`pYgFnamf6LN7@;;dV~GV`Ob~}z`8?@`EWkXAI!Tl`_bNY
z`4wb;_QCXLhw>rvwR0UW<g+9B2)^*G?0LCy(%0`2?3<h~n16PI54_uVsK-w5#U7oz
z8YlbGA9>`<N60l^_G%w&|La@|ANu$Fnq2Dz6d%}w_0~Mw&&l`jNMHTq$9VLgbuJpO
z*MkA;y1babm5uD!`Cf?DP4-D2@ryn4KlQ{F`GI&Muau{{&Q8T){wpr<U)MwA5A+k4
z#dqfu?%Nkzznn`rCm;_$A(BgPJV5b9J^pK*H_y&d<SW|YfqeSm!8&3*=1G11kSk6Y
zm-?Q|h&ST2`GxiHlBh*d46qC9uXV?I>%M(^n4dTgWQY8XeL-=7AFyL~Cf`zTo}X^L
z<qNOR)fb}s{GLBs=bRtPKOukQXYcvv?&mD|i~0IKbnAMje0T5p`kc3s=X_3o&IiQ>
z<CQnackIu|6<?g|Iaeg#d9JeW+c=+eU%tVA#VfoyKg7R!;wpaBr<ZYio`SEFC+>CF
zgIA3le6B%mh^_3<zTJK&TnFs$<umL~9MaBv`sBIqxu5(iTyH!#v>u6z%KDXGv3K!M
zTxCD@`QF!5-#Hk$^z$C7af|EnLiL?5Id9Uh=PKfT7_W_6K5IR(zM6mX=q*olA9~&-
zuQv|)EV+1B&v}ac);xJ{_FXR=z1EU{n0Fw=(0<q-<d40=P(NkkmY47g>%08Pee)-7
z@K56uNBH}KANs<k$9;E+KI<+&6&J+0B~M;;qkSjL)hF(-Q*n!*i?jNdXFA_%h4h8y
zIee~s=my8_c)^Wt41RY({t;rB_wfgD9dGtU)(t3*LE|$H*X`StA-^RbZ^kE1xUSp^
z$>neA>xcgM5Z|EogY|>&=U4~WMW`S7_)yP!VgG8J;*s}u_^r5Op1gl$ey!`y2Sem1
z`WY%{T<l7Ijt|c<!tvr++!gon%l=&#FUW`F@DFwt+70`%UMPqCSkFQ^^k+xn8~@5e
zafiLhJLEr}cglCzFS{1!oL_iuD~`$E*frkqA`i1qBiH&C+Q0t!O?V$)+8d92%Q!tJ
z7tbLc<-7b^9%vj;{sOJ5e&3=juW%l(%s;*N<M*iEn}nXfdM+FKum0s{`jr>4;}AnR
z%Hp8-qKtR>CqCJw@u_bf#7q7I<x#h$J3Vyf?yskN-?`)?NB?s6;CFh@xnZ^YC;sn<
z0qdaO`4}fZ2@${aaG!tClYaaWvLndu__6Vj$FA8A|8!sfc+%cD%#(bL9?qZH8+{=;
z@>OMaYriKS6Bqay|1po^g#4JF(gz>Vc#N0+#>L)TSKs`LpZF0+<P~@i<14+5(>yDy
zAI=wkAiENWw1dv!<ss%7>W^RG+kNpE-||Lf_Cim3gyUDwI3Ydt2l1GN#$z7wV7}cq
zuljM{c-5nSXz#|se#q0#{)K(JZr+XCd@CD2`PL6{UK!%ue#^Yz3+k7ANH2(o@V<G)
zkA95Led~?up`P^Rm&(>_^0nvJA)0@3+$YC6Ek0N;!~V1jQGIBil7;q{)+_NqT(@q8
z`R&Ln_nSCvsksJRSC1cCC$mtz7iYD%zKNf%E89=gS9@{IIh%g1(_uYw#BKeuBhT^p
z2Y(Vb?dQpHo&8(ytl#cuwKE=Z*gPonV{-VvcA*~p!hDD;+Cy<3zvSYRzV2)1x-vQX
zF>mrH?X5%Jn-V{u{>6FmT>j?0bn`1e^FBM?{eDAPJKv!gvG18DjX(UgRQ`6{)5p#I
z;B6xYl;wf!)b}Cyh2P)O>#|E{OxR|*5d-vMcjWoqv-<3jeJIlxf6D9=;uHTNs%QQ~
z`%tEjvUxR+`k{yUqc8dFg*<ZTp{zX~mGNvG>hlBgmFY{qyoJ8nhjRIg`_@5mz_`pm
zdHRFmLc7S1v(Wz7xai9s@P&8xt$W5{T=EF@@u4igbZ(xt-dfkh6+A$G3&lZb-k|e6
za-nrpKk_f_+-INSHNA~jd+VUzGsr8*wa$B=Ond7syVK4(U>#S_d7^cleCsYA=q2Cy
z_U4~o_wL^<(ihs1``(xK9YFaMJ&m7y`+N84OP+HO`WY9bzkI~JWTEF-@<E6nNWOe4
z3)y3c^a|0sk%j!pa{=p$^Eq*Vzv0V${^q#~KV#?O6+dK`>{~tSqxIbRuee1nJNMj$
ze$Eg1FTK2HtG@Sdtyk75>(C#*yU;zWoHlyE^EKaz{l*@BtIYSwdHX!Ka*pKO={rXq
zbl#Ekyf*mvyJkP~-0ZKsZ}#B#uARGw_53b^A2_%1dokxauKT?TJ9gdqJbj$wL;85H
zm|Xq%9f30W>O<G<6YNj;xj4grJn!}V7xFW9z}~&D#ZJt-`;b3}{nE#CH}Q-8i;v<&
z=tuMt58PL_pYVGX`w`zehx{3m>--iv=M{gA55L-rBgPHA|Au$#CV$ee`sDF*aZA7a
z8~@^hIA#9G;lFs+pYhO7o+^&0FV7{<edDE<ew^=`2lHuut=q;cUf_rR<mwm73-Dk(
z>`;H~PJi_F-lDunzTsTVzDYc^u7!3_fA1;kU*61q@IW5D&8zV^=V8yrD_+X$#9L@R
zw%?Y|7#|+#fp77~ef@{|20QmV4)f%FNOogg!!teDC%gB&njZRhZfo6A&-~$ueDlPQ
zJcqPCkz;-H+}nCX4xW_Bu@1?f>~oaq>$y9B4CT0Q9b^Z_%}&@8Ir?!vflp|h^jF_H
z%0A2wzVYfhyM_GEJX;@>!+9`X=P>kh9eQ8DJ`A7i!#IsgKjM>of&Z%KceDIh|E`ly
zfBtM9=#58t4SU6__#lrm|Ky5)?2es@L)JUvBxkUF@LiJ7-`Kl#!T8uscwar^pr3K_
zD`=f3PyAsQ{6~FuqMdm$5A03c7pLX*@@)Bl>*5)^bDoQTX#Jrl`FJ*Ner8?36Mprl
ze|kBG^E(OsupjO4MLxTQ;ypgdF%ROV{6#zycj@oC`7mDf+^3&<c%Y~GA(tO$Z$GAN
z9=z{jT<nQ_yc?(M>Weq(k&9>Z$=|(?g;#lqe)$ji<SH9KeXMW%kz9IM-#w?okABqS
zudcHXW%u!Hp0tPjf}W7Qh~MOzKl3Aw!!RCMN6asK!z(+)hjTCa27TyjT@=6U$LQs{
z^F#5OoD*LkeaMx&E!HRghyCf-`sKd;pYuoYojtHKhzH|0PUE*8*pITy@Vv~rCH`B-
z$&pt$hovt)jT2Ayv)*Ggzw|Lqd7k#+`7>U{;SkBCKl?Wh_p{`%7xl?kpFH)<M_x2<
z;)}S#@A<d&i$A+=zSQUU=GnYk$02?7j|co)xAp6~b=Gz39(nwgT=AD4?z0>7q71cX
zS9m8!9Kf4-VE5$Uhu-2Dzhd|N-n`*Myi&H#@ndm~Ug{YqJ8_?0^rb(&wTJQu<IrC#
z6c^bE`L2t%_%KfBJVZWX9<3ANoN<zCUg)cy_UxDb>{UMOxemLqPr#%3vCgv(*VUIF
ziywGrN6z0}5BsxjkteTX=j4d5=F_;@hcaHpMZD32zS`M0)8GC;Ugms5{>R_R7cb41
z`sSTJ;xRNXc5Yv8e(8&Ea`0um{Ffc;-|tN2ckGNj{o=tq8@K+97ccl@51t>|xAJ#!
z4L{ad@lBj3hdrv#9_S0nA)mdnYyIPc9PP-ruW+Ayaf*KIjvcwKKlV%?d698hM?F_H
zul8x==|{dJj%J~L_&Hv+mv4z{;=*8jn)7$T)w3S5D{({I#IrcZZtz0B>()Q=>8XG5
zkbHJ5ek+r!A9<m&@riHbyH4LwuIt9DzB0W-#EX8EwWpUnUO(~%<&Y2dLQnn|)_0$u
zLh%G&c(pD-^_{bdH};?6ONjQ1^r0Vn4&$hNl-<ZfH`wKpqn})B)Brz~|H|LRVgAWa
z_|KXL{qu)k+I`f3c;h_=?f7|!;*_}UKK~RK_@nj8b@ie7;-Bn5JRp}}^H1_Z)Xw_G
zzVI!6<K29kKl)lXA-V3G5BiIH#x2f=sJ?NsL*-Ck_RSyom2(yAC1iil^}%t@iC4y<
zfA_7o>|XrGr~Em*&Tk;z@S+UKH{S3%`FIZHXony3C(kz?dJoE<^ZgwDs~%pRi?bKJ
z<4IY4NDt@1{9b+in`iv92jhk08J9A<^}e-y$2o@i#;@@kFM0M4)@}OGkKTA74;o)+
zXX3Sdhko?3UTLpi`J8pv`iBqWCP#dR^oI0=p&y!0$j;!hCoOi#vp*d*0Lg*k7Ze{L
zJ)k%S=?&R8WZ#gUkUc?iS|K}<f6Bj{Yl#2iDt*-t<1l&7WwX$G5Lqa{HoxqO9f{lQ
z(slL}+O_At)?0ZwIqjnN5j@|sKOl$Rt{ad1Anae9b*^c=%AWW6J)-A&-Yei&pa0Ra
z$3JlXqJ5qR^FMl!M?Sf}vq*o}y?3L%@j`LKc$KwNk6h?|0O)-PJn(P(DS05Yu82qM
z+`PMPe+k7sDBfEqtfTs+2m2DIZa89U|Eh=0-{-qr8*la9SDt*!Tz%Ki)qm~btIX5q
z{k6Qvo*+I#bRT*@Bt&-J3h^JJafiq*p!GtYV}0R=p3C6Rc?=%$tiG~&uzwO)LV4ng
zd8aRbuwQ4t?9cprUO^vdy<yk-(T*PMj=kBB*he~dah)7;$-|F$rhjq6ee-D?(7H%}
z^8l@@=84|of%}j>SYPOcXMSq^8^k&JlzEU3d2h#iSK8B8oZ{E!Pnmq<B%l9!9}92h
zLw)P8b|I1jy?^<kkKez-2PZ7pm-U{P@Bi`pZ#?qoRhOK-Kp*sd3GZWy*WPRM{a$g?
zeBuY<$NcaU`$RmNC++Y6`GL3$*&%({MQDHOi9h(FpZeavQ=h$=AM2p|=G%M?%DM0A
zb*{MQ;S@jCP4*=3wZ8j1=J*iT@s)-0CMfTN><}6UWMB5<;-d8mPlNOG^3F%j^U71H
z{S$knw|JL@_VM=X^rQzp`MZ6EGCNXVS$+!5M~L!J^27=Cl<DERe&mgi{&+JFp&r_c
zYxIZY(2E?X9^Uwma}4tJYn<lII_3O|U7IKC0=ux@$+!5Sc&6-}O_{y0SNiBjzx;t;
zyYBhB=it^keq?{BY<;mlkqh~wb%Fo#dwS8^x{FWqibuQ}m-%#l<y?wC*;m=Ov43Uf
zPkzT?e%<$8kbdZ4zvH}1efDGjWgSy?p3Q%)_r}BT&4)Zj-lPo8zql<P$Q#6I_r*=}
z>Blbh?|jU;FWy`y-#LZxh39Yh;77{lTRZ+`eh1^gjdyK(@LMJ%;+yq_-jF=VF7crr
z9`r*m_C^oCqjFz-V2}2(>;iB4<tOai`iy_;ka6;B@<P9mKe0b?5%S|Kq!$#|#TV~2
z4C*r{A7hv5@4e(x?|JwmBL~DA`rsEY{7!q<*^}#@(~xI8_pWo&hWp<=Yw+(Lsc$~a
z7k&Ama|i1O{n)#4@K=1O7up9qf##_dihuls9U7<c(4Sq=k9^~R?2z9m(;x46WY6q}
z|7oY5Jc^yFPd~il5pQ_Wk8zj>?dfOy{2AZ+Z-s+?Gw1J6I4_XDS+6{og7$^2kbm0u
z)7v^>9*jdh>l?%$z1Rc)wB9=>w(eQ??bocw_;O!du@2i8l4spjkDnW#c{Y#qvR+tk
zl*v(_KE|oNcJwk180rzqQPvOpkk1$={UQD27wYjlcICSI^u(94`GVTvk6z*)zi^%Z
z@(*_7KD*V9pNI?W&UN+VhxEXU@v%$y!}Soq>X|?C^>19(Av~HF^N4?C{L#lc$S&nA
z{1PA9Ti=~S8XtY+J5aoFzC6hDoPCmf#kwj^8K-s6x}?lM&7(L@5B_TZh;Qc!#z(IF
z34i7#>l|A>^G|Q{zz*<7p8Lw|mOarEhV#$f*$02(zwSfx70#bHXdLQ|p7VEx@JEg~
z7h-shDSn4}rhc58v0HX+KM##hJLr5OMEtdj_$><g1H_+sH($=nT&K5vhxuYB>|A~z
zUz1OGt`p{K@&fr3JNKS|`QoSiN*v%H=8J!659KNJ(=WfyLhmt&!}f>rM{@Z+yYjw_
zGWqsz_DjZRJoGnC`y=lM@;i1&j{8u2GtcB{$L~GYQ67x%bACsHFL@>T>_A*LKI5eq
zf0E~jYsSlu%pd*u9iHW@^l_ct`P~{j72n(!|DgRhInX{#d}qJ<XE*wR<g#z;i1G0U
zd98NN8Q2lM_0K-|nRreQ_3&@KVt>k^T<z(_zQcOXsXT|Jzp`<%596XQ`S>$V{o6<A
zM?Gl1$)g{;CJ(A_Jmv*&?A+f`W3S>IKatN_|Kxk@pB(MjuXx0M=_CHx2RLsM@4Rm%
z&iSs8_o<-!5WmXoUH``Ky7MV^&Q8n&Kd|1A!{41l@>Bk8-o+h!us^(tZ{j8S^iUS}
z%%k}gSBwYW-q&!SUl@<|*>(N9ZoKj|<KS1?sjoddFn)T3dKovq=nwIQ2mP~i^CO;^
z4?NL>ojOmYmpCG?#;ftD57nnP{vrQ`@=)&wK>Ka^obf>V(u<!PCw{{E)+zpJ-=LoP
zRAzs8Hhvh!d49`(#5ccBavkD9nO~B_@5MQC@R^13g)9`m@I}sGyqmL+F;3^j%IuXt
z)8Ba1ClAtxe~V}AH00BD{sj36xsW}vo6&Rfe$O#|4}*QV?|G^HvF|t7FUVWvQS4tH
z?0iHX>^wp{Jcr2M_&0y#r`8Gf4YlL%;xfDCfBF+2`L+AvlX+uz>_fb=Ub}DI@P4m(
zW53EFI<N6v6YHt(Ht>J@*$~YS4EJfiXQCe%*2`NqPWH{8JqP5E;tf8n2l&t)|MmlZ
zzh|9xp1?oEEAy>?d6egXP(SuX=8Ij)gXCkb^DBOBeCA*LV}I65<7B_=P@KSvdhCo}
z>qq{pf8)oWynApw!8r@LkeqN_%GOKr=xP4M7k<Le#5HB}WBic6DciS*mw0ktKX}C-
zeX`o=hu<5Act~IIojm<P{h23r>OQ&J<41jZ&{KQ%qRjr}gM;=r=kFTgL)_B8>*PUl
zAU`2DyiO1EK|VAuP<`{Q9)B=?`e~<(clEV*UwiG;Gaq<&9ly%jsb{|37su%V$ycAh
z@e6(_4;EKKf6>nRnT6H?>xeie4!AFl=|`M$U%YjnUi1{N*$+Fl4wGjd)Z;JWhy4|M
zcb$BQM}CW6`%rO9JTeY(jXw0pPgr){xXIC88M>~2?dXL!d@8$7KjRZ$#5;Mj^BL=G
zn3p?WaSjW$mnT_|y%(yTJWc-QyZ!Pw=cV!i?{SfDUVWcTfA$UPu?P8_b7%RS_{NUx
z^YtfBp|^g-BYB*C2cGnU7vpe_`qdGCIq%_jjq0O^^-COu&KJb15ZN&_pOD-T@e;~o
z7a_`D*q`U)?9e)o5B}nL4m*YNm@H)P<eDeXag^C-i0+5zJ;*HNk6FmR>~Em`gYmNm
z*R5N4#SawM&Aa%;&ditZUFwJ3x-Wi-GthXH*(X0Ihg|yl&Zc>EpPtI<tH*!j{dmw$
zJ@2DhH^m`x@rXC`i5GH|wdbeof}fc`&xQVW?t{Pl^a(E~=0QLFNj_qH=EZeromP)t
z?(5%udeX<adx+wX@2Z(6{TVmDmFela{z44LOFwp^AMr$cagM!NUs|E|A>ZB$@cm8e
zmj3Z5uQEUEiQEv`3;VMU7>{wNXWrDa9+^Mw*r#>Ib@QXWa|S4mii67HoxE?5|2gkX
z$wTO=ti60od=~G;1?PhN46pJ!JbK?no`g4c0-ci>H?-cFC;WIW<Q!4{F7BBJda?(+
z<Ij7f>_L3-9Lc$!b&WshhhNYa8khQbYlZ9}3qyaFSHo8Fv+m=;e2SC!7x%>Z{eOMx
zh3nlqBMEsR-+Kq_UwiRMUgW-c6)%k6d9nF--o?Jmk9zo4#+P$G<6;NCi|KuP`ib}K
z)jmX?Xg_5BLw(piUX4%x?2mlcjaz%?h1OLlF5pZ5{D2(&^9OOnIKBTx4m+~G8>jeU
zyyhL>_R0Lnc+5L~>4AUxnJ44Yo<GsoeCtQtQRb)Qh%5LpPW-S}agg8IhlzXE2l=IQ
zHTfFEn|+`-YM+Eh_T;|$)<1S3UpGGf$sg$DI{OUgNf|Hra2=20p6mR`JZYyM|FIvD
z2eDu6#NQC{ZXVbl`%*S8>%6=Ge|R-t<^j5IAM3hvOZh=4&%CfJdMKM0c18{~zR-@v
zKYqZz`3t%HHRO$aJgLtQ#YyK;^o8Vl|4qE64}TP=#KW+jctk&bC2tbn?a!?rS%{Y`
zBoE?Cynx1sSJ&AYd*|=wmEQcsytxjwXK%{p!@P@+c%=`$m7P1tpYRR&qqxA{jLZD%
zkDk!HK=L5Hp?>Jy3fY<d#R)t@a>Mly&&Fk)P+xm!-0tI%U6D@@aX&=om(B&`%i@mv
z=GT0SyUOB-_~HCj{A3T}v^XZ-lOs<Tujp%DAbIYyYe=5@_@%cvKo4d7vM2swULm{E
z&UJoAKS)pa&9gEz@7533)n~8b13x04Js1Z&U>E#JJMyhxP<wLquU{xXcD~3C*&QD7
z?|kyFlQ&vsnMYRc6L-vu`LI4In^$q){JO4ezRWwh@(t@IIp&YOwL)^(t$n8VAl>Ir
z`eE1dN%@2QrTmEh7`Ob#`JntueJCIEyBzO3c`t&V`cpOzzpwLqIy{l1U+>e=2Ve9s
zPI8UI{K(t*kNWOIcE{euJNcCRc!cbX-C3Xcg?Pk1`KS8fy6SxSecyZ6FZTGvE6M0N
z?@90zenl^Fz&^zMn;&*<p5<5gpg;d-FZ5&A+VP+8`IT|8Z+!7%?b(6%*2I1L73ag^
zh<G7Bm``Q=;laEZ4}UbT#zRlzG;j2_Zp+KWN%>nA;?w*ahxu|oLSMY`JN>wBp6Dyz
z6nCtjcs73aYaAi!pWT>Ob{eAoq56=%kY7P|2<Z>Wf#NzePmq4zD+-ZZd6NC2^@zR6
zuh|QG<VXC$x?>*A3;Fz<9kEO8%nN%kAKH=Ye1QGf-;ry7fM@f^Zmm!Bk|&ur=Z@x|
z9OruUWXI(4JNCwp_?3Ca7ds92Mecj<q%59W$LOK|P#(XgkGOCA>W9~jTYd@Uoz@ZS
zkLTX>v;K%<)-UsbUwY%od6+zgUe24uGv^@Uwf5$ZUHKlP@whLp@(<Vfo%O|h;R_Gy
zk&6%dY3J`~TTiXu+8ZCf%$NIk(64o#-^nA`r*T-<jE|l3ck=?p6KK2;&mpRBK9$*<
z{L4B>9zB%Vw|V97#zlYjB2M8?oX|g>^}}xYA$x(=1><mTfOmc#=4s*^yT^}q=969M
z&-{hRZmpZ_OM7vh9OGp#<nc@Q>1F)%5_j}toa`SjuHzXRzj;w#8G4V!ed{+p^p7|1
zQTTgptFAm^%kSRzdIJ5OInS-mzijkp?&}-b=e)vo?`e{wo$)>Q)d}bR=!Q2E*R_M>
zIsfpSQ9E+ggW9`KZ+d!fR9XM*&ickLtON2|`5xZnOZ4=2^sKA&p}*&?(EaWHw(%>+
zPkk*h4)3F~N9(`!+5XbH?7HVL;<o3{;y-(42l6`mUw+R%*^%+8$4>NTUX4>b^27^B
zU-`7W0ng%=I4m!+PI=zPzUiyIycmD(n_uguI7u)5Y8}Ck@$ftGjbGtGKIB~4x*<-9
z3*s*x<dGrDv#b~54_?iy`8Gc9p_^~~iF4uzG_UTnb9uTlUdd%g+Nn=(ddlD2#~c44
zN7=mM3$OUWo9}p$FArrm_NC@c|K#CK|LmOp=82tJAGNdIY6taa9M;h=@3StjC;5sz
z)9)U{Yp5N&lz-rxJwx}6M}NlYKEBD<o<F#5y`YagO1&&(uf~ZN=b-dApW+^#<q7(C
zuIjpdfcElq>$kiNZ^j?`0Xt!*tq?EJb4mI;FK4&v(Zjs3A9@-mUg*WX$$|V5`Yw>~
z;2iqKHE-W>zL5jBj@n@DSKgRwz<qz8-FJUo_q}!H&9?mdm-m=9a`5l9yHCD))Bkkt
zv-@u~*8sUYzH^z)b~$|H;NNMNCpvdwXYxh%VEp!p-ZSQ3>dU|6vGnn~M*9T$xqXQD
zTb$oHFBG@^eItHrKSdtBjK}Y{<*Tl1Cyuj!WjwG`zuVG}ID!xBnf{Gq?>m=#<mg|{
zPU){-=sMn<&l)#<T_<0eebS4*;;DStJVEindaEDnFn#1(?nCpdp7<zFkgth5=9fNr
z7O&WWby>S`-q@x3?z@f`<5Q-$e(A5CGX2Rl@7hD{=+6$=5yYSR@)_qw#_7A`>_U6%
zg83Dnv@<^X>c_gFES`y*+PP0JcCTOkg!RPdaGk;<Bv;wE<V}7TO+I7?^dtx0-ZQTo
z;>)^c-LOvZ3wFdGLc{~RmhZ7MeBhNl_6_k(4!PuL&+pl*`sNAm#>emVOCRG1?MHv=
z(Mwr>>`y<+<jFhOv-k<s579XELr?alY@FszzsmG-JqwM?`3&A&7l*}7`4aj3UEH*e
z(T{x@m%Ph5E#H#Q$$#X_&N1X;@<8jDdfJ;W_wjH2!JoVbU+Nnd{p?TV`SK%W`wBSN
z&z$c$TK}zwcyL{u!~?&xK0<k*@zWo_A>zZn3$jb=t~g7svU>P;F5y02*)Lv<A8&Yp
z;rl!G{pP`bQ@-mtx4hNwU7VXJL;lOo*egA)lhz^Uu_5w5c5l6tm&+H$6>*9Eh-=~&
z|8-uZo%K(BW$Tst<hoBk<K?&fj-8QXzRaiV_6Ov!EB%n?KL62<eQL)q<k{?kowFaj
zK=Wf<;*W8YPY?B7H!k|n+kJ9eH-2#uKhXY;T<apcmRA@jG(PenyK}ySUv_|3^Jv`?
z&mjHyCw|R`{TUv~rH8V-6#sb9zk2w_qj{kZR9}1J;%COIp7}Fh_!TGcX}tIrU(_Rq
ze)Kip@alzcoW8}vi}aZv@j_g2jwOE)Px0dSM$T{8xp-wBmYnbA+m>E-bf0rr{_EUU
z+!BwqcOGSY#xIWH>CV0Wu>LuH^Y{6E0X?jb?ACdv=fTd4jnlg6oKk#e-`blO{%hU9
zk8$f4I*+rS8xIsO<qiBpKFw~}sr&5J^Ah<6dGZ+FI}q2!Z*f@sCQo~I%g))AeUQAs
z`pU1=gXRmrc(cwy^TwW?XIuZA!{ZU(&MUPe&wgKj)+=_)?$|wapMU6If6l}8M^D$;
zi9AI=o-^PHidXU;cCXCd*r~F5_~0M>RJ_oRUhI<o{Eoi#bPkRW_Ntxu;5xpbcIxqe
z=gH24^+S&S<wu^A@JnTJ%X2dEM0|1Ia|PobjGuG<o}Ij&J?kG|%H|)M2RxHw9d#}t
z&ReI%3-`^J{@Jho`Hg<q3%??tpQ>j*{chX%`K37M?~>4iT@J4I)1KVmrBx@S@+fhW
zztWSRxX-Tm6??TViAVB1>yv$ycqHB^Tj%T#l=(Y7<-7XBm;DAi@O}{e*sFQSLi}VQ
zKVk3im*;=xYim3|e;>ZY2mM?B?34JB=c3jn&++hV9{7j(@f?dC^Fz=7l<CVZ<um4m
z-QX9>XCS`$fqKg1xz1i)R}b&(J*@9O<d4vG_IJ(=tKC2Ge@6^BZ)bPbJNX(v(q6yT
zQ+{fnEWRqU2kVl!CNA1v;uCNB!H4-|7y1?F*|9SJWjBL7zWBj$Bd(m7vVU=bp4LZp
zCBEq&f9ylP#cs)CN6Ol<bJy{0zU6Ir6erE2^-Ubar@X^FS%>k*zqRKN;uF2_BEFap
z`v&U}|Fy2*Aslxo-}vxLFZ0R{@UA}O&*G9e&YrZxgYlRz^9*NRdF-wSd~ohQ<C32*
zdGe|o?K@$vKIl9G`n!KwXuhHEinT)P5FVlTI3PZt_cuC4^Vu%SGg@INmw!X{Z~YQ~
z#7}5_hwR0D_R8PcEqUU&_O6Ha?z(+}`_Of0Ut+!1kN7DL>DP7Rwm!?7$aT(S9kR}o
zCqI!FkSjhKhkXtC;-mY}J_|aRFh2V$<8y8;o|#X-cfgnDiJpskpVRl|Jm+;^Kh`Dn
z<w^QgR^NWmbv*i=g>kqJ@kS0bKgJ`z$gk`}$zi|vRiAybOT1*+gMO`R;=Fj`xsiMg
z+Mjz*EzB3hS!kZcN%gH8)^TNaVqFqP@xboXr?2$|AM!QrvygwYD{{s6yok>b=@p{+
zf%3|{*h(++)C!H)`fr_AhSpJjNpE&%-dq<~oF|H#>a$b!C$BOt^P;}8cx1l9`^KX`
z^Fgk74e1N{ONj0pXXuam6;JdV_Jcp?J>-fH>O*pkQ+wkR2gC*aD(i=Q?eVXkJWC!b
z9;j#j$#b7S%a8aOzWKj?jMw*j$j2)^v&wjPpFYa;Fn@#m&G~MO@yUDW<@+wyIq}E3
zMm{@bx8j$1(4YCSj>?Cv!`54I-h7*9a>YCKoM)I1_4z43F)!hKE1L&=hddgu^^RRx
z-{c|IG3~?oAxA&>6(8_RANhmzLO<^7$Gn***R|6hJD|ULa-Dql$t4e8#zDU8=D|F%
zd-DhB1IdT_h2{^6chLHph3pU-xBC2-J>gyaVn@b<Fa4W8^0kBb;aA=Z_5K9j)x(SW
z_;%lZRJ_n{s3-rXr+U^6{F-m_T-TrL<U{*7`$Oxx{hvHld;3Soj>$1j^5ipkx8Jfa
zvwyQsbxy0DyvICvjxE0HH_Qvvb6*}MPTB|h9+~}^deHtA572x-`*vu*4nzISBgDJ*
zMIm0Z(C-GI@wP&Lj~;5zF8R55V*Xthcf@=4!tS8^5dYRAyxK4G7x4kV{DT}Q4v6#g
z=fCWmU6ady=n*3O3DLS1BK=ySd>}-27-C*KqK|l5Hi}Qs{<B@o;>+_)^J#n`%J0J8
z2gh<`ATlswgmWYDPCmg;y-zNFI=6EE<DA1emHP5*W$*dRtGo}-kL6kV(cW_vdP4D<
zeDPj>DlZcc#Chjl_W8z3j(p#^ymuk*r>FOv?Tg64qxfn)v>y_WtUvUWkK@NUAbIvP
z;d-Qh<0N1Fv`$;!tfSUx_1yn)GHK?hyXG1EUXJ__&(_rttxxKkC;2n|j00LH!gVi{
zuYVZwU_4o<oxA~>C*!BDeJ+0SN?+rH?uR_!BMY<i&3j#cA<D1ipPqwx-Yc&%PWH(U
z<!$l|cEqog<=yf?ejtzcJWSrFA9D1gEPr>tsJ-V=>hTMCqkRFr*oS#`-+3{62=h2)
zev*aq)4Yfec5YuA>d79RZ_?lXoZpx?`mtNjMT`$0A>uC!^_zvpL0|hi$Zqgso}u@9
zVCWy(hiDw^APe=Ah3qN|*<XmQo?B-1SJZX;H}k^Z_@DO~z3=4thkD*0V>iw#=qEn-
zUcLFWueTqB-Xr1P{7n3T;XLx&EOed#?c*VRq3>p8^-m8x(U*Pld+QFnvMz|P)-k+=
z^WZ$gx?sIv-`exfFwV;t^vl2G1LT=c?VU%6pW?pxG(K@B9Df)G$RSsp$D{B6TSwG)
zpI_0#{NvlXh4xTA=SAWSWG}{pPwR^FYv<m^rM-Ez9?}~h#>-E<e}YHv>6jPusvIIe
z^c@Q0^8S|j2oXR04?pm#r4PUT^&RHv;~x-zA%=S%Jn<L2Lg#C~>(L6ulU8UxLd=S*
z;W*kYo7XHf-&ttBLbRT@!dCsVKfLj4b`~N#^qwrggx0IDU-mQ@w_e|S*At(3D6yWh
zTk*todU;=+zUEJV+KEr>OMCXM{^{$^vB>cUj2}C0_5S7hm+v2W-1sr$`$xU`@3=+B
zEirDX{t?HFAGh?I<u&?8_K)0RhgFXqKkn=kHl4Ea$k3j<F3*}U??Ipc(dFq-#cm0@
zfA%_8AHL6RNzvG?UUhGO#iAp&|K3+_NIS*4jql*IZu!!gt3IB#i&^7nS58<zOK#EE
z@Bil8-hb{AlasQrXnnKf9k=uvZymA6qe;a7%tID$?Oy--HoyMpjz|3?X*wQ%_cl+z
z@A$itqOn~*v+Ccv*`GJR`0KNhsxV8Bc>VtvxZ{K!=6(6{zot#ay&gK_`+LoGZ(21@
zn)>tto4s&N(h7Io@8};SANpOI7t6NG6MneBj(=EuL|@bJj3t*p{QbWj*;h4oJ0GWA
zG=9rVb~z=9SS$}~UUm5Gmu+`<+AS{k`@;`Ebcx&3rsKHXKm5$Rf4nJeDz31>{i{#<
z+yiOEf9Ijw_|vL=S^J}^<*u^U>4%+n-^3)uR^#vXeo^D>R*$0ESH1r!O7EiOgmzR_
zx#)3(^1grf|E#h3HS_dE><tg2uG?zoofe+Auc`R0Z9ll+J&VlU*9t?s>$IHre7{rq
z-M0I{uQqsP``<p6bc@~U5!?5cfvFokcl7u>Url<2-Hv|p8wXzb^g4e}nugsz-|CjP
zqS_agcNL{imYnc;PAqr!e648SvgRwi-tBowR(t&vwI3pf+=!8ZJPhQ`7t5}fN7q=V
za|ZH0uj;&qd@ZXwH!j*<tJ1ryemZ?Wr!MV^p3kb~<$a!6*LqbyFD|RUvivKQn=cxU
z7*VwymepU;&(Vs0{#(|#do8!Bc#+qhyRG-lGd?xq!mFpJ!v$}=>B)7s`OB2FEDY!4
zf8P1c<)`j2Gc6m7I-d#U95r#9>)*QTl%&@f@*2y{U|^LOjz4v|(X-NCV^M4sy-r4X
z?Y(zZbe_sC9{l*p|5)R$htj4Xy@xAad&*aL`PJGlBu&LfuHSRo5l2r;dX0JQXSIDF
zUwrq~UroBjy#19`uh;vBa9rKSv#RzZW%cv(PyhP#({K4>QWS>#<}GJEH|q9!-u{cK
z_xlMyU;l~urzS;VQF@gvCp-`8Rk`eOX31Y|-TSwkb?_5u({cYV&RY7)m7YqQhAYk(
z{pzA$dpa!(v*x|i>v^B=Pn_?!$0c_@nih@k+V7%sA31Dt3d^4Fg|FJ+m#gnTH61G0
ztN2z_zfIX^g?e6h%%r=o{mI2iQ*pTeex=iRS@(9W&S&ecf2iL+E37f$=H;JDBld=e
zrsOf9-TY&ojhEeG><ejAv8nm{P|2&R8b@fidCQONxW;xj-gidYbS&z;r)W8O=b`L+
z?S3ESgU^iq^vD13aMCWWcHRrC&A;rQlf3xJ3pQBg;D6qk<i&P>uhH(f>r%hl-+8p#
zzHIwj<9*vbI`4unrV%^iq3rny<^OcdPapr~ytC7$V%hJ<lr8`ByN}s#;!ZQtUSVht
zMVG(+z!~W=?|dX}Dwg%WSJiU!p4+xtuV{NA?{kk{I(@}&yqt84<kb!LKVzQzSGeQ#
zq*vG}AKR|-%}*?F)bym+*s1<yKQ9<6@;l{sxU|drJFU9ctIO|(iq>Pog1`OZsu$mv
zR*j+kPyFUx3%-4=NlCA;sPF5B<LXuU;gQdL;l<k*80a;2`(E-Jvv=NRy{leKJH@>H
zPygMiqyFcLS!r3=ss43shu5b3uBE8w6^Czs$i)L!zMNKtcy20|eJ)Uz{9c_iWcAyt
zb)@Kif4AU`D=mEZd;<}OCJ#m1OV#p<{@rg?^1}D1npQ4)UqfEi_7nE^rAdFE`?+7e
zkw)wd4+pGp(62u9_Ynh4$DwXN{N}-<e|XmEN4}V}!oz=b%$L{R`I#gy`kq)(*y?@2
zqVDH?4`8U(>-F!*y6wNI?=KG(z3cMcN~e0rUz&<Vzc<#cymp@>w|YJ`aKhUj+WVGg
z(pK1Rzv{R9e7W8CD)rM;Y<IkMtKYQ!S*UMW@2^)Sx0T=4{eG|PS)N`tHtoAD?Y>`H
z*ZNiQ&VIW6y9wHd-_>e*Iq&=8O|57CtHO5Q+s~?>_kH`S#?`8y9mmdk?gy_;N}GmZ
zzR{F&RlhIG%G<krkEh*zN337Y-+yT8d*Jjg$}YnA*6sI#dfmS4^PG^cx|GX$uc|7!
z`!4s(&)<95?6j^});xFncMr?bt0=iUoVUuK{<zEYX;bl{J2(CIVGB)9n~F_+uZ`YC
zt#iGSQ}lPIqP+IL=jHocWnocqGRlbzL<S-Qk%7oSWFRsS8Hfy&$-oEpU2DVV7oL{p
z#k}tgw0f^6>wd4Ew}<@>mG3n+^}J=M@E_V$R=KF(O%^RD>-ksHuIII<GGlF&9U15~
z164gYFRP!T&ONJ|ho-#`lBI7|zpt#ipSpg3wyeKX)GN7F&udxz)zx0Aif47{$NMiK
z)<wU%>k<C`tXqAv_BBnvUiaT8ME&ypJ}mN73{Q34_tz!PWW~{<e>dLl?=HH1zpUu?
z?do5*`n~#oVz>SJ9Zy+!_oknibnT)~B*O*E;;~)1S#c-t^|dbj&2C$+IWwsn=GDK`
zdPVQ&I@P17cDwxS?s@jQbM!z}Sd`v%m-B<a{{F0|zOZOt)9`CYJoo-FqZaRLDz5R>
zeeZbi-HZ1%70<r%@qtx7zgS<>@w~qtx7wa-E!o#J-1_`+J6w9yVtq}=4Q_wrxP|6g
ztgkNEZM?4Qcj%34>PF8UJ^s#DlTPtn?>&9YiB~<F^a@x0^6uyFf6bFg#NP1G?f!=y
zMjYM@6#YI=l-Jqc)s25Yu(Q7(h_~VG#l??)!YBW{;&-=wJ*^sxAIZN<9GL%OSHEMC
zdHRM5&NpkNt2SS3R9{gTp8pp8T%lL;!f|z4{&#C?e(B_GX6^apo)0B;#iG9RTz5HD
zKY!}fa?9c|uRh&=KejIIyZtV;_IY{Bs#n(c3d;H&d?>f<?;Mu>ou78)|9-;H*MDOE
zsYy{-^!ZViot1UHsPFTarC-@{%gU#U(z~hOwXDlJQ8zv2-gw;$m%nCO+9~G!9cbO^
zbvoX<wQI`Xc?|XJ)pxOqii3IO)a5<HZuQ9fJMOyG%lh4TyVvv1cenNGx^Hi#SJro#
zn{qwx_oKV5m)AbpeV?yx^@qxH<f8T?S#nz0Lsq+@uDAMoZcVxW$n|?pJL2eR32en{
zr}rmKeR_e-UbrTyJ9aw0cH14d^crs+vB#rH)3DWgR`mULd9A8`r+)`+sI&`z*BHx_
z?!NXX7bj)&Q1&@*+4Af9cY@m0tJCwwFg}&dH~CecpM~SBs(i;>Upw*6g`Q4e+5L9<
zd)Kn{Dq4Q{yeXD1I`@&oCZ~~y|IWh-8{EJ8q|ZH&HWiEd4pNlU-go5j)#~qL6&>fw
zmUqXo6F>aT)#e|l3K!Vm(I@}3>H-5z#RHd}vht)u78q!SMbDeOsZ;Fs@0R6l->H7B
z+TpY4?^u!BE;e=j$kMyj@3PwEr|SCMp63tM_EpbAr~Q3?_c8lT+-XMIE_T{q-ge9V
z{_w*OUE=n%shBrzcUmtmk72#=`)y4txBH#JaQs>2Zl6bF$?f)fw}02M+x}zwW-w6n
z_oAY_(+@lEzKKcXA@cC19W)~f{7cJD^NPB*``5^N`@i+p-frQQ-}`)0cl^btf4b;|
zcfFJ}6^rIKuhn!K@7sND*lqp1<7v0vLNj;%##Z+|pEMP{zf=~s%ctj$O~tys7n!9`
zyYtd*{kndaw<x{J`g<Z($t~-5oaA<EKTT=h>AAo0wflZdyTAVo>vy}Z_NslW=KyuT
z->IF{tzE0<VNJQ;)Zdx<9e>ksuZPa~{$6w4n--1YXj!=UZ%+E+kw-4s*DV&MU)8^N
zDoXDxIo*DjFt#6F-?z`2@2q`K*7YoVbiL^But#}oY<J^*XQX*~XzKc0w%&LAY0sN>
z+3}?`3*VT%^ET^U^<vs|Z0g^Mj(V^1r$6rUd|DL`;&sHxKx7~?Fx(l)%8!e_KHPbZ
z<BtqP1|kD>VW6t-0k`{It5|>5(pRqZR65josEZ%8t4C3ObNae-dLDnk__5<w?_a)u
zh5nJpjUO|<f7F}*j$3rx662QYA92k1aZA5hUZa0x|Hv(NSoPTP<IX-|(<v*DY*+7=
z^%r<<o^PCX;>2{QVoL&ARJic<@2vKfPu!3ejV-!u{Ox_e{rQ27=YBNl6zfL7Ny|O?
zY`W|1NxPUej{p71$Jd#0+Ffa<m?gL9>+O!W-TFoAn<ejuOU!<9zbjuzBK~I{vUqFv
z`ZGuDwbjuFPDmpDXCB(+Gpqi$9>4JG+fAFERE1f3#OwdZz{2UhyPdMoRY_BE$LD@_
z)z#NbPOHWnpZor#&nz=ypcS^ePL^$#uO4&I`WGDZbka1O^!4{o-stU<lB%)W`MCI$
z6+b)u7tf{<i{+uHb$XS#fBCCTAKovC*jpZ&zK?G=&gZS)ZTxA~{)68i`QR}hcsy-7
z-tdmTO&(tS`832<<L~zV&EMZ_rFGvkqOV)*R*$0ESDhCWrFYSCmiX|_b8mgwl(cFr
zdK{s=)eqiy=ka}``y%#+hvnDb>$aIwW+qL=RVFRH=>4;&B&{&CyH3k_&tW^2-)+0I
zKX=;^i!3;*uUqU^kJ!Gq47~O6MLx0q4KveTVYj2-eEn;lSm4OVlcr&}=K<aFR#f|L
z$Dw^u`eeyD?fWmyv(gVn4MeOL4_WgSUN7q$N<T%PqmUahGLVOXYd*f|x1adX!)dQE
zkFK#!=M3b1Uf+2S`C3->+$`^S^VVDIJ^PP8bEW%|rs7wx+ivkq-}_?HRBS5m^e*c8
zORwad@Q0Nb+wHO|(xzio@h4w295I4@bc%V;d&;hN)8jLLxcBi7Cq>~sYu~%W!PB2j
zio&Aws+*ju;z3?}?zUdl@1^v*zp~fka6YPve|PS@{HSLSe>E){Tb-j<bwA`amYcyq
zUC!Z(Vymk6Ld)vs4||<){ka=FpEMQA`n||p*ZtxC<wm}eHU*2a=cdZ3YTX&?{p6kh
z^R~V3k|UOXBk30N_E%QDqVuD6<%Q!qYyKnF{N2pg(@yc}#|G9~_1#aWO~t(L8Ro4Q
z@|(B3`B{tHaMAjINsGq3{TEfQ=y+9>US-Sq-6l8wWu2oQOnZ%Gk26bt_+DvK%jZ5f
z_0F5NoR&5Xx8LiHlfE(ki)mSyHSe8XKlS>vcfIVC7t&T(*88R5bN<7>zvavy+&Voi
z3n!oQ<uA{Z%ua_2_WB&H+y2X%=d$IN<quWKZOZ<(EIm7omvwLF{kklyyZ)hmMd#sF
z$t(K(vR=vS^!aG3cHul1U2baLOYf#W?{E6=^QzXrQ@gJ!Pbj;es`<{_U#Ig@bh~zc
zZ`|qn7Vg7Zm0Rr>!~5b|UMxF*FH3&e{u0VB+TKEWWu0RT7r7yyWtTgBeo&Woz4~2R
zQT^tPZ)Mdh+rEl^?q0RLUOyM8d;ewGPf@?uEK9#`<#zi#yxaDj>fdUgUe*0h^{=}>
zwbHMu-+y(=?{H~X^*U5me+N&x@?)>Bd3Q3@uqr<Z`w8v8ta-~TH}CJ}TGb23)vIz{
zz8Bo>c|f=CDTL=GS>???eCY~%eB!0FEbR7qZr$43XSG^~>vBJD9B5UqTYGM|J)WD2
z@x96JzL)R4%wDY{MfaQa{5)Ro&G(Iq=DlcpsaoDeAMW4gqywiVO~a~wH`}y+s<xl7
zzi#u_y0p*xe7A1(>cU^QdekldFIpddlXpETI<G9tZnO5Ad9Qan&b<B9tzNso`>I?0
zqMm#9N={kt2UjJxTfgab`!$!?@ybsh^K#l$4Bw~eb-CU5y;_|Eh3DB><#zj3zuo79
zS^Z~SZ#BNU-_IIXQ?9o=zjdqcTq<HQ=X>q)C{9$3^7OK?SHH{XwBM$ki)86rRorJk
zMSUlyZt<~Oee&*u>r$`X@tXgtu&BRll;t<!^{U3zs-LR-x#)h{T@UM0zpQhlP;S=q
z;7->s?Xu*?>;L)h%az4TINwbxxB8t^-uq#EtJ^tcyZV%Ut`zcBmvYm8m(;6s*u4Fg
zHP79CuevO~ijrIQcf`w<U$^gfx2sR5zpJfFyQ225QBJeJn_JiUYd20_u-A-Vzhm!_
zeZw8w<*lf5n0EWks^9K;Ox@}?^*b%}F6wz_*8I9&_4$9R{^(Woc~H~j;h`w}*tZur
zZ_RyPN}7r@AG~?`pLd;|3>B=)@2jfTqm@0k`h8q@KRh>&<-N8#`oIZE$U|LzN7b$#
zt?VW1e!IVuYPBwPd%xXrmsP)M>tL3?O*_w~Z_)2P_ex&g^Y?cBu&Vm<?51Ph`Ok`z
z@p@!nC@|2}eNEANS2dU2=0kO9U$=9sqV!pQ{k?9RIb~*2H!S+^-}G8u+0W;yl3&z!
z;EI-0RXnRpKUw)`UG%HF9`;$VTYdB9*G;Qe)boX^<TQ=nsPE6Ny84>QX;XOc{%YN@
zF7G90#nGbkuXgj_@SMBX<+|BVyZW^I{#9M-mz^iK`<`}Fen*<6XWsMOvg>8VoxIn(
z{eE*<?eps2X}zM~Z|qc$qS{sc{&!J&*IiE0>vvW1s(x>w=>8U5;2k?XwE8PaQ?Y11
zdL^%{_+GZ$Uhy^5`|UPf_j>=4)xU9l^}6jA-}JpNCY@qc@ja`by6iW))uX8Qy{nef
z?*2Ti&kiFFZwBh}{CBwPQPuhVQ0-@UdvWoj@73QsEPf>aE-@=FYwGo){dwW^-rY`F
z=&A%}z4zO+>qX7)zgtuDOGTeg7A-I9J&d7tz1`<S-PZ5aa?ARA3VHSE_B$4JX<yXy
z<h;CP)hp|}#O*$(>$HB^&l|0Sb-`}^K5Sof9L=(`vaV;HPqll!Ed9!s+o|8QYFCur
zP0JJMTNi)tRF9(a-B3=u-$@JWm;IhXS@O&N-dk1j%i^bOxvk<=S@(;IgL&mVIsg7+
z{`U(D_B9<>KX~Jv$M=oyYZ{*Yx!aCdWWiB=-D2MP?zUdl&z*|uFI-n*`9FUzBCmav
z^__yM<jT96j%B}R-^w1s=e$Li>vDe5tsde2uobVJ-VeW%Q`d5*<7>BF(e_xCyn`oQ
z`LWm6ygR81+vT;Y`r&@A-ExfYz4<*>RqJ?J{S<Xx-YYpp<;}g8)9L$mVSFmfPC~ie
z<_Yb#564+mIXu@dyWdXFd&<_UX!&KGD^(@8s&mEI&%j!%zWeDE4@JKRQMJ6H-z%zG
zUZ?S?F713bBMXa;b7jlB==zWE_rYyP4pfC(ukoIjCtp6#z);7c=dIiPGVgbXx_xis
z{bR4&aPf;?PCCV`a~OPv`Bb-Ma@)mT?N`d`x7B<8?ebG~{Z6095A}9c&qJsEZT{g)
zSJ>kdFQx5br~Tz^cg-<tF7tzjUQ3#ao&Jt7Z@auahV{bpfToq(eJ&b~KdapB_nouk
zc6)vDDPR8bJjv{|Q|z|?*uEJIl>MAF%J0o{ipYCCcyC4&_?MR5?$7Gh{$C^O?f*J`
z-cxnErhgBm+jEz&eO{~S^t<Fme=on4yu9OSw_aWU-e+FD@}9SKTCZI`^X}*BTCY={
zs&3crymVW?>UbXeoB6}NkAFDH%R{&4>8;v#Yd=kC-|72(#@FuqBHcdMef5}w*1zDO
zr<0~(*7t6ka=mW*g--P;`uD%Rk{9N`O)2+^uWtKo_r2e|^~KS$u+!(oVY{OAt9svB
zl-^l#I*miI-7kN&>BIXaMe&d~ZyYN1vh30Is(vRD``Lf|nJe9wggoRu&lqm?T0Mt7
z_qnNe-n8YkwCULN-~Du(FZH^8QR_<Ca>VPZ@U4$8@`?3tn3+cGEe~(IZnr7t-!^(+
zxMI=v-sy9ls@fGj-k~b5s`p0o@>i98#D41eccY7%m+<=wvAp!6_s^b^)P;wl=6Sfu
z$s0#o)hj#CCciFuZ`HpKjQ#CA_tuw9NwaupWuJAu|NRgD{F9?D9M#t?p1$s!5sp7#
z{Md1;_pi{uV*kkF#*Z1_KkChY$1OT;iE&Hyk2q%hxTW7LuhBoUf8-WBta|MDac7^f
z>6Dd67S7Ld$4x(O{hwT&3{~uwuyNo0`MOW;aCcHPcB@z2+fQHR;WI|x@=)3-)@^(<
zA9^OaaG__@b}?%l6PMg`g-`!>a@r|o$u0W&1FPM>{(|3mC@Bkz);CMu)w}F5=9LqF
zl|=l{JY@0K?)CW(pL_DhJ~c6EI_~!Q`!0LxncI`1v0Xm1>hJN`5~tpB*7HeKn59R&
z{(lU->-clloPGN}X;bl$lh+z~;|ae`tHv$=diVi5_s>XL;RdIjJ#oXij!*Jp*>-vB
zMZ5lDt~F;TO~czivg|c)+jL}K)!6NPjN9g!rKY^|pGm}Gd01xL73&;(`9IQb@v^fv
zJ@hA!{55SlrdzG{v4<w!kv0`iKjoI!Cf{{;+EhIB>tnvT=8gBJMPs+|r&ar%9-i_1
zGTTi{nvT~zzU+;2@A!NYVyp3Ydw<7$?w|WBSKps@i{0u`RQtSnN>TNS(z|Fm8%-H`
z>FIxeI;k3q9!Dr|_Fbn=y5}#`lZd_H;jbUPXujtskLYVE-n8*oZa!`MQGKm2w7X8r
zdGotY<#*ez+qkcNw|d0(y=7p=J6^f!$S;oW>lJo8`l8nJs^y&V<E0<p;e?;1onp7V
z71jQXgKxfI#|3Uqy2YaO$&yp_^Ta5xnCG-v^A%oy<&-^N8MXY4NztgE6_(s|s|Q}2
zoD_xRMvM&PVW4Qc?o?hLU1Ocj8MtihiQAub@dIfq?7WA3EkE|dcbtFKgMUk#hOKz3
z`~7y`XQ^BLrrArV@0Dk4{m(O&yFP6y*4@7I>KD#SS>=4uaKwmi`_DV>STEoD*EF8G
zwV%B0^V)N_^_Je{w1pRa<@t2D;1e6|{`U`UJu@u}!}&OB?NJk6y6Kg)Y+UEso37pb
zs+mdASkykbTR9=GvD^#>PI>FsPq==Q$!V{#D7Jb%j&A#J+Ik!6n|J-pTaR5d?VOk1
z!xc|_@bY;l-1ck|u{$2}+RxTA4_Rxg$EGCRV&4ABs@H3K3CGoKUevU>Rkpq%zj@1L
zy<eNR|Dx*Mv(z@%ZFS`2q$n&(ud?MlbJC@YY`*jN(q3cP<IIwO-)S%W^|pz3rA^18
z-%qYeUfJ)pXU%)3*Ylo>m-V}=&mDW!w`b1(rxcbIf60Au-=#jh^-DLWLj{}ud)uad
z7oq9DlWO|!hr89k(|B3;c6Tgy+RR_g_ehe3b=N=C@9Bw`jojjx=hCL(e!HHz>E^$E
zK5aTK@P%#exO~4C(xNe(=c3C?PTTc&yDsxo8nId)7M}IfSN?d;lr&;K9y+!AEx!5g
z-@Iqbdy}SP);iVe>z&R^(e1vq;`w7&KJnosFU~#JEqARn_3@-#96RFbJ%6?L6G>id
z_x$vp`_DRa%TGO*<i&*^9sPw<A9y}#8ZNQmyi0uJYmX&WVbS&$%DZa%!K>eQ){|+(
z&Ugs<EW7-(e@yt&ALn~6Z7P0r(!_h#JMr1Hsn~1Y%eJq_mN{>Si#B~EZ7O#9J;1uP
z>-4*it=g4kKU?4P<0TjS&h(^PEK9#`<yQ5bxU7CU)qmRu|NfbM-ugn)YwT42HRgZ&
zL5E-dSlU!Pa*>bp?{@X0Y1!B*zr&?n+0X5&lK<FIhi<*ZEq_mk8s>dYS#-V7{)>8_
zqEk6Vor{Fy>Q#A#yH`2y%sUnv=ryK0?DfIbU!0b9ir>5I__cREa%$QsF8se6t~1wr
zr>Cv3Q~mO`yJGA2%)V)#>FH3xc|ZG(?|$y->4iVPDEfH_d9CI@@BJHgJO7}g4}U(b
z3NQN1V%xvtT{F_A;v1iSb=rBqn2|>8j)z{YBSrUn&7ZGWWAO(T9EdnHc_`Xms+RYq
zmrg!)ixr<tn}$u@r_+1JIeUHjxF_yRi^i($C+zRAH6LB|lgG?Xn~saVZS4M^z3TO}
zX}H{qKe_y}551N)9h=5`yZZ8*G51|_#UK9mLedI1`2FLbpE>j8Brmo)hi&(Hc&oqr
z;QqI+{`Hpkynj~GDK0Skl+!2dF(v5~+kKy)F7>~D&w)4X{;{b^Q?V{QcB;n`Q&xZF
z!z(_YG!@T#cFm)Y_{UR8Q?aV|T*~UF({tFe+TA(Vv~#|9@#AS;{N>XhnsV4%pH8zd
zubfWnO}g`d8>YTCHO-5|?LDF|KDX7F=YRNA+Ei@%`_ox|5wHLExKR{8z51S^JiTn(
zf8;k#xZ;R?(x%{@KfLywyWe?3(sZoqcZ2L_m80K2=6!GZbJ8t-XqPp&{oPBqB|`<P
z{(FYq_IK#M_incI*bCCUX#T6hZO@&*|NQs;G3geM{@JpZoxAs6lU7*OxLWnI*$!im
z|Mf2(Oq+&z|L#lO>V@CaDyy9J-G$S~Eb-7IXZ<tn6pKE8kMcTuu95fqYIUtw)ZcLo
z<6En^Tlf1*pE=_n*X{AQq$~{is!Ms-?RT2{nRQ=Cn~Hhg59(FDvgWzlys#|2ijtf3
zefz@IwtG2kDz5i|5B%lr-<g#*6(4-hLbKL6@s+fx*!1_>w%cmncRn<3MjEj<JY2i&
z|NQu%%U?*#!gk~2@_Sr%@$WZzJ{|7ZE^nRw{lTzZyZvU>FUy{)lH1hhI`m$D@vX1A
zZ^K8^ZZT_qU9a1BO6Yak8jt++8;d`b3|GWMQTW8vNjpwHa_)hq;y<4_XQg9j%`-4m
zaKooJKK}>X&oeOGu$4WZc>a0E?eoroG%tSru-zVCeZ);^#Q)Ai+28dpOMWYR$+~~W
z^4I=(g&QAE^J3ojUk-WwrrR(4#S>|}IPa5d95LVcr_)~JP|3@i`u@uevwywo`m0Sz
zn~r(Ei_mTUT2}j_-lOTYoV;-;D^AAik%6JWKv{n;uqwGlJwK>wE{Do@Om^Jo{<*($
z_5JBk!(OjTS^ekD$D2~G)py&Pa=)zijJlOuRXnRpKUsNoUG%HF9!KxK>uKk#|3Ct}
z)pw|VFTJdNRatVE8TsIbE4=H8G-BkT3_Ix6?_pZ?+w1QZ>pv@w7X4kn-9CTcWlueG
zdy*GlJL&KpM{IjfQWw1I3vYSq&Ua5rnu=!}eDei6E^u?wG<@z$C+~jJ#{Wpl!ZUCA
z)R#Vd%!^50{Pm=Vj{U{G(~_p*xwns7;MtpIB)!J0xRdw#L2uh@)f-1mPU?nv_3yOa
zt!I93;(eE2kraiU>QPj?zyEXh$!qL4-#}GZl-_lhv(n5L$Nb<Ai}W=OZ<?|F?Z<p!
z(Y~hQy(4~e(RWv0xUZ?W`s?R>WVxR&+}CuR@WNk?xa8VJ`<jLa&3fawzkXwpzNX_h
z-?;c||GV45eRaWZ<8|GBPf8!-YV}>TR`*v}a?`CIcx`eLG4jxf9k9cQ!<&IG?R3Ps
zKR#$$8nL%LEVknTU*75JX=%jHco^PZT>R*(`h8H?Pw^uOB(8SjN3S|=-o6n-1ZTYC
zm8*{Y;^@AjFg!1x|M0mdf9z8elcuBR(A{ESO^q+M`*)D5u3y%7P3tB%^t-alof_`q
z-}%M#zAMg4ipIS943+mCi{dXYZ&~%i-z^tae&E!3Hn?TU32E6__IHy$@r47Qz5J2C
zq;<h-9$)sxxp#a%DGJ;DUVf|kS$0;|^`hcrS^AYNcj{A1y?udgpH7O#qV%r&cf7Ln
z+xM%V-QasCUY9f#i&|eoIg=J#`{C!;eJt%1_nUW~g-5^l@w92^dk{^>OD2!FaQW{%
zlQs?W{w}!adPT*-ymB^v_LdjF{QBGjO~?LOtKN0!&7%gIhC3|2&R*+GeIxA_^Uim-
z^~(Caf7x;$Snc-p7yQmcNmH?`?|z1I^V&yK|Gp2s|M9n-zxcxSQ<J9SukOG8iJ9Mf
zAt?)6*+bdy19bX#sOr}4;@5um#)eNlk-%2Gc6$Fi=iK_4ac5nY)*U+?U%Ty!+OJnF
zr|k7Gl;19|Rn_k{K9}{ofU4x4^|c2U+2LdNB~8PubB?B5@AmJ?^t$~sCtbS8<~x5c
zZ8~=Pd?1WZe?EV+Z=L+rJCm~UUB{oZ=Iq<=NvpzeoK=<k9=dVg4YzzYfo1o*$x45D
z-}p~FoVJT)>s7S;vY%g-EkE6AwU0eC@s2cNwLE0~U7x00uPR=a)z9*~-+21b&)k|e
z6<a-TE;`PYEpL-)Q^qg1>D&WV;d9epef_%=<{D@!eq{PVkItAp*FY;QdfvMIPEb?6
zr<wJ=1bh~~E|c3XmK7g+CAZc0q}%1E>iTv0zD3>isCphc?Jw_lTElvs_LsL^um0Xx
zQT?Vn?DfIbU!0bfg?V`l>n(clyMEO-FfnZ^w)?(oIR32i*)Q()$A64{GU*hv<aT?#
z+wX&P+kb4|3<egMeah(*_L!1Hj66g;DBcg65e5FG<tcCd`U%%>GC3_8yWJ0W+Wuc7
z>+S!xo_WYxTRk=<sXNx~{gQU|DVpEBR@3S4XF9d(PTS=jPrLQH%{P19zG%A%<+aOa
z)4o?3>U+(xpF3&k&;32AJGMJ7-PT{=3)|dr`F<~?Llv9;-Hlzh-)Zh=)_ox@3%j+S
zrnJB8tW6L7$s>PF+ePDR_wOG*a`IXuZ#?1GX{XrK=bQVF{Kg4a9I;Q@R9toJ(>oq9
z_IGL1G5mdBQ_F|$d+%mTkG&vm8rJ>2@GSi%>~qM-Z;!Y&DH_GmvhdR9&wKoV0~YV=
z7K_ra)AuPmZC8}uS#rYtLoD~^{A8n5uDk8~3q71P72ExNN0vRh9^Zeg_Wj3@?_Rxc
zQuls8I%(oP>z(*)nuXuH>-e>IJ#uQ=bd3A!a`)%0_Wj{^7VBCTudBifMt$kyODr*R
zpj*6ViH*+Q>#{e}Ug2^p{^at@KJ;4JRNUf|w@f+e;TO~4ibdOd(a%APmRIz6hpN1)
z_D^~F8?OEt$348`^KhI0E4F^m?3?zPo<?kshrDsLRlTym<3fJc?_jFF-syWGvE7N=
zpLOvAY0*5ivd_BSFN%lL*PS!M@du0_J8t#<75i7}A9>vPG2{D3z4`CBMaL~MZmIqe
z$BZAh^qb{1`bYMU++v4Sj~zem>=QPfvhv8f@)Pxn3`7PZ1CfEqKx7~?5E+OJL<S-Q
zk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtNuG7uSv3`7PZ1CfEqKx7~?
z5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtNuG7uSv3`7PZ
z1CfEqKqDA<V9osxU3RXgld|xYuYK|6Td%k?Eeng%d+g3Xo^R3qSxLlt^3Vupib_TX
zA_I|u$UtNuG7uRU0t|fV&M7O+xc|j;sN!AMt-8|&`%g;R#UU`YaR`xt$iOgTVAs3%
zJ!Fq(rl#<WKi{(V9q)Z3X@z;?e_7`|RmsImh*RfV?9#_qel(5vpLrOnZglN&C$D|)
zk-tq~Rd!g{enP*fd%0a6v;4bVIa$19$*;@xEWL;7^{jcV%XssyyLGKs7ar@T$534_
z<G3ONk%7oSWFRsS85q6{$g5kupOclBwK|V~=j=Vt`SkA7lWviou$8`}Z#jLLUmiYB
zUn_K85aRH)h&b}dKx7~?5E+OJL<S-Qk%7oSeg<0Q<?(*GzvF1dZ~mE#H6sI&fylt{
zWWae<*7;G<zdvggx8wcZuA9S+|KaKNanzB4$UtNuG7uSv3`7PZ1CfEqKx7~?5E=N-
z3~bUrd%w4wyI5b{al0eFcisV~FWDEdnLPYw>x)+-1CfEqKx7~?5E+OJL<S-Qk%7oS
zWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtPEDh7P7WVm2eyvBYa1CfEqKx7~?
z5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtNuG7uSv3`7PZ
z1CfEqKx7~?5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtNu
zG7uSv3`7PZ1CfEqKx7~?5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvq
zA_I|u$UtNuG7uSv3`7PZ1CfEqKx7~?5E+OJL<S-Qk%7oSWFRsS8EBt@)7PETX8Zx;
z$BtXQf2IDF`$rx(e$4p(QE&b`Zqac|j9aRI#4+Q?E&XPBjsB7SBe&RL)nmtxJNtx9
zr>s1({T^eBt{7PGBX8St$EByFMPpY=#m1F0u=8f??Q`6WS!ugixuOMkc+k7Q@`-aU
zc|I)*3syIZYZwE|ePHTomp(VwKwd1%4nw({j`{hqGtPN1EgBm(JyG494D9gflMi|0
zq^U{7_ISuiZ>%=V83^NZEEkiP6=hbdtb4(hFN~X!G#wZ2|H^^)Uh-PfRD5Z(b-wnw
z_xvf1`0qRvW#_|H&d%%I`H>_4_WPt=tU4d6%luUBe{~u6aM7b(KZx}^iyv9M6@4CC
zwY<FXVYv8H9A{);C@}Di_5bwQcYOTuq$pf#lM&nRHRArHY|M(Yd9T-H?uy!fR-L!T
z{@Ob?EQ_yR+p*^jMW1hl<I3wNVY^t43=BsG@~(qJwO-kCF7B0$dE;K*dfn>VYkTQb
z|FYKKcIB3x7ndzR?>RtS>-E}xo7R8cI9k;Haj3~@>T|iO^)5SZk)PL|W4-nm`0Q41
z8*%6lUQfG4^Auux+7ws2EI;V<{66HR-Ex>e701-jlvH*eGc;Lie^gCh_uJhE4E6df
zpwl=PwyTP-c7HDs)_+@a@*|HfI4fy|oyHZsmW4&1;|^DOdFQ97{b1hyhD*IL&xz&k
zejiel-4rzsMd_1QPN(rYt6koHdtI-n_|&bOUh~lO{=4OWsJ5@myf&5JqT+d!)86_V
z?mL=Z4)=RaD-U=3txFt@di-|=n!3(Jz5n~~5&rwEcYinX|KItK{M3V=PA$D@`9{<9
z?Q~vx-7f3=CSL#d-zyIlK6-6$vHyB8P}ch&QEoH7TTw4-jM|jNKw0k>l%?-bm)q;}
z%ue~OOJ3fo{;^$Tpa=$r+Ie_9-znyNC$z1h!fQCLZp)qSr|Yt>Xf^)5Hym^S_|LqU
zW?{GMf2aHv-EOGG!Lst?vg8&OUy8PiR`S~YT}N5<i{h^-a$3zx)%RP)(WcyQ>U-cr
zEq<3Zo=)wsYJbY=uc^O3Dr?=WN^V)acWRHZU1Xq63^XNgYNcn^x`XSo^4DI;?KZAu
z?Td%o^;ULO)%`5{9ctHm<u66`o0Z?i>qGN>omTT|eaVYO#gk6sLc4y@YWzd(eqElg
zMm@^;JELXuT9kheH#vFZXRH?)sFH!9;)k8)3%zdFZhldvZM@l6yZUuozv}UY{q@=&
zy6wNI{@ktIwA()K_n!0CYq$ScznuMa+4*l%<PWvD*6n$4tM(yZu^bubhJkLMGj!8;
z6>XdPJ_^0N&9h?r_P)0h`HFnaIOoB%C_7m7lCN*R{#hdj%EBsQXzZiD47B?kDb_FM
z?^d?T)AQc1uVqEes$`()=PXS*Z^)}>Rqw^tt)D7wKK3<y7zq8Y)!$7B?|*ZxHRpb8
zjgftD{|y%U!M(T6)0c%~Z{BJDj}9!<*K2I|cbZw_iP!)8d`Eu%ou9n<XI<+Jw|&a+
zSqb7uha&?`ecoTR-c6nRFn_D6JXF_#s`!ijH1qxCaNax*tQ$5po^`5s);V3L@?yL0
z7+AdTZ>yd2-O&S`VqMplqMkoiC8ut2DX%_ZoNZdUs`)LepXEL<^|VW$n`@vI?(_Sv
zKeW|yqx<THMeh$fm6sJoi=MZp`A5`uXym`G=CRxN7@O8UD-W&udRE*SD%V@tU0v=E
z6~5y*h6V$9^Y6OWtLt;Iyn1z-N7k)fUH9k1MX$W}R#m;B5goHf(QW>A@Z~?=;){pQ
z)z>Sm$_~ozr*3huDt)poGu|0a3{+)rv7i6`_cn??zZ-7y>h7~;=aF?Ee_iye>prwo
zy~6jWyDgWUZ}m!kQT~@@$90XjMUA^xa_Y)!UcEZ?>$2Nr*>S7uW#w^Uzgcmjl^uq^
zvnop8EPZ-?J>;z}<)V0RC8sXqje7KkfuiCh?TeCEm7KiagKM`QJ|ad2A_I|u$UtNu
zG7uSv3`7PZ1H+erR_8BGxgVY@4Br`zBaaMJ$3S>K8_Opj^2SM1lgLBlVJO={7-6dV
zJw;XLu=)wd8OzOPAn$YDy4Gvzy3&+=M3$c6{vwwD`}>bY^OH3%O}XBRkETV{Zr@9b
z?N5H>(FJEEArDQngQ#zN477>^O}pPJPil{ySUWP%TL$uePomZLsq_Ahuif`}s;b}0
zufzN8?n`<*%dz*!Kx7~?5E&?&fnL8)AN&8)XW#Mh$CI+>xhe8{bzamIexjbmFmU-<
z=luGa`EN;z!eX?FQX&KGG0^RIZF=3lDd%bJ%}V#R^PY347w-4lEmytnwma{=w$rfx
zvfroZ-h!fp|HJ@)=@dgdXiB-=cu=?c?Z%6`)~_3{!%d&M+C?k9%8E1pDMk+EYh|z7
zq5PrDWgJ^%ATkgc7)lIO_58l<e%gKBU$^>Ae=bz@c}l1KRqeZ-?z>~VG53%E%!_GJ
zJalR=L#18Oc3;%IlqIL=@kV*oFwpdMFzWw*?{^f%L#Oe#Zuwm+eTw3%DRNrz(X0D)
z=dY+J-0r&D>A88k{ghST@27i(;W=qj%AL-8t9C`%aZ}}-zU~~&#~(0$?6}qYSMFb>
zf8=rF$BgeE_2$3h79F?5xTX3>95a60(r=d6=pWfXa*G{SJ$C%KvrpJ`%E}`PANStZ
z?77QIf4V+tDn31Jy%j&V(_={%c1ze^uYTaNUtWE4S~O<qb?P189&<zTd{Q-L>C@@;
zFP(7a`X4>&?xa)fwBNGYt#$sW)3?9l#iU)#(r2%=Mm>1!EBB_IVwT*ZuXj3sdD|7O
zZ<f5G{dcIz%i^uw>s80Cvif`K`f<|_T=t2yXl$3ytol{kUzQ&6`u{PIH{Qj1|DG4*
z%{$w*_p<G>({;Po?eg+nbiH=-rm+5l*WEPzJM&FXA{NU-QGOE6S5wRF#+lmH?{Z(!
z>-ba_pGDbumYkmRR?xg><tLry!N>gQf%O)>_?fgS%$lF7ug_X^?RA&C{*1ILta^NT
z`^&PYvajEE<j3zFx97qGo#Mo^zw+2so6J9ug&%ut(Lc_+$*XC*xY&kkeCY1wrlp-?
zQGV2E|FK<n^VcqadHt+ay}a?em7KEf7q#w}rC(mTMem!2i#+`Hdfz~gPVu5e&OBnD
zU%iwTh2Oh#^Rai${YF|g&YX463L70bEA13J<*Vp+MdcMgUH@CZJ@9A$Ov=JzZ@OUY
zvs0f+BG!wCy!Kd>orQ9&(_1|=<FW@%emKp-R_kTBZsZ+TtoQE>WW85V*7bkyy?3S1
zE_+q(9RvT~b5;BJy7u$$Y!qKAYn=_{7X7@q>Fal^b)@e1>*_Cg_3BkzEvnzL*1@Xf
z=AEx@>*bA{-HyMg`R!Itx8n%ghw;4IawvbOl)L36Z~MIV9P72mKvC!XS#m;cy7lwC
z?X&i?u6OzzHQeup<*c}!SMRKPVLu_S!?o-jqZMY(cZcuYcglC(cvp74PW2DlZF2MM
zbC$hiu7Rdv+2`G5%im#@u@C<AvFT}1XnxDKzoO-5`ghLvRkGHv@Or2H;I%9)n$O`X
zFKhh{uNQS*QI<XAm0R>VRlD*+J#RexYxlnR#o0-x*!1^yi+(O0^4n=S%s)FVciMla
z?TX%KcdAF#?K)kjif&i+xb;``y!T38)_$<=*URS1Jog&Qnuku~Yu<KcJx?lo-m~Pl
zn|Ji8e$(@lQ2$=>SX94R_B33s_d0%-)qkt`E34nK<dzk$%a%J-&JkMKQ>XXKvKMmm
z<^!=_vEL^uD=!?Xa@lLUSa$px?(*B6kGj+!D(g;D{k`h-q-?tz?s1^o@s~Y6q5Mws
zvAVQts^8PQ=yT+D<%RZ8bh%x9hI@VQT@{T*J%<hD6m2i$@sDniys9v7K2`O)ANJQN
zAK|%Xugab3AHJW^v~phi>b73UZ&B-gXoppo!*PY>y!KoaA9az_ZQfe9_PydYw5MLj
zmvFuAv>cAB+j6Jxdv^N%><|7w_TD?diX!{pCW|N{AYxV!5J3bM6%+P|xGE!ptO=A=
zF^w2l6ctRE5HVmHRK$#8RxlzeH6ZA$N=_0b=P=|UD8JWtecZ2SDLdTmd%N$Pe|R`m
zr%s*vRCRUVzI}Tp|FivZt;-bnrLDiH=hf{xc*ao`Z~t=O2aX##4@9-krh0v5Yadm<
zZgJ5oDGY1ho&52_vco@L?SxI+`PA?8d$u^4D)*beSN|SQo_ySq&lWpTv3lN+9;530
z>coYqo=4)+-tBiaajo}zJrMSNiLm0i?dx9aZRg#f>&$k3)AB#Rcig5&Sm(EFixYG`
zhSe@E=Z{qL@VYO=&Cl<?9asOTKSyOcKIfD2Wo|w6u!rXu8|8VB{mQm4QhxlLt|Jet
zzs#{Q=(-I{H^wiS>gOx;NgBg?9yjbbUOz{ry5HK`2mKvey4ywdxvZqb^}5dD>KC_p
z@_V0K?!(gNyyC{7`>x;fk>B%#-}ur$tn*&bI5MB4+44FjX|~_>7L|Wo<~^$M+2RJ>
zS5j>k*5_ikI7vGmvSlB$t+%ktng6hk-=O-(tzFXAeb9WF|8!AazeR1QYh1Q|!aiR8
z#`F8RyZM-ggWevr&T*sP&lRbDKC*ouX^Wfc`Bk>(^M%yY@8=o(ha9=-(KBu<m1Cn{
zeew5uf5Fzjt(<<m(suvER_-@{nTKsXP;OgCwsJY&ZnoFs(=AR|^{h9qoN!*h=T~=`
zWF+PO7WKXzb^NfdcWiN_UU9MIzNU7#e?KOwc~13sh$^qJ;>PuP6R&xB-S^|_$N4d-
zw4J|`QZBD!l4d7GXIq}O>!YxaSFfMX|9I}f<NIw>JugJnJ}lk*o?kN6uPx80^i5Y>
z+i{VUa=)J^ZT^28mzMX+d%aFbmmljfZueEwoew&M#jxJTo~bzLT7T*4$M*R<uI1cE
zOqF4+XX42B;m6(fdOa1kU%Jk_VfPdCK9Aqy$m_ZedLD3#A2hDtc45y?#^PD7XWXuL
zGF>0`ajo(G`tL>C^zpktB;B9$rC)}1eEaP;Q~rMUQ}u7u?BG!Yjb9zYp5L(6X{O@1
zT_1R@m%7^aKP?Y--bjjWdmmr_<S7TQGjZd)d%bY?4exI@*}~HoR~cEM#SW8nX6n9<
zTfcHY%XZs#{z=!klYX5m&y%U@mUKOA`AePRZfE*)d(i%iw@y9u#9OwSq%!EZwcRm8
z-ump{CRH8cazER)kJ`R>3|ek`e%Drh;h|5>X*8>JPF!evP9v&v+w;gl%cIIO>bOxq
zcU0lBy?f5CI@5t}&k?x2FEH-)VLgwMY8>0X<F(xHeW`T$d(A(npY3x*(DI;pgw@XN
zb7I)_VdWk7IB~!3&({2c9*1#hmz4XUZh59_zYn{gR6ke5rF~pJ-;hVD`*pgWAM@J}
z<4W~<EbMjcH(s`!58U>pxYqk!chaAA7&UG8J+<81-xZB|ev<lot!&FX=<}7@?mxf3
zjrurAHGb6RkyP`O?K8DM%RJodRIj6w-ahQ>Fu6|S^7$myJi^k+wjOg!T|0hWg$g-w
z<Li}vynEYg<#LkBxO|@!G>@piZwVUTwvOG(qdG6her@lsk>j=bQSSErkk|UK_7CEO
zb^lQ678g6L@wvr``uy$pzDU&mVb8CdzJBY&&d)7gSkKkE^+SEs^XoUhyss0M&3IE~
zSox$oPO8_NnQkwy<HNH3uFtsnbN=<qsLwO$iXYYWThwvWrH||<)&0-bzR%!>$96xu
zT7l%p4d?86@10c(Ql*S5ZMXP8-t+hAdDm7hu(9qVZ$Efn`8^A4w6&Avz54HY+zYQ(
zD)36#UfMss+;#D%Ep{)kvFSUL^2g8Et-!`dJ~_74@>?ntxY71^aw)gXd)Vb+?duvR
zp1(u$y04~sp7r`U7(c(~Pd9(p>m`rxGxMsK&Ce|lTRY0-`6H-p%OmXl!*4vdc}z+@
z>lpgg!K;6&$7Q<OdmT@ze%zN!8d)dlVyZgFwLSY}Qkky#*M5#Y?7YhsJLr(q^+MiB
zIWM~98T5Ka?nmi5uX*h!=(uD%InH$1@;-gk{RaKK=rymnJlDkWtMvA~VpuxaKL4b<
z-0S|Ke!<>96_sw*b>i_pg4|EjZM*HSUiZ<s`04j|{@z2t?_k^*_WUO8b9vbLhOMV9
zURdYJq{Wff%hI;#6_;{8k1GHC`8+Q3<CaIzd9%fp&m#u4gT5yuF72Xz?}tr)xALUz
zk1}8Db*$yR+irSLpXzy&IJWClzj;f4se{eV^!X#(eut*xDctb*ZW|ocAu~DzJwI!^
zKVOffst5afQt0+MIcR-Y&zJj+gYItcQ}X)0W9$7Z%XNf2AFJ)}{!OO!V*R)=)qNse
z?cK%^v_4(=C$%5Hy4tSOlTz;X{dZjIZS!J#u347bo{zSb7tI-2y4{je1zyQ>&_U_;
z{B_v%QXe-v=(wJp{Ll8swJuW-HiPElwViFfWm~!RI?3gGLv--__cYwD@7(H>`aW~F
zamBqp=sMAMW#&A%ZMq3vhopXvaoJZw*D2=xv5!VYzx%Vz-|Kado8O`l&uu7Et$abc
z7?rNXjm!0MNzWrL&rb%;!|&&1)}M_@-$#?ao}+rsFI{n?x}RrTuau{2d{Or!zo$)_
z9ko7D$4|=rH`aZ+81#7^ZC9rMCF3_aVd;?S^C)p|?{+;C_Ii(cKaug}^@-d59`<-`
z@odj0C8a!F&&}KPljTXjj&{pmuIsqjZpT%+>)rM>TYb9o)czj$@|IVBcD(8k^mBVj
zZzuQ5ptjpS6;{1GZ)eKR^!HeC|Gp{R`2}6KNo^<3^ZlFsK9c<I`JMCX<sP0}C%1H;
z{Py{|Wq$iNuS#CEyc)TsdgtfW{LQYPTROLNn@)8e%+GtdZ_8!1ONXvQjix7Dbxw_W
zCR-W0U>eYq0?S7|u>0fl$`k~pr%ux7)aSnH&8iM7a{SVkV)5kFrKWV8=me!LZrJ73
z2Y0`!?DZQ>R17;`jh9&kqFN{Ej{9J@G2d2iyupOcux97p)@wT6^N(Lmx)@g8*&3&D
zmC3IySiIRJm9Ne`yxfMjHkovD`ro>&yyDJPjvIS#Ij7mTdo6anGF#VKNgdbvfra%?
z+q}k6hO0wKT~9i0RUk|SUOJ`fbMM@`$|Rk(`#@o8RT5&kUC)GFUlKHp(-^Os|M<bL
z&e(30nbDz1r`bPM=vB5L+Zm@dA7w7bHd*u8!{5wwQe~7mYV7D0Xte0myxnhE>tu?l
zj&RV5;k`%i7~~|5S1l`hO8fa{P(Y@to3>X4lB7Ux-(F{%Wt*L_nIsJ~*X&i`%7b=)
z`pK8SHkoGjDy$<@1yli5Kow90Q~^~$6;K5-r$D-%tH|64wP#g86;K6K0aYM<3UsM>
zP2XcaEiy^vPw&mUd*}<>Os45|#;&>kyn|QWx74Icul<!2Kfk|w$~OPFe9zORU(u!|
z`3sz|neKIb$A$M5H$P*B6Bg5*f6)GF?f&ZJAMLZ=$uxuJpK7~P-tTm8jq}z!NoRvY
zUupf=iL0HYG3w`PQe7Xiz1RB^U;TTJ4<0D9!U>yU@3U_4uI_Qo=5{S-IoZNNb=&r<
z_04i8shm}@<1KTX*-p~<YSqG;O=rw=lFB27H(h_<@jp8ym2w}<7JK)u6V4rT(hOr`
z(#HAply##<O)qki#=l+MYuy#ctap^V)*<S3SmX0i6H9Gc=OmR;Uq?n2zv5MkPHj7N
zoul-oL(t>kwk^%xy8GaA1xe>s*X;lBKfYM)#En7cCDZLXJ=){`9TycjQS<d_*B<+R
z^@WaC4xf5RpXDRRI$jyK_24z1YyN%CPhC3BbJE4A*V{|CeEROP!CM?<Fdd?v9~u9F
zvYq#AIIhr1Dvvy>_r06fzv(2E#V7Auzs~w~CR3E_L2Xqa4hlRnvf-kc)#jU28Hcg#
zDu=amKW%tRp>Z<BUFGDZ?D+og4}0RRaVFFBlIyN~uAMXWoQqp6caq9Rqu<=`lOHCT
zsJSccwCukMn11j7P^!f$C+VF0;r=~4{`{enZQS_yr)}pRH_N1(|6L*f`Cs1~_9z}Z
zV4jmH?p>+#*5N0tG^tXKOKnv^6;K6K0aZX1C{?OV-oa<LJnk2x%uXG2jjIBxfGVI0
zr~;}$>I&H2Ywfi>?B~kU6)))Dmx@a}x98=xKAPWyp=<Rw8@6lw>L8O;zVf$UN1Xn7
zp-C#g{<?U|p{GqVap9IxFXvC2xX|&+b_34dcx{V0j*Sf`3}3h3BTF1No^#~cecB9J
zZ~W5l_f~%2+k4cA35_lp`?aG~9fGWb9d{f(WA)jSjIw0ukopy^w+OFX_Wr-#cxb7U
z6ng#pZgKbH`YCF%3c~t3Bw_!(%S^{>-J|Kek(X{YN#TYX&R0+D`-|~P=09Cb_jPEx
z#^u&err(2I()y2j-z0uizrPFn?|Wq`Ub^m6Mb!`UoK%K2?{52>)+a-Orq4h7TBFAo
zn56RlsULOxVaIY4S0-Z>XqKvgDxeCe0;+&2pbDr0s(>n>3aA3AfGVI0{8s_D=W1mA
z9yz_w9DM9r$BhqtRn)HKSxX(iY~KFi$9qp5<7ApuFFtG5lyfJTq_fJ_>W_UrWR6K1
zx9(eT&%)M=Owu{C`5SldIDLspD$g#v`R6@%Uu=@jHdnrL?ZjJ(j2r*EGwT0Tf$URY
zaqD4qIvw+blT=pe_2Sm9PwX_QGF>w}`GxA!E1WsSB%OzSy|G%;Z#SE4<EhntoN?#{
zMJ8^n)@0Z#+vgRVY~i%y>I{Fo>Q0k%uAg=JOPBUuV&cYh?Pqc6=ZekGUbJY-N+TJU
z{7PCLcRQE-{l>zPMyWdJI?&^wr0xT@eLw8qqxgNk@0bhUd3pO+Lyb~(h;kj+*1wjk
z0;)jt3V8i{m|^)vKcaZ_VD`W8_@5Vl>?Dnit4w}v!Q#y(seEwdNgXe5JJBSSwNC!~
z%Re8sZ$Z*IW83mirXIUnLDG0w|5G1mGiA2|Ww<(|x+@>>-QzvqJ?keYZcKGtwC_^!
zn!d+;T4XZChR2n^;pRbm6(pT2uI*c=R->u~N#o~JS}tqYtVTglvW}FhfGVI0r~;}$
zNl@VCR$uMk?!K)~wy@eA3mY`pzRC%UC1DNd82_uly{*@DyyqXknz(YrlaJiC`O9(z
z%5Zh~?+VcWm1G4vzjoX8)9+sCgv~B<FR1=Ok0K{(Mm=|N`+d$mdsl7IcGOZ67QZ>;
z<h}PBHq(T~sOLk*59>YgGHz1rDxaNN`GbNLPS^}<{6`+MdhesVmMusc<NAImuX)Yt
zw7qVN)f-IMO!Yl>>HhnqVfB~p_l`!@U!VJ`H>*0V$VnQ#{+*Pt{leM@y~auPcSmt)
z@Adt`ar0X~>Ve%KpI4?JReGK8RQ<wU?;7uftInx0&$#K3uIEKkT_4)MWGT>i@H=-r
zb<Q*=Dkk;s7`f#g)hc=El&a6YbL%P-cSbdS88=<~X4w5ibv<s2D|L&Do$CFT5$8>K
zfBd_foTM^opPR_{wUyoeRn&BYlT=1^K1)~JRPUQ4y}jJOzIe0i)V@V)Ow_DCxcgOQ
zuit3WMLEuN+2_=KYS?ZE&U0YY{SK<%=){^W%M|!!)Om#+-|cf?-0QFIan0s-EoV8(
z^y(1SedkQYv0axYr9ABG|G32qd!LIsUQ~Zy$rd-M*SX~F_3wg(J#VRAXWHf?Ddp%H
z_VG?!zs%I{d!p*s?R@4}pLF{tWgVwG&veg2(Efs6uLtc<+htaPpWd5y_s|!%nWWP1
z{SmKfr}};dul=Z>DiDPNsjgpbUs(UN&1)RhL3P;GI*8(!)tIV)DxeC)Ndd3V)5q2C
ztJCf~byDqO6BXk$SDH)a6iE8tNz(jN`+HZJTY1UqxzT`q=g#c7-6&NDT?cv`=sw_Y
zA4t}2uh~ikQhh#L+pjJ)rQ<{=RUOjxylYhb_<hbyu0v`2{T@yI!+kzGuJ=zeotIpv
zYO4aOfGVI0r~;~hDxeCe0;+&2pbDr0s(>n>3ZzGY{>^^xC%=1s=e&Bkb#m+GmhO|^
zK0mk2Z~x|1$*Y!EBezuV{JfgK+4XZv=az2Msm_D>c`x^ExvX~SQt$7*_U9v}O*DRa
z;g&MR=U4d1p&fRIM~h~)pE1+Hwo$hG@QvBu)OmZj$rNpQ%JL7MT6z8%W4D>Y&l-R5
z@OR}4p6j~yrS<zR`XlePTbA^zP-e~_ddl|Lj30mEP>+uIx$&Y(8?S1%V4Xv|8ygiJ
z^w4w5oGP`h?|$3SYaDFq=|_@!^1)`@E#4T~{<7=Wo0%6ZZ8Ndu+W+BRFL(X*!seTd
zEiU7ro_?fFIsR^>y^V}tlDM>EypVE8xn$$k>pCp!IR1}yFYBdEzcL@j?UyH9aoy(m
z?=3g9BR=yrV94qY_dQjyz&39(5A3lgbRY9vnT3C>A9SD}Y)HS79M`hk)=rkwA39PG
zDVOoFX)kp`57sxF`&^@U+LkFctmopdx|Ml)N0|cb`rmwf!N8`a3MlXO=-Z{8=YBCz
z_D5XWOPg{?|McVL&pb08bitOi=|KMw{UjM5amf#o4}MUtE9Q-Ssh7l$a{T4`q~0s#
zzR7Xy#;Jwpc6_f@^@76R&jU7+HzaRtNIdrI5o4M(ssGMC1?9&-{lO<a_b#BFB<*E6
z<7eKm85ix)51V<S9i(2i$DeZYz+aMjbcZqz+Trg;)}b5wnKiQxy`^S>jp?~;(_W%d
zfOW$<z?SQna@)G19<okZ?~rmyj$``8A5u=9*ifD?Ild^zCO^g@ZOSFdgX>T7NR{+M
z9zn@|Wg~gZdZ`O~LdK0vzMKy!my~hoSI)b%$y3I&+2p}_ROU-OSx!8z8yHunh(3^Y
z!1^M8X)_MSiH#rECD;%@;zGucjUS{Q+WaVoGC#^~BtGr1A?>J_`N{QwKg1u}uJ>7A
zw1em<Nj~JsJVMIZmmqPt4x}7^^2Fx&!iM-kSx)}M$4?SJDAy6TB>hRs<DC6omb3mL
z^)f#hA3e}n5`W@gqXV{$ws<yww{r4;=nN@`=mCk}v(5bjZaS)JfsOd3i|9n2jE8dO
zoqQ<AUzU5d(MkFfSCTx!BI{I=e7us+!;pOeq9^+dHpd}0dSJ`_koLq4o6IZwC%P~X
z*pPBb;xJFtOIyaF9lCRz(;k0FJp5@-J>+v2`wgTWq(9=&FSaCcX%F#}?Xe;GVN2TT
zZSnn<%Y11Msb^i14<ujw{gU>O_SiPs{3s_sN$QEq{80{B9~_5L2W;X%*$&-^i%nd4
z{IVZVjvvP(He^4*hU7<ojFUViS@*IWo4B_A=m$Tphp-{zV0@7FHqu_!Q*I;vvYvh+
z^)`R%@#pggwyc-)C~exwa&#lUyf4c605U%ieOZT$NA^SitV_9GXeWsu?I3=#UamWN
zJlbs9QO`b4e~|v5EXRh@k8;R3AoGfiZrIdIQcr(25)aCGXa1Q#`eB~2=?~iaqn!1@
zxGA@_$6xx92b6yJL;AB(`mql1r=ER?a!5TyPlzA;9ez;ePd_$YWE}h;_1N^oyfN?8
zQ;v<U<n2b{(=YK@M^TaE6tW*-M@{-;{OE*^=)iGAxg>hCzjFVUd6ID`m!w_PBv0;x
zNJia%+~PCuaNZZ~e%h;ZxdJ|yF@GE%><jEOtOwQ+Htq0dolwsF6IZTB;>rDm_VmN~
z-z}bOM|>EUd0`$HH{)bpAi6=uA2#V9qN8m-W&ik7&iR3INPNc6yh!3lIsP`{$NIvL
zICA|kF7`e0BaV&i*L?3U+jAb_xS}8OAg-Ku@`d<8$|1)k^)}+qeke))<j44kNB(kN
zr9bWHkGQmh%maEu{2_kW^haKj=n08Szu2-J`AW)lM7?Z}O~3frh`%j9?cD0|^Go`t
zoz3R>mGh6kB=bW%@~0e|bprYPLOlE>sV8srW`5}xGT+3J<(#irUy{V7KaOkGCHc?~
zHvWtkn|P4^<nc&+;?nQE@Bi5N*DbsM@%$G13;m)e{X+D`#?Osj{cV2eZtDkK{N~5J
z)1Lkz<&bjnz=oVx(VO*7{^*OJESLR|AO6%s@~{y-sK+K>H_{KpA6wF`Uiza4L^swm
z`atO?<4N1?`i$cT(l7HyKGbuZV9R=0&OQO<@rk{6!Ba=oe75u-*U#hu3x9tu4BL(5
z!SN2c&hSgxL*g_3REgeD#+UPt&3t1+^1)`^a6Dnt9vgp2jvqIYw<O0!rYXlI$Bzz_
zL&|B7zMNk;|4IK;$7?%&<vNji%XZ{Jev*tYYRWuVk1`+WM_%$e9X}gcuhe4`kNL!g
zoNvevQZMuMYLf?)?dXSk`oqS*(B(N<t^d#ONz$bQ>k2YoUP-&CNFFji>x*^8x~JT>
z9;t`451Z_3@_f(vo9hOyBiNV8i}tcy9{1Ri#Icd<0UN!pKU;gwD~zA(FUlD=Bn~!|
z<;0Qo_{n<w&|Q*wXMbkB*=X~p+}4iu%Y2{<`Ei}e`4K<JdZwSCw8bHBTU_#>o;a*K
z>g9UHhK!Rqka*+~l;q);w*K7W$npEN88`jN_Ocv*bb}|pe(lh1_4oSY_aXSPk8r)m
zJkSsA=?8y^KQ?6jLXJO4`XN63LGq*?%6b_G8?ygn6NmLC$3fnZ<Bs;yUzSUod?D)z
z+eY>UTfNtE#_Q&f-ZnqmK1iN&T-XrZAUctsv?(VaY#Z@o9Qa{V52+^~<{45iNqfIU
z2ijqC+(>=#^Ll@Warz~?lZTBQ59}A%_+!g@$|3%Y526S2fxnHMA8Ch9tVc<yFZHx1
z54Sk9lSD`2K;khkvYoW?Cof6jG9GLw^=181583~)C2iv&p3DoIIOM^6QjQ;SC}%vj
zdi;pTaf;2lBX1d>xWth*^FTfGg)QS@6GygZJTf2J5g%KUej)u}L*h^mnO|&4+C%EG
z$p_ok591D-<ReKQ@;D)1>CZf*njifVhkmdn(T(d5NIC8K9WOSN@fZj7P{zZS?Xl4f
zn>ZXt?6-U#!{+#s#~tgN^@dG6=7n;IKlPCOrA>LNWW2;d2OH6gcH||=xEMG4AM=hs
zHtUJF=!z~f9yWP1AH*X*{o_x$8}WzaXKP0uvR>vxJSg*{9HJZJ!<OyPOV;D(mGmq9
ziOYOo<41n9hqm!couqABf2<E1-P+meS=ZDvK3hMO<1gD&58eEz$DjQTT`0E^y;-Nk
zgEAg*sV`|n5A?Jhzvzixay{GDJ@Ke#J#qek96#6`U(Bcco=5F?DZo5I=2MdSm-{s3
zP+n*7xlU5XXWdZFdV<8KfBdjn-;@)N^^PBL@Mql8roAjDp6rJ><O}7v>7RV?qn!KN
zHgaE^`){^9$VcW+Imfq+>?@G{27hegQ*NX5!<LkK#p`I%Y*JN#^~(BZJyH%Sm*lv_
zkK@PfJjL;4BXK3=`lMghIsU|@p1knm`h??&^D>nE(2n@j6Q6p@vC#`#)>CdH<FK`p
z<>Vtt9P)wqVcUqGq}%#q9q_#?`z3MElldV}DEA+^PRW;Y>Cbcgyx)ud>5uax?WiyO
z{dtL@AMf=*g+{v-u)i}7>iJy@e&kI)<i|W=<3}F&W0ME{GLEy(IQ8@{TdEf@9_q=T
z@vQ#grU#n*T=kFlqha%XRqa>L?6dl@Dt~-$Wy_Dako=fG=9h7?jybNdU#qtA#JO+R
zDv)$rKXIvd+ZQ<hLdMTN11S$m;xk_|FZ71^yU}kx%s1;m+OnMWPTr8~H%Q))yjXXT
z_K^N*M?C5!nUA=Vag!gmB;$pw2W&~!fh@<C@ukgu30W>}sdu*6>_^Ot@;~oA4yps|
zhW6O3V~%U+bsfp}>>I>ke}e2!Q2Mc-aGa2bw8=jx(IKgH8#j7Gu6wxNL4WGeo%vuM
zxgOzqhk0OsV}7{4q&?;MlP`L+ZivJ64cA#*NAX@*eiuZXYu|ad{Oz6gD4-qJhmwrX
zmKWvHk8;LgBYw0eA9Q9s=#TEiqrD`0<1h7<{@Cb)&eEn_5<l|7X8xEj>gf+sF8#1s
zU&Mp-17%+L$#QAS_SllL9qW&N<TzwGwk;0*5eI+DWjk!n_t;)Ze2C8IPQS#*#vgj^
z&#NE0N`Lx?<R$$wWs_IX{%P+<beELthWVr%oBhT{xBl6e<UYgkEJ^>AV|yjzl>V$!
zN%~2Z^oMTPkbbeflJP*+JGx^N4?juSKlQe9>LKx=EXO7vY~quz?KmVZezHCF5S^Jf
z%AxF+crriwgZL4La><~_t*ss7pdXtb<@k}WB=wT`F%Q%e*NuMtWjy+o<u+UPM|)W=
zZQ?OrNVyw{XKP0u#KE6+M?1=~p)6;Ar5?(4NPm*7AL2l+>mb*S*!W3u9Z5OY>9mu?
zpK`7vA#oTF<C8=`N&JW-$$sFM<imV0j-(QuY-GGaiB58Uux<TP&N_vZvu?<Xa%@S~
zaZnPUyls@nnVZeHWWLyAFZjp5Z{K_MAAkQx?jtfj_2f%E^B_sRJkQY{e`%ASB;$e9
z%l7n}F4BM0{h=3>{ZY>PK_B)fbdzMC!e7#BJo-To=AFE3`BIJ_WW8bI7dFF=j}GXF
z9@sXLhoo)1($B3Q>g9Y>PF}1Jn@zbS<EB05TWtLCgOsz*WIojUC2=IfiX-D<GcVYZ
ztZU|#;{-q2L&_oLl8g&KDE-h`mg7$xSubt;WjS_K-)E2?`YBWUo+MrRvCf%4$bQ7S
z@tP0oSRPNbr=GZw{@8!mr?AP3^)2@=>KPB$i;#BY0qLK9h|6^&*ZIUJ9)7gP=5rAF
zQ%`&RsK<u*k+0N|c;qk1_+`1Y>7Vg1PHg6byde46@}QplvAyOc^TZ||{X*&?<@AS5
zy<gIvc(g|+$oj!%9biMsSzpY9Y>&<NH$0Cj<H&e??^O8v^Gtj{1$nLrJ(+iG=1USk
z@|FJNMc%Z>pY_D?PkV?zwk;lhlC<+0AAcz0vECUs<HBY<w1>1~+_t#*v7X$>`lDaw
zo%r+*iO>3##838%ANvlr^ru{sJSoQ}PsljXmHDQg{IMZ<VB6*!KgfI#7t$UQSCTlC
z(;l1gU=vTaLl?@ulDN{Ja`J(cOUm}lkJobPXR}#X)H7~L>LKm$qntdcXMIA-!zTTb
zkBm>br1Z1dvK=<#w#^^q%s25VCk`YY#2-@5@rbUnJ;y2MFZze{4~Z*n%H?rFxf?m2
zrJtNH@@JejN<VBU+fgnl^O3gHfpJid{^%*^LFPd{{mAhXhdlA4oPMQEIh6I-HZl$<
z<H&Zb12;d~;Rh*)loOY6V^a^wll9HKaGs?cKl;Judu-`PJ^qsPFXxARIj<03=0ksO
z<hq4?h)e(EWs5^Q=|?$nAb!+iL-K*LoN@A80N1VTJJdt+<~o`C9LkcV0PBEt!uo=&
z7uE?j>x}0NIo?>mkadbJNjzCjJ4ih?<oJP<<Im?o_A$u*0r6*?lrz4<-#;JXN8ZHe
zxksLN#1G2zP+V-r8CE{%%ea{jbb_3RgZ9h$hV{#QunzhBjvsljP9WzEh#uGwf5`Z3
z#E*W-gZ#*w_SpE-p8jPy$1CHe9)H?PQV*Fg+B1F|Wjk!?PdOwH@@2e?ACd=p(GF5C
z{jq6>jX&ks<U!u#iOu;QfAl7<B>6$&V9WT_L#{)mA2x9y<&wm~Pv*%u7(ep>sSkQw
z(qGW_NpHt|Fc0Y(Pyepp|8mL08;p(YJCOOek@diQLaqzg$GOg=J}5cPShtY={4!`B
zskXD}8C5ygE0U~3N%nhLj^6A?lK4|DZPvNlx|i*U$NaNT;|FbdQ=ctFM~**iH=?7B
zvc2@<dQ6sMGq2c?`Qo_YIOV#Ta>gUq8F>*8%K4PX5Ah)N^arUYADbWL__4o{FC;GO
zRbD4hZX<D6&$jmXL8%AjHrnEn7wxc_H_9RPa@^EYZX<DtBimuK{v=rstP>m2hj`=q
z-1cUb$LbZhk$IpUWW5uYJS3?PO2)&uW&DzAbKF9XPv!+(SXby6H*#EX9FRA8GA{fm
zw^6pY`B5(Y-Sm?F=*{}39@|Fz$b<EbO}$sjcGzCYIOvad(x%)mWgKki77u-F`w=>`
zPMAl^*|#9|96!XxhTO-)wvo7!Qdh=HUXrpsdGWcE{_!Io<h&+HKa3kv4vB-_*s`8-
z@?)IRk8)do_*2jI5@bHH88_`%H{=7QpNuc}8}y<-{D}i8hg=Uq%58BdhkV{V_sxfA
zZ5p-bAMewlo_?{Jf7(%x9@ylAO?yf9SJoHr_2K>?*Ll1*i2Jw1CmuHAXCBa%JjkDR
zkol#ayr_rrb6nUq{kjqTCCOis{AIbcnMd@4l%p^4u+f=vN%Uo0v?DIZF>#nL`VUIt
z*hn5xlYIg*ZpLM^DYub$LCO9IgT{%f9qWbl6IDNMahWHtBwyyutsVMHKhA^XCvCTK
z>SbKYp{&QI-Y*#s$2IY!9&$YRLEcm1c6@St<0tnK>Ltlf<|WH%had4F^$<TvbYXpv
zuZ`@ZTrXhT{AK^>NWbKRAN7<=%65!behW_=xlSmT<o$n9QT7*=&3r)4r?&aAm9rkW
z-e<mT<a&qxY~(r$e;aLivH#lIc`Zkue}3^^pPD%b6fhrhy#%${M_9kC+pzobnjhD#
zUj2jm$?-`WU5Jk^*!ZIxHh$PpwxirDX-{6XqZ}KO7x}PnQ4VD}?TJf&*wEIVa?S%X
zzO1J|NIiLDQ!hz9c|qxiO+BO>QZDmn9PHy9_xQ{9=tjAXw1+l7#tG5K=1;y5KWySq
z4sGM6o_3Pt5AnyAl<jS{Z1=ZThaI@&k{Sif1D_*hf7rxj+{B|CQqR1ye{+7IoPM!s
z$9ahTntICV2ck1J`Lb?^Lws!FG5?(B@N*;m;zwM@i7m%ZJ)~ad&p5~ry&?73v_~)U
zl7864#U>8AVMF{N<G0mQE{QJWDeG-E<0EhU$Vb}rL!J=bi6@C}l+%uQl*6O0nBMJ?
z+m|}@AC!y>k_QYL7k$VVTgF2d#!DWwBM(__i^n>mz3h*A+7lmAFWX_$&dm>h$)N3I
zoUm=i&-ge_Bw3F(k{9uq2S|Ku;xoUHebGj)Gug+u9%ntta_T9U<bEReV{PNJ#pnI<
z^g|r_v6WNL=O6SyKS;lvKV*LBL_GW@$xF7UUr0aL_>&LiHnI*_@65X-@yH(?WPjL<
z2R)fLIc{v;o6qlQ=?Bt3>rs+%kT-cyPJ2kXG8ykp(0S5%YPMiq%GXg*tt0jw$bL$H
zl>4QvJ>}4ET$?}fpxkG%*<U%YP|p4e@gp9T{i7G3`)ChkeBwg4^923LIM_DQ4&p~%
zkUY?XabR;k<a}+5M||`^FRr)fkLwu9xh^L@{X+5~9{pnzSL#SSnLqxL_{sCW%!7RC
zpM1DpL{F{{AlDuAi(c5ofz;C<anXZvsW<V+8=JUpWLyw`Y#aGIM?6QseHZ+_#$jEM
zH#*QhD4Ad45g$FVsh31g)(19z%m+5{ApVf~B;Mt%J9l04WVHgWBRNi3KlriE@T1)s
zpFPs)={F86nAH4*o2ITmpn&fK=#OziuG8glLSE>FjXyTm#T<9ow1c!`J&+H?k2sXm
z4@3u!qr%^RmxwOZL)sCKerQjB(hq;i<$O|3JX`;=oPPP;2<rmUFQgwnU&!@<-t@z9
z*R#$218zF1Y60t_@OPb9H~7<@zk|;2IML5W@`L1$p2UZY1JX{CII<j@dPun>>lr`R
zJM##shqiKA&weJ4clKfSC*tAnM%uCe(Vp?4FY_q%Bu`2F(1m$*%O}-(@|STaXS}Ry
z^2Q%O?4U$HNpz9NF*;Mv`U%><tsUcs^h=)f56P2!@S_|u&$1mh?a&v0Y~}$Qay+w-
z5f73-e)K~;Z2YlhJ>}>^J$X=W)74fkk1H83Xk1%=loJ=qa_m&e{*!7PZ6EIa3f3>{
zn&X7yK$7E8wxd1eP?k$u`mqk>`e*%6E-Bkl@3vlO=jP|No_@WO<DNVq<E5T>)XRK$
zj+=7nkIuBG9#YSBgX~xO;ZHgHJ~s7ozVV}8(k(yfkIgzIKK)|LyeP*H5)U0E@uNR!
zb9_+G@x*zD^PnWxJ%ztt56d{%^v87;aVUrM59uHBxrg&5?HE6LQ4Yz&jm)DYaUuON
zALQ$o=#VPqeA{d}E}LJv%h@+%ew1@uqc`Qu2R7{~XP(#}D2J5eM?cgPk9t{-Eosva
zKN*iasHZ(P{`ip>#2+#aY>2-k^^ks8mmEhjALc{);ZNT9aeT0Ea6C(!c#!=lRWfeI
zfeqPL&;dWjfgj`J{tNpW_4q^f9okVYiB9<0h>nmr_+!ibu_1os$NdZTTgd!C`lUbS
zh53T`$^NJ(p0v@^mJfMB;y~g-@?gEu4l<9jKWP)6cG!~iOF8RG#>0lhg?x^;)#F$A
z`}r3?`ePkSvQ8<-mZYBjm3f5hGuV(gwtC`1?z>{AO8SSy!zM5GIm*c&(k>`DzR*v`
zp<nW$9)I#6jwJg9elnh|96uZ7?<KJ5U+QLy<5tdi-RKs_Ydv|R3*$k5^duk3@uwf+
zFi*t6mj1+}T(+m4{;=tfc9dh|&wj%`M?UP&_|q@-ltbd-M>+kH59bSP{A?r-@{^<<
zQcs?gW6OTXn|z@3!^WR@Q1*i@^I@H!1ND%2oNp+H#Kn)iY&QDi2k~b-l+%xmtQYd<
zJR<!$e^E}mpp^514v>5p7esIRB_HZ(&-fwrP_AqIhzIe<hV%m&5BU*~_FPwBL*g+W
z#!WeKu<>I)u&E~x)(JLo<haR0Qntg!kN7sS|Jmv(C$5e7*+@I;(F3BBB<tLbwBz_?
z{OE}dSr5deo$b1q_>ee|aycHZ?}*QJ9;AQP9X6z#IF!?$v?+)5Ltc<^;>S49i+S=(
zjw`?Ig8H+r7!NkI%`@frF@9`xfXp9y@p-54_v?7B`_Y$qqaSFCPe05jao8_t56M@O
z_~^;;O*<&#Vnh5O{m6RqM1SfjXP<=B6BirO4xK5N#1B$0^Q4@0iEV2~Ui2^hXwP_=
zFZ^kbAGDPdAL36uY{ms;Ipd=r;$h>@I-_0^e_4)A9Q?2$<xq};aybtCsF&-7_>g?i
zUy}9Daf%%mvR|Z{4?1$(Lv-{?bj2V2u_5bB5}hS^9tRfwzF)eL`Q&*zerHZQ)(d`=
zvo6?wS?}nA&H42Gz1RMH#I%Wq@l%gK`f|O_eo8r{J@L4H!=HI34)G!Nv?C7rQ13?K
zLHx;sa`MKOBu~Z#$&2<j;s+TA{>&5ok|$(dvB@8s_UyyNk=HwvL(b#a#KVS^lRu;$
zvfjxDotQ_;$y1JB){`IWn|k7shdlnsm%OmijdJFdddeZ?<jFXwXCBE1e@H#;$rpdw
zKl=>vi9_D>52^P`;>dozwh!BnaYDug(Sv!93z=8)hLl6*oq9?9Bxxtf`mqsz8IOKh
zchsXV?V!||c98yQkIivN9Lj0OyizYI{i&xNc|+>m$oWtD(I5LH`wZhk2g)UBkG@bI
zU*yX;$b<Imhv-B-B(9C}xWrGEV^dFFkp3X~Qcu2+a!CHvvmU62^v`|-sfWa49wg;_
z$a-wz;Lm(h&v+s82g!$V*=*V~9`r*8>WPDmA8{c%kw5Jq<KX(hHXha&>zelTPe1g_
z^&r=q#OJz`@4F5-b!XA#qxLP>IOYD1BVRo14-%ieA^l-P^2Ub5p&rV5sjr+j#s$%v
ze6S(q^baW~57rZUVWT&BLUhCrQcpiN%K4M+7`Ge;`ryxX72`k`$hwkwkT3m^5Aj(?
z>@$8z-tzcCck;uA<SA`g?$#cEuOyyV%KXqxmeW7;BZ)q=^Gfn`^CvIJypk8|2-_`Q
zSoQP^=|3nr4rotYNpxc#I1aEO<3u;&QqFOMp2WqE{ek@io8yr<ltcQLWE{k2UxL&_
z`jPpOC*xuq<U>9E6Oa7?Tk1wR`OzL5(hvUVLww?5;}6LXGCs(-s3#6Okq0CXY}#Yf
zuN)8Uh|f6ir#<@8KjR}F?eUkS9~<!}4*fBHbiu}-c;t=EzRiA5I~fO?@wlyD`XL^q
zU-F_H9T^vT;ScE_f8yYe?$l%BM>+k<eFQ)HBQ7K!w#<k8@x!KHY}(Tw{;Wsx!w;f6
zc~edv)MHamIplK{WW2;BFUn;+;xf;)<M)chr9bX3N#aL&;qUiV=}*RIpQ9dKS)b&E
ze$-3dDVOmWuOxZ6k@vFm`(12)C(3m>`J-1<WFF~Hl5sMB=m)7M4!WVQB=z{q_LMUp
z*tBCmqMrF7A9<X}_UvQCll2^5vRsaXJY>1FS$A%v9sNS`bfe5i_Q$yC2jWLPq#QpR
zWnRn+^9u1J4z?U8^>QBYlO&Ff#Iuoh#Kn(tnTM=rypVN>E&b_-b&5V5U)Yd1<U=_m
z4sq#^c$Cu~oA%gpocKfb1=)`FlIQ?w=SJ>B;K%tv#+NqvMMc@4Uz__X5M4RW<aHYJ
zDEpD^I9?(3w1+Yd`a<T3b;G_&J#ip;(=WCp^^kgONIB$lF!j==+(zP%uO#~=dO@iZ
z<-~*dQ4eK5l;h8FMmaY1<OQk6U&bdN%H=w;wZq>lZE?t3*2}zQJ$W;K*plRDE0^QN
z-;Kn>51TxROF4ejV^c3ld;B2tip_o{uNyfYIsT-L?vnUJ+Hs!1hKxs&dfLf&)U!V`
z4s1xdB>N2I#G^mXd#n#h^n>WiIHbSY;r`A&RXy3y<#B`U7T@Maxs9wBxA;-l%lzn%
z@^s}%fAk-Bq6>LL%Gr0Jthd=g$8Foc*e}`VxPK9}A9=j^jl=#*zL0%do`1-fa>m2?
zhH@E~>qg?+NPJ1Z<KjAlcFZp}x{*KU2gtgS=MU-`FSaE0#G#&YNV!)sE}K7bZDhQV
zI5vM-&%8p$M_$;F_BQg~G3hV!qdon~eDP=fvA*!99#Rj9Pae!W@n|RGkq6f$_)!ii
zw-G;xA8}<l^<K$1A>#;2bYedx58|PxB<(mJ@h2}y*<R{_E|5HA9BleyKf<OS5{G>d
z(oayz@z`vxcZ2q0YsdH`(ZMTi@!iU8<8>=1FX_iTLAU<-U9eIWPzC<00M|uwU*@=j
z9A6x7kaCVsxzDq2<ByG>Y==!glH}``e&cYR=(iu6Km8LI+vblx=*H&;c|PKLMfwq!
z`GH)op)=*!kaFUZFC;(6d6o8*^Lro4A?u3s9sZF1vFVp~*yP7~jqitPM}OGp0$Gnz
z7ur!zzw`&m6VeVcKiK3!{*ZF%$8{Eze)!W5c|q#Q4^j>(hs+DQP%cS7=p*ZigCF(S
z)U&?HM-qQX97vwb7wu&|wymD{Q2NmxQqOToJM@rbzR8#N&^Eu+(=Q~xtjDGw;!3*x
zomk@ANFFw#hmGWCqb)AwjMGN9=OoxSY-C+Z+Vr5@7EhK>+cKhehZlcxZX0*;jH~NT
z`Gea!Ki2rN%cnUu^7)i~%tqE5_1I9Jk60g4Uu=k<)SY_Du_ZZ9Wx2G8BPsoCwrpp!
zX)o*9SE*+m%ls*Kql|+sDdSKt%cae6B*&4eP5-ty)Jw8X(2@3(yODKj^ONUA{Agz*
ze$eJ;E0^PxHgO@xpCs||lQ!kdGxJNi>{r_8O?&FG<$i??r629-pL+U9m5h&mh(~{r
za%d|@Pq%n-ec}&gKlBG_kBvV>PwL&)Eq>%bcvO`NuUFZ(fWL1`d+KeJaj?nDM&^NW
zNk94{9yWQ|);aa~_iy(5XVLub`JMCX<<`xums`3|e*65~GQa(sS0%4nUX9#Rz4P;G
z{$|(DEuCAsO{Y2!=I6cKx8<_hrT_a$R{y68r~;}$!WEc0v2)phjmJ64%<7QvWu)n=
z0;+&2@P`5~wHeXqr(<4ql=0BvkBQO$#7cp6haK_V-=19Tz~>L?RsG|m<~fIdQNR9@
z^A;#ohpbo!u{!898C5_PPz6*0RX`O`1yq69D6ogQ;j%@8mzbcuug>V)p?5DbC5^F}
zKutsy2tk3T%3RoZ(zNn9ZmjU`{i}}dKgNW`@gLtmwNn2bru5mpj~j8(*3cM&BpN~$
zPz6+hL@97umBkf)SvbQC?|#zye}6gskL$83lP{byVAynLzwHgbDmVL!|6$j)|EisP
z^gQF0`5kKBQgi$=6E%Zgm)hC|y-q!NV?p26TUHnwxAr}t`P~;RFqz`$dxtk0v+N5K
zHp4p4hZX0?xBDH~WY5{g#<0ilHD1FS4V>-Ajx*^Z$9LQr)^kp^E57^7&Wf|0pv=^9
znXUasz3<00exfc0G+C<xw)02W<rkgPXhn;!e>IXlhR&+>NRurlD6Ip_vWl>uXUZ~p
zti$Q^!O&IZddzp?PU`@)LKR4`0%4y=!itw*Nj3E}D)4g6edZURG1SDBm7be3?cDR`
zIAO8pantTD?!H^j;=F>7XC1f135zvvn{x9>rI(mg`R%&>rhWg^ViOjxer3<0ZHA3D
z>0%m}u6Csgr~;~hDxeCe0;+&2pbDr0s(>nxK?S^ij~>)-Q^#2cbvWl$6BTXG&1m_q
zpC?MHE*U&Tv};vB6;K6K0aZX1Pz6eq0&#ziHP!s)Ts<S_(jBF8{IWzX0v)U>pbBJ#
z0$<(T<h|Eh6&vXFd2X-IIVJV}0(lR;=Bpv&`i*fUQ$5dQd;6?7ICe3Jwfp|9@<kOl
z8900X+=c5tFES;S94ArnEoWKb71vERHv0V?m*3yF+3r(&-6!$0{hslvJ0B{pGjF|#
zD{a5e3|j8>cdME5vz^z1jyLRfN#AD%_v`Y~U5BnV%0KI{i&KpjsRF8iDxeCe0;+&2
zkZ1*Jec0^AJ$_tg-Z^W|eWz^PVCBP0j?P_o&vxUNZr8<pUUwt+!C<pR4?f%Npv}g{
z)AyfQF#Y;%#*M_2O!TB`@~VI;pbDr0s(>n>3aA3AK$;Y|pw;nBt~h3{QN~k;G%W+|
zMio#6RDsYH7=7b`ZH^rOh>4or%U=9Usho*UxhFTSo_G49|6%QxMdR-J@)L9S^WXJo
zd-a5rx#zZ}osQaLzEP?Uaaad#2kz+SFM6W)^JQ|B|IxusFRfPvvQdF1SN0oH{O)q&
zm44rs=J)-z`+j>y%`=`^=1AVw`}0b3N4{=SrF^b2KmV22Z)iW$q{_)h_L%T|r*$T3
zKJ@C4%L}?MH(~Lua+eOMyllQn7q4tysmjL5^GvpI(cDSZ9%`}Nghjo6^mhGt@<E%v
zX<K2riHg~{*JtyXc2?M}&0}XTHA>atw{?)sYhA~t3aA3AfGVI01XVyjFX`2GdmdEQ
z^L(m}ss29C7S~qJ?*@~~RP{<)d)x0jGgWT;9i!j!O#KeCr1cwgf7EuWfGVI0r~;~h
zDxeCe0;+&2pbDr0s(>n>3aA3AKuJ)bYvH`-hRvGhK)2_%%!i+@t$Nrb6O>W^E+8rK
zOTyYp!WcjKYRr3Aww!ND3X?D=|2N0|yNv#9+1zO+C@*>Xv99eOndF4UEiYD|b84Rz
z4*cqn8$LSxxI3J%dGDod?%1u&G?P>|S+~dQFAU6glEze@6OT*#4SkQi_R(8rn56RC
zqh2dtYv1W6sVp8n=;|pmCY!kM|L&>T^#5C$E<CYX!3vWq^B%nGqn&vR9r*kqy{dnF
z)I0~u_t;1E-foSX%)hsoO~+I=kL9e^mzF#I%qd2xI_NsE9|zgI2k6+`75MOtC-&^}
z>IUP+21gzCzyYH+8|AKbxbc(Cb;^Ff*eG|cgZoNU4^==F@T<V1wcbASgvFao*p%;A
z$Tr)r&g!oUr~<(i81vanKizlqN|Wsju8_9OW(7vyJG|MLWnY+VqwViMW?T8zz6UhF
z`+@~V=}m`hK9+TCs(>n>3aA3AfGVI0r~;~hn*ynR?sJPz{hX_3<XpO=RF2Y{4$oY%
zXXEu3ZgrHZLmbxu$F))wPz6*0RX`O`1xk(r!@Hlf{@-6tH(u%Y?{v8N*RJ^PFFPyF
zcCw9eSwY$IeNtTX)I3xHRUmu?s!YCc%79_h9lzYe+;G{V!Ap!XGdhI75;S6*74ZB1
zYW(9orfkUVrma`auKRhBW8=iHdpy4Q;MGpLXdbG2b+56@95;rQ&zXmhzx|_4E1a;%
z@3obG{{6b@ky$;mVXx9Lr~;~hDxeCe0;+&2pbDr0Q7a(--kh|{Ozib^!xwkYnL4p^
z*@2D6IZD+**MS}fx)1zkABcK?4lDjIFP3|^S>9wPX$))p8b=jS1yli5AT<Tto=5Zh
zoI+gvk6m-exo5wz#8Ijaf3*%$JDU8r=D7hkruw@#ZJ(X*jR|XB{jXeY5hn$%y7Qsp
zI`h_>q%uxgYA%^qU~s=KFWq(M>i>D(^}afzbBEr&$S73@T?hYr9Ay5;^_RiGx)tB)
z+xBZG)AX0MdaDA_D$u{#@A31y=XcJlms>CQh}_bB^4sU<mig`9yefIs@@nLk>Yblg
z^EbPGZt2|8Z93I?FhB3*zAcy4E*))l8eJ7o1yq4FDv<l$#V`Eu+%%)itPW|Ma_veL
zPz6+hNEPVV{p|U#%~@oW@z5dC$=2Y36<GeY^X`mm)*JZ!vCXRO*mIh5)|h=>c;}f>
zO4T7N)<NI{L>sCCs(>n>3aA3AfGVI0WVHg{AAkHoUp~9Y3Can*n-mRsugEEB%<4s<
zgDY_g+<5E4XL6tX#kld;X%{puuD`|!i^m=~XZzI`>{(Fbt`W<Q98)GVmbeMj0jmP4
zK#5o2`YtWLxvIqshx^*~%N>2)x!25hhB+;pobdj9)33(Aetl`{?Ehiqm(S=v{)X9(
zR|ehxzr3O4*gAzvolG(4b?S;Krf1t*HaRv1y?zd@-S6LHFz;cJ^%FI{uBUQ!+Lm3@
zVw+>*No!lRf1vUfC*2IXu7b8Z^Oa8v7q?pP#GRWi{^y8&w{AD7((Ss`t)B1uGR^I8
zUGUVNdo43bXQuY=O!r&jFEFx1fUwSwu5pGI4tRP~`GW5SxgHD3ESUz^VTAo$hHHme
z;xMk<?&Z##zS@a9vqWwkhAQwU1>$;r_opP)Pz5XsT=Lr4clRq^;l!1dR&T%Zx^JgA
zVezX5_iX>T?=~}b)1e29UobW_-rxMro%@_u?4-&vBd&hD)=Qh5uo(2dL(q1XWv0cd
zfGVI0r~;~hDxeCe0;+&2pbDr0S*AeTzo$-DeglS|`04?3HaN<7>X2oJoeoD8Pz6*0
zRX`O`1yq69Dv<5(Et*j3qQn0-qEdlh#&*IqF;zenPz5XsRGfd@BmI|_&VgR<cX<6B
zh}YjK#nsRD+_2a2msCH?a!Y)%e6A+6--V_4euqru5%v3#`+j`Ykh4Dc#@JY;*M=dF
zeDtn~inizeZTo`Xa<BVsru@SGJ)vKH((P|s&td<*A!xiz|E?<4{=;5R*%B}DCq^BB
zDxeCe0;+&2pbDr0LV@#+IIr^ig~g_9<?+Y#XtKr1*H4|Xc>ZPEjHFvT;@fz_PowYO
ztJFs0#uw*YcG|6{tTfQA-&JexxU<UE(gkk3>e<sq{<BlL0!bm0nyP>*pbDr0s(>n>
z3aA3AfGV&H1<soMY5x9WCK+Wsb=bv}X^|?R3aA2!P+(Pu8WVpRy4;DHk8~Te^zC~-
zH4R&Qzj;;N8UI6`i%Ru-SK3cnx!;v{oxRCXst$2j2X32z+x^#)R_~^l)~f>9sDSPH
zKU?{8gN9FfeSDGeO568Fw(=I0N}u2A$+88KVZFCp{?2B%)6O`txcLN=D!C4bnq0Sp
z#i*X|w8gcptE85P{TyY`=NC&_yQrW4&&Itzo5$3l)wzcry==XsR2_a>2id&Vb!@7D
zDxeCe0;)hz1%iHF@cTK<?Q?adA3nUa%6V@(LFxB*N@0C&&QzSZ{yl=I^2*fjGj09K
z-zlrzq-e-{MNV9G2)aLNJ5@jxPz6*0RX`O`1yli5Kow90Q~^~$6;K6K0ac(RC?N0O
zyFK5Q>T@|!-A7JZ+>)@ik}yWwbBx(qo`gC1zd81+cHPFGZ(L)7a>NsPUGF(^rwNPl
zxn<jP&Z(Bm=a`db^B$n2(d%>XaraA_?lRw`t#_HP?fu7@D*wNGYBv2ptmpGr{r>(k
z=r*7IYW(Zhm$uF}O4Z@Fb&yTRR5p(#Q}6rNevd!s%V!rkszWk#$mTsj$L6lU_-9*O
zcJ4*v9XI-YKDVU#%loTp|M~aW?kiC}Q~_1MuL4ospOT-8v(5Iav-+z7sz7iBO3L+M
zaD}vGHY<>=>pxS!pG<eZXTI`j;o?^79i=xNviVrnv8e*8fGVI0r~;~hDxeCe0&WVV
z`hB=teCqwa*QccWhxJ^U#%We<$DY$1)gfFRIIfkdfGVI0r~;~hDo}D1@cTR5u&=+e
z^*Oe<tf08QZ_#gFanV!rPz6+h@D*@-KHOG+(%M$-AE>;=QD#Pm@K=IHjI#nkf2SJf
zF=az;Ztv+2`?}pNp4aEm!g|iic76z2?)P(%`d`qvxc(YPb%<7nY}l)G461-CpbDr0
zs(>n>3aA3AK-3Bh`*!F@m$faGW8+&l*Z92V`rQkH-rvx6`u;9`kDI>7E?f73sQ2fv
z;wSZe%VCXQ<ER3vfGVI0q^3Z&K9?rPm95t+I<9a(kEM1r`ER{qis{+*mQ9WullJ#o
zG~cuD?pM6RNs12sSFW~*lLC+K{(XZl-Y9mG$~bAMxny1ee&-$Zxh8Gr?S0g$Q+DZ;
z`6Jg~2CF*MnE1=k<xZyQFKhKy1)^1;f3x4?=XcNVoL4XRh}`<QrTgT!&(AIM+rN2L
z@~Y+4$Su`7Kd<I*_WxIzTe?lBIuGXOz1+9uvf8Di%}%4M0;<3+75J#eHMg!F@RH+~
zy9}m<s(>o+|10p#=LO|kcAaNR3U%650aZX1PzC%ckbB#A?e941C!-8k2Y<7m-l~8q
zP!bjBa^N|=->mhev2pu{w~uIDYq`lZOXA$<nEzLSlJa>{$C=&t5B_&$Yl$kL3X~KD
zCj4;aDZ@9ecjC^HG9NllRUk?QxSp$Y^xLa`Y_rk?<$>ot|K;1uiwul1XN|22WV-^p
zb<8jQ#m-qyN#(US*ZHtl@7X3Qrt7}Z{_jmcpzZCeO<1hgpu$-b2d_0jdDP^b!`e0<
z<7A4fo1Xnd;l?E<Dsn!HiXCdr@4vj#N=K<WBw-zRJx@lp&uZLg&tq}XH>&+y#`U_M
z)lU^r1yli5AUz69I(BWd(E}Hlq%!E|scdb>s%i4(`FmW^dy$hA2EBiou6F3E%nk*v
z`C;2n7cL*`q{{42BfL>Wb^oTM#tmy`Y|rEQEe|UnuW`cuob0t<^-~4HP+-k>BcC79
zW{aczvkqZs8%B(z&P!C!`Lr4I(KXl1UFM{lSG`<i^mWZwJ4)5T&pPn?JNs<&4`UD1
z7||<GQlAIO_HjqOa{PWii<^J5JGyMXb@F7R%x)cm9tUB6zSnqver_S?yl6XBKow90
zQ~^~$6;K7DQefNZMTI%HZ*{_ERQYOLRX`O`1>&kew*9_k;ir|yKU(PvN12)qab2%*
z%PZ*p(^Q`$3#+}Gp4Hwh{@|ZS&2wx#<j5oD%)4lT6O;{qJ>Zzd8^;+pE<OK~BaXUb
zkco;>Jul`quBhshc3(3r{evElLE9z$JQH->nQoV<-!<F%b=ya^UKLOUO0EL>efF-u
z&n~$uLPxF&Bu|07DN{xiRjQohl}^XAj_K07LQY)S`iLQ~pHx^iCn!_>oRuwqcR;?s
zAyanH-*+B4sc?D2eK(k-koWH;mGeJ7xbxo&7n!6otoMAyB~GgQSgP%lcP5NB<FM|B
zX`JkSkGsyV&hP*IJBQKkN*Y}i2&O>Nzb7I8u)cRqcbw&&n!oYM;OUN62Ak#Jt<rse
zG<c@rv}8Rcl||(*zhidGZ;kS=bqM==(sUmuQT3Op`$J*%oAm4Mu<kpCvm0s@RX`O`
z1yli5Kow90;-f&+-%rJ7q?(B;pbDr0sz4A0lKvjTX_f9ge|x2kPPQ<r-_=AFH>|&>
z5M=HCRh!1M52&#0qH;OPzt-W4;a|UX*0%CFN#kG5{a@7`eD9SfKe+r?CoU}0d$)P_
zKC#WoG?V&%V9oo_@0-rF9$|lPl<s&*|9*u0y*}6YS9_(_sRF8iDxeCe0;+&2pbDr0
zs=$8~==jKWk8Bva&7_-A|6PxyJr^1E_ncA159_{Zrs8=0d>L22sOB|Yaie;^)o)ze
z`Jj2bEj2%Axygh@zwb{=*F0!H$^Jcbs`Kafd>qyHaOsL0m%npl+jp|1-1fe<xRqy%
z-a3x#P~eSUZoA^a&&%W_mAOCN)_K*?iaAMThYJQjapT-dIZ5Zrn=fl!^_j{!O4Z@F
zb&wsqqkR;q{;tj*S!+}QRX`Q+rhx5z{aXIlf6pLneI{N$^UNV_COgVtI(S>{ne&SJ
zK2GLFmaLw`K5jD|&+q+EH~)0MZzx^;`Tah_=5Jf=Uf-9LZa=sAj(dH&pGOY6Kb{Yc
z8*S^@t=x88q*^ZfwatUudza<&6k)wLE7k9<Wc#q5BaphOtqQ0Ds(>n>3aA3AfGVI0
zr~;~hDxeCe0;)i&3dHsI1%C6A?{`jB7j3T!r~>W^_&rY5zs`rfde1hhgX)mvbs%?!
zLykOR&b*5jI8ig*zu%DKR6E%UxP9Lk_j>t#Xi{wXeQDC{xIb6nmS51nj~thFaeLmv
zZ$77wtNZsRUAGt;lfA2I_Nss?;6;HZZ=S!$6}=ZZ*+RGHvAk%aKB_>rDNwILg|j9O
zUTealyuTf_9rblzw#~PWM-@;7Q~_1s&kFc`ADa4i*?jBd$tG+a{yZycsRF8iDxeCe
z0)MSQs_(VQbo=ce-aevrt>q?N4Ep`yU(Z7LwSMzV^?UbJe<$ttddTnhi$Q<C8P)f%
zQNNEg{1ug*5&f>aRNo)rc0XO~E02G)(icwHI{4iW<LV#vec^P)_xt_2`WKXM*>#=?
zs>8{3A2?z0&?QdT%+9ldeN<6D|0v_zGtnAVKow90RDn1tkS*_nO!fUxnQou1>)5FJ
z@q2$GQ~v2b_R@8qFY11xzW)<-{8aB3CcS;w>)!VMyttHyJzkBM9tGmEeoAT{ar+%`
z(0%u+m#d7vuK8*wDGdAmaopo&`g>Sg|4~0@oND~AcHy`k=TYZlo9}d$NA<mIQsaiT
zj*}Xvf3x2+&+ne!Ij>%B{oEsSOZUldpPyUiw}11h<W<Y7kz1;FeqPPr?E1N-b4$1B
zROi9`yqEj7Tvoeu=;N;6ZF*6u-QP0V%2WlrU~RYBTTYx{!e**@#l3y4+Pm+w$G4N6
zpp4u2-fK|q-Ag}RZ2ZzTj>~)AdhEE{o^gWG7B}qj7uL^Qy?OVICMt%VuPxrJyXRHD
z@cQ14a@RW8bn{#O<=VIEmVIiwNjirux^i^S&$gJb=~qvi|M0`l+V|B*mYQ_YmWP)A
zM}ZffopxCHdq+A+<xQ6@?J<7D5+~ifrRSG5PuaNAxG``4Jzl&0(IVrOQLoEtr+s<2
z)BbxWX&nCAG0$x7Jkv=xQ=N|v@1MHy=B4W$Ww<(A)@9A|<yS8-sj_&*{?|_}yWAw5
zYx}={;lSJ0nWXZbQs>s}-l@nWmE(4{Zcz2!g(hsKy8pPf?|<yk#rHhC$Vobv{Zy)R
z-QxL<q}%vYt#AK&_vaq1zs97>RPzX{{q?_|(k{1fhDjI0$~){h>mKTM+PRf~H0fs8
z<B;(N-FoW3`~EW6QD#PmN&`)wZ8w+6Nh;SByxK4C)v7sel<O|2?e+Kz8b8%`gO0xW
zicyy>aZ+Wfd1(90Dj=^r(ru?Y`s<oczhvI$rE`+T7s_98#^esAbAmEe-NI_$?vYov
z9=u_hlPbf?#}?<7ZR4L^cF@0#GG02^=1Z2}=Nx<Oz`xBiVUvD7y!ayXYLkslSR}4e
z74V`!zxLhQw!dzLlWBU%Regdhu;+<qoIbSC9yx9d-h;Q6Veg~1cm)+InDOt-H(@ci
z->RuSyDc$EWmr07Do*=uZ{Pja6Z1?`*>B<{<KJvK)=4UF{A#b-h23{JN#{{3wsxsh
zYl{;#z3qmjO10SM=@pONG~Efx{p$7j@rG`rolJB1>z$s-{baNWi-lMBUixd>Pn@v$
zPKRH+4Lj&nC)?Qnvz9#?emv1cO|R=YRlip|S3mEvj_aM0!l?ICIUk)GR(fvKlj}{?
zT>r|tVRino$%M@l7oIXLcT2I8X-ZwyPKE+SdB?xk?D@VX(+o?k7X9Dsa{RzWCaL^5
z=g}K(x_-4uDnFTV*cC-HH<+-<y14(jy^lM-OR<v_l6OgErH7gyeMp^6CaJuCkN!P=
zIew$bG`+5$#^>b^Xk7m_lPbOT7nR?lz4p84s=Zd5Oi_;OrKZL?<mZJZDAV;jg{b;@
z;`rOA%&xe=ghi>ZS39cf6R-V;<#%Jv>TAFFWt<6%VdWKdoc3>B+ViH*+nG!=>T%lQ
z-*HKgg6pd<F-hnBH<h}-*Y~SU(%9<V#*K2@uQE~5Ht#{pYxg_g%t_1NH*OsJT%)?P
zhmUi@=6mh-`R9J;7ddeMnHOHYqv={FDkk-N09)Qkf3LvWt}nj6VEQzZG`86Hs}&2r
z8EKNru<q|CHBQj|GVbl#?fGfX!ULxn8{?k8%rEHs8#3K4Ti%1v=(7&DUw_0(Cv3`j
z4%_bk$^Jd2UH+YuRF3Su?v%RsO>vUO7ji!v{`9G<9HloMg0B0sZrQK-9w*OtlFkkl
z4j#Gpp|hO0G3dO6-ELF+j%})LnP9x~ruo~8doG=9{Bn=ZN8EAM12c_R{`mX_b(`)v
z!+53bc|luws`r;O)xP;hD@N{q=L(Z9hP~coyrjK<BX0RhJ)^dpjs2|p2|4SWq_WG<
z?+Z>DQ0yd?^|zKU)2?OdoJ=w5_2u<<u5tBC>T}F8??%mU+_rDKpG??{YW-YzW}CH7
z-&AZ;WmNg48u#$7cMYn!X`{&$gXSOh{rsTug65y}=ZY^}eP->~mXC3wW>9@gO1o5l
z&m5Qb2Rdh8H+%35C)-$jZTVFVX3cWaO}YNPem;)M@2n$>Ck)sz%cM#<u1wii3~c#p
z!v#f7rkU#f`c%Ka^m@PG$S>Zwpzw`lCMq6r-q3TGHJNL&g**DKA2qwphQ$1SG_2?A
z4lKBA<%8pwm~?U3qw}_ZJZ!c}D$Cvb^rS~V8*Y-ul77D{s`J*c7CBGuyJoGU%uXGm
zS~uy6o78<)-p5Ls9oF;IGETbIjjf**>!y^-Z?x1To#}cXY}oxI{dsNj5BmF+q_pGs
z@%s6D=h$iclskN(@yaK!JAF;TD{GC7hhBL?%YOT>GhTVwi(ix-_v>2YMz4SO!RB}0
zh{Ki_?LOHF$|jpy{P^^aZBEiS^3wT_T)$welT^m-^Ml`fUK;X_x%sncPEtAah+*T(
zHeco>l~-Q#TIGJr7CA{}ug89z@<G3iPF#5BtSfRxt@z3D${}z4c+LlxPBk`qeLqdm
z_XqiXFPY!#{Tr`nc7DSai%e4a&X_8Ld*#eBNoCUi-i+7#>T&f8`|q@*J6_m-*VSu0
zuG6Dt(ysev{`}?Ug|+56NoTsg-)8-U^*o!`&y7j(3;RAu)bX0MYdN5I@3F=!ng4V#
ztmnrj{P4~RH}~0M+?cL$x%HFO-!Xe#M{)D>`g@X8{Z3hPU-PMxb~uvTj~&rt$sOf$
zf->oU2af!sdf!aC;wJU?mE_%NLiryiR4i~(rQF|A{X2V^ZtwN}a!^01S6u9#H`jjV
zri1s+Nh(jM^m*(1*4N01E2Em{RG+tuDzC8O&MtnufBOp;o1}8YceQSKYw9|aRPOck
z9gn}cv)Ck+H=j6t%)W>2Fm4=L`mF<VpZ&!s!_}eX^|`YSziWp{D&0OurdscJpB#CA
z;VBcE40lR8{puFh-_h~gugyQz=jJorzWBPMSM@sTTa#3V^?bN(ekng@z`lnM{QF`j
zE+j9n-;ZXiA3B7^u-@OBsW`Uhu98|F_1_zbYkar$-1vpgJ2#JBV7xM{-{oX`9KX+l
zB;Eg-dv1Mi$H6O%S9*OfaNEC)IqKdoi;Q2!<?ng;%{#9CLGN#7x?QIKo<vyvKUr?(
z!c(ejHf|g>>8f!}hi)=)qwRU=q?Pw6Dy%oI`yAs&xAU&ns{%<@phKUgr+#_HGLuw>
z^}S-c<}j@1t}-3x^j4#5|GQwlNtL(s{IcdL8&?{yeB+EW$~1jurEz1>>lMG>=cMa%
zbhdrY3hO>yrsI73`G7mite$J)#&qpxap~vT?$uhqb?HV&GA{YWJrDUgJJo#enDb~(
z&6B?|*~Y4;FFHJb&=ivtW}E&xo_HNsJI`5}|3Pkni5u^!a_rq32Y+ocP1`;i*8L>n
zY<~CH13ox<tMSUYqu0LvNvEwQF66ytNoAjXXUuqTpJJ0Vp4GPE1MOefY@%XP?<)k&
z`?4-;mM_10fyp#&`;OQ0xIQ1?HUFS~1AATYZqt*Bov;`*kFeT(a%lHyLq_eDlP-pp
zciiLLQ|jEB-8&VTq_N_V5lec^`_&|s4US*&kK1aNE=VdHf4JhJ$@`QlNIE}j_}E?>
zca|zh8eeL4#(-@{mo7*;<EDG6`}K=YKXBN^bKi5~M#dHNJz%u!H}!+@=Pn*<GR36)
z{@yLmev^BT-|_TxlQd4gtNOT08%{Ps$vRY)Yz2a@A8nV~@h9~=uG!90KY9H-qYS1)
z$v%q1AAPMs2j_pfW}Znm!yn0?BYORNtV!_;`|q5LDLZ_@%||VAV5KYD-7$RXCno7^
zwW;3I*W`Teq{=_9rsP|N2lcym!R5<MT$!}zfP>~MpO=W*4!Uwb9{I}ZgZh7O!lu`J
z+EseA=gr^$;3S=4z0b?5ZZ^NZ=YQSsgXK$2*!22)pH%&B**5;!We5G+M8#C!n-X?=
zufNCh>KFF&j<$6cRryWxw-@(ZI@zSksPc<C?y`@HKDhp(EhcP+mG_}v?^C7WxyzlT
zvGIBN0~*(V%_NmkJx`eG^Kz*kr<rPh%mEDto!)hlNjh8X`_+mC-;6X#WBa!*?RnGZ
z?M%8E);{PpPM`Jb>K*>==O*c#e8-Bhm1Y(?N#lO?di;1px6w|j^g7?E`bGVB(W8og
zO0_O4j<5B#lT_Yx+0q{4M=Wuorq}h6l)n#V%QNYJSC9O|dViQLj@x<&`@W2fm(=%o
z$-L$7^xbp~T7Pe|gPLBn=@%1s1|6T@cF*+d-14?d%H$-CXU@L=)W^ndbJ9h>x~9v2
z(dwNAgRa~wCn&?Z@0qDMN&DPHzDfT)k*WC(tKXo1KQ3<Vg8rQjxj&VA<M!Tb%WXGN
zv)W6hsL8AyCS8={Oqb2y%Yae$8}#>?QRfwQ{CheawP^4D%bcWgpGzJe^?sk99c6lT
z*ze6p&))C(`A$+f{Iz4A+1z=ilT`ZsJIQhLuifu}Gbb&7-`E)TJ{NVobRP%B*Op(^
zVAd=rsVuy@_tIb6e&V<>?0Ktq`^a*yPAE1(`TpiBhL8Mmt8rtg=Z~v%^^sdlP!6hi
z#gN@cZZ|=BcE__9&$?}!kwniH`#inkv74qlVUf6gnY7=7y5;Nke4Jl>((T``eYdvl
zuUp|L|Exp0=OJi+^<J6yd53>3bo?@Ce_q=)>U8(}<JYb?NoCZ3x5ulF((mma1NXhT
z>NY2-^!xX-<@jy(+f|$FxvtkdCn#-kQ!QV+=Ak;rf4JNU%2fN;_Q_Bn)9<U*{=1!a
zZsi}1>X4l}Bx4u&Z<g}+v|-tqdfxNDBg^dnE<Cf%+NW<SHgTuh-|dWheS_ZRuWs<(
z4igr=t{Q&l>6i02o%Y4%pO!l5qSx`1G{307m-4Hpd|oVRcBcNWp5K1`&P%HP=CEH*
zIKA=;CtLYWtL=5C)?DKxm7|+?oLseDkrNeDT|Y@_AN23XGd{n+8<_deH*Y;*>nJBE
zi}H?tui5i`P12a^@5nOUKC0hEq&x1=BZiGD+kBambe?s~e$DqddA^f0^1N2kI4ys~
z^*yW1b;2h5XjHt<IriFtf177gWmx$|_1*$o+_3Vt#Toltqq?()k8_ln)#3WP8ccZQ
z<yB5nS=jQWV+NMq=y;`VJyLEv9<)5#_otxm3(rnFto*$r9p#^O=zGW5xmWjH@7Q?K
zhc7LD|Cd!x(s`hB_I0xd&v25;xIG{0Hy?fn6nC;;r;Giz_P_7bX6sF=4Epy$!)ll6
z-%ZVQ`?&sn`*h_M_Ihvr(Tb6~-?_rXm0^!JTjS;awD6>7E*<JbO|SaN-_5ICSZ`eS
zIVN2?B<=4M*z!&FJ-pgJoc9^qbZpW8%`V3eTx68~bnv>5y7}FBMYHo8wpe7M=9UJx
z*X;C4>6}c{wojz1Jk#GN#jSt0^%>XtR_``$l-qulNtOMZ{XW9{?)jbb>g67p+aR}e
zpZxauxn+L)H?K-wwY(a+rF!S*)%?w_pIbV&bem3f9?Z{sxo^v5wM)lUAI(b@Pz6*0
z-wF(<l>bnNbEX?*W_9pAaoRu?Pz6+hm@05n`FF?MGN@vXG9Eg_beWX+w4VF!y!Fm~
z;|%<%<pZ-$U;lx#<drY#?2$XwC{>4SSO+EkKvw}&0aZX1Pz6*0RX`O`1$L=G&TZe;
zG9N84LD};9Yj3?_(?qAFahGChp(>D21;%u)*6g1Ll+AHtyMKPtzuffioUoWsxioE6
zKow90Vy8ftK6iC3x3+xFlJ*BLe0k(b15X_CK)Z*IU-E}dIqdk!`EB0$YMJrM2mk)g
zsuPY~Wc+gLn+q;z)nvZ$%ICk@^7l2PSDUbS`>=Ygf6Q5F{PLCiPuQ=nv%=UodG`2+
zRyJK=GR4O#ulo95AI>vjv;Se|{Jf)Zn2DMl%B=5K<L{G<8~6Ks@137_8*Oa#x~{_Z
z8{GPi<rjak#H5R19S>2*>EHCzQO$0znv*K`y1eTd)gE2w1m)&8Hr-Rb)2k-aJnZI8
zbzVGro=G~7t5vtl{$ocvC5^E=;cG&H6iC;3+P2>;?^yWA;j1d=1ZAK>C$Ldo^E+F8
zyJ3P87858zpp?75W*%+RXsL-i0}W0asRF8iDxeBProcrVYJ6S3+j0|EesS#^H4Zyu
zh6#&*yZ4{3+&pB8nR3llgTG(6-3g0pd+k{JW2-`wDl7Nz_u}k#3r$$eSyN}PK`#}U
zbTP7ps)1DjRX`O`1yli5Kow90Q~^~$6;K6Yr9jx<x662K>P>p*<ifQkDqcPGmzqys
z{IgMJw+^v7XfzpBKow90Q~^~$6;K7TSb=mskCmx@c+ZSqX7QBk&{P3cAWIcEugO2E
zJ+{|k2Nw2R)b^V%hZr}uJ@A&DD=Pi$xUuQRmm3Va<9m}-zPNEl)6XW4aU>hJ8Bp`G
zQJ)y)u64-L!=+TS<t=-iJ7T$krOv#d&zQwUrld0J-^Z+Yc<0z_CVb=A$nPb+vY=0;
z8fSd7%6R3@IcK)peBLZ0*=6*tf4`wj#T?HZc-qLd^Ts+|1f%}FUaIj=C~SDwB@ce?
zghkst#-;p{uS#uhw6s!=WL)<7y&h_@?7MbFPEsk?zuN!V7qazyspC=wQ~^~$6;K6K
zfs(7hbN_g&VuK@hI=3w8GxoII=URD4rNK*ndTN~`>Gr!~x9iBk&(&=BUCRl^jq-UV
zxA*MKsCL|aFJ3U)xbe|D_g=EH!(1a-a#w_oToq6SQ~^~$6;K6K0aZX1@UB4X`rFGK
zom*s-@zlZlOsc0UpbDr0@ls&x!?(@;wL{gMsL69td%SSK*?TS?>_{&E{)EDQugx=c
zH_ZH{{8zJ5=E&Z4YgcG+rBSL5aaad#M}5%0cayE{-1O3VRUjJ`IIYncEi1O3Z@jWW
z)tq0Bx_qti$|(mNGy3#pD~(s)JL=7$2lbifNKVapwtS64*E*^4E$5I^D(#r+z_6Y_
zllewv-`0NHvYDGp<%Gq0Z(n>!+keb7>Ebuf{V?X3!7EI*Fz3{UGv~hbvk8ko4sUbB
z-q$ZN%D>j(qbFZ`dV9fK6BV;@ug~T&J^lH)Cw9tTYLu$OZ|fkN*Sd~P6;K6K0aZX1
z2&%yNgHO2Z_yxP=_@(^4uA4oeQvO37&Y5m(O!aq;^7n&ocBbBc5mvu$&*NrWecawR
z==bk^+Wa@nd!={f&lVYFv^oUcAGMt-pbDr0s(>n>3aA3AfGVI0r~;~hDxeCe0;+&2
zP!beabID7si*}puz&C$twy5&g<#KEsy71!W2VPfTg3|3h5+$X+B&@9@jPW0<tKWHZ
z%T=bNFbQ+=e{<}1aKD~4KHD!RC?Bi5>g#`fIM0N|Mzu?Sa6qfo2B!M`rmg+JoyX1D
z?a3uhQu)fjd9AiS`lXXJ?*HD$|G4^-g(m5Izv|b+pK7+yB$bs9X)?6ftT0LC?ZfJ|
z{xN5zi3|Vlo|;YnU)8k2{CE0(;iSq_%lA5{+h?V6Aipn#^6#3sJ%^(8yMF(7w(62i
z$5b|t#Y`Q1!AECLGfLG#*Ma>w$mTsj$L6lU(C^pIdAn1Q<HmY79x!H7*|IsxUF%Rd
zeUFoFe|v+Y+_etwD^WdE0ad`S0<~(LeD7TktZ~BTJqH};Y;e|@Y^7hF)n64*1%fLu
zdE>30nqD6}+0Nh!Y0GR@pkKF-PkjD`)lRnY>XtLs)*L$D*l7E`Xwu5P{yj|f`_JDG
z&E{iS$EFIX0;+&2pbDr0s(>n>3b-k7?A*nZcb1yr1f^Si>a+E|-cgM!toQk7obeB>
z{D17-cf3~Bu{UrIs30muLnOf-#D3A(V%<-S9l@5w*lYBN;<XTatf)a_qKGvGOT^wH
zDwbp)m1t10U_(@-2?PP@y}Z{?_&)dE$-D2vfdd@K`p2JZ?OC&CX3fl+wf24v+YQf?
z9qQH&;yPinK(auxK(auxK(fHUj|IB!(6;##Jw|7Xigoq-H+<s1$G)<(>x&2dV$^j5
zezf|KY(>ZVV$UQ;vOuyx-CN+ewf7&o-tVSowZ)-3^c`{GdV}(W%f}9N?+J-xeYU`B
zoemoH%RT;)*B0xuPAh|KU4|dD(VeGG%8PKFFVDHE{jQ_)Wrd6Sd-b~dJ3TvXylLm}
z40t)KD?H(n+cy}t)0cU|WQY3X2P;FaN;)J9Bnu=9Bnu=9Bnu=9Bnu=9ENu&1)pd=t
zpPKhkR)oGQYT2^Q)lY2EvXn5{A^AaygOmsUnFp43ey%J2bJlqH{VwZI$W}D0EBzCn
zWPxOXWPxOXC9^=K_hc)}-=E!}Y41^kdQ8d^COiCJez0WGRQu6I|DHkYZ}-sA8ytN2
z;5=cn!_x4B+IwyqQJ*YO^gYaq(qEr!nPgf17C5WR@I5YU{nfwJYo~nD{i(xSP011_
zJ0w5&*Em@Iky|?nuDHCqO`{h_=gS*wCv~Du7FfC#xOMx5{Cjroc~W(|4qJBEszcRv
zJrC>Iq2V|Gt!`euc6F-`4X)`~z3w;d%{x?esOorZ+iQDP_v*gam`$pdu5^ievOuyx
zvOuyxvOuyxvOuyxvOuyxvOu!HN^60xvtI2ryTznDVY0)@><24t04K$g1(F4l1(F4l
z1(F4l1y)cCG~RaL8E0SlWxl*|1vOt1J6RxEAXy+;AXy+;U@2H&!Na{9u0CZ<zOvz>
z)>};a>z#Y{+xqk|SzY5ZEvNkU^>wFYOB*L$@brVnJUJn+E1WfVNXveEew8O&K6a?9
zbzEM2Cft4Sw!41xb)Inf+F@C(zdTwAmvo=0x~l0VTMZbUudFzI(@XBW=ag}IUE}VL
zjC*PQT_$FgaJz$#{q&Z1r)KpPJMA`YV$;t*$!d!`&ujbEpl*}1Ma4>foSr8O{7(zC
z?|Rs!m+k&}Uf1|P=ScsbERZaaERZaaEU>&SFuw1>`*pkdzAWLA*<pFFW12r%U};$3
zud@$t(z(&dd{Ob*p9~pv$QE<6CB?pDe%fU0q)8cE8oDK($pXm&$pXv60?X<?H_cPG
z@6(b!lRfM5{p^bBPs<}`ra6-Zk_D0lk_D0lk_D0lmWBm({PQzst-Z$TrP|`(J3id%
zySuGcsx8hxvfo1!<~A%<!Y$s|?%b{Snwc#swz&02e|&r5ye#4WZHJ}dLy2dyK(aux
zK(auxK(auxKz+8r!9PFbo%0ur%2!nU{?k3izP-&q@}-USc@2|n$pXm&$pXm&$pXm&
z$pXm&$pXm&$pZD!0{8ap^5d45eVr#vcBo4~sE@wBBFeP?(6*;GKWA*7Fxg>g_`!-8
z6N@7At!;n*y;qtI%oZ0HWsXFXERZaaERZaaERZaaERZaaERZaaEU+S2psxN7-pclO
zOjopj*Q~aGuW?!XW^LtLQu<}BUz%^lu)vD?yWS*s`hENI_<j3|5!Fe`WPxOXWPugW
z0$sLyYS_Uazmq3ix^~!XbjJy&T%YF&lO2|&AFOy8DM_3xkSvfakSvfakSvfakSvfa
zkSvfakSwq~EU={io?jYYH~;?b@>u)2oU^9w6Qe%d^yR#^c-boVbiSh9gltjq+U{??
zKf2G@Y-wY!)(1D=;Dn*szbn?IZzNvH0?XS1E86!~N#0j`&2BL%Pj;x^c8GDkxZis&
zZXG%I{k^Uq)p~4RTWs2Rr;$C59+}q`v&$ap)bYrV@`%w6shup4ERZaaERZa)JS}j`
z1-Eox<A|YoU18^{&I2A<b#R_A*`Yr9!6j?F-0R{^-p>;*TRSYz+?M7}7DyII7Fd}r
zaBI~sZ_3wuKL2;c0|)<L+mBy<FHe~4@Qoj=%(4EBZcF+*_r<+`t*dcM`a9BH-W_&q
zlluqd^$k}!@~kNx9~hO_ckHy=w24hW|0G*nT-oC*>6a{!ERZaaERZaaERZaaERZaa
zERZaaEU;`Wu%v%?ZBfThA35#QH7=W&|GVI_S-mt%vOwKfpx2Qfy!G_HbF;d}+W!8e
zw(k}GUEjl0(z|ZFB5_F;NES#INES#INES#INES#INES#INES#INES#INES#INES#I
zNES#INES#INETSpEpXPl&H8Nc-pG7e;i)}_ANbr}WAc?9>)J2&>Dh6|nzqC8#l`ym
zyNOEj)wMl~_`QA62DkjY$LKua|F%Psy;6U&Kz*>lkO}*pKY#tnS;D1jho^V_?Yj5x
zF*SqrVeOL)$pXm&$pXm&$pXm&{~i|T)uL*zAB=lDuPq*N-p&v7{`85w2rGS$Twkoa
z|IoIlHa}-<Uf*%8!`Esx{DJZLii%76_cY@8`r<!Hj%0z_Sm2(+$J}<w?NjoE|F<1#
zW79;NERZaaERZaaERZaaEKr{<(DiR0{^{gvhUZHQr#;ztSl7wJ@+HOkY{Mj5vOuyx
zvOuyxvOuyxvOuyxvOuyxvOuyxvOuyxvOuyxvOuyxvcUh(0`WVNW!3)w$&&i&p9Q+k
zTDt#EeSP^mzPE0_(DpsM_B^S&U5BkYY~7*ix}Jyi?9lL=|5i7zUc0(ghX&X5tX}t<
z_U0X`I#hK$w(Ye&t9y0dYs@B94UV|{{;Te4_CenHi@qbyJ$YF6-rjRQe&vnNvoqK0
zdUUs8BeJt^8++gbpO49Ue{|9oFF)3>^zV+3-q`b|X`3}I{prpJo_eBvgOc{v`!sp)
zwdy|kAHRLheh+QkFJFDX+u!Wh?6`8<^B3DS`(*r96SJ$H-)o1n`n;N<n|l2t4<B;j
z)7hx|54~=mj*Uv2{@|J}eb1^Yo%{a%uk=0otL$$_JllSsd*^1ex)0v+<BeLD`XAD4
z)};gA&y2%Q|A05Tjp_CFpzQj`+ON~+(E(*7U!=F*<lzU}&^US+$4`u_e<@+bkjVN$
zcDHVOHhf^sCjXd{vA;HY{;>PszUn!>^Q3b8@&oJTc_}-o+jmZzUiL@phmZLqdM<yE
zuXoeq+I`*j?OgxSv)Y{dm!_ZO$Dh~Z#G$JVDfeqXb$+)8ue@?t`T5gNXB*wJU_}1h
zHqSPHcFvew8^73Z{EH84y<pIc@#S$RpE-8R);mwj&2Kz3Kg1V5bjC+HjDwy>#8*GQ
z?8$EIiJ#~C#k1%5p%a(TIz#^9`KbM_KJNQdXXf$}L<fy0r+q>{?AqeQUEla*d<OXy
zKjWwTR$e1N{^X<ISx2^i|JLJ2W%Nc*Uj6J7`{Vkd$G-^Ck)M3l5At{GtZh8Gj3<xr
z^wl<wA3optoa<h{=&Q0Hn+Gndti~V6Wjz-c`SmjsXW#y(Uyk~hyuc328kdzl(>yDt
z1uj`*!N%P$-m)am*;ln+{yZyf(tAcGZ^)10NB-jf#>;b_8>ip+(T~o#aP{e%mWK6t
z?6#AetXFz^RD*k7-+1HF4?8}4+C6>Sls?~e=Lbq_u3N%?-|4q})p6LoY-Ph}Z~F>V
zkH-B<zwt4Dc`k2zj<5BRx7AD55wf@O_^#~89*~{bMSt{b{YC4=j^ctH)e-n5{AazC
z?67q0pl(s`sqf@nbz3}_s5g(FeBN%;cAc0Hz3I>y^Ddf{tE<#?Wx9puCT;bjcu^mT
zFLk6kSiaIvKKTubQ~GG*Pp*h@-u&)ve{Zq*w9I_QlgB=WkNMECqc%I!U)y;0h3q+e
zz?%QreCCwgIgq_09lq$v4adGbtK&~5&dWWAW4i76(T~oZQ$BB!n?B^TUXVN^oBn;&
zM(0n@$>lluPi)`*>D8W}T=oz4AQyYemw!3A{~-$=nw-@Ynvefzvlsu4enLL$>^b{*
zj$d3)b~8_G>yL<E<gaahcGIRGImvB*agId?@x>3~<2icqy7%*g&fERuIhlE4JVttL
z$iK)RksQ!IAWre|-2D2{kq?p|vXAGH4}R<c(M6;OyFvX>oJ2cBdi;%p#-ZmoMHv0e
zIA|O?{tES@(+}k%C?1SM7u)DPhx(y>C7)Oq>n8v62Yloqe9SNI$*C=mna_CRAp2V%
zenB3{4?Sl;h#!9-H@V_|N)L#RoJB}K_A#%1<Fv`4pS<SPj}AX=XuZhE-}tdQSlc}G
zM$gXdq@O;<YqPs~A^og#)W^Ep*C0KNM`yhCmRIP(F6w@D{^?haec||~jZ6G^i$Bjf
z=F`rrl_DC)PUa<#d8`M$%<DP(kk@nev|bP&`qGQM_^~(t_T0SqqqlBQzj4~ox@(I&
z^Rm17$!Y!Lb8YkCZ+vX)hvX<~>xb@7oYS4ZU;XZ!=eBP<JF~ufUAq3<Ex(-kFZUa-
z_dMr;pKi1u^PbLm+<P_qu=knh@b|vWIP<BmA|G;?cVzW#&3AaGK?zz{c3>}Ueie~l
zS-<!^`U$@1<pc8@Z@rC&*4a4zVBPW2Ccl1sw2Lr~CztgnKY5_%_AT|V=k!Mxk>198
zE-vaC$;FT9Iq!piTyyrsCN=QcPW@}1`=~(;KU24P4&^WWyvNZtAEaNjH~U&2a>@7n
zjvt`Y@Afv%BTuUjBN|r;<2vy}a(f=<r>}l=5v?<SVIT7tkDuod9lq%G^Go{=e?w;+
zjC976&vW+Fj^p)1cDB9|tt)%P_|@hI>=BWi_NPiHZ{SNV@<4RZ`WvTBPHpq!3)y3N
zqd1Cn1-sH8ic4+!#kO_ir%=E7U=bhoWmn_Tvvbk9us1ZHI?T8g9q9w<X?<4|<X_~C
zZT-s%qaC%KD~j;v58Z$77SGHm-<#Pl-Mgvx)#uPYT3f^~jx%5M3-scGo$U|uwz#FY
zaq=vb|K$zi^y4dTq4*W&@<>Gd#J@H^+E70<j=cEOUmHEtk1sx+M?@!|Xv5fVf3+X8
zlYaG-^|0?kbq+fFzI8AkInCp_`SjC|ozQu%pIstf_AEkv6yq0t<cojVKNh~P;Lo1d
zS7iT6NPqjBxQOG>#s1jFC-$fIKfjBuWQR(1neqHC`fY5Zvwoh(Hah$v`f?mBd>?0>
zJvSa-am$~ee1u-yYkMA1K8PsqppVbxjo4q;ZS%7mz2r~xv5Wo5b9vQs{m#LWUfy+I
z<9-M|eh@vn*dN>USl;OV>ES~@X+83x#wGVw?wQ;}dGG2T3jJ1JKl{PBF-=O*e)Q!(
z>~0+w7mMPzl7Gi><`*}{i$nXF_|vZq)py1lhrSXT2gMWl%p>mAbNt495z!eR?QcH&
zH9GQ|FShmLPfjQf@z)>Gb10r8U26aHcOEO69mIb`^{;#p^Qkt!6A%0<+5sK^^&CBa
zH7|bRfgBLMxZvm7(0q|TKGznnONud%rFQ+a0KecT@?2bh@AJ^B`|K;u<@^Ypaq>R?
zP(J1F&V#3oI^ny!Z9lvGyBFUX;S0&FEzV%k^P+gc$MYzk=k%tpcvi2m4|*8sjjL;v
zSIASuADw(=-N~yil6N4!F>l8E0CenV-D5pN4$q-}<F)aH@+A9OKlfhzk6sZykL#vQ
z9`;#O<j?GH9DipA&tt!K<b&RLc3~&;u#5HeK39FLUp?)8v*$5>*cZA75ueaKpnFJd
z&m*4xz4dw=^wiuu_WSP8^Vmio`D;URxQ}#is!fkdD9;tqu{(Q0{v$70H)x&3n{%4;
z3_n*NSSR1*InRkd<2<*{-m5;i&Ib1k+-lVly3Kx3`m+DHM&;i%J5NIM<AbmG$H#XB
z-uG+!Juf=^jQ2f(-;MA;^P#67efd2<cMc>U`{RqBb&v=6IkaBd(0F#lUtF1Y+%Z>A
zxajq!CHj$<eC%_^U+@3^Ilo%H<a-|HS^CpY-i7+<txeA8Px1#kZT_Y|B0jO7pX=wB
z{8`>44|@4Y{?hiGe)ws}dB{b6{E9HnLoWNc{4XDfbAABrC**<T*Ctn#(|SR9Nt^sp
zZ~gKP`Q>4D$H)3Y^3g+GW}Vo_c>Kj3dGO=co*PGh^xEt}5Av`(dVC@N;$B<+Lx->V
z@Mj<O0DZ-==W)E}?7+U{B)9d2<YOm(%s!suLtg#zzI;Ir{p_gUIQfWO@Q-@YgM8#d
zj~^r-KIE_-><^<{_|wPuC@;Aoe&&JXVh8k|+eeJ|TwX4cN4`a8oPC45^1b`Wi0I{c
zh(DwUdVJXl(uZC3(=Ym0TrcsiAM$I@#lPp`2kPes+WbZvKW+1fGbnzF5FLziSV!}F
zPCojW-*fVz(=RUM2mRU*zlg@GE1>xzqK8qQNN+xTW8H$Dy*!8d<w5+6Lx(?oAv*F#
zdi;$?ACX@`c}P5Hd(JNUt+#%8mp?%BM)~7B@(469#7{mjK0ZfBPVq`!^xEj<EpbdQ
z@r@6@`0!8lfVxEfGG4w>cgr{Q(J#M2&mlVF&7-aUMQ>jDN<TjIu^-5D5sicKxq8|8
zLR(%Y2RYT<@+~^+VSTNWc|B)W{q&=U^+b;@V*R`gOj5CcxU+8dDS1x(L2)bJpp%FA
zJO3#{@eS=`@;1IuzkDtq%G3DB*Z9dd#>*etp5sRk`r%7|bk0rAQS63}-sBKx_=+2H
z7>`anqVo(izw=$ZCo+H3i{12V+t<uXPj*MI?OZ26TQB;t17sh50{Njj67ti8D~AR6
zKmXzX{7(GBqJI7f`4#^7@<;RG%U{J2J`u&0arkMI4}Z_i?>RpB8y~TV9zXGipMHGw
zlOxKdAH6oYBH|ZO9T4*izUH+q`so+Zy2-ETBi1$w<Q05lzKL~^{0rqV`v5-nL3H@y
zr>)ME|I|(Fr*2bUsn3mzsQxq#;s@hA+V(s9J^t(j)xGq?SHJQ4)t6BH>UVbPQu7&y
zKO`SJLiR_;{?-Q?&#(QC3f9(t5`F!!fP5m~Sbu&2<r~k@>sRN<HxPgMPu-)<pU}&P
z@^PF`Ka|JtvG1s#$mMyF+>sx>%<s8<!E<$@Hhq16Ku>Z)dXh^&xz*F&KeC%T6kl}S
zN1}uFV|>t?&;00&gZ0Bpm!1^X$@*!tul0w~fB40sBL9Hw&#$y2dQL9Lk6;|9Z9U<P
z_4{tH`}tKh_(T*(=;L$tiguxwxYI@_pRgA?@vlu^^bntj)<wT|B{|WnD~uzbHjMnp
zrJugmLtIrt>ttM!-D7`IyJ$Y~@40oA4=SO0fPK_o>ML~&I(3G+RNbOpQCFyAwACZV
z(K8}F*j<|))p73kd{6N6#$Dgsy;;MO?-SIC){{S|#~?Xj%uM89N9)SY5j}^-L;B-K
zK7Jeh4jnt0Pro>`F3@^LIoLs6X<dCsbN=OnPuhOx2Ib%1&$<0SnjL@cYBf8L?|o@a
ztA;iB+b6B7en=m36d`@(iAr?QUXdQZ*sm?_$yY>Q)Q^w#gs~rg7@yPEIzsCO<xOpf
z-uy*~9vY{;tTE0X+v0=XQ7`nc(sOo);$?C5Hed8Z@|Y*0{3U+XZ{kRN<D+h|ud4@*
zQ#URy@^ALmj`2a?s1N!gjDBxFiT48dMn170Kk`BAiC$c3M>LOq`a$-<S3fKor$5eb
zJt6+kIP+<vkBClu6!|ee=7IExh`y-bbM~_y+Qy+Hr|0_7MY)UgLC+4__(i?Mk>_##
z*tY-iPi^rH=^LM`Z`=pj|LlwS#<qTZ%|kwV>W_5#A$=h~g!rO2zj)NIKJdH<#h-D}
zF6@9GxgfsA#Ws6+PJeokTYnKozL74<XFl_r$Gr5$M?ZRc^NVO7{Pde2;s@;;;zb<9
zbro;yD-P^8<|B7(n-8MXHlO+RLw2E`@$6ZI^k%mrI_pSo^JtTg9Om~N>W}l_Pae<X
zy6Q)duYP*qqwP7ljVBL2#(8d>er@9*InkL1z30$;+T`T7#-Y=X9-{LcJ$@16xT5j;
z@zaL*6d^lAJEA97<fqM#@QL)!iTvL@zBh4BbuM<UHa~me>%D_{(bM00ia&8}|1-{c
znOykeCqJkop?Y(dEl-;I-Yu(??~$C7$!i|{@tlo6L?89gMi)^$)5CN9+VUQ{h~mS#
zYMWm_`&b|QxqfzJ2je2We&Zo~X&VRin@1b!=O^UH*S!3MKN=tBG0yi_?gR9DFU5b+
zxhL^{O`DzE7eV~+aX%wY+#f;XAUVhfeXnJHZT!urpS+&Oap=iwe*NTSXMT^KT<j$N
z<qv$&X_JRvSug9TZT;98AATSn__sI^f8<85Z65vn#&h%XKW)$HXI!L1kH5BY+7MrS
zw2dzs=Q%#w?1vA&=%9Y<z`o)Q9lp`7_@iSF^wuTzqqknr{*o|RAXy+;AX#AXEFk~N
z>-IbOUS77Z$j_eJZ|(Q!(Wy_=8R+pdPQ9T%Fi!nopHUY;`%sj}JoXRt_!@^UqUY$z
z0nO(*dir>-&N097#>IZmE1`KH`Hj<0F8jQGZO@I5ZT<FddKs_ne$u%sp2OVJc+c)$
z-@U!}wCLQ==r_(ik9nNu^t%@{zkYPkb9~K*eo^`0XS{aQi@f-oAL^&S{y2`jkq&)a
z2jlgdhrC6IzvtS<;|tl#c~c&Ub)Yu-i0X889zN*k3H3YY$MN`}gZdX0`G0*!`J3Gq
z7wrp6I$nOL#K(S7iB8@rq9eCFhTgh)9<hiYdy+5m!xx=>ZBfy<IKRBRq9OY%s@(SV
zBKjhJ*0*R}QNR5$KDQ4?dCUXVG3fELpSu@vU*q?45P$q&9A_T)ALc=?jm~_Ke9-sQ
zp5tpgI&z@*Tw6T@<8!~a@!gR&I&zCA`B&Z+moYBTspt3wde61hvrry_u^(S?Ykp|o
z@Em<5<Y(4R{Y4JX@zD<#uc}X@R&)!fZ`51rF!iu{OuxDls>`(Tf!-6C2Z}@E)m2cO
zp)*c=6rq1-%Xf*|?ww;kkuN-#M<Si`Vp0B*@7zzzKlng-Puu)Z{=<(vMJV4Jhffjx
zvg)U&d_*6}-u5%{NAx=w=sPU+zy1qfx^?j0&or!YzGvU1h4kYO_8I%G=lljn`NRi5
zKjVatS3j=(f}CEkuJy1kb<Gz){OvoC9qc>yBlPw?^z4b=erg=_yb_|vKho<bPxMdz
z3q9ut##w)Dd?T7)ouLhlS4W8tc82l|{}NyN$w_V)`QpE#?`ui^|7`*1NB%F)w4pi_
zz4Idf;lKRTxQNC_G_QVjw0X7Tb8@)XkuT*>^@8!_F)m`{Z=C$%Ie9&&cM;kL%&%>{
zer;&JC}*X%aS{Lb>ZjI<VgddV*OTA#cXa$-JtQvF$NJF~q4<mbW?uc~MbCddC%1V#
zhx*AyPIW8(^rR2eZ$5mr@h|Exk~`9)GcUwHB0E9!kbLxYo}mwZ=%H~DqrBSk3zWa;
zgATH{`N#puM=$bw9_7*x@m*2kFUeUyEMR}-KT8YUQ!HzAu3Ogg%Ddjvxo?5a3-Y*i
zKre4;lizq~o|qrx4}740c}D*5+<g4NxyXFzJU2hS@|frJgX%u|yT@_=<(|qskp7E{
z_>fcHX1Ay}z4RL&+w>$C{zYx$J&#!QoIK=({Kh%hIn}w-Ia9wjyT^0C^~RT-tebO*
zdvEmC6B=KH&N=9v>)l)6uU^sb+^y|7q<7TU{hfKB`RPR;{Mak<V{d5Rw=R(1!$_xY
zVCP8Z?`hx%(LwR7&Y+Ly=CL09);RYf`tcFZ>_T7m<3H9D9X-vf4$>wcdTsJSe#buM
z!;if3ll+06{N%wG;&1+l>|j3mQX60W&^^9?x4}JsjBoP#9lHB{{l>`?_<9aKhknQI
z-;=iwYVUU1`e%-4*|^5<;63MWeQ)UYgZ@8mRQ_GO-${5bU(Y(~yxTfXZB(Nl;)f5p
z{eD+lUZbbqW%%C4JuN%Bmo^W3MmuPm$9Fd32Y>!VA9Th=<ZsYC{D5EZ1N!lM<09HO
z^z&P7a+uG4gntq82Ykq-A3x*S0X=((FZMP*BDo;FAinYrdqMdM>KD)Qj5w9Q$%%hN
z`mrl|>!wZ4i1^{-xi)>#u_OEHM;DQOA-kF<(#L*sn8)+j#z()l`K)j3_gp_Yp?-W}
zlmmbL{1JUb&(TBk#<qSK$Ky|qXgB@_`6)lqj(&u%ap=u&oagA-l|1}~{PK}DJL4Pq
zm<JvG;&^t8r%-Z>hbS-p;<E^i!%sV+@v%S7kIsIqO&*Akc_Sa~i0sHNko`P2j{Tv2
zNDgwL)5Z@Pht7II@>o~xh~zbn9q=`edE<C={M&Q<_^E#UDj_-8#XOb9ReH`I>ZT$z
zpZG$ju28S2i_}k^vzPJO_-cC&<GxKF_7wm0K(F6A*iY0+==hU(fb@gLnV+AK6MyTA
z-gqb<(VKk6X`_ey#W?eMZa%1gagknr@5|ogWN-Y*#h%7{4lDJOpMFLB$W_Eo9qhUF
zvcCA6kH1(Sdh(Nq#^J*r#(Pe0{n{}0%g>%iqz@z)I{fMjvI~2X6RJz)HF-|I^@7fu
zo*So*bq<g9ZEQO?qt|w>73a<!l@K4}%?F*6%tKzNE_W{w?@!#z6nz)v-`_iJ@Ag0H
z^y^hhz9;kVmqY#N-Ou<w)pr)Y%Rm?9(T@7ilitvM6MFX2PjB|nM$fM9Puy?ej~<<I
z=F!i8V!!A3L43@oO-|$J?YZ&jVn2EHNA%pi8GFM>uMUm-aJ--J{L-D)eB-pI8kg@U
z^o!RbL}$Ev0QVAp$7H;H%l>Ykfcl}h)3)!pufWfEXy1(OB6@W87pPxek&m?_#^?4K
z@|$1&(+}Cn{)3Jk<URD}lmE2oFMra*IC6WAue`=zq4D(LhsKdx+c<RWDX*c2?4r$|
z$;1B_6_4Cx#GG%v*P?WIr<w0`KWF2Tzwar(Y0Hn=@`*g;UVqH+H5dH!%~eb0mlxfi
zi$nL~@~Zr7{F9B>xT1C2<|X;zgm1q)_^AHhE%kb^(aQtIwl0mBRNd?s*KSan)qU`m
zA8*vM^yrN}Z<@AQ(-OI@8#&2iU9A&;;Sa_^&-saU#fM*L<E!mCKE}(Z{4vt;4}L)}
zZOE?dPhaaoZ+7QzF^}^jdDl36j5jZg{p2ABq(3{GpPlUo?4pehU;DXvA^T{Pj~?{W
zPcCiz>1$lH3qIsyPi^vwNBYrUd~18o5BQIM_7Df=f%@&U<YNaI?GyE9hu9z6_*e%h
ze$ksZqIvWi5Al&VqW!hmIr7schqm>N{p5&xkprspBC7KuqNA7JE9kE)B%gVr9`vUt
zL<fzJ$X>C(Qaj2eKSA@Vm&pam4ap%cwV{4>Rz&$&+=(~wBcAljzfgZf&*cH}rybFA
z^73csIV6|o5ygS}M_r))vu{Cmfa(eMazCIxb>4Jdwoe&{k9pN&#(TeruX<P8eFJ*)
zMLy2$=Fuh>dj0MZJdfvk^!V|MsE_@c-st3M<Js5#<~e$PV?DL`k9==_<LnpY#aBQ7
zvfto~9<mG6AJ-B8C`at~96Cp>Y-qi#546snvjb#TcH<}330i0SJmeShguDoiv;I&%
zkUz-FFVOMl_*{N5p8WKJ@{6`OAs4idJI~11tq-62hnu!*Q9dV-i(SY|j*B<%JL-$$
z*Dm2>|Fn<L+qyt<&?BPf?0^sZn~!|vV<&QYZoGbo9zT9#oHo>NKJ|tDlU?nP{K9kl
z>giXGec||~jZ5}p_BT!&vbTMa9OR}4xy-A+GQas^J*{pq4xh0b{{8&w1DcfFw?#hq
z8BebV*V*8nfm^LwQeUwXI^*zlPPE?@q4nn%{8B$Vh);2g&N%)B^~30o?2I3O;77*c
z$IsalJv#iwsW#M)k2Z8KW#5q3<TLHKpXi5iKKVnuuq(aUft{m&@+*2e|InWwm!Chd
z^@2e&#^>~?D`dCGFY3pi=@Io3KM}=;c!%s0*PUO8ThH~Y8=!R&C-}RcktfIxt&`{S
z4L`+)edw#-IP!`AZMHu5p!uh+QNqV~@oXP(PbB`Vzjc?t>`U&i$Zvd<qy6dI^lSd*
zTIKfu?z^l5e|B$Xz3fNi|J{Gw_};@4)+xEi(&o1@e(;6l7KhM!MY+ib)uYxuj*s)0
z52A<Wh2l!vIzj#s>Bxm|MDiLJ>lFR`1AU~^ACbJ#{`w=mb+xZ(ix+-rf04)7L7Tkn
z0_kIZ$S&HDp7<KC9qk+SA(wu9q5KKu8S!L)BOgBIhy2UFARjuX#`nM00~(K?e)^FY
zqSJPcG_SVvAaqW2{)}k<({G%%b1eS&*$>1MJD`K;oKLmQ1JT1M7kP@tdu}}Yv$OSP
zKhLc*JJQ2A&tY5#ZO`#9YV+^fqWI#E_P02X=j2j<kc*ziSwHKlE$+~>hxKAV@868)
z$Hwtz^N|mVSM%v72RrGvj`n|cpeH&=Zu1yVKJiLU^byHp9R2m9m-nn!C0tSO6VxI6
z%z9XVej*;BIEC^HG)^4i2k}8?{j@!oC-8;lMQ0rUg`Uep)*C%O<cw&3dCxeAzczmQ
zVNt(vklg6?lY{;Et9-!k*^wUT_2Wk_ZD<~C^2C16^+WS|ZhlA(>wsQAx`_0P{o43I
z@`!W%^+WN_p7=#%uf;|8$xyzH^JrUtzYlrg@7a@s2d-NFo&z29`xEyBzRUFQH0p=e
z$2#Ee?-$V%T}1M`HzzN<S_i1#eAdN&VjbvX9C~fz&{?<WAL7FPVPCS(cpecy`<4C1
zc>MI+7xdFd8=9B=^o8`(#)qEJ^C-7A{`?)ix>&xqf6{||<|U7LBU%^vIIb^$(a-O_
zN8vyEqyO+f^yVWMzS__@&*g3H$cJ9+g^zw{9DejRU(~}mcEFb$+W12K?17Kx<b;u*
zHo1(ig!tPh*&RPfAN2ToUQ`G1L-gAC@LT#r^y1Gv;?y|t@40d6N%qnY&97~|`SFzx
ztUo%>W50dVI>6{R>}fot7rD)+9py70J<JpRl3!V8^P?A+_{$6IO%C~nJmke6(mT#W
zPV$)_S{LJrkeuW~r;QIfxTyW)jL3i4#dCIHH~t~7M#RT>82j0my*-C<oPPRh<D+dH
zL?7wUdyYOL`Ov{gSKBr|&^!^@+d8l}`K_0J=sDES4v^gZM?ZS!DEZ3%s^9$P!H>VF
z7oh$)F3yLae)H>R5BBlgJlYY-tzR3`58~(k%{cwq<e{H7#2-EW`s4XL_N#ByXAqrs
z<m)*!Kh$rWeyBg{>$&x?pP<u5hmZLnxg#2f4%+|F>&Ks+jAwssbg>^F<MDTY0Id@`
zb*Sh1q4x#i0$=^!-+1ri{i^pS(7iW0?_K2!dBnWp*}j3!dl>I4$f0c>^SWR6-UXV+
zdzFaZJE_m<&CZ^CFBjz?5A^=Z`=)5`qIIFSbuurxJl9VTa(T~Xy!d8s{=^>SAfLQJ
ze|gWm_;?=qXhZzPL9~Z{^hIs@L^^ixoE{LpHhIJsqz^j%)?b_b(32NhC-RsNJ$az{
zv|&*{eaQulgHcZNTTl8JkKTCm(@PubHy(fe^oE|xv;2%4`n4hdmH+Iok)GTU(c>5M
zJ^til2l7QU9zSxB51LmyqVeXBh~9h=*&V9))!)W@ZXL|4A9}9MkLgbzb}~;y{(;VT
zejC>ZeSEH;KS6Ym-0VX?ZS&h->}&K<KgND_f_#Twf8?jGvws-}^+Wwoe?)oLe9ouj
zWq;>5<J~(Nr(fH=^pE|XL-V?yG@tV-|8ovDANia^A$t7L83*|>`$PTKBevPiIU7Io
zu%G$;9UpY~Lh&hoT5o(HdPrW+q2~}?5u!65#(sLUGyUkL4I>|P_>za-kscl7Pudaj
z!-u@+^z%1zdLGxgXg=%gzTJ8mPY!$`dH4-KEt<zT_a^S`#1mv!Xnn|8)^qZCyG`45
zVy<o8`hw)8hxL+g<9?{0y~UA!`GP&H7xeeEtc&*~>}ovw>$fiW#`*0N_Iq+ge)1fo
zul<M}Ap7teery~%@nnBuH*%8CbM*Y!bNnK*gYo={Uz>;C<fA9_`{sz|qc5bdHahZ<
z!}@x^rr$b~N56H6{Ng;GTX*9k9lk|z<+*X_^h5LUH~A&Xk4`^(=r`Uv<71uFWiZxP
z{6}0t`dgQ{UyB?5%5MCGA9&A5Z+y_le)G|TJp7j(=*!OZ<p<jM>t`==<D*}l$<F4(
z2R*seo9u|*a~Q|D-+=TLSH@}6AJUK9`nAas+xnsB_>rR$l8c;>9gLT6BAOSXhvLR_
z_9Lf$^!g)uPF`(DZu#83kvP$hj-7p%Y5hI-y{_*X#hr5p`#^HgkKWq!q?fk%A&>7z
z*_k}_aIc_EZs#D#|H%#U)kYWj;cGs6>o?vw_BB5`_dU*&+Uy$XjnfawX<lev`yBnW
z<9PBye4u?D7QIi6`p}E~+D9M%$Mf19yiN(d=lGL@oci78L39z#!w>0!j@=--h}M@J
z<b=lKFFt&iAJO;E=-ijmN85d}`%>Tc%csV>*JMXXZhndn{+^Q`o$=(+Mz0_8SI_Zb
zpK=~vSig(^a{qZfP8_=0kX+o0YyJxL%O??w=!)i(Z}72x=7IQ<j~}2HPvQ-|=W!l<
z%?tHI`P{tt=%<%FX&w2u`R()Sh;rSt(C+dp{?-M5{p3VXPUyLL*}-#d@r4f3gS_?!
zbqTrDPwGRc{y`7fkKE)W7kiR}eCX(JesuDYb!T^cjWe%t_&|D+6XIh$I^(R5`Pnxj
zdWf%i$i)ukqkps`IXou^e#YZxK60_Ieu$rO?57`+haAS6A6?OS&yA0C=11o_`q+=2
z95Em83*#c<Uxf0iapsB0zSc8hd>+T+&rjG>+q&wH{NjA<B~HALGv0jiDL*m}oq3`C
zh+Wa!-^gKqiuCweH+hqu_*oD9_+7M{Hu>mf9(?hWubdOn$zPtk2XGFB@*{g0kH7iI
z>z+tIdU_eBA6>*apSHYhe<44N`MQYCda!?#r_68RzC~_&LgUcWLmM4CYDXkz5k`LI
zEwBH=dIw#UH`05qpWOUHe?;~|ACX?<(#A)>`J=z+cb{NAi|YT>zjQ5N|FkaFPk+oq
zp2IjUw#6ZO`AxnrLUjmw@@PYO2Y>QH{Sont@|o9jZSz6T@zth}Hu>x?5MR&jThKfZ
z9sbaBh^`2Yi%1?ApVzh>=f_t+dF|)w9QxZw(L?rTM{>zO_7Qaa5uJIp(P=|;<hGuX
zA3q~sMC0*=abA4Q%P%2%aul^=zxj-d^02dWwl+HZf$`4a=+u+?tpmFl=UxF_oX@^!
zKXI-!KfX{M=Dr}-XP%o^o1MuK?d!R9u%7y%I)(o9ArC){`={S8;ZH7lMS5+|*^}Jl
z@f^K%7gv#<pG5oO2l2tTu8>{HMNaZW|DiX(6#vkA@=N;R&pzzFxOy0Gz4@IsJ=}{q
zr|`pge^(?2`<t(bu2MffMRe?HJ;+Hf_lEqyy6|r(ZsR=cZ=7|Zm-sQxen1}UO-}mQ
zC-6hZf6!ZZ{GfU0g${p+ujkM_^e^g1=Q%wivIlwOkxJw6;dkPJeZ)hI5Bj5r#zA!K
z#y`-rGdr;xq(8m%)7Sd(5A=|Jke@;Q>;%n&KRL}C+vKO0an5z}eG&Q&%y(P%Gw*fD
zWxRfJ(jTH{2l_hyk<WPaOZT*}@^Y8&?8owVCG;J-JQ>^ieLrs=`+_)w{8O8sieqsk
zZ|c{M@@tD*`xLq8M{a*#$#eboEp6|M=;8esdE_r``3YTYldlL@Ue8;(HRI3xSUsSh
zKl2A|C=TpD{8zia<I26dNxNi$#j}8Y!M>%=QO~IB#Jl~+euhr`$NEh@gpWF1J!hOe
zBOj@g@sqdE8HZjSMt=1h`Sfc?xyWZ6M2|l@81M1Phrf9te(Fbb?4jR$#<81u<!}7R
ziyyg+i*>8ISif=Pw+}>QZ|kBB$z#3rL(k1e&q`<<I(EUAKcLf34$sy7(cbK(pPkuH
zo80(VFUX!ye?)vBxsB5%7o=B|$NC!&)p6onJ*aLJ*Y<bkE@=JPuL${#^?~Ya^=kAR
z&wW27zo>`h0re_<$PLAVHhOZgTWyiP_?ln7WG8yai_Yuna^vueXuYGqM)|ebm)_PN
zpXgWY88ONc+vY1mcDBCOk-pIPZO&i((0cPv_GCZ((0q_SQSP`t#z%kR-+r%cetwD%
zJ<vIqp)0Zve>1N*TT(QiIG11K1%ALE$*Iks`5SsDzZ%ECte5%uJ^Mzqj@FAEoo{{L
z;rHzRZn?il?(Y#l{>H&g8uwna<nQB8xO2}_Z`fz862HRN`jJDMoZ`{<(vUptV%^DY
zJb&ZIzNe-KIoOB&_$PT{n?FRk(G{U_{LA{17m^#IBd<37@guJ|A)olQzU<0g>}h`d
zjbmSJe9`&$&(M+6dhvJuYP`JYx$hh7*WyL}W51U#`8|L4T?BgZY+c2h=jJyLlpom(
zA84F*^dI_JpSUhX{(%n0{ldJ^{N^)`{rQ1@@}P(OhrZg@(>m+NM}HBb$A^FEC#N<<
zk8kYPrdQO%IDGiI`5<}FLC+yNcIS8OkH2<A{z2ZNIAVX}t-pE24~+fjp*Z1h__G&%
z>7xz#kGys3_6yIgJ-hZisk&W<tvhVfq3XJxhxP2x@SFcuH?Lm1x>bh;*YvDj_nY?S
z9jZE1bv(B1wLPnQb>C~uCRGjo^YHucx^3r?*{n5=x$O5Jj>_)dbDh84c;L`%)h&l?
zKXd52*_<6tKjz1qzL#IH;aWYfy6^dHPMc;YwOHr&{Mmm0*}K(@&&za=-<?18)fF?d
zK2Kcn>pQRgGMh2H)7>Mld8_<<<4$Y8de`iU*-zd-f97Y?CS~8>wEN{pclvMKX3ghc
zX@2CV8NA@iQ_eZ=50kSZysQ0TW53vYN(R64yAE?2t^ZLLanI_V&+gTBOjZdG{p3$O
zKXStO?8#l9eQBrD=auE#V!i1jE_r%E2JblT#>S^E7?l64^!Qr0pZ!KYX{(QFx?cZ9
ze$gG>FZt_``?C8EZ}`RPFAmCXxas1ffAhs(vm>A1a@xR0`jv5$b@%@2j6ZyuM?B##
zuWZtE-4k>2ciHlemxhe)U)JZi8STcecIL;~{pU9L^m}c_me=XPYoFhAPWypbrR$5~
zGf$cH`(^`jXq{VJ@VnIp_nw%2dDnzzZ@P0>_UJpa-#%xzA^C`Nh7B5Y!0Q<~U-<U*
zCtY*fEBQ6!H!L0X>idOw@!cnNxp3Z;oc!i%clvP;e7gIS`CG@0?eNJ}<MaD3ZT^>+
zn~unjnA~;h^q)-0+w}X{Coi5kG@m&7k?!XnKPhkd;@|f8?7azj`yF4q`}GIE%um>8
zhd2NF;;{1b>l!vc>Z9$)<<^7U_?<T7ul$SO<4@nWubZ*O%>!RA%SR4KPV0Q%na6(8
zaX^o9{P2JCiYxwZ-rIJ(VvEP0n2_Bt`Mn=>{$xhkZ}j5>+1-5ji_=GLc=z`IY%wvX
zf5#5H>@=m%3;CMA?7z>0Eyrbh_1?bv{gGpH>tJ1e-|3K1v;H(SAM?cL@2=T&aQ5oB
z275N2JhU8F>^5hk4tKQP_lxX<J<ndf>g^A*XorLDe{SPeH;u^nDLQd*%HzG>I{&mj
z89)F0h3Z+=od)OpS3mv4>pKm)wj8k5D;a&R?zi?y_n!BUa(-b4$bZS#toKLXJK@BS
z^3%@S@bk0AeU@MOyH&f5{oXs}_+ekj4(w=s#=iaOTkSf3m2Ekr|AV)T8J=0k*EgDQ
z`2DYp&EyaCljq)b--LH&Wea+>J@7k+Pb}v#d5|9LX5IM<`NhfWTh80@M}HoXf4%0Q
zen)LNAs;(v?Bo589GQzt@e9eP4UL2NLvbMP&%61a9&gMTl|%6%e-xp0*!!K*zBlaf
zw{pI;p5pGtmhbM+ZI@a3s3EN<^ys%Bzv+?JvUfI~pYM8M*PGY4d~QBqm%p}{^xVY!
zrIYsi@wTT-Ea!`tZ`+~sYdg#?=ZCd#yR+fc)h6frwwb^4P7SA&kscAnQ|yoW?ElQv
zy{;TPBZuM$(i4&w@=I9G8w>Xdeh%fch{o@H<oUNw+F(uw`L{TZxXO{=IrOsMP0!f}
z;!ke&kmvXVv=18}QT+UNc+Yj`U-o_m`Ef*XZ=AerJ?_5vl%^m3^zjU;8;Vd|Rzi8C
z2rKQQ>3RLOfV=|L4c5DK^WL3~=sK`m&+<e2kGyPuk`HMukLtHS>X+Z_WAgK-zw11&
z(}Xd3+b%8gK1WT?^SOH*f5}19a{IP?ywS{kzCLu1vE};1b8$NFfS;ZJ?ydhQ=S6y&
z2cj=R<BCum))wuX;(u{bol=DKQcrGw`qR%Id(qc9wC+$ohVmVhXCS*nc7-vYRcaU6
zS%2U0mv$Sl@0c8t6WVW}e#n28H##SZpZ*8T+3wi2KP<=FrW1$dy{dZW&PD19avR5<
z>PL0%;SK-X>f|RsE}ys5BhIDP%eex7ek?xa1M=G6)Nkf{>GSO56UR*{*PF(PZ}l3!
z<T&xA-t*>9%=0y`8?@QJRbQ6tZuv)@X+P_Aa@OGdR`aqQ^UJUAz0>qE+7}`Jg!a!O
zBqx+VApe8v35c!;*&oV_5!F5FQdqTXo1dJr+q4XdYq;-i)Av2?`sw9!l6fHihvE~8
zf2cm7FFr-6u8*jWm-qRLb55*#`Ga_ncjPzwkGOD-!q+~<e|8=G=2@dJe5UMI_~WaN
z#b5s(=k9yh9^V@FFZU1XKIdxq^B-?K?8m?Us+?EZ**U*m<Hm0+*lR`(#fLn_?&cSV
z#~!+W*X^!pQgSYSx#cz&&3I{E_V^)>t$F+r4NBsQeWQPiTXn2^Abw{($>UyuU&yok
z9DUm{P2T=;__WNv;(TX5c}l!Hw~2G>F0R!R^ssK~2IuX!`#ihhCZ(yF^(B{krmolZ
zcyj9|6Uyi7kB&WS)i(bzs(c<5KN01P+M@HA^VwsYzJBiuGpFX(UHu}z%Dd3HLmWf)
z6MvAuKz;+QBjiUA-^E3KQH0{iJ;I?Qp8n1Ohm6YM7cF)i@YEWU%5frYtfzC7IHRXH
zV>f=se&~1H`?^bh@%ZE%S_g>Ey6_)<j=z0X+_1lWmVJ%4Z^&!*ovYs7c<<85!^`Jy
zejtwUH(ozKfX0~*k^|!7KDr3`10*lB4)D{Pk8bnS{I4_U-Ud2%+rQ)&$nN$>@<Vw7
zM!DrhC~rga7a_ld=7sDA`NItte6qzZhfd9%H{7qt@9HQ1<h(!Z;jISEoiZ)wpZtm+
z#`_oZx(~FU*@xv%`zd^WM3;NpAMtf2PudUMm&u#%H`z;F<+=IgU-PLy$Ww&)$qV?H
zmt5{|@Ry%GH=p$<mo|SA5Aqnf@G-A?gkOu3GoNbn;k}Q(Q;th{M!Y&7X^VgJ;p2Xo
zU&>476Tj+RZR3rj7dm+Xow(xP@(_I^dapnZaURirZ2zSXebr&|H@nN%_Hp;co~zTy
ziB3Br{i5FVhy2I+nf;8HugoXjV|?Jp?h*MDWFHvM>(Tz|Fz<~be{~Y%_b}2IwdIwn
ze3N~*e|l2+{!%-l{A1s>ukvg6=hoYMJANWg_=kK4tNVZT<Z};ADZhtefA%&XKaVKy
z$|Kg7--<7CnAbS*74M<x2l3&r#<3?qGA}yvL;FW1RG&foijaOqNd6+UK1E1=XkTV;
za>Al{>Z)IUmshMS|Fs_OHPlJ!es;1?TQ^7#dPd|w_I3V1KYn4K_x^xiLv+x6D5Rgb
zD2i+JmC!u&rx(4o)nC?kQBhun?8FZ08}?QIifixTy%!`0{lzJH^s7_sKk9jUiDUO*
z^z>Z)E-v{w{?-{E`>OrNJ}A%gPxTo6AU~16)h+USME+}k=0Ei0@8}|uga6P|yt4zG
z*5=M$O}3p=e!puz@<y!WcgyPe87B{av0;yy<?k8TS=@_5IQXIuhYc81Rr6uvd$${S
z_JZ<#3%lOB{_L;c=-!}49dEt)uXy6${6-zXFQ9cYUL5i({-NIIclKd^YhLjLhi%$+
z^fs5yD!<2c-g4if&Q_N=pJ`iH@ucq3@4iDl#jgCrIfFjzZ+~Wo#s_@Xe(fI)&x%l9
z;GdBF*_l1n8{)`0OFWVf%Gcr)iVJaKoyafl+i&&UU$=OET=}~z^_=^uR*(E@y|XWv
zogLNqlxF9>HmCd@ko!3HV`uX62X+>>?w_omJYc-I#rKFVT^emZWP16#GW6Dw{^p7K
zhTiT+oomQ(bJZ>DwEkjb#xB;?dHkH}o_*GS@Uu+5wGXIY)Vt(oZ~Bqbdjoci_BEgT
zMfckK-*WtB4L)p8^4$EZoz>)w*=zjwcW3yX`J2CQ`oP3SCH7!X>%rgnt9xYjc28=3
z-79+k;e5$|#O0;W9D3c<E^{+=KYOq<`S{Zlf7rZ7w>KM>e%pP!ww=~&TpIP#)dO2T
z(WKP*haK;}^3tl(@DE$=b<bf{CEwLeoz<w@b>Cg3^vBnp`ta2Y8kWu&GWd<{zdJWu
zcc05Qd-3=MnQ`xSdTh%soBvndcxg=M?8*kEm%H}rw9kf(O6)~n^$5SP&s!J%;JnO!
z_GRP6Gr!?)4?j6~;1QpW%h}t0FR#mc)`vZ{<qdhky$U|QZ{dgLa~~^j@k9RLeCnRk
zc~L&{{@Hg#_78cv!@*y5d~4v`^7lUUjroSZ;w%5lqxN-ed4rzx;_vJQ)j9Muzj4rb
z`BZ)+7rNWB->z{`uhF@4p!WgJ&+d8bTb`@C*^@og@#;Q)q7EaM{H5Q1YhL-oz6RYR
z$rDh#iDU04#V`Mk`=$E@`ON*1@em(*O8ooYNPNR~U!L>$W!p3;*|&`+-|2gg%*Sju
zBcqRbim+1sMgAhRfA}7^2>FNji>SU?QsgfY`7?A*s4L`O%L?U3c@y$OZF$Q3Zuhz3
zfW7kzb~@|P4`!C@5$`R<nY^x^k>AxJ(ci?++AVvu-1N88%JmRGvah=zk~id^=pXXh
zqN2P8<%dcr-qk7g5%FlhhV-|;LHnD0<a;Fkre7ZAm-^8=KgfHY!%FeXPw<Cvob`~;
zor9qLL7rwu&#fDI?QizGMMe2oJ*8gqdk1xu{g8gz?85K8@8);vS$6QA1YdE3ul~4B
z>I(Nr-ZQhS_mlip{^O7QK-+%qJ6v@QKf<TmzE^MGbf*S2_IJ-Kq4k3JMPxt7pKFWO
z-+OF+A>J?D^~eif{c>gw<r&Bi)qCm$a@BWSZA$ZF`d>XKhw`KRtd4~I)%i)iFCNr)
z`sHtVl0Ep9^S8Lw&%b@Y;(d#@dI?%baW4KLKeSHP*?Ot3J(nlM&yu2eD?)Wq5vmi5
zkX?(AU5ij2DME2x3DudkMf-Fmlou<Z{jj#^`~}4~|CJx*L3x$k<SqGuKk`p`gx@(|
z$lv@~{7k#;=EryGJHLDnBEF!wk@wj}9>mW*lJz%UJZl?gpK>0s4>>Q0Yx~vSa~JHk
z-o2AD=O5!QZgcdd4_)(LzhheXdno$`l-Je4?2Eti12muWEdAsKb-jBsa-g?gTOaoV
z&iCf!zv%g;`+eULum?NIpY9RG^QjMy@AHQPCYOKrgP;4!xc>4T`R)IZU%SWq;)ZQL
z8-B~w%>C)Z2ey7<w<o?;I<oUeBWgCUDqY&T-!Xf=TUBB|=PY%v^PGB^z15AzO+4vO
zcicC6YWaRv-C|z%FwS=o>8YNN{;FRdc8~0w?%d69+<&M$)idf+_Zs$hb&UM(9wg>>
z{w^-;BlfxOKl^l>4JZBg?@&+p_NM>X=&dp3-(k{&e^_tdDOgwEbvlp9@8U&X(XTB|
z<$HO`IageeLtRRK{%T(1)YtNlb(bgT;d~`t(0dNuYm3LoUmelrg#%`+|I5)Cf96->
zP#p0~`G!8m;b)vWhQFhCPU7$U-?_>?hWkq>&Q|%L<q1!GZ+7M$bIPN?ntx*Ns?t3t
zJUF!335`pa{rQE{ZtS~SX<&zMud&*BjY{q@&EsBj#3nbs{@QDe%J-e_QPH^v)fQj$
z64#!)pP?^0_FzZ%G5A|Q^V3uO(x1QJ$G_ww{$QWc=11}gKIDvfz<B$Dc$63HpHTeE
z3*I;R?gM%cCZDYP#<i`l|9DQm;YUxt|L$guYQ`OM&>w%Zb;BC()#M517T<fQTg9n;
z%YIBA`G`EehnC0bEiZ{{b`m%05Blq8A4pF7n|M=~$P2!6!PmLhy|?p^`wsOudz%lk
zyLGef>|j3k738<B@~=FHAA7r>mGAssMBXEp`F+o3zp($C4?X?8=drHjQ#X+-+EX2<
z?xAn=H-5xF@S~^s)FIZ(zO(r~^FBCmtGVTK8ve$+2b15#C3{#${zGr$(Xos5u>UzP
z@hf$0v}YX8uc0_l53?Wti^!hj^W1pP=|LWLi}6W+_K*kp6@MPs<&k#R{c3j3zt{`i
z<sTe;b+fnUm4E+B59>@0dDT9{p6XS4u#fz}j?np){pAhy06VY~zcrrx{7;^+Pup+p
zuj&?cf_(vhdE~~gcl}PIwhc@2xwtovHapul*rgB`{?6C>sW%`yh$H>tMm$CIoc$m>
z;z!RIN7hfje9bS7=Qs8R>tQ@P{$zgjv3iVrkiDW^@ROhDjSj!EUoNz-`S>kF2g%1S
z^r8<mulkmq=<j(%c9f^p#pI?xKU0tDM@L_JX_H4AMmdaA&+v1I4||Jy^UIs~n1>xA
zdf%=dRA>2aKwT$3#n12i?>}wJZDwW%Y&-tYKc71%lV3cC`o*>LEB@+c=TqkzzdJI{
zdDeGr@`3O1@$ufqKJR^u_Ym@d`$YA+bCL5iWJh%@KF;6qdwTvLZ;{V9buBx{bK2}i
zU-P=(r#Jc72VJbM=_UU;w^&c-4ROuBMSP7Xr}eUa*44QlKY3n$i1zpQNa!KHAU&Lm
z+?%L3=x1JjTeio-zZ0e2ryqVr_oAu4eprBiif8-5(nfhz{qH+2_pa(V`<D7myvxV@
z4ZZ!^ccSVQb*}GGj5kgjy}HXj8S7AWfamr(`-!&wMx2Tp@kJl`Q{EEK5skB7$)omf
z_h#&9U-I0(#s1`?H+$2IKJqU<p0gj+&tB-EbyUAPZ;(s90qaMUBq{330`iJ;tvGUz
z5dBp>=KO4bi}A%j(eo#Dpty?hV_#HvMKn%*YJa!ixqlVMP+k)6zJH11?2{w&?@qq(
zl=<cF9^KEG&pn%ZOukkJm{(m$fA@gp?<U{7xU~OStCa60-7`Y>r|K-{0d4chJN6Ix
zNFJ5XJf}B%(!;%``c1u4gq8fz`QG{a$^0*mtvYO0`S)bb<K7>*Pw@WJ_m<vIxYzJ~
zX0I{(PTKp%S-HGr9^<?RwBN{s&iU@I-hIDMo89l8R{lF@_AAfjNAlZ`oJ;(k7aw|)
zkKFDl-3u7+y8!P~*v0#DZTROS`}94e|NPwVnDDo6x7zlmwX(0L=k`~KA0(%?`>tqD
zdRrIwC)N#r{OtSW+T`p9hO}wXq$XnYALmyX_uFONj`C@{*Y@6pyyDaShx;J+DDtg5
z=>A#1{N#IY@hablOL@{cL|q_%%e%&@1LPs+2=<gOicp;Z#e;iec}D!<%fBP?V|>`l
zJq@H6{*avVo_i_hM)}&`Gm$5u^OkY!D*ssrd}6=0^A|dKmK^f8y3TyoL%x%j)P2^G
zzUb|H=AplR+B}e7%JcS9=STauI+fkko#uDH@9#sZAH{b>aYC+@1?lDcK6SSGhu`y0
zdb6WEz`pLu?FaT1bo@a5Z65rrgSPyGKYzlHyyg{u)*Ze4!GGi_aT8H}r9M*ksPFjg
zD%}>mu*vMIlJC~}sXQmILG_$^McpPop|}!v=*0=UIcLc$)(M?_NnXz(J;e_>y{D!x
zB$w~0#i9I5ekdP6bxW6dgI?=+@w|NUmG}1i%YM_#e+NijwO^}S?C17tZHVreb|*Ex
zvE95}-6D?piM*trT>Oh)LUlkyex)w+UdBDN`cpnKue!r{`l>tBanSzb_n3z@yX)%l
zk5|=@cl;k-|6ZTFSE+G6@&4I7^o+<}{EK|zNZk|TSUjr(pu8Y1?T6x-|L{}wf;cx0
zfA2-vm*0?w|GH1*Z|o}1XyZ>_-%0shncvO%P69vorM`y}=kgLc;(Ycu^P^MW@?&;l
ze|+sv>L~V=Kg6B=ntkOfb(B0OZsh^v`6qvpH_6W)&X4k%d}+UyH`vR3<aWMeZ}#=O
zJM|&^nvXvGo4)v4fBNW`|Kwxxq4&O|2-%4}<Wc(AM|?NP4(w?@^2pQt9^!*9yPD5<
z_M)ftrl<L=kABE+ymw`9`derFocBxKGnf}Yb~PV4(UG4W$!9&u&tL3s>RI)u^-|9o
z&mQ#BPcPqv+27Qye)q}m?7!mM`!w|Ip&vcFpu>;flApbxaYZQq*{|f4zHg4MK4`nD
z8o!T_ALUPI-;ht_%}Qt=l3(%X|Iq#*pUQ9Mx1Y*u>Ms0G{<1~s@rDg+<Ujj2dT}5R
z8mDbPp^tixzT}h7=%b#9@}Z}X`Ro(;+lSR%>;cPpdf~rsLSB5$i{84L2OVTjdDuCX
zUgT1j@Ehxo9v}JBdD*<|<{pRr#jW=X>@JVFmvN89U+4+Fe{rusPW1?To1b2OPfQPX
zf#hdTbnI&XwEpPTb@;Qd^-`a>|CFcc?fX0VPrh=W=>29yc~QOPUR9p7Z&)Y$rFo3E
zKhVQ^LHXD`QEvN`dW8S*1ANQ=yzt-WL2rKfp4{f;FX|3?k6oQ7?R)fRNB7m*^s-Lu
zLmtRpwMFt3q4P0676;;t|MDO4#&7r|yV#G|mw(dF`q=l_haI4Olb!fC<TsGLp}Y*~
zMLy_UCm)#4y4%<3Pk#Q&pS0Db{{0tv;G@11PyEP!FF%ln9KL6;{^p5wjB)g!r#AnS
zS0MTQJqdLL{^}3kgNO(5V!U-m&tBFE9rX9R)fwsvhz^Pmc7gmUVo@EUpPUiZMG@&y
zgyKnl7BBV#`AFM)Irs1G4SX*o&-$*{`QN>Qao$&pOMWSSea9<)@bw;%JW%|r3!vYN
zpc}dOCL{LyY<Avn+pT`la{UFl^ZSE89Jk$5hc>F&aoD$dZ+6gHH8XCR{hNUeS1<ov
zli%$A>>oG3W0msXA@Tc}Xm52fe}L{q#gqFTb-udUcVOyf=K<sJH?R1>-+c0v@1NWs
zshj=%6>(}_^N2(BGvs%^BU1lZZ}l7hw{G?Y`;G6u@G+m?X{qbvX>sdbhd=vnOxry2
zs(n%3Vi$H4SMJx1lLy7Q{KY;4$K1DL-xl*Tzx(*ZMLQkx$3_iGP@KzuzROZ)=(jG;
zG1iqo80Y?q{p4%sHhPICaq1q~zGA=fI{^KDCuu#!q4gA>>}#J?Ke}gg{!o{x525!k
z>RxrIdBv&kM){9CY2UFP){A`dyngF$pRq3X7kv3O`|&sa#ShWR>)QOC-0BH_Ee_>9
zcBGH*8_6q<ox{l~ANsu;J=m50<~5J{)BZ^g>&Nc+vmg1$;T&9EzlHz)oOoeB--D_n
zoagbQuYJQfe#oBoXYpeHaGy+m{$n5W`wzcoq!&L_AMroGm%-2P0NKfTM%}6I@&2Cu
zy{}Zi!#I!lp`Y`X`NXGuXT8mDf8=lUGEQD+XZNe((|M45<oA65KVdh|>Fc~HPaCJ7
z|C`r$#PSyZkPrER^AUU5@7b09^q{Zx;UDf7*wgtEKXJj&+0XB|`6K(vXYArRe)6z9
zV_n^=#Pt?G@~ilkZ@mB4j`Q<Vb%=OzpDoYnmk;DMZRZJbDbKO1{l>m5&x$YWEdTQZ
z>*G5wd7M7Z6Y2o|WxU^0^B4O&e_}uWEf336_H}vG_tx^Dd<@ya_wx2(`!D(VgLve>
z?CGA9Ui_lyyK{0or#Sbj56La>+P}5gS)B4q{wFTbsgwD;_;oIZ#`72F0?*BdzqpWx
z<wyC2pGSJ<T;B)yJ8bqN{v+<?J#^xmeEd>B|Fy5Ehv=bh<$uPBS9Ii%kIm0c_{d-E
zYF}Z0dXSSpp%X{s<>%<dTU>v3F)zEbH<V}OJ@KS&;6LcK%}YOiNPh9)Ils04TOabt
zOZ=X`{J{M)yV-Z-b$&zsc2_MJwC$U-Gk)zJ+kD2eoBcq&;NBFy`q_N|v`*|tKlPyV
zj5a^<-jqJ^ce&~jc9qA~IrcSjs?VYQTwL;3`nsRz*Zdt{`xrUc7m90bb`~GjS3f&y
z%MZqz$2fBGGtc?)qN4HQT>cV2@+*Ifbo^a?VBb-H(l_csKl6AGsUG5I`q|ra_HrKM
zf8>!D<TLB*x%HH1#Syfh^8@>=dFjc1>`xwkBOl2R{FNX2T|RrLQ}E?S?5IwWf7BPo
zL;c3{8~%kK`_Rk0^k4_}#HSL9XZ-Lnk9p-m{ZL+%N8-LgKYm3%{-mFt^wDp<_!l|l
zOYvnrwAqhb@`&gB3C8jI<q`a#IzT?>C(gO-!;i!{{oFr_M}Fem%b%@}x`Dsy_x%hx
z-4D`JUU1JsKHu&59@#qDzsN7o_&qtf>F<6<9${~E<Ryps#fLm)|4?tLADzGT<7c0O
z<X}JVfgt&<7x{g+%|Gm);@y60f0AF^`x@`ND)#3O{7Swuzr4vWb{_J}Yin+sk>hV1
zI`cSB%Zubh@40=Jp7Jq%_EY(RUpt@4zwVRS&$-X{`Rw36+If~gdS6RUc2=*GhrIgf
z<2m`P55Dwe5B|ZfaUH}3WFKvQh>!WmVSU)mdXiTgva|J<=hR8&x9@4=M=x!9;p@3N
zkbLBiD6jD+<Ji$W?lJfQ`P7N>gSwww?l<H$_H_QR&)HX?IFmQ*!}2ARkKB{WgYt`e
ziiq-w{fK<@^>=o}xw=+<l`r`}zqseBJ?8%Q7xQ!9oA6`u^IQJ9-D8{Gbn_*RYSho@
z$i=P^#V<egTpc0a?3eiQ6YIb)@nL`Vvflj0xlmn6UgP<b{o1+7e&(LQ@3*|yai5`W
z{fxJ7(g&J{9Q;y0ebK9D>~HQX-21qPF~5D%`xX0%{muIna`QuN-*@`H&^-<Q(0z^f
zWZsMMQ|DRtwa%UVA6hr|w!iXk`=b4p9{AJKI%(sBzkcJ$Pd|SL(!Y;DZhXC$qOW@_
z_fxx1{P2no4{cgvU;OFiJjy;09s7Bo=bnqc?g8;3m+|Th^LuYdKku!4hYwd<_&da3
z{yZCg+TN>_=tp1Qp^{6RJ;}!o;({FXgx;I_4$$u%y|3ecekUpJyzkrm^6KaNw_3GC
zU-c7v;p?8AACp@if$|R&e^6eB@;_96LiGVu??d|ow7)}n3CiD4o`Ci(NH0haNN$L(
z63Wl&8+D@JKl)wfQNKQ8!4LPGS^jsDJh!hx_j2-E5&90I2;CFA?{&}X-0U7o-g3|F
zy@vCG=lqeM@@ILnwkR(6r*(6m@3}Y<ck-ULJnkGJ51_-x`)cFFy?)=xh%<b>?-sZ8
z)b@Ve`$G9jK9vvYMLx*B=99nV1>@x($o`NV-n)o@`w2g?fAV|vmvg&1)_T~t)nm?G
z-jDHDaqPauzRcgW@nJ9DlgrEU+I#cwx@PG72Bk$s`5aoOi1H)ke-YU|B0q!HD<c1k
zXuYBP0Ow5ShjLx^msd7vy6%a&dd_<q>x94eJ@y~_G(S*RqE~l_b9Di~WH-Olvd{SU
zvFs!K&i6d*${y_JJ(cm~<VWfn{^7nszczWTvv^Pkq4!>tJoKX%dj7{x#S1yq`B42#
z9(4MhAL+;L<dV;<zc`a0JXfEoW2}q(q)k8d6u#mRU;8w_VMlSvKIUZ)e4urWXdWoO
zA$jFp==(?O80!-Cx4g|B@*X?$b0{A{@nL_L_pGaPhrGxx>L2%??8M%2etAUxXJ_}L
z_*!TF!H@VQJ=qhT{LTNIr;Mi$Kc+7}*42K)F7|)=xfhm?oF9wOIT$)WLiZWax<Gl!
zxm_HIV|9$W#rp>Ours~H0ev8P`<=LTuC~tF&Y}DxJ}0Ms`=jswt&8}F;*7spM`)b3
z_2vismp`HBujJqd+Q#E&J<U`0lZAgbU)y-|v5UIZdWZ}8L3~(0`HKGXg7vpgpi`gO
zuf&b@^gRW7{M1j*hvLdSzLSt|pt!Z~{pQzy|Isrq&n^GGL+4?6%Q$(;{^GnPU)yK=
zJ#OED%X{P^zjLMaat`6&(Z1>f@h%UDH}BiE`H%Q^PI8Vx=ibu&vF}{SWnXYV=04Op
zjbC_wDBg`1&*Bxj&#@j*es!P2U#yq>kDfn9`=Ymxc(2dD_?7%9j=V27k2vSg>}Gw$
zjl837B$v3ge&P)sf8gJ~gI166b7<Y6`a;`!@)PH4>#k04j&;73$DQZp8UH>b|B+9j
zyyE;pp16;xgWbcLm*4Y0^3sF6P+l)W`4+N^c<}uK{rR!{@1E7~<jH5>Lg${@@8t3G
ze$#zEyZU!Y+>b!x*~NJKtabDJl<)ugZ||SKN{!!pvWt1><KD#Ioo5g0hR*xRi0oep
z#fN)t{=g67{%PImX<uazb|$AbRNrdTi{IG4qJI3qK4E`BZ+`k(cl)_~CJw9%q!<5y
z_F?fL{wtw<#y;YH<_D{vGVzpY4NLY9`<3r5<-JO1e>2{GNpE!nIh_B*5%j*v`Ck5p
z&Z)-hm%rVo_`XmcbRKZOV;}I`_f_O4r|0sjethIr`?vYo!@T5Ecgo}Bx34?@st@oZ
zhy5HMerF$o>L~ky^8x<kisxc_*^kgc^z1-Re2q6hf4AT9Yk4N-IeAdN88xK!gdY7C
z<mzj2V?Xh}PTPD?oUj)=vIjlHG5OJJTYq%WJo1$Iw2zy|yymm+?CO4yAF#8w^Oo<5
z@kM8!w;tBdeF;AiU*5xdU+?{`-;Y@*`<pnn4)U=4X<gJk^0>UDUmfP2kR9C<tH0In
z>PPv5{}<h3lf(XH|D~_`ll<y0`pNV3l#lHT&~t6;&)>))&b86;6Z@6AR^3M)`-ywn
z$cLXrWN-FoZ}#xs-+iq8);>!==K$~f<32<W>k?63uPv&V>k8Qos_*%OdR{&!ul?J(
z)pttP+5WBGQSYj+{C-rPV@KZ&sE3_{#2dZQ8OI--r`5^okvKp9^`1;!LY|1m`_9LA
zA?j_<$;F@0`M$^b&U53n$uCa%yL{z7+j?0KdCI!@e!{xQ6TTB<FXu`8%!7`<n}?qE
zBhTdr=P!D?kN5qIc;Y9<^CPIPx4!bO`T3Fef!2$^iF0{~ABaoqZydhnf$C%HVP50#
z@f^m#ld2APUUVMueT;RsZ}S^@h5Ygsdj3XEai>m)`uQ1uAg?+<#)I*Y-0UD8)FbHB
zDdr`=`o?(c<s4?;F;BEJ`S_dtihr^LIryo5dTPt-?n}r4otOEO{Y0C8iXSMBJ!fzG
z9KW^R;zZo2H|XDN!7l5z_@H48JF~C#q!&NqzwD_VL2utTpY`Ey;)Z{)mwxw8?gzxP
zes=I&K30#4d-vm>>zBXW_p2AI7ymb}@38m<`>P||&+!-QXFMdgJZAkMKJ1`hT=RGH
z+rPBsHFlskKJqdB&BHGI9G!hyzE)p4SMY0joqqC&dpY-u=)5nm->`@ML>r1Xs19}R
zau4XdpdJ!W_9Oaeix>XO&g>wc)5|#ZqH~D&!(W>}S%3Z_e&}OewbcvqKSU>gh&Rvi
zweQP+=;XH|bS^1E@`+<~=9hQaksjnVuY9EL(jV#VoBrMhI)5MC_s{a2@1n_JpCq5}
zOU(=Ulk*z%eU5XT^EH2S{uT$m%i%x#U%bgH;!Rw#r~D=^`CUYHm~|De>K}e4AIZPy
z@Fx#{#n*HB-MZKZ%<Fq|>muJ9r=LGspAEY2*y)j{8`M}=`@HpzD2~04_Wg-_0{1P}
zC!+O)&btsFc>_A<*A~?m^j9Z{7y7ZI{ZHKZ_Z-pLcjj&O>QM(@(Xht-@g2`ioZaf&
zRm=Y_DSx-#{7^mQ_XPe9|A&q5-EQF73v%+=|J70M0r;zW()rbTi!1qtpYTWd-~MKQ
z;`j81>LPUnJMlO3Xv=%@jQT|$unyv#yz-fG{FJ}T8|oBo_Hyo!m-)SOhIo?i=x^VM
z=X`N!Klfalz1T&3n}?r3@r{l@$S3w+eA%BK5P$obx<a1NZ$9g*KCzzkm!FI?zkQg$
z(bIhDH2$Y9p^yAYUwqVo_IrH1Csb$X*LHtHA9CBztqVELi=N!Zi3fJ0H@mTidm_*6
zANcbN{pQ1m|7nv`{>K*`e~IhQuf?C|<a6&JUc{yNa$XQu{7D?iyG1B2+0Q)eCl1XU
z`5SK>di8=lCtvUn>!hCa`xEx$XX1i==*4c3d{Ex;yC!Y<UcN%_-&rAt^Be!>|NOvy
z;(SLwdD49Pp>-p-eT09oyZp`{JZD$-=C9<_7Ps<|eMZ|p4b@l972?yr@B0Aj1?}7F
z2ImF*jCU_7FS`fyen7qNJOKHd`UD?(h%<Q|AMYRNq2BPlfpY}EkL&Dw35|Cy^B$3$
z{FHw{=NbA&Bp);nq+dku7rcLi^or=cI{#^Z!lO;kY+O~NUb4@dNB(6W`!mF^2(6p*
zvUv8MlU?{pMEgX<Sa-&@cq~G8Q>Q?FLoRLc0mY;BjdAS#6@TJC{8e4;J&^N_xMN?4
zzkI{4z6Ud(^ArE}o)3TT>BwU~`tm>V8utrwvx7MIKA67ZlRW77k@|yw$|KtN>-XNz
z`(p1G$?1GyACmv<Bla8iu>V^RdCfR-m`7gm-d!GIXWtvKbHoFN4t{jg0~(dT-zTp+
zjh}f>uT5_MPLckI_ECNj{l@-F59@5du%9j}nm_uZbz1MUA#3;Q(X51@{Y^coj<cVM
zZ+!m+7k+;!Kd?9d760tTzSh-uhxRG!CJ&&Om-(xH@2R2s!2WJOS0}QsdXT@6)Ayg&
z!#>U**vCDfc=4XpeXRJDPrYX~kNZFR`rQ*0XXYU{Kj&9|M~bgHf?Vj_%Xpu!9<VR5
zr}Yzu?CRc9{E0_?>U_g*%m?LV_s#aNN@yL5&^;8CXQ2JE6521|BdhiAefJ$xGIrp1
z^1AOS#FfAQiXO^izRzWEzeizj?|-d>{AxeoU(Rdn@4i@ki4$>(9v^wreEgjpzJHP@
z)luZ-m+q~c>)Bg9;hY8SKm3Ous5_j)p>roa)p5r0JIFtr3$3I1`5FGs!}1$Dva@{Q
z{Gkq!PvmX!<s4|7c;NT;ado44%<H-LEaY^4mUrya^mcx=uUH51Zy(_Q?1J8NZTSu#
z=T`U4+UWdyO;G&`)t~O)oHPCV9{8eD7ov9$NFRAS?x)_<+Si=VoG0nY4#taXe#qYJ
zDt`Etw%_e*yXTfS#IfJOi5vBo`bgYCa#&|_`2N6sBmL;XZqB#v9qfb7xBA(|e(5<o
z@_WeN`GvZ{IC)=P$3LCpjTaB{qxiEA+xNwTJZ3(5Kt6-?;K$~%-r|y7o#UbPaelNf
z+rNFMqt3Cu>U(*VJobP2*f{y${)<l$;$MX9rH=ADdG)w?{Qg7U*RO6MXGHeqxAqV6
z#`^)|ts8r)@9B?D9N;e=)raPBUjVJE{e-``zu;f`?T?V(LG=~HAL?iCO6a{7{^X*s
zeLy_QPwd41AbRIE$Zy3ZyBeo0e~MG_5z#$^_bBeKeJ^95lUM8m^sqmm6Zi5Geo!8h
zztGE1_7QF8ZS#`D{f>O3pI!9FapX2HIgGPz>Ns@-|APEVo<gV1k0AOoZq(t9*86^u
z$s167+DF8be)9PplK57aigW(L|Db(Gy@@Y>GoD|nvz+gYgYGfZ$MU*y^c7F`3HDO&
zsTZ}aFFw|@Z2yJ-9+mnA%J=qF`z1fHKKNK4{Y6MG=Q?o?<sIvzo|UiU)wo~C1L(xJ
z_lNc)aYjCN<fryQacdtJZ}eps`QiVucOLLo6xSaIR4k|{MeO{ov0~T6wpWd)c@}IK
zC05X2iyAw|Sh1Hx(MaqSqtRHgyU*SgqbMLMz1jeKEdN*T_ujp~yXVdB-uoW!;rRK;
zZ_k-CXU?2Cb7ppS?>_8Co_5LS%tP#CUNet~3-r=|>?4lI_r&YbA8dE`>jOIGXrKHA
zBFB2e?@4-Wf5xw#IPm9y`xU-}!F%2OzJ&S2eGI=Vp&s)G|MI>%_4$4H1Do}{{;dZW
z%q@F)|4UC@anan?7rlP*|83qm$M=J7KIqPIiyyRLj_=$s4+`fSg9a?}Uf+8I&J*Y#
z_t&f|d|#jYM*a>De&cr}oEH#x<a^=`zcXJz%_EE(`y}q0SjRZ8z%R6e9>$sZ%Ka63
ziF5i1&3>ME<UWUU6y_KEYwn9^kGzOIoR?V0yh9)H$b4r1ihl9|`3QTkk9os=1M`J^
zhCg}U!~FvGGS3wmH<0^DXyzA)9UyrQ#J?c@0-4Vsc@xxq0`Uf72S^^`y&^@%f%#4x
zP>=B=UT6nD@cfl^f%Ogl60gJ)$aoOnj637U`h+}ujeG1yj{L|tV<+nnam9GB599oV
zeGU6B{LOtI>mKtFz4)7P0~uHLS>$uhXDDa=0BN7^fwE3=-_E`YKVl#Kp&zu%II-?9
zUigP~jC$k`+69>}_?`PU<{R^eyh0q{FZjd-`xMp%?kk|VUn39DK5<Dr(k}7G{)u@=
zenF4cXWv8|(hm1YtascGu|5!|tRL7*zU8}J<Y(+;-$nc|4~Z}KgY<{>i+PDY+K1-8
zjefFT@%<g_Bwj)M1JZAhaZ_Z!!FdnQNr@-k^CFHl9+6}JLYxub%m>a}SU0IhyZDiK
z#18gd+)tn%KIbgF|G@kpZqUzs;k*@n<Z<p>*q31+?J=Lw!#t)xAojEGXS`T%i68P9
z>jL?V`HbD<FXlDzLf+w=lXzvk+21hU#3S>9a^jYGgFH0I`6~Np;*4_gHvM6~v9ILs
zk+J{ece?D$IA`H}h5aS<kW-rH2;r|*eWmB!?Q;B%l6@=s_`Nab^0ar~=EI-d^za`8
z_NO4f2gZKd;T)gyA>`=~erBCxoXP9lFLDmX{9wN0AM9dZO~1GwW#7&I7MkZLJjcdv
z@+0jsKIp?Aw8Qxndf~HQW;~E1@AJEH@)-T4ebx)&k8@tm<=Ee&4`g1mE`#)g`xW%i
zZ`N0y_c1?+1L`sV$j?0QCvS6p#yZHjlQ+nZoHLUrXrJ|x=Z53~+GV}N&-jP@tmpjf
zqj^rrch>NiBI`Z*kGLdH^4yyD)IpvHlFx`2{K5JS60h`=^_F!NnsK9_<P+irdGZZ;
zkLLxfmpuO=|1iGfUDg%yH{~Gxqh0npw2z&{19IHoq6b>-<^31z<9wU_3-f{dJjRuN
zaesyVj3fI+)-U3gcw@e>Zh)+l$gwY{-`tOJe}w;7Z?KPb1G^Xx@;~bqc2S>s!FVFa
zz6qNBm4)=1_Spxq9<k052RuJ#{(`KZ*v&X_f699S#2tB>@ne5S98({8<_+`3LgI>b
z9Asa?I>GZZ<}Ld!&Z*dUG2Zxv{RDBLeq+A_G7cd79nQ1)-SeLPx-Z;ymHBIVZpr(0
zd`F${(){i27lXG}`YGfbn%^~$j|N}))1&L%)uEO=$M3HB9Rbg!XrKDn$Gjk~GQO<4
zApRr&G0xE3KjUxey}sq%>;L}fj<xJN+0XJ^j(Ng!SoV=Tf93v^=kDBB5O>Tw?!y=-
z#*=d{&eIux%2^MYSHvCrN#;HKW#%F6(hv5<ycbM7a6bzYcleopaDRXwIUisilLv@z
z{7e2}zG5Hs8CT8;@HhRYf8;g%MSYNZAa>#h)>G`lzswi>LH@yC_?7*?&NuQU^BMiv
z&v;`O`H=RR-^hXZllVuTa|+@ZBrlM!v7b0mq(5pGaYjGUPre~9Vh8pyj`SbD(ht@b
z$~pg|J^0i^FL{~v86WhKC-lA_yJ;VL(ZhIQ2lg}Gj321_(96Dn^{KTXeqx`>xRYme
zUf^Fv^nmQ=@dM+4zi5a1Y><46JpNZ?e8}s>0r#=w74{*-F?ykCk9o)bhkY$_^aFYN
zPrM)pQXgd8nTH_#!XK)ic*kz?0rp@Aex-fn$*1_sLhAF}nfXn7<XP51^m85xpY?$}
zMPB9{f%m=WuZ|!4Qq~Q|51RFZ{Ts-AH~zpr@&xNE$a=;4z&L`83-atEi7&>L^Bv+I
zq&{-&59vSa8TFB4|B5{82S^^Hf7nm{;l7eM!vDlI<3L<6Z}1cHtc&D3;*D`Y9zOAd
zKKw}jreEY0?u)qJVjs<Zf_7N1I7eXosmFeYa|aN+m=CNU#3lPr{%!;NQ$_4yyyz$Q
z&5RRr_?0-MUE&?P7;o$&&KXzM8{!GXfA|-_5|_+Z#(}tD9m3D}k9`;SC&)2A#2d){
zIyCh_?sGYZ;QeQib29FK=_iPPLH1qvl{hBO=`a3f-jUBK$6x4WT<|}BC*Pt6KM-g1
z8~+iXN)xBdOOQC=KAt$IJ?z8|`o;c^d5OPteMS%c!(QTn{Tb^KenOsg7(Lj_JVqY9
z#3hJ7(MLa+AB-RVp&!f#{KPyVeu;D96F&aKKg1{cDF;7Y^vA2_bYC*Z`6cJWoa^cN
zF6S`355V8$;d^+*G3U3$H|NltPhWrPrhRUhuU(G!AUOBqoRM=T>|s80K0tpt=e&Q$
ztv&7v7RYhl%Y36gd60f`Ud;13>X8o^Kjty#c=(%fBMw+ch(GFcpUXN<o+NHruR-=1
z+}|_bc~6{uIrq(!b6?H<How#7T$cNE?#nqx;JzGv#3^x)etzf6eun<AZ=k=N*YVt%
zIAwkjcf4nTKN(;A%y|>zO&oDf$+<TBbjFu?$ofdToKvBP_jAz4_b7Qkjr}@#4tt1i
z@;dF3_lRTmN9?=lAAIH!_vfs`>>t@@;b+!c@+A3)`*vu`Sr6D(qmOe6#({Z+pIHBR
zK7v2#7voD@bH2yCAs@1SvoB)Zzz^(C(Mul2Kg<ut4L>U~e;F^%)fqSDJLl7kKX&0)
z<`3VWV131}%unVy<4Ql-U$UO^yn;N>cof!;{QHIY3p>dF=qJzOXXXoh<QYHOW8Gn#
znGg5@|1kdek-X3PK|aD?ApHT^_o0vVhH)pZu#0-==ey$UL&-NF<AYy0uOUA%ZtQQE
zSKPlbzZf6#9OK3~u<qhlXzXS^R%Bhbko>?n@%)r{Wu0bxnHS_GX!b2UPhmdu+>1D&
z9iCH>C&_oj0qrxdXpg*0KN&A*`bj)PQ;s}-hObC|f{#AN2|oKi#tS|)^AtO{FJL}E
zv(EGTAilrD{W|tw2YSha%uDVIiEq|3)-T4Jc}9K%$*-&{%opq@{?N~O5eK}dNjv0U
zo;wpK#4iY+JWO6-{8<-i5Bu5Av;MK~V!y`oPVz0!l__UG$i9r{iHtLLVZZhd8pJN-
z@jJ-A5C8Dq5c^c-0qX|)LG~-$cd%b$9Pk(Y$4`t4^~r<GXAnP<M~N@un|(es`zjFs
z(T<KQ@yY!gagQIc3!L)gIsaYqKMUpX1MgY!_gX;m2*|$OLgD~KzYE#-E7C5=c!1=w
ztVq5~i^RPPnO`9L9FX}>e^|HSGhd-W;trZTK)xbQ*r#Z`5Qp5)unvO69rZz;zY)LW
zOI?4-XT%}<Q0|+sn|^cOfIRym;+l00f3Oclj(ku5X$R!~jQu6+I>>zwe&T+L`zY?4
zkYgQT9fHq3hWj=4Dd^!jGWw8*#=ks|q#d63qDSxNkXO5*xqrh>kaJV^SIF}|5A8Bu
z<R{M0S?^hw83*zx>nica{t`a>ZS;VQFL{&v$2!iwiTAoWSK~c2_Up9EeHZhK^@n(Z
zPy7)l)W<KZ&%`79SkCXsx6BvTeUSY&^VdTBhds<U=BtI+3lg7-^h1&P0J451Me>G)
zR)3X`|L_C%pX5#UQS1kJZc9ATKK?@wNPLhNnD?xY^p|}A<4T;6w}=C1)dNlc$VZG1
z^HRqF`*l8HKjRNge6aqrpTJ(qsfS;f&meN-Z}`-Mrr-3Ba`toh5&I}-T(OUG^x_BP
z$miU@Gk?&}y3M);&3Z^Z_HFFrShuK$Ke!K~UFHGffgFBPntsqfkocfI@*4NY*vI>;
zte1=f{=@IwcM~7vH}3b)Pkgf<fM%Sihke8)cF`W^G0Z>CMOdF1H}V|gL)<f;u#0iW
zulNNTzv55&i9c8;v77r%#+5vP9>#%oLB^Nok;rpjh#Yx<^#Xm!lV9;4_v4h~PiXQ0
zeq$b?2mOpY{iK|BC?`L`XT4-yMj!nJS@%Kq5g>e!bsNObApNwEbr<A(hWKPXXFtfg
z%D#a(WE_}>>OcIzK8O7n?<eayk)EgGSK<<T@EiL&#)JJY@j`v#mGMTNeGW+cvL2Af
z*=NyD#t~#3m|v`i!~^lhehEJzPoAV*@+IR<+>tkt=Y9^o(A?icqYq@C#{C-ha(_hL
z06CZE+@JhTJ=*7foczN&0n#3M9V8Eej1x$EApHP&P6XoDq=+6D;#Ui?55#XS<oO(k
zpLo7QTod=?6XF&meu!J*8AJ~VpZg5-GjGtte4-xrPqYW0`OUd8^MgFYeG~h2>M?(4
zPm%RRk@|}0cOiBuGT#;PvkR$5o&mLAE*ihMkaZYj9MYot5t{GWqmT8%Lj8TPr@aTf
z2imn`KCyo%Kl0p}b0CoUOnyd=a}b_SbKlK-k>m@WNAvyy@6EHHBQLSfV1L8!UfF+=
z?|2T#xeWVF>|!5HUSpjgZ_{7)_1tIBPxf=H`>aRoJCP?}vtMOD%JVqR$=Elt|I~Gj
z`wi}EK<pupbN@~`cCik#FJqs{zKMMw`#1JC+)p9L?@YPhrvLO8zbVol^4QC`;&=KF
z;s?eRWc)$ew~%%%BtBFh^6V?vhaiVNApODL$P;HeF7(er?n|;F<EO|znEQ3!6JXpa
zC!Xm)^AA507xYWp;l2)kQO<ss{WkM}_K6qt(?0QueLTP8{+0VK&Mk;{+F^c>cbFgC
zAF!XrUhaz&vCBg4dy*p0h43r$k$nsCgI{=#!M>Zk#`!z-IJaa!%X<#wCB{t=zgUPJ
z77|zF3GSaj;(+mBd_m5OK;pxN%vbG?(%jctsN=0PajHoFEVSaKT72#+@h|Tm5KpY<
z-2Za!MEoGnx=B1TE;^pvKk|M8=XX3`XS`T%nWxNuQ1ct@fvle({Q$`eAbAI*zx+;{
z{Gmv^(=OwVKgk=+ALb=A?@5w}Sl3twn4i3t#C<922Yl8?T}K%w%2{Wq$9)g>5ckXr
z{Kve5Pk)hTJcw`n!a7B~;&0*rKl2_3@kKrLH}OTDq&?1eIJX07k2s?~d5H63;*NbR
zH2W{kXQ@wrXrFT-wHF#W{D^+ew>eiQ4_e4~+m`Bg(#aEd>5%)&($D{4+Y#+^tZTf-
zLfkT6I=($9x8tAY&++~P?_YyUe|6N^y*6Dihknp{4@2*Hn9sZq!}<hrPQ-f_F64aA
zg^YtD@t}x3S&{dYEadrdTC~pHlutfXWIhn*<T3KEg~;=r5d6(|dO1&J{OC7v!Tcc3
znQydD9*56-<M}G{pZP$0<QMW5@kQQ1KkL_38|5Q%$O%IZ8Mu1Qx;4FO+MhpUk0CW3
z>i-?sd0>};-D}zn9x|{;J-tRv`<nK<?7!-HLk8Y>;kMIOZr|?8i{9RS;kTX-7hU_#
z+owJCL}1|!&kq0X=c7IeExcyGOFuYf-ASQ^>)$<Y!$qf#37QMP-*2^E+szyiC?;+1
z;ngl0bJ8JqhKi34f9;7MKmK0;YWdzjzH#jl$4v-a=(5A|yNi!r#xX1Y>i(sM<pzEm
zTDW(Infvdx_Vmz&#K$tTANa?UcibF+H#{};>;-yG2|!oBDA)1KN<Y`_wtIHz|859&
zxogL9zh3jR5Onp6a_ny|bd4YOy3iU&_`No{XyV~BzX-u)_WFJA(YHPpBt>f+bbcu9
z>L2A=&qXu;&Kq%1ufDs#6<Ub?tl00+Z|0w|{LH{Y>}V;}e!Q~Z=yUhIWOiU7^U6Z<
ziG|j<E1&Ul%@eB~_?DgUk+Tpv3zLqMrC<H8G=8v9{id{KH+)NPYv)`01dFeJO-ozz
zRrzf2UHU8auS>tHoOP_ZQ2nCx_X7@mV5tK?3N5VEE?3^L;skr}pKE?=JDNACPh4of
zn1>+kyAXK`Y1cxvQ)%s&(wb+K)_$g?ExVKQHQpFcknvEYT+xab_}c%hw3QcLU;IzM
zS!Tl5frZ-NthA1EQaWpYt$OrxSg(a%tm*VsaMy(E26a04I}?$A@uKg4`22vMf<6Az
zt>>&iwabBfEqvg2BZhV;(C`&q@nwyV#dq1Qan@RBYh0?uxAK;YpERyreoPwI*0?HP
z?bdRoR~+!ze4pGgF$B}*N0rz1mDcu^e&W2lw)^8^Uj!Dq;&}HRR_nLr?>pu|jT5!U
z+7DZN9WS-N(($fz-fMrMubA-l?N@KGNUepe9~QC?(D8;=zbdVMsuu0?H|r|-+cj^=
z+ls^!NFLWXb&X%ra+jTH%MbqW(#@Y+dV!p)p4vzJfI6;9Bd5qXDPL*Tr?lFq^f&kJ
z)BCk)Q_9W_=(p}Kt$B`KtK8yKPf^=d+R_J~_ysi&(mxlfKb6+;P}<55?0Z@Y$sZki
zju`pM*muIDXxZ=LlSec!STy;;!gY7K@TQH9{v@za<40-DJ1&|$U2W9(QCj0iX~r=t
z{&CVvlg7>WerVyG<M;10{j))#g>OEz+%wDm-*Z7$%xbsRBR?t<ci#`Vuy)q{U)G&d
z#N%zRE<59y4IdAh3%6MKu~q+Y#GJ5Nc+eWl)ULP0t6^50(0$8xE3Gs>u#miIp~hiz
z(Trc#d1v`Y<HSX?f2$OEe$-r;)K6AFTzT9oPugx$xu*`8|MOn!jR+N8<3zdUQSybZ
z11&|T^_wLRU+<$_^je4B-LKu#KNOyOTH}Cz3kz|Zf8NdfY%N%@FX3})%Wi9av{k<L
z(;8pqiG_O3rnD;#DYx`naha6Q{CA=HV{Xx!r<8WBtFHXzs&B3D@Lln4jU)A4sQZ8G
z{zvckbib~C(0*(DsGpTq{VrPb4)g1(!GHhz^1Y`Q&bzsv;T!~n&-+E2pZ;j=!)yE;
zzy}B1_|m9-_G_QBkaG&qlC$`<@4{<d>$TM0_kV9b$Kd{&@r2g#hOcPZr+lpkO}pBj
z(ynnOUs*^yit0a&UudnDmPSunRQsUSze?jj<Q0iiP}c{gRlkeYap-sa*US8F^n5vx
zxF-%Q)cLOY1KP@O%GY+ZA6gEr{)2YKrRKMj_dfZh`;VOzuC><AA1>XqV{Of(exI$j
zeD~TOhjiFt$TdsW?l5-B&3AmVWbO7ReDmR~kv(dc``c-K58HOB+M9;2HuZx07A;`!
zo0os>;F(L;?i${`%O;!lsNG@jZtFhzb@xI&%6C|J(aV?W(6x5wKks(dy7w$m3%|?u
zFHKo_!cw(oym{N9M-A^%d+)fR3%xpL>DpBuUi+KAYjmsay~gLk&f~k)?)u4y&2O(+
zx}f)WiyZXSDOWF5+x5GrcRc;mu7&#85B}!1H?IEMQ;QYO;oshD?V}#OanahLJ74x<
z=i@rn-n?S>ySI9&W9@C@cHQZ}3(Z%1#TVCJe&h-rYOfu5;wLL!*`ao$AO8DH?-SeC
z-g@5r$ISk)UG29Q_WR+BTfYxq-|&pli!3%PJpP-Wn~eT?WMQ979tVjF3$67Id5t?=
zm#p>HwV%;))ti>q_)^-X&vlNf<>VDm_hb0qg_i$Zd{=*}T~0hG>NrAc+$gQ%1Fd<g
zrD?TOY27y|oz(9dFV&*)V{;?pY@zl?;|AJgC(l1zezo!#eA=^6?M+L&`k`?Iz0-$N
z58ZY8S7EiWrSgaFZ<NOWiZ1)`KhO6pBt9&(#s&XZiaP(HwZ9haifb!gUH+pSd$OW6
zzu{+H-(2-5ryropUskzm9kJ@cch#rdil-sJn{s%~j`P<lvhQ;t=b;u_?|CHUGv6$9
z#g8@5lJXgE3mFFsiANV&^<8{d|EpDQ^&h^=PFJ}VpBA6Ew$N&i-|JY&@2y<Oymz71
zPNjV7`6s_WvXJp|q17)do?QOYaw}h0`r)e{rLBI$M~@42zhLDHmDBq%Er+(wC9HLo
z=dV`0SIW2EOHjV+cXE`|Pw=N3_vrh<nA-yjUHS3FKmRXS{lf{Ng;u=6XZ$V1ZVM+Y
zyYI7$d@wb%(0cC!K6Y5B@o3oz-_m3GCoSLF7h3UIZNA1`OVLT^k?TDjD__7Lci0k-
zY&>aJ;6mgrblIWhs>gb7WzPKH&j0QEPX(ZBep+#<ac9MC*7XQ|+K$rd2aC4i%#w%i
zikqbK!`io0$|n!GP}kF>w5$IuRql#YSGnaMm;EliuKLVNU0+=FwA{+u@U8tcd@CQo
z*L6y1%YSM4?fS2D|I*W6EIdD`H2#y;x9rsXbm=9X_grxMv7rk!4=JtVt+dK3?aC{R
zn}y`Br0DV^`_!bU^G4$-Ep6@JlJc$m58u_V&o&RfUA_O8K~mK7L6?3VC+q@Q4>P0j
z-52tc%RbFVO1tdRc(CZ#?(ex~pIzGJEads`n1z48TEEr43D(|!(q)exF}v`4F6e6x
z8vn)G2h9vNU-0|qrpy>0bhzNE$N&80jUoKUA3S=$3vZuN;P*T5*3Z|SF)>7Lt4^m3
zetY`61zP#^7e0P;A@X2adF)!f|GPU~@Zj)}`slNedgNc^s1L$Viqr#XM-l&mlw%M5
z(RPs|Ke2CBxpw*YVaxdW^#(nb9QBX`u@}@l?V?@t$|~1&Non$eg*rd0O|$Q?ka)1r
zn*Z>%ze@l5H)F1wz2(ngwb44if`88I6OKA^k559-HU8Bux6TJ$eD;ABy4F$horTO#
zMaDBL{^_Y*MlV<UP3S`1Un%X{Z&>}Wbo`ReJ1Y;s?{fRvEADp8)F3HZ<CK-3v>)ns
zrLBCSe9jk}3ti)?aa?WM<xeZGC*|vTwbE7|fM2P7R^Es2^1o}oYP?jNc8#~oe%Cm;
z=CR)YRXQG=<GK2!?WU!(=J}SYhaW6-^|z($wAyKI`z-xE*4*>{%dVUiT4>GJr0b1q
zK3V$VYg}9BJ@766!bi?R*5$OQ^H}YM*7Ja@^m0SSoObr9pNAHbk1bStmA3o?-`ZEf
z*K=si<5~{w+V37TeZ_0m+~%uL5&yPUOghf4^(ATln0JhqqMk1@KP=R|s_~+<-q%4}
zc}V$Ie&M|$%TH>D(wBbl<o{0TJu0wJ{bJFoS7~eB!Pohqw3UC@$F($8Yrk&A7ksPV
z&CREMjdK@GzOxX$7P|5R@nWLmJCw|$o&WxqWfu>=cfKF&%D<Fbd5C#lDO&pBTjRny
z;zE9Jrsw(4)3-lqtqxDjDSQtIKK1w=6EvuD`ujA>;iHH1Z;;=M>F=rFBS(GsDi5vY
z&`FWsmrXhGAGcg{>CXYbR{`li{Zgdg$YDQ-9+y4*-VHmbfAKdzzkA4r-v+b;svQ=M
zJobUeQ%-vz{sq;3^jOIKJLS3#ur6Bft;2W4`?23Xy~(Kir<XmSZP)FDg@#RiyzIOG
zw3`&I{6v4ybKgFDy>jo(9~b@(0(|htPD_lsZP@rixz<O{wH_e{y5wocg^xV6Xuk^%
z8dLW74$#kdX2s<$f9kPo!fA!S6M$Yt`UT=w=$1m(6-C#2t@|a5w)E(JR%tz-RNB%D
z-!-3H<?O#L<o?z|mwujGDC&KYMYE1tXxR<lwZ2>Zf^V(g@Y%OmXxVM?X;*R3f<vF5
zeEZ0<znh}@jCJ6_z7N0rpD8l~Q2k)NCro|^S2*PRTTc4b>;NQ>Tl3bMAMmaHC49bf
zhkp7;zSaDqw3QcC&Z1rU5PLGCu@AQFPRn=M#eTp-^^4{!7tJ_>R(|0An7{wQ_*?Tr
z`N$_l%a1w_m_H!#Kz;fJTKBhVmn8?Ec0lZcW<D@J@Y5oCEp)|?mFKK__zC}k^v8v~
zZ;|x(qNaWMd1tk!_T%SvCyv@?mBl}D=3VM1w$t%G6#9~)^}9EhKXhI}PrJSGcNMPs
zmOlr5xKZa{?>MRKK8O1g&=tST4-1Je3pc;{m4k-u`CSNVK2rO+PX%?J=y>Qnr(Ym>
z1Vpcev}gH+xCC{4;Df3MJCdT-*Z!*C*%yJj?@$^!Q2V3fqxRr03$=dIej}%P$zPzx
zgR4FERiLiJe3u7A4pjY&vxVv}wO4847F7GxA386!J+({Am8M+_@dL;>fI6>Kzxq$(
z*F|gp)lO|s<&{?dsC_P4=abUpb5Qdie9-DY-!EEj#pTa<aEsaDwGW+s(evjoP|J6Q
zcDw22>wfOrz3^S2K6};vwC^EZ3g0K9obMcgHBax;^PQW!7QSc1_lKwt&3BIA7wR2<
z>whlaruf|=^kNs^NunO*)Q5k`$QRDJaQ7~S?=juH#Ier|{9y6g9iN(S*0GB%Q48IE
zf!|zy={cPX-&MM?L$ATTZtq<9j?v#Q`1fnWJ1$iCz7fB3S3bXs=RAV@V}8HOI1x{r
z*YQ2N{+o^JcI$Qv<~Z-<ypg<tyq;Hb{s*cY@*wk{IL9958%Uf|4l*Cnhke*hf1v3H
z?Wtb)isWnZ0!aJx(?aZ1q(1HGe1*nN{6YUIhmRb2m;A}PYRy0D(=QPF7$<1@0S&?j
z;ZvV};!pAw{YZ-RPm%s7MeMN98c+DdxrNAq_{Bo}lN1>z^f9kN_B|GAK37`zOG*=0
zS&?<YLf1H2_2A<N3$6J8A3YW#XCd>?LXBUgUGqW5D=n?<K@%^o@z#FvTohElpdVDf
z>ba!0ukspq+Md#?U;TyMpvvPHP~%GFRlmxsecB)Ge_DC<m+DjdXrJ>lko|*&>?<t9
zUs=&=7e4zOMefTL*STVgo8Q=^T`kCXgX$NZhpd+%>lmo(3GoiHeuFxXblfx#(k`g>
zDXnpnHcnk}tNMu}5IaHmp!yp=sOMg+x2$`tgP^Pb+8>Qe?GNMTLi$lwU(V%qe05&B
zXsxfb+DTl3_!Y!&pxRHIf?AFqP|G#GwSQ{A*4O^1KAj&bPuzi8t~7F>mTUX?1yuR8
z_G$iL+(6o~P}`#)pxTFCkbOHyJ&^jW4<3_wK-Wi}BfC(a+gD1Hr!3U`n3lHm!guwT
zJYu2iIiap^N?Y$KCgp4Un$PGbm@GQlj?`U;FF$yTuzT)5Kc73`*)zHokn@+8LM#7S
zeD)0<oBlfzF8`4KL0xY={l>LN95<ot{kTf~!up#Obsy~Mp=U49bBgo5!j|%r?t88E
z4!+7Ot@~!Bb>CWT+7;jA3q`flMeBNwe=JmarPYrXopjz=a;|<^_264}Dc{wORc`qY
zzEuxCa#_(8kFNOAaV?|sf49!kgP$#=pP;p4v-+XqWYJ}I%ipu3eG4uBTm4qPWj}nK
zztDQ$3T^GjlyCVRzSUp&*r!N2X!*zDyT$|iEL8udrCsZh_P5%!)vu)OSoPtj^@qx5
z^$&6DLhN**r6;MMtoXC|=p~P+UtIRO{DwZOU+|Gr#6AnHe!*wHT1dYwL{D>}WtWSe
z)E>)k@LhUbdCOYAHrxC8z2EM&MD1Y<e0=e{J1<!~{+`dT*!tyewYR;t+`yCfU$(aA
z!2NFNbk1_MZ+^V)7h^wOrhxa~-ec&h6PK<%{<kM}+u+I`wG$t|?~Ffh?Z*4}?fk#L
ze%8JApC@+=UOuyX?Fru=ebU7@bgP~8r#r{LzEii_5j(8#(BHrAT85M__OGXZJ-B!G
z+LcD^bXe`s?zPYt^!)DD7iV>=ReWHv1@|BQU6<OwEgW{e?u@$reYe^4J0I>`yW)F$
z_IvY%PPMz=IO*1t)>*K2gMq<8Yb-op?Ru;4cKnG~{uFM%{BL(1-Sx}x(uMxG>hgPk
z65jKA$HUg`Iy(I0#AlCrcene(euIZT_u$qq1;1VA{V5B)Jt}zVug~u`px?OQ-ZQs2
zYpbpQ8}y!Y`k-|-doLI~VEsKmKH`G{B1bv%3R;o+)N8-p#}}@)%J_ml$`w`a|Ay`U
z?q+{~r%(@lAniWB{o$)#ckg4t&bM4K;>qyY5bSf_a<6~zzrP3c8>F7LkKT2*y!Fja
zn@kH<-09PucfDtJ0bdxp$<dRBe^L0}VgIRHZ~X2{p9OuMS!4X@6Fv>@zu>JC+I=vi
zaQ}vW`~|{Sq})QyUrO^H3&{Qkbmddd16{~-O+|eUmXszx(k|zf@WG_#yvSRq&(ok?
z`3k!%wAxcX`4QB0&>AP6e^{t`pz*7s)nELfXvw?y+OE?2T--%l>xSBE(Jp(fde*qZ
z*XJin>wHyO=M%Kc-sURTae}t?4=y{H2gE%H--XC4TJ|L6t3IWZ>d7jn?J4c@3+1l%
zt$72V@wL#Ef6?Q@sbgL{xKqbx0_F|qvY&oh=xR5s{d~WPd}^VNo6=T2_?i!tcFix9
zSK3vdaj?+xJNCO!&r_6EdzH5Ga#Fs_K3Du%^8&t0KXRKd+T*+tw@nK{{OTGX*L%0t
zz5%}aGcApsSuyGS#81}!pNntBuku~~c9pBXYSZ|c{w9rcm9zY=a!S*_3$>m_Gp?Xz
zm-1chTjd>Roc_cDGslN69P$2=?_4+S#XxbF1<tweSJOuZS#iIcAANT8tuq1(pS$Gg
z)7q^!F0k;YMIN~Fu&*8oEOhlt%Pl=t|17(b@~wVXi$8JuO)vQA!}~)E=X^D0_uD?Z
zJV=ZGeeJ@trmi|IaG}*Md{_O4pIP*yK{F-<7HT{yZRMM^{mm->;m)1^d;h36!b<U#
zPKRH0{av#I3#W9S^6Nz=e-kM7nYGP;aXZfnEc|R_$JO>a@YBFT#>s^%fBF3Ze_m~B
zsOakVup3sIf8tuRLkq2OSH8|;r8n5<#R2;+_j#ze^-m{v`1rt2LKj-|nYh?xrT$}2
z>iJ_}A##da{bP%sM;|^paG}<-=x6#IdgN9!-U-1CkLh{h@&|noxUk<HEA_hP$)5uY
zt^VscShQuYi*L1S@hyE{{`0nra%X=ZSm??_R(tR@?<nn>Kd$ki+?uBcoYZdkQZwcR
z)kZ7dz<<5_7vZdR+UNec;_EA%dB>zs(aOInr?k#rr8UoFr8Pe>Z(OMMDs7!lSDSCy
z>1wCTsqgLh@{DIgMQi^9-xbG{Yy4ZZl`mZP6?$%<wCg_A^}W)x<?Pp!qMmyy?Rvj@
zZkN06<I|QWeQ&$#z9+1?+F?J2Y0>)rjm3ZXq(|1+_m%dw7P`*KT<`bm{hQLR_~-k6
z7Fy3g;9K8~ZY_LQ|L1o3c^x)eaO;b{395}(zIW)3>o5F$(A-EIG#9$^lm7lf>D>KK
zp7!uR7R*_={euHWom#ti&cYMk$bGry{2g;yk>C5LMSjoN+_>AvoBZSaZ5GX0$nOBF
zjhwS7a(>!YG3z|g{_^+U_PgY;j(yJlK79Pqi<UqBz)uT*r;+c~u|MNGywD(W{M|qP
z4k0xA8|72)%RlsAZ<~W(4%#}p@<?kdx9)e6@|jmj*J0wth1PyHsa=+w=53{|{Ygvl
zn=3vTCq>tL8LhRv<KXS5Z8-J&u(^@%GqhE-{=U?_!msweTIYw0PI~{`R^(j$b=^l=
z<t^R+E55k)@*`L1Q235rrL>;cShV%ra&Ggjd19@T@U!MUt3A(ud-sLkdOoz~NwxWD
z-v?27>+f)R{+8`lT4{XH>iw|nzRxc5!PL-ozGs!Y_UXD_G}r!B^>RO{$oz4wH&*_J
zPyYL*Amf<!oO)i7XWSKA>b|d~+hM$uVx{vy@3UQWbKlpY-+HetD}Uck3--P0*@bh}
zLf8D#cuh+mWc*#%YVpa(Sy7)eRhxFjVWs7j&Ks+Kwf$h}OM0%CR!-j&hjzUW?Rvg}
z9t-)-xb?ju*K-i79(;b6XQ9<U-FI6wzXP<8@8``ew4S5ubBDAv{;d>U&zG#YflvQ<
zj&<-`)B1e2&SE+6pc8L6Y4Hx7a(oA#_NbQ>U3uJ<Pp$b3-*q4EI<IimclFbXZ}_hI
zj7PPR_FY)%b1K*PQqKO&${X;ld;?$e1^PkPe01FhVuyt;e`2SFuKPJxyy|_T(ysow
z+IPh(e$I;4J^;SuU)pocSIUvM(B)@Wxuq9Ac39|IpRM?TZ~3WOd{-Q~;-u1YD<5JP
z<Ac9J%ir)_cIkXl8a;ChEx*8bjX&|@Le^JB{0^q|1NA`uel+c9enJj(m8(7I1+`ph
z##{MHV-I{6jhyNyo~?01uOj+f$h^?{@QEAjKn}h58+rKjgS<^W<U#oOAA}F0N0D}D
z*M;b7Zlr&eqV|XWTF5wAXvGnHtDeQD9;oA^v@5@&$3oZs&2_)v8V~ii(n;e+`>nL4
z$K@Z@qqN#<(N_FecEPvwCFNV=S}nfSFZeG1y2@SeBU<rk`A6p!`IhI=yqBckBjUb{
zaiM?Bh0Ghi>!PUV$ojhpmB((@jk$&Ri~COt>A!{a$3m+fe8yYR(g)v)tE7Bayz9KQ
z;#=cbX~q}3LHrNG2k{SxTvD{`z#i?7tH0O<qDK*Xt@SAB__*Y-15|mnOKF!MRiDzd
z3#vZ#r?#)O)_29HOJ4OWt@X8B>9q11e@d%<Eyu5*whtd<{|vh1iBnL=H|_ocJ3!T^
zwA!cT7Oj4RR(bq@JuQvezohh_34PvPsl&{Wc=+bteR{t(ZA!p>`NM0@zW>r)zX~`f
z;&~D0pF9`goRWPD`Hp#oe*A}h_)%$<r=Eq>w-7s$qQ;}r#Ic2HCw86l(hGN;b<N~p
zv(1*<^7N-Z4cLd_zqR+DblIav)P3h+ztQLJdr1)~S0wMTPy6(yhZgzU(H{i-4g^|}
z@xdN#PigErcFxVmfA&rJ-^HLG7OLIQ=Z!e1SKr;=3W)<n^jZ5Qi%<VtXpNt1{%SlZ
z&A4PmSDX@07P{&)zbs@wL0+@`oR*JYb)1#%*mK0lSH`{*axSO+RC$ZG##iTOR+|27
z)#;SMZ%=<WBo6SKg_d3L(PN?d!=lLxF0|s*HGi0&F0{r6zTVeCTmFO3Iz&DNv1ibN
zL!X~~`^bPe*Kvk#p~^$^_sA`D%|G?G=38jm1NHB6@%{zKIt`-7LiAdw{UC26uQYr`
z?XS`*Uv1hI$7#!nTNg4f2Yt9v=U?wQDX@@yt4Ka@A@f1=vCF<n{p`|D|FMhtYN5tk
zR$Aw?#+!?_;(@%Pi2p(L3-bhjfMs^(|4xjKw`C7})*%;ad|5R9QU5ES_@RH;gB+;+
zQ5roK;;*Erez5dfd>y|+zvS=uFy0`3un>PJTKQS~p>`>){=v_xM`?{ir7b^ceCWKW
zHm!DBwEEXYTm6Htezj;F4`|Cy_*T3qpYa6IpB3p}TD1BPAAK&r&+T&TO^Q5USY*)R
z!}q!Tf8pUTpWEr19^ZsopZ64XUS}OAt!L$J<r61ak@&Pw?NXY&Xraq4>;4D6rC0fC
zSG8#!e`xFkdCq3})8!BJx{&os(N&Lf=C6g!1FKwrf0OpT|Fn4lyA|=b3+Eiaf2Zl6
z4GOD`u6`327HWT>$uEkWzba}TQkrwz=0deoX|*>kZTSzr&I6@g`xLd;qIrI<&sUXp
zwWoPpY5HZMwx_g}f0S?Ock$I;r8(cRQ0-M3KU?U^qb~bh<!X=8R==vv=f2uwya%lH
zsipavFUaE-y4F9vKU3P3$6fP7<5_9lUn%XHKQ8}kf0VA&Kj^dkUn$>ZpQ{}EnhRa~
z-?Zy*T6x#{=hEl$U!~)x^<DN<yIkj+((J!2wAMxVuKBO!x}Vqbw6y9|+La%y@7bxJ
zEZUWy@RNlaA4<FYq3!GbNNIh~QEC02heccWBQAS&U#PV9$D*|!wD#Ackq33%NlUx(
zkd|9{Gwr?(|G4~4xixNS`8v<4MeFzP7@w@)J)qx7k$6cv4q4@CPciL!ZT-&Q%3I2}
z&MT~OO&TwjU#$37e$wx`TkT}!Cyf^?KQb@1KhUoH;%c9HVWDf?v;4w(;A-DxzqQ`M
z$3GSh`Q4PmYj&K!)<T^xEluk>t+dPk`n*eNSAAE0aMgE>uN7C7@-4gc98GDfUQ#~u
z4&?n)&~=~d>Nk2V)bUkX<40-NeTvHu);xpn>Yu9|Jr=s^>v(6S)vwS_|LnI~ukB`z
zD7@e0nqOM((r@KUmtQP@!l!>0y7as90PR?)aY}oZJbc}MD9t*N6<zCr+U=sPc|EuI
zu5qIr`xqCHaj=m37P|ap=}pSF+D*!Lwc|Qx!A}+@wZoEg*=OYo_>8NC$Yn+Lhm13R
zaM?|{WmmQM*kvJhSXeDPEIpQ8N%@xEYVobOgzt)*O6{N>3$1a3Z`FhEvY&Xi(CVM{
zJzn^(b;l~V?5LEl@n)TO!ne*5D9?&KRnvVP<7J_1UrxC`7fVX(dvQu9)nm!&xu4Rm
zeWT?U-G5rNrQedb#^2&6{m!=4UQ)i*kEDFH|Ei7h5jo_9A%_fHy{1>qdNu9OAF{`g
znhy2<4(vRz%fRk6?FJ7S*rT3aqo#dL`(5^5^}HbiZ@h5ZX)CuUo;`XGcn^3Fcn^3F
zcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fv<VOF{?JKl-1O+U
zaBg9naFRFDd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Z
zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Z
zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Z
zd%%0Zd%%0Zd%%0Zd%)p=P1+9`_{f|wp~t5EF!Q0>v8QeLQJ5BQdE}3;4`2J^uu>ee
z!mOb$^%)m>^nR$11IM)X#ohzn1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP
z1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP
z1KtDP1KtDP1KtDP1KtDP1KtDP1KtCz#sl}A@Z-YMk9jVr7H)TT?$OS>z8oaQi`F@3
zg#-F;5PI}}XwEp8c;LNnJlJV`=&@z}&}tH$H^h737wCat{8`)Ye!=+AV|D%TK<9s-
z`|i&V2W=Jqb=K@Z4qo_{psnMMKMvgC+2dabJ$gUX$H6Z!#rrOM4|orF4|orF4|orF
z4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|os!LOihV
zAC4L^;pla9mExjL{(16A|J@>&7KctcX}&HOZ;{K2bB>-m<j#A3otxXZ;1P@OaPoS;
z$$`I+@bTSjV;-1r&FC?U>@Xv6;cExI{pax5FN3tWOqWN``R4tpK}%s9bH6v&d%%0Z
zd!W^O;Dz~K{9^F1rE*|vHF$}Bcipngm1Dzd<9@?V@3r?{qeG9Sez<i`pU3~!{k_nm
z_k)jvd1oH9*8FbSwP5hvO?M8P7PJ%&KJ}uL@7cY5ZeB;r-=6P1;630y;630y;63om
z@jy%6=lya#>U-)v;630y;630y;630y;630y;630y;630y;630y;630y;630y;630y
z;630y;62b9J>b8aUES|ydw*8PpRIAOc$>Tjya&7oya&7oya&7oya&7oya&7oya&7o
zya&7oehD79YNLEFhnz6vkb$e$tXH#sP5bkQ>@lRKL;b%4I}hwKuzO9r!9xc2sHfMc
zX<yTRm;F~gZ^*zKFWh$8%I&M>qa9cI&%t+pFf&Ms^S^xZxsUBUCCG}?KUw^;m;N#*
zs1()S-of!V?)vZ-frU71hxR9I`tvt0gsla!%VVPlPTQ*H<=wBJ8G_y~={%tGs{hf0
zwteZbnPF?eAJ;kQtR2SAp97h%>3Zp<o7)5HJaXsW6PI}}Y%Qqydd%=n7fif%T<F5%
zx{TX!<eeXctqon{=UQL?aNM_Zei-&m05XqUc*E@d{@Ho<L4ga&BkFgT-PJ1B_1#5R
zt6i^W-t@pXD|Z=k_x|q%9_OwfT1#|Ybmt|@b-BM?4jenG_k)+8@nWF()>q^A-SD0N
z1&V*Z_4FO@{^w6YrN}<0T1bB^RD0LmvG<XWb^0!7E#oq;*x_2QYLChdd+DihL2E<e
zpZ!N`%a_<^A^x<==au-HSNd_$c4urk=Zs0AqTYAt{{5Noqdj+>JuS3wUhz-0`l0*%
z2VNR8{INGj1TJ*tv9#se4SFs7<4->ain_i+r`50W{mvM3#hcwa<P>jyeZ99o+w{i(
z)N-Y{A7cH4Cf_KMKhvV?{!PoRJO^Lz-z-|~S32oBK)Y#C+iNbmTKlY))}whKZCt6m
z6$kLOy|naAW0pL&_SOY+a|^L|zQ=BQ^u?9i<+9@3vY&R(n>cBUwZHG2%Zjf2?kd0Q
z_`UwwV}(w+ta$R&7oM2E?-IGB$hhtGQqR?{S!`T*`ha04{h|NF@UOET`S!>c#)g|;
zedcn5Hhn*wZ?7vy51UowQx0DI&%6GT`|a0(BKr#Tf)9@Cc-bQ(r-mSQWW^W8AG`Zs
zXU&((il^T3j}AX{`d^S0M_<=*m41&coXd(&bp2xV@3#LbNQ;)g(yj+8ultv@w65n#
zX{$e3`P*J_&jG(Va*-U!KF7kW>x}LblzwxOO<q3gq^`NN$i7E$(k46Y^1z0R<t)_m
zh&z72-$H*n{o6n>X`VZ@=Vg~%_I#IITGV}o_5-@_Ww-og{p)9hqc8u@=uso5g>T(3
ze3{9M&j=S;fA`6sUN<e=e8lW8It-W=9(nALA9}1kt#JRc<VHu{xb@<b3p9D^n89z|
z_d@RpA>|K!x8b1=k9{+=FzdLx>S^5Wc;fHJFR|yS5VW4BXuY)bU#?o@+f5&OICP<o
zkJ4$|t(H7-(Nf4htWspXvyganq2^;~jhD3cPV0BvDYGsb5rS6zYK^0wBeXPa^$))K
z(;7F6j~_wRr?l#Et*hT1^3P=+T5EE+(|fBQdRy=D;rJ_FxpK`NCWkZ6d3Mhg7M~Qp
ze&73Dw|#7Cc;D<_UC`~!Vjk4^x^#mBw!M1fl<>~CcDgvY^Sc0~-oHQm<ekF?%q-}8
z=fPQTT(Q=qu-i(%yM28Bso{RNez^Wek4`G&<8@{&yY^8#ObwIb);9-h{9MyLcj}MN
zZoAWnvBAKvhu(bN`qP39cY1%J-H*OM*yrJ&*4m=~gz(8@H+Z$jzdj7_d-UN4F5cta
zu+u*GT>0%G<HC<_n{nxRvnGV(^>wy;{;s#5ey*@z{rZF_x-L0#Lb%#PmtA|souk7`
zMqK`MhY24Ro>y<svG2#r^dBFtIp3D6y)tW3xZt*b{`hx0ObNGtX6AiguQfIN^YQl%
zA9&rQaPPg=A2#fm(c!we*E-&M)-B-=eNOGU&zV!hg+J}K-@le?mkajo(&yPRlLGLG
ziT}6bLOV<fU-;*+4UZr9b_hM`%sYp^e%;i<x^#Hw7f(5Nvya1ehduPp{MU^SU+Zwz
zmBZG5KHT=5Mep8o<fKA9>eC<g*YtaZt^T&@`eR=Whi-As=<{!VJ)F96_gAMpKdSJ&
z{fwo%4Lj-LcMI{s_%6QvB}1=$bV3N;(SE&`|L?IW;SKw~F@EB;F9dh|_P$yBJU2Yp
z@XEsuUh{=9g>`f4pU%JWnAy|ACx6(vXUAP%3;*}|UDrN$$wvhwzF&U!+r@6|KRMj(
zfl0^Cm^HZ&j|0zs`0Eo!PA-h!J@fB$_Hk=X3{O1x(kDC3o)XSJ<=>kgT>Dn|W!FzH
zKYje@LVrG8;+>J}9Q$E7rFQ+EBM*NuT;*4PUFMwL;|j?8N<B6^ebjyH-V=Vd*?Q-0
zw&}?5ktg0dYV30(!e^&!-eKQI$A!b+oP6beYtIOO-um~eeLr$~NSuR=%Y@wjb$kA;
zQH40XYpEmt^5(naLgMnqZLZ#I!95-io>}#i4Ub-Bf!w>ldhe9m22Tk(Ucc6Z2VDD1
zu<>c@J+kv@9|sTLyVCCOJ^y`xzFl`;>eIeIjSYJ*bN9Fpwz<E6<Uz&}zmiArOV2Lb
z9XfKYDdECre?MT)ZZCxE?e<Ead#7F*-2A{<ThBP=gTi_2{`>sy;BUu`4Ds7@KOZ)_
zZ}?Ino{7gRdJp}2yIDm)s$a>AiVrS%MCU73e>_xFIXw@ACJxxYk+=4^>A90G-C<gI
z{@yoj+Gni^g}7nAM?DL*eTzPN`yJMP>eP3F^Ox&&@Jm082%o>_seNbdFf&~8yKi54
zc-%W7`Hu5#^6v7h4mfbbkrNC1&#|Xp_VH<J&n(b8{^!-<L(eS675R{SLOF4WzSX<$
z-*ffrCKU1y<JqJ0ns=Z3*z}NcXyzyP%P+jU%tAjbHnEV8w|e^RuWx^BM)>M(3;*rj
z7d{GS9eCz}{U4hap0e9f&)&Pk^fDx_*=Jv~@;)6dx^8lz-oHLy?9JyEpI+FHGj316
z{qrt&UpFN@=!|Q(Ty?<o0<Qn@ck6e)etIGAuujo$`a^!e-yr%KXY$a>?N&bWlryJ=
zYyWNMVWal^sDP}0^pAcses^y3$S!N1_d<AU*FMjl-hWcz{nO1Rbe(zN;#0!aK5KW}
za_`LuAM0`Y(MyheDIhOiwecxGOzAx_MDDZWZr-=H|CBJ8yzcGyoiU-1M_3P8KRDOm
ze&x=Gul;uO@$ZL27F&MTv1iT<7ykOA&W9fJen|hYn{h;r_RienyubYX=;RP&{(|J)
zou6Oe$Q`bmSXjSU_sIjSd#pROM}G&M^zyNfEjGT8H^@t@x2(s+3+ol_;CI$l?4>^A
zs;K)QXxeXX<UGJa*S^fkGw`i*zO4Id^k&7Z?OOHln?6^XTeM|owfL_4e4bM&64#2@
zW!={?zZIzmy6j3?zE0mY7P<J8`ErWZIhpdY$A!dAQq*%a7p>=BNomfF$gkvS_Rp-Z
z?8jJ#$!n}*>}yzu&g=Et3$`CnJlA6VW8e4j2LJ1P^`2u3>)O^!{eAU|M?4g=?`9wI
z=vUipd+4n3A^U9B;f3Bh`{4cl^>LVVoadI@v5$@zK6$I9a@9iSjfIR;TGaC<S3J+H
z^36U!V&e(3I_4~_)UKrSGppPdf8FW0>p%G*$hw|otw%m3E}%6(E3NlEoZo`zQOs&*
zYpKWkiZ1+V#UT?;Jm&kLQcN4SuJxeOa-QR}->o)se+;JWe_DBcKHJu4?t`<UwO+XR
zoXb=jb^NqnnunlW`N8UstDaSEJvU>1R2wb(E9L9+t!mS*=iID27V^76_KzUvVIa>-
zIltkYNz2)%!Uxd{!bdO2xe#(7^4c!6qOOa?F?zBh`Ivr!oWr3PKD71&KJ^$MXwD;$
zLmzrT_{f3qL4E($qLaqE=8>c{_JP<3%{VIGDn}k<oVj0t&p9gNhoA8yaZUU&9?<Bc
z{<Nzm?lS1n@rCCW#4qiE*iAbw#4g%_W}fn#hVs^m#BWmMIT(J!PmDuWWLy<3zcF6e
z0b&p1Y$5WL(=IgSpru#qK_d^M7bH$9#ee?ikLM2k=_B)Z<>%xMJ^Pl$r-h)lNB^)3
zM2_=o<TTEavk-kC^PBNQA8}*(i+bq69t)9Ai<Z8mdDB|Y$an0cI7ilfBXWw&1N;PX
zJ_W+B7C!#m9~L-w+?Ro(o`)o*UH6Mw%d_s+taX&}0WCQfKdD_VJzAdi{*~XUHy2vt
zhabVDenF1=QT_b`>lS&6yrS_>Immnhu?zjoQ}pY6M2{l+kmo!Ly&&U>e)L%T1}&$5
zw1-~&#eEm|XPhgOpQuN^W&gx}le|p3oI~Iz(0ZPweDq*9h&|YS$BA8!->}>C09;~l
zx1P5Rex~q#&;EaZa?`UGpBiE>_20XB#5<duIi>LYo%dMiH+uSSw9wI8Uoxif{tx}c
z9t+v$;y2o5oUQjMkwY&u<F4m~$e~x!T4!4epZ6!yVr#KKs~!AKWNx8rJs_??@}wg3
zL6LgQ1InS9Z_wz4&-}OcDa;ECX&*hzLuf_xDBq&V)0ER5dUPH#f02W>5IyYokb{qY
z+J_G!??Uujm~<bTRgU-XJeGbxEvY}U<{#!mrAR!3#GfMN7OH$!+7<6}tDOCh3)$yc
z$i6KpvQJX9^t<@XbE_ZhM?vyFNWLQ9v!CQX8om`LtPjYcmpEoUVZVnS__Rko3)LRl
z136a!Imdyvkoq9@L1P~@NWZOe<d`?mmb~({ztAA^l_K#ouOjz>a|?BzE3NaEeS`~{
zw@HyWOpDgORQ<{QrJ{ANgkI~pn(A@U`hH%u>ACfOmA0F7A7agCjX!8t9O*bFUGI>y
z)-jDEOAbE$()%>#k>+pK6Y?+dk39Pi)+^*J<op3tzu`|s*L=nf`iVUH=?6$ZEu=pn
z{RY+V7OnEwr}AnC_JgdKE@VAZeM+l;t3?xEX_51~=0Yn^T71oaY5U{lEAznWhfH2>
z-3>ks+A7|A%DU^GGk9{)R&n^n|5|9NyV~bGdO!F$Xuf$+S(Z#*Tlb&Urum&-bD{3z
z+Y0Tv@2XaLwf1kSr^iyyEY)e5*=675&3aDj>$R@GgV9#UBYEa)l(h1l-iM{ttMZpF
zQ?tZ_JNC%6Rcx;JVp>Y?VN0*@)M}S4kn`9&{Gjv0qGw$C%Icq9^K{r&k>~nJk@E%4
zgZO<5&#igh#Q78FSe)bNIhLLaQ4i#Kk=8>W^*Oh5(a3@5h2|WJ=Pf)>0y!7MF51(2
z)Q69L_#ovLO+Aoyk%y)nK6>FR@>~u+dO-O2$I?5u*I~Y=WZ|!ly5PX$5Bf3yPr2l$
zuP%7`>d=K9Hh$~sRWANAP$cg8eK>g0eZ79loqTqnsOvQ6y`YvubAGDGxjHy##=n<)
zZQ#T}apxs3`{S+~d=|KHhd->dNY|am2W4pdoxcI&5BPLcx6cBQekkhw0sT_Gi`H|r
zpJqOG%JDak4?yw`sK4vwJGh|wQ)%o3neQNe2lalA_yBp{0HRmd6Uu3a{HFO6y~trd
zH1(jFzx<vCKJ96J#*y`aJWPG;K@a(#eu3yA4`CMwAA}E*e>5M_F8YzDpVR{>ryexl
zaRB+w4|$I7`B09Xlv_0IT1b6R<**OL|H#3ozm&sg{iU4uf+>e)-BLd7fQ&b^g^ar*
z_FyOVv4ij65nqfG_A~zIA<rX^KKQf)QV$wr9)ZjQMdms08B$)bGyiw~povS`g-^TK
zi5}$nZW;1CXT+a+zJcBFk#iyPw9kHye(L-B<W=O+uQY!LLix03A^o793uzzJJPJ+R
zR*H;=$7FvmvaR}weF*D0{b1e_N6?hR2Z>*3MSgF<yr-Nz4BcFaeOZxt=R)!dXw82Y
z|Al_t##}t&qaZ6L^`kW&j0?!R1Pu}wpo>pA<3K!7A4CqC^$t7Ws~j{)Uetb~kNp68
zsfYjIQy&_;;X^a7^bdX5OMTi0k;5MN{0<g6DYAdi_PC#CAI5$Ty~tCpH1(0Ukp60a
zR9^KX526ReAJ~nZApNGC_Mk!dpv9+Lk@3QQ_BY&L%(v`H^Y?uIzlHCzVn1;L(k{P?
z#(v~!hxVaC_LKMvK8StPS2@a&#~%EN9n1&nqmTBXu^)ZxbG034^y3feYkTmi2ht8S
z`%mlyu@6K)^%ZGP`K)8u4Km+ZAE-w;G)RA`hg~kjF6_tuAbb!$?Xusc9K;@e-=|0(
zomcVfGZ&iD@2#%6Y9Z^tBKv-j{g(^%Jv-JJka2}pw0`HT&)L*orCoNq%C$YUOUsp3
zdt9{IgP#=Dzgg*db$<AMc@KCGcn^3FWb=ToL)E6)Pqa4F?`x-}S*N+rt`ymyyZof1
zob>zKDp%<^>;9-(blP#?KBZcy`?}Ufb3X)XJa~HE@&M-^iq`L_J-_kqTzWrL*AJX8
zabBZH`Mfpzd~du5ya&7oya&7oya%El7_wuo`+pBwG6zP>+Tu0OM=M3$7yVM{)*1uV
zI;TKS+VhG^<y)&?y<Oe|-UHqP-UH3!0p54diaf7yp}wDyl;%5f7V3Kou72=478mk<
z6{zniaV`aNZqDx&c%O{-v3QS3QQueM{Ur8x$kV>k$WaeD+Qok4v|RbTzXehs+CqI!
z!ux$`QTsuEu>+(Z*pFS5^SdUH@ln+0Ym~$1_ZhT9xgz7JamC-`QCxBJe_YZ2#(#y1
ze?9%3UG_WT)ljkTmf!vG>S`YaAivjeq270>eM&Qrd7l#`&cQbaU$pJi1>Xt4V_)tw
z=7&BH1fc#tPy2&E`MV{wd&sh5ugVRc7(nxUk$x~Pij*t%>e{#O(rdpFfL-^#<JB3n
zUJXF_`u7mZ{BrfzTR-*a%fUB!q<!W)aZ6kfhak^kiBo9mp$|k4`gzY2r2hTejOt(W
z_)Q`64?gWE?zI0(n{B-ChXKfV(jI!iuQ%Um>kEdxUHE$p$bsY;+I^#D>7x%AGd{q7
z@R=WK5B2dAGzcFgZY<>eTSfl91nuK@X!KjOi*MP3pFrA$R>WV5`rQG|C(y)?BI5~a
zx!M6u{<M(%PQMf>S42O^`{W>eMb%@`^wSy__{f3UUZpg<Hu4#G*Ps^;KXcEeb0B$6
z-_t-oDL%M;x1-<dIDZa2_>OO8^jd3tXyLkt{yg&ZTc!t!<OPs%WgetO>bp?mGAsSo
ze9wP7U#?>=Epq;n6?NXO(|3(UE<R<xTmkdHBPHHk{-B(7jd=3tJx~n~5MON-d!F3q
zw_o<_kZY^B>{SQfb@~cjb8QtT9yF=%j+=fPv{g*HuC`XWo})%wzVnaohHVvD*V;N(
zBNuxe**&nu$6Fk|*7Zx|+A3~y`HH=6xw%KqW4s?+k-_&I+A8XOLtCX?<3u_4KW!ED
zxkFo{3-_aXf7(W9&P&=VG7mh?+aBOKLt91H@6G3S`Me!HzIWaO-UIa>xVQgm?;QAY
zw_Iz%w9ge<i~QW)_xWh{9`GLU9`GLU9%u;<@cw|lPo(b^@;i9mD+G~)1`FSd`_HSx
zhn_h-<b4R<mjHRMi1$rE<k6?^*C3}ze-!l`&3bQzcC{Qn-=|hYo_2We20JKMy!{VT
zuibdDX9Gpv6HJP{7sGpG>KEGOIUVnx%`1rCLHu7Bw~daxaqGn=hs-HGk956neaDGi
zkKeG{^Z?ZNA*)Thp7&Yrdn+G%75NTL+V8CJ{%NJ?@^@Ohsx6OyK;lN<1JO9z_JVs3
z_|1`v<iM^6{%_=0Z+6OodLFBPZ-n<lL5(YYe_zL&`E8-b74Ln5%pXwmoX#iYKrLro
z0eL?cBtL?@PY&{aHAp>B$K#R_mp|QM!iOQq`@x{T_p0@kCU1hQM<8~B)Ms5&B#ta3
zk8A#h*7(uC1J5`xo`;01JiE>YFBjgIXZ(2|mhsyB>NA%cwCVey`iD5eF6(`I;*{rd
zDyQ=uJFpl3f|M(&9*f2fkbWpyd4qXIee9|>T5*t;kDV@L99t`DK2<uY-Io8zGg*;+
zt0I0Uo{4|rh4z#tU%*$So}$Vrt#Q87+Y3zCe!yFSqRubXla|I0pzbg6OH#DvH+<Rw
zANp>?LmwXdW(YDKAaaW9-gw;I9Y>4_K=n7Y)`KSh@%MHWHD1UsNs<1d7urJhEf#8g
zD6QkZ-JsXPKmPP{plHoAi?8i4E}$ze7*~*ez9N6;jQM4u-dAeBSsyJ_dEx}r^0c(R
zw@AE!_zk4rAnQDM!vgo5+IQIO0PMZY-Qzyk=Kc^wFR1sQnty2*)O>;;zy)qvWuHq&
z-4cMr1MMklJZQYDJuVtM6!{$&{Q#};(tfi4qn?Etr%FGw>_2{WNblzZP}@<vHLs|>
zg>#q_e?NYSJx3LO2dDP&J3mF0gJyo~cqy&+ShUuIR(be}I_}u1sQz=&^piLRbzGFz
zIIWb%j+R30&+s=VU%B7fGeSk(pINkyPqk>qV{W6~heKQbfY0BVON*Kpm1aKz>Ucuy
zIIzC}lg2ImNsHJAQeW+YMh<lS-QA@7tGO-5c(_o<x!Sa@CvA<^b)mUvU6(4Qn=7wd
z=Ns^KeN0LdcbrRs?6cV~>-m@Nw|QP+p{~!Wk9IhZQ@m>9Q+}Azdt#`VcD(2x*j#Z{
ztvKL(-Zh?S`{7z=$<r48xY5);_jT!#vygSug_eHh6XzD{`ew<&$FD9wP|kXm6kX#*
zIrsHye_9$lU8v_q(A<ZU4-~5v&)lDJ?hk#^pvSIU_`5fX=iB?fF@EB;F9h6Ikw3^2
z<evdU-uvS1GbV=UgHK-II}6vWyibRVuA3YxQVwc4H0?u^KX@Lp|KFe7^sL3FhU7KY
zf4=X@`@@V6;|oo^Q%~>rXb=7L6V&s5_>7zD_d?iXA$BRw9Pzt8tKRxq;6h6;e5*eG
zv5@|O=yRcd_rTH*-_l39BI5}%-$9kPXz~MyK2XPlb%1hc3z2gn?StsU@2)&gzgr5E
z=IKi1Tz;b*KeaUK`vOX{4!Y3gC!J49Yrm~{*LGZai*oC`u&%gFTdwvgO}}(rLbLx+
zRDUtQnhS}`v}pO+;<N6xHe6@_p`+*AyF{+q$ajt`Bwn+krC0eHPf2NwOK9Sdc!9Ri
z%JaI8vYse1Uo1TqAN$z<f$-BJ^BGKAKdF4u@!|Vcx*ud;%6pvbOW7~NSEL-&`|cn5
zoZ54rGmHCL<eZrQ{*&hYd6m|60GjdF=jRrUJZQ}?_{8PhLhQ(joc~rE@xO&x+aoW5
zdY^33_@}kUzxu`UAN@aVtD2X0zkX&2T6q~het<>~baNr=fQ4z>r5~W`E#znU4xY+c
zH1=5KyazxY=G>L@5onO}T+T_5qdxhFa@G&}q3Z_pq%-dv`ucTKLw;8PAN{&6p%3I-
zh5kZozp<b7kMj+Xb&LM+_v_%JmvY+2PM-TSp4bB)nsXxb@jfc3cJqB~kaF6GR-~Ng
zHqh8hp1}_M1C6~}4_XnwFpeN{=vSmY#vP;{h}{;dUh08XInNEX9Dh@u7HQXo<h{9t
zt$jUdE%s^Lkryl^{w*ZFh<Eg8oN2pS{?-k{mzliyjKc4#$Rp^dKKWQ_=4~0ux`TZn
z^M&^5Cwh>x5PK}dFCh9Y#2yRD%dG|3w=@@8`!0)LZNF%ovOloKk2oQI6e$N8XX1l6
zLch|?2aQ+y%{~W2PV2ER&~n;^kNxn`L;bO*U-t27YtIZp<gN7rJ?KXteqdgMtluDV
z7HYfLlN8xcaE`=z9_KaWapFSvD`{!;TFAJjMaD<{QEeJOfPXmd+c`fB`zFBOj4$@U
z=l4vEEB9OKU-W`4y}uZ;*z&uMJ#%Krxh;R^g7~99?7yLzx6oPruJJ^>@bRBw(tP~W
zibE!xc+B^Kbq>pOfvomtt%sj2{P(1DgHNBE6IhtE|CU^}<BYggz1qGt&y&8-XVr(V
z>n42W7w2t?*va1?0ac$xYrm78(_vp$#BLBh+CQc72S|NI@^;ex(Y_+{*+P}iN~hIN
z+*S+cT)ELcUmVgt*Ic;AJ<Ha1UVGtObK&(v{_^gb?H0*37rwZ`{Ht%@cd=Y^;m>ov
z9rf{%opa5FXKa4_@0MR@sa$j6dwq{Pf4QX=%{3QZa!lvr@5*(`S*Y_Lntdav>oI&#
z=OOpIAn$vDD$n^Fh#biN8DxI~!Ux$`fanFaee{5wM}pknfb<KbT~PN4{N5X6e+cUL
zHFck$_Ue0K*a6}<kbZ-z&qZ^e1G0W&KS-Ro^l3S9XQ8eS>ff|9zstv7kaJm(_y&;!
zIVT0RU+4jKUDoID(5%DkKMq@Zg{M}#Y=Io*&mVEtoBw{XLr$NsLn8;e_9xV{5IvQm
zrN7#I?r$t~^#lKBMIC3QtF`~|dfJ8u^gR%bhqSc*9!>Kj`3cl=@)bya1K}q{&3mka
zpq677h+I<SJOX6h2I&XLJ^<8y!MD)0zOb%Z$a$!R+?SDWG#*$#K+c;$wNKko|0}Ki
zcKL_(9@PDa()7<l*ZM@c)@Sx>f7z$7UsL2959C~nzfa40#P^l(4}Rf!y6S}{E)}sC
zwDvpjIhRn>{US7e0yU1H=@0Rg6g6(DP3!n_UfEKp^GIpNzuHJVREoq6NPe}@HGaBo
zpjT1#lgBJn|6SFiUHgW<4>@7TAp=*hS-)n3n)c@p*<(mehx&g9b{^PeVE3AKgNF?4
zQBSW?)4rztF8i-~-jIQPFWh$jRok^|pX$F-@lq<H$Si1IEGb?ksu&S+UdF1fuN7Ai
z$8)w2D<GO0Na4lSYl>NE{ibwhEi-L03$nHpSI=rBEx%N^Rl8IwCFNG?mPpxTqj*zl
zn<{fQR#9cAb`hwsu?nk;F-?+XO_ed~s>kv2TAL_J*KScdqB3$>EYEOd>#<4_@GX%H
z_Qh36s~&k!b;j^xSoO6Gwkot}v$|PeX|ipJsES<H>k+b1bu6=?$S!HD+0?(O>lzgd
zw$T@p9p80ZJhp45EwR$EOjCtXV6bC#w=~nT&Ei{QP(8b?mMwnuT8`6HL94M<qs+*4
z<g(rhA@hdCa}8lji76h}d(X?eE6(+CyjJb7qQS+hQiCmisZ`5S@ixVC>ntlRD5YA~
z@OieiW^k&dGA-ZoxuwUF&%n>zMCOVajLcVwZ!(4(!zypJX`#l_hJW*d^$@u%5Q#PC
zD<_txwd57bB}+J_0VPsR5iSj`@}_jRNLi~lutmI3#YVC^)UIcjl$mXa8f<euS@Kqe
zwDM(kH!@5$hI}e>QY)`(G<BWS6(T)LUDvKqEU9~R5ox02GFlMxkRn<ft(TTBUaYHL
zZ&SpsqQc@;w809lvntG5FJfvHR7AA-OYZ^ifv5)(&aYUpSY}lDT)Va`v-k-Obyatj
zCu}BxSe06;l9~htH?xS<5sTl<=CpeCQNKj5TfO0l46N@_>RzR;kjNA-)HXbje<^9h
znvMKRN#;z5D~RJ|8*<UqKngFl;!5kMZZJUw(<ZYZfzrexRwHTor43rOOQlj$&7w4k
z4NW!NRGCwEMU|b}MWDjQDy%NXG)a~<RmP~R9>>dTZK5b$yG7}U%E)E0Jj0c($0|v{
zw?s177gr^%dgMXX8N-iZ)z>oEs?eg%>SlqZ$+jh;Dsow`N61FivCM`dyQHyZQ~#!}
zYg91UMqf;JeAjL9*shtj#7f69O%+Cg!H(74(oD-Xi*JoV_3XA<w)oX+IZjgrt;SZ3
zGNad}ZO50ybIlP+DPC%Ur}7>!yjWLim$Xvn6%8(4m6~etOQl+tinl4ATW48mK`GU;
zhR?IDHG@+vm1+5w&n-Qcd<K5zCNfvdU}U~ZOE%fqR>QyfX6qqxSs=>$lb6y0gKaBH
zT|rwE+SDg<HtEvfDz`PaP)Vydu!a0kwPvz9)UIcjlo@S^8f<27QZ?qpwDg+|l(m?R
zVYC~|j%7NUO<gB-1&0AGR@`O{Z_-0qdzH0b#F#3mh-mYd-UHqP89Wd>4P%)Zlvlrs
zWtqiKXo0J`t2|*d3B;<@QkB#sFnBJB%<~<rf?wxNAEoY9>IzkGy;sp%Jdl4WX``Br
z{Yy#aOo%Ip<7FFE(bPZ+FSX)I>!)roK?Tz$vmk-e#3EKBY5AoMTD421Qc}&LG>HvO
zHQZF0Q+Gv`o!Uj9!p17BF2*!TmNiwzsH+~w%WG|-C|$cn>4?h6WwAWNm957rNx-*6
zGT0YaC9QhoLDd<<k73o<GT5rnqRr}Nfu+f|C88>FS+7UPM%A&*h9bM9v1U{MrmkyL
zFxW<4Om=+NZSmNynYP4A$1+V7MuEYO)!oue%QlN|jY0M7wpzCM)oVFUQw6QYR*f>F
z*QIU8m&9|;5lJatYJsQn9x%LES8A8EQs)&7E?$+IYVk{@T9%5pDV|$rS!qEj)v|`q
zv#m9QQ!SNg`IgTuJ(he1e&!}JSIl5!zDi3r+1OUYzxihCA#zzD%KMX-(gK5RD@t8K
zTNK*VCvrCF(%>q$HMdYnt2Z!{AKGBAY?JMhGT#`Y2AkQNRE;??E&XN#Wi4i7812Tg
zW0{U-Q`bpd!C^p)6}MT#oAglDUS+KpF{TPCBHH|=_kj06)B_3USFBhpGpc;9U0aq}
z{1`*yMB{kTs;=^=Y-QI~YN<+UD(g#g)}H4(SOw?Jn?6e2tJD>$;Cip3wRj-^QqsmX
zoA{TK%$X2Z5XZ|lsG_NX6kck@mDW$)V1f#!O=dv?rHMtXM$+<28?<VdN~NTlMQIWn
znrgVIGN<l}Dm%4{K!uG}SY3>1k}PYgj8RuTj+fWkL{YkSi_#I5k;`IvhAUf-Rg!>j
ziDa-Zu1Z?<$b+ggh9ASKuVt`Rp+%e3%>ql4ZA(N|<g#9mkd3NinGHpDNn_2X{!Lxi
zs9>;-zL@OzuG`|VT{CTom5ybaDvSby9jm*gnU-x9-x`DJ*=@CK@vGNzoTds|jjbAG
zMz2fTjxUMlnj?}%99gi`oL@3!d&DkjrT!`!T)ZkZ)#8^*wJa5HQ#`lMveJT5s$~tI
zXIpCqr&=o0@-3fRdMx=2{LD>cu9(5de3h1Lvazj(fAh`OL*%kRl=mkur3D7tR+PGe
zwkWizPvmUUrNLEhYi^;ER&QV?Kg`8mWo_E7XP1=u#t=2w%-*DG%!z5~HybExF&o2Z
zH<lgCbTpf~PU;E{16r)O%^KdMhqCr6YrTjuRZtPp<}bYmya%ElNI1V@#bTLJ<#X-Y
zvdrSg7#b%U$BR~Vl}BYOyRK48RZ>$~Uz)S_Jm0}8IB(wcQR-f$u22Qndljw41NoPd
zHmTXPIbTX@Exq-2#Cag*k8j6`dEQ*|02vx=aWfhkF1GIDCHn{zyvPMuP+7*H#4JD{
zBbMl`BT8sXXCk45x~w)b@DrMqT63J<A}=x_3e3VqGJmGF$XA#<9w?4$X>=F(g^Rgk
z-!Gxnrd-@q+1&R{yT)wLV5DVf9p5TS*nxa2O-M%@sK0Kc3;)VRsiMZhdS3CWC{f2N
zW0x`V;ugQC&0jU)fre!#@9%n7Hn_8K?DLmRRQ5&Q1F;@x2pvl5L#(*FDJepvVr(9*
zDa>j#K>jUQ6-lNAUcK@nQM@YUs|iK9;+4uo=tM#h`SQ!kHI(a|_c(bkHA4NrxUI;E
z>oihvWknsu(t>>9qFCmyya(p42O`^sIzy@-qSe<pcm0*R^?C<KmXg%<eFg7<R^)*u
z=5u|KS!@=xqQ3nSOeq<M62k0#GAXmVFNpC#qfd*TEndY43oDJRRBb*oZ&b_LmZ7!!
zDy7vLr~<2SAyUofIj_0e9;hF~;^9TLwKucAVgNRCL(Q$eVHB)JnxU5sDw{#C)diE^
zd^Xetb@M4V`;#wdftK?xC2d->S=N`5EF+`*=(R2e5hs+2MG7Kg5GxbQtgmktM3r-G
z3da@1@lX_xRt@1dkix4|Sk7r^Mc9yFye08mvyN4el5dIEcdfoG4+Z1O<;yt*;l>LQ
zh?3FkXsxvTXx;cxu{2skSgNs(vJ@2g#j7U7i+ZBs(Q9ovX<5A-|ErEzx+%qZU9t5F
zOoxJkT>V9fY>_Bll}MM0M~bQ?AJYJ;!9uG=M%-0FjEkZWoMc#9uY{^6@{If?#b6Un
zG>FvwQz}|YMXOsS3Hb3Mb=-!6;y=;m)UeXB63SB0DAZ6|E?>tfFR^M^qLzkybB<z5
zt7%!2`XYsq%V-->w$*)+$#<k4tP+bKscK!?RgID;J9=#yS0WR!!r&Y1NX5uyD&a`I
z$fc>mC@|P{8tsxub7hxSt+f1l#mVYKq#~CQ!{WJ(+M6nisNm8NS4#_Gg$oqaVFNF&
zpiw48iHa&sV=<mYv;-mQ4T)SwE~TWN%>P#qDqJ+^h)_s2V1ysJjJ8@L9z#T{MT#5Q
zF@*(I;i9~?Vo^k6i7yiI>x2wVW<gr{Mn#Ra8vYdpkcuVt$6BM>(9lp5q7Jz%;cL&M
zLQP*cme&8<aFrJ>a*GY-iyGRk|3_&G67%zdMfu`YQ~K?)GKKLb6`CnrAO{)+aqKuA
zYD#@t&ux@$_&0LdSWw4k<VB?#c`^KGb*}4?>q2Xd43l3kSMpz8zJwPctYV9wui8{G
zWgx97-Gcl=({&eh2T=tQ6XJVSE>_NItm)*}3&aS;u*4)tOVZ-UHkZZ-Q7ojG)mYl_
zZ(gt-B9{e%ibc#z6&a0J2HRGYsDdff&}d0%38B`N(3WI`U%V_{;U9EGN??I8MKP?r
zv{_;n(2Sy1P#XCnzj#Gi0pp9}x#}8+l5{~)Y_jpHNYhLyr$`mAkZiyxyYV{8iHcf$
zv%k^Gznp8+O?58NGNlV%G71cKp_;#FdmhL%rdB13-;`U7ItDu$!-#;ilBbnTaa2Wt
zS9cLFU(c=kzo9UH-9SXI^Hrm)f_$`vyj%$@F;Swifrx3g%(O<q#y)6Mg|a3KOsPUG
zL&9JesueCQL4$9wP3Kzj#R}yT<(x)qOZibMdfgyf#>n^A;Fq!Uj55Bc(egT4SxuUw
z;<1(MGzeeFMx$bx)U^uBR5$9PT7gP_QEaJjRk#o>G5i=-L!}76a9MXzxULr}A!6tT
zMfN{h-IgpXsi109S42(ZGAdD(>%4BW39F!-U(RVXut|PnVf5eP&_%g<C3&b<9Jy@T
z5ePP7DOFaJVlA-vwi!l^O4&yJG!CI}mZ~DZctxE?tdMV2H=Su!u=o*cO%bcW;@9ge
zEi0uWZE|S|l=I6u=+H(R`Q`F?PI*C5k#&`}PKkU8QQv|xl+IgWu*+?1*FOJJ(q=Wk
z&iYbPxz=({sX?VwMA|HfbUaojmKjl;?9$W{R}jZb){IJtrUp`YTE$6MTHo0~1_9G1
zvmiqwW)-WEwEWTrtlFhgDXC^rn#6{t8g8o0sk@@ePVFL4VPh3m7h{?v%bF@<)K!n;
z<+V0Zl&;;PbVOz3vRIzs%GP6*B;Z>j8SIOzl2$$Ppz4g_$FS;a8EjQ((PnkCz|v&f
z5>XYotk)xCqv}{@Ly=w5ShJ~rQ`a>r7;K|2COf|Cws>sUOj}~5W0|H3qrhOt>TYSK
zWt+vf#-Mt3TP<7s>a`rFse)Ext45j8>(aL4OX9iah@=!RwZKz(4;WsoE452nsq=~k
z7q3c9wfLn{Elb7Q6wj@*thAt%YFWeQ+18rDsg}yLe9PyS9!ow0KXVhAD`qe<U!^6R
zY;3FH-+Z(65V<T6<^9P^X@SAE6{W7AEedVw6FHl7X>gU>np>!()f?DCeyCbASsiNE
zvrEd1Hbf0Jvp1<4b7ETh%?8R^%*HU<jb+C&9nGe$le&V#fEFumvxYb6p{%{iS}$Tu
z6;wpD`AhEs?}4ZX63(w!u~=qQ`CPlUEVK9t4Ruv_l_zW_fmoGVs*;)n2G1ptdA@^H
z@aw$kqtv}hU7^JFmDMg$<%zlAf%=z{uIgI<E4;e@2CiQBZqjd>^X;U$&_my(_kj0+
z_dse7_@yni%X|gzftL1wpQbJCao^rA$OC?QwSmF*gVY8F$u!ZAp6QxaRbY)%T7IkQ
zRO>aya%=0g*jwg3;62dN9`Jdhr9Ix(+Vc_8)<(?R>pkE-;630y;63om^nl;9{4&Sg
z_d4Q%{3nxsQ?t2$DJkL>f9XBoJ>WgyJ>Wg?3-mz4);O|1j$GDxqm=pux~I9i>>b-&
zp7naY2fPQo2fPR75fAvKavm8--wp2p?}1;O2P)6smWH%^>vq=CQ>CFY)f)u{yGqvl
zg4D!0GLBbpNt1;IDSzQTP<0RZ&@){+FS=%4?2|cwzCb(>8U4tm2zk<bAhQRou(J4>
z8}b!PJ>VT&>JJB%e<^A6nk}-vlw>(P%8y<*@ldl=ENXeNq<9r6k7d@En+1`|g-cU7
zt{{$A&=h-N)ewFIDZDy;<(!5_gbfMC8e%ZbI#xkSz9nAYwfeF=6s#+kFXt458!tp4
zN=C1vwbJsVb>m0H(r67~sm40WQc&a<ubL1q>WPX+ueIfbW$o(a_+NFz(oHGO>x!*c
zU^)~O<mxX<WQ#=cszka}JW^CG`IrV!4HjB0GUBcZVq6r3;3UJ+dL>jnk!R#DDF&Np
zqCuqYpHk6MDq7ttNx+X6spB>j6#t1fr-qf5l~9&~Mxln%a``$=d5KlS616nsn{yOf
zT20HE)E6m?Tt?f7vaRlmOui%aV3k<>NLB07u4<G-+0kpuxDuI&6$am6M=C}xQwc}v
zMJ`PhMuEYu(`c7Onk&1sYNh4ZD^6A?A{Due7#7cM)ZSECL<N_IxLR5eD_o$U4jXuJ
z1&uN(N>o&78jJBHq9q7XZ%E`iaw#SCWd6T`P~oCMM}$JM0VDj#Wwg~2@fadnEmGXb
zjwvj#3K!+A6^kMoOMH=tUngW}G7HkmH!5nZ)$p$<fK)88Kh_%6hK7ck5Ov6931532
z6>9prv9$i*hO4}Akt+>0wA%PjUNpi<$uDWEzHX_;QmV9>__8vER4NrOP=yQRK%*dz
z9mhjWsZZ;<jnWPOMlMSWN~9Wzd?V2+ao6>{GJYe&<kwqO@?T!Qgcl*KVvC=z+Eg)R
zAgw9gg8V|$br<!YwWeu7;R{(OT+V6K;N;f}#0bT(#3V>d(&EQ9m&OQDETov#SlaM!
zUa%e_mj!}~Ma)YT8I4y4+g6mQf+^I{Xh~@aq1KhqmSluqyewYfA9O`ZV1Y43F|53_
zSz;E@jG|Uh8u=o>ctu$O<BQ_C>KccVbU{&Uvhk`&(@ZI+NENS;Y``eH@jA+hiduZL
zzxm2$Jy*p9ja$0Y@-0&=J(hf($#GTSf%Hsi__0b1-Ii=YSKNiwu*J{R2o1+8jpa5n
zV~ZNJ6-(+?Li8&}3!>L~xfqt{E>&pB%TU*e)!P)Yt5~skRjRt4Dk+N+(QB<{m05i1
z$CbzN)Gq3m@~H_Tm8X?WacP$eT!US&rSAU*W%=s{B6^*#8f6vai~1WCmC$L35{(T+
zOtWRCH44TzT_{sCN(xE{m299534>j(!V)z220QPK<m!tR$~BgA8uga)qg3>|DX$b;
zEw$9DU&hZ@ga7~Ry?e}V?Q<6NYnv(oj3ND_k;tKlzZjwrQB<VuE(Hp7Lr<IvlmaEi
zg3&}FBqeDSON>ST0L?E1Le$id5ETOb1Be>Gl2mK3c&G_RiLnMbG!>-Cp`udnInUhp
zT<cluIlb?@_j>mCu9>;!nrr5swcckRzUSMYSyLPF3~i1x)7#>qFyT`u-CNLVQF4(M
zEi3bAzZ0F@b7xX%@CM?dgu6Q&_lZ=3CT1agnaeSA$=pzywUSJ7GsFEh--4rL!Zn;w
zm!9dKSPND}$eXw%^ED>SsEVi|;pF;Itz$&7HChTRJtq*3#{d<#!k~W4i!<|KG)4u1
zhgT(T#8Ejs0xCdwnif(}0^y`tb2Umm%|(P2UlG%on+m@26<HG@guJ6Qg<k^dFavX}
zQvi6K@l&6rA4>XJ@4xW;P!e2Qm*}u{!5Kjn6VBm;4Ve{T?FqKfZ0>)_hczg15Jz-6
z8s*gZm@9upoGi{n$YJ(#)lodx_ICv-qFY!MM}!0d4{w7WI+>pRx%1B4`JtqFaTLSl
ztqDWSx?BLA=QMMIWXL{S>q<m{0_s}Y>_!Wwj7z1~D3_zm4!oxfloN5u?dHL1u0%*c
zNSs~y16ebu0#T8Iw!4r<s&KMRy_8ID?vY`1{E4FnA%ijqPo>Dkr67~x?4#idCrtrx
z0H?y_9wMi@<U$FM0Ps-6nbdA^QPokLg!k1XR&vu|Ju2#3xg=pOpwE*!%^edMma#4&
zoCBGeI>*+m*`!g6wSWdmesZS_h5B-mJFJl+W3o+na&y;0Z-QKw(gHV0hum4aMJg$l
zUF^aK2`ItDQN9RC6<;R-3ujyAZ>O2nOo1yc7}nf^&nM5MZC|o<Wlzb0P7S%MT$GxS
zgLvshwN<f3wN0N}1g)sjJ#xo*W<jEZJIrM2P6-t@Uz1xCg(ZPD$t9gn*n$fHahj^U
zlwYtEpftj@bSrH+v({~npNM<w;f#Of>7k9#pK7J8h*Vz0iO`T5Xd1~v#6)=8?#K|6
z@DxsCNtUllB$HeZtT3#y0uT?tSZG%PByrP{Ex|lpi4im5;f81XF|Cv_$ZV5pAZ5c0
zOjw)Q&INO9{J_JroZCLtqw{Jmh+OedxZ+bCa9TXSP6zAYqMBm?5dcGF#76~3`D5~~
z#4xjVz&SR&6#?3a>sp<I8YBFwCf2wYZBYT+6stuhefY=vm<>{UH*(ra$R5R~icu~f
zNpc-`p!Cu<N+$I@=vA|dK-M?O4%^qlD$P!cGi+(Jj-F5G_~NAc9AsY=Gvs{lN)Dq1
z*&h@|5sE8$PO!i%_uy>p)<8lii4s%+jQ|6{s{|qf9Kdk7h#v(jV?~<HguqBH6$@~1
zTUvifX3mdzhPZJHHW|d5@F|qDwctoGj*?(rD--2I*XZP)JCjg@HxL&k+}+{0Poxqw
zF$>|#T#lJb=7!R&m1L5e8Sc0F791rLuHl5b^i21}TCgHQ-oz!DuQ6dpRYVO5C)bB+
z9V3dZ(NbXPIe~CI2B<g&&tq&I8z`?5hM<DL!?PN(H{z%qE)x|XJWUHJD1mU&thpMc
zp5`LLim!-i%uNMf`HHNG5JKM3n!+yub(n!U)+qqI&gdUX`oi~LbnS<dR@D<ZH*jv?
z+`zeka|7oF&JCO!I5%)^;M~BufpY`r2F?wf8#p&`Zs6R&xq)*7=LXIVoEtbdaBkq-
zz`22Q1Lp?L4V)V|H*jv?Bi?}iO-Wz${)^8KCGC3*^b=pZKlrsz^<<qJunl~4kJd+b
zV7t?=!us~zc@<-A5Al{}-9E{gJvVS};M~Bu0k8pkt2Ya1A0d2h<RiT4=aKKXft$X8
zP%*;qH}<J?*A4Iu)4TTSZk`PEQe#Em@bo{8^uM<&W$t=gPBF9r+4x=e{s2id{he@h
z{G<5tU;5r@zEkGhz{_j^eFwbEG#^sebF&}P-FyXSsSi|*Ucq?h2(M%V=acbDZp1mp
zxq**+15bD>(hntl@%wk@hmt<-hwVK5YuUgPzU^PjQO`MEegpovx%2WfofDiJI5%)^
z;M~Bu0sCpm?75K-e*=22#pLUUfAsT&&)dLFZz(qwKktlZr%M~)8>x5mys2edv5xT0
z4<(fqy+9!MBHfV}nBi=BZs6R&xq)*7=LTMI1N@WV?w=1|-A|cypc=j4R?pVw2F?wf
z8#p&`ZeZSk{v`FtpR4FYNuT+=@qa|7*EfB7d^YJzuKjcp8Z+bBU;l4(e*Q{xcsP*>
zFXda1@-UH#-G$^X(P6ICI;dj8Ih?Sew<6qRf-N-L4gx?}gAxaEM5m)sPK{5u@>j&k
z;%pF5>L^}2cGVrmRWMOC(XCE+Q7S=}@`O7hoBd2Ce`k*WV@cv}@r(8a;)V2)PBS;8
z!z-h~L)?qsG%#tiK2QLi=QMMIWXL{S>q<m{0_s}Y>_!Wwj7z1~D3_zm4!oxfloN5u
z?dHL1u0%*cNSs~y16ebu0#T8Iw!4r<s&KMRy_8ID?vY`1{E4FnA%ijqPo>Dkr67~x
z?4#idCrtrx0H?y_9wMi@<U$FM0Ps-6nbdA^QPokLg!k1XR&vu|Ju2#3xg=pOpwE*!
z%^edMma#4&oCBGeI>*+m*`!g6wSWdmesZS_h5B-mJFJl+W3o+na&y;0Z-QKw(gHV0
zhum4aMJg$lUF^aK2`ItDQNApfs_6G{+K@#?Z2gDBkj2AJFH0wwp@b*5$hG*;L3l}I
zlK63DPYHpBu%>+EIkv=ZT#3OCkF;uJ>SFtxqHD4u9=TJ+Ve5|HVIjxaL=iWkmP-k4
zKb|d3kx8xvxxoCO8nRI<lo-`Kglit95q_bJ(pG9#OB}b@?96)r?k$F+-{{sQTc-&)
zg#%m>#GG&{Y>FF53YRNlBD`&PWQa+43T-}x#qc#D$wlMz6Op<WTcfZjf$*XPNvvP8
zC5ZNohnNWuH@q^~kWHDnNt0?It%4btnA=%A7tHa>LyW=fQ#}gENaPqEEyas?Jr5y~
z56_rvGc0~aG8N&eN2a?E>n;_#!f=2HfDs+yD1r15A53a8X6<BiY&eo{`7P*5y@P5H
zjt1)0C$txBQ32Z&i$=zOO)@nbVUVh?%$&}}NAXE~l+#RUt`H0_Hx$3*a*+MG4CN4R
z^<BQh1|Pr~H;UP*%M4pOOE|XCF>}Hvb^B693(c;LD0A(LYfV6GT*>3eoNT!xXKU{U
z5<*FopbBUN7yw=+5E0-2UbnqHUV6k<#T5~;isvZ%%w2%(aTdA)3qHxr`4P|1<~TFG
zEglLJK7}$=3tBBoF4Cf9WghK!qLX{>OeziDKwOk?cZcIXkxJ0SEQBv}Ic6@I8%ncQ
zl1Xl6xZmbmaFk5Ah7;=2Gu;zw!HNiZ6PIMZ#)KJF5j7;7Tpy}+j3~B7OM#{51j6wc
zpyE~-)Q@>_W<HF@s37q0s>F>rDu+iv1qe^mLJCSCoHT2$MyaQ{h_K=-Vj6Q(!B@T_
zYa)b@ceJMPOF$iFV2*VP0IxHC>a+B(O!|`dKl}Vp5?ouC=&*Ic89@~j&f$a&nH6E}
z3AWH|?tjULH7IcqM|3(G<<$6?D}P0tEY3v8VfJ&?ZFpDRVcZcW@;t*wyuicTpodPT
zXMgG&1#_Zyss9$N;V#ft)2h9c3UEQ7R26On;T1MK^u72^1CuuE0|n4|PBSM+hU~Mo
zu0#|lpsuCOZnR*^xKwJ5ayiQEz<bI-IT4rKZXT@WN`wT2#MzZUkTruU5EUtCy9;Tg
z3MbptOUdNs9vMc*pE!CDGAM)aREk_&3NjhaJ{qoY(i8v(a4JmhA#$oqE|dTX01ri+
zN$nOFRUO4icwbFoB{vP$qoTf*OA_V+`aG%A+%bV+8S4_lIgpvDb8O9;O&YaW3uut!
zCwIzFs4pkE!x||vCfkH3H+L=cCdg$eEpU@`$ep!Yq>^IU#V&l1fD%j`<%^J1@pTfg
zaJFUscA8nu6u8oYVa+Z0eDX}%_9aVK_LLmx)R4Q%MX3onh?ibeTNP_m+w{3b(26SE
zBX^8v79={j!%U{`lu%*wHMuoWSQ2QHT+;c3Ew}&>r>V+I`2|Y>N+VoLx6+m~Yu)Df
ziMY2O&iGfJ9@+@~saD#GNaaPG2o0%$rjaZ}OoX@X<{_lGmzdBgoW@E*N+gq953De(
zvH}nfz*uNk0VHwLk}bhJU5ODh;o*j7`!TJQG01F_Y9M8M#+^A~vX|^$f#8L=X!ofe
zomb;xfPRP{8sbH4Z7LE`c*bO#VevDPsR)PjYcZ_5RJfx#(msl#QIHPekCD3)!_3;r
z=GgF71ZX3!YjqB4jPR?PSmR!_MFnhAtQMK{;UDW`Hc0K=$Z0Dfdla84M!9?>$#vX;
z(o5SYnbh;3SIsH{S>Gf(Y+nnjG&?EIu%*#DdOo4!i<9bekbPCmkn_DOIgA!$e^3-f
zD6Zr=!2+|~gR`|;0|}udN>Bwf0t^7J5{L+J0K?@XeiW>X6=^mT0wcLpEWp8SY5gUc
zIX~hV;>Io5WDsw{r%=w;f+NK^N`iT<Oq35@qmz5?OhOIbKwOk?cZcIXkxJ0SEVR-T
zUdjzeiPSE+$NPGp+^8xB#k@A(g6lv^L+m_;wPU*$tcVnB;*!kQxF;D^5j7;7T<L?Z
zMig74rN+{80^xWJP;n~^>c_k|Gap7{R1kQ0RpLe*mBS;T0)(e&Aq6E6PMS4Wqtw$}
zL|E|^F^##Y;45E|H4#F{J6cotC7=#7FvmItfY%xQLrFjT{g+<*p`=yyM9vMI8#p&`
zZs6R&xq)*7=LXIVoEtbdaBkq-z`22Q1Lp?L4V)V|H*jv?+`zeka|7oF&JCO!I5%);
z1H1nn=ONXTd~V>}z`22Q1Lp?L4ZPF_^lwV~()T~-*oTt%|NE9M<MUGXLq0l1Ir+cC
z$zK6)Tso*HxWqDUt;_*+5MFo}NrNDxav4oCy=Z>p2ve_&@~ON=nG#bWf60YTdc`YE
z<82Z~!%G+aW!V@=UW_k;6Im+s*z<5%d($6<*4(6MkRfhO&C=X5Boy?cxdW+FLJrTz
z;tY4Ufn-41hi+pkMj?ynh0nx-be=eqNJ=*A5NCbGfNKsQv`F4&&9C}4Lz+(8u|zFz
z6q&%sVI^IbElpKNsIRIQEm&bBD~%&wT&?RPQ&Y=lwZ6%6aa}}?vY+}4ffRknuC7WR
zdOr3zy6umifQXVu1bN26Zj`IlOj-(5gjenB=GH)Tel}QuI{A>=h}BqWy37jJ$lDW;
zYbfZx^n7BtLd#GG>r`xMTQWNW-rK_4Es2RZvve!KwRJ`0K*0U_Eo7C=vAA3qblv1O
zq6@zOmuEZD(%Mfg^3%G{r>5y;x|{aGQIgzIw`*(N%27L_VcOKf>n#cSyGcIEhNP4*
z!3OsSY4-{Eo|OBNySpwN<1)jZZaSa8id!`z<#;J|je9t>!@IrsU0KygmNt1<{CI_R
zJW|P{(s=ERJUl&)bjvQvP6R8~YVI-G^qmf`XJ^4Q)6S}swy}Y|E8WYpR`dk!sM$d&
z-*LP6Hp6SNa1nAElt%bV9!uUmHm}ot=Lcavih!36qBoo0uOOGD=v|42nk=lkDZ#CV
zLU|ZVuqjUqvLBno*OR1iw_ddNOB|~wBC+?qGPU^o${);{&e0rdIw+l;kMD*gn!0K5
z%Z#aD3bp=fnX3z@IW|LH&L+c?ytQe!o+PKuwV;<-c38e8G!DI-@T#57LGONsA*gGj
z{gfNWYm-$GMM9qLep66WdsRP2(>t`zWOcJml2eK@{2<UN_=h+XZC8l9>?+5kox(;l
z>w<iwx;@XiV^z8aU5lrab4kglqR3Yhx0Ki(<s6OtnAfIm-&SX_%@lj5IX9=;UWAUb
zUtxo}ks3p6XTg+vis};gRDfnP+}}@Pac1g7o4oY}6aZYi?IPK27zdbMDi+J!)|;|J
ztWL8LC1!GQfsqchATcZJgpcO7_?`ZC_rBr@=*35LtoX>CeDKJzb@uqJiYDY=JJ5<1
zp(mMt9#+FR&lU<@c-#1s_9Uij<jHtmpOrVvBTpr-In-U-#)rNA<r>4u`XlliskfVr
zcO>%q;u=|XbqAH!J*xO30Y#k9ka7=ut;1eMz!AAHFN}{-<i;`VP;$tU?E@*1vY8RN
zk>RDu?V>4CE?nE=EF>}Az4$>I4X?v84d$~N4ai3tmE_xEsHIG65akRaOD{F-PZ#}A
z($9JSbB}!}iQ1j74`BoJz?`ph1Lp?L4V)V|H*jv?ZEiq*9@n4Evnh$UdA4)va|7oF
zwl<*O@ba6)_#NR^$(cJhaBkp18?f)k2g#qZ=LTNg2F_>x)!mYFpmPJ~2F?v!-avly
z&i>`flW}g~+`zekvVrw2WJOBB8Jrt9H*jv?+`vb-f#qig{ZP`+egE?w^Pwbas9)TB
z8Q|`>;PD;qJ~9%#;wbcyHTlpQ5}k3#$K5G|vvY$FK8CC4_XO#>sOA1yfqpI#?9*6H
zdLz}a7f1#gBBVnfid+~Au!FMH+yd}q{!UXJ&Dq-!Ssu?Wn&<Bt#T0U$#i`al8+veR
z?=mD=l`bZ009%1QXNw;>D-i+Xk<RTelOIalrd8Cl)02QnRJomPwQmIP)bW^%d{0mf
z-CnLWyBvqJ3wi=~+yML;?b;qBNH6}mRoty>*QUMg0OUUWy_LF1+(`>YT0(C*@`$#n
zHwN+{^UP^U8w<!4FXrp0JD#H#8}V#>;|AJq>Y-Wi5>UehxN!<CPv0B<(AoP$mvhY7
zjMT_Uk8*S$w3G2Qx<r3EmtfwUlZ+=`DRSyjXGlluwwU@h3vIa!O2H$qxc8dbKHb5i
zt@0NU?(oQEY->+X(#3HhVSA6YjJSs$8Rl;vq|c5vFY}UL!C3Ehj25(t4SQ`JmQ+PC
zbhQ?Nbj8N%a-k7f=erZR(|e%}r1y8e?R0O(SN?{DxDex3Ypub#PJ9v;{cPpU!20Gq
z^QqzqFFUav{nnOs>>91wk}s?gSRC2$zTz<Njie$i!d5o5A-JS05NlCzs6o6UnE11-
zl|%s}H2MSqrop}KHG0$=uC2Ym9dDPA<WibKgUmYM_3U_B;3=%2PB(Kjp5Yh)v>uOe
z;xuki%mMQKL{(T%!Bul<j_Ya5Kq*)%7C?03$^>T(NBn?2H*YEup+;VV&ydFUezDWm
zHfQwW8=%`LUwj=GvUC~Wjr<KsE;(Y!yv7$z%Tv;|e>HdmS*_^hNVj0nWH+P=vtOQ4
zJo#c<zDr|zRQxCpcc>MAaodB9t(Z0R2@;>?_NR(HLeAHl+W>uU@q3s)CgVGGoFRW%
zKo~LE8bRn>FySgBM5oiE!ZIY4z7+Qyady0zzw%_p6@R_RI0)Z|W<q}B9{6-tKV!vZ
zzl2qsZrp)a%rVJCqIyOfmR*g>D5OY9R}GJhYG;^ZA!gKkv3wm1GSq}U>j_5&W%w1J
zl0va-PMUsL7YgX!o_;9l=e_^3Yd@5<E=Xpi;q3ryHHc@+({?DMTO2kr;YD^4;uJBR
z?ZJw;j;9<^Cndl)%v>>IDdMZU5rs#@h;*9_6wwvo^aNXIwn+4Wd_Rv7(hr}4AO9i7
z0E==UO6oGqM-WfQwt#4=*#OKV9eUB5S0eeUjhB;-6(>~_p2O_tnxnYaa$#3JVG)|g
z)&kcHhf<QRq=>?EYLvU@EO)lpILGBlL!KDEW8Q1qNySWbMM==rzRFE=RnHnCg%(*6
zqjoV%50^FwLzRNcGsk4dD8-0Sh^(*jggWMYqp?*@*H9-Um|$toa%)fNlrqkc<wP_l
zFCh1#T&dZVZ>&0vZc58X&R?mEwj3nsC)W{;YAVr-Y9&&wAf8FEH66m)x4Z>`v$n7}
zX-#BYx#Pl-v>v|%c_p*YL>Vwr1Zu*qX8o+nF&gtm5VNqTvUF<GV9S--hX5)T<Guqq
zp;|tOim9B7NUVj!aI7H($>pI-%$m-Tlo6s|LBQz5Df*@N$Yzmr9}*mH%8xoJrmTpt
zr$N%KFjIw3h4jj*f+(DV<eFZB$gMC%jEsv}r$|GG8RtkO#Z7>~({<#DVqY#koG4Wi
zl^T(g#*jdwd=&|0sc{7vq#nXJ)Mq@IBh)cCL6_julgqFO>1-Y(2eL}SI^}OY9)kO5
zU#nn@S_oipYRazSWD-R?8F_>w$s8jxnJ-$XzLZFi(=*PwQbUXx@K_wWD&!d!9ubYq
zd49uW9(l&sLO}?&6Okm9TRjzUm6s+!3x%r{!{pf5qAF@Uqg@7Jy!a}v_zChHIW!y%
zB!*e)Qdh3!oMa}Jq^+G=t#T<RiJEE-vb{)@No9(Ga5w7i3`Uq<pdf_06C}xrp+!d(
zBTX4e{khCy>QrBrPG~d|uuv=*)?8+#Ai0&46RR%>UD;DYpj9i)TL(!Al}L!V3toB=
zr!s29X@~e(;wUa$PoY9=EC_jd9w?srF(el$O~4F?wo%R2B1SbuM!t9iK{ZDS`G@Ms
zqX&$&=$c&43TL@8)r*uVD%{x+h7@Odz93vVD2;GjMipFTrNbI#8?v|%9n5?YvsAGZ
z<kyHfuU1m}SZE?|-~Fs%MFpec`dSgmU&MKisznj1+t0)beWpPhXdSI5>64tFXCX3@
zVLK|P_2%ePpwgP*qYq#Hu|b5R6g9*YoK3pp3Tu&aOqgwd$xM^(nF9@Sc$J*+aKlRq
zT$eIcC-?FUX2VRAU`X*bJtH0uhkT305-2yJ++StM!{AIh2L%FT!-y$JBB~4z#ALEi
zWEE9?j^xp#{@YKl*GN_HJ;4yk2%<bP?3zu0xso@Q5r-|W$akJlxp$-qjEKDPH8^xt
zT#PM}vtFeXmdK}K8DfoBkr7C)R&3aaxi9rMkFb;<E~og{#gTG~QB8^X{OdHj>Wc))
zWv$TAJ|dN_;!0M1E#inS4vU*`ccYm@R;7D(+~fpT)-c9b$HO5_5HC%ABMZlw+KEW~
z%4znM=>@^kv}6h`py6?l4^)Shbf}iB2&u$6pdeZy5_U)VZqgl<yQ&g-7A-cTtt6y6
z(Hrv*&w~M4it?q13*_QBJGoj)$~v^pYm&}z0i1E0Xi%9wS;F!cMI0;7a?DDY%t2Xt
zDT_n(4RgzmQ)gIoTr7o$T;5%tK<`k+g}?{(L=03L@|h|mkteL(fE`;4KnlR9Ug%yl
zoIu#K&@vS37aO`Llxqh^GJIvG5Iw<*%%xVLykxY83xp$&w)db3N-&sGEV%h3ka)@=
zWA(~RBz48CIBPJhYKf@x7Q^or^nNiq%0Jwj0FK3x$O}c^LvdPgq+|f%ke~uV)|2YA
zM3YxA4B2PBZgd+jx@Zf@cwQ${RfRKkKu)$PRgsHwo&YA(MXRJn0vbkPB`BWh%m_Rx
z7_*BIMxEp-%$+hG3_|Bfu<kmwXefyh*%ln>3D<B!kI7JR+$)K+(Fpj~k2JTSn8Z1@
zBtwb_aa^rXDzsVxOOA+3Nf5t?HC{tO`3zX?usIf3zuT-xXCeVsp^H$cSRitk&yJu>
z<NV+_uD<gn!2}tUuX3Sr_{9!g<}b}_CZZjli%;CZ!>g1Vaa7KStDPi51uagXCWeuJ
z0^tl@=r$ClatbaSmE$|cMRL((LM9gx3eLf}2pB2z(mVt?%^e>`(is;dPZe45C}%8r
zBE3aWrK!<IG`XpPX-J8x8=V#T!c9`BuIBPkLx?_<^qJ2a%JllCPmj+g{ru}cos?z;
zZW;nG7{;{7gqM_Bkn%8*irt0eF41AG)H<l5!wIW+jl97z*vpdR3htyNyC`uGNA&5S
zbC{-q3dU&&2C1FGdCMkr85bdwi#lAnaj;yD1-fRV62Kq<*>#2`%5~Xah*tnN=`F<t
zkT+JEslaK-N8?RnYiZV}383?wW=@a{F^e<AsM!*6AV|SY_zH`4P-X{?DBJU)svhsr
zWq;ICsqzbprfz{2zy^m2AzVn?Af^@;6!tKql2SMGvL6k|Lm7Z8ykr-kcG-$>DaZgA
zo(1s%4&eR>kPH(^M>xvNLG}}0T;(q~Jb%N?5Yjt}ljh-?zMw~CXpjT&@rm+kvX2_T
z4&6Ft5s%|1J?-fko529G_LYA7CITa2JgXymVBrO^76~KqRH)#Q4lgzA-2>EKzz-Kn
z;>~e6Th<EVf;OolKIhX?vzE_TL5)Vy_~G|T=A=JF(X!!KAG4EA12fd%T=_U$l<CTz
zHVQP3G9AH>_-vf|>6mV121DnG5DoID0@DjVdvPH`B#n;ImOKPPgPtyVqBc~lz{^%A
z&oX=`6t|EPrQjN%G{P^GQ5u!*zzd}W+V@6%RYA(BN#6(+?!p^V15Gnnh?oe+x4{&L
zm0~QhP!k}o)+V8XBtGl{5&+hrn`{G>8PhCxc2VaZ@6j=mC&b~aVhJ&tW~|IZ1LBOH
zawl@z9u;U2o9yRG#e<!b<_F0TvOSs^`~60%fl^D!uACK-ApoA?_*FzVcV?D6Ol&qD
z$qR+K37`UaRaPP!@#|IdK9pwJ%sHAX5|Nd8=h)d+B{aTv+^q!UC(T{%VWArfO1T6D
zszc*cWaL?`bs7b!O%vErI?M&y<P6R+u4bmty@(tS++Pt=%_zn8s(6ivvm!kPsFQ@L
znWc1k0r+r7mb{M`PyukWuKi41#9WZ$?5Dj=3u2Z~^EDL!9RNq65zIzLJb(kZ+192^
zI_6L$Q4tPrp2sPckOGS)i;@VR$}L3nd*OEMef_*EJL6)O%PmXReA8V4n6c7f%{9+r
zv#ybyNU}#4fwr$IA*8G)e!`P$yauw=etQ>@{p}iNnjpCi(N;R8P|<{IIH9415-YwU
zqJr?~k)Evts*B1Iz9<+`Y>gm!PFUHD<IzFItuUw`bK%S~v;e3e@bG}d1n=Mz1=3LB
zD4xdhQg<{t$5w-I%ZJ_1j2iaV!WEfHrm*n!&wPRORM~_@tjW(%;g^7zGcd=q8pli$
zCS6lFeElnve*XJkaO^`#^hcGCdhwrwK1zJiJ=E<0xxE)Vz%<~2Zc??cpbnAxJ<;7a
z=@SRKcPsi%yXJLe+4!!8-Ean$j6^&o6D!M!FI1%sXk!*4Wr0as!f0(?OS=e;>WiEY
zgP?obeHh#8kp+QW>XIiVYOrUT803)GOP7wO*HbsqeWSZXY3t}t=U}f{^`_pl*oP7y
zUCK89#yhlUsrIZMu|;?1j2CD}+E3gvjRqIGH6<e2B#uJXx=wji#rfM0h|}01Zf%>Q
zN_|XZXWvjia-I#DB{fdEU^fAGmOC)Zp`*Oan9u9__)d@SaNyp*Y3iW|ck`tDzNL&0
z^8TPhA$)UNFP@xlJm0>zd6Y-*@+02LfujdC{j{FkN8hCl{cEO(<7^0jq$#-1c4?43
zw$~PX*js(QM}FR9^G`k>_8q_BbdSCzTi=^t?0%#k6^SdFfZh*<@AXh_t>V%Fr1lnt
zOC2<SRg0^Nb`Erzw*43_vn{~o4ll-#Uwo(QjgxGz3W*=v*fBQhc@4gFUVv*<Uti6*
zi}!j@dVPo6{I%Z)dP861UAStdJ3W@^4eyEWRVTi+NS^4u&-DgR-xT)W0o&}YC%CG9
z;~>ZMHhxCa4<-GA_rLJihmveD*y4XPgKJ>su41SQYyuNV=@$H{&BmbK&z*UClsJmd
zx=)GGkWXoTRg>2>@#~sq&qB+@CD>E_42}(Dw(Gf->-eu1dqpQfqab|eQZJV$Msjz`
zKcA*EzJ*bDF9#e3z`JGdNxZ`>Py#0Id4LaH3oQ}hkKM;py6$WOXaJ61bk+?adS%@!
zdTuwF`mQ~bSaPY+6fUu^a3)&f7=Dbsj%QuT-b?GzI(g=+d>_!AMW)YmXooFwM}WHv
z9%=q6xra7#h<xYcZuZ_x-g%HWBF8Jyy^u++2X{J(O5aKCmC3xC!@Zh8?s|jbRN>vN
z!2=}gY&SQ!TA&uzh-O@oI{9spshc%5i_IE7Yy&yi=2GoTy7!n3X(0Ugd65#rk9PnS
zYl9U5rH_JjfO(ENw)P0O{Ate(oEtc@f%+X)BdB9Ewk2pvEZ?NIpexH<@=*ODvbXq%
zopxa@IF^L)yY27c4Zx(RFT#P$i^F-|?7#S6=K$vhnho6J8O<8m=XUy`q+j^{%dh=V
z(juwH$awbUU9uugNnJ+J!%K`A^w^z5CRg#Au2HHeiMBM6p<7fEYr;94u)$stmY84*
z&89ZUWhz}@O(eo2wH&dpfZ=Af^Asv7I+Pn<g;D}$uSke=1matb+8U0C0Yj6VL9<Pi
zq@Gc$$Qb)DF)<M@@bE@y39^(Y+}UF9&+*>`L7G`NH(vY;#0%*oon|ioRnM$%P0ZSm
zb9`^S*^;$1Yb_HsQsFtxoQ;%GTI+S}ctl(Z6j0aFW?Mx8nKCYwnxpK`yb!Le7Vjwo
z<wRVfn7qEKayUW)LS1DHF-tt;i$+N+G*>0q+$?HXwlEJ802axVOm3b$!=q*j7nM9I
zqHo|+c^9@5FcD-joc;bR;D~tUBH0a`u|wos7(SE$2>=g4oQZC$hpLX^q`a>tv67o6
z>rqkP$|VVN0ezm-Y3`W7u#9yH;T*`!)H$|h%_faHtp98-o=!>xB^lQ;5r=17FCy|t
z3uEOaJS9U#du^CpmS{;hp~H2)I{_IJ6i0;-#Uk9hHG~EtF#vBT^v?MWWD)s)Td3>0
zK)->lQjRzcVEkdtWnK!BTS+;w`hw7vJtYJhTAH%?>L88=KSAuy^(GK_cvMiMs@CTe
zU6U2@$ek(<TX*~p3pvgvins~2+_3Tw)spAYaBLOiTG07HO<8eOV#Qf9UzDGr?4vZo
zwRS6QIkVPoju(6)Org<jdmfv^d@ahg!}pMV{zkW|Y@H_H6b^7j5Ocz*uqkdJDO|3I
ziSV}Fks&7GDYW?%7Q@$skUPEvQQQi{f&_p~F6zzORRBq>QnDqmCap!xgohj6{6M74
z+@?u2kXFGAOw8>po(tx9<t0XPnK#E$+@teqtkHg4bvyaM#EaP4R3xJCjL9~`>6Z;7
zmor42V=k|WmePV;j5Drn4xj=U&WiY`;3$8L+m#q*c0SdNw<16raji#dmxjth_*G4;
zb1yof66ULwXB$QGtmd5iL2B<tPJ7L8j^b0rD5sgyTp<`<ZYX}q<skcW8Ol|>Mx?PK
zJqD<ggsGXOG`%#QVM#{bV|98yq2r5_>T{5N?U*6wmoAJJWPeZ;MJTT1Il%(6+=H{V
zTLTH9BuY>PGy)6&uM&s|Z~*H~*e@RitKy1?ZpCwyedaEJP6qG=R$##=nI_*r{!O9P
z*$CZ=(i=!7#j*r7ddr!U9M51>q71^LviG9SJ$5{zG70ik#R@AmNVvPhF`umkRyN0m
z1ACC3Rc<H+GM5zFd<%{S6RzQe#vmxM;wvI56_1EL3G+23%&7Qzctv7T<QXx`MU$z@
z$P=6!*Ayy16@Y&U_PjfFj1l3?LKOfN1Rh?MxDiL?@Cc{?;c43OrzFD4SIRfF3^lmR
z)m)976b>S+_=;?+tb9e*L<k}8XpQhu!&z;PbqWBlGy4C&?aSZ4zxG2(tLi1>@^}kl
z0WS48^XCTM=my$ny#-(BfIQ1<8t8@g@R+tQyV%iv8ej_b&^`K?jl6*B(c>NE$C0kD
z42@UgeNcya{#MWRm7O16$Kis~?n%L`7&3;C-@-{8DDPqlZ9f5rXSwBPk1e73XPOW>
zy*(R{Pb1|UxTorMa`K~;{g}2#`v)$nRuAc{%*)6<wJvx2Xc^6#8|aGl>{8i9x}y18
zDEYw8F301mo|H^-Jy>s+vgQPUXWALR)(yPc8#o4g+-z^}R}pvH#odJ3C;GAKZOaq<
zxHoV<av%5oKTrRp4V>R-Pg=CJu}>S!P0!mfZ{!una<iUt%bwWkriI_MjGKyY8v88%
zC^vBNfxAe!xHkQplHR}nMb~~P>Gm0XNGuhZ@Vx9^zgcgksY{pWuryKQqKXOUaKak3
zZRpneZT0K2a3+|T`|Kvpg(E&%dz2q7udh@MBmi&#$3S=S9@*uQ+IKnp2QBguZ66)A
z$9uGW7A{ej`~sLPObFpZ8o$kf!X9QcrII_wsF4j$7G(eqU?`JY0Ut@i*<b-k02s}T
z=<WEB3=@k?VM>s%+x(I(N+q)upsJ(zR*!=@)S6^K4#0>GaX2Yzu%)!r%)lJO?jJ(p
zqdM%>H_WzVb`k<Gd<q}O7iGG#r;P%I->TDzE{7_8nYe?|YxU@ms|k<XSvvNa4(+fc
z2hw}s#7S<_UJe#HRDkfPyyPJTgkRD-^B>~|<Z0onJ~+$svjDaykO1)f&2K6gpWZE^
z;BFoowa0sO3<DEw3bSDbCeS7h7i^sYHF-K7rxGvXr;oC?As!KD7HvQUFuawBj|w1v
z8PI{Q#4xi)o@2xBIF3se2R1IO6dW!{wY|Hch*Ob4|8jaX3sT=1P&Uij<P6TSw_g$o
zNNydyBv>kt@Vv*wsJKO4U55=mfHQ6sv(`l|fT1+xlO9ZhL=X2U6bJO_oQJ|Jnf)SC
z5+x`!3jhPaRuwx%M1TYM_!eEr9h$8KF|2qsS)+-Ik_ey5EkyKt;q!ra>Ke_gjq7$v
zEC8lz23L4#>s(tWVyHUd$!!O?7JQWM{&@~e&{A7)w4ZPdCp5-H34}-G>cJYAST(Lk
zp3>y{@oug>d_GhVczFA!JIddudJ_wdgc1n9X^L9>z!uj?UsF>L^w>y0l=O?<|KejG
zO8V6F8gG5}L*4}`KmPW2ei5QB@52*ZVi~v2bwnJ57urSAAjqg(Hnv;JM<CM*r(C&`
zPvtdKEqpDcVEK9A&@D{kc$<WAJMN;7))+uuWMv9HAP>(VUD6NX>osd{`YMp6#Frh6
z2D}Sy#6yjNprB{mfz&BMJ^7{h5a56Oe3~6*s@w>fT4nS+aVGA|r>nv>vL=VxOA!ye
zPQB{OQTg0D51pAr*6JQ3Dz8xtx$)&mQXMZ6j#tD|LJjz(6!E8`D4#r2Q>%wi#V}k|
zHFK2x)L_)7=tFjOQ+yBYV)un^liqEnIWEyi!580QuV&g{*8)=EXJDi~qtd;wvn0&|
zWk_wrJXV@6(Wf>j!nuZ5GJx6Y`NZ&EbUCO)?Wa*`+cGvlTHg9<!IhRaz9e@`OqJKx
zO)|Sf?$>W2Yiy3i<-(xrCbtn?_yxE;+mV*merl1Q)_p!TO)t~kv=@$&<c_*sTkBSi
z+7S)YrWRgrNoYI;>mfEIrGyDKxIaj{Pr&!2>1&g0^sk*ah%yYF&tJv$=|wR}#ww+r
zzK26QyxWWKUsk)TdKT{dLnpCEjqm)TMmwRM(+~P*`!x})SgXfa_RpaSw|id7PBUGi
zpZUuh*t^obJnP8sG!|gHOO~`$!Yo#Ng&CiIo=r}J(g>ebo*CQw$A*oY#e?yKFds$0
zOM~mp=JzYeWhr`B;*ll`t8PkgtD#UH#u9AG(}L{BCh_$oY22+Bt^E?m>WN6~eXmR{
z{=V`Dv!-)2hnfybXXoR)A&I7L8vHV2Dwsm8zg1>Zy>rOR*{8`{n|50&+guBJnPrFN
zTSDW|%L%XA*&OulXBdLICfZNAalAHJ6;UMQ>FzfLHMLjub2Pm}>r7TR+ax)qD8ml|
zoq~UeGtqX1xXZ3`Oxh`IG_x+qN2=TNoI6&fYtXfLIysk=j4FzJHE~Oc?NQFr$S?5f
z+v+T~nPTrW=MAUTarP^0FgH@;Zfs}4?b=gRm$;__G@If6eiDl_QzzQwtuLSe;M#2$
z$!^0q!1PkFSmw6glpSJqnvEzilZy+Cbf5)^Sy?B1G`Gd?^tZeB6|VsuD?V~3A3Snw
zojrc5q6zue4zyxL=t<_Eht)97vxPzz-ZuWEJ&EZWc`}~YXXOp^$WzH{4t3YI@nLU&
zxyEp^{)qfW>g{IZ9f`cYxJFi8-9e>wk1D=MKoKW2q};<^>#&y*a6~T53*%!Hxp53T
zlpL~T`#?&hY-U7mWO!+EyJ(7(3)l8I3rS3OFMg0l!|Sk2gZZpR1M-nZCHb}(YAMqi
zL^*@V(n}5d(}g~i^qJ2i&GGuCPmj+geZ{d)CsAK{QRhpqJ)EI|X^{ypHElu4!$c}}
z7m~X~hq+Sgpo$6SaKeV(ig1%-a5tALxHGd*;vkOb^U)x0CeCxhB@q#;t}<Kvs#7CT
z>ae~CK#>5z0St}|*<cuhb6xf=e~Z0IZz(Q71mV?DAB{Ipp0zaV0|n4|PBSM+hM2_}
zV$^JjQjlDey~<)8l-YqJ%Jzh)s>geD*&nqus@Zvr1z!NGfC(X7Nc$ir=N1(9Frz7@
zzY0;YlyG@*PzK-t#)p{NH6fCWTQn8H0+0YOni<ijDI|nsm`F0hQDzRZpZFJAq5~}o
z70~__9MMsn1P<5q1wASgf*gR4Pb6K|D{25cbnBQ!<~Z6e83T&7fa<azI-F@!4!Xlt
zH3}hs132o!J3x)13m(t}@WzgC3TOd%>pqOLaFvwOH9YSxkfoO6TY64BACKAD(*Xj7
zNAROQ`wY}i$DCUC=t6gzJCn{4eb;cOi1jF#zrv%`s?8Q@3Q%GPzJs!~yP~6Npl*~#
zxz~p4QBHIF2ISWARqyh0vlqa+f&_r|@oKUSBrRaVCb-q220h-RV+NRLQ<x1iFo71Z
zw;*`oEt-9*N0;JVK%><_Y0r{fIV&Rl0iI#;Gm=L~c$z8G-G_CT3SD71Km@>u4snz~
z`iKuEH5s#ZvN<*!$wR&u?sN_d9WJ;y$c7?LMW+5~MClf!0_UKkbePw)$r+quT%DN7
zt&e$0uv8%7d5?)vaf`aT4jX&`XWS@et&3U!Lutq-J(vWE9_~>nzS37j4vX+bIuo<6
zJ8RS+%S!+Qz+UFFAtJy5>~hhdIZaNwkUNwuL8h<C1Jn@OEDD-=h<-18KJZRmqp>Z)
z7@T6m>>Mj_F|$Z`#kSMbH4tkwcmvJUhSEUWLw4D_h;+icu{1$)+wwNwf}?!GHJs3B
zgAyygBBFxu=rPZfGRwT4+aX)~OM#V{b-hZSU_2qvfpY`?ZWs{9yf`x-Mq^YEcz8f!
zf_LzV0%<656raX=Z6P|&Ynv0yQ^qd=70kdK1JM=sLj5a~zT*8?o*zo`rOf;45*;?z
z)fiPwIENE9gjR&rkHK{yiCn>*nS~MuaYUbw26;1ao)a#Kh*))%+2U878i`Ve^*sQJ
z1ON_TaAe2^!x)_FvTylY>`i)0aRDL-ua5d?ym|7hrCA>+fX;K8IYBbSEY1+4W=oWU
z<eKbN7VDtQ4jfUoCqz{}-lNO@sHIWO&SNb20$2r12;oB72QfLfps<G-O{x0FfGOef
z+tYyx0OLbU?V1ot#x0tPU;#(~7|o36(-aayGE5{H;V3f)*-!inEzyA%g$ii@3XbR~
zP6CH(`hp&n2|*6P$0w35>lHPC9lCYQB6A#Vmy7|$T0nK#4;{|5DF@x*sv3n5zyTa}
z;T@nx(FG4^0(fIbI0du-ymcSOS-46{=^CE*7syh}@hv?ko{z`u?CAi3!Xx-mpM3`E
zr(;g7dvu{Y&7Dc-h`wvMQ^a}{%wOSAYSm_oGzBQJ1K&Ye+Fj96HBdK7qulGk(Ik2}
zGb?RHgu4rGNDXAoCyST}$G0H~OEH#Ms0k=;5>mBAbOB5NBmk`EYO)O^EnvbXxYeQt
zJ>H{Z2AF75m<=;9fflg0Ab8;|ntiHAm*QPOqt!rZ&yrm^D<b^?o?-Dbl1E2)nkm!W
zhjo_<U12yt1i*+6ag;#%hz}+;8MAh>IW`>0L%tX8bPfw0F1R?zh9XWyrv7O}=@z5{
z=b)o>nAfz)8JuHWotVk3k9kS3R3PDbkBL!ni@LfF8+-s~+$d(Pi&_9fX~-u%m;{L)
z?olYd(pN+di||D{6SJ>7Yt$giO8^7FUgonQBESLca?zkUO-{OyJCrR!rmx8Z)DYS%
z3YvL{elL7J@J?N$u`R(EoMOZ594l}!vq*Tww$s!#5NkAe1I^Ti(m>locG<g#bi%u_
zG(mFP@;2XsqkO_OoX}{45-YwUqJr?~G0&7T%e<c3AzS-Pft8tcy-J>7JR#75a|8Zv
z7!b$2I5QtcV^k1$ctB!;ckqb<X((|NPh)whJDQwht3kNs!|rEB4SQ?hicBR_2>CEE
z3FPo-P2rb-RA*q0bLUOp3-v=uU-|y4&JQK|Qs&ili4L3VYK$r-oWltlLMy`R$KX1U
zM6Te@%tDESIHJ!-gS?qI&k2`AM69~XZ1Jm3jYO%#`W^s90ssdvI5K2|VGPc7*|+>H
z_9nfhxBwA^S4Vv`-aL8M(yR{@K<7EloFExu7H5c2vn5JFa!vLsi*-<D2aYJ)6QZgf
z@6ly{)Y7PC=P?$10jvThgm5A4gP5FKP}sward0i7z?5+L?dd=Tfbk)wc1?&R;}%Ut
zumB_gjAlmkX$lD;877j9aFm&Y>?i((mgqo>LIt#c1xIugCxOE?eL;`Pgdhjt;}c1j
z^@<w64&6FtkvWdGOU8g=EugyWhYn}jl!NYYRgFRj-~f)g@D5O;=z<3{0lcvzoB~<^
z-ntLtEL<g}bPdn@3uLL~_?Dg%&&OkS_H=+i;Sv0(&pre7(=n&kJ-X1H=FX&ZMBg>s
zDPlbe=CAN5wQ93PngW#Af$yL!?XKvk8mJqkQSSBNXc9e~nU%I8!rg^8qz1C)lSNE~
z<J*vgr5H;r)C3ea38~s5x&S5s5&+h7HQ5G|7BFEG+-gyS9`DgH15C6j%!V15KnvJg
z5WMgf%|6wmOYttC(Q2TyXUVRd6_Ne`&#?Fz$)h7Y&6Mfx!@5g_t}q-R0$@akI7%RW
z#0Qg_j9EL`92<`0A>RvkI){Z07hD`<LlLJUQ~xxgbPG~}bI?&b%xl`@49+pGPR!)i
z$Gjw1Dv<EJ$Hb_(MO|Ho4L*P~ZWOcDMJ<4#G~|;WOoBuY_b3!!=_?|KMff6}iP_hk
zHENLMC4d27FZ0<D5#RuJxoFUwCMR9U9m<v<)7RtyY6xu>1<gD}zZX6qc&Dz>*p^@n
zPO)Kjjup6=StPt-+iB_=h&39#fo5t$X`t;PyX;*=I^o?|njpDtd7E#+Q9j`sPH412
zi4|WFQ9*e0m}g3vWnRzikgfftz{<?JUL{X3o)GB3xdDGS42WZ1oS6@!F)9c=JRmW_
zJNQI_G?X}sr?I@$9Zk-$)gavRVfQnmhP}0LMW&J|gnXEo1af$^rtnKZsxvUhx$~y)
zh5DhSuX_L0*M2Bzv(K44H*jv?+`zeka|7oF&JDc%4fwB|#!*^7gKz)1ukr+2E8fbz
z{9*LZ(|JfV>A+EH=K(hU)Ls>G1M(&I>*Un7k|Zp=U<_=57eMI<54F?V*Z}|kymjBS
zN?xbjdv!O8&(s|^!0W!F-m8(4{|CH{^T8g|4tUx}Y@mH?TJRCwPt9`!=LTMS1LwE<
zOW%RB|G9y41Lp?L4V)V|H*jv?+`zeka|7oFUf~AxLrGu#{%aojp`@FCOL{@zLi*6&
zMXM;5;E-4*-$x}YV*i?6um&X%j>m!-k{VwZ6fdNqIwg+QB0j=5o8KoHt;5?onX}xv
z(;TKp*8KZOBb6gv3O54t%vvjrXBZxAA*v9<txvK$s5>av@7~4T@!=Apt|4Zf+B*AK
zT8hXBPaU1&4kayMYe83)Fzhx0h<t!*5}*XY7pg3c?FHSoFx4d2w>)YxyoMq=aZ)Fg
zL3k=X?%qih$sqj5pm)%?gT#@E7w(JpKi(Wv0C+j=)c|gQGfT4_RHd{`0xiMT0&0}~
z+PKd+>JC@cD0CNGdH}c%ut`0ju$~bV19+|P=DZO@y}3C|`e?>5yG*kr6j*SsOlWfP
z(~?QqM-|RdM@+sh2k@h=nPZ6eK2~-IHtY-%9o%6t8p;B^!V!fR((KhlMB&wfgeSLf
z5MD@!gDq8*L-^t52Mg}m_n_86`o_VK62j3t!$ym%Q}+2yvf`KSG9n)IU|%GJ@VmXS
z!8u`PSRsk2ZVGd8lWHKXfEk#`t+Ic(z<1qLHqhwRqX^!{8?6RP8eKM@{|a9ukubvB
zMu*6)!Y&ZCIf|#f?&^DxZ{wJ?9&>Csl81aR-02(^I$V$jIb4eiB4h8>sgJoqYQIHu
z+DgbC#ixo<o_`K0f=GI+E#(fsqEUERq(?L3^q;o@csF3;F|$3n-Fk>JKnYgvZUpH7
z9KeU)7LMvnZzH>pcbJF<d*lGzlP>N|W6Mx-ol<Wry#EpIs(4qid}oj_!c(nKj96{S
zS!&cKHz{3ot<3CH0ASn=6`%^h2k<oA2zL|L(pQ7PGoZv#Je?Gk!bQS@Br5KH=2AKX
zTT)$a#S~c+)T&2qh$h=W#x^yjEX*00W7Pq0bO}#Cl=L<4f628UN<z&to_#xG%-~+7
zIXs-mgqQLyNO_n@#qL6Km*_B8Y8_ND;T%rb&|491GQk#_Z3h7$tU-x`IHJ?hD5u7!
zTlp*EWN{{fg=E#w5Z4$mDH(r;r?{)`u+3JJCTtzT;f>l7WGzp)v&G(@v;RULcg|Cg
zp9h$8m(#l#AzPY75MEHpdo<oOu$E?hpa440Y32mUkbSn+m52fb)U~wPjTTH9mrAWs
zE=QRicuyH9C*qRZ&4blkiI9MhIJ@!(vSv^Pq9O%tcOi{b;bfb7DVf~dBg5$U6Gsn1
z24xVQN|B38K_<i5N5d6PngZYePKC)mL{4?dg%ThE;Gu{!somnDs-rjw@2g3y<fg%T
zRMfX}Ny1z}pC@&iJ0>tJV_iZx2Qo8tj;)z*QVGKtm8^Yz7XQ31Q#c_sG*MWFh2T^u
zeXLtUPnBz{;IH1Q$sE)$=8D4bRDJ;j2XIL4p01uYm7x~6$EeU>ht*$>$-l8-J)#^w
znN`|wrogcU!<t+05g@!IGD-ZnvZsVVLs(NbU(t1N9wa5^dJ_mdJSwYE5`sCcSPLuS
zkvml!w(j^H7IK_T6mb)3xnboWswL0=1t2l#%Z&uxHK-{ou1c)9D5tCP6H;81Mz|Jl
zr7dUH`pxkZac@1GaVI>q5&BcDv=x!ci#QP)QUgsRS%{bjZ`&OiViKOhX)MX|Rf%Mh
z>wy)9RaOAv0T>JIDu5(zTCydWrz<gHCOq8mY(J)zG6tD#QVpbR&$u%uO!ku9D-gW!
z7VSROqw{K94A2kpLqoiXtxZKD3eT8qGn{_eP&g7~W=X@ug*tffqP{7DS|7#HC`bqK
z$FN<AVP@@sb8L7k0<;m=wK@kiM)*}ttZ^^eq5`%lR*Q`Pn#8(<4N`kIayp7y>eMr{
z$uCMHpqbKKkq&Q`SVf$%yT~ghu{s~KI?WCnWB_N}C}wRCjc3@>S;Dc6jywyC(WU!R
z#W}S$VwK$UfX}riAU3Y#ab!-m+=R2WT>}ZBBuY>PGy)6&uM&s|Z~(8{-X1SKVyohc
zh*-sQlzrwd!1g!`U4aFkWaj*cXJ~Vrncfx;g$bWR8L9=X79|&H(Xuj+_B+wZJ$ELR
z25%rPO1QhjF`umk)@6<j2lgO6tK3iuWM)N*85<Y(wU%-6-O+@Z3D<B!jl5fCe?yn%
zmeMz0qh&_L&xcaP<)CB4EEi3tG9yoLZd_BS0963~CD`-s)G<bcGYeG!R1kQ0RpLe*
zmBS;T0)(e&Aq6E6PMS40M>SsPG1Nkwgo8=m%I+{_b>u3yMn`K3zXaX53qb||a5tTP
zDCw8H|E2f-P}17%B7*0ROn6DI1yg>X3st&Ahi$h(6%)?kgl!8q*<0%y$(tcK6HLr4
zcN6Er5g)BR%8!=USE>dQ062hSpu2dF?D9zMyBz+57Ws&_kB-{oJ=#7Cm#9mA0ZbMq
zgm58^-{wGJ4>Oul$sJ?V$Ob2iG5`lKl*z4tk0jx2umB_gjAlmkc6>;NiAAO`CCJxp
ze#sW4lGzGS)lqz_$H5$GO)?+{U_^&FoD?<KQd(+eV2)w;4<YeU9ro%QW?M2l2>}>B
zg^%NlGF{ozMuEa_)#*f+LzTWv+`;IzdUVLugh%cy9s5j&c36@F={<1bBsXa<2a6mk
zKzLMM@{j_;FX^56k8uO?wD46QoaOmh0NWEt0C@i9Hx-Oe?-o&THxG^4<2^ctfr&PS
z*)RhWXcLDEw$6Z>JROfyi5KzHN4dS>f#Cra!0=WgJ}Q8GeE6-BIXB0y3K9VJvJb?;
z$HbXqNAGkx%vCp&&f4uHx#ZUJH4#X7mufEOffj)CcZyNW)?S%eGe`MU4QgPb=l5ux
zE=33FsS>(bGW(^VBuY?d761l-ttwWBhyVvLyhVr~1%*+97*@QRtkJ|pNrX@379#q+
z@VE`#32lm*<-0>;IKdSlbERqxu5i)TxwcNkP<6tS+YWFo_$b|L=XtR@v<QS>HRxT%
z&;(Cuf|@@8KYTt^5O{d|=0D2cr+TtJ?hVj~l0Nf!<1_yB`le5h&nEq{W1mjq!gA4t
z+Vd*^A#Y*q4P>f`%|DM0afQuJtSiDri4QV_$WEv5_S|Qwq!KA(jq5I#kHB+rAL`)$
zAtiTTs*RBnR28q3*yXOL?#p-;eQKbM${5KH#pCtGJ~^d@SH+xA;a=_4p>#v9c#T8D
zDx85OBLUZ#<=77<d%1K<X}Hyva%&Z%=|4%zGhnrSEL?v%^Y{*zXCTJT#WFGR$2Z)p
zuiu12(M0`j%v6a-S`<mV)b#ug^%i@bPGt1Yut{aiMx-%$)S9$NPGr$VTiRY^PlSWL
zrs{g+Il08LlLdYSz)d@(`So`5QI3jj^KZOEdzNd@>JeLXcg}c$cBK8pEz^u(q{r8c
zUW!HZ66<n)8qPvgNO^gH0Ao1_SF~h4v^898TXGZ>RakX-?Uxx)d^<5F+a~PIn$V4M
zH<a51yCz0`#yV&tZ0N2HF*41IaLZXiwAKRqwjwTxKJE4RR*&uQSoNpfhFy*L;n>Bz
zMB=ATNiVuVef+OA^_=M>$t@g&7t)IcU=?L1c|)ncLX<k9nP*B2*gKFWx8Gu{vaw~3
z2Lv$~H7^Y$Ze}V_DS3P_56K;;!LOR4o;@VeVeQ6_TP{C;1j{^S6SF)4mLFia?%m4?
z?fr@}P#w39E{A1p-o9T5vIFmL_<=LaJ;*`!eFy1raLgPo;KwKs-o1r8?KL&q()lRa
z->$lt-5m6eeJ&2N5acOe;D-R4TX;yvH!8Y35eyt&AKlw!0|Ykk!kd6qhzLFUrrtWQ
zS;n**4#6rq>Voktx#FfQLYIK6hP<li(j>d{+iRI;+h*@>`W&L(_AocLJ#hH6x8&~V
ze~i6Da5p77+rp@n{M{07f;i5T&SYn0=ye>GMv>nG=ne4qd(LPI-4%q>y?oI95#2|r
zKWOM}S>5zDDP_mu`Qd>}0Ey9uTgmrXdH(n*NG;eEzC9fp5mO+4HY6jr+*uEG=1HFt
z%T~c|mkFF#_N;rQJFw%+C__u=OP3iS-8;YTF4)Yce`V4yd;iOieJJUn57|xPbn8Bl
z*Hex{KVN(x`}#nSVYkQZ(~rrGqx;)=Uib~K{G>SrW0iU?GI2y%af*|xyr!U#7B$v%
ziaBPbV>86%gfp&l$^;`QF_PRx+Np(lT!HT#>#%aC8(8{=aL}E7`#e0+S@1F&i2GWf
zZpYnunfX43u3IbK%Dwz<?05V$37T}^D7Etd+GT^C8;~!tUni#uNs_Sef-$fKUI3*d
zJk(BaV*~utj_#XQ$?MeLuTt5$-i_tmHT5BFx}1d`vG~<ZN!@LMk5N4^o(wQueX8_k
z>KKVf5MIUS5rP-q%t`SsQsS@f?Fy^rhll8v#L(5&7D#t!V|BUEh^+J7iQMU(8+h3b
zq+d3SLGJ&EUZZF=W@)bQvs-qTC>aMZJ}fFgcy8vtlWFuazEhsy_!Lj*b#)6Ka?hlD
zI9E?`yfBTE+~YcTt$6e#8~T)jX_&a{=})ms8z}do1h8@+3znaWW&^U**lc(Wd4i5`
zr_#rTI@EJ#yP_MF)H047a%W;aZgZ#dy*EJjF#ho3-s%t5iwE$*jUFw}cL6ndz$sbs
z0lMf25P#eZ<Dq_>I6RLgp3qr8l=REr|BCZNNtbTN)ALWb|MXMyn|<Sc`VD!RhV~ek
zefkvKIJ&=`=jHrR(vD{(ZtNhBpv0(t7ip&!>Tw0WbF9P4oo-<18^S?%_U-fVL}$Uv
zY#{DyeYzcY=Vj*m7`kq)cq{kvyV2iG@Q`TIfuq#U16=++OKw2E#D1Ne+E$W;g%^y0
zE${*;9pRyNdK(+ypLTTLv`Sv5+<SF5iqF&?H^A$@quxtOm9x+z7QebFsk_bi7}W#g
z$pF*Ur%G?8j*)l-;Z=MdA$Z}<oD}aOCI0H(uCQu;c!+LE3|(z)fpmv9R+kHn$U5Jh
z$erG~ftTGt`enly<o=K7HHubamgWjSyJdHYl5qgz!=eI&=VtCZnMN<;JLL(EPw|9a
zSGV9H_e{EnbM+L*3)48sJ+5=tibqefp-(B8hKakL{uH~kfpQ;804w*gVELJ7HXu8V
z&4$;IC+G-wDt%n2Lp^u4E4oohE#t@`cP7^3Hg_uDdjoV2;}0+Ht^Q!WcmN;V=+W|g
z7f_Q2oRTFUpo@+G@yE?D9_q)5!}Dn337z#rNx$O#*S_ixB|Tvib~B#X%39vvzO_Tp
zy!W5^{)+d%Q&%az?7{pe`sa;tsZ-8pKRr|io3+K~82QG%lg({zk-V4AEi>Fwi{?jo
z>eW+vG?g8452ziW+6bToz*cp&y+#+*BsjS=*%V$o?44w4`#Xsu8HCp%wxfG!Y<Kvu
zvu|gAcv76X=H97VD@vdx*jhl1lAqiuL!rK$<PIllWaJ_rAcQmJ1$=af7Z|0sl)<|3
z44ZAOACVC}p}>MOp+_4Pl|y`7;d3kw)>^1<41gyCi4N|tSOdxeyuuNM7t-vZ#8G^b
z{kTK4yf?KKbO?l_cZP=FK<jF@No#gN7GUkM4?#i*U*DSd^*LErY+$+xIm~`2_b9w)
zfPRP{8sbG9V_ofSOCD-<h<x?HhgL^j;8-wm{cNf7vFfSOUia`ohYQjqhih>`WH$bl
zOpw|ep3`Yc_-r8lBtB^Ji8JTB1Eq8O)Bor;5D&<7beNcUcM6}}Zk@UeP=b}a8$mh%
z2Qd8Jy%&G?d2S`5fv5MeTif*Y(L6k2)~K(26X*wXkb~Ps=dO5-w1t3{s<z}THR_VP
zZ_pxy%8@F<kq(>>2myFK*>2)m`f3n(29!98r<0;mc;jF}5*2qpb19vHqf#qhku^cB
zdenw!vJGVHNo`@yz#OX%fTK%z`k|z+eg7-3{ZP`LW_vheCQOS=cyS0^*s@b$aUr=&
zbXZYa0#q^K98TEKS`lt=3|?l+!axePMk677uRi1<;W!m8y<8hA>>;U&*GMEfY!crN
zkpRE}4DK6bZ_iZrcdZ6$ag*LsT!09|t7AMGFPoFr*3zsG6hP-W&72?^VisqJQL`mV
zL2^y@DvNbcW(STa+Y_Ry9`Dg*f7H^bX6G>$d;zQiCWLSy?Sq({TTs};jHXomW5ASf
z`R(aI1%UA(rglw;B;yuMMX&%Q0E}iv^l1tSAsHr;jBu2hgX|~%g_h_*i$Vpoe+5T$
z6eoehHGM&k%7h>X;NufXm-UJozz*FyW|29LwoAr<VlAM$?1v6#+LVLta8->$2;cyY
zy6_HAqv(PMGy%M^Bb)+S0N%O}<1AbyrF0F?`wL{L<@lDK6VJzEcJ_3DK;aSmsLwtF
z_0ut@);+q=o#xJ@b41@Y+$mx`3g)lyD79*{MVbPX*n#h$EbXr7s2ZpnrBUuGyeVoR
zo@zgin0h!fD{V!Dy9;kf4P?zHi<k(<w;>5jF_u`U2@qFnlTbktA9eu=0PDG$Yy*`U
z(=2y(QRg1-(J_)I#Nn%A39;EVV_~u|0~4vZxO)qN7hVyfpilMaQoIXjv>GT4BH5L*
zBGMn=85Tbyd31!QnKIpdSa+$=6@~*u0F3AmM+u~l_+V0#F>5EAW5ba=<a^;x=djS>
zf{TM}DB@IP>YqlGZb2$=4mwJQc}<&~!8yj&iJ9E`n3n`g1rnb3m>3nesH^L+!3S{0
zjbhfis0A>ThJ4b4Ns#E_9);p7eMRK32w$W#G5fl+Mh&vO1TX;XWj-4s0vy0D7Y&-z
z<fIF^L)j8!`kFjI4WZ4VpqYp0_rm7`@6<IK+Y*eyDK^Z`u>uz}i-cEfJ55~!u||V8
z&`fP84YWOEm%WQfC%hX=6C}4SZ}TlU$|qdI32oaHa4lF7DL{CV+oGTx!jo+6F9lX+
zcA^7#-da?l0#pI`yJ0{a1K~_eTRc<{czC4{20Qpffi#pjil+g+)E!OEvDF~l@?rNg
zqlUe;a7Ct)DTI8Om;`cow5IS&K&mq^$GP*S?}hrIq+j{|S6%y|q|H7rko2RyK<_)Z
ze9BSAZ}Fbe=gzzNh~(_{n0@-ux$%p2zu~y&|GsU<^2&`JL5WfQF49gd)Z-Q5-Ejj;
zZ=epkvu~fT=&a``TN{Y`TAyym-Puw(bGK|@YsFi+TYBE7R{uPmheVSO9Hn*+vGJ$&
zs*oFyFR@=Ir?!<OVc`X1U<<qeN=JC8o!-U<_@^D+H?5M_DfeF8jp8$P#|`kh@2K~Z
zQspf4h{dmNO6qPiK1TJxcrw6r^{LXEsbeG_L3kCPM+jbcGbhEnNQu9?w=1lgA0DDx
z5<^#8TOi$`jn(BsBeKqSCvvBEZs27%kbc=P2D$$udX1vhn5DVG&u-aWqGTMv_^_w|
z;klXnPNvbz_)d9(<5N7L*VQd}$UT$p;aok%@xnAta*ylWwc^o}Z0J)8reWf)r$5Cm
zZJ^wT62QuRELeUfnhnTKW3%Bk<Ow>$ok|}U>QK*}?TT(xQp-4U$eoGxxXqo)_uc^A
z!}!CCd#gWKFCM@LH+r-@-v!j<0jFfi2k4?BK>TqtjEDMh;_y70ctU6WP|~k@|EoWg
z4<%h%>Y1+_c)~+g8=lej2F@F@?q_2sl1hZ+=3jUgICXe5-u(PwEzPEB+)X)lQ#^CJ
zlyeR`_tZMr`U&*ca5|a6CvoQq$3NQQHP3eREFTG9Z&^0tpZD?(PXJmX!tc=Xlz7+%
zUgEAjsppg0JnYtd48neo+*wZDXF9aQ7P%w9-35;{f0f)r8#zRNmB&4YdcX$m_E@Np
zn4(&O2OI%QKDDzp+cJ34H<^j`hIhAvJV5epv)w3k>1dbox3xEQv!-UTS;L2IAONPK
z(|vmzC^xzUPdZELDQ><#X@;}Sxq**%1L+%Z?4U+)DMm93iImty+ElnId8q9n@>S!!
zu_9dvnRYk(L8Nfcg|yS$v%v?hmG_tZ5B%8Y@!seL?(vLfjqLME=tD`L`NMyX|4NZh
zebcAMXOq6}*r$`E^3r;Rb>VtRMDrI^u1=3nRqlV)Sw~$RuBLB~z+hMy6xf(u;at*j
zuEu6@k9TYxP<$5GdX>%?^QyV7EE{(iw!;}%G7`S$xje_>3suDiv@r{jvVay<6Rpjw
z`MuGo-b;PsJl)goQ`lZN7ZfD%di9Z`#H2(OJ-?tLeEgb4qEC1{&~aI>$od!k()J>I
zBJ5hzRFlG+dMBL^%00T2ZT^jSXwPQ-xjW!w-BFf9=wF~6=sr`smKVCwTgF5&)lTgY
zW<p0TXtHm+QoPv`=-(BdcG_u=89Vr9>_%N(T^i#tW2OesJ#>_p8S{BvAKyuD{p&aI
zfq5;7Ll5re(0ntL@%SfTW0V`83aE?@g>WeJ@Wk~S*gsl+aO~D!*>``_!^|run#vBh
zy+gt4c&ZLe5!Cd{h2-90C)Y|`n4(R`ajxz9(7NdfZIAG<0bsXwX^=hUTMPJL51afI
z2(L%>p*B*^Ip!>nNdT+R{oAiS1Q}DH)QRT}#G2&B<=g>3?)u^#P!Vd3luLV5E4GqZ
z<mo<bgPO|aP7mdEr<E$bJ&#|+RYP7?bhoo?-=2B4ZT8)Ux{uK9-<5Ta>kdD!uxifH
zG4@=2OKjh@uJLlC_z11DZ!X9hcTwM;TJ@I$m&bJ*0E5DFcfTwI8YjHDf$jU^*-ecu
z>6M3_qIS6YE(K+TX9e=GZ#?q$(903=So)?gXqZznwzC30E7&{a#d$W4tAAzE*S-HW
z*L)}`I*)f-`@wI#Ji8r$5TnvY<9DvfHG<vCh|O~i13G(=GzipHdCj9Ra?TN^UK!=_
zw0QW!l$Z+nOD=TMD_-eP-X>u*x^y8+u3EwEa3hi*#;{oGxl%4!d((HpkfK3GduwWz
z=BB=s0WZ(E1BeUEx-`0GhUI^BylERtc@L;UoQyqI&KUL9Oz7AkfshviR(vDkBl)?U
zxa!*sl+lT_-C9kPxKwsVAr-OVX*ptqRVgd4T$Sadln{O?P5i!RRiem-Q_E$lmkk$I
z%^YPv^%x2%`jB1S72ij@*nOefB)3^uj&n%&8ou}rdo|MryB3fNKLaD}8I|sZoh4}&
zC_`!^=CRUri9WSK5zaNdk^#(C&nJfWqRT-YYCnxi+m^8b((=|<3$C=Z@g=!iVye8h
zZj#v@a=(5HSz~i7E*Az}H@S`I!Y{z(*^acd_EU@ewC?k%X?mINroC{KBzM&9+FG}A
z)Q)JFHns42OG4u*SP!uwDJ4v>!TmwneFDBGO<$X2qkrwZL6l+WeEuq~PcMo=GFB<|
z^gSHf;oV+*|5S}+X@`f!k5^d7BbD~3^)vGD^f=NjyC^#mtXQkZ3hbXNwsClUwv1~F
z&pNb$={6(U<>zIp<ae40Yr<Xa(|6o1e$(egA^064r$K3izvNlu-DC4Qy*=`SFds$0
zO9#=L&F@!`%Tn~N#6wLMR^61~RzsmYj3wBVrv=%MP2%fG(zshMTKgrA)f17}`(Bw^
z{C(vQW=-d44mBN=&d$eoLlRBhH27u4R4|2Ff3?ijh0`3HAunf>;Yr@wv|CS-)8<;x
z%Pc!A-x3;!UQT$`&gP(ZKf@5zHPL>`jpMb+s)!;XPj|m5sHwfGpQGs=T4%Dl*(S*;
zMHzk&=oI`zoQbw8#9elkW71AxqnUL<K2qJD=iIR>U4yR0)5*D{WK>b)tBG4mY>#q|
zMt;m|Q@3xcv)E>ez0;hVQ*AFo$Jwv2!Q4oVA-1z%$~{GOiF+zQvl;I1C$Ts)b)rq)
z`T`07uHAN#>^6)8OfMCSWp3+D*&$Y^*@zM|xwybc2U?Jrm36{Lb6fmQf4h5M@dWhZ
zBRW=m<W4?#<k&iU{8mL1@~<6e#fs3A%s&sSVVq|Rg)Y2p{7HKf(>3yBJg?8n8|IOx
zlGhySu5IJP-u`lp;bi?0`Hj@u&Bi+td3|w>th%~`O6wj~e35`6PH0HEhrQNeFC*ZH
zT$mTe$0%~+7<MQ*WXbk{lt|gkh}_8V(&Tp06e$<3?Qs^8nC@QuAdQCCVVMT=S&atd
zBaKS(Z86kRrZtFi29c$g8uq7)ekkeJy#M;6A4<{@w4V)BdXIq8d|`(MAz-|`jt&$w
z&CN(D;6QZ7HH4(r;%qr1Y3EyBMzSbdqUP8K4LS2<-b|iLcgJ+U>W|SRD`%fl6w^y)
zAgVDG_L!MUS2oq#D{aD=DwSguFmY?>UDz2HDOCZ_Mr?<a$piQK9U6}L+?B&ot{{8l
z^UEF0)9FTHe3=z7t)i^_@E#gQQ>s;U?IEFoZ0VA}N=;78K1mYBubd>n;>c#y%Q&sk
zFyeXE9TlWJV{}q`B`iSG)ube1j}wvD4!I)@N;m?F?TZ|R7?X;8HD+T88dY2`cTP^l
zI3>zUm@_QZtTR0&E*;ZyCsoii)f2*Kj3XYcOa9?>8D}$~Dy~#rItv4qa2qH?QZ}yg
zqH_JmjHL+IE^GrANso!7A~J@T8AWA;DD2^TnA7T0GUwUXT=u=y02OB=5xI&E%g?44
zOtZ;U%C{N<RgUweSQ2y{UutqYugE#2mYS63bh;<=O_gcteX;V@Dnf=|!*QR062$2B
zFhbIqC>S%@gwABrxNzQ-8Kgc*+I&X!V!WxI`w%6In$M+jLpiTJ*-Tp@U7}v8s;~84
zNnM3&7R<5vU3y*88_OYeLYbc-pVCf~ayGOOgPv5hsuZVyTzauJT*D?|bSvp%Q@Lcb
zMsVG8N2j|o)-p0_larW)<I1#3t!2u|RHEyP%jqHxvm{hda3LsESD$&NV{#g^#_;u(
zzAv&Ovp6T-DxwVh*p6?0dW}Bf_%+A;4iTADGqIer&tY9VBaI*j$cN#{B*nxWE1j1+
z;Rz!a9dLY!9XXCwl}C-{sd5=ebJUA++*4#35>;u#J-Dl^Lg-xKy{MFCqB`zHic+pI
z0hS`NOf*@;sdV`<DnxY2tlTFcab#J`6<68VT<CkG)7IAla9UPr*RdUTc}oTkUcxPZ
z%<AovbKh_e&>$CXE)5HROheHIAQ!6~$j~w_zGUY^Rc7C}Zc>%mvA*8TAW#dD4{XEJ
z!8RBrbx=LkN+Vf7dwi456kS;5GdPqBaf(ZB1Qe%&v9%E@X-y9%7o>_@U3@005;d_J
z#|f~2T)krir3tJI*-RHXma1<GcZ<>#ao-5!F{HChVL%+5NL&;tDWq6)?2>?D?2w<n
zcsWv{SLN#htvr?F@*-Ul@`g})jXdpc#+I0R6|J^8iCyBcl9EV~V*{lNFOf5Z2O;o*
z%9cjbYK?2w1S~e5^pLuNF`PN-J8pck%pRG{r(li^W2;FrOdJ>YhvUAb2_dlivK(oR
z7gL0u*;-1_qGAeC#cAOrz_h4d6hcR^xN;K?+0#%AUAMkMMLaDoaKe&nQ$^j}hca5Q
z)D^&3rE95P-(4220#(>pj<RDHL&<iSRr@gl5YGfSDnVi*oG5c{j+0EVD=8xOBJ|n;
z5vrMn;7oH0#{#oP+H0*&tVLWsg7S8?A*tW7gJBA0U1_7*ILXfCS{*QX7MUfI`DQaV
z`I+K%B0(wt8qV}H09VqP*&$hd5tZCz86eQUOPI=l5QEUfOb$`66-ZQMjJ=8z&wrZO
z9D|DJm5Lb}7y5}N!Ns3Sh&=Mfmr6s}+Kz}vFTIM1q<Ji;W97y?oA-&a$wRXmZUl>2
zYg}?{*nB?aFx$qDN;(A(K*yre^m0X&pmt799^q7tf3P;|i&SRznwcT~C7$JJno&WG
z(n~|TN!1RX%tZy}kp{n5ki3zTGmNXU7?d`IFtuo8J@uK$HKxYaYBj?RwF`Iii-8P`
zfMc1qKs*QMX<Q$fJUs`Plq9K+g+-s^g{}miSZah~O}T3{7Lg%K{mCRj=_!ac)}~D4
zs^CkpG%gdGrq*(gaUqyqYMC`*DKFDSm8nQQ7pqqGP27Y={YV>w9P$O=fI|UGmIb0x
z3gs%*rm|uE?kyG8in@1S)P2SYyZEA_@{$G4T3azUR*jiQuwmB5VzYv$8<%Q~Ls|0a
z6&TuAVlcgnCqd^8ha2(5+Hqf^NRG)gi8Hl>6pgDTsvEcql8OjX$@E3vZ!>mPtX)b6
zS?JVpmpYNGZb)*NAWMbiJz1y~svvcg%Q$V;_}5z2;|<2Nd$sZ`7iug$O32lhkMEcx
z%1iO6F_mGtLM*!uurxvPr5{TA`uD%~*oTsI!qc1^I5%+T4ahg_okdT%a|7oF&JCO!
zI5%)^;M~BufpY`h26p}8Q%TNsr&Bz!a|7oF&JCO!I5%)^;Q9^3PqN}ge7&YuK0wm7
z@8BFbE21J?{$%_PGoBLX2F?wf8)!G+?>_2SY}Shwz%w^EEUL;6X>h}AyGz@_)5+R=
zRo_9W1&@yEKs9=F?>DE<>p6Gy=I+}$@SE9yes`VUWN+q<eTakVhmwBn`(JnEhmz#^
zwWrwZ_?;fR9v|4<-WqN(k+toCZi6~$Fv8fHU}1vu`Q%fDobr4`w|HQH1Cq8RS8|7s
zeQ00y;l`=Gf9NscyrO88GMC&@L__VXRbG&}S_L>M4F0IGX$qUJ>=nmAv%+m=wuD)t
zWS>h6Fl#y%154BDuEo(hKu>)U!x|4NQIu9)iN=+)d?BBJ{EBiB6-8~3OBG_HYMI5t
z1F#aCtmsKLl@VKJQY%p&(S0t<qgUT~^wHJk)q;bEaF^gSjX{a&WSX0YcUtcdO<(4j
zHN3OjtWAsz{Sz;qg2vG=$rK7hJgy|0t5&m7rb`URef1<B>maFaPR2kVfvHqw)l_;0
zi=P8&&PqHC=PH3!J3Q0!ax(5C*4enBK(>QS@KJ_T<dbgi@;YRhV5f3ixNoi1>8!~G
zdP?Urs4pVK31GNfM1AV3#B}saFOqQKj!#MICnVzv=((H~KO<5qDw@4C){>V21fEVs
zv+&Dt0|eFf(-Cw)OZo{NAi|_8;kXmlj>rZEpAbq=HofLJrHU)nR!)!asB}SfW-*SP
zWEG@4K-Q2sV_c0HAez6j?@wKsa1Od*c<LfWbMPo^*(ae~oPBO$fU;H8g;>uAXHsm-
zaOzoKBOaCGRlCcv2Iw#GGvcWY2l0>M8!hgX$`c_!-JncKu-uM2jZ|bt!{`}a;UvaB
zyt0;PyG?qK=BFy9F1fiW^|NmldMag`_86PxMGUH?WlD<dhvY0GDth&q9$hqKkOjUT
z;N$B;<WzY(2mCyO93~Z;GFw^_zbcvqCV%I0rm7{RrVDsaK#m2jc3QcG{wR%!5+Z><
zMx3BB-dNOw25$(~v#@E-We{ycUg~LVu-U;8H-PX=!Z;Bh^M+z4dQ2p<^~PuDBlGb-
zJ>$-S)liMpd9^zSWL_K9&cId!GV*?^D8F^^ds_Ea$8R=TK$ZTYeQ7|JLtdh~!!Uzu
zxrfn>%nZ)BC_BRADTzw|$V(G)f{aRY?#E?5hqNLMrRbO=6cxG>##bH40;{27`VyS3
z<0Um_hLs<YVb+ZoZL+AmHhdZI5^(eYv)*;}N2`wV=P3`xPi=N+vWdJ7x*}Wx;7RVS
zhW4gJZ!z)qL~jWE4_)N_gp-PLPAcHJxr*OTsK}N`*1pK0cO-?}5r{*>NTB^eL+t&b
zy3~?u?C5-Ox)99Q;fxo>{0~(uFQrtRK>MPM6%K3Ns0StFhZ*K7#8rhrh*hyHPekJK
zMOS`>f>fAFw$VJJ?$ew)6<O*_SME<9x#<T%ipo$z78+-?L})T*y=nvIIg2zEaZma<
zu61lu9gPtaRuNNFb(dntrj@`Vq2>b3OG{12-Q~X8o%g6BI-(b9JE{$Nix}~Xl=5m5
zSMG*%5)ol&2bm^aWv~SBbR0{HwMWy`SA4q0He<Kkl@r`q>vUQjLky=3DS?87X}vnu
zR1#Jbs~Cs`ldLd}Z%LDvn6k!TtHCg1A9PW1J{QNtFjPc1rkuR8=i<h!Ry&Y@%9cD~
z^fIfihI;6fWHMhk$K%*oo+G~05NVu?139c6VXq}|ZeU%ckX45)i*pz1L>z%=0uE>r
znZ?Ju=&%ykOS5vdKBcpqSNnLo+F4*v>U^x7du3maE{ODvE1%BP$f`eT;uVuq+E_w~
zT=G<g7zyQ2TbR5oiTUbsMtY1|5|EkP$Xp{-l2M4O>TEO?dL%d|qP=eQxsMEg-IeC%
z<LFUoAVTEDU=j9}DoglAC+oH9^Qsa~oa<g`F76(K=_w|(=4CRvt^E(v7_*Hd_&LL&
zq4l^#5Eh@rjx3BEc?I+Gh3i$%<VL6;O8Ry0fBlsoN@B^}sPpTH9fL7Ui%fV)rUfYv
z6RFr;NbV9H=1Q%DDkhx62^)GV!cC6B-CVBV&dfrIgE*qkM}xeXIL`@}L`1B*%53qg
zPK`vV!}=ZoMFIc^FgP+~gJBHLb=kN4E%qk8rMLhQgjYv>G~PUU*3zsG6hP-W&72?^
zVisqJQL`mVL2^y@DvNbcW(STa+Y_Ry9`Dg*f7H^bX6G>$d;zQiCWLSy?Sq({TTs};
zjHZ<Q;=#01!XULM18@N2Lrm?O5J|=@nu=fnNB|hkjOf!85<)UeBpKl-GY8pE{0lA7
zffj`dX#Wb1=qOGChim$R9+e3}4#3AJk}m5NHGmzub<84j9Br430mWKCb=eOc&a^29
z-QlVlg%H329ChIxphnRJ4`>2-V@Egzv;e$yAI4d@N=oS(p7$5XQp@ozJtv-z$L#Fs
z0D;0I_)(vI2I{9{POW=%p*zi;N#}^ZYq(RydKAoG;ZbVUW{WfhD6s?IL0Q^e(NQ%}
zH%g=2>%q|^dN?yHZAFB;3vWmbWX&gwm<Y$WAqh({mRP6>C~gu`wMBFROaLSRtmkU7
z4J0jK!X~)Yq6R(Qqhkh`Xj7OCGcbV`u(u$1;Vqhdsz;aNT|lGNKxxmCT{$Zv{Q;g~
z@iUS~M|hek)7^)4mkM2BI6wrzhz@a-K>COeCN&wecCtA(9LYnz7w&Wp3mq=FILL+~
zPDQ5vX+-H3qyp!lqjZ?pw8<HqV_cn>$*qrhNw8EP;dzgVQE`j9x(*wB0B77NX03}_
z07GfWCq0-1i5~7zD8ABHL=KDaMLH9+uRCkhAj?Yt1HfM9vmqkD0qk<opgB!Wx{y1R
zEkUNQ$ph37+AIp1d5C^5d_M3_U8Aur!5Exk!|WU@a51w;c*VBU)HM)mG<XBe)P~YP
z+e3EQyNGncyRkGua@+DY--4rj!Zn=GXoC_fz9OQ6@aQqmlrqb_p4%Z?`%8h9nRUHN
zo?tv7(1CLU{%#l$$GkW*A4X$T5O{b%VuE+@i2`XTaTHHud8s>^oMWp&xaGs{XGRTs
zYvGDaB~u9bFfj?_@Mulpmw;4fV2*R=P2UUkLrK5>{Wt9VP}0BmZD0DAf6`}t>i7TT
zf9?PHn?LDqy*ur`J^aQ${HgEx*8l8pe)8JvYrB8{#E<%dU-}dO?B{;xYq9_0&-r6N
z?wh~t|9W@Yeh)wNy}$c+efKy1<xk$a{msAcU-+J1{bk?#$!TA94`2P6pZ<^hsUQDW
zKRNBWhweG}kN?Kk{e^G&`~SOlug(7V@4n?5KKC2`!n@PH%{~0nfAQb?p+E91|JA#D
z+duo?{)RvK+rR7kJ~{2{*u#(hfv^9mzw2}U+9#)dU3<9q_uZ-YnD?`Duea^>pZn_H
z{smv~ng9Kh(_XiSfAGiuvA_2#|K$Jj$!oVC>z^0=z*qg2f9CIe_22#Ew6cdk{focs
zH~!&2`nNtg?YW1y{2YAu=l+ua^k4qrAA0vJ`!m1ir~Q9_&-eZHPu{Bie}3%m{28D9
z`G5A4)4r`e{Iq}jbAS17{g%J@$xGYU_U~oh$~+(1?@#{VXMg(p|KYFy!FQ+qxc2bU
z|DO0q{{!Fttv~Dce(F2#ZU4<b@caMWKlGRWr%#UBXZ^b4>?6O<`zfFO6F&9BKlTUS
zomTd6{(Iq<{rAG>{?GlNpU*fypK`trobLnY`@s1=FuxBR{r$wrzwY;8_ukvz{jTr+
z*MGw2|Lu3rwm<Xt{eds}F+ctfeR6O6#ozy3U-eZ#=Ci)@rR}|bAN!C0nP2l0|BXNJ
zZSNj!ul@6_m)<k#|LA=^TK}(q>3{f7e$D6nm3OB-_mH21xBBPn===1w;9vCHe#=+<
z<nQ|*-ktV!@8S3Sqrc;C{J1awFFtveJ^%dm{B!6>{yF@A@~uDPH~qmM_^a<ud!Idg
zte^Y4_T1n9$v6I?pYq$j<K1b`J)F<M`F(J{51j7<=lj6xc^|mfpCi6Co$vkgf9p?t
z`#<|P-ktW`!%IE~*Z#Wttn2^2l5;O_b1!f0=Z|yV=lu6VzwclD5x@6qzx*$~JMFoL
zt>@sa{P!y#>zx11-~P5g@u&aj|M>2-=N_K%9GpKtKj-J?4|UJq@zegufBBn#;Cntf
z?c3VJTmSE|U;E#C^4913+28PeU-AF`roZ;?L$fda-?N|nU&r4o@TmLnBYyY?|FQ4*
z=0E@LL$l|fE1Z9>u=VE(=kswsALsLN{`u?C_kr_s-mm*}-sk<1`*VIC>-;>{Yx+Fa
z*MIZB{}cbE-|^?(-P@ku@8|pR`QKmN@4r7g|9g`Dzb85Wd#>}p=X&IS&viau=kxV3
zJYT>5PyIXp^?&r!zyIB7&pn*a!TEh~z7JgYK5+he;r#W&$MEZg^ErR)bN->f|DJyj
z*7^5f<-Z5(d>+o{;d~yB|NVgX{`WQ?y53j*;Gg{7f9T78_><FqTzmME|NJ-p-(UZG
z{<n7@iame6Jb%7Cf4)5a^X2({bAI3C@0<Vm7yK)K{L??~Q{Q>oV|#e1KmWP+-<N;t
zy^jBH!L|3{+rRBc{=heX?q`4JY2Vf!&d&#)pAY;HKOcC$552ATp_jg2kN*FTochoC
z|LvT6KKJ~L|F6^eob}JykNgk*`EUKouleimUdsN4Z~Xhd;1~X=A9{D%a}V?!{LuIQ
z?%(y@-}sk5IqiM-aP+?mKKbY0Cvg6K0_Wc+@LJyoe)FIH%uo0;f8}$&^JBOF)bISl
zulb&z`ZwR5_ShaC_4iMmx*zu6FLvJF^Zx$UZ~2R#oX>$g2mj~a`QzXA`Cs$DzB}#z
z-`<&kMRBBWd_aurn#6eGePATUYZWj?v#}73*<=EV8m|P81<?_+F)J8@S@YK@poSda
zcx*Ht;JGTI5*a*FK@_8LqbO1F1XK<Wlw%NG#HBOq^F;s6F+)#xRrl-j_`MH3FkSVl
zx4u5Ay9yk%=mXy8ndP<o{r<Vqvr)TERBdYO^BhY3T`HOFSJ&?v(Tei6E(GuMjHCSX
z`#3tErr-b$@PTsuK&|tLoj&ewnsqRP)>ivwN0i3S{)s@r0X|S~KTzvBN~jms;hO%i
zlt|?Rt*)Q=sNU-xUGqARD<Hh9RDQts*p^=&`3d^@O|wf0d6L7VZMtLg_L(UZ9C-G@
z)?(jlvm2futmu2keUA^Hd!ItB;}F?>ZgJZ;4^geE+UEg!vVN@J@8NZPP;KiM*0w&i
zy+4V+m*ze%YaG=wFBR&cu6f4n)t<Z_&roh<pI0jB>D;I5m-i?X9BShOd*wBJFQ3Q1
z%fmJejvl&~LZ#*K<aM^zerEN(L#>vJ_4_=wmXFux-%;)z_1-pJq~HJ!w)j9R$5W-=
z3wd(CO3g1+-F{gaZ$vrR8vjLmpgmZ69)Wc~@^=4b^3YiwJ1)6Oq0(|V>OExAwzv`s
zm4<`8=E39bJyxa4D^*@&J%6jMd^`UXn-X~A7*(s{xFzjz%UYfo@2J^$hg|PHD0O_O
zZohc$LD}aXz#sh8I8Tt)eEn;;>wXt+6i{ANZR;ggcOM;}E4z*Y>-?=fzh?dS|C+S;
z=Phn|1S&5FwdGZ{)Q9)?{84_s=4+_#x?Ag?Yh^x8)KgpYucE!t-g3@2-08X8Ivb~y
zQP!2$eu3*Jd9R<qTOQt6&l2ld^0=<)lh1AA=WkbS%1Ni-piK^PJ->kc*fu{TxY>tW
z9u3&k_2ZpHD~jiprFmXiDvlox+R-*-ND+a81AKsffIMJZ9_Z@WUiYxo4Z@4E@?6`K
z^K9d1@%hK);-Z!J^((9Al9zOrf5u>=pTSk{r}x)8gd2w$^^QU`G&D3eG}G%sjRt4o
z=Pr5&y@U7I=fjMFpE~Ju3%zn(9RBwO5Z(zqYDrd<cb-}+KP$eeZ8@~oZm2H^fB*=9
zfQ3M9k01DZJ~+bfAOHd&00Mjxz;V{nZr}h80s;5{K9Iu)7RvHP_!_>J!`H0(Oun9`
zu065Zo+uacnKk)L>0Zg2UzO&ewf(c!{wR;df#3iR0s;5{K7bF<50D3t2apGl2YAf`
zmfkCW<dql7jeIXhzL&p8VvitTKLW^`7PjC}I}Y}9D83m4<R^f9U|}v52l-o&?>$Pz
z9p59ZkHHf>1p@E^d;lM4w;x#Cj|c)F00JNY0w4ea>=3~9iKQ*Tp=u87I2I)V0T2KI
z5C8!X009sH0T2KI5C8%83E)11rH#M=90UUJ0eoPK4=mIWK>!3m00ck)1V8`;KmY_l
z00ck)1gau{`z{tb-~bK+0r&tufDh0Ql*$8^-iN4F0!sN#uKjAYT~Tik009sH0T2KI
z`3T5$UL)6j<RgbYfdB}AfI11tm9N#dA9cC`wE_VUkb{8Q@*(PDagEkF$kBJO4-fzW
z5C8!X009dDt)F+G{_uf7K<z%Tv@#+HfB*=9fLaKs{kUbX`l+S;Q4bIR0T57%fW7iF
zzSq*C-~bK+0r&tuV9^JbR?$vGKSV!du^+;h@FjeS^CaW}<N@RX<N<r-0ZZ>f1OX5L
zfyxQkEC0y-y_MGxpMU@efB*<cK|t=jj{R9$4IIEhAOIh*<pWD;5FZhMKiTpp+5_z&
z5P%Qh1NZ>_0C@m;0C_-rdBEa&M5zfNpIV550~;Kq_BSiv2Y<2QFSG;NK_CDhzz6UF
z`T_C)@&NJx@&NJx@&NJx@_<r#z|#8=L4Yj+O7#cqmn|Q#UK+Xl!a4_(4+KB}1VBI=
z1mrsIVLui>fde=Q1myC8g*GAxfB*=9fZPP+%JaPLNAC8({y+c(KmY_l00ck)1V8`;
z_#?pU^$yBy@ftXQgFpa2fDhmU^aJDp<N@RX<N@RX<N+S?fTfiXLEv!$TF+lg-2cbJ
z7;N-2xa$4&6ZH<^#$iUiqYw=Z4UG-W^tw=^!CCmZi{3%+=souNFr#7hLa!6AIqE*n
ziGH=fluw~G<KP~<tYul9{e(3Q>H1^+|D>c5R&?=#XD=n(%Al<1!^7rRip)6_D$aow
z-_*7otoTjq<>(U}e06uvTmrQ(huZcR{Qcv{&F(QPe42TcP$-`q+`OZIz4Yo{Dz!#T
zadIgwD<n`hI7r>T_&#a+4bHE{&#%EnU0iV9^!Ry`wLEy~5A@x`S*EF(L~Ckod#e5S
zGn*I|r|rB!py0r>59GeCW$iw>wdKL}9vjzt*0uvH_Z@LvY1?(>$VophZ%|T5)S_d8
zV;qk;WfQ2jIMi_+{9Ig{gH$O!5o;dd8fPL>YfO5%<HSP+#9IA(lSAH1GYTnG9uARX
za)QR3D<n`0hh6SLsbwAW2^1XQ1M~yr0ptPX0ptPX0XFhLH)E`4aZoOS^2MQjQTOkA
z-Y%hXYf({j(y~K!Jf&8tua}knzVMKaCziG^BKFp;i#uk=7vvLbRcStz?SIhxvF(mc
zUO;%&DF4ut2mK2u6dZo<gEtzzb#DHS%ak`g8yHo-`oRgps<JZfgmP5<+<nC9hQ`e=
z9wu_>!cBkaZ+6S1P;jt^4_Ldu|8C!PJ-2105ni?5$pfaK5or`!GY;0C@8kDvyG{Z(
zaI@w58oq&V;2Z1jOR1&bX!UuUsHAx>w#v<+P;gMv2cwo0ItJFyq)>2x56}--%L7Wk
zr=a%t2T{M;y)XQTpH|Ofd8zlzPwR!bFWE<+;2?(&<bIzO`{Q$7sLI+6<2yAQ-(igt
zYv<3_mIw3g)@;6AaC)2*FZ1T0oP5rA^+M+{Aw%+ip=wjD&&T^@{a4o?`HW|eLIMQ`
z_yGL?dBC<jfb-MG&riXFO&+}5KSnv_nHNTB&h~om$lCe+@cV48XM}cPqg|AKE`Zhj
z-*g$V);#APg@OZofPR2HU|Sx*d&*V6rwl&egLU+<j$n0p0P{*~YhG!4wKr?e`Jg=V
zKA%-xp5guTZOMLhqw3$GP;h_`&<~IYY|8_<KX2Xrd8NnOuzzL7Rn;<o#G`sV`5YmB
ze)dwrtqckd-~b<>AJjAtypuR7u<VqXK&|7T)Hs{Ho>#*6@;MJR^&aQrzH}$$MW4UY
zu--SvN~qQq{f*E5hIXsE-FV4EC?^}o4ZNqp=6f1w7qm-(DWAdzeE1+@ijzxeSs_tM
zAI*yfm1cf9-Z?7eL$&W0&mqbC91?hgH`d>=-}*!J-NRX?shLFU`e<AmYTf!MTECBs
z`eQr+<0%+Vs%;)PT-@Rp-O3CC1qUtq0OLW*j0aV=-aYTz%a{L>YRtMvcu}oAN6*{)
zW30|UgK|axdg;}@6db?-K48BesP#Bk)91ab`#s*r9Z`N|$Dc7?#>aSBuG{ptjWPEq
z6dd4#%KL%VpMO)!xGdHIQg$66t@oqq>W}j#KF*uaPF1%PYxld=dcBW&SvT(ucva1d
z_4_8T-#h0rb6g4GO{=@#rnS#`8+Su_G4D0zy~e!PQs)7+zo&%yDf>Pq_=A5{?*p@b
zzK`<BJC55uW`$2PuM!HCmV;WxPi(CRjzer6hpHQg;PrFd->drm9{7MyS)KhvZ9dS-
zd7{+!A8h?TR-bd{ZCn85<#Rj-?S^*aaU8OyzGXGPS5_`tuS@ZFRnNPF>vn0c+wFDU
zZR>fQ*8Wba`x=$yA#AtL`=3&^8>`=g`|Z-+Z?~2w-V3PNdjWFYXJNHH>~;Lc_u{^+
z8t==(*YLHNujM+9NVOlA_}rDf6LX1L6!V-ZGtZe?xStJ-DqsEJ1W}7({sU#^KVbDd
zALXjq^&mKbQ;|7`*476YXQ+Cd0q>hv{k}Q)fDhJZvETYE=wIkxHR)ftj#B123f%3>
zoz-z;l#7pbS(R$1UG71tWgYVg6dd4#e8T&BA}GIkgln7$?HFqvhu7kr_YV0wn`lMx
zTzA#ab<54?%`5X0TJ7%eBNQCq1M~yr0o(FGZO7T<{(DyL8_Qj9?62x^ajma!9{Dh%
zWsmiDDYRxB^3DwEX-wKdS=09%cAY)i>i}UzUFyzU`iH0j%8HH*I&$UQ_=A*pO$|C9
zlGQYiO09Ko1P*?`OBUr_gAW$B-8Azwg(}M->#J{uw;YvBp!VfpTIk&V-IwwxZ(6*4
zZsykHRLYv_@Q-T(FZ|<^{_`%BP^mRFa!b>{$D~mx8yvVcsfr7~;qvwuBlHId6dcs-
z0~L2+S*ljA)4rH#9j;T>)pN$WtkIuerCL*|?=SHFedXcQc<4|Pfr10iK3KW^`mA64
zOQ~EsKICD*qS6hNS1pge+ig!y8iDe~;gi+gQ7ux=P^ES3*Kr<C-7KO|>p1l9yXJ>s
zsRxK!biuHziEEo5Ay92`@ck&(WnPn0L@71ReKuj-kvmjs9hY*vjqA7B#9D0@pQ1}@
zkWZoVaA<zz!kim#<r1ic!@!l>eNzwrn?k_>K0rS}9zY&I9zY&I9$+I6eAmtIPe!ln
z1j-kO+~(OYH1jB;a%<_Lb0gX|K1HS0wq4t~EFW}-Dy>ot#tzS{7vE}M9<{ehRbE!U
z=h~h5y*wQ*5?=K{)`>QeFJ7ZiaQM9swvP{9J8*su;Z67ZPVRQxxsbA|tmG|}qiO>0
z_@3zJU3{L(rB#<k{hvMva6kIxgft4Z4~Gv%oeP;6dY!PM9w(RXSlTg}@T%bhSEbdR
zc%49N#v!WR8r`JLbv&Wq03V<qAP*o9AP=yU2jYjOP%<EsYE4gj&MR-*;tqv^gOWaQ
zd1g-egYrBI1qUtrASu5?_vf;6DU>Y^Kiwpax+i83rL_O>vbiJ2QUV2ss(cW-EzH5|
z=>iI6lSAv-Q%AQxA4Q<xAcqf1BX<mG*C(4o!NC?Elyx5Ztoy+N3gwGKz*qCrXKqcQ
zysCpwgM<Nb*D2IK9QL$L$Tl1*rfSoV1{|K*aBC@nf`dJL@W!0id~)A6Q?)6M?~fnf
zZw&VO^ZlYj1PTuD0r~;*fNgo8v+F3&j1ffy3Jxs$pmdgL<u^xe5hysY;Dd&7e@u1m
zvyV!xysz(q@~dN=9<{XNTXkpu^nIry0)3oA`HBv&Bp%EpP;h_`&<~IYY|8^!Cj{$+
zsCk_b_#6I~!{4{3IrnzBQ%0cRP#Yf{cAU`RRIAMds!a}t^WC@hDKrwbYD(jlP2=K=
zDHI&w1M~yr0o(FGaf82%@a<GYp*7>sdwJHWhELz5P;GHYF!fmd*~}|MEsFK8*;xM?
z?SOW`dh}_O*M2ay*YSttB#A(+<FI)|-EA9+Z&10l)o6zc(~5MSP;d}@;Ig)TO5?B0
zlovhF_s7*_QW4d<?rM;*Bs?;oLcu{zKIpY{Zqb<8W(oy|n)qNw_MF_y7m^7S9N+`=
z1LOhQ@_=cJtJ|#sH;GcZDXhh%|8v<uq2M6+V0OCGXD2fX2~=7R^-hItiC<hmq2K@?
zpdav_2c}0GhJVyJk!oE{Ee8B_y<0JXTF1f1e`8l4=i3wt4)6i`0rG%tc_8o1pq|F0
z9TW-<TJ(X}+dp4U3%f(0@^W}C{cjQP?7vCXro7El5*3}>bK!sPP_1Zr^xbZIa?%Jd
zdTjTwHM$X}C=?vv1M~yr0o(Gx`ppMS6LPK+wdth^9>inW9a>vWZ1Z4Nr@<Eq6dY{v
z!IzP(*WDlO<OyY)L*&DZmOa+rrBHBy4=V2m=>@OU*`9cXYE}DpZC-zZ`Fp~<#tx>3
z|LFB^3RQ-~S5pTLYtj5Bg^F|d^x?%hBQIoAC^+!!gT0Hxn%q8gm9U~(xt}d{J;%RP
z_w(4ZyQpS0<YB;~(hYRHMs(BSd9zwPC?c%tv<r=j2Ya6-yeT=lIG`Raq)-coxCLi?
zXFY8qP;mIY4|o}eMLGEzH)plIQm4MzcI8Lq1PTt;_yF@^YjZwqsoPy`<KL)H?Vf9B
z7N4R^YLHK%(sJ;d-nh{R8*fpqYLCL#C)dB2O`vRW_{KNgDb=T(s!fNdm+L#fT|lAW
z03T?lA9PRdopWi$O{#WH)~!uwZ#qGh*4E+8pZW8x63Uv!4_x=uF7tT;1qb*5{Q!Bu
zwmfia!F#Dwepo|z(c1@~ir?P-27yY;A$;Jfw7L_o6R0#C5>glSHGiB!rB*!8TC?X_
zrFvg3wspEs({srbDi4QxKe+jZo-8G5)x15+Xw&a*Q>Zc=zW#R9Nv~1a1X?wRPnxx*
zt5WV#duwXY`H-xpd6YH9dIng}peE}Xz$fqtd{X)PFO$5dhF&n|Q(iRssl72ykvE81
zlo#$OC!77`gHh)~W`<rTYSH4|qfc(jzDl9;a(H9TYd*Q}n<<nn4!ECKv-^qQBn>Cy
z{1(RhAD9VLoWsV_rw=&luTdyCzz66D$OE?Jf#U6RGq)zEQr0zoXbL3*GAVDmF6h_S
z`^-B-q2Qn<ADAw`-16?e3ls_tJo&(XZI{TUxj6(14)8$^`L0dy7p9Xtq2Qn?A1oMl
zHF0h8BSb4YxS!LuR|0MkC>tCSySMj<i^!r-aDWfc50D3J%LARi$Ou0gbCvR<QPC6n
zzkINqD6LZ#?a7NsG*c)W98Qm)bt=DEI?<{w`Bz=tZ~u(JMn8kAexiPo-XYvL%&2!1
zVz6<nv5Udq5HPz!CL0<Xn(1|+MuW5PSr@&7ep1B`A1<u;oxmkdI$hvWAuJQZav}Uv
z2;T}}g%DN>;X5I$62fXBtP#RmAw&xyMhNSKuwDoogs@Qvn}o1g2>%ko79ngE!Zsnq
z3Sqkt;)Jk62s?$aO9=5o_+ALRg^(ZwlMoVxutx}cg|JTu`-Sj>5U3D-6he{^4hSJx
z2nU65NC+tvA)>MOl%Sv~A5MPb!--S;{~A1P`uN^Pk&SL*8{Ne=dWdc8BewA+v5kGj
zHoh#jv7a%b$^S2WFVU?PbrSjAib{!Wt*DjA){1J0Y^|u5$kvLAiEOQ?naI|Ps)=kB
zt(%)@-P}a$<|bM<H_^JeiPp_cv~F&qb#oJ~o119e+(hf<E?PHt(Ym>d*3DhCZtkLW
za~G|fyJ+3qMeF7+S~qvmx_OA!%|o<q9-?*g5UrbsXx%(S>*gU^HxJRedAw{4yelM#
zQX$+E!u^U6AmrE?J~P59vS&q7oh2mI89s)8=yitK-eaFPe5QB!?X!U2eqo+rz9G~Q
WW(c2cG%T$6|HXo!8GbWn^!zWIX7f1!

literal 0
HcmV?d00001


From 42815d7683318dc22340e8e83e2da2a7d19f8716 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 12 Sep 2023 18:00:38 +0200
Subject: [PATCH 046/142] match original

---
 icu_benchmarks/run.py | 32 ++++++++++----------------------
 1 file changed, 10 insertions(+), 22 deletions(-)

diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 5b00d31d..a729fb1b 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -5,11 +5,10 @@
 import sys
 from pathlib import Path
 import torch.cuda
-
 from icu_benchmarks.wandb_utils import (
     update_wandb_config,
     apply_wandb_sweep,
-    set_wandb_run_name,
+    set_wandb_experiment_name,
 )
 from icu_benchmarks.tuning.hyperparameters import choose_and_bind_hyperparameters
 from scripts.plotting.utils import plot_aggregated_results
@@ -44,14 +43,19 @@ def main(my_args=tuple(sys.argv[1:])):
     date_format = "%Y-%m-%d %H:%M:%S"
     verbose = args.verbose
     setup_logging(date_format, log_format, verbose)
-
-    # Load weights if in evaluation mode
-    load_weights = args.command == "evaluate"
+    # Get arguments
     data_dir = Path(args.data_dir)
     name = args.name
     task = args.task
     model = args.model
     reproducible = args.reproducible
+    evaluate = args.eval
+    experiment = args.experiment
+    source_dir = args.source_dir
+    # Load task config
+    gin.parse_config_file(f"configs/tasks/{task}.gin")
+    mode = get_mode()
+    # Set arguments for wandb sweep
 
     # Set experiment name
     if name is None:
@@ -73,7 +77,6 @@ def main(my_args=tuple(sys.argv[1:])):
     pretrained_imputation_model = load_pretrained_imputation_model(
         args.pretrained_imputation
     )
-
     # Log imputation model to wandb
     update_wandb_config(
         {
@@ -109,22 +112,7 @@ def main(my_args=tuple(sys.argv[1:])):
         )
 
     if args.preprocessor:
-        # Import custom supplied preprocessor
-        log_full_line(
-            f"Importing custom preprocessor from {args.preprocessor}.", logging.INFO
-        )
-        try:
-            spec = importlib.util.spec_from_file_location(
-                "CustomPreprocessor", args.preprocessor
-            )
-            module = importlib.util.module_from_spec(spec)
-            sys.modules["preprocessor"] = module
-            spec.loader.exec_module(module)
-            gin.bind_parameter("preprocess.preprocessor", module.CustomPreprocessor)
-        except Exception as e:
-            logging.error(
-                f"Could not import custom preprocessor from {args.preprocessor}: {e}"
-            )
+        import_preprocessor(args.preprocessor)
 
     # Load pretrained model in evaluate mode or when finetuning
     if load_weights:

From b07152058b1c8bdb48e0234916659ea6ab92a9e3 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 12 Sep 2023 18:00:52 +0200
Subject: [PATCH 047/142] readded tft dataloader

---
 icu_benchmarks/data/loader.py | 129 ++++++++++++++++++++++++++++++++++
 1 file changed, 129 insertions(+)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 5644ec69..8bf38fef 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -172,6 +172,135 @@ def to_tensor(self):
             return from_numpy(data), from_numpy(labels)
 
 
+@gin.configurable("PredictionDatasetTFT")
+class PredictionDatasetTFT(PredictionDataset):
+    """Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.
+    We also need to feed the model the variables in a specific order
+    Args:
+        ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.
+    """
+
+    def __init__(self, *args, ram_cache: bool = True, **kwargs):
+        super().__init__(*args, ram_cache=True, **kwargs)
+
+    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
+        """Function to sample from the data split of choice. Used for TFT.
+        The data needs to be given to the model in the following order
+        [static categorical,static contious,known catergorical,known continous, observed categorical, observed continous,target ,id]
+        Args:
+            idx: A specific row index to sample.
+        Returns:
+            A sample from the data, consisting of data, labels and padding mask.
+        """
+        if self._cached_dataset is not None:
+            return self._cached_dataset[idx]
+
+        pad_value = 0.0
+        stay_id = self.outcome_df.index.unique()[idx]
+
+        # We need to be sure that tensors are returned in the correct order to be processed correclty by tft
+        tensors = [[] for _ in range(8)]
+        for var in self.features_df.columns:
+            if var == "sex":
+                tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+            elif var == "age" or var == "height" or var == "weight":
+                tensors[1].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+            elif "MissingIndicator" in var:
+                tensors[4].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+            else:
+                tensors[5].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
+
+        tensors[6].extend(
+            self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(
+                dtype=float
+            )
+        )
+        tensors[7].append(np.asarray([stay_id]))
+        window_shape0 = np.shape(tensors[0])[1]
+
+        if len(tensors[6]) == 1:
+            # only one label per stay, align with window
+            tensors[6] = np.concatenate(
+                [np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0
+            )
+
+        length_diff = self.maxlen - window_shape0
+        pad_mask = np.ones(window_shape0)
+        # Padding the array to fulfill size requirement
+
+        if length_diff > 0:
+            # window shorter than the longest window in dataset, pad to same length
+            tensors[0] = np.concatenate(
+                [
+                    tensors[0],
+                    np.ones(
+                        (np.shape(tensors[0])[0], self.maxlen - np.shape(tensors[0])[1])
+                    )
+                    * pad_value,
+                ],
+                axis=1,
+            )
+            tensors[1] = np.concatenate(
+                [
+                    tensors[1],
+                    np.ones(
+                        (np.shape(tensors[1])[0], self.maxlen - np.shape(tensors[1])[1])
+                    )
+                    * pad_value,
+                ],
+                axis=1,
+            )
+            tensors[4] = np.concatenate(
+                [
+                    tensors[4],
+                    np.ones(
+                        (np.shape(tensors[4])[0], self.maxlen - np.shape(tensors[4])[1])
+                    )
+                    * pad_value,
+                ],
+                axis=1,
+            )
+            tensors[5] = np.concatenate(
+                [
+                    tensors[5],
+                    np.ones(
+                        (np.shape(tensors[5])[0], self.maxlen - np.shape(tensors[5])[1])
+                    )
+                    * pad_value,
+                ],
+                axis=1,
+            )
+
+            tensors[6] = np.concatenate(
+                [
+                    tensors[6],
+                    np.ones(self.maxlen - np.shape(tensors[6])[0]) * pad_value,
+                ],
+                axis=0,
+            )
+            pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
+        tensors[7] = np.concatenate(
+            [
+                tensors[7],
+                np.ones(
+                    (np.shape(tensors[7])[0], self.maxlen - np.shape(tensors[7])[1])
+                )
+                * stay_id,
+            ],
+            axis=1,
+        )  # should be done regardless of length_diff
+        not_labeled = np.argwhere(np.isnan(tensors[6]))
+        if len(not_labeled) > 0:
+            tensors[6][not_labeled] = -1
+            pad_mask[not_labeled] = 0
+        tensors[6] = [tensors[6]]
+        pad_mask = pad_mask.astype(bool)
+
+        tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)
+        tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]
+        return OrderedDict(zip(Features.FEAT_NAMES, tensors)), from_numpy(pad_mask)
+
+
 @gin.configurable("ImputationDataset")
 class ImputationDataset(CommonDataset):
     """Subclass of Common Dataset that contains data for imputation models."""

From 778c619025d58d7e487f715b5a6e9bc3ecba5c05 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 12 Sep 2023 18:04:24 +0200
Subject: [PATCH 048/142] readded ordered dict

---
 icu_benchmarks/models/wrappers.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 4b0fb7e9..3426b71b 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -10,7 +10,7 @@
 import torch.nn as nn
 from torch import Tensor, FloatTensor
 from torch.optim import Optimizer, Adam
-
+from collections import OrderedDict
 import inspect
 import gin
 import numpy as np

From 3bcb969104a9f108450ff13faf2ac10eba9600f1 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 12 Sep 2023 18:04:43 +0200
Subject: [PATCH 049/142] merged with dev

---
 icu_benchmarks/models/train.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 2f7401f4..d1b8dfb8 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -190,7 +190,7 @@ def train_common(
                 pin_memory=True,
                 drop_last=True,
             )
-            if model.needs_training
+            if model.requires_backprop
             else DataLoader([test_dataset.to_tensor()], batch_size=1)
         )
 

From 9d4f4a7dffd9a30361bb6b1176f103a95270844a Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Wed, 13 Sep 2023 13:39:35 +0200
Subject: [PATCH 050/142] changed if condition for loading models

---
 icu_benchmarks/models/train.py | 18 ++++++++----------
 1 file changed, 8 insertions(+), 10 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index d1b8dfb8..8e3cbfed 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -126,7 +126,8 @@ def train_common(
         train_dataset
     ), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
-
+    test_dataset = dataset_class(data, split=test_on)
+    test_dataset = assure_minimum_length(test_dataset)
     if not eval_only:
         logging.info(
             f"Training on {train_dataset.name} with {len(train_dataset)} samples and validating on {val_dataset.name} with"
@@ -137,7 +138,7 @@ def train_common(
         train_loader = train_dataset.to_dataloader(
             train=True,
             batch_size=batch_size,
-            num_workers=0,
+            num_workers=num_workers,
             pin_memory=False,
             drop_last=True,
             batch_sampler="synchronized",
@@ -145,7 +146,7 @@ def train_common(
         val_loader = val_dataset.to_dataloader(
             train=False,
             batch_size=batch_size,
-            num_workers=0,
+            num_workers=num_workers,
             pin_memory=False,
             drop_last=True,
             batch_sampler="synchronized",
@@ -153,7 +154,7 @@ def train_common(
         test_loader = test_dataset.to_dataloader(
             train=False,
             batch_size=min(batch_size * 4, len(test_dataset)),
-            num_workers=0,
+            num_workers=num_workers,
             pin_memory=False,
             drop_last=True,
             shuffle=False,
@@ -178,8 +179,6 @@ def train_common(
             pin_memory=True,
             drop_last=True,
         )
-        test_dataset = dataset_class(data, split=test_on)
-        test_dataset = assure_minimum_length(test_dataset)
 
         test_loader = (
             DataLoader(
@@ -199,13 +198,12 @@ def train_common(
 
         else:
             data_shape = next(iter(train_loader))[0].shape
+            model = model(
+                optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
+            )
 
     if load_weights:
         model = load_model(model, source_dir, pl_model=pl_model)
-    else:
-        model = model(
-            optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
-        )
 
     model.set_weight(weight, train_dataset)
 

From a15fd405b886764e2868689567567deec60a72da Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Wed, 13 Sep 2023 13:57:14 +0200
Subject: [PATCH 051/142] reformatted

---
 icu_benchmarks/cross_validation.py    |  6 +--
 icu_benchmarks/data/loader.py         |  3 +-
 icu_benchmarks/data/pooling.py        | 56 +++++++++----------
 icu_benchmarks/imputation/diffwave.py | 71 +++++++-----------------
 icu_benchmarks/models/layers.py       |  2 +-
 icu_benchmarks/models/metrics.py      | 16 ++----
 icu_benchmarks/models/train.py        | 38 ++++---------
 icu_benchmarks/models/wrappers.py     | 78 +++++++--------------------
 icu_benchmarks/run.py                 | 42 +++------------
 9 files changed, 92 insertions(+), 220 deletions(-)

diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index e68610b4..bdcfdd43 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -37,7 +37,7 @@ def execute_repeated_cv(
     cpu: bool = False,
     verbose: bool = False,
     wandb: bool = False,
-    complete_train: bool = False
+    complete_train: bool = False,
 ) -> float:
     """Preprocesses data and trains a model for each fold.
 
@@ -101,7 +101,7 @@ def execute_repeated_cv(
                 fold_index=fold_index,
                 pretrained_imputation_model=pretrained_imputation_model,
                 runmode=mode,
-                complete_train=complete_train
+                complete_train=complete_train,
             )
 
             preprocess_time = datetime.now() - start_time
@@ -118,7 +118,7 @@ def execute_repeated_cv(
                 cpu=cpu,
                 verbose=verbose,
                 use_wandb=wandb,
-                train_only=complete_train
+                train_only=complete_train,
             )
 
             train_time = datetime.now() - start_time
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 8bf38fef..d09ff055 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -186,7 +186,8 @@ def __init__(self, *args, ram_cache: bool = True, **kwargs):
     def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         """Function to sample from the data split of choice. Used for TFT.
         The data needs to be given to the model in the following order
-        [static categorical,static contious,known catergorical,known continous, observed categorical, observed continous,target ,id]
+        [static categorical,static contious,known catergorical,known continous,
+          observed categorical, observed continous,target ,id]
         Args:
             idx: A specific row index to sample.
         Returns:
diff --git a/icu_benchmarks/data/pooling.py b/icu_benchmarks/data/pooling.py
index 7eeb2949..48689d35 100644
--- a/icu_benchmarks/data/pooling.py
+++ b/icu_benchmarks/data/pooling.py
@@ -16,16 +16,17 @@ class PooledDataset:
 
 
 class PooledData:
-    def __init__(self,
-                 data_dir,
-                 vars,
-                 datasets,
-                 file_names,
-                 shuffle=False,
-                 stratify=None,
-                 runmode=RunMode.classification,
-                 save_test=True,
-                 ):
+    def __init__(
+        self,
+        data_dir,
+        vars,
+        datasets,
+        file_names,
+        shuffle=False,
+        stratify=None,
+        runmode=RunMode.classification,
+        save_test=True,
+    ):
         """
         Generate pooled data from existing datasets.
         Args:
@@ -48,10 +49,10 @@ def __init__(self,
         self.save_test = save_test
 
     def generate(
-            self,
-            datasets,
-            samples=10000,
-            seed=42,
+        self,
+        datasets,
+        samples=10000,
+        seed=42,
     ):
         """
         Generate pooled data from existing datasets.
@@ -65,8 +66,8 @@ def generate(
             if folder.is_dir():
                 if folder.name in datasets:
                     data[folder.name] = {
-                        f: pq.read_table(folder / self.file_names[f]).to_pandas(self_destruct=True) for f in
-                        self.file_names.keys()
+                        f: pq.read_table(folder / self.file_names[f]).to_pandas(self_destruct=True)
+                        for f in self.file_names.keys()
                     }
         data = self._pool_datasets(
             datasets=data,
@@ -101,15 +102,15 @@ def _save_pooled_data(self, data_dir, data, datasets, file_names, samples=10000)
         logging.info(f"Saved pooled data at {save_dir}")
 
     def _pool_datasets(
-            self,
-            datasets={},
-            samples=10000,
-            vars=[],
-            seed=42,
-            shuffle=True,
-            runmode=RunMode.classification,
-            data_dir=Path("data"),
-            save_test=True,
+        self,
+        datasets={},
+        samples=10000,
+        vars=[],
+        seed=42,
+        shuffle=True,
+        runmode=RunMode.classification,
+        data_dir=Path("data"),
+        save_test=True,
     ):
         """
         Pool datasets into a single dataset.
@@ -144,8 +145,9 @@ def _pool_datasets(
                 # If we have more outcomes than stays, check max label value per stay id
                 labels = outcome.groupby(id).max()[vars[Var.label]].reset_index(drop=True)
                 # if pd.Series(outcome[id].unique()) is outcome[id]):
-                selected_stays = train_test_split(stays, stratify=labels, shuffle=shuffle, random_state=seed,
-                                                  train_size=samples)
+                selected_stays = train_test_split(
+                    stays, stratify=labels, shuffle=shuffle, random_state=seed, train_size=samples
+                )
             else:
                 selected_stays = train_test_split(stays, shuffle=shuffle, random_state=seed, train_size=samples)
             # Select only stays that are in the selected_stays
diff --git a/icu_benchmarks/imputation/diffwave.py b/icu_benchmarks/imputation/diffwave.py
index 1ba9b705..c2f8ecf1 100644
--- a/icu_benchmarks/imputation/diffwave.py
+++ b/icu_benchmarks/imputation/diffwave.py
@@ -47,9 +47,7 @@ def __init__(
             **kwargs,
         )
 
-        self.init_conv = nn.Sequential(
-            Conv(in_channels, res_channels, kernel_size=1), nn.ReLU()
-        )
+        self.init_conv = nn.Sequential(Conv(in_channels, res_channels, kernel_size=1), nn.ReLU())
 
         self.residual_layer = Residual_group(
             res_channels=res_channels,
@@ -68,15 +66,11 @@ def __init__(
             ZeroConv1d(skip_channels, out_channels),
         )
 
-        self.diffusion_parameters = calc_diffusion_hyperparams(
-            diffusion_time_steps, beta_0, beta_T
-        )
+        self.diffusion_parameters = calc_diffusion_hyperparams(diffusion_time_steps, beta_0, beta_T)
 
     def on_fit_start(self) -> None:
         self.diffusion_parameters = {
-            k: v.to(self.device)
-            for k, v in self.diffusion_parameters.items()
-            if isinstance(v, torch.Tensor)
+            k: v.to(self.device) for k, v in self.diffusion_parameters.items() if isinstance(v, torch.Tensor)
         }
         return super().on_fit_start()
 
@@ -107,15 +101,12 @@ def step_fn(self, batch, step_prefix=""):
             )
 
             B, C, L = amputated_data.shape  # B is batchsize, C=1, L is audio length
-            diffusion_steps = torch.randint(T, size=(B, 1, 1)).to(
-                self.device
-            )  # randomly sample diffusion steps from 1~T
+            diffusion_steps = torch.randint(T, size=(B, 1, 1)).to(self.device)  # randomly sample diffusion steps from 1~T
 
             z = std_normal(amputated_data.shape, self.device)
             z = amputated_data * observed_mask.float() + z * (1 - observed_mask).float()
             transformed_X = (
-                torch.sqrt(Alpha_bar[diffusion_steps]) * amputated_data
-                + torch.sqrt(1 - Alpha_bar[diffusion_steps]) * z
+                torch.sqrt(Alpha_bar[diffusion_steps]) * amputated_data + torch.sqrt(1 - Alpha_bar[diffusion_steps]) * z
             )  # compute x_t from q(x_t|x_0)
             epsilon_theta = self(
                 (
@@ -126,16 +117,12 @@ def step_fn(self, batch, step_prefix=""):
                 )
             )  # predict \epsilon according to \epsilon_\theta
 
-            loss = self.loss(
-                epsilon_theta[amputation_mask.bool()], z[amputation_mask.bool()]
-            )
+            loss = self.loss(epsilon_theta[amputation_mask.bool()], z[amputation_mask.bool()])
         else:
             target = target.permute(0, 2, 1)
             target_missingness = target_missingness.permute(0, 2, 1)
             imputed_data = self.sampling(amputated_data, observed_mask)
-            amputated_data[amputation_mask.bool()] = imputed_data[
-                amputation_mask.bool()
-            ]
+            amputated_data[amputation_mask.bool()] = imputed_data[amputation_mask.bool()]
             amputated_data[target_missingness > 0] = target[target_missingness > 0]
             loss = self.loss(amputated_data, target)
             for metric in self.metrics[step_prefix].values():
@@ -180,9 +167,7 @@ def sampling(self, cond, mask):
 
         for t in range(T - 1, -1, -1):
             x = x * (1 - mask).float() + cond * mask.float()
-            diffusion_steps = (t * torch.ones((B, 1))).to(
-                self.device
-            )  # use the corresponding reverse step
+            diffusion_steps = (t * torch.ones((B, 1))).to(self.device)  # use the corresponding reverse step
             epsilon_theta = self(
                 (
                     x,
@@ -192,13 +177,9 @@ def sampling(self, cond, mask):
                 )
             )  # predict \epsilon according to \epsilon_\theta
             # update x_{t-1} to \mu_\theta(x_t)
-            x = (
-                x - (1 - Alpha[t]) / torch.sqrt(1 - Alpha_bar[t]) * epsilon_theta
-            ) / torch.sqrt(Alpha[t])
+            x = (x - (1 - Alpha[t]) / torch.sqrt(1 - Alpha_bar[t]) * epsilon_theta) / torch.sqrt(Alpha[t])
             if t > 0:
-                x = x + Sigma[t] * std_normal(
-                    cond.shape, self.device
-                )  # add the variance term to x_{t-1}
+                x = x + Sigma[t] * std_normal(cond.shape, self.device)  # add the variance term to x_{t-1}
 
         return x
 
@@ -254,9 +235,7 @@ def calc_diffusion_hyperparams(diffusion_time_steps, beta_0, beta_T):
     Beta_tilde = Beta + 0
     for t in range(1, diffusion_time_steps):
         Alpha_bar[t] *= Alpha_bar[t - 1]  # \bar{\alpha}_t = \prod_{s=1}^t \alpha_s
-        Beta_tilde[t] *= (1 - Alpha_bar[t - 1]) / (
-            1 - Alpha_bar[t]
-        )  # \tilde{\beta}_t = \beta_t * (1-\bar{\alpha}_{t-1})
+        Beta_tilde[t] *= (1 - Alpha_bar[t - 1]) / (1 - Alpha_bar[t])  # \tilde{\beta}_t = \beta_t * (1-\bar{\alpha}_{t-1})
         # / (1-\bar{\alpha}_t)
     Sigma = torch.sqrt(Beta_tilde)  # \sigma_t^2  = \tilde{\beta}_t
 
@@ -319,14 +298,10 @@ def __init__(
         self.fc_t = nn.Linear(diffusion_step_embed_dim_out, self.res_channels)
 
         # dilated conv layer
-        self.dilated_conv_layer = Conv(
-            self.res_channels, 2 * self.res_channels, kernel_size=3, dilation=dilation
-        )
+        self.dilated_conv_layer = Conv(self.res_channels, 2 * self.res_channels, kernel_size=3, dilation=dilation)
 
         # add mel spectrogram upsampler and conditioner conv1x1 layer  (In adapted to S4 output)
-        self.cond_conv = Conv(
-            2 * in_channels, 2 * self.res_channels, kernel_size=1
-        )  # 80 is mel bands
+        self.cond_conv = Conv(2 * in_channels, 2 * self.res_channels, kernel_size=1)  # 80 is mel bands
 
         # residual conv1x1 layer, connect to next residual layer
         self.res_conv = nn.Conv1d(res_channels, res_channels, kernel_size=1)
@@ -355,9 +330,7 @@ def forward(self, input_data):
         cond = self.cond_conv(cond)
         h += cond
 
-        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(
-            h[:, self.res_channels :, :]
-        )
+        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels :, :])
 
         res = self.res_conv(out)
         assert x.shape == res.shape
@@ -382,12 +355,8 @@ def __init__(
         self.num_res_layers = num_res_layers
         self.diffusion_step_embed_dim_in = diffusion_step_embed_dim_in
 
-        self.fc_t1 = nn.Linear(
-            diffusion_step_embed_dim_in, diffusion_step_embed_dim_mid
-        )
-        self.fc_t2 = nn.Linear(
-            diffusion_step_embed_dim_mid, diffusion_step_embed_dim_out
-        )
+        self.fc_t1 = nn.Linear(diffusion_step_embed_dim_in, diffusion_step_embed_dim_mid)
+        self.fc_t2 = nn.Linear(diffusion_step_embed_dim_mid, diffusion_step_embed_dim_out)
 
         self.residual_blocks = nn.ModuleList()
         for n in range(self.num_res_layers):
@@ -416,14 +385,10 @@ def forward(self, input_data):
         h = noise
         skip = 0
         for n in range(self.num_res_layers):
-            h, skip_n = self.residual_blocks[n](
-                (noise, conditional, diffusion_step_embed)
-            )
+            h, skip_n = self.residual_blocks[n]((noise, conditional, diffusion_step_embed))
             skip += skip_n
 
-        return skip * math.sqrt(
-            1.0 / self.num_res_layers
-        )  # normalize for training stability
+        return skip * math.sqrt(1.0 / self.num_res_layers)  # normalize for training stability
 
 
 def std_normal(size, device):
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index bd844d64..c5fd7550 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -743,7 +743,7 @@ def forward(self, x: Tensor) -> Tuple[Tensor, Tensor]:
 
 class TFTBack(nn.Module):
     """
-    Big part of TFT architecture consists of static enrichment followed by mutli-head self-attention then 
+    Big part of TFT architecture consists of static enrichment followed by mutli-head self-attention then
     position wise feed forward followed by a gate and a dense layer
     GRNs-->multi-head attention-->GRNs-->GLU-->Linear-->output
     """
diff --git a/icu_benchmarks/models/metrics.py b/icu_benchmarks/models/metrics.py
index d46a22fe..4c2ffd2b 100644
--- a/icu_benchmarks/models/metrics.py
+++ b/icu_benchmarks/models/metrics.py
@@ -49,9 +49,7 @@ def accuracy(output, target, topk=(1,)):
         return res
 
 
-def balanced_accuracy_compute_fn(
-    y_preds: torch.Tensor, y_targets: torch.Tensor
-) -> float:
+def balanced_accuracy_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor) -> float:
     y_true = y_targets.numpy()
     y_pred = np.argmax(y_preds.numpy(), axis=-1)
     return balanced_accuracy_score(y_true, y_pred)
@@ -63,9 +61,7 @@ def ece_curve_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor) -> floa
     return calibration_curve(y_true, y_pred, n_bins=10)
 
 
-def mae_with_invert_compute_fn(
-    y_preds: torch.Tensor, y_targets: torch.Tensor, invert_fn=Callable
-) -> float:
+def mae_with_invert_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor, invert_fn=Callable) -> float:
     y_true = invert_fn(y_targets.numpy().reshape(-1, 1))[:, 0]
     y_pred = invert_fn(y_preds.numpy().reshape(-1, 1))[:, 0]
     return mean_absolute_error(y_true, y_pred)
@@ -76,9 +72,7 @@ def JSD_fn(y_preds: torch.Tensor, y_targets: torch.Tensor):
 
 
 class BalancedAccuracy(EpochMetric):
-    def __init__(
-        self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False
-    ) -> None:
+    def __init__(self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False) -> None:
         super(BalancedAccuracy, self).__init__(
             balanced_accuracy_compute_fn,
             output_transform=output_transform,
@@ -87,9 +81,7 @@ def __init__(
 
 
 class CalibrationCurve(EpochMetric):
-    def __init__(
-        self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False
-    ) -> None:
+    def __init__(self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False) -> None:
         super(CalibrationCurve, self).__init__(
             ece_curve_compute_fn,
             output_transform=output_transform,
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 8e3cbfed..febfcb59 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -27,9 +27,7 @@
 from collections import OrderedDict
 
 
-cpu_core_count = (
-    len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
-)
+cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
 
 
 def assure_minimum_length(dataset):
@@ -103,28 +101,16 @@ def train_common(
         else (
             PredictionDatasetTFT
             if model.__name__ == "TFT"
-            else (
-                PredictionDatasetTFTpytorch
-                if model.__name__ == "TFTpytorch"
-                else PredictionDataset
-            )
+            else (PredictionDatasetTFTpytorch if model.__name__ == "TFTpytorch" else PredictionDataset)
         )
     )
 
     logging.info(f"Logging to directory: {log_dir}.")
-    save_config_file(
-        log_dir
-    )  # We save the operative config before and also after training
+    save_config_file(log_dir)  # We save the operative config before and also after training
 
-    train_dataset = dataset_class(
-        data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"]
-    )
-    val_dataset = dataset_class(
-        data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"]
-    )
-    train_dataset, val_dataset = assure_minimum_length(
-        train_dataset
-    ), assure_minimum_length(val_dataset)
+    train_dataset = dataset_class(data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"])
+    val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"])
+    train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
     test_dataset = dataset_class(data, split=test_on)
     test_dataset = assure_minimum_length(test_dataset)
@@ -198,9 +184,7 @@ def train_common(
 
         else:
             data_shape = next(iter(train_loader))[0].shape
-            model = model(
-                optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
-            )
+            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
 
     if load_weights:
         model = load_model(model, source_dir, pl_model=pl_model)
@@ -243,9 +227,7 @@ def train_common(
     if not eval_only:
         if model.requires_backprop:
             logging.info("Training DL model.")
-            trainer.fit(
-                model, train_dataloaders=train_loader, val_dataloaders=val_loader
-            )
+            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
             logging.info("Training complete.")
         else:
             logging.info("Training ML model.")
@@ -273,9 +255,7 @@ def train_common(
     )
 
     model.set_weight("balanced", train_dataset)
-    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
-        "test/loss"
-    ]
+    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
     return test_loss
 
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 3426b71b..fdf7bf13 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -23,9 +23,7 @@
 from icu_benchmarks.contants import RunMode
 
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
-gin.config.external_configurable(
-    nn.functional.cross_entropy, module="torch.nn.functional"
-)
+gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.mse_loss, module="torch.nn.functional")
 
 gin.config.external_configurable(mean_squared_error, module="sklearn.metrics")
@@ -90,9 +88,7 @@ def save_model(self, save_path, file_name, file_extension):
 
     def check_supported_runmode(self, runmode: RunMode):
         if runmode not in self._supported_run_modes:
-            raise ValueError(
-                f"Runmode {runmode} not supported for {self.__class__.__name__}"
-            )
+            raise ValueError(f"Runmode {runmode} not supported for {self.__class__.__name__}")
         return True
 
 
@@ -165,9 +161,7 @@ def finalize_step(self, step_prefix=""):
             self.log_dict(
                 {
                     f"{step_prefix}/{name}": (
-                        np.float32(metric.compute())
-                        if isinstance(metric.compute(), np.float64)
-                        else metric.compute()
+                        np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
                     )
                     for name, metric in self.metrics[step_prefix].items()
                     if "_Curve" not in name
@@ -208,10 +202,7 @@ def configure_optimizers(self):
 
     def on_test_epoch_start(self) -> None:
         self.metrics = {
-            step_name: {
-                metric_name: metric()
-                for metric_name, metric in self.set_metrics().items()
-            }
+            step_name: {metric_name: metric() for metric_name, metric in self.set_metrics().items()}
             for step_name in ["train", "val", "test"]
         }
         return super().on_test_epoch_start()
@@ -317,9 +308,7 @@ def step_fn(self, element, step_prefix=""):
             element (object):
             step_prefix (str): Step type, by default: test, train, val.
         """
-        if isinstance(
-            element[0], OrderedDict
-        ):  # check if the data loader is the one for the TFT nvidia implementation
+        if isinstance(element[0], OrderedDict):  # check if the data loader is the one for the TFT nvidia implementation
             data, mask = element[0], element[1].to(self.device)
 
             for key, value in data.items():
@@ -360,9 +349,7 @@ def step_fn(self, element, step_prefix=""):
             else:
                 data = data.float().to(self.device)
         else:
-            raise Exception(
-                "Loader should return either (data, label) or (data, label, mask)"
-            )
+            raise Exception("Loader should return either (data, label) or (data, label, mask)")
 
         out = self(data)
 
@@ -372,31 +359,20 @@ def step_fn(self, element, step_prefix=""):
         else:
             aux_loss = 0
         # Get prediction and target
-        prediction = (
-            torch.masked_select(out, mask.unsqueeze(-1))
-            .reshape(-1, out.shape[-1])
-            .to(self.device)
-        )
+        prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
 
         target = torch.masked_select(labels, mask).to(self.device)
 
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
-            loss = (
-                self.loss(
-                    prediction, target.long(), weight=self.loss_weights.to(self.device)
-                )
-                + aux_loss
-            )
+            loss = self.loss(prediction, target.long(), weight=self.loss_weights.to(self.device)) + aux_loss
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
 
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
-            raise ValueError(
-                f"Run mode {self.run_mode} not yet supported. Please implement it."
-            )
+            raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
         transformed_output = self.output_transform((prediction, target))
 
         for key, value in self.metrics[step_prefix].items():
@@ -413,9 +389,7 @@ def step_fn(self, element, step_prefix=""):
                     value.update(transformed_output[0], transformed_output[1])
             else:
                 value.update(transformed_output)
-        self.log(
-            f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True
-        )
+        self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
         return loss
 
 
@@ -462,15 +436,9 @@ def set_metrics(self, labels):
 
         # Regression
         else:
-            if (
-                self.scaler is not None
-            ):  # We invert transform the labels and predictions if they were scaled.
-                self.output_transform = lambda x: self.scaler.inverse_transform(
-                    x.reshape(-1, 1)
-                )
-                self.label_transform = lambda x: self.scaler.inverse_transform(
-                    x.reshape(-1, 1)
-                )
+            if self.scaler is not None:  # We invert transform the labels and predictions if they were scaled.
+                self.output_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
+                self.label_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
             else:
                 self.output_transform = lambda x: x
                 self.label_transform = lambda x: x
@@ -505,9 +473,7 @@ def fit_model(self, train_data, train_labels, val_data, val_labels):
 
     def validation_step(self, val_dataset, _):
         val_rep, val_label = val_dataset.get_data_and_labels()
-        val_rep, val_label = torch.from_numpy(val_rep).to(
-            self.device
-        ), torch.from_numpy(val_label).to(self.device)
+        val_rep, val_label = torch.from_numpy(val_rep).to(self.device), torch.from_numpy(val_label).to(self.device)
         self.set_metrics(val_label)
 
         val_pred = self.predict(val_rep)
@@ -550,9 +516,7 @@ def log_metrics(self, label, pred, metric_type):
         self.log_dict(
             {
                 # MPS dependent type casting
-                f"{metric_type}/{name}": metric(
-                    self.label_transform(label), self.output_transform(pred)
-                )
+                f"{metric_type}/{name}": metric(self.label_transform(label), self.output_transform(pred))
                 if not self.mps
                 else metric(self.label_transform(label), self.output_transform(pred))
                 # Fore very metric
@@ -590,9 +554,7 @@ def set_model_args(self, model, *args, **kwargs):
         # Get passed keyword arguments
         arguments = locals()["kwargs"]
         # Get valid hyperparameters
-        hyperparams = {
-            key: value for key, value in arguments.items() if key in possible_hps
-        }
+        hyperparams = {key: value for key, value in arguments.items() if key in possible_hps}
         logging.debug(f"Creating model with: {hyperparams}.")
         return model(**hyperparams)
 
@@ -645,9 +607,7 @@ def set_metrics(self):
     def init_weights(self, init_type="normal", gain=0.02):
         def init_func(m):
             classname = m.__class__.__name__
-            if hasattr(m, "weight") and (
-                classname.find("Conv") != -1 or classname.find("Linear") != -1
-            ):
+            if hasattr(m, "weight") and (classname.find("Conv") != -1 or classname.find("Linear") != -1):
                 if init_type == ImputationInit.NORMAL:
                     nn.init.normal_(m.weight.data, 0.0, gain)
                 elif init_type == ImputationInit.XAVIER:
@@ -657,9 +617,7 @@ def init_func(m):
                 elif init_type == ImputationInit.ORTHOGONAL:
                     nn.init.orthogonal_(m.weight.data, gain=gain)
                 else:
-                    raise NotImplementedError(
-                        f"Initialization method {init_type} is not implemented"
-                    )
+                    raise NotImplementedError(f"Initialization method {init_type} is not implemented")
                 if hasattr(m, "bias") and m.bias is not None:
                     nn.init.constant_(m.bias.data, 0.0)
             elif classname.find("BatchNorm2d") != -1:
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index a729fb1b..90af2c27 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -64,19 +64,11 @@ def main(my_args=tuple(sys.argv[1:])):
     logging.info(f"Task mode: {mode}.")
 
     # Set train size to fine tune size if fine tune is set, else use custom train size
-    train_size = (
-        args.fine_tune
-        if args.fine_tune is not None
-        else args.samples
-        if args.samples is not None
-        else None
-    )
+    train_size = args.fine_tune if args.fine_tune is not None else args.samples if args.samples is not None else None
     # Whether to load weights from a previous run
     load_weights = evaluate or args.fine_tune is not None
 
-    pretrained_imputation_model = load_pretrained_imputation_model(
-        args.pretrained_imputation
-    )
+    pretrained_imputation_model = load_pretrained_imputation_model(args.pretrained_imputation)
     # Log imputation model to wandb
     update_wandb_config(
         {
@@ -90,11 +82,7 @@ def main(my_args=tuple(sys.argv[1:])):
     log_dir = (
         (log_dir_name / experiment)
         if experiment
-        else (
-            log_dir_name
-            / (args.task_name if args.task_name is not None else args.task)
-            / model
-        )
+        else (log_dir_name / (args.task_name if args.task_name is not None else args.task) / model)
     )
     log_full_line(f"Logging to {log_dir}.", logging.INFO)
 
@@ -117,9 +105,7 @@ def main(my_args=tuple(sys.argv[1:])):
     # Load pretrained model in evaluate mode or when finetuning
     if load_weights:
         if args.source_dir is None:
-            raise ValueError(
-                "Please specify a source directory when evaluating or fine-tuning."
-            )
+            raise ValueError("Please specify a source directory when evaluating or fine-tuning.")
         log_dir /= f"_from_{args.source_name}"
         name_datasets(args.source_name, args.source_name, args.name)
         if args.fine_tune:
@@ -127,13 +113,9 @@ def main(my_args=tuple(sys.argv[1:])):
             name_datasets(args.name, args.name, args.name)
         run_dir = create_run_dir(log_dir)
         source_dir = args.source_dir
-        logging.info(
-            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config."
-        )
+        logging.info(f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
         gin.parse_config_file(source_dir / "train_config.gin")
-    elif (
-        args.samples and args.source_dir is not None
-    ):  # Train model with limited samples and bind existing config
+    elif args.samples and args.source_dir is not None:  # Train model with limited samples and bind existing config
         logging.info("Binding train gin-config. Note: this might override your config.")
         gin.parse_config_file(args.source_dir / "train_config.gin")
         log_dir /= f"samples_{args.fine_tune}"
@@ -144,22 +126,14 @@ def main(my_args=tuple(sys.argv[1:])):
         name_datasets(args.name, args.name, args.name)
         hp_checkpoint = log_dir / args.hp_checkpoint if args.hp_checkpoint else None
         model_path = (
-            Path("configs")
-            / (
-                "imputation_models"
-                if mode == RunMode.imputation
-                else "prediction_models"
-            )
-            / f"{model}.gin"
+            Path("configs") / ("imputation_models" if mode == RunMode.imputation else "prediction_models") / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
             if args.experiment
             else [model_path, Path(f"configs/tasks/{task}.gin")]
         )
-        gin.parse_config_files_and_bindings(
-            gin_config_files, args.hyperparams, finalize_config=False
-        )
+        gin.parse_config_files_and_bindings(gin_config_files, args.hyperparams, finalize_config=False)
         log_full_line(f"Data directory: {data_dir.resolve()}", level=logging.INFO)
         run_dir = create_run_dir(log_dir)
         choose_and_bind_hyperparameters(

From 54563e4419ca4bf27af74f3c295db5c9b7b98205 Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Mon, 18 Sep 2023 14:41:45 +0200
Subject: [PATCH 052/142] added scheduler removed lr

---
 configs/prediction_models/TFTpytorch.gin | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index b10fd4b5..27e2cf25 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -8,16 +8,16 @@ train_common.model = @TFTpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-#optimizer/hyperparameter.lr = (1e-5, 3e-4)
+
 
 # Encoder params
 model/hyperparameter.class_to_tune = @TFTpytorch
-model/hyperparameter.hidden = (16,32,64)
+model/hyperparameter.hidden = 16
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =(1,2,4)
-model/hyperparameter.lr = (1e-5, 3e-4)
+model/hyperparameter.n_heads =1
+model/hyperparameter.lr_scheduler = "exponential"
 PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
 

From 739920286bf09324e8469fa29d5b4d5a753d5c27 Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Mon, 18 Sep 2023 14:42:03 +0200
Subject: [PATCH 053/142] removed lr as a required input

---
 icu_benchmarks/models/dl_models.py | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index cfc538c8..e46f07b5 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -408,7 +408,7 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         historical_inputs = cat(_historical_inputs, dim=-2)
         future_inputs = Tensor()
         if t_known_inp is not None:
-            future_inputs = t_known_inp[:, self.encoder_length :]
+            future_inputs = t_known_inp[:, self.encoder_length:]
 
         o = self.TFTpart(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
         pred = self.logit(o)
@@ -423,14 +423,14 @@ class TFTpytorch(DLPredictionWrapper):
 
     supported_run_modes = [RunMode.classification, RunMode.regression]
 
-    def __init__(self, dataset, hidden, dropout, n_heads, dropout_att, lr, optimizer, num_classes, *args, **kwargs):
-        super().__init__(lr=lr, optimizer=optimizer, *args, **kwargs)
+    def __init__(self, dataset, hidden, dropout, n_heads, dropout_att, optimizer, num_classes, *args, **kwargs):
+        super().__init__(optimizer=optimizer, *args, **kwargs)
         self.model = TemporalFusionTransformer.from_dataset(
             dataset=dataset,
             hidden_size=hidden,
             dropout=dropout,
             attention_head_size=n_heads,
-            learning_rate=lr,
+
             optimizer=optimizer,
             loss=QuantileLoss(),
             hidden_continuous_size=hidden,

From bf330d4256197fadc8f050aba112e8ed91a4c82c Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Mon, 18 Sep 2023 16:53:02 +0200
Subject: [PATCH 054/142] removed extra test dataloader

---
 icu_benchmarks/models/train.py | 17 +----------------
 1 file changed, 1 insertion(+), 16 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index febfcb59..9b5246e1 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -112,7 +112,7 @@ def train_common(
     val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"])
     train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
-    test_dataset = dataset_class(data, split=test_on)
+    test_dataset = dataset_class(data, split=test_on, name=dataset_names["test"])
     test_dataset = assure_minimum_length(test_dataset)
     if not eval_only:
         logging.info(
@@ -238,21 +238,6 @@ def train_common(
         logging.info("Finished training full model.")
         save_config_file(log_dir)
         return 0
-    test_dataset = dataset_class(data, split=test_on, name=dataset_names["test"])
-    test_dataset = assure_minimum_length(test_dataset)
-    logging.info(f"Testing on {test_dataset.name}  with {len(test_dataset)} samples.")
-    test_loader = (
-        DataLoader(
-            test_dataset,
-            batch_size=min(batch_size * 4, len(test_dataset)),
-            shuffle=False,
-            num_workers=num_workers,
-            pin_memory=True,
-            drop_last=True,
-        )
-        if model.requires_backprop
-        else DataLoader([test_dataset.to_tensor()], batch_size=1)
-    )
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]

From bf475975b24bb5511b9c930462f6a032791036d9 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Mon, 25 Sep 2023 14:01:41 +0200
Subject: [PATCH 055/142] added name to tft loader

---
 icu_benchmarks/data/loader.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index d09ff055..871dec1d 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -465,6 +465,7 @@ def __init__(
         max_encoder_length: int,
         *args,
         ram_cache: bool = False,
+        name: str = "",
         **kwargs,
     ):
         data[split]["FEATURES"]["time_idx"] = (
@@ -475,6 +476,7 @@ def __init__(
         data = data.get(split)  # get split
         labels = data["OUTCOME"]
         features = data["FEATURES"]
+        self.name = name
         self.data = pd.merge(
             labels, features, on=["stay_id", "time"]
         )  # combine labels and features

From 69c0287f2d787ec5a1f098646b462dfc110fea46 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Mon, 25 Sep 2023 14:03:40 +0200
Subject: [PATCH 056/142] reduced cv repetitions to 2

---
 configs/tasks/common/CrossValidation.gin | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/configs/tasks/common/CrossValidation.gin b/configs/tasks/common/CrossValidation.gin
index f3efd9ef..519ad3dd 100644
--- a/configs/tasks/common/CrossValidation.gin
+++ b/configs/tasks/common/CrossValidation.gin
@@ -1,3 +1,3 @@
 # CROSS-VALIDATION SETTINGS
-execute_repeated_cv.cv_repetitions = 5
+execute_repeated_cv.cv_repetitions = 2
 execute_repeated_cv.cv_folds = 5
\ No newline at end of file

From 41967796eabc2c4afb74b3d2abd46e61bc85b880 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 26 Sep 2023 14:07:53 +0200
Subject: [PATCH 057/142] added explain flag

---
 icu_benchmarks/run.py | 44 +++++++++++++++++++++++++++++++++++--------
 1 file changed, 36 insertions(+), 8 deletions(-)

diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 90af2c27..345d7948 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -52,6 +52,7 @@ def main(my_args=tuple(sys.argv[1:])):
     evaluate = args.eval
     experiment = args.experiment
     source_dir = args.source_dir
+    explain = True if args.explain else False
     # Load task config
     gin.parse_config_file(f"configs/tasks/{task}.gin")
     mode = get_mode()
@@ -64,11 +65,19 @@ def main(my_args=tuple(sys.argv[1:])):
     logging.info(f"Task mode: {mode}.")
 
     # Set train size to fine tune size if fine tune is set, else use custom train size
-    train_size = args.fine_tune if args.fine_tune is not None else args.samples if args.samples is not None else None
+    train_size = (
+        args.fine_tune
+        if args.fine_tune is not None
+        else args.samples
+        if args.samples is not None
+        else None
+    )
     # Whether to load weights from a previous run
     load_weights = evaluate or args.fine_tune is not None
 
-    pretrained_imputation_model = load_pretrained_imputation_model(args.pretrained_imputation)
+    pretrained_imputation_model = load_pretrained_imputation_model(
+        args.pretrained_imputation
+    )
     # Log imputation model to wandb
     update_wandb_config(
         {
@@ -82,7 +91,11 @@ def main(my_args=tuple(sys.argv[1:])):
     log_dir = (
         (log_dir_name / experiment)
         if experiment
-        else (log_dir_name / (args.task_name if args.task_name is not None else args.task) / model)
+        else (
+            log_dir_name
+            / (args.task_name if args.task_name is not None else args.task)
+            / model
+        )
     )
     log_full_line(f"Logging to {log_dir}.", logging.INFO)
 
@@ -105,7 +118,9 @@ def main(my_args=tuple(sys.argv[1:])):
     # Load pretrained model in evaluate mode or when finetuning
     if load_weights:
         if args.source_dir is None:
-            raise ValueError("Please specify a source directory when evaluating or fine-tuning.")
+            raise ValueError(
+                "Please specify a source directory when evaluating or fine-tuning."
+            )
         log_dir /= f"_from_{args.source_name}"
         name_datasets(args.source_name, args.source_name, args.name)
         if args.fine_tune:
@@ -113,9 +128,13 @@ def main(my_args=tuple(sys.argv[1:])):
             name_datasets(args.name, args.name, args.name)
         run_dir = create_run_dir(log_dir)
         source_dir = args.source_dir
-        logging.info(f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
+        logging.info(
+            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config."
+        )
         gin.parse_config_file(source_dir / "train_config.gin")
-    elif args.samples and args.source_dir is not None:  # Train model with limited samples and bind existing config
+    elif (
+        args.samples and args.source_dir is not None
+    ):  # Train model with limited samples and bind existing config
         logging.info("Binding train gin-config. Note: this might override your config.")
         gin.parse_config_file(args.source_dir / "train_config.gin")
         log_dir /= f"samples_{args.fine_tune}"
@@ -126,14 +145,22 @@ def main(my_args=tuple(sys.argv[1:])):
         name_datasets(args.name, args.name, args.name)
         hp_checkpoint = log_dir / args.hp_checkpoint if args.hp_checkpoint else None
         model_path = (
-            Path("configs") / ("imputation_models" if mode == RunMode.imputation else "prediction_models") / f"{model}.gin"
+            Path("configs")
+            / (
+                "imputation_models"
+                if mode == RunMode.imputation
+                else "prediction_models"
+            )
+            / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
             if args.experiment
             else [model_path, Path(f"configs/tasks/{task}.gin")]
         )
-        gin.parse_config_files_and_bindings(gin_config_files, args.hyperparams, finalize_config=False)
+        gin.parse_config_files_and_bindings(
+            gin_config_files, args.hyperparams, finalize_config=False
+        )
         log_full_line(f"Data directory: {data_dir.resolve()}", level=logging.INFO)
         run_dir = create_run_dir(log_dir)
         choose_and_bind_hyperparameters(
@@ -177,6 +204,7 @@ def main(my_args=tuple(sys.argv[1:])):
         cpu=args.cpu,
         wandb=args.wandb_sweep,
         complete_train=args.complete_train,
+        explain=explain,
     )
 
     log_full_line("FINISHED TRAINING", level=logging.INFO, char="=", num_newlines=3)

From 0e31937f370140da8b523445a9a0d378ad63d151 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 26 Sep 2023 14:08:08 +0200
Subject: [PATCH 058/142] added explain flag

---
 icu_benchmarks/cross_validation.py | 22 ++++++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)

diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index bdcfdd43..5315e586 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -38,6 +38,7 @@ def execute_repeated_cv(
     verbose: bool = False,
     wandb: bool = False,
     complete_train: bool = False,
+    explain: bool = False,
 ) -> float:
     """Preprocesses data and trains a model for each fold.
 
@@ -80,11 +81,15 @@ def execute_repeated_cv(
         cv_folds_to_train = 1
 
     else:
-        logging.info(f"Starting nested CV with {cv_repetitions_to_train} repetitions of {cv_folds_to_train} folds.")
+        logging.info(
+            f"Starting nested CV with {cv_repetitions_to_train} repetitions of {cv_folds_to_train} folds."
+        )
 
     for repetition in range(cv_repetitions_to_train):
         for fold_index in range(cv_folds_to_train):
-            repetition_fold_dir = log_dir / f"repetition_{repetition}" / f"fold_{fold_index}"
+            repetition_fold_dir = (
+                log_dir / f"repetition_{repetition}" / f"fold_{fold_index}"
+            )
             repetition_fold_dir.mkdir(parents=True, exist_ok=True)
 
             start_time = datetime.now()
@@ -119,6 +124,7 @@ def execute_repeated_cv(
                 verbose=verbose,
                 use_wandb=wandb,
                 train_only=complete_train,
+                explain=explain,
             )
 
             train_time = datetime.now() - start_time
@@ -127,7 +133,10 @@ def execute_repeated_cv(
                 f"FINISHED FOLD {fold_index}| PREPROCESSING DURATION {preprocess_time}| PROCEDURE DURATION {train_time}",
                 level=logging.INFO,
             )
-            durations = {"preprocessing_duration": preprocess_time, "train_duration": train_time}
+            durations = {
+                "preprocessing_duration": preprocess_time,
+                "train_duration": train_time,
+            }
 
             with open(repetition_fold_dir / "durations.json", "w") as f:
                 json.dump(durations, f, cls=JsonResultLoggingEncoder)
@@ -135,6 +144,11 @@ def execute_repeated_cv(
                 wandb_log({"Iteration": repetition * cv_folds_to_train + fold_index})
             if repetition * cv_folds_to_train + fold_index > 1:
                 aggregate_results(log_dir)
-        log_full_line(f"FINISHED CV REPETITION {repetition}", level=logging.INFO, char="=", num_newlines=3)
+        log_full_line(
+            f"FINISHED CV REPETITION {repetition}",
+            level=logging.INFO,
+            char="=",
+            num_newlines=3,
+        )
 
     return agg_loss / (cv_repetitions_to_train * cv_folds_to_train)

From 008aa846beb2943929cd689121c43857b8830166 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 26 Sep 2023 14:08:21 +0200
Subject: [PATCH 059/142] added explain flag

---
 icu_benchmarks/models/train.py | 68 ++++++++++++++++++++++++++--------
 1 file changed, 53 insertions(+), 15 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 9b5246e1..4d5b4d7f 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -1,6 +1,7 @@
 import os
 import gin
 import torch
+import pickle
 import logging
 import pandas as pd
 from joblib import load
@@ -25,9 +26,11 @@
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
+from captum.attr import IntegratedGradients
 
-
-cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
+cpu_core_count = (
+    len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
+)
 
 
 def assure_minimum_length(dataset):
@@ -66,6 +69,7 @@ def train_common(
         torch.cuda.device_count() * 8 * int(torch.cuda.is_available()),
         32,
     ),
+    explain: bool = False,
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -93,6 +97,8 @@ def train_common(
         pl_model: Loading a pytorch lightning model.
         num_workers: Number of workers to use for data loading.
     """
+    with open("data.pkl", "wb") as fp:
+        pickle.dump(data, fp)
     logging.info(f"Training model: {model.__name__}.")
     # choose dataset_class based on the model
     dataset_class = (
@@ -101,16 +107,28 @@ def train_common(
         else (
             PredictionDatasetTFT
             if model.__name__ == "TFT"
-            else (PredictionDatasetTFTpytorch if model.__name__ == "TFTpytorch" else PredictionDataset)
+            else (
+                PredictionDatasetTFTpytorch
+                if (model.__name__ == "TFTpytorch")
+                else PredictionDataset
+            )
         )
     )
 
     logging.info(f"Logging to directory: {log_dir}.")
-    save_config_file(log_dir)  # We save the operative config before and also after training
+    save_config_file(
+        log_dir
+    )  # We save the operative config before and also after training
 
-    train_dataset = dataset_class(data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"])
-    val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"])
-    train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
+    train_dataset = dataset_class(
+        data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"]
+    )
+    val_dataset = dataset_class(
+        data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"]
+    )
+    train_dataset, val_dataset = assure_minimum_length(
+        train_dataset
+    ), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
     test_dataset = dataset_class(data, split=test_on, name=dataset_names["test"])
     test_dataset = assure_minimum_length(test_dataset)
@@ -119,7 +137,9 @@ def train_common(
             f"Training on {train_dataset.name} with {len(train_dataset)} samples and validating on {val_dataset.name} with"
             f" {len(val_dataset)} samples."
         )
+
     logging.info(f"Using {num_workers} workers for data loading.")
+
     if model.__name__ == "TFTpytorch":
         train_loader = train_dataset.to_dataloader(
             train=True,
@@ -127,7 +147,6 @@ def train_common(
             num_workers=num_workers,
             pin_memory=False,
             drop_last=True,
-            batch_sampler="synchronized",
         )
         val_loader = val_dataset.to_dataloader(
             train=False,
@@ -135,7 +154,6 @@ def train_common(
             num_workers=num_workers,
             pin_memory=False,
             drop_last=True,
-            batch_sampler="synchronized",
         )
         test_loader = test_dataset.to_dataloader(
             train=False,
@@ -144,7 +162,6 @@ def train_common(
             pin_memory=False,
             drop_last=True,
             shuffle=False,
-            batch_sampler="synchronized",
         )
         model = model(train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode)
 
@@ -184,11 +201,16 @@ def train_common(
 
         else:
             data_shape = next(iter(train_loader))[0].shape
-            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
+            model = model(
+                optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
+            )
 
     if load_weights:
         model = load_model(model, source_dir, pl_model=pl_model)
-
+    logging.info(
+        f"train_dataloader with {len(train_loader)} samples and validating on loader with"
+        f" {len(val_loader)} samples."
+    )
     model.set_weight(weight, train_dataset)
 
     model.set_trained_columns(train_dataset.get_feature_names())
@@ -223,11 +245,14 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
+        check_val_every_n_epoch=1,
     )
     if not eval_only:
         if model.requires_backprop:
             logging.info("Training DL model.")
-            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
+            trainer.fit(
+                model, train_dataloaders=train_loader, val_dataloaders=val_loader
+            )
             logging.info("Training complete.")
         else:
             logging.info("Training ML model.")
@@ -238,9 +263,22 @@ def train_common(
         logging.info("Finished training full model.")
         save_config_file(log_dir)
         return 0
-
+    if explain:
+        test = next(iter(test_loader))
+        test[0] = test[0].to(model.device)
+        pred = model(test[0])
+        ig = IntegratedGradients(model)
+        # Reformat attributions.
+        test[0].requires_grad_()
+        print(pred.shape)
+        print(test[0].shape)
+        attr, delta = ig.attribute(test[0], target=0, return_convergence_delta=True)
+        attr = attr.detach().numpy()
+        print(attr)
     model.set_weight("balanced", train_dataset)
-    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
+    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
+        "test/loss"
+    ]
     save_config_file(log_dir)
     return test_loss
 

From f9ac9ea40bef0288d94e38ff439b40924a022c0a Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 26 Sep 2023 14:13:14 +0200
Subject: [PATCH 060/142] added explain flag

---
 icu_benchmarks/run_utils.py | 227 +++++++++++++++++++++++++++++-------
 1 file changed, 184 insertions(+), 43 deletions(-)

diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 85b676ac..f6e54e41 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -23,35 +23,134 @@ def build_parser() -> ArgumentParser:
     Returns:
         The configured ArgumentParser.
     """
-    parser = ArgumentParser(description="Framework for benchmarking ML/DL models on ICU data")
+    parser = ArgumentParser(
+        description="Framework for benchmarking ML/DL models on ICU data"
+    )
 
-    parser.add_argument("-d", "--data-dir", required=True, type=Path, help="Path to the parquet data directory.")
-    parser.add_argument("-t", "--task", default="BinaryClassification", required=True, help="Name of the task gin.")
+    parser.add_argument(
+        "-d",
+        "--data-dir",
+        required=True,
+        type=Path,
+        help="Path to the parquet data directory.",
+    )
+    parser.add_argument(
+        "-t",
+        "--task",
+        default="BinaryClassification",
+        required=True,
+        help="Name of the task gin.",
+    )
     parser.add_argument("-n", "--name", help="Name of the (target) dataset.")
-    parser.add_argument("-tn", "--task-name", help="Name of the task, used for naming experiments.")
-    parser.add_argument("-m", "--model", default="LGBMClassifier", help="Name of the model gin.")
+    parser.add_argument(
+        "-tn", "--task-name", help="Name of the task, used for naming experiments."
+    )
+    parser.add_argument(
+        "-m", "--model", default="LGBMClassifier", help="Name of the model gin."
+    )
     parser.add_argument("-e", "--experiment", help="Name of the experiment gin.")
-    parser.add_argument("-l", "--log-dir", default=Path("../yaib_logs/"), type=Path, help="Log directory for model weights.")
-    parser.add_argument("-s", "--seed", default=1234, type=int, help="Random seed for processing, tuning and training.")
-    parser.add_argument("-v", "--verbose", default=False, action=BOA, help="Set to log verbosly. Disable for clean logs.")
+    parser.add_argument(
+        "-l",
+        "--log-dir",
+        default=Path("../yaib_logs/"),
+        type=Path,
+        help="Log directory for model weights.",
+    )
+    parser.add_argument(
+        "-s",
+        "--seed",
+        default=1234,
+        type=int,
+        help="Random seed for processing, tuning and training.",
+    )
+    parser.add_argument(
+        "-v",
+        "--verbose",
+        default=False,
+        action=BOA,
+        help="Set to log verbosly. Disable for clean logs.",
+    )
     parser.add_argument("--cpu", default=False, action=BOA, help="Set to use CPU.")
-    parser.add_argument("-db", "--debug", default=False, action=BOA, help="Set to load less data.")
-    parser.add_argument("--reproducible", default=True, action=BOA, help="Make torch reproducible.")
-    parser.add_argument("-lc", "--load_cache", default=False, action=BOA, help="Set to load generated data cache.")
-    parser.add_argument("-gc", "--generate_cache", default=False, action=BOA, help="Set to generate data cache.")
-    parser.add_argument("-p", "--preprocessor", type=Path, help="Load custom preprocessor from file.")
+    parser.add_argument(
+        "-db", "--debug", default=False, action=BOA, help="Set to load less data."
+    )
+    parser.add_argument(
+        "--reproducible", default=True, action=BOA, help="Make torch reproducible."
+    )
+    parser.add_argument(
+        "-lc",
+        "--load_cache",
+        default=False,
+        action=BOA,
+        help="Set to load generated data cache.",
+    )
+    parser.add_argument(
+        "-gc",
+        "--generate_cache",
+        default=False,
+        action=BOA,
+        help="Set to generate data cache.",
+    )
+    parser.add_argument(
+        "-p", "--preprocessor", type=Path, help="Load custom preprocessor from file."
+    )
     parser.add_argument("-pl", "--plot", action=BOA, help="Generate common plots.")
-    parser.add_argument("-wd", "--wandb-sweep", action="store_true", help="Activates wandb hyper parameter sweep.")
-    parser.add_argument("-imp", "--pretrained-imputation", type=str, help="Path to pretrained imputation model.")
-    parser.add_argument("-hp", "--hyperparams", nargs="+", help="Hyperparameters for model.")
-    parser.add_argument("--tune", default=False, action=BOA, help="Find best hyperparameters.")
-    parser.add_argument("--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint.")
-    parser.add_argument("--eval", default=False, action=BOA, help="Only evaluate model, skip training.")
-    parser.add_argument("--complete-train", default=False, action=BOA, help="Use all data to train model, skip testing.")
-    parser.add_argument("-ft", "--fine-tune", default=None, type=int, help="Finetune model with amount of train data.")
-    parser.add_argument("-sn", "--source-name", type=Path, help="Name of the source dataset.")
-    parser.add_argument("--source-dir", type=Path, help="Directory containing gin and model weights.")
-    parser.add_argument("-sa", "--samples", type=int, default=None, help="Number of samples to use for evaluation.")
+    parser.add_argument(
+        "-wd",
+        "--wandb-sweep",
+        action="store_true",
+        help="Activates wandb hyper parameter sweep.",
+    )
+    parser.add_argument(
+        "-imp",
+        "--pretrained-imputation",
+        type=str,
+        help="Path to pretrained imputation model.",
+    )
+    parser.add_argument(
+        "-hp", "--hyperparams", nargs="+", help="Hyperparameters for model."
+    )
+    parser.add_argument(
+        "--tune", default=False, action=BOA, help="Find best hyperparameters."
+    )
+    parser.add_argument(
+        "--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint."
+    )
+    parser.add_argument(
+        "--eval", default=False, action=BOA, help="Only evaluate model, skip training."
+    )
+    parser.add_argument(
+        "--complete-train",
+        default=False,
+        action=BOA,
+        help="Use all data to train model, skip testing.",
+    )
+    parser.add_argument(
+        "-ft",
+        "--fine-tune",
+        default=None,
+        type=int,
+        help="Finetune model with amount of train data.",
+    )
+    parser.add_argument(
+        "-sn", "--source-name", type=Path, help="Name of the source dataset."
+    )
+    parser.add_argument(
+        "--source-dir", type=Path, help="Directory containing gin and model weights."
+    )
+    parser.add_argument(
+        "-sa",
+        "--samples",
+        type=int,
+        default=None,
+        help="Number of samples to use for evaluation.",
+    )
+    parser.add_argument(
+        "--explain",
+        default=False,
+        action=BOA,
+        help="Provide explaintations for predictions.",
+    )
     return parser
 
 
@@ -80,15 +179,21 @@ def create_run_dir(log_dir: Path, randomly_searched_params: str = None) -> Path:
 
 def import_preprocessor(preprocessor_path: str):
     # Import custom supplied preprocessor
-    log_full_line(f"Importing custom preprocessor from {preprocessor_path}.", logging.INFO)
+    log_full_line(
+        f"Importing custom preprocessor from {preprocessor_path}.", logging.INFO
+    )
     try:
-        spec = importlib.util.spec_from_file_location("CustomPreprocessor", preprocessor_path)
+        spec = importlib.util.spec_from_file_location(
+            "CustomPreprocessor", preprocessor_path
+        )
         module = importlib.util.module_from_spec(spec)
         sys.modules["preprocessor"] = module
         spec.loader.exec_module(module)
         gin.bind_parameter("preprocess.preprocessor", module.CustomPreprocessor)
     except Exception as e:
-        logging.error(f"Could not import custom preprocessor from {preprocessor_path}: {e}")
+        logging.error(
+            f"Could not import custom preprocessor from {preprocessor_path}: {e}"
+        )
 
 
 def aggregate_results(log_dir: Path, execution_time: timedelta = None):
@@ -132,10 +237,14 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
     # Calculate the population standard deviation over aggregated results over folds/iterations
     # Divide by sqrt(n) to get standard deviation.
-    std_scores = {metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()}
+    std_scores = {
+        metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()
+    }
 
     confidence_interval = {
-        metric: (stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list)))
+        metric: (
+            stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list))
+        )
         for metric, list in list_scores.items()
     }
 
@@ -143,7 +252,9 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
         "avg": averaged_scores,
         "std": std_scores,
         "CI_0.95": confidence_interval,
-        "execution_time": execution_time.total_seconds() if execution_time is not None else 0.0,
+        "execution_time": execution_time.total_seconds()
+        if execution_time is not None
+        else 0.0,
     }
 
     with open(log_dir / "aggregated_test_metrics.json", "w") as f:
@@ -159,10 +270,14 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
 def name_datasets(train="default", val="default", test="default"):
     """Names the datasets for logging (optional)."""
-    gin.bind_parameter("train_common.dataset_names", {"train": train, "val": val, "test": test})
+    gin.bind_parameter(
+        "train_common.dataset_names", {"train": train, "val": val, "test": test}
+    )
 
 
-def log_full_line(msg: str, level: int = logging.INFO, char: str = "-", num_newlines: int = 0):
+def log_full_line(
+    msg: str, level: int = logging.INFO, char: str = "-", num_newlines: int = 0
+):
     """Logs a full line of a given character with a message centered.
 
     Args:
@@ -175,7 +290,12 @@ def log_full_line(msg: str, level: int = logging.INFO, char: str = "-", num_newl
     reserved_chars = len(logging.getLevelName(level)) + 28
     logging.log(
         level,
-        "{0:{char}^{width}}{1}".format(msg, "\n" * num_newlines, char=char, width=terminal_size.columns - reserved_chars),
+        "{0:{char}^{width}}{1}".format(
+            msg,
+            "\n" * num_newlines,
+            char=char,
+            width=terminal_size.columns - reserved_chars,
+        ),
     )
 
 
@@ -185,24 +305,43 @@ def load_pretrained_imputation_model(use_pretrained_imputation):
     Args:
         use_pretrained_imputation: Path to the pretrained imputation model.
     """
-    if use_pretrained_imputation is not None and not Path(use_pretrained_imputation).exists():
+    if (
+        use_pretrained_imputation is not None
+        and not Path(use_pretrained_imputation).exists()
+    ):
         logging.warning("The specified pretrained imputation model does not exist.")
         use_pretrained_imputation = None
 
     if use_pretrained_imputation is not None:
-        logging.info("Using pretrained imputation from" + str(use_pretrained_imputation))
-        pretrained_imputation_model_checkpoint = torch.load(use_pretrained_imputation, map_location=torch.device("cpu"))
+        logging.info(
+            "Using pretrained imputation from" + str(use_pretrained_imputation)
+        )
+        pretrained_imputation_model_checkpoint = torch.load(
+            use_pretrained_imputation, map_location=torch.device("cpu")
+        )
         if isinstance(pretrained_imputation_model_checkpoint, dict):
             imputation_model_class = pretrained_imputation_model_checkpoint["class"]
-            pretrained_imputation_model = imputation_model_class(**pretrained_imputation_model_checkpoint["hyper_parameters"])
-            pretrained_imputation_model.set_trained_columns(pretrained_imputation_model_checkpoint["trained_columns"])
-            pretrained_imputation_model.load_state_dict(pretrained_imputation_model_checkpoint["state_dict"])
+            pretrained_imputation_model = imputation_model_class(
+                **pretrained_imputation_model_checkpoint["hyper_parameters"]
+            )
+            pretrained_imputation_model.set_trained_columns(
+                pretrained_imputation_model_checkpoint["trained_columns"]
+            )
+            pretrained_imputation_model.load_state_dict(
+                pretrained_imputation_model_checkpoint["state_dict"]
+            )
         else:
             pretrained_imputation_model = pretrained_imputation_model_checkpoint
-        pretrained_imputation_model = pretrained_imputation_model.to("cuda" if torch.cuda.is_available() else "cpu")
+        pretrained_imputation_model = pretrained_imputation_model.to(
+            "cuda" if torch.cuda.is_available() else "cpu"
+        )
         try:
-            logging.info(f"imputation model device: {next(pretrained_imputation_model.parameters()).device}")
-            pretrained_imputation_model.device = next(pretrained_imputation_model.parameters()).device
+            logging.info(
+                f"imputation model device: {next(pretrained_imputation_model.parameters()).device}"
+            )
+            pretrained_imputation_model.device = next(
+                pretrained_imputation_model.parameters()
+            ).device
         except Exception as e:
             logging.debug(f"Could not set device of imputation model: {e}")
     else:
@@ -223,7 +362,9 @@ def setup_logging(date_format, log_format, verbose):
     logging.basicConfig(format=log_format, datefmt=date_format)
     loggers = ["pytorch_lightning", "lightning_fabric"]
     for logger in loggers:
-        logging.getLogger(logger).handlers[0].setFormatter(logging.Formatter(log_format, datefmt=date_format))
+        logging.getLogger(logger).handlers[0].setFormatter(
+            logging.Formatter(log_format, datefmt=date_format)
+        )
 
     if not verbose:
         logging.getLogger().setLevel(logging.INFO)

From 3bf60c700f8ca8492099234630e8e1860767bbdc Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 26 Sep 2023 15:44:39 +0200
Subject: [PATCH 061/142] cleaned conditons and added flag for pytorch
 forecasting

---
 icu_benchmarks/cross_validation.py |  2 +
 icu_benchmarks/data/loader.py      | 24 ++++----
 icu_benchmarks/models/dl_models.py | 96 ++++++++++++++++++++++++------
 icu_benchmarks/models/wrappers.py  | 86 ++++++++++++++++++++------
 icu_benchmarks/run.py              |  4 +-
 icu_benchmarks/run_utils.py        |  6 ++
 6 files changed, 170 insertions(+), 48 deletions(-)

diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 5315e586..bcf2f5f5 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -39,6 +39,7 @@ def execute_repeated_cv(
     wandb: bool = False,
     complete_train: bool = False,
     explain: bool = False,
+    pytorch_forecasting: bool = False,
 ) -> float:
     """Preprocesses data and trains a model for each fold.
 
@@ -125,6 +126,7 @@ def execute_repeated_cv(
                 use_wandb=wandb,
                 train_only=complete_train,
                 explain=explain,
+                pytorch_forecasting=pytorch_forecasting,
             )
 
             train_time = datetime.now() - start_time
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 871dec1d..5dde7809 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -172,19 +172,20 @@ def to_tensor(self):
             return from_numpy(data), from_numpy(labels)
 
 
+"""
 @gin.configurable("PredictionDatasetTFT")
 class PredictionDatasetTFT(PredictionDataset):
-    """Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.
+    Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.
     We also need to feed the model the variables in a specific order
     Args:
         ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.
-    """
+    
 
     def __init__(self, *args, ram_cache: bool = True, **kwargs):
         super().__init__(*args, ram_cache=True, **kwargs)
 
     def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
-        """Function to sample from the data split of choice. Used for TFT.
+        Function to sample from the data split of choice. Used for TFT.
         The data needs to be given to the model in the following order
         [static categorical,static contious,known catergorical,known continous,
           observed categorical, observed continous,target ,id]
@@ -192,7 +193,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             idx: A specific row index to sample.
         Returns:
             A sample from the data, consisting of data, labels and padding mask.
-        """
+        
         if self._cached_dataset is not None:
             return self._cached_dataset[idx]
 
@@ -300,6 +301,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)
         tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]
         return OrderedDict(zip(Features.FEAT_NAMES, tensors)), from_numpy(pad_mask)
+"""
 
 
 @gin.configurable("ImputationDataset")
@@ -461,11 +463,11 @@ def __init__(
         self,
         data: dict,
         split: str,
-        max_prediction_length: int,
-        max_encoder_length: int,
         *args,
         ram_cache: bool = False,
         name: str = "",
+        max_prediction_length: int = 24,
+        max_encoder_length: int = 24,
         **kwargs,
     ):
         data[split]["FEATURES"]["time_idx"] = (
@@ -480,6 +482,7 @@ def __init__(
         self.data = pd.merge(
             labels, features, on=["stay_id", "time"]
         )  # combine labels and features
+        # self.data["sex"].replace([0, 1], ["Female", "Male"], inplace=True)
         self.split = split
         self.args = args
         self.ram_cache = ram_cache
@@ -495,10 +498,10 @@ def __init__(
             min_prediction_length=max_prediction_length,
             max_prediction_length=max_prediction_length,
             static_categoricals=[],
-            static_reals=["height", "weight", "age", "sex"],
-            time_varying_known_categoricals=[],
-            time_varying_known_reals=[],
-            time_varying_unknown_categoricals=[],
+            static_reals=["height", "weight", "age", "stay_id", "sex"],
+            # time_varying_known_categoricals=[],
+            time_varying_known_reals=["time_idx"],
+            # time_varying_unknown_categoricals=[],
             time_varying_unknown_reals=[
                 "alb",
                 "alp",
@@ -596,6 +599,7 @@ def __init__(
                 "MissingIndicator_46",
                 "MissingIndicator_47",
                 "MissingIndicator_48",
+                "label",
             ],
             add_relative_time_idx=True,
             add_target_scales=True,
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index e46f07b5..5dd067f4 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -15,7 +15,7 @@
 from typing import Dict
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor, cat, FloatTensor
-from pytorch_forecasting import TemporalFusionTransformer
+from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork
 
 from pytorch_forecasting.metrics import QuantileLoss
 
@@ -28,7 +28,12 @@ class RNNet(DLPredictionWrapper):
 
     def __init__(self, input_size, hidden_dim, layer_dim, num_classes, *args, **kwargs):
         super().__init__(
-            input_size=input_size, hidden_dim=hidden_dim, layer_dim=layer_dim, num_classes=num_classes, *args, **kwargs
+            input_size=input_size,
+            hidden_dim=hidden_dim,
+            layer_dim=layer_dim,
+            num_classes=num_classes,
+            *args,
+            **kwargs,
         )
         self.hidden_dim = hidden_dim
         self.layer_dim = layer_dim
@@ -54,7 +59,12 @@ class LSTMNet(DLPredictionWrapper):
 
     def __init__(self, input_size, hidden_dim, layer_dim, num_classes, *args, **kwargs):
         super().__init__(
-            input_size=input_size, hidden_dim=hidden_dim, layer_dim=layer_dim, num_classes=num_classes, *args, **kwargs
+            input_size=input_size,
+            hidden_dim=hidden_dim,
+            layer_dim=layer_dim,
+            num_classes=num_classes,
+            *args,
+            **kwargs,
         )
         self.hidden_dim = hidden_dim
         self.layer_dim = layer_dim
@@ -81,7 +91,12 @@ class GRUNet(DLPredictionWrapper):
 
     def __init__(self, input_size, hidden_dim, layer_dim, num_classes, *args, **kwargs):
         super().__init__(
-            input_size=input_size, hidden_dim=hidden_dim, layer_dim=layer_dim, num_classes=num_classes, *args, **kwargs
+            input_size=input_size,
+            hidden_dim=hidden_dim,
+            layer_dim=layer_dim,
+            num_classes=num_classes,
+            *args,
+            **kwargs,
         )
         self.hidden_dim = hidden_dim
         self.layer_dim = layer_dim
@@ -136,7 +151,9 @@ def __init__(
             **kwargs,
         )
         hidden = hidden if hidden % 2 == 0 else hidden + 1  # Make sure hidden is even
-        self.input_embedding = nn.Linear(input_size[2], hidden)  # This acts as a time-distributed layer by defaults
+        self.input_embedding = nn.Linear(
+            input_size[2], hidden
+        )  # This acts as a time-distributed layer by defaults
         if pos_encoding:
             self.pos_encoder = PositionalEncoding(hidden)
         else:
@@ -207,7 +224,9 @@ def __init__(
         )
 
         hidden = hidden if hidden % 2 == 0 else hidden + 1  # Make sure hidden is even
-        self.input_embedding = nn.Linear(input_size[2], hidden)  # This acts as a time-distributed layer by defaults
+        self.input_embedding = nn.Linear(
+            input_size[2], hidden
+        )  # This acts as a time-distributed layer by defaults
         if pos_encoding:
             self.pos_encoder = PositionalEncoding(hidden)
         else:
@@ -248,7 +267,17 @@ class TemporalConvNet(DLPredictionWrapper):
 
     _supported_run_modes = [RunMode.classification, RunMode.regression]
 
-    def __init__(self, input_size, num_channels, num_classes, *args, max_seq_length=0, kernel_size=2, dropout=0.0, **kwargs):
+    def __init__(
+        self,
+        input_size,
+        num_channels,
+        num_classes,
+        *args,
+        max_seq_length=0,
+        kernel_size=2,
+        dropout=0.0,
+        **kwargs,
+    ):
         super().__init__(
             input_size=input_size,
             num_channels=num_channels,
@@ -263,9 +292,13 @@ def __init__(self, input_size, num_channels, num_classes, *args, max_seq_length=
 
         # We compute automatically the depth based on the desired seq_length.
         if isinstance(num_channels, Integral) and max_seq_length:
-            num_channels = [num_channels] * int(np.ceil(np.log(max_seq_length / 2) / np.log(kernel_size)))
+            num_channels = [num_channels] * int(
+                np.ceil(np.log(max_seq_length / 2) / np.log(kernel_size))
+            )
         elif isinstance(num_channels, Integral) and not max_seq_length:
-            raise Exception("a maximum sequence length needs to be provided if num_channels is int")
+            raise Exception(
+                "a maximum sequence length needs to be provided if num_channels is int"
+            )
 
         num_levels = len(num_channels)
         for i in range(num_levels):
@@ -295,6 +328,7 @@ def forward(self, x):
         return pred
 
 
+'''
 @gin.configurable
 class TFT(DLPredictionWrapper):
     """
@@ -316,7 +350,9 @@ def __init__(
         quantiles=[0.1, 0.5, 0.9],  # quantiles to produce
         static_categorical_inp_size=[2],  # number of catergories
         temporal_known_categorical_inp_size=[],
-        temporal_observed_categorical_inp_size=[48],  # number of categorical observed variables
+        temporal_observed_categorical_inp_size=[
+            48
+        ],  # number of categorical observed variables
         static_continuous_inp_size=3,  # number of static coutinous variables
         temporal_known_continuous_inp_size=0,
         temporal_observed_continuous_inp_size=48,
@@ -325,7 +361,10 @@ def __init__(
     ):
         # derived variables
         num_static_vars = len(static_categorical_inp_size) + static_continuous_inp_size
-        num_future_vars = len(temporal_known_categorical_inp_size) + temporal_known_continuous_inp_size
+        num_future_vars = (
+            len(temporal_known_categorical_inp_size)
+            + temporal_known_continuous_inp_size
+        )
         num_historic_vars = sum(
             [
                 num_future_vars,
@@ -370,7 +409,9 @@ def __init__(
             hidden,
         )  # embeddings for all variables
 
-        self.static_encoder = StaticCovariateEncoder(num_static_vars, hidden, dropout)  # encoding for static variables
+        self.static_encoder = StaticCovariateEncoder(
+            num_static_vars, hidden, dropout
+        )  # encoding for static variables
         self.TFTpart = TFTBack(
             encoder_length,
             num_historic_vars,
@@ -408,11 +449,19 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         historical_inputs = cat(_historical_inputs, dim=-2)
         future_inputs = Tensor()
         if t_known_inp is not None:
-            future_inputs = t_known_inp[:, self.encoder_length:]
-
-        o = self.TFTpart(historical_inputs, cs, ch, cc, ce, future_inputs.to(historical_inputs.device))
+            future_inputs = t_known_inp[:, self.encoder_length :]
+
+        o = self.TFTpart(
+            historical_inputs,
+            cs,
+            ch,
+            cc,
+            ce,
+            future_inputs.to(historical_inputs.device),
+        )
         pred = self.logit(o)
         return pred
+'''
 
 
 @gin.configurable
@@ -423,14 +472,25 @@ class TFTpytorch(DLPredictionWrapper):
 
     supported_run_modes = [RunMode.classification, RunMode.regression]
 
-    def __init__(self, dataset, hidden, dropout, n_heads, dropout_att, optimizer, num_classes, *args, **kwargs):
-        super().__init__(optimizer=optimizer, *args, **kwargs)
+    def __init__(
+        self,
+        dataset,
+        hidden,
+        dropout,
+        n_heads,
+        dropout_att,
+        optimizer,
+        num_classes,
+        *args,
+        **kwargs,
+    ):
+        super().__init__(optimizer=optimizer, pytorch_forecasting=True, *args, **kwargs)
+
         self.model = TemporalFusionTransformer.from_dataset(
             dataset=dataset,
             hidden_size=hidden,
             dropout=dropout,
             attention_head_size=n_heads,
-
             optimizer=optimizer,
             loss=QuantileLoss(),
             hidden_continuous_size=hidden,
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index fdf7bf13..ff70db34 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -23,7 +23,9 @@
 from icu_benchmarks.contants import RunMode
 
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
-gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
+gin.config.external_configurable(
+    nn.functional.cross_entropy, module="torch.nn.functional"
+)
 gin.config.external_configurable(nn.functional.mse_loss, module="torch.nn.functional")
 
 gin.config.external_configurable(mean_squared_error, module="sklearn.metrics")
@@ -88,7 +90,9 @@ def save_model(self, save_path, file_name, file_extension):
 
     def check_supported_runmode(self, runmode: RunMode):
         if runmode not in self._supported_run_modes:
-            raise ValueError(f"Runmode {runmode} not supported for {self.__class__.__name__}")
+            raise ValueError(
+                f"Runmode {runmode} not supported for {self.__class__.__name__}"
+            )
         return True
 
 
@@ -161,7 +165,9 @@ def finalize_step(self, step_prefix=""):
             self.log_dict(
                 {
                     f"{step_prefix}/{name}": (
-                        np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
+                        np.float32(metric.compute())
+                        if isinstance(metric.compute(), np.float64)
+                        else metric.compute()
                     )
                     for name, metric in self.metrics[step_prefix].items()
                     if "_Curve" not in name
@@ -202,7 +208,10 @@ def configure_optimizers(self):
 
     def on_test_epoch_start(self) -> None:
         self.metrics = {
-            step_name: {metric_name: metric() for metric_name, metric in self.set_metrics().items()}
+            step_name: {
+                metric_name: metric()
+                for metric_name, metric in self.set_metrics().items()
+            }
             for step_name in ["train", "val", "test"]
         }
         return super().on_test_epoch_start()
@@ -236,6 +245,7 @@ def __init__(
         epochs: int = 100,
         input_size: Tensor = None,
         initialization_method: str = "normal",
+        pytorch_forecasting: bool = False,
         **kwargs,
     ):
         super().__init__(
@@ -255,6 +265,7 @@ def __init__(
         )
         self.output_transform = None
         self.loss_weights = None
+        self.pytorch_forecasting = pytorch_forecasting
 
     def set_weight(self, weight, dataset):
         """Set the weight for the loss function."""
@@ -308,6 +319,7 @@ def step_fn(self, element, step_prefix=""):
             element (object):
             step_prefix (str): Step type, by default: test, train, val.
         """
+        """
         if isinstance(element[0], OrderedDict):  # check if the data loader is the one for the TFT nvidia implementation
             data, mask = element[0], element[1].to(self.device)
 
@@ -321,10 +333,15 @@ def step_fn(self, element, step_prefix=""):
                 if key == "target":
                     labels = value.squeeze()
                 data[key] = value
-
-        elif len(element) == 2:
+         """
+        if len(element) == 2:
+            """
             if (isinstance(element[1], tuple)) or (
                 isinstance(element[1], list)
+            ):
+            """
+            if (
+                self.pytorch_forecasting
             ):  # check if the data loader is the one for the pytorch forecasring implementation
                 data, labels = element[0], element[1][0]
                 mask = torch.ones_like(labels).bool()
@@ -349,7 +366,9 @@ def step_fn(self, element, step_prefix=""):
             else:
                 data = data.float().to(self.device)
         else:
-            raise Exception("Loader should return either (data, label) or (data, label, mask)")
+            raise Exception(
+                "Loader should return either (data, label) or (data, label, mask)"
+            )
 
         out = self(data)
 
@@ -359,20 +378,31 @@ def step_fn(self, element, step_prefix=""):
         else:
             aux_loss = 0
         # Get prediction and target
-        prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
+        prediction = (
+            torch.masked_select(out, mask.unsqueeze(-1))
+            .reshape(-1, out.shape[-1])
+            .to(self.device)
+        )
 
         target = torch.masked_select(labels, mask).to(self.device)
 
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
-            loss = self.loss(prediction, target.long(), weight=self.loss_weights.to(self.device)) + aux_loss
+            loss = (
+                self.loss(
+                    prediction, target.long(), weight=self.loss_weights.to(self.device)
+                )
+                + aux_loss
+            )
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
 
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
-            raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
+            raise ValueError(
+                f"Run mode {self.run_mode} not yet supported. Please implement it."
+            )
         transformed_output = self.output_transform((prediction, target))
 
         for key, value in self.metrics[step_prefix].items():
@@ -389,7 +419,9 @@ def step_fn(self, element, step_prefix=""):
                     value.update(transformed_output[0], transformed_output[1])
             else:
                 value.update(transformed_output)
-        self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
+        self.log(
+            f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True
+        )
         return loss
 
 
@@ -436,9 +468,15 @@ def set_metrics(self, labels):
 
         # Regression
         else:
-            if self.scaler is not None:  # We invert transform the labels and predictions if they were scaled.
-                self.output_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
-                self.label_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
+            if (
+                self.scaler is not None
+            ):  # We invert transform the labels and predictions if they were scaled.
+                self.output_transform = lambda x: self.scaler.inverse_transform(
+                    x.reshape(-1, 1)
+                )
+                self.label_transform = lambda x: self.scaler.inverse_transform(
+                    x.reshape(-1, 1)
+                )
             else:
                 self.output_transform = lambda x: x
                 self.label_transform = lambda x: x
@@ -473,7 +511,9 @@ def fit_model(self, train_data, train_labels, val_data, val_labels):
 
     def validation_step(self, val_dataset, _):
         val_rep, val_label = val_dataset.get_data_and_labels()
-        val_rep, val_label = torch.from_numpy(val_rep).to(self.device), torch.from_numpy(val_label).to(self.device)
+        val_rep, val_label = torch.from_numpy(val_rep).to(
+            self.device
+        ), torch.from_numpy(val_label).to(self.device)
         self.set_metrics(val_label)
 
         val_pred = self.predict(val_rep)
@@ -516,7 +556,9 @@ def log_metrics(self, label, pred, metric_type):
         self.log_dict(
             {
                 # MPS dependent type casting
-                f"{metric_type}/{name}": metric(self.label_transform(label), self.output_transform(pred))
+                f"{metric_type}/{name}": metric(
+                    self.label_transform(label), self.output_transform(pred)
+                )
                 if not self.mps
                 else metric(self.label_transform(label), self.output_transform(pred))
                 # Fore very metric
@@ -554,7 +596,9 @@ def set_model_args(self, model, *args, **kwargs):
         # Get passed keyword arguments
         arguments = locals()["kwargs"]
         # Get valid hyperparameters
-        hyperparams = {key: value for key, value in arguments.items() if key in possible_hps}
+        hyperparams = {
+            key: value for key, value in arguments.items() if key in possible_hps
+        }
         logging.debug(f"Creating model with: {hyperparams}.")
         return model(**hyperparams)
 
@@ -607,7 +651,9 @@ def set_metrics(self):
     def init_weights(self, init_type="normal", gain=0.02):
         def init_func(m):
             classname = m.__class__.__name__
-            if hasattr(m, "weight") and (classname.find("Conv") != -1 or classname.find("Linear") != -1):
+            if hasattr(m, "weight") and (
+                classname.find("Conv") != -1 or classname.find("Linear") != -1
+            ):
                 if init_type == ImputationInit.NORMAL:
                     nn.init.normal_(m.weight.data, 0.0, gain)
                 elif init_type == ImputationInit.XAVIER:
@@ -617,7 +663,9 @@ def init_func(m):
                 elif init_type == ImputationInit.ORTHOGONAL:
                     nn.init.orthogonal_(m.weight.data, gain=gain)
                 else:
-                    raise NotImplementedError(f"Initialization method {init_type} is not implemented")
+                    raise NotImplementedError(
+                        f"Initialization method {init_type} is not implemented"
+                    )
                 if hasattr(m, "bias") and m.bias is not None:
                     nn.init.constant_(m.bias.data, 0.0)
             elif classname.find("BatchNorm2d") != -1:
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 345d7948..00fc3ef4 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -52,7 +52,8 @@ def main(my_args=tuple(sys.argv[1:])):
     evaluate = args.eval
     experiment = args.experiment
     source_dir = args.source_dir
-    explain = True if args.explain else False
+    explain = args.explain
+    pytorch_forecasting = args.pytorch_forecasting
     # Load task config
     gin.parse_config_file(f"configs/tasks/{task}.gin")
     mode = get_mode()
@@ -205,6 +206,7 @@ def main(my_args=tuple(sys.argv[1:])):
         wandb=args.wandb_sweep,
         complete_train=args.complete_train,
         explain=explain,
+        pytorch_forecasting=pytorch_forecasting,
     )
 
     log_full_line("FINISHED TRAINING", level=logging.INFO, char="=", num_newlines=3)
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index f6e54e41..94f95c61 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -151,6 +151,12 @@ def build_parser() -> ArgumentParser:
         action=BOA,
         help="Provide explaintations for predictions.",
     )
+    parser.add_argument(
+        "--pytorch-forecasting",
+        default=False,
+        action=BOA,
+        help="use pytorch forecasting library ",
+    )
     return parser
 
 

From ec011c8e971266ced2487d9102ef2019ed5ac9bd Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 26 Sep 2023 16:27:48 +0200
Subject: [PATCH 062/142] added pytorch tft built in methods

---
 icu_benchmarks/models/dl_models.py | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 5dd067f4..f55c9a9a 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -512,3 +512,28 @@ def forward(self, x):
         out = self.model(x)
         pred = self.logit(out["prediction"])
         return pred
+
+    def actual_vs_predictions_plot(self, dataloader):
+        predictions = self.model.predict(dataloader, return_x=True)
+        predictions_vs_actuals = self.model.calculate_prediction_actual_by_variable(
+            predictions.x, predictions.output
+        )
+        self.model.plot_prediction_actual_by_variable(predictions_vs_actuals)
+        return predictions_vs_actuals
+    
+    def interpertations(self,dataloader):
+        raw_predictions = self.model.predict(dataloader, return_x=True mode="raw")
+        interpretation = self.model.interpret_output(raw_predictions.output, reduction="sum")
+        self.model.plot_interpretation(interpretation)
+        return interpretation
+    def partial_depenecdy(self,dataloader,variable):
+        dependency = self.model.predict_dependency(
+        dataloader.dataset, variable, np.linspace(0, 30, 30), show_progress_bar=True, mode="dataframe"
+        )
+        # plotting median and 25% and 75% percentile
+        agg_dependency = dependency.groupby(variable).normalized_prediction.agg(
+            median="median", q25=lambda x: x.quantile(0.25), q75=lambda x: x.quantile(0.75)
+        )
+        ax = agg_dependency.plot(y="median")
+        ax.fill_between(agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3)
+        return dependency
\ No newline at end of file

From 2f1b74ab63ed1a3b23975472e515e25eac71ffa1 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 26 Sep 2023 16:31:21 +0200
Subject: [PATCH 063/142] added coma

---
 icu_benchmarks/models/dl_models.py | 29 ++++++++++++++++++++---------
 1 file changed, 20 insertions(+), 9 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index f55c9a9a..69933aca 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -520,20 +520,31 @@ def actual_vs_predictions_plot(self, dataloader):
         )
         self.model.plot_prediction_actual_by_variable(predictions_vs_actuals)
         return predictions_vs_actuals
-    
-    def interpertations(self,dataloader):
-        raw_predictions = self.model.predict(dataloader, return_x=True mode="raw")
-        interpretation = self.model.interpret_output(raw_predictions.output, reduction="sum")
+
+    def interpertations(self, dataloader):
+        raw_predictions = self.model.predict(dataloader, return_x=True, mode="raw")
+        interpretation = self.model.interpret_output(
+            raw_predictions.output, reduction="sum"
+        )
         self.model.plot_interpretation(interpretation)
         return interpretation
-    def partial_depenecdy(self,dataloader,variable):
+
+    def partial_depenecdy(self, dataloader, variable):
         dependency = self.model.predict_dependency(
-        dataloader.dataset, variable, np.linspace(0, 30, 30), show_progress_bar=True, mode="dataframe"
+            dataloader.dataset,
+            variable,
+            np.linspace(0, 30, 30),
+            show_progress_bar=True,
+            mode="dataframe",
         )
         # plotting median and 25% and 75% percentile
         agg_dependency = dependency.groupby(variable).normalized_prediction.agg(
-            median="median", q25=lambda x: x.quantile(0.25), q75=lambda x: x.quantile(0.75)
+            median="median",
+            q25=lambda x: x.quantile(0.25),
+            q75=lambda x: x.quantile(0.75),
         )
         ax = agg_dependency.plot(y="median")
-        ax.fill_between(agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3)
-        return dependency
\ No newline at end of file
+        ax.fill_between(
+            agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3
+        )
+        return dependency

From 86d719a3a8c12755e018f06080b6817092e2a2b9 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 26 Sep 2023 17:00:07 +0200
Subject: [PATCH 064/142] added attention explantation for tft

---
 icu_benchmarks/models/dl_models.py | 14 ++++---
 icu_benchmarks/models/train.py     | 67 ++++++++++++++++++------------
 2 files changed, 49 insertions(+), 32 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 69933aca..d4fccccf 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -12,12 +12,12 @@
     StaticCovariateEncoder,
     TFTBack,
 )
-from typing import Dict
+import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor, cat, FloatTensor
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork
-
 from pytorch_forecasting.metrics import QuantileLoss
+import matplotlib.pyplot as plt
 
 
 @gin.configurable
@@ -521,15 +521,18 @@ def actual_vs_predictions_plot(self, dataloader):
         self.model.plot_prediction_actual_by_variable(predictions_vs_actuals)
         return predictions_vs_actuals
 
-    def interpertations(self, dataloader):
+    def interpertations(self, dataloader, log_dir):
         raw_predictions = self.model.predict(dataloader, return_x=True, mode="raw")
         interpretation = self.model.interpret_output(
             raw_predictions.output, reduction="sum"
         )
-        self.model.plot_interpretation(interpretation)
+        figs = self.model.plot_interpretation(interpretation)
+        for key, fig in figs.items():
+            fig.savefig(log_dir / f"interpretation_{key}.png", bbox_inches="tight")
+
         return interpretation
 
-    def partial_depenecdy(self, dataloader, variable):
+    def predict_dependency(self, dataloader, variable, log_dir):
         dependency = self.model.predict_dependency(
             dataloader.dataset,
             variable,
@@ -547,4 +550,5 @@ def partial_depenecdy(self, dataloader, variable):
         ax.fill_between(
             agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3
         )
+        plt.savefig(log_dir / "dependecy.png", bbox_inches="tight")
         return dependency
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 4d5b4d7f..c697bb29 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -19,7 +19,6 @@
 from icu_benchmarks.data.loader import (
     PredictionDataset,
     ImputationDataset,
-    PredictionDatasetTFT,
     PredictionDatasetTFTpytorch,
 )
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
@@ -70,6 +69,7 @@ def train_common(
         32,
     ),
     explain: bool = False,
+    pytorch_forecasting: bool = False,
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -97,21 +97,13 @@ def train_common(
         pl_model: Loading a pytorch lightning model.
         num_workers: Number of workers to use for data loading.
     """
-    with open("data.pkl", "wb") as fp:
-        pickle.dump(data, fp)
     logging.info(f"Training model: {model.__name__}.")
     # choose dataset_class based on the model
     dataset_class = (
         ImputationDataset
         if mode == RunMode.imputation
         else (
-            PredictionDatasetTFT
-            if model.__name__ == "TFT"
-            else (
-                PredictionDatasetTFTpytorch
-                if (model.__name__ == "TFTpytorch")
-                else PredictionDataset
-            )
+            PredictionDatasetTFTpytorch if (pytorch_forecasting) else PredictionDataset
         )
     )
 
@@ -140,7 +132,7 @@ def train_common(
 
     logging.info(f"Using {num_workers} workers for data loading.")
 
-    if model.__name__ == "TFTpytorch":
+    if pytorch_forecasting:
         train_loader = train_dataset.to_dataloader(
             train=True,
             batch_size=batch_size,
@@ -163,7 +155,15 @@ def train_common(
             drop_last=True,
             shuffle=False,
         )
-        model = model(train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode)
+        if load_weights:
+            model = load_model(
+                model, source_dir, pl_model=pl_model, train_dataset=train_dataset
+            )
+
+        else:
+            model = model(
+                train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode
+            )
 
     else:
         train_loader = DataLoader(
@@ -195,24 +195,20 @@ def train_common(
             if model.requires_backprop
             else DataLoader([test_dataset.to_tensor()], batch_size=1)
         )
-
+        """
         if isinstance(next(iter(train_loader))[0], OrderedDict):
             model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
+        """
 
+        data_shape = next(iter(train_loader))[0].shape
+        if load_weights:
+            model = load_model(model, source_dir, pl_model=pl_model)
         else:
-            data_shape = next(iter(train_loader))[0].shape
             model = model(
                 optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
             )
 
-    if load_weights:
-        model = load_model(model, source_dir, pl_model=pl_model)
-    logging.info(
-        f"train_dataloader with {len(train_loader)} samples and validating on loader with"
-        f" {len(val_loader)} samples."
-    )
     model.set_weight(weight, train_dataset)
-
     model.set_trained_columns(train_dataset.get_feature_names())
     loggers = [TensorBoardLogger(log_dir), JSONMetricsLogger(log_dir)]
     if use_wandb:
@@ -263,18 +259,29 @@ def train_common(
         logging.info("Finished training full model.")
         save_config_file(log_dir)
         return 0
+
     if explain:
+        # actual_vs_predictions = model.actual_vs_predictions_plot(test_loader)
+        # print("1", actual_vs_predictions)
+        interperations = model.interpertations(test_loader, log_dir)
+        print("2", interperations)
+        pd = model.predict_dependency(test_loader, "age", log_dir)
+        print("3", pd)
+        """
         test = next(iter(test_loader))
-        test[0] = test[0].to(model.device)
+
+        for key, value in test[0].items():
+            test[0][key] = test[0][key].to(model.device)
+
         pred = model(test[0])
+
         ig = IntegratedGradients(model)
         # Reformat attributions.
-        test[0].requires_grad_()
-        print(pred.shape)
-        print(test[0].shape)
+
         attr, delta = ig.attribute(test[0], target=0, return_convergence_delta=True)
         attr = attr.detach().numpy()
         print(attr)
+        """
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
         "test/loss"
@@ -283,7 +290,7 @@ def train_common(
     return test_loss
 
 
-def load_model(model, source_dir, pl_model=True):
+def load_model(model, source_dir, pl_model=True, train_dataset=None):
     if source_dir.exists():
         if model.requires_backprop:
             if (source_dir / "model.ckpt").exists():
@@ -295,7 +302,13 @@ def load_model(model, source_dir, pl_model=True):
             else:
                 return Exception(f"No weights to load at path : {source_dir}")
             if pl_model:
-                model = model.load_from_checkpoint(model_path)
+                if train_dataset is not None:
+                    model = model.load_from_checkpoint(
+                        model_path, dataset=train_dataset
+                    )
+
+                else:
+                    model = model.load_from_checkpoint(model_path)
             else:
                 checkpoint = torch.load(model_path)
                 model.load_from_checkpoint(checkpoint)

From 00d78dd0dc29313aead9479ac3fbf5554546af98 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Wed, 27 Sep 2023 14:07:31 +0200
Subject: [PATCH 065/142] changed forward method for tft to allow captum
 explantations

---
 icu_benchmarks/models/dl_models.py | 24 ++++++++++++++++++++++++
 icu_benchmarks/models/wrappers.py  | 16 ++++++++++++++--
 2 files changed, 38 insertions(+), 2 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index d4fccccf..88517aea 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -2,6 +2,7 @@
 from numbers import Integral
 import numpy as np
 import torch.nn as nn
+from typing import Dict
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.models.layers import (
     TransformerBlock,
@@ -508,10 +509,33 @@ def set_weight(self, weight, dataset):
             weight = FloatTensor(dataset.get_balance())
         self.loss_weights = weight
 
+    def forward(
+        self,
+        tuple_x: tuple,
+    ) -> Dict[str, Tensor]:
+        x_dict = {
+            "encoder_cat": tuple_x[0],
+            "encoder_cont": tuple_x[1],
+            "encoder_target": tuple_x[2],
+            "encoder_lengths": tuple_x[3],
+            "decoder_cat": tuple_x[4],
+            "decoder_cont": tuple_x[5],
+            "decoder_target": tuple_x[6],
+            "decoder_lengths": tuple_x[7],
+            "decoder_time_idx": tuple_x[8],
+            "groups": tuple_x[9],
+            "target_scale": tuple_x[10],
+        }
+        out = self.model(x_dict)
+        pred = self.logit(out["prediction"])
+        return pred
+
+    """
     def forward(self, x):
         out = self.model(x)
         pred = self.logit(out["prediction"])
         return pred
+    """
 
     def actual_vs_predictions_plot(self, dataloader):
         predictions = self.model.predict(dataloader, return_x=True)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index ff70db34..c60cd6a5 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -343,7 +343,20 @@ def step_fn(self, element, step_prefix=""):
             if (
                 self.pytorch_forecasting
             ):  # check if the data loader is the one for the pytorch forecasring implementation
-                data, labels = element[0], element[1][0]
+                dic, labels = element[0], element[1][0]
+                data = (
+                    dic["encoder_cat"],
+                    dic["encoder_cont"],
+                    dic["encoder_target"],
+                    dic["encoder_lengths"],
+                    dic["decoder_cat"],
+                    dic["decoder_cont"],
+                    dic["decoder_target"],
+                    dic["decoder_lengths"],
+                    dic["decoder_time_idx"],
+                    dic["groups"],
+                    dic["target_scale"],
+                )
                 mask = torch.ones_like(labels).bool()
             else:
                 data, labels = element[0], (element[1]).to(self.device)
@@ -369,7 +382,6 @@ def step_fn(self, element, step_prefix=""):
             raise Exception(
                 "Loader should return either (data, label) or (data, label, mask)"
             )
-
         out = self(data)
 
         # If aux_loss is present, it is returned as a tuple

From e877c5ffa5189efca46337f149bf3ad53f5c8de2 Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Fri, 29 Sep 2023 14:55:06 +0200
Subject: [PATCH 066/142] cleaned files up

---
 configs/prediction_models/TFTpytorch.gin      |   4 +-
 configs/prediction_models/common/DLCommon.gin |   4 +-
 configs/prediction_models/common/DLTuning.gin |   2 +-
 icu_benchmarks/cross_validation.py            |   8 +-
 icu_benchmarks/data/loader.py                 |  48 ++-----
 icu_benchmarks/imputation/sssds4.py           |   3 +-
 icu_benchmarks/models/dl_models.py            |  38 ++----
 icu_benchmarks/models/train.py                |  48 ++-----
 icu_benchmarks/models/wrappers.py             |  81 +++---------
 icu_benchmarks/run.py                         |  42 ++----
 icu_benchmarks/run_utils.py                   | 121 +++++-------------
 11 files changed, 101 insertions(+), 298 deletions(-)

diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 27e2cf25..13c7520a 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -12,11 +12,11 @@ optimizer/hyperparameter.weight_decay = 1e-6
 
 # Encoder params
 model/hyperparameter.class_to_tune = @TFTpytorch
-model/hyperparameter.hidden = 16
+model/hyperparameter.hidden = (16,32,64,128)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =1
+model/hyperparameter.n_heads =(1,2,4)
 model/hyperparameter.lr_scheduler = "exponential"
 PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index f41f2104..7c72b17a 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -14,8 +14,8 @@ base_regression_preprocessor.generate_features = False
 
 # Train params
 train_common.optimizer = @Adam
-train_common.epochs = 50
-train_common.batch_size = 64
+train_common.epochs = 1000
+train_common.batch_size = 256
 train_common.patience = 10
 train_common.min_delta = 1e-4
 
diff --git a/configs/prediction_models/common/DLTuning.gin b/configs/prediction_models/common/DLTuning.gin
index b4d13e12..1a2b7bb4 100644
--- a/configs/prediction_models/common/DLTuning.gin
+++ b/configs/prediction_models/common/DLTuning.gin
@@ -2,4 +2,4 @@
 tune_hyperparameters.scopes = ["model", "optimizer"]
 tune_hyperparameters.n_initial_points = 5
 tune_hyperparameters.n_calls = 30
-tune_hyperparameters.folds_to_tune_on = 2
\ No newline at end of file
+tune_hyperparameters.folds_to_tune_on = 1
\ No newline at end of file
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index bcf2f5f5..412df3eb 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -82,15 +82,11 @@ def execute_repeated_cv(
         cv_folds_to_train = 1
 
     else:
-        logging.info(
-            f"Starting nested CV with {cv_repetitions_to_train} repetitions of {cv_folds_to_train} folds."
-        )
+        logging.info(f"Starting nested CV with {cv_repetitions_to_train} repetitions of {cv_folds_to_train} folds.")
 
     for repetition in range(cv_repetitions_to_train):
         for fold_index in range(cv_folds_to_train):
-            repetition_fold_dir = (
-                log_dir / f"repetition_{repetition}" / f"fold_{fold_index}"
-            )
+            repetition_fold_dir = log_dir / f"repetition_{repetition}" / f"fold_{fold_index}"
             repetition_fold_dir.mkdir(parents=True, exist_ok=True)
 
             start_time = datetime.now()
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 5dde7809..0e46cc51 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -3,15 +3,13 @@
 import pandas as pd
 import gin
 import numpy as np
-from torch import Tensor, cat, from_numpy, float32, empty, stack
+from torch import Tensor, cat, from_numpy, float32
 from torch.utils.data import Dataset
 import logging
 from typing import Dict, Tuple
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
-from .constants import FeatType as Features
-from collections import OrderedDict
 from pytorch_forecasting import TimeSeriesDataSet
 
 
@@ -105,15 +103,11 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id].to_numpy()
-        labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(
-            dtype=float
-        )
+        labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
 
         if len(labels) == 1:
             # only one label per stay, align with window
-            labels = np.concatenate(
-                [np.empty(window.shape[0] - 1) * np.nan, labels], axis=0
-            )
+            labels = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, labels], axis=0)
 
         length_diff = self.maxlen - window.shape[0]
         pad_mask = np.ones(window.shape[0])
@@ -121,9 +115,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         # Padding the array to fulfill size requirement
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
-            window = np.concatenate(
-                [window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0
-            )
+            window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
             labels = np.concatenate([labels, np.ones(length_diff) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
 
@@ -179,7 +171,7 @@ class PredictionDatasetTFT(PredictionDataset):
     We also need to feed the model the variables in a specific order
     Args:
         ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.
-    
+
 
     def __init__(self, *args, ram_cache: bool = True, **kwargs):
         super().__init__(*args, ram_cache=True, **kwargs)
@@ -193,7 +185,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
             idx: A specific row index to sample.
         Returns:
             A sample from the data, consisting of data, labels and padding mask.
-        
+
         if self._cached_dataset is not None:
             return self._cached_dataset[idx]
 
@@ -333,12 +325,8 @@ def __init__(
         self.amputated_values, self.amputation_mask = ampute_data(
             self.features_df, mask_method, mask_proportion, mask_observation_proportion
         )
-        self.amputation_mask = (
-            self.amputation_mask + self.features_df.isna().values
-        ).bool()
-        self.amputation_mask = DataFrame(
-            self.amputation_mask, columns=self.vars[Segment.dynamic]
-        )
+        self.amputation_mask = (self.amputation_mask + self.features_df.isna().values).bool()
+        self.amputation_mask = DataFrame(self.amputation_mask, columns=self.vars[Segment.dynamic])
         self.amputation_mask[self.vars["GROUP"]] = self.features_df.index
         self.amputation_mask.set_index(self.vars["GROUP"], inplace=True)
 
@@ -363,15 +351,9 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
-        window_missingness_mask = self.target_missingness_mask.loc[
-            stay_id:stay_id, self.vars[Segment.dynamic]
-        ]
-        amputated_window = self.amputated_values.loc[
-            stay_id:stay_id, self.vars[Segment.dynamic]
-        ]
-        amputation_mask = self.amputation_mask.loc[
-            stay_id:stay_id, self.vars[Segment.dynamic]
-        ]
+        window_missingness_mask = self.target_missingness_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
+        amputated_window = self.amputated_values.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
+        amputation_mask = self.amputation_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
 
         return (
             from_numpy(amputated_window.values).to(float32),
@@ -470,18 +452,14 @@ def __init__(
         max_encoder_length: int = 24,
         **kwargs,
     ):
-        data[split]["FEATURES"]["time_idx"] = (
-            (data[split]["FEATURES"]["time"] / pd.Timedelta(seconds=3600))
-        ).astype(
+        data[split]["FEATURES"]["time_idx"] = ((data[split]["FEATURES"]["time"] / pd.Timedelta(seconds=3600))).astype(
             int
         )  # create an incremental column indicating the time step(required by constructor)
         data = data.get(split)  # get split
         labels = data["OUTCOME"]
         features = data["FEATURES"]
         self.name = name
-        self.data = pd.merge(
-            labels, features, on=["stay_id", "time"]
-        )  # combine labels and features
+        self.data = pd.merge(labels, features, on=["stay_id", "time"])  # combine labels and features
         # self.data["sex"].replace([0, 1], ["Female", "Male"], inplace=True)
         self.split = split
         self.args = args
diff --git a/icu_benchmarks/imputation/sssds4.py b/icu_benchmarks/imputation/sssds4.py
index d359fb09..25933b68 100644
--- a/icu_benchmarks/imputation/sssds4.py
+++ b/icu_benchmarks/imputation/sssds4.py
@@ -414,7 +414,8 @@ def calc_diffusion_hyperparams(diffusion_time_steps, beta_0, beta_T):
     Beta_tilde = Beta + 0
     for t in range(1, diffusion_time_steps):
         Alpha_bar[t] *= Alpha_bar[t - 1]  # \bar{\alpha}_t = \prod_{s=1}^t \alpha_s
-        Beta_tilde[t] *= (1 - Alpha_bar[t - 1]) / (1 - Alpha_bar[t])  # \tilde{\beta}_t = \beta_t * (1-\bar{\alpha}_{t-1})
+        # \tilde{\beta}_t = \beta_t * (1-\bar{\alpha}_{t-1})
+        Beta_tilde[t] *= (1 - Alpha_bar[t - 1]) / (1 - Alpha_bar[t])
         # / (1-\bar{\alpha}_t)
     Sigma = torch.sqrt(Beta_tilde)  # \sigma_t^2  = \tilde{\beta}_t
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 88517aea..7070ffac 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -9,14 +9,11 @@
     LocalBlock,
     TemporalBlock,
     PositionalEncoding,
-    LazyEmbedding,
-    StaticCovariateEncoder,
-    TFTBack,
 )
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor, cat, FloatTensor
-from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork
+from torch import Tensor, FloatTensor
+from pytorch_forecasting import TemporalFusionTransformer  # , RecurrentNetwork
 from pytorch_forecasting.metrics import QuantileLoss
 import matplotlib.pyplot as plt
 
@@ -152,9 +149,7 @@ def __init__(
             **kwargs,
         )
         hidden = hidden if hidden % 2 == 0 else hidden + 1  # Make sure hidden is even
-        self.input_embedding = nn.Linear(
-            input_size[2], hidden
-        )  # This acts as a time-distributed layer by defaults
+        self.input_embedding = nn.Linear(input_size[2], hidden)  # This acts as a time-distributed layer by defaults
         if pos_encoding:
             self.pos_encoder = PositionalEncoding(hidden)
         else:
@@ -225,9 +220,7 @@ def __init__(
         )
 
         hidden = hidden if hidden % 2 == 0 else hidden + 1  # Make sure hidden is even
-        self.input_embedding = nn.Linear(
-            input_size[2], hidden
-        )  # This acts as a time-distributed layer by defaults
+        self.input_embedding = nn.Linear(input_size[2], hidden)  # This acts as a time-distributed layer by defaults
         if pos_encoding:
             self.pos_encoder = PositionalEncoding(hidden)
         else:
@@ -293,13 +286,9 @@ def __init__(
 
         # We compute automatically the depth based on the desired seq_length.
         if isinstance(num_channels, Integral) and max_seq_length:
-            num_channels = [num_channels] * int(
-                np.ceil(np.log(max_seq_length / 2) / np.log(kernel_size))
-            )
+            num_channels = [num_channels] * int(np.ceil(np.log(max_seq_length / 2) / np.log(kernel_size)))
         elif isinstance(num_channels, Integral) and not max_seq_length:
-            raise Exception(
-                "a maximum sequence length needs to be provided if num_channels is int"
-            )
+            raise Exception("a maximum sequence length needs to be provided if num_channels is int")
 
         num_levels = len(num_channels)
         for i in range(num_levels):
@@ -333,7 +322,8 @@ def forward(self, x):
 @gin.configurable
 class TFT(DLPredictionWrapper):
     """
-    Implementation of https://arxiv.org/abs/1912.09363 from https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/Forecasting/TFT
+    Implementation of https://arxiv.org/abs/1912.09363 
+    from https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/Forecasting/TFT
     """
 
     _supported_run_modes = [RunMode.classification, RunMode.regression]
@@ -539,17 +529,13 @@ def forward(self, x):
 
     def actual_vs_predictions_plot(self, dataloader):
         predictions = self.model.predict(dataloader, return_x=True)
-        predictions_vs_actuals = self.model.calculate_prediction_actual_by_variable(
-            predictions.x, predictions.output
-        )
+        predictions_vs_actuals = self.model.calculate_prediction_actual_by_variable(predictions.x, predictions.output)
         self.model.plot_prediction_actual_by_variable(predictions_vs_actuals)
         return predictions_vs_actuals
 
     def interpertations(self, dataloader, log_dir):
         raw_predictions = self.model.predict(dataloader, return_x=True, mode="raw")
-        interpretation = self.model.interpret_output(
-            raw_predictions.output, reduction="sum"
-        )
+        interpretation = self.model.interpret_output(raw_predictions.output, reduction="sum")
         figs = self.model.plot_interpretation(interpretation)
         for key, fig in figs.items():
             fig.savefig(log_dir / f"interpretation_{key}.png", bbox_inches="tight")
@@ -571,8 +557,6 @@ def predict_dependency(self, dataloader, variable, log_dir):
             q75=lambda x: x.quantile(0.75),
         )
         ax = agg_dependency.plot(y="median")
-        ax.fill_between(
-            agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3
-        )
+        ax.fill_between(agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3)
         plt.savefig(log_dir / "dependecy.png", bbox_inches="tight")
         return dependency
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index c697bb29..1cca8970 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -27,9 +27,7 @@
 from collections import OrderedDict
 from captum.attr import IntegratedGradients
 
-cpu_core_count = (
-    len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
-)
+cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
 
 
 def assure_minimum_length(dataset):
@@ -102,25 +100,15 @@ def train_common(
     dataset_class = (
         ImputationDataset
         if mode == RunMode.imputation
-        else (
-            PredictionDatasetTFTpytorch if (pytorch_forecasting) else PredictionDataset
-        )
+        else (PredictionDatasetTFTpytorch if (pytorch_forecasting) else PredictionDataset)
     )
 
     logging.info(f"Logging to directory: {log_dir}.")
-    save_config_file(
-        log_dir
-    )  # We save the operative config before and also after training
+    save_config_file(log_dir)  # We save the operative config before and also after training
 
-    train_dataset = dataset_class(
-        data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"]
-    )
-    val_dataset = dataset_class(
-        data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"]
-    )
-    train_dataset, val_dataset = assure_minimum_length(
-        train_dataset
-    ), assure_minimum_length(val_dataset)
+    train_dataset = dataset_class(data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"])
+    val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"])
+    train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
     test_dataset = dataset_class(data, split=test_on, name=dataset_names["test"])
     test_dataset = assure_minimum_length(test_dataset)
@@ -156,14 +144,10 @@ def train_common(
             shuffle=False,
         )
         if load_weights:
-            model = load_model(
-                model, source_dir, pl_model=pl_model, train_dataset=train_dataset
-            )
+            model = load_model(model, source_dir, pl_model=pl_model, train_dataset=train_dataset)
 
         else:
-            model = model(
-                train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode
-            )
+            model = model(train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode)
 
     else:
         train_loader = DataLoader(
@@ -204,9 +188,7 @@ def train_common(
         if load_weights:
             model = load_model(model, source_dir, pl_model=pl_model)
         else:
-            model = model(
-                optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
-            )
+            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
 
     model.set_weight(weight, train_dataset)
     model.set_trained_columns(train_dataset.get_feature_names())
@@ -246,9 +228,7 @@ def train_common(
     if not eval_only:
         if model.requires_backprop:
             logging.info("Training DL model.")
-            trainer.fit(
-                model, train_dataloaders=train_loader, val_dataloaders=val_loader
-            )
+            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
             logging.info("Training complete.")
         else:
             logging.info("Training ML model.")
@@ -283,9 +263,7 @@ def train_common(
         print(attr)
         """
     model.set_weight("balanced", train_dataset)
-    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
-        "test/loss"
-    ]
+    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
     return test_loss
 
@@ -303,9 +281,7 @@ def load_model(model, source_dir, pl_model=True, train_dataset=None):
                 return Exception(f"No weights to load at path : {source_dir}")
             if pl_model:
                 if train_dataset is not None:
-                    model = model.load_from_checkpoint(
-                        model_path, dataset=train_dataset
-                    )
+                    model = model.load_from_checkpoint(model_path, dataset=train_dataset)
 
                 else:
                     model = model.load_from_checkpoint(model_path)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index c60cd6a5..8bffe10f 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -1,16 +1,13 @@
 import logging
 from abc import ABC
 from typing import Dict, Any, List, Optional, Union
-
 import torchmetrics
 from sklearn.metrics import log_loss, mean_squared_error
-
 import torch
 from torch.nn import MSELoss, CrossEntropyLoss
 import torch.nn as nn
 from torch import Tensor, FloatTensor
 from torch.optim import Optimizer, Adam
-from collections import OrderedDict
 import inspect
 import gin
 import numpy as np
@@ -23,9 +20,7 @@
 from icu_benchmarks.contants import RunMode
 
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
-gin.config.external_configurable(
-    nn.functional.cross_entropy, module="torch.nn.functional"
-)
+gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.mse_loss, module="torch.nn.functional")
 
 gin.config.external_configurable(mean_squared_error, module="sklearn.metrics")
@@ -90,9 +85,7 @@ def save_model(self, save_path, file_name, file_extension):
 
     def check_supported_runmode(self, runmode: RunMode):
         if runmode not in self._supported_run_modes:
-            raise ValueError(
-                f"Runmode {runmode} not supported for {self.__class__.__name__}"
-            )
+            raise ValueError(f"Runmode {runmode} not supported for {self.__class__.__name__}")
         return True
 
 
@@ -165,9 +158,7 @@ def finalize_step(self, step_prefix=""):
             self.log_dict(
                 {
                     f"{step_prefix}/{name}": (
-                        np.float32(metric.compute())
-                        if isinstance(metric.compute(), np.float64)
-                        else metric.compute()
+                        np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
                     )
                     for name, metric in self.metrics[step_prefix].items()
                     if "_Curve" not in name
@@ -208,10 +199,7 @@ def configure_optimizers(self):
 
     def on_test_epoch_start(self) -> None:
         self.metrics = {
-            step_name: {
-                metric_name: metric()
-                for metric_name, metric in self.set_metrics().items()
-            }
+            step_name: {metric_name: metric() for metric_name, metric in self.set_metrics().items()}
             for step_name in ["train", "val", "test"]
         }
         return super().on_test_epoch_start()
@@ -340,9 +328,7 @@ def step_fn(self, element, step_prefix=""):
                 isinstance(element[1], list)
             ):
             """
-            if (
-                self.pytorch_forecasting
-            ):  # check if the data loader is the one for the pytorch forecasring implementation
+            if self.pytorch_forecasting:  # check if the data loader is the one for the pytorch forecasring implementation
                 dic, labels = element[0], element[1][0]
                 data = (
                     dic["encoder_cat"],
@@ -379,9 +365,7 @@ def step_fn(self, element, step_prefix=""):
             else:
                 data = data.float().to(self.device)
         else:
-            raise Exception(
-                "Loader should return either (data, label) or (data, label, mask)"
-            )
+            raise Exception("Loader should return either (data, label) or (data, label, mask)")
         out = self(data)
 
         # If aux_loss is present, it is returned as a tuple
@@ -390,31 +374,20 @@ def step_fn(self, element, step_prefix=""):
         else:
             aux_loss = 0
         # Get prediction and target
-        prediction = (
-            torch.masked_select(out, mask.unsqueeze(-1))
-            .reshape(-1, out.shape[-1])
-            .to(self.device)
-        )
+        prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
 
         target = torch.masked_select(labels, mask).to(self.device)
 
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
-            loss = (
-                self.loss(
-                    prediction, target.long(), weight=self.loss_weights.to(self.device)
-                )
-                + aux_loss
-            )
+            loss = self.loss(prediction, target.long(), weight=self.loss_weights.to(self.device)) + aux_loss
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
 
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
-            raise ValueError(
-                f"Run mode {self.run_mode} not yet supported. Please implement it."
-            )
+            raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
         transformed_output = self.output_transform((prediction, target))
 
         for key, value in self.metrics[step_prefix].items():
@@ -431,9 +404,7 @@ def step_fn(self, element, step_prefix=""):
                     value.update(transformed_output[0], transformed_output[1])
             else:
                 value.update(transformed_output)
-        self.log(
-            f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True
-        )
+        self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
         return loss
 
 
@@ -480,15 +451,9 @@ def set_metrics(self, labels):
 
         # Regression
         else:
-            if (
-                self.scaler is not None
-            ):  # We invert transform the labels and predictions if they were scaled.
-                self.output_transform = lambda x: self.scaler.inverse_transform(
-                    x.reshape(-1, 1)
-                )
-                self.label_transform = lambda x: self.scaler.inverse_transform(
-                    x.reshape(-1, 1)
-                )
+            if self.scaler is not None:  # We invert transform the labels and predictions if they were scaled.
+                self.output_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
+                self.label_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
             else:
                 self.output_transform = lambda x: x
                 self.label_transform = lambda x: x
@@ -523,9 +488,7 @@ def fit_model(self, train_data, train_labels, val_data, val_labels):
 
     def validation_step(self, val_dataset, _):
         val_rep, val_label = val_dataset.get_data_and_labels()
-        val_rep, val_label = torch.from_numpy(val_rep).to(
-            self.device
-        ), torch.from_numpy(val_label).to(self.device)
+        val_rep, val_label = torch.from_numpy(val_rep).to(self.device), torch.from_numpy(val_label).to(self.device)
         self.set_metrics(val_label)
 
         val_pred = self.predict(val_rep)
@@ -568,9 +531,7 @@ def log_metrics(self, label, pred, metric_type):
         self.log_dict(
             {
                 # MPS dependent type casting
-                f"{metric_type}/{name}": metric(
-                    self.label_transform(label), self.output_transform(pred)
-                )
+                f"{metric_type}/{name}": metric(self.label_transform(label), self.output_transform(pred))
                 if not self.mps
                 else metric(self.label_transform(label), self.output_transform(pred))
                 # Fore very metric
@@ -608,9 +569,7 @@ def set_model_args(self, model, *args, **kwargs):
         # Get passed keyword arguments
         arguments = locals()["kwargs"]
         # Get valid hyperparameters
-        hyperparams = {
-            key: value for key, value in arguments.items() if key in possible_hps
-        }
+        hyperparams = {key: value for key, value in arguments.items() if key in possible_hps}
         logging.debug(f"Creating model with: {hyperparams}.")
         return model(**hyperparams)
 
@@ -663,9 +622,7 @@ def set_metrics(self):
     def init_weights(self, init_type="normal", gain=0.02):
         def init_func(m):
             classname = m.__class__.__name__
-            if hasattr(m, "weight") and (
-                classname.find("Conv") != -1 or classname.find("Linear") != -1
-            ):
+            if hasattr(m, "weight") and (classname.find("Conv") != -1 or classname.find("Linear") != -1):
                 if init_type == ImputationInit.NORMAL:
                     nn.init.normal_(m.weight.data, 0.0, gain)
                 elif init_type == ImputationInit.XAVIER:
@@ -675,9 +632,7 @@ def init_func(m):
                 elif init_type == ImputationInit.ORTHOGONAL:
                     nn.init.orthogonal_(m.weight.data, gain=gain)
                 else:
-                    raise NotImplementedError(
-                        f"Initialization method {init_type} is not implemented"
-                    )
+                    raise NotImplementedError(f"Initialization method {init_type} is not implemented")
                 if hasattr(m, "bias") and m.bias is not None:
                     nn.init.constant_(m.bias.data, 0.0)
             elif classname.find("BatchNorm2d") != -1:
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 00fc3ef4..40496145 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -66,19 +66,11 @@ def main(my_args=tuple(sys.argv[1:])):
     logging.info(f"Task mode: {mode}.")
 
     # Set train size to fine tune size if fine tune is set, else use custom train size
-    train_size = (
-        args.fine_tune
-        if args.fine_tune is not None
-        else args.samples
-        if args.samples is not None
-        else None
-    )
+    train_size = args.fine_tune if args.fine_tune is not None else args.samples if args.samples is not None else None
     # Whether to load weights from a previous run
     load_weights = evaluate or args.fine_tune is not None
 
-    pretrained_imputation_model = load_pretrained_imputation_model(
-        args.pretrained_imputation
-    )
+    pretrained_imputation_model = load_pretrained_imputation_model(args.pretrained_imputation)
     # Log imputation model to wandb
     update_wandb_config(
         {
@@ -92,11 +84,7 @@ def main(my_args=tuple(sys.argv[1:])):
     log_dir = (
         (log_dir_name / experiment)
         if experiment
-        else (
-            log_dir_name
-            / (args.task_name if args.task_name is not None else args.task)
-            / model
-        )
+        else (log_dir_name / (args.task_name if args.task_name is not None else args.task) / model)
     )
     log_full_line(f"Logging to {log_dir}.", logging.INFO)
 
@@ -119,9 +107,7 @@ def main(my_args=tuple(sys.argv[1:])):
     # Load pretrained model in evaluate mode or when finetuning
     if load_weights:
         if args.source_dir is None:
-            raise ValueError(
-                "Please specify a source directory when evaluating or fine-tuning."
-            )
+            raise ValueError("Please specify a source directory when evaluating or fine-tuning.")
         log_dir /= f"_from_{args.source_name}"
         name_datasets(args.source_name, args.source_name, args.name)
         if args.fine_tune:
@@ -129,13 +115,9 @@ def main(my_args=tuple(sys.argv[1:])):
             name_datasets(args.name, args.name, args.name)
         run_dir = create_run_dir(log_dir)
         source_dir = args.source_dir
-        logging.info(
-            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config."
-        )
+        logging.info(f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
         gin.parse_config_file(source_dir / "train_config.gin")
-    elif (
-        args.samples and args.source_dir is not None
-    ):  # Train model with limited samples and bind existing config
+    elif args.samples and args.source_dir is not None:  # Train model with limited samples and bind existing config
         logging.info("Binding train gin-config. Note: this might override your config.")
         gin.parse_config_file(args.source_dir / "train_config.gin")
         log_dir /= f"samples_{args.fine_tune}"
@@ -146,22 +128,14 @@ def main(my_args=tuple(sys.argv[1:])):
         name_datasets(args.name, args.name, args.name)
         hp_checkpoint = log_dir / args.hp_checkpoint if args.hp_checkpoint else None
         model_path = (
-            Path("configs")
-            / (
-                "imputation_models"
-                if mode == RunMode.imputation
-                else "prediction_models"
-            )
-            / f"{model}.gin"
+            Path("configs") / ("imputation_models" if mode == RunMode.imputation else "prediction_models") / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
             if args.experiment
             else [model_path, Path(f"configs/tasks/{task}.gin")]
         )
-        gin.parse_config_files_and_bindings(
-            gin_config_files, args.hyperparams, finalize_config=False
-        )
+        gin.parse_config_files_and_bindings(gin_config_files, args.hyperparams, finalize_config=False)
         log_full_line(f"Data directory: {data_dir.resolve()}", level=logging.INFO)
         run_dir = create_run_dir(log_dir)
         choose_and_bind_hyperparameters(
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 94f95c61..b23aae68 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -23,9 +23,7 @@ def build_parser() -> ArgumentParser:
     Returns:
         The configured ArgumentParser.
     """
-    parser = ArgumentParser(
-        description="Framework for benchmarking ML/DL models on ICU data"
-    )
+    parser = ArgumentParser(description="Framework for benchmarking ML/DL models on ICU data")
 
     parser.add_argument(
         "-d",
@@ -42,12 +40,8 @@ def build_parser() -> ArgumentParser:
         help="Name of the task gin.",
     )
     parser.add_argument("-n", "--name", help="Name of the (target) dataset.")
-    parser.add_argument(
-        "-tn", "--task-name", help="Name of the task, used for naming experiments."
-    )
-    parser.add_argument(
-        "-m", "--model", default="LGBMClassifier", help="Name of the model gin."
-    )
+    parser.add_argument("-tn", "--task-name", help="Name of the task, used for naming experiments.")
+    parser.add_argument("-m", "--model", default="LGBMClassifier", help="Name of the model gin.")
     parser.add_argument("-e", "--experiment", help="Name of the experiment gin.")
     parser.add_argument(
         "-l",
@@ -71,12 +65,8 @@ def build_parser() -> ArgumentParser:
         help="Set to log verbosly. Disable for clean logs.",
     )
     parser.add_argument("--cpu", default=False, action=BOA, help="Set to use CPU.")
-    parser.add_argument(
-        "-db", "--debug", default=False, action=BOA, help="Set to load less data."
-    )
-    parser.add_argument(
-        "--reproducible", default=True, action=BOA, help="Make torch reproducible."
-    )
+    parser.add_argument("-db", "--debug", default=False, action=BOA, help="Set to load less data.")
+    parser.add_argument("--reproducible", default=True, action=BOA, help="Make torch reproducible.")
     parser.add_argument(
         "-lc",
         "--load_cache",
@@ -91,9 +81,7 @@ def build_parser() -> ArgumentParser:
         action=BOA,
         help="Set to generate data cache.",
     )
-    parser.add_argument(
-        "-p", "--preprocessor", type=Path, help="Load custom preprocessor from file."
-    )
+    parser.add_argument("-p", "--preprocessor", type=Path, help="Load custom preprocessor from file.")
     parser.add_argument("-pl", "--plot", action=BOA, help="Generate common plots.")
     parser.add_argument(
         "-wd",
@@ -107,18 +95,10 @@ def build_parser() -> ArgumentParser:
         type=str,
         help="Path to pretrained imputation model.",
     )
-    parser.add_argument(
-        "-hp", "--hyperparams", nargs="+", help="Hyperparameters for model."
-    )
-    parser.add_argument(
-        "--tune", default=False, action=BOA, help="Find best hyperparameters."
-    )
-    parser.add_argument(
-        "--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint."
-    )
-    parser.add_argument(
-        "--eval", default=False, action=BOA, help="Only evaluate model, skip training."
-    )
+    parser.add_argument("-hp", "--hyperparams", nargs="+", help="Hyperparameters for model.")
+    parser.add_argument("--tune", default=False, action=BOA, help="Find best hyperparameters.")
+    parser.add_argument("--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint.")
+    parser.add_argument("--eval", default=False, action=BOA, help="Only evaluate model, skip training.")
     parser.add_argument(
         "--complete-train",
         default=False,
@@ -132,12 +112,8 @@ def build_parser() -> ArgumentParser:
         type=int,
         help="Finetune model with amount of train data.",
     )
-    parser.add_argument(
-        "-sn", "--source-name", type=Path, help="Name of the source dataset."
-    )
-    parser.add_argument(
-        "--source-dir", type=Path, help="Directory containing gin and model weights."
-    )
+    parser.add_argument("-sn", "--source-name", type=Path, help="Name of the source dataset.")
+    parser.add_argument("--source-dir", type=Path, help="Directory containing gin and model weights.")
     parser.add_argument(
         "-sa",
         "--samples",
@@ -185,21 +161,15 @@ def create_run_dir(log_dir: Path, randomly_searched_params: str = None) -> Path:
 
 def import_preprocessor(preprocessor_path: str):
     # Import custom supplied preprocessor
-    log_full_line(
-        f"Importing custom preprocessor from {preprocessor_path}.", logging.INFO
-    )
+    log_full_line(f"Importing custom preprocessor from {preprocessor_path}.", logging.INFO)
     try:
-        spec = importlib.util.spec_from_file_location(
-            "CustomPreprocessor", preprocessor_path
-        )
+        spec = importlib.util.spec_from_file_location("CustomPreprocessor", preprocessor_path)
         module = importlib.util.module_from_spec(spec)
         sys.modules["preprocessor"] = module
         spec.loader.exec_module(module)
         gin.bind_parameter("preprocess.preprocessor", module.CustomPreprocessor)
     except Exception as e:
-        logging.error(
-            f"Could not import custom preprocessor from {preprocessor_path}: {e}"
-        )
+        logging.error(f"Could not import custom preprocessor from {preprocessor_path}: {e}")
 
 
 def aggregate_results(log_dir: Path, execution_time: timedelta = None):
@@ -243,14 +213,10 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
     # Calculate the population standard deviation over aggregated results over folds/iterations
     # Divide by sqrt(n) to get standard deviation.
-    std_scores = {
-        metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()
-    }
+    std_scores = {metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()}
 
     confidence_interval = {
-        metric: (
-            stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list))
-        )
+        metric: (stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list)))
         for metric, list in list_scores.items()
     }
 
@@ -258,9 +224,7 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
         "avg": averaged_scores,
         "std": std_scores,
         "CI_0.95": confidence_interval,
-        "execution_time": execution_time.total_seconds()
-        if execution_time is not None
-        else 0.0,
+        "execution_time": execution_time.total_seconds() if execution_time is not None else 0.0,
     }
 
     with open(log_dir / "aggregated_test_metrics.json", "w") as f:
@@ -276,14 +240,10 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
 def name_datasets(train="default", val="default", test="default"):
     """Names the datasets for logging (optional)."""
-    gin.bind_parameter(
-        "train_common.dataset_names", {"train": train, "val": val, "test": test}
-    )
+    gin.bind_parameter("train_common.dataset_names", {"train": train, "val": val, "test": test})
 
 
-def log_full_line(
-    msg: str, level: int = logging.INFO, char: str = "-", num_newlines: int = 0
-):
+def log_full_line(msg: str, level: int = logging.INFO, char: str = "-", num_newlines: int = 0):
     """Logs a full line of a given character with a message centered.
 
     Args:
@@ -311,43 +271,24 @@ def load_pretrained_imputation_model(use_pretrained_imputation):
     Args:
         use_pretrained_imputation: Path to the pretrained imputation model.
     """
-    if (
-        use_pretrained_imputation is not None
-        and not Path(use_pretrained_imputation).exists()
-    ):
+    if use_pretrained_imputation is not None and not Path(use_pretrained_imputation).exists():
         logging.warning("The specified pretrained imputation model does not exist.")
         use_pretrained_imputation = None
 
     if use_pretrained_imputation is not None:
-        logging.info(
-            "Using pretrained imputation from" + str(use_pretrained_imputation)
-        )
-        pretrained_imputation_model_checkpoint = torch.load(
-            use_pretrained_imputation, map_location=torch.device("cpu")
-        )
+        logging.info("Using pretrained imputation from" + str(use_pretrained_imputation))
+        pretrained_imputation_model_checkpoint = torch.load(use_pretrained_imputation, map_location=torch.device("cpu"))
         if isinstance(pretrained_imputation_model_checkpoint, dict):
             imputation_model_class = pretrained_imputation_model_checkpoint["class"]
-            pretrained_imputation_model = imputation_model_class(
-                **pretrained_imputation_model_checkpoint["hyper_parameters"]
-            )
-            pretrained_imputation_model.set_trained_columns(
-                pretrained_imputation_model_checkpoint["trained_columns"]
-            )
-            pretrained_imputation_model.load_state_dict(
-                pretrained_imputation_model_checkpoint["state_dict"]
-            )
+            pretrained_imputation_model = imputation_model_class(**pretrained_imputation_model_checkpoint["hyper_parameters"])
+            pretrained_imputation_model.set_trained_columns(pretrained_imputation_model_checkpoint["trained_columns"])
+            pretrained_imputation_model.load_state_dict(pretrained_imputation_model_checkpoint["state_dict"])
         else:
             pretrained_imputation_model = pretrained_imputation_model_checkpoint
-        pretrained_imputation_model = pretrained_imputation_model.to(
-            "cuda" if torch.cuda.is_available() else "cpu"
-        )
+        pretrained_imputation_model = pretrained_imputation_model.to("cuda" if torch.cuda.is_available() else "cpu")
         try:
-            logging.info(
-                f"imputation model device: {next(pretrained_imputation_model.parameters()).device}"
-            )
-            pretrained_imputation_model.device = next(
-                pretrained_imputation_model.parameters()
-            ).device
+            logging.info(f"imputation model device: {next(pretrained_imputation_model.parameters()).device}")
+            pretrained_imputation_model.device = next(pretrained_imputation_model.parameters()).device
         except Exception as e:
             logging.debug(f"Could not set device of imputation model: {e}")
     else:
@@ -368,9 +309,7 @@ def setup_logging(date_format, log_format, verbose):
     logging.basicConfig(format=log_format, datefmt=date_format)
     loggers = ["pytorch_lightning", "lightning_fabric"]
     for logger in loggers:
-        logging.getLogger(logger).handlers[0].setFormatter(
-            logging.Formatter(log_format, datefmt=date_format)
-        )
+        logging.getLogger(logger).handlers[0].setFormatter(logging.Formatter(log_format, datefmt=date_format))
 
     if not verbose:
         logging.getLogger().setLevel(logging.INFO)

From 25fedafff7c49d699250df64fd8b18579bc7568f Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Mon, 2 Oct 2023 10:07:20 +0300
Subject: [PATCH 067/142] added predict_mode

---
 icu_benchmarks/data/loader.py | 41 ++++++++++++++++++++++++++---------
 1 file changed, 31 insertions(+), 10 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 0e46cc51..82807635 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -103,11 +103,15 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id].to_numpy()
-        labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
+        labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(
+            dtype=float
+        )
 
         if len(labels) == 1:
             # only one label per stay, align with window
-            labels = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, labels], axis=0)
+            labels = np.concatenate(
+                [np.empty(window.shape[0] - 1) * np.nan, labels], axis=0
+            )
 
         length_diff = self.maxlen - window.shape[0]
         pad_mask = np.ones(window.shape[0])
@@ -115,7 +119,9 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         # Padding the array to fulfill size requirement
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
-            window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
+            window = np.concatenate(
+                [window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0
+            )
             labels = np.concatenate([labels, np.ones(length_diff) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
 
@@ -325,8 +331,12 @@ def __init__(
         self.amputated_values, self.amputation_mask = ampute_data(
             self.features_df, mask_method, mask_proportion, mask_observation_proportion
         )
-        self.amputation_mask = (self.amputation_mask + self.features_df.isna().values).bool()
-        self.amputation_mask = DataFrame(self.amputation_mask, columns=self.vars[Segment.dynamic])
+        self.amputation_mask = (
+            self.amputation_mask + self.features_df.isna().values
+        ).bool()
+        self.amputation_mask = DataFrame(
+            self.amputation_mask, columns=self.vars[Segment.dynamic]
+        )
         self.amputation_mask[self.vars["GROUP"]] = self.features_df.index
         self.amputation_mask.set_index(self.vars["GROUP"], inplace=True)
 
@@ -351,9 +361,15 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
-        window_missingness_mask = self.target_missingness_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
-        amputated_window = self.amputated_values.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
-        amputation_mask = self.amputation_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
+        window_missingness_mask = self.target_missingness_mask.loc[
+            stay_id:stay_id, self.vars[Segment.dynamic]
+        ]
+        amputated_window = self.amputated_values.loc[
+            stay_id:stay_id, self.vars[Segment.dynamic]
+        ]
+        amputation_mask = self.amputation_mask.loc[
+            stay_id:stay_id, self.vars[Segment.dynamic]
+        ]
 
         return (
             from_numpy(amputated_window.values).to(float32),
@@ -452,14 +468,18 @@ def __init__(
         max_encoder_length: int = 24,
         **kwargs,
     ):
-        data[split]["FEATURES"]["time_idx"] = ((data[split]["FEATURES"]["time"] / pd.Timedelta(seconds=3600))).astype(
+        data[split]["FEATURES"]["time_idx"] = (
+            (data[split]["FEATURES"]["time"] / pd.Timedelta(seconds=3600))
+        ).astype(
             int
         )  # create an incremental column indicating the time step(required by constructor)
         data = data.get(split)  # get split
         labels = data["OUTCOME"]
         features = data["FEATURES"]
         self.name = name
-        self.data = pd.merge(labels, features, on=["stay_id", "time"])  # combine labels and features
+        self.data = pd.merge(
+            labels, features, on=["stay_id", "time"]
+        )  # combine labels and features
         # self.data["sex"].replace([0, 1], ["Female", "Male"], inplace=True)
         self.split = split
         self.args = args
@@ -582,6 +602,7 @@ def __init__(
             add_relative_time_idx=True,
             add_target_scales=True,
             add_encoder_length=True,
+            predict_mode=True,
         )
 
     def get_balance(self) -> list:

From e120465dd3bde19bd22522e0e0f06306b895bac5 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Mon, 2 Oct 2023 10:43:00 +0300
Subject: [PATCH 068/142] added pytorch flag

---
 icu_benchmarks/run.py                    | 44 +++++++++++++++++++-----
 icu_benchmarks/tuning/hyperparameters.py | 31 +++++++++++++----
 2 files changed, 60 insertions(+), 15 deletions(-)

diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 40496145..7302aec8 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -66,11 +66,19 @@ def main(my_args=tuple(sys.argv[1:])):
     logging.info(f"Task mode: {mode}.")
 
     # Set train size to fine tune size if fine tune is set, else use custom train size
-    train_size = args.fine_tune if args.fine_tune is not None else args.samples if args.samples is not None else None
+    train_size = (
+        args.fine_tune
+        if args.fine_tune is not None
+        else args.samples
+        if args.samples is not None
+        else None
+    )
     # Whether to load weights from a previous run
     load_weights = evaluate or args.fine_tune is not None
 
-    pretrained_imputation_model = load_pretrained_imputation_model(args.pretrained_imputation)
+    pretrained_imputation_model = load_pretrained_imputation_model(
+        args.pretrained_imputation
+    )
     # Log imputation model to wandb
     update_wandb_config(
         {
@@ -84,7 +92,11 @@ def main(my_args=tuple(sys.argv[1:])):
     log_dir = (
         (log_dir_name / experiment)
         if experiment
-        else (log_dir_name / (args.task_name if args.task_name is not None else args.task) / model)
+        else (
+            log_dir_name
+            / (args.task_name if args.task_name is not None else args.task)
+            / model
+        )
     )
     log_full_line(f"Logging to {log_dir}.", logging.INFO)
 
@@ -107,7 +119,9 @@ def main(my_args=tuple(sys.argv[1:])):
     # Load pretrained model in evaluate mode or when finetuning
     if load_weights:
         if args.source_dir is None:
-            raise ValueError("Please specify a source directory when evaluating or fine-tuning.")
+            raise ValueError(
+                "Please specify a source directory when evaluating or fine-tuning."
+            )
         log_dir /= f"_from_{args.source_name}"
         name_datasets(args.source_name, args.source_name, args.name)
         if args.fine_tune:
@@ -115,9 +129,13 @@ def main(my_args=tuple(sys.argv[1:])):
             name_datasets(args.name, args.name, args.name)
         run_dir = create_run_dir(log_dir)
         source_dir = args.source_dir
-        logging.info(f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
+        logging.info(
+            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config."
+        )
         gin.parse_config_file(source_dir / "train_config.gin")
-    elif args.samples and args.source_dir is not None:  # Train model with limited samples and bind existing config
+    elif (
+        args.samples and args.source_dir is not None
+    ):  # Train model with limited samples and bind existing config
         logging.info("Binding train gin-config. Note: this might override your config.")
         gin.parse_config_file(args.source_dir / "train_config.gin")
         log_dir /= f"samples_{args.fine_tune}"
@@ -128,14 +146,22 @@ def main(my_args=tuple(sys.argv[1:])):
         name_datasets(args.name, args.name, args.name)
         hp_checkpoint = log_dir / args.hp_checkpoint if args.hp_checkpoint else None
         model_path = (
-            Path("configs") / ("imputation_models" if mode == RunMode.imputation else "prediction_models") / f"{model}.gin"
+            Path("configs")
+            / (
+                "imputation_models"
+                if mode == RunMode.imputation
+                else "prediction_models"
+            )
+            / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
             if args.experiment
             else [model_path, Path(f"configs/tasks/{task}.gin")]
         )
-        gin.parse_config_files_and_bindings(gin_config_files, args.hyperparams, finalize_config=False)
+        gin.parse_config_files_and_bindings(
+            gin_config_files, args.hyperparams, finalize_config=False
+        )
         log_full_line(f"Data directory: {data_dir.resolve()}", level=logging.INFO)
         run_dir = create_run_dir(log_dir)
         choose_and_bind_hyperparameters(
@@ -149,6 +175,7 @@ def main(my_args=tuple(sys.argv[1:])):
             generate_cache=args.generate_cache,
             load_cache=args.load_cache,
             verbose=verbose,
+            pytorch_forecasting=pytorch_forecasting,
         )
 
     log_full_line(f"Logging to {run_dir.resolve()}", level=logging.INFO)
@@ -161,6 +188,7 @@ def main(my_args=tuple(sys.argv[1:])):
     log_full_line(mode_string, level=logging.INFO, char="=", num_newlines=3)
 
     start_time = datetime.now()
+
     execute_repeated_cv(
         data_dir,
         run_dir,
diff --git a/icu_benchmarks/tuning/hyperparameters.py b/icu_benchmarks/tuning/hyperparameters.py
index 212a7753..cf3d0c17 100644
--- a/icu_benchmarks/tuning/hyperparameters.py
+++ b/icu_benchmarks/tuning/hyperparameters.py
@@ -36,6 +36,7 @@ def choose_and_bind_hyperparameters(
     debug: bool = False,
     verbose: bool = False,
     wandb: bool = False,
+    pytorch_forecasting: bool = False,
 ):
     """Choose hyperparameters to tune and bind them to gin.
 
@@ -67,7 +68,9 @@ def choose_and_bind_hyperparameters(
         return
 
     # Collect hyperparameters.
-    hyperparams_bounds, hyperparams_names = collect_bound_hyperparameters(hyperparams, scopes)
+    hyperparams_bounds, hyperparams_names = collect_bound_hyperparameters(
+        hyperparams, scopes
+    )
 
     if do_tune and not hyperparams_bounds:
         logging.info("No hyperparameters to tune, skipping tuning.")
@@ -78,17 +81,26 @@ def choose_and_bind_hyperparameters(
     if checkpoint:
         checkpoint_path = checkpoint / checkpoint_file
         if not checkpoint_path.exists():
-            logging.warning(f"Hyperparameter checkpoint {checkpoint_path} does not exist.")
+            logging.warning(
+                f"Hyperparameter checkpoint {checkpoint_path} does not exist."
+            )
             logging.info("Attempting to find latest checkpoint file.")
             checkpoint_path = find_checkpoint(log_dir.parent, checkpoint_file)
         # Check if we found a checkpoint file
         if checkpoint_path:
-            n_calls, configuration, evaluation = load_checkpoint(checkpoint_path, n_calls)
+            n_calls, configuration, evaluation = load_checkpoint(
+                checkpoint_path, n_calls
+            )
             # Check if we surpassed maximum tuning iterations
             if n_calls <= 0:
-                logging.log(TUNE, "No more hyperparameter tuning iterations left, skipping tuning.")
+                logging.log(
+                    TUNE,
+                    "No more hyperparameter tuning iterations left, skipping tuning.",
+                )
                 logging.info("Training with these hyperparameters:")
-                bind_gin_params(hyperparams_names, configuration[np.argmin(evaluation)])  # bind best hyperparameters
+                bind_gin_params(
+                    hyperparams_names, configuration[np.argmin(evaluation)]
+                )  # bind best hyperparameters
                 return
         else:
             logging.warning("No checkpoint file found, starting from scratch.")
@@ -112,6 +124,7 @@ def bind_params_and_train(hyperparams):
                 debug=debug,
                 verbose=verbose,
                 wandb=wandb,
+                pytorch_forecasting=pytorch_forecasting,
             )
 
     header = ["ITERATION"] + hyperparams_names + ["LOSS AT ITERATION"]
@@ -126,7 +139,9 @@ def tune_step_callback(res):
             table_cells = [len(res.x_iters)] + res.x_iters[-1] + [res.func_vals[-1]]
             highlight = res.x_iters[-1] == res.x  # highlight if best so far
             log_table_row(header, TUNE)
-            log_table_row(table_cells, TUNE, align=Align.RIGHT, header=header, highlight=highlight)
+            log_table_row(
+                table_cells, TUNE, align=Align.RIGHT, header=header, highlight=highlight
+            )
             wandb_log({"hp-iteration": len(res.x_iters)})
 
     if do_tune:
@@ -141,7 +156,9 @@ def tune_step_callback(res):
         logging.log(TUNE, "Hyperparameter tuning disabled")
         if configuration:
             # We have loaded a checkpoint, use the best hyperparameters.
-            logging.info("Training with the best hyperparameters from loaded checkpoint:")
+            logging.info(
+                "Training with the best hyperparameters from loaded checkpoint:"
+            )
             bind_gin_params(hyperparams_names, configuration[np.argmin(evaluation)])
         else:
             logging.log(TUNE, "Choosing hyperparameters randomly from bounds.")

From b4d6b426a6e00096ad2d6bdafd6e7d9a0d6be116 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Mon, 2 Oct 2023 10:43:46 +0300
Subject: [PATCH 069/142] added LSTM from pytorch

---
 icu_benchmarks/models/dl_models.py | 95 +++++++++++++++++++++++++++---
 1 file changed, 87 insertions(+), 8 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 7070ffac..ab77b3ea 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -13,7 +13,7 @@
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor, FloatTensor
-from pytorch_forecasting import TemporalFusionTransformer  # , RecurrentNetwork
+from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork
 from pytorch_forecasting.metrics import QuantileLoss
 import matplotlib.pyplot as plt
 
@@ -149,7 +149,9 @@ def __init__(
             **kwargs,
         )
         hidden = hidden if hidden % 2 == 0 else hidden + 1  # Make sure hidden is even
-        self.input_embedding = nn.Linear(input_size[2], hidden)  # This acts as a time-distributed layer by defaults
+        self.input_embedding = nn.Linear(
+            input_size[2], hidden
+        )  # This acts as a time-distributed layer by defaults
         if pos_encoding:
             self.pos_encoder = PositionalEncoding(hidden)
         else:
@@ -220,7 +222,9 @@ def __init__(
         )
 
         hidden = hidden if hidden % 2 == 0 else hidden + 1  # Make sure hidden is even
-        self.input_embedding = nn.Linear(input_size[2], hidden)  # This acts as a time-distributed layer by defaults
+        self.input_embedding = nn.Linear(
+            input_size[2], hidden
+        )  # This acts as a time-distributed layer by defaults
         if pos_encoding:
             self.pos_encoder = PositionalEncoding(hidden)
         else:
@@ -286,9 +290,13 @@ def __init__(
 
         # We compute automatically the depth based on the desired seq_length.
         if isinstance(num_channels, Integral) and max_seq_length:
-            num_channels = [num_channels] * int(np.ceil(np.log(max_seq_length / 2) / np.log(kernel_size)))
+            num_channels = [num_channels] * int(
+                np.ceil(np.log(max_seq_length / 2) / np.log(kernel_size))
+            )
         elif isinstance(num_channels, Integral) and not max_seq_length:
-            raise Exception("a maximum sequence length needs to be provided if num_channels is int")
+            raise Exception(
+                "a maximum sequence length needs to be provided if num_channels is int"
+            )
 
         num_levels = len(num_channels)
         for i in range(num_levels):
@@ -529,13 +537,17 @@ def forward(self, x):
 
     def actual_vs_predictions_plot(self, dataloader):
         predictions = self.model.predict(dataloader, return_x=True)
-        predictions_vs_actuals = self.model.calculate_prediction_actual_by_variable(predictions.x, predictions.output)
+        predictions_vs_actuals = self.model.calculate_prediction_actual_by_variable(
+            predictions.x, predictions.output
+        )
         self.model.plot_prediction_actual_by_variable(predictions_vs_actuals)
         return predictions_vs_actuals
 
     def interpertations(self, dataloader, log_dir):
         raw_predictions = self.model.predict(dataloader, return_x=True, mode="raw")
-        interpretation = self.model.interpret_output(raw_predictions.output, reduction="sum")
+        interpretation = self.model.interpret_output(
+            raw_predictions.output, reduction="sum"
+        )
         figs = self.model.plot_interpretation(interpretation)
         for key, fig in figs.items():
             fig.savefig(log_dir / f"interpretation_{key}.png", bbox_inches="tight")
@@ -557,6 +569,73 @@ def predict_dependency(self, dataloader, variable, log_dir):
             q75=lambda x: x.quantile(0.75),
         )
         ax = agg_dependency.plot(y="median")
-        ax.fill_between(agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3)
+        ax.fill_between(
+            agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3
+        )
         plt.savefig(log_dir / "dependecy.png", bbox_inches="tight")
         return dependency
+
+
+@gin.configurable
+class RNNpytorch(DLPredictionWrapper):
+    """
+    Implementation of RNN from pytorch forecasting
+    """
+
+    supported_run_modes = [RunMode.classification, RunMode.regression]
+
+    def __init__(
+        self,
+        dataset,
+        hidden,
+        dropout,
+        optimizer,
+        num_classes,
+        cell_type,
+        rnn_layers,
+        *args,
+        **kwargs,
+    ):
+        super().__init__(optimizer=optimizer, pytorch_forecasting=True, *args, **kwargs)
+
+        self.model = RecurrentNetwork.from_dataset(
+            cell_type=cell_type,
+            rnn_layers=rnn_layers,
+            dataset=dataset,
+            hidden_size=hidden,
+            dropout=dropout,
+            optimizer=optimizer,
+        )
+
+        self.logit = nn.Linear(7, num_classes)
+
+    def set_weight(self, weight, dataset):
+        """
+        Set the weight for the loss function
+        """
+        if isinstance(weight, list):
+            weight = FloatTensor(weight)
+        elif weight == "balanced":
+            weight = FloatTensor(dataset.get_balance())
+        self.loss_weights = weight
+
+    def forward(
+        self,
+        tuple_x: tuple,
+    ) -> Dict[str, Tensor]:
+        x_dict = {
+            "encoder_cat": tuple_x[0],
+            "encoder_cont": tuple_x[1],
+            "encoder_target": tuple_x[2],
+            "encoder_lengths": tuple_x[3],
+            "decoder_cat": tuple_x[4],
+            "decoder_cont": tuple_x[5],
+            "decoder_target": tuple_x[6],
+            "decoder_lengths": tuple_x[7],
+            "decoder_time_idx": tuple_x[8],
+            "groups": tuple_x[9],
+            "target_scale": tuple_x[10],
+        }
+        out = self.model(x_dict)
+        pred = self.logit(out["prediction"])
+        return pred

From ec8346d15c8b478c15fc36c1148b96ee75dee30d Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Mon, 2 Oct 2023 13:27:10 +0300
Subject: [PATCH 070/142] added RNN implementation from pytorch forecasting and
 changed dataloader to fit both

---
 configs/prediction_models/RNNpytorch.gin | 222 +++++++++++++++++++++++
 configs/prediction_models/TFTpytorch.gin | 106 ++++++++++-
 icu_benchmarks/cross_validation.py       |   8 +-
 icu_benchmarks/data/loader.py            | 173 +++++++-----------
 icu_benchmarks/models/dl_models.py       |   5 +-
 icu_benchmarks/models/train.py           |  51 ++++--
 icu_benchmarks/models/wrappers.py        |  79 ++++++--
 7 files changed, 500 insertions(+), 144 deletions(-)
 create mode 100644 configs/prediction_models/RNNpytorch.gin

diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
new file mode 100644
index 00000000..0e47111f
--- /dev/null
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -0,0 +1,222 @@
+# Settings for TFT model.
+
+# Common settings for DL models
+include "configs/prediction_models/common/DLCommon.gin"
+
+# Optimizer params
+train_common.model = @RNNpytorch
+
+optimizer/hyperparameter.class_to_tune = @Adam
+optimizer/hyperparameter.weight_decay = 1e-6
+
+
+# Model params
+model/hyperparameter.class_to_tune = @RNNpytorch
+model/hyperparameter.hidden = 10
+model/hyperparameter.rnn_layers=2
+model/hyperparameter.num_classes = %NUM_CLASSES
+model/hyperparameter.cell_type='LSTM'
+model/hyperparameter.dropout = (0.0, 0.4)
+model/hyperparameter.lr_scheduler = "exponential"
+# Dataset params
+PredictionDatasetTFTpytorch.max_encoder_length = 24
+PredictionDatasetTFTpytorch.max_prediction_length = 24
+PredictionDatasetTFTpytorch.time_varying_known_reals=[]
+PredictionDatasetTFTpytorch.add_relative_time_idx=False
+PredictionDatasetTFTpytorch.target=[
+                "alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
+                "MissingIndicator_1",
+                "MissingIndicator_2",
+                "MissingIndicator_3",
+                "MissingIndicator_4",
+                "MissingIndicator_5",
+                "MissingIndicator_6",
+                "MissingIndicator_7",
+                "MissingIndicator_8",
+                "MissingIndicator_9",
+                "MissingIndicator_10",
+                "MissingIndicator_11",
+                "MissingIndicator_12",
+                "MissingIndicator_13",
+                "MissingIndicator_14",
+                "MissingIndicator_15",
+                "MissingIndicator_16",
+                "MissingIndicator_17",
+                "MissingIndicator_18",
+                "MissingIndicator_19",
+                "MissingIndicator_20",
+                "MissingIndicator_21",
+                "MissingIndicator_22",
+                "MissingIndicator_23",
+                "MissingIndicator_24",
+                "MissingIndicator_25",
+                "MissingIndicator_26",
+                "MissingIndicator_27",
+                "MissingIndicator_28",
+                "MissingIndicator_29",
+                "MissingIndicator_30",
+                "MissingIndicator_31",
+                "MissingIndicator_32",
+                "MissingIndicator_33",
+                "MissingIndicator_34",
+                "MissingIndicator_35",
+                "MissingIndicator_36",
+                "MissingIndicator_37",
+                "MissingIndicator_38",
+                "MissingIndicator_39",
+                "MissingIndicator_40",
+                "MissingIndicator_41",
+                "MissingIndicator_42",
+                "MissingIndicator_43",
+                "MissingIndicator_44",
+                "MissingIndicator_45",
+                "MissingIndicator_46",
+                "MissingIndicator_47",
+                "MissingIndicator_48",
+                "label",
+            ]
+PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
+                "MissingIndicator_1",
+                "MissingIndicator_2",
+                "MissingIndicator_3",
+                "MissingIndicator_4",
+                "MissingIndicator_5",
+                "MissingIndicator_6",
+                "MissingIndicator_7",
+                "MissingIndicator_8",
+                "MissingIndicator_9",
+                "MissingIndicator_10",
+                "MissingIndicator_11",
+                "MissingIndicator_12",
+                "MissingIndicator_13",
+                "MissingIndicator_14",
+                "MissingIndicator_15",
+                "MissingIndicator_16",
+                "MissingIndicator_17",
+                "MissingIndicator_18",
+                "MissingIndicator_19",
+                "MissingIndicator_20",
+                "MissingIndicator_21",
+                "MissingIndicator_22",
+                "MissingIndicator_23",
+                "MissingIndicator_24",
+                "MissingIndicator_25",
+                "MissingIndicator_26",
+                "MissingIndicator_27",
+                "MissingIndicator_28",
+                "MissingIndicator_29",
+                "MissingIndicator_30",
+                "MissingIndicator_31",
+                "MissingIndicator_32",
+                "MissingIndicator_33",
+                "MissingIndicator_34",
+                "MissingIndicator_35",
+                "MissingIndicator_36",
+                "MissingIndicator_37",
+                "MissingIndicator_38",
+                "MissingIndicator_39",
+                "MissingIndicator_40",
+                "MissingIndicator_41",
+                "MissingIndicator_42",
+                "MissingIndicator_43",
+                "MissingIndicator_44",
+                "MissingIndicator_45",
+                "MissingIndicator_46",
+                "MissingIndicator_47",
+                "MissingIndicator_48",
+                "label",
+            ]
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 13c7520a..63bf001d 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -10,7 +10,7 @@ optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
 
 
-# Encoder params
+# Model params
 model/hyperparameter.class_to_tune = @TFTpytorch
 model/hyperparameter.hidden = (16,32,64,128)
 model/hyperparameter.num_classes = %NUM_CLASSES
@@ -18,7 +18,107 @@ model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
 model/hyperparameter.n_heads =(1,2,4)
 model/hyperparameter.lr_scheduler = "exponential"
+# Dataset params
 PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
-
-
+PredictionDatasetTFTpytorch.target="label"
+PredictionDatasetTFTpytorch.time_varying_known_reals=["time_idx"]
+PredictionDatasetTFTpytorch.add_relative_time_idx=True
+PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
+                "MissingIndicator_1",
+                "MissingIndicator_2",
+                "MissingIndicator_3",
+                "MissingIndicator_4",
+                "MissingIndicator_5",
+                "MissingIndicator_6",
+                "MissingIndicator_7",
+                "MissingIndicator_8",
+                "MissingIndicator_9",
+                "MissingIndicator_10",
+                "MissingIndicator_11",
+                "MissingIndicator_12",
+                "MissingIndicator_13",
+                "MissingIndicator_14",
+                "MissingIndicator_15",
+                "MissingIndicator_16",
+                "MissingIndicator_17",
+                "MissingIndicator_18",
+                "MissingIndicator_19",
+                "MissingIndicator_20",
+                "MissingIndicator_21",
+                "MissingIndicator_22",
+                "MissingIndicator_23",
+                "MissingIndicator_24",
+                "MissingIndicator_25",
+                "MissingIndicator_26",
+                "MissingIndicator_27",
+                "MissingIndicator_28",
+                "MissingIndicator_29",
+                "MissingIndicator_30",
+                "MissingIndicator_31",
+                "MissingIndicator_32",
+                "MissingIndicator_33",
+                "MissingIndicator_34",
+                "MissingIndicator_35",
+                "MissingIndicator_36",
+                "MissingIndicator_37",
+                "MissingIndicator_38",
+                "MissingIndicator_39",
+                "MissingIndicator_40",
+                "MissingIndicator_41",
+                "MissingIndicator_42",
+                "MissingIndicator_43",
+                "MissingIndicator_44",
+                "MissingIndicator_45",
+                "MissingIndicator_46",
+                "MissingIndicator_47",
+                "MissingIndicator_48",
+                "label",
+            ]
\ No newline at end of file
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 412df3eb..bcf2f5f5 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -82,11 +82,15 @@ def execute_repeated_cv(
         cv_folds_to_train = 1
 
     else:
-        logging.info(f"Starting nested CV with {cv_repetitions_to_train} repetitions of {cv_folds_to_train} folds.")
+        logging.info(
+            f"Starting nested CV with {cv_repetitions_to_train} repetitions of {cv_folds_to_train} folds."
+        )
 
     for repetition in range(cv_repetitions_to_train):
         for fold_index in range(cv_folds_to_train):
-            repetition_fold_dir = log_dir / f"repetition_{repetition}" / f"fold_{fold_index}"
+            repetition_fold_dir = (
+                log_dir / f"repetition_{repetition}" / f"fold_{fold_index}"
+            )
             repetition_fold_dir.mkdir(parents=True, exist_ok=True)
 
             start_time = datetime.now()
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 82807635..a0c6f39b 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -6,7 +6,7 @@
 from torch import Tensor, cat, from_numpy, float32
 from torch.utils.data import Dataset
 import logging
-from typing import Dict, Tuple
+from typing import Dict, Tuple, Union
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
@@ -461,8 +461,12 @@ def __init__(
         self,
         data: dict,
         split: str,
+        time_varying_unknown_reals: List[str],
+        target: Union[str, List[str]],
+        time_varying_known_reals: List[str],
         *args,
         ram_cache: bool = False,
+        add_relative_time_idx: bool = False,
         name: str = "",
         max_prediction_length: int = 24,
         max_encoder_length: int = 24,
@@ -481,125 +485,82 @@ def __init__(
             labels, features, on=["stay_id", "time"]
         )  # combine labels and features
         # self.data["sex"].replace([0, 1], ["Female", "Male"], inplace=True)
+        # List of column names to convert from boolean to float
+        boolean_columns = [
+            "MissingIndicator_1",
+            "MissingIndicator_2",
+            "MissingIndicator_3",
+            "MissingIndicator_4",
+            "MissingIndicator_5",
+            "MissingIndicator_6",
+            "MissingIndicator_7",
+            "MissingIndicator_8",
+            "MissingIndicator_9",
+            "MissingIndicator_10",
+            "MissingIndicator_11",
+            "MissingIndicator_12",
+            "MissingIndicator_13",
+            "MissingIndicator_14",
+            "MissingIndicator_15",
+            "MissingIndicator_16",
+            "MissingIndicator_17",
+            "MissingIndicator_18",
+            "MissingIndicator_19",
+            "MissingIndicator_20",
+            "MissingIndicator_21",
+            "MissingIndicator_22",
+            "MissingIndicator_23",
+            "MissingIndicator_24",
+            "MissingIndicator_25",
+            "MissingIndicator_26",
+            "MissingIndicator_27",
+            "MissingIndicator_28",
+            "MissingIndicator_29",
+            "MissingIndicator_30",
+            "MissingIndicator_31",
+            "MissingIndicator_32",
+            "MissingIndicator_33",
+            "MissingIndicator_34",
+            "MissingIndicator_35",
+            "MissingIndicator_36",
+            "MissingIndicator_37",
+            "MissingIndicator_38",
+            "MissingIndicator_39",
+            "MissingIndicator_40",
+            "MissingIndicator_41",
+            "MissingIndicator_42",
+            "MissingIndicator_43",
+            "MissingIndicator_44",
+            "MissingIndicator_45",
+            "MissingIndicator_46",
+            "MissingIndicator_47",
+            "MissingIndicator_48",
+        ]
+
+        # Convert multiple columns from boolean to float
+        self.data[boolean_columns] = self.data[boolean_columns].astype(float)
         self.split = split
         self.args = args
         self.ram_cache = ram_cache
         self.kwargs = kwargs
         self.column_names = features.columns
+
         super().__init__(
             data=self.data,
             time_idx="time_idx",
-            target="label",
+            target=target,
             group_ids=["stay_id"],
             min_encoder_length=max_encoder_length,
             max_encoder_length=max_encoder_length,
             min_prediction_length=max_prediction_length,
             max_prediction_length=max_prediction_length,
             static_categoricals=[],
-            static_reals=["height", "weight", "age", "stay_id", "sex"],
-            # time_varying_known_categoricals=[],
-            time_varying_known_reals=["time_idx"],
-            # time_varying_unknown_categoricals=[],
-            time_varying_unknown_reals=[
-                "alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",
-                "MissingIndicator_1",
-                "MissingIndicator_2",
-                "MissingIndicator_3",
-                "MissingIndicator_4",
-                "MissingIndicator_5",
-                "MissingIndicator_6",
-                "MissingIndicator_7",
-                "MissingIndicator_8",
-                "MissingIndicator_9",
-                "MissingIndicator_10",
-                "MissingIndicator_11",
-                "MissingIndicator_12",
-                "MissingIndicator_13",
-                "MissingIndicator_14",
-                "MissingIndicator_15",
-                "MissingIndicator_16",
-                "MissingIndicator_17",
-                "MissingIndicator_18",
-                "MissingIndicator_19",
-                "MissingIndicator_20",
-                "MissingIndicator_21",
-                "MissingIndicator_22",
-                "MissingIndicator_23",
-                "MissingIndicator_24",
-                "MissingIndicator_25",
-                "MissingIndicator_26",
-                "MissingIndicator_27",
-                "MissingIndicator_28",
-                "MissingIndicator_29",
-                "MissingIndicator_30",
-                "MissingIndicator_31",
-                "MissingIndicator_32",
-                "MissingIndicator_33",
-                "MissingIndicator_34",
-                "MissingIndicator_35",
-                "MissingIndicator_36",
-                "MissingIndicator_37",
-                "MissingIndicator_38",
-                "MissingIndicator_39",
-                "MissingIndicator_40",
-                "MissingIndicator_41",
-                "MissingIndicator_42",
-                "MissingIndicator_43",
-                "MissingIndicator_44",
-                "MissingIndicator_45",
-                "MissingIndicator_46",
-                "MissingIndicator_47",
-                "MissingIndicator_48",
-                "label",
-            ],
-            add_relative_time_idx=True,
+            static_reals=["height", "weight", "age", "sex"],
+            time_varying_known_categoricals=[],
+            time_varying_known_reals=time_varying_known_reals,
+            time_varying_unknown_categoricals=[],
+            time_varying_unknown_reals=time_varying_unknown_reals,
+            add_relative_time_idx=add_relative_time_idx,
             add_target_scales=True,
             add_encoder_length=True,
             predict_mode=True,
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index ab77b3ea..99ce7cb9 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -637,5 +637,8 @@ def forward(
             "target_scale": tuple_x[10],
         }
         out = self.model(x_dict)
-        pred = self.logit(out["prediction"])
+        print(out)
+        print(out["prediction"][-1].shape)
+
+        pred = self.logit(out["prediction"][-1])
         return pred
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 1cca8970..9f7aadf9 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -27,7 +27,9 @@
 from collections import OrderedDict
 from captum.attr import IntegratedGradients
 
-cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
+cpu_core_count = (
+    len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
+)
 
 
 def assure_minimum_length(dataset):
@@ -100,15 +102,25 @@ def train_common(
     dataset_class = (
         ImputationDataset
         if mode == RunMode.imputation
-        else (PredictionDatasetTFTpytorch if (pytorch_forecasting) else PredictionDataset)
+        else (
+            PredictionDatasetTFTpytorch if (pytorch_forecasting) else PredictionDataset
+        )
     )
 
     logging.info(f"Logging to directory: {log_dir}.")
-    save_config_file(log_dir)  # We save the operative config before and also after training
+    save_config_file(
+        log_dir
+    )  # We save the operative config before and also after training
 
-    train_dataset = dataset_class(data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"])
-    val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"])
-    train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
+    train_dataset = dataset_class(
+        data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"]
+    )
+    val_dataset = dataset_class(
+        data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"]
+    )
+    train_dataset, val_dataset = assure_minimum_length(
+        train_dataset
+    ), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
     test_dataset = dataset_class(data, split=test_on, name=dataset_names["test"])
     test_dataset = assure_minimum_length(test_dataset)
@@ -119,7 +131,6 @@ def train_common(
         )
 
     logging.info(f"Using {num_workers} workers for data loading.")
-
     if pytorch_forecasting:
         train_loader = train_dataset.to_dataloader(
             train=True,
@@ -144,10 +155,14 @@ def train_common(
             shuffle=False,
         )
         if load_weights:
-            model = load_model(model, source_dir, pl_model=pl_model, train_dataset=train_dataset)
+            model = load_model(
+                model, source_dir, pl_model=pl_model, train_dataset=train_dataset
+            )
 
         else:
-            model = model(train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode)
+            model = model(
+                train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode
+            )
 
     else:
         train_loader = DataLoader(
@@ -188,7 +203,9 @@ def train_common(
         if load_weights:
             model = load_model(model, source_dir, pl_model=pl_model)
         else:
-            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
+            model = model(
+                optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
+            )
 
     model.set_weight(weight, train_dataset)
     model.set_trained_columns(train_dataset.get_feature_names())
@@ -215,7 +232,7 @@ def train_common(
         max_epochs=epochs if model.requires_backprop else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="auto" if not cpu else "cpu",
+        accelerator="cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic="warn" if reproducible else False,
         benchmark=not reproducible,
@@ -228,7 +245,9 @@ def train_common(
     if not eval_only:
         if model.requires_backprop:
             logging.info("Training DL model.")
-            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
+            trainer.fit(
+                model, train_dataloaders=train_loader, val_dataloaders=val_loader
+            )
             logging.info("Training complete.")
         else:
             logging.info("Training ML model.")
@@ -263,7 +282,9 @@ def train_common(
         print(attr)
         """
     model.set_weight("balanced", train_dataset)
-    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
+    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
+        "test/loss"
+    ]
     save_config_file(log_dir)
     return test_loss
 
@@ -281,7 +302,9 @@ def load_model(model, source_dir, pl_model=True, train_dataset=None):
                 return Exception(f"No weights to load at path : {source_dir}")
             if pl_model:
                 if train_dataset is not None:
-                    model = model.load_from_checkpoint(model_path, dataset=train_dataset)
+                    model = model.load_from_checkpoint(
+                        model_path, dataset=train_dataset
+                    )
 
                 else:
                     model = model.load_from_checkpoint(model_path)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 8bffe10f..074bd06a 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -20,7 +20,9 @@
 from icu_benchmarks.contants import RunMode
 
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
-gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
+gin.config.external_configurable(
+    nn.functional.cross_entropy, module="torch.nn.functional"
+)
 gin.config.external_configurable(nn.functional.mse_loss, module="torch.nn.functional")
 
 gin.config.external_configurable(mean_squared_error, module="sklearn.metrics")
@@ -85,7 +87,9 @@ def save_model(self, save_path, file_name, file_extension):
 
     def check_supported_runmode(self, runmode: RunMode):
         if runmode not in self._supported_run_modes:
-            raise ValueError(f"Runmode {runmode} not supported for {self.__class__.__name__}")
+            raise ValueError(
+                f"Runmode {runmode} not supported for {self.__class__.__name__}"
+            )
         return True
 
 
@@ -158,7 +162,9 @@ def finalize_step(self, step_prefix=""):
             self.log_dict(
                 {
                     f"{step_prefix}/{name}": (
-                        np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
+                        np.float32(metric.compute())
+                        if isinstance(metric.compute(), np.float64)
+                        else metric.compute()
                     )
                     for name, metric in self.metrics[step_prefix].items()
                     if "_Curve" not in name
@@ -199,7 +205,10 @@ def configure_optimizers(self):
 
     def on_test_epoch_start(self) -> None:
         self.metrics = {
-            step_name: {metric_name: metric() for metric_name, metric in self.set_metrics().items()}
+            step_name: {
+                metric_name: metric()
+                for metric_name, metric in self.set_metrics().items()
+            }
             for step_name in ["train", "val", "test"]
         }
         return super().on_test_epoch_start()
@@ -328,7 +337,8 @@ def step_fn(self, element, step_prefix=""):
                 isinstance(element[1], list)
             ):
             """
-            if self.pytorch_forecasting:  # check if the data loader is the one for the pytorch forecasring implementation
+            if self.pytorch_forecasting:
+                # check if the data loader is the one for the pytorch forecasring implementation
                 dic, labels = element[0], element[1][0]
                 data = (
                     dic["encoder_cat"],
@@ -343,6 +353,8 @@ def step_fn(self, element, step_prefix=""):
                     dic["groups"],
                     dic["target_scale"],
                 )
+                if isinstance(labels, list):
+                    labels = labels[0]
                 mask = torch.ones_like(labels).bool()
             else:
                 data, labels = element[0], (element[1]).to(self.device)
@@ -365,7 +377,9 @@ def step_fn(self, element, step_prefix=""):
             else:
                 data = data.float().to(self.device)
         else:
-            raise Exception("Loader should return either (data, label) or (data, label, mask)")
+            raise Exception(
+                "Loader should return either (data, label) or (data, label, mask)"
+            )
         out = self(data)
 
         # If aux_loss is present, it is returned as a tuple
@@ -374,20 +388,31 @@ def step_fn(self, element, step_prefix=""):
         else:
             aux_loss = 0
         # Get prediction and target
-        prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
+        prediction = (
+            torch.masked_select(out, mask.unsqueeze(-1))
+            .reshape(-1, out.shape[-1])
+            .to(self.device)
+        )
 
         target = torch.masked_select(labels, mask).to(self.device)
 
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
-            loss = self.loss(prediction, target.long(), weight=self.loss_weights.to(self.device)) + aux_loss
+            loss = (
+                self.loss(
+                    prediction, target.long(), weight=self.loss_weights.to(self.device)
+                )
+                + aux_loss
+            )
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
 
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
-            raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
+            raise ValueError(
+                f"Run mode {self.run_mode} not yet supported. Please implement it."
+            )
         transformed_output = self.output_transform((prediction, target))
 
         for key, value in self.metrics[step_prefix].items():
@@ -404,7 +429,9 @@ def step_fn(self, element, step_prefix=""):
                     value.update(transformed_output[0], transformed_output[1])
             else:
                 value.update(transformed_output)
-        self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
+        self.log(
+            f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True
+        )
         return loss
 
 
@@ -451,9 +478,15 @@ def set_metrics(self, labels):
 
         # Regression
         else:
-            if self.scaler is not None:  # We invert transform the labels and predictions if they were scaled.
-                self.output_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
-                self.label_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
+            if (
+                self.scaler is not None
+            ):  # We invert transform the labels and predictions if they were scaled.
+                self.output_transform = lambda x: self.scaler.inverse_transform(
+                    x.reshape(-1, 1)
+                )
+                self.label_transform = lambda x: self.scaler.inverse_transform(
+                    x.reshape(-1, 1)
+                )
             else:
                 self.output_transform = lambda x: x
                 self.label_transform = lambda x: x
@@ -488,7 +521,9 @@ def fit_model(self, train_data, train_labels, val_data, val_labels):
 
     def validation_step(self, val_dataset, _):
         val_rep, val_label = val_dataset.get_data_and_labels()
-        val_rep, val_label = torch.from_numpy(val_rep).to(self.device), torch.from_numpy(val_label).to(self.device)
+        val_rep, val_label = torch.from_numpy(val_rep).to(
+            self.device
+        ), torch.from_numpy(val_label).to(self.device)
         self.set_metrics(val_label)
 
         val_pred = self.predict(val_rep)
@@ -531,7 +566,9 @@ def log_metrics(self, label, pred, metric_type):
         self.log_dict(
             {
                 # MPS dependent type casting
-                f"{metric_type}/{name}": metric(self.label_transform(label), self.output_transform(pred))
+                f"{metric_type}/{name}": metric(
+                    self.label_transform(label), self.output_transform(pred)
+                )
                 if not self.mps
                 else metric(self.label_transform(label), self.output_transform(pred))
                 # Fore very metric
@@ -569,7 +606,9 @@ def set_model_args(self, model, *args, **kwargs):
         # Get passed keyword arguments
         arguments = locals()["kwargs"]
         # Get valid hyperparameters
-        hyperparams = {key: value for key, value in arguments.items() if key in possible_hps}
+        hyperparams = {
+            key: value for key, value in arguments.items() if key in possible_hps
+        }
         logging.debug(f"Creating model with: {hyperparams}.")
         return model(**hyperparams)
 
@@ -622,7 +661,9 @@ def set_metrics(self):
     def init_weights(self, init_type="normal", gain=0.02):
         def init_func(m):
             classname = m.__class__.__name__
-            if hasattr(m, "weight") and (classname.find("Conv") != -1 or classname.find("Linear") != -1):
+            if hasattr(m, "weight") and (
+                classname.find("Conv") != -1 or classname.find("Linear") != -1
+            ):
                 if init_type == ImputationInit.NORMAL:
                     nn.init.normal_(m.weight.data, 0.0, gain)
                 elif init_type == ImputationInit.XAVIER:
@@ -632,7 +673,9 @@ def init_func(m):
                 elif init_type == ImputationInit.ORTHOGONAL:
                     nn.init.orthogonal_(m.weight.data, gain=gain)
                 else:
-                    raise NotImplementedError(f"Initialization method {init_type} is not implemented")
+                    raise NotImplementedError(
+                        f"Initialization method {init_type} is not implemented"
+                    )
                 if hasattr(m, "bias") and m.bias is not None:
                     nn.init.constant_(m.bias.data, 0.0)
             elif classname.find("BatchNorm2d") != -1:

From e3f6058d7f6238e9b0860e00c4311c8fbb82419b Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Mon, 2 Oct 2023 13:28:06 +0300
Subject: [PATCH 071/142] added RNN implementation from pytorch forecasting and
 changed dataloader to fit both

---
 icu_benchmarks/models/wrappers.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 074bd06a..bdb75f93 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -340,6 +340,8 @@ def step_fn(self, element, step_prefix=""):
             if self.pytorch_forecasting:
                 # check if the data loader is the one for the pytorch forecasring implementation
                 dic, labels = element[0], element[1][0]
+                if isinstance(labels, list):
+                    labels = labels[-1]
                 data = (
                     dic["encoder_cat"],
                     dic["encoder_cont"],
@@ -353,8 +355,7 @@ def step_fn(self, element, step_prefix=""):
                     dic["groups"],
                     dic["target_scale"],
                 )
-                if isinstance(labels, list):
-                    labels = labels[0]
+
                 mask = torch.ones_like(labels).bool()
             else:
                 data, labels = element[0], (element[1]).to(self.device)

From 6c05b225fbaf004687cd2fda3b20cb0dcc29b6fa Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 3 Oct 2023 15:49:15 +0300
Subject: [PATCH 072/142] changes to get RNN to work

---
 icu_benchmarks/data/loader.py | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index a0c6f39b..3eea2698 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -535,10 +535,13 @@ def __init__(
             "MissingIndicator_46",
             "MissingIndicator_47",
             "MissingIndicator_48",
+            "label",
         ]
 
         # Convert multiple columns from boolean to float
-        self.data[boolean_columns] = self.data[boolean_columns].astype(float)
+        self.data[boolean_columns] = self.data[boolean_columns].astype(
+            float
+        )  # changing boolean to floats to allow input to models
         self.split = split
         self.args = args
         self.ram_cache = ram_cache
@@ -572,8 +575,10 @@ def get_balance(self) -> list:
         Returns:
             Weights for each label.
         """
-
-        counts = self.data["target"][0].unique(return_counts=True)
+        if len(self.data["target"]) == 1:
+            counts = self.data["target"][0].unique(return_counts=True)
+        else:
+            counts = self.data["target"][-1].unique(return_counts=True)
 
         return list((1 / counts[1]) * counts[1].sum() / counts[0].shape[0])
 

From 5c3d6a3b4487d2a5d0e1dc2bca2436bcfdd5a22a Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 3 Oct 2023 15:50:12 +0300
Subject: [PATCH 073/142] changes to get RNN to work

---
 configs/prediction_models/common/DLCommon.gin |  2 +-
 icu_benchmarks/models/dl_models.py            | 37 ++++++-------------
 icu_benchmarks/models/train.py                | 10 ++++-
 icu_benchmarks/models/wrappers.py             |  5 +--
 4 files changed, 24 insertions(+), 30 deletions(-)

diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 7c72b17a..99251ea2 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,7 +15,7 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-train_common.batch_size = 256
+train_common.batch_size = 64
 train_common.patience = 10
 train_common.min_delta = 1e-4
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 99ce7cb9..ca491c13 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -12,7 +12,7 @@
 )
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor, FloatTensor
+from torch import Tensor, FloatTensor, isnan, isfinite
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork
 from pytorch_forecasting.metrics import QuantileLoss
 import matplotlib.pyplot as plt
@@ -497,16 +497,6 @@ def __init__(
 
         self.logit = nn.Linear(7, num_classes)
 
-    def set_weight(self, weight, dataset):
-        """
-        Set the weight for the loss function
-        """
-        if isinstance(weight, list):
-            weight = FloatTensor(weight)
-        elif weight == "balanced":
-            weight = FloatTensor(dataset.get_balance())
-        self.loss_weights = weight
-
     def forward(
         self,
         tuple_x: tuple,
@@ -593,6 +583,7 @@ def __init__(
         num_classes,
         cell_type,
         rnn_layers,
+        batch_size,
         *args,
         **kwargs,
     ):
@@ -607,17 +598,7 @@ def __init__(
             optimizer=optimizer,
         )
 
-        self.logit = nn.Linear(7, num_classes)
-
-    def set_weight(self, weight, dataset):
-        """
-        Set the weight for the loss function
-        """
-        if isinstance(weight, list):
-            weight = FloatTensor(weight)
-        elif weight == "balanced":
-            weight = FloatTensor(dataset.get_balance())
-        self.loss_weights = weight
+        self.logit = nn.Linear(batch_size * 24, num_classes)
 
     def forward(
         self,
@@ -637,8 +618,14 @@ def forward(
             "target_scale": tuple_x[10],
         }
         out = self.model(x_dict)
-        print(out)
-        print(out["prediction"][-1].shape)
+        out = out["prediction"][-1].reshape(1, -1)
+        nan_count = isnan(out).sum().item()
+
+        # Count the number of non-NaN (finite) values in the tensor
+        non_nan_count = isfinite(out).sum().item()
+
+        print("Number of NaN values in the tensor:", nan_count)
+        print("Number of non-NaN values in the tensor:", non_nan_count)
+        pred = self.logit(out)
 
-        pred = self.logit(out["prediction"][-1])
         return pred
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 9f7aadf9..a7bb5e7e 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -98,6 +98,8 @@ def train_common(
         num_workers: Number of workers to use for data loading.
     """
     logging.info(f"Training model: {model.__name__}.")
+    with open("data.pkl", "wb") as f:
+        pickle.dump(data, f)
     # choose dataset_class based on the model
     dataset_class = (
         ImputationDataset
@@ -161,7 +163,11 @@ def train_common(
 
         else:
             model = model(
-                train_dataset, optimizer=optimizer, epochs=epochs, run_mode=mode
+                train_dataset,
+                optimizer=optimizer,
+                epochs=epochs,
+                run_mode=mode,
+                batch_size=batch_size,  # check why validation set is needed here
             )
 
     else:
@@ -241,6 +247,8 @@ def train_common(
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
         check_val_every_n_epoch=1,
+        gradient_clip_val=0.5,
+        gradient_clip_algorithm="value",
     )
     if not eval_only:
         if model.requires_backprop:
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index bdb75f93..163ca900 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -270,6 +270,7 @@ def set_weight(self, weight, dataset):
             weight = FloatTensor(weight).to(self.device)
         elif weight == "balanced":
             weight = FloatTensor(dataset.get_balance()).to(self.device)
+
         self.loss_weights = weight
 
     def set_metrics(self, *args):
@@ -340,6 +341,7 @@ def step_fn(self, element, step_prefix=""):
             if self.pytorch_forecasting:
                 # check if the data loader is the one for the pytorch forecasring implementation
                 dic, labels = element[0], element[1][0]
+
                 if isinstance(labels, list):
                     labels = labels[-1]
                 data = (
@@ -394,9 +396,7 @@ def step_fn(self, element, step_prefix=""):
             .reshape(-1, out.shape[-1])
             .to(self.device)
         )
-
         target = torch.masked_select(labels, mask).to(self.device)
-
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
             loss = (
@@ -408,7 +408,6 @@ def step_fn(self, element, step_prefix=""):
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
-
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
             raise ValueError(

From a5d2a9981c6281858edff2677101d03de3670e3b Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 3 Oct 2023 16:06:39 +0300
Subject: [PATCH 074/142] changes to get RNN to work

---
 icu_benchmarks/models/train.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index a7bb5e7e..23744ee5 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -150,7 +150,7 @@ def train_common(
         )
         test_loader = test_dataset.to_dataloader(
             train=False,
-            batch_size=min(batch_size * 4, len(test_dataset)),
+            batch_size=batch_size,
             num_workers=num_workers,
             pin_memory=False,
             drop_last=True,
@@ -238,7 +238,7 @@ def train_common(
         max_epochs=epochs if model.requires_backprop else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="cpu",
+        accelerator="auto" if not cpu else "cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic="warn" if reproducible else False,
         benchmark=not reproducible,

From 56b0deb465ca90c21d5d05c8c0ef5bcbf9470255 Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Thu, 5 Oct 2023 17:26:04 +0200
Subject: [PATCH 075/142] removed grad clipping

---
 configs/prediction_models/RNNpytorch.gin |  4 ++--
 icu_benchmarks/models/train.py           | 12 +++++-------
 2 files changed, 7 insertions(+), 9 deletions(-)

diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 0e47111f..74ca1ef3 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -12,8 +12,8 @@ optimizer/hyperparameter.weight_decay = 1e-6
 
 # Model params
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = 10
-model/hyperparameter.rnn_layers=2
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
 model/hyperparameter.dropout = (0.0, 0.4)
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 23744ee5..f8cae6af 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -158,7 +158,7 @@ def train_common(
         )
         if load_weights:
             model = load_model(
-                model, source_dir, pl_model=pl_model, train_dataset=train_dataset
+                model, source_dir, pl_model=pl_model, train_dataset=train_dataset, optimizer=optimizer
             )
 
         else:
@@ -247,8 +247,6 @@ def train_common(
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
         check_val_every_n_epoch=1,
-        gradient_clip_val=0.5,
-        gradient_clip_algorithm="value",
     )
     if not eval_only:
         if model.requires_backprop:
@@ -272,8 +270,8 @@ def train_common(
         # print("1", actual_vs_predictions)
         interperations = model.interpertations(test_loader, log_dir)
         print("2", interperations)
-        pd = model.predict_dependency(test_loader, "age", log_dir)
-        print("3", pd)
+        # pd = model.predict_dependency(test_loader, "age", log_dir)
+        # print("3", pd)
         """
         test = next(iter(test_loader))
 
@@ -297,7 +295,7 @@ def train_common(
     return test_loss
 
 
-def load_model(model, source_dir, pl_model=True, train_dataset=None):
+def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=None):
     if source_dir.exists():
         if model.requires_backprop:
             if (source_dir / "model.ckpt").exists():
@@ -311,7 +309,7 @@ def load_model(model, source_dir, pl_model=True, train_dataset=None):
             if pl_model:
                 if train_dataset is not None:
                     model = model.load_from_checkpoint(
-                        model_path, dataset=train_dataset
+                        model_path, dataset=train_dataset, optimizer=optimizer
                     )
 
                 else:

From 8880adef1c76a9c3a98f14896e51cc0e0388f348 Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Mon, 9 Oct 2023 11:13:59 +0200
Subject: [PATCH 076/142] small edits

---
 configs/prediction_models/RNNpytorch.gin      | 5 ++++-
 configs/prediction_models/TFTpytorch.gin      | 2 ++
 configs/prediction_models/common/DLCommon.gin | 4 ++--
 configs/prediction_models/common/DLTuning.gin | 2 +-
 icu_benchmarks/models/dl_models.py            | 6 ------
 icu_benchmarks/models/train.py                | 1 -
 6 files changed, 9 insertions(+), 11 deletions(-)

diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 74ca1ef3..63c1de4e 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -11,13 +11,16 @@ optimizer/hyperparameter.weight_decay = 1e-6
 
 
 # Model params
+
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.hidden = (32, 128, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.lr_scheduler = "exponential"
+
+
 # Dataset params
 PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 63bf001d..a067fc20 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -18,6 +18,8 @@ model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
 model/hyperparameter.n_heads =(1,2,4)
 model/hyperparameter.lr_scheduler = "exponential"
+
+
 # Dataset params
 PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 99251ea2..2cfadcfe 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,8 +15,8 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-train_common.batch_size = 64
-train_common.patience = 10
+train_common.batch_size = 256
+train_common.patience = 20
 train_common.min_delta = 1e-4
 
 # Hyperparameter tuning settings
diff --git a/configs/prediction_models/common/DLTuning.gin b/configs/prediction_models/common/DLTuning.gin
index 1a2b7bb4..b4d13e12 100644
--- a/configs/prediction_models/common/DLTuning.gin
+++ b/configs/prediction_models/common/DLTuning.gin
@@ -2,4 +2,4 @@
 tune_hyperparameters.scopes = ["model", "optimizer"]
 tune_hyperparameters.n_initial_points = 5
 tune_hyperparameters.n_calls = 30
-tune_hyperparameters.folds_to_tune_on = 1
\ No newline at end of file
+tune_hyperparameters.folds_to_tune_on = 2
\ No newline at end of file
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index ca491c13..0646e2f2 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -619,13 +619,7 @@ def forward(
         }
         out = self.model(x_dict)
         out = out["prediction"][-1].reshape(1, -1)
-        nan_count = isnan(out).sum().item()
 
-        # Count the number of non-NaN (finite) values in the tensor
-        non_nan_count = isfinite(out).sum().item()
-
-        print("Number of NaN values in the tensor:", nan_count)
-        print("Number of non-NaN values in the tensor:", non_nan_count)
         pred = self.logit(out)
 
         return pred
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index f8cae6af..b7192370 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -246,7 +246,6 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
-        check_val_every_n_epoch=1,
     )
     if not eval_only:
         if model.requires_backprop:

From 869a6112431cd75f04176215f5ce9c6588e8a14e Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Mon, 9 Oct 2023 12:13:33 +0200
Subject: [PATCH 077/142] changed 0/1 to categorical vars

---
 configs/prediction_models/RNNpytorch.gin |  99 +++++++++++-----------
 configs/prediction_models/TFTpytorch.gin | 101 ++++++++++++-----------
 icu_benchmarks/data/loader.py            |   7 +-
 icu_benchmarks/models/train.py           |   3 +-
 4 files changed, 107 insertions(+), 103 deletions(-)

diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 63c1de4e..ad049c6d 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -125,55 +125,7 @@ PredictionDatasetTFTpytorch.target=[
                 "MissingIndicator_48",
                 "label",
             ]
-PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",
-                "MissingIndicator_1",
+PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=["MissingIndicator_1",
                 "MissingIndicator_2",
                 "MissingIndicator_3",
                 "MissingIndicator_4",
@@ -223,3 +175,52 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "MissingIndicator_48",
                 "label",
             ]
+PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",]
+                
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index a067fc20..442d44e8 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -26,55 +26,7 @@ PredictionDatasetTFTpytorch.max_prediction_length = 24
 PredictionDatasetTFTpytorch.target="label"
 PredictionDatasetTFTpytorch.time_varying_known_reals=["time_idx"]
 PredictionDatasetTFTpytorch.add_relative_time_idx=True
-PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",
-                "MissingIndicator_1",
+PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=["MissingIndicator_1",
                 "MissingIndicator_2",
                 "MissingIndicator_3",
                 "MissingIndicator_4",
@@ -123,4 +75,53 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "MissingIndicator_47",
                 "MissingIndicator_48",
                 "label",
-            ]
\ No newline at end of file
+            ]
+PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",]
+                
\ No newline at end of file
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 3eea2698..e8e8aaa4 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -464,6 +464,7 @@ def __init__(
         time_varying_unknown_reals: List[str],
         target: Union[str, List[str]],
         time_varying_known_reals: List[str],
+        time_varying_unknown_categoricals: List[str],
         *args,
         ram_cache: bool = False,
         add_relative_time_idx: bool = False,
@@ -486,6 +487,7 @@ def __init__(
         )  # combine labels and features
         # self.data["sex"].replace([0, 1], ["Female", "Male"], inplace=True)
         # List of column names to convert from boolean to float
+
         boolean_columns = [
             "MissingIndicator_1",
             "MissingIndicator_2",
@@ -540,8 +542,9 @@ def __init__(
 
         # Convert multiple columns from boolean to float
         self.data[boolean_columns] = self.data[boolean_columns].astype(
-            float
+            str
         )  # changing boolean to floats to allow input to models
+
         self.split = split
         self.args = args
         self.ram_cache = ram_cache
@@ -561,7 +564,7 @@ def __init__(
             static_reals=["height", "weight", "age", "sex"],
             time_varying_known_categoricals=[],
             time_varying_known_reals=time_varying_known_reals,
-            time_varying_unknown_categoricals=[],
+            time_varying_unknown_categoricals=time_varying_unknown_categoricals,
             time_varying_unknown_reals=time_varying_unknown_reals,
             add_relative_time_idx=add_relative_time_idx,
             add_target_scales=True,
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index b7192370..c1d96240 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -98,8 +98,7 @@ def train_common(
         num_workers: Number of workers to use for data loading.
     """
     logging.info(f"Training model: {model.__name__}.")
-    with open("data.pkl", "wb") as f:
-        pickle.dump(data, f)
+
     # choose dataset_class based on the model
     dataset_class = (
         ImputationDataset

From ce0831cb53229fffab73c97f5e6ad3997f78036a Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Mon, 9 Oct 2023 12:19:53 +0200
Subject: [PATCH 078/142] changed input of rnn

---
 configs/prediction_models/RNNpytorch.gin | 99 ++++++++++++------------
 1 file changed, 49 insertions(+), 50 deletions(-)

diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index ad049c6d..63c1de4e 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -125,7 +125,55 @@ PredictionDatasetTFTpytorch.target=[
                 "MissingIndicator_48",
                 "label",
             ]
-PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=["MissingIndicator_1",
+PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
+                "MissingIndicator_1",
                 "MissingIndicator_2",
                 "MissingIndicator_3",
                 "MissingIndicator_4",
@@ -175,52 +223,3 @@ PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=["MissingIndicator
                 "MissingIndicator_48",
                 "label",
             ]
-PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",]
-                

From 2828a3ac359b81a8e98061d9f1954a81bb8a998b Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Tue, 10 Oct 2023 16:06:50 +0300
Subject: [PATCH 079/142] latest changes

---
 configs/prediction_models/TFTpytorch.gin      |  54 +-----
 configs/prediction_models/common/DLCommon.gin |   4 +-
 icu_benchmarks/data/loader.py                 | 155 ++++++++++++------
 icu_benchmarks/models/dl_models.py            |  29 ++++
 icu_benchmarks/models/train.py                |  44 +++--
 5 files changed, 168 insertions(+), 118 deletions(-)

diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 442d44e8..68290c1d 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -26,56 +26,7 @@ PredictionDatasetTFTpytorch.max_prediction_length = 24
 PredictionDatasetTFTpytorch.target="label"
 PredictionDatasetTFTpytorch.time_varying_known_reals=["time_idx"]
 PredictionDatasetTFTpytorch.add_relative_time_idx=True
-PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=["MissingIndicator_1",
-                "MissingIndicator_2",
-                "MissingIndicator_3",
-                "MissingIndicator_4",
-                "MissingIndicator_5",
-                "MissingIndicator_6",
-                "MissingIndicator_7",
-                "MissingIndicator_8",
-                "MissingIndicator_9",
-                "MissingIndicator_10",
-                "MissingIndicator_11",
-                "MissingIndicator_12",
-                "MissingIndicator_13",
-                "MissingIndicator_14",
-                "MissingIndicator_15",
-                "MissingIndicator_16",
-                "MissingIndicator_17",
-                "MissingIndicator_18",
-                "MissingIndicator_19",
-                "MissingIndicator_20",
-                "MissingIndicator_21",
-                "MissingIndicator_22",
-                "MissingIndicator_23",
-                "MissingIndicator_24",
-                "MissingIndicator_25",
-                "MissingIndicator_26",
-                "MissingIndicator_27",
-                "MissingIndicator_28",
-                "MissingIndicator_29",
-                "MissingIndicator_30",
-                "MissingIndicator_31",
-                "MissingIndicator_32",
-                "MissingIndicator_33",
-                "MissingIndicator_34",
-                "MissingIndicator_35",
-                "MissingIndicator_36",
-                "MissingIndicator_37",
-                "MissingIndicator_38",
-                "MissingIndicator_39",
-                "MissingIndicator_40",
-                "MissingIndicator_41",
-                "MissingIndicator_42",
-                "MissingIndicator_43",
-                "MissingIndicator_44",
-                "MissingIndicator_45",
-                "MissingIndicator_46",
-                "MissingIndicator_47",
-                "MissingIndicator_48",
-                "label",
-            ]
+PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
 PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "alp",
                 "alt",
@@ -123,5 +74,6 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc",]
+                "wbc",
+                "label",]
                 
\ No newline at end of file
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 2cfadcfe..c1600089 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -14,8 +14,8 @@ base_regression_preprocessor.generate_features = False
 
 # Train params
 train_common.optimizer = @Adam
-train_common.epochs = 1000
-train_common.batch_size = 256
+train_common.epochs = 1
+train_common.batch_size = 64
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index e8e8aaa4..1de09cd2 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -487,58 +487,109 @@ def __init__(
         )  # combine labels and features
         # self.data["sex"].replace([0, 1], ["Female", "Male"], inplace=True)
         # List of column names to convert from boolean to float
-
-        boolean_columns = [
-            "MissingIndicator_1",
-            "MissingIndicator_2",
-            "MissingIndicator_3",
-            "MissingIndicator_4",
-            "MissingIndicator_5",
-            "MissingIndicator_6",
-            "MissingIndicator_7",
-            "MissingIndicator_8",
-            "MissingIndicator_9",
-            "MissingIndicator_10",
-            "MissingIndicator_11",
-            "MissingIndicator_12",
-            "MissingIndicator_13",
-            "MissingIndicator_14",
-            "MissingIndicator_15",
-            "MissingIndicator_16",
-            "MissingIndicator_17",
-            "MissingIndicator_18",
-            "MissingIndicator_19",
-            "MissingIndicator_20",
-            "MissingIndicator_21",
-            "MissingIndicator_22",
-            "MissingIndicator_23",
-            "MissingIndicator_24",
-            "MissingIndicator_25",
-            "MissingIndicator_26",
-            "MissingIndicator_27",
-            "MissingIndicator_28",
-            "MissingIndicator_29",
-            "MissingIndicator_30",
-            "MissingIndicator_31",
-            "MissingIndicator_32",
-            "MissingIndicator_33",
-            "MissingIndicator_34",
-            "MissingIndicator_35",
-            "MissingIndicator_36",
-            "MissingIndicator_37",
-            "MissingIndicator_38",
-            "MissingIndicator_39",
-            "MissingIndicator_40",
-            "MissingIndicator_41",
-            "MissingIndicator_42",
-            "MissingIndicator_43",
-            "MissingIndicator_44",
-            "MissingIndicator_45",
-            "MissingIndicator_46",
-            "MissingIndicator_47",
-            "MissingIndicator_48",
-            "label",
-        ]
+        if self.data["label"].dtype == "bool":
+            boolean_columns = [
+                "MissingIndicator_1",
+                "MissingIndicator_2",
+                "MissingIndicator_3",
+                "MissingIndicator_4",
+                "MissingIndicator_5",
+                "MissingIndicator_6",
+                "MissingIndicator_7",
+                "MissingIndicator_8",
+                "MissingIndicator_9",
+                "MissingIndicator_10",
+                "MissingIndicator_11",
+                "MissingIndicator_12",
+                "MissingIndicator_13",
+                "MissingIndicator_14",
+                "MissingIndicator_15",
+                "MissingIndicator_16",
+                "MissingIndicator_17",
+                "MissingIndicator_18",
+                "MissingIndicator_19",
+                "MissingIndicator_20",
+                "MissingIndicator_21",
+                "MissingIndicator_22",
+                "MissingIndicator_23",
+                "MissingIndicator_24",
+                "MissingIndicator_25",
+                "MissingIndicator_26",
+                "MissingIndicator_27",
+                "MissingIndicator_28",
+                "MissingIndicator_29",
+                "MissingIndicator_30",
+                "MissingIndicator_31",
+                "MissingIndicator_32",
+                "MissingIndicator_33",
+                "MissingIndicator_34",
+                "MissingIndicator_35",
+                "MissingIndicator_36",
+                "MissingIndicator_37",
+                "MissingIndicator_38",
+                "MissingIndicator_39",
+                "MissingIndicator_40",
+                "MissingIndicator_41",
+                "MissingIndicator_42",
+                "MissingIndicator_43",
+                "MissingIndicator_44",
+                "MissingIndicator_45",
+                "MissingIndicator_46",
+                "MissingIndicator_47",
+                "MissingIndicator_48",
+                "label",
+            ]
+        else:
+            boolean_columns = [
+                "MissingIndicator_1",
+                "MissingIndicator_2",
+                "MissingIndicator_3",
+                "MissingIndicator_4",
+                "MissingIndicator_5",
+                "MissingIndicator_6",
+                "MissingIndicator_7",
+                "MissingIndicator_8",
+                "MissingIndicator_9",
+                "MissingIndicator_10",
+                "MissingIndicator_11",
+                "MissingIndicator_12",
+                "MissingIndicator_13",
+                "MissingIndicator_14",
+                "MissingIndicator_15",
+                "MissingIndicator_16",
+                "MissingIndicator_17",
+                "MissingIndicator_18",
+                "MissingIndicator_19",
+                "MissingIndicator_20",
+                "MissingIndicator_21",
+                "MissingIndicator_22",
+                "MissingIndicator_23",
+                "MissingIndicator_24",
+                "MissingIndicator_25",
+                "MissingIndicator_26",
+                "MissingIndicator_27",
+                "MissingIndicator_28",
+                "MissingIndicator_29",
+                "MissingIndicator_30",
+                "MissingIndicator_31",
+                "MissingIndicator_32",
+                "MissingIndicator_33",
+                "MissingIndicator_34",
+                "MissingIndicator_35",
+                "MissingIndicator_36",
+                "MissingIndicator_37",
+                "MissingIndicator_38",
+                "MissingIndicator_39",
+                "MissingIndicator_40",
+                "MissingIndicator_41",
+                "MissingIndicator_42",
+                "MissingIndicator_43",
+                "MissingIndicator_44",
+                "MissingIndicator_45",
+                "MissingIndicator_46",
+                "MissingIndicator_47",
+                "MissingIndicator_48",
+            ]
 
         # Convert multiple columns from boolean to float
         self.data[boolean_columns] = self.data[boolean_columns].astype(
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 0646e2f2..c492f6a3 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -565,6 +565,35 @@ def predict_dependency(self, dataloader, variable, log_dir):
         plt.savefig(log_dir / "dependecy.png", bbox_inches="tight")
         return dependency
 
+    def forward_captum(
+        self,
+        encoder_cat: Tensor,
+        encoder_cont: Tensor,
+        encoder_target: Tensor,
+        encoder_lengths: Tensor,
+        decoder_cat: Tensor,
+        decoder_cont: Tensor,
+        decoder_target: Tensor,
+        decoder_lengths: Tensor,
+        decoder_time_idx: Tensor,
+        groups: Tensor,
+        target_scale: Tensor,
+    ):
+        tuple_x = (
+            encoder_cat.float().requires_grad_(),
+            encoder_cont.requires_grad_(),
+            encoder_target.requires_grad_(),
+            encoder_lengths.float().requires_grad_(),
+            decoder_cat.float().requires_grad_(),
+            decoder_cont.requires_grad_(),
+            decoder_target.requires_grad_(),
+            decoder_lengths.float().requires_grad_(),
+            decoder_time_idx.float().requires_grad_(),
+            groups.float().requires_grad_(),
+            target_scale.requires_grad_(),
+        )
+        return self.forward(tuple_x)
+
 
 @gin.configurable
 class RNNpytorch(DLPredictionWrapper):
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index c1d96240..11ca6a57 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -25,7 +25,8 @@
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
-from captum.attr import IntegratedGradients
+from captum.attr import IntegratedGradients, Saliency
+
 
 cpu_core_count = (
     len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
@@ -149,7 +150,7 @@ def train_common(
         )
         test_loader = test_dataset.to_dataloader(
             train=False,
-            batch_size=batch_size,
+            batch_size=1,
             num_workers=num_workers,
             pin_memory=False,
             drop_last=True,
@@ -157,7 +158,11 @@ def train_common(
         )
         if load_weights:
             model = load_model(
-                model, source_dir, pl_model=pl_model, train_dataset=train_dataset, optimizer=optimizer
+                model,
+                source_dir,
+                pl_model=pl_model,
+                train_dataset=train_dataset,
+                optimizer=optimizer,
             )
 
         else:
@@ -237,7 +242,7 @@ def train_common(
         max_epochs=epochs if model.requires_backprop else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="auto" if not cpu else "cpu",
+        accelerator="cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic="warn" if reproducible else False,
         benchmark=not reproducible,
@@ -266,25 +271,38 @@ def train_common(
     if explain:
         # actual_vs_predictions = model.actual_vs_predictions_plot(test_loader)
         # print("1", actual_vs_predictions)
-        interperations = model.interpertations(test_loader, log_dir)
-        print("2", interperations)
+        # interperations = model.interpertations(test_loader, log_dir)
+        # print("2", interperations)
         # pd = model.predict_dependency(test_loader, "age", log_dir)
         # print("3", pd)
-        """
+
         test = next(iter(test_loader))
 
         for key, value in test[0].items():
             test[0][key] = test[0][key].to(model.device)
-
-        pred = model(test[0])
-
-        ig = IntegratedGradients(model)
+        data = (
+            test[0]["encoder_cat"],
+            test[0]["encoder_cont"],
+            test[0]["encoder_target"],
+            test[0]["encoder_lengths"],
+            test[0]["decoder_cat"],
+            test[0]["decoder_cont"],
+            test[0]["decoder_target"],
+            test[0]["decoder_lengths"],
+            test[0]["decoder_time_idx"],
+            test[0]["groups"],
+            test[0]["target_scale"],
+        )
+        # pred = model(data)
+        saliency = Saliency(model.forward_captum)
+        attr = saliency.attribute(data, target=0)
+        # ig = IntegratedGradients(model.forward_captum)
         # Reformat attributions.
 
-        attr, delta = ig.attribute(test[0], target=0, return_convergence_delta=True)
+        # attr, delta = ig.attribute(data, target=0, return_convergence_delta=True)
         attr = attr.detach().numpy()
         print(attr)
-        """
+
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
         "test/loss"

From fec90cfc1d966dd593ce8781c96b1e6bc83be012 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Wed, 11 Oct 2023 16:01:19 +0300
Subject: [PATCH 080/142] added quantus explantation and plot

---
 icu_benchmarks/models/dl_models.py | 100 +++++++++++++++++++++++++----
 icu_benchmarks/models/train.py     |  40 ++----------
 2 files changed, 95 insertions(+), 45 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index c492f6a3..a5f7f555 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -12,7 +12,7 @@
 )
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor, FloatTensor, isnan, isfinite
+from torch import Tensor, FloatTensor, zeros_like
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork
 from pytorch_forecasting.metrics import QuantileLoss
 import matplotlib.pyplot as plt
@@ -580,20 +580,96 @@ def forward_captum(
         target_scale: Tensor,
     ):
         tuple_x = (
-            encoder_cat.float().requires_grad_(),
-            encoder_cont.requires_grad_(),
-            encoder_target.requires_grad_(),
-            encoder_lengths.float().requires_grad_(),
-            decoder_cat.float().requires_grad_(),
-            decoder_cont.requires_grad_(),
-            decoder_target.requires_grad_(),
-            decoder_lengths.float().requires_grad_(),
-            decoder_time_idx.float().requires_grad_(),
-            groups.float().requires_grad_(),
-            target_scale.requires_grad_(),
+            encoder_cat,
+            encoder_cont,
+            encoder_target,
+            encoder_lengths,
+            decoder_cat,
+            decoder_cont,
+            decoder_target,
+            decoder_lengths,
+            decoder_time_idx,
+            groups,
+            target_scale,
         )
+
         return self.forward(tuple_x)
 
+    def explantation_captum(self, test_loader, log_dir, method):
+        # Initialize lists to store attribution values for all instances
+        all_attrs = []
+
+        # Loop through the test_loader to compute attributions for all instances
+        for batch in test_loader:
+            for key, value in batch[0].items():
+                batch[0][key] = batch[0][key].to(self.device)
+
+            data = (
+                batch[0]["encoder_cat"].float().requires_grad_(),
+                batch[0]["encoder_cont"].requires_grad_(),
+                batch[0]["encoder_target"].requires_grad_(),
+                batch[0]["encoder_lengths"].float().requires_grad_(),
+                batch[0]["decoder_cat"].float().requires_grad_(),
+                batch[0]["decoder_cont"].requires_grad_(),
+                batch[0]["decoder_target"].requires_grad_(),
+                batch[0]["decoder_lengths"].float().requires_grad_(),
+                batch[0]["decoder_time_idx"].float().requires_grad_(),
+                batch[0]["groups"].float().requires_grad_(),
+                batch[0]["target_scale"].requires_grad_(),
+            )
+            baselines = (
+                zeros_like(data[0]),  # encoder_cat, set to zero
+                zeros_like(data[1]),  # encoder_cont, set to zero
+                zeros_like(data[2]),  # encoder_target, set to zero
+                data[3],  # encoder_lengths, leave unchanged
+                zeros_like(data[4]),  # decoder_cat, set to zero
+                zeros_like(data[5]),  # decoder_cont, set to zero
+                zeros_like(data[6]),  # decoder_target, set to zero
+                data[7],  # decoder_lengths, leave unchanged
+                zeros_like(data[8]),  # decoder_time_idx, set to zero
+                data[9],  # groups, leave unchanged
+                data[10],  # target_scale, leave unchanged
+            )
+
+            explantation = method(self.forward_captum, self.model.multihead_attn)
+            # Reformat attributions.
+            attr, delta = explantation.attribute(
+                data, target=0, return_convergence_delta=True, baselines=baselines
+            )
+
+            # Convert attributions to numpy array and append to the list
+            all_attrs.append(attr.detach().numpy())
+
+        # Concatenate attribution values for all instances along the batch dimension
+        all_attrs = np.concatenate(all_attrs, axis=0)
+
+        # Compute mean along the batch dimension
+        means = all_attrs.mean(axis=(0, 2))
+        # Normalize the means values to range [0, 1]
+        normalized_means = (means - means.min()) / (means.max() - means.min())
+
+        # Create x values (assuming you want a simple sequential x-axis)
+        x_values = np.arange(1, 25)  # Assuming you have 24 values
+
+        # Plotting the normalized means
+        plt.figure(figsize=(8, 6))
+        plt.plot(
+            x_values,
+            normalized_means,
+            marker="o",
+            color="skyblue",
+            linestyle="-",
+            linewidth=2,
+            markersize=8,
+        )
+        plt.xlabel("Feature Index")
+        plt.ylabel("Normalized Means")
+        plt.title("Normalized Means of Attribution Values")
+        plt.xticks(x_values)  # Set x-ticks to match the number of features
+        plt.tight_layout()
+        plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
+        return means
+
 
 @gin.configurable
 class RNNpytorch(DLPredictionWrapper):
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 11ca6a57..5f707551 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -25,7 +25,7 @@
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
-from captum.attr import IntegratedGradients, Saliency
+from captum.attr import IntegratedGradients, Saliency, LayerIntegratedGradients
 
 
 cpu_core_count = (
@@ -150,7 +150,7 @@ def train_common(
         )
         test_loader = test_dataset.to_dataloader(
             train=False,
-            batch_size=1,
+            batch_size=batch_size,
             num_workers=num_workers,
             pin_memory=False,
             drop_last=True,
@@ -269,39 +269,13 @@ def train_common(
         return 0
 
     if explain:
-        # actual_vs_predictions = model.actual_vs_predictions_plot(test_loader)
-        # print("1", actual_vs_predictions)
-        # interperations = model.interpertations(test_loader, log_dir)
-        # print("2", interperations)
-        # pd = model.predict_dependency(test_loader, "age", log_dir)
-        # print("3", pd)
-
-        test = next(iter(test_loader))
+        interperations = model.interpertations(test_loader, log_dir)
+        print("attention", interperations)
 
-        for key, value in test[0].items():
-            test[0][key] = test[0][key].to(model.device)
-        data = (
-            test[0]["encoder_cat"],
-            test[0]["encoder_cont"],
-            test[0]["encoder_target"],
-            test[0]["encoder_lengths"],
-            test[0]["decoder_cat"],
-            test[0]["decoder_cont"],
-            test[0]["decoder_target"],
-            test[0]["decoder_lengths"],
-            test[0]["decoder_time_idx"],
-            test[0]["groups"],
-            test[0]["target_scale"],
+        explanintations = model.explantation_captum(
+            test_loader, log_dir, LayerIntegratedGradients
         )
-        # pred = model(data)
-        saliency = Saliency(model.forward_captum)
-        attr = saliency.attribute(data, target=0)
-        # ig = IntegratedGradients(model.forward_captum)
-        # Reformat attributions.
-
-        # attr, delta = ig.attribute(data, target=0, return_convergence_delta=True)
-        attr = attr.detach().numpy()
-        print(attr)
+        print("IG", explanintations)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][

From 391c7f5a10d18402f97fd12ca467688f5d098a4d Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Wed, 11 Oct 2023 16:07:19 +0300
Subject: [PATCH 081/142] added back gpu line

---
 icu_benchmarks/models/train.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 5f707551..0b54a4b5 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -242,7 +242,7 @@ def train_common(
         max_epochs=epochs if model.requires_backprop else 1,
         callbacks=callbacks,
         precision=precision,
-        accelerator="cpu",
+        accelerator="auto" if not cpu else "cpu",
         devices=max(torch.cuda.device_count(), 1),
         deterministic="warn" if reproducible else False,
         benchmark=not reproducible,

From 8746996d8c5217bb51b579672fa69256eb61ab2e Mon Sep 17 00:00:00 2001
From: "youssef.mecky" <youssef.mecky@guest.hpi.de>
Date: Tue, 17 Oct 2023 17:13:21 +0200
Subject: [PATCH 082/142] quantus looks like it is working

---
 configs/prediction_models/RNNpytorch.gin      | 99 +------------------
 configs/prediction_models/common/DLCommon.gin |  2 +-
 icu_benchmarks/data/loader.py                 |  7 +-
 icu_benchmarks/models/dl_models.py            | 45 ++++-----
 icu_benchmarks/models/train.py                | 10 +-
 5 files changed, 34 insertions(+), 129 deletions(-)

diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 63c1de4e..91ea89f1 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -13,7 +13,7 @@ optimizer/hyperparameter.weight_decay = 1e-6
 # Model params
 
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = (32, 128, "log-uniform", 2)
+model/hyperparameter.hidden = (10, 32, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
@@ -75,54 +75,6 @@ PredictionDatasetTFTpytorch.target=[
                 "tnt",
                 "urine",
                 "wbc",
-                "MissingIndicator_1",
-                "MissingIndicator_2",
-                "MissingIndicator_3",
-                "MissingIndicator_4",
-                "MissingIndicator_5",
-                "MissingIndicator_6",
-                "MissingIndicator_7",
-                "MissingIndicator_8",
-                "MissingIndicator_9",
-                "MissingIndicator_10",
-                "MissingIndicator_11",
-                "MissingIndicator_12",
-                "MissingIndicator_13",
-                "MissingIndicator_14",
-                "MissingIndicator_15",
-                "MissingIndicator_16",
-                "MissingIndicator_17",
-                "MissingIndicator_18",
-                "MissingIndicator_19",
-                "MissingIndicator_20",
-                "MissingIndicator_21",
-                "MissingIndicator_22",
-                "MissingIndicator_23",
-                "MissingIndicator_24",
-                "MissingIndicator_25",
-                "MissingIndicator_26",
-                "MissingIndicator_27",
-                "MissingIndicator_28",
-                "MissingIndicator_29",
-                "MissingIndicator_30",
-                "MissingIndicator_31",
-                "MissingIndicator_32",
-                "MissingIndicator_33",
-                "MissingIndicator_34",
-                "MissingIndicator_35",
-                "MissingIndicator_36",
-                "MissingIndicator_37",
-                "MissingIndicator_38",
-                "MissingIndicator_39",
-                "MissingIndicator_40",
-                "MissingIndicator_41",
-                "MissingIndicator_42",
-                "MissingIndicator_43",
-                "MissingIndicator_44",
-                "MissingIndicator_45",
-                "MissingIndicator_46",
-                "MissingIndicator_47",
-                "MissingIndicator_48",
                 "label",
             ]
 PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
@@ -173,53 +125,6 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "tnt",
                 "urine",
                 "wbc",
-                "MissingIndicator_1",
-                "MissingIndicator_2",
-                "MissingIndicator_3",
-                "MissingIndicator_4",
-                "MissingIndicator_5",
-                "MissingIndicator_6",
-                "MissingIndicator_7",
-                "MissingIndicator_8",
-                "MissingIndicator_9",
-                "MissingIndicator_10",
-                "MissingIndicator_11",
-                "MissingIndicator_12",
-                "MissingIndicator_13",
-                "MissingIndicator_14",
-                "MissingIndicator_15",
-                "MissingIndicator_16",
-                "MissingIndicator_17",
-                "MissingIndicator_18",
-                "MissingIndicator_19",
-                "MissingIndicator_20",
-                "MissingIndicator_21",
-                "MissingIndicator_22",
-                "MissingIndicator_23",
-                "MissingIndicator_24",
-                "MissingIndicator_25",
-                "MissingIndicator_26",
-                "MissingIndicator_27",
-                "MissingIndicator_28",
-                "MissingIndicator_29",
-                "MissingIndicator_30",
-                "MissingIndicator_31",
-                "MissingIndicator_32",
-                "MissingIndicator_33",
-                "MissingIndicator_34",
-                "MissingIndicator_35",
-                "MissingIndicator_36",
-                "MissingIndicator_37",
-                "MissingIndicator_38",
-                "MissingIndicator_39",
-                "MissingIndicator_40",
-                "MissingIndicator_41",
-                "MissingIndicator_42",
-                "MissingIndicator_43",
-                "MissingIndicator_44",
-                "MissingIndicator_45",
-                "MissingIndicator_46",
-                "MissingIndicator_47",
-                "MissingIndicator_48",
                 "label",
             ]
+PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
\ No newline at end of file
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index c1600089..fa4b440f 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -14,7 +14,7 @@ base_regression_preprocessor.generate_features = False
 
 # Train params
 train_common.optimizer = @Adam
-train_common.epochs = 1
+train_common.epochs = 1000
 train_common.batch_size = 64
 train_common.patience = 20
 train_common.min_delta = 1e-4
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 1de09cd2..d25174ce 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -591,10 +591,9 @@ def __init__(
                 "MissingIndicator_48",
             ]
 
-        # Convert multiple columns from boolean to float
-        self.data[boolean_columns] = self.data[boolean_columns].astype(
-            str
-        )  # changing boolean to floats to allow input to models
+       # self.data[boolean_columns] = self.data[boolean_columns].astype(
+       #     float
+       # )
 
         self.split = split
         self.args = args
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index a5f7f555..7e4d7c8b 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -516,6 +516,7 @@ def forward(
         }
         out = self.model(x_dict)
         pred = self.logit(out["prediction"])
+
         return pred
 
     """
@@ -595,7 +596,7 @@ def forward_captum(
 
         return self.forward(tuple_x)
 
-    def explantation_captum(self, test_loader, log_dir, method):
+    def explantation_captum(self, test_loader, log_dir, method, target):
         # Initialize lists to store attribution values for all instances
         all_attrs = []
 
@@ -607,40 +608,40 @@ def explantation_captum(self, test_loader, log_dir, method):
             data = (
                 batch[0]["encoder_cat"].float().requires_grad_(),
                 batch[0]["encoder_cont"].requires_grad_(),
-                batch[0]["encoder_target"].requires_grad_(),
+                batch[0]["encoder_target"].float().requires_grad_(),
                 batch[0]["encoder_lengths"].float().requires_grad_(),
                 batch[0]["decoder_cat"].float().requires_grad_(),
                 batch[0]["decoder_cont"].requires_grad_(),
-                batch[0]["decoder_target"].requires_grad_(),
+                batch[0]["decoder_target"].float().requires_grad_(),
                 batch[0]["decoder_lengths"].float().requires_grad_(),
                 batch[0]["decoder_time_idx"].float().requires_grad_(),
                 batch[0]["groups"].float().requires_grad_(),
                 batch[0]["target_scale"].requires_grad_(),
             )
+
             baselines = (
-                zeros_like(data[0]),  # encoder_cat, set to zero
-                zeros_like(data[1]),  # encoder_cont, set to zero
-                zeros_like(data[2]),  # encoder_target, set to zero
-                data[3],  # encoder_lengths, leave unchanged
-                zeros_like(data[4]),  # decoder_cat, set to zero
-                zeros_like(data[5]),  # decoder_cont, set to zero
-                zeros_like(data[6]),  # decoder_target, set to zero
-                data[7],  # decoder_lengths, leave unchanged
-                zeros_like(data[8]),  # decoder_time_idx, set to zero
-                data[9],  # groups, leave unchanged
-                data[10],  # target_scale, leave unchanged
+                zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
+                zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
+                zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
+                data[3].to(self.device),  # encoder_lengths, leave unchanged
+                zeros_like(data[4]).to(self.device),  # decoder_cat, set to zero
+                zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
+                zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
+                data[7].to(self.device),  # decoder_lengths, leave unchanged
+                zeros_like(data[8]).to(self.device),  # decoder_time_idx, set to zero
+                data[9].to(self.device),  # groups, leave unchanged
+                data[10].to(self.device),  # target_scale, leave unchanged
             )
 
-            explantation = method(self.forward_captum, self.model.multihead_attn)
+            explantation = method(self.forward_captum)
             # Reformat attributions.
             attr, delta = explantation.attribute(
-                data, target=0, return_convergence_delta=True, baselines=baselines
+                data, target=target, return_convergence_delta=True, baselines=baselines, n_steps=20
             )
-
             # Convert attributions to numpy array and append to the list
-            all_attrs.append(attr.detach().numpy())
+            all_attrs.append(attr[0].cpu().detach().numpy())
 
-        # Concatenate attribution values for all instances along the batch dimension
+        # Concatenate a‚ttribution values for all instances along the batch dimension
         all_attrs = np.concatenate(all_attrs, axis=0)
 
         # Compute mean along the batch dimension
@@ -662,9 +663,9 @@ def explantation_captum(self, test_loader, log_dir, method):
             linewidth=2,
             markersize=8,
         )
-        plt.xlabel("Feature Index")
-        plt.ylabel("Normalized Means")
-        plt.title("Normalized Means of Attribution Values")
+        plt.xlabel("Time Step")
+        plt.ylabel("Normalized Attribution")
+        plt.title("Attribution Values")
         plt.xticks(x_values)  # Set x-ticks to match the number of features
         plt.tight_layout()
         plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 0b54a4b5..cabdb8ba 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -25,7 +25,7 @@
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
-from captum.attr import IntegratedGradients, Saliency, LayerIntegratedGradients
+from captum.attr import IntegratedGradients, LayerLRP
 
 
 cpu_core_count = (
@@ -269,13 +269,13 @@ def train_common(
         return 0
 
     if explain:
-        interperations = model.interpertations(test_loader, log_dir)
-        print("attention", interperations)
 
         explanintations = model.explantation_captum(
-            test_loader, log_dir, LayerIntegratedGradients
-        )
+            test_loader, log_dir, IntegratedGradients, target=(0, 1))
+
         print("IG", explanintations)
+        interperations = model.interpertations(test_loader, log_dir)
+        print("attention", interperations)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][

From b7be9f6f201e77f06b1b4c4a98a8e5db696cdd19 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 24 Oct 2023 11:36:28 +0200
Subject: [PATCH 083/142] added deeparmodel

---
 configs/prediction_models/DeepARpytorch.gin | 80 +++++++++++++++++++++
 configs/prediction_models/RNNpytorch.gin    |  6 +-
 icu_benchmarks/models/dl_models.py          | 27 +++++--
 icu_benchmarks/models/wrappers.py           |  2 +
 4 files changed, 108 insertions(+), 7 deletions(-)
 create mode 100644 configs/prediction_models/DeepARpytorch.gin

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
new file mode 100644
index 00000000..9c7b8116
--- /dev/null
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -0,0 +1,80 @@
+# Settings for TFT model.
+
+# Common settings for DL models
+include "configs/prediction_models/common/DLCommon.gin"
+
+# Optimizer params
+train_common.model = @RNNpytorch
+
+optimizer/hyperparameter.class_to_tune = @Adam
+optimizer/hyperparameter.weight_decay = 1e-6
+
+
+# Model params
+
+model/hyperparameter.class_to_tune = @DeepARpytorch
+model/hyperparameter.hidden = 2
+model/hyperparameter.rnn_layers=2
+model/hyperparameter.num_classes = %NUM_CLASSES
+model/hyperparameter.cell_type='LSTM'
+model/hyperparameter.dropout = (0.0, 0.4)
+model/hyperparameter.lr_scheduler = "exponential"
+
+
+# Dataset params
+PredictionDatasetTFTpytorch.max_encoder_length = 24
+PredictionDatasetTFTpytorch.max_prediction_length = 24
+PredictionDatasetTFTpytorch.target="label"
+PredictionDatasetTFTpytorch.time_varying_known_reals=["time_idx"]
+PredictionDatasetTFTpytorch.add_relative_time_idx=True
+PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
+                "label",]
+                
\ No newline at end of file
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 91ea89f1..84eb6dd9 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -13,10 +13,10 @@ optimizer/hyperparameter.weight_decay = 1e-6
 # Model params
 
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = (10, 32, "log-uniform", 2)
-model/hyperparameter.rnn_layers=(1, 3)
+model/hyperparameter.hidden = 2
+model/hyperparameter.rnn_layers=2
 model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.cell_type='LSTM'
+model/hyperparameter.cell_type='GRU'
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.lr_scheduler = "exponential"
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 7e4d7c8b..f5052b0c 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -643,6 +643,8 @@ def explantation_captum(self, test_loader, log_dir, method, target):
 
         # Concatenate a‚ttribution values for all instances along the batch dimension
         all_attrs = np.concatenate(all_attrs, axis=0)
+        print(all_attrs.shape)
+        means_feature = all_attrs.mean(axis=(0, 1))
 
         # Compute mean along the batch dimension
         means = all_attrs.mean(axis=(0, 2))
@@ -650,10 +652,27 @@ def explantation_captum(self, test_loader, log_dir, method, target):
         normalized_means = (means - means.min()) / (means.max() - means.min())
 
         # Create x values (assuming you want a simple sequential x-axis)
-        x_values = np.arange(1, 25)  # Assuming you have 24 values
-
-        # Plotting the normalized means
+        # Assuming you have 24 values
+        x_values = np.arange(1, 57)
+        # Plotting the featrue means
+        plt.figure(figsize=(8, 6))
+        plt.plot(
+            x_values,
+            means_feature,
+            marker="o",
+            color="skyblue",
+            linestyle="-",
+            linewidth=2,
+            markersize=8,
+        )
+        plt.xlabel("Time Step")
+        plt.ylabel("Normalized Attribution")
+        plt.title("Attribution Values")
+        plt.xticks(x_values)  # Set x-ticks to match the number of features
+        plt.tight_layout()
+        plt.savefig(log_dir / "attribution_features_plot.png", bbox_inches="tight")
         plt.figure(figsize=(8, 6))
+        x_values = np.arange(1, 25)
         plt.plot(
             x_values,
             normalized_means,
@@ -727,5 +746,5 @@ def forward(
         out = out["prediction"][-1].reshape(1, -1)
 
         pred = self.logit(out)
-
+        print(pred)
         return pred
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 163ca900..f8dcbb42 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -397,6 +397,8 @@ def step_fn(self, element, step_prefix=""):
             .to(self.device)
         )
         target = torch.masked_select(labels, mask).to(self.device)
+        print('target', target)
+        print('pred', prediction)
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
             loss = (

From 89943455ea959030b9e0998771b9684612ac5e2f Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 24 Oct 2023 12:17:43 +0200
Subject: [PATCH 084/142] deepar fixes

---
 configs/prediction_models/DeepARpytorch.gin | 63 +++++++++++++++++++--
 icu_benchmarks/data/loader.py               |  6 +-
 icu_benchmarks/models/dl_models.py          | 62 +++++++++++++++++++-
 icu_benchmarks/models/wrappers.py           |  3 +-
 4 files changed, 122 insertions(+), 12 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 9c7b8116..43c6e914 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -4,7 +4,7 @@
 include "configs/prediction_models/common/DLCommon.gin"
 
 # Optimizer params
-train_common.model = @RNNpytorch
+train_common.model = @DeepARpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
@@ -24,10 +24,59 @@ model/hyperparameter.lr_scheduler = "exponential"
 # Dataset params
 PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
-PredictionDatasetTFTpytorch.target="label"
-PredictionDatasetTFTpytorch.time_varying_known_reals=["time_idx"]
-PredictionDatasetTFTpytorch.add_relative_time_idx=True
-PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetTFTpytorch.time_varying_known_reals=[]
+PredictionDatasetTFTpytorch.add_relative_time_idx=False
+PredictionDatasetTFTpytorch.target=[
+                "alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
+                "label",
+            ]
 PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "alp",
                 "alt",
@@ -76,5 +125,7 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "tnt",
                 "urine",
                 "wbc",
-                "label",]
+                "label",
+            ]
+PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
                 
\ No newline at end of file
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index d25174ce..642d678a 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -591,9 +591,9 @@ def __init__(
                 "MissingIndicator_48",
             ]
 
-       # self.data[boolean_columns] = self.data[boolean_columns].astype(
-       #     float
-       # )
+        self.data[boolean_columns] = self.data[boolean_columns].astype(
+            float
+        )
 
         self.split = split
         self.args = args
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index f5052b0c..b66fe3c4 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -13,7 +13,7 @@
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
 from torch import Tensor, FloatTensor, zeros_like
-from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork
+from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss
 import matplotlib.pyplot as plt
 
@@ -748,3 +748,63 @@ def forward(
         pred = self.logit(out)
         print(pred)
         return pred
+
+
+@gin.configurable
+class DeepARpytorch(DLPredictionWrapper):
+    """
+    Implementation of RNN from pytorch forecasting
+    """
+
+    supported_run_modes = [RunMode.classification, RunMode.regression]
+
+    def __init__(
+        self,
+        dataset,
+        hidden,
+        dropout,
+        optimizer,
+        num_classes,
+        cell_type,
+        rnn_layers,
+        batch_size,
+        *args,
+        **kwargs,
+    ):
+        super().__init__(optimizer=optimizer, pytorch_forecasting=True, *args, **kwargs)
+
+        self.model = DeepAR.from_dataset(
+            cell_type=cell_type,
+            rnn_layers=rnn_layers,
+            dataset=dataset,
+            hidden_size=hidden,
+            dropout=dropout,
+            optimizer=optimizer,
+        )
+
+        self.logit = nn.Linear(batch_size * 24*4, num_classes)
+
+    def forward(
+        self,
+        tuple_x: tuple,
+    ) -> Dict[str, Tensor]:
+        x_dict = {
+            "encoder_cat": tuple_x[0],
+            "encoder_cont": tuple_x[1],
+            "encoder_target": tuple_x[2],
+            "encoder_lengths": tuple_x[3],
+            "decoder_cat": tuple_x[4],
+            "decoder_cont": tuple_x[5],
+            "decoder_target": tuple_x[6],
+            "decoder_lengths": tuple_x[7],
+            "decoder_time_idx": tuple_x[8],
+            "groups": tuple_x[9],
+            "target_scale": tuple_x[10],
+        }
+        out = self.model(x_dict)
+        print(out["prediction"])
+        out = out["prediction"][-1].reshape(1, -1)
+
+        pred = self.logit(out)
+        print(pred)
+        return pred
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index f8dcbb42..f78792a2 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -397,8 +397,7 @@ def step_fn(self, element, step_prefix=""):
             .to(self.device)
         )
         target = torch.masked_select(labels, mask).to(self.device)
-        print('target', target)
-        print('pred', prediction)
+
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
             loss = (

From 986e11fe915b568e264140d74670b97d957eb658 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 25 Oct 2023 13:36:25 +0200
Subject: [PATCH 085/142] minor changes

---
 configs/prediction_models/DeepARpytorch.gin |  68 ++---------
 configs/prediction_models/RNNpytorch.gin    |  65 ++---------
 configs/prediction_models/TFTpytorch.gin    |   2 +-
 icu_benchmarks/data/loader.py               | 121 +++-----------------
 icu_benchmarks/models/dl_models.py          |  16 ++-
 icu_benchmarks/models/wrappers.py           |   8 +-
 6 files changed, 50 insertions(+), 230 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 43c6e914..94ae90f0 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -13,8 +13,8 @@ optimizer/hyperparameter.weight_decay = 1e-6
 # Model params
 
 model/hyperparameter.class_to_tune = @DeepARpytorch
-model/hyperparameter.hidden = 2
-model/hyperparameter.rnn_layers=2
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
 model/hyperparameter.dropout = (0.0, 0.4)
@@ -24,60 +24,18 @@ model/hyperparameter.lr_scheduler = "exponential"
 # Dataset params
 PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
-PredictionDatasetTFTpytorch.time_varying_known_reals=[]
+PredictionDatasetTFTpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
 PredictionDatasetTFTpytorch.add_relative_time_idx=False
 PredictionDatasetTFTpytorch.target=[
-                "alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",
                 "label",
             ]
-PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+PredictionDatasetTFTpytorch.time_varying_unknown_reals=[
+                "label",
+            ]
+
+            
+PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetTFTpytorch.lagged_variables=["alb",
                 "alp",
                 "alt",
                 "ast",
@@ -124,8 +82,6 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc",
-                "label",
-            ]
-PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
+                "wbc",]
+
                 
\ No newline at end of file
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 84eb6dd9..ab42884f 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -16,7 +16,7 @@ model/hyperparameter.class_to_tune = @RNNpytorch
 model/hyperparameter.hidden = 2
 model/hyperparameter.rnn_layers=2
 model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.cell_type='GRU'
+model/hyperparameter.cell_type='LSTM'
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.lr_scheduler = "exponential"
 
@@ -24,60 +24,16 @@ model/hyperparameter.lr_scheduler = "exponential"
 # Dataset params
 PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
-PredictionDatasetTFTpytorch.time_varying_known_reals=[]
+PredictionDatasetTFTpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
+
 PredictionDatasetTFTpytorch.add_relative_time_idx=False
-PredictionDatasetTFTpytorch.target=[
-                "alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",
+PredictionDatasetTFTpytorch.target="label"
+            
+PredictionDatasetTFTpytorch.time_varying_unknown_reals=[
                 "label",
             ]
-PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetTFTpytorch.lagged_variables=["alb",
                 "alp",
                 "alt",
                 "ast",
@@ -124,7 +80,4 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc",
-                "label",
-            ]
-PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
\ No newline at end of file
+                "wbc",]
\ No newline at end of file
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 68290c1d..a6523c6b 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -76,4 +76,4 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "urine",
                 "wbc",
                 "label",]
-                
\ No newline at end of file
+PredictionDatasetTFTpytorch.lagged_variables=[]
\ No newline at end of file
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 642d678a..e02b2e76 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -465,14 +465,17 @@ def __init__(
         target: Union[str, List[str]],
         time_varying_known_reals: List[str],
         time_varying_unknown_categoricals: List[str],
+        lagged_variables: List[str],
         *args,
         ram_cache: bool = False,
         add_relative_time_idx: bool = False,
         name: str = "",
         max_prediction_length: int = 24,
         max_encoder_length: int = 24,
+
         **kwargs,
     ):
+
         data[split]["FEATURES"]["time_idx"] = (
             (data[split]["FEATURES"]["time"] / pd.Timedelta(seconds=3600))
         ).astype(
@@ -487,113 +490,17 @@ def __init__(
         )  # combine labels and features
         # self.data["sex"].replace([0, 1], ["Female", "Male"], inplace=True)
         # List of column names to convert from boolean to float
-        if self.data["label"].dtype == "bool":
-            boolean_columns = [
-                "MissingIndicator_1",
-                "MissingIndicator_2",
-                "MissingIndicator_3",
-                "MissingIndicator_4",
-                "MissingIndicator_5",
-                "MissingIndicator_6",
-                "MissingIndicator_7",
-                "MissingIndicator_8",
-                "MissingIndicator_9",
-                "MissingIndicator_10",
-                "MissingIndicator_11",
-                "MissingIndicator_12",
-                "MissingIndicator_13",
-                "MissingIndicator_14",
-                "MissingIndicator_15",
-                "MissingIndicator_16",
-                "MissingIndicator_17",
-                "MissingIndicator_18",
-                "MissingIndicator_19",
-                "MissingIndicator_20",
-                "MissingIndicator_21",
-                "MissingIndicator_22",
-                "MissingIndicator_23",
-                "MissingIndicator_24",
-                "MissingIndicator_25",
-                "MissingIndicator_26",
-                "MissingIndicator_27",
-                "MissingIndicator_28",
-                "MissingIndicator_29",
-                "MissingIndicator_30",
-                "MissingIndicator_31",
-                "MissingIndicator_32",
-                "MissingIndicator_33",
-                "MissingIndicator_34",
-                "MissingIndicator_35",
-                "MissingIndicator_36",
-                "MissingIndicator_37",
-                "MissingIndicator_38",
-                "MissingIndicator_39",
-                "MissingIndicator_40",
-                "MissingIndicator_41",
-                "MissingIndicator_42",
-                "MissingIndicator_43",
-                "MissingIndicator_44",
-                "MissingIndicator_45",
-                "MissingIndicator_46",
-                "MissingIndicator_47",
-                "MissingIndicator_48",
-                "label",
-            ]
-        else:
-            boolean_columns = [
-                "MissingIndicator_1",
-                "MissingIndicator_2",
-                "MissingIndicator_3",
-                "MissingIndicator_4",
-                "MissingIndicator_5",
-                "MissingIndicator_6",
-                "MissingIndicator_7",
-                "MissingIndicator_8",
-                "MissingIndicator_9",
-                "MissingIndicator_10",
-                "MissingIndicator_11",
-                "MissingIndicator_12",
-                "MissingIndicator_13",
-                "MissingIndicator_14",
-                "MissingIndicator_15",
-                "MissingIndicator_16",
-                "MissingIndicator_17",
-                "MissingIndicator_18",
-                "MissingIndicator_19",
-                "MissingIndicator_20",
-                "MissingIndicator_21",
-                "MissingIndicator_22",
-                "MissingIndicator_23",
-                "MissingIndicator_24",
-                "MissingIndicator_25",
-                "MissingIndicator_26",
-                "MissingIndicator_27",
-                "MissingIndicator_28",
-                "MissingIndicator_29",
-                "MissingIndicator_30",
-                "MissingIndicator_31",
-                "MissingIndicator_32",
-                "MissingIndicator_33",
-                "MissingIndicator_34",
-                "MissingIndicator_35",
-                "MissingIndicator_36",
-                "MissingIndicator_37",
-                "MissingIndicator_38",
-                "MissingIndicator_39",
-                "MissingIndicator_40",
-                "MissingIndicator_41",
-                "MissingIndicator_42",
-                "MissingIndicator_43",
-                "MissingIndicator_44",
-                "MissingIndicator_45",
-                "MissingIndicator_46",
-                "MissingIndicator_47",
-                "MissingIndicator_48",
-            ]
-
-        self.data[boolean_columns] = self.data[boolean_columns].astype(
-            float
-        )
+        if (len(lagged_variables) > 0):
+            if self.data["label"].dtype == "bool":
+                self.data["label"] = self.data["label"].astype(
+                    float
+                )
+            columns_to_lag = lagged_variables
+            grouped = self.data.sort_values("time_idx").groupby("stay_id")
+            for lag in range(max_encoder_length, max_encoder_length+max_prediction_length + 1):
+                for column in columns_to_lag:
+                    # Create a new column with lagged values
+                    self.data[f"{column}_lag_{lag}"] = grouped[column].shift(lag, fill_value=0)
 
         self.split = split
         self.args = args
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index b66fe3c4..12a7213a 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -516,7 +516,6 @@ def forward(
         }
         out = self.model(x_dict)
         pred = self.logit(out["prediction"])
-
         return pred
 
     """
@@ -723,7 +722,7 @@ def __init__(
             optimizer=optimizer,
         )
 
-        self.logit = nn.Linear(batch_size * 24, num_classes)
+        self.logit = nn.Linear(1, num_classes)
 
     def forward(
         self,
@@ -743,10 +742,11 @@ def forward(
             "target_scale": tuple_x[10],
         }
         out = self.model(x_dict)
-        out = out["prediction"][-1].reshape(1, -1)
 
-        pred = self.logit(out)
-        print(pred)
+        # out = out["prediction"].reshape(1, -1)
+
+        pred = self.logit(out["prediction"])
+
         return pred
 
 
@@ -782,7 +782,7 @@ def __init__(
             optimizer=optimizer,
         )
 
-        self.logit = nn.Linear(batch_size * 24*4, num_classes)
+        self.logit = nn.Linear(4, num_classes)
 
     def forward(
         self,
@@ -802,9 +802,7 @@ def forward(
             "target_scale": tuple_x[10],
         }
         out = self.model(x_dict)
-        print(out["prediction"])
-        out = out["prediction"][-1].reshape(1, -1)
+        out = out["prediction"][0]
 
         pred = self.logit(out)
-        print(pred)
         return pred
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index f78792a2..7a72c4ca 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -391,13 +391,19 @@ def step_fn(self, element, step_prefix=""):
         else:
             aux_loss = 0
         # Get prediction and target
+
         prediction = (
             torch.masked_select(out, mask.unsqueeze(-1))
             .reshape(-1, out.shape[-1])
             .to(self.device)
         )
+        print('output', out)
+        print('output', out.shape)
+        print('pred', prediction)
+        print('pred shape', prediction.shape)
         target = torch.masked_select(labels, mask).to(self.device)
-
+        print('target', target)
+        print('target shape', target.shape)
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
             loss = (

From 0ed2fbc6cd78ce1cbdd855919f722435b395567a Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Wed, 25 Oct 2023 13:40:16 +0200
Subject: [PATCH 086/142] added XAI metric flag

---
 icu_benchmarks/cross_validation.py |   2 +
 icu_benchmarks/models/train.py     |   3 +
 icu_benchmarks/run.py              |   2 +
 icu_benchmarks/run_utils.py        | 128 ++++++++++++++++++++++-------
 4 files changed, 105 insertions(+), 30 deletions(-)

diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index bcf2f5f5..8c0c02b1 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -40,6 +40,7 @@ def execute_repeated_cv(
     complete_train: bool = False,
     explain: bool = False,
     pytorch_forecasting: bool = False,
+    XAI_metric: bool = False,
 ) -> float:
     """Preprocesses data and trains a model for each fold.
 
@@ -127,6 +128,7 @@ def execute_repeated_cv(
                 train_only=complete_train,
                 explain=explain,
                 pytorch_forecasting=pytorch_forecasting,
+                XAI_metric=XAI_metric,
             )
 
             train_time = datetime.now() - start_time
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index cabdb8ba..806b0c20 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -71,6 +71,7 @@ def train_common(
     ),
     explain: bool = False,
     pytorch_forecasting: bool = False,
+    XAI_metric: bool = False,
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -276,6 +277,8 @@ def train_common(
         print("IG", explanintations)
         interperations = model.interpertations(test_loader, log_dir)
         print("attention", interperations)
+        if XAI_metric:
+
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 7302aec8..4d9fb2f4 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -54,6 +54,7 @@ def main(my_args=tuple(sys.argv[1:])):
     source_dir = args.source_dir
     explain = args.explain
     pytorch_forecasting = args.pytorch_forecasting
+    XAI_metric = args.XAI_metric
     # Load task config
     gin.parse_config_file(f"configs/tasks/{task}.gin")
     mode = get_mode()
@@ -209,6 +210,7 @@ def main(my_args=tuple(sys.argv[1:])):
         complete_train=args.complete_train,
         explain=explain,
         pytorch_forecasting=pytorch_forecasting,
+        XAI_metric=XAI_metric,
     )
 
     log_full_line("FINISHED TRAINING", level=logging.INFO, char="=", num_newlines=3)
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index b23aae68..cc4b7960 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -23,7 +23,9 @@ def build_parser() -> ArgumentParser:
     Returns:
         The configured ArgumentParser.
     """
-    parser = ArgumentParser(description="Framework for benchmarking ML/DL models on ICU data")
+    parser = ArgumentParser(
+        description="Framework for benchmarking ML/DL models on ICU data"
+    )
 
     parser.add_argument(
         "-d",
@@ -40,8 +42,12 @@ def build_parser() -> ArgumentParser:
         help="Name of the task gin.",
     )
     parser.add_argument("-n", "--name", help="Name of the (target) dataset.")
-    parser.add_argument("-tn", "--task-name", help="Name of the task, used for naming experiments.")
-    parser.add_argument("-m", "--model", default="LGBMClassifier", help="Name of the model gin.")
+    parser.add_argument(
+        "-tn", "--task-name", help="Name of the task, used for naming experiments."
+    )
+    parser.add_argument(
+        "-m", "--model", default="LGBMClassifier", help="Name of the model gin."
+    )
     parser.add_argument("-e", "--experiment", help="Name of the experiment gin.")
     parser.add_argument(
         "-l",
@@ -65,8 +71,12 @@ def build_parser() -> ArgumentParser:
         help="Set to log verbosly. Disable for clean logs.",
     )
     parser.add_argument("--cpu", default=False, action=BOA, help="Set to use CPU.")
-    parser.add_argument("-db", "--debug", default=False, action=BOA, help="Set to load less data.")
-    parser.add_argument("--reproducible", default=True, action=BOA, help="Make torch reproducible.")
+    parser.add_argument(
+        "-db", "--debug", default=False, action=BOA, help="Set to load less data."
+    )
+    parser.add_argument(
+        "--reproducible", default=True, action=BOA, help="Make torch reproducible."
+    )
     parser.add_argument(
         "-lc",
         "--load_cache",
@@ -81,7 +91,9 @@ def build_parser() -> ArgumentParser:
         action=BOA,
         help="Set to generate data cache.",
     )
-    parser.add_argument("-p", "--preprocessor", type=Path, help="Load custom preprocessor from file.")
+    parser.add_argument(
+        "-p", "--preprocessor", type=Path, help="Load custom preprocessor from file."
+    )
     parser.add_argument("-pl", "--plot", action=BOA, help="Generate common plots.")
     parser.add_argument(
         "-wd",
@@ -95,10 +107,18 @@ def build_parser() -> ArgumentParser:
         type=str,
         help="Path to pretrained imputation model.",
     )
-    parser.add_argument("-hp", "--hyperparams", nargs="+", help="Hyperparameters for model.")
-    parser.add_argument("--tune", default=False, action=BOA, help="Find best hyperparameters.")
-    parser.add_argument("--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint.")
-    parser.add_argument("--eval", default=False, action=BOA, help="Only evaluate model, skip training.")
+    parser.add_argument(
+        "-hp", "--hyperparams", nargs="+", help="Hyperparameters for model."
+    )
+    parser.add_argument(
+        "--tune", default=False, action=BOA, help="Find best hyperparameters."
+    )
+    parser.add_argument(
+        "--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint."
+    )
+    parser.add_argument(
+        "--eval", default=False, action=BOA, help="Only evaluate model, skip training."
+    )
     parser.add_argument(
         "--complete-train",
         default=False,
@@ -112,8 +132,12 @@ def build_parser() -> ArgumentParser:
         type=int,
         help="Finetune model with amount of train data.",
     )
-    parser.add_argument("-sn", "--source-name", type=Path, help="Name of the source dataset.")
-    parser.add_argument("--source-dir", type=Path, help="Directory containing gin and model weights.")
+    parser.add_argument(
+        "-sn", "--source-name", type=Path, help="Name of the source dataset."
+    )
+    parser.add_argument(
+        "--source-dir", type=Path, help="Directory containing gin and model weights."
+    )
     parser.add_argument(
         "-sa",
         "--samples",
@@ -133,6 +157,13 @@ def build_parser() -> ArgumentParser:
         action=BOA,
         help="use pytorch forecasting library ",
     )
+    parser.add_argument(
+        "--XAI_metric",
+        default=False,
+        action=BOA,
+        help="Compare explantations ",
+    )
+
     return parser
 
 
@@ -161,15 +192,21 @@ def create_run_dir(log_dir: Path, randomly_searched_params: str = None) -> Path:
 
 def import_preprocessor(preprocessor_path: str):
     # Import custom supplied preprocessor
-    log_full_line(f"Importing custom preprocessor from {preprocessor_path}.", logging.INFO)
+    log_full_line(
+        f"Importing custom preprocessor from {preprocessor_path}.", logging.INFO
+    )
     try:
-        spec = importlib.util.spec_from_file_location("CustomPreprocessor", preprocessor_path)
+        spec = importlib.util.spec_from_file_location(
+            "CustomPreprocessor", preprocessor_path
+        )
         module = importlib.util.module_from_spec(spec)
         sys.modules["preprocessor"] = module
         spec.loader.exec_module(module)
         gin.bind_parameter("preprocess.preprocessor", module.CustomPreprocessor)
     except Exception as e:
-        logging.error(f"Could not import custom preprocessor from {preprocessor_path}: {e}")
+        logging.error(
+            f"Could not import custom preprocessor from {preprocessor_path}: {e}"
+        )
 
 
 def aggregate_results(log_dir: Path, execution_time: timedelta = None):
@@ -213,10 +250,14 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
     # Calculate the population standard deviation over aggregated results over folds/iterations
     # Divide by sqrt(n) to get standard deviation.
-    std_scores = {metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()}
+    std_scores = {
+        metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()
+    }
 
     confidence_interval = {
-        metric: (stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list)))
+        metric: (
+            stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list))
+        )
         for metric, list in list_scores.items()
     }
 
@@ -224,7 +265,9 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
         "avg": averaged_scores,
         "std": std_scores,
         "CI_0.95": confidence_interval,
-        "execution_time": execution_time.total_seconds() if execution_time is not None else 0.0,
+        "execution_time": execution_time.total_seconds()
+        if execution_time is not None
+        else 0.0,
     }
 
     with open(log_dir / "aggregated_test_metrics.json", "w") as f:
@@ -240,10 +283,14 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
 def name_datasets(train="default", val="default", test="default"):
     """Names the datasets for logging (optional)."""
-    gin.bind_parameter("train_common.dataset_names", {"train": train, "val": val, "test": test})
+    gin.bind_parameter(
+        "train_common.dataset_names", {"train": train, "val": val, "test": test}
+    )
 
 
-def log_full_line(msg: str, level: int = logging.INFO, char: str = "-", num_newlines: int = 0):
+def log_full_line(
+    msg: str, level: int = logging.INFO, char: str = "-", num_newlines: int = 0
+):
     """Logs a full line of a given character with a message centered.
 
     Args:
@@ -271,24 +318,43 @@ def load_pretrained_imputation_model(use_pretrained_imputation):
     Args:
         use_pretrained_imputation: Path to the pretrained imputation model.
     """
-    if use_pretrained_imputation is not None and not Path(use_pretrained_imputation).exists():
+    if (
+        use_pretrained_imputation is not None
+        and not Path(use_pretrained_imputation).exists()
+    ):
         logging.warning("The specified pretrained imputation model does not exist.")
         use_pretrained_imputation = None
 
     if use_pretrained_imputation is not None:
-        logging.info("Using pretrained imputation from" + str(use_pretrained_imputation))
-        pretrained_imputation_model_checkpoint = torch.load(use_pretrained_imputation, map_location=torch.device("cpu"))
+        logging.info(
+            "Using pretrained imputation from" + str(use_pretrained_imputation)
+        )
+        pretrained_imputation_model_checkpoint = torch.load(
+            use_pretrained_imputation, map_location=torch.device("cpu")
+        )
         if isinstance(pretrained_imputation_model_checkpoint, dict):
             imputation_model_class = pretrained_imputation_model_checkpoint["class"]
-            pretrained_imputation_model = imputation_model_class(**pretrained_imputation_model_checkpoint["hyper_parameters"])
-            pretrained_imputation_model.set_trained_columns(pretrained_imputation_model_checkpoint["trained_columns"])
-            pretrained_imputation_model.load_state_dict(pretrained_imputation_model_checkpoint["state_dict"])
+            pretrained_imputation_model = imputation_model_class(
+                **pretrained_imputation_model_checkpoint["hyper_parameters"]
+            )
+            pretrained_imputation_model.set_trained_columns(
+                pretrained_imputation_model_checkpoint["trained_columns"]
+            )
+            pretrained_imputation_model.load_state_dict(
+                pretrained_imputation_model_checkpoint["state_dict"]
+            )
         else:
             pretrained_imputation_model = pretrained_imputation_model_checkpoint
-        pretrained_imputation_model = pretrained_imputation_model.to("cuda" if torch.cuda.is_available() else "cpu")
+        pretrained_imputation_model = pretrained_imputation_model.to(
+            "cuda" if torch.cuda.is_available() else "cpu"
+        )
         try:
-            logging.info(f"imputation model device: {next(pretrained_imputation_model.parameters()).device}")
-            pretrained_imputation_model.device = next(pretrained_imputation_model.parameters()).device
+            logging.info(
+                f"imputation model device: {next(pretrained_imputation_model.parameters()).device}"
+            )
+            pretrained_imputation_model.device = next(
+                pretrained_imputation_model.parameters()
+            ).device
         except Exception as e:
             logging.debug(f"Could not set device of imputation model: {e}")
     else:
@@ -309,7 +375,9 @@ def setup_logging(date_format, log_format, verbose):
     logging.basicConfig(format=log_format, datefmt=date_format)
     loggers = ["pytorch_lightning", "lightning_fabric"]
     for logger in loggers:
-        logging.getLogger(logger).handlers[0].setFormatter(logging.Formatter(log_format, datefmt=date_format))
+        logging.getLogger(logger).handlers[0].setFormatter(
+            logging.Formatter(log_format, datefmt=date_format)
+        )
 
     if not verbose:
         logging.getLogger().setLevel(logging.INFO)

From d42bd7338d948dce40708dd5a5e05cf768bbe3d4 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Wed, 25 Oct 2023 14:40:30 +0200
Subject: [PATCH 087/142] added XAI metric part

---
 icu_benchmarks/models/train.py    | 68 +++++++++++++++++++++++++++++--
 icu_benchmarks/models/wrappers.py |  8 +---
 2 files changed, 67 insertions(+), 9 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 806b0c20..1caaf9c6 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -6,6 +6,7 @@
 import pandas as pd
 from joblib import load
 from torch.optim import Adam
+import quantus
 from torch.utils.data import DataLoader
 from pytorch_lightning.loggers import TensorBoardLogger, WandbLogger
 from pytorch_lightning import Trainer
@@ -270,15 +271,76 @@ def train_common(
         return 0
 
     if explain:
-
         explanintations = model.explantation_captum(
-            test_loader, log_dir, IntegratedGradients, target=(0, 1))
+            test_loader, log_dir, IntegratedGradients, target=(0, 1)
+        )
 
         print("IG", explanintations)
         interperations = model.interpertations(test_loader, log_dir)
         print("attention", interperations)
-        if XAI_metric:
+    if XAI_metric:
+        batch = iter(test_loader).next()
+
+        for key, value in batch[0].items():
+            batch[0][key] = batch[0][key].to(model.device)
+
+        data = (
+            batch[0]["encoder_cat"].float().requires_grad_(),
+            batch[0]["encoder_cont"].requires_grad_(),
+            batch[0]["encoder_target"].float().requires_grad_(),
+            batch[0]["encoder_lengths"].float().requires_grad_(),
+            batch[0]["decoder_cat"].float().requires_grad_(),
+            batch[0]["decoder_cont"].requires_grad_(),
+            batch[0]["decoder_target"].float().requires_grad_(),
+            batch[0]["decoder_lengths"].float().requires_grad_(),
+            batch[0]["decoder_time_idx"].float().requires_grad_(),
+            batch[0]["groups"].float().requires_grad_(),
+            batch[0]["target_scale"].requires_grad_(),
+        )
 
+        baselines = (
+            torch.zeros_like(data[0]).to(model.device),  # encoder_cat, set to zero
+            torch.zeros_like(data[1]).to(model.device),  # encoder_cont, set to zero
+            torch.zeros_like(data[2]).to(model.device),  # encoder_target, set to zero
+            data[3].to(model.device),  # encoder_lengths, leave unchanged
+            torch.zeros_like(data[4]).to(model.device),  # decoder_cat, set to zero
+            torch.zeros_like(data[5]).to(model.device),  # decoder_cont, set to zero
+            torch.zeros_like(data[6]).to(model.device),  # decoder_target, set to zero
+            data[7].to(model.device),  # decoder_lengths, leave unchanged
+            torch.zeros_like(data[8]).to(model.device),  # decoder_time_idx, set to zero
+            data[9].to(model.device),  # groups, leave unchanged
+            data[10].to(model.device),  # target_scale, leave unchanged
+        )
+
+        explantation = IntegratedGradients(model.forward_captum)
+        # Reformat attributions.
+        attr, delta = explantation.attribute(
+            data,
+            target=(0, 1),
+            return_convergence_delta=True,
+            baselines=baselines,
+            n_steps=20,
+        )
+        a_batch_IG = attr.detach().numpy()
+
+        x_batch, y_batch = batch[0], batch[1][0]
+        for key, value in x_batch.items():
+            x_batch[key] = x_batch[key].cpu().numpy()
+        y_batch = y_batch.cpu().numpy()
+        metric = quantus.FaithfulnessCorrelation()
+        scores = metric(
+            model=model,
+            x_batch=x_batch,
+            y_batch=y_batch,
+            a_batch=a_batch_IG,
+            device=model.device,
+            explain_func=quantus.explain,
+            explain_func_kwargs={
+                "method": "IntegratedGradients",
+                "baselines": baselines,
+                "n_steps": 20,
+            },
+        )
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 7a72c4ca..ec467b75 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -397,13 +397,9 @@ def step_fn(self, element, step_prefix=""):
             .reshape(-1, out.shape[-1])
             .to(self.device)
         )
-        print('output', out)
-        print('output', out.shape)
-        print('pred', prediction)
-        print('pred shape', prediction.shape)
+
         target = torch.masked_select(labels, mask).to(self.device)
-        print('target', target)
-        print('target shape', target.shape)
+
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
             loss = (

From 46eea23667ace3a69138245e11f25f41386ea4c7 Mon Sep 17 00:00:00 2001
From: "Youssef.mecky" <youssef.mecky96@gmail.com>
Date: Wed, 25 Oct 2023 15:09:19 +0200
Subject: [PATCH 088/142] changes for quantus

---
 icu_benchmarks/models/train.py | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 1caaf9c6..bf6ab4f9 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -272,15 +272,15 @@ def train_common(
 
     if explain:
         explanintations = model.explantation_captum(
-            test_loader, log_dir, IntegratedGradients, target=(0, 1)
+            test_loader, log_dir, IntegratedGradients, target=0
         )
 
         print("IG", explanintations)
         interperations = model.interpertations(test_loader, log_dir)
         print("attention", interperations)
     if XAI_metric:
-        batch = iter(test_loader).next()
-
+        batch = next(iter(test_loader))
+        model = model.cpu()
         for key, value in batch[0].items():
             batch[0][key] = batch[0][key].to(model.device)
 
@@ -316,7 +316,7 @@ def train_common(
         # Reformat attributions.
         attr, delta = explantation.attribute(
             data,
-            target=(0, 1),
+            target=0,
             return_convergence_delta=True,
             baselines=baselines,
             n_steps=20,

From 16cc5c81b46623a94603fcdb1f8f5f2944203647 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 31 Oct 2023 14:34:17 +0100
Subject: [PATCH 089/142] added faithfulness correlation manuelly

---
 icu_benchmarks/models/dl_models.py | 108 +++++++++++++++++++++++++----
 icu_benchmarks/models/train.py     |  67 ++----------------
 2 files changed, 99 insertions(+), 76 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 12a7213a..6f321709 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -12,7 +12,7 @@
 )
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor, FloatTensor, zeros_like
+from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss
 import matplotlib.pyplot as plt
@@ -603,19 +603,19 @@ def explantation_captum(self, test_loader, log_dir, method, target):
         for batch in test_loader:
             for key, value in batch[0].items():
                 batch[0][key] = batch[0][key].to(self.device)
-
+            x = batch[0]
             data = (
-                batch[0]["encoder_cat"].float().requires_grad_(),
-                batch[0]["encoder_cont"].requires_grad_(),
-                batch[0]["encoder_target"].float().requires_grad_(),
-                batch[0]["encoder_lengths"].float().requires_grad_(),
-                batch[0]["decoder_cat"].float().requires_grad_(),
-                batch[0]["decoder_cont"].requires_grad_(),
-                batch[0]["decoder_target"].float().requires_grad_(),
-                batch[0]["decoder_lengths"].float().requires_grad_(),
-                batch[0]["decoder_time_idx"].float().requires_grad_(),
-                batch[0]["groups"].float().requires_grad_(),
-                batch[0]["target_scale"].requires_grad_(),
+                x["encoder_cat"].float().requires_grad_(),
+                x["encoder_cont"].requires_grad_(),
+                x["encoder_target"].float().requires_grad_(),
+                x["encoder_lengths"].float().requires_grad_(),
+                x["decoder_cat"].float().requires_grad_(),
+                x["decoder_cont"].requires_grad_(),
+                x["decoder_target"].float().requires_grad_(),
+                x["decoder_lengths"].float().requires_grad_(),
+                x["decoder_time_idx"].float().requires_grad_(),
+                x["groups"].float().requires_grad_(),
+                x["target_scale"].requires_grad_(),
             )
 
             baselines = (
@@ -642,7 +642,6 @@ def explantation_captum(self, test_loader, log_dir, method, target):
 
         # Concatenate a‚ttribution values for all instances along the batch dimension
         all_attrs = np.concatenate(all_attrs, axis=0)
-        print(all_attrs.shape)
         means_feature = all_attrs.mean(axis=(0, 1))
 
         # Compute mean along the batch dimension
@@ -689,6 +688,87 @@ def explantation_captum(self, test_loader, log_dir, method, target):
         plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
         return means
 
+    def faithfulness_correlation(self, test_loader, attribution, nr_runs=100, pertrub=None, subset_size=4):
+        """
+    Implementation of faithfulness correlation by Bhatt et al., 2020.
+
+    The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
+    (or 'fidelity') with respect to the model behaviour.
+
+    Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
+    the average explanation attribution for only the subset of features are (linearly) correlated, taking the
+    average over multiple runs and test samples. The metric returns one float per input-attribution pair that
+    ranges between -1 and 1, where higher scores are better.
+
+    For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
+    or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
+    test point and the average explanation attribution for only the subset of features is calculated. Results is
+    average over multiple runs and several test samples.
+
+    References:
+        1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
+        explanations." IJCAI (2020): 3016-3022.
+    """
+        if pertrub == None:
+            pertrub = "baseline"
+        similarities = []
+        for batch in test_loader:
+
+            for key, value in batch[0].items():
+
+                batch[0][key] = batch[0][key].to(self.device)
+            x = batch[0]
+            data = (
+                x["encoder_cat"],
+                x["encoder_cont"],
+                x["encoder_target"],
+                x["encoder_lengths"],
+                x["decoder_cat"],
+                x["decoder_cont"],
+                x["decoder_target"],
+                x["decoder_lengths"],
+                x["decoder_time_idx"],
+                x["groups"],
+                x["target_scale"],
+            )
+
+            y_pred = self(data).detach().cpu().numpy()
+            pred_deltas = []
+            att_sums = []
+            for i_ix in range(nr_runs):
+                # Randomly mask by subset size.
+                a_ix = np.random.choice(x["encoder_cont"].shape[1], subset_size, replace=False)
+                print(a_ix)
+                print(x["encoder_cont"][:, a_ix, :])
+
+                if pertrub == "Noise":
+                    noise = randn_like(x["encoder_cont"])
+                    x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
+                elif pertrub == "baseline":
+                    # Create a mask tensor with zeros at specified time steps and ones everywhere else
+                    mask = ones_like(x["encoder_cont"])
+                    print("Shape of mask:", mask.shape, mask.device)
+                    print("Shape of a_ix:", a_ix.shape, a_ix.device)
+
+                    mask[:, a_ix, :] = 0
+                    x["encoder_cont"] = x["encoder_cont"] * mask
+
+                # Predict on perturbed input x.
+                print(x["encoder_cont"][:, a_ix, :])
+                y_pred_perturb = self(data).detach().cpu().numpy()
+                pred_deltas.append((y_pred - y_pred_perturb).mean(axis=1))
+
+                # Sum attributions of the random subset.
+                att_sums.append(np.sum(attribution[a_ix]))
+            print(np.asarray(pred_deltas).shape)
+            print(np.asarray(att_sums).shape)
+            correlation_matrix = np.corrcoef(pred_deltas, att_sums, rowvar=False)
+
+            # Get the correlation coefficient from the correlation matrix
+            pearson_correlation = correlation_matrix[0, 1]
+            similarities.append(pearson_correlation)
+        return similarities.mean()
+
 
 @gin.configurable
 class RNNpytorch(DLPredictionWrapper):
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index bf6ab4f9..d8858250 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -272,75 +272,18 @@ def train_common(
 
     if explain:
         explanintations = model.explantation_captum(
-            test_loader, log_dir, IntegratedGradients, target=0
+            test_loader, log_dir, IntegratedGradients, target=(0, 1)
         )
 
         print("IG", explanintations)
         interperations = model.interpertations(test_loader, log_dir)
         print("attention", interperations)
     if XAI_metric:
-        batch = next(iter(test_loader))
-        model = model.cpu()
-        for key, value in batch[0].items():
-            batch[0][key] = batch[0][key].to(model.device)
-
-        data = (
-            batch[0]["encoder_cat"].float().requires_grad_(),
-            batch[0]["encoder_cont"].requires_grad_(),
-            batch[0]["encoder_target"].float().requires_grad_(),
-            batch[0]["encoder_lengths"].float().requires_grad_(),
-            batch[0]["decoder_cat"].float().requires_grad_(),
-            batch[0]["decoder_cont"].requires_grad_(),
-            batch[0]["decoder_target"].float().requires_grad_(),
-            batch[0]["decoder_lengths"].float().requires_grad_(),
-            batch[0]["decoder_time_idx"].float().requires_grad_(),
-            batch[0]["groups"].float().requires_grad_(),
-            batch[0]["target_scale"].requires_grad_(),
-        )
-
-        baselines = (
-            torch.zeros_like(data[0]).to(model.device),  # encoder_cat, set to zero
-            torch.zeros_like(data[1]).to(model.device),  # encoder_cont, set to zero
-            torch.zeros_like(data[2]).to(model.device),  # encoder_target, set to zero
-            data[3].to(model.device),  # encoder_lengths, leave unchanged
-            torch.zeros_like(data[4]).to(model.device),  # decoder_cat, set to zero
-            torch.zeros_like(data[5]).to(model.device),  # decoder_cont, set to zero
-            torch.zeros_like(data[6]).to(model.device),  # decoder_target, set to zero
-            data[7].to(model.device),  # decoder_lengths, leave unchanged
-            torch.zeros_like(data[8]).to(model.device),  # decoder_time_idx, set to zero
-            data[9].to(model.device),  # groups, leave unchanged
-            data[10].to(model.device),  # target_scale, leave unchanged
-        )
-
-        explantation = IntegratedGradients(model.forward_captum)
-        # Reformat attributions.
-        attr, delta = explantation.attribute(
-            data,
-            target=0,
-            return_convergence_delta=True,
-            baselines=baselines,
-            n_steps=20,
-        )
-        a_batch_IG = attr.detach().numpy()
-
-        x_batch, y_batch = batch[0], batch[1][0]
-        for key, value in x_batch.items():
-            x_batch[key] = x_batch[key].cpu().numpy()
-        y_batch = y_batch.cpu().numpy()
-        metric = quantus.FaithfulnessCorrelation()
-        scores = metric(
-            model=model,
-            x_batch=x_batch,
-            y_batch=y_batch,
-            a_batch=a_batch_IG,
-            device=model.device,
-            explain_func=quantus.explain,
-            explain_func_kwargs={
-                "method": "IntegratedGradients",
-                "baselines": baselines,
-                "n_steps": 20,
-            },
+        explanintations = model.explantation_captum(
+            test_loader, log_dir, IntegratedGradients, target=(0, 1)
         )
+        scores = model.faithfulness_correlation(test_loader, explanintations)
+        print(scores)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][

From 85216b77ce7c1250fdbd82cba9487b1f2db62df8 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 31 Oct 2023 14:54:46 +0100
Subject: [PATCH 090/142] fixed a few bugs to get the faithfulness correlation
 to work

---
 icu_benchmarks/models/dl_models.py | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 6f321709..1db5a8a1 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -12,7 +12,7 @@
 )
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like
+from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, from_numpy
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss
 import matplotlib.pyplot as plt
@@ -704,10 +704,12 @@ def faithfulness_correlation(self, test_loader, attribution, nr_runs=100, pertru
     or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
     test point and the average explanation attribution for only the subset of features is calculated. Results is
     average over multiple runs and several test samples.
+    This code is adapted from the quantus libray to suit our use case
 
     References:
         1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
         explanations." IJCAI (2020): 3016-3022.
+        2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
     """
         if pertrub == None:
             pertrub = "baseline"
@@ -738,25 +740,27 @@ def faithfulness_correlation(self, test_loader, attribution, nr_runs=100, pertru
             for i_ix in range(nr_runs):
                 # Randomly mask by subset size.
                 a_ix = np.random.choice(x["encoder_cont"].shape[1], subset_size, replace=False)
-                print(a_ix)
-                print(x["encoder_cont"][:, a_ix, :])
+
+                # Move a_ix_tensor to the same device as mask
 
                 if pertrub == "Noise":
+                    # add normal noise to input
                     noise = randn_like(x["encoder_cont"])
+
                     x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
                 elif pertrub == "baseline":
                     # Create a mask tensor with zeros at specified time steps and ones everywhere else
-                    mask = ones_like(x["encoder_cont"])
-                    print("Shape of mask:", mask.shape, mask.device)
-                    print("Shape of a_ix:", a_ix.shape, a_ix.device)
+                    # pytorch bug need to change to cpu for next step and then revert
+                    mask = ones_like(x["encoder_cont"]).cpu()
 
                     mask[:, a_ix, :] = 0
+                    mask = mask.to(x["encoder_cont"].device)
+
                     x["encoder_cont"] = x["encoder_cont"] * mask
 
                 # Predict on perturbed input x.
-                print(x["encoder_cont"][:, a_ix, :])
                 y_pred_perturb = self(data).detach().cpu().numpy()
-                pred_deltas.append((y_pred - y_pred_perturb).mean(axis=1))
+                pred_deltas.append((y_pred - y_pred_perturb).mean(axis=(0, 2)))
 
                 # Sum attributions of the random subset.
                 att_sums.append(np.sum(attribution[a_ix]))
@@ -767,7 +771,7 @@ def faithfulness_correlation(self, test_loader, attribution, nr_runs=100, pertru
             # Get the correlation coefficient from the correlation matrix
             pearson_correlation = correlation_matrix[0, 1]
             similarities.append(pearson_correlation)
-        return similarities.mean()
+        return np.mean(similarities)
 
 
 @gin.configurable

From 119dbcf20a92e525732e26aa2f7be1faf89cc99f Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 31 Oct 2023 15:49:03 +0100
Subject: [PATCH 091/142] added pytorch forecasting wrapper to make it easier

---
 icu_benchmarks/models/dl_models.py |  98 ++-----------
 icu_benchmarks/models/train.py     |  19 ++-
 icu_benchmarks/models/wrappers.py  | 214 +++++++++++++++++++++++++++++
 3 files changed, 231 insertions(+), 100 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 1db5a8a1..17dd9260 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -11,8 +11,8 @@
     PositionalEncoding,
 )
 import matplotlib.pyplot as plt
-from icu_benchmarks.models.wrappers import DLPredictionWrapper
-from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, from_numpy
+from icu_benchmarks.models.wrappers import DLPredictionWrapper, DLPredictionPytorchForecastingWrapper
+from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, is_tensor
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss
 import matplotlib.pyplot as plt
@@ -464,7 +464,7 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
 
 
 @gin.configurable
-class TFTpytorch(DLPredictionWrapper):
+class TFTpytorch(DLPredictionPytorchForecastingWrapper):
     """
     Implementation of https://arxiv.org/abs/1912.09363 from pytorch forecasting
     """
@@ -534,6 +534,7 @@ def actual_vs_predictions_plot(self, dataloader):
         return predictions_vs_actuals
 
     def interpertations(self, dataloader, log_dir):
+
         raw_predictions = self.model.predict(dataloader, return_x=True, mode="raw")
         interpretation = self.model.interpret_output(
             raw_predictions.output, reduction="sum"
@@ -542,6 +543,8 @@ def interpertations(self, dataloader, log_dir):
         for key, fig in figs.items():
             fig.savefig(log_dir / f"interpretation_{key}.png", bbox_inches="tight")
 
+        self.model = self.model.to(self.device)
+
         return interpretation
 
     def predict_dependency(self, dataloader, variable, log_dir):
@@ -688,94 +691,9 @@ def explantation_captum(self, test_loader, log_dir, method, target):
         plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
         return means
 
-    def faithfulness_correlation(self, test_loader, attribution, nr_runs=100, pertrub=None, subset_size=4):
-        """
-    Implementation of faithfulness correlation by Bhatt et al., 2020.
-
-    The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
-    (or 'fidelity') with respect to the model behaviour.
-
-    Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
-    the average explanation attribution for only the subset of features are (linearly) correlated, taking the
-    average over multiple runs and test samples. The metric returns one float per input-attribution pair that
-    ranges between -1 and 1, where higher scores are better.
-
-    For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
-    or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
-    test point and the average explanation attribution for only the subset of features is calculated. Results is
-    average over multiple runs and several test samples.
-    This code is adapted from the quantus libray to suit our use case
-
-    References:
-        1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
-        explanations." IJCAI (2020): 3016-3022.
-        2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
-    """
-        if pertrub == None:
-            pertrub = "baseline"
-        similarities = []
-        for batch in test_loader:
-
-            for key, value in batch[0].items():
-
-                batch[0][key] = batch[0][key].to(self.device)
-            x = batch[0]
-            data = (
-                x["encoder_cat"],
-                x["encoder_cont"],
-                x["encoder_target"],
-                x["encoder_lengths"],
-                x["decoder_cat"],
-                x["decoder_cont"],
-                x["decoder_target"],
-                x["decoder_lengths"],
-                x["decoder_time_idx"],
-                x["groups"],
-                x["target_scale"],
-            )
-
-            y_pred = self(data).detach().cpu().numpy()
-            pred_deltas = []
-            att_sums = []
-            for i_ix in range(nr_runs):
-                # Randomly mask by subset size.
-                a_ix = np.random.choice(x["encoder_cont"].shape[1], subset_size, replace=False)
-
-                # Move a_ix_tensor to the same device as mask
-
-                if pertrub == "Noise":
-                    # add normal noise to input
-                    noise = randn_like(x["encoder_cont"])
-
-                    x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
-                elif pertrub == "baseline":
-                    # Create a mask tensor with zeros at specified time steps and ones everywhere else
-                    # pytorch bug need to change to cpu for next step and then revert
-                    mask = ones_like(x["encoder_cont"]).cpu()
-
-                    mask[:, a_ix, :] = 0
-                    mask = mask.to(x["encoder_cont"].device)
-
-                    x["encoder_cont"] = x["encoder_cont"] * mask
-
-                # Predict on perturbed input x.
-                y_pred_perturb = self(data).detach().cpu().numpy()
-                pred_deltas.append((y_pred - y_pred_perturb).mean(axis=(0, 2)))
-
-                # Sum attributions of the random subset.
-                att_sums.append(np.sum(attribution[a_ix]))
-            print(np.asarray(pred_deltas).shape)
-            print(np.asarray(att_sums).shape)
-            correlation_matrix = np.corrcoef(pred_deltas, att_sums, rowvar=False)
-
-            # Get the correlation coefficient from the correlation matrix
-            pearson_correlation = correlation_matrix[0, 1]
-            similarities.append(pearson_correlation)
-        return np.mean(similarities)
-
 
 @gin.configurable
-class RNNpytorch(DLPredictionWrapper):
+class RNNpytorch(DLPredictionPytorchForecastingWrapper):
     """
     Implementation of RNN from pytorch forecasting
     """
@@ -835,7 +753,7 @@ def forward(
 
 
 @gin.configurable
-class DeepARpytorch(DLPredictionWrapper):
+class DeepARpytorch(DLPredictionPytorchForecastingWrapper):
     """
     Implementation of RNN from pytorch forecasting
     """
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index d8858250..e9f563da 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -271,19 +271,18 @@ def train_common(
         return 0
 
     if explain:
-        explanintations = model.explantation_captum(
+        attributions_IG = model.explantation_captum(
             test_loader, log_dir, IntegratedGradients, target=(0, 1)
         )
 
-        print("IG", explanintations)
-        interperations = model.interpertations(test_loader, log_dir)
-        print("attention", interperations)
-    if XAI_metric:
-        explanintations = model.explantation_captum(
-            test_loader, log_dir, IntegratedGradients, target=(0, 1)
-        )
-        scores = model.faithfulness_correlation(test_loader, explanintations)
-        print(scores)
+        # print("IG", attributions_IG)
+        # Attention_weights = model.interpertations(test_loader, log_dir)
+        # print("attention", Attention_weights)
+        if XAI_metric:
+            scores = model.faithfulness_correlation(test_loader, attributions_IG)
+            print('Attributions faithfulness correlation', scores)
+            # scores = model.faithfulness_correlation(test_loader, Attention_weights["attention"])
+            print('Attention faithfulness correlation', scores)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index ec467b75..892477d5 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -438,6 +438,220 @@ def step_fn(self, element, step_prefix=""):
         return loss
 
 
+@gin.configurable("DLPredictionPytorchForecastingWrapper")
+class DLPredictionPytorchForecastingWrapper(DLPredictionWrapper):
+    """Interface for Deep Learning models."""
+
+    _supported_run_modes = [RunMode.classification, RunMode.regression]
+
+    def __init__(
+        self,
+        loss=CrossEntropyLoss(),
+        optimizer=torch.optim.Adam,
+        run_mode: RunMode = RunMode.classification,
+        input_shape=None,
+        lr: float = 0.002,
+        momentum: float = 0.9,
+        lr_scheduler: Optional[str] = None,
+        lr_factor: float = 0.99,
+        lr_steps: Optional[List[int]] = None,
+        epochs: int = 100,
+        input_size: Tensor = None,
+        initialization_method: str = "normal",
+        pytorch_forecasting: bool = False,
+        **kwargs,
+    ):
+        super().__init__(
+            loss=loss,
+            optimizer=optimizer,
+            run_mode=run_mode,
+            input_shape=input_shape,
+            lr=lr,
+            momentum=momentum,
+            lr_scheduler=lr_scheduler,
+            lr_factor=lr_factor,
+            lr_steps=lr_steps,
+            epochs=epochs,
+            input_size=input_size,
+            initialization_method=initialization_method,
+            kwargs=kwargs,
+        )
+
+    def step_fn(self, element, step_prefix=""):
+        """Perform a step in the DL prediction model training loop.
+
+        Args:
+            element (object):
+            step_prefix (str): Step type, by default: test, train, val.
+        """
+
+        dic, labels = element[0], element[1][0]
+
+        if isinstance(labels, list):
+            labels = labels[-1]
+        data = (
+            dic["encoder_cat"],
+            dic["encoder_cont"],
+            dic["encoder_target"],
+            dic["encoder_lengths"],
+            dic["decoder_cat"],
+            dic["decoder_cont"],
+            dic["decoder_target"],
+            dic["decoder_lengths"],
+            dic["decoder_time_idx"],
+            dic["groups"],
+            dic["target_scale"],
+        )
+
+        mask = torch.ones_like(labels).bool()
+
+        out = self(data)
+
+        # If aux_loss is present, it is returned as a tuple
+        if len(out) == 2 and isinstance(out, tuple):
+            out, aux_loss = out
+        else:
+            aux_loss = 0
+        # Get prediction and target
+
+        prediction = (
+            torch.masked_select(out, mask.unsqueeze(-1))
+            .reshape(-1, out.shape[-1])
+            .to(self.device)
+        )
+
+        target = torch.masked_select(labels, mask).to(self.device)
+
+        if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
+            # Classification task
+            loss = (
+                self.loss(
+                    prediction, target.long(), weight=self.loss_weights.to(self.device)
+                )
+                + aux_loss
+            )
+            # Returns torch.long because negative log likelihood loss
+        elif self.run_mode == RunMode.regression:
+            # Regression task
+            loss = self.loss(prediction[:, 0], target.float()) + aux_loss
+        else:
+            raise ValueError(
+                f"Run mode {self.run_mode} not yet supported. Please implement it."
+            )
+        transformed_output = self.output_transform((prediction, target))
+
+        for key, value in self.metrics[step_prefix].items():
+            if isinstance(value, torchmetrics.Metric):
+                if key == "Binary_Fairness":
+                    feature_names = key.feature_helper(self.trainer)
+                    value.update(
+                        transformed_output[0],
+                        transformed_output[1],
+                        data,
+                        feature_names,
+                    )
+                else:
+                    value.update(transformed_output[0], transformed_output[1])
+            else:
+                value.update(transformed_output)
+        self.log(
+            f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True
+        )
+        return loss
+
+    def faithfulness_correlation(self, test_loader, attribution, nr_runs=100, pertrub=None, subset_size=4):
+        """
+    Implementation of faithfulness correlation by Bhatt et al., 2020.
+
+    The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
+    (or 'fidelity') with respect to the model behaviour.
+
+    Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
+    the average explanation attribution for only the subset of features are (linearly) correlated, taking the
+    average over multiple runs and test samples. The metric returns one float per input-attribution pair that
+    ranges between -1 and 1, where higher scores are better.
+
+    For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
+    or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
+    test point and the average explanation attribution for only the subset of features is calculated. Results is
+    average over multiple runs and several test samples.
+    This code is adapted from the quantus libray to suit our use case
+
+    References:
+        1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
+        explanations." IJCAI (2020): 3016-3022.
+        2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+    """
+
+        if torch.is_tensor(attribution):
+            # Convert the tensor to a NumPy array
+            example_numpy_array = attribution.cpu().detach().numpy()
+        if pertrub == None:
+            pertrub = "baseline"
+        similarities = []
+        for batch in test_loader:
+
+            for key, value in batch[0].items():
+
+                batch[0][key] = batch[0][key].to(self.device)
+            x = batch[0]
+            data = (
+                x["encoder_cat"],
+                x["encoder_cont"],
+                x["encoder_target"],
+                x["encoder_lengths"],
+                x["decoder_cat"],
+                x["decoder_cont"],
+                x["decoder_target"],
+                x["decoder_lengths"],
+                x["decoder_time_idx"],
+                x["groups"],
+                x["target_scale"],
+            )
+
+            y_pred = self(data).detach().cpu().numpy()
+            pred_deltas = []
+            att_sums = []
+            for i_ix in range(nr_runs):
+                # Randomly mask by subset size.
+                a_ix = np.random.choice(x["encoder_cont"].shape[1], subset_size, replace=False)
+
+                # Move a_ix_tensor to the same device as mask
+
+                if pertrub == "Noise":
+                    # add normal noise to input
+                    noise = torch.randn_like(x["encoder_cont"])
+
+                    x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
+                elif pertrub == "baseline":
+                    # Create a mask tensor with zeros at specified time steps and ones everywhere else
+                    # pytorch bug need to change to cpu for next step and then revert
+                    mask = torch.ones_like(x["encoder_cont"]).cpu()
+
+                    mask[:, a_ix, :] = 0
+                    mask = mask.to(x["encoder_cont"].device)
+
+                    x["encoder_cont"] = x["encoder_cont"] * mask
+
+                # Predict on perturbed input x.
+                y_pred_perturb = self(data).detach().cpu().numpy()
+                print(y_pred - y_pred_perturb)
+                break
+                pred_deltas.append((y_pred - y_pred_perturb).mean(axis=(0, 2)))
+
+                # Sum attributions of the random subset.
+
+                att_sums.append(np.sum(attribution[a_ix]))
+            print(pred_deltas, att_sums)
+            correlation_matrix = np.corrcoef(pred_deltas, att_sums, rowvar=False)
+
+            # Get the correlation coefficient from the correlation matrix
+            pearson_correlation = correlation_matrix[0, 1]
+            similarities.append(pearson_correlation)
+        print(similarities)
+        return np.nanmean(similarities)
+
+
 @gin.configurable("MLWrapper")
 class MLWrapper(BaseModule, ABC):
     """Interface for prediction with traditional Scikit-learn-like Machine Learning models."""

From d03deedc2e2d58c6844ccdf19e88345e1c3d4760 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 31 Oct 2023 15:54:16 +0100
Subject: [PATCH 092/142] added pytorch forecasting wrapper to make it easier

---
 configs/tasks/BinaryClassification.gin | 2 +-
 configs/tasks/Regression.gin           | 1 +
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/configs/tasks/BinaryClassification.gin b/configs/tasks/BinaryClassification.gin
index 492a12eb..f3a7790d 100644
--- a/configs/tasks/BinaryClassification.gin
+++ b/configs/tasks/BinaryClassification.gin
@@ -16,7 +16,7 @@ train_common.ram_cache = True
 
 # DEEP LEARNING
 DLPredictionWrapper.loss = @cross_entropy
-
+DLPredictionPytorchForecastingWrapper.loss= @cross_entropy
 # SELECTING PREPROCESSOR
 preprocess.preprocessor = @base_classification_preprocessor
 preprocess.vars = %vars
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index 5cf3f8d9..6b6ac607 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -16,6 +16,7 @@ train_common.ram_cache = True
 
 # LOSS FUNCTION
 DLPredictionWrapper.loss = @mse_loss
+DLPredictionPytorchForecastingWrapper.loss = @mse_loss
 MLWrapper.loss = @mean_squared_error
 
 # SELECTING PREPROCESSOR

From 3c78359a3d3a059d44827f371dbb411e55640afe Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 31 Oct 2023 16:29:24 +0100
Subject: [PATCH 093/142] added data randomization test and fixed few issues
 with wrapper

---
 icu_benchmarks/models/dl_models.py |  93 ---------------
 icu_benchmarks/models/train.py     |   4 +-
 icu_benchmarks/models/wrappers.py  | 179 ++++++++++++++++++++++++++++-
 3 files changed, 175 insertions(+), 101 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 17dd9260..08a1e75f 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -598,99 +598,6 @@ def forward_captum(
 
         return self.forward(tuple_x)
 
-    def explantation_captum(self, test_loader, log_dir, method, target):
-        # Initialize lists to store attribution values for all instances
-        all_attrs = []
-
-        # Loop through the test_loader to compute attributions for all instances
-        for batch in test_loader:
-            for key, value in batch[0].items():
-                batch[0][key] = batch[0][key].to(self.device)
-            x = batch[0]
-            data = (
-                x["encoder_cat"].float().requires_grad_(),
-                x["encoder_cont"].requires_grad_(),
-                x["encoder_target"].float().requires_grad_(),
-                x["encoder_lengths"].float().requires_grad_(),
-                x["decoder_cat"].float().requires_grad_(),
-                x["decoder_cont"].requires_grad_(),
-                x["decoder_target"].float().requires_grad_(),
-                x["decoder_lengths"].float().requires_grad_(),
-                x["decoder_time_idx"].float().requires_grad_(),
-                x["groups"].float().requires_grad_(),
-                x["target_scale"].requires_grad_(),
-            )
-
-            baselines = (
-                zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
-                zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
-                zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
-                data[3].to(self.device),  # encoder_lengths, leave unchanged
-                zeros_like(data[4]).to(self.device),  # decoder_cat, set to zero
-                zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
-                zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
-                data[7].to(self.device),  # decoder_lengths, leave unchanged
-                zeros_like(data[8]).to(self.device),  # decoder_time_idx, set to zero
-                data[9].to(self.device),  # groups, leave unchanged
-                data[10].to(self.device),  # target_scale, leave unchanged
-            )
-
-            explantation = method(self.forward_captum)
-            # Reformat attributions.
-            attr, delta = explantation.attribute(
-                data, target=target, return_convergence_delta=True, baselines=baselines, n_steps=20
-            )
-            # Convert attributions to numpy array and append to the list
-            all_attrs.append(attr[0].cpu().detach().numpy())
-
-        # Concatenate a‚ttribution values for all instances along the batch dimension
-        all_attrs = np.concatenate(all_attrs, axis=0)
-        means_feature = all_attrs.mean(axis=(0, 1))
-
-        # Compute mean along the batch dimension
-        means = all_attrs.mean(axis=(0, 2))
-        # Normalize the means values to range [0, 1]
-        normalized_means = (means - means.min()) / (means.max() - means.min())
-
-        # Create x values (assuming you want a simple sequential x-axis)
-        # Assuming you have 24 values
-        x_values = np.arange(1, 57)
-        # Plotting the featrue means
-        plt.figure(figsize=(8, 6))
-        plt.plot(
-            x_values,
-            means_feature,
-            marker="o",
-            color="skyblue",
-            linestyle="-",
-            linewidth=2,
-            markersize=8,
-        )
-        plt.xlabel("Time Step")
-        plt.ylabel("Normalized Attribution")
-        plt.title("Attribution Values")
-        plt.xticks(x_values)  # Set x-ticks to match the number of features
-        plt.tight_layout()
-        plt.savefig(log_dir / "attribution_features_plot.png", bbox_inches="tight")
-        plt.figure(figsize=(8, 6))
-        x_values = np.arange(1, 25)
-        plt.plot(
-            x_values,
-            normalized_means,
-            marker="o",
-            color="skyblue",
-            linestyle="-",
-            linewidth=2,
-            markersize=8,
-        )
-        plt.xlabel("Time Step")
-        plt.ylabel("Normalized Attribution")
-        plt.title("Attribution Values")
-        plt.xticks(x_values)  # Set x-ticks to match the number of features
-        plt.tight_layout()
-        plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
-        return means
-
 
 @gin.configurable
 class RNNpytorch(DLPredictionPytorchForecastingWrapper):
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index e9f563da..eadb252f 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -272,14 +272,14 @@ def train_common(
 
     if explain:
         attributions_IG = model.explantation_captum(
-            test_loader, log_dir, IntegratedGradients, target=(0, 1)
+            test_loader, log_dir, IntegratedGradients
         )
 
         # print("IG", attributions_IG)
         # Attention_weights = model.interpertations(test_loader, log_dir)
         # print("attention", Attention_weights)
         if XAI_metric:
-            scores = model.faithfulness_correlation(test_loader, attributions_IG)
+            scores = model.Faithfulness_Correlation(test_loader, attributions_IG)
             print('Attributions faithfulness correlation', scores)
             # scores = model.faithfulness_correlation(test_loader, Attention_weights["attention"])
             print('Attention faithfulness correlation', scores)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 892477d5..c22d9bcb 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -18,7 +18,7 @@
 from pytorch_lightning import LightningModule
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
-
+import matplotlib.pyplot as plt
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(
     nn.functional.cross_entropy, module="torch.nn.functional"
@@ -559,7 +559,101 @@ def step_fn(self, element, step_prefix=""):
         )
         return loss
 
-    def faithfulness_correlation(self, test_loader, attribution, nr_runs=100, pertrub=None, subset_size=4):
+    def explantation_captum(self, test_loader, log_dir, method):
+        # Initialize lists to store attribution values for all instances
+        all_attrs = []
+
+        # Loop through the test_loader to compute attributions for all instances
+        for batch in test_loader:
+            for key, value in batch[0].items():
+                batch[0][key] = batch[0][key].to(self.device)
+            x = batch[0]
+            data = (
+                x["encoder_cat"].float().requires_grad_(),
+                x["encoder_cont"].requires_grad_(),
+                x["encoder_target"].float().requires_grad_(),
+                x["encoder_lengths"].float().requires_grad_(),
+                x["decoder_cat"].float().requires_grad_(),
+                x["decoder_cont"].requires_grad_(),
+                x["decoder_target"].float().requires_grad_(),
+                x["decoder_lengths"].float().requires_grad_(),
+                x["decoder_time_idx"].float().requires_grad_(),
+                x["groups"].float().requires_grad_(),
+                x["target_scale"].requires_grad_(),
+            )
+            target = self(data)
+            print(target.shape)
+            baselines = (
+                torch.zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
+                torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
+                torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
+                data[3].to(self.device),  # encoder_lengths, leave unchanged
+                torch.zeros_like(data[4]).to(self.device),  # decoder_cat, set to zero
+                torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
+                torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
+                data[7].to(self.device),  # decoder_lengths, leave unchanged
+                torch.zeros_like(data[8]).to(self.device),  # decoder_time_idx, set to zero
+                data[9].to(self.device),  # groups, leave unchanged
+                data[10].to(self.device),  # target_scale, leave unchanged
+            )
+
+            explantation = method(self.forward_captum)
+            # Reformat attributions.
+            attr, delta = explantation.attribute(
+                data, target=target, return_convergence_delta=True, baselines=baselines, n_steps=20
+            )
+            # Convert attributions to numpy array and append to the list
+            all_attrs.append(attr[0].cpu().detach().numpy())
+
+        # Concatenate a‚ttribution values for all instances along the batch dimension
+        all_attrs = np.concatenate(all_attrs, axis=0)
+        means_feature = all_attrs.mean(axis=(0, 1))
+
+        # Compute mean along the batch dimension
+        means = all_attrs.mean(axis=(0, 2))
+        # Normalize the means values to range [0, 1]
+        normalized_means = (means - means.min()) / (means.max() - means.min())
+
+        # Create x values (assuming you want a simple sequential x-axis)
+        # Assuming you have 24 values
+        x_values = np.arange(1, 57)
+        # Plotting the featrue means
+        plt.figure(figsize=(8, 6))
+        plt.plot(
+            x_values,
+            means_feature,
+            marker="o",
+            color="skyblue",
+            linestyle="-",
+            linewidth=2,
+            markersize=8,
+        )
+        plt.xlabel("Time Step")
+        plt.ylabel("Normalized Attribution")
+        plt.title("Attribution Values")
+        plt.xticks(x_values)  # Set x-ticks to match the number of features
+        plt.tight_layout()
+        plt.savefig(log_dir / "attribution_features_plot.png", bbox_inches="tight")
+        plt.figure(figsize=(8, 6))
+        x_values = np.arange(1, 25)
+        plt.plot(
+            x_values,
+            normalized_means,
+            marker="o",
+            color="skyblue",
+            linestyle="-",
+            linewidth=2,
+            markersize=8,
+        )
+        plt.xlabel("Time Step")
+        plt.ylabel("Normalized Attribution")
+        plt.title("Attribution Values")
+        plt.xticks(x_values)  # Set x-ticks to match the number of features
+        plt.tight_layout()
+        plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
+        return means
+
+    def Faithfulness_Correlation(self, test_loader, attribution, nr_runs=100, pertrub=None, subset_size=4):
         """
     Implementation of faithfulness correlation by Bhatt et al., 2020.
 
@@ -635,22 +729,95 @@ def faithfulness_correlation(self, test_loader, attribution, nr_runs=100, pertru
 
                 # Predict on perturbed input x.
                 y_pred_perturb = self(data).detach().cpu().numpy()
-                print(y_pred - y_pred_perturb)
-                break
+
                 pred_deltas.append((y_pred - y_pred_perturb).mean(axis=(0, 2)))
 
                 # Sum attributions of the random subset.
 
                 att_sums.append(np.sum(attribution[a_ix]))
-            print(pred_deltas, att_sums)
             correlation_matrix = np.corrcoef(pred_deltas, att_sums, rowvar=False)
 
             # Get the correlation coefficient from the correlation matrix
             pearson_correlation = correlation_matrix[0, 1]
             similarities.append(pearson_correlation)
-        print(similarities)
         return np.nanmean(similarities)
 
+    def Data_Randomization(self, test_loader, attribution, explain_method, nr_runs=100, similarity=None):
+        """
+        Implementation of the Random Logit Metric by Sixt et al., 2020.
+
+        The Random Logit Metric computes the distance between the original explanation and a reference explanation of
+        a randomly chosen non-target class.
+        This code is adapted from the quantus libray to suit our use case
+
+        References:
+            1) Leon Sixt et al.: "When Explanations Lie: Why Many Modified BP
+            Attributions Fail." ICML (2020): 9046-9057.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+
+        """
+        a_perturbed = []
+        for batch in test_loader:
+
+            for key, value in batch[0].items():
+
+                batch[0][key] = batch[0][key].to(self.device)
+            x = batch[0]
+            y = batch[1][0]
+            data = (
+                x["encoder_cat"],
+                x["encoder_cont"],
+                x["encoder_target"],
+                x["encoder_lengths"],
+                x["decoder_cat"],
+                x["decoder_cont"],
+                x["decoder_target"],
+                x["decoder_lengths"],
+                x["decoder_time_idx"],
+                x["groups"],
+                x["target_scale"],
+            )
+            y_off = np.array(
+                [
+                    np.random.choice(
+                        [y_ for y_ in list(np.arange(0, self.num_classes)) if y_ != y]
+                    )
+                ]
+            )
+            baselines = (
+                torch.zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
+                torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
+                torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
+                data[3].to(self.device),  # encoder_lengths, leave unchanged
+                torch.zeros_like(data[4]).to(self.device),  # decoder_cat, set to zero
+                torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
+                torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
+                data[7].to(self.device),  # decoder_lengths, leave unchanged
+                torch.zeros_like(data[8]).to(self.device),  # decoder_time_idx, set to zero
+                data[9].to(self.device),  # groups, leave unchanged
+                data[10].to(self.device),  # target_scale, leave unchanged
+            )
+
+            explantation = explain_method(self.forward_captum)
+            # Reformat attributions.
+            attr, delta = explantation.attribute(
+                data, target=y_off, return_convergence_delta=True, baselines=baselines, n_steps=20
+            )
+            # Convert attributions to numpy array and append to the list
+            a_perturbed.append(attr[0].cpu().detach().numpy())
+        a_perturbed = np.concatenate(a_perturbed, axis=0)
+        a_perturbed = a_perturbed.mean(axis=(0, 1))
+
+        # Compute mean along the batch dimension
+        a_perturbed = a_perturbed.mean(axis=(0, 2))
+        # Normalize the means values to range [0, 1]
+        normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
+        correlation_matrix = np.corrcoef(normalized_a_perturbed, attribution, rowvar=False)
+
+        # Get the correlation coefficient from the correlation matrix
+        pearson_correlation = correlation_matrix[0, 1]
+        return pearson_correlation
+
 
 @gin.configurable("MLWrapper")
 class MLWrapper(BaseModule, ABC):

From 8972c12c817032c52b9cd4ee3da2d4a44cdc6e19 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 31 Oct 2023 16:32:07 +0100
Subject: [PATCH 094/142] removed check for pytorch forecasting from DLwrapper

---
 icu_benchmarks/models/wrappers.py | 38 +++++--------------------------
 1 file changed, 6 insertions(+), 32 deletions(-)

diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index c22d9bcb..a89ed8d8 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -333,40 +333,14 @@ def step_fn(self, element, step_prefix=""):
                 data[key] = value
          """
         if len(element) == 2:
-            """
-            if (isinstance(element[1], tuple)) or (
-                isinstance(element[1], list)
-            ):
-            """
-            if self.pytorch_forecasting:
-                # check if the data loader is the one for the pytorch forecasring implementation
-                dic, labels = element[0], element[1][0]
-
-                if isinstance(labels, list):
-                    labels = labels[-1]
-                data = (
-                    dic["encoder_cat"],
-                    dic["encoder_cont"],
-                    dic["encoder_target"],
-                    dic["encoder_lengths"],
-                    dic["decoder_cat"],
-                    dic["decoder_cont"],
-                    dic["decoder_target"],
-                    dic["decoder_lengths"],
-                    dic["decoder_time_idx"],
-                    dic["groups"],
-                    dic["target_scale"],
-                )
 
-                mask = torch.ones_like(labels).bool()
+            data, labels = element[0], (element[1]).to(self.device)
+            if isinstance(data, list):
+                for i in range(len(data)):
+                    data[i] = data[i].float().to(self.device)
             else:
-                data, labels = element[0], (element[1]).to(self.device)
-                if isinstance(data, list):
-                    for i in range(len(data)):
-                        data[i] = data[i].float().to(self.device)
-                else:
-                    data = data.float().to(self.device)
-                mask = torch.ones_like(labels).bool()
+                data = data.float().to(self.device)
+            mask = torch.ones_like(labels).bool()
 
         elif len(element) == 3:
             data, labels, mask = (

From bd8c2e5f8520d836be64770d689fa648e094ceaa Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 31 Oct 2023 16:37:51 +0100
Subject: [PATCH 095/142] similiary function from captum

---
 icu_benchmarks/models/similarity_func.py | 292 +++++++++++++++++++++++
 1 file changed, 292 insertions(+)
 create mode 100644 icu_benchmarks/models/similarity_func.py

diff --git a/icu_benchmarks/models/similarity_func.py b/icu_benchmarks/models/similarity_func.py
new file mode 100644
index 00000000..88d19a9a
--- /dev/null
+++ b/icu_benchmarks/models/similarity_func.py
@@ -0,0 +1,292 @@
+"""This module holds a collection of similarity functions i.e., ways to measure the distance between two inputs (or explanations)."""
+
+# This file is part of Quantus.
+# Quantus is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
+# Quantus is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
+# You should have received a copy of the GNU Lesser General Public License along with Quantus. If not, see <https://www.gnu.org/licenses/>.
+# Quantus project URL: <https://github.com/understandable-machine-intelligence-lab/Quantus>.
+# Quantus project URL: https://github.com/understandable-machine-intelligence-lab/Quantus
+
+from typing import Union
+
+import numpy as np
+import scipy
+import skimage
+
+
+def correlation_spearman(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate Spearman rank of two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return scipy.stats.spearmanr(a, b)[0]
+
+
+def correlation_pearson(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate Pearson correlation of two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return scipy.stats.pearsonr(a, b)[0]
+
+
+def correlation_kendall_tau(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate Kendall Tau correlation of two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return scipy.stats.kendalltau(a, b)[0]
+
+
+def distance_euclidean(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate Euclidean distance of two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return scipy.spatial.distance.euclidean(u=a, v=b)
+
+
+def distance_manhattan(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate Manhattan distance of two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return scipy.spatial.distance.cityblock(u=a, v=b)
+
+
+def distance_chebyshev(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate Chebyshev distance of two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return scipy.spatial.distance.chebyshev(u=a, v=b)
+
+
+def lipschitz_constant(
+    a: np.array,
+    b: np.array,
+    c: Union[np.array, None],
+    d: Union[np.array, None],
+    **kwargs
+) -> float:
+    """
+    Calculate non-negative local Lipschitz abs(||a-b||/||c-d||), where a,b can be f(x) or a(x) and c,d is x.
+
+    For numerical stability, a small value is added to division.
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    c: np.ndarray
+         The third array to use for similarity scoring.
+    d: np.ndarray
+         The fourth array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    eps = 1e-10
+
+    d1 = kwargs.get("norm_numerator", distance_manhattan)
+    d2 = kwargs.get("norm_denominator", distance_euclidean)
+
+    if np.shape(a) == ():
+        return float(abs(a - b) / (d2(c, d) + eps))
+    else:
+        return float(d1(a, b) / (d2(a=c, b=d) + eps))
+
+
+def abs_difference(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate the mean of the absolute differences between two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return np.mean(abs(a - b))
+
+
+def squared_difference(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate the sqaured differences between two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return np.sum((a - b) ** 2)
+
+
+def cosine(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate Cosine of two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return scipy.spatial.distance.cosine(u=a, v=b)
+
+
+def ssim(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate Structural Similarity Index Measure of two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    max_point, min_point = np.max(np.concatenate([a, b])), np.min(
+        np.concatenate([a, b])
+    )
+    data_range = float(np.abs(max_point - min_point))
+    return skimage.metrics.structural_similarity(
+        im1=a, im2=b, win_size=kwargs.get("win_size", None), data_range=data_range
+    )
+
+
+def difference(a: np.array, b: np.array, **kwargs) -> np.array:
+    """
+    Calculate the difference between two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    np.array
+        The difference in each element.
+    """
+    return a - b

From d4a55ff2074851ec4b4b07f11e5546da527f56f4 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 31 Oct 2023 16:42:34 +0100
Subject: [PATCH 096/142] used similarty functions already defined

---
 icu_benchmarks/models/wrappers.py | 21 ++++++++++-----------
 1 file changed, 10 insertions(+), 11 deletions(-)

diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index a89ed8d8..9640dcd9 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -19,6 +19,7 @@
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
 import matplotlib.pyplot as plt
+from icu_benchmarks.models.similarity_func import correlation_spearman,distance_euclidean
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(
     nn.functional.cross_entropy, module="torch.nn.functional"
@@ -627,7 +628,7 @@ def explantation_captum(self, test_loader, log_dir, method):
         plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
         return means
 
-    def Faithfulness_Correlation(self, test_loader, attribution, nr_runs=100, pertrub=None, subset_size=4):
+    def Faithfulness_Correlation(self, test_loader, attribution,similarity_func=None nr_runs=100, pertrub=None, subset_size=4):
         """
     Implementation of faithfulness correlation by Bhatt et al., 2020.
 
@@ -654,6 +655,8 @@ def Faithfulness_Correlation(self, test_loader, attribution, nr_runs=100, pertru
         if torch.is_tensor(attribution):
             # Convert the tensor to a NumPy array
             example_numpy_array = attribution.cpu().detach().numpy()
+        if similarity_func== None:
+            similarity_func=correlation_spearman
         if pertrub == None:
             pertrub = "baseline"
         similarities = []
@@ -709,11 +712,8 @@ def Faithfulness_Correlation(self, test_loader, attribution, nr_runs=100, pertru
                 # Sum attributions of the random subset.
 
                 att_sums.append(np.sum(attribution[a_ix]))
-            correlation_matrix = np.corrcoef(pred_deltas, att_sums, rowvar=False)
-
-            # Get the correlation coefficient from the correlation matrix
-            pearson_correlation = correlation_matrix[0, 1]
-            similarities.append(pearson_correlation)
+            
+            similarities.append(similarity_func(pred_deltas,att_sums))
         return np.nanmean(similarities)
 
     def Data_Randomization(self, test_loader, attribution, explain_method, nr_runs=100, similarity=None):
@@ -731,6 +731,8 @@ def Data_Randomization(self, test_loader, attribution, explain_method, nr_runs=1
 
         """
         a_perturbed = []
+        if similarity_func== None:
+            similarity_func=distance_euclidean
         for batch in test_loader:
 
             for key, value in batch[0].items():
@@ -786,11 +788,8 @@ def Data_Randomization(self, test_loader, attribution, explain_method, nr_runs=1
         a_perturbed = a_perturbed.mean(axis=(0, 2))
         # Normalize the means values to range [0, 1]
         normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
-        correlation_matrix = np.corrcoef(normalized_a_perturbed, attribution, rowvar=False)
-
-        # Get the correlation coefficient from the correlation matrix
-        pearson_correlation = correlation_matrix[0, 1]
-        return pearson_correlation
+        
+        return similarity_func(normalized_a_perturbed,attribution)
 
 
 @gin.configurable("MLWrapper")

From edd4147a554121d7ce0915e5b05c4eddf3024e99 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 2 Nov 2023 14:10:58 +0100
Subject: [PATCH 097/142] latest changes

---
 icu_benchmarks/models/dl_models.py |  15 +-
 icu_benchmarks/models/train.py     |  17 ++-
 icu_benchmarks/models/wrappers.py  | 227 +++++++++++++++--------------
 3 files changed, 136 insertions(+), 123 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 08a1e75f..7fb7fd27 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -494,7 +494,7 @@ def __init__(
             loss=QuantileLoss(),
             hidden_continuous_size=hidden,
         )
-
+        self.num_classes = num_classes
         self.logit = nn.Linear(7, num_classes)
 
     def forward(
@@ -533,15 +533,16 @@ def actual_vs_predictions_plot(self, dataloader):
         self.model.plot_prediction_actual_by_variable(predictions_vs_actuals)
         return predictions_vs_actuals
 
-    def interpertations(self, dataloader, log_dir):
+    def interpertations(self, dataloader, log_dir='.', plot=False):
 
         raw_predictions = self.model.predict(dataloader, return_x=True, mode="raw")
         interpretation = self.model.interpret_output(
             raw_predictions.output, reduction="sum"
         )
-        figs = self.model.plot_interpretation(interpretation)
-        for key, fig in figs.items():
-            fig.savefig(log_dir / f"interpretation_{key}.png", bbox_inches="tight")
+        if (plot):
+            figs = self.model.plot_interpretation(interpretation)
+            for key, fig in figs.items():
+                fig.savefig(log_dir / f"interpretation_{key}.png", bbox_inches="tight")
 
         self.model = self.model.to(self.device)
 
@@ -630,7 +631,7 @@ def __init__(
             dropout=dropout,
             optimizer=optimizer,
         )
-
+        self.num_classes = num_classes
         self.logit = nn.Linear(1, num_classes)
 
     def forward(
@@ -690,7 +691,7 @@ def __init__(
             dropout=dropout,
             optimizer=optimizer,
         )
-
+        self.num_classes = num_classes
         self.logit = nn.Linear(4, num_classes)
 
     def forward(
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index eadb252f..b7c966c2 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -275,14 +275,19 @@ def train_common(
             test_loader, log_dir, IntegratedGradients
         )
 
-        # print("IG", attributions_IG)
-        # Attention_weights = model.interpertations(test_loader, log_dir)
-        # print("attention", Attention_weights)
+        print("IG", attributions_IG)
+        Attention_weights = model.interpertations(test_loader, log_dir)
+        print("attention", Attention_weights)
         if XAI_metric:
-            scores = model.Faithfulness_Correlation(test_loader, attributions_IG)
-            print('Attributions faithfulness correlation', scores)
+            scores1 = model.Faithfulness_Correlation(test_loader, attributions_IG)
+            print('Attributions faithfulness correlation', scores1)
+            scores2 = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"])
+            print('Attention faithfulness correlation', scores2)
             # scores = model.faithfulness_correlation(test_loader, Attention_weights["attention"])
-            print('Attention faithfulness correlation', scores)
+
+            scores_2 = model.Data_Randomization(
+                test_loader, attributions_IG, IntegratedGradients)
+            print('Distance Data randmoization score', scores_2)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 9640dcd9..45c44771 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -19,7 +19,7 @@
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
 import matplotlib.pyplot as plt
-from icu_benchmarks.models.similarity_func import correlation_spearman,distance_euclidean
+from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(
     nn.functional.cross_entropy, module="torch.nn.functional"
@@ -534,7 +534,7 @@ def step_fn(self, element, step_prefix=""):
         )
         return loss
 
-    def explantation_captum(self, test_loader, log_dir, method):
+    def explantation_captum(self, test_loader, method, log_dir='.', plot=False):
         # Initialize lists to store attribution values for all instances
         all_attrs = []
 
@@ -556,8 +556,6 @@ def explantation_captum(self, test_loader, log_dir, method):
                 x["groups"].float().requires_grad_(),
                 x["target_scale"].requires_grad_(),
             )
-            target = self(data)
-            print(target.shape)
             baselines = (
                 torch.zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
                 torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
@@ -575,60 +573,60 @@ def explantation_captum(self, test_loader, log_dir, method):
             explantation = method(self.forward_captum)
             # Reformat attributions.
             attr, delta = explantation.attribute(
-                data, target=target, return_convergence_delta=True, baselines=baselines, n_steps=20
+                data, target=(0, 1), return_convergence_delta=True, baselines=baselines, n_steps=35
             )
             # Convert attributions to numpy array and append to the list
             all_attrs.append(attr[0].cpu().detach().numpy())
 
         # Concatenate a‚ttribution values for all instances along the batch dimension
-        all_attrs = np.concatenate(all_attrs, axis=0)
+        all_attrs = np.mean(all_attrs, axis=0)
         means_feature = all_attrs.mean(axis=(0, 1))
 
         # Compute mean along the batch dimension
         means = all_attrs.mean(axis=(0, 2))
         # Normalize the means values to range [0, 1]
         normalized_means = (means - means.min()) / (means.max() - means.min())
-
-        # Create x values (assuming you want a simple sequential x-axis)
-        # Assuming you have 24 values
-        x_values = np.arange(1, 57)
-        # Plotting the featrue means
-        plt.figure(figsize=(8, 6))
-        plt.plot(
-            x_values,
-            means_feature,
-            marker="o",
-            color="skyblue",
-            linestyle="-",
-            linewidth=2,
-            markersize=8,
-        )
-        plt.xlabel("Time Step")
-        plt.ylabel("Normalized Attribution")
-        plt.title("Attribution Values")
-        plt.xticks(x_values)  # Set x-ticks to match the number of features
-        plt.tight_layout()
-        plt.savefig(log_dir / "attribution_features_plot.png", bbox_inches="tight")
-        plt.figure(figsize=(8, 6))
-        x_values = np.arange(1, 25)
-        plt.plot(
-            x_values,
-            normalized_means,
-            marker="o",
-            color="skyblue",
-            linestyle="-",
-            linewidth=2,
-            markersize=8,
-        )
-        plt.xlabel("Time Step")
-        plt.ylabel("Normalized Attribution")
-        plt.title("Attribution Values")
-        plt.xticks(x_values)  # Set x-ticks to match the number of features
-        plt.tight_layout()
-        plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
+        if plot:
+            # Create x values (assuming you want a simple sequential x-axis)
+            # Assuming you have 24 values
+            x_values = np.arange(1, 57)
+            # Plotting the featrue means
+            plt.figure(figsize=(8, 6))
+            plt.plot(
+                x_values,
+                means_feature,
+                marker="o",
+                color="skyblue",
+                linestyle="-",
+                linewidth=2,
+                markersize=8,
+            )
+            plt.xlabel("Time Step")
+            plt.ylabel("Normalized Attribution")
+            plt.title("Attribution Values")
+            plt.xticks(x_values)  # Set x-ticks to match the number of features
+            plt.tight_layout()
+            plt.savefig(log_dir / "attribution_features_plot.png", bbox_inches="tight")
+            plt.figure(figsize=(8, 6))
+            x_values = np.arange(1, 25)
+            plt.plot(
+                x_values,
+                normalized_means,
+                marker="o",
+                color="skyblue",
+                linestyle="-",
+                linewidth=2,
+                markersize=8,
+            )
+            plt.xlabel("Time Step")
+            plt.ylabel("Normalized Attribution")
+            plt.title("Attribution Values")
+            plt.xticks(x_values)  # Set x-ticks to match the number of features
+            plt.tight_layout()
+            plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
         return means
 
-    def Faithfulness_Correlation(self, test_loader, attribution,similarity_func=None nr_runs=100, pertrub=None, subset_size=4):
+    def Faithfulness_Correlation(self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=4):
         """
     Implementation of faithfulness correlation by Bhatt et al., 2020.
 
@@ -654,10 +652,10 @@ def Faithfulness_Correlation(self, test_loader, attribution,similarity_func=None
 
         if torch.is_tensor(attribution):
             # Convert the tensor to a NumPy array
-            example_numpy_array = attribution.cpu().detach().numpy()
-        if similarity_func== None:
-            similarity_func=correlation_spearman
-        if pertrub == None:
+            attribution = attribution.cpu().detach().numpy()
+        if similarity_func is None:
+            similarity_func = correlation_spearman
+        if pertrub is None:
             pertrub = "baseline"
         similarities = []
         for batch in test_loader:
@@ -679,7 +677,6 @@ def Faithfulness_Correlation(self, test_loader, attribution,similarity_func=None
                 x["groups"],
                 x["target_scale"],
             )
-
             y_pred = self(data).detach().cpu().numpy()
             pred_deltas = []
             att_sums = []
@@ -701,9 +698,20 @@ def Faithfulness_Correlation(self, test_loader, attribution,similarity_func=None
 
                     mask[:, a_ix, :] = 0
                     mask = mask.to(x["encoder_cont"].device)
-
                     x["encoder_cont"] = x["encoder_cont"] * mask
-
+                data = (
+                    x["encoder_cat"],
+                    x["encoder_cont"],
+                    x["encoder_target"],
+                    x["encoder_lengths"],
+                    x["decoder_cat"],
+                    x["decoder_cont"],
+                    x["decoder_target"],
+                    x["decoder_lengths"],
+                    x["decoder_time_idx"],
+                    x["groups"],
+                    x["target_scale"],
+                )
                 # Predict on perturbed input x.
                 y_pred_perturb = self(data).detach().cpu().numpy()
 
@@ -712,11 +720,11 @@ def Faithfulness_Correlation(self, test_loader, attribution,similarity_func=None
                 # Sum attributions of the random subset.
 
                 att_sums.append(np.sum(attribution[a_ix]))
-            
-            similarities.append(similarity_func(pred_deltas,att_sums))
+
+            similarities.append(similarity_func(pred_deltas, att_sums))
         return np.nanmean(similarities)
 
-    def Data_Randomization(self, test_loader, attribution, explain_method, nr_runs=100, similarity=None):
+    def Data_Randomization(self, test_loader, attribution, explain_method, similarity_func=None):
         """
         Implementation of the Random Logit Metric by Sixt et al., 2020.
 
@@ -731,65 +739,64 @@ def Data_Randomization(self, test_loader, attribution, explain_method, nr_runs=1
 
         """
         a_perturbed = []
-        if similarity_func== None:
-            similarity_func=distance_euclidean
-        for batch in test_loader:
+        if similarity_func is None:
+            similarity_func = distance_euclidean
+        if (explain_method == 'attention'):
+            Attention_weights = self.interpertations(test_loader)
 
-            for key, value in batch[0].items():
+        else:
+            for batch in test_loader:
+
+                for key, value in batch[0].items():
+
+                    batch[0][key] = batch[0][key].to(self.device)
+                x = batch[0]
+                y = batch[1][0]
+                data = (
+                    x["encoder_cat"],
+                    x["encoder_cont"],
+                    x["encoder_target"],
+                    x["encoder_lengths"],
+                    x["decoder_cat"],
+                    x["decoder_cont"],
+                    x["decoder_target"],
+                    x["decoder_lengths"],
+                    x["decoder_time_idx"],
+                    x["groups"],
+                    x["target_scale"],
+                )
 
-                batch[0][key] = batch[0][key].to(self.device)
-            x = batch[0]
-            y = batch[1][0]
-            data = (
-                x["encoder_cat"],
-                x["encoder_cont"],
-                x["encoder_target"],
-                x["encoder_lengths"],
-                x["decoder_cat"],
-                x["decoder_cont"],
-                x["decoder_target"],
-                x["decoder_lengths"],
-                x["decoder_time_idx"],
-                x["groups"],
-                x["target_scale"],
-            )
-            y_off = np.array(
-                [
-                    np.random.choice(
-                        [y_ for y_ in list(np.arange(0, self.num_classes)) if y_ != y]
-                    )
-                ]
-            )
-            baselines = (
-                torch.zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
-                torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
-                torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
-                data[3].to(self.device),  # encoder_lengths, leave unchanged
-                torch.zeros_like(data[4]).to(self.device),  # decoder_cat, set to zero
-                torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
-                torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
-                data[7].to(self.device),  # decoder_lengths, leave unchanged
-                torch.zeros_like(data[8]).to(self.device),  # decoder_time_idx, set to zero
-                data[9].to(self.device),  # groups, leave unchanged
-                data[10].to(self.device),  # target_scale, leave unchanged
-            )
+                y_off = np.random.randint(low=0, high=self.num_classes, size=y.shape)
+
+                baselines = (
+                    torch.zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
+                    torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
+                    torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
+                    data[3].to(self.device),  # encoder_lengths, leave unchanged
+                    torch.zeros_like(data[4]).to(self.device),  # decoder_cat, set to zero
+                    torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
+                    torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
+                    data[7].to(self.device),  # decoder_lengths, leave unchanged
+                    torch.zeros_like(data[8]).to(self.device),  # decoder_time_idx, set to zero
+                    data[9].to(self.device),  # groups, leave unchanged
+                    data[10].to(self.device),  # target_scale, leave unchanged
+                )
+                self.train()
+                explantation = explain_method(self.forward_captum)
+                # Reformat attributions.
+                attr, delta = explantation.attribute(
+                    data, target=(1, 1), return_convergence_delta=True, baselines=baselines, n_steps=35
+                )
+                # Convert attributions to numpy array and append to the list
+                a_perturbed.append(attr[0].cpu().detach().numpy())
 
-            explantation = explain_method(self.forward_captum)
-            # Reformat attributions.
-            attr, delta = explantation.attribute(
-                data, target=y_off, return_convergence_delta=True, baselines=baselines, n_steps=20
-            )
-            # Convert attributions to numpy array and append to the list
-            a_perturbed.append(attr[0].cpu().detach().numpy())
-        a_perturbed = np.concatenate(a_perturbed, axis=0)
-        a_perturbed = a_perturbed.mean(axis=(0, 1))
+            a_perturbed = np.mean(a_perturbed, axis=(0, 1, 3))
 
-        # Compute mean along the batch dimension
-        a_perturbed = a_perturbed.mean(axis=(0, 2))
-        # Normalize the means values to range [0, 1]
-        normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
-        
-        return similarity_func(normalized_a_perturbed,attribution)
+            # Normalize the means values to range [0, 1]
+            normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
+            score = similarity_func(normalized_a_perturbed, attribution)
+
+        return score
 
 
 @gin.configurable("MLWrapper")

From 176120ff66424326f374d7d2aed4d706ccd4c350 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 2 Nov 2023 14:31:46 +0100
Subject: [PATCH 098/142] changed attribution to take into account all
 timesteps

---
 icu_benchmarks/models/train.py    | 14 ++++++++------
 icu_benchmarks/models/wrappers.py | 17 +++++++++++------
 2 files changed, 19 insertions(+), 12 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index b7c966c2..4037f480 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -272,22 +272,24 @@ def train_common(
 
     if explain:
         attributions_IG = model.explantation_captum(
-            test_loader, log_dir, IntegratedGradients
+            test_loader=test_loader,
+            method=IntegratedGradients, log_dir=log_dir, plot=True
         )
 
         print("IG", attributions_IG)
-        Attention_weights = model.interpertations(test_loader, log_dir)
+        Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
         print("attention", Attention_weights)
         if XAI_metric:
             scores1 = model.Faithfulness_Correlation(test_loader, attributions_IG)
             print('Attributions faithfulness correlation', scores1)
             scores2 = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"])
             print('Attention faithfulness correlation', scores2)
-            # scores = model.faithfulness_correlation(test_loader, Attention_weights["attention"])
-
-            scores_2 = model.Data_Randomization(
+            scores_1 = model.Data_Randomization(
                 test_loader, attributions_IG, IntegratedGradients)
-            print('Distance Data randmoization score', scores_2)
+            """
+            scores_2 = model.Data_Randomization(
+                test_loader, attributions_IG, Attention_weights["attention"])"""
+            print('Distance Data randmoization score', scores_1)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 45c44771..0f1db2bb 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -572,14 +572,19 @@ def explantation_captum(self, test_loader, method, log_dir='.', plot=False):
 
             explantation = method(self.forward_captum)
             # Reformat attributions.
-            attr, delta = explantation.attribute(
-                data, target=(0, 1), return_convergence_delta=True, baselines=baselines, n_steps=35
-            )
-            # Convert attributions to numpy array and append to the list
-            all_attrs.append(attr[0].cpu().detach().numpy())
-
+            attr_all_timesteps = []
+            for time_step in range(0, 24):
+                attr, delta = explantation.attribute(
+                    data, target=(time_step, 1), return_convergence_delta=True, baselines=baselines, n_steps=35
+                )
+                # Convert attributions to numpy array and append to the list
+                attr_all_timesteps.append(attr[0].cpu().detach().numpy())
+            print(np.shape((attr_all_timesteps)))
+            all_attrs.append(np.mean(attr_all_timesteps, axis=0))
+            print(np.shape(all_attrs))
         # Concatenate a‚ttribution values for all instances along the batch dimension
         all_attrs = np.mean(all_attrs, axis=0)
+        print(np.shape(all_attrs))
         means_feature = all_attrs.mean(axis=(0, 1))
 
         # Compute mean along the batch dimension

From edc4ad2f0778fe74ba9f957012a1834e329edfa9 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 2 Nov 2023 16:13:32 +0100
Subject: [PATCH 099/142] added attention method for randomization

---
 icu_benchmarks/data/loader.py     | 11 ++++++++---
 icu_benchmarks/models/train.py    | 30 +++++++++++++++---------------
 icu_benchmarks/models/wrappers.py | 29 ++++++++++++++++++-----------
 3 files changed, 41 insertions(+), 29 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index e02b2e76..cb2db4be 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -487,9 +487,7 @@ def __init__(
         self.name = name
         self.data = pd.merge(
             labels, features, on=["stay_id", "time"]
-        )  # combine labels and features
-        # self.data["sex"].replace([0, 1], ["Female", "Male"], inplace=True)
-        # List of column names to convert from boolean to float
+        )
         if (len(lagged_variables) > 0):
             if self.data["label"].dtype == "bool":
                 self.data["label"] = self.data["label"].astype(
@@ -544,3 +542,10 @@ def get_balance(self) -> list:
 
     def get_feature_names(self):
         return self.column_names
+
+    def randomize_labels(self, num_classes=None, min=None, max=None):
+        if num_classes is None:
+            random_target = np.random.uniform(min, max, size=len(self.data['target'][0]))
+        else:
+            random_target = np.random.randint(num_classes, size=len(self.data['target'][0]))
+        self.data['target'][0] = Tensor(random_target)
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 4037f480..e1e70da1 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -271,25 +271,25 @@ def train_common(
         return 0
 
     if explain:
-        attributions_IG = model.explantation_captum(
-            test_loader=test_loader,
-            method=IntegratedGradients, log_dir=log_dir, plot=True
-        )
+       # attributions_IG = model.explantation_captum(
+       #     test_loader=test_loader,
+       #     method=IntegratedGradients, log_dir=log_dir, plot=True
+       # )
 
-        print("IG", attributions_IG)
+       # print("IG", attributions_IG)
         Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
         print("attention", Attention_weights)
         if XAI_metric:
-            scores1 = model.Faithfulness_Correlation(test_loader, attributions_IG)
-            print('Attributions faithfulness correlation', scores1)
-            scores2 = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"])
-            print('Attention faithfulness correlation', scores2)
-            scores_1 = model.Data_Randomization(
-                test_loader, attributions_IG, IntegratedGradients)
-            """
-            scores_2 = model.Data_Randomization(
-                test_loader, attributions_IG, Attention_weights["attention"])"""
-            print('Distance Data randmoization score', scores_1)
+            #    F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG)
+         #   print('Attributions faithfulness correlation', F_attribution)
+            F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"])
+            print('Attention faithfulness correlation', F_attention)
+         #   R_attribution = model.Data_Randomization(
+          #      test_loader, attributions_IG, IntegratedGradients)
+          #  print('Distance Data randmoization score attribution', R_attribution)
+            R_attention = model.Data_Randomization(
+                test_loader, Attention_weights["attention"], "Attention", test_dataset=test_dataset)
+            print('Distance Data randmoization score attention', R_attention)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 0f1db2bb..71cb6f40 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -579,12 +579,9 @@ def explantation_captum(self, test_loader, method, log_dir='.', plot=False):
                 )
                 # Convert attributions to numpy array and append to the list
                 attr_all_timesteps.append(attr[0].cpu().detach().numpy())
-            print(np.shape((attr_all_timesteps)))
             all_attrs.append(np.mean(attr_all_timesteps, axis=0))
-            print(np.shape(all_attrs))
         # Concatenate a‚ttribution values for all instances along the batch dimension
         all_attrs = np.mean(all_attrs, axis=0)
-        print(np.shape(all_attrs))
         means_feature = all_attrs.mean(axis=(0, 1))
 
         # Compute mean along the batch dimension
@@ -729,7 +726,7 @@ def Faithfulness_Correlation(self, test_loader, attribution, similarity_func=Non
             similarities.append(similarity_func(pred_deltas, att_sums))
         return np.nanmean(similarities)
 
-    def Data_Randomization(self, test_loader, attribution, explain_method, similarity_func=None):
+    def Data_Randomization(self, test_loader, attribution, explain_method, similarity_func=None, test_dataset=None):
         """
         Implementation of the Random Logit Metric by Sixt et al., 2020.
 
@@ -746,9 +743,16 @@ def Data_Randomization(self, test_loader, attribution, explain_method, similarit
         a_perturbed = []
         if similarity_func is None:
             similarity_func = distance_euclidean
-        if (explain_method == 'attention'):
-            Attention_weights = self.interpertations(test_loader)
+        if (explain_method == 'Attention'):
+            for batch in test_loader:
+                print(batch[1][0])
+
+            test_dataset.randomize_labels(num_classes=self.num_classes)
+            for batch in test_loader:
+                print(batch[1][0])  # generate random labels
+            Attention_weights = self.interpertations(test_loader)  # test_loader works by reference
 
+            score = similarity_func(Attention_weights["attention"].cpu().numpy(), attribution.cpu().numpy())
         else:
             for batch in test_loader:
 
@@ -789,11 +793,14 @@ def Data_Randomization(self, test_loader, attribution, explain_method, similarit
                 self.train()
                 explantation = explain_method(self.forward_captum)
                 # Reformat attributions.
-                attr, delta = explantation.attribute(
-                    data, target=(1, 1), return_convergence_delta=True, baselines=baselines, n_steps=35
-                )
-                # Convert attributions to numpy array and append to the list
-                a_perturbed.append(attr[0].cpu().detach().numpy())
+                attr_all_timesteps = []
+                for time_step in range(0, 24):
+                    attr, delta = explantation.attribute(
+                        data, target=(time_step, 1), return_convergence_delta=True, baselines=baselines, n_steps=35
+                    )
+                    # Convert attributions to numpy array and append to the list
+                    attr_all_timesteps.append(attr[0].cpu().detach().numpy())
+                a_perturbed.append(np.mean(attr_all_timesteps, axis=0))
 
             a_perturbed = np.mean(a_perturbed, axis=(0, 1, 3))
 

From 2be507674b1d17fbfbf090a61c5401001974d548 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 7 Nov 2023 13:34:21 +0100
Subject: [PATCH 100/142] added baseline for faithfulness attribution

---
 icu_benchmarks/cross_validation.py       |  13 +-
 icu_benchmarks/data/loader.py            |  56 ++-----
 icu_benchmarks/models/dl_models.py       |  33 ++--
 icu_benchmarks/models/similarity_func.py |  16 +-
 icu_benchmarks/models/train.py           |  56 ++++---
 icu_benchmarks/models/wrappers.py        | 192 ++++++++---------------
 icu_benchmarks/run.py                    |  49 ++----
 icu_benchmarks/run_utils.py              | 133 +++++-----------
 icu_benchmarks/tuning/hyperparameters.py |  26 +--
 9 files changed, 199 insertions(+), 375 deletions(-)

diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 8c0c02b1..7d3f217e 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -41,6 +41,8 @@ def execute_repeated_cv(
     explain: bool = False,
     pytorch_forecasting: bool = False,
     XAI_metric: bool = False,
+    random_labels: bool = False,
+    random_model_dir: str = None
 ) -> float:
     """Preprocesses data and trains a model for each fold.
 
@@ -83,15 +85,11 @@ def execute_repeated_cv(
         cv_folds_to_train = 1
 
     else:
-        logging.info(
-            f"Starting nested CV with {cv_repetitions_to_train} repetitions of {cv_folds_to_train} folds."
-        )
+        logging.info(f"Starting nested CV with {cv_repetitions_to_train} repetitions of {cv_folds_to_train} folds.")
 
     for repetition in range(cv_repetitions_to_train):
         for fold_index in range(cv_folds_to_train):
-            repetition_fold_dir = (
-                log_dir / f"repetition_{repetition}" / f"fold_{fold_index}"
-            )
+            repetition_fold_dir = log_dir / f"repetition_{repetition}" / f"fold_{fold_index}"
             repetition_fold_dir.mkdir(parents=True, exist_ok=True)
 
             start_time = datetime.now()
@@ -109,6 +107,7 @@ def execute_repeated_cv(
                 pretrained_imputation_model=pretrained_imputation_model,
                 runmode=mode,
                 complete_train=complete_train,
+
             )
 
             preprocess_time = datetime.now() - start_time
@@ -129,6 +128,8 @@ def execute_repeated_cv(
                 explain=explain,
                 pytorch_forecasting=pytorch_forecasting,
                 XAI_metric=XAI_metric,
+                random_labels=random_labels,
+                random_model_dir=random_model_dir
             )
 
             train_time = datetime.now() - start_time
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index cb2db4be..fdf4dc68 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -103,15 +103,11 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id].to_numpy()
-        labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(
-            dtype=float
-        )
+        labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
 
         if len(labels) == 1:
             # only one label per stay, align with window
-            labels = np.concatenate(
-                [np.empty(window.shape[0] - 1) * np.nan, labels], axis=0
-            )
+            labels = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, labels], axis=0)
 
         length_diff = self.maxlen - window.shape[0]
         pad_mask = np.ones(window.shape[0])
@@ -119,9 +115,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         # Padding the array to fulfill size requirement
         if length_diff > 0:
             # window shorter than the longest window in dataset, pad to same length
-            window = np.concatenate(
-                [window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0
-            )
+            window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)
             labels = np.concatenate([labels, np.ones(length_diff) * pad_value], axis=0)
             pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)
 
@@ -331,12 +325,8 @@ def __init__(
         self.amputated_values, self.amputation_mask = ampute_data(
             self.features_df, mask_method, mask_proportion, mask_observation_proportion
         )
-        self.amputation_mask = (
-            self.amputation_mask + self.features_df.isna().values
-        ).bool()
-        self.amputation_mask = DataFrame(
-            self.amputation_mask, columns=self.vars[Segment.dynamic]
-        )
+        self.amputation_mask = (self.amputation_mask + self.features_df.isna().values).bool()
+        self.amputation_mask = DataFrame(self.amputation_mask, columns=self.vars[Segment.dynamic])
         self.amputation_mask[self.vars["GROUP"]] = self.features_df.index
         self.amputation_mask.set_index(self.vars["GROUP"], inplace=True)
 
@@ -361,15 +351,9 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
-        window_missingness_mask = self.target_missingness_mask.loc[
-            stay_id:stay_id, self.vars[Segment.dynamic]
-        ]
-        amputated_window = self.amputated_values.loc[
-            stay_id:stay_id, self.vars[Segment.dynamic]
-        ]
-        amputation_mask = self.amputation_mask.loc[
-            stay_id:stay_id, self.vars[Segment.dynamic]
-        ]
+        window_missingness_mask = self.target_missingness_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
+        amputated_window = self.amputated_values.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
+        amputation_mask = self.amputation_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]
 
         return (
             from_numpy(amputated_window.values).to(float32),
@@ -472,30 +456,22 @@ def __init__(
         name: str = "",
         max_prediction_length: int = 24,
         max_encoder_length: int = 24,
-
         **kwargs,
     ):
-
-        data[split]["FEATURES"]["time_idx"] = (
-            (data[split]["FEATURES"]["time"] / pd.Timedelta(seconds=3600))
-        ).astype(
+        data[split]["FEATURES"]["time_idx"] = ((data[split]["FEATURES"]["time"] / pd.Timedelta(seconds=3600))).astype(
             int
         )  # create an incremental column indicating the time step(required by constructor)
         data = data.get(split)  # get split
         labels = data["OUTCOME"]
         features = data["FEATURES"]
         self.name = name
-        self.data = pd.merge(
-            labels, features, on=["stay_id", "time"]
-        )
-        if (len(lagged_variables) > 0):
+        self.data = pd.merge(labels, features, on=["stay_id", "time"])
+        if len(lagged_variables) > 0:
             if self.data["label"].dtype == "bool":
-                self.data["label"] = self.data["label"].astype(
-                    float
-                )
+                self.data["label"] = self.data["label"].astype(float)
             columns_to_lag = lagged_variables
             grouped = self.data.sort_values("time_idx").groupby("stay_id")
-            for lag in range(max_encoder_length, max_encoder_length+max_prediction_length + 1):
+            for lag in range(max_encoder_length, max_encoder_length + max_prediction_length + 1):
                 for column in columns_to_lag:
                     # Create a new column with lagged values
                     self.data[f"{column}_lag_{lag}"] = grouped[column].shift(lag, fill_value=0)
@@ -545,7 +521,7 @@ def get_feature_names(self):
 
     def randomize_labels(self, num_classes=None, min=None, max=None):
         if num_classes is None:
-            random_target = np.random.uniform(min, max, size=len(self.data['target'][0]))
+            random_target = np.random.uniform(min, max, size=len(self.data["target"][0]))
         else:
-            random_target = np.random.randint(num_classes, size=len(self.data['target'][0]))
-        self.data['target'][0] = Tensor(random_target)
+            random_target = np.random.randint(num_classes, size=len(self.data["target"][0]))
+        self.data["target"][0] = Tensor(random_target)
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 7fb7fd27..b26bd6b0 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -149,9 +149,7 @@ def __init__(
             **kwargs,
         )
         hidden = hidden if hidden % 2 == 0 else hidden + 1  # Make sure hidden is even
-        self.input_embedding = nn.Linear(
-            input_size[2], hidden
-        )  # This acts as a time-distributed layer by defaults
+        self.input_embedding = nn.Linear(input_size[2], hidden)  # This acts as a time-distributed layer by defaults
         if pos_encoding:
             self.pos_encoder = PositionalEncoding(hidden)
         else:
@@ -222,9 +220,7 @@ def __init__(
         )
 
         hidden = hidden if hidden % 2 == 0 else hidden + 1  # Make sure hidden is even
-        self.input_embedding = nn.Linear(
-            input_size[2], hidden
-        )  # This acts as a time-distributed layer by defaults
+        self.input_embedding = nn.Linear(input_size[2], hidden)  # This acts as a time-distributed layer by defaults
         if pos_encoding:
             self.pos_encoder = PositionalEncoding(hidden)
         else:
@@ -290,13 +286,9 @@ def __init__(
 
         # We compute automatically the depth based on the desired seq_length.
         if isinstance(num_channels, Integral) and max_seq_length:
-            num_channels = [num_channels] * int(
-                np.ceil(np.log(max_seq_length / 2) / np.log(kernel_size))
-            )
+            num_channels = [num_channels] * int(np.ceil(np.log(max_seq_length / 2) / np.log(kernel_size)))
         elif isinstance(num_channels, Integral) and not max_seq_length:
-            raise Exception(
-                "a maximum sequence length needs to be provided if num_channels is int"
-            )
+            raise Exception("a maximum sequence length needs to be provided if num_channels is int")
 
         num_levels = len(num_channels)
         for i in range(num_levels):
@@ -527,19 +519,14 @@ def forward(self, x):
 
     def actual_vs_predictions_plot(self, dataloader):
         predictions = self.model.predict(dataloader, return_x=True)
-        predictions_vs_actuals = self.model.calculate_prediction_actual_by_variable(
-            predictions.x, predictions.output
-        )
+        predictions_vs_actuals = self.model.calculate_prediction_actual_by_variable(predictions.x, predictions.output)
         self.model.plot_prediction_actual_by_variable(predictions_vs_actuals)
         return predictions_vs_actuals
 
-    def interpertations(self, dataloader, log_dir='.', plot=False):
-
+    def interpertations(self, dataloader, log_dir=".", plot=False):
         raw_predictions = self.model.predict(dataloader, return_x=True, mode="raw")
-        interpretation = self.model.interpret_output(
-            raw_predictions.output, reduction="sum"
-        )
-        if (plot):
+        interpretation = self.model.interpret_output(raw_predictions.output, reduction="sum")
+        if plot:
             figs = self.model.plot_interpretation(interpretation)
             for key, fig in figs.items():
                 fig.savefig(log_dir / f"interpretation_{key}.png", bbox_inches="tight")
@@ -563,9 +550,7 @@ def predict_dependency(self, dataloader, variable, log_dir):
             q75=lambda x: x.quantile(0.75),
         )
         ax = agg_dependency.plot(y="median")
-        ax.fill_between(
-            agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3
-        )
+        ax.fill_between(agg_dependency.index, agg_dependency.q25, agg_dependency.q75, alpha=0.3)
         plt.savefig(log_dir / "dependecy.png", bbox_inches="tight")
         return dependency
 
diff --git a/icu_benchmarks/models/similarity_func.py b/icu_benchmarks/models/similarity_func.py
index 88d19a9a..3ff96474 100644
--- a/icu_benchmarks/models/similarity_func.py
+++ b/icu_benchmarks/models/similarity_func.py
@@ -140,13 +140,7 @@ def distance_chebyshev(a: np.array, b: np.array, **kwargs) -> float:
     return scipy.spatial.distance.chebyshev(u=a, v=b)
 
 
-def lipschitz_constant(
-    a: np.array,
-    b: np.array,
-    c: Union[np.array, None],
-    d: Union[np.array, None],
-    **kwargs
-) -> float:
+def lipschitz_constant(a: np.array, b: np.array, c: Union[np.array, None], d: Union[np.array, None], **kwargs) -> float:
     """
     Calculate non-negative local Lipschitz abs(||a-b||/||c-d||), where a,b can be f(x) or a(x) and c,d is x.
 
@@ -262,13 +256,9 @@ def ssim(a: np.array, b: np.array, **kwargs) -> float:
     float
         The similarity score.
     """
-    max_point, min_point = np.max(np.concatenate([a, b])), np.min(
-        np.concatenate([a, b])
-    )
+    max_point, min_point = np.max(np.concatenate([a, b])), np.min(np.concatenate([a, b]))
     data_range = float(np.abs(max_point - min_point))
-    return skimage.metrics.structural_similarity(
-        im1=a, im2=b, win_size=kwargs.get("win_size", None), data_range=data_range
-    )
+    return skimage.metrics.structural_similarity(im1=a, im2=b, win_size=kwargs.get("win_size", None), data_range=data_range)
 
 
 def difference(a: np.array, b: np.array, **kwargs) -> np.array:
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index e1e70da1..f7a38790 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -6,6 +6,7 @@
 import pandas as pd
 from joblib import load
 from torch.optim import Adam
+import numpy as np
 import quantus
 from torch.utils.data import DataLoader
 from pytorch_lightning.loggers import TensorBoardLogger, WandbLogger
@@ -54,7 +55,7 @@ def train_common(
     optimizer: type = Adam,
     precision=32,
     batch_size=64,
-    epochs=100,
+    epochs=1000,
     patience=20,
     min_delta=1e-5,
     test_on: str = Split.test,
@@ -73,6 +74,8 @@ def train_common(
     explain: bool = False,
     pytorch_forecasting: bool = False,
     XAI_metric: bool = False,
+    random_labels: bool = False,
+    random_model_dir: str =None
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -173,8 +176,12 @@ def train_common(
                 optimizer=optimizer,
                 epochs=epochs,
                 run_mode=mode,
-                batch_size=batch_size,  # check why validation set is needed here
+                batch_size=batch_size,
             )
+        if random_labels:
+            train_dataset.randomize_labels(num_classes=model.num_classes)
+            val_dataset.randomize_labels(num_classes=model.num_classes)
+            test_dataset.randomize_labels(num_classes=model.num_classes)
 
     else:
         train_loader = DataLoader(
@@ -206,10 +213,6 @@ def train_common(
             if model.requires_backprop
             else DataLoader([test_dataset.to_tensor()], batch_size=1)
         )
-        """
-        if isinstance(next(iter(train_loader))[0], OrderedDict):
-            model = model(optimizer=optimizer, epochs=epochs, run_mode=mode)
-        """
 
         data_shape = next(iter(train_loader))[0].shape
         if load_weights:
@@ -271,25 +274,32 @@ def train_common(
         return 0
 
     if explain:
-       # attributions_IG = model.explantation_captum(
-       #     test_loader=test_loader,
-       #     method=IntegratedGradients, log_dir=log_dir, plot=True
-       # )
+        attributions_IG = model.explantation_captum(
+            test_loader=test_loader,
+            method=IntegratedGradients, log_dir=log_dir, plot=True
+        )
 
-       # print("IG", attributions_IG)
-        Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
-        print("attention", Attention_weights)
+        print("IG", attributions_IG)
+        # Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
+        # print("attention", Attention_weights)
         if XAI_metric:
-            #    F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG)
-         #   print('Attributions faithfulness correlation', F_attribution)
-            F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"])
-            print('Attention faithfulness correlation', F_attention)
-         #   R_attribution = model.Data_Randomization(
-          #      test_loader, attributions_IG, IntegratedGradients)
-          #  print('Distance Data randmoization score attribution', R_attribution)
-            R_attention = model.Data_Randomization(
-                test_loader, Attention_weights["attention"], "Attention", test_dataset=test_dataset)
-            print('Distance Data randmoization score attention', R_attention)
+            # ra2 = np.random.randint(low=0, high=101, size=24)
+            # random_attributions = np.random.normal(size=24)
+            # F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
+            # F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
+            # print('Random noraml faithfulness correlation', F_baseline)
+            # print('random unifrom ', F_baseline2)
+            # F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
+            # print('Attributions faithfulness correlation', F_attribution)
+            # F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"], pertrub='Noise')
+            # print('Attention faithfulness correlation', F_attention)
+            
+            #R_attribution = model.Data_Randomization(
+            #    test_loader, attributions_IG, IntegratedGradients)
+            #print('Distance Data randmoization score attribution', R_attribution)
+            # R_attention = model.Data_Randomization(
+            #    test_loader, Attention_weights["attention"], "Attention", test_dataset=test_dataset)
+            # print('Distance Data randmoization score attention', R_attention)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 71cb6f40..90177122 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -20,10 +20,9 @@
 from icu_benchmarks.contants import RunMode
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean
+
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
-gin.config.external_configurable(
-    nn.functional.cross_entropy, module="torch.nn.functional"
-)
+gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.mse_loss, module="torch.nn.functional")
 
 gin.config.external_configurable(mean_squared_error, module="sklearn.metrics")
@@ -88,9 +87,7 @@ def save_model(self, save_path, file_name, file_extension):
 
     def check_supported_runmode(self, runmode: RunMode):
         if runmode not in self._supported_run_modes:
-            raise ValueError(
-                f"Runmode {runmode} not supported for {self.__class__.__name__}"
-            )
+            raise ValueError(f"Runmode {runmode} not supported for {self.__class__.__name__}")
         return True
 
 
@@ -163,9 +160,7 @@ def finalize_step(self, step_prefix=""):
             self.log_dict(
                 {
                     f"{step_prefix}/{name}": (
-                        np.float32(metric.compute())
-                        if isinstance(metric.compute(), np.float64)
-                        else metric.compute()
+                        np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
                     )
                     for name, metric in self.metrics[step_prefix].items()
                     if "_Curve" not in name
@@ -206,10 +201,7 @@ def configure_optimizers(self):
 
     def on_test_epoch_start(self) -> None:
         self.metrics = {
-            step_name: {
-                metric_name: metric()
-                for metric_name, metric in self.set_metrics().items()
-            }
+            step_name: {metric_name: metric() for metric_name, metric in self.set_metrics().items()}
             for step_name in ["train", "val", "test"]
         }
         return super().on_test_epoch_start()
@@ -334,7 +326,6 @@ def step_fn(self, element, step_prefix=""):
                 data[key] = value
          """
         if len(element) == 2:
-
             data, labels = element[0], (element[1]).to(self.device)
             if isinstance(data, list):
                 for i in range(len(data)):
@@ -355,9 +346,7 @@ def step_fn(self, element, step_prefix=""):
             else:
                 data = data.float().to(self.device)
         else:
-            raise Exception(
-                "Loader should return either (data, label) or (data, label, mask)"
-            )
+            raise Exception("Loader should return either (data, label) or (data, label, mask)")
         out = self(data)
 
         # If aux_loss is present, it is returned as a tuple
@@ -367,30 +356,19 @@ def step_fn(self, element, step_prefix=""):
             aux_loss = 0
         # Get prediction and target
 
-        prediction = (
-            torch.masked_select(out, mask.unsqueeze(-1))
-            .reshape(-1, out.shape[-1])
-            .to(self.device)
-        )
+        prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
 
         target = torch.masked_select(labels, mask).to(self.device)
 
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
-            loss = (
-                self.loss(
-                    prediction, target.long(), weight=self.loss_weights.to(self.device)
-                )
-                + aux_loss
-            )
+            loss = self.loss(prediction, target.long(), weight=self.loss_weights.to(self.device)) + aux_loss
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
-            raise ValueError(
-                f"Run mode {self.run_mode} not yet supported. Please implement it."
-            )
+            raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
         transformed_output = self.output_transform((prediction, target))
 
         for key, value in self.metrics[step_prefix].items():
@@ -407,9 +385,7 @@ def step_fn(self, element, step_prefix=""):
                     value.update(transformed_output[0], transformed_output[1])
             else:
                 value.update(transformed_output)
-        self.log(
-            f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True
-        )
+        self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
         return loss
 
 
@@ -489,30 +465,19 @@ def step_fn(self, element, step_prefix=""):
             aux_loss = 0
         # Get prediction and target
 
-        prediction = (
-            torch.masked_select(out, mask.unsqueeze(-1))
-            .reshape(-1, out.shape[-1])
-            .to(self.device)
-        )
+        prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
 
         target = torch.masked_select(labels, mask).to(self.device)
 
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
-            loss = (
-                self.loss(
-                    prediction, target.long(), weight=self.loss_weights.to(self.device)
-                )
-                + aux_loss
-            )
+            loss = self.loss(prediction, target.long(), weight=self.loss_weights.to(self.device)) + aux_loss
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
-            raise ValueError(
-                f"Run mode {self.run_mode} not yet supported. Please implement it."
-            )
+            raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
         transformed_output = self.output_transform((prediction, target))
 
         for key, value in self.metrics[step_prefix].items():
@@ -529,12 +494,10 @@ def step_fn(self, element, step_prefix=""):
                     value.update(transformed_output[0], transformed_output[1])
             else:
                 value.update(transformed_output)
-        self.log(
-            f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True
-        )
+        self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
         return loss
 
-    def explantation_captum(self, test_loader, method, log_dir='.', plot=False):
+    def explantation_captum(self, test_loader, method, log_dir=".", plot=False):
         # Initialize lists to store attribution values for all instances
         all_attrs = []
 
@@ -587,7 +550,7 @@ def explantation_captum(self, test_loader, method, log_dir='.', plot=False):
         # Compute mean along the batch dimension
         means = all_attrs.mean(axis=(0, 2))
         # Normalize the means values to range [0, 1]
-        normalized_means = (means - means.min()) / (means.max() - means.min())
+        # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
             # Create x values (assuming you want a simple sequential x-axis)
             # Assuming you have 24 values
@@ -604,7 +567,7 @@ def explantation_captum(self, test_loader, method, log_dir='.', plot=False):
                 markersize=8,
             )
             plt.xlabel("Time Step")
-            plt.ylabel("Normalized Attribution")
+            plt.ylabel(" Attribution")
             plt.title("Attribution Values")
             plt.xticks(x_values)  # Set x-ticks to match the number of features
             plt.tight_layout()
@@ -613,7 +576,7 @@ def explantation_captum(self, test_loader, method, log_dir='.', plot=False):
             x_values = np.arange(1, 25)
             plt.plot(
                 x_values,
-                normalized_means,
+                means,
                 marker="o",
                 color="skyblue",
                 linestyle="-",
@@ -628,29 +591,31 @@ def explantation_captum(self, test_loader, method, log_dir='.', plot=False):
             plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
         return means
 
-    def Faithfulness_Correlation(self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=4):
+    def Faithfulness_Correlation(
+        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=2
+    ):
         """
-    Implementation of faithfulness correlation by Bhatt et al., 2020.
+        Implementation of faithfulness correlation by Bhatt et al., 2020.
 
-    The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
-    (or 'fidelity') with respect to the model behaviour.
+        The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
+        (or 'fidelity') with respect to the model behaviour.
 
-    Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
-    the average explanation attribution for only the subset of features are (linearly) correlated, taking the
-    average over multiple runs and test samples. The metric returns one float per input-attribution pair that
-    ranges between -1 and 1, where higher scores are better.
+        Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
+        the average explanation attribution for only the subset of features are (linearly) correlated, taking the
+        average over multiple runs and test samples. The metric returns one float per input-attribution pair that
+        ranges between -1 and 1, where higher scores are better.
 
-    For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
-    or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
-    test point and the average explanation attribution for only the subset of features is calculated. Results is
-    average over multiple runs and several test samples.
-    This code is adapted from the quantus libray to suit our use case
+        For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
+        or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
+        test point and the average explanation attribution for only the subset of features is calculated. Results is
+        average over multiple runs and several test samples.
+        This code is adapted from the quantus libray to suit our use case
 
-    References:
-        1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
-        explanations." IJCAI (2020): 3016-3022.
-        2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
-    """
+        References:
+            1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
+            explanations." IJCAI (2020): 3016-3022.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+        """
 
         if torch.is_tensor(attribution):
             # Convert the tensor to a NumPy array
@@ -661,9 +626,7 @@ def Faithfulness_Correlation(self, test_loader, attribution, similarity_func=Non
             pertrub = "baseline"
         similarities = []
         for batch in test_loader:
-
             for key, value in batch[0].items():
-
                 batch[0][key] = batch[0][key].to(self.device)
             x = batch[0]
             data = (
@@ -726,7 +689,7 @@ def Faithfulness_Correlation(self, test_loader, attribution, similarity_func=Non
             similarities.append(similarity_func(pred_deltas, att_sums))
         return np.nanmean(similarities)
 
-    def Data_Randomization(self, test_loader, attribution, explain_method, similarity_func=None, test_dataset=None):
+    def Data_Randomization(self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None):
         """
         Implementation of the Random Logit Metric by Sixt et al., 2020.
 
@@ -743,22 +706,15 @@ def Data_Randomization(self, test_loader, attribution, explain_method, similarit
         a_perturbed = []
         if similarity_func is None:
             similarity_func = distance_euclidean
-        if (explain_method == 'Attention'):
-            for batch in test_loader:
-                print(batch[1][0])
+        if explain_method == "Attention":
 
-            test_dataset.randomize_labels(num_classes=self.num_classes)
-            for batch in test_loader:
-                print(batch[1][0])  # generate random labels
-            Attention_weights = self.interpertations(test_loader)  # test_loader works by reference
+            Attention_weights = random_model.interpertations(test_loader)  # test_loader works by reference
 
             score = similarity_func(Attention_weights["attention"].cpu().numpy(), attribution.cpu().numpy())
         else:
             for batch in test_loader:
-
                 for key, value in batch[0].items():
-
-                    batch[0][key] = batch[0][key].to(self.device)
+                    batch[0][key] = batch[0][key].to(random_model.device)
                 x = batch[0]
                 y = batch[1][0]
                 data = (
@@ -775,23 +731,21 @@ def Data_Randomization(self, test_loader, attribution, explain_method, similarit
                     x["target_scale"],
                 )
 
-                y_off = np.random.randint(low=0, high=self.num_classes, size=y.shape)
-
                 baselines = (
-                    torch.zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
-                    torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
-                    torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
-                    data[3].to(self.device),  # encoder_lengths, leave unchanged
-                    torch.zeros_like(data[4]).to(self.device),  # decoder_cat, set to zero
-                    torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
-                    torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
-                    data[7].to(self.device),  # decoder_lengths, leave unchanged
-                    torch.zeros_like(data[8]).to(self.device),  # decoder_time_idx, set to zero
-                    data[9].to(self.device),  # groups, leave unchanged
-                    data[10].to(self.device),  # target_scale, leave unchanged
+                    torch.zeros_like(data[0]).to(random_model.device),  # encoder_cat, set to zero
+                    torch.zeros_like(data[1]).to(random_model.device),  # encoder_cont, set to zero
+                    torch.zeros_like(data[2]).to(random_model.device),  # encoder_target, set to zero
+                    data[3].to(random_model.device),  # encoder_lengths, leave unchanged
+                    torch.zeros_like(data[4]).to(random_model.device),  # decoder_cat, set to zero
+                    torch.zeros_like(data[5]).to(random_model.device),  # decoder_cont, set to zero
+                    torch.zeros_like(data[6]).to(random_model.device),  # decoder_target, set to zero
+                    data[7].to(random_model.device),  # decoder_lengths, leave unchanged
+                    torch.zeros_like(data[8]).to(random_model.device),  # decoder_time_idx, set to zero
+                    data[9].to(random_model.device),  # groups, leave unchanged
+                    data[10].to(random_model.device),  # target_scale, leave unchanged
                 )
-                self.train()
-                explantation = explain_method(self.forward_captum)
+
+                explantation = explain_method(random_model.forward_captum)
                 # Reformat attributions.
                 attr_all_timesteps = []
                 for time_step in range(0, 24):
@@ -805,8 +759,8 @@ def Data_Randomization(self, test_loader, attribution, explain_method, similarit
             a_perturbed = np.mean(a_perturbed, axis=(0, 1, 3))
 
             # Normalize the means values to range [0, 1]
-            normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
-            score = similarity_func(normalized_a_perturbed, attribution)
+            # normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
+            score = similarity_func(a_perturbed, attribution)
 
         return score
 
@@ -854,15 +808,9 @@ def set_metrics(self, labels):
 
         # Regression
         else:
-            if (
-                self.scaler is not None
-            ):  # We invert transform the labels and predictions if they were scaled.
-                self.output_transform = lambda x: self.scaler.inverse_transform(
-                    x.reshape(-1, 1)
-                )
-                self.label_transform = lambda x: self.scaler.inverse_transform(
-                    x.reshape(-1, 1)
-                )
+            if self.scaler is not None:  # We invert transform the labels and predictions if they were scaled.
+                self.output_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
+                self.label_transform = lambda x: self.scaler.inverse_transform(x.reshape(-1, 1))
             else:
                 self.output_transform = lambda x: x
                 self.label_transform = lambda x: x
@@ -897,9 +845,7 @@ def fit_model(self, train_data, train_labels, val_data, val_labels):
 
     def validation_step(self, val_dataset, _):
         val_rep, val_label = val_dataset.get_data_and_labels()
-        val_rep, val_label = torch.from_numpy(val_rep).to(
-            self.device
-        ), torch.from_numpy(val_label).to(self.device)
+        val_rep, val_label = torch.from_numpy(val_rep).to(self.device), torch.from_numpy(val_label).to(self.device)
         self.set_metrics(val_label)
 
         val_pred = self.predict(val_rep)
@@ -942,9 +888,7 @@ def log_metrics(self, label, pred, metric_type):
         self.log_dict(
             {
                 # MPS dependent type casting
-                f"{metric_type}/{name}": metric(
-                    self.label_transform(label), self.output_transform(pred)
-                )
+                f"{metric_type}/{name}": metric(self.label_transform(label), self.output_transform(pred))
                 if not self.mps
                 else metric(self.label_transform(label), self.output_transform(pred))
                 # Fore very metric
@@ -982,9 +926,7 @@ def set_model_args(self, model, *args, **kwargs):
         # Get passed keyword arguments
         arguments = locals()["kwargs"]
         # Get valid hyperparameters
-        hyperparams = {
-            key: value for key, value in arguments.items() if key in possible_hps
-        }
+        hyperparams = {key: value for key, value in arguments.items() if key in possible_hps}
         logging.debug(f"Creating model with: {hyperparams}.")
         return model(**hyperparams)
 
@@ -1037,9 +979,7 @@ def set_metrics(self):
     def init_weights(self, init_type="normal", gain=0.02):
         def init_func(m):
             classname = m.__class__.__name__
-            if hasattr(m, "weight") and (
-                classname.find("Conv") != -1 or classname.find("Linear") != -1
-            ):
+            if hasattr(m, "weight") and (classname.find("Conv") != -1 or classname.find("Linear") != -1):
                 if init_type == ImputationInit.NORMAL:
                     nn.init.normal_(m.weight.data, 0.0, gain)
                 elif init_type == ImputationInit.XAVIER:
@@ -1049,9 +989,7 @@ def init_func(m):
                 elif init_type == ImputationInit.ORTHOGONAL:
                     nn.init.orthogonal_(m.weight.data, gain=gain)
                 else:
-                    raise NotImplementedError(
-                        f"Initialization method {init_type} is not implemented"
-                    )
+                    raise NotImplementedError(f"Initialization method {init_type} is not implemented")
                 if hasattr(m, "bias") and m.bias is not None:
                     nn.init.constant_(m.bias.data, 0.0)
             elif classname.find("BatchNorm2d") != -1:
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 4d9fb2f4..70fbfa43 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -55,6 +55,7 @@ def main(my_args=tuple(sys.argv[1:])):
     explain = args.explain
     pytorch_forecasting = args.pytorch_forecasting
     XAI_metric = args.XAI_metric
+    random_labels = args.random_labels
     # Load task config
     gin.parse_config_file(f"configs/tasks/{task}.gin")
     mode = get_mode()
@@ -67,19 +68,11 @@ def main(my_args=tuple(sys.argv[1:])):
     logging.info(f"Task mode: {mode}.")
 
     # Set train size to fine tune size if fine tune is set, else use custom train size
-    train_size = (
-        args.fine_tune
-        if args.fine_tune is not None
-        else args.samples
-        if args.samples is not None
-        else None
-    )
+    train_size = args.fine_tune if args.fine_tune is not None else args.samples if args.samples is not None else None
     # Whether to load weights from a previous run
     load_weights = evaluate or args.fine_tune is not None
 
-    pretrained_imputation_model = load_pretrained_imputation_model(
-        args.pretrained_imputation
-    )
+    pretrained_imputation_model = load_pretrained_imputation_model(args.pretrained_imputation)
     # Log imputation model to wandb
     update_wandb_config(
         {
@@ -88,16 +81,12 @@ def main(my_args=tuple(sys.argv[1:])):
             else "None"
         }
     )
-
+    random_model_dir = args.random_model
     log_dir_name = args.log_dir / name
     log_dir = (
         (log_dir_name / experiment)
         if experiment
-        else (
-            log_dir_name
-            / (args.task_name if args.task_name is not None else args.task)
-            / model
-        )
+        else (log_dir_name / (args.task_name if args.task_name is not None else args.task) / model)
     )
     log_full_line(f"Logging to {log_dir}.", logging.INFO)
 
@@ -120,9 +109,7 @@ def main(my_args=tuple(sys.argv[1:])):
     # Load pretrained model in evaluate mode or when finetuning
     if load_weights:
         if args.source_dir is None:
-            raise ValueError(
-                "Please specify a source directory when evaluating or fine-tuning."
-            )
+            raise ValueError("Please specify a source directory when evaluating or fine-tuning.")
         log_dir /= f"_from_{args.source_name}"
         name_datasets(args.source_name, args.source_name, args.name)
         if args.fine_tune:
@@ -131,12 +118,9 @@ def main(my_args=tuple(sys.argv[1:])):
         run_dir = create_run_dir(log_dir)
         source_dir = args.source_dir
         logging.info(
-            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config."
-        )
+            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
         gin.parse_config_file(source_dir / "train_config.gin")
-    elif (
-        args.samples and args.source_dir is not None
-    ):  # Train model with limited samples and bind existing config
+    elif args.samples and args.source_dir is not None:  # Train model with limited samples and bind existing config
         logging.info("Binding train gin-config. Note: this might override your config.")
         gin.parse_config_file(args.source_dir / "train_config.gin")
         log_dir /= f"samples_{args.fine_tune}"
@@ -147,22 +131,15 @@ def main(my_args=tuple(sys.argv[1:])):
         name_datasets(args.name, args.name, args.name)
         hp_checkpoint = log_dir / args.hp_checkpoint if args.hp_checkpoint else None
         model_path = (
-            Path("configs")
-            / (
-                "imputation_models"
-                if mode == RunMode.imputation
-                else "prediction_models"
-            )
-            / f"{model}.gin"
+            Path("configs") / ("imputation_models" if mode ==
+                               RunMode.imputation else "prediction_models") / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
             if args.experiment
             else [model_path, Path(f"configs/tasks/{task}.gin")]
         )
-        gin.parse_config_files_and_bindings(
-            gin_config_files, args.hyperparams, finalize_config=False
-        )
+        gin.parse_config_files_and_bindings(gin_config_files, args.hyperparams, finalize_config=False)
         log_full_line(f"Data directory: {data_dir.resolve()}", level=logging.INFO)
         run_dir = create_run_dir(log_dir)
         choose_and_bind_hyperparameters(
@@ -177,6 +154,9 @@ def main(my_args=tuple(sys.argv[1:])):
             load_cache=args.load_cache,
             verbose=verbose,
             pytorch_forecasting=pytorch_forecasting,
+            random_labels=random_labels,
+
+
         )
 
     log_full_line(f"Logging to {run_dir.resolve()}", level=logging.INFO)
@@ -211,6 +191,7 @@ def main(my_args=tuple(sys.argv[1:])):
         explain=explain,
         pytorch_forecasting=pytorch_forecasting,
         XAI_metric=XAI_metric,
+        random_model_dir=random_model_dir
     )
 
     log_full_line("FINISHED TRAINING", level=logging.INFO, char="=", num_newlines=3)
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index cc4b7960..6caf4bee 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -23,9 +23,7 @@ def build_parser() -> ArgumentParser:
     Returns:
         The configured ArgumentParser.
     """
-    parser = ArgumentParser(
-        description="Framework for benchmarking ML/DL models on ICU data"
-    )
+    parser = ArgumentParser(description="Framework for benchmarking ML/DL models on ICU data")
 
     parser.add_argument(
         "-d",
@@ -42,12 +40,8 @@ def build_parser() -> ArgumentParser:
         help="Name of the task gin.",
     )
     parser.add_argument("-n", "--name", help="Name of the (target) dataset.")
-    parser.add_argument(
-        "-tn", "--task-name", help="Name of the task, used for naming experiments."
-    )
-    parser.add_argument(
-        "-m", "--model", default="LGBMClassifier", help="Name of the model gin."
-    )
+    parser.add_argument("-tn", "--task-name", help="Name of the task, used for naming experiments.")
+    parser.add_argument("-m", "--model", default="LGBMClassifier", help="Name of the model gin.")
     parser.add_argument("-e", "--experiment", help="Name of the experiment gin.")
     parser.add_argument(
         "-l",
@@ -71,12 +65,8 @@ def build_parser() -> ArgumentParser:
         help="Set to log verbosly. Disable for clean logs.",
     )
     parser.add_argument("--cpu", default=False, action=BOA, help="Set to use CPU.")
-    parser.add_argument(
-        "-db", "--debug", default=False, action=BOA, help="Set to load less data."
-    )
-    parser.add_argument(
-        "--reproducible", default=True, action=BOA, help="Make torch reproducible."
-    )
+    parser.add_argument("-db", "--debug", default=False, action=BOA, help="Set to load less data.")
+    parser.add_argument("--reproducible", default=True, action=BOA, help="Make torch reproducible.")
     parser.add_argument(
         "-lc",
         "--load_cache",
@@ -91,9 +81,7 @@ def build_parser() -> ArgumentParser:
         action=BOA,
         help="Set to generate data cache.",
     )
-    parser.add_argument(
-        "-p", "--preprocessor", type=Path, help="Load custom preprocessor from file."
-    )
+    parser.add_argument("-p", "--preprocessor", type=Path, help="Load custom preprocessor from file.")
     parser.add_argument("-pl", "--plot", action=BOA, help="Generate common plots.")
     parser.add_argument(
         "-wd",
@@ -107,18 +95,10 @@ def build_parser() -> ArgumentParser:
         type=str,
         help="Path to pretrained imputation model.",
     )
-    parser.add_argument(
-        "-hp", "--hyperparams", nargs="+", help="Hyperparameters for model."
-    )
-    parser.add_argument(
-        "--tune", default=False, action=BOA, help="Find best hyperparameters."
-    )
-    parser.add_argument(
-        "--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint."
-    )
-    parser.add_argument(
-        "--eval", default=False, action=BOA, help="Only evaluate model, skip training."
-    )
+    parser.add_argument("-hp", "--hyperparams", nargs="+", help="Hyperparameters for model.")
+    parser.add_argument("--tune", default=False, action=BOA, help="Find best hyperparameters.")
+    parser.add_argument("--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint.")
+    parser.add_argument("--eval", default=False, action=BOA, help="Only evaluate model, skip training.")
     parser.add_argument(
         "--complete-train",
         default=False,
@@ -132,12 +112,8 @@ def build_parser() -> ArgumentParser:
         type=int,
         help="Finetune model with amount of train data.",
     )
-    parser.add_argument(
-        "-sn", "--source-name", type=Path, help="Name of the source dataset."
-    )
-    parser.add_argument(
-        "--source-dir", type=Path, help="Directory containing gin and model weights."
-    )
+    parser.add_argument("-sn", "--source-name", type=Path, help="Name of the source dataset.")
+    parser.add_argument("--source-dir", type=Path, help="Directory containing gin and model weights.")
     parser.add_argument(
         "-sa",
         "--samples",
@@ -163,6 +139,19 @@ def build_parser() -> ArgumentParser:
         action=BOA,
         help="Compare explantations ",
     )
+    parser.add_argument(
+        "--random_labels",
+        default=False,
+        action=BOA,
+        help="train model on random label dataset",
+    )
+
+    parser.add_argument(
+        "-random_model",
+        default=Path("."),
+        type=Path,
+        help="Log directory for model weights that is trained on random labels",
+    )
 
     return parser
 
@@ -192,21 +181,15 @@ def create_run_dir(log_dir: Path, randomly_searched_params: str = None) -> Path:
 
 def import_preprocessor(preprocessor_path: str):
     # Import custom supplied preprocessor
-    log_full_line(
-        f"Importing custom preprocessor from {preprocessor_path}.", logging.INFO
-    )
+    log_full_line(f"Importing custom preprocessor from {preprocessor_path}.", logging.INFO)
     try:
-        spec = importlib.util.spec_from_file_location(
-            "CustomPreprocessor", preprocessor_path
-        )
+        spec = importlib.util.spec_from_file_location("CustomPreprocessor", preprocessor_path)
         module = importlib.util.module_from_spec(spec)
         sys.modules["preprocessor"] = module
         spec.loader.exec_module(module)
         gin.bind_parameter("preprocess.preprocessor", module.CustomPreprocessor)
     except Exception as e:
-        logging.error(
-            f"Could not import custom preprocessor from {preprocessor_path}: {e}"
-        )
+        logging.error(f"Could not import custom preprocessor from {preprocessor_path}: {e}")
 
 
 def aggregate_results(log_dir: Path, execution_time: timedelta = None):
@@ -250,14 +233,10 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
     # Calculate the population standard deviation over aggregated results over folds/iterations
     # Divide by sqrt(n) to get standard deviation.
-    std_scores = {
-        metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()
-    }
+    std_scores = {metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()}
 
     confidence_interval = {
-        metric: (
-            stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list))
-        )
+        metric: (stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list)))
         for metric, list in list_scores.items()
     }
 
@@ -265,9 +244,7 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
         "avg": averaged_scores,
         "std": std_scores,
         "CI_0.95": confidence_interval,
-        "execution_time": execution_time.total_seconds()
-        if execution_time is not None
-        else 0.0,
+        "execution_time": execution_time.total_seconds() if execution_time is not None else 0.0,
     }
 
     with open(log_dir / "aggregated_test_metrics.json", "w") as f:
@@ -283,14 +260,10 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
 def name_datasets(train="default", val="default", test="default"):
     """Names the datasets for logging (optional)."""
-    gin.bind_parameter(
-        "train_common.dataset_names", {"train": train, "val": val, "test": test}
-    )
+    gin.bind_parameter("train_common.dataset_names", {"train": train, "val": val, "test": test})
 
 
-def log_full_line(
-    msg: str, level: int = logging.INFO, char: str = "-", num_newlines: int = 0
-):
+def log_full_line(msg: str, level: int = logging.INFO, char: str = "-", num_newlines: int = 0):
     """Logs a full line of a given character with a message centered.
 
     Args:
@@ -318,43 +291,25 @@ def load_pretrained_imputation_model(use_pretrained_imputation):
     Args:
         use_pretrained_imputation: Path to the pretrained imputation model.
     """
-    if (
-        use_pretrained_imputation is not None
-        and not Path(use_pretrained_imputation).exists()
-    ):
+    if use_pretrained_imputation is not None and not Path(use_pretrained_imputation).exists():
         logging.warning("The specified pretrained imputation model does not exist.")
         use_pretrained_imputation = None
 
     if use_pretrained_imputation is not None:
-        logging.info(
-            "Using pretrained imputation from" + str(use_pretrained_imputation)
-        )
-        pretrained_imputation_model_checkpoint = torch.load(
-            use_pretrained_imputation, map_location=torch.device("cpu")
-        )
+        logging.info("Using pretrained imputation from" + str(use_pretrained_imputation))
+        pretrained_imputation_model_checkpoint = torch.load(use_pretrained_imputation, map_location=torch.device("cpu"))
         if isinstance(pretrained_imputation_model_checkpoint, dict):
             imputation_model_class = pretrained_imputation_model_checkpoint["class"]
             pretrained_imputation_model = imputation_model_class(
-                **pretrained_imputation_model_checkpoint["hyper_parameters"]
-            )
-            pretrained_imputation_model.set_trained_columns(
-                pretrained_imputation_model_checkpoint["trained_columns"]
-            )
-            pretrained_imputation_model.load_state_dict(
-                pretrained_imputation_model_checkpoint["state_dict"]
-            )
+                **pretrained_imputation_model_checkpoint["hyper_parameters"])
+            pretrained_imputation_model.set_trained_columns(pretrained_imputation_model_checkpoint["trained_columns"])
+            pretrained_imputation_model.load_state_dict(pretrained_imputation_model_checkpoint["state_dict"])
         else:
             pretrained_imputation_model = pretrained_imputation_model_checkpoint
-        pretrained_imputation_model = pretrained_imputation_model.to(
-            "cuda" if torch.cuda.is_available() else "cpu"
-        )
+        pretrained_imputation_model = pretrained_imputation_model.to("cuda" if torch.cuda.is_available() else "cpu")
         try:
-            logging.info(
-                f"imputation model device: {next(pretrained_imputation_model.parameters()).device}"
-            )
-            pretrained_imputation_model.device = next(
-                pretrained_imputation_model.parameters()
-            ).device
+            logging.info(f"imputation model device: {next(pretrained_imputation_model.parameters()).device}")
+            pretrained_imputation_model.device = next(pretrained_imputation_model.parameters()).device
         except Exception as e:
             logging.debug(f"Could not set device of imputation model: {e}")
     else:
@@ -375,9 +330,7 @@ def setup_logging(date_format, log_format, verbose):
     logging.basicConfig(format=log_format, datefmt=date_format)
     loggers = ["pytorch_lightning", "lightning_fabric"]
     for logger in loggers:
-        logging.getLogger(logger).handlers[0].setFormatter(
-            logging.Formatter(log_format, datefmt=date_format)
-        )
+        logging.getLogger(logger).handlers[0].setFormatter(logging.Formatter(log_format, datefmt=date_format))
 
     if not verbose:
         logging.getLogger().setLevel(logging.INFO)
diff --git a/icu_benchmarks/tuning/hyperparameters.py b/icu_benchmarks/tuning/hyperparameters.py
index cf3d0c17..7c02735e 100644
--- a/icu_benchmarks/tuning/hyperparameters.py
+++ b/icu_benchmarks/tuning/hyperparameters.py
@@ -37,6 +37,7 @@ def choose_and_bind_hyperparameters(
     verbose: bool = False,
     wandb: bool = False,
     pytorch_forecasting: bool = False,
+    random_labels: bool = False,
 ):
     """Choose hyperparameters to tune and bind them to gin.
 
@@ -68,9 +69,7 @@ def choose_and_bind_hyperparameters(
         return
 
     # Collect hyperparameters.
-    hyperparams_bounds, hyperparams_names = collect_bound_hyperparameters(
-        hyperparams, scopes
-    )
+    hyperparams_bounds, hyperparams_names = collect_bound_hyperparameters(hyperparams, scopes)
 
     if do_tune and not hyperparams_bounds:
         logging.info("No hyperparameters to tune, skipping tuning.")
@@ -81,16 +80,12 @@ def choose_and_bind_hyperparameters(
     if checkpoint:
         checkpoint_path = checkpoint / checkpoint_file
         if not checkpoint_path.exists():
-            logging.warning(
-                f"Hyperparameter checkpoint {checkpoint_path} does not exist."
-            )
+            logging.warning(f"Hyperparameter checkpoint {checkpoint_path} does not exist.")
             logging.info("Attempting to find latest checkpoint file.")
             checkpoint_path = find_checkpoint(log_dir.parent, checkpoint_file)
         # Check if we found a checkpoint file
         if checkpoint_path:
-            n_calls, configuration, evaluation = load_checkpoint(
-                checkpoint_path, n_calls
-            )
+            n_calls, configuration, evaluation = load_checkpoint(checkpoint_path, n_calls)
             # Check if we surpassed maximum tuning iterations
             if n_calls <= 0:
                 logging.log(
@@ -98,9 +93,7 @@ def choose_and_bind_hyperparameters(
                     "No more hyperparameter tuning iterations left, skipping tuning.",
                 )
                 logging.info("Training with these hyperparameters:")
-                bind_gin_params(
-                    hyperparams_names, configuration[np.argmin(evaluation)]
-                )  # bind best hyperparameters
+                bind_gin_params(hyperparams_names, configuration[np.argmin(evaluation)])  # bind best hyperparameters
                 return
         else:
             logging.warning("No checkpoint file found, starting from scratch.")
@@ -125,6 +118,7 @@ def bind_params_and_train(hyperparams):
                 verbose=verbose,
                 wandb=wandb,
                 pytorch_forecasting=pytorch_forecasting,
+                random_labels=random_labels
             )
 
     header = ["ITERATION"] + hyperparams_names + ["LOSS AT ITERATION"]
@@ -139,9 +133,7 @@ def tune_step_callback(res):
             table_cells = [len(res.x_iters)] + res.x_iters[-1] + [res.func_vals[-1]]
             highlight = res.x_iters[-1] == res.x  # highlight if best so far
             log_table_row(header, TUNE)
-            log_table_row(
-                table_cells, TUNE, align=Align.RIGHT, header=header, highlight=highlight
-            )
+            log_table_row(table_cells, TUNE, align=Align.RIGHT, header=header, highlight=highlight)
             wandb_log({"hp-iteration": len(res.x_iters)})
 
     if do_tune:
@@ -156,9 +148,7 @@ def tune_step_callback(res):
         logging.log(TUNE, "Hyperparameter tuning disabled")
         if configuration:
             # We have loaded a checkpoint, use the best hyperparameters.
-            logging.info(
-                "Training with the best hyperparameters from loaded checkpoint:"
-            )
+            logging.info("Training with the best hyperparameters from loaded checkpoint:")
             bind_gin_params(hyperparams_names, configuration[np.argmin(evaluation)])
         else:
             logging.log(TUNE, "Choosing hyperparameters randomly from bounds.")

From f5e8a3d66cd4c72e157aaf9c190e79e91e024809 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 7 Nov 2023 13:54:20 +0100
Subject: [PATCH 101/142] added option to load model trained with random labels

---
 icu_benchmarks/models/train.py    | 28 ++++++++++++++++++++--------
 icu_benchmarks/models/wrappers.py |  1 +
 icu_benchmarks/run.py             |  1 +
 icu_benchmarks/run_utils.py       |  2 +-
 4 files changed, 23 insertions(+), 9 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index f7a38790..68511625 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -75,7 +75,7 @@ def train_common(
     pytorch_forecasting: bool = False,
     XAI_metric: bool = False,
     random_labels: bool = False,
-    random_model_dir: str =None
+    random_model_dir: str = None
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -259,6 +259,9 @@ def train_common(
     if not eval_only:
         if model.requires_backprop:
             logging.info("Training DL model.")
+            print(next(train_loader)[1][0])
+            print(next(val_loader)[1][0])
+            print(next(test_loader)[1][0])
             trainer.fit(
                 model, train_dataloaders=train_loader, val_dataloaders=val_loader
             )
@@ -293,13 +296,22 @@ def train_common(
             # print('Attributions faithfulness correlation', F_attribution)
             # F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"], pertrub='Noise')
             # print('Attention faithfulness correlation', F_attention)
-            
-            #R_attribution = model.Data_Randomization(
-            #    test_loader, attributions_IG, IntegratedGradients)
-            #print('Distance Data randmoization score attribution', R_attribution)
-            # R_attention = model.Data_Randomization(
-            #    test_loader, Attention_weights["attention"], "Attention", test_dataset=test_dataset)
-            # print('Distance Data randmoization score attention', R_attention)
+            print(random_model_dir)
+            path = Path(random_model_dir)
+            print(path)
+            random_model = load_model(
+                model,
+                source_dir=path,
+                pl_model=pl_model,
+                train_dataset=train_dataset,
+                optimizer=optimizer,
+            )
+            R_attribution = model.Data_Randomization(
+                test_loader, attributions_IG, IntegratedGradients, random_model)
+            print('Distance Data randmoization score attribution', R_attribution)
+           # R_attention = model.Data_Randomization(
+           #     test_loader, Attention_weights["attention"], "Attention", test_dataset=test_dataset)
+           # print('Distance Data randmoization score attention', R_attention)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 90177122..20d13a25 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -703,6 +703,7 @@ def Data_Randomization(self, test_loader, attribution, explain_method, random_mo
             2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
+        print(random_model)
         a_perturbed = []
         if similarity_func is None:
             similarity_func = distance_euclidean
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 70fbfa43..b8d8a98d 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -81,6 +81,7 @@ def main(my_args=tuple(sys.argv[1:])):
             else "None"
         }
     )
+    print(args.random_model)
     random_model_dir = args.random_model
     log_dir_name = args.log_dir / name
     log_dir = (
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 6caf4bee..4bc3985c 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -147,7 +147,7 @@ def build_parser() -> ArgumentParser:
     )
 
     parser.add_argument(
-        "-random_model",
+        "--random_model",
         default=Path("."),
         type=Path,
         help="Log directory for model weights that is trained on random labels",

From 78146cc077d82e7289b25f088e5fc6d632aeec36 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 7 Nov 2023 14:44:04 +0100
Subject: [PATCH 102/142] added shapley explantations

---
 configs/prediction_models/common/DLCommon.gin |  2 +-
 icu_benchmarks/models/train.py                | 74 ++++++++++---------
 icu_benchmarks/models/wrappers.py             |  6 +-
 icu_benchmarks/run.py                         |  1 -
 4 files changed, 44 insertions(+), 39 deletions(-)

diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index fa4b440f..8f5d648e 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,7 +15,7 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-train_common.batch_size = 64
+train_common.batch_size = 128
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 68511625..3b35dd73 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -27,7 +27,7 @@
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
-from captum.attr import IntegratedGradients, LayerLRP
+from captum.attr import IntegratedGradients, ShapleyValueSampling
 
 
 cpu_core_count = (
@@ -259,9 +259,9 @@ def train_common(
     if not eval_only:
         if model.requires_backprop:
             logging.info("Training DL model.")
-            print(next(train_loader)[1][0])
-            print(next(val_loader)[1][0])
-            print(next(test_loader)[1][0])
+            print(next(iter(train_loader))[1][0])
+            print(next(iter(val_loader))[1][0])
+            print(next(iter(test_loader))[1][0])
             trainer.fit(
                 model, train_dataloaders=train_loader, val_dataloaders=val_loader
             )
@@ -277,41 +277,47 @@ def train_common(
         return 0
 
     if explain:
+        #
+        attributions_SHAP = model.explantation_captum(
+            test_loader=test_loader,
+            method=ShapleyValueSampling, log_dir=log_dir, plot=True
+        )
+        print("shap", attributions_SHAP)
         attributions_IG = model.explantation_captum(
             test_loader=test_loader,
-            method=IntegratedGradients, log_dir=log_dir, plot=True
+            method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=35, return_convergence_delta=True,
         )
 
-        print("IG", attributions_IG)
-        # Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
-        # print("attention", Attention_weights)
+        """
+        Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
+        print("attention", Attention_weights)
         if XAI_metric:
-            # ra2 = np.random.randint(low=0, high=101, size=24)
-            # random_attributions = np.random.normal(size=24)
-            # F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
-            # F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
-            # print('Random noraml faithfulness correlation', F_baseline)
-            # print('random unifrom ', F_baseline2)
-            # F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
-            # print('Attributions faithfulness correlation', F_attribution)
-            # F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"], pertrub='Noise')
-            # print('Attention faithfulness correlation', F_attention)
-            print(random_model_dir)
-            path = Path(random_model_dir)
-            print(path)
-            random_model = load_model(
-                model,
-                source_dir=path,
-                pl_model=pl_model,
-                train_dataset=train_dataset,
-                optimizer=optimizer,
-            )
-            R_attribution = model.Data_Randomization(
-                test_loader, attributions_IG, IntegratedGradients, random_model)
-            print('Distance Data randmoization score attribution', R_attribution)
-           # R_attention = model.Data_Randomization(
-           #     test_loader, Attention_weights["attention"], "Attention", test_dataset=test_dataset)
-           # print('Distance Data randmoization score attention', R_attention)
+            ra2 = np.random.randint(low=0, high=101, size=24)
+            random_attributions = np.random.normal(size=24)
+            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
+            F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
+            print('Random noraml faithfulness correlation', F_baseline)
+            print('random unifrom ', F_baseline2)
+            F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
+            print('Attributions faithfulness correlation', F_attribution)
+            F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"], pertrub='Noise')
+            print('Attention faithfulness correlation', F_attention)
+            """
+        # path = Path(random_model_dir)
+
+        # random_model = load_model(
+        #    model,
+        #    source_dir=path,
+        #    pl_model=pl_model,
+        #    train_dataset=train_dataset,
+        #    optimizer=optimizer,
+        # )
+        # R_attribution = model.Data_Randomization(
+        #    test_loader, attributions_IG, IntegratedGradients, random_model)
+        # print('Distance Data randmoization score attribution', R_attribution)
+        # R_attention = model.Data_Randomization(
+        #     test_loader, Attention_weights["attention"], "Attention", test_dataset=test_dataset)
+        # print('Distance Data randmoization score attention', R_attention)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 20d13a25..6220c875 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -497,7 +497,7 @@ def step_fn(self, element, step_prefix=""):
         self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
         return loss
 
-    def explantation_captum(self, test_loader, method, log_dir=".", plot=False):
+    def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kwargs):
         # Initialize lists to store attribution values for all instances
         all_attrs = []
 
@@ -538,7 +538,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False):
             attr_all_timesteps = []
             for time_step in range(0, 24):
                 attr, delta = explantation.attribute(
-                    data, target=(time_step, 1), return_convergence_delta=True, baselines=baselines, n_steps=35
+                    data, target=(time_step, 1),  baselines=baselines, **kwargs
                 )
                 # Convert attributions to numpy array and append to the list
                 attr_all_timesteps.append(attr[0].cpu().detach().numpy())
@@ -592,7 +592,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False):
         return means
 
     def Faithfulness_Correlation(
-        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=2
+        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=6
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index b8d8a98d..70fbfa43 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -81,7 +81,6 @@ def main(my_args=tuple(sys.argv[1:])):
             else "None"
         }
     )
-    print(args.random_model)
     random_model_dir = args.random_model
     log_dir_name = args.log_dir / name
     log_dir = (

From 48dc500d833795a28f1102f8d76157168e6f5205 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 8 Nov 2023 15:29:50 +0100
Subject: [PATCH 103/142] latest changes

---
 icu_benchmarks/models/dl_models.py |  2 +-
 icu_benchmarks/models/wrappers.py  | 11 +++++++----
 icu_benchmarks/run.py              |  3 ++-
 3 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index b26bd6b0..455e333c 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -641,7 +641,7 @@ def forward(
         # out = out["prediction"].reshape(1, -1)
 
         pred = self.logit(out["prediction"])
-
+        print(pred)
         return pred
 
 
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 6220c875..beebdb36 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -537,7 +537,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             # Reformat attributions.
             attr_all_timesteps = []
             for time_step in range(0, 24):
-                attr, delta = explantation.attribute(
+                attr = explantation.attribute(
                     data, target=(time_step, 1),  baselines=baselines, **kwargs
                 )
                 # Convert attributions to numpy array and append to the list
@@ -546,9 +546,10 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
         # Concatenate a‚ttribution values for all instances along the batch dimension
         all_attrs = np.mean(all_attrs, axis=0)
         means_feature = all_attrs.mean(axis=(0, 1))
-
+        self.log('Attribution Featues', means_feature)
         # Compute mean along the batch dimension
         means = all_attrs.mean(axis=(0, 2))
+        self.log('Attribution Time steps', means)
         # Normalize the means values to range [0, 1]
         # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
@@ -687,7 +688,9 @@ def Faithfulness_Correlation(
                 att_sums.append(np.sum(attribution[a_ix]))
 
             similarities.append(similarity_func(pred_deltas, att_sums))
-        return np.nanmean(similarities)
+            score = np.nanmean(similarities)
+            self.log('Faithfulness correlation', score)
+        return score
 
     def Data_Randomization(self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None):
         """
@@ -762,7 +765,7 @@ def Data_Randomization(self, test_loader, attribution, explain_method, random_mo
             # Normalize the means values to range [0, 1]
             # normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
             score = similarity_func(a_perturbed, attribution)
-
+            self.log('data_randomization', score)
         return score
 
 
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 70fbfa43..182a1814 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -191,7 +191,8 @@ def main(my_args=tuple(sys.argv[1:])):
         explain=explain,
         pytorch_forecasting=pytorch_forecasting,
         XAI_metric=XAI_metric,
-        random_model_dir=random_model_dir
+        random_model_dir=random_model_dir,
+        random_labels=random_labels,
     )
 
     log_full_line("FINISHED TRAINING", level=logging.INFO, char="=", num_newlines=3)

From 446ceaaaabc77d38f7b3e4aae9bdb164223a48e9 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Sat, 11 Nov 2023 02:02:21 +0100
Subject: [PATCH 104/142] latest

---
 icu_benchmarks/data/loader.py      |  1 +
 icu_benchmarks/models/constants.py |  4 +--
 icu_benchmarks/models/dl_models.py |  3 +-
 icu_benchmarks/models/metrics.py   | 21 ++----------
 icu_benchmarks/models/train.py     | 26 +++++++--------
 icu_benchmarks/models/wrappers.py  | 53 ++++++++++++++++--------------
 6 files changed, 48 insertions(+), 60 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index fdf4dc68..6287810b 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -501,6 +501,7 @@ def __init__(
             add_target_scales=True,
             add_encoder_length=True,
             predict_mode=True,
+            target_normalizer=None,
         )
 
     def get_balance(self) -> list:
diff --git a/icu_benchmarks/models/constants.py b/icu_benchmarks/models/constants.py
index 45af8271..49555d42 100644
--- a/icu_benchmarks/models/constants.py
+++ b/icu_benchmarks/models/constants.py
@@ -65,10 +65,10 @@ class DLMetrics:
     BINARY_CLASSIFICATION_TORCHMETRICS = {
         "AUC": AUROC(task="binary"),
         "PR": TorchMetricsAveragePrecision(task="binary"),
-        "PrecisionRecallCurve": TorchMetricsPrecisionRecallCurve(task="binary"),
+        # "PrecisionRecallCurve": TorchMetricsPrecisionRecallCurve(task="binary"),
         "Calibration_Error": CalibrationError(task="binary", n_bins=10),
         "F1": F1Score(task="binary", num_classes=2),
-        "Binary_Fairness": BinaryFairnessWrapper(num_groups=2, task="demographic_parity", group_name="sex"),
+        # "Binary_Fairness": BinaryFairnessWrapper(num_groups=2, task="demographic_parity", group_name="sex"),
     }
 
     MULTICLASS_CLASSIFICATION = {
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 455e333c..4e2a9a42 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -641,7 +641,8 @@ def forward(
         # out = out["prediction"].reshape(1, -1)
 
         pred = self.logit(out["prediction"])
-        print(pred)
+        # print(pred
+
         return pred
 
 
diff --git a/icu_benchmarks/models/metrics.py b/icu_benchmarks/models/metrics.py
index 4c2ffd2b..5428db61 100644
--- a/icu_benchmarks/models/metrics.py
+++ b/icu_benchmarks/models/metrics.py
@@ -1,37 +1,20 @@
 import torch
 from typing import Callable
 import numpy as np
-from ignite.metrics import EpochMetric
+from ignite.metrics import EpochMetric, Precision, Recall
 from sklearn.metrics import balanced_accuracy_score, mean_absolute_error
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
 import gin
 import torch.nn.functional as F
 
+
 """"
 This file contains metrics that are not available in ignite.metrics. Specifically, it adds transformation capabilities to some
 metrics.
 """
 
 
-class QuantileLossClass(torch.nn.Module):
-    def __init__(self, quantiles):
-        super().__init__()
-        self.register_buffer("q", torch.tensor(quantiles))
-
-    def forward(self, predictions, targets):
-        diff = predictions - targets
-        ql = (1 - self.q) * F.relu(diff) + self.q * F.relu(-diff)
-        losses = ql.view(-1, ql.shape[-1]).mean(0)
-        return losses
-
-
-@gin.configurable
-def Quantileloss(self, predictions, targets, quantiles=[0.1, 0.5, 0.9]):
-    loss = QuantileLossClass(quantiles=quantiles)
-    return loss(predictions, targets)
-
-
 def accuracy(output, target, topk=(1,)):
     """Computes the accuracy over the k top predictions for the specified values of k"""
     with torch.no_grad():
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 3b35dd73..3da76809 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -259,9 +259,7 @@ def train_common(
     if not eval_only:
         if model.requires_backprop:
             logging.info("Training DL model.")
-            print(next(iter(train_loader))[1][0])
-            print(next(iter(val_loader))[1][0])
-            print(next(iter(test_loader))[1][0])
+
             trainer.fit(
                 model, train_dataloaders=train_loader, val_dataloaders=val_loader
             )
@@ -278,31 +276,31 @@ def train_common(
 
     if explain:
         #
-        attributions_SHAP = model.explantation_captum(
-            test_loader=test_loader,
-            method=ShapleyValueSampling, log_dir=log_dir, plot=True
-        )
-        print("shap", attributions_SHAP)
+        # attributions_SHAP = model.explantation_captum(
+        #    test_loader=test_loader,
+        #    method=ShapleyValueSampling, log_dir=log_dir, plot=True
+        # )
+        # print("shap", attributions_SHAP)
         attributions_IG = model.explantation_captum(
             test_loader=test_loader,
-            method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=35, return_convergence_delta=True,
+            method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=35
         )
 
-        """
         Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
         print("attention", Attention_weights)
         if XAI_metric:
+
             ra2 = np.random.randint(low=0, high=101, size=24)
             random_attributions = np.random.normal(size=24)
-            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
-            F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
+            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions)
+            F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions)
             print('Random noraml faithfulness correlation', F_baseline)
             print('random unifrom ', F_baseline2)
             F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
             print('Attributions faithfulness correlation', F_attribution)
             F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"], pertrub='Noise')
             print('Attention faithfulness correlation', F_attention)
-            """
+
         # path = Path(random_model_dir)
 
         # random_model = load_model(
@@ -316,7 +314,7 @@ def train_common(
         #    test_loader, attributions_IG, IntegratedGradients, random_model)
         # print('Distance Data randmoization score attribution', R_attribution)
         # R_attention = model.Data_Randomization(
-        #     test_loader, Attention_weights["attention"], "Attention", test_dataset=test_dataset)
+        #    test_loader, Attention_weights["attention"], "Attention", random_model)
         # print('Distance Data randmoization score attention', R_attention)
 
     model.set_weight("balanced", train_dataset)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index beebdb36..c87bf161 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -156,23 +156,27 @@ def on_train_start(self):
         return super().on_train_start()
 
     def finalize_step(self, step_prefix=""):
+
         try:
-            self.log_dict(
-                {
-                    f"{step_prefix}/{name}": (
-                        np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
-                    )
-                    for name, metric in self.metrics[step_prefix].items()
-                    if "_Curve" not in name
-                },
-                sync_dist=True,
-            )
+            for name, metric in self.metrics[step_prefix].items():
+                try:
+                    value = np.float32(metric.compute()) if isinstance(
+                        metric.compute(), np.float64) else metric.compute()
+                    self.log_dict({f"{step_prefix}/{name}": value}, sync_dist=True)
+
+                except (NotComputableError, ValueError) as e:
+                    if step_prefix not in self._metrics_warning_printed:
+                        self._metrics_warning_printed.add(step_prefix)
+                        logging.warning(f"Metric for {step_prefix}/{name} not computable: {e}")
+
             for metric in self.metrics[step_prefix].values():
                 metric.reset()
-        except (NotComputableError, ValueError):
+        except (NotComputableError, ValueError) as e:
             if step_prefix not in self._metrics_warning_printed:
                 self._metrics_warning_printed.add(step_prefix)
                 logging.warning(f"Metrics for {step_prefix} not computable")
+                print(e)
+
             pass
 
     def configure_optimizers(self):
@@ -284,9 +288,11 @@ def softmax_multi_output_transform(output):
         # Output transform is not applied for contrib metrics, so we do our own.
         if self.run_mode == RunMode.classification:
             # Binary classification
+
             if self.logit.out_features == 2:
                 self.output_transform = softmax_binary_output_transform
-                metrics = DLMetrics.BINARY_CLASSIFICATION
+                metrics = DLMetrics.BINARY_CLASSIFICATION_TORCHMETRICS
+
             else:
                 # Multiclass classification
                 self.output_transform = softmax_multi_output_transform
@@ -483,15 +489,15 @@ def step_fn(self, element, step_prefix=""):
         for key, value in self.metrics[step_prefix].items():
             if isinstance(value, torchmetrics.Metric):
                 if key == "Binary_Fairness":
-                    feature_names = key.feature_helper(self.trainer)
+                    feature_names = self.metrics[step_prefix][key].feature_helper(self.trainer, step_prefix)
                     value.update(
                         transformed_output[0],
-                        transformed_output[1],
+                        transformed_output[1].int(),
                         data,
                         feature_names,
                     )
                 else:
-                    value.update(transformed_output[0], transformed_output[1])
+                    value.update(transformed_output[0], transformed_output[1].int())
             else:
                 value.update(transformed_output)
         self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
@@ -538,18 +544,20 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             attr_all_timesteps = []
             for time_step in range(0, 24):
                 attr = explantation.attribute(
-                    data, target=(time_step, 1),  baselines=baselines, **kwargs
+                    data, target=(time_step),  baselines=baselines, **kwargs
                 )
                 # Convert attributions to numpy array and append to the list
                 attr_all_timesteps.append(attr[0].cpu().detach().numpy())
             all_attrs.append(np.mean(attr_all_timesteps, axis=0))
         # Concatenate a‚ttribution values for all instances along the batch dimension
+        print(np.shape(all_attrs))
         all_attrs = np.mean(all_attrs, axis=0)
-        means_feature = all_attrs.mean(axis=(0, 1))
-        self.log('Attribution Featues', means_feature)
+        print(np.shape(all_attrs))
+        means_feature = all_attrs.mean(axis=(1))
+        # self.log('Attribution Featues', means_feature)
         # Compute mean along the batch dimension
-        means = all_attrs.mean(axis=(0, 2))
-        self.log('Attribution Time steps', means)
+        means = all_attrs.mean(axis=(0))
+       # self.log('Attribution Time steps', means)
         # Normalize the means values to range [0, 1]
         # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
@@ -689,7 +697,6 @@ def Faithfulness_Correlation(
 
             similarities.append(similarity_func(pred_deltas, att_sums))
             score = np.nanmean(similarities)
-            self.log('Faithfulness correlation', score)
         return score
 
     def Data_Randomization(self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None):
@@ -706,13 +713,12 @@ def Data_Randomization(self, test_loader, attribution, explain_method, random_mo
             2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
-        print(random_model)
         a_perturbed = []
         if similarity_func is None:
             similarity_func = distance_euclidean
         if explain_method == "Attention":
 
-            Attention_weights = random_model.interpertations(test_loader)  # test_loader works by reference
+            Attention_weights = random_model.interpertations(test_loader)
 
             score = similarity_func(Attention_weights["attention"].cpu().numpy(), attribution.cpu().numpy())
         else:
@@ -765,7 +771,6 @@ def Data_Randomization(self, test_loader, attribution, explain_method, random_mo
             # Normalize the means values to range [0, 1]
             # normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
             score = similarity_func(a_perturbed, attribution)
-            self.log('data_randomization', score)
         return score
 
 

From 73d39095cdee74e123adad5229cb6d29ddd26c5c Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 13 Nov 2023 13:36:13 +0100
Subject: [PATCH 105/142] removed label from input for tft

---
 configs/prediction_models/RNNpytorch.gin |  4 ++--
 configs/prediction_models/TFTpytorch.gin |  3 +--
 icu_benchmarks/models/dl_models.py       |  4 ++--
 icu_benchmarks/models/train.py           |  9 +++++----
 icu_benchmarks/models/wrappers.py        | 18 +++++++++---------
 5 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index ab42884f..334def2d 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -13,8 +13,8 @@ optimizer/hyperparameter.weight_decay = 1e-6
 # Model params
 
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = 2
-model/hyperparameter.rnn_layers=2
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
 model/hyperparameter.dropout = (0.0, 0.4)
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index a6523c6b..abb81f6f 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -74,6 +74,5 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc",
-                "label",]
+                "wbc",]
 PredictionDatasetTFTpytorch.lagged_variables=[]
\ No newline at end of file
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 4e2a9a42..5f1d287e 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -640,10 +640,10 @@ def forward(
 
         # out = out["prediction"].reshape(1, -1)
 
-        pred = self.logit(out["prediction"])
+        # pred = self.logit(out["prediction"])
         # print(pred
 
-        return pred
+        return out["prediction"]
 
 
 @gin.configurable
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 3da76809..018c8bc3 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -287,18 +287,19 @@ def train_common(
         )
 
         Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
-        print("attention", Attention_weights)
+       # print("attention", Attention_weights)
         if XAI_metric:
 
-            ra2 = np.random.randint(low=0, high=101, size=24)
+            ra2 = np.random.randint(low=0, high=100, size=24)
             random_attributions = np.random.normal(size=24)
+
             F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions)
             F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions)
             print('Random noraml faithfulness correlation', F_baseline)
             print('random unifrom ', F_baseline2)
-            F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
+            F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG)
             print('Attributions faithfulness correlation', F_attribution)
-            F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"], pertrub='Noise')
+            F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"])
             print('Attention faithfulness correlation', F_attention)
 
         # path = Path(random_model_dir)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index c87bf161..41eee23b 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -509,9 +509,11 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
 
         # Loop through the test_loader to compute attributions for all instances
         for batch in test_loader:
+
             for key, value in batch[0].items():
                 batch[0][key] = batch[0][key].to(self.device)
             x = batch[0]
+
             data = (
                 x["encoder_cat"].float().requires_grad_(),
                 x["encoder_cont"].requires_grad_(),
@@ -542,28 +544,28 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             explantation = method(self.forward_captum)
             # Reformat attributions.
             attr_all_timesteps = []
-            for time_step in range(0, 24):
+            for time_step in range(23, 24):
                 attr = explantation.attribute(
                     data, target=(time_step),  baselines=baselines, **kwargs
                 )
                 # Convert attributions to numpy array and append to the list
+
                 attr_all_timesteps.append(attr[0].cpu().detach().numpy())
             all_attrs.append(np.mean(attr_all_timesteps, axis=0))
         # Concatenate a‚ttribution values for all instances along the batch dimension
-        print(np.shape(all_attrs))
         all_attrs = np.mean(all_attrs, axis=0)
-        print(np.shape(all_attrs))
-        means_feature = all_attrs.mean(axis=(1))
+
+        means_feature = all_attrs.mean(axis=(0))
         # self.log('Attribution Featues', means_feature)
         # Compute mean along the batch dimension
-        means = all_attrs.mean(axis=(0))
+        means = all_attrs.mean(axis=(1))
        # self.log('Attribution Time steps', means)
         # Normalize the means values to range [0, 1]
         # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
             # Create x values (assuming you want a simple sequential x-axis)
             # Assuming you have 24 values
-            x_values = np.arange(1, 57)
+            x_values = np.arange(1, 58)
             # Plotting the featrue means
             plt.figure(figsize=(8, 6))
             plt.plot(
@@ -601,7 +603,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
         return means
 
     def Faithfulness_Correlation(
-        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=6
+        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=3
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.
@@ -688,13 +690,11 @@ def Faithfulness_Correlation(
                 )
                 # Predict on perturbed input x.
                 y_pred_perturb = self(data).detach().cpu().numpy()
-
                 pred_deltas.append((y_pred - y_pred_perturb).mean(axis=(0, 2)))
 
                 # Sum attributions of the random subset.
 
                 att_sums.append(np.sum(attribution[a_ix]))
-
             similarities.append(similarity_func(pred_deltas, att_sums))
             score = np.nanmean(similarities)
         return score

From 25446047f514a36bbe69f6a0a2d3397e53c05989 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 14 Nov 2023 14:20:35 +0100
Subject: [PATCH 106/142] labels for los not shown to RNN and deepAR

---
 configs/prediction_models/DeepARpytorch.gin   |  9 ++++----
 configs/prediction_models/RNNpytorch.gin      |  4 ++--
 configs/prediction_models/TFTpytorch.gin      |  6 ++---
 configs/prediction_models/common/DLCommon.gin |  2 +-
 icu_benchmarks/data/loader.py                 | 10 ++++----
 icu_benchmarks/models/constants.py            |  2 +-
 icu_benchmarks/models/dl_models.py            | 12 ++++++----
 icu_benchmarks/models/train.py                | 13 ++++++-----
 icu_benchmarks/models/wrappers.py             | 23 ++++---------------
 9 files changed, 35 insertions(+), 46 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 94ae90f0..d4962081 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -13,7 +13,7 @@ optimizer/hyperparameter.weight_decay = 1e-6
 # Model params
 
 model/hyperparameter.class_to_tune = @DeepARpytorch
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.hidden = (4, 64, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
@@ -26,11 +26,10 @@ PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
 PredictionDatasetTFTpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
 PredictionDatasetTFTpytorch.add_relative_time_idx=False
-PredictionDatasetTFTpytorch.target=[
-                "label",
-            ]
+PredictionDatasetTFTpytorch.target="label"
+            
 PredictionDatasetTFTpytorch.time_varying_unknown_reals=[
-                "label",
+                "label"
             ]
 
             
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 334def2d..0e623606 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -13,8 +13,8 @@ optimizer/hyperparameter.weight_decay = 1e-6
 # Model params
 
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
-model/hyperparameter.rnn_layers=(1, 3)
+model/hyperparameter.hidden = (4, 64, "log-uniform", 2)
+model/hyperparameter.rnn_layers=(1,3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
 model/hyperparameter.dropout = (0.0, 0.4)
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index abb81f6f..5ff809a5 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -12,7 +12,7 @@ optimizer/hyperparameter.weight_decay = 1e-6
 
 # Model params
 model/hyperparameter.class_to_tune = @TFTpytorch
-model/hyperparameter.hidden = (16,32,64,128)
+model/hyperparameter.hidden = (16,32,64)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.dropout = (0.0, 0.4)
 model/hyperparameter.dropout_att = (0.0, 0.4)
@@ -25,7 +25,7 @@ PredictionDatasetTFTpytorch.max_encoder_length = 24
 PredictionDatasetTFTpytorch.max_prediction_length = 24
 PredictionDatasetTFTpytorch.target="label"
 PredictionDatasetTFTpytorch.time_varying_known_reals=["time_idx"]
-PredictionDatasetTFTpytorch.add_relative_time_idx=True
+PredictionDatasetTFTpytorch.add_relative_time_idx=False
 PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
 PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "alp",
@@ -75,4 +75,4 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "tnt",
                 "urine",
                 "wbc",]
-PredictionDatasetTFTpytorch.lagged_variables=[]
\ No newline at end of file
+PredictionDatasetTFTpytorch.lagged_variables=[]
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 8f5d648e..8e4149fd 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,7 +15,7 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-train_common.batch_size = 128
+train_common.batch_size = 1
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 6287810b..0f16300b 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -3,7 +3,7 @@
 import pandas as pd
 import gin
 import numpy as np
-from torch import Tensor, cat, from_numpy, float32
+from torch import Tensor, cat, from_numpy, float32, min, max
 from torch.utils.data import Dataset
 import logging
 from typing import Dict, Tuple, Union
@@ -498,7 +498,7 @@ def __init__(
             time_varying_unknown_categoricals=time_varying_unknown_categoricals,
             time_varying_unknown_reals=time_varying_unknown_reals,
             add_relative_time_idx=add_relative_time_idx,
-            add_target_scales=True,
+            # add_target_scales=True,
             add_encoder_length=True,
             predict_mode=True,
             target_normalizer=None,
@@ -521,8 +521,10 @@ def get_feature_names(self):
         return self.column_names
 
     def randomize_labels(self, num_classes=None, min=None, max=None):
-        if num_classes is None:
-            random_target = np.random.uniform(min, max, size=len(self.data["target"][0]))
+        if num_classes == 1:
+
+            random_target = np.random.uniform(self.data["target"][0].min(),
+                                              self.data["target"][0].max(), size=len(self.data["target"][0]))
         else:
             random_target = np.random.randint(num_classes, size=len(self.data["target"][0]))
         self.data["target"][0] = Tensor(random_target)
diff --git a/icu_benchmarks/models/constants.py b/icu_benchmarks/models/constants.py
index 49555d42..093d15bd 100644
--- a/icu_benchmarks/models/constants.py
+++ b/icu_benchmarks/models/constants.py
@@ -64,7 +64,7 @@ class DLMetrics:
 
     BINARY_CLASSIFICATION_TORCHMETRICS = {
         "AUC": AUROC(task="binary"),
-        "PR": TorchMetricsAveragePrecision(task="binary"),
+        # "PR": TorchMetricsAveragePrecision(task="binary"),
         # "PrecisionRecallCurve": TorchMetricsPrecisionRecallCurve(task="binary"),
         "Calibration_Error": CalibrationError(task="binary", n_bins=10),
         "F1": F1Score(task="binary", num_classes=2),
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 5f1d287e..1cd782a5 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -636,14 +636,14 @@ def forward(
             "groups": tuple_x[9],
             "target_scale": tuple_x[10],
         }
-        out = self.model(x_dict)
+        if self.num_classes == 1:
+            x_dict['encoder_cont'][:, :, -1] = 0.0
 
-        # out = out["prediction"].reshape(1, -1)
+        out = self.model(x_dict)
 
-        # pred = self.logit(out["prediction"])
-        # print(pred
+        pred = self.logit(out["prediction"])
 
-        return out["prediction"]
+        return pred
 
 
 @gin.configurable
@@ -697,6 +697,8 @@ def forward(
             "groups": tuple_x[9],
             "target_scale": tuple_x[10],
         }
+        if self.num_classes == 1:
+            x_dict['encoder_cont'][:, :, -1] = 0.0
         out = self.model(x_dict)
         out = out["prediction"][0]
 
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 018c8bc3..be8e638d 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -283,23 +283,24 @@ def train_common(
         # print("shap", attributions_SHAP)
         attributions_IG = model.explantation_captum(
             test_loader=test_loader,
-            method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=35
+            method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=20
         )
+        print(test_dataset.reals)
 
         Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
-       # print("attention", Attention_weights)
+        print("attention", Attention_weights)
         if XAI_metric:
 
             ra2 = np.random.randint(low=0, high=100, size=24)
             random_attributions = np.random.normal(size=24)
 
-            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions)
-            F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions)
+            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
+            F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
             print('Random noraml faithfulness correlation', F_baseline)
             print('random unifrom ', F_baseline2)
-            F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG)
+            F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
             print('Attributions faithfulness correlation', F_attribution)
-            F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"])
+            F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"], pertrub='Noise')
             print('Attention faithfulness correlation', F_attention)
 
         # path = Path(random_model_dir)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 41eee23b..ca3093bd 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -316,21 +316,7 @@ def step_fn(self, element, step_prefix=""):
             element (object):
             step_prefix (str): Step type, by default: test, train, val.
         """
-        """
-        if isinstance(element[0], OrderedDict):  # check if the data loader is the one for the TFT nvidia implementation
-            data, mask = element[0], element[1].to(self.device)
-
-            for key, value in data.items():
-                value = value.float().to(self.device)
-
-                if value.shape[-1] == 1:
-                    value = value.squeeze(-1)
-                if value.dim() == 3:
-                    value = value.permute(0, 2, 1)
-                if key == "target":
-                    labels = value.squeeze()
-                data[key] = value
-         """
+
         if len(element) == 2:
             data, labels = element[0], (element[1]).to(self.device)
             if isinstance(data, list):
@@ -474,7 +460,6 @@ def step_fn(self, element, step_prefix=""):
         prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
 
         target = torch.masked_select(labels, mask).to(self.device)
-
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
             loss = self.loss(prediction, target.long(), weight=self.loss_weights.to(self.device)) + aux_loss
@@ -544,7 +529,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             explantation = method(self.forward_captum)
             # Reformat attributions.
             attr_all_timesteps = []
-            for time_step in range(23, 24):
+            for time_step in range(0, 24):
                 attr = explantation.attribute(
                     data, target=(time_step),  baselines=baselines, **kwargs
                 )
@@ -564,7 +549,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
         # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
             # Create x values (assuming you want a simple sequential x-axis)
-            # Assuming you have 24 values
+            # Assuming you have 57 variables
             x_values = np.arange(1, 58)
             # Plotting the featrue means
             plt.figure(figsize=(8, 6))
@@ -603,7 +588,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
         return means
 
     def Faithfulness_Correlation(
-        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=3
+        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=6, features=False
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.

From f435717f49ad1124dd7ba5f14fa9c3ece5c1cd44 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 15 Nov 2023 15:49:30 +0100
Subject: [PATCH 107/142] added shap, sailency explantations and logged values

---
 icu_benchmarks/data/loader.py     |  2 +-
 icu_benchmarks/models/train.py    | 66 ++++++++++++++++++++++++-------
 icu_benchmarks/models/wrappers.py | 31 +++++++++------
 icu_benchmarks/run_utils.py       |  5 +++
 4 files changed, 75 insertions(+), 29 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 0f16300b..57b2df55 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -498,7 +498,7 @@ def __init__(
             time_varying_unknown_categoricals=time_varying_unknown_categoricals,
             time_varying_unknown_reals=time_varying_unknown_reals,
             add_relative_time_idx=add_relative_time_idx,
-            # add_target_scales=True,
+            add_target_scales=True,
             add_encoder_length=True,
             predict_mode=True,
             target_normalizer=None,
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index be8e638d..6778d642 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -3,6 +3,7 @@
 import torch
 import pickle
 import logging
+import json
 import pandas as pd
 from joblib import load
 from torch.optim import Adam
@@ -27,7 +28,7 @@
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
-from captum.attr import IntegratedGradients, ShapleyValueSampling
+from captum.attr import IntegratedGradients, ShapleyValueSampling, Saliency, GuidedBackprop, LRP
 
 
 cpu_core_count = (
@@ -275,33 +276,68 @@ def train_common(
         return 0
 
     if explain:
-        #
-        # attributions_SHAP = model.explantation_captum(
-        #    test_loader=test_loader,
-        #    method=ShapleyValueSampling, log_dir=log_dir, plot=True
-        # )
-        # print("shap", attributions_SHAP)
+
+        attributions_Saliency = model.explantation_captum(
+            test_loader=test_loader,
+            method=Saliency, log_dir=log_dir, plot=True
+        )
+        print("saliency", attributions_Saliency)
+        attributions_shap = model.explantation_captum(
+            test_loader=test_loader,
+            method=ShapleyValueSampling, log_dir=log_dir, plot=True, n_samples=10
+        )
+        print("shap", attributions_shap)
         attributions_IG = model.explantation_captum(
             test_loader=test_loader,
             method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=20
         )
-        print(test_dataset.reals)
+        print("IG", attributions_IG)
+        Interpertations = model.interpertations(test_loader, log_dir, plot=True)
+        print("attention", Interpertations)
+        attributions_dict = {
+            "attributions_Saliency": attributions_Saliency.tolist(),
+            "attributions_IG": attributions_IG.tolist(),
+            "attributions_shap": attributions_shap.tolist(),
+            "attention_weights": Interpertations["attention"].tolist(),
+            "static_variables": Interpertations["static_variables"].tolist(),
+            "encoder_variables": Interpertations["encoder_variables"].tolist()
+        }
+
+        # Path to the JSON file in log_dir
+        json_file_path = f"{log_dir}/Attributions.json"
 
-        Attention_weights = model.interpertations(test_loader, log_dir, plot=True)
-        print("attention", Attention_weights)
+        # Write the dictionary to a JSON file
+        with open(json_file_path, 'w') as json_file:
+            json.dump(attributions_dict, json_file)
         if XAI_metric:
 
-            ra2 = np.random.randint(low=0, high=100, size=24)
             random_attributions = np.random.normal(size=24)
 
             F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
-            F_baseline2 = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
-            print('Random noraml faithfulness correlation', F_baseline)
-            print('random unifrom ', F_baseline2)
+            print('Random normal faithfulness correlation', F_baseline)
+
             F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
             print('Attributions faithfulness correlation', F_attribution)
-            F_attention = model.Faithfulness_Correlation(test_loader, Attention_weights["attention"], pertrub='Noise')
+            F_attention = model.Faithfulness_Correlation(test_loader, Interpertations["attention"], pertrub='Noise')
             print('Attention faithfulness correlation', F_attention)
+            F_saliency = model.Faithfulness_Correlation(test_loader, attributions_Saliency, pertrub='Noise')
+            print('Saliency faithfulness correlation', F_saliency)
+            F_shap = model.Faithfulness_Correlation(test_loader, attributions_shap, pertrub='Noise')
+            print('shap faithfulness correlation', F_shap)
+            XAI_dict = {
+                "Faithfulness_correlation_normal_random": F_baseline,
+                "Faithfulness_correlation_IG": F_attribution,
+                "Faithfulness_correlation_attention": F_attention,
+                "Faithfulness_correlation_saliency": F_saliency,
+                "Faithfulness_correlation_shap": F_shap
+            }
+
+        # Path to the JSON file in log_dir
+        json_file_path = f"{log_dir}/XAI_metrics.json"
+
+        # Write the dictionary to a JSON file
+        with open(json_file_path, 'w') as json_file:
+            json.dump(XAI_dict, json_file)
 
         # path = Path(random_model_dir)
 
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index ca3093bd..32e05423 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -20,7 +20,7 @@
 from icu_benchmarks.contants import RunMode
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean
-
+import captum
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.mse_loss, module="torch.nn.functional")
@@ -513,15 +513,15 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
                 x["target_scale"].requires_grad_(),
             )
             baselines = (
-                torch.zeros_like(data[0]).to(self.device),  # encoder_cat, set to zero
-                torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
-                torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
+                data[0].to(self.device),  # encoder_cat, set to random
+                torch.randn_like(data[1]).to(self.device),  # encoder_cont, set to random
+                torch.randn_like(data[2]).to(self.device),  # encoder_target, set to random
                 data[3].to(self.device),  # encoder_lengths, leave unchanged
-                torch.zeros_like(data[4]).to(self.device),  # decoder_cat, set to zero
-                torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
-                torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
+                torch.randn_like(data[4]).to(self.device),  # decoder_cat, set to random
+                torch.randn_like(data[5]).to(self.device),  # decoder_cont, set to random
+                torch.randn_like(data[6]).to(self.device),  # decoder_target, set to random
                 data[7].to(self.device),  # decoder_lengths, leave unchanged
-                torch.zeros_like(data[8]).to(self.device),  # decoder_time_idx, set to zero
+                data[8].to(self.device),  # decoder_time_idx, unchanged
                 data[9].to(self.device),  # groups, leave unchanged
                 data[10].to(self.device),  # target_scale, leave unchanged
             )
@@ -529,10 +529,15 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             explantation = method(self.forward_captum)
             # Reformat attributions.
             attr_all_timesteps = []
-            for time_step in range(0, 24):
-                attr = explantation.attribute(
-                    data, target=(time_step),  baselines=baselines, **kwargs
-                )
+            for time_step in range(23, 24):
+                if method is not captum.attr.Saliency:
+                    attr = explantation.attribute(
+                        data, target=(time_step),  baselines=baselines, **kwargs
+                    )
+                else:
+                    attr = explantation.attribute(
+                        data, target=(time_step), **kwargs
+                    )
                 # Convert attributions to numpy array and append to the list
 
                 attr_all_timesteps.append(attr[0].cpu().detach().numpy())
@@ -588,7 +593,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
         return means
 
     def Faithfulness_Correlation(
-        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=6, features=False
+        self, test_loader, attribution, similarity_func=None, nr_runs=2, pertrub=None, subset_size=6, features=False
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 4bc3985c..0f559e4c 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -213,11 +213,16 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
                     with open(fold_iter / "val_metrics.csv", "r") as f:
                         result = json.load(f)
                         aggregated[repetition.name][fold_iter.name].update(result)
+
                 # Add durations to metrics
                 if (fold_iter / "durations.json").is_file():
                     with open(fold_iter / "durations.json", "r") as f:
                         result = json.load(f)
                         aggregated[repetition.name][fold_iter.name].update(result)
+                if (fold_iter / "XAI_metrics.json").is_file():
+                    with open(fold_iter / "XAI_metrics.json", "r") as f:
+                        result = json.load(f)
+                        aggregated[repetition.name][fold_iter.name].update(result)
 
     # Aggregate results per metric
     list_scores = {}

From ce078a1f305e39ab43ee3b73ca1a66777877b398 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 20 Nov 2023 13:19:09 +0100
Subject: [PATCH 108/142] renamed dataloader , rescaled target,  removed target
 scale, target center

---
 configs/prediction_models/DeepARpytorch.gin   | 21 ++++-----
 configs/prediction_models/TFTpytorch.gin      |  2 +-
 configs/prediction_models/common/DLCommon.gin |  2 +-
 configs/prediction_models/common/DLTuning.gin |  2 +-
 configs/tasks/Regression.gin                  |  2 +-
 icu_benchmarks/data/loader.py                 |  8 ++--
 icu_benchmarks/models/train.py                | 46 ++++++++++---------
 icu_benchmarks/models/wrappers.py             | 18 ++++++--
 8 files changed, 55 insertions(+), 46 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index d4962081..27652b96 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -1,4 +1,4 @@
-# Settings for TFT model.
+# Settings for DeepAR model.
 
 # Common settings for DL models
 include "configs/prediction_models/common/DLCommon.gin"
@@ -22,19 +22,16 @@ model/hyperparameter.lr_scheduler = "exponential"
 
 
 # Dataset params
-PredictionDatasetTFTpytorch.max_encoder_length = 24
-PredictionDatasetTFTpytorch.max_prediction_length = 24
-PredictionDatasetTFTpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
-PredictionDatasetTFTpytorch.add_relative_time_idx=False
-PredictionDatasetTFTpytorch.target="label"
-            
-PredictionDatasetTFTpytorch.time_varying_unknown_reals=[
+PredictionDatasetpytorch.max_encoder_length = 24
+PredictionDatasetpytorch.max_prediction_length = 24
+PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
+PredictionDatasetpytorch.add_relative_time_idx=False
+PredictionDatasetpytorch.target="label"
+            PredictionDatasetTFTpytorch.time_varying_unknown_reals=[
                 "label"
             ]
-
-            
-PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetTFTpytorch.lagged_variables=["alb",
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.lagged_variables=["alb",
                 "alp",
                 "alt",
                 "ast",
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 5ff809a5..ebd26e4e 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -22,7 +22,7 @@ model/hyperparameter.lr_scheduler = "exponential"
 
 # Dataset params
 PredictionDatasetTFTpytorch.max_encoder_length = 24
-PredictionDatasetTFTpytorch.max_prediction_length = 24
+PredictionDatasetTFTpytorch.max_prediction_length = 1
 PredictionDatasetTFTpytorch.target="label"
 PredictionDatasetTFTpytorch.time_varying_known_reals=["time_idx"]
 PredictionDatasetTFTpytorch.add_relative_time_idx=False
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 8e4149fd..fa4b440f 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,7 +15,7 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-train_common.batch_size = 1
+train_common.batch_size = 64
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
diff --git a/configs/prediction_models/common/DLTuning.gin b/configs/prediction_models/common/DLTuning.gin
index b4d13e12..ec0240e2 100644
--- a/configs/prediction_models/common/DLTuning.gin
+++ b/configs/prediction_models/common/DLTuning.gin
@@ -1,5 +1,5 @@
 # Hyperparameter tuner settings for Deep Learning.
 tune_hyperparameters.scopes = ["model", "optimizer"]
 tune_hyperparameters.n_initial_points = 5
-tune_hyperparameters.n_calls = 30
+tune_hyperparameters.n_calls = 15
 tune_hyperparameters.folds_to_tune_on = 2
\ No newline at end of file
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index 6b6ac607..1234f707 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -26,7 +26,7 @@ preprocess.use_static = True
 
 # SPECIFYING REGRESSION OUTCOME SCALING
 base_regression_preprocessor.outcome_min = 0
-base_regression_preprocessor.outcome_max = 15
+base_regression_preprocessor.outcome_max = 168
 
 # SELECTING DATASET
 PredictionDataset.vars = %vars
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 57b2df55..68109f4b 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -429,8 +429,8 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         return from_numpy(window.values).to(float32)
 
 
-@gin.configurable("PredictionDatasetTFTpytorch")
-class PredictionDatasetTFTpytorch(TimeSeriesDataSet):
+@gin.configurable("PredictionDatasetpytorch")
+class PredictionDatasetpytorch(TimeSeriesDataSet):
     """Subclass of timeseries dataset works with pyotrch forecasting library .
 
     Args:
@@ -498,8 +498,8 @@ def __init__(
             time_varying_unknown_categoricals=time_varying_unknown_categoricals,
             time_varying_unknown_reals=time_varying_unknown_reals,
             add_relative_time_idx=add_relative_time_idx,
-            add_target_scales=True,
-            add_encoder_length=True,
+            # add_target_scales=True,
+            # add_encoder_length=True,
             predict_mode=True,
             target_normalizer=None,
         )
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 6778d642..c1306d14 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -22,7 +22,7 @@
 from icu_benchmarks.data.loader import (
     PredictionDataset,
     ImputationDataset,
-    PredictionDatasetTFTpytorch,
+    PredictionDatasetpytorch,
 )
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
@@ -111,7 +111,7 @@ def train_common(
         ImputationDataset
         if mode == RunMode.imputation
         else (
-            PredictionDatasetTFTpytorch if (pytorch_forecasting) else PredictionDataset
+            PredictionDatasetpytorch if (pytorch_forecasting) else PredictionDataset
         )
     )
 
@@ -277,31 +277,37 @@ def train_common(
 
     if explain:
 
-        attributions_Saliency = model.explantation_captum(
-            test_loader=test_loader,
-            method=Saliency, log_dir=log_dir, plot=True
-        )
-        print("saliency", attributions_Saliency)
+        # attributions_Saliency = model.explantation_captum(
+        #    test_loader=test_loader,
+        #    method=Saliency, log_dir=log_dir, plot=True
+        # )
+
+        # print("saliency", attributions_Saliency)
+        """
         attributions_shap = model.explantation_captum(
             test_loader=test_loader,
             method=ShapleyValueSampling, log_dir=log_dir, plot=True, n_samples=10
         )
         print("shap", attributions_shap)
-        attributions_IG = model.explantation_captum(
-            test_loader=test_loader,
-            method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=20
-        )
-        print("IG", attributions_IG)
+        """
+       # attributions_IG = model.explantation_captum(
+       #     test_loader=test_loader,
+       #     method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=20
+       # )
+       # print("IG", attributions_IG)
         Interpertations = model.interpertations(test_loader, log_dir, plot=True)
         print("attention", Interpertations)
         attributions_dict = {
             "attributions_Saliency": attributions_Saliency.tolist(),
             "attributions_IG": attributions_IG.tolist(),
-            "attributions_shap": attributions_shap.tolist(),
+            # "attributions_shap": attributions_shap.tolist(),
             "attention_weights": Interpertations["attention"].tolist(),
             "static_variables": Interpertations["static_variables"].tolist(),
             "encoder_variables": Interpertations["encoder_variables"].tolist()
         }
+        variable_importance = np.concatenate((Interpertations["static_variables"].cpu().detach().numpy(),
+                                              Interpertations["encoder_variables"].cpu().detach().numpy()))
+        ind = np.argpartition(variable_importance, -5)[-5:]
 
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/Attributions.json"
@@ -310,26 +316,24 @@ def train_common(
         with open(json_file_path, 'w') as json_file:
             json.dump(attributions_dict, json_file)
         if XAI_metric:
-
             random_attributions = np.random.normal(size=24)
-
-            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
+            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise', ind=ind)
             print('Random normal faithfulness correlation', F_baseline)
-
             F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
             print('Attributions faithfulness correlation', F_attribution)
-            F_attention = model.Faithfulness_Correlation(test_loader, Interpertations["attention"], pertrub='Noise')
+            F_attention = model.Faithfulness_Correlation(
+                test_loader, Interpertations["attention"], pertrub='Noise', ind=ind)
             print('Attention faithfulness correlation', F_attention)
             F_saliency = model.Faithfulness_Correlation(test_loader, attributions_Saliency, pertrub='Noise')
             print('Saliency faithfulness correlation', F_saliency)
-            F_shap = model.Faithfulness_Correlation(test_loader, attributions_shap, pertrub='Noise')
-            print('shap faithfulness correlation', F_shap)
+           # F_shap = model.Faithfulness_Correlation(test_loader, attributions_shap, pertrub='Noise')
+           # print('shap faithfulness correlation', F_shap)
             XAI_dict = {
                 "Faithfulness_correlation_normal_random": F_baseline,
                 "Faithfulness_correlation_IG": F_attribution,
                 "Faithfulness_correlation_attention": F_attention,
                 "Faithfulness_correlation_saliency": F_saliency,
-                "Faithfulness_correlation_shap": F_shap
+                #    "Faithfulness_correlation_shap": F_shap
             }
 
         # Path to the JSON file in log_dir
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 32e05423..3b956900 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -529,7 +529,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             explantation = method(self.forward_captum)
             # Reformat attributions.
             attr_all_timesteps = []
-            for time_step in range(23, 24):
+            for time_step in range(0, 24):
                 if method is not captum.attr.Saliency:
                     attr = explantation.attribute(
                         data, target=(time_step),  baselines=baselines, **kwargs
@@ -593,7 +593,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
         return means
 
     def Faithfulness_Correlation(
-        self, test_loader, attribution, similarity_func=None, nr_runs=2, pertrub=None, subset_size=6, features=False
+        self, test_loader, attribution, similarity_func=None, nr_runs=2, pertrub=None, subset_size=3, ind=[]
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.
@@ -654,16 +654,24 @@ def Faithfulness_Correlation(
 
                 if pertrub == "Noise":
                     # add normal noise to input
+
                     noise = torch.randn_like(x["encoder_cont"])
+                    if (len(ind) == 0):
+                        x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
+                    else:
+
+                        x["encoder_cont"][:, a_ix[:, None], ind] += noise[:, a_ix[:, None], ind]
 
-                    x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
                 elif pertrub == "baseline":
                     # Create a mask tensor with zeros at specified time steps and ones everywhere else
                     # pytorch bug need to change to cpu for next step and then revert
                     mask = torch.ones_like(x["encoder_cont"]).cpu()
-
-                    mask[:, a_ix, :] = 0
+                    if (len(ind) == 0):
+                        mask[:, a_ix, :] = 0
+                    else:
+                        mask[:, a_ix, ind] = 0
                     mask = mask.to(x["encoder_cont"].device)
+
                     x["encoder_cont"] = x["encoder_cont"] * mask
                 data = (
                     x["encoder_cat"],

From dd3bdac833db983ee9117d0be3e4131970d94261 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 20 Nov 2023 13:19:48 +0100
Subject: [PATCH 109/142] changed gin to match DL

---
 configs/prediction_models/RNNpytorch.gin | 18 +++++++++---------
 configs/prediction_models/TFTpytorch.gin | 16 ++++++++--------
 2 files changed, 17 insertions(+), 17 deletions(-)

diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 0e623606..75233dca 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -1,4 +1,4 @@
-# Settings for TFT model.
+# Settings for RNN model.
 
 # Common settings for DL models
 include "configs/prediction_models/common/DLCommon.gin"
@@ -22,18 +22,18 @@ model/hyperparameter.lr_scheduler = "exponential"
 
 
 # Dataset params
-PredictionDatasetTFTpytorch.max_encoder_length = 24
-PredictionDatasetTFTpytorch.max_prediction_length = 24
-PredictionDatasetTFTpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
+PredictionDatasetpytorch.max_encoder_length = 24
+PredictionDatasetpytorch.max_prediction_length = 24
+PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
 
-PredictionDatasetTFTpytorch.add_relative_time_idx=False
-PredictionDatasetTFTpytorch.target="label"
+PredictionDatasetpytorch.add_relative_time_idx=False
+PredictionDatasetpytorch.target="label"
             
-PredictionDatasetTFTpytorch.time_varying_unknown_reals=[
+PredictionDatasetpytorch.time_varying_unknown_reals=[
                 "label",
             ]
-PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetTFTpytorch.lagged_variables=["alb",
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.lagged_variables=["alb",
                 "alp",
                 "alt",
                 "ast",
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index ebd26e4e..94a42740 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -21,13 +21,13 @@ model/hyperparameter.lr_scheduler = "exponential"
 
 
 # Dataset params
-PredictionDatasetTFTpytorch.max_encoder_length = 24
-PredictionDatasetTFTpytorch.max_prediction_length = 1
-PredictionDatasetTFTpytorch.target="label"
-PredictionDatasetTFTpytorch.time_varying_known_reals=["time_idx"]
-PredictionDatasetTFTpytorch.add_relative_time_idx=False
-PredictionDatasetTFTpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
+PredictionDatasetpytorch.max_encoder_length = 24
+PredictionDatasetpytorch.max_prediction_length = 1
+PredictionDatasetpytorch.target="label"
+PredictionDatasetpytorch.time_varying_known_reals=["time_idx"]
+PredictionDatasetpytorch.add_relative_time_idx=False
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.time_varying_unknown_reals=["alb",
                 "alp",
                 "alt",
                 "ast",
@@ -75,4 +75,4 @@ PredictionDatasetTFTpytorch.time_varying_unknown_reals=["alb",
                 "tnt",
                 "urine",
                 "wbc",]
-PredictionDatasetTFTpytorch.lagged_variables=[]
+PredictionDatasetpytorch.lagged_variables=[]

From 2defb3799bca3f2bbec504d012f947997434e3b1 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 20 Nov 2023 15:23:17 +0100
Subject: [PATCH 110/142] changes to accomadte for los attribution, code more
 reusable now

---
 configs/prediction_models/DeepARpytorch.gin |   2 +-
 icu_benchmarks/models/train.py              |  85 ++++++++--------
 icu_benchmarks/models/wrappers.py           | 101 ++++++++++----------
 3 files changed, 95 insertions(+), 93 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 27652b96..bb18cbd0 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -27,7 +27,7 @@ PredictionDatasetpytorch.max_prediction_length = 24
 PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
 PredictionDatasetpytorch.add_relative_time_idx=False
 PredictionDatasetpytorch.target="label"
-            PredictionDatasetTFTpytorch.time_varying_unknown_reals=[
+PredictionDatasetpytorch.time_varying_unknown_reals=[
                 "label"
             ]
 PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index c1306d14..bf06ac0c 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -28,7 +28,7 @@
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
-from captum.attr import IntegratedGradients, ShapleyValueSampling, Saliency, GuidedBackprop, LRP
+from captum.attr import IntegratedGradients, ShapleyValueSampling, Saliency, GuidedBackprop, LRP, FeatureAblation, Lime
 
 
 cpu_core_count = (
@@ -276,38 +276,49 @@ def train_common(
         return 0
 
     if explain:
-
-        # attributions_Saliency = model.explantation_captum(
-        #    test_loader=test_loader,
-        #    method=Saliency, log_dir=log_dir, plot=True
-        # )
-
-        # print("saliency", attributions_Saliency)
-        """
-        attributions_shap = model.explantation_captum(
-            test_loader=test_loader,
-            method=ShapleyValueSampling, log_dir=log_dir, plot=True, n_samples=10
-        )
-        print("shap", attributions_shap)
-        """
-       # attributions_IG = model.explantation_captum(
-       #     test_loader=test_loader,
-       #     method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=20
-       # )
-       # print("IG", attributions_IG)
+        attributions_dict = {}
         Interpertations = model.interpertations(test_loader, log_dir, plot=True)
+        attributions_dict["attention_weights"]= Interpertations["attention"].tolist()
+        attributions_dict["static_variables"]= Interpertations["static_variables"].tolist()
+        attributions_dict["encoder_variables"]= Interpertations["encoder_variables"].tolist()
         print("attention", Interpertations)
-        attributions_dict = {
-            "attributions_Saliency": attributions_Saliency.tolist(),
-            "attributions_IG": attributions_IG.tolist(),
-            # "attributions_shap": attributions_shap.tolist(),
-            "attention_weights": Interpertations["attention"].tolist(),
-            "static_variables": Interpertations["static_variables"].tolist(),
-            "encoder_variables": Interpertations["encoder_variables"].tolist()
+        if XAI_metric:
+            random_attributions = np.random.normal(size=24)
+            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
+            print('Random normal faithfulness correlation', F_baseline)
+            F_attention = model.Faithfulness_Correlation(
+                test_loader, Interpertations["attention"], pertrub='Noise')
+            print('Attention faithfulness correlation', F_attention)
+        methods = {
+            "Saliency": Saliency,
+            "Lime": Lime,
+            "IG": IntegratedGradients,
         }
-        variable_importance = np.concatenate((Interpertations["static_variables"].cpu().detach().numpy(),
-                                              Interpertations["encoder_variables"].cpu().detach().numpy()))
-        ind = np.argpartition(variable_importance, -5)[-5:]
+        for key,item in methods.items():
+            attribution=
+            attributions_Saliency = model.explantation_captum(
+                test_loader=test_loader,
+                method=Saliency, log_dir=log_dir, plot=True
+            )
+
+            print("saliency", attributions_Saliency)
+
+            attributions_lime = model.explantation_captum(
+                test_loader=test_loader,
+                method=Lime, log_dir=log_dir, plot=True, n_samples=10
+            )
+            print("FA", attributions_lime)
+
+            attributions_IG = model.explantation_captum(
+                test_loader=test_loader,
+                method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=20
+            )
+            print("IG", attributions_IG)
+        
+        
+        # variable_importance = np.concatenate((Interpertations["static_variables"].cpu().detach().numpy(),
+        #                                      Interpertations["encoder_variables"].cpu().detach().numpy()))
+        # ind = np.argpartition(variable_importance, -5)[-5:]
 
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/Attributions.json"
@@ -316,24 +327,20 @@ def train_common(
         with open(json_file_path, 'w') as json_file:
             json.dump(attributions_dict, json_file)
         if XAI_metric:
-            random_attributions = np.random.normal(size=24)
-            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise', ind=ind)
-            print('Random normal faithfulness correlation', F_baseline)
+            
             F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
             print('Attributions faithfulness correlation', F_attribution)
-            F_attention = model.Faithfulness_Correlation(
-                test_loader, Interpertations["attention"], pertrub='Noise', ind=ind)
-            print('Attention faithfulness correlation', F_attention)
+            
             F_saliency = model.Faithfulness_Correlation(test_loader, attributions_Saliency, pertrub='Noise')
             print('Saliency faithfulness correlation', F_saliency)
-           # F_shap = model.Faithfulness_Correlation(test_loader, attributions_shap, pertrub='Noise')
-           # print('shap faithfulness correlation', F_shap)
+            F_lime = model.Faithfulness_Correlation(test_loader, attributions_lime, pertrub='Noise')
+            print('shap faithfulness correlation', F_lime)
             XAI_dict = {
                 "Faithfulness_correlation_normal_random": F_baseline,
                 "Faithfulness_correlation_IG": F_attribution,
                 "Faithfulness_correlation_attention": F_attention,
                 "Faithfulness_correlation_saliency": F_saliency,
-                #    "Faithfulness_correlation_shap": F_shap
+                "Faithfulness_correlation_shap": F_lime
             }
 
         # Path to the JSON file in log_dir
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 3b956900..5898c8fa 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -528,39 +528,36 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
 
             explantation = method(self.forward_captum)
             # Reformat attributions.
-            attr_all_timesteps = []
-            for time_step in range(0, 24):
-                if method is not captum.attr.Saliency:
-                    attr = explantation.attribute(
-                        data, target=(time_step),  baselines=baselines, **kwargs
-                    )
-                else:
-                    attr = explantation.attribute(
-                        data, target=(time_step), **kwargs
-                    )
+            # attr_all_timesteps = []
+            # for time_step in range(0, 24):
+            if method is not captum.attr.Saliency:
+                attr = explantation.attribute(
+                    data,  baselines=baselines, **kwargs
+                )
+            else:
+                attr = explantation.attribute(
+                    data,  **kwargs
+                )
                 # Convert attributions to numpy array and append to the list
-
-                attr_all_timesteps.append(attr[0].cpu().detach().numpy())
-            all_attrs.append(np.mean(attr_all_timesteps, axis=0))
+            all_attrs.append(attr[0].cpu().detach().numpy())
         # Concatenate a‚ttribution values for all instances along the batch dimension
-        all_attrs = np.mean(all_attrs, axis=0)
-
-        means_feature = all_attrs.mean(axis=(0))
-        # self.log('Attribution Featues', means_feature)
-        # Compute mean along the batch dimension
-        means = all_attrs.mean(axis=(1))
-       # self.log('Attribution Time steps', means)
-        # Normalize the means values to range [0, 1]
+
+        all_attrs = all_attrs.mean(axis=(0))
+
+        features_attrs = all_attrs.mean(axis=(0))
+
+        timestep_attrs = all_attrs.mean(axis=(1))
+
         # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
             # Create x values (assuming you want a simple sequential x-axis)
             # Assuming you have 57 variables
-            x_values = np.arange(1, 58)
+            x_values = np.arange(1, 54)
             # Plotting the featrue means
             plt.figure(figsize=(8, 6))
             plt.plot(
                 x_values,
-                means_feature,
+                features_attrs,
                 marker="o",
                 color="skyblue",
                 linestyle="-",
@@ -577,7 +574,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             x_values = np.arange(1, 25)
             plt.plot(
                 x_values,
-                means,
+                timestep_attrs,
                 marker="o",
                 color="skyblue",
                 linestyle="-",
@@ -590,10 +587,10 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             plt.xticks(x_values)  # Set x-ticks to match the number of features
             plt.tight_layout()
             plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
-        return means
+        return all_attrs, features_attrs, timestep_attrs
 
     def Faithfulness_Correlation(
-        self, test_loader, attribution, similarity_func=None, nr_runs=2, pertrub=None, subset_size=3, ind=[]
+        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=3, feature=False, time_step=True, feature_timestep=False
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.
@@ -656,20 +653,16 @@ def Faithfulness_Correlation(
                     # add normal noise to input
 
                     noise = torch.randn_like(x["encoder_cont"])
-                    if (len(ind) == 0):
-                        x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
-                    else:
 
-                        x["encoder_cont"][:, a_ix[:, None], ind] += noise[:, a_ix[:, None], ind]
+                    x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
 
                 elif pertrub == "baseline":
                     # Create a mask tensor with zeros at specified time steps and ones everywhere else
                     # pytorch bug need to change to cpu for next step and then revert
                     mask = torch.ones_like(x["encoder_cont"]).cpu()
-                    if (len(ind) == 0):
-                        mask[:, a_ix, :] = 0
-                    else:
-                        mask[:, a_ix, ind] = 0
+
+                    mask[:, a_ix, :] = 0
+
                     mask = mask.to(x["encoder_cont"].device)
 
                     x["encoder_cont"] = x["encoder_cont"] * mask
@@ -697,7 +690,7 @@ def Faithfulness_Correlation(
             score = np.nanmean(similarities)
         return score
 
-    def Data_Randomization(self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None):
+    def Data_Randomization(self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None, **kwargs):
         """
         Implementation of the Random Logit Metric by Sixt et al., 2020.
 
@@ -740,31 +733,33 @@ def Data_Randomization(self, test_loader, attribution, explain_method, random_mo
                 )
 
                 baselines = (
-                    torch.zeros_like(data[0]).to(random_model.device),  # encoder_cat, set to zero
-                    torch.zeros_like(data[1]).to(random_model.device),  # encoder_cont, set to zero
-                    torch.zeros_like(data[2]).to(random_model.device),  # encoder_target, set to zero
-                    data[3].to(random_model.device),  # encoder_lengths, leave unchanged
-                    torch.zeros_like(data[4]).to(random_model.device),  # decoder_cat, set to zero
-                    torch.zeros_like(data[5]).to(random_model.device),  # decoder_cont, set to zero
-                    torch.zeros_like(data[6]).to(random_model.device),  # decoder_target, set to zero
-                    data[7].to(random_model.device),  # decoder_lengths, leave unchanged
-                    torch.zeros_like(data[8]).to(random_model.device),  # decoder_time_idx, set to zero
-                    data[9].to(random_model.device),  # groups, leave unchanged
-                    data[10].to(random_model.device),  # target_scale, leave unchanged
+                    data[0].to(self.device),  # encoder_cat, set to random
+                    torch.randn_like(data[1]).to(self.device),  # encoder_cont, set to random
+                    torch.randn_like(data[2]).to(self.device),  # encoder_target, set to random
+                    data[3].to(self.device),  # encoder_lengths, leave unchanged
+                    torch.randn_like(data[4]).to(self.device),  # decoder_cat, set to random
+                    torch.randn_like(data[5]).to(self.device),  # decoder_cont, set to random
+                    torch.randn_like(data[6]).to(self.device),  # decoder_target, set to random
+                    data[7].to(self.device),  # decoder_lengths, leave unchanged
+                    data[8].to(self.device),  # decoder_time_idx, unchanged
+                    data[9].to(self.device),  # groups, leave unchanged
+                    data[10].to(self.device),  # target_scale, leave unchanged
                 )
 
                 explantation = explain_method(random_model.forward_captum)
                 # Reformat attributions.
-                attr_all_timesteps = []
-                for time_step in range(0, 24):
-                    attr, delta = explantation.attribute(
-                        data, target=(time_step, 1), return_convergence_delta=True, baselines=baselines, n_steps=35
+                if explain_method is not captum.attr.Saliency:
+                    attr = explantation.attribute(
+                        data,  baselines=baselines, **kwargs
+                    )
+                else:
+                    attr = explantation.attribute(
+                        data,  **kwargs
                     )
                     # Convert attributions to numpy array and append to the list
-                    attr_all_timesteps.append(attr[0].cpu().detach().numpy())
-                a_perturbed.append(np.mean(attr_all_timesteps, axis=0))
+                a_perturbed.append(attr[0].cpu().detach().numpy())
 
-            a_perturbed = np.mean(a_perturbed, axis=(0, 1, 3))
+            a_perturbed = np.mean(a_perturbed, axis=(0))
 
             # Normalize the means values to range [0, 1]
             # normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())

From 4c8ee4a8a736c00f1d1c1bec84a4c3795db9aa2a Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 21 Nov 2023 16:23:49 +0100
Subject: [PATCH 111/142] changed to allow for timesteps and variable per
 timestep attribution

---
 icu_benchmarks/models/train.py    | 99 ++++++++++++++++---------------
 icu_benchmarks/models/wrappers.py | 67 +++++++++++++--------
 2 files changed, 93 insertions(+), 73 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index bf06ac0c..c0a055f5 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -278,47 +278,66 @@ def train_common(
     if explain:
         attributions_dict = {}
         Interpertations = model.interpertations(test_loader, log_dir, plot=True)
-        attributions_dict["attention_weights"]= Interpertations["attention"].tolist()
-        attributions_dict["static_variables"]= Interpertations["static_variables"].tolist()
-        attributions_dict["encoder_variables"]= Interpertations["encoder_variables"].tolist()
+        attributions_dict["attention_weights"] = Interpertations["attention"].tolist()
+        attributions_dict["static_variables"] = Interpertations["static_variables"].tolist()
+        attributions_dict["encoder_variables"] = Interpertations["encoder_variables"].tolist()
         print("attention", Interpertations)
+        model.train()
+        XAI_dict = {}
         if XAI_metric:
-            random_attributions = np.random.normal(size=24)
-            F_baseline = model.Faithfulness_Correlation(test_loader, random_attributions, pertrub='Noise')
-            print('Random normal faithfulness correlation', F_baseline)
+            XAI_dict = {}
+            # random attribution per timestep
+            random_attributions_ts = np.random.normal(size=24)
+            F_baseline_ts = model.Faithfulness_Correlation(
+                test_loader, random_attributions_ts, pertrub='Noise', subset_size=4, time_step=True, nr_runs=10)
+            print('Random normal faithfulness correlation for timesteps', F_baseline_ts)
+
             F_attention = model.Faithfulness_Correlation(
-                test_loader, Interpertations["attention"], pertrub='Noise')
+                test_loader, Interpertations["attention"], pertrub='Noise', subset_size=4, time_step=True, nr_runs=10)
             print('Attention faithfulness correlation', F_attention)
+            # random attribution per variable per timestep
+            random_attributions_v_ts = np.random.normal(size=[24, 53])
+            F_baseline_v_ts = model.Faithfulness_Correlation(
+                test_loader, random_attributions_v_ts, pertrub='Noise', feature_timestep=True, nr_runs=10, subset_size=[4, 9])
+            print('Random normal faithfulness correlation for variables per timesteps', F_baseline_v_ts)
+
+        XAI_dict["attention_faith"] = F_attention
+        XAI_dict["random_faith_timestep"] = random_attributions_ts
+        XAI_dict["random_faith_var_timestep"] = random_attributions_v_ts.tolist()
         methods = {
             "Saliency": Saliency,
-            "Lime": Lime,
+            #  "Lime": Lime,
             "IG": IntegratedGradients,
-        }
-        for key,item in methods.items():
-            attribution=
-            attributions_Saliency = model.explantation_captum(
-                test_loader=test_loader,
-                method=Saliency, log_dir=log_dir, plot=True
-            )
 
-            print("saliency", attributions_Saliency)
-
-            attributions_lime = model.explantation_captum(
-                test_loader=test_loader,
-                method=Lime, log_dir=log_dir, plot=True, n_samples=10
-            )
-            print("FA", attributions_lime)
 
-            attributions_IG = model.explantation_captum(
-                test_loader=test_loader,
-                method=IntegratedGradients, log_dir=log_dir, plot=True, n_steps=20
-            )
-            print("IG", attributions_IG)
-        
-        
-        # variable_importance = np.concatenate((Interpertations["static_variables"].cpu().detach().numpy(),
-        #                                      Interpertations["encoder_variables"].cpu().detach().numpy()))
-        # ind = np.argpartition(variable_importance, -5)[-5:]
+        }
+        for key, item in methods.items():
+            if key == "IG":
+                all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
+                    test_loader=test_loader,
+                    method=item, log_dir=log_dir, plot=True, n_steps=20
+                )
+            else:
+                all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
+                    test_loader=test_loader,
+                    method=item, log_dir=log_dir, plot=True
+                )
+            attributions_dict["{}_all".format(key)] = all_attrs.tolist()
+            attributions_dict["{}_timesteps".format(key)] = timestep_attrs.tolist()
+            attributions_dict["{}_features".format(key)] = features_attrs.tolist()
+
+            print("{}".format(key), all_attrs, features_attrs, timestep_attrs)
+            if XAI_metric:
+                faithfulness_timesteps = model.Faithfulness_Correlation(
+                    test_loader, timestep_attrs, pertrub='Noise', time_step=True, subset_size=4, nr_runs=10)
+                print('Attributions faithfulness timesteps correlation', faithfulness_timesteps)
+                XAI_dict["{}_faith_timesteps".format(key)] = faithfulness_timesteps.tolist()
+                random_attributions = np.random.normal(np.shape(all_attrs))
+
+                faithfulness_timesteps_variable = model.Faithfulness_Correlation(
+                    test_loader, all_attrs, pertrub='Noise', feature_timestep=True, subset_size=[4, 9], nr_runs=10)
+                print('Attributions faithfulness variable per timestep correlation', faithfulness_timesteps_variable)
+                XAI_dict["{}_faith_variable_per_timestep".format(key)] = faithfulness_timesteps_variable.tolist()
 
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/Attributions.json"
@@ -326,22 +345,6 @@ def train_common(
         # Write the dictionary to a JSON file
         with open(json_file_path, 'w') as json_file:
             json.dump(attributions_dict, json_file)
-        if XAI_metric:
-            
-            F_attribution = model.Faithfulness_Correlation(test_loader, attributions_IG, pertrub='Noise')
-            print('Attributions faithfulness correlation', F_attribution)
-            
-            F_saliency = model.Faithfulness_Correlation(test_loader, attributions_Saliency, pertrub='Noise')
-            print('Saliency faithfulness correlation', F_saliency)
-            F_lime = model.Faithfulness_Correlation(test_loader, attributions_lime, pertrub='Noise')
-            print('shap faithfulness correlation', F_lime)
-            XAI_dict = {
-                "Faithfulness_correlation_normal_random": F_baseline,
-                "Faithfulness_correlation_IG": F_attribution,
-                "Faithfulness_correlation_attention": F_attention,
-                "Faithfulness_correlation_saliency": F_saliency,
-                "Faithfulness_correlation_shap": F_lime
-            }
 
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/XAI_metrics.json"
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 5898c8fa..576913a3 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -541,13 +541,15 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
                 # Convert attributions to numpy array and append to the list
             all_attrs.append(attr[0].cpu().detach().numpy())
         # Concatenate a‚ttribution values for all instances along the batch dimension
-
-        all_attrs = all_attrs.mean(axis=(0))
-
+        # print('b4', all_attrs)
+        # print('b4 shape', np.shape(all_attrs))
+        all_attrs = np.array(all_attrs).mean(axis=(0))
+        # print('after', all_attrs)
+        # print('after shape', np.shape(all_attrs))
         features_attrs = all_attrs.mean(axis=(0))
-
+        # print('feat shape ', np.shape(features_attrs))
         timestep_attrs = all_attrs.mean(axis=(1))
-
+        # print('timestep shape', np.shape(timestep_attrs))
         # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
             # Create x values (assuming you want a simple sequential x-axis)
@@ -590,7 +592,7 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
         return all_attrs, features_attrs, timestep_attrs
 
     def Faithfulness_Correlation(
-        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=3, feature=False, time_step=True, feature_timestep=False
+        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=3, feature=False, time_step=False, feature_timestep=False
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.
@@ -643,29 +645,38 @@ def Faithfulness_Correlation(
             y_pred = self(data).detach().cpu().numpy()
             pred_deltas = []
             att_sums = []
+
+            def add_noise(x, indices, time_step, feature_timestep):
+                noise = torch.randn_like(x["encoder_cont"])
+                if time_step:
+                    x["encoder_cont"][:, indices, :] += noise[:, indices, :]
+                elif feature_timestep:
+
+                    x["encoder_cont"][:, indices[0], :][:, :, indices[1]] += noise[:, indices[0], :][:, :, indices[1]]
+
+            def apply_baseline(x, indices, time_step, feature_timestep):
+                mask = torch.ones_like(x["encoder_cont"]).cpu()
+                if time_step:
+                    mask[:, indices, :] = 0
+                elif feature_timestep:
+                    mask[:, indices[0], :][:, :, indices[1]] = 0
+                mask = mask.to(x["encoder_cont"].device)
+                x["encoder_cont"] *= mask
             for i_ix in range(nr_runs):
                 # Randomly mask by subset size.
-                a_ix = np.random.choice(x["encoder_cont"].shape[1], subset_size, replace=False)
 
-                # Move a_ix_tensor to the same device as mask
+                if time_step:
+                    a_ix = np.random.choice(24, subset_size, replace=False)
+                elif feature_timestep:
+                    timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
+                    variables_idx = np.random.choice(53, subset_size[1], replace=False)
+                    a_ix = [timesteps_idx, variables_idx]
 
                 if pertrub == "Noise":
-                    # add normal noise to input
-
-                    noise = torch.randn_like(x["encoder_cont"])
-
-                    x["encoder_cont"][:, a_ix, :] += noise[:, a_ix, :]
-
+                    add_noise(x, a_ix, time_step, feature_timestep)
                 elif pertrub == "baseline":
-                    # Create a mask tensor with zeros at specified time steps and ones everywhere else
-                    # pytorch bug need to change to cpu for next step and then revert
-                    mask = torch.ones_like(x["encoder_cont"]).cpu()
-
-                    mask[:, a_ix, :] = 0
+                    apply_baseline(x, a_ix, time_step, feature_timestep)
 
-                    mask = mask.to(x["encoder_cont"].device)
-
-                    x["encoder_cont"] = x["encoder_cont"] * mask
                 data = (
                     x["encoder_cat"],
                     x["encoder_cont"],
@@ -681,13 +692,19 @@ def Faithfulness_Correlation(
                 )
                 # Predict on perturbed input x.
                 y_pred_perturb = self(data).detach().cpu().numpy()
-                pred_deltas.append((y_pred - y_pred_perturb).mean(axis=(0, 2)))
+                pred_deltas.append((y_pred - y_pred_perturb).mean())
 
                 # Sum attributions of the random subset.
+                if time_step:
+                    att_sums.append(np.sum(attribution[a_ix]))
+                elif feature_timestep:
+                    row_indices = a_ix[0]
+                    col_indices = a_ix[1]
+                    att_sums.append(np.sum(attribution[np.ix_(row_indices, col_indices)]))
 
-                att_sums.append(np.sum(attribution[a_ix]))
             similarities.append(similarity_func(pred_deltas, att_sums))
-            score = np.nanmean(similarities)
+
+        score = np.nanmean(similarities)
         return score
 
     def Data_Randomization(self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None, **kwargs):

From d3ea3d32485bf5060cec4da018df3c857861be49 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 21 Nov 2023 17:28:16 +0100
Subject: [PATCH 112/142] latest changes

---
 icu_benchmarks/cross_validation.py       |   5 +-
 icu_benchmarks/data/loader.py            |   6 +-
 icu_benchmarks/models/dl_models.py       |   4 +-
 icu_benchmarks/models/train.py           | 125 ++++++++++-------------
 icu_benchmarks/models/wrappers.py        |  76 +++++++-------
 icu_benchmarks/run.py                    |   8 +-
 icu_benchmarks/run_utils.py              |   3 +-
 icu_benchmarks/tuning/hyperparameters.py |   2 +-
 8 files changed, 102 insertions(+), 127 deletions(-)

diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 7d3f217e..535d9619 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -42,7 +42,7 @@ def execute_repeated_cv(
     pytorch_forecasting: bool = False,
     XAI_metric: bool = False,
     random_labels: bool = False,
-    random_model_dir: str = None
+    random_model_dir: str = None,
 ) -> float:
     """Preprocesses data and trains a model for each fold.
 
@@ -107,7 +107,6 @@ def execute_repeated_cv(
                 pretrained_imputation_model=pretrained_imputation_model,
                 runmode=mode,
                 complete_train=complete_train,
-
             )
 
             preprocess_time = datetime.now() - start_time
@@ -129,7 +128,7 @@ def execute_repeated_cv(
                 pytorch_forecasting=pytorch_forecasting,
                 XAI_metric=XAI_metric,
                 random_labels=random_labels,
-                random_model_dir=random_model_dir
+                random_model_dir=random_model_dir,
             )
 
             train_time = datetime.now() - start_time
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 68109f4b..75722831 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -522,9 +522,9 @@ def get_feature_names(self):
 
     def randomize_labels(self, num_classes=None, min=None, max=None):
         if num_classes == 1:
-
-            random_target = np.random.uniform(self.data["target"][0].min(),
-                                              self.data["target"][0].max(), size=len(self.data["target"][0]))
+            random_target = np.random.uniform(
+                self.data["target"][0].min(), self.data["target"][0].max(), size=len(self.data["target"][0])
+            )
         else:
             random_target = np.random.randint(num_classes, size=len(self.data["target"][0]))
         self.data["target"][0] = Tensor(random_target)
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 1cd782a5..ad73a479 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -637,7 +637,7 @@ def forward(
             "target_scale": tuple_x[10],
         }
         if self.num_classes == 1:
-            x_dict['encoder_cont'][:, :, -1] = 0.0
+            x_dict["encoder_cont"][:, :, -1] = 0.0
 
         out = self.model(x_dict)
 
@@ -698,7 +698,7 @@ def forward(
             "target_scale": tuple_x[10],
         }
         if self.num_classes == 1:
-            x_dict['encoder_cont'][:, :, -1] = 0.0
+            x_dict["encoder_cont"][:, :, -1] = 0.0
         out = self.model(x_dict)
         out = out["prediction"][0]
 
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index c0a055f5..18341a99 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -31,9 +31,7 @@
 from captum.attr import IntegratedGradients, ShapleyValueSampling, Saliency, GuidedBackprop, LRP, FeatureAblation, Lime
 
 
-cpu_core_count = (
-    len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
-)
+cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
 
 
 def assure_minimum_length(dataset):
@@ -76,7 +74,7 @@ def train_common(
     pytorch_forecasting: bool = False,
     XAI_metric: bool = False,
     random_labels: bool = False,
-    random_model_dir: str = None
+    random_model_dir: str = None,
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -110,25 +108,15 @@ def train_common(
     dataset_class = (
         ImputationDataset
         if mode == RunMode.imputation
-        else (
-            PredictionDatasetpytorch if (pytorch_forecasting) else PredictionDataset
-        )
+        else (PredictionDatasetpytorch if (pytorch_forecasting) else PredictionDataset)
     )
 
     logging.info(f"Logging to directory: {log_dir}.")
-    save_config_file(
-        log_dir
-    )  # We save the operative config before and also after training
+    save_config_file(log_dir)  # We save the operative config before and also after training
 
-    train_dataset = dataset_class(
-        data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"]
-    )
-    val_dataset = dataset_class(
-        data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"]
-    )
-    train_dataset, val_dataset = assure_minimum_length(
-        train_dataset
-    ), assure_minimum_length(val_dataset)
+    train_dataset = dataset_class(data, split=Split.train, ram_cache=ram_cache, name=dataset_names["train"])
+    val_dataset = dataset_class(data, split=Split.val, ram_cache=ram_cache, name=dataset_names["val"])
+    train_dataset, val_dataset = assure_minimum_length(train_dataset), assure_minimum_length(val_dataset)
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
     test_dataset = dataset_class(data, split=test_on, name=dataset_names["test"])
     test_dataset = assure_minimum_length(test_dataset)
@@ -219,9 +207,7 @@ def train_common(
         if load_weights:
             model = load_model(model, source_dir, pl_model=pl_model)
         else:
-            model = model(
-                optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode
-            )
+            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
 
     model.set_weight(weight, train_dataset)
     model.set_trained_columns(train_dataset.get_feature_names())
@@ -261,9 +247,7 @@ def train_common(
         if model.requires_backprop:
             logging.info("Training DL model.")
 
-            trainer.fit(
-                model, train_dataloaders=train_loader, val_dataloaders=val_loader
-            )
+            trainer.fit(model, train_dataloaders=train_loader, val_dataloaders=val_loader)
             logging.info("Training complete.")
         else:
             logging.info("Training ML model.")
@@ -277,50 +261,51 @@ def train_common(
 
     if explain:
         attributions_dict = {}
-        Interpertations = model.interpertations(test_loader, log_dir, plot=True)
-        attributions_dict["attention_weights"] = Interpertations["attention"].tolist()
-        attributions_dict["static_variables"] = Interpertations["static_variables"].tolist()
-        attributions_dict["encoder_variables"] = Interpertations["encoder_variables"].tolist()
-        print("attention", Interpertations)
-        model.train()
+        # Interpertations = model.interpertations(test_loader, log_dir, plot=True)
+        # attributions_dict["attention_weights"] = Interpertations["attention"].tolist()
+        # attributions_dict["static_variables"] = Interpertations["static_variables"].tolist()
+        # attributions_dict["encoder_variables"] = Interpertations["encoder_variables"].tolist()
+        # print("attention", Interpertations)
+        # model.train()
         XAI_dict = {}
-        if XAI_metric:
-            XAI_dict = {}
-            # random attribution per timestep
-            random_attributions_ts = np.random.normal(size=24)
-            F_baseline_ts = model.Faithfulness_Correlation(
-                test_loader, random_attributions_ts, pertrub='Noise', subset_size=4, time_step=True, nr_runs=10)
-            print('Random normal faithfulness correlation for timesteps', F_baseline_ts)
-
-            F_attention = model.Faithfulness_Correlation(
-                test_loader, Interpertations["attention"], pertrub='Noise', subset_size=4, time_step=True, nr_runs=10)
-            print('Attention faithfulness correlation', F_attention)
-            # random attribution per variable per timestep
-            random_attributions_v_ts = np.random.normal(size=[24, 53])
-            F_baseline_v_ts = model.Faithfulness_Correlation(
-                test_loader, random_attributions_v_ts, pertrub='Noise', feature_timestep=True, nr_runs=10, subset_size=[4, 9])
-            print('Random normal faithfulness correlation for variables per timesteps', F_baseline_v_ts)
-
-        XAI_dict["attention_faith"] = F_attention
-        XAI_dict["random_faith_timestep"] = random_attributions_ts
-        XAI_dict["random_faith_var_timestep"] = random_attributions_v_ts.tolist()
+        # if XAI_metric:
+
+        # random attribution per timestep
+        # random_attributions_ts = np.random.normal(size=24)
+        # F_baseline_ts = model.Faithfulness_Correlation(
+        #    test_loader, random_attributions_ts, pertrub='Noise', subset_size=4, time_step=True, nr_runs=100)
+        # print('Random normal faithfulness correlation for timesteps', F_baseline_ts)
+
+        # F_attention = model.Faithfulness_Correlation(
+        #    test_loader, Interpertations["attention"], pertrub='Noise', subset_size=4, time_step=True, nr_runs=100)
+        # print('Attention faithfulness correlation', F_attention)
+        # random attribution per variable per timestep
+        # random_attributions_v_ts = np.random.normal(size=[24, 53])
+        # F_baseline_v_ts = model.Faithfulness_Correlation(
+        #    test_loader, random_attributions_v_ts, pertrub='Noise', feature_timestep=True, nr_runs=10, subset_size=[4, 9])
+        # print('Random normal faithfulness correlation for variables per timesteps', F_baseline_v_ts)
+
+        # XAI_dict["attention_faith"] = F_attention.tolist()
+        # XAI_dict["random_faith_timestep"] = random_attributions_ts.tolist()
+        # XAI_dict["random_faith_var_timestep"] = random_attributions_v_ts.tolist()
         methods = {
-            "Saliency": Saliency,
+            # "Saliency": Saliency,
             #  "Lime": Lime,
-            "IG": IntegratedGradients,
-
-
+            # "IG": IntegratedGradients,
+            "FA": FeatureAblation
         }
         for key, item in methods.items():
             if key == "IG":
                 all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
-                    test_loader=test_loader,
-                    method=item, log_dir=log_dir, plot=True, n_steps=20
+                    test_loader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=50
+                )
+            if key == "Lime":
+                all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
+                    test_loader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True
                 )
             else:
                 all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
-                    test_loader=test_loader,
-                    method=item, log_dir=log_dir, plot=True
+                    test_loader=test_loader, method=item, log_dir=log_dir, plot=True
                 )
             attributions_dict["{}_all".format(key)] = all_attrs.tolist()
             attributions_dict["{}_timesteps".format(key)] = timestep_attrs.tolist()
@@ -329,28 +314,30 @@ def train_common(
             print("{}".format(key), all_attrs, features_attrs, timestep_attrs)
             if XAI_metric:
                 faithfulness_timesteps = model.Faithfulness_Correlation(
-                    test_loader, timestep_attrs, pertrub='Noise', time_step=True, subset_size=4, nr_runs=10)
-                print('Attributions faithfulness timesteps correlation', faithfulness_timesteps)
+                    test_loader, timestep_attrs, pertrub="Noise", time_step=True, subset_size=4, nr_runs=100
+                )
+                print("Attributions faithfulness timesteps correlation", faithfulness_timesteps)
                 XAI_dict["{}_faith_timesteps".format(key)] = faithfulness_timesteps.tolist()
                 random_attributions = np.random.normal(np.shape(all_attrs))
 
                 faithfulness_timesteps_variable = model.Faithfulness_Correlation(
-                    test_loader, all_attrs, pertrub='Noise', feature_timestep=True, subset_size=[4, 9], nr_runs=10)
-                print('Attributions faithfulness variable per timestep correlation', faithfulness_timesteps_variable)
+                    test_loader, all_attrs, pertrub="Noise", feature_timestep=True, subset_size=[4, 9], nr_runs=100
+                )
+                print("Attributions faithfulness variable per timestep correlation", faithfulness_timesteps_variable)
                 XAI_dict["{}_faith_variable_per_timestep".format(key)] = faithfulness_timesteps_variable.tolist()
 
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/Attributions.json"
 
         # Write the dictionary to a JSON file
-        with open(json_file_path, 'w') as json_file:
+        with open(json_file_path, "w") as json_file:
             json.dump(attributions_dict, json_file)
 
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/XAI_metrics.json"
 
         # Write the dictionary to a JSON file
-        with open(json_file_path, 'w') as json_file:
+        with open(json_file_path, "w") as json_file:
             json.dump(XAI_dict, json_file)
 
         # path = Path(random_model_dir)
@@ -370,9 +357,7 @@ def train_common(
         # print('Distance Data randmoization score attention', R_attention)
 
     model.set_weight("balanced", train_dataset)
-    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0][
-        "test/loss"
-    ]
+    test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
     return test_loss
 
@@ -390,9 +375,7 @@ def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=N
                 return Exception(f"No weights to load at path : {source_dir}")
             if pl_model:
                 if train_dataset is not None:
-                    model = model.load_from_checkpoint(
-                        model_path, dataset=train_dataset, optimizer=optimizer
-                    )
+                    model = model.load_from_checkpoint(model_path, dataset=train_dataset, optimizer=optimizer)
 
                 else:
                     model = model.load_from_checkpoint(model_path)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 576913a3..6385b677 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -21,6 +21,7 @@
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean
 import captum
+
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.mse_loss, module="torch.nn.functional")
@@ -156,12 +157,10 @@ def on_train_start(self):
         return super().on_train_start()
 
     def finalize_step(self, step_prefix=""):
-
         try:
             for name, metric in self.metrics[step_prefix].items():
                 try:
-                    value = np.float32(metric.compute()) if isinstance(
-                        metric.compute(), np.float64) else metric.compute()
+                    value = np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
                     self.log_dict({f"{step_prefix}/{name}": value}, sync_dist=True)
 
                 except (NotComputableError, ValueError) as e:
@@ -491,10 +490,9 @@ def step_fn(self, element, step_prefix=""):
     def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kwargs):
         # Initialize lists to store attribution values for all instances
         all_attrs = []
-
+        method_name = method.__name__
         # Loop through the test_loader to compute attributions for all instances
         for batch in test_loader:
-
             for key, value in batch[0].items():
                 batch[0][key] = batch[0][key].to(self.device)
             x = batch[0]
@@ -513,11 +511,11 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
                 x["target_scale"].requires_grad_(),
             )
             baselines = (
-                data[0].to(self.device),  # encoder_cat, set to random
+                data[0].to(self.device),  # encoder_cat, no cat variables
                 torch.randn_like(data[1]).to(self.device),  # encoder_cont, set to random
                 torch.randn_like(data[2]).to(self.device),  # encoder_target, set to random
                 data[3].to(self.device),  # encoder_lengths, leave unchanged
-                torch.randn_like(data[4]).to(self.device),  # decoder_cat, set to random
+                data[4].to(self.device),  # decoder_cat, no cat variables
                 torch.randn_like(data[5]).to(self.device),  # decoder_cont, set to random
                 torch.randn_like(data[6]).to(self.device),  # decoder_target, set to random
                 data[7].to(self.device),  # decoder_lengths, leave unchanged
@@ -531,29 +529,23 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             # attr_all_timesteps = []
             # for time_step in range(0, 24):
             if method is not captum.attr.Saliency:
-                attr = explantation.attribute(
-                    data,  baselines=baselines, **kwargs
-                )
+                attr = explantation.attribute(data, baselines=baselines, **kwargs)
             else:
-                attr = explantation.attribute(
-                    data,  **kwargs
-                )
-                # Convert attributions to numpy array and append to the list
+                attr = explantation.attribute(data, **kwargs)
+            print(attr)
+
+            print(attr[0].shape)
+            # Convert attributions to numpy array and append to the list
             all_attrs.append(attr[0].cpu().detach().numpy())
-        # Concatenate a‚ttribution values for all instances along the batch dimension
-        # print('b4', all_attrs)
-        # print('b4 shape', np.shape(all_attrs))
+
         all_attrs = np.array(all_attrs).mean(axis=(0))
-        # print('after', all_attrs)
-        # print('after shape', np.shape(all_attrs))
+
         features_attrs = all_attrs.mean(axis=(0))
-        # print('feat shape ', np.shape(features_attrs))
+
         timestep_attrs = all_attrs.mean(axis=(1))
-        # print('timestep shape', np.shape(timestep_attrs))
+
         # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
-            # Create x values (assuming you want a simple sequential x-axis)
-            # Assuming you have 57 variables
             x_values = np.arange(1, 54)
             # Plotting the featrue means
             plt.figure(figsize=(8, 6))
@@ -567,11 +559,11 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
                 markersize=8,
             )
             plt.xlabel("Time Step")
-            plt.ylabel(" Attribution")
-            plt.title("Attribution Values")
+            plt.ylabel("{} Attribution".format(method_name))
+            plt.title("{} Attribution Values".format(method_name))
             plt.xticks(x_values)  # Set x-ticks to match the number of features
             plt.tight_layout()
-            plt.savefig(log_dir / "attribution_features_plot.png", bbox_inches="tight")
+            plt.savefig(log_dir / "{}_attribution_features_plot.png".format(method_name), bbox_inches="tight")
             plt.figure(figsize=(8, 6))
             x_values = np.arange(1, 25)
             plt.plot(
@@ -584,15 +576,24 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
                 markersize=8,
             )
             plt.xlabel("Time Step")
-            plt.ylabel("Normalized Attribution")
-            plt.title("Attribution Values")
+            plt.ylabel("{} Attribution ".format(method_name))
+            plt.title("{} Attribution Values".format(method_name))
             plt.xticks(x_values)  # Set x-ticks to match the number of features
             plt.tight_layout()
-            plt.savefig(log_dir / "attribution_plot.png", bbox_inches="tight")
+            plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
         return all_attrs, features_attrs, timestep_attrs
 
     def Faithfulness_Correlation(
-        self, test_loader, attribution, similarity_func=None, nr_runs=100, pertrub=None, subset_size=3, feature=False, time_step=False, feature_timestep=False
+        self,
+        test_loader,
+        attribution,
+        similarity_func=None,
+        nr_runs=100,
+        pertrub=None,
+        subset_size=3,
+        feature=False,
+        time_step=False,
+        feature_timestep=False,
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.
@@ -651,7 +652,6 @@ def add_noise(x, indices, time_step, feature_timestep):
                 if time_step:
                     x["encoder_cont"][:, indices, :] += noise[:, indices, :]
                 elif feature_timestep:
-
                     x["encoder_cont"][:, indices[0], :][:, :, indices[1]] += noise[:, indices[0], :][:, :, indices[1]]
 
             def apply_baseline(x, indices, time_step, feature_timestep):
@@ -662,6 +662,7 @@ def apply_baseline(x, indices, time_step, feature_timestep):
                     mask[:, indices[0], :][:, :, indices[1]] = 0
                 mask = mask.to(x["encoder_cont"].device)
                 x["encoder_cont"] *= mask
+
             for i_ix in range(nr_runs):
                 # Randomly mask by subset size.
 
@@ -707,7 +708,9 @@ def apply_baseline(x, indices, time_step, feature_timestep):
         score = np.nanmean(similarities)
         return score
 
-    def Data_Randomization(self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None, **kwargs):
+    def Data_Randomization(
+        self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None, **kwargs
+    ):
         """
         Implementation of the Random Logit Metric by Sixt et al., 2020.
 
@@ -725,7 +728,6 @@ def Data_Randomization(self, test_loader, attribution, explain_method, random_mo
         if similarity_func is None:
             similarity_func = distance_euclidean
         if explain_method == "Attention":
-
             Attention_weights = random_model.interpertations(test_loader)
 
             score = similarity_func(Attention_weights["attention"].cpu().numpy(), attribution.cpu().numpy())
@@ -766,13 +768,9 @@ def Data_Randomization(self, test_loader, attribution, explain_method, random_mo
                 explantation = explain_method(random_model.forward_captum)
                 # Reformat attributions.
                 if explain_method is not captum.attr.Saliency:
-                    attr = explantation.attribute(
-                        data,  baselines=baselines, **kwargs
-                    )
+                    attr = explantation.attribute(data, baselines=baselines, **kwargs)
                 else:
-                    attr = explantation.attribute(
-                        data,  **kwargs
-                    )
+                    attr = explantation.attribute(data, **kwargs)
                     # Convert attributions to numpy array and append to the list
                 a_perturbed.append(attr[0].cpu().detach().numpy())
 
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 182a1814..1faadab6 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -117,8 +117,7 @@ def main(my_args=tuple(sys.argv[1:])):
             name_datasets(args.name, args.name, args.name)
         run_dir = create_run_dir(log_dir)
         source_dir = args.source_dir
-        logging.info(
-            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
+        logging.info(f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
         gin.parse_config_file(source_dir / "train_config.gin")
     elif args.samples and args.source_dir is not None:  # Train model with limited samples and bind existing config
         logging.info("Binding train gin-config. Note: this might override your config.")
@@ -131,8 +130,7 @@ def main(my_args=tuple(sys.argv[1:])):
         name_datasets(args.name, args.name, args.name)
         hp_checkpoint = log_dir / args.hp_checkpoint if args.hp_checkpoint else None
         model_path = (
-            Path("configs") / ("imputation_models" if mode ==
-                               RunMode.imputation else "prediction_models") / f"{model}.gin"
+            Path("configs") / ("imputation_models" if mode == RunMode.imputation else "prediction_models") / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
@@ -155,8 +153,6 @@ def main(my_args=tuple(sys.argv[1:])):
             verbose=verbose,
             pytorch_forecasting=pytorch_forecasting,
             random_labels=random_labels,
-
-
         )
 
     log_full_line(f"Logging to {run_dir.resolve()}", level=logging.INFO)
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 0f559e4c..944295ea 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -305,8 +305,7 @@ def load_pretrained_imputation_model(use_pretrained_imputation):
         pretrained_imputation_model_checkpoint = torch.load(use_pretrained_imputation, map_location=torch.device("cpu"))
         if isinstance(pretrained_imputation_model_checkpoint, dict):
             imputation_model_class = pretrained_imputation_model_checkpoint["class"]
-            pretrained_imputation_model = imputation_model_class(
-                **pretrained_imputation_model_checkpoint["hyper_parameters"])
+            pretrained_imputation_model = imputation_model_class(**pretrained_imputation_model_checkpoint["hyper_parameters"])
             pretrained_imputation_model.set_trained_columns(pretrained_imputation_model_checkpoint["trained_columns"])
             pretrained_imputation_model.load_state_dict(pretrained_imputation_model_checkpoint["state_dict"])
         else:
diff --git a/icu_benchmarks/tuning/hyperparameters.py b/icu_benchmarks/tuning/hyperparameters.py
index 7c02735e..1458a7c2 100644
--- a/icu_benchmarks/tuning/hyperparameters.py
+++ b/icu_benchmarks/tuning/hyperparameters.py
@@ -118,7 +118,7 @@ def bind_params_and_train(hyperparams):
                 verbose=verbose,
                 wandb=wandb,
                 pytorch_forecasting=pytorch_forecasting,
-                random_labels=random_labels
+                random_labels=random_labels,
             )
 
     header = ["ITERATION"] + hyperparams_names + ["LOSS AT ITERATION"]

From 54a6f18e97a54a4078554faa67614607dffcd66d Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Fri, 24 Nov 2023 14:41:20 +0100
Subject: [PATCH 113/142] latest

---
 configs/prediction_models/DeepARpytorch.gin |  12 +-
 configs/prediction_models/RNNpytorch.gin    |   9 +-
 configs/tasks/Regression.gin                |   4 +-
 icu_benchmarks/data/loader.py               |   9 +-
 icu_benchmarks/data/preprocessor.py         |  13 +-
 icu_benchmarks/models/dl_models.py          |  15 +-
 icu_benchmarks/models/train.py              |  70 ++--
 icu_benchmarks/models/wrappers.py           | 391 +++++++++++++-------
 8 files changed, 329 insertions(+), 194 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index bb18cbd0..1b7a9ca0 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -23,15 +23,17 @@ model/hyperparameter.lr_scheduler = "exponential"
 
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
-PredictionDatasetpytorch.max_prediction_length = 24
-PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
+PredictionDatasetpytorch.max_prediction_length = 1
+PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_1', 'alp_lag_1', 'alt_lag_1', 'ast_lag_1', 'be_lag_1', 'bicar_lag_1', 'bili_lag_1', 'bili_dir_lag_1', 'bnd_lag_1', 'bun_lag_1', 'ca_lag_1', 'cai_lag_1', 'ck_lag_1', 'ckmb_lag_1', 'cl_lag_1', 'crea_lag_1', 'crp_lag_1', 'dbp_lag_1', 'fgn_lag_1', 'fio2_lag_1', 'glu_lag_1', 'hgb_lag_1', 'hr_lag_1', 'inr_pt_lag_1', 'k_lag_1', 'lact_lag_1', 'lymph_lag_1', 'map_lag_1', 'mch_lag_1', 'mchc_lag_1', 'mcv_lag_1', 'methb_lag_1', 'mg_lag_1', 'na_lag_1', 'neut_lag_1', 'o2sat_lag_1', 'pco2_lag_1', 'ph_lag_1', 'phos_lag_1', 'plt_lag_1', 'po2_lag_1', 'ptt_lag_1', 'resp_lag_1', 'sbp_lag_1', 'temp_lag_1', 'tnt_lag_1', 'urine_lag_1', 'wbc_lag_1', 'alb_lag_2', 'alp_lag_2', 'alt_lag_2', 'ast_lag_2', 'be_lag_2', 'bicar_lag_2', 'bili_lag_2', 'bili_dir_lag_2', 'bnd_lag_2', 'bun_lag_2', 'ca_lag_2', 'cai_lag_2', 'ck_lag_2', 'ckmb_lag_2', 'cl_lag_2', 'crea_lag_2', 'crp_lag_2', 'dbp_lag_2', 'fgn_lag_2', 'fio2_lag_2', 'glu_lag_2', 'hgb_lag_2', 'hr_lag_2', 'inr_pt_lag_2', 'k_lag_2', 'lact_lag_2', 'lymph_lag_2', 'map_lag_2', 'mch_lag_2', 'mchc_lag_2', 'mcv_lag_2', 'methb_lag_2', 'mg_lag_2', 'na_lag_2', 'neut_lag_2', 'o2sat_lag_2', 'pco2_lag_2', 'ph_lag_2', 'phos_lag_2', 'plt_lag_2', 'po2_lag_2', 'ptt_lag_2', 'resp_lag_2', 'sbp_lag_2', 'temp_lag_2', 'tnt_lag_2', 'urine_lag_2', 'wbc_lag_2', 'alb_lag_3', 'alp_lag_3', 'alt_lag_3', 'ast_lag_3', 'be_lag_3', 'bicar_lag_3', 'bili_lag_3', 'bili_dir_lag_3', 'bnd_lag_3', 'bun_lag_3', 'ca_lag_3', 'cai_lag_3', 'ck_lag_3', 'ckmb_lag_3', 'cl_lag_3', 'crea_lag_3', 'crp_lag_3', 'dbp_lag_3', 'fgn_lag_3', 'fio2_lag_3', 'glu_lag_3', 'hgb_lag_3', 'hr_lag_3', 'inr_pt_lag_3', 'k_lag_3', 'lact_lag_3', 'lymph_lag_3', 'map_lag_3', 'mch_lag_3', 'mchc_lag_3', 'mcv_lag_3', 'methb_lag_3', 'mg_lag_3', 'na_lag_3', 'neut_lag_3', 'o2sat_lag_3', 'pco2_lag_3', 'ph_lag_3', 'phos_lag_3', 'plt_lag_3', 'po2_lag_3', 'ptt_lag_3', 'resp_lag_3', 'sbp_lag_3', 'temp_lag_3', 'tnt_lag_3', 'urine_lag_3', 'wbc_lag_3', 'alb_lag_4', 'alp_lag_4', 'alt_lag_4', 'ast_lag_4', 'be_lag_4', 'bicar_lag_4', 'bili_lag_4', 'bili_dir_lag_4', 'bnd_lag_4', 'bun_lag_4', 'ca_lag_4', 'cai_lag_4', 'ck_lag_4', 'ckmb_lag_4', 'cl_lag_4', 'crea_lag_4', 'crp_lag_4', 'dbp_lag_4', 'fgn_lag_4', 'fio2_lag_4', 'glu_lag_4', 'hgb_lag_4', 'hr_lag_4', 'inr_pt_lag_4', 'k_lag_4', 'lact_lag_4', 'lymph_lag_4', 'map_lag_4', 'mch_lag_4', 'mchc_lag_4', 'mcv_lag_4', 'methb_lag_4', 'mg_lag_4', 'na_lag_4', 'neut_lag_4', 'o2sat_lag_4', 'pco2_lag_4', 'ph_lag_4', 'phos_lag_4', 'plt_lag_4', 'po2_lag_4', 'ptt_lag_4', 'resp_lag_4', 'sbp_lag_4', 'temp_lag_4', 'tnt_lag_4', 'urine_lag_4', 'wbc_lag_4', 'alb_lag_5', 'alp_lag_5', 'alt_lag_5', 'ast_lag_5', 'be_lag_5', 'bicar_lag_5', 'bili_lag_5', 'bili_dir_lag_5', 'bnd_lag_5', 'bun_lag_5', 'ca_lag_5', 'cai_lag_5', 'ck_lag_5', 'ckmb_lag_5', 'cl_lag_5', 'crea_lag_5', 'crp_lag_5', 'dbp_lag_5', 'fgn_lag_5', 'fio2_lag_5', 'glu_lag_5', 'hgb_lag_5', 'hr_lag_5', 'inr_pt_lag_5', 'k_lag_5', 'lact_lag_5', 'lymph_lag_5', 'map_lag_5', 'mch_lag_5', 'mchc_lag_5', 'mcv_lag_5', 'methb_lag_5', 'mg_lag_5', 'na_lag_5', 'neut_lag_5', 'o2sat_lag_5', 'pco2_lag_5', 'ph_lag_5', 'phos_lag_5', 'plt_lag_5', 'po2_lag_5', 'ptt_lag_5', 'resp_lag_5', 'sbp_lag_5', 'temp_lag_5', 'tnt_lag_5', 'urine_lag_5', 'wbc_lag_5', 'alb_lag_6', 'alp_lag_6', 'alt_lag_6', 'ast_lag_6', 'be_lag_6', 'bicar_lag_6', 'bili_lag_6', 'bili_dir_lag_6', 'bnd_lag_6', 'bun_lag_6', 'ca_lag_6', 'cai_lag_6', 'ck_lag_6', 'ckmb_lag_6', 'cl_lag_6', 'crea_lag_6', 'crp_lag_6', 'dbp_lag_6', 'fgn_lag_6', 'fio2_lag_6', 'glu_lag_6', 'hgb_lag_6', 'hr_lag_6', 'inr_pt_lag_6', 'k_lag_6', 'lact_lag_6', 'lymph_lag_6', 'map_lag_6', 'mch_lag_6', 'mchc_lag_6', 'mcv_lag_6', 'methb_lag_6', 'mg_lag_6', 'na_lag_6', 'neut_lag_6', 'o2sat_lag_6', 'pco2_lag_6', 'ph_lag_6', 'phos_lag_6', 'plt_lag_6', 'po2_lag_6', 'ptt_lag_6', 'resp_lag_6', 'sbp_lag_6', 'temp_lag_6', 'tnt_lag_6', 'urine_lag_6', 'wbc_lag_6', 'alb_lag_7', 'alp_lag_7', 'alt_lag_7', 'ast_lag_7', 'be_lag_7', 'bicar_lag_7', 'bili_lag_7', 'bili_dir_lag_7', 'bnd_lag_7', 'bun_lag_7', 'ca_lag_7', 'cai_lag_7', 'ck_lag_7', 'ckmb_lag_7', 'cl_lag_7', 'crea_lag_7', 'crp_lag_7', 'dbp_lag_7', 'fgn_lag_7', 'fio2_lag_7', 'glu_lag_7', 'hgb_lag_7', 'hr_lag_7', 'inr_pt_lag_7', 'k_lag_7', 'lact_lag_7', 'lymph_lag_7', 'map_lag_7', 'mch_lag_7', 'mchc_lag_7', 'mcv_lag_7', 'methb_lag_7', 'mg_lag_7', 'na_lag_7', 'neut_lag_7', 'o2sat_lag_7', 'pco2_lag_7', 'ph_lag_7', 'phos_lag_7', 'plt_lag_7', 'po2_lag_7', 'ptt_lag_7', 'resp_lag_7', 'sbp_lag_7', 'temp_lag_7', 'tnt_lag_7', 'urine_lag_7', 'wbc_lag_7', 'alb_lag_8', 'alp_lag_8', 'alt_lag_8', 'ast_lag_8', 'be_lag_8', 'bicar_lag_8', 'bili_lag_8', 'bili_dir_lag_8', 'bnd_lag_8', 'bun_lag_8', 'ca_lag_8', 'cai_lag_8', 'ck_lag_8', 'ckmb_lag_8', 'cl_lag_8', 'crea_lag_8', 'crp_lag_8', 'dbp_lag_8', 'fgn_lag_8', 'fio2_lag_8', 'glu_lag_8', 'hgb_lag_8', 'hr_lag_8', 'inr_pt_lag_8', 'k_lag_8', 'lact_lag_8', 'lymph_lag_8', 'map_lag_8', 'mch_lag_8', 'mchc_lag_8', 'mcv_lag_8', 'methb_lag_8', 'mg_lag_8', 'na_lag_8', 'neut_lag_8', 'o2sat_lag_8', 'pco2_lag_8', 'ph_lag_8', 'phos_lag_8', 'plt_lag_8', 'po2_lag_8', 'ptt_lag_8', 'resp_lag_8', 'sbp_lag_8', 'temp_lag_8', 'tnt_lag_8', 'urine_lag_8', 'wbc_lag_8', 'alb_lag_9', 'alp_lag_9', 'alt_lag_9', 'ast_lag_9', 'be_lag_9', 'bicar_lag_9', 'bili_lag_9', 'bili_dir_lag_9', 'bnd_lag_9', 'bun_lag_9', 'ca_lag_9', 'cai_lag_9', 'ck_lag_9', 'ckmb_lag_9', 'cl_lag_9', 'crea_lag_9', 'crp_lag_9', 'dbp_lag_9', 'fgn_lag_9', 'fio2_lag_9', 'glu_lag_9', 'hgb_lag_9', 'hr_lag_9', 'inr_pt_lag_9', 'k_lag_9', 'lact_lag_9', 'lymph_lag_9', 'map_lag_9', 'mch_lag_9', 'mchc_lag_9', 'mcv_lag_9', 'methb_lag_9', 'mg_lag_9', 'na_lag_9', 'neut_lag_9', 'o2sat_lag_9', 'pco2_lag_9', 'ph_lag_9', 'phos_lag_9', 'plt_lag_9', 'po2_lag_9', 'ptt_lag_9', 'resp_lag_9', 'sbp_lag_9', 'temp_lag_9', 'tnt_lag_9', 'urine_lag_9', 'wbc_lag_9', 'alb_lag_10', 'alp_lag_10', 'alt_lag_10', 'ast_lag_10', 'be_lag_10', 'bicar_lag_10', 'bili_lag_10', 'bili_dir_lag_10', 'bnd_lag_10', 'bun_lag_10', 'ca_lag_10', 'cai_lag_10', 'ck_lag_10', 'ckmb_lag_10', 'cl_lag_10', 'crea_lag_10', 'crp_lag_10', 'dbp_lag_10', 'fgn_lag_10', 'fio2_lag_10', 'glu_lag_10', 'hgb_lag_10', 'hr_lag_10', 'inr_pt_lag_10', 'k_lag_10', 'lact_lag_10', 'lymph_lag_10', 'map_lag_10', 'mch_lag_10', 'mchc_lag_10', 'mcv_lag_10', 'methb_lag_10', 'mg_lag_10', 'na_lag_10', 'neut_lag_10', 'o2sat_lag_10', 'pco2_lag_10', 'ph_lag_10', 'phos_lag_10', 'plt_lag_10', 'po2_lag_10', 'ptt_lag_10', 'resp_lag_10', 'sbp_lag_10', 'temp_lag_10', 'tnt_lag_10', 'urine_lag_10', 'wbc_lag_10', 'alb_lag_11', 'alp_lag_11', 'alt_lag_11', 'ast_lag_11', 'be_lag_11', 'bicar_lag_11', 'bili_lag_11', 'bili_dir_lag_11', 'bnd_lag_11', 'bun_lag_11', 'ca_lag_11', 'cai_lag_11', 'ck_lag_11', 'ckmb_lag_11', 'cl_lag_11', 'crea_lag_11', 'crp_lag_11', 'dbp_lag_11', 'fgn_lag_11', 'fio2_lag_11', 'glu_lag_11', 'hgb_lag_11', 'hr_lag_11', 'inr_pt_lag_11', 'k_lag_11', 'lact_lag_11', 'lymph_lag_11', 'map_lag_11', 'mch_lag_11', 'mchc_lag_11', 'mcv_lag_11', 'methb_lag_11', 'mg_lag_11', 'na_lag_11', 'neut_lag_11', 'o2sat_lag_11', 'pco2_lag_11', 'ph_lag_11', 'phos_lag_11', 'plt_lag_11', 'po2_lag_11', 'ptt_lag_11', 'resp_lag_11', 'sbp_lag_11', 'temp_lag_11', 'tnt_lag_11', 'urine_lag_11', 'wbc_lag_11', 'alb_lag_12', 'alp_lag_12', 'alt_lag_12', 'ast_lag_12', 'be_lag_12', 'bicar_lag_12', 'bili_lag_12', 'bili_dir_lag_12', 'bnd_lag_12', 'bun_lag_12', 'ca_lag_12', 'cai_lag_12', 'ck_lag_12', 'ckmb_lag_12', 'cl_lag_12', 'crea_lag_12', 'crp_lag_12', 'dbp_lag_12', 'fgn_lag_12', 'fio2_lag_12', 'glu_lag_12', 'hgb_lag_12', 'hr_lag_12', 'inr_pt_lag_12', 'k_lag_12', 'lact_lag_12', 'lymph_lag_12', 'map_lag_12', 'mch_lag_12', 'mchc_lag_12', 'mcv_lag_12', 'methb_lag_12', 'mg_lag_12', 'na_lag_12', 'neut_lag_12', 'o2sat_lag_12', 'pco2_lag_12', 'ph_lag_12', 'phos_lag_12', 'plt_lag_12', 'po2_lag_12', 'ptt_lag_12', 'resp_lag_12', 'sbp_lag_12', 'temp_lag_12', 'tnt_lag_12', 'urine_lag_12', 'wbc_lag_12', 'alb_lag_13', 'alp_lag_13', 'alt_lag_13', 'ast_lag_13', 'be_lag_13', 'bicar_lag_13', 'bili_lag_13', 'bili_dir_lag_13', 'bnd_lag_13', 'bun_lag_13', 'ca_lag_13', 'cai_lag_13', 'ck_lag_13', 'ckmb_lag_13', 'cl_lag_13', 'crea_lag_13', 'crp_lag_13', 'dbp_lag_13', 'fgn_lag_13', 'fio2_lag_13', 'glu_lag_13', 'hgb_lag_13', 'hr_lag_13', 'inr_pt_lag_13', 'k_lag_13', 'lact_lag_13', 'lymph_lag_13', 'map_lag_13', 'mch_lag_13', 'mchc_lag_13', 'mcv_lag_13', 'methb_lag_13', 'mg_lag_13', 'na_lag_13', 'neut_lag_13', 'o2sat_lag_13', 'pco2_lag_13', 'ph_lag_13', 'phos_lag_13', 'plt_lag_13', 'po2_lag_13', 'ptt_lag_13', 'resp_lag_13', 'sbp_lag_13', 'temp_lag_13', 'tnt_lag_13', 'urine_lag_13', 'wbc_lag_13', 'alb_lag_14', 'alp_lag_14', 'alt_lag_14', 'ast_lag_14', 'be_lag_14', 'bicar_lag_14', 'bili_lag_14', 'bili_dir_lag_14', 'bnd_lag_14', 'bun_lag_14', 'ca_lag_14', 'cai_lag_14', 'ck_lag_14', 'ckmb_lag_14', 'cl_lag_14', 'crea_lag_14', 'crp_lag_14', 'dbp_lag_14', 'fgn_lag_14', 'fio2_lag_14', 'glu_lag_14', 'hgb_lag_14', 'hr_lag_14', 'inr_pt_lag_14', 'k_lag_14', 'lact_lag_14', 'lymph_lag_14', 'map_lag_14', 'mch_lag_14', 'mchc_lag_14', 'mcv_lag_14', 'methb_lag_14', 'mg_lag_14', 'na_lag_14', 'neut_lag_14', 'o2sat_lag_14', 'pco2_lag_14', 'ph_lag_14', 'phos_lag_14', 'plt_lag_14', 'po2_lag_14', 'ptt_lag_14', 'resp_lag_14', 'sbp_lag_14', 'temp_lag_14', 'tnt_lag_14', 'urine_lag_14', 'wbc_lag_14', 'alb_lag_15', 'alp_lag_15', 'alt_lag_15', 'ast_lag_15', 'be_lag_15', 'bicar_lag_15', 'bili_lag_15', 'bili_dir_lag_15', 'bnd_lag_15', 'bun_lag_15', 'ca_lag_15', 'cai_lag_15', 'ck_lag_15', 'ckmb_lag_15', 'cl_lag_15', 'crea_lag_15', 'crp_lag_15', 'dbp_lag_15', 'fgn_lag_15', 'fio2_lag_15', 'glu_lag_15', 'hgb_lag_15', 'hr_lag_15', 'inr_pt_lag_15', 'k_lag_15', 'lact_lag_15', 'lymph_lag_15', 'map_lag_15', 'mch_lag_15', 'mchc_lag_15', 'mcv_lag_15', 'methb_lag_15', 'mg_lag_15', 'na_lag_15', 'neut_lag_15', 'o2sat_lag_15', 'pco2_lag_15', 'ph_lag_15', 'phos_lag_15', 'plt_lag_15', 'po2_lag_15', 'ptt_lag_15', 'resp_lag_15', 'sbp_lag_15', 'temp_lag_15', 'tnt_lag_15', 'urine_lag_15', 'wbc_lag_15', 'alb_lag_16', 'alp_lag_16', 'alt_lag_16', 'ast_lag_16', 'be_lag_16', 'bicar_lag_16', 'bili_lag_16', 'bili_dir_lag_16', 'bnd_lag_16', 'bun_lag_16', 'ca_lag_16', 'cai_lag_16', 'ck_lag_16', 'ckmb_lag_16', 'cl_lag_16', 'crea_lag_16', 'crp_lag_16', 'dbp_lag_16', 'fgn_lag_16', 'fio2_lag_16', 'glu_lag_16', 'hgb_lag_16', 'hr_lag_16', 'inr_pt_lag_16', 'k_lag_16', 'lact_lag_16', 'lymph_lag_16', 'map_lag_16', 'mch_lag_16', 'mchc_lag_16', 'mcv_lag_16', 'methb_lag_16', 'mg_lag_16', 'na_lag_16', 'neut_lag_16', 'o2sat_lag_16', 'pco2_lag_16', 'ph_lag_16', 'phos_lag_16', 'plt_lag_16', 'po2_lag_16', 'ptt_lag_16', 'resp_lag_16', 'sbp_lag_16', 'temp_lag_16', 'tnt_lag_16', 'urine_lag_16', 'wbc_lag_16', 'alb_lag_17', 'alp_lag_17', 'alt_lag_17', 'ast_lag_17', 'be_lag_17', 'bicar_lag_17', 'bili_lag_17', 'bili_dir_lag_17', 'bnd_lag_17', 'bun_lag_17', 'ca_lag_17', 'cai_lag_17', 'ck_lag_17', 'ckmb_lag_17', 'cl_lag_17', 'crea_lag_17', 'crp_lag_17', 'dbp_lag_17', 'fgn_lag_17', 'fio2_lag_17', 'glu_lag_17', 'hgb_lag_17', 'hr_lag_17', 'inr_pt_lag_17', 'k_lag_17', 'lact_lag_17', 'lymph_lag_17', 'map_lag_17', 'mch_lag_17', 'mchc_lag_17', 'mcv_lag_17', 'methb_lag_17', 'mg_lag_17', 'na_lag_17', 'neut_lag_17', 'o2sat_lag_17', 'pco2_lag_17', 'ph_lag_17', 'phos_lag_17', 'plt_lag_17', 'po2_lag_17', 'ptt_lag_17', 'resp_lag_17', 'sbp_lag_17', 'temp_lag_17', 'tnt_lag_17', 'urine_lag_17', 'wbc_lag_17', 'alb_lag_18', 'alp_lag_18', 'alt_lag_18', 'ast_lag_18', 'be_lag_18', 'bicar_lag_18', 'bili_lag_18', 'bili_dir_lag_18', 'bnd_lag_18', 'bun_lag_18', 'ca_lag_18', 'cai_lag_18', 'ck_lag_18', 'ckmb_lag_18', 'cl_lag_18', 'crea_lag_18', 'crp_lag_18', 'dbp_lag_18', 'fgn_lag_18', 'fio2_lag_18', 'glu_lag_18', 'hgb_lag_18', 'hr_lag_18', 'inr_pt_lag_18', 'k_lag_18', 'lact_lag_18', 'lymph_lag_18', 'map_lag_18', 'mch_lag_18', 'mchc_lag_18', 'mcv_lag_18', 'methb_lag_18', 'mg_lag_18', 'na_lag_18', 'neut_lag_18', 'o2sat_lag_18', 'pco2_lag_18', 'ph_lag_18', 'phos_lag_18', 'plt_lag_18', 'po2_lag_18', 'ptt_lag_18', 'resp_lag_18', 'sbp_lag_18', 'temp_lag_18', 'tnt_lag_18', 'urine_lag_18', 'wbc_lag_18', 'alb_lag_19', 'alp_lag_19', 'alt_lag_19', 'ast_lag_19', 'be_lag_19', 'bicar_lag_19', 'bili_lag_19', 'bili_dir_lag_19', 'bnd_lag_19', 'bun_lag_19', 'ca_lag_19', 'cai_lag_19', 'ck_lag_19', 'ckmb_lag_19', 'cl_lag_19', 'crea_lag_19', 'crp_lag_19', 'dbp_lag_19', 'fgn_lag_19', 'fio2_lag_19', 'glu_lag_19', 'hgb_lag_19', 'hr_lag_19', 'inr_pt_lag_19', 'k_lag_19', 'lact_lag_19', 'lymph_lag_19', 'map_lag_19', 'mch_lag_19', 'mchc_lag_19', 'mcv_lag_19', 'methb_lag_19', 'mg_lag_19', 'na_lag_19', 'neut_lag_19', 'o2sat_lag_19', 'pco2_lag_19', 'ph_lag_19', 'phos_lag_19', 'plt_lag_19', 'po2_lag_19', 'ptt_lag_19', 'resp_lag_19', 'sbp_lag_19', 'temp_lag_19', 'tnt_lag_19', 'urine_lag_19', 'wbc_lag_19', 'alb_lag_20', 'alp_lag_20', 'alt_lag_20', 'ast_lag_20', 'be_lag_20', 'bicar_lag_20', 'bili_lag_20', 'bili_dir_lag_20', 'bnd_lag_20', 'bun_lag_20', 'ca_lag_20', 'cai_lag_20', 'ck_lag_20', 'ckmb_lag_20', 'cl_lag_20', 'crea_lag_20', 'crp_lag_20', 'dbp_lag_20', 'fgn_lag_20', 'fio2_lag_20', 'glu_lag_20', 'hgb_lag_20', 'hr_lag_20', 'inr_pt_lag_20', 'k_lag_20', 'lact_lag_20', 'lymph_lag_20', 'map_lag_20', 'mch_lag_20', 'mchc_lag_20', 'mcv_lag_20', 'methb_lag_20', 'mg_lag_20', 'na_lag_20', 'neut_lag_20', 'o2sat_lag_20', 'pco2_lag_20', 'ph_lag_20', 'phos_lag_20', 'plt_lag_20', 'po2_lag_20', 'ptt_lag_20', 'resp_lag_20', 'sbp_lag_20', 'temp_lag_20', 'tnt_lag_20', 'urine_lag_20', 'wbc_lag_20', 'alb_lag_21', 'alp_lag_21', 'alt_lag_21', 'ast_lag_21', 'be_lag_21', 'bicar_lag_21', 'bili_lag_21', 'bili_dir_lag_21', 'bnd_lag_21', 'bun_lag_21', 'ca_lag_21', 'cai_lag_21', 'ck_lag_21', 'ckmb_lag_21', 'cl_lag_21', 'crea_lag_21', 'crp_lag_21', 'dbp_lag_21', 'fgn_lag_21', 'fio2_lag_21', 'glu_lag_21', 'hgb_lag_21', 'hr_lag_21', 'inr_pt_lag_21', 'k_lag_21', 'lact_lag_21', 'lymph_lag_21', 'map_lag_21', 'mch_lag_21', 'mchc_lag_21', 'mcv_lag_21', 'methb_lag_21', 'mg_lag_21', 'na_lag_21', 'neut_lag_21', 'o2sat_lag_21', 'pco2_lag_21', 'ph_lag_21', 'phos_lag_21', 'plt_lag_21', 'po2_lag_21', 'ptt_lag_21', 'resp_lag_21', 'sbp_lag_21', 'temp_lag_21', 'tnt_lag_21', 'urine_lag_21', 'wbc_lag_21', 'alb_lag_22', 'alp_lag_22', 'alt_lag_22', 'ast_lag_22', 'be_lag_22', 'bicar_lag_22', 'bili_lag_22', 'bili_dir_lag_22', 'bnd_lag_22', 'bun_lag_22', 'ca_lag_22', 'cai_lag_22', 'ck_lag_22', 'ckmb_lag_22', 'cl_lag_22', 'crea_lag_22', 'crp_lag_22', 'dbp_lag_22', 'fgn_lag_22', 'fio2_lag_22', 'glu_lag_22', 'hgb_lag_22', 'hr_lag_22', 'inr_pt_lag_22', 'k_lag_22', 'lact_lag_22', 'lymph_lag_22', 'map_lag_22', 'mch_lag_22', 'mchc_lag_22', 'mcv_lag_22', 'methb_lag_22', 'mg_lag_22', 'na_lag_22', 'neut_lag_22', 'o2sat_lag_22', 'pco2_lag_22', 'ph_lag_22', 'phos_lag_22', 'plt_lag_22', 'po2_lag_22', 'ptt_lag_22', 'resp_lag_22', 'sbp_lag_22', 'temp_lag_22', 'tnt_lag_22', 'urine_lag_22', 'wbc_lag_22', 'alb_lag_23', 'alp_lag_23', 'alt_lag_23', 'ast_lag_23', 'be_lag_23', 'bicar_lag_23', 'bili_lag_23', 'bili_dir_lag_23', 'bnd_lag_23', 'bun_lag_23', 'ca_lag_23', 'cai_lag_23', 'ck_lag_23', 'ckmb_lag_23', 'cl_lag_23', 'crea_lag_23', 'crp_lag_23', 'dbp_lag_23', 'fgn_lag_23', 'fio2_lag_23', 'glu_lag_23', 'hgb_lag_23', 'hr_lag_23', 'inr_pt_lag_23', 'k_lag_23', 'lact_lag_23', 'lymph_lag_23', 'map_lag_23', 'mch_lag_23', 'mchc_lag_23', 'mcv_lag_23', 'methb_lag_23', 'mg_lag_23', 'na_lag_23', 'neut_lag_23', 'o2sat_lag_23', 'pco2_lag_23', 'ph_lag_23', 'phos_lag_23', 'plt_lag_23', 'po2_lag_23', 'ptt_lag_23', 'resp_lag_23', 'sbp_lag_23', 'temp_lag_23', 'tnt_lag_23', 'urine_lag_23', 'wbc_lag_23', 'alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24' ]
 PredictionDatasetpytorch.add_relative_time_idx=False
 PredictionDatasetpytorch.target="label"
+            
 PredictionDatasetpytorch.time_varying_unknown_reals=[
-                "label"
+                "label",
+              
             ]
 PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetpytorch.lagged_variables=["alb",
+PredictionDatasetpytorch.lagged_variables=[  "alb",
                 "alp",
                 "alt",
                 "ast",
@@ -78,6 +80,6 @@ PredictionDatasetpytorch.lagged_variables=["alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc",]
+                "wbc"]
 
                 
\ No newline at end of file
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 75233dca..3969df0f 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -23,17 +23,18 @@ model/hyperparameter.lr_scheduler = "exponential"
 
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
-PredictionDatasetpytorch.max_prediction_length = 24
-PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24', 'alb_lag_25', 'alp_lag_25', 'alt_lag_25', 'ast_lag_25', 'be_lag_25', 'bicar_lag_25', 'bili_lag_25', 'bili_dir_lag_25', 'bnd_lag_25', 'bun_lag_25', 'ca_lag_25', 'cai_lag_25', 'ck_lag_25', 'ckmb_lag_25', 'cl_lag_25', 'crea_lag_25', 'crp_lag_25', 'dbp_lag_25', 'fgn_lag_25', 'fio2_lag_25', 'glu_lag_25', 'hgb_lag_25', 'hr_lag_25', 'inr_pt_lag_25', 'k_lag_25', 'lact_lag_25', 'lymph_lag_25', 'map_lag_25', 'mch_lag_25', 'mchc_lag_25', 'mcv_lag_25', 'methb_lag_25', 'mg_lag_25', 'na_lag_25', 'neut_lag_25', 'o2sat_lag_25', 'pco2_lag_25', 'ph_lag_25', 'phos_lag_25', 'plt_lag_25', 'po2_lag_25', 'ptt_lag_25', 'resp_lag_25', 'sbp_lag_25', 'temp_lag_25', 'tnt_lag_25', 'urine_lag_25', 'wbc_lag_25', 'alb_lag_26', 'alp_lag_26', 'alt_lag_26', 'ast_lag_26', 'be_lag_26', 'bicar_lag_26', 'bili_lag_26', 'bili_dir_lag_26', 'bnd_lag_26', 'bun_lag_26', 'ca_lag_26', 'cai_lag_26', 'ck_lag_26', 'ckmb_lag_26', 'cl_lag_26', 'crea_lag_26', 'crp_lag_26', 'dbp_lag_26', 'fgn_lag_26', 'fio2_lag_26', 'glu_lag_26', 'hgb_lag_26', 'hr_lag_26', 'inr_pt_lag_26', 'k_lag_26', 'lact_lag_26', 'lymph_lag_26', 'map_lag_26', 'mch_lag_26', 'mchc_lag_26', 'mcv_lag_26', 'methb_lag_26', 'mg_lag_26', 'na_lag_26', 'neut_lag_26', 'o2sat_lag_26', 'pco2_lag_26', 'ph_lag_26', 'phos_lag_26', 'plt_lag_26', 'po2_lag_26', 'ptt_lag_26', 'resp_lag_26', 'sbp_lag_26', 'temp_lag_26', 'tnt_lag_26', 'urine_lag_26', 'wbc_lag_26', 'alb_lag_27', 'alp_lag_27', 'alt_lag_27', 'ast_lag_27', 'be_lag_27', 'bicar_lag_27', 'bili_lag_27', 'bili_dir_lag_27', 'bnd_lag_27', 'bun_lag_27', 'ca_lag_27', 'cai_lag_27', 'ck_lag_27', 'ckmb_lag_27', 'cl_lag_27', 'crea_lag_27', 'crp_lag_27', 'dbp_lag_27', 'fgn_lag_27', 'fio2_lag_27', 'glu_lag_27', 'hgb_lag_27', 'hr_lag_27', 'inr_pt_lag_27', 'k_lag_27', 'lact_lag_27', 'lymph_lag_27', 'map_lag_27', 'mch_lag_27', 'mchc_lag_27', 'mcv_lag_27', 'methb_lag_27', 'mg_lag_27', 'na_lag_27', 'neut_lag_27', 'o2sat_lag_27', 'pco2_lag_27', 'ph_lag_27', 'phos_lag_27', 'plt_lag_27', 'po2_lag_27', 'ptt_lag_27', 'resp_lag_27', 'sbp_lag_27', 'temp_lag_27', 'tnt_lag_27', 'urine_lag_27', 'wbc_lag_27', 'alb_lag_28', 'alp_lag_28', 'alt_lag_28', 'ast_lag_28', 'be_lag_28', 'bicar_lag_28', 'bili_lag_28', 'bili_dir_lag_28', 'bnd_lag_28', 'bun_lag_28', 'ca_lag_28', 'cai_lag_28', 'ck_lag_28', 'ckmb_lag_28', 'cl_lag_28', 'crea_lag_28', 'crp_lag_28', 'dbp_lag_28', 'fgn_lag_28', 'fio2_lag_28', 'glu_lag_28', 'hgb_lag_28', 'hr_lag_28', 'inr_pt_lag_28', 'k_lag_28', 'lact_lag_28', 'lymph_lag_28', 'map_lag_28', 'mch_lag_28', 'mchc_lag_28', 'mcv_lag_28', 'methb_lag_28', 'mg_lag_28', 'na_lag_28', 'neut_lag_28', 'o2sat_lag_28', 'pco2_lag_28', 'ph_lag_28', 'phos_lag_28', 'plt_lag_28', 'po2_lag_28', 'ptt_lag_28', 'resp_lag_28', 'sbp_lag_28', 'temp_lag_28', 'tnt_lag_28', 'urine_lag_28', 'wbc_lag_28', 'alb_lag_29', 'alp_lag_29', 'alt_lag_29', 'ast_lag_29', 'be_lag_29', 'bicar_lag_29', 'bili_lag_29', 'bili_dir_lag_29', 'bnd_lag_29', 'bun_lag_29', 'ca_lag_29', 'cai_lag_29', 'ck_lag_29', 'ckmb_lag_29', 'cl_lag_29', 'crea_lag_29', 'crp_lag_29', 'dbp_lag_29', 'fgn_lag_29', 'fio2_lag_29', 'glu_lag_29', 'hgb_lag_29', 'hr_lag_29', 'inr_pt_lag_29', 'k_lag_29', 'lact_lag_29', 'lymph_lag_29', 'map_lag_29', 'mch_lag_29', 'mchc_lag_29', 'mcv_lag_29', 'methb_lag_29', 'mg_lag_29', 'na_lag_29', 'neut_lag_29', 'o2sat_lag_29', 'pco2_lag_29', 'ph_lag_29', 'phos_lag_29', 'plt_lag_29', 'po2_lag_29', 'ptt_lag_29', 'resp_lag_29', 'sbp_lag_29', 'temp_lag_29', 'tnt_lag_29', 'urine_lag_29', 'wbc_lag_29', 'alb_lag_30', 'alp_lag_30', 'alt_lag_30', 'ast_lag_30', 'be_lag_30', 'bicar_lag_30', 'bili_lag_30', 'bili_dir_lag_30', 'bnd_lag_30', 'bun_lag_30', 'ca_lag_30', 'cai_lag_30', 'ck_lag_30', 'ckmb_lag_30', 'cl_lag_30', 'crea_lag_30', 'crp_lag_30', 'dbp_lag_30', 'fgn_lag_30', 'fio2_lag_30', 'glu_lag_30', 'hgb_lag_30', 'hr_lag_30', 'inr_pt_lag_30', 'k_lag_30', 'lact_lag_30', 'lymph_lag_30', 'map_lag_30', 'mch_lag_30', 'mchc_lag_30', 'mcv_lag_30', 'methb_lag_30', 'mg_lag_30', 'na_lag_30', 'neut_lag_30', 'o2sat_lag_30', 'pco2_lag_30', 'ph_lag_30', 'phos_lag_30', 'plt_lag_30', 'po2_lag_30', 'ptt_lag_30', 'resp_lag_30', 'sbp_lag_30', 'temp_lag_30', 'tnt_lag_30', 'urine_lag_30', 'wbc_lag_30', 'alb_lag_31', 'alp_lag_31', 'alt_lag_31', 'ast_lag_31', 'be_lag_31', 'bicar_lag_31', 'bili_lag_31', 'bili_dir_lag_31', 'bnd_lag_31', 'bun_lag_31', 'ca_lag_31', 'cai_lag_31', 'ck_lag_31', 'ckmb_lag_31', 'cl_lag_31', 'crea_lag_31', 'crp_lag_31', 'dbp_lag_31', 'fgn_lag_31', 'fio2_lag_31', 'glu_lag_31', 'hgb_lag_31', 'hr_lag_31', 'inr_pt_lag_31', 'k_lag_31', 'lact_lag_31', 'lymph_lag_31', 'map_lag_31', 'mch_lag_31', 'mchc_lag_31', 'mcv_lag_31', 'methb_lag_31', 'mg_lag_31', 'na_lag_31', 'neut_lag_31', 'o2sat_lag_31', 'pco2_lag_31', 'ph_lag_31', 'phos_lag_31', 'plt_lag_31', 'po2_lag_31', 'ptt_lag_31', 'resp_lag_31', 'sbp_lag_31', 'temp_lag_31', 'tnt_lag_31', 'urine_lag_31', 'wbc_lag_31', 'alb_lag_32', 'alp_lag_32', 'alt_lag_32', 'ast_lag_32', 'be_lag_32', 'bicar_lag_32', 'bili_lag_32', 'bili_dir_lag_32', 'bnd_lag_32', 'bun_lag_32', 'ca_lag_32', 'cai_lag_32', 'ck_lag_32', 'ckmb_lag_32', 'cl_lag_32', 'crea_lag_32', 'crp_lag_32', 'dbp_lag_32', 'fgn_lag_32', 'fio2_lag_32', 'glu_lag_32', 'hgb_lag_32', 'hr_lag_32', 'inr_pt_lag_32', 'k_lag_32', 'lact_lag_32', 'lymph_lag_32', 'map_lag_32', 'mch_lag_32', 'mchc_lag_32', 'mcv_lag_32', 'methb_lag_32', 'mg_lag_32', 'na_lag_32', 'neut_lag_32', 'o2sat_lag_32', 'pco2_lag_32', 'ph_lag_32', 'phos_lag_32', 'plt_lag_32', 'po2_lag_32', 'ptt_lag_32', 'resp_lag_32', 'sbp_lag_32', 'temp_lag_32', 'tnt_lag_32', 'urine_lag_32', 'wbc_lag_32', 'alb_lag_33', 'alp_lag_33', 'alt_lag_33', 'ast_lag_33', 'be_lag_33', 'bicar_lag_33', 'bili_lag_33', 'bili_dir_lag_33', 'bnd_lag_33', 'bun_lag_33', 'ca_lag_33', 'cai_lag_33', 'ck_lag_33', 'ckmb_lag_33', 'cl_lag_33', 'crea_lag_33', 'crp_lag_33', 'dbp_lag_33', 'fgn_lag_33', 'fio2_lag_33', 'glu_lag_33', 'hgb_lag_33', 'hr_lag_33', 'inr_pt_lag_33', 'k_lag_33', 'lact_lag_33', 'lymph_lag_33', 'map_lag_33', 'mch_lag_33', 'mchc_lag_33', 'mcv_lag_33', 'methb_lag_33', 'mg_lag_33', 'na_lag_33', 'neut_lag_33', 'o2sat_lag_33', 'pco2_lag_33', 'ph_lag_33', 'phos_lag_33', 'plt_lag_33', 'po2_lag_33', 'ptt_lag_33', 'resp_lag_33', 'sbp_lag_33', 'temp_lag_33', 'tnt_lag_33', 'urine_lag_33', 'wbc_lag_33', 'alb_lag_34', 'alp_lag_34', 'alt_lag_34', 'ast_lag_34', 'be_lag_34', 'bicar_lag_34', 'bili_lag_34', 'bili_dir_lag_34', 'bnd_lag_34', 'bun_lag_34', 'ca_lag_34', 'cai_lag_34', 'ck_lag_34', 'ckmb_lag_34', 'cl_lag_34', 'crea_lag_34', 'crp_lag_34', 'dbp_lag_34', 'fgn_lag_34', 'fio2_lag_34', 'glu_lag_34', 'hgb_lag_34', 'hr_lag_34', 'inr_pt_lag_34', 'k_lag_34', 'lact_lag_34', 'lymph_lag_34', 'map_lag_34', 'mch_lag_34', 'mchc_lag_34', 'mcv_lag_34', 'methb_lag_34', 'mg_lag_34', 'na_lag_34', 'neut_lag_34', 'o2sat_lag_34', 'pco2_lag_34', 'ph_lag_34', 'phos_lag_34', 'plt_lag_34', 'po2_lag_34', 'ptt_lag_34', 'resp_lag_34', 'sbp_lag_34', 'temp_lag_34', 'tnt_lag_34', 'urine_lag_34', 'wbc_lag_34', 'alb_lag_35', 'alp_lag_35', 'alt_lag_35', 'ast_lag_35', 'be_lag_35', 'bicar_lag_35', 'bili_lag_35', 'bili_dir_lag_35', 'bnd_lag_35', 'bun_lag_35', 'ca_lag_35', 'cai_lag_35', 'ck_lag_35', 'ckmb_lag_35', 'cl_lag_35', 'crea_lag_35', 'crp_lag_35', 'dbp_lag_35', 'fgn_lag_35', 'fio2_lag_35', 'glu_lag_35', 'hgb_lag_35', 'hr_lag_35', 'inr_pt_lag_35', 'k_lag_35', 'lact_lag_35', 'lymph_lag_35', 'map_lag_35', 'mch_lag_35', 'mchc_lag_35', 'mcv_lag_35', 'methb_lag_35', 'mg_lag_35', 'na_lag_35', 'neut_lag_35', 'o2sat_lag_35', 'pco2_lag_35', 'ph_lag_35', 'phos_lag_35', 'plt_lag_35', 'po2_lag_35', 'ptt_lag_35', 'resp_lag_35', 'sbp_lag_35', 'temp_lag_35', 'tnt_lag_35', 'urine_lag_35', 'wbc_lag_35', 'alb_lag_36', 'alp_lag_36', 'alt_lag_36', 'ast_lag_36', 'be_lag_36', 'bicar_lag_36', 'bili_lag_36', 'bili_dir_lag_36', 'bnd_lag_36', 'bun_lag_36', 'ca_lag_36', 'cai_lag_36', 'ck_lag_36', 'ckmb_lag_36', 'cl_lag_36', 'crea_lag_36', 'crp_lag_36', 'dbp_lag_36', 'fgn_lag_36', 'fio2_lag_36', 'glu_lag_36', 'hgb_lag_36', 'hr_lag_36', 'inr_pt_lag_36', 'k_lag_36', 'lact_lag_36', 'lymph_lag_36', 'map_lag_36', 'mch_lag_36', 'mchc_lag_36', 'mcv_lag_36', 'methb_lag_36', 'mg_lag_36', 'na_lag_36', 'neut_lag_36', 'o2sat_lag_36', 'pco2_lag_36', 'ph_lag_36', 'phos_lag_36', 'plt_lag_36', 'po2_lag_36', 'ptt_lag_36', 'resp_lag_36', 'sbp_lag_36', 'temp_lag_36', 'tnt_lag_36', 'urine_lag_36', 'wbc_lag_36', 'alb_lag_37', 'alp_lag_37', 'alt_lag_37', 'ast_lag_37', 'be_lag_37', 'bicar_lag_37', 'bili_lag_37', 'bili_dir_lag_37', 'bnd_lag_37', 'bun_lag_37', 'ca_lag_37', 'cai_lag_37', 'ck_lag_37', 'ckmb_lag_37', 'cl_lag_37', 'crea_lag_37', 'crp_lag_37', 'dbp_lag_37', 'fgn_lag_37', 'fio2_lag_37', 'glu_lag_37', 'hgb_lag_37', 'hr_lag_37', 'inr_pt_lag_37', 'k_lag_37', 'lact_lag_37', 'lymph_lag_37', 'map_lag_37', 'mch_lag_37', 'mchc_lag_37', 'mcv_lag_37', 'methb_lag_37', 'mg_lag_37', 'na_lag_37', 'neut_lag_37', 'o2sat_lag_37', 'pco2_lag_37', 'ph_lag_37', 'phos_lag_37', 'plt_lag_37', 'po2_lag_37', 'ptt_lag_37', 'resp_lag_37', 'sbp_lag_37', 'temp_lag_37', 'tnt_lag_37', 'urine_lag_37', 'wbc_lag_37', 'alb_lag_38', 'alp_lag_38', 'alt_lag_38', 'ast_lag_38', 'be_lag_38', 'bicar_lag_38', 'bili_lag_38', 'bili_dir_lag_38', 'bnd_lag_38', 'bun_lag_38', 'ca_lag_38', 'cai_lag_38', 'ck_lag_38', 'ckmb_lag_38', 'cl_lag_38', 'crea_lag_38', 'crp_lag_38', 'dbp_lag_38', 'fgn_lag_38', 'fio2_lag_38', 'glu_lag_38', 'hgb_lag_38', 'hr_lag_38', 'inr_pt_lag_38', 'k_lag_38', 'lact_lag_38', 'lymph_lag_38', 'map_lag_38', 'mch_lag_38', 'mchc_lag_38', 'mcv_lag_38', 'methb_lag_38', 'mg_lag_38', 'na_lag_38', 'neut_lag_38', 'o2sat_lag_38', 'pco2_lag_38', 'ph_lag_38', 'phos_lag_38', 'plt_lag_38', 'po2_lag_38', 'ptt_lag_38', 'resp_lag_38', 'sbp_lag_38', 'temp_lag_38', 'tnt_lag_38', 'urine_lag_38', 'wbc_lag_38', 'alb_lag_39', 'alp_lag_39', 'alt_lag_39', 'ast_lag_39', 'be_lag_39', 'bicar_lag_39', 'bili_lag_39', 'bili_dir_lag_39', 'bnd_lag_39', 'bun_lag_39', 'ca_lag_39', 'cai_lag_39', 'ck_lag_39', 'ckmb_lag_39', 'cl_lag_39', 'crea_lag_39', 'crp_lag_39', 'dbp_lag_39', 'fgn_lag_39', 'fio2_lag_39', 'glu_lag_39', 'hgb_lag_39', 'hr_lag_39', 'inr_pt_lag_39', 'k_lag_39', 'lact_lag_39', 'lymph_lag_39', 'map_lag_39', 'mch_lag_39', 'mchc_lag_39', 'mcv_lag_39', 'methb_lag_39', 'mg_lag_39', 'na_lag_39', 'neut_lag_39', 'o2sat_lag_39', 'pco2_lag_39', 'ph_lag_39', 'phos_lag_39', 'plt_lag_39', 'po2_lag_39', 'ptt_lag_39', 'resp_lag_39', 'sbp_lag_39', 'temp_lag_39', 'tnt_lag_39', 'urine_lag_39', 'wbc_lag_39', 'alb_lag_40', 'alp_lag_40', 'alt_lag_40', 'ast_lag_40', 'be_lag_40', 'bicar_lag_40', 'bili_lag_40', 'bili_dir_lag_40', 'bnd_lag_40', 'bun_lag_40', 'ca_lag_40', 'cai_lag_40', 'ck_lag_40', 'ckmb_lag_40', 'cl_lag_40', 'crea_lag_40', 'crp_lag_40', 'dbp_lag_40', 'fgn_lag_40', 'fio2_lag_40', 'glu_lag_40', 'hgb_lag_40', 'hr_lag_40', 'inr_pt_lag_40', 'k_lag_40', 'lact_lag_40', 'lymph_lag_40', 'map_lag_40', 'mch_lag_40', 'mchc_lag_40', 'mcv_lag_40', 'methb_lag_40', 'mg_lag_40', 'na_lag_40', 'neut_lag_40', 'o2sat_lag_40', 'pco2_lag_40', 'ph_lag_40', 'phos_lag_40', 'plt_lag_40', 'po2_lag_40', 'ptt_lag_40', 'resp_lag_40', 'sbp_lag_40', 'temp_lag_40', 'tnt_lag_40', 'urine_lag_40', 'wbc_lag_40', 'alb_lag_41', 'alp_lag_41', 'alt_lag_41', 'ast_lag_41', 'be_lag_41', 'bicar_lag_41', 'bili_lag_41', 'bili_dir_lag_41', 'bnd_lag_41', 'bun_lag_41', 'ca_lag_41', 'cai_lag_41', 'ck_lag_41', 'ckmb_lag_41', 'cl_lag_41', 'crea_lag_41', 'crp_lag_41', 'dbp_lag_41', 'fgn_lag_41', 'fio2_lag_41', 'glu_lag_41', 'hgb_lag_41', 'hr_lag_41', 'inr_pt_lag_41', 'k_lag_41', 'lact_lag_41', 'lymph_lag_41', 'map_lag_41', 'mch_lag_41', 'mchc_lag_41', 'mcv_lag_41', 'methb_lag_41', 'mg_lag_41', 'na_lag_41', 'neut_lag_41', 'o2sat_lag_41', 'pco2_lag_41', 'ph_lag_41', 'phos_lag_41', 'plt_lag_41', 'po2_lag_41', 'ptt_lag_41', 'resp_lag_41', 'sbp_lag_41', 'temp_lag_41', 'tnt_lag_41', 'urine_lag_41', 'wbc_lag_41', 'alb_lag_42', 'alp_lag_42', 'alt_lag_42', 'ast_lag_42', 'be_lag_42', 'bicar_lag_42', 'bili_lag_42', 'bili_dir_lag_42', 'bnd_lag_42', 'bun_lag_42', 'ca_lag_42', 'cai_lag_42', 'ck_lag_42', 'ckmb_lag_42', 'cl_lag_42', 'crea_lag_42', 'crp_lag_42', 'dbp_lag_42', 'fgn_lag_42', 'fio2_lag_42', 'glu_lag_42', 'hgb_lag_42', 'hr_lag_42', 'inr_pt_lag_42', 'k_lag_42', 'lact_lag_42', 'lymph_lag_42', 'map_lag_42', 'mch_lag_42', 'mchc_lag_42', 'mcv_lag_42', 'methb_lag_42', 'mg_lag_42', 'na_lag_42', 'neut_lag_42', 'o2sat_lag_42', 'pco2_lag_42', 'ph_lag_42', 'phos_lag_42', 'plt_lag_42', 'po2_lag_42', 'ptt_lag_42', 'resp_lag_42', 'sbp_lag_42', 'temp_lag_42', 'tnt_lag_42', 'urine_lag_42', 'wbc_lag_42', 'alb_lag_43', 'alp_lag_43', 'alt_lag_43', 'ast_lag_43', 'be_lag_43', 'bicar_lag_43', 'bili_lag_43', 'bili_dir_lag_43', 'bnd_lag_43', 'bun_lag_43', 'ca_lag_43', 'cai_lag_43', 'ck_lag_43', 'ckmb_lag_43', 'cl_lag_43', 'crea_lag_43', 'crp_lag_43', 'dbp_lag_43', 'fgn_lag_43', 'fio2_lag_43', 'glu_lag_43', 'hgb_lag_43', 'hr_lag_43', 'inr_pt_lag_43', 'k_lag_43', 'lact_lag_43', 'lymph_lag_43', 'map_lag_43', 'mch_lag_43', 'mchc_lag_43', 'mcv_lag_43', 'methb_lag_43', 'mg_lag_43', 'na_lag_43', 'neut_lag_43', 'o2sat_lag_43', 'pco2_lag_43', 'ph_lag_43', 'phos_lag_43', 'plt_lag_43', 'po2_lag_43', 'ptt_lag_43', 'resp_lag_43', 'sbp_lag_43', 'temp_lag_43', 'tnt_lag_43', 'urine_lag_43', 'wbc_lag_43', 'alb_lag_44', 'alp_lag_44', 'alt_lag_44', 'ast_lag_44', 'be_lag_44', 'bicar_lag_44', 'bili_lag_44', 'bili_dir_lag_44', 'bnd_lag_44', 'bun_lag_44', 'ca_lag_44', 'cai_lag_44', 'ck_lag_44', 'ckmb_lag_44', 'cl_lag_44', 'crea_lag_44', 'crp_lag_44', 'dbp_lag_44', 'fgn_lag_44', 'fio2_lag_44', 'glu_lag_44', 'hgb_lag_44', 'hr_lag_44', 'inr_pt_lag_44', 'k_lag_44', 'lact_lag_44', 'lymph_lag_44', 'map_lag_44', 'mch_lag_44', 'mchc_lag_44', 'mcv_lag_44', 'methb_lag_44', 'mg_lag_44', 'na_lag_44', 'neut_lag_44', 'o2sat_lag_44', 'pco2_lag_44', 'ph_lag_44', 'phos_lag_44', 'plt_lag_44', 'po2_lag_44', 'ptt_lag_44', 'resp_lag_44', 'sbp_lag_44', 'temp_lag_44', 'tnt_lag_44', 'urine_lag_44', 'wbc_lag_44', 'alb_lag_45', 'alp_lag_45', 'alt_lag_45', 'ast_lag_45', 'be_lag_45', 'bicar_lag_45', 'bili_lag_45', 'bili_dir_lag_45', 'bnd_lag_45', 'bun_lag_45', 'ca_lag_45', 'cai_lag_45', 'ck_lag_45', 'ckmb_lag_45', 'cl_lag_45', 'crea_lag_45', 'crp_lag_45', 'dbp_lag_45', 'fgn_lag_45', 'fio2_lag_45', 'glu_lag_45', 'hgb_lag_45', 'hr_lag_45', 'inr_pt_lag_45', 'k_lag_45', 'lact_lag_45', 'lymph_lag_45', 'map_lag_45', 'mch_lag_45', 'mchc_lag_45', 'mcv_lag_45', 'methb_lag_45', 'mg_lag_45', 'na_lag_45', 'neut_lag_45', 'o2sat_lag_45', 'pco2_lag_45', 'ph_lag_45', 'phos_lag_45', 'plt_lag_45', 'po2_lag_45', 'ptt_lag_45', 'resp_lag_45', 'sbp_lag_45', 'temp_lag_45', 'tnt_lag_45', 'urine_lag_45', 'wbc_lag_45', 'alb_lag_46', 'alp_lag_46', 'alt_lag_46', 'ast_lag_46', 'be_lag_46', 'bicar_lag_46', 'bili_lag_46', 'bili_dir_lag_46', 'bnd_lag_46', 'bun_lag_46', 'ca_lag_46', 'cai_lag_46', 'ck_lag_46', 'ckmb_lag_46', 'cl_lag_46', 'crea_lag_46', 'crp_lag_46', 'dbp_lag_46', 'fgn_lag_46', 'fio2_lag_46', 'glu_lag_46', 'hgb_lag_46', 'hr_lag_46', 'inr_pt_lag_46', 'k_lag_46', 'lact_lag_46', 'lymph_lag_46', 'map_lag_46', 'mch_lag_46', 'mchc_lag_46', 'mcv_lag_46', 'methb_lag_46', 'mg_lag_46', 'na_lag_46', 'neut_lag_46', 'o2sat_lag_46', 'pco2_lag_46', 'ph_lag_46', 'phos_lag_46', 'plt_lag_46', 'po2_lag_46', 'ptt_lag_46', 'resp_lag_46', 'sbp_lag_46', 'temp_lag_46', 'tnt_lag_46', 'urine_lag_46', 'wbc_lag_46', 'alb_lag_47', 'alp_lag_47', 'alt_lag_47', 'ast_lag_47', 'be_lag_47', 'bicar_lag_47', 'bili_lag_47', 'bili_dir_lag_47', 'bnd_lag_47', 'bun_lag_47', 'ca_lag_47', 'cai_lag_47', 'ck_lag_47', 'ckmb_lag_47', 'cl_lag_47', 'crea_lag_47', 'crp_lag_47', 'dbp_lag_47', 'fgn_lag_47', 'fio2_lag_47', 'glu_lag_47', 'hgb_lag_47', 'hr_lag_47', 'inr_pt_lag_47', 'k_lag_47', 'lact_lag_47', 'lymph_lag_47', 'map_lag_47', 'mch_lag_47', 'mchc_lag_47', 'mcv_lag_47', 'methb_lag_47', 'mg_lag_47', 'na_lag_47', 'neut_lag_47', 'o2sat_lag_47', 'pco2_lag_47', 'ph_lag_47', 'phos_lag_47', 'plt_lag_47', 'po2_lag_47', 'ptt_lag_47', 'resp_lag_47', 'sbp_lag_47', 'temp_lag_47', 'tnt_lag_47', 'urine_lag_47', 'wbc_lag_47', 'alb_lag_48', 'alp_lag_48', 'alt_lag_48', 'ast_lag_48', 'be_lag_48', 'bicar_lag_48', 'bili_lag_48', 'bili_dir_lag_48', 'bnd_lag_48', 'bun_lag_48', 'ca_lag_48', 'cai_lag_48', 'ck_lag_48', 'ckmb_lag_48', 'cl_lag_48', 'crea_lag_48', 'crp_lag_48', 'dbp_lag_48', 'fgn_lag_48', 'fio2_lag_48', 'glu_lag_48', 'hgb_lag_48', 'hr_lag_48', 'inr_pt_lag_48', 'k_lag_48', 'lact_lag_48', 'lymph_lag_48', 'map_lag_48', 'mch_lag_48', 'mchc_lag_48', 'mcv_lag_48', 'methb_lag_48', 'mg_lag_48', 'na_lag_48', 'neut_lag_48', 'o2sat_lag_48', 'pco2_lag_48', 'ph_lag_48', 'phos_lag_48', 'plt_lag_48', 'po2_lag_48', 'ptt_lag_48', 'resp_lag_48', 'sbp_lag_48', 'temp_lag_48', 'tnt_lag_48', 'urine_lag_48', 'wbc_lag_48']
+PredictionDatasetpytorch.max_prediction_length = 1
+PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_1', 'alp_lag_1', 'alt_lag_1', 'ast_lag_1', 'be_lag_1', 'bicar_lag_1', 'bili_lag_1', 'bili_dir_lag_1', 'bnd_lag_1', 'bun_lag_1', 'ca_lag_1', 'cai_lag_1', 'ck_lag_1', 'ckmb_lag_1', 'cl_lag_1', 'crea_lag_1', 'crp_lag_1', 'dbp_lag_1', 'fgn_lag_1', 'fio2_lag_1', 'glu_lag_1', 'hgb_lag_1', 'hr_lag_1', 'inr_pt_lag_1', 'k_lag_1', 'lact_lag_1', 'lymph_lag_1', 'map_lag_1', 'mch_lag_1', 'mchc_lag_1', 'mcv_lag_1', 'methb_lag_1', 'mg_lag_1', 'na_lag_1', 'neut_lag_1', 'o2sat_lag_1', 'pco2_lag_1', 'ph_lag_1', 'phos_lag_1', 'plt_lag_1', 'po2_lag_1', 'ptt_lag_1', 'resp_lag_1', 'sbp_lag_1', 'temp_lag_1', 'tnt_lag_1', 'urine_lag_1', 'wbc_lag_1', 'alb_lag_2', 'alp_lag_2', 'alt_lag_2', 'ast_lag_2', 'be_lag_2', 'bicar_lag_2', 'bili_lag_2', 'bili_dir_lag_2', 'bnd_lag_2', 'bun_lag_2', 'ca_lag_2', 'cai_lag_2', 'ck_lag_2', 'ckmb_lag_2', 'cl_lag_2', 'crea_lag_2', 'crp_lag_2', 'dbp_lag_2', 'fgn_lag_2', 'fio2_lag_2', 'glu_lag_2', 'hgb_lag_2', 'hr_lag_2', 'inr_pt_lag_2', 'k_lag_2', 'lact_lag_2', 'lymph_lag_2', 'map_lag_2', 'mch_lag_2', 'mchc_lag_2', 'mcv_lag_2', 'methb_lag_2', 'mg_lag_2', 'na_lag_2', 'neut_lag_2', 'o2sat_lag_2', 'pco2_lag_2', 'ph_lag_2', 'phos_lag_2', 'plt_lag_2', 'po2_lag_2', 'ptt_lag_2', 'resp_lag_2', 'sbp_lag_2', 'temp_lag_2', 'tnt_lag_2', 'urine_lag_2', 'wbc_lag_2', 'alb_lag_3', 'alp_lag_3', 'alt_lag_3', 'ast_lag_3', 'be_lag_3', 'bicar_lag_3', 'bili_lag_3', 'bili_dir_lag_3', 'bnd_lag_3', 'bun_lag_3', 'ca_lag_3', 'cai_lag_3', 'ck_lag_3', 'ckmb_lag_3', 'cl_lag_3', 'crea_lag_3', 'crp_lag_3', 'dbp_lag_3', 'fgn_lag_3', 'fio2_lag_3', 'glu_lag_3', 'hgb_lag_3', 'hr_lag_3', 'inr_pt_lag_3', 'k_lag_3', 'lact_lag_3', 'lymph_lag_3', 'map_lag_3', 'mch_lag_3', 'mchc_lag_3', 'mcv_lag_3', 'methb_lag_3', 'mg_lag_3', 'na_lag_3', 'neut_lag_3', 'o2sat_lag_3', 'pco2_lag_3', 'ph_lag_3', 'phos_lag_3', 'plt_lag_3', 'po2_lag_3', 'ptt_lag_3', 'resp_lag_3', 'sbp_lag_3', 'temp_lag_3', 'tnt_lag_3', 'urine_lag_3', 'wbc_lag_3', 'alb_lag_4', 'alp_lag_4', 'alt_lag_4', 'ast_lag_4', 'be_lag_4', 'bicar_lag_4', 'bili_lag_4', 'bili_dir_lag_4', 'bnd_lag_4', 'bun_lag_4', 'ca_lag_4', 'cai_lag_4', 'ck_lag_4', 'ckmb_lag_4', 'cl_lag_4', 'crea_lag_4', 'crp_lag_4', 'dbp_lag_4', 'fgn_lag_4', 'fio2_lag_4', 'glu_lag_4', 'hgb_lag_4', 'hr_lag_4', 'inr_pt_lag_4', 'k_lag_4', 'lact_lag_4', 'lymph_lag_4', 'map_lag_4', 'mch_lag_4', 'mchc_lag_4', 'mcv_lag_4', 'methb_lag_4', 'mg_lag_4', 'na_lag_4', 'neut_lag_4', 'o2sat_lag_4', 'pco2_lag_4', 'ph_lag_4', 'phos_lag_4', 'plt_lag_4', 'po2_lag_4', 'ptt_lag_4', 'resp_lag_4', 'sbp_lag_4', 'temp_lag_4', 'tnt_lag_4', 'urine_lag_4', 'wbc_lag_4', 'alb_lag_5', 'alp_lag_5', 'alt_lag_5', 'ast_lag_5', 'be_lag_5', 'bicar_lag_5', 'bili_lag_5', 'bili_dir_lag_5', 'bnd_lag_5', 'bun_lag_5', 'ca_lag_5', 'cai_lag_5', 'ck_lag_5', 'ckmb_lag_5', 'cl_lag_5', 'crea_lag_5', 'crp_lag_5', 'dbp_lag_5', 'fgn_lag_5', 'fio2_lag_5', 'glu_lag_5', 'hgb_lag_5', 'hr_lag_5', 'inr_pt_lag_5', 'k_lag_5', 'lact_lag_5', 'lymph_lag_5', 'map_lag_5', 'mch_lag_5', 'mchc_lag_5', 'mcv_lag_5', 'methb_lag_5', 'mg_lag_5', 'na_lag_5', 'neut_lag_5', 'o2sat_lag_5', 'pco2_lag_5', 'ph_lag_5', 'phos_lag_5', 'plt_lag_5', 'po2_lag_5', 'ptt_lag_5', 'resp_lag_5', 'sbp_lag_5', 'temp_lag_5', 'tnt_lag_5', 'urine_lag_5', 'wbc_lag_5', 'alb_lag_6', 'alp_lag_6', 'alt_lag_6', 'ast_lag_6', 'be_lag_6', 'bicar_lag_6', 'bili_lag_6', 'bili_dir_lag_6', 'bnd_lag_6', 'bun_lag_6', 'ca_lag_6', 'cai_lag_6', 'ck_lag_6', 'ckmb_lag_6', 'cl_lag_6', 'crea_lag_6', 'crp_lag_6', 'dbp_lag_6', 'fgn_lag_6', 'fio2_lag_6', 'glu_lag_6', 'hgb_lag_6', 'hr_lag_6', 'inr_pt_lag_6', 'k_lag_6', 'lact_lag_6', 'lymph_lag_6', 'map_lag_6', 'mch_lag_6', 'mchc_lag_6', 'mcv_lag_6', 'methb_lag_6', 'mg_lag_6', 'na_lag_6', 'neut_lag_6', 'o2sat_lag_6', 'pco2_lag_6', 'ph_lag_6', 'phos_lag_6', 'plt_lag_6', 'po2_lag_6', 'ptt_lag_6', 'resp_lag_6', 'sbp_lag_6', 'temp_lag_6', 'tnt_lag_6', 'urine_lag_6', 'wbc_lag_6', 'alb_lag_7', 'alp_lag_7', 'alt_lag_7', 'ast_lag_7', 'be_lag_7', 'bicar_lag_7', 'bili_lag_7', 'bili_dir_lag_7', 'bnd_lag_7', 'bun_lag_7', 'ca_lag_7', 'cai_lag_7', 'ck_lag_7', 'ckmb_lag_7', 'cl_lag_7', 'crea_lag_7', 'crp_lag_7', 'dbp_lag_7', 'fgn_lag_7', 'fio2_lag_7', 'glu_lag_7', 'hgb_lag_7', 'hr_lag_7', 'inr_pt_lag_7', 'k_lag_7', 'lact_lag_7', 'lymph_lag_7', 'map_lag_7', 'mch_lag_7', 'mchc_lag_7', 'mcv_lag_7', 'methb_lag_7', 'mg_lag_7', 'na_lag_7', 'neut_lag_7', 'o2sat_lag_7', 'pco2_lag_7', 'ph_lag_7', 'phos_lag_7', 'plt_lag_7', 'po2_lag_7', 'ptt_lag_7', 'resp_lag_7', 'sbp_lag_7', 'temp_lag_7', 'tnt_lag_7', 'urine_lag_7', 'wbc_lag_7', 'alb_lag_8', 'alp_lag_8', 'alt_lag_8', 'ast_lag_8', 'be_lag_8', 'bicar_lag_8', 'bili_lag_8', 'bili_dir_lag_8', 'bnd_lag_8', 'bun_lag_8', 'ca_lag_8', 'cai_lag_8', 'ck_lag_8', 'ckmb_lag_8', 'cl_lag_8', 'crea_lag_8', 'crp_lag_8', 'dbp_lag_8', 'fgn_lag_8', 'fio2_lag_8', 'glu_lag_8', 'hgb_lag_8', 'hr_lag_8', 'inr_pt_lag_8', 'k_lag_8', 'lact_lag_8', 'lymph_lag_8', 'map_lag_8', 'mch_lag_8', 'mchc_lag_8', 'mcv_lag_8', 'methb_lag_8', 'mg_lag_8', 'na_lag_8', 'neut_lag_8', 'o2sat_lag_8', 'pco2_lag_8', 'ph_lag_8', 'phos_lag_8', 'plt_lag_8', 'po2_lag_8', 'ptt_lag_8', 'resp_lag_8', 'sbp_lag_8', 'temp_lag_8', 'tnt_lag_8', 'urine_lag_8', 'wbc_lag_8', 'alb_lag_9', 'alp_lag_9', 'alt_lag_9', 'ast_lag_9', 'be_lag_9', 'bicar_lag_9', 'bili_lag_9', 'bili_dir_lag_9', 'bnd_lag_9', 'bun_lag_9', 'ca_lag_9', 'cai_lag_9', 'ck_lag_9', 'ckmb_lag_9', 'cl_lag_9', 'crea_lag_9', 'crp_lag_9', 'dbp_lag_9', 'fgn_lag_9', 'fio2_lag_9', 'glu_lag_9', 'hgb_lag_9', 'hr_lag_9', 'inr_pt_lag_9', 'k_lag_9', 'lact_lag_9', 'lymph_lag_9', 'map_lag_9', 'mch_lag_9', 'mchc_lag_9', 'mcv_lag_9', 'methb_lag_9', 'mg_lag_9', 'na_lag_9', 'neut_lag_9', 'o2sat_lag_9', 'pco2_lag_9', 'ph_lag_9', 'phos_lag_9', 'plt_lag_9', 'po2_lag_9', 'ptt_lag_9', 'resp_lag_9', 'sbp_lag_9', 'temp_lag_9', 'tnt_lag_9', 'urine_lag_9', 'wbc_lag_9', 'alb_lag_10', 'alp_lag_10', 'alt_lag_10', 'ast_lag_10', 'be_lag_10', 'bicar_lag_10', 'bili_lag_10', 'bili_dir_lag_10', 'bnd_lag_10', 'bun_lag_10', 'ca_lag_10', 'cai_lag_10', 'ck_lag_10', 'ckmb_lag_10', 'cl_lag_10', 'crea_lag_10', 'crp_lag_10', 'dbp_lag_10', 'fgn_lag_10', 'fio2_lag_10', 'glu_lag_10', 'hgb_lag_10', 'hr_lag_10', 'inr_pt_lag_10', 'k_lag_10', 'lact_lag_10', 'lymph_lag_10', 'map_lag_10', 'mch_lag_10', 'mchc_lag_10', 'mcv_lag_10', 'methb_lag_10', 'mg_lag_10', 'na_lag_10', 'neut_lag_10', 'o2sat_lag_10', 'pco2_lag_10', 'ph_lag_10', 'phos_lag_10', 'plt_lag_10', 'po2_lag_10', 'ptt_lag_10', 'resp_lag_10', 'sbp_lag_10', 'temp_lag_10', 'tnt_lag_10', 'urine_lag_10', 'wbc_lag_10', 'alb_lag_11', 'alp_lag_11', 'alt_lag_11', 'ast_lag_11', 'be_lag_11', 'bicar_lag_11', 'bili_lag_11', 'bili_dir_lag_11', 'bnd_lag_11', 'bun_lag_11', 'ca_lag_11', 'cai_lag_11', 'ck_lag_11', 'ckmb_lag_11', 'cl_lag_11', 'crea_lag_11', 'crp_lag_11', 'dbp_lag_11', 'fgn_lag_11', 'fio2_lag_11', 'glu_lag_11', 'hgb_lag_11', 'hr_lag_11', 'inr_pt_lag_11', 'k_lag_11', 'lact_lag_11', 'lymph_lag_11', 'map_lag_11', 'mch_lag_11', 'mchc_lag_11', 'mcv_lag_11', 'methb_lag_11', 'mg_lag_11', 'na_lag_11', 'neut_lag_11', 'o2sat_lag_11', 'pco2_lag_11', 'ph_lag_11', 'phos_lag_11', 'plt_lag_11', 'po2_lag_11', 'ptt_lag_11', 'resp_lag_11', 'sbp_lag_11', 'temp_lag_11', 'tnt_lag_11', 'urine_lag_11', 'wbc_lag_11', 'alb_lag_12', 'alp_lag_12', 'alt_lag_12', 'ast_lag_12', 'be_lag_12', 'bicar_lag_12', 'bili_lag_12', 'bili_dir_lag_12', 'bnd_lag_12', 'bun_lag_12', 'ca_lag_12', 'cai_lag_12', 'ck_lag_12', 'ckmb_lag_12', 'cl_lag_12', 'crea_lag_12', 'crp_lag_12', 'dbp_lag_12', 'fgn_lag_12', 'fio2_lag_12', 'glu_lag_12', 'hgb_lag_12', 'hr_lag_12', 'inr_pt_lag_12', 'k_lag_12', 'lact_lag_12', 'lymph_lag_12', 'map_lag_12', 'mch_lag_12', 'mchc_lag_12', 'mcv_lag_12', 'methb_lag_12', 'mg_lag_12', 'na_lag_12', 'neut_lag_12', 'o2sat_lag_12', 'pco2_lag_12', 'ph_lag_12', 'phos_lag_12', 'plt_lag_12', 'po2_lag_12', 'ptt_lag_12', 'resp_lag_12', 'sbp_lag_12', 'temp_lag_12', 'tnt_lag_12', 'urine_lag_12', 'wbc_lag_12', 'alb_lag_13', 'alp_lag_13', 'alt_lag_13', 'ast_lag_13', 'be_lag_13', 'bicar_lag_13', 'bili_lag_13', 'bili_dir_lag_13', 'bnd_lag_13', 'bun_lag_13', 'ca_lag_13', 'cai_lag_13', 'ck_lag_13', 'ckmb_lag_13', 'cl_lag_13', 'crea_lag_13', 'crp_lag_13', 'dbp_lag_13', 'fgn_lag_13', 'fio2_lag_13', 'glu_lag_13', 'hgb_lag_13', 'hr_lag_13', 'inr_pt_lag_13', 'k_lag_13', 'lact_lag_13', 'lymph_lag_13', 'map_lag_13', 'mch_lag_13', 'mchc_lag_13', 'mcv_lag_13', 'methb_lag_13', 'mg_lag_13', 'na_lag_13', 'neut_lag_13', 'o2sat_lag_13', 'pco2_lag_13', 'ph_lag_13', 'phos_lag_13', 'plt_lag_13', 'po2_lag_13', 'ptt_lag_13', 'resp_lag_13', 'sbp_lag_13', 'temp_lag_13', 'tnt_lag_13', 'urine_lag_13', 'wbc_lag_13', 'alb_lag_14', 'alp_lag_14', 'alt_lag_14', 'ast_lag_14', 'be_lag_14', 'bicar_lag_14', 'bili_lag_14', 'bili_dir_lag_14', 'bnd_lag_14', 'bun_lag_14', 'ca_lag_14', 'cai_lag_14', 'ck_lag_14', 'ckmb_lag_14', 'cl_lag_14', 'crea_lag_14', 'crp_lag_14', 'dbp_lag_14', 'fgn_lag_14', 'fio2_lag_14', 'glu_lag_14', 'hgb_lag_14', 'hr_lag_14', 'inr_pt_lag_14', 'k_lag_14', 'lact_lag_14', 'lymph_lag_14', 'map_lag_14', 'mch_lag_14', 'mchc_lag_14', 'mcv_lag_14', 'methb_lag_14', 'mg_lag_14', 'na_lag_14', 'neut_lag_14', 'o2sat_lag_14', 'pco2_lag_14', 'ph_lag_14', 'phos_lag_14', 'plt_lag_14', 'po2_lag_14', 'ptt_lag_14', 'resp_lag_14', 'sbp_lag_14', 'temp_lag_14', 'tnt_lag_14', 'urine_lag_14', 'wbc_lag_14', 'alb_lag_15', 'alp_lag_15', 'alt_lag_15', 'ast_lag_15', 'be_lag_15', 'bicar_lag_15', 'bili_lag_15', 'bili_dir_lag_15', 'bnd_lag_15', 'bun_lag_15', 'ca_lag_15', 'cai_lag_15', 'ck_lag_15', 'ckmb_lag_15', 'cl_lag_15', 'crea_lag_15', 'crp_lag_15', 'dbp_lag_15', 'fgn_lag_15', 'fio2_lag_15', 'glu_lag_15', 'hgb_lag_15', 'hr_lag_15', 'inr_pt_lag_15', 'k_lag_15', 'lact_lag_15', 'lymph_lag_15', 'map_lag_15', 'mch_lag_15', 'mchc_lag_15', 'mcv_lag_15', 'methb_lag_15', 'mg_lag_15', 'na_lag_15', 'neut_lag_15', 'o2sat_lag_15', 'pco2_lag_15', 'ph_lag_15', 'phos_lag_15', 'plt_lag_15', 'po2_lag_15', 'ptt_lag_15', 'resp_lag_15', 'sbp_lag_15', 'temp_lag_15', 'tnt_lag_15', 'urine_lag_15', 'wbc_lag_15', 'alb_lag_16', 'alp_lag_16', 'alt_lag_16', 'ast_lag_16', 'be_lag_16', 'bicar_lag_16', 'bili_lag_16', 'bili_dir_lag_16', 'bnd_lag_16', 'bun_lag_16', 'ca_lag_16', 'cai_lag_16', 'ck_lag_16', 'ckmb_lag_16', 'cl_lag_16', 'crea_lag_16', 'crp_lag_16', 'dbp_lag_16', 'fgn_lag_16', 'fio2_lag_16', 'glu_lag_16', 'hgb_lag_16', 'hr_lag_16', 'inr_pt_lag_16', 'k_lag_16', 'lact_lag_16', 'lymph_lag_16', 'map_lag_16', 'mch_lag_16', 'mchc_lag_16', 'mcv_lag_16', 'methb_lag_16', 'mg_lag_16', 'na_lag_16', 'neut_lag_16', 'o2sat_lag_16', 'pco2_lag_16', 'ph_lag_16', 'phos_lag_16', 'plt_lag_16', 'po2_lag_16', 'ptt_lag_16', 'resp_lag_16', 'sbp_lag_16', 'temp_lag_16', 'tnt_lag_16', 'urine_lag_16', 'wbc_lag_16', 'alb_lag_17', 'alp_lag_17', 'alt_lag_17', 'ast_lag_17', 'be_lag_17', 'bicar_lag_17', 'bili_lag_17', 'bili_dir_lag_17', 'bnd_lag_17', 'bun_lag_17', 'ca_lag_17', 'cai_lag_17', 'ck_lag_17', 'ckmb_lag_17', 'cl_lag_17', 'crea_lag_17', 'crp_lag_17', 'dbp_lag_17', 'fgn_lag_17', 'fio2_lag_17', 'glu_lag_17', 'hgb_lag_17', 'hr_lag_17', 'inr_pt_lag_17', 'k_lag_17', 'lact_lag_17', 'lymph_lag_17', 'map_lag_17', 'mch_lag_17', 'mchc_lag_17', 'mcv_lag_17', 'methb_lag_17', 'mg_lag_17', 'na_lag_17', 'neut_lag_17', 'o2sat_lag_17', 'pco2_lag_17', 'ph_lag_17', 'phos_lag_17', 'plt_lag_17', 'po2_lag_17', 'ptt_lag_17', 'resp_lag_17', 'sbp_lag_17', 'temp_lag_17', 'tnt_lag_17', 'urine_lag_17', 'wbc_lag_17', 'alb_lag_18', 'alp_lag_18', 'alt_lag_18', 'ast_lag_18', 'be_lag_18', 'bicar_lag_18', 'bili_lag_18', 'bili_dir_lag_18', 'bnd_lag_18', 'bun_lag_18', 'ca_lag_18', 'cai_lag_18', 'ck_lag_18', 'ckmb_lag_18', 'cl_lag_18', 'crea_lag_18', 'crp_lag_18', 'dbp_lag_18', 'fgn_lag_18', 'fio2_lag_18', 'glu_lag_18', 'hgb_lag_18', 'hr_lag_18', 'inr_pt_lag_18', 'k_lag_18', 'lact_lag_18', 'lymph_lag_18', 'map_lag_18', 'mch_lag_18', 'mchc_lag_18', 'mcv_lag_18', 'methb_lag_18', 'mg_lag_18', 'na_lag_18', 'neut_lag_18', 'o2sat_lag_18', 'pco2_lag_18', 'ph_lag_18', 'phos_lag_18', 'plt_lag_18', 'po2_lag_18', 'ptt_lag_18', 'resp_lag_18', 'sbp_lag_18', 'temp_lag_18', 'tnt_lag_18', 'urine_lag_18', 'wbc_lag_18', 'alb_lag_19', 'alp_lag_19', 'alt_lag_19', 'ast_lag_19', 'be_lag_19', 'bicar_lag_19', 'bili_lag_19', 'bili_dir_lag_19', 'bnd_lag_19', 'bun_lag_19', 'ca_lag_19', 'cai_lag_19', 'ck_lag_19', 'ckmb_lag_19', 'cl_lag_19', 'crea_lag_19', 'crp_lag_19', 'dbp_lag_19', 'fgn_lag_19', 'fio2_lag_19', 'glu_lag_19', 'hgb_lag_19', 'hr_lag_19', 'inr_pt_lag_19', 'k_lag_19', 'lact_lag_19', 'lymph_lag_19', 'map_lag_19', 'mch_lag_19', 'mchc_lag_19', 'mcv_lag_19', 'methb_lag_19', 'mg_lag_19', 'na_lag_19', 'neut_lag_19', 'o2sat_lag_19', 'pco2_lag_19', 'ph_lag_19', 'phos_lag_19', 'plt_lag_19', 'po2_lag_19', 'ptt_lag_19', 'resp_lag_19', 'sbp_lag_19', 'temp_lag_19', 'tnt_lag_19', 'urine_lag_19', 'wbc_lag_19', 'alb_lag_20', 'alp_lag_20', 'alt_lag_20', 'ast_lag_20', 'be_lag_20', 'bicar_lag_20', 'bili_lag_20', 'bili_dir_lag_20', 'bnd_lag_20', 'bun_lag_20', 'ca_lag_20', 'cai_lag_20', 'ck_lag_20', 'ckmb_lag_20', 'cl_lag_20', 'crea_lag_20', 'crp_lag_20', 'dbp_lag_20', 'fgn_lag_20', 'fio2_lag_20', 'glu_lag_20', 'hgb_lag_20', 'hr_lag_20', 'inr_pt_lag_20', 'k_lag_20', 'lact_lag_20', 'lymph_lag_20', 'map_lag_20', 'mch_lag_20', 'mchc_lag_20', 'mcv_lag_20', 'methb_lag_20', 'mg_lag_20', 'na_lag_20', 'neut_lag_20', 'o2sat_lag_20', 'pco2_lag_20', 'ph_lag_20', 'phos_lag_20', 'plt_lag_20', 'po2_lag_20', 'ptt_lag_20', 'resp_lag_20', 'sbp_lag_20', 'temp_lag_20', 'tnt_lag_20', 'urine_lag_20', 'wbc_lag_20', 'alb_lag_21', 'alp_lag_21', 'alt_lag_21', 'ast_lag_21', 'be_lag_21', 'bicar_lag_21', 'bili_lag_21', 'bili_dir_lag_21', 'bnd_lag_21', 'bun_lag_21', 'ca_lag_21', 'cai_lag_21', 'ck_lag_21', 'ckmb_lag_21', 'cl_lag_21', 'crea_lag_21', 'crp_lag_21', 'dbp_lag_21', 'fgn_lag_21', 'fio2_lag_21', 'glu_lag_21', 'hgb_lag_21', 'hr_lag_21', 'inr_pt_lag_21', 'k_lag_21', 'lact_lag_21', 'lymph_lag_21', 'map_lag_21', 'mch_lag_21', 'mchc_lag_21', 'mcv_lag_21', 'methb_lag_21', 'mg_lag_21', 'na_lag_21', 'neut_lag_21', 'o2sat_lag_21', 'pco2_lag_21', 'ph_lag_21', 'phos_lag_21', 'plt_lag_21', 'po2_lag_21', 'ptt_lag_21', 'resp_lag_21', 'sbp_lag_21', 'temp_lag_21', 'tnt_lag_21', 'urine_lag_21', 'wbc_lag_21', 'alb_lag_22', 'alp_lag_22', 'alt_lag_22', 'ast_lag_22', 'be_lag_22', 'bicar_lag_22', 'bili_lag_22', 'bili_dir_lag_22', 'bnd_lag_22', 'bun_lag_22', 'ca_lag_22', 'cai_lag_22', 'ck_lag_22', 'ckmb_lag_22', 'cl_lag_22', 'crea_lag_22', 'crp_lag_22', 'dbp_lag_22', 'fgn_lag_22', 'fio2_lag_22', 'glu_lag_22', 'hgb_lag_22', 'hr_lag_22', 'inr_pt_lag_22', 'k_lag_22', 'lact_lag_22', 'lymph_lag_22', 'map_lag_22', 'mch_lag_22', 'mchc_lag_22', 'mcv_lag_22', 'methb_lag_22', 'mg_lag_22', 'na_lag_22', 'neut_lag_22', 'o2sat_lag_22', 'pco2_lag_22', 'ph_lag_22', 'phos_lag_22', 'plt_lag_22', 'po2_lag_22', 'ptt_lag_22', 'resp_lag_22', 'sbp_lag_22', 'temp_lag_22', 'tnt_lag_22', 'urine_lag_22', 'wbc_lag_22', 'alb_lag_23', 'alp_lag_23', 'alt_lag_23', 'ast_lag_23', 'be_lag_23', 'bicar_lag_23', 'bili_lag_23', 'bili_dir_lag_23', 'bnd_lag_23', 'bun_lag_23', 'ca_lag_23', 'cai_lag_23', 'ck_lag_23', 'ckmb_lag_23', 'cl_lag_23', 'crea_lag_23', 'crp_lag_23', 'dbp_lag_23', 'fgn_lag_23', 'fio2_lag_23', 'glu_lag_23', 'hgb_lag_23', 'hr_lag_23', 'inr_pt_lag_23', 'k_lag_23', 'lact_lag_23', 'lymph_lag_23', 'map_lag_23', 'mch_lag_23', 'mchc_lag_23', 'mcv_lag_23', 'methb_lag_23', 'mg_lag_23', 'na_lag_23', 'neut_lag_23', 'o2sat_lag_23', 'pco2_lag_23', 'ph_lag_23', 'phos_lag_23', 'plt_lag_23', 'po2_lag_23', 'ptt_lag_23', 'resp_lag_23', 'sbp_lag_23', 'temp_lag_23', 'tnt_lag_23', 'urine_lag_23', 'wbc_lag_23', 'alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24' ]
 
 PredictionDatasetpytorch.add_relative_time_idx=False
 PredictionDatasetpytorch.target="label"
             
 PredictionDatasetpytorch.time_varying_unknown_reals=[
                 "label",
+              
             ]
 PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetpytorch.lagged_variables=["alb",
+PredictionDatasetpytorch.lagged_variables=[  "alb",
                 "alp",
                 "alt",
                 "ast",
@@ -80,4 +81,4 @@ PredictionDatasetpytorch.lagged_variables=["alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc",]
\ No newline at end of file
+                "wbc"]
\ No newline at end of file
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index 1234f707..75041763 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -15,8 +15,8 @@ train_common.weight = "balanced"
 train_common.ram_cache = True
 
 # LOSS FUNCTION
-DLPredictionWrapper.loss = @mse_loss
-DLPredictionPytorchForecastingWrapper.loss = @mse_loss
+DLPredictionWrapper.loss = @l1_loss
+DLPredictionPytorchForecastingWrapper.loss = @l1_loss
 MLWrapper.loss = @mean_squared_error
 
 # SELECTING PREPROCESSOR
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 75722831..96ba9695 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -10,7 +10,7 @@
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
-from pytorch_forecasting import TimeSeriesDataSet
+from pytorch_forecasting import TimeSeriesDataSet, GroupNormalizer
 
 
 class CommonDataset(Dataset):
@@ -463,6 +463,7 @@ def __init__(
         )  # create an incremental column indicating the time step(required by constructor)
         data = data.get(split)  # get split
         labels = data["OUTCOME"]
+
         features = data["FEATURES"]
         self.name = name
         self.data = pd.merge(labels, features, on=["stay_id", "time"])
@@ -471,7 +472,7 @@ def __init__(
                 self.data["label"] = self.data["label"].astype(float)
             columns_to_lag = lagged_variables
             grouped = self.data.sort_values("time_idx").groupby("stay_id")
-            for lag in range(max_encoder_length, max_encoder_length + max_prediction_length + 1):
+            for lag in range(1, max_encoder_length + 1):
                 for column in columns_to_lag:
                     # Create a new column with lagged values
                     self.data[f"{column}_lag_{lag}"] = grouped[column].shift(lag, fill_value=0)
@@ -501,7 +502,9 @@ def __init__(
             # add_target_scales=True,
             # add_encoder_length=True,
             predict_mode=True,
-            target_normalizer=None,
+            target_normalizer=GroupNormalizer(
+                groups=["stay_id"], transformation="relu"
+            ),
         )
 
     def get_balance(self) -> list:
diff --git a/icu_benchmarks/data/preprocessor.py b/icu_benchmarks/data/preprocessor.py
index 69b90f5b..9da25308 100644
--- a/icu_benchmarks/data/preprocessor.py
+++ b/icu_benchmarks/data/preprocessor.py
@@ -117,7 +117,8 @@ def _process_static(self, data, vars):
             sta_rec.add_step(StepScale())
 
         sta_rec.add_step(StepImputeFastZeroFill(sel=all_numeric_predictors()))
-        sta_rec.add_step(StepSklearn(SimpleImputer(missing_values=None, strategy="most_frequent"), sel=has_type("object")))
+        sta_rec.add_step(StepSklearn(SimpleImputer(missing_values=None,
+                         strategy="most_frequent"), sel=has_type("object")))
         sta_rec.add_step(StepSklearn(LabelEncoder(), sel=has_type("object"), columnwise=True))
 
         data = apply_recipe_to_splits(sta_rec, data, Segment.static, self.save_cache, self.load_cache)
@@ -205,8 +206,11 @@ def apply(self, data, vars):
         Returns:
             Preprocessed data.
         """
-        for split in [Split.train, Split.val, Split.test]:
-            data = self._process_outcome(data, vars, split)
+
+        self.outcome_max = data["train"]["OUTCOME"]["label"].max()
+        self.outcome_min = data["train"]["OUTCOME"]["label"].min()
+        # for split in [Split.train, Split.val, Split.test]:
+        #    data = self._process_outcome(data, vars, split)
 
         data = super().apply(data, vars)
         return data
@@ -284,7 +288,8 @@ def _process_dynamic_data(self, data, vars):
         if self.filter_missing_values:
             rows_to_remove = data[Segment.dynamic][vars[Segment.dynamic]].isna().sum(axis=1) != 0
             ids_to_remove = data[Segment.dynamic].loc[rows_to_remove][vars["GROUP"]].unique()
-            data = {table_name: table.loc[~table[vars["GROUP"]].isin(ids_to_remove)] for table_name, table in data.items()}
+            data = {table_name: table.loc[~table[vars["GROUP"]].isin(
+                ids_to_remove)] for table_name, table in data.items()}
             logging.info(f"Removed {len(ids_to_remove)} stays with missing values.")
         return data
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index ad73a479..57107b60 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -14,7 +14,7 @@
 from icu_benchmarks.models.wrappers import DLPredictionWrapper, DLPredictionPytorchForecastingWrapper
 from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, is_tensor
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
-from pytorch_forecasting.metrics import QuantileLoss
+from pytorch_forecasting.metrics import QuantileLoss, MAE
 import matplotlib.pyplot as plt
 
 
@@ -510,13 +510,6 @@ def forward(
         pred = self.logit(out["prediction"])
         return pred
 
-    """
-    def forward(self, x):
-        out = self.model(x)
-        pred = self.logit(out["prediction"])
-        return pred
-    """
-
     def actual_vs_predictions_plot(self, dataloader):
         predictions = self.model.predict(dataloader, return_x=True)
         predictions_vs_actuals = self.model.calculate_prediction_actual_by_variable(predictions.x, predictions.output)
@@ -525,7 +518,7 @@ def actual_vs_predictions_plot(self, dataloader):
 
     def interpertations(self, dataloader, log_dir=".", plot=False):
         raw_predictions = self.model.predict(dataloader, return_x=True, mode="raw")
-        interpretation = self.model.interpret_output(raw_predictions.output, reduction="sum")
+        interpretation = self.model.interpret_output(raw_predictions.output, reduction="mean")
         if plot:
             figs = self.model.plot_interpretation(interpretation)
             for key, fig in figs.items():
@@ -700,7 +693,7 @@ def forward(
         if self.num_classes == 1:
             x_dict["encoder_cont"][:, :, -1] = 0.0
         out = self.model(x_dict)
-        out = out["prediction"][0]
 
-        pred = self.logit(out)
+        pred = self.logit(out["prediction"])
+
         return pred
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 18341a99..4a18deff 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -260,44 +260,37 @@ def train_common(
         return 0
 
     if explain:
+        # local_index = np.random.randint(low=0, high=batch_size, size=1)
         attributions_dict = {}
-        # Interpertations = model.interpertations(test_loader, log_dir, plot=True)
-        # attributions_dict["attention_weights"] = Interpertations["attention"].tolist()
-        # attributions_dict["static_variables"] = Interpertations["static_variables"].tolist()
-        # attributions_dict["encoder_variables"] = Interpertations["encoder_variables"].tolist()
-        # print("attention", Interpertations)
-        # model.train()
+
+        model.train()
         XAI_dict = {}
         # if XAI_metric:
 
         # random attribution per timestep
-        # random_attributions_ts = np.random.normal(size=24)
-        # F_baseline_ts = model.Faithfulness_Correlation(
-        #    test_loader, random_attributions_ts, pertrub='Noise', subset_size=4, time_step=True, nr_runs=100)
-        # print('Random normal faithfulness correlation for timesteps', F_baseline_ts)
-
-        # F_attention = model.Faithfulness_Correlation(
-        #    test_loader, Interpertations["attention"], pertrub='Noise', subset_size=4, time_step=True, nr_runs=100)
-        # print('Attention faithfulness correlation', F_attention)
+        random_attributions_ts = np.random.normal(size=24)
+        F_baseline_ts = model.Faithfulness_Correlation(
+            test_loader, random_attributions_ts, pertrub='baseline', subset_size=4, time_step=True, nr_runs=10)
+        print('Random normal faithfulness correlation for timesteps', F_baseline_ts)
+
         # random attribution per variable per timestep
-        # random_attributions_v_ts = np.random.normal(size=[24, 53])
-        # F_baseline_v_ts = model.Faithfulness_Correlation(
-        #    test_loader, random_attributions_v_ts, pertrub='Noise', feature_timestep=True, nr_runs=10, subset_size=[4, 9])
-        # print('Random normal faithfulness correlation for variables per timesteps', F_baseline_v_ts)
-
-        # XAI_dict["attention_faith"] = F_attention.tolist()
-        # XAI_dict["random_faith_timestep"] = random_attributions_ts.tolist()
-        # XAI_dict["random_faith_var_timestep"] = random_attributions_v_ts.tolist()
+        random_attributions_v_ts = np.random.normal(size=[24, 53])
+        F_baseline_v_ts = model.Faithfulness_Correlation(
+            test_loader, random_attributions_v_ts, pertrub='baseline', feature_timestep=True, nr_runs=10, subset_size=[4, 9])
+        print('Random normal faithfulness correlation for variables per timesteps', F_baseline_v_ts)
+
+        XAI_dict["random_faith_timestep"] = random_attributions_ts.tolist()
+        XAI_dict["random_faith_var_timestep"] = random_attributions_v_ts.tolist()
         methods = {
-            # "Saliency": Saliency,
-            #  "Lime": Lime,
+            "Saliency": Saliency,
+            # "Lime": Lime,
             # "IG": IntegratedGradients,
-            "FA": FeatureAblation
+            # "FA": FeatureAblation
         }
         for key, item in methods.items():
             if key == "IG":
                 all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
-                    test_loader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=50
+                    test_loader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=10
                 )
             if key == "Lime":
                 all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
@@ -310,22 +303,33 @@ def train_common(
             attributions_dict["{}_all".format(key)] = all_attrs.tolist()
             attributions_dict["{}_timesteps".format(key)] = timestep_attrs.tolist()
             attributions_dict["{}_features".format(key)] = features_attrs.tolist()
-
-            print("{}".format(key), all_attrs, features_attrs, timestep_attrs)
+            print(np.shape(all_attrs), 'shape of attr all shoulld be 24x53')
+            print(np.shape(timestep_attrs), 'shape of attr ts shoulld be 24')
+            print(np.shape(features_attrs), 'shape of attr var shoulld be 53')
+            # print("{}".format(key), all_attrs, features_attrs, timestep_attrs)
             if XAI_metric:
                 faithfulness_timesteps = model.Faithfulness_Correlation(
-                    test_loader, timestep_attrs, pertrub="Noise", time_step=True, subset_size=4, nr_runs=100
+                    test_loader, timestep_attrs, pertrub="baseline", time_step=True, subset_size=4, nr_runs=10
                 )
-                print("Attributions faithfulness timesteps correlation", faithfulness_timesteps)
+                print("{} Attributions faithfulness timesteps correlation".format(key), faithfulness_timesteps)
                 XAI_dict["{}_faith_timesteps".format(key)] = faithfulness_timesteps.tolist()
                 random_attributions = np.random.normal(np.shape(all_attrs))
 
                 faithfulness_timesteps_variable = model.Faithfulness_Correlation(
-                    test_loader, all_attrs, pertrub="Noise", feature_timestep=True, subset_size=[4, 9], nr_runs=100
+                    test_loader, all_attrs, pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=10
                 )
-                print("Attributions faithfulness variable per timestep correlation", faithfulness_timesteps_variable)
+                print("{}_Attributions faithfulness variable per timestep correlation".format(
+                    key), faithfulness_timesteps_variable)
                 XAI_dict["{}_faith_variable_per_timestep".format(key)] = faithfulness_timesteps_variable.tolist()
-
+        Interpertations = model.interpertations(test_loader, log_dir, plot=True)
+        attributions_dict["attention_weights"] = Interpertations["attention"].tolist()
+        attributions_dict["static_variables"] = Interpertations["static_variables"].tolist()
+        attributions_dict["encoder_variables"] = Interpertations["encoder_variables"].tolist()
+        print("attention", Interpertations)
+        F_attention = model.Faithfulness_Correlation(
+            test_loader, Interpertations["attention"], pertrub='baseline', subset_size=4, time_step=True, nr_runs=10)
+        print('Attention faithfulness correlation', F_attention)
+        XAI_dict["attention_faith"] = F_attention.tolist()
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/Attributions.json"
 
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 6385b677..dacfbfb3 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -4,7 +4,7 @@
 import torchmetrics
 from sklearn.metrics import log_loss, mean_squared_error
 import torch
-from torch.nn import MSELoss, CrossEntropyLoss
+from torch.nn import MSELoss, CrossEntropyLoss, L1Loss
 import torch.nn as nn
 from torch import Tensor, FloatTensor
 from torch.optim import Optimizer, Adam
@@ -19,13 +19,13 @@
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
 import matplotlib.pyplot as plt
-from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean
+from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson
 import captum
 
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.mse_loss, module="torch.nn.functional")
-
+gin.config.external_configurable(nn.functional.l1_loss, module="torch.nn.functional")
 gin.config.external_configurable(mean_squared_error, module="sklearn.metrics")
 gin.config.external_configurable(log_loss, module="sklearn.metrics")
 
@@ -160,7 +160,8 @@ def finalize_step(self, step_prefix=""):
         try:
             for name, metric in self.metrics[step_prefix].items():
                 try:
-                    value = np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
+                    value = np.float32(metric.compute()) if isinstance(
+                        metric.compute(), np.float64) else metric.compute()
                     self.log_dict({f"{step_prefix}/{name}": value}, sync_dist=True)
 
                 except (NotComputableError, ValueError) as e:
@@ -431,21 +432,8 @@ def step_fn(self, element, step_prefix=""):
 
         if isinstance(labels, list):
             labels = labels[-1]
-        data = (
-            dic["encoder_cat"],
-            dic["encoder_cont"],
-            dic["encoder_target"],
-            dic["encoder_lengths"],
-            dic["decoder_cat"],
-            dic["decoder_cont"],
-            dic["decoder_target"],
-            dic["decoder_lengths"],
-            dic["decoder_time_idx"],
-            dic["groups"],
-            dic["target_scale"],
-        )
 
-        mask = torch.ones_like(labels).bool()
+        data = self.prep_data(dic)
 
         out = self(data)
 
@@ -456,15 +444,17 @@ def step_fn(self, element, step_prefix=""):
             aux_loss = 0
         # Get prediction and target
 
-        prediction = torch.masked_select(out, mask.unsqueeze(-1)).reshape(-1, out.shape[-1]).to(self.device)
+        prediction = out.to(self.device).squeeze(-1)
+
+        target = labels.to(self.device)
 
-        target = torch.masked_select(labels, mask).to(self.device)
         if prediction.shape[-1] > 1 and self.run_mode == RunMode.classification:
             # Classification task
             loss = self.loss(prediction, target.long(), weight=self.loss_weights.to(self.device)) + aux_loss
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
+
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
             raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
@@ -487,56 +477,70 @@ def step_fn(self, element, step_prefix=""):
         self.log(f"{step_prefix}/loss", loss, on_step=False, on_epoch=True, sync_dist=True)
         return loss
 
-    def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kwargs):
+    def prep_data_captum(self, x):
+        data = (
+            x["encoder_cat"].float().requires_grad_(),
+            x["encoder_cont"].requires_grad_(),
+            x["encoder_target"].float().requires_grad_(),
+            x["encoder_lengths"].float().requires_grad_(),
+            x["decoder_cat"].float().requires_grad_(),
+            x["decoder_cont"].requires_grad_(),
+            x["decoder_target"].float().requires_grad_(),
+            x["decoder_lengths"].float().requires_grad_(),
+            x["decoder_time_idx"].float().requires_grad_(),
+            x["groups"].float().requires_grad_(),
+            x["target_scale"].requires_grad_(),
+        )
+        baselines = (
+            data[0].to(self.device),  # encoder_cat, no cat variables
+            torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to random
+            torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to random
+            data[3].to(self.device),  # encoder_lengths, leave unchanged
+            data[4].to(self.device),  # decoder_cat, no cat variables
+            torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to random
+            torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to random
+            data[7].to(self.device),  # decoder_lengths, leave unchanged
+            data[8].to(self.device),  # decoder_time_idx, unchanged
+            data[9].to(self.device),  # groups, leave unchanged
+            data[10].to(self.device),  # target_scale, leave unchanged
+        )
+        return data, baselines
+
+    def explantation_captum(self, test_loader, method, local_index=None, log_dir=".", plot=False, local=False, **kwargs):
         # Initialize lists to store attribution values for all instances
         all_attrs = []
+        score = []
+        r_score = []
         method_name = method.__name__
         # Loop through the test_loader to compute attributions for all instances
         for batch in test_loader:
+
             for key, value in batch[0].items():
                 batch[0][key] = batch[0][key].to(self.device)
             x = batch[0]
 
-            data = (
-                x["encoder_cat"].float().requires_grad_(),
-                x["encoder_cont"].requires_grad_(),
-                x["encoder_target"].float().requires_grad_(),
-                x["encoder_lengths"].float().requires_grad_(),
-                x["decoder_cat"].float().requires_grad_(),
-                x["decoder_cont"].requires_grad_(),
-                x["decoder_target"].float().requires_grad_(),
-                x["decoder_lengths"].float().requires_grad_(),
-                x["decoder_time_idx"].float().requires_grad_(),
-                x["groups"].float().requires_grad_(),
-                x["target_scale"].requires_grad_(),
-            )
-            baselines = (
-                data[0].to(self.device),  # encoder_cat, no cat variables
-                torch.randn_like(data[1]).to(self.device),  # encoder_cont, set to random
-                torch.randn_like(data[2]).to(self.device),  # encoder_target, set to random
-                data[3].to(self.device),  # encoder_lengths, leave unchanged
-                data[4].to(self.device),  # decoder_cat, no cat variables
-                torch.randn_like(data[5]).to(self.device),  # decoder_cont, set to random
-                torch.randn_like(data[6]).to(self.device),  # decoder_target, set to random
-                data[7].to(self.device),  # decoder_lengths, leave unchanged
-                data[8].to(self.device),  # decoder_time_idx, unchanged
-                data[9].to(self.device),  # groups, leave unchanged
-                data[10].to(self.device),  # target_scale, leave unchanged
-            )
+            data, baselines = self.prep_data_captum(x)
 
             explantation = method(self.forward_captum)
-            # Reformat attributions.
-            # attr_all_timesteps = []
-            # for time_step in range(0, 24):
+
             if method is not captum.attr.Saliency:
                 attr = explantation.attribute(data, baselines=baselines, **kwargs)
             else:
                 attr = explantation.attribute(data, **kwargs)
-            print(attr)
 
-            print(attr[0].shape)
             # Convert attributions to numpy array and append to the list
-            all_attrs.append(attr[0].cpu().detach().numpy())
+            print(attr.shape, 'shape out of captum should be 64x24x53')
+            stacked_attr = torch.stack(attr).cpu().detach().numpy()
+
+            # Compute the mean along the specified axis (axis=0) and keep it as a numpy array
+            """ score.append(self.Faithfulness_Correlation2(x, stacked_attr,
+                                                        pertrub="Noise", feature_timestep=True, subset_size=[4, 9], nr_runs=10))
+            r_score.append(self.Faithfulness_Correlation2(x, np.random.normal(
+                size=[64, 24, 53]), pertrub="Noise", feature_timestep=True, subset_size=[4, 9], nr_runs=10)) """
+            print(np.shape(stacked_attr), 'shape after stacking')
+            attr = np.mean(stacked_attr, axis=0)
+            print(np.shape(attr), 'shape after averaging over batch should 24x53')
+            all_attrs.append(attr)
 
         all_attrs = np.array(all_attrs).mean(axis=(0))
 
@@ -561,7 +565,8 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             plt.xlabel("Time Step")
             plt.ylabel("{} Attribution".format(method_name))
             plt.title("{} Attribution Values".format(method_name))
-            plt.xticks(x_values)  # Set x-ticks to match the number of features
+            plt.xticks(x_values, ['height', 'weight', 'age', 'sex', 'time_idx', 'alb', 'alp', 'alt', 'ast', 'be', 'bicar', 'bili', 'bili_dir', 'bnd', 'bun', 'ca', 'cai', 'ck', 'ckmb', 'cl', 'crea', 'crp', 'dbp', 'fgn', 'fio2', 'glu', 'hgb', 'hr', 'inr_pt', 'k', 'lact', 'lymph', 'map', 'mch', 'mchc', 'mcv', 'methb', 'mg', 'na', 'neut', 'o2sat', 'pco2', 'ph', 'phos', 'plt', 'po2', 'ptt', 'resp', 'sbp', 'temp', 'tnt', 'urine', 'wbc'], rotation=90
+                       )  # Set x-ticks to match the number of features
             plt.tight_layout()
             plt.savefig(log_dir / "{}_attribution_features_plot.png".format(method_name), bbox_inches="tight")
             plt.figure(figsize=(8, 6))
@@ -583,10 +588,27 @@ def explantation_captum(self, test_loader, method, log_dir=".", plot=False, **kw
             plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
         return all_attrs, features_attrs, timestep_attrs
 
+    def prep_data(self, x):
+        data = (
+            x["encoder_cat"],
+            x["encoder_cont"],
+            x["encoder_target"],
+            x["encoder_lengths"],
+            x["decoder_cat"],
+            x["decoder_cont"],
+            x["decoder_target"],
+            x["decoder_lengths"],
+            x["decoder_time_idx"],
+            x["groups"],
+            x["target_scale"],
+        )
+        return data
+
     def Faithfulness_Correlation(
         self,
         test_loader,
         attribution,
+        local_index=None,
         similarity_func=None,
         nr_runs=100,
         pertrub=None,
@@ -594,6 +616,7 @@ def Faithfulness_Correlation(
         feature=False,
         time_step=False,
         feature_timestep=False,
+        local=False,
     ):
         """
         Implementation of faithfulness correlation by Bhatt et al., 2020.
@@ -617,93 +640,197 @@ def Faithfulness_Correlation(
             explanations." IJCAI (2020): 3016-3022.
             2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
         """
-
-        if torch.is_tensor(attribution):
-            # Convert the tensor to a NumPy array
-            attribution = attribution.cpu().detach().numpy()
+        def add_noise(x, indices, time_step, feature_timestep):
+            noise = torch.randn_like(x["encoder_cont"])
+            # In-place modification to save memory
+            if time_step:
+                x["encoder_cont"][:, indices, :] += noise[:, indices, :]
+            elif feature_timestep:
+                x["encoder_cont"][:, indices[0], :][:, :, indices[1]] += noise[:, indices[0], :][:, :, indices[1]]
+
+        def apply_baseline(x, indices, time_step, feature_timestep):
+            mask = torch.ones_like(x["encoder_cont"])
+            if time_step:
+
+                mask[:, indices, :] -= mask[:, indices, :]
+            elif feature_timestep:
+                mask[:, indices[0], :][:, :, indices[1]] -= mask[:, indices[0], :][:, :, indices[1]]
+            # Apply mask in-place
+            x["encoder_cont"] *= mask
+
+        # Assuming 'attribution' is already a GPU tensor
+        if not torch.is_tensor(attribution):
+            attribution = torch.tensor(attribution).to(self.device)
+
+        # Other initializations
         if similarity_func is None:
-            similarity_func = correlation_spearman
+            similarity_func = correlation_pearson  # Ensure this function is compatible with GPU tensors
         if pertrub is None:
             pertrub = "baseline"
         similarities = []
-        for batch in test_loader:
-            for key, value in batch[0].items():
-                batch[0][key] = batch[0][key].to(self.device)
+        if local:
+            for key, value in test_loader[0].items():
+                batch[0][key] = value.to(self.device)
             x = batch[0]
-            data = (
-                x["encoder_cat"],
-                x["encoder_cont"],
-                x["encoder_target"],
-                x["encoder_lengths"],
-                x["decoder_cat"],
-                x["decoder_cont"],
-                x["decoder_target"],
-                x["decoder_lengths"],
-                x["decoder_time_idx"],
-                x["groups"],
-                x["target_scale"],
-            )
-            y_pred = self(data).detach().cpu().numpy()
-            pred_deltas = []
-            att_sums = []
-
-            def add_noise(x, indices, time_step, feature_timestep):
-                noise = torch.randn_like(x["encoder_cont"])
-                if time_step:
-                    x["encoder_cont"][:, indices, :] += noise[:, indices, :]
-                elif feature_timestep:
-                    x["encoder_cont"][:, indices[0], :][:, :, indices[1]] += noise[:, indices[0], :][:, :, indices[1]]
-
-            def apply_baseline(x, indices, time_step, feature_timestep):
-                mask = torch.ones_like(x["encoder_cont"]).cpu()
-                if time_step:
-                    mask[:, indices, :] = 0
-                elif feature_timestep:
-                    mask[:, indices[0], :][:, :, indices[1]] = 0
-                mask = mask.to(x["encoder_cont"].device)
-                x["encoder_cont"] *= mask
 
-            for i_ix in range(nr_runs):
-                # Randomly mask by subset size.
+        else:
 
-                if time_step:
-                    a_ix = np.random.choice(24, subset_size, replace=False)
-                elif feature_timestep:
-                    timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
-                    variables_idx = np.random.choice(53, subset_size[1], replace=False)
-                    a_ix = [timesteps_idx, variables_idx]
+            for batch in test_loader:
+                # Move batch to GPU
+                for key, value in batch[0].items():
+                    batch[0][key] = value.to(self.device)
+                x = batch[0]
+                # Assuming this is a method to prepare your data
+
+                y_pred = self((self.prep_data(x))).detach()  # Keep on GPU
+                pred_deltas = []
+                att_sums = []
+
+                for i_ix in range(nr_runs):
+                    if time_step:
+                        a_ix = np.random.choice(24, subset_size, replace=False)
+                    elif feature_timestep:
+                        timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
+                        variables_idx = np.random.choice(53, subset_size[1], replace=False)
+                        a_ix = [timesteps_idx, variables_idx]
+
+                    # Apply perturbation
+                    if pertrub == "Noise":
+                        add_noise(x, a_ix, time_step, feature_timestep)
+                    elif pertrub == "baseline":
+                        apply_baseline(x, a_ix, time_step, feature_timestep)
+
+                    # Predict on perturbed input and calculate deltas
+                    y_pred_perturb = (self(self.prep_data(x))).detach()  # Keep on GPU
+                    # Change dims as per your requirement
+                    pred_deltas.append((y_pred - y_pred_perturb).mean(dim=[0, 1, 2]))
+
+                    # Sum attributions of the random subset and keep on GPU
+                    if time_step:
+                        att_sums.append(attribution[a_ix].sum())
+                    elif feature_timestep:
+                        row_indices = a_ix[0]
+                        col_indices = a_ix[1]
+                        att_sums.append(attribution[row_indices][:, col_indices].sum())
+
+                # Convert to CPU for numpy operations
+                pred_deltas_cpu = torch.stack(pred_deltas).cpu().numpy()
+                att_sums_cpu = torch.tensor(att_sums).cpu().numpy()
+                similarities.append(similarity_func(pred_deltas_cpu, att_sums_cpu))
 
-                if pertrub == "Noise":
-                    add_noise(x, a_ix, time_step, feature_timestep)
-                elif pertrub == "baseline":
-                    apply_baseline(x, a_ix, time_step, feature_timestep)
+        score = np.nanmean(similarities)
+        return score
 
-                data = (
-                    x["encoder_cat"],
-                    x["encoder_cont"],
-                    x["encoder_target"],
-                    x["encoder_lengths"],
-                    x["decoder_cat"],
-                    x["decoder_cont"],
-                    x["decoder_target"],
-                    x["decoder_lengths"],
-                    x["decoder_time_idx"],
-                    x["groups"],
-                    x["target_scale"],
-                )
-                # Predict on perturbed input x.
-                y_pred_perturb = self(data).detach().cpu().numpy()
-                pred_deltas.append((y_pred - y_pred_perturb).mean())
-
-                # Sum attributions of the random subset.
-                if time_step:
-                    att_sums.append(np.sum(attribution[a_ix]))
-                elif feature_timestep:
-                    row_indices = a_ix[0]
-                    col_indices = a_ix[1]
-                    att_sums.append(np.sum(attribution[np.ix_(row_indices, col_indices)]))
-
-            similarities.append(similarity_func(pred_deltas, att_sums))
+    def Faithfulness_Correlation2(
+        self,
+        x,
+        attribution,
+        local_index=None,
+        similarity_func=None,
+        nr_runs=100,
+        pertrub=None,
+        subset_size=3,
+        feature=False,
+        time_step=False,
+        feature_timestep=False,
+        local=False,
+    ):
+        """
+        Implementation of faithfulness correlation by Bhatt et al., 2020.
+
+        The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
+        (or 'fidelity') with respect to the model behaviour.
+
+        Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
+        the average explanation attribution for only the subset of features are (linearly) correlated, taking the
+        average over multiple runs and test samples. The metric returns one float per input-attribution pair that
+        ranges between -1 and 1, where higher scores are better.
+
+        For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
+        or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
+        test point and the average explanation attribution for only the subset of features is calculated. Results is
+        average over multiple runs and several test samples.
+        This code is adapted from the quantus libray to suit our use case
+
+        References:
+            1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
+            explanations." IJCAI (2020): 3016-3022.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+        """
+        def add_noise(x, indices, time_step, feature_timestep):
+            noise = torch.randn_like(x["encoder_cont"])
+            # In-place modification to save memory
+            if time_step:
+                x["encoder_cont"][indices[0], :, :][:, indices[1], :] += noise[indices[0], :, :][:, indices[1], :]
+            elif feature_timestep:
+                x["encoder_cont"][indices[0], :, :][:, indices[1], :][:, :, indices[2]
+                                                                      ] += noise[indices[0], :, :][:, indices[1], :][:, :, indices[2]]
+
+        def apply_baseline(x, indices, time_step, feature_timestep):
+            mask = torch.ones_like(x["encoder_cont"])
+            if time_step:
+                mask[indices[0], :, :][:, indices[1], :] = 0
+            elif feature_timestep:
+                mask[indices[0], :, :][:, indices[1], :][:, :, indices[2]
+                                                         ] = 0
+            with torch.no_grad():
+                x["encoder_cont"] *= mask
+        # Assuming 'attribution' is already a GPU tensor
+        if not torch.is_tensor(attribution):
+            attribution = torch.tensor(attribution).to(self.device)
+
+        # Other initializations
+        if similarity_func is None:
+            similarity_func = correlation_pearson  # Ensure this function is compatible with GPU tensors
+        if pertrub is None:
+            pertrub = "baseline"
+        similarities = []
+
+        # Assuming this is a method to prepare your data
+
+        y_pred = self(self.prep_data(x)).detach()  # Keep on GPU
+        pred_deltas = []
+        att_sums = []
+
+        for i_ix in range(nr_runs):
+            if time_step:
+                timesteps_idx = np.random.choice(24, subset_size, replace=False)
+                patient_idx = np.random.choice(64, subset_size, replace=False)
+                a_ix = [patient_idx, timesteps_idx]
+            elif feature_timestep:
+                timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
+                variables_idx = np.random.choice(53, subset_size[1], replace=False)
+                patient_idx = np.random.choice(64, 1, replace=False)
+                a_ix = [patient_idx, timesteps_idx, variables_idx]
+
+            # Apply perturbation
+            if pertrub == "Noise":
+                add_noise(x, a_ix, time_step, feature_timestep)
+            elif pertrub == "baseline":
+                apply_baseline(x, a_ix, time_step, feature_timestep)
+
+            # Predict on perturbed input and calculate deltas
+            y_pred_perturb = (self(self.prep_data(x))).detach()  # Keep on GPU
+            pred_deltas.append((y_pred - y_pred_perturb).mean(dim=[0, 1, 2]))  # Change dims as per your requirement
+
+            # Sum attributions of the random subset and keep on GPU
+            if time_step:
+                patient_indices = a_ix[0]
+                timestep_indices = a_ix[1]
+                att_sums.append(attribution[patient_indices, :, :][:, timestep_indices, :].sum())
+
+            elif feature_timestep:
+                patient_indices = a_ix[0]
+                timestep_indices = a_ix[1]
+                variable_indices = a_ix[2]
+                att_sums.append(attribution[patient_indices, :, :]
+                                [:, timestep_indices, :][:, :, variable_indices].sum())
+
+            # Convert to CPU for numpy operations
+
+        pred_deltas_cpu = torch.stack(pred_deltas).cpu().numpy()
+        att_sums_cpu = torch.tensor(att_sums).cpu().numpy()
+        similarities.append(similarity_func(pred_deltas_cpu, att_sums_cpu))
 
         score = np.nanmean(similarities)
         return score

From ab115fdee0b4f2a685ac219978fb83c0a0b2e67c Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 28 Nov 2023 18:16:13 +0100
Subject: [PATCH 114/142] cleaned and fixed aggreagtion along batch

---
 configs/prediction_models/DeepARpytorch.gin   |   6 +-
 configs/prediction_models/RNNpytorch.gin      |   8 +-
 configs/prediction_models/TFTpytorch.gin      |  15 +-
 configs/prediction_models/common/DLCommon.gin |   2 +
 configs/prediction_models/common/DLTuning.gin |   2 +-
 icu_benchmarks/models/custom_metrics.py       |   2 +
 icu_benchmarks/models/metrics.py              |   5 +
 icu_benchmarks/models/train.py                |  99 ++--
 icu_benchmarks/models/wrappers.py             | 458 ++++++++++--------
 icu_benchmarks/run_utils.py                   |   8 +-
 10 files changed, 320 insertions(+), 285 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 1b7a9ca0..9a61ca84 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -8,17 +8,17 @@ train_common.model = @DeepARpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-
+optimizer/hyperparameter.lr = (0.0001, 0.001,0.01)
 
 # Model params
 
 model/hyperparameter.class_to_tune = @DeepARpytorch
-model/hyperparameter.hidden = (4, 64, "log-uniform", 2)
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
 model/hyperparameter.dropout = (0.0, 0.4)
-model/hyperparameter.lr_scheduler = "exponential"
+#model/hyperparameter.lr_scheduler = "exponential"
 
 
 # Dataset params
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 3969df0f..aa2398cb 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -8,17 +8,17 @@ train_common.model = @RNNpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-
+optimizer/hyperparameter.lr = (0.0001, 0.001,0.01)
 
 # Model params
 
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = (4, 64, "log-uniform", 2)
+model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1,3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
-model/hyperparameter.dropout = (0.0, 0.4)
-model/hyperparameter.lr_scheduler = "exponential"
+model/hyperparameter.dropout = (0.0, 0.9)
+#model/hyperparameter.lr_scheduler = "exponential"
 
 
 # Dataset params
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 94a42740..4f2d078e 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -8,16 +8,21 @@ train_common.model = @TFTpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
+optimizer/hyperparameter.lr = (0.0001, 0.001,0.01)
 
 
 # Model params
 model/hyperparameter.class_to_tune = @TFTpytorch
-model/hyperparameter.hidden = (16,32,64)
+model/hyperparameter.hidden = (10,20,40,80,160,240,320)
 model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.dropout = (0.0, 0.4)
-model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =(1,2,4)
-model/hyperparameter.lr_scheduler = "exponential"
+model/hyperparameter.dropout =(0.1, 0.9)
+model/hyperparameter.dropout_att = (0.1, 0.9)
+model/hyperparameter.n_heads =(1,4)
+
+
+
+
+
 
 
 # Dataset params
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index fa4b440f..2760e187 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -19,5 +19,7 @@ train_common.batch_size = 64
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
+
+
 # Hyperparameter tuning settings
 include "configs/prediction_models/common/DLTuning.gin"
\ No newline at end of file
diff --git a/configs/prediction_models/common/DLTuning.gin b/configs/prediction_models/common/DLTuning.gin
index ec0240e2..0c56677c 100644
--- a/configs/prediction_models/common/DLTuning.gin
+++ b/configs/prediction_models/common/DLTuning.gin
@@ -1,5 +1,5 @@
 # Hyperparameter tuner settings for Deep Learning.
 tune_hyperparameters.scopes = ["model", "optimizer"]
 tune_hyperparameters.n_initial_points = 5
-tune_hyperparameters.n_calls = 15
+tune_hyperparameters.n_calls = 20
 tune_hyperparameters.folds_to_tune_on = 2
\ No newline at end of file
diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index ddb5d37e..d1dbc271 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -67,8 +67,10 @@ def __init__(
         )
 
         def mae_with_invert_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor, invert_fn=Callable) -> float:
+
             y_true = invert_fn(y_targets.numpy().reshape(-1, 1))[:, 0]
             y_pred = invert_fn(y_preds.numpy().reshape(-1, 1))[:, 0]
+
             return mean_absolute_error(y_true, y_pred)
 
 
diff --git a/icu_benchmarks/models/metrics.py b/icu_benchmarks/models/metrics.py
index 5428db61..7301d8d9 100644
--- a/icu_benchmarks/models/metrics.py
+++ b/icu_benchmarks/models/metrics.py
@@ -45,8 +45,13 @@ def ece_curve_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor) -> floa
 
 
 def mae_with_invert_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor, invert_fn=Callable) -> float:
+    print('y_true', y_true)
+    print('y_pred', y_pred)
     y_true = invert_fn(y_targets.numpy().reshape(-1, 1))[:, 0]
     y_pred = invert_fn(y_preds.numpy().reshape(-1, 1))[:, 0]
+    print('y_true', y_true)
+    print('y_pred', y_pred)
+    print('mae', mean_absolute_error(y_true, y_pred))
     return mean_absolute_error(y_true, y_pred)
 
 
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 4a18deff..1c739b90 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -24,11 +24,12 @@
     ImputationDataset,
     PredictionDatasetpytorch,
 )
+
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
 from collections import OrderedDict
-from captum.attr import IntegratedGradients, ShapleyValueSampling, Saliency, GuidedBackprop, LRP, FeatureAblation, Lime
+from captum.attr import IntegratedGradients, Saliency, FeatureAblation, Lime
 
 
 cpu_core_count = len(os.sched_getaffinity(0)) if hasattr(os, "sched_getaffinity") else os.cpu_count()
@@ -57,6 +58,7 @@ def train_common(
     epochs=1000,
     patience=20,
     min_delta=1e-5,
+    gradient_clip_val=0,
     test_on: str = Split.test,
     dataset_names=None,
     use_wandb: bool = False,
@@ -242,6 +244,7 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
+
     )
     if not eval_only:
         if model.requires_backprop:
@@ -260,82 +263,50 @@ def train_common(
         return 0
 
     if explain:
-        # local_index = np.random.randint(low=0, high=batch_size, size=1)
-        attributions_dict = {}
-
-        model.train()
-        XAI_dict = {}
-        # if XAI_metric:
-
-        # random attribution per timestep
-        random_attributions_ts = np.random.normal(size=24)
-        F_baseline_ts = model.Faithfulness_Correlation(
-            test_loader, random_attributions_ts, pertrub='baseline', subset_size=4, time_step=True, nr_runs=10)
-        print('Random normal faithfulness correlation for timesteps', F_baseline_ts)
-
-        # random attribution per variable per timestep
-        random_attributions_v_ts = np.random.normal(size=[24, 53])
-        F_baseline_v_ts = model.Faithfulness_Correlation(
-            test_loader, random_attributions_v_ts, pertrub='baseline', feature_timestep=True, nr_runs=10, subset_size=[4, 9])
-        print('Random normal faithfulness correlation for variables per timesteps', F_baseline_v_ts)
-
-        XAI_dict["random_faith_timestep"] = random_attributions_ts.tolist()
-        XAI_dict["random_faith_var_timestep"] = random_attributions_v_ts.tolist()
+
+        XAI_dict = {}  # dictrionary to log attributions metrics
+
+        # choose which  methods to get attributions
         methods = {
             "Saliency": Saliency,
             # "Lime": Lime,
-            # "IG": IntegratedGradients,
-            # "FA": FeatureAblation
+            "IG": IntegratedGradients,
+            # "FA": FeatureAblation,
+            "Random": "Random",
+            "Attention": "Attention"
         }
         for key, item in methods.items():
+            # If conditions needed here as different explantations require different inputs
             if key == "IG":
-                all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
-                    test_loader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=10
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score = model.explantation(
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=50, XAI_metric=XAI_metric,
                 )
-            if key == "Lime":
-                all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
-                    test_loader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True
+            elif key == "Lime":
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score = model.explantation(
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True, XAI_metric=XAI_metric,
                 )
             else:
-                all_attrs, features_attrs, timestep_attrs = model.explantation_captum(
-                    test_loader=test_loader, method=item, log_dir=log_dir, plot=True
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score = model.explantation(
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, XAI_metric=XAI_metric,
                 )
-            attributions_dict["{}_all".format(key)] = all_attrs.tolist()
-            attributions_dict["{}_timesteps".format(key)] = timestep_attrs.tolist()
-            attributions_dict["{}_features".format(key)] = features_attrs.tolist()
-            print(np.shape(all_attrs), 'shape of attr all shoulld be 24x53')
-            print(np.shape(timestep_attrs), 'shape of attr ts shoulld be 24')
-            print(np.shape(features_attrs), 'shape of attr var shoulld be 53')
-            # print("{}".format(key), all_attrs, features_attrs, timestep_attrs)
+
             if XAI_metric:
-                faithfulness_timesteps = model.Faithfulness_Correlation(
-                    test_loader, timestep_attrs, pertrub="baseline", time_step=True, subset_size=4, nr_runs=10
-                )
-                print("{} Attributions faithfulness timesteps correlation".format(key), faithfulness_timesteps)
-                XAI_dict["{}_faith_timesteps".format(key)] = faithfulness_timesteps.tolist()
-                random_attributions = np.random.normal(np.shape(all_attrs))
+                # logging metric scores
+                print("{} Attributions faithfulness timesteps ".format(key), ts_score)
+                XAI_dict["{}_faith_timesteps".format(key)] = ts_score
+                if key == "Attention":
+                    print("Variable selection weights faithfulness featrues ".format(key), v_score)
+                    XAI_dict["Variable_selection_weights_faith_features".format(key)] = v_score
 
-                faithfulness_timesteps_variable = model.Faithfulness_Correlation(
-                    test_loader, all_attrs, pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=10
-                )
-                print("{}_Attributions faithfulness variable per timestep correlation".format(
-                    key), faithfulness_timesteps_variable)
-                XAI_dict["{}_faith_variable_per_timestep".format(key)] = faithfulness_timesteps_variable.tolist()
-        Interpertations = model.interpertations(test_loader, log_dir, plot=True)
-        attributions_dict["attention_weights"] = Interpertations["attention"].tolist()
-        attributions_dict["static_variables"] = Interpertations["static_variables"].tolist()
-        attributions_dict["encoder_variables"] = Interpertations["encoder_variables"].tolist()
-        print("attention", Interpertations)
-        F_attention = model.Faithfulness_Correlation(
-            test_loader, Interpertations["attention"], pertrub='baseline', subset_size=4, time_step=True, nr_runs=10)
-        print('Attention faithfulness correlation', F_attention)
-        XAI_dict["attention_faith"] = F_attention.tolist()
-        # Path to the JSON file in log_dir
-        json_file_path = f"{log_dir}/Attributions.json"
+                else:
+                    print("{} Attributions faithfulness featrues ".format(key), v_score)
+                    XAI_dict["{}_faith_features".format(key)] = v_score
 
-        # Write the dictionary to a JSON file
-        with open(json_file_path, "w") as json_file:
-            json.dump(attributions_dict, json_file)
+                    print("{}_Attributions faithfulness variable per timestep ".format(
+                        key), ts_v_score)
+                    XAI_dict["{}_faith_variable_per_timestep".format(key)] = ts_v_score
+
+        # Getting the interpertations using pytorch forecasting native methods
 
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/XAI_metrics.json"
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index dacfbfb3..db2f77b8 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -21,6 +21,7 @@
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson
 import captum
+from captum._utils.models.linear_model import SkLearnLasso
 
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
@@ -478,6 +479,16 @@ def step_fn(self, element, step_prefix=""):
         return loss
 
     def prep_data_captum(self, x):
+        """
+        Prepares data to be fed into captum and generates baseline as well.
+
+        Args:
+            - x:Batch from dataloader
+        Returns:
+            - data:batch data in a tuple after being prepared
+            - baselines:Basically zero tensors in the input 
+        """
+        # captum requires gradient and float values
         data = (
             x["encoder_cat"].float().requires_grad_(),
             x["encoder_cont"].requires_grad_(),
@@ -493,12 +504,12 @@ def prep_data_captum(self, x):
         )
         baselines = (
             data[0].to(self.device),  # encoder_cat, no cat variables
-            torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to random
-            torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to random
+            torch.zeros_like(data[1]).to(self.device),  # encoder_cont, set to zero
+            torch.zeros_like(data[2]).to(self.device),  # encoder_target, set to zero
             data[3].to(self.device),  # encoder_lengths, leave unchanged
             data[4].to(self.device),  # decoder_cat, no cat variables
-            torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to random
-            torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to random
+            torch.zeros_like(data[5]).to(self.device),  # decoder_cont, set to zero
+            torch.zeros_like(data[6]).to(self.device),  # decoder_target, set to zero
             data[7].to(self.device),  # decoder_lengths, leave unchanged
             data[8].to(self.device),  # decoder_time_idx, unchanged
             data[9].to(self.device),  # groups, leave unchanged
@@ -506,14 +517,126 @@ def prep_data_captum(self, x):
         )
         return data, baselines
 
-    def explantation_captum(self, test_loader, method, local_index=None, log_dir=".", plot=False, local=False, **kwargs):
+    def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir):
+        """
+        Plots the attribution values for features and timesteps.
+
+        Args:
+            - features_attrs: Array of feature attribution values.
+            - timestep_attrs: Array of timestep attribution values.
+            - method_name: Name of the attribution method.
+            - log_dir: Directory to save the plots.
+        Returns:
+            Nothing
+        """
+
+        # Plot for feature attributions
+        x_values = np.arange(1, len(features_attrs) + 1)
+        plt.figure(figsize=(8, 6))
+        plt.plot(
+            x_values,
+            features_attrs,
+            marker="o",
+            color="skyblue",
+            linestyle="-",
+            linewidth=2,
+            markersize=8,
+        )
+        plt.xlabel("Time Step")
+        plt.ylabel("{} Attribution".format(method_name))
+        plt.title("{} Attribution Values".format(method_name))
+        plt.xticks(x_values, ['height', 'weight', 'age', 'sex', 'time_idx', 'alb', 'alp', 'alt', 'ast', 'be', 'bicar', 'bili', 'bili_dir', 'bnd', 'bun', 'ca', 'cai', 'ck', 'ckmb', 'cl', 'crea', 'crp', 'dbp', 'fgn', 'fio2', 'glu',
+                   'hgb', 'hr', 'inr_pt', 'k', 'lact', 'lymph', 'map', 'mch', 'mchc', 'mcv', 'methb', 'mg', 'na', 'neut', 'o2sat', 'pco2', 'ph', 'phos', 'plt', 'po2', 'ptt', 'resp', 'sbp', 'temp', 'tnt', 'urine', 'wbc'], rotation=90)
+        plt.tight_layout()
+        plt.savefig(log_dir / "{}_attribution_features_plot.png".format(method_name), bbox_inches="tight")
+
+        # Plot for timestep attributions
+        x_values = np.arange(1, len(timestep_attrs) + 1)
+        plt.figure(figsize=(8, 6))
+        plt.plot(
+            x_values,
+            timestep_attrs,
+            marker="o",
+            color="skyblue",
+            linestyle="-",
+            linewidth=2,
+            markersize=8,
+        )
+        plt.xlabel("Time Step")
+        plt.ylabel("{} Attribution".format(method_name))
+        plt.title("{} Attribution Values".format(method_name))
+        plt.xticks(x_values)
+        plt.tight_layout()
+        plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
+
+    def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=False, **kwargs):
+        """
+        Generic method to combine pytorchforecasting data loading , interpertations and captum to generate attributions
+
+        Args:
+            - dataloader: pytorchforecasting data loader 
+            - method: The explantation method chosen
+            - log_dir= The directory to output the plots
+            - plot= Determines if plots should be done or not
+            - XAI_metric=Determines if XAI metrics should be calculated or not
+        Returns:
+            - all_attrs : Attribtuons of features per timesteps
+            - features_attrs : Attribtuons of features averaged over timesteps
+            - timestep_attrs : Attribtuons of timesteps averaged over features
+            - ts_v_score: Faithfulness score for attribtuons of features per timesteps
+            - ts_score: Faithfulness score for attribtuons of timesteps averaged over features
+        """
         # Initialize lists to store attribution values for all instances
         all_attrs = []
-        score = []
-        r_score = []
-        method_name = method.__name__
-        # Loop through the test_loader to compute attributions for all instances
-        for batch in test_loader:
+        ts_score = []
+        r_ts_score = []
+        ts_v_score = []
+        r_ts_v_score = []
+        v_score = []
+
+        method_name = method if (method == "Random") or (method == "Attention") else (
+            method.__name__)
+        if (method_name == "Random") or (method_name == "Attention"):
+            if method_name == "Attention":
+                Interpertations = self.interpertations(dataloader=dataloader, log_dir=log_dir, plot=plot)
+                timestep_attrs = Interpertations["attention"]
+                features_attrs = Interpertations["static_variables"].tolist()
+                features_attrs.extend(Interpertations["encoder_variables"].tolist())
+
+            elif method_name == "Random":
+                # Generate random attributions for baseline comparison
+                all_attrs = np.random.normal(size=[64, 24, 53])
+                features_attrs = all_attrs.mean(axis=(1))
+                timestep_attrs = all_attrs.mean(axis=(2))
+            if XAI_metric:
+                for batch in dataloader:
+
+                    for key, value in batch[0].items():
+                        batch[0][key] = batch[0][key].to(self.device)
+                    x = batch[0]
+
+                    if method_name == "Random":
+                        ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                                                                        pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
+                        ts_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                                                                      pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
+                        v_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                                                                     pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+                    else:
+                        ts_score.append(self.Faithfulness_Correlation(x, timestep_attrs,
+                                                                      pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
+                        v_score.append(self.Faithfulness_Correlation(x, features_attrs,
+                                                                     pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+
+            # Faithfulness score for attribtuons of features per timesteps
+            ts_v_score = np.mean(ts_v_score)
+            # Faithfulness score for attribtuons of timesteps averaged over features
+            ts_score = np.mean(ts_score)
+            v_score = np.mean(v_score)
+            return all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score
+
+        # Loop through the dataloader to compute attributions for all instances
+        for batch in dataloader:
 
             for key, value in batch[0].items():
                 batch[0][key] = batch[0][key].to(self.device)
@@ -521,74 +644,61 @@ def explantation_captum(self, test_loader, method, local_index=None, log_dir="."
 
             data, baselines = self.prep_data_captum(x)
 
-            explantation = method(self.forward_captum)
+            # Initialize the explanation method
+            explanation = method(self.forward_captum, interpretable_model=SkLearnLasso(
+                alpha=0.4)) if method_name == 'Lime' else method(self.forward_captum)
 
+            # Calculate attributions using the selected method
             if method is not captum.attr.Saliency:
-                attr = explantation.attribute(data, baselines=baselines, **kwargs)
+                attr = explanation.attribute(data, baselines=baselines, **kwargs)
             else:
-                attr = explantation.attribute(data, **kwargs)
-
-            # Convert attributions to numpy array and append to the list
-            print(attr.shape, 'shape out of captum should be 64x24x53')
-            stacked_attr = torch.stack(attr).cpu().detach().numpy()
-
-            # Compute the mean along the specified axis (axis=0) and keep it as a numpy array
-            """ score.append(self.Faithfulness_Correlation2(x, stacked_attr,
-                                                        pertrub="Noise", feature_timestep=True, subset_size=[4, 9], nr_runs=10))
-            r_score.append(self.Faithfulness_Correlation2(x, np.random.normal(
-                size=[64, 24, 53]), pertrub="Noise", feature_timestep=True, subset_size=[4, 9], nr_runs=10)) """
-            print(np.shape(stacked_attr), 'shape after stacking')
+                attr = explanation.attribute(data, **kwargs)
+
+            # Process and store the calculated attributions
+            stacked_attr = attr[1].cpu().detach().numpy() if method_name in [
+                'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
+            if XAI_metric:
+
+                ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                                                                pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
+
+                ts_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                                                              pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
+                v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                                                             pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+            # aggregate over batch
             attr = np.mean(stacked_attr, axis=0)
-            print(np.shape(attr), 'shape after averaging over batch should 24x53')
             all_attrs.append(attr)
-
+        # aggregate over all batches
         all_attrs = np.array(all_attrs).mean(axis=(0))
-
+        # aggregate over all timesteps
         features_attrs = all_attrs.mean(axis=(0))
-
+        # aggregate over all features
         timestep_attrs = all_attrs.mean(axis=(1))
+        # Faithfulness score for attribtuons of features per timesteps
+        ts_v_score = np.mean(ts_v_score)
+        # Faithfulness score for attribtuons of timesteps averaged over features
+        ts_score = np.mean(ts_score)
+        # Faithfulness score for attribtuons of timesteps averaged over timesteps
+        v_score = np.mean(v_score)
 
-        # normalized_means = (means - means.min()) / (means.max() - means.min())
         if plot:
-            x_values = np.arange(1, 54)
-            # Plotting the featrue means
-            plt.figure(figsize=(8, 6))
-            plt.plot(
-                x_values,
-                features_attrs,
-                marker="o",
-                color="skyblue",
-                linestyle="-",
-                linewidth=2,
-                markersize=8,
-            )
-            plt.xlabel("Time Step")
-            plt.ylabel("{} Attribution".format(method_name))
-            plt.title("{} Attribution Values".format(method_name))
-            plt.xticks(x_values, ['height', 'weight', 'age', 'sex', 'time_idx', 'alb', 'alp', 'alt', 'ast', 'be', 'bicar', 'bili', 'bili_dir', 'bnd', 'bun', 'ca', 'cai', 'ck', 'ckmb', 'cl', 'crea', 'crp', 'dbp', 'fgn', 'fio2', 'glu', 'hgb', 'hr', 'inr_pt', 'k', 'lact', 'lymph', 'map', 'mch', 'mchc', 'mcv', 'methb', 'mg', 'na', 'neut', 'o2sat', 'pco2', 'ph', 'phos', 'plt', 'po2', 'ptt', 'resp', 'sbp', 'temp', 'tnt', 'urine', 'wbc'], rotation=90
-                       )  # Set x-ticks to match the number of features
-            plt.tight_layout()
-            plt.savefig(log_dir / "{}_attribution_features_plot.png".format(method_name), bbox_inches="tight")
-            plt.figure(figsize=(8, 6))
-            x_values = np.arange(1, 25)
-            plt.plot(
-                x_values,
-                timestep_attrs,
-                marker="o",
-                color="skyblue",
-                linestyle="-",
-                linewidth=2,
-                markersize=8,
-            )
-            plt.xlabel("Time Step")
-            plt.ylabel("{} Attribution ".format(method_name))
-            plt.title("{} Attribution Values".format(method_name))
-            plt.xticks(x_values)  # Set x-ticks to match the number of features
-            plt.tight_layout()
-            plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
-        return all_attrs, features_attrs, timestep_attrs
+            # Plot attributions for features
+            self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir)
+
+        # Return computed attributions and metrics
+        return all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score
+        # normalized_means = (means - means.min()) / (means.max() - means.min())
 
     def prep_data(self, x):
+        """
+        Prepares data for custom forward method 
+
+        Args:
+            - x:Batch returned from dataloader
+        Returns:
+            data:Tuple consisting of the tensors of X in the format the forward method needs
+        """
         data = (
             x["encoder_cat"],
             x["encoder_cont"],
@@ -606,9 +716,8 @@ def prep_data(self, x):
 
     def Faithfulness_Correlation(
         self,
-        test_loader,
+        x,
         attribution,
-        local_index=None,
         similarity_func=None,
         nr_runs=100,
         pertrub=None,
@@ -616,9 +725,24 @@ def Faithfulness_Correlation(
         feature=False,
         time_step=False,
         feature_timestep=False,
-        local=False,
+
     ):
         """
+        Calculates faithfulness scores for captum attributions 
+
+        Args:
+            - x:Batch input
+            -attribution: attribution generated by captum,
+            - similarity_func:function to determine similarity between sum of attributions and difference in prediction
+            - nr_runs: How many times to repeat the experiment,
+            - pertrub: What change to do to the input,
+            - subset_size: The size of the subset of featrues to alter ,
+            - feature: Determines if to calcualte faithfulness of feature attributions,
+            - time_step: Determines if to calcualte faithfulness of timesteps attributions,
+            - feature_timestep: Determines if to calcualte faithfulness of featrues per timesteps attributions,
+        Returns:
+            score: similarity score between sum of attributions and difference in prediction averaged over nr_runs
+
         Implementation of faithfulness correlation by Bhatt et al., 2020.
 
         The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
@@ -642,137 +766,41 @@ def Faithfulness_Correlation(
         """
         def add_noise(x, indices, time_step, feature_timestep):
             noise = torch.randn_like(x["encoder_cont"])
-            # In-place modification to save memory
-            if time_step:
-                x["encoder_cont"][:, indices, :] += noise[:, indices, :]
-            elif feature_timestep:
-                x["encoder_cont"][:, indices[0], :][:, :, indices[1]] += noise[:, indices[0], :][:, :, indices[1]]
-
-        def apply_baseline(x, indices, time_step, feature_timestep):
-            mask = torch.ones_like(x["encoder_cont"])
             if time_step:
+                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing='ij')
 
-                mask[:, indices, :] -= mask[:, indices, :]
-            elif feature_timestep:
-                mask[:, indices[0], :][:, :, indices[1]] -= mask[:, indices[0], :][:, :, indices[1]]
-            # Apply mask in-place
-            x["encoder_cont"] *= mask
+                with torch.no_grad():
+                    x["encoder_cont"][idx0, idx1, :] += noise[idx0, idx1, :]
 
-        # Assuming 'attribution' is already a GPU tensor
-        if not torch.is_tensor(attribution):
-            attribution = torch.tensor(attribution).to(self.device)
+            elif feature:
+                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing='ij')
 
-        # Other initializations
-        if similarity_func is None:
-            similarity_func = correlation_pearson  # Ensure this function is compatible with GPU tensors
-        if pertrub is None:
-            pertrub = "baseline"
-        similarities = []
-        if local:
-            for key, value in test_loader[0].items():
-                batch[0][key] = value.to(self.device)
-            x = batch[0]
+                with torch.no_grad():
+                    x["encoder_cont"][idx0, :, idx1] += noise[idx0, :, idx1]
 
-        else:
-
-            for batch in test_loader:
-                # Move batch to GPU
-                for key, value in batch[0].items():
-                    batch[0][key] = value.to(self.device)
-                x = batch[0]
-                # Assuming this is a method to prepare your data
-
-                y_pred = self((self.prep_data(x))).detach()  # Keep on GPU
-                pred_deltas = []
-                att_sums = []
-
-                for i_ix in range(nr_runs):
-                    if time_step:
-                        a_ix = np.random.choice(24, subset_size, replace=False)
-                    elif feature_timestep:
-                        timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
-                        variables_idx = np.random.choice(53, subset_size[1], replace=False)
-                        a_ix = [timesteps_idx, variables_idx]
-
-                    # Apply perturbation
-                    if pertrub == "Noise":
-                        add_noise(x, a_ix, time_step, feature_timestep)
-                    elif pertrub == "baseline":
-                        apply_baseline(x, a_ix, time_step, feature_timestep)
-
-                    # Predict on perturbed input and calculate deltas
-                    y_pred_perturb = (self(self.prep_data(x))).detach()  # Keep on GPU
-                    # Change dims as per your requirement
-                    pred_deltas.append((y_pred - y_pred_perturb).mean(dim=[0, 1, 2]))
-
-                    # Sum attributions of the random subset and keep on GPU
-                    if time_step:
-                        att_sums.append(attribution[a_ix].sum())
-                    elif feature_timestep:
-                        row_indices = a_ix[0]
-                        col_indices = a_ix[1]
-                        att_sums.append(attribution[row_indices][:, col_indices].sum())
-
-                # Convert to CPU for numpy operations
-                pred_deltas_cpu = torch.stack(pred_deltas).cpu().numpy()
-                att_sums_cpu = torch.tensor(att_sums).cpu().numpy()
-                similarities.append(similarity_func(pred_deltas_cpu, att_sums_cpu))
+            elif feature_timestep:
+                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing='ij')
 
-        score = np.nanmean(similarities)
-        return score
+                with torch.no_grad():
+                    x["encoder_cont"][idx0, idx1, idx2] += noise[idx0, idx1, idx2]
 
-    def Faithfulness_Correlation2(
-        self,
-        x,
-        attribution,
-        local_index=None,
-        similarity_func=None,
-        nr_runs=100,
-        pertrub=None,
-        subset_size=3,
-        feature=False,
-        time_step=False,
-        feature_timestep=False,
-        local=False,
-    ):
-        """
-        Implementation of faithfulness correlation by Bhatt et al., 2020.
+        def apply_baseline(x, indices, time_step, feature_timestep):
+            mask = torch.ones_like(x["encoder_cont"])
+            if time_step:
 
-        The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
-        (or 'fidelity') with respect to the model behaviour.
+                idx0, idx1, = np.meshgrid(indices[0], indices[1], indexing='ij')
 
-        Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
-        the average explanation attribution for only the subset of features are (linearly) correlated, taking the
-        average over multiple runs and test samples. The metric returns one float per input-attribution pair that
-        ranges between -1 and 1, where higher scores are better.
+                mask[idx0, idx1, :] -= mask[idx0, idx1, :]
+            elif feature:
+                idx0, idx1, = np.meshgrid(indices[0], indices[1], indexing='ij')
 
-        For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
-        or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
-        test point and the average explanation attribution for only the subset of features is calculated. Results is
-        average over multiple runs and several test samples.
-        This code is adapted from the quantus libray to suit our use case
+                mask[idx0, :, idx1] -= mask[idx0, :, idx1]
 
-        References:
-            1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
-            explanations." IJCAI (2020): 3016-3022.
-            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
-        """
-        def add_noise(x, indices, time_step, feature_timestep):
-            noise = torch.randn_like(x["encoder_cont"])
-            # In-place modification to save memory
-            if time_step:
-                x["encoder_cont"][indices[0], :, :][:, indices[1], :] += noise[indices[0], :, :][:, indices[1], :]
             elif feature_timestep:
-                x["encoder_cont"][indices[0], :, :][:, indices[1], :][:, :, indices[2]
-                                                                      ] += noise[indices[0], :, :][:, indices[1], :][:, :, indices[2]]
+                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing='ij')
+
+                mask[idx0, idx1, idx2] -= mask[idx0, idx1, idx2]
 
-        def apply_baseline(x, indices, time_step, feature_timestep):
-            mask = torch.ones_like(x["encoder_cont"])
-            if time_step:
-                mask[indices[0], :, :][:, indices[1], :] = 0
-            elif feature_timestep:
-                mask[indices[0], :, :][:, indices[1], :][:, :, indices[2]
-                                                         ] = 0
             with torch.no_grad():
                 x["encoder_cont"] *= mask
         # Assuming 'attribution' is already a GPU tensor
@@ -781,7 +809,7 @@ def apply_baseline(x, indices, time_step, feature_timestep):
 
         # Other initializations
         if similarity_func is None:
-            similarity_func = correlation_pearson  # Ensure this function is compatible with GPU tensors
+            similarity_func = correlation_spearman
         if pertrub is None:
             pertrub = "baseline"
         similarities = []
@@ -795,13 +823,18 @@ def apply_baseline(x, indices, time_step, feature_timestep):
         for i_ix in range(nr_runs):
             if time_step:
                 timesteps_idx = np.random.choice(24, subset_size, replace=False)
-                patient_idx = np.random.choice(64, subset_size, replace=False)
+                patient_idx = np.random.choice(64, 1, replace=False)
                 a_ix = [patient_idx, timesteps_idx]
+
+            elif feature:
+                feature_idx = np.random.choice(53, subset_size, replace=False)
+                patient_idx = np.random.choice(64, 1, replace=False)
+                a_ix = [patient_idx, feature_idx]
             elif feature_timestep:
                 timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
-                variables_idx = np.random.choice(53, subset_size[1], replace=False)
+                feature_idx = np.random.choice(53, subset_size[1], replace=False)
                 patient_idx = np.random.choice(64, 1, replace=False)
-                a_ix = [patient_idx, timesteps_idx, variables_idx]
+                a_ix = [patient_idx, timesteps_idx, feature_idx]
 
             # Apply perturbation
             if pertrub == "Noise":
@@ -811,32 +844,45 @@ def apply_baseline(x, indices, time_step, feature_timestep):
 
             # Predict on perturbed input and calculate deltas
             y_pred_perturb = (self(self.prep_data(x))).detach()  # Keep on GPU
-            pred_deltas.append((y_pred - y_pred_perturb).mean(dim=[0, 1, 2]))  # Change dims as per your requirement
 
-            # Sum attributions of the random subset and keep on GPU
             if time_step:
-                patient_indices = a_ix[0]
-                timestep_indices = a_ix[1]
-                att_sums.append(attribution[patient_indices, :, :][:, timestep_indices, :].sum())
-
+                # patient_indices = a_ix[0]
+                # timestep_indices = a_ix[1]
+                if attribution.size() == torch.Size([24]):
+                    att_sums.append((attribution[timesteps_idx]).sum())
+                else:
+                    att_sums.append((attribution[patient_idx, :][:, timesteps_idx]).sum())
+            elif feature:
+                # patient_indices = a_ix[0]
+                # feature_indices = a_ix[1]
+                if len(attribution) == 53:
+                    att_sums.append((attribution[feature_idx]).sum())
+                else:
+                    att_sums.append((attribution[patient_idx, :][:, feature_idx]).sum())
             elif feature_timestep:
-                patient_indices = a_ix[0]
-                timestep_indices = a_ix[1]
-                variable_indices = a_ix[2]
-                att_sums.append(attribution[patient_indices, :, :]
-                                [:, timestep_indices, :][:, :, variable_indices].sum())
+                # patient_indices = a_ix[0]
+                # timestep_indices = a_ix[1]
+                # variable_indices = a_ix[2]
+
+                att_sums.append((attribution[patient_idx, :, :]
+                                [:, timesteps_idx, :][:, :, feature_idx]).sum())
+            print(y_pred.shape)
+            print(y_pred_perturb.shape)
+            print((y_pred - y_pred_perturb)[patient_idx], 'error')
 
+            pred_deltas.append((y_pred - y_pred_perturb)[patient_idx].item())
             # Convert to CPU for numpy operations
 
-        pred_deltas_cpu = torch.stack(pred_deltas).cpu().numpy()
+        pred_deltas_cpu = torch.tensor(pred_deltas).cpu().numpy()
         att_sums_cpu = torch.tensor(att_sums).cpu().numpy()
+
         similarities.append(similarity_func(pred_deltas_cpu, att_sums_cpu))
 
         score = np.nanmean(similarities)
         return score
 
     def Data_Randomization(
-        self, test_loader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None, **kwargs
+        self, dataloader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None, **kwargs
     ):
         """
         Implementation of the Random Logit Metric by Sixt et al., 2020.
@@ -855,11 +901,11 @@ def Data_Randomization(
         if similarity_func is None:
             similarity_func = distance_euclidean
         if explain_method == "Attention":
-            Attention_weights = random_model.interpertations(test_loader)
+            Attention_weights = random_model.interpertations(dataloader)
 
             score = similarity_func(Attention_weights["attention"].cpu().numpy(), attribution.cpu().numpy())
         else:
-            for batch in test_loader:
+            for batch in dataloader:
                 for key, value in batch[0].items():
                     batch[0][key] = batch[0][key].to(random_model.device)
                 x = batch[0]
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 944295ea..7a67f3c5 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -15,6 +15,7 @@
 from statistics import mean, pstdev
 from icu_benchmarks.models.utils import JsonResultLoggingEncoder
 from icu_benchmarks.wandb_utils import wandb_log
+import numpy as np
 
 
 def build_parser() -> ArgumentParser:
@@ -238,7 +239,9 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
     # Calculate the population standard deviation over aggregated results over folds/iterations
     # Divide by sqrt(n) to get standard deviation.
-    std_scores = {metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items()}
+
+    std_scores = {metric: (pstdev(list) / sqrt(len(list)))
+                  for metric, list in list_scores.items() if not (np.isnan(list).all())}
 
     confidence_interval = {
         metric: (stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list)))
@@ -305,7 +308,8 @@ def load_pretrained_imputation_model(use_pretrained_imputation):
         pretrained_imputation_model_checkpoint = torch.load(use_pretrained_imputation, map_location=torch.device("cpu"))
         if isinstance(pretrained_imputation_model_checkpoint, dict):
             imputation_model_class = pretrained_imputation_model_checkpoint["class"]
-            pretrained_imputation_model = imputation_model_class(**pretrained_imputation_model_checkpoint["hyper_parameters"])
+            pretrained_imputation_model = imputation_model_class(
+                **pretrained_imputation_model_checkpoint["hyper_parameters"])
             pretrained_imputation_model.set_trained_columns(pretrained_imputation_model_checkpoint["trained_columns"])
             pretrained_imputation_model.load_state_dict(pretrained_imputation_model_checkpoint["state_dict"])
         else:

From 64fb77f823d84537eb58814791a1f32fafba13eb Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 29 Nov 2023 13:53:41 +0100
Subject: [PATCH 115/142] added data randomization test

---
 icu_benchmarks/models/train.py    |  40 ++++----
 icu_benchmarks/models/wrappers.py | 158 +++++++++++++-----------------
 2 files changed, 91 insertions(+), 107 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 1c739b90..1775c0a7 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -263,6 +263,18 @@ def train_common(
         return 0
 
     if explain:
+        if random_model_dir is None:
+            random_model = None
+        else:
+            path = Path(random_model_dir)
+
+            random_model = load_model(
+                model,
+                source_dir=path,
+                pl_model=pl_model,
+                train_dataset=train_dataset,
+                optimizer=optimizer,
+            )
 
         XAI_dict = {}  # dictrionary to log attributions metrics
 
@@ -270,24 +282,24 @@ def train_common(
         methods = {
             "Saliency": Saliency,
             # "Lime": Lime,
-            "IG": IntegratedGradients,
+            # "IG": IntegratedGradients,
             # "FA": FeatureAblation,
             "Random": "Random",
-            "Attention": "Attention"
+            # "Attention": "Attention"
         }
         for key, item in methods.items():
             # If conditions needed here as different explantations require different inputs
             if key == "IG":
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=50, XAI_metric=XAI_metric,
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=50, XAI_metric=XAI_metric, random_model=random_model
                 )
             elif key == "Lime":
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True, XAI_metric=XAI_metric,
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True, XAI_metric=XAI_metric, random_model=random_model
                 )
             else:
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, XAI_metric=XAI_metric,
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, XAI_metric=XAI_metric, random_model=random_model
                 )
 
             if XAI_metric:
@@ -305,6 +317,9 @@ def train_common(
                     print("{}_Attributions faithfulness variable per timestep ".format(
                         key), ts_v_score)
                     XAI_dict["{}_faith_variable_per_timestep".format(key)] = ts_v_score
+                print("{}_Data Randomization distance ".format(
+                    key), r_score)
+                XAI_dict["{}_Data Randomization distance".format(key)] = r_score
 
         # Getting the interpertations using pytorch forecasting native methods
 
@@ -315,15 +330,6 @@ def train_common(
         with open(json_file_path, "w") as json_file:
             json.dump(XAI_dict, json_file)
 
-        # path = Path(random_model_dir)
-
-        # random_model = load_model(
-        #    model,
-        #    source_dir=path,
-        #    pl_model=pl_model,
-        #    train_dataset=train_dataset,
-        #    optimizer=optimizer,
-        # )
         # R_attribution = model.Data_Randomization(
         #    test_loader, attributions_IG, IntegratedGradients, random_model)
         # print('Distance Data randmoization score attribution', R_attribution)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index db2f77b8..ada7d150 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -569,7 +569,7 @@ def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir
         plt.tight_layout()
         plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
 
-    def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=False, **kwargs):
+    def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=False, random_model=None, test_dataset=None, **kwargs):
         """
         Generic method to combine pytorchforecasting data loading , interpertations and captum to generate attributions
 
@@ -583,16 +583,15 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             - all_attrs : Attribtuons of features per timesteps
             - features_attrs : Attribtuons of features averaged over timesteps
             - timestep_attrs : Attribtuons of timesteps averaged over features
-            - ts_v_score: Faithfulness score for attribtuons of features per timesteps
-            - ts_score: Faithfulness score for attribtuons of timesteps averaged over features
+            - f_ts_v_score: Faithfulness score for attribtuons of features per timesteps
+            - f_ts_score: Faithfulness score for attribtuons of timesteps averaged over features
         """
         # Initialize lists to store attribution values for all instances
         all_attrs = []
-        ts_score = []
-        r_ts_score = []
-        ts_v_score = []
-        r_ts_v_score = []
-        v_score = []
+        f_ts_score = []
+        f_ts_v_score = []
+        f_v_score = []
+        r_score = []
 
         method_name = method if (method == "Random") or (method == "Attention") else (
             method.__name__)
@@ -602,6 +601,8 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 timestep_attrs = Interpertations["attention"]
                 features_attrs = Interpertations["static_variables"].tolist()
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
+                r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
+                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name)
 
             elif method_name == "Random":
                 # Generate random attributions for baseline comparison
@@ -616,24 +617,29 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                     x = batch[0]
 
                     if method_name == "Random":
-                        ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
-                                                                        pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
-                        ts_score.append(self.Faithfulness_Correlation(x, all_attrs,
-                                                                      pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
-                        v_score.append(self.Faithfulness_Correlation(x, all_attrs,
-                                                                     pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+
+                        f_ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                                                                          pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
+                        f_ts_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                                                                        pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
+                        f_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+
+                        r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
+                                       explain_method=method, random_model=random_model, method_name=method_name))
                     else:
-                        ts_score.append(self.Faithfulness_Correlation(x, timestep_attrs,
-                                                                      pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
-                        v_score.append(self.Faithfulness_Correlation(x, features_attrs,
-                                                                     pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+                        f_ts_score.append(self.Faithfulness_Correlation(x, timestep_attrs,
+                                                                        pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
+                        f_v_score.append(self.Faithfulness_Correlation(x, features_attrs,
+                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
 
             # Faithfulness score for attribtuons of features per timesteps
-            ts_v_score = np.mean(ts_v_score)
+            f_ts_v_score = np.mean(f_ts_v_score)
             # Faithfulness score for attribtuons of timesteps averaged over features
-            ts_score = np.mean(ts_score)
-            v_score = np.mean(v_score)
-            return all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score
+            f_ts_score = np.mean(f_ts_score)
+            f_v_score = np.mean(f_v_score)
+            r_score = np.mean(r_score)
+            return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score
 
         # Loop through the dataloader to compute attributions for all instances
         for batch in dataloader:
@@ -659,13 +665,15 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
             if XAI_metric:
 
-                ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
-                                                                pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
+                f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                                                                  pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
 
-                ts_score.append(self.Faithfulness_Correlation(x, stacked_attr,
-                                                              pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
-                v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
-                                                             pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+                f_ts_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                                                                pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
+                f_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                                                               pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+                r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
+                               explain_method=method, random_model=random_model, method_name=method_name))
             # aggregate over batch
             attr = np.mean(stacked_attr, axis=0)
             all_attrs.append(attr)
@@ -676,18 +684,24 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
         # aggregate over all features
         timestep_attrs = all_attrs.mean(axis=(1))
         # Faithfulness score for attribtuons of features per timesteps
-        ts_v_score = np.mean(ts_v_score)
+        f_ts_v_score = np.mean(f_ts_v_score)
         # Faithfulness score for attribtuons of timesteps averaged over features
-        ts_score = np.mean(ts_score)
+        f_ts_score = np.mean(f_ts_score)
         # Faithfulness score for attribtuons of timesteps averaged over timesteps
-        v_score = np.mean(v_score)
+        f_v_score = np.mean(f_v_score)
+
+        min_val = np.min(r_score)
+        max_val = np.max(r_score)
+
+        r_score = (r_score - min_val) / (max_val - min_val)
+        r_score = np.mean(r_score)
 
         if plot:
             # Plot attributions for features
             self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir)
 
         # Return computed attributions and metrics
-        return all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score
+        return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score
         # normalized_means = (means - means.min()) / (means.max() - means.min())
 
     def prep_data(self, x):
@@ -846,29 +860,22 @@ def apply_baseline(x, indices, time_step, feature_timestep):
             y_pred_perturb = (self(self.prep_data(x))).detach()  # Keep on GPU
 
             if time_step:
-                # patient_indices = a_ix[0]
-                # timestep_indices = a_ix[1]
+
                 if attribution.size() == torch.Size([24]):
                     att_sums.append((attribution[timesteps_idx]).sum())
                 else:
-                    att_sums.append((attribution[patient_idx, :][:, timesteps_idx]).sum())
+                    att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :]).sum())
             elif feature:
-                # patient_indices = a_ix[0]
-                # feature_indices = a_ix[1]
+
                 if len(attribution) == 53:
                     att_sums.append((attribution[feature_idx]).sum())
                 else:
-                    att_sums.append((attribution[patient_idx, :][:, feature_idx]).sum())
+
+                    att_sums.append((attribution[patient_idx, :, :][:, :, feature_idx]).sum())
             elif feature_timestep:
-                # patient_indices = a_ix[0]
-                # timestep_indices = a_ix[1]
-                # variable_indices = a_ix[2]
 
                 att_sums.append((attribution[patient_idx, :, :]
                                 [:, timesteps_idx, :][:, :, feature_idx]).sum())
-            print(y_pred.shape)
-            print(y_pred_perturb.shape)
-            print((y_pred - y_pred_perturb)[patient_idx], 'error')
 
             pred_deltas.append((y_pred - y_pred_perturb)[patient_idx].item())
             # Convert to CPU for numpy operations
@@ -882,7 +889,8 @@ def apply_baseline(x, indices, time_step, feature_timestep):
         return score
 
     def Data_Randomization(
-        self, dataloader, attribution, explain_method, random_model, similarity_func=None, test_dataset=None, **kwargs
+        self, x,
+        attribution, explain_method, random_model, similarity_func=None, dataloader=None, method_name="", **kwargs
     ):
         """
         Implementation of the Random Logit Metric by Sixt et al., 2020.
@@ -897,61 +905,31 @@ def Data_Randomization(
             2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
-        a_perturbed = []
+
         if similarity_func is None:
             similarity_func = distance_euclidean
         if explain_method == "Attention":
             Attention_weights = random_model.interpertations(dataloader)
 
             score = similarity_func(Attention_weights["attention"].cpu().numpy(), attribution.cpu().numpy())
+        elif explain_method == "Random":
+            score = similarity_func(np.random.normal(size=[64, 24, 53]), attribution)
         else:
-            for batch in dataloader:
-                for key, value in batch[0].items():
-                    batch[0][key] = batch[0][key].to(random_model.device)
-                x = batch[0]
-                y = batch[1][0]
-                data = (
-                    x["encoder_cat"],
-                    x["encoder_cont"],
-                    x["encoder_target"],
-                    x["encoder_lengths"],
-                    x["decoder_cat"],
-                    x["decoder_cont"],
-                    x["decoder_target"],
-                    x["decoder_lengths"],
-                    x["decoder_time_idx"],
-                    x["groups"],
-                    x["target_scale"],
-                )
-
-                baselines = (
-                    data[0].to(self.device),  # encoder_cat, set to random
-                    torch.randn_like(data[1]).to(self.device),  # encoder_cont, set to random
-                    torch.randn_like(data[2]).to(self.device),  # encoder_target, set to random
-                    data[3].to(self.device),  # encoder_lengths, leave unchanged
-                    torch.randn_like(data[4]).to(self.device),  # decoder_cat, set to random
-                    torch.randn_like(data[5]).to(self.device),  # decoder_cont, set to random
-                    torch.randn_like(data[6]).to(self.device),  # decoder_target, set to random
-                    data[7].to(self.device),  # decoder_lengths, leave unchanged
-                    data[8].to(self.device),  # decoder_time_idx, unchanged
-                    data[9].to(self.device),  # groups, leave unchanged
-                    data[10].to(self.device),  # target_scale, leave unchanged
-                )
+            data, baselines = self.prep_data_captum(x)
+            y_pred = self(data).detach()
 
-                explantation = explain_method(random_model.forward_captum)
-                # Reformat attributions.
-                if explain_method is not captum.attr.Saliency:
-                    attr = explantation.attribute(data, baselines=baselines, **kwargs)
-                else:
-                    attr = explantation.attribute(data, **kwargs)
-                    # Convert attributions to numpy array and append to the list
-                a_perturbed.append(attr[0].cpu().detach().numpy())
+            explantation = explain_method(random_model.forward_captum)
+            # Reformat attributions.
+            if explain_method is not captum.attr.Saliency:
+                attr = explantation.attribute(data, baselines=baselines, **kwargs)
+            else:
+                attr = explantation.attribute(data, **kwargs)
 
-            a_perturbed = np.mean(a_perturbed, axis=(0))
+            # Process and store the calculated attributions
+            stacked_attr = attr[1].cpu().detach().numpy() if method_name in [
+                'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
 
-            # Normalize the means values to range [0, 1]
-            # normalized_a_perturbed = (a_perturbed - a_perturbed.min()) / (a_perturbed.max() - a_perturbed.min())
-            score = similarity_func(a_perturbed, attribution)
+            score = similarity_func(stacked_attr.flatten(), attribution.flatten())
         return score
 
 

From 42ef29946c9c2b3322ac567302a5eb1b0e35462b Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 29 Nov 2023 16:29:10 +0100
Subject: [PATCH 116/142] added mask for feature ablation

---
 icu_benchmarks/models/train.py    | 44 ++++++++++++++++++++++++++++---
 icu_benchmarks/models/wrappers.py | 11 +++++---
 2 files changed, 49 insertions(+), 6 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 1775c0a7..ac144025 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -280,11 +280,11 @@ def train_common(
 
         # choose which  methods to get attributions
         methods = {
-            "Saliency": Saliency,
-            # "Lime": Lime,
+            # "Saliency": Saliency,
+            "Lime": Lime,
             # "IG": IntegratedGradients,
             # "FA": FeatureAblation,
-            "Random": "Random",
+            # "Random": "Random",
             # "Attention": "Attention"
         }
         for key, item in methods.items():
@@ -297,6 +297,44 @@ def train_common(
                 all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
                     dataloader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True, XAI_metric=XAI_metric, random_model=random_model
                 )
+            elif key == "FA":
+                # basically need to specify a mask on which input values are considered as a feature
+                shapes = [
+                    torch.Size([64, 24, 0]),
+                    torch.Size([64, 24, 53]),
+                    torch.Size([64, 24]),
+                    torch.Size([64]),
+                    torch.Size([64, 1, 0]),
+                    torch.Size([64, 1, 53]),
+                    torch.Size([64, 1]),
+                    torch.Size([64]),
+                    torch.Size([64, 1]),
+                    torch.Size([64, 1]),
+                    torch.Size([64, 2])
+                ]
+
+                # Create a default mask for non-targeted tensors
+                def create_default_mask(shape):
+                    if len(shape) == 3:
+                        return torch.zeros(shape[0], shape[1], max(1, shape[2]), dtype=torch.int32)
+                    elif len(shape) == 2:
+                        return torch.zeros(shape[0], max(1, shape[1]), dtype=torch.int32)
+                    else:  # len(shape) == 1
+                        return torch.zeros(shape[0], dtype=torch.int32)
+
+                # Create a feature mask for the second tensor
+                num_features_second_tensor = shapes[1][2]
+                feature_mask_second = torch.arange(num_features_second_tensor).repeat(shapes[1][1], 1)
+                feature_mask_second = feature_mask_second.unsqueeze(0).repeat(shapes[1][0], 1, 1)
+
+                # Create a tuple of masks
+                feature_masks = tuple([create_default_mask(shape) if i !=
+                                      1 else feature_mask_second for i, shape in enumerate(shapes)])
+
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, feature_mask=feature_masks, XAI_metric=XAI_metric, random_model=random_model
+                )
+
             else:
                 all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
                     dataloader=test_loader, method=item, log_dir=log_dir, plot=True, XAI_metric=XAI_metric, random_model=random_model
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index ada7d150..cb928a02 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -489,6 +489,7 @@ def prep_data_captum(self, x):
             - baselines:Basically zero tensors in the input 
         """
         # captum requires gradient and float values
+
         data = (
             x["encoder_cat"].float().requires_grad_(),
             x["encoder_cont"].requires_grad_(),
@@ -625,7 +626,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                         f_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
                                                                        pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
 
-                        r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
+                        r_score.append(self.Data_Randomization(x, attribution=all_attrs,
                                        explain_method=method, random_model=random_model, method_name=method_name))
                     else:
                         f_ts_score.append(self.Faithfulness_Correlation(x, timestep_attrs,
@@ -638,6 +639,10 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             # Faithfulness score for attribtuons of timesteps averaged over features
             f_ts_score = np.mean(f_ts_score)
             f_v_score = np.mean(f_v_score)
+            min_val = np.min(r_score)
+            max_val = np.max(r_score)
+
+            r_score = (r_score - min_val) / (max_val - min_val)
             r_score = np.mean(r_score)
             return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score
 
@@ -647,7 +652,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             for key, value in batch[0].items():
                 batch[0][key] = batch[0][key].to(self.device)
             x = batch[0]
-
+            
             data, baselines = self.prep_data_captum(x)
 
             # Initialize the explanation method
@@ -913,7 +918,7 @@ def Data_Randomization(
 
             score = similarity_func(Attention_weights["attention"].cpu().numpy(), attribution.cpu().numpy())
         elif explain_method == "Random":
-            score = similarity_func(np.random.normal(size=[64, 24, 53]), attribution)
+            score = similarity_func(np.random.normal(size=[64, 24, 53]).flatten(), attribution.flatten())
         else:
             data, baselines = self.prep_data_captum(x)
             y_pred = self(data).detach()

From 7c87164d1289d3ca4976f8691586b7df15b3fb20 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 4 Dec 2023 16:30:39 +0100
Subject: [PATCH 117/142] plotting , added HP range , fixed distance range

---
 configs/prediction_models/DeepARpytorch.gin   |   8 +-
 configs/prediction_models/RNNpytorch.gin      |   8 +-
 configs/prediction_models/TFTpytorch.gin      |   4 +
 configs/prediction_models/common/DLCommon.gin |   2 +-
 configs/prediction_models/common/DLTuning.gin |   4 +-
 configs/tasks/Regression.gin                  |   4 +-
 configs/tasks/common/CrossValidation.gin      |   2 +-
 icu_benchmarks/models/train.py                |  60 +--
 icu_benchmarks/models/wrappers.py             |  55 +-
 icu_benchmarks/run_utils.py                   |  60 +++
 output.txt                                    | 475 ------------------
 11 files changed, 142 insertions(+), 540 deletions(-)
 delete mode 100644 output.txt

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 9a61ca84..5135035f 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -8,19 +8,21 @@ train_common.model = @DeepARpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = (0.0001, 0.001,0.01)
+optimizer/hyperparameter.lr = (1e-5, 1e-3)
 
 # Model params
 
 model/hyperparameter.class_to_tune = @DeepARpytorch
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.hidden = (40, 120, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
 model/hyperparameter.dropout = (0.0, 0.4)
 #model/hyperparameter.lr_scheduler = "exponential"
 
-
+train_common/hyperparameter.class_to_tune = @train_common
+train_common/hyperparameter.batch_size=(32,64,128,256,512)
+train_common/hyperparameter.gradient_clip_val=(0,0.01, 1.0, 100.0)
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index aa2398cb..0b88c670 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -8,7 +8,7 @@ train_common.model = @RNNpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = (0.0001, 0.001,0.01)
+optimizer/hyperparameter.lr = (1e-5, 3e-4)
 
 # Model params
 
@@ -17,10 +17,12 @@ model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1,3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
-model/hyperparameter.dropout = (0.0, 0.9)
+model/hyperparameter.dropout = (0.0, 0.4)
 #model/hyperparameter.lr_scheduler = "exponential"
 
-
+train_common/hyperparameter.class_to_tune = @train_common
+train_common/hyperparameter.batch_size=(32,64,128,256,512)
+train_common/hyperparameter.gradient_clip_val=(0,0.01, 1.0, 100.0)
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 4f2d078e..18eabb6a 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -19,8 +19,12 @@ model/hyperparameter.dropout =(0.1, 0.9)
 model/hyperparameter.dropout_att = (0.1, 0.9)
 model/hyperparameter.n_heads =(1,4)
 
+#dataloader
 
+train_common/hyperparameter.class_to_tune = @train_common
+train_common/hyperparameter.batch_size=(32,64,128,256,512)
 
+train_common/hyperparameter.gradient_clip_val=(0,0.01, 1.0, 100.0)
 
 
 
diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 2760e187..bfc48ef5 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,7 +15,7 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-train_common.batch_size = 64
+#train_common.batch_size = 64
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
diff --git a/configs/prediction_models/common/DLTuning.gin b/configs/prediction_models/common/DLTuning.gin
index 0c56677c..b631dc89 100644
--- a/configs/prediction_models/common/DLTuning.gin
+++ b/configs/prediction_models/common/DLTuning.gin
@@ -1,5 +1,5 @@
 # Hyperparameter tuner settings for Deep Learning.
-tune_hyperparameters.scopes = ["model", "optimizer"]
+tune_hyperparameters.scopes = ["model", "optimizer","train_common"]
 tune_hyperparameters.n_initial_points = 5
-tune_hyperparameters.n_calls = 20
+tune_hyperparameters.n_calls = 15
 tune_hyperparameters.folds_to_tune_on = 2
\ No newline at end of file
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index 75041763..1234f707 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -15,8 +15,8 @@ train_common.weight = "balanced"
 train_common.ram_cache = True
 
 # LOSS FUNCTION
-DLPredictionWrapper.loss = @l1_loss
-DLPredictionPytorchForecastingWrapper.loss = @l1_loss
+DLPredictionWrapper.loss = @mse_loss
+DLPredictionPytorchForecastingWrapper.loss = @mse_loss
 MLWrapper.loss = @mean_squared_error
 
 # SELECTING PREPROCESSOR
diff --git a/configs/tasks/common/CrossValidation.gin b/configs/tasks/common/CrossValidation.gin
index 519ad3dd..d0c4a704 100644
--- a/configs/tasks/common/CrossValidation.gin
+++ b/configs/tasks/common/CrossValidation.gin
@@ -1,3 +1,3 @@
 # CROSS-VALIDATION SETTINGS
-execute_repeated_cv.cv_repetitions = 2
+execute_repeated_cv.cv_repetitions = 1
 execute_repeated_cv.cv_folds = 5
\ No newline at end of file
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index ac144025..14517da7 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -77,6 +77,7 @@ def train_common(
     XAI_metric: bool = False,
     random_labels: bool = False,
     random_model_dir: str = None,
+
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -127,7 +128,7 @@ def train_common(
             f"Training on {train_dataset.name} with {len(train_dataset)} samples and validating on {val_dataset.name} with"
             f" {len(val_dataset)} samples."
         )
-
+    batch_size = int(batch_size)
     logging.info(f"Using {num_workers} workers for data loading.")
     if pytorch_forecasting:
         train_loader = train_dataset.to_dataloader(
@@ -244,7 +245,7 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
-
+        gradient_clip_val=gradient_clip_val
     )
     if not eval_only:
         if model.requires_backprop:
@@ -280,12 +281,12 @@ def train_common(
 
         # choose which  methods to get attributions
         methods = {
-            # "Saliency": Saliency,
-            "Lime": Lime,
-            # "IG": IntegratedGradients,
-            # "FA": FeatureAblation,
-            # "Random": "Random",
-            # "Attention": "Attention"
+            "G": Saliency,
+            "L": Lime,
+            "IG": IntegratedGradients,
+            "FA": FeatureAblation,
+            "R": "Random",
+            "Att": "Attention"
         }
         for key, item in methods.items():
             # If conditions needed here as different explantations require different inputs
@@ -293,12 +294,8 @@ def train_common(
                 all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
                     dataloader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=50, XAI_metric=XAI_metric, random_model=random_model
                 )
-            elif key == "Lime":
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True, XAI_metric=XAI_metric, random_model=random_model
-                )
-            elif key == "FA":
-                # basically need to specify a mask on which input values are considered as a feature
+            elif key == "L" or key == "FA":
+                # for Lime and feature ablation we need to define what is a feature we define each variable per timestep as a feature
                 shapes = [
                     torch.Size([64, 24, 0]),
                     torch.Size([64, 24, 53]),
@@ -322,17 +319,17 @@ def create_default_mask(shape):
                     else:  # len(shape) == 1
                         return torch.zeros(shape[0], dtype=torch.int32)
 
-                # Create a feature mask for the second tensor
-                num_features_second_tensor = shapes[1][2]
-                feature_mask_second = torch.arange(num_features_second_tensor).repeat(shapes[1][1], 1)
+                # Create a feature mask for the second tensor that includes both features and timesteps
+                num_timesteps = shapes[1][1]
+                num_features = shapes[1][2]
+                feature_mask_second = torch.arange(num_timesteps * num_features).reshape(num_timesteps, num_features)
                 feature_mask_second = feature_mask_second.unsqueeze(0).repeat(shapes[1][0], 1, 1)
 
                 # Create a tuple of masks
                 feature_masks = tuple([create_default_mask(shape) if i !=
                                       1 else feature_mask_second for i, shape in enumerate(shapes)])
-
                 all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, feature_mask=feature_masks, XAI_metric=XAI_metric, random_model=random_model
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True, XAI_metric=XAI_metric, random_model=random_model
                 )
 
             else:
@@ -342,22 +339,22 @@ def create_default_mask(shape):
 
             if XAI_metric:
                 # logging metric scores
-                print("{} Attributions faithfulness timesteps ".format(key), ts_score)
-                XAI_dict["{}_faith_timesteps".format(key)] = ts_score
-                if key == "Attention":
+                print("{} Attributions Faithfulness Timesteps ".format(key), ts_score)
+                XAI_dict["{}_Faith Timesteps".format(key)] = ts_score
+                if key == "Att":
                     print("Variable selection weights faithfulness featrues ".format(key), v_score)
-                    XAI_dict["Variable_selection_weights_faith_features".format(key)] = v_score
+                    XAI_dict["VSN_Faith Features".format(key)] = v_score
 
                 else:
                     print("{} Attributions faithfulness featrues ".format(key), v_score)
-                    XAI_dict["{}_faith_features".format(key)] = v_score
+                    XAI_dict["{}_Faith Features".format(key)] = v_score
 
-                    print("{}_Attributions faithfulness variable per timestep ".format(
+                    print("{}_Attributions Faithfulness Variable Per Timestep ".format(
                         key), ts_v_score)
-                    XAI_dict["{}_faith_variable_per_timestep".format(key)] = ts_v_score
-                print("{}_Data Randomization distance ".format(
+                    XAI_dict["{}_Faith Variable Per Timestep".format(key)] = ts_v_score
+                print("{}_Data Randomization Distance ".format(
                     key), r_score)
-                XAI_dict["{}_Data Randomization distance".format(key)] = r_score
+                XAI_dict["{}_Data Randomization Distance".format(key)] = r_score
 
         # Getting the interpertations using pytorch forecasting native methods
 
@@ -368,13 +365,6 @@ def create_default_mask(shape):
         with open(json_file_path, "w") as json_file:
             json.dump(XAI_dict, json_file)
 
-        # R_attribution = model.Data_Randomization(
-        #    test_loader, attributions_IG, IntegratedGradients, random_model)
-        # print('Distance Data randmoization score attribution', R_attribution)
-        # R_attention = model.Data_Randomization(
-        #    test_loader, Attention_weights["attention"], "Attention", random_model)
-        # print('Distance Data randmoization score attention', R_attention)
-
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
     save_config_file(log_dir)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index cb928a02..9fe0c682 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -19,9 +19,10 @@
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
 import matplotlib.pyplot as plt
-from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson
+from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson, cosine
 import captum
 from captum._utils.models.linear_model import SkLearnLasso
+import os
 
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
@@ -543,7 +544,7 @@ def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir
             linewidth=2,
             markersize=8,
         )
-        plt.xlabel("Time Step")
+        plt.xlabel("Feature")
         plt.ylabel("{} Attribution".format(method_name))
         plt.title("{} Attribution Values".format(method_name))
         plt.xticks(x_values, ['height', 'weight', 'age', 'sex', 'time_idx', 'alb', 'alp', 'alt', 'ast', 'be', 'bicar', 'bili', 'bili_dir', 'bnd', 'bun', 'ca', 'cai', 'ck', 'ckmb', 'cl', 'crea', 'crp', 'dbp', 'fgn', 'fio2', 'glu',
@@ -604,7 +605,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
                 r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
                                                   explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name)
-
+                print(r_score)
             elif method_name == "Random":
                 # Generate random attributions for baseline comparison
                 all_attrs = np.random.normal(size=[64, 24, 53])
@@ -641,9 +642,9 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             f_v_score = np.mean(f_v_score)
             min_val = np.min(r_score)
             max_val = np.max(r_score)
-
-            r_score = (r_score - min_val) / (max_val - min_val)
-            r_score = np.mean(r_score)
+            if method_name != "Attention":
+                r_score = (r_score - min_val) / (max_val - min_val)
+                r_score = np.mean(r_score)
             return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score
 
         # Loop through the dataloader to compute attributions for all instances
@@ -652,7 +653,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             for key, value in batch[0].items():
                 batch[0][key] = batch[0][key].to(self.device)
             x = batch[0]
-            
+
             data, baselines = self.prep_data_captum(x)
 
             # Initialize the explanation method
@@ -695,15 +696,16 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
         # Faithfulness score for attribtuons of timesteps averaged over timesteps
         f_v_score = np.mean(f_v_score)
 
-        min_val = np.min(r_score)
-        max_val = np.max(r_score)
-
-        r_score = (r_score - min_val) / (max_val - min_val)
+        # Random data score
         r_score = np.mean(r_score)
 
         if plot:
-            # Plot attributions for features
-            self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir)
+            log_dir_plots = log_dir / 'plots'
+            if not (log_dir_plots.exists()):
+                log_dir_plots.mkdir(parents=True)
+            # Plot attributions for features and timesteps
+
+            self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots)
 
         # Return computed attributions and metrics
         return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score
@@ -912,11 +914,19 @@ def Data_Randomization(
         """
 
         if similarity_func is None:
-            similarity_func = distance_euclidean
+            similarity_func = cosine
         if explain_method == "Attention":
             Attention_weights = random_model.interpertations(dataloader)
-
-            score = similarity_func(Attention_weights["attention"].cpu().numpy(), attribution.cpu().numpy())
+            attribution = attribution.cpu().numpy()
+            min_val = np.min(attribution)
+            max_val = np.max(attribution)
+
+            attribution = (attribution - min_val) / (max_val - min_val)
+            random_attr = Attention_weights["attention"].cpu().numpy()
+            min_val = np.min(random_attr)
+            max_val = np.max(random_attr)
+            random_attr = (random_attr - min_val) / (max_val - min_val)
+            score = similarity_func(random_attr, attribution)*53
         elif explain_method == "Random":
             score = similarity_func(np.random.normal(size=[64, 24, 53]).flatten(), attribution.flatten())
         else:
@@ -931,10 +941,19 @@ def Data_Randomization(
                 attr = explantation.attribute(data, **kwargs)
 
             # Process and store the calculated attributions
-            stacked_attr = attr[1].cpu().detach().numpy() if method_name in [
+            random_attr = attr[1].cpu().detach().numpy() if method_name in [
                 'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
 
-            score = similarity_func(stacked_attr.flatten(), attribution.flatten())
+            attribution = attribution.flatten()
+            min_val = np.min(attribution)
+            max_val = np.max(attribution)
+            attribution = (attribution - min_val) / (max_val - min_val)
+            random_attr = random_attr.flatten()
+            min_val = np.min(random_attr)
+            max_val = np.max(random_attr)
+            random_attr = (random_attr - min_val) / (max_val - min_val)
+
+            score = similarity_func(random_attr, attribution)
         return score
 
 
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 7a67f3c5..1fb89843 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -16,6 +16,7 @@
 from icu_benchmarks.models.utils import JsonResultLoggingEncoder
 from icu_benchmarks.wandb_utils import wandb_log
 import numpy as np
+import matplotlib.pyplot as plt
 
 
 def build_parser() -> ArgumentParser:
@@ -254,6 +255,10 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
         "CI_0.95": confidence_interval,
         "execution_time": execution_time.total_seconds() if execution_time is not None else 0.0,
     }
+    log_dir_plots = log_dir / 'plots'
+    if not (log_dir_plots.exists()):
+        log_dir_plots.mkdir(parents=True)
+    plot_XAI_Metrics(accumulated_metrics, log_dir_plots=log_dir_plots)
 
     with open(log_dir / "aggregated_test_metrics.json", "w") as f:
         json.dump(aggregated, f, cls=JsonResultLoggingEncoder)
@@ -266,6 +271,61 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
     wandb_log(json.loads(json.dumps(accumulated_metrics, cls=JsonResultLoggingEncoder)))
 
 
+def plot_XAI_Metrics(accumulated_metrics, log_dir_plots):
+    groups = {}
+    for key in accumulated_metrics['avg']:
+        if key in ['loss', 'MAE']:
+            continue
+        suffix = key.split('_')[-1]
+        if suffix not in groups:
+            groups[suffix] = []
+        groups[suffix].append(key)
+
+    # Define a dictionary for legend labels
+    legend_labels = {
+        'IG': 'Integrated Gradient',
+        'G': 'Gradient',
+        'R': 'Random',
+        'FA': 'Feature Ablation',
+        'Att': 'Attention',
+        'VSN': 'Variable Selection Network',
+        'L': 'Lime'
+    }
+
+    # Plotting
+    num_groups = len(groups)
+    fig, axs = plt.subplots(num_groups, 1, figsize=(10, num_groups * 5))
+
+    # Custom handles for the legend
+    handles = [plt.Rectangle((0, 0), 1, 1, color='none', label=f'{key}: {value}')
+               for key, value in legend_labels.items()]
+
+    for i, (suffix, keys) in enumerate(groups.items()):
+
+        ax = axs[i] if num_groups > 1 else axs
+        avg_values = [accumulated_metrics['avg'][key] for key in keys]
+        ci_lower = [accumulated_metrics['CI_0.95'][key][0] for key in keys]
+        ci_upper = [accumulated_metrics['CI_0.95'][key][1] for key in keys]
+        ci_error = [np.abs([a - b, c - a]) for a, b, c in zip(avg_values, ci_lower, ci_upper)]
+
+        bars = ax.bar(keys, np.abs(avg_values), yerr=np.array(ci_error).T, capsize=5)
+        # Modify the title to use the second suffix
+        title_suffix = keys[0].split('_')[1]
+        ax.set_title(f'Metric: "{title_suffix}"')
+        ax.set_ylabel('Values')
+        ax.axhline(0, color='grey', linewidth=0.8)
+        ax.grid(axis='y')
+        # Modify x-axis labels to show only the prefix
+        ax.set_xticks(keys)
+        ax.set_xticklabels([key.split('_')[0] for key in keys], rotation=45, ha="right")
+        # Create a custom legend for each subplot
+        custom_labels = [legend_labels[key.split('_')[0]] for key in keys]
+        ax.legend(bars, custom_labels, loc='upper right')
+
+    plt.tight_layout()
+    plt.savefig(log_dir_plots / "metrics_plot.png", bbox_inches="tight")
+
+
 def name_datasets(train="default", val="default", test="default"):
     """Names the datasets for logging (optional)."""
     gin.bind_parameter("train_common.dataset_names", {"train": train, "val": val, "test": test})
diff --git a/output.txt b/output.txt
deleted file mode 100644
index c197bcbd..00000000
--- a/output.txt
+++ /dev/null
@@ -1,475 +0,0 @@
-{'encoder_cat': tensor([], size=(64, 6, 0), dtype=torch.int64), 'encoder_cont': tensor([[[ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        [[-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
-         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
-         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        [[ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        ...,
-
-        [[ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        [[ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205,  0.4882, -1.0575,  ...,  0.0569, -1.9736, -4.7018],
-         [ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        [[-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067, -4.7018],
-         [-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067,  0.2127],
-         [-0.5397,  1.7179,  0.7785,  ...,  0.0569, -1.9736,  0.2127],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]]]), 'encoder_target': tensor([[0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 1, 1, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [1, 1, 1, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0]]), 'encoder_lengths': tensor([3, 3, 3, 3, 3, 6, 6, 6, 6, 6, 3, 6, 6, 6, 6, 3, 3, 3, 3, 5, 6, 6, 3, 3,
-        6, 6, 6, 6, 3, 6, 6, 6, 6, 6, 3, 6, 6, 6, 6, 3, 6, 6, 6, 3, 6, 3, 6, 6,
-        6, 6, 3, 3, 6, 3, 3, 5, 6, 6, 5, 3, 6, 3, 3, 3]), 'decoder_cat': tensor([], size=(64, 12, 0), dtype=torch.int64), 'decoder_cont': tensor([[[ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205, -1.3078, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         ...,
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        [[-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
-         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
-         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
-         ...,
-         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067,  0.2127],
-         [-0.9524, -0.5048, -0.2978,  ...,  0.0569,  0.5067, -4.7018],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        [[ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205, -0.6483, -2.2605,  ...,  0.0569,  0.5067,  0.2127],
-         ...,
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        ...,
-
-        [[ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.3433,  0.4882,  1.6016,  ...,  0.0569,  0.5067,  0.2127],
-         ...,
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        [[ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
-         [ 0.1205,  0.4882, -1.0575,  ...,  0.0569,  0.5067,  0.2127],
-         ...,
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]],
-
-        [[-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067,  0.2127],
-         [-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067,  0.2127],
-         [-0.5397,  1.7179,  0.7785,  ...,  0.0569,  0.5067,  0.2127],
-         ...,
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000],
-         [ 0.0000,  0.0000,  0.0000,  ...,  0.0000,  0.0000,  0.0000]]]), 'decoder_target': tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]), 'decoder_lengths': tensor([ 3, 11,  8,  8,  9, 12, 12, 12, 12, 12, 10, 12, 12, 12, 12,  7,  6,  3,
-         8, 12, 12, 12,  2, 12, 12, 12, 12, 12,  5, 12, 12, 12, 12, 12, 11, 12,
-        12, 12, 12,  8, 12, 12, 12,  3, 12,  7, 12, 12, 12, 12,  9,  6, 12, 12,
-         4, 12, 12, 12, 12,  7, 12,  7,  5,  4]), 'decoder_time_idx': tensor([[ 10,  11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [ 17,  18,  19,  20,  21,  22,  23,  24,  25,  26,  27,  28],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [ 15,  16,  17,  18,  19,  20,  21,  22,  23,  24,  25,  26],
-        [ 45,  46,  47,  48,  49,  50,  51,  52,  53,  54,  55,  56],
-        [ 48,  49,  50,  51,  52,  53,  54,  55,  56,  57,  58,  59],
-        [ 55,  56,  57,  58,  59,  60,  61,  62,  63,  64,  65,  66],
-        [ 11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21,  22],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [ 36,  37,  38,  39,  40,  41,  42,  43,  44,  45,  46,  47],
-        [128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139],
-        [  6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16,  17],
-        [ 72,  73,  74,  75,  76,  77,  78,  79,  80,  81,  82,  83],
-        [ 86,  87,  88,  89,  90,  91,  92,  93,  94,  95,  96,  97],
-        [106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117],
-        [ 24,  25,  26,  27,  28,  29,  30,  31,  32,  33,  34,  35],
-        [ 67,  68,  69,  70,  71,  72,  73,  74,  75,  76,  77,  78],
-        [  5,   6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16],
-        [ 39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49,  50],
-        [ 38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [  8,   9,  10,  11,  12,  13,  14,  15,  16,  17,  18,  19],
-        [  7,   8,   9,  10,  11,  12,  13,  14,  15,  16,  17,  18],
-        [ 62,  63,  64,  65,  66,  67,  68,  69,  70,  71,  72,  73],
-        [ 34,  35,  36,  37,  38,  39,  40,  41,  42,  43,  44,  45],
-        [ 26,  27,  28,  29,  30,  31,  32,  33,  34,  35,  36,  37],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125],
-        [ 38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49],
-        [125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136],
-        [ 33,  34,  35,  36,  37,  38,  39,  40,  41,  42,  43,  44],
-        [109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120],
-        [  6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16,  17],
-        [ 23,  24,  25,  26,  27,  28,  29,  30,  31,  32,  33,  34],
-        [ 61,  62,  63,  64,  65,  66,  67,  68,  69,  70,  71,  72],
-        [ 38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49],
-        [ 10,  11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21],
-        [ 14,  15,  16,  17,  18,  19,  20,  21,  22,  23,  24,  25],
-        [  6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16,  17],
-        [ 59,  60,  61,  62,  63,  64,  65,  66,  67,  68,  69,  70],
-        [ 35,  36,  37,  38,  39,  40,  41,  42,  43,  44,  45,  46],
-        [ 38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49],
-        [ 25,  26,  27,  28,  29,  30,  31,  32,  33,  34,  35,  36],
-        [ 57,  58,  59,  60,  61,  62,  63,  64,  65,  66,  67,  68],
-        [  8,   9,  10,  11,  12,  13,  14,  15,  16,  17,  18,  19],
-        [ 42,  43,  44,  45,  46,  47,  48,  49,  50,  51,  52,  53],
-        [ 69,  70,  71,  72,  73,  74,  75,  76,  77,  78,  79,  80],
-        [ 28,  29,  30,  31,  32,  33,  34,  35,  36,  37,  38,  39],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [ 46,  47,  48,  49,  50,  51,  52,  53,  54,  55,  56,  57],
-        [ 37,  38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48],
-        [ 30,  31,  32,  33,  34,  35,  36,  37,  38,  39,  40,  41],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [  5,   6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16],
-        [ 57,  58,  59,  60,  61,  62,  63,  64,  65,  66,  67,  68],
-        [ 80,  81,  82,  83,  84,  85,  86,  87,  88,  89,  90,  91],
-        [ 44,  45,  46,  47,  48,  49,  50,  51,  52,  53,  54,  55],
-        [ 10,  11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21],
-        [ 92,  93,  94,  95,  96,  97,  98,  99, 100, 101, 102, 103],
-        [  3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14],
-        [ 10,  11,  12,  13,  14,  15,  16,  17,  18,  19,  20,  21],
-        [113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124]]), 'groups': tensor([[173],
-        [283],
-        [107],
-        [ 34],
-        [388],
-        [429],
-        [126],
-        [376],
-        [265],
-        [213],
-        [314],
-        [431],
-        [185],
-        [360],
-        [247],
-        [203],
-        [245],
-        [222],
-        [299],
-        [ 43],
-        [190],
-        [181],
-        [345],
-        [ 16],
-        [278],
-        [452],
-        [395],
-        [518],
-        [564],
-        [266],
-        [ 76],
-        [187],
-        [464],
-        [266],
-        [527],
-        [ 11],
-        [315],
-        [ 42],
-        [513],
-        [105],
-        [468],
-        [313],
-        [188],
-        [525],
-        [201],
-        [554],
-        [559],
-        [551],
-        [181],
-        [549],
-        [335],
-        [382],
-        [369],
-        [436],
-        [389],
-        [145],
-        [475],
-        [442],
-        [ 43],
-        [527],
-        [207],
-        [554],
-        [331],
-        [239]]), 'target_scale': tensor([[0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.],
-        [0., 0.]])}
-(tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
-        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]), None)

From 08c3c31a3179e49e4906dbff13b23f60e4e946b8 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 4 Dec 2023 18:20:22 +0100
Subject: [PATCH 118/142] added stability metric

---
 icu_benchmarks/models/train.py    |  22 +++--
 icu_benchmarks/models/wrappers.py | 151 ++++++++++++++++++++++++++++--
 2 files changed, 155 insertions(+), 18 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 14517da7..b22a6438 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -281,17 +281,17 @@ def train_common(
 
         # choose which  methods to get attributions
         methods = {
-            "G": Saliency,
-            "L": Lime,
-            "IG": IntegratedGradients,
-            "FA": FeatureAblation,
-            "R": "Random",
+            # "G": Saliency,
+            # "L": Lime,
+            # "IG": IntegratedGradients,
+            # "FA": FeatureAblation,
+            # "R": "Random",
             "Att": "Attention"
         }
         for key, item in methods.items():
             # If conditions needed here as different explantations require different inputs
             if key == "IG":
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score, st_i_score, st_o_score = model.explantation(
                     dataloader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=50, XAI_metric=XAI_metric, random_model=random_model
                 )
             elif key == "L" or key == "FA":
@@ -328,12 +328,12 @@ def create_default_mask(shape):
                 # Create a tuple of masks
                 feature_masks = tuple([create_default_mask(shape) if i !=
                                       1 else feature_mask_second for i, shape in enumerate(shapes)])
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score, st_i_score, st_o_score = model.explantation(
                     dataloader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True, XAI_metric=XAI_metric, random_model=random_model
                 )
 
             else:
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score = model.explantation(
+                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score, st_i_score, st_o_score = model.explantation(
                     dataloader=test_loader, method=item, log_dir=log_dir, plot=True, XAI_metric=XAI_metric, random_model=random_model
                 )
 
@@ -341,6 +341,12 @@ def create_default_mask(shape):
                 # logging metric scores
                 print("{} Attributions Faithfulness Timesteps ".format(key), ts_score)
                 XAI_dict["{}_Faith Timesteps".format(key)] = ts_score
+                print("{}_ROS ".format(
+                    key), st_o_score)
+                XAI_dict["{}_ROS".format(key)] = st_o_score
+                print("{}_RIS ".format(
+                    key), st_i_score)
+                XAI_dict["{}_RIS".format(key)] = st_i_score
                 if key == "Att":
                     print("Variable selection weights faithfulness featrues ".format(key), v_score)
                     XAI_dict["VSN_Faith Features".format(key)] = v_score
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 9fe0c682..96ffe0ce 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -594,6 +594,8 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
         f_ts_v_score = []
         f_v_score = []
         r_score = []
+        st_i_score = []
+        st_o_score = []
 
         method_name = method if (method == "Random") or (method == "Attention") else (
             method.__name__)
@@ -605,7 +607,9 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
                 r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
                                                   explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name)
-                print(r_score)
+                st_i_score, st_o_score = self.Relative_Stability(x=None,
+                                                                 attribution=timestep_attrs, explain_method=method, method_name=method_name, dataloader=dataloader, **kwargs
+                                                                 )
             elif method_name == "Random":
                 # Generate random attributions for baseline comparison
                 all_attrs = np.random.normal(size=[64, 24, 53])
@@ -620,7 +624,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
 
                     if method_name == "Random":
 
-                        f_ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                        """  f_ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
                                                                           pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
                         f_ts_score.append(self.Faithfulness_Correlation(x, all_attrs,
                                                                         pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
@@ -628,7 +632,12 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                                                                        pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
 
                         r_score.append(self.Data_Randomization(x, attribution=all_attrs,
-                                       explain_method=method, random_model=random_model, method_name=method_name))
+                                       explain_method=method, random_model=random_model, method_name=method_name)) """
+                        res1, res2 = self.Relative_Stability(x,
+                                                             all_attrs, explain_method=method, method_name=method_name, dataloader=None, **kwargs
+                                                             )
+                        st_i_score.append(res1)
+                        st_o_score.append(res2)
                     else:
                         f_ts_score.append(self.Faithfulness_Correlation(x, timestep_attrs,
                                                                         pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
@@ -640,12 +649,14 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             # Faithfulness score for attribtuons of timesteps averaged over features
             f_ts_score = np.mean(f_ts_score)
             f_v_score = np.mean(f_v_score)
-            min_val = np.min(r_score)
-            max_val = np.max(r_score)
+            # min_val = np.min(r_score)
+            # max_val = np.max(r_score)
             if method_name != "Attention":
-                r_score = (r_score - min_val) / (max_val - min_val)
+                # r_score = (r_score - min_val) / (max_val - min_val)
                 r_score = np.mean(r_score)
-            return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score
+                st_i_score = np.max(st_i_score)
+                st_o_score = np.max(st_o_score)
+            return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score, st_i_score, st_o_score
 
         # Loop through the dataloader to compute attributions for all instances
         for batch in dataloader:
@@ -671,7 +682,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
             if XAI_metric:
 
-                f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                """ f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
                                                                   pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
 
                 f_ts_score.append(self.Faithfulness_Correlation(x, stacked_attr,
@@ -679,7 +690,12 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 f_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
                                                                pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
                 r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
-                               explain_method=method, random_model=random_model, method_name=method_name))
+                               explain_method=method, random_model=random_model, method_name=method_name)) """
+                res1, res2 = self.Relative_Stability(x,
+                                                     stacked_attr, explain_method=method, method_name=method_name,  dataloader=None, **kwargs
+                                                     )
+                st_i_score.append(res1)
+                st_o_score.append(res2)
             # aggregate over batch
             attr = np.mean(stacked_attr, axis=0)
             all_attrs.append(attr)
@@ -698,6 +714,8 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
 
         # Random data score
         r_score = np.mean(r_score)
+        st_i_score = np.max(st_i_score)
+        st_o_score = np.max(st_o_score)
 
         if plot:
             log_dir_plots = log_dir / 'plots'
@@ -708,7 +726,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots)
 
         # Return computed attributions and metrics
-        return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score
+        return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score, st_i_score, st_o_score
         # normalized_means = (means - means.min()) / (means.max() - means.min())
 
     def prep_data(self, x):
@@ -956,6 +974,119 @@ def Data_Randomization(
             score = similarity_func(random_attr, attribution)
         return score
 
+    def Relative_Stability(self, x,
+                           attribution, explain_method, method_name, dataloader=None, **kwargs
+                           ):
+        """
+    Computes relative input and output stabilities maximization objective
+    as defined here :
+    ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
+    Code mainly by:
+    Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+        """
+
+        def relative_stability_objective(
+            x: np.ndarray, xs: np.ndarray, e_x: np.ndarray, e_xs: np.ndarray, eps_min=0.0001, input=False
+        ) -> np.ndarray:
+            """
+            Computes relative input stabilities maximization objective
+            as defined here :ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
+
+            Parameters
+            ----------
+            x: np.ndarray
+                Batch of images.
+            xs: np.ndarray
+                Batch of perturbed images.
+            e_x: np.ndarray
+                Explanations for x.
+            e_xs: np.ndarray
+                Explanations for xs.
+
+            Returns
+            -------
+            ris_obj: np.ndarray
+                RIS maximization objective.
+            """
+
+            if input:
+                num_dim = x.ndim
+            else:
+                num_dim = e_x.ndim
+
+            if num_dim == 3:
+                def norm_function(arr): return np.linalg.norm(arr, axis=(-1, -2))  # noqa
+            elif num_dim == 2:
+                def norm_function(arr): return np.linalg.norm(arr, axis=-1)
+            else:
+                raise ValueError(
+                    "Relative Input Stability only supports 4D, 3D and 2D inputs (batch dimension inclusive)."
+                )
+
+            # fmt: off
+
+            nominator = (e_x - e_xs) / (e_x + (e_x == 0) * eps_min)  # prevent division by 0
+            nominator = norm_function(nominator)
+            # fmt: on
+            if input:
+                denominator = x - xs
+                denominator /= x + (x == 0) * eps_min
+                # fmt: off
+                denominator = norm_function(denominator)
+                # fmt: on
+                denominator += (denominator == 0) * eps_min
+            else:
+
+                denominator = np.squeeze(x) - np.squeeze(xs)
+                denominator = np.linalg.norm(denominator, axis=-1)
+                denominator += (denominator == 0) * eps_min  # prevent division by 0
+
+            return nominator / denominator
+
+        y_pred = self(self.prep_data(x)).detach()
+
+        if explain_method == "Attention":
+            with torch.no_grad():
+                noise = torch.randn_like(self.dataset["encoder_cont"])*0.1
+                self.dataset += noise
+
+            Attention_weights = self.interpertations(dataloader)
+            att_preturb = Attention_weights["attention"].cpu().numpy()
+
+        else:
+            y_pred = self(self.prep_data(x)).detach()
+            x_original = x["encoder_cont"].detach().clone()
+            y_pred_preturb = self(self.prep_data(x)).detach()
+            with torch.no_grad():
+                noise = torch.randn_like(x["encoder_cont"])*0.01
+                x["encoder_cont"] += noise
+
+            if explain_method == "Random":
+                att_preturb = np.random.normal(size=[64, 24, 53])
+
+            else:
+
+                data, baselines = self.prep_data_captum(x)
+
+                explantation = explain_method(self.forward_captum)
+                # Reformat attributions.
+                if explain_method is not captum.attr.Saliency:
+                    att_preturb = explantation.attribute(data, baselines=baselines, **kwargs)
+                else:
+                    att_preturb = explantation.attribute(data, **kwargs)
+
+                # Process and store the calculated attributions
+                att_preturb = att_preturb[1].cpu().detach().numpy() if method_name in [
+                    'Lime', 'FeatureAblation'] else torch.stack(att_preturb).cpu().detach().numpy()
+        RIS = relative_stability_objective(
+            x_original.detach().cpu().numpy(), x["encoder_cont"].detach().cpu().numpy(), attribution, att_preturb, input=True
+        )
+        ROS = relative_stability_objective(
+            y_pred.cpu().numpy(), y_pred_preturb.cpu().numpy(), attribution, att_preturb, input=False
+        )
+
+        return np.max(RIS), np.max(ROS)
+
 
 @gin.configurable("MLWrapper")
 class MLWrapper(BaseModule, ABC):

From 4893f3ea976c37a4a1ea95659d718308dc7bb9b4 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Tue, 5 Dec 2023 17:32:38 +0100
Subject: [PATCH 119/142] attention ROS RIS

---
 configs/prediction_models/DeepARpytorch.gin |  2 +-
 configs/prediction_models/RNNpytorch.gin    |  2 +-
 icu_benchmarks/data/loader.py               |  6 ++-
 icu_benchmarks/models/dl_models.py          |  2 +-
 icu_benchmarks/models/train.py              |  4 +-
 icu_benchmarks/models/wrappers.py           | 55 ++++++++++++---------
 icu_benchmarks/run_utils.py                 | 25 +++++++---
 7 files changed, 61 insertions(+), 35 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 5135035f..e01083ff 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -13,7 +13,7 @@ optimizer/hyperparameter.lr = (1e-5, 1e-3)
 # Model params
 
 model/hyperparameter.class_to_tune = @DeepARpytorch
-model/hyperparameter.hidden = (40, 120, "log-uniform", 2)
+model/hyperparameter.hidden = (4, 64, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1, 3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 0b88c670..486e5b21 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -13,7 +13,7 @@ optimizer/hyperparameter.lr = (1e-5, 3e-4)
 # Model params
 
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = (32, 256, "log-uniform", 2)
+model/hyperparameter.hidden = (2, 64, "log-uniform", 2)
 model/hyperparameter.rnn_layers=(1,3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 96ba9695..1b3e72ad 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -3,7 +3,7 @@
 import pandas as pd
 import gin
 import numpy as np
-from torch import Tensor, cat, from_numpy, float32, min, max
+from torch import Tensor, cat, from_numpy, float32, min, max, randn_like
 from torch.utils.data import Dataset
 import logging
 from typing import Dict, Tuple, Union
@@ -531,3 +531,7 @@ def randomize_labels(self, num_classes=None, min=None, max=None):
         else:
             random_target = np.random.randint(num_classes, size=len(self.data["target"][0]))
         self.data["target"][0] = Tensor(random_target)
+
+    def add_noise(self, num_classes=None, min=None, max=None):
+        noise = randn_like(self.data["reals"])*0.1
+        self.data["reals"] += noise
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 57107b60..ff66c4da 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -476,7 +476,7 @@ def __init__(
         **kwargs,
     ):
         super().__init__(optimizer=optimizer, pytorch_forecasting=True, *args, **kwargs)
-
+        self.dataset = dataset
         self.model = TemporalFusionTransformer.from_dataset(
             dataset=dataset,
             hidden_size=hidden,
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index b22a6438..943a533e 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -342,10 +342,10 @@ def create_default_mask(shape):
                 print("{} Attributions Faithfulness Timesteps ".format(key), ts_score)
                 XAI_dict["{}_Faith Timesteps".format(key)] = ts_score
                 print("{}_ROS ".format(
-                    key), st_o_score)
+                    key), st_o_score, type(st_o_score))
                 XAI_dict["{}_ROS".format(key)] = st_o_score
                 print("{}_RIS ".format(
-                    key), st_i_score)
+                    key), st_i_score, type(st_i_score))
                 XAI_dict["{}_RIS".format(key)] = st_i_score
                 if key == "Att":
                     print("Variable selection weights faithfulness featrues ".format(key), v_score)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 96ffe0ce..50ac9c4d 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -605,8 +605,8 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 timestep_attrs = Interpertations["attention"]
                 features_attrs = Interpertations["static_variables"].tolist()
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
-                r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
-                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name)
+                """ r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
+                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name) """
                 st_i_score, st_o_score = self.Relative_Stability(x=None,
                                                                  attribution=timestep_attrs, explain_method=method, method_name=method_name, dataloader=dataloader, **kwargs
                                                                  )
@@ -649,8 +649,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             # Faithfulness score for attribtuons of timesteps averaged over features
             f_ts_score = np.mean(f_ts_score)
             f_v_score = np.mean(f_v_score)
-            # min_val = np.min(r_score)
-            # max_val = np.max(r_score)
+
             if method_name != "Attention":
                 # r_score = (r_score - min_val) / (max_val - min_val)
                 r_score = np.mean(r_score)
@@ -944,7 +943,7 @@ def Data_Randomization(
             min_val = np.min(random_attr)
             max_val = np.max(random_attr)
             random_attr = (random_attr - min_val) / (max_val - min_val)
-            score = similarity_func(random_attr, attribution)*53
+            score = similarity_func(random_attr, attribution)
         elif explain_method == "Random":
             score = similarity_func(np.random.normal(size=[64, 24, 53]).flatten(), attribution.flatten())
         else:
@@ -989,7 +988,7 @@ def relative_stability_objective(
             x: np.ndarray, xs: np.ndarray, e_x: np.ndarray, e_xs: np.ndarray, eps_min=0.0001, input=False
         ) -> np.ndarray:
             """
-            Computes relative input stabilities maximization objective
+            Computes relative input and output stabilities maximization objective
             as defined here :ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
 
             Parameters
@@ -1019,9 +1018,7 @@ def norm_function(arr): return np.linalg.norm(arr, axis=(-1, -2))  # noqa
             elif num_dim == 2:
                 def norm_function(arr): return np.linalg.norm(arr, axis=-1)
             else:
-                raise ValueError(
-                    "Relative Input Stability only supports 4D, 3D and 2D inputs (batch dimension inclusive)."
-                )
+                def norm_function(arr): return np.linalg.norm(arr)
 
             # fmt: off
 
@@ -1043,16 +1040,30 @@ def norm_function(arr): return np.linalg.norm(arr, axis=-1)
 
             return nominator / denominator
 
-        y_pred = self(self.prep_data(x)).detach()
-
         if explain_method == "Attention":
-            with torch.no_grad():
-                noise = torch.randn_like(self.dataset["encoder_cont"])*0.1
-                self.dataset += noise
-
+            y_pred = self.model.predict(dataloader)
+            x_original = dataloader.dataset.data["reals"].clone()
+
+            dataloader.dataset.add_noise()
+            x_preturb = dataloader.dataset.data["reals"].clone()
+            """ dataloader = self.dataset.to_dataloader(
+                train=False,
+                batch_size=64,
+                num_workers=8,
+                pin_memory=False,
+                drop_last=True,
+                shuffle=False,
+            )  """
+
+            y_pred_preturb = self.model.predict(dataloader)
             Attention_weights = self.interpertations(dataloader)
             att_preturb = Attention_weights["attention"].cpu().numpy()
-
+            RIS = relative_stability_objective(
+                x_original.detach().cpu().numpy(), x_preturb.detach().cpu().numpy(), attribution.cpu().numpy(), att_preturb, input=True
+            )
+            ROS = relative_stability_objective(
+                y_pred.cpu().numpy(), y_pred_preturb.cpu().numpy(), attribution.cpu().numpy(), att_preturb, input=False
+            )
         else:
             y_pred = self(self.prep_data(x)).detach()
             x_original = x["encoder_cont"].detach().clone()
@@ -1078,12 +1089,12 @@ def norm_function(arr): return np.linalg.norm(arr, axis=-1)
                 # Process and store the calculated attributions
                 att_preturb = att_preturb[1].cpu().detach().numpy() if method_name in [
                     'Lime', 'FeatureAblation'] else torch.stack(att_preturb).cpu().detach().numpy()
-        RIS = relative_stability_objective(
-            x_original.detach().cpu().numpy(), x["encoder_cont"].detach().cpu().numpy(), attribution, att_preturb, input=True
-        )
-        ROS = relative_stability_objective(
-            y_pred.cpu().numpy(), y_pred_preturb.cpu().numpy(), attribution, att_preturb, input=False
-        )
+            RIS = relative_stability_objective(
+                x_original.detach().cpu().numpy(), x["encoder_cont"].detach().cpu().numpy(), attribution, att_preturb, input=True
+            )
+            ROS = relative_stability_objective(
+                y_pred.cpu().numpy(), y_pred_preturb.cpu().numpy(), attribution, att_preturb, input=False
+            )
 
         return np.max(RIS), np.max(ROS)
 
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 1fb89843..577607ff 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -291,6 +291,7 @@ def plot_XAI_Metrics(accumulated_metrics, log_dir_plots):
         'VSN': 'Variable Selection Network',
         'L': 'Lime'
     }
+    colors = ['navy', 'skyblue', 'crimson', 'salmon', 'teal', 'orange', 'darkgreen', 'lightgreen']
 
     # Plotting
     num_groups = len(groups)
@@ -303,23 +304,33 @@ def plot_XAI_Metrics(accumulated_metrics, log_dir_plots):
     for i, (suffix, keys) in enumerate(groups.items()):
 
         ax = axs[i] if num_groups > 1 else axs
+        # Extract values and errors
         avg_values = [accumulated_metrics['avg'][key] for key in keys]
         ci_lower = [accumulated_metrics['CI_0.95'][key][0] for key in keys]
         ci_upper = [accumulated_metrics['CI_0.95'][key][1] for key in keys]
         ci_error = [np.abs([a - b, c - a]) for a, b, c in zip(avg_values, ci_lower, ci_upper)]
 
-        bars = ax.bar(keys, np.abs(avg_values), yerr=np.array(ci_error).T, capsize=5)
-        # Modify the title to use the second suffix
-        title_suffix = keys[0].split('_')[1]
+        # Sort by absolute values of avg_values
+        sorted_indices = np.argsort([np.abs(val) for val in avg_values])[::-1]  # Indices to sort in descending order
+        sorted_keys = np.array(keys)[sorted_indices]
+        sorted_avg_values = np.array(avg_values)[sorted_indices]
+        sorted_ci_error = np.array(ci_error)[sorted_indices]
+
+        # Plot bars
+        bars = ax.bar(sorted_keys, np.abs(sorted_avg_values), yerr=np.array(sorted_ci_error).T, capsize=5, color=colors)
+
+        # Set titles and labels
+        title_suffix = sorted_keys[0].split('_')[1]
         ax.set_title(f'Metric: "{title_suffix}"')
         ax.set_ylabel('Values')
         ax.axhline(0, color='grey', linewidth=0.8)
         ax.grid(axis='y')
-        # Modify x-axis labels to show only the prefix
-        ax.set_xticks(keys)
-        ax.set_xticklabels([key.split('_')[0] for key in keys], rotation=45, ha="right")
+
+        # Set x-ticks
+        ax.set_xticks(sorted_keys)
+        ax.set_xticklabels([key.split('_')[0] for key in sorted_keys], rotation=45, ha="right")
         # Create a custom legend for each subplot
-        custom_labels = [legend_labels[key.split('_')[0]] for key in keys]
+        custom_labels = [legend_labels[key.split('_')[0]] for key in sorted_keys]
         ax.legend(bars, custom_labels, loc='upper right')
 
     plt.tight_layout()

From 38ed6e335727f58fe5a36f8fad7525889f626b70 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 13 Dec 2023 15:06:31 +0100
Subject: [PATCH 120/142] added condition for ROS and RIS

---
 configs/prediction_models/DeepARpytorch.gin |  19 ++-
 configs/prediction_models/RNNpytorch.gin    |  18 ++-
 configs/prediction_models/TFTpytorch.gin    |  20 ++-
 configs/tasks/Regression.gin                |   4 +-
 configs/tasks/common/CrossValidation.gin    |   2 +-
 icu_benchmarks/data/loader.py               |   2 +-
 icu_benchmarks/models/train.py              |   8 +-
 icu_benchmarks/models/wrappers.py           | 160 ++++++++++++--------
 icu_benchmarks/run_utils.py                 |   2 +-
 9 files changed, 138 insertions(+), 97 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index e01083ff..b093027e 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -5,24 +5,27 @@ include "configs/prediction_models/common/DLCommon.gin"
 
 # Optimizer params
 train_common.model = @DeepARpytorch
-
+                         
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = (1e-5, 1e-3)
-
+optimizer/hyperparameter.lr = 0.00046
+#(1e-5, 1e-3)
 # Model params
 
 model/hyperparameter.class_to_tune = @DeepARpytorch
-model/hyperparameter.hidden = (4, 64, "log-uniform", 2)
-model/hyperparameter.rnn_layers=(1, 3)
+model/hyperparameter.hidden = 116
+#(4, 64, "log-uniform", 2)
+model/hyperparameter.rnn_layers=1
+
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
-model/hyperparameter.dropout = (0.0, 0.4)
+model/hyperparameter.dropout = (0.286968842146375, 0.29)
 #model/hyperparameter.lr_scheduler = "exponential"
 
 train_common/hyperparameter.class_to_tune = @train_common
-train_common/hyperparameter.batch_size=(32,64,128,256,512)
-train_common/hyperparameter.gradient_clip_val=(0,0.01, 1.0, 100.0)
+train_common/hyperparameter.batch_size=256
+train_common/hyperparameter.gradient_clip_val=0.01
+#(0,0.01, 1.0, 100.0)
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 486e5b21..050f11a5 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -8,21 +8,25 @@ train_common.model = @RNNpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = (1e-5, 3e-4)
+optimizer/hyperparameter.lr = 0.00016
 
 # Model params
-
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden = (2, 64, "log-uniform", 2)
-model/hyperparameter.rnn_layers=(1,3)
+model/hyperparameter.hidden =  42
+#(2, 64, "log-uniform", 2)
+model/hyperparameter.rnn_layers=2 
+#(1,3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
-model/hyperparameter.dropout = (0.0, 0.4)
+model/hyperparameter.dropout = (0.3976, 0.3978)
+#(0.0, 0.4)
 #model/hyperparameter.lr_scheduler = "exponential"
 
 train_common/hyperparameter.class_to_tune = @train_common
-train_common/hyperparameter.batch_size=(32,64,128,256,512)
-train_common/hyperparameter.gradient_clip_val=(0,0.01, 1.0, 100.0)
+train_common/hyperparameter.batch_size=128
+#(32,64,128,256,512)
+train_common/hyperparameter.gradient_clip_val=0.01
+#(0,0.01, 1.0, 100.0)
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 18eabb6a..19130460 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -8,23 +8,27 @@ train_common.model = @TFTpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = (0.0001, 0.001,0.01)
-
+optimizer/hyperparameter.lr = 0.001
+#(0.0001, 0.001,0.01)
 
 # Model params
 model/hyperparameter.class_to_tune = @TFTpytorch
-model/hyperparameter.hidden = (10,20,40,80,160,240,320)
+model/hyperparameter.hidden = 32
+#(10,20,40,80,160,240,320)
 model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.dropout =(0.1, 0.9)
-model/hyperparameter.dropout_att = (0.1, 0.9)
-model/hyperparameter.n_heads =(1,4)
+model/hyperparameter.dropout =(0.37, 0.38)
+model/hyperparameter.dropout_att = 0.0589
+#(0.1, 0.9)
+model/hyperparameter.n_heads =2
 
 #dataloader
 
 train_common/hyperparameter.class_to_tune = @train_common
-train_common/hyperparameter.batch_size=(32,64,128,256,512)
+train_common/hyperparameter.batch_size=64
+#(32,64,128,256,512)
 
-train_common/hyperparameter.gradient_clip_val=(0,0.01, 1.0, 100.0)
+train_common/hyperparameter.gradient_clip_val=0.0
+#(0,0.01, 1.0, 100.0)
 
 
 
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index 1234f707..75041763 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -15,8 +15,8 @@ train_common.weight = "balanced"
 train_common.ram_cache = True
 
 # LOSS FUNCTION
-DLPredictionWrapper.loss = @mse_loss
-DLPredictionPytorchForecastingWrapper.loss = @mse_loss
+DLPredictionWrapper.loss = @l1_loss
+DLPredictionPytorchForecastingWrapper.loss = @l1_loss
 MLWrapper.loss = @mean_squared_error
 
 # SELECTING PREPROCESSOR
diff --git a/configs/tasks/common/CrossValidation.gin b/configs/tasks/common/CrossValidation.gin
index d0c4a704..519ad3dd 100644
--- a/configs/tasks/common/CrossValidation.gin
+++ b/configs/tasks/common/CrossValidation.gin
@@ -1,3 +1,3 @@
 # CROSS-VALIDATION SETTINGS
-execute_repeated_cv.cv_repetitions = 1
+execute_repeated_cv.cv_repetitions = 2
 execute_repeated_cv.cv_folds = 5
\ No newline at end of file
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 1b3e72ad..a0e51dbc 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -533,5 +533,5 @@ def randomize_labels(self, num_classes=None, min=None, max=None):
         self.data["target"][0] = Tensor(random_target)
 
     def add_noise(self, num_classes=None, min=None, max=None):
-        noise = randn_like(self.data["reals"])*0.1
+        noise = randn_like(self.data["reals"])*0.01
         self.data["reals"] += noise
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 943a533e..7097571b 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -285,8 +285,8 @@ def train_common(
             # "L": Lime,
             # "IG": IntegratedGradients,
             # "FA": FeatureAblation,
-            # "R": "Random",
-            "Att": "Attention"
+            "R": "Random",
+            # "Att": "Attention"
         }
         for key, item in methods.items():
             # If conditions needed here as different explantations require different inputs
@@ -342,10 +342,10 @@ def create_default_mask(shape):
                 print("{} Attributions Faithfulness Timesteps ".format(key), ts_score)
                 XAI_dict["{}_Faith Timesteps".format(key)] = ts_score
                 print("{}_ROS ".format(
-                    key), st_o_score, type(st_o_score))
+                    key), st_o_score)
                 XAI_dict["{}_ROS".format(key)] = st_o_score
                 print("{}_RIS ".format(
-                    key), st_i_score, type(st_i_score))
+                    key), st_i_score)
                 XAI_dict["{}_RIS".format(key)] = st_i_score
                 if key == "Att":
                     print("Variable selection weights faithfulness featrues ".format(key), v_score)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 50ac9c4d..b3c76a40 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -605,8 +605,8 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 timestep_attrs = Interpertations["attention"]
                 features_attrs = Interpertations["static_variables"].tolist()
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
-                """ r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
-                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name) """
+                r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
+                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name)
                 st_i_score, st_o_score = self.Relative_Stability(x=None,
                                                                  attribution=timestep_attrs, explain_method=method, method_name=method_name, dataloader=dataloader, **kwargs
                                                                  )
@@ -624,15 +624,15 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
 
                     if method_name == "Random":
 
-                        """  f_ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                        """ f_ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
                                                                           pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
                         f_ts_score.append(self.Faithfulness_Correlation(x, all_attrs,
                                                                         pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
                         f_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
-                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100)) 
 
                         r_score.append(self.Data_Randomization(x, attribution=all_attrs,
-                                       explain_method=method, random_model=random_model, method_name=method_name)) """
+                                       explain_method=method, random_model=random_model, method_name=method_name))"""
                         res1, res2 = self.Relative_Stability(x,
                                                              all_attrs, explain_method=method, method_name=method_name, dataloader=None, **kwargs
                                                              )
@@ -681,7 +681,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
             if XAI_metric:
 
-                """ f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
                                                                   pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
 
                 f_ts_score.append(self.Faithfulness_Correlation(x, stacked_attr,
@@ -689,7 +689,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 f_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
                                                                pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
                 r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
-                               explain_method=method, random_model=random_model, method_name=method_name)) """
+                               explain_method=method, random_model=random_model, method_name=method_name))
                 res1, res2 = self.Relative_Stability(x,
                                                      stacked_attr, explain_method=method, method_name=method_name,  dataloader=None, **kwargs
                                                      )
@@ -914,9 +914,22 @@ def apply_baseline(x, indices, time_step, feature_timestep):
 
     def Data_Randomization(
         self, x,
-        attribution, explain_method, random_model, similarity_func=None, dataloader=None, method_name="", **kwargs
+        attribution, explain_method, random_model, similarity_func=cosine, dataloader=None, method_name="", **kwargs
     ):
         """
+
+        Args:
+            - x:Batch input
+            -attribution: attribution 
+            - explain_method:function to generate explantations
+            - random_model: Reference to model trained on random labels
+            - similarity_func: Function to measure similiarity 
+            - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+            - method_name: Name of the explantation
+
+        Returns:
+            score: similarity score between attributions of model trained on random data and model trained on real data
+
         Implementation of the Random Logit Metric by Sixt et al., 2020.
 
         The Random Logit Metric computes the distance between the original explanation and a reference explanation of
@@ -930,8 +943,6 @@ def Data_Randomization(
 
         """
 
-        if similarity_func is None:
-            similarity_func = cosine
         if explain_method == "Attention":
             Attention_weights = random_model.interpertations(dataloader)
             attribution = attribution.cpu().numpy()
@@ -974,39 +985,58 @@ def Data_Randomization(
         return score
 
     def Relative_Stability(self, x,
-                           attribution, explain_method, method_name, dataloader=None, **kwargs
+                           attribution, explain_method, method_name, dataloader=None, thershold=0.5, **kwargs
                            ):
         """
-    Computes relative input and output stabilities maximization objective
-    as defined here :
-    ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
-    Code mainly by:
-    Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+     Args:
+            - x:Batch input
+            -attribution: attribution 
+            - explain_method:function to generate explantations
+            - method_name: Name of the explantation
+            - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+
+
+        Returns:
+            RIS : relative distance between the explantation and the input
+            ROS: relative distance between the explantation and the output
+
+
+     References:
+            1) `https://arxiv.org/pdf/2203.06877.pdf
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+
         """
 
         def relative_stability_objective(
-            x: np.ndarray, xs: np.ndarray, e_x: np.ndarray, e_xs: np.ndarray, eps_min=0.0001, input=False
-        ) -> np.ndarray:
+            x, xs, e_x, e_xs, eps_min=0.0001, input=False, device='cuda'
+        ) -> torch.Tensor:
             """
             Computes relative input and output stabilities maximization objective
             as defined here :ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
 
-            Parameters
-            ----------
-            x: np.ndarray
-                Batch of images.
-            xs: np.ndarray
-                Batch of perturbed images.
-            e_x: np.ndarray
-                Explanations for x.
-            e_xs: np.ndarray
-                Explanations for xs.
-
-            Returns
-            -------
-            ris_obj: np.ndarray
-                RIS maximization objective.
+            Args:
+
+                x: Input tensor
+                xs: perturbed tensor.
+                e_x: Explanations for x.
+                e_xs: Explanations for xs.
+                eps_min:Value to avoid division by zero if needed 
+                input:Boolean to indicate if this is an input or an output
+                device: the device to keep the tensors on 
+
+            Returns:
+
+                ris_obj: Tensor
+                    RIS maximization objective.
             """
+            # Function to convert inputs to tensors if they are numpy arrays
+            def to_tensor(input_array):
+                if isinstance(input_array, np.ndarray):
+                    return torch.tensor(input_array).to(device)
+                return input_array.to(device)
+
+            # Convert all inputs to tensors and move to GPU
+            x, xs, e_x, e_xs = map(to_tensor, [x, xs, e_x, e_xs])
 
             if input:
                 num_dim = x.ndim
@@ -1014,29 +1044,24 @@ def relative_stability_objective(
                 num_dim = e_x.ndim
 
             if num_dim == 3:
-                def norm_function(arr): return np.linalg.norm(arr, axis=(-1, -2))  # noqa
+                def norm_function(arr): return torch.norm(arr, dim=(-1, -2))
             elif num_dim == 2:
-                def norm_function(arr): return np.linalg.norm(arr, axis=-1)
+                def norm_function(arr): return torch.norm(arr, dim=-1)
             else:
-                def norm_function(arr): return np.linalg.norm(arr)
-
-            # fmt: off
+                def norm_function(arr): return torch.norm(arr)
 
-            nominator = (e_x - e_xs) / (e_x + (e_x == 0) * eps_min)  # prevent division by 0
+            nominator = (e_x - e_xs) / (e_x + (e_x == 0) * eps_min)
             nominator = norm_function(nominator)
-            # fmt: on
+
             if input:
                 denominator = x - xs
                 denominator /= x + (x == 0) * eps_min
-                # fmt: off
                 denominator = norm_function(denominator)
-                # fmt: on
                 denominator += (denominator == 0) * eps_min
             else:
-
-                denominator = np.squeeze(x) - np.squeeze(xs)
-                denominator = np.linalg.norm(denominator, axis=-1)
-                denominator += (denominator == 0) * eps_min  # prevent division by 0
+                denominator = torch.squeeze(x) - torch.squeeze(xs)
+                denominator = torch.norm(denominator, dim=-1)
+                denominator += (denominator == 0) * eps_min
 
             return nominator / denominator
 
@@ -1046,32 +1071,29 @@ def norm_function(arr): return np.linalg.norm(arr)
 
             dataloader.dataset.add_noise()
             x_preturb = dataloader.dataset.data["reals"].clone()
-            """ dataloader = self.dataset.to_dataloader(
-                train=False,
-                batch_size=64,
-                num_workers=8,
-                pin_memory=False,
-                drop_last=True,
-                shuffle=False,
-            )  """
-
             y_pred_preturb = self.model.predict(dataloader)
             Attention_weights = self.interpertations(dataloader)
-            att_preturb = Attention_weights["attention"].cpu().numpy()
+            att_preturb = Attention_weights["attention"]
+            # Calculate the absolute difference
+            difference = torch.abs(y_pred_preturb - y_pred)
+
+            # Find where the difference is less than or equal to a thershold
+            close_indices = torch.nonzero(difference <= thershold).squeeze()
             RIS = relative_stability_objective(
-                x_original.detach().cpu().numpy(), x_preturb.detach().cpu().numpy(), attribution.cpu().numpy(), att_preturb, input=True
+                x_original[close_indices, :, :].detach(), x_preturb[close_indices, :, :].detach(), attribution, att_preturb, input=True
             )
             ROS = relative_stability_objective(
-                y_pred.cpu().numpy(), y_pred_preturb.cpu().numpy(), attribution.cpu().numpy(), att_preturb, input=False
+                y_pred[close_indices], y_pred_preturb[close_indices], attribution, att_preturb, input=False
             )
+
         else:
             y_pred = self(self.prep_data(x)).detach()
             x_original = x["encoder_cont"].detach().clone()
-            y_pred_preturb = self(self.prep_data(x)).detach()
+
             with torch.no_grad():
                 noise = torch.randn_like(x["encoder_cont"])*0.01
                 x["encoder_cont"] += noise
-
+            y_pred_preturb = self(self.prep_data(x)).detach()
             if explain_method == "Random":
                 att_preturb = np.random.normal(size=[64, 24, 53])
 
@@ -1087,16 +1109,24 @@ def norm_function(arr): return np.linalg.norm(arr)
                     att_preturb = explantation.attribute(data, **kwargs)
 
                 # Process and store the calculated attributions
-                att_preturb = att_preturb[1].cpu().detach().numpy() if method_name in [
-                    'Lime', 'FeatureAblation'] else torch.stack(att_preturb).cpu().detach().numpy()
+                att_preturb = att_preturb[1].detach() if method_name in [
+                    'Lime', 'FeatureAblation'] else torch.stack(att_preturb).detach()
+            # Calculate the absolute difference
+            difference = torch.abs(y_pred_preturb - y_pred)
+
+            # Find where the difference is less than or equal to a thershold
+            close_indices = torch.nonzero(difference <= thershold).squeeze()
+            print(close_indices)
+
             RIS = relative_stability_objective(
-                x_original.detach().cpu().numpy(), x["encoder_cont"].detach().cpu().numpy(), attribution, att_preturb, input=True
+                x_original[close_indices, :, :].detach(), x["encoder_cont"][close_indices, :, :].detach(), attribution[close_indices, :, :], att_preturb[close_indices, :, :], input=True
             )
             ROS = relative_stability_objective(
-                y_pred.cpu().numpy(), y_pred_preturb.cpu().numpy(), attribution, att_preturb, input=False
+                y_pred[close_indices], y_pred_preturb[close_indices], attribution[close_indices,
+                                                                                  :, :], att_preturb[close_indices, :, :], input=False
             )
 
-        return np.max(RIS), np.max(ROS)
+        return np.max(RIS.cpu().numpy()).astype(np.float64), np.max(ROS.cpu().numpy()).astype(np.float64)
 
 
 @gin.configurable("MLWrapper")
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 577607ff..c46f7b0f 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -258,7 +258,7 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
     log_dir_plots = log_dir / 'plots'
     if not (log_dir_plots.exists()):
         log_dir_plots.mkdir(parents=True)
-    plot_XAI_Metrics(accumulated_metrics, log_dir_plots=log_dir_plots)
+    # plot_XAI_Metrics(accumulated_metrics, log_dir_plots=log_dir_plots)
 
     with open(log_dir / "aggregated_test_metrics.json", "w") as f:
         json.dump(aggregated, f, cls=JsonResultLoggingEncoder)

From a39b5c418b62fe163bbd621502f0cbabb6fdc191 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 13 Dec 2023 15:52:02 +0100
Subject: [PATCH 121/142] added condition for ROS and RIS for attention

---
 icu_benchmarks/models/train.py    |  4 +--
 icu_benchmarks/models/wrappers.py | 43 +++++++++++++++++--------------
 2 files changed, 26 insertions(+), 21 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 7097571b..b22a6438 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -285,8 +285,8 @@ def train_common(
             # "L": Lime,
             # "IG": IntegratedGradients,
             # "FA": FeatureAblation,
-            "R": "Random",
-            # "Att": "Attention"
+            # "R": "Random",
+            "Att": "Attention"
         }
         for key, item in methods.items():
             # If conditions needed here as different explantations require different inputs
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index b3c76a40..5c0cde89 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -605,8 +605,8 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 timestep_attrs = Interpertations["attention"]
                 features_attrs = Interpertations["static_variables"].tolist()
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
-                r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
-                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name)
+                """ r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
+                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name) """
                 st_i_score, st_o_score = self.Relative_Stability(x=None,
                                                                  attribution=timestep_attrs, explain_method=method, method_name=method_name, dataloader=dataloader, **kwargs
                                                                  )
@@ -638,11 +638,11 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                                                              )
                         st_i_score.append(res1)
                         st_o_score.append(res2)
-                    else:
+                    """ else:
                         f_ts_score.append(self.Faithfulness_Correlation(x, timestep_attrs,
                                                                         pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
                         f_v_score.append(self.Faithfulness_Correlation(x, features_attrs,
-                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100)) """
 
             # Faithfulness score for attribtuons of features per timesteps
             f_ts_v_score = np.mean(f_ts_v_score)
@@ -681,7 +681,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
             if XAI_metric:
 
-                f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                """ f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
                                                                   pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
 
                 f_ts_score.append(self.Faithfulness_Correlation(x, stacked_attr,
@@ -689,7 +689,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 f_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
                                                                pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
                 r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
-                               explain_method=method, random_model=random_model, method_name=method_name))
+                               explain_method=method, random_model=random_model, method_name=method_name)) """
                 res1, res2 = self.Relative_Stability(x,
                                                      stacked_attr, explain_method=method, method_name=method_name,  dataloader=None, **kwargs
                                                      )
@@ -985,7 +985,7 @@ def Data_Randomization(
         return score
 
     def Relative_Stability(self, x,
-                           attribution, explain_method, method_name, dataloader=None, thershold=0.5, **kwargs
+                           attribution, explain_method, method_name, dataloader=None, thershold=0.5, device='cuda', **kwargs
                            ):
         """
      Args:
@@ -1008,7 +1008,7 @@ def Relative_Stability(self, x,
         """
 
         def relative_stability_objective(
-            x, xs, e_x, e_xs, eps_min=0.0001, input=False, device='cuda'
+            x, xs, e_x, e_xs, close_indices, eps_min=0.0001, input=False, attention=False, device='cuda'
         ) -> torch.Tensor:
             """
             Computes relative input and output stabilities maximization objective
@@ -1032,11 +1032,15 @@ def relative_stability_objective(
             # Function to convert inputs to tensors if they are numpy arrays
             def to_tensor(input_array):
                 if isinstance(input_array, np.ndarray):
-                    return torch.tensor(input_array).to(device)
-                return input_array.to(device)
+                    return torch.index_select(torch.tensor(input_array).to(device), 0, close_indices)
+
+                return torch.index_select(input_array.to(device), 0, close_indices)
 
             # Convert all inputs to tensors and move to GPU
-            x, xs, e_x, e_xs = map(to_tensor, [x, xs, e_x, e_xs])
+            if attention:
+                x, xs = map(to_tensor, [x, xs])
+            else:
+                x, xs, e_x, e_xs = map(to_tensor, [x, xs, e_x, e_xs])
 
             if input:
                 num_dim = x.ndim
@@ -1078,12 +1082,13 @@ def norm_function(arr): return torch.norm(arr)
             difference = torch.abs(y_pred_preturb - y_pred)
 
             # Find where the difference is less than or equal to a thershold
-            close_indices = torch.nonzero(difference <= thershold).squeeze()
+            close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
+
             RIS = relative_stability_objective(
-                x_original[close_indices, :, :].detach(), x_preturb[close_indices, :, :].detach(), attribution, att_preturb, input=True
+                x_original.detach(), x_preturb.detach(), attribution, att_preturb, close_indices=close_indices, input=True, attention=True
             )
             ROS = relative_stability_objective(
-                y_pred[close_indices], y_pred_preturb[close_indices], attribution, att_preturb, input=False
+                y_pred, y_pred_preturb, attribution, att_preturb, close_indices=close_indices, input=False, attention=True
             )
 
         else:
@@ -1115,15 +1120,15 @@ def norm_function(arr): return torch.norm(arr)
             difference = torch.abs(y_pred_preturb - y_pred)
 
             # Find where the difference is less than or equal to a thershold
-            close_indices = torch.nonzero(difference <= thershold).squeeze()
-            print(close_indices)
+            close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
 
             RIS = relative_stability_objective(
-                x_original[close_indices, :, :].detach(), x["encoder_cont"][close_indices, :, :].detach(), attribution[close_indices, :, :], att_preturb[close_indices, :, :], input=True
+                x_original.detach(),
+                x["encoder_cont"].detach(),
+                attribution, att_preturb, close_indices=close_indices, input=True
             )
             ROS = relative_stability_objective(
-                y_pred[close_indices], y_pred_preturb[close_indices], attribution[close_indices,
-                                                                                  :, :], att_preturb[close_indices, :, :], input=False
+                y_pred, y_pred_preturb, attribution, att_preturb, close_indices=close_indices, input=False
             )
 
         return np.max(RIS.cpu().numpy()).astype(np.float64), np.max(ROS.cpu().numpy()).astype(np.float64)

From daffaa2b41bae8078eb7de8eb1544c0508e9a32c Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 25 Dec 2023 00:25:38 +0100
Subject: [PATCH 122/142] added flags to add explantation methods and metrics

---
 configs/prediction_models/DeepARpytorch.gin |  2 +-
 configs/prediction_models/RNNpytorch.gin    | 12 +++++------
 configs/prediction_models/TFTpytorch.gin    |  2 +-
 icu_benchmarks/models/train.py              | 13 +++++------
 icu_benchmarks/models/wrappers.py           | 24 +++++++++++++--------
 icu_benchmarks/run.py                       |  7 ++++--
 icu_benchmarks/run_utils.py                 | 22 +++++++++++++++----
 7 files changed, 53 insertions(+), 29 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index b093027e..bcc6abf6 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -14,7 +14,7 @@ optimizer/hyperparameter.lr = 0.00046
 
 model/hyperparameter.class_to_tune = @DeepARpytorch
 model/hyperparameter.hidden = 116
-#(4, 64, "log-uniform", 2)
+#(4, 120, "log-uniform", 2)
 model/hyperparameter.rnn_layers=1
 
 model/hyperparameter.num_classes = %NUM_CLASSES
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 050f11a5..6c0f2e15 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -8,24 +8,24 @@ train_common.model = @RNNpytorch
 
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = 0.00016
+optimizer/hyperparameter.lr = 0.00041
 
 # Model params
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden =  42
+model/hyperparameter.hidden =  214
 #(2, 64, "log-uniform", 2)
-model/hyperparameter.rnn_layers=2 
+model/hyperparameter.rnn_layers=1
 #(1,3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
-model/hyperparameter.dropout = (0.3976, 0.3978)
+model/hyperparameter.dropout = (0.244, 0.2441)
 #(0.0, 0.4)
 #model/hyperparameter.lr_scheduler = "exponential"
 
 train_common/hyperparameter.class_to_tune = @train_common
-train_common/hyperparameter.batch_size=128
+train_common/hyperparameter.batch_size=256
 #(32,64,128,256,512)
-train_common/hyperparameter.gradient_clip_val=0.01
+train_common/hyperparameter.gradient_clip_val=100.0
 #(0,0.01, 1.0, 100.0)
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index 19130460..c7264ed5 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -27,7 +27,7 @@ train_common/hyperparameter.class_to_tune = @train_common
 train_common/hyperparameter.batch_size=64
 #(32,64,128,256,512)
 
-train_common/hyperparameter.gradient_clip_val=0.0
+train_common/hyperparameter.gradient_clip_val=0.01
 #(0,0.01, 1.0, 100.0)
 
 
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index b22a6438..73116f46 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -25,6 +25,7 @@
     PredictionDatasetpytorch,
 )
 
+
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
@@ -72,9 +73,9 @@ def train_common(
         torch.cuda.device_count() * 8 * int(torch.cuda.is_available()),
         32,
     ),
-    explain: bool = False,
+    explain: list = [],
     pytorch_forecasting: bool = False,
-    XAI_metric: bool = False,
+    XAI_metric: list = [],
     random_labels: bool = False,
     random_model_dir: str = None,
 
@@ -245,7 +246,7 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
-        gradient_clip_val=gradient_clip_val
+        # gradient_clip_val=gradient_clip_val
     )
     if not eval_only:
         if model.requires_backprop:
@@ -281,11 +282,11 @@ def train_common(
 
         # choose which  methods to get attributions
         methods = {
-            # "G": Saliency,
+            "G": Saliency,
             # "L": Lime,
-            # "IG": IntegratedGradients,
+            "IG": IntegratedGradients,
             # "FA": FeatureAblation,
-            # "R": "Random",
+            "R": "Random",
             "Att": "Attention"
         }
         for key, item in methods.items():
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 5c0cde89..b97d1306 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -242,6 +242,8 @@ def __init__(
         input_size: Tensor = None,
         initialization_method: str = "normal",
         pytorch_forecasting: bool = False,
+        explain: list = [],
+        XAI_metric: list = [],
         **kwargs,
     ):
         super().__init__(
@@ -262,6 +264,8 @@ def __init__(
         self.output_transform = None
         self.loss_weights = None
         self.pytorch_forecasting = pytorch_forecasting
+        self.explain = explain
+        self.XAI_metric = XAI_metric
 
     def set_weight(self, weight, dataset):
         """Set the weight for the loss function."""
@@ -605,8 +609,8 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 timestep_attrs = Interpertations["attention"]
                 features_attrs = Interpertations["static_variables"].tolist()
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
-                """ r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
-                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name) """
+                r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
+                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name)
                 st_i_score, st_o_score = self.Relative_Stability(x=None,
                                                                  attribution=timestep_attrs, explain_method=method, method_name=method_name, dataloader=dataloader, **kwargs
                                                                  )
@@ -624,25 +628,25 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
 
                     if method_name == "Random":
 
-                        """ f_ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
+                        f_ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
                                                                           pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
                         f_ts_score.append(self.Faithfulness_Correlation(x, all_attrs,
                                                                         pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
                         f_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
-                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100)) 
+                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
 
                         r_score.append(self.Data_Randomization(x, attribution=all_attrs,
-                                       explain_method=method, random_model=random_model, method_name=method_name))"""
+                                       explain_method=method, random_model=random_model, method_name=method_name))
                         res1, res2 = self.Relative_Stability(x,
                                                              all_attrs, explain_method=method, method_name=method_name, dataloader=None, **kwargs
                                                              )
                         st_i_score.append(res1)
                         st_o_score.append(res2)
-                    """ else:
+                    else:
                         f_ts_score.append(self.Faithfulness_Correlation(x, timestep_attrs,
                                                                         pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
                         f_v_score.append(self.Faithfulness_Correlation(x, features_attrs,
-                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100)) """
+                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
 
             # Faithfulness score for attribtuons of features per timesteps
             f_ts_v_score = np.mean(f_ts_v_score)
@@ -681,7 +685,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
             if XAI_metric:
 
-                """ f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
+                f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
                                                                   pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
 
                 f_ts_score.append(self.Faithfulness_Correlation(x, stacked_attr,
@@ -689,12 +693,14 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 f_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
                                                                pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
                 r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
-                               explain_method=method, random_model=random_model, method_name=method_name)) """
+                               explain_method=method, random_model=random_model, method_name=method_name))
+
                 res1, res2 = self.Relative_Stability(x,
                                                      stacked_attr, explain_method=method, method_name=method_name,  dataloader=None, **kwargs
                                                      )
                 st_i_score.append(res1)
                 st_o_score.append(res2)
+
             # aggregate over batch
             attr = np.mean(stacked_attr, axis=0)
             all_attrs.append(attr)
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index 1faadab6..bd65c935 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -56,6 +56,7 @@ def main(my_args=tuple(sys.argv[1:])):
     pytorch_forecasting = args.pytorch_forecasting
     XAI_metric = args.XAI_metric
     random_labels = args.random_labels
+
     # Load task config
     gin.parse_config_file(f"configs/tasks/{task}.gin")
     mode = get_mode()
@@ -117,7 +118,8 @@ def main(my_args=tuple(sys.argv[1:])):
             name_datasets(args.name, args.name, args.name)
         run_dir = create_run_dir(log_dir)
         source_dir = args.source_dir
-        logging.info(f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
+        logging.info(
+            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
         gin.parse_config_file(source_dir / "train_config.gin")
     elif args.samples and args.source_dir is not None:  # Train model with limited samples and bind existing config
         logging.info("Binding train gin-config. Note: this might override your config.")
@@ -130,7 +132,8 @@ def main(my_args=tuple(sys.argv[1:])):
         name_datasets(args.name, args.name, args.name)
         hp_checkpoint = log_dir / args.hp_checkpoint if args.hp_checkpoint else None
         model_path = (
-            Path("configs") / ("imputation_models" if mode == RunMode.imputation else "prediction_models") / f"{model}.gin"
+            Path("configs") / ("imputation_models" if mode ==
+                               RunMode.imputation else "prediction_models") / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index c46f7b0f..020238e4 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -123,29 +123,43 @@ def build_parser() -> ArgumentParser:
         default=None,
         help="Number of samples to use for evaluation.",
     )
-    parser.add_argument(
+    """ parser.add_argument(
         "--explain",
         default=False,
         action=BOA,
         help="Provide explaintations for predictions.",
-    )
+    ) """
     parser.add_argument(
         "--pytorch-forecasting",
         default=False,
         action=BOA,
         help="use pytorch forecasting library ",
     )
-    parser.add_argument(
+    """ parser.add_argument(
         "--XAI_metric",
         default=False,
         action=BOA,
         help="Compare explantations ",
+    ) """
+    parser.add_argument(
+        "-ex",
+        "--explain",
+        nargs='+',  # '+' means one or more arguments
+        type=str,   # or any type you need
+        help="List of XAI methods to run"
+    )
+    parser.add_argument(
+        "-exm",
+        "--XAI_metric",
+        nargs='+',  # '+' means one or more arguments
+        type=str,   # or any type you need
+        help="List of XAI metrics to run"
     )
     parser.add_argument(
         "--random_labels",
         default=False,
         action=BOA,
-        help="train model on random label dataset",
+        help="randmize target labels",
     )
 
     parser.add_argument(

From a9ec05eaf237e0f27ae3c91b7aa5e462d28193e5 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 27 Dec 2023 19:36:59 +0100
Subject: [PATCH 123/142] changes to allow easier run

---
 configs/prediction_models/DeepARpytorch.gin |  68 +++-
 configs/prediction_models/RNNpytorch.gin    |  81 +++-
 configs/prediction_models/TFTpytorch.gin    |   2 +
 icu_benchmarks/cross_validation.py          |   4 +-
 icu_benchmarks/data/loader.py               |  12 +-
 icu_benchmarks/models/constants.py          |  14 +
 icu_benchmarks/models/custom_metrics.py     | 421 ++++++++++++++++++++
 icu_benchmarks/models/dl_models.py          |  22 +-
 icu_benchmarks/models/train.py              |  46 ++-
 icu_benchmarks/models/wrappers.py           | 248 +++++++++++-
 10 files changed, 857 insertions(+), 61 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index bcc6abf6..04d8011a 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -29,16 +29,63 @@ train_common/hyperparameter.gradient_clip_val=0.01
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
-PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_1', 'alp_lag_1', 'alt_lag_1', 'ast_lag_1', 'be_lag_1', 'bicar_lag_1', 'bili_lag_1', 'bili_dir_lag_1', 'bnd_lag_1', 'bun_lag_1', 'ca_lag_1', 'cai_lag_1', 'ck_lag_1', 'ckmb_lag_1', 'cl_lag_1', 'crea_lag_1', 'crp_lag_1', 'dbp_lag_1', 'fgn_lag_1', 'fio2_lag_1', 'glu_lag_1', 'hgb_lag_1', 'hr_lag_1', 'inr_pt_lag_1', 'k_lag_1', 'lact_lag_1', 'lymph_lag_1', 'map_lag_1', 'mch_lag_1', 'mchc_lag_1', 'mcv_lag_1', 'methb_lag_1', 'mg_lag_1', 'na_lag_1', 'neut_lag_1', 'o2sat_lag_1', 'pco2_lag_1', 'ph_lag_1', 'phos_lag_1', 'plt_lag_1', 'po2_lag_1', 'ptt_lag_1', 'resp_lag_1', 'sbp_lag_1', 'temp_lag_1', 'tnt_lag_1', 'urine_lag_1', 'wbc_lag_1', 'alb_lag_2', 'alp_lag_2', 'alt_lag_2', 'ast_lag_2', 'be_lag_2', 'bicar_lag_2', 'bili_lag_2', 'bili_dir_lag_2', 'bnd_lag_2', 'bun_lag_2', 'ca_lag_2', 'cai_lag_2', 'ck_lag_2', 'ckmb_lag_2', 'cl_lag_2', 'crea_lag_2', 'crp_lag_2', 'dbp_lag_2', 'fgn_lag_2', 'fio2_lag_2', 'glu_lag_2', 'hgb_lag_2', 'hr_lag_2', 'inr_pt_lag_2', 'k_lag_2', 'lact_lag_2', 'lymph_lag_2', 'map_lag_2', 'mch_lag_2', 'mchc_lag_2', 'mcv_lag_2', 'methb_lag_2', 'mg_lag_2', 'na_lag_2', 'neut_lag_2', 'o2sat_lag_2', 'pco2_lag_2', 'ph_lag_2', 'phos_lag_2', 'plt_lag_2', 'po2_lag_2', 'ptt_lag_2', 'resp_lag_2', 'sbp_lag_2', 'temp_lag_2', 'tnt_lag_2', 'urine_lag_2', 'wbc_lag_2', 'alb_lag_3', 'alp_lag_3', 'alt_lag_3', 'ast_lag_3', 'be_lag_3', 'bicar_lag_3', 'bili_lag_3', 'bili_dir_lag_3', 'bnd_lag_3', 'bun_lag_3', 'ca_lag_3', 'cai_lag_3', 'ck_lag_3', 'ckmb_lag_3', 'cl_lag_3', 'crea_lag_3', 'crp_lag_3', 'dbp_lag_3', 'fgn_lag_3', 'fio2_lag_3', 'glu_lag_3', 'hgb_lag_3', 'hr_lag_3', 'inr_pt_lag_3', 'k_lag_3', 'lact_lag_3', 'lymph_lag_3', 'map_lag_3', 'mch_lag_3', 'mchc_lag_3', 'mcv_lag_3', 'methb_lag_3', 'mg_lag_3', 'na_lag_3', 'neut_lag_3', 'o2sat_lag_3', 'pco2_lag_3', 'ph_lag_3', 'phos_lag_3', 'plt_lag_3', 'po2_lag_3', 'ptt_lag_3', 'resp_lag_3', 'sbp_lag_3', 'temp_lag_3', 'tnt_lag_3', 'urine_lag_3', 'wbc_lag_3', 'alb_lag_4', 'alp_lag_4', 'alt_lag_4', 'ast_lag_4', 'be_lag_4', 'bicar_lag_4', 'bili_lag_4', 'bili_dir_lag_4', 'bnd_lag_4', 'bun_lag_4', 'ca_lag_4', 'cai_lag_4', 'ck_lag_4', 'ckmb_lag_4', 'cl_lag_4', 'crea_lag_4', 'crp_lag_4', 'dbp_lag_4', 'fgn_lag_4', 'fio2_lag_4', 'glu_lag_4', 'hgb_lag_4', 'hr_lag_4', 'inr_pt_lag_4', 'k_lag_4', 'lact_lag_4', 'lymph_lag_4', 'map_lag_4', 'mch_lag_4', 'mchc_lag_4', 'mcv_lag_4', 'methb_lag_4', 'mg_lag_4', 'na_lag_4', 'neut_lag_4', 'o2sat_lag_4', 'pco2_lag_4', 'ph_lag_4', 'phos_lag_4', 'plt_lag_4', 'po2_lag_4', 'ptt_lag_4', 'resp_lag_4', 'sbp_lag_4', 'temp_lag_4', 'tnt_lag_4', 'urine_lag_4', 'wbc_lag_4', 'alb_lag_5', 'alp_lag_5', 'alt_lag_5', 'ast_lag_5', 'be_lag_5', 'bicar_lag_5', 'bili_lag_5', 'bili_dir_lag_5', 'bnd_lag_5', 'bun_lag_5', 'ca_lag_5', 'cai_lag_5', 'ck_lag_5', 'ckmb_lag_5', 'cl_lag_5', 'crea_lag_5', 'crp_lag_5', 'dbp_lag_5', 'fgn_lag_5', 'fio2_lag_5', 'glu_lag_5', 'hgb_lag_5', 'hr_lag_5', 'inr_pt_lag_5', 'k_lag_5', 'lact_lag_5', 'lymph_lag_5', 'map_lag_5', 'mch_lag_5', 'mchc_lag_5', 'mcv_lag_5', 'methb_lag_5', 'mg_lag_5', 'na_lag_5', 'neut_lag_5', 'o2sat_lag_5', 'pco2_lag_5', 'ph_lag_5', 'phos_lag_5', 'plt_lag_5', 'po2_lag_5', 'ptt_lag_5', 'resp_lag_5', 'sbp_lag_5', 'temp_lag_5', 'tnt_lag_5', 'urine_lag_5', 'wbc_lag_5', 'alb_lag_6', 'alp_lag_6', 'alt_lag_6', 'ast_lag_6', 'be_lag_6', 'bicar_lag_6', 'bili_lag_6', 'bili_dir_lag_6', 'bnd_lag_6', 'bun_lag_6', 'ca_lag_6', 'cai_lag_6', 'ck_lag_6', 'ckmb_lag_6', 'cl_lag_6', 'crea_lag_6', 'crp_lag_6', 'dbp_lag_6', 'fgn_lag_6', 'fio2_lag_6', 'glu_lag_6', 'hgb_lag_6', 'hr_lag_6', 'inr_pt_lag_6', 'k_lag_6', 'lact_lag_6', 'lymph_lag_6', 'map_lag_6', 'mch_lag_6', 'mchc_lag_6', 'mcv_lag_6', 'methb_lag_6', 'mg_lag_6', 'na_lag_6', 'neut_lag_6', 'o2sat_lag_6', 'pco2_lag_6', 'ph_lag_6', 'phos_lag_6', 'plt_lag_6', 'po2_lag_6', 'ptt_lag_6', 'resp_lag_6', 'sbp_lag_6', 'temp_lag_6', 'tnt_lag_6', 'urine_lag_6', 'wbc_lag_6', 'alb_lag_7', 'alp_lag_7', 'alt_lag_7', 'ast_lag_7', 'be_lag_7', 'bicar_lag_7', 'bili_lag_7', 'bili_dir_lag_7', 'bnd_lag_7', 'bun_lag_7', 'ca_lag_7', 'cai_lag_7', 'ck_lag_7', 'ckmb_lag_7', 'cl_lag_7', 'crea_lag_7', 'crp_lag_7', 'dbp_lag_7', 'fgn_lag_7', 'fio2_lag_7', 'glu_lag_7', 'hgb_lag_7', 'hr_lag_7', 'inr_pt_lag_7', 'k_lag_7', 'lact_lag_7', 'lymph_lag_7', 'map_lag_7', 'mch_lag_7', 'mchc_lag_7', 'mcv_lag_7', 'methb_lag_7', 'mg_lag_7', 'na_lag_7', 'neut_lag_7', 'o2sat_lag_7', 'pco2_lag_7', 'ph_lag_7', 'phos_lag_7', 'plt_lag_7', 'po2_lag_7', 'ptt_lag_7', 'resp_lag_7', 'sbp_lag_7', 'temp_lag_7', 'tnt_lag_7', 'urine_lag_7', 'wbc_lag_7', 'alb_lag_8', 'alp_lag_8', 'alt_lag_8', 'ast_lag_8', 'be_lag_8', 'bicar_lag_8', 'bili_lag_8', 'bili_dir_lag_8', 'bnd_lag_8', 'bun_lag_8', 'ca_lag_8', 'cai_lag_8', 'ck_lag_8', 'ckmb_lag_8', 'cl_lag_8', 'crea_lag_8', 'crp_lag_8', 'dbp_lag_8', 'fgn_lag_8', 'fio2_lag_8', 'glu_lag_8', 'hgb_lag_8', 'hr_lag_8', 'inr_pt_lag_8', 'k_lag_8', 'lact_lag_8', 'lymph_lag_8', 'map_lag_8', 'mch_lag_8', 'mchc_lag_8', 'mcv_lag_8', 'methb_lag_8', 'mg_lag_8', 'na_lag_8', 'neut_lag_8', 'o2sat_lag_8', 'pco2_lag_8', 'ph_lag_8', 'phos_lag_8', 'plt_lag_8', 'po2_lag_8', 'ptt_lag_8', 'resp_lag_8', 'sbp_lag_8', 'temp_lag_8', 'tnt_lag_8', 'urine_lag_8', 'wbc_lag_8', 'alb_lag_9', 'alp_lag_9', 'alt_lag_9', 'ast_lag_9', 'be_lag_9', 'bicar_lag_9', 'bili_lag_9', 'bili_dir_lag_9', 'bnd_lag_9', 'bun_lag_9', 'ca_lag_9', 'cai_lag_9', 'ck_lag_9', 'ckmb_lag_9', 'cl_lag_9', 'crea_lag_9', 'crp_lag_9', 'dbp_lag_9', 'fgn_lag_9', 'fio2_lag_9', 'glu_lag_9', 'hgb_lag_9', 'hr_lag_9', 'inr_pt_lag_9', 'k_lag_9', 'lact_lag_9', 'lymph_lag_9', 'map_lag_9', 'mch_lag_9', 'mchc_lag_9', 'mcv_lag_9', 'methb_lag_9', 'mg_lag_9', 'na_lag_9', 'neut_lag_9', 'o2sat_lag_9', 'pco2_lag_9', 'ph_lag_9', 'phos_lag_9', 'plt_lag_9', 'po2_lag_9', 'ptt_lag_9', 'resp_lag_9', 'sbp_lag_9', 'temp_lag_9', 'tnt_lag_9', 'urine_lag_9', 'wbc_lag_9', 'alb_lag_10', 'alp_lag_10', 'alt_lag_10', 'ast_lag_10', 'be_lag_10', 'bicar_lag_10', 'bili_lag_10', 'bili_dir_lag_10', 'bnd_lag_10', 'bun_lag_10', 'ca_lag_10', 'cai_lag_10', 'ck_lag_10', 'ckmb_lag_10', 'cl_lag_10', 'crea_lag_10', 'crp_lag_10', 'dbp_lag_10', 'fgn_lag_10', 'fio2_lag_10', 'glu_lag_10', 'hgb_lag_10', 'hr_lag_10', 'inr_pt_lag_10', 'k_lag_10', 'lact_lag_10', 'lymph_lag_10', 'map_lag_10', 'mch_lag_10', 'mchc_lag_10', 'mcv_lag_10', 'methb_lag_10', 'mg_lag_10', 'na_lag_10', 'neut_lag_10', 'o2sat_lag_10', 'pco2_lag_10', 'ph_lag_10', 'phos_lag_10', 'plt_lag_10', 'po2_lag_10', 'ptt_lag_10', 'resp_lag_10', 'sbp_lag_10', 'temp_lag_10', 'tnt_lag_10', 'urine_lag_10', 'wbc_lag_10', 'alb_lag_11', 'alp_lag_11', 'alt_lag_11', 'ast_lag_11', 'be_lag_11', 'bicar_lag_11', 'bili_lag_11', 'bili_dir_lag_11', 'bnd_lag_11', 'bun_lag_11', 'ca_lag_11', 'cai_lag_11', 'ck_lag_11', 'ckmb_lag_11', 'cl_lag_11', 'crea_lag_11', 'crp_lag_11', 'dbp_lag_11', 'fgn_lag_11', 'fio2_lag_11', 'glu_lag_11', 'hgb_lag_11', 'hr_lag_11', 'inr_pt_lag_11', 'k_lag_11', 'lact_lag_11', 'lymph_lag_11', 'map_lag_11', 'mch_lag_11', 'mchc_lag_11', 'mcv_lag_11', 'methb_lag_11', 'mg_lag_11', 'na_lag_11', 'neut_lag_11', 'o2sat_lag_11', 'pco2_lag_11', 'ph_lag_11', 'phos_lag_11', 'plt_lag_11', 'po2_lag_11', 'ptt_lag_11', 'resp_lag_11', 'sbp_lag_11', 'temp_lag_11', 'tnt_lag_11', 'urine_lag_11', 'wbc_lag_11', 'alb_lag_12', 'alp_lag_12', 'alt_lag_12', 'ast_lag_12', 'be_lag_12', 'bicar_lag_12', 'bili_lag_12', 'bili_dir_lag_12', 'bnd_lag_12', 'bun_lag_12', 'ca_lag_12', 'cai_lag_12', 'ck_lag_12', 'ckmb_lag_12', 'cl_lag_12', 'crea_lag_12', 'crp_lag_12', 'dbp_lag_12', 'fgn_lag_12', 'fio2_lag_12', 'glu_lag_12', 'hgb_lag_12', 'hr_lag_12', 'inr_pt_lag_12', 'k_lag_12', 'lact_lag_12', 'lymph_lag_12', 'map_lag_12', 'mch_lag_12', 'mchc_lag_12', 'mcv_lag_12', 'methb_lag_12', 'mg_lag_12', 'na_lag_12', 'neut_lag_12', 'o2sat_lag_12', 'pco2_lag_12', 'ph_lag_12', 'phos_lag_12', 'plt_lag_12', 'po2_lag_12', 'ptt_lag_12', 'resp_lag_12', 'sbp_lag_12', 'temp_lag_12', 'tnt_lag_12', 'urine_lag_12', 'wbc_lag_12', 'alb_lag_13', 'alp_lag_13', 'alt_lag_13', 'ast_lag_13', 'be_lag_13', 'bicar_lag_13', 'bili_lag_13', 'bili_dir_lag_13', 'bnd_lag_13', 'bun_lag_13', 'ca_lag_13', 'cai_lag_13', 'ck_lag_13', 'ckmb_lag_13', 'cl_lag_13', 'crea_lag_13', 'crp_lag_13', 'dbp_lag_13', 'fgn_lag_13', 'fio2_lag_13', 'glu_lag_13', 'hgb_lag_13', 'hr_lag_13', 'inr_pt_lag_13', 'k_lag_13', 'lact_lag_13', 'lymph_lag_13', 'map_lag_13', 'mch_lag_13', 'mchc_lag_13', 'mcv_lag_13', 'methb_lag_13', 'mg_lag_13', 'na_lag_13', 'neut_lag_13', 'o2sat_lag_13', 'pco2_lag_13', 'ph_lag_13', 'phos_lag_13', 'plt_lag_13', 'po2_lag_13', 'ptt_lag_13', 'resp_lag_13', 'sbp_lag_13', 'temp_lag_13', 'tnt_lag_13', 'urine_lag_13', 'wbc_lag_13', 'alb_lag_14', 'alp_lag_14', 'alt_lag_14', 'ast_lag_14', 'be_lag_14', 'bicar_lag_14', 'bili_lag_14', 'bili_dir_lag_14', 'bnd_lag_14', 'bun_lag_14', 'ca_lag_14', 'cai_lag_14', 'ck_lag_14', 'ckmb_lag_14', 'cl_lag_14', 'crea_lag_14', 'crp_lag_14', 'dbp_lag_14', 'fgn_lag_14', 'fio2_lag_14', 'glu_lag_14', 'hgb_lag_14', 'hr_lag_14', 'inr_pt_lag_14', 'k_lag_14', 'lact_lag_14', 'lymph_lag_14', 'map_lag_14', 'mch_lag_14', 'mchc_lag_14', 'mcv_lag_14', 'methb_lag_14', 'mg_lag_14', 'na_lag_14', 'neut_lag_14', 'o2sat_lag_14', 'pco2_lag_14', 'ph_lag_14', 'phos_lag_14', 'plt_lag_14', 'po2_lag_14', 'ptt_lag_14', 'resp_lag_14', 'sbp_lag_14', 'temp_lag_14', 'tnt_lag_14', 'urine_lag_14', 'wbc_lag_14', 'alb_lag_15', 'alp_lag_15', 'alt_lag_15', 'ast_lag_15', 'be_lag_15', 'bicar_lag_15', 'bili_lag_15', 'bili_dir_lag_15', 'bnd_lag_15', 'bun_lag_15', 'ca_lag_15', 'cai_lag_15', 'ck_lag_15', 'ckmb_lag_15', 'cl_lag_15', 'crea_lag_15', 'crp_lag_15', 'dbp_lag_15', 'fgn_lag_15', 'fio2_lag_15', 'glu_lag_15', 'hgb_lag_15', 'hr_lag_15', 'inr_pt_lag_15', 'k_lag_15', 'lact_lag_15', 'lymph_lag_15', 'map_lag_15', 'mch_lag_15', 'mchc_lag_15', 'mcv_lag_15', 'methb_lag_15', 'mg_lag_15', 'na_lag_15', 'neut_lag_15', 'o2sat_lag_15', 'pco2_lag_15', 'ph_lag_15', 'phos_lag_15', 'plt_lag_15', 'po2_lag_15', 'ptt_lag_15', 'resp_lag_15', 'sbp_lag_15', 'temp_lag_15', 'tnt_lag_15', 'urine_lag_15', 'wbc_lag_15', 'alb_lag_16', 'alp_lag_16', 'alt_lag_16', 'ast_lag_16', 'be_lag_16', 'bicar_lag_16', 'bili_lag_16', 'bili_dir_lag_16', 'bnd_lag_16', 'bun_lag_16', 'ca_lag_16', 'cai_lag_16', 'ck_lag_16', 'ckmb_lag_16', 'cl_lag_16', 'crea_lag_16', 'crp_lag_16', 'dbp_lag_16', 'fgn_lag_16', 'fio2_lag_16', 'glu_lag_16', 'hgb_lag_16', 'hr_lag_16', 'inr_pt_lag_16', 'k_lag_16', 'lact_lag_16', 'lymph_lag_16', 'map_lag_16', 'mch_lag_16', 'mchc_lag_16', 'mcv_lag_16', 'methb_lag_16', 'mg_lag_16', 'na_lag_16', 'neut_lag_16', 'o2sat_lag_16', 'pco2_lag_16', 'ph_lag_16', 'phos_lag_16', 'plt_lag_16', 'po2_lag_16', 'ptt_lag_16', 'resp_lag_16', 'sbp_lag_16', 'temp_lag_16', 'tnt_lag_16', 'urine_lag_16', 'wbc_lag_16', 'alb_lag_17', 'alp_lag_17', 'alt_lag_17', 'ast_lag_17', 'be_lag_17', 'bicar_lag_17', 'bili_lag_17', 'bili_dir_lag_17', 'bnd_lag_17', 'bun_lag_17', 'ca_lag_17', 'cai_lag_17', 'ck_lag_17', 'ckmb_lag_17', 'cl_lag_17', 'crea_lag_17', 'crp_lag_17', 'dbp_lag_17', 'fgn_lag_17', 'fio2_lag_17', 'glu_lag_17', 'hgb_lag_17', 'hr_lag_17', 'inr_pt_lag_17', 'k_lag_17', 'lact_lag_17', 'lymph_lag_17', 'map_lag_17', 'mch_lag_17', 'mchc_lag_17', 'mcv_lag_17', 'methb_lag_17', 'mg_lag_17', 'na_lag_17', 'neut_lag_17', 'o2sat_lag_17', 'pco2_lag_17', 'ph_lag_17', 'phos_lag_17', 'plt_lag_17', 'po2_lag_17', 'ptt_lag_17', 'resp_lag_17', 'sbp_lag_17', 'temp_lag_17', 'tnt_lag_17', 'urine_lag_17', 'wbc_lag_17', 'alb_lag_18', 'alp_lag_18', 'alt_lag_18', 'ast_lag_18', 'be_lag_18', 'bicar_lag_18', 'bili_lag_18', 'bili_dir_lag_18', 'bnd_lag_18', 'bun_lag_18', 'ca_lag_18', 'cai_lag_18', 'ck_lag_18', 'ckmb_lag_18', 'cl_lag_18', 'crea_lag_18', 'crp_lag_18', 'dbp_lag_18', 'fgn_lag_18', 'fio2_lag_18', 'glu_lag_18', 'hgb_lag_18', 'hr_lag_18', 'inr_pt_lag_18', 'k_lag_18', 'lact_lag_18', 'lymph_lag_18', 'map_lag_18', 'mch_lag_18', 'mchc_lag_18', 'mcv_lag_18', 'methb_lag_18', 'mg_lag_18', 'na_lag_18', 'neut_lag_18', 'o2sat_lag_18', 'pco2_lag_18', 'ph_lag_18', 'phos_lag_18', 'plt_lag_18', 'po2_lag_18', 'ptt_lag_18', 'resp_lag_18', 'sbp_lag_18', 'temp_lag_18', 'tnt_lag_18', 'urine_lag_18', 'wbc_lag_18', 'alb_lag_19', 'alp_lag_19', 'alt_lag_19', 'ast_lag_19', 'be_lag_19', 'bicar_lag_19', 'bili_lag_19', 'bili_dir_lag_19', 'bnd_lag_19', 'bun_lag_19', 'ca_lag_19', 'cai_lag_19', 'ck_lag_19', 'ckmb_lag_19', 'cl_lag_19', 'crea_lag_19', 'crp_lag_19', 'dbp_lag_19', 'fgn_lag_19', 'fio2_lag_19', 'glu_lag_19', 'hgb_lag_19', 'hr_lag_19', 'inr_pt_lag_19', 'k_lag_19', 'lact_lag_19', 'lymph_lag_19', 'map_lag_19', 'mch_lag_19', 'mchc_lag_19', 'mcv_lag_19', 'methb_lag_19', 'mg_lag_19', 'na_lag_19', 'neut_lag_19', 'o2sat_lag_19', 'pco2_lag_19', 'ph_lag_19', 'phos_lag_19', 'plt_lag_19', 'po2_lag_19', 'ptt_lag_19', 'resp_lag_19', 'sbp_lag_19', 'temp_lag_19', 'tnt_lag_19', 'urine_lag_19', 'wbc_lag_19', 'alb_lag_20', 'alp_lag_20', 'alt_lag_20', 'ast_lag_20', 'be_lag_20', 'bicar_lag_20', 'bili_lag_20', 'bili_dir_lag_20', 'bnd_lag_20', 'bun_lag_20', 'ca_lag_20', 'cai_lag_20', 'ck_lag_20', 'ckmb_lag_20', 'cl_lag_20', 'crea_lag_20', 'crp_lag_20', 'dbp_lag_20', 'fgn_lag_20', 'fio2_lag_20', 'glu_lag_20', 'hgb_lag_20', 'hr_lag_20', 'inr_pt_lag_20', 'k_lag_20', 'lact_lag_20', 'lymph_lag_20', 'map_lag_20', 'mch_lag_20', 'mchc_lag_20', 'mcv_lag_20', 'methb_lag_20', 'mg_lag_20', 'na_lag_20', 'neut_lag_20', 'o2sat_lag_20', 'pco2_lag_20', 'ph_lag_20', 'phos_lag_20', 'plt_lag_20', 'po2_lag_20', 'ptt_lag_20', 'resp_lag_20', 'sbp_lag_20', 'temp_lag_20', 'tnt_lag_20', 'urine_lag_20', 'wbc_lag_20', 'alb_lag_21', 'alp_lag_21', 'alt_lag_21', 'ast_lag_21', 'be_lag_21', 'bicar_lag_21', 'bili_lag_21', 'bili_dir_lag_21', 'bnd_lag_21', 'bun_lag_21', 'ca_lag_21', 'cai_lag_21', 'ck_lag_21', 'ckmb_lag_21', 'cl_lag_21', 'crea_lag_21', 'crp_lag_21', 'dbp_lag_21', 'fgn_lag_21', 'fio2_lag_21', 'glu_lag_21', 'hgb_lag_21', 'hr_lag_21', 'inr_pt_lag_21', 'k_lag_21', 'lact_lag_21', 'lymph_lag_21', 'map_lag_21', 'mch_lag_21', 'mchc_lag_21', 'mcv_lag_21', 'methb_lag_21', 'mg_lag_21', 'na_lag_21', 'neut_lag_21', 'o2sat_lag_21', 'pco2_lag_21', 'ph_lag_21', 'phos_lag_21', 'plt_lag_21', 'po2_lag_21', 'ptt_lag_21', 'resp_lag_21', 'sbp_lag_21', 'temp_lag_21', 'tnt_lag_21', 'urine_lag_21', 'wbc_lag_21', 'alb_lag_22', 'alp_lag_22', 'alt_lag_22', 'ast_lag_22', 'be_lag_22', 'bicar_lag_22', 'bili_lag_22', 'bili_dir_lag_22', 'bnd_lag_22', 'bun_lag_22', 'ca_lag_22', 'cai_lag_22', 'ck_lag_22', 'ckmb_lag_22', 'cl_lag_22', 'crea_lag_22', 'crp_lag_22', 'dbp_lag_22', 'fgn_lag_22', 'fio2_lag_22', 'glu_lag_22', 'hgb_lag_22', 'hr_lag_22', 'inr_pt_lag_22', 'k_lag_22', 'lact_lag_22', 'lymph_lag_22', 'map_lag_22', 'mch_lag_22', 'mchc_lag_22', 'mcv_lag_22', 'methb_lag_22', 'mg_lag_22', 'na_lag_22', 'neut_lag_22', 'o2sat_lag_22', 'pco2_lag_22', 'ph_lag_22', 'phos_lag_22', 'plt_lag_22', 'po2_lag_22', 'ptt_lag_22', 'resp_lag_22', 'sbp_lag_22', 'temp_lag_22', 'tnt_lag_22', 'urine_lag_22', 'wbc_lag_22', 'alb_lag_23', 'alp_lag_23', 'alt_lag_23', 'ast_lag_23', 'be_lag_23', 'bicar_lag_23', 'bili_lag_23', 'bili_dir_lag_23', 'bnd_lag_23', 'bun_lag_23', 'ca_lag_23', 'cai_lag_23', 'ck_lag_23', 'ckmb_lag_23', 'cl_lag_23', 'crea_lag_23', 'crp_lag_23', 'dbp_lag_23', 'fgn_lag_23', 'fio2_lag_23', 'glu_lag_23', 'hgb_lag_23', 'hr_lag_23', 'inr_pt_lag_23', 'k_lag_23', 'lact_lag_23', 'lymph_lag_23', 'map_lag_23', 'mch_lag_23', 'mchc_lag_23', 'mcv_lag_23', 'methb_lag_23', 'mg_lag_23', 'na_lag_23', 'neut_lag_23', 'o2sat_lag_23', 'pco2_lag_23', 'ph_lag_23', 'phos_lag_23', 'plt_lag_23', 'po2_lag_23', 'ptt_lag_23', 'resp_lag_23', 'sbp_lag_23', 'temp_lag_23', 'tnt_lag_23', 'urine_lag_23', 'wbc_lag_23', 'alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24' ]
 PredictionDatasetpytorch.add_relative_time_idx=False
-PredictionDatasetpytorch.target="label"
-            
-PredictionDatasetpytorch.time_varying_unknown_reals=[
+PredictionDatasetpytorch.target=[
+    "alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
                 "label",
               
             ]
-PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetpytorch.lagged_variables=[  "alb",
+
+PredictionDatasetpytorch.time_varying_known_reals=[]
+PredictionDatasetpytorch.time_varying_unknown_reals=[
+                "alb",
                 "alp",
                 "alt",
                 "ast",
@@ -85,6 +132,11 @@ PredictionDatasetpytorch.lagged_variables=[  "alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc"]
+                "wbc",
+                "label",
+              
+            ]
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.lagged_variables=[]
 
-                
\ No newline at end of file
+PredictionDatasetpytorch.target_normalizer='multi'
\ No newline at end of file
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 6c0f2e15..a1804a1d 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -5,42 +5,89 @@ include "configs/prediction_models/common/DLCommon.gin"
 
 # Optimizer params
 train_common.model = @RNNpytorch
-
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = 0.00041
+optimizer/hyperparameter.lr = 4.4815375287835764e-05
 
 # Model params
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden =  214
+model/hyperparameter.hidden =  42
 #(2, 64, "log-uniform", 2)
-model/hyperparameter.rnn_layers=1
+model/hyperparameter.rnn_layers=2
 #(1,3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
-model/hyperparameter.dropout = (0.244, 0.2441)
+model/hyperparameter.dropout = (0.397, 0.3971)
 #(0.0, 0.4)
 #model/hyperparameter.lr_scheduler = "exponential"
 
 train_common/hyperparameter.class_to_tune = @train_common
-train_common/hyperparameter.batch_size=256
+train_common/hyperparameter.batch_size=128
 #(32,64,128,256,512)
-train_common/hyperparameter.gradient_clip_val=100.0
+train_common/hyperparameter.gradient_clip_val=0.01
 #(0,0.01, 1.0, 100.0)
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
-PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_1', 'alp_lag_1', 'alt_lag_1', 'ast_lag_1', 'be_lag_1', 'bicar_lag_1', 'bili_lag_1', 'bili_dir_lag_1', 'bnd_lag_1', 'bun_lag_1', 'ca_lag_1', 'cai_lag_1', 'ck_lag_1', 'ckmb_lag_1', 'cl_lag_1', 'crea_lag_1', 'crp_lag_1', 'dbp_lag_1', 'fgn_lag_1', 'fio2_lag_1', 'glu_lag_1', 'hgb_lag_1', 'hr_lag_1', 'inr_pt_lag_1', 'k_lag_1', 'lact_lag_1', 'lymph_lag_1', 'map_lag_1', 'mch_lag_1', 'mchc_lag_1', 'mcv_lag_1', 'methb_lag_1', 'mg_lag_1', 'na_lag_1', 'neut_lag_1', 'o2sat_lag_1', 'pco2_lag_1', 'ph_lag_1', 'phos_lag_1', 'plt_lag_1', 'po2_lag_1', 'ptt_lag_1', 'resp_lag_1', 'sbp_lag_1', 'temp_lag_1', 'tnt_lag_1', 'urine_lag_1', 'wbc_lag_1', 'alb_lag_2', 'alp_lag_2', 'alt_lag_2', 'ast_lag_2', 'be_lag_2', 'bicar_lag_2', 'bili_lag_2', 'bili_dir_lag_2', 'bnd_lag_2', 'bun_lag_2', 'ca_lag_2', 'cai_lag_2', 'ck_lag_2', 'ckmb_lag_2', 'cl_lag_2', 'crea_lag_2', 'crp_lag_2', 'dbp_lag_2', 'fgn_lag_2', 'fio2_lag_2', 'glu_lag_2', 'hgb_lag_2', 'hr_lag_2', 'inr_pt_lag_2', 'k_lag_2', 'lact_lag_2', 'lymph_lag_2', 'map_lag_2', 'mch_lag_2', 'mchc_lag_2', 'mcv_lag_2', 'methb_lag_2', 'mg_lag_2', 'na_lag_2', 'neut_lag_2', 'o2sat_lag_2', 'pco2_lag_2', 'ph_lag_2', 'phos_lag_2', 'plt_lag_2', 'po2_lag_2', 'ptt_lag_2', 'resp_lag_2', 'sbp_lag_2', 'temp_lag_2', 'tnt_lag_2', 'urine_lag_2', 'wbc_lag_2', 'alb_lag_3', 'alp_lag_3', 'alt_lag_3', 'ast_lag_3', 'be_lag_3', 'bicar_lag_3', 'bili_lag_3', 'bili_dir_lag_3', 'bnd_lag_3', 'bun_lag_3', 'ca_lag_3', 'cai_lag_3', 'ck_lag_3', 'ckmb_lag_3', 'cl_lag_3', 'crea_lag_3', 'crp_lag_3', 'dbp_lag_3', 'fgn_lag_3', 'fio2_lag_3', 'glu_lag_3', 'hgb_lag_3', 'hr_lag_3', 'inr_pt_lag_3', 'k_lag_3', 'lact_lag_3', 'lymph_lag_3', 'map_lag_3', 'mch_lag_3', 'mchc_lag_3', 'mcv_lag_3', 'methb_lag_3', 'mg_lag_3', 'na_lag_3', 'neut_lag_3', 'o2sat_lag_3', 'pco2_lag_3', 'ph_lag_3', 'phos_lag_3', 'plt_lag_3', 'po2_lag_3', 'ptt_lag_3', 'resp_lag_3', 'sbp_lag_3', 'temp_lag_3', 'tnt_lag_3', 'urine_lag_3', 'wbc_lag_3', 'alb_lag_4', 'alp_lag_4', 'alt_lag_4', 'ast_lag_4', 'be_lag_4', 'bicar_lag_4', 'bili_lag_4', 'bili_dir_lag_4', 'bnd_lag_4', 'bun_lag_4', 'ca_lag_4', 'cai_lag_4', 'ck_lag_4', 'ckmb_lag_4', 'cl_lag_4', 'crea_lag_4', 'crp_lag_4', 'dbp_lag_4', 'fgn_lag_4', 'fio2_lag_4', 'glu_lag_4', 'hgb_lag_4', 'hr_lag_4', 'inr_pt_lag_4', 'k_lag_4', 'lact_lag_4', 'lymph_lag_4', 'map_lag_4', 'mch_lag_4', 'mchc_lag_4', 'mcv_lag_4', 'methb_lag_4', 'mg_lag_4', 'na_lag_4', 'neut_lag_4', 'o2sat_lag_4', 'pco2_lag_4', 'ph_lag_4', 'phos_lag_4', 'plt_lag_4', 'po2_lag_4', 'ptt_lag_4', 'resp_lag_4', 'sbp_lag_4', 'temp_lag_4', 'tnt_lag_4', 'urine_lag_4', 'wbc_lag_4', 'alb_lag_5', 'alp_lag_5', 'alt_lag_5', 'ast_lag_5', 'be_lag_5', 'bicar_lag_5', 'bili_lag_5', 'bili_dir_lag_5', 'bnd_lag_5', 'bun_lag_5', 'ca_lag_5', 'cai_lag_5', 'ck_lag_5', 'ckmb_lag_5', 'cl_lag_5', 'crea_lag_5', 'crp_lag_5', 'dbp_lag_5', 'fgn_lag_5', 'fio2_lag_5', 'glu_lag_5', 'hgb_lag_5', 'hr_lag_5', 'inr_pt_lag_5', 'k_lag_5', 'lact_lag_5', 'lymph_lag_5', 'map_lag_5', 'mch_lag_5', 'mchc_lag_5', 'mcv_lag_5', 'methb_lag_5', 'mg_lag_5', 'na_lag_5', 'neut_lag_5', 'o2sat_lag_5', 'pco2_lag_5', 'ph_lag_5', 'phos_lag_5', 'plt_lag_5', 'po2_lag_5', 'ptt_lag_5', 'resp_lag_5', 'sbp_lag_5', 'temp_lag_5', 'tnt_lag_5', 'urine_lag_5', 'wbc_lag_5', 'alb_lag_6', 'alp_lag_6', 'alt_lag_6', 'ast_lag_6', 'be_lag_6', 'bicar_lag_6', 'bili_lag_6', 'bili_dir_lag_6', 'bnd_lag_6', 'bun_lag_6', 'ca_lag_6', 'cai_lag_6', 'ck_lag_6', 'ckmb_lag_6', 'cl_lag_6', 'crea_lag_6', 'crp_lag_6', 'dbp_lag_6', 'fgn_lag_6', 'fio2_lag_6', 'glu_lag_6', 'hgb_lag_6', 'hr_lag_6', 'inr_pt_lag_6', 'k_lag_6', 'lact_lag_6', 'lymph_lag_6', 'map_lag_6', 'mch_lag_6', 'mchc_lag_6', 'mcv_lag_6', 'methb_lag_6', 'mg_lag_6', 'na_lag_6', 'neut_lag_6', 'o2sat_lag_6', 'pco2_lag_6', 'ph_lag_6', 'phos_lag_6', 'plt_lag_6', 'po2_lag_6', 'ptt_lag_6', 'resp_lag_6', 'sbp_lag_6', 'temp_lag_6', 'tnt_lag_6', 'urine_lag_6', 'wbc_lag_6', 'alb_lag_7', 'alp_lag_7', 'alt_lag_7', 'ast_lag_7', 'be_lag_7', 'bicar_lag_7', 'bili_lag_7', 'bili_dir_lag_7', 'bnd_lag_7', 'bun_lag_7', 'ca_lag_7', 'cai_lag_7', 'ck_lag_7', 'ckmb_lag_7', 'cl_lag_7', 'crea_lag_7', 'crp_lag_7', 'dbp_lag_7', 'fgn_lag_7', 'fio2_lag_7', 'glu_lag_7', 'hgb_lag_7', 'hr_lag_7', 'inr_pt_lag_7', 'k_lag_7', 'lact_lag_7', 'lymph_lag_7', 'map_lag_7', 'mch_lag_7', 'mchc_lag_7', 'mcv_lag_7', 'methb_lag_7', 'mg_lag_7', 'na_lag_7', 'neut_lag_7', 'o2sat_lag_7', 'pco2_lag_7', 'ph_lag_7', 'phos_lag_7', 'plt_lag_7', 'po2_lag_7', 'ptt_lag_7', 'resp_lag_7', 'sbp_lag_7', 'temp_lag_7', 'tnt_lag_7', 'urine_lag_7', 'wbc_lag_7', 'alb_lag_8', 'alp_lag_8', 'alt_lag_8', 'ast_lag_8', 'be_lag_8', 'bicar_lag_8', 'bili_lag_8', 'bili_dir_lag_8', 'bnd_lag_8', 'bun_lag_8', 'ca_lag_8', 'cai_lag_8', 'ck_lag_8', 'ckmb_lag_8', 'cl_lag_8', 'crea_lag_8', 'crp_lag_8', 'dbp_lag_8', 'fgn_lag_8', 'fio2_lag_8', 'glu_lag_8', 'hgb_lag_8', 'hr_lag_8', 'inr_pt_lag_8', 'k_lag_8', 'lact_lag_8', 'lymph_lag_8', 'map_lag_8', 'mch_lag_8', 'mchc_lag_8', 'mcv_lag_8', 'methb_lag_8', 'mg_lag_8', 'na_lag_8', 'neut_lag_8', 'o2sat_lag_8', 'pco2_lag_8', 'ph_lag_8', 'phos_lag_8', 'plt_lag_8', 'po2_lag_8', 'ptt_lag_8', 'resp_lag_8', 'sbp_lag_8', 'temp_lag_8', 'tnt_lag_8', 'urine_lag_8', 'wbc_lag_8', 'alb_lag_9', 'alp_lag_9', 'alt_lag_9', 'ast_lag_9', 'be_lag_9', 'bicar_lag_9', 'bili_lag_9', 'bili_dir_lag_9', 'bnd_lag_9', 'bun_lag_9', 'ca_lag_9', 'cai_lag_9', 'ck_lag_9', 'ckmb_lag_9', 'cl_lag_9', 'crea_lag_9', 'crp_lag_9', 'dbp_lag_9', 'fgn_lag_9', 'fio2_lag_9', 'glu_lag_9', 'hgb_lag_9', 'hr_lag_9', 'inr_pt_lag_9', 'k_lag_9', 'lact_lag_9', 'lymph_lag_9', 'map_lag_9', 'mch_lag_9', 'mchc_lag_9', 'mcv_lag_9', 'methb_lag_9', 'mg_lag_9', 'na_lag_9', 'neut_lag_9', 'o2sat_lag_9', 'pco2_lag_9', 'ph_lag_9', 'phos_lag_9', 'plt_lag_9', 'po2_lag_9', 'ptt_lag_9', 'resp_lag_9', 'sbp_lag_9', 'temp_lag_9', 'tnt_lag_9', 'urine_lag_9', 'wbc_lag_9', 'alb_lag_10', 'alp_lag_10', 'alt_lag_10', 'ast_lag_10', 'be_lag_10', 'bicar_lag_10', 'bili_lag_10', 'bili_dir_lag_10', 'bnd_lag_10', 'bun_lag_10', 'ca_lag_10', 'cai_lag_10', 'ck_lag_10', 'ckmb_lag_10', 'cl_lag_10', 'crea_lag_10', 'crp_lag_10', 'dbp_lag_10', 'fgn_lag_10', 'fio2_lag_10', 'glu_lag_10', 'hgb_lag_10', 'hr_lag_10', 'inr_pt_lag_10', 'k_lag_10', 'lact_lag_10', 'lymph_lag_10', 'map_lag_10', 'mch_lag_10', 'mchc_lag_10', 'mcv_lag_10', 'methb_lag_10', 'mg_lag_10', 'na_lag_10', 'neut_lag_10', 'o2sat_lag_10', 'pco2_lag_10', 'ph_lag_10', 'phos_lag_10', 'plt_lag_10', 'po2_lag_10', 'ptt_lag_10', 'resp_lag_10', 'sbp_lag_10', 'temp_lag_10', 'tnt_lag_10', 'urine_lag_10', 'wbc_lag_10', 'alb_lag_11', 'alp_lag_11', 'alt_lag_11', 'ast_lag_11', 'be_lag_11', 'bicar_lag_11', 'bili_lag_11', 'bili_dir_lag_11', 'bnd_lag_11', 'bun_lag_11', 'ca_lag_11', 'cai_lag_11', 'ck_lag_11', 'ckmb_lag_11', 'cl_lag_11', 'crea_lag_11', 'crp_lag_11', 'dbp_lag_11', 'fgn_lag_11', 'fio2_lag_11', 'glu_lag_11', 'hgb_lag_11', 'hr_lag_11', 'inr_pt_lag_11', 'k_lag_11', 'lact_lag_11', 'lymph_lag_11', 'map_lag_11', 'mch_lag_11', 'mchc_lag_11', 'mcv_lag_11', 'methb_lag_11', 'mg_lag_11', 'na_lag_11', 'neut_lag_11', 'o2sat_lag_11', 'pco2_lag_11', 'ph_lag_11', 'phos_lag_11', 'plt_lag_11', 'po2_lag_11', 'ptt_lag_11', 'resp_lag_11', 'sbp_lag_11', 'temp_lag_11', 'tnt_lag_11', 'urine_lag_11', 'wbc_lag_11', 'alb_lag_12', 'alp_lag_12', 'alt_lag_12', 'ast_lag_12', 'be_lag_12', 'bicar_lag_12', 'bili_lag_12', 'bili_dir_lag_12', 'bnd_lag_12', 'bun_lag_12', 'ca_lag_12', 'cai_lag_12', 'ck_lag_12', 'ckmb_lag_12', 'cl_lag_12', 'crea_lag_12', 'crp_lag_12', 'dbp_lag_12', 'fgn_lag_12', 'fio2_lag_12', 'glu_lag_12', 'hgb_lag_12', 'hr_lag_12', 'inr_pt_lag_12', 'k_lag_12', 'lact_lag_12', 'lymph_lag_12', 'map_lag_12', 'mch_lag_12', 'mchc_lag_12', 'mcv_lag_12', 'methb_lag_12', 'mg_lag_12', 'na_lag_12', 'neut_lag_12', 'o2sat_lag_12', 'pco2_lag_12', 'ph_lag_12', 'phos_lag_12', 'plt_lag_12', 'po2_lag_12', 'ptt_lag_12', 'resp_lag_12', 'sbp_lag_12', 'temp_lag_12', 'tnt_lag_12', 'urine_lag_12', 'wbc_lag_12', 'alb_lag_13', 'alp_lag_13', 'alt_lag_13', 'ast_lag_13', 'be_lag_13', 'bicar_lag_13', 'bili_lag_13', 'bili_dir_lag_13', 'bnd_lag_13', 'bun_lag_13', 'ca_lag_13', 'cai_lag_13', 'ck_lag_13', 'ckmb_lag_13', 'cl_lag_13', 'crea_lag_13', 'crp_lag_13', 'dbp_lag_13', 'fgn_lag_13', 'fio2_lag_13', 'glu_lag_13', 'hgb_lag_13', 'hr_lag_13', 'inr_pt_lag_13', 'k_lag_13', 'lact_lag_13', 'lymph_lag_13', 'map_lag_13', 'mch_lag_13', 'mchc_lag_13', 'mcv_lag_13', 'methb_lag_13', 'mg_lag_13', 'na_lag_13', 'neut_lag_13', 'o2sat_lag_13', 'pco2_lag_13', 'ph_lag_13', 'phos_lag_13', 'plt_lag_13', 'po2_lag_13', 'ptt_lag_13', 'resp_lag_13', 'sbp_lag_13', 'temp_lag_13', 'tnt_lag_13', 'urine_lag_13', 'wbc_lag_13', 'alb_lag_14', 'alp_lag_14', 'alt_lag_14', 'ast_lag_14', 'be_lag_14', 'bicar_lag_14', 'bili_lag_14', 'bili_dir_lag_14', 'bnd_lag_14', 'bun_lag_14', 'ca_lag_14', 'cai_lag_14', 'ck_lag_14', 'ckmb_lag_14', 'cl_lag_14', 'crea_lag_14', 'crp_lag_14', 'dbp_lag_14', 'fgn_lag_14', 'fio2_lag_14', 'glu_lag_14', 'hgb_lag_14', 'hr_lag_14', 'inr_pt_lag_14', 'k_lag_14', 'lact_lag_14', 'lymph_lag_14', 'map_lag_14', 'mch_lag_14', 'mchc_lag_14', 'mcv_lag_14', 'methb_lag_14', 'mg_lag_14', 'na_lag_14', 'neut_lag_14', 'o2sat_lag_14', 'pco2_lag_14', 'ph_lag_14', 'phos_lag_14', 'plt_lag_14', 'po2_lag_14', 'ptt_lag_14', 'resp_lag_14', 'sbp_lag_14', 'temp_lag_14', 'tnt_lag_14', 'urine_lag_14', 'wbc_lag_14', 'alb_lag_15', 'alp_lag_15', 'alt_lag_15', 'ast_lag_15', 'be_lag_15', 'bicar_lag_15', 'bili_lag_15', 'bili_dir_lag_15', 'bnd_lag_15', 'bun_lag_15', 'ca_lag_15', 'cai_lag_15', 'ck_lag_15', 'ckmb_lag_15', 'cl_lag_15', 'crea_lag_15', 'crp_lag_15', 'dbp_lag_15', 'fgn_lag_15', 'fio2_lag_15', 'glu_lag_15', 'hgb_lag_15', 'hr_lag_15', 'inr_pt_lag_15', 'k_lag_15', 'lact_lag_15', 'lymph_lag_15', 'map_lag_15', 'mch_lag_15', 'mchc_lag_15', 'mcv_lag_15', 'methb_lag_15', 'mg_lag_15', 'na_lag_15', 'neut_lag_15', 'o2sat_lag_15', 'pco2_lag_15', 'ph_lag_15', 'phos_lag_15', 'plt_lag_15', 'po2_lag_15', 'ptt_lag_15', 'resp_lag_15', 'sbp_lag_15', 'temp_lag_15', 'tnt_lag_15', 'urine_lag_15', 'wbc_lag_15', 'alb_lag_16', 'alp_lag_16', 'alt_lag_16', 'ast_lag_16', 'be_lag_16', 'bicar_lag_16', 'bili_lag_16', 'bili_dir_lag_16', 'bnd_lag_16', 'bun_lag_16', 'ca_lag_16', 'cai_lag_16', 'ck_lag_16', 'ckmb_lag_16', 'cl_lag_16', 'crea_lag_16', 'crp_lag_16', 'dbp_lag_16', 'fgn_lag_16', 'fio2_lag_16', 'glu_lag_16', 'hgb_lag_16', 'hr_lag_16', 'inr_pt_lag_16', 'k_lag_16', 'lact_lag_16', 'lymph_lag_16', 'map_lag_16', 'mch_lag_16', 'mchc_lag_16', 'mcv_lag_16', 'methb_lag_16', 'mg_lag_16', 'na_lag_16', 'neut_lag_16', 'o2sat_lag_16', 'pco2_lag_16', 'ph_lag_16', 'phos_lag_16', 'plt_lag_16', 'po2_lag_16', 'ptt_lag_16', 'resp_lag_16', 'sbp_lag_16', 'temp_lag_16', 'tnt_lag_16', 'urine_lag_16', 'wbc_lag_16', 'alb_lag_17', 'alp_lag_17', 'alt_lag_17', 'ast_lag_17', 'be_lag_17', 'bicar_lag_17', 'bili_lag_17', 'bili_dir_lag_17', 'bnd_lag_17', 'bun_lag_17', 'ca_lag_17', 'cai_lag_17', 'ck_lag_17', 'ckmb_lag_17', 'cl_lag_17', 'crea_lag_17', 'crp_lag_17', 'dbp_lag_17', 'fgn_lag_17', 'fio2_lag_17', 'glu_lag_17', 'hgb_lag_17', 'hr_lag_17', 'inr_pt_lag_17', 'k_lag_17', 'lact_lag_17', 'lymph_lag_17', 'map_lag_17', 'mch_lag_17', 'mchc_lag_17', 'mcv_lag_17', 'methb_lag_17', 'mg_lag_17', 'na_lag_17', 'neut_lag_17', 'o2sat_lag_17', 'pco2_lag_17', 'ph_lag_17', 'phos_lag_17', 'plt_lag_17', 'po2_lag_17', 'ptt_lag_17', 'resp_lag_17', 'sbp_lag_17', 'temp_lag_17', 'tnt_lag_17', 'urine_lag_17', 'wbc_lag_17', 'alb_lag_18', 'alp_lag_18', 'alt_lag_18', 'ast_lag_18', 'be_lag_18', 'bicar_lag_18', 'bili_lag_18', 'bili_dir_lag_18', 'bnd_lag_18', 'bun_lag_18', 'ca_lag_18', 'cai_lag_18', 'ck_lag_18', 'ckmb_lag_18', 'cl_lag_18', 'crea_lag_18', 'crp_lag_18', 'dbp_lag_18', 'fgn_lag_18', 'fio2_lag_18', 'glu_lag_18', 'hgb_lag_18', 'hr_lag_18', 'inr_pt_lag_18', 'k_lag_18', 'lact_lag_18', 'lymph_lag_18', 'map_lag_18', 'mch_lag_18', 'mchc_lag_18', 'mcv_lag_18', 'methb_lag_18', 'mg_lag_18', 'na_lag_18', 'neut_lag_18', 'o2sat_lag_18', 'pco2_lag_18', 'ph_lag_18', 'phos_lag_18', 'plt_lag_18', 'po2_lag_18', 'ptt_lag_18', 'resp_lag_18', 'sbp_lag_18', 'temp_lag_18', 'tnt_lag_18', 'urine_lag_18', 'wbc_lag_18', 'alb_lag_19', 'alp_lag_19', 'alt_lag_19', 'ast_lag_19', 'be_lag_19', 'bicar_lag_19', 'bili_lag_19', 'bili_dir_lag_19', 'bnd_lag_19', 'bun_lag_19', 'ca_lag_19', 'cai_lag_19', 'ck_lag_19', 'ckmb_lag_19', 'cl_lag_19', 'crea_lag_19', 'crp_lag_19', 'dbp_lag_19', 'fgn_lag_19', 'fio2_lag_19', 'glu_lag_19', 'hgb_lag_19', 'hr_lag_19', 'inr_pt_lag_19', 'k_lag_19', 'lact_lag_19', 'lymph_lag_19', 'map_lag_19', 'mch_lag_19', 'mchc_lag_19', 'mcv_lag_19', 'methb_lag_19', 'mg_lag_19', 'na_lag_19', 'neut_lag_19', 'o2sat_lag_19', 'pco2_lag_19', 'ph_lag_19', 'phos_lag_19', 'plt_lag_19', 'po2_lag_19', 'ptt_lag_19', 'resp_lag_19', 'sbp_lag_19', 'temp_lag_19', 'tnt_lag_19', 'urine_lag_19', 'wbc_lag_19', 'alb_lag_20', 'alp_lag_20', 'alt_lag_20', 'ast_lag_20', 'be_lag_20', 'bicar_lag_20', 'bili_lag_20', 'bili_dir_lag_20', 'bnd_lag_20', 'bun_lag_20', 'ca_lag_20', 'cai_lag_20', 'ck_lag_20', 'ckmb_lag_20', 'cl_lag_20', 'crea_lag_20', 'crp_lag_20', 'dbp_lag_20', 'fgn_lag_20', 'fio2_lag_20', 'glu_lag_20', 'hgb_lag_20', 'hr_lag_20', 'inr_pt_lag_20', 'k_lag_20', 'lact_lag_20', 'lymph_lag_20', 'map_lag_20', 'mch_lag_20', 'mchc_lag_20', 'mcv_lag_20', 'methb_lag_20', 'mg_lag_20', 'na_lag_20', 'neut_lag_20', 'o2sat_lag_20', 'pco2_lag_20', 'ph_lag_20', 'phos_lag_20', 'plt_lag_20', 'po2_lag_20', 'ptt_lag_20', 'resp_lag_20', 'sbp_lag_20', 'temp_lag_20', 'tnt_lag_20', 'urine_lag_20', 'wbc_lag_20', 'alb_lag_21', 'alp_lag_21', 'alt_lag_21', 'ast_lag_21', 'be_lag_21', 'bicar_lag_21', 'bili_lag_21', 'bili_dir_lag_21', 'bnd_lag_21', 'bun_lag_21', 'ca_lag_21', 'cai_lag_21', 'ck_lag_21', 'ckmb_lag_21', 'cl_lag_21', 'crea_lag_21', 'crp_lag_21', 'dbp_lag_21', 'fgn_lag_21', 'fio2_lag_21', 'glu_lag_21', 'hgb_lag_21', 'hr_lag_21', 'inr_pt_lag_21', 'k_lag_21', 'lact_lag_21', 'lymph_lag_21', 'map_lag_21', 'mch_lag_21', 'mchc_lag_21', 'mcv_lag_21', 'methb_lag_21', 'mg_lag_21', 'na_lag_21', 'neut_lag_21', 'o2sat_lag_21', 'pco2_lag_21', 'ph_lag_21', 'phos_lag_21', 'plt_lag_21', 'po2_lag_21', 'ptt_lag_21', 'resp_lag_21', 'sbp_lag_21', 'temp_lag_21', 'tnt_lag_21', 'urine_lag_21', 'wbc_lag_21', 'alb_lag_22', 'alp_lag_22', 'alt_lag_22', 'ast_lag_22', 'be_lag_22', 'bicar_lag_22', 'bili_lag_22', 'bili_dir_lag_22', 'bnd_lag_22', 'bun_lag_22', 'ca_lag_22', 'cai_lag_22', 'ck_lag_22', 'ckmb_lag_22', 'cl_lag_22', 'crea_lag_22', 'crp_lag_22', 'dbp_lag_22', 'fgn_lag_22', 'fio2_lag_22', 'glu_lag_22', 'hgb_lag_22', 'hr_lag_22', 'inr_pt_lag_22', 'k_lag_22', 'lact_lag_22', 'lymph_lag_22', 'map_lag_22', 'mch_lag_22', 'mchc_lag_22', 'mcv_lag_22', 'methb_lag_22', 'mg_lag_22', 'na_lag_22', 'neut_lag_22', 'o2sat_lag_22', 'pco2_lag_22', 'ph_lag_22', 'phos_lag_22', 'plt_lag_22', 'po2_lag_22', 'ptt_lag_22', 'resp_lag_22', 'sbp_lag_22', 'temp_lag_22', 'tnt_lag_22', 'urine_lag_22', 'wbc_lag_22', 'alb_lag_23', 'alp_lag_23', 'alt_lag_23', 'ast_lag_23', 'be_lag_23', 'bicar_lag_23', 'bili_lag_23', 'bili_dir_lag_23', 'bnd_lag_23', 'bun_lag_23', 'ca_lag_23', 'cai_lag_23', 'ck_lag_23', 'ckmb_lag_23', 'cl_lag_23', 'crea_lag_23', 'crp_lag_23', 'dbp_lag_23', 'fgn_lag_23', 'fio2_lag_23', 'glu_lag_23', 'hgb_lag_23', 'hr_lag_23', 'inr_pt_lag_23', 'k_lag_23', 'lact_lag_23', 'lymph_lag_23', 'map_lag_23', 'mch_lag_23', 'mchc_lag_23', 'mcv_lag_23', 'methb_lag_23', 'mg_lag_23', 'na_lag_23', 'neut_lag_23', 'o2sat_lag_23', 'pco2_lag_23', 'ph_lag_23', 'phos_lag_23', 'plt_lag_23', 'po2_lag_23', 'ptt_lag_23', 'resp_lag_23', 'sbp_lag_23', 'temp_lag_23', 'tnt_lag_23', 'urine_lag_23', 'wbc_lag_23', 'alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24' ]
+PredictionDatasetpytorch.time_varying_known_reals=[]
 
 PredictionDatasetpytorch.add_relative_time_idx=False
-PredictionDatasetpytorch.target="label"
-            
-PredictionDatasetpytorch.time_varying_unknown_reals=[
+PredictionDatasetpytorch.target=[
+    "alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
                 "label",
               
             ]
-PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetpytorch.lagged_variables=[  "alb",
+
+            
+PredictionDatasetpytorch.time_varying_unknown_reals=[
+                "alb",
                 "alp",
                 "alt",
                 "ast",
@@ -87,4 +134,10 @@ PredictionDatasetpytorch.lagged_variables=[  "alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc"]
\ No newline at end of file
+                "wbc",
+                "label",
+              
+            ]
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.lagged_variables=[  ]
+PredictionDatasetpytorch.target_normalizer='multi'
\ No newline at end of file
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index c7264ed5..de79f61f 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -89,3 +89,5 @@ PredictionDatasetpytorch.time_varying_unknown_reals=["alb",
                 "urine",
                 "wbc",]
 PredictionDatasetpytorch.lagged_variables=[]
+PredictionDatasetpytorch.lagged_variables=[]
+PredictionDatasetpytorch.target_normalizer='single'
\ No newline at end of file
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 535d9619..69db2960 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -38,9 +38,9 @@ def execute_repeated_cv(
     verbose: bool = False,
     wandb: bool = False,
     complete_train: bool = False,
-    explain: bool = False,
+    explain: list = None,
     pytorch_forecasting: bool = False,
-    XAI_metric: bool = False,
+    XAI_metric: list = None,
     random_labels: bool = False,
     random_model_dir: str = None,
 ) -> float:
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index a0e51dbc..49c898b7 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -10,7 +10,7 @@
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
-from pytorch_forecasting import TimeSeriesDataSet, GroupNormalizer
+from pytorch_forecasting import TimeSeriesDataSet, GroupNormalizer, MultiNormalizer, EncoderNormalizer
 
 
 class CommonDataset(Dataset):
@@ -450,6 +450,7 @@ def __init__(
         time_varying_known_reals: List[str],
         time_varying_unknown_categoricals: List[str],
         lagged_variables: List[str],
+        # target_normalizer: str,
         *args,
         ram_cache: bool = False,
         add_relative_time_idx: bool = False,
@@ -482,7 +483,13 @@ def __init__(
         self.ram_cache = ram_cache
         self.kwargs = kwargs
         self.column_names = features.columns
-
+        """ if target_normalizer == 'multi':
+            target_normalizer = MultiNormalizer([EncoderNormalizer(transformation='relu') for _ in range(len(target)-1)] + [GroupNormalizer(groups=["stay_id"], transformation="relu")]
+                                                )
+        else:
+            target_normalizer = GroupNormalizer(
+                groups=["stay_id"], transformation="relu"
+            ) """
         super().__init__(
             data=self.data,
             time_idx="time_idx",
@@ -502,6 +509,7 @@ def __init__(
             # add_target_scales=True,
             # add_encoder_length=True,
             predict_mode=True,
+
             target_normalizer=GroupNormalizer(
                 groups=["stay_id"], transformation="relu"
             ),
diff --git a/icu_benchmarks/models/constants.py b/icu_benchmarks/models/constants.py
index 093d15bd..702fe7f7 100644
--- a/icu_benchmarks/models/constants.py
+++ b/icu_benchmarks/models/constants.py
@@ -26,6 +26,9 @@
     MAE,
     JSD,
     BinaryFairnessWrapper,
+    Faithfulness,
+    Stability,
+    Randomization
 )
 
 
@@ -85,6 +88,17 @@ class DLMetrics:
         "mae": MAE,
         "jsd": JSD,
     }
+    Faithfulness = {
+        "Faithfulness_timesteps": Faithfulness,
+        "Faithfulness_features": Faithfulness,
+        "Faithfulness_feature_timestep": Faithfulness,
+    }
+    Stability = {
+        "Stability": Stability
+    }
+    Randmoization = {
+        "DDR": Randomization,
+    }
 
 
 class ImputationInit(str, Enum):
diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index d1dbc271..ee942a2d 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -6,6 +6,8 @@
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
 from torchmetrics.classification import BinaryFairness
+from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson, cosine
+
 
 """"
 This file contains custom metrics that can be added to YAIB.
@@ -132,3 +134,422 @@ def feature_helper(self, trainer, step_prefix):
         else:
             feature_names = trainer.test_dataloaders.dataset.features
         return feature_names
+
+# XAI Metrics
+
+
+class Faithfulness(EpochMetric):
+    def __init__(
+        self,
+        output_transform: Callable = lambda x: x,
+        check_compute_fn: bool = False,
+        *args, **kwargs
+    ) -> None:
+        super().__init__(output_transform,
+                         check_compute_fn, *args, **kwargs
+                         )
+
+    def update(
+            self,
+            x,
+            attribution,
+            model,
+            similarity_func=None,
+            nr_runs=100,
+            pertrub=None,
+            subset_size=3,
+            feature=False,
+            time_step=False,
+            feature_timestep=False,
+            device='cuda'
+
+    ):
+        """
+        Calculates faithfulness scores for captum attributions 
+
+        Args:
+            - x:Batch input
+            -attribution: attribution generated by captum,
+            - similarity_func:function to determine similarity between sum of attributions and difference in prediction
+            - nr_runs: How many times to repeat the experiment,
+            - pertrub: What change to do to the input,
+            - subset_size: The size of the subset of featrues to alter ,
+            - feature: Determines if to calcualte faithfulness of feature attributions,
+            - time_step: Determines if to calcualte faithfulness of timesteps attributions,
+            - feature_timestep: Determines if to calcualte faithfulness of featrues per timesteps attributions,
+        Returns:
+            score: similarity score between sum of attributions and difference in prediction averaged over nr_runs
+
+        Implementation of faithfulness correlation by Bhatt et al., 2020.
+
+        The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
+        (or 'fidelity') with respect to the model behaviour.
+
+        Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
+        the average explanation attribution for only the subset of features are (linearly) correlated, taking the
+        average over multiple runs and test samples. The metric returns one float per input-attribution pair that
+        ranges between -1 and 1, where higher scores are better.
+
+        For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
+        or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
+        test point and the average explanation attribution for only the subset of features is calculated. Results is
+        average over multiple runs and several test samples.
+        This code is adapted from the quantus libray to suit our use case
+
+        References:
+            1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
+            explanations." IJCAI (2020): 3016-3022.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+        """
+        def add_noise(x, indices, time_step, feature_timestep):
+            noise = torch.randn_like(x["encoder_cont"])
+            if time_step:
+                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing='ij')
+
+                with torch.no_grad():
+                    x["encoder_cont"][idx0, idx1, :] += noise[idx0, idx1, :]
+
+            elif feature:
+                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing='ij')
+
+                with torch.no_grad():
+                    x["encoder_cont"][idx0, :, idx1] += noise[idx0, :, idx1]
+
+            elif feature_timestep:
+                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing='ij')
+
+                with torch.no_grad():
+                    x["encoder_cont"][idx0, idx1, idx2] += noise[idx0, idx1, idx2]
+
+        def apply_baseline(x, indices, time_step, feature_timestep):
+            mask = torch.ones_like(x["encoder_cont"])
+            if time_step:
+
+                idx0, idx1, = np.meshgrid(indices[0], indices[1], indexing='ij')
+
+                mask[idx0, idx1, :] -= mask[idx0, idx1, :]
+            elif feature:
+                idx0, idx1, = np.meshgrid(indices[0], indices[1], indexing='ij')
+
+                mask[idx0, :, idx1] -= mask[idx0, :, idx1]
+
+            elif feature_timestep:
+                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing='ij')
+
+                mask[idx0, idx1, idx2] -= mask[idx0, idx1, idx2]
+
+            with torch.no_grad():
+                x["encoder_cont"] *= mask
+        # Assuming 'attribution' is already a GPU tensor
+        if not torch.is_tensor(attribution):
+            attribution = torch.tensor(attribution).to(device)
+
+        # Other initializations
+        if similarity_func is None:
+            similarity_func = correlation_spearman
+        if pertrub is None:
+            pertrub = "baseline"
+        similarities = []
+
+        # Assuming this is a method to prepare your data
+
+        y_pred = model(model.prep_data(x)).detach()  # Keep on GPU
+        pred_deltas = []
+        att_sums = []
+
+        for i_ix in range(nr_runs):
+            if time_step:
+                timesteps_idx = np.random.choice(24, subset_size, replace=False)
+                patient_idx = np.random.choice(64, 1, replace=False)
+                a_ix = [patient_idx, timesteps_idx]
+
+            elif feature:
+                feature_idx = np.random.choice(53, subset_size, replace=False)
+                patient_idx = np.random.choice(64, 1, replace=False)
+                a_ix = [patient_idx, feature_idx]
+            elif feature_timestep:
+                timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
+                feature_idx = np.random.choice(53, subset_size[1], replace=False)
+                patient_idx = np.random.choice(64, 1, replace=False)
+                a_ix = [patient_idx, timesteps_idx, feature_idx]
+
+            # Apply perturbation
+            if pertrub == "Noise":
+                add_noise(x, a_ix, time_step, feature_timestep)
+            elif pertrub == "baseline":
+                apply_baseline(x, a_ix, time_step, feature_timestep)
+
+            # Predict on perturbed input and calculate deltas
+            y_pred_perturb = (model(model.prep_data(x))).detach()  # Keep on GPU
+
+            if time_step:
+
+                if attribution.size() == torch.Size([24]):
+                    att_sums.append((attribution[timesteps_idx]).sum())
+                else:
+                    att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :]).sum())
+            elif feature:
+
+                if len(attribution) == 53:
+                    att_sums.append((attribution[feature_idx]).sum())
+                else:
+
+                    att_sums.append((attribution[patient_idx, :, :][:, :, feature_idx]).sum())
+            elif feature_timestep:
+
+                att_sums.append((attribution[patient_idx, :, :]
+                                [:, timesteps_idx, :][:, :, feature_idx]).sum())
+
+            pred_deltas.append((y_pred - y_pred_perturb)[patient_idx].item())
+            # Convert to CPU for numpy operations
+
+        pred_deltas_cpu = torch.tensor(pred_deltas).cpu().numpy()
+        att_sums_cpu = torch.tensor(att_sums).cpu().numpy()
+
+        similarities.append(similarity_func(pred_deltas_cpu, att_sums_cpu))
+
+        score = np.nanmean(similarities)
+        return score
+
+
+class Stability(EpochMetric):
+    def __init__(
+        self, output_transform: Callable = lambda x: x,
+        check_compute_fn: bool = False,
+        *args, **kwargs
+    ) -> None:
+        super().__init__(output_transform,
+                         check_compute_fn, *args, **kwargs
+                         )
+
+    def update(self, x,
+               attribution, model, explain_method, method_name, dataloader=None, thershold=0.5, device='cuda', **kwargs
+               ):
+        """
+    Args:
+            - x:Batch input
+            -attribution: attribution 
+            - explain_method:function to generate explantations
+            - method_name: Name of the explantation
+            - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+
+
+        Returns:
+            RIS : relative distance between the explantation and the input
+            ROS: relative distance between the explantation and the output
+
+
+    References:
+            1) `https://arxiv.org/pdf/2203.06877.pdf
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+
+        """
+
+        def relative_stability_objective(
+            x, xs, e_x, e_xs, close_indices, eps_min=0.0001, input=False, attention=False, device='cuda'
+        ) -> torch.Tensor:
+            """
+            Computes relative input and output stabilities maximization objective
+            as defined here :ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
+
+            Args:
+
+                x: Input tensor
+                xs: perturbed tensor.
+                e_x: Explanations for x.
+                e_xs: Explanations for xs.
+                eps_min:Value to avoid division by zero if needed 
+                input:Boolean to indicate if this is an input or an output
+                device: the device to keep the tensors on 
+
+            Returns:
+
+                ris_obj: Tensor
+                    RIS maximization objective.
+            """
+            # Function to convert inputs to tensors if they are numpy arrays
+            def to_tensor(input_array):
+                if isinstance(input_array, np.ndarray):
+                    return torch.index_select(torch.tensor(input_array).to(device), 0, close_indices)
+
+                return torch.index_select(input_array.to(device), 0, close_indices)
+
+            # Convert all inputs to tensors and move to GPU
+            if attention:
+                x, xs = map(to_tensor, [x, xs])
+            else:
+                x, xs, e_x, e_xs = map(to_tensor, [x, xs, e_x, e_xs])
+
+            if input:
+                num_dim = x.ndim
+            else:
+                num_dim = e_x.ndim
+
+            if num_dim == 3:
+                def norm_function(arr): return torch.norm(arr, dim=(-1, -2))
+            elif num_dim == 2:
+                def norm_function(arr): return torch.norm(arr, dim=-1)
+            else:
+                def norm_function(arr): return torch.norm(arr)
+
+            nominator = (e_x - e_xs) / (e_x + (e_x == 0) * eps_min)
+            nominator = norm_function(nominator)
+
+            if input:
+                denominator = x - xs
+                denominator /= x + (x == 0) * eps_min
+                denominator = norm_function(denominator)
+                denominator += (denominator == 0) * eps_min
+            else:
+                denominator = torch.squeeze(x) - torch.squeeze(xs)
+                denominator = torch.norm(denominator, dim=-1)
+                denominator += (denominator == 0) * eps_min
+
+            return nominator / denominator
+
+        if explain_method == "Attention":
+            y_pred = model.model.predict(dataloader)
+            x_original = dataloader.dataset.data["reals"].clone()
+
+            dataloader.dataset.add_noise()
+            x_preturb = dataloader.dataset.data["reals"].clone()
+            y_pred_preturb = model.model.predict(dataloader)
+            Attention_weights = model.interpertations(dataloader)
+            att_preturb = Attention_weights["attention"]
+            # Calculate the absolute difference
+            difference = torch.abs(y_pred_preturb - y_pred)
+
+            # Find where the difference is less than or equal to a thershold
+            close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
+
+            RIS = relative_stability_objective(
+                x_original.detach(), x_preturb.detach(), attribution, att_preturb, close_indices=close_indices, input=True, attention=True
+            )
+            ROS = relative_stability_objective(
+                y_pred, y_pred_preturb, attribution, att_preturb, close_indices=close_indices, input=False, attention=True
+            )
+
+        else:
+            y_pred = model(model.prep_data(x)).detach()
+            x_original = x["encoder_cont"].detach().clone()
+
+            with torch.no_grad():
+                noise = torch.randn_like(x["encoder_cont"])*0.01
+                x["encoder_cont"] += noise
+            y_pred_preturb = model(model.prep_data(x)).detach()
+            if explain_method == "Random":
+                att_preturb = np.random.normal(size=[64, 24, 53])
+
+            else:
+
+                data, baselines = model.prep_data_captum(x)
+
+                explantation = explain_method(model.forward_captum)
+                # Reformat attributions.
+                if explain_method is not captum.attr.Saliency:
+                    att_preturb = explantation.attribute(data, baselines=baselines, **kwargs)
+                else:
+                    att_preturb = explantation.attribute(data, **kwargs)
+
+                # Process and store the calculated attributions
+                att_preturb = att_preturb[1].detach() if method_name in [
+                    'Lime', 'FeatureAblation'] else torch.stack(att_preturb).detach()
+            # Calculate the absolute difference
+            difference = torch.abs(y_pred_preturb - y_pred)
+
+            # Find where the difference is less than or equal to a thershold
+            close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
+
+            RIS = relative_stability_objective(
+                x_original.detach(),
+                x["encoder_cont"].detach(),
+                attribution, att_preturb, close_indices=close_indices, input=True
+            )
+            ROS = relative_stability_objective(
+                y_pred, y_pred_preturb, attribution, att_preturb, close_indices=close_indices, input=False
+            )
+
+        return np.max(RIS.cpu().numpy()).astype(np.float64), np.max(ROS.cpu().numpy()).astype(np.float64)
+
+
+class Randomization(EpochMetric):
+    def __init__(
+        self, output_transform: Callable = lambda x: x,
+        check_compute_fn: bool = False,
+        *args, **kwargs
+    ) -> None:
+        super().__init__(output_transform,
+                         check_compute_fn, *args, **kwargs
+                         )
+
+    def update(
+        self, x,
+        attribution, model, explain_method, random_model, similarity_func=cosine, dataloader=None, method_name="", **kwargs
+    ):
+        """
+
+        Args:
+            - x:Batch input
+            -attribution: attribution 
+            - explain_method:function to generate explantations
+            - random_model: Reference to model trained on random labels
+            - similarity_func: Function to measure similiarity 
+            - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+            - method_name: Name of the explantation
+
+        Returns:
+            score: similarity score between attributions of model trained on random data and model trained on real data
+
+        Implementation of the Random Logit Metric by Sixt et al., 2020.
+
+        The Random Logit Metric computes the distance between the original explanation and a reference explanation of
+        a randomly chosen non-target class.
+        This code is adapted from the quantus libray to suit our use case
+
+        References:
+            1) Leon Sixt et al.: "When Explanations Lie: Why Many Modified BP
+            Attributions Fail." ICML (2020): 9046-9057.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+
+        """
+
+        if explain_method == "Attention":
+            Attention_weights = random_model.interpertations(dataloader)
+            attribution = attribution.cpu().numpy()
+            min_val = np.min(attribution)
+            max_val = np.max(attribution)
+
+            attribution = (attribution - min_val) / (max_val - min_val)
+            random_attr = Attention_weights["attention"].cpu().numpy()
+            min_val = np.min(random_attr)
+            max_val = np.max(random_attr)
+            random_attr = (random_attr - min_val) / (max_val - min_val)
+            score = similarity_func(random_attr, attribution)
+        elif explain_method == "Random":
+            score = similarity_func(np.random.normal(size=[64, 24, 53]).flatten(), attribution.flatten())
+        else:
+            data, baselines = model.prep_data_captum(x)
+            y_pred = model(data).detach()
+
+            explantation = explain_method(random_model.forward_captum)
+            # Reformat attributions.
+            if explain_method is not captum.attr.Saliency:
+                attr = explantation.attribute(data, baselines=baselines, **kwargs)
+            else:
+                attr = explantation.attribute(data, **kwargs)
+
+            # Process and store the calculated attributions
+            random_attr = attr[1].cpu().detach().numpy() if method_name in [
+                'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
+
+            attribution = attribution.flatten()
+            min_val = np.min(attribution)
+            max_val = np.max(attribution)
+            attribution = (attribution - min_val) / (max_val - min_val)
+            random_attr = random_attr.flatten()
+            min_val = np.min(random_attr)
+            max_val = np.max(random_attr)
+            random_attr = (random_attr - min_val) / (max_val - min_val)
+
+            score = similarity_func(random_attr, attribution)
+        return score
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index ff66c4da..0966dec7 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -12,7 +12,7 @@
 )
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper, DLPredictionPytorchForecastingWrapper
-from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, is_tensor
+from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, is_tensor, autograd
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss, MAE
 import matplotlib.pyplot as plt
@@ -475,8 +475,8 @@ def __init__(
         *args,
         **kwargs,
     ):
-        super().__init__(optimizer=optimizer, pytorch_forecasting=True, *args, **kwargs)
-        self.dataset = dataset
+        super().__init__(optimizer=optimizer, pytorch_forecasting=True, dataset=dataset, * args, **kwargs)
+
         self.model = TemporalFusionTransformer.from_dataset(
             dataset=dataset,
             hidden_size=hidden,
@@ -486,6 +486,7 @@ def __init__(
             loss=QuantileLoss(),
             hidden_continuous_size=hidden,
         )
+        self.dataset = dataset
         self.num_classes = num_classes
         self.logit = nn.Linear(7, num_classes)
 
@@ -575,7 +576,10 @@ def forward_captum(
             target_scale,
         )
 
-        return self.forward(tuple_x)
+        # with autograd.set_grad_enabled(True):
+        result = self.forward(tuple_x)
+
+        return result
 
 
 @gin.configurable
@@ -634,9 +638,11 @@ def forward(
 
         out = self.model(x_dict)
 
-        pred = self.logit(out["prediction"])
+        out = ((out["prediction"][-1]).squeeze(dim=-1))
 
-        return pred
+        pred = self.logit(out)
+
+        return pred.unsqueeze(1)
 
 
 @gin.configurable
@@ -694,6 +700,8 @@ def forward(
             x_dict["encoder_cont"][:, :, -1] = 0.0
         out = self.model(x_dict)
 
-        pred = self.logit(out["prediction"])
+        out = ((out["prediction"][-1]))
+
+        pred = self.logit(out)
 
         return pred
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 73116f46..e1566713 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -73,9 +73,9 @@ def train_common(
         torch.cuda.device_count() * 8 * int(torch.cuda.is_available()),
         32,
     ),
-    explain: list = [],
+    explain: list = None,
     pytorch_forecasting: bool = False,
-    XAI_metric: list = [],
+    XAI_metric: list = None,
     random_labels: bool = False,
     random_model_dir: str = None,
 
@@ -154,6 +154,12 @@ def train_common(
             drop_last=True,
             shuffle=False,
         )
+        if random_model_dir is None:
+            random_model = None
+        else:
+
+            random_model = Path(random_model_dir)
+
         if load_weights:
             model = load_model(
                 model,
@@ -161,6 +167,11 @@ def train_common(
                 pl_model=pl_model,
                 train_dataset=train_dataset,
                 optimizer=optimizer,
+                explain=explain,
+                XAI_metric=XAI_metric,
+                random_model=random_model,
+                test_loader=test_loader
+
             )
 
         else:
@@ -170,6 +181,11 @@ def train_common(
                 epochs=epochs,
                 run_mode=mode,
                 batch_size=batch_size,
+                explain=explain,
+                XAI_metric=XAI_metric,
+                random_model=random_model,
+                test_loader=test_loader
+
             )
         if random_labels:
             train_dataset.randomize_labels(num_classes=model.num_classes)
@@ -246,7 +262,8 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
-        # gradient_clip_val=gradient_clip_val
+        gradient_clip_val=gradient_clip_val,
+        inference_mode=False
     )
     if not eval_only:
         if model.requires_backprop:
@@ -264,21 +281,7 @@ def train_common(
         save_config_file(log_dir)
         return 0
 
-    if explain:
-        if random_model_dir is None:
-            random_model = None
-        else:
-            path = Path(random_model_dir)
-
-            random_model = load_model(
-                model,
-                source_dir=path,
-                pl_model=pl_model,
-                train_dataset=train_dataset,
-                optimizer=optimizer,
-            )
-
-        XAI_dict = {}  # dictrionary to log attributions metrics
+        """ XAI_dict = {}  # dictrionary to log attributions metrics
 
         # choose which  methods to get attributions
         methods = {
@@ -370,7 +373,7 @@ def create_default_mask(shape):
 
         # Write the dictionary to a JSON file
         with open(json_file_path, "w") as json_file:
-            json.dump(XAI_dict, json_file)
+            json.dump(XAI_dict, json_file) """
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
@@ -378,7 +381,7 @@ def create_default_mask(shape):
     return test_loss
 
 
-def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=None):
+def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=None, explain=None, XAI_metric=None, random_model=None, test_loader=None):
     if source_dir.exists():
         if model.requires_backprop:
             if (source_dir / "model.ckpt").exists():
@@ -391,7 +394,8 @@ def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=N
                 return Exception(f"No weights to load at path : {source_dir}")
             if pl_model:
                 if train_dataset is not None:
-                    model = model.load_from_checkpoint(model_path, dataset=train_dataset, optimizer=optimizer)
+                    model = model.load_from_checkpoint(model_path, dataset=train_dataset, optimizer=optimizer,
+                                                       explain=explain, XAI_metric=XAI_metric, test_loader=test_loader)
 
                 else:
                     model = model.load_from_checkpoint(model_path)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index b97d1306..b0cd933a 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -18,9 +18,10 @@
 from pytorch_lightning import LightningModule
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
+from icu_benchmarks.models.train import load_model
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson, cosine
-import captum
+from captum.attr import IntegratedGradients, Saliency, FeatureAblation, Lime
 from captum._utils.models.linear_model import SkLearnLasso
 import os
 
@@ -242,8 +243,7 @@ def __init__(
         input_size: Tensor = None,
         initialization_method: str = "normal",
         pytorch_forecasting: bool = False,
-        explain: list = [],
-        XAI_metric: list = [],
+
         **kwargs,
     ):
         super().__init__(
@@ -264,8 +264,6 @@ def __init__(
         self.output_transform = None
         self.loss_weights = None
         self.pytorch_forecasting = pytorch_forecasting
-        self.explain = explain
-        self.XAI_metric = XAI_metric
 
     def set_weight(self, weight, dataset):
         """Set the weight for the loss function."""
@@ -408,6 +406,11 @@ def __init__(
         input_size: Tensor = None,
         initialization_method: str = "normal",
         pytorch_forecasting: bool = False,
+        explain: list = None,
+        XAI_metric: list = None,
+        random_model=None,
+        test_loader=None,
+        dataset=None,
         **kwargs,
     ):
         super().__init__(
@@ -425,6 +428,55 @@ def __init__(
             initialization_method=initialization_method,
             kwargs=kwargs,
         )
+        self.explain = explain
+        self.XAI_metric = XAI_metric
+        self.methods = {
+            "G": Saliency,
+            "L": Lime,
+            "IG": IntegratedGradients,
+            "FA": FeatureAblation,
+            "R": "Random",
+            "Att": "Attention"
+        }
+        self.random_model = random_model
+        self.test_loader = test_loader
+
+        if random_model is not None:
+            self.random_model = load_model(
+                model=self,
+                source_dir=random_model,
+                pl_model=True,
+                train_dataset=dataset,
+                optimizer=optimizer,
+            )
+        else:
+            self.random_model = None
+
+    def set_metrics(self, *args):
+        """Set the evaluation metrics for the prediction model."""
+
+        metrics = super().set_metrics(*args)
+
+        if self.XAI_metric is not None:
+            for metric in self.XAI_metric:
+                if metric == "Faithfulness":
+                    metrics.update(DLMetrics.Faithfulness)
+                elif metric == "Stability":
+                    metrics.update(DLMetrics.Stability)
+                elif metric == "Randmoization":
+                    metrics.update(DLMetrics.Randmoization)
+
+        return metrics
+
+    def on_train_start(self):
+        self.metrics = {
+            step_name: {
+                metric_name: (metric() if isinstance(metric, type) else metric)
+                for metric_name, metric in self.set_metrics().items()
+            }
+            for step_name in ["train", "val", "test"]
+        }
+        return super().on_train_start()
 
     def step_fn(self, element, step_prefix=""):
         """Perform a step in the DL prediction model training loop.
@@ -449,7 +501,6 @@ def step_fn(self, element, step_prefix=""):
         else:
             aux_loss = 0
         # Get prediction and target
-
         prediction = out.to(self.device).squeeze(-1)
 
         target = labels.to(self.device)
@@ -460,13 +511,112 @@ def step_fn(self, element, step_prefix=""):
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
-
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
             raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
         transformed_output = self.output_transform((prediction, target))
+        if self.explain is not None:
+
+            for method in self.explain:
+                if method == "Attention":
+                    all_attrs, features_attrs, timestep_attrs = self.explantation2(dic, method)
+                else:
+                    all_attrs, features_attrs, timestep_attrs = self.explantation2(dic, method)
+                for key, value in self.metrics[step_prefix].items():
+                    if key == "Faithfulness_timesteps":
+                        value.update(
+                            dic,
+                            all_attrs,
+                            self,
+                            correlation_pearson,
+                            100,
+                            "baseline",
+                            4,
+                            False,
+                            True,
+                            False
+
+                        )
+                    elif key == "Faithfulness_features":
+                        value.update(
+                            dic,
+                            all_attrs,
+                            self,
+                            correlation_pearson,
+                            100,
+                            "baseline",
+                            9,
+                            True,
+                            False,
+                            False
+
+                        )
+                    elif key == "Faithfulness_feature_timestep":
+                        value.update(
+                            dic,
+                            all_attrs,
+                            self,
+                            correlation_pearson,
+                            100,
+                            "baseline",
+                            [4, 9],
+                            False,
+                            False,
+                            True
+
+                        )
+
+                    elif key == "Stability":
+                        if method == "Attention":
+
+                            value.update(
+                                dic,
+                                timestep_attrs,
+                                self,
+                                self.methods[method],
+                                method,
+                                self.test_loader,
+                                0.5,
+
+                            )
+                        else:
+                            value.update(
+                                dic,
+                                all_attrs,
+                                self,
+                                self.methods[method],
+                                method,
+                                None,
+                                0.5,
+                            )
+                    elif key == "Randomization":
+
+                        if method == "Attention":
+                            value.update(
+                                dic,
+                                timestep_attrs,
+                                self,
+                                self.methods[method],
+                                self.random_model,
+                                cosine,
+                                None,
+                                method
+                            )
+                        else:
+
+                            value.update(
+                                dic,
+                                all_attrs,
+                                self,
+                                self.methods[method],
+                                self.random_model,
+                                cosine,
+                                self.test_loader,
+                                method
+                            )
 
         for key, value in self.metrics[step_prefix].items():
+
             if isinstance(value, torchmetrics.Metric):
                 if key == "Binary_Fairness":
                     feature_names = self.metrics[step_prefix][key].feature_helper(self.trainer, step_prefix)
@@ -476,6 +626,10 @@ def step_fn(self, element, step_prefix=""):
                         data,
                         feature_names,
                     )
+                elif (key == "Faithfulness_timesteps" or key == "Faithfulness_features" or key == "Faithfulness_feature_timestep"
+                      or key == "Stability" or key == "Randomization"):
+                    continue
+
                 else:
                     value.update(transformed_output[0], transformed_output[1].int())
             else:
@@ -575,6 +729,86 @@ def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir
         plt.tight_layout()
         plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
 
+    def explantation2(self, dic, method, dataloader=None, log_dir=".", plot=False, **kwargs):
+        """
+        Generic method to combine pytorchforecasting data loading , interpertations and captum to generate attributions
+
+        Args:
+            - dataloader: pytorchforecasting data loader 
+            - method: The explantation method chosen
+            - log_dir= The directory to output the plots
+            - plot= Determines if plots should be done or not
+            - XAI_metric=Determines if XAI metrics should be calculated or not
+        Returns:
+            - all_attrs : Attribtuons of features per timesteps
+            - features_attrs : Attribtuons of features averaged over timesteps
+            - timestep_attrs : Attribtuons of timesteps averaged over features
+            - f_ts_v_score: Faithfulness score for attribtuons of features per timesteps
+            - f_ts_score: Faithfulness score for attribtuons of timesteps averaged over features
+        """
+        # Initialize lists to store attribution values for all instances
+
+        all_attrs = []
+        method = self.methods[method]
+        for key, value in dic.items():
+            dic[key] = dic[key].to(self.device)
+
+        method_name = method if (method == "Random") or (method == "Attention") else (
+            method.__name__)
+        if (method_name == "Random") or (method_name == "Attention"):
+            if method_name == "Attention":
+                Interpertations = self.interpertations(dataloader=dataloader, log_dir=log_dir, plot=plot)
+                timestep_attrs = Interpertations["attention"]
+                features_attrs = Interpertations["static_variables"].tolist()
+                features_attrs.extend(Interpertations["encoder_variables"].tolist())
+
+            elif method_name == "Random":
+                # Generate random attributions for baseline comparison
+                all_attrs = np.random.normal(size=[64, 24, 53])
+                features_attrs = all_attrs.mean(axis=(1))
+                timestep_attrs = all_attrs.mean(axis=(2))
+
+            return all_attrs, features_attrs, timestep_attrs
+        # Loop through the dataloader to compute attributions for all instances
+
+        data, baselines = self.prep_data_captum(dic)
+
+        # Initialize the explanation method
+        self.train()
+        explanation = method(self.forward_captum, interpretable_model=SkLearnLasso(
+            alpha=0.4)) if method_name == 'Lime' else method(self.forward_captum)
+
+        # Calculate attributions using the selected method
+        if method is not Saliency:
+            attr = explanation.attribute(data, baselines=baselines, **kwargs)
+        else:
+            attr = explanation.attribute(data, **kwargs)
+        self.eval()
+        # Process and store the calculated attributions
+        stacked_attr = attr[1].cpu().detach().numpy() if method_name in [
+            'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
+
+        # aggregate over batch
+        attr = np.mean(stacked_attr, axis=0)
+        all_attrs.append(attr)
+        # aggregate over all batches
+        all_attrs = np.array(all_attrs).mean(axis=(0))
+        # aggregate over all timesteps
+        features_attrs = all_attrs.mean(axis=(0))
+        # aggregate over all features
+        timestep_attrs = all_attrs.mean(axis=(1))
+
+        if plot:
+            log_dir_plots = log_dir / 'plots'
+            if not (log_dir_plots.exists()):
+                log_dir_plots.mkdir(parents=True)
+            # Plot attributions for features and timesteps
+
+            self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots)
+
+        # Return computed attributions and metrics
+        return all_attrs, features_attrs, timestep_attrs
+
     def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=False, random_model=None, test_dataset=None, **kwargs):
         """
         Generic method to combine pytorchforecasting data loading , interpertations and captum to generate attributions

From 1534c13ebe15532e6c282f26e18e94c2b9bea1b0 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 27 Dec 2023 20:32:50 +0100
Subject: [PATCH 124/142] fixes

---
 icu_benchmarks/models/custom_metrics.py | 35 +++---------
 icu_benchmarks/models/wrappers.py       | 76 ++++++++++++++++++-------
 2 files changed, 62 insertions(+), 49 deletions(-)

diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index ee942a2d..9b84a18d 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -243,7 +243,6 @@ def apply_baseline(x, indices, time_step, feature_timestep):
         # Assuming 'attribution' is already a GPU tensor
         if not torch.is_tensor(attribution):
             attribution = torch.tensor(attribution).to(device)
-
         # Other initializations
         if similarity_func is None:
             similarity_func = correlation_spearman
@@ -287,16 +286,15 @@ def apply_baseline(x, indices, time_step, feature_timestep):
                 if attribution.size() == torch.Size([24]):
                     att_sums.append((attribution[timesteps_idx]).sum())
                 else:
-                    att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :]).sum())
+                    att_sums.append((attribution[patient_idx, :][:, timesteps_idx]).sum())
             elif feature:
 
                 if len(attribution) == 53:
                     att_sums.append((attribution[feature_idx]).sum())
                 else:
 
-                    att_sums.append((attribution[patient_idx, :, :][:, :, feature_idx]).sum())
+                    att_sums.append((attribution[patient_idx, :][:, feature_idx]).sum())
             elif feature_timestep:
-
                 att_sums.append((attribution[patient_idx, :, :]
                                 [:, timesteps_idx, :][:, :, feature_idx]).sum())
 
@@ -323,7 +321,7 @@ def __init__(
                          )
 
     def update(self, x,
-               attribution, model, explain_method, method_name, dataloader=None, thershold=0.5, device='cuda', **kwargs
+               attribution, model, explain_method, dataloader=None, thershold=0.5, device='cuda', **kwargs
                ):
         """
     Args:
@@ -441,19 +439,9 @@ def norm_function(arr): return torch.norm(arr)
                 att_preturb = np.random.normal(size=[64, 24, 53])
 
             else:
+                att_preturb, features_attrs, timestep_attrs = model.explantation2(x, explain_method)
 
-                data, baselines = model.prep_data_captum(x)
-
-                explantation = explain_method(model.forward_captum)
-                # Reformat attributions.
-                if explain_method is not captum.attr.Saliency:
-                    att_preturb = explantation.attribute(data, baselines=baselines, **kwargs)
-                else:
-                    att_preturb = explantation.attribute(data, **kwargs)
-
-                # Process and store the calculated attributions
-                att_preturb = att_preturb[1].detach() if method_name in [
-                    'Lime', 'FeatureAblation'] else torch.stack(att_preturb).detach()
+                #
             # Calculate the absolute difference
             difference = torch.abs(y_pred_preturb - y_pred)
 
@@ -484,7 +472,7 @@ def __init__(
 
     def update(
         self, x,
-        attribution, model, explain_method, random_model, similarity_func=cosine, dataloader=None, method_name="", **kwargs
+        attribution, model, explain_method, random_model, similarity_func=cosine, dataloader=None, **kwargs
     ):
         """
 
@@ -531,16 +519,7 @@ def update(
             data, baselines = model.prep_data_captum(x)
             y_pred = model(data).detach()
 
-            explantation = explain_method(random_model.forward_captum)
-            # Reformat attributions.
-            if explain_method is not captum.attr.Saliency:
-                attr = explantation.attribute(data, baselines=baselines, **kwargs)
-            else:
-                attr = explantation.attribute(data, **kwargs)
-
-            # Process and store the calculated attributions
-            random_attr = attr[1].cpu().detach().numpy() if method_name in [
-                'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
+            random_attr, features_attrs, timestep_attrs = model.explantation2(x, explain_method)
 
             attribution = attribution.flatten()
             min_val = np.min(attribution)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index b0cd933a..92a364a2 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -518,15 +518,16 @@ def step_fn(self, element, step_prefix=""):
         if self.explain is not None:
 
             for method in self.explain:
+                print(method)
                 if method == "Attention":
-                    all_attrs, features_attrs, timestep_attrs = self.explantation2(dic, method)
+                    all_attrs, features_attrs, timestep_attrs = self.explantation2(dic, method, self.test_loader)
                 else:
                     all_attrs, features_attrs, timestep_attrs = self.explantation2(dic, method)
                 for key, value in self.metrics[step_prefix].items():
                     if key == "Faithfulness_timesteps":
                         value.update(
                             dic,
-                            all_attrs,
+                            timestep_attrs,
                             self,
                             correlation_pearson,
                             100,
@@ -540,7 +541,7 @@ def step_fn(self, element, step_prefix=""):
                     elif key == "Faithfulness_features":
                         value.update(
                             dic,
-                            all_attrs,
+                            features_attrs,
                             self,
                             correlation_pearson,
                             100,
@@ -573,7 +574,6 @@ def step_fn(self, element, step_prefix=""):
                                 dic,
                                 timestep_attrs,
                                 self,
-                                self.methods[method],
                                 method,
                                 self.test_loader,
                                 0.5,
@@ -584,7 +584,6 @@ def step_fn(self, element, step_prefix=""):
                                 dic,
                                 all_attrs,
                                 self,
-                                self.methods[method],
                                 method,
                                 None,
                                 0.5,
@@ -596,11 +595,11 @@ def step_fn(self, element, step_prefix=""):
                                 dic,
                                 timestep_attrs,
                                 self,
-                                self.methods[method],
+                                method,
                                 self.random_model,
                                 cosine,
                                 None,
-                                method
+
                             )
                         else:
 
@@ -608,15 +607,17 @@ def step_fn(self, element, step_prefix=""):
                                 dic,
                                 all_attrs,
                                 self,
-                                self.methods[method],
+                                method,
                                 self.random_model,
                                 cosine,
                                 self.test_loader,
-                                method
+
                             )
 
         for key, value in self.metrics[step_prefix].items():
-
+            if (key == "Faithfulness_timesteps" or key == "Faithfulness_features" or key == "Faithfulness_feature_timestep"
+                    or key == "Stability" or key == "Randomization"):
+                continue
             if isinstance(value, torchmetrics.Metric):
                 if key == "Binary_Fairness":
                     feature_names = self.metrics[step_prefix][key].feature_helper(self.trainer, step_prefix)
@@ -626,9 +627,6 @@ def step_fn(self, element, step_prefix=""):
                         data,
                         feature_names,
                     )
-                elif (key == "Faithfulness_timesteps" or key == "Faithfulness_features" or key == "Faithfulness_feature_timestep"
-                      or key == "Stability" or key == "Randomization"):
-                    continue
 
                 else:
                     value.update(transformed_output[0], transformed_output[1].int())
@@ -780,23 +778,59 @@ def explantation2(self, dic, method, dataloader=None, log_dir=".", plot=False, *
 
         # Calculate attributions using the selected method
         if method is not Saliency:
-            attr = explanation.attribute(data, baselines=baselines, **kwargs)
+            if method_name == "IG":
+                attr = explanation.attribute(data, baselines=baselines, n_steps=50, **kwargs)
+            elif method_name == "L" or method_name == "FA":
+                # for Lime and feature ablation we need to define what is a feature we define each variable per timestep as a feature
+                shapes = [
+                    torch.Size([64, 24, 0]),
+                    torch.Size([64, 24, 53]),
+                    torch.Size([64, 24]),
+                    torch.Size([64]),
+                    torch.Size([64, 1, 0]),
+                    torch.Size([64, 1, 53]),
+                    torch.Size([64, 1]),
+                    torch.Size([64]),
+                    torch.Size([64, 1]),
+                    torch.Size([64, 1]),
+                    torch.Size([64, 2])
+                ]
+
+                # Create a default mask for non-targeted tensors
+                def create_default_mask(shape):
+                    if len(shape) == 3:
+                        return torch.zeros(shape[0], shape[1], max(1, shape[2]), dtype=torch.int32)
+                    elif len(shape) == 2:
+                        return torch.zeros(shape[0], max(1, shape[1]), dtype=torch.int32)
+                    else:  # len(shape) == 1
+                        return torch.zeros(shape[0], dtype=torch.int32)
+
+                # Create a feature mask for the second tensor that includes both features and timesteps
+                num_timesteps = shapes[1][1]
+                num_features = shapes[1][2]
+                feature_mask_second = torch.arange(num_timesteps * num_features).reshape(num_timesteps, num_features)
+                feature_mask_second = feature_mask_second.unsqueeze(0).repeat(shapes[1][0], 1, 1)
+
+                # Create a tuple of masks
+                feature_masks = tuple([create_default_mask(shape) if i !=
+                                      1 else feature_mask_second for i, shape in enumerate(shapes)])
+
+                attr = explanation.attribute(data, baselines=baselines, feature_mask=feature_masks, **kwargs)
         else:
             attr = explanation.attribute(data, **kwargs)
         self.eval()
         # Process and store the calculated attributions
         stacked_attr = attr[1].cpu().detach().numpy() if method_name in [
             'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
-
         # aggregate over batch
-        attr = np.mean(stacked_attr, axis=0)
-        all_attrs.append(attr)
+        # attr = np.mean(stacked_attr, axis=0)
+
         # aggregate over all batches
-        all_attrs = np.array(all_attrs).mean(axis=(0))
+        # all_attrs = np.array(all_attrs).mean(axis=(0))
         # aggregate over all timesteps
-        features_attrs = all_attrs.mean(axis=(0))
+        features_attrs = stacked_attr.mean(axis=(1))
         # aggregate over all features
-        timestep_attrs = all_attrs.mean(axis=(1))
+        timestep_attrs = stacked_attr.mean(axis=(2))
 
         if plot:
             log_dir_plots = log_dir / 'plots'
@@ -807,7 +841,7 @@ def explantation2(self, dic, method, dataloader=None, log_dir=".", plot=False, *
             self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots)
 
         # Return computed attributions and metrics
-        return all_attrs, features_attrs, timestep_attrs
+        return stacked_attr, features_attrs, timestep_attrs
 
     def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=False, random_model=None, test_dataset=None, **kwargs):
         """

From 1586fbb583d6661ee771f5080f9a2681d4b579a9 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 28 Dec 2023 15:53:25 +0100
Subject: [PATCH 125/142] :Revert "added condition for ROS and RIS for
 attention"

This reverts commit a39b5c418b62fe163bbd621502f0cbabb6fdc191.
---
 icu_benchmarks/models/train.py    |  4 ++++
 icu_benchmarks/models/wrappers.py | 34 +++++++++++++++----------------
 2 files changed, 20 insertions(+), 18 deletions(-)

diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index e1566713..cf94843c 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -290,7 +290,11 @@ def train_common(
             "IG": IntegratedGradients,
             # "FA": FeatureAblation,
             "R": "Random",
+<<<<<<< HEAD
             "Att": "Attention"
+=======
+            # "Att": "Attention"
+>>>>>>> parent of a39b5c4... added condition for ROS and RIS for attention
         }
         for key, item in methods.items():
             # If conditions needed here as different explantations require different inputs
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 92a364a2..0f86e6cc 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -962,7 +962,10 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                                                                pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
                 r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
                                explain_method=method, random_model=random_model, method_name=method_name))
+<<<<<<< HEAD
 
+=======
+>>>>>>> parent of a39b5c4... added condition for ROS and RIS for attention
                 res1, res2 = self.Relative_Stability(x,
                                                      stacked_attr, explain_method=method, method_name=method_name,  dataloader=None, **kwargs
                                                      )
@@ -1259,7 +1262,7 @@ def Data_Randomization(
         return score
 
     def Relative_Stability(self, x,
-                           attribution, explain_method, method_name, dataloader=None, thershold=0.5, device='cuda', **kwargs
+                           attribution, explain_method, method_name, dataloader=None, thershold=0.5, **kwargs
                            ):
         """
      Args:
@@ -1282,7 +1285,7 @@ def Relative_Stability(self, x,
         """
 
         def relative_stability_objective(
-            x, xs, e_x, e_xs, close_indices, eps_min=0.0001, input=False, attention=False, device='cuda'
+            x, xs, e_x, e_xs, eps_min=0.0001, input=False, device='cuda'
         ) -> torch.Tensor:
             """
             Computes relative input and output stabilities maximization objective
@@ -1306,15 +1309,11 @@ def relative_stability_objective(
             # Function to convert inputs to tensors if they are numpy arrays
             def to_tensor(input_array):
                 if isinstance(input_array, np.ndarray):
-                    return torch.index_select(torch.tensor(input_array).to(device), 0, close_indices)
-
-                return torch.index_select(input_array.to(device), 0, close_indices)
+                    return torch.tensor(input_array).to(device)
+                return input_array.to(device)
 
             # Convert all inputs to tensors and move to GPU
-            if attention:
-                x, xs = map(to_tensor, [x, xs])
-            else:
-                x, xs, e_x, e_xs = map(to_tensor, [x, xs, e_x, e_xs])
+            x, xs, e_x, e_xs = map(to_tensor, [x, xs, e_x, e_xs])
 
             if input:
                 num_dim = x.ndim
@@ -1356,13 +1355,12 @@ def norm_function(arr): return torch.norm(arr)
             difference = torch.abs(y_pred_preturb - y_pred)
 
             # Find where the difference is less than or equal to a thershold
-            close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
-
+            close_indices = torch.nonzero(difference <= thershold).squeeze()
             RIS = relative_stability_objective(
-                x_original.detach(), x_preturb.detach(), attribution, att_preturb, close_indices=close_indices, input=True, attention=True
+                x_original[close_indices, :, :].detach(), x_preturb[close_indices, :, :].detach(), attribution, att_preturb, input=True
             )
             ROS = relative_stability_objective(
-                y_pred, y_pred_preturb, attribution, att_preturb, close_indices=close_indices, input=False, attention=True
+                y_pred[close_indices], y_pred_preturb[close_indices], attribution, att_preturb, input=False
             )
 
         else:
@@ -1394,15 +1392,15 @@ def norm_function(arr): return torch.norm(arr)
             difference = torch.abs(y_pred_preturb - y_pred)
 
             # Find where the difference is less than or equal to a thershold
-            close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
+            close_indices = torch.nonzero(difference <= thershold).squeeze()
+            print(close_indices)
 
             RIS = relative_stability_objective(
-                x_original.detach(),
-                x["encoder_cont"].detach(),
-                attribution, att_preturb, close_indices=close_indices, input=True
+                x_original[close_indices, :, :].detach(), x["encoder_cont"][close_indices, :, :].detach(), attribution[close_indices, :, :], att_preturb[close_indices, :, :], input=True
             )
             ROS = relative_stability_objective(
-                y_pred, y_pred_preturb, attribution, att_preturb, close_indices=close_indices, input=False
+                y_pred[close_indices], y_pred_preturb[close_indices], attribution[close_indices,
+                                                                                  :, :], att_preturb[close_indices, :, :], input=False
             )
 
         return np.max(RIS.cpu().numpy()).astype(np.float64), np.max(ROS.cpu().numpy()).astype(np.float64)

From 8b39c6ebbd3d3a606ddf6dd422d37d60cb41e995 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 28 Dec 2023 15:59:26 +0100
Subject: [PATCH 126/142] reverted changes

---
 configs/prediction_models/DeepARpytorch.gin |  68 +----
 configs/prediction_models/RNNpytorch.gin    |  81 +-----
 configs/prediction_models/TFTpytorch.gin    |   2 -
 icu_benchmarks/cross_validation.py          |   4 +-
 icu_benchmarks/data/loader.py               |  12 +-
 icu_benchmarks/models/constants.py          |  14 -
 icu_benchmarks/models/custom_metrics.py     |   3 +-
 icu_benchmarks/models/dl_models.py          |  22 +-
 icu_benchmarks/models/train.py              |  51 ++--
 icu_benchmarks/models/wrappers.py           | 285 +-------------------
 icu_benchmarks/run_utils.py                 |  20 +-
 11 files changed, 67 insertions(+), 495 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 04d8011a..bcc6abf6 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -29,63 +29,16 @@ train_common/hyperparameter.gradient_clip_val=0.01
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
+PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_1', 'alp_lag_1', 'alt_lag_1', 'ast_lag_1', 'be_lag_1', 'bicar_lag_1', 'bili_lag_1', 'bili_dir_lag_1', 'bnd_lag_1', 'bun_lag_1', 'ca_lag_1', 'cai_lag_1', 'ck_lag_1', 'ckmb_lag_1', 'cl_lag_1', 'crea_lag_1', 'crp_lag_1', 'dbp_lag_1', 'fgn_lag_1', 'fio2_lag_1', 'glu_lag_1', 'hgb_lag_1', 'hr_lag_1', 'inr_pt_lag_1', 'k_lag_1', 'lact_lag_1', 'lymph_lag_1', 'map_lag_1', 'mch_lag_1', 'mchc_lag_1', 'mcv_lag_1', 'methb_lag_1', 'mg_lag_1', 'na_lag_1', 'neut_lag_1', 'o2sat_lag_1', 'pco2_lag_1', 'ph_lag_1', 'phos_lag_1', 'plt_lag_1', 'po2_lag_1', 'ptt_lag_1', 'resp_lag_1', 'sbp_lag_1', 'temp_lag_1', 'tnt_lag_1', 'urine_lag_1', 'wbc_lag_1', 'alb_lag_2', 'alp_lag_2', 'alt_lag_2', 'ast_lag_2', 'be_lag_2', 'bicar_lag_2', 'bili_lag_2', 'bili_dir_lag_2', 'bnd_lag_2', 'bun_lag_2', 'ca_lag_2', 'cai_lag_2', 'ck_lag_2', 'ckmb_lag_2', 'cl_lag_2', 'crea_lag_2', 'crp_lag_2', 'dbp_lag_2', 'fgn_lag_2', 'fio2_lag_2', 'glu_lag_2', 'hgb_lag_2', 'hr_lag_2', 'inr_pt_lag_2', 'k_lag_2', 'lact_lag_2', 'lymph_lag_2', 'map_lag_2', 'mch_lag_2', 'mchc_lag_2', 'mcv_lag_2', 'methb_lag_2', 'mg_lag_2', 'na_lag_2', 'neut_lag_2', 'o2sat_lag_2', 'pco2_lag_2', 'ph_lag_2', 'phos_lag_2', 'plt_lag_2', 'po2_lag_2', 'ptt_lag_2', 'resp_lag_2', 'sbp_lag_2', 'temp_lag_2', 'tnt_lag_2', 'urine_lag_2', 'wbc_lag_2', 'alb_lag_3', 'alp_lag_3', 'alt_lag_3', 'ast_lag_3', 'be_lag_3', 'bicar_lag_3', 'bili_lag_3', 'bili_dir_lag_3', 'bnd_lag_3', 'bun_lag_3', 'ca_lag_3', 'cai_lag_3', 'ck_lag_3', 'ckmb_lag_3', 'cl_lag_3', 'crea_lag_3', 'crp_lag_3', 'dbp_lag_3', 'fgn_lag_3', 'fio2_lag_3', 'glu_lag_3', 'hgb_lag_3', 'hr_lag_3', 'inr_pt_lag_3', 'k_lag_3', 'lact_lag_3', 'lymph_lag_3', 'map_lag_3', 'mch_lag_3', 'mchc_lag_3', 'mcv_lag_3', 'methb_lag_3', 'mg_lag_3', 'na_lag_3', 'neut_lag_3', 'o2sat_lag_3', 'pco2_lag_3', 'ph_lag_3', 'phos_lag_3', 'plt_lag_3', 'po2_lag_3', 'ptt_lag_3', 'resp_lag_3', 'sbp_lag_3', 'temp_lag_3', 'tnt_lag_3', 'urine_lag_3', 'wbc_lag_3', 'alb_lag_4', 'alp_lag_4', 'alt_lag_4', 'ast_lag_4', 'be_lag_4', 'bicar_lag_4', 'bili_lag_4', 'bili_dir_lag_4', 'bnd_lag_4', 'bun_lag_4', 'ca_lag_4', 'cai_lag_4', 'ck_lag_4', 'ckmb_lag_4', 'cl_lag_4', 'crea_lag_4', 'crp_lag_4', 'dbp_lag_4', 'fgn_lag_4', 'fio2_lag_4', 'glu_lag_4', 'hgb_lag_4', 'hr_lag_4', 'inr_pt_lag_4', 'k_lag_4', 'lact_lag_4', 'lymph_lag_4', 'map_lag_4', 'mch_lag_4', 'mchc_lag_4', 'mcv_lag_4', 'methb_lag_4', 'mg_lag_4', 'na_lag_4', 'neut_lag_4', 'o2sat_lag_4', 'pco2_lag_4', 'ph_lag_4', 'phos_lag_4', 'plt_lag_4', 'po2_lag_4', 'ptt_lag_4', 'resp_lag_4', 'sbp_lag_4', 'temp_lag_4', 'tnt_lag_4', 'urine_lag_4', 'wbc_lag_4', 'alb_lag_5', 'alp_lag_5', 'alt_lag_5', 'ast_lag_5', 'be_lag_5', 'bicar_lag_5', 'bili_lag_5', 'bili_dir_lag_5', 'bnd_lag_5', 'bun_lag_5', 'ca_lag_5', 'cai_lag_5', 'ck_lag_5', 'ckmb_lag_5', 'cl_lag_5', 'crea_lag_5', 'crp_lag_5', 'dbp_lag_5', 'fgn_lag_5', 'fio2_lag_5', 'glu_lag_5', 'hgb_lag_5', 'hr_lag_5', 'inr_pt_lag_5', 'k_lag_5', 'lact_lag_5', 'lymph_lag_5', 'map_lag_5', 'mch_lag_5', 'mchc_lag_5', 'mcv_lag_5', 'methb_lag_5', 'mg_lag_5', 'na_lag_5', 'neut_lag_5', 'o2sat_lag_5', 'pco2_lag_5', 'ph_lag_5', 'phos_lag_5', 'plt_lag_5', 'po2_lag_5', 'ptt_lag_5', 'resp_lag_5', 'sbp_lag_5', 'temp_lag_5', 'tnt_lag_5', 'urine_lag_5', 'wbc_lag_5', 'alb_lag_6', 'alp_lag_6', 'alt_lag_6', 'ast_lag_6', 'be_lag_6', 'bicar_lag_6', 'bili_lag_6', 'bili_dir_lag_6', 'bnd_lag_6', 'bun_lag_6', 'ca_lag_6', 'cai_lag_6', 'ck_lag_6', 'ckmb_lag_6', 'cl_lag_6', 'crea_lag_6', 'crp_lag_6', 'dbp_lag_6', 'fgn_lag_6', 'fio2_lag_6', 'glu_lag_6', 'hgb_lag_6', 'hr_lag_6', 'inr_pt_lag_6', 'k_lag_6', 'lact_lag_6', 'lymph_lag_6', 'map_lag_6', 'mch_lag_6', 'mchc_lag_6', 'mcv_lag_6', 'methb_lag_6', 'mg_lag_6', 'na_lag_6', 'neut_lag_6', 'o2sat_lag_6', 'pco2_lag_6', 'ph_lag_6', 'phos_lag_6', 'plt_lag_6', 'po2_lag_6', 'ptt_lag_6', 'resp_lag_6', 'sbp_lag_6', 'temp_lag_6', 'tnt_lag_6', 'urine_lag_6', 'wbc_lag_6', 'alb_lag_7', 'alp_lag_7', 'alt_lag_7', 'ast_lag_7', 'be_lag_7', 'bicar_lag_7', 'bili_lag_7', 'bili_dir_lag_7', 'bnd_lag_7', 'bun_lag_7', 'ca_lag_7', 'cai_lag_7', 'ck_lag_7', 'ckmb_lag_7', 'cl_lag_7', 'crea_lag_7', 'crp_lag_7', 'dbp_lag_7', 'fgn_lag_7', 'fio2_lag_7', 'glu_lag_7', 'hgb_lag_7', 'hr_lag_7', 'inr_pt_lag_7', 'k_lag_7', 'lact_lag_7', 'lymph_lag_7', 'map_lag_7', 'mch_lag_7', 'mchc_lag_7', 'mcv_lag_7', 'methb_lag_7', 'mg_lag_7', 'na_lag_7', 'neut_lag_7', 'o2sat_lag_7', 'pco2_lag_7', 'ph_lag_7', 'phos_lag_7', 'plt_lag_7', 'po2_lag_7', 'ptt_lag_7', 'resp_lag_7', 'sbp_lag_7', 'temp_lag_7', 'tnt_lag_7', 'urine_lag_7', 'wbc_lag_7', 'alb_lag_8', 'alp_lag_8', 'alt_lag_8', 'ast_lag_8', 'be_lag_8', 'bicar_lag_8', 'bili_lag_8', 'bili_dir_lag_8', 'bnd_lag_8', 'bun_lag_8', 'ca_lag_8', 'cai_lag_8', 'ck_lag_8', 'ckmb_lag_8', 'cl_lag_8', 'crea_lag_8', 'crp_lag_8', 'dbp_lag_8', 'fgn_lag_8', 'fio2_lag_8', 'glu_lag_8', 'hgb_lag_8', 'hr_lag_8', 'inr_pt_lag_8', 'k_lag_8', 'lact_lag_8', 'lymph_lag_8', 'map_lag_8', 'mch_lag_8', 'mchc_lag_8', 'mcv_lag_8', 'methb_lag_8', 'mg_lag_8', 'na_lag_8', 'neut_lag_8', 'o2sat_lag_8', 'pco2_lag_8', 'ph_lag_8', 'phos_lag_8', 'plt_lag_8', 'po2_lag_8', 'ptt_lag_8', 'resp_lag_8', 'sbp_lag_8', 'temp_lag_8', 'tnt_lag_8', 'urine_lag_8', 'wbc_lag_8', 'alb_lag_9', 'alp_lag_9', 'alt_lag_9', 'ast_lag_9', 'be_lag_9', 'bicar_lag_9', 'bili_lag_9', 'bili_dir_lag_9', 'bnd_lag_9', 'bun_lag_9', 'ca_lag_9', 'cai_lag_9', 'ck_lag_9', 'ckmb_lag_9', 'cl_lag_9', 'crea_lag_9', 'crp_lag_9', 'dbp_lag_9', 'fgn_lag_9', 'fio2_lag_9', 'glu_lag_9', 'hgb_lag_9', 'hr_lag_9', 'inr_pt_lag_9', 'k_lag_9', 'lact_lag_9', 'lymph_lag_9', 'map_lag_9', 'mch_lag_9', 'mchc_lag_9', 'mcv_lag_9', 'methb_lag_9', 'mg_lag_9', 'na_lag_9', 'neut_lag_9', 'o2sat_lag_9', 'pco2_lag_9', 'ph_lag_9', 'phos_lag_9', 'plt_lag_9', 'po2_lag_9', 'ptt_lag_9', 'resp_lag_9', 'sbp_lag_9', 'temp_lag_9', 'tnt_lag_9', 'urine_lag_9', 'wbc_lag_9', 'alb_lag_10', 'alp_lag_10', 'alt_lag_10', 'ast_lag_10', 'be_lag_10', 'bicar_lag_10', 'bili_lag_10', 'bili_dir_lag_10', 'bnd_lag_10', 'bun_lag_10', 'ca_lag_10', 'cai_lag_10', 'ck_lag_10', 'ckmb_lag_10', 'cl_lag_10', 'crea_lag_10', 'crp_lag_10', 'dbp_lag_10', 'fgn_lag_10', 'fio2_lag_10', 'glu_lag_10', 'hgb_lag_10', 'hr_lag_10', 'inr_pt_lag_10', 'k_lag_10', 'lact_lag_10', 'lymph_lag_10', 'map_lag_10', 'mch_lag_10', 'mchc_lag_10', 'mcv_lag_10', 'methb_lag_10', 'mg_lag_10', 'na_lag_10', 'neut_lag_10', 'o2sat_lag_10', 'pco2_lag_10', 'ph_lag_10', 'phos_lag_10', 'plt_lag_10', 'po2_lag_10', 'ptt_lag_10', 'resp_lag_10', 'sbp_lag_10', 'temp_lag_10', 'tnt_lag_10', 'urine_lag_10', 'wbc_lag_10', 'alb_lag_11', 'alp_lag_11', 'alt_lag_11', 'ast_lag_11', 'be_lag_11', 'bicar_lag_11', 'bili_lag_11', 'bili_dir_lag_11', 'bnd_lag_11', 'bun_lag_11', 'ca_lag_11', 'cai_lag_11', 'ck_lag_11', 'ckmb_lag_11', 'cl_lag_11', 'crea_lag_11', 'crp_lag_11', 'dbp_lag_11', 'fgn_lag_11', 'fio2_lag_11', 'glu_lag_11', 'hgb_lag_11', 'hr_lag_11', 'inr_pt_lag_11', 'k_lag_11', 'lact_lag_11', 'lymph_lag_11', 'map_lag_11', 'mch_lag_11', 'mchc_lag_11', 'mcv_lag_11', 'methb_lag_11', 'mg_lag_11', 'na_lag_11', 'neut_lag_11', 'o2sat_lag_11', 'pco2_lag_11', 'ph_lag_11', 'phos_lag_11', 'plt_lag_11', 'po2_lag_11', 'ptt_lag_11', 'resp_lag_11', 'sbp_lag_11', 'temp_lag_11', 'tnt_lag_11', 'urine_lag_11', 'wbc_lag_11', 'alb_lag_12', 'alp_lag_12', 'alt_lag_12', 'ast_lag_12', 'be_lag_12', 'bicar_lag_12', 'bili_lag_12', 'bili_dir_lag_12', 'bnd_lag_12', 'bun_lag_12', 'ca_lag_12', 'cai_lag_12', 'ck_lag_12', 'ckmb_lag_12', 'cl_lag_12', 'crea_lag_12', 'crp_lag_12', 'dbp_lag_12', 'fgn_lag_12', 'fio2_lag_12', 'glu_lag_12', 'hgb_lag_12', 'hr_lag_12', 'inr_pt_lag_12', 'k_lag_12', 'lact_lag_12', 'lymph_lag_12', 'map_lag_12', 'mch_lag_12', 'mchc_lag_12', 'mcv_lag_12', 'methb_lag_12', 'mg_lag_12', 'na_lag_12', 'neut_lag_12', 'o2sat_lag_12', 'pco2_lag_12', 'ph_lag_12', 'phos_lag_12', 'plt_lag_12', 'po2_lag_12', 'ptt_lag_12', 'resp_lag_12', 'sbp_lag_12', 'temp_lag_12', 'tnt_lag_12', 'urine_lag_12', 'wbc_lag_12', 'alb_lag_13', 'alp_lag_13', 'alt_lag_13', 'ast_lag_13', 'be_lag_13', 'bicar_lag_13', 'bili_lag_13', 'bili_dir_lag_13', 'bnd_lag_13', 'bun_lag_13', 'ca_lag_13', 'cai_lag_13', 'ck_lag_13', 'ckmb_lag_13', 'cl_lag_13', 'crea_lag_13', 'crp_lag_13', 'dbp_lag_13', 'fgn_lag_13', 'fio2_lag_13', 'glu_lag_13', 'hgb_lag_13', 'hr_lag_13', 'inr_pt_lag_13', 'k_lag_13', 'lact_lag_13', 'lymph_lag_13', 'map_lag_13', 'mch_lag_13', 'mchc_lag_13', 'mcv_lag_13', 'methb_lag_13', 'mg_lag_13', 'na_lag_13', 'neut_lag_13', 'o2sat_lag_13', 'pco2_lag_13', 'ph_lag_13', 'phos_lag_13', 'plt_lag_13', 'po2_lag_13', 'ptt_lag_13', 'resp_lag_13', 'sbp_lag_13', 'temp_lag_13', 'tnt_lag_13', 'urine_lag_13', 'wbc_lag_13', 'alb_lag_14', 'alp_lag_14', 'alt_lag_14', 'ast_lag_14', 'be_lag_14', 'bicar_lag_14', 'bili_lag_14', 'bili_dir_lag_14', 'bnd_lag_14', 'bun_lag_14', 'ca_lag_14', 'cai_lag_14', 'ck_lag_14', 'ckmb_lag_14', 'cl_lag_14', 'crea_lag_14', 'crp_lag_14', 'dbp_lag_14', 'fgn_lag_14', 'fio2_lag_14', 'glu_lag_14', 'hgb_lag_14', 'hr_lag_14', 'inr_pt_lag_14', 'k_lag_14', 'lact_lag_14', 'lymph_lag_14', 'map_lag_14', 'mch_lag_14', 'mchc_lag_14', 'mcv_lag_14', 'methb_lag_14', 'mg_lag_14', 'na_lag_14', 'neut_lag_14', 'o2sat_lag_14', 'pco2_lag_14', 'ph_lag_14', 'phos_lag_14', 'plt_lag_14', 'po2_lag_14', 'ptt_lag_14', 'resp_lag_14', 'sbp_lag_14', 'temp_lag_14', 'tnt_lag_14', 'urine_lag_14', 'wbc_lag_14', 'alb_lag_15', 'alp_lag_15', 'alt_lag_15', 'ast_lag_15', 'be_lag_15', 'bicar_lag_15', 'bili_lag_15', 'bili_dir_lag_15', 'bnd_lag_15', 'bun_lag_15', 'ca_lag_15', 'cai_lag_15', 'ck_lag_15', 'ckmb_lag_15', 'cl_lag_15', 'crea_lag_15', 'crp_lag_15', 'dbp_lag_15', 'fgn_lag_15', 'fio2_lag_15', 'glu_lag_15', 'hgb_lag_15', 'hr_lag_15', 'inr_pt_lag_15', 'k_lag_15', 'lact_lag_15', 'lymph_lag_15', 'map_lag_15', 'mch_lag_15', 'mchc_lag_15', 'mcv_lag_15', 'methb_lag_15', 'mg_lag_15', 'na_lag_15', 'neut_lag_15', 'o2sat_lag_15', 'pco2_lag_15', 'ph_lag_15', 'phos_lag_15', 'plt_lag_15', 'po2_lag_15', 'ptt_lag_15', 'resp_lag_15', 'sbp_lag_15', 'temp_lag_15', 'tnt_lag_15', 'urine_lag_15', 'wbc_lag_15', 'alb_lag_16', 'alp_lag_16', 'alt_lag_16', 'ast_lag_16', 'be_lag_16', 'bicar_lag_16', 'bili_lag_16', 'bili_dir_lag_16', 'bnd_lag_16', 'bun_lag_16', 'ca_lag_16', 'cai_lag_16', 'ck_lag_16', 'ckmb_lag_16', 'cl_lag_16', 'crea_lag_16', 'crp_lag_16', 'dbp_lag_16', 'fgn_lag_16', 'fio2_lag_16', 'glu_lag_16', 'hgb_lag_16', 'hr_lag_16', 'inr_pt_lag_16', 'k_lag_16', 'lact_lag_16', 'lymph_lag_16', 'map_lag_16', 'mch_lag_16', 'mchc_lag_16', 'mcv_lag_16', 'methb_lag_16', 'mg_lag_16', 'na_lag_16', 'neut_lag_16', 'o2sat_lag_16', 'pco2_lag_16', 'ph_lag_16', 'phos_lag_16', 'plt_lag_16', 'po2_lag_16', 'ptt_lag_16', 'resp_lag_16', 'sbp_lag_16', 'temp_lag_16', 'tnt_lag_16', 'urine_lag_16', 'wbc_lag_16', 'alb_lag_17', 'alp_lag_17', 'alt_lag_17', 'ast_lag_17', 'be_lag_17', 'bicar_lag_17', 'bili_lag_17', 'bili_dir_lag_17', 'bnd_lag_17', 'bun_lag_17', 'ca_lag_17', 'cai_lag_17', 'ck_lag_17', 'ckmb_lag_17', 'cl_lag_17', 'crea_lag_17', 'crp_lag_17', 'dbp_lag_17', 'fgn_lag_17', 'fio2_lag_17', 'glu_lag_17', 'hgb_lag_17', 'hr_lag_17', 'inr_pt_lag_17', 'k_lag_17', 'lact_lag_17', 'lymph_lag_17', 'map_lag_17', 'mch_lag_17', 'mchc_lag_17', 'mcv_lag_17', 'methb_lag_17', 'mg_lag_17', 'na_lag_17', 'neut_lag_17', 'o2sat_lag_17', 'pco2_lag_17', 'ph_lag_17', 'phos_lag_17', 'plt_lag_17', 'po2_lag_17', 'ptt_lag_17', 'resp_lag_17', 'sbp_lag_17', 'temp_lag_17', 'tnt_lag_17', 'urine_lag_17', 'wbc_lag_17', 'alb_lag_18', 'alp_lag_18', 'alt_lag_18', 'ast_lag_18', 'be_lag_18', 'bicar_lag_18', 'bili_lag_18', 'bili_dir_lag_18', 'bnd_lag_18', 'bun_lag_18', 'ca_lag_18', 'cai_lag_18', 'ck_lag_18', 'ckmb_lag_18', 'cl_lag_18', 'crea_lag_18', 'crp_lag_18', 'dbp_lag_18', 'fgn_lag_18', 'fio2_lag_18', 'glu_lag_18', 'hgb_lag_18', 'hr_lag_18', 'inr_pt_lag_18', 'k_lag_18', 'lact_lag_18', 'lymph_lag_18', 'map_lag_18', 'mch_lag_18', 'mchc_lag_18', 'mcv_lag_18', 'methb_lag_18', 'mg_lag_18', 'na_lag_18', 'neut_lag_18', 'o2sat_lag_18', 'pco2_lag_18', 'ph_lag_18', 'phos_lag_18', 'plt_lag_18', 'po2_lag_18', 'ptt_lag_18', 'resp_lag_18', 'sbp_lag_18', 'temp_lag_18', 'tnt_lag_18', 'urine_lag_18', 'wbc_lag_18', 'alb_lag_19', 'alp_lag_19', 'alt_lag_19', 'ast_lag_19', 'be_lag_19', 'bicar_lag_19', 'bili_lag_19', 'bili_dir_lag_19', 'bnd_lag_19', 'bun_lag_19', 'ca_lag_19', 'cai_lag_19', 'ck_lag_19', 'ckmb_lag_19', 'cl_lag_19', 'crea_lag_19', 'crp_lag_19', 'dbp_lag_19', 'fgn_lag_19', 'fio2_lag_19', 'glu_lag_19', 'hgb_lag_19', 'hr_lag_19', 'inr_pt_lag_19', 'k_lag_19', 'lact_lag_19', 'lymph_lag_19', 'map_lag_19', 'mch_lag_19', 'mchc_lag_19', 'mcv_lag_19', 'methb_lag_19', 'mg_lag_19', 'na_lag_19', 'neut_lag_19', 'o2sat_lag_19', 'pco2_lag_19', 'ph_lag_19', 'phos_lag_19', 'plt_lag_19', 'po2_lag_19', 'ptt_lag_19', 'resp_lag_19', 'sbp_lag_19', 'temp_lag_19', 'tnt_lag_19', 'urine_lag_19', 'wbc_lag_19', 'alb_lag_20', 'alp_lag_20', 'alt_lag_20', 'ast_lag_20', 'be_lag_20', 'bicar_lag_20', 'bili_lag_20', 'bili_dir_lag_20', 'bnd_lag_20', 'bun_lag_20', 'ca_lag_20', 'cai_lag_20', 'ck_lag_20', 'ckmb_lag_20', 'cl_lag_20', 'crea_lag_20', 'crp_lag_20', 'dbp_lag_20', 'fgn_lag_20', 'fio2_lag_20', 'glu_lag_20', 'hgb_lag_20', 'hr_lag_20', 'inr_pt_lag_20', 'k_lag_20', 'lact_lag_20', 'lymph_lag_20', 'map_lag_20', 'mch_lag_20', 'mchc_lag_20', 'mcv_lag_20', 'methb_lag_20', 'mg_lag_20', 'na_lag_20', 'neut_lag_20', 'o2sat_lag_20', 'pco2_lag_20', 'ph_lag_20', 'phos_lag_20', 'plt_lag_20', 'po2_lag_20', 'ptt_lag_20', 'resp_lag_20', 'sbp_lag_20', 'temp_lag_20', 'tnt_lag_20', 'urine_lag_20', 'wbc_lag_20', 'alb_lag_21', 'alp_lag_21', 'alt_lag_21', 'ast_lag_21', 'be_lag_21', 'bicar_lag_21', 'bili_lag_21', 'bili_dir_lag_21', 'bnd_lag_21', 'bun_lag_21', 'ca_lag_21', 'cai_lag_21', 'ck_lag_21', 'ckmb_lag_21', 'cl_lag_21', 'crea_lag_21', 'crp_lag_21', 'dbp_lag_21', 'fgn_lag_21', 'fio2_lag_21', 'glu_lag_21', 'hgb_lag_21', 'hr_lag_21', 'inr_pt_lag_21', 'k_lag_21', 'lact_lag_21', 'lymph_lag_21', 'map_lag_21', 'mch_lag_21', 'mchc_lag_21', 'mcv_lag_21', 'methb_lag_21', 'mg_lag_21', 'na_lag_21', 'neut_lag_21', 'o2sat_lag_21', 'pco2_lag_21', 'ph_lag_21', 'phos_lag_21', 'plt_lag_21', 'po2_lag_21', 'ptt_lag_21', 'resp_lag_21', 'sbp_lag_21', 'temp_lag_21', 'tnt_lag_21', 'urine_lag_21', 'wbc_lag_21', 'alb_lag_22', 'alp_lag_22', 'alt_lag_22', 'ast_lag_22', 'be_lag_22', 'bicar_lag_22', 'bili_lag_22', 'bili_dir_lag_22', 'bnd_lag_22', 'bun_lag_22', 'ca_lag_22', 'cai_lag_22', 'ck_lag_22', 'ckmb_lag_22', 'cl_lag_22', 'crea_lag_22', 'crp_lag_22', 'dbp_lag_22', 'fgn_lag_22', 'fio2_lag_22', 'glu_lag_22', 'hgb_lag_22', 'hr_lag_22', 'inr_pt_lag_22', 'k_lag_22', 'lact_lag_22', 'lymph_lag_22', 'map_lag_22', 'mch_lag_22', 'mchc_lag_22', 'mcv_lag_22', 'methb_lag_22', 'mg_lag_22', 'na_lag_22', 'neut_lag_22', 'o2sat_lag_22', 'pco2_lag_22', 'ph_lag_22', 'phos_lag_22', 'plt_lag_22', 'po2_lag_22', 'ptt_lag_22', 'resp_lag_22', 'sbp_lag_22', 'temp_lag_22', 'tnt_lag_22', 'urine_lag_22', 'wbc_lag_22', 'alb_lag_23', 'alp_lag_23', 'alt_lag_23', 'ast_lag_23', 'be_lag_23', 'bicar_lag_23', 'bili_lag_23', 'bili_dir_lag_23', 'bnd_lag_23', 'bun_lag_23', 'ca_lag_23', 'cai_lag_23', 'ck_lag_23', 'ckmb_lag_23', 'cl_lag_23', 'crea_lag_23', 'crp_lag_23', 'dbp_lag_23', 'fgn_lag_23', 'fio2_lag_23', 'glu_lag_23', 'hgb_lag_23', 'hr_lag_23', 'inr_pt_lag_23', 'k_lag_23', 'lact_lag_23', 'lymph_lag_23', 'map_lag_23', 'mch_lag_23', 'mchc_lag_23', 'mcv_lag_23', 'methb_lag_23', 'mg_lag_23', 'na_lag_23', 'neut_lag_23', 'o2sat_lag_23', 'pco2_lag_23', 'ph_lag_23', 'phos_lag_23', 'plt_lag_23', 'po2_lag_23', 'ptt_lag_23', 'resp_lag_23', 'sbp_lag_23', 'temp_lag_23', 'tnt_lag_23', 'urine_lag_23', 'wbc_lag_23', 'alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24' ]
 PredictionDatasetpytorch.add_relative_time_idx=False
-PredictionDatasetpytorch.target=[
-    "alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",
+PredictionDatasetpytorch.target="label"
+            
+PredictionDatasetpytorch.time_varying_unknown_reals=[
                 "label",
               
             ]
-
-PredictionDatasetpytorch.time_varying_known_reals=[]
-PredictionDatasetpytorch.time_varying_unknown_reals=[
-                "alb",
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.lagged_variables=[  "alb",
                 "alp",
                 "alt",
                 "ast",
@@ -132,11 +85,6 @@ PredictionDatasetpytorch.time_varying_unknown_reals=[
                 "temp",
                 "tnt",
                 "urine",
-                "wbc",
-                "label",
-              
-            ]
-PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetpytorch.lagged_variables=[]
+                "wbc"]
 
-PredictionDatasetpytorch.target_normalizer='multi'
\ No newline at end of file
+                
\ No newline at end of file
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index a1804a1d..6c0f2e15 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -5,89 +5,42 @@ include "configs/prediction_models/common/DLCommon.gin"
 
 # Optimizer params
 train_common.model = @RNNpytorch
+
 optimizer/hyperparameter.class_to_tune = @Adam
 optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = 4.4815375287835764e-05
+optimizer/hyperparameter.lr = 0.00041
 
 # Model params
 model/hyperparameter.class_to_tune = @RNNpytorch
-model/hyperparameter.hidden =  42
+model/hyperparameter.hidden =  214
 #(2, 64, "log-uniform", 2)
-model/hyperparameter.rnn_layers=2
+model/hyperparameter.rnn_layers=1
 #(1,3)
 model/hyperparameter.num_classes = %NUM_CLASSES
 model/hyperparameter.cell_type='LSTM'
-model/hyperparameter.dropout = (0.397, 0.3971)
+model/hyperparameter.dropout = (0.244, 0.2441)
 #(0.0, 0.4)
 #model/hyperparameter.lr_scheduler = "exponential"
 
 train_common/hyperparameter.class_to_tune = @train_common
-train_common/hyperparameter.batch_size=128
+train_common/hyperparameter.batch_size=256
 #(32,64,128,256,512)
-train_common/hyperparameter.gradient_clip_val=0.01
+train_common/hyperparameter.gradient_clip_val=100.0
 #(0,0.01, 1.0, 100.0)
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
-PredictionDatasetpytorch.time_varying_known_reals=[]
+PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_1', 'alp_lag_1', 'alt_lag_1', 'ast_lag_1', 'be_lag_1', 'bicar_lag_1', 'bili_lag_1', 'bili_dir_lag_1', 'bnd_lag_1', 'bun_lag_1', 'ca_lag_1', 'cai_lag_1', 'ck_lag_1', 'ckmb_lag_1', 'cl_lag_1', 'crea_lag_1', 'crp_lag_1', 'dbp_lag_1', 'fgn_lag_1', 'fio2_lag_1', 'glu_lag_1', 'hgb_lag_1', 'hr_lag_1', 'inr_pt_lag_1', 'k_lag_1', 'lact_lag_1', 'lymph_lag_1', 'map_lag_1', 'mch_lag_1', 'mchc_lag_1', 'mcv_lag_1', 'methb_lag_1', 'mg_lag_1', 'na_lag_1', 'neut_lag_1', 'o2sat_lag_1', 'pco2_lag_1', 'ph_lag_1', 'phos_lag_1', 'plt_lag_1', 'po2_lag_1', 'ptt_lag_1', 'resp_lag_1', 'sbp_lag_1', 'temp_lag_1', 'tnt_lag_1', 'urine_lag_1', 'wbc_lag_1', 'alb_lag_2', 'alp_lag_2', 'alt_lag_2', 'ast_lag_2', 'be_lag_2', 'bicar_lag_2', 'bili_lag_2', 'bili_dir_lag_2', 'bnd_lag_2', 'bun_lag_2', 'ca_lag_2', 'cai_lag_2', 'ck_lag_2', 'ckmb_lag_2', 'cl_lag_2', 'crea_lag_2', 'crp_lag_2', 'dbp_lag_2', 'fgn_lag_2', 'fio2_lag_2', 'glu_lag_2', 'hgb_lag_2', 'hr_lag_2', 'inr_pt_lag_2', 'k_lag_2', 'lact_lag_2', 'lymph_lag_2', 'map_lag_2', 'mch_lag_2', 'mchc_lag_2', 'mcv_lag_2', 'methb_lag_2', 'mg_lag_2', 'na_lag_2', 'neut_lag_2', 'o2sat_lag_2', 'pco2_lag_2', 'ph_lag_2', 'phos_lag_2', 'plt_lag_2', 'po2_lag_2', 'ptt_lag_2', 'resp_lag_2', 'sbp_lag_2', 'temp_lag_2', 'tnt_lag_2', 'urine_lag_2', 'wbc_lag_2', 'alb_lag_3', 'alp_lag_3', 'alt_lag_3', 'ast_lag_3', 'be_lag_3', 'bicar_lag_3', 'bili_lag_3', 'bili_dir_lag_3', 'bnd_lag_3', 'bun_lag_3', 'ca_lag_3', 'cai_lag_3', 'ck_lag_3', 'ckmb_lag_3', 'cl_lag_3', 'crea_lag_3', 'crp_lag_3', 'dbp_lag_3', 'fgn_lag_3', 'fio2_lag_3', 'glu_lag_3', 'hgb_lag_3', 'hr_lag_3', 'inr_pt_lag_3', 'k_lag_3', 'lact_lag_3', 'lymph_lag_3', 'map_lag_3', 'mch_lag_3', 'mchc_lag_3', 'mcv_lag_3', 'methb_lag_3', 'mg_lag_3', 'na_lag_3', 'neut_lag_3', 'o2sat_lag_3', 'pco2_lag_3', 'ph_lag_3', 'phos_lag_3', 'plt_lag_3', 'po2_lag_3', 'ptt_lag_3', 'resp_lag_3', 'sbp_lag_3', 'temp_lag_3', 'tnt_lag_3', 'urine_lag_3', 'wbc_lag_3', 'alb_lag_4', 'alp_lag_4', 'alt_lag_4', 'ast_lag_4', 'be_lag_4', 'bicar_lag_4', 'bili_lag_4', 'bili_dir_lag_4', 'bnd_lag_4', 'bun_lag_4', 'ca_lag_4', 'cai_lag_4', 'ck_lag_4', 'ckmb_lag_4', 'cl_lag_4', 'crea_lag_4', 'crp_lag_4', 'dbp_lag_4', 'fgn_lag_4', 'fio2_lag_4', 'glu_lag_4', 'hgb_lag_4', 'hr_lag_4', 'inr_pt_lag_4', 'k_lag_4', 'lact_lag_4', 'lymph_lag_4', 'map_lag_4', 'mch_lag_4', 'mchc_lag_4', 'mcv_lag_4', 'methb_lag_4', 'mg_lag_4', 'na_lag_4', 'neut_lag_4', 'o2sat_lag_4', 'pco2_lag_4', 'ph_lag_4', 'phos_lag_4', 'plt_lag_4', 'po2_lag_4', 'ptt_lag_4', 'resp_lag_4', 'sbp_lag_4', 'temp_lag_4', 'tnt_lag_4', 'urine_lag_4', 'wbc_lag_4', 'alb_lag_5', 'alp_lag_5', 'alt_lag_5', 'ast_lag_5', 'be_lag_5', 'bicar_lag_5', 'bili_lag_5', 'bili_dir_lag_5', 'bnd_lag_5', 'bun_lag_5', 'ca_lag_5', 'cai_lag_5', 'ck_lag_5', 'ckmb_lag_5', 'cl_lag_5', 'crea_lag_5', 'crp_lag_5', 'dbp_lag_5', 'fgn_lag_5', 'fio2_lag_5', 'glu_lag_5', 'hgb_lag_5', 'hr_lag_5', 'inr_pt_lag_5', 'k_lag_5', 'lact_lag_5', 'lymph_lag_5', 'map_lag_5', 'mch_lag_5', 'mchc_lag_5', 'mcv_lag_5', 'methb_lag_5', 'mg_lag_5', 'na_lag_5', 'neut_lag_5', 'o2sat_lag_5', 'pco2_lag_5', 'ph_lag_5', 'phos_lag_5', 'plt_lag_5', 'po2_lag_5', 'ptt_lag_5', 'resp_lag_5', 'sbp_lag_5', 'temp_lag_5', 'tnt_lag_5', 'urine_lag_5', 'wbc_lag_5', 'alb_lag_6', 'alp_lag_6', 'alt_lag_6', 'ast_lag_6', 'be_lag_6', 'bicar_lag_6', 'bili_lag_6', 'bili_dir_lag_6', 'bnd_lag_6', 'bun_lag_6', 'ca_lag_6', 'cai_lag_6', 'ck_lag_6', 'ckmb_lag_6', 'cl_lag_6', 'crea_lag_6', 'crp_lag_6', 'dbp_lag_6', 'fgn_lag_6', 'fio2_lag_6', 'glu_lag_6', 'hgb_lag_6', 'hr_lag_6', 'inr_pt_lag_6', 'k_lag_6', 'lact_lag_6', 'lymph_lag_6', 'map_lag_6', 'mch_lag_6', 'mchc_lag_6', 'mcv_lag_6', 'methb_lag_6', 'mg_lag_6', 'na_lag_6', 'neut_lag_6', 'o2sat_lag_6', 'pco2_lag_6', 'ph_lag_6', 'phos_lag_6', 'plt_lag_6', 'po2_lag_6', 'ptt_lag_6', 'resp_lag_6', 'sbp_lag_6', 'temp_lag_6', 'tnt_lag_6', 'urine_lag_6', 'wbc_lag_6', 'alb_lag_7', 'alp_lag_7', 'alt_lag_7', 'ast_lag_7', 'be_lag_7', 'bicar_lag_7', 'bili_lag_7', 'bili_dir_lag_7', 'bnd_lag_7', 'bun_lag_7', 'ca_lag_7', 'cai_lag_7', 'ck_lag_7', 'ckmb_lag_7', 'cl_lag_7', 'crea_lag_7', 'crp_lag_7', 'dbp_lag_7', 'fgn_lag_7', 'fio2_lag_7', 'glu_lag_7', 'hgb_lag_7', 'hr_lag_7', 'inr_pt_lag_7', 'k_lag_7', 'lact_lag_7', 'lymph_lag_7', 'map_lag_7', 'mch_lag_7', 'mchc_lag_7', 'mcv_lag_7', 'methb_lag_7', 'mg_lag_7', 'na_lag_7', 'neut_lag_7', 'o2sat_lag_7', 'pco2_lag_7', 'ph_lag_7', 'phos_lag_7', 'plt_lag_7', 'po2_lag_7', 'ptt_lag_7', 'resp_lag_7', 'sbp_lag_7', 'temp_lag_7', 'tnt_lag_7', 'urine_lag_7', 'wbc_lag_7', 'alb_lag_8', 'alp_lag_8', 'alt_lag_8', 'ast_lag_8', 'be_lag_8', 'bicar_lag_8', 'bili_lag_8', 'bili_dir_lag_8', 'bnd_lag_8', 'bun_lag_8', 'ca_lag_8', 'cai_lag_8', 'ck_lag_8', 'ckmb_lag_8', 'cl_lag_8', 'crea_lag_8', 'crp_lag_8', 'dbp_lag_8', 'fgn_lag_8', 'fio2_lag_8', 'glu_lag_8', 'hgb_lag_8', 'hr_lag_8', 'inr_pt_lag_8', 'k_lag_8', 'lact_lag_8', 'lymph_lag_8', 'map_lag_8', 'mch_lag_8', 'mchc_lag_8', 'mcv_lag_8', 'methb_lag_8', 'mg_lag_8', 'na_lag_8', 'neut_lag_8', 'o2sat_lag_8', 'pco2_lag_8', 'ph_lag_8', 'phos_lag_8', 'plt_lag_8', 'po2_lag_8', 'ptt_lag_8', 'resp_lag_8', 'sbp_lag_8', 'temp_lag_8', 'tnt_lag_8', 'urine_lag_8', 'wbc_lag_8', 'alb_lag_9', 'alp_lag_9', 'alt_lag_9', 'ast_lag_9', 'be_lag_9', 'bicar_lag_9', 'bili_lag_9', 'bili_dir_lag_9', 'bnd_lag_9', 'bun_lag_9', 'ca_lag_9', 'cai_lag_9', 'ck_lag_9', 'ckmb_lag_9', 'cl_lag_9', 'crea_lag_9', 'crp_lag_9', 'dbp_lag_9', 'fgn_lag_9', 'fio2_lag_9', 'glu_lag_9', 'hgb_lag_9', 'hr_lag_9', 'inr_pt_lag_9', 'k_lag_9', 'lact_lag_9', 'lymph_lag_9', 'map_lag_9', 'mch_lag_9', 'mchc_lag_9', 'mcv_lag_9', 'methb_lag_9', 'mg_lag_9', 'na_lag_9', 'neut_lag_9', 'o2sat_lag_9', 'pco2_lag_9', 'ph_lag_9', 'phos_lag_9', 'plt_lag_9', 'po2_lag_9', 'ptt_lag_9', 'resp_lag_9', 'sbp_lag_9', 'temp_lag_9', 'tnt_lag_9', 'urine_lag_9', 'wbc_lag_9', 'alb_lag_10', 'alp_lag_10', 'alt_lag_10', 'ast_lag_10', 'be_lag_10', 'bicar_lag_10', 'bili_lag_10', 'bili_dir_lag_10', 'bnd_lag_10', 'bun_lag_10', 'ca_lag_10', 'cai_lag_10', 'ck_lag_10', 'ckmb_lag_10', 'cl_lag_10', 'crea_lag_10', 'crp_lag_10', 'dbp_lag_10', 'fgn_lag_10', 'fio2_lag_10', 'glu_lag_10', 'hgb_lag_10', 'hr_lag_10', 'inr_pt_lag_10', 'k_lag_10', 'lact_lag_10', 'lymph_lag_10', 'map_lag_10', 'mch_lag_10', 'mchc_lag_10', 'mcv_lag_10', 'methb_lag_10', 'mg_lag_10', 'na_lag_10', 'neut_lag_10', 'o2sat_lag_10', 'pco2_lag_10', 'ph_lag_10', 'phos_lag_10', 'plt_lag_10', 'po2_lag_10', 'ptt_lag_10', 'resp_lag_10', 'sbp_lag_10', 'temp_lag_10', 'tnt_lag_10', 'urine_lag_10', 'wbc_lag_10', 'alb_lag_11', 'alp_lag_11', 'alt_lag_11', 'ast_lag_11', 'be_lag_11', 'bicar_lag_11', 'bili_lag_11', 'bili_dir_lag_11', 'bnd_lag_11', 'bun_lag_11', 'ca_lag_11', 'cai_lag_11', 'ck_lag_11', 'ckmb_lag_11', 'cl_lag_11', 'crea_lag_11', 'crp_lag_11', 'dbp_lag_11', 'fgn_lag_11', 'fio2_lag_11', 'glu_lag_11', 'hgb_lag_11', 'hr_lag_11', 'inr_pt_lag_11', 'k_lag_11', 'lact_lag_11', 'lymph_lag_11', 'map_lag_11', 'mch_lag_11', 'mchc_lag_11', 'mcv_lag_11', 'methb_lag_11', 'mg_lag_11', 'na_lag_11', 'neut_lag_11', 'o2sat_lag_11', 'pco2_lag_11', 'ph_lag_11', 'phos_lag_11', 'plt_lag_11', 'po2_lag_11', 'ptt_lag_11', 'resp_lag_11', 'sbp_lag_11', 'temp_lag_11', 'tnt_lag_11', 'urine_lag_11', 'wbc_lag_11', 'alb_lag_12', 'alp_lag_12', 'alt_lag_12', 'ast_lag_12', 'be_lag_12', 'bicar_lag_12', 'bili_lag_12', 'bili_dir_lag_12', 'bnd_lag_12', 'bun_lag_12', 'ca_lag_12', 'cai_lag_12', 'ck_lag_12', 'ckmb_lag_12', 'cl_lag_12', 'crea_lag_12', 'crp_lag_12', 'dbp_lag_12', 'fgn_lag_12', 'fio2_lag_12', 'glu_lag_12', 'hgb_lag_12', 'hr_lag_12', 'inr_pt_lag_12', 'k_lag_12', 'lact_lag_12', 'lymph_lag_12', 'map_lag_12', 'mch_lag_12', 'mchc_lag_12', 'mcv_lag_12', 'methb_lag_12', 'mg_lag_12', 'na_lag_12', 'neut_lag_12', 'o2sat_lag_12', 'pco2_lag_12', 'ph_lag_12', 'phos_lag_12', 'plt_lag_12', 'po2_lag_12', 'ptt_lag_12', 'resp_lag_12', 'sbp_lag_12', 'temp_lag_12', 'tnt_lag_12', 'urine_lag_12', 'wbc_lag_12', 'alb_lag_13', 'alp_lag_13', 'alt_lag_13', 'ast_lag_13', 'be_lag_13', 'bicar_lag_13', 'bili_lag_13', 'bili_dir_lag_13', 'bnd_lag_13', 'bun_lag_13', 'ca_lag_13', 'cai_lag_13', 'ck_lag_13', 'ckmb_lag_13', 'cl_lag_13', 'crea_lag_13', 'crp_lag_13', 'dbp_lag_13', 'fgn_lag_13', 'fio2_lag_13', 'glu_lag_13', 'hgb_lag_13', 'hr_lag_13', 'inr_pt_lag_13', 'k_lag_13', 'lact_lag_13', 'lymph_lag_13', 'map_lag_13', 'mch_lag_13', 'mchc_lag_13', 'mcv_lag_13', 'methb_lag_13', 'mg_lag_13', 'na_lag_13', 'neut_lag_13', 'o2sat_lag_13', 'pco2_lag_13', 'ph_lag_13', 'phos_lag_13', 'plt_lag_13', 'po2_lag_13', 'ptt_lag_13', 'resp_lag_13', 'sbp_lag_13', 'temp_lag_13', 'tnt_lag_13', 'urine_lag_13', 'wbc_lag_13', 'alb_lag_14', 'alp_lag_14', 'alt_lag_14', 'ast_lag_14', 'be_lag_14', 'bicar_lag_14', 'bili_lag_14', 'bili_dir_lag_14', 'bnd_lag_14', 'bun_lag_14', 'ca_lag_14', 'cai_lag_14', 'ck_lag_14', 'ckmb_lag_14', 'cl_lag_14', 'crea_lag_14', 'crp_lag_14', 'dbp_lag_14', 'fgn_lag_14', 'fio2_lag_14', 'glu_lag_14', 'hgb_lag_14', 'hr_lag_14', 'inr_pt_lag_14', 'k_lag_14', 'lact_lag_14', 'lymph_lag_14', 'map_lag_14', 'mch_lag_14', 'mchc_lag_14', 'mcv_lag_14', 'methb_lag_14', 'mg_lag_14', 'na_lag_14', 'neut_lag_14', 'o2sat_lag_14', 'pco2_lag_14', 'ph_lag_14', 'phos_lag_14', 'plt_lag_14', 'po2_lag_14', 'ptt_lag_14', 'resp_lag_14', 'sbp_lag_14', 'temp_lag_14', 'tnt_lag_14', 'urine_lag_14', 'wbc_lag_14', 'alb_lag_15', 'alp_lag_15', 'alt_lag_15', 'ast_lag_15', 'be_lag_15', 'bicar_lag_15', 'bili_lag_15', 'bili_dir_lag_15', 'bnd_lag_15', 'bun_lag_15', 'ca_lag_15', 'cai_lag_15', 'ck_lag_15', 'ckmb_lag_15', 'cl_lag_15', 'crea_lag_15', 'crp_lag_15', 'dbp_lag_15', 'fgn_lag_15', 'fio2_lag_15', 'glu_lag_15', 'hgb_lag_15', 'hr_lag_15', 'inr_pt_lag_15', 'k_lag_15', 'lact_lag_15', 'lymph_lag_15', 'map_lag_15', 'mch_lag_15', 'mchc_lag_15', 'mcv_lag_15', 'methb_lag_15', 'mg_lag_15', 'na_lag_15', 'neut_lag_15', 'o2sat_lag_15', 'pco2_lag_15', 'ph_lag_15', 'phos_lag_15', 'plt_lag_15', 'po2_lag_15', 'ptt_lag_15', 'resp_lag_15', 'sbp_lag_15', 'temp_lag_15', 'tnt_lag_15', 'urine_lag_15', 'wbc_lag_15', 'alb_lag_16', 'alp_lag_16', 'alt_lag_16', 'ast_lag_16', 'be_lag_16', 'bicar_lag_16', 'bili_lag_16', 'bili_dir_lag_16', 'bnd_lag_16', 'bun_lag_16', 'ca_lag_16', 'cai_lag_16', 'ck_lag_16', 'ckmb_lag_16', 'cl_lag_16', 'crea_lag_16', 'crp_lag_16', 'dbp_lag_16', 'fgn_lag_16', 'fio2_lag_16', 'glu_lag_16', 'hgb_lag_16', 'hr_lag_16', 'inr_pt_lag_16', 'k_lag_16', 'lact_lag_16', 'lymph_lag_16', 'map_lag_16', 'mch_lag_16', 'mchc_lag_16', 'mcv_lag_16', 'methb_lag_16', 'mg_lag_16', 'na_lag_16', 'neut_lag_16', 'o2sat_lag_16', 'pco2_lag_16', 'ph_lag_16', 'phos_lag_16', 'plt_lag_16', 'po2_lag_16', 'ptt_lag_16', 'resp_lag_16', 'sbp_lag_16', 'temp_lag_16', 'tnt_lag_16', 'urine_lag_16', 'wbc_lag_16', 'alb_lag_17', 'alp_lag_17', 'alt_lag_17', 'ast_lag_17', 'be_lag_17', 'bicar_lag_17', 'bili_lag_17', 'bili_dir_lag_17', 'bnd_lag_17', 'bun_lag_17', 'ca_lag_17', 'cai_lag_17', 'ck_lag_17', 'ckmb_lag_17', 'cl_lag_17', 'crea_lag_17', 'crp_lag_17', 'dbp_lag_17', 'fgn_lag_17', 'fio2_lag_17', 'glu_lag_17', 'hgb_lag_17', 'hr_lag_17', 'inr_pt_lag_17', 'k_lag_17', 'lact_lag_17', 'lymph_lag_17', 'map_lag_17', 'mch_lag_17', 'mchc_lag_17', 'mcv_lag_17', 'methb_lag_17', 'mg_lag_17', 'na_lag_17', 'neut_lag_17', 'o2sat_lag_17', 'pco2_lag_17', 'ph_lag_17', 'phos_lag_17', 'plt_lag_17', 'po2_lag_17', 'ptt_lag_17', 'resp_lag_17', 'sbp_lag_17', 'temp_lag_17', 'tnt_lag_17', 'urine_lag_17', 'wbc_lag_17', 'alb_lag_18', 'alp_lag_18', 'alt_lag_18', 'ast_lag_18', 'be_lag_18', 'bicar_lag_18', 'bili_lag_18', 'bili_dir_lag_18', 'bnd_lag_18', 'bun_lag_18', 'ca_lag_18', 'cai_lag_18', 'ck_lag_18', 'ckmb_lag_18', 'cl_lag_18', 'crea_lag_18', 'crp_lag_18', 'dbp_lag_18', 'fgn_lag_18', 'fio2_lag_18', 'glu_lag_18', 'hgb_lag_18', 'hr_lag_18', 'inr_pt_lag_18', 'k_lag_18', 'lact_lag_18', 'lymph_lag_18', 'map_lag_18', 'mch_lag_18', 'mchc_lag_18', 'mcv_lag_18', 'methb_lag_18', 'mg_lag_18', 'na_lag_18', 'neut_lag_18', 'o2sat_lag_18', 'pco2_lag_18', 'ph_lag_18', 'phos_lag_18', 'plt_lag_18', 'po2_lag_18', 'ptt_lag_18', 'resp_lag_18', 'sbp_lag_18', 'temp_lag_18', 'tnt_lag_18', 'urine_lag_18', 'wbc_lag_18', 'alb_lag_19', 'alp_lag_19', 'alt_lag_19', 'ast_lag_19', 'be_lag_19', 'bicar_lag_19', 'bili_lag_19', 'bili_dir_lag_19', 'bnd_lag_19', 'bun_lag_19', 'ca_lag_19', 'cai_lag_19', 'ck_lag_19', 'ckmb_lag_19', 'cl_lag_19', 'crea_lag_19', 'crp_lag_19', 'dbp_lag_19', 'fgn_lag_19', 'fio2_lag_19', 'glu_lag_19', 'hgb_lag_19', 'hr_lag_19', 'inr_pt_lag_19', 'k_lag_19', 'lact_lag_19', 'lymph_lag_19', 'map_lag_19', 'mch_lag_19', 'mchc_lag_19', 'mcv_lag_19', 'methb_lag_19', 'mg_lag_19', 'na_lag_19', 'neut_lag_19', 'o2sat_lag_19', 'pco2_lag_19', 'ph_lag_19', 'phos_lag_19', 'plt_lag_19', 'po2_lag_19', 'ptt_lag_19', 'resp_lag_19', 'sbp_lag_19', 'temp_lag_19', 'tnt_lag_19', 'urine_lag_19', 'wbc_lag_19', 'alb_lag_20', 'alp_lag_20', 'alt_lag_20', 'ast_lag_20', 'be_lag_20', 'bicar_lag_20', 'bili_lag_20', 'bili_dir_lag_20', 'bnd_lag_20', 'bun_lag_20', 'ca_lag_20', 'cai_lag_20', 'ck_lag_20', 'ckmb_lag_20', 'cl_lag_20', 'crea_lag_20', 'crp_lag_20', 'dbp_lag_20', 'fgn_lag_20', 'fio2_lag_20', 'glu_lag_20', 'hgb_lag_20', 'hr_lag_20', 'inr_pt_lag_20', 'k_lag_20', 'lact_lag_20', 'lymph_lag_20', 'map_lag_20', 'mch_lag_20', 'mchc_lag_20', 'mcv_lag_20', 'methb_lag_20', 'mg_lag_20', 'na_lag_20', 'neut_lag_20', 'o2sat_lag_20', 'pco2_lag_20', 'ph_lag_20', 'phos_lag_20', 'plt_lag_20', 'po2_lag_20', 'ptt_lag_20', 'resp_lag_20', 'sbp_lag_20', 'temp_lag_20', 'tnt_lag_20', 'urine_lag_20', 'wbc_lag_20', 'alb_lag_21', 'alp_lag_21', 'alt_lag_21', 'ast_lag_21', 'be_lag_21', 'bicar_lag_21', 'bili_lag_21', 'bili_dir_lag_21', 'bnd_lag_21', 'bun_lag_21', 'ca_lag_21', 'cai_lag_21', 'ck_lag_21', 'ckmb_lag_21', 'cl_lag_21', 'crea_lag_21', 'crp_lag_21', 'dbp_lag_21', 'fgn_lag_21', 'fio2_lag_21', 'glu_lag_21', 'hgb_lag_21', 'hr_lag_21', 'inr_pt_lag_21', 'k_lag_21', 'lact_lag_21', 'lymph_lag_21', 'map_lag_21', 'mch_lag_21', 'mchc_lag_21', 'mcv_lag_21', 'methb_lag_21', 'mg_lag_21', 'na_lag_21', 'neut_lag_21', 'o2sat_lag_21', 'pco2_lag_21', 'ph_lag_21', 'phos_lag_21', 'plt_lag_21', 'po2_lag_21', 'ptt_lag_21', 'resp_lag_21', 'sbp_lag_21', 'temp_lag_21', 'tnt_lag_21', 'urine_lag_21', 'wbc_lag_21', 'alb_lag_22', 'alp_lag_22', 'alt_lag_22', 'ast_lag_22', 'be_lag_22', 'bicar_lag_22', 'bili_lag_22', 'bili_dir_lag_22', 'bnd_lag_22', 'bun_lag_22', 'ca_lag_22', 'cai_lag_22', 'ck_lag_22', 'ckmb_lag_22', 'cl_lag_22', 'crea_lag_22', 'crp_lag_22', 'dbp_lag_22', 'fgn_lag_22', 'fio2_lag_22', 'glu_lag_22', 'hgb_lag_22', 'hr_lag_22', 'inr_pt_lag_22', 'k_lag_22', 'lact_lag_22', 'lymph_lag_22', 'map_lag_22', 'mch_lag_22', 'mchc_lag_22', 'mcv_lag_22', 'methb_lag_22', 'mg_lag_22', 'na_lag_22', 'neut_lag_22', 'o2sat_lag_22', 'pco2_lag_22', 'ph_lag_22', 'phos_lag_22', 'plt_lag_22', 'po2_lag_22', 'ptt_lag_22', 'resp_lag_22', 'sbp_lag_22', 'temp_lag_22', 'tnt_lag_22', 'urine_lag_22', 'wbc_lag_22', 'alb_lag_23', 'alp_lag_23', 'alt_lag_23', 'ast_lag_23', 'be_lag_23', 'bicar_lag_23', 'bili_lag_23', 'bili_dir_lag_23', 'bnd_lag_23', 'bun_lag_23', 'ca_lag_23', 'cai_lag_23', 'ck_lag_23', 'ckmb_lag_23', 'cl_lag_23', 'crea_lag_23', 'crp_lag_23', 'dbp_lag_23', 'fgn_lag_23', 'fio2_lag_23', 'glu_lag_23', 'hgb_lag_23', 'hr_lag_23', 'inr_pt_lag_23', 'k_lag_23', 'lact_lag_23', 'lymph_lag_23', 'map_lag_23', 'mch_lag_23', 'mchc_lag_23', 'mcv_lag_23', 'methb_lag_23', 'mg_lag_23', 'na_lag_23', 'neut_lag_23', 'o2sat_lag_23', 'pco2_lag_23', 'ph_lag_23', 'phos_lag_23', 'plt_lag_23', 'po2_lag_23', 'ptt_lag_23', 'resp_lag_23', 'sbp_lag_23', 'temp_lag_23', 'tnt_lag_23', 'urine_lag_23', 'wbc_lag_23', 'alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24' ]
 
 PredictionDatasetpytorch.add_relative_time_idx=False
-PredictionDatasetpytorch.target=[
-    "alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",
+PredictionDatasetpytorch.target="label"
+            
+PredictionDatasetpytorch.time_varying_unknown_reals=[
                 "label",
               
             ]
-
-            
-PredictionDatasetpytorch.time_varying_unknown_reals=[
-                "alb",
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.lagged_variables=[  "alb",
                 "alp",
                 "alt",
                 "ast",
@@ -134,10 +87,4 @@ PredictionDatasetpytorch.time_varying_unknown_reals=[
                 "temp",
                 "tnt",
                 "urine",
-                "wbc",
-                "label",
-              
-            ]
-PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetpytorch.lagged_variables=[  ]
-PredictionDatasetpytorch.target_normalizer='multi'
\ No newline at end of file
+                "wbc"]
\ No newline at end of file
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index de79f61f..c7264ed5 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -89,5 +89,3 @@ PredictionDatasetpytorch.time_varying_unknown_reals=["alb",
                 "urine",
                 "wbc",]
 PredictionDatasetpytorch.lagged_variables=[]
-PredictionDatasetpytorch.lagged_variables=[]
-PredictionDatasetpytorch.target_normalizer='single'
\ No newline at end of file
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 69db2960..535d9619 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -38,9 +38,9 @@ def execute_repeated_cv(
     verbose: bool = False,
     wandb: bool = False,
     complete_train: bool = False,
-    explain: list = None,
+    explain: bool = False,
     pytorch_forecasting: bool = False,
-    XAI_metric: list = None,
+    XAI_metric: bool = False,
     random_labels: bool = False,
     random_model_dir: str = None,
 ) -> float:
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 49c898b7..a0e51dbc 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -10,7 +10,7 @@
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
-from pytorch_forecasting import TimeSeriesDataSet, GroupNormalizer, MultiNormalizer, EncoderNormalizer
+from pytorch_forecasting import TimeSeriesDataSet, GroupNormalizer
 
 
 class CommonDataset(Dataset):
@@ -450,7 +450,6 @@ def __init__(
         time_varying_known_reals: List[str],
         time_varying_unknown_categoricals: List[str],
         lagged_variables: List[str],
-        # target_normalizer: str,
         *args,
         ram_cache: bool = False,
         add_relative_time_idx: bool = False,
@@ -483,13 +482,7 @@ def __init__(
         self.ram_cache = ram_cache
         self.kwargs = kwargs
         self.column_names = features.columns
-        """ if target_normalizer == 'multi':
-            target_normalizer = MultiNormalizer([EncoderNormalizer(transformation='relu') for _ in range(len(target)-1)] + [GroupNormalizer(groups=["stay_id"], transformation="relu")]
-                                                )
-        else:
-            target_normalizer = GroupNormalizer(
-                groups=["stay_id"], transformation="relu"
-            ) """
+
         super().__init__(
             data=self.data,
             time_idx="time_idx",
@@ -509,7 +502,6 @@ def __init__(
             # add_target_scales=True,
             # add_encoder_length=True,
             predict_mode=True,
-
             target_normalizer=GroupNormalizer(
                 groups=["stay_id"], transformation="relu"
             ),
diff --git a/icu_benchmarks/models/constants.py b/icu_benchmarks/models/constants.py
index 702fe7f7..093d15bd 100644
--- a/icu_benchmarks/models/constants.py
+++ b/icu_benchmarks/models/constants.py
@@ -26,9 +26,6 @@
     MAE,
     JSD,
     BinaryFairnessWrapper,
-    Faithfulness,
-    Stability,
-    Randomization
 )
 
 
@@ -88,17 +85,6 @@ class DLMetrics:
         "mae": MAE,
         "jsd": JSD,
     }
-    Faithfulness = {
-        "Faithfulness_timesteps": Faithfulness,
-        "Faithfulness_features": Faithfulness,
-        "Faithfulness_feature_timestep": Faithfulness,
-    }
-    Stability = {
-        "Stability": Stability
-    }
-    Randmoization = {
-        "DDR": Randomization,
-    }
 
 
 class ImputationInit(str, Enum):
diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index 9b84a18d..4e082d29 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -6,8 +6,6 @@
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
 from torchmetrics.classification import BinaryFairness
-from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson, cosine
-
 
 """"
 This file contains custom metrics that can be added to YAIB.
@@ -135,6 +133,7 @@ def feature_helper(self, trainer, step_prefix):
             feature_names = trainer.test_dataloaders.dataset.features
         return feature_names
 
+
 # XAI Metrics
 
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 0966dec7..ff66c4da 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -12,7 +12,7 @@
 )
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import DLPredictionWrapper, DLPredictionPytorchForecastingWrapper
-from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, is_tensor, autograd
+from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, is_tensor
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss, MAE
 import matplotlib.pyplot as plt
@@ -475,8 +475,8 @@ def __init__(
         *args,
         **kwargs,
     ):
-        super().__init__(optimizer=optimizer, pytorch_forecasting=True, dataset=dataset, * args, **kwargs)
-
+        super().__init__(optimizer=optimizer, pytorch_forecasting=True, *args, **kwargs)
+        self.dataset = dataset
         self.model = TemporalFusionTransformer.from_dataset(
             dataset=dataset,
             hidden_size=hidden,
@@ -486,7 +486,6 @@ def __init__(
             loss=QuantileLoss(),
             hidden_continuous_size=hidden,
         )
-        self.dataset = dataset
         self.num_classes = num_classes
         self.logit = nn.Linear(7, num_classes)
 
@@ -576,10 +575,7 @@ def forward_captum(
             target_scale,
         )
 
-        # with autograd.set_grad_enabled(True):
-        result = self.forward(tuple_x)
-
-        return result
+        return self.forward(tuple_x)
 
 
 @gin.configurable
@@ -638,11 +634,9 @@ def forward(
 
         out = self.model(x_dict)
 
-        out = ((out["prediction"][-1]).squeeze(dim=-1))
-
-        pred = self.logit(out)
+        pred = self.logit(out["prediction"])
 
-        return pred.unsqueeze(1)
+        return pred
 
 
 @gin.configurable
@@ -700,8 +694,6 @@ def forward(
             x_dict["encoder_cont"][:, :, -1] = 0.0
         out = self.model(x_dict)
 
-        out = ((out["prediction"][-1]))
-
-        pred = self.logit(out)
+        pred = self.logit(out["prediction"])
 
         return pred
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index cf94843c..32358d29 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -73,9 +73,9 @@ def train_common(
         torch.cuda.device_count() * 8 * int(torch.cuda.is_available()),
         32,
     ),
-    explain: list = None,
+    explain: bool = False,
     pytorch_forecasting: bool = False,
-    XAI_metric: list = None,
+    XAI_metric: bool = False,
     random_labels: bool = False,
     random_model_dir: str = None,
 
@@ -154,12 +154,6 @@ def train_common(
             drop_last=True,
             shuffle=False,
         )
-        if random_model_dir is None:
-            random_model = None
-        else:
-
-            random_model = Path(random_model_dir)
-
         if load_weights:
             model = load_model(
                 model,
@@ -167,11 +161,6 @@ def train_common(
                 pl_model=pl_model,
                 train_dataset=train_dataset,
                 optimizer=optimizer,
-                explain=explain,
-                XAI_metric=XAI_metric,
-                random_model=random_model,
-                test_loader=test_loader
-
             )
 
         else:
@@ -181,11 +170,6 @@ def train_common(
                 epochs=epochs,
                 run_mode=mode,
                 batch_size=batch_size,
-                explain=explain,
-                XAI_metric=XAI_metric,
-                random_model=random_model,
-                test_loader=test_loader
-
             )
         if random_labels:
             train_dataset.randomize_labels(num_classes=model.num_classes)
@@ -262,8 +246,7 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
-        gradient_clip_val=gradient_clip_val,
-        inference_mode=False
+        # gradient_clip_val=gradient_clip_val
     )
     if not eval_only:
         if model.requires_backprop:
@@ -281,7 +264,21 @@ def train_common(
         save_config_file(log_dir)
         return 0
 
-        """ XAI_dict = {}  # dictrionary to log attributions metrics
+    if explain:
+        if random_model_dir is None:
+            random_model = None
+        else:
+            path = Path(random_model_dir)
+
+            random_model = load_model(
+                model,
+                source_dir=path,
+                pl_model=pl_model,
+                train_dataset=train_dataset,
+                optimizer=optimizer,
+            )
+
+        XAI_dict = {}  # dictrionary to log attributions metrics
 
         # choose which  methods to get attributions
         methods = {
@@ -290,11 +287,8 @@ def train_common(
             "IG": IntegratedGradients,
             # "FA": FeatureAblation,
             "R": "Random",
-<<<<<<< HEAD
             "Att": "Attention"
-=======
-            # "Att": "Attention"
->>>>>>> parent of a39b5c4... added condition for ROS and RIS for attention
+
         }
         for key, item in methods.items():
             # If conditions needed here as different explantations require different inputs
@@ -377,7 +371,7 @@ def create_default_mask(shape):
 
         # Write the dictionary to a JSON file
         with open(json_file_path, "w") as json_file:
-            json.dump(XAI_dict, json_file) """
+            json.dump(XAI_dict, json_file)
 
     model.set_weight("balanced", train_dataset)
     test_loss = trainer.test(model, dataloaders=test_loader, verbose=verbose)[0]["test/loss"]
@@ -385,7 +379,7 @@ def create_default_mask(shape):
     return test_loss
 
 
-def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=None, explain=None, XAI_metric=None, random_model=None, test_loader=None):
+def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=None):
     if source_dir.exists():
         if model.requires_backprop:
             if (source_dir / "model.ckpt").exists():
@@ -398,8 +392,7 @@ def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=N
                 return Exception(f"No weights to load at path : {source_dir}")
             if pl_model:
                 if train_dataset is not None:
-                    model = model.load_from_checkpoint(model_path, dataset=train_dataset, optimizer=optimizer,
-                                                       explain=explain, XAI_metric=XAI_metric, test_loader=test_loader)
+                    model = model.load_from_checkpoint(model_path, dataset=train_dataset, optimizer=optimizer)
 
                 else:
                     model = model.load_from_checkpoint(model_path)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 0f86e6cc..b1844d3d 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -18,10 +18,9 @@
 from pytorch_lightning import LightningModule
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
-from icu_benchmarks.models.train import load_model
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson, cosine
-from captum.attr import IntegratedGradients, Saliency, FeatureAblation, Lime
+import captum
 from captum._utils.models.linear_model import SkLearnLasso
 import os
 
@@ -243,7 +242,8 @@ def __init__(
         input_size: Tensor = None,
         initialization_method: str = "normal",
         pytorch_forecasting: bool = False,
-
+        explain: list = [],
+        XAI_metric: list = [],
         **kwargs,
     ):
         super().__init__(
@@ -264,6 +264,8 @@ def __init__(
         self.output_transform = None
         self.loss_weights = None
         self.pytorch_forecasting = pytorch_forecasting
+        self.explain = explain
+        self.XAI_metric = XAI_metric
 
     def set_weight(self, weight, dataset):
         """Set the weight for the loss function."""
@@ -406,11 +408,6 @@ def __init__(
         input_size: Tensor = None,
         initialization_method: str = "normal",
         pytorch_forecasting: bool = False,
-        explain: list = None,
-        XAI_metric: list = None,
-        random_model=None,
-        test_loader=None,
-        dataset=None,
         **kwargs,
     ):
         super().__init__(
@@ -428,55 +425,6 @@ def __init__(
             initialization_method=initialization_method,
             kwargs=kwargs,
         )
-        self.explain = explain
-        self.XAI_metric = XAI_metric
-        self.methods = {
-            "G": Saliency,
-            "L": Lime,
-            "IG": IntegratedGradients,
-            "FA": FeatureAblation,
-            "R": "Random",
-            "Att": "Attention"
-        }
-        self.random_model = random_model
-        self.test_loader = test_loader
-
-        if random_model is not None:
-            self.random_model = load_model(
-                model=self,
-                source_dir=random_model,
-                pl_model=True,
-                train_dataset=dataset,
-                optimizer=optimizer,
-            )
-        else:
-            self.random_model = None
-
-    def set_metrics(self, *args):
-        """Set the evaluation metrics for the prediction model."""
-
-        metrics = super().set_metrics(*args)
-
-        if self.XAI_metric is not None:
-            for metric in self.XAI_metric:
-                if metric == "Faithfulness":
-                    metrics.update(DLMetrics.Faithfulness)
-                elif metric == "Stability":
-                    metrics.update(DLMetrics.Stability)
-                elif metric == "Randmoization":
-                    metrics.update(DLMetrics.Randmoization)
-
-        return metrics
-
-    def on_train_start(self):
-        self.metrics = {
-            step_name: {
-                metric_name: (metric() if isinstance(metric, type) else metric)
-                for metric_name, metric in self.set_metrics().items()
-            }
-            for step_name in ["train", "val", "test"]
-        }
-        return super().on_train_start()
 
     def step_fn(self, element, step_prefix=""):
         """Perform a step in the DL prediction model training loop.
@@ -501,6 +449,7 @@ def step_fn(self, element, step_prefix=""):
         else:
             aux_loss = 0
         # Get prediction and target
+
         prediction = out.to(self.device).squeeze(-1)
 
         target = labels.to(self.device)
@@ -511,113 +460,14 @@ def step_fn(self, element, step_prefix=""):
             # Returns torch.long because negative log likelihood loss
         elif self.run_mode == RunMode.regression:
             # Regression task
+
             loss = self.loss(prediction[:, 0], target.float()) + aux_loss
         else:
             raise ValueError(f"Run mode {self.run_mode} not yet supported. Please implement it.")
         transformed_output = self.output_transform((prediction, target))
-        if self.explain is not None:
-
-            for method in self.explain:
-                print(method)
-                if method == "Attention":
-                    all_attrs, features_attrs, timestep_attrs = self.explantation2(dic, method, self.test_loader)
-                else:
-                    all_attrs, features_attrs, timestep_attrs = self.explantation2(dic, method)
-                for key, value in self.metrics[step_prefix].items():
-                    if key == "Faithfulness_timesteps":
-                        value.update(
-                            dic,
-                            timestep_attrs,
-                            self,
-                            correlation_pearson,
-                            100,
-                            "baseline",
-                            4,
-                            False,
-                            True,
-                            False
-
-                        )
-                    elif key == "Faithfulness_features":
-                        value.update(
-                            dic,
-                            features_attrs,
-                            self,
-                            correlation_pearson,
-                            100,
-                            "baseline",
-                            9,
-                            True,
-                            False,
-                            False
-
-                        )
-                    elif key == "Faithfulness_feature_timestep":
-                        value.update(
-                            dic,
-                            all_attrs,
-                            self,
-                            correlation_pearson,
-                            100,
-                            "baseline",
-                            [4, 9],
-                            False,
-                            False,
-                            True
-
-                        )
-
-                    elif key == "Stability":
-                        if method == "Attention":
-
-                            value.update(
-                                dic,
-                                timestep_attrs,
-                                self,
-                                method,
-                                self.test_loader,
-                                0.5,
-
-                            )
-                        else:
-                            value.update(
-                                dic,
-                                all_attrs,
-                                self,
-                                method,
-                                None,
-                                0.5,
-                            )
-                    elif key == "Randomization":
-
-                        if method == "Attention":
-                            value.update(
-                                dic,
-                                timestep_attrs,
-                                self,
-                                method,
-                                self.random_model,
-                                cosine,
-                                None,
-
-                            )
-                        else:
-
-                            value.update(
-                                dic,
-                                all_attrs,
-                                self,
-                                method,
-                                self.random_model,
-                                cosine,
-                                self.test_loader,
-
-                            )
 
         for key, value in self.metrics[step_prefix].items():
-            if (key == "Faithfulness_timesteps" or key == "Faithfulness_features" or key == "Faithfulness_feature_timestep"
-                    or key == "Stability" or key == "Randomization"):
-                continue
+
             if isinstance(value, torchmetrics.Metric):
                 if key == "Binary_Fairness":
                     feature_names = self.metrics[step_prefix][key].feature_helper(self.trainer, step_prefix)
@@ -727,122 +577,6 @@ def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir
         plt.tight_layout()
         plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
 
-    def explantation2(self, dic, method, dataloader=None, log_dir=".", plot=False, **kwargs):
-        """
-        Generic method to combine pytorchforecasting data loading , interpertations and captum to generate attributions
-
-        Args:
-            - dataloader: pytorchforecasting data loader 
-            - method: The explantation method chosen
-            - log_dir= The directory to output the plots
-            - plot= Determines if plots should be done or not
-            - XAI_metric=Determines if XAI metrics should be calculated or not
-        Returns:
-            - all_attrs : Attribtuons of features per timesteps
-            - features_attrs : Attribtuons of features averaged over timesteps
-            - timestep_attrs : Attribtuons of timesteps averaged over features
-            - f_ts_v_score: Faithfulness score for attribtuons of features per timesteps
-            - f_ts_score: Faithfulness score for attribtuons of timesteps averaged over features
-        """
-        # Initialize lists to store attribution values for all instances
-
-        all_attrs = []
-        method = self.methods[method]
-        for key, value in dic.items():
-            dic[key] = dic[key].to(self.device)
-
-        method_name = method if (method == "Random") or (method == "Attention") else (
-            method.__name__)
-        if (method_name == "Random") or (method_name == "Attention"):
-            if method_name == "Attention":
-                Interpertations = self.interpertations(dataloader=dataloader, log_dir=log_dir, plot=plot)
-                timestep_attrs = Interpertations["attention"]
-                features_attrs = Interpertations["static_variables"].tolist()
-                features_attrs.extend(Interpertations["encoder_variables"].tolist())
-
-            elif method_name == "Random":
-                # Generate random attributions for baseline comparison
-                all_attrs = np.random.normal(size=[64, 24, 53])
-                features_attrs = all_attrs.mean(axis=(1))
-                timestep_attrs = all_attrs.mean(axis=(2))
-
-            return all_attrs, features_attrs, timestep_attrs
-        # Loop through the dataloader to compute attributions for all instances
-
-        data, baselines = self.prep_data_captum(dic)
-
-        # Initialize the explanation method
-        self.train()
-        explanation = method(self.forward_captum, interpretable_model=SkLearnLasso(
-            alpha=0.4)) if method_name == 'Lime' else method(self.forward_captum)
-
-        # Calculate attributions using the selected method
-        if method is not Saliency:
-            if method_name == "IG":
-                attr = explanation.attribute(data, baselines=baselines, n_steps=50, **kwargs)
-            elif method_name == "L" or method_name == "FA":
-                # for Lime and feature ablation we need to define what is a feature we define each variable per timestep as a feature
-                shapes = [
-                    torch.Size([64, 24, 0]),
-                    torch.Size([64, 24, 53]),
-                    torch.Size([64, 24]),
-                    torch.Size([64]),
-                    torch.Size([64, 1, 0]),
-                    torch.Size([64, 1, 53]),
-                    torch.Size([64, 1]),
-                    torch.Size([64]),
-                    torch.Size([64, 1]),
-                    torch.Size([64, 1]),
-                    torch.Size([64, 2])
-                ]
-
-                # Create a default mask for non-targeted tensors
-                def create_default_mask(shape):
-                    if len(shape) == 3:
-                        return torch.zeros(shape[0], shape[1], max(1, shape[2]), dtype=torch.int32)
-                    elif len(shape) == 2:
-                        return torch.zeros(shape[0], max(1, shape[1]), dtype=torch.int32)
-                    else:  # len(shape) == 1
-                        return torch.zeros(shape[0], dtype=torch.int32)
-
-                # Create a feature mask for the second tensor that includes both features and timesteps
-                num_timesteps = shapes[1][1]
-                num_features = shapes[1][2]
-                feature_mask_second = torch.arange(num_timesteps * num_features).reshape(num_timesteps, num_features)
-                feature_mask_second = feature_mask_second.unsqueeze(0).repeat(shapes[1][0], 1, 1)
-
-                # Create a tuple of masks
-                feature_masks = tuple([create_default_mask(shape) if i !=
-                                      1 else feature_mask_second for i, shape in enumerate(shapes)])
-
-                attr = explanation.attribute(data, baselines=baselines, feature_mask=feature_masks, **kwargs)
-        else:
-            attr = explanation.attribute(data, **kwargs)
-        self.eval()
-        # Process and store the calculated attributions
-        stacked_attr = attr[1].cpu().detach().numpy() if method_name in [
-            'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
-        # aggregate over batch
-        # attr = np.mean(stacked_attr, axis=0)
-
-        # aggregate over all batches
-        # all_attrs = np.array(all_attrs).mean(axis=(0))
-        # aggregate over all timesteps
-        features_attrs = stacked_attr.mean(axis=(1))
-        # aggregate over all features
-        timestep_attrs = stacked_attr.mean(axis=(2))
-
-        if plot:
-            log_dir_plots = log_dir / 'plots'
-            if not (log_dir_plots.exists()):
-                log_dir_plots.mkdir(parents=True)
-            # Plot attributions for features and timesteps
-
-            self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots)
-
-        # Return computed attributions and metrics
-        return stacked_attr, features_attrs, timestep_attrs
-
     def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=False, random_model=None, test_dataset=None, **kwargs):
         """
         Generic method to combine pytorchforecasting data loading , interpertations and captum to generate attributions
@@ -962,10 +696,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                                                                pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
                 r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
                                explain_method=method, random_model=random_model, method_name=method_name))
-<<<<<<< HEAD
 
-=======
->>>>>>> parent of a39b5c4... added condition for ROS and RIS for attention
                 res1, res2 = self.Relative_Stability(x,
                                                      stacked_attr, explain_method=method, method_name=method_name,  dataloader=None, **kwargs
                                                      )
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 020238e4..969b4f4f 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -123,37 +123,23 @@ def build_parser() -> ArgumentParser:
         default=None,
         help="Number of samples to use for evaluation.",
     )
-    """ parser.add_argument(
+    parser.add_argument(
         "--explain",
         default=False,
         action=BOA,
         help="Provide explaintations for predictions.",
-    ) """
+    )
     parser.add_argument(
         "--pytorch-forecasting",
         default=False,
         action=BOA,
         help="use pytorch forecasting library ",
     )
-    """ parser.add_argument(
+    parser.add_argument(
         "--XAI_metric",
         default=False,
         action=BOA,
         help="Compare explantations ",
-    ) """
-    parser.add_argument(
-        "-ex",
-        "--explain",
-        nargs='+',  # '+' means one or more arguments
-        type=str,   # or any type you need
-        help="List of XAI methods to run"
-    )
-    parser.add_argument(
-        "-exm",
-        "--XAI_metric",
-        nargs='+',  # '+' means one or more arguments
-        type=str,   # or any type you need
-        help="List of XAI metrics to run"
     )
     parser.add_argument(
         "--random_labels",

From 21bf3513c4ce16e3f51114804f9c30029ccb2343 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 28 Dec 2023 16:27:55 +0100
Subject: [PATCH 127/142] added choice of normalizer

---
 configs/prediction_models/DeepARpytorch.gin | 66 +++++++++++++++++---
 configs/prediction_models/RNNpytorch.gin    | 68 ++++++++++++++++++---
 configs/prediction_models/TFTpytorch.gin    |  1 +
 environment.yml                             |  2 +
 icu_benchmarks/data/loader.py               | 15 +++--
 icu_benchmarks/models/custom_metrics.py     |  2 +-
 icu_benchmarks/models/wrappers.py           |  5 +-
 7 files changed, 136 insertions(+), 23 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index bcc6abf6..9407222a 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -29,16 +29,63 @@ train_common/hyperparameter.gradient_clip_val=0.01
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
-PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_1', 'alp_lag_1', 'alt_lag_1', 'ast_lag_1', 'be_lag_1', 'bicar_lag_1', 'bili_lag_1', 'bili_dir_lag_1', 'bnd_lag_1', 'bun_lag_1', 'ca_lag_1', 'cai_lag_1', 'ck_lag_1', 'ckmb_lag_1', 'cl_lag_1', 'crea_lag_1', 'crp_lag_1', 'dbp_lag_1', 'fgn_lag_1', 'fio2_lag_1', 'glu_lag_1', 'hgb_lag_1', 'hr_lag_1', 'inr_pt_lag_1', 'k_lag_1', 'lact_lag_1', 'lymph_lag_1', 'map_lag_1', 'mch_lag_1', 'mchc_lag_1', 'mcv_lag_1', 'methb_lag_1', 'mg_lag_1', 'na_lag_1', 'neut_lag_1', 'o2sat_lag_1', 'pco2_lag_1', 'ph_lag_1', 'phos_lag_1', 'plt_lag_1', 'po2_lag_1', 'ptt_lag_1', 'resp_lag_1', 'sbp_lag_1', 'temp_lag_1', 'tnt_lag_1', 'urine_lag_1', 'wbc_lag_1', 'alb_lag_2', 'alp_lag_2', 'alt_lag_2', 'ast_lag_2', 'be_lag_2', 'bicar_lag_2', 'bili_lag_2', 'bili_dir_lag_2', 'bnd_lag_2', 'bun_lag_2', 'ca_lag_2', 'cai_lag_2', 'ck_lag_2', 'ckmb_lag_2', 'cl_lag_2', 'crea_lag_2', 'crp_lag_2', 'dbp_lag_2', 'fgn_lag_2', 'fio2_lag_2', 'glu_lag_2', 'hgb_lag_2', 'hr_lag_2', 'inr_pt_lag_2', 'k_lag_2', 'lact_lag_2', 'lymph_lag_2', 'map_lag_2', 'mch_lag_2', 'mchc_lag_2', 'mcv_lag_2', 'methb_lag_2', 'mg_lag_2', 'na_lag_2', 'neut_lag_2', 'o2sat_lag_2', 'pco2_lag_2', 'ph_lag_2', 'phos_lag_2', 'plt_lag_2', 'po2_lag_2', 'ptt_lag_2', 'resp_lag_2', 'sbp_lag_2', 'temp_lag_2', 'tnt_lag_2', 'urine_lag_2', 'wbc_lag_2', 'alb_lag_3', 'alp_lag_3', 'alt_lag_3', 'ast_lag_3', 'be_lag_3', 'bicar_lag_3', 'bili_lag_3', 'bili_dir_lag_3', 'bnd_lag_3', 'bun_lag_3', 'ca_lag_3', 'cai_lag_3', 'ck_lag_3', 'ckmb_lag_3', 'cl_lag_3', 'crea_lag_3', 'crp_lag_3', 'dbp_lag_3', 'fgn_lag_3', 'fio2_lag_3', 'glu_lag_3', 'hgb_lag_3', 'hr_lag_3', 'inr_pt_lag_3', 'k_lag_3', 'lact_lag_3', 'lymph_lag_3', 'map_lag_3', 'mch_lag_3', 'mchc_lag_3', 'mcv_lag_3', 'methb_lag_3', 'mg_lag_3', 'na_lag_3', 'neut_lag_3', 'o2sat_lag_3', 'pco2_lag_3', 'ph_lag_3', 'phos_lag_3', 'plt_lag_3', 'po2_lag_3', 'ptt_lag_3', 'resp_lag_3', 'sbp_lag_3', 'temp_lag_3', 'tnt_lag_3', 'urine_lag_3', 'wbc_lag_3', 'alb_lag_4', 'alp_lag_4', 'alt_lag_4', 'ast_lag_4', 'be_lag_4', 'bicar_lag_4', 'bili_lag_4', 'bili_dir_lag_4', 'bnd_lag_4', 'bun_lag_4', 'ca_lag_4', 'cai_lag_4', 'ck_lag_4', 'ckmb_lag_4', 'cl_lag_4', 'crea_lag_4', 'crp_lag_4', 'dbp_lag_4', 'fgn_lag_4', 'fio2_lag_4', 'glu_lag_4', 'hgb_lag_4', 'hr_lag_4', 'inr_pt_lag_4', 'k_lag_4', 'lact_lag_4', 'lymph_lag_4', 'map_lag_4', 'mch_lag_4', 'mchc_lag_4', 'mcv_lag_4', 'methb_lag_4', 'mg_lag_4', 'na_lag_4', 'neut_lag_4', 'o2sat_lag_4', 'pco2_lag_4', 'ph_lag_4', 'phos_lag_4', 'plt_lag_4', 'po2_lag_4', 'ptt_lag_4', 'resp_lag_4', 'sbp_lag_4', 'temp_lag_4', 'tnt_lag_4', 'urine_lag_4', 'wbc_lag_4', 'alb_lag_5', 'alp_lag_5', 'alt_lag_5', 'ast_lag_5', 'be_lag_5', 'bicar_lag_5', 'bili_lag_5', 'bili_dir_lag_5', 'bnd_lag_5', 'bun_lag_5', 'ca_lag_5', 'cai_lag_5', 'ck_lag_5', 'ckmb_lag_5', 'cl_lag_5', 'crea_lag_5', 'crp_lag_5', 'dbp_lag_5', 'fgn_lag_5', 'fio2_lag_5', 'glu_lag_5', 'hgb_lag_5', 'hr_lag_5', 'inr_pt_lag_5', 'k_lag_5', 'lact_lag_5', 'lymph_lag_5', 'map_lag_5', 'mch_lag_5', 'mchc_lag_5', 'mcv_lag_5', 'methb_lag_5', 'mg_lag_5', 'na_lag_5', 'neut_lag_5', 'o2sat_lag_5', 'pco2_lag_5', 'ph_lag_5', 'phos_lag_5', 'plt_lag_5', 'po2_lag_5', 'ptt_lag_5', 'resp_lag_5', 'sbp_lag_5', 'temp_lag_5', 'tnt_lag_5', 'urine_lag_5', 'wbc_lag_5', 'alb_lag_6', 'alp_lag_6', 'alt_lag_6', 'ast_lag_6', 'be_lag_6', 'bicar_lag_6', 'bili_lag_6', 'bili_dir_lag_6', 'bnd_lag_6', 'bun_lag_6', 'ca_lag_6', 'cai_lag_6', 'ck_lag_6', 'ckmb_lag_6', 'cl_lag_6', 'crea_lag_6', 'crp_lag_6', 'dbp_lag_6', 'fgn_lag_6', 'fio2_lag_6', 'glu_lag_6', 'hgb_lag_6', 'hr_lag_6', 'inr_pt_lag_6', 'k_lag_6', 'lact_lag_6', 'lymph_lag_6', 'map_lag_6', 'mch_lag_6', 'mchc_lag_6', 'mcv_lag_6', 'methb_lag_6', 'mg_lag_6', 'na_lag_6', 'neut_lag_6', 'o2sat_lag_6', 'pco2_lag_6', 'ph_lag_6', 'phos_lag_6', 'plt_lag_6', 'po2_lag_6', 'ptt_lag_6', 'resp_lag_6', 'sbp_lag_6', 'temp_lag_6', 'tnt_lag_6', 'urine_lag_6', 'wbc_lag_6', 'alb_lag_7', 'alp_lag_7', 'alt_lag_7', 'ast_lag_7', 'be_lag_7', 'bicar_lag_7', 'bili_lag_7', 'bili_dir_lag_7', 'bnd_lag_7', 'bun_lag_7', 'ca_lag_7', 'cai_lag_7', 'ck_lag_7', 'ckmb_lag_7', 'cl_lag_7', 'crea_lag_7', 'crp_lag_7', 'dbp_lag_7', 'fgn_lag_7', 'fio2_lag_7', 'glu_lag_7', 'hgb_lag_7', 'hr_lag_7', 'inr_pt_lag_7', 'k_lag_7', 'lact_lag_7', 'lymph_lag_7', 'map_lag_7', 'mch_lag_7', 'mchc_lag_7', 'mcv_lag_7', 'methb_lag_7', 'mg_lag_7', 'na_lag_7', 'neut_lag_7', 'o2sat_lag_7', 'pco2_lag_7', 'ph_lag_7', 'phos_lag_7', 'plt_lag_7', 'po2_lag_7', 'ptt_lag_7', 'resp_lag_7', 'sbp_lag_7', 'temp_lag_7', 'tnt_lag_7', 'urine_lag_7', 'wbc_lag_7', 'alb_lag_8', 'alp_lag_8', 'alt_lag_8', 'ast_lag_8', 'be_lag_8', 'bicar_lag_8', 'bili_lag_8', 'bili_dir_lag_8', 'bnd_lag_8', 'bun_lag_8', 'ca_lag_8', 'cai_lag_8', 'ck_lag_8', 'ckmb_lag_8', 'cl_lag_8', 'crea_lag_8', 'crp_lag_8', 'dbp_lag_8', 'fgn_lag_8', 'fio2_lag_8', 'glu_lag_8', 'hgb_lag_8', 'hr_lag_8', 'inr_pt_lag_8', 'k_lag_8', 'lact_lag_8', 'lymph_lag_8', 'map_lag_8', 'mch_lag_8', 'mchc_lag_8', 'mcv_lag_8', 'methb_lag_8', 'mg_lag_8', 'na_lag_8', 'neut_lag_8', 'o2sat_lag_8', 'pco2_lag_8', 'ph_lag_8', 'phos_lag_8', 'plt_lag_8', 'po2_lag_8', 'ptt_lag_8', 'resp_lag_8', 'sbp_lag_8', 'temp_lag_8', 'tnt_lag_8', 'urine_lag_8', 'wbc_lag_8', 'alb_lag_9', 'alp_lag_9', 'alt_lag_9', 'ast_lag_9', 'be_lag_9', 'bicar_lag_9', 'bili_lag_9', 'bili_dir_lag_9', 'bnd_lag_9', 'bun_lag_9', 'ca_lag_9', 'cai_lag_9', 'ck_lag_9', 'ckmb_lag_9', 'cl_lag_9', 'crea_lag_9', 'crp_lag_9', 'dbp_lag_9', 'fgn_lag_9', 'fio2_lag_9', 'glu_lag_9', 'hgb_lag_9', 'hr_lag_9', 'inr_pt_lag_9', 'k_lag_9', 'lact_lag_9', 'lymph_lag_9', 'map_lag_9', 'mch_lag_9', 'mchc_lag_9', 'mcv_lag_9', 'methb_lag_9', 'mg_lag_9', 'na_lag_9', 'neut_lag_9', 'o2sat_lag_9', 'pco2_lag_9', 'ph_lag_9', 'phos_lag_9', 'plt_lag_9', 'po2_lag_9', 'ptt_lag_9', 'resp_lag_9', 'sbp_lag_9', 'temp_lag_9', 'tnt_lag_9', 'urine_lag_9', 'wbc_lag_9', 'alb_lag_10', 'alp_lag_10', 'alt_lag_10', 'ast_lag_10', 'be_lag_10', 'bicar_lag_10', 'bili_lag_10', 'bili_dir_lag_10', 'bnd_lag_10', 'bun_lag_10', 'ca_lag_10', 'cai_lag_10', 'ck_lag_10', 'ckmb_lag_10', 'cl_lag_10', 'crea_lag_10', 'crp_lag_10', 'dbp_lag_10', 'fgn_lag_10', 'fio2_lag_10', 'glu_lag_10', 'hgb_lag_10', 'hr_lag_10', 'inr_pt_lag_10', 'k_lag_10', 'lact_lag_10', 'lymph_lag_10', 'map_lag_10', 'mch_lag_10', 'mchc_lag_10', 'mcv_lag_10', 'methb_lag_10', 'mg_lag_10', 'na_lag_10', 'neut_lag_10', 'o2sat_lag_10', 'pco2_lag_10', 'ph_lag_10', 'phos_lag_10', 'plt_lag_10', 'po2_lag_10', 'ptt_lag_10', 'resp_lag_10', 'sbp_lag_10', 'temp_lag_10', 'tnt_lag_10', 'urine_lag_10', 'wbc_lag_10', 'alb_lag_11', 'alp_lag_11', 'alt_lag_11', 'ast_lag_11', 'be_lag_11', 'bicar_lag_11', 'bili_lag_11', 'bili_dir_lag_11', 'bnd_lag_11', 'bun_lag_11', 'ca_lag_11', 'cai_lag_11', 'ck_lag_11', 'ckmb_lag_11', 'cl_lag_11', 'crea_lag_11', 'crp_lag_11', 'dbp_lag_11', 'fgn_lag_11', 'fio2_lag_11', 'glu_lag_11', 'hgb_lag_11', 'hr_lag_11', 'inr_pt_lag_11', 'k_lag_11', 'lact_lag_11', 'lymph_lag_11', 'map_lag_11', 'mch_lag_11', 'mchc_lag_11', 'mcv_lag_11', 'methb_lag_11', 'mg_lag_11', 'na_lag_11', 'neut_lag_11', 'o2sat_lag_11', 'pco2_lag_11', 'ph_lag_11', 'phos_lag_11', 'plt_lag_11', 'po2_lag_11', 'ptt_lag_11', 'resp_lag_11', 'sbp_lag_11', 'temp_lag_11', 'tnt_lag_11', 'urine_lag_11', 'wbc_lag_11', 'alb_lag_12', 'alp_lag_12', 'alt_lag_12', 'ast_lag_12', 'be_lag_12', 'bicar_lag_12', 'bili_lag_12', 'bili_dir_lag_12', 'bnd_lag_12', 'bun_lag_12', 'ca_lag_12', 'cai_lag_12', 'ck_lag_12', 'ckmb_lag_12', 'cl_lag_12', 'crea_lag_12', 'crp_lag_12', 'dbp_lag_12', 'fgn_lag_12', 'fio2_lag_12', 'glu_lag_12', 'hgb_lag_12', 'hr_lag_12', 'inr_pt_lag_12', 'k_lag_12', 'lact_lag_12', 'lymph_lag_12', 'map_lag_12', 'mch_lag_12', 'mchc_lag_12', 'mcv_lag_12', 'methb_lag_12', 'mg_lag_12', 'na_lag_12', 'neut_lag_12', 'o2sat_lag_12', 'pco2_lag_12', 'ph_lag_12', 'phos_lag_12', 'plt_lag_12', 'po2_lag_12', 'ptt_lag_12', 'resp_lag_12', 'sbp_lag_12', 'temp_lag_12', 'tnt_lag_12', 'urine_lag_12', 'wbc_lag_12', 'alb_lag_13', 'alp_lag_13', 'alt_lag_13', 'ast_lag_13', 'be_lag_13', 'bicar_lag_13', 'bili_lag_13', 'bili_dir_lag_13', 'bnd_lag_13', 'bun_lag_13', 'ca_lag_13', 'cai_lag_13', 'ck_lag_13', 'ckmb_lag_13', 'cl_lag_13', 'crea_lag_13', 'crp_lag_13', 'dbp_lag_13', 'fgn_lag_13', 'fio2_lag_13', 'glu_lag_13', 'hgb_lag_13', 'hr_lag_13', 'inr_pt_lag_13', 'k_lag_13', 'lact_lag_13', 'lymph_lag_13', 'map_lag_13', 'mch_lag_13', 'mchc_lag_13', 'mcv_lag_13', 'methb_lag_13', 'mg_lag_13', 'na_lag_13', 'neut_lag_13', 'o2sat_lag_13', 'pco2_lag_13', 'ph_lag_13', 'phos_lag_13', 'plt_lag_13', 'po2_lag_13', 'ptt_lag_13', 'resp_lag_13', 'sbp_lag_13', 'temp_lag_13', 'tnt_lag_13', 'urine_lag_13', 'wbc_lag_13', 'alb_lag_14', 'alp_lag_14', 'alt_lag_14', 'ast_lag_14', 'be_lag_14', 'bicar_lag_14', 'bili_lag_14', 'bili_dir_lag_14', 'bnd_lag_14', 'bun_lag_14', 'ca_lag_14', 'cai_lag_14', 'ck_lag_14', 'ckmb_lag_14', 'cl_lag_14', 'crea_lag_14', 'crp_lag_14', 'dbp_lag_14', 'fgn_lag_14', 'fio2_lag_14', 'glu_lag_14', 'hgb_lag_14', 'hr_lag_14', 'inr_pt_lag_14', 'k_lag_14', 'lact_lag_14', 'lymph_lag_14', 'map_lag_14', 'mch_lag_14', 'mchc_lag_14', 'mcv_lag_14', 'methb_lag_14', 'mg_lag_14', 'na_lag_14', 'neut_lag_14', 'o2sat_lag_14', 'pco2_lag_14', 'ph_lag_14', 'phos_lag_14', 'plt_lag_14', 'po2_lag_14', 'ptt_lag_14', 'resp_lag_14', 'sbp_lag_14', 'temp_lag_14', 'tnt_lag_14', 'urine_lag_14', 'wbc_lag_14', 'alb_lag_15', 'alp_lag_15', 'alt_lag_15', 'ast_lag_15', 'be_lag_15', 'bicar_lag_15', 'bili_lag_15', 'bili_dir_lag_15', 'bnd_lag_15', 'bun_lag_15', 'ca_lag_15', 'cai_lag_15', 'ck_lag_15', 'ckmb_lag_15', 'cl_lag_15', 'crea_lag_15', 'crp_lag_15', 'dbp_lag_15', 'fgn_lag_15', 'fio2_lag_15', 'glu_lag_15', 'hgb_lag_15', 'hr_lag_15', 'inr_pt_lag_15', 'k_lag_15', 'lact_lag_15', 'lymph_lag_15', 'map_lag_15', 'mch_lag_15', 'mchc_lag_15', 'mcv_lag_15', 'methb_lag_15', 'mg_lag_15', 'na_lag_15', 'neut_lag_15', 'o2sat_lag_15', 'pco2_lag_15', 'ph_lag_15', 'phos_lag_15', 'plt_lag_15', 'po2_lag_15', 'ptt_lag_15', 'resp_lag_15', 'sbp_lag_15', 'temp_lag_15', 'tnt_lag_15', 'urine_lag_15', 'wbc_lag_15', 'alb_lag_16', 'alp_lag_16', 'alt_lag_16', 'ast_lag_16', 'be_lag_16', 'bicar_lag_16', 'bili_lag_16', 'bili_dir_lag_16', 'bnd_lag_16', 'bun_lag_16', 'ca_lag_16', 'cai_lag_16', 'ck_lag_16', 'ckmb_lag_16', 'cl_lag_16', 'crea_lag_16', 'crp_lag_16', 'dbp_lag_16', 'fgn_lag_16', 'fio2_lag_16', 'glu_lag_16', 'hgb_lag_16', 'hr_lag_16', 'inr_pt_lag_16', 'k_lag_16', 'lact_lag_16', 'lymph_lag_16', 'map_lag_16', 'mch_lag_16', 'mchc_lag_16', 'mcv_lag_16', 'methb_lag_16', 'mg_lag_16', 'na_lag_16', 'neut_lag_16', 'o2sat_lag_16', 'pco2_lag_16', 'ph_lag_16', 'phos_lag_16', 'plt_lag_16', 'po2_lag_16', 'ptt_lag_16', 'resp_lag_16', 'sbp_lag_16', 'temp_lag_16', 'tnt_lag_16', 'urine_lag_16', 'wbc_lag_16', 'alb_lag_17', 'alp_lag_17', 'alt_lag_17', 'ast_lag_17', 'be_lag_17', 'bicar_lag_17', 'bili_lag_17', 'bili_dir_lag_17', 'bnd_lag_17', 'bun_lag_17', 'ca_lag_17', 'cai_lag_17', 'ck_lag_17', 'ckmb_lag_17', 'cl_lag_17', 'crea_lag_17', 'crp_lag_17', 'dbp_lag_17', 'fgn_lag_17', 'fio2_lag_17', 'glu_lag_17', 'hgb_lag_17', 'hr_lag_17', 'inr_pt_lag_17', 'k_lag_17', 'lact_lag_17', 'lymph_lag_17', 'map_lag_17', 'mch_lag_17', 'mchc_lag_17', 'mcv_lag_17', 'methb_lag_17', 'mg_lag_17', 'na_lag_17', 'neut_lag_17', 'o2sat_lag_17', 'pco2_lag_17', 'ph_lag_17', 'phos_lag_17', 'plt_lag_17', 'po2_lag_17', 'ptt_lag_17', 'resp_lag_17', 'sbp_lag_17', 'temp_lag_17', 'tnt_lag_17', 'urine_lag_17', 'wbc_lag_17', 'alb_lag_18', 'alp_lag_18', 'alt_lag_18', 'ast_lag_18', 'be_lag_18', 'bicar_lag_18', 'bili_lag_18', 'bili_dir_lag_18', 'bnd_lag_18', 'bun_lag_18', 'ca_lag_18', 'cai_lag_18', 'ck_lag_18', 'ckmb_lag_18', 'cl_lag_18', 'crea_lag_18', 'crp_lag_18', 'dbp_lag_18', 'fgn_lag_18', 'fio2_lag_18', 'glu_lag_18', 'hgb_lag_18', 'hr_lag_18', 'inr_pt_lag_18', 'k_lag_18', 'lact_lag_18', 'lymph_lag_18', 'map_lag_18', 'mch_lag_18', 'mchc_lag_18', 'mcv_lag_18', 'methb_lag_18', 'mg_lag_18', 'na_lag_18', 'neut_lag_18', 'o2sat_lag_18', 'pco2_lag_18', 'ph_lag_18', 'phos_lag_18', 'plt_lag_18', 'po2_lag_18', 'ptt_lag_18', 'resp_lag_18', 'sbp_lag_18', 'temp_lag_18', 'tnt_lag_18', 'urine_lag_18', 'wbc_lag_18', 'alb_lag_19', 'alp_lag_19', 'alt_lag_19', 'ast_lag_19', 'be_lag_19', 'bicar_lag_19', 'bili_lag_19', 'bili_dir_lag_19', 'bnd_lag_19', 'bun_lag_19', 'ca_lag_19', 'cai_lag_19', 'ck_lag_19', 'ckmb_lag_19', 'cl_lag_19', 'crea_lag_19', 'crp_lag_19', 'dbp_lag_19', 'fgn_lag_19', 'fio2_lag_19', 'glu_lag_19', 'hgb_lag_19', 'hr_lag_19', 'inr_pt_lag_19', 'k_lag_19', 'lact_lag_19', 'lymph_lag_19', 'map_lag_19', 'mch_lag_19', 'mchc_lag_19', 'mcv_lag_19', 'methb_lag_19', 'mg_lag_19', 'na_lag_19', 'neut_lag_19', 'o2sat_lag_19', 'pco2_lag_19', 'ph_lag_19', 'phos_lag_19', 'plt_lag_19', 'po2_lag_19', 'ptt_lag_19', 'resp_lag_19', 'sbp_lag_19', 'temp_lag_19', 'tnt_lag_19', 'urine_lag_19', 'wbc_lag_19', 'alb_lag_20', 'alp_lag_20', 'alt_lag_20', 'ast_lag_20', 'be_lag_20', 'bicar_lag_20', 'bili_lag_20', 'bili_dir_lag_20', 'bnd_lag_20', 'bun_lag_20', 'ca_lag_20', 'cai_lag_20', 'ck_lag_20', 'ckmb_lag_20', 'cl_lag_20', 'crea_lag_20', 'crp_lag_20', 'dbp_lag_20', 'fgn_lag_20', 'fio2_lag_20', 'glu_lag_20', 'hgb_lag_20', 'hr_lag_20', 'inr_pt_lag_20', 'k_lag_20', 'lact_lag_20', 'lymph_lag_20', 'map_lag_20', 'mch_lag_20', 'mchc_lag_20', 'mcv_lag_20', 'methb_lag_20', 'mg_lag_20', 'na_lag_20', 'neut_lag_20', 'o2sat_lag_20', 'pco2_lag_20', 'ph_lag_20', 'phos_lag_20', 'plt_lag_20', 'po2_lag_20', 'ptt_lag_20', 'resp_lag_20', 'sbp_lag_20', 'temp_lag_20', 'tnt_lag_20', 'urine_lag_20', 'wbc_lag_20', 'alb_lag_21', 'alp_lag_21', 'alt_lag_21', 'ast_lag_21', 'be_lag_21', 'bicar_lag_21', 'bili_lag_21', 'bili_dir_lag_21', 'bnd_lag_21', 'bun_lag_21', 'ca_lag_21', 'cai_lag_21', 'ck_lag_21', 'ckmb_lag_21', 'cl_lag_21', 'crea_lag_21', 'crp_lag_21', 'dbp_lag_21', 'fgn_lag_21', 'fio2_lag_21', 'glu_lag_21', 'hgb_lag_21', 'hr_lag_21', 'inr_pt_lag_21', 'k_lag_21', 'lact_lag_21', 'lymph_lag_21', 'map_lag_21', 'mch_lag_21', 'mchc_lag_21', 'mcv_lag_21', 'methb_lag_21', 'mg_lag_21', 'na_lag_21', 'neut_lag_21', 'o2sat_lag_21', 'pco2_lag_21', 'ph_lag_21', 'phos_lag_21', 'plt_lag_21', 'po2_lag_21', 'ptt_lag_21', 'resp_lag_21', 'sbp_lag_21', 'temp_lag_21', 'tnt_lag_21', 'urine_lag_21', 'wbc_lag_21', 'alb_lag_22', 'alp_lag_22', 'alt_lag_22', 'ast_lag_22', 'be_lag_22', 'bicar_lag_22', 'bili_lag_22', 'bili_dir_lag_22', 'bnd_lag_22', 'bun_lag_22', 'ca_lag_22', 'cai_lag_22', 'ck_lag_22', 'ckmb_lag_22', 'cl_lag_22', 'crea_lag_22', 'crp_lag_22', 'dbp_lag_22', 'fgn_lag_22', 'fio2_lag_22', 'glu_lag_22', 'hgb_lag_22', 'hr_lag_22', 'inr_pt_lag_22', 'k_lag_22', 'lact_lag_22', 'lymph_lag_22', 'map_lag_22', 'mch_lag_22', 'mchc_lag_22', 'mcv_lag_22', 'methb_lag_22', 'mg_lag_22', 'na_lag_22', 'neut_lag_22', 'o2sat_lag_22', 'pco2_lag_22', 'ph_lag_22', 'phos_lag_22', 'plt_lag_22', 'po2_lag_22', 'ptt_lag_22', 'resp_lag_22', 'sbp_lag_22', 'temp_lag_22', 'tnt_lag_22', 'urine_lag_22', 'wbc_lag_22', 'alb_lag_23', 'alp_lag_23', 'alt_lag_23', 'ast_lag_23', 'be_lag_23', 'bicar_lag_23', 'bili_lag_23', 'bili_dir_lag_23', 'bnd_lag_23', 'bun_lag_23', 'ca_lag_23', 'cai_lag_23', 'ck_lag_23', 'ckmb_lag_23', 'cl_lag_23', 'crea_lag_23', 'crp_lag_23', 'dbp_lag_23', 'fgn_lag_23', 'fio2_lag_23', 'glu_lag_23', 'hgb_lag_23', 'hr_lag_23', 'inr_pt_lag_23', 'k_lag_23', 'lact_lag_23', 'lymph_lag_23', 'map_lag_23', 'mch_lag_23', 'mchc_lag_23', 'mcv_lag_23', 'methb_lag_23', 'mg_lag_23', 'na_lag_23', 'neut_lag_23', 'o2sat_lag_23', 'pco2_lag_23', 'ph_lag_23', 'phos_lag_23', 'plt_lag_23', 'po2_lag_23', 'ptt_lag_23', 'resp_lag_23', 'sbp_lag_23', 'temp_lag_23', 'tnt_lag_23', 'urine_lag_23', 'wbc_lag_23', 'alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24' ]
+PredictionDatasetpytorch.time_varying_known_reals=[]
 PredictionDatasetpytorch.add_relative_time_idx=False
-PredictionDatasetpytorch.target="label"
-            
-PredictionDatasetpytorch.time_varying_unknown_reals=[
+PredictionDatasetpytorch.target=[
+    "alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
                 "label",
               
             ]
-PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetpytorch.lagged_variables=[  "alb",
+            
+PredictionDatasetpytorch.time_varying_unknown_reals=[
+    "alb",
                 "alp",
                 "alt",
                 "ast",
@@ -85,6 +132,11 @@ PredictionDatasetpytorch.lagged_variables=[  "alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc"]
+                "wbc",
+                "label",
+              
+            ]
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.lagged_variables=[]
 
                 
\ No newline at end of file
diff --git a/configs/prediction_models/RNNpytorch.gin b/configs/prediction_models/RNNpytorch.gin
index 6c0f2e15..8f8132e0 100644
--- a/configs/prediction_models/RNNpytorch.gin
+++ b/configs/prediction_models/RNNpytorch.gin
@@ -30,17 +30,63 @@ train_common/hyperparameter.gradient_clip_val=100.0
 # Dataset params
 PredictionDatasetpytorch.max_encoder_length = 24
 PredictionDatasetpytorch.max_prediction_length = 1
-PredictionDatasetpytorch.time_varying_known_reals=['alb_lag_1', 'alp_lag_1', 'alt_lag_1', 'ast_lag_1', 'be_lag_1', 'bicar_lag_1', 'bili_lag_1', 'bili_dir_lag_1', 'bnd_lag_1', 'bun_lag_1', 'ca_lag_1', 'cai_lag_1', 'ck_lag_1', 'ckmb_lag_1', 'cl_lag_1', 'crea_lag_1', 'crp_lag_1', 'dbp_lag_1', 'fgn_lag_1', 'fio2_lag_1', 'glu_lag_1', 'hgb_lag_1', 'hr_lag_1', 'inr_pt_lag_1', 'k_lag_1', 'lact_lag_1', 'lymph_lag_1', 'map_lag_1', 'mch_lag_1', 'mchc_lag_1', 'mcv_lag_1', 'methb_lag_1', 'mg_lag_1', 'na_lag_1', 'neut_lag_1', 'o2sat_lag_1', 'pco2_lag_1', 'ph_lag_1', 'phos_lag_1', 'plt_lag_1', 'po2_lag_1', 'ptt_lag_1', 'resp_lag_1', 'sbp_lag_1', 'temp_lag_1', 'tnt_lag_1', 'urine_lag_1', 'wbc_lag_1', 'alb_lag_2', 'alp_lag_2', 'alt_lag_2', 'ast_lag_2', 'be_lag_2', 'bicar_lag_2', 'bili_lag_2', 'bili_dir_lag_2', 'bnd_lag_2', 'bun_lag_2', 'ca_lag_2', 'cai_lag_2', 'ck_lag_2', 'ckmb_lag_2', 'cl_lag_2', 'crea_lag_2', 'crp_lag_2', 'dbp_lag_2', 'fgn_lag_2', 'fio2_lag_2', 'glu_lag_2', 'hgb_lag_2', 'hr_lag_2', 'inr_pt_lag_2', 'k_lag_2', 'lact_lag_2', 'lymph_lag_2', 'map_lag_2', 'mch_lag_2', 'mchc_lag_2', 'mcv_lag_2', 'methb_lag_2', 'mg_lag_2', 'na_lag_2', 'neut_lag_2', 'o2sat_lag_2', 'pco2_lag_2', 'ph_lag_2', 'phos_lag_2', 'plt_lag_2', 'po2_lag_2', 'ptt_lag_2', 'resp_lag_2', 'sbp_lag_2', 'temp_lag_2', 'tnt_lag_2', 'urine_lag_2', 'wbc_lag_2', 'alb_lag_3', 'alp_lag_3', 'alt_lag_3', 'ast_lag_3', 'be_lag_3', 'bicar_lag_3', 'bili_lag_3', 'bili_dir_lag_3', 'bnd_lag_3', 'bun_lag_3', 'ca_lag_3', 'cai_lag_3', 'ck_lag_3', 'ckmb_lag_3', 'cl_lag_3', 'crea_lag_3', 'crp_lag_3', 'dbp_lag_3', 'fgn_lag_3', 'fio2_lag_3', 'glu_lag_3', 'hgb_lag_3', 'hr_lag_3', 'inr_pt_lag_3', 'k_lag_3', 'lact_lag_3', 'lymph_lag_3', 'map_lag_3', 'mch_lag_3', 'mchc_lag_3', 'mcv_lag_3', 'methb_lag_3', 'mg_lag_3', 'na_lag_3', 'neut_lag_3', 'o2sat_lag_3', 'pco2_lag_3', 'ph_lag_3', 'phos_lag_3', 'plt_lag_3', 'po2_lag_3', 'ptt_lag_3', 'resp_lag_3', 'sbp_lag_3', 'temp_lag_3', 'tnt_lag_3', 'urine_lag_3', 'wbc_lag_3', 'alb_lag_4', 'alp_lag_4', 'alt_lag_4', 'ast_lag_4', 'be_lag_4', 'bicar_lag_4', 'bili_lag_4', 'bili_dir_lag_4', 'bnd_lag_4', 'bun_lag_4', 'ca_lag_4', 'cai_lag_4', 'ck_lag_4', 'ckmb_lag_4', 'cl_lag_4', 'crea_lag_4', 'crp_lag_4', 'dbp_lag_4', 'fgn_lag_4', 'fio2_lag_4', 'glu_lag_4', 'hgb_lag_4', 'hr_lag_4', 'inr_pt_lag_4', 'k_lag_4', 'lact_lag_4', 'lymph_lag_4', 'map_lag_4', 'mch_lag_4', 'mchc_lag_4', 'mcv_lag_4', 'methb_lag_4', 'mg_lag_4', 'na_lag_4', 'neut_lag_4', 'o2sat_lag_4', 'pco2_lag_4', 'ph_lag_4', 'phos_lag_4', 'plt_lag_4', 'po2_lag_4', 'ptt_lag_4', 'resp_lag_4', 'sbp_lag_4', 'temp_lag_4', 'tnt_lag_4', 'urine_lag_4', 'wbc_lag_4', 'alb_lag_5', 'alp_lag_5', 'alt_lag_5', 'ast_lag_5', 'be_lag_5', 'bicar_lag_5', 'bili_lag_5', 'bili_dir_lag_5', 'bnd_lag_5', 'bun_lag_5', 'ca_lag_5', 'cai_lag_5', 'ck_lag_5', 'ckmb_lag_5', 'cl_lag_5', 'crea_lag_5', 'crp_lag_5', 'dbp_lag_5', 'fgn_lag_5', 'fio2_lag_5', 'glu_lag_5', 'hgb_lag_5', 'hr_lag_5', 'inr_pt_lag_5', 'k_lag_5', 'lact_lag_5', 'lymph_lag_5', 'map_lag_5', 'mch_lag_5', 'mchc_lag_5', 'mcv_lag_5', 'methb_lag_5', 'mg_lag_5', 'na_lag_5', 'neut_lag_5', 'o2sat_lag_5', 'pco2_lag_5', 'ph_lag_5', 'phos_lag_5', 'plt_lag_5', 'po2_lag_5', 'ptt_lag_5', 'resp_lag_5', 'sbp_lag_5', 'temp_lag_5', 'tnt_lag_5', 'urine_lag_5', 'wbc_lag_5', 'alb_lag_6', 'alp_lag_6', 'alt_lag_6', 'ast_lag_6', 'be_lag_6', 'bicar_lag_6', 'bili_lag_6', 'bili_dir_lag_6', 'bnd_lag_6', 'bun_lag_6', 'ca_lag_6', 'cai_lag_6', 'ck_lag_6', 'ckmb_lag_6', 'cl_lag_6', 'crea_lag_6', 'crp_lag_6', 'dbp_lag_6', 'fgn_lag_6', 'fio2_lag_6', 'glu_lag_6', 'hgb_lag_6', 'hr_lag_6', 'inr_pt_lag_6', 'k_lag_6', 'lact_lag_6', 'lymph_lag_6', 'map_lag_6', 'mch_lag_6', 'mchc_lag_6', 'mcv_lag_6', 'methb_lag_6', 'mg_lag_6', 'na_lag_6', 'neut_lag_6', 'o2sat_lag_6', 'pco2_lag_6', 'ph_lag_6', 'phos_lag_6', 'plt_lag_6', 'po2_lag_6', 'ptt_lag_6', 'resp_lag_6', 'sbp_lag_6', 'temp_lag_6', 'tnt_lag_6', 'urine_lag_6', 'wbc_lag_6', 'alb_lag_7', 'alp_lag_7', 'alt_lag_7', 'ast_lag_7', 'be_lag_7', 'bicar_lag_7', 'bili_lag_7', 'bili_dir_lag_7', 'bnd_lag_7', 'bun_lag_7', 'ca_lag_7', 'cai_lag_7', 'ck_lag_7', 'ckmb_lag_7', 'cl_lag_7', 'crea_lag_7', 'crp_lag_7', 'dbp_lag_7', 'fgn_lag_7', 'fio2_lag_7', 'glu_lag_7', 'hgb_lag_7', 'hr_lag_7', 'inr_pt_lag_7', 'k_lag_7', 'lact_lag_7', 'lymph_lag_7', 'map_lag_7', 'mch_lag_7', 'mchc_lag_7', 'mcv_lag_7', 'methb_lag_7', 'mg_lag_7', 'na_lag_7', 'neut_lag_7', 'o2sat_lag_7', 'pco2_lag_7', 'ph_lag_7', 'phos_lag_7', 'plt_lag_7', 'po2_lag_7', 'ptt_lag_7', 'resp_lag_7', 'sbp_lag_7', 'temp_lag_7', 'tnt_lag_7', 'urine_lag_7', 'wbc_lag_7', 'alb_lag_8', 'alp_lag_8', 'alt_lag_8', 'ast_lag_8', 'be_lag_8', 'bicar_lag_8', 'bili_lag_8', 'bili_dir_lag_8', 'bnd_lag_8', 'bun_lag_8', 'ca_lag_8', 'cai_lag_8', 'ck_lag_8', 'ckmb_lag_8', 'cl_lag_8', 'crea_lag_8', 'crp_lag_8', 'dbp_lag_8', 'fgn_lag_8', 'fio2_lag_8', 'glu_lag_8', 'hgb_lag_8', 'hr_lag_8', 'inr_pt_lag_8', 'k_lag_8', 'lact_lag_8', 'lymph_lag_8', 'map_lag_8', 'mch_lag_8', 'mchc_lag_8', 'mcv_lag_8', 'methb_lag_8', 'mg_lag_8', 'na_lag_8', 'neut_lag_8', 'o2sat_lag_8', 'pco2_lag_8', 'ph_lag_8', 'phos_lag_8', 'plt_lag_8', 'po2_lag_8', 'ptt_lag_8', 'resp_lag_8', 'sbp_lag_8', 'temp_lag_8', 'tnt_lag_8', 'urine_lag_8', 'wbc_lag_8', 'alb_lag_9', 'alp_lag_9', 'alt_lag_9', 'ast_lag_9', 'be_lag_9', 'bicar_lag_9', 'bili_lag_9', 'bili_dir_lag_9', 'bnd_lag_9', 'bun_lag_9', 'ca_lag_9', 'cai_lag_9', 'ck_lag_9', 'ckmb_lag_9', 'cl_lag_9', 'crea_lag_9', 'crp_lag_9', 'dbp_lag_9', 'fgn_lag_9', 'fio2_lag_9', 'glu_lag_9', 'hgb_lag_9', 'hr_lag_9', 'inr_pt_lag_9', 'k_lag_9', 'lact_lag_9', 'lymph_lag_9', 'map_lag_9', 'mch_lag_9', 'mchc_lag_9', 'mcv_lag_9', 'methb_lag_9', 'mg_lag_9', 'na_lag_9', 'neut_lag_9', 'o2sat_lag_9', 'pco2_lag_9', 'ph_lag_9', 'phos_lag_9', 'plt_lag_9', 'po2_lag_9', 'ptt_lag_9', 'resp_lag_9', 'sbp_lag_9', 'temp_lag_9', 'tnt_lag_9', 'urine_lag_9', 'wbc_lag_9', 'alb_lag_10', 'alp_lag_10', 'alt_lag_10', 'ast_lag_10', 'be_lag_10', 'bicar_lag_10', 'bili_lag_10', 'bili_dir_lag_10', 'bnd_lag_10', 'bun_lag_10', 'ca_lag_10', 'cai_lag_10', 'ck_lag_10', 'ckmb_lag_10', 'cl_lag_10', 'crea_lag_10', 'crp_lag_10', 'dbp_lag_10', 'fgn_lag_10', 'fio2_lag_10', 'glu_lag_10', 'hgb_lag_10', 'hr_lag_10', 'inr_pt_lag_10', 'k_lag_10', 'lact_lag_10', 'lymph_lag_10', 'map_lag_10', 'mch_lag_10', 'mchc_lag_10', 'mcv_lag_10', 'methb_lag_10', 'mg_lag_10', 'na_lag_10', 'neut_lag_10', 'o2sat_lag_10', 'pco2_lag_10', 'ph_lag_10', 'phos_lag_10', 'plt_lag_10', 'po2_lag_10', 'ptt_lag_10', 'resp_lag_10', 'sbp_lag_10', 'temp_lag_10', 'tnt_lag_10', 'urine_lag_10', 'wbc_lag_10', 'alb_lag_11', 'alp_lag_11', 'alt_lag_11', 'ast_lag_11', 'be_lag_11', 'bicar_lag_11', 'bili_lag_11', 'bili_dir_lag_11', 'bnd_lag_11', 'bun_lag_11', 'ca_lag_11', 'cai_lag_11', 'ck_lag_11', 'ckmb_lag_11', 'cl_lag_11', 'crea_lag_11', 'crp_lag_11', 'dbp_lag_11', 'fgn_lag_11', 'fio2_lag_11', 'glu_lag_11', 'hgb_lag_11', 'hr_lag_11', 'inr_pt_lag_11', 'k_lag_11', 'lact_lag_11', 'lymph_lag_11', 'map_lag_11', 'mch_lag_11', 'mchc_lag_11', 'mcv_lag_11', 'methb_lag_11', 'mg_lag_11', 'na_lag_11', 'neut_lag_11', 'o2sat_lag_11', 'pco2_lag_11', 'ph_lag_11', 'phos_lag_11', 'plt_lag_11', 'po2_lag_11', 'ptt_lag_11', 'resp_lag_11', 'sbp_lag_11', 'temp_lag_11', 'tnt_lag_11', 'urine_lag_11', 'wbc_lag_11', 'alb_lag_12', 'alp_lag_12', 'alt_lag_12', 'ast_lag_12', 'be_lag_12', 'bicar_lag_12', 'bili_lag_12', 'bili_dir_lag_12', 'bnd_lag_12', 'bun_lag_12', 'ca_lag_12', 'cai_lag_12', 'ck_lag_12', 'ckmb_lag_12', 'cl_lag_12', 'crea_lag_12', 'crp_lag_12', 'dbp_lag_12', 'fgn_lag_12', 'fio2_lag_12', 'glu_lag_12', 'hgb_lag_12', 'hr_lag_12', 'inr_pt_lag_12', 'k_lag_12', 'lact_lag_12', 'lymph_lag_12', 'map_lag_12', 'mch_lag_12', 'mchc_lag_12', 'mcv_lag_12', 'methb_lag_12', 'mg_lag_12', 'na_lag_12', 'neut_lag_12', 'o2sat_lag_12', 'pco2_lag_12', 'ph_lag_12', 'phos_lag_12', 'plt_lag_12', 'po2_lag_12', 'ptt_lag_12', 'resp_lag_12', 'sbp_lag_12', 'temp_lag_12', 'tnt_lag_12', 'urine_lag_12', 'wbc_lag_12', 'alb_lag_13', 'alp_lag_13', 'alt_lag_13', 'ast_lag_13', 'be_lag_13', 'bicar_lag_13', 'bili_lag_13', 'bili_dir_lag_13', 'bnd_lag_13', 'bun_lag_13', 'ca_lag_13', 'cai_lag_13', 'ck_lag_13', 'ckmb_lag_13', 'cl_lag_13', 'crea_lag_13', 'crp_lag_13', 'dbp_lag_13', 'fgn_lag_13', 'fio2_lag_13', 'glu_lag_13', 'hgb_lag_13', 'hr_lag_13', 'inr_pt_lag_13', 'k_lag_13', 'lact_lag_13', 'lymph_lag_13', 'map_lag_13', 'mch_lag_13', 'mchc_lag_13', 'mcv_lag_13', 'methb_lag_13', 'mg_lag_13', 'na_lag_13', 'neut_lag_13', 'o2sat_lag_13', 'pco2_lag_13', 'ph_lag_13', 'phos_lag_13', 'plt_lag_13', 'po2_lag_13', 'ptt_lag_13', 'resp_lag_13', 'sbp_lag_13', 'temp_lag_13', 'tnt_lag_13', 'urine_lag_13', 'wbc_lag_13', 'alb_lag_14', 'alp_lag_14', 'alt_lag_14', 'ast_lag_14', 'be_lag_14', 'bicar_lag_14', 'bili_lag_14', 'bili_dir_lag_14', 'bnd_lag_14', 'bun_lag_14', 'ca_lag_14', 'cai_lag_14', 'ck_lag_14', 'ckmb_lag_14', 'cl_lag_14', 'crea_lag_14', 'crp_lag_14', 'dbp_lag_14', 'fgn_lag_14', 'fio2_lag_14', 'glu_lag_14', 'hgb_lag_14', 'hr_lag_14', 'inr_pt_lag_14', 'k_lag_14', 'lact_lag_14', 'lymph_lag_14', 'map_lag_14', 'mch_lag_14', 'mchc_lag_14', 'mcv_lag_14', 'methb_lag_14', 'mg_lag_14', 'na_lag_14', 'neut_lag_14', 'o2sat_lag_14', 'pco2_lag_14', 'ph_lag_14', 'phos_lag_14', 'plt_lag_14', 'po2_lag_14', 'ptt_lag_14', 'resp_lag_14', 'sbp_lag_14', 'temp_lag_14', 'tnt_lag_14', 'urine_lag_14', 'wbc_lag_14', 'alb_lag_15', 'alp_lag_15', 'alt_lag_15', 'ast_lag_15', 'be_lag_15', 'bicar_lag_15', 'bili_lag_15', 'bili_dir_lag_15', 'bnd_lag_15', 'bun_lag_15', 'ca_lag_15', 'cai_lag_15', 'ck_lag_15', 'ckmb_lag_15', 'cl_lag_15', 'crea_lag_15', 'crp_lag_15', 'dbp_lag_15', 'fgn_lag_15', 'fio2_lag_15', 'glu_lag_15', 'hgb_lag_15', 'hr_lag_15', 'inr_pt_lag_15', 'k_lag_15', 'lact_lag_15', 'lymph_lag_15', 'map_lag_15', 'mch_lag_15', 'mchc_lag_15', 'mcv_lag_15', 'methb_lag_15', 'mg_lag_15', 'na_lag_15', 'neut_lag_15', 'o2sat_lag_15', 'pco2_lag_15', 'ph_lag_15', 'phos_lag_15', 'plt_lag_15', 'po2_lag_15', 'ptt_lag_15', 'resp_lag_15', 'sbp_lag_15', 'temp_lag_15', 'tnt_lag_15', 'urine_lag_15', 'wbc_lag_15', 'alb_lag_16', 'alp_lag_16', 'alt_lag_16', 'ast_lag_16', 'be_lag_16', 'bicar_lag_16', 'bili_lag_16', 'bili_dir_lag_16', 'bnd_lag_16', 'bun_lag_16', 'ca_lag_16', 'cai_lag_16', 'ck_lag_16', 'ckmb_lag_16', 'cl_lag_16', 'crea_lag_16', 'crp_lag_16', 'dbp_lag_16', 'fgn_lag_16', 'fio2_lag_16', 'glu_lag_16', 'hgb_lag_16', 'hr_lag_16', 'inr_pt_lag_16', 'k_lag_16', 'lact_lag_16', 'lymph_lag_16', 'map_lag_16', 'mch_lag_16', 'mchc_lag_16', 'mcv_lag_16', 'methb_lag_16', 'mg_lag_16', 'na_lag_16', 'neut_lag_16', 'o2sat_lag_16', 'pco2_lag_16', 'ph_lag_16', 'phos_lag_16', 'plt_lag_16', 'po2_lag_16', 'ptt_lag_16', 'resp_lag_16', 'sbp_lag_16', 'temp_lag_16', 'tnt_lag_16', 'urine_lag_16', 'wbc_lag_16', 'alb_lag_17', 'alp_lag_17', 'alt_lag_17', 'ast_lag_17', 'be_lag_17', 'bicar_lag_17', 'bili_lag_17', 'bili_dir_lag_17', 'bnd_lag_17', 'bun_lag_17', 'ca_lag_17', 'cai_lag_17', 'ck_lag_17', 'ckmb_lag_17', 'cl_lag_17', 'crea_lag_17', 'crp_lag_17', 'dbp_lag_17', 'fgn_lag_17', 'fio2_lag_17', 'glu_lag_17', 'hgb_lag_17', 'hr_lag_17', 'inr_pt_lag_17', 'k_lag_17', 'lact_lag_17', 'lymph_lag_17', 'map_lag_17', 'mch_lag_17', 'mchc_lag_17', 'mcv_lag_17', 'methb_lag_17', 'mg_lag_17', 'na_lag_17', 'neut_lag_17', 'o2sat_lag_17', 'pco2_lag_17', 'ph_lag_17', 'phos_lag_17', 'plt_lag_17', 'po2_lag_17', 'ptt_lag_17', 'resp_lag_17', 'sbp_lag_17', 'temp_lag_17', 'tnt_lag_17', 'urine_lag_17', 'wbc_lag_17', 'alb_lag_18', 'alp_lag_18', 'alt_lag_18', 'ast_lag_18', 'be_lag_18', 'bicar_lag_18', 'bili_lag_18', 'bili_dir_lag_18', 'bnd_lag_18', 'bun_lag_18', 'ca_lag_18', 'cai_lag_18', 'ck_lag_18', 'ckmb_lag_18', 'cl_lag_18', 'crea_lag_18', 'crp_lag_18', 'dbp_lag_18', 'fgn_lag_18', 'fio2_lag_18', 'glu_lag_18', 'hgb_lag_18', 'hr_lag_18', 'inr_pt_lag_18', 'k_lag_18', 'lact_lag_18', 'lymph_lag_18', 'map_lag_18', 'mch_lag_18', 'mchc_lag_18', 'mcv_lag_18', 'methb_lag_18', 'mg_lag_18', 'na_lag_18', 'neut_lag_18', 'o2sat_lag_18', 'pco2_lag_18', 'ph_lag_18', 'phos_lag_18', 'plt_lag_18', 'po2_lag_18', 'ptt_lag_18', 'resp_lag_18', 'sbp_lag_18', 'temp_lag_18', 'tnt_lag_18', 'urine_lag_18', 'wbc_lag_18', 'alb_lag_19', 'alp_lag_19', 'alt_lag_19', 'ast_lag_19', 'be_lag_19', 'bicar_lag_19', 'bili_lag_19', 'bili_dir_lag_19', 'bnd_lag_19', 'bun_lag_19', 'ca_lag_19', 'cai_lag_19', 'ck_lag_19', 'ckmb_lag_19', 'cl_lag_19', 'crea_lag_19', 'crp_lag_19', 'dbp_lag_19', 'fgn_lag_19', 'fio2_lag_19', 'glu_lag_19', 'hgb_lag_19', 'hr_lag_19', 'inr_pt_lag_19', 'k_lag_19', 'lact_lag_19', 'lymph_lag_19', 'map_lag_19', 'mch_lag_19', 'mchc_lag_19', 'mcv_lag_19', 'methb_lag_19', 'mg_lag_19', 'na_lag_19', 'neut_lag_19', 'o2sat_lag_19', 'pco2_lag_19', 'ph_lag_19', 'phos_lag_19', 'plt_lag_19', 'po2_lag_19', 'ptt_lag_19', 'resp_lag_19', 'sbp_lag_19', 'temp_lag_19', 'tnt_lag_19', 'urine_lag_19', 'wbc_lag_19', 'alb_lag_20', 'alp_lag_20', 'alt_lag_20', 'ast_lag_20', 'be_lag_20', 'bicar_lag_20', 'bili_lag_20', 'bili_dir_lag_20', 'bnd_lag_20', 'bun_lag_20', 'ca_lag_20', 'cai_lag_20', 'ck_lag_20', 'ckmb_lag_20', 'cl_lag_20', 'crea_lag_20', 'crp_lag_20', 'dbp_lag_20', 'fgn_lag_20', 'fio2_lag_20', 'glu_lag_20', 'hgb_lag_20', 'hr_lag_20', 'inr_pt_lag_20', 'k_lag_20', 'lact_lag_20', 'lymph_lag_20', 'map_lag_20', 'mch_lag_20', 'mchc_lag_20', 'mcv_lag_20', 'methb_lag_20', 'mg_lag_20', 'na_lag_20', 'neut_lag_20', 'o2sat_lag_20', 'pco2_lag_20', 'ph_lag_20', 'phos_lag_20', 'plt_lag_20', 'po2_lag_20', 'ptt_lag_20', 'resp_lag_20', 'sbp_lag_20', 'temp_lag_20', 'tnt_lag_20', 'urine_lag_20', 'wbc_lag_20', 'alb_lag_21', 'alp_lag_21', 'alt_lag_21', 'ast_lag_21', 'be_lag_21', 'bicar_lag_21', 'bili_lag_21', 'bili_dir_lag_21', 'bnd_lag_21', 'bun_lag_21', 'ca_lag_21', 'cai_lag_21', 'ck_lag_21', 'ckmb_lag_21', 'cl_lag_21', 'crea_lag_21', 'crp_lag_21', 'dbp_lag_21', 'fgn_lag_21', 'fio2_lag_21', 'glu_lag_21', 'hgb_lag_21', 'hr_lag_21', 'inr_pt_lag_21', 'k_lag_21', 'lact_lag_21', 'lymph_lag_21', 'map_lag_21', 'mch_lag_21', 'mchc_lag_21', 'mcv_lag_21', 'methb_lag_21', 'mg_lag_21', 'na_lag_21', 'neut_lag_21', 'o2sat_lag_21', 'pco2_lag_21', 'ph_lag_21', 'phos_lag_21', 'plt_lag_21', 'po2_lag_21', 'ptt_lag_21', 'resp_lag_21', 'sbp_lag_21', 'temp_lag_21', 'tnt_lag_21', 'urine_lag_21', 'wbc_lag_21', 'alb_lag_22', 'alp_lag_22', 'alt_lag_22', 'ast_lag_22', 'be_lag_22', 'bicar_lag_22', 'bili_lag_22', 'bili_dir_lag_22', 'bnd_lag_22', 'bun_lag_22', 'ca_lag_22', 'cai_lag_22', 'ck_lag_22', 'ckmb_lag_22', 'cl_lag_22', 'crea_lag_22', 'crp_lag_22', 'dbp_lag_22', 'fgn_lag_22', 'fio2_lag_22', 'glu_lag_22', 'hgb_lag_22', 'hr_lag_22', 'inr_pt_lag_22', 'k_lag_22', 'lact_lag_22', 'lymph_lag_22', 'map_lag_22', 'mch_lag_22', 'mchc_lag_22', 'mcv_lag_22', 'methb_lag_22', 'mg_lag_22', 'na_lag_22', 'neut_lag_22', 'o2sat_lag_22', 'pco2_lag_22', 'ph_lag_22', 'phos_lag_22', 'plt_lag_22', 'po2_lag_22', 'ptt_lag_22', 'resp_lag_22', 'sbp_lag_22', 'temp_lag_22', 'tnt_lag_22', 'urine_lag_22', 'wbc_lag_22', 'alb_lag_23', 'alp_lag_23', 'alt_lag_23', 'ast_lag_23', 'be_lag_23', 'bicar_lag_23', 'bili_lag_23', 'bili_dir_lag_23', 'bnd_lag_23', 'bun_lag_23', 'ca_lag_23', 'cai_lag_23', 'ck_lag_23', 'ckmb_lag_23', 'cl_lag_23', 'crea_lag_23', 'crp_lag_23', 'dbp_lag_23', 'fgn_lag_23', 'fio2_lag_23', 'glu_lag_23', 'hgb_lag_23', 'hr_lag_23', 'inr_pt_lag_23', 'k_lag_23', 'lact_lag_23', 'lymph_lag_23', 'map_lag_23', 'mch_lag_23', 'mchc_lag_23', 'mcv_lag_23', 'methb_lag_23', 'mg_lag_23', 'na_lag_23', 'neut_lag_23', 'o2sat_lag_23', 'pco2_lag_23', 'ph_lag_23', 'phos_lag_23', 'plt_lag_23', 'po2_lag_23', 'ptt_lag_23', 'resp_lag_23', 'sbp_lag_23', 'temp_lag_23', 'tnt_lag_23', 'urine_lag_23', 'wbc_lag_23', 'alb_lag_24', 'alp_lag_24', 'alt_lag_24', 'ast_lag_24', 'be_lag_24', 'bicar_lag_24', 'bili_lag_24', 'bili_dir_lag_24', 'bnd_lag_24', 'bun_lag_24', 'ca_lag_24', 'cai_lag_24', 'ck_lag_24', 'ckmb_lag_24', 'cl_lag_24', 'crea_lag_24', 'crp_lag_24', 'dbp_lag_24', 'fgn_lag_24', 'fio2_lag_24', 'glu_lag_24', 'hgb_lag_24', 'hr_lag_24', 'inr_pt_lag_24', 'k_lag_24', 'lact_lag_24', 'lymph_lag_24', 'map_lag_24', 'mch_lag_24', 'mchc_lag_24', 'mcv_lag_24', 'methb_lag_24', 'mg_lag_24', 'na_lag_24', 'neut_lag_24', 'o2sat_lag_24', 'pco2_lag_24', 'ph_lag_24', 'phos_lag_24', 'plt_lag_24', 'po2_lag_24', 'ptt_lag_24', 'resp_lag_24', 'sbp_lag_24', 'temp_lag_24', 'tnt_lag_24', 'urine_lag_24', 'wbc_lag_24' ]
-
+PredictionDatasetpytorch.time_varying_known_reals=[ ]
 PredictionDatasetpytorch.add_relative_time_idx=False
-PredictionDatasetpytorch.target="label"
-            
-PredictionDatasetpytorch.time_varying_unknown_reals=[
+PredictionDatasetpytorch.target=[
+    "alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
                 "label",
               
             ]
-PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
-PredictionDatasetpytorch.lagged_variables=[  "alb",
+            
+PredictionDatasetpytorch.time_varying_unknown_reals=[
+    "alb",
                 "alp",
                 "alt",
                 "ast",
@@ -87,4 +133,10 @@ PredictionDatasetpytorch.lagged_variables=[  "alb",
                 "temp",
                 "tnt",
                 "urine",
-                "wbc"]
\ No newline at end of file
+                "wbc",
+                "label",
+              
+            ]
+PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
+PredictionDatasetpytorch.lagged_variables=[]
+PredictionDatasetpytorch.targetnormalizer='multi'
\ No newline at end of file
diff --git a/configs/prediction_models/TFTpytorch.gin b/configs/prediction_models/TFTpytorch.gin
index c7264ed5..37b2dd1a 100644
--- a/configs/prediction_models/TFTpytorch.gin
+++ b/configs/prediction_models/TFTpytorch.gin
@@ -89,3 +89,4 @@ PredictionDatasetpytorch.time_varying_unknown_reals=["alb",
                 "urine",
                 "wbc",]
 PredictionDatasetpytorch.lagged_variables=[]
+PredictionDatasetpytorch.targetnormalizer='single'
\ No newline at end of file
diff --git a/environment.yml b/environment.yml
index ff548770..2232e1a3 100644
--- a/environment.yml
+++ b/environment.yml
@@ -33,6 +33,8 @@ dependencies:
       # Fixed version because of NumPy incompatibility and stale development status.
       - scikit-optimize-fix==0.9.1
       - hydra-submitit-launcher==1.2.0
+      - git+https://github.com/youssefmecky96/pytorchforecasting
+      - git+https://github.com/youssefmecky96/captum
 # Note: versioning of Pytorch might be dependent on compatible CUDA version.
 # Please check yourself if your Pytorch installation supports cuda (for gpu acceleration)
 
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index a0e51dbc..8acc1f88 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -10,7 +10,7 @@
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
-from pytorch_forecasting import TimeSeriesDataSet, GroupNormalizer
+from pytorch_forecasting import TimeSeriesDataSet, GroupNormalizer, MultiNormalizer, EncoderNormalizer
 
 
 class CommonDataset(Dataset):
@@ -450,6 +450,7 @@ def __init__(
         time_varying_known_reals: List[str],
         time_varying_unknown_categoricals: List[str],
         lagged_variables: List[str],
+        target_normalizer: str,
         *args,
         ram_cache: bool = False,
         add_relative_time_idx: bool = False,
@@ -482,7 +483,13 @@ def __init__(
         self.ram_cache = ram_cache
         self.kwargs = kwargs
         self.column_names = features.columns
-
+        if target_normalizer == 'multi':
+            target_normalizer = MultiNormalizer([EncoderNormalizer(transformation='relu') for _ in range(len(target)-1)] + [GroupNormalizer(groups=["stay_id"], transformation="relu")]
+                                                )
+        else:
+            target_normalizer = GroupNormalizer(
+                groups=["stay_id"], transformation="relu"
+            )
         super().__init__(
             data=self.data,
             time_idx="time_idx",
@@ -502,9 +509,7 @@ def __init__(
             # add_target_scales=True,
             # add_encoder_length=True,
             predict_mode=True,
-            target_normalizer=GroupNormalizer(
-                groups=["stay_id"], transformation="relu"
-            ),
+            target_normalizer=target_normalizer
         )
 
     def get_balance(self) -> list:
diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index 4e082d29..baecb3e1 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -6,7 +6,7 @@
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
 from torchmetrics.classification import BinaryFairness
-
+from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson, cosine
 """"
 This file contains custom metrics that can be added to YAIB.
 """
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index b1844d3d..95040aac 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -1072,7 +1072,8 @@ def norm_function(arr): return torch.norm(arr)
                 denominator += (denominator == 0) * eps_min
 
             return nominator / denominator
-
+        if not torch.is_tensor(attribution):
+            attribution = torch.tensor(attribution).to(self.device)
         if explain_method == "Attention":
             y_pred = self.model.predict(dataloader)
             x_original = dataloader.dataset.data["reals"].clone()
@@ -1090,6 +1091,7 @@ def norm_function(arr): return torch.norm(arr)
             RIS = relative_stability_objective(
                 x_original[close_indices, :, :].detach(), x_preturb[close_indices, :, :].detach(), attribution, att_preturb, input=True
             )
+
             ROS = relative_stability_objective(
                 y_pred[close_indices], y_pred_preturb[close_indices], attribution, att_preturb, input=False
             )
@@ -1124,7 +1126,6 @@ def norm_function(arr): return torch.norm(arr)
 
             # Find where the difference is less than or equal to a thershold
             close_indices = torch.nonzero(difference <= thershold).squeeze()
-            print(close_indices)
 
             RIS = relative_stability_objective(
                 x_original[close_indices, :, :].detach(), x["encoder_cont"][close_indices, :, :].detach(), attribution[close_indices, :, :], att_preturb[close_indices, :, :], input=True

From 31c81767bd4dc32ef15894db6a557bfa02293f0c Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 28 Dec 2023 18:48:39 +0100
Subject: [PATCH 128/142] small fixes

---
 configs/prediction_models/DeepARpytorch.gin | 1 +
 icu_benchmarks/data/loader.py               | 6 ++++--
 icu_benchmarks/models/train.py              | 6 +++---
 icu_benchmarks/models/wrappers.py           | 4 ++--
 4 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/configs/prediction_models/DeepARpytorch.gin b/configs/prediction_models/DeepARpytorch.gin
index 9407222a..e7dbf54b 100644
--- a/configs/prediction_models/DeepARpytorch.gin
+++ b/configs/prediction_models/DeepARpytorch.gin
@@ -138,5 +138,6 @@ PredictionDatasetpytorch.time_varying_unknown_reals=[
             ]
 PredictionDatasetpytorch.time_varying_unknown_categoricals=[]
 PredictionDatasetpytorch.lagged_variables=[]
+PredictionDatasetpytorch.targetnormalizer='multi'
 
                 
\ No newline at end of file
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 8acc1f88..499bdc6f 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -450,8 +450,8 @@ def __init__(
         time_varying_known_reals: List[str],
         time_varying_unknown_categoricals: List[str],
         lagged_variables: List[str],
-        target_normalizer: str,
         *args,
+        target_normalizer: str = "",
         ram_cache: bool = False,
         add_relative_time_idx: bool = False,
         name: str = "",
@@ -509,7 +509,9 @@ def __init__(
             # add_target_scales=True,
             # add_encoder_length=True,
             predict_mode=True,
-            target_normalizer=target_normalizer
+            target_normalizer=GroupNormalizer(
+                groups=["stay_id"], transformation="relu"
+            )
         )
 
     def get_balance(self) -> list:
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 32358d29..0ced60c0 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -282,9 +282,9 @@ def train_common(
 
         # choose which  methods to get attributions
         methods = {
-            "G": Saliency,
+            # "G": Saliency,
             # "L": Lime,
-            "IG": IntegratedGradients,
+            # "IG": IntegratedGradients,
             # "FA": FeatureAblation,
             "R": "Random",
             "Att": "Attention"
@@ -331,7 +331,7 @@ def create_default_mask(shape):
                 feature_masks = tuple([create_default_mask(shape) if i !=
                                       1 else feature_mask_second for i, shape in enumerate(shapes)])
                 all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score, st_i_score, st_o_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, return_input_shape=True, XAI_metric=XAI_metric, random_model=random_model
+                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, feature_mask=feature_masks, return_input_shape=True, XAI_metric=XAI_metric, random_model=random_model
                 )
 
             else:
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 95040aac..ae63ab48 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -1105,8 +1105,9 @@ def norm_function(arr): return torch.norm(arr)
                 x["encoder_cont"] += noise
             y_pred_preturb = self(self.prep_data(x)).detach()
             if explain_method == "Random":
-                att_preturb = np.random.normal(size=[64, 24, 53])
 
+                att_preturb = np.random.normal(size=[64, 24, 53])
+                att_preturb = torch.tensor(att_preturb).to(self.device)
             else:
 
                 data, baselines = self.prep_data_captum(x)
@@ -1126,7 +1127,6 @@ def norm_function(arr): return torch.norm(arr)
 
             # Find where the difference is less than or equal to a thershold
             close_indices = torch.nonzero(difference <= thershold).squeeze()
-
             RIS = relative_stability_objective(
                 x_original[close_indices, :, :].detach(), x["encoder_cont"][close_indices, :, :].detach(), attribution[close_indices, :, :], att_preturb[close_indices, :, :], input=True
             )

From 07cbc3cf9c413fe2c848ab32746a2c9e62f6cf09 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <Youssef.mecky96@gmail.com>
Date: Thu, 7 Mar 2024 11:35:56 +0100
Subject: [PATCH 129/142] formatted

---
 icu_benchmarks/data/loader.py            |  31 +-
 icu_benchmarks/data/preprocessor.py      |   6 +-
 icu_benchmarks/models/constants.py       |  10 +-
 icu_benchmarks/models/custom_metrics.py  | 194 +++++-----
 icu_benchmarks/models/dl_models.py       |  10 +-
 icu_benchmarks/models/metrics.py         |   9 +-
 icu_benchmarks/models/similarity_func.py |  14 +-
 icu_benchmarks/models/train.py           | 106 ++++--
 icu_benchmarks/models/wrappers.py        | 462 +++++++++++++++++------
 icu_benchmarks/run.py                    |   6 +-
 icu_benchmarks/run_utils.py              |  73 ++--
 11 files changed, 616 insertions(+), 305 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 499bdc6f..1707a2d3 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -3,14 +3,19 @@
 import pandas as pd
 import gin
 import numpy as np
-from torch import Tensor, cat, from_numpy, float32, min, max, randn_like
+from torch import Tensor, cat, from_numpy, float32, randn_like
 from torch.utils.data import Dataset
 import logging
 from typing import Dict, Tuple, Union
 from icu_benchmarks.imputation.amputations import ampute_data
 from .constants import DataSegment as Segment
 from .constants import DataSplit as Split
-from pytorch_forecasting import TimeSeriesDataSet, GroupNormalizer, MultiNormalizer, EncoderNormalizer
+from pytorch_forecasting import (
+    TimeSeriesDataSet,
+    GroupNormalizer,
+    MultiNormalizer,
+    EncoderNormalizer,
+)
 
 
 class CommonDataset(Dataset):
@@ -483,13 +488,13 @@ def __init__(
         self.ram_cache = ram_cache
         self.kwargs = kwargs
         self.column_names = features.columns
-        if target_normalizer == 'multi':
-            target_normalizer = MultiNormalizer([EncoderNormalizer(transformation='relu') for _ in range(len(target)-1)] + [GroupNormalizer(groups=["stay_id"], transformation="relu")]
-                                                )
-        else:
-            target_normalizer = GroupNormalizer(
-                groups=["stay_id"], transformation="relu"
+        if target_normalizer == "multi":
+            target_normalizer = MultiNormalizer(
+                [EncoderNormalizer(transformation="relu") for _ in range(len(target) - 1)]
+                + [GroupNormalizer(groups=["stay_id"], transformation="relu")]
             )
+        else:
+            target_normalizer = GroupNormalizer(groups=["stay_id"], transformation="relu")
         super().__init__(
             data=self.data,
             time_idx="time_idx",
@@ -509,9 +514,7 @@ def __init__(
             # add_target_scales=True,
             # add_encoder_length=True,
             predict_mode=True,
-            target_normalizer=GroupNormalizer(
-                groups=["stay_id"], transformation="relu"
-            )
+            target_normalizer=GroupNormalizer(groups=["stay_id"], transformation="relu"),
         )
 
     def get_balance(self) -> list:
@@ -533,12 +536,14 @@ def get_feature_names(self):
     def randomize_labels(self, num_classes=None, min=None, max=None):
         if num_classes == 1:
             random_target = np.random.uniform(
-                self.data["target"][0].min(), self.data["target"][0].max(), size=len(self.data["target"][0])
+                self.data["target"][0].min(),
+                self.data["target"][0].max(),
+                size=len(self.data["target"][0]),
             )
         else:
             random_target = np.random.randint(num_classes, size=len(self.data["target"][0]))
         self.data["target"][0] = Tensor(random_target)
 
     def add_noise(self, num_classes=None, min=None, max=None):
-        noise = randn_like(self.data["reals"])*0.01
+        noise = randn_like(self.data["reals"]) * 0.01
         self.data["reals"] += noise
diff --git a/icu_benchmarks/data/preprocessor.py b/icu_benchmarks/data/preprocessor.py
index 9da25308..486ab0e8 100644
--- a/icu_benchmarks/data/preprocessor.py
+++ b/icu_benchmarks/data/preprocessor.py
@@ -117,8 +117,7 @@ def _process_static(self, data, vars):
             sta_rec.add_step(StepScale())
 
         sta_rec.add_step(StepImputeFastZeroFill(sel=all_numeric_predictors()))
-        sta_rec.add_step(StepSklearn(SimpleImputer(missing_values=None,
-                         strategy="most_frequent"), sel=has_type("object")))
+        sta_rec.add_step(StepSklearn(SimpleImputer(missing_values=None, strategy="most_frequent"), sel=has_type("object")))
         sta_rec.add_step(StepSklearn(LabelEncoder(), sel=has_type("object"), columnwise=True))
 
         data = apply_recipe_to_splits(sta_rec, data, Segment.static, self.save_cache, self.load_cache)
@@ -288,8 +287,7 @@ def _process_dynamic_data(self, data, vars):
         if self.filter_missing_values:
             rows_to_remove = data[Segment.dynamic][vars[Segment.dynamic]].isna().sum(axis=1) != 0
             ids_to_remove = data[Segment.dynamic].loc[rows_to_remove][vars["GROUP"]].unique()
-            data = {table_name: table.loc[~table[vars["GROUP"]].isin(
-                ids_to_remove)] for table_name, table in data.items()}
+            data = {table_name: table.loc[~table[vars["GROUP"]].isin(ids_to_remove)] for table_name, table in data.items()}
             logging.info(f"Removed {len(ids_to_remove)} stays with missing values.")
         return data
 
diff --git a/icu_benchmarks/models/constants.py b/icu_benchmarks/models/constants.py
index 093d15bd..224355eb 100644
--- a/icu_benchmarks/models/constants.py
+++ b/icu_benchmarks/models/constants.py
@@ -1,4 +1,9 @@
-from ignite.contrib.metrics import AveragePrecision, ROC_AUC, RocCurve, PrecisionRecallCurve
+from ignite.contrib.metrics import (
+    AveragePrecision,
+    ROC_AUC,
+    RocCurve,
+    PrecisionRecallCurve,
+)
 from ignite.metrics import Accuracy, RootMeanSquaredError
 from sklearn.calibration import calibration_curve
 from sklearn.metrics import (
@@ -14,8 +19,6 @@
 )
 from torchmetrics.classification import (
     AUROC,
-    AveragePrecision as TorchMetricsAveragePrecision,
-    PrecisionRecallCurve as TorchMetricsPrecisionRecallCurve,
     CalibrationError,
     F1Score,
 )
@@ -25,7 +28,6 @@
     BalancedAccuracy,
     MAE,
     JSD,
-    BinaryFairnessWrapper,
 )
 
 
diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index baecb3e1..84b074cc 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -6,7 +6,8 @@
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
 from torchmetrics.classification import BinaryFairness
-from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson, cosine
+from icu_benchmarks.models.similarity_func import correlation_spearman, cosine
+
 """"
 This file contains custom metrics that can be added to YAIB.
 """
@@ -32,7 +33,9 @@ def accuracy(output, target, topk=(1,)):
 class BalancedAccuracy(EpochMetric):
     def __init__(self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False) -> None:
         super(BalancedAccuracy, self).__init__(
-            self.balanced_accuracy_compute, output_transform=output_transform, check_compute_fn=check_compute_fn
+            self.balanced_accuracy_compute,
+            output_transform=output_transform,
+            check_compute_fn=check_compute_fn,
         )
 
         def balanced_accuracy_compute(y_preds: torch.Tensor, y_targets: torch.Tensor) -> float:
@@ -44,7 +47,9 @@ def balanced_accuracy_compute(y_preds: torch.Tensor, y_targets: torch.Tensor) ->
 class CalibrationCurve(EpochMetric):
     def __init__(self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False) -> None:
         super(CalibrationCurve, self).__init__(
-            self.ece_curve_compute_fn, output_transform=output_transform, check_compute_fn=check_compute_fn
+            self.ece_curve_compute_fn,
+            output_transform=output_transform,
+            check_compute_fn=check_compute_fn,
         )
 
         def ece_curve_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor, n_bins=10) -> float:
@@ -67,7 +72,6 @@ def __init__(
         )
 
         def mae_with_invert_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor, invert_fn=Callable) -> float:
-
             y_true = invert_fn(y_targets.numpy().reshape(-1, 1))[:, 0]
             y_pred = invert_fn(y_preds.numpy().reshape(-1, 1))[:, 0]
 
@@ -138,33 +142,25 @@ def feature_helper(self, trainer, step_prefix):
 
 
 class Faithfulness(EpochMetric):
-    def __init__(
-        self,
-        output_transform: Callable = lambda x: x,
-        check_compute_fn: bool = False,
-        *args, **kwargs
-    ) -> None:
-        super().__init__(output_transform,
-                         check_compute_fn, *args, **kwargs
-                         )
+    def __init__(self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False, *args, **kwargs) -> None:
+        super().__init__(output_transform, check_compute_fn, *args, **kwargs)
 
     def update(
-            self,
-            x,
-            attribution,
-            model,
-            similarity_func=None,
-            nr_runs=100,
-            pertrub=None,
-            subset_size=3,
-            feature=False,
-            time_step=False,
-            feature_timestep=False,
-            device='cuda'
-
+        self,
+        x,
+        attribution,
+        model,
+        similarity_func=None,
+        nr_runs=100,
+        pertrub=None,
+        subset_size=3,
+        feature=False,
+        time_step=False,
+        feature_timestep=False,
+        device="cuda",
     ):
         """
-        Calculates faithfulness scores for captum attributions 
+        Calculates faithfulness scores for captum attributions
 
         Args:
             - x:Batch input
@@ -200,22 +196,23 @@ def update(
             explanations." IJCAI (2020): 3016-3022.
             2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
         """
+
         def add_noise(x, indices, time_step, feature_timestep):
             noise = torch.randn_like(x["encoder_cont"])
             if time_step:
-                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing='ij')
+                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
 
                 with torch.no_grad():
                     x["encoder_cont"][idx0, idx1, :] += noise[idx0, idx1, :]
 
             elif feature:
-                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing='ij')
+                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
 
                 with torch.no_grad():
                     x["encoder_cont"][idx0, :, idx1] += noise[idx0, :, idx1]
 
             elif feature_timestep:
-                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing='ij')
+                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
 
                 with torch.no_grad():
                     x["encoder_cont"][idx0, idx1, idx2] += noise[idx0, idx1, idx2]
@@ -223,22 +220,28 @@ def add_noise(x, indices, time_step, feature_timestep):
         def apply_baseline(x, indices, time_step, feature_timestep):
             mask = torch.ones_like(x["encoder_cont"])
             if time_step:
-
-                idx0, idx1, = np.meshgrid(indices[0], indices[1], indexing='ij')
+                (
+                    idx0,
+                    idx1,
+                ) = np.meshgrid(indices[0], indices[1], indexing="ij")
 
                 mask[idx0, idx1, :] -= mask[idx0, idx1, :]
             elif feature:
-                idx0, idx1, = np.meshgrid(indices[0], indices[1], indexing='ij')
+                (
+                    idx0,
+                    idx1,
+                ) = np.meshgrid(indices[0], indices[1], indexing="ij")
 
                 mask[idx0, :, idx1] -= mask[idx0, :, idx1]
 
             elif feature_timestep:
-                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing='ij')
+                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
 
                 mask[idx0, idx1, idx2] -= mask[idx0, idx1, idx2]
 
             with torch.no_grad():
                 x["encoder_cont"] *= mask
+
         # Assuming 'attribution' is already a GPU tensor
         if not torch.is_tensor(attribution):
             attribution = torch.tensor(attribution).to(device)
@@ -281,21 +284,17 @@ def apply_baseline(x, indices, time_step, feature_timestep):
             y_pred_perturb = (model(model.prep_data(x))).detach()  # Keep on GPU
 
             if time_step:
-
                 if attribution.size() == torch.Size([24]):
                     att_sums.append((attribution[timesteps_idx]).sum())
                 else:
                     att_sums.append((attribution[patient_idx, :][:, timesteps_idx]).sum())
             elif feature:
-
                 if len(attribution) == 53:
                     att_sums.append((attribution[feature_idx]).sum())
                 else:
-
                     att_sums.append((attribution[patient_idx, :][:, feature_idx]).sum())
             elif feature_timestep:
-                att_sums.append((attribution[patient_idx, :, :]
-                                [:, timesteps_idx, :][:, :, feature_idx]).sum())
+                att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :][:, :, feature_idx]).sum())
 
             pred_deltas.append((y_pred - y_pred_perturb)[patient_idx].item())
             # Convert to CPU for numpy operations
@@ -310,40 +309,40 @@ def apply_baseline(x, indices, time_step, feature_timestep):
 
 
 class Stability(EpochMetric):
-    def __init__(
-        self, output_transform: Callable = lambda x: x,
-        check_compute_fn: bool = False,
-        *args, **kwargs
-    ) -> None:
-        super().__init__(output_transform,
-                         check_compute_fn, *args, **kwargs
-                         )
+    def __init__(self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False, *args, **kwargs) -> None:
+        super().__init__(output_transform, check_compute_fn, *args, **kwargs)
 
-    def update(self, x,
-               attribution, model, explain_method, dataloader=None, thershold=0.5, device='cuda', **kwargs
-               ):
+    def update(self, x, attribution, model, explain_method, dataloader=None, thershold=0.5, device="cuda", **kwargs):
         """
-    Args:
-            - x:Batch input
-            -attribution: attribution 
-            - explain_method:function to generate explantations
-            - method_name: Name of the explantation
-            - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+        Args:
+                - x:Batch input
+                -attribution: attribution
+                - explain_method:function to generate explantations
+                - method_name: Name of the explantation
+                - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
 
 
-        Returns:
-            RIS : relative distance between the explantation and the input
-            ROS: relative distance between the explantation and the output
+            Returns:
+                RIS : relative distance between the explantation and the input
+                ROS: relative distance between the explantation and the output
 
 
-    References:
-            1) `https://arxiv.org/pdf/2203.06877.pdf
-            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+        References:
+                1) `https://arxiv.org/pdf/2203.06877.pdf
+                2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
 
         def relative_stability_objective(
-            x, xs, e_x, e_xs, close_indices, eps_min=0.0001, input=False, attention=False, device='cuda'
+            x,
+            xs,
+            e_x,
+            e_xs,
+            close_indices,
+            eps_min=0.0001,
+            input=False,
+            attention=False,
+            device="cuda",
         ) -> torch.Tensor:
             """
             Computes relative input and output stabilities maximization objective
@@ -355,15 +354,16 @@ def relative_stability_objective(
                 xs: perturbed tensor.
                 e_x: Explanations for x.
                 e_xs: Explanations for xs.
-                eps_min:Value to avoid division by zero if needed 
+                eps_min:Value to avoid division by zero if needed
                 input:Boolean to indicate if this is an input or an output
-                device: the device to keep the tensors on 
+                device: the device to keep the tensors on
 
             Returns:
 
                 ris_obj: Tensor
                     RIS maximization objective.
             """
+
             # Function to convert inputs to tensors if they are numpy arrays
             def to_tensor(input_array):
                 if isinstance(input_array, np.ndarray):
@@ -383,11 +383,19 @@ def to_tensor(input_array):
                 num_dim = e_x.ndim
 
             if num_dim == 3:
-                def norm_function(arr): return torch.norm(arr, dim=(-1, -2))
+
+                def norm_function(arr):
+                    return torch.norm(arr, dim=(-1, -2))
+
             elif num_dim == 2:
-                def norm_function(arr): return torch.norm(arr, dim=-1)
+
+                def norm_function(arr):
+                    return torch.norm(arr, dim=-1)
+
             else:
-                def norm_function(arr): return torch.norm(arr)
+
+                def norm_function(arr):
+                    return torch.norm(arr)
 
             nominator = (e_x - e_xs) / (e_x + (e_x == 0) * eps_min)
             nominator = norm_function(nominator)
@@ -420,10 +428,22 @@ def norm_function(arr): return torch.norm(arr)
             close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
 
             RIS = relative_stability_objective(
-                x_original.detach(), x_preturb.detach(), attribution, att_preturb, close_indices=close_indices, input=True, attention=True
+                x_original.detach(),
+                x_preturb.detach(),
+                attribution,
+                att_preturb,
+                close_indices=close_indices,
+                input=True,
+                attention=True,
             )
             ROS = relative_stability_objective(
-                y_pred, y_pred_preturb, attribution, att_preturb, close_indices=close_indices, input=False, attention=True
+                y_pred,
+                y_pred_preturb,
+                attribution,
+                att_preturb,
+                close_indices=close_indices,
+                input=False,
+                attention=True,
             )
 
         else:
@@ -431,7 +451,7 @@ def norm_function(arr): return torch.norm(arr)
             x_original = x["encoder_cont"].detach().clone()
 
             with torch.no_grad():
-                noise = torch.randn_like(x["encoder_cont"])*0.01
+                noise = torch.randn_like(x["encoder_cont"]) * 0.01
                 x["encoder_cont"] += noise
             y_pred_preturb = model(model.prep_data(x)).detach()
             if explain_method == "Random":
@@ -450,37 +470,36 @@ def norm_function(arr): return torch.norm(arr)
             RIS = relative_stability_objective(
                 x_original.detach(),
                 x["encoder_cont"].detach(),
-                attribution, att_preturb, close_indices=close_indices, input=True
+                attribution,
+                att_preturb,
+                close_indices=close_indices,
+                input=True,
             )
             ROS = relative_stability_objective(
-                y_pred, y_pred_preturb, attribution, att_preturb, close_indices=close_indices, input=False
+                y_pred,
+                y_pred_preturb,
+                attribution,
+                att_preturb,
+                close_indices=close_indices,
+                input=False,
             )
 
         return np.max(RIS.cpu().numpy()).astype(np.float64), np.max(ROS.cpu().numpy()).astype(np.float64)
 
 
 class Randomization(EpochMetric):
-    def __init__(
-        self, output_transform: Callable = lambda x: x,
-        check_compute_fn: bool = False,
-        *args, **kwargs
-    ) -> None:
-        super().__init__(output_transform,
-                         check_compute_fn, *args, **kwargs
-                         )
+    def __init__(self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False, *args, **kwargs) -> None:
+        super().__init__(output_transform, check_compute_fn, *args, **kwargs)
 
-    def update(
-        self, x,
-        attribution, model, explain_method, random_model, similarity_func=cosine, dataloader=None, **kwargs
-    ):
+    def update(self, x, attribution, model, explain_method, random_model, similarity_func=cosine, dataloader=None, **kwargs):
         """
 
         Args:
             - x:Batch input
-            -attribution: attribution 
+            -attribution: attribution
             - explain_method:function to generate explantations
             - random_model: Reference to model trained on random labels
-            - similarity_func: Function to measure similiarity 
+            - similarity_func: Function to measure similiarity
             - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
             - method_name: Name of the explantation
 
@@ -516,7 +535,6 @@ def update(
             score = similarity_func(np.random.normal(size=[64, 24, 53]).flatten(), attribution.flatten())
         else:
             data, baselines = model.prep_data_captum(x)
-            y_pred = model(data).detach()
 
             random_attr, features_attrs, timestep_attrs = model.explantation2(x, explain_method)
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index ff66c4da..6141079f 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -11,11 +11,13 @@
     PositionalEncoding,
 )
 import matplotlib.pyplot as plt
-from icu_benchmarks.models.wrappers import DLPredictionWrapper, DLPredictionPytorchForecastingWrapper
-from torch import Tensor, FloatTensor, zeros_like, ones_like, randn_like, is_tensor
+from icu_benchmarks.models.wrappers import (
+    DLPredictionWrapper,
+    DLPredictionPytorchForecastingWrapper,
+)
+from torch import Tensor
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
-from pytorch_forecasting.metrics import QuantileLoss, MAE
-import matplotlib.pyplot as plt
+from pytorch_forecasting.metrics import QuantileLoss
 
 
 @gin.configurable
diff --git a/icu_benchmarks/models/metrics.py b/icu_benchmarks/models/metrics.py
index 7301d8d9..738c5523 100644
--- a/icu_benchmarks/models/metrics.py
+++ b/icu_benchmarks/models/metrics.py
@@ -1,12 +1,10 @@
 import torch
 from typing import Callable
 import numpy as np
-from ignite.metrics import EpochMetric, Precision, Recall
+from ignite.metrics import EpochMetric
 from sklearn.metrics import balanced_accuracy_score, mean_absolute_error
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
-import gin
-import torch.nn.functional as F
 
 
 """"
@@ -45,13 +43,8 @@ def ece_curve_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor) -> floa
 
 
 def mae_with_invert_compute_fn(y_preds: torch.Tensor, y_targets: torch.Tensor, invert_fn=Callable) -> float:
-    print('y_true', y_true)
-    print('y_pred', y_pred)
     y_true = invert_fn(y_targets.numpy().reshape(-1, 1))[:, 0]
     y_pred = invert_fn(y_preds.numpy().reshape(-1, 1))[:, 0]
-    print('y_true', y_true)
-    print('y_pred', y_pred)
-    print('mae', mean_absolute_error(y_true, y_pred))
     return mean_absolute_error(y_true, y_pred)
 
 
diff --git a/icu_benchmarks/models/similarity_func.py b/icu_benchmarks/models/similarity_func.py
index 3ff96474..fd931e7e 100644
--- a/icu_benchmarks/models/similarity_func.py
+++ b/icu_benchmarks/models/similarity_func.py
@@ -1,9 +1,15 @@
-"""This module holds a collection of similarity functions i.e., ways to measure the distance between two inputs (or explanations)."""
+"""This module holds a collection of similarity functions i.e., 
+ways to measure the distance between two inputs (or explanations)."""
 
 # This file is part of Quantus.
-# Quantus is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
-# Quantus is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
-# You should have received a copy of the GNU Lesser General Public License along with Quantus. If not, see <https://www.gnu.org/licenses/>.
+# Quantus is free software: you can redistribute it and/or modify it under
+# the terms of the GNU Lesser General Public License as published by the Free Software Foundation,
+# either version 3 of the License, or (at your option) any later version.
+# Quantus is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+# or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
+# You should have received a copy of the GNU Lesser General Public License along with Quantus.
+# If not, see <https://www.gnu.org/licenses/>.
 # Quantus project URL: <https://github.com/understandable-machine-intelligence-lab/Quantus>.
 # Quantus project URL: https://github.com/understandable-machine-intelligence-lab/Quantus
 
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index 0ced60c0..e7ed5d68 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -1,14 +1,11 @@
 import os
 import gin
 import torch
-import pickle
 import logging
 import json
 import pandas as pd
 from joblib import load
 from torch.optim import Adam
-import numpy as np
-import quantus
 from torch.utils.data import DataLoader
 from pytorch_lightning.loggers import TensorBoardLogger, WandbLogger
 from pytorch_lightning import Trainer
@@ -29,7 +26,6 @@
 from icu_benchmarks.models.utils import save_config_file, JSONMetricsLogger
 from icu_benchmarks.contants import RunMode
 from icu_benchmarks.data.constants import DataSplit as Split
-from collections import OrderedDict
 from captum.attr import IntegratedGradients, Saliency, FeatureAblation, Lime
 
 
@@ -78,7 +74,6 @@ def train_common(
     XAI_metric: bool = False,
     random_labels: bool = False,
     random_model_dir: str = None,
-
 ):
     """Common wrapper to train all benchmarked models.
 
@@ -282,22 +277,39 @@ def train_common(
 
         # choose which  methods to get attributions
         methods = {
-            # "G": Saliency,
-            # "L": Lime,
-            # "IG": IntegratedGradients,
-            # "FA": FeatureAblation,
+            "G": Saliency,
+            "L": Lime,
+            "IG": IntegratedGradients,
+            "FA": FeatureAblation,
             "R": "Random",
-            "Att": "Attention"
-
+            "Att": "Attention",
         }
         for key, item in methods.items():
             # If conditions needed here as different explantations require different inputs
             if key == "IG":
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score, st_i_score, st_o_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, n_steps=50, XAI_metric=XAI_metric, random_model=random_model
+                (
+                    all_attrs,
+                    features_attrs,
+                    timestep_attrs,
+                    ts_v_score,
+                    ts_score,
+                    v_score,
+                    r_score,
+                    st_i_score,
+                    st_o_score,
+                ) = model.explantation(
+                    dataloader=test_loader,
+                    method=item,
+                    log_dir=log_dir,
+                    plot=True,
+                    n_steps=50,
+                    XAI_metric=XAI_metric,
+                    random_model=random_model,
                 )
             elif key == "L" or key == "FA":
-                # for Lime and feature ablation we need to define what is a feature we define each variable per timestep as a feature
+                """for Lime and feature ablation we need to define
+                what is a feature we define each variable
+                per timestep as a feature"""
                 shapes = [
                     torch.Size([64, 24, 0]),
                     torch.Size([64, 24, 53]),
@@ -309,7 +321,7 @@ def train_common(
                     torch.Size([64]),
                     torch.Size([64, 1]),
                     torch.Size([64, 1]),
-                    torch.Size([64, 2])
+                    torch.Size([64, 2]),
                 ]
 
                 # Create a default mask for non-targeted tensors
@@ -328,40 +340,72 @@ def create_default_mask(shape):
                 feature_mask_second = feature_mask_second.unsqueeze(0).repeat(shapes[1][0], 1, 1)
 
                 # Create a tuple of masks
-                feature_masks = tuple([create_default_mask(shape) if i !=
-                                      1 else feature_mask_second for i, shape in enumerate(shapes)])
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score, st_i_score, st_o_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, feature_mask=feature_masks, return_input_shape=True, XAI_metric=XAI_metric, random_model=random_model
+                feature_masks = tuple(
+                    [create_default_mask(shape) if i != 1 else feature_mask_second for i, shape in enumerate(shapes)]
+                )
+                (
+                    all_attrs,
+                    features_attrs,
+                    timestep_attrs,
+                    ts_v_score,
+                    ts_score,
+                    v_score,
+                    r_score,
+                    st_i_score,
+                    st_o_score,
+                ) = model.explantation(
+                    dataloader=test_loader,
+                    method=item,
+                    log_dir=log_dir,
+                    plot=True,
+                    feature_mask=feature_masks,
+                    return_input_shape=True,
+                    XAI_metric=XAI_metric,
+                    random_model=random_model,
                 )
 
             else:
-                all_attrs, features_attrs, timestep_attrs, ts_v_score, ts_score, v_score, r_score, st_i_score, st_o_score = model.explantation(
-                    dataloader=test_loader, method=item, log_dir=log_dir, plot=True, XAI_metric=XAI_metric, random_model=random_model
+                (
+                    all_attrs,
+                    features_attrs,
+                    timestep_attrs,
+                    ts_v_score,
+                    ts_score,
+                    v_score,
+                    r_score,
+                    st_i_score,
+                    st_o_score,
+                ) = model.explantation(
+                    dataloader=test_loader,
+                    method=item,
+                    log_dir=log_dir,
+                    plot=True,
+                    XAI_metric=XAI_metric,
+                    random_model=random_model,
                 )
 
             if XAI_metric:
                 # logging metric scores
                 print("{} Attributions Faithfulness Timesteps ".format(key), ts_score)
                 XAI_dict["{}_Faith Timesteps".format(key)] = ts_score
-                print("{}_ROS ".format(
-                    key), st_o_score)
+                print("{}_ROS ".format(key), st_o_score)
                 XAI_dict["{}_ROS".format(key)] = st_o_score
-                print("{}_RIS ".format(
-                    key), st_i_score)
+                print("{}_RIS ".format(key), st_i_score)
                 XAI_dict["{}_RIS".format(key)] = st_i_score
                 if key == "Att":
-                    print("Variable selection weights faithfulness featrues ".format(key), v_score)
-                    XAI_dict["VSN_Faith Features".format(key)] = v_score
+                    print("{} weights faithfulness featrues ".format(key), v_score)
+                    XAI_dict["{}_Faith Features".format(key)] = v_score
 
                 else:
                     print("{} Attributions faithfulness featrues ".format(key), v_score)
                     XAI_dict["{}_Faith Features".format(key)] = v_score
 
-                    print("{}_Attributions Faithfulness Variable Per Timestep ".format(
-                        key), ts_v_score)
+                    print(
+                        "{}_Attributions Faithfulness Variable Per Timestep ".format(key),
+                        ts_v_score,
+                    )
                     XAI_dict["{}_Faith Variable Per Timestep".format(key)] = ts_v_score
-                print("{}_Data Randomization Distance ".format(
-                    key), r_score)
+                print("{}_Data Randomization Distance ".format(key), r_score)
                 XAI_dict["{}_Data Randomization Distance".format(key)] = r_score
 
         # Getting the interpertations using pytorch forecasting native methods
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index ae63ab48..f22ecf85 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -4,7 +4,7 @@
 import torchmetrics
 from sklearn.metrics import log_loss, mean_squared_error
 import torch
-from torch.nn import MSELoss, CrossEntropyLoss, L1Loss
+from torch.nn import MSELoss, CrossEntropyLoss
 import torch.nn as nn
 from torch import Tensor, FloatTensor
 from torch.optim import Optimizer, Adam
@@ -19,10 +19,9 @@
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
 import matplotlib.pyplot as plt
-from icu_benchmarks.models.similarity_func import correlation_spearman, distance_euclidean, correlation_pearson, cosine
+from icu_benchmarks.models.similarity_func import correlation_spearman, cosine
 import captum
 from captum._utils.models.linear_model import SkLearnLasso
-import os
 
 gin.config.external_configurable(nn.functional.nll_loss, module="torch.nn.functional")
 gin.config.external_configurable(nn.functional.cross_entropy, module="torch.nn.functional")
@@ -162,8 +161,7 @@ def finalize_step(self, step_prefix=""):
         try:
             for name, metric in self.metrics[step_prefix].items():
                 try:
-                    value = np.float32(metric.compute()) if isinstance(
-                        metric.compute(), np.float64) else metric.compute()
+                    value = np.float32(metric.compute()) if isinstance(metric.compute(), np.float64) else metric.compute()
                     self.log_dict({f"{step_prefix}/{name}": value}, sync_dist=True)
 
                 except (NotComputableError, ValueError) as e:
@@ -467,7 +465,6 @@ def step_fn(self, element, step_prefix=""):
         transformed_output = self.output_transform((prediction, target))
 
         for key, value in self.metrics[step_prefix].items():
-
             if isinstance(value, torchmetrics.Metric):
                 if key == "Binary_Fairness":
                     feature_names = self.metrics[step_prefix][key].feature_helper(self.trainer, step_prefix)
@@ -493,7 +490,7 @@ def prep_data_captum(self, x):
             - x:Batch from dataloader
         Returns:
             - data:batch data in a tuple after being prepared
-            - baselines:Basically zero tensors in the input 
+            - baselines:Basically zero tensors in the input
         """
         # captum requires gradient and float values
 
@@ -553,10 +550,70 @@ def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir
         plt.xlabel("Feature")
         plt.ylabel("{} Attribution".format(method_name))
         plt.title("{} Attribution Values".format(method_name))
-        plt.xticks(x_values, ['height', 'weight', 'age', 'sex', 'time_idx', 'alb', 'alp', 'alt', 'ast', 'be', 'bicar', 'bili', 'bili_dir', 'bnd', 'bun', 'ca', 'cai', 'ck', 'ckmb', 'cl', 'crea', 'crp', 'dbp', 'fgn', 'fio2', 'glu',
-                   'hgb', 'hr', 'inr_pt', 'k', 'lact', 'lymph', 'map', 'mch', 'mchc', 'mcv', 'methb', 'mg', 'na', 'neut', 'o2sat', 'pco2', 'ph', 'phos', 'plt', 'po2', 'ptt', 'resp', 'sbp', 'temp', 'tnt', 'urine', 'wbc'], rotation=90)
+        plt.xticks(
+            x_values,
+            [
+                "height",
+                "weight",
+                "age",
+                "sex",
+                "time_idx",
+                "alb",
+                "alp",
+                "alt",
+                "ast",
+                "be",
+                "bicar",
+                "bili",
+                "bili_dir",
+                "bnd",
+                "bun",
+                "ca",
+                "cai",
+                "ck",
+                "ckmb",
+                "cl",
+                "crea",
+                "crp",
+                "dbp",
+                "fgn",
+                "fio2",
+                "glu",
+                "hgb",
+                "hr",
+                "inr_pt",
+                "k",
+                "lact",
+                "lymph",
+                "map",
+                "mch",
+                "mchc",
+                "mcv",
+                "methb",
+                "mg",
+                "na",
+                "neut",
+                "o2sat",
+                "pco2",
+                "ph",
+                "phos",
+                "plt",
+                "po2",
+                "ptt",
+                "resp",
+                "sbp",
+                "temp",
+                "tnt",
+                "urine",
+                "wbc",
+            ],
+            rotation=90,
+        )
         plt.tight_layout()
-        plt.savefig(log_dir / "{}_attribution_features_plot.png".format(method_name), bbox_inches="tight")
+        plt.savefig(
+            log_dir / "{}_attribution_features_plot.png".format(method_name),
+            bbox_inches="tight",
+        )
 
         # Plot for timestep attributions
         x_values = np.arange(1, len(timestep_attrs) + 1)
@@ -577,12 +634,22 @@ def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir
         plt.tight_layout()
         plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
 
-    def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=False, random_model=None, test_dataset=None, **kwargs):
+    def explantation(
+        self,
+        dataloader,
+        method,
+        log_dir=".",
+        plot=False,
+        XAI_metric=False,
+        random_model=None,
+        test_dataset=None,
+        **kwargs,
+    ):
         """
         Generic method to combine pytorchforecasting data loading , interpertations and captum to generate attributions
 
         Args:
-            - dataloader: pytorchforecasting data loader 
+            - dataloader: pytorchforecasting data loader
             - method: The explantation method chosen
             - log_dir= The directory to output the plots
             - plot= Determines if plots should be done or not
@@ -603,19 +670,29 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
         st_i_score = []
         st_o_score = []
 
-        method_name = method if (method == "Random") or (method == "Attention") else (
-            method.__name__)
+        method_name = method if (method == "Random") or (method == "Attention") else (method.__name__)
         if (method_name == "Random") or (method_name == "Attention"):
             if method_name == "Attention":
                 Interpertations = self.interpertations(dataloader=dataloader, log_dir=log_dir, plot=plot)
                 timestep_attrs = Interpertations["attention"]
                 features_attrs = Interpertations["static_variables"].tolist()
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
-                r_score = self.Data_Randomization(x=None, attribution=timestep_attrs,
-                                                  explain_method=method, random_model=random_model, dataloader=dataloader, method_name=method_name)
-                st_i_score, st_o_score = self.Relative_Stability(x=None,
-                                                                 attribution=timestep_attrs, explain_method=method, method_name=method_name, dataloader=dataloader, **kwargs
-                                                                 )
+                r_score = self.Data_Randomization(
+                    x=None,
+                    attribution=timestep_attrs,
+                    explain_method=method,
+                    random_model=random_model,
+                    dataloader=dataloader,
+                    method_name=method_name,
+                )
+                st_i_score, st_o_score = self.Relative_Stability(
+                    x=None,
+                    attribution=timestep_attrs,
+                    explain_method=method,
+                    method_name=method_name,
+                    dataloader=dataloader,
+                    **kwargs,
+                )
             elif method_name == "Random":
                 # Generate random attributions for baseline comparison
                 all_attrs = np.random.normal(size=[64, 24, 53])
@@ -623,32 +700,82 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 timestep_attrs = all_attrs.mean(axis=(2))
             if XAI_metric:
                 for batch in dataloader:
-
                     for key, value in batch[0].items():
                         batch[0][key] = batch[0][key].to(self.device)
                     x = batch[0]
 
                     if method_name == "Random":
-
-                        f_ts_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
-                                                                          pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
-                        f_ts_score.append(self.Faithfulness_Correlation(x, all_attrs,
-                                                                        pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
-                        f_v_score.append(self.Faithfulness_Correlation(x, all_attrs,
-                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
-
-                        r_score.append(self.Data_Randomization(x, attribution=all_attrs,
-                                       explain_method=method, random_model=random_model, method_name=method_name))
-                        res1, res2 = self.Relative_Stability(x,
-                                                             all_attrs, explain_method=method, method_name=method_name, dataloader=None, **kwargs
-                                                             )
+                        f_ts_v_score.append(
+                            self.Faithfulness_Correlation(
+                                x,
+                                all_attrs,
+                                pertrub="baseline",
+                                feature_timestep=True,
+                                subset_size=[4, 9],
+                                nr_runs=100,
+                            )
+                        )
+                        f_ts_score.append(
+                            self.Faithfulness_Correlation(
+                                x,
+                                all_attrs,
+                                pertrub="baseline",
+                                time_step=True,
+                                subset_size=4,
+                                nr_runs=100,
+                            )
+                        )
+                        f_v_score.append(
+                            self.Faithfulness_Correlation(
+                                x,
+                                all_attrs,
+                                pertrub="baseline",
+                                feature=True,
+                                subset_size=9,
+                                nr_runs=100,
+                            )
+                        )
+
+                        r_score.append(
+                            self.Data_Randomization(
+                                x,
+                                attribution=all_attrs,
+                                explain_method=method,
+                                random_model=random_model,
+                                method_name=method_name,
+                            )
+                        )
+                        res1, res2 = self.Relative_Stability(
+                            x,
+                            all_attrs,
+                            explain_method=method,
+                            method_name=method_name,
+                            dataloader=None,
+                            **kwargs,
+                        )
                         st_i_score.append(res1)
                         st_o_score.append(res2)
                     else:
-                        f_ts_score.append(self.Faithfulness_Correlation(x, timestep_attrs,
-                                                                        pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
-                        f_v_score.append(self.Faithfulness_Correlation(x, features_attrs,
-                                                                       pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
+                        f_ts_score.append(
+                            self.Faithfulness_Correlation(
+                                x,
+                                timestep_attrs,
+                                pertrub="baseline",
+                                time_step=True,
+                                subset_size=4,
+                                nr_runs=100,
+                            )
+                        )
+                        f_v_score.append(
+                            self.Faithfulness_Correlation(
+                                x,
+                                features_attrs,
+                                pertrub="baseline",
+                                feature=True,
+                                subset_size=9,
+                                nr_runs=100,
+                            )
+                        )
 
             # Faithfulness score for attribtuons of features per timesteps
             f_ts_v_score = np.mean(f_ts_v_score)
@@ -661,11 +788,20 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 r_score = np.mean(r_score)
                 st_i_score = np.max(st_i_score)
                 st_o_score = np.max(st_o_score)
-            return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score, st_i_score, st_o_score
+            return (
+                all_attrs,
+                features_attrs,
+                timestep_attrs,
+                f_ts_v_score,
+                f_ts_score,
+                f_v_score,
+                r_score,
+                st_i_score,
+                st_o_score,
+            )
 
         # Loop through the dataloader to compute attributions for all instances
         for batch in dataloader:
-
             for key, value in batch[0].items():
                 batch[0][key] = batch[0][key].to(self.device)
             x = batch[0]
@@ -673,8 +809,11 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             data, baselines = self.prep_data_captum(x)
 
             # Initialize the explanation method
-            explanation = method(self.forward_captum, interpretable_model=SkLearnLasso(
-                alpha=0.4)) if method_name == 'Lime' else method(self.forward_captum)
+            explanation = (
+                method(self.forward_captum, interpretable_model=SkLearnLasso(alpha=0.4))
+                if method_name == "Lime"
+                else method(self.forward_captum)
+            )
 
             # Calculate attributions using the selected method
             if method is not captum.attr.Saliency:
@@ -683,23 +822,61 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
                 attr = explanation.attribute(data, **kwargs)
 
             # Process and store the calculated attributions
-            stacked_attr = attr[1].cpu().detach().numpy() if method_name in [
-                'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
+            stacked_attr = (
+                attr[1].cpu().detach().numpy()
+                if method_name in ["Lime", "FeatureAblation"]
+                else torch.stack(attr).cpu().detach().numpy()
+            )
             if XAI_metric:
+                f_ts_v_score.append(
+                    self.Faithfulness_Correlation(
+                        x,
+                        stacked_attr,
+                        pertrub="baseline",
+                        feature_timestep=True,
+                        subset_size=[4, 9],
+                        nr_runs=100,
+                    )
+                )
 
-                f_ts_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
-                                                                  pertrub="baseline", feature_timestep=True, subset_size=[4, 9], nr_runs=100))
-
-                f_ts_score.append(self.Faithfulness_Correlation(x, stacked_attr,
-                                                                pertrub="baseline", time_step=True, subset_size=4, nr_runs=100))
-                f_v_score.append(self.Faithfulness_Correlation(x, stacked_attr,
-                                                               pertrub="baseline", feature=True, subset_size=9, nr_runs=100))
-                r_score.append(self.Data_Randomization(x, attribution=stacked_attr,
-                               explain_method=method, random_model=random_model, method_name=method_name))
+                f_ts_score.append(
+                    self.Faithfulness_Correlation(
+                        x,
+                        stacked_attr,
+                        pertrub="baseline",
+                        time_step=True,
+                        subset_size=4,
+                        nr_runs=100,
+                    )
+                )
+                f_v_score.append(
+                    self.Faithfulness_Correlation(
+                        x,
+                        stacked_attr,
+                        pertrub="baseline",
+                        feature=True,
+                        subset_size=9,
+                        nr_runs=100,
+                    )
+                )
+                r_score.append(
+                    self.Data_Randomization(
+                        x,
+                        attribution=stacked_attr,
+                        explain_method=method,
+                        random_model=random_model,
+                        method_name=method_name,
+                    )
+                )
 
-                res1, res2 = self.Relative_Stability(x,
-                                                     stacked_attr, explain_method=method, method_name=method_name,  dataloader=None, **kwargs
-                                                     )
+                res1, res2 = self.Relative_Stability(
+                    x,
+                    stacked_attr,
+                    explain_method=method,
+                    method_name=method_name,
+                    dataloader=None,
+                    **kwargs,
+                )
                 st_i_score.append(res1)
                 st_o_score.append(res2)
 
@@ -725,7 +902,7 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
         st_o_score = np.max(st_o_score)
 
         if plot:
-            log_dir_plots = log_dir / 'plots'
+            log_dir_plots = log_dir / "plots"
             if not (log_dir_plots.exists()):
                 log_dir_plots.mkdir(parents=True)
             # Plot attributions for features and timesteps
@@ -733,12 +910,22 @@ def explantation(self, dataloader, method, log_dir=".", plot=False, XAI_metric=F
             self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots)
 
         # Return computed attributions and metrics
-        return all_attrs, features_attrs, timestep_attrs, f_ts_v_score, f_ts_score, f_v_score, r_score, st_i_score, st_o_score
+        return (
+            all_attrs,
+            features_attrs,
+            timestep_attrs,
+            f_ts_v_score,
+            f_ts_score,
+            f_v_score,
+            r_score,
+            st_i_score,
+            st_o_score,
+        )
         # normalized_means = (means - means.min()) / (means.max() - means.min())
 
     def prep_data(self, x):
         """
-        Prepares data for custom forward method 
+        Prepares data for custom forward method
 
         Args:
             - x:Batch returned from dataloader
@@ -771,10 +958,9 @@ def Faithfulness_Correlation(
         feature=False,
         time_step=False,
         feature_timestep=False,
-
     ):
         """
-        Calculates faithfulness scores for captum attributions 
+        Calculates faithfulness scores for captum attributions
 
         Args:
             - x:Batch input
@@ -808,24 +994,27 @@ def Faithfulness_Correlation(
         References:
             1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
             explanations." IJCAI (2020): 3016-3022.
-            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for
+            responsible evaluation of neural network explanations and beyond."
+            Journal of Machine Learning Research 24.34 (2023): 1-11.
         """
+
         def add_noise(x, indices, time_step, feature_timestep):
             noise = torch.randn_like(x["encoder_cont"])
             if time_step:
-                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing='ij')
+                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
 
                 with torch.no_grad():
                     x["encoder_cont"][idx0, idx1, :] += noise[idx0, idx1, :]
 
             elif feature:
-                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing='ij')
+                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
 
                 with torch.no_grad():
                     x["encoder_cont"][idx0, :, idx1] += noise[idx0, :, idx1]
 
             elif feature_timestep:
-                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing='ij')
+                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
 
                 with torch.no_grad():
                     x["encoder_cont"][idx0, idx1, idx2] += noise[idx0, idx1, idx2]
@@ -833,22 +1022,28 @@ def add_noise(x, indices, time_step, feature_timestep):
         def apply_baseline(x, indices, time_step, feature_timestep):
             mask = torch.ones_like(x["encoder_cont"])
             if time_step:
-
-                idx0, idx1, = np.meshgrid(indices[0], indices[1], indexing='ij')
+                (
+                    idx0,
+                    idx1,
+                ) = np.meshgrid(indices[0], indices[1], indexing="ij")
 
                 mask[idx0, idx1, :] -= mask[idx0, idx1, :]
             elif feature:
-                idx0, idx1, = np.meshgrid(indices[0], indices[1], indexing='ij')
+                (
+                    idx0,
+                    idx1,
+                ) = np.meshgrid(indices[0], indices[1], indexing="ij")
 
                 mask[idx0, :, idx1] -= mask[idx0, :, idx1]
 
             elif feature_timestep:
-                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing='ij')
+                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
 
                 mask[idx0, idx1, idx2] -= mask[idx0, idx1, idx2]
 
             with torch.no_grad():
                 x["encoder_cont"] *= mask
+
         # Assuming 'attribution' is already a GPU tensor
         if not torch.is_tensor(attribution):
             attribution = torch.tensor(attribution).to(self.device)
@@ -892,22 +1087,17 @@ def apply_baseline(x, indices, time_step, feature_timestep):
             y_pred_perturb = (self(self.prep_data(x))).detach()  # Keep on GPU
 
             if time_step:
-
                 if attribution.size() == torch.Size([24]):
                     att_sums.append((attribution[timesteps_idx]).sum())
                 else:
                     att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :]).sum())
             elif feature:
-
                 if len(attribution) == 53:
                     att_sums.append((attribution[feature_idx]).sum())
                 else:
-
                     att_sums.append((attribution[patient_idx, :, :][:, :, feature_idx]).sum())
             elif feature_timestep:
-
-                att_sums.append((attribution[patient_idx, :, :]
-                                [:, timesteps_idx, :][:, :, feature_idx]).sum())
+                att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :][:, :, feature_idx]).sum())
 
             pred_deltas.append((y_pred - y_pred_perturb)[patient_idx].item())
             # Convert to CPU for numpy operations
@@ -921,17 +1111,24 @@ def apply_baseline(x, indices, time_step, feature_timestep):
         return score
 
     def Data_Randomization(
-        self, x,
-        attribution, explain_method, random_model, similarity_func=cosine, dataloader=None, method_name="", **kwargs
+        self,
+        x,
+        attribution,
+        explain_method,
+        random_model,
+        similarity_func=cosine,
+        dataloader=None,
+        method_name="",
+        **kwargs,
     ):
         """
 
         Args:
             - x:Batch input
-            -attribution: attribution 
+            -attribution: attribution
             - explain_method:function to generate explantations
             - random_model: Reference to model trained on random labels
-            - similarity_func: Function to measure similiarity 
+            - similarity_func: Function to measure similiarity
             - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
             - method_name: Name of the explantation
 
@@ -947,7 +1144,9 @@ def Data_Randomization(
         References:
             1) Leon Sixt et al.: "When Explanations Lie: Why Many Modified BP
             Attributions Fail." ICML (2020): 9046-9057.
-            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai
+            toolkit for responsible evaluation of neural network explanations and beyond."
+              Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
 
@@ -967,7 +1166,6 @@ def Data_Randomization(
             score = similarity_func(np.random.normal(size=[64, 24, 53]).flatten(), attribution.flatten())
         else:
             data, baselines = self.prep_data_captum(x)
-            y_pred = self(data).detach()
 
             explantation = explain_method(random_model.forward_captum)
             # Reformat attributions.
@@ -977,8 +1175,11 @@ def Data_Randomization(
                 attr = explantation.attribute(data, **kwargs)
 
             # Process and store the calculated attributions
-            random_attr = attr[1].cpu().detach().numpy() if method_name in [
-                'Lime', 'FeatureAblation'] else torch.stack(attr).cpu().detach().numpy()
+            random_attr = (
+                attr[1].cpu().detach().numpy()
+                if method_name in ["Lime", "FeatureAblation"]
+                else torch.stack(attr).cpu().detach().numpy()
+            )
 
             attribution = attribution.flatten()
             min_val = np.min(attribution)
@@ -992,32 +1193,37 @@ def Data_Randomization(
             score = similarity_func(random_attr, attribution)
         return score
 
-    def Relative_Stability(self, x,
-                           attribution, explain_method, method_name, dataloader=None, thershold=0.5, **kwargs
-                           ):
+    def Relative_Stability(
+        self,
+        x,
+        attribution,
+        explain_method,
+        method_name,
+        dataloader=None,
+        thershold=0.5,
+        **kwargs,
+    ):
         """
-     Args:
-            - x:Batch input
-            -attribution: attribution 
-            - explain_method:function to generate explantations
-            - method_name: Name of the explantation
-            - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+        Args:
+               - x:Batch input
+               -attribution: attribution
+               - explain_method:function to generate explantations
+               - method_name: Name of the explantation
+               - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
 
 
-        Returns:
-            RIS : relative distance between the explantation and the input
-            ROS: relative distance between the explantation and the output
+           Returns:
+               RIS : relative distance between the explantation and the input
+               ROS: relative distance between the explantation and the output
 
 
-     References:
-            1) `https://arxiv.org/pdf/2203.06877.pdf
-            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+        References:
+               1) `https://arxiv.org/pdf/2203.06877.pdf
+               2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
 
-        def relative_stability_objective(
-            x, xs, e_x, e_xs, eps_min=0.0001, input=False, device='cuda'
-        ) -> torch.Tensor:
+        def relative_stability_objective(x, xs, e_x, e_xs, eps_min=0.0001, input=False, device="cuda") -> torch.Tensor:
             """
             Computes relative input and output stabilities maximization objective
             as defined here :ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
@@ -1028,15 +1234,16 @@ def relative_stability_objective(
                 xs: perturbed tensor.
                 e_x: Explanations for x.
                 e_xs: Explanations for xs.
-                eps_min:Value to avoid division by zero if needed 
+                eps_min:Value to avoid division by zero if needed
                 input:Boolean to indicate if this is an input or an output
-                device: the device to keep the tensors on 
+                device: the device to keep the tensors on
 
             Returns:
 
                 ris_obj: Tensor
                     RIS maximization objective.
             """
+
             # Function to convert inputs to tensors if they are numpy arrays
             def to_tensor(input_array):
                 if isinstance(input_array, np.ndarray):
@@ -1052,11 +1259,19 @@ def to_tensor(input_array):
                 num_dim = e_x.ndim
 
             if num_dim == 3:
-                def norm_function(arr): return torch.norm(arr, dim=(-1, -2))
+
+                def norm_function(arr):
+                    return torch.norm(arr, dim=(-1, -2))
+
             elif num_dim == 2:
-                def norm_function(arr): return torch.norm(arr, dim=-1)
+
+                def norm_function(arr):
+                    return torch.norm(arr, dim=-1)
+
             else:
-                def norm_function(arr): return torch.norm(arr)
+
+                def norm_function(arr):
+                    return torch.norm(arr)
 
             nominator = (e_x - e_xs) / (e_x + (e_x == 0) * eps_min)
             nominator = norm_function(nominator)
@@ -1072,6 +1287,7 @@ def norm_function(arr): return torch.norm(arr)
                 denominator += (denominator == 0) * eps_min
 
             return nominator / denominator
+
         if not torch.is_tensor(attribution):
             attribution = torch.tensor(attribution).to(self.device)
         if explain_method == "Attention":
@@ -1089,11 +1305,19 @@ def norm_function(arr): return torch.norm(arr)
             # Find where the difference is less than or equal to a thershold
             close_indices = torch.nonzero(difference <= thershold).squeeze()
             RIS = relative_stability_objective(
-                x_original[close_indices, :, :].detach(), x_preturb[close_indices, :, :].detach(), attribution, att_preturb, input=True
+                x_original[close_indices, :, :].detach(),
+                x_preturb[close_indices, :, :].detach(),
+                attribution,
+                att_preturb,
+                input=True,
             )
 
             ROS = relative_stability_objective(
-                y_pred[close_indices], y_pred_preturb[close_indices], attribution, att_preturb, input=False
+                y_pred[close_indices],
+                y_pred_preturb[close_indices],
+                attribution,
+                att_preturb,
+                input=False,
             )
 
         else:
@@ -1101,15 +1325,13 @@ def norm_function(arr): return torch.norm(arr)
             x_original = x["encoder_cont"].detach().clone()
 
             with torch.no_grad():
-                noise = torch.randn_like(x["encoder_cont"])*0.01
+                noise = torch.randn_like(x["encoder_cont"]) * 0.01
                 x["encoder_cont"] += noise
             y_pred_preturb = self(self.prep_data(x)).detach()
             if explain_method == "Random":
-
                 att_preturb = np.random.normal(size=[64, 24, 53])
                 att_preturb = torch.tensor(att_preturb).to(self.device)
             else:
-
                 data, baselines = self.prep_data_captum(x)
 
                 explantation = explain_method(self.forward_captum)
@@ -1120,19 +1342,29 @@ def norm_function(arr): return torch.norm(arr)
                     att_preturb = explantation.attribute(data, **kwargs)
 
                 # Process and store the calculated attributions
-                att_preturb = att_preturb[1].detach() if method_name in [
-                    'Lime', 'FeatureAblation'] else torch.stack(att_preturb).detach()
+                att_preturb = (
+                    att_preturb[1].detach()
+                    if method_name in ["Lime", "FeatureAblation"]
+                    else torch.stack(att_preturb).detach()
+                )
             # Calculate the absolute difference
             difference = torch.abs(y_pred_preturb - y_pred)
 
             # Find where the difference is less than or equal to a thershold
             close_indices = torch.nonzero(difference <= thershold).squeeze()
             RIS = relative_stability_objective(
-                x_original[close_indices, :, :].detach(), x["encoder_cont"][close_indices, :, :].detach(), attribution[close_indices, :, :], att_preturb[close_indices, :, :], input=True
+                x_original[close_indices, :, :].detach(),
+                x["encoder_cont"][close_indices, :, :].detach(),
+                attribution[close_indices, :, :],
+                att_preturb[close_indices, :, :],
+                input=True,
             )
             ROS = relative_stability_objective(
-                y_pred[close_indices], y_pred_preturb[close_indices], attribution[close_indices,
-                                                                                  :, :], att_preturb[close_indices, :, :], input=False
+                y_pred[close_indices],
+                y_pred_preturb[close_indices],
+                attribution[close_indices, :, :],
+                att_preturb[close_indices, :, :],
+                input=False,
             )
 
         return np.max(RIS.cpu().numpy()).astype(np.float64), np.max(ROS.cpu().numpy()).astype(np.float64)
diff --git a/icu_benchmarks/run.py b/icu_benchmarks/run.py
index bd65c935..5c2d3bac 100644
--- a/icu_benchmarks/run.py
+++ b/icu_benchmarks/run.py
@@ -118,8 +118,7 @@ def main(my_args=tuple(sys.argv[1:])):
             name_datasets(args.name, args.name, args.name)
         run_dir = create_run_dir(log_dir)
         source_dir = args.source_dir
-        logging.info(
-            f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
+        logging.info(f"Will load weights from {source_dir} and bind train gin-config. Note: this might override your config.")
         gin.parse_config_file(source_dir / "train_config.gin")
     elif args.samples and args.source_dir is not None:  # Train model with limited samples and bind existing config
         logging.info("Binding train gin-config. Note: this might override your config.")
@@ -132,8 +131,7 @@ def main(my_args=tuple(sys.argv[1:])):
         name_datasets(args.name, args.name, args.name)
         hp_checkpoint = log_dir / args.hp_checkpoint if args.hp_checkpoint else None
         model_path = (
-            Path("configs") / ("imputation_models" if mode ==
-                               RunMode.imputation else "prediction_models") / f"{model}.gin"
+            Path("configs") / ("imputation_models" if mode == RunMode.imputation else "prediction_models") / f"{model}.gin"
         )
         gin_config_files = (
             [Path(f"configs/experiments/{args.experiment}.gin")]
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 969b4f4f..7d57bc2a 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -241,8 +241,9 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
     # Calculate the population standard deviation over aggregated results over folds/iterations
     # Divide by sqrt(n) to get standard deviation.
 
-    std_scores = {metric: (pstdev(list) / sqrt(len(list)))
-                  for metric, list in list_scores.items() if not (np.isnan(list).all())}
+    std_scores = {
+        metric: (pstdev(list) / sqrt(len(list))) for metric, list in list_scores.items() if not (np.isnan(list).all())
+    }
 
     confidence_interval = {
         metric: (stats.t.interval(0.95, len(list) - 1, loc=mean(list), scale=stats.sem(list)))
@@ -255,7 +256,7 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
         "CI_0.95": confidence_interval,
         "execution_time": execution_time.total_seconds() if execution_time is not None else 0.0,
     }
-    log_dir_plots = log_dir / 'plots'
+    log_dir_plots = log_dir / "plots"
     if not (log_dir_plots.exists()):
         log_dir_plots.mkdir(parents=True)
     # plot_XAI_Metrics(accumulated_metrics, log_dir_plots=log_dir_plots)
@@ -273,41 +274,48 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
 
 def plot_XAI_Metrics(accumulated_metrics, log_dir_plots):
     groups = {}
-    for key in accumulated_metrics['avg']:
-        if key in ['loss', 'MAE']:
+    for key in accumulated_metrics["avg"]:
+        if key in ["loss", "MAE"]:
             continue
-        suffix = key.split('_')[-1]
+        suffix = key.split("_")[-1]
         if suffix not in groups:
             groups[suffix] = []
         groups[suffix].append(key)
 
     # Define a dictionary for legend labels
     legend_labels = {
-        'IG': 'Integrated Gradient',
-        'G': 'Gradient',
-        'R': 'Random',
-        'FA': 'Feature Ablation',
-        'Att': 'Attention',
-        'VSN': 'Variable Selection Network',
-        'L': 'Lime'
+        "IG": "Integrated Gradient",
+        "G": "Gradient",
+        "R": "Random",
+        "FA": "Feature Ablation",
+        "Att": "Attention",
+        "VSN": "Variable Selection Network",
+        "L": "Lime",
     }
-    colors = ['navy', 'skyblue', 'crimson', 'salmon', 'teal', 'orange', 'darkgreen', 'lightgreen']
+    colors = [
+        "navy",
+        "skyblue",
+        "crimson",
+        "salmon",
+        "teal",
+        "orange",
+        "darkgreen",
+        "lightgreen",
+    ]
 
     # Plotting
     num_groups = len(groups)
     fig, axs = plt.subplots(num_groups, 1, figsize=(10, num_groups * 5))
 
     # Custom handles for the legend
-    handles = [plt.Rectangle((0, 0), 1, 1, color='none', label=f'{key}: {value}')
-               for key, value in legend_labels.items()]
+    # handles = [plt.Rectangle((0, 0), 1, 1, color="none", label=f"{key}: {value}") for key, value in legend_labels.items()]
 
     for i, (suffix, keys) in enumerate(groups.items()):
-
         ax = axs[i] if num_groups > 1 else axs
         # Extract values and errors
-        avg_values = [accumulated_metrics['avg'][key] for key in keys]
-        ci_lower = [accumulated_metrics['CI_0.95'][key][0] for key in keys]
-        ci_upper = [accumulated_metrics['CI_0.95'][key][1] for key in keys]
+        avg_values = [accumulated_metrics["avg"][key] for key in keys]
+        ci_lower = [accumulated_metrics["CI_0.95"][key][0] for key in keys]
+        ci_upper = [accumulated_metrics["CI_0.95"][key][1] for key in keys]
         ci_error = [np.abs([a - b, c - a]) for a, b, c in zip(avg_values, ci_lower, ci_upper)]
 
         # Sort by absolute values of avg_values
@@ -317,21 +325,27 @@ def plot_XAI_Metrics(accumulated_metrics, log_dir_plots):
         sorted_ci_error = np.array(ci_error)[sorted_indices]
 
         # Plot bars
-        bars = ax.bar(sorted_keys, np.abs(sorted_avg_values), yerr=np.array(sorted_ci_error).T, capsize=5, color=colors)
+        bars = ax.bar(
+            sorted_keys,
+            np.abs(sorted_avg_values),
+            yerr=np.array(sorted_ci_error).T,
+            capsize=5,
+            color=colors,
+        )
 
         # Set titles and labels
-        title_suffix = sorted_keys[0].split('_')[1]
+        title_suffix = sorted_keys[0].split("_")[1]
         ax.set_title(f'Metric: "{title_suffix}"')
-        ax.set_ylabel('Values')
-        ax.axhline(0, color='grey', linewidth=0.8)
-        ax.grid(axis='y')
+        ax.set_ylabel("Values")
+        ax.axhline(0, color="grey", linewidth=0.8)
+        ax.grid(axis="y")
 
         # Set x-ticks
         ax.set_xticks(sorted_keys)
-        ax.set_xticklabels([key.split('_')[0] for key in sorted_keys], rotation=45, ha="right")
+        ax.set_xticklabels([key.split("_")[0] for key in sorted_keys], rotation=45, ha="right")
         # Create a custom legend for each subplot
-        custom_labels = [legend_labels[key.split('_')[0]] for key in sorted_keys]
-        ax.legend(bars, custom_labels, loc='upper right')
+        custom_labels = [legend_labels[key.split("_")[0]] for key in sorted_keys]
+        ax.legend(bars, custom_labels, loc="upper right")
 
     plt.tight_layout()
     plt.savefig(log_dir_plots / "metrics_plot.png", bbox_inches="tight")
@@ -379,8 +393,7 @@ def load_pretrained_imputation_model(use_pretrained_imputation):
         pretrained_imputation_model_checkpoint = torch.load(use_pretrained_imputation, map_location=torch.device("cpu"))
         if isinstance(pretrained_imputation_model_checkpoint, dict):
             imputation_model_class = pretrained_imputation_model_checkpoint["class"]
-            pretrained_imputation_model = imputation_model_class(
-                **pretrained_imputation_model_checkpoint["hyper_parameters"])
+            pretrained_imputation_model = imputation_model_class(**pretrained_imputation_model_checkpoint["hyper_parameters"])
             pretrained_imputation_model.set_trained_columns(pretrained_imputation_model_checkpoint["trained_columns"])
             pretrained_imputation_model.load_state_dict(pretrained_imputation_model_checkpoint["state_dict"])
         else:

From 041b38b7200bc9ea9791e0885f1d26041f29ec47 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <Youssef.mecky96@gmail.com>
Date: Thu, 7 Mar 2024 14:56:41 +0100
Subject: [PATCH 130/142] formatting

---
 icu_benchmarks/imputation/diffwave.py       |  2 +-
 icu_benchmarks/imputation/layers/s4layer.py | 12 ++++++------
 icu_benchmarks/imputation/sssds4.py         |  2 +-
 icu_benchmarks/imputation/sssdsa.py         |  2 +-
 icu_benchmarks/models/custom_metrics.py     | 14 ++++++++++----
 icu_benchmarks/models/dl_models.py          |  2 +-
 icu_benchmarks/models/layers.py             |  6 +++---
 icu_benchmarks/models/similarity_func.py    |  2 +-
 icu_benchmarks/models/wrappers.py           |  3 ++-
 9 files changed, 26 insertions(+), 19 deletions(-)

diff --git a/icu_benchmarks/imputation/diffwave.py b/icu_benchmarks/imputation/diffwave.py
index c2f8ecf1..437303ed 100644
--- a/icu_benchmarks/imputation/diffwave.py
+++ b/icu_benchmarks/imputation/diffwave.py
@@ -330,7 +330,7 @@ def forward(self, input_data):
         cond = self.cond_conv(cond)
         h += cond
 
-        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels :, :])
+        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels:, :])
 
         res = self.res_conv(out)
         assert x.shape == res.shape
diff --git a/icu_benchmarks/imputation/layers/s4layer.py b/icu_benchmarks/imputation/layers/s4layer.py
index 1441a834..fd8a53a6 100644
--- a/icu_benchmarks/imputation/layers/s4layer.py
+++ b/icu_benchmarks/imputation/layers/s4layer.py
@@ -685,9 +685,9 @@ def forward(self, state=None, rate=1.0, L=None):
             k_f = r[:-1, :-1, :, :] - r[:-1, -1:, :, :] * r[-1:, :-1, :, :] / (1 + r[-1:, -1:, :, :])
         elif self.rank == 2:
             r00 = r[: -self.rank, : -self.rank, :, :]
-            r01 = r[: -self.rank, -self.rank :, :, :]
-            r10 = r[-self.rank :, : -self.rank, :, :]
-            r11 = r[-self.rank :, -self.rank :, :, :]
+            r01 = r[: -self.rank, -self.rank:, :, :]
+            r10 = r[-self.rank:, : -self.rank, :, :]
+            r11 = r[-self.rank:, -self.rank:, :, :]
             det = (1 + r11[:1, :1, :, :]) * (1 + r11[1:, 1:, :, :]) - r11[:1, 1:, :, :] * r11[1:, :1, :, :]
             s = (
                 r01[:, :1, :, :] * (1 + r11[1:, 1:, :, :]) * r10[:1, :, :, :]
@@ -699,9 +699,9 @@ def forward(self, state=None, rate=1.0, L=None):
             k_f = r00 - s
         else:
             r00 = r[: -self.rank, : -self.rank, :, :]
-            r01 = r[: -self.rank, -self.rank :, :, :]
-            r10 = r[-self.rank :, : -self.rank, :, :]
-            r11 = r[-self.rank :, -self.rank :, :, :]
+            r01 = r[: -self.rank, -self.rank:, :, :]
+            r10 = r[-self.rank:, : -self.rank, :, :]
+            r11 = r[-self.rank:, -self.rank:, :, :]
             r11 = rearrange(r11, "a b h n -> h n a b")
             r11 = torch.linalg.inv(torch.eye(self.rank, device=r.device) + r11)
             r11 = rearrange(r11, "h n a b -> a b h n")
diff --git a/icu_benchmarks/imputation/sssds4.py b/icu_benchmarks/imputation/sssds4.py
index 25933b68..205affa4 100644
--- a/icu_benchmarks/imputation/sssds4.py
+++ b/icu_benchmarks/imputation/sssds4.py
@@ -298,7 +298,7 @@ def forward(self, input_data):
 
         h = self.S42(h.permute(2, 0, 1)).permute(1, 2, 0)
 
-        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels :, :])
+        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels:, :])
 
         res = self.res_conv(out)
         assert x.shape == res.shape
diff --git a/icu_benchmarks/imputation/sssdsa.py b/icu_benchmarks/imputation/sssdsa.py
index 55764104..fb384efe 100644
--- a/icu_benchmarks/imputation/sssdsa.py
+++ b/icu_benchmarks/imputation/sssdsa.py
@@ -389,7 +389,7 @@ def step(self, x, state, **kwargs):
             if self.unet:
                 for i in range(skipped):
                     next_state.append(state.pop())
-                u_layers = list(self.u_layers)[skipped // 3 :]
+                u_layers = list(self.u_layers)[skipped // 3:]
             else:
                 for i in range(skipped):
                     for _ in range(len(self.u_layers[i])):
diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index 84b074cc..536a7e5c 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -194,7 +194,9 @@ def update(
         References:
             1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
             explanations." IJCAI (2020): 3016-3022.
-            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for
+            responsible evaluation of neural network explanations and beyond."
+            Journal of Machine Learning Research 24.34 (2023): 1-11.
         """
 
         def add_noise(x, indices, time_step, feature_timestep):
@@ -319,7 +321,7 @@ def update(self, x, attribution, model, explain_method, dataloader=None, thersho
                 -attribution: attribution
                 - explain_method:function to generate explantations
                 - method_name: Name of the explantation
-                - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+                - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch.
 
 
             Returns:
@@ -329,7 +331,9 @@ def update(self, x, attribution, model, explain_method, dataloader=None, thersho
 
         References:
                 1) `https://arxiv.org/pdf/2203.06877.pdf
-                2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+                2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for
+                responsible evaluation of neural network explanations and beyond."
+                Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
 
@@ -515,7 +519,9 @@ def update(self, x, attribution, model, explain_method, random_model, similarity
         References:
             1) Leon Sixt et al.: "When Explanations Lie: Why Many Modified BP
             Attributions Fail." ICML (2020): 9046-9057.
-            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for
+              responsible evaluation of neural network explanations and beyond."
+                Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 6141079f..98282bd4 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -324,7 +324,7 @@ def forward(self, x):
 @gin.configurable
 class TFT(DLPredictionWrapper):
     """
-    Implementation of https://arxiv.org/abs/1912.09363 
+    Implementation of https://arxiv.org/abs/1912.09363
     from https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/Forecasting/TFT
     """
 
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index c5fd7550..944f4b89 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -804,9 +804,9 @@ def forward(self, historical_inputs, cs, ch, cc, ce, future_inputs):
         x, attn_prob = self.attention(enriched)
 
         # Don't compute historical quantiles
-        x = x[:, self.encoder_length :, :]
-        temporal_features = temporal_features[:, self.encoder_length :, :]
-        enriched = enriched[:, self.encoder_length :, :]
+        x = x[:, self.encoder_length:, :]
+        temporal_features = temporal_features[:, self.encoder_length:, :]
+        enriched = enriched[:, self.encoder_length:, :]
 
         x = self.attention_gate(x)
         x = x + enriched
diff --git a/icu_benchmarks/models/similarity_func.py b/icu_benchmarks/models/similarity_func.py
index fd931e7e..bc49850f 100644
--- a/icu_benchmarks/models/similarity_func.py
+++ b/icu_benchmarks/models/similarity_func.py
@@ -1,4 +1,4 @@
-"""This module holds a collection of similarity functions i.e., 
+"""This module holds a collection of similarity functions i.e.,
 ways to measure the distance between two inputs (or explanations)."""
 
 # This file is part of Quantus.
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index f22ecf85..f5456a08 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -1219,7 +1219,8 @@ def Relative_Stability(
 
         References:
                1) `https://arxiv.org/pdf/2203.06877.pdf
-               2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+               2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation
+               of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
 
         """
 

From 2ab11a50b000f74a9ba1a6651c32eee5833cf693 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <Youssef.mecky96@gmail.com>
Date: Thu, 7 Mar 2024 16:04:57 +0100
Subject: [PATCH 131/142] formatting

---
 icu_benchmarks/models/custom_metrics.py |  99 ++++----
 icu_benchmarks/models/train.py          | 292 ++++++++++++++----------
 icu_benchmarks/models/wrappers.py       | 107 +++++----
 3 files changed, 277 insertions(+), 221 deletions(-)

diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index 536a7e5c..d1526e6d 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -145,6 +145,53 @@ class Faithfulness(EpochMetric):
     def __init__(self, output_transform: Callable = lambda x: x, check_compute_fn: bool = False, *args, **kwargs) -> None:
         super().__init__(output_transform, check_compute_fn, *args, **kwargs)
 
+    def add_noise(self, x, indices, time_step, feature, feature_timestep):
+        noise = torch.randn_like(x["encoder_cont"])
+        if time_step:
+            idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
+
+            with torch.no_grad():
+                x["encoder_cont"][idx0, idx1, :] += noise[idx0, idx1, :]
+
+        elif feature:
+            idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
+
+            with torch.no_grad():
+                x["encoder_cont"][idx0, :, idx1] += noise[idx0, :, idx1]
+
+        elif feature_timestep:
+            idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
+
+            with torch.no_grad():
+                x["encoder_cont"][idx0, idx1, idx2] += noise[idx0, idx1, idx2]
+        return x
+
+    def apply_baseline(self, x, indices, time_step, feature, feature_timestep):
+        mask = torch.ones_like(x["encoder_cont"])
+        if time_step:
+            (
+                idx0,
+                idx1,
+            ) = np.meshgrid(indices[0], indices[1], indexing="ij")
+
+            mask[idx0, idx1, :] -= mask[idx0, idx1, :]
+        elif feature:
+            (
+                idx0,
+                idx1,
+            ) = np.meshgrid(indices[0], indices[1], indexing="ij")
+
+            mask[idx0, :, idx1] -= mask[idx0, :, idx1]
+
+        elif feature_timestep:
+            idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
+
+            mask[idx0, idx1, idx2] -= mask[idx0, idx1, idx2]
+
+        with torch.no_grad():
+            x["encoder_cont"] *= mask
+        return x
+
     def update(
         self,
         x,
@@ -199,54 +246,8 @@ def update(
             Journal of Machine Learning Research 24.34 (2023): 1-11.
         """
 
-        def add_noise(x, indices, time_step, feature_timestep):
-            noise = torch.randn_like(x["encoder_cont"])
-            if time_step:
-                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
-
-                with torch.no_grad():
-                    x["encoder_cont"][idx0, idx1, :] += noise[idx0, idx1, :]
-
-            elif feature:
-                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
-
-                with torch.no_grad():
-                    x["encoder_cont"][idx0, :, idx1] += noise[idx0, :, idx1]
-
-            elif feature_timestep:
-                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
-
-                with torch.no_grad():
-                    x["encoder_cont"][idx0, idx1, idx2] += noise[idx0, idx1, idx2]
-
-        def apply_baseline(x, indices, time_step, feature_timestep):
-            mask = torch.ones_like(x["encoder_cont"])
-            if time_step:
-                (
-                    idx0,
-                    idx1,
-                ) = np.meshgrid(indices[0], indices[1], indexing="ij")
-
-                mask[idx0, idx1, :] -= mask[idx0, idx1, :]
-            elif feature:
-                (
-                    idx0,
-                    idx1,
-                ) = np.meshgrid(indices[0], indices[1], indexing="ij")
-
-                mask[idx0, :, idx1] -= mask[idx0, :, idx1]
-
-            elif feature_timestep:
-                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
-
-                mask[idx0, idx1, idx2] -= mask[idx0, idx1, idx2]
-
-            with torch.no_grad():
-                x["encoder_cont"] *= mask
-
         # Assuming 'attribution' is already a GPU tensor
-        if not torch.is_tensor(attribution):
-            attribution = torch.tensor(attribution).to(device)
+        attribution = torch.tensor(attribution).to(device)
         # Other initializations
         if similarity_func is None:
             similarity_func = correlation_spearman
@@ -278,9 +279,9 @@ def apply_baseline(x, indices, time_step, feature_timestep):
 
             # Apply perturbation
             if pertrub == "Noise":
-                add_noise(x, a_ix, time_step, feature_timestep)
+                x = self.add_noise(x, a_ix, time_step, feature, feature_timestep)
             elif pertrub == "baseline":
-                apply_baseline(x, a_ix, time_step, feature_timestep)
+                x = self.apply_baseline(x, a_ix, time_step, feature, feature_timestep)
 
             # Predict on perturbed input and calculate deltas
             y_pred_perturb = (model(model.prep_data(x))).detach()  # Keep on GPU
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index e7ed5d68..a911a7ac 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -119,6 +119,7 @@ def train_common(
     batch_size = min(batch_size, len(train_dataset), len(val_dataset))
     test_dataset = dataset_class(data, split=test_on, name=dataset_names["test"])
     test_dataset = assure_minimum_length(test_dataset)
+
     if not eval_only:
         logging.info(
             f"Training on {train_dataset.name} with {len(train_dataset)} samples and validating on {val_dataset.name} with"
@@ -126,87 +127,24 @@ def train_common(
         )
     batch_size = int(batch_size)
     logging.info(f"Using {num_workers} workers for data loading.")
-    if pytorch_forecasting:
-        train_loader = train_dataset.to_dataloader(
-            train=True,
-            batch_size=batch_size,
-            num_workers=num_workers,
-            pin_memory=False,
-            drop_last=True,
-        )
-        val_loader = val_dataset.to_dataloader(
-            train=False,
-            batch_size=batch_size,
-            num_workers=num_workers,
-            pin_memory=False,
-            drop_last=True,
-        )
-        test_loader = test_dataset.to_dataloader(
-            train=False,
-            batch_size=batch_size,
-            num_workers=num_workers,
-            pin_memory=False,
-            drop_last=True,
-            shuffle=False,
-        )
-        if load_weights:
-            model = load_model(
-                model,
-                source_dir,
-                pl_model=pl_model,
-                train_dataset=train_dataset,
-                optimizer=optimizer,
-            )
-
-        else:
-            model = model(
-                train_dataset,
-                optimizer=optimizer,
-                epochs=epochs,
-                run_mode=mode,
-                batch_size=batch_size,
-            )
-        if random_labels:
-            train_dataset.randomize_labels(num_classes=model.num_classes)
-            val_dataset.randomize_labels(num_classes=model.num_classes)
-            test_dataset.randomize_labels(num_classes=model.num_classes)
-
-    else:
-        train_loader = DataLoader(
-            train_dataset,
-            batch_size=batch_size,
-            shuffle=True,
-            num_workers=num_workers,
-            pin_memory=True,
-            drop_last=True,
-        )
-        val_loader = DataLoader(
-            val_dataset,
-            batch_size=batch_size,
-            shuffle=False,
-            num_workers=num_workers,
-            pin_memory=True,
-            drop_last=True,
-        )
-
-        test_loader = (
-            DataLoader(
-                test_dataset,
-                batch_size=min(batch_size * 4, len(test_dataset)),
-                shuffle=False,
-                num_workers=num_workers,
-                pin_memory=True,
-                drop_last=True,
-            )
-            if model.requires_backprop
-            else DataLoader([test_dataset.to_tensor()], batch_size=1)
-        )
-
-        data_shape = next(iter(train_loader))[0].shape
-        if load_weights:
-            model = load_model(model, source_dir, pl_model=pl_model)
-        else:
-            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
+    train_loader, val_loader, test_loader, model = prepare_data_loaders(
+        model=model,
+        train_dataset=train_dataset,
+        val_dataset=val_dataset,
+        test_dataset=test_dataset,
+        batch_size=batch_size,
+        num_workers=num_workers,
+        pin_memory=True,
+        drop_last=True,
+        pytorch_forecasting=pytorch_forecasting,
+        load_weights=load_weights,
+        source_dir=source_dir,
+        pl_model=pl_model,
+        optimizer=optimizer,
+        epochs=epochs,
+        mode=mode,
+        random_labels=random_labels,
+    )
 
     model.set_weight(weight, train_dataset)
     model.set_trained_columns(train_dataset.get_feature_names())
@@ -241,7 +179,6 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
-        # gradient_clip_val=gradient_clip_val
     )
     if not eval_only:
         if model.requires_backprop:
@@ -260,18 +197,14 @@ def train_common(
         return 0
 
     if explain:
-        if random_model_dir is None:
-            random_model = None
-        else:
-            path = Path(random_model_dir)
-
-            random_model = load_model(
-                model,
-                source_dir=path,
-                pl_model=pl_model,
-                train_dataset=train_dataset,
-                optimizer=optimizer,
-            )
+        path = Path(random_model_dir)
+        random_model = load_model(
+            model,
+            source_dir=path,
+            pl_model=pl_model,
+            train_dataset=train_dataset,
+            optimizer=optimizer,
+        )
 
         XAI_dict = {}  # dictrionary to log attributions metrics
 
@@ -324,21 +257,11 @@ def train_common(
                     torch.Size([64, 2]),
                 ]
 
-                # Create a default mask for non-targeted tensors
-                def create_default_mask(shape):
-                    if len(shape) == 3:
-                        return torch.zeros(shape[0], shape[1], max(1, shape[2]), dtype=torch.int32)
-                    elif len(shape) == 2:
-                        return torch.zeros(shape[0], max(1, shape[1]), dtype=torch.int32)
-                    else:  # len(shape) == 1
-                        return torch.zeros(shape[0], dtype=torch.int32)
-
                 # Create a feature mask for the second tensor that includes both features and timesteps
                 num_timesteps = shapes[1][1]
                 num_features = shapes[1][2]
                 feature_mask_second = torch.arange(num_timesteps * num_features).reshape(num_timesteps, num_features)
                 feature_mask_second = feature_mask_second.unsqueeze(0).repeat(shapes[1][0], 1, 1)
-
                 # Create a tuple of masks
                 feature_masks = tuple(
                     [create_default_mask(shape) if i != 1 else feature_mask_second for i, shape in enumerate(shapes)]
@@ -392,24 +315,18 @@ def create_default_mask(shape):
                 XAI_dict["{}_ROS".format(key)] = st_o_score
                 print("{}_RIS ".format(key), st_i_score)
                 XAI_dict["{}_RIS".format(key)] = st_i_score
-                if key == "Att":
-                    print("{} weights faithfulness featrues ".format(key), v_score)
-                    XAI_dict["{}_Faith Features".format(key)] = v_score
 
-                else:
-                    print("{} Attributions faithfulness featrues ".format(key), v_score)
-                    XAI_dict["{}_Faith Features".format(key)] = v_score
-
-                    print(
-                        "{}_Attributions Faithfulness Variable Per Timestep ".format(key),
-                        ts_v_score,
-                    )
-                    XAI_dict["{}_Faith Variable Per Timestep".format(key)] = ts_v_score
+                print("{} Attributions faithfulness featrues ".format(key), v_score)
+                XAI_dict["{}_Faith Features".format(key)] = v_score
+
+                print(
+                    "{}_Attributions Faithfulness Variable Per Timestep ".format(key),
+                    ts_v_score,
+                )
+                XAI_dict["{}_Faith Variable Per Timestep".format(key)] = ts_v_score
                 print("{}_Data Randomization Distance ".format(key), r_score)
                 XAI_dict["{}_Data Randomization Distance".format(key)] = r_score
 
-        # Getting the interpertations using pytorch forecasting native methods
-
         # Path to the JSON file in log_dir
         json_file_path = f"{log_dir}/XAI_metrics.json"
 
@@ -423,7 +340,146 @@ def create_default_mask(shape):
     return test_loss
 
 
+def prepare_data_loaders(
+    model,
+    train_dataset,
+    val_dataset,
+    test_dataset,
+    batch_size,
+    num_workers,
+    pin_memory,
+    drop_last=True,
+    shuffle_train=True,
+    pytorch_forecasting=False,
+    load_weights=False,
+    source_dir=None,
+    pl_model=None,
+    optimizer=None,
+    epochs=None,
+    mode=None,
+    random_labels=False,
+):
+    """
+    Prepares PyTorch data loaders based on the provided datasets and configuration.
+
+    Args:
+        train_dataset: Training dataset.
+        val_dataset: Validation dataset.
+        test_dataset: Test dataset.
+        batch_size: Batch size for data loaders.
+        num_workers: Number of worker processes for data loading.
+        pin_memory: Whether to use pin_memory for faster data transfer to GPU.
+        drop_last: Whether to drop the last incomplete batch.
+        shuffle_train: Whether to shuffle the training data loader.
+        load_weights: Whether to load weights from a pre-trained model.
+        source_dir: Directory to load weights from.
+        pl_model: PyTorch Lightning model (used for loading weights).
+        optimizer: Optimizer for the model.
+        epochs: Number of training epochs.
+        mode: Run mode for the model.
+        random_labels: Whether to randomize labels for the datasets.
+
+    Returns:
+        tuple: Tuple containing train_loader, val_loader, and test_loader.
+    """
+    if pytorch_forecasting:
+        train_loader = train_dataset.to_dataloader(
+            train=True,
+            batch_size=batch_size,
+            num_workers=num_workers,
+            pin_memory=pin_memory,
+            drop_last=drop_last,
+            shuffle=shuffle_train,
+        )
+        val_loader = val_dataset.to_dataloader(
+            train=False,
+            batch_size=batch_size,
+            num_workers=num_workers,
+            pin_memory=pin_memory,
+            drop_last=drop_last,
+        )
+        test_loader = test_dataset.to_dataloader(
+            train=False,
+            batch_size=batch_size,
+            num_workers=num_workers,
+            pin_memory=pin_memory,
+            drop_last=drop_last,
+            shuffle=False,
+        )
+        if load_weights:
+            model = load_model(
+                model,
+                source_dir,
+                pl_model=pl_model,
+                train_dataset=train_dataset,
+                optimizer=optimizer,
+            )
+        else:
+            model = model(
+                train_dataset,
+                optimizer=optimizer,
+                epochs=epochs,
+                run_mode=mode,
+                batch_size=batch_size,
+            )
+        if random_labels:
+            train_dataset.randomize_labels(num_classes=model.num_classes)
+            val_dataset.randomize_labels(num_classes=model.num_classes)
+            test_dataset.randomize_labels(num_classes=model.num_classes)
+
+    else:
+        train_loader = DataLoader(
+            train_dataset,
+            batch_size=batch_size,
+            shuffle=shuffle_train,
+            num_workers=num_workers,
+            pin_memory=pin_memory,
+            drop_last=drop_last,
+        )
+        val_loader = DataLoader(
+            val_dataset,
+            batch_size=batch_size,
+            shuffle=False,
+            num_workers=num_workers,
+            pin_memory=pin_memory,
+            drop_last=drop_last,
+        )
+
+        test_loader = (
+            DataLoader(
+                test_dataset,
+                batch_size=min(batch_size * 4, len(test_dataset)),
+                shuffle=False,
+                num_workers=num_workers,
+                pin_memory=pin_memory,
+                drop_last=drop_last,
+            )
+            if model.requires_backprop
+            else DataLoader([test_dataset.to_tensor()], batch_size=1)
+        )
+
+        data_shape = next(iter(train_loader))[0].shape
+        if load_weights:
+            model = load_model(model, source_dir, pl_model=pl_model)
+        else:
+            model = model(optimizer=optimizer, input_size=data_shape, epochs=epochs, run_mode=mode)
+
+    return train_loader, val_loader, test_loader, model
+
+
+def create_default_mask(shape):
+    if len(shape) == 3:
+        return torch.zeros(shape[0], shape[1], max(1, shape[2]), dtype=torch.int32)
+    elif len(shape) == 2:
+        return torch.zeros(shape[0], max(1, shape[1]), dtype=torch.int32)
+    else:  # len(shape) == 1
+        return torch.zeros(shape[0], dtype=torch.int32)
+
+
 def load_model(model, source_dir, pl_model=True, train_dataset=None, optimizer=None):
+    if source_dir is None:
+        return None
+
     if source_dir.exists():
         if model.requires_backprop:
             if (source_dir / "model.ckpt").exists():
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index f5456a08..5792a5bb 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -901,13 +901,13 @@ def explantation(
         st_i_score = np.max(st_i_score)
         st_o_score = np.max(st_o_score)
 
-        if plot:
+        """ if plot:
             log_dir_plots = log_dir / "plots"
             if not (log_dir_plots.exists()):
                 log_dir_plots.mkdir(parents=True)
             # Plot attributions for features and timesteps
 
-            self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots)
+            self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots) """
 
         # Return computed attributions and metrics
         return (
@@ -947,6 +947,53 @@ def prep_data(self, x):
         )
         return data
 
+    def add_noise(self, x, indices, time_step, feature, feature_timestep):
+        noise = torch.randn_like(x["encoder_cont"])
+        if time_step:
+            idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
+
+            with torch.no_grad():
+                x["encoder_cont"][idx0, idx1, :] += noise[idx0, idx1, :]
+
+        elif feature:
+            idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
+
+            with torch.no_grad():
+                x["encoder_cont"][idx0, :, idx1] += noise[idx0, :, idx1]
+
+        elif feature_timestep:
+            idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
+
+            with torch.no_grad():
+                x["encoder_cont"][idx0, idx1, idx2] += noise[idx0, idx1, idx2]
+        return x
+
+    def apply_baseline(self, x, indices, time_step, feature, feature_timestep):
+        mask = torch.ones_like(x["encoder_cont"])
+        if time_step:
+            (
+                idx0,
+                idx1,
+            ) = np.meshgrid(indices[0], indices[1], indexing="ij")
+
+            mask[idx0, idx1, :] -= mask[idx0, idx1, :]
+        elif feature:
+            (
+                idx0,
+                idx1,
+            ) = np.meshgrid(indices[0], indices[1], indexing="ij")
+
+            mask[idx0, :, idx1] -= mask[idx0, :, idx1]
+
+        elif feature_timestep:
+            idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
+
+            mask[idx0, idx1, idx2] -= mask[idx0, idx1, idx2]
+
+        with torch.no_grad():
+            x["encoder_cont"] *= mask
+        return x
+
     def Faithfulness_Correlation(
         self,
         x,
@@ -999,54 +1046,7 @@ def Faithfulness_Correlation(
             Journal of Machine Learning Research 24.34 (2023): 1-11.
         """
 
-        def add_noise(x, indices, time_step, feature_timestep):
-            noise = torch.randn_like(x["encoder_cont"])
-            if time_step:
-                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
-
-                with torch.no_grad():
-                    x["encoder_cont"][idx0, idx1, :] += noise[idx0, idx1, :]
-
-            elif feature:
-                idx0, idx1 = np.meshgrid(indices[0], indices[1], indexing="ij")
-
-                with torch.no_grad():
-                    x["encoder_cont"][idx0, :, idx1] += noise[idx0, :, idx1]
-
-            elif feature_timestep:
-                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
-
-                with torch.no_grad():
-                    x["encoder_cont"][idx0, idx1, idx2] += noise[idx0, idx1, idx2]
-
-        def apply_baseline(x, indices, time_step, feature_timestep):
-            mask = torch.ones_like(x["encoder_cont"])
-            if time_step:
-                (
-                    idx0,
-                    idx1,
-                ) = np.meshgrid(indices[0], indices[1], indexing="ij")
-
-                mask[idx0, idx1, :] -= mask[idx0, idx1, :]
-            elif feature:
-                (
-                    idx0,
-                    idx1,
-                ) = np.meshgrid(indices[0], indices[1], indexing="ij")
-
-                mask[idx0, :, idx1] -= mask[idx0, :, idx1]
-
-            elif feature_timestep:
-                idx0, idx1, idx2 = np.meshgrid(indices[0], indices[1], indices[2], indexing="ij")
-
-                mask[idx0, idx1, idx2] -= mask[idx0, idx1, idx2]
-
-            with torch.no_grad():
-                x["encoder_cont"] *= mask
-
-        # Assuming 'attribution' is already a GPU tensor
-        if not torch.is_tensor(attribution):
-            attribution = torch.tensor(attribution).to(self.device)
+        attribution = torch.tensor(attribution).to(self.device)
 
         # Other initializations
         if similarity_func is None:
@@ -1079,9 +1079,9 @@ def apply_baseline(x, indices, time_step, feature_timestep):
 
             # Apply perturbation
             if pertrub == "Noise":
-                add_noise(x, a_ix, time_step, feature_timestep)
+                x = self.add_noise(x, a_ix, time_step, feature, feature_timestep)
             elif pertrub == "baseline":
-                apply_baseline(x, a_ix, time_step, feature_timestep)
+                x = self.apply_baseline(x, a_ix, time_step, feature, feature_timestep)
 
             # Predict on perturbed input and calculate deltas
             y_pred_perturb = (self(self.prep_data(x))).detach()  # Keep on GPU
@@ -1289,8 +1289,7 @@ def norm_function(arr):
 
             return nominator / denominator
 
-        if not torch.is_tensor(attribution):
-            attribution = torch.tensor(attribution).to(self.device)
+        attribution = torch.tensor(attribution).to(self.device)
         if explain_method == "Attention":
             y_pred = self.model.predict(dataloader)
             x_original = dataloader.dataset.data["reals"].clone()

From 1f904dccb4419b9d1b11ad7e1c6feaaebc87b678 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <Youssef.mecky96@gmail.com>
Date: Fri, 8 Mar 2024 13:58:33 +0100
Subject: [PATCH 132/142] cleaning up and formatting

---
 configs/prediction_models/TFT.gin             |  25 -
 configs/prediction_models/common/DLTuning.gin |   2 +-
 configs/tasks/Regression.gin                  |   2 +-
 configs/tasks/RegressionTFT.gin               |  33 -
 configs/tasks/common/CrossValidation.gin      |   2 +-
 environment.yml                               |   1 +
 experiments/benchmark_classification.yml      |  39 +-
 experiments/benchmark_regression.yml          |  31 +-
 icu_benchmarks/cross_validation.py            |   2 +-
 icu_benchmarks/data/Checking data-Copy1.ipynb | 582 ------------------
 icu_benchmarks/data/constants.py              |   4 -
 icu_benchmarks/imputation/diffwave.py         |   2 +-
 icu_benchmarks/models/constants.py            |   9 +-
 icu_benchmarks/models/custom_metrics.py       |   2 +-
 icu_benchmarks/models/similarity_func.py      | 288 ---------
 icu_benchmarks/models/wrappers.py             |   2 +-
 icu_benchmarks/run_utils.py                   | 207 +------
 saved_dictionary.pkl                          | Bin 20818976 -> 0 bytes
 18 files changed, 63 insertions(+), 1170 deletions(-)
 delete mode 100644 configs/prediction_models/TFT.gin
 delete mode 100644 configs/tasks/RegressionTFT.gin
 delete mode 100644 icu_benchmarks/data/Checking data-Copy1.ipynb
 delete mode 100644 icu_benchmarks/models/similarity_func.py
 delete mode 100644 saved_dictionary.pkl

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
deleted file mode 100644
index 09fcf537..00000000
--- a/configs/prediction_models/TFT.gin
+++ /dev/null
@@ -1,25 +0,0 @@
-# Settings for TFT model.
-
-# Common settings for DL models
-include "configs/prediction_models/common/DLCommon.gin"
-
-# Optimizer params
-train_common.model = @TFT
-
-optimizer/hyperparameter.class_to_tune = @Adam
-optimizer/hyperparameter.weight_decay = 1e-6
-optimizer/hyperparameter.lr = (1e-5, 3e-4)
-
-# Encoder params
-model/hyperparameter.class_to_tune = @TFT
-model/hyperparameter.encoder_length = 24
-model/hyperparameter.hidden = (32,64,128,256)
-model/hyperparameter.num_classes = %NUM_CLASSES
-model/hyperparameter.dropout = (0.0, 0.4)
-model/hyperparameter.dropout_att = (0.0, 0.4)
-model/hyperparameter.n_heads =(1,2,4)
-model/hyperparameter.example_length=25
-
-
-
-
diff --git a/configs/prediction_models/common/DLTuning.gin b/configs/prediction_models/common/DLTuning.gin
index b631dc89..04fb8ff6 100644
--- a/configs/prediction_models/common/DLTuning.gin
+++ b/configs/prediction_models/common/DLTuning.gin
@@ -1,5 +1,5 @@
 # Hyperparameter tuner settings for Deep Learning.
 tune_hyperparameters.scopes = ["model", "optimizer","train_common"]
 tune_hyperparameters.n_initial_points = 5
-tune_hyperparameters.n_calls = 15
+tune_hyperparameters.n_calls = 30
 tune_hyperparameters.folds_to_tune_on = 2
\ No newline at end of file
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index 75041763..6ce4c0f6 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -15,7 +15,7 @@ train_common.weight = "balanced"
 train_common.ram_cache = True
 
 # LOSS FUNCTION
-DLPredictionWrapper.loss = @l1_loss
+DLPredictionWrapper.loss = @mse_loss
 DLPredictionPytorchForecastingWrapper.loss = @l1_loss
 MLWrapper.loss = @mean_squared_error
 
diff --git a/configs/tasks/RegressionTFT.gin b/configs/tasks/RegressionTFT.gin
deleted file mode 100644
index a1ff4d55..00000000
--- a/configs/tasks/RegressionTFT.gin
+++ /dev/null
@@ -1,33 +0,0 @@
-# COMMON IMPORTS
-include "configs/tasks/common/Imports.gin"
-
-# DATASET CONFIGURATION
-include "configs/tasks/common/PredictionTaskVariables.gin"
-
-# CROSS-VALIDATION
-include "configs/tasks/common/CrossValidation.gin"
-
-# MODE SETTINGS
-Run.mode = "Regression"
-NUM_CLASSES = 1
-HORIZON = 24
-train_common.weight = "balanced"
-train_common.ram_cache = True
-
-# LOSS FUNCTION
-DLPredictionWrapper.loss = @QuantileLoss
-QuantileLoss.quantiles=[0.1, 0.5, 0.9]
-
-# SELECTING PREPROCESSOR
-preprocess.preprocessor = @base_regression_preprocessor
-preprocess.vars = %vars
-preprocess.use_static = True
-
-# SPECIFYING REGRESSION OUTCOME SCALING
-base_regression_preprocessor.outcome_min = 0
-base_regression_preprocessor.outcome_max = 15
-
-# SELECTING DATASET
-PredictionDataset.vars = %vars
-PredictionDataset.ram_cache = True
-
diff --git a/configs/tasks/common/CrossValidation.gin b/configs/tasks/common/CrossValidation.gin
index 519ad3dd..f3efd9ef 100644
--- a/configs/tasks/common/CrossValidation.gin
+++ b/configs/tasks/common/CrossValidation.gin
@@ -1,3 +1,3 @@
 # CROSS-VALIDATION SETTINGS
-execute_repeated_cv.cv_repetitions = 2
+execute_repeated_cv.cv_repetitions = 5
 execute_repeated_cv.cv_folds = 5
\ No newline at end of file
diff --git a/environment.yml b/environment.yml
index 2232e1a3..b48befc7 100644
--- a/environment.yml
+++ b/environment.yml
@@ -28,6 +28,7 @@ dependencies:
   - pip=23.1
   - einops=0.6.1
   - hydra-core=1.3
+  - quantus=0.5.3
   - pip:
       - recipies==0.1.3
       # Fixed version because of NumPy incompatibility and stale development status.
diff --git a/experiments/benchmark_classification.yml b/experiments/benchmark_classification.yml
index bc1c55c6..61763ffa 100644
--- a/experiments/benchmark_classification.yml
+++ b/experiments/benchmark_classification.yml
@@ -17,29 +17,26 @@ name: yaib_classification_benchmark
 parameters:
   data_dir:
     values:
-    #  - data/YAIB_Preprint/aki/miiv
-    #  - data/YAIB_Preprint/aki/hirid
-    #  - data/YAIB_Preprint/aki/eicu
-    #  - data/YAIB_Preprint/aki/aumc
-    #  - data/YAIB_Preprint/mortality24/miiv
-    #  - data/YAIB_Preprint/mortality24/hirid
-    #  - data/YAIB_Preprint/mortality24/eicu
-    #  - data/YAIB_Preprint/mortality24/aumc
-      - data/YAIB_Preprint/sepsis/miiv
-      - data/YAIB_Preprint/sepsis/hirid
-      - data/YAIB_Preprint/sepsis/eicu
-      - data/YAIB_Preprint/sepsis/aumc
+      - ../data/mortality24/miiv
+      - ../data/mortality24/hirid
+      - ../data/mortality24/eicu
+      - ../data/mortality24/aumc
+      - ../data/aki/miiv
+      - ../data/aki/hirid
+      - ../data/aki/eicu
+      - ../data/aki/aumc
+      - ../data/sepsis/miiv
+      - ../data/sepsis/hirid
+      - ../data/sepsis/eicu
+      - ../data/sepsis/aumc
   model:
     values:
-    #  - LogisticRegression
-    #  - LGBMClassifier
-    #  - GRU
-    #  - LSTM
-    #  - TCN
-    #  - Transformer
-    #  - TFT
-       - TFTpytorch
-      
+      - LogisticRegression
+      - LGBMClassifier
+      - GRU
+      - LSTM
+      - TCN
+      - Transformer
   seed:
     values:
       - 1111
diff --git a/experiments/benchmark_regression.yml b/experiments/benchmark_regression.yml
index c13929eb..1f9176f3 100644
--- a/experiments/benchmark_regression.yml
+++ b/experiments/benchmark_regression.yml
@@ -17,25 +17,22 @@ name: yaib_regression_benchmark
 parameters:
   data_dir:
     values:
-      - data/YAIB_Preprint/los/miiv
-      - data/YAIB_Preprint/los/hirid
-      - data/YAIB_Preprint/los/eicu
-      - data/YAIB_Preprint/los/aumc
-      - data/YAIB_Preprint/kf/miiv
-      - data/YAIB_Preprint/kf/hirid
-      - data/YAIB_Preprint/kf/eicu
-      - data/YAIB_Preprint/kf/aumc
+      - ../data/los/miiv
+      - ../data/los/hirid
+      - ../data/los/eicu
+      - ../data/los/aumc
+      - ../data/kidney_function/miiv
+      - ../data/kidney_function/hirid
+      - ../data/kidney_function/eicu
+      - ../data/kidney_function/aumc
   model:
     values:
-     # - ElasticNet
-     # - LGBMRegressor
-     # - GRU
-     # - LSTM
-     # - TCN
-     # - Transformer
-     # - TFT
-      - TFTpytorch
-      
+      - ElasticNet
+      - LGBMRegressor
+      - GRU
+      - LSTM
+      - TCN
+      - Transformer
   seed:
     values:
       - 1111
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index 535d9619..c19076e5 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -23,7 +23,7 @@ def execute_repeated_cv(
     train_size: int = None,
     load_weights: bool = False,
     source_dir: Path = None,
-    cv_repetitions: int = 1,
+    cv_repetitions: int = 5,
     cv_repetitions_to_train: int = None,
     cv_folds: int = 5,
     cv_folds_to_train: int = None,
diff --git a/icu_benchmarks/data/Checking data-Copy1.ipynb b/icu_benchmarks/data/Checking data-Copy1.ipynb
deleted file mode 100644
index 9abe4d48..00000000
--- a/icu_benchmarks/data/Checking data-Copy1.ipynb	
+++ /dev/null
@@ -1,582 +0,0 @@
-{
- "cells": [
-  {
-   "cell_type": "code",
-   "execution_count": 4,
-   "id": "6b471a18",
-   "metadata": {},
-   "outputs": [
-    {
-     "ename": "ModuleNotFoundError",
-     "evalue": "No module named 'icu_benchmarks'",
-     "output_type": "error",
-     "traceback": [
-      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
-      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
-      "Input \u001b[0;32mIn [4]\u001b[0m, in \u001b[0;36m<cell line: 1>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01micu_benchmarks\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mdata\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01msplit_process_data\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m preprocess_data\n",
-      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'icu_benchmarks'"
-     ]
-    }
-   ],
-   "source": [
-    "from icu_benchmarks.data.split_process_data import preprocess_data"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 6,
-   "id": "9786d96d",
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "0\n"
-     ]
-    }
-   ],
-   "source": [
-    "print(1-1)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 7,
-   "id": "b9f28515",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "import pandas as pd\n",
-    "dyn=pd.read_parquet('../../demo_data/los/eicu_demo/dyn.parquet', engine='pyarrow')\n",
-    "outc=pd.read_parquet('../../demo_data/los/eicu_demo/outc.parquet', engine='pyarrow')\n",
-    "sta=pd.read_parquet('../../demo_data/los/eicu_demo/sta.parquet', engine='pyarrow')"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "35342dde",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "dyn.columns"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "c6f3e4ae",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "dyn.columns"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "64c217e9",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "outc.head()"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "87724d04",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "sta.head()"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "fddac54c",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "sta.columns"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 3,
-   "id": "265af37f",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "from typing import List\n",
-    "from pandas import DataFrame\n",
-    "import numpy as np\n",
-    "import gin\n",
-    "from torch import Tensor, cat, from_numpy, float32,empty,stack\n",
-    "from torch.utils.data import Dataset\n",
-    "import logging\n",
-    "from typing import Dict, Tuple\n",
-    "from constants import DataSegment as Segment\n",
-    "from constants import DataSplit as Split\n",
-    "from collections import OrderedDict\n",
-    "\n",
-    "FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target', 'id']\n",
-    "def ampute_data(data, mechanism, p_miss, p_obs=0.3):\n",
-    "    \"\"\"\n",
-    "    Generate missing values for specifics missing-data mechanism and proportion of missing values.\n",
-    "\n",
-    "    Parameters\n",
-    "    ----------\n",
-    "    data : DataFrame\n",
-    "        Data for which missing values will be simulated.\n",
-    "    mechanism : str,\n",
-    "            Indicates the missing-data mechanism to be used. (\"MCAR\", \"MAR\" or \"MNAR\")\n",
-    "    p_miss : float\n",
-    "        Proportion of missing values to generate for variables which will have missing values.\n",
-    "    p_obs : float\n",
-    "            If mecha = \"MAR\" or \"MNAR\", proportion of variables with *no* missing values\n",
-    "            that will be used for the logistic masking model.\n",
-    "\n",
-    "    Returns\n",
-    "    ----------\n",
-    "    imputed_data: DataFrame\n",
-    "        The data with the generated missing values.\n",
-    "    \"\"\"\n",
-    "    logging.info(f\"Applying {mechanism} amputation.\")\n",
-    "    X = torch.tensor(data.values.astype(np.float32))\n",
-    "\n",
-    "    if mechanism == \"MAR\":\n",
-    "        mask = MAR_logistic_mask(X, p_miss, p_obs)\n",
-    "    elif mechanism == \"MNAR\":\n",
-    "        mask = MNAR_logistic_mask(X, p_miss, p_obs)\n",
-    "    elif mechanism == \"MCAR\":\n",
-    "        mask = MCAR_mask(X, p_miss)\n",
-    "    elif mechanism == \"BO\":\n",
-    "        mask = BO_mask(X, p_miss)\n",
-    "    else:\n",
-    "        logging.error(\"Not a valid amputation mechanism. Missing-data mechanisms to be used are MCAR, MAR or MNAR.\")\n",
-    "\n",
-    "    amputed_data = data.mask(mask)\n",
-    "    return amputed_data, mask\n",
-    "class CommonDataset(Dataset):\n",
-    "    \"\"\"Common dataset: subclass of Torch Dataset that represents the data to learn on.\n",
-    "\n",
-    "    Args: data: Dict of the different splits of the data. split: Either 'train','val' or 'test'. vars: Contains the names of\n",
-    "    columns in the data. grouping_segment: str, optional: The segment of the data contains the grouping column with only\n",
-    "    unique values. Defaults to Segment.outcome. Is used to calculate the number of stays in the data.\n",
-    "    \"\"\"\n",
-    "\n",
-    "    def __init__(\n",
-    "        self,\n",
-    "        data: dict,\n",
-    "        split: str = Split.train,\n",
-    "        vars: Dict[str, str] = gin.REQUIRED,\n",
-    "        grouping_segment: str = Segment.outcome,\n",
-    "    ):\n",
-    "        self.split = split\n",
-    "        self.vars = vars\n",
-    "        self.grouping_df = data[split][grouping_segment].set_index(self.vars[\"GROUP\"])\n",
-    "        self.features_df = (\n",
-    "            #drops time coulmn and sets index to stay_id\n",
-    "            data[split][Segment.features].set_index(self.vars[\"GROUP\"]).drop(labels=self.vars[\"SEQUENCE\"], axis=1)\n",
-    "        )\n",
-    "\n",
-    "        # calculate basic info for the data\n",
-    "        self.num_stays = self.grouping_df.index.unique().shape[0]\n",
-    "        self.maxlen = self.features_df.groupby([self.vars[\"GROUP\"]]).size().max()\n",
-    "\n",
-    "    def ram_cache(self, cache: bool = True):\n",
-    "        self._cached_dataset = None\n",
-    "        if cache:\n",
-    "            logging.info(\"Caching dataset in ram.\")\n",
-    "            self._cached_dataset = [self[i] for i in range(len(self))]\n",
-    "\n",
-    "    def __len__(self) -> int:\n",
-    "        \"\"\"Returns number of stays in the data.\n",
-    "\n",
-    "        Returns:\n",
-    "            number of stays in the data\n",
-    "        \"\"\"\n",
-    "        return self.num_stays\n",
-    "\n",
-    "    def get_feature_names(self):\n",
-    "        return self.features_df.columns\n",
-    "\n",
-    "    def to_tensor(self):\n",
-    "        values = []\n",
-    "        for entry in self:\n",
-    "            for i, value in enumerate(entry):\n",
-    "                if len(values) <= i:\n",
-    "                    values.append([])\n",
-    "                values[i].append(value.unsqueeze(0))\n",
-    "        return [cat(value, dim=0) for value in values]\n",
-    "\n",
-    "\n",
-    "@gin.configurable(\"PredictionDataset\")\n",
-    "class PredictionDataset(CommonDataset):\n",
-    "    \"\"\"Subclass of common dataset for prediction tasks.\n",
-    "\n",
-    "    Args:\n",
-    "        ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.\n",
-    "    \"\"\"\n",
-    "\n",
-    "    def __init__(self, *args, ram_cache: bool = True, **kwargs):\n",
-    "        super().__init__(*args, grouping_segment=Segment.outcome, **kwargs)\n",
-    "        self.outcome_df = self.grouping_df\n",
-    "        self.ram_cache(ram_cache)\n",
-    "\n",
-    "    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:\n",
-    "        \"\"\"Function to sample from the data split of choice. Used for deep learning implementations.\n",
-    "\n",
-    "        Args:\n",
-    "            idx: A specific row index to sample.\n",
-    "\n",
-    "        Returns:\n",
-    "            A sample from the data, consisting of data, labels and padding mask.\n",
-    "        \"\"\"\n",
-    "        if self._cached_dataset is not None:\n",
-    "            return self._cached_dataset[idx]\n",
-    "\n",
-    "        pad_value = 0.0\n",
-    "        stay_id = self.outcome_df.index.unique()[idx]  # [self.vars[\"GROUP\"]]\n",
-    "\n",
-    "        # slice to make sure to always return a DF\n",
-    "        window = self.features_df.loc[stay_id:stay_id].to_numpy()\n",
-    "        labels = self.outcome_df.loc[stay_id:stay_id][self.vars[\"LABEL\"]].to_numpy(dtype=float)\n",
-    "\n",
-    "        if len(labels) == 1:\n",
-    "            # only one label per stay, align with window\n",
-    "            labels = np.concatenate([np.empty(window.shape[0] - 1) * np.nan, labels], axis=0)\n",
-    "\n",
-    "        length_diff = self.maxlen - window.shape[0]\n",
-    "        \n",
-    "        \n",
-    "        pad_mask = np.ones(window.shape[0])\n",
-    "\n",
-    "        # Padding the array to fulfill size requirement\n",
-    "        if length_diff > 0:\n",
-    "            # window shorter than the longest window in dataset, pad to same length\n",
-    "            window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)\n",
-    "            labels = np.concatenate([labels, np.ones(length_diff) * pad_value], axis=0)\n",
-    "            pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)\n",
-    "\n",
-    "        not_labeled = np.argwhere(np.isnan(labels))\n",
-    "        if len(not_labeled) > 0:\n",
-    "            labels[not_labeled] = -1\n",
-    "            pad_mask[not_labeled] = 0\n",
-    "\n",
-    "        pad_mask = pad_mask.astype(bool)\n",
-    "        labels = labels.astype(np.float32)\n",
-    "        data = window.astype(np.float32)\n",
-    "\n",
-    "        return from_numpy(data), from_numpy(labels), from_numpy(pad_mask)\n",
-    "\n",
-    "    def get_balance(self) -> list:\n",
-    "        \"\"\"Return the weight balance for the split of interest.\n",
-    "\n",
-    "        Returns:\n",
-    "            Weights for each label.\n",
-    "        \"\"\"\n",
-    "        counts = self.outcome_df[self.vars[\"LABEL\"]].value_counts()\n",
-    "        return list((1 / counts) * np.sum(counts) / counts.shape[0])\n",
-    "\n",
-    "    def get_data_and_labels(self) -> Tuple[np.array, np.array]:\n",
-    "        \"\"\"Function to return all the data and labels aligned at once.\n",
-    "\n",
-    "        We use this function for the ML methods which don't require an iterator.\n",
-    "\n",
-    "        Returns:\n",
-    "            A Tuple containing data points and label for the split.\n",
-    "        \"\"\"\n",
-    "        labels = self.outcome_df[self.vars[\"LABEL\"]].to_numpy().astype(float)\n",
-    "        rep = self.features_df\n",
-    "        if len(labels) == self.num_stays:\n",
-    "            # order of groups could be random, we make sure not to change it\n",
-    "            rep = rep.groupby(level=self.vars[\"GROUP\"], sort=False).last()\n",
-    "        rep = rep.to_numpy().astype(float)\n",
-    "        \n",
-    "        return rep, labels\n",
-    "\n",
-    "    def to_tensor(self):\n",
-    "        data, labels = self.get_data_and_labels()\n",
-    "        return from_numpy(data).to(float32), from_numpy(labels).to(float32)\n",
-    "\n",
-    "@gin.configurable(\"PredictionDatasetTFT\")\n",
-    "class PredictionDatasetTFT(PredictionDataset):\n",
-    "    \"\"\"Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.\n",
-    "\n",
-    "    Args:\n",
-    "        ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.\n",
-    "    \"\"\"\n",
-    "\n",
-    "    def __init__(self, *args, ram_cache: bool = True, **kwargs):\n",
-    "        super().__init__(*args,ram_cache = True, **kwargs)\n",
-    "        \n",
-    "\n",
-    "    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:\n",
-    "        \"\"\"Function to sample from the data split of choice. Used for TFT.\n",
-    "\n",
-    "        Args:\n",
-    "            idx: A specific row index to sample.\n",
-    "\n",
-    "        Returns:\n",
-    "            A sample from the data, consisting of data, labels and padding mask.\n",
-    "        \"\"\"\n",
-    "        if self._cached_dataset is not None:\n",
-    "            return self._cached_dataset[idx]\n",
-    "\n",
-    "        pad_value = 0.0\n",
-    "        stay_id = self.outcome_df.index.unique()[idx]  # [self.vars[\"GROUP\"]]\n",
-    "\n",
-    "        \n",
-    "        # We need to be sure that tensors are returned in the correct order to be processed correclty by tft\n",
-    "        tensors = [[] for _ in range(8)]\n",
-    "        for var in self.features_df.columns:\n",
-    "            if  var == 'sex' :\n",
-    "                tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())\n",
-    "            elif var == 'age' or var== 'height' or var== 'weight':\n",
-    "                tensors[1].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())\n",
-    "            else :\n",
-    "                tensors[5].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())\n",
-    "    \n",
-    "        \n",
-    "        tensors[6].extend(self.outcome_df.loc[stay_id:stay_id][self.vars[\"LABEL\"]].to_numpy(dtype=float))\n",
-    "        tensors[7].append(np.asarray([stay_id]))\n",
-    "        \n",
-    "        window_shape0=np.shape(tensors[0])[1]\n",
-    "       # window = self.features_df.loc[stay_id:stay_id].to_numpy()\n",
-    "       # labels = self.outcome_df.loc[stay_id:stay_id][self.vars[\"LABEL\"]].to_numpy(dtype=float)\n",
-    "        if len(tensors[6]) == 1:\n",
-    "            # only one label per stay, align with window\n",
-    "            tensors[6] = np.concatenate([np.empty(window_shape0 - 1) * np.nan, tensors[6]], axis=0)\n",
-    "\n",
-    "        length_diff = self.maxlen - window_shape0\n",
-    "        pad_mask = np.ones(window_shape0)\n",
-    "\n",
-    "        # Padding the array to fulfill size requirement\n",
-    "        if length_diff > 0:\n",
-    "            # window shorter than the longest window in dataset, pad to same length\n",
-    "            tensors[0]= np.concatenate([tensors[0], np.ones((length_diff, np.shape(tensors[0])[0])) * pad_value], axis=0)\n",
-    "            tensors[1]= np.concatenate([tensors[1], np.ones((length_diff, np.shape(tensors[1])[0])) * pad_value], axis=0)\n",
-    "          #  tensors[2]= np.concatenate([tensors[2], np.ones((length_diff, np.shape(tensors[2])[0])) * pad_value], axis=0)\n",
-    "          #  tensors[3]= np.concatenate([tensors[3], np.ones((length_diff, np.shape(tensors[3])[0])) * pad_value], axis=0)\n",
-    "          #  tensors[4]= np.concatenate([tensors[4], np.ones((length_diff, np.shape(tensors[4])[0])) * pad_value], axis=0)\n",
-    "            tensors[5]= np.concatenate([tensors[5], np.ones((length_diff, np.shape(tensors[5])[0])) * pad_value], axis=0)\n",
-    "            tensors[7]= np.concatenate([tensors[7], np.ones((length_diff, np.shape(tensors[7])[0])) * stay_id], axis=0)\n",
-    "           # window = np.concatenate([window, np.ones((length_diff, window.shape[1])) * pad_value], axis=0)\n",
-    "            \n",
-    "            tensors[6] = np.concatenate([tensors[6], np.ones(length_diff) * pad_value], axis=0)\n",
-    "            pad_mask = np.concatenate([pad_mask, np.zeros(length_diff)], axis=0)\n",
-    "\n",
-    "        not_labeled = np.argwhere(np.isnan(tensors[6]))\n",
-    "        if len(not_labeled) > 0:\n",
-    "            tensors[6][not_labeled] = -1\n",
-    "            pad_mask[not_labeled] = 0\n",
-    "        tensors[6]=[tensors[6]]\n",
-    "        pad_mask = pad_mask.astype(bool)\n",
-    "       # data = window.astype(np.float32)\n",
-    "        tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)\n",
-    "        tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]\n",
-    "        \n",
-    "        return  OrderedDict(zip(FEAT_NAMES, tensors)),from_numpy(pad_mask)\n",
-    "\n",
-    "    \n",
-    "\n",
-    "\n",
-    "\n",
-    "@gin.configurable(\"ImputationDataset\")\n",
-    "class ImputationDataset(CommonDataset):\n",
-    "    \"\"\"Subclass of Common Dataset that contains data for imputation models.\"\"\"\n",
-    "\n",
-    "    def __init__(\n",
-    "        self,\n",
-    "        data: Dict[str, DataFrame],\n",
-    "        split: str = Split.train,\n",
-    "        vars: Dict[str, str] = gin.REQUIRED,\n",
-    "        mask_proportion=0.3,\n",
-    "        mask_method=\"MCAR\",\n",
-    "        mask_observation_proportion=0.3,\n",
-    "        ram_cache: bool = True,\n",
-    "    ):\n",
-    "        \"\"\"\n",
-    "        Args:\n",
-    "            data (Dict[str, DataFrame]): data to use\n",
-    "            split (str, optional): split to apply. Defaults to Split.train.\n",
-    "            vars (Dict[str, str], optional): contains names of columns in the data. Defaults to gin.REQUIRED.\n",
-    "            mask_proportion (float, optional): proportion to artificially mask for amputation. Defaults to 0.3.\n",
-    "            mask_method (str, optional): masking mechanism. Defaults to \"MCAR\".\n",
-    "            mask_observation_proportion (float, optional): poportion of the observed data to be masked. Defaults to 0.3.\n",
-    "            ram_cache (bool, optional): if the dataset should be completely stored in ram and not generated on the fly during\n",
-    "                training. Defaults to True.\n",
-    "        \"\"\"\n",
-    "        super().__init__(data, split, vars, grouping_segment=Segment.static)\n",
-    "        self.amputated_values, self.amputation_mask = ampute_data(\n",
-    "            self.features_df, mask_method, mask_proportion, mask_observation_proportion\n",
-    "        )\n",
-    "        self.amputation_mask = (self.amputation_mask + self.features_df.isna().values).bool()\n",
-    "        self.amputation_mask = DataFrame(self.amputation_mask, columns=self.vars[Segment.dynamic])\n",
-    "        self.amputation_mask[self.vars[\"GROUP\"]] = self.features_df.index\n",
-    "        self.amputation_mask.set_index(self.vars[\"GROUP\"], inplace=True)\n",
-    "\n",
-    "        self.target_missingness_mask = self.features_df.isna()\n",
-    "        self.features_df.fillna(0, inplace=True)\n",
-    "        self.ram_cache(ram_cache)\n",
-    "\n",
-    "    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:\n",
-    "        \"\"\"Function to sample from the data split of choice.\n",
-    "\n",
-    "        Used for deep learning implementations.\n",
-    "\n",
-    "        Args:\n",
-    "            idx: A specific row index to sample.\n",
-    "\n",
-    "        Returns:\n",
-    "            A sample from the data, consisting of data, labels and padding mask.\n",
-    "        \"\"\"\n",
-    "        if self._cached_dataset is not None:\n",
-    "            return self._cached_dataset[idx]\n",
-    "        stay_id = self.grouping_df.iloc[idx].name\n",
-    "\n",
-    "        # slice to make sure to always return a DF\n",
-    "        window = self.features_df.loc[stay_id:stay_id, self.vars[Segment.dynamic]]\n",
-    "        window_missingness_mask = self.target_missingness_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]\n",
-    "        amputated_window = self.amputated_values.loc[stay_id:stay_id, self.vars[Segment.dynamic]]\n",
-    "        amputation_mask = self.amputation_mask.loc[stay_id:stay_id, self.vars[Segment.dynamic]]\n",
-    "\n",
-    "        return (\n",
-    "            from_numpy(amputated_window.values).to(float32),\n",
-    "            from_numpy(amputation_mask.values).to(float32),\n",
-    "            from_numpy(window.values).to(float32),\n",
-    "            from_numpy(window_missingness_mask.values).to(float32),\n",
-    "        )\n",
-    "\n",
-    "\n",
-    "@gin.configurable(\"ImputationPredictionDataset\")\n",
-    "class ImputationPredictionDataset(Dataset):\n",
-    "    \"\"\"Subclass of torch dataset that represents data with missingness for imputation.\n",
-    "\n",
-    "    Args:\n",
-    "        data (DataFrame): dict of the different splits of the data\n",
-    "        grouping_column (str, optional): column that is used for grouping. Defaults to \"stay_id\".\n",
-    "        select_columns (List[str], optional): the columns to serve as input for the imputation model. Defaults to None.\n",
-    "        ram_cache (bool, optional): wether the dataset should be stored in ram. Defaults to True.\n",
-    "    \"\"\"\n",
-    "\n",
-    "    def __init__(\n",
-    "        self,\n",
-    "        data: DataFrame,\n",
-    "        grouping_column: str = \"stay_id\",\n",
-    "        select_columns: List[str] = None,\n",
-    "        ram_cache: bool = True,\n",
-    "    ):\n",
-    "        self.dyn_df = data\n",
-    "\n",
-    "        if select_columns is not None:\n",
-    "            self.dyn_df = self.dyn_df[list(select_columns) + grouping_column]\n",
-    "\n",
-    "        if grouping_column is not None:\n",
-    "            self.dyn_df = self.dyn_df.set_index(grouping_column)\n",
-    "        else:\n",
-    "            self.dyn_df = data\n",
-    "\n",
-    "        # calculate basic info for the data\n",
-    "        self.group_indices = self.dyn_df.index.unique()\n",
-    "        self.maxlen = self.dyn_df.groupby(grouping_column).size().max()\n",
-    "\n",
-    "        self._cached_dataset = None\n",
-    "        if ram_cache:\n",
-    "            logging.info(\"Caching dataset in ram.\")\n",
-    "            self._cached_dataset = [self[i] for i in range(len(self))]\n",
-    "\n",
-    "    def __len__(self) -> int:\n",
-    "        \"\"\"Returns number of stays in the data.\n",
-    "\n",
-    "        Returns:\n",
-    "            number of stays in the data\n",
-    "        \"\"\"\n",
-    "        return self.group_indices.shape[0]\n",
-    "\n",
-    "    def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:\n",
-    "        \"\"\"Function to sample from the data split of choice.\n",
-    "\n",
-    "        Used for deep learning implementations.\n",
-    "\n",
-    "        Args:\n",
-    "            idx: A specific row index to sample.\n",
-    "\n",
-    "        Returns:\n",
-    "            A sample from the data, consisting of data, labels and padding mask.\n",
-    "        \"\"\"\n",
-    "        if self._cached_dataset is not None:\n",
-    "            return self._cached_dataset[idx]\n",
-    "        stay_id = self.group_indices[idx]\n",
-    "\n",
-    "        # slice to make sure to always return a DF\n",
-    "        window = self.dyn_df.loc[stay_id:stay_id, :]\n",
-    "\n",
-    "        return from_numpy(window.values).to(float32)\n",
-    "\n"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 9,
-   "id": "504fd6c4",
-   "metadata": {},
-   "outputs": [
-    {
-     "ename": "KeyError",
-     "evalue": "'train'\n  In call to configurable 'PredictionDataset' (<class '__main__.PredictionDataset'>)",
-     "output_type": "error",
-     "traceback": [
-      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
-      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
-      "Input \u001b[0;32mIn [9]\u001b[0m, in \u001b[0;36m<cell line: 1>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0m train_dataset \u001b[38;5;241m=\u001b[39m \u001b[43mPredictionDataset\u001b[49m\u001b[43m(\u001b[49m\u001b[43mdyn\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43msplit\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mSplit\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mtrain\u001b[49m\u001b[43m)\u001b[49m\n",
-      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/gin/config.py:1605\u001b[0m, in \u001b[0;36m_make_gin_wrapper.<locals>.gin_wrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m   1603\u001b[0m scope_info \u001b[38;5;241m=\u001b[39m \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m in scope \u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;132;01m{}\u001b[39;00m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;241m.\u001b[39mformat(scope_str) \u001b[38;5;28;01mif\u001b[39;00m scope_str \u001b[38;5;28;01melse\u001b[39;00m \u001b[38;5;124m'\u001b[39m\u001b[38;5;124m'\u001b[39m\n\u001b[1;32m   1604\u001b[0m err_str \u001b[38;5;241m=\u001b[39m err_str\u001b[38;5;241m.\u001b[39mformat(name, fn_or_cls, scope_info)\n\u001b[0;32m-> 1605\u001b[0m \u001b[43mutils\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43maugment_exception_message_and_reraise\u001b[49m\u001b[43m(\u001b[49m\u001b[43me\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43merr_str\u001b[49m\u001b[43m)\u001b[49m\n",
-      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/gin/utils.py:41\u001b[0m, in \u001b[0;36maugment_exception_message_and_reraise\u001b[0;34m(exception, message)\u001b[0m\n\u001b[1;32m     39\u001b[0m proxy \u001b[38;5;241m=\u001b[39m ExceptionProxy()\n\u001b[1;32m     40\u001b[0m ExceptionProxy\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__qualname__\u001b[39m \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mtype\u001b[39m(exception)\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__qualname__\u001b[39m\n\u001b[0;32m---> 41\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m proxy\u001b[38;5;241m.\u001b[39mwith_traceback(exception\u001b[38;5;241m.\u001b[39m__traceback__) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;28mNone\u001b[39m\n",
-      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/gin/config.py:1582\u001b[0m, in \u001b[0;36m_make_gin_wrapper.<locals>.gin_wrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m   1579\u001b[0m new_kwargs\u001b[38;5;241m.\u001b[39mupdate(kwargs)\n\u001b[1;32m   1581\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[0;32m-> 1582\u001b[0m   \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfn\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mnew_args\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mnew_kwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1583\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mException\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m e:  \u001b[38;5;66;03m# pylint: disable=broad-except\u001b[39;00m\n\u001b[1;32m   1584\u001b[0m   err_str \u001b[38;5;241m=\u001b[39m \u001b[38;5;124m'\u001b[39m\u001b[38;5;124m'\u001b[39m\n",
-      "Input \u001b[0;32mIn [3]\u001b[0m, in \u001b[0;36mPredictionDataset.__init__\u001b[0;34m(self, ram_cache, *args, **kwargs)\u001b[0m\n\u001b[1;32m    113\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m__init__\u001b[39m(\u001b[38;5;28mself\u001b[39m, \u001b[38;5;241m*\u001b[39margs, ram_cache: \u001b[38;5;28mbool\u001b[39m \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mTrue\u001b[39;00m, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs):\n\u001b[0;32m--> 114\u001b[0m     \u001b[38;5;28;43msuper\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[38;5;21;43m__init__\u001b[39;49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mgrouping_segment\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mSegment\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43moutcome\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    115\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39moutcome_df \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mgrouping_df\n\u001b[1;32m    116\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mram_cache(ram_cache)\n",
-      "Input \u001b[0;32mIn [3]\u001b[0m, in \u001b[0;36mCommonDataset.__init__\u001b[0;34m(self, data, split, vars, grouping_segment)\u001b[0m\n\u001b[1;32m     66\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39msplit \u001b[38;5;241m=\u001b[39m split\n\u001b[1;32m     67\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mvars \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mvars\u001b[39m\n\u001b[0;32m---> 68\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mgrouping_df \u001b[38;5;241m=\u001b[39m \u001b[43mdata\u001b[49m\u001b[43m[\u001b[49m\u001b[43msplit\u001b[49m\u001b[43m]\u001b[49m[grouping_segment]\u001b[38;5;241m.\u001b[39mset_index(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mvars[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mGROUP\u001b[39m\u001b[38;5;124m\"\u001b[39m])\n\u001b[1;32m     69\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mfeatures_df \u001b[38;5;241m=\u001b[39m (\n\u001b[1;32m     70\u001b[0m     \u001b[38;5;66;03m#drops time coulmn and sets index to stay_id\u001b[39;00m\n\u001b[1;32m     71\u001b[0m     data[split][Segment\u001b[38;5;241m.\u001b[39mfeatures]\u001b[38;5;241m.\u001b[39mset_index(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mvars[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mGROUP\u001b[39m\u001b[38;5;124m\"\u001b[39m])\u001b[38;5;241m.\u001b[39mdrop(labels\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mvars[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mSEQUENCE\u001b[39m\u001b[38;5;124m\"\u001b[39m], axis\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m1\u001b[39m)\n\u001b[1;32m     72\u001b[0m )\n\u001b[1;32m     74\u001b[0m \u001b[38;5;66;03m# calculate basic info for the data\u001b[39;00m\n",
-      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/frame.py:3505\u001b[0m, in \u001b[0;36mDataFrame.__getitem__\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   3503\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mcolumns\u001b[38;5;241m.\u001b[39mnlevels \u001b[38;5;241m>\u001b[39m \u001b[38;5;241m1\u001b[39m:\n\u001b[1;32m   3504\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_getitem_multilevel(key)\n\u001b[0;32m-> 3505\u001b[0m indexer \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mcolumns\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mget_loc\u001b[49m\u001b[43m(\u001b[49m\u001b[43mkey\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   3506\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m is_integer(indexer):\n\u001b[1;32m   3507\u001b[0m     indexer \u001b[38;5;241m=\u001b[39m [indexer]\n",
-      "File \u001b[0;32m~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/indexes/base.py:3631\u001b[0m, in \u001b[0;36mIndex.get_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   3629\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_engine\u001b[38;5;241m.\u001b[39mget_loc(casted_key)\n\u001b[1;32m   3630\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mKeyError\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m err:\n\u001b[0;32m-> 3631\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mKeyError\u001b[39;00m(key) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01merr\u001b[39;00m\n\u001b[1;32m   3632\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mTypeError\u001b[39;00m:\n\u001b[1;32m   3633\u001b[0m     \u001b[38;5;66;03m# If we have a listlike key, _check_indexing_error will raise\u001b[39;00m\n\u001b[1;32m   3634\u001b[0m     \u001b[38;5;66;03m#  InvalidIndexError. Otherwise we fall through and re-raise\u001b[39;00m\n\u001b[1;32m   3635\u001b[0m     \u001b[38;5;66;03m#  the TypeError.\u001b[39;00m\n\u001b[1;32m   3636\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_check_indexing_error(key)\n",
-      "\u001b[0;31mKeyError\u001b[0m: 'train'\n  In call to configurable 'PredictionDataset' (<class '__main__.PredictionDataset'>)"
-     ]
-    }
-   ],
-   "source": [
-    "train_dataset = PredictionDataset(dyn, split=Split.train)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "c7f48f70",
-   "metadata": {},
-   "outputs": [],
-   "source": []
-  }
- ],
- "metadata": {
-  "kernelspec": {
-   "display_name": "yaib",
-   "language": "python",
-   "name": "yaib"
-  },
-  "language_info": {
-   "codemirror_mode": {
-    "name": "ipython",
-    "version": 3
-   },
-   "file_extension": ".py",
-   "mimetype": "text/x-python",
-   "name": "python",
-   "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython3",
-   "version": "3.9.13"
-  }
- },
- "nbformat": 4,
- "nbformat_minor": 5
-}
diff --git a/icu_benchmarks/data/constants.py b/icu_benchmarks/data/constants.py
index 4e4a9c34..031f35ca 100644
--- a/icu_benchmarks/data/constants.py
+++ b/icu_benchmarks/data/constants.py
@@ -15,7 +15,3 @@ class VarType:
     group = "GROUP"
     sequence = "SEQUENCE"
     label = "LABEL"
-
-
-class FeatType:
-    FEAT_NAMES = ["s_cat", "s_cont", "k_cat", "k_cont", "o_cat", "o_cont", "target", "id"]
diff --git a/icu_benchmarks/imputation/diffwave.py b/icu_benchmarks/imputation/diffwave.py
index 437303ed..458c2c5e 100644
--- a/icu_benchmarks/imputation/diffwave.py
+++ b/icu_benchmarks/imputation/diffwave.py
@@ -330,7 +330,7 @@ def forward(self, input_data):
         cond = self.cond_conv(cond)
         h += cond
 
-        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels:, :])
+        out = torch.tanh(h[:, :self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels:, :])
 
         res = self.res_conv(out)
         assert x.shape == res.shape
diff --git a/icu_benchmarks/models/constants.py b/icu_benchmarks/models/constants.py
index 224355eb..cd03bcd0 100644
--- a/icu_benchmarks/models/constants.py
+++ b/icu_benchmarks/models/constants.py
@@ -19,6 +19,8 @@
 )
 from torchmetrics.classification import (
     AUROC,
+    AveragePrecision as TorchMetricsAveragePrecision,
+    PrecisionRecallCurve as TorchMetricsPrecisionRecallCurve,
     CalibrationError,
     F1Score,
 )
@@ -28,6 +30,7 @@
     BalancedAccuracy,
     MAE,
     JSD,
+    BinaryFairnessWrapper,
 )
 
 
@@ -66,11 +69,11 @@ class DLMetrics:
 
     BINARY_CLASSIFICATION_TORCHMETRICS = {
         "AUC": AUROC(task="binary"),
-        # "PR": TorchMetricsAveragePrecision(task="binary"),
-        # "PrecisionRecallCurve": TorchMetricsPrecisionRecallCurve(task="binary"),
+        "PR": TorchMetricsAveragePrecision(task="binary"),
+        "PrecisionRecallCurve": TorchMetricsPrecisionRecallCurve(task="binary"),
         "Calibration_Error": CalibrationError(task="binary", n_bins=10),
         "F1": F1Score(task="binary", num_classes=2),
-        # "Binary_Fairness": BinaryFairnessWrapper(num_groups=2, task="demographic_parity", group_name="sex"),
+        "Binary_Fairness": BinaryFairnessWrapper(num_groups=2, task="demographic_parity", group_name="sex"),
     }
 
     MULTICLASS_CLASSIFICATION = {
diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index d1526e6d..fd88699f 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -6,7 +6,7 @@
 from sklearn.calibration import calibration_curve
 from scipy.spatial.distance import jensenshannon
 from torchmetrics.classification import BinaryFairness
-from icu_benchmarks.models.similarity_func import correlation_spearman, cosine
+from quantus.functions.similarity_func import correlation_spearman, cosine
 
 """"
 This file contains custom metrics that can be added to YAIB.
diff --git a/icu_benchmarks/models/similarity_func.py b/icu_benchmarks/models/similarity_func.py
deleted file mode 100644
index bc49850f..00000000
--- a/icu_benchmarks/models/similarity_func.py
+++ /dev/null
@@ -1,288 +0,0 @@
-"""This module holds a collection of similarity functions i.e.,
-ways to measure the distance between two inputs (or explanations)."""
-
-# This file is part of Quantus.
-# Quantus is free software: you can redistribute it and/or modify it under
-# the terms of the GNU Lesser General Public License as published by the Free Software Foundation,
-# either version 3 of the License, or (at your option) any later version.
-# Quantus is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
-# or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
-# You should have received a copy of the GNU Lesser General Public License along with Quantus.
-# If not, see <https://www.gnu.org/licenses/>.
-# Quantus project URL: <https://github.com/understandable-machine-intelligence-lab/Quantus>.
-# Quantus project URL: https://github.com/understandable-machine-intelligence-lab/Quantus
-
-from typing import Union
-
-import numpy as np
-import scipy
-import skimage
-
-
-def correlation_spearman(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate Spearman rank of two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return scipy.stats.spearmanr(a, b)[0]
-
-
-def correlation_pearson(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate Pearson correlation of two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return scipy.stats.pearsonr(a, b)[0]
-
-
-def correlation_kendall_tau(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate Kendall Tau correlation of two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return scipy.stats.kendalltau(a, b)[0]
-
-
-def distance_euclidean(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate Euclidean distance of two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return scipy.spatial.distance.euclidean(u=a, v=b)
-
-
-def distance_manhattan(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate Manhattan distance of two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return scipy.spatial.distance.cityblock(u=a, v=b)
-
-
-def distance_chebyshev(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate Chebyshev distance of two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return scipy.spatial.distance.chebyshev(u=a, v=b)
-
-
-def lipschitz_constant(a: np.array, b: np.array, c: Union[np.array, None], d: Union[np.array, None], **kwargs) -> float:
-    """
-    Calculate non-negative local Lipschitz abs(||a-b||/||c-d||), where a,b can be f(x) or a(x) and c,d is x.
-
-    For numerical stability, a small value is added to division.
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    c: np.ndarray
-         The third array to use for similarity scoring.
-    d: np.ndarray
-         The fourth array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    eps = 1e-10
-
-    d1 = kwargs.get("norm_numerator", distance_manhattan)
-    d2 = kwargs.get("norm_denominator", distance_euclidean)
-
-    if np.shape(a) == ():
-        return float(abs(a - b) / (d2(c, d) + eps))
-    else:
-        return float(d1(a, b) / (d2(a=c, b=d) + eps))
-
-
-def abs_difference(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate the mean of the absolute differences between two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return np.mean(abs(a - b))
-
-
-def squared_difference(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate the sqaured differences between two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return np.sum((a - b) ** 2)
-
-
-def cosine(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate Cosine of two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    return scipy.spatial.distance.cosine(u=a, v=b)
-
-
-def ssim(a: np.array, b: np.array, **kwargs) -> float:
-    """
-    Calculate Structural Similarity Index Measure of two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    float
-        The similarity score.
-    """
-    max_point, min_point = np.max(np.concatenate([a, b])), np.min(np.concatenate([a, b]))
-    data_range = float(np.abs(max_point - min_point))
-    return skimage.metrics.structural_similarity(im1=a, im2=b, win_size=kwargs.get("win_size", None), data_range=data_range)
-
-
-def difference(a: np.array, b: np.array, **kwargs) -> np.array:
-    """
-    Calculate the difference between two images (or explanations).
-
-    Parameters
-    ----------
-    a: np.ndarray
-         The first array to use for similarity scoring.
-    b: np.ndarray
-         The second array to use for similarity scoring.
-    kwargs: optional
-        Keyword arguments.
-
-    Returns
-    -------
-    np.array
-        The difference in each element.
-    """
-    return a - b
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 5792a5bb..79f13ef4 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -19,7 +19,7 @@
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
 import matplotlib.pyplot as plt
-from icu_benchmarks.models.similarity_func import correlation_spearman, cosine
+from quantus.functions.similarity_func import correlation_spearman, cosine
 import captum
 from captum._utils.models.linear_model import SkLearnLasso
 
diff --git a/icu_benchmarks/run_utils.py b/icu_benchmarks/run_utils.py
index 7d57bc2a..dac9730d 100644
--- a/icu_benchmarks/run_utils.py
+++ b/icu_benchmarks/run_utils.py
@@ -16,7 +16,6 @@
 from icu_benchmarks.models.utils import JsonResultLoggingEncoder
 from icu_benchmarks.wandb_utils import wandb_log
 import numpy as np
-import matplotlib.pyplot as plt
 
 
 def build_parser() -> ArgumentParser:
@@ -27,132 +26,39 @@ def build_parser() -> ArgumentParser:
     """
     parser = ArgumentParser(description="Framework for benchmarking ML/DL models on ICU data")
 
-    parser.add_argument(
-        "-d",
-        "--data-dir",
-        required=True,
-        type=Path,
-        help="Path to the parquet data directory.",
-    )
-    parser.add_argument(
-        "-t",
-        "--task",
-        default="BinaryClassification",
-        required=True,
-        help="Name of the task gin.",
-    )
+    parser.add_argument("-d", "--data-dir", required=True, type=Path, help="Path to the parquet data directory.")
+    parser.add_argument("-t", "--task", default="BinaryClassification", required=True, help="Name of the task gin.")
     parser.add_argument("-n", "--name", help="Name of the (target) dataset.")
     parser.add_argument("-tn", "--task-name", help="Name of the task, used for naming experiments.")
     parser.add_argument("-m", "--model", default="LGBMClassifier", help="Name of the model gin.")
     parser.add_argument("-e", "--experiment", help="Name of the experiment gin.")
-    parser.add_argument(
-        "-l",
-        "--log-dir",
-        default=Path("../yaib_logs/"),
-        type=Path,
-        help="Log directory for model weights.",
-    )
-    parser.add_argument(
-        "-s",
-        "--seed",
-        default=1234,
-        type=int,
-        help="Random seed for processing, tuning and training.",
-    )
-    parser.add_argument(
-        "-v",
-        "--verbose",
-        default=False,
-        action=BOA,
-        help="Set to log verbosly. Disable for clean logs.",
-    )
+    parser.add_argument("-l", "--log-dir", default=Path("../yaib_logs/"), type=Path, help="Log directory for model weights.")
+    parser.add_argument("-s", "--seed", default=1234, type=int, help="Random seed for processing, tuning and training.")
+    parser.add_argument("-v", "--verbose", default=False, action=BOA, help="Set to log verbosly. Disable for clean logs.")
     parser.add_argument("--cpu", default=False, action=BOA, help="Set to use CPU.")
     parser.add_argument("-db", "--debug", default=False, action=BOA, help="Set to load less data.")
     parser.add_argument("--reproducible", default=True, action=BOA, help="Make torch reproducible.")
-    parser.add_argument(
-        "-lc",
-        "--load_cache",
-        default=False,
-        action=BOA,
-        help="Set to load generated data cache.",
-    )
-    parser.add_argument(
-        "-gc",
-        "--generate_cache",
-        default=False,
-        action=BOA,
-        help="Set to generate data cache.",
-    )
+    parser.add_argument("-lc", "--load_cache", default=False, action=BOA, help="Set to load generated data cache.")
+    parser.add_argument("-gc", "--generate_cache", default=False, action=BOA, help="Set to generate data cache.")
     parser.add_argument("-p", "--preprocessor", type=Path, help="Load custom preprocessor from file.")
     parser.add_argument("-pl", "--plot", action=BOA, help="Generate common plots.")
-    parser.add_argument(
-        "-wd",
-        "--wandb-sweep",
-        action="store_true",
-        help="Activates wandb hyper parameter sweep.",
-    )
-    parser.add_argument(
-        "-imp",
-        "--pretrained-imputation",
-        type=str,
-        help="Path to pretrained imputation model.",
-    )
+    parser.add_argument("-wd", "--wandb-sweep", action="store_true", help="Activates wandb hyper parameter sweep.")
+    parser.add_argument("-imp", "--pretrained-imputation", type=str, help="Path to pretrained imputation model.")
     parser.add_argument("-hp", "--hyperparams", nargs="+", help="Hyperparameters for model.")
     parser.add_argument("--tune", default=False, action=BOA, help="Find best hyperparameters.")
     parser.add_argument("--hp-checkpoint", type=Path, help="Use previous hyperparameter checkpoint.")
     parser.add_argument("--eval", default=False, action=BOA, help="Only evaluate model, skip training.")
-    parser.add_argument(
-        "--complete-train",
-        default=False,
-        action=BOA,
-        help="Use all data to train model, skip testing.",
-    )
-    parser.add_argument(
-        "-ft",
-        "--fine-tune",
-        default=None,
-        type=int,
-        help="Finetune model with amount of train data.",
-    )
+    parser.add_argument("--complete-train", default=False, action=BOA, help="Use all data to train model, skip testing.")
+    parser.add_argument("-ft", "--fine-tune", default=None, type=int, help="Finetune model with amount of train data.")
     parser.add_argument("-sn", "--source-name", type=Path, help="Name of the source dataset.")
     parser.add_argument("--source-dir", type=Path, help="Directory containing gin and model weights.")
+    parser.add_argument("-sa", "--samples", type=int, default=None, help="Number of samples to use for evaluation.")
+    parser.add_argument("--explain", default=False, action=BOA, help="Provide explaintations for predictions.")
+    parser.add_argument("--pytorch-forecasting", default=False, action=BOA, help="use pytorch forecasting library ")
+    parser.add_argument("--XAI_metric", default=False, action=BOA, help="Compare explantations ")
+    parser.add_argument("--random_labels", default=False, action=BOA, help="randmize target labels")
     parser.add_argument(
-        "-sa",
-        "--samples",
-        type=int,
-        default=None,
-        help="Number of samples to use for evaluation.",
-    )
-    parser.add_argument(
-        "--explain",
-        default=False,
-        action=BOA,
-        help="Provide explaintations for predictions.",
-    )
-    parser.add_argument(
-        "--pytorch-forecasting",
-        default=False,
-        action=BOA,
-        help="use pytorch forecasting library ",
-    )
-    parser.add_argument(
-        "--XAI_metric",
-        default=False,
-        action=BOA,
-        help="Compare explantations ",
-    )
-    parser.add_argument(
-        "--random_labels",
-        default=False,
-        action=BOA,
-        help="randmize target labels",
-    )
-
-    parser.add_argument(
-        "--random_model",
-        default=Path("."),
-        type=Path,
-        help="Log directory for model weights that is trained on random labels",
+        "--random_model", default=Path("."), type=Path, help="path for model weights that is trained on random labels"
     )
 
     return parser
@@ -272,85 +178,6 @@ def aggregate_results(log_dir: Path, execution_time: timedelta = None):
     wandb_log(json.loads(json.dumps(accumulated_metrics, cls=JsonResultLoggingEncoder)))
 
 
-def plot_XAI_Metrics(accumulated_metrics, log_dir_plots):
-    groups = {}
-    for key in accumulated_metrics["avg"]:
-        if key in ["loss", "MAE"]:
-            continue
-        suffix = key.split("_")[-1]
-        if suffix not in groups:
-            groups[suffix] = []
-        groups[suffix].append(key)
-
-    # Define a dictionary for legend labels
-    legend_labels = {
-        "IG": "Integrated Gradient",
-        "G": "Gradient",
-        "R": "Random",
-        "FA": "Feature Ablation",
-        "Att": "Attention",
-        "VSN": "Variable Selection Network",
-        "L": "Lime",
-    }
-    colors = [
-        "navy",
-        "skyblue",
-        "crimson",
-        "salmon",
-        "teal",
-        "orange",
-        "darkgreen",
-        "lightgreen",
-    ]
-
-    # Plotting
-    num_groups = len(groups)
-    fig, axs = plt.subplots(num_groups, 1, figsize=(10, num_groups * 5))
-
-    # Custom handles for the legend
-    # handles = [plt.Rectangle((0, 0), 1, 1, color="none", label=f"{key}: {value}") for key, value in legend_labels.items()]
-
-    for i, (suffix, keys) in enumerate(groups.items()):
-        ax = axs[i] if num_groups > 1 else axs
-        # Extract values and errors
-        avg_values = [accumulated_metrics["avg"][key] for key in keys]
-        ci_lower = [accumulated_metrics["CI_0.95"][key][0] for key in keys]
-        ci_upper = [accumulated_metrics["CI_0.95"][key][1] for key in keys]
-        ci_error = [np.abs([a - b, c - a]) for a, b, c in zip(avg_values, ci_lower, ci_upper)]
-
-        # Sort by absolute values of avg_values
-        sorted_indices = np.argsort([np.abs(val) for val in avg_values])[::-1]  # Indices to sort in descending order
-        sorted_keys = np.array(keys)[sorted_indices]
-        sorted_avg_values = np.array(avg_values)[sorted_indices]
-        sorted_ci_error = np.array(ci_error)[sorted_indices]
-
-        # Plot bars
-        bars = ax.bar(
-            sorted_keys,
-            np.abs(sorted_avg_values),
-            yerr=np.array(sorted_ci_error).T,
-            capsize=5,
-            color=colors,
-        )
-
-        # Set titles and labels
-        title_suffix = sorted_keys[0].split("_")[1]
-        ax.set_title(f'Metric: "{title_suffix}"')
-        ax.set_ylabel("Values")
-        ax.axhline(0, color="grey", linewidth=0.8)
-        ax.grid(axis="y")
-
-        # Set x-ticks
-        ax.set_xticks(sorted_keys)
-        ax.set_xticklabels([key.split("_")[0] for key in sorted_keys], rotation=45, ha="right")
-        # Create a custom legend for each subplot
-        custom_labels = [legend_labels[key.split("_")[0]] for key in sorted_keys]
-        ax.legend(bars, custom_labels, loc="upper right")
-
-    plt.tight_layout()
-    plt.savefig(log_dir_plots / "metrics_plot.png", bbox_inches="tight")
-
-
 def name_datasets(train="default", val="default", test="default"):
     """Names the datasets for logging (optional)."""
     gin.bind_parameter("train_common.dataset_names", {"train": train, "val": val, "test": test})
diff --git a/saved_dictionary.pkl b/saved_dictionary.pkl
deleted file mode 100644
index 8a15fe6a723a8fdaa797ab67c82361959fa6c0a4..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001

literal 20818976
zcmeF4349Fa8^<4;AQ4#`B65{Qq$oLP-8$=TrOu))N`esg5ofi8)>WZO)KSN(TWQ@b
z#e}-=s546Ai2J^8|98p$TidSPnVsW(XP^CizJ4}4^B%wVop)y5dFOqTJSUdS1}{Ee
zq!dri-n&;^$8IT>|K@Dav`LKy^=hRg`}K(H)-EnFqHXtH@e%EN#dVEON%pQ8*E_EE
zi~mnZDKX^59-ghccIcIoT=1WpckI?XzE`)nE{PFc<GRIli0_q{lAODGm+oyl*L(Rr
zWkO1FK-M0uyL4=m81esIQj+tu?$fPD$F`lj#J6r^*~hYR{%(D`_89OoF<tv~>D@7|
zSFgAMB-GZu;@fubme{*jpSHa%8)kp8VM=mNlK2;!Oh}1=VJS7TwMiL~(m17fOtzS8
z^&WQ^o|2F+IeWX_1A4?;_R@8HJ7v_!mm7E_XRDG@zkdC?&&lV-KQYidrOl}7S7v5|
zXVE%Q<g1_qxp=Rd4rnIrJsn8N0TP}NDmT)B6!iW?2TE|MoepRsp`#9H;ZdRvq{j4x
z4&<a737>T~57L30yfaJ(TBG<#9f+j_*;gqFjnaXXY#FTsskt{v2U_4h$%iux|3L?G
zkaMvP<e>jD9VkK2Y8^PkqP043mLJKsVrcT44v<$pXhuA7g+p6(Ae8Z2bsz^5NVu-D
zDoqE7Ha&Rjpbn%Y^RNzR;O7%Mkei>*=s*mE?&&}dGIM$W?UW1f07O0y{wUx9Ec0oQ
z2Pn-5qX!5jdtnc72NJ?OK+Aj(;Q@}QN49Z>W|1D?h}%&fpmoy9c>oQZs^tM%p>G2Z
zK+pr4>p(wJM-lWuOdYt|*aM`ZZ!^+PAhLK+uQj<AQjxM9JwR&S>f`|kdO%AZ$eH8;
zw9+ft16X49L=T{m^(0-aP~<xga0UOF9w0SO*Lr|G4sG@Tcc$GR9)Qs4L9Xi_pcK6y
zdH{`#d+GsNrJv3d5KlPor7Y|6d4k&Xs^AH7QLv6D$VEasPmqhcT|Gez<39HU<fR8+
zPw)h#h@0pMYV*MiPe7daAZd;#uu9qDUVyyx;9zqva7S89@B*3`lHvt&^5A<fK+pr4
z>wxEUFJKYx<z9fG2d>nCBcu+=!7&o9Gh~wO<m4O)Uo7XzzCuY}=LNJdmZar%{BX|;
z)L?^`H#p)F*#_~3E5rwy`Q7LZ_UKsL8|=}bgEt_rdeD$~;wp(fy#Zm%gMi`QKsyQF
zd4n9}TI>x<60*u0#87*WH?YFeE8d_rx<B>?SBVME1#&aBXf9BSD(~h3SNSm}7a*>C
zaPos(AcPs?aslGJ2cvf60%!P|Z09O-a`=E&nVr`M<X~E)52(%4XdieT8$R~|t?}+i
zA5e<4F+RXL<;VK~qK*fTC-{Jz#4PXuXPLL%2c)L^ZXa+*s$K8_F)VoO1Beeid9FSe
z{Bwg;7#imWt#D~zZm`E?@|zmW8kQTJ;RmvvSV~RL4K%QKMs9G1xE;B{9{cy?2IN%_
zKFQ1tjwo_BH^{||$GHK?4coH$%Fbv{U(h;z{Ct5X@&@>V)JzHS1>~g%r;WZKB@5s7
z1y=a3rZ2EWTpeGK8osqJI7{16zF?1b<Tn=ZPWA;>sJY4)+>syF`hr+)ZubRO$vEu`
zuj4fNT`JCyun^}w$a{`l3)o@SHD5s3_TbSQdJseRB6?7Z8CCV5Cg;2BK}uE((t{kF
zn63w*6gBI?9xzJ}2zo$s9Vov*4|4GHLOrNOo*(s~7N+HTaCg?P)`M7@{GtbhJ`b9b
za*>0de$|7L^xUKeE%EbK(#|620mXG-(l$NF!TjxdK+pr4>p<U~dO*;F*VTdNd-R|M
zGWO}g-I$%G2Q}G9;-`U9$MrxPUuWn+3{y_&K@K*NaOGmaIXx)FBodd}G(N8f<W&!<
zUDSgZ`drZi3pBf?2dz@#rXCPwJ?I-|0C(nRl7E)yTa1*KMi@l$foR8rvTqtdDc&Y=
z$;qCQ29T3-Zy5kFK<iS<vzEk({I>^Dr41k#X=Mz6Xw!q(at1)q1Dfl=)rtl{&;y$5
zz}}h$V2!4A3_uf^?->Ab9}jZH7yzNegOwi`Kq=Ofxa6jNeFJELLnI%sV^0Ip-+`#^
z0oTv~tWvHi>C@5x{XaGU@_G+SwKRa*OlxBRXIb6B0BX_oYXiu|%~=LO*zh3d90MTI
zdQf$q0TAtbFlLnjl%~Tk29N@NjR6pCd2sr710d)D&2_*l(*RPl;k*G@VE8=)AkKTR
z;)wwe`8_D<<p&769z53jfgBvq;|GXx9_;q>0|Y&wxelDo>jz5kQ-B{J=z*g;kSoLw
z#Bes8)MFV)d(#ixl@Znb0D0Ae??dx}SW1WI16G(^J|7^$JgD}5K2QR$rTGAH?7`Rj
z@&PNHxtkBfvRCg9_UQJ8KPb(<vi?9jODg#T@~Q_v5l;~GfaW@IrMf@R#*C)^AQeYq
z{Q;5JgJo^}!4;-<@&_>l^!Ep)nK-~7lw`lzA6Tc>9DhLQ@}S0ge-KJUsy`sgdr<9q
zexRA5H}V4_zXvz(<p)Rn;U54r@pec6XoVYP1AsN=R|x<bc!LW7A^g#R)MJ5s&@2Fi
zQa3IDw8GPG0)R&H&k6vc9NruNw6S$x0MN`;C;(FPPSFA&H3NGU03rBnE&x(v++P5Y
z|MuY9^a9`x{CKqhAkKM^{Z;|c0)IXz0EluPEYt-8BEJWVh|jcOGz5ZBwnhYkSo|sk
zf>>&jeOcgPl|WF6#F~L12M1~ef|R7a9|&^Oyj~zEMT-W3fXM4XdZR#~m8<On0a4F`
z2R#EpN~#SC1mvX$b$<&4p#<&-1UZS{8wdz`Kyw}V>d!zxUiILY%s`NmySD;C3I>!8
z0%wSQCkR-jY{eizwB<q1x<Q~ew@4b~=J6*%AQjW%gFp!yd=>;^>G4GnNWs9dL4c_1
z!O<B(V2?G6f&dZcLFu1@z!e^?3j#U#D>VqDqTG=nP@Ccpg8)GfXs!b@as&hIG|LwZ
za#5*pFbJhViC~b5Vs8b5SpKXK3}U(eVK7L+q*lQ|1J`>6gVwl5@`Q-<;MXC+AQv;n
zlQxq=jxPxY1U(Q^2PUr$21H&Dj_nHuweh|g3_=+CFc@fO!;@f;i)>yYpfs0qg@D{#
z%pC&GFw8du5Z64Irw;+8=pGRQuHaoh1k_?>l@M^2(p5u%CF;H#0`5xo_d@_t-h-Bn
zLO>2;KMVmSXxWye9}(@rFTFxQ3i9?20XcY(6arFF`b%;?KwfyTWoigW#pg3aKn;vb
zLckrE^HT^Q>UrR^Is{mv{n`-F5<%-iKu&^5_;OKlV+hDW<gO4v{_Vl--65bw4jc#p
zcjNbiA%Mu^LEqyc;3^eQg#dycxLOCkJ`(~6dO&j>D0MXiw8XX>As`2t_d<YmibE){
z!Z4CnDH)O@6kK7nPbkPuGKoWr{N)!4T4HrjC`ie&BB7vV!pesNjkK>83Q}>oMkpY3
zc#u{z6r^TJqfkIzdN99nD9A;XmZ5;Smj@%?E(o-;t6o8nf?2_Z06`CEt^-SA3xSp~
zPb>tkkY{rtK%DbHpH>L8f`Q}(QO|=e7YYH*G`?L35cELqI?&^BAyAuioe`vDXOIyP
z^g!x5FsP;x+?kA~Mo^PX62G$;yBk3*GRKf~s>sUyMqq`JIST{J6v$f`5Mdq!63-Cr
zdoVe^Ft`HOu`syGfxU$RvG2j3rwaoMl!+<=&d~A0A|NM2`xXJQ<egLm<RaVQA|MrM
z&x!zo9?)C|PE;%kv=hh`1%w_C@;5CC&NB0pqCgAXdlv<{I5CDCM+o&ERGd;25OzI?
zpI#K0Wd=#RlWd<;6cBYi&^;>(h&mqR$Por=aosx%Xkd>a3|QxBR2WE2^S8qQ`P+jj
zmBWB0{z?b~+G*aO_$LALhJ*p4oCkx(g@KkCO!CJX?Wcx;l-&D13|wW+>@c7SpC7|O
z3U>S&2CR_yXBdcOL}nNu^my?3#V`;<<n=IMfp+?0Ae4Br58}KBt=}jHLWw2&h$Y{<
z#Q>4dgKO1_f!ds?Sqx}oU)y3J1$D+2110bqUku!tk#mZHE6n-17$C}du!%1QQuAbU
zF>rV49xDb~=Ki%}Acns8ivjYt2Nk`;L24?KFcIfHNG5(MMUPzJAT>Qnn22&7q`wgk
zG|?<99B5(jTj3xl-71HJ60B_+4#?jg#5D^CnyAw~9HeA@zi@Dd=0n3lZt{-{2Uj@U
z^-VzB$AgZ8-UKBmd*w}V2Uh8dgIshl76(@`H7*Wvk<_dd$U(PHN&$i%{Id?+A5{wE
zq~G>ZAO}e&$uYt>-<nH<yRpAf1h_M4B>(Qr*5(mF1LX%rfUDeH76Ax)V66kiHbel8
z{FxR3tPpZ20%&K%kqD5Ju&WW^DvusU0OGm_m3$+CX2u6Z0)igUTnF|AMFK6nTRsw;
z<xqo2KwS4=NApNP*zsT|DKEsmJ=irM5)j88d^R`|q+;UgNN|PzzeED<_-u{@gf0)}
z)rbPAxK=X?5cGiNI*<_?1!~f`Qxs^Ozq&+$8h8(k0t7vvxekmO9|g$Y9(?qD6wpG>
z8BxFzNo%4&s~k&<0t7vvsSeb-5e0}k9(1`A1*|jZUKF@9({n_F5*T8l!4>v=91Tj)
zDLxvIe|xaHeKaVExpOqQ8(Th)2Cd=$RW!&&*HO_R2SIzv0L!#LR0dpO?b$LQ2M0;G
z&JdBSEFiD<Aewk2B{d6_1q3~yxef%CEDNkLF)SOr_@wksX_M^LrccK%y*qYGOi9k3
z*rj9J_>>7LF%Xk2CfkUV#wjs6%fAVElJoV5>((wVF`{*sj%^Yn;(GOp8$k95ZQZ)Z
zfc~xfbnDTvZRalWt?SpUX8EOhT-(m^?MOh$Z)EM(wtKJmti2+7ckCM9F1}0eI1+%q
ziS^5i9my_re{I>+xBiP*Lz8t~E2NAXY57t-#xucKzkdC?&&lV-KQZt{O7R*xl5k66
zW3u&58E(l?Lf(Y@F)%D8p}@=C$S?ICcNm_MAS$#`)sJ!>FWs{IYyLH=w&km-XIyd)
zq8TN|AMhrcbb3>uFHz(7w_6#ArhXVUHy_d1+)XbB5QQznqJxQ;zgXSx4Wg!U=hqvF
zW?U=a8Adc|{(E)bBx;NuFs>v~>qn{hQUezqE=?khO-m1rwnYB5PusVNn5$P=Se~fq
zahL0rh-R#q{?5BZle+93SdFN$*z=7wiKd>2^m>n|^}|}aW50{7A7e>;gKtyn6EWx9
zl-`J_DfO1IDbbAnMLT><G%32tk`_dbm%h5)ifHN&t17l38r$OB;CQ0oUtntj5%cak
zxw{ZG4eQ^yJJF1)3n%v?nso2j$v#AlKj$hofM{xm%AG$a8f)ybY%o#lhgSdOPM_T2
z+a70*>OF1garQX0SD&?(AE}Q|^yk0ZFIIeEC7rhC&Ho>-Ss9?U{kM1C|9|?`K7SsY
zc7G`G(ERP5s>6wz>OT8&6wwS_<j!QGNx!u=j3a97J+9dVqN$}ePWz5%Z060=Q;5Qh
zA|<9<H0#4`Ik297jnpG^vk$wQNm-cke6?ye(Tr^?AI~G2l$2R*5mD<$>3XJCysq9~
zIBI8T&lm0Kv$pT7q|@$sRAcRj)a4;{c#irr9NlL*y05|XzP`W6Re!DhH}=|Bsr{Ea
zJZt@vD_pJh)7dL$sqd#9_IF8iz3%8b%UCgB<PuUjQg7DYyOe0`qJI7>EtRG1g66*v
zF^3#8uO(_a=zV@Y(TwCur8g2ys?|AhGg0HS@7HW2n!0-T^PNOvyFab9mnamENKPYS
z&S-VuAW_qlZvu}J&1jU`>NwG)TsP;OBx?OAeVzCJ>#eK=$Q{4x`4rf1O-p)dNr%w>
zer!jSZAXm17ruOoRF>3%jiWP(#=iY!zYCVib9Ke~%S6m`&w5-ZYHF3g?k%Dj0ky~7
zC7QIa@8JhTjU(rWJ|UWFI@;zr(b$Jx3w4&Kfy*mi&ra0bsna{&L`_A%ADEkH#_?Sn
z^AJr+d6F|9QDeQ*^#h2eW^a`eL^O88*z{1M(2qA7iJBvCv@c53biueJoM?u*(d`mM
zlRg<*u{2R*-W7wRh^Fp5v!yK2*f0HamnRBUYBjDz)O@GU<f=qXOXi(45zUA{5>}IF
zQbDiIb%+{|R9IG*XzDi|@7E(5``*;5jfg_FU0*gKYF_hr$45j>y-FEc5X~t0X|q;D
zlg^Bp7Dv=LZNuqyL{pnwFP=a&*4Nm*Gf~*ua8);==0QUq_atg6zr0!>q8ZmukLXV{
zX+gfd1Bn{rYWfc*ni|xn`B0+PkKFC5^|_$X>pa@iXT9H4>OM{@ekh#}rQ;J?&u!}w
zYw2}#|E1P-Ir{kD&%>O4p6@32H`HF2WL02V&(F5)7Om}o=;vu};=fYuq*C#Sz24M(
zea?D5ELDFKmM7!qSBcs;sq2AM;Yk&av*TV}z3$iC`BS3y2Fvl)b_V@;)ephrlx4cF
z!E&LqTx`cJsr@{S+WSp6?FWxNFxNcH(jPwh@VSvh%{6kCP9|!4Qa<q;q8Td_){G~b
z)OE`9iA0UzJ8MlLntJll=xIb_C%tuGCQ)eEGH^Cgv)7nb^N5-@uAj4zXvTnRmlhLE
zDpM$WDN*C)2K`nLO`V;zel^k9mdiZW5{3My-(OGEyeIFtRHCNgH4gtyG~?agq1%Wi
z-JjEDCsE_lLksp0O-;ygeLvCI!sXvNNEFgL3_L>AJbv=VbfTu19XU@D&G2|s?-bFb
zbtO~I5;gW|k$#?NYUyO-Wumd?*0sMz6lPvse3Pj8qk^~Z5H%SZRD3`*WBcI2kBKG?
zS-RyJQDdc3J~|!Q|CV3l>_lUKtUlR`D6~sF=|j{UIwvd-QPZJ=o$?aR7?XWjexgaW
z%iRwoYJA?lYADgvUnYH7h-hq&?K_GPg%S_*gj+Q0gZ&;N-@}pV{D$o)K2Lnz^F-F$
zQCiy_TGQp|eM%f(;ojpb+`is2avkIL<>cyn&v?Cs<G09tk<#P0Sijz4<-Qf$M{)gn
z=$km-{y(3FgYyddE6iW?OPF8(n%~7veN^4otOU^BUntfmvHOXgzX_jX|M$EKQjMDt
z{GJvg^;@ZN@U%Sd#r|-*`@^-`ZfZ4Oa?MkRe#vv)Cv;vSq32I5H#W;nuJ_ZmULUlc
zKWFD3Wz`2vw{!CYOLc##wBJg{CzVh59A4h%@LH`GTFsZx`wa8}uS1=gCrRwMZmeJG
z)Gzd19p7!|D;C->AlLn+b^bG~7w%Oru)JSi-pu=xl>yqH599MQb?3Phd;Ny<)5$YG
z9gc6r_wuAz*NfHuFXqp?8k@)7-Z~igJ~_^-`p<b)wRYZOtsRm&y;6r~o$ue?y!?_s
z_sPnD()T|~`KDAntj#OuvM(ytoBz-IFK|A5SLe@i*{`g1a-|9f=N)rx9Xwk4m9aj&
zzJJl-Ix8aEA+f%T-4EZl$XW;5&U->~j~A8uxe#^h;$)p4vA(+4zPL$!#r|#A)?dNn
zOq%12H6K|^uh9D-rSqqBe3JL4P};AJzDm<RiC!<l$@#8nDJNRe;p)15TFHkw=ce>V
zBs13al4!{nE$MKzp4j>;bR4*OH_pfUx^-fey1$USKFS?_(){=8zDXis?Rf&O8}{nD
z$YSfe)b&g1@Wei!q$M4Ap7^l4nG}Yz<5gHIH`dZC^nP5i_oTCoRC-*1TF-Y=e^^zq
z4aswCi*tkHEqP~T-q);R0<D#kR`bQ$eFc^S8~sK&&Q#iQrp(v-tPDtfUSP!+R?>;@
zbxHGHm(t~}bbRjeoQl<PKv*AMUMJD&1Fh+D_jQuQmW$Z^SRIEcb$+D|&)I%qwc@3<
zUrA~|#D*t!KUR+?R{Q(ZzR#Dco=6pr((l2rnr>F#b5AWr*?J4bBa;6)-<wwIgI4n4
z?&El{9ADpG&x${+q!arurRl#+UA_DHPEzQx-JjOlYbku2^9(EV4`RK2ef#6^eO5c~
zo1w4r_-gHMX2oB*(kV7vvHM}a!GFIGjeg2=oTIPu_-buGlYVP$7)MMt_8!-40#UW^
z+tn&>wc@qy-*!12Spi5re#F&ztz6Dm);v5)_ZwpWo4o6+*v`Kp^!*F5^-Ao1c>VL=
z*FUDZ&%PW*Or`a_;_Q98tMxG>p!T@C#0Q)|+PV3ol|BxYj*ph{U+$Xk|GyWKWdLp0
zMb>ddtd);a=@lD~*!}SQj_3CamdbN=#rn%c%yZ9rTqkO3mA~#SqIiD)v{|cc<of1}
zbLV&SpdpWYlI>09msjgUG~@c|5&bRuFUYrdAW>snO@A%Vvuz$fd1JaI|1mpqp0woO
z<59g+mi(_PnR3>W|2{3!&l628oou{JH1-^>58=xC5abYj@uBrRpVVdVz-lCgQa^vq
z%C_d-Cw27RrMu?)|JSvQyzeac_(!$sfm-p}_8a5;7{cbqP<mcrq5Ya#z0PFqcwsHQ
zLgyvXdj6c9cZJq@;IMq$>-jf6r#Q2vtmQuVOaI*Ei9(fHjVlo~-{~{CDpAvtc_&Ro
zGvbeg)g+o!(5rJDqQ)bx%>!<nw&8R;5>9H9>%|j@#`+q&cP0v38?Nd`6#Fr-A0wDl
z2c^azV7rg){y*PWvW_Fi=hN8kFDB=sq%wt~ml8EzZqRQ9(bU;V>sJ$vZMn>2Em6pS
z`u+7p&3p2WOC@R=UgPlZL^Iy)9lDKZ()~GYb`mu%J+xpC(bR+-*Y^{REnNPcgG3>{
z!@wg%vHyBlw36+<(EB&m?{AWlGOH~j)lKgHL32i{12oKYZ7TZxz})1x$~eAjV;-VO
zDNl0dBWkQC?L26B{ruK}nZ#FU*fMZ7QM1>WR`ZCOHm;wukZ8t$YnOzyqa>skA9s7M
zMC)?~ED!g%K6d+Z&N2YootNu$xP5(a@_Db)J}DiatK(`}ORx35fYo$6dXGk{`EoRV
z#rFCNeQ=NK8soF?*K8w&E%tej()C2?_*`w@t?i|wc1P&(Db;gg^pnlw9LtT(a>Man
zW79(|{bBuhqtVhI7I~w6QA>Z=1>=%%OMjTT(d`mMlRg<*u{2TZ2k$%GbbZLI3V`J=
zPx<5VMdx`0%Z1Kz$=J5?@jOyE_`J})pBHNN_+&m`Ss7sVdV%%yH~2QCKFNc-&!cP(
zIc8o<d@vpKKEIx5MslUn8;K^>>YTWlsI&cUWYZTPX6v()^}${8X}t8+?N%hSxPF9l
z^A$T9XUb~5VKv>f_G9C`UpOyF3rqiDNqOf5!Q&S1%kg~(a($2Q!MQY!7vG0?*^+Ag
zz6P)NT^nD5^Ht#c#Q0tr=`Z+Y(wBvF<h{5a+jkTp3MC%q3Abq02isA3o{w7BB|x8F
zf8CG%yxt$N>o(y$<~WZz4fCAh@r}nf_Iv!h-$Sd%g;w&Vm3(OBdAiHxkre>m@5p<<
zL(4jBV&6-0mwqFck0)Hu?EX+<qWRlBRfiLG_Bk^8PS^D?mIqzsVaoH>s@bGaSQ{^m
z_2^|ilFJ9V!gY0BQ0w|vxGu)?jWtOjS>IcB*YdKR4?(K;t)0ETh|RotdWyyT8AVD=
zCt_}f>k4?2@m*S(w+@dhyk23UKNagK)>9VhDb8E!-1>#?ay-+PPukLBdferDC6Xw)
z?|bk%`f07bmclO{k&Nx=zn`Ddwtfm8r+B@C$0_~CDX!Oy>ovbg8f#<hfN>>>ik+ts
z+k0u-d-PLUKe4~~)%TjQeqsI6()AhEXROa!s?YZ3aW&>$F*u4mzf0Xo*}AG(2Ouls
zNwK}a_Cm|;1=eq@-&(HU)_KjTJs;va<G9|er4gp=y#PEf;CTVh3;0~d)%A0&^KL2K
zo+urk*yka}VlT&kvzA`#cy*kgw4eV<OX0O$(EJx7Lg%5za>w(mGv`^g>WNzM#-^o*
zMq48P+NbT?M9kHzEG$n{E92-aS+jgt=?BI3zV_awq`X*LPwqt}(@;IrT7FEaw~S3m
zrqo)G)ZNEBV11J3`b#Ze#l|akzu4SOF9%q5rmX+YUc2Hh`ILJeX!KE@c0+6PZ(=>c
zdg9Fey}QhxWm+=tq{*^wO#G3snnaTddUdWt)OZBvUB`LXao%-TuUlw6-_V*ad><Fr
zSI%QO52ZM_zOw4|I4c24^&6|6UxoVrwVqF{=PxO`$&wZ%ldhgWjX&oqHh_ea+M#mi
z&xytwyDS@QG4I$kQS+AC&MRgK%JNa__;qz$0zUU)^SO_4<owVlq@Jakj<$JDH1?s_
zLLE8J!SagNvlBIU>huoH^ShChzWB&}oG2BKtM5%)#|h!_h{t0PIeud|j7<-<9MAoD
zqtS9aN8V^(l&GWk!LYu&*dLi?pxo^ewin}?O(0&UmfAS&JEFGNQ8+)VH0N)r+5@S=
zF@LeT-y0+_Q@Qi&jYMgmXWH5M*3cgp+g*w7ziitLEEi?U1$}<iXSw~5JN%^i@71-;
z2W9=peLbX9Jb2wD?79of<<;ZbG^~H)?!=UgstYIgB5J!1tx|p|6%Xsj3;HCi{{E|9
zJj+12+kaQ<87t*1cRJ+`-+Eo(F6D*&vH>HPkV=($la78_R_Zs_8|Uf`mZyv7XPNp<
zSX*b%Jol`}byAs3tzOP=iuH=FdWGddS9!3yo~Z3{o)J)c++9++lJ@mI{D7!h<07zr
zWBu0Bdmvb!|GPdb<%?4B*xtvxTfADGheWM%l-p0%@<n@kX-}WC>xXM4ALg8!(i@S?
zSl{p0YQAtjluD&HlEO`@)j4r9QRB1k*K8x2x_bBXokU~1YipfDlBE|P+MgrQ)^Cfx
z{r7daz31=$dVY!hqw6n|N<?eFiNpEZvkm}S+YLNE@%a3&ejlakg;Md@_7g~UpGW!e
z?d<)B?R7g|-@4cJt+VfEQMX<fu1|vNlUUkdM&>u9{r!L2_fKdWM?w4b0&VG&Iz3W{
zr?vO!Y{%0}bw7{3<9*tH-KS+VN^NzVRF0%vH|Lxrit{w8J5M7Xr!S9Ft@uD|K4?vs
z*!c?8%8y#{+UtL!Ej?OU&qizWc5D4SFZMVPyB}-oBw#-RUE>W@uXkAqp!GSU?Q;m#
z{jxJ}E3N4gdtGH6hb8oR4*G!o4cOm6#+}^hGr6i|-a=d#V$MS1L&kt>mlhLEDpM$W
zDN*C)2K{hd2r};Rmq}k1(vfkGJ+|*CLKI3o%oA?WtPd;aWm@NNWz|Qe(ydfHLdVgl
zl`eOS7yFaGy4{Ktdg>3WDz+hNZ{AC_>V>`X!Tz<cR?Q}bVmr<N%L(s4)VTj(wcfFs
z?%3~Q>&I9U-{9Mn`b3l(hl+k-`=Uns!rD3~xbCHE{TtZMU^}CQb_VP5%X%!g4|0bu
z)%>Tly=QE@4j!vMvYKw8*YR4(r_}k7Iy@X-|Ln_A#8e#rC2XESt=2=;^OcnVcY7Wv
zm+x|gtCe-VTzwx2{dLd#S1fn;EO)KV+vV)M2CUV0*3#=}T#?rDBX-=ad3T-MU5FWu
z*6EV!y<%LqHf!DatO`iWbvbLt1Fh+$HC;-N=f`%X)q#VgGMJ`(6L{2GDTvZ?yj#bC
zS(8Rvy5y>_?v`Ii;}cmw?_+u5eWMnhcRPFE_^<i!vJL=uX}^_j&#--=yM1B4d>B1`
zS@WH>^g8Ol!u}rb^Xo~WW+clqZ!Xpsy4(Gvq|9oINFf<3zFapK%Zsh@61zS&ma}`5
zv)0!4p{3r^k`8nADhtb##8~?w<?8QK`hGF#x7LPnBwxnf<C;w%np$e(wC{-GbFZxD
zUdt<9&ras?Fn8+ojyF+L(eDT5CYo`4*Ty_VlTx1K%tzE%uXOzYqN&+ir34X;-7q#i
zlqmG$jYguP`%9|SYFx>(KhCq#Y98^yv~j(v^Q@2(eeuEmD|`<B&*x=0&-9V7n&kLQ
zD(Ka@4pDqB(zW*@jY|(L*h73wO~`S5KhfC2<=;6-6w*83dM1_@7}qnoKa|*D{&r8*
z;Y3Y!rG1aXVzK4J%6aZ(YS~GzR{Kk_9OUWmpxsy1)0dS1R@y6j$G_@6+L`wk`#V-X
zo<|A?`|)2te;jS&vuHn_XiuN*yyW)UEv4qAqHbJ&Rsphp-chRjXsz$GrVG!z|9##S
z`#e@_&*|iT&cMoX!Ad%5pBG53{v_M$W~Gi_R_g_;=@vTA5w6oj*E&r&|E95fXUlVz
ztqoUoBWfNr<Z(};rt-_H^&y&Z{q%_bM3WZe+dGh`F|MZnV4|r(eVPv?8hc=_c^FX`
zefZo+qUIVoOD7XGJt?2~4bhC132Vj^P3k)3`9z}D58gird!9_2Z?vTc$0_~yy!qDg
zw@O`q;J5%B7l7+Qk@H>ca`ywtdQZmZ?W=|o#r2?^TaSgd{&a`;aakuq)vjmMjvt?&
zm6%S-z})P^?q;H<JlZ%9XH_d^#>2SFa~docdB&UIe89ry16J+*fUE;Rwev0O=R5Ro
zYXb2A=f8S&{wut1$NP2`@7wWu=;ieg?LN?+J{<S;>U<5@UySdyig~_g4BuI63VF_#
zdh*feX+&cuy>(!w<+)$OmVvX0n!UzQ@m{MbW=GDGBt01(kLsNwnzXKD%2}euJ}uJE
z6SaQuK10}ZO51g}m>pk}RRP?#{AJ3&yB=5Wn(x>a=LW}HGf&iQk0*b5`Yo;2qOU5Q
zr?LtF{a4BV8AUYt{4FEk)&6)q&gE%8@wv~d^~^YOe&`cY*-}kM+dL<V<74F+AFKND
zoRt8z&s(bd!20vNW7A}v7scWDr?v~4|3V7U9CFOOmZ<5Vyz~C14s0BqN#cop`^$b8
ztd++S|F2QCJuLC}j7!czG^51$1KvcFPHzhIC2IWsb}IwX)DOev<|7)LyXoZsqByV4
ztM$iXiRB}epD#m`E<DB8ug=PVRDL*{epY>>HQltP3*UFf_nmQGRdSw|=>0OSJby|p
zoou{Jo<qf+Ti5=Y<$2W1tBY?EHGfp__8p=oLxYMBh-PdbJoquuq#;YUJR@qXbjn9(
zdEelcU*qgVV}GnZ*^4N&OFZdA)Eqh|EDurBp@W_B63rNseOZ2@_`X8X4j&VnlA@a|
zX+c!(_jE}VFFs0NSHz~Jhelgs#`_+(=r?4wKj-DqA^m-WT;XP1E8rPM0#BO%Ufnl|
z8e<2HD@impaM9t?L}@?nND^OsSZj}1O*hUP(a(RSrSR15Kf1bN{biC2^W3u@*NK{1
z<*$2-sO|F?R@+gb^OKXTzxc4$4&d>G$CD=3AHm~KmG^V9Y?rLx+uF}7SGcTSZ)Q|o
zIJp-wC+Xg?lYNL9f6i5G08#4ZJH_i0N!D}1;{eC6xZ^tT?pp8NHQ)aq?`$*P7yY24
z{bQvZSV<?&@B4CnYo*_p#QO32_5}TTy+6+0chTk}ZRx>zN3}8E87=+&Sihxdw^V!Y
zII95h`H8&mgVBDR(2_n{(qXP%Wnp=e7~0O4wE06@dT1*zXRljr+kdruRV!ZF#~b1E
zpVU^zN##h&b#u;1qQ>6~Up__jU+*_*e0+-1$A{AKu`>Qo?D43SpGw8U+PF=z<z}k;
z?8{NaRDA#a<@@hS)i0&uv8~5y#f#S=H&0KonC~X@l*l$tHmd@=d%p@E$HMvtY4eM=
z^tgI|ueJP`Qg0cXl1ydvSN45C+x`k>uFJA2fNJfs+VSIil=vPcdGA`T^-x%8Us*{f
zj#rX*yb{&}Vb`bBz3-s#`4sC5<Ml=6_hVS8AFQMk&wuis|I8WktgEHf`o&7VSV<=y
z-)tV=N&EU9en2XjZ2jl`_}9M2iS_&S*J0?-fBVC{FIgE-t^ZQu{zK-=qrtZ+^+_ho
zIX9&@B8vU-($1%i_2a+WG3z)Q+WnC_eNu;Ko%clP`AJmIM^*x~avs5USK4;h-N*R~
zorliUtCXPyd9IdG^3!Imh$fvGGcAs&aoUE{?TDr}xn4YhXsoY1>$8}ilu!JIgrBi8
zVa<4=){j<>Z>{CW-P<cx%afLL(~=IM@5!;2Zr0M9dLq*6J(6ggUlivT)x`XwTN|$G
zMy_YfgN8irNz_z+d9^;4>zwPSNAxF3%Q_s)w-Z?zVC8y|dA~K%Z;0ja`u;hw?|;)a
zo*K*9J<8em?E5v_NMWb0-u-+h(b(=!YwfibK2cWA4|pA1d+$<9`7M&@IykEiVEMnC
zC&+q8sO<+W=@8lvfcK{o9hb!;lHGBh>ew&+bC)Oge^8}X<4Q!$clu1OO4PJu-boYD
zjQAsAHHos`kDt_a%JYdNA7a;u#P&hhbsPF6%r9r(2SHz@@s)WWvoe6!O+IbkCSKrm
z(}&&7#0OIz_nwzG>#)H4>)j&fVa+m7?REu^XFQ(0E%zZszaN;}avySh*Ty`S`;e3;
zIrCYr<Li~KA3!uUd#jWnqOlvsrsH!ua`28md~PK1!(1b0>13j|<C$E&4@xRiD0(Rg
z*Lb->zZFDNXD6*+O*FRUGLN-HA^++3*Aq4G$vZBUsA+hO!@m>Fc(-@xHlj)Q=d{^L
z)VTD}f;~i26LMVNPc*i0`F9Qyh4c;sj}SGFpS&@hs3~Sg&XYtlJRa3MMKoz$$&|B1
zjeT0ApC_7HI@x%cXzaOl?XMAqnO7IzBx?Ss;O#p^O@;;)9}vyhK6vnBqDe!RZh1!3
zSm~6HPDl2?<<~en(bylWPxc}T?GjJ=5H*L+3Cly&bm(BGyhJm`WM7t_Xj1KR_XCL<
zpSQ0XN;LJCNnaKs8rx(0jv_>%#KSz{7R~x#JNoK*NG@OG3Rmgz%=rGpG4ondn5Khl
zkHf|KD9wF3v-4wNf3mFo$#{I>@rBQ8$aSilKd-@d0NVj6+W}hZlj`ZpN&w!^zIs1v
zeSg^5@rkOJqr~$LxcmE8tly{7+P_NcaYAdlGFD7~=UtMxq%M00RwIh@LE(H*_<V%)
z&j<BsK9uzL#~zq#9!3<OkKpr>XiFom_H~dY4djE@)2>}ltKC2RvJ~B?9NjIysXwf$
z*oI^}w#B)@@s`Z{V}Fa3{Vk?e`Rm>yw$ic=d(tOEE0(s556HV>a1_zhooBX`wOl{o
zIQauJE!R&CTL#W1YW5n_Y93M3#`SX+63rNJ?b2eRI6ef&hkQ(`Nm6u^B`t^=FMV~p
z6;W(QZQcWQ@O+q9_Tq!rhp%2AqA&7Zf1tne`kVAyYr{BVva$ELW)p~}mfAS&JEF0f
zuB{h}<;G^YN%mg3gYyZ_L+RDI4!J%x9;vXbF45F)I^M5GH1@rzRT~k7Y`eZ}Le#wG
z@s5v((l!qzj@y=I+%~q8*iO1*{{_~2I_f=^10CgHZl==vby*caEBE!P*Jt!|Za(5c
zZ0@F)11w%_85SK(#QeqTes2&}Yh4C>k4c&Lm@GwZ`Dmp*u^mtNGI!d|ANG@XPTXuQ
zWDVD2t(3F1s90-nSW7o+>DAmkp;-l>T0K`geu3lTq^_@0hlkg}&%PW*Or>o;Y`gO$
zOY}U0_IgBn`tW^%%xa5BAsH*;`vmwt0eNnS?-SsBOQbzW4P11%G|||!^iV3^Cs1Zw
z@<g9bFV9#hy;ItJk&>Lfb^BiNJyYr@j!0>f5FV2)Mi-N9L`vh77@g(cgggm(6Y|Hv
zu#|)XmhZ*M2HEO8?l3$hAv9UnCOT!*$d`ZiNY3_oO8xrv>pmx+7yrZniC|Rq#~ofq
zXZPQ1Ss%L_i21=9=nEASfOhvmYqVRom@{-h0tfzXA|wb5gFpw+fqy#i_u)nU_s<{j
z|L6cZfDWJo=zxk2;Q2{K^D$~#9iZ*7pe;Shm7DwKTse<1CO2`w%?>6(*%h?0o75M~
zCOUu)pabXtI)Dz4X$SuKK||01bO0Sd2hag@03ARF6mj66mtZ)eLJ`ZQh=kr^qc>$z
z_>N&P>%c!NCbOBqgrfs2IPlMd4KFKLC^2_QBi2lJDQV0nI)DzK1Ly!cpg{*%zK~%#
z?Ha5~s$}W)b>Q{Cs*+%g)Hw&R=$!Kjw?YS0Z~zZa6->t{-OvFyIMUo8i*8tPna?r<
z)slfEGG7fa1L%PC4zP9PvXxrtOAAA!?mZiuDrxhT{sLnz!OmMsN!BPyhV#DRmgoRF
zfDX{&z`q*MfBk`$l+u+Bw27|rz|^1v=m0u^4xj_*06O4C4p>$hBFhTtVm$u2tk8&*
zPg9B~2lj~T)-EnFqHXtH@ev)nwTtf`pBT|5E-^kOIalj$@%>uAZ~0rwgp`E1<m}e}
ze?fNiY!}x%F6G}V6M2wjipa7=4J!P-EK$NouP#c&Sr#R-M4mITcie#19owZOd-m?w
zl|-4nOI(}yE-CS3gJjQcapX_+`wX`%fz&oR=)V%$E3R7yOG+DC^dG59=u<yAI|;m3
z@01uw_Dt;Ey+;aJG0F13_#P=S*<J)1ZwWcsvvuzQJyMdrE!k<`%MxX>ck8b4z2jcw
z?^BYdxZb^cC8m&NnsT&m-zBcYi~kmg>(Zrrzr@z<`t<11v29%M_|`8{o|rN+u}_<1
zuiCY$HEG(o)`uwxIZ6y6yB1IIPRL#SH48Zv8dd$u%xv&1S|^Hp6?7mM?^V+Q&7{4j
z11UK`!V^N}Mmms!-k<0|2`;tM0Zk-y)B!C#O4NbWn7+_~oKz#>v(Dy0I*^lhhUq|S
z6d$Prv6LYDDn+4DI*^hrqjexP_a^B;3*0C9aE9SO=s*r~F4ln@^k1d}B?wxr17}#Y
zRtL`VBiU9AO@7k>@~Q{Th$pUaXp0VnGJdNL<X{2`*Hu=f=>XBD2X7tJfs|w()&ULt
zd_o6u^V1m}h+)t@9mqjuP7k1+aseKI$mhWy1w4RdJ`M5!r5R!L0HI_r>;djTLYN0=
znGYg7z!CMxHqOv2(gPfEJIVvJPFgt+pn+4hJU}b-ZQub2dO&j>=tt@(f*y#e16Lb+
zfK>EtM%oEP77yyRCf7nLQnsT9NX=WFJODutXsH7^lRSV{dL?@RORS#g0W`9nq>B}b
zeCGkK;6Kv?q~_^b53t9f%^u*+wEM#Y5IQ}`b=?D$qW41&ppkJ;JwU7U(|H2o3Fp0(
zWnDf`P@7&AJV7oB*6{?nNNDE?a#6RdCx~I(=bnJP^x*3Wo}d(Q6Fos~KA7PNi1QvK
z&G7_QDO=nNke41DZ0-f_NQ((xKodh!yg*JKeD4JadO&j>@SN@iEaJV~3lQ|cl{#>Q
z)FC-IM#6Q5OtPJvoFn0j<viI}D9P)*fELD*w7iZV?s<V4Z1D01M_eM?Al`6=_&_ti
z8@<6E9gBN|JsNcI2IN%_8WK-jC9$VBAZ&RMFx(qxC*eD9kb_){y+KJrR(XRMYVYv|
zR#<w)8?;9E$KK#7F~PY&Zl)H^1xiun-CW=*KgQ$&#B~o&evk`<Fk@UUK%Dnr)Q()>
z3}2J&TxCuUAJ8hZ^ZI}sOpEjZwRsxt1FvJl=RTk{-W}-!N|83k2Uw^4cppI2@!;_U
zACQxn1wP;`^S1kd)O6qN1MWz*3qBx*1&@6I@qs7L)#rkLZjcH?<J_PXE)C2L_P9)b
zQ-fK<a)UGcK(-T0sp+|a2KLU#4bBj^BRAM%|DN1{yz0RxnYqCcMegPXxtQ@dHz2uT
zTQ*<W8SUu{TBnbnFVIBZ0AG-rDIvaqy!7C-(HEp-;oH8z3g6ZA1(t}b;|o&5xAp~R
zX*<dn?9q<=#sc2SzQ77KSNVcF^21tR5X;T&zTheur+wjdoF>0Z#TgP7;+zM0&yi~Z
zJIuP~3kcgDJbFV9V(4B(4{9-^svgwje0M!a$%;XGkb@J`^&pg@W<A&gX6XSz4`{9f
z<rnBd4t`#!2eruaqaM`4v|JDF&id7Q5KEI^^nlRkK~qvLa`4lydQg&{oAjV1e%?yj
zSp+?xxDHI(rUyBgzg-UqdO&j>=(|%72zv0kI`Djt9<)HlK0UY_v(xmTCL2loG*Ife
z9%$q13_XZp$|*g_!6p)}Tnsp;2c?)q;!>N&=k<WR>Or-OdJsdOD|%poX4mwfRZ85{
z1EQ=4eZvgk&iqXB&k}u$k@C_AgGfFQ?RZf3O#>*!+axYI*;CR0a#HRs10V)yT}pY@
zk~oq7_8_XX0pudBi~$gBdJtRA00??Oa~-%^(Etc~Kyw|~Thjon(X@^MXd?4H10e3>
zL9Q4BAar=J@&f}X#d;E#+_bN604;Ec<l}YhX+Zir5Y;{48XACA$~7f@S{k7L#|A)N
z??I`S22h)6Z4BTnt2-D#ExLYf0J*q1%K!))9^{;307P03s?IY2qJ0m>tTKSoboj*p
zQoye<0HQ4qPXBHI1U;a+4tQl6Kx#IeHvkI^zh?l%c@I`RF#saJ2PM7y0AbgI$9g}I
zgX4Mp08!3^-F|+6pa(S9fwOu2KnZ>d@B;)ra8w6!h4_IO&W4kEECXq8`hmMLqPia-
zuX^x(Xg&~2>F|8O3X{v{14Nhy)!xqsO5n9LA0UoB_<CPHV1+Yx^MP3Q>ixkU-QMsA
zrP){3A82PuC4WF(_24Js34$KbTnDaH_XpaT(bONL;z+DNAo6;!tc^do!qiUwAclbc
z{-87y2l#`M>^J)Z>(rX#4+vcz)L8EiLWxNA2Sj-fs$I_yG&A%@en8~+;O4#j;D|r`
z1Ar#p4haCQaHDJhu*Uo<0YC$9Z~-8MKN^sFERYYH1%OcM#sz>@c=}BM&`ADS0U(sa
zn*)G0w(bi6nz;%EKx*D8S^%VGV6Oro1fR_XKx&Np3jp%p9(<c#0NjBeuNDBrIS;bm
zDgavG&j$qnQO<*fx<EkW_h1q6nHG$OKoH8-h(HjFUxh#rOHHyb3p}h62uhJyGZ5t9
zK&?QKlC<{&L2jDY3k0QT(I5~Ic|AyP6bQ6(wOt?}>Ur>>XCO#PwLyV^y!4>%Z-F3`
zz#V}gC-Hj&0YMLFt^;5F83@R$9{iFS2vTzQRv<{hfYL$W46*M70jrd)7zBv6Jm^_B
z2-M~lNrT)x{v-&bVp@C<C_#hIf<P=iz6b&-7&tZv5OqB`IwJ_|v1U;aAmThI{ZkOQ
z!lQLTAP0Y?27y$RI}!wHQ~Y5NAm{<jbznx0V4$65`GP?%Disa}p)@EF3{p|-tzZz#
zpA~{ZEcZVQ1}T`-Di~<sdaq#68uv(^5OE&-IwTn6V#avVW>U!UCBcB82SVz=<ki7|
z$m_weeZinM-Zz6m2tywR1MO^h5)5*Y%_{_y=2ET@keiFSL%<n^`Gx@Eng{drA)pl9
zBSOFxyvv7xTCA)R0?typY6!4I-FHL4UCI7_2q4OP(6Uho$U*FfA)o{;+miGnqCNPf
zR|rT!-rgY~2M>}$Kq^XqNzMnz3lFwT4FRe6d}au!fpJL)xC3*33IRkt4}4aK086xA
z8v<G)Xk7@%NiYdtE=q0;0Xc}=6#~e=J-EF)1hmM310mpU{C+S55P3Z4dprbOrQ)d&
zK+pqM>%iA%LI6PzXs!dLu7-e?*mffX<RJ522(V6Z2nALcM$#%JLvn<ID~$FD1-VHk
zacGgh{6axXtPTnVDOpw|6tqlO`B0#d_SHf`DlXRu1%wU{(rSi+)GTQf3dl<j<~I%n
zxv0`I6cG3FVC35cfmU|aD+p3BE4UCK=mE`jU`cEt&@$$Ug}@c^Y%T<da~|l^3V~KI
zkh~!3dC=uTA)uMYw+jJ+9>`q>dOR)!YLl)rf|TqGG6I4gNL>d8)ii=TlhM=&YLZFf
zcNSxJBdA5@7?MsES-IZ`tS~ZXVPKg8c?$y~%!5GU8KQj;CdU^BSKvAp23I+-w=f{~
zJ^1r<VPJtWQANNRI(}FL<YZ{yA|RH$lZt>`WIJ30q$2HE5kSxbn(M%ciba8T0=c4q
z(Bnb=rbWS7W`0r>XrX)Wq97M1#*pI(q27avQ;Gt@t_ShcivqLEAZd4!?Q@C(qOJ$J
zXGH-~$AcU>!ayyqdxrrH>@kD^>pYDL1F32Lb{HUkdoZPP7|_ID31L7x&HEGoBw*f<
zFhG>^V9>ZQ&@zKb{#c{^)G&~ed*6qFtE`zF1~lRGV;D%mj$gxo6%zjp1F?+A3<HE7
z4?e#b24aZ39tJGXPG1a!5>NI)ocEyh8^u5<v1A{y<a@UmAo6)|ty(cqn-evQ0gdcy
zTMVS2&e&q01b*X-fjcvDPBCzWIX@QzL^%&O@x?%Do@_1#?oQof#X!s4zg7&y(D!~Z
zK>qfiqIWn*O=S`$;=BjR#4n}jkt-aerY8v#QO<+(H^PA?nuUb}Ei8U39OR^1<#14f
zwN1kT`P+lIX5l~+b-IUxl&tR;4$jbgXgJ7C{&C^p3WvMC35ff6&~eb4paf;Fyb12W
zDqV4qiw?%(;3}rZ#X&BTnw0`M==Mn|K+uDK)`9z@N`aj8+g=LfAn7DIMi}Q?b7^ol
z_BV<EcP5SG-<{doJOXH-{GbSMmAlI#06`C|b)eXW2%wQa(;|QsLJma$?Tk1Q0df*{
zH3D4a(ZdKpT=$@oZzRyn_<%@2&;y$5z`meJpoMqKM}o5)Y7hyC>mKZA9tj9L9_%FL
zg}Ao|y9Pu8;@E@F21kNaOk5oauF(INNT3~`&5?l6<-xofQ6Lr9YDNKq9?)C|GGe1Z
zO&WKK0<H5`mncvJ?}1T(pa(S9fidHw0QuX4kG_urTF5yg3Rog(O%!OAV`))<pa(S7
zfm%1B08z(-E_b4Ubq3vw0(WM5j%ZK<LrgTd!k&+#K?yp=M+5S24_3F21|>0fjs|yQ
z%jeOcHT=Jd2D#`uDjMV<Xiphnnf8aufGey$TL$FdAPLtQB65`l<n<mz6OW{%W`VMR
zpa(S9fuNFQfi)&Jg_oa{-YIPo!ebyNTTHeQDUDNNbe4Y;@+9O<$R7j4QW6SSz8AOr
zT<>v*;VB8B$-4FxQbvt@`Dc#=WBvN|>pmx+7yrZn32#*Oqnt-p`0`85?7a$>o3)jb
zET)!~LsF|h;Xom+bG*#@Ln8-sTGKUC7w*lWKg7wcl9jv2fb=aI{?)%pDsN9k220Iu
zkB&A>Em+|Q=d7xo@Kv9Vx|VP5ab7AFnxEEr7x@P1FL11C+4?0*7a9AIb5@=8r~ibv
zKTgkcn^Q__j8B<#BYYRHmOAT;tJ~#RIQ+?-GE+FQ>T3K8)-}z1b3_KOmRi%_?5zpu
zYr;?S?xuqS3auCze1TV6t>r+wmQUtSE%7_W+bZqp#`H<!K-+Jkrg?hJ<Atg;vHv@i
zCFwIJ^qP8)mrK(<QjMv82YGvHO;KR~ur(#?9p%L8{6fALKX`uyC$(DBhbpZD5yM(8
z@LhY6XH6$`th76=;aT31O6!(-p^3eZQ7gqyf0z)oJNytYRHaVu`%~`kA2(<dua-)k
zA6mj|G;&Xyc1wTa#VRfNu%_c$+STEMGXCON)!q4PuS`9#mDieD%g0s?=SKdVZSDnL
zZnce9=yylwC+qJ*!?kXI;`Ws3Ge5b^E3MY}#i~^O1Y+Z7wO>xEbl2m@-r02IGN-m~
zzBeNJ#GKO{>#ElEg;en>-Crj5xW)Zm??Br8Q|m8&c8-@?Umq*}N{a&<nrxjjq2?uC
zth$<y`5&#UkoL_fj#ZuBu+j6V3YR!n)t--1oxha|$6kLS?SH>nd)4Z@-<;>Qs@0G7
z+t~I-CNG!T*54k}?$2u8;|}kvHY*BK7Vf;u)2<s!uU))-|1F+2we?@E?{^9<+)y&~
z1}~SYl`qlpN;Dqq&#@0~yw~Om$Ew;sH&yz(=<*Y}WAFGa;oWPV<<!!&GQnee?K;V^
zrmIb>{M#&E!P~0HV!+)La)8zAcd65D&u*dP258&9`JLF-dmr88gsMG1)cX4t{h7~<
z1NZT)>ZA>;?&chq&bz8=`H%6U15!9ZTl*+BJ!+4?6dJ!=_jO8#d-MCKjJNik;9XTI
z>N1AGjsxS4%-FPG!DgN{EqqWHn(b^R$GX1zOrJUN{e7IftK3J7TJMF(oiDNBiyhyr
z^nTMGA3F5t!5^H~RPOsVd-cNJ`<Y7reRPc=L;YRbIk~jS>RpSf_e<xLQnBYxd*vkc
z-&sFTj(qygx%6z;IMy|I&&<=Y$=5j6bkE45m2=eF#G$I|!}8f(0!MT^$FZv3+jf>J
z*l#oEth#zX>#qAteg1GZyx4v*YrAJZ&ud+s^@FzUlEQ11n|5G7ue2)dgY~@vt*cbN
ze%#|i(!M3vd7&y*{I=ofQajxKqR<{rDzy#Y+3!^gwzzzF&3;a(YUR8WyFTLV>Ah*3
zRy8vHNLr1DfARLzT2F!xl?m&4;StB0sx@9mY`ndi&YWF8`8dbAN<9uksqm!gXR{tR
zs4pFmyT6at@a>g<h2A;d_~Pkfj#gFcd{@f&+vC-nE^%tBy>gT5?^|1xI9YbuDV|lW
z7Zx=uXx}BCRdqCOLauU`E8IVxv^-yE?Vr4}Dpxv$hI^#q7u)g{J<O@4Leop@_G0Jd
zaCX0*^)hOvSDVYZsx8+K4!`oVuSqUlIA@yiey_)z)Jki)ThnDQmO0U*dv=qpK3r4$
zgRa@UOsuNjw`TnB<vmTPFMY6<w^}OKya7V<Bldn<t^J&xr$?v{N^Q?-|K7KCKY#yP
zkyMU#b+-OE+D@tZVjEBC^&Sttg<pBrG^Bcmyhke^=3P}=Kb-yU?DafuC+z-*7QdXW
z2e=(NAjW}p?XJAJ^2q^SZWTK3r@igP<PrDz+kx2HU-(qIw1i_-3-nE|?0;e?&#GoV
zdUBXscA95Z<+{I9I$THB!L<K=vDBMCzRyqPSyfl(Dg3)dQ}=%pXP<d!3dv^Dn)Y43
zL_hJ(XPi})a#`!+(=dO|&C`!Qz22vpt8V=`BJv_ni;iw{WPJ3)E4<vQR6ArF&$bW0
zn^x3xmlvwG;VAuGsrS8{jYq2df4KE#>9&z)IA>L=cv%aFmiB=)-)z%cVeQ-XM^z6p
zDW$gYOZ}a-_fcH!U#DgN-i1y-dHmTiUM+RC9?hxP`1r;q>v*lG&~-pv^^dmw?e(+L
z`ggy|p2y0cJ;|fK>VsPK%C;Xz>U6N$zoqqbt2ItWs(3RewkUaV%^8kWb@e!7b)IQQ
z=@$B2#p<Kb{iNDn=)8i~^wOf`3Ora=uFWw{Dy1b~N~Pn+z(qeVzWE!+nmQUEqm+Nv
zcxuh6RlEP@yBzDP^!jU7|4`|4z}|b<O8?#Jj@Wsrn<eQzZ^R$vT~%v3Xw~7rW^h|R
z%ivhoN=?VK-;{ZSV@(}hA4>f@t@RMo#kK=VJ^z;4CH;`|x*wNZ<Xlx(=ev+PUt+^!
z{XC%bJg8lM_n2Dj>}8%66<aT)-p@FG>*68nbDPvsSMwv+^DwFUY19g*(6ko2KcD^!
zM~k{z&PwecII?b)nCz!GXH}_m*~X(*y>c~PxyE_f`Xl!Hf|Z+Q7kKLqFO`=6s+4*9
zsyjTZO6&7nN9l6Z&&bN}j@G>uThGPrw_-x=iS1XP;G|Nq>5zLrrTWQ)<|7~f`7h<7
zZ}O;g`tZZpb1T;Uc9XZKLd)CLb-9Jc$LjV%=ar>(d#UD)v=@%j>(0;`pR@Jc)&3-*
z`#YO1-0r^}kZauff5+zduMxvqF7RD@k!M9u+@3Oh<|mhVvFg|nb+?$YB`|Wq`83{<
zs+-Dx$KE=)YJ(D;?r>`9zf**Nd7T4~$9Mhi$8J}6+LV=c$TpofYp+^;_nY&)v--NS
zQS&!P{T5RFE_WVTdk;t}=@ol@BX@mp6uwl?GabdF^zW?KE5o^wKWCeJfmd2xEvE(R
znr6N^B7>)0X<yH>y(x(M>exL_tU4-3+RMk;^exKv<LIjTYdou3G0lJE?~||bLRD=0
z=BQrC{d=9`W!4`WIheDj*7${v`*wC6tw%>2rWUMlgrimEn#axRtJIpa$<?_>pKRw@
zS681qO6@C#_o@Sq-lq}z-W6@z4{E79aqY|j&Q*QYHOJrj>}>zXmda;%T2-!bWJ1Gr
z?4)a3IJ+)~QrkJZZar<sk<j$mo9EuO(gX-DQ|+5yzP!o{RafJ+-QUqT6??zihQsQ2
z+jRg~&4*g)u9>}8!E&><a_*wG^ItgoUGDM`x^6LT&o`t_kE7>=wEpgFeWHE4^2z+E
zC4Q%Pt*EQVue0+OTDP;tCwIH*>Nqv2_qXQTR+oF}x15O2<XO>*rz-iJZFqzis<!#Y
z?`%2X?E5Fy?W!dn_V;~wj<ct7k7suEy%4GQ|MGg?-<Hn1!D~gO_DAXP8a^$$uzri%
zyj&WX`SJX3&9`~fmp*jr8?yX4+~Qf)jNN1A9?ZPWE3MYEgwXyS>-KA=F5H_#e~1&R
z<2|Po8u&>j=SZ)e%9m^A<O`ftm7e#9miVmYM61fx&+RC`)^#_urfQ9Q6&vr9(z`D<
z>35ZrOC8mN)uvVcZ5FTK9jVZHnH<Hd)bF(STR2Le((6{q9p8o~Tjxxud5LFD<sQ$Y
z)_QY;0}8De7<_>js!|`nw2jx0I)C=Us}~kED`?*(9#xnR_WYFkzAkvr%+s;S*ErVo
z{OKBb^Gturv8vYl7*<v8es#ISmwWwQrNXz~f3v2J?)Sz1Zm%7+{d>-~Gn&Lsz0Iku
zw&@c3-O=?n{(gV|xIvqEseMqoA6e@7Y~M4K`@6I6FS^TiQkSQ@hbOe%lxq9mu7^FS
zQtvWHi+Y{!&?jMgHj}g3JvpslrC%R&w5YT1nb5jjj)lXY+$l4KV_nldQjMv82YJ?1
z=)5zw+e_6h*oGtbI0U)EmwH^G(C}=1wEZsCIy!QPbMg0`Et(9w#j&Dg>z6EDWb8vu
zsD7KfRNUxQH#wndTdq=lXQlm;D!)pFBlq|srSgICDC<C2V!eA?=Vmi0rE=#^?7Sq7
z_LDo^o>TYli`jaHS6c1mi`MrEgr+NP{;BmBKRd@eQhPo~{dcXetEzPVq?)J7*>I$;
z53Yvi@lIsXyL-=atn0QKJ%T(&KH=n6xyJn}6>j>xWvd4Cy~(?q+B;61T^~Vg`AGe|
z-`Vuiy4~2yyQa1uc9&;W3;Mh>=}Mk^JgsW4o=f%j8hHy}Kd@G3Qc8z3+@5jitfz@J
zJ$$Es+Z<nbnv_<l%GusL548SW?(vV-`NJHILze1yR-gX~eeYhX?WL;s)^P6T`Lc4Z
zcK10)TA@qg@h-EjbF`_W`JZV0-TJ;a>$>^gi0Bh@PIIiN+~XFl#}z7F4oq*@=y_Cy
zOB}08mz%;<BlP(xE!)%jKCZO{%Dq|eosWV~b5biU<3OzWL`4VGbl_l*pQ}Hse~*(}
zZS&~r_;^;v5m?hHbpB1-?OmO(SE~K3_tC7WRQabhoUKDrv$emwlapGl$8FDg88y?Z
z&E?!h-+iXfocR7ejuw@wTwQH<h3;=VA1Q0!)p~DQs(6K#7j4_qI<8!3x`g%zZ#NJ6
z>iweEd9_sT{&?H)#jf+H^nP-Wvy?l0d+R~7^7oh{b=KUze1m6IrJi3@>F}B*={;}6
zALZ53=67OS?|pQW6RM?*zdc^P=@Q49mR|4i;9K~WcUI>Y^1b-M`zttC)q35JN|yst
zT^Fhyj@<9}$Q8b;#}zI6i(L<p_Wgv;BWG{>8sk$Y-3Z^sORYsK&e{Ia@v}UuDs=vD
zTDM;@q4vb~t50yO>YkBBE9a=UiBnp|o(%`>uRO}D(H@>weY!hfb8tm(ldD>9Tj6ql
zlzqUdrDDrR=zhz8eE!L#owqnvRjKh_-rIJTD%fu`Csu`)i@rs}zxp>x<yljy#$Bo%
z4r{*KORw5K+KXT4@nGHF(YO!VuWKEpFRe`Q*j~F%a;|FYyo-E;^cOf=wfd-*$~BKU
ztKq6u4))@u{kfjl^DnLY&8gV<_{Jvdcvm&0(U6?8syydd(@IUpwBM9@gOf__)oWJ&
z-ZWo&{!w|Ka`yDp{HJeuz3pL=OXc!S?)M%(Xq&sjcVBPdSkaQVQ)is@+sU)8D@Im*
zlIsZ1v8F=jqo8g3(@Sf8R;BVOUT&4EJ+OvL>wAsPrc3E@9)6WQkCi`rl9yTo`-iP5
zS??$(wd#kQ*ZsKUBIm4%?ccH1Ct9@5@iOZVjU3F0Ri)=+v5hY&yjHns2ln%<D(mb0
zN_~7Q72fDJN5)4#yu!Pxv>eB_>6lR`_Vn@P7kH`k<Lxc#{5b3$9~x=!oEUIdv_APb
zF}&>Ww|Q69HouO(yZXMc*!G^5{iHr$sukYo8bOBoyS8(zYN2T@c7Hzo7mhU*JKn|7
zap~&snRlle8*jSCYenxg>oq&6`3;^`eU^CVuN%!y@vLjjmD8I1cJUm~ik6<2<Le0@
z-r!l)ZDp>koj&gx&${xPKJLbCyu?eb^~Z1Ov?b;;PrJI>o;%wAFa3@+Avc%ttn2x!
z#<4lZF5?`jBR`b>y~>oYq9@ci#<Qx9=IN#V_py)LlzX~<CnuFUsy`Qcy|c4js~jd)
zweoMt*FMdW$)T#_!(RuD=g(Hl;FVHG_1V$yjYjTi({AZcJS%E%e3jPzZk^|X6;=AZ
zIjQ3l`aZFBdwa)M;pA0EYi+*I)28;)>+0WGsi(H-lzaRGhQH|N#W#QB?D-(}Ih?)y
zVrzflQ|ZzYPAl5MXKiBNn-_UjG<C)o*(R;M%%h?YSP$4e4qUw+5WD}V{MQ%x1l;DO
zR<Y@^?#Funo4xcWAKI07SG!X@?fToz`}2dpJ;}4G-{k1B@8ju5c~*64sf*FWyPV}&
z)#Y=!?d_Hy<E2t-yFja|z0Umo4?oLseB)od(%PzY%BNkv$>eEKSI1k?wttP<xkl&d
zdXBfJQXdbC8YeaVb;}-(b#3BbxIjIx-*|V^5wTqdT>Rl7XRA`z6QSW<ivF}1=x%dT
z>G$TgpRP5H;aS%LrM3?)`0ENzs0#I8sqN?H>2NoHkK4SnT4wj@0oOL&;%QOCnUBK$
zDu07#T}KXI{`}bU%e<>9RlQl_m(YLv(R;kw>S}&wee}k@_FvrOrP4<$3iOZmy3Nb2
z(XARTtzP>UFSW9IoJ6g;ldan5o4l*)YJIZqUv<RjXDM|~aIEV)C2pR5yVW7yRkiwJ
zPe(tTH#o6#1~0YRix2<q9u7pNA4#k6@Go9&)ynfSNAuFK@_UojyB1aNm(J0qQXh4!
z%^z$py{yiAXq#?Ujt_h3WMv$iy>#Yma&_*}C);^e)!sM{*8bkB>CD;nlaF(9snB-R
z-n_c3{9SB6wosqMZZC8kyx4fG`#C$01ufgzj+<lcJFW93JDV=+cvGp>S|7!#tM4V-
zi+5{_5+}<}JH<<__Tp3e?{Xj4O6N!E_?o>nA$?8wY2IlSxz38*<s^6buFhxZXn)VC
z`}f6cJ;ST5tj#ZAFTJ#_7iTX$tX<DLO0V4Oh}gD|5yM(8@LhY6ms)M(QTluNuS%Jx
zue!streiDbn%aKYU7mGi&39{hdrZ4Ot9g$*yiirTeq(&EIuJOrZk3qqr#RZR)~s5!
z`)|I>v8qzd%dB)bom;iJy7$TzPAT2cWb2#>H81h5>Z`Tf`EMWhxRA7O$#tGKow@4P
zk0T;4@=~euj>x;Ez4rj!_3!e!{deeEd0s3wAAdXX`v2K`P9@dfJ>H2ddUx+xPHG+g
zYX-OFvkcBzm3#e5doD}8zO%h}<o-LW>lbP*-HyuH-a1P5{w}tCadv%T>vm$-EtIPN
ze(R9bZ0+yv<m6IE<zoF^>h^%O@T@*qRj%^TYPeS4SXZ^yledkRbsufhpHw>M$?tM6
z;e@JfI8uGDur&JD)RH$i+O*$C-s`_R{fuK(M|Ix*^YB5}IJH%(dLmain+JQw{P{x$
z&#KO=|9i;ixiWaARIYNc4R^j@n!eT6!#u0{e9*6piy9B}tZUB>lMjC|@(3@t+Lpi4
z-xr_QS8!s^>pUxZ;`Ws3Ge5b^%cV;Bzj|uk_-T*taI9&LyFQ~=zHyslRj)Ug_eb6Q
z7dh7T_Z@wYuh#A0gzCLMLw}$9)n(3}maRSMvR|329P2u0OPL#sd;Z0-s>L5yYUtrP
zg=bBrny1IvaO}0$Ij450xccaAj&;3wu>RUL?e}v+)t)a>|GjMelBJ7`eaI=TQs)E1
zLkBDlMBTpDb4|HRysN6Tn@JVk>{0u~ADmb%xVctj-=L?QQ2n)U`3JY}+~wp}rTt_z
zzE3xu`t#U~Y$m0ZRlhLZ=)gZ6FqfQtdePNw9I7xM{^_BZ|2ukLg_Ylz%$kz0u;lL?
zt2*9uN}++DWOAs&e6X$awd<Z-KQm|#Pm9{7&(-fzje}D?oW6J4%;wMXIVe}D-#4;`
z=W0D*W&c}6(%-K)>H;rT)z0UQGO(v>wKSep?N;tV5Z@q!*NUqB-mK90zp2sc&%UQl
z@vQ0Y<Q-Et9z4pkuAk1TF)ruWpLtfb+VV1$6DQo}Tvb=w(WB8jFRpu5z{INlIy%bG
zsFuORs<vD|IQ+`bz9zX<Y&{8^Uv9j|$C<p`YE9qB7>_TjF3e^UtM#5;3NF#>0Vh;1
z6zsmF@y?r^Rjui<YGh2C3KR7=d9i8@N31qVe1B!Gw=;NDVLpiU+jc)w_L?zqpP%Ej
zqQ~<8(7aRV5l$=mLEGFFzWaIuN2^Xgw*02~!{I#Z8uj+0FUD0}!Aq^S#~Z8Phc6n_
zWoVAeyjYbg-=XDa*P7ew6t9*#%7>x#U$rtWo#2&LN9l6*_vd#TC-={{j8|KoO|QM}
z;yb@v&dacY7pnHc)9T;bM<<MXQh7Jen#R7peCL_{zjLf==RZokQG41!o>l#I@ekRu
z=St(1Qn`-56HmTgd*|d8&XMMwTz${)>GL_EdTD85J-^sA&RLzj?sTn^ADrie>ND?~
zpD*rljdNBz-Pl*OXpbwLP@SJ^+Jf;Du5+lb_d#fV{cO&!tGS!|IjPhd|LBzLe}wiv
z#W}0%m%bO5`t2{g(rQhwSPi`SSp)A0S2?LwXnNu!CuS6UaDjKE4_CkE(WK91o)!J!
z)0)*bd@zS+Rju<Vu&UnkCl-o+aGYmNtA9Q;zG4E;v97D1mmTWu{~O1uvihFolIKT`
zx89k~v8sPOX?ecT+CO=#+MZ9i_4jhlYx#f|t10^`HtE{)0q?4QJ8bCvj*sv0N~vdy
z-U~C|y3dPMTOS>LZ&H5M(CDBioZ9LropSvy)pez-;W(SW@hcX$T)cEQCzU!IuT=9V
zTElTRUa{K=onKyTJhbnZ6ke;`v;+Hjp(<6rGAFhud2!7dPHQUnyy?z{FI7279nPbp
z4O0tNIKr{2uFemq_Wt(PFOcf*V(W`s^G-M$uGn<5wqMMVI%{rUzQMby^FLZyA?=$}
zoU`h^ZD*;1{Wf!`?&(8F^$vNDRzA#YP5bO@|HqcfXLwY3e2}_*oz%bF&|4k-OsuKi
zw`TnB<vmSuY3tys4N7#n!%3yRyMCN7*nERiOIzn%<Qt^Fz=>6<(`~KCHO8k*x)HvM
zr(NZ0&#XSMrc$pnDOGrb0}8De7<_@3T3IU}YkHM_zgDjGkR63Pp<|`pX${ZvN+~Pt
zfunWp<o-RaOz_xVyH0XiQC(_>+g}vg!?CJj*BfwkKWlr0Dz5`l^;;<wj@bSWd0lip
z^abmhX1+NhgGYVU2iKi_8NzZe{gxB)nLI0+HviQ6i=Un2Syflp%d1+j#pT0m_H$Bc
z`DFgo62DVCt7<qm^5<-GFYv6Y(C5bPw!KvCoL0jT`kY<p_Ez7Nx*v2lo~J)d2-+Qf
zh-XzJ(~qRpc=#93n#T5vS=&APd0r{C)<2;-V&<V86KfbuQmfE((6)V}t51tX%znhF
ztzvzt7Zx=uXx}AXE86U>3F&LXPxGv+tJk&G{l(S`XZPz_FQaC9wYi*7wHLm#;}q;|
zS1s|dzwg6yoLm~aKH}`@y=k0K-P)qW$+FW<@vNw|{;;an`TgWld-d1R-#w~sYM1@3
zn;fgk`ugOopBHH9w{uj_rT)F*sY*U)8y?|VRjJO8Qu|tHT8rJEPydBuRnt9Ejj4VI
zd8JkAa&Q!$)YlD;;uHG&&DyJ0-~Hx1FSj~MSGfWYmX&LBjH5*zrB~?hN{^e6Dn3@<
zBe161(KrvOeivFFXt^G6HXTCyOKIJH(uP%cbB;^rg_f7t?Nb^J$vLaabB<M=QpT&%
zxv)DNt17nKSlw@Qn<L|+A70^EQ@QT<l?vCdvgfh#XHW7{sjKxbXOpXQk3QMXyQ&`V
zL>9ff_bjKjE}z{ca74Fr9P3)3Z+d0_6GM4cRjT@p;kedbIQx89Yuio#b4lF$n><al
zDJ}P{uBJoZqTyfto22qe>9=)1fB#yMRL)gZeLbdy!=KzKGldhI4ypTzln#&9`#opV
z#rpFRq3aCVUSICcUwdWhfvvo&YOlY1i~h`K#)12IXI1F?TSBiVq<T+U=yNG&$1U2H
zkI?UfTIx<*J9B_jN`<CR==RaA8ZNC~`xcM-dLNX2p5khJV&B(LdOxA{)w;b>?-N+#
zv(@*f+}}TL&?a7PwT&0Qi*rD!_FArZ_Kf@=duJY?Q}zGxS?meP7UIcHh=gq2CtEU+
zE!kp1j4dT1im?-wG?HvnmaJu0iSt;;He<;;q%rp0SR(81Oy>T+d78)F&b{}Xd*Abi
zx954zJ?C>?=X1`z_pF7hrE9QGvPw@(z1Z#VoR=l7nrBq~>S1Y*N@l59^VrOeSMBq%
z+Hm@6!GEt(``oBCTu_^8`Fu6+NM`HxuU>1DeSBK4sG4PT|JtiE{w49I|Me7ez7cg@
zm2CI&w?<!la@L{$jC)SGy?5`eo4-qTt2w0>IM%#-pD*(M>06Sh`u#v(xhbK4OHgY$
zu$ot+HlDy9%;uM0zMUJFF3VoYs47|Wcb(xs4j*%ch^CG17M|O%^;IIet}XNHg3UXx
z5UW+mFSP#7s^3nIC(!evR=f0u7j@mHY<J-*GdI6^=r|Eo)%M5o8qVna4!f@pRR65{
zUT?kS^}42;lFD>=?wDk@D&^noehUBimF$s!XSM$(t9VvjDsr=WgL9ISzSuk6)1TbE
zEjdM7T?=m)cPmK}MMa(e3cJOgzif5+$?c@S)|i)J+|434h+5tLQLnKBf4V`Gv`G0U
zM!quWCNZi79;-O<eAd5+zn-|#Wx{iB{6&;>Vf?hOdVTvB(W_s*{llkguiYS8RV`1o
z;rz?ZpNH9>N2T3T{M=i&B(t=%XVvW$>&{9}(?P#vt5<Q*b;)jhvHJ&yx_@^`QqnAg
zPle21ct)~H{r%eN&!g^EJQ*K+QG!0qK`w{)vQ3t1p3LYKRZEv>SR<yu4M|k({m#Z>
zJ!>DAL|3!M<IRd!_J6bV)fev`DN`z~S2S(E_}Zx>J3_rmTHtn;9NXW#E7_&JPF#q&
z`a)i>zn058?d}Wf^Ld@3M=t$vD0{WsUQsm5jyZi|8b0ZDiq^{h%%$DGhImEQ8?&qK
zAzRXUMb#Rw{86mjr#Zc5tJ?d(;m>bguGjv&B)SHgN2r#A)^HbhT^KPi;ZMn^YE73_
z`1Bp7H4FdtswBEzJo;p&>EE1|MAbm+R7Ck+?>aNyt`qg#6RXa%kkirI{>bzGL{qKn
zxp)mPpSx~U$L*iCZqA$ygcmis?nmqOdi`JNx=O_hRD3gUqGY!6wr_<J7v6qXyHAzc
zaIEsT=e?h5-fQ@mB&sg`s_ukNEB=sJsmxXDe*R{ie`j^PcJIe7^EXR9hbO%Cg!hUh
zsyb!-X8!QPU8}dgEZL=I`Il0>f%cCV<$G59O<Em~+0RF2$1nQ0n9_K>(chKoooK(G
zDP4U2tK?gvRsHj+bv{s<@ijiL*7XKat6Jv^X$>D}{?(G80YhUm{!O&%=he1-+iCc9
zqE@?pm~YAbpesbL4!x9p)uP##iCRtA{@JWOC$18$8b5pS@{c}DB1YBUf0RDU>9u(8
zbmWeOmx!o3FL?f#8d)w9)~dwM|M$0h{@kzK+*>4&`iE!zS?#)bz2S`7dwYD3ujjve
zo@iC0>8icnZvRnhr|99FvM;3GeqXKoc4-Z-wSSYh@M`<f^oHwQF!9dq{s##(bq;p-
zBe6<fZ~atzozZsK>V3pmr^o7CC8HeezFySz$FN(B_2X9ix7p`0>5bp2`5L_JkFRw;
zkdmJ)er`+u={JMCqN-KpZ&o_L-)VEI;ID^?s4D8YuKsqb-3LgiJ}Ldq+k7{p_WV%r
z@iVWzx>OQHtvW}*taMg*m*;4|U?~w@wT^?x;pP1Up`(5Zo^_g7#X+fE<L&pUEslF%
zj`~Y-ntqwN^xI>WCrP4eOs)5x3LSMv5>+=1_62vkzExtaywxAI_N>@-=+!G#u1i*{
zUGt1sr7LRvmF;G=E?upkSryJH_fxIoLw1KZt9)x7=)1PV<r4&|j)Pg{AX>au&u26{
zy&45tpWFA#PDwP?dhV7S-fVxXowiqjO5bggL{+o+h<47izr9+=FIjUm+8@Wto`E~7
zjn2?Gmhh@ppC`#WJ<;cx*fmbf%D?^NF?+s$p!iwekL7++URCWm6Gp>b>hWIpm4!mR
zPS-Q*x3+FJ{EFnS{{F4^bHA@%Xc3?Fs>E9T?a@1L&**jjb|`)R{_~Sde_NGAMAa(0
za*SI4%|)VC1MyHhAC7gn5+S`BbpPp^WR<e^d%NNr_UHM^<1XHi0%@Rf=Iwi<`&RIB
zz1H)wc?%z2_mjt}o?R$G&E=rC{!QDM=eaqjb`h<r^|@Z}ygj#Er?&izK2NAk2jQGe
zK<#s#)^Ju)V|`9@sYxYIK9e(zS4qvvpD6P>wf1}R5?<@RKD>rEdY+qE{;M|L@@*R}
zs#*66v0LSR1uKswM;AVEl!&TUjmN7kU#;QNe7kc;*rsDdt7=Whs_=UIF|^vJc3%lz
z!-+BvT**#W=Z|YGf7ap6>hGAdY~769&t4-=Q@#6O;JP}MrtIEH)N;_<p88+My1tmT
zFBfY&5g!Qrn?U1HvF$r9`jJ4*<?wIjTK<13u~zo1Q}0SzRVfXl<DX_<&)YuJyxSw6
zW3N_~{6p{WM*E}vum7_C$FE*2<ddvYv(EdpI^HGo@3)+|^@=2_?yTv3;;W&jB(qiR
z`fbs|>8-z7uUnEK%f>(6IYIQQ*7Ved$M~|j<I7Ix%WA$&`eqkq9k{cZh^AJJ)2OY_
zZVR{S#>y%U=H8I3)-0p@*It$JFUhEyU0zm?7y0K$l=0a>c9vNkdh^SmpNXh?>p&x4
zffx5nyr{o__<v?KE+A?={yemPp15ZD_4MQ7iCL=E9#-M~-*<WYthQf1P&lRWGNs?&
zs<I;O@xk{ct2DI6x{hg|O_Houqw7IguLrU#KBMax?Yf@zzAjeB$9n#p+W1(tqgvw$
zt@vEw>v6{=)~dGumG^L3_Y37UywP>YTK&o?;aNS0IE9B<`H<%w8dbCGL%bj$0s*(|
z)3ob{qpQ#KKS}heS!HR~KG**7nq|LJ<K6Xq-p_k&J0gj$+jG`k=G(oA=vAfi<@Nd1
z->%;}o-e-a&TgXhJ(sufC#~yC)rKFpf7-e^b2bpSs9G+8!pY}0veH25SoOV9J-6z2
ztJdMLPDfr(X7qi;uIsHT-zUk(D=U>xBD`zl;!n~QEqq3@TdgV=t?96?KX%3U$JI$w
zr?)&uL{+`>$9TC;RPKT7wtIes)_N?5*ZW>9+I4b%jcU_PNo6`bcT5spH~clS{2wzD
zC8MgAr(N$;gK`yIIw0?9!dkViw^j<TmZN_-UdF}z?b7=>eTQky!oR&L>D8A%*q7t+
zzfVeT)j;z~c>g|M&Cb~mRNF_KqVjv8+Io;Tc6ws?*o%^=nrBq~>S1Y*N~~2aKUU!a
zJ>QA)y;<}9tcusFb79O%=l1v>U(bK{JP}m`&3_i<dtUdQQfrr4{l2W)N8aPny050z
z{HT48(Hd{_$Si3tzjJ_Ct*;z<e)Zl_gCx;ZX`kNsP@l*5#2&({D!mV~{#}&)+4^kl
z^!vtY$B5{<G0U<3o&h_EX!`1;8#~7hTt`GzyUuIY+TQaT-mdW&r2AF;FU1l`AP%hW
z2Uf=?%0A>qx$s)Awc52gdf>2E3tu9lY4`P^-xfW7j)<<ZANrRYFY8(T?b7<$=za-W
z)AJ9{>$zQ4+vU%Nx6fsVx0&<!uV)C<S`MPOpR)a|_T^Qo7k0g`D%qv?cV6ZR>9s?Y
z{*FL)u4%Su=2x}P5>Zutei2pW_Cs`SP^8><PwhBML{qcI=K{s6)J}@J?~iP^(tAUo
z{2Se8gZ1@F`QZI`qwDLHt`Ag>ynX-6w{zptW!WnkRaW~dXN=iex<KCzk|_Gpy(}jt
zyt+dYRe7(cYVDh!XWmZ>l2%Egsa5+!vrebrl-Aonp1M*JUH#Xo+Wq;&gx00b{BlfE
ztD=o-D%mSayRX+yqx(JZ_WYn^Pw#?>cW(DTNX%B*4z=}0^l(<+=VZQCX`ucX(Z08P
zzQ57(Vzr-^-SITKTX=56)>nyWIy&d2d5`6~O03fR-QJ#-aO)a@KHy*#ceBzF&EM^^
z*JwK_Uq8Cn-JDGp?<Gd{+SymH=6~ULVwL8|+5ES~lg>+4Y1Ia|iYys@Tw<kTmk-MP
z>9tFe=-RQv*E>(G*h#EX*>COE>a{awcq`2@;#RerKhk#xflaMT9;q<pDAB8G`!D=y
z1{&`(>wEw71AQ)VJ4=r3Z{C%ZRPA{kz22BUqv6;?n)h>1I!_>w{Z9XWNJ!T$SGVmV
zP;DIi&y7~vS5R&LoZ9+tG@R1<OQ-ui@&^JU5m<Y()zM#1?<K7CwHD>SO#Ey&F<Z@^
z*XQ338~rT%M6Er;@9*37gjZA*ZC@byoOzjlAT@g5Q@TD-x^DTN)%@;Zf1a;A?&1xJ
zm3BS5=Jit_ruRBs?H<pP(>J=`8ZY%FOY25|_G{)N{%W>;*9y*EQ!4*uBAQmpeRFZ8
z_6H@=)#~+Stkbi4{sHUsjJ6+kU!VKK_-Z-+7(ztVXZD2T4?ccEGF#1VM-l%60uySb
zkcjFX)jpzFSoyH$!-7XfHHiufN&Yt?Yeboda`k_1^Ty*1|M~oKMD~cBVL`*AB625x
zE)f<S7Tn-==Mhm6FORNyKnndI*XW<i)lL5BegEyT>8K`@9KSmw)yTwkc^;?gnS8Zv
zHmbg+MP@xoH9Yeh=W<aE+Bm#?UMkU_tmykRRd1zJt36Z`F6K^Cglgmu_39U=>Iwg7
zTq&yZpH=Dk^32;;h9(-mD?Weu<itOHtiuabqH9*4TbZi&cK1tFsU|F)TJa^Sk=^44
zyiC<o^#0n~RDFLum!Te2`A;;v!@mx17?y1O>t9ASq!OKeU3_DzUf&hZ8&nhe74F=E
zYGnCl3tLn5oE>(xEmhz5%d51f8s7Tkpm(SS<;cA$f=cxES2A~}>K)#%Nl&T?HRgWP
zi)!Rw2aoij>iH>C(T}M5I#=uVG1YKS_r-&#%747_GuuA1<M*#;N`B9qJ!<P=Ab)mR
zepuH}@JYS>5`~K>cKY8p|G&PbWPsQ9KhS;u|LwQ?{&{f9jUkj_^p~+UMo{(Ezc+L=
z)r64ewvMG5xvE{Zaa29M$2I?os;~6gDPL0!KarF;iE7ZaLM5jrYs#On<siEK4)LSt
z=1qG>)3Wep9kzT1)r6R3w`Wt0j6Cu3JgV}a)qZAIx<=nGl-e2I^ToS;a{DfdodNHo
z4r@PHmj~<kl=?H2-m{e6YhdrA`-_b7mHTf3wXdxCvyM;pKeOU>T0dQ&a%TNLt=Qki
z=y_e~Im=Te*QkZmIebZV;}%g3pVv3XvSerJFsIc@D$)54M*m9HyC>tR)l?J4RxPuZ
zYGj>mA8erNxi|Tj7^=P%+wX6s8s776oj9sNCCZN7MI}0+?e0BPy_3GkbAW0>V_(}t
zR3kGb%{)R?{<Hcy@Bg2-QVcLVecSV?vEP~(dwH=#Ykxo55o_BK&-!Q19i`6V8_=Zu
z398{Q4DEY5*?BH3U44#9^sM8dm#BK%=B$5(YC^8M<E~SU-0|VQn^Zlce#n1^s?U3%
z{e7z8e`lB*5={TIq{^kVRHM6gt(cLjxA5cvnW-im+O{?;)ySwj>7S(PX;7wNE~>t?
zZKIx|8ouW9`219Z`Vx<aYV>oLI~AttJ?&XojA}x3<Etg9Mz$PMr3_V1_N9Z$QT1&-
zwy^@$@S!;}SEd?Ny-t&=RHLu;`KAU{@50$fyi^n3*<YkK)yM)Fy1hcxv;W1#^{M*4
z=yIb0)$n@X)@V#MD0th@W>lkpxxM90s@`6uv$dw0Q0lGbZK*~c`)tbFR6SGHBzC0g
zYj&wb1l91zJw3Zo4cgRb`MXr32M)gdK2>k!B`^1(ns6y`WIw8rbDoSFK-Keh?Hq%s
z`kw03Y6w;N&+K+p9v9U5oX5L;^82P$?{QAyVRb&NPEV_!`}>h-_A0%9+4Wowj;W7_
zsUFXF$@>kr&m}1i%<KKxzun@s9nd|V<`Vu^wUbt*qxX5!@&24LKFq4W3FRsH{gqMs
z#@Y{9#m6chwewz$KKDEB{>iAlK{-C$&Va|mJT%W!HtD?v<-%vV_|IFidpwQX_f41f
zgNN^)6+Jw;KYYx-lcT6c*GgYzELHEF${&0|HDOuAFXO33_Ly{k0##43t#u|*^&Po2
zW(w8tiO=qyPBo}en>;h9MrZh}?QE*vwX0{&rJC^3#j^{jMm}G#{35EJbFcSZO4T<b
za`g(T;cXU&{z^3{XX2}?sYb_UALpa$9Z_rFda4O8_0AtdHS)&H_FJiX7VVuAOVt;V
z?$S=G;m=gAxQA*`eCGlCsYZ|gW^FuGZ`hXfN2n%*-fD1^YUFRFqK;Gb^l2S`imI>7
zSkF1C;U|CVbdhS%^a~4;s7Ak8;OaH1-fXW|xk)u)^PoYusYVW7wDBHQPt~K3g@n-S
zuRPHtE!FV(HNVL~HK^kUM;@aZoquMLtW>>w_jJupHQ}?gi*r(qtXt_u9;%-EooeK#
z>RUN+XhEvs?`_^vh-y&Dzq1xg)|5Z=d+1z;W77Q%?I^|*A2psRdOON%yThwpO6w^x
zzrx+~D_rllf;`8#zMPD%d&ct?=5NvWBCF?biT=DL%6lu?M}7Tz;HHn;|L>=va9;sm
zZG6E?8?XQD?-IZtZReU|0Pp>UdO7L6PVfFr7{^Y1Uj<h4W;CzU62!l)ng`FzcrW_H
z`R)&QYP;#we3`XR9eA1Z+^2P4BCYpNl$)64X4d+2r~8A`^QU(IQBgU-Zgu+uvwFX@
znzz;IvEqbrcyq_$o$41(&6n2q3~<16sG5C}^v>&sf7#<-;BF6h|NV-!_6wNxekt#N
z2ETBZU!c4nUEad;Nyz~3<HHy~vv;3Mz0YshpU#~9=`g<$>+)Ef>m_Rcm+<#phwWn@
z=o}2JC&#|3|K3;CY4<JB+9B5VvW`#Q??2GK{LIIFQZium`;S%JtV&0;edP>uvB_`#
zf8T$N`{5hiKg%$$lyh=f#e;pvjGcqWOTRMw;nDqzis!88Y=`u6*Lxk-Tcn%=9blc1
z+4Dus9v9;7T%45qBmC81`{ELRMgO+3^H)&M*wi!8IErSk*7qQ*^JjH>%==R~?bilZ
zHtmz{^CDF4cg;&V@nVP3bNif<kLdL4;v3V<$mb>Tk}qEDFzP4%e6`L4kKT^`SRZvx
ztX1z9X8WVr@kjnpuYPfwh-l*pI5+I!bCLD>J8S=99iQIuNnY&0eWGd4Xj&L*=c|ZT
zZlc+%^?ls3b<!yyt)3TP*Z8jQ`{h;I(>#Z_J~`-}<h%>A?`uji1v#ypoSHAu-YZZJ
zV)Pqfo+;aTro#JuN(NZp7ewJAik(>3#b#ZX)#Yt<dTtq~5_KLB{NcfK60aP1wae}2
zB<U>|z1N9450iC%S;wcgU)Zj6dF@wX%|mZ|dao0;K8f1jXZL$PD?ecskJam7M73Mg
zb?&~>oSnB&qU_jz@Au{ue{f1Z+&+&7<@o6SdQtd@Vki19+4NufR&2k&l@@w<&%1Tv
zk_+GOJ|oKhgYer&w?B&OSp%#$16Omn%Kgov@HNX$z47Y34*dqHuZIRtbLu&`n!{CY
zKO<MQ%QlWe_Vgat{41(<zqi{}-gc$y-@k3xj+6*k&mS?mua)6^rR>9Fb-y9{-^@K{
z#ee?|t?MuJ`jy`6@cfhd^N+Xwy`iHiR9^2ZYTwh1`eQ<_y5p`>4%k0h-Tu*5*F&q*
z<7NJrTl4+@>q1fh@b+9J&m$78e5|rpZ#sIf!~Gri_tVMFb7AS~b5x>d9S^-k)!Q~_
z{VP;)e}Aia+hF>9b4=a+J$m5a+warx-pWf}?n5=<QsT&d$=A<$GHw7>&)c<gIC-D#
z=<(mIjZe;h*p~E1lJg&WtHIIa{Qp)e>UeVg`?QWfMb%ektmhon@RK+n!pQj$w2J=o
z$LoDQvU}Wsm#KxU$6r&j?YQ-%O6xA&n(zOAt`%gxv)S_>?eYV=()I5*#{L-E_Q$Y#
zUtz8NnofPr6s^99X0O(LNt~WPwfnB{x(^)6$6e#!7^gV4u|o2D@X#EYD^m@sUZ+V_
zs?pc_d{cv}cj4?KUaAT2>@QNAYGi>7-Cm*U*>7wgaL<%Ai5+P?zGjz7L{JTX+|#og
z)u2s{mcL6C{TS%S$V=V9s`&?K_tEbEd%co8j~wIEX!jS;`%&cc1<Nm@>N)p%-=$Q2
zGa^^7pc>w0ap<p9gK{Rmx|(WqZ1!<Js@@T`_N}Ly@KW#mF;pXO%xu4vs%O#OIk8lI
z5$P`Nq#FKA<%)Z#2E}(Cu%9aWuZNenvfbDE{wDhSO=RSWm*-J;GrNB<I-%`u9`?ER
z7M?sHGp$z%hqkTFN;NX-PWmUQdK$3Z2My29&+eX1xdt_AlV=9i=nS8=olVudcJ<7;
zR1-eBcvee0N@M!x&u!yMypA)VJlyeo?D}#}0f2Vr!8skS_Xm~Zy;gIwIz6NFYDKeG
zelHN!Zl!fJPR*Cn{1yM_D{yefbB*WT<X>WFVe1|5vD#0pPS0rjF1MFT?T*&^$!eS!
zJjJZ%C^s?74fDM|kI$dnAJ&(6JjwlG&t2|RIJrOUv}a+l<o>Ye##c*HjchrjN*Sv1
zAKp7%dOl=|1EBoPDSy;2KJOza7e31+A!ga_*|cylUg+-eLZ{YG;q#S}0b!pPM7RI-
zFQXdLJh*)y<>-6|qkp9wyn8a9T1_=!Y}GPrsYcf6_Q3|KYWv-2(0~4f^=Fm-;Ff%P
z&JMfUmSz^`N2uGcSZSWAsD2}=-Msc=W8W|A3(`8d|FD#~`+}g};=LT}A?Wix*1;K?
z$BXq44_Z>Ezt`Y--`M;T>{o&H#8_8G`wLc199l4hu8Vtb^OizXgG&CLwOF#I{GlB+
zXM9vXmjIj}{oD^ekLIKI+$QW}j(yB|*yj}W8}%Fg9{=t4aB979O1_+u52uW$8!nHO
z2=KmR?)wfe=dkHrm*kdyBg6Fx=b7CYLLo+f8CzonRkh>D;Lg|cFv^3k^6+LIwtNOH
z6w&5O!;c>Lky#wfir475pz`@wI2Yso+S;^`<aNt#EieE55Lm6ZR{MMrej+JxQZoE$
zg-TAP65Sl<3S^}7yPUFb9qJXHS48Mfg`dJtMetMXTdMB-Lbuds-tx(tJ>J{hFIA;R
znf=~_=jgk2;*tx$MA@-uNB=v1%G>!VsHb?|K|STap5nY_oY!2OHrAf-kH(dvs&}78
zwD)Y=d+=n-6aB>xuWN>X!M~jJd<K7pKRd~v1MTDL$-Z<@IXZsl+sfIwswovvl=-A+
zFVJ2%xxIjY!@r&E-}1iZ+};my&N$ASO>TraTNi-)0`3dAFJN58==r(wzFSteCswDY
zcRa*X^uhdZ(d?Dyt7CuCzB!g97hZ=styWUex(_wV9rsx^_gTCA#IAJ1cg5!~pPcxo
zk9Bx~N_5TYb1PGI$~?N{tR?@6(hrLEzHZziT3(`^C-;w&ImpkPmLIS0isucQDZ9=i
zb^Ck=_>(!$Uv_cTo37sL!ZW{dE?4r!ob}%YYFFHnPqX)d21j$+4X5qD2|t0KsCmD4
z%l=v3g|m-%lh2KLXMd5}R3i&y==KU#&wlK?j(yj$@4C_F7GCc+yxN8JxHw-qYw~?4
zO5OR&w)=640ao=J+n!&G`v0AtPp9WEvV60Jt!XBW-akD*Wh(j+jmOuyTDOm>hI_g%
z9+V94;1pf^miq52mK;^`pH=nO=)44s`-nO2;~Dis{yWsqeBJ}??^6x`JHy-%dY=ng
zQsq)ws?lA$R^)kqH)`oWe`ePctI{#LZd#rvg!+j3_!O<b;cGsR&!1eM`x1{QxjsL4
zxl>`PO6$Sk?*{uLQvjOXE}^{`*ZeEW+E;q*l&`7!KSyDIRyOx<R_y_+c%na9(f4T@
znYYrZ)gG$6?=!7-zcuhN*zPiV|MG7)P%hS#3phW_*=!zW#~=Aaz52=fLCJq+KMz@z
z4xYQTJ$Ip89<JBk;r*KQq)-xS%>AYpRsVBnt>R@>I-=JXaMI?#|LGS`0cdvnZ`993
zDQC0oG&_Fzxxg*ug?`ywqZU%9@+I-nFDr_F!*A624a(Et{n?~`6Vc8YjGlEo^b&O@
zZ`%j^o5HX7@+*`FU*#d{d7}RHJRw)zao4GHMeg`;-%YA^&5MA4!@r%h4g~(3x<6aR
z#j14t-{akuu2c6Rv8x=-<|$gfc(<2#`_!Hv?v#8)r(YM}m}W*^-|y6XVLz0rW!BQd
zjjYq{gAG(Y_a^@mL)Eup`~9s{!+SdGoI{$WfBtwMN8+vD7Ti*Q4i7Z`{-5zn^pCDS
zN1ce*eiOz0+fyn4uk8lvC+cTv{XSOtg;nYJ_Y*LC&*Qv)t9}3Qe{RR~t-GFY)vnLt
z?z}FXPlEGFlG|WG_BZ5x{lEYE6W-=g@cz8On|-Y9VI7~-*3tRTr)Tv(5AJwROYJ=^
zp|P*+A?h5FnUZE6p^AMP?cJvl_4GkKbqWVw^TDfKdiN`|D?fIn8>s(@H+!6No{iJ?
z?RNToUT;0nd!1<KB%mLGulWYH`(26wypA*ak3-nbE5N?3yxOJrxk{dgrFA?99MIo@
z{subl<XWF^Y9#Mlh;t!k&ZQg@KDu~z0oBOo3zlC*)pPFkzBm_x&U;)racIF1I`8ql
z&07jl4J!F})?&$;@+ZoDnb-YWMdfIf-BzWebsmjfcDXHG^d}9w+Lji&@B8Id+EWd*
z@1<RS5vY97zcy_73|c7u^9)c<c>l2D{X<m06V>kUufrRLB^&?xmr)I=STzq7ywJYb
z(Y}aw&I!)FG}ga?b_VT?6WSU0@dH0Ln}gZ$v)ccZw{^z;=irITQB=FNKF2#HpRDu2
zIzG&=zc+L=g^KxK+V&ZAsvp{(uM`8^HXdjeceCPk%DG-f*CT<iJKtYX?(Qsir|sLN
zc3%V0{9QDAmF5*WEkAna-9~SJC3ANQL+P9@R_ls!Zf(lB^C=F<%X7JC^?_G=d9};x
z`TS^C+V0*%oxwZli#!Kpr=ZG9eV6Bf$)@pUms$Smw)`s1PZa&WkMhKOqZ7uv)xJ0W
zXFt4@3gDLZ+v@fV?F--Si}2+mX#Ew9yJ+?*^<SaCC*!Hrv``bqnzL^%{Dtp!KQi*f
z%kyX<d8#}(Hy7n4R(a_?9~<TDj&gR|`98e(EiZON*Q`FbGBrl-hcv6d&+7NZ$W`sK
zjidSU^d8szE2_TIYo~lo72{qh<6cXuTuMv#@rdr)wPHr9-oldyWTu*MXxrMXR3oGA
zq<@mCr$L#9xv2WmwvBp<YWSMZ<MUGu>PtKxs=E72s@G{!HTin%v(k1p<=|br+SYwm
zXo>#whyE*!!~c7{4Es#)FH)P<&&UE9y1hab>mrS<i}WnoJ13TM^hKn*w3BN1GnFgu
zp&At58RwZKx4<~h<i-$+A^OYM8Y8HB>$6?Qk&HO`Pn7%I18c>x*Qx!bC<k-;J9y`6
zyL~AJh|*pKs{gig46yGn`a720o=pn}{rHdGKaRKgS-jUL-tF_>mpo9rW!1h^+|BDx
zDL~QhJ64q+ul~-fUAXV2e&5wQ9_zGmI<v<aM5z~|*vb38KxXwP`9C*XReweG3sLRX
zy3Y~LY2xdgChWiI>DeuLoMls^<?m9B9ys{+`&7M^m%QAEYQm+&k^QJf&UrF!09DW1
zwQ~%j>U*kBt07dwch8C*PBm!EzLTS<M%PMTW-L|joys44K{a7n#4qEiM)sI=e*#tc
z5AP4!#*=yD#+yBur<D4>`SSc*tDZkFF97oba2^!B-_@;jBM&|A$#cI`jr>$`9+bNC
zSa|DCSA364xe?m-JY#qI7(XjHm6kzt^QJwcsd}?I<360?R>I7OamzRj%Eg@dCfE;H
z+kU{dtq(}40Jhz4MZe#H-=+x40Q+A(y#E#6+wtBm;(I%uhaNl+@y>yF`!Mh8;r$xW
zUyOCFddBxW#kSU&M92AjM{bRoLN$Eiv%9A!kNY)hlV=9i=nS86v98q{wk7=$YEMGw
ztp-P_M*dbR>Nr(TpVslGsLFqM&(Jnb>3{B)u=8tD9Kf~ZZ&Lo<TCd!i@9@?q2fZU_
zo~mo>Qx0?bEoHReYSVoxr2xR+CjQe3Idc3hA=ktG@u=tKw4WIFdDzc9qkhPLhdP_j
zd!YS&s+b>Z&iq*0>vM_$cHg&b=OFrc-oYuR?u(*0|Ea^ARx4>CM&~;i{VP@P9&`8o
z^$loJ{sc`Y{Dq-?Ps`4eoc_<{>V_t#|Nh%!(@{+*IevFWs*#E7@;pw}Gx=)UY*c+s
zi_ChGYIx>1&gG(teRUr8pJXJ-f2?>uh>cx*mY-jpk^xpc)Y>m9H(u@L)h?`e#(HP$
zt4i<FjMkSqW&FukW~}EN9ft}(`CF%p$>UMeFDyu+8vSO0tJkP{v%OyBCe?(^g9hEE
z8aa5;#(PvfRgXRvlDyvF$`eh}QVpM9^P3D*gF1e2<T0wz`DYf%O4Yk}PuJ{J6Fy73
zI44!CS18=M1;rFuzS+XoRLx$eOOyEL&+6xj@Llow%O|Id_a0aD8;aVW^PqOHe{V1=
z-h_*}(-fhRNB&T+esQXv@Q=ooqUy^tZ(kXzyw^Kw;y-_)wMU}bjeR5f=2(_ocy{+6
zU0AyM95o|)*749wRK0C;*1tm4fBZ$%c2w*B<TUI5{E5~Mpgy5KIpX{g)IVF+=cHg~
z*6%I%^O_Z}=+B!8HRgWPi^7Tg>)??-R6RdsD*6#s?)E#y^9hsloKO!if5naGz`ND&
z-J0+JpYQB9-xoaiX#Yeh2T|<A{=N^+x3+q{B>dyi?FslinvdG|F5WouW)Joqb;f>Y
zy!7|OzuB}~wyir(DL{;$n7bZ~_j<yMeZ1HaU9<Y!%G4O%?w7pr;msc2%1iBYtAG1%
z7gxK|<$b;p#(#Wm4^iic%#<|q2vyJeXU-j^`k(bp4xgW5b^Wk9JyGWW>8+1e@w6%(
z(dKRHEjMrddqYQ4s9690VEwyQ{$*7<{(fv%x_AyrN}QAo-zEE$m^M!~#ev=4uY!84
zt$&a=UcA|3^!?sx`SJR$c;29yO6X_pdO-jF3SpkhQXIgxcG>Rqv5pe!DCxRuv(7^i
zrF|8}PRv&_cfJz*K-=>vck3N2zMsNh1osz{*T;zBAEMZa`=7b@zvu*W&ed|Ne~FSW
zQS3zh7PEdw?)Y%uP3mN(^`G}8k6gzI|9<py82F^lM|dtN8L+MYl5zi`@a6IPmr)IA
zCZf}?i*HO7{qbz~(}sVf-j2!hXn5zt+CJ9t$@`vIy+4WV`A9LqDfbbyyKLKCx6kv{
zx(}VVSLtl6>9|@#skfT9r5btcvng*=^-Nil*paHQ*`*Q@RKp)P=X@6Loys44LE}$Y
z7V*n?s`8&x>bKMK<M#GS)bixTZeHxrx=v0syG66t_s4S?>QSSyzbN(>b;SOnn;I>D
zmp;!#4;*~^eX8EdOJ44i{G4+sab!QLyqv=!d^?en0a2b8h3D;%enXVQqx<LduK(t3
zJ~hhO9p&t~H~E(sTG+l7+wX6s8s776ojAGhsfu!cz;kfjxJAk3H;>VCaEcG0{2%NS
zBv%Q$dGKO~)_wrIpE9bKCCZL<<382lLvv)VOyB>4s@G{!m1^|0KHt=!>RmYdh?i=@
zJNt{&rYd?req@hH_b1SN=shPA?Sr=GHt^EMOYM3PaAm_)c#bI<z;n}M9bTX;@Z8k2
zXEf#D&Fb!bc~hzd-mkao+=n#<pxx~X>NDzd#^m>q!jlJNPJRzLv~6wH<oA%MJL#WH
zevWTYreQ9szO-$lo}wDQ=JR-r)6vR1X5YzClt*-}^kv3U^`FmV^gSr@`GVya(Re-Q
zUhlh<s&7W*>J?PO+bj<Km1<DV#8+2SjgHMe&PUZdqSn6kR1;q6oj-<Z<c*o_w^H>i
z+B+wfsxKnlrJYp6pQ&7N57nUf&I9&SjUNBa+IXtour2A2P)!KE)!-=A$lppu9jEH)
z(>neXRbQF0o^w>gPyW{FBGsVj7ZxN@jefJh)oWC}*<P=5lWM}|L4$5njU2pa<2|aL
zsz)CS38B|td7?>Ls^Rl%ev^S}P{$9BJVrG-|I8v;se1SB>6)Et!e?n0=cF20x6+L~
zR6X}Q)yPlPw{qgpf>gub+q|U^)u57pXDybjDSv23AHENn#l@_6t)9<}^$!Q5f2D=#
z-6QrqT=*lK_jF<R$3lOysr|{QU#MRguc6PWE+4N!JAih8WjnyDKiO_qiUD|^efWJ=
zUO%jMexmK=$avoYx39kv{XLb}{#9P<39oh~ES*~MC2Cw`_qYKsQ^kHz*bfThBeZ}1
zsXncS(Ek4L-Ls;HQ^oiQ#z)F0H{y0b2PNA;|L{C*?0MSm{^19u=$>+PTYi1tFR#*`
zW;(p}$wBWVXFdn|TUhqDc-!Wze}!V@<sA0NmP4wPNuD2&ed(ZbRDD~IZLE;|{D67#
zyQe2VKQ(HTX9m^i44<`~P1U=0^~||c6F$0lb^%q)55fG97Sv55%QsutnyTmQu&ZsU
zq8;^H2dXeWOp*Qbhv&nGpAW&s+~*JQHJ5MXs&?7NQOKU&<C=d()mM7$l&`6VpD=b_
zD9TODa$~lxT;YC#eJC?@dxbuqdiKA#xIR_i7hP^Npc-E9+Zv6j1_f^$+Kg)SFSob6
zNtL&KC^2uF&Ae^2lV~U1*na`P=fm$&4t$hDbaR{5*QGdsQ{L-s_h;~&^(18wp81V)
zxsq9I99}*zmFQ1a^nIGDUFR}j9g{Wdn39V;`OhisiT`}U2f1@@{?Jd}?Sl<+Av@TQ
zol?$nQHj>xh-SBF_Bw8#(3AqO&Cl&lU*r5Z*8a*mK0F8C8#<ao<!wLgfcqpf8qeU}
zk9fBa>j_T0JdYNVrwY~+U_AjHH^h1Z?6*YQ17Dtb`^r!a-xZ&qi}eK7%uAk7J>`ZK
z)jO*FCs7gg8-6e{s(nPUu;8$eu;7tVO`^g=lK+j!8j(FBXIRkisEFLjpG(jH!3}PA
z9uXBWeRRzOQfNvv5G4Y&Cxi@}SU5&<nwHBjb>iz8zm?2ZCBA0=-lgc`3T12kEd^4u
z(?|Ru00JOj7l9XURn9l4+g)Op%K0u@<JyX2y>1e%x-eg^F6A%XBBJW511)2>pSn-%
zR=FIYd=T(7%{Q&qcRu1YHA;7S+udot{_8r?sz%cdbbYoF2M@Oly+G_z*)Fr6)AhzL
z+Xn>!5C8!X009sH0T2KIJp^t%)+9&a?8k`cDr$d)Y`0zgMM#&A{yIH%7uSFQ2&7J6
z&*-qH>%=8VdNuV})qiVO{}pfPswc*G$Z_9&9)0WXWdd~<2fn=$)`0*BfB*=900@8p
z2)LDi=<`~BDm-E0ftPm^qw3c3N4`M-1V8`;KmY_l00dGefO)s6bHTSD00JNY0?sDz
z{h`&;<Ie~C^y-PNad}>NFT@802P^%cU*ffiTdLiYpxQV%yEQ}ZKmY_-63{z;kfjZX
z3j{!bErH{8zZjgb=`RTi4p#aBTQ5dzq7!(n>`Ub$bEWY@#o*xYZ%XTFA`fqS{O2Xd
zr9irN!>CU~hVGRD>GPHT`fzQ>UnQ-ob^eplc}7mXUTJ*H-yeAYtaV*8@A-j!+6b7v
zkF3^qCs258xk6|V009vAFM-jMvi52^|BPg}9(^fW(<;6Hlvt^MPA4^Qy6@tQi;~@H
z_WcF%TeBa$zklH~-vk1^KjQmGzh8^Ohi~tLb)pjp)IY`gdsg>v(Rtf#zt(;rgm=ci
zfGEd@%Q4-S3H{b&I!&Nra<ID{mD88^t#q2^j$G8FUjM5^Ryn^=5C8!X@F&nFqQ{Q$
z^{+{0tADyyeGW8UrS!d3_8}b*NR0q5`&y@F8&BWy_WZ^Z8`uH@AmB^_O7q<CySe98
z$4$*2p0&bZVpQ$&PXBogtnJXd4vV+zoYv1M`u^`)&-vsnJnMRf_&`7p0i*XV|LgVG
zAo4Y=#>IJy$LccS?S8M<9;=@dtS%>{r;UJr`)Jo^ZC3v`G^=%D|BV^nfB*=900@8p
z2#8KVw0&84;a#-GwH3#D-6VD?FC1V82!H?xuqGhNdNJ53=DI)D{;Cuo>*oZe^zi$`
z1W-@;uBS@lDG!&X@+YJ9O*sX7L#Uk(Ag?D0q((WJb-mW{f3xh6<F$I6h}ZPgdcD#-
zeb&FLwF}|IAh7B{%h>Iw?i05t@1>%)zVen2z4HQCUB?>_*yAn&TIT_{iw<(@76N+r
z>tnqykAHkh@4c*kcS|`VpCI4_0)hH#@cqO4q&SU-6MQ$dJn-6|rPVI2&(mtdE1hql
z_jgv$4|d0+mSdoBsre+nvyOo1>zV`kmDcaI66c+p*8H0lUMp9GcQ*n5_i(kJ-AxU-
z2LTWO0T2KIcMve^d6n0A1C^@U=ObRi>7DNxDF680*#zWyer{J@_Ag~+hug~&`Pa5i
z(`n;LR^`X&-GSA7JGJo`wHMc$L%^)(M<r|R9^XRxIB!$QNAWudfPgy)`0w*<_h+l9
zYn_jS@DI+Dv#NZwuFFLD-D$r5>pB4kXL0!c&}!-N=YxHs>xr##d0u!g#3!1X#oelQ
zl$Uhuey-4)KJWA5l<ZTxo|e~eypCU5ZI@l;W7odatkUJ}d~CbgW4rt(P`X;nSM5GD
zR)rJ&eNgTAg(%^yZjX(or`7-c!`s!4@S3jOe#$Est#%ppYo}kYm8aV9df!j1zD{j^
z5Do-HCZKm-r{3#CX7woByA)kqp=_<cCB3S5J+jg30_92X{w};-r*{5JAP%g*clvXK
z)%ml!--ES3AU+V#LO`h>+^N3@Vr12PT%_}0pCHVC*EauME3ZI)9_V}C>wBAs9y`X@
zzb1*Mc0I=iDrc?Vi^kh*d*yT4v~omv5C8!X009u-k$~0xAFR{kk*OA&^!E4KeVx&M
zW3#T8*)PnOt4sMyw}?|!&L<QE+(<z0_@R8AXzP0e%@4%)`KHzS&PUAVAez5uwbv|u
zZfu*7XAqDHjGmOWSJU}tBq%r-@dK^nKu&${5j|g8%Mszt9WMo6ckz|)%TRL(sGT?I
z)^O(1N9?>3(E47E@Yd`@$C|TS&Phi;TgGlbb)SHPIu3f<W!~p0>dhCf0|5{K0T2KI
z5C8$U5|HO*Ku3KVGIXyb3J16PEAkBjsS%LZ2|>RYoUrLH2^_$Ig&(9=*YKTv1bCT;
z!RlN(R`J+ZB1jnoKmY`sNr3n9AJ{i_;_Df|mB7JW9L#QQ%-+`n@r${?4$2MXCgQ$O
zf&A9&b`<f0fEog#Kfjr^j#bonmCD?%ed<`HE6O;jQ|+{Rzh)(W#_wVf5dA)5wcqoS
zo*3Q*D?z{=1kBn;4e^FOT_-L{0taw_A2`1r*mbus>pVug@`H3i00cn5Z3M9XTiyC^
zr*U$dpCiBe2wV>R{loxrTe4d9S*v`NQ}%7rdoHvxA$vlII!-Cu%}YPGQ9I(M<`2(W
z;V?0(_IMFebqfKb{%QC1c9oOS_Q9;{S)H$=#MP|vTcvdII|#5Ku;KOm1E2pk*b4=R
zNBIE@U;B3)tlDS9(<DdX?8k^v4t77kAbmFPN#Mvg$A7nhYX0xmc^7Kfun42Iz8FnY
zZ$Fz|*NNW0X%{!%(>3bHW?gUAIV5J;t2JJ;&%e@Yk7(g}uUBrdPc*-Qy&%8?0khV>
z@sKyzz%K!_{LZd;`6cZR%UGWiWLA7y-|u+~KkmEFqi@~4OrVeE!0Nrq9kmy^1OX5L
z0pkP$?c>Did*c=&1`se#K=g4C*7nOp%^shW(@|1Z?Tx7M`13LAeaflv`t!qQ5coF%
zz5N3JjsyP(0jCf!Yn<Myc&+YFv#Q)#^@G_J59{Ytr^jcNUm+b3009tC8t*l0J)hO_
z>V4kl<vO$0C7K<t)%^>BYP8jMn3bMhe+<M0--7@MIFEqQetfg<3r6F2UWp+`AOHfq
z5pdc$?^gAn@m7wo2L$XSz}q|o)-06zkwp33>V7G`>FMnUP`Zw_U+@~A(f#Rp%@^zf
z0T2KI5C8!Xu%1A=Efe~!$#k07t=8upDS`k9fB*=9fN=s=^-~8*$9N79gP8=3j?0*7
z8{!555C8!Xu%5uA=1uoqoN-aIO0CZ!QUn1I009sH0T2KI5coHNeu>v6ZmD)pvRnUc
zla>Du)SqSb_g3avPg+Lj|M*|8#|B&@9)WaHnKxS=2=O{ijrOOqx}H~$uxtMeq?@%@
z)A?s4vpMMXJFEJc0;QwYZmajlVV$1Ryk)h&i!u)#c8>ZqWawTA9Kb<DKTz8)`-juJ
z9)#EHjGkkwbiMz3n9<Kl>GA%%+WVl|aI$}j^1fr%x#uzmy{h#5syAMx>)djFot#f4
z9!7sRtDRL_A4Lym7C)=wHR?zH*ITtus$0@w{r>5e`279SZJ%GbRXN|FZg&Z5bsI<I
z7X*?C@V4)Tl0B@(F+`0=_F7NVeA8-u=Oady_xEKb`|MuF>Gt%^ey*4O9hy1;v**#I
zo~r%d8r?q=*M~h_CoW0S%0cP=%=&k|o)&04*68<o?cwD*t>tZ2cwTtG4r2r^hyH$I
z0J$x(Qe)OAU#)kZrE<*rLYV#hq&NO9MHg2nTkCI06czP-PPUu%ebYZaecl-q<o5R)
zyYrtra#53d{jU;Hm6v{bz5cKFI=%N9UanIb5BL9FFDI?ni9X-V==+D>>$P$=8lKhj
zpd61;d;G6A>%GQ5UjKCcKac;mYV88!E=#<qzkT=&0xSvm*AL!5>uqm&yH37evn<O;
z#U;=85><_ktJ!`1qsk_=-*~TwO7_YA4Yg)HmRX!r^G!Rx6Q$n^c8b+ciSj~u>FHNj
zy8qerJ1_lJM(q&wIn}Ilj1h0hu!)6ZBqa`JJqICPb<eR{+dZqFCH0QKy7fAv<zw_d
zp!IsA_TqXr1VnqTmdE{hSG}cU^!#wO*9WrCDz1Uj$M+!MMgmIfHQe$$tFl*m|Iqt8
zEBln<5xu;px80rQ>%XoO-c*S@tKUVRZzDg?v*ICoJfhAcgxzA#1Jzo7X4f-T{k=+d
zvPuW>WNTWb_n(qh4s!c%wJQ0A-M<IwH?_(x(f3Ib{kc|j`|VzrVKjZa`CC=btxCtL
z`Ibug!F@)F(1YLA?%(9J{m^*pM}R$i_V1uvtSuLHtrkLwLcpr$0XZE}xWG;j009sH
z0T2KI`v}-Qztd>{l_=M19e1)jywUSfaQ(r<EkiF5Rvbj}^FVg8;-D0deJv?c7J-1)
zaa&%)+dXb$G<~IdjNjSbKLYXOWnVO{b_B|gS>KD$w!ud82=I2Uo!sARm6}J^nwSI4
zGqUS@YjW-URMf7|a!WYpm$<ugXx8(MRq@*0ujQ2V?WzTsFWV$X;q1qV(f41sT_31l
zP?Yb@ZXfNc&u&kb*Zw_LcKM&b!0TsL`(|jZr%svIYt}qXt?k3Ye7U-mzjTWPQmyeJ
zJP3$LK<|DCqFu-8K81K5URP@M`W>XV?eU+N9G7GcqVG2$+b`$GEETbBhRq-V0w4ea
zAOHd&00QO_uxow=(hYmMPF#{CItRD+3&`hALLYs;JL`Rx&1(_OiP@>+$)e9U&^jK)
zTX?h14>YG11di1y^L~&IvG*Sh?4kI@2m#T~DKe5*Tnhpq00JOjKLOGESG2Alv!6KM
z7Rcl7tktadFstGfy&uo6^<#W{JFMe{0I&19S=$w;U1i<>%Q`+@{6woAqRmJ3w^!?Z
zM=Kw*!&|*Sn_c%gtJ5{R+*rjg%KA9iIicU0Os5Gr7~`O~-UNCsQ2JhLdZNGYn{B^V
zo(K;DAOHd&00JNY0-_SoJ3mv@{K9S!;E@2UeMJ1@G5fs<@jGL^g4yM0b)C~YA6c!P
ztjC8CUymK*>tB<2;h@$Jw1x|`zYxBE^!`g)`DiT{w}&@tpAE$OV1E|3`xo*L0?7pI
zy1)C+uVDSzs?yauFT_8*RdymB5C8%71X$183lyJvix3h7{zt&-_3Zyk8UF(T?F8(4
z&Sagg-uGwi#N2TOZ|9-;+hcbrSUsQFuJo*)pYER?FZ<qzI`3ES_>SAJv%7s^oqmgR
zgQN0g&FB?fSwA<}6`y~9lh)6y{X>-aSgmIkCBLu}1V8`;KmY`sM8N8PQeM+DyG&V+
z-y*({VG|3-NLo3VUCzA4@86H(_RqYP5A1O!0bcq`mF(b+zS2Gd_<hu;Aw%~{-~bM8
z@&i`wl-~8DqFu*|SD<-WZvCDWSHuGX$pnnfCjdnd;E90ReXd12N5U$5c`8fT0s<fa
z0w4eaN(9g!h5jg>`lFQm6~BW32!H?xfB*=vBOvOyfZ6-WnH9g@^_IL{Cy%fBdm=A(
z`rBpqXRY(z5dOh@LA%Qh>4N|W@JE2xdFgJk%kA@$k^c!354^ma80Fyh_3~=-&tD77
zmqi1^ADi`9n;@^<D(d^RRd(~`HRfBRb$<e<g=e+zfYtHf{afF9J#cddH>>NLQhMgM
z)JOpYKmY_l00fi?7+nW~>!(h9J>$2M+c_BRKg0EG`XRwFb&kp{p}G*`zWY4-*4@hl
z3JxOpfjX~7C=dVv5YSJ+>iKd?`wPjxtG!Qa?eEpP?no(q{p7{I^7})prN^HS_Cd9A
zQ0iwE{dcjw8rFjV2!Mcl2`D}9@cO%ZOUzi#)$YHe6i)5)x>C4>`EqqBf9V#1x|4%3
zZ^qRi00JNY0&XK<*LiTfr#qoSiYG+%j%puKEG#4}G%R>zRFkN%kmP?OvPNW&$Qc$i
zJSrl0^5+s^!C}D-Zg(CL6)}Bu%>z=X+x$p(ewRJBr00Plal~mlW@ekb^T+QZW~*-S
zj4o6vH%7@I8V~>h5a5nLwk;XTjrs7hWS7c7I``=O^^a+9Nm}*LdnC`%xNDNAdM{}5
zCr!>INp`DT4p2S_3}01!Y}x4(h||<4-7#egzU<w0jA&J(=?1!f%AUEW@2ojX>{8jT
z3K4(oi2gp2@TRhTP!IqC5C8!X009sH0T9qbfVbycz4p}364Cdqug?%s^-hP2BTkg~
zgNUxbpMQE#;@mxwsM@Yw=)SyjPDxg&o;r(bKmY_DO~B}V;L))w|N7YZ(3L;84e`2N
zl}U*a!l*ytdUKx7az7ih=JAj-#4HuVJ7FaVfB*=900@8p2!Mdw2>c#2V|(U&TO?66
zq*0Dab2luOjH=tp9r*<T5C8!X009sH0T6gJfn1A1)~=cxLA2_lW5ur^00JNY0wBOQ
zfzSa(uGD^gDbcG+>$~y0GxsmByIj8-d_1Z^ol^v=jRW7U5v&6N5a5YG!(aN9ZM!T%
z(yBb=0k(hu2s}*Swf$XZ&fXtSpx|JoA3W@Z_~dp1qOWgroBdY%tJddE-oo#0-?~ME
z_Y);iHLBOSmi;SSkpijJ{$Mn{3Lm$BdfStyB(-WZJzUQhfuWxj`zfsZHpwil*&}aa
z^V;(yqskXG=UcNgQ?~1+VlNV?GdMWkW5vp0wN>929+dwiVWqt0K9BTT<`+M>AVJ0E
zz^i}3uG9&*?LFLW`Axml@hu3jA+YiZ-``)fJ5J2j&eLigj$3q+7*#g8c5f_&_U-SJ
z#giaG!NJXbP`l8D4I|TEl%QgA;H3pO+TY3R_3MWe^3BY+N|Kcq-mn7%K!862Ge#HR
zobQVmVz%0q^ZrwwI<>IgQAw-XWiQeNfn)-2#vHzw^u$%kZsol+JOjsOKR@6+;YE4p
z1^Yk%1jHv$t$f0rGsk`>qNvm6?Xw!svda@E*FUwk^N0gPR@QcywV$=s@y?vOAZOj6
z!^A1d%5Uw8NAz}tbv;9TAfSi9`z;#8eUNa4uu?rXh<we^OGm1XU3Xk^np#~ZaX$^M
z+v~mbUeWdG!a*G?6in;2O06y@q^FI*@$qwtO`m#5vRk!T{ol}E-FW8R(V6!Wqx#>N
z@eK%o00@8p2!Md-1oZC5XxDYT@E$g<^v>})W)i!U7Y?uk1V8`;SQ9XNeGKBaW*ri*
zoLT#;Qhf5fZdR(49)5qAK!u1uc0_-lNZhLFhgZ+tI7=eoRY}L;4c<L?iUiV!OH=vN
zxmSv%Thrq-(W}ZSAcUI;VE^|A`@gF#bF1gKskIB?#2_%_P|FK%H%=gKQQk|XLb{^0
z!(TfnDJgIHuxq}XoGx!XV2`^9__l31_re#KiCT3R9pu(61n$?llWl6+W0F=qwtiT5
zQtzf@mhOJ~)`pNbk|eX$E#-`Sf`AhUoLk-D&k_~)5z*D`{Rx~FzY}~nwLENIS^vA&
znw}$~t6A&SweEAFHhhsOC5IPUKqS$%W7M-N*CnkadNnnl#CO&axR$MCh4eoiB%<n=
zvISrEZaYSds#fAt_RKwfXU$o{s|Fgcv-*3jToK;g1oWOO?QUwwJqUmR2!H?xxP!p#
zLBCGFd-$m277bLYb=FsH^?A#ML=+vqs`}Wn(<czE8Yutx-q{2yR1W?$^2T+^?fTR1
zL01noKOs3q?cRqK>GQE4uG(_Q`a{v8S^u~xLA7x(dUt3SHgMGCO2>#*YSdm_Zw>*o
z_6bz7)~@yJNVmw9+OIDqR&h}B1N;sGAOHe(6X30%#VQ{1x(VpT5hqIgLDX`vs(e`O
zJBxVO?&Axd*7DT4j)2$jYu^}~eQV4WNi@y2B}2I}A6}MNt5xkNFX`x=Pigmcm=|x&
z`G0}t*W0!49dGI0FL1N+pfjIKR;%6h&uID^7C!&6H|jUSi`vy5e|2n%r`=>9;YFvM
zE?cR1xl58-Rl5$ss&H!U<~7{Zjf)nxSa@BsO8dY0bJGr`FH2^rS?gxaig(Q2nzsvj
zZb*97u66|JI-7vnejw4qrI}Iovq|N*6Rm1i>5H=ekXiXhydc06fnym?-p}~X8ObW;
zDeo+8F}g3W)#rK2=>=l&Lf&~_e!4nV5?$5Kg9yZd_4i|Dw#hqx{4N4z!$I%5L%r9r
z_6NiV0$K?4s~tCV%*+eKZ5oKtoePT=_J8{l5k*V2`Y7w;buJT8Rqeiq2&ZkINSRk4
zKM(YM`!4xz7wf!2;#GMWkI-s|+UG{r;R2Pj*6-D>&u~jPc|L|#jtCC|AOHd&00KM`
zV0HeLUGeb9REtfAmU|XFd+vlJnpSUlcH&E$4okeL7WN1W0w5qJ0i*l>6xz4HPZm#t
zB#Qp^@!!XqZrCM>uFo!QcX{ycy^?4;W_zdQ_j|`nPF17zDbV#|wl1(51lSOmIdwtK
zx<Q8t6da8BLDYmUiS<^TlhkU%U;348yDUL+i;8kz-a4tpvuVohlAz!KKd`<Yu(_)v
z79|45x=tHYf7NaRO`U@hKm6`Y0*6lCNecV=3V}MCL-QJWH@7)>nLxn-e(=9~ptn`G
zYQ4JNboAy6*MR^CfB*=900@A9TM3ld)wt8v>z4>rEDmn<SL7Q6QX|l;RJohO8XqB0
zaDX4Ms|TsoHGF3ufoiM1FFYv!NdkSCgMB4}ltBOlK){&<j$I31^5@m#1PTuD1Agj(
z*{#iwYi`Y2d?JBB!2y1Ndf?P{!0dJu@q>UG0=0(CYP9*S{X|qvKfHSG##s`HsCw>|
zV(HfOI89ioQkjoFpS8~uBXWC1)!_D_MaExw!V9&BL!TBqcjOq5$!nD=`7?e80T2KI
z5C8!;5D@*`RoLI*&k_~)5pVzp_<?=(zzz2b<Piiw00cmQJprrFn@4&#D-SyJxnws7
z_MU_ook*Z??}BGzw``KER<l@+DO>Pm@3vz^bk#e~W)@eY@wN*aIO=kxV?<OHb==wC
z?g|lq?1=t8kr-8@{($R2;86rRPpfq}ZqZ31sy-@jX8!it+hw*@%XWeArt&$EMpcPj
z`mIyvO)hnnuvVpX@jD2xAkZw={CTYswh$;dJjxGP_}ahYcrPhKxrL?95-1-W;=h~R
zHn{&$0tE;7fqC`d-)*3p|6jc`<HX?_8;RRg4I36=w4OI<G|iAkIV#QFuvijB2UpG9
zqV@e~NmLy+uJq3FIc5^*!yJtIv02yG&Jxl0t*_4z)@qi$TH~FvXYT1cYt9n0RI5Fr
zg+Kpe*O|i#ClRQ<97OXQ*b4$Y5HLCp7U+5&@&+6DC1BR`hF$UUOWGZliE?iBSI4G!
z+D-NmVfDr{?~cyAmwY9dj+xme@BHz*h*ipJp96Q)UgQ!4KmY`c6JXWu28zeHg@^$J
zj1!n#|J2&fBMuPg0}e9L3h9d04u9>SWR}Y5DCy>v^}l<q={e$7RpNu+K>!5w6Bxd#
z`q;A5ClIZwKSf*#0{jxF^`0km+WB=vR9!!$kZ)$jRgzU|RY|_Kzw6A|`{RkI8fbhA
z->1Hg&MLn`Iv@Z7AdqH8+0Q1G-%bLl|N381KhK=HAZOj6!$efA*&}aa^V;(y(R9PY
z=RfvF{YI?PKs8#l@zy~0-sy00#EBAr5Um=B3%&;d5O5xWp`R7|DXjZ8Nw4z$e&)Oq
zLykZI1b8E`;PU%jiuC`JI7PKSH|h<~TRFlW5U`T~Z}S*fvk26GB-;0S+dHeTGizTD
z)_%cjeC@mByIrjF3Q4c>nlIP|0w4eaAOHd&U_AlSad?Aw51t}+tMxfYiXZ?2AOHd&
zV4Q&7_i|R(8P5S?Fq6QsuG7ZUU$vX)RWofv+#mn~AOHf^6KMELzp`zYB}i7O^*Ka}
zAOHd&00JNY0w4ea|0YmkSL04wuU{f|>%VQX^8dV?TWe*W^`w>i*_btthnyj-RF4g~
zMmz$Zfn&3uA8?*HP5af38#-p@1!A{~=Mk{j%>?XTPwp1>ddGSEuXEeEYsjxV&)Y+}
z+gt8#X_t{ta|zge&Y_&U&eLigj$3q+xK(8idezr2)8udJpO*ru-guR+vufWRUeZzG
zVf6Q37R5J9x}49;TCG0MrgotV8%CzTC|RwdoRh0$r%^xhzdq1;3HZKx?#5XXiB=Bn
z!Um4IT<I7QRjaM~zVM*@CyCYS@0a)t0w4eadI?+)E_bv2)N4c(Z8z@EeASv>BW_i_
z<?Ph!?0)}td;05I<tf^z_&Uifb*kU8%2%)7bN$*e{Z7d&Wo0Mpctk%JEYSYJynpXM
zUT5}ay?*1i`k<B*FZ(2_wHM*|CQzr(f}BNOyFys0*|}|&YyP}e30sI(HOsEA1|N?q
zQ0ElkO~-8SwETYWcu5pJK7LNI=~E9$qH6zQhaays_;<-H)$6~Bwbq||cR&uWsJi{z
zXS2`Fo!u*{?#vwX(zRco@QSKge%?Fx<HOm!cB@|h*Lz)sg)enDx8^Ewi|Xa1^*Yh#
zb=$S?m{!gRZw-OLRWr9}eLq^_MXkv-QgJH*(fZ4sGEaBeb4z+27!pUU^1H&l3!aVL
zvPlwEXAk;y`rX4vB~evtIg9$<EZZ$=Ii;Vpw&Q?7CnZt!&~nd$XV0CGSgBUOH7mSR
z+I{51-gPG65B`U1Y$QhgfY*7l=Uyq6ZcUHVM0A~5yl3@8FJ2^8>3c~T$}KE)mbhJ6
z?eAe%JXX*DW}P1I=g2DAr}bPfvz`|N#k(_8w(F&0FA}Ia4uLqT{r<K6U1!eTA5W}O
zwRRz#xdg6dD_J4^PX~!8YWKb*YSY(SYFg*l=nWspKC9Z>K<VRq5O5=bA4A(zxf6Va
zxLsM5eeEm}ec$@}3}LOT>{E*8nVw(mf2HdUNmOmuE_7esIj1DvREayQ--mcsMK0)(
zL`2mod*+_Lv*xU2<DRUzn-$Ob*7KX5K5|}yf&<TfV3z;O@v>S6BgYp=<@QH3bxObV
z$|rn(f6?wZv04N9125m-xv*$q|F<s@f3?a!+<)x7|A;!@M7Eo^=W)^Ov8tZCEuAzo
z%6>Mf{B~lMipp2*Ik0A}<F+f_t&>_jo2J|@2?`GI1Agj(sO>21PMyH{6FvHWlKW4|
zDox!Ad<y~~00JNY0w4ea|06KFYRk+^eU~M>l$UcT&9+0deFAeW3R$~qZUk|P_Bu5=
z{TKZ<NTO<>^FT!T-l~4C|FvxR9|-)HK&@f38f|`SKM`Gr9BO&t?Zyd2R2BWcQoZHb
zi7#zBEQzjB1xsf<a`6vIH2wYj(}NP{?vbq0|F$If#!Uofj4r-8-xo227uCAn&}eu!
zl{NAT0w4eaAOHd&00JNY0w4eaAizI?aT87^{E#n=SFL_jac{j-<8Ml$>!*j>U7j)G
zzQkI!miU;OGe7TI{j?NFwZ@0=ARr<E-p|itWuMvS?&*!6mwsB<k$S)Rn3-+z&L6*v
z@Tx}p%Xzup?0H8bw#~2^1V8`;KmY_l00cn5JOaF&CkH$DI^R!idC2F5xm8046uDCS
z^`*ovwQBzmqv`0~w>i*t=Cz3C#C-n8t}}-fP9jcItIk0&E1h$zJN#Lq;yxm(n$>?~
zPAv!=>+?tQ&V8)wP6D-ugH`)q@SaX!55+G=2&5lgJ$K_Qi9{4Nl2=>{0w4eaAYeZM
zQP<zfc2_H(aOcdi--%kapE%zZsNGLgYq&b=tG4>Q<w7E=ivHeJ>pf5CwDaqT)ylWG
z!#Z9Fyx*ch+y@C)h|{%HtB<liUgt9L*B>%|vT1zIJrb|Vi=Sw<L-hXS<~8zeZgcQ5
zQL9?{m>oXrf#7b>jXFWBQu(>i>1uYlv5J3Pt2{*;6<;So!NCoFpts&AozEWVcdhAJ
zb)L7V>1gGN@E`yJAOHd&00JN&DuJ3k@+LN~Jx|iAqUINNg8+{N0=?%M{l4?GT8HBn
zog`3ju+tB$u5-gzRUcb+`UIj?)!G?o{yn~*Kd)8779u(awSJ&A+_NR8ZaF{cm;^PK
zgVu6!dw9>lvDwcLI8UJFaB#bSA^#wdOrUm_h`w)qeTFzqtt#FA#STASYw+)qUbV_j
zqyqvVpq_wL{WIz<LP!w!9|5b*tCG|CU&{C&2xupucmGx1uG72!f_7r=xZ<_9%WSKb
z?E>Ld?JfnY{r}me?Jf`Ay;rhY@7KALZED(Ml3q1>e(`&0KmTO^<cpHsdM{}5Cr!>I
zNup`(LKilSOn*@lRr^nQ>eRw|M<uJ(?)HUs`n>E{ZPX5<<NUl_Kc;NKm%ZDL5xvUV
zKSYV|eu0~n2c7v`GOD8F7j}XG2!H?xfPj+-RHz*MY2=OTlH1iuWs1CjfJ{Kt^{r;v
zEpvf_fSCkD*+<_@272Od(z4TgZ<opFg@Ob80QEp_2lTWhsa<pITKJMbuO25(*VOW^
z=eugFzAro||49OUz(Ef~Tmu4u2pFA587N+S4+3@&I3yk0?JImhQc~=*m-;??c9a>?
z1OX5L0T2KIy#z|^YTRk-^-BZ_4m|jQUO&clAOHd&00JNY0@?}eD^hya_MstO(e%X~
zGka#Z8sZgA(>?#mre$01N}}s;r*cMBIhMie?{U2J6Z+d__h+Ny25!CH?s7x=AOHd&
z00IvaC^Dtw@IniS1O*2peh|_qN2R$N7E91ab9lJ+;gdgsLgC*|nbBgm1U1G%@4k;}
zuk$ZAr+;4ggzxV!+8rlu)diQ|?^2}ypF}jBIdwtKx<Q8t)Lssv?PITZzbD?V8?(LB
z^83BxC0^9&_X*?=1V8`;KtK!vh4$_5lf{!DLBYX@AN2Y?*RLJZ@06h6z@s0qYK<|^
z81syIm}jik4_Wy=;sF5=0D(XRtlpRY$2GTREk2PzlvM3}Gp*-2zL%7t+`>|4iCR@U
zM<Ebbr+@!u=9h+4FMC{qYU7}H-*8c`bGqLne;@z?AmCI2qMlpp)bcPpU##YroIlZ{
z|0lWsl#FVxQ<Kww(QktUbteb2{T=ax00@8p2sn$tr-#~Io-yLS1bvvpgbFF;9MwCj
zeMGUa(6BUN!6T!ZM1_SU{~M7tB6~#6u%O{l5xJ8;mk0|E3vO_`^N6U3>7#2NkV2j1
zB_`&k_#2Chzgc=aaheXg-hRcU_*i1LnpiZdqQ0DGL1e28J4hf^6$t`?00@A9fCQQr
z&D3*7Xe_ZyWn_BqpILT4b$-~w?L<_)-f+mGC1qoY-71#@ln(;;=Jh?arTcHhX=;?N
z-g;s*-9Xoy{d}wxzigLX?YC08vVBkx009sH0T2KI5C8!X&_kg6obTrK8M2*-u77Mh
z+dJW>SR$%EQ}ED@duii{=vwvrUKMr?izTA!H~nrUt!)`gtWrI77T16P2t1mA-u`d-
zx<|*Z{Hy5qQ&#pXlM*9@QokX7w`P8WKWDR_m&EW+SP23k00JNY0w4eaAmBCvGY|hb
z@}t%}iQCm}<&OM<00@8p2!H?xfB*<Qnm}Uk+L<rz+eWnNqhrOdAOHd&00JPuH-V4q
z7pXe%`X-`RFZAe<=1if@1PTsT`hlqHQSG*y@74&`fdB~bMBuUFhk8D_c_Y!PJmmql
zfB*<QOrZIq?+Sfddjo-jgOz^puovQ!+X*xeE7`SuzBL3Y1_!JCb<4n#T{^wFortRP
z`XN#E_OKQSncmn&0;$#hU^KnoW3HS`s<V@*Rio+QdcFvREUi?xYq_0-cjb$k^R3A^
zrPgnyN5&DTGdMWkW5vqh{Ee&bR$CHBSShc$-%zh<l`Q4s2vlqiy!sdHN}WKbUfcRV
z-6oDeoyj5fQpdL-z=nYLK)ntn=WiosE31AiHo10hEc++qs55Hx4gv)SH~T@w?o&cP
zYP*9##pJ+C3;q!qJ7VIp9mHws-yg^eZ`c6>Aiy7i)+hTmN$k3gn5}l@{QV<+<74M+
zC0f-kdyy^(BopA}J&*U&$dm6_muypZ5?+*dUa$`YKtOx~_n#b?W^1J%BvJHA_S+X@
zAOBf$iq^hXqRr6a-w~r~mnT-w{k>=WZ$3#@)^@ZTy(UfTv^$CD%FA;YD?9R^NwazO
zpF4=CDtdn%>w1RxKtK-x*7aYH4I*E|>U<I2;<36+{%k&^M9k$4M071Ouh;w&bvF{L
z)ar6VdfEtF&a)u0RfZkJZq;V>e?zl+-uiFM_yz<(00ck)1VBJ^0%o6M$2xvqc-y`1
ziWd&B0|Y<-1XvSb_5O-@oH<|GXt}cXSEcyI*DSKGONAYTl`5r&-ybG0{MjleUkTn$
z+^W2dSNPlWO}|@7Yg@*WK>BcLDu3E@cJJm~GdB~xs+<BsxQRf)-ctsA=7}M0SG8qs
z^*TSbb|IV?1Xw*Mi4qU*rK0wHU*7Vuee<2HK4~nmOL^k~d)!5!!KDgP*dK93t-6a2
za_bfXbMk&Nc<c5UqE*`tOY=*}-`6Cw^wZCJbsm!~)N8i7rJRvZ5O4wk-uI1GYai?R
z>Q0Z(3BH?J9$2k|vO6BL`V(eneth`UEPqI>wMCtN2ljemB++&K2R{w!6FFGYtEu@U
zzO#-%yl279ceidLqN>(?Xaa@TO5E=DPAPq}+C#1JYUPUX?k3Q*Xr`VsLSu<mbvHHS
z9t1!D1V8`;+(F>u`bDY^yuOLJMFW+p)^QWN!~fR#rO#>?-9of#p#0-|XA=;8zbjVu
zt39t&lyGXtDG*MK^GZ=pe3p~ZyMxw#7^C5h+KcPWArPqF-01g8)*6i;*X!G_-)e43
zet_RW00cn5ZURQvH?zLpDjs(4yC-V;R+W!i*H0tg?p{CdU;csCsf+f#+I2y^h0{7e
zNNsql+EHH8Y1Dt)mGT8P5z%x=<gG6I@@*udYW^*6-OpFeM?}@fiXZCv<mQdU?RvWM
z#69EgY#~;w|NTg`&qn)iSzm8gdz`ym^Vc&}KO*s>A(>7+_3p%HylVBes~dWiC|%HN
zmUex*;gDwUJmpoZe;=Ng;e|~(y`t*P9mD!n%#z7#m9BVe+mHKCg?i1>{5Ph(J>hb&
zR}`%p`*pXsYG0T1s$J~}(secgrF9ri`CV`Sn^~o=_Px>WaAxHn@qz$P1gx%SJmsCG
zEk^gRx2r!e5CiXldL2s6-$q1N|9w#deGbHd_4kuQ$IqJc?FIs6!{OZuRob>$zn)m7
zto;G;fq)hQHxEA5aZ=9>#BCagQRy)w-~6e`Rw9Zf2Ctp@;=XM}R26MJRy2D9`FWu4
zdH4Gr%kF=(=98^NG!3hLecizuF~n*ORL)wzA9TI_ic9ga1p0u3{C=gCBf^6K2!H?x
zfB=sKL|?zjBU3FlU2iyK(UP*UL^SQxYg_-P+r$yxR115A1pyEclYr5Er&;aKfq2%W
zX`OZ_aXJU1{>J<D<)-zYm6&%YVXb1eF0dK|*brd7?ica7^Ze|^ANE>sG}B6=Ruhsk
zzBfH)4RMQ#^8WnCwzItxeu^bfaDX3JUk}*a)e(yl0p8zt?6yydAAWZx0nDdx$Gj3z
z_Tz<}S579?*-5~`8V-6}b-VA^O6lv(7p?;V5C8!X009sH0k;xZJ2l7ZDPb`LDi#N~
z`YZAc0;v(uyI&5jTM*eQ!wv!t;E<Xhq*mAPoqYs&neWM4zpZ^Gf|Nl31VF%<1eC^K
z@Ozrgv;W*dzyTbr^aHb78>j4}i+tp$Givk>0uJB+KlpDwFuNT^{2-u)z?t$VYo-ZV
zDv7GI%J#px?eoz@RPFmk+0h>aFD0y0smxz!{z8qWZ)}r9)$DIe*=m0=M}peJVN~N5
zN7FsEnpmYu{*2#200ck)1VF$I1h9@A>&UT=-1>FosCUk)cW$^>Adesb0w4ea><QFO
zSA5H;5jzRgT^!hZ5@K{B0oMDfs*TSqmX8%b)bq*B8;R(u_xvlfxEhTY_aA57e~kJA
zt_Oif5fHt-cvRlZ{O$6<_2m}t$?WA#y9|0~_?F(mUZbkS&a87WmD0uUAi#n^pmW^u
z{de>F4B1Y=;Q<F0zV`1p#!Wm?=VpV=1j+}8-#WkaS?!`*2oxOP2j<m-f46~Z{+}bm
z<$QnqzLmI5)v#d^W^(BGS#!SKK-8+yG@mJW=*GRYaYPi&viqs?!xnBQqAH%>Qh$y$
z>c?hXZ`Qd^X4$JX-rdr~*JiC*OUzQO_J|g~!KDgP*dK8OYA*-T{08=d01pJzj#q09
z$3xy=1HS~$4{Y{b-WvTRrzpRq-C^1BuP3$|wj@sCU8^i@xie(?Cz7Z-JJY7u%CGsA
zh^D0y>K&@lV?D7-XB?m1xYz70#HhNX_9B-c00JOjoPb&Vb#lDMEkq0;V4Oh6(n@u^
zmfK074>-s~y$9-bC^>%{F-zril$4i!*zK}Ii4T4U0T9qnATfCD%oq1<BU)8|intO4
z_$46v`R!G|?^R*fuvlW1T2+#wmy1*8xf$(;RC~QueuZ>E00cnb!s!M>E4{IV1k!0G
z7e{wqvYd#nHxEA5aZ=9>L{yy+yQ)v4-kXSMdZ9;;G-nEJCRS;n8a@2kDkonF-cCf<
z+rwHUWO`#8(W-&C;Cm1N0p}6eP_Jo~Eal^fUe!8Z(dzKdD>38<1VDf{0;ela+%xXZ
z7UC3b(K=z;?ULJw*~(iv!X6N?lfb8MXU!E|d@JEaS+jWmNZ<I_Ia`V7x}to<Tj%d>
zCA{m@A9p^JH*y<ct*rfm*Z7Le>oxyG-Hk-A@|rK$1p*)d0w4eaAYeU#K1p3m1i!zN
z*sa#*94UeT2!H?xfPirVR;@D+l#cNnAO<rDXkEAAAKpyc5H|>b00@A9^#m^GSrFMO
z!wzDVTAxFt2m&Ag0w4eaAOHd&@NWXV%-i|5O;-M&mwnW&%(I@fW*+`=<VUS{5>~3m
z23#W^fw0=w*B!hOL!72&oqMS@UhzBv7Q2~1#qLu=KWe*!*rje^pK(g9-%5{+BSy7T
zuWkLGZWBkK-~c~BJus^sa7(+4e40xjPrhSavQ61Z>{2;*tk)Td7N5*Px6a!1YK>kS
zrJz8nJK{%QnEy<g&9nd9LENH`reXY7v+ft;YZh78rNRzkmDVVe>D0=N8;I3v*LxD{
zbd7t`!!d}yFBNb0xBYTOT+>Edh^VS{e=k<yANERo0s#;J0sRE_PsmYc)aV^V6m`pU
za-e+bFKK69DT?1Z)lQ@R{Yuw6(`(t~%Ix=1c9yb@iI;gadhM9Man;>wOX7%Vs#ITD
z|E||-<ns(!t6qC}xz2x`on4=K;Q>305fH6@b)WOyygoy=6Rm2DQsApux0kHq`?!9Q
zsspcYBBE%#XTi;Pw{9Y$s@dmW1ops3|Dtt%AaCK#8V|B6UeTTp{q5EJe9Zc~M|(N@
zU#;G6pLKd#&s}CuF!dO%S~qM~I;{O#ZG5TI#J6@6_&w&z$)q|v2`_3l52WvG0<7;-
zX2+-XKFE7`tGvtXazXt4XC>y{Ndj@uTF%}B^*WTCzm15hrN@kX^QR_TiD=5pc@ui=
zV9i^r-<lPkw|)%Rb1P|W%UA*q;NV6-*gmsf$D%z~6W&zR{&s)6t$yyg-f+mGC1qoY
zQGKT1p&R$o#t~7~?D;EZ$1nQ4F4p6>R>v1;o@Ai!&3ax86fZCRdZO47h@;x?wT|l>
z4X@TNgfo{wmfcUCAGUBi5k;F8&D3*7Xe{Af^_H4h<7fWy2C~n89NMhUfzrqKAmBy<
zPrbA$tm3WB#O=ze?5lp?tHQ2fv4pj<vQH_VEdxt->GbAyBB~mFKR4@oCGM<#=Y2d=
z$v*kLhZPUe<7vL=yF#DV-aw$>z_TBiUC%nijJ-Gi!x&;zmFyI?Uv|*-_A4&M#}d(X
z@$E0mL><^ptX3tzu={uUzUZ%3+0}0Jnl!D`?j%rqI0TwEB+B>j8)JTBRXw*V9jp2q
zl=8!B{hiu)l<eX4_YL)$R>@L6jwq?<&mXYg8PAa@M`xEKuk9r4(nsLOMokwNy1J2A
zrTVPIRUiNYAOHd&00M3z@cxm$@v(EZ61y~PW1c^=dd^5r*N@IFT2r|94M`LYJ<;Ts
z1vRfoPSNNg%V)IQeM%Bl%hVpccHNh!BvCbYj``>Qu6|51Tiw)lBd^*C{8g=aXz@RH
zNusOtN>G6nJy%Pj>Xr6`iZ8qTIT2NNJeQC=^P*)$bj{!KMBQso{7OXAdXrYvjmW;1
zSf$!q8eHKd0%on-&>OGS>l;NcOWxLFiO!Q3`&X2Yc<cPVtpxgj125$QJIo;<>O5ZA
zZl}y&kn>}gn&Vx{vF6BdIo}_@ZzZCtsQWfqwa>R%>6pF$iP`b1-QQSmIOUp$5M~e%
zWuCcSI|J=M!`t_)-e*JWZd(89Z$6?{owEM0R6@N&HF~TktW|5BQ<_JuH9mv~0TBr-
z{;=1AqnTC`r|aZtPhE(7e+}VHXZL(5ZOc<WBC3vDv2#+{n(K*EHRkuT=dZoLk?^V)
zdh|$hrqE_0y4J~3@T<YQwh*&a#I_kWg8&GC00@8p2!H?xm`4EnV;I{X!|nX^_Io?!
zI{$r!@p(aHs|-7cKL^(H3e0N}&57A5{_eDV)J<1>%cv1M3G`tO=G20~v7W9xanHCr
zTL{!14m+0J|7Oi6TZ!8>u!rIoBLpgTpAz~}+Z{v{HIi3c3j!bj0w7>N0sr?<(LO89
z*Yf|}e&T#v5M3o=Z_wIxgtaamQSxa2r#28#bwcc_K8<>BBBE;L8x>Zzo4$!yt$ces
ztmB1%UHdm#oo=A@R#u;ztTjDe{6woAiNR}UzPN82af@o@V|I8}>mt;~<Nv(p)X!#@
z8>{%8vOWd*z&T%P&iT??Z|3CvWboGQF+{6sP0#J~F_iMJl_$c300@8p2!H?xfPkn3
zs*PxQsKDUwC9NuIeqlEV@JPVyeSi>uk#$`v>>%J^4+pF3oZfl$YVCB(xPn{rsn!p)
zhI86F0bcW?wOrgDe)zLhPQDVnoj}du;CBB){y`v_K+~d`dd>)qC01+qIp59eGh{n4
zORXy1K=X*yes7hXNCyN!Ks^Dg`X$s`gpeTcKLYzF<ft=h^bTT`{x4<x4+OLm5cRwS
zR(5MA=8h}W&L_4joZY4HPVa$JyC2EvwMzG9TAROf_xxU~_3Mf0H*9K<+pAZ5t(RiD
z70>C_t5q6~XjC^(cCXzU-}ON7n`6>>Mbp;hYc>5l?r%v{9rR-xPiWqAlGSQ=`@%Z?
zj8kg;R(fO{5nc73r^(xOM&A!u`-dp;{n)7K;zCz95~C_geqkpFfB*=900=mVK<P0f
z-~6e`R^oPbQkfzzARrSkYyN<!@ycAFAOHgV5vaQCwb2DGOd(KkfFGb9$n5~DmP%<~
z4X6Chir4jqLl!M58%v-MII!Z1ct9YTfZBP`fx>|z2=GLpW`~4a+0L&cN?M~#rc*07
zZXnQyIq+1LumuD_00ck)1e6H$N$Of6`2C#(3JyH@fs()CcMt#p5C8!X00DLcN+r}g
zRHMgwBAPy5{iSqo%wIx8(?Rdms#xRkpNZ(&v1X4?{+_d#_<N#N`wS?h<8PPUpIe{o
z+a$5;Is(<k!R~TH`XB%TAOHdn6L9Okt;jdqa}Rl2U(BmL57$0?@+WY3#+(*CUiwzD
zTj#t`x?!&lb%>}c&)+ml{maeipGDs{!l?aP&jaQqyw&^1BE8i2?XCKKuL`?{#S&3e
z@3|1XUB}z`#!kOaAb%hL0w4eaVi1V={p|T`?{6efa4_Nrc<xXAxgT8k;KHgkF8;>i
z;%}DTPN3i*f*-K*d&C0*AOHe^2v~g{VxV>D{@<(Z|I^xUxn*F<E}h=oPSom(@)2*H
zzqgeH(m-6D{(V^O>+25Qh#^pI9E|QC&inOF_j}|I1V8`;oJv5{{t>5^huQgJwf??t
zy5d_#jo3+y>hCdEPA1jaNuciJV79*_eh>fw5C8#Z5ttCWs!yZdn+Vj8Lxq%bj_MuN
zKB8Dyny|EC!6T!ZM1_SU{~M7tB6~#6u%O{l5xJ8;mk0|E3vO_`^N6U3>7#2NkV2j1
zB_`&kS?sj?Nd-0&r|G<0Rofg{vYD8zCKipVsPDgjzRc-}%_NYjiUff`00cllKms?y
zW)FV)V?yjw8QH5nnz!yYcRA6j2T#}hz1#d>iKu#`=bmmYe5;AwDwhM44+85)-Z=T{
zP#<xc8m0T*l(F5L{T4&Cs?l@<UEl1_<-N1M97F6<*{;u8cIY^%@@B%D%JxA)00ck)
z1V8`;KmY_lKo5aUH(q+FZLb(2x^B$Zqr{{Y8;PizxGZO-85d#*)Lsr&k2~q9v$zHX
zK;Y2??hOr{k#<ZB(W{S+UHR9eXIo9U-E<>yyDF0sBgALj`X5};GKN4o<6v?>tr*@3
zD?tDRKmY_l00ck)1l&fz?0Lmz$DjE9J>UPackXdtjQRiXd=OFyIfO{$u&tG_%xBYv
zZig-8v{<L@wyh3SqR6QdC2>1Nv=zzOc96n#tM0qgCY@8HBS}dIl2Sh&c7C?`HqBf!
z*EMrp^LhSrzwVj2uJ`r6p4a=j=9+8fGt*<*WD!^Mt=*|F1Rwwb2tWV=5P-n$32cA&
z+;Z2fpD1K?_f|Oy0SG_<0uV5pK%-`RwXV2FiHNGTZ>rLxNApsF!Xe2xIQUNs2Q;rV
zS)g(p%pQ&K4gwG`i9pL^YM$3)$^;>+Ce;C7AOL}#37mN6=@mYj^o~H`kYpU}9EEJ<
zJAwTUF8ZeXK|-K*a7a2{+rKenX@|w7!d6|j;OxT}H!Kyl>c#8-^VRV;lnOshIv(Pd
z=l32;zx{E`BfnV$bdMv;N{@fFxcbk(+m5MbQOmwJ2VcCtdSU%Z0?h`8>_@C!H5`6=
zVY7oamkL#CYVC*DZ!vT0oJj(;n}ezGg<rcSz<pr3`@pvG+I6e57XoSsB=x>CRqqd0
zQ|tV-!u-tc`5En}ru}5!^G^C`@5k50&6E*5?(eis&F>v&OyP|m5P*Pr1Zv-2=f}bS
zED^djt(upvJZ4JnFO!9=rsXf?Lf{tyy5Cn5*1qYjQ965{OX{9DMVO+d^TIy}K)`+i
z=YKZv-`$^nPuQZievi}7?}tasd;R!j6GU81D-s`h{;+LZzIe=WRka@_Ra=edT2v~m
zSIfW4;{Be?sYia^ZE8uWuvPoq)&IgdT}y?ns=A+H4+2pL?6`LBn5oUC2vr)z2b=Tw
zJ%4Ze{-oPP)$;{u+wtCa(Y>8?J5in-fesIK+T3WsWRbSYdHsJ=tBT|QYneR|fB*y_
z009WtPC(c5V!G^_!aM5sJyqp1g#&&-00IzDO(5)ZE~GpCuZ#aZS-?S#gX(w<vgh|+
zo2dJP%H#OX1X`YU><5Ft9xw7$mu~B@$3dMZ3RCss@++=s-lIhL>CUYgye0b{S#SE_
z5=fCx1mZp~l$HEp+x*~@CLDdjil5(`pOA+2i*$An$hY~d>8%p>c@xv>u}{gotG++;
zX_1zi#sh!yi$LEtcYZSZib*1@<`*4mn=b^GuD<e#trwRFSsiiIq78?=veD6{`@Z+*
zH)n6CnA5HK($3Tq0$Cu?_nh<Bb=`l8uw9e7{?AG|S%}?T)!~wHy&jsgZL+Xkbv+l{
z^pb=7A3A)NqgtEa{^LCdFC607t{=?&q5tDOJ`hoLS24-H<OulP*Rf^)C7+M#Sp83v
zL|m06UUI?%14l2IC`{F!Cm%ZIqvk@St+HH6pWg&tZvXm=9j=}zWHrC3QF{nL00Izz
zKz<O&x9@cMwd%$DZ=KzG*hFEAdf%TDcH6K1?9VoVtgIi$t%vRFvtIw>@*sb*&vB;B
z+tlv^nC@5Hv%^oHbzbxIDHB9e8rNU)>ml&M==JYhcIIfIT7$f`^}Ff#E%)8~pg7<-
z1Rwx`v<ZZLUOTCDN%2VgeUP@6pH%zQzNt!&9?eSy3Wt1-1MfPMt=qOfU#x4LD5-JO
zlyW9?TGHY8>pm2=X`^l}ORsJ}Uf8NjntrtSk?qC8R^9$Y?E~w+UMlikWxr3TD!uK$
zkJmMRN41;Qczkbn=Yvb0`phv!-#hx6`llR!bS|tuQho4QM>jYkr%TJ+)%bxa!>i}Q
z>Vq}r*4+E}iaA?#<F=zOzkUDDouu^PPv^ZN&U)L?rGH<2_D93|4iL8Jwte3#SpRIX
zh^lFgBb1wM0*$x7^;o}qCyT6W)bH@<YJFYru~uzYSN*XI0h0(so!>{zXHwl&`V#j&
z^vg!>-)P9%Ng`?WW6-{MY^#1{N`>tj_P&ZA2i5y)_iog#_>qYMRl}j(idDI(O(%+^
zRCPRH4+1g-Cbsx@ul0va5c#Hlj85HWa^=^DOcJ)Js_zJ<W#^LV&+k}&*w62sLA7x3
zi*vvI56`dgbnANKg{f-mJZD(ZqJFRaYrIHW{n}Y}f9dA4&bsE-Ndn!OgDgkVLjVF0
zfB*zcB%tc~m$dA-O#A#xG_GDeQOIgZ)m9_A7L|%1)%9FeQok2aU5=^OZ*DxG5P(2-
z2^{>Vg#(&b`uTfSd7`ClN8EqW(x0!#E!Uv)tf|N2zGv3-{JQ#4c1Iv;27zD#ryluv
zx2Ywi0)<1IIC#Hw*6^akw>e?e8T-2~PtDlj<ct3Kuz3wjn#^@f*Za4xne~r_6CD%|
z#6j}?Aoz?;3Kaxw{r!DXeyd=N4f_aeAJ(vMZbpefv(16`0Lr}w5MKGgE9!_H^>>I#
z<%?P`@<9Ls5P$##AOL}UCD8WxKkuKLUMx_%IOJ=*Qf~<Cih%0pY_WIeC!?>JB;b(W
z9Cp>$*q0uGq~9~+SDvolWkj``UW-sN1Rwx`Y!dJr|Jk3;d*9#=_g~&OsC%@r<-JV!
zY1ezSXfN8!r1!q*9!IecfiMK#U4QfPe++v|*s8XEf6Db+_WdP6>4Vz5-MCd}pStQL
zVXJN~c<J6ly9t4&!y)PSdk4ic#~}a#2tWV=`9L7*eM<82y;Qyz%lC2_2Ywv~^5I-T
z9U%Y#2tYtR0aM?Hj9+@cH>5t2u$V;xQNMQ@H=iz+QNPD-TE5}ucOTd2pi&{Ly12%*
z+y0FqOFJz7`MayO>T`YmdC7Z|gzE2cCvC6q6*t`Y%&C(_T#XwK<cGlS2*i!gwDa$-
zZaR<E*!a&q>VCV~F-`Y=v*whp-+bc4)gX3lYo9pi$X7oUsx_!wjzd5Nf#IhYHalqZ
z&+|SM4!em16|wfeHi$Q~`W^3`%#Zg@($49&^Z$;4Vg6rMdi<-!)hCI3(=cpQq$yc)
z>F8Cb4i{lHZke_{XHatRE%l!nT_S8%+juACZ`?T6m0#ETq%MDD?QU*2r_JTX?+aZj
z>yK^exo$Ayy1}-%!CwfNKp^gYY`^>_)D0iZC2-y?t*&Y_e3ZzFnoBxAJUit2*5|hG
zP~@1dC$`$N_T%@zD{R$vD^}&EHk~MJ)1=oKTeNF?=++j+BCh5~@1>RyfB*#I6KK5s
zt;hP^J6Y(`_#R>b0`UnX_5D0s%5jPI>G$%4^#e<UE_KTZ(tLl9VGtjVLjVHN2@J2_
zV&>L4lZ32BFOi%OFqeR?`F>J%OIIE<CHI%fA}LL(B@bLae9XnC6$@Ln<<*zH{Mkt#
z3fuMJ5%XR@e%S<}Ta$__%7Fj`AW&;+zbl^=4+uY9-ulZ6=brzXuw8#z*Y~>7myH*;
z>h@s``{rhp2;213BR}spwWL%erG7oSWkK%b;=RTT+x6j+wMRCu@bf%GQvJBF9|91_
z9)Y;;FYtRV(Ukq!YcXmB0SK5zVCl9FdmPkxqR5I~Tz<th&3lvx-D+Ap;tvGUB%tf}
zfl=+MW?}2^iM@W;zP-+mga26~Ojlj&K&s=yRC^^=TaD;iR4SsXsrAAy2tWV=5P$##
zk|*#=t<(Ry_}`O7+M0aLDG>q?fB*y_5TC%XsfVwtP^DO;rSWTk1)T)qzBf##Z`g(a
z1Rwx`<O$S%_JS3ky){83rODTj5+MKq2tWV=5P$##{+EC$@7wubpOXCFl<#6ES?4Ha
zEgQVOxW}}~LX}4GfjssR_;$|ea~HfZ$H|&*sylA`-TRLcX{&t^fyenwAgTA;_?43{
z>>ucQN1vv%$BDSw;h|2O8x5E&P&g0=^aEYvK)#I2)Kf13RoACc?YOmT)NHTT7569+
z`Km65s5)f%8y8&o-VDc2quLG17q#7O%@@QY?(x2D?)+r*6_bQ&P5S#P_fMGnx3O1^
z5lO49apm<pZX9{@NBtcl`F8y~PTSwtkktFqyybY~l5Ge;00L17B=tUSRpsbf53t2<
z)OOBlzPRtvBmWIIK6C115tYOC`^yih`0ym5OS2lcs_LbC+*0MIYCEdVCu!L+<vAeJ
z{fHVjy555t)gM#x>H2P*tX)%h;72?HQRfYQ`QmvKoV(qunbTf7vP8&gu;pke4to8@
zE4I!WFHF<6RSsxSvG;gki?%$b=6OA)Oc1uJ?)Hxson^}CdXHgJc9)eN|7vmdNy1jO
z{r*AH_s#suiyHq?^QHZMskr5P@23TAN981&FK)fk&hPh}sP6r;oJpSyfuMes^zmeB
zO)2?Gz?9$r<*Oe_#hvYZb)S!Q+m~(Um)pPoVu!0I3S0H3&pNMp`jiR6I-Pdy?Opes
zD1Ne3Yxi!{uK1COLY2zKt*-Q!d_JmU^*>D#C>)3b`a#;`z~FM#j&8W>17VuBFCN>f
zUzt*2yDl5Kf1@F5e_kg}NDuzg!U4@IO%`$0_IY;F>l>7}^LtfvUx(<nFZ=wCuKA*$
z-Al&xdT7qJ$pQ_>!H;9u{Yl^F3+or@^b**%?|TL7pDh-)Xj1RLjayFCRvTWw#mud9
zCJEi@=b!F!UHg9JvmXNaNZ{rBCf4rPeS*k$Rn_*V?lZaa>q909)vC(BAUmfFy63H@
zM-~fPwd8~c2991ZQJAJd+*KWq`yH3G^KXBm_JMU@FBPhI*ly>%TUuS!X80(9!olP?
z(B02eJtrRI=fM}RuU=SxlCVwF9=BDM@Ao{g?)^b=k@j)F`y5I6mG*PE`CdNvL3X|m
zl5aec>gP$7lhpY&s6MCu;rsPpepw=t(jb3KJ??k@@$Ogs9>28hdEe*7_UGNEmXr!O
zz#+{zFm;^7uV@73ciGzYviC-aq%@kB<bnVMAOHafKp>w8=w5#gvcG=)88aK*_KA~q
zZ8q<Yt1lY6z_CToxNYjTtq*?SWJN!yyXE0`&l>L7su#Ynvi#WG8;-4d!NX1d)Afvd
z9Nn5vV>fk`C$RF^lLl4)?tRC09oOUYqsA^AC2ZA+E&knW{UH;Ct-9p!L7OffJ6_nX
zy5IMw%YM?|ZIK^o$dM%iroEq9_wO0K_HA#=+Ba3{(W7~(Fhy<WX^Q_z|IX6%@=R$L
z{Ln+d_U9AYzcFNKhs8g?r^9kh`}rL`(G_g9bn{tfU32RsVXOMR@89(O-uakq+etka
zsjD2_<5h6aBZUqEao>AndVcS7tfp?uzCXt={iT0xa^hE~y&=Ntf=j0NKcvejVY}XY
z(%<T|sXs!fR@pum^!`v;d!&bejRa14<mNNFKK7x=x-Q+jclSn5PY|Z*#rtoa-FnzW
zVXKxLd`tajMwf`Js_E+(?Vs!O&r9B$By3mR^HUqgW_*SK1Rwwb2tWV=5YR`!wC85<
z=lLD$5BvE&;M48U4i9zO+-SgLk+15$-#0B){k~6K_Ef#6%eA9#MAT!s&t3g5oYS>b
zWKFl<Uw%l%hbIYJHSO>9#4TUf`^oh50)MNfuQz7hzpSof%ht<3LOBwLfNksaaq3E5
z2tWV=5J;au-*e7i*LD9X!d8`CS4p3^*&ehk$emog*La~?A09F9_2ZXK5VoqS?{cSQ
zr_aAc<Lbo|Mbc_^wBwyA1bS6EXYZp=xnE>me>$>n`}q&{6J9m__eiG1iL4)e`n^11
z{lF4otIB>4T$YdS^tyj{<h6fs`4!hR?@=PMuDaVzmHnjW8<dyMJP+=_a*x_?qJHN>
zwmkR#Fk0B`pv|Si%R!bW=^+3C2tWV=5P*QK1jY^DtNM3`jTEwKYklz>0wxmB^&F7t
z?}x&_8*Y5&)X5_29Fp#HQJ+f<>!+^wbz#?x@4bhO1KD)4&#jO?+V7}j+a=%A|8DiY
zS1lX;fk5?e$oKf7{t)<uKw8hclvHgsqH9s9(4|SW?!hNbIQoPYlSNcb%1_FH00hDl
zNb34Oyho&jz;6V0TswEn)MiseQu<rT{09Mf0;RL}xuouiQ-tjrwZF*|%a0to-s7Um
zZrZKTs>yvN?T<J#mz1`A=Fn%)J>{re()#E9THjRZjstR0_0faYmp<Kc-&|DP=a%||
zrcSP$OIvrG`Oue7t@zrpP0Rf0j-_AL9_!eu4ffdh+Pe$h5=m>?<BRI@(^^kbU9PJ0
zN7{Dce&0!TeAr^I*O?!Fc6Zm;ow#aCeen|l5P$##Adn>jQGd^uwR~A>Q|bZ%mw>JB
z^VQ|I%LRpiP68LN|Ib&)-%u({SDg%^*nZ^4L#w>e@pXa1fjFQaxZ^;Sv1C_ymJQxs
z++*5gk#*fw-J|UDzIP<wx1%s54+Q)O^lfwJC!?>JB%-RHUG_sD4FV5s{r!D||4}T0
zH0$4Mk6OPp+Ke(G009U<00L17==y#gc2EE7;(t#TaFF8=HIB&#0SG_<0uX?JJb~?Z
zH+=N<TVEEo>FHnnu;i~jpK@%|3GMg3cFwtjh3)!NY2O>`ZXGJT5x1kmniEzPP8Mma
z*RQm<FW!Ia?AF633RI3m+U-X95P$##Ah0um;q_b0+&X8HK;aN44zm7Sm#X^h-22$_
z5*XLx^P|Qt93|4$QwH7h*3%=4g{}H<$=V~ESNQpR+M3kcZdu>{X<gszMqf5w<f|?@
ze9)$g$Bq}aX`j3LUpS|0sX)`^p!z*1y6x%u9bCTK&H8zQ`au8!5P*Ok1THSW;+p0?
zN(2grIB~FS<o=C@teqrKIG7j*sz&2{zkZ?q<%d*!c#^P%gQ~d44g??o0Y3tld_JmU
z^*>D#Nok{QElaO%KVAfB*!vYhcD%<Q`s(1ldYt>A2&+NA%k|@$_5I6AkAJne`p?%5
zs1%2+|1K-*aZmjq009VOl|b#As`Tj5yj0kxS#1y9^}1wSuZQMrn=In0s^8n_`hGEX
z^?n~(6<^))j(rF~00I!m7J+svR^_HPohZ<qIXrRlf7`jJW6|}6^$IE$R4ORjtEgpB
zLAhW47FI8;S-5{enI1)j2mG>Kzo2YE*$cL|?paj$LhmLEobuU<5{<Q4vb4^QveQ;N
zuI}66gdQFHuXbErT(#qfk((>zqUyKv-oEgKx!*fkRgF!9X$U|70uX=z1Rwwb2tWV=
z5P$##AOHafKmY;|fB*y_009U<00NmN&}HVG-><#0at_@!hfFs<m4g5TAOHafKmY;}
z2=uRi!KtG6C?~E)C_2?FRa?~AGH3W(!cS|>{B*?d#nT<zbbZgNU!2x$fs-{&HCQMd
z0uX>e9uj!1Wz9SKRaokva7Z!^^02E>We7k30uX=z1RxNJK-6`gNQIFJ0uX=z1Rwwb
z2tWV=5ZEn&Z|Yt7PKA?J3R(U8**8qPYs7lTtBam!^wgdA{@?^@r+>Wm@1bipIkxJe
z!7sLJ(|d!HuX@R!J5Bg%+&3bu9x=N0nYVxUl`utR<6Snr>UvhMad#`H>3o6O&0)8L
zF$eR8!1NZ=e)r~O(}k-oSC$lwY&%VO^^CLjtvR>aU=da;&E44Pq6Hs2vKn=rE~$LC
zY~1_4Kiss?Nn7(KK&dDMAn^YQbgo^Yx%2ou$2R@{6|n4|6Z?*Q?vSaQowW3eH{a`7
zH0(!5R+Gl;_q@2f|Lf9M9-OiCL*eSy@sF1tyK1rU>e*v&?)!M{kA<ur{=>|!Bln!@
zpk{GMIu6q+PxiXpwDeK0=e@jDzrVZg-^H97&uw**()Uj&J)mI6PmZe(JwK-Zg2r2%
zr1aH0|9(P~hGlX=+GX)qA1%JVY%c5SjjyD)-Sd}jRdxNU>UUX5bEZ59KmY;|NS1)#
z`NI9qi&y>5Q|@<QT=nMj+xD9;zJJcp$;VgU;Mk^q^|WQb?Z0PJeZMm)d(Fz6QSHmi
z);YGTZT(f1zsZi;lm-C^KmY;|fB*#i2}E7j&w4(8VjM|>fV*yh?)+Wdtn)}C)+iGK
z5XdhAXFvS?g#Y|!ri1><!It%C{H(L1?6j4R=^XNFG@-T-*gb)h4w+wc;7wbEta`1s
zeXe?gn)@tq(0ClYT-i38zyt4|cj4VFD;KAwX447p-QLcTAANM$dM9Z;aX{Tpd(B$!
z=+f{6NC^Q5s3zb(AD^@)^&GxzIeyQPw5tAUwZGrG$w9+$P#ve(g8&2|009U<00OcE
zo}YHheHZnfDNs1%Qyj?l6I0Xg*Lv!|2KK95tV&I-Cw@Ty0uV@xfcu?X+kZb8*MHso
zzg_?2ao7D=tvIZ@^$F|Oj$@B&IJN6Yk(Ac|w*GzM&Edj!9X`Cq$8TS;$dT2s?Ls;T
zKmY;|fB*y_U<!eze_4Lp`EUOuY*+7lNZH=?`&murVoJZj4+ua20uWF`U|XLld%yJj
zCV}eVpr)0~v(ohHmgk@O?q(6B=6Q}s5Ku>;^~rA?HFfF)k(MrfVN~(N5hVgmhJ(7u
z!4d=@009X6F9F%#b>~a^j$<!5c#mew1*(_B*NdEgjjZ#YaP@y<Gl~C$zN>A@@uw@)
z|LTsHS2<y|S?z<~o7{T2<EKfmwW$o#?g!Y!&IVuI=lWEu`&^ps>i7K&+xL6FcNMih
zy4u(8J0xNEyYJOZ>w61s`L=6V-lnBfJ)Q;Gvtg{q$NVMW_nn&q``y~<pOe;#eA6d}
zb{X6Bs<|TH)UOBVem>u?T=qi%0{@r5<NLQ7e{S(|;p(3EU3Jab!&eGZ_5ap@JrIBZ
z1R$_00=>TZ+mB~H_N74KU{W0HDpuK-9RihxU9??H{9d4NAP(pU?l@rD2)AO#jqA3Q
z`AVR0AP(&52d0f9__KQgjqiAJzhR$#E>OEU?7o#b3IPZ}00OoXm~~~>D&sGjC-Plw
zZ%6!x00bZa0SG`KI|RD-|JR*w)?V)f>7230ox1tCuN>30|C!4!8aio)gYKF`b|Tlz
znylD!&ZN~<mJ3<Eds3U<pM2xj0@cI8%xDSk)wrk5Z}?c@5`n@Yi*XRXCy>%a0;WC}
zY9cFqN}s^tKg{epa?hzwSgka7W2cK2eC*hw=@Z6-ED`X&f5EivZ|Yt7PKA?J3f1cV
zo-Hf-cKge2?sfVKfx^LrILOi%NnPwA@O9}c56)Qnp+Mn~g*dq4`RV1a-|Ks^<${a4
zSHF5`WaWL>$-EC6&T!`a^X9#OKWH5@?)XanU(ZkW%oNk{)cbyJG3ByVrw-mCTpiT4
z<e=3jZ4~z&KJnI9M{N*Jjhk1DYPZ@~C!O3g_s$Uu1quh^ApL%DPpe0cJ88$K!W1oN
ze9ddc1HTYi(}(Z6f5T}fd?HXgIHYs7rBn#mNuc8D2Os%EtwkcL+F5J7hJYRd-<3P%
zW^uuo|I?qca{t%Vdgy6QtU>?+5P$##AOHaf<O>1Qz5|FqkCz_1YO&D8A@1+u`kb(+
z_ng^33f<~{9~IsCd$I6<$Nzpm*p_(6m-DIV^*p6)vjw*nRLhyJGnO6qpS~ZI&6%p-
z?SK5@Qw~_|*sgaBuh-;^6IVI5=${rnfBf(3ee3w?qQNh=Ytws!W1G6;#)<d*d18a{
zKRK?}{kxb`<GHO43J2nVe!w`e`*FY>M^Oks00Kb-4te^>3%ZLhoFEOd!*K{e00Izz
zKo$w`9+cg`2j$XhFYY*P+A@LK$>H_Jf0+FGrZtXj+N{hO)xNxJorA)GIG`V-Jq~1X
z?n9j*009U<00Izz00iO_Sl_ei7pHYw;K*wHlCb~*2tWV=5P$##AOL~i2#gwh{oyBc
z{7InUIQ&MA{~!=dK=u9c>NW0e<usiyY*X9cv-rf&E@PWsHCNcGwtvsU_U|^o`^o5b
zGrAW$s2qpKx87c(^;_!%x_b`6XChMkLg0Ymf2nuj{G|eggG?OIdmvyVfmv5}tup?i
zdE$ae_Y8fn@d7j@4$ofHWXF_u-geT~u=7vb=iB(d?4qHQW;nKSuyKsVX9z$50uX=z
z1Rwwb2tWV=5P$##AOHafKmY;|P*0$0i#l8841Y^R)&BJ_I92o><)o!i{Wc}vYb|Tu
z(XYZ%Cv81p{O!jUZ!c49n$F&1_6vXgezVAm_Bp-z^(Bio3Ee6?QtaE|gdQFHuXbcr
z_V<{6>C=jTKYvWyUu)*4BZe=Y?nG5p<FjmBksbnJ2;6)4#9Ln-wLze8_}@5~G41kw
zN=L2`u6A2^{jlLDyzgXH2i<o_m0=A&cT}rs=d*nEC(Ky*OBz+@Yg_Du<q|jD^cK^8
z_vU5OMbhf^`<r@KzEk0(l|r|A%Viq^5P$##AOHdT37E2uZJQs>Y9I98<kribq_tpe
z+ed~^d(|;T!~ULybaKDDHl3IKBOf(@Kz0bUeXe?gn)@tqRH^M`Y@1JH&EFFT)a|s_
zto4qqYTI{tZ1cCujkVeq+_ubdHRyYoe#dX@RQra?%a%L3b;;5?JIYR5>Bwr_akS#<
z2Os%EtwjROHix)zl6L;2--oCAeQIg*qkITJ00Izz00bZa0XYI+ebcB`gRj;)C>%_V
z1G%WypSt{2e}A%Z*RMoaJ+R-co&Gs#t;iQ$S+mz)hgW{aF-=Q{-f-@?o~wjz)gOV;
zN;vJtvfDm<ey(GhMyn0Ek|OYjX75#-S~%W8;XoYN*AM25J?_-a&wb^f>2gSFyrvvG
z2z;`j-7)KCtr4o#j@siT1Rwwb2tWV=5P$##AOHafKmY;|FrPrPeMT=>e^Z%aU1~l7
zJcIxQAOHaf*iYcPV;gSm{`B`wQffajY5)NUKmY;|fB*y_009U<00IzzfS16Y_g!_(
z*~3?gu<xPt=4TrM5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb
z2tXhi0<X2Kc}KqrOC2;Chh$obQcNN6>_ts>OnK*R2Q{06De;0I5P$##AOHafKmY;|
zfPi`eooiQU?mRxv(XHw|$07tE0D)`}IO&l2MF-xrMY#Ihk+*!(Ve!`@Exq*Gi#tx6
zwoItj#&<lq->^?V7rOPpd$)CX_LXwQep=A@n%9a4ej(6oa+opg@_kB2t`NGl*EfIr
z@yy4*6et{a6$jZkJ5dV=KmY;|NQyxJGnZX7bkYpR)ukhwe=>Z%5OKB2jkVeq+_ua?
z;b3AMB-Ott2Lcd)00bbAB!N|zzkkxLEx#2TZauqBpDr8zzXq*CCg~$(K>z{}fB*z+
zBT)D6Vor_cwmLy-TfJ2KJ7?^1r*3}kD@V1SJ?P<2-`?X>v3308rN^#X?5ne`>{?~~
zMe{^f)NP*^W=(9rXOEv8+w`Qe8^*NCRmeqExBe&uAOHaf<OhLCC-=;~bHqY{!htyW
ztsmq^@28d!&`ZF59va>G`S7Sb;<nZqXYE^aZneR})wuS^4*>{300Izz00bZa0SG_<
z0uX=z1Rwwb2tWV=5P$##AOHafKmY=}Cvd@}dxpN(c!AKR*FO4co!7UP&AIx}^JDri
zXuQQSMfdG+LXVF9S361R?puSS5J-oBuKtf*z2j>-(L$*ZNS8p>7In7F8UB_)v&~`9
zfm_d>JEKVCo2J{lC_OHLrvLcxs1rY2FLdi?4JWs5ebgo~?y2(|K32Ho|7-Occeiqy
z&KIusuYbX*qW35#YkKJHv-*`iyt6P><HiyBApn8w5MaKV?fEM08?Ak_)5EC=1Rwwb
zvj}Xw{HC_A6mA!0d4)F+fB*y_009U<00Izz!0rfK^pBfw`1|}FLbp~Lu;l0&?`;#Z
zx;x)F1_1~_00Izz00bZa0SG_<0uX?J-2~=!DgNQz9#e!@Yd5~_u;#bU7Smfy``w$D
zO*dAf(w{Ny@_kB2t`Mea)cTMQ0uX=z1fmhxyLIRL?tP$4u~!!#bw$6<hx{n~^!7?;
z-QU08x5BFnH(apb<7G>PpSs^sOIn-NKIpy4t(QAd^|(5d&K9SaI#KoAPe!+!(Y@G7
zTMO2<ePsBwR~_3lDtA-zb*^2Zx%2ouCuxl~Dv(Pzfyv*V|5lTazY}rwisz@7zkaXp
zMOu34wHJ4sHf@=Rs=7I29|91700bZa0SG_<0uX=z1Rwwb2xOJOi393(+H2N&$26TY
z_PA3wKlhd6)vvy3RI9;PYaP?I?R^dJ{9TzfjxGAnv)@_$%m+U>akb4~>P>j4bc2K1
z$sw!b9QA?#1neWws6xx>eOfFNX{&wY@E8IRfWXcK;=WJznlr9AOdK=Rv0bA+hhR#+
zSyy(gGXA1@B45-L?yes@jy<m7)UG3iZTfZTD-X_C`k_#zJC81GK>z{}fB*#a5a@LE
znF~j(_|eH1)k6`h5P$##AOHcg2&DZTr>HcxTyRnM>Q^rn=&u}>EUmMn?6j4Rsrvpo
zLnj|!eS?F-A<Z}l8gCxodV7u5Z><waYmi^w<BdlSzwwwKHabE2#d{5|?eXR&$4^av
zZs<>oo<IKg^}cm<>7v0ewrkUSgQH8+;=8wV<VPP}w%*Biy{^%q+3g2?>!9gyc)8NC
zHL8r;=mcrdI2QJ}cbsG!0uX=z1Rwwb2xOhW^V4p*@1ou_MOgiMk@K&Sb>0)U>&lwF
z{yMz!GY-0Y4q1<X>IVS`q)*_9p<Tu{y=ty__^$gmoOZ$|s1FDCcQ&f;W4Z5<KI5!?
zYtF4USlFuPeDLvqhUaDoRE~qH_lQ1wQIj20-g(<WchA8calb2f%FW_}F(RtE<7QMX
zf7RcgY~1xLfx;n)IB?@0g#ZK~po&1+^VLlUHlBUap=FE1>V=2aUD9WJg<?P5)@RDz
zFFn6Wgw>bMyZ&F@_FN^xYS4FaZr}3JD}#D{FHksyi37JE-`M7z?!Q}7rueo&o&Qn4
z(K?}8-)?`|&Am=vAyn%<tsXh<q#d6MKXtE<Q8T|grK(-rI7T4=0SM$ffd^-IJ*?X;
z<%(@l)$P@+%o)|bylkBlS8L7ubj0w*)1B8^*1V%%g{7g@&4>QV&sGk8znf~gvZQEa
z+iAj2tG1}KWzO)ogl-MnZ{BP7L$}Y*C|m66x@w&kHQA#~aga9s%ktaKfBPq4oBFkj
z?EdOC?r!BYoiEZ>Tl)`w+dcl}OZ$Z#M@bitz!N9`H-AybqU#Im6;vvyTu`=GQOlx&
za=-j7tX^2NaQ}ibJ&Fns_+`6(LD_<`7i?|av#9We-c1%b<;y&G=ZeN<_ITXUt-pJx
z`n>YzED`al`&;fxAHLRQ?3?qQq_o~`Yw8uuEuYJpPU*7tx;3ZoaCGZ$Etk%JZv6Ra
zI$o(ORkr`CD><4a)%`Y_7xw0AIB;d>e;n|o6ISi5DW0n!AUi&&PVd$sYQKN8${F3R
zTr%0QRaNthYR_#4)AipCAMJ4Gg>!{htE?{DcHyBDMOHLwe=t2?-1zdFrv>egTdrSz
zZyYAQ?fP}_c^l_7c>G%tReN1}*=<w5SSeJi>z^vsHO{N@4?DXj5Vn8p-ikd(qsFT}
zHPqv2RBo#B>8VxvRu^s`_{oslmpV~3{Tl4-!McBrUC_0`ddF66R;^s;CeyxmvZgz?
z(r;Ty%};+TG1`A&uOp&aFH<IHzN+hZRDa!ks@l)BC#zxS52mMUcHmP-SHEGS2vSw&
zGu3vYo<B^>7j~VLbib}E%l6l>=~cyN*z!mR0h0(sUAJ}f#jQ(=CXaSH@x@BHd{;k?
zrk%Hf{4st0<LBSvjnD7XzvtH?tEy{W@7nc?-?00G;)>(DUq1=US$Ddy*RN62>Gt2x
z{;gl0-f_e4zZYq#>iZ)}+w&WDliKgsuDbR|y&snE`TXKiS34zT_s1dUE@)k`Y|c;p
z{IKQv-_~_-*Y8^wzBuvdXSX<U)vq1B`&F%HCS~W~KVEapwOcnjvg)n3-}a_U>izAx
z!P|vyO^TPUa#Wo!{n{a^{i@d8up{>zqPuQmyJqEj(3bPN>*s@KoUmWRch)$ztEzD$
zsqvC>%pl-*ePa54Rr7|l?L=KSGA&=+^RMpS;9eKWs-ItR>zj6dRp$%cc1(+NRsQI@
zo^tK_#jPp(T|afHDfKkP4^?rSmK|MjlC<5luJdHqxxDFR&zq*ExBdCxfy2++w`!Yh
zj;$*D9t8J(ysNtAP40R?Ry3{ij_vCxVf{~wZ`^XzI{(D=TQ+~v=k>J8ljY)<KI(dp
zpS`&IRgd$2{scWA=a(;Nzp3*dS--M!KJ=?c(E5ej&hGpJ{q;R#K^&rvCqZ`O9`~;I
z#<k~{KdtNJxa9`*hp@-pev`HajqkF@qt-|D`OFl1LGj@73QFhKkK*n(g_CW5xb5cZ
zwx_lq{6lf&Vn2Op%jmls9#O8?Pp9;_ug1HxwhBLu`W$l5?>^%m59?Rda^l9fs{BFc
z4L1&=YE(Yne7f2rYQ0qFi#l&ol}~kj(z54`8`ayYc<cV1psx4nrq!-t%k|Fl-R+=p
z&h+DHt>2kiZrFIxl};BIS$|c{1LE2VJI@W<&%7LD?;q((@AtbG)Av96@PO`D+)^&5
zN_C&V{OqebuX*DuscqSHSX0ug8t<ZVjypb)Kiv9lTAZTB1^JR85ca$sHeE7a|5i%S
zI;5)O?)64o{mt`V|J$eLP4k_1%M|;m>g(LN_PqWly&d*>4$|rQ9h52amZ*Nj{r$8l
z`E7sh(RTmi#zE5KreAro_e*U_?>+DP)nD~~RqM#u(evE})4Ar$yhYWzgDN{oaaTRw
zgtZs-dDgJ@b)_@yeN(FHDSIAORjzGz;$Cm7%5OS1*?7{OUN^6(*MDi{OX~cR^>X6I
zRo3|UJqPc%KQ2Gh@`qiQHYHuy`{}l%i#o6L=1VKCyz>LqeyF-WN!yO!cw*~*Z~WP|
z?OorpZ9C|>U)}esb<clY`?BM>?0HJ{cqMC3wtlkqb*EPyXCGX+smAmlsug?njHOlT
z)$Lur*iWObb9nhAeSISvucoB;z7NH_?d7U_JMHtks`8VLKet_4zx7(jN>@}Wmerv1
zP}t+HALz~mOnI-(&R)rHd9M?qZp)Xv`y95-`|wxpeb~Br?LNBn9}W9Ff7ItK!aipb
z<VW1!0R-9iJ04VS+~@wo@^RCJZHJ`Bebo^vdp+ZqK5oBNU9YJ2R9_z@ZO^?9PfDX+
z$EB4oYFwz!m-c-bw?2OFJ8<`V+sTb*UC$M{_HEU;<DlyNZabT<vhAH#eN0);k@X|$
z^;=LrZ#zUi?*x@&`f=Iz2ukla&vN$%<#&&V?N?sDVe{!a52S6^d*1VId*fdBcG&g^
zo6hyu?|n#qzt0NV@7j&3Vf~MrPS!uw=Vk1vT^9;k&zEg))xR6`>t|v2%dYdQO7D07
zPW68GdP9}!{%)nLeOsR!QSEoweyA#b!`cb!m#%cGIVWvT*L~2ayuI^T)3@W!2cmNF
z%QtiKmDBfIU!mAfWyimPqZf~#TW*U8Qa^vv+V5URrKM@r+tzYT?dR@wyKO2P_wKwo
z?)AH__;c^;=~7*OyLRKAC&(Y|x-;s0D(ZS@*nG16GirJ_{!CHZ<2bGRe{tK-)bpz<
zzvA)<n?Gtlbn^w3W7|B#&tKd7dD6eLao6YlzEj}cA2d%3T4(obH^2QsaqD+nRsYVa
zouGEiTD@ial{;=mRlm42ZGYJN3EpwQyX{`bxtex8Z9fnAjfb-PW!ph@zmM8~OwT9F
zRW`lsct!fxx{Q5uzN3qSZ2NiBM~!E%9kwBmCIQ*=kL~G$u4nv?r`blkD%;Ob_M82E
z09)E8>iNyoe7ovHGW%r5J=b>7@7c0;+^x4)b<N{+*_CZ?H@&ysY^x>U_dJU1bA#UZ
z9JpSFeXhk_FLCpwbw4NSeMw#Uf}X>*?YJ&ZN!d*+F7vfqUC)&yW!L*&prp3laWtxW
zf1jg!+xGS7sQyQ7pQQ4+^DxyK<iGBCOUu5g*K>Y;scv_#y`VUVd%wk7p1U1XZc@ji
zjvJ=r({;UwUA^~fWcj-NCTaB>mxA`I?!RfBFZ{|?T|Ym2ri?SH{E&_NsOkN9Mz!ac
zFYfu#l>D}=<NNs;^t>3y^?r}dlsJy-N7Cn4Rpseghx8jqz58|jURKtw*Du*^zw^E6
z``z_>Th(;FLH@-Zce0XS)jUzwj_v&X<~3>UmmN1%r4Rc%!KBj3@;7asDeJ#!>HV&|
zbnVwYj+ho7LH@*z2UX*O*G^LRn}f=ceSde_&qsMVdbi`YcT)LPT{oF-C+axoZC6#>
zN!>q>TaK#wCS^x=90t`RXg(Em{S)@M?(=J0`$3#cJMR9TQMbC|x?laHj#J)zve$iK
z)BE}D?hiW8n|j>!->YG{WF?)q{<^k<?t5~)Y}<v0P84A|c>BNZZC!C;y4|4jd)(t;
z&;N1L+5Q}^uKP5m+x6=oww-@W=N#mp?(sF>?fb>K-+ozMQOCcaa>AakZAq8beBG~|
zW%mcQSH2y0&l9p7qo$9$j*xHpWjP1^E>U)zan~nQYg*$-TIEJPznGRU>~~I0NvE5i
zZ2u&EG6YnQS4rE`MM?Ml5ZAt~^Qg4W4?*R+@!-zaRI4fN<<=90fD8e5+)<@L{ZZ9%
zZyn>V7u%M9#8F#6e0}sP5mkepv+^F-bsjfm{FL=0>(|q|+S~27s?Wno+0(@(s$JQ2
z2{*l}@jhugNwu@Ca+2odmZz&ew%K)g*sikot+Rf<vbBD;^&hXl?)k)4m7VvQmfrSp
zEbV^lz0L`{ownam%TKGH%Eqbe`OWt9?sZXERaIZvb}==*s&%EL?F2m^Wy^8X&Won`
zlh(Mat6Wv<rBT;COv`6lym|f6eLp{{{jlddZ#vW3-|LU6eixOis(fDVy0*j4yYoGr
zF8+RYz1RC}NBh2cRBqmJ+}3SV=dpfqA9ug^e4uJu)p;ezj_&q~%QtR*Rqc@#J8^lc
z&TslSr)qwcH9M-~ASruM*HiuSCFOTg<)l6D)>Tj4-+i&|{1A1XYFa+Oe#!nvJKTBU
zTwyy0UE_gkSJnEa?E7z3y*DT6aYxtrLAD)q`62uLLpQw}Po}6YKG+TSK6~$XF>Tup
zS~tl0aZ~!S*ALa>y*r=t>(^o1*|if_z4d*0$^4q%HQh7kr!}|qXjA^PJ#ta?t{Q)s
zSff+bT-sXiwl(z%=9bUNs_gHGW#0oEwmsusUxlql+;Vi~A2|HHeXF+F=GdlTxqH)j
z`>E<}Z+W(F%YIM9wDd_oFQuzI-S<0G+qZq)SXVn{#ctU1Yuf2lkMCZ4s-AOo?Nof}
zkyjQqS|OsUYe!bYuIqT~pZ2!j{NI-SVXsq6O{e-kXiz+<E-z_&raqsZFMcI`J_?HG
zpyMr?Jlg5R7c1qG)~MHmvfp<h{ebRQ+)^$F2R$78`j_l|o21hF`5SeAQhw_yM|K`3
zyUr}@x2%0{`k?yA&hxzXgSKVs8}_;&Y`UOzhM?=?wAK|WSD1V5U(ep;cs0mh+2gY7
zXF=b6@~#`}UXS#a7qlJZf6#H=_0erVsD8E_myL7N(g%%8rX7zuABdYz*E}`oyp^xV
zWzW}Xr4O3dn0{P#z0|byw)YR!>rZjVIo<iwe!kVT^KV-IsE)g|?FGe^>bOq3zN*Sk
z`aVEX<^5q>=PKvkzS{9qzw>fh=M&ZO>Do(r-CtLEvg<=`dY6M~s*6)nc3r;cuir6M
z6<?;zuT|B@YbUAk-?Vaq<|jeNO{<TrKi)Wqx-EO1Z|e9Iv_2g6co0Wj^O>M>Z9A@-
z(-}*v)T`UOe6gPf{k|ydeWkS0`OTkI?@wwzqr04-@!gi=s{4s5d#cA9ReqUbM|HnU
z+McTQ4(zOGTxO5Q9a#>#$4Bgkdw-qmyjwQC?Dc2T<ACb&qK?n1&o_SdWbeQG*$KNp
zDi88ipI82zB|?^isn@4z`xQ1`{nE*fd*1YZc2$o*u06l<lHM;n-pHm8TOV(G*s?7<
zuZr5wWXmxvz3%(oro@YE|1dSZDLjMx&>dGv+t(cr-tk!1bLQT1lHS(M5&Hu_8FKql
z$B#qMc`m45yT|?P>)IdWx8L!BqZf~#TW*W+s{7tSzx97r>-m1yA?|$BPfeL8#2p`O
zSx;Bh-f7v9y&r2z`lRE-ufNOgkJ>N%#;LUTs~&Hn@=={n*ZIg+yQ=wO4+3Eb=)SMy
z=Y#6=oOhg7J#M=8y!G*xYx}laPutYnF5Ye1>J#K|+&FXj=$_}f<pimF{hF`p<rCEY
z9N+ml4KF|0ZNKY%)%%0|@>`ckdVi3gwjGZ;&(T#+`tQ#DR~3)A=S5Zdb;XS;yRy$Q
znU-Gl^+S+7RosK@*gpTUb^XQbchvKRY`&m)&~-d0-t+aiF5X_dY2A;ATdwzh9oxIU
zkExqW)OHNIo=N)o4PE8wdQKxLyRzT4F(rM{@1^sOtM0a{d3@S-qRt0h4!SjHJkfPL
zXgwrqxk1-Cx_R1aKkD_nY4ugr-;=Ia)Oe4ZFQ^}?I-WF7+saeLIjSAoc;PPuc10j-
z>mEV#8CA!3#ZqQpQrA^(IjYX{UOTeuf2O66dcBeF`Mj~N`?(5l+y!lW{foODb-r&}
zKJU7JzHNKsKv&#)%Sn1$_WK#$^hx{cEzkDx)wF&u>yPc@f^2<BulD&A+v7*o_~RYd
zWVeInIUN7>eY?6i=$apR?aIc5n_g9%2HA1T^ZK3Ew(F0pe(fLRpPSFM6ID&SuC>LV
zsJy-TvUdIsid)-`n;wr@i632@z5UJB?YQl%s@}TVDXd+$9)9QjxSXQqcm0a1ri_n4
zzvuV6{}^<i+Ap8GKZswDebvW<{E2%!>UY}H&gWfsvu#`VcVTtgpVDRRb!$%F;UuMT
z<J>R5dmR&`QOCs~yWZnb?<>o$8-+dJnU>DWE9|yAz9*%A@#eQbY(2uJbIb9Yr}*uc
zjU!d*U4Nshd%s?_%K8(wzeJtqrIjyiy>z9E+i#<e-=^eK)z6Z)qq;w)WzV+xrPtrI
zxWz3u>3CILp0^#m=aabOx!-v_>-)VNgSO+IKUC!p@>ADw)$0;b?S-vhRQrDU!j@}W
zI@$S_ZS@Q5Z`kW1Z@RGdNe2N4KmY;|fB*y_009U<Am0h7S{Ko6N7s5--1ZNucU=2^
z`BkrTs<Ib%z1T0m9|!iEL7?WA9&O5hwnr{0y{pC_Cf4XwHK$rt(~SB&2>G7hr+?3{
zg_lFncx(D`RqtVtwWIq!c-;G7QS+<ruW{|^ny2Wp>vz5l+8>vb?)<X%RZL5-tG!im
zH;q@2Ke~9@VpmnWV+R5dfB*y_0D<TPqF(R#<x@QmiC!}^b9$c}vVB|k^}KHTs`=}R
zI@iyr{Y#f$rrHg<4m0hz-*{@vepUBv{Op(-XZQsH2tWV=5P$##AOHafKmY>C641R)
zg#AZ5+<D<#;pO12OWCTb*MZ_jy{hk7MqMXVozK>|@cQkyE@jJp)p4GdJzLg8{rt2g
zUc7$B9T(z`i{ys@1Rwx`WC?iZ`&rrcp5Ih$C)+yam!^tq*6o<zl6VLKc>+oGTY1ma
z=J2i;*tQ+_`c5{#TKZUn00bZafvgkodoQak`%Ql@fS-Tf_aE4{opgL;y&qCP2*?m<
zIB;d>e;n|oqg!QsA}s_U009#SB=sG{pt>eKuM6_WJuds6Dc$L(^ti9ayR){6Al2<h
z()N>%kGSOpegBo?r+3`&`|pJ=4$TgH>geh>Y!s+o4(>RalqNmy_>~v;y55xM8NBV5
zv|qZP=ajuaWJ`M6?(-#eKh`bB8`nv1>v|7xz1!B*E0|k8XREsHpnE+=w|(8;C)aKN
z$06q~XkD>v&URJ({%hU#-1yg}14l0&KeyZ#VVnBh|MBkkn~x>6f72!P{`TD9?IP=H
zTm0y{j!et0?boeI^+#Rh#9jXhn?LM%Agn!G(}m?Ao6hg|aI*Vl%T<*=?mB}h`E_6C
z>$b0JoJq>AZR-VD^Ec^rwW#Bdt>=%R?{(O6T-EisY3rCt_vfhbAC%8~J*sOvX<k9)
z1#t;F{^haD4mo_UDmhy<sD5FOCw(3;y*!tj>~*%S=PA?tcKPd4-ScwY_H~W#`DRyF
zd<EIHopaE6+0^5*>no<Eckd(Hs=DX@LH*9Q<5A~HN!Qc%@hr%HQ`Rek{D>Ma-h8t6
z5rdxlwdMFeHE){lyj!MNwaWI#p!uV$J=yfS&wH}|c+*Eck9h6q+V=LRxZAq?G~Mpi
z*KXSX-itnSvaY7im!tMGQ}UV0Rh3^s=ZUnB#~t@f$)DBpp(*`Tl^>>`&w~6*yT67l
z-;{nC)sLX-y{sM&ia+o1u;ZRLU)VTGD_z)n#Z70*I-P7iyzQB_ZQZ}a&~0DUcQMkk
z)2v##&P}F$?_^Ea{d4Sst_{{ZUX9y-vU)zEUe)K7KWB;1%|UkjH#L3G`nTV4Z+}dB
zTle`um3`IomAK<x()puaPne!h)%A7O>?HO4t*&`iQ2$8!xh}u)SJ!^o`i1>nLtKAl
z&*!rB@Jk<5ZrbCND*uAwKI!8@^-^_QcfFFH7sV}4mPgort?K?vzS!}$Lr}f_&bzAT
zpI&=jKUHs=dfgP&FI(q(-gXbWtvVjO_JaC_>BqhGQMGMroVk9>wwo$$t{vI(z3E*$
zVbyQ`Y5M-O=fC-0zN&e!tevoNEt@XtdDSnUw_hf;?T%xnYS?)@=`Oo%>K7|T5C^|@
zRy98P#dqBO-u@{&4ya1+^&@Co71y-v$a3@J;TLzh_sjYpH@)feps@Z0oqzn=*|hyZ
z^|0l5-0?H*{I+nm-A}*%7j*uJI_~+E=eIv<Ir);$kDGUY+~+5~_Snu(0=mXKFHPC^
zsf0}*biE^cJnZj3z4cSI9d^As?R0+kaoB&tehuGQ<CxAtRlG*E6Eq)5>UdV>6}qpZ
zUHehnJ1Ae+>r!vJAiGH)PkLUK_H|v{^5gO{C4XEFZvLd>)mz@UE6@7qk7riQsn#i7
z)?T;f^c_ytRMq=Yv9tT{zo?2wZ@kC-uEa}QM&I4=h;qe#D*HYiH@(+SzioHiKvy&_
zv&Z93);RdJd))n5<78@mqPA1i@2x~F&y;++)}u_fo7DBLu67Q)UKci9)cDW$e6s7E
zrln7+pFaBVfbLh^QZARY#{JzxP=0U!R`vT(ubm)2f{uIrRJ9#ct}VxP{ccdVU0eA2
z`KgMJyIyLWCN+;%)!(ybC#iX$u5x7CL$-hD^4rg@w?2O9Y}ubw|JGHGY4g<}f4ujr
zY~7aar*3+0|B>C+JwKQAM^*ZP!_V8dYMX72sTvg5LC51>hktwKQ@zhPx@>Wf_PYDh
z6|<Ue5pi`<(O%t0)LSKz)@mc_UitDzAB(6u=+5r<J-X>p5mj3sJLZlFSC<M|t<<5}
zT{k>3U#QaZHAj8j<HR|_Hm&l|M>jtBw~3Cen%4buRpY3vop0V*|NNtOydqTV1K*b{
zUDI=!uw4&+`GLZ!8#ju$y79^KXD(f~S)`?-@9%T*!?o6ks45#*SxG-)*`X^7JFgRQ
zH7h)*2LvDh0SG_<0uX?JnFP|hj!mmvzx$VA_Xn-t$sSMpeg39IWm@qNw_JC<-d0Vj
z{bl<_TItjNeW+W0QtcvJ&aDeyoOtxJTO8Zf)cZ$a{nCAZPq+QB^|U3OZeCvdvi;UC
zeOlve(&hTqKkELt^M<tZtNQ(RSbnO?leJ^Zb0c9ndefQm{FK+PsM~S-g{tRJZMBp3
z^<>oYz0b+pzU^`d+YgfZ{f4e`;`Up={9*aI>EgzXuIDe4va73old@~d`7Ej*NzZFl
zmlw6(rsPYipXn+`S3h&d$E>St9Ln+tdp#j**IoBZTEpJ=uq~bG>qhSUFez=(<k3zi
zzE~+|y3TpC>vaqEtDZyQ5G4*&z3)iYj_i1EYWl3aXDn;)1-q-pA12o5R5fR-nttAK
z{fo<2cK&H<dR^Bcy6x(^E=$X<_dJpG^?_e`UcY3w{p^O_FI%2Bz3K0nPRc(o7t^+r
zx;|1}j&0Y=S@l=deY~K45%&FTs?!De74^8S&ojDy$Gs1fcK)Qsf6CeQcPFwk9;zDu
zlD3mH|DbU)t@*#|at}Ul<Gcose=EG|*5B{>DtBB>TUGN&+MaHHZadnde(@G|zwY$}
z?3Zo3@X(1uHHW18ov!lyo?EhIzv_0?ZBJJ_CT-WWb>FbhHF(p-eP5Yr`AusVRsO_{
zZ|{7<8waLst9~xd&z_e{R=2~>$5f|Fi>GPvEbEV7ye73jZo8>oCr#U)Y3I2he`Mp$
z)by&yqonMa$}_BANsTkMmlJoKbn_<_m#WM0u0!hH4jaE=)9Lm%tM=WvO<Pr6U)gFW
zZT^&>9|YpAKZMQirJ3|Pthao>?K75EsaLmm`QjiAnx}h@%g%>_#w|bpv%Wv<`QOxZ
zy4T_T?0e%+)%?Y4H?8fg$Guy>pm8DV$6XF-sqW_llD4laUc$}`!lp}#OIUwQN$2(h
zzj5D|{i@e5blY>!b1r{B)$NC0x%skR)pN38f8P=|T~gy!+U2DEJhoeY+&D0$zxc(M
zZvUqAxUa^$v$hJ`H0gB&UFF4%U-Iwz_X1(LB>h}F<&9m?wZVEPX%4E#QC-(hw%dK;
z<o}W`>R5DrVZDON1yu^l_9|*wR8a1hzlGHcYZmTbP^L#w;Q_yF*DokrQ1*hYt$P*~
zzR<hL0;jy{cFL^1rcbr)yXVnMMb`A~qT&w+G@C7SYubVF+|mobxnlA{;iqZ)NBIzd
z00eXqxb*h-yNxd>n@daGdR?<__~{o8EuWLs505_Z(pn?S<!sfbw*Tv{TfX?oNn71^
zK+Pbad)}3n{kU{%R(bM)(|Ru!vKqHszx>yJ-EVM*J>M2-sq0tAPdBvcTy2IhO<n&`
z2tWV=5P$##AOHafL?K|>{5GmT10QW!r`C7F9b0w#nd_HKxTnOiT~~iy>ChvZ%yMki
zr1?kbv*dvQ1a?m#?t0wrtp^{KT~Etu`e0Iaq{!Oy2-GLsb0&5~CtgAT0uX=z1Rwwb
z2;>_9+piy0U615jyHj5XKmY;|fB*y_0D;{Th&oT+y;Y7v00Izz00hh?P;Xr2%kQi_
zUqn^E-*vNp_rH6GPdq+R=WY3(&+O3%?;rpHlL$n8ZqlSW;0pvGurq;tp9fL@XzzK5
z%WdZ<WGmkZJlOBX<KF2#NuYLcNIG5zHktR+=E}vwR*m}mRa5c>y>}<+c!*n`>~m#t
z%Ok&81fs5UC!NnMYIp+y2tXh@fw<R=-u%%$(VH`9zQpl#-{%;XzuqXpx@rR5k9+L$
zZWWgcG@Be$#|!o#@V^A2-shCfr)pj2e|_hFc>=b+j|{)<exICa?RQ<x3rCbGKhHt!
z<RCxt>B-@}KF`Xw_kMvMdRT=31hPOt_Ie;{`m|U@{oO)Z{!%UkejyMwZdK<qy*0e^
zowT=2=Y@X|fPno3Ok0-;@+WKSHLAYTkya$O`s(r)Z{D!NaaFY+Sy@lX+H>AXKNqOF
zyr|corsPxI&#(u9C<IhLcM`=1oAan1*G#dObep7oy>4syNw*W_$q~rPd_z_LkfTRx
z2tWV=5P$##Y#?CD^-EYkOkt|}eyyxMQ#jxU1Rwwb)dW=ieg!)lT6L~AL%?BY4yxld
z$e!%^Jt%!pc^u!FfU5N{TkWX6Z=~8@5Z9esGk7b=UeNL25=fCx1XRbnt@gsUdD8b$
z!umxzI|yXuIkd3#Fuhg6j#sADW9yhR4)0l1A(xh##sh!yi$K_QbvIpp(V@2aLLlgR
z&+mAZArt<+y~1iom(FYa#Cc05eB$WVd}(Lu34tsS2z%XNdOE*xmHln|?s@c5;pLEp
z*xgkfy!EtoJLtYl)Z<~Vuk$tCyYG*xyJqDI5mk2;lk7{5K+@~Ms>@S#|6G>1>Eook
zZjiRhawUC!6NuWM^P3vAhX4d1009W(2Z6NbwMm!n*Q#N^XJbmbsNeDQt3Ug*O~CYZ
z^rZZYdR=5{zNENOPCDOthr1cv<DMOoUgwPKFZuNl&^3Py@;2&yJMvxl%@vau3R^e?
z#R115009W3O~ADAG$|fQ&jYhkUQ+F&`nd}16@NIO*=zv^yEp`WH^G+Ux~{iuwVTv9
zYDzh_uAkX{-nNxHew(r0YI{4F)_>C)k6*v_rF-X|vPaGoz54hX%a590Jr`D+|FLoF
zz3NrVnW7u+8okHNwtMGn)y8{wI_a>cHFHVnqJ4kb`0U^I%jwelF75i%wZjj{*`iH)
zUblAmbNl6@YFgt6<z|~ezW(kjY`t~0e$vn1MJ-QP{jm!HlL#bzU&N%ktMtXz>qkEZ
zw${h%ci8dYkAv#{s=k+totAZKeK*_*+rFyOsg4KiK|qFpt@nZA`t8R^Rlmkg<;(A^
zJYT>eT@HS6?zexhRrj{(-0?fdRK0)iDP>!ovc<7Y$GqO)sfVsAlS^9t+F5r05z9&n
zt8O3Xpuci(=gYDjNe=-CKmY;|Fp+?1&wa@HV<J--pKP69Z9jAUk_q>eIHsu#d(uJx
z0(KH;ci6W*&U#>z6IEY*d*2ps-1xm?i@yJ_HUBBuf2(7=Ms2UO^Q~-p>LG{y`3EPf
zx@>%naWe)laa60FqYGX`Kn(#~es_kSl@2|k$t=fI4zlm_da&P($Gy{glE@d0`hLNr
z^W{r>P|y78f3~!@nzK4qf(V5Doj|^(3*yJ|Y!cwTB6hu3Bx-v-)~(wib(((Rpl~1#
ze(MKOM|Jo4D^oRUy~qav2tWV=5P$##@|A%1ceiZY@m{lhjaTXofn5=><vs|0a(~XI
z=L2@t*VvaHfg^6bY|`vr3q;!b`H{01)_VC9VTz{LB9sgP2tXj41jaNuV)>VU{#c-J
zAP&sw2f9a_q~9+|d3Jq2lJ=s#Z2I1*U%cubN3joqFa&JB?<V{H5Yy7D<{#AN{~1_e
z-@$uU%-O0vW?c3AyDIIKL(}1KXsc6tI`=PglG31f<~Rf(009U<ARh?$T~D%~b&)Kr
zix9`eaT0Ny59bQ%2muH{00Qa>n6{3LKW+Q&dGu1Dn}hmD!eSN)=#E1_`?^@FzE72w
zJwI+i`*m@RYd0;9s>%)GZp(31>lSh2f&38I9f36ub>Hj$a?>4Kb$4~sdCb)BTqZT{
zsD4kpYfopJVH*N2fw<ojCV#)d9rk=%WSxT>FRFBU^+8`ga?~mZHG{+X*M9oBc;GV!
zg#&S*uOG;cpIJYz_|;FAQ`Gdb?@^6fj+<{@<0sBrGT{@)7B%g=ma_i%#aF)XcjM7C
zjryMPpnRs(Q}=gvOsS_XKk~ibno@sV?TcLqWSM}j_YG#58Fh<K;HEwU3I?2WkI0I~
zCxV4I1WcKS1o<I*eQ#QN+3Q_X(zp8R@)mF2u);}7RXukTr%#X<0uX=z1Rwx`Y!R^K
zdE2D?bX!;V^M7tRLE7=t4Xrv?n<4U5gZOY90uYE!An5+0-|^@rk`n^v5=d&Ep{pFf
z{;F$#QZ1?L{!&tQgVu9xIiA!!G^x0v90))F0zu>N5z9&nt8O3X*se*P?<i;I?@tEB
zm+IqwJzCd1&=$LXT-Xl*2xN~y+~+5B-v@|mKYJ}kjUWI4(+G4w?y<|eRa`DmIS!_^
zBmO`jO#-G}C#q(l`*)DopZC+|%EiLvplh8;bzGQgFYbLvQ|pCa5P$##AOHafBu^mj
zx~r=E$=93`ApijgKmY>q3ApPVX=(f#U_mE=xbK74=^M5o009U<AbA2w-=|N$hLi{a
z2tWV=5P$##An?Bg;;sv*%KyJU@xOTluB&<Bh%)8pIjYpW8sd=-0$F+AvbP>-t+RW}
z)zN0ygn%vrs^$w(?dYnPX?DHiMAYrL&;63WZQnhQUMk=K2jalLADA|N;*XsK;{HxP
zZhn`Vs^?N|wc~P#szLLWsCI+$*~&Yr-$6X$9`{~1tJ)5V57o!h`d)3)_vdw$7dMW)
z`J-O{k?+E9u9&<~csay9@8xTLZ(Ons0SG`KDgp1fZ2ET8d1qGgMQ!J-=3|{R+v}WJ
zjayaqQhnX&wWrEY)poKno>Pw}A2_Y|V&TUjYTW3W$3*qVlzh6l$l5i92Y$pOAlvRi
z>En45oIC1zW3c6DDc*g5RNXZzR|wP8Z#*`2e^6YReq8r;Otk1MQ|9DZcmHL<0X1^A
z>g%_@bno0#_Q=_)UsarX{-0~@nX^@|KEB5Cqvlu7rL9rpKWaWx$H}mM$&N=+ImzY=
z>p$sszOItxO!{O9$o4bc>6587rQ|CC+s0+nuN!3jODgWX>&Uik>w15CQg&tAIjQgG
z>MF;y?>B_?N4Ec(mR|L9Sh8`eD}7S^lyY|d9w|KF@lNkaf;iYY4s>7lMYV6+eBRIB
zq|d9{&s@J`!aXHUTn#(VFg2ZDd)Tr+>A0c1oqtatJO9)*U-Yv}d+pp_K|K68hTR|a
zJQg)ySieZ8mq1WlnRYyCyd|A4YO6(k&PFz$pMOcsAN<N^KLqlTz_nlZ8{A>fw?)3I
zs<ziP-mCI2$d0Mkonie->O3M_P7rrh$7TI8ExmicNfi&<?Z{p)lfLrhcUGP+;E*f_
z-Th3}{e>VuZH*h#u9u_wlh*ktC@#`I?l=EQ$}e5>TI`1VJ-F;Ti)ra;=Wy-p$0@0P
zo>V!i&gVgPg6iYPTi5;|e@s2@UY7-_-+9W`{rSEwM*UeIHF15|)NvBOq7m3@)xB*x
zcl^#tN~3v6E(kyX0uX=z1oDYMTI<sJR&Lb$2IP}_Zr6{O?eV5--7YD+e(j=qe?E=f
z)K#8<t@pvbeh1wjF#Wiy@8Mu)K*4}>?h#Qr*n0m$^?4)6p8QBdjw}%fx<8oo@t}BB
zecbl8OlqDRwjA4en&N-b{m%6AOlcSV&_f{VJkOMTw!MGN&)>A?&3d9M*y^FJPU-2~
zzs#{!P3Z^u=7(QDvSq*S=f#419w~GXIAU2zVb$&9oUH1=M_bmZ_1$pCbX9deZT`o`
zt@o-|Ef-c#m~-n1|2U*V&UP(q{p1gyY^#`4t+IVC==T}2_DBx_8wr?x|3#I5s@EB@
z_v`Zxwm+{B<Uj4K=DC)%Iodcj<1+*x009U<00IzzfIb3YpNAk_y!U8im0!?xWY&)d
zt*3JQvCF$vTrN~|&^IFLv8?L;iD`CBf8Hv{Kilu~uX(8ZUiX)q?x<Ehy};k9X>pxz
z{si@td^zqPp&W@rz_xi{oVt=10uX=z1kxwq{avnY+p^cq=@U2GgRsvX%BH*Z^ZQ2M
zbZM2Gt(u!yF#O)>RdTj!#fJ44*H~LAm$aH4?RaMj0o%?yQT_FM-s!tPmS1z@{zHVR
zYD%2Q`r*ghmi@Bp0<wH`r<e6VZu+d=NA%W9ce|;wKj!rYPd#*1nH&m-e29an{U+*s
zLbg2bIKuYM&#lPvBs~Nm009U<00I!OmB9Jee)_q1;4?>7ZLKeULjVF0fIzkh$UYz7
zrVmR+_j|RlpYC@a!^VMZx_o<nI<5N3woAUJx8=U5t$yZvd{KW0{6fH;XC<xgzCWt&
znw2YrE={U+<DO5G@{@8P0D<rXlA70q_lT4b_>F+B^QNEO-%92`2*?vidVZ;^yr|E`
z$`i|v98q~Dl`rj9Q1yOg)9u84Z-Xu0{|NFk?)4=3i$5IDY_^c)kkokMEhp{qMRoa6
zpPwdQ<;(A^JYT5h5cfJ(b$r-j?}!Cge%bfvADy^rOMUSZ0uX=z1R#(l0=m|LY_prC
zHl;2Qa0zTbbN!MD_mnv3uN+)1C<Js8u;usaIvGT<Z3{R2-1&V3@PNmjM=uq+;vmXc
zva3APs}K6}k)u{QS=U|FJ<7hg>*HC=AB7=#AmB&9d%sh5+s`ihA&>@vp+CMm;k?f(
z=Yn+Z{_8(J;ik$tbY~7}v>9bW00Izz00g2E&~@F9UDlPNtSd#0WAZ@&0uX=z1Rx+!
zAnx<vY31KK=8VI87FEdErmw!eZ;Ll>{NAx$XI*m5ph`s>9dE?N{oR|Y{9eD(-d6oP
zn0&F9cDqqN1Rwwb2<%MY&e=;R&F!&5pm2y22Y&D8VgJtW6WqD?vE?Oj#IlmYs@umo
zX{+h)B?<CR_j|#;?UwcJW#em%n=yEaldt+%w{C~jY5Ij@oAUcBGk%9=+d4<qo}V_Q
zK3P9cP(KJj00I!Og8=hGGv<f3x5G_+1{4f9=N^H=!NfRFH5$wQPLA|XK5$y^#R3lL
za8MQZ*nt29AmB$J>Gids_qDsn!`=r9x}W4dF8h36(E6_**R1cC<w*JstvXkmq5OU1
zLGhXN@vO%^^@9KeAdpo8wqEaNwLNs#OV#h?Rb5Be8h`k$W}QrTykj2%5P$##vPEFv
zqb=*y`fj*`?#$tdlmFY!MIDQ-FRWKkrJ!m-*<M90iwes9^0%;hVa>w*3(E8;Dm>tq
z?fL~}3(8)wwRO*;!WVisS>Tk<R+La12tWV=5P$##AOHafKmY;|fB*y_009U<00Izz
z00bZa0SG_<0uX=z1R!7`fu|c(8@T=Dr2@?_4i<(gE<*qU5P$##AOHafKmY;|fB*y_
z009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*!v6X5q1>HfZA
z>wp$(KmKTi@Kfypptu0Pdid>ETs-C=009U<00I!mAc3LJZ2Muv*m(lAn?nXelPW;~
z0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00fi~_-t&u%d5BhSfFrF
z5eLcw6GISy00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz
z00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<
z0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00IzzfHDHFlnxrQWbmFj6b{6J
z{ry1Mz<?nLKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHaf
zKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_
z009U<00Izz00bZa0SG_<0vRRnr{hMyytHVxK;b|fn9~n*4>wOYs5Wr>%S#0c2jYN!
zkkxTO_c)4u2!tUp^qFlxj2Jsl*s3>o>HW;*l|B>buN;Ef{G7|Kym;@ss^(BQ*boOn
z@yu}uKmY;|fIvPFV4ll-pX(j2_ZSD_jRW~`uAq((fB*y_009U<00Izz00bZa0SG_<
z0uX>ewh6FK9B-W%e)+{O+oJ)sg8&2|U><>dJ<pog5_kjw2tWV=5P$##AOHb*0=nk&
z@}7_b0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z
z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=
z5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwbmw>MC3x&1o`hh|K
z0{RF%y!y?tzx&_~k#*Ha5X%sN00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|
zfB*y_0D<fhc)CHgf!kkRDooMrlBQ-5fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb
z2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##
zAOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|
zfB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKp@)$s-0i1-K4%1
zb0{2C#6h-u3$=rQ8UiPdDQP#i*+gNgs;MPbAOHafKmY;|fB*y_009U<00Izz00bZa
z0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z
z1Rwwb2tWV=$r5<FLA8O~UtTKmU6Z9wX%I+;z|CEHKXZAd&jk7_hh(TzO1=}Y<$P*d
zJK4ewKbbc&Z%*#ILDe|*+1PfMS8w;R$hxZH#k_In`Limn*nt29AdoczL#n-V-lpe=
zI6-Ra`$N<AD{JvUeINh<2tWV=5P$##AOHafKmY;|$R2@`tygaw)c6O-HodpY^nc&n
zuzb!o-EqjwtG2CK?_@=@*CX`QDC_qxb=6N#YhV=u5P$##AOHafK)`kaHNJi5FGn6Q
zU8JqOaudf?YcxloyXIhfoXPtC!Xd-jcfR5yfx>|}pdT;}L>mWW$6eAx00Izz00eXp
z_+Z$$TkmfAf<WOACl0Q7>GrbkO!<pI;b3AM=omY(2>}Q|00IzzK)M8mKC|tI5o70x
zq*V4fu^_+FrAO%yfB*y_009U<00Iz@Bd~Qqi?ttrv_iyH-ut%md*9#@jfY(Hk{SXK
zfB*y_0D)Z*c;e*$?k(zAbbVpHf~p036qM~%)Uv3c+%JC%s~6TR+`phqkD|f@e%Y>H
zP`04#1zTJ9EGm4VcasH9`CZv$9|Rx(0SG_<0uX=z1Rwx`L<zjnwM~Qb`+p#0RhN?k
z&O2pR;~UH8bm`t5?^wF*m{pFS>Z%uZApijgKmY;|fB*y_009U<00Izz00bZa0SG_<
z0uX=z1R!7{fwRi=X*uVGl>$}6!NfMkCkQ|Q0uX=z1d=7NcE9ao)*QD=pm0bt4w8*}
zN`n9dAOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1R$WFz;nm8uTydKYLTz1
zpYYBl_`R2f{kwl}O;UVzZY{PT009V?OW?&O*Y_*8d6DDQ)$5-5ZqU@34w?-P=0+f%
zWtl+R*G8?Y(P5~tO&{Oq?l&uUnj&o18y;wLR;`~FiMX2O7}Zg?1NydUv7~T{6IXTk
zhD``S00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fI!v>*#2I*thW#K
zg8&2|009U<00Izz00bZa0SG_<0uX=z1Rwwb2<(c$MR&EF^_{4iQ>CRJ&HQ?Mr3yJ)
z^`8T;eq-OW);hLnzpoEkGj6}-!gdY%UGA=8n0*j{00bZa0SG_<0uX=z1Rwwb2tWV=
z5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0`?O)t4yDkb6!{}@<r`$F=_w-
z2tWV=5P(3^1nzmh&Fvj?LmksKX)cs!Hi0o6o3;C?%LfjcZ4PEfA>Kg%0uX=z1k5DR
zY5CbL4|uFn&UXFJ+HO6D%&(k7;gHX9uw>q_9z%LB5$LZR%p7&`3Ib*k$d~(})H6Ts
zf1B02?D59-@4Dds_AzUYTP5J2hJ!sl7f&Go0SG_<0uX=z1Rwwb2tWV=5YR;+?0vSV
z@87R@Rj&>!znCXfX;eQ<$!GieecRVYt*g;ts4!J^^%Lwu00Izz00bZa0SG_<0uX=z
z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=
z5P$##AOHafKmY>a38?ygc~<QVYdB=+$d6|V-5S2dNC^Q5KmY;|fB*y_009U<00Izz
z00bZa0sRDYuZLkjnRUl};jF($Pyz%X009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=
z5P$##AOHafKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHaf
zKmY;|fB*y_009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_
z009U<00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<00Izz
z00bZa0SG_<0uX=z1Rwwb2$)FVjjnAPoZtTg5m!yD8$Lk*0uX=z1Rwwb2tWV=5P$##
zAOHafKmY;|fB*y_009U<00QX}czmC`->lqeicqcT)+&DKZC@L;u11HUB5jRd6D&Xg
z0uX=z1Rwwb2tWV=5P$##AOHafq)A{{!y!XQemqm4>2OH1?I{}q5P$##AOHafKmY;|
zfB*y_009U<00Izz00bZafpiE|ysB4+m0!#gx;5?RjI&yPI#ENZwiD?0^+9XK?YCUm
zs;Ym_fxU3QlR2wQpO$l8SSgZH+xr>*LjVF0fB*y_009U<00Izz00bZa0SG_<0`?PF
zGH+OqA-$If6b^CXVC{a}$E-PSm9Rza@45M11Ks!iu^;XE3eF>DokwgLkJI)uzeiJQ
z4*>{3AUuJz$7@^44^MRGlzVr)W9hPERynBM9CmJ1wp0<YrT?q;Qx!k#Kwwt{RL3j!
z!mW$$sx8>}8-b*r+s=A9zY*m>2tWV=5Qsv6c}=)^O;}D*f6sTr18vT#_0u8|SEIxL
zc_07*2tWV=5P$##AOHafKmY;|$QFSE#$Mawi>_sJw(1*Q+cY@8{|5qvLlSY2t=>;<
zAOHafKmY;|fB*y_009U<00Izz00bZ)N8qOAgU&qixb=>$D*Ih@IXa|<00bZa0SG_<
z0`?NP=lM3bcgziSlGfGhp80Oj)R_({$06wbW%foQo<jfv5P$##AYeNI)$geY`#W9R
zx#2$qAOHafKmY;|fB*y_009W(2Z8gh8P>SPx{5heFNgeyCTa-*y9q>n?;iOMS~G6H
z<-#@&c1J7TLjVF0fB*y_009U<00I!m4uSU{{Q8`&|6c7RrAIWndH(&awmM1cT`zv|
zpR+z0Dx&Ih$F{FiarA0oy9PaPZt8K{xu>;mlh#Oe|Ez9r-t+i}j%uCqm-BLCKbt9R
zSKH&+^*`=&p=rfu+U2_Z;%eIUpnM2G00Izz00bZa0SG_<0uX=z1Rwwb2tWV=5P*OU
z1mfP$NGiV#ER*o@hg%PN;pRK5<f3X4K2in*AOHafKmY;|fB*y_009U<00Izz00bZa
z0SG`qHG#OlFHn_Vb-l0$0SG_<0uX=z1Rwwb2tWV=5P$##AOHafKmY;|fB*y_009U<
z00IzzK$$YvUR$~A)aO<@XtEs2HLCl@Gd-6$s2m6SECe6`0SG_<0uX=z1RwwbSpsi#
zZPVcV{vQb2RJNX^hX4d1009U<00Izz00bbA?*yJW`M(5;Iu>1DSg&A@f@%e2dlj`T
zDk%5M-@@vJH4FDIDAS{;@PJ>o>lc(QD0{)y);)^~U+CRrfm6Or-MPJfzohL7p<8=B
zzNqG2|5_$gHJEEJDZPH+>#dJzw@_qF|9bh74n6l+C3LH*c8_W&sD4rHnv!qJGkyAB
zvEVBaR@3(TmSfJn_~ynlgepzDK1rAV?BSJ$P2Xd&2&+l^M|oxum{nuiF)v-ZL8PTF
zjgPMB|Lv?ltrN0(%EyOQnceGKVXIbHzv-#>+O8LAtJ@Byt6zLPv8eIe$G`No$ch%%
z`a{bRfBZsZP2<MbGtX?jVUK0&99fN9u3!H1dv|$s$fqAWX{qa%U%yCuzw4hWRW&d4
z+EL{nc6Lu7>~-?)t=MxkY<<1y?5Uw1PrW>n-qusA^sTOLani=qW-k#@HT@dw?7>4V
zo-bMb=@*WzDm#y{CH>B=^xM|4wJ)uI@9cR(xBgaQwEx1+yQ5hzli2#Ycl|!9ziz&$
z<Ayh%Yfo0cslMUG1FEldY}d4|7gUuyqGG?gZ#19a*se$KnH&7$y;B`qHEI5`{WWZQ
zRqHlk%Of2GOd=5Xd=|GZX|EHRQobL@I=37?aqK6L30u^=zGchy?+;!1@s8Wpima;c
z`&54M8+N~L|FUjBEN9*6+FgA6Id|0EBy_26|NZQLde%8RYIP`^OG}rJ>HWgUE<ZW8
zYTD~Ze(kDj|L>k$*!PbGpF3&kx+k}-T6pSgVVn9r7pJS8lCs-l&zUpM>oUjjQ$Ihp
zU3O8^Q}_GEF<nnR>7Jz@zPQ1$RaLFy`?aHYzh9i2y8pr2UtId}lwu*P-g^6OkNmLj
z&967!FQ;4Qv_HP=6Sr-0Y*kg~f4_D}YQL&=x3uhpeXf~wa?gQ><rVfmnd#}0`n`y%
za>72}ZF)LYago$`NjYW^Nb7t<x!HQoI_!DQFP-k*5Ov+&&oA5K*ZVw+soQ?>V*7sG
z@#7b_rtHu6`X}8EruOf!ewn(CB<q*%xc9Sf%5wo}`ytz2rln8%`F+*p-Mf9<c@yf5
zal-0p%SSwYN&6v=tH;;>#~YQ89wAh#ZTFRe{7w4&;+CgsUB+}fZhdXju;*LV>5}H8
zdfYPAo+_@*ZvU+N(pR=RvZ|_luN~FAz4qPhpmkF>ovfNZeuec<_wS2Lx9_%lP@MQ3
zPkWuh?|LBY{-F7~-*bp*zlX!Gd_R9x@3-~2TUGJu_8at90$UG%YR(~LCyT6U(DQ(9
z`v*PmuZv66^Gne6m9FDa?^CGG=eC=xLBDtOdk!${{;0njbk_yKs_eQ!*m<bm`9EmC
zDd&lx^Sm1eQFXtkM^vBIf149myZ?BK=+$h86Qr`wl|`+W>U=@*qWie&`lMwq?Db9B
z=~T}<z4pS+^JUYe)vjU7^;@@BHNQ>EPT2ENzN8Bq54zInnzzdO>+Sd6ZNKYb)AxHh
zr1g6>Z@K?2fBrR(F07QZP1h|rqsQIv?3q)gZ<ZeS$cOXx$=Ryj_)2O!?RiJq<wxb5
z)_Iq5_5AL^%Qb3TkS`em-g!mb?PR?EtrSz%Tf+MBo9Dm&7j*qEdmX1cz1RPwx6?k)
z#VudgI&)HXlYXz7>hi*_Uz?t8bItoMeDcjjj;%VX&oh7TaNAPHR*n0<gQUkzzw$O5
zH~XKZKkS{$nl9S!s*jKQ{s+fZzxu1*@3&6L{$#$JW;^G6S?7&<A759#q~=em$D6SB
zqCWo{*1oQErp?1t)l>GprmD)d%}(0yWpm3nom<%FWs*v#n^(~DXQmy`%6UC)u5sf^
z^?E|u_F8=O=blafwA=}+arv2+U-mqjc6wXaYyA9Hbv{Weu5RA@l@<#=Sns&1+K&ly
zzOT@~%WB7V4Z2TZ+i~xAcWm2^dOc@KKJR^E?{?7rw4mpkgYIJ{^_-OM??<a1uVn4X
zuE)sQ*PT8nj%1Jfy;m@(-$fliy?l~>zp<+K5+`jZ=s8a>SKZsHfA5no_S|;y>vuu>
zRoyRe{XlmnV9LB|XRqY9qW+%En@_&v-RF=!|B^nL@2uLwdH2z!|7h6rR1i<U^LbeN
ze)(1Hm;Ifpn?9`nQPatOFCp#qUvK%UBQz`LD^*-%<HNM{LGwS=$K8Hwi>iu~xORfZ
zciqQb|AU?b@f#n4uJ5BB4_XH^{kZJ<iSG1nJDaZF_0XX0wCba3ew4HwUE_(j9io;K
zb-eJ)7Zg9LkNf!%b$<}Yu<_%aKl*KZ-;d+Hj`!R4KCc$^J|)@Xe)nP1-tWi5mi=zO
z^;-{&`nxLEZd6U`xglBq{5a*ye%<qwxb}nQg|h7(WH0Erw>-c1Mh)Am=ddxg%H)Fd
zm1_^}SLX6^xgf3CzRG2trj^SD=^Jm4?0oJs6>>q^ZPr7Tet%EJoS&{(*6E{{Hdf5V
z)r0mswO+*wE9X?H?0YQ=Hva9&gIib1>C(M6J&-%2`VUUNtG8eHZEtkWd%$UQa4sqJ
z^CPMKLD%7?AJ?@G7M1`0<6k)I{XHw@Y|%=8Y1QG?J{z5=8kLh@zMyqN+2=RH{*K(w
zpS1ShanHuXR-aWqmzJhgZ(GX^irch~M_mU<JD+Snch@E2UdQQ*Ki$t=>GIpPn=jX!
zrqwg*xZ!u*XWM=^{!G!J^-<H0yW?ci8n^vyeSb+<zvJ==n?LG!;^qq~N7waO(spfm
zuE5XFr2Dt~`%J(4Jl_35<7Civo9z5FC~p0ZKf8AJhu_Tq*0Ei6%|C<IqodyMl%4;Z
zHg9wNiCTX*pI=;>wm<B3S5Uupxw)ElJ)_=_G$o(QF|NwCgWq{6>V8$fe|GI8J)iI^
zPnN4}`l!z<lkeP%Z*Dw8sOBKse%|!5^By<7*ACkdNRxo-er3A7sOP;j+h|v1+wz<l
ze&z=Mc<)rlHV*Fj$~JY^$Gp0$J|wd*==$1w+_mlXBkS94y}jyQN86%7?^E$R?yWc5
zY6<wQN6B92d*jgc(!DMXQg>a~%{Qn0@nxU5ZIhF<I<wz+t400Ijw$NCe>^Mtw^~Dv
zy0O_B$5!?Ju7m9^j}H0tV@DSU+va&*f1_R>#m(oAqfu2?d-~b^px?38zuk9%V~R$7
zZd5j3)b>d#pZC13YCFh(UHA2{`+UjjPrq=&atQlg;Hc?>;=ucRNZIXsw~sq-LcKAL
zSA)t;>bUnhm+h4w@3?KPNQ;B2@hxdPx_S9sC+glGm80tZYudV<>iYTFGyVR9tba+h
zqaRQ2^+DWi+4G^P>1}_%x%>WiTh-R}7}syv>n>B%+xmOv<zsrkFtW=}j%xKAN4@)X
z&qKo6_xdHfo%T5B_9s~ldp&7tI@$N=n3g{5?=iBTF0J<%+0HfX?;3f_mmN1%rT4oJ
zoz{NY_xQ-(x7D3q_xDS5+aL7%f4$T1j~hi&+V0}p&$;9Ov3K6#Ru$LZzn7wb2zFu=
zdy8VP5#v6#SZ?fMei}6fqh73N(AZ*&pvFpau`AKUuGq!CFLp(dCQ_BAs3;<cf`GiQ
z51Hg~WL#6$v~#}y+;yj|S@W5)r|o^#{i=jq%)ymA*SH|-*>8`ze>vu!HkH`Y^g9|I
z);R6dP5)L`i6v!@vRn0i+|>N3_x!3Q-1a+3a=mi7Q@NKz*}+xbP<q6BuD&F7t{-Y$
zxZywc`SAJsJ}y#~q3W9}AJ^{`>ZTXp5medj`~1sQzshdmZ#vJH<Gb2tn_k#2OVe|i
zRO5%*_mlal+|)f<gg<t}AwB<9=yC`>PgB)vwfmp3*%8~{v9PY<OYV958v{nH->TwU
zs_rl5;xFHWsuj1Yf63o(N!?9-XCn1*+xtAJr1QOB>Z)JW&Q<<5-Z=l5c9Xv-bfxY0
zf^_xE@p8RSaYx@4ea}C9P7zwF>e<vk*8ZSvzP^9ATHJ3o<!Bo})%6f%m+O5t+w>~V
zw&RC<4};n-qRxM9@7>wH2Sn9Vs^Q{2P{`{?u3rCo8H$7N_MqyY@Q+Kt^}Zo>4&po9
z7d_i^FVl3nuD@-cOTN>I_K^BI-qk))PDl_iA>jJGOgVk1eHpoos_#zf`iI{4cg@fB
zy(QgqED=wx=UR7%YUi=?5%tb3MZcZ4&9YC<*rrV-H@VyXTfKJ0)cT68<Ua4RwZrtD
zU@Gx_pXd43PyHUKuW+v4=~MkqsV{%-`k;Q_HFWs+o%dX{%4zi#Q)zm>Ozl@zTX-(t
zNAq>RGL+s>_cVRw<9naew|>)cAhr1N`EF>p?fJUx{m5EPXSxrnyKb+Rcw$|oD{FoI
zEtfyneidtd&z-Mn|E|{fs^`70?~=RjBig3ts=rX@*}n6!z0aYWj(T5;3a7qLLETsJ
z6|VcdljEg+ua)b0WII3QN+)yo<>nesV(IZcz6kxk_^b5F_ug`G=X#Gqwp;hTrPSlu
z&NF<cQ?+|ne|dev<>sEZ$I|D@*EPNTeW6-+yYl2tFP?X)%GAC;l8Qat?I4ss?(>pR
zc8AJ`n~$kJU*&YYcS)sBb^KD<RW1AJshxA_+GF~i)zspLS|9ov4}@N>^_0Ax3*|N)
zUtQzt?qiz9Gi~2JA053`<1Swp-14kLm!A2}OWzizQnaU3r0RSk7k}ILL6MH<can2y
zC$gR0Z+-Ci4No0gy2@1gv1|wTd5GYo+JDnM&*WN{%H^f*_pNlpV;u+gT~OG{N5=eq
z^V@#^UExdWe%HYCUZ{v~de7XnoVvfm>-xT@ZaxcsyZ`RnuQa1bP5QdG=v%Mtd9U10
zx!v63Tx@;vb1FBt@Ar2@>33cKgif!!pO9+2Q0rjd`B3|Rw&C3G-TBhTJ-*BO#CMib
zk+$cA)cq?FE|)*HTX&wHYCO~ZrqtqdKUb*s6zl#JSH8K;pQ_DX?)FOQquMhSZu@+8
zO=tT&v;E#?>gn!Yw$I&9tx{LXRhsUfR7-qS|5BYV$M#p*YwOQ-PFSsaOxN$O>ow|m
z;i{iXPvtL%bI&)a=Mq&PvHGWKACBw%z*Y~nu2I=->aS~`IG53tT=VK&_2_a+Jzlk~
z6H+ga?R>yBov+`O^|k*;>7~Zc*x{z*bgcMX?~laN!}UHQ^z86Rm%FByrX0B5M-_Tp
z$2s5WYjs?4{VphXdee5AdVKkJ7G*bgyH?%5%FRFZ`L!?mRQ0s+*-zH@{PN>6wB&A|
zLf?bCzq;SI7fdU2mp{9EukW7R=F`$t3J$4B-F<~r<86J;h^=~D_0J-A`Q7`6UHa`#
z<4ae`)xQ3?pu;Y!T=iKQTXL00)^pB^`|Q|rz=X2a<lp*@oATanPZYNDvVVTM=VeQM
zT85VM_T2IF57+#lbd_>Duyr4^^46Q|d;5c>u6*j$MMv+u$(v<r^1TZNeR1I`lS@}g
zalGt+^@pr?O}~O$Qu*c<F0U6{-PC>#wVo2|d0pvhZ>s5I?d!?;y3WH~*D<Q~impG`
zIT3YE87rK7yotR}#ht%yJ4rQOtainfZ|?88Qs1#mJ^Y&$F5d6DhsG7TOYVCqvOd%H
zp89z)WuNVHk^6b0(q}pztMuslyUtTWr{{i7hSo=&)2ANp>pO+I?Nt?zx(`T&tMpRq
zI@@qnexd!Tb*XE(Zh1_fL#f1<*Nv3hc3l~IoaN5X^tomlpW9wreN_D;eB27l-afhH
z;$T}(wVF;Ik9~2F^9k)veIBK<&vrhMS~}f*Y$~3wd$=mSt-k!~`sYshvNV-cIb8SK
zRrXekzw7uwy>F9Bxb63RQb`wjz08##)jzqzJ0CY>*?q5RTXB`P{eo+L#jS_?Ic3X-
zJDuw~x!Tfm^MyVT&>^6jS9AHRY@oh_X&Nr?v-%nz<nq|=Po<j9bRQ{|_;S6-?o{Tp
zTz0tbLkK>udZ_dwA4+nKH@WJO`Q&~MAKHGIpQ$_4I7ZnO`gi12`B8S}7B0^Zbltwn
z$-RE&(&ua56iTmdz50&ltKD$vjrF}8u6#q!pSieGdbof8CA7X}zq+K-wd33(cd7e3
z2d?p`_TVdA{N4h$?9t)A14qB`W#LQuzDMhex2!kU->XaI9+9mbw)a_6OK1DNiz!cC
z-v9Bl78WXL`ui)X#?SS8tio>7-)l=HzOX-(l*d)!PW`*Vy7rm=j-ZGyIQW*K=iwqh
zY9CcKUQ_#()cLt@p1JC$-Xl?kr&eC7edvaVE*I4wsBqD*Or@%yar=knnOlC`yiN7x
z#tZg>z>)}1&;Q)QeeZJ%{YzqL;%B>_w@v43|HX9NG>y--f3DJF%d^(g$!$H<d6#ZJ
zvVE@ZT<=-Pc5|JZ`qm@HS#Bw}18)67pPWx<x9Pe@jvv}yIltWI9kRW;>*iGBg}NW)
zD<5k8pK7@8^S)5}L*Iwv&QJIE?5TM2xMu36^rs#!e@`ekx9NJtHoj@Q5b=GDpR!(&
z-|*!JZ+KU?S!HY~>=9D+-WauB5c61}$5yKHhyTRW`n3DFRB^E7EvJ*`U%9*G`8eF0
z&;I3ziKQ<N-0!i*(x>WYuKsd=%hm0=-r_2^c<#iKR5`>xMrcXdr5he<okQhAwKsWx
zPPi%V)Vzk`MCs=ar|fYJ&-Hhix43TXRXa6&P~<A9_Mduq==G>;J6EOGy_Y}n+V?*?
z>d}@JQ#s&~$IsmB##R+wdEt8N%s6|{(iKzrw*zl`<<s#W6}nROJ|E>D`+be6lg}7(
z*X-66Q#rlQzi-{{%yyMfQq|s6{=S|Y2M*bz&+fZSDRiZ{7aCfsxaj)(S~uwCFXHLm
z7f{`EOx0h^uVP8p@3F<+cdOO>|GCxe1FzX>cA2|0{cd4u@oTmHrB-h7oDkC0KG*MG
z#!Bxxo(q1uQun#+JD#q+-0?!s=T!OWmP<ZIb9L+H%RP_O&DS>GQU@G*+Wl9~E_9`B
zKB=d(-KQ7ja;<k?=PS^=`20%LQ)<%p^TN0ORK9;~YsZHF*yqFN@B6s$C2iYV?0son
zyLF!%sm7!BD|N%E^CpBpc>IQ^jxE8Vx;ePcgOL8Ge}_kRT$bbMj$^sRduim?t@@tz
zX_1T6Jy#1A&vbsleV#1Zr|r5cbUN<yHK9*%u$51Co3h8~)!J56(pE2(ZbrvhcXsOd
zVUe2r&E;oZ`>*js3vNmAki$*KQPqBztAFV8wpjOnbN^j$VZYptT-~wXQwjAv3zd(k
zr(7l1`}3;z`c?j}=QULNSm(pK{nOVxRn9-wy=L2dsd`m~%jHhZo$5UtUAuhU&*ZA7
z+|D(qd*js}F6w>O#v@L8b#8}>t2Dj8DaZfypC?=}>Av|zEScK6$<!Y1^;KwnkF~yL
z-2wZ&RQQsjy+baUfbV-lOU^aEuj@UO*y(fKpAUWRWZTlXzo#Me$@XyfKkoagvc0j}
zy~tm-OV_<z%b#z);iDgx)gs0Ib*{4W&W$$=K5$yeC3V|@X}k-*-L|~+*NY0<@7$~M
zbNxM*^M?OzsjV(-UGXJXU1|FbTJN`Xg_4i9{LN9<j$NuE<T59H+q6QzkIP)-DmVS|
zjjr!ZE_0P_mc8-Wb^ko7^d(j8m-07#&oZ_6cl2%1_x!`>6uC;>_f=Dkw_dx2<Hs-G
zp%PnC?VLMY%yV*;rr!bM=BLsZ>U_#qKC#XbRXlz9bDx*n{(e^MbQE9H`$N9tQ}&vM
z>wecEwRo=gMPjE9^?g{=_Tc+H3U2+O_q%-MC+-u7`l8<RP<^LDcl;6iWw}b-a;6$j
z)U&R<z3o{acANNNnX9zDHxWBs&+U)A@0l($%3S3i7j)QVm8(80xukA;RmJ1FHyQd~
zv?@QY^2FA|{ag;M&vZSUYJ6Ywi%@!Fw`1G<UDv~^=M?w(qp3db^U7HIOv~*WpX;8Q
zs~xe^tL~-d;vbuzYyMpAFqR(GxTx}1%^Q9B>)sR2EuLz=>gw-W525#aUG1gPH}+Zc
zu7`G?Ttfad2VdVEsdc@geVWc+L)~W*_Q^8TzMPydx7+kSxov!IyG`|-b?DMFzj^7~
zB3CKymxYq1d_%=gb>D-s%XECB;&bu#wI1fOkJ3YNLijcZJ+{+Fr6~u&-S_<>s^6p@
zF7}B9FCjy}C#A|y*b_?ny4M7~KmEHOxvwKapX2-P55<T94<ESEH-)RD>{RVTrS9)*
z-cc)hsdix+&i$Ug>b%a>F5CFIKQD6qy*6QYuHW4ec2oD)QV$RPJeBja?YF6>GyP7S
zZG5Wyq4qsRK9oHoJhl7PEA(vh+cO5Ys8DjnRY&f#>Srx0)gn6|H)Pp;uW4J!MSd~5
zYwz<1wy)$WSK4f!?tlNqa+O@=1=qf{{QmE+P=WNzL5!23r0w&EJDux$D!#tQXR24O
zx6tcr)BRi1_}uHiRP>2?OfE9?cYEY^9op?{9Tn<)Ud~6XA6=!W2k!5=ht?<afjiFc
z*{jZnQuP<@-<Rys;l2Y$zwl*|o7DZ@Ypm~OihQ@<{K83t?wnOryNva{1>1acz5k@L
zJGcHT+iQEDE|qklp69yH0oU`^Q0Xad@_xE->vAxS$F*KF&6g{_>pnv&=|w%~DwqB0
zl1kT(bBkQ1>Ap~^@l$EfrgqrQYi!fmjz84+ooaaQ_acP-TzpOUNqxr;{hp(3`=`>W
z?*DN6+rBUFD;>p!3g`OWr&xNZ^67@>_P(lbdj&Vs-@)J-Cu8f0eZN`o_a#GpPtR9A
z;ysAmrTC6rEa|I#QF_I_&0J-!?Jo8Ao2uPD-~F8Gce#bzo)?BLr|G^!Zt)lVcK_YC
zUui~>y5!<&dcSzeQ<wLD{H%pV?$UN&i%MtOAF1!|nCdf)KW?Sp4PEc+xrHxjYL6UW
zKG)NA%X)mdxz7uI*_pcD(Dy!L<(KMvl*5-Fyy0ElW|fd>a&Y}#pzrjlwrf+ne1E^v
zbsXm2=MSxqn(yd_yK;=3-gkfG(r?;-Qj7o3t!^K9%}%q+T9c;jTlf4!)@!?O5h@+`
z{9L6kmG@vw=e2TuR+~H4c`aAIxt<#cyM6Bi$I`Dl7jf-BvC4yduOGR3{p)4!98!4>
zx!R$-4v^#dzDGvs4}H(fHb1WQfzabR?uAYt>V1UJ`7Cw7p{Lz{<?JF?dGqHZ2F-YH
zX2C7F&dVr0;vOxxr25DBo%dX{%4zi#O5VL}pSz!0rLICr>U^)~qodbq+~w<nlDfa&
zPPJExyBw~|pL%bbTBlz0-LreYw82w_uGGCJCg#s7Nv$tb^HW`aY92)0%jCM>LbdzQ
z;il_r+xWiA<@z0k(CJNiQ1dhq9{P7@Lgy#S$tC4_u>H=6@AvRzJGtGu<7h7NsOL$w
zg>$zDS)cB6KGk?~dAQu6$3O1;OrPgm+=L!#-bIBAJ-SkNUW<5U>ws$K6?%Rs^3!eC
zrtwy~;jxZ``z|QB<QbbCvHcwj>nfpTsNcD=&4*f#bA_wUajVT=-WL&W?&l)6KDmF&
z`~0EYvK^G$c0WFqbYeeWSE_hW`$xXQxz@S0riYr>s_Hp4e`>!rmvG(h5vCfC>p2ZQ
zzvwt~#nFX|gR1}ey4S19PvxJ=b1hXnRNQ2L*WZhj^}$`61VW7qzP{h4d!DQtPxm{z
zzFJEz-}BSunR>kYw%zjR4in}TwI+SFUtPVv#vN0=q4o`Y=i~c5HKAX(eof=q-VZg+
z$9H_!d8Mg+zT1^={nYtXF5z6~wx;#Z9iQu-maNBgT+SuF@BRV(6Hn{Y?&Ff{JVEX+
z;{3x_Qumfr`!tllY^TaCp39JHv>9>f%(B`zsOsI-Uw560cy7P1V_Q$Q=V7+_rIOBd
zUo+SAq0Vi5)id|`sO@^tHl6FbEtT}S&lkDod8z90T|aXF_VqibuHW^S?c{RjvR^>i
zsd~>Vb$@a1GFQp<9M0uE57|ztU2=tU_4jJi6Kfv9HGea$54rtOZm#yHt4A)6@8`E~
z{kq@vN;RHty@iS=w-4D}E#H~5-B*)&n7X-sFH@z*)E=(*)cxAn;kx&|Rq<S(W1-7$
zyKfLXo$u!b^e;YtQgMh~|GDPRHIAC<vE5gv(xp0IGPR33m(&f9y-(rVFHGB2wZwOQ
zZj1EQ!oycNxb|gz*$2HK00JOD5um;c$Q4f6f$$#^sFv@ExW1qJLn=RZQS<s#!sYd+
zYClW%=bC45>G5?g?>gVlExp_xOx?Mam$LWAJv)w%eD6IMw^+C3Dnr$FY(BdC;_`DK
zmOE6vnC6p9zohI8b<T-=_By!74)2sQ2d;ThDtf5rEO)qRzg5N89T#(trz*d#zv}!p
z7k{qysLQWf?sNDmhb~7sUhH|5$X~Y0*6n(qF4y;1xa{_|PXN7_ZZ+|gsmAZq>h^=w
ze(%-Po>ay~>K-f=9(v!>bpMHpFUI3sCD*(pmLAvVSM2n;wi{vh@Z|?@cvrVsWv<fp
zzA%+e_c@e%Jk$PTT3%{jNrao~J#fe#eRkhvN)cLC%Xz0<j@0L?p`TCk{#&YUYJQeW
zxa~boxgEo8_IohYcZGa~n~wLX#!qEG-PI1)_v}+k&$T}kOHb%^wQYW`&qdRHlT_p9
zx_+t_yTy7|eopBA9<%RwvR$TbY9EUVH_cCuzv@ccZ_s+br7N{AWiF|?Q(woa?4#Om
z>fy2OJ#pvDHIJkEMKNN)!v}8kO%YmBcG-rf{`&)zeLWu?y;kEcUl)*ZIOP8QQ>uOC
z5`OdNBL>ZQZ)TC3Ol=&^<-6*XovM8m@thd9((i_@_x0RDmZ9rIm0vFXPGzU>-^CU6
zn2U6spHb(YRCw<9T0@^BapmW0A2Rg!GjF_c{xR()e^KNrbDO8t+V72pDzDgYG?lrJ
zpR)bcvVY874tYM2`nbXMyHb=MnM<|0#l5D`G8Rv29hG``?0Yh<^+xr(+`?nmhiU${
z=jW-V3q4+O<>zaC6iTn_J&IaM?>bJ3^rrWoQ;E-g-bd+Eo!8d7zv;TqHNI%Kv8C(&
zWA5o=^GE)*LBMsq&`mGX3_Y&P`K#Qz`{>m3Ob$1l=Tq+&sO*P-+5}EL_|JQuz3x|K
zXeqD%xYiq~JZICcbw2Clai!MWd?f8O*}gxfO2@S?VXMd2bC~;kOQGJ!P~F4f^5?!^
zELw5s^L6h0bbp5<m3Un?I$^-l>pxZs{wgV-V{*Gy`PPa**Z9TNzlEM!doIT=uj@QC
zmHwg{SAE+TyZ&tRua@%&uJu?fJ!0I;RqF0zr5caxIT%Y%sCygS<0y5HxmLsbZT%n8
zO}(1BH*V0dOhfyI`b(P*YHDcl{olrpjVm;++)&rMsj<`d?u{Di8|wF+e|(>&#{QS@
zI=X0Cx5awTmkWEnQ~H)$ZTtD7BX9eFoXG1ooYwRHaQ4>|zbQgVVL#;JdNk#rTd%(2
z4Q>7EL%*#XU8tn5d_(E&vD=PAE_i)PsVhU-rOM~BLrxfU=&UIPmsI6bt^VWpdt%M&
zCwx)p%4+3+a_JHHcT<}i&Yu3qB9;_Z+RiiXoO0|N_df7m3F*c`)I;i0RnM{7udjSn
z^O4;Bo!eKxK?Co5W_np?DxEW4vI5ijE_Zxk)k&|M@^s4`nhRU$s~rpbxFpxSP}aj`
zAN2e*fnWdW-$PFr^<_!PpVpF3=%y`>9JBx0Woat&DWM-rkNEArH(osJgCdsnQ>yq{
zZRgn{UHlUK&;ofJhnsi*#Fpm|Yw3Tk{TFV!-AU7@6~5*Fa<uzzx5syWwM&Pg1tsmY
z*VS>)19kP;+AGpgJaof_K2xcyKlOOJ^_6No+jT~2>1@ZhRMLg!Y1&_P<8$%U%@6TF
zASD92=QXN$u4VDH|C5_u6~~qD`trVurcN(%k-F!-sl`imzQbjg={}HYd{z6^^-q1?
zoVtCwoPEdhJ?{FpU!}it<5j)ZdS!kYOP(|HjYaL=`lf`m<DjZnU;Zns-nhqw@60J<
zN#k=$({oKv@m1$DzUnEIUTXfPvSa51ZfiAZ+Uz2AdHZ^G<%v`06uC;OessfC^(gzR
z)>Ena+wPmo<yN_OIP=nbt{b(m@GZIadve#Ks)tbix!)&({Xczw-8Y}q*JHZ&`Ccz^
z>(?z;>hT8bxzrVNhR-ftWvKBI`J_Q0_ve1@_NP9s=-OvHpBCwS_XctA%2cZCYJB^p
ztuLA|ugF!Vx-OILQoW~;yFd3lM%Jflx2gF19=B5|XDW6~czKg&`n^7{gpA7}*8aKc
zIg9JM0_l0yXRrRMcbDeJz9@5*rsp{F{L0sR3}PNyEi%?~XYlRo&sb*c{G!@rEWWP!
z`szQFURS$K(}%v#8!JCkE~@yp_N(-{`pfGXSGUTZ%|BbXYrjF$N|jWlqx^lJ`-Rf)
zdz}_rKXng5)U)Znk%(`*Phs01<#eX^9_9Fy{i=Q`+o9`L>ErU3&+&BaGL5Huzaq7G
z)N@=nT=W~r#RSB7P;HW2&;6$X$o)=s>z*^1##24lQ}uV<r{PNfqkc==kEHe^sC_S6
zyYD&qi_PvIw6N5bzV18Q+D~x_6;3^0f7BifJ-Eg_-S=!v-!m~Rj<@Ks$0ctC3#{0*
z{$s`W{e4~eJeA8GxH{RgGo8n$7N1(L)>=4Q!?0B3>%73!9O}72&eMtcoo;&W=d;W~
zSHJIgv7Q&M`BLp9mvGg4Ri^%O`wHcj_t|oBb3G4jIh(Hk5nu27Td8oeZ5K!vh5+|C
z6~^lS<>LDM44wXemM{K~%HQ|A$5g*;e`t3q<C?7<@;-TPZfajK^>D8FPHa6~`##WP
z_Pa2Gm+L-3=y6k(pRe<i*m~vr$*S^m`*WR(Ku?(8*-MRcYWGKc*(1i$ROGO8ciQLB
z>0cJUCHHuvtIzgcp00jh@lu;la+T9`+~LYMH$ArNEberva|?B^L^mJbylls_Sn0Ui
zV=R5K_=$X7+m)~ROfGs|`K6ZLbpJATeD3}B*!rk@qPpR++Lg?Y+YZxtUaIkP+oz`N
zjlF+c>-kgTg6=$16;Iy(i{;k6XJs9ad%!Z(XF6UPs`nn9I)60{SG_kteWyj$jwyc`
zkn6dGX?kj$O*LGsCsb1A>f0S_J;$AI==v7*VtP+d#HaeDZn&@m@`nUcSr`A%Dsvat
ze9b)PpH5*qUL$^(-&xIt^G~-f|Dn3)JJkGG_1w{Yzg;yCi0w~}JFe|amyc<@P~$##
z3zheSY~9p%HmGp!a@pz=^Q2s5=>1;PeiA#rZM)!3XWP%X(uwj=>utH;iszEAc1G=!
zr5^5!i>+Q!&&f;E_ctRIh?<|I8gAQusC05Q*p5%!_s~LpUryw+*We?&{d#oU3YYxg
z>-jf5J8tQUN?vfyogGF^tE*7*k+VmgxY|ec6-utU-wqqxHMwQQR8Br-={Js^+M?o1
z3cjHw^}Bg;yZ7z3?N_RF!d}($)b{*H=(Uxm=PS1HP3=>ibJU8z>A6to_^Ng+{7vio
zrdgZ6zsInpn{B0>pX%PX?tYMNJk`FC?4Q~^#kL=-(&_RN@wm^0WPJ@g^!V%aiLENN
zF6Dlqa_fE%k1L+aj!^#cIy*PFsvo-g+w!-azr;!>_EoH<Eht4gs{L8pNv-o0_Qdjz
zy$>9#+_~nP>i3^Q&xg74Gj0Fk{7|*-)b;n({(PU)`LbK+mFJakhxtB~?YXm9U-*)`
z{mC_+Xn(0lw0paD?^=JZ^ZC@%yYkUZui6jN^{4VlowKfezQ)Z^dgXmL-)?b#mpa!M
z;jZglvA?P-sdlRjr`C;`g!|ferPk}V;imnBYu!QV(H%cTJn9^q3Rkwv!~#TlDM|5_
z*L|^zN!3s4;imma72mWT<a%^<i|0nQNmH(-@nhe6ME?8Uc=4<c3c<m&{^a=7_^Aq)
z^}r1RF$vi2Yt>r1n02(Iyj}P6kp9Rq`>$P=;IL#4x%O|>?~yI351IIxo+}F7rs;Bz
zFUl=T+x>@F>1@vz<Z{E!OF-;z%ChA)x4wGDs8-FoQvP1Iuu^wFROBzxalgNyt1q>2
z+P2@Mny&Hfm$trW!n`6^S!=(i7`wcx`A07Py5kOaJTZ>iN--XXlD^JyZ1sj($Jo|S
zsC2&0S1J2L@eiGk>Kr0ff8Xl`svYQ_gR9~x?^RP72MSl&or}McZzlH#Wc#Ue-c-YV
z_tVhhq{vUMA8j{t|Mkt+R6ld^l<LI&^J<a0_W@nwsep2?%T3!&s_{duGfeZ*eGcpH
zbEO(jdt6QQxo*48OD$b$`;WTz_<BC5#!=ZnmHTS4U8e3-$7jke)%~1Y{Bu7i)O}7>
zd%RfdGpc{29-dl1H61tE;wLu0F@3JN-bdz6pZj<d`nlk{AErJI$@aPOU>g@CMPh#{
zwlwXpvDcpx#veIz8852seGOH4Oxu@;|D%d!@`!yOK{tP<;s#UWI?ht*se+K{ndW7w
z?f1y`P@GZ==ibj&=@aMUs`7K)QxN%ZOO@Ti-}St~*Zv!&SJ*F0({)R#@pb2Sa`}C^
zb?rgCCBKIua|yj)<f~nzuGjZET3$E$a&P<Q#vk4D?y8ks<=IbM{pb~=+Er>zs``g`
ze>oSadmfi+JbAvGn>+P(qH5<Vf75ni`@Ki1o>g{G^UqYnxz;15dQ|mn>aV-ckxD%3
z_eEU82j9N_jAh2oFHAYw#?SS4Fk<a5t2nF16>7aLkMq<zftqjH-mj9^E4psVo>aqa
zIi;2^mvanZr+0p;4hsLM1YDnsvERe=1uxh4lSA1}JwIcGi}v7azloX$gbEksgaiQ-
z0^I$Q`@MCQKDmslbHZx%?|fv7uDdQhvapq=`v|IaldHeHuFA!2yYH1tI$!%}qTimk
z&P@}y_@H$qH|hHOf&<zQ{?(Zima1?`TYjqFJx<-9>-V2y=^20d%#&6b(Y}(ayuW|h
zaoEUZDxoCzx#Wty)_(JM!<Mb2B5gUE?gy*xb5^Ur_x&rT`-kFwY_2l*_G8+QQj0Ht
zXE8Uo*l&&{cf4ZEfk#iMYc`eC`E{z{y5~jQ@l5Z}Qun{BEnKXNbS2mNESL2*WoIty
zYS~V%dW4=61|2$UN)elbYQNI-J2SfRxyNZ;eWv?|s`%BiKdUMqmp?WBQQ_2l#Ps>h
z6<_yxC&!E3&qL)O`#l($yD#^u`|Ys7U6Wf@a+S9Gu%_vJ@mJ}^zC`057rry6jJ3~^
zD}UK;U-vIu^{U#vtG};#O{n`MzVmVA>zdwI{p70Gl_z)lRPQ4TySV1RspxUN$05>l
zw}Vjnxaaes><*O=wJ+d%Kh{?{sdBl7Q|t27!nyV<V(UqbtFAq%JP&N`aP2Rtq?gw(
zvCb!L^A+}odjC#$zuGk3Jtu#$+5LkSmb$Y2t_Oay?F)-aZh7Jv?H;;wTz#{t^!41b
z)vG$6$i-jGGjo;HK78uosf^F2cGy18<aD<E&UZTQ_c+A<kBYO*MdendH}#kKQTr^e
z^H0<9#WX+N^``E;QI1FPk;6^x*PWNhcA2_mJ5_F7o}uHZ>`mR@w10~DazEvEo3@|S
z<IDGfa&vP(cVg+wW&SJd3_Xr<=a*YMSJ^A-gIl_OhapsY?(tpLC(c_^k+$c}y7MKu
z{IT7s&r_=1J~6N2l2m)jooDR(a<!hnY9GLtKh>V8@L1~u+x}dw?aNpD2&LC_|D;;t
ztNNGe{#mH_Dtm4HbFW9!{$V<9iE~I@DfXRhCHFeP)DF{oFsAyc_|*9kwSI~9-n?qw
zY1-bX_@?zk>661v`;&-I)td-6-G^}<pH%gdi@$66k^a?(ep@%XaOGgyZez!{oe#LC
zqsDjJaPIv{U460k^L6ueZP%gF$6BXy<*V91hyR!NT{Lxi;hTeQf7OlWI?nk{&(&^i
z^~9P-o8GTP{7s(e_xik2mjm@4D0jQIZGWky6YG~+lk)mdxOL|z-0`^Er_d+XvAIjz
zej_-fCPjP8Rp#EFg#EVbHaQ(v`{LdwwbjQ}9$8Pdo|94Ka_#4o{?x<ec3|t)J+Bb)
zV$Jude0A>~QSm5_uI)~hp2|13aO(WUG@ROhqUJNE=dP-H^5rj{pT1@6{brHBYdL+@
zgRWk#b0MnT#tP>iZ>ar*+``54Fjr}MKAc*7>N$?^-d&m>`=a#4ftruy67GARFRyE>
z-L3oll;d$fZ&dnB=bb7&y8fw+o0L7a`x{7i>>Kw!@LuW4!L)tq#@B7Hs(7yF87jTr
zo)gLp!v^@CAL^EdYrPU$kM8$yL_GPqqU+B6xlHYkxQ6TAm+&3W_x+Jj`o%s->e6)I
zEcN)l+owDpbDwJqeR4jb-MZsyYVln69aBl~I?oj8sr3MNxGg`G9b)}1$|FnN_D9V_
zReDr*34hmlGgo@o@j;X)b?IxrE0kW*o^z9Ozw_;GJ?{@^e?9S=!d2S#i&W=rt~})Q
zwtTqL?R>y(ttL&IT|h48z|9x>KtP8;==GDZNvwOgq-s9L?Jw*XGW7nL@BCc%m13nA
z^`VlfKHqKaa^);I+2*gCPUJ($#<ySE`l1Q*idvO2pHTaj)sm0Q&(zKJ9HR74c0F?T
zs1sNFsJ;?gQg-GRZoBW|tDLd!0aE$qJ|3j9p7hmjxb%kleHCB%_}-6_^>cBj^ytp3
zQjO<2FHI%A?{${YPqhbM;kM^CR65`7lFHvT+!t?IuROlw;?}*NmTEleTrih#TfWr2
zC^0VOE~$BL>fyQ0!+nqcvVM776>d3Ru5R1sg5cm=#%ib3y)@N&pE?hq#s%LzbJcI#
zzfvnN)jo8?Lzj!%ccj8q{e|)mwO$L&GrRoy4H|ggGt<jx^VjTv!C(;hF#&3v%q^T6
zzq!JHOw!CJ);y9sUuyn{@O^K*c-99+wQ-PJJ6FHr(j(jF?2bL(V9Gz2d)c=808~1z
z_C@J&Zf}kXDEp{z$LyVxQ~S25hr6E3#7=L!zmZBhs(mB;l&4$n&|HASk~vWOt-9f^
z17WE5wtVG7J&#il4>e!1Z6~4D6~ew$p1Z=HR3!Jhx>odY)t~NooJ+h={W?@WsdEp-
z!FN8s&ZoHgDfBERpnI<BD<0}03;|#Lof=Q6(uH29Q}(!q*UEee<qY#Xe7-olwsTc_
z-S0HouHQ`eBW>69ruTd+H(u3itykukwIY4()2Q^S+MCLsd%hy_F@0Xey3Z}_wUwsj
z)QxXz57L1E2-MZtKF?B1$2~rW)|VPjU3;S4Q)Rk0B7Zj#V?rDPy3c#x@#2)eu1;=;
zvD~`l=l&jmZob3zKB>nRqgys}N!@&H<C)$ww#_H?bhh886XlY>f4Re%m)>*TsD*{D
zl+#y>+xK{=t6$VpYSMInC$;#v$F$^oOoEGTdy8Gqx^~Cf@8Qaqi*v1hPY~s<?e9sV
z{=?S4?zk++Q=K#9>Yv;1{+=`QjYaL=`ld`pn(m8nw-eDmechYl*6V9MCOFti*ZU~8
z`T9DiPgSoeC)0A6?!)ND4|R_|c0RuDdxh3(+Rmu>6JFlrnSQU&D`QFSec0IgxSxZe
z^~IWh=<d_m#-pBtzQU>gC&IUR<i$BFztygpTMGR`+VYE)?)LTS$`hx~DNLm(Z>lo)
zb+NDSn{(9{_x#?sKGXSNwZ#|pKKS<aXDl;ze&H%z`NZy@w&#thq%$3_Q;YAr{zZF-
zTrvUI{!67_a<1`xb?>dE8ZY#{L$3U6-@mX;XWP<z)w^sD_jt@bp33(6o}cOJm+jJZ
z$68<c%2#*)G1Yk7?O(SYn8p+L9Bl1X<!Ac7mmFXBy*Jx<w)=eI-c2r&>-iQ-k8jSZ
z_DlKu+V2acSND08YCK=_m|D|IwR7%pYF(9jxM~07=BLsZYkcO)H&(r?c>3};Z9i0e
z?&l*{eNuWTzErtGhg0^NhI8EuhMqPzoIU-Ig)Rr%{T9=7rtKk>eRo|uY}bpar4#q=
zM14{F394~Ncb@EPza*4iv0sp@6!oktW6evr^3@#|5zp*7Lh<6h*QeWFRq?2CO*cH1
ze$PGcg}$GD&rrpO>Ib=mtIC&)KlgYON}ns=6VGV((4FJzn`L>=$zN=C|Dc6sEGeI7
z<mwK+AIkOJk6d{RyQ%$?)WdbxL#f6KwVv{ok7}K&@>k6pefftTS3>3It3A8cgX#X2
zYkXIGsq~@dC%NaNyPuazJk|J=n!jkDrt6zf&uw9!EPcIqA?ua%<#zjeZpwPO?Kahy
zO24$VL+r1IlBRq^#lLIpw?`cC=G-zj$^AYj7jIwVD3^Vdo>22qU-{hn`>yS$k7?1I
zn*1+!rtjmbzGK3*PZ3LxXczzG^`C#_=Xfl)Y?p7h>v|`4dQqRbNwMFRs|>w9^xY30
zIeXNJt9?{oag~&vxwX@3v)A?9#Wp?FE=<F@pHK4oPq<C(vW;)LpO9L7+jfyky3R+o
z=(_9DBa2+6ul?sxdSi39?YF6>)4kW@8jmV}=yeWdj|h+bT{$^_T~5dJSoid82Yy~w
ziwympGUV60OY>u2l(}-qb-kFI-C~>!B}2_KsC;zC;nd^F^%i;^E7niGWGd?hZaZ@O
z{y5d2xWnak9op^seK}Qn-QRQKj%RzmE2k6nklMb4sXd{`N$&h&pGVYc{-XVRlb+8t
zRp*_d{A0J9Sl=-Zoo{aIkX+ZPD!W7VCtvxPK2KAN&;7hr>C=4<sM4qAPmRyY_NEKM
zFdi2l!))22r`9iQ?a6JuT5a~)eg`qNbgJiN>i(hUQ?>(8=y@hrew7=q>b2G@^UK_2
z`&|$GWZM@Om9A3WcZux|y$^2N{;71KzHh<3U+k)n;zEUoem{%*-X&E&-Egk&_Ed`=
z!Oisf=-QrBNl&epe1{AEzNByaee0L+-@v`&kPBX)Qkrt`)xIdbsjXj3?J*q}Q;F}p
zU;5S`dcMk?AJ@K@N)OfVatpWZH=)aEYiBO$Qkj=gcG&KN%IQq+y_i0CQ;na>``J{#
z@ZG<CuZvynr}}{^+%;d_^xXV(^R2dcrt{KV<E!Q~vHe5sEBeZZ8b@;vH+@crj-Tpt
z-qbGD^MuRab=;0UPNb5*E63RBP1|Lx_{#e-ResMi`*IsLaNYN7Wwr8pGBr2%{9L6k
z^z&AgpDBO2KC8{m_1uo7$Fx3Eiyw=d?|fD37}x%z8eekt=e`F4ef!>c@vILD6$fAM
zv4+;GdcINqvG1AXaxX^Jer)|+$Gy<$bKehA+0T8B8(N>QbD2=<XJ7L#s(<M2%jn))
zP|a_Iziv52d{=2YpVPIE;zqScQ@eEI>)P)t9>v8qUy7>;r`mn!@Bw=+b;X?FvrFI7
zcez~qW9amzJXFs|*`I3XxrcK<zvX(MzHh3U5Bpxv%64+Qxw*#H=Q^IpN^d(3akmFq
zpY1v-m2`4>xZJw)pH$<Swl^+rLXT=Z)%6#8bfxZkA+>n1$5Yq*L#<b<Eg#)>Z5)sJ
zygSzYwbb8J5B2<W?Vq`%SG6;0A3b)s+&^Qj!(=-sx7fGHRhrJ@RXlEAudY0C>YO69
z<o-Q4N?&U89$R~;`L?Q_Q}a*#`KD{1yib*zJJxtzYx%1B-#sUPvDy8D7M3cx;}v5L
zJbFT1v#C_&r}8&lFV|Xp6*t-6bv_~MgS$2f*v=2AbTUm}>m%;<q3*pVxqMu1)A@F+
z_|-BlbGM&d^l`OUTRo|8674wmc_Ou5;|jN}Zz>)2JW+*HdP9f1wrf*+Rlm!Ti+}9<
zxm@{k?Q5CVKX-iY-)Ry0bmvuwcfz1UXH6-<;fEYt$5oMD?k}$2wN>q>2!Gq}|H^i%
z++zL$x$~=CIt(qUjf1M*UHz+#Q>x{$t*61Kc3bwNSC?%zmA3h%l8$SglA9ja{aoGj
zzUrBKe=qcUJk~xMcfPr`d)eMl_e`jJ>b}B#w`;k7>)r!3jVI>Qwo<l}%gsFw#nPAR
ze1@`%Yd<BH9@BoHiXY2<*L;1Ad$L}x{$6c*RQEG|`*Yp<HmwiUzP`#|_r4TYJh?o&
z-^EHb9@jo~EIsl#65B1e5817HKIG<ayB{s{Fm?NGXQuX0@pb#X?{c~7SB<~A{;uoV
zmfPI=>KUV2HB)lP_@ldC@lKm&SIPCef2Mkd?R`>@Ek?I&wv`kgD!j4Hq6_OrEY)l(
zw_jt(y3dYm-OMF#{e9Q=)5o-E&Q0e2yR6Z^Yf25Ip2Azt`@`8^PyD9fmb&A9>hZYR
zLvHUU$o6__e(9^_dapJ8l18=Ac0Mhqi&_X|4gw%h(*#1DkJL0|i^A)!(^HS<x{pAi
zlh2=1bw7OcDd%1`euZWsb@#XBd2MXB?mZRLc&WGJ)a>J0=LtQg^~D{Z>gS=tWA`uH
z{JGBmV(AH0zp?Ytt=~}hsHl8m^HSxX+IVbhPpJ8~uY7ddbE@%l=V7_T^Ht9(y{i6c
z>Q9YB)e>&Xg^Hi*dPmnTuJ1-uoJ2VHyj549uY1F~`Rm3LoJ?h`d83@KsokphuJ<sx
z&nI)!C!cfV;?C_}7iF(#$EtQN`@8PfRcm^g6Xka0V7s2MO~?Iw(tX~f8qc-<Vy91a
zy`^iHZ97gSUF`aI&0n5Zak;5+vs%JU->Wk%x9)t(R4?~)Kv$nQhp0A5eQ&zf!cF@{
ztoW|uXDaD!_g8es7sUJFca8)PYF*|l+;qH8HGb;%tZUUiv7eButd?^GsvKe*f?V?W
z?3)W!Ke>d9??1}ViB#q{vR$Tb*Kyo5KRJG?`(L5$lDVYj&h>fc+P;zgi8ZgE@I{dd
z2dZDJGO5ely0L94p(SOPZFnx<!B*MXutSf(PM_GSlB+y#otq|Z@j>fKu5$7*OTTgS
z)E1Ro<*NJbu)$rETUK(HXFqZEqgRY-SAqO#4%9dq`+GTD`K#`;i1MrETdv<l;Y!cN
zA(kE$e=dJfkGV*$a@*?hwI3WxFZEr?S_{8<pHI54Jz`N2N`|UuU-_hRZ==@iNM#>S
z*kOAf&h;IuRP?Ccx5(Y!bzf1QPw2j95G$VVa|WSb=2EThSo4TjJVpKHD!JBMwt8Ia
z&2_v=Eq$#1W1Fv-7vwH&%O5*ktowId`R3kkLfP*+PX0GN*Y^hl>k0jS_kcZ@x?;}o
z*(I0s{rdu;^oRZ)?|<8E<bQHs*SXps89%bG5dyKEr=j;BxU19mcipM?0=dhRi$34n
zed~AqeP-MA^81Qh>xpFEPs^nNJKCAcJ!xNdI^uIqZhM}cYC2!{Ex6|uq4s^`agy7u
z`~I(PJkg+a<EI)g^nQbDe*b;$i3(6iU*n+AtH`!yyt&31*Y+T%=N>mw)tAb7j_W;@
zR6306Ip^Cx)89)=C4Sv&eQy~0cH8FMWh(b<Z0)dJmm=MPpKSZWqO#UG?0ChP1CO3i
z*X&zzkE6EV6RGxe{kHxO>84&y-5WP(Xy34GL;a;q2Q@Xc`2KHW$Ho;JS8k~5-PG9W
zd-p~S^$qp=&Og3SQ)B<jcO6}{<gO>^0|5{K0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!XP!Q<1^?w4IdNp-#+@N9Eh7Jw&mo^>L)X?Jlzl|LmS7=<hp{{pR
zW2f)k8#UB7)bBh0_&!aI{V(5jbkS0w7TzEL0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JP82Z7zs`t-#Q|9MjhnFohF
zS{1AW0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!^0-gWS
zrtg4;@g*cU=(K~MFKD=N+m`Q?xye6n@ZmKNo;k5}m3qAju|Qxkfd{{Py}t3y2MS&J
z-aof()9LU@1tp((?b1=_wfL-nG~<AF2?8Jh0=W`6W7AD`X?5R-GIg0NU)T)-AOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAW(e-9^CPT>(1;xxvVxhvhCT+-M0MHQpoD7Ig|(l
zKmY_l00cmwRtapm{ww#-KjgzQHF?9<4J{_On^Ky}OLpnIaNox!mqMmeua|GK$;hAg
z8dpNb<&a8yfE^$J0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JP8Gl88Sy?T26oEZhT++p<xdd}<kQ4ve7KWfM2ZXf?{k(=DRQ{SD>So`r(
zSLWRIV0-=q<_u`Mw9hT$O5d{URZmPDyz=NWRhfTH`YA!vgO|1F*m`)8tNi-HGmcvB
zkT*(h>4!G7f`A@@`<Lss^T)TWScxU|tVApj009sH0T2KI5CDN%C(x<siH#@yZEBgj
ze5>0DtvdCeT0-h^sP%S_^79h-r1u?boHKAtnY%n@`7U)2o;j(smAq7;4Fo^{1V8`;
zKmY_l00clF1c7I+T4&8m_MBWo))t2lZ3}sT00@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2>2&(_Ki2+vdlH_
zm$}P-E?D8V9)FosQqq*@a&vw^;P9VKElp+a-=C;;`%T*k;)4JPfB*=900@8p2!H?x
zfB*=900@8p2rMShb<?N5zT~xurKvpakJn!I%<m?Zlw9-l?GEj+a8e1GHV5<v5C8!X
z009sH0T2KI5C8!X00BP)&KYn>m-g38C?UZC?Ew8C^>M(@C<v_}00JNY0w4eaAOHd&
z00JNY0wBOk;F4YXF5LIA$t9#62j1F%HV^;-5C8!Xh(q9-7j}Gcjb0;5Q+ZvFQEzQ{
z*!WUc_BpX-=lvg>ScZ~uXd+h-009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0b2t1
z{Nu$}p1-C|MI{gIbLDdzpS4sal)P>5O|y5~U_qghm!AFP+M66Px9}xxIU*ehfB*=9
z00`trpyx>oe(|?Iyk0`)#UaPG2Ae?u1V8`;KmY_l00d|PxBsQ%U0d(`LCGcO-f`~u
zUG|$;+R78RJpYn~_kK`9f&<z?_<lf-7-lh^8`!jS?>#4$Dru$-u|WU?KmY_(1U9_p
ziG81F7+P}4Tb^FF#S_cDT-K^oapv})u;9_b8(cE6)RkR-w&=J^UYk@xRtE>}_5^((
zkRpL=Px$L`hu<)@G?k((>l}T<6)m59uCSFqn|o#bxII2AbY=P7{`arF{g*|xO2Jhn
z2cNS_myU;jQE*FDe((nY5C8!X009sH0T2KI5C8!X009sH0T2Lz5CpdV<IKYs9x}Ov
z1P8PObw3Ev7m)`DfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2-px<Va?%h|L$j#%T%RJnUDqqKmY_x2wdN)GGygXClt2w
z=TBZV=b??;R=8y4sGql6aX?+gR!;uo$%p>>qi+h|(u8q{xZUpU@bPVZCzp`m5TYG~
zXwxa>@r#QezW(RyzgR%#%^{^W1zSJ>1V8`;KmY_l00is_bl-E}o3maXTiQzdoRJa)
zKmY_l00ck)1V8`;KmY_l00ck)1Qrv}{e7GnH(l2Esc}<FT{&^j_1fP5_0*D*s0R=L
z0T2KI5C8!X;3lyC<LABE`pKsXZpm#Y^wkD|+cx;Wv0V-uQPzsAjao%1K>!472=t$J
zSce{enOx>3ZOV!?AOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaY69ICoZj<+Gbfj(QfPeX&NeMtpYcgi?NXE%5(Gd11V8`;KmY_l00ck)
z1V8`;KmY_l00ck)1V8`;Kp;GUFPC}n&?(oxTtI>Y+5!3j#(|$62f~lg$PolU00ck)
z1V8`;KmY{v3C#Vx-0relMwi@D-vY!00T2KI5U6DW^H=-su*3d-P!USjGHaB*dI*f`
z+<U2$hL0^x<=z*re!<@#m{^9E)k7X70Ra#I0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!Xh(Vz3E(hK*;?AihCEx0HLaR>wr<SS97z~jo2!H?xfB*=900`(1IPH(u
zUiQrICY7OO-Gy`RUhS7}m(ipI=alu29rM)hrk1rLrQ*X00w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&00JNY0$vD=A6|K2{Yxg7q2z-*zHr@{-6xm2^6&Q!Sow$>
zCY8RW7r0Og0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHdqf$m52Jbd=cV@gvgseuy&KmY_l00ck)1VEt12`p%F*S)VcPbn$+
z_6h4Yb?QH*G?g`u7sUqw5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5J-T)_-XUHZg9z@(pIi@=@%dNSU9n)b(ug_
z!UPZi0T2KI5C8!X009sH0T75vpvTUC-SLK@&zHH&R$Wfqyx&D*N>>@RGLSh4fB*=9
zK-Cf$+w&j89y+0IMOSV<aM-Rb-)L2V%#*|XvsP=^;H8I4NN_+q;Ohrf`+z{fKmY_l
z00ck)1V8`;KmY_l00fdIaMwQl|9ROxWnnAj!cpJJMfj4xCoFS>8w5ZA1V8`;KmY_l
z00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY`E3A{G=?M{8WOesy}`U{5d
z`o*bJN>h32j%)OI`gc=G$RBdhtuMp_0TAFL@a%D`zq~<*>kCS*Gj`gltE~TKp(`(&
zIcE8{CeA2y<<jk*eE6h(^9z;aD={>I00@9U-UR0Ccm1u!ep5?Ij=g8dmFrzHwS@G`
zA@9}%>p>t-0z0mCVw*$We!cW1uf2A)8}HfV#nM*hNflOu00@8p2!H?xfB*<k1p00L
zpBPQOnz}b`(9ofwV?+I=O$RkKwD|sSW5>o78dq+p>)q7Y>3jD^4fPH6`_4bUPg7(6
z%Xb}Jw50SS90WiB1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;
zKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l
z00ck)1V8`;KmY_l00cl_2?Vw{>xo})cgeH@5*&2e!6E0}weY2*-!5{K7d+B5X3@Xj
zD_!Lhx)8iTKuO?VPxQX_s*64^bmf`TPXFYAe|=q0a`@LDJ=<d4xdo&d2W3Se00ck)
z1QI9k#<K%=yKv8uMd~tfj4%%bKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)
z1V8`;KmY_l00fK(T>1HO7jJRdtEH*@<>RCOcIqy}N+FFhMPv{F0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X00D0V)_HhV
z-Rq;qmcHd?pPh2=+^63tQ<dHTLoo<|00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!KG%5P10Xbq4kvGqQxNEe<u)z)?^T009sH0T2KI5C8!X009sH
z0T2KI5U3FX_ng~p`FrLKDRPlDQY|PZ2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zWJ6%$b<@|Gz1_qzcbN@v7zzR)00JNol)#tUE_l96kIAK}yzz+<uiduoloB#e4nZ3k
zG6Mk+009sH0T2KI5C8!X009sH0T2LzDk9K-yM6E5Yv{z%mwaW;v=@(kXL1RdHis%|
z(kKcDfPjJk&Oh=z|3JOfR=p{@FT6nj1V8`;KmY_l00ck)1geOD={oO{etYlpY~z%&
zR%8`bHHrcPAOHd&00JNY0w4eaAmE9><*U5Z<=B>!N=Q2no~i|E0}=3bzpK{t2C6P(
z0|Fr6k-#xueb#%saYM_{a??jYd}X!vqe@@WBQme5_d4~eRX==peA#*Pv)5Q-rwOEd
zb-)1|UEAjUQdi!5%1)cUamxFpuB4u8M~xb=<!wDCm5|_&ay#(aX`vhhKmY_l00e5B
zz&a1ls(XFZ*ix2*H-EF;j{h20Qc`ywOU0}4YC!Qp00ck)1V8`;KmY_H5jeGesS{Q{
zZct$=SGoM$V&J%LB_$&j3)zAI2!H?xfB*=900@8p2!H?xfB*=900`7Lf#+9P{=|bj
z^(wgJ<dc7S)(g+RQrgPRK0D>!xlg}QLV^R@LHK?k4~H#IANJ-Ur;jRaC3RnY>Dl`{
z_+9&vrLNp>=28o;?m4#PlBVU7`M?bVAOHfX5O{9tW;^_9>4gQC+;7t8Z!d09U(uBp
zesb)9x-nlC)gn`=SJ(jpAOHfjPGE~-r?cj~)wi&fE7kq6?}Y1LE_CG)3)h@@^87c8
zYL!%by1L)|74Ccd^MYGa`63(yKmY_l00ck)1V8`;KmY_l00ck)1V8`;d=jAE=R>&O
zdyn8$-JE>(LFfhn5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!XNRPnftGv|Z*p`z@m85>pwpPN^tGU{=XshdP+U~t~CY7nm+N>3n
z8U#QfB!OM$uRUYyLnaos^44wI518`N7X_ER=iF|~-!pGWVJrLYH22lsTij6imLcm1
zd4T{3#3hjGJP~%8&Ew;?BIFJNAOHd&00JNY0wBOeVD;4>==S6`w->gO%MR!P0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X0D)>DF!q}(54h)mj|*;DEqqXpumnE6
z?$+s(F8;Kr6&aQ;asvSn009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sHfgck1*?>h|d)@v)$t{1F3tS)o0w4eaAOHd&
z00LYDj``}d-rJ2ETAIoQGk^Q>eeaDfLrbo5Lk|dm00<0wc(>1|{Ax_;Tdwo)th(1n
zjV*mkszwK&`G-3@o-nZtB`JFl4gw$m0w4eaAOHd&00JNY0w4eaAOHd&00JNY0<j2u
zx$T1IyY!e`QgZb(mwS44i^*l`G8RYV3j!bj0w4eaAOHdsffI*K|7wk2-B*T|Lw4G7
zgN-K-D`H8ROaCvP==i6$<4aS?eIGxRKDqpGg8&GC00@8p2!H?xfB*=900>kUfn|HW
ze9b*~4=Ziu!>6w^u;-YOrAi()YhLr@JKrsR$?B>Nlm-Mq00ck)1V8`;KmY`4n}F~4
zc3t(au>9m>um1Pw(zmSbR)Erj00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x<V4`@9tY1@<-m7JO7=PF
z`u=@Bf4wx7IkAAPAOHd&00JNY0w4eawMSr+ao6`e;L?eOFIjs;QDP7P0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5U5@Pr*5{<RxRFrzl2=Op?Yf?B?AEv009t)Nuc*u9iF?lc(>G*{l9pk<Dc4&
zFCoDJ?Ew7%<G|u^Am$j3yg>j2KmY_l00ck)1V8`;KmY_lfQvwj(}%rz$myd>SGm`z
zSFQTtyW>k;Ie7Co+wJ(Tab>QOi$C;$00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xq(`7~?!9A1Y_@X6
zm;BYvla71!s!kPKnI17%1Ogxc0w4eaAOHd&z(Jtj*8kDc)T^m`;|2{K8<uORzqIL~
zriK>Z|84BpxI*K~4RyVn8asXO-l(Czp?=@_$M<P!?0@;Lql=bxkAA+<YHQv*qVO%d
z{Ne7AyH6foauoV|9Vaziy!5+!eO!c+!v1QLQ+nMzv~%zArEkeq@3wk=F=Usi$A37j
zR7qRCsiZS)m$B_OwU>$?yF8)uH|?*X?L&TP5t#E>kDae<|8fya3K~>DuGa8a+~s^l
zJ)|yIION&-_vXwjWT|?7=juOm=FfYs_43TJ+N5jy+P-qYEo&asvSKP-^HrrEc+vwq
z9aq1wh$V$x@6YM<(5No6OIzuyp9uT7q_6cTmwnLl(**Xq?Y!Q*{QUERl0U5_pU{0@
z-t)NmD^4j*Wj-bJV`<;NU3PinpQjYDq@Pm7*DBBZOx^KI@Iwns?MM7D>%AY=(*Inc
z)^Y#K(e6Lpc4DXf|ED_asmfJH%`SpVYl9aj<Qk7{%?Q<geCPAlw=2!KvfrGNlBV;9
zQ1M%j?%I9ilb@Bj%iP!3w(C9HbhiDLN_Sn`_V+*Z&VtfaQu!bp1X3jM$xX*S^4|F`
zmbTKhOt$$}TRIiTx31lC=w9Q_DsqwX{y5xAJ~s%><%Igqo#r-7D11p(`_=Z3wJxca
ze6>OQjp(a=3B7);S8ES-uS=Cr=<$L(zZpk9alvk{%_(9@b<O$-5$!dWRJEhj{Z;2P
zerhRwtxvc8>q=J-t!qwQQu|5ChcnHybz79_eYx22ZTH)Bd4$?O%sroH7A?Q&Zr^=V
zx=P)8@fEMy*6+ELC-?R*?2kPj)@uH~&!<A^*F9&Y;`tgMp*Kwex%PXi9p@6B%RV?|
zr)xX%-8*dCgUT-3^+PJ@T<a@TdfR$SJ)Q6QwW{5w;!lkyuHmVaGZ{NFKNm`UpEq^;
zxb7R~uE*DVFVO4fI~T6c0p0Ye&lgPXi}l>eeSFZhKNertd}FOYT<tbZUn}FWDHm0I
z+j|TueXjnx<F;-*l|58{SEZx;bH5KS?DxG+jjf-0u8Ml5+6i|!*F788agbWqsPx<N
z;3~IDk8BrJ|F+>2w^YNq&Z|Vfi7i#_z}DZkJmQ?U+N5|+sHEUU-S6;qpGsv{s{Xd`
zlX0gL^(G{>-%mZqsd-we<E-wy*!F(0>iHSkpQ=C64y+~9{<G>nt?Y07KDc$kcoW8c
zKJb<%o;I#|j7ksHUv$GS>CpDmV|Q3o`jWo(d*aosUe2M%&rtc<&ReTBonCA37fZLF
zo5nM>_?}iBPnvh!x1~z@nqSL$b<3$c-;+7q-QmK6cmHL}=Gaojxz_w$zcXQ*UT$}`
zZn=Gha&vztGSqt3mb0ABmb31-fp}rQPsYvJwp}1y7y|PCxvM*j)&I*ycR$B9-uUj1
zzA$g|xrM9zpXH1Hqvr9(`yNdDGRl1A0t9yVTk$V{|MvFQ&ACcn`vRf#Qv3d?hjZ_%
z$I{0=-wdrU_P(fV{<`a8RlLB%rZPKu9$zhPxqg)Zm;>Z<3%K>ZYY$F~IfdGHMm|5h
z-%g!pbA>YxH`?lK&yBP*@D$PZ9Nt$tu40<5v$^ukO%L~el&!whxurIq`{rd^kFnCJ
z?j7Xj&-EO%)#KW(Z1YVuooT;!<(FD|YX3B~@KoE8tzD+=PZd8_yON&=+;*tSo14GP
zDR*~n{AGKo``f99Q|lk!`$3`J*D;;nnU2HU<CUpCxm{9jQ~kc<bGI{9yH=j3{^z{b
zcE3XA6Z$+V)IBBce5ms-nX7NN={S*UeEHnPw_DVUs^3xmT>W3z0r^7$R6h$9{zI$G
zU8%N{+~?u2KkmIw^PUhV#kBr&Hh=N%WtwBjqt@xQf9LV-n?uW<o&UVt@qcR9tdjF?
znmcouqHVLSY=7PSN4kw#sySE5-5R;ZBWm73g>#q7R-fyA*VyT8*HyOZxaJ9=^w_ow
z?sQat@EtD7L+xM4`$oR)8hFwJI~`ZQu*g-WcK#yjId%Ew@eeJ0=!S0#T^Xx9a(~Fh
z?W<kN^<aDNihKV)biE9Ee%N_0cWzPPlDE7z_}eYMSgN9u16~~R>^Xf}RVb-zkL|q0
z^*tjhz3zCx9gpG>x_xoyC;F|5+df?nKC;Irbrn-7^xDc)=YgU2?NZH$nkU8%SG{*4
z{7u{Aou}P<?&>EjDr}{kpXyvd_xw|Lovy1-H=f{^o762|sCcUSuQh+UU#Q%=?VT&0
z%8pR}x%DsEUSI9lmVc_}#VWg8$01uzs&v$RlbW~7;gmiRuCgbVf9U%qa(=POool|L
z9i=8it!udRF&&r0^VoHt*jM}Wy}t5gx6sR7ztE@mzP9cDsqgQUg!*2eXn(0l+xkc)
zojiXI<#w&V)Sj!{_PFxVP4Bv2M&;xB{7ogj?Y)FrO(*J&I=>L%)bpR}UsQTIoNBkq
zaNpl~3#8w*->cG#dRIy6eGFB&ypE^ba((-9tL*jWuWXoU1?aBVOy@zdOX;f}h0?3)
zx2FE4^}zicq@D++;okbI>3D7m)-b-<7lDj>|6lMotiB&bw;eZrqzE1$5RrgtzF4jP
zrq3r4KVlUvIcwAB8{${4I%;;2%zPv9`mnwClxn&qcOrdX)A3MfHl5$+9$%DOmbsNn
z%o9zeTyD5|38?m;wyzv;%bEwZtmsO)J^t59&2d@Wn-=Lpo!`j$n6^*f@nx=X|M0uZ
z;8EQ?O!rk&iyy0figDChYL44n^MCH|Gv%Vsx_$yBG@YmG#;44W)o*;~%YA;}TOU{d
zR_Y-dg6nv$s!t-@V8OcmCDnLhKd)M3s+=jiT>A-?Ufgd9EtTU2H9ukQ!M5j9zS1!l
zPpyvPoLRWoXYegmK&gIQtKq)u!?%84{XCT3+@Ig7c^>@3><@8&Z${`-jia)^ypJ2p
zE!$=44%Oe0&kx^I^1Y7stsnKdxIU?MjIZ$A`KLDTG#xj&;`=_&Gu5wq|6Ihwe1~_w
z6S`mco`?Gy|9t6neQt7%3$mV6e&^iu{d-^EmC@BJbClh=KS!(0e%tX+u2(L%?e|ba
zr{gMbC_SosU|jyxI8NnbeGVY)Py2qpy6O*A-(2~G-e*zSeO=r3_doQ`g3?vW-;3gM
ztLzs3t{h{{`-T0oG+pPU8eg|w<nsG+i+xeZn}>GpJ-&41pgV8nis!3cq^_519F46<
z^}Wel{B7s+xu;Y0k66!5+kAcPvxL%1tpidI7yGQKN>w{o`9HI0`Av8G?we9q+Rm?a
z^~>>MkN39u$GU&6s%KMwuK7+VJ+AXW-SnaM_i~v}$abdI?p5_F+bz3IpNG`^$o6@k
z>T|}IU8d(5h`-C!<3F5Mrq03i`Gfdj2)O<(r0(yb`HJU@QmE&AuKBpO8(--~dvHC6
zQKe61eJaWc2?8brROcDC^Es71ZhyIquKH6wZ&KOy;A=<jc<e)O7q-%rm&y)Te|epi
zi`$f6Zt+#?WLJODZ*$`z>~(GLw(EXXI@5g#5ue(hNHsjy`LVEDHIMY|FY?b#Qun}2
z!+mo$)o<I+Lf>D^Ex%OmV^em7`h6+n)A6Kv$9-GM9H{z9J>2&EHI;Pjubcl!w^2(q
z$CA|Z&~|)IHJ$7H)i%9q{;Klly5F6R9<eael&Q9_Sa#*IZy?)gyU&JnTMpf8+*w6x
z98~)w{~U1mgAcF$R>>{JzJ#fi+o7pjRZjTx+(VZ+y55_j+DEK@r1`k}Q8d%`8ah7L
z{fk)jBJ%wacuyX-^U74x$^F%rTQy(S^*60P)BP;Ozj*0)_xiZt;t*>-&7H3}cSuc|
z_6ObgT=g&O2{qnP^99#%U5>uuQTsl%7Ve9)tzOsm<vV?D=OL86T=Qew`i`BBtNcPw
z?DBBuPt6D9=M=Y_yF5al@8@Z1^Bu|_-E~Cj@qDeneCy@fH;AQ&dd_f#i~j3Na^34T
zy`RDrpUZw-J-*J_bn_SSeA!{U9`MyZZT0$|7yH)l`+4nKzim58B^@;nOf{UVUTyoG
z?{s3no=b}FE~xH5>7GY%*(39#<~OeMG*i2MwR_unwrP7eEw3(LTl+;iU9Qx+CRVtv
z{nZjr?x)=Dr`!H@rK^Y5HCKyNjjy@*bFJr~r)&3-PkvVFa-g1n2oLie4BdH<X*}KM
zYwUR3<GZXcm35D>gL_|F)+d)gwp+}@a+TFGui`2P)t*epsnp_&^Eb%5NA5m(co~|5
z={nLiKGmM7@L2O~+xb~6dwu6?%OjO^q5D}X`KkJsIFCt9s_eD(Ppy2W_L!cRnd&z^
z&lL0QT%|mpr`)DIsrb|S&)NLNyO(KJ$%}t>&aj8uwP{w#ek=av@890ux;eCz-^&*6
zH7~vDibHOmUkWL|x4gxb_1k>XsztLbRr`Zq?X&B@pFN?zSx8m9OFFdu^w=F1m9k#h
zE#e8O(kJ}Ixr!|1a;e;MI?AomqwDYcx$j%Q?f1G<OBd@q%C(xW?R>yBov(cs-F;>5
zc&W@&Rd&QW7c|Y+^gK7U_^#{x*y(-WYx1q1`+4R%&iPIs`&=Z}z9`o|F_#|GaW3`v
z)OwD)U8|m3@Q<^<ZOUCW4)`ALRr<Nxr_e|3OXL!6yYD18q$Xp}n~?wF?-Z%}qp3ev
z`x@W<(HG`zKDUT1xymE!QH{IR?r++j<oL1X&0ObrayzhgOTQCudQOr`e7SyY-4sV(
z^L8rV+`>)gk5qiBKB@GnKc8cbN3rsC9XIyrdhn4wKB=qN%Izx$+_L6DEi1Ni#eMeq
z`IYOoskqAZ_t<ybyU(<)q$at?8{73ps_D4)jbiEX^&AVOH}&zHvX6S6r5^6<{I1sZ
z#u^W}@}=6nDxCXyqta)(E>`K$^&jx!kY~^7)2fn-Ol=-wYR{dg-FxopCoC#*m8SOy
zb>r){S5-XM=ekO-uYEaHexdzcxl!qL%R`;#r5di=zC}FS^OscD8M<~+`ya01sob-v
z_4y&nztL)I-aDeO<-j%1Rq2t(V_*Agaz3Hmq3_Le<tOf^=Ps%H@Ve!r^l*pU-p^C5
z8%6(8J>OL6WPeq9%AY#VGhJ``itjo;P#jVTm+Lh*cdGMn$}Vc1!X3^vF4^`Aibu7C
z+w$Q~C&qcm#lO47%@_JWK!<?u=c&pDuJK;zk>CFcHSe-rx2Bqo`?(%UpQ@b+|5SO~
z+U3esaB<Z`r5E{7vRdX>w&jrd_#T(5O~1^~)NR`jP18|!t<m?p)BB&?syVi#?946P
zw!Ql*CpB)y3b!4<Qb`w!o1Cw&cEhDNmHM)^Lv>!`%b$xorAMr1a+S8{0kxWrY7f4`
z#r+Cy>8t+<y{_*Wkm-E^m-PC&C*@nO?R}h7(vkI_$@_C}9{<q7hi>?`FqNs_L+7#&
z=L(C@6+&~6_hEgzrD65IPOf$oTaWEMz5jyn{7<ZV0#v=I_GN7M0Yo~_II~r6`+QHW
zz+B^H=yHkqv8}XS2Lzi>g$9~geyZ^<lz(QpU@izOk$~-fWUA?O-#2lMw?vo*p1$sD
zsPy`pxB4D$<#PsZH#M(D_;HWCcm9i|%t3DIw(B<CbYeZCD`op!-E#SJb^96zxb<?a
zOJqHAJJ4-^rt#!_Lc8U1o4P~WE9aN${6g0*+jVIw>2e+4gx$X9CBF2V_S013``!ol
zt)J`nROE5Z)XhEL4Xw}gye_r)p}xan+Ac(VG2Z!>BEQh>N0pzjM@ZHC!s7S7R8pi9
zdTb^4{I^#0+47dtr7~}V9hY6+_~$8w%t3eF94nr$`CMqdGIuVw+~0C_tM*5A{kh5=
zdwdt=P^}+yd2q)QcDYK^{Tg5KZSOTx_k*e5<D~d;t#5SoQ1-Zn%fI)Ui(CADZm!bz
zI^4H@>iaj`;jZmmmEL#zp~f4oaMN<A)^(J>Sog(}w&PQ1dsXdC<<C8@5&4+5FY5W}
zD_q!XD|OE|P2-u`!#%G{MW3&Eo~~ZkdncjSlgMxJ_c~K)zpi$0kI$j?<=P%~$A?to
z$>$65^Tc(Y?0cS{s(zUx<qqB7RQXYM`U*Fl_xOshTR!gmb@Nr7o7wv7=4%_zw13*>
zlX^ONoYi%Ua>?J(xvp*d`yYB|L8&WE*IBv5H*H_O;){ApP3pF@RO4}vv(={0w!Nv&
zheG-5+I{iQ&KdS_yEe_aO0Io9U*i_^cI`g$$<In~$b&<w&#7v$OLttB<K?=4qq3Xp
zevL|xD&K1Hw_T5Nr4#Klm-(Kr=a1lEE2(kKRxg##7Y|px6c5{QiiaF-`kfNj^C?yO
zVBgL7&-cts%gq(v_FUgK-Q687Jb3qCwru8>ayxQ$$9`U^+GVc(q5D&+`I*iGd_U(y
z?Tg6y==RT0_g-AjsY9oidFr}#xyOp9T36)a@9Vk2RUfII*Q@OE#WNSZqTadYdxM@I
zcHYaKTU2tD15bKjr{n4u7P-n?#xa$hzUL#Ny+baUK&|o*J-_14&vySLb~@kfP1e8U
zR!{r7?tiJ$%Q?rc4=R7zUYTz!w``ZLTX$WPOFZszCA2>7^|)?3FpcNh9#r|IK5y3j
z{);?+4&9DwJwMm>Nu^h{U&^1-i*Ucc+d)0wQxA_dzmW5#+BtW)oPTa^)&85xpPQdb
zU+&|L>w8AF=~X;^`J0|=g^vHP-dF7NXopFKZ%OgheXk)@Jj!0v@X+<jm0zys>cVcW
z^8reaX?sY0z9QR~+jCoGuPB$OFV%aI!e7;obiZ$>Ixon@U-Y9~rKo3JsrwzQRO5;E
zotqT>Emx`9C%5(2ZLg|$w)x8GQfcS5-``6uU8wo1uY6MDX={(^_sLR=&wbtyN}p|g
ziFB%Yi0*vMSG-*2bCjK-?|-T4)6}0@pQRqI+E2{IKi7GiusiqtR8bCHshT(X@*m&*
z(HG`zKDVegX*y2n#&@j;)xL!<e^+~{^zym{?j^s!pWFQb-SH}RJkdT)*HyOXRHo^K
zeX^wHdtBjizT9qK<Cd(K+ip{Rx$WD@_J;1KrhG%iH|@Wv#OJO*F5cXB34L7lQF^$>
z*;sl)?+fVWC%Eg*t8L?@zJEa3C!TYHmyoIKF9>@=$yoh}D_{A!n~Pg@E}x5kXwJU-
z1K0T;Wv8!wci(zb86REk5aWMnNwo{paQVB)u5Q)$dQ9!Iji1|ojcjl3`{qZN`)s(!
z{S(VvrCcAmxo3~x;r73+`9&E^+V<O2({X)&n9@U)-*%m4n~t(ag!`T^nChq2Z6f^G
ziCg!-ZT{CqwMbJ=vE$pe%T&^Zp7(L*mkP&fw?mARp`;k!VoTTYg}c4m>XYj&^nQi!
zeH^iGUu}|WA10KZ(ER0g9op^der&9J#a!o&q8`Nk++1boagsYf-{UUyZ#i_Yac31W
z2hsk0$x!1tcRsr9JNEC+g?^49zq?27K6!Xaagh62?ryGg-&lH5{ce16c4<BcY}a?8
z)9K#Bk>hD%FpkD`-ejCHQvzS>2<Ywf(5No6OI;4Ru2W$5@|*7V-8W@y4x#6*w!;tC
zco|wxtbLJG^QGp?p~GGG%T3eUwtp&JsQGm0c^j1<#f1tF-A_#4yQ1P#<<kw%eO=0J
zzu>03|0Cjs9>09$$Ia7JpWyFH`r;N!Z?(KP%)MW0s?S&ZqV&r947s^O%{RF7F`cg=
z{>4kbyVu7B7YE<tBh~M63%7k=B6K;q`)O+WZ0jL(I%>V78!qOvx-zx;vbAT%kxyK(
z+iP<QZfV*dsd_TiXByx4{L<7OIX<;6R)x!Ye7U*qB?&#g>`Yy6*Ee>*^O5sDEUR6n
zI{y#-{Nm2fbl)Mj_}sj>^-<5UT*7^~3)gX2JSTFMT<bztj<M5oaqz82cU_QbJm38}
zRsE*zTels_dbyt?-1<Vzw{m&D$#%;1S#56a_7qxQ=zA>O`RR@;xy4gG?_B#&ZX76k
z<?#mY&Yk8qOelPDP_@h4{B7q|rsu-k@qO>7a_Oh~t!cRHxEDG-_r8BB`h1Nmq4b8D
zulvf!*Emb{kM;N1cig+rw60Kcz>7nkJ*Q8rimRmZ5#ifc4!C8_gIZQ7Iq3Oe=e^vy
zMMWh|`<HHerP={C->1^c;kxIQ)OTB`dm7aDNnG_)<)HXe=|woz?n8%D+_=Jhmy3Iy
z=UbmC57B-o$yHye>Y>VS8ZOrZ_5PY_9_i~jDfG&Aa=TUKg1_GT^07H_w+C5YZr?SP
z?Ul>J<re)jS4qul<Zv!-LXYcnM(EL%q2C+j%1?DKEEj(+4zcub?NgcR(QVhJ@#J-!
ztJ`*e(>0x|-M-UP>sqdG*KstL^z!p47q{F$sda|0aM=#ZoeN(sJ5@Yl&F`uGI__{T
zj<NN~^MBc`s^`@Fx$77DdhC2<`<Dw-4zb&Ft>mxje_Zn<S&u3|mA~tJwc69Gxc$ig
zu%{|l9W}eK{(j((6vCqx2~g*Pw&60%H;;d4;X^lkTX0KT`&8+4#}~PLTyE3zmss(A
zulu2Yqt({DcSPZf16O;s)#IyNrh3IalJVUiePQ0_bBkPMsQo9~`liz9j=OR^N^j_J
z({+&?-_%}tU7nje*Y_4>yG_p}RPjygpF6&<eN*mx)v4(7#S40W_&sc{`u<UWaa>2M
z))B(r_xe27dlRtxxJTYQ|HYEx@T2z1<dNGvAy<2CtIKp9BjVfUmr6Q$eqSqYQGdRr
zuX>i(F|O{=`&!n|X~+2kcfP*ck8k~<`g^SV2)@fF_itbCVYu>>?c{QY{;t2T{Cw3L
z^xizQbMNt`iUZfWn`^xqT2H9^!Cd{l+Vr^2n_{Ootq-wJqATw_?cQ@&KVeZ3O3LL?
zjT^c6yFRyLr{@}<x%5!|E%Z9ZSAMqhR=NL}x_v*7Ozq){Z_6`OI@fuJd_HFCR-KQ@
z{#^TlruK2gcl{m6Ev~HJ=95+}nr-FX9WFe0_g}VbHkG#dsnWS}h?Sn}ITA~cYd^P5
z&oz#k>f!3QvYt@mvhRF!=iyX5?tM*DeXjTYP1_aXgMbME>fBd1+$2xL1pyEM0ZsyI
z^!@Ji{wKF;_ANPWt_5AO))U<M%JXbfH&xfV&)L}V<nLhW-uL&lA84zWS~u8+=d%CK
zWvA;t{J@hQ*y*_Xg$0*1?U&r~seT?R+_wG3PRG?QW9bQ1zrOE3QTpZmZ<X72eCAFU
zs$NX<;U0ft>5J9hxbmgid+On?`;4ijcb!+5rZ?@ks`$F|;oRe?%5UwT*!^YGeSWU`
z)aA#OpQ#@1=dv$-x*X+rsf>fNUAAu5dkL}T6RG6OHNOb0Cl!8Nc8GSY+YiL|G;)=x
zjEkyuovy#>Jj^vdbq})E!hPS<@U1`h`bvG>NZIFW915k^SNqg`4&@%twEvlwSGRsl
z^~&D|4dqT{{lIO9>vP$6dfj%!74MP`Z9hGBhec)TQgzQcH~;q6&3~lZsHK`sWp3q6
z&E8n|*-gj$RO9=8Zzz<0?)x;c_1V5JkV?8x_3oOF>-Zt|eR7wk^XOFK$G&gImA_~|
z@^d2gI3)6y?J{+{u5V29ljEl{F2=S)=8~E_HlJF}-}O0|N_wvE3{m|e)c&BH4`r8a
zIM==d^lUkFuW@G;rW}0lKl#?rH4lOwz3<(sI8o;%sfLUBZzxHPlcDcJgw9WtU-dnU
zRQCCD`Q2%ioxbNqq8@XRuJdcx_djyIZzJqZW#7fr4%7LNZhX~q(3ihyxl@au>hoV^
zmu@>qJs#IRLoPkG<7lqG`=zp*`yN6n`dr69c|H+pUT&KYSHGb2$Xu$`ts0Mg`^Vy$
z`tyUbFZQ_OT5qxK=gvR1c^_qu>+>R&^i)4dH9Qr5wsuhM3*kL>zOwzxh3hye(%aTk
z>~yB@bEX#G^&Ug5rWfrtwoI)(g|;U)f8<{q1VXjzAJGi`orWJ7uw)P2-(lIO>%m9%
z_@u6)D^>eiDu4335lc3|>V3Js?-onXo5w%2@Sz*NEka9qKPXoDedSyA4gWs{jJ+<1
zoqz0l=6+tv{W7%M*FKBxcO7)&xxPQ5o8I=kO*fr<Pt?^d`h%;U*y*|Zm2Z9Go`>MC
zlD_YAs`9NBf7^3NIUV<N&DXe^n%>ys23P*6^cU6h-@ARG*W*n2g}#sLD?fQW%FWF^
zeumbU8*kZO-FbAX@wnaxqx5i(qoLbrF8TG_`ah(bdNp-#+@N8(hUFXTFKs%gsiDR9
ze;YeCuF$x0LtXEt#!lb6H)^PFsNZ+~@qL;a`(M86=%Qual>eKw&6RJzTl$vgkDmU>
zS?9b~ay5+WZa?#{Cp<l*3?(mn<*nt$Pxzp$HrcR8_ieVEHL3J1x$50kPoLRaZhrCQ
z6HAq})tgGXYu0@5)onVzUFyo%c25}9>8wMy99eQn)AICu{;tlgPaad&iZtbe_`C#u
zar7Er9e?vHg{^e8=BVxWnRd(#Bg<G)*#B;;bH?pB`d_7~JY~hN4x9C-m&#n_jorun
zq4%Zll(D6#ht%bt&R=H1nq9}1vV3>@b$4I>%m<2Uk(acaa`ab6JX}_rbZuYHZNBh;
z9b4a3m`c}tRq1WVZLV~}u2knM!ago}@YVxQp8NY@CAZ|V4|;x@z>N0)9J2i0_m!0V
zX)XDL{$cKbSI$4~-O^OXD&c#F?Dy<i{f3vOlH1--#@_ekpdRm)u_SkSQqebL`L)&>
zy7AD`R;E%8VaK?YAG@*ZF(b-crD;FQCBCR1S#m$ORr=(7P2Cq9{=_K*R(YeOq^VvJ
z-}HIF6<_3UE4j`CZ1X?6?%k)ly*IARUFLpXFz5YWow(^8W6RuS?(OKzqUYEh+m0(k
zN!xx)r911;b>{5){wt-cr1C*H2&72h@>{30U3B5#(pI{b>D`ZC-1(Z9rk1rXRUG&J
zblHFRy65d87pa;Tr|zHXJ|>r47hZhZ8CU#iLg`DY+OMwvcWa-t)cNfvl#q5DbUFKu
z*J0GAzkB7%$)zvp+kTb)K3`rmVdjxdC8QmP*yF#dUVZuR^wBvRx9$065ldcwZE=0S
zEuScJlOzA}$@tc<4J&2otDZvX-D2{zhi~2fwL&FTc3jwE%@Yp&>f0i9`S*z{T>ar^
z9~ZgG?j8Ryu}|0L(p9SJQTCsB{dtF;wrFITi+ui;xh+SJeW5g!a=Bx>f3wt#3wG?c
zY9&{>)yz*PKG12kimQxW54q3tW1SDX=F4@C5?hZvKgiYXTA!iQ=YH-Y>~Hz+ZC`rq
zp7%>v8G9T>{$UAl?Z<`IbMbC154mfX_siVnXE%2oI)3}nW$yBn1-lfV-Zr|lmA+ej
zpYi8!GWF3nOH-+`YyQzoZ*<Jg<I7xSF7wsBFIm`amGvf-)h62<^2gh*nK`laB~|S<
z75~@P9sE(7=O>oFWGdxM#SZHJUFzYpI*++>g^^>*+-0oy16}X+9{co)|GsgXQDyFO
z#u2>+eA9YNnX8=J|Bk))>N&2=RhstSt^f6#Yo2QTdLiU*dcQY*nJ@oRa?AP4ZN1lD
zE_$%678&cg6Z>2=RQ|E}y5{@$iOb))|Ma)Z(9+dz)AXk6yj<g(a#6)UX1h`C>#qL1
zFqJBOuKov{e$X`&`VK9*q{<%ZeGpYT%Ky!&o4>N|6_d)`<=?+~`JT@QO)0BYj@a<1
zqkGMnQmUk=XWMn^eoMW!!f%=<m!>lJ`@^Q?MSQz<k}5aRZy*;Fpq?|ihIfA8$jhEw
z{n<h#Mg3FXA@jA5uKFGw<^PvKzr4QZo#V^UGSvEo%16|jkgoSGsrGGqerWq#)=lSo
zd`;y!PnA=&16#S?)XtS%ziv?pB`==7#u=xb(W0U&ZOg@-&bA+LrQ<G7EPd2?;~LJj
zUvT5uS0B6XQ6HAR<j~{hU3A!nQ%Y_bt6p{URgGud975~6X1k-`96oz;3AvbqZhP<*
z&)4^{P3={+_t5^_^D3FcR?mEQZ}ZsUMQpk21z*?iIcZdpo7`^Nj8-R~GpXQ`L$02D
z*>;yqD{_@``wHcDoqxni&o%F{<*YltAfDg*RlfMzwhN>SLx9>hG7S%7_5X6A_93~#
zhYi2vyHDp1EnVgREMNQ|HSYWR9)_)6+5XV(i*7vN+JEfzdYP;2vgc*5e=+ceB3HTI
zW*hGD*;CIIxysH*Tyf5V?k^U(%UIuS;L3M{{!g#G@7pgIxynt~{_F6Id%RlM%FyGc
zD!;?~p1Obi$e&fJO>VsEshf{Kc-e}Os`7LDQ_pjR+kN-FR=9Dk$DzmW{XpN`xW^k^
zeYU?>sH@*sywvt%xyrd$ee(<F{bh8aE4lK`P0ua$>#VeBR$V2wOr4wUdfrz$-@JbP
zi+f)=Z0uV_F7mszPZ@o0$H$7?<;D-stY4?w3q>sH+OAf;d4pRfH4Q4NMVhvMSAMCb
z$M;U{zIV#S?ZktwIrqIgUo2f^tac^eJK(nC+)kH{IP!s4%iN{-E>*QiU-O<&dVSrm
z52cs8pFv+70;cc($FWkKEBAeURsW~%bKCB3$|d{t10NkR`@#>4Q1TCLKYHZg8SfXl
zN%^@iyJfDv-MN4FPT1f2AB)Of>r5;|Nl`CU|MDF!?120sf!?hew;I&(gEF@Kp;hKC
z*YB>2^yWGLbPC_?8v5_*-1_7(Wwmnn>DJ{x^ly(}x9qLIom5it$}{FqAGrSHQYDu<
z_NLp1ocTd1OWogpq`tH4`h88?=OOhyN!Ry5mcRb*znpmZlu}pfzNh5t_myP5+%1&r
zzNqW(W^k9wR-bPFNi|;R@9Md}S06gPZM)!3ryAEo`HS*U??K4xdoeHYHNUleZlsbf
zwcn8z^_;pieSX9$&yeNUT5IUWL(ANx>At0Ie7PQuzH;5YKW~~=#+FA8e&K}SuYOd9
zk{@jH<(LmPomO(m+n;-`uGj0+N|hXU!j<!1{{2TKB@b_R_>}|aeppiSpw~9}^0=#~
zltSu$7eJmz>!#C<cimIJxU|i+Q_Ij&^*ue|@B7}QZ~da*is!j3<>$VxTj;fwzUCFO
zUQ_#0o$rUTE46uzY|pP>y5Z0ByG<-r(zL!^&&lNc-dyLm?eBbaOd(7897g4SqR(^v
ze!t?Rf|4ukvd6+%2Y*sf^0;qyoc{A8W)zfsvC{ADe{3_mP{{`$_}%ob2Y+3-%AJSz
ze)OH@ZwoH@o8ET}xbD%qimu${+1}58x=V|SFS%Cdng7>m%a@8;mvX;QxjQU3_<%qC
zVvR~Dsj?%KzwhV0Z~eCC`BXYv{<l7U@2`(}ylo{HxzBfx{a@DyT2*YNEhkkvYQ9J9
z8_D66J`t|6Czk)%o`-&Z!vED(V#!$L&NW}teWXzFO}{IWN_^Abbrj>d-2YA8zS^Jf
z`<}k+7J3C=$UNSw?lb;^mCkwm!@`$*?GLjy{oB;hg>Na^Un)|}7a(st<BC5`D0Mlw
z)?Z`ajc<MKy+LKQN>@I*>0N(^kjf|Z-^0|kkL&l7V(AI}oHm^|Qu_j?`l$F+yH$l(
z%kLWcIya`;H&xH7aMRz}8~Kk<#<zZLSm|4uewWa+9CCbByG+erKF3qJsriCDZpAL=
zWv{%o-1rF}l(|ak_j+{0P3wXCyNhFn4;;Dm*egmdDc%!^Els(a#-H3~*=~0&cXtVy
z1_#snljF<dqpn-l12+i7Brx#jo9}tY&d-;%A}_ynO4~&j4lYe)%sN_9-mdS#AboZH
zUMY8dnx5CB9)C%F$i(l^uU7lbikl5Db*0cP-&6a4?A>|TUDNqL{t$_T5TV9WR7<Iv
zsiEvcsUcPKSj15C*j7;$^H8d#1W~irSXxuXyrafMLrh6TLPmm$h(yf3*XL?p&*%EB
ze$MX<&pGFLcHI9w?`!S7*1hiazK1pJefBe_myJ8ca{JKEdfo5)YsGZjQEn^)mCh}y
zy`QevX<6&<(X?CXsq5Ar@SUkytMQsQ54iXrKc1d-7Hsc(pq;J#a$8+G_hWZY&x(qN
z-gnKAbys;W3v3M!_r7=T+2=p}QC3tO^3Q$F8!~)G)@qbT>&2;i9s9Rg-CoICg?k-0
zxyRN+U(7oTK78dzT}O?6ByTp>YoAuNYt}xblK!=Rm$oSN$8>pfzvr&GHZKY``+U8U
zAGCKrQ*QlQsoz%JZ`S|$Zk60hzk@B>{jcplYw=INGbt|`R{Q(kdh1pDJhjpF7GH9&
z4;P&^DX$e99lxzw2UXioqsLcHUZcMA-C5+E@%5#)9WrT1-fS$#XP=vgF4%X{psZ1N
z%H=!!dgW~<W<|$wi+{e&4^Ew)6%}t;<)YQP51Nq`70dk|xhU7C>~`dy-+63mR&?BB
ztsh=IV3)VEqG9)ww!PtqZ6{}q#&UTyz25)P4K6$8hB0|ZW2#rJ<x1ZVX!iXZ4m$er
zvmbpos})=Iz3^uB^Z2#D-|pzeMrB3C6<*u<`)hsqSoT%N#a9{d%&Di|pS2n%Y;^C#
zC;jd3Szu#$s5Ngks(!iiqM}@H?|E6H+c)|>nvmDN_pI9c9j2nSZ&Y4L<s7j5V)wm!
z(!!;V#=#4`dg`WE&R41yH~HmS=N@}xw^B!8%41rt^?i|!s^3}8vzo0(z5J$jZB)71
z-??^ly+)0<RdVXJbG7XbSZ_qHE`NA8D~FZdr>XWmu*sM9IB9{I?`Gxj$Y0NM^Y$OT
zmvvTbmAxC)&&1!o_0tD?f0i{Hhy3b|hrV~w=UF+uJNvq@oZhad-&3u%+)=;dq<@MX
zoqrqEf3@FhO2^-<@~ey9efdTQyqE_zhKE-FKiX#bMQ4|L-<HcCd*b-<S+nq`=PW&Q
znH|Sw&BnD?Ja5GvZyT2t6-U4Mlb?O}{;}Ct4a>#H1<T(4<T}d?%Zq}Qz6aCU?;mpF
z38&vP_sG1c_}1uK-+%Pbk$Gpua(-U!daIt3Oh593$IrO!oU9za{QVU+?{nO3S&H+m
zIM<GgZgp=~H2nK_4;%UHzpl?(jpgiI>-F14&V6$Kk;Aj1;IA%uc(Ly+^hDk)Jb$09
zrYyPLzw%mnthau@i{9Pv;@{6zDjM$azz0{YvC#abR%1Dy%l&_cO3(SP`S_;cPd)pm
z9A0_x=cU2-Zl9OKN`9Aq|Ek(@wfoPsUQvG+UDS0!s&}=&Yws-eF5myv{mwe)xx86e
zE<WRW$1<?{lY<uPdghp{Ufk@I59iu){D`chuvzRabKBm(y?5ycvu5M7f0*#w_0D`V
zD;k#jze*i_oj+KwH5yOY^_w#{d+mw5qtQ46gMnHY_~ps-f41yKqw_{#tKPRx$JpNQ
z{vP$&rrFz9kIIXJ?R~!2W9Oe;eZjPGd9~R6z&~C0>~XK>jmC2SPq4FJua>u>)|-FP
zAt&AX$oRa`C~kkX>k6-pdVO40H2m>je{uQ+$32_Xi=~@h@A0iop2-@88-2d&7JasO
zIja@#UTV!HpWXDutZ2CIZw~nFt2e)x6&15b{<y-bi;T!xjjOG{?z`u0@_g1_Z1s0F
z8vXl~M*YsVQF2zi=Ke*u9rk!uEB4#sS4YnJ$$X`1@#nW)cFsKy_9=Z;aP!CVNpG#P
zc&W4FZ!VcSVDkYBl!}UzPTOjwPbV&1Dk}cry*sClxxHJdsQBU%EB#}GUoKQCD(>^j
zKhHS+=0!^#h3DOK=>k{nHa@Qv+q?hjsP>Ki9!s_IzF7W<?kD#en^%j?{y&nPO}>3m
zwOH+Y_Lb^Y`oBg+xj%Nq;wS97=1Y0eu(LexYp)(f@snojdEWWAEwjqUcjhTh+<p8y
z=MQ^1Z!|Xgc}BJJZe8rkVe55&Aa5_0d;hCw*He2px_st*6Zf65*ByDiSn2=!HT%1c
z<>Xg8-=+2T`TjO<9x(8wytCq56NgV4IC^YebWC|~R=Ft8UCQa%s`X^E_%A27sQ$b4
znV;{n*tf>#MaM=zf9WWBLnp2N#0tL~pLaGa=lAWs-mLG#H0n8hB{|jZ)9bC*UVhu#
z{=DvYO?-6KFS5Y@@X%g98{NL#|4Z&H*Bkv^q>h%C{=c4j%Z>JfR>^DhywobaTh&jc
z|5q~qJ$o&8_zN?$W?{3&*{pe*+|K^{%g}GokstMYKJP3%H2eEa&6Z#3c|)V_H!B~I
z+iaf7$!`=tt(H@6Jng++>+eXK^}C4?=j=LRyOAGdwIYA&EZAA{T)leKyDsf4?K*m$
zAN}9BOt&Ln$?N5zv&3<#N2C8QTW5(EdZgp6^gFzVR{wna4ZGc#)r*znxA*?mZ!hqd
zV{aaoH4AIiuU4K8-|>RqjJ)BWd9$(F^PYO=lW6yU?|;Nyt(eAXqsvA6-gG5>_kQM)
z!(MpujXW@T2tSB8kO%5rpSHSPqw{+zui5L1TIDx;y^xN-qsy)Kr|9qhyFc|Dp;1`L
zpIUux*y#S69iJip|K2a)rPc9U$)79TuXcag=z6v8&(*K>JiI9NciepH9>d-oofQ??
z=c|V0;;Yf|(kOZ5^sV=Lqvx+y%d7R=uU7xARsPEReq+YLS3Hr`i?!-kseY~cSCZH4
z`<2d{I_rJ=-V__HZz?~^bN0Dgqx~o3{d1EG9-EZU={(fC9%yg7X8qr)Mx7tbX>yvy
zccbMt`uTpV<u#g*kXPIP{cW_p+neW_-M-f6){U;;Xud<<GjH^GcwFAud1$m>&nfe5
zXSZ9c^=tJxx&9jMpX8BWbgVbt@X*<?((kr9`~6RQtg!WJzvxmbI-dFbzB5*O?t`qo
z*eL&Mb-pXtU!%s8%Gvejuif+ObKc6nYWT*Ff3o3-v7cr2Vzb^yt|a&KsXu#q(lhU8
z?Zwm&>MgJL=G1wPJaK$hEmj&|N8PXW{ibsDEAiN@dB0M>MZ4eVbBjv<H}0j2-x|~H
zsps=@SnKz*TK)gATlIhII{N?j=zrmC|Cg88@}jI>Z13+a5BcLNS3NZH(Y#)4)bnfq
zXR)(kqw~!vd;a67zb!T-pHuLTfd`&<#Ptv6Ma8QY?z!_b!|u(Cim$vd=%uws|1K{o
zZn53p9(?jU*JMS-m*+Y0)eT4eIWIbzkD7%ecmBzP%U$w9UM=o*+~gix4}CE&DtbS*
zR&4ZqQ%+v3>%ekys=eN-?`1ZspN{%of1~>C`j>SdyZh$X@}lDDKkT~mjhnufHyW$;
z*K+mheNR!la&nvX-O_UXwfcIW%P!hr+@2HjW?{AC>2~p#`LW*|mlq9deNIv7d!Oa*
zH|t!dRdO5s|0+{?jec*SUU_dU`J+W{x?)6LbZnJhw6~u|{~w2rmUqaBC!BuI+#~a5
zVVaL?&0p2lPy1<gd8yHZUs~gd$$4PM@G$C@m6o~nkCXFqxZsHI{q%|RC*+lI&-wN|
zf6zyh^H$-jhn@fF<Ne;qYsItXIb^Lzd%vEiX#H63{|{ID|CKBKf3I@?AD~jZ&T_xK
z`QgF^j~m_ZlDqTHj>|1GZTH33d@L&})_OkCS?afXU-4w=drOR*xn!xmSZ`dEd(Wez
zuQ%!(g`9f%X>_~iUpVuNw?7)52mXhLM)Mo;f`MunX!d%hoDQ4qbj-nT_g=2lC`|u9
zP_6$Lx7zyYcrM@n)&0&o=ec}N!Dss{x7_emhUZ1Y)GlecsL$WZ>HXH|Ti<{5(2;q)
zh@VRTU$MRS^-~G|eMRr#W4=Es|Eglj&pB^={A%C%dz1pJ<00j**7DT7j{V!LZm;Cc
z#!7nqb%FQpIrM8wmO2_M=~=Y<^**0y);P;Ox9aHYMVt3iecL-<HL88H{H9TI8}+_x
zN69JozJIf>H;VU4a$Y*{tea*oJtiwE{^X2ve{<K8qqCyp=j$&!-_jE&WX;A({?M%Z
z&H7zWv%eG8-rosSYTqorQ@Pv!V#f={{&9F-G`wKh+n-!#nPGWRar!c^oWJl(ujEC?
zdfz8p<&J4{eRG#bvU0fk$Cp2IL$4RIqGPl3P_yOFy5=umKlYb*W<|rlfA_GF&;INB
ztW}uiRps;d+4}jvZ~taibi8BW!Fvt7;rXmq#6z>O+<O$wzFz5^u-xx2JqIm^A6;|(
z)>|F(YF-YT&1X^N*SinwY~%gS;^(}$_iaz+?ZsA)tJ1o*(*3a`7C&LvHDAhW#c`LN
zwCm5md?{}>wwljo_1CQ5rJcXeR#TST?q7LFVYB!Oxxqjy87TU`Fx9_Nzay+9r@b~R
zw_a=3^=AFvB;~)+<$Cx1)wb*C=TAlLKlT5<TkOBin$ySU&B9vaBGs?G<@C8*QOkS$
z<J47F`|O3hvtzT~OGy2oS@D<l)9Ug?vmU?s-A5<p^<t~_P5W!~_qSRt@6!8LKBvcp
zlk%csx!)rd<@)f0K0a-al_usz!%Dw*E6V+%@t*3NzF%E$In8Ueme>Em(+7Qc+)G(e
zv9tf)u9AM|UNy1jm#03HHydm93-v2K7iiRd{gwN@ZmX};2OfC;B7eDjoe_D8m42^p
zUrbKDnC6ctS2zFG%XiwY()U%1cE8^5dUn=!?R~z~?DnT!zWs$4Z}oH@_<!?|##>R#
z?LD_?)VRz2&Z$+`D~%)lPDQKAmGr9i{;-J;zMY>sI_qex^nQK0_X;cd!BL-Wn!SDX
zsJt9j;wk+;R<p~M^l9(?YQH<w+3NLwbc4%|xnWFRR4nRuD(PMBy{KkgZ?B)1YyaqX
z7Jql?&-#=q;XV(YlV7&^BBfgK+jotB?X<-gDOHQ7?7Z{Wf4a{irCRaNXS_3foeSnJ
zRf{Vh^yX*BUo=mtR@~v#{)_&6tL~+a!o_=i>(8&B(5qA{{`{He|2%Mo-lbMyIe#m6
zz0vCg{XDhb^zWTFX;fA%-o4bCOFp~li&?9YzV+hQcAvHQr{9^BHyUfzr`qoe9+w?6
zdc&Tt=0(Ad{(so*{odde8-DP^O;&v)FDhRB>2BZL<C8b?qTyfqFaP=<E}fDG1`qZ4
z!QRh2a@Y$`zL6IduY2gBF2_FoW)9Q&q`lwY+~v0yK6lz)lk%crt=|Dn>!&<cDtDIO
zPgc^iS@+K#bMkj5-8D4tXsmX8r(C|nuUFn?V%8|Ew%<md=TzHoz4gu<we%X7e9^m9
zE!NvlXKz>U`IK?py4aP&*6aR24r_hxRq5QZsQ2srZdjw+u~TQmKM#0ep}7vbFE1*#
z>i4(J>ZjTNN3C9e+x^Kwi*-G7Ox96Yt=_9{`P--88#FL0Do*Wn?X!KhzBO+zUa;Q2
zLtpq{LLL}Ae3kv6X#2>b>051HPV4Qn)oCxTzuartoPy1MzOdC?gC9Am|ER1{xJBuw
z<M#Xe$gH#AEsI@!&Hg=yXVv1jR$FO{f8X~|)+pqk^?on2)$Q8*|FFun$3wl?ZLNzB
zSa-b{Svjm%Zl!kheutyk?V5EyRf&)Eezng_)m!U<FK!t;`h~2~SnGE-+FSpqMP}?V
z{g;pBt;S~iRjcHue$?J_x%b>!b-lg*6z%K#UQRjQI_fz_NB@7LX7yifd^YQM-)aBN
z8fRydn~uA)mCMCnQLZ;@{%(|=TD|^nZ-0kxG<M(ThL6m@s>trU&iML|7g_CvyuEn-
zuReWe(6=Y#_2T8f{=<PQ>@zv<C|vrAZ{OYb;y3c5Vn^}TsD7KxJAV)SzZn=hY4s;o
z_}%!tS@?hV+VNMtA5c_0aQG9CT-Wo_ylB|!=j-))+RNWxVe>x6-Ig^PQ+?`v-=$o;
zM(+<&ee0d)8{IC|tJ?C+*ET-<d)JT7=M=2<yKtSYextu5T&*9}TJNvNU;m@Y_by!8
z=zv|%x___Qv0Q)4ZFT9~kKH{zs~3NG-BUlf?urkxX5kZaee&18%{?QlglYf7ZW!?A
zP4<|O7X|D6zENjs=l7h=!gBsw>H3e)y8dsM-qod4R6OURCtld>su@{P@wRPy{Bzuw
zT}nm8!!AAbo-dx5r_|9{$<G^o|IGO&?mJ_zJMyC9l^1_r8hr2ed9~PR{>sUFVcg%(
zJo)e4OGU-C_quB6s*m(8wHnLu*{ti0{ylA_@1WIszfs?nt0boqzeE1H&v`?J&&WCp
zR?@fD`^{R<G+S;_zEfIG&sP0TyjFi0)8)<mp1bDSyeK%K#{#?DxX?{`(ebK<d+z+q
zuzPd%nsx6k+rIIm@mZ}{?f0l^{r_j3t^NkzUhKzHuN#sT71_C3Z1lNftK~KNd@|%!
z_gr}EGe6&Dv2Tsf8->;SYq@$mtv%?-kNQ2Iw-?*{o<g+$+Y=`r^m2}eX1!0`D!Jv}
zJCEx(F4OJESMuPYxjdBH?^OGLqMB8&+;hyLUH{Qz1CRf7mvMQkuu<nD=iPJZ0$1%e
zJ})X3{X6qT|DE#LAFR3h>F18ii;A`GGs>+Sn{~b1|D`Va^-BL=vflgcwZj+7AJP5f
zUSsoCVSDwh)V@~#{;Jlm^&G3C>;K@56IbbZ%+Rc}Bmb=RJ&9KRzdHJ*m_72x6;@qj
zMAmFf`)zdjyz_5cW|fcc%$tpk*1J}Dt@>SPv-)ZF_m5Kfwes7l`h$+{dC{Xkcr`B?
z)|%fc)o&EvMgRY<AIzL~)H5fH%ZiGP`ff?1<?yd&|G%d@1|E3c5!XMM7ZnFBwfdIV
zZ}M>4b5Fc=mgkN1X%%k$_5y!7_U2((U~743^!G?wEpO$0zcJ(BE1t+&h0TtKpWgEB
z>L>L2Bx^5L(zCt4&+Vx8&HDfKt(M#9-=|b7uaaNXdf#)@W@D{-)m#6CZomKbQX9?6
zii$TJboAqAKl*M~RQ%+7UEUt@`LwL4`0yjoEVR$S_p_qn`)@At*w>Djl64f;8@HXc
zU8~+Ftu+2tpTAeq$NkBd_Bd&QneS$u1rM6F;X6z1H8rb*+bs9uk-u2zy=+dwO25PG
zZ1>OGX_Y(oK73|YRJ`?_4|^?t*37J^*xvm~w6AWTCr`Et$KHJRW<NUTwXCDDz2~xB
zJ~?B`dUsCDIvf+jmHWTaO}pOY@3o$=>o;d^_SzHqoQAJ1diUiU9q?ja6m0c*Oybrp
zFD|#twA~k9^RcYaxcDjqo;mf@`?FT#khL!ydc(%IWbH+UX!Ux((fu|0|E#rIUQxbJ
zQc2G~mtC~MxIHK4MM3`GEIj-D)5cEQdq7?^toD0u?Ok`ZxBbIce$;i;=tuH79b5H%
z_D1#7-ruc7`^NnK50B0I-gv9zp7HgiwjDBQNZxF$b{<$b+yCVyw!A3o>{w}^Rqnf^
z)$&klz283ZuZ6~peLZh4*4l5Q>OVSQ$cQVi?_Mf8zWKL)?_IKck5XZAv)j8ZcGn`M
zHw%jAkAG>i6|Y;MRCL^G+KAiEyrx&FXn6S6J>J~^oCQj)!g}-1xmQh^KH&5>vZCS8
zZ|>IXhPz(LijK?Nw)b!EUHZYSz1Xbx_3M>uUDztD*Z(WEJMjKR{&M*`Bl52rR^p}7
z{c695*r<BV{(gC-@s_(^^xv;GT7NtpzT*YI8F|A$^J;PWktaNU#%<?hMaM6eKWfx1
zz3<LC3i018oW9H}=P&%yD|u0|)$iRltG{Ob9;#Zo&GN5Ga_5>jeA2+tWAmb7tG=t!
z=zh9g{AGUZH^=2g#d6;bD$4a{`+KFhE86|R-+26@4Ne@9)rzei=hMBG{8yJLle0M$
zPq}=DU$4B)#H_Pp*9WIOxx*=wvi9N<_q})di1|Lws>L23%-VGOd1hus$LF`(^57kI
zoRT#fYmLWB=MlBatF?Z!<G5CR()#s&2PBo#ta9%mGv~cv!TCy!!e;5&+2vOI`@Evp
zJLJR@PQPdFk$GTadANN4SNA*Xoagd66<huOYOVdX>b;1f_tV}y-0a^GH(P$AzAxKe
z9y_{yQQjAA*8eANmE5oD{{yJizpIV=RBn5p6F0kktG>6=sJzx{IjKG8v~sn6-C67H
z^Zjk!JYe8Uc~P;^&ucr2ymH@3Zg2d|Y5$L*()m{0Z|r;Tt)4$RTYqmX`J+W{x?)7$
zQP^4XQzbnb{r~9OoA+AXe(||)eRY9vzMai!c;A-GAA92X@mW!^*5{wq*8lGD!xuc}
znBiHo@kcA)^28pa|CKcwE5(QJrB=e<EPl?5d*AkCUI{N6*XM@q-Wisc!<R08YfQJN
zp3f^`wdeJf-e;|p&(izV=C`xvIb^Lzd%vES!%FX^H0%A6O25Z$@Bi(q^nZuidu~$c
zchlA0ujKcwzF)0hwYOfg$JOj}mqyF4m5;+W8oTdv!$;<oa87>?*XZ{-Q#(}pJx00r
zshj=(A5%RWU7p(O+GqQ0eQO@ru{?ab(Arz|KkfFsR&2CCr1DN4_0F9qocL5;R7}Uy
zsB&jLXRTJRTF>Vi#Tz;8Juf4tT5NBB9o4>Cdv?}(wfb3StKU)Qg4OzUt@R!~Z@b>l
z^qrJ73Onlik*lr0?z`u0@_g1=u~pxd`{6sMojmHUN!gr^<<7?%6`#$@pJ{w{)bICF
zJt~zO{hhy7%d6Ed%JI?v(G4y;=7uqOtFc+<apmN;>i=anyPrzuapmmX*{(M$AGKQU
zUdK)DvGvdw^P=L!-N&zU{;-$xz~=H$F8(S#2P*emr_%Lu{gu1k>iMf^{gvY_y}s&}
zzkT|>K?AdTF_jnPuO<VpyfEmcwMYLhZ#1TCHtRg1-1CHbuUC40;eI;4MwK@^<-@tQ
z96uuK?6}^WQ|CGI#PM0talf8#+_}@#*Rrn~Hp*Y?mDB2dV7dM}>O8v9{kH0T@OtfC
zsa=nqe|Gf+)5hhsVkNn)zJKX;$DMK5{m<t0Vy*g~{==?2-?-^(c~P<1--T(F{4bV2
zqWj6c#^&wCTJ7KH`px32-1~Gzzg}y;iu(2C`S!lwsC^nQ*7NmZrFp#GbHY^5a^<G|
zAGFUl&ECFxR37p>mWOh7Z}0V@e>b5h`qyixdf$hvx4+KPuHHOc^mhCA-f!I58x78z
zjm@%id;gzF+P>U=y;;|rweD@T+?2<nm0PtRXx4djCArPEM;hOieveh}{qy(PYRZz^
z{VT5*Q#~71ezwnY%MD*;cwTg@_xz}%+f~{Rb@uzuzi{RkZ+|pAZ!Z>gJyP#;vs&9V
zD&FTbIkmoT*V*dVYL7GLo4D_cz3#}X#S0faZgjs(?#^3<>3AyrPPVh(Z`S&`RdOr+
z4!YI%Q~yZIoqfKmR=;MSW2U?*H#?s6(`u}j&sMc-@4QlJA5`i7TF*{j|7X9Ql~uxf
z-#hp0^B?{wYcJm3Z{jl7e(lSw)wsr#JMO=IweF>Q@w}V%dElT|dX(zL9Ul1Lsx=mx
zzf>#EJ9y)npC8_<)F^!D#%<Slv)@9cqTx~V9r;B6T^1}A6?gx^!+RcdTdz{9vDNuj
zf3<$cy4w2b_hyP(ZuR=9S^b^;!J4a|e(t!uUgR%dHJpFZAt&AX$oRay_*LVr*6}`k
z<wspdjeaC=FV@;`)DH%NfncCf46J$dHNAVTJ9nws*eHF!Dsq1Q((daou-W5z(Q(My
z7Y@B)<6E+Nv68(nSoZcO*I8y*-q|sIZhX`tGxnJN%g6FcSnd9$RrN;h{F4WlyX1wu
z(b%g0A2O%()9B~peYe<uoi(SA&pQe$#d)jlAD107dc&Tt=0(E=M||(6Pn<s?haF|d
zdi!maeVhH=jY{&X{SI$u`M-CqvP-S~wD<W&XK7#Qd(@r%eo_8^c{x4n%~#RxKkJ2B
z<F;J=nXhep`1h_KozLl5Zv6FLZ`JRHo7GRFzK2vTe||?cr(mP=P_;a!^(ws&(5m~X
zUiFrLdGh?9ExXa^yk6|==Ml};uT}54G`gRT`YvU&`W1K8#=(!CK4bOgFP)qh6*pdc
z`hvZ0e<N=+R?DAy?TxQj7rp!PjShG*ZxuHB|7MIGvG@tQuK7}4RBRRR&FZJ9ywkhU
z{@7}Hjs8C^t(N!J=v&`^^w5!cqfvgG)3Mg?F4W3Lwa<HMmD{TCwWs~m`u|>vQoqs9
zX&NoBQTy*|<y^4dzC&O5U_u_)8Xi0ksue3e$7uHb;Rk(u+8!%S%!`VRK9{LhUak4Q
zv)147sqemj&>tq|ft`zoK9^mz!MHsq=0(M#%scJPuhl+h?(FrN_4|Tm%kAvXZL8JK
z_b}6W;_u%*Y~-{5x<2b@O!ca@+}`I}wf2|RU;X3DpShve3)xp4Q(ih-xz_b_XQ^NG
z_Z^#^r&}do9&ImHii>B4^}qG&_r8&}8Y}5l@BQ{ZN9pMH^?qNgv$X3h?;+N!$AR}R
z@|Vlk8Iczi>wW*Av$bpW^Or{T*WT}l)!M#No~-nL{fctG-0zc(zw3ijp4{P-Nm+`&
z+I5B3M!h~R>uB8l$w7;CJ#$P}RLmav;|i-TG9qghvSU$jyCX*Kd+X-UW%Xiv|4-eo
zyS#kgE+b#f16#{Owa-y&t+)1{e|*7x-JZ@`g<BulccTTjdm}3<e!l*)^DRAbLRJ(!
z>+4@%uiqP!vZCXlS1*}0rT0r&XTeYRo%cHjY%(UV73=*@OlNJE{$6-aDYy6kT&3%Y
z`Fs3v<zq`dOPv+Fee2#qC$Icj)@)39nN!Oh^<A!x;<MKG$U0m7YTpZLR=xJxZH3o%
z{{C8DK9<cX*y#Tg+-!M2K5OKd`QPuK7ai{yc;I<QT>oHRG`w}OD~GMu{eirwSn2yJ
zasNNxp=ocP)V{O7Pf||5TK606t>5V9!{x?RwCk;Y-;>{VG&XDfQxv%!&42X&)nK4i
z|DP;nvsM3JLqBQ%<;wRwvD=`{w-}dI!oBb5wf)fd-^r@QiKF&gdAFZ@kX4J7{!c(T
z{pa-S-7fwzKlYpB@}go#J&!(oqp|xwH+*F7`v7y=cV*^uzc;@2uCt5Y9uFObXTN{i
z*lBwY$Qy;#{-3aB)w^h1pBuJ&XIS27?6LFDuD)Q}xV%x=-rs*z+rHWFzc%`P2651;
z{~y<^e%AZhokza-;?r4C@S)W|-+sexH)i-Rht>YBq~3bv`u(b0A9vYFyZ-#km-6;v
zqyNurt?$PawSGrEe@y#r@B6)-C0+(CwfdIVZ}M=SVzc#Y)^~TBE%)RHcYo-wm;5R#
zI<C6qZ=Zf|(7>#yShVMOslJ`%_a$FdJsSNTmvZtNokyGf|6GeI|B|(5Gd<wg0s9}a
zV&6V}7w$XPIRkba(0A_t`Flw3Ll!<{(Y{^I8gR(s|5N@>-?{qEwe6<|pFQA^YtG$l
zRMxf8b_#jHKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t%yA6-bjtcC
z&Hvf7JTQ2e6a8T79`Ehe_oVmpqT(EvaAH8gKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&
z3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ
z!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`
z1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9
zFc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t
z%;5}Na{qBNj=XV19vD1?9~37J{(Z)l&rM(Dku0z=Jj~%~zUtvUwfCW`EwShOd0_Am
zeo%WHR2^thF&GF2f`MQl7zhS}fnXpQ2nK?IIh=vF2MoSz@=EXIfz|Oahr2@zFBk|0
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nITkfti=C
zdf0}eN9KXSLudAb4h)g#Iv5BBf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl
z7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>
z27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?I
zU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}
zfnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl
z7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nGtuz`;{?`}LbkeV!E+PyBqlX*a(+D+>%BI+Gt1G#!N6
z!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DZk!z?%P9e%kz3zL|Aata>0(F&GF2
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>
z27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?I
zU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhR$
z#=v9qzdGdnKTXO4|NnVt7~vr(7zhS}fnXpQ2nK?IU?3PMFawvYJzMz!#}3&4kQMtb
z+;@?_bDc9_#{qrk{-3{x^gd+aLl*7Z<*WgREdD>`@ARFk?_Aq{dhpo;4!P#s%|>Ni
zyIk|vsoj<y_(cx4`tB-I*8R<A`7-Zp`N}_!__Ph5JnW1I$87g;zT{z_oU+hjGxN=F
z9rK4XzV%^S*>&iCiU+QL!L`p^FgP!V+b(q7Ekj=)pAX*Py2XFibx8ilZ`^e48ViqX
zD?c}U%MmNgKRI7<hh8&!e&@~n_gC$)<;f2Y&o{by!s*-0d@1if?7odI-Q>}%R{Y%|
z#}EJek6+HpA$i-Zd+w(5eLgN*c!MRzTsGwKHodZC*BJ9z-&gZhUf5&XhfaPsKVkbb
z?%R0kgtl=WeRQv(U;OpH{GMT(9>4zyPh=COY&7+fLq}%6`qkmv_n&`4_OF?%A9cWY
zCT0DP@7{I7eN(d=*1O`q>yLRWyJgUnKD#{iUiS8H4qRx`VbikPuk82YTi>0TZT-dL
z<5pYu%WUdrcO81^qaU_i?{ejc%a)lwE&KG|ej|SS`**S<e$Z><jW7M6O%C4iyW3^k
zp8M$~cV@30Jaxdtg`dy9cGM?39<=@)ZR1*Z%3XI2S^dQfsyA!hh5L_KZ*sQ5p<Q~N
zboMK4c+&I3`%K&Xt(=|UpD#S{sXIQKm`}O-@JAL~{gpPJesb@KKRvY5n{B+)hknZs
zf8<ZQJwL9^uI!8-`o3}Kxm#{Ecx<-cvqSG%?XDr&b^qA5-?SwsX5=Z;mz~*@Kcz@t
zXgvI`9Fhy!IqiSrx3@iTwXv_a>8n0{cU}65e{}8rLAK-fzcu5iTi<O%dKfRnON!*N
z-zC?~_;|u0W83_bU*Ma4`LX`>qdsIWe5c56DXukOtpj&@<-iQ4_m%ncg>OB+^8SB*
zwawqyk-p;Ln1}bc^u2$7+!hz~7e}Y>)&KG1mYI?DxccYYoPN#cZSk@79j|<Cxre%!
ze!AV~o89#9E~VA??%w5xyM2`1xy<dWUU~Yb+1VGozQa<V&Q<#EnEP(Lu*>u|e`8mO
zXa2#i?&C*ZQQtVr)yKd4<clMAB~L%>if3^O<!SYH_~X>M`>!`8KjP4zZ@&4MY58Th
z?6lLMx8KjzhuXQX{xNSqa@B`x&1%ad`n_vlzb*rMK9HFw=qK*@llts$+~S-+^E-UV
z2Pu-H9n>G>Pvwvt`oPqV_-79&Zd2q>^<uwotar#E4}F={i~PGBnup3^rGBejW8oct
z_+Za@migBE&uulUZN9?4JgI(qUD^1??tQ^Vz5n=GTU?7D{>abhV_p>>uFEUt<xyic
zov_!2A7=F7m*W2Hu|t<x`Pfg|@`&s5wSJ)M@)X1ayz{w7_x*j(uBE5mTWiXb7j`Xa
z|6e%U5aQH&MLdf`^WDFfy?^l`XT6>qx45*PvF=INL+YofJ)ZTKBK}jfeiCoJpWA8b
zC+`f)%nQalYxeqtowWP-yT^a)$<42Cn{WAp`QoOm=T+05n7fU~wO`w5kD0fBk^S_l
zDGT0v;Kyz2sc$?o;rRzw{363UyQjFt>gPXr<bhpEFul)y<q%&fnwL_fKNJt}kIN2R
zbkHVqm&}tZ&DwPGr&s(t<CiNOH{*9(Z}CQkKmLtZer<j{=IEn_j{D_Yr7J(*=!X42
z{Up;5|AFQQ=sJ{Fp?nNe|6qT4f?o8uUKM}VU)I_D#rjJ>{8_%HpLtduqNn@({L~%4
zarpIDyw|q=7gsOtaLMW;_IoFX+TjE0H%0j~^#gg0pR*VH$xHH^>+*_m;fXwX-aKi)
z!Vkz*pI(!e`Nn}i8}&l&I{V0T(0%y~pY(z3WnYvcyU|O&;0Gz<(K?Gfd4T`&Z#>}D
zxYPMgd-+0Kh)+BkH~(b^>rwoWLoTGZ`3tJAU%cvHe7g_Tv#xbrfBlAUbLAdCoVO%z
z=nslJ>*f^k22<YH1M<^KNWT=xN#`|X^WL$C_nLL{xu557=sUx|cEOEb<|$sW%5}FL
zy6H#xKi^t+|5ZMo-WI>sZ?219>t5$4+C%vPUitU!U)kyOxl8QFPRiy5e5Yu>=ePVr
z+(CW}%|qgb9m&yeipFhSAiufpyhhw9LwQnuvwoKs&C||-)F)ru$QMU$xXMmzY&<N7
z^0WM59uRl(E)>tuJi%X~yaMr*BK?d*Ue>?+@-UQt%OSr`5r37CovTIuAwP*n<ALs5
z&!zI^J$a2i`5QUfvzz{mi=U~FPyS6ly|vd5p2dgz<}Ll|C-o!u*~h%Yk0E`Wn`ozh
z<Hw8f@JIDhKGdTx<WI)KKcRN~n105ANBW2>>n&yau?NHxJ@l&|{X_Nej5mm{YLTAu
z!_9r}UhaUM-p=QJX2cHXPW~j{W#J#}wfgU;{#V|%UttgD9bf$M_)WGx=HLHyerf#j
zdYb>tFY1{W)N>sVX`HZ+`5J%h1Lbdi!tc%F+UbXV_)|IL&*e~@*NWm#KD7Tz?VI+?
zF3^2`&X0^6KX|cTwNF)_pW}P-UDqA`{Y9?H?IXn%{glZwpQJp~kDc)@9>{_0Kri~I
z^2Mok^i>}p+L^zMbJpyA@TQ-wyxP!nrf1II*ulJMzP0}ueq8^j7Cq(dOg;U1o{;iq
z9^@bNG4InGkIKe{2lw?ye|o5gXZeXe*g<|$KeYpXTo*^y;o`@5=&2w5%Uj}7zvQY9
z^$WE(U(g@&6MU#=ov1(hLjAeVpY`iLeq4vf$G`AHAM+@=@)^J2|0%L3KKZwP=>_$R
zPlyNkfuGUGe88XOVfyfUe(1jT_)qI=M=m|Y4|Jb>?ThK<d6@ajc-h0aszvtFpK%!v
zejz(3v#b934Y`$2yY`~>Ex!|2{94&K=}q6%uPW)yzPoPS<>C2unU>p^vCAh5tup!2
zCEsX!&TOAz{lY%_QHJ;>uM)~PDe@2eD3?R_g{fWSIVgV1q5Tm|>&XYuxFI`1`3#C{
zn2ujwf%g5*Eu24Ai|oa3tOv!5c~Ts?Z@!Ul_y>N;<tNsao_9GPus(Dip$ye`ZlE6j
zN&Sz%@qhY=6Z0p&jWgw2-0>Uh0P@oDiyw03bN=Ccl^<A-L3x&b<l_(0$9#~kFFe=f
zH`Xc6x8)7`$S341`p}lk?)K~VGkR<H>ksd{W%d21WO%ZUwO%C0`iFm7Z+UJ&K0WE@
ze8uyDYLTBo{6Kn%H-4_Z`{paWe|pI)eNNl6%V7JQv*ugl+iN^9E9bYXpLh7Y+gvrX
zt?asemi?1Fsw`gYljK|HX5z&DjXeHtp5^DzxcE7LG*9DO-eNy_)3~+g=lqj={)xZ8
z4&L>{>zD3XTDi*|L$-duTWOQ!mfzuw6S}uO=b3fK{@ee1=6t2W<DTk2=FdG#$4+~2
zg^P#I+g5f?>O9oGZ+ibOk8QQY=b3Xp_u0#J`LFFc)a>^Yp4#{u_q^Qmv$pph)|&l(
zh4@STpWJdNKJhAV;8&TSkz>6hPg(cik6od9?3yC|<rncRAGof6NH6wH$AKU9l-ZqK
z+?OAX!+mA#_=|Sr(o5c<&#c+|G<j4#*X1AcwSJ)a3J>h&ITt-W@8AdW5B(s$<Wu93
z-<<Cnhj@bG+B|mJxf7o0f57{>`I6nOyT!M9<~#YrdnWP<)X(ih9v?mJ=hNHPZ|=)0
z@{j)b8^7Rh@)MMYl;tP>Y#i3}>bWn!(nnkvFC>S)DbkOfq4fznsi!?X_^0veH?=pr
zdoIEb=3#Q=O?i?$@}c!=qmkbFrI#}O@R6dpk-yw$rxfKGs6X?CxDy}HyyG0qI>GZr
z{j;CvL7vZBk3j9Lv*}T8-C$hi3%udS`5C$LP9^l-gS=$?vu5Wpc}l*Nzsuzt?d1>k
z@gk3^2g$>;`s`%=B_CEo{Zv9cLV8z=+NH>!szu|3<_~!uAG2oLg&$hqy6=5r`zh;_
z7fyNSuLr!;wZxvDH!AZhe&jws=2tNFe}3Cz`?F8}<%-=(<arLwf7n?ZkYiqR-}sEn
z`p<QGu6^Mlzkg$$?j`4F<Z5Re<d5zBm!rEZF;9s+>oEP=cZ$mt?FW|H_>vQP4xiQb
z9<P0X`;R>|e9sl`p4s+ZH~IEq<jODRVSYzn?Uc2{FM0gLI@5l~b2Pl+i5%_3mvND=
zzC5NKxyCJS$zfM~c<-RiKWD!`w?+Ts&pGL}kJ{eL(l0suhTRq!x5;J;9oDVwy{`ZA
z&)M&H=~sTUe!t|}$wwV}NsrPu-`~7nzj=C<_z}N8<<w)}Tz$bF{~IT>*Wv1sr#`*B
zUorl|_n*G&sQF6t|2!Y|z=#R+w3WRN08=^NIA&J2+i&kyGA?n2Z}uf$|N0ew+KWH?
z3H_2IPe9|;uYO#I?9D&<iFoYw{EH*M-GBblueN^j;raitNa>P~-|I19_pg=yGT(>a
zUh{*cN{{ci=*T~O(!0dIo_B}`^~EE7<tzRyzq$^MTON@A<U`}Kzrzc;#wR}PQ}AM(
z@-bfMD_-&L`a!S%qx<G-%=KUS@$%)@e){YGUa(|8D&E*vJMRO+!}jgE?}um4+qQ3=
za?M_AU-i>@O8oKP%iVhZc6ZI!wx1Pm8(#C`tl`UdDe-6V!C%GsWhYJTd&gg9Wby)g
z<AvYQm*3G35BOvcelJedlfT*3`q=tP-j*l%kN(qh0rM#2XVyjLTl*h*lODz|{^WIj
z%Kr4Ck30rl=V#=Rlk&*k>fy(H1^E@frLXzKb^NfacIxv}aV$Q>lXZr5BKh`{_`|dJ
z{q%=V>p|D?@ld}LzqQJ?T}#eEjGO-Yrx*Vdk9by(J(TG!j?G){v%7iAzS=nSYaHx`
zKk*OAXLt7FKibiUd^{-gbN$hm9^%nDK;FZjdg2e#Lp|qJ?vv|2`QnFP7zcgX3lHqe
zzl_5<DnE3cpW=@_^e11Cr+qp<r1{Et?4#`q?em=@C`0EGP(5fr4atZ0&(QfmiuTp`
z<InWApJyL=lwZqZcywRhlux1hP(OI42ei*LZtckDr|R368>hIZm-rVS{FFT7Fy3kr
zPxgKMhdlmcy!3THMdPF=J>)t38=v;ZLl1Fo+}gPh$<eR8s*Gpnw)UOwLp+FIb``JU
z&-mGeUgE=b_D_-Dx}MGl<jJqf@~ZLb$9!rY=dbwjyaxa7Lw0hWLNC0DL-+ZovVJNd
zp5+}p8lV20Q;-kk7f28BB@UeXsFz}TuBUy9<mun|@Pj}1^$W>UPka6+-=(M>BrnaE
z@&)<ix{gO>cH}SYp}l_arGI7T*`9lOUMt@8hhOdW%f8MjttYj^k2qA1{&=FN=hE&&
z{ZCzaiJxCJ?~JzJ%ZpF*5;^pv2fwn8ric6LL-o~T2le@j@$(n{rJi=FJZ1Xhi5&7>
z7ti>G<h#%A`~dIrA9UU*4_haSTPR<eC**zWWAe-|%Fy~WMe_pGUy8<G3E3gFA3mXc
zNnh>oL?3pSkNF|HXy0h8G*0)~w-Vw5;vq%%&10U!7%voGkiSFErSQi;&HMbe9P-N)
z^<NJ4R}RaytM<BaK=BRD<K-|Nk23$VUMh$Dv>b~6a>#$nVI>~ew;YzUr|ZsJDq*?t
z@mK3A{wiPS$9~m54ZqgA`j_{_J#^jvQa<(mqvxfbN6O>;j-L2b#yh>0&41-k`&!ZY
zFn^Ud_?LL(ud`-9PceS$OXD&gWzU_hzsaYkamu&s0Oc+7i+=5w>4|@NS-y2XVjRlm
zCF@b+BZt18^Rk0DHV*Tb^Evx>^;5(%e@u~omP39-&vJ-A`YPiE|EYd>Vn_MATtDe`
z_No@~MNjKSW$R4uIY4n}9O8oCk>foI@u(~=trwy7AvxAv-rLeoiu!}r>nU1K)5rZ(
zo;<?d<jSk!N_%?JU;e`*e^yT4#}J3?Ew1UsAI#gX>tC5%{m}!mm-V$c;z!~`J^IoM
zidXAVWqxj-EMM7gdN0WMoX_FY`1qIjr?1~>Sof$$Uv|O^KFCSuE#vW?kM|7OlN|mD
zT^AqT!!Zv2%P#Eey8RP>GH!OjyL!f{9=oxZb}-E=?5{p~t~<ZrH`?<%`n!+Ew7*n7
zIZ!|BY5!`SWnEx>#GjnUxUN2ba-Tn%m(p`t?bRa(-{wK%rZ=AO2Jw>efFJT8dy}gj
z|7BnAZ}As?q&@j*p4Sh#`~=_lhwh7C^EWg<>KE!iy-#2BgmVDxeCKcatp^U>Zl6!u
zelO|0Uwn#3=N;tqxp$k(9~j?dFgfh|k7xS*ewSsx%u{4nNWYFo`!{yCUuO?~DSs+&
z{?Su=UUb!hgI621>x;WD_P@B~eUtkxd%*(#b-nArV;`S9?*fD0J-+M4x1HLn4gdN5
z<yYI_7xNDu_S7ryzx<E++uD&c<Lskuy7b8IgD*br{U;V#V4lJH-E)ZzCOq@4-h+=`
z|JFmh-S)pIZt-kfP(1Mi$lswnD$laJ^DpZF^DI<<RL`5<I^fg~a^H>IbLh5<J^a@X
zbJzW@!S8g*{lSN;-u%63a}8d5{`>d+?1NA9tGeHM?j76DGx*^3dJnqxi@67Xx!6jF
zuYG6t!Saf{WSuDP%zK`X%7^NU7iH@`a+TTL`quoZKk*5z&pp@j{4Twp%CnC5{KoS*
z=(#(2*4f6Vp6i~^doJkx4)wj?WZmhxFgfZ?UwzS88(cF_37V%<J@lI*InZ-{NKfdw
zEOZ^>DMk7~&-u!sakCTNy=Sy@>7r2&ec8R_c?VwP2maxl50V2tkGId#ulDLe_4pmc
zyZOjGWc_Tt<+?aFKZ<j5tY^I6OpbNAylnk#-KHP&iacPwCI0klokE^<pm|pu$xrHQ
zpU#iwS^M8KUfI)gTz-uY^OXDikloqAxku{Ho_F(4_GMS`&rigSdB%Kg{cPO?*$3i5
z-o~S{JcrK|@oJrHJp!Hkm@mlHj^9`h@B^5(V?W5x$W=dWFF)&#e(GD_<0bVo?I630
zCwalVF0S~WybiTXk$+jIk;}f2p60!-+noE+h(F9|`+b7=F)nc=4(-duoBg_Z)_e@j
zyXyPhmU(o{c{4U!Y~;tec^9f@9Y-&5&wkbe{3y*s%G&vz2Yuuvev!rjf6!jPshzE#
z)t3*%3AwKGBXX?Y<PYum15{6bQQvw&SwH1aKD~3Bh1dD>(;wyHNWPSpjN5(V7dPVG
zxb2Vm3;$pza>S23j}PPHr?nzKcP=B2=qJv^wQ~aFhQ_CVek70Kk3Siw>&^$r6CZfM
z4^*E&s)sM~`MG?8Z}s%&K0ES1`iV<C8Hc#=dna)xp50frK9)z6?av^+<qvkJk21aS
zAP?b>UDU%1z3{+p#-)G#neWNt$ND8leDY8JBR=^9dD_uWyf|0j-{wc>Tk?<o>8~Gl
z6gS3eT>Kdt4}H|9ue=~W%=`Ee_xck*p1;_4>mN__)h{$}Tj$90?&Bjp|E3SUt%Iyz
z%HaV!44gUtpB5-V`r@S=;sf$?^DBMWNgR<Up4bn6?1u+seuY2zTYgVbzG5%_$p7RG
zamN3Z*~>WSFTUsx<t04GH_F=aFaC#bJn2_Gap%7N_z%RZagr~e*)OvfJ|X?ZBYBV>
z<g=Uc(Mvt|>Fv7y=qJAPYdp|-q5Q_bp>>@8$SH^By;_l9TVLA$+ee6ZdCIw{^@jbT
zdY;$GbF-4)ccpexPn_VBJ=|w!_DE6wP+#20C-n86w|wmVL8zW_*oWCi+fP}4<ITRx
z`1mJ%=wV-@f9>=OwRhh>3_4F^7kSnD#@7G#i}sb|sYfs8K>Fp+#$z0iJ=mYWvo};v
zT*z~-lc!(fXFq71<SMhf>(2S<FW;0y@gxrQFK^JZ)rcqaBmE%%x4x1m%rDxxF0RP6
zU%|V0vtGfAe56di^-9{Fo%Kr|Io5f`VSZvC<8s}+VZKskAJ?s;?N{Yj_vL%{_3ys*
z4t?c!{h2@U=)Q58mz5zqkZZm*|Kkxa*4^|a2S5B8zwALT?aa&cXLtRv6HN23GQTiy
zi8JvaZ-^V$)f2zYd9PjL>b1JR(xde2pZDy)#qtZ5tOMmWc~M@Gcdd)k@s=Bpamfef
zCwdtl|CPU>ex28}o%7EAJ+SjPzt@m&)z_cv>X~oJcMhzc>*f{bf6m$I>vw<7gQ4%f
z(APP;`COSj&RghvfBn}EJxYH6XPz)WIM4Qb!&E=>iFw5QN{&1uew6v6{LfCVn<ve)
z>@3doXWmoR-ul>jh(CxA`+0KhbH$<f7N_FK{@OZQUKF?1$Kp}^tEWt^d{+teFRmef
zz;BB31b<IaUc!q!Xq?93yc_@GBt>y*odo$ayNW~W3;mMEUh=r-9o|3Ij=j~BAL(uV
zBfj{b`32&UJZL@4KaB@p+OsFS$n)ts#klY(-^sJqb@HwJApYrvCwzF%(RjtLJW8&5
z#;YIuNA|-nzm#|7d8od9fjnZ~av#bM>@P0-eh(jb64&_GFMHC@dVrpOXKY-~FT@`n
z@j$NYo|jl(>Ysn{H+mYcd}tiniyQvWF7g+<S;z1%Jh2-&<`HqFUw&^K+8d|m{rtyy
z3X~7%=lA>O5A!>}QBQki<F=2MPxTAsCw7!qQ-1NRK0ClPugUu<@-u!_4viDywGtYC
zCFCD4)e}FEeIfgoL-VhFEuO@ob2IA%^SZpje*B2N<u&WRe{A3Tfo$KNrBqMjlArMI
zc{O{=uf9VE#RdDw+xqeS4Sr?)NN@dF2N)lH<X>?MtydxYLwN^UN5RxT^qZn}kGSO5
z-g|`P$z$RlfASLhxDQ=tNAnE1<}G<b9LWQaKIHKy*NsCy*PnTWeEp~Gv`^zeJNqAb
z!g@#<$~X3R+Nm$Us4vgDZXQwJdd|FJzSmwn@I&h+_sMtNIw!5~K0nYu{mUV~=q<i$
zMRwA^@v;|t^E>`4F4$YY+ClB0{-Acs<narBAikZ;xv!lv{pbztXXFd|=uh5}XWZx4
z%GL+u(o0@&oj%5g5AyYEeA-!$umd?M-}0k;s~!1ieb@D4zp5VOSN0R))%uBE_EF^V
zAO4Fsa>^mSYDIeC4^QH+S*RZ<ZyK-Pu~`2bm-?wc(BF9!9{H8_<&eDsXESi+fBxUJ
zc^u+X{HJK2<T|@5<6pdq7xJNgQe@X!k>Avc^nvCddh<_lB5#Uo<1`LD*vH|+ed`13
zDe|lr<xlgXex1XJANd7O@`-uQeevSlMm=)qp`CS)>()0Zuf}Cw!IyEmZhXq*@*8{@
zH&jnPl|S_Z%@^e0EgfIlpK<Ds9Qy_P2<uY&0Ot($bI#vj`ktZrj$G~4C*OVNC%^e}
z#Xs~PG*`*~760_bzj4wZPp-4Ka~J-^Pw_^sb-yzIhUOFd^jWi?Uqbe_AECGPwspGS
ztyovm%lj(g%X*R@vm<|Yzr9Gl-%(g6dw%43y>+v7t#zmR<cKHZ;&1H6zl}$H@CSbH
z{GWfQ$KLAW9kK_%<p=r`r{u^#u3LA}8`4|8g!~IXelOs862y~wcqAWR__H2z9luaL
z@?Ga2?n8M|-ef=b?K|WV*YT%)islW!n{ZxEA9C5#dxh?kuRnfg-0V^<TK8J-@*DoA
zKmL%S@yP@7Ex%0lh;>zS8IaGN2Z>Mlo4xppc;esg^F#i?p6tsH)OTP1;zr((Z}cbM
zlcRt0CDdO2l2`O2KakTb<lhi4wW9qBdl;Xxcr#Ad^@}I>&BKr#Dj_|f=R)Gu`JTKj
zPpI$zse2x4zpw0kz;{ON^ZBE76J%fLyp<fNANk3;n;on>@Z-Am4!`6-{1MN}_#-DR
z>)(Cr5ZBpDUKLN~XXSEODNC@EIMSc8@uq0K#SWei@WWIN`<Fw!rDz>kD;ig|==?p!
zYTK)4JzXuzXZ0dKu%0T1vyydlwPpK>O33czkpGs$a?fXKwR^SoE5(KNSc?3*5{gIb
zH2GE@G2hwm$|IhOrq?|eNYOkcZnRStr}jVM8c*WU`GMz}=4XCwyz-Rx;y?8#?d1)6
z%8Sl*=)-UH$KUKn(|LnF#z_ty>8VVPIHRBb<XhKM`Nn7c>~|IF(T{w*nE&{__|y+Q
zQ*@s^{$jkT|KiU)h<AMBp&Xi*_^I^+|2Hr3XZs}kDev2u7u8p`P81g@8h<*jjxO`t
za)_t)B3@H0*RSjRuN?B%a>(8(@)Les3Hb>*`m?Tvu9rjgpzlt=lxO?<YEi$n`jt3r
zHJaa>h4O~;0r{m`<cH$dxp6uVk*B`=B=4G^Qhv3wFQG5_@|Jqe{m3!j%BT8)^0EAG
zUt%BQy#;xk9lVzz-`baW&(!lK_32~YOxu~CAUXEm&Og1^MPKu`{fz78bNe9v?!3tS
z<=n{mnfINXZ^`fCQT#Z6a}KJ$_*WLM?Bu$B%p=ZWwKrcVtEWHnhu`HudC<J>eGLBQ
zJQz>xq8<O_XY!u)DgWSC=6%=Ax616p&*{Zq_*shj<)_N@(!c)11K#9s_07ZLRJ_oK
ze)z%%yJ?R{`at=gANl<$JF}Pb#?-Fv>yMr2X&p;1{TXlC-uTmY+T)GAVakv3rtQ2R
zLXLJ&eedhB1HR-V??V~4c)^cx(gQE#yKbE9Xg!?zxBl^}9=_;lKTocGiSg=}-uh(+
zc9D-~&E9|VH{(uua9^I3_r#HPwENmSzcF6dAzoATJjlMv_?+j<SK6@$q>tw~@~icT
zvbb>0PTn<to!V{bfnVhQkD_%BetmC2{)Nu<@lCGhM&!bw?+pLi1vh@#){p+=FXMsy
z53;BJ#f!X<`hocd`d-tPXHPzS=-P7+R)&w<`{GMyf4y7Vb>DHKxBoY+zVAWd*ZAcf
z{1_*F<sZmD_`Nv5zj5&oaff&QBA(ob;*C5g9@Ha;|7$1Dt4~k&wS(f#c-7Mmx(>x}
ziuUQ+rTI<%Fb=%Z!+0V8!IyXy&(L+_){inh)Q8$x*FonA{1J+O{%xPl?~O-XKzZ0X
zF~3QX-uPipda^e^y!q1ox7~KYJcHHKPX6;=tGt5`a+Iwvpmn8njB<+lN!Kss)~odJ
zoW*)IMeAU8fa=jxK62kVs_Wi+_j@n(^zV1Vevjq72lvzV^w-Y!$IRR0DEt3q%6{KP
zAI~j32iG6uXW~H~k!RT3`du7ZZ_9V=E&uXY^2`JBqI%6j`+{l_@8T1R3;Pyvtv>(e
zkL)XcoL}-6e&zdH`mvsY__p73ekJe8ulkoKyss%w;M@Gq&aT^cv6KCkeHj0M<jOnj
zZ2ch5yDq<~Z#|(O_vyv&<SqJ@v!i%S<?&DL%)5Bhue>6E%SZgu^%Ui6X#Dnt5MTBk
z#>Ia0V?XxA7d@1nXNXJk%$txt^<p|+b|p{0t~)2B7dz4in$L~19P)=0#hLgrZ{r8A
z?&FVM<}vM3bYI-lo1IhCkNrtH9_@|8{H!1A8NA4&kRIZkzuJe$L(XCNuYBXae&jFm
zQ*__FTMknml<5hrH=I{F7vVQh`&yBGt$+Bf>-<#Nc@00&zx>8;T~|)!sINbMq#pU=
zn*Xt<cKkwr+VdlR;XZT@gJ=EN7m!O2=L70FS25q<j~|<t_&5Hbcw=Arp5E%=#r)#F
z>#6?wbAIIgi1d9;{Te5^;tfBp8@GPh1ByrX5l8xA2dKXFo_WZ4<S8hw@T#76)@RlS
z)`#@f4qCriZ&=S-KhqD^8dp(&&St>6#Jb76F0YAa-vPB=vaYf&vyQaBht567bDplO
zKkFa$opaGkSv~72?d30VE+2^h^uBph9H}oJo!i;x$S35xE}vSD$phwDdBl8e-j`4G
zo6ZBSi+lIY3)-isKlS8&`9a<y&v@iz`APruCKt*l%GMY1Xu6(|cg+{(uN3W1_*q)!
zXZD-wrPtke-bO#?mg#-%cU<i1)ek;??!o-me!+duBlqch#nDI4{dw;Dv+22?{f+C+
zVIjHDIbtPb7xObe<j?FVZpDwX_)%sL^R#sy`|%%`+RM5KiW711(umFX8hm}Xw(tM@
z9=`LW{=1&K$>_7X4#qq5{~4xrTmC1L7V9!tzF`-B;r$N&4DkxBgQ56<@>n_4A7o$1
z&mcYV?3|r^c4))hdY*V=KJVaFW+Oj^>-V|$je)=GRf6<@><;x0owq}N2-SmlgY<^{
z3W_I4uN19=tnb97yd<9~v!}cwU$Gm1Q;$EB<2rwq-?dYQ;()))FYKzUAAZUIt)JN;
z)zf*8eCWP@joW;uALkYJ*Y;25Q`gOh^wiG$hj;y$AAMKEdfU8XKEn@r&a?0=|9CDZ
zPss!9mgbMO2CQ}9POlu8rD)vZ!*ggn<4-+&i#O{U<8a@49^!!?s&8I1ZuQKE*4Nq@
zmvQrR`9?eZn?JSVZ^lQC`I!GYe}v}swpHqt|M@+ubxkGYH}rt?FNfAK&dJLm|A71e
zns4xK-24Gg>?&^Y0r4!}^@qQ5$gb9-_~7T_S)B4aJkd|w&>uh6h3rc&>sR9;pZ?+%
zFZ$6xz3`+S|8gA~Cx1xwb<V6lJ&lJScn;zFP2@sxX1!$IlP7)0-#Xv?U|r?B*}S4m
zj=W~xw+?ci<9rq3RXfksp!@WePw?!zxMUak+I{wAZ{x?K_WVqK=V$!89Lj(4miV7F
z`~3xeq|D#Nhx_a){^TXs)rayMl=t|f{^U#jibp8U$yJZvKzfP;yoxv1@yUMdrR@BR
zJ@Bp_`-uZ`^&_6lXW|Nyt9{yzeD%bYdfwkKKI?erTl}AY;m0`5i~Lo5k;_l@@4Ei9
zH&5yZ|N2Y)Q9IYwC)fD!ryjlWfJdl3G@sFn9Dd6G_!E8ksrKTJeU<5{fAPSc)-B?b
zUi?aX{^mYC&CB}bcjAYh&g*@r>ZYvcRnwlByW||u`<mP5o1Ze@EnP}}4{u%Oe9-r@
zd<W3;Ugv?%2|X8tzW=q!a?9^<#tGd^=5hT``RJfwi!QrR33|@#_i^S8evqyweV3HJ
z#w&lx58f-aj#Q6+>`jjRp-i6f`aYQPs3+gaTl95aBmVTqKa5YFWB+s>7RSzyyidlz
z^yfPNbq-HH<R|nnKI1Yk@OLOqDxv+IJg02_a?W9%mY3uW`$hHblhS>$`Q5nX0plkZ
zPbu10k_+kQd{X>cH>l?v#yY_I%YF$@>?&W$uk^F-5a-&vk0(fP`58at(#O6Urq{`H
zom~0YINc{lzo~xaLwZ1YOx{&?KW)eUzSnHs1?k5g+S8L=*pI%}hsK$nv++~?;}v@E
zTs)?EpMTj$TVKi}<`a2Revm($%izg4&HMT_9z1H#&-jsiVZI<&oI!SFPj-RgkUsk7
z-;jQgA8W^c<~#EozVVaV6aVb1f9rJlPy8$UUW|1-zQiHa&Uo3?x)}fB+j!_HKSJ+I
znXlQO-Tj|Q_w9c?kM!IQdSA<YD_=w3!8E^_|Hy;(SH8pTeJ^?czw{bAbeWZp{UoEm
zbpgHF@U360)bFD{CHCQ0&XM>de({Na^C0^}eDPcU#Gl!X{_0s5neW7fe%V(){0Q1t
zK>OQ9qyAImN1oTx1ApQa4|uE<^(+3^iGQ;r{;Xf*Iq@M5$dSj)AI{t5UGXeGo5$7H
z-v8;7pU(YV|C1*y^hGWo$=Cj`vU=or?qOeyH{+6@mH8Jr;$3^Z*!Rdo(0r=Ccrd@y
zn||gW=NI_UUVp9|C;p5dU-)%hJ$YCE_*Tz*=lFG9J?Ht_^AlzL7?=0f7ybUg=|`+T
z?_kek*dygj8L|uh@F0(~3qAOs`s|5Ec7^=ZI*(j-$2WP(+Tlrk<5wRq?5mzK{j6V%
z8>(*|<N1pA)-m)_PnkaS<rm73zU-`D^}DZl(aXzR-DR-*{KEM3!|v|$M{&Y_zGsVf
zsJ-|x9{n3PJ?QT`6yL@r?(j@6{TWB9H-A?T>R&rZ@6_MelfAU3KfcXd^iKVsUFAde
z(O&-GC*oN>a_MWl#z`OJ&>ud{o5q7z@g#oqqaHrwAL{{SahvwXZ^(z@oqY1-1Nv&O
zfBb7#E1LJLC!9~%KP%hk+xMz(-C&(#-ZZZ|*D|kKznGt`bF7!m`|M&pW4%BgRKHf_
z5Bw9dhqz{c=UeQQ^2RRs7nkhAj?j9?c_Y4ECx<`b$$k1bm&dE~H~SF#0%haTuYAp(
z>}_4<Ij4OG{mElz`$GF8zw@S#=S1QQ|M;c{-uS!xr#?Q!mv;D4CYK!dQ<OK;^(B86
zXZXdFapFOL{E|J}_Cd4v8_L$fDQX}1$&LrDe@7lX1P}k?2j&lXULKH-p?qQeAl~H#
zc|#s`{%KzG99sV3fAX)g_@|e7N_+XkefdRRl}F5b=0ke9p3Y118F|K?qH`<xA8+J3
z-_ReF&*?3%(F0nK$;;Mn(E8Up%=$@L-i6lN?z5Y99($6*p8B`{!jJditQW0w$y4SJ
z_|snhD~@<@xfzGdHCX+Nk9+@#MHZN+?fpD@k)xe;r~cT%y2AR#IJCpJcKC+I&%fyH
zI(g~5l9r7V(%(F1zUCM7<#*<FdePT7$fKX@?B%|8uH%hA;7fbsbe*5l8-L{E!+q$w
z@$27s)YFf+S05kjiWl*NKl0SK{<3~c*ZJbqILN^_KIK1o$#~SW&Vu-~zQUt*oqp6e
z4)yKBtaI(R^viDi1t07n&ssm?MSVQDuAckWS@h)p>`!0)$;bH7zx?ODLf;kjJlJz!
z&r6**d+rA1YtN@WhjyMUziRKf<gD56pD1hZ9Nc-Yb8qsj8|mfxtULDK{@*j_Yx_Tw
z<Xew>@6dOS+KZkGd)`fN{L#<&l<BLT{ao97QnS~`>f2Y6<GkGSAa-!yzEj!r1p2ZE
z<TvJFym`KXCw4Zk+xK~|Ks)+a_bYp@Ko98m|DIorzwrI1?>cI}lKy)=|KiAR_n*J?
ze4m3i9{jIGOY0vt_|!hzEm6X==Og@uTzQJ#%JL6fYxaAj@{RXjJSPw*;#ht4#G(DC
zc$Js%E5F-s%BP+;k}JNUe))m;mJf_WK9+}|@u&7sW_Ny~EU(Bz`ekSBt>e6h!JqZ}
z&G$F&*KeL)C4MW;;UeQU*=(W1x|JaNd0zywFTaQ6(4T(zg8Y@f#)o(H_$zcD%9D74
z?$b-0L-O%*$+eS@I`onrC2`81*qc9+%YV(Y=0)|~*IqsMmCfgnfAdr8WqxIR^fV67
zot=~D7n)DWqo3#D<cTwJ!H>z&p8fctaY1_O2XE>dAAQXW^i(hHPdn#+>3%1jPt~Uf
zy{t#9cf>cI$rV@p!Tjz2FXM}z`q!`c^Z%5+*X6o6h4S(P&+oO!^m}`^{l65?k<DvR
z-enK(OYw8_v;K^Seo%hLlk4J>Ua6n6H<ah(L-NFTYFF`To)nM#jUL*|f8s~HC|hT#
zM=$G2`#O5c`_9v}qd#Q#S;=><*<C;Q#xwoc2O5vEag(RLe#z0l{AwL%ed2fd=3%^A
z&uAy!`Gq_pudA=@K6D-ONBN(>@K61E@4`6vo&MCPH~q*bH<hD49*mEF(F?i``MEMV
z?808wMR-tFpMK^m^Jj|W(ntTs!H?8)AMzXe>W?1$hCitX<vsgLXg}rrz&=Y{L3zNw
zj~}G_FY)BOLVa;(9M)svSbVB4e(ks27Z>!BZ|JQb`Z^zSZsI(~`H=mnJfT1HGkNlu
zanl1DH~#5IPKxO{NTo76o3GOT^eZpv51;se@~txeF<;8Vcqa!R)(QLwKk0q@aq{r1
z9sASA_^jjci%<ISXYGyG`0GXbk!RfWGj6<9>W7}td&kPwA>^^E`N{sq`UOAqBS)DW
z{$Rgoy=7fyeS=?d%ue(Wuhu>CiFKQN<vKkezeth4LHWK`G@f$EKdME3Rx9GM5}N1D
zS9oMs{s+@M&mQuDa|HjFPksH6&mY*ueRd<)b$QiwDF2m1{sZ+3jT5GRp)8N6Cm$$_
z3vnrqtl#*v_^pKW<Im2Ayzl8bjQ*WBh!eciL!O`?f8wXcXP$GOsekhID^6gg`J%Jl
z?>Bs#EBE-}ylv+V&LyCIjCGdvAb)2kehTFc{wyx|Ej!eU=0$!_Uw$rL<W>1vK9GOt
zBmN=3Vt@7NOTPGk;u-Qo@~j`VS5G{U58b!^wO-}7^vARDThH-3@hTq01*8wY%%h%v
z;?4fr{@gsIUwmm#o_6jV4>{`TkA3A!?a9TD`}D_)dgga}lCQqs3(-qGb~8_>$S&j=
z4<6Ed$p4`HgMapszsNTZ?VJm_5AkFDk@n9|_$j%NpR@|)V{yU%Azs9tak3l!Qh(-0
zcyJ#du9L^FQ@cAC6Ic8MU-JH}+2;|Sr?8JaXuZcz=*LgU;eXokfA!c+yr_o{{d<1J
z-p0*8)raJ`FR#0<U-#*y%s$33w)bC-?y|%@CGo>f&PlX`#x3s}FTb#FHP1lujnnxB
zedGc8l6-oyYl`g459#S#K;D6P<OlS?lX~g-qH_xU;W;hi1M*wrw{LSFl1G2`6c6+w
z&+{4l;??}BJ$d>!e&fMAp4}H`^m5;I>o591_H@2Zp8NcV-qsz?r}bkz+IxQy5A=4f
zBoAvRF4zYj_-1cq_2_B;fIsb1evJn&)?4ITj~Opij~_yQs~!H0C-r~jt$R-T+rU#6
zZhIc1zBr<fvUQC(Pf<Po=X^+<nGZbgx8KI2^*2AkGk(ccPo8w&{K>BPSC*&rXFkQ3
z{#?g9|8n0v!p^SqFM5%qe}3pa3-gWk+L0?B*uj1Ck9PF&oQYqgXdPvLql`cEh<JqT
zSP88wobRRkEayn_5P9+&|L}VPW#^>oi8E*)N1p5aNPphHp%>I%eSWD7>Cb;%SKm31
zdi<XKwZk`mCWn63FXl(}e4k#N%Uk60Fa9jQxKE!H)#K;<j~wwrA910b`}|ZMP*1)y
zpU5-fQ{3oJ|MI4M4ZS}AU6<E92eDry2Y*-oefw8-I(_bvvgaIrKWZQAKEBJLbBiw*
zTj}t%@9aL9Ux`!bJ;n6*ZuW8f-nltB-edIp7QZ9%`x5ct+}$~hI8+uFl~CL{M=sZ1
zJ^ee+_a3J*y<Ml5cKR`1WxN<yDhE%ltDp9(z2^|>t4AM*Z}*kShsMJm<jPZc6L0+5
zI9)d{u?t@11@+_saju^G%8-93(~ti0yL=>`^`n1z4L{I*@{LRU7?1ky;{_UzyvhFL
z>mSNz>|_6+pA^kYDdG(e^uUAb@)Dku>1Vx(2jejB_)aIe%IuB@e7Q~zd+;krKYZZF
z{9%0bhT21X85flAn}zmm*0tipyaCN8^0?=`<}34y`NTY69O*j9KF0c$JoU{--t#lR
z=$}6Ph5hMoUQYF*FMaf9UL=>j$+ym92Y$r=%)jD`-|!E9Vco239@ZbewR4@G^hq&2
z$Dwa3hd<JneDkC6;a|V%8+UqN*?SHATzmJG^~-KhJ?-hE9b`xLhV;}<9OKJ%JR1+x
zueg+_<wN-e$`|ZSK9m=f`3pO^Pd|tU<4w`NhJEE5{)J!bS^h>(dZefw9_eSCc)_c2
zxK4lkkT0&Zv+fWt`f;CJ*P;9AI9#VcJ8EbA_=0#gUhViJq>u6O7yT(i@dwF)`h)a=
z`h)s`{0X`ajW<PkkzT%EW*yH?&KaO~&N<{^?_pVQ$Y%%aFmUGle_EhqKdimDweNQR
zs~>qdeGXt=nKk<ym|XeYzDIj@H4m9*^h1wSuJJ>8Mf@2bIo>mNogU=#OYy)T`I$20
z52=5XC!W|Duly47!!#dQ5AhrGhB7_yN}hWBiClUaAHC!YWqzz3x#Y=b`th7beb<d&
z`*QK3zZCIHu6fTm_?2@W{dyj*9>3uq_!i%uKcpz`oEPzD@}PR2XYfBrFPOH|o<7En
zFJ<-?FZ@+Ke!*W|XIFBJpIsookmu99pWZjm+E@BLsq5OO<#e95p7y;9^&x%I{`r@F
zp!&+@JMvsl%dV@hOrAKQhx@L}1MX`FU5E5{ACe2{3Gt;qB$pn}O<h-y+;p7m0@bHy
z8lTGQ(N8(;mmGeqp6@<ruk1cdc{UFHq-FB7qc1(k)n0w*I(_B8lppoTSJsc~$|<@p
ze#NyiRF59yq~mbC9MTUjcr$MP<vMgfMf-8Qz_gySanOVQ`f*N_%HgltrAUtU&TsT<
zT=aF$mG-Nh-(@?`avrIyp8lY76VC_ab$M9+m8Z>L^0WEC^9;`&%=4b(nMbT+l&!0*
z7p<$H@48xNc@Ase?z-Rgd2iqILFdHQlh%jYr)BMZZ`Zl7b*cN-tN6m7dE5M}9X-kQ
zdsOqkde(n<GJfk>a(zEG)t_GMVf_g4Y&~L~r|kKW{-ApF(~t8V=-dbL17+)9`jCq^
zX#Cpq6J_JahjoT}P=EC0KlJflobO9Ryh8mzdO+(pNFOK;ApM~6Li;AjPawa5<j{|Q
z>)*ObJS(SY9mucM_Z}Ag#l8MgIpzWFmBk79<}3NdI*>f=%_G*O^yDw(Sa)0B@)Py=
zU3y*K;wO0J@BD~;#EX99Q9K!kvT<9t%df^Q4xG2R&wkK+i;rqiKiad0>z=RMzvv&P
z^_AVnEBTNe*(aR`mDz_r?4F`|(LCunf#08)hdpnzJ~MA>XWeNXPY-&MV;=V$*|=Rd
ze(MBsJrCB-JPye*9_^v~)(QCGS7|-<#XUdeSLO19`cOP;FQ1jmbM8C$Op*QgL0ZPQ
z`^v@#U007B<91(JexRTF)-B{f^`Uy&Lp(q{;SI_cDevm1^6+n+pbWJmPyg&IKNzR7
z@j!C4SEdgnSN=0D{cC5O)-U9eqYSm@&-7F#Pkr+YzoM7^$)lfj54|CIP<>_3iBiPB
zddloX4x|q>9;jU<<bNqz&pH>S4}S1vewCN?M=xkxX+1pRn;!f~KgLB5>tp&s&*iMk
zt*foKy!Qvq!|cqT^t;yV?_BUZ<002|>pXUK-S-)e8o&MhH;-MQ<h@(^`klFb2)XpK
z-Zd`uAA4x{o-5ovGsCxbzUxjOeEI%@b)EN8yoaj3-?OSuzHvczqnGRSqK|c@{Rm#I
zcgfcuec3_2<#*o8aUVZvT&Tz2<X?JNhgfIg8$b3Z)?bkQ#XGdmgyx4zXnf`BndhN-
zDn;iS=IvV1x>6iKaa#`AMV!d5%ISUA*~9qhZyjP?2gzl3^N)N1T_@jppZn~tU+>*n
zmpT8j-m-494omNQK50E^{q6kUK7l>hnf&y5hW7L{e)_s@Jp97C(>PqGhce_p`gc7=
z<2DYzJGXAdhkjEu9^=Q8@wlGmdwIgXqg+2JKln91_wk+TXT0fkehB#;z359n`zig=
zL%;N45A~GkVI5B&>wEE$qITrzPg(roIpv@KiWl)|f5zX)qp$J0&(7rF2M^W>%AOz5
z$946r2gJSm#=(E^YTWpP{Fnb2hqyN$a>#cbx{e>u`{WzrQ5G-SLH^@>!MOB;NBpKJ
zPeXbeC;#A=?4e%DH$QWqzuUjcE3WIub!F$g&VS_xa$JY(!GHD7Z|F^5_Qo%JE8|t!
zJS}eVs-5}G@2=RV99pmV{gic!b)J6st95{S)-}rXQ_p#!>-4kkwXWh<#;ZT=oCD#J
zUbAMucVJzUwzK{r$9U;s{X>p*9X_ldwX?3Yz9Ls#lWTm|X?P&dx>$eE__Y_8)}z)d
z;!7Oin}3Nzexd9fpT2l>u7+oE%TM{4di;^U>sNfThkn>sJ<mbpMe?9_-XB-SkNnTR
zc*l=86@U8GpK-gcp7n>gRMs!v?F+2a@l1~K8c*7ve67BGB~IyS-=kml;1}eiNMHV;
zzVlZ6si!|^{^SSh>EAp`Z)N=$hx+oYJWihLc*ckE;x*+@T&Zt7DYB#M%Jd-L^=eTb
z@;-!fQqK#$cWplQ|B(ED6ZY~R`j8);KX8S9-3G6+Th9}3%;z2K_Yw!M*L%>lU(7xD
zw2{LuJAX{i!N+a6=u7Vn>^XSX%ij6zlF#;NyFTAdudg<IM)$$XJ+=M4Jx`i<@Zp2@
zT4<+TdJev9)+Wn7dBA)PL-U+-SM#88_<t|XY1M=3n_nS6;-BVE_7O+?Q+@5lvHB^R
zr{o7H|I6plI>`FRd@2voM;_x};$FU#2bINxdCEFdfBas5@-01}{*6<-v4eFCy~L%s
zqYpjk>HNdEAfCjH_W)<jo`3aEKAzGzWheFMW&HSY|NpUf=fQec<sHDYlLRCQ>l6fB
zAPPc}L@88<`-4FlNGY3W6)}K=L~#(2DHRQ%tte1sF(8a=(>jPnG}00ju=l#4vUI9I
zSsbfi*#t%tR9=y-<IMZHj$h}@``!y8qt8G2&N;txp7ZSIIp_C#Z^RY3^wR^r$%7B)
zTmEQ1y*NMn-3vdW4|za-BDZ<$#IM-FeEzSWT=qxw(C5$m9?BC?Kh$qNlux1k5E=*R
z4U!M?w-%}!<VAibKST4~FR?p5q4D-Fe#T$u-FrxJBL0l?UQ&IejXuATSJV;wU7Mew
zgT8w;aibozUs+GvI`WUYk9@}A4~m->qAQ;mr|y$a#3enmmv|D_;+~xJ%dYs;W`F*}
z&;0I`pIKL2+E3IEzSHqNj(Fwo{L#KDP6wgCL(Xr+rFp3*^}q4%m7R0c(cUNc9_6+v
zJI~zc6>~fH`0kzE&+;qt{J!#$?KkZ0e9??<yAM72k|*}~?cYcC_ptZ)z_%_q<R_bV
z{=ZJY(^QALclUSC(et|x_r~t&eeZ)m>$u-0kM{}w?y}$g_`6BkejjQbZS}jq2e$Q;
zmDlgNW@cwy>xl#7{Xga8rBDA4xc3Fd`G3#p)8DK0_dPwQC;ZVP`}(`O<iHPo`a8Px
zf{y3(p>DVCsq?qH?56!@ch)oC`GJ3sQ@^(N2;TSbANpWde0V=&ymk2lzwv#NxJQ?N
z_}+<r=*4sKA)nJPyIR-ZMe)6m@y4+mdhQ?a<J=706SYu&u^+fMlV{bT_H%jOc}L!o
zAN;?$@|^u!Ji8~-Z~ymPeIb8A`Ayv*FUgbY1>@vB-)rRWYxT>Y&KrB2eD9GrES@&z
z_tk!9>HqEYy$8Nw)=l!g{O+C|U+NtDJUN{w=~KV`johBwzt9;L)f4hLe(hiO8EtxY
zAI-k%4t&|KjptX^v45&3(8rJG`~zS1ZFLO#^ur(NgMO_~uj&)}Q+MN={^;AeRXwfW
zd}uv&s{LIZp&oL-g@5ajkKD$qALv)UgZv&J_)|xiXPkQ|b%M5awfQx@$Uo$gALvj1
zL`R%k2Os#;#<%(jpXiW7{mvioi5@?oPxR1HZ}Jc5p3;2w!XLY{m*>{OpZWOYZ~PLU
z_$D{|@*nb}V;uVE7;k;^<Y)Yn(>&)y<K;{1&@Vf)pZWY$J*-XN)-_K&Kyu+<8=_-<
zZTFYt)lWWhqeIW^Wj*$!PkONL;L~{HJf{cp(Oa*7zPCSFhaBu|K0eW>M}9>v{L^=~
zt?#*Y`Mq@@{Xy%K2fyryuIKvc4d3ELKmOSPrXI<QzB&WH>JRgs+wf%_>&w^Z@o#?3
zuK3jEPwdDK)LGV}S9YX-dNYn6Sl4>S;ZNIm^vOwX`y@V7Z|rD&{w9y(m%P@eAM^Q-
zdr0=im%4!5<YEW?<T4Lk>)9Xp4}P@i-+i#}4)lw6{si4CTgUq-e5ixf^U(c+dsX|L
zeM3Jz(KGwH2XKEte)fXKvmZ2mSR}7FR+otvc42>dGoD|m-?HA2Z}IJXhkt%VpOZ!B
zaOW52F82q{d(N-UdEN`+(>;T8gma+xroOZ9_4BlIJAJ#)M%Ouy9~kd^j$h|f=S$~4
zd>ik*r18-CTs`N!NY10b`sQ<f_V8mP=s5qm7ZB(C9RJ>n@^A8x-?`E`(K!-bay$PT
z2g!p^^UYINLG__|&b#<UhhF&szj9uXN5mnNr{p2~q<uvEi;EWGhhLEkf8vDx><{cf
z@8+qG(W8HIqGz1>;tSnGeg?%e#22(KWG5)Dp}HT6N5~JL`wZy+H-i3uI>`P|K8O4c
z(l^96G!HK7f42+iAJPNlFOdI1dV%shRDVGGU?M-X-`S`5qx>wt%ir?4`Rwd|K|g;N
zC+Z0II{b<r<wbUO4wpy813#5#=*hay+4e#A1L_+4r~TBv<35v~<UxMLf1TUWXJ_NY
zkNhalu)q9locl-jj=mSS&&ZqN(KzdQpJD%&fAFO~u)aK{?LNePkoVZ~7yF^3pPlH1
ze%UGWMEX0s(F1<WLr)#%x%*{#jhyW1IsWvs13u`BKJ35zR-Gb__`iPsOW)!JU3wyq
zd<DfX|1@8HOkVTxZ(k%o`Q#^hWiRse`t=5WAwQP~#J}~$gZcd0K1vSvW$wkqsd@Nf
zFZRuP*ShF>Pe%{(fw~Gk&*cT{ny+mfeynd@^u-rH@|<4PY4R?4j6+v{B0l63@`^Wp
zgdRJ}qx_N|8qbeC$7d_f2CbX&(T{oJ6Ovood6C|resN+vIpuTm7^nV{x5W=V(Vu<5
zda3_xlgIgmKCS2dLh^x6`eqMywNIFbPx~7`urGO!N$=`qa=E|9KfRfcK05NceHtC>
z%QsM5h<kdlPtb>WCl|l8K05LY|HCJ{u@gJfA2gqS&EqHJFrS?An11ukbFZMD!7o36
z_5=GGzUYVj$qm`ZdibF)^Z2JcgJ1e&Uwp{Z>N5Hf57t*#S)X34Pmk8s?_LET=98EF
z`mK*o>yp?0WF7uXo}_CZRFCoxacN!S=)rjUG|qkv?c>(vKkP*h?#tOpd}#BZ98YfJ
z@Qq*oW}flRb?lPQ=@I|*N1yDVU%cukzwfNn;rxkzus{8n$G_NL-pcizlb>y*Cw%fx
z`H4O74e1fz;zK_q4?B|^dXK`d=viKqU+5EG;=(+Bpr60;3wFZ~KO{dt=lA$vXL93*
zp5$x(F0QPH9y?o4JN*KEdNrQDw2i|z`N_+k8Gq)BBXZ#z(jU8uAM28z{_S7J*&peT
zKG>Dr=|Nsb7uwhGrA|VR{>h17{Hqi2iH`l+IQHXT{EfWOc=Pek-`T->^ld(U^S`ve
z`QqO^e36@<TZi1@%slmuah{`R9J}Gqc>Ix%o{VE>`Iz3U!_UxTKlcLc&flHK@ohc&
zz(>+|Pom!7cj$^|@hCpf<B#4?k;6E46kq(nbN$-Zkw3N3w_oVD|7iQWTmCMI_@{Sq
zk5Bh<`1Bl~^g*xe%@6q#d*RQz>`R~YWgY9{hh8DR*^NECuXZj(-#Gj`Pm4P!{>di~
zXj?b^!ub{*-*-WED*qEN&^&pHf8z(=p6eH<=F4yDIO{sc8AmVVV84tPbs|)c;KTaz
zAU%=GI@a~P*T3f=C;J%ZJZv5IMUP#`#b3}D7xY0+`?dJ@9>_h2{O*1NAMQ)!8_%uB
z9_l}SD!-<G<By)4$K`8rZa=^meY1yr>l{lj_@YmKU>)*$?%d=aMw{O7$$!zKUwnys
z<MA&q=$8lZ%fHzd|N6;^KE93TKlIKo@P$uw`JH$+pC6%%U-R%oesXBjulqxBg)i%&
zix22If42|W2i$MSllu8h=1X#T4($v0r(gSydGwv<1a#yB`qPiT{T_-t<LwXpom}=W
z^sJNeTE}?L(W6ItA!n{j@A~Z@iN^60ZGJ|7@|N#E@uMF<5FL4fKX@<9pEDl$m3o<;
z={HfH@I9;i#4hp}`5-;=KmOr+V0vdS^L&TF-{npFrSUs0SasjIr>)bmBR@bFJ$b}D
z`oce?2Z(R>RgcgQJ$TM8__V%wU@v+xFVQ@5@K@`5ZXP?*ul(VAVfhz-=<#cM^Lr0^
z@|_burDye^JZv2Q<G04+!{4b^Z)vj+JCF~b=!g^RTL=GsuSsv_rM^9<59>kug#5v7
z{1#p32W@<4(+j>6Q$F(eZpnP|qH7&;XtTe5^02e_dCtG;6Zt`Xp#JRj>n-((Iws@Y
zdrWk59s8R&wyrqDpMLS99#vl^dN1bt5A}(8_EU8Uc_4b;yFv1MpNO7$kY9)w`GOtA
zD?783_kHp({vo=?%Y%?y^eqnA(f*7M<R9{Z^~tBs)n*6!^F0Z__nywV$T)~UazXDm
z>6ab&nfS9lx!6}eMW5dqpZXxTddoQZN`1zz{1F}c@EnR~b`huOTE}zsAbz0xVdL;)
zJVcK^@CDhAKB4jUaqA31am=1sPq35vgx<)nj?cVq|Ag$~UI4%Byr}=34f)xfKjOnU
z{20%#&^2D$IDSc=?2ixU9QH@oeENp;$8Vu^{f<li#1~{Ae3%c7!-xA9@|&lf^XQBJ
zk=uB3n#V8k59Mw1o$J}(Im*64pZr9<f<C^~HS(8v#_=2T(YN2=N1Umb_z%D2xA;{r
zSeG5hp)DTRfxh^IcuT~;@%Ujc_QXH>@_=#ppfCAcKmDN3j_A`1dXW9`X*_$Y*Tk=V
zihb~*zA_Fy&#h;k{6&AB;|H=A)X#3{lh3;JXS{veeraE(H{<9N9na~VA6t*#&^I~E
zXHRmYgRb%PO)tjN54(D9evZ?I*2kayQ$9i$qATBuch9ZwewN+Yft}Pr^a=4pZu(#^
z@eTPSdFT~C<e@)w=}kYrt*`C5_;a3@&-t@FXZ@519qZ$}h2)0nck}tHe*Ovh9lfy|
ze#yx`=Got^<NQFs=8-><oy8wM=tI9aN%S1z1ONDi`l0Xd)uZGWkMw4~c&8V7Fps}N
z@>mbD6J&4oxpRbl$v)uw9lyU;SFnTqIMH{=z8_VG^LOVVzqjxC9YX*Ak(S@g$ESJD
zXYP%Cuj)QT+_NkB=!0LYf5@p{{f`g7OYr-A-_QCzf_&p#2FWFViBo=wzP32EU*lK2
z(}VME(p8trAME9Kw(PF{@;eUC>A^UDK!5Zl&ybH^oUiBue~_I1t^q&cPvRHizlGvc
zTx&!9FylnL7|&nyyVs!qMDZzJ>0Lkc-j-gSyTvuXK*#+_qI!{^qG!JK?Ju6=o4xcy
ze#777HFh)KI*a<}9MAD%eP}#HH{1H5d`@2W$EQ5SZy`UGXQ207@+beow|wzzp?yz2
zSI<NKmh~yW^W1odo;p^WpXqnrhx`g%c4i0dtPjM4e*8gk#ShS9f2hBO{8E0=hU|!*
zw*A5So<sS@y5a)TAN{(|q&Iqn*{^Lr{TuJ^6rl^f_cBf$gPwkfzPdwO9<eUFprc=&
z(YC%ew4SzcDTjXYcy3+K`Hwa}@DK5AeQnRhskZuuKSS{i<$K7_A-{&s%}}24+&bvV
zQ&8Q&-ynZf=joS6#gRHoenMB9KKVN&4?WpW#R-3h))#Naw-BG``yO1K<L?LA_x(QO
zAK!A*_uumJO*-<6=hiyvH}9*Qqny{=!#khx7j{X%(NBL#mwovYyNWaY{7D<~3wB^H
znB(+AcEg{xeURSh-9GHO@z6N@TZjF$&BvcQ03GYN*HV|dkJnEx`qhi-a&@e>{T$+h
z{^&tJKKQF~&Ku-{#`ClEcXY@rUg<fXtGDrEee>13+W4Ty)HnKx;+j6?iA2xQ)mCp<
zAAR;Ehv)Vic_;NwUMP=P7yUu=#0NyrJn>6T{0@ucH{O2k`y=<HF!vvAdBr|u|C5j8
zAN1{W{$8*5(eje_X~wI+ymzB_NDtQc{?_?}y^Ql-+Wv1Iv@Udhh4LKaKaicF^Pc?5
zp6J8Ox8&n@>@QC_|A{Mp0mYH^vTYpn+<N#oUp**Z&4coXcKSJbp7;9q7Uq!`@=xRO
zCr_#~;GT2t`NFv;uGj6o!^@7kdA%2m|2;_gu7%Fi{!XIx`6;{bL*ENb6y*_mWKa8n
z_Y~G+XUPBg-5@lth3XgU$s>c%Jnwbsm0bQFg*wLjHup8^9Qy`3p4)fOgZ2yli4VW4
zr(boT-x2$JkuALAO=n$s%wubhCHi{^_#&se4Z06X<lil9*=4f*>?qHu|L9A<e9!+p
zch4f;)pz<ce#Nc&gdN=%n9nZm-Jp8X@88wM{6*bFf8L*pfBC?A{K-9p`YHWcy+BTU
z;0t|m={cka`yahpN4}DmAbHpYJ@**wLT}__AM)UvpWws#<WtAWquR#nhx`YU6JPky
z4}Je>o_)}H)BXmnZ$5eCS9#4i`<(9`{aqVvaVbvSXREvHkNnC!-*=;LUC5uH{Q%0x
z<f31GV;`Y6dKM4v<=ERedbBR2FLs_NS|9SS7P4Op=`|5O`JsjC1IXS`T(ywjv`~H<
z7VRhEmmffUL2&{3A58o3M`#_0PyUeTIpnvHU16@zKOs9o^>GW?3HJK+o3^|we>&HC
zE*{v+dno4xafE+)OucSgi$8uQKKYG0pWOV}`2d<X2%Qt~1N~iK^nEABf5-=&cc6Zl
zavImdZyvhg?VtGdyw2bCP%qh6{hb>1m3qbZ0RDcp{)=CE;i6f0&hE~>_}-)5dG_qi
z->uR1_m}*AZhx=b-*>Y=@h|rl-ZRJ>>}H%g0{_lW^1M9G@6}=IPVYJ8S@R%z)`9jh
ze&bxre(E58B<`I9*voq-=V<YoekG5{_xy(6%M18{>Qe82{aq}7A0A%0@4Tx=4w=)j
zBUFE*#~;Kcv>!oyLGxN@{f!U(=TCoqZ0$~6E<gJF<m}}h)P9#pzxd|o@{T;izr62K
zH_@lOq;92us6OM5><`&p{YGE%mON)Ye7EA-b8#)77|$=%N9rrcAL&J#zv3tN18u*n
z<&XT?y$Qr0e%;e}4n5a~)*+Yq*3)l1)Q>;+N#<D(f2l8Y=oK0-UK5Rjp8K7HHpHj(
z*h_t%e9{Mdk&hkuC)CgG(C;g}NAg~YKifaaYd?hcGy9-%<R%yN{Fx%XrCi#O9>o{F
z%*Pk~4vXUXxkB=|7iD*GKrVVs`O>enA$sCn{5y}3)3_GmGm(FauUwa%=->XSU)(we
zqsNcM1G?<$x%)AE>xcXU-{>2!EgsF2@8xrGgARGk#}|I{IXTIVj&&ivJl8fa(Yo|=
z={bk(z4wY4W9q0x^)<BrI_J1IvCrA(-N(4+rFZMwKcW2}UFQOI6~2r^w}t2=qB{u5
zGg(af_#T#?IHNE7g}7D+Iv<F0`B=Z-hvs;B+W1pWTYkci=RY+n?|Yx(?^k=z;{P@D
z|Bv{8*o@bPJG|rAd(Qa$)G_1n<^MUczP9{jzk%eF7a;%U_u^4}7>BO7qDS-T**JR8
zHok@G2=Q!P@s6*wtGE^a=4ogCL(e?)^h5dE`N#h=<bA7qQ1{~6&fVyG4@STAgueP1
zpZIcalTX!&@(wxEuIdkV;kW$6?|sDsdi;c6vj;uWzyC*vUG1CrfcoVv@<aPCeM95%
z&)?9aKj%8*$b~=k6ePbm5!aqWdZ9OU9i$&}nrD5_>65=}(<A=Lhd=R6UgPlNzSFwN
z7drT`K6?1zKm3#5c#bZ+Kz_lF_~9qcrR+;Sb}~;J=DDB0droia0`J4M$s?Zd3yp*5
z(wDd<hv(*tXL0LX>D-B~cs1YtYCip-Lm$@l96C??yV~~gMg8wr-P_<t-mqWNBl+YP
z{6hCa_;v2Lo_id18u|DaJ?ZCf{LB89D8I_1?uY!oKpo`Xj2!M;tjGT3B^Upb=h9!b
z#U1@XamMfY3wy~&{(h6ZXgz-GIX}=w$9~zu*7s)F&u%T`Z-dZ#<6)7$>|^|we`;IT
zeH=e_pQG)$dGy5oIZm92EAMI0b^qkvUwnw07RpOaJpZ;&ZT86O5&r8w&Hb14VcJD}
z4XSIi4sW&PAN!~}PFz6a=mmPtj?S(4BZv4xN8Vv?{K+fqC?3SEyrEzHr_Jv6CHY3Y
zYs)v{&OGr5(f8as{EdA4i~RB)6!+xd*Va!o4&U-l@<Sfa(fvcCxbZs#zoY5(-!Jh~
zaU<UNh3DGh&p6NdKRQsq^|bjBy6gn`oA}i>-+q>K(S!1W_Zr?$c#q=!hVNXA7x(rt
z?*+`)kDo+-hEMeby@)G$4_)tvmTrCEjyKO*zq1bh-S0#1Y3$>J(EEM&6%fCW{vbKf
z#}7L`S14brAAD!PZoVVr=gx2Vqi6TN;vR|@`NBPnJZGOb-o2K*A)eLuP#%*n<Z*e~
zx_*zSjXvZ@>fv?1cgyWtEnmAc&%F>iAwOYX^XY+{kloe$+T_O%{}nIlmhpP;a|ax7
z)q`tB$*Z2Q&uF8ouCq_km;9<tukyV*PMiMR)8dz1z5ge#@$3ripX`rcc0&*U=-6M3
z!-wb2IokAW95jx;<Xgy|_G|v(_mlqL7;(W5$S*#`w{_8R@8LevcQu|{pC93aUgT~5
zqfPI=3$wm`g&#Y&v!Cz1=!M?c$G)pBfW7{H$3OXN$|X;@*VXo(-|wXDOZ4F0o&4-Y
z9{I`smiYF6`<K)I>EOTUyg&ARrMlC8WgqfBmpt|r`v<gtkYBv&m&fIG?|tnH@`ruW
zdi;ofyl3{De)XfTAJQZL(2owpFTb$9eLy@w@<V!s=0kEr`!ggzBp2k@kUvA`d`J#x
zT%!FBs=LJ{y`n>Z{FPqGA%5vezj=_q7?<^pw)xO{&^U-Lls6&&hUy&1Um*VIXR;`N
zK>0?!E3Q3fCwUG(_6OtL?~zYlKu6m?V4VGDQU5;3bNR-4)}e3w;tPFzp@)t(I@;(P
zhhNXNa~wL*{s)~4@ngSsUbDYL`N}+P^sIv()Q>-UP@h@P`p#$0ckDoqiS)oO=-^X7
zKIlu^zJy=$h(Gl3i68wr&$`ffa@&Xa2RitH^bYN_^kbi~zPcBm^yj(wBp<!l|Fp#u
ze$kUp(UV7v>&17H_A~RzMgRCkhd%Jb-|3H?*`Iv&C3JJ1d`Lce<B#MrKaqU&BEOSI
zKfCIuXMSLP@<RGF9{>0vH#yX^@(jB4=|0LlZF!Xb=+%1sO5G3D6_9?cM}Ge0xpM-2
zIEPqY{Hu$}k#y*XU(f@(+U&~jtV2Hf;*Zu5U-XR+{>ZM@hx*Y`zv!obd}v#T{pl6T
ztN0#-_F44QLF&ci2fyM*J!iar`#iezjK5w#-syw>$p`T_EPC%iAN<lixqE&7Bfi)d
zU;HMKeU0b8>QVBLm;TIC=faj;?y=N`{F0ySx$@2pUUcoO&b<kAKdA1KZ_-ZqbG{T`
z__u$tpZeRn_(B(7&RhJ*yxIGne8T${&K&=~mA>%l|6Aq<=&~F7`1Ae=fA}$<UCc|&
zd?~K8Up|uWp|}+P@(DZBFF)gl{KmeGp0>P!zI{sDI>vj?-9r1%ut?rS`eZMN4|Yiu
z$Knc|9GB~9%Qq86{*%bQiS#@O2kFtimPk*D-d_$v{t4~B{7)XkpMAT9{1Oh*FTZP{
z{i%hSM~7|md;TIm#2I8y<M^3)qc8K(l^@-c*f;nseM0$WSi}#sKSKM%Ae5)wTUeie
z<6m3;v5xWfW$|d=mhbH|_5<tMkNjOa_h;6F=0oes%lLun0f-K?Pe6P@a>LB?^aagJ
zbYJS;(LKNK3EW@$F3I~&aqfE_ah%V6zr#<_buP0H*eCGgUQ<34U&itOtn*rR#WU^a
zH{yh!LG-oxz2DV{EB5Cf;uVTpZTU%?KG64lh&%|BPkx{u9r3sR&U^mgW7{q0<Oh5t
z(m#8MEBVxOdU)hxTby~tVe`9uzd?R^)pK;oA)e$fcHlSK_#~(OhMsc&z^^<gPtuP(
z{f6_-KjzinSg-S(e9md&+usN9K8s%InZLS6SGS2H_iWnuv#vZ$4sz+INBN9?(ea#I
z;zk>v>1X(47xw1Q{L_96@deoh9sI}(?3s4dPw)I-SQMAf7Rt}+K!2xH9%o<ghv-**
z<vk+1Tztx<H=S_)1|7Yyi~Wgz(Tn`;`!#m+Udj6~{wF@P{T_~;<tKJF-uoc;{p`K`
zjVJB${f|r^;|EZj!r$hTn>Rl2hz+~<zjw~^*S&b%ZsCpR?|<s%n~x)T_&I&ihkV2j
z_%(j$U7K9;toH@-k@?1nBl@xLK>5J=&if1R{gR%3?=R?+UGOVTt%Ltz{ggiV3w_Bm
z<b&cAzxd);u(iJV#(Dp0UH->j-lvKK{%#&R-s6*BT%*e_@<rA;$*=XHJZ(RqFY6cw
z<tuti6mR%zA$oAoIQ&>2iVN?Fq380aar8t^$nKs)egxSCiVw(6&^U+=w2u2%eB&4Y
z?4#avkIJ8oM~@w}<tzGjZ%eO`K8)v&zLP+Y-{Lp>(X*bq6@7f0Z#+GaJLmHg<E;yw
zYtcuKJm`@JKh{%ML3%>Zc<*P$wS7o_HV$9vd3xZd=;Bk`xYRR#^z#G!!t@97TZbO3
z<9nRslbq-oug<~`dytF2vWNN=zv$3c()WH8(&Hc`7gR5b2l~~fM}B2J{(vv~GoBvF
zZ9UJ?=V$u0=@mcFK4Lt&kbUTt|Dj7i@{4}{NDhduIOGS}W@r19dFX4aZ}CZgkp0oe
zC;spc&7)7_^rK^+^IeVmVdwu7PdfJg?bn^%`99J<>3-SY@ld}y_oy?~t=jIf(TD0g
z<Dfdw-`Rlfzw^2KapOF9&!Vm-5B_`?>3hNC$NL4()q#*6Tc{p&USW6kvG1u%?0fuL
zT=n|zdDIW~U;XH)yYUB|FQ9uCeup3aD!%c<f5-#XW&F>+4%Ny0!+RBa=0}Nshv+*m
z{$gE!cZ6LYxah5mFF0sMM{nwEdZh>F4|WjO>N-fj5M9VmA$<)(_kaA9T=ZjKq&M*n
z$%7Aa;e%fJn{$D>m^?{`|KQ&^^YQ8Zn0!pn=*zFpN#cw=5Pf;v`s$}V-;j^o#zTCV
zXI*+FhkkmaZ*=ho_0zxg#05V`r`L~DZSjgf{Afde&&Yd0{*m{_@|<<dhvHveqYwIa
z-we&emvQ8_4u3(9y`4At19{oeeot@e1>@u!br`*(Ydw0l-#YiAo9pmD{!9*SbriXv
zJcciMk)B$4{`uN(zgK4tLi^1h8|i-#+LxYh$Zwu4<X`N`-`P_?eTkFY-@GqzZz%ub
z(|q*AiMSQl_I-H~KjKvWV;BBpJo=vN7oYmo?f8%<JQuIz@myVP9rNh}zxbj@=QRD!
zE$BkypgP?-F46li<LMdwTu<FD@5sya%5L<9FZ(a~<U#UV&%L5NFaOi4JnWt_k$n8d
zet>WLr}goPKK|?j<YX^(j6820ZS?sYev__o^hZu|+dr&leg0z{{;Y$p=k{IaA@zoO
zSKGP7JarAe)FsYK>RsnA>v%6pFXTvczpNgxPxAx(p<};Q=ZP=t!;F7=VNZ1nd#k&k
zIJEDQOI+|LZFXS?`GUTz&mW-s4EAavI>RD46U7<-q1P6YFOlBRYoWM;;u6XeiQdc0
zNAjO_=vzEFFIk6vJvR@M7ajfN!Vf#PP`rE3<huZWhsb*{^znx;e2~|8@BiEnJNL=s
z#@YAW=lGo+`tF^?6TRz4Uw&q1bjeLm*2g!0q<{N_?=zhjoGZn@eCHe}?(84db?)|e
zRplpo_q!@>c>x{y-ghP9OCI!HvAlsk{^SXHPCVgT-jpZN#kaVZPyC+Ty@q>J{@UxG
z&*%fc_#vl!Y@F{TolEd3f1zVM|6p%^DStUn^K1R&^ZP{f{0>W=#2=Jz_y_;vH~4}2
z-AC%@7m4-(@<8i}7k+Gi;~(tDZ^+5N$*G?`=mkA?q)*6h{NFm-=CKpG`Ga-Tx9T1D
zfA&T4;TvE2@tyH4-`Llz!+-U&C%)~o@*BR$Z63MM%{q`h*-?DRGyIufc;77E<wyI4
z_4$E)R-RLz_+5|soE-d>UcB!x&i-J(z$ZP+v-}TT^zh;Qz(4r~Ims=K(Kp}v&~xkI
zlYimUd}tgbKQs=q{~$y!G50%mLqFFg5Bd0qe)}8w=+AS0D38c%{Epw!bNabB<e%co
zd}v+qn(N@hz5&tK7T5BUe&+yfdBS@G{$d~S93SdydXNu{v#xw%e<Tk-%IE5I^W`6U
z<k#duR~+ZQ;qUf{Z}vgYb9@<Ref(ILe(m$}F+I9(G!8%RGd(B2{T3hc8@;MG>@VVz
zpDybEPBinUIN>M!iC=1yll@w9;>&#eo5vph9}wf^Y5I^y#UDA;<@^qP{>Sg^H}nI|
zqbGbp`New{D9-88czRc#X?u^T4)vZ%8$a?je#CX6@#K;h^h0#)qx5GyeX*bY2>Sbf
z{$8T`-gEW1--S4T_)gY$md5+8dGqV1UcA+N7j*2)5B&ez^rhYrXY!AF$hloysvFIl
zEXp5{A0*1dlST2?!dx%g;w4dD%lyDk@k^ijA^q}WZTgVEjAuu7;7{VqeVV#Uzr27h
zz97Ga{FOi0U&RN0=t&*Kzwim=QGB2)FUxo2M_(MJ-UqeuKM3s;6Ge5XdQ&{A@6?gL
zXU#fCd>h9vogd|4bt812E-%v${jrnz@&Q!WnrB`7ndiBBja<e-`Pn#i9lHAY2fg95
zg;_tN3(;$#bz4XulSO>Aken@~zZTk`TF6d=utk5O{ru8B22}5$hadZ*Ha#us*F*fq
zzR6$YBl%5S$`kHq2BEy9ew8oOTk^GhB#(OTeS!K0-|pY!NqHST>&nCSP5DlJD=(QZ
zPnwrVpZq~x!7lib7x*<h(XaXLW%SEO>?lvjH|igCi1F6*T;1WEA&*#xzd-u89)CfP
zAMg|U$CtbZ?RWAeI`nB@;eYffzv-8M@QHuVQ!aWn4jM-;^YO{Q{ho{*?CE|{zx_rV
zUGb&gI`Sibvo3!c7Rf^&kp9@my5dFLX|uO@&{kJi2cP=M0rith{V$Hy$?`w@;6uN7
zwXXdLUFbXw#p|$WUHMI2hi~<?JjYM**Rqpw=#t+&dWPsj=W})<AOB+?blv~Ti`Jzl
zcJbZj)Ll2-;FO)$?ruE#q+Q>C|D193dop}Ehg(k`CbxAUd+`(BRXR7aH-9uf<Iy<%
z?62QEnDJ$v@!|p<^56sV2gt9*bE3F!VUBOLtq1u@qI(B^3O7HxH}mq#XLju7+{cgA
z|Lh6*k9GK){Gm>w2mSO&9(M3N^{?N(q4DgaZJhnnKE}S<(E9uwU+j$z`K(7C`=)vL
z#;5(tbMxpg_08Yie>vBQ2j|jWyk}BxxJNVIeBZIQ>MP#^_%4v$eYfqN1&T*<xMwqu
zU+^#bgZxX}(_5nRvwY~D)O`7jyzIz+@(ko3{K0&FfIs6~NU!wH57-x~XVgLZ?d#5o
z_BAMOpznJi`S5MugYNazef-aJe0Z+kJbXjz8Hay#pnmhntzR4B%X9L0E{=>7=i*Y{
zWlwyHJAA2YQ(k#jT+0jM3w`zWu*i=Rt-}xHYyIRS4>S&{!?n$W>Nexl*Yr%E>}j6o
z>;R2({!mv#_s{Ox?HBTyeMcVJcXam?=YDQVCx7U7jxrt{_e}DM`r5kkkLU7=dV?L<
z)A@ni&JoUs_ID^x<4b+z93vk%7aE68bs%}wY4R>PjZ1W{V^{AT*p=V%2YIlC^g(|0
zoqQ+Xy0@b@dXR7J_lfj|4ov&7ll{Uv+EBiS<g$)7IjjTm%U<~9AKJsBeZ%_Rldz9_
z+O(Uv7uWoO9mJWq!#8`8(>|H~?2kTq_!;}6Lk@N@U%zqA51#Ws@0lQe=>>ni_}v11
zSr?xWpY$l-lN;u9`5)if=<yHy@yl%MAB60bXdQVA`d-UEFQ2+!vETc?LEAX{m%MBL
zGR}UN?~Tki4;|w@#~1zbNAo<l9y;z-a-8S(73<&|eb2QM(baFfeN*0GZ+u$cdiFi@
z(PKaJQ+{o6s%;*++Tu^V<5S!E`mGDe!!PYi;!?l3(iTtRP~7@XS3XY^ALhCDL|<Hr
z2mUR8$+PxBd|HP-%#+9QEwA`~Pd_^L9s1W#UgM#2PVS>g-@5MO>^IIq_HlCH1OMcO
z{7hW2CqKi#eS)3&H~p{&f8k&J23kkF^Lx+rqsQOGuXq%{)-zxH=!g2zBd2-dz<BbL
z!*k=!rw{rekNSpw(4|jpazJ`T-+OiZTc6$NL)$ofXWV-a$RBbZzsCo?XB@GQeU)E|
zN6-18du4H8pXJBm)%jU`syp5H=~o{>^T?$?+xU?uoQJIgJvUx_T1Q-pKlQNn=u6vt
z_Yl4Z(T@*w#Itea_8fh2j9+bfW)Jp}573bh^+Woj4|X!o`Neqs)_0CV$2fMTH~Mt$
zuntrgk;A^g@A$VkQy+WZhA-<o@41)J|HWNDv(JI=o7b_&>y{t)(GULcS0m6m;zfRt
z&%~X6<1=2}hqTZ<D6hyf;*<WwxqRXt!#PWx>mE-2L(jZi&pLh=Zyq_#Pn6#vyFq^0
zLUzQLJW9{x;qUYdjZ3}Z2Yo0{L2^R#V9HA_D6c_!h5X0w?)gg#@d?>G(LI=c>(L{7
zTHkuci!*5dp;z?v<HvLRBfVJ<vLpTB%lO=Htiz7{&N`6Z@k2g!7CQDx{KDjiAK=e?
z>)_LU3_8&I<iaPq>OFcvAJP;1u@nD;_5=FCujkfhU-a_1{?voEeZxBR>N$FzL;jyA
z?u|1aTF*G`MDr8T!5>5qKiceoK6>UQJ#BiUU-s7?7V%-9vJbMaesxHq{aL+`^OKIY
zeGl6AJV%GW;xorvzlFvp-{evscyFUW(Q|yMTiDY&+UlcTymuO}Pu_FX4R^h7YNyVi
zKYvfhd*dtjop<%fA#=J@=Wln}P5aI6<n8e~u3snUM{oSS(`J`%d;W(XAMv{`b&>Z@
z^v<v3Irl{TlD~;_b`g*AAOAH^oBhZS*&984v8Or@sz<f`|IF+wkL5lr527m{+E?)B
zp3S`zdg?j%=jul1U3I1Rkm^8nwR%Y&r#I`{AMoiM=lmn@LwO#OSKgHme2<B~eD3_f
zuJQ#s_(YF?us@_XZF;r7Jk~<*^IAwBEyUk5Me-)HKl%AH|Mh-L9-#;Nq+jFo(~tLd
z=o^oYd?LOwuIN=gz^|b7T4-JV#=q#3|3T~13%}xL#-YnU)OGkE5BZE2$MS)=5l8Z#
z{A|2&{91gn3pCI7{?NUbaq^<P3dx%&Z{f!}^vRx(z7y%iJruvEH}Xvs=~;fmhkp6W
zbABjas8hr@l)v!Fj{GE1e96Nt^nNkXzH8sGuZc(Z9WcjR&wDM;A-`ik{M*lsv)_Av
z$-l{`O;6&0pVGT=__4nD;}41Uo9x$?r_f;+cC~NtPvf;c7tfGCh<~VV;0N-W{Y1Wl
z)>(YUvX9^Qs#)Fs8(p~l5ywp%Lyv!YPA>bu{FZ1Oe<d%z(8ouP!xz1oM<3*ocdch0
zv<|;Cj~+cohyKL5arh($JE8}T*ESDb&-EKe&)WDmp1oV>cdh(e{_|a=e3Ev`w*8CV
z@kKw*rM_>0-g82J>^lyqp2erQcAp~->DT(6i)-{f@Ad0@$e+;RAI6j0bNN9Vf7<5L
z3%;#`AM=d!JW>3hD<8^l_Bs0?lyAvt|B+{nlQ*3|<VoYv(YCLX7ytZ$o!z&Rmt68O
zKhaM<ZRq}%9Q3AcCx`Qe{B!Nn<G0*q@9E>`wKKOkcf+f8nbtXfs3TsoX7{B_XROmX
z$N9UM{%!~Tdq0T}{$(GKALIk^55<{%#C-lmFZ`OE`qex38~yT?@%9;ZhSr7bf?xSV
z{?QNR8z{eM%Tq0+ANo*Vqr;!X7yjtOb8_+*<Hech`svxap6lmd@*ZR_`u6{3L+5sW
z=X<ZbcVM?fexq*ne!<@@5oi33zdOIW*HDk@m$&Gf9-z4M+<QcEY~Qr6&>KJS9P*d+
zANH^xTi^a;{mmZQ<Ib<Y`|;6ShhNc?=kfx6`H6M-2~?Nk*L;3pzP5NwdH4hW<`@27
z4f?^i_4zOU`MdG-fUb4%g&%e`-#GKhg-`R*H4h)=$rtQJul&S#a`6j(qhFic*41Vo
z^w~8Lee&@y{Od=L9@!V_SJ&E~*jfC`ht5;>E#ufvT|yq1bq;-b&JOJ3dpYsIzUmhI
zu!np^PvnBmC+Y!cUtnMRfblT(fKT%vzvB1!L`NMVpOO>*-Y5EAN4;SmlfTp-_FMgs
zzUY;n*o!>qum}C&*L?Q?^d>GKJ(_1ceW=H@6X^+G<ZU57tjlibiXZ!*@%ZcY>sxlB
zFLtCS&&|i5Hoxbm@_{^rk3?}ve~`bjhxpJo&U#6gz0AWOz9Ih-ui}|p#_>z{#PrQB
z{7pYQCVFl^=8yPfCvv8I?8<IwC-m8i9@rt{m4C7``t+r3oVY~?zmUGj&Cl>f9)4y#
zKFt%4;?Q`{(c|yxC3-cVT<mE+{nEdF>yTeR`RH4lo*_BVg~pTDc;EN?of3VK+xo_9
z<5L@7<n&zIe26dn7;k-f3Vr@bFVJ)An@|7H_mlQ5b}^ryi6?xw5PyGY<X?&G(?WjU
zLh`qe-2Q&FddBZ~)Di56jyg&n!yo<HC;6Lv*6Z&B)|K!0E&mYr?tl0be{c_DJbTNp
z?s4o7_>`aIef5C+oa>2y^W=YR{))c^tM)qY>bs|msrUT;Ngb+=)UWNi_W}ByOVpdr
z4d&~2f6i{|Uvzxej=t}!%yVAzcQ(;g*ZO-(^oB0})U)`K_xO|V#MpsfId6N8zJ1s4
z`|!o@%tK#3BsYKc9O9S#yf4HzI~eEtGwaenKE)mW-7}d-59aen^$0saa-nBFd0J>b
zzUhJf=p8+BIDahu$uTcm`l|(<bFaUf<$G6p@g1FUzU%cn3jW1!$RWR6blFd5zVO_q
zMy-!O<Jp6L#1XynOaAD6;SCRccZ1U}d3tp5+Fv>3iJw0{Y8-vAzjf$`-K}SR@*0Ov
z{6O=qXB~9-3Hd$e5A^3bxy-k|c*DPV5ZCl?9y|E|1>`sR3t!sq6Xj=lRoi|m&)bLX
zqxN^>wB<$nuX*aH^k@5rb*zsbI{335y5<{izp`)Wx8FebTH+bs=o+UEz@PmAAI8y(
z=jK_*{jhQVe^=w7{LEhBTs+(F@a>!{A42^4ZrJZW-QVFa(L88hw+{W-f9aQfwCPhn
ze@uI-6X->qfQ~xW`&{GI5$LH8wCM|<^u|Bv$$ISQIsWzYC+{IVH=iG>OXYoX7-yb!
z(8ag$)>RjY8~x~O+uy8Xymiop`tf6ZXdJrGbMyHjq-T8K-#qf#m+-G&TOL4H-cl#v
zo18HD(vJ>)$U{zbopJQ*xjc-|q-)+pkssp+f9Ug5DE^Eyuh*}$$w6P}wD=PL>}Fp?
zU)^SXc9Wm1r~bx|esvL~SL>-q@M&Go@$ET2pniOsFYnM7JMu&Mg1_JkKjc#f=qIN<
zO<sI>uFap>*?O6moJ-J?_wXeT(`&D9Z+x-0{K!v@mw&BmJbThN{gK0S_fWZ>wsqK9
z{<UvtL-rvz``A~}5%0N<_3)_<Ql~kesGG>4KJ;9@B%iPk|Du0!!Jmwi=gen6^Peg5
zSNjXUBERQ@5P#<L4|Obi@C$ABAP0Rw^{V=r{OF4#{p<<(qkYqQpB<TV!RqU#b@;}Q
zar90v>Pz>C?p5s9zU%WHvwF-p{P9=*XdUlm#Ibq}zx+a6nwR_==ef4^_!WDa58YQ7
zPcNR!m(CmH($6mZK>V>MJIh<vF%Oz=oPK=k=TGSK2lj{AFAt#u@rNG%=#8A>Rh*cI
zAM45s`tbqjQQLT^AAS19pLO(;%RJAG$FF|lpm;Mr(R}iFt{>m{fyP7QJ$DW<-*avD
z=a2jwy%stj%R^8er+?>R=V$i`&^?8Dc~5Y}j`Oa%_^D}Q|Gse31rOdnW9;)Aue{{x
z2j~7CqT_u2gHwO8+eWuPHtPPvdEI&2J(}^*ICY~ov@hAW$VXoHLfX)MpSr|-06O$(
zytqK$zNhUz13!Mhz<%-rJM&+3_=WtBuJz5cpR3a#|CC?NBWDYJALaiSmX8zV2lbSC
z*WVR#KDTe!N9+rJN9$hOeD#oXtMT$5{dwOE?GNfPc|$!(58}xFAis+v`GCFELE=Y0
zbf3n**_+>_AG=>-PkzAO{Kk5T<dUz%fqY_pdBZqy%m2kQKeDd<M_iIq-sh+GYx<!_
z`ZnG^Zhy`1v(>HW(m#8M2mWgxVsC!qUWNau*CD@2KFsH*#*;_g0{K1UZ^rX8=OF9a
zXT=deG>`n`A`iLwt96qOd!tM5`q2|_>OJeDXC68D8^4hk<zx90pOD|Vx0Y|@GyaMm
z`RR#YyPwie4#;2lsrC6cI{aB4#{XJ3-nQQ<_x^emU+%qM`{=fpZnt{sn0tHi?eArX
zOZ3qtuYU2E{ni(E@&P%-kGLfddg4}|VF%|q`o|~z<KH~`gY=AlbkPy7;!xgVPkhrS
zKJbsO{fWGuqvQJpa?1bsgU*@!Ry;euh-beO^!=sh@<q1otKtXBU+#bHKjKc@<A-14
zU;Lp*59$eVC~t^&ZF!3S(~mgCxA-+*zQ(V%_>y1f3qAXRJdu3*Zk$}sgTC+ZJ6d)z
z4*&E{F6exj`#*Wq2ax{p$v-BF&IkNOobpTlub#mV`}$s39>))TvbXyZ`O19pqTjmw
zNt>VX3vp*2WG{Y#zWvobf%l2(8Fsf%XhU+y3)=L-pU6die$QUUYtxtc+Wdtd@iXya
zJUv15*~2(=p>dF3damvJaBcFjgLULb$PVmBUyy#RhfnjMe8NxVZFJDLud+LT@Z335
z8y(|4N7r-lgkO9ahaUgJr+#$heaL^&&2jt|f9B~oUfVe6d7}GS{W+doiODBD*vG|*
zerTWTtvqbshkrS@qbK?$m-Y1XH+-UR9J^fn$_p3Gx^s3%@8)Y;k6$Joa*0#=G~ask
zgbw@S9~zhQ@F9LY=imGR%ERVC<M9VQ=lA$f&)H|Ji+}UTO@G#*7j5!G^$5MPi*pjp
zJWF5B@6J8^#yJB$=Mn3ehrV+qK0VjAK6%VjUs}ihAWq~*^R1J7+9%~Peg&Nq(KBED
zf-bv|)4qo<bqsy5gZ1qX^rt@dTs)DFpWq99`;PWsUvx|FyDy&7nde+dUU?i}_EGf~
zd9qG3-hSt~y39U_KjZO@zWqoWvNt=>gE*i+_OU*_dtb(1^czPH>VN)bJ$7afdSaKf
zx4ehGc;uhPv!n9}eHd??`zqtuf#2baKGYH9!iV+c3-a~iyG-|BlSTPtSj3OK=U&u#
z&;BIN<yU!=9q5l-><;OLpBX2g^DBB{cjK(bKK#x&^E_t{{$w3?=SSv=4@fW8ht7ZO
zh7P|IFZK)c=$HSX$3MiS^9em!M?P_0AdmIrEp2Fh`XVoW=*x4cpS|hLy7CkK(u02M
ziv#+>CwcLM4n&9lh)4Y4-+hPs6n^WuxOEPfkL6+K5&MUI&pP&X@$SA^+>?*|@~v_F
zQ=HJJdlK^7ALUVT$UoT8y&C_fXZ&c(v(_gky3jdJy@W6KSMIyK@6qpf=EttS=B1y!
zVXaOb>3x9j-1xD0aG!ur`qD2h(5GLh&S59{8TziyehBde`5%9yPkxWSyav@>{L6hD
zKT)sCm(K0<Z=Cm5=JQ|oZQiG`3wvpMFTnq`JvZNTD4&x9l3V-zAAI?eFCMvI4Dwqj
zUdRLO<Kh)v`l1JR6j$PnUgR6=sNdutD6Ykccyezj{>8g`OFlARKR*;V;#ho}=RQH6
zGT;3Ze~0!z`#Qh0&#;R&J=+f;zN{mkLh(eO_8al0-@bwm`wRZWk$Wv|f7f4o&!@h5
z*I&N+=@H*?h<9!GUhe1YpY}0-m)X6b@1flX(zD;i`<<w`rGM+XuO$b4(7*goAN)+4
zeAbm$e6Ql3+264;P97xJAVkOiUyUC5tq0kcAF7AsEoh&B_-AkZ{Q47T9zWy6OV{oA
zDZSDQKIIR3h5Vksia&mQ^N9~#_V!z+cl2amU~hc0pY<X8T8EvCmv7|*`Gw#6-r!w_
zOh4q&lh^9RFaGc?U*Sib7;hcpAUW<nW6gK2IBIUU^_#DL``7<tLFYX#z0!-gC5L<`
z{{MTgJ6`pfsW0rvOAh*HfA@;^vz>R?=alU>S+86Fu3dJU|E`%Ge%X&+#F78cQ2Y&h
zUTBSDk3@3XKNIP{h3u5*JPGLu+TUUFm2G;3;%%bnI|%ujzMZF``bIu9PM(yH`47JM
zh4_W`Kl#mgZFRJ~?7fP*ik#k4ndiB95FhGQd5GRU7r*j9diF2*1)uh3`@Md37<-{>
zJ^8_Xhdf|kx4ya(fA#})vUThiz5clz(hvL5r+OTk5An^8`tj}m6Gh*1`BMI2cbI<7
zKE}&85Pe84`ZC{h=W+IMej$%IW>5PH{?S2S9%$u{R(~rmWWRmX{-UnI7du1t=11Ol
zt7G}Kb=0r=wfU2A=0o+g{A)h_8RxnBUHxhu<K%Vy_6==u1?5e4;5YnO{mmcPN&IP(
zQ=44sdwxcr*29PR<7ed8Z(oG;DX!!z{>Xmv0sTSq>5Kl@nH})wx%032&JX(W!#?Do
zfBHaQo;1#U`l5erD4)xV#_?-<7N`7LJd)de0zbePy4GQDc~pM19{saF|AFG&c>9=q
zWq<SB-wV*@f9!9*b>&a}&_1ai761CtHy_%c%|lNc-}0S2!msu7XZakG!#N=HIXj~x
zZ?da?a@fbT@ekFZ<nZ_G$e~WKj`~$xiwl0kkNFqBh2mEHp`&d+zthi8$!VVP;@Ul)
zIFoPC!KZr;=XP-`o{dvyqi20}G{43#zO>~D@k4%mqi217XS}%NFW%>p3t9*N=39^5
zjFZp#rTj-u{w)5`L6=|Si`@9No_XTD7r%3(Kl9B)UmJ=O>zGfU@@ejG`sFL*$zy%;
zh(qhr2Ycz~AMA%Oa$67I(D}o7e3&QSh+lQExOA^2AKNd)3A>6zayehRH{_@86FnDC
z=FuPjgzSk={q#f6kbLZBeb4zJd!S=Hq;Ko-Yy8qHduh|(Ak4ZF-|~nyJ6KnK#%Cga
z^^0F|M=pAy2l-Qc(-S{uH}|OM$z$y2{*=Fw$GYenXTE!6>#1+}7yj@|5BTz3Y_{D$
z<J&#9de;4g?~Hs`0m<Qa9{BYgtbXqwuDo&S#}C}%=}~m)N!$C1$IrfV!H3>5Wo*{M
zv;JqZ3#W{U4|E_t+}G0^`-n^XBss}L|NIx<=;M!lpmEk=H}#@@kiPLrKjz6F{E&Uk
z=ZEOixB5^WC?5Ek{l$EK0j;k-h2jWZex%<x@q>SLHa*KH^uZ4NR(_*T<Iy1>dCWJ@
zc=Y+dw!CA!b)1Lk#X68((WiHIVJC8m7yZzCB>rPQL?6HW4<G2`o82LK>4jd&mvq>j
zpUB6?(HHv0LH1UM@(cPg55M*e&+XIn!>&o69=tE0Cv?eSJ^IzpuJ}U_@_T5!b<ovz
zZsqst2K>k$;!V9Fe%O~E;uF$4{nDHD=?7Y0-e(W>2D>K8tMtGQ{FEQdkL(5QAM&)k
z?spXSCwUWneyLyHm8a!H{!E|zSew4+huy@9d@0}3cOrh!QRm8I{Dyt`DLt6SFUUzA
ze6x>n>J)ZnA8md>PIBR&A3@{fWpeAMH~GeT{0bfGl8gPtfqFUp%sh6mZqmhvJm5L>
z9fQ9|wR+}@|Ms4Zri|_Ij$`jR<MUI;E<NY4z4u-*W9(!1?KJJU>!**s^^!OKeEK!h
z#*Ewa_0tc2(^gZ*{QWKdh@Loy?8T4M{^TbglxM{c{^bRJ!jI)u`x)e?_8G|D?8E=G
zA^P~yk1uv(7k2gBc=<#>JGJ^fXJ>j5FZ@tHe?wnCI^LH$m!e}H{iYnAr=0l0Kfkpe
z{@6+0v95mO@hhJj2hmI9Px!=_Ho4fBeB`k%IgE3!Bwsj>$~VwBcBcOp;t$#nApRge
zAwD7bA^qifuG*8+0DZb=(B`-DG<%AF*zya0g`W6vpX_@CZHTToKwqBA^QHVM{+!#?
zan5<rzTr6(7xp*wtj}NNduV-i06&&b<!y0_E<W^I2Oad?YuXp=JMz5x$NqyK`#iqQ
z)3%N_`5}I+t4?Ge^#?y8hdjcc=+`{^mi<(oguVWKqWXy(zCWZ7@}wU538W`_A(#2;
z9CYbjKl+fKjd!m>PX1#(>p<hsS4W$tE>w5On<)=D_!EBcr~Y<NU>$Y0ws^6w`Qk+#
zP9AjeYhC^3^9yv1!!LUq&+q7oUh!*vbc|=0MEQoE(B+5XG!Y$k%6*Q1;Gcc)g<txm
zfBwGkwrgJWi)B;C@1fM0<~i4B(?9>GuSEVtPulK<<xNOF^3o6cq}`m)*adz0$GXsb
zexgl2{!I>kh#%|7EA|6+cRuo5o{+DMPgK`rK2wJ|_u9YIA^ZmN3;9z%wZ47N{)s<%
zO&-)QpIR3m(Ef{0IJsT_&u5l8nSFd8ub%YYnjfim?ceePzm?aVH@tV{x4x6&_nt$V
zcaE2zoWtZ7>w1n~>+?(g!Vkp<G@rjhaUkCDZ(aKhy7Y%X=Sy(}`6<5mul4vpzsC=M
zHlKd9+0l6><C<O*$)89+!=ikY$Zyz%-1KH$`zig>m-*}<9~wvR@&iARhvaSkN*?(U
zU3RlynU}~O=p}#Xq9fmHTaQ2S6ZF($#^FmoB)4(ar(gcS{``r)`4#`6C;jYUJUZfu
zzO~s&eBy_m^Ii@=__PlB(MOMd60NVUq8Ib&7ysyq56{!S+Wf4A<|oR7_W!gadqU>{
z`A{6NgZ)x|(67yI^eqnLZTZ%BnBMdI?$dX!zGwA22LIo%?{WNo!S4;|+qo2aJ_zxV
z$S(%rp!eCHd;dHLlaFEB{E0ux!~B(=<PCKXJK0aYzYq`NlK-Oby+`f`##!HZc>_Iq
zm52Eg{xW`(AO3_MdFa9ViOw6oyJ8=F^!n#o`ynJ3d7<^miGQd+5j}is(}Vbjo<raL
z$+z+-Ijm0~+KJ9b;#mC4C-Q^5XukUr?L>4Td(s#ClLtCK$*0a&+Vmm6IUjlMe8S)C
zcg}szM>(JV$ZH+^(+8v{a<hYZ;!WPAudI8t(?42m>m~aCZwFz{%Qn6f#a|0s{I>eV
z56n0n)K0!z?UbX%55F81Tl@_>KlLaNw2*(baL~H)*kqA@`M-M7e(yQ`va9v<lN%lV
z#-nE(x{&<xn)&+W5pD4-4#lTDJ6S~E{i)xr`Ob-dkjL*loD1wf_(G2!=#Bs4lYI1{
z-tnAX<*l@nwtP%K5Wn=Q?!#XTjfePw>I>if%7@nV-K9Fucc1$G?#lP&>L~9Ky{B+*
zz%I@O@;kZZ-N_<(#H-(ju?zoXM|R~Gp7RH7_SNs4;<@?i7Dx{DV)~o92VKwU*>m#f
zr+;=CU+?_eKDF5+t4H{Q`R>uk>-%?l;Ft0hdqDKaK_BE656<`a6$j26<m~n9Iedw`
zH=K9=F|Yo{dY!iOhx;G;#y3AxU!jXH{-_`S^uS-pgHG~gJ|v&-!P&)g^x2;tS}5*X
zD1M=Q2^Xzcboc>x9zK$I?b747+-C3To!>Pi@*DDyn;e7CIgXw93FM#j3_T~0^<l;<
zyFvcTzsZ4rex?nbYuSNcq2upnvM2tn2l*BMqj&sk^FwGJ{_t&`MEb{HBL5>Feype6
zLj1~S{I-SY;ESH*Bl#ocC#U>&&^}M@_SuK0bmXBAam}x^6Rq#NB=1Mut9b9@JfPpX
z#d!BA#@QFem3uV%uX*l~_>=wEy|4Qud6B&SewlN+eb)U8`^!iEUc2A7@_+j+y|W{`
zIfv7m-y`_F7`usEf3E?b=&=tJAAXM|PveJw`@0tYZh?67yE1;e@96F)&i&k!&iAGC
zsekJ!E3e;k&CJg4>)k)m3;9o-zujdw?KitS`Sc&{x5?vEJ9Okbzu%`nbpbuIhdPBn
z(gS}s-hB*u<fCVLknf=V0P;6+W50Dj<oxWtBR_PGHjbY;ziT^ZyGKyJ^H=S>M<Aa(
zA|K;V-gZxAJ~^$Q{Lv$M=?Oh{5EuAkN8|aAxZ%I@20!5+<fR{WGhaPw|HYs6p*+D4
zjC1ef-orR`mHMAwu{-_Azw|@T_?AD}McX*>F8<Xq@{n=%UG}sd`;pUl<H*6DIo>$>
zb)K;vzm-?X5BaCIariJF8VA|Wcy=@%-`>B7A9Bg(?CH7upKbi%+dAm+e`p+j2ci9H
z5aOG>*5?Q8PcF!RjbkVFa~>4W>I-q_eE`2j*F1Fsl<yMNU+||-T=C9xzV_(IGe!BN
zh4P5})<XLRRR2N0d);&8og2L9+F6~vqhH>#zxn$L@{j%3-z~~|onGvR#@P?8<L{;H
zv|!bJ=bpAs*Sc@?duIE;d?`<QuhT;EjLU!U>93i0<`X0Iqdpkt^XT5p%P*hVJ#f)m
z7hiDDjPd=Mz7o|R{7gNh-9mLuBL9`2_&xpeclQ|D@}lvOKZ_Uf&Cldv-z}*p)L~Gb
zHIM&^S9ICe-y?$bpw7wnZ}ubm1hl_;uP9#eEw0Dw^a~zXch%vKjqn%xqz@?Xy5BL*
zerFx)(}Q;IYvvn=-?VSer$_sreSyB+LqPP@0rcd#{3ySXPd>z#dF*Yzy4XFK_0)^>
zhc5o<jo<PMc}tzy>))gB1NADo$z>is@XPMxz%P32hHv@E`u-k*cuJHP_>cNu9Fd#f
z@dJpSyhCs1<C{ECe3F+x(f6SJH=pOeRqfx^0Da4!>_gw;k$=l?>SFtec=LUjIN-++
zeQ^loQ^?=MBYONHzmHJYI)Cs(^wf{m$G1AjIawUi5;@hyzWel^K|QJ7RL9YiJfL1s
z57=*=!^tb&?Ca`8b+9<HKKcCpStxIk&-q3h-}v?1y^8uZ$Ela;o8P3L*jMPmc?iGu
zY4isnd&!UN%Rksp-jiSH%f2X|svG1-`%#W}AHq-Z&EK_s4`tuuuh!+a_%cpCU|n^c
zarElmhMe-bI?H<e*LdUbZ$AIRAAfQF<(H8D(X+2w&pdLV?|z4z{G0x?(ZwIsFK&%b
z<Y)30zZSpZ5R%t8er^AdN3_}5c>JR$4j{YXhrRiM{qC6}IR+v5=*jyYcBCirq9=~+
z6UN&Q=!KlnIBouc4`?0v7UGlN$Q$lcwXF~NGkN4|>${ipU5oYE+jlMQo3-i1eS*9!
zZ<$XndXuj_w-2GCU!5qg<Cp&U8$<_R5M6wclicz;x!jB6Lpzb*laHUL-sK5)Q~$|p
z_C4s{*ZEO9(Q|arRnJ1}LwP9Cc-Z3CI?%X8c^Ex;(>U`z*Uw*#hsl4he_k-&bEqF0
z5A{QN$n!*W)$#UCZRcC}@#Ldtau`o;_Y2zcy!;NWhfn*PeT*F9!M<a@eTiJgJ6G60
z>@Qh=;cHm5K0m;({Em)(d}%}D=+n6l|K=^~zc;2IsLo71sK3#n2YJdqChp}2>xg&t
zqx@>TwmhP(uFST4C7;Om_D$>Bchrx@tJm-iom=g<_(tD2_qv{2&%UL;HO}`g>O6Y0
zf0&03ztnbKL62O{(dt=t<TvbSAF&?3&9h$W4}Ig=&-)AeIJxX=^vq7s`uN8uy;`3f
z;#EEoZ;9ejUK9uRL;gdb<kfF~!iV}Gk==5h`K^BILhFlLNKSQ33)xjXLiUhP=usUo
zQN$Oz{D7Xs1-a-;zj5Yy&L8yC8-B%~bFqAb5BF{S7aeGQ_R_X4zV!1~ZS|8n2;Yz%
z`Ll75Jl4ZEq%U!x?K}$UUmS=tbfEo8e9CwHO#YGg#2vIAy^H%?kAC%|pLq!%;@SQ|
zUgu^gUY)<)@8Q>dC4I2Fe*Du1`Pm;mcEP`X_nz)Gt;62rq-X0HM-KPz=*pAiur59F
zW5^Ha%{ulI^Vx}>?H`am*b(3S!20wjpYUsPiAR3S->u7@^v=%o;huzD$%TJ*rzh*l
zU*^#(dhS)^efxoR)sgH8jZ?3w6VWkF{i)9Me$V?n^SswH4%&x|(~qv_{=XIP=hS~K
zq<{N4eX+aWQTYA3bB}t=IZoYW-_kbD--Gh^oy3j2h(3D8d+r?Rdo0M0oR8&Ob)Iu6
zIsDGfKJM=T`2K5~H_bWzlF_xsKKb`+u3P^FGsgd3)eqnC<Q50tH*;*sT8ppy;Hk66
z4t?2Y{$|yJIb*&HAOE|S|Mlh9oU_Z<M*Q6Yd@X<DNxOXiBh$yk2R-4Jzl$gP2l?sK
z?`PcqvY-1CavEoS{6O>B2RfhAAN^>L%h~_?o$O{lyZT-T-`3N%ui?{sPV1PDuD{#B
zZsd6FquXA(-Rh}ho-bIn*LhdpJ!Om>eh)+rXdQI;9X`<Y_YsWKf68ggPx$ftr$*^}
z!@oNBk$3O3))@V0cgGz+^{&^?9^+T;JIIZ$dHxQXzklZMr18g}JbLB-J+}3%aliY|
zRkz%F@}@J#*Y*1Z{;<c%_a1q};%Q^_$RF@69@qsx$)9*+H)wrvK>tv@vk$xA&${}3
zf1*xfKi{qCSHIv>9GDN)yX1$~h3G;1CUlO3_9ZA^LVgOxI}|7IC2MwHx^%`m9pooa
z{<FUPAg-+IyHUu`&0{x+4z!OZ%Cm{;3H2|#@+<3zXZw7we}AN|Vi)`x2leBNUesgW
z6S=2QXVEwR(B>cXq8}al@I2)n)K(9vAIL`!>}g$f6*=u&{D+^26Yt;bC*;Mi`&)98
z%YK3%^sVQ-jgEQ@9pl-R9PFTez@L8i-t=ZZxzObg>Qi+mddBf@`=x&L?59wDj~;!C
zUvbV}^hZAWuwTm)>RR<V|J2U<l)a(z5)>EEIl*)HNzSj{Z>w*i_3^15<zM0leQ`^F
z^tsOWZn=G{<!g8Di{u%0%=dcijgI=5f4YA)j$g4m{gK0W+w|dkZFZzz$R5x=22_u$
z|9$_aen!XnMZN7D<J`l~(WMvT$i@Homw1BgBY*OHdLRe=Y11FHK0m>Sb)6IF760VZ
zFK?NT9y`(xdytPj(7BRa{E6Kle;^;a=+LkDWfy$WA9?U+Udrb_%yay+Ytk1t)~9EB
zB_9;m^eDbOXFv0egY>LzJ#x_xe(_Ha)}tSOM4#xeFMXH~t%II*<W1*ndW6>X+&V2Z
z9<rD3@LPy4h`xQBKcFig$>-)l=U?%}KJ;%seR{9qUVwd}`xkzX9{by8=#9SFhrjv%
zB%!>L$Zyz_-2BA8fp7N*?ghjldCX@=bhM3^H?-L=`H_FD=e=nQQ_u9KpFI4UezozV
z9`t@!+rGt5)I;<Ft;;U#lqhZ>J3xFVdhQ+w(ko;S$R051ng^}ZLU91;585Xa-7~em
z6W6ca7GL&z^`^Ly_rx)KibHjWJgY9#uWn-ph`xO$QC`EJc~HF3kGRF3=kg@GKy;j!
z)Klb?ztorLc`l!j+kE5X8AvbcJ@vM8zjLp;&w0&!NU!Pd<VF|&=->;#;)^`=Y#rxV
z=M(&Su3jaNdkpsyo)1Fuv=Co|P@T}i<b$4uMR__=e6&!Uv@r9;WZUe^PVCI@=-0lf
zjc;xDLj1;f$RGF*|8?)kA3Wzr^v~bi8+mSB_8<qe4*wG;o<rwqazpFl-+X$4^e->d
zkNC0<d&(pBBkSXnUfIWfEq?9q)<YK`=E-CB8~mDQU%-!j-2UoZYClUq_Wzta-=Vkb
z;s;-I*0(qByqA$b`GbCZ(33WM(!V?}@6xaDSe#SHjSuyg^O$|beUyDb-jkoz3+g-L
za(`h5b&=m)c|XOkAw7`Kdi>4%F8iGO5r|Lvq6ck!=Kez-`Zv%1$8Q!j{%!<%{Kq|x
zdocW~!_))rb@;P&-BXcA99V~b*i+oQ*CwZV_~!5Y2A}xgC+aDF!p@K#l5c*4Kl<RG
z^k_f9AH9$Vo#cx?+-vZ6dBM8;S^Z_7qZj(+@AM+y@?&~pKl;Z%J&7~#tMNxZa;RtI
zaeVPRe6c5a__zBTh;R1AkA8VeUi6$@`M2lvqu+S<11&WE4~5>xLi&eq=+3(1+NWoY
z|6bBKb-TYK;dh$ioxj;1_@Vv8zGgqww(qi=I*nbS_i)~Od5^=M-n)r=@x)J@C-@zI
z<(K@$IOluxjpt|95r^UlJ^JA9{14I-zb3!@Y9Chzc+W_0#>?N<)s`<HxwZK<^c-I;
zBoBEZy7-`fdV=%<`2);##SyfxLHP<Fke#%Rlh5Q)bi_TsRmb3mAJLQX<kD|l^XP-!
z>4hHYK^z!|4*xKYzgeGN_zmO-kRH$_XX+vS%6#(SAHV!G<*^?6=^yl~4wE<OjU41B
zH$I^J0@ah&hsHy4LHdBkLHt8}LGvK~Aw8%Qq4<OD!T2Tqtdsg-e|+*2erdcsP0!wo
zxG$Ce)tUAy@=O%1kAM3oKKbwF{om=+lXdZnFMdg0awVd}A6n?%G12-h9HfWjC)@Ng
zERr{oURtOwX(9b4^0OA*Vf)2(3;Dw!bnludA3*UUep<-hDR-;g(%(e;Q*W*H2K9@l
zi6T8gd0?`b@?=}wv77f1ei!AupK<2-BinD-+xemyo!{FWyZV}!e)5L3x(__P@Dt0g
zn%d#py6z3s)qb~W{@rJ+`OXzb&F$<vaPb+-K7QY;W_9X4e971HGJR{azq%sj<&XSd
zn_uD|zx-1EMPI&g9w0A2k?+tW7rk1~eD?SKsrv+RkB)VGF9PKu>o4k`<MEF^dcz{U
z(igeu71FnJ0aS-Tbr)1O%g5-bThuksJ_^0xf#^W|+t-~N=%<D31PAHMILO}iNA)9l
zA-xy}-5069>4SdAVZ8UD=o&}QkX_VE=4)FQx<A!6AG#;OC$t{<)EDIN92y7d3-UWC
z4<w2YD2@{OZ6dpg^F)8QNIXIL0(w8%!lWyY!rXu5QHZYf<VPsKdvB+%aj#;W`&{?F
z<k#Oq^WpFP-KPHE39??Ib=XgvJn|!CKl{(w7vFo-JI|io`Mqe`@x+sky?^_4XLpI#
z^WFX+bZ$v(>4kr`F!eybM0#jp%B#(9?2DN<wZ)x0YQM6NxSw=S=(&47`?C2^f1>Yq
z6TL5Tp8|bnD9=FoLECv(eu6p9c=^e9;>LS!p7$~6o98(`-A6<BU-Blp+R*pV=(t~_
z5A@tGd2XI{%!lTuT;{p2w@*O(fHrx_pJ*N5a~scY+W4S1^sQ$<;4l2v{Qy7Yf8y7E
z?z!=hpPHxN{lD+l%y*A&efLW02>vZ@#F6pjNpycnfAq#4>_#uft6Su4e9#vppZyhI
z#+wg~rw{tHkEw^~MLsZI9i-oUZF;vJK8zzLJsIzL&eKNU?}g}1eW2b!pZ`I6X9s@i
zxw!CLeDP=BYg-4h1OFf|dq8&4HXo|r$VCougU`&Le{OB(8~YN!wBPa*`yqSqYwugd
zt$5(q&V%9sns1-Bo;bCx@%qi<$LgGX&i~2h+(tff7-zoq@UI^o^Yo*qZr6`bbet<a
zH_!U=HMt==+P-gRFR1?YeLKDT{~pM%Up?;}>Ymhk&v)c!^hU4v<{#bzcz^C(X5WF%
zf5zFb<QH`8bJpVz(7Nn}AMe%l%Wv$5FL}*#?{SPHxApb&Gv9%TD}31>z4zk>_9gLa
zKlh$o{fmz8FvyYXySI|x`L}&Q+j{2Hll6@gH{OqvTR!q0-916C|NANO`MtUN20i-$
zy^r1X?d3oC*EyYgdT1SV=~upT-pM@3e(w2<ckksKV!yRt+5gy?{q@7lALLMX$z$pt
z^z760&93IFuh9`-;z6B(4}9}m^Q=qH&e`~|Pg`Fc;!oW1Bk$*()1AZJcj7DS67gz1
z-xJB_`0@L4>+m=Chx{GC^zYm%|FgS%t8F}gVt;Y#x&4ej{M}Z5BX6^>b;O1HPjRbX
zUedPCb6xKo{na<0^RtH^8(~L!@cSe2O#k9ezx@h|NA%E7^xQrtZry{}KeeIvhV(D5
z$}9M$Pbg2K>psYQ^sK|K@)+bF>^)I*kIPRX{Xu?|$dB+LFZ<nqe)71Fu`YSZBhJh>
z4jN~Ffc9m11|Pmh6KD1}<K<U*MI6f);>G@q?$WIf-0|jF>v!^oekgBPS3Z@u<O%w5
zE|WLWF`t~|BbRmLVR3AKlQ;N<zw2+Fb<88b{H(6QFaG6m^XS!nO@H<U^X+rDP1$+o
zMz5ILiErcZLtpl{MDnOZ<N^AUZ}AVULoe2W>I-t}mlx@W-N>V^;kTb1nRCJF>!x+u
z;#(Z!*Eo4V{h^NMf8HBHafgn)B45xqeefIc#-E_?Fwq?r$u}(W7kXvymOa?bey7d8
z>Na#Cdi>EouOI5?7ux(e^=N(L@h_f?Px;gn?8Of<pUQ9cH@~wM&+MX3$bM~n@hAQZ
z#ZR6GwehQNrGI$}og61F%)^gyo<nqu(~o~B-}sJ|oz$i5A#V8rdg2y6>+u72GhcqS
zFSwUN*Z!ftRTn!ixzDCI`-i-WK05pgeRk(B>@MHoAEL*u;)CAUjbG9mJ=%B3BX7tL
z_DOl&{T;pNcW>srOrLrGZ$0~#an=>L#^INqjFZ>s*>ie@=sRz)13uBqxaSx2Xn$ZQ
zbm-l4dczO;=0SQ*M2EkkC(odZKYmL8`uQdOkwe>f_nz#@f6;};JFjb-&kxae-pcqF
zkH(2pdgSl=p*W`x$o`(=Lwu*+=+At5!Kb*hpW@eZ{K#MG0Q|GB=i-AOh{Hs52cdmp
zqDZfU(C>s3^Sx}gogaop=K@F{{8!%6reAT-j)|V5$FJoBc2K_?$DjFy_1GQiPeiAM
z{9zCdvMawHgh^-6x}HDR=l)vnkD~!~Xrg?{zvR_Kb&Yruht6x_NZlq6+K1(7^TaKb
z*X`f(fc;UP)o=es7kZ9=ZG8Kkl{_z>yWgP?=P`Ac`0||rJv!Gy>-75f&+dQdH|rq!
zv!1*I{oPG`y8n{bjf47~2b>qR(a}#1?W{ZHJ@%LH$YVY<4x$UyKm5h{2GT#o?;w=V
z`4i+%Ei?|Q4<I>V(oxqW8khdg{~&$ye|~Nrxzt_qHh%dH{i>t+A-d>6af!~b$Zjq4
zyEQ2O)VJCYfAWR;Gf^E1&9kqQ8y)h|8+!C<p7`dM=s^8Y_I6Ln@A-#1T>gQ^CDKc)
z4rK2?7Sc}(-KP&ie%iv+%d_2PuVJyb^00j${^i`xJ+}8u?r~CH=XiRuAE>MSzZ>dY
z`C8sHkABq2>TCLNujjtrxy8MrwtG738}I#_-?zH2B%k@@^Lqk$%J;9{XIjVmGVe=`
zqepu8|DE|=t@-%(9crTQQRxGc3w=mm_Cx!aebBwDeszYn_d4cl(}R7S9kuP}?!}B}
zUntM#e(Ijn{>NX?ckg6>kw^8*6XbyM47%nIi}I@Z+VW`bf7u?izu1>r*y2;(O>EK0
ze)||b4vVdP-RkEjiR{xt{MvV*xK0#zEgX;c3*Wofhc9|^WZ1TS)4s~S_DSfueb+us
z4s{j(f%a$nEWdJ3>APLu>q7nN9rM-S^r~MRK=CH;^6x})xDUmz{T`Cv{R27tJ#2iF
z+j!smx<^9~T2K9h5B^KP>_T4OO|YN&+Ux{9ch1qq7rM~jy<i7&(U<$({JxaDz6Yfj
z>uRG*kC6RZ@h1Pvv+T(q<O}n)oiq58=jO?;&X4l9{KOvUnV*>a)01<i{Sf`hVy=tt
z7Lwn-V?X0x?Ak(lkPrAnB7YMn=<##@ZJszW4~h@p_iFPW{=|Ph=eOwOeDe~~F&^?~
z=N^AQ%irnPf1?YxKjOG)<NsfRzhnQlC7*uJd!{`(>N)gx>HYmSe=i>TJNy2wJbM0~
zyuTZgbo@UK{tiC=ytg68tp5LB;9iAY`GNQlU;LHd>*wcCeDFW}E&ujhKYHR%{baoT
z)cIGv#*fv@@&P`f_u=kUjK6Z<c~_4dGN%)d==J*dI@wPi`-t`MiEnkQHo3$p{+$b~
zXMOA1w~R-}IC40zWZtsAe)G`5H@Tb(tVcfd*+D&lPyNPgTUWn2UOnSnfKChX+d^_$
zhkf`9xyhAiUGM1=*#}zB_j=@Tj)Lm5Jm<99&TY=4<L9%@F5mY24?o_l<JiY}lpX0?
zKYHl%Kl%|b#%q&<UVU$89lv9QzPqFs=QYSq__i-3;_K|i{fc<Su`3Q+ws6V%OEy?C
z^`k5HU$JD`Z~k32Z`pdwHe530BP*6|{G0a6mrPwUb??WIIC;ggi%!|?j*;o=>B7=L
zX`nPv8u$~{z@kI9+~(U`Ts>0wyL$K&HFt?s8Ym5v21)~^fzm)}pfpe#C=HYbN&}^V
z(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v21)~^fzm)}pfpe#
zC=HYbN&}^V(m-jTG*B8S4U`5-1C!Li>Mw2e(3)kBj6PR**A@S8`ss&1Fgj73q<A&0
zG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v21)~^fzm)}pfpe#C=HYb
zN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v21)~^fzm)}
zpfpe#C=HYbN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v
z21)~^fzm)}pfpe#C=HYbN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5w
zX`nPv8Ym5v21)~^fzm)}pfpe#C=HYbN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=
zlm<!zrGe5wX`nPv8Ym5v21)~^fzrT~CyzYvD_@%XtI=Wcpe;AK_`BymG5UPLGGS?;
zG*B8S4U`5-1Eqn|Kxv>fP#P!=lm<!zrGe5wX`nPv8Ym5v21)~^fzm)}pfpe#C=HYb
zN&}^V(m-jTG*B9ts0Qx%(fnJ#{H<F@2H_rioOkYe3s;R4KDT<9s6A`^#58cm7xsDK
z%3W8F7Cyi9Ffl9E=+Z!Gpfpe#C=HYbN&}^V(m-jTG*B8S4U`5-1Eqn|Kxv>fP#P!=
zlm<!zrGe5wX`nPv8Ym5v21)~iHSq3loY(u%lr^J;rHAr^-^Rh<$f*&ffzkjCocZ)G
z-gdxx_m55%(J3qqlm>>^z~BAvmZ$9a|LlEvyie8o|2dotWuC868M+!SB}qxvz8c68
z;g&L7Ly1U8qe_JaTqOx5Ls3M^P$IHTNkxStGvS<L9uAIq`2Fsy-sgMoqvx)*r_X2a
zbDn?B`#fu{&)#dl_Zpt{JkMJD<PlrYA0Et^my*JOFdz&F1HynXAPfit!hkR!3<v|l
zfG{8o2m``^Fdz&F1HynXAPfit!hkR!3<v|lfG{8o2m``^Fdz)Xhk>58JLSLfmRWXo
z#rUu)cNGSN0bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&
zKo}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV
z2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801841Bt*+HbF2yjlqg4}P=*Cv8v4
z2G78mF89u!cHwbbP<Y6W?I8FbBzF)7gn^vGK*L^D?|-lD78PsEDg4TF3InlVpn1)C
z*LVD9h6*djf&;mSFdz)%YzF=|ZTxFRroOENo$-*fTa7%uF!0|Dymju#KbsZXVTT3(
zyLKW81HynXAPfit!hkR!3<v|lfG{8o2m``^Fdz&F1HynXAPfit!hkR!3<v|lfG{8o
z2m``^Fdz&F1H!<6GSH&pKR5j2<Vsu68xQ}<sr)Yt2m``^Fdz&F1HynXAPfit!hkR!
z3<v|^WZ>lz6I<l&IY$XblZSA&Dyfw)APfit!hkR!3<v|lfG{8o#DamZpP$yJ&cHpk
zDgN~)HLQN=gLY6@d|bEAW4G?M{lXbz-W~bRuXoxx6=Ttc<Q~F+Fdz&F1HynXAPfY-
zK<bpSCk8B=rGSm6R;+aGu@qZ}-P+t!p?>Eyn?rhj+lY+??ySCjmn|3#9)f6Va!p}C
z7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4P
zgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&
zKo}4PgaKhd7!U@80bxKG5C((+VL%uV2801&AnXie950*eUwMD9DgL^;?ZnTAercOx
z>%V;VdGgX#$}QGP`RCTX6?Q78*md!}bI*+4tvEc>y-MSm<8~-hys=L0q7x48Ri-%X
z@5O%GS^9u-ic8e$uLjOPsyH0JHFZPH=T9i7cticYwWn4%txU1#rw^{E^5j9qVfVHL
zb}Y(wPPxUml9nH7KKZCJ#p-9)|6@@9v&s~wEV}jCldqjqg3;t5?Bk=<PZ$segaKhd
z7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4P
zgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&
zKo}4PgaKhd7!U@80bxKG$W9Er^!bPL@2-<>`;FO&BY75KKo}4PgaKhd7!U@80bxKG
z5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@8
z0bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd
z7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4P
zgaKh7UJTr^_qMAauK0@*j5ZJPYCrM5?@fJYkG`wv3R_Tkkai&bK*j-Q<3PN}Z@IrP
zAPfit!hkR!3<v|lfG{8o2m``^Fdz&F1HynXAPfit!ay__DAx3c%4dokv9lXSgMF!`
zFdz&F1HynXAPfit!hkR!3<v|lfG{8o2m{%Nfp@C^@hn517}~zmUz1BDmrPE4ZD`Y>
z$+`ahw^QLx#X6NtPIzT#r_%qHuS`x%PQ3e6$5)4T8q>e-T9rHdFcI*Rbi6)!z^vW}
zl_@SC@nY4r9d;-V-+6b%;NI_VP^Osnx8j9===-f5RSfv`$mMf515Z``YE`L>O*X}`
z!;?mCE}miMbj;z+Lmpok5C((+VL%vg#lYxWc0S*&Z+<IY=&I84BVj-o5C((+VZb2+
zoSo^`2M;x_G<c_-Q;_39(to!{I&4-_To@1rgaKhd7!U@sF#|VMYgY7)$1bv*Vm78k
z9xX%+JaV>O_a>!ID&Vr~pO|yed)pL{`Vk>-ys4o&e!<4`{EWy4b{cWtnSKp2Rp*Dg
zZZEg(EE_jq>i?(Zbn6dpeRlKuU74lcJ(y$GZyXO%PY+R7lBx&;!hkR!3}hn)pl?8b
zzp+E@YchXLv`ph4Q7)y=`b8J-K4SZYK~oI<O@~p}AF@mRgKiIUePKWt5C(*S|7M`e
z%yQqIeda<-P<V)CJNWNbD3UNB3<v|lfH3gq48VRN=*}M-6mNh2wh|KZKXXat9kWyI
z?2bDBU0CwzdkRavq#n`U4=(Rd-d~i@GeG;L<D<yJfG{8o2m``^69zc{c#lR)pK15-
zUYo;;x$F0Tcl~@t5&m}$aV`=4UDRZbmz<K0_*Lj5f}&Sv2faGzc8sw(>f;2RH}G6R
zW30~83hMoUH*S8U>4=9jMP2+D%}0Bu>0y(@5;unSjrS<%Z=7Y+_(XdM3ggX3P|p>$
z>pu0sP2K9WnyyUo+*fsaJU@P`3L5L)wsq3Fvg>UQ!^X4eyvjXK(CL^D-D0fy*69WG
zIk<TpdKH(?rJKG_({i!)zna!7YUQ|g81MPS>;0o<C(_lI_D@}ZMn0WaZkX~?O1&pC
zjL-GPbvWeudr;|Azo)!}MW*$HjQ4TT?bQpu)MUSHiuip{=yu$q;|Jn#ZY;C#p$)Z@
ztf2Az3eA6++a#A2G-7{QSkUeJlHdJre?`>!q5aWRj{0-w)cSkXld}8oquwqspT-(Z
z=Q;TIfUrw*`l2pfrw8Lcej(}$gx?vctB=$Cp!8x~Hz@s3uA8&YmCbJN+f)yda{e$7
z^?6t7pS{;f@_uv1`{hQnSo;r5@5j~GZTgz~Tc_52Znqs(bo<;<`5n&R6#V@hQt;sQ
z{RQp+4cqm-@`c|AZLgQ>$Hp(>7tgV_J7u@~p<g+jfAKp(VIT+w!qU%R$A$gXFVxwY
zx%{xi4{-I9bZ6hC7ruVyE4%yoV)cAcYfrJ_C8yjch`Cs<DGWq~fw1)RZhfW=S^Cp9
zJL(t~ebeuHgrz<E{akWA*}v{(|9a5<Lf1aCTU>|f^Kg2*)A+5NUf%Ea3xAKU-3FcS
z_jAehgaKiIXTa+^!Zbcv%KiRbta$Ojw}5EpAl_Q=t5u~kHfg^%8x&7r@ncCn!gYRo
zc7Gp!`AOq-9k260_xq^C^>X!>^z3;~RIJA%X-C<!9l3q4v2K51y`NkDOn>L}xpCci
z&BJov<N>pKA5<I<Zda*M(L?;Ir(2%5-=SZ`8oByNx-cLNM27*d<3>>Ss-UkMoi`*k
zjRgZ<`$1TKAGH5ht{<DYo>;W`DBmM2{Y_Voaa~`VbHPXHwxICvpY1@`ZsMIE^?2d1
z*55BXf?Vg%&t-|)oaT2wcl-RJ*ZHg0`TbtydcXUroSqLL<)eMB(fUVgJkaK`m3FK#
zEc&|FdEV`O5cKb)Uq-uMhV}gg-EOnX^_}|PX?bp+kYDLR+yCYI@r_3Y|Fm7jx_{Al
z23jAla=g}6y8U<HIj62G{2Hf%x(?n?R4`WU$SohS>W8NLY-u@sFEA`f>lam-_D|io
z$mNIi`LQoJUP$?ksvN8FA?o+x#^12McdouP-K(A#w%^m|%PBI|=cyd^{zvN{)_E!3
z_2SyQS9_}N_uPFYT^JAs&SwC61ID$meJ_C4=lmy-6k#C#44@w+O=w!@zK&;=pzsie
zcA(RjB>&Ij*o2p0d5QmsEj0*|0qH-%_8;L__-*&_UOQ^-AjmQ2Kd&7tf2iC?&ab(6
z_Yqs#L2TQ>fA)3xKmH8Be&`nC?XQsgA8K4_@J?IDL;PD>v}%yu=Y?p!jri5p^t<YQ
zm5bk#=8QX|U;F*EZs$+Gf9LmeMfG}q^+&nRC6#x~PPMZu4^eI3Qg2~E7!U@8foL&+
zcxS|&L>>LsU$fiwqSZR3w!(lgAPfit!hkRkW(IWro_OUya<*OfCZ$d)pj-VmUiFvz
zn)wq=c?xqMlDY{4!hkylg8JUC+xa4_zsqj+40#@Z&%2E^pYD3Q{7M)Q2801&Ko}4P
zgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VIb-Zlxu%Q
z+vo4eXRVk&{rcYq<+C7n&OKsKnd14ZuwvACm--6>QD6Y)stCsRJUx8B4Dqj>(tcSq
zI`xpEh<vYz?~g_WqtJJxj>3R2APoE`1F?<^*WC|w^E!V7^?P0YB|qBx7GnMWf@-Hh
zUl-@g0J-}uAMs+<wH<aSid=b`j_>{Ay*$x=?7kNsNU?Po7QHbn-;0)dB#qo$Ji`_q
z!p#HvH5~`2jPW4o_gVW)`o`vqx7$F!?+1NHFlR9Udx}@=G~&K9{TgH{ApM@QTN(Wg
z{Y)@gdTp$FleCLywTq~}FTdI|#<f@>)-O7Y>U)uT=ZyZ3bGBlQZqI+=^eVsly|nwB
z(e7jQo}IQ!*xh1PzwB}!@SI(q$Ln*zPHKA2o|N<V+`^p7C)ynl_O)2^;n%u?yN=iX
zDZl^UdIs8$;8&|kWo)vA2jRibcA#5-arR}b6S;h(LmvZ1ZM@^geMuJvgn{TVfbj>2
z^|e=w4g*qCVL%uV2801&AbJc$ZQT&o`bO`~OO1sAVL%x0mjT$PfLIs$g<&24gRU1B
z2Nm>vas62Pm*RCl_yJ7YEBF!o-~J}y#~DBVj>>X1VL%uV2801&AZIWD{Y~@^XL^U<
zeBnJ%^!q-3U9aYk^+qrFFpniDJjAjc#A;ObtDnT{Iz1XKeWu;Rd+o3x#|P{Q*$u<S
zKUbev^((oLFdz&F1HynXAPfit!hkR!3?w9ExAVs3ojPNH+Wmr(oz*Jmmgejfl5S=v
z&$%r}o;*wpM7_TvKF})$HJ|9N+wQYd9sj;$jfyv-P3r0!>wLe@dyV(?F1o#X!Izrs
zw}VE!Pg9InoMgQ38}xhd>-t{(K7NfartL|}2?I_Th`Jsk^*3AhIOXZ9=7W#Ccguwq
z(Cc{sVgJ2TeoV{P?7e)=gnaoe!T9pvWK5K@{xcA-{!gF8>pH|#Z>IMZSr`xogaKhd
z7!U@c&wyWcot$=^SiO19UT(T>i&r_Fz0p+vlyc!buRK=mU+yCe2m`Taz^gwz*1wPH
zJQ=U~Io9{(-cPL4@!a@Mtxivo{AiuWkM|oE#5ww0JLk^RLV8%9L+TNXe-gaLeuj5;
zN9y;W^E;j+R_LtuQd$@g2801&Ko}4PgaKhd7!U@8f#@(m*IObd517^apfcrQ`G^;*
zuI;cxadsF<Uop7%`y0;VC*D0oXHN*5e>mQJ$^HK~1Na_l)cQN@!5r%PiR1@szZHJ}
zoZqPRv#`{c>NT`p@osPbdz6*j?9KrE;z8?QvEtF``B1T{S6J@jl%Jr>=YQss$~$JK
z+F?cR9N?hR(N2QqU9P`=(Z#!u*lu~id-4l|YENOiZdm)dX}#jjr(f;O^g4b$FYZV6
z-=Jg~=MpO!b_uTtzBu&jz0m8p9S`Mq-sThGPx$kqkGO3=rt@CB%XuBYgSN9^9UXQI
zng_qGPv<3_Y^q0iwR5=bXR-R-QxDwKtxl`y$}4hwbNYzW^q|HQzj=0B&$_p5owTm(
zdYePP>lv%-c^!wn{@(5TaQeI4N0udZPdu&Sg?^2rezyy5T#)o|en$@P#qV|y@9S~>
zIo|d1dVe}TQvYGlGRJFB>0bHr>-W%K{N@+ejTidGPgt)LFJ5)`4J%*SrC#9QFFi@F
z^XG9tVd2H=y502mVYT0c<vp0zBPe~r?RB%uxF^q%GyMdZ?_z~mH*y%YcIdZWkL#I^
zhg><l&#3V!^$o{3#Pv^I`p@}}ZMgoL<5~NJxb{iYqx#;_U!5ZQrzz@qk7|BY-#e{$
zRQZ>Bhugj`?<d;thmI3&<*>B_&i|ZK)5G?DboHU_&{PkJS2<eWsLE9D6*=7Y;h^4+
zX<V6GyJWn7ryGCbm2Wz3#GAjM?jJASgTAlnd-bav=0U%%C)Wvw-7CAl2eey?v3_oP
zf4}9R=Ix-a8*BdMelp&~V_#5M+Z~OMgsj_dN$vhYjSq6&-`9zFFR?=Ueq$}mdv|6Z
z@Y*hfZ)bc{zQQ`bOTB*AccNyW=j;Y<-)B(I@3uZs`5o=&*-ehhd&p_^<jymVRk~m8
zJl1++yzdvQ_k;V4e0Khx>Xr8?zmS`EBt0DQ7+&@Kpns2c=@tDNf1F;&>p2fze;<_p
z%k;WByN|B@)cmpD=mj6<vD_jze#38!`$P7MxSm@KO0PD(E;lZlzHhEvPSdyKd2@ze
z0RA?&{T;v4jgzMN+<2GW+9Nkkg(Y3bcUbC)_U9J8>Sb<!&$aXHn(kE()cv0F=2Xu$
zEsyU*g%x!?A>Xh1yVW;M`R2;Gy>G04hxt{|83XjaJ3TkS^gIUBa?WZWp3=H`k?Li1
z9Oe28eLsH5Z!iCPQd-hVh3oxvcEAVYtL@GDejjAN#xL$VkDP7Sy-BH)ip!5xI{4u9
z4De)%c_;j>^uqB&Z0>qAePrj6&lWzAp=f%nWZE8Mb>HlEAJg^`bh+@1j8YR}Ko}4P
zgn=+Kpqr1Jt_#9e-gNyT<qkEjG<c^SYaZg=Uc%frrEWX}x^YYLy~P&`f5P7neZRHC
z+74hp(%Ii=K7Pj`KjK6)2lgGx17yU_z)p$sVSifGKVJCN_zUNJu{y6s@y0R>AKFkm
z$?E)2*X`xDoy|I*kF)b4owJ99wf_ZOFSJ7r!5fD`*H5l5^P`jbG3?{1)XyCQrg4+*
zt}egQF~H3aNQWOWyN%a$-gEkV@wU5oy}zzqNWL&23<v|lfG{8o2m``^Fdz&F1HynX
z5Kji2uG9T0AJ1p@a&ND8f}noy<(cHS!hkR!3<v`W3EX)Ql730$9kWwyQy!!(2?N4_
zFdz(gV}SZqIT`U<eyz{sIucLhB%UZL?ZsPfjLxqSZ^FeDBb^&ha%wtw6HFd3tM|dI
za~S;MBRV}uYHGp&#)(+()06wfXMbc=dG9i_+;?Z6xzGv<vA@C;G5-Q}<16|ZMd<y&
zsOVu_J)_#ba%#P~_ozz`ny<0LlSXbXo?!=tbe|XcCHf%{=|J!TIbMi!rYNG`PSHeH
zkTSvm82G;5gxPZ*-=cu=>IZV)6@z=fzrlE4Xsq}`I}sEHauNg34{|C76(1$n{c2UI
zj7@gDdB{n9IXgaUrxEv^>DM4rYZr!|;C3A2&aFkd+v`T<ciFM+$TJB8u$$2HePD0U
z^-nsk;CDG4<JGcY$Aey>dw-JeEv^K73r55L6|a6N_Z0?&0bw902Do_~dVmhO`K$Js
z^o`9IZ?}QC4tGDK2lbqCT{8wo-?H=hZhiAxg6?=QV>$bm;>HC@|8sp8EduLlr-=9?
ztgF2u>LZvv8Nl}>ql%{c7f~*IzD0Q|VL%uV2801&Kp4ml449t#E#<aOT32?x9Tpz4
zL)(*Q5C(*SKW88+cB?<Hy`%~Q!7zaDV8jZ8c?!8|c4xq=Jv`pOcl&-)`mA4c@$Msb
z*bx5Scq9DM9ENq?56ikIsCu|<@A5lgAj}MK^9bUe>3H}=`?338cp$~M{XSvlM(QRE
z2m``^Fdz&F1HynXAPfit!hkR!41fXbJJuo2Z*&?*jq><C;;P-E?mS2?AN6pHy83eW
z<<hC$0`Egm7!U@80bxKG5C-DM0G*%1B2#~c$llHk5&ncflkfS&ZzPgC3j@M{Fdz&F
z1HynXAPmHl0mPLF*6;uB`uWO~hghw%oZ9!{_i*~VsFjBwg4!{V9u(^KvC#TZ8SjUt
zLxx`ovQxzMX+6F25zjG7?kx<sWx#3wk@A23ognb<bsUl32lKsM;Wb!ZgB};qzXXK=
zVL%uV284m^%s^1z>ovV@cIGyg&(^hAt+cNB57~aB>3a<8{(Gs{pPyqLFJ9t(zC|qi
zq?3Ea%8s2~?}PKBog(@{Sa=EQoDS}~VR>$;hs4X7#LMa0iIXumCzXx&yqi<+pL8Nu
zg_>vcTNGFHyy?*;hcC1^%t>wY-_Od~&0+7=^?S*e`NNO-L)wwFqbRo{jF&phZtXm$
z>K9I*JJ#==TX%*n9r0SR<|n7zFQ@6V+2wgTz1%cCr?hW*PSG2qtT#&g_11p5bIHA?
z!#^ST=YCA5`mdDt_d5gGMbGj3Ibz+;{I0)TFWL;my8Zcmzj*5-um=EZ{#bAHf)Ddp
z*q?`U`yitnWYhIgtjc-SL*(~gttyqV$+z!2gq=@}e%QYl@BZlY{@r)Zz2KeqcG%qS
z@O*+fg8{$iC%K-iBf{8obnP;Xb(T8Cj)AcG-E{SVUhNm^?q`~hIAcz~kaTbM)Ufl1
zbsoi7t><+2@j9aD;zwwH*xy%ntzVBuOP^`?@LpR`c#wABe?Ra#j>_-N7<i}pALWKV
zF|>WBzb2PVE|r}4+R&y$lXLz1Z>PeYighZPobbxfPNn}XUzwbkoOt)Cj;{{wG^T&u
zwJLYQhD{YtSNVFq9W)*vzi#S)9~RiU$Etg9%fe;Hx7a!?|Lpw<!?!Q8V}%6@zq_``
znuWFwU$1dR>yoekYMbJkTGut|Tj3?!6p!EYb-y3a&a^pfa4bEoPy1E2DUQla**~Z1
z4>qvY_73TXDt@GNXj)&C)4lhg^G~k$KG%+OyKN4GzCU;UHtjwxn7><!&EY>@{%P@-
zBe&T))b$sd&($YZ>Amm0yTrmTQf<G`uX-Q2qiA~5<BOCI3(TJQ*umPr*c_Ix*lE~P
zcdf94!jC4-O`E=Ky3HZ_Gte)7a>{k;?l1I8m#demSmAY#FG{U@*+H8_ue_&qyl_L?
z!@KNQ<Lehzx$=Tr(z53Ly5H^?G<WiT+bND3liF|O&^2~caoXm|PZ#~i4m;lHwH?Lk
z_ermPQtIkg_t+H6bbo2gi<8%??1G~{_;TUt_O_kfkhT-LzR>BHb#Fhnf8gTv3iy|@
zgKM-Yx<u)a-q$brk#|P*xwz0)6)(hd%<S5r{N2rxEVr0haODR*o1as_ghkW7oVn?o
z;;_~C6%PKG_pG8g^O?%i-y4^yfT$nEWj8l#+Of_V<rIJ1mcH)tsYeyCVWl-Sw+!E_
zbcpo^hqQiFuJY>TBk$b1+2-)sTDxlH>wQR>;@%G%cYEflT$Wel)-6c?$2FT<yp#Bw
z(qaDcS2UTi`h?P9)xq^{$e*0ca*IQU6{ys&)iR|+%Kw3S+bVvsWP`246syIP<t83f
z6zTP1CHE|U-<Z!!F14K^S3gd#r|EvpgT2qSn)E^0ZAyn+ech(lERwI+vsdi5<ArYP
z9roWX>G@#FgemK6w-~nPo0fn4r{zCbYsVU)f9SBm-D?_;JN>QAA=P*0<^AT9zd!b}
z(qYZ7OBZ=y;2vd)-2HXwbX?KN*Gy^r<R^W8%Uajb_M-c}U-|UDPG!uWI&^wnlyCf2
z`^yVV&#-l<TNgGs@m0mt>*m;|i2GOkxPH^aW%k)lafn@@cKrouwpSc=IQO&`ZC5JC
zNbJ(EV^Ksq<nWO_gZADsal37bPRAFzKBe++4L?qtF!z|E*mG>xgcqx(+NP-UCnQb(
zAi376jke!N`F1PA|D!{W|1y`f>p6eRCfhCQ?B}NWy7xfs?0(6kp4i)XV7+u3Sfl;5
z{rBeEYMY{|o&A?H4Q8xQ-^q^a7+2h4m6lgNeB+pXwpVm}|D5tWzt#n~&P#o6efZMd
zn-s+=>(zG`%vz^(XxeY%{eE=oAFug3JyMtNm7kk$sr<$h|FTt1M~=r0)35%nT-RN8
zRB*yo%O3n`dAc1nJUk#H?dWTV?d*zoZ0u9u`)e}npm0h3E@Otb-D$f;-FS`sdTkzD
zddI2L$`sRzmhE9ZutWKc4VLYn`gqB0Hiw-5q)_^Z`3)c6rJUkN3B3p1P-BYCAvb=|
zbWX24aLMld<DT7Zo8sKJ2M+wl`W-gK@5Zbty?OtBo1$sE(CJ}#o&!rx^y|B2hti?W
z{><gmcqPB=18%P~b6%l}<JXO_og(#f#Y^_OZuk0ow3}F=Q@gw0<#qO(u;$-VdED_=
zK00Vqr0-3qmvZ^^I$mW{y|rJ;x?)W)eoJx4)uT(9-gQQ7+Ni>gZtePE`JG?=fqRcM
zook<AO`q3!V9&u%{-m5DH{YyYIIqGJeUE3wUD5u|$=~cgmV0EU{Wh@Jq`~Fqm(8@7
zeBZ3j(t6A6$EGH>T7COryVu;icmDH%lUdhCd{;`Brjx2<X4pF9?BiU$X!@58XFT>$
z`-3(`E}qUbo!$p>!(QKgaIV?`TZg+=7hO3c@v!~($DS+sRJolt;#crqseZZT^6Tey
zf9_D0KNRn?-5bN&bQr%+Rk`4%(<Lw1qm~@%RpaR;X<6|xXvfq(3%|xSYwJy#{qu_R
z;#CX2`M3IcQnx7})sydAH#51yrTbJCem1XEu@xKCeFPJx_Gz4Z#dy2tj~!~f`P}OB
z@Tx0T-_!TScNK?-*T8jvn5XIQpx4p({|(jZUb*j;)e5*Y<!IW}_i|f6s(&7Cb7E?G
ztwRd9q{on9cWzmrvXHE^sQnu65r}#Ni*%e=r&G^w6p-qZcs`1FK48zwUTN2;-Y-gr
z)IMh_!_LPc<;_%1%Cq3<6QkECr%3&x)PL$$9(C#Iw9ZquYsb}TtrxBR?a@`rDRx;n
zDCOzfeo#8xQoUWJO0Q2*I>b7>Zr!TWu0A?lnW9rWCEBl3G<^?UN=-?qS@=uk6luSw
z^Sa2{n*pc&FD&I9`@7xx$>a39p}ViS=e8@qQXE2G;gH%B@630s@wrEjD&Ww)rw5Hc
zx;ZPJ5#u1@U|#>C_Po@w+Y}J<Ek)>G(1S3ZVE-KEgNpauzjI~p-HJn9FKhg8=RKz$
zI;h+t)<aJBH*obD^2jACD_7pGD4Nd0PWQd&;)|T-<NdhB8iOj&JALm;J1C^>$Z0*P
zzDH$h57o(>z0Pes2xlj-N*5k=?VDR|ugJCA39B}om{L7e0iQg6?`wOe?pMISEy_Rq
z_KJsX=;0fx7k{|XKbG4%q;WZja{!jf=u~Iwki9mL#-Y%Cd^E1(Jih)|A7h;iJsya4
zB;GsP58|bv7ebH3{!sWoXnZ>Luj9TztRsO)2cGx)KjKHZ_vCjvo;#}OwLgUA_p}|v
zO6KDCOw*mtF9$}g9F)@bC)+L3^(iO+FtAs<_2myLQ{=A0rBi*!RK_@@L(ZPVrDI&8
zb{SI{`$ct#`{=OEji0SOUVgi6ifKb~73pDZuuYM+J37z!C1YO7E{OTVG%nY)oNm10
z_UqI1uYP{}>aEpID;=7~RR^{2nwu}5yZ*9*n-1l&DC(Z4#l;QI)!B1Unc~sr*Y=(B
zc`l2?$6m=_`P*X|$`oriD_whTdZyyA<I*;1_oXLV9M10Ex?i_4c`T=h_82SV^!4nP
z&he=mw`u;6-YwGJUzK4~{9)$f!*)NGVROjc-)p+*d}dmXtB-CT<bJ11*R3a`l26-7
zc9V7bFW$51J^&ggr<+%~d5iAP#k$lKp?91150s;Extv~yeG*(;EM0GL<!HLDKkD*P
zFAnK@Y&)oI6E$_0twZ?RO_7UB<kDfEG)1?1FZVu?UZ{5etGgE1YXj#ezTIcS&V#m7
zM86UIodLRz$u6?4z2g1fFs$pzyH{Uead^?f{geJaH@8KxaF2<{)vN^T(?#`0H!qi|
z9{S~~;(dB#s9lr(GHTox8Ch}Azf5Rx*Vhfxv(Bx7-vRddo(*0Y_RpbPv-}UR`$4~_
z>vY7CaQ;{96UFaf&qtgE#$VVA5QhT0D*VjU4^w93GbwK#-=%=Mcm(9ra~izL-LJgh
zwWP%bt+x(0nDWV>3$1cDJd<#^*^#Vu3;f@BZagRS8t?`?Mv=#gr#=4UU>kN8*!On~
zx#~d67IW0Rd6UL8{`>4Ke;51;;1}}=<_nCY*msBV9`;t)lTp7mzkT!A+Sh(lXcwp-
z&Y^>S8TMQ38=X`8vf_mrZp}JJ4|ZhKAMqcE7dkL9@t&STepNuUGt?jcC&Z;!XuNUV
zrvuV$#Ph=s2R?zYW9ty@*(rh_AoXLR9|OUwDWbn}m=*u@M?4gGML$G4!E>QqpuZuW
z2=52=LOyV4@`yg|P9<cmr?0dY|Lgfqrxe;%ywR;c$*<qx{YD+B|Hy9{<AYa3{|*W<
z-uR7CZ66p1qKeQTvI}~(C&arQ!f)(1I(;taLmWas3JPIAfPSY#=s!B7Jm~bFSmo=^
zsR(O6)-8S^^zB$9^meaE+ZSZh{R=3k+owSDz3yA4`!FDz`pb0Xa6O8M^Lyx2`T9fV
z9kSgbcy@|(T_Uo(`2Vo-=CnOX`PedGy6@ez9PBNzLK=4%l<b#%H~Finqt0|XoOLb`
z;-_IB0~SoXWcpuwr=7R|jP5tW?>W47$Ba)Kughoog>>CX*UP5*sN3^PIlnd1b@pr1
zcFdK7AB97m{)_8U{Q<J>d<p!XL&V9NB3EC|F9tuRj%V(=u<vr{blv4tPc`+kaeiE{
z>GV7jD(mc8Scf>}iI(TaN!VL;NaMda+3kJkI7emfen{7?SH0$Q&jr5&hlsa}?JUpV
z&9py6tsI?Scn18<F^%Vta{v2$s+`J?*ZCzXzt_cS#X28$kf71+XoI*1x2QXxhn}+l
znbw!a!#S1V$JHV96`dcAu16pvZozNV^)F5@M>_Ok?4vV9`X2m}sogMa=S%9&!HHGA
zUwj9pM{(<^c%}Qr7vfi`eSJi~mnLs*ooSbA)4fLS5B?8Ri;Qkxw!`-G&JoR*Ikn&m
zLyy?l-`?WRx;Hew>5vUXKAm@{9SiZl;FYe|Xgn(XSBM`&+!#fAj?k9OoBOw|nxTNY
z_%H6<0HnWhs>FtrGAET+<m^mHce|e-zk~gYL)xzJd^$8;H=Fw55f8>8;`k5;M$zl~
z6ThePFVZ=r{6$?3YCH|<y0BAb7u4ASP4nXwr$^%}FkZVw>R;vTs8LJj?A>0|ssA8W
z@_pA%__FS-19rR-dVy2)%6q(je`NCOBepz~*IM3pYW2^DC1fGuiEn&x^6+A3QdLs5
z*`Ma$yHkyrdSrj;F(<R)y>TuY;{FhKhPYYiY1m%`eFFXql*{+SthJBUJFI|RZhNZd
z#H;r!4o%0Yes|XR?yVsw9P3-U^(D=Z_x!Cp-<;-SUs6zrICE3PIiMVx&R-}ua^lqs
z_7=a$!g)Nv4XtXmKC?1C>l{*?yLPt3y2V#5IF!}S(cX19;(^znT2bVPn)mJIJ37xO
zX&pTF%XK{$7qx~rnw-{mey%LU`CoK=S=g^<ol3J#D`2BM&32Y~^unxt;se^3{Jh6E
z2NdcJzXcHc<lk#Dul~ocBv};4=Na8<*F_gtz;j*hxV3xZ6N*FZ1JGgdUz0i<ZFy30
zXgaQSs61@o(sRcYhq`_>qjSx9SHF?V;_&Xz)4Q$Ql-uGE=cfW!4_~tH!5fb$9iDt{
z-H0U}b}9}{&l%<Fg><aPbV$EbjP^tk=kRif^%94)eL)^DY}R9!KAB)8FYj!R$p3%)
z#n6t>ud&|=h<zg6Ha<D#qF0Y+^*{8>*ZbezKc(3UrNfqw&wTooXOCs=L&G|ldyZJ8
zQ@aEwQ@<+qSphL$a;R%>)XqWWX|K;Im@zw5aftC#huk_G>AL=`%cpi7x3ca&PV@bW
z`<3fB`(9{v;mc`zb6TFpw?M`^nLEcG>9CW)Zp$He&J*X)gIyU2{S8Qe2mcMwwEv^L
z>AC^suny+Vx4}BV?fMn*W}Kg$yRK<m8TWgn<J^7@vA>o>*yU+^MEs8qVZWz{btC+U
z^!or*#y(bz2hg(%6n=MYku?jmzWac9p)9+S=6rhkT=A08AEJsh4k&1ujvHZ@P1}{}
zcZ5viq;Xy=km{jS#y)n6+&ai<`mht-d+sk?$hy5r&y>|^`K|jOTGF@VLrGZ({}Nqa
z_AR@hZ_6e}vc5Bn_-Bj<7>D3j!npGB$~JpmQ@@?}U0=)>KyLiw(m6dnsC1kM!6D{X
z9b(+|in@6j?H&F=Al@5?cz>}%?woiooxa~#?FZuYuCS*=|M45qzJT=lT>GZ&4sk+2
zdL1h3_6_LXhpyky`%xMCC=m8_Ao`gOIXeZ;)#31h*RQzZo?ZtOhd3vO&L5DeT_Wo|
znLo}w=I%@F7B5YyF|umS0}80K=jrSx$Va>3#(m^dq~ijW@tiv3o*!|11^&{q;_D4A
zv~Iqma{Y!Yav7uBeMucJTh@Bi6ZtG1&bzGi%C04HTRMb(gLN0a(*~S5?y?<|4j#2R
zI~45rK*a0N_W?hZDdM~#9pXLc5YI;u@l|KliW4^%I$|Rp3j2XE@4#=3xGBWnVVnY@
z91uJLb@~AEaeW<f>tx62<lI!mnNfVU!`5zPemtBNKL$Q@sEe<Jy~8Vhcyn93Qfj8X
z;Qn(b_dRti3$fmIx-Nk|#w}iPeaeUe&GK0*`~Q04@a0L?qymd8R2`PjT65FjdWSAa
zw9s#0Z>Mn?h?7OTpyM^{_CTDwh4ZG+FQ6A;JR6hn%&_q<Ct3q)mh087`(_0_ALq|e
zKXCc&)f#0q&QLgi4&MO;KJme>LLXJwndLXY_rY+!73u-MHR7zXPQbbW>4@*jvKy4Y
zZ_MW<m)clo;diiOz-|wF1D$WNFA4Ji^mf>RFkT@J5`J3PTQHxa{-|G;UF3r=7oKi!
z+jvg!L=o?SBI*UyjguH}qKeQHb%^#)k<*uuPWd&JQ6JbNP0{W6Kz~QiCxzXRL&^&$
zLqDa6{^S>;9|7rnOJ&5T=n(HkhxdH*>w|R<A5}o~U!du}2;F*DxBr$qm!3<<c!_b)
zEmD5A?(5dF#<)aFhtP8<B0eZ<96z{a;j-gfZ1gkCvly?|zIocdX>wkR!wKJ%Ut8?$
z3`G&=q5|o2=wzM#;Wgi>-r#q6`usSrheO@GK(7b6Rq>Zr6mOeoQQSJNSg(Bz&MNq^
zIHc=&$lSPublo^h$1N(;{XBI3_bQ{FUXj`lq9RY4_IUB8b&FZ1i1jJ`o(H}MjdM<+
zZvmTqS>VRfo$^{0Qo8-MP^}^s+9ii>_XR>v4hp&Y`jxJ;Yf1ipf2YqcKDhB#mwxN&
z=XMoacqZ#RVmNmih;;>ry7S?1ZXt)5pD98wGDW}IuV2@p>*c7)-1`bT9dWL1k<-_8
zdMoT56fq9Zy>(>y8>}>?LtTGxn~(TuzuKqnI?$utqVD>-b&5_ef}Ki-yAxk6xx4Oh
z#UZ}GY>IRp0{aaR_8C*;zEei+GddY|G96N0IsF!PJsrXyqQfQM^-1XXY>HY^y~@8T
zd~rf;*wAF&vONnG>>scbBEA{@AN*{nR`<$%udG(M4(x%5Ck4V@2`n^c?7-SrtW`RE
zw$`p%`FbBxbiUE~;gAn|pAKOkq=-0vw}|KUin{fg+x%&xPF()*!-s9)r~B2aj-8L%
zK<>He`=BycA1*Eq?I>vE>cyqQFXI+<<84^-59a-JL)GP{6wWh)-3$ApF)tM9II&Kr
zp5G`s|G<7h*V**>I6DAM&$5$ertF_n^#>c{JN%D8*zr020&#}Gp!FoL*Qet;)(P-O
zxkbt+;;T5s_h0-%_(S|cw8wbkdA$D7Uj0gk9W`DEzhKmn`UTy}bUbq^yY-tmy-!f(
zb^RCmHivHO1-o!iNY|lU`=Ik4CsVtLU$QQ~oX)>g*4ee_Jj}_6yLF1X^Fp|F57M35
z)8QZYijL!B>;kpxFG#c9w)dcZhj^T*BJ~GTSr^|K_585gE8X^!pnj*DuaJ*8Ac}}f
zpon-N9n$ORWb}J(-Hr5W-(TCh|CFP4p5irH-!?ywMNt>0iglCUh;tM;<jx~N{3M+p
zXnYIZR}wavwi8p?bUfC@MS9J@bmqn>cU^nh#<~-BAhb)wfdifTqY*zxQD>iXn!oeF
z)B&j-OIowLx9-=iOdhN1;CeUYPtIk{e5UgB_r_&rofinZKlUeb<ItV?jx|2_=uvg-
zvyrC~>*Tk9ush)UyuIGKcSfTEClt;*KsvC;3+lwXkLI$#C+>&1Vh#}x%pr{n<K#g<
zj~zaG%@Lc%&6>6w#5;4Si<?2bEuDwC`XU|i)*QnA#G&c)ar`4)*PmfW<mPGJzREkR
zALv*3>ti;M_7m>;kdFQ0I@FDuG=5OGZqSVbI{vwJhTrMB^$Lxbrn2t$$j5qxBGp$R
z({&n^b@pf7x`_5SPDXtw>e`Ff_M!VdcYlnd9KsL8p{_sro$qxW>i6$){!h@zy`Qk8
zJJl~t%j5aFl-v4olb4HH6xXG9y6MH63t2kk;v1k>aELfz4!L=Xrqgi@GG~8>zr`(b
z=U>ru8h7ee)~!3R4&mYzXgc&G=+T_r4&$uXaR&1%)$7py@O+4WpmtQ$1Bi7dMd}}i
zT@!voAlfU|yO<Y&&}(q60eIpN{yyjfs3#EZQHLDQ;05|P5c6>7x+fp%bK`ECBI*Og
zxKOv-v5M~(-K$_<!1`8))c<#$ys5(JDqpX+slVN+jPpKqXzX`N=soC$8dGelFF{7T
z@QUD<BB#$FUAHe1`DjnEM*2LaGWs_~s=u4cx^;stpX*;VosJ`%OrHxf#u2{|{9yg2
z#d_N+ez9bO&FN!x>sFn1_0j1{hurVorh^ZRTW%5KpAPXpDC)))(|p9AU3JCkd-}fk
zt^%6+qv`oLkg>0TBI2ucsEg;~`Y+;YOp#uP(<5j)7l)#oC$O)AL*4v}d|m(1*->yE
zirl#lG#z$5ityK(>ZROwm7uriQ0E^&KKiXwG_~JTesprye&s**DeGkDd8U{)Bv+9h
z)&@H$<n~4A(sljXYg`ZZF;S%BCo50w(>V8v@ix~kW0w9$1-WrXm#&*vVy%Cf-Vgd@
zWbxI@N8Y)2vyF9I)R3E(a$361-WycD&JG>+eAD?dR^`}nKHv;a|K|LQG~MmGg8Lop
zYZS3ArwF|e_7xq{_&&(+`vP@(4wp~$Z^*DK>5yChA|3Xxpb+(q73$U@Uh`>tr83t0
z6sg@iR<hf;e*BJp-w|@LNrTJJFPmutah?;9wr6@S0{psMzePH?o-$phbLCKPr}~!8
zpJ+P1k2;)tT8p+T6^A%~ilT1#)}2#F^B;L<RG*6rZB;ttzF&M^JHBRfi+2)#Q{4VQ
z@UF98arxMv%OS>FzYzNl-6H+|l1|3Cm8N<YXAj|i|8}Wozj&kZUfUEApF)xIr|8mk
zaXiS^wezs$I~_-ymS482()3&3D`~x2r^~w!=e^Ln>~Ei~{BCKIb?Ly=XKvnhQl+Jr
zuYFyCUlsO8fJfM4peG>i3HEmQfuIjyyoFzvelO>~`+xd0<Hqk5_21C(-l@!;>*+Kd
z{xYYi8`rt~X&o!I%=dkQ^_Ry!Z+GdYsmkd(W#!sxnV(cEVo~gKL)+T_`l^%#d~3vs
zQOj>GVNs;tdE#W)C%JwDd!bjP<Ah%_S8t@BHxIr1)8a2jZnI6(xp5TnZ$TmI4}|^N
zwB9IZYWL@!GpKZ~-_UeCzbU$%-?86U@S1yWyYegLl?STV2Q8bn55LM`9@HT{XM@UA
z?`CD@`=`ZT9I^NF<%boAeGe>cl$>-}xkak~({p#J+_2J`np=kNRi>D*Xxf)EH=R>n
z$5E<}=w!MMr?S)e8hR6ly6Zvz(jj+!n(lUf<>KABbgn<q`h;EP&IyT1I_xG<#h~X0
zuh)nD))YB?n@j(3Tl%`oryf;0?6|Z|+I{JXS>FTRbN``1*R43KID9_k)?pvdm}*nR
zIV)}v=a8AA&d+U{&#m7OZyFSG;~dhVZ|e~2ZHn+O&~z$eJ+4E#4(DWD{0GN3O{epx
zPR9NUltX+jknW$L^9SOI;a9>smLm1D)8~bqgW6qm?TB8-RMy4saQQSY%T(s<eMqN#
zQ5o%<;!nTUEZMeAX{$q-ln2vxmCC}_g-6!v{a#TE2s<_6_P5t6+dpaEd>e@G5dcd+
z_IjV{i+-^&E+Kvn`Se^nocE?f8pjWR1jn~kH}6WdaivYy&+A*xYqVzbHk%^iuQ_}?
zWpa<-PUf?8*z=k)yZ_q1u*Ko7Nq-qN?u!h?A?L4q@zdG=%6HoyTZha0PObj=ump=k
z*r7SRZ{5t~3YYFv93sAwqUkwo7&kcNzH`L+9ca2Pt^@W}Al5JVemO;)CyDb7b;#+P
zoS%XAOOeCDfK$IV{Hj12zs||<yE^q>>*BqGViEq^pivj^7uI}g{|t+a_CUu=zhwBS
z{6d{wj?2e$3JL>G7{L09>WlE30<k|0i2VZ4cY)BGf%rZ3!(qK-iZo7|KBI2^8uff?
z7cKMN>2DuNO1Ir2*6pT<b&%iaRNn|od8hFJrsZinkC$xvy>py9Zi?Wc=kgnS9=z)t
z`@H9?*lE~Pcdf8#+!?ipK-T#uXur|fM_``;QhNopGf<h^_k#U2+`8Rudc}M0-?_5)
zZiRE_DRS%Gw4!BuSP$$_6iw}!^mmYP?mZCxJsRHz`xoUGGPkbAyhD-do2KIw_DSlH
z+H+vnq5fuSUp#MqyL(OJai_n{`ranp2ZrZDoHm`e=s5$_u2170FkkmhdFaWvmj9wC
zQad>1ou0!xCGRU!caI)x1F??^?VCf62gGv$X`E!mE85OjSZ$pG()vP%o<Wh`m$nng
zRR4j@*`-XskA-+P4ym8|$>aCFwrA>o1?0{vnO>vJ!RFic8OH<Z#ywp=c!~<@o+s#h
zQ@gs?deHHKlQCa9MQ*)`bj*`cMH<&)D&ssuQ^a`AAsxp}W$t}*>AL;;Tt2r>pzAuP
zGS^Q`?QDn-c8mMF4JvSJldY=m?dSFnT)aN(JKNBMu#Roh?&E^_yQSEeuTS)9Jz#8)
zRGT9Fa&+E-tUGTO`Sf|X?@ypUnCB@{{rZ<HZ`--M`g$7(eVjw^;udlKDR)jhmrnJ7
z>@NQ>uvfeF<qs<8^YmQcq-wK2&A<2me+T}H+VfJ!Zd2%|&`W{PV`0~UjQBjX59l|D
z&qaU3e1vrP>!IgC@5ObYAECcMuLn|lJ*OWc?hVg@bsNT4v}5Q?_>LCt54{=Rm&SeR
zc#HWD<>@+s`tRvFl+JVD8HoA<u|B1GIrzoCa_GN6)Pwr<sK12nC%}6~zoK!)7{^SJ
zdtWpi??;E=g`#PFP!9a)5bfM8q8*tc9k(IVbsN$tQeLQxb^}Cv2GV#`*vqi42GV{(
zWxDPI??B9pD2IMW5&1y0OCW8Z=tn@O=WSwKaEjD!!?k0KgLIwQ^`~XC(&pTtfRs;8
zrtN~$vuNCgPR2L_q~jQsX+02ULy^WCL56;#L+E?pgWB;;Wy}v@McS_2%FyrPg|KHq
z&xO4U`WWIn;7`SQJ;(=QT%hw6WNN2`9t6zy!>qNB);p|#3##N@vgxX$3P|^%Zppm4
zf7_}VS?AY7hFy{(^eSM7%EJaOJ$Fo*BI>~*jjy3{$)|ss*RO7pr9+JS(64}0kKx)2
z^eY`=e5U#bmAQ2$(oNTcRIlM==%YG>U7q6s>Dv00zi-UvC70UZgG0=VI;8VGCu6;b
zd7VT0KB&yiCp6tyKRWSM#nkKO*rxUgl!M+u5$%^ls&DIL?m1}xfDC<!@(X>1;`$4I
zN}YN8ZwiR_gnB_AeWcW@<r`+Eoae{Fyo>W);Wr0j9nR?=^!;)&orkfH9rHEp6sRx8
zc{~T=x6zL=?x5X3rtgimXR5zKrur5qqy2D*b(~|oWcU7Y&+fK$_H3+k-6Gv*Mr9fw
z?o?j!kFTyN|KtH%hxA+*Dx2zEu_~t>mw)U~<IU$*+wm%g`6as`9WUI<XdkdAc|~sj
zliU7Ce}{b>7#BID^Mg)CJUz~tfnC=tf?rd-=bK+2tb6#V0uCBKvHcqZ4lCfpy>lfM
ze&ehH)^1k1_T2PL1%y5fe<wZP2mX?*d2mc>zmY@N*nZ1+Ufq5Pr};WN0p??hruJ~m
zcTqu|{u-5htT&^EuzN)np+^Rd7!QI%Q+uzj9H*ai>C~?1R_65Spwc=0+-o}gO{Pfg
zNPfwWy^_E3x5qM+Qw*woy1g!^Pemm?s^<it-eI*@o7M~Gutg2AKg}z;?RP={&g(jh
z{vO|hU3A5*Bc_keV*!UIkLc6xRDy+l$@Dx@e76*Kl@a}3n!L4jrp@89o0~Q5Sm%ra
z;`_5e#6eR2bmJAZOF5O{H#0@*C)3H?{7=(ir*n$B`_p`GpC*@Hr0b$~lMWTJbm)|K
zS|0HeI;77N7MZ@!sK~GjJ4M=0{gzGjz^IgieO=Jm`Idh31^-J_5O&L`qSN^-tmWYk
z%c<!1_)f3q)(^q`&gpzc%Oh^UDVo~D>GQ{1rq3TWncD56CeyfKud=Q^IrgL9Q8c~j
z@kJ^uzvKG7r0<{lc*$*c&^#dCI98}@&$@iX<2XgcznCH&H+6P+8g~R4`yD7c^%uc!
z8x(SJSX}zFHbYC*z2!&S6uIkg>4@7lMO}O$&FA8^xOAub18qM#8S!nV=(IlQFI@jZ
zd>fGS1I@j4WceGcH02kbZhh}v3AOTA9CGz_o6g0HL?u1w_$jCDiuNm~GS~l+?)1Gl
zwKM4YCC&GHKGyl&={^rCJLSi(@>n0y{XU4>2h#Xb#AD<A0_k&M92~McsmJ|=*4RLL
zJ;;dH(Z%=C^E`ESYVLiTp6iF_Fh!^Fq_lik?Jha3Ub=oA_I#(~XspU}`k7mO!0YdV
z(#Pq!KCzO6dLF;78<luQx^E_GGB@u<CEawLAJ%eI@AN8r9p~_ST^v39<tPv2;u4Tf
z^_W=8h%=(+qGP`q@S#)X>kpZC$maBD`aN$>9`y6r;iK0av1uGvzRalwUl@ACrikw~
z?OI)Q<&4C`HsU3y-45r00IB_iTi4?piiRgP=lZ$921OC)N6_ohb5ByN7EhL&cu)a*
zHh5v!KZkBr9OAqgQ>5o5>g3c@Z5n?x>xkkI?UO^NdJTOaoQ(I$A^H(T@ah!lJ{p~@
zvrp>AP1un*%sPLj|Ly%#nyonR94F*cyQyC?XGg&~8Fb%Iy!3UhU3yKA$~>;CCyhTd
zmA&#8)bFwHH>wD`P}Gse*SeLtc7phFQ>6M3)f=d+vp4C^i{|!WanIv59r{pCL%Ofq
zsSG_Y*67zd+-tkR@3D?BMNS`$cRIco<`uyUhe5?5a`%gBI?liH3lE*!Iwg1A0#@tJ
zd+%zoKX2APjjj8-wX88N(JJ=DoR;lh$YuT3@Z-b@bC0Q3#a~)cyltXYX+y`$mn{Cj
z?{Ov+EIj;djoj9IP3G1A_?0AUY0A;Gsqf{sI$pM{^{6NES+||c)9TIUxvihpU-D^Y
zrMy;?61Q!9{pB<2*ZtQ&JZbS6^<kk-14;}(!lM0D-v(cNnBM|6%F}FTnMW_QIDF^B
zq*D|6=e3^d+ri#^^a2)(|C-d{Xv>p|!x^1x&b#`JT$T=X`#58j@6=yP%X8}x({!v?
z+#)?6z^!b$e)p>!=Lewa^xQC0nbTi!KAKy+@3VQOimlk7bcl2EgGR)Km}1cTG)(J(
za<EG|MVyDuq0{T3yi@xGcb<hV?iJ_uxkb*tqq7Uqe0olzuDw#e-TJjSJEkt3@{My+
zfI7X4jxRcydygFdNax-w_gzoj`G8*YF>YhrgPx3YRWNUW|0kEXZo1*YO^UV;Zhh0R
z(wdrEhVNCTxcQw5uf9;@B1?xj&&MfN>EF0klles~idSzPlzw633oH()p3B95JGCp*
z^3)#5`6G}{>jjysA5Ax1Uvl<rE*<sr8%^u0EjRsx<XWpX+BpA~BG<pYrgP_MMkO8R
z5&DhrgF-*Xyo+%i=iSor6XO>2WW<T-_KVZ`3jKy+c3)4sJ=ezcca%>~#(M+m;>xK1
z%&Uy^|DED_{os%9lTd!pK2yJ*a{R;(zbWY3u;(Ja7VAjx0KZ4tkX%K2SQ~6y7jc#F
z*U@z`^{?p07u;Wmh|6@0-#4qXwB9ls&kMx!c}2vD0uj%E{h;6zNZSkL3+JzzBJPL#
zQ{?#9JvXk)A$<-iyPbc*o9=uMr~L)xY5PIEuv65<>*~f4-8^BMPkDDMqdx<wK1gL<
z`$0b9qv`mI_v|;)`<Tjz$MK3<{qMC;N?rZx9@}d<zx}zUe#5Yrd$z;YZe@NvY=636
zt?JnM=y~`;iILxw`tG2u!*Xr9*U0@rrX4Hf>Z?oVz7GjI2=rKrrt!0|^K*#rK2SW@
z<&ImsH$I^_bUNM!Ro--bV9QPI(>V8v@ivWP3!CiJju^J`Ue~|z`aSk1#R~nJ7ufc$
z$-sOcHuO5L)8A`oh*`Ey56x-$F(Y1XBn<oq1Jpj~xBMTs`mcYfUEQk;zY7rSLBH|8
z9^(HC#a<k-_w(h4RdzvLJWN>g>3gAacAHPUuEW94mg?;)ReF7j3JP%_9fnQsigkS`
zKiG#Qc>TQY&mFRLJW&3e%CUOix_b33yP$8&CP(aeqpqG&%cuToQ#q*T<o1KbI~{R1
zQAK=TKHeBqyYcI~oPR7T>GKoc?lWQMK|59me>8_)`z_87!hTOvM4Sgj8lR|>b^0H@
z{(17S0+sr;TBh8dLyb7rSfS1?@0Z<G=dbXqKRdlnQ2GwedkhMLiZ8--sh<$`ZyM);
z_JRG;Xt&tsj_-@$d%8fByJl^@Nwa@mVI!^u@m$zni8u<n-v)6jhzCIXLjAaTo1Sxv
z_*AEe=LWic&xqRz3b}Jix#vY3oeuH-oTA@;HEv(3E`4~T$!UG(=dw6F(W~`<u{~04
z4t3A*<F@p5mrp&aIPAEzP1=3w|F?e&GWJW2ZJt_aPvLwPaNmhJJLY|r&!SjlX^9@M
z-kI0p@R#S?mzZ=eMFG)HX`Bg-M}yyd-n>a;8vlK^LY(=iOHZeDp0Zs{SheBAl<KKj
zi0`H1yF$Pp2KH*VzWhN2nIh(Q^jnHp|Izx<{W2R`)oOiaWx4|5yg?xS?i<~gaNaz=
z`<i=hyYefAaRukA0I5HT`dcxc;r{pz5|H}+Q6A`YKP~1x?B@fT#v#%1p2~E;IhA$$
z9xCr@UgE~}M-+!2CG;M2Lyajmhe`Q{eA%Mw&o+l#oE&%w8cp{p_+4&9gC&jjpDbqe
zd2?EFr;7?&-(U0KKaX8`p*8sZ%{M%FF1PjZ^xNxwTQ|X?y!X%CUt;hZN7aHVd6#Uu
z>Zr<J{)#3uR-aJM*4kAoU++U|$)R2~o?eosXuLDzp!tmI^`Q?qMX!DzzkiSKzj?)2
zzehSQJC)t~{jl#KXpA+VLEUfK>vIZb%uZDltBtOC<+E)w74WN{-@bZlwbRNJyWI9v
z&xu#>R}^*Ug6aI0X+_KSupZc<bcpY?>k#uZho3I0H@bPbOvNGAJsjTsd3v{%n{rzm
z7Va_exSEw<afmp34mo?%+BZ+zH%-oKnIhGjAan7GNXPy|4j*}ERG*6rZB-oNyabB)
zKBf+J=Zb8qGqYgL2}u@*RDY#1=3gD+ya!;j7IXf3ONpYESJbWlqL$C;hkmDf9WTAw
zPq}qHO~?A}&V0ujpL_JE(xGnrr1|^14JvSJldZ}W(eF*s>Hg*2iLaL2UH7;$#RDS~
z@98<@SJm!~VQo5$->1c%m%Y-iQN3SOPDiS5J)d&xu#aa<wSlnjpND^}x4CsVm(JO@
zxOA`mll~rd$f%=NJwH~z*Tp}$&8P2G=O?21rq56B6KmP)_!R5<()(c_U{sOL!%>r?
z;uq)Ysq+hjHQ#SNHD1@#`Pajq|L9MzOdi?&0?RAXam}ghb)A6U)BWb03_plp2s?6A
zk?yAsN_N^VP(D_1-dX2LWOS-Cb;w@ZR1XTO9G)M@#dqlTt;7F@>$pXn8=*t^>2&)v
zIC~$`5jUYj#8vo>y757`zY%sdw|vvzg?&D8T0h-;jaq(C<ECHNrEv$QGG~8?SGr$(
z(Cg6t7FC(rt(?k!J-2TB4?4fsTldarRNzEbTma%!aQ-yT#Y9{=>>#kuQNOn?u8p3b
z#I5W7;x{bcT?3CeUljIce0PcSL&Lv}xLw4{AZ`l&I>h5ZMm(mD$E0erKh3{)r-~Ko
z>f<#(R(>ke^)zRfpy@O|Pvpi=e$wYRoAQ9TW4}?Sult?vwViXnr{81aWSsZq6rIkK
z+&SvHbltjw<|B^XDf+!0a(X@PTwGnc={`Qc%5m+DOXvK$^js?{<2-4KxKGohPSut_
zQ{1{FrR|x*EemDgT?thxq-@P&oo;>aT?w`FSPT30tW#;$X*M34o9{5+a!BJgI2rYI
zi$VDpX?>jfBk22r%&lvk+WV%pUbObNM^`CR<m^32=hknC>y8(~-sBeD?l%j|@96j)
z71>np3wt^E1EPu*8gE?p>40=QYN*qD!k$mpS9Ctnt^alTbl#7O%=ytw(<zTS8T%&V
zjk<dJm0$AdpXT+en`8k8?yI!pq1<^bAij_06pzg8a8;G>k1LA0_;BQ#KA+!qVAK0L
zUEk@-)9;c~dDo=Bj2ibvhH{JaJEk}vm_t(@qE?RXe>RmTO?$j})4Ij1k$10~R%CvO
zENqhh;pAJY70CKtV81(SeD~In6AE@o=-XHi;5<3lS@8V~*e!L~^q*tD`DI;#rNb+(
zPZ?35Sw3s)xMIEbH8`tGQRi3W@=f<IzkJgjtIEHS$I>C};7-wLdvtp~zti8jolkV@
zDBbq~{EEXuKcVYBq;tr9mnzoj)GmxT0S-C;371aWovA$!b~RnRt!W$z{hk$->3gE}
z=49%Rf=s_dMrAth(e_IHppa=?iQDrDac=w*$JRab?cCkADRTZTtRqa{FUke&Z<Ol~
z+I!2y?Ka1Qu763s*|~_|DX6$Jx$f`re;9{xj)f^!>DRb*TDk9(4)I+Kr-*Zibm&&^
zIBz}o!Iuk9x3_I--=N>w(8-9eqzFBnL#O)Z&mV8DHT3>%who=zr_bXhGiCprsz2CV
zI{1g32XQ7kJh0?MzrI^`D9*o$bi`Ze5bfBhz8tUerh1msa^Qht$Hl!GEx5dx#bN8h
zBWv}3uc&qZLreOWd?<-UdR@O{zt6+)>v~j=^-Ffj8_t0;MflJBM#PVKMVza|A<o^w
z`B0{a^R7&huGb(VZiVwRBOU%zuLyrJ=ZB@~x_J@#V-lVjHvZ*Ai$i+OmQKdGVevwo
z&*c?i|J7l|d+y)4viELfiu9Z^oeaC3DVpjDZvDhg_nRVK(G<CJR=9IbkY4xO9fRgh
z-fsiZKZ8Qk=kdE7ofkNJ9<`78C1V@{zWzn+d8uQ!si>f-eVCS`=Nxh}{hdyx@6)Y}
zcwxT~^B0GhKRBf4HR<Fk>(zG`%vz_sBAo}_@=1T^x1JRA^>p_7?4FPJhxI3iIImQP
zQ(sf(e)#q`+Z3stfQuJ$J3pAl9iSY(8*Yji-*t%VJ4N^*bqF3gJ2;mPzkn&yxEGzw
zje9g5b`j1V%%z*o*ID%(+cn|Es;M^CCxZXY2mB>LA@wJ5GS^;dy3_p#@K;f!aTPzz
ze0<pM$1-dV={dnvp3%AHysO{HWtk!uKW&=+(5dqEhs--<n<Dl}afox$IIP{QbnUt6
znaV339*~iC^tHn_?4CgCm!z_89?I(PKX$0`=5woU$cxM0pIB~PDeJE0@4i>IdqL~w
zJ1W<2xFVND&zq#{s}!rnljSBJR1`-{J+i;_n3GC}^xSl;w=n-^q0irC_G{k*u9H>I
zKmO6Z2W~%~>H6C=E|=EJR8HyAbW)Yf3_H6ZH=af%-7kAHu4CE`oR&iz*f*<ZAG~7j
zImhQ8a(XTo-(s5XbUs9RZl4cL$NpRBYaG)0n%d>UQqHOUE^Ot4Ubk}l+H~pAw@v-y
zD5sk*Ozk;d_t$XWMd8v-?KE6JqUqc^$TS`CO_<MhXu7_letKiM-nNQgEZJb|*3ZaC
zyoqCbK+gwLCQMmp<Gdga>A4J$ssDIfhm>3`uQ;WE%ll5P{`s&3Yt2oA>m9lzF$*&b
zuKb{9^K*)#X?!o@Zs%QAdS%xVxh);ub}~<^H=F0SbU5OH*PmKZ<cQKC7jNR!FJgKg
zGRirfkKLBvwYuoa8HtDOs3Ki=#!7a}x7*)w=R3KLJK^Hu-KNJ%zsC5@;VqTN9e?Gc
zgLYU@_uUBYI|eizdH|gdb#l)JFAV$V(5=c8IeQ0}e(R*ZU*vl7w9=u@f6e7zX)XTO
z^PNs9Q>_2V{7(ioOR!9F_3$P89=!3G(jn*H(xqSax6f97w=~HzMW^F~PA?2wKGvmC
zN9bY2e@*IewB<>y9q??4b&IcBa7a;v{-i@{cY}Pm&55b$wGJsA>b@sp8i(UHKB-Xp
zi1`g4-=*9lo>zzTduM*hPWvU@4-#wHX}m5iZ@Ql1%F%r^PG#z!H<hV>Bx<tXb_=xI
zayLAaaJbnK<rK?|d?w}1<GYksg#Hq5G+vjFS~)1C?N4@C%j3M<s3NDoM=hOOmqjH#
zYJ8)<Ue}vkoDb&@M!I9W+xN}tEUmZ9INkz&5%^bhi1MZg`?FIt)z=YM%f$`QbgzC#
z+)rosF}0Hv9N)R+bAvN%rwBj34&Am}E`FD$rxqFAzHEo>wkcA-yiV5nyOGcJ6J2|T
zf6Y|iGi~=M$JxDg>D)egn(kE3r*?Intn*iJ`KEr1tbIv_{sQQiu|o9s?1ng}*>B|f
z8|E_|()Jy+9CSPKdwrE%uZQs)@evfcxF(wZPQ4j>eymZ{0{&dBf2G<p3R-^SvYVSV
z?O5lG8Z>@l`!@z0X7Q6Rn@`%2H<zWujEPUy_-4ih7Kbe#pZWAF&mL1w5$(It<ivHg
z%apJdTzbdolf_F~Uw`-g=`TMnU_E#JWd%1K%4OlZs@UJW{?3<fX@C6@wd3$R<$G1R
zz&iQdx)Dn{>{PJp(fxM!t(%!#;nIByh`2%yKgu=x+h%u`v`{~a@B^D7&a>AcjhBEu
zH7J}puTaJD>qgja<9n%JhLb6OR5sm*>el|i`D^@6Kh~_l-Ya$&vnX==L_1U-HgM^=
zV@ii~yzyJ6_oweCEb<Qnd$n6%{-6R5XkYU49^V{L6lwgfPNw@0{FY7kQ<#?X>wW0<
z*}2X4tA4uc(DSkUmfenvroRh|U!K+jywm;J&}XQh7v~~EA4a?<>?+XXu^z)Z-IOO?
zInEEq%@^1o)wBG4V?Hmr)JFUg_65L?L+vP7j{`BE)BOmDTLm9FG}R-%Y&hewhuR;s
zO%d<WEt<v|pd8Nefu3cGRPW(r-F{fF`Ezd_S^fqqO_`!wKM>j(MeNIWieCM?S$2;4
zUB(P=yVIt!>A4VY<14uHn@rPn{(j_F8(s6tXWM2fr{@vk`(LK`;_&C5eS6qxMRCvl
zhX!4@;;f>`y>DH=rTf;X?9~27%fn7+iipeAq0YWY^PTRK<;owpqiA~5<BOCjV!q-K
z^Mq409q(y5ya!#oaGJ0Cu8z}u!~;>J^};-=LtR`S%`bCFyPos6Y&!4UIn4VU((CAC
zoqZbXYo|!{C+JZ+)a?tW`CLDE<egD{E-tiHQPlZEkdOWH*w<)^NqH7LePZ+)r9<u<
z5t<HrkyC`erhET1pW~BD*V)THEYxX0iQz|-Q~X=f+<ga%?685l@rfHRX}a$GLFChZ
z&dH{JEtKnfU}>Y|q{GS-@m)xY35%wEIdjuFWr`SgImG(L6uEue98XB+=Aod{abF#B
z@yhW^r{e%8d)@yPw%_Z<&9LRuei^h(^;=G+@t>RwdkB!zE4Xx>zmewC@mMG8?8LCc
z0I9y|w~YJvg)_%pwqw%4quPCZbUgtXdZH=1<wG}q`kl|&y<(lt&6lR>I{i9U`SkrD
zZprQbF#OKco)GkYa=NeIscfo$P<vNUGU7b;Kc1ZH)k`k7bbb|6KZo1;V!YuH_jk&>
z*Yecv)NAL0Yu<V9oZ=96exUtS-v(cNnBQ`X)xN*Bb^j?xZ4PyDmiVqM#j<~Wy+GeO
zi576$>vIZb%uZDlb^DB*=5LsO^>^jE?y^ne(CPXbGRB+Mo%i0=Vt-!CDeB@rXg=0g
zrf7P;0{knc=k}nS&VI<{bLU=j={h@qyElfl=`enuGClVi<(%#-;o?xx&nTMqd+aZx
zei)s9TW6O*ewQ-6>x|g6QE^zN`%7b9oV-@)5b-@g*y|}e<&`T>+XKFfphMI1dfzxz
zVna%qlZr$1bEn9yW9hjJruH_j+@PPw4j;Ydh>h>I!{6BA&bl`=zUh!XqTfrCx3<pA
z@)IE)`uOuHw+{Pw##DP+o1vxZ-tyyls9Wdh_KQG&;*eW!A>B02UUz=p+<ATHKU8_0
zO%eWN9dhFs(m6d@cU?};;NAyK$NS`vTQ?&e?M8>#ze3UUoZsm+${cLIZJ$k%`t2d>
z+7<G3dWX(`sEdPzpN{Uw)Y<ouuUkLkIXIl3_;#NOI}h3vX&eYA<GQACdnjkR?}V1a
ze8VB)KDqIkiw_w(tU#rHt(GZM<m`Q>c3=1lsGpq1*-(G`OH*o$tXlJcf?nu1!r$Q)
z{jQhk^)OyW9nrr9|N9)`Sn)x}<EY9`&ld<wd8!8lEpy`w_0LiJ8@F$t`+k$@dYRfM
zAe*ieu-@hn`&h8f20}095OIAv#Puk0*X7n9x^&L&%gtY0{Dsah9_xIa-spEe*Z#ex
zL%%^BEY^F7KY>j9%SRLErcGZq-3G$Wf%Oj3fz&@n=YMXW=e*7Xdk%i`Cq?~Zi01{u
zE{pbv_?EjrPw%#JQ*Numi9Ox^b#S{X_~zg0=Sltl-}9+)MeCBU|7z=yTNi1^VZ-a=
z{`fvMhi>OL>}#fob~EIWOIB8{yj{7)UnaD;>+6Q;iXz$*hloEhMSKsB!`~WyoH$|b
zG36K1dBs!)ufrQnPU|~Amt~5kalxkVNoO~t`P9w~S=-Jpta9ZAx1`x9$F)=4z5trf
z%>%l0tS>pF?Uc$Gzc|#L3ykrULtQ)4U7zbObYBA}Cok`8kI0{{IK+6%q4vDB{e@oX
za`jTh>0L<Y-Vfq4IK+My`rQWI`D(~#?|Z0Mji;BSW$jDl?6*iqKj%=Vf6;u-E>F{`
ze?cd6?SM<?c-N)t&e23Z_Mey{ea||1<};P2zc(&ZQCv_Z?~+Yd9aTEy&Q+%VS59vA
zeT9QR<~^%`^m`;!-ktbr$=!92D;?_M4|LD<myK6nb4&UeMKOQ*E1Jw$eL|Tc{InGN
z-C5(iw}zZh98y0HWYc{)M`m`os>=7r6-Ca@NYk;OghQ?$$0}X-UE`%GN7JUhm)kN$
z_(OqooyW<sdS2Xz+K;oF%*~@=OUHW8D>_~GB0ekX$gSJFrlX&jqV9fJ&vU3dPg<AH
z#*zQw7lD6-BDYSY>83nz<#hf)E}y%e({x?Dk<<Qz@|>N*Y5ihVKIn6>bbdZ=of@lj
z#GjcW=jZ0^?z(Y+=5yEK(sh1+dVSsdp!sOmrpU!#aOtoUa!Bo+uy2^6Q+u?IPrvi&
z`w5%Oy_cw^`~BYN^}LSPx_Ilb=fh9t6}|Ep@89d<-@=kl{WwvRb@5!;HQ#CbHq~qB
zbxqF^iB~yJ&-OcAXAeTY(|w{;uhGd)_o31Encd{5jaTQ@_x5x92QFT(xP1(vPd~V2
z;j-gf?DL*0;^~NApy;*##3~*yR`;RnM3E;yUGyJ2?4WosUGIjyoLjpN_S3`*!{#U8
z>cf4P#BI9QagX~w^$Sy(?myt-IFOG0a@_Yav7f>%(l`N~O!t>Trsp*3u7muBCpPE$
zxxt3>_TwVnl|vd|WGXvd$8+acBi;1ATsclZMjVS*<l=vkj`sjWx+!w@8?JvK9s6T-
zi1<q#>edllJ!v|3KG%8aBNOlGIpkM$fI;e4(#fXZp~ZN>A$L9v(w)xZ*q^3DE)E;%
zy7^IOho|QUQkkApN@d;pnO={|*w3g#wC}LsvmLf}EA!)FJ7^60d84)-bIONld;<MW
zSay?j-wAV?f8c``Zn@y964qb)TSc2~Eo2@0Y~-oLI{B?i8#-RTWO1T}?`dG)CB|tw
zpWJu<PoHMo_`QO@gMC1BUV?tt@v>#DM?I0xqB!xhHdC*Cp@>D1wimbOSJQZ7$TY5Y
z=-$(V#vk2$-gyjke-WO?6lwfnQ1XD9<$AU2zF7fvanJZp1cx{uBWUFA&*}F_NBe+G
zk?P-0W%xav+F5os$o1a*ITu+vM8Dz?`lnk2KYk<CBcdjo`bkg@{OQg=;o`e#I@UK>
z&pO5QPOt8Ib;mwsiWrwok?W6KIz3ka>p%{5=MBT&p<Cx+pBu*?_8)M_truuIcm9Dc
zK40f2(*C~M>`(LW-Kntuf+EK+(y>2~qG>z8Jjo$_?~v*K2~LLIKoR{7$i<KA_F>R`
z-99FoubV&V`AbyBegqxr&abBY=*}TSztEwseNw%L%ADUl_xgr6ENZ<?VP7}qUG!@p
z+9BpYd~Xc)Eu24w?^n@%3pd`K@#NF34k^U7!rlje8tfA|zZCISh!aDc8uo(%ss4)V
z!2gYX)R-TDSg+7^G8fN>?*`EFs4o!jk=j*sc01@V9KxQ3>(l+hI$39@LjMJV?^q-5
zZ;JHXpQy>)xnErRWq<o@<#$Vytlw@;UAO4tQdX_^-=06DO@1ri53|-jTJNyJItlYA
z_VcZ|X>h$mmn2%)S9X2oxUu&BQwsYj@!d(ZFKQ=3`=b7eU6cMYYTOqY3W$AHIM*K6
z#XODc<2s1b0K&e9bD4m!pWu8i^nd)0&d&u3zq_``nuRt+r|lW@B+zsniE?ybIb@7a
z6ro3%w*R2ZVf^Qi*57YgHy<N^VZWYrD$P2rbja<GFZuLO^ZM0IvMyS<f70LQ=4SD$
zpWnWEYqisg!ymV$ue*HeQAM%G3+lwXkLI#;c){GIsi&(YSR8WSe}p}RL*2fA?tV1=
z;oiBD3cqnyQN(z{A*UCbrmyV(>xsjcCt2}EuD{W^V=BAdx5wR&OUL<ug5P~`uG#^c
z^1$&%(=p!w9iNwfEIqAH`&G8?dr-*d{5#xvLtHxaTMjw>mZn4hHbtx-IMn&iO!GOt
zh^BMT&!w9_r_=AU(sO*U9`TB_z3%>7?*=E|JYlyS_1UNIEIDrD90rQ{GN%@NVdxQ?
z!}IK?ZRUKk$~tKG`~%Tm-6Hfhidg3X>GSy|Pkia7jU`&`vpGb6*CF&;in?<MO!K+(
zN_FY<oET2##t)=({n%+b9e+6)-#O5sE{;xT_w}03)t}lm{c5jnuXCQABBN8CsYCYK
zu}&Y-yG7djt1@hyzs$u6?pj@R<&4C`whl2q=#ct3bmevVLGQ=K_3^%S=ba!Q>r5Tu
zypIl*hYehM?wI0`p0AJkgax-;e*L`e&mFSeqSNP~c7>?P)E*sG*;GH&m2>+3{VorC
zmfr~baCSq^9xmyPpZuiHZ?<1NI9=Cp<>~xQ=V7NZ^`lUk>fffaX}^ZvBKZ6FQnDKl
zrtQFKIqIJlxlqON>qgjq^WgNopnPhP(e2B2*ls(;`{!JLy5*q@Ew@PHJ2-jI{f7o!
zx8kgFi`@61Yd0%hdv1EBGDXwz)HL3c#wBnv;#qZQs?YxNeESlU&ZQ_0d%bn<j79}c
zC{yJ2JD`2}jfk6|$gM}<hvg9GOH$;{J@cB5^Ts)Z-H9Un#iq#VCAv5q(|r2fOvHWY
z5dL!xx$j&dojY&YX*z4y{C3Bnxs&(XutQMf{1aTd-|HLG>v88lo2G+r4$-f{k5hft
zw7lteXt{Ezr{CE1ibEB*oxRA4I_l0D3~N5;zlut_Zl9cKJ~!@!Jst6hLE|e!7JoM2
zmqL~fP4ykR4_qhf^hO#V?N+Ax3nxEYYges&y$`7|A8h<!$Mahh;$)ydB90W_wSzu@
z@fC4tK=>~)@8i5^*x%?n2YNo;mqqvQQW^1iK&Rsp>Sc=bd{@ND>d<Y!qw|wf|6;7l
zo1Xhj<KJ|8EB2Q-MY>M~GF>0|Ez{?N%zYPtrqlLKW%`|C+CHcpm3dEBPw10Qk@DeG
z*7-}k<~yyAuDos^KjN0@I#3tC#jcAq=VsCIl9TE9h`41S7pK9cV}5drnEyQE?rZM3
z?aHqdjenpr#wUuLJpk#LXWgPM-kYxH-O5(!!lSNzbE{1e=SOhJeb)`?bbO>T^bH+y
z?MOF&a{GpmPVF^em389=@@br|X}bd-u}03Gq#Z|Z9a;VcD@|eF7Kc<{jhC!@o}8Ay
zVQK2lSp`qqnD03JbiZ2FvGY;eFVw{!n&xA@<rMLqEFktd`Gwdo#bHiqXPo_$OV{aF
z$d46&^qf$)a;)kr_Ynq!0bzhL0Q+OmNbf81|7O7Ld=j?b(S1`+<^Ro6LIM|`5W4jG
zUB(P=yVHh$maeloneOk3w@mF!zR4eax$ty*+va#cI_$wZg#NBWoOh=~zt26xIVW74
z57O}+aEh=4n8qET9Cv<?({`ov_uzgygdLB5|Cag<>A7drj)rqy*zX_qx%J^ocW+Yo
z9xIUUTfctylEROVzR<$Ham1Zb{~@*8;Ji7UkAZV;DAG6+_yd7>4r+I$b`kUkAojJv
zkB93O>iBY-7arJT!wyW5v+L3C{6fZkIK;UdurC6sy&N*0OXvUdE8lb<uGezit8Xja
zw8S|De=*(%?10qIO5Y3J2dMMgn0{Z1?!$)6#YNEhms_86>3GjhQKye`_cKkeF{tvq
z)Az2lbqKqHQ$#=3A=(EJ`{U?7UYu`({sP3g)j%%p1LvPnesDf6ZC~tsa7WSfrpFg4
z+-F1e;tx0a$8sBp?_|OsLGOq0p3V!f(_y@zeh&C+xN~BVPW@xhw<zMgt+Q&yiJJ=@
zv4NC_sL72V?!4#JLkCr?5dMtEUddni+hZBZDT3cv?;G}dKey{FI-Yu!!CzF7n|GY1
zL+{d|E>0q-eBC*xZu6<0=v8*x9=-M_zkctw{`7Zl`!{}fwE4At=X{>a0{*)H`iCbi
zKBIuxACLBm`45Qo9fx#&rt1+JPe$V+sD8!Whf7DiA&{<XPMls)^H{@;Hju6txpt~M
zX8`%!`UB~1<4)*0lauMWH&oWyC(b(uXKdGm7ptb)G#%$Ix<&AX_KkQ_4(a_N)A!5C
zv|VyC^h1hX?MYPsgG|p$qcZnANC$sF-Fd5sYvcNlE*<rzNcnOqbL(it9Ru;)Zqe&|
z)cv07^E%n7-I?+Zxo_D8eOop;VtYmIoEk3OseXrb6h)n0Woq}pxDXir*rCRo&#ktp
zolh74MeQh5rq^>TQ~L~+(O%r5X}{C)LG!Wh)gkOcPW55@jzf%}*&VrYmP>ca2hKm%
zA=SHdGN%^=o$hoVi%NN?{U_GtsU8%x%=w?fmhN=lb7;$V9@um6lRqigM@><;ehE4s
z>*%N=_Wi{goz5d+Est~TqmHnTWEb>$PDi|d&wXb&Z0Ycmxy7LMwOC#Mj*Wc^e1A=b
z9W~_kOGGUl_5eCh&~<fGWv_Z0<pr|oxrx}%!mSIrbkpl|`|`q;4nGKxp2Nt=Iz3bu
z|7V&{<1ujF1Bc@}q~vOO#VN%hzHg?(<$b4C|9n`2<reWBG^ePGcSJmiDb`88yU<%L
z^H@%i`<{+29p8=9p^ks#Bd*OYB0kU*x&7i8UpO2$cv;OZkDs<<ja+@aro$f*RXowF
z^?<QGQthZ>*yahR^)a=hhqW9%=O?N%^+QBO*7+fkuk+`c=Ih4GsE&iQ-dsE|o}bg3
zOwX(R*X?h$nfz8k%PEe}GrHBTi!QKq*y7>_=j!Y^sGK4<ultog@3PV>yOzjp>5%)5
z9+!S+zGIEgJ$h7`BHm+EkZX^o>6{-N`!jTib&To#!d4FPE>T0j`dz%Q6IJ^Q>P_c|
zu**UFNn>?=?AHhjA}*y|o9;Doe~@W&`_pJT_OUp{={3q6Y`$%uZHlJ;d9UAJ)Wx$Q
zANr@?=(oKc^~3%*pwo31)hDoTC#;D5F+n5c#jD(<Oz%1)Hf>aXBNrEjbod)Mq;_vz
z91z_{rjxn2Uax$*{XO);sA9bR-fnrv?{Z&v>*(#rQf;ru`Eh8v)A<YKO~-4e@u*xp
zGSabc$Em(zI_^8&cT3Boex_&|XN~9AA>xARdv+>AAL7t!JUaJ#`h3ME4K6>wY^LoL
zx$i;Kbf>)N+CArQ*ZmIpPWubWbMqHXce*bD<x{#eom3?=!}bd?-g!m(KBFR^8+PmC
zFTH=#jurZCzoF+zL{%Qq@1@CGTW8v_La*)Juir=Yefd>yzj&1E2m}9%0jeK5l{tNa
zrqepOwTrurm!<covZ+53<q#jM``!qbZ{=Fi|Lbjs6uuXN^N$gKH{l<PM{F*;*5>9#
zTK}9PN5xOh)iYjp<pINHJ$C7n3D$GhUsiC_p<I@!pH}xh2WmfpUDFgz*O$8WYEEhY
zgWunL!-MB?TR9cu{hmi%dOEH1l<mqd#5@u&yzjGlrHZZCp!`PNzIvzmSmy_g^c)zc
zGW<{+((?zh<aQqy%-=1=#<_JIQa>8{L%jX%Uhn_No<VzWnYi7K3gZ1z`xTYp#|aAY
z{-c82cW-p*I3JjzQ#(6X-fw>-#tocHNs*pU<y6MGPC)LvQr!27bn#cXE{E_7a{flA
z`;@V+;xKERyz#TO$IEZGsZ8xIPUB_pzBuIK*|_uu$I{dKv|nXAMZ|kh47z{o+Cy0L
zo$liaYx$_gv7_ENr!T6=*;%5J&W)Q<NvHFpu}sci#yQMDdfv3(<0#?{1Ro!^`>_lg
zJiJr=52lAcF|>WBzb2PTE}fkC+R&y$lXLz1Z>PeYighZPobbxfPNn}XUzwbkoOt)C
zj;{{wG^T&uwJNvY?}@$MvAOos{YLFnYypQEMH5F~IN;gCDpvTfEic^JaOnY+-SEMO
zK04F&xjiaqbohFGFRpuUzTYs$7yK?|@bca}k0?{zGPv_6U$)J(z2e!vb#D3gt^>AH
ztUopRnP1lLvvoLR<NaT5=(5+=p;Nkb?IG;>rtcZ$O!-+ld)w1%_MT9&!s~}u`lRr}
zLn>Zq%8S=>cP_s+*L9g+*lyA5{jX_r{p#YUKD6V7sQ>WNSN89>eUfsDJGbns_-*%J
zZ9#85;5~#Lf4*p7-{ueOR}{luZ>gU@41D>?)vY@1nx&kg>D^x*w7S-^mZy{{2EEQ`
z_g>;Bb06QTIK*?++O_P9H>>VYZZRl5AN3RTmw|ZcMPBbaW$^S`&lcFNbm%Wz@w=K^
zy((Tcb~~P-zCC{Wb$O-Nj@e%EmS0DI`SOKZ?4a<u8rutWxpBVDA)Y6@p;P^umJb^r
z-20RC|J}ZW=7(!1BM!X!#jZC~RM5!X-!%QFzwfB}@ja<ZhxB@;_93_9tk>Vs`Z|@H
zSD9SzoeK{sujur?D4(lxM!j1;Ibu7-2F=d)yR+6|+burQ?xjc1T)NX1{GA7#eKsoj
zPVLiSEAO^F)8DP1T;^P>RtIezq8|$S%RuSJj&zx}b(``V{jH^3Ehq-u+HLSVul1t#
z)Tqgq*8iYt!t@o&D{?#rt*`oh{f+q^K9qiXw;eT{U2;g<HN`XR?219zGjLtnesy*>
znm_Q+<U3mI{Z54)XN>#p=GU%UsKSPn|9Hu*jx~D!&f=$4P#CZ0nlWh5C;6tP+fMQG
zt-t=f@PikXSB(1lkLvxgmSf#+l7>H8YgN}ZDqhI>?ZTeky-w5lbCcIAim`rwazAGb
z^eeeO<2^e~{Xh2ZJ^r$>{2%{}8B9nTXC;+NR1QVNH+xG-Db##BC_WBRluA<aaY{tx
zkYoBvlxiFbh4G<izdy(^A#yIljMIo2%ot<F@7Lq=e(t{Ct5)l<*V=onnf=fE^}bKn
zecji6-Pd*9_qx~G3p*xB?X>0JzUHNSjB7U)C*H8xj+=aAUMnvq$zxFYO~0>|H~vVf
zKPdSR`yEQ43<>mlAEYkrouf4c%RVO_6un;0>$ERJmxnZ(BY^|YI`r5te{5E(X?Vbu
z>rGo_(u`Kq@yFX;a@dc4@KkGI$EsHh`<550=4(*&y)RmKeDAa~S6}Uq4|Wzb{LaKF
zJFWGp2Rc<@l3smkvwc7Gma}eb4H}ESr<Ju`+I_RK%WvECiXU$9%>A9bm^4pi=NG*{
zU6*!A`~IfoH|0K8<g@B~u0#6S?ip`+>wd3$`FS^Vnu>YPEB8D2{O`Q&HRIbu%qNGk
z{YO@M@gBs2n$9YZy44%h`BcB<Q1^Q`N&F3Jf139_VD+n-*IwVxi0zB=S@iy9eBR&G
zblLYQ>qhV1sT*!FdCfn!l91k1^<H<?bmM!6bx$|_UKsw<-v6j;`H2TFdBulrn9=DK
zt(zhKj|A$z?yBN1Y5t$sW{(wi*kDd;i2tt~Ca=EzT_5`C^Q}Q+(!5obU$q~snttB6
zy>9i>;<Bpdy(+pV?YhIB=iD@_lNGDtBQM=S?-TO!TXy}Tb}#Z#^?io2p4V6X{55S|
z7cGzSu3OsqM8vA%oPN<wIyYKa^9QYWRlPTtcU)P&mr|E{S?{anl|#Skdr!1!IOzDI
ztp0m_|4{pW=}YX#ccYtzWyL*F9$C}Fdji?+2{ctsS@j@my}ZwF^1knv^&TC2Qxwax
zo4wL0x?f0YmvtV?s@Jc3By)~Be1pYLAKUH~%NqB>N~bRIZc#pZ{XXEJ+n0UMw%_!U
zo(nfEf8~w7KeEeX(_2l&vf|D`(dqRbTwUz<8kg@oZDhvNodpf!cToE!@2uw_!+L}L
z9iyTA2P^*#5BTb&-LIbAstZp4?23DiU+<Y#)z~jP6#L1#uZ{I0oqng?zv>ms`p%W{
z_xhY5@AIXy`c30oKXt*nd{??I@+itj(s@l)`9=B6+U}=UFFoy=ard{Hf=&5;N*bSo
zK7YunA4&C^{(Xtaf8O)(B)K#-KfV`I_w>YRPhRxEfzP#u`2WgbQ2H15fo1a%-@6{B
zv;B_sP)`3|PQC7j2Hn2bb+({=KgM|WI_$iaHl6vqPF5txem_UiU*7!#e(Qo+zdw_;
zUR}o3FFv-PIB&)iOV4gK4b$$E_e*~IY3lbsBH#V~z7Y8hv7QpB+xcDE`9@NCS?5gk
zoTYzo)igVv#BWpg*G1o#D%$QJGe7)qD1q`M(69Sc<#GH!l#=+)d;dD`cOzr{vd2}F
zUe<kRUF&7tx0O{dNq_2^UsNy3ZrAU9L+r2D@g)5&W3TU%HSK$-k>5`paqHRVZGBg#
zF4*fl_p5!BV_Ldje8qn1YS+^ED$CwSI#E7F(>JYs={nzk^Szz2Fz<I->sG(tzsFcs
zfA<~x*|VqY^lYc;m?n>+dS0|$+4t3p@=<lWem$qlJC3sMQx54=^E<3*&*k#Y_q<C_
znYQd^Q#(a5Z+}g#pC<RFlvn*bybJ61ZQ{7I>iMsqe{hL)SDMx77iQ&aSnpp;px5^U
zliL65-qLfkemAc!^?H47m@bjQ*2wza(_p#$=icM}<9?_AnWJIOzmq_G54JApFFtzM
z>c9WS%udrV?{~ES-P{baEUW$LwO-X@%=`Ysu>O&Uer&tHwR)99+Wont@~ZVAjec3*
zjZW*Q-{p1t{&Ze`s@`uV(eHJh>iWA6k*}ir#$M%<wf%-WUpsH)`WIS7F|D5uY_#n<
zr*HaPr&s*aA8x<#bw|wZ3<~f2%4L(5-tGBLQ?Y2g7xg)7I!!OCE>GIW%h&%h$Gvu?
znHSyM$%<*uW0dEOSCi_W{J!&77<b#;PT3gmVdPCGeSeW$%f_<q0S}r^UOC75Wy^J8
zp%?F$C%vzn6hF%!)OXLy>bI=te(YRU9KY1&Gfy~oPNytPy7!fqzuE?0UT5afk9Hy!
z^^aw@t9m{5`W@A>+aGfDpEmyeC+D`(FzJ1RrsT)>-KBe-@8j?N_>_A-H@%Y=d+j&2
zkA5m^I?`!sdi&X1pSs1eFSKA&$1y0rXRQC5kty$(-&$CZp5(=<=jXEfHkk6QSI+HZ
z#RopI^!!in@nWZHT;bx2_ulQ`u_OJ)ov&H`6W1-b=t$FW&9TpKcGl`+N1Be?9Pr2Q
z9KP40BMS=lx_<Ni?!5l)o$%?I(_VP2Qx)co>(@Nvx?@+K^<1YaH13GKK7Z+V`=Y;V
z7sowJH%kJqTK%QB9rf^-c2-Qgx6|wSds_RnbAY7t&8qU#_-R^sUEbp=%17F{Ot0l-
zzn9Q-dR48fe$gG&`s5|j>}gVY(dWHc?|;VknbO`*Tu{r+Q(e*dUfGy7-n{oeno>XM
zd1GDktNt!SnjU2R{k{0R18M2Be*0CP_C5W&muEd6Z%Vzo%R7m`yuSyLw|?I9nzHH_
z)tj{S>wBqLv2OM`ZG2^w$L|Ua(^vd(gJ<sVq{$&k{$<B02W&Lqw0n-fwbOJQ)IO=K
zeyckF?HAp$@8uPxSJd9e@5nYao!2kdrT&z&*V=yfM=xoWjY)Pi=DQC&XpDU2O*d6<
zo6dJx-$@!2o$pMXveR0hdZ5!U%=<md1-*XL<F9^wCn#$_dF^Id_1%xI3+6p{E&80j
zs&;<oV~7v0J$c)QT7!~9zwdjL)nC^0-=unF*@Z|a-n(x~x~h9qS?NYO#&p@|BURC>
z+xIQg_$m4vr|3P2IIm@;2W3C=_G`Rp`{DTg*O=~CT$!dny_S>55Em>7RApb2=vFPg
zUiCbwebs!I^}H=FouqLM`gfko>ObpwVv-!n&QFSShxvEUIHfbSl~)e&ed?;yW$Qz)
z^opKW=8a48{*Fc3^VMF<^N!m+fwX%?buW+c<-(FqS|=NS;GHjj{F*T%X_)mr!ePCC
z|K8<VN4#Q_y`P`gDGT$yFWzhYvhE91MJMX{UYmY+<g%^qZpH6L$8=J>9rOGB94FeR
z$WL9<gL*C>=f7w=uYVtO{k;A>Z++j@`pw?=P1^aQ$2)m3>w6LEEk5JKQ>XuGey3@;
z&8lbIHTzxjI#px*UQJOwPimKS|5(@Wwp8^zuWGt!=cP?4&ud=>UBBvkBT4IQm_Jwv
z#P{%$(sh0RZm`npz4x-7llPm>${T%uWS7ULx9WywWXl`o+_>-Uk94wP*5{nPzTaI|
zd*dA9Py%@*khCu?dhVU~xp>j~d#u0bZ=O0}d>i)q{l?#)w*G6^oiV>PZL`fkebw`0
z+jYllZaenIgC2RNm4??|`oO+-OrO(=nDseU)%A+b|Df9q`g_Y&_aD#M%1X0MRh@U%
zm2T1RXH?a$tlt|*+8<PvUzG2pdl+f?Wy?S7?+nznUQ_P1Ch^(p^Pal2_uaCr-zQA^
zj%mNX2esudxBA-6cR$hT7uIE-()e(`-&8EyKK7elU4Fl?C?8qx+1IsRzusq0>gTrA
zm)r7~=caaGe4c&e==T&NCfzqmd+&A7@oC=sC}qWQ;{GB2YYBYv;7OB~-*;LI#&cBn
zHj>V3Vt&-)q;%4Ao+SOQYkqtlQCB`AUrBLl-usbtt3U73Q>HDu+0;(cG3onSdGq6U
zOv?J*(4>1dY0t&ueXl{K%YM$C^t+r@=ck<`^je-Y&t>IjeGk9vdjHxRrstkH;k-l7
zIqSt%U9qfvZqa=~S?$tiqMRCK>s7z$l{MdKbT0Vm+({3%o@y1vyt7sGJ>fyOE6Xkp
znof)to0?AAXAC-D{2kBf((LE%2dsU?Pgi-clZ403UTn!jKl5;>sp$QtsxYq|D!ab>
z+e7@H--#|;zUURb-xTk2SCuXrmk-J>WWAS`w_aKMu%h!l=yqkF%hLY^1=Hf`s-G`Z
zCC93te@_4GihGV<@0nIpu&i^bLDNb49rU{AfB(tDFM9O5x3!vz{d(U$$sQE-=dqoA
z0c_g7D?SgYI-U01vuJr$`jnTSqWhpEJsnj3eg~ibowvPae7h*#HtWUj9rDpd+JnNn
z-^)(pZ^=8a8~^^z$F`e_8y@i0NxNS?yVW#2W78$pnREKHtzL0Z>o%?5@h9)O@vWb~
zqg57qzIpG|4Y!!Q=AT<dG2Tlknob*kEbsMu%};Ez#|k@aFsId29CZ9$R{vGM=TJ5M
zyx-5OOZ}pDB(B4<(?#vjuw8F<JV}04->c{s{if}=>-wFetb7gXzERozR^>m6zJJ^A
zc4gc5Nbe(8?zit>=KrZx77prOLe}RodCw=Seh=0@ATJI|&avOJ-di0Mop^6FDV^00
z#q|}_@!omU(shZS7gj!!-ZQ8wKk9GZ^nG8sY|_%ZJ>TgSd%X{k^`3lP-wQ6Aud1KR
z4(abZZDhvN9dbDF;3cp4&<!&>c`<3d{cYAK|8V-4r#e|N-oL9Voy2di`DO1b`bDqk
zInS_NIp^(V`^BoBmsU+zTpOPcylJVqJKa3?PaPPa(<jB{@%gUz<Klheq<e2g?}ett
z0cFqkkluq!eeUfeouM3tawt;|dW{E)wy)}*PEmfeOR|$q&rf^LBWrnC-$_WKllT0s
zZuOJ=bW`%hCr!l|SM{1^7wU@rdTyB1PmD9u(nY^(7uzNEll6Ql)=P_n)9%G)eeO|q
zy)?dxmZ!z{Wv>VOmZoFV<AkF8Cq0iXTTk-REBk!u{J-9Q(-l8HxYcwl8aL&Yf4}NG
z-_43;pX2vRFYS8>O(~y$=%$<e?tM>odd0HzChzZ}R#m^(@s$1EX3}>8%g(RLpEq@X
z63=UkrhC;V?cX-(j1Ru?*hSk-#~*KZ$zebG!BeeXkzA6ns{50%zeA4x)5gF5<lI&o
z#&&f{m;D?i?e`y=Uf!?gjCIkEUio<UX;WYMx%C%qHyu}-u;W2fA78XxH|%wOV*93i
z7a)z#ro9iC#P7suPhRxEfzP#?idlcBw668~{kwH#_m_0u)3p4f?mKXc4_y0Xt166k
zFfHBG^@#7H?;WSfrL6L#=NL)(N&RNkm$KJg)%1GZKMlHl)9&#{eyj3VNqU?$KkxfG
zv3^oNReet@slTl6YscUFPCFM&dJiJ&drL|7CyB27q;balr1x6up6~Y{vSQKkl>MBj
z-}H)(J8QeNeNn&4oBExK$oKXW=goLx>DjHiWBfhUrl+@`z4fVEEc-&MDvWlws^=83
zpR{z*?|LP*tLnK<)pXPD`81`xEWN5)-jROM`%h{5JuI(Q0`~WH$E@*vUF$XVUQzrW
zLDlK9o-Y<%HD&c*HXZxIvaqOp%5InT9DHFbFI&Dz^s?>`7u0%5a*6p>pUYOYzs~Ci
zH@ozNy(g|ard<_Axk{(qN6%Z1Zojc#_dJIE{QLLT)8;+txp`84(fdd7TqrGF^t>pZ
z$LCFxS5@aG=;|j8)6U};^zta5qVH`itnH%wnwl;v&hM2@)$+^w{%kyNDw<9@_e{$#
zDu*;a>$*<!en+Wp_3LupTU2lQ)vnj~n+DxJ>G^t9`9<d?@Ats`ZckG&@4l_B^^3m8
zleb+_y(+8S;xkS>b^5R7chaz`-?7X~x7YU&)7oc^OVaAadtfnrz?JJwTV>LWR@1QP
zd&YHXmo{$;dU@J?kGek(%*tO?=d1mqn{+N!_xz&w#f!E}@`GniIPcJN&U&%cbWA%x
zOe#;3OWFAoZ`f?dO};U&)o(0&97B3T3H)6Gd7t|w`J;YIGwbiyB-JbW9nzxrzw);0
zw>+!rFX=mHW#td*T&5~{4e1Ug&`$~Ey+1Lm|Fe0Y-~aJeuXD7Z$T92tVZ(Y4uRVF&
zhgu7Y9HRbLb)TvzUHl!i>(~vBdi_h<z2dKS+V8o!FCEwJ7gn{8@7H-wuX<k9cNpl!
zf`;*Znki?mwf*joUeYRxgO*R)bBnr{m;JrNLDS3nef_%BJ89P)_B`jNS)Ha~*6+X8
zz1|)>e|^t+uX?1@RE)o0R5abI9i4pP{nIYJ?~YDWF@C4MY3aQ88;aI1sweIpSB?1V
z^*x!Q@`>k#_<?En8q><rDT+n!wJ+@Lipo2->s2r7*1oF!&!Fh$ea_VH`h${7+5Psq
z?@nqT*PC=z-}|bHZqa)e+8KARSoB=0tafSq59;1c8edt<>+=3!R(_M}rQKgkDo>)f
zpypTgUHhu(_WL_)!~TB$o=;4_rbP}#&qs^4t7`paeXk(CFO>BjjPsAI<1V_-EBbqH
zdD~?@hp6gzUVEk6uY22h^)>0fRndI?=e5Iw`VJ9(i{jFkzvBEe-n&@4Y|P6~*56TT
z+W9)W-&Bmx`I?rl>wCayd{zBDuzu0!$D4|2@0r)NylLw+@*CgJjsB!p{VmI1N4{e^
zYy6tEUa#j2QI2Wps-N>@r62jM>V3!9PyDX^pwf}Aq;yt!W~~?dNqVn9T$c9SD{Fb|
zx9oJ@?*^3Je_iMm{SIVV{9kqK(tDqL@#5_T4eRnddu7d2*>sxrUVh}a>EBDM`h9?+
z`V!-(ey5A_-L!UP-J8sNFYwakkKgHzWoLATxNyoLZ`?Jk|M4mJd~SMYVV6VM_O<Na
zQQKzKGwz!Gu6doRG3kByvhoi+>(FDr{IOZBvT*Y1+u!w}pFZCj6mGclwev=<f1y<r
z%ew!94m~c)Pv6@e@YPAXUp>23_Wc}kXc}fcr|WmUUe805+UK304{WsUI;U^?Tqi5W
zbKb1!wC|!8eg0dtU0LVJdFkZ+zF%GI=hd_OzH-^5rFVP2lZ1ngmy*^~+4;SGhdOKf
zUiq%-ygn=4cn_&x>8kIQ*Zp}!-gRBHe*FGt-P1+=Oi}stt6kY~;(|i2EcqwBFHzV0
zrk|7dD%Z4oK6%TVer}rgytr=l2lf58vij}y9&lRwvh3V~N~h@a_N?v5t>4eh()ug=
zcRTw{uixi6P3@2V<;AM*hradX*PZ>cD`vK;#-r{#aElLI`(&$EjB=~$yIgtc7X6M_
z+Wz;)+g)<lkACn}D`LEFQI~Yuc$!jPmOP`rmYq&J=S?b)e0aa4*Y9|ZUuyH2C!9N{
z(<>&uKUXx)>$gAbwZE!<N2*_Rt3G#&`kOVKv>t}}OJ4Dz8)kH>mP6h4t9kj^@|RnE
z?dH3m=nNX8zE^z?SoM9#s_Eyodr`jM+~L&cw%%d<NKs6B4!6&RZ~WSUN50T0iYLGC
z{1wLCHn)=(zx0ROZ+zVmvpZ>+_MF;#4Mj2U`onLWpQ_VIzjISGzifW1UjOK)@tt<g
z(eLs!e)E<WU8jR?ciDX#O!?L;=XRQkN%phs{Ji>**RKq^e%+t@X5~-3kaxa{)-U_J
zaQ&v2_j?_6t6y~9n%b`DdE%hkReg?JHT|-`bJ(l?U31&9Hy-rJGp$~+-}WY{zovfY
zCGws39%|j{C%vaW%)j;Q^R~XL(=R#n>z-xZ`boN%ns$#i>A8UXtlv*iFMgk<U-Mpd
zKQrz*_iN9Mf3}kqllFy4?`y?#8}jTGkDI;Nl81ig;Z9T0?`uT)G$kF^L(|e}>x!MI
zJ0`8i$ajn{%X)s3^n4>P{Uka``FWo|m-QY?*7vH^%lf_2I6ryQ{oaS?<u7mjs-Ful
zDD<<QgVm+ppwBh4_McZ@)$dpQtiOm^$JcAU*zYi1OajqQH!YnON0nV(G|v3>^A9eu
z?n<*dX}I`|6HlG~tNES0=y`Kdyzki0o;_u!XFIUh_fxVyXD|ErDT}Rwe{QwcrXL=;
zY^%FlWh42fVXyb7>voQpm!G=6-)=t;?Le>R39<cct1q|ZG0#oyz_k8ic}z$8F>Sw4
z7Lrq3S#|l{y2w}6>Ab%)QMdZ>cNB-|DLbw8sRuey4rRw1Wv$1obgEtlS<fTV>ZS3S
zc0Qj}Ui9}-^7^-=`q6I3bk^U;NqTOPlwbCHYA3Jw#%~|?;5n_TF@E<XO)v76Pupzs
zPha)?*mhYs@rKQI+~gbcS`qW|RkXfy{~-?McgEA~THVW=t~W96?RUE9x~*%wyx+Yj
zT0d?4F%GLboyL#yqT{WqUD5NQvfB0fzGcz&O?j@3&$8q_q?7Iap`n}xBPVt;;#x<%
zVw1g}pV!HXS-%7AT&h=0dY)hOz3;s3ik^?=?I*2%oVT>^3Z<3jJ@=~m9=Z277Boyc
z|LAvq*548C*K=+CM7{2Jx@x^{Dt+_ts?}e5+ffgXX*U(?@*JcnA7#IPGAMdYv+MM~
zDvaN6YDzk*{#1P)okl;&za~8oXlj1H&((VEuj%KSk^i)~z3$~n`>UqpXWh4|mv$d0
zsXVJ6$a^1NedkemF=?Nd_Iq^~es;%s*W7z&D+%NCP|r)U${}yPqThFq?eg|pR{ehY
z?|Y|ixW(i(|J<qyv+fgntylK_4fJ|_9+9-3ist9Fqj}$V>Gl3-RqfO4PhHEC?0;4H
zMeR?~dCRL`W!KNU9_m`Z*K=*{>vpbE)^~2R(n->P-=EEj<O%=qrV~$F<K}BSRby3t
zxG3EudDb<*=s1(wl{L<$e3v?j&$8cpB-gxgZW4b@%O8L8o*Uo#`8!&JLh`Q)v)*^D
zYrVYBGwWL4K6_ADm42nsO*_x6dwEmOkNfpKfwJ$tqvtttQ!(i|MOpc2>#k|#MbFdY
zI5+;lJ750zHDgAy;-a7X@=soS&)AWynD@LO?Yr8syy*SRq;`3~N9^3XDCRxq&bkkb
z^~!qhIqFYZ+H;~_-!G2s`|SsdzNeMeE=_*1Jnen{SikJ_RmU#9_sJJ8-mVIJ<(IzY
z#a{c(+CDE|Y4x-AyRg;EdhS;DdRgz8#Cq{P#j?`}o^|N4U;fyvR#UO)J+!RtW~~34
zkty$(-x}h-mcztpPhRxEfzP#?j=g?=I<0-Lc01{OxGDKb?~mur?^TY2Zr|(ggAUu5
z^E;~xsvP^R-*vS=y?!S!Yx{GLy<p@0mU^r;XgqBByJwuzncB*VX?k+)r4Q_T$MiX^
zrs5wqdwA{1+dkB4D&~DIS-1LrPp#<~zqgW<PFqiD<z;^_aM1Mf#&v_P-|utCvis}x
z_n4dB{-;+jJ?)xt_qT@FUpb`NwO-5fu9LKT5Lv%xm-L=|(R->@x6Au{u5R_KmQU66
z)1K!PE$?;x4640*tNq5h>2+Ryn*KeRvfhIjl-?%w+pm3OQvYS;_xe6jzuT8J?noys
zP8gP7@xu+CxxYgWQGN?+y6O9*$bbCZ&!(i)?&BntSFN{A`z}c2w<`JM-N)6fe%|}I
zWy`VG_cY4lr|$P`>iRrtVc}~~?+uit-;s{^r{7ri`!ZR-b2+GbM=$-ADLc%5s#P_{
z?`tfm>7@59qJ9@mSM|Ix@>AA5jg24u{YBqh`{7nrjC`b}7Z&?&oJsN+)IG_%#fweh
zC+&OjN%LH{{6RlIF022feQQ(l-?Zaf|FrDfGde@8Rt`niU)nw8r1HFe_~m0yTjD$W
zJ>44Ozm`MR`RjGxk+q*;y#-qWRsBA5zob_5J?W%&S-<;H)%dEKk0IR&r`>b>t(^r$
z4oUm?toc>xZ&tcRzcX2OyCv_uZv6W<AKQNG$*(*6Wmn8>H5L06Pi39oy!F!dn@Qz)
z`{`Hx4R^kF-pKVYwDMwJ{`z(Qq^f@U^<AE-`zh;Qect&myZ#xQF0sy>)1PgXjalD6
zYFfRr&Uc1%?tjM->t4}Hl0#X~W6EAHNq(u<{IdBR(i=+P|CT`3_gIJZ#-4rI-hXWs
zmBarn^F03--KVDYQTOLhz5d?9!qz_T`zvYsQM9})|BOz*_0)GZn~qJ_|6cDiW^Mob
z)7F3Ox-;gt7Bnone)G1=dQP5HFK>TY>kpqVSNnW<L6>h?-_hHC_SUCvvFr=2A^vMQ
zWL?h-TfISj7bWZc$FzD?&sV?WvpAlj`@p)itNI+gYWh_@hbT%n>PJjxjYk*MdQl#E
z(?#bcYrDMslvO|J{z+Z(dwrj~*Y^P8@3K~%?)5%h(e`Ek?qE~t_4|7u!~VAV+Rb-A
z(JAY@9C_)KWk>VUsp|Q1+0V-dMXxIP^qX$4@pfLlT2Shj^}X5!MIKd+!~1f*VpaSN
z=?*2(1PKi4z0!X7JLjmwH(31ivF)N*)m(n+h+EG-Z|l1{Rby5BFQ{~T-M`nReb(=?
zCe=%u-(J6KQTF#ZlIX2D_W8}uT7B$D-rp0<JHEWXbC~zMp{(C2L`VP93k&x8J<zg$
zm+7+mHkk6QSI+GW3e(nkzslq9mo_E6;{#Wozt=KjMw*KKn%}hNpLHqk_x-t~{`xg9
zdHY#V?y(H&yJ}_iTh@D1gQDZU>VkstJFr!!^RC;Z`eohQsym&e@#oDy=KOb$zv;RO
z?WW^xvtIn(As=0&-7995-=O?P(mmQ<^Rs?0AZxv>a*Xw=zBiPWe&nlYy4QPrN%7Re
zl3(;b#h}~e{T_7P>pR!Yx(@RGu1VeMS9R{0ReojFYkEA|>vPdTw{QCI%?&EfDXZVI
z_NRlQQ+L0R#$VC*m($vnWgqKK=c;3u-uvW>7jI|9yyN-UTYmoDla`p-O2fQ#i=O`{
zJ@?9*?>SJvG4H*{SU-*5vhRZ@Jts}ePkOGF_jzHgpZ8vOulLIP)xPfMNLl$yvk!GG
zZ|Zma@SXI%<*M`Z+WWfJPx_8j)A9$sj~VvA;TDtE{Bw&OvYsOjx?ZpQ&$_PbynJQ#
zb9Jk?pu7*tpT&7g+duZJJo1${-Rpgiy#2-c@pl>*)^wCh-O_clFIoBN_j_Aq_gB{Y
zOoOJAb-vQ-rLBv+-;*n=ei~mz%ahJ6nv(zP=O0{R-IZo_nu;gwy2GC5+%&7xRQ%$j
zhpqnmZ_Mm86`Lye$ahoLYvi-vdXr`+%l5DRrk57qCXK(>IKAxlD{u7ukzF2}-fAjN
z|Llr;j$iMYR#liZk9Et>`}=oE_3e{;#mz1~Veg6Sj%haq^Zq_c-ReKF%^oZ4u)&;G
z)3IN_A5?ZfgZfTNul<%She6Rh;L7!;tukpwt1g)L-dkDsjp{-t?K%F!US7BOA&sAS
zU!<xuJ_cn+`rU88&-2K)Us&~fTm7P+^}Um->h(HbRoy4=H{G&-ce7vg`hEUVm3@xq
zN=->O{X3gw-@jZ?>Gf(SwSVulGgn{jj}LZw#eT2*UhmuFZD00#fK9s>m&ETDH?FhD
zy3bzJ=@-)5L1Vwq;p^7lpyfa8f8mfo()oSX{H*6>QJVe!owC?}QabDXkF@t(VtL*;
zDeXPtB>orl{H*gctatoQcPu-jLk`~ieB{0-R{z~zw|Am`#&p(rqhI%0(|!kXLD&1?
z_=XZ_k_5^=&r4hXWtGpqth4W7KYXNB72dSuk%xY4yT7%{LOPN4_o3=quV3fyW%c8C
z`^Y~D*ZIP>Ke_41G3}}_so#Yqe~+ENzURDGJ<{nF)8g&C<z?Hyy3(u5d3dkqqjgy?
zNq($q{_6VPdm3N8elM^t?d{W=ih0jB>Q=vB&u^>hC-3hf)UAHM?r~LZM~C$PM*=6k
zZB!B`?sMWUyT4(p6}MVxtFcF%_~8?`TIBD4cYoRL6L(*Jt1;g^arYJfo_^z2W49Xn
z!MUF~{KVZ)|Mt7?Y>gZ9$cziV`G&8J8(C$&%`dtAo#RF>{M*ypJbcU|BL_Xc{winv
zZ0yMU9{I{cyI(e@i|2j%d+)pMy<<im+~Xq$Pgrkk7uA<9y~5d3e|5>MF(cZYa?zTv
zIQx;?I$v7lkdb$vJH4~x^Glrh`(q#M;@cKq{GjVLc)0V&W!}Bo5i=g`?7h+sr@naj
zjLyBUx@Yl&wtKqs{&O$hZ<R04?)>sicl_l)=g;bFv%zv--{{wKy7@1E?J4iP{9Vs<
z{@03Ed}q<C9`D>gd84II{n4|XpZ{dPJGcL@XS(S-zIWi*ohHucEWX^hk6eBCW1Y2T
zZ}p|4&wQYB=QTgs;dMLR{ZIA2@%R&4|9Qd%ozCO0+x7gP-P!uc)XC4Sd+Y73dGk-3
zyw<9JX<c~WZV&Hs?KQ0nfAaWd*Btn>&V@_fddGoh9n<>c$vd3%g^4${)|`39O>g<y
z-K|MSZMD|%7e3Hh<Nv&O>~}vtt#$eF*PXP?;SaUeJ$sR*Z<zL2>*_NOJL3b(J=!|%
z^xy6I(l<TQdeyg&dgjB||FxU{j!Q3JW$~{(+4|Y^l~3tx`Aip2IsfbXz3<ZLE&LsF
z=Q*!gf9+el{Yv8te&M_au6p5uOS|Ja;kj3D@{9N1+k%gN^TyvU``7zh(0Gkww_p7C
z=1biCcz2xoLHFttA6sGE9v8HhI^}<Dc>3HK9eCZCZ%^FnQ&YQoAPwoq4`2GtN8dZ)
z(GFA|(<|Jw>W@yp<G#+bPan0(*1!F07vJ@V*IakTTkq{m-QZJq|HnIT=txII7k~67
z>WO~ojWmAcOG9!aNAjZw<{NH*>S7!I|Az=&{m>2_h|bKDzIDSkU!2<28~G5uZ+&-(
zoo+ex!Itrpmvt#0T94!-|DCV=+?#iM`pT|c*Iabk3%7sx(oV!!&pbrD`pmby`xl$v
z(!mE*Pr3f_ElnQQpLP&E_=W#|{mh>oeP@?{^sS4Caa_u+?<IHr{5$qPXr#N2<eOjf
zV11~MuKw{y-qO%`jf>vLe024ve|C#KlEx1-K51xN_%#po{EQQRd&8^Oe5&)+AN~4_
zRsZmG7ujvN{5i8XzvtVtx^_0!x30(mUH!->7ko(5JA6ZOQBICfd-5b-<Bj$7t3H2*
zKjXkR`yf9eI{J~G`?5v%-|h=z+i=}wulnWpKekAF{U5yMzb-y)d|UdY;~rXm&I4oH
zS1+~2dsg}8m~MUP*DSmAY8P+#V(Y(_I`+a1cbnIej^z)&c*4CKkIZT9H)r`TUG~K6
z);C^q%Yo1B|3Yi}%eHy+t6R@+eQ>dfTV8tn+?IOMJ00`-`!Ag{rX4YBz=qy=FWfeH
z?(NTY`@^qt`0iOB-Q>XWv%BLlj+Mqun)sh*Ki}G9+7H&=?5i`oe$see_Vme%Zu*K@
z-SHWhe&8cV%y{7Ai@(sNZ(QW|rXxRn&VMd5rXAap2Ris6NBql|KH<hG6F=Me$GoAV
zoLtbAR`11=cR6I=DdXC^t#IhHJKE#A^G|Q+6|^4ceMI`lt}55gx?&gPTVLAA*KbzT
z9)4?N-OEmT?1j#&etp5VE3P!R^XSD-uX5+w&v#Kg2xC3=M?84TeXCx1#ZR7UiDPcM
zpmqMY&Um65Z?T(Dyar=Fdq4MQo8Nilyhpq7-uR2>KXlEJf9uA5;yQjE%14J^gZfnt
z<%=KGXWx(b&d$I6(&ukz@dxHje5RaVFrVVuZhADX72ly_e)tXRj9yxor>y+0SFX6$
z_^y6RvtRU-{<3Gv<x49!kI*{PuXfgleyrmp#D7HV(KxMR>k@tavl}<Pa_omcy6mFu
zcdUN!<+n{;tZS#yM-P37E_)Z(y?W?Kt7pGJPt2S7rziA=UehmWdP>ghtGGELzYWFH
zv0ks~7(XP@<A=n5kbR=Z<RPEE5(khU`B+!hkM`CLd&qu9WM{<<@Saz{_@*ll{#$pS
zM2`5;zjcQX^C~_u4)yS3{7||1mX7V|0evA4Y2(HxKH|8^Q#<`>XFjwSM?v)TXMW|Q
zBW)bUuUx;7{NlLqNj~H+ADVwiFVKVL`NVnG-Mi#}KGZ$;5EmNv{Bc`u`JE3;>z<Q{
zGtJvFulw!Ji|p}a_nc#uw|?aPb8ddF1Nl98=_bcNe(_OrJAd2sL$8_jl^43W{~K32
zdz0m7b>nvDE}Q*u?w)^nXl4f=;$r(l=QYr{pmP;z_0SU+qbIGN^zKVsec^=F9_gs(
z+y<&geo1L^r7!pGwZ}4FTm7omrSIDDSD$>~4;_BbIf(hQzT~r~)`RmexW@IL-(k|^
z7dw|<`Q(h-Zk^Y;=)(1<e*Zmly7T0mfqbJ}@d=HG{<8<s-Wk7g`jQp#0nHz@P9Fcu
z8;`zV%%8gFU+<jt&@B@;o&1mUtkHcv`=&iP(+7GiZJio#QKT>Q41Md39!2`(ryR<M
zv0ble`kq8zxp5`+r`&v7A8{Pg5tHb~eDj9>y664;{IyoNwR?_(F1aU-E7HxHw$7X{
zTW=9R{+^qT+Gor8otamkx5jdp&+jOgK6=YzuDEc$7d!SB?52H!ew=HoXWd%I;t+Hv
zUH+xRKRWh>E?w(iJ@Sb1AvfzrJ@G)S&mQ1|J&>=xcJkQ=^&vhZz1Tl`Nys1acaVQ&
zN1$?Pc0f8J|7*Wx|0bWG=7-s9@vD3&j)3xwLmFNAP(FV2BW-_%5Bd0x_2W471GN|5
zYwtOMa&fwH`(6B^7g0Gx$38_GqQBUK_dYlF=ou|(<I~RmL^~+o`HXQzv=2g0{k<Oj
z#I9f8@#$79w?9)~KlV%Zi}KZr{o)fEpLDD*tsW$w*p8g&2f5QbDBhqS`aut!B-D<6
z;0qr8-%p&p&6!7b_ITMBKeGGQBi-j>);W6**%$Ufyv_eW>-ptterDV*-@m3K{uh5)
zckF;TUA({^vQG#9@+;?WxBYV+=gIoV5B~TQc2~SC{*Ue0$9FD&+NVc;_|!i=hh!(k
z$LK=z`G+%K@yb7Ly4TE(_!@eSsUEucMc;E!^`y-Qf5Q(-N94C+zO?q@)#xYH;}0Wh
zZ=aDDoe%U2#kb@r&Q8i#4%MTVP&thG)`NAT9P)GYQ`)*^ALLsv_AUHtH-GfpPMSSr
z|C}$~{p6iD&0p`??)koac9ETAAK6{!Ow#x5`}W&kx!>&W`MP}eiQVCE*fVL!9zgcc
z_}MpfrP){G(Z6#*{iCNp`#g3w31d6+qP=-x@8Ud4n}2EcE~#Jn<Y7MXFRh;aDt*A0
z^2i@P(5Fx68#npM*RS&$boHZO*sVXRm-3<Z`o|afNUJZcKl32oGCvWO(<kGB=#dA$
z^k@E{aaE1{pXVR;8J>5r^AXkOFWG1Gq?7DC`yTB+f1#av`j7e2P<<HbC#BU-LiAze
zQ+{kOAOG4x`sEzRy&KO_tP{_3q}yMZv;K~kKhu3K$v&~G?5sGJeHPEFZ~blfADg~$
z^>v<VLGjz3JI~u;mz8F;KC|em6VHC~zV3Mp{dZr4ev2RR$sXVO)BpJRKA%{;orV+6
zTJ@L<H=oe{!W%YydfA&6Z=d$#(_g#6mlkWYyKmod+h4!uibdO=e|+ZdUp@NiUB|Z}
zzM<!-<YL_8e{<&HYy9T+4msmrySQ%M+hRA^i?}|!_Hp$7mh_5$eDlRqK6y(^8hv(E
z{}I^-ajyRHFYaYGoSU$p;$8RH#D(Hg_uQOMC>J-XCq6L_@r8ItTD&~+)QyY2>`(W1
z#J%_z$Nb{9)7O8^qJQb$Z=;v|?vHo9?1UeWo!Py2W}QI#NM7X0U$e8$6_vZUB%fb^
z>@qa(`iuR`H?P*Ic`<J1TjXe8L=NWD{A%x<6F<t)(_Ve3oq50qJ&E=7kFL1V{My$@
zn^*ecoX~kF{bL8L6Zu^n_1omv&3FHC6rHb^h4hu2=%uuIk}pj@kp9s(dSt&F<!wE8
z$2mHF^GZ(kvC`f8<T83sR5|%U_6a{p`mUVcvA(6_dZGVu-1MG5QP2E4C%XMJ`(Jne
zV>7$=OYs@yVqD5S4<MKBI7j6W$H%U^2diKAWBD)qiTl}qb_n0n%8gSRJtz*8CJ*`$
z(fW=^F7y?~_350~`iy$ZF4JRlp!(4HLqzMHJ#g;JPw2<`Fdpl}ype}_R&Ktb`5|xf
zV7~Ybafkh+dg2fBfF7jB_>oqh-@so)d}?Pr((0?HoLpjBy(FYJ?6>)&C-PyOZ|$`+
zZ|0NVH_y)P&9n0qbj1zgef8;0OzW4PIfqgYY7g-f`-%GvY5N!T^^0HqsIQ)U`m29(
z>+0|5J!SN)yU3^Y0Ld49a@9XRrD5D>;S-ui{m?Ibvt#s)ekqUkLwjhx@JB9?Jwh+m
zi+mavy@ljyA8P*u>7Rbphv<^8@j~M<kK}LO&{a==<~R1EKk_%f__yyfe(Od(?aZ6{
z^itaR^@mUO>^r?@<GfWqwEsm{d-`R(^2sG4z9Sk3dC?2{W4)utPl|u&q47cdM81vF
zd4zTEJfFNEIhY50(Oc_6TqG^-pr?>tL3E8Dl1oH(mpqc>Nl%rd!_L!Ja>So?s9b&d
zkw4?rUO(!Q7erS(X-JOzV3f0UsJ?#5Gv+Ht2VW4s_>)f$jT@q;A9e$M`bh5RnP>3<
zL^t-Y9sZzsweIBWkA9#RF^MiZ)(=Eix%SXL!@NZK7$5o2Cuwp<AEFP{gW97H<9N|C
zU-*@#|M+*V#?H`3`l_CBCiQE+@vokC<bXf*<U?_%eHO%*xKTW6AEUgB&b#dgN7261
zy=wb7=X%cT$ODoyzO7esGEQmZk?xiUe&mxE{>`^?>m3@uw07n>=A(x%bfNL77g4>K
zuU%HO?=X+(vMcD2Gqf%tyF;J&2WUM*`RoZg(&))oKcf2dU0S`EmM^}+PfYW_5FLm=
z?WOT+UE_;BIS(R7=UwEk9lenc(J?R5&coD$`gaddTL0*<YtCbx!#c0i54z47*)!)I
z?342qe#ZK+54Qh>^6ii9pV<@rLhIPNvi@OGK0S~1<inS)yZ?!|d}oRFoo~MEKaT(L
zOWOFc&xYzpOrwuK@*^kXH9mR(t#@*eZ+;@0U-J#EQ|XA}ef8+I^<y0<hw4e|UmB_h
z<>MpvuUx;{No%K`@|Z?f{m7s3#rEi^FU~;+DmPE&(Ym(I&`Uymh_j@Pm)(i>De{dU
z{6+nsNBI58n)BW`>5>U;&$YbA<M|eJKVtl=M~*n@g+<$*SM71y{nMX4V$rteR^IpV
zoI{%4>ld=4ah}M}x`EaiyTiU|4~?I^$&>uze$4(2Dwnpev(HUJ^!XQPo{ZN#nIHM)
z8)|19(orAui(mW9q<rO`^BRwFYo{N6%y_1*{k`RW{qXGWeKh?>w9j?V*tvs#r4QTa
z)+zh`<oWLRE!+pw-aRt>;6pxs$<4jAxSpINI43a<e52#NO8NNE-grEpWFMR#K=W+=
z=q3FZ53&>VKpI`+g4O~3p%+j*r<~ug4%q?a_`<LD^7V@!`Qi|Cj8lE}<V#0B)t8SC
z^~nR@P~1(9_+b}}13mFhx83M>JM4S-edm#oy@}6f?B6_>(;t3~hukKw+Madj_nzq9
zGo=SmKll+xiO=w_Kl5t*;x~vN=sA?<4(NMt7mCZec6{`EEbb8*w>XGAp>asN$H0!;
zSNdKDxr&QoyyZET{UEyNk(Y7#PK@V++8K}agMW1J9phbT_3T^mE8l$3PxEShK>8Td
z>MOTy%pdx(o%-~Rzhme4FLs~ZWrx{)=a2k>bP}Q)>pRCqU%$89zs^UdzJ0N7oW0j$
zd(Jvz=7jD&8uw}5`SW|PocZRLv?2dPU&%eryM2hb+B}(O>&SXBPx6!G8uO(g`LjFL
zzkQf=j8EA6cz>I{FrU$%@h|+O_3oa7e)LOk=odR_{N!p~&}Z_rK0T*qC(VQXAUU#6
z^w<3!<8nWPJ#^m$ALa|9hrV+p<MQ6qUw`(71NM9Am=W;-|IT0Y@7@cw4(KO;uO2$m
z{J%H^{aC)m>$m&z3vYV1`#mrI4SoITU;HEPa9`f~r3d_>@v}?jfxTlt*=zI2KB})=
zzWYMT*=_dT`5b!uBYVrfy4Ocv_!IucIaIv=puO_}bk5ml-yi+?yR%yE6?*RlzVD!m
zHv7R=bGqg3<MA8r3F5>3t(eBIa|%dK>dV&;zLxyl#xJgR+tb~9kH$w%_#iLy#6Ot_
z^GNQ}=EXWO589&#*$L~9{#f_ymo)ppK3bRjb?hI%__SW-tIr<uuj~x{VsDJg`EcaN
z`nNx1_u~7a#>K9(v+NQ3D4#uo)+;%%Gw8ZUh8}xreD1TcTgt`X=qSerzSv3jQNH=p
zfAlZf$ybm6RF1AR`s5|8yl!aT$v-RdkLr^<l%EyJnLb8j-^D@VB+rGtkE@*jTIuu4
zO?mGsi;S={key)<y^rEOmx#XWwA}vBF1h_K&$S@C>D)$Id-t<rT6=bq-Bn*XdeYvL
zf$S#!rL{9o`A~i3U-{A{pFM5zv)%7r@f-Ma-%9&ve?52N542<F`6>244aG&;v(M~3
zWFO;xOq?ZO{Hi|t0of(%*8WhrcvSqru89lTC3Y3sZ-{HzEA*^C`P!k!PI=xTPLa>P
zNTbWnuv>9GvB%oQb&4PB+&J*1Uvh9hL_XxGe|7=NCl}+>PQS*%&dL`z^2^ST#CiC#
zF0C)?m_EliiG2`%h@0rI_=A3m7sO-Y3-w|<Y4xLi(@%az9Kg@>i;%w;r%0PW@+A*`
z*?f_oeXe~_l%sNbf<AqdW=F`+ILTQYNq+dYUfB<Fg6xI*5y{ax2tL?%{IF~4L*szr
z3FD=Q<VgRGI}NQ1bj&X#7k14!>?5N65)aZhdPq+6&-mD9Xg=6;_Bzf(oDb#vk99<@
z=+a|$+kE4jK8t7Z&0af~@*K#$EcNL*xyV<~x*<RPiLdcPPV~$?&@c70V}JPvc99*I
z&z?bXtnnqGeX;#AI|r>>=NR@Q^b686dJ2_8`Qm=<Abl5~X~&Kzm(~tlNY9-!*(X_V
z{1Ja^zrud;Q~U?CKGETip>=7V)YA@K<CBj*xj_CI|MZLATSwM|cuqV<e*76e$&uY<
zm*q#)9zAj<|Ex&9^p>2=hjBV55hu}i^0eNp7wgh`QD1*ah@SfR!3Vyectkz<<_Xd#
z`(g9Oznd?}U$77C89QKq%@4CL<b)r7#CY+GANE0gboED`>=VACp5a?R^cj7q9OBP9
zq5tZ$v*hR;11cv6c9=Zaarw?S>8p7*zw}GJh|z!W589(=eAcmX8K-{ei?~yLbd;m7
zeMEd&2kcmsFZ%c)2Pj`XDDJgB@QW{e;>-N8FWS*Ne90#-s2zIh#kBhP(T-iEC-xus
zq@V1nb7gkUc#YpU?Hl+T{*WCMrzT-+r#*W~U(KI>txw~_m+=^HM1C(Xikq!#>tCE?
zpTe)i{W!ZJo`I1LJ0so^M?(56uCmUpAAW-ztXF;>zv5JWLAi5q>xbQ7XT`71Da6P4
z(+)l3bY3h!&MUn%9(Kfd=(RMY|MUVKNWS7;^5t)gj~}9E{G@iqrM<LsUiOrn)H7c4
z)~|AMGcU^d9sQ_BFQ9hx3o3{DLl1W4J^DVUaX1$@K7NH=;WybeezxDpUWm_}JD6AT
zo^@*;tRLf2-}-{)6Oy-mmv{tPAL38?5AE-qgIGuQb>awmYu?_n%}M9JWc{&i<@Tfa
zG4Im&G46==o%BH5Wj*UZj+0){PjZTB`zhtt0sV&Tjd9X*<@C|K(P!o8#_^*=PV5gm
zE^cGDjoWz9)t`Lr%mX_wO>WwuD{X!tJ;A?z$;13Wa%R`rQF=_T&6Dv)`t(6Qd6BDm
zgZjsp`s8AK^bT4N`Y{j2k8k~;gFbo0{^^Z;XujEP^q~2{5BcEBc<^Wb<VQNj=N_qZ
zP3uX0^sPJfpWEv8eXoE2nC|a?kXsx-xyNzQU*D%^SLhY~q<ue&y`jJO)Gv&3a?a#D
z&v@83@{MR*_|dO?^`(>0c$4t@6MwSRM!#RA?RPj<J@S_;egCN!S_d8S<$dp7|M~9s
z%*1KdGx?;Uc_n9Z^!!FSWZ(4%-Ai^a2)^}>nHR3H|G2i_Iq)4Fdg6BuJil^Zncg}7
zGe78&gY$OZ@gaBmNv`P1XBS|cANph6u-Dr2ug=}XSK<Zx5$)K0a^=VMLw@8*-sIz)
zIoc2Lj(E>JqGSBzs6TdsUz0DbAMx~MCq8lMM}F~Qi+){r*jvwC`BigU?1_0mhdk^D
z(KSElA938;hplks{1(2=<Fz;Mv*p=mjcv0l?z__i_R{>|!+K^v*g0{E_N^~2dCei)
zF3~>q&wH&l`GbqLmpx$d(++y}rR_}~K6m6(FPYH&y&&?ZxAcg7*$edDqxX9t_#iL#
z&V2E+?2WjL9DQ%gdbcj5$8LYzVT;{0zWvp=&v|j19hPcKlNXfl_f7mx)5#xMcI9s^
zHm>b{AO6+%oh<dOTl#4p?N1>7trz|fKYl00IPSe@<=uYy$dX;ZK`!(rt}E*+V%!Jo
zm;L8g*+u#&?yyg@4~z7r?cdaMeq|r;JOlsaWPYK1`wsM!YtLU=@6r+d{uBEsZsPao
znRUT$u@mOcIGp21vv=gguGv?SH$LTy<Je7cpY=nZtY`LE+Pp#Irf2MdahY%S3tE5d
zGI>Dd^p9O(_vi~hNq^-V2foa&{@68kN<HILU))Dd${{)6iyY}SIpB-E*N%RX8+ntH
z`NOw)<44>Vk!DxL;rN8)2=xc!x}~q;D`=eL56RK|86Q27hU7qB$(`OPXTQivJ>}{v
z*Pn5czjYSp#W>BIafnOw6Op{gS3JRAh^NxfxnxA=oz6?0A4gQ~+#(H~n>r`;oGH?k
zc7CfL&m*0aI_Jf|^KI?bcb*!@C9ND?_Sm_Q^GETFG`q$>v5(I0AUf=*_VW1)e#N-d
z$DeaT=T`EKn;bm9H9qoC&p7l8_0JE<M+bdr=h66(ufLea7u4Q!N%yhoiTS55zFSTo
zv^QV)faGLejmtT|@k;BDoQwlKbf9+Xi*NW}aijek|16&!^&W=b^YA?0eVKS3=pH*e
zu&^V2hSrUBCT<g_ITwfGB=MGb(0b+%#qHKTdd{!-GkyXcepFig1o4M|$bR61y}&oR
z`W3%<exkj7r~L~*LI3P?>{Fn99eq|$njYJyiHF2F_7~1E#EblHS01D1_2NZ-O}TN9
zYm8IKgTLjs%?o*uk2pb`g&sf4kMqw^oG8u{e~5edckR*BpEyQ5CVtp^m&Z2x>L$yy
zxA@J;H?RGhm$vVE*BPsw_VH!9`#SG)vmf@&?8>}b-*(U=$1T$Ko;Q2M{<5p=iM01z
zq}_vNpV{ST|JX(Ln0;oa*zf4y@#Fa`I<dWc_0hu@`;R_9K<|9V*?w0(d6A=Y6>+8c
z5QmB5#JBd}^j+L#9pHoC5q}s5e$5a6VO?A2>_Jh)7k=?iZy~)Ee_A)zUqtJLoW#A>
z2eeM4(P3Y$5B9)1XJ5s?_CL@*gdNb1eHMqY|Lg`opuKj^>)By+p?2m0J;+X*Z}&RH
zP2x4E9ec!on16gj`)_<e<@)E3=q-DuAO4NqWoPLdI`}bOe47{ZC12bX`InDR<B%5L
z%GWP`^l!ePa|-)T`zq&e&Jpd4={^0XxAc&F^cQhqkE^O0zkQ;6sNSO#*ZN)rzvTU8
z{>tx!?!U%jE3W*FF(Z)wil{yG9<}#+yyxS+PJS)km*s!hbK~ImjK_0(alCxrr!daQ
zr}tF7cj|pp<BsLzgTCKk!ngj67ae{bpZu$FL-rf8_v|sdBhAi;Z^U2XGj^IC%8KO8
zZ;B7Z0m}K?I8WjW``74)<wJJL{*b+6C)siJ>?_%4c7;7QZgz<O6@Q4&*)M)sd@jy(
zp2Lngr_|m)fIak_LOuT5Jdr<ph%Wyto;413HzGTME<ZsI<R#uASN@Xx$deq&8y!fW
z)Yo1+{v49O=TZ^z8}rlB<d+r29TCMFX=wjq+-Ye4VSGs_9!o<07m{~G^3IFwM;g-4
zBxG-rFlqeq#e+%6KENb-$+s>diZke!G^EFAX#KEH#!Zg&()gfq>)N@vakJ~zb5~zS
zzZXgmyYoEyy-w>BJ^KLrC3bC>5C886w_J6R_J@!8;?{TEHLu04v47$==M>J>*g5sp
z<DcA5a898==zKuhzK=b|FS_=L_EYR0yA18?BYo}IWAQ7yY`pqK$9SCYm@n&@UDluW
z_F3kYe94<UjRPNZU$*G}+kIhdo18tjj(W@9h}YR+_Lcsy6A_&wboG4nyH4ov6HtHr
z1!Q;Vy?Sx~fR1%!e;CI#=(PAvJSKh<r-?JgY2r<3_DEb5<0NsPILJ8#v>$^p9=D(6
zzr~Z{SbX4DT#8R|DL#v${=~uJKj&KdS1$gA<ftAwtM5Gl`vmX%i_^(bT)z6dmigEQ
zJI?FK$Dj0|^{OuYvpde8*>U@HaWMPME<)>=AL7^eC;keuLy&#pAK9zep5NvtwbxF*
z{@G>cDbiiP81A8K57~2eT$<f<?t%~4r9bMAyZNJjn7sI7<0n6MQv3nMpZJ5?kuP~@
zCr#h9gXE2Ga^|n}ORo6kCyWQO=f<Hv`I!&)p8awz<@p`^=eZyM!mk(~`#i*rmOk}I
z&vphS2l^BBN!q%hNA!=r(KC7{&JuUU_>6sW&M3~Z&lW$~U!!9^i~Gcr(&(~p&IRoA
zrNwFD-595;haS7?Ih^+T!I$$H=l#x$odZ~B&S$M3aX-7nE?8IQ$2!v9`V_Z`Pn?@u
zf7Ycqk-rw#&^z<ZKG+X-?eD1F)vtQa6Fh$shu{a_&XMrLPKb}B$pu|{h7a;&pUt0g
zaXPyN>8bX{r(8eU(_^Tf_WF|#trLFGda;gT+I+)UkH0a`^uoHKchdGT#%rAHcjSwH
z(LeqMCVl@>94D?5zlqbJ^Izw$;zRMGb7}eX(K#;^XN!y31>-hP#%ui6q4D#p^hcW9
zA$gcjs2n|dpkL3W^{XAdkxyUfoqF^@TtHv6*MCIq@xw0Z$Gl6kxA;L{8hz)M+S3>I
zA_>V)xpuIdKN??2Ysb#$mp(${Fka&)Kgf=Fo@U;xb8_M*p!*f>ox?xve)K2D9W<f)
z{)Kw>`_MUu=h)6`#Ov;X@FV)A*UnL-={J6%a~JC%`c>_Xmwxh7=qi`*T$^6=@9c|w
z&-0Z_^ZWd%aYn!G-Y$R4zsiU5p*Vs6hx~#5@Sz=g%F$77o}70<zmMVmgYz8wFzwA_
z+;=(G5?A2YeFFLRZN_2#$PFDRt~O5cray6h&09q4pB+v@<HEoA$oS>smz?nj>8bIN
zn|o5`1$}ZR_jvC_+#t>n|3H2N@<05GeCV7d)>A(z-?<6&98J6hJuefliHn>E#kBFU
zW8y0H#CPH@<Bt7_&z$pkt|!gU#(LsA^q~HvBeK_!UlI30@uYpIxJQ~li2g%dtv-7n
z{Ruy>AAUps{6)8)@OxP?pWLlK@tJuskLIhpuO2;Lz@L8UE&Y-vU;Sl8{!sh?^~Vp8
zgZlU-cX~xH=n4H5ztb1<98o;RZ;LbV1@WaGz92hAkF<m0CP>cwK0QE(Jn-W=i*kN3
z3E9^q<nO(o&wjC!?7!z4_L=NB`zRgdD($@9^Q(yLkaJ_}*?v%3JN4-S{bnE84ejXz
zJ&1aP9>lNu?4W+oL&yG*zDZjT?4$K!T|)b2Xq?WMozJsh_#_u}Jx_3-gZ(icd>9Wu
zfG&QdttWO%J9Y`5#>p>{4<sk@VTZKCmwxaetv-Dc&sn$Dmvv~p@YCW#dK%*x@veB-
zeP?lk^<$lhXT`bbi6`8f5&wz@yiX}!7w^dzPq^<ZPKU-n#GwTKE&=NuvJ2uX>y8~`
zC*-qJ;sN%J9kTxTA$|*5cj6xLi*?8k@^8*H(BuF3dHZSQ);D_g`{Hx!S-y40KCpY@
z8z|p+*%$k~$Or!QWB-Cb&kgKP@rQ4A*!a<7SNL=JkxzD394QVm4$tS=W%Ht6^u#ah
zBfhocPn~n%mpx{$(WM`d9+4+K!l!!l1R9q(7C-bV@<YG*C;m!5=;7Zy&?`tz%IS~u
z1@_*4)Og67{EQ=_@j(0-mwf!vNAb0|R@^PV;aBa0J@4j!`C0VHM_fQ|>;Obp+!4=5
zoufK0jp(_#b64fgQT2~+=ddvLtAEew)e{dXH-6~8LgWWu?*GQ~W4}M6+&MCO_SO0^
z4r%p_TfT8=Z(ijapK+lB@#EgJ_XnZhFEAeT=nX!l^-o{S3q7F+>Z>OXv2S26oqvdf
z#4+{}&X1g<h>M)Jh^L&-vA^s9v|f~3x7HK?N1xdv{!g6AKCx?_8>4TZ#ZEd0muAQK
zFZ)4ugFTne?y@KRi1<<5$ZxO{#?4Maegj?f@Xep<&v_kryJyCZ89)2Qk7@6mQC#dC
zSzJpW=z;N=XY(7M2l3zhwej$);s;1B_R0Juxj^~>wKHGrKRV`JK6}jXpbORGSILk6
z<e#O<mEN$+kiJ@X^p%~qkF{S@U;GByeQ}<&b%7o~!p}td);)U#oufKuHLvv8Jkr1J
z`t*Eu^nA%Yn}6$ye96oF(Zh(oALN{Zy`evm5A^wM_2>tElBRd;1U(UdusifYIsUC*
z<6sZeXICOx-}o1AdM<{4{i!eCxbUx>+{r;&e8ZmGH=Aec+qu5=9qqh!?2Y>DzWp>i
zWZs-FX^*b^toU;N#Ln2SOGDqUVz=<G93A~Y{ou#BfOCKPZye-+Z|x$IFSIV`8NE<0
z&5x*OKG`wzZ9bv>89mb<dh}5~dh{8cZvN={rC0P)J@nW;{uSCcuv6?4`y<~zhuw0Y
zfZt~KpnUe4y|oX(k2L$Mo^tK^1NHE)K04MnyMQ0#)J}cl!>@gWe$-Qru6>sN`3rJ^
z{J8luZ`z6LoUcZIz`lz|`I%Uc{}4~|SH>&;FdlYNT+6RPakBfk{F2ih^I@LI9UtUq
z-1H{Kx5g9iDI1@0xZgn!$<6%nZ}c8oSK>eG*tw%~a`6v3);T)#kY1R7a-bjd4AKMo
zK`!jOe#B$!vpA4njeN;B4s_HLC-N`kK>x%~?3Q@R`0!8O>g$Kx@C)_NFY=${uAIEl
zBS-NXePSogGx?c+NU!Lh^<|%dK7F(<#T9Y<<V4>gxuhXE(i3TVU|s50fB1HuPyW^q
zJ@b6Qc&%%3hxi74d|0panVh@*kG=<@op~oWbon>(vrhP7@ddpWkJ1DCB67hG{-J!x
zpLO{f^?&pjeR@UzV9ZZiXW}_}Y#icZag%fsioc<CLmt)xv>%W0i~Xm4cx>-{%(`M9
zjNAN}XY(25Opf9j^G*KZ3-T~7{NUg76Y&>)5PymD#C6gz#)s-b_k798cSXd-;y`u~
z-}IfGF)sGN`jbx{)|YjJj`axTTc6ggG`pd`^~fGXWcT<(@r>uF);~L_KXDvCue~(;
zq8-1lTz&m0*FV0DgMEz1-r*~@k97D;^FTkzO?~{!N1wekE^!Mv8J9FV&e4qDK3m*j
z{L<R%-@2kV=2bebH~Rp7S$rlgV3*i4_K3f?-sD5;(SE@Cm*%I@MNgc-u35j{L$p7$
zzN$ugWL)IL{)-DB`)1!{y^GhAkUe6joKuhkWPjOddPx7-2mXxSThB2qj!7z~m&#*(
z`T93L{qWz`5B{tx?TsU1ERXA!J(CZuJL@bhom5Z%#%I0oYxI^M;&&kX&rg^?`vB|0
zI)Ki*=#eyi5>K+P#vRk-8OOoC8Mk>+k3Eq#etJa@%r}1%{aK86tq<$Sx?#7_VTbrV
zaey?rlc%(KAy@r~5A_GplU7gv;xzTxQS+i-e2A;ew|FC>ctM=%{7Brv&OvsNy){01
zL4M>UZ9c>c_%|=~$@(OJ^~EFTNULu?@xi{+Z+?JXXGiI|{>f9EN8aWIinomepX`b8
z!$?OO-}+^LA^U4x;~#1V@uy$%>H6Ezb8TsSvya+Ae9}MtK>mMWL-R&H<j#K43wkVG
z;>XxScHBA;x3PQD>}ouR<WIz3;ydx1{+tIWH^215{*^tY-_{3xF+O}brxPDKm-9T>
zb7kMf@|@c9dC!BP=f|E8L(h|yqvyG~^SO_|=cc3f*>ZkIJLiS!>1U<SFE{1At1L1C
zwa4EIXHWgrC9}qi{Bo_|ec<Dt8#^NHJGjvEDE^6j%saW6Z}UW6&L^EaN<;a^^}a{G
z^3d*=jTs>a_}8Dk;eh>K`j33$SML5nyf;7|vA+J4N2E{ABk3i-PtT0gb9?K6-jW|W
z^3jo|uj1f{_!;79pC0+)Q=Oq4hI0739<WRHLCyo@^E2XV<<iR8d+XdfPC|Ch`eqNL
zm0Rb|8Q4jDi1WoM;w$F~>an}pL*q0a^u&++G!zG+r$6%R`p5r%{mh>oeP>7i`ay@>
z_-*`raIae*JAB-@5q6v(f$T0n<Q^IO>${NB>^c9YoqFz(q3gRa>LvMAew!cXuie)|
zmtWzpJRjpvwU71OvlB1y)9wcwH$L=-fBHkOw5MO_;|JgJ(T)8?zY^2Vqhj1A{*%vN
ziL2;^{gC?h8R}~<zIMJN-h|4f?R&)O_GR`#o-a5@voCT^Za?Ha$9_rLep1^0!}+>%
zH2L<U^5gRWzw5OBZ$G-&7k;{Q`{kG1dCHTwEY*JJ&+ol*=9^#A_FeIKf6Kne{Vw-2
z>`(2x$x$3jzT`<B<YQi)kCTh}63?+C<RR@GM1RgB#5L>yy8zh@_4UI}(I4j|>=1jX
zf93k87wngDI9H$_<YND!oqdV=(&Xd0gYyIE9Mrj!b42sv9E6^_&q6=Yr&sjNd_wzm
z^MD`vU_SMW4n5Z{)}s&f7hld*ilTn)<H?8KTF>NTe()WiyWzw6fc1}m>k$9sX}pl0
zJJ(gu{aEXb9Py_;JL0_{^5;M4H~%hvi1yol)jq-f2WkGren?t9Y4#nuPvLw~yb_<|
z@Q3`WeUp6=Kgn-H<@~Mo{4F}@M^vAG<1hJhbkyVT`7QPM+b9Qq(fN!tRG%NVF6a%t
zu&&v27|)fR3!)=Uuh5I@!MURINB7z26?<=;D`&UNvwaV}p;z{y;)7UE+#xO?7x4pq
zLYE%Ka{1_(Pxa76hyKR-Q4c-y1J$R`#&4YTTinhMiMOo>?Zi>i^g}<!gCFsf@zFzi
zOh3J+Ykr+SvoFq_>33W&=;@DK%m;bs58_LE>&CdqiGG-mI9}scE^WN%if^q$<A`Ye
zqhnmgp?`8lPuxpR&U@G)=UDa;(0=2>zdgOp!^bQ#V!g8K*0puXZ}2<TDSsm_<R6u5
zALAPF479%z*Ek;&pFn;-o==LK#Sh{VaZ2omzY_N;kNt`##6J;@m)&zu*Zp03?tNLm
zFXY~a=f?4US#}LQ{#-mIPVgK@xj2HI=6Bg!s2#r9VeMkR`9;q<HNWh7ymv!C=np#L
z7JP_Llp6;<!k>0fef9L?d|G@Wo)JIE7q57ph7bDy^NqfJl5$9o#69Xo{gI}BU40(C
zpGD8b>-a;DervB^aiVxnx$$VPp7mlK$v0l*@~tEM*&j%I{-&OE+~49$JN#Ss=vu$#
z(Kx#6bW~pWi*iNRIPe9HGs-{Kw+|;L^MXElaXjpZ`r4C=@o5+H#V_`A&X?@xoG;m5
zil^-Z?Hld$j5m(Yy0X6_XK0+}jsC%S4w9C}kNv551OMbgAMj1i%FR3ZlB0b3z#d4G
zD>*qYv@aoV{5hZFzvzQ`P)~dMYrdT)nqPj-x<C(~<j%gatJbgeO%B!vJvJ`-VZ8Xj
zuX<5W*j;=;agjJgd<4ab;z@ClxDvV#W8Wtph4P(?sPB1?eEfTkX<XfLzw&c$-tFlt
zJL;J~<*`1!gw9!DjLYJ=r?^raDLxaIi3_F0pV~ojpfoz#L;F8*r2VG255{t7{h%+c
z96kKXhxCFUrH9h=)OujQ>`S0|wf@bEd9$vpOY=bwp!tpKm3~C}`avIm?1BFH8)@|E
zGrIbhPjAI*?#+njd<W0_$bRqO#glhAWZx;{y5B$a`}R8?`@ZonuC~<3t=o0hyl3LF
ziAYcFckD0aTfg?Z<YykFqkPzjXurq}eft)6!T#C0&_6k*`HQ-i(|>*ivcv2>KR`d&
zH}Yp!*-i4X--hOsUDb}=wVv$j%_sj!-sZtP^FPwotM~=~#$!Fu1AZm0SN>Rghz>i;
z9`pC=(JOSV%Q!yqgYiP+vHp!yJi|_j*UhK-qj%0D&5Lse`|^n9$$E%(Tsga|J^ISE
zmsXyJ^a84{AM~VQ9G7}Yh(7((5B*Xe(R?)x=@t7<zvxN7F#6e;7RN*96XHnwaVXxi
z-<OX2eCG?0Uvr*-zWu*_IkbNk=R)fbzs{HBORJxR*0cKdHL*T<$rtB{m&9ex(VX9j
zyR2L1`r;<&dnV|j=RD9lfzB(fBYsnw{Sj~KA3sq4-}v(JufBTvxDof@#Ba_C)iW;i
z@TtCh{HU+oy*F|Yk1BURZmVDX^x@g}jU93Bh%UZix80lW_{)FJpVe{BDSpw8J>u6Q
z@?Z9+_O1L56wmNO;z0gLT*u#t|Fjo}@~iBEb9?#xCI8RP^8fsta&aTO>>j1(WBj9d
zf!|jjU3z0Y-F@<zZ+Z7GHov8Vf932py%Wcw%ii&)<f0#XBJDX!jCZU9dMLipUfe-0
z_%Q$YG;Z;Sagdk(#h>^fXY<3Zt8bi8{2`4_tnb{^`AVdVU;AA9fAq=K{6qR}zb{Ur
zx7yR!m=?!Fdai$ZKrZM?(`#s6p!QI`yl6cdukqjm-}*Hl^2LYzm~#*EpjYD3y!`cA
z-~8fNJer2$3iB7`!7iCk@hv-O{Opi)JU1jydTo6`eCtnpdd?oPKh~rEj9*%N{X%j^
zSDYxVUDVSk7y76_e3O$nK-?hi@%+p4N9$UB`#gxQ{>&5oh4>>+`S{abenj=OCl~UP
zubuiZtzP60J^AQB{TMGfK;wb(BbrC`p!+YL`+MGGp2TmS>zEJczw9u(f^YrfpFTOy
zl`qZS@H70c{V#n=L(iR{_yjsHvk#?jN#jFDJ$4g4&yCQNPd?&gsGjFp>dR+OjGsNV
zU*@0WtH%zrOX}m#IP_yY^39|E^`kvLjbFS+p7JB&BcgsFf1$nS5YomcjbC)}3(+Mf
z^yEY3(&$0@PR{htdStKZebi_6**cH*fxcUh{43;_p!qV7#zj8(Bv0`uzM*>j6}dw5
zN>Ayt=eOpceDI|nBu8{1yUGvRr}1mzaB;Eyi~UyI2Z?jli|2RFN5sF<MUnsV{3!0b
zod>vo;2fHL<?rm<q5ZUT1N(mWM(qFPyBDB6d+lC>{Ky}9FX(7~?N96*vyNN)SU)L0
z)=Nv<XC>9cuX@tz%a=~V*e=#%2iYO^%0AV;iv1F=Sl`gPkH~+B10nmN9Li6tN8har
z`YvA@^26+(xQHDWm$2{R7UPBLDQEZ5V^^X6$eDfSuh~W8cP=NN-6a?DRUYZ`XZ)mo
z%#(KPIDaAz5+6i4vP<q0v4853uX%&=^@G0W?dn1HN`3XDmFri&dBzvL&_DXlPtnm%
z8eM7Yz`EhL%@==ZUc@2v$o#9<^~0m@gQJH}@w7M^YH$9e&A)oqujd==1av;;{EvO$
zAC1?#M~~mOPn6G(YiFN{u6*&3cqyJ^=}&$9M#R78KF)je=ls&SmUsZ)`en!Q8<D-!
z4#smidO`2ZCp+PsNt#|j`RJ09e0ByM@-biJ@4l(~sLs*I)p*RC@v?tONMG3%`W^GF
zL-xu!8nmv|qwmmqi*{2#=vYtqz?XQK-Q&-!56J#nzwDrP<Zd45D|+N){-Ao&<V^06
z9$8P&e6dsPBzciv+@F)1c2Jz?Ig@;L5k2+~I^WaI^C)qz_llG|M{u5_K2$H3ODCaz
z(9sX{{*tu!6JVr^4t}Jed`QmZC=ETI(~owMp0xUrp8Wtc4=}dZ4^$rcz-JPge~3T&
z6zS=Y9+0!N`udCc=1V>IbktLBJmf)M>iM2CebygDS3P{|7aeKsp#GqI{E(OVf%t*e
zt+;@_5r@TnGyB7yX{Wxla^u3E^<q8PcS85z-G_2t-tWKew!)#)?r4u|pLE<q%g=dW
zY!|%;s=e{*U%%uEm6IEJn_u<ti$27s{v$@d?YE7~JdryjM{<Sa%1`h+{EdA6Lw)gB
zjL)U%0mKKSFA>Q(3F)Wu*q&ciPER75C-V}~{E??|NaF(?cEr3wdMHgF=z;e5Ll+u{
zaTzZ@V*6NMJ^nA!MGvZvj`~o3MD58bVj5lJhjBin^=mxRNk~2s^<#f$z9U`Z!ar1B
zx$%kvtvBWDp7=(4c7%Sg<8fVE-|U2P_CB_YdZ%2z_k!4$dAGjpphu2dq|F|_^^Tbr
zuCf2Pw%?K1<FxyyKYPTY-E#H(uCg?guf4SLr12ORzF@3RUJ<P?`U1_rwDqTb?AQ6T
zb?1D7{bJ{oJO7OHt)6`=`w2a#l}<x?MxXHGyg>i>La%Gj{XTB^E*m?~uA>{Dhv>&T
zVHfB<{)`tL>w!F=`#bh=-W#-^Q$FQ_*G~HC1>?Hkp|WqZzp(Fcug8AaeJSUm_QCd%
z_Jf`yd*1B62fF&PA9Y@DUnFhc8uRUsl-p<N-*W)_YRIq9+a&##57`xZ9}zvMpI9#`
zt)IA#*ca<cxi~krQx0Q0@wWVk>>7;a=tAZAOhWab_Ru&WIg-12Sux57y@=+6|B{ZV
z+`d*CTDM7l(|(A*i^$GEe$9H1X#GR$Up;B|BO*Hn*(?0xSNs}LeW={|VLaz>9%B7V
zJMT~*q7&)HeC3clAUQ}!R9+O><w!TCi_-6RJAB7}<-=HBmo&fZxwYro_FGA)9({)D
zi!0<iM|8i$d98DIdhEFlKgEwJho1YO<NOlRZ~Bi8jQz@o;zZwN5=Y91F)n1!*<Joa
zTKudYzafqmZ>ncME*<-gY5dryM-*4GZ_)nI%eX(6HlO4h<=rixdB#m|`P$trb{_H*
z>^45wRsFE{%Gp_HUY*}aJCBy8CsA*t^{2jjQQ8@&aXGgpH|^=GI4;_4_e;>TPVg_^
zx)BG9i`>sf*K<*6{L*XVu&&W1XY+=xw0hz~aiVpBzBKf_1A6`=j`Tc992$`vO2^TC
z20qhJyh2aut^EdcZ$!Dc#(rMDdh}VGLa(8G>w<om4|du-v!nF3YmdYC7UZKxkEQ7^
zz3$52{%mx=*5zX~UwifSgO2A!)(L&qPTF{5JMHCbhhKI|JRzTcLwvCdp3C#g?pvWJ
zZCvIhw#UEuGfrqds|WEL`^P_Ir=fn(F+Ox4zR4@89zM}E5A>9sbxvemK>Dq|G?WkZ
zLth~|MWm<ljUz4JxYg56zV#pDX>qmZyb(PwWN+=`<wMUI?bjo+hY%fzjx@wiOtY)#
zdA^>6N%iEr|0nIazkKH!v0wGolZN<F4wdW2IH7%&^Css!^qqcLr|bp&Rv%hF{2l)o
z{UE)A%GqD>13OHwlxs)t;<}WtfAz)5`0Lu;Q9CT&)sJ;VfAN7o`TVVP^y}gibR(h%
z^@C4*NaGWdllTEY_>jh*^$EpQFxHD{?I3@`zS0x=!~VIa#1F97=F2$I@qVX!D866i
zJVm?^`{9?+V;|`meRuAQKlYfuYbR}-`n7(b`4dlw^Vl77G{5Snp?!<}j{f|Pu=gF_
zvdu~7zGVHeZT-`G@*rpHoBf8?uXADN%+7_$ja;C6cJ9&fQ_juJ4|>YU+q^novtLWn
zck(087_XDN{VqRY9_YFCVt;|2b;ZwFN6w|uwXPzfkDl{p_nM^L^Agv)@5Nqw&)4}e
zyDbg1=Qq%G4vwC4a{2g)bjZnhIlsvNDu?Lc-}v}*@}P&-H@%~m<i$Q(7vxXQ)-}6l
zKc>EQPX8f0tiE}~Cp|H4a&`W%TswTyM{?zNlh8at?Y!4&p3Dn*;)C4f^Y`Kh_ax<;
zfBm5^Zjd&A5dY}7U+EsAe$ZDh=4%(}CAIV3ZbbV8{+|C}m&76R)uU(Zw=|@W^pL)Y
zANVVFmmL$w%eNn87wv~4eSD$IenNgpoMk-H=<~DI6%^MPAGxvX{Hgxwh53=D2NA6c
z^&!9LT*dcUJ?HSg2s^J_8hY;H`$wMNI;V3U;2cDM_<;1=`hwP(amt7Cq4A@ehU_If
z#cuISo-^_%>?0JXYR|5jZ~j}FUeil@pkH<h9q4@<{*K)f2e51Iak9hwCjV$${DX0^
zv*JYg#;KhDme0QGPd~=R?i;st53Mif4&n`RXUEA0J@JP1&mNGMxXS!km)5y_^0WT%
zVSPh(L3__5&9C_pf8i&#Lr*`}we>EJW>?q;<A(euIgqb;qL1`RxpCr;J=70=*f0FE
zOHe=Tu6p{RkLsbXzH#8An;!js1^+Lf9vhGG(I0f26Gfy)^d4F-QSX&QdM{1?tta&P
zd-|)sa`sqz_4FU>p^uJrDZUg}uw&>;vy<Wz{7JLB><NCPjT^uC!Kbu!#14on#2M~`
zh%eZ6{Yx7s{*5ojALwR9@fi8&C;AQY(LWU58wV7}8He$ShrA~u4pXijIiY7f;ym-G
zp6>)n^KZss-i+6Hr0I?J_=EECr$6$C`i11AANqxV_kVnch<v1>?-+TXmfWO$$I17J
zd{@YOx!T9~b@k^vN4{%hK8)YE$w@pxzR-M|C-WD-e}ixQL_XAy_0Yo~{WCw(5FgU^
zdE_r`AJ5O)@AF^IwZ)y{WalpYlQe%P&W-uvedYWby5jO!Pe1DONBkx~!~gP2%K0&V
zpMT;n*eiYmDu>#kD{e6!s6XQ%7yYBh&x-?`R~a89M|_Gqj90#K<6l0+mo$0eGtx6&
z^`-f3ajf<b-{helzQkkZ4Sn(<U-a~gKlO}<e8n^5svdgy7DtFnyL^p4$3ura^)KCx
zw`M(b%fwA5x0IXL*e`neGj95WKlRXu;v4afcopB|r@ngPVSI_l$sgb7DAyl)P<+|N
z(ea~?5A!6RG#~VsUeI^*OuqOtPJG4u$d`COo=1^`af(O9C*n(Ssqv8mInf*Y56G_a
z1N;KNC(dxbXaCDiv&YKCBicuF?j=6spYS10g5ooAFgtI4q`Ujj(eG*7Uy{4}=8s&=
zyR`Xa*VqGkVSSLV`EZ__hT>yr9?&s=^hR1aH1APg=#hH%8?iol7De-D-p!+VlaBgC
zAGCwwJ$j{{eT{KJ<EBUAdj0EHTqhnC7m1U^D^UFAdnggbY4W8bDi<$`_k6EYoTnZX
z*NKnt3+0oOI7xhGeOgcAS#mKi<SVV7eUSZ<{7Bcl(0B62r}H%LeR#g&-kj%Y($4AF
zRrV7<)>{%f$8)~t98bTV_lOtS1Lue0H|KofN9TO_^!&^BVf3q>ar-{H-_`N^QOf0`
zE8n@L@6|Y;`S3Ac-1?5Y=C%A@kMS9==XCfp4t%-)B#j^Q;Ji~m?nmm+y+-#ZeeWIO
z;|p)t^yy`9TD%Rt7i4@B&RX@D3pbz8_S}yD<>$m1?6&!qhWs7-OmEDmdzRuXc3wQk
zj*Caxe|F!zK>C1ha&taxKIkzyniunFezaGvALZ_o8i#(QJ)gby=6$w2`>e5T=)NfZ
zB7gJfdF=@|PMP@G&X~4&)gQj0^@BdT`tdta-p3ld{c(pacGvj!Pu86G#z~h<XrKA3
zPptE$7nf+$lOOH&hu2?q;^N)kN%cFbo4sx0>zCPQ$@W*@KIg@4c37(0-tVbuPyhb3
z`_Z2qchH1x{eAYn{)LfuEZJ^-ami~A*>;KcyJzoy<LM_a)xP_w$98_r@0Muad(q0f
z{qm6|yL8d}+%7MB>ad?L_D}Djd4J9RQJg_P!~-!d6fZ*c#L4s!itD4EDyNsy){A!1
z^xAsRzu!k=|L8G2vwu}zTyK47$6k8AuD<<%_@CXzxAkkC;hX=qp2er#es}%+JN7?l
zq=hc~!hV|<{5TJw*VY&N$-dEJ^P7b1B)f$kd6;+OH%{XcSMj^flf<*)YVk9FEPm#9
z`9JZI{Fs(6E)!phi|{9&6nC<t;!Sqg{=_`eGxI0Dw(k*F(_{TPH<v~~@}WQZ=xe8(
zo<zRXH;y<ye0Rq?djAT2@}j5E`nCTu-<}sa$ByTd()MZQ(fpV%`!46k^dO@8F|HGr
ziNDm7PyZr4{LnZ07*YHjf6s+ntXJ}3r`RF(#Cmm~nxAG@tSfN~d#k;BT+Y*!+xJ<I
z^64W!r1b~!CGN8B*%@&mzQud&DEq3tI4h1vyd@4`uk=G6>{%R_diaUu@!V5deSC>u
z^pB3Ta`T{H<5my9((1|AFZ$+>-=HV_mibd(n!m*ly5c?aNH5H@ex&gQ$r*k9@zeC1
zopCMz)iVy~c(I-J5!<P!Ur1lch213=<3=y)yEJ~(N1r^{H}s8D|M*sJ9qJcd^q}_a
zE&t9wpvP~J139YC&N~kjSD-5{;7|Am`Ru%W<5O-t_%SYgsK*~gI_Tq9JL8u&KKf!_
z^e0Xz3;7*!1^aKG&tKW!J5P!J6TjpW$H~61gX-}=`Za%0KhCH0Prl~Ud^^92^0toY
z5sdR}9<`^3Ft#&)))Bj6{jhWPsrrNFU7SEJ)|>IGXZ+$8e4!_;oqQPe*gP2*KGfGQ
ze(0yTO&VS68Xv|@PV5AG0NGb|i~Kydf#eL;hwQt3ihBI3eEjf7{EhL@XL4X)`3d~Q
ze$Bt}@E6)gd!(G+nGf>Ahj!#lj`)Mh=~2>rs&9Vv1C_Ho=(0!J;e$M(cG8d>^dq0#
zwTt~kdDB1hVII2c{J6CbTj9$2E$1N49h`$Wk8|GPe!6{|a~b{Ohdj-X_?o>GpNoT~
z)f2aiw_~|_@;x^fXS3JR=&{q{b9U8vx&4asBjaO_#c}MFxQSe?E9=tyTi@&@I}F)p
z$gW3Bs;^yleT=@}NI&TxyA$<Ez1Uuw9@0DGG=6CP@jJ$ij<n|{;t~6Mb~~oU3)a1L
zi!OcBKfP0K9`J3x=q>r!&(RO(0iK80pGI^Z!2i=D{o5Bh-{*hr<L#@_b)MllfHXez
zLr(lYeW4HNE4MD7^<muBlXU^@gN;+#dUhX0d}LjFj)Fe9*smKGv_9zvzR8CiBmel+
zPekqM4}E5@orCc+{H=8X%_liW{WU)Q(Pwdn`Ld4bDHN~JV~CD*#=cn3)+MyQA^xo!
z>w!Os7~5$d+mWO8((36Cil4+u>;ZjcAEhIb4|$P~c~LHnZ}W^#a>H-jC(s}3-@3Iv
zts8zLo<q`8ek$rGzXs8fMxUNrANXJgtRs47{ix55;SWFd5fSyLALBP3_DDbMkhFe0
zH)RLd6-d6u3DLtpWDnRM`4Q1E4&yftX&CFtN0)r{Yh6S7;JlMvfY!VE(0JHq{YKQU
zak01Njhx-PgzT=k2Orv*C-Y{UQLl`PKG0M8Y`yVk`Y{gktUh^1Bya7^m;U$_<Kchk
za~!wl_@0}4zVE)MbtZ0%_ZHk&aR1W1(TMI9$k)!j1!<@sd_nw4n?L#1k@=T4e);r-
zoZQbaPWLhJZG7^{2cPm|T05v7IT(lg7y5_z$DjJ*vn2n)uJb4SgZ0Kf(QkT3AB>Nl
z(a)%7^0nuG*+KpBXVwAackmlgJ#n&mjPpY7^7YSex$j~<@TdHodgP5BeKbGPYh3^N
z9VShFvEzPCq))%d2R(W~@7XQ$WIp+8`%!U>dFPisul8J?{P=xoC|-f|g8h;vcWH=@
za{a{eZoRO-B5(6<9QcW-+;dRqzCwKN;W>lzdKl};w?Bn3ZY1}J=0QF_p?v2>(#nlf
zT0Mx5q_lI1SWo>Vq+jH3zi1vIKWAUTUhv2CmY<~W?1gsxzV_^c=aMmvE<_Lg7{9SQ
z5I_7TJHRjV<LK%~oJ}s|AkL=m<OkU`=TP(;+ApiGo__2@*h&3C{NWcGukk_lT-rQJ
zv)|?eJ@W*`8RAU!pz{BZy*rP#tE%z^oCE?$|DcShAYucClA=ut6mpJD2?|s+fJS+O
zN+Z}|1B#;HKs1Pe;)}|&0VPiqc@?A)p(N}JK2ak`D@1yTKtu!;4Kf7t2E+4h2NQPA
zJ?EbG-uX{vRyWsNzcstH_CEXC5B=Z_f9>Qjukn#uuX=pR&o9y+^ROr5$syj$leF{m
z;;MKjzN!~*#0~h$N6imc`GooPw_o_lN6lk?RJ_oxzjJ}S3_tbC`dg<wN&e^eFz|sl
zImIpf@G(B}Grx9t;m>~UBfQKbZ$*taFC3KRd#HBnCWrm8F75cB#v7*``KzZV@(uZ#
z-+5`bkLvZq$9<4_jMI)kKE~^(J?^LTiF1bgZDn?)%>Rhn{2@Q4onF&Fc*z^-f&Hde
zsD0pv`3F=yFrU1HpJm77a85MN{U(1PualoRfAO#Q@@wQ`-{gX)`B456UviM&{i!%l
zF83_*0(qW1S6s$dzUny;eki{9;m=>nFU2wOM|^VMr#;@oxxW+FQR|O>oxg{l^@ywT
zC)7UEH<VoX@E`mJ`=-zOnP0te^ba+jeD;sNnjc=`I2v(<pFAnz2)9V-t?^MFdV%VP
z!j(MqT3J8y&`)J{kK(5tjedb&p}*`+o<lE{;YdF7*?;RcKU%IMoG0+p9{C#2-tni;
z%I1%L0#30Vp7<MQzxAh2{DpnzU-TmfKFav<1N4gBq3n;IM@rAtvlDW$7w!BvJK@*G
zW%i*TI}`8Chr-`_*p+qiL;RKJP4tQ%^&Xt(tJ?WN)ca!2dHkk)8$bQk^E>p4Jp6Cf
zzH7Ii_MQINS7qzBuk->f`_I3B+A4>hSBrYB{eh3Jf8wi)sU!Ad*eHFYKX9v}@Qsun
z!IhrT>!|PY9#lKN@TCV)Kj;n0&*^Wxbws}O2qnLMsCxXAQT6(x^h|s7Q?Z{id5u@!
zZM~ugotnqKqod_Zj#kHwmcM@HQKo-Y<=9`pNXgBfVtcc)@y+_h_9%a=?XiEa%ltsI
zdErw<<ssr7JM|o!f0qyO8~h8(FUXJdL)j%t9{!KLYiHM}{^B#L9v=J$$}d=#vU&JF
za--%&t&f~=(;t8RBGu1$@}cC>PuaTgw;$Gl!hv3^CqKU4f6$-ZnMXhJlMi3xB7g1p
z*dJwinQ_(^`CA{kQT+8Mr!tB!KFZ{<E;NoapZjX-aNg$E?4$gPo}<<&Uxl;r?vvaP
z$V*Z8Jnl*Gp$GWMqva3m);i>CaDb!xNcH4}hkp1g!^1re8t;4J{Sf)c)r`UszVa6N
zv*!i&N#3li-TSECbM^c|K8>IG_%rkA=iDzZ_ue4B-kbFMIlrrm&vBIX*KgUpKi)le
z^~Y+S6M0Y3b29I7s#o@VJijmX_fWom+xs^?bI+;uE9YMKpFe$QmGbwles6m1om=m7
z&&<{8%6^aP?+l`T*X;M7`o(eYy!qVU9(>@c<=<~LPJeQF57qBy{XW+3485P~@4*`9
z_u+mY?{}@J`QQK-zhhLdY+g8;U%h?M-|u>z<Lt9_=<jz++MVn0#pfM!_FlT%CR6J6
z-E+YB`zXJE!pHrkdw2J$?9csdypMH{Tt)Sd<K1ubTl@}x$Zw#YAL#G=@BV;)<45p`
zeny<($Bj3?c;cQD<=35q**8DpJ!bbn?A`C!+((M<@)LfN-y|QuNUoTVx%ahB>(|e{
z4nNH=!be`_9!<MA0C)bCy!h%b{<$wEFaFje&Y<qm^h4nbr&tyrQE`l((qqpVyl3Kh
z0jl2n4C>>14ZgGH{RZ<W(_{R+4?(VIN4^uM-g^-E(m#CJkGLl;qWD;^_bIGPnS7}A
z8t*=fUb$CB=|kLi_Ct@^6{?*cvRitfp8nH&?evPC=||7$Jv)UTKAv;(AM_DF_M;#E
z%I+2U9pgL?<oDpn&uRBO$o{FPr*K!VpL`m156G{`D^TYR{pA<#p;7*czcL>^qaV=@
z=#lk$?nQ6u2|VqyeYpPr>^JxDx2#@A{Z2%j6VH{!J#le|uWk0j3r}0MF5b#ljMLxu
z6qLOuBYuj9@;bj?5+C(9-h1BW^BoM|Px;`&uRXf$v^DGUEx&hS5A4M`*Yl{jpUUz|
z_q*~z`K35gr3dm;`z)_iFCL-nhyK||_uMEwr6<l2`qOjv%8uv*dqvrk{%Ev2{zF`$
zPs;l97vdX#z+TDWT!b3WFOUm0zkd9n{S^<LZ^SWil--F}?)Sx0{lr;*lYfqWJldsr
zq8=3w#EEF<?AX40USWUW<2)Jl0G`Hq-@|)H`fKO!_({|{`5p4AN9n2kK;`G`24x@Y
z1?BhHhu?4TquSXYyJ6SPWzJiE-^HIeCpy>R=iDb>bk9J4=ruj2XY|VXTN#xn$xrds
zUg~4xJB#Y+uir7inV#4O@k~BLujm;nPr}DJg5QXgKG;|BnVr!u)Va;y-Q;hb>t^o%
z`v3lq%cs`kd6ECNFYscIaDx-Ogp>a4RK9FHd-hx^o-4^Gu8A-7%XwW~BB$TedLNS?
zp+Dxg-_9fQW&VsmbPnPV<u!1(etKaa_yd04IDP?t{#HFb=GVkQW%IxpwI9}ndfu*I
zTtE4Z*KYqJ)$SZ(|MY{i_ZyV?5&lKJekgxJKkzlL=cLNkiOT2bBR^tZc^=BYXqRWe
z4PWcTA0;0=V%hlEpM2t$xNrZf_MPA3*X)~n3HfUk-sFQL3UBKqpLy($@zz!9Pvd(;
z^bExZg+r7N?(9{+7)O-lb>fFSQy%NPIOdTj+IRbH-Qt=!C|>a=_?VyH5wGA*F8%S7
zFY`<Agd=%Sd9-!$fBc@Z{9L|EF7v6k9^buCh99}$X5H#h?aKTye#$6(Q2qF8`2jiQ
z-EbtIdC3*mfgftTvUySR$Fh3!7>~L)be?i=s2+8H8Sjh5HTR+7v+o+ZSH)M{HlH}@
zT!+Sf?n&Ky%KMx@<SD2;(s@K4A}`S{Z}EE-W$nHX>pn*wF3)jqqh3BMPdaLg)t>e2
zV<(OH-mEgY^pk(fyHNR&GP#_$tyft)xy|RE%K04*>h-T`*U!0+T;@Zamm-yipw4Z|
z;sokGNIY?mAaBvmFZ28IW${?t<A?Y~aXFUx&oXW_zT@tEz`ofN`|`a9{*0dTckbcU
zyU$S8Pdz^TPV|5L2s`GN$tgbY+x$7ofALTF@we86I^S5g{_6EZ$<1$>$Nb`>xF~Ox
zzt|W3?5B8&+E?p0zc?hXlCRSf{KP-=il_9zeT}??{>g{wG3xw=(r^9c+3r_ky>|8N
z$az}+jk+Ipev*fqpFLGkdA$CpbAt0fImks$a`0dLn%_f^uS$OR1<r-a@!mjLKYkxo
zZytD9H$Nf|RfZR8z0RBXXh*d>=fr;M^>?pDAKZtaaKs0NGd{82zM%BbJkD?YCp?^E
z?X&Ur*ZNTP^p;%YF%Lc{eS$mx?EGgRB6WUH&)>&&Xpj8O19xR|%HN}(cOK<0#253h
zf7HCFdC0FG-o~NMALL?>&LMCl2m6S52K{#*AN9(7@)&W0-xDX?8>r`hW0_yX7mnf&
z{_<e?JwDD0)=3}WRfV_twA&~CnZB_D`%jPQuXTzKF<z)=FZhwqInw!+-q?To1t;-V
ze1fa|L>v?!<=tz(XOB}}`Oat5m%Q|n{l56Er&4)Wr1C;}qx?|b=RS`fxzCd)8z(=d
zU-CZV@HJjNKJrEN`ZuHQBkec**@=4l!)~G;spseTMf~)qPwZA1Wxw(Vc8k(4agZJI
z>*64P$$!#E{+*x2kG$k3w{^pTob*pUzlo3WC_bohaG}TS!Ft&p{bv`kA3NtC*r|EQ
zjaolSZWMp^Tk;LRgCVDW##xVgb}z1=>=`cB>D<YWMQY#dC!G0l?bbuzyieshhWl&h
zZGSJo{jtAq;M{=5zaO9<>KyGoA^lPFILD*T|GwvkpZA8G8~A(Y7td4O{}}JNFaD@=
zzvsE`58U&+SJ1A%dhc7A&v}Jk6`$-Ie=J_PcXXcce8)bx=Wu`F{)3#>;U43!yWjYc
z#jl%Gw{O+~Pk+||AN}EkA9?Aod_Y;8frogDuW~#u(YMpSb@~ftd|<V@b%;OCan4E3
zMf8oF*2iye^xl8@!LHAlGIGyW#~u5Xy{C-CdxfQk?sLo<+fE+Q@B9ZJo_*iRQ%3fD
z|MmBOW|dV&+&35}Z$Z`LBmRg};-Pl&M*MScBhHCa`Wf$DM_C+G7H5su&i?reRJ`Y(
z_$BYF$lI(#oF=b(C+i@eb*T3}9ryU|k=+l<N8}yuHQXDxuW)X6U&5c$KXFk$WIfu&
zBY7Tw%fE_G>W#x!Tu~N(jTbji`e{Dn-6v=#C%)_g9_4x(e>a%@xi>Igybw?1b>!ja
z<t^mkPxwvwo%c=oG5f}^dj6vfKld!;B%k&14>8Zz&QIe*53CpN{JnnU<Ja6@@ee56
zwY%@;SDfRW^VyF)fPKiL<XO%`>;~T2QF)hqOa7%Em9ItWeP!=0@t59PM%|0SQ(Wce
z#a;dqKH?2r`Ca#laQ7Tr{KelrB8sp55s%?79vDv_$tO?X-|<EDGY%i^C^^h;9jJa0
zAN=JB`ibkt!<jw6UtCxAyoJ8G=VSNs1NiHwo!#<-sCb8x%X;mXJOU+`dF(eiw98BI
zv+r=FNAfoLgMM&@pY`z*sCmRCaZ3D<x1;={_`r|RD|r{b@EI$WPm8DggL^yvKA!ud
zKaq#a@AY?oz#pLI!H55F-|4)^5Ai$hBV&Hx9Iak|?_=?Y@(J^ho1fI*eXG0yKks9Z
ziyu`cAActf^P}*9k2oT}!5f7qy!3P4Q;$D<$;HmdfgegA@v}bXD}2R&<CN*IdjXVQ
z>8IYh;R844Hut>r(K%fnAP&pZOWJs@WIlP1yvF%jf9=+3K6ui9l-%&ApZLRD{FFz?
z@8oUeIvRP9=PvRjc_jZuZvKG(6z};1eEDg9gkR+cJr`1MUjB)^<U*~RpXcX{vp#x+
z(hqr2#9tZZ2jzeKHa(X2kRM<BZXe~5^p0NhGx8Su1ZR37KXXq3U-DU(dmT8^2YBH}
ze)t>5F2xD!GY`J}zxW1!d5-b+M_i-_@=tnTJbiLr!5?4xsDC`~(--??AFRvz<-PU|
z9^@jobyz=s#_^Zx`6GCnS3CWn7xu@#(^K-%XX8-kJkMw8CHeIue@PqfJ;BL&K_1Co
z<zaAkE};+fMj0M(8*Lt7etyHc<i+w0@3DJM&fm!U<YmUmpHTPQ;<9)v4x`%j7r&j;
z^{ZMRJG3wK7q0R%=Wpw>f9wN}_8+xx^oBnW$Hh5uMEnvr=%;b^jel2Wm-d<dyN|VR
z<kXMfMD-U>`5FGL)YC@1B$syQYE(V@(jOmwpPgH;xGldDSNUJ@GUfqr)DNDhcH`tb
z@)UmFeC8)ltT)bnMY)twaRXIvym-T}&>Q|I&aZ47y&{*geT;b_+{HWVgdd#bEArS_
zw!ik%Jj&!JC;gPa^S9<zPp*gq{e%-7?4Nu~nO?yefB2J&Uc!ak#=}QH@q!(bgM9e#
z=Wro6iZ8pNrzpPmll=DCKH<-=k&nE_M@kO$aJMe&C!cy{?Qmx|<YH&WnMXhC(cig=
zKX;$(p4a<<?psms3%Z{bH_StCozKJ%W%nHNC4N($!{0`#o&V(b?YsPspLGt2=O_4;
z`xyQXt9o|DpW0{C^APtG;;Q=(@Ab;_oDbv=)-7HrTc<oqJ|WLn#z&k%?E|~B-_F(U
z``w>MsvmuD?iT0W|3^Ob#&Zn$0Y3D}^9=Ro!_V)!JnyK=E5=$cp34vA*QkAF*Zi;h
z2Y9ffD$2gBkNsG`=NItOFYb%`anwDeGV1<Pp6I^Jb1~nq&@PX4|H(e(+x!l^<Zb*p
zzvDiV{n0=A$p0EIuS4az?124yJ}J)N>l`m$@-ybKKk(E}PVpK)a*_is?h)Cy=acS#
z;i{hf!hxUX|LKvq$6xbH`k|hS%jfasr|7Rd-@5RVAKIrF|KX1>zoIPuh$r}qi|&Wa
zCyv6E9XZeOQ}TZE!9AW|<l*R8?LLkdKlpo;K3IqM_EG!dIf&=qo|~Yan|O|*|2Gf%
z<%|E{{MGC4+kDo;8(#fS`ttk!Yvh1utWp2fAt&v-%F%1o{ob0M&~N)6sr})%%KWHt
zo)8z^Td;HS6cuO09s5S#oa0dU5GX&YY=7B5eRi%EN7)t1f4T4XoZj;{{P+pyVCQFk
zOPuvQh<wIJ8s*|I@uk1)o!??l=8gJRRpwvJYd-!*y>m1_MsD{2o_}~QQZ=9aLLN(x
z_#JxY9L3%vWheT}r}=C3_@ewD|A>;89nxcd%sB<s&cE=l`02+Fq3k|VcEJv`TetP`
zb0{463w}{vAdgbdAM3B(^9l9xcJ1n|lYOK5lUG@sG@f5~k1Ec4-mD)!{6D>B*X)Wt
z@DFfA;V7=_=blHN?EXi-%umF!c*ib`!;d|fho2Hh*fIIZ#h$IxeJcN_?EDhP$FhFz
zYf<;B&L?m(-u&d@hf#i;pLV~?p0zs{@k{)pIBOpL@iQ+<K4tyzx4-V?_!sN6F8jb=
ziVN!P7k_SF?Z5rA4tj{6eT0+qqI&XJw{iBB9H@1}ReWWC;xeipbx%=6`A5|G+4$%O
zOFP~8-7h`p;U5?5(>a`+`mulWlaqgTZ^wV3@@wZ1{hU+cd5_<6KC%vePJT!);7$MS
zw|ii}pOi1jujEtm0{SD5@*O_kMQ~21C*~JN#RGm&-Y@^9-}IUu(GT*=tHdK^af!cm
z4=2v?5A@u74B|h#G!HrLvp8iP^hKO?PNNU@ojmkT**In6)zdd|&v`<gK;QIJrdRk@
z>9>1J_b%d^I4%!B;aEjW`wjOz?7=!v_5m+)z@J{&2mXM3_8YZc`4wv4;40qA55#+M
z8@}Wt7kTide<;69U+ACy@I~nledj;5qv8#JO;5GQ{7(L$o&QDoVf^G1><kr0Q0GVE
z@zc)EQTH73YUd95wDIiRx+0G3o85^w<cQRM&=XX>{ZzKE^c=N+=0~-|*FBJY2w&$`
z^4eePz>lAVBV72=xIgTLUg3i;KMEK5lrp^K7tS^I7arD6pPy3S_+12hm8Y0b9%Nqa
z<TM{S^h4>Pex-gi&ex?}jk5hf;bK2xe|`a9a-sAv@{9OskJNnTw_p4asvhO9mFX{e
zOZmb*gLdm7KRIHVzlIa}^{?{t>Z|4<2fxODmGd|BrmSzo3plY)c1G{WsjMG)=^_0`
z`B{FFUXvdVD0_r6{KX&qP;%=BNAZ=s<VW?x7a!x)(+BI+Ph7-L88t7t%KC=hzz-h!
zm;4)fZ`4Em$*Dg%_%VJ@JNb>*U%fJ1$c?IpJNrT5L|*H`-*}YYp$GaIr=NOd^2B-h
zE$esg<-hrJevyB3PUZJe{P;g*_C>!v2NQ4IkNZ85`#tA(@mRgGdolU2`v7I}+kJpI
zF77GIuRWKG^LXyT-uP4d4?p?}cln+BCVF5W`4@54c}4!|97HeqC-c*1a>&Q6OI#xl
z|3#naoqH>JxxAko@-yx93x9e{FZnI<!iS%tulN}UKYD20@Uc#~X-7*w4L_jX`tjkX
z=q;R`|Lqq!$t`c^=jGel&F>z<{QSIqr$_t@e4T@x!`!#gBj;%6CG+A#ZhB^Zb|*gQ
zCy%mjc1j-WF-|*N-~eCeIqN{3pPW<dJ3Ur!ymK}^(vLpD9X|BJy5Y@V8Bd?BAAj<|
z-Ms|4oMWxi?<=E!q(}5v{>Y!&FZ<69oonc$eA)c&SKt%vRoVLRC5QET-j9!cv2N!~
z)V#5u{&2(3IL}v|Pn}oY`_M0P*-!Rm{nn-3`n*S^tlhrZSAL1U@@wQKr*k!Y*q3<8
zufhRUPw)6s>#(2fj34B`*cbhzFY3if_thwU((bucj9cQMe%@OZcikt6o8qnh{06zj
z8~SS<;tD&|j`BnF-#Ao1>!QE*O}v38ySG0m{OK3J$&TzRy^8v2|M@NP!usVso=32E
z_s!xVdsa3dy^Q{vzSCFwgz`J^zy8g|^jkLAa0E4u9hi@vS~vS(KkQXMG}<?NF%O)r
zgZ%6PwO)3vzx5bLZe^6;wV&j(-*Ftd)WZQ)4+s3{i8x0e#XI}YPUN2`eefJY{v!TF
zO3&q?#<N#^_zn4-{@U5G_1Pc(hksNLANbi9=L7a8Zm>gkjk0HPmVFv$f9<z+dL&-h
z2l{EB?Ng+1uzq<xoaE2)FZ&=)vtRq6-uOt3cRt{M-527=53qauoqzZ()VU@4TYi_G
zq2{$N>okwwOTS|Kmn~lZ@U`k^9r4-iCcSX&^6w3O=iIk!{ecy0*3nxpc=kcxU$$0V
zf9EgygrE1w{O;7c&F_6N{?GbRzelw$dJi{c&uhdd?{B!LG>`B9h+CcuIp^6g{OB|N
zr<d%4f1-ExU%PsCLf_@f{JVPkB!89f+b8E@f0v1WLD?I76o<qw_9rf|V`X;FAMoSk
z7pGCrdHFm1;H4fO?3BM$j&{tiiUZ^jw>>{}p7FfUd!4cD`wi;7XX^b)&o}uCagBWZ
zjdhXN{>np)llS7sukmN}NIiYv7sV0#30FAUFXt!w%|DZ?idq+(?FYW}M!sv^_*tKQ
zfrsa_@N`a=CwosjzL#r%yszs0RsG>=|Ge+3AAKbUDju@yXh-Zxy?*SNe~EFEUcv{C
zai8ESZjqDTqVh)TB)@s+k$&_)=AZF>AZ7i`=RAdv_g7K-XP?-K=ksx&^h5Qln$PdO
z*rE3Ttebu+!$sby9)IUA)Vk@Tb*V?y!v`+>B|D(^_L(1w^UF8wvvt@v{G(jtU?=v4
zoT&3J`QQK_<McBhIZ^Ai&&DZh?=^L9<xfi*o_Fzg;*z+gKfl8tqT(BW#$NU3Px)tg
zCu$z^^OvZ4<578pcGSGeo=+%iNAW||$NAOkM{aT3dSbhHq}@322)?mi{^5QNAN;Kc
zmG9_>ADqmuEN)sSiof=_f8@sB`pJQh@#fXvc=hB~Zbr-Tjpu*jyF8vh5l8rY=U4uR
z-(%1GtluB-$NVX~)t~?2m*lPVpFN0E#`7zvJV+ToeE3Q7i(AH7FaG>BN?v{tKjYA-
z7x<Hl-z6XWkNd^1nvWb&KJD<K@ASs{%|}j@pC%_h5g+qw*WbF($WOibP~+jK9Q&J3
zKXR$JE@k-QXPkNUS0=B0P;Wf^%xk=Tpcl$0yztSlirPo{hO+#IK3FHd>?g`?{wPnZ
zCy#m*e-uv2k(!@;v1}i-E2HY+%ucntr*i(5_d1u#JH#oJJ<<z&<wNE(UjAg?*^_zQ
zXY<qeI&U~vYNuE70rSAmy4@SNS8?y+9?p4IKfiZyKj3#F?#tclxcBg!*Sg6a_lZ2@
zL-nTz`0)SuXgAJzEK>QNe2^aN=X}E6@I%RCUQ|7~l(nPc7yioPo&7W)e*BYlDw9{e
z^|2@UtN!8|`Ot`m@y4-Nl)mbx3{Ub~CqCpMCw&z++{eo|<TK9wWxmnKAJ~~VW}JHb
zl+D+S+AsEqzw@zkuyZl{kq_cee&ZvR4?QK%Z@dqs9Upx86Mn)y0Do;?>=(bTT|G)K
z@a6wdc_BR1(_8+=c>axD$98%_|L~zV^jW(yJtB9rb<s!j%dg9QXy_5Wfj6AtP7aiQ
zSO@>1j9P!B>{h*dEdJWLj9tj%<?G^s`%?F1^w4_fKR&4SnMc_<sl==C-iLbQ@vRy!
zpN50=kqgewNAwtf_R5~h{c6~SeGzZThtg;3fS-1hURyuD<|mJS^xwMUJf)nC^SO13
z&(>S6tD%4T!N)!;lgqr;MQ)VbsB!iyu7~{WmL0NR_10~Cxz2|C_OFW4PkI~ur+Y^F
z>pUnQaZdF7Ts$B@eIu{$-1?5U?@r5G$bqkNxi8`O3-Cp)AHK?{edE`a>5=_I;mfaD
zhkE_27a#JV`l0G0B`0eAp0C8RdX)d;ulZ$Ey*PyO$FX1Jr+=jQ^83#D;+k`ZcJ=0q
z^P3mNpPyGYUKz!YA6M3&9a#r|Xgqn9^;4z?%2m|;hw(@MZinZ5_uG@p@0}}8JL{fv
zFIqln#COX*XZXQ+FWBqD=dCj0dl|kzgUT=TSN2^R{G21*XS>&xA6l>baDTV%`&<5G
zkGb!jJYqlXpZ9n8J^P})U(|WZKHFDz$KJi~!oTyM?2&yt2iR}<jy%~saE$qn`*P2R
z;0%A1J!towhu>{R`GwI^`7+w>xF}bYo1aC^BknrK8^@2?7yXp&pK=)&8u2dn*H8Qt
zzr;E6h&Sd}Z+xWqnNOMDMztGX%G=08w8JM-a@Z&BaKMkB<nQ=5_2lEH#53_qImR*V
zsCBUq<JgD!<9@=MzTn3$y%$7od{FPfIM>-nI9MNj@_r!v$c-O=MSlIrhsq<!s~?=m
zi;w=+1#f=Bevl)U^(P;k`6YbKL$6T!pxwIRL*G#AF^>Mj*FMSv>7ViR5+CjO>8C#S
zXLssR^~T{JsdnR(%^PunhkCfGhZ`y`*k}7?pOoz<J)j@%>);1h`LgvWJ7<~~bxy*U
zUeJ>$kFw`9&b8!o?qj#+!8e{a#6j`Yx+66&eHSmyi!VOpLiHmz`K(*Jdh$AV*iZdY
z{fuMJ>`6QRalQDE%R1q2J;podu><GJsMp>v^PZgdK)es<Jv4rezEtTu{g>bHWBRja
zzqe=ak?K!w_Zd-se#QM?+z0s~+}IJ^^><%jeb%AB{jm@B+4<7F0Sb3~<g1aApZ$}E
z{qqy{Q@kOs{_3rZ9b3Qg#yR(!M?Zdxep-)y`~W_29(a(0T^n!zytfx|BA4}8zk3gU
z3U2g^p4dP7O0UePOb&eDW8ElR%KO2_`Jde6vVU-a8@%XCiNn-;AG>kR#fAE~Zu?+8
z<R!oP$c+!YQT@n6f9Z4Cudyz6^py2B+Uc8dao#A8`Agd1LqGUK^5}=Zekl3jNFMva
z-sq40Ag6KUG#~t9x!m7y9|s@vM|`XN2K%sXlsx31H}ufD<(2YL{#u;(KCb6%%G$jL
zt*qVixHzt=tlsx7QT6zu_$!;=b3yCz996#Vd8)Gb!7juJ^~$JxM>$gVPjB#5ugo6R
z>re0K4gG3Htv~LEvVD!X$m8IFS|@)P*BQ&!74cR!4_xszAAS)BI9Y$RTXrr_!H1sf
zubtk*mmSOd<Uy#s&pin0e#3o;denW9douTR+TDvNqmeH@D7mZ)Uhw6otP?))kk`xe
z)a#EgTu|-i6F>M>W#j3Cc#9gZ9i<N_Ijz_J*$?xY&;H3<wc~@*Bk`R6nKx2;!Y-{7
zf9>|oeWdqVl(o|rRK5E*dO~kJmnT1JU&RII5dPo!L%VwW?mQ52({3L4T4y|GX_vRz
zC)EBqH`%{P-8ZUNCJ#B?huUZDsQK`LqkZKcl<6PJ-=OrCpH`+1u^&EB&tpG&C;vj#
z<L~!5?vwa$zsup@wfp|5^Ah{DZ}x@0z!AR6>|K9)gc?t;^rx5hNj-a!hmcD?WgP1L
z8}?vd*@^kR$3jkd6TbGremjTC6C%~_{E9C<CpSMrAM7tmUzGU`@}S!BLD`#mQ1fdi
zSET0G4_{Os%KzgV_eUP7-8}Z4pRhlX(kJ?*zvnmJqxJqQe<<EnQT@C>uU{;CPGLS}
z{XMVXkNNEqr^{aar*EIMq@V}Zfm$ER?^vfYKP1ja|I2^jOAd0O)=hr(p3A{g*}CyV
z;b8sx`Fr=Cx2dn9<iOWHdcJ1eaDY>!?3n(rd+Wo`b3^Myt(W}dC6}^w<Lf;!`IxeH
z`=y>-_R0Cne$qd<;$z>G$wPj>r}DfLg&Um7sh&LICH%Z6B<`wjM*qh}OTPNhwQJXR
zJLZb(7u~*Y{q`ewJEXSuI_3AueUGz>idXV}Wqwwi@_S(Y+#AYcozuLRDqnN&;vC`o
zM*KUf-8n|w7azr2_h-Iupx(U&KM?)2bGo?3Kl8WpaMXBy*?TJ9ca7^4FWgVTC;DOY
z<4X>46y?A8d*jSwyto8^cxtyE<H!v+d5e1o@{rqj@dv*073;;%{L1ca=><v-{gw5%
zF1Xuo>%iZBSik3U@-n~U_I{6Zx3c$>-A~BV{0<x+dQM*R(P!=YYbPJQR;KUrLw1G_
zT&*|S5q|8^eE~ko`0EEB^RqMhVt?sB`#`m$)@i@w({M76yq7%W<Y!Rrv8>)Y=nMJL
zIInT!u&*dPMLqAZukeDmeW1tohyP(O@L^x<*SU+G%eUh=_Ql_s7d0>YU=Q$z4|_uO
zGaj{n=0(kC9KAG;^~Jo?y}7)ef06G;9GrLHVmt~TIKoYSi}G9W_jh5u=jV3`;tHI6
z&sw|hAQ*r3#*^o~al@%~=M3M&@H+_e^4H|HUVhd3<eTy;WzVVL=APMl-4meBcW^<S
z3;0d=*(df)F7?K%k7es358NY#L#(&&RpX6AJtyR6*rRh_tJHS~t&g6wGq~W(zO?hJ
z{3Cw$U!G}SmF+KT-<-?DdsIJs<d3l)e^h^E`J?fuJUQm4`fHcRR+Z()@_GH`&&~nz
zXH-A)lUM#5&msEDpUGoh)Oo|azH@<U_x_@K{k1!Xm=|v3wQlVwxyeWW`9FH1o*g?+
z+CTe@!bd;Ue!<N-g1q`^C!c#6a*`jVhvXqIyXD8C{_<b3o_~+^)+N63$KtuTql}76
zvA;MM+x1hAkMoXvP+lSr^jw*r5byX?e7#SOM*ku|pa=Yn^@&&J6KBn5yt4k{w!D*{
zD*HE{w~9lN8ZXY!Pw|7lra$}^$`4vEN<Q&Ne>Ca={a_FD+I>2`rPs>rQQ0}(Jw7|7
z=g#}qL%-dBDwC5wvqO5%4p8?B?nT^(l9%5iC%MUmvOm=QhyKd^6F-YDyM`b8XJ^)_
zoxZVm>qgz%Dx>s6S-;2!ZgABPRj(g@aE$fJ?r%`*Q^q$^{PiQRa_mnZwHxnzz>lya
z{)fMiKN&|(e!zO@3%li4oL`&^`3voFe)bssgnH*Z=OldiMP>d(-lac39nVY7WzJdp
zIVZu(IP3EJQhBs-))gs!=85vj*YL3({oqKBSa#2a8b^L*{@K36$NsP@=PTzW`(PXz
z`{B=@Yqx%SW__r2lZU(!Fa7XShPVFO@iUKp){DZQd{yJ{iS6cv7aYt-Pa_`2S%)(D
z>9cn8lN)ueqZh{6e|(j-tG8eFl^+lfwOb#&+zXKxB`5yoMddBN=jUD{+9Nrw6Lk(|
zfAmFvc*Xhkhm&#E;rDs=%Y3N)M6DAgm+uq$JDvV6mG_{W&#gDgjX%7tM?F2VKl*96
zf9mxU|Ku&=KY#9iU)+`l$Y-?6SHyqRdB!;P@(6s@i}U!4W8#W@M?Y~-euG+<xF=7O
z_c)h}-_|R>iIe6fhdjx6WpbeS;ZF|d5AoD}g?tsI2l0ED^7(kLNPpccp!OT(&y?v+
zr0xgY6VX%k?A3j!`)K{VALl;WcR}g7zq^Y1`=Y+zsh@i+e|PO?*BrjvS>K&pcaP=o
z;2Lk-<o|ifajX4$)w;ji=I_1vdqwWWd{@(X$KT^6zxnVduXWl7aTqS%AGB`v&JQ@B
z$sdf<-#W~%pK<oVcTL^j_`a$B^ngC`cg~CaEWgKJ@XPc{9$|j|ksq>ta>2*C@RPTZ
zgMTECdB`Jv#C7OzUHb9U)*(*GNBKQ*P25r@r+YK}%`bkq&(mMM^_fSWL_RdGkG$-V
zyyVB%^FQZq{hZ@b{KO%A*gf1)cC=vn<1g9kqBZKu?8NteQFiUQ6F)%D`2+h;FZ5HU
zPwa?3^E>)0>qo!i{+QP|_CR0M(=+<*c?rJuSG)22f_0N$zD0h1n_aLoRD7pD^uhev
z*%$pXA9?7d{?>~R`OOpYG7j~ekv!(7N6PG9-b+8}FL}iq=LmS1&v^XRn}@yFcl%--
z{`jiLkG{*-$i+Ui`+HyFxjaHX;2f_$#&z}LJ}U1})=&N*A8<dVzdS-e6d!ql=NS6S
zZ<<l_$ZMSEt=D++;qSghJIe0Zwf&;+?2Vn#L;FTw?LVqrypWHh^w7T2Gy6?H_;>ng
z-_`Te>cs_o;yCs6h`!N3_QO7{7Y_V`b&EIbTR-c^hyKHdT`QYc8BXM7PsZWLPoUP%
zuHk}@^|Dj<qwo-i@qwT5_`+SgxanR=o-EGEm&HG2=NnZ1&u%=26d&=Y|LjVhPOnjR
ztH1LHKfrIW7v~p#!+FPjsQH{PJnwMsz}Gzb8HW$be)S`#`PnVGjB}pS-};sDg#-V9
z@>}lX;SC2k!@b<M#_y2quX}s?K%eLfeWstj(<h#z?8-SwJQlazFW^hANcl(9`ceB`
z%HQ~olrkJ5rI+;9y!0MEo<Hb^s+R}C%{YAE$WFZ%z>mU7J3r{Ym0ad!$Kp<vJ<wNr
zhSF<#W4wB}@gMwy^;A*kLgzvGCMqseQU0=u(mRwtric78%Fpxb;+=8)x_<n=cJ;<7
z>*t&&zTy|>$4@<K-^qvKV?J>~ybwpk4gBCH4)FVE>?c0bJLB{tw|UHO9{UJ?{pG{t
zSC3k!vi|Djz2dKW`HTF<IfP$v&PDA{%!9~foyu^xKJDyOS$}e&<{=k3{QVV_yw-v8
zo8%@BeS-&mgA;j_$!UL4>tH|fHSNmu01ocUoloOD=2K5!=m-7cNBA{<Onl*2^y3HR
z5vcs6^ox!775y$g{I7Y`!-2fyl&6}9f5!*44*Zm3yR!9>lb;nwtVf*IUdFBY_fK2p
z(DQ2gi+j!;;=b|v;e*QG%qNcMFV1KeC*>FBl~<S_AJjPg<bB%9b$w^6SO4EvU0djf
zpZWB2A8vm4<njXjOZgjo%?BT4=P38y;-kDy9w`44C(R>|lb@mTIQQ)0u>4Iw&(-~|
zz<W&O@jRX!s65g<<cj%_{+<Vr2Y&8Z-Nzn!+MN&lW!}{KyWg?iAC5bFwesG{{geA!
z-#b)xkBbj{QTpzEZ+gbxh;RIecq=aOZ~PCx!f%O(sC<(B(qnqTPMxF3VO+a3`VaSf
zaImlRfWFgTW%@<$jb}e7d)ALW(;M$`&;!qhl<7~@d-^FqWOwGFN3owc2*1+bHQpB@
zH$K*B{np8!x}O(E#Up;4U#3^|%s%7K|JomM1?4ZzPd;+N$-12%;p=&sarB#>$)n|A
z^q4#mCvv0og59!rdKCA;zNts~HB`PI{fYIW{2a=kDeH&ADVE{py`n10P9pVQN35s6
z*5iF7`=USnCx4VrnH{0>GI^c*cKYeL760kEmGjWz`3rX6@yaQ6_x1X#kI%83r{#U-
zb)PTK<Tst8<(Jxx_Z&|dfBo@|d8l)nybfRaC%eQ~ey%Kk#h)DVF8rMzls$K|ZujWg
z-5;38{=mUF?Z%PceDZ1hv>T^v9s0vVo-BWqU(*ZtkdGYx{<L$l=Zo&W$tTY77yKXm
z#7kwAUKp>QyyB*I@+)gc#Ygc{T;<oq-B>oC`0c#s+zUtRty+)%<Z%wgAJxvUMtPO_
zVR6d&M%-~O5jW(^#)})qlN(O<N&Z4F<Tbu)Lm%9a(J%VPp5UfFj(48)9E`oOZ+6Ge
zc#iLRqCAj4;<xCfapF6DqSw}MpUKCL*%A4z3oh)7{MKun=Xva!pR!I=yLsS(4}6`k
z)x)1(vkv<e`?DAIaFhQki@V~nJW3o_ub=y3aaBGdua}RAOU^&q@kRNQRw?~v2lSl&
z%a`PHW!@G3PMo}3{vtoJZhZI?_Tv0a4phG992fH==QQyjuJT=RUw%Rk=P={+lP972
z%QNhQcKeIJyvn}HFXV6P<&miKp!c$!3sK**R`&aG@1@IkQ1yO~0vCA`z4E+4o~GTp
zjk6B<75=Ds{0`smVf@|DEf&A`t$UrhqEK&t;N`rEUyKv{1v{jd^w2ttmuIpo`8oXQ
zG5us$?8rIG^GxRtddmNilON$P#TVxRa<Xsw1ZQ&5bNWRt_apG4Pu8W~dX(8A|Hl8u
zxG9eF!`1<JewCl)SLMb0yYc3Qr}d!piQTau`lX$HDx>z5{o~6ovePOm{}8t$9{e9l
z-%#}^zp0&na~~kjU<bZukNQ6S!h^s3+Mn+_X(WC}U%mdQdf&NMf7iD+|M*AOnKZ)Q
z%zM&qCtrU4wkrzXnb7XL``Z1k4V5QW<z4ih-$tFw*@1dE(L4E_ybEP_>`)$p(ns@~
zhrOAfov;u3$q&*~c13TL=^KB@Ppaq7=qEqKuY2ChztekmkB{-r`6xf39i{gudx-qZ
zhaW$JuX^j#j^fYH@=N%dPd~UtdGy1Vzr%<9+830ZDF4l$lFK^qC#U^1Pm~j1^1?ws
zc#D(P!H?TV{;L@!pK<u(OI~_O4)t)N&-}gq`dL4mjk7-TE0cpi*6#Oi^qJiDho81C
z_2PoOn!j~!BA2|*dW^Rpa1qz-11j%PM#%{ecE~>MlXmma2kXa&T>J<5<$?T<_3%Ud
z8Gj^Cb`NMg8uMQJ#*fKQ$nCicJs>xIATK}e{$r;X9<%6%dseGgofqUa_E8=o{}D&s
zKZyVK8|AN!kM|q&B2xJZzVZ@sDWmvDxvX29kJSBv_!cSo@u{NnB5?^$?iEn+O8$UG
zImA8V#7}WiUP15ZQ+Yq$xQC~A^fBrmzV;E0@;~k5M9Hh{o<)E09iI9tyGODg@+Rvd
z5B%*bs$Twpk38Kx<|QXPVrOuHlXjH-!p(a>%EpOD%1`OnI4^6D^D5hK`$-Pt*t`DP
z&8Ht4<t)qOX;66ZFZkmVsrBlQFTE}0YP_H2-1^|GH*B{2h{p=wN$~wR=i6BCe2MDk
z_f5Xb@{T!sFWqgEDfQ@=`E7BEe-!t`EmWQ$4v4?{qw<XyH=XOn8PDg$Pw`5<@9v>+
zvQBxU{`fjK<0qb>;<&hG{njO}h`aWe|DcEL5#@i_t9+i_tJj}iup@kpEA{11XI#D2
zkvG@)eR6mY1V2>0b=YU?5FhQg{S{~RhlhQZPa03p<(=ZGd5xE^7>}QN{IrutKkeja
zfAVSl@QZq=9NVp5JH17#>dmL$Xla};@>kXm4)|-YqE-FKYn*=OQI1r9{glzF{_3Or
z>Wz;ywi~ZsRlE8~<M@b!{7F2Omx%N7CscgZPn?zaM1IQdbB#kiA9H{19$sE)UeEL7
zr=BaI{JiHAei!I|3P1OR@{dU63+m;m)+Y~DmN)QU{1!arN%(qxX`QJ2M4p5%f5;E=
zZ|;xyRb}I?(|mA*JO1(?>ym%-v-VfndOWZ4oW}mh+xU5Sn1_7uBM%%R^*qOO8Tte-
ze8^>;v1~m3@qCDWc&><_{18s|&%Uw)^1xYsPv1QErQhVnKjv}vQ@$m>$Xh%o5pU#s
z;;6jJ`OY|g-}zIVGftc*4}H^)uXg3wPCo78w0NaIy+-jZ<6Gl=q8>kS6%}vEZ@-$+
zc&>$)I4l11yY3C#GsvqVr5E&p9;l~Z?33MTSEe5*z2TRX@ncu)EK+u@zj@fV`T0Bc
z!_L_WeMa5$D|;TR%rCGDam9S>Up;$Lk2<%J+w&E8z)5=*6^E=Zmere=oT&Lw{=oe3
zM6Jhu$xGP<{f_%WZ{Thp;Ocq0=lR<8=im5K?aKB;8Re&x@e`*#=STTf^PyGq#d($W
zkCZ-%^KfV1;!CV|k7!?%%_pwVBle-c{W8w}*jM)Ly&d_7_blj<^ALY3PSa2P@Mllv
zSC5jD-iR0a!-YMe<WWY&OYsjL?$PY8eSjZ+C_UHTcNV<o=KT@z*EsylPw&Y`AHDZs
zJUcL6yZi%A<f50>Vcqso9>bpKH%bosz<=`#<Q0dU`%v*kS)T4b!FvbJ<>E{{*Zc0A
z^S}E5_adI#q3V54UfFZO^B;V8_I)Q$89_a7bRRCikVm-BiIg1X@m!G{o)>z)c+?iF
zJ?q)WP8#uCQ9Vjd_a3p{bIHixJnn-cAGpC4p6+GDdCxUa?Z(+3_bI6NQspafqZjeK
zP7j<fmE~)mCz8wgO&+IgKiv1qi=6-5^WZ~H_fG2LeU<#i`ceAfp2_+nKYUT|-733x
zvJUqj@PIS<*#}(I<D-m<<NR#&lj5<wRQ@UtwU6>~d91u#+4+zEiGD$zKpyj==EJX)
zqw#wt@@YrOLC@?ve<P1omN(1)^`}49sf?OOfBJ{7{i9d(9py)?3$FModrlSSH4nS6
zUi_^~KhHPS>xY_0JG+DnKF(S27k8uF<YB-3Iz99r0DH18=2teaIDwKI{_s(6J^DMh
zC|6PYhC2V?M?Ux{TbFejr(HR&3!d~Bo^c)4NiX2fzm?+}zk`eCxh%JbfODjL8kHBz
zhxt+YxBDLVQ>gutr|O4BKPfLm=_R}JzN5Gk<GJTPsCHER((ZYQ{Kb2U>cu%!e8eYG
z<Kzjb=QP?;`A2LwPX57fsOM+Zi<9ILC*_smF8Q^KqvAKd_{BKQ?(GLY@@#pq{2ulE
z+rHXAczezl$BUQF>Bh^$>7)2gpPVP$gV=BQq4Y_ao{%e!vv0=5{FELUhssCgu_*lD
zipF~TKp*Tgy^J*OqjvgYzs*BW#3}XkgIw}Sb}6pN*Z6()sQ%i;C-Dwn^N~Zm*G_-s
ziQ<O1A}))M%G%Y7Z{m!&kDq?V>n}eNm*}1MHpDgU^18@RyF3czC!10C3i_F!Uu{O+
zL&SdWxlwk8vZv_xBcEnvxT4Ma^8@Zp*gZMO=Una{MEr@=e#l2rc|2UjF?gwu{Gz|C
zvLAXcZ)!&QS$1SRdy?m(?M3~GGs-!Pe%N2<6MDt2<vsEs`_FH%TYBR?8~TO6`uMy9
z^5$g+@?Cn#U#OS=Mru5}@cucuwWIPq_HJDCC(71iJZc>3ew&|?FOtW7K7844ytlIt
zDEk-Z-B<DtsCyyzKd5+&x~KBqk^3v}&!hU`BYyiGi?|%;)z5q}Uyv7QN97;#31xYc
zdr#|iPbrU-@4-p@r)T27d`CG_ehoi)5`N;axGj&6FYsI9zB2pa$53(BIg-A!Cw3ri
zuqXN^&d0c`EIy<14`ueG?A*(~`3d~!KRL;xj9;`rW%hy(zTyj<#2<cMo&pE;aC7cP
z$!(l6yy3>*kYBtPS3Spwd60c!=k`NA>Kx#_z^+mHsvTwL$|(I;)(=1P+6TDWC;ME|
z@Lnu@wc|%0;mHo<k@TJZ&~NdZ-q|<%09XCx)$TV@d4c!O^^-UAN9x(Fa{)i%IRL*S
zZ{Zi*v-3mFA^d`Kh_ZR{SI>X&FZ_e~@O6I-C+n0aS&uS4@T0f%+r3<Q&)T?8Wmnqm
zhy60I^9DJshrG@uaJNqKSqDDG@t5}B{Ny4Be`(#3k8+d?&d#g+GCeR3esO*FQ9tV-
z4}6TbUjC-^V~zKR*@JP`gP*eb@WUTvUvOf7=4FTYDzh8wgadp~=VSQe&)(T5e;^-<
zeq7$O(M9Jkz4PtU>gQ~D)uYolduIL2RW|tj@4vWy-QQI{d#8(LeEv1-)csD}`)i(W
zitC<>dOw4IlRxph%J8C>*2k{dyS##Z%SYr9?3+C++b?+1FJ;tzvlIJjJUQtv{EerF
z=9PD#_L2Oa|0!E%TsJ#pujVm7{wVzTPvgl&4*l^{PcHc*zw39I>gAL2M)zXYLk{zz
z=0){$Pel&x)@h%$qxKu6C*+jpqVi+%D_gg9n4h0lM(w+Gq3|THb>U+@^r*^@;IDt=
zuim=xqu<K>4L?Jl`3-)D|KR8OLGfMx=+BJf$H@sNc50v1!^M33u)JSco=8r3S-*A4
zFWkfF$FJy5KT-10V>H$~4>@<SNA_iXqouxI;=3i@KZ)<5xkq+C?S5GvDX-MeeX{=i
zu=iy2Gv52M&iCz7af_d1pYZVf0ItobeL&eIKI+lP&pE}pK->`r#YcIwI4Ey+zHt6=
z58z(F^K<tSo(G`L56-{d=keZ-_k(=bM*Q?%nSC*zdEC#)2iY~fC%=C35a$*0dH;z%
zh|}(!+;f=My$Zdecg|bh!^1~^=PTzKW%kOy$TP)Z{)Zit+r0{^-a5#|&+v!rRa`f}
zdj6I_q0jsWKKkL${?*Gr;lS=tc@Mo{ug-(^3H9C=oUIR)Z_yj`dLPm}sQl0V+HdXd
zwd|+&(NO*b?x=O4?2Y}AU%h^Do#rKve*6{r*^BY`D$`SxoN%jJ2R+gbr%27`yym{a
ze(>Xw!qGb6&0q01_*sv89d_kj80DXwE78&)G=9IRY(DG2hd#iGA5w2UaJ4V=6n^ww
zS%2$M4`+NL&dT<WoNy*T9NiDQ2Zy)!F5v`UehI$L<Id^ym*0XbzstWmH}F^d0so?&
zUs0C#%TM?deu!T*pK<2XZXWU1Ii26)_xM-*JQq-g8$H2?KaS@~_GzEQSMd{G?zQkI
zC%LQ}l}9SetDS4eAwPt-GQRMNc?~)ECHl@E^QZKmURh_P;t@OWy-N9=_lo&t@9ENe
z`Wx*+J}eI)54#~he%i@vy!D9>-XrE`<9e;ncayif=}+fxanvg1-_28&$D-Ehy?=gy
z{jx`tU*M0}A^k(yp><im@%%!$uM5vUXUl2V7wlM`N1v6QGs<yiO<DIB_ugNN_ssN{
zAE?rA`i?JZ9DftzhJ23xI#;3A!9SUw9pFz+_wdU69r;jl=%=3C@FxeVK2o^4heDkj
zt<OG)ciOEBAM>H=BZZHC^n$)Szw)>I3;%5#xlrflNa-*A7LUX?`s7@}Kd^Ul(QEeU
ze23!a?>NSJ@pB*NJFWW3Gsuam*H1p>_bb|+W1L@%H?MVw1N;rYq+Xmh4?lq~e}wwo
zi~jr+`TWkuc<p|dVqN^ObC-VBr$4^@H{9h5<Tsvt<TcK@jz6Ur@Z#U$NKWz@FV7-}
zxMUptc5aZLq2y)P^xwHqJmHVnz4M;uPVxiok&5^D@=xw{l-alO#>oT3Z+7S$C4Uje
z`AO?>ew5e9uf%`+^~2XanDz5J`bEE{%-`}q{F}0MqWp&W<XPrX<`?Bz{13d$10VhQ
zJ^p3Ts5~O+tCoE3@!!_EjRU(C?<1w}^hz1!FX*{@GWkV0uJQbt-lH*ZD($AxU%Pp#
zsJI#P_AEac0dYtiLB**^V_cF)%g@CddMZDORK6|Fi1X@EdA)lJ_cy40F7>MM9M(BN
z9;zKb`7M2q6iV*z!~_1H-xW{zYyLFqGdZHZvJ;fP>nFbzU!t8^m-vZ~cqdNq|7ARG
z{GA|W^0O~g91$<XBXQFCTKv@y)gNx+nE1{=G^6w(?kim3pxymCy>jlOpW+<-qo2w#
zFO|pfhxWyM=8yYiUr>664?U63ME%rHz47cpzNL)H&)5h1gopEm{>t<nrBC$7IfH%E
z8+zuROx`L#=0}_d<e~hB`wROc50#&y@>qE=%3kO*`*aT`52UB;*>geqZalqpzGIj4
znjO+V{pmGa%p;F<?`A&Fd7UdgM~=^*y`SkhwC@|Eo|lo2{iF1Z-qJUE;v7$pt&jd#
zC!Ce(1--=AJo>W}RK0QhElOYM89C`kr0%WQt^MbB;BG(oFZ}7hbAabND1P#5<#>Ne
zzwCFM4}bPTK4p}Ao-?ELWVF<Ca`zSVlwTLGl<AN4IX~MElplaM9QjXp^4s*Gii+n|
zl>akcKl8Oqqu=Kr@w1Ld&5vr2<H$??*iU~{T#ESO6Y*DeKY$O44{D$ABfoj!p&vf%
zn*6bBpVhN(euqC{@2L7n@xzB-;)nSil)u%^PU*8aM$g$Dd(@A-_?Xu`+SS|FC_j1h
zGoBpQPac%q_5~mH?AQL;AN%Kd0lSRz*<bC}Z(aNyKgQ3K!~Epo=iDpAb@G4s<0~#1
zk1u)5Cw`;e6VlIkev>^|cf^%E#*ruTH;?D^k&3sd_eZ=>qQCc7^y9ac`E}#f!=0Xb
zzUTcu`9aJJ@P{v4?3aAq`r&B(#z%Rq2QDc7D0$(|uC$vkQtkFVmeteyXm9+4c6zSN
z4(TC&a3N2`MZNykgAYFusXV5N;)`0ZxXYf*!_KXTJlgpa{+B%RWB0wv;(>U_-@r{7
zuJ|dluW0Z3X-AE-PWs{eY#k{5_TC(QjP}JY=r26wJK7^PKO8;(HjaLy_&UE@hx37W
zB@UzTbS{!lxo1J$+k5Wmp2hd2yl>?k<-KY9==p6tcNy;-<~aqb-g{)?hxwx3IKO#r
zq3k&dY8}pb__=RV_B>mE&vl&loZIF1?h)l7&cAS`&&G+1+T9DtZ}juL$$H3*KR)7S
zoS(d&Uy;{2mVU4+avEnI?d}nbbANz(e!<?H-`KzVE$d^4&ZqR*Im7wLdw}8(Im|0w
z@|(^N;tl_avOD&IKRMWs_|KoPOa8(6UHoG2%HHcRAG`J52fKD&W5?#<cb)$|cXBTU
z5B<bB_*q9Y$`6WP{JAo|_J`l6fArSAlLx-2@%X8aW#iq0*cbb&UO)RRAJC4!e)35D
z_;2?c@*e&ib-$sX-w#CUJ!<D$@Au-XoxSpl?wQyJ+)LWHKW1O>f+t+;bHvLy)Hr$~
z|736Y>Cc|nfjA}akQdNfR6R<6wex@0@BW0Jf)hVT?@)e#JmjVi+SOY(N<YZQ4%j!G
z>63Z*9r+sypGesSdo`bXC;0_@%6^USE0GK3SB)>{U;dkGH<)#1t(?D6#t$CRe?-3c
zk<)%DYv;fD_n7aA8|vNLsW;9#_#NvbuleaUJ+}_=$bB3>_{$gQnK&m7$v?$cWqFbO
zL0P^dACl*ohd&fQ$>&_|Js9W0=ts$KUr^(%Q{G^m{D<|iA9<fVfStLwwhsBAd;`@_
zer5gEiyu7gFaKqK)x(7y!^gVfI`FqH>yT%naI|iDj`gAT%RRsSzz1IH?UVY^QaGXb
z#d+u(oXG`W@8KvLPcCspSsqJoQTeRAT0i@29`6ys32x+rGs>>uf)Bptr$6k>I_xk0
z@UUL}?I%4!&4=O(4|y@a&2R8u{3E~bIS4<>--rj|2`W!imdA;w;*EI9-}7hqiZkMB
z>?aTO{LQ&zy+5DzmA5TitM0sJU*Kk6*r$A1J}f^H53Pe7@Zp#EA@<0Q#dH3M-%|Fx
z((^t1#0UODf9upA<zMv|kLV5m68#N5HJ;r3sBz@tm#iNZm*V`TJ~i?+`eU3t3WdA4
zO7G~Kd;^a3MLeW;_}D*sLa*o_zUHyd^p0H^XFlWUBR=FaAANEz!k^yDH}Qj?{bo1T
zrN1&h=EdK-=?A{-L;hvm@Q)O(=0WK#J!2Q<hf}n7d9~-(&VR-^cUm7l?2G;?TUWFf
z`UOY**pK*&k9zwl-qLsahT31a+kg6vKm8}iSgCWe{q<ZL{?1u`hwQxV{0(>a)6RX)
zfzE&W8*d)^Okep;^5EnAPG9vio;{F9yK{<llF$7-{`k@x_}llW$K+$z=3&?P(pUVW
z{h5zl&{OlWhghaJ{E7CGHvA*IbS}`Yo}YCd(C(Z>AN6D3^pans=i2dyn{`@0yyOG)
z8Xxo0!_sc2-uu{%b1p8ltB;iZl9L}GuX^*tk$oz|nS5|S#WD72oOR=GU#(9(A_v@+
zW4wf`e&mXj9$6<n<tOQ<cKDLd{?K1`2Vc0e59=|ndh63}oH9K5J^h{k+;clG$#3P$
z@(Sk=`6wz6_xlEUs(eabYv1_u$FJZ1=w+u&sf!Eh*$2F%UD9{^$9{~Dei|R`#+!%!
zng_N2`WdI)I28YqHvEG5jYILbKkD&A#W(RzUdkR&{ZaXX`Ahs8a+$ZJ;d@o3d=sqR
z`H($0kFp2nIC}0}Lf_?ADE)ya{iL_3{bHZ`^K<kHKYAfgG+uvxjQ>>Tm&6HWek1x9
zenkGrzwu-ElbfEAhkZN8(O+@WdDeQ|kGeN?e}LK_&)58&9Op&9C-U4)*>f@e&F^<S
zAH>)5GX3yH@#Xi(;kn)pU)$`57oN6i{hc?T``d#LT(y4fom=m7&&<{8hn{xl1Amz}
zweI;FKKRlj=UO<~ck#_WD#M-sV9)f5zM%9EjsAi^F|Yk0k9Z(%y2sE@{_R}q9>je1
z-G0+I>q4WyWe?hor(foWgLOD3>&ITK11|IprHAyEJ;Q}O{3|<PC)OqIlgGZezcCNo
z;0ZtLiF&~g@l*E4e$oT-*)RB!UmRuUu^iVSpW#2uV?XS#e1ZSuFWqanCxf^9M0m&#
z?HB!lA6(^|{Fw8Nyw|zG`9R)8zu;y*dP07A4te2DFX*>+alkt8;}_`(T=`x2%l~{g
z%KI6<gXKG4)6Tl*+>4e^8u47ud&<7E=s91+1Apf>=Q{fdPkPM1kdxmbzkLuF@O2K0
z`K@!I`vdEg|H_L|eDQnrBZpjd`d8L2{i?s0_VwG|zv-EKPOZ<}|Mma<AD2(9FPQ%L
zOZK{Gjk<lcPI$!ily=?tt|Pla>7Db2@$f;_!w0UY{ilEIP`orBD(|C5{1AUY&-e{`
z>AqS&eE38Dh2NyNu^vBmDh{(_)IGa%g?{d9$?18Ad8~_lqx^+9jUP(hNX=uN+S#r3
zqVN<4_%VEp=fC+^@@w~=i8wAV;1|Rj`)Zx^6Ta|*3;Dzg^BYHA`-?xy@0gFCi!1bx
zUZeb^JWrYb=O5TV`=H<OG(Y_&2fuC|)@vQg)@z)y^`X|okD|`K?2&!)W7cII#>u1I
zql*)0#L@c1zbL=>&hDc;(cg(@{4u%tL;j3k<Y)CGzxYgl$qhGlU?1oyY8~{@{upn+
z@ug4l2>N8b@HGx@<fKRBw;p`(hok$2W|Z9Yh`%NmdFiG8_M0BapU9`5_1kBDfuC@`
zK>0Uie%3hi!d<(3gg;Y{+Bg0LKXTen_NgDgDqrK*J@0jXa$n^9qMiP*L)5;)hu%ax
zR)!aUD!=tT3G>3ie%d#9DVrBA@=W~2P4+<#=mkANwWHQ)y!F8wF6@jSH_m?OM=o*F
zK9zp#oKJrAwHH57Be(waUpycOJA(t9l~MAS`{lV`<GmYtO>TVI0XenfV;|I8C;1{C
zaOEf9&R@epnLQfEpWsiQ$f;d_ah^WuFK&xp<l%ShFTZ3R+Qk#`6VCjX{`?BNqj&6?
zT;?age%2w5$unbo79Zrl;tDy%5pl(N3Dw^{H>y3xD}0>8oD1UfDdzystGsvWeNy{v
z{nkgn;7AY1VVx-4?Wgh0sC8OTGb%rFo~e3H<~>??ct0MV#_0z??WprDJ+trRP|wex
z?2x}u=BMbl{-}C5vMcussQ4s~*cWoy7kXKRYix%T+~h-%KRr<|zfw;Q`2~B5R69P#
z<3~TO2c`GqQ??%U^ou<9%Q>Ci^E3Q6zrnAPo4=KZ+9&&Kznu&1t9Ik?H7|d!pLk$C
za*6}|6hAG#@UQeNmgPODxUHXfr`|eHe2u5C{3QPz?VcX<C;CPF@WUS-;x9h~KX}Lf
z5qEiv{_rA)eV}h}l`oUqx|GF1_>jZ=@(k^8HoyJVPrQV){8E4Ok_!&jp<P^8kJ2OK
zBmU$xzkJU=unYGe@jW_m)jHvUAHBjyKkL+P92~8mecDg;Xw~o4#6fXS8TDR>`!jLD
z`xSAV_`$D=AIkiXIKnR(CmzcW_&@z)zJR~@Z9F;ES5f{PU;dSxD1S*F{#My|_2fmx
zk0=M6qrA!}oKgFMs<&UtsP|*wOkV2~2gDiiz`9ZEg_HH-V?V81KY1Lg9bfhMNBQuN
zRDW`qU)-a2%Jjv)+YjyPo9!1pvw!ZB?T6p{!e2ks`%LB)@8Ll1xL#$rMQS|U@v$!J
zB`5tNf7Bmk{qP|-YCpvZ_KT9!K3PW<#n1j)2l-KQ*$?u?c4hldZu`tG%!8`ezsiqJ
zlzP8Ajns4L_*~q*Nfq^+-23$SxW_h+`yS7+wWIE7{<`~(A6fjmNp<f9=^v@TtD#;Q
zRWE*uuc-V${vc14UpN<Pmj}u-Q2kN$Wj@#Vo|k%j<TI#zfPC^wd62wVKl;u7=r?`m
zpV$pMV=w%LJYdi%{E{AU^M1ShqKG)iZ{?Y>J(k5;@y0pO^Dp{F|K+jH8T6C>mg6s&
z{@k;c-B8F!#0~ifIoONyi1GAYUhe+Od4&EuU$AfIE9X$>NqN5c<R#X{9<_@<;;VSi
zj@*;r!wymF6sOpyb<59;XOGV9aOHoT%i!dmg#UtzeRm!~;R!eWl*uRGFu%C%K1#p1
zPW|jNocJGn`6qnj2gcFE$S>-V{`kTZ?r=nnw{P*@i~i6X{#RK$>i$C6d5yk_BlJtV
zeM0H4{cyf?Z>!!uDtqq2PCfT^-ZUS(vH$G3*LjZTRDNgYeMskQ_k5nOSeJW1{jJmU
z3;jJ0W9Rk_zV=CddEe4_PQfnio45@><Fxa4_L1K~=@0v4XZ)G_0sC!V$&0V~_(l3n
zpXs-D{+xdEA94Nkql%K3p2God^a*bKrgiX7`oYU{5#K|XXC5@|Q}^CfU#0Fjp*&MR
z&kf~?@=5jH^Po@OC!sf<1A0y(KlT2mdmjDhBYi{J7aHrWTf2DazQ_Knr?1A*3;NB@
z@x#YH7)Otci*_OJwl3@Sz6t7mG`PUa`)vFW`BC$`--C<$R`Qsi{E^1*Lx>aN4a(n(
zC*rldH^v|F+xN4y%XgiR<9T0M9uTQ~8h>$ITsKeb=bY;tEuI<2&gBori>J!?TR&<X
zyp*+fn!*Y7J2Pce9ziZ;c-U|E4ffNztykWl3=jL@zQ%sx0|)nL`a1{dk6*-ze<i<s
z#eE5X=l)q<shyv;K6)%K5qF|r^?cTNc_9Cfk9-3kc_TT*1MTuG_|QjGoWY;m;*0V2
z4X*HGpKw=?+9!7D??3uJjCy{--+!eq=3~G1#lG2J`e%QY@m0?skWW8ypyVTu@7vIO
z{bJegDamX7)?@wn8V^VP$Y;E9<VCFq#g82Nkry7yRn&Ox*6(+y?9zNF`RyC|^|x>8
z$%nF2c-UY2hgz>Pe)?-i;R#3U(%*QL{Mu3LfJda{VGrzv-m*KC{jyhiAu8XLugSl}
zYk7lu`IqyBa_lFb8z-NYf5-#mGs^no7b!XAqw-JutxGvld1zHX>#=_40yNg^ALX=8
z)VSDBIj&28c@RFR_3MY~udF|OtwTTU_JKUc+eiD2S~tGVXZXw0l;zXvmEF(khnfea
z&*Fu$yodg?3-&GUh<m7d6d(44KfcQNDx>DrAH}z-T|H`Clq;6;C9iR4<QL`SXZTBH
z{80SRUZ4AOzTP0fpRf<lA)O=Sh063%nLf(zw5ylDI>*aTjo0oTK)bkQe)XvM<(@<y
zY+mc&@A)~@cL?Pz;<N8r$Q#6Sb|J2qpWozH`Az3r?WlTw(D~YX(D*q2Dmw@3&!0N?
zJ3pZA6Rlr;#MODjxxjwdPy1w?bAa=Mc7I>cb2Gn7MfrR8Cw_nGzQB4=_c5q?@yGq1
z`v-AI+=$dTRJ_3#KjXwf{p<sIwA)u@aglz|w;1Q;XY5aYBCk;&_sw~Xz1k;sp^WNh
zoPA_h+KorGlan5y<g;G+n{}$Ue*IDRv?yGatKz--=x4~Op8TFiShw@1b1VJje^BRF
z_ulw8PaE%kL7aqVyl-R|S8qId&KozJT3<YW!R|X=Ii>FJ<Er;}b$5E<F^g`vXSKTW
z2N!<r(QT)#SwDNHi)MWOHS5$rJoC{Nv){UQUHx~?eaqG#Sg~e($xAQU?~C7hM)^BL
z<JqD4{C-ir`%r$#@9EeF%FeXM{?=_?W%dq-_`7ud$fN9eyx+yc4Ig~n>#EnE-0YJ*
zt7rG-gNu2TcRBmkAOGp4|HRerJn1>Vjz2q*f6+7XMBXLNIOjSK(pP#YuF*638@pz=
z=F^`&;KP3OGY`4>Eq2Nt_$BKS2bIZTT`><eFUn5fV!V2B#=5NAy7Y^b9}<V{vvKf1
z$qAn*zx_b9D?68?>g5ymSG#)m2lT->_X+6v4?aBmzLTen_}#VdCA;TwAEdu~WO&J2
zlp}=;sy}MIqovN3sOJ>ofV{-}2JDAERlV=4UVO06RdJL4dQRYZ2RX>69v^W;{Gk8x
zBk@?f_#m#(?|2_izde_u*Z7;CJ;BL1{lp*hdOusdcWSVo_S^okH~FzVjves_C2jmJ
z-Sb7yY2Zjcc_e$XKjMXP>`{L>dS1mJ8z)~=_MA#to-7Yz*X*qsC9m_nxNiUKCkj9N
z4M+Rq`J+5t91>sUyPi9Gj;Wt`>U%KydmibWfQoC@Pv7a2_idb$JdakdpM9{O^prl*
zH|IG0*#-T92mPU+^cGI2aVWjeP9F2IPn3N8hd3r4!$<zEpLKgqjSs%o%igWqIQ~dn
z<Zr}J_JNXvKV(nZ`9YLkiVMo@+50Ofdl%RBqu1JF+4C^xXZ*xNls)hl^pu{*vU6D#
z<*&?Z9`ds{R6IhRQ;oMSeh!73cDPtKIZ%3t(r^1-(#H46*a<z2l>X3PwAAm$?_BYL
zFaJ*u#S8X@>Q8U+mp`#5`s*A+PIhXXc9ecwhxLn_{5=1zo&1r%dHElFQTp#3X?}i<
zov~Z<@oU<RW7qh@i9giOI^3ry%TrK(*8149d;;Zv#5Hk@-{#+)d-*4Rf?t4x`S~yF
z)DAazSf8@}r3Y|PwjTKC4>z>z*T@UVL0*0h4%Tg5aPb}qf9>7{z7c==25<A>ubv(l
z4<Gp9uRlLdUinAd5A(=3Q2p#He&%-$F)z8{zz^_G56*hSX3LLwtneKr)OVOr-&=CN
z@O(7hM|e+yKjdHai~do*q}}sMR6OE``BU-1dk*G9`EmW7e>|5(Js;K{sd=r-y5)WR
zU8$#y_w>HBeIqlTfB5`&9J2ZD8}0t=-6zkRKWF~#Q=a(uko6AP;E-qTKIzl*4|&!T
z<(KR}dH2bC{_UNgnSaRXpMB+$+A8wuq)iZb&zjHQZq9v6i=^%0&;;IWo?r+pm_Gmb
zD<&_lb&5_f_*!1>8;?I`=G<4^RU~~f4lM)O=74Qh-Fe;(3$H8MrEN}I4x7HomD~N|
z3pdsJMG5lvjXwACn|8adHqmImaF6rahop(XAR{n$jZYl?v*kAxt<phWQ_41a1ddz#
z(jzXJbz3o38ue%N5@gRrAQ4Ce5`iZWnE!7FUj67VC)N8!-~6ksUUBaM&!`V8{lar<
zGjDo)z51}ypKti#^VYrOFSSV1Iue0IAQ2dJ1oF8}YtL;4eLsg)zHZ&W4=WxMP42PY
zdrvgHa=nQ_BG4RxR-Y@zd(lBHuRdw{rk7o{xEM6rd`GfJKM?3vewBQm{N6+}Jo>Ts
zIcFk}2qXfDKq8O`Bm#**B9I6q0*OE(kO(9KgN#7C&!-1@O=FR*T|9DJYtQ-nh1Xbg
zqZebljZ3rp(@wXyyPjTP{lpxD`ur*7YVUVAt>QM;-+36{X}d{&tv&Ziyb`a`{LZ{r
zaV6(T1QLNnAQ4Ce#xnxLns>E(F3oYnanCcT`!t??L9VTJ1O`=4Q?94?bFJ_2#2zu;
z^=R03x7Ga^w)LgF!;OIZ^x<AbN}mWM0*OHP5m>arcfRoE_3x@x(b=1=vt_a5)>^M<
zl^la!Kdkpy-IBZenwv8u0*OE(kO&L{0<GQ)^t=BcVAe?)x)sNhZ@w2bn)jkoPf|~Y
zQBOMQgOgVxkO(9Ki9jNd2s}Lkp3jYyK0Wu|)wcROp;r6%ZY86e>HZt<J$veTkJj3y
zo%;I=qm#COy~e87ogUvmBy#3NAQ6}-2;8vuYrenMx{uU`mBzm}Ic(+Y=kI#rUKcE_
zjg>aj*WXTCHgC_l_trX1<MaJya%6iVkO(9Ki9jNd2qXfXLZH=qnnCTq?11{ho98X4
z^@?^1|Ku|<5a{$hSF8Lc2Fkh0L?97J1cncRLCLFz54pjVc>Q%h{`$YYcv8JrbTCm!
zsk)6otA9^0`|mvOhK1J^y~d&2dXW4Rf$@w$ukLl?Jm);}p>2P?%Se%QFmY(6^PSGC
zTfKK_*1wtj+5Th%I=!FCuQ%U~hx3znvw6w5cBvW0yZ>(HXR`g$BQBYBTR|U^CIX2-
zA~4Yq7;PMwXt?BhJBvWO`C7NfjaL7My<cZ{Gr1)Ki9jOoR0Qt$(zJzZyzQ^GVWrQV
zy~lt4^&^w(y{1n^vbCCFiBIi*C#N+uvUeho2(*JhyY@L=$Bp$neXZ`nX#IGv-&g3?
ze4|}&Rlip8&Hn#DV7%IEzt)v{l?ZeKfqp%g?3RCTmG4+FZ1+3;W1Tn0CjyB;BCz_T
z<(po1)#74MX}|KLoHxIZ{B++(?w7}?I!}(*`hMY@^NueB<k!7kRjcu1wDN86-08=>
z;+wy#b($`iKL7YDCNHiH3SDvP@t@lDv1PUKLOcEawPE3R<f5Oh^^+f-Su~^lx<?q)
zytmbRi0uE5-$jnrk7mC_U}Xr5^*LJA_?6B1L|ec7V9wjy-v>6+gYn*8Wgl7p{O#u4
zx3owc1_OujW{0ib7j^r6_E!1#YCqEc+Oxl^{b_ZdCrbZ3xAo_HHsb~NR___J|8P9l
zs`8`bwLbN7IP`L&?^EW5?d64u+vsrXmw$iBH-7%Hm%sYK+Mv)AzVW|qe%D^>*GEe~
zcIs&_yY0>E)VoE0aLTVP{FjX$u8o!s3Xbi{@s(G;Xop2hFRe`!I%v-U$9n$lmmkS_
zdy7D?zRzr)_<hMvmxn#S?AGoN3ohMSfAUQP1_^;)pJxVX?I}wlkO(9K!;HX0IoEb-
zox{A>PsrZtdk^EK|FG{{%AW`f0s^=Ec;Wl6*<)J08GYbiw|?)bhi+6KE#3O!Z$G-;
z()H`(h3+_YtKHtX<<xqwX{U7LkZqnf>zTW(RUdEk`uV$_xYq?sYwgm5-gw(l`<?jf
zBI#h^&?&udmCu~_?f>taUvg)WG;!#p9vr;%u<hq<ytL>Qee;*roB7Aj+*NdowrW?|
zKM_a-5`jb@5l93Qfr!8_-nRP54}JUYTDx@mCRc9vi!a<%8y33vK66fe<tImK?b3r6
zT{nNtZ(LsMH$8041+xx1>7v@O(Jij|+wt%J=C#FmrThH;iVbEQeMZq~%5G=hu=6Jm
zTy|rTbhvSN-wB(1?DDVPTqI2#QV)7(2mEN#M4+Dt?0wWjGm8~>7D*F_Vbp`S@4V>a
zFP-}9B58MVXw^r3<kGhux%pYQ6cdHEichEdA9&*(+dsJE?jmV_aCrOAXYa7jm+mf-
zj&B?;+yBp7ed{fk7fBO`)PsN8LA!D6l(#(ZLvwDpy%<#5?z(bZBCs+9KD+tthd=m}
z+lqGS%6NCOZT8-^9sj(;okh~b;R!w1_l(nLoUr(=fup0<`TP!aXTJkITHLc=B9I6q
z0*S!b5ZM2)_inMz`);pwi;mne<AU$~*Pm<MqGMZq`=cLz(?y?t-S_URwM&QXcOc{C
z_q}g;`_=z`(vf!-!$ybwJ(@2cKI?<W?R!VjZQ8y?=g35$bp$?n<;xzJbIhSdtF-l*
zx;tXn;!%|Mr25}K^!z<<FNU3V+dp=T|FC>-FUlG9sNHgRt#fd=zxAB0XS{ITyNf}g
zgM&;;k_aRMi9jNd2qXfDKq8O`Bm#**B9I7-9f59rZ>d*&_rIjr;gUDpUG$0`wd4~Y
znY+!c#YCaITzukIkNoK7+IXWgZvWp$?)c@0YQsubJ^#DMe)Lnntqm(}7x%GyyyTDw
zBm$j6;9ZY@e!GhoEG?2I4ygxe2k8ge^8=mo$F1_&;JMR}dBrz>SL-xwRgUbR2qXfD
zKqAl{0$X0Qe4{t4esz&_V&Tx9w&bWpAQ4Ce5`jb@5l93QfkYq?NCXmrULnw{_f>M9
z-o78yX}o)-k2z0&5O~3bw|sWiv#%|ZCJw0w-Lr%KSa;6aE&_kw=yNZ>X}9ZYXuEuJ
zTq2MNBm#**A}}5h7%$&T8<cen%XfXgdh&iBxbW^fit$3bz5mhHyY*eZI8T-nfkYq?
z=oSLq@?*(&IKSW8?|ti*o{yIA2VPX)_PF=`rWo&Zw0LB{@rl46ryX?E%g#Hu7<M{)
zvvsyCmfTuvm5$F{%~d4=iNMMb==A#(!_IGISY=xxkO(9Ki9jNd2qXfDz&{b__VcFX
zf9VmI%(|_ZC^*Ew_mSmeX3l-pT}8ieSbfs+O)tA@agj7}7>{}|ca2XR{j=pa6-g6^
z@t_C)%*>wrzlpBjPe!AArPJ4&@uTM){)bwxY4=u?oD+eGhrpow%Zaxmx#C115$H7n
z$DTN{)gGU_w@BI=4!z!|oG%d=WCRB7{(6wHO4$;DL?97J1QLNnAQ4y@0)Kwi56_yi
z>fdU^N>4v*hpU!9e@cB=>Gxhg_2Oq;@yA-X>B@9qWo^B>$0WlxtL{AShK1J^?b1%2
z6UdV^5qJWDe!V}_t$BN8*S(r&*#Y&1H_uy8OFEo5q?egk2z1M@HRC(6P|npR0*OE(
z&>VqYzi&3)=UXv%-gj2)^H^=@=&8qlYS+h>)mHR~HrLGTkq8VQ0?qQ!Z13;yk`n*K
zzn{O0OZ{89{&ngef1>kgHFCaZ&F617=f0&yziF#-Xa7WCWeD8yrD+S-c-voVqoph3
zm2HVYB9I6q0*OE(uyO?2eUB%O>(zS%<87V=)8`+5#pK1cr2WAm>hZ9Z`*lyTa<7uD
ziNJ6n&@Ve0lzE41w^G_fAQ9*X0`2;%e#n}04lV+n`aR7Jd%x!UYpwf8ZM1Z75lP7s
zfkdF+2<&^t=`&7Pd{;3j^uQbM*#5yKcNd+a2QNKr`*|BLEryluebhrUixqbk?b3eh
zSI(aZBm#**B9I6q0*Sz&BXGq|Z~FX;Zn>*AUg*Ex|J>ux*y6fkywYy#Q@ii$c1x~q
z|J{;7TW_~;9Cp6#o(D%eZrJfj`G*sMcKuJM$3^toeGZi4p8j_=5#Mf?b0405AF6Qa
z_I~u6f7QDCUEgn9a{ff17YK|tUiaG`do_Q*f8SU&e=l?)=NLW&p7-@1ov`*7uPKI=
ze);fOA3Sc~JBneWz4AxhnrF2452NijR?DgKd#%=wRgN5=2qXfDK&KFB)qiFG!T8Qd
z?nCawAbjs8_3$5h*eQRUd=h~~AQ4Ce1`~nN-batM-(apjrAh=kgFv_L4aYm*&Y+)M
z5`q3AFkb95qBtnu;fiwgt2|!zIoHwJdtrl$SGVop#r2K%IpCWc*2hYl$vLdw#XIMb
z4{iJFT}Fywr~T6JW_pqBi9jNd2qXfXLEy+mKV9o5KRmPOHf<*DSnm^C9p6lzY)=Fd
zfkYq?NCXmrL?97(dIb7)Pta`Mr<XQs6M;k^5l95aD+1mAE>QAczUgIGEiMv=#3A)y
z>~=6(KfL19<3F|QW6NrTPTT!ETBF4w`y~R2Kq8O`Bm#**H3A>n_L^-rIqx^cM5Wcp
z{G;V7uYA!Ci<VwmYn8T}-?Tce+5Kj(+neDzXzla=?ZB%a{pF;3r|6)qZ&>7MW?#dy
zj+C=K1bXeKa=yWQkG(zZ8T+U~`MpZJa-=@@N1yY%A1nX<%-Ah6IV1v!Kq8O`Bm#**
zB9I6q0*OGo2poIj$X0uN?%pEl%5i8H-yD|+Bm#**B9I6q0*OE(kO(9Ki9jNd2uuV7
zTK&FCr~0@1+&ss9;^?0(zo|$ZI)lSRu<=}DB9I6q0*OE(kO(9Ki9jNd2qXfDKq8O`
zObi5;9Z+9*^SlMMq}|0~VrWyYG7(4w5`pd`(CYUUvj5KWZdiC-kvNPEhhG0axBHfq
zGb93uz_26Gt>^y3j!Vj)2qXgiLm<DavGVU~M11<UQ#o@YkO(9Ki9jOIeFS=aj}qtW
z{t9!3L?97J1QLNnAQ4Ce5`jb@5l93QfkdEx2*h*ncq=!14%$EDbLK=K5l93QfnFdm
z=y9!Coa%P__49W<ajy%O*7{9nZ??{s#gbcVNfU=b)Pu49-SckSaTWhw_}H8y5l93Q
zfkYq?NCXmrM4<Bs?6=;?m+v`agL=F4^z~-^=sAb~q1KFcUh?Fa2uzyPsqeyeo6myj
z^N+t`^5R;nbkM$=(W?BT?Vq}w2qXfDKqBxD1P0~#LheT*kO(9Ki9jNd2qXfDKq8O`
zBm#**B9I6q0*OE(kO(9Ki9l-z#P4!VlyYk*W$$hyFeuL%QjTM0&VAKg#aMCZw!S3)
zL?97J1p19Yx8K)F{+Awc$*kK7IP`nJbN)mi5l93QfkYq?NCXmrL?97J1p1A@r$2hw
z=70bDW3^$USM2nr&%fxFyK3E{{oa>e&%f^(r_VTH@m)pI#3A({?I8U?dwyWY9Y6l*
z9T#0+v`eeLSCsXiIQnPHZz=`_ha+aaZJP`K@|&Vn+UvNU^Cbd_Kq8O`Bm$j4VBx!7
ze9>$7{8KGy;xI^h&<Wj69*ICAkO(9Ki9jOIUj%+|%C9c`|LmQ4{FhVu|68>Zl8hxa
z49XJumL=wX%UC~zA|_js$`(e5M7AtrFD64-QVe6unx#9HZOYc7MY~odm9$CyejYqO
zGmegPFX!CL`+fb>t8?GyT<1Ekb6wZD&bjaR{n=)_?>rMW-f3*JG51dLOc?e%<+jh6
z(q&V6&l>~0w~8I^Tjh{jCp<sS*CF-tf_a+{(N`D{2801&Ko}4P(r2LC9Xnk!@}E<E
zTX@2RLwa{w_=PW+4<6FDK+!-L5C((+VL%wjKLZc!|H~Nqb??`)>uzO#DBHfQ;;sGK
z^ee0M)4yHobltjZy|N0o^y^yxr}7?U70W6f|3jBP{klGSd#lM_<q9uOIIMSJzydER
zv_*N(TPqKL>iBQGr0|PIy<1ILGTSRQ9MSiwmFEsx=G#V<CtLF#RnCipw;#7f+flv_
zqsq-nJKeZ+*VEQ}+Zor^nA>BrXA9FFPtyJC=2dq+Y5k<XY%%cuvK2e*x$l3xv=BC)
z?|#>_9Ufflhs9~{j=6a4ym`JY?DfXIb5E%`(GQ9>>(t!${^qm%sF)V-YQM}w*0J%$
zxApx;&+mK{<E|G;<%>c4i94RGwvTIHVdDe!<Myj-KW49UcE7Cq!(Lj5Ub^)?=yf$7
z*eau+Ep)XXwWXiP`F#e0uG7CSB|mqA#(TA1en#j<)3o9#{kj=d`cl>Nm$2ouM!>wN
znAJQds9s_D&r5xso!H7NOmcMn2fc2R&;6kC)Oxf}8dW{QmV@S#rg~xfrGC<Ux#};d
zA4fe`aNTaqxLS2OPut*!#e0rCzuB0!t9)A+_xXg$PJ`ZOpP6?w@rfVW#-RRSs%Kj-
zktYl!!GN0AhAqd{(bc(1-sI|dbiJP1mam?Vhb`OIC#bxj&t39bFK9jywVuv?gI*7M
zK9_X8p!Sn?y`Vh1u4gO1*2!_LzfS(%(~j(4<Bg48Slng6oL<$gU+%@lsD7E1cBSSY
zVb?wC+uFyO_EUMguk*{iT>rZI8{52c$7UmRs2fMvc#~HBNA5ah$}MX)E7-=Q@_1Y4
zZMu0>uiNS`>vGg_>&kEY@AO}NX}yAL%sO95>8G2Aq^@UW{aLf~v7c|)bFCj1lir6&
zf4_d;GOzpayz3n`kGGZA+LEK&UxvM&RJ@davcW)B)~&F5B$ZdV?x$GR!MJhEwKsHZ
zd`tT$s;%?FqE4^4<F(yBtR3Zj-OjuIw&!1!r?cCj@6)-i=lUF1H~zTi*1GmV*9rCf
zN2jOl>-SzcYR8tRuJrPXLHW&ky`<un8V|zeS89D%db^GTs{FX)2>U*X)c^JGy2ARW
zI)ARpy11U#>$caApz=)b6&B0=pnPPlURv=eX&!j~T<mCzS2bS<TDH}{D#v!5P~}rL
zosUv4EULdl%xl^8J}_<hrhI~VkE&<d^B(qGNUd+mmv#I_U1xOiY+YCNGL0|l{h;x)
zSnJ_8X=7Bs)wPQ{9-X|rUsvOWT81yvxx3oVR1TWgMSZ@jo_{1&ww?DxmB+k+rgbcC
zd=7gqK>e`uYqG1T_k-G_S}$sUw%e(Csb%aSDt0T>um8={wF@d9ari5(23)yiAt<Wt
zb>-eiRk`rMS1J{PV$^ZjZkJS?(e-0{JZZ_Z?H{VVuyIFMKd;YY(wf(x{i=C>v6NZ&
z1b_bCW>9-@RWIuP(9|xhzNY-k)Ss>UCOx<x)E`wIO#6tW?&p2qmUsPaotvokl~mdG
z@0vcW)Oqd1nd`l<Xxi7Q_ig1XQF)HKpOtp``1=Y`?Lyim!+>tz81;HGTK~6~pm|5!
zdg^{LYWcsK|NK9boUH8&RQ;1K+sYG=qua;3x}NuS%M~w4$+fjFN?T6e?I><MPP={F
zxSVv{)b(d8|IeD7wBF0-JVfyK|9+V8vuS?v`hId)y|cD{splfD%DOll^m^31BWivX
zR4-TcvKA+7jWg^#+UhrNa=t&|$+x;SUG1ffS?7)YY*x@UUvcyAq}!X$dGc~U?RXs5
zuF}@ucK>1RCN6(jYj2xR*YfiIoL1FeeXlTTIW4<V&yk|)p<8#u+LKO>>+7c9O@+-@
zb@f%g(<+CJo5fx~>3qm!&nCMt`ITurY8-P_4tvfKcAp+KUfF7Ang{j1t$uCEG1*Dj
z`?~klO!j5+2i<*@2b15}zOSCM=cR0#SCv=S<-G8gH9eV^sOz&We?SkxUo&8edr95@
zwN~Lvrg#zbe)yh$zeU#kQS>>icVWN+FE2d&zIFZfu5MqayZ&1VzrGSQKUVWdwXBOH
zLH$zKp6hX+KLj18T2Gfxay_nIVV`pbea@TI`yjuzruZctAJ{=q)X52YJ^o1SNYMFp
zUaqTg!gV>QePpGc9V>}1CG5F}?fQ(s^m`lC?%8*!*D0I)U5)o@IV*M&RS$T`3Np`G
z$FP6zC9Uz8?7P(NtMg!buA;kd>s$gkQGd@9wViHUQLmfg592gp(Dh%<bEDd^?R{WW
zJ>s@A?W0WVrtSN-_ZRrBF0V-IxwdWl(aB@{K~cw3sa!YtL)!OE@ly9ZNq3z*Hz=xc
zR<*0R{g~oB_k-%I(<kipsBuDV7q(wh{jm9@s*mk*(0M8Lda7Tj`#;Lqb5f}4m2^33
z-kFtluGYb*^N^I?*y<;)-$&(1*Uq$_hmDV>^_lxRebTy)UTI-k<IZcpw*1kyoUEJ&
zhSejgpT}JfT-hJ_g)4NWH*#I+EpmVV`%+QQlU(zv?mJzTvA?XKDL={jebfFm?ERqg
z9(Vp++jH9T<MI=<z3yBx=yg*+S-1b7`lXe3+v+#yIBnHK@1TBUIww14ZIerzRax)r
zFlgMNp2==?_Ze5-5gwApp!T5FQ=bc(%C_v-Bq!|t%p}irUuTzDyKWi>_f6}JdL9*a
zoU~V;f9w1gy>u9MzOz<8Y(A%}A9fy8e~w$u8n3qW3CnBP`Z~Lo>p%Y;1Xn!R@*{P=
zb#cseU6oU8Wn2CjRZdhsS+7bB<DS#TZJ%@=9Cd!9+LfzzN%JQCZTkC_#mdX0SL`wO
z<5ks0<PFD+t5v7-v<<$Bas4ptb2Ivjx{l?w9phH<<B^?*bUw1uh@{bUzNNb#^}QN(
z+^)-E<Bw~7T+7#uH>_Q$^NZf5a}n<2pXKcar?Wmp&1YP-)9tsCx~{X2sMk&Vxvbxh
zo1aVjn)}}0e3l=Shp2Xzm3C>_GjTL3hRx67)(>hAN!3d#zDAv&q}sXiBa{9&G@W!s
z@1d1PxI$O!wMninKhElPSM!ftlO2#BMn&>U6;1snbw923!B)R<<9^oKyLxWoT5jCu
zkf!#w>?N$6sJz+s->C1ax@u>8d^&ka<v~gHlT`kw<CAsI7Mk*nyxdpU=e(8k!n01#
zr2NlSKkE5oUdyidaV<CO@0x<v*Tqrwder@a?RKhOVavAGg{0*qwXVAACu{qVu=+&J
zlZ&lg*!_jnzqZEQ9+SPS*+JO2X{)|2KT7L**mc`ge2gkLuXz}H2hCG#)uUfm7<C@g
zY8N+;2^tr=73$aj=IPpnpvd>L)5fs#;>xaq$~Dc0?eFrrl9#ph(v?0@*FDobN4+05
zFHWm|)bXX&&UGA9-yaQI4!XXY&JE+PC$8J4HUCljb@hBhmCL;9^t3fE^kW-!dUHK$
zoTXjZeS)p}=&z#dc#@R-pyvbX_+6FLTF;8LUsrjPPOdAv$hzF1`>Lq*Y=0kICr>x8
za^3Ckj^pyH^N*nW)~wVEng?gKUf$P{tn!Lps-5q5?twR)a`HT1hqUXt#rH$LTeQMA
zVN^df<(s<v-Sl2s(ECx}BM!SyrGDIa5j5{q>#6e}wH){F0w&#F-7mT-FKRfi#zl8l
zErdm#ee159z6+Jr`?~SCzOJ+LNA5ah$}MX)E2wC)Pt|`-=j^6+E$VwFrhawzUD>~D
zxlz}bthLk4OIEMjKJQZH>HJf!r^^8Op9znuHSqZ9FMr~zsLpfJWmoKx$C)s$Kbv^Z
z>V3|gvxeA1)b-ZX&eikeuyPrvpfC`I0bQOHbRVtObDh^CzgnZeOnYdnVICNFol*1q
zte1nv*&(NmpVxi%B40(*`HOnL#Vg$pZqu+z!8XQSFKMr{V^`NzyDzq~Y2Ax@pV*|@
zS6<3-^Dp9;EsT2Z<H|2w%eBQ@T5{s9Q&HR7iceAdkJ?T>&yHKp+PWTfJY{>nHRjq6
z$9X~hKkR-&>W|v7<*6$@*E~e^i>$PZYQI@&NB)>sbiLlXmcRG3Bm38QW22WAuH1M=
z`-&gE;OkK5|E_r{Hu+)mt)S~#vD6D1&)N6e+Arlb?x}ibWjzV2hdPeDmQByQOXa>U
z&ZV`!>H5{Z&l>c)Ze0v}-S+zfruR5p-4E;cN!QP6KEZilSj_5p(92f5vvqD0RZiG>
ziCW*4UnV6t>UntBc5&lI*!c;Xmu9UVewcUE<r8_kZhM`#C69bt^~1RK9M?~E?QN|q
zCOPag<0Aa37&Z>su5WAqTr6^I^BY%Q*5jLM&v`GK;z7{+apyg)_HpB!+CHsuxa!yR
z_n%Df#m2qwYFw_r8<6!lrs@|r@5xI0yypEu<7`&zCB?g|cz|46?}>!v!`8eeotH*E
zU$WiK*1k}c6Lp=yp0a}KxbjxEZ7)g7%gXu~Rgb*xb8N-KtoX62N7!}7RG)nRo+Hn1
zHm2<=U&R)W9CgG$8&n-(LR<Etj#u4psby8)u;p{sHo3G}mGyp53>!y-@}iC_XgO^D
zqV5As<)C~99jC5d(D6#WF&D3$H_x}lL)3Vw`l;=5P=CxzJ)Pa><+>?O<mG<Ud?>AU
zN$0J$`^$@6nDo>=r;B@CeIA+BvhDb;%8R-V>Dt9zA9Zp~*G+cBeRce4my>#bM*4a3
zt!_<Mdnylc<A~bcWKX7jVb<^4IzJUT_s%_~=0ra%55%#g&}7f1`?~j|U0;v-9fh#(
zSLda^?ff!nd3o8d+169%Uuo@&O#SNgje0$)IOnP#Q{JlH4|}e-X6Iu+->~OeKPbkv
zPm_LO^Z&5=q`h96o>Qv%pju8kZqgrlrOMwnt-BrDuiH`|3_5@6ykW<xd|erTLoZvX
z)87=gxUY`ORJNseUgfCvo!2scl2^=|{Y1?N;<hubUs3Oy;zX(5kNZ8Zq}oTFcU`-r
z{gL&{73MXr2i04bHw3*-K4S~x#yQvR(~f6B=f!qCSN;`MuI=%qEzcA;ZQFIReBY*2
z$D#8ZlO4IbZ`(eR=SttC<wo6K!9&(iXMcIQ9yC8Hm3p@9%XChd)O}myQRSqa|E1M`
z+<6OnPH(%Ot-MaPgQ(@V{dfAWzO-H;C}uUk(e3}kUN`X=R$kEhrhfBwKW<!5^C8oF
zfogrV9CqD|tAAG8^L!|6jLT2lyiWH#Ea>&P`e(hp?Rb;6z1doKkz<NSS-&53|0nHw
zx0<qKwilNN>?p4oH~wa&eb9IxwO-gb5H?TATK%Z|GTZBoPF~n`S0^Xv^`xGg+Uh6l
zI-OMf3Oc^9`e&_v(EVcAdO__~S1)e7*2#?;S5577{Uv?fq-WgwLC2%+hx1Yn8(*TX
zhjH_{wA;Hrmy2tsw%Uiyr!rfA@b=@jXgkWc^`43=Ibr=T>-BBfp-zq(k6rPh_LFou
zZa$D!`>^?cR_n+8&M56&#SNWYo!^+Q>->oPUAI37`uxt6Pt%WzCi_Us&eeY7#s?k$
zS-Y<Ot~_t$wDZWQ{U>b)w&pdeU#M{)D`ngJx~TlvZfA<KrFP#ozamc<2*Q9a&$4x0
zr-SYD4O6~n>%Kakp!sNC>e=EcX*ue=sO7wTPb8?Gaq|NBQTG+L%k*PH-TtyvuB-e+
z&D%@0oyuR>a#H&plU`YwKa-w$yKg#=(%pA8&Y<6+PP(3shrC=*JI^BTab-8Q<mP={
zwY5J9D#sS@COMDXb;^`m)@)XYi&4MR7q^}3?++H6e3L!6y1%I5yc!qXS+$Ti{IF8z
zwG(Hq_gtZ^_X|w>sHpF?q}?tn`Ds)=(&jJi{!RXF>p4u?a^jwkq}@KR&p&LR>zd^0
z*2k#vA#3fT+M$jY+t<_DZ<*q#Zv8d!X_|lA{1uD5sQFmd#&2t1ob<Tj>TQa%y8B`O
zeq7Z0N%NVv{-W|~t6kps5`Be%O)*gK!S5TtefbYwv0zf?_M74|%(bNSq@SSYzG^*F
z{;114ZCy9@ANRg)otNu#dR4oAxu@enwe`69n5n&K|D)bl^^03h>Up{Aet2FN6kY41
zlOI$+T|M1-l4=JgyNh~X?I-DS-1wWczG{D}Kj*F7`=}}x9{5V7LS8ZL^X$0&o6fbe
zdOxn;W~F`J{8iOAE9V4R`CU%b_p)^BdtR<b9e2=n@Sv`LLCbOXKSA50Pgde4`hYsV
z;a7*`cgTr~S&6%O(<3V1YCFz3iUqODV!^E2r7k|??Yb-bSN$!oWi`(WTedaswDUT~
zq04jA@>leVi<ADa=GBHb)G3%S>h~Ulwll>K-T9U2y2*Y_{3X4vnDmKzUw2NO_3L?m
zzgrhS@^U@yIaAp7aqV2!zRkM#o_@4b^@0iaKjed@CtbLG!8Kn0X}g^d+_+sKEQZ}*
z2L1i8TF=&d?@{H1%{!<cwZEwCboQpZZd&ic-lreN9~8s(W2-*;M#Z4#F=6Y4)hDPQ
zCY?vb?Qi>btG+mHYK0MLq56JB*s>|_%lds&UYphXY3+w>_v`9>ASt=F;&jq-&RN^!
z(q>iG`*}t6d|h2n)caw}VdJ2#e%$pTZ2P$JO6Ad1R`b}j$|gNs-4A;1;<{eY@!4LF
zgZ8U?|0(MAp#4Uz7j}OVwV$lF3!B$dpK+<bcMW^~6*TY9TRq!(R8)D)Q`Ru7|7ERy
z-t1A;*LD6J)!x!R#|^4?*6cB~J~{b&8e#Kesh^L(lTvK+6gAJ4b~X3CzxgcR#6!@y
ztk#Q~H-&8%_WU7g{jhq4t#8{elRoz{^`|?>(Or)^Pr7#4r){2H$us$7)cf!d6+fQZ
z_u&RdS1jl-s$5e$-MprK-PQ9Y6CdjRwDWiRH^l+l_ru0nTlK@f4-&V2*z=vB^;J8H
zTeih-R5`ZoD{Xmk<BRQc>9F!_JqHW>J+8RVYfSCKt~(}qLGPR1bLGCP_YKwfkhXt>
z?LVyjQePc+*2+8|$Sb<aqixMo)O<Uwc6qe}TY59^<fU<uI1m-%_8+%>(Dx~`TF+H{
zNJ?(hd^Tu1>|Jov3`F(YpzSu@Yx-ARKW}s`KlJmp35WGA3|Qc$-A;Y(>$)7((!%D?
zLF=pf1dU@s>)Ga0)i>z*iR*f*UO~%At!qj3<2uf`mT#M<pmq?pUfcQ8){g4Gb)j3K
ze*JHruKgP{@uc=gyokFG3K|zo^>ojh(9?wKIgPFTN>Xy-?jzD}@0usoeuI|d+PAb1
z_CA7+r?lRWx7BY}z8@IW-fh*(>Ul#@y>#}dyKaj&lN{4Jo%Q>+^ZTIkOuX8@Z+ktn
zJ+Jc`^=>s~$!yOxns_zkbxGeh=^OTb()!!#PtAw(R@TL7+t<U|gKa-{EiZ5LuG5!!
zXTD6&(RF{%HfUa$_I@gE|F+k!q~$TsK~ZNndAV*n-;n!bE?zrto^P86+x8LEUUl_s
zok!~aK3?3o61Kgqd?sx<Nyk6i{khIFl9C^E|DV-*uAfV}k{_2(-8q^mo{;zHFs$9F
z-=#=9uGs2d_xwb6UHATkj%RgUvK60>zOc<7-+6PYpI3x0+j&V+^6+bQe$_JL%o?hC
zs^z43jM@);>f)F>9$nt(>N@%r8`_Frs+_p<U|J7t-;b-WslBUwq}b%Dc9ivU(C0R3
z*TX);`i&ZQU6s*K#iZt+eqHsa%QJ0XXB=svtvo(0Ibnat7q@;|^JMB*w_hr?>n5Ip
z-q+=|ajysMKW@FW-&c>@f7E@6>vpWisvWQ&1$FDL+D=#2<sZ~n*CG1TokzG@Cv|e+
zNws6!&n=?LQ}uFP*3D;LuB+#dc`57cGcVVZ-Y>fLgP`v*nChAI%<H+28rSn$PRqYc
zb`kczZU58p5>~IM_4Bgc==991U!!+k{TaQZ=37~5$2ii$sD7K(c1g{vY5eRrl0uU{
z+)qlMp#7NgD((lBmzR3FbvLWm^TMl6PuqD@((-igA<Fg6YbVZJ@7dykIGGgM(mSr4
zw4RsP?pK{R^4z3Q7k_MB=UlbeFs=MIYQL_}^;P+*T_;`UIf8Ap<riV)*or&M1J6MO
z@11)}&53@}JQT}w<)nFw+aLDt3X@v5ZTFLx^*Qc$+w;;sFV6?jGwgbn_4>B+hNR`?
z#h-P0CUssBbREb`+%Vah?!BU*JcO;M>o4r}u>IPqulu_?dApvJKe_6MbtbQ<dmby-
zfBv44YaVp#jH~NW<4;zfr<nALdM=f<c47aHjIMs%eQMs@N40-bJI=vOej9fk%G&<K
z#Ea>^$$xF%XP#`KX?#KN+tR04<iJ~tSGphEreT#rP>lQhE-UR_=N%^b@b3!qzK)pU
zd|vLS9WPwjmq~8kpL6K)yR2Wg{d^#4d3o_`ou01auF3E6c0X$Vne;eK=c{?YpS69D
zP9Ix5tM9Ft%C`DVTTWDeGqnrrkEV4rth})GT|d_etB+bgs-38ET$L9!oLA$bJF6CK
zp({PJF8AD81CO8n@+W@UsLN+;U5|S%=c>Iau4Vmx-26Q4{i&&c(>Y4o_f7ie<-Th>
zj~j>6p69sz+p>eWJf__~soy=u?Z@>#-nD#N>vrDcgthOW<5T6D_KRuXkDC_+Z6DXq
zY_$)&UsC6@*vdiYFRS&=S=;2&W>wbvc|}`xoi#Z+dx`5muGhn)<p*6?q~7-9wrD%b
zkIO?|<AkcW&c968!>)g}>+}7KtYh4LoZ3Ea+|#vpl|P#Hp-IO{wZEwEg(4?z97(G^
z{%HFgCaheOedzAH;!WotTxWbHw8blSW(!SmJ?#CUal`a}jw$cd*@5nQ&~?EyPbNDv
z@nw>4YHxZ^Quo}N>n8oVk9}tagW6TxdalRo>N%1txncRheoSccgS_82tp{Q6tMj0X
zb6nTO5nJnqPL8Ul?Q+s~VR}zQy>D8#isin#4+&bfZ4auvyz{G$$5uIPyvbVqu<N=l
z-jG9Fjf%(#il+55tM^TLeb(=XwX>|%k9vL)w_Q-X4qMN5UK?~DXuDq4;;E|7m~pl0
zbe^`sca401L(Su|8qcEYrL$XG`AJ-V)wNgs)Af7GI{B&|rgGfB50`d(bsu3W>-dP8
zpTup)`BG9C)h{#GZo*-`3j-E-#6=Z#^PIKouJbxwys~{=7q{j5oL<$gU+xJHKl7mT
zKU3bQ<~>2ny8d+6b?0+({pauQ>UdMH$6YT}c~UM-2GsqQt+L9dDgH#gA2dz{t*6WT
z)$8i`Y?b5sbyDqh&vEj0y;Rp7TlN}NPSEEvuIpv(oHVLFarenV+ee*mQ@g11Y?7Pw
z{h&PR>V?%eDsNe7S1j?@Ro)cV9^%$l&v|T>qvm&6Zztyt$(+B~vP0EB;+_lZ;yBl%
z=HsfKx^h(c($3EBYh3fFoA<ca<J!HnuRQ#z<G=AtJVc$RsCiM;{<6}JxG8wf?w56c
z*t5-pt9dp(7Y=*hRz4k6PFD6qarKBAH%hHtT6ufi`8Bo2zsXB%#}DL1J*T1_dfLXk
zoj>dJP0K#x_M7$nx=BA<e^1Hw`yWZkv-KXJNsev5FBW;G{e-LguFh9X^AgssS$~p3
zRe#g^oAvuapMz($o++**ecxoirIuH#<5m49Y}r;^(#Z*{kFEMi+iBST;;u8Q9fU0h
z{d-7zPdl=IjW;%WaWUvQf$e&6*D>}}QBk-5FkO$@pRQe4`;huSzb_h>hqUq)TX8h3
zoVamK*FLTvXQh4Eb6Q>fwDU)uUO}&iwX0&OZz~=Y3m>k=VcItYz3)mtkt+-c1Hyob
z0n>NiY~45MA@_gHfbP5_Z`b4Qw|~vbZ%cXM<KvN?hjc!&(g-kUo|ct;1$wyJ7nsf~
zitT<}zpy<I%x_xwThC7CO5Ax#tG#+I7j*xTm3m3dc2>q~(!=zgLsIvXnwPNs=+5!-
zcHMRyPg<Vq@3Xj;ug1;1mtFIv$`9+ey86YI-|5Dq8-LX6VV?_Ut$y6+U$hU(pK0B7
zbw91VI&8l|c~tAU;w`LPRj;UJ-oK8DVR={UN9{K&?R4=k>({fwpDjJ$JE$KJPxFed
z*0-SiQIB)JsF?J3Lv8bh-3iv*_x|Ry{G@rXJ)cR-OPVj+{i*A@lz;v`Eb+0}G3@#s
zmXDz4%wgpwb*`xP6PB;M)DQZ;T3Ypz&NI|`&uTg8b5vLTspoZhDd+th)^<J-R9@J4
zpVj)V{2^(%>UU@JUJm-+N>=N|{e65=?SuY)N>?xJxoKAG2aS7KtCv+i!|Fwzt*#SE
z$5CB>s$OBsao-mR+n-wh<EecgZg6zPLaC$5OV-L+UC&H<1wALoYQ3b^Ti5;AdQaKp
z7q;Hl4=cx{SJeA)<8#{WUE7sw`FTIL*5wU(xt@3a(cf15OIl9S=aZ&588&~<YJKcN
zMaHlDj$2mFPuLd;Zu&di%wt|KuAjtxJ|4IKwA*Lx{3UIDlFmC_=Yg)}XLbLn(<|xu
zb=9Bg{2=N3I=)TUUH^Ugq~)jeK7i|f<F1oxJ`whuBI)``?N?M@inZ+8K6Ucb;#1up
zNjco#)rdQ9#kRjoJ1;i%uiCe1-OtPYy#Agd@Aj|Slax0dKjX$p+9&-!sjdE!_D@qB
z40}IrUY(Wpw&pXa9PBo2OnRT4wf@rPU-~Z%29k>NDw|2?QELCXa?rT1dmg8|KIso@
zUTt_ooq~$G`ljc4I{l0NdW%PnI^v%Vs*Z4lLD$Ks_0swshx&epE&t2P{-N0Ip7FVQ
z?vWR{@Q@bT_TQlEjjMXL;(#j0mA<a!s^hR-R^xEoGJcTMa}`(pWMv%*st0*9{Oi!Q
zz3TccmFsc+hxWG5&0YCP=Hw3E^!skM<EBYoTtCZ-zv$wRYyHBWuV$rwQorAiYhR}J
zy65b+ubcGXKI>ahjQYN%$*+^TpOhWB8jtOLNz(GpS=;2&W>wbvVbNrFs{PuIqfyU+
zbnTMn-Bo`N?EeeX{kr$-*mbwE?aOv3t9Wa_HvP&f{q%3wI$gK!TCc3aE&aOI|Eau3
zS;ex7$N$i!Prt5@-rj1mSK0PFm8`s;{U@E*Yvl}IaML`LtYOH&g#lqe7!U@80bxKG
z5C(*S7z{jjZj&=!KjKX<YZzlli8k#}<?Fe1n~(8|6>BZ}Y|fyA$NPD~L?>T569$9<
zVL%uV24XVsM!l8w?_1o<H(^@qaLnOJb74Rj5C((+VL%uV2801&Ko}4PgaKhd7!U?B
z$3UkmY7A}g$TVNj4iA}|;Di*C_cu5F{$^HrNyscprkj79w(E_<zFg=F+U22S3`hnl
z3<v|lfG{8o2m``^Fdz&F1HynXAPfit!hkR!3<v|lfG{8o2m``^Fz~w!{O!V-n-3W@
z!PBAo`;r5iUVQ91izoSlrNINnlh^lNW>$E7#MljUd{=nE*0*iHzR6r)P<Z%XJHT%Q
zg@K<Kxc%<u-mS9!3okGD_7Uf=8e8{czf`exo0c2=!>9U!!oyE?AW=*h5C-fs@Zt1v
zwYv?O<>w9UGZ9SU78B0C_p7rucsd-jy8iTrRcClQ)a~<i*SBfE?1kqaf5jKfCJ(`^
zMrsKI!hkR!45$ozzsn`v_dIo;pERm^NLd&V284lPVL*Lu7`0q1e8@P20bxKG5C((+
zVL%uV2801&Ko}?$29kb{qEvX5aTW&yw%#8t4mM;I!a%kd2>bWYvZcLfBMb-w!hkR!
z3<v{xW+1P>!_(=V^zTUXJolozFpv}ju6{3z+@!`K{RjiXfG{8o2m``^Fdz&F1HynX
zAPl%+Ag#aSbn>^uIu95(+YgFvMlVu@0bxKG5C((+VL%vg&p^`e)k=Tqd=G8qdzG3V
zG3PVC)KT@Q-p$*6(*CCTe%h$&C1qitlo{YTZ_-%GoXgmS0bxKG5C)2efp*tU`()J(
zYrL=+^n0CAzsnQ0UGdC_j6@g^284m)WZ;sAFL}31?;(Cxu{c>ze^kfcaouL4<}CAq
zVyF3gT;I6zaNmTb`u9WBai>4SZZrsbzT>8Mktz%b14%H@Vc5RAw;4Lh7yOxrBsiBo
zgaKhd7!U@80bw8=2Ka74I=%n4SN={((0y0l>iu@arG2Gg?MEN1)OVtv6)f%9kkJbR
z!hkR!4CIZ0pubBD`uoAK^~fW^yzwXc3IoD`Fdz&F1HynXkTe72O@cLsHh5&3pEVwm
zwg>6&X9j}qYoy+!^LnkE;R_GKLzeBJ)Z$v+o=*kUm*;eXmvrCr)Oo(}AUud2==~tA
zxUc^G1^K>9I^TB@-i3Gn@Tq=WJ4h>zNWak-xO?N+yX&0$p)Z(k9-_^@v=Rn{0bxKG
z5C((+VL%vg&4B!VHQeubUC)Qe7Y2j@VIVF8|5$m%W-a%uQg~#1z5A<-tE9!7m+$k}
z-7o&u`|}gKJox3E-x%@LE2n?;<pB%4q_F7$1IixQvO+;H7!PsnM%w>A1M_BHdfbD9
zXZvCCw#Sbi*=x!Vo}e2Zet*`A=kDJwoVodsK@&U^R@!Ucs?BN*^zw#T`CYH*VfS~|
z#p8jcIg;<%G&}vdE64j;!_u4$8MTRljs3b0+ivA2ezBry{BmCy2*ZH--donnao+>d
zwSVZP2Y&3`V76Z@ICy=h!>+9`+ZPlbvS0^cVvy7o2801&Kp6Nf12w+f_ut3dv(9_&
z+$Lwde#D!R@$Sy=54+`%*}kCgAa)>rKpYr-)b9OPU)jemcAU_9@=2#3{vTh5)86ho
z{<WdweZhD<{B{MCEC0*D?VXQ$^Uw7w7QnuLKC;OZcU3Nc|7tLA;>MRN6}F$d_t;0T
zt6Z2hbNO*6^sbm19{u9GV@BQhzAq>|h#i>yVC8$2njSIdGe0YM;Vq{v?fc+tzu2(L
zC)eyff5rq~Fdsbp?;0ro7Y2lZ-)G>si~iZ}-iv1XQStYs%Uxlhv>14!-pcy-E$-#F
z8vA*ZMlVkC!1{ffk2-hbI8TSBbCH>sU)HbF1+#p?P4iG%Rwknr2801&z&!&|-%~fe
z-!*1+{pk&>&hXMgSIoKoPP8k%MXoR)3}l~yw~sh~)!4cp`+~{v5cYkT#w`k64*L8%
z&xGDzx7eY7<5gZ*%)W(+hQfd_APoGPfvo*b_}AJ=Nf;0YgaKhd7!U@80bxKG5C((+
zVL%uV2801&Ko}4PgaKhd7$_|UqQ3tl?d*QvQuq`;EA^e|iyh?C4oYj)kkJYQ!hkRk
zoq>UE>+IWW=3Fl-e)aa%We+V}?CCJNX3|m^5C((+Hw=6@eO&EsLuUD5(Tz4DRTvNk
zN|AxEzsu8o|2*mIr8rkI=Hh1{E8qFDrAP74SUyMS`uEGFB0rzAFIuO~K-|BV9ydOl
z^0m7+j=j6ixgYvEwB_I4ZKDsZ)oGz;Lfzk~+@}4q7oLCo6<@Hld5G&z(th9F+YBA$
zXPpPval3c(cAvDrX}+I#bj=g;d-k7nUayri{Gdp?sAy~55ji!6Hh5&3FFZu!A*%gK
zJ7GW=5C((+VL%uV2801&Ko}?;2I8LEO8ad6J)~lpkJ9l#8L2QJ3<v|lfG{8o2m``^
zFdz&F1HynXkRJvv*!s5Z*EgB#3vQZ+{LG|iTFeXteJ@Jtg?o?87GK3|dong*Kp4mt
z1M>IWlKFdX!gDe_>%6Pfzh9^0MXn12!hm}QYAyO~&Y*+G`+32r`<bxc4~kmf{ftQu
z!hkR!3<v|oz`&W4mfv~E+!sAT;UNh-C<dPiGlpV$@3_>*5oXS%t}q}B2m``^Fdz&F
z1HynXAPfit!hkR!3<v{p7}({4TGJL*``S+n<LEAJ-7^sN`84g^YbZSk1HwR77zq0Q
zWz_EthkdUn>idXczmFRAJLJU4yy46WkB=C;VUC{^X2rgv&U4c3vci|>v1ta9@*C+V
z**GP93*Skui=}RFn~tvOUyb^{bz1GxXDnYDWaa&(e6c3l3IoD`Fdz&F1I5X}#sjuG
zZ0JLOEaVmYp7HjLZI^CSh>OKJyD}<aKo}4PgaKhd7{~$x1KZZwx7W<MUQo<}&Z2`b
zAPfit!hkR!3<v`*8Mtio%Rlt@ndL=Am%52;VL%uV2801&Ko}4PgaKhd7!U@80bw90
z17Uxs6;!U&69$9<VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4P
zgaKhd7!U@80bxKG5C((+VL%uV2801&Any#M{his9za7?jz_{6dv0>irLi85~gaKhd
z7!U@80bxKG5C((+VL%ut4F-aKcQJ2&-%j+MbY8EOGkoDec*v6-lt$E*Q5HJ`t=qKR
z;2%EK7c34Qirrpie8PY*APkfm151bRzs*TqR(e74;!ipZ{&1UbJh0gw1Djra=?`95
zyyv@xjb<)h?Hzx|b(@WvvrI5!Jaibg@9u4ej`9VChdkLqQt>;jzjGY4-=ui3)lb9P
zk3Lwb??gW?9?<mSW6xPU$=6}r{b<<l^-(|QIVAN>s2g9<>p}G~y{{J39@TnjzmF&V
zKR#mYhB^KPTi>?*`X+O8hu`mVN%uWZo#!Wow&I+~5e9^T|1$8m3ukUVWY7e!)bPIy
z$p8MAfiov9zw?l}FM2_-)BHWIZ(Mn}FZeSLaH-~7QXW^k+mKnlYaZZVP#6#fgaKhd
z7!U?3R4|<*abLI+284m)VL&~fE1r=>8_C_r_BrnJd7pY=F<NiZN*E9ZgaKhd7!U@8
z0bxKG5C((+VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV21=0u)9()RazC%X
zYm2IPDcY*dF-PUS)Y{qPJPlcUp111uf7hDt2?`H+u!A&wMEVp4ij9Gw-@9;Cuh{sM
z@dyLLKy(I<Iq~sMg^%v{1ce8&ga7(LbdQpj!hkR!47g#Sck_0iw7+S-ZyR5qa7^Qo
z|Nhc*jpNT~d*}}<H+Y~MmPD#BAPfit!a%7qAm3H}`Mav6%0sEzy^LEJNP~f_zK@dz
z7t*INAPl%+;F5<gdACdNA--#LqrFHi76v+0`Fd{M=3~69V^-H?(W}PL29Heh^Ug!D
z__K^d7!U@80bxKG5C*c(K-|BdC+)NK_u#R(fo<#T+iT`ruUIka@3i1YP#6#fgaKhd
z7!U@80bxKG5C((+VL%uV2801&Ko}?{1`he_4l{>rF~=87hKFLZ9T|%-APfitaTwt5
z0jG_+G1OZ0*_=TKkM{+I2eE_F^#fgekn6%g78$Vp{fc6fm&G{~orD2lKo}4PgaKhd
z7!U@80bxKG5C((+VL%uV2801&Ko}4PgaKiov>8bHyEZ1alh!A%{h4@_`@(=QAPfit
z!hkR!3<v|lfG{8oWQBphT{v^|A%iA(#foYF4ng`K)_K6V*?v+yT-|@tdA(N7@CAj3
zB<&z8@m%x}27Y27FXyRYzbl&5I#MkCh&PlL0}X3G`e3EL6MaG9LF_>MpxEO;X+?J#
ztuP=A2m``^Fi@NfRGm}t!x#KjUa??t&XSDEH3L63X#dA=+t2no&EMnt#+8Trc|p^8
zo@?GjzAzvR2m``^Fdz(Mhk<6pkMGs)=J&m<p($UH``P@vhuN|0WHmXU>BYyMvv`s(
zC_IQAh#wSd97r~*mTqqk?ElNy`*rWvvFmPSJCyBMR`J$;ZTgi}`sv@Ub-HfdwO(0;
zTl#ga|5JI7vWjIDkN=@dpMG5*y}i|BuX19{4S1>fzD=vl_HA{Dyrl5#?FM-}5B!f`
zY&ieT@4GBNsJG`DZOyyL*)#+DH*Vkj`U58TL2=VklDqoS+q*RG{m)^(Elf`HZ1nVa
z>*4PWT(ZCqib1u0`_soSI&<(?uT(H7FL|%mWBrdOcI&p#OAFy|)pG~+X|ZmuS8A9R
z?`pryL)P)#MPqK8(RY=nV%GH*{p>IhHLthR`){Sx9Z<FIpk0^yuF$lOqmiwA$+jHy
z6ST{~kZSer7}}v)A#XUk&&ds1U;CqHLc46GCoSr^Kw27axP0*BJI-6~g+<%?9)94x
zZMUty&Pxkzk0&egF5I)xP3xb2$BT+t8HebRPX?mqZ(-ZTWoKTaj<fIDWtwNgsQpyh
zYM*ZFj$h*?jZyWbovr;j^1|XD2i$n?7abP+rHZch8DZrX3;$u`O4jN}%`c<23+t~^
zKNtf+pU;G?7p!-wl{N!wj~%(gkU5LJs2ErOu<<`@^-bqwLGSB$%-ebwHcn)veyKcP
zbTv-ZetVBRf6&M&E4`o?HU1>sPJRB9m9olzT4mRKq%|(-R~QJxfa(0k_5G;ltZBE?
zo%4j@Ug`=1Sz{n;`<Af!oIJYAn&C56_+c??^N_dwl&Wvk^E0(w*5*a@5eC9AkhSyh
zsQNs)_{h_@n7Gu>3##!bujQR9tXtQ)$6P-yhOt?xn|}tP`gz*zf}TTXy<S{@EtdAK
zJ*nqIedbSZ_pKKeUF(&W{IvZyX#aU#k4<{JK9|zjRk6ei(>QeFFV*XL$;U;{Z0?`J
zo|8%aXy=P2-ol=9=Dq&h`<6~Qt68OjU^pJqTIaf-eZ^ZFZkg>BJEp~N+<xC1TXE=Z
zA5|$NjdArZmG*-!xN~Kz9u*2f(H1YJd|G$kR{yFT_4!WRvMN7ld5hMU&wBErm3~;P
z+VQmQ4j8-47tA*g*h$Ssoi=ZNa~D5v_{f&Mo^5r{94{~Ue>Y$DZoL*Wyu9O_o2QgD
zXgSd<Rm5)ue`cV6r<PY-vh`<vsiAKE*R)<ybC*lL^0SHuyg%nZOLzX#FICjdo2%<V
zp9f^M-u5qV^q%T7-p?!U^Tbz`rl0kyr$bls?n>@<UvK=l<F+fjpy*1EuySpGHz=&U
zu=-}T{)TDqZro?oC%!9Oc*7BU?bBj}7Z<b2U*7e4VDU%Or}kgz7aQL8;*MW+E3EVd
zf94_Yc4|xiyydAU9&+f2_dom6H=!+lY|F{(bG2f5zHW<u+jbFDURHi59##)qzqby`
zlhhLigaKhd7!U@80bxKG5C((+VW3zTxOwoSV=t;xxgaP!h#lm|52`%>TICB`Z1h}X
zvBWtUhcJ)?16euWPGYRmM{owRVi%>RhwgU+NnH<aA)Bt5x96(1_1!cp`Mv7;?=ahz
zU#jnqo5r1%1**S8vaLth=dn^h+~4Jn%9rc!b-0!v_C3(7)wk_ur7q7DU$b(*)aOaY
zFAN9+!hkR!3<v|lfG{8o2m``^Fdz&F1HynXAPnS*0o~t2%ldWoxsR=~t+_LOZ#?Y%
zq@K@9KfnIFEw+BwnzkI%cNc8gy~z1B14(^%_t(8jNf`Jw1C_Sgr`x*Y*LYF!*V3dU
z3<v|lfG{8olqLf=oxS3~W+Q*_Ojw%i$f$(@VL%uV2801&Ko}4PgaKhd7!U?vAnShz
zJ*a-Jo)^NCpfFI33}o&1d|~ym_4~-Ma~t*dqDi$28aG|n3(Kq2FKq^*{=T%dS(DKV
z1HynXAPfit`C=gK-)&@cruRsT?LOmHvFAg5=1*_;t(O;^*Qn#{yLOr8g++DTQWgeW
zGLZIr@@oHKf2Se!Z@l-54vT%^p*VQ3^}DK4lM^>aR_%D&b_a}I=7&W!e(B14*X>Zf
z=fvec*mQE!zqcB;*e^AVJ5R3Lulrk#8XKl>@NA<=?_$g28P7pip1kbOjpzA!!9IJe
zc;?Dx(|uQHvV)+1uYISFo3E~YT!lhVH1(U*{rjt&*5uQTWBk10vFpaX_RahmUa{b#
zx3(M8ztsvaZ`gQb^RxbZ>RQhgKC)%6XItGf$4eSN`KHJBcb_rY4~w?_cb_M|sx<wq
zSG}yD?eXd4&D>%3_q%tjP!J5qgN{$z*NbJJoA<wi@!ahfj;MLWLO-wAcFFVeww*f9
z&nqrCcl_6#H=p9?9mBo{m6iIg-fs#TFPFaf@!;JyF7ngHp!(|S^{czpm}dPJ_=3su
z(DUNe<Gn_{pI5Z?K4efiZ-4suMQ08k>k0nM!|S`>`=@HJ&i3<)uHOd@D_<8MU0<Jl
zMZ;&`9J0vI3U1N*@>x$_w9*fXS<9otzUPs(`eRO9cJ|KQKlk&Fj}AQUvDRNr_w$Bv
z^Lf|tB5(4qx?siU6<)7YNDEDTEqvs{WgX7_&KJxE4?))*wVr97quyWK;FEU;H2A>F
zIz}B|TJ2g68}Qi<A6F{4!nDRE{kml!>2nIVqp48gtUiNIfAF0Jo^5ok_rZVLwc^M}
zzVVVq-MGfxQSt8^{xZ|gJ7(>Df@hY@Y4Y$HvwcB3JS4pz()o*AkH<h#=N@@~&&BrN
zgU^dRoplBM`|MHArK(Kbd)A+aPw{PGUijAOS>^fHDqqlIqnB6A%JWv89y&X>b^YZR
zU#t33#TC9_HhHi$&mu<{5C((+VL%uV2801&zzqXe|ND|FhhH_$%NvIMjzd!Q-OQ3m
z6$XR>VL%uV2801&Ko}4PgaKhd7!U@80bxKG5C((+VL%uV2801&Ko}4PgaMljg#CU%
z()DeQRb&YR!hkR!3<v|lfG|)B4BYkOwgdNR_Pv)?to!bcTV2<8iB~MRWYKxgHaume
z7Z;QI-KFh*Z0`?>%}3MzuCOaUb-zorb=}o>P+iIGe)bh_ZMbE&mv;>MJKn6-OX|I7
zSN$YqH?HCVa@*a0K+irSm-u<bq`y1bVfQm;-PmltZyV!&?@!l$`<FL*PxTq^2gT7H
z4{dn)=5ze8h&^Qm7uP*v%h$L4%8!bDCZGG+`z;svY2l2?|EO0se2$-2?BA*76_;%N
znV(m@r^~1P&aVHSpEvw)i|gARdB+4VE@s7#RXuF~-QTsxj@)6$oW)*`^*^52t=q!L
zIOpamWer+R^x~pAUQ_vvZJv5yw|+DHsCe&{w>2O2>?}X8xcZ3~C!DZigr9fJ`}gAV
zx=u|z`0^K9w_fCjMU!8~y&pG@WUalceOOX*i{<a=X6+onq>uM-wc$f{{?5xQCUs7b
z*M2r!K6vsS=PmcLj&b9;seRP%>}9Q8(DQQ>A4%W0bq+3a=KN>r&R_arc`&U5Vegyl
zNB8&H^Kv~bzfwP%U1o(hlRca6pFFzDn&C56_=3Vi96NYm@ki6A_Fw6j8tVLO&;@s{
zY}KPeA#149H}3VIe<v|(_40ncmUsQ<zJKf~t9Kpe+d|v#F^Rn4?+skCzz@oUtM_t5
zZu{ofA27ie9)yP^>>zGF5Vife;|FZp?uE&oZH%f9?TY1`UmXX}xwDROpTno!e${ga
z^=Ywgu9sI#o4>gI+kPHMT*xY>?a#Wu$Kv|B?RkzXFKheVxca>P>EjojIe4t6L)-ok
z^tn;i>)HOjPG02gTw&e1&OPS(D%!RaktYlY1HynXAPfit!hkR!3={(c+gx^ci%%Z;
z!gGzqz*RAgA*;WSv(0h!-S>ES!wX;gdB-0XU-Q`gr+)3{9kce{a9Dk8{ax|2<eXgf
zPwNls`;?zoJn@i2N4)>pm%iZd^AI+U_Pl)3rK20I@$!nh4m;7mW=Z8jSPaWk)cRr1
z#iRB^yLa2(e0`m(-}m!^d+)XM<Vz0u$j>Y8^Tbz`rl0kymsgDY97@+7dyI-rCpZ0j
zt6_`%sHp3=`~LHv?r`EuzHQX?C)e}EK-m3JUVgu0Th9istr=2r`{#UDxLMoRpL=B9
zn_k-Z*fqybKID&YczMTHuWC@?^=BvfS;xHQ0Xn_!uX0+GPdAS7U7;FX|J?N1SK17y
zQm~C`KS`Gt*QwX&;5!z3dBY>NJf`#T+6%nAW7_Y1MeYBrK7&qw@SO!-SWMc!r9WXH
z4F<CQew$9epuZckRnOM%veTGj=~EaG28xS;-XqT+G;+#H&lT$4!;X7B?0foY&6@4s
zL#8zjTkDp{iT6F1wB}#>6$XR>VL%wjHUpP#yzkPpF8{_4i>^kM)$hM-{oZjEe%N`g
zpI6Mvch6nfjV-x({rfO!>1~U5lbp{7e7*h5^XB+@#YYDo_gL$%ru%uvq~f#59$eqI
z{r5Lq%R6A=!xy%BWxnSMUGpvSg#lr}Jp*rz-tMg{2Q2jSieZ00k(K&4xBT+pQ`^q+
zlS22FBt85d1J5j()8yeZX8TFw@98A>gn?3HV5bvro4D;IQ~jv8ZO4fZTu^<n9~Rp#
zd4AruQ|I}vvDD^M#w!d61IaV+RoP)j?mBOsCm4>0<Yy|IJyz{_+I9zwUFHjx77y99
zMt54(eRs#LuIszR%R9c%f7f%mO&RZH9bNJ4&O$`8O$M_5yF!zGVc$#5di}89ok+WW
zUf(ZGn>Snihy8n(S*vgB?+nI`iK{-SxZ8IXw=LuqpWL=X-HQ*XU&t%I`oZ=q>U8)+
z!8OLUlVWdwa{Cu5-o19D2WD-&qnFjLcl|jZ`-0(k$l5%LKEi-7P`nIO>oDT6HC4a&
zqGItf^Xri<ICuQloj0H2Cyl?>B6_LGJ?Hrw+m7_p#^@uEmcoEAAPkfi18aACcBe6y
zFY|-qT^paTck!BKzMvZ(O3V6Wv@j6#JDhp_{(D~aw*B4Jq~$%caA?a0pDpt8hI#!S
zx2kuaJytw(WwYsiUU6}QPu?BS-~%sjIQxo*&%QZik)KyozgNsUVGHy6`}??hhvhG+
z-|3qA39GlQ`WvRbyK$dUpZIBG(DA&!`@Mgv_Udfk76$G2?Cl15I}iMiZwrI=8?|1K
z^*^52t=mE`C`Rok>+M!OcTk@e>*ji4(bl+h-yP5U^%Kihb!c+l0xxZ}HBYwXeDY0?
z@9#civY%H>`<y>+|JA%FYV>H**K>tY?d86IKDz(F?bmy0VN|{I)^5(tQ_32&oan_x
z-j4~2qdOkj@bb;)_=2U$!(InmwOPBreB%`x{&lO4FJJfBcb=dd9&E?gmv&ou$~Aj`
z>*o!FKBtOWZ|4f@)^+YN*SCeHeZA}ZamQghKgEvYeuq4%_PW218236n<PEd-y8-lx
zJ0H6Cd7B^fjqCSe+eiH_NZfXJoP5$rH`ac_PYPZAJwxVAFq;g7^@pt0xBd4pg38P0
zY!;tZOJDqW@NOFy`FX(;4>@$i`=5R3hsCVVT;Be^Wm5Xunzt{<wjR~^naYK{;OEs|
z`1i_2RSS8=eRsI}!ky2oQHY8UF50j2BjeV1w(yMSzdm~D8|%Eh;<#D+|L|6g3WdDl
zmG@nI(?OqBE94bNv|IV{Noy+=T%oNvSFCbef43jrvKAlK{jElg4bwMxdByh+JnOA#
zW#4&uLsx#TlZ*X@#U~dZdHNO;m-@CbtR7jbZ|l1tapgqi<)*V&9N28+4_>k0q>%#_
z9(l$xFYma^<%9MtJi6YCi&5<&Xgl4xwCTD|{*1~0s8=?8jvp1T8SvV~_Fv8SU7_yp
zg`2KNecqQ=yRhHE&RTs}-wQ;p&VJ>(Fdz)rXW*za4{vnchDlytanhFk%HFthofj1C
zGq`Dqw%?yfTHZ$=UA=d&hlhGe<EGhkcXdtc6*YIc<SXA6+SW9wzdLHuJMQy(+wD#I
zhQ0see#ceq+4vbRZ@8e=iPw#&`--1eeE6|r&R%`MK+iRXwTEpVxNF1AGvD^}g0`Ns
zq%Eh;uKjv$@$5V=?0iM7ANPDH?e@0M4U?7^b>8yYu2TQEzc_EgG%s)XR9TZD&o7wc
z<rO!q+yAgtizj-vFs}WD{oSdqeq24I{jV9w+WUK=kKNy82%m2GY&yB=-&+k^>}L({
zuX0+GPdAS7^N!nXQGe{Aw@mkv!sF%+Yuj+rYkpkpb7ZF_hn_y#%PUq{G{5%~ZRUA-
z#iV{W<*J{!zxz+Sec1DjUuPv!ips#k8;;m(pB5uL+Za_J+9mz`nSNd2on@O>e5~U}
zPw@A7*#Cwe?caLhNiQ#$mwY*@T|fF<)x(Bfwa9agI{iBCve)veLza53QP;m*FD(X=
z{(F5<zcb<beepN8dFp}P`pxk3hTFfq(R-@Tc;6P<vi_jI3k_Q@srAS7o?+JSXZ81=
zw(QF!$JO`Tl9J1FW_6r5{{8%V)sFqbPa9ReT$i)%ZzlbcUT3AhIsaL@^Ox4X!yEK_
zzOMbi*53s)>3PnN*Y;d;>1r=;c+~|fKCke4r9$4Z!r;$44H;9TkT{wQBjo+C`JJhL
z_p`5fYr`$GJrzT$6}h_lcOCLp&g=6E^d9x@%@6nK|CN_l-22!L54Je(GcT`r*Thd>
z{G!DiFD;DhxEXEu(%ZW<?)}eUzH2lMKk2_mWYQ<A_k-ruN%JZFRjoT{*X2GuJh1qq
z=~Mf!^aX{7VzvWaoE&t)ohw`Qs89%sLC;kOywrT(rd4M9aq-<nV{V(#ca^8&;RoK^
zcH8RfJlCjuf7W!p(pLL)TX*~#&lc+DFRAOMbE~-bUFS6-KUn@4^!&<Jy`=0{*H7H|
zRIKgcEA8JYm3sfhl9%Rf9<hH{7&MQ{TD`pNYjwXz%$pq+o4)M3iv@LdmY3^!pNGd=
z4}Wjqk_EnN47v`u)*~tTI{OXEL)dz8|E?bG!}e>dzUg^nUhZGGXQP|eKmCrESIql9
z-`2i5Z8<^n#H`hG^*c13+^F$~zhfo%eV642_4bnHK{uaauWz_~@Z>wrTkd5At33Z&
z<qKMD^pe7;_HSyJcAlimpIl!z@shRsw(|^I<BR(|BklIK&!yot@B78P>JR^}(6+rK
zEiY^5q*3)rYTeTH<I28Fa&`L1b+_kRX|Gdp`%gN~*zV7@eVhC)Z}+3}uZ}b9{$}m5
zBX<}wXR#L)!@l>U{!YPE*1fOgJzQ=0ke$Et!eZF{bJX>Xc45!yqV5yo#uwVF_G2pZ
zzM5{H@^ZcR$nyt{oU+ohjVAuo`=)b3wZ83gkM%#E*sa?_&lSeiH|lwB+<t?$H^sBC
z_k;E;^@IUoKo}4Pgn>LU5OtrUw#(C)MOR^<*co6ynKg{Me+b(yEAK&t)!Ss3S)0$W
z`sDSw7<#MgUf#-K?J=wMgTDV_@+Z^%u;-OouOF1RtkmOu+q`1f^XIJAk80PPPYZ7P
z_vNm5F~v2xFAN9+rN=;8zw>kbKB~w+^7JhxF7;jWkTs79`+h~#`mV>l_g+g+zT}XP
z{JfwmKEl3FQtI_x%|qVgX63xkmL6F>cQNS|_1wU<y&Zntp$kqtrD7qk=!&1Lo;L*5
zEA08RZ63qEPo&njrB7HMlCJOi{bSeiljh5If3Eq_$+u<SB1aey2FwiT{4nlybI;kV
zvForC{cDy~F60&88(VScZ68%B1V!B}M16h}x1Fv1P^rmL^F+0*n`gN$3~Y)4-d_>?
z^>_Jg^RX%JgItUHyR)#r{~O)s<OZ#;{m~QLG!H@M>ZdwE*JEA1p#4a_;%30r?=zEP
zE$Hu*gVqb0Z@L;^P(9PG=lbtaxRPHi>rUL?DZ8@YpmMYF9k`(SN<Fs>L_KeIt5tOA
zQQxZ!+oS1w0a?HA`hB6K<VSt(thNg~|9Pn&G_T5AJ-+iIxNcqN9&>$HJlOi3lWl)-
zHD23#iacT9e;J6HZ%Mlw@BN~~Vm~Mk|I2xp{|A+?)(g{z)D;GVfu9*r{Yc8;;#}VN
zGW}iKyxtdo76ycY^cm>+P@nnJ+kNZh4d*rLIQy<$rg^S0{Ta%p2CmK%Omc0#hvJ%5
z-TMr#uB+q8`gyuZzr5vXs=l`D)Rr98KGbrtJa3LV4wa9vWn1UIIyqtWu~pyI_qA=w
zjr+Z$Qhnb)tbLT)d*w3T_RX(9V1i%jJcRWVTlJq=GN;MIXUz6(qxu|Q%8y@k=HRhj
zUU{%(Zz4w+5C((+VZaRoarY^u(%udG6)LFbGp@?6^e<MqarfU@X&*O#EY|szaR~#$
zfG{8o2m``^Fdz&F1HynXAPf{I1FrVtiz}Sb`-|^q`nE7C`=<BPlDZ#Nf3;oQ_ls0{
zQWge;0bxKG5C((+VL%utZU&Cp;r(kz99q2)6vMI+^!sXI<!+ev?#6vaec}rW4=(J$
z)%yju<+|Fhi`?Pw4P3IoH}PP5K1)Sj*m$4S`nuoy$o1y?Hmx$-7apSVkkmP3+WlnJ
zuSKuoV8GVz6ywS%4mM>J!hkR!3<v|lK(-jzw&TPHE~viP7fgqTrjwigz16VAzF=wa
zVDkP4{WcXh+iI$pS8VXwnjsapf6mvTY5ZBazx~S_y{Gz&_w$NP?m5rj*mk7v3bQg#
zqK8`s8a-XB)jM53_40~m^xXcL7hWCZ3knZn2kHAk*!K>u{`VzU4!>%gS1h==!6)wy
zXz+nota$GD0o%5FVX_w$-L7ULJueJ&-u1%~_wPO1OA7PCi|A>Wfg@k4xcHeH=J|r@
z@bHfpAL{VLwM%_L;US70nBFs!`#-;rXDe#j`YvkT{*GkY`o{geVyU*bW%nXS7!U@s
z&p_{y=MNe=Wu@mDRVI@DPNmwvu53H*NBw)<ruXx8`pb1;AV~&n<+pL=Bspg3OBfIa
zexHFy|9aF_x2zoF<qdy-RN1}zz~YalPwl_b7fhCi>{>)=Xl6TK7kO3d4%&6OUury*
zhPTKlg#lqe7!U@80k;f1u>UXe`*rWvvFmPSJC^NKR`J$;ZTgi}`sv@Ub-HfdwO(0;
zTl#ga|5JI7vWjIDkN=@dpMG5*y}i|BuX2TVPWz_gRmZIJS5$Aj#U*F2_51x{TAg-}
zuPVp0&VQ`aS$|*cfAmrPw~l}Pd;h|z|9SiW)~@iIH@Ro4L%;gYf9#IUy6#nHMLBZ+
zq%~L6zHO)V{>H1@c0S_cmF4aBdvE+-du%h>yMF%NyL9>2Fki))2kmp~;}vK5wO`n2
zzxF$Q>hIn4qF%fHb*SIsk1bn_{dBO`eadV54xKpL-|vkR_nv#w0{{Nyiw;<|>uA4y
zm3jYu=<N@@i|Z8L-Q&Qap>el?-*5i#TXQ@e&b{`Ynjasq&>MZo1$#EW=sWLeZ^)-F
ze({C3&0{<5_E@Efo_|-zE*IW5tX#kIhfbb5;o1@Yp)>cn`rT(Hdz~8;#?EOo&Fit_
zcCSpY^|jagPj6MKeb6%R?mPGCy<fYP-l0`a*?8!y-+S*hd-=p3JFNF^*rCcEr!QRN
zb$@2cr#D~oV>y1>{fe#H-dCZp^s>=|9vZdLJ9nE6Z=QQzrNTp_59)k+&x!?a-GE*1
z>a@WtyYRzn&mFhcd-t{>yG<NZp>WTQR}Jo6^+)gX>t6rkxa+2R*H?Jp+I#jH@0~Dx
zvvb})cfR-d>5Z#xeEA#i{tXv(x?{nt^6_2x>^G-Z8@SX1H-G!FGndX>>@^)Q`-biM
zFE7XQj^AnTEuUZJUHtt)T`Rmi*=HWXKR51q%bWh3ay!9pm@o8!AN1gU`TYJOpZlu*
z)%%Rx?9_`!l#hpbLqFR6`^IW*r`7({qyD!Ke%*N6PnP&9Vqf52UfZMI%$gH@^bLy3
zz8ThW(BQAi?SOdz*S+)M72oYP!(*JtZSliPFV?y3TkowyuY3L3d*+tUSGgXFA0dbK
zApKB&ZG7M7M{PCaH~u&Il$VBoJL3x<JAd&1%HC>p`ZS;UKI*~APwnu|o#lGK|C{^%
za7X=*KPdM@#sOj%Dxw$2I8|i+j=kdgQ@7b=xrZHs&$O@C^V)mYd%OOz!U^|H`LF%s
z2lx?wbK0~O6AmhT@4fQKq?V^{R<XeSQEU6o7&C73!Xaat-*w+b)e2AT`P^5%`&TU-
zeMpCSU8Ytl{PWEl|NcyqnuWS|?t1X0qbnESiMTPV=1IGCAM!(a9AQ4t55Hy}ut(Zo
zx8&h_&t6us!1!qo&(tSwa38;CUXjOq-PGdp@w+{-)Q5lOwffcG)RF)C%73Ew33v4z
zzRvH`_0Gv(cA4+DUwuyRs_VY?X-8bgk3n@Ec>mNXU#$GoqVn|vd!k?Lfpr0W(TDYb
zdiW=C0bYo|#0SO;Pb$I#e1Z5iNIyEHo+=;v#2!KT0KdAV_KX9s7~`vmJ?fBg>kxa0
zihtbe^~SB5tnzdS-#Ww(bQm-ZV*ZIY%+n56)gSpz$Ir{-5cWZQgm2oh?w}8L#5_@t
z^`U${_{ATHN6fcc<~sV(KX!wC;IGt2U*Zz&UmxR*Z?j>whkhV(wtef)&o6v?od*&p
zK>Qy$Iy+$=RKH!m^{j<Ezqri*;^0<wcm30Hf7txzuWRteH|6U!aiZb_Be&Y>)~Wt8
zx3Bo}UyEn?>#NnfaY@UCzM7Y@z7t38eQLm-Ki2u$CvM>H^rs?zts?Wt{F8UVKl8fH
zdx!sf$IVxG%rE&Dc_i1#pT0h>?t<1kRw#73=ZWKMy|Y>2vkQi|c%o3Fa9M?q2hVO&
zsX$!se0b+JXP>#&yL0jdJ*U4~vGDZ=uV20D@+t+^5fxcCK<p7@-O(ZARgrv8#aHid
zc+QYlYZO520c1Qs%vt)wCHwv0UH;0w&om#i(nHR_I*mPL@7dpZAmdh%^^b84d9nIQ
zQwkN!*F)kxSbK4=;XAZk=7abrNInBD?ofODb?ra*K>P&6kKpl}o;|NScAM26*O}i*
zway#5>Z2vz+9iX(Xn5m_a{s#Ju2!{Qxq4oC9!@{R8T24dQ)V3*u-%6@jqf$j$37Vc
zc3pm7lMmAm{=t3*zEs3-i6bC!iTI_)ANXPX_zn37?eP!fQXhXhqxF8n?*Dvt`TD>*
z1t0LvK7sqRM-TKQE|Omn=kQ~Y{Z7Lk-|f2e_R0knvFp7%jjVIg36%@1f7qjn@W}Yk
zn|A1XS<lN3y#2<_3dCWqGmep!%Z6RMOV#qc@y@n&wzy^U%?dS+?RfPAtv7nyM}PWb
z{KRK?pnvYqxoYVRn;%xUK)wi)=cxY3dP0Bb!91|8a~)*e!2fAae}^17b?1Xd*D8;1
z@D0LGqi%;Bd+0&63o0^?AbBvze1Q132^p`7$iXi0Py86aBM!0d=*~U3kDua~<SF<q
z>mzv{`Lep7q<`?(X8)@B`t7yK_Z!o{TzGen=c^PNefN*oADCUOeBXh6skruEpWZ%d
zVeN9egAexeI^J3Ln19wk>;iuxzJrXPILEk&f5b`bjrA73Xn)KZEpJ~t`Y!*b!&hy5
zv(p%le1h{K=8JahBk&LEasC2gH`HU@WS@aufY?3jIey4G4-zLp>|aIlW5$I(Dl#u1
z?R1D;s|ZgfWL!Z}wIB2(UV+3z%ETMu6zzz2*a`Pphgg@GC+=e}v{UOdkMN8A(=Txs
z{g^+>@P<E<x026d&(xzG_d)ys{nhoKe1!YRVZQMv%GfRO1^vqP_~qP_dho^i4^NDj
z{*lkR5AT%GgZM?ij1PYV(T8~;-qH{988>{g{~!(%C*dt9qVM<*D>og}cW(JU<b~x|
z&pT<aRptAcKOME(#Led}_M1!?->F&C8RhlRhxkGqVq740$hkIU_DRHj_(C7z5cY^4
zsOvZJmwbqImHi&b`7rq|>-9D_UVOHHdF=x6i1iygWBzqW{-s0We^?~mw7O_o_4(g_
zU4FideX5AQ?CY=_&e5@N{2d<QwS3+A<=m9Km3G7fcp~qG7wndG>z*@D+PTUPGt2i0
zoZFKRGOzd__D`99kjr&=R>#lzCV4dS*{?8e`bTf}ap()*#Anud_Cv@o@Bf!`WXAo6
zy~n&d`uGJN$bOLZlJ$Xo)68wESNzlJ6(0L;<^%gjE_^c{;um_u59j5~7yZB^`xMp%
zc)?!qTjZCI^Oxry*co;~zvM;KM=tSAjeE>DdJ>NqH}{c4JY*ciAL6Up4|2GUeEc6C
zkc-}w>7TenzuX7WAB0Ej3xrSX61^B7^^mI~{)2qvfXGpi`2ykXg>4_Z@RJtf%bx=x
zmw3)|IId$y=tX=WAB8{S4|yEpWd4zh9_qOm^|3?hGk+?gAM=Uau!tQo@01xAi2pKQ
z%q#YRy`Wc8NPdf7lE0uIafS7kbr_z|oB3tFbjUt}=N{xG?Dxp;m`Cz4@)q=<f7*fQ
zO<rk2)-x5cC-|X19b!kIS`Qv|NInRv^;G{PPJ!qP5~o3U1<@14PC@1qWSwWf%DNL2
z(E}v!<9Qc)k;majAbBM7kKMsDb~N|#SMNQh)qmfc;GBW{n)3x+nfmC3z99V)Pv9RO
zu|wXYAg*EW@JT*FUO-&H4~bvo<?6VJH>~5xK_B7@egxvb=*zh9uM;cnz2LxZ-}&SX
z=!;#U2YCwml8+ESxKAF5-y`Swv2D+u*l1DtdjseX542;QfJgMf?#Qpmd&do${&efx
zS9|1jtlQ|tIXusCf+G0}NPY*>KlvZ)EAfbYj(K4{cyYnVnvZ|E!JoA6-w(We!hhcf
z!k&p=*e!@%;1|pz^UV7J_y_fgr`RWU#`{0akIFap!uW_2*fDVn9@r0&cXCdFU9iqG
z@9e+ehr9*3r#9&E{r{a%tMF?5Yd_g=X4OL7Y9EcQzhm_R`6K6gGuqwNsNrTc3dg)&
zrAM7+)e27!IqAwhmenYKub}<7mg{c&_D2tWi2ILx_xO+PE?-mrybt}j&Un=K8Q5=N
zXRLeZjlE(ACM2#ff5bohmbgj2MLp&f`v8eY_&10j(;x8{gdgUc=Vi<{*YPWm`Nob|
zw}?yl5Bo97*c1G6A70^&eu#gxQ`bx6QXe}bpJhEGeo>F><mEhHq#w>bv19z2`6e#$
zypD5s_+md#9*mywgq^V;!#;^?jDz?B61V#d?Rx*dH?Qz{{)@d3Cn)3Ztb^OPS>E!n
z(^i(hk3yOFgWah0d5#UT&aqD*E@7Y8Gj@ku?zir9z#dKS{I>jk7~&Ssy=l*LZPsbV
zg<Zn~{jgsE>5p?N_(EU&fH+A$PP}AaLw)o_Kl&y9QV+jC9@wJc<Q8|{zrOtaAI3pG
zL_O?}IIkl4s%^x+i9`4!>jL8=KT`Fg46lq2-c%%xsz|^1J8=cSW*wnSJ^YaN#23}S
z@P810#Q(8Z&JR^&o`{pIXRHInGtSM?(}d`W-YOEuSbvEp#4-BE{-{s<p^RL3Q}yD#
zS<Xk{3FJ8%{g@D5O$e_b;|q$|iHeLjD3Vu#<hAfl9?bf%U*9HAbvbHX`TNG?i#)Gx
zH)wpvyH;;j*ldS&vma_#y+EErUUPky!Y<X0-mE~L%KId=JFn(ZvxaU{t8n>cZ=N}M
zzdD5(pYQj|2WM_x{@x_#)jVJ4c_#D3`!~FY$vVjTkDa0qda^E3kNIak#9p6nw5s}N
zM{Q9cKjV2T`$+tU>+plVoU^WZxlgC=2URIhpFEKKOGWa05Wk~6@nBkqzpVePU4=r`
z$4)t6#fHk|^6?w&ow&mIup^#hGC%Omdd)mhCZ3Q_p*MQ)J}vTyBk03*_*Z$sFOd&2
zkMO~K!UOlQC-mn15Bg(%IF}}$W*p=<cOP@w{uAEbqWrz6QP+K2wrkHC1>`e-_`$#q
z-d~q(zGb1-A*Z*x^}ekOy|zC1?&gneSva9>%?qC#U9UWU$3Ln)II#O+&(|)XKl0cY
zV5ii>Kj`Q5(f>E<nr2%S;P2+!9=~bjjawG5JDv-HJnup7`7c&}e$DDF3-}GZ!y|gZ
z2k#B=o(cT$-ZT3Yc>VO-KRmf^;eX$AYudh6^S3tCDu9byol$$>nCj(m13i%kGCq*&
zyjKk`@bTu$H@w?-S>3`_^{?8qLA`o~-L_d<|E3+cF8`eae#AI;yJx^NO}cMeAg(aK
z*fDm)eiuInsgK{F5AVScr&>=wZ}R$!wkm&5kbMGqHt`1gEZ66j#eZvmQJn(*cm15X
zuRi~1&BB^}_FTQgebo#9^k%H7G`?zKpP8MXsQXm4@_i%k)%<N$y)pYvty%bdUdNA)
z>A!j5lX`DF{#pOJ<-D{!@xFD7wy9jcAEh1hjooawN0Ys)p1O4bd!*iJ5AAr{v>|oM
z<2Cb&pJNBa_1<URessIlTa@o>iI4p5M17A0J7hl$&%Cdw+70hl5^u0~e$T?Z(=YiA
z=QZSmJU1n8WPN0v;Q18mKKV3pmHiKSiJFh_ybV2=U)~=f9ucqcPwa&8VMoM6uH%RB
zM|=E-b(ncl<2QB>599&p#X7*g5WB^1Sce%Gah&(-h*RiId*m|?cqjkE51AM0F)#3g
zUaarv%X-K7;R!v75A0KsM}5|P@+antbqReK2kS5M%(-FplmBq?k&mt_|9uzhGxpEC
zVXySVeGvJ~FMb5y_zUyLIPh=Q9r9#&f;Z*``(Qq>8~UL>=S|E9>nl96p3o1-Jir_M
zFdq12Jd}yAtPji&^NC*Qfn8wN%s2X~?XYL$l7Ar=eUS^YkLLV=a|h0~xDRp;0J6Uf
zitJn2U$B2yk$e{383*=5JY?SCh52B=!npBs_RpLbfb3U^tN1T@A^n4#Z?MlMzT*Gv
zbEwb08{UYwAo{XTL|^vv=)pNuP(&Z(qbFtL&>!)ixCPP<{qRHPgZ(GtK@Rn?H}ruY
z^dKH_pX<nBzCpDf=NGg?Z;*3Ll}F}>b8z~@ZfK8R;MdHjikt_bAJ2{8nfS^672df{
z89j;btk*m*Vw{|xb6$kKfXIO#>>GXXL-;`s^GQD@L~qt(<`cVt5By(6^*uTK5`D_`
z{N;BCoGXI(73W5jIhR-C6Y-3A!1&;ad50Hx1Mz$8h5Fbh`ZEt8*BL+Z)p0W4_!;`a
zBk>A;(Hs4c&pgu)*FmnsBYbh6{<scbltJd3I1Vr9LmWU~t}{O7gZ{`%(GSE<_}xDJ
z;fKsS{HQ!)2k6Uu@qCkc#II<NJp76Aai4KfkMU6t{oxt^!XDt2IEftEv##M^TqkZ2
zPsz8C!}>^j@)X8_UhFH#_le{99rBSwKEwJ<Kjcm1IUwg7Dw1D=j7vr4#f0dG9icz=
zhaZtou@57Eg*TA;@Xm8{<iVqDgl`kFewk3^4Zl+nduAOYZt+|TB>w3Td&V!ZbL<WK
z#{X2L-f!@i_)ol4k$#dw;txn1(IIxHB7RKXu0!lwhvY3fWPWvM;_Wx{p?sG_^;@;f
ze#V44dnPWER}n|a*U3Xz_uvVB*(c%8+*kd8e2_RuKiE5Y2JZ*(oD{o<N8*rb@2n53
z8?3iFq<vWAJc{Qh%scN75a-l+f=`}<@_dx%+4wcjgRy_!XJP#KCp@Tr!FA?=`6C{q
z2X@E25^p&#V}7wK=9hC6=8^L-^y0k^)(PSrdf`Xx`#A@qo{HEB^Qt2K>kzwA<F#65
z-jL6^4f9Bz!8oxi)-&vwddM+hSo>qWL=W{`f$wFozXI_i?27jW(3kfY-s<?|p?myu
zdHL_Jk<af}IM?7hW%MDRL>}_V6V)=;d7pv&S4I4T_V@|oQvHbWVTZieL4Vi<`ZJIC
zGxNeaN|`vr`xV$L@ss$4f6+hZ8~8ozKK6}&5Jy0+<8RDAeoQ{Y`-a4G>=PvJ@O&4<
z|9P*6eu)RfamJ57pg-ef93bc7>^o@Bee`7>nP1*hVIIf})qZ(z2zzH9u|MpSdB$$A
zEBpYvfhWd8nYhb5GXLBsju3ws7xj@(98lK{c!C$&F^?epQ6Ia2FZxAa)=%_AKjbri
z*d_JwCy?>7AI9$KAHDDs<_%t$f94V7Je>0w=9hJc`}h@hj(qfiH^#yIu?~UMXJ1U2
z_V7X(gm2~#zkvtZ5nqUdtOxMNevonC$HX1@f=Bf_^NilyhcA$MAufUB0q~E0%rkMH
zxQ+erJH}g|e(Ck=_y3{%eP6x@h+lDjfIspcu-YCz_ZZe`Sm!I&_$tC5i2SsX{Tud9
z{z6<Pui-fY^MSo$Pn59-=8N|PS)a)JIG@DMh?BhM2+!CbevIAW56mw}d}ln!2Z<-d
zAMBF(V?80B5qGdF&W%{_m_NqH`<BEZz8}eWcrO!PK<t_K&f()PC-mypWqHLBFTSvH
zbi17@j2M1j?TNeGUws6Ag8sCJC&thI5qYfN_#5-5`YCdegFUcr!Y5_)p`MDw8R7)@
zS?|%4{T$E3cwVFC`8-eHxuRO;{2k=^9`(7-xfK0!A7p<3GXLyD(2H{|71?)y>V370
z9-z9vM_-V7*gc59>JYie;dv&{KY9O(b3UFI5^vd8FhA&_BJl*i=!Y`>gS2Pe1hH4*
zC;1TmhX3P-<X`v$`3mh>=g6aIk6iqP_(R+w4$ux{ohH5!H|P(&Xb+OVARj$h&+sdd
z`~kbG{>yWB&LNmTo?Bzb#Alwn!#{q(^Kj+|`OF`PU*bQ!|A2jPUXFd>m#j0y4dN^7
z1o1)5W8g#OlX+!a#6OVxAaR<0K+YMc$Gk9a#7C}E#xAI*+9&eRo4f-%Ay4N%<D?z@
z&@bx;c0_;VIVy5a&N&wE)v)gJz6a|$eg`uDDsrwtJ>C~&enIxV{C<!9I_oC;SjL0D
zF<$s#TwF&kaTq!93opz!`zidJJdkymyzcP(D&5nu!Wy6Z%oFyCe&~x{=t(|;J)kf9
zGRo+Of3qHvmtY6PC)R81j=Z1#$YZd3;t%UG{zZG%b?#$ttpBXb*ahn{ai06+`S_6z
z;S0X$4`e)yAD&rvh>z4`K8U;QAK;yO<Z<l(Sr6Ffkay_xHeJWg86We7-4GwJ5As3m
zmvsg^<s1t8V_hR3X5UYlc);&#Y0rKIIn=}duy^ibpX8zHbyW}i7rn7p#)rK!-^f#W
zKo0(kJoG^xey$?(&-n<5zW4|03I4}?;V<|x`obgYF@A$y%p3Mg{6=5;0cnrj;{V8p
zFX91y$$HQFh96U(_Q=KWv1{TANFD_*@Q+_02mhhWyul;mg+EZO&$u~<!|zysm~Z5<
z&p;0A%QXkwdiKBG+vu~8V!micyg(lMa=yZTp6l=ca-I3bo|zBgpALDB4YKb?E~wfs
z@rv<oKKuQzF8F$l&wQZ|<HGJ}$9y24I7u0VC+vuE6W@vBD$);jMqI;g@N4`K{U{T!
zh(E+7ko)jLo&a8b*q;w>zwPSsdFMLy(Tj5!t`i49{2l%1SLF}8LasV~)&uxt-mriA
zqa8fZFGxS=fxVzVdSTbdp^SVGJ!pr1AnoV}K0wwj^y9h?S@*&s`vT$s`xyM3^LcgM
zCSPG4C;y-w_lb|#H@pysh%dxt{0cuIE>oZRP?7$qN1Q+};y3Y!yo&Q+5Ix!Vl4rBO
zCf|ZLt`pxG59<MC;s<)sKmDOUNPGOBb1;x~j&%XK%p-XkaULG1hg^6;ALQacAa+d|
zp3xWoWW4B4ef&$ESL!kU$OExE`eQv+?Fcy_enx(!L-H&9kba1d=$|yAKY0&-2Z!~R
zxOUNJb?4ml^bbDsMmzW>&cHkQJnIttGH>`B^GP0$ow44~5AEQMxJAFzLqFmx{K7wR
zj&|ffl!*%<`8w+!<0W52U)F2x6F2Ce^^$!m`qB@25qG)Ib?k!i<0q84&-l@seF*bF
z|K#2D!+hW$*ggCpmwM!p@JfHo6ZO##L@x7=-o)?n`@j4h9{iB_L0n-yCQk-APvzW;
z-=mQypa=1gxXO7I@d_T`8NSe0Mf68c;sf?TnekC3UXqV-e#iM5=Xk6GoZE2DOT6bh
z7yQ0Tt*>6EKgLgfTm6^wM8?CpH|IXQH_CY+=S%R8{**aiq#ftZ^b6mdM{@oMFXRb~
zm-^(h$RTg%I}^+safSNW4fcmU!H??a*a_<$>pJ}*pL~cog8vgID6=0SzH&ZAd|-c1
zJf=N<#(tmYMED)`uv7ZuK6#`GiC6Rk;t!NnWZffv>5%>z5Bd=Y*+&!qi8sH&U%q2W
z-hq8Ezr;J{hj_%gz;*Iw&JD;1h!ey|;voJ^JI2BL0N5?}nFo;d5&HzOL*gg?46n>D
zymF3%y~7XdBJm0NtiQ}7>ofY}XPk$#Zm_<gANd>g$=mS@;u<`{AN}GVjFWlbd=}nl
z&p9}9>4)bE)aM)&ga`Vk9_{HLyI@?jC;ky{@C){@*e!9BaS=b5Z|sKgGY<IWJOq81
z7uHwWfy_Va9dV5LXPnqSagXs4$A~l75Be~E{EzX&8|xeS4*a1%cDM4dqkHf0bhQ!a
zh5dl=#Xf_55dO@#@E81%cubl8$*YMkv<Kmhe3m%C{NN`bc^CErGH#H#uOhse5Z*xe
zGa>$ALgFE)>WMvq%rD4(Q-?fv<UaF79OOM))^*|r_Jmzx-`FqfBKE@jx$s24j30T#
zBk~FSi1CoGfaFi`L);+lun)%`h=<gpJ$8nF(H=gqKaltWs`HJVgZKxC|A52;6;(fB
z9O%z@;2nABjX#iwtLrmy0zaZ1NIYbHM-KHt_9f(1th2-=_@RIF#V;8L^FVvz19Cyt
z52&Zc2l!Hv{RH_O>oj&t8F}P)#3%A6)^X&}59=XvI1eQcV11zs(w_B+e#r~TE4hw-
zCWJ5aC9a3YU-CBW2K&NJ@o%o<-}oo~f}PWj_{IK;>$C&$C*ls*LGI(%_&@hS`cunX
zC+=hK*d_OgSMY}&<2Tql@sE0(gR)K{4?Vcf_{+!1-%0(2fARw2BJo${8#!D@9(o~%
z`9L4$i8A|Z^aQah)>Gz>ctBoG{J?)HWB1g@{&k2xVUc(a59o;>x<dQ`#4m_foCk0p
zy|5?r#r~KVo~PjV*cW!EL*`5HX9loicm?qrcvX>kV*kUs%RIq9c1nBv4dlLxI{P<W
zC;ougnF$#;NPbN_@(ui)@!-GgJE+e-2p-7;Xh*z49()m>;FCB(JvA@DPgsxeWBdXh
zST9*W$>W$uknymuMsMyDH;I?56RbDn&E(TuXJ5|x#Qfvm$RnOpW<Ka2ez;G4c!fvo
zgglG93%-~q<^j2kk8?TXGfvKFSf>~d?I^<o<3<kk*her9{E#woXb+$8hFu|#GJZro
zc!LM*3HxGwhd+3s%>ILVoWro5l;;7ztjpXdo)aI43)mC$fqvwB*dO+a9g_cH*R;n^
z@sFs8pCFI?n0ygrpUyc7*U^h}QO;Yqj$ZgL{IVX>Kl6tj;HRv2%s2iHuh<Lz0>Urn
zJml~A0sY`N*cpC{eDnmVkKKXX$KJ3r{EYn)c@gz!hu-8#@I`;<4PWqpol?e*u{Xv6
z60eDu?0cA3<_%QmnfRh2cC5blgIxAQ+~@DMV=weWefCx80Y5>J`K3PmVHe1SXBDwO
z<_EilH|!OZyvi*D_yOw$ah`aK-Q#!Er_6o)k@Xk5U>!sr{=&RtH^?O(5r0%U#0Bh#
z_bou~vp%99a)?*->k8=?Wc*Q4^&{dP{tX{oCl2C|tkd`%$U0&|&hd!n@D0!8C&YQ`
z(T?*+-cw?q#`zuJIpn)CJg>zb*)P)%`4q_cC(nmL@+#sLNFIcK*gtt0_Q$$_y}~PH
z?1y<KZs89g^Ui(j9lp5Ec(7M^Q|DhTo1z4M0iXDxsuz0T$CSy>7&q4$H*%OC9n!Cg
z@E;b@GbnO?rDE8A)%K<`>t<Bscc=K34vC*(k@p`=$UN&1d)6WTs3Lx68%=&e-k~CX
zYeMpS;xF-*eG2&;ahd%G@s~J2p2_n8)xXJKsZSnCndh7MDRB-zWL}Beymwh14}N*S
z2*1Y9nGft1yW%==68|f2|I7Z3=TF>6FXW*g`(ThbNdADHv?HE@#1s4wf2JK}<RJ(C
zXit6CL-J<$!0#A8<6-{|AMi~-tRpJ2eo_y+Ag-$R%XQ`fy;(1?8^)y~_pwLz7vu-*
zr|5_E4*SM_=?^=kzB)d&%)Uw;SK4KG(IGs<MdnLI>^Lf7uR7G}jeX#sCS*Rz|B=f+
z6jb9WaS$X<>X5h!s`5-_+J&_%)A(K8Cogt|%rnnT@z<cpyyG7{U!sg0;ulE2<O`f*
z6E|6}h$ozj^ZZ)v7yn?rBc3Cl_{}<meAYYSEqOn2ntX)z#CMSQ**Kr%xg+}+>d_zl
ztLp;oRYX4D3*kK=#>F_vtB?<}j$-eeH*&tnc^~gLseZ_L74_IBbDg-4UvVGgJXMVc
z+*i+Gk%J!Gr#|Zva@ilUzO(-E{WiYqz`SxEM}6`M+F^h234fexs=q(s_xxO^J@4It
z<S*=3d0&|KYgD9P?2|HjQYL=^@fYlsage7Fr?7Y83i|Qg1Ny<Q*v}%5{z2x8^K_p7
zkiW9NlTQ#IxK3P9;{f>&{zCuc7vv?#BOk{8@khqZa|7xzE|7Nk7xM|f@JRe(y{0}$
zef~}f^8xRShxkHVgGcO~I7d6?gLpuEz`i-ZfG6ySaq}J}`Vep6oq8bm;eqiIcbGT&
zC;l_f=!tz17ttF%iIddhd;>q_Jcu$#-pzZ!w5K29IPr&l6!8Onu_x?;xTfZ}tY`Qc
z^Fn?@oMIlyV~A_;!ak7uv<Hb($iogm@+|C@{V4Zohdj;$IsXUo3-Uv?J#vV{tQ+_n
z<DxzO!FmJoe1Z4z`JUJgM<0C7z14p7`F<1k`MZ+rv-s`=a>(<TA9#gN#)o|BF>fIL
z#JCv;_QLqo^$}jVk9-h)m?z}H8-7GR?xPp;Nt}UK>Vfb8(w{1q>mYi7#A$c{@o)5D
ze3WTVf7l25aUbOGMx%#1KIVh#*f)B>1A1_ucHAev5w}>+utyL(A-{n~p06?<t~2lW
zH}L?zRAfD&9%bgAGV4a0mdE{d-fI;{FfQsb9^zoR-&UFT?}y(0z(Wr8IBy`XvQAT0
z5q?;2iSLw&v&aEyN8BZ!VqF7iPd(x?_tA&-fO?#7(jR#kJdjTzpL~Yjg&~LM25NhF
zWIZ8Y<M&zoehcJxS>%1>ZS;#C=*M-CIE+82@fAP9jyaE^-TZ%Sc(woX{~s@3+oRsh
zniI>v(~Er)|9I|;{o*gKkbZ+A^NgKhN6ZI&QIB<&de{x~hMjVbspePJfBC&0c1`>u
z&amFn5B|-3P$r%*f9S_LK-^%y=;wvySI;|XuT?(rhIvOW_0{<yPSPK7m34~vKtISs
z4!={zK3Gq=k3PsJzEPiXQN|Cc2U3r5&7V4<(yBWukKj7-l{iiR@X0tChl<2&ka03U
z<ZFzF_rl?e@zIX!#AV_(_0SLdfM@D)9Yl|)h(Ckqg}&%To=TZ|l+h1FKkn<0^;Jdu
zh4IrKKWE-puh18Mkw;lY#=|(!n|W4|^AGZ8;w1gE4uYJ!(2o7c|6}jogZwJXJAm(Q
zHk-}OXaN}-XhT6os3=e_0_`3sjImaOT0>=|;Sx!!DH>pEB(Wk|AcTtnQ^jGF5^iB=
zv|ymNmHly(mMB3Y<)To6A~z*QluKfg{&POZnXmKC`Ryikn95(@Ip@6Z^FH_IdEayP
zxAtf9hJBH1c_!<UeU3a2)id#dj{JaLB6{{|B}_fImam|7(Q!>b(EO1ieT!TBMEcf$
z7{~tXz@PXzbYFf*I>xCF<TVaI`Y3ufj=jydp84#=j-EsMWMAt*{(}#Cb{=WIaq5kI
zuzIJinGeOE=k%(6(l>jHr+n@_#CMjS)5~C8C)H8<bFLsByuYIleIGr_pW+$6@(2Fd
z8yZg@_vyhp`a*W57yOY6Kjd>x=bX!U{hD>mgZ?hhy7b~XKY}BTt5?SV|JH!MDC?s>
zN1yNS-Hg|FskizYsBd&WA-=>P`p&(q<D6FC<h;>5<IGp*-AB)J=LhIHmlePAqVdif
zJ=YhAe{xti`IV2n7lQJ!dLhrCM?UxEC4HN|8s<28*gD3cE1&4Q&?Sd{SUn=IIwSvB
z*Lv>LBYE71{EOdMH<5nyU--Z``t;+u`!MN}gFfsJ_$mI}H{W>raz0BA^NizX^p6g|
zgziK8f;MyxOh5GHJY0N=L;D+Y+GmMda^P2;U{C&ELh~Ryc@KB?GiUcd{H>mjb5-YT
z&H<dOIuGzWiufZBdy7+Y+rNre$c~WzLw1Ag2UA{t2F)uWy~-!{1?Y+k==~=oH}v~E
zkUv8Dfb0SJ0~GI&+z|gz+(YX^aR=pxM13GXAwRy^*>n0r7m6F>@j+gEk_SC#p7o*m
z=(C4&f4{roJtF-XpZuve_96Zsusq@Yn)&!(cl<YW>i>%<&$Exb!r$4MzgkCp$cOex
z`c&&f^~`$s)Mv^U>}s4onZEVC)`iwdv<^M!`{_6BO<wcpf&A#$7wgaEIemEJhxPwE
zB%g8QK!+UWk=wbhd_zBp=HXL*qYw1;^Uf6$b*c6t-J&?*pU`=937uEiAC{2)A^9Nv
zvK#xl7N_iMUgj10kuR*rUM1uQ;>>%0C|>A^z1W96`Hywj-T4B&T93VrS0}`eb0PY%
zpCuRhJ%{|#I_`T9?RWL?jV?R$H}}cmy|?dHypJFk)bFe3(D`%EM=xFcxrLjW|F5J!
z@jle|UdHQN)Omi;^ao$rWuIeqSv@Ro%g6dOb`}4w`Io$EpTWP_U!HMa9yMPc=cnxI
zeF%G*Z~fdil==ww#g{xS{=G+HZ+##;@+Zh|_%A!NAAd0(8VBV=^Z2p4zz?DGByoa{
z{RV%8=;Bu#%CF+l{TyfC<vF|P!`Ov<;?H>Ycdbs+Ptr$6UPNE~vyXE&d5%BPoA1W#
zx7Bxfhd=NieTaTre}aE~fw<&<>Z7<w|Hq%_5PkF60bTR>J2}J)JFu_$;tZ;1_!r-q
z7wMf{_>cLn`3L`VP4DdMT$KLE;qR@TCx{>NvJd{nDgUvaItA5f|G&rYZTj6)|Mx}S
zKIYy-Hk`9=(-1##?&p24ya445?}wc;%HQUDkL<bs-~Q6_)r)R>a#D5U<U5{w=j2ya
z?hn@IvCdna8;fV}znxFwm%R9s2b^z<4|z-<#lG^1cv82}b6!e+?yFzcGhe*&19?op
zX`DDFpL1$?Lq3$B&@+#}@LzswJ^nyG=ssj0e=q3o2>CnYmu;v{IZtry{j7bva|0+Z
z8E2oQZu=fey|>So@5GV#H_vsVK4R7IagSbl(&k~fbjs`Yzw)Rp!`73>^b6#<{I=IU
z^uW<Q9dz^$`U`m}<6b=DPrpKb{YKK0AMr(x=;&jt@7i<g;5WzFk6Isn>*jd;Sw|g_
zm*rP_&`0y194FtC&vz~K0`aH*SciPhgM3G)Uhpe&k`J=0ar_HE@{v9e|LP?>i&uK_
z9?ChJ{<q2RcdU8;{OxZa;!ou99U{G;Pafms1$t%|`9L2@f9z>o`h&U*>BBWY;Kxwk
ziw}I7ub-edb&%fGGyJiO|8vQYt*>8_C&{bcs+Z33*?}KgUp*y<x}e_Bzwz`=4*3zP
z7y4)@|9YNv%QaMo)Khg=y+Kc$n8&}=W9av{)qQ-Zljx%h>46;T1-$1&Hy?0$HNKJ`
zoj21Hz0x~>q*r|{JGdqXI?z0J<^S&UKXGV1^ZA+jsct~_hw3SGe``_vC-S4*Z;4lN
zKwfb<Qq(VtH~CwBl`q&s{gj8q9lhaC9v1)dr+g?6@Dta@laD{igYuMl#<`Dg`N8kN
z@H=^%-;mdH{kVOQI%Pl3kI^wse?VXM7wDUp`oSlCd2T+w)H&}l=uuv!2lxG74t<sH
z{?W0n`RXBkse9~^{Nan=pl6)^>E_q|_2CN-?(e8;-p|kn{m>_Wb|2C|d&tB1P<QBG
zUiO^)#)%W>Lg=~A4z9_`PuYR}>D_q#X*_za$%S9@<R|?EeX<j}p7VF>ySARVF%Q44
z@nJl=o~Ix1Gk#=Ueuhuu&?A@e;*}narw4L7ze8U>KnJo<%16KSU|q;>@$Ej;KjO#w
z)}sgX`LX;$9(G}WNdNpa$Ma9}<J0=qvv2WSyxBixyt%LMqDOWjr#Mn4JV!6<sye3r
zs^|JN@uz;sx9sG(IF}F5v#$M~`SPzg#us|->%T{e=;^zB-=_ZQ1DpfdpF1bgS9z{q
zK{x5pqyEKn{Vk*y=LgOe^bO{z-y=nS#y{oJjB9-8tMmi*OU^U+lllCg|G6)I_>J}T
zYu0nVpf4jo{g6M;f9wmKADid<ROfcii^yYN;haQY?VM|{KEJX)`O&qW^C5h&AAg}|
zdXd-YB~c!eXV{q^xi1d*opJQ(exmah{@7ss`zGRvo$MFfr#H`?o5>IMEs6TzMEU3y
ztZv!kpzE8zhk8XHzuo$fAH9!~U-i%GseZ<L7x_W|A-|}f;#ZxMr=as^&lA-nb<eeP
zbaC%o-unvUtmC}ewR%Htd5j$DLX)rj-hGYEx$$Lv^qrT;&(1&cyhPloH}bzeQa*Q{
zrQe16E&Zgv%y@Lz4XUT|KYHw}uc8m*<l&|7+4!lxr~5nhW%dE)sXx}ym$~LQ_V=#!
zp%5K*LLVJS&+LUybxi+9&-zj466zLuuF=s)<1_0@UemKUp;vt9ztP7hzr=_2^i%kD
z-!()Z>QBV8arQ^9J!fa@@kf5H56Jb|CDAzYSclyD2=;Z&?(8WJ=+QX+J2~`$_!JlN
znEUjv&k+CU&>z0+Yu%?G@dx!=_~&2ps`DB4=2ym>PmlCQZuv!Ds*goqU!?D(C;J}n
z3E5Ttm8Z$;IlH@W9r0$qeT#YfT)gMibKbUL$ag-@i=9XNyI|MyIFu)y<Gy&qS+h@j
zYk%eMjs0GT{gH9{YWZ6~sa~A--YKv8hsgsKw7&PkuAQF`Ox>{dl9_|m!{2-Ig83`@
zD|uc&j=tZi*;>TM%w=m%IRAx#>eeyu`1-=##y0Ddiyznr(674VIemHVTwa}JM{?1>
z`sulK@k4*;(gSSbss6irblAsxdi!GZU9*=ui*NGclb_HV^nL+<_F1mI_t>xJKl^$g
zdrbv>&&&_#kKDdL<sbGV^eLZ^C-;N)bN2J%M}Ok)vGoP=M_#k9`XV2Qdv>s2v>%mE
zttVfww{_5$f6RmGBRSNq%op_2)ZN-2<WXoneuyqRqstH2LtmBX|HMJ`<Ppd(FFf;O
zSDrC;(-35L`6c5Dzxd+U^onot8xMVV4gJ4Z<Dod<&;D;Nl$ZFo?_>BkyVJisiBEp(
zntjEgdZ~{PfAlR5*bVB#`Jr*vp<jH{3zT2g6R3WPPkO)VtAp=;u<EVYPyAVz-&kLs
z@SMK&sn9z7#QRU{k{4ayHJRr*KO(pL`0_hU?4_Q{H}o!F7_T1lH~z@q<r9C$>H96;
zBl&)cKZ$qeZuE`*#>bxB_>p}!Hg#Pd=f{xUpE>*klfJZSuwoDM`3GEi(X`{QS~<R{
zuj(ItvKKz&ef|sSU4NPRj33*t@NXy|>09{$dh(n&(U;OU6mR-O^-w>CzCO>o^1XeR
z`T9kDqJ5qIQQxkwC69ia9Q;`P@)vPOF8;=D^ynJWr~c5jc%={P(nH3JYkcr4c4l{R
z%D?dMx%<Y6JJ+5=eQ2V*AYRdNEnfKD^h1w(e9l9C6@K|EJzAf?)4zU^9P+(9jb6%!
zfBJ&vS=Y7a*1@0rLoRgjo9l`%&)Gfw!8pkN)`Rq?-k^g|eU16<L+9k^=wrmUeYp2V
z@|!po7y23V?E}SwICNf-^ZB)P#lJYVu50n`zW8<@+Ru_hz7X&D^4`w5iRbbI{^T+8
ziBtB(kMXX_i+^@c%yW>oYk5V!!XNuW{OVWP%RDH*dTzdW;79CaUrmqlC4J-1ee%;Q
zdhCTRzrh#%@i%g~re}WRIpn{tM~e8zuXX8*9`I{i37zk<3qNrF>;0<p4tld+aZPW|
zXYfHj&z<w=i|n(+zw-@#OaIO_^)dLhFG5G$L-ELe@CC(9qI?a#?;9z~XC)LT-6Fe=
z7_WVB^=D4sG^9S1ki6=o{LU`Pr+lvu7N4%wQ~%FKf51NKnR-Dk^$I=x2s`pq&)r9_
zglT8rgF=1*`A?#}AwT>6!*?C}LUhE1yuzQ2YeW6Be#U-SzhfUk5Aq#7n9mRG3&`!7
zz4fp7Rj>GmxRY1d2e#FVm+N!>I<M#3SFYW&uX0`h^+C>`^kw{-e%y!pH2osG@nil!
zTGa39L-8qp%8&M$^1FEEr~HE)kpGGwbqJ~h{05>ce#Eu=_>}Lh@40#k^#kf1eW}al
z^Lyj<d;HG+3epEe&wP2wef)YZe&~muU8}R=QXe3G)I;a@=&7H^IfryUNYDBJaf!a~
zXYdR0=e+}Z&iS2Bm~VahPGm3R)N|(`__5D6o?qz;=!rkEhk59@uP(6<eeid3>0>;{
zpZMTUke~dfP#uwHoe$6heTi4{upfQUCw-Bhe#EPO4825t&cFFB|Kiv9pf}^37w|Ln
zUcTe!=*th#IP}B;d-G%R^HcNXSNTv~6leHJRKNJQK1khFAB{`Ix4iGXSAFnYzLuvT
zdY)TXpGD93fZ`4vdDe6DohPHOPje2E$d2qyF7)vW*$Z9u)cWGjdqwg={=lx{lit`*
zp4T_w$9Wom!M|&IVt4EDCvm1O7|(y48{6NhkMyh`)nDVEzUdKN{@UoT{v8(l7-xO+
z*@3?4gZ+)C7k_tZ9($0(`7}L?fAiUmocb8+<CDJWm%qx}^kM(zoD5z1BqzDd!#Dq;
zXZnNW5_iVy<HRNYotw~`@eS60uS`Dkz4wqWtcR~={*SNx?`dcLY{<2B_1DfRoL~6v
z+qLf>_^bSFzrkPB3;hz5AJtd&z;k(B{ZNPW73f2GSpG+!T>OXK^fT#S>NNilf9jci
z693{)(0cfX^rWse<LmfR|Jjqj!sOF6z2h(Sv*7e&uN#;(w$q0C3V(N#=Y8tG?^XB*
zedrf_$8Wqo*WXo`XTQpyTzl@mb4uqu&UuWpj=T%`3v{17-ACW|G3Xl4AC1QczoZZO
zIiIulE$3YI`ukS(H1DC<PkhKf?8jf+w?1^g85iF}@N?@lbnD-H5)Yos6Y2uL%lxdK
zWM1R<{8#=o9$E*=pYoNutM22+d(n9d<{h-?%z?`L4)xFbMtRP2{Ss6spnSio@4z>0
z*fOTWf9V(c9?^6D?(cErNA;e+`@3)7#fb;w{XLuK=v&u1;-d}ulRWNPp5XudnxEsF
z|L3~m3_bio`I~>Di=I4z4*ujB@m`XT9`iYV>?5pa9qW3IUv$_3AI5u*o;b$0>y(>b
zp}140AimQs?4!QYC%)JTeen;~75o`TzcA|se#9v~@dHS&^g@1m!Jl>TZ#=)?=e{dO
zH&I<mv>tu2FL}w2KjYYgeO=?%eR1jk!LXBS>!;tipYqZRR6p=-{k&G6+~)^S-oT%}
znEgs9?~fMM1<3E9`rvyT^&tCc`yi;V5+~{kzw{l5b?pb((f-AGsJf!gsB83vU+d`e
z_3Q4dr|On<^yl~yM@gUk_$5EnFR&-S$#Kr}`3b(n7d!HQb|;_viOyr>4|L4K2YU2C
zKJ$z-5AsL)a^E$2_B-rop6B`%{G($WyYU-z%roBpt%NBLKCP4Vj7Ohd<V*I!4>Zm-
zdqVNh_+h<INIvYZ$s<nqeWG(z=UvXj?3bYZ6#i1L<i>CAU+`}}|6ot^-8Y_nJ;y)4
zq&Mh3ySTOvq&IT7hQ{H8J=G8VSclx~V_kM|ADx_+*XZ$Mh+YZNf%t^vg7|^thwPCk
zzxi%P|156hVd#9v`Ix-x`zxp~_k87!3-8{1?AVID_7my?e{kRan%%sQht@Zq{~K=|
z^x4rm8Bgqr9u$Y_9(&Sr)(QI)?<2@>o_ZnQ>tFc2yuvQ*ZM;5<-Q8E8_={_C3em?0
zdftoZAMKlsr$6;V|DvB&$MCBTqGLa0zICB><yY~EFZ*$F*_RuSuJ7fIZ$o(lpW=gl
z`Gsra`2#=XZ<$}kpZ6q*^o1^eMGt*;<#)!Lk1l@Y8RO8!5B>5Bbm-AMercRMPEY8V
z=NcLh%|n-exew7}Z~F)S0?ii(;>mc=jYrot{}B)PLWkVwKy+N2hmPm?x1MYBpz(?3
zxlfPyb05+l`?)sWdia3Wac!LW<a5nm=tn<=Km0)S=-YbcLw+vbh&$`Evvu$T%@+s!
z9ewN42fF-~T;dE}`XrBW<TBrS_%F?;XLRsuU2=Mku5skF4x}&FQ2mhy@>;$zKI!Ab
zeTc6Tnn!Oi_0JBFK3%h)dGu<WYluI7>Av~=09psVl!qMnV0Y&N=0odw?wVcg2c3J1
z3-|HGPt0@Qxrz50{L}hSKJxwqAKrWN1Nu_G+}CgEWAs=2$bEg1zRK^`IQR8;C7%2K
zL!aq)*?qs{?>_Wv`a%6E`ub9Tr{Z^d(e-yyzFX6uvM+l=`zPlC&Ns|={y-k@GxdA)
zrk_@a?Z3oV>Q5hOJiR+-p=W)rJi(6k`{+V`L67X@yoR2PXCHQ^Z+wvtU2#D!dgV|2
zlYIEq*N8`X3(9l&p>O@2_W|A`Kzf4Kh2((NgZPE?4bg$tf&3BTAF?NucgYLoA^Agp
zDxUQ>`fm2}-c_EJr^v_8$*+%x`ek;pzH6wzm!GUBKe=z5`{vUxImstq%UAr`dtK;0
zx#-__YxKxp<uz!&JZfF~@f?abc@?@(&+H>EAUfpl7R|?B#+`E>aYdi{Bl3txbkL<I
z<IJNs&&iu;9RK52iGE+iduZ={mfk;e-7y#SG`|bt_ayy3r1N^`>->oStBd^4`#Ik$
z@I!iICv}kDxpsap9?%g7=Htga_OuQ@=}%rX-+0J>`7?W&@3}ngd_Z22zahJr?^>Q_
zFY8#>bNoW%_z$_o6Tjd`d7X6lA3D(b<gl*!;?1@3(0KJ%o`&@38mf!*Ykv>%LyxZY
z2kNA{>KZ+0eTW{vp<ntTFFwt4Z9Zgg{$ZSX5I^d_I)sk%FLI*8Z|L8(e#!a`-|L?`
z@cG?uA4V5F`bO7!=-?k;5dUz8GmiY1&wgXqDv{mriN1C54fzkW9z?GV*)dVy=sR?B
zs|(^r99Wk==rxfZ<N@dP<lsmA%6)Y8DZZER|2cfus(*K$tDp9LhVR_`{|@KF^v&-4
zfjz{Bar_X9Bl`09=+>RN$D8kY{L6in?@ZnIy(j-Bm%h<=alWg=w|#?k2kYNgqz}*W
zed^afd+q+Kem-O$;@sDI_BZx3=Cc#}klfZwz0(8w{?5j}!~O*Sklnl|<qz!Q`x}12
zPsxdn?{>tc@%ZGw^lm;q$jAH@lA9f&b)a}?L;l6zApX=PXnbP!LG;xv>MO-LBzFlj
zzRByp{OP*kr+$y9pU`LMlk`DQpU95p$v=?(<PZIleX=~KU)P7}gP}ZUedrwD{@8lt
zl+W<PPTBXGhYx)Ty5{A5`ePsbWM9c2tb-qZrGKJVh#z`Fhu(c3;CmDHqG$Sq<ksil
zhd-mMuF0>EUpQxS?gZJ(dgOD>PspJ!hWcLD-VdP<>5E>-VgE@U{OHTw_g)Bke}q5l
z>)+{>-`O|kHU5o9haBuqAMC_mVDjz0bzM9EBe(U<w-2$dYj$Bz@r5t<>6ibDW9!hH
zYx;ov(>~01j`lz7>3chJLGeU?<aeJRd4I)z;>LG0<i{`cJvV)ko89Px-V)_$h#&Un
zZ%};Ex44o=*oD6FjW7JTk01P!pI#yRpyNJ0+i!dBd$uF*+O>De+Xp&VeEsAtZx}be
zbMpB8{^Bj`$9MXc|NExuj)@)pu=)5?zx7K{|3Xgs(4RP8Rkup$?~3(1&hvaHpf9kV
zeU@>4cg^2L=&$rg&V%)h&Xe`4eh*Clq|eYtndjV(o{ZDC>2L6-|MYud`cnP0Ykil$
z8}s``?jO2p|MQp39as7LVE^}qoaoWJ-)HiF_|UN)edu@T4W^ysXY#5$&h6w~_r;(4
z{L;CU-{rG^^7r8CyK($Ve7SEO`AeTF&g41g=I)zE4s|B$r}6p#exje%k3fAfL|>e$
z|MCgG#fv^%o;6PV$`A4Z{o1$5JI15KZsOTE@>vIe=;{~fMf~MF_K_zN`H^#I_LUd(
zt@J9c<WF@TU*ZoR^e8{m6S?tEzs{fKSADnNUo{>bdf`v_qaXHG56F{<ZtlCRLynQ6
zx~YFthtGcI?EZ(p)zjH|@vB$-<g&hwzGT)LrXF$P!rqSm+d&WN^>g1`_LEbO8q-<!
zCwI*~<1780A3gNlzUR*!(>d_O6`y`;ufERBOD5g7+eL$&1;6MWT5;pA>c?@r%>J9}
z_nOeul|^gsdE*~09N#(e$ZuYJ<7E>%5B>06e|+a@leQU>&$+R4Wa!+OoW4JI?Yne+
zw)1FuWOsE{Jh2yl(ht$Q_>}Jwov-k3ap^sU_~2*uRs2o95C`@z@|yADQr<*Y9O|R^
zseX(9@pt*cdlr0qUuRu;6o2|`{jBxLfq!(J2gn!BXIvZSy)S;f2UKtIL4Q!3vY&DC
zw>&_9_$0qLVNc_YBM*JhyFQ+u@JoIt?)?51f1`hSPkxiH<#~RHPkBgw)SnxVULv{F
zt9&jVAU>hIPfw6PLHQSoD@Yz_J(zsCrWgGIzlQWePVph`^fCN`ALw7ii++VX`WStJ
z^^M0j`r<<z@eg&*IQ;`Z<6q{v&rj49`Gnm3lbre?e26P?r2nM{dWGy#Li5xcazW=W
z=8I!UU-U0u(X0F_zW6zNkW0Ol=f#0Iw7xhLfBe>d$=}`ZbACe~>MuLAp>tt=Bi_U}
ze<qJU$v#AX@4oK_oa_3#8NYj_FSefVYWS1=mi?1`!`pA_pFd^aan0Xf`g_cuK63n)
zegA65=Dzb;`w-{bf41-P2{$d8Sp1!I`#pIX|Mp4F(Vc7CuUXIE**b@|&+)qt_BZx3
z=-3CbgYQJ_+w5!6wXXTz6ZoBA`%mi|XJ191&^Yf4=$+pD&YShoBbWKc<JbPrb9!QT
zeq*1DKj-z{)9^F)aDT9V|G_S<$zgxYuHMJczwy?!FZR8Z{G=Y^Ltaw%$<43zQ~C$<
z>Bat^o#>N%`U`$&oO-ISu%7%6ofDPNIhlH(4$IH_M(1GqIdwrkl5fZ_|LL>jJNe3e
z{e`?N|H)(WmA=<H>ZN(c<41n?|GDH>{k1;a-&4QwWB>Do51-W6>Am5U2iET}rh`v;
z)_v!K@@>kYe#`Ud<CmV*HGU)Ct3ULto~o1jL;b3HYFxKyo_fm9tb<Q>WIy$vUDQK$
zj$h(Ke5wc5@m!ydA9=r7ck|rY=RWc1ka^-q{iQE`13LJ?7k%o3=v)8IUi?Y_M<32H
z@NFK{@31qy$lvmScsCE?mw)3E|M*iEpmq4a@#M3vxTZ&PSdZS&abF%FpKJV~BY(>m
z_EqA<IQ;|v(MQlb{pq8~hkt&EKZt+x^{3{`x5hz!i9SBX89gSy^kiRPK9nb1%b)aN
z9y;XEXBlTcy~#J`8}B(iLUiqa{9O~e{G8tSk9pS72hzXs)=52?Cm)+{UGuF&zv$76
zb24#nerX)K{0Ck9iD%ED`hs8fr#E@swQ=+=KIq4K_!l?yU>^C6hveX=?xW+{IPaCj
z6FphSy8MNIna>aKLvQSWj%)9MjHj<0M_<;ZNBmibKE(|>`T_H;hY$NNa>zgClbc=e
zi7)H36Mb8c-&xOn{JYN{>K3}z#gFlME$;A}@q<1@-+1z)Yh8M!H*tf1^TmPZ;sIUb
zoHtkpozyG+=5za5=bfJG*IeV<d9ZylKcxqJLH!o`@;QF>!Sa40d+YD*C&V}X@DqKu
zeoQ~5UZZEedM^)&=aC}$$)^wE7yMBDllR33J@R|}k;{HXK2hKC?LK{u7X6<s`&;`P
z`x^a=`}#@euk@|n>mTha(P2mZkN33v&AF()Mn7ghZQp7?Z2$YoIn{3CrVlj#zs$LY
z|JO72viHsS+2%2oeWrQnplg3>|7pDUdWRgfWW!V2kEveVd#@L!E*Y%Q@m`X@I$v?#
zk$P?VpnAV--{`$7y7ue#>G-f8H_!WE^U$#mgtITa{GAiO@vHw!XP)y8dUoIczvEZx
zjXb4}$TOLz#DO}ej<_yi#vy;;FXCUk@F)3J+{+8%9UbHOrS<q5BoDfY=0o`!dS40g
z1I>ft9^wn~8|XX&djAUfHRMmwz6i=E(0MVm|AO=h=@q8FjDzHY@-!p|#24gWQ2l}G
zcOt*lN65?knEjkr_`4Ey(|L{j?LPk&Cs~i3v*3sSs(<Q-@09GX=_~i?{DeIEFm*}&
z5ij~1^+P>&{^6WjK9~2^RdI!$a|rR`JjHpR`RpY>lG`}>kzW~SJ^m!`sXNXCy*E&2
z@kM_A!M^%*`qJ<E&WRq>b^2mY_tiVk<r#dlkGw^0`hv#gHM;r(^~iTt=!#GI8y)wZ
zW6B%oLUF1;<)4i{pZUbiQxEy+u>17D5BL!}_{E3(2gN%(=riS8^D|G%Q{-X?*Yry+
z^56^K^v&P+1wO=yYksW$`o2nBvb#KvE`2+n#y`JdKYEb|Jog?(en(F~;2aBo-b3gU
z2cy5xk3szy{?S3-b9SUR<M7K)#_12qMIZVk@1tCs?^^z6Px-_7A^XTT{E0t^dwQg2
za#{~s&wTVe#|Jx*Oa3w6IC)y$qBn>>xm@#m{wV*;C+PBTc^RMf3)ZndyOdCV<nP|I
zibwkubxnSBzUaQbkbT&}e%1H6-qXq-uIaDQL;X9p^yK^weQ3P8M{e@yGwu7yi*I?&
zI`+ZzAdfhYS6|qPKFGn3tiw<EFFKyni+qbu`a@s6qG$Y?rytS(m}h@Q4t<U9nb-k6
zXdQ8bFZcCN@+<xGSJ&z@d(oS9_`mV=W4!vpPWm0=@niqU|JfIx)}=4{w;nz{S1-(?
zKXG9`KKUI!#VJ1d6?*t~&0ocv`^mraXML)^lN|VVU*Bmw{`o68@I!C>24DEl@6rdk
z$PM|S`R37war}$k(e<1?@lP+-<wy7zH{`;%?>F^(;suhMzsVc?LY(5q`t)s|yseS`
z@X3DSk)82@fAZ5SKcz48UDGT7BR@OH<NN|W>xoPDgPx;jUH<C(O!<nQ)faYmp6KuO
z*%cjp@jL6{1AY4|^X&i3Cx`jovqN>rbL+8#Yj(4LWIy>A|M)}C`kq@a`NltedCpG!
zmR+H7;=nk3uq%1-ML*)i`sxxtlm9(uZ*^1N;jjEuJfVXR*XZ$od5ymDXFYz6F8TQ(
zJ<+%4ko@M01OCWAt&cDI#Rq!iWlwbFLwdyzdytF#uF-+)WPS9}V{d-wK0mX*K1JVx
zE;;l!>?p4210D3~%Xh8(!#Mtz_7Erf3x389@;!afw{`gsd!y$XU-F&!<yY#3@#d>z
zN6h%d$#d5CcH}SX(wA}Mv_JG*e$ao>i|;7-1^tqb9rOYCwT^hP4mtU!`}_-<XPou;
zxBi@7jb|VGTz!vzh+NQqiQMAC{=|HJ0XpbIeyC5UFPQrRa^VBt<Z1fQ`u}36^VY#1
zy6BJtIv0@d?0?a-UxWMrUG{e#h)?UerYHQ<lYGjq__trQ4*SZJ`1YJ$@gbkk6Mt|m
z9_{~~BjS@DydR=Zbet!U2VLg{jeh+;Nd5Qg@)y6RCw?aHiEnv?T=eZ+k=*Xvhs%rf
z$)47sKl9LYUf?`MUS%isyW#ss7wk4?`PYZ}4}H*YV}}=xKV$5Mr-#W+K6J&W_;O!-
znCCwIiW7O(eRAU8c>{eI=Ugw>CkHv{l^)S!AO1s+&NtD;r~L0ezn3RH7iYdJlo#cH
z`I;WpN#oE#m){wuPPorr>IOT|Q|=Sck<aO0e|rC079Vu#cV8Oj&z|d3)B|+v@4dIz
zhmyxS<aggSeUQg}^+3I&SNs^qzvzXX6V0P{blFkgPyhM|@~Myd59?W9U%?*yi9gDl
z;(}l4x71a9C5m5i^I!bRukOn);tv1(!+Q9j2Y%<d{3xH2N8jQ-I6Fdi<}d8$zH4&w
zH|y#*ji+z+Gavu_2R-$RACb#*{(!Fe5Z|tIoOsX&)3<X0hz|SV16};%i+yt(`{p?I
zMUOuuzv9(6>#!p~<OlK`|F$l@qfd|c6-VqzK79cH5GUkn{OYxrefaJJ)(p$v^d@es
zLr!(VeSN-l-A_L-A9@bSn{v`0x#`Dx;*0*sjgEfAdj;~7PyN$J8K=K-{%9Wh@||($
z(4+C_xh4;J-A6|pslW1p_4Hf%OzY|I^w;<?529z^0=+NR*PttpLUzEve4!rFyKCdD
zL+|=3$gcc@J|MZsfet>c%l{JP4f&d1(1*T}e!Z9Wp2>GP-oyEBid^0ULhtdQJ}2Mv
zdY?rO{E(Y|#HDNR>-5?BQ+^}9(MQj7er&$@@!WWJhw_oQ(|7S>^F2p5(R(-Z@FO2W
z_MjJbgZOt1#RGq_P8-sf{iyN!TJecra`G?N>Z-iVuc3ZXexP6U@h^_a>Akq;>`Y&&
zCw3%9`WZX0n{iOw@jLo;pMAXN6%U?UA78#xvM;nQePmpdn|{#4k2p}5>4hC~9=Yg|
ze^?ja>YVY$LH!whn9mOOmFkCUeW7?DH@le6PpnTq{$f75{Di!o>znwqKEOKkl>M~+
z3Vr<we<LqDu#<gaqIvkENA{;jej*OY!LF`7$G>=VpB>RdpZxp?l8?U0L$CJd`W5{g
zJ<z9mmwD22c4l|$TZf*=Nv^DCp7Rse)-jLY(3Ac{UoU=hUGv;$Cw4&3IO~#wob+iP
zx!DO_$Uo@C`uZLG8jr4Z%y*4H{CJL@?`qM7=&IZLLwxaHenGGF!cXka^{WtH<l{%s
zeg;~PTq%z{j=x4fJxdo}xc8c0`#V4K(T6y3tq#hQ{8|6Re)v%z*&lszSVDeEZhag-
zv5#_}pFw&SALxh+^!Ype<I{6-i!S@f8;RDlK9paqUqXE}#J6?%6IAD*zZdsiJO0R(
z{P0WkjhE--598R+_h}G)$Pd|9AD~Y0NAaew=4bi<-!uAt6uR#_e)8+f(ew8X`VZf`
z`<>#!`nx&%n>@Ewubd<C1M=y=@om50d=sjN;wS%ZKtAz(zWn2RS<lI9zk#mx#T$O<
zfqwWkzY?#|Ijp#6Cw60hazgepkA3WO=pSFkS=YJ0{t7+$mE87G=<y$@pJ8WuVqf_K
z^79f}pPu<Q{-OEuEB|n99{V@r>-QSI(@d0q>=)Qg97Fx1ysHn9AH}!4Deon+v$~RU
z@CW?6HqJbL0@>L(*XFZFUeh1Cd5s=K$F+6v0qwWdb$-bY`Hk;#+0E~K$a|2#i9>eu
zcjNqLTcP@)PN*+1>yYv8%VYcpsw@1*H9vB$<2gj1fASCcnE#6x_JQupPwEYN;ul?d
zbd7!q&2K~gg|0lzuKZA4lXu9WenRofF3!K`k$#PHpFPAYyRn1u=!s+Jy6Bk4PuzFS
zZu%d3$1i%;(+{A7ALH@seT{kKv7T%G<Gqz@bge^9^xS7}*Ze2t;Yai#FBtFKj(qxW
z`hffc$|q2Lg7|~#KFmBMf63SU4azh6Z*|t+O^Ivy$~=B5uklyU`7^orJydV?C;0Gp
zD%O_|{r!afhjY>r(ziUI4#~^r$tU!-a=#PT@Ai$ID|uX;k|#0y3;hngTi5#{el2e}
zxAA?x_2|cZ>yjVe{EM9EvKRfczj|mL`jubYhv?u3(wF)2r{6!)UqXFb>YW_Wc={)o
z`}pEF<fR{eNKg7L^YZ-ze`hz}pIBeLHVz;7r4N56>zcg$9lz{>Z+dL<#kCKv{><r{
zhOFng^Gy9Wy5bMNe#eX5=o^0!UE}a453nnFp}fbB<v;IL66FtQygDUciWhlN9<iSB
z)-}&_{_UE*n{oAbSvk&lD9_{9e0-@J>W=v97WLEe6FT(b9LBo%@pq%@96kG8LGNch
z_dSF26m{2o6u+P4IX=`|=d|_-&SRkZ2z?LYdlUSsyUvG=lMf;JA^(Kz4pZO!1&S|r
zlNZc~>>$sm<F5IUI1v}}jrCyWSL4tV|MD&L9HJ*4)qC?H{-E_-lc%|^e~-&J{SWlr
znDyz+dp~swKJuM?zINxalPkaHs84A_dCKoBkdOcBm(@RZ8Y$X;$#41!dbY0fB<n){
zuHR`x`Yxe7jSl<5l6?Fm=iy)7k?++fb|5!8>@U91QO6;^<Sp~8<2gF&nEF)0pC2~;
zjH%a*>m=eAI%g7h(DOuoTSD=<txz74Pkh&ozVBAVk$yxx@h^VvzH>tN#T&bbXMMkO
zDt2_vr{Cod`c3|#-_fr)e{;?#?^zFj)<d6P8>f!(8}BjgYw#gH*-L#UKeVp*XY#sr
z=-YWE6o=N6-_#lMvI{-CW;b-ikGhOsaf>hg4FAwi7zgzo4c2{%bs>7t_n-2%I%}V-
zj@!rUpYVr&<MaXQIlGH*c4l|}Bkski{Sdtx?;KqJ&F`EWK>CFG8mN!Kw?4;rg}#Rn
zSM1Dg;)z}8-8lXUjiaxTA~_(sB~<@P$R85<CwWRp9>^~sJ(N(ME1`Jr7VRtenLGef
zPJJD;KZE=dX8lnIq5Y8S`xgKH=Pz8aT?Os8;I7Yo{-&jC$5jv?Q2vGD9?IWPJeN>D
zg8at!67q=mM$J0)-#Ms1;+|fei|Nz(weO42WoPzm=zM(Wp<9>#U|8I<lfKD#_o4f+
z<fnNbzx)*1KZqCkOT5V|CA80i_=56C3H^SU^Lh15z2R5#p}K5;@BCjM<9y%mTYLYi
zPo6gax!s>#G^R3-fBRi*b&&qpQNPO$?$eX}sK4Z&;wbBZ^A&vSFU)goKDzt^s$=+q
z{ET1ASMGa1W#4Ll$&c|N9w56y<I$xL*HFLfe9Jkd-#dYRhdoi>ieGWa?@EYo$c`m+
zj>=E;hvL*eNWPW-)Wt-9H=;k_Z|I@JPuW}l=v>Xd!FyQwjKAs&#E<)sUyvI;e&ibR
z7xr|WNU!|BbN&J4XK^JD<Ii<zU#>5P_!5Wg&);0j-}+vB8i!wUxE43idO6NIiTY~&
zGNcFjO&y{KeTezVr*-fNlTX*a4>ZnvXuk#V2aUrAee)-Fb4@O&9`b+a9816LTo9k?
zlE1_7y936FD|JnM($8B5>UXym^~w0;PxR!z{?D~~i(mEDz61Ylh+d-g+7O?RT}y~>
zNWaj1eFOWjpXcOE{nG<I+JBIPKG@OtMB~&4*XCK*e$I3Jx$iqc<H<qao};U7lFxn$
zAI6i%HAL5Qd_i)e=bF9f&3UtV<Z_+-q2nB!Jm$F%(Z{F$!E^pVUl2X|{Ks#9dHZV*
z8{g4iB--af_15_abe>_p-+NUr<xl7D@|5xR8~RKAnELFwdaHjcq5dWj|83|!)M!z^
z2=S@Dx1sU!xP6Oy=b9g~J3nGy`gWgR$PePo_qOgkmykcam+}8coL{IH@~GdtHs1do
zbnShOdLs_d!6(0#f83Yv^%3-;PU{QwH5r%uP@Jo?=FzJ-L0?>Z?i!l!d=GtfTb<Td
zpzAq)^+oy<=YII)50KpY7Kk75>AO(x1@v|N&^r1<{PUyi8$2g()-U@n&&`+r(B%*M
zin-f8z4zL~dOQ5ue%^R}hrG@&Jm<gsL|%mE%lDAI#fSOq4be&eBv-~^%3&S(SiQoR
zzvuSeQGDV{T#}#O$n8G*u8l(%ALyHJoOyX4AL7?K=wuv|&${#>ub}U_JmNWitjEsu
zjURgBFZ$86hq~%MIrF~!!oTn>Zs`x&_Z!bY@M|1D%KA-iagPqfuX*^TC-G)Hd*H*m
z@`8LMKN{znUganA#ku%*pMKZ{n(w?o-a<z{ly~fh<tO>oJ{Vp4fapT+JDk(VgXq$?
zy62kT$V*TiOZ2`#UqVjr3FKSnTh5o9i}8Q{El;X@<TXw_s1xE(+_4M$`fiZE`3rrk
zJMOb5KQWF!IS0=1{Jezp)P~LzO2`k|&^dg!C_eQ|{EqzMgFWbnzWKlWU>-W+lK#Yn
zaq2ia;?R8O2hQ2ViFxQG%1b5GZ$sZD^4qk7c!K)$HhUP~hV-h=(x-D1`H}vd@7YiA
zXMKlp^d{ek5BY)~^+V)we&@M)`bgK-p?`ehLp>o6dy04Q?7lvZ+}2Bet;?_2i=6nD
zPayqT-#YwU9!NAFk_R36nBR(*|7#R4<hLKCFLudmd66DnlPm4p_*uNyre}Ib&wBhC
zA5i{)?xW{g-GbJ^pZf2-+Wyi0SKiHdwXb&W?HoMslRJ^#@k@W~fM4|&s)PRTExxVq
z{Ujuh`E7`vyawe5$S#oG&_{<nzbTZjtxqrd4A=4$em!SDekCsLYwUaUlj2qU>n~lK
zr_azg+mGm*^ac7EeWZM0J@o~B{K+@+ojmCO4e6`o>8vNd`{9rLk6q<s_UF&?m%M>5
zexZ+Wp2IHAhxCv1qMt~)=~+Isj%$6MYx^hqF`k~(8Sit+<-A{gB){j0o~u9T(+hjx
z-?ey`FFjY!ApY3jICW7T;jj7|=ZDTG*u%Q=1HW)BK8)uF>LmTrC;G;D-^V}j<62&p
zhxwsv{z<R=9bM1mW%~?vQ8)0Ze`Ig_W$_1nH$o5I%Uef$7$+}!k1vnNEA%2R^a<>S
zKjYOA*NK~EoOks2Z~abXUEf!FZ|hvveeb8yVF&vxa#&Y>_M9J(-#GK=M?ErLy~%l=
z(~~@FzWefne4$Q5&z*mphn{)(GoF7!`3JHuRPUkp+K?YY`hnzy^b7er#0MlF)Tbn>
zcTk>|4~+Mo-S?OBy!<0R)eZ3n$te#&&&iSKIh223j%Obze?#>YvOAOqpnO0-qeb}+
z;)nj+hvb#7$*VsQ|FFqZ_5Vpi-@ev!eSvy|uDZkS@;CnQ<Gy*u8z;_;qetWA?-D)F
zb38iMQ@7{`l9wL!Td5D@^ljwl2kIs{Jcsfze%-eYyI4nG0`W`EMDo#V+TDEg=>>Yf
zgD?6Jcld+&<{$2}6F#ix8rtWx6MrKIJ)#TgnO^Xh{Gf+Fa*2QXDBTyw=%SZ-!usTr
zZ^@@VnNKeEa}I|t|DZ?lfe-p(H}mC7bghFg`IdgwXX}`czH!ic=J9)U)s05}`a4bX
zupfEI%O2LjuW{^PefpM<$;n^!_x3gVV&|{?G40>1JN@9}KRa&IFu#H789Dh4x#*ET
zJl8Mkr}Yc?W<UKVJCobB_%x5+tivwquRelY>`Jfb(kH#3OCIykcb}Z>sJ`=ebj;_M
z##<Ni8-A#-!EYO?v*yz?eyods@)*yb^>6xE{V)F5-#*5D@|uS}#9z1Qe3w1hS^wI5
z!zmA}-(gILAG06-kbn4RB74v)KFLom_OY)Jr}jy(B$xdGf7I{m*X<kZAEEK&rB{eg
z=sOm4*@OMbr{DH}hkd8Y-~67E^JL@o$-cjJt?#!Va{ho1Xy4@jgtCvnM-lJzjbHcq
zb0WWBPx(YXWUq`v<Dq%z$s6>@zwBSpQ-`hRT*NiMLmwUV%{Q-v`05tPohWbd8;GuW
zNR&UIJi<PZUJ~WUk)r&RNWY2n*@p6e3Dv!BQQd7r{*b7?>67?Lj?3$mU%yd8epkZM
zJooW6Qsjr|@GpH$^5Z_g;y3&h9jJb(SLvro-}7!!9)Ri~6jyDi9^g|xm)G2v|J5ON
zP+pXmbG~?1H;j|#y=U=vE6&s8e~1rAzcAyG9w5FTz94xad7*Wn`%qm=ly~em#kc*S
zb82y_FBVt6%Q2qcqUYS4AF2ngooA!x+(7@#f1JbWU-T{NA1w71TYH~h@qc_l^z{|=
z?(b{po&DIA|G3W|ApceW<pD^a^aAApdZ9o3(;vQOE?aZL`7aDq<gqUQfc(HZ&Nt;%
z&+(5gKWV-<ssH~EI^I9=3wgcKOZ|6(jUDRW55y<GqgVZdeuaPXZ-}0CGjFcEXxi~t
ztsGx@uI`{Oe@*$_AH8qK{l<0ld+xJ?{ER>SAilg`Q$O)XKTw>ozdX%8{EnTCryt|V
zi7tBV&)@h5Kf=HH<kUZ?bNnmiHV?nBxnIA3XLoY&8{@l0ak;Hfym}AgyxloEJ$a8J
zUy4(H1OM`pozMCHRZ}W{%FpPZ{Orcxm)<{f-7y#Sbohz+>K?zg52G*k<(Gc1NqzG7
zpS~kef31JyUAy*9dHX;|+(Ui>{oV1B(>{9gDbrtF`90d5mfbyk-HO*%P+sCU@(6w*
z|G+mr@JD)AKk<P-=)QfZ{iFSxarW!_0{brWj6;tc>aY3y*tPv1y5`}#TR){2cGuTs
zJkuBb*@w^{e}JX+jprZog=_10zauYr&aUp8ul^fH4s^WdcFo=}`IYCq$F@I|pX4R$
zh&Or4H9iyR8~-J=4m;61JCGY6?&HHcuEn4C?)oI_xF$Ec^x;~*fcPt6%7+fbC&Uj_
z$D!{q*#r7c-rt$X_v)8%_?9m`m#^5*b9#gP5FPrIuf5Mf58voHm$eQ$^1Sn7bj)|(
zH59M<0DW`Px4!<veDagacPZAxukrdYh)?vb%Wv=_zV#=@nTL-3f^qcYx$(xKgFoZR
z<GX15@}C^fUg(n>AN0mf?vpd`TMr$2U?+T;=ly|o(1-Yj^pSqYekH^Y)K}0O`CX%r
zf6u)SArJkr2fyM+?nCRlhSrl$tOL~<es7-j>4$#3zZ0kYf;{*W->%Vx^eK<<AM1He
zALc>pswc)lbl9K%<wg9t@0y-nTMu2R&gsXTAKOQq_rQl1JiM~EQWxQ1{X74@i_yR8
zH}o<37SHdW_`&O+|F0i6zwfDE$?@hn2Y3FVKXzWP?{VG@opbv;wv<c%<b9X^OTUI+
zeUyIAee)namXMs}(0{9U*$=3X{8~QIzpLx6_2v30^61;yg}$ITPyZIr>Mefwnf?@t
zOMDs+$*Z4cU%wl}KG6Qb|0!@@#{T@%zS#Q=|EGf<oEOlKb-n+QFUYT7fa()WJJBo5
z=cQ|YN*?Ef=&(CGi(_%0eP8O8UF;vpFMj#6IG{iNO;7Ig7k`(;9@asJJnZVbEBh4V
zAUWyZz7o=h{jfa2?~Q}__j_&5EyxXx<7f7%P~7la>(RUG#=q-xBlp<@9pm^Vy^#mr
za?Vw+zi(Af<@;rR&R_Wrd-%IGe(rZ<=*@k6@MH4R2Y$`txAqh4WWU5O<p<xHs;B&v
z->P%+g!*K@`se(@e#3Wp_?9otr#JWQYw%~E=RQ5S)>ow6@hjh{pY%}5^Do!?^v4g|
zkerZyM~mz+Qsh5vXrKCWL;BlR$iMidcvP?Le-ib(&cmJai97yI&-NGg6Z%E{u{_A0
z_z<_`m4Ecd{6~I~C**-d^;!Hu@}noOB;Vx2xBMgz%Ln*H4<F(LpZJkC=v|*rkNC46
zu&;4V4t!e|TF3hIh+p+wza`Jfi{wLJezzYYpM3AVn045lUep!qtFQD;Kh{NGT~?>a
zLq6lkqu-`i->pFFsE7FDhxA8od0kz$j{1jRbzXnYKK#e;y6T(MCD-UX&-ME^>Z|W{
z^;P;L^Ym$+vx{-|5BetkgTF_|zrVj=Z~CEs`x|!QulhCqL=JXlC;b!@$L#7F>gTPa
zUPE=5oxC4UKW2A!;iu{vd%ETi(77M~Dk1%Mi|mxhzr}~&V<{njPoy_?DItAB`h)C~
z=sCLlT0e&#`NTi{(YxpBf_20jyLiqn{I*TshpyWH{3Ub8Rel##e9Nc&nZNQQab-U&
zPs@Yk*LT{7$RqN&xY7@s#}CO*9_z>(zE@y>^y!oQ^n?%B_9^1pefseGoAQOY$G820
zeVcO&`R|jD{cz=qFHflaKLhXA{qDqVFJ1h|s(U9_{%?Wb*|I+ozuw1feC*kcAK7Q)
z5WajzYG1Q*$Ax!qK6Y#+ufJ*btyAXC7*oA?!&$RWduxBiZmZYrbj5}z`z!OE`}mzt
z-$C#*{-qD(zx>+!ApT1a{6?Kq2jnUF(D@a6v!CmdKj4Gks|)g@`SLn{@ZASKoQKel
z^D!tM<ZFHgt&bn~#VbAWuSDbIVJP3}6U7a?n8!cK!yoylI>*2GmAYphJ&8By+>ieF
zm+M4&moM;Pobye7u1=BzU+hj!^`4*USDfRj$DXq%ew_#L!?cS!&TsV*<e_)#$OHUU
zKbrZ7e&r8!9$)OOKDckZI?ccMh4uKG=k_K1MW0TO?7$D$-8KKryvPr&Pfoup$FAn3
z{_&^Ih#TlRKKY+|NFU}yc7pU-LUG)N)=#7ddBHmLVxMT=Mjm;R-;tM{jAL*6KYDhJ
z4t?-T`xXArZ}HDh*k7N>Ui?}<)^Dq~^0n_O<r(ie_^rGtZ^%2|8;C!CC;pv_@<Z~7
zFLFS9=~K+tmxwd@5C8Ir{3B1P8{~IB%8$u;%)N(fIA`6aA<x~HU#v$T@}K<X+PaXQ
zoJZ2*VEtZbpNsdrdd}N64B^{3z4K%1(y#u;_ptb-7jnwC^vrJJPM+f@P`sJPF8Fg_
zp0+MI=|Mk4KM#NJ$qVMM=&y`Z*U>Rg-{gGv`^O#e%X?<_H2-Ijf7*A8D}9W95Z~fS
z|7QPgy!g=%@l$!oJg9#8z6v_O7oX%~FLFbF&!X?suer7!{>XuU$X-w$GLIf0Iq@k!
zk(Yht33(8%8b0pPOHbN7JnIcpk2rB*Z^wIg-%omP@A|7x+&bZ(zPEY!z!O(|`l-G8
zI_CTRw53yCxBrz#Z5e*%@DEJ-(yGCV{`oKe(btMw`H=s)R`1nk{)L|TBSm(H^qHv6
zjuzFM5~?pHOnzMRCvh&%sx$nDp3LKS<o2At`H^eum~S4G|K%Tjf_2D+Zqh>+zvdao
zPxS%Dp=&=%4)Q~BYaQ`GKJmg&&C@^7FL})`A^zGBzavHab^VC<*v?yg7vlVm-|08?
zqxR|Y3OmZ9&Uu}S>u=bZ{@I&-(9?H8&&84agucA*+=pGs>3oPh#_5mHr+4(shwNQK
z>zbcv-4gQG5~5#1`YIv4m5`hzblx>uEUlaO*>!7?9@(FNLC>wD9;To0Q+_Mn@h#5f
z3;m00^}G$$FLh3TDSzk>^?%Nb^o7=y|Kurszk2ALN`GO$pbnz1ztT_3Tk5ZUf^`$=
zNnWIPe5l9z7pUKtPsArW^ltyfU+sg{0rcc2d5IkIk$gt)o|6YX$gcQA$9noYb;~$)
z1V8pG@;LhR!w&XQ*44N3TY8ljtcxDI(j)(1cm4}g&+;xmq!;f+(c!<&7s<<B#^FzX
z6kqs~U+mZLZ+%D~C8U>z?pe1E96h%)MBn;C_D0wKhMw8MzRkYOJaNW<_;n+H{dY9t
z$ok@&e)SFN1N+jGJSGp(tNZ)}icdIL?=wqiUkLdR`PiMD{G5HnJH3#L9oX48`gRT3
z1*UziXMWn(wR39rMBhB|Dn92en0L^kGY6{M{(jv%c3&{Q^8UoW&flMj4|Mc5P+wsj
z)HjGTdbU4Sf9Tz{dHjI=ohQ<-`{vWXzQ(zd@$x4Bkw5sYI;c*HOX&Ted5OMXC?S7u
zL;8gDlE|;sC;RH{&-~k2!>3NF)H8MXi~n}!;%oPuP>E;rpg0z<uGz0Smp<UsHLt(?
z*<pHTC;1ki^zWRHU5(Rci+g%!2lK4M4%SoujAsw=h@SaS-o_`taDIZ0e8YdNV;=hA
z%zJ+42l^EGNj(sM{1ji}R~{B`)+3j^DxW4gH<Z8S1Lu0;5W0^~<K;i{$wT6dzo{4U
z5WOWj7xey&obnz$%bWa!eb^m(KPyhm7hn9%{SrD4ASb)p|M5@q;LE!18)topj=sw{
z{>#4PN<^O?jQ5?Ae5Q`#kG%3UeK;pG-*|jM>zZ#I{>bk>`sP_*-zs0DCl2Hnbwj_%
z599&s>1*++Zx)~Oj=V08(8ou&nDyATxaIfk;=c9C2lW?Ff9P6XfZ|1dz_)#`b=kqX
zuGs<7oB8}l-SYm}ck=Rpb9egGXQ?mphwr$Z3(I@r-TM)Uu66VwgLVH#Z{#6AKY;Rq
z{A3(-U!2pc{6jA5$>&i1Q&-8WpV23JuD?nYfBG?bU4O&w_F4R^g#1Um@FVh~PrvAU
zE^a+f)DL8xQ76<n>(dLq_#;gEIZmAO7xfylkLTi%pYf}15nuFV9zWq9Str;d`JrF_
zD(?6XeeySStm8iU(5HX(n%<0~ckA#2dO??*=*o-ab$;Tz0sRL0`U~&#=u5v?vak1;
z*5!B5di2bn$mg0I^u;gup=*6fx5!S3^ymLe!=8^`y7+SoHw}BA<M$K4`jwyGK696z
zPX2zq_d@>9opWdB()Jh58Pt1p-M1ck`f~Kpk>Ay0eXVhRFU9)kpl=*{>V!VSbLZ*&
zU479<k;D4>2mQR~@{xE%4}J8|hw46h=BbPFKmRe$bNoQh$%!w1ioR=lmbcKsk2v+c
zfOAcL?Y_8HKg5~1Qg6lG)qDN(56<|)*eX%n+t)d_aE-n^A)f39)GOEWj``wO{E2tp
z3#f<A5v-4%_rmg?c!laH^j>^m>V~zK%p9zseXM!%n)ThM*JtP5zU-3QwhYN*@&!J9
zUx+{c#?I`>&iXw5D}S*&`tnca75;}le-nS=%{=jn5BJ4~=ll^}{>rcUhkP$TxS#b)
zA7@?9#gqC$FZiN=cCjDQ|I4f5oj>85T;j?5Jbokp;6t1mpZcR0{6ci(WpY~2bA6_K
z2-%VR(D?>`AQ%4Vh27|lJ;Xgf6R-3J<q!HckN<hD?=VjOF-{)9m*@NhvcEiFK0lWS
z%olHv-t}#nU*rkjwc;1b3(hy(rw{zOPaY^9*n_?Bg^qaO5BSG7f7J)OPkw$%@8}r^
z@dfFbK8<q?(L>*UNuA&?=F9u`Tl7btzEko(!*|vnoba|suit)RMW4n)a`N+x2mbDU
zv+u8ehmGI2Va7Rou_Hd%OWoH`THkZ;o#==DJ(rinqx^yo{gP{a2!5;&jcai8+}Y<o
z@#v7a<6k8dC*sOE2Y<JpWq1Ao@j*`LzHyKqVA{vL5~2ge6>RiS|Gpmoajr<8CBzrx
z4~fp*&=>#Kw+|<uI;9@Vb5OqE7yN{OSYQ7je;ZF9(0%jK7dOsj)GvJC*L-?%pPyJq
zUXic(sd?x?{P0ifx;D-=dgLd!Yx;EGy7G|w@(jNs2frZ?dh`$FIs7<h)&E%6eAoQL
zIM@7zypVq674gKcjB}o8ygVgt@$I>J)`#dC2i;F3zx(LseRS{z&GQ_Z&(B<=k6(IC
zJ-gqA;@Wo~_7UoizC*rC^xXL*zc3#f=l#1p3+;QX59Mibp$>bFZle8?@z#N!<3~Mq
zKhbl1dY*iVfANZc>(Zb8fE@Nu{*KK4%Q@{Q=Ty6mn?6wazGm9|=XQT~(U?jfVn2m0
zw61)lPh~IPK|3$i&)YxnGy5<8#2@7kc~oA_cs5=?2k|E!_>cSY7PL>*zxrJy`%!VA
zzTn?J$Nmw2zVnt}#H0P8{fP59c6YwuoIro$y(D=&_gxNs>Ce2k@_ShL*Y`m3_&X7G
z%6X4@>?FUM=ehg%r3c7<;)dPiCHjH#E4f^IUu0Yf>9>T|FJX!9Nbif|M0$~T#f5SF
zOT6P#Jro!8j}E(&+kNBdFYmi%NA@yaeTVv^5|Yn8oWIEf_T>;AerG@L|Eu#ic|!ey
z{1e~q8!vx~AAODfL>}Y6#_J2jr~R9KI@I@g&d<^J93A-?k~>k}(m%=b^vWJE=ey>&
z>_G4AA<vH#=~MpWXZ+Utx5lsQ@7nm8{K#IAUyB3&kA5P(>I+J!P9@T_`t|gkH+}s1
zQ~N7)=tVtM-_=d_RhQKj^~Q7b^hf#&^!d$T{r)uT89iEu{*99-A%E1Tic5JH%CpUT
zo_c?eK01)!TNkoBJJJ*X6j$g%apLd5AidEG`NgYvfb{y(^3{uOdva2R56|cIeEZ6^
zd-gToQL&SLh<s3dvKu<`2m18KPw5}JFQ2%Ucg3eTH;#Ucr)T5Xl^w+seWFYM$*=L`
zag83-f4CNx=$R+q;@^1un`e9peGkB2$cr9++t4|#@5A`NI*NaM$vepxJwo%m=Xc*V
z)Q_p}=J6NuS&u)kGxQye=j@4ZekQM4r=c7FpPpaQ2Y%2oAAi<Oeapl0ruD^%cyul#
z&qDhieYbdY4lB;(0r^%w@Lc{x&wTRf3w+PyzWoXL+0l6%{qZ~cpnoXuIp6TTk2v)G
zoH+8Gjygm>c}9OP&-y(KdZRCL(gVKf2fy+e`N&0XNDlMZk6ifSm&WUx@I!9<NOcfB
z{CQ6R(S!VyKk8Tb8UKUw4-_A+`GN8Lz%@O{tG=g@KRg!)#^G0<bB%v>%XuaJ@*nvU
zzve;fIQN!U_z(Md4{G1)dk1yQ{vFc0xDiMGuAIG%V_$WhU+Wv#-FGGYj{N+9-1<}c
zfbtFb?Mu<&$Mi)%@|SV^#&dQy&ihg0#hHDNb)4hM1L*Qc`WIK^;}7ghF4xfhg<ta@
zae}_QD*o|@FaH<By6Ajh=)w2>!MQtB#?yy!;sAP1PH{s%b&|dDg%9J66F20L$IZtF
z{qO_!=WqN`zR~v^FD|V|pX^OO{j>Y{XAktnLn1!;AHASw9sJ9e_OblKxr^U}_r3!?
z<Jir8{eb%A`$PSX{z=}F|M(fZxNhPi{-16mXT5LJ*Q9^&GkT+UDDTrd<lpQkKhPsY
zpMNDic1OpxxaKENo#ZF>ulB3_K;MF2c{IP1l^>ye$o`ES-j~(-6vyl=zmiY<`A&#j
z{0q{%c#{9=)4o-ok!XGN^LbvA2Yu)K;*(tDX9sqK>_xx&eB<Q{`Y_KpeDFtcLNDUQ
zbC}QRSzPdY{sGY=muu^~Zyo%5PQTu3)4O@(k}vp?b6)k7p81<|1pDR^>M!Y?{;Wq%
z&+%uTxbU2v*w1tNBNw^(5&s~s{L8-L)Hr#<dibSZ>${I%bpv1G#5#Fjf32UUXY1fY
zoO*Ado>?D1>|&n#{24#i5x<arJa<iB^hq!Bn*9ef9$mk`<M(>#OMb(z_)b1u^GAqJ
za+?qNv-!q*4z15W@lAi`^Aq-BKWM(+brkpN2)|?xaSM0()ce;Qdep?u=cgZW^}8>b
z+%Yck`e~0XeCO_yJCI+E7TGzG{3X<Pl~7!kkX{a$_P_z>UG&T5|I_PR^t;Ya=+k>J
zeY5;PPxdRW?W>^vMV{2(;75NbZ;D@j&(GCw@9nKCzw4Kr$H;fCJvU$e!w0$Wsjs33
ze#7r}Ui|76Ke?>0LqGaG@8z81L+A3oEA)RVoZA~`Kjb?`c7gdm#XigLlv>B{9_a)9
zU72-_^Ls1k;9tFTA3gk$+j}Ye8P5;1Z{=^s(X(rEkQ0CS;s3^4$8+-f|2zCP*CoI8
zyjQcX`tLqJK;L<j_3>rCb@j>i6YQXmhxFii3CU$&!SC#2#V<bobV=WZC-l5Hj1RwS
zHFvwG_g;HgZ|6OS{Q3JIzp=ODcjf(lmcOI$z7Ed4XS>Hfea*(^?<@ELJ$euB{hq&L
z@SI<#J=|wU&;31x-|0fv-(BzndSM@a$}jD|=!M?Iqv!15JwN}mj%)Ps$$t11&+J0K
zp8Go#`b&E3i#~e%)_vFhzkqy7PWq%5b+uH#^h4@!)+u>8^R&E(u6lrueW*UleAn`?
zzEU1`jjn6-<!^G!>rfnfuKy6X=;;^G!<W8=9;~ar80Wrm`ZxWGy6*jzagcxUSAHgM
z@GtSLuCg;ENA7#+w}kBByA^cQS@Y<}`5L+Pr|6hZzwD<D`%YHhXq@%<6XbX7YQFx@
zHNEHy<Y|2K1M;w=b@0vq)O~h#jbD73rw*WNJ@oM98eilxUcPn>(dXB$%@=3ZN&40$
z2Xr5uM9-o1Aw82DAGtn1b|0VU&?9+`m%q?A4;nA7-N#2_iH>pfB7f5d`L`CuL9<@{
zf2imI|K#Cs?n7}!e)q|T??m#+FZxjVM?T_T>Y_a3nxE>E^n=Es2h}4epP7#zc8AV8
zp?UPD&y){Re)|CWf!4#fzCixfhw10=LvHp#$F+Tq{RaEEPp{~b-+LtY<u!84kLZXy
zal<ds=V#WZujHTK8E3xzm;RT0=+JxeLvH-Ai##h&^J{g(ee09YwRQ2qPLQAJH{}ie
zl)g^Cgdh0`zt%%9$I-X^&)>}7S`?qnI`wxu_`**cCST~1OC4l?^;{oF4$q-|g8p9o
zh<p7k{p+tY?%5xo{_a!0!Y@9_tDkaSr_XeL@7jAzeVq4~`VRcb58}o8{90VFH@@?|
zN%|4Fj1w2+qEG&89q~f{N!K;{;?sU2^O14ri!1SDe#1}wT?su{4?o7apYpIXf8}rV
z?YaBv4Sw)x9sHq(4|35TzmuoYlkcqWIpnv{IM?RWt9kgdKE6G7-?*gXIX=y!AM2A#
z-Lej(ADH*ix1Q&&%|jomv!0vhIl0(_zo2UzG@qRG4EaIsZ~fkaeK2$mKtJ-4^9tv5
z3r;`wx`A0^J5SuR%N|D_Jn7dEefx9s{a(0p59epj4Xp3n1Imy7&lhyQrXK0LoL}gx
z#D#k0nw-u{=!xF=jk@7{fPSs#JkN7-d7mKvu&?;mm#{DX<bQr<p7@0Fg*>jm;OF|S
z68aqk-=(QnP~Flep)cRt->MJL_cr$J-g~GU>WaGKy@m4xeYo}2D`=ljZhaj+st@+n
ziQ<rbp?bv*;z~TS7r&ur{iJ@EKico8*Xp7^PyV%DqWI<y)|EH;xB7_>e(c(O>+x@X
zDE?9});CU_lb7T%{O}v^ADo-0^YqA0{8Bz65B{vr54_*yKd$+azL)<P=X)^U6*_<5
zzxeiE1fnPY>4D#g3;HLgxa6n!NjmNuN1y!MI?frbBd+8Z@`yij(~IZy=)HHw9scP<
zoLh%I&BLdBL_U3@bF8FieeX40ql<6$f&7=h@mKyWUey_SS^ZEiTszM<p5Lht_Ce^1
zE92y6@hz_L<-YoYfB8l{8L!_|UyRQ@BTnQ!<IzP=eG+HliJ#dYlPmK@)*b!3`_{#u
zYv?||bKiV^@4Y&C*<YT4_L=q_-p|{&@FVZ%eb=eJ8JF+D=|kM}b9Kw_>SWy6e~1(L
zfV_Uc$9m)tkK|Vu$t^C-^Sw9!U?=As>XN*w&d9gU0bGk${PPR>MxOIr{-j_0(GR`)
z&W4`Ub11%D%U9$_7k~1dJd@}-e;O&u8|0H0#Jl+Roc&<tWqIFx`sVlin|;*#%qQ%i
zE?SRYiEr_4J$(;*%46~+Kar2f?>T*<=e}$C#5MWkcYebU@xh+ZIQp_KIWs=^llo-5
zdIQnRaqgpUJUv^FzRAPB`T%s2zWf0jIqP$1`bb0%AN0kK(c?et&5zi@e#<z0qUX6!
zMhBY5-{l#6SdV_vFXVZMj&=DVG>$%uQ-|=6Kk?x{`tF+#@tLSTTbF+58Qny6I`trq
z=*fDn@$36H^XU)2?C!q!FkYM)$B+Fzf^p{4v-|uJU*vPm5BUK)5TE!Ur{|D8Tw8~}
z@d4@GbL-eQq2s=B=&~nytxs<I8g&&tbga+MT$7I-@GDNCegr-G<S*=ne|ca1ww^wb
zU7Z)Gzw({@E|0TEB0kL5x1oa%dXo=*XQq$Qztbl_v2P)tdHNapp*QwYzpZCJxygZV
z<Ir>dAV1?%AFEE|hyCcwwe{GUobtLp#kKPc^YLMyOCQGL1OLty=->M6OdfpO?~u!S
zuK5qX-FMDGFZ9SC>=)SIef~#2{Ncwq_t7VxdO=Qhr62sUKY!r=`cr*0zU60p^JD#w
zxME*^YJK`J&VA$fv-o8n{$^dz%|ox5U!U*u|Iwm+(k;q&-p@H_b^d9;Yags0LHgAP
zvIo1u#=qnD?Dz|P@Dp(*KYI@8Q+{9<c^DeU@8}<2@&&uW>@V;mj>M_?{1QLnpMBLu
z{FukD(X*eSFLlR#*ZO#J8BZ?jy6@V$`XqY7H^eW*H~r{$(bM1JPd+kE|81W}UU5sm
z5MB9=J@^~H^na**kLWu-eU(0dp2%U}>%9g4RnO!H&#f=6`5Sup=D+l=UyvX1<J?Fd
zbMDJ7zE6~w(087MfAo#RH$O+89P|XekMVz2>CJf-K2G2AuRF~e*V{S$z_Vv9zhld=
zzpHjG;{WjKC;5T=?^-@(SNb$g{&1cnKI92{lo!MubZ)8s@jt&u0gV%<;+mi6lf;d>
zgg<t7E=K?C!B1WD7xVd<=j<*YtA|iNcFm8V`TPiaF29R+dDQy+i@#jE-Ma7Ga_(z7
z&^)L=l^?9*n*X8iIX;aOch)B-yF2I6Z}@u`$ZqllJ2;1-PjZP*^!Xim_@_Ll|KP92
z<4eBfpYovm2hrn?{FPs*r}7N{F%I%O>sTMsgZcDto@;uLr_9H%`}$DV>JB}5e`1{a
zA-=@DdGx6cc%S0^0KKYr`XKks<6rbjZtFUSrQeKW_U0e*9lcst9v~-v{k@#?M*jzf
zU$!AR{BDEqlAVV-Cx!G+PJHth@#OCbAi3#@pVE)NclDfqh;w@P9$&tM?2Qk8#{c}E
zHRGL&>j!e4IEDCEC*%qI;?sS3MxOP5#r=J;zQelohF|wx(-*q#L-u!0?|Uil@%g=Y
z6KD8yzF_=}_fGim6}RtDsYB%ApRS$j^F#l?)!$*F=UiRA;dg!y)9;^jyI&}c;|GcO
zrRQyh>;&ZlNZ&B|qF<QL*-75*7TJS;57xgg#_u8fc@K5(vWKsE*P<7O#J{*xFT_7S
z?Azrj@!+{UYacAH%d6y*$K982(2)<tA$`f8`cQh-&)eTS_YimJ7;hc?sITtpyT~se
zqlZ7&<fd2eljKe7(}(@J^ALH)`p|gS^veIBe1TtfVJC9pJNsqNy*Dr)|4{yrZ>?{>
z{0Z^pdno>nj(*TM?}yQqKg`pwh&O!Tk3EupDSwpit4D3<djWAH@A4;du^WD&dJ5%t
z`AVIkC)e^@-baW2T+;{4=khlFTHkzjFi%}_&E61y^uUkV$8-M4-}tZb{K|d&izmph
z%@a@Nd2W4q0DWk`YQH1y$!Q;K|DZpye>LBCbnbiqNN@B?@4maTE=14!R&|@5j6+ZS
z@DtbarTj^c@(`pK&)Jh6b3Xl;k6-&O`)2E-E1$EI{08l#`G<Xwc#`MUHK^ZXPjwAF
zevtbE@oT<yq5j)`LA}z4sk`bX`slh=@6gc~;Zq#Yi@Zi({Jn|${>QJo<Uj5lGEe>?
zhw}{d#GQPKu6#tl=HrhX5dVqx6X<%*uGX=ymRI<%ICEa^yv6wvKa-dFE4lcWbv!2*
zee!E`jdy+_Ud55|;>~k<HBNksd-34D_4Qr&Wlwy#Pfqm7qyFVJe|4YS_M?8sME-!*
zk@wXhd_Z*gS;}v|e1RWwlhZu;7+wBtym9=|y6kP7`R<!XPww+W{%JnG$xSc((tUp5
znmqLFxpnY`e{pBLb@)B`*^M0JmT&aCp0l6z>0du8ZuH&GtISi+#3jBxN7r}pzJphn
z_`7p(d7u8p6~16&hx-5RH1)gwoih1>e&`#2;=*(KG!C*qWN&)LC%dqdbzIA{ZJ2!}
ze;~K(CO&HUUGtYj`IO!Hr+k$9kWc8rbMw?``N@5KgL&xiJ4i2{)1!IpB5zvXc>1SL
z{*O-PPxI|3@UOpCFZe6J<KO&}yzGr1eG0$!z62fTyWRuYztI=}!l(FgzN~-r-IMnk
z=%a@YIm9J8_+$rtnY!uR9X;dKZ~4n}e98|gzxd{7^yfT}-K>jmd4@kg^yMdfv!D0r
z_~H)`efisb`NTPez7d+|99}%qKmTAq{Q)`fho1bve>|se{%>7&bj{A}r_RcQ{L^^x
zLSFKtV?KV&V{h+`!~^~k*~fkPlAqIu_2~h>uJJ=|=PK$^-Zzgto|9YOPe17KKjZMF
z?^Zuuvpf0Br!Rhlo;<_<_@#Y@{|AeX@3`ePeGa+E$<Mq`vYt4_2b2$?`RX3Kx#qw8
zQT*|Da?lt1(K|n3ck?0pnrA(HhzEN0od4QCTHpOd`N#VZ`Vcqj7JWc`sz3C?p7OZ+
z_T%R9bNNmju%CD$hx#oqy0-7+x6Vb#K~M6#^F-t1W&T0m{EGhgA3oWSe$rn2AJQkj
z)OXkV4*Ab|)<>WI)ff54eEwj(`Si;E<gkuw^XS7hWIyp~zIe0_KEx&Z_>o8X2l`MS
z0`q%?-WyJNVEqnbI!Da-#L08k_crLgO8$<O^CZ8=RjNk~pZ-r}{kyZ|AwSd~L3ZW0
z<Rmx$;+O0wkH`o5bL*Oio;*U|NuS=?$Mf_zeuIww*?o3(pI^xH;)Fk_2hMB6rTgjx
zI^q{S=PKTFp@VPw<tO+<H|4~a_0YpN{^(Ua$uIIa{>YK%Qt0ZBp!#FKVO@4{t^UcU
z`cHN@9v|%Gy%RlpU#<^;<g^aGK>h-)5BWQk*P#25T_F2F`>ov1y>dM|4e&>P;CzvP
zIUf~2{EB_}leiZz{1!duko*pcTPP0stvtp*oXe~K`Udq>pX1s*_xZ8s@}>AyC;2D8
zhUh`{Tt1ec#lL>VwLEP+`Q&kQ)i-^X`RWgT@ozl7#h-kSu4{a<D?Z(4XLeAR(mwhE
z^U>3{kc*u7*Vp0KzCoWwUV2iW)phxfAK;5Wu(!JGoPd2jH(&p5y!-ew&+kgGi+TLX
z`sBkWRFBjRc}+hl9_f)j)B|y8p7$v1?|UQT*@b?c2k2|Wsqc^ML-ZxS_eKZ*_#-#{
z;zOKS&$$Y_s(a2)ojdryVE9t6`47MJeo5X+l(+D~p7Iku64goa`922XgFeL}Kj5E!
z&)K<>?`nOg?3zD$k4TTM*~5KuunRwv7v%%~=6uUO#(jB<AJP+f*p>ar2R)b1#ECp3
zUXm{U$R*zNiI85<lb6u7K7OrZ9KV6&v_CYS{pELdmA}xl4$OJ_CH^HY#j}1%{PI`%
z)xOd^eWYvk)_pj-s{h+FOMk*|)CqQSe#Nf(botIa{%D?lMIM&-<PZLd9;A7GDIUa;
zdHAtDy7C`%pPcCOcYaN7c};#7x7HIEuB{^;)MNS)FV?ZHeh?Z*fB2D4`9He&6o2ZD
zICI_s={3<hiPr5F*(ni!@)morgMP>Q@{fK>UhrIg=a=+He&<;FIQDVg$$sq0pRA`p
z(U<W%h@QTK-N++f<+&k#(1r3O|HKzP@+<ucI_%1?tPAx^-W#wZzwv)I$ictZN8jQ9
z^!Qy2^zBa|`Q>4B<zaTgx4ghV=!+fLm7eUATsyC3PxeKJ9QN(j^&PK%Kp&Jy&n0B1
zM0rqOqmT3+Dv^Jy^YRKgq5Dvt;XnM-zJ|Zhqr4)|(vR<g^po_cF4Oy9{e7<Qe#n6?
zM5hh)w~73@4f$CclCur<9V12lrVo?f^^x)k`^sbD(Rlkrb&_A9%g^bBU$VdWlSj<c
zk6NGqSQnqx)pzE7^U;@|(7^{j6XhNL?wUT>k$ezc>p*p%p5$>z&P3y_i%;^|x0*+f
zP&}cdUg6jJ>`PCH_9gZi@`iljJ-$3;J#oig5PkNBnLo%$uQ2yp-V5Nze#_rO;)5Kn
z)pPuKU*$U-=M?0nSA5!!Ie!p8@;5!;TR)WkYQBA4_C2oItqs{d(ch7@q54`vago@T
zx762eG4ri`R=1e^m9F(ku#{Kz6WyY|A+f|q-dCrf_%C7FJFn3xVM*?8@8h4p^Jn=*
z-)o#cTb-9z<vV#z9_RPyyY^hZB_F;^dQgYNqx>Z=i1*PV`s9&k<Wc;Q!}}Tco&U)r
z_~57XM<4ux9_gXEu77t5ALMoo<!N?PCp?GrEw8y&?-JEf_F`YiFYqfb%Gdfd=bQRq
z{sVpABu~gU`h0xY578ev=vm$!En1)7%e(A{56DkE=NIO?PtWva-=*%@m#{x1hy9Yg
zBtGbg|H;4lXwUhne%pQvU3M@|9N?F{ke|EeZ}h<);s8DMhF;L8A9V2HIsf2i=+Kkz
zgO2{a`M>z1851jhMn2>D4}GCWpW@Sf`lEOI7SH($KJjBeYJL2a=+US7=&?WkN+|wI
zC@(<$8>Ih4?+x*nNRL@x>8B0(qw_p|0{uN1KZE4R^9Oz-ukthRtsy&z7yd!-sb|k~
zJUaZqc?CU@%X28+ptzwg*Zcum-!*>eF%jSVn!TX!HLOd2=;N=1^0WOVeV35@P&~1V
z?=RSyUy;*1a=MmpjhDB{kw`Dz7ulD2Z)D%0ZrYa^?|qOu=Y5d<k8Ayjyykt5enno9
zCyhtPyo3Ml%`Xn${+ep_x}C1r@MM3*PWlr0hCcc47l$5sv2V@T$~l|Av+=&q??o67
z#g%-(&#mjbL;fhw`M>ex@^{$kk#jZn;eY%KU-}00_=WQ}_V@R>^1pTO*!PB`=1=|Q
zklzC^4t?iz=J{O$b<z2zIQIS#zv!vo^g?gq&3*C3KgBmWtwT@NVNa+YK<CxzFZ_%@
z`>xLUi+wNuGGE=+$Jv+jSNEL<y5_&`i+_FN|FU=H@m~%5AOC*)R@x&=h%Avpc4a>w
zUBoZsx^!i^QK{_2B~%n$*|#D$%f*%KCDB6XqjGIgRNB)*rS;qXt)$=MK6rMoagTFm
z&dfP8pYQqS`})qAnfLoW?|ILA_W7`$$heVl8{;LulMlI&lXeI__|cB(FYp8Cg>fSG
ziJcLqUKn3c5A@gU|DiYZB(@vY7nnaGpE#aIpXiBnl$-Ss<iZc3584fSXMKouZtM}d
z>|@YgXvef8_H$?#=!f<VJ?xh86Le{p>}TP(vF~kuS9m`cdgLP=exCZKpTbV4PwW81
zZ=pxhArIpL{8<o^4tvF}NKgA<U6yeO<9X6iAJj8p@>4(Lqdm|Lw(0K|y-`o-B@CG-
zh9P!KIj~piop|J={s~i$=!5#BKVp19KTZF`J^=d(=o!1@y$BG0CnA0vd|=v@x6d9{
zoQsRJzaXSv!7m4)C<pHcgS;10x!$IEPqnU+gYLvDM>l;TS2fq9!<gDvw<^i;`!4kV
zj5B*I8}LZm(&8NXIluGH(^HTCbLHY(+0egrxaru6xij9Y*tPZ%+vR4@T>MbiH!J4g
z`@o-S*Xmuf^0x0cKmB$8n?Kk-H|(*!FB>?l>TmFl`QHtD^SNr<^7C#6{D#%-_T^8P
zRLTV*<zZaRxMWO8%MslgRN6+D_j`~7eIPgQDv5YUr(A>WPOrReyhFc^zJri{oAp!t
z8}>~*rrl!Kv={Q>r?F@JGyVua1wGcIp-VpcYuXup4ulW>1p16eSr_MiGxY<~{ut-b
zF6jsG$JjsV;E#VIp7ival#_mh@8}bM1LDunE8|MWbI`%BAV2$%jI#*Sk0CeZVm`_H
z1<1*GjdC!qzz@<7F-{=NI0E_TU!jLxQEtMFJE2Fpkr(=upYwK<3;odFL6>^L4`V;{
zpNuydAHxs5Qf~S&<YK&oe(~p&lkcqaF@B|<_)eI40e+kGBJ#c^{(^dB{7rrFz7cYO
z=!yD94!$$*MBmgOblJbAoTNiPQBmx-vhc4pfZxNvVCUE^c1=9~l5zgF`uXKuW9$RJ
z$hZZ>Z!tcEK6F9a1%3c|$j5wu`4;nW{4oBF@g(bQv`-K@k%M-Pmf%Bv{5$g_P+b3G
zT!x)8pJ6^gd&GY-o~PXOgN#e*x9IO!ufWgXPeIaCKK5nlH^p@neh-lMf_Mj;c@gEO
z-YGBrAoD-;&OC|!l70rgux^DMVThe!2h3mazr^D=(I@^0|3ZI(o#RjOzoZxa6@0K0
z<^|Xbi2kv2`d|3b|6qTtv%sHm9r9xDj8E~O%$Hd=p&Za>{6RlPJod)88G7gs`=Ect
z|3e3U&_Qne1AdzI6vmg(Lq6m|KgdJ-ARhVXKZB5VN_(SSVz11nK+=o$NqfV-(Fb-%
zKE@T;JARaOv<JRpH|U4@7ZHBCk#d5Z+n~JY3;#>I1}O(&?2_`Mcl-@>2{RvNK8`<t
zF8&r|Jpj9)eevG*HhcS}-$5S675EwY1<FPFsYmD_C+khDOQRR!p-VaFUkIZQ{2Ts2
z{7$(^$GI=)pg-Crev5G{@{k|>BM<dNKY%}Gos4#Yo%5ak4WvIs9_V7vjMr$_=%4jw
z{2567fy@h;A2F{E3PT5EpN#c6{33||Vq5}}4;0f84ny+8pY-B)!pMajgoziw6OUel
z;>C1geUKl-pOX&1OBi|}<A2sa=wGM@@`K1p7&_Pme2@cV9zwh4JARUU(4&2jj(Q+W
zKZG32r;wX|iF!kB@<&DHImk~xg}%`n{Tclz<pPnH^%>R&un**?|3MER<!64!xPx*q
z?qoblf5dnNKSeu7zof$tksd#Uf5s0%2R}}~2wnO|#!>KLJb>Kj1;pR6UdcEFME}e;
z;D`UiAJV_#*YLaa3#6ytz^~%}MZZaU%FDPMJ&5BD_@fsPy)v#Mj2t2#*0141zk-~E
zkstZ^jy_p0V_v~{pMHV<n|>R;a()IopkKf~>HjDP^-6s+&ZFLGkK}{jHu}H3GXWjs
zKtGHF=$D{F82d#3)IZ2~_<_g?G9RGbfsE4`#{?neg&v5U*gJY+oFa?(6Z{bE7`;$W
z%s1didnO*ek{|g<hurYPpJETROZ+<H82mE+0E8c5=z!Q0>Cg-EP;b~da)ac@9;kov
z(VpOsey}_I4fRGlKo87A36qZbCHe!QM|#>d>j0Dsy2wL+!hDx?Mb-(RN0@mtdPHvM
z<B!=NqWtI$yP-U+kAn11q(?5olovVibI?OS$b(-b9sZkohd$#1`ZekoI_MESV}G<`
z_)`w#!~T&AebB$4FTSH^kov+tNiWt5<A3Z9KMh^vgfDgr9m+*{Nl&@y*XS4E2Vxhr
zf69;Eu{ZdkKk5^?un)>j{m^cp!+3;sUhzBiKz(B8<YRnA7&@e5yh?w9pQ-Z8eJAoU
z9%9_b`aJ0w@30?=--a&ZIOZ$(y&%M%s2B8yebDc6?t}3h^DxF8^vCqy;yjggQGUOQ
zdZ#}nJ>vrSVkhunJOH9E>KXe$?|g?J`)rJ_C>Qj(p8|j8rQEaQ{1@+5Avg1Q?45kf
z+X++8*fsqD{(^BN`p4hEhw&El$WK2(IhZeVZUj9dCvp)F9sDfg81^5r3+Ph+=oPwz
z=`R?+FrGvIl!J9%<iSoD=W$L3IaoKQe%c;T@09l4iu34)bz;f`Kkfs8=t(@shdk6b
z@A>j>2KCRnG;-H^?&Pu8&8(0|&U4=_UHU++%K3ihZthuONyYr7FZkb7n_MZ6KF}BE
z1EE8CLHO}r#nn45Ib`0+#d*T4n^BLfTSK37U&QmB@)HL6j$GtJ9>%qdPpA*%Cm;2J
zol+0vBOZOw&ur^ohM&{qyFV`Skq7&w-O|peXWBje0myh4IYG(;(m#XS`m<lo-B50j
zaSBL#0kKaIKL<hwWSt3Qp95qa5PhIO<cAJ^5u}}xp8O#4faq03^e-a&VdTdiLF@`U
zpuX^L)I0IWOFe`53;K2RialXp^xODN{1xLg%1wIo2Y=|okM!7SRAfAhT|yVXM!y3e
z<fY!wCvsB%_($p+{o`*bC-Zy!F8QI4|3WVKus)4^(52kyiT<8^)C1!@=<uC-rCj7k
zAJC=$KyLUW5B#tz^hOx_#I6`8GyX$gj0@=3kemHZ_@YnxSNccB=cFeLUHW(Y339Lw
z12P@}u@}atjPJk?`wVP2<@2H(<3QFu8UKNd>zL;-&qdGF2lE#E40^<0aBhlm9sUhO
zZpw-Nsdwf9@FC26f_g$9&|y3ef5NO=;y=(g>yyZX{u!53e%2wG|Fe$CxS4$D5xvu%
z8CSxGdM6&eW0%w?<)U9A9ph>8Q!esh2lR{R4?hZ`U*v-id|7XyJf!10VfsDzz#si#
zPmEUyBR~BH{UYs!_D=a&mjTfSdL<q8NBe{?eqKc6Vf_n)F7qhLjh>-HIZ2P5Ko_~t
zEA>o0pg;5hQa{)i^w1alM7gMU^gujzNBvM=AoM67`bBQq4R%iXkqdi(KXjNs6AwQS
zd59-J?F__!ksic8kt+z{2O>9qh40ACcl<bhAH-g%Pmua1AAGPQ^h`NH_~PH713%Ug
zs7L66_&d(&g2amm9rTDC*dO+TenISsFm`~QP!15iQf};-`i4L8e1{K+{(=yDB|Txv
z#rg>Cj&?-7QC`|V>ETBhKS{qxdk671)Engy5xO9L4HVN8Mm`bY3sN4C`T(gn5PJaG
z-xiT_q7RUE3bKzNBIn8(hXf(x3dRSF^BBj{PVk?MA80q!Gv67HLXY+-(t#dyf)KkR
zKXwg5pZ3Ie#zB+=JH_9F(8qs@<5=WkzCk(B8{;qRn{h09K|b=KAI970iFFO^opB!h
zGyFl;<7g+yNqZw5d>BU}2lESY-35IS&peE<xb8rhc-C#8OMDoj&mcq(VMx6OA>(Th
z{YORotccWq5MmEOi2cSz`cv$T`X(&43*<p=+CTM<p0Fe4d(;!_OQaXqnaGblqi5tH
zjJ;x~q=!BTJ<`)&K<tcpHTDd`4?7|s@zA9`;m0X2=~&N1ANVtna)7iC=upqJSH=zW
zXVfEdpfBo)b`M>~1<Yr$f5s2=YmD0%PxGDrh<;=9{1uz8JAAXpxM9>QpDmpK+#Wf`
z`;?P<hcEt@bq@B2Sa-rcn7=c>L@&@s&(st0G5@E2@z>ZjdLxYgg%5h;JL$1E%E35~
zaRT|MPs)S;M-TA94p}Fo{jk1+JtGJ5VxP2I`aSdmqJQ{cAJ`G`&=nE8$9`y6{2m_P
zu~XLTSU(~jzN}xNPxQcd%1!x@n{|2QWq*-&f&F6#)H{4wkD{Go&)6O7OZYk3H~K)o
z&<9D+JehXMJ}Bb`%1?hxy%DB<*~ep@6#K!RSZ{<La-bLVkNoJDeF)kG@z^tA^uYWQ
z{ZKyI9py%^v{UMd`omt3llnwX_B)s-Fz=vV(I53j`x3w7r|?_YH{)&U1wY5QTpYKH
zVbYsI&JBY2QIPRI<9Os_oX@=i<f9$nN4EKa+PM!K9rO0K-_ynZXy2sA4hYjO2@}sa
z6hBS9igd99+9BiKZT-_P`!m=neh@pQJhTVe3-%6O>Ki+P4|YlzddNllCLMkZ`@#+>
zKkN6ryNF*v4(MQ~l#_JuB|Z8gJqTa=jUdEt(2s-Uhc99D1|lDbK0wODyoml4q#p+<
zA9@BwI`|#x9ebw!SYJB)@dID()MArIJmtY3aQ+2<&N&$|p7xJC$cbF&hcN8|Ik5-E
z^YDWY<0||O^LP-x$U!?skBlG4j~=K`(!mcq#g1r~*c17X3x6W!r(Doy9iH-$4?lu@
zAbtnL4}+wm+#qs;*dvJCAoUJnuORjZqIZyVAoYx0gXj(YQZBxu4-mgYyG4G!Q}5^(
z`OpvYVGq<hVb-Hqmq8v`q~0k9{Ve)GUiur-BR65}0{x%|^d}<a=U!S6Qa%xpGYHX7
z5Mtj!i2PBJ`VtYp5QNlc5Q=&ujDP1mSP%x$<vZ(yK}h|Fp}2lX7<&e><DmNH`>ab(
z`Eg?Fq8xS>hLk5RV$WcZ{e*qTKglBdpkPqFMExEnC+(K^uXrzn`%(i3SNY@gTyYM&
z#!nH?zBTnmz3;bgQL9I%|FmuX4<FJ~PcI+z#)3QhSITid=Fyw3zqZTi+vh}kLoUX9
zoO{H-vi^#nrM<z2dcZ&Oop)ETPxvA)?Vt3d1L>bh2hzVGAN?ErNymJHb^#ya`Of?W
zI`}=#6*9h|o|va$$M`w;^BsQ@72%Ivp=bDl^rOu4K<0ZO>og#Kjd?8NP5O8G4Ulqy
zqMx8XMMS^wBR@z#28#L7FNj`1^a5f}%!imCv97`SHIQ`7gPG?u?<bz`%$uRhe1vr(
zkn!t3M_jx0fbFX0S$Cp7h$l=wkac<ZvX04k7=1z)B%XXCAL<(<Ul3wvAnjX3`~(=(
zPvUPyBp>ZkMBabGPC(9`fV^uIgv?_={30m&d;BcOIsxzBaQ_j6F6r@4v}4|D0U2+x
zz5pE&k&E;o>%bs-2tw{PitP-3$PZEvoFh8^oJTkPuv1Bncch31yR6^&^TTFT%ZZ3z
z<sF(Z#D9v2orsA44nq1}P_)0OVag#Qex81lb(A1~Px~Ps_iX6*2@}scAYuF=NWUZ^
z=XXV9zKS1UK8|1EJNpBy`!K&{|B3#GeuH&L_FJHXKFH6y1L@$){D}UDbnv79VtpB8
zeTa2$)<xkjmIM0mhc0xOZ=)CJlAip?g?^}i^Z{Mg(-=1q7S}z{C;EU7VdzmF=#dZP
zJ95AWzsY$W?1lAw#!=WO>(L<nE9W-pcd<{=uRx!1BVop4jH~Eh7%vbHKjgvwi6=}v
zauUX#C_l*fgLJev#uey^^yr;>K!4~PJ(G`i4<GbP`-U&&K#w5fE5e+267^0z`ltRV
zH~s;8rW~Y49`s20(L4T^a#DWsQ9kUR@gjPo{n3um8~Yio3uCYN2mA%?ig=Lxl%H`r
z^)K2XVNpKhr(BeS^0Gc7`p@zfMjxyfVn^tQc=U-Mr9Sa@*dczEb^~3)Aat=y>Ye(+
zzQ~Wigg)`u4f*h&#Pgl@Lis2MVf;Pv!54d_obV?exj^VZmwebK@?ig@LvHv(kMz)`
z9%y$U^J&^4{^IW=d)@!)fz`IHKcH{?6#MM>1^k=sh(8whgOQtf?1TDZ{6H8zW54Jb
ze?flu@E!j|eV}jrFa06@9y`E~vF}7W(O$6^v7JyZ&PgB#VfHN<x6`gTS3tR`Z}?L#
z{3-Wx**AwT_QJXi{XgYmoX-3gdt<&$dh|yBhn}EMza`GI&@cTL@#um63%=}kVHfnD
z;{2KQH|8JA8z?u3eyAV%C+r40!cQ=tq#mg!<Y#`)`VRAa%E3Myc8b3so_35srhJUI
zNYD8K<_C;7_>MiI7yKo9#IL{yI|6AR(4+rgT+KX}dZ#@2JN)4`y&ZeVJ<k`<^;tJz
zJVm*mUi18x);E@U)EoYs^%3@w@!yn-cY5%v$c^37ZYV$FN9-GU(F<~+AKEeYi@vZI
z@?o#&gK-t@0DD6&#?i!6@5E!@q{AO!57-}aP`;$cJG3hhKZ3tu9fNiSQcl)e=<m4a
z&bkBZBh)i|K)z#-AoG6epLWZ4_=3oZT+DZP4}p4My@mB9+9m!M|Ak#pZy@zWdy+-;
z2R{&gAbw3m><s&YKl;SZ39}EuIvMx4s4v!OxVORn2#9>>kM<6K{Gq5H+A)3%#4pnh
z$%p@;zL>|s=g<qQA3f@)>N)752at5|!LQ?2s5idD7ym{WebP>$$9REuihm&={U~zd
zFX%t%xADjL8RSPF$cg{KPY{M5<IHXS#xM8K;g5Xu=d5F*U*y1l(OwvTh)BDpKgQp~
zhjxT~ln*+@V?XEx`I(<$|JX6|<DZd}bzJ1(_q*{sygMTD#oy3Ri1Oh-SO*7@Yny!(
z_3QTVA(K4X1@*@Gl=`IokPl=Wgg&Ao{xK>t9>AWF6MMo=8E+yt=ZINnWZa7V)9*r`
zb^;&TIp5I>-x*&p9-ti(k3RtUj{N8my7(dDX;;`6@7yrX1F1j86{HvaE`FQu*e7;@
z{ZfzEKkb9@3h~%8egk^w7k!{V{2TGKbCD17QGU{a)DQ8rSM-KH@tcf;sVByD#5117
zPf-up74-0b@S#4*&$yKRKKB2pXVSr+`4e`@{DyScC1LEC@gVZkpU^Mh52+WvQ$P3%
z<fUFHH+m53jd=w8828i8kQaS%{uw^xXWUFV_>R3$Pn>6k5AB+EkDTb2{s6k@3%RH-
z+9L=Z_%ffTzeGOLBPV?EBiIG{Ax!(ho{$SWCqAfL<O473U**R(ch$;^NWIgZ(HHfL
z9ib1-`%^yT<-Hs9PI<6f_Pfc)d=opMy)n-sJ@Y8S@MpgS{c>JeED!x4^@Sgzy!b)-
zMZTj~>I*+YJyY-44Ro=8^1&BGpTr{{<8%BJ{tbUcI{H=Y06#`N^M3Ntu0+JIhaqx-
zj6cGVekKfq_R)j(_xa9wpD-jn@`JQL>VtJp+9~yqK1IZ?@xS<0{4@Rv`q(k{iXGD)
zXh$N7^+p(eL5RJDA@&x8_=7OSUcoTF#3w!P&-kUI0p=GX(*LraihVP#z%ChQVqdfq
z;`vVdqQ7Ds!FSp{{W@Xd$%h}uztZ0lreCF>1IbT0;Da2LgZ$8Ay@!5*eh51Fd+t?}
zkMT4}yP_Wi(G%kn+B5PoFM}@qKlvFCG7m*Q#<z^m`3{ng@gwD+zl1*X1nhwR3w`02
zkqi2i8>C+b=@&utgI<~UfY<>DzcA#yANfG~0g(L%Q1su7FF@#kVn2kvfuv)eNjW)3
zNIk4Q`Sd&M4cabGn0^es(!Vk80%>QoKgJ_`r<~9O85bfS<2~e#iqu~aeqNzjlTBMo
z@*w<>mwKfA(LW=nh|DLD7sPKc{v(XOX<w9^dSiSGAKE2h5W46S#CzjMX;0J}<6!I`
zWF99XdLh4v#HWS$qab9xHw<b2L5M#xg<<1hG5*<0y*8yax8L@AqC54vrmcTtjU4AX
z|L=%bs=aq_^&H50AnlOf7iD~Xcc16)TH9;;ZTs)o5BpZ!3uAwP{d&%AGLD8G_lWpC
zQ^w(h`8`n1X|f(qI@YyGU;TuOn&0zioo(fXf7u}iZK(HXl^pAS7d*8`pIHNIZ~Glp
z_zc=-eU*tVYUR+=*o}|R`TXmexzhF@Tz!A#9k<aX%)J!$Y1k*=K7@$KgPuU~yej8b
zsSoxYSU2W7>$U7VF%M%N$~$;fetCbI-%VtF7dyr7>A&bV**Ag@<pl8?tpCw2STCgC
zh7aR3{2cy`F#ZAl&<{fTm#7Fo#z%zluVOzchVjcmi2Okq)ZZ|E5s`id{~m<+HO5a-
zk?~Gcq`ixXy#*n1)9--Tk%;)GAf!FWBI}~`kJLLz|4Ki}`V#%9Xb;d~{!MzmGd|<}
zQ|?R9zcLQz-URd+r?St8UXhRX1TyZ%4~YFUVeFprz=!b(<z}4_JEooRPB!Zfq@#Su
z2ST5D3--@=3Hr2m`Y+bg7)P*fhWz*o=1YW`mw@nPTu3_T@Ey6(3v%(D_61$?K@a_~
z{zDl4=mYv7?T-2t5r2*z@PniyKl|>4nFrF36Hh+qqG$Xz<5~O%{)qe{Vt;&Rf0TVf
z`bV*UfL~OEKFE9_2=PzM|CygMAEJJO5c|MRgU;{M4zL5-Fa7|;j)<rJ>9<Hn`(u8<
zd<j3ocaZ%a<RBe%2!o^}e^5I76aI{JeCPKacrTyzcjmA7HNwnSxW5Kn{0YCufFI-@
z9P6D$*FSxK-)>uc{3`S*5B_W0{IuzflN!FT!Y4oLFxWT!5q^dC#kwx-i0{}rNc+J4
z@L$+7@qC9K{tdrQdh8YcoYTRdGrp$3V;l`KKI3;pNY6c4#*5Hn9L#t81^$5km~_aA
z+{i~5ze4*Yp7nj`z=!cB`AAQgbmT{G_$~C#`6a$nKGGvM>6wSX4}CHYrye-(ft-9t
zFNEO>KjuBmPoRUIgOG9uA>~9~#zoi*<)&OBQV!+;B2qsf{Gf~8n6EYIFsAm^tx9s}
zjeC7zh+RS#IVlhN6_N1^c7c4@8%R9mARqlE_JG|&2Y%Ei?FjiU{QCG!eO7Mqz!f))
zd-18UTepSLFNnQu%fI2WJ1cEk;^X&0{6fDYe;E<?y{T`z&JA1c*|L7i;(Pj@+qY$f
zfB)@VyK|k+JGU(A-M4eye}@~jEN)qR*5*!s>)ZLcdz(-5N*GTY#~Njq-Eq$JQcoA7
z(_d-UpzF62y`-?$wAN#G>9E)<G@RCRQuW;qn&X*9UG=U?r~9k{e(wY3`$4Em*Oq(>
zw*GbK(!;*-qGH-|>u$HQ<%y~%Thn!K{mviS_L}KM#i;t}-0`Gkl|TH%PYT<%+qmM%
zEpz;|FsfeSrn_dZM&mxMzrt6=xctAq`+`Xiy;nJx79#(NS6*CWV2$l^vUqo!W2;Z8
zRWau(zWLDMPYvu)l1m$@2itK%$<6motg+M+(RR5r`SfT&bv&wv`){f2|I+IHtM2i!
zX9~l{8Avpv=dMS*SZ9H!ipMUia`tJX=Xtsj`7+e20o8g=JKqDXH(v3>6?45X)T^!Z
zsD1Rw4@b8CVzDoaX-T|sT;*FXcxs3*i@N3g=h^S{>-X7*URtQTJhsr=du^2;Yu~)o
zS4CUOfgEePZS8tR*Bbc@?a~AFBVtnZB=VIFM;&L%(k1;-Z#3I<@u=^;q!IqA*k^L<
zrHA*K<i$l*{%Q3G8Q)*`$Fk!CThov1J!kTWXJ7YR#iZ6_;iKAavwXj=0aN|J?{m#!
zZPCDihxR-7zG2HfRg7DX`E46r_|fzMetE|rJ$C4F*OT4}@B6Oc+qH`>e*C(zzH8Wc
zShp_+UpUz>RJ?NTfG4gjnd+ND(S8r?^4g2@Z(ra?#T);!vF^})miV?~TKmFbd`11I
z6+Uc5pAp9#yw)>?X_ZTrude+W*XMz0_Xn!=r<yPBIz`<4u8bE%IfLfws`ZfNOGX2a
zz1FwYo9!R;$~y*W`@HYoTC?{Z8$4IB`)-4JuAaQ!lf@u?r1hQkBv(<_JUK1>XM8ga
z=+HpF9Va~Imu&T2$E)UiQ*p%`TYXn>MR)(rc5iL*UB_q7yYBc}2N&hzq7GZmXt<++
zs>j`U)BZh+@~&ZAJJ0g#@fW)EXu8Ghdu_*oHxB>7lSR{VrA43nqPFAFO)jjt@1dn$
zq2Y)JZ)pEUlhvLn46>`P&wO^()B7#-)5h7SJutJ;q`AH+G_@}jM0diymsH*`<SSnm
zgX&55_koYzaNJ{6H~6k%w~KFmJwJ4V?>bKS*A<_%Z@b=i4KI1A!M{FiU+QO==^>6C
z$ex!kw0=LS^62hAqS6!hzX~-RRUfvclf6HsD}B;>_+!Hl7q4r$UEVdUbMvqh2Om&5
z9~E`!HE5qY?7J@hA|2iHB9VU7aM=CIp!fy{^}GG0>C-&dQTIJ&+vtVe<Hb(G?vL9V
zZ=1fu<cWG;%=UCae!=$dy8IRGCc~dKpi8f=(Fv-5+t-~x828Vni>j{llE!XN^}e>r
zWy`#@F{s{5eeZR_ty5-BUG2GwFQ5I7+T#vf@41R$>k_8w-4xyYy}Qo6`m<$zR5T@r
zEM3|C1KsIm<+U}v>33wq#<5Z3P0t&mzF#Su&enQIt|0x}KK>4q&$aQNX?dydv{2W5
zD3OkAc}(kBH~pyhzhw20RQlH@Hh<{$8f$#lFzs_6y7<=|)@)G28&~>qF)4cy`M7#7
zd#`3Q8qPUymLC^IeMAjM9nWX!<mNxN=qsunWa%;usAxd<d^w2To*#7Uc3Yh<{kW)7
z#w^c&G~oJq!8PwcQsK<EH~OyP^?MGwVCvw_eunAl;j?@C-*nr>n|xQ%^>%#b0X^%l
zyrj%m#s7G;bpMOnpV^vUoX;es<+S(tGX8e#w`TR=Rz0Y`FDaWZDL*9gQQgND=_DN%
z{dn48UFDa3Z%h|o*XSH|{Xxf+)?DSMjjriUlsD;dn#f;wIO+1l<rjAD-j(<{E0<O~
zW%Y0`EsSeNuD+`kB)6>H#MPf|`9DA9<1=3GzR`09&*)h5%(HLZ>Sg$AJ)~t<tzMe*
zT(fSQJy5qCX^&Sk{{MS_JM6t`+vB6YTW)K*xb3V^-z|_mPlF!Q8jnRi_eDD7wFPzm
zzL72EN=l9yn-6=p+co1o*D&myc2fF-&rN&X^2Q0@6y;sVSru<z_@{%4^GPB4%g{~@
zsP2#I&WCow?uQ4(Pbj(hzKJ!KdT|jsY{j_cvn9Xj`$DGZPj22~&WPHJyrj@pJIv&A
zTLZes)j{-hYboeGNZIdk-@(*fE>(KE?+GPM@1DAib}8!gotG5a{*H+#pQ>Hx&Zp~l
zLsaR+Esv>ro9n-0VB7CT#I5I~@(1;ILEn>-OXTC~{(`CddZK<*!%6p(w(~pvm1YgP
zeml`~4aNS})o|GFJ*wKdYQCU*TB_ew<uIL3_I!tGdf9S3Uwgu+83%0gGPF|<Vdr6F
z>pSkdX|nuWO&|B1V0p>E-Nt=}ow0p|TzSW+cCPBzRr5uaD@*skH4xWd2Ic?X+`_*W
z>bw0x<qFqGx+%iW9}u6el)vRJzH60q__q>eUw%&m)}AwdrOBla*DlGO;TYcC=Gf{}
zYE{fd#ow!&G~?Pd=-HxLuQi@&G+jnl*Hc`VTlRfX(wnx2F#4+dLt?%#K1s)`j`MBF
zCwm`OSNgd1Wc$1K$Txq?6{~vOwaSylxcZ5jKkj*|Fg~*9>TON$`ufoRN6u)qvFjpV
z7FB<@EX*!U$A_JVg#TeH`iwZ{;I&><oPAgSXZqc_&I`k^-vLNFUiSTrsOizGZj?O-
zW@~!YeKPXt)@w%Z*BS`>y*$(NHdpBXT5I8<p!5B<e7EHsMz|*9rLb-O8?@doerF!*
zDhAnkyxJ@(QvIE$xO~O@s^b}vj;mqu9;U11NK0-}e_8l{-|sbDKeF!@m^#mC3td;|
z{kG13WaZ-hf<nQ#?N@f33csNJdR^beb^5f!lv~7oukASS#^FDBu3*?YR#3cH?kwz<
z24vrj4cnJ7eP3kScBAg`CVuDJg575g@OvLH-!Bv#SAEo9_IrK0?<%r>nc*dsHw^j8
zH?0TP?LO{3Zq@ws5~}x>gVuS%z902QvrQL|`rZpeQ7%{gL0q}1CtESB9#!MR%A5AM
z+Z2CW&dWrV!}R&wFgu7kuL~PDit%B5lZrp@ySLWtea8mRRV=jod);>%)N}RZ^`2{3
zdBh_Jb^2zlX9~m4iI|F4-AC5ler;LDMGoEN3!(?V47+_f_`=D4R6Qi6Pt)rmNzsq0
zS6kC1r6*H-Y=5u3yvP@KKABd2+4Hzj(+Bxc@%xQ`*;sezK1=+xQItzI?CQB7*X0gd
z*Rno7;dj&@d#!J)H`_nxW%#omT;JEXtsZ6lrmpn+U%zmNjtzhGY{9NGzG+hR>rI}k
z$nPS#iXT;4zi!NJTRhj%7JHVJBkubHy8MDjNA~=ZE$P!fH!tc@46EMLOFEw|=R=~(
zam`+h#(i3Ug)fWwdv~3C^=Hd`SrqG4SJ;(xd0pjAicZk)gX;R8R5>y}$!ehD@khQ|
zfA}gtDKuTW2CMr#vS*czzHT%vcg<nV1~t5KrJpngl`CmE(&iVW2i@P}zH^b~Z#`z0
z4vW3C^q|`w;?fKHJpk9fhrKVZD}H9B<M%kL-D=-eJnH&`jw!9V$}evi)J~3FR^{x|
zM$hxYP<8!ZJm(rWta{$kmV8m)U$Hga1Fbh+@xv8!JzFuz&UJnFUVG#7p4}_ulfnxx
zf9!;B8<*r=!=&^h^3he#X^*F6`R~29%8#{gUh11d&J~zCe<RYh_5HXoxsqOog`Y0F
zhfdV<465S-<M|Rk51{Lv3*Iev9Zlbx&`m#V{2Mo(c9o%B8ZgygCPmjSZOTg%wV%n-
zz4M2*y=Ho0dY}fnU3}~7`Jo&9LPg4%VWxpJHSp|t*BxK$;G(=~6wAoCUlm<@SKzgY
z%^$kG#u{H0dk*-&qF3Kt<7cR*hor_cY4MRg7Zf!;?e@k2m1gWU;U}-Wpl$b}@9kf$
z?dTnfa&htLiCa3Y+V3YXDRlMRGjglmU$!Nm>btzQ<a7011ySy(_vJ{JRzH#!UtQ-E
zP1A{bUf0%ivgd$pOP`i~>f$T=uCZ<D)1G(9zS~QB)pyv^&L_Tm6c*O??iJ(y4DHf@
z>-%c1%YXVS%^Gz5cB1Dx>N-zp8y#Ksa^?X&>#w||%-4;&?<FKnuf^iGKC5)mkG|{p
zk8}UAa8TvVzN<Ly^pE#`w9QX`(x}=Ss;=`Vm9O5gzdicYffe#`aqxs*7xbD^oKFi?
z>qj<U)b(Rq)3MHCJIdBe*!;lO_^9vM+MX`zd$+cyGrf+cN?*1;kUs3)L0$9awCJen
zQ_L6CzR+V@m{xykE8n#0A&9Sd{@u24kX^g_U045M>b+^xban5q2HCZ%-^1P!HNCEF
zN^horuO_MarY=6*YjPcR=_6@6uD@%X)c&q1e*oPTcRW<`$>a)oS25}Pv8ME7svJ!Y
z`}><UCzs@-;+X$?;;-v_l;mX*d$I-N?zgAypW@1&<u9z?kBnNMy5^6zv|~{YUHg>K
zL9cPgfx6bi#qx%Q<DRdQeQ&_F^t$w(Hl3u#lcxBjtv^-%s`H+(b{RKbwLG@w)73Af
zMJH?@#Fh9xKj_x&wmM(>uAplBv?ZVEb+@GGhpmgM#v6YRrsu9lyjW*}7ne_5e;$@U
z?mHv4<d6D3T$nzRj%PmLD$ZHCwAv}FhkF_RtcNiBvF-e=tUR{-u10y4L$<!+rVq2{
zuz1&gr_0sz1+sFdHJ?%C>&m<#NKWzGR@87%`%=}T>3l)*#{F*k{DUd_w%#i)FLLR&
z_d*-zx;F37{SIeFZ}t9F_xRY$^pNQxNjr!;9#GA1>UU^V<sskodk(r_>fp_Op<&&j
z@11Z|r7gbgxaR#wDxCTDM&DI*{oPgF_qJW9hhMN2d8Z{UOlur&+kMQS^#)hJKlWPR
zR&Ta{&<jG{?IG-aYggi<?tc|px}e|J)@>)U^mOlw29>kCe3#XKQt6Lwa$(JV4=weg
zqHVvkHlgI^`zF>{>bZ`(^s38l)^)Gde@e%qT-q47Jo{eTao~-^fADl+QooB>Ui@Oz
za_V}=Ev`M0-}HUsr07TW`(f$OOAvlA?w?H;RbA=n#vp##_wu2Exc4}s=3jZ-io5zw
z{LynA)9%l8@xQcvov)USD$3c0x?9~xmDaBtbK4frbv&bE%`?xwb*tw(HaleZKh*13
zoO2aj|2>uNyAA5Odh&YDRs8&vkI#6$`$o@oH0_UM{iti_Kty>}_XA}0q?%rK{-w%?
ze75}_QkXogUYhe<vu>L`S23)7w#S>&JNBrGhpp%{;+TWida4-2SJ&@ni*$ncCH=kh
zq!UkB*1lrS6eeAsr1{D2-^luR(z||-#q|EDE&7RjkH(h#uI&HHwm;j`M_s2^O(*)t
zuFrgS)zkYe^mU`iC+YBj9qV^HVpMU?HMD*F7gfHvzh|0Oe%1MQ()OM4%QTQ_Ak#pm
zflLFL1~Ls~8pt$|X&}=;rh!ZYnFcZqWEv=J4XB=vH>H)dejhK4FYl+jiecqTD&Cac
z=%PEa_ngTio_*bO6_dU%tV{1<bX=_u<QC6uxEfCC-JUReFdc9DoPloo*qJFbt#9b-
z`rXQ;=_J*En&RWyyQx9)s@8kO<Bxo`{_s_P5N_RkS?i;Rt@B;S3p$?LW!j}{eOJ-d
zcZVkY>xxg>w_Wcih4@nu8y$Y(ca^KJ^IgZ+CN_WQ_8Mz^SrqkQDxCB<9Db(ClhG~T
z8qjqvHfb%&zEduyxBdKT()x6jUsCTgh<setS0;C+f&bP3<BJT7E`I#Fv3^uN{I?EO
z{ulcrSHn?%f5O&uQR_QP_v`i{+PrPQ+bPNyHyro(_iV{;>fLhPbnmI#XqTc+-+8(*
zt{u3N|C-smOjvt!t-LCxT~5{W3%2CbZMSLDv;AJAD4#8MAS;J0_j}{Yp?e=RX?nWa
zTU<XXzE9&?SoI#VE%}ntx2l~a^*%rH#r<8A2CMr#vS*czzATD-kq1=${ciF>M-_Fg
zqn9^2s`?i5slMxDYrf*!{Hfh1m3f7N;`+I5;jnj=U5W4ZRPSq>T(->9jbeLqH5~W%
z33Qcva`P5*M$}&9rG;VTNNc?_Os=T$^y_J3RC$BasoHB0y+Zw-mc7R1JK^3-DsLF_
zmG3GhHUBWhC+XjvGu8fr&drM7RnL(J{f=7DcU3;3yr#lQ-FI-EkF34Qu6Kr|hrjIk
z__X?c<W)se`{c6nL_J45v(oW<oYii%?+U8E?`=yyTh>KY<#6?WoGtYqR}NS8kK9SE
z$Lr>k^!}A@eqwtU`;)l8XQKM9rkF46cWt7+PZ~8{7{9Q1QIBC^u^!S2hskLwK4`v?
z^!KRc(|yk%Dqr3A0Hg9%O(*MbRr%PSuWh@HE1ujk$CpK2<<kATkQu#S|K6sqdeTkD
z)%S0Z`@k-*y*U5&1-`2ocO5w_zpA|?-LH!LRONOppRW5KHHS4D)bPfYepC#47a{xp
z`|m~+ihgeV%f`Ax_gUf>8oJsKxFWZ1|JrlF{}sLZ?i$}UG+kf1_c3JW6}r>M-EU>Q
zrwg%TSLP`P9@_8R`-Uy|bR&9j4fnma<G>q-|KQn%rp|S_RzJuq-p?u2u&MJ3L3CBm
zS%l@I-gKd^`!uR_(po1s#n+U7Ns4aRxertEy6icl^ZOczdw($L{Ho`)eqRge$LhY9
zUnul+)lAsBt8MYJ=jd!p&+lEUqVDq8@~(%f9J=?XlcuLyZd>x1)|W0nSEzJU{Ycb&
z=q;&nbyPmO*N>8>XWM#DRC%J-kFIpspDZ>W*6qu|7f$wdqpTdZrPqC~Qk9;nUR38f
zw&oMh3%DBArC(LMh?-ATzd`hE8HWYQ5wz|a_Feb=`;4BMcMYQIHOPL$zJK(|4@b8C
zVzKWk?$vBY!#U^8@?Aw&^^V-O=snENb;s*or%9Tg?0G)j`U|3``a4Rt<coX%O*Maz
zK23cO8(+xA>z0%Ic&0H-Zd30<sL~CSS2q5Vmm2)*!}g`VYv{^-@E|#pI)^B052o9j
zZu+|BwQ18qFS7Nenm+0AO2)6zg&$2H;G+lCdUG`&_U#(F{$8r?bC7ZEZ`yXF?(rsm
z=f}mU_vcNgdwSxQPOJ9&$#WgWcMe<&cb)M~ld4~D@-j?U5BK)3)^_v`MY-~Zvh(Gn
z($7Baftiga&GlW!q~EWMdyk0xrtfc=rcXIsMb&i{Tk@HDuP-UOvil3-_)^#Xz@T#`
zvVKw456PB4NKV=Bvh^GF9;jG9vf-r9cgxBrn?CBftn!}j%maGXUwKKH?+X6o+<z<_
zRC%*+JIdNC>Eq77qvnU7h)KN%A@a?_Y8p7|`h$)st+~o~6_b8X#kSw8MV_eb*0yxI
z$Kz_P;J3MS%LP5%?<)q;+k0)5A8X&d)OQt4&wo|v|JG8>eCc-at*_^YZtz`2-hmgj
zlfQS@xmSO-%ugFdxlD&;=WVv7PikE=<MU3xexH5lrKJbc?O^bPUKjM5Qk;*AVRE{*
zAC0`a-ls~N4(B0~#xT2(jaR+55|l6Ke6Orrs_9L=E0fWkRDHLD=6GRxh<g9m)^uUz
zvpwF_`#@3Ys<vBO^PRD1W#1b6RnMEou=2O9IH2X!hquosg<<j#FS}1<I(^iAF4=u4
z+tbVL2d14~b$ues=k|O4<*okklOoE_e}l%`;&)y50fOG&iTZmq%=c7L^_{z@@<mN2
z*4u~&Z)pEUlhs~Q*z1B@r_7wX+B1c!zoTt?zNq)tO{beb=89E4?po!Ap=mip`l4MY
z9oFTi3XKl+VC(!is$8mm-<EuZVpp=`jI`5Z|E5s&U8%J5MZM2;bdw8f?t5sdmlm2D
zr`SffO@p2-s`Xmqxq_8PJaSN{Z`OK+f=P|PP4NjE5857IqwKOf&Y527nZ~I7fUD<#
zM7e{`uSL~^Y&u=@I9a~B(ue(CM;1S4@`z_&_e|>{?QvsV{)IB$S6w$tJD)4(D1!X6
z?cdYh@6p9S>i%9>x~|WBcGc7SE%a?caU5VetgHVO=_GaU$hP}fq8@Ay>pnlHtGuSi
zDW>Uj4?-8B*P!><g1(#fA4$=FZDR9>Zm+S%PYcn%DU7<FZfm+i88@e04pa6OM%Pxm
ziYiyoIR)FltNvc6E%~DEN0ireVeQVfc-41w#B$1>qm;cTCi}aBVe>>WUe~w|I<oo*
zx?d6XyX-oPto}(aTMt3$Mf(X057@DOw<AUsJO7T`?6Hjx=sL1OPB(_>PmB+H?^#T*
zJ8XJf7FS-?b%?lpqvqH3duh;#%D23w!|%C{;(6kz;i&zOYPzU$Wa;|0db9n5UZ#g%
z>p}JRO>E0&`uAqy`fb<kPS^TyTs@|pzr4MJoRv4(_pg(-Z`pZvQt6|<Um7<Zznhmf
z%D%%FW+zGQH=E)!?(~oMezeU`zN>iEoNp?wcw?*YDqg?mpbMrB-t4=Mb%(xp!cmpB
z_+c2<p2Fhep0l$pzpn9KQgmD&U#ES}P#1sG>yjD$)%#c7<6}>k9>V;ZZSkhxNy+HD
z{rlav)t_zcC#pRCcAW5-U$WIt8eJVPBX`x~ZoFy#o<(_Ak#fYvsCJfAx~T88#Z70r
zJV~v;=;C8aAGW_YXq$Yd>cKU-uJ1#O^1D9om6hMLy}5oa)0BR)_vzwkR`q-b`QmZ^
zPwsWyYxSSfu_$LNy1w6d#w$%OeYkc>&Q-j$eVwnCjVj6|jsLMwz5j8ZIKyt;ZmaX9
zZ%Pk(Yb}@>{;<$7XkR;6GjcJe-jfrhNNWE_<Ri)v7FNB-VM{*Q`FfB(Oug$AMmOo-
z^PBLmD?Vx8cD?U9es)j)n{K;!lOKdZ^=#XBSI0}D+`8|jBu%f`A-n&fUdQ5GhGzAk
z+Wu7g?Lv9q(3WyUt-rYMiA3d-r7LU=n0iMmDXr;VPq=8mle?FUsF({wQ{@YqpV{_Z
z*ZX1_od;$%nl#s!)kE8M8&^EJWsa|krtCK=U0eEF-Ww42zg!C^<sYKfuWY(~ukASS
z#^FDBvM9?Zsq{hbD};TQmCyEftHb0o{hoDF^bb6=-?{e<Tkge0+4|R=-qi0d*z%5&
zZaHT3-1UeT>n!jxG^+>R!-$Kz+sm5|9sbn74kbC=s9VmAUZ#Qn)_|+$h+UKWzjZG2
z|FO%eoPFBpd7dnW@vwbgYT%<c9QRn&4St5v^ssgFWv!1Mw$6`=VfLXKzyFalT5asQ
z$k&BQy>|?s8~?Jg?$CXf_^zYt=P8rYzwPhUs>-L@ezSa;1~Luk(}1b-kf!Of9&Q@-
zX}Yi|g{=>pir01C&~$Btttab_7t0g(9%tJ5<F22jonQQ2^|ZsHeTZRQ?aEa;qCRB9
zvg;qR>1E~7l|Jm9@3iCNzQZb>o0km-ojcCH|Ngn`Oh1`^e(&7AE%u?ykIK@Ky%!!e
zeNyLdGCrgJvft~|{V+X5t+$|bvUXJ{>ATMOrb*SWH+gZfQ1oG{9PB&9MN{%+bTbXO
zpn;Xgt+=c2#2>vv$1^(CJoD^Zw|c5*sxChm_s^z_s;=}*qbd0dm2S|u#<lNp&*6vF
zuj>7Yr1P;a9yab3<86DlwCgjUUG?;S3w_<_`nd;L`D+eqHmKo^D}7ar`g^}o)9J2H
zS$bjZJgNA&<N37n>pH)cG@YdCcS6a{_f4#^)UzFB+mEUAN!5!iAJgO2r0B=JZ<OVa
zcW*20{czp<RnHx&=96u=LFrB1TeFRBQ2AxQ>wY&OEqc$_o-k_00h_$I7*&6zC!Ki8
zvi22oNu%!jWzc(K@|lg!7+xu-3!~~$On1;*?_J)iahaDkigKxjRsA9PTz@Y*uKc~G
zwH~uehsB<2DEoea?ddtM7_}Xo_uX4-_P%3-=PDlE<ieW!9$M<@#;E$!l`ii4U}@)1
zdVN(lKiT(uY)fzY{J&}XvhUW~mfp5?dt2nu<sWUQBl}K?Y<Z&k8(F%t`xv^?yLvu8
zOm0)_vbyN%?l;5eF+Z{$lX`c`G@rQhxU}-S{vMO;Jr-T*W&ORa>7(vj+n%n__Ge_r
zHMXTs%b&^eZPTD<i)y{rc&cb?yOx#fuoZnq9CPqm&owNRcF%omvEFR?9aB***>K$P
zmn=Uqy)E<dwB-=>Vp}+Do@Z;k>2*5O^le+$Q<X=Tp2FJ2QP&@IOli$ke$p5wmuh_e
z-d*Qj{n;|#RlMY-2LJl7eW~v%rhT3@E`QbY)wbkIx*eP17q=eEOa3rF8TVa}wDZqz
z+vviNrVsFgFsS^x&q>Cm=i0dzQC`#Y)1Z2^?Ypk`_;mTFw9jqG@;CMSV@c6X`ksp^
zex~|u({x=O*S|Kg`9rtYSmV2nrtDFbuI%pvJbT`C$JaWzD4(Gldg!;~gvb1nt-dPS
zHhu_`=Yo#scA0kRTHiDxN7QpcuBX$ro)(o3`b`RLKOYlSzO?6?aqCO<ceQQH7xx{G
zwDNBpSNWC;o*LpOg&n%w^`v*g`@U{;?cEmS4O*vh<-4i()(eepTt7+vu=Af$<CETx
zPWn8<>8~_v(DmDio-W+~rPcda-Q!~~LpeQQw`pV4@l>IvOI!aL|4aj!1~LuUrhz+l
z@BH4<>Q!<X%IU$jdd%d>G>~baa5bQNJs|2lIc_>vYlQc7Gfeh(kHhRG?(a^7<qzXu
z=<%7oCu{GbzSI6)H~5Qk$%budCqZ(XmQQ?VJqt(sU8YP=g`p?*_gqES?J&si>Hcoo
zk7o3L)<C}_f6+wWoBFov+^}W+mV31<zNhcGeOp%e_utO7JJ;#FbIYRMeLL6vceqi@
z;+Dl{ZSM59zMY@DxA`=$Br!cc(5PFT^{1EQbTPbl^uF6)KKN}vDJ&T|dd8l`oBTq<
z!AF1DZ}P}3zG>7|@2Yfa`cp4|u*NVi2vzCYl5bA;BNiOD+Z-<{rY-mD5AOY^H7_ss
zb)jr|s;#JW$7A<R^IgSPe|mWT>spNUqheJ3ocd6;ml|KQ-b)I*&KY{@PmL#gX<<~o
z#7$S{q{}LvKX|dPigEdu&Hi-F#~n6%X(95f^6y_~^ABByuJ&BT>mRxxe|*L7ytI*e
zupNW+WV_s%e0nsX+Ar(j{#z<zch8+TZ--Lf6o&N+NOVNIk;m-t+6GS*UpuK~>-+z@
z%F~U=m!Vz_+`aJC?ia3F<GF@YnoPZ<>zd_W80ytldelBz@x+`z)+@@(Vp<Y^vaqcG
z)qUo9vZ!0$1{DXr(R^MPFD=wv9$V-QKdt#gi%yv5tD-ICK#qx*)vvI;+mF7hIBeRr
zgH9b;lrJ=-enfn6#mb4Fv|8nxMv<><IO;e`mM-b<-}8s%eupo;q!IqA_~&O&J8bP^
zll-`-%0H=feOviNUB@+@E@-{nweM#%9logAoeMozF>X7Q&9B;SvwXj=fd+Hx)}FC>
zv8Nk<pKBItiw15OdEpMt4_)V}V%&0=TEB9QZjc@u*J(TS?A{yv49)7H(9Tha<=JIn
ztrktkmieYov|ri&<e)=t?ECaS(>>cUZ2d5;cv1gpg-wtD(<+xL-_chN81rJYQZFth
zb^gH=pJhF+Jm;>zj`Pw&)%sJ-m-f8QwfQje2CciR)<c#r84b8L-Zmv~khX0f*M!N}
zcE$6<-W|Tcm&G7`r1c#?<0|S}w@*v|8Q)9;Iy7K<o{<#&w9Xl;^3{D#D=m7NAJx$c
z%O8!X_Zw_Wcg8cd#=h3{3*U82YyH4BzNTB=BV(THF!;gMzAT!SD=qr6>)C0gcXeJ9
zCb#MJv>>~(#gFS6uLRLCweG8%Zcshx{@&-#FDLBOv?T8;9<gxPdHzG&<ugo256@KX
z(fIwcpZpBd)kD`S2R>D0N=e>TJipN)8x{<yl#hzTdpEB4<lHS@hH`q)-G4--_x_0=
zeQ{W?*?ys8RDIZ%E@-@D`g_uP(Dm*`(sa_=k5WD7W2#<tucNBUsd^vE^u3&<=&Rln
zcO_rgz6f?>+j?14dBXO=ZHo^&FXP(xu=SrpkEh*a__GFVYiCKx6IB1>hunJ4TQB_R
z*^1AdH@-)$>J@TH<4bL(zuBYyRxfP~sy9>L>o>h}eA(hM&s9`?cgvQ1VdH;O^=^u;
zZ2xLX4q3XY=U8;*la<%j^rro%uJ?jW--DCg?~a<@)_O;-pmm#zuiSe79aY!)u3}R2
zepmQV-)W(#b#B@6RIj%CrE?o@@?62F>wB_#NGko(@AC)kckoi*HIy9}*_vM0`eIV{
zBJz=a$F9)Qi~5KfcCEc-^2+(isCJO0%QT>(0n_tB=<6DXsgyCx^B)b^`u<#0xm4d5
zwI!eIc?R3kM_muGJzcjaT9rNd`5NC<bbTBZcb)G)9xdJff_^u`_3z?(VNzO-`aW)!
zF5d4lWcqVkf2!{}x}Gm7KP2)|J@+8eNjfb1dkd!1>ngwO?{ewl>l&Tc#+=sW*fz!a
zw9z%aiSnAh4<OPP`Rfj!anf6Rb!xHRcLi;qPezq5?e`ht+L5Vu)zYFXJCBa5Kil$$
z{SKUM@vfbFPfFgj>`Jr`-ExTiVir#4_kBdYhusshJ>K*@$29%8?M$rCLJ9xa`<AA!
z+_Kb*is&(F`n1y_uPx~6`!BZWFRgc^;_?l<my)!8kA0$ecJHGW`mUq6?_xTfbUkJK
z<TSAB$jcWG9A71$HX^m^?_89Z^FOw{Cl^-^Th7baE(h%*!%PD@H4t|HR9C!Cng2~o
z*SQzj-wzej>n@inJyZ9glA^1tKBLl!J1<E(zp7p6&X<%u1o4Sm9@+i;@{-=Rb=0`^
zoK$|<eb%Jp68X5g?rQ42Zc#s~VgJrOmNxvND5ncue{a<`KT_W03432k)y`G(1@+^q
z-&N%>oi8c>DqD_kX79W3q^D}+GPF|<hdtE($$vb*T`mY^>pSi_5Lte%rq{JjoHiZq
zy_&+f_EaeOquP1Y_f>SIiz-)^?tg2*w)us4&dv6)>Nt*k|GQS~`5JcqfOvZ({#lYX
z6K5X#(YQ*vw9qzv*e;K)^Hf*YKV{!{wbed!y>}T#$8`A%WgJ*u%3<0b!szQ7mxb|3
zIzDWD&i44U`l-0}l=ObBD!-`j6N%~K>L+UcxaY9L_zZZx{s~RWiu10ctMC0Fx9aaN
zspbo_3)At;<AM<WvhT7+eOHV5u&o%k-3R4Iueverxl~*7%i5cmUbkK|dcW2{kX`dV
zY~E#id|Kx}#dDUng~jn#TH#+?qyCUBeh@jL?wcgFA8(3J(EK-U4XdtS#N{jISDhb;
zbX*OK>yxgQBQ3e3`hh}8r~00r>H3kqH>&IRBW$H3s}Eb!r}Zw9ZoX-a!%Xpw+kT_|
zUNPz5cYdQoHY^xYDQ^l@*NIH!que5zeqTj5{qjDp66>$Lg!gE-<3WACt&)q1VeiqI
ze%~VMx@c1AY#mRi%H_&?n?Z7#9v21iEwA6ZUODioDpN}GuHv=l^=SX&jTQ4rBmIl&
zI*_<tA2l3x9Y-9OM$K1f>8K}LQPqCJ@`aT*ss1%-KDO-_gvld&k61Q+m>ooYH$gUC
z7~iDgRoBmK$>-`h+#tDi?JFitN7nvj)2q(k<K~l<lk~dVuWjQF-Q^3S7qs7(eILH4
z+MNr%AUy=>Q}=srNsnJl@e3P=*dA}|`M9WZ>9)(H>EZWX$F$$qR^`t=jcyF`qvH3l
zb>pPsMY&|damPzp{<^hiY+mfi>LKazx^8}9>n*zDIro;~!~0*?VyrK&XSx=)tsZ6l
zrmpn1jUQz1wb^!0Pq&{-nqHxt)6;#PBPn`u`z>95L8RlFe@EW9`iz@j)T0=d-9Jbx
zy>07(Ve+V+*HB$|5$jb~Sap7ybUt0>O^S}`@n2Hq$oM3y0aNS2L3B-*&eXWZwA?}Z
zGW|WMTwUi3J@u!?lReil>Ubwe54ykGHa~S;o}}lKy6qt@y`=YPOz~rWV;aN8*;)Lc
zH=57u;@PH$pmq}VT&wClNcXtfw0)bVAGJTVHC<GDNh@8Do$LCZ)IKVFO!XJ0=|(+w
zp*x+fdQQuJRr%`nvm!n0Mf`3mZ2F#?Df+bAq*0gMLx*>g!qC?JA=!PyEd3D+j@xaH
zr%MmEjeo=BF}-h0yUNfm4VZdIAt}0cX;WSrSJ%%>t&_*4OAVNMkK>uDJsQ7X_LHAB
zQqBxB4Wy|7)B8)%7t6?ZRdj7%>X9+ebr}5MYF`y?c|Sr`{<QO%UKcG?`m|eH>@O)f
zZ2O&n#&y~bJ-hb?KQ5Mx96e*t;!S>1nD#g;F8_~KJTd2w^@{SYqN#CoQgkoAa_jwf
zR9)x0j<V~urr)Uwqn~ttt-5cJbiNl?tep5st5v?M7`OjOJHP9GAnAP--TYL~QRvDS
zwhkT?Z|ix4Fu7#c7j>P7W89yiT^i6;ziHEnyH1=|e%<dQq(!e#*DG}OGp2qY-F3a|
zKJSzkJ>9<#kTkut-qTg>4NaZ1(nYt=onKDascA_*E~dS1s9Ha=`O=;T#^tZN?<AWq
zuD_S97gyIMO!c=J-ODPTKX|b(tB0`l!L;IK+XLyt&IO9Uo90?r*L-gDl>^4S*sRnO
z5k01bX}uF+E8n#0A&9T+du>UlxAmNN+H%DmpQM#v_r6b%UAy`{Y~A1VK8h*5EqL(S
zYb$x1eN!mA-kx@P-R&SPdSUw@w#KJj|8e;zRo|xk0d!U0$Fe1#u666EbY#~t(oQcs
zf0aFNLwfAV7K}UpNNaosUtR07Y0;q^u3}nxi_2G+zHMp8q8z&SuUm~AS~8-2<=heN
zMjo@nYa6_@Q7ms*IPSj5uxZy0I(1}G-c@vU-CmSC?!HXg`IEX|pqr2CdEKyf88=>b
z+-7_Fw8z15`KNW>qP+7>+poptAJ^_|$*<}s!}5vyWwwUn`tz{-y2rI?(TjS&I7}Z&
z$ES6kR2Sc<dbc&5?se)g`>|#HQdN$)>jYW;pMSqfRS&B5rkXFzp2Omkx_7FJ5AO~Z
zI(B=aRoRoDukl?$aoxhTu<Cj9p!Sv4I4g{Aki2ofhn)wsH9qcsZ<hap{P7jP^HlYq
z+b-^2_-gkHSFQ011(QD47nfg{|H<OZ`d{5=o|ow%(}T@+5O+MFnqPGs7M4#{9`fnl
zU$TYXE(>e5Xgapcw;gTKm#Q3L=Z9^NH+{a@G=1IsCHMtfQTMo7mR?%(e_iLCbkWgm
zcWKiL+jrL89?~waF8-?P$*$)MvKQTUB1=#AecQ0{rz`QQ>&>$IPb$4^`?c-7l&t)=
zrq|`iblHt9_tE2)NB4cBr1sTxjju#HaqoAel|QJTl>Hvpp2%;jACZ+yR)0b1(Mu5O
zdM`qx6T~n3UOqHn`gh2nU%lGym(Fdt$#WfT>DS}R5%xQaX~pYqbwBpLrRgiTEcIMR
z(QmsNzG38rJ2XFZo#!g1{r*c_{<_b*CQUEy`2kz<o3?*hKN{Dr($24X-a%GRs_A9h
zi7FrRG465&RqM}|e75)%RXI$JC$L9Vtl#v?@nwt4JXH+h8};24*>pktlK!r02T}7S
zU7n=*^{=z}hb}``dr>i}-o$igG#$RE+MNqMSJCvnE^+-e3-7!A<%8e$RrTP?ctG|%
zi0$ch-M0*)Bl^dnb>_I=MLtP~(^{8`%UAXOf^GTY>LIQCs_TDA+jqwA*BbEe++%6O
zFN$)u<M<)B-t*QAKYFfWuR7<=`e9K?&NY1H%C+z8esys!Db#g-=hs#qJtVu1lvH}%
z^ML3YF`IID@5c3>oV&%#@Mk@k)%Jg=i}U327XFVWe*Zt+=d(n5y3W;!bVNR3VbkXZ
zg0###!@4ktkLmBO%=-&n&airxwKLiCYNS_PKa$O-`tA(*&{y1bi?ICg2}4)cX?4jh
z>POZd%6ocIf3AjA?{VnP=c@jK<W@Zw5=1ZVcU#BZVgB6qc-8et-TC6$hw1#P?ZK9O
zNv{W*t`F$j-oC=*V;t-{>b_?hM2~t;3xnjc^}Ff!36iP@Th7<X%3<rcC8}JuzDF)A
zm+Jll^Zy`}J(p)|de`T}y7t48rjzu0NJ0L(LB&CDG@sYSa~;L?{ixv}z3BdKdjBAZ
zzV32l^fC?nTLZf0w@GVL^?f_?sh)?6dX778x}^2#D!;Ve!EjYyncSHM{#ye<=fJb?
zSDtg%U&nbtdiZZ0s{9}JyJD`yn|=?~G=0Vc8NRhwrxxq|sCp1>-nMzCC|}%gTK5ff
z^G*Brp`-FQ^$tK>J8&ict|KpBJaBxKyeg($PSbwhG=16gcebYwT0gh-yRGZ=VRD(?
z?@Nk)TJv&Ue0A;T$MvI0y>n`ckFEY8s$5C=WmP-TeGW2+9`Z%iTTnXHbysoU9(h37
z{bVscbW~B-x#;pnM^)cqzNq_zuA~dQ$K+bP?e}q9v5PP{!_F(J#*6LE)o|SRU38V(
z)H`WW>4ud<_qtP5dSP-!jSp)lNySH%Hz-}y`V`Xz(JR#NNuT@D#V=^Q?aKF{alR|x
zZ8^Uew_QfHN7eakRQjY-t$$g+th(>y`gz%;<d?Ne)$wIeKKRSlbJ#kosd(g7McR*U
zl$9r{zHOh^iSn6R=S+&O?)Q<Brl)ItK506-_5;$UlT^Q9ijQmSUdSu@A=kp<{VTCQ
zxqIQO-7j3V##6<p_u0bIiSuAF9CV&f^?TfV^r9Za!m9fxy7Ps}X)0da2TCg(wS2nI
zF-7GYb^I1JT~xlR>16v0RX(=o3-fP5@w&<-?u!;m*p_`&UG=1!PEzw1Q+$$M$J5Qv
z^mtI!UZVC7rqZd(?OHxv?IP;=ebP0q({||Dy*Kz7mQOt-bsjWnKDyUWGkTX*Jb&<F
zA3f;yue$rwr0ALI_n@n~E)kSZcAlj>ecbamvipN!>9OON+Dw15NBymy?P%)v1JHwO
zD6UJ|5@y|3H@a3o$eVP%=+=8wdO`cxroOA5vk1#az3IZV)@k7zv~K3=ch%pqwI!df
zcZX!@=(6XG&hKl$^m>+V`i!?_@%Ocmer(d`BwXRAt7gLbQQP99j$>?17k5AEnW{Y+
zzhCx~uL~z$R=>jXZa@02;w}qowP-rF%y$)8kIL}Z>vQNsm!B(CI=auri1g5#?sFh<
z>De+5P?aO@`<iLxmz}S~O^^M_V%&3WLHT9nur0mp{J>OtRlNl53%L4S_x^<I^i=I4
zYChF>DTC<Su5Z>43mxO`%cY&)^!`GS{f2!XeDs(7CXd|WyNb*sTt(N<HJREslBFAF
z=epx<?-#=43p;lp#_QH!5WO(DZI9Qz?jMz2kUmX)*WI7#mNTjGP?+3N-<3%!U6{PG
z@nd(-oj7lYQr|T+{l2bg`U@Vs_S#C`X5Tc*+JmX}1XFa=o)_ri&p8p<`ch5r%06n4
zoTlE_NQ$m%y}6ptmiuw#O%CyXdfaea`wME%vfsshu(ZN)^=V6f*V{!}^`G&*Ab)(t
z?>zL7*1l)jd{y5iF`du0^I@v;B<-Jctv{%i*Oq)qzu%>sU)=M2Y3EPc?sfB%-S-#A
zm)h>FKT~V$YfZoK#c`(Wd!^#KdRSOi{xEt`<7L-dqP`m^){ksB>Uy}Ybh7fvrf1xf
z;jhPmqCS$kFRhD@E9aGE?UnSV$DPoJpNOwNxc8sdyu92ujUwMHtfm21&u1Wa)P7S}
zI@xg#=}q5@5&g0&VbycoLHX2L!EbY!sxQ;~g3vd;&uN;zuKr#(o!?rDnJ=#G+l%s|
zcTqdC^D<NEMY&9eb@e}K(Q&0eDKt5Q?!%b+Zn_=lI?o&?r|stvgXDAdKCmdaDZd>?
zH_R?%<HPvL#s{6BmX%92ebVnK>gH#9UBNVc-TFwH9`kNjG41^~)pl!ZzLs?=e>nA=
z%6ZcmR{qz<oYv*oHpTg*FialegZlZjzDK>+B<`oV9%dcR6_njy%+fc1XweDt{7esS
z=>b0}qU`y}sOR-{o%;xSPxpv+BahkPwGCbv%9c-8dR^s*j#zJT-yM<7uS+jMbadaR
zvxS~5{UG;a!ch15u`qg~T_+vZbxyI+=ui)7t%K|08?+AO%J-!1Q<>(&K9wto{hPw-
z)poyhZo^HUYZ&#OcGPjAsdPd9L)W=TTj)fs4_nfG`ot;w9C=7lu27KixvOZ)I9XMW
zu<@*}_^ACr+Uwq?`A5}*Y`P$SE&E-Tudei_&Ix36Z)y6<Ela(qdXT-hVOx4v`w>y@
zpmU?Pf7g9~D~z71|8aF+*wlI7r0Is$tE=%z`$JXx56jo>iB@G#e!j*xjkdMJFum9o
zulqcMuJV~$mq?1P>U=6FA9{^?&y{px>;JCAr`^uczbQ1m{%e|kp`1rZdtbtoeTC6=
z#Sa9@ne@0x7e8C}r&Z-}z1^Gki(&1~wRqL>yjV`z_pW5uErQOu1bvrXrw&>Vlchs?
z+5IeWy+<}2B!AF%S^0FQ7wso341d}6YunNX%?ILs57VC*A9ij)Os_i}WS6?W$CX#L
zABoF1YW}F_;iIOD%D23wb9LQOlw0*YU0lBumY@F8RaDh$)O<;;Z@J1R>3s@Q{8Z;_
zQS-(1+i9<(#O1H+er#Mlrk&rm@o*;3!~0*?Vys^%dPv&7ZC$TQYJVfHU(}Ufb{#3q
zPT22q6~pG&w#SR_V!0MB6n`4kzlF7@u=up@^_$}Bs^4%`?yz~jEAguHJK6V?vh>BX
zdmpvXkE@3;zh+y!YxmQl%Io_3g}TPUg+|A=_7hc}^BWzqVZo3}`J}PWonKDascA{x
zReYvukH+tp{p7of!=_z3=+u!#`M4O>&XP(O_AZOAc+=$(-`li3Y)T(zob=XSom#B-
zGmNGOQ}y5)-KgsUuB6j-?lNgQs_Pb}?Ja3N=;miiKf3n+(xM~#yA`_isJbpfK3(!W
zane6eukdPxoUQmkqi%K9pI(x46<duQS~8-2<(#WHe#ou&y!FD5Uec(`zC}7ozmr{F
z`H6Zr71njW)pj~T@22R|tLpeGET1S}TKkBi9AV)geY^5q_PlkFK4kAr=t?i1w>2Hs
zJ?>76Uf6zTP(9oBUG_ect?8dTZ+wqh)hpy&#h~}Cv+s@TJvn!ar%Mm2?N8O;%J#!)
z`7>AeMy<cN^MR;*vUG*50aN=NNog(WeVwT3qMj2lRlY~YJlA3HgR6bpv1H`v8G9CQ
z@?AsQ_cf!+$L}($qAB~0N>}z>0#oT3ce(yfvnl<Ebfea<Y&zL{3Z$3ilT`Yk_b<Y}
z%gV<-j;koUzGQ29UF&(W^{+d<DgDaM6J1H~YP)ky?y&uVsPVeni>cqS)GcR5FVnz(
zYryvTWueIT-#VB1KkEAtV!ALMyDY5LqUqQ&-&Tw||IO0Xuduw^kJ9%6!t6seo_FAM
zVbbs3xN;9NOitJDdj-8atNXhv^9xxyRoictFVjG#0eu?Kb?+-_I%TszUGs5=&7Nu0
zr|H6?6xP3+ir2M{VY)U<txpEg70VNp&Xw=F)~VB?W81o=Xdhx&oS&Dsu&58&aNKiF
zVfkg{(UrdKNv)fd)~S@Y72}?Rjru-GP`aeor(8K7_!~Lr)U7>Z^J34{cbc7-yW*fX
zn$PRv{ia^B@4u@TU4B%Sj`Q|rxV<O!KB{RxaoeM;T@_0DxOyXhq3FX@IbJ)dW$XL@
zy2^`-rsT`$W*Tro13~kLLi_$wo9S=%sK3=yMN@Sd_1sn5bf)AlRJu{`yV#zN@q{Xd
z)vxUSYFg<z?=Jf;Z`eKVq~lG!udiw!uI39{*9?o-y$+q$egJ&K+Idp(Nw+6m{Mf&W
z+fS#JKdJgn`dpeSKV5bamrhdkBFo3NeMD7xOq~ZvimtBnqG{35-H)bCkNz|%lx?>`
z>206K=#np}{IcK0d%doPb?xUw)t{=Lj+@WayV1JGYf<$mrnBYUL{To)aNKo*q|ZOt
z#?RIJU#8YK(xw}=9mL(wN;|))_aUR|PggqK_hpi%7q`F6@=v||!5YInRXyk$uck#u
zwcNJkV?EAQG^Iyf{?T?ivi?@KJf}WX?WM-otoOoDmTp?>Flmn)RQZR!2W6@q>7uK;
zKCPP1ws}gBJi6{vrB27u-$RSLZ&6<J%ldm;)2qIVX<NRu=l#0*yY^nJD6cEOzZ!Sm
zU~9XUl}q(q`?T|=HICNJmvNq0Z-uhnpLRP->vs=r<15;iZQ<0%@z2y6`&!d4d|5eM
zJs*kOaoeXY`Qz><l0U3nptqzkOfJ><wAP<>@s0X>6t<?r?p#G%+P|tCw%UcPTuGfr
z)6GZs`I<0#VSZ9}U%|HYVSXg2{9*Z$j!)`*ttmdb=ZPY{pnA0Jd*eE7ho0SggC7@l
zjkj!>S4WjYtasPKy7tl1rsL{<hpBV1rs;B@({)VyzJkbK_I|ACdQO_YtL-?d+_LLe
zVtU)g4`K4C&iiEZA%|=}SJQ{hqjblk-=xsgIp(D4rfpww>&x}uV?ch<kGc}pHLpsW
zj;VRPX}Y$pAIZv-^tzC%{Nnma^6T0+!EYsnN&ilR?(->O=a5~AkGk%pD;;*5Hb$L4
z6>7S)^`G%C9~#j0-ar@~QJYVk^v}~Pyjmfb6pH+V!f}5mAgTO8@)!E|wCB&R@(*jD
zs_{v+->`O7=yso#FVlca8qodTNm9R$tm+5T%7;w^-P?5KyXoK0Fhw8z2cc=Z2;y5_
zzPn=Ig(~M`Ppq%B&yyAMuA?pOB}k8^<x6`%DC3{~?pmt9J7@TJ$V`s|TyH<7#%Cg3
z(|$Cg|FZ`A9r=qU`rg#HUFU`^_iDL!%i??bp4+!&g@6C;T)T6f&O5g(>fN_<-G7H0
zwJdH~eAebpf9u=%xqF*W^GXua<In-;e>-5$F}^N_Tkh}Hwy4b%FDdN!Y*D`tp5EdW
z8h-s&uhZVXdXr}wb=A8nojKK;%qu<edp`(O>DrR-`Wyb#?D6|XdQmZLxz8`V^`LXU
zU+n2Z+44-;x%J3%?p)xxinI3rvT5~)=Xy~ws(#M+=IJM=RaogKh4qSOAG6nii~O`O
zs$SxzyR+8GEiZrfJ6{#!^1tnr!}fh}Tv0A9ME?E$&~W*=CB->eoP6qWfBv|4QO;Go
z;e@pd4n46#E^VY9Y{w?`dOUP&@A;mHw#%K#r$+<rir&5AjjI=VvZ#mqZ>dz<{9wOv
zofdnhutn*K%U55r#0x?s+OOTxwmsUt@2g_7$DW&ZOi3S4HzHq#dNnZM?{ybma_{rL
zYxw!ok58LE<ZUku^=d0UY6H4{`byhspL(*Gmc+HY9{5HsKh2j#-SW1*c<92FH;?wx
zLfz%Dg<hA+m1<TwbfK?`wv+=ouAO*e&+4aC%e#tuPkrgWL#kEH7aCGOBEIucgXzZ}
zS0is4MZU7(!Ot|m?b-QX`C%wam-M&ndgXo}RaoF9jqq2+&aId2y!e1`y|}2#zrkI*
zoqPCBAA98$?;H5iJ=YwN_tQqzd4_HIra!j&tMSt&`L5zux7<H?pPPsKNug@H&GP-e
z2I@Y&cIL=VUwgXo_qk@VwrF6|6XUo0y4@sC72}rUxCSG~KT%_zSKcv5k9De7u5-uD
z-+Qj$$2E8Fwx;Ym&ozv?Z|Q2fBW`%-xSMh_JX0vz@B5eanKt>DRen_LaPg%TmmRs<
zw;jh`KJb>ix|RC2p{W0~!tczP`?r^;uJTM_TIEvZdt3K&dp7=Psvj5su-l_+n&#*G
zu434FhUs|K`cut!+x1gVsoU-qKP{Z~NTp_<?KsYN4TI+Cs`ZfNOGX0|_I#}DjqT=p
z<sE~xy=3gJ?dtjKJy-Ff#|FMP<g2ZoEC%T#t?z4Jxbp0K7ESbB#qPIXzOZ4@Cw^Q^
zOaB?)OanSJ&|*WsyIwq?B=0(wZrJ_VT$Pf%E9k9kbLwgTsF-&h|Gs|vD=N1r$;U+<
zww%#$M+2L-eEh(_mKWz;!&fR?by}S|#d+7U<;MLc-S%mPe0fLHEoRDlqwlW0w#<`7
z({iOne`KXOCq8k*LeF*_HUFpmTi09V6&kKwvh!}$+AQ@<VUS&=bq`GBJ9hIs9h&aH
z*f)jSx4r4wOW$4PCxu-)Jl*_^QKh~t2Gx`9@89>j?fg2QedD`|NuA3v&F9AAJMMkE
z$2GodIRAlzia)yd9Y4c#^>9M_#%<nTKi+c{W!KZg?zd#||N9=0?*1bxy{PXaxSlSm
zK5R>u);W))^)PM0-Myv`tDJWYdw*B;nVX)doR5mS^g5vah>!RBXr3>Ny7-H9uGs#N
zg`IBy&eMe={itEpcYA7lSLLEdW|exb<8$jS|8QozAG|`vXKtQV>ylHadqIeusGcWI
zJKtG74}R~ER#QD!u)Lp}PP_a;?LhWB?Iy#YH6VL$y-?Bz)xYiKOY2^sE$b^m^=9h(
zimM*zc>nP;Jy%h7UBk8Xrs~}k-Kh68NM}k8S-NW)%o|YUv$=j+C@Zh6>DOJq^GWBg
z|I{~y55MsFA3HxY*pG{wCp|J}$`@-qQ&?rz_&tt1f3>F@ZLN3YYBT$rKVGxTYEQ(c
zKl{I{PU^kRa}_V0x3;w2g{!;_f7S!_ofe*S*$JaZ96sCAjk4t#-0r@MpLu<j?+VuI
z{Ptmg8#2=mLs>l}mA-85GVj$I%YE1IFDuSDx3vEf-*wcb_oVDa<m2jkYTE1Z5AM?a
z-Z3XG@G~q-J&5fj3uhY8se!Xj|7>xudFy=BShxLSHEVAw_2Z(>vX@6%y_!@QxBin&
zzN@(CxQ^ePaDJKZD!Qt7<TmXmlA{0i`G;)&<eD{phUw^`Q2K#F9e0cG^(WO=+P)8(
z$*1=2e5N<Odb|9B>4TRn`_i)olRDRxlphlLbbj=oNB(fhWKR}FI!T93-6vC}tE>F7
z{B`kljgHulrxkWhZ=$@Pcd0P_&!eXMNukJJcX-l8J1v=b@Ep$-+-t|?=giu&z;hip
zJ$S{Y3iUto(!#iQB-{TLTKc&9vn_w;qlUlq^{|qhD_HONC2y{~y+SU-U+W<)yZZ9v
zJ%;bvyf_EymZMPDS6nZb?D*TZ^t#SVq(vugJImI+e_r=SkEizEec-U!U;DNr^4fyB
z+qEs_N=lBQ1J3_;z@B4#*D$P|B&9$246k$Cb5{=D>SrjY2lSVrof=4My-wtdc2w6b
zM0%=W<ggVB<sO2m`!}ZPCar&4?J$$eZ4D%KZ$-D3bj=6imS1<dROy{xcI!dse81Rp
z6$@?sUUzv^?Lv1x)p41wd~wTj@Vx7HZGYr;dENNvipdAQ;#bVOil?lreAM{fC3#yg
zZapWJf7~^<7C%(<xfd0al1t>{>OO$3eqN*_>PIy!(v2H-wVse$b-tZ;KHdFwTJ*yD
zSyek%&8NFwRpl_9FKnG!SG;UFf_?`x`@Xbok9P0-Y3pHP{}C@tK47dDgtGM=cU?-B
zpR4KP-V3rNzv=yTSI%phmNTlI#~uI4=8q~@mhOLR;QA`7U#_<Hdrvq1Z*JjV(;7Fb
z@(mkrhQ)_#B;6EI*N4-U^0(Yg_X}b4e=AY;<@Yqu@!6t&A3VLqvkmV&{p=5quTqqA
z4L^Ua!^<U&H+xa>_v$9ixTNl(m@cDg|B~fXyAO)qRPSAxwudnKQTyGnbYXmwj@Pvg
zlop*ZeVdApTTia+i>sbnimRWf`QxtVhVfDLS6M#2_iV%FZMx&b?80<>Sb5;z?_URx
zfB59}UQ~=ao{vf|t#phhbR&AzjcN6dy7|WKPj%}xqxWkKg!MZ~&EHJ%(KT-RwbsHz
zy7-!+6ISm@#fNJ$UJ6rorAjxdUUm0Vw$O`Ln?*&o%rEYnJ*4uu@yk6~j9Y$N@^fAa
zv_&pg<j_6NvYnplIM8%`x^|z~w7g;S$SgkHcd1;{f86#fJ8psBVIw{namKJR&lJk`
z`=--VZV^?-scGelIzATDiRI41ZfQWaUkmdKvh-xrtBz-~e9?YyF06iSk2l><i~g^?
zhfS@MB}Lct_#lWry@cxTN~`{^Tf3rnuXy9?MP3+I+x%d^ah(=>V9^D~z4Q4CbNq6L
zRFtJSW#`r-&$)AfryIkHTleX;Ge>s%+AB2N;o+Cp<lcVCa|KQJ$6<DG$|ddkyz}yG
zKMZAmA3cn3Qt>Y>Z+usenM*uZvC`ZnqvrKr<rNCf{Q9!LtRAz{a}8^JSLLEdW|exT
zu+HL!n@-xh)Jqyw>p^$>y{-GXJsW>C)wcz8moJFktv`M8<*P3*@-vjvLy$hdy5;`C
z``kR-*M;Z(ZP-I)RsZF?j_a=9`K0sLf9ku6YhSqX?0XhX^j$^&oc&LoJ8gpRDpvlX
zQIGBVF7;i<a~AHj)8jiV^IgN|XO6q-!|hi1x-rO){^q+a>j%FPOzy+G?K_>j>Ap9-
z43pJE+lz-TTzT_o&sE&>o;CM0efC{1!*ulE`tN(%W;b$PQ(x-4-+iamlp{8Ju402B
z58S@2$tKShT>kla|LA(!I?q*1>t2g)zCGT4VsOKi8$H+Xz!s-GP~4&@r;7SK!>=iw
zblC}`M;t!ea}BF@tGM9Q|C{Eeg}>I~Z$nqDf8dd`_Wr@Q9cym<$EFWoSnRuks`Fyi
zdFXFzR?nAF^MBgEb-h)d?Wm`;$x$)AewbWIGEPZqVCl5M#b@<i;U|Tr%XYzbwTk}I
zv?#9|P0PLGst)gWsJFyV8iUI9+8?*Sc=rR>`mSTmKOb{g$)RifFbvXz?(ehq|FUWI
zhv#}3>e0ibC&q90b-PJkQmES=;?gsHz9%XAXZ1Yzy+c|}^-SZR&Rl%Y;rp-iUB$m#
z(ERcfOIQ2l4TIWAlX^WKI=1(GFAP^M*?G5WZI*hn_}se7Kb+a_2TvEizUBF?$1GXo
zxr)avE`F=uH%mQNv2*KXJ1;)qThCSuvU6SE7cR}`k2<kZJ}G>1ZO`2feW@hx8cthq
zcdx0#D(90%UG+So&6$6{WyosZ6;xfPikh$K4(AV>e$01%z0PkR_O~H3jbqq(XItYx
z?^0p<pGQsilSW;351ql!G{5cH`Cs{AczKrzjcX4b;Ms=D?tiP+8J9ioyN0&yL#xWu
z;I7@yJ$$E+Jy-CE8{RqYrrZqA6w<CTv`YiJ?{y_j&n|7sOLP4Ve`@yleIq^B(DXfM
zRr=I`?C;e?P2Z^MZntzB^1YunQqBxB4Wy}oOLjQy%EMM}^-ZH##y{-#=$fYa`MxS9
zy-yaGU)=jwY5Obq+kUPmNWQYU%e+@>EcXi)>vtOb`UfYL`nr*J`|NI|cb@i4m0Wqj
z6Kd5PJ@xHsIa~0l9slrH{=KTXxOnnTpS*ra|Ejs9@bq05wyZI$O3qb$`=AYHPCKho
z&Q%<B<dOFuJ*6b)Dw_WN&7|nN_AXA+_vCGVCl~p;-+uYRhDD$FuA-~=gONLIJ>T~D
z`W-*&ch|DdeOEE<eXF+D%&2(XyEA-QjJiH#Te=hX`Qq@tE#~=2Bjf%I?b5(~17Eu5
zngjB_>-hQ8k58LE<ZaJ2R9)A$E#EUY&#HCFsnfke!@D|mc<hCyv%I)y>U`*djb^N@
zIc~G>D*ko<fAxOq<R5+4@$V=1|F&(t4Zf>5Xz2-yZ{KC9pERoWhH=-=P3M39)OUN-
zYP;Td9aZZ`HlOYDtT6fdom*$_{Aa%PO=DcUm8}<3=YLGo-Ex1ownc5Gc&=ko_hn4;
zx$F5o2ETaqSDtMs+a5?iZO1G2I(yUtKM4OYXWEpb?_2J>iuH<TAG6nii~P8FXRVW4
zUjFWPzKG~CEiB!z`?I+!CHV}0)<atL5XAS*1x2sce|VX1D{_9JP_RqoN;Rt-y3mh{
zCw*A4Y~TqSJl9b7dsIPo?dta%i|@Gi?H<?oL1_A&L{oZG{aur&`CPkKtNXpywCK&5
zc+=RGTUPq6VbuF)wx+Aw{;`_1H<kLP(Ud=c?us=({`Ba$lDw<vt!#7ZY5%C0PYO+q
zqg!m~ch`#tl;m~e-`8({MdcPH`KY*R-Th}@(Xk{ii`bJb7<b=S_IGu3r{{ObT}9RN
z=~=#b=livtm$RiI3v%3c{nS(HwtL0bg&lT2yXv<`Z1sCJsW5K+C!3UU(QzHWIpO>=
zKQ0Qz4Gn%g@RYkBxM8BFiniU~mX)W)MSopBWd0X^hUw}-*YElyO-I#kLK|kx1WCQO
zkrW@(@2r}p@A|nh<PTccvi1AJFMR&T&d&_?T}4-Z?=(!#HZN9fIq>igJ>97K&PdYv
zV!HFt?1l1szr+7_-u-pY`^F2yP=zF%U}FCfFHAmQtmi7uf8e0vkM4cPcNKN}_b~H0
zuEEIhPt=&_xr*64$a>yEc14e2_8b;}?)2erkN$eK=L)+1yJAu0x9xsIm^`N6nYNXF
zQ2UB|PSsTYAbI0{HzluWx??XNc*|YgO8pGe(Sz>ai_*Q{mNvbi1J3_;z@B4#S8%^}
zOWXEn_r9NDriaW9+-(PO#{;VQ7x<gTTz&fnPdBQ{L%zy?+Wpb<mag#&4gY!8>i4f&
zxYoBFw{LsXwU@rT%6AoA|J^Uw$H(iMzp&HaXRY^jA%4MDO#5EO$VzigeBy?MUed^Z
zrY#t@f96_z*!@M_?ct^6jqmC)bBX6VM!hdlXz8}B{^OrJ&Hcd>QMaAQ($l@(ZTs&p
zN0smVDSzK_`-j(fvM8(nq|&=~4^Nc0)mPnax~%0|&vmRccgd)E{a1N$QJ3BP<Lke@
zGN8s<FKvul9^L(7jqj>l^vJAI&oth$>y`U`RAGUq8~56=`8l(;Ebv@IUHyJsdm?|G
z=XctA_+DE)*RkKf4j%vT$?Ltih+cxw)Omp*e%bf(p@F3CPn-7lNzr$GoP_+%K3I4`
zS=k29Hq_nfY%8CtJPn3CaQm_*n>^RC$J<W~Zn$!z=PLFYUHae=BR6=i;*znuwyWo_
z_guvnzwBIX`6p{V*AYK#3T6FhzXw0tvS89=-&I_9{mv(yzy4ES7G?FMn*N0LjoZAx
ze!LeIRr!!_%6p^luD!O*a|I6@@yUoYhLw4);>n-pnpPaQ&T|!Yzt06d>`@gvw_djM
z;sd_*R56IJsrONX_$B>)*JCegTYpA{oGDDYJW2DDedl4o-|H^C<lg6fS#)iGR+Lxu
z{fDUeCSA1Cl8Fb;@zTbs-6}3P_5Y@Mu4CB!NL%7v*Rv@9Ntc~4dc@(gJyR(9$GGoM
z1?3m{Bpq%%?A6@|+_2Ge4Ids;tI51A>pj=8^y8+*pH*1zxrX<T{CdS*H<x*`n6!Oo
z{C=&0px=kK?fabSP3Dyz`MvKd#(kGMt^B6%?}_xsU2|*kLq(r^aq-ty9z8T==hh?7
zxpRSM8|r?~Ai74(rj(rr7J7QK+Wrsqw12lWZT|n!#P9!S`u^ykhU+SIKXb0H8)e^r
zyXw*&FaG_ORlX?<((-YOi{I+^%~DSn2Jtcdy-B?u4;|ZkzL#OTdI+oMPT$vg<g;f=
zy{K4a*7!Y+Jb$$(i&4*OlkUFRLn@COzuc3>{@*V@WuNnwd8&xM>USFa`UfYL`lb**
zVaV^wmsb?^V`?9zQ0a>Lb2TjMZ%n0kRewQp%f3TR`XG97zo$J84&wiLmkQJWJZie{
zDw_Ts)2Q_0+K1`<uCAvc_xqRinKt>DRen-vx;~(9%RZs19Cz<Mu-(4*&GTHx*S9>s
z^_V4#Jkv<Mr-gMEH{5j6-ld*1SiN$cJ8u5ot78b0ss~s5*S6QpsCeDGGkjOkRlSRH
zhyAX5P<&W_x%#yH(9`!`@4Jq=-it|#&epSkT=mk+Ykk+S+C{A=4C%7kPYQ$l_2Vbr
zmK!p4gYP=3zCRVD7v0}Yts92X)m@H^UZ#P6YapoKl>MHxHsik674`m%m@cW`r$}0#
zuJXJ4-FI3|Ibx%iG`gy<Ozunr|E+=PkFEY{{Ip4ahH`rNZyl=qUnu7%RPWQ-l27-$
zPUVH3X!C=Xp0N1#U6%T`peSG5ur2Sj#+5_7cV}z3?ZrbEuDp4)ryJwifh+kZ&7S)B
zA17{?SH-l;IcwCc2BVvmd9LE6^VXKuyKt3PDA=O(#O13mS>m~ZvgZqJNq@w&dykm;
z@e0p%bbTBY_1tV&y10JS^?5?YlFugXvvQ&5D#l%BjGAB7PF7s?K*#%!pXr%K<ePEC
zf&IVzVxzB$7i?Fn=r2u+^0N3w#ZCQw7+RDEkq3NyQlBOr?p*1sB6L*o-S(Z%-E`j@
zUWRgdP}R4X&((MIg5+-a#<jCfT(Ze46fB<hX7{^ySmR}ARu8hjhY&Tr*xp<X$2~`<
ztK5^G7{A@u?Iw9?VOTl-FlXA7qwib(|LmQ2xE96P#t$IaP(<uX)YuCuL9s;LqgX)^
zQIlxIf`x-6mMGX`#i&FSF#3rG#n=_GSKI>%f<#4%bVZs7NIf7$@VmUsorIBLc4lX0
zXWsYzhns27JoB5fv$Ol|NqwuzudM0U-Prb)JqC?Qi`K0E%JMzhY2wE1jvkljs?1)g
z#$WJ<as6h^`80`5SKf2O*%hacPGVJG??bq%XV@b>#*TaWmBg32ntxe;e_;CkJKf(=
zZuv;fgm*fONOU#p_xzA=_(N~^8MWI7iLRnN^{CssUNJn8anK#Fs`D`2bkyIi&#;rL
z9_^QA)uMh^4S!8_r`xt}+jO&!65Fcl&yj~-zG3GrmZmD5d-&~F&wl9pG#d4EUgqjN
zuBiJ3TlJoJ%ieG9bJCb3Hofh_^Sk$HJ~4@1b+5;v<(FNjLXXe;owuUPi}O{yp5*<z
zWxl?{YWqAj@86#hIjX|8_XVc*RDQZjFKS;W)|<6`bMAR*<Kxfmk=oL%^(K|~KDWym
z->0@TYrUrVs_y@1vv1ptO0ueHJ)xIdI(bg=T>bA-UA5INSAQR}XuBV6w9O5RmQG8G
zR$KjK+fJ;1uL*wZ{$4G#{L^o}?zBxm{vvgyrsL($wf=I~D>q%57Nw?s$5yV_HAA00
zdxN1#UTtvAo-bBbk4>{`#UBoNZCIZ#Q)+o~DEj%RtNKG*-%HIO+v}_E_Y7>+=lbsz
z$o_tgX}s<4Z|mCIef_OF-|+JE)Kpt+egC57)l*ViigA2k_ugX<s`J;RTooK*)6<q-
zyH}g(Gm~=B*!l<ly83%%GP_0nZbCdKcOAa%@$LFOamnY2ZAHJ?Qq%WNY?WL7w~ote
z|7uWDD*9@#)4uA}cwic<7ImGqHJ)i_?F25r;*<k>-2F}xi|W2VhWu-%P<-ly5$B$@
zaNxXzx;h72J%jJ&$D*HizNg=go!&}g)f=zBV8*T$LsC;cebN^vU%JJ@#FlRVY@5Tn
zJ+&f<Rb{_FizOa^S149}YsPki);W8A;!16;bERs>_4jYU&GbDRmRs}t(4oKYC-blI
zor~^l{*N~jUyAXktK#3QwH4O&BeeXk-s{%2Ys$s;_q2+(SJXLz?RstfZqamHWG%OS
zqa)i)Tsbw-Rhj*~?_*uQ-omFhsF|L)Qq%A4<^7$x@A9#Jp9g+F`Q0A3-85lV;!0)X
zR~3(QqF6Q7a{~CC_4k6I=Bw#*QR_Ru8J-rcs&Q^R-PZf^vDmTAPi7DEB39kG{p_=E
zYX4s9TV*_5#oInlw(X~{^-|@x>G$4Tl~ehvn~uN#8Jjk5c-rgZ4w{z6s;2MP>6X(S
z58~gW2|es<op4<bo;T_CmuY@q=Yy<g&F8VE{T6GviTBk#@Z&}wrm?B(=QpnH+de<d
zJO2K*#f^6jJZx&3SMz=kz*hPF-X8My@vkpUsKs)y)t|Jp_rJOwu_Q6&P_FYp_jfA3
z@^4%9%6?s5o{+|_!|N>Q)8mC1sV}wlyQ|oL_nP-QU3B@^7hSXKxJSQDW7V;ZI&R<k
z-KlA;s@wnK{_@yAUj5z6Q$9&u>Eg>bKm4>ML(@>y*5Aqe)rEr|xNXj`G^=L)orl;z
zy5Z5A|L&bOH&0D`sjh#z`Lli(x7_pn^TdVkT|A>EjYZRT7ahFo;st5wsykkh|HR`y
zo_Xuni&NK%`a~UeW$kgle|KE!TE%)9HY4!Cir3C>zRS!sR(<8PeO3&)bW!R{`%bNR
z?pNLBq`Ip6T{hkPTb|c_qXx_8r{%6@v*4Gd_WWy~Jyv`=I<>93&-qOAr)_pQdh*!M
zQmS1JzZ@xUV$S-*krx~`e_<M{esIB4oo~Bgek#&2{eFw>a-v_Z!&&dcm@n?{6WH!Q
z@%zE9!>;GS0sHU&$=2PLrFm85r|s{W__Fs_ms9WA^3*R9su>5>xX?|{`d#U)`DEXb
z(e=wT|BK@XezUa4!X#F$KW)7~-Zyc6605rUy|ypAvHyMf(c>R|<fN8!lcLmg-iBK4
zf%)fMck{XHRJ&GH`(H{s&3g6JjjDaA${(cnYk&8dkG?Q54Mh*x>79T5z5dsUN{jy8
zknQ?yJ@@xz$MyL+?fZF7LC$OEjCE2}-M?p$^_)}Jt}MT7J<s}HP*#6s`S7~}U6t80
zjTif^`|eX7`%A;gi7$O&<>NmdH0zthm2R-&Zx4LfVP;ZlI&aqITW@gs?4)ST%GFi>
z%Iu2c+IQIZ?^I){5B*^4`@Til@wGm>s>k*BseRe+xXC^<I$S?3iA6E~UFq4Ej$h}p
zH^!&2Y1ZfTuD<h<)voOK7qOma%KR?+@Am1|Z|m<=hhoR~cc5(7*J1jm_kH~6{3KR&
z^<6R1Zr1C-M#~;sKA_LcG<KD}50#aJZa!PzKi93-R(_s8t^NGT7oS+a;|9O+_flh5
zU%$_^U61emxas;MJ1-4A{^pI3Xm!CG)6&@0*7u%5dyicAefYfjO~=2g+$kr_ytC;p
zi&NM7^zow__wTbf4MkO4LQB7K^wmQyJbGSYOKr6;l;5+O`eFNf1>k!87mqwRw)VU<
ztG<2u2@iLEd0`s6W^J#leBkd&WB+}vtMhl?c8hvWf_hzzFO}V1*9?93><xw{v8w7l
zuTayA`a7()>TO@~M#mQ~`6h`~J8yN-&Vy=CO<bvJ-o=(K`@21=c(I+c$M&mr;vI9-
ztcu^+WmVnx7q_i7>7ecR{3gw+*RC4!>2ptilZtA3Upjr;C)+)L<dih4=KXH)18p~*
zJmt}^(yS_b4@?!`;F>*OtgIfJ=2db0sKOr|wrZ=3R@JK9_P4Y8F225YwEF%51IO3z
zT&L2us`!cNs&fW0zwNNA?<c`-UVlx0=a;wMtmQlRy{6$c&o57GY1VqnEnlx+z5UVl
zgT78<SJU6$X8m55X}-MIk)q4n`aYV>j_u!*l$t%)->ddz-}GEEYq`AZwynR{SCk!F
z_k(#k7F|B}{WJX!`~JuG_@2^ruUdoddMyo2dp*46jCOZDp1Rghp8L1h`u;`DtEZ${
zRW(jb@4rGRx99HT=WX=iiZoWedY8d>{NE8jq_OJ2U-i79)AcLU*!7wnpP4ah<qxS!
zRr5)ezSC`M{=4(B#{cGu*ABU=a(ohtX8jINsQF~y2Z}BJ^e5JD^;FXViLS=}ygyd|
zRqI-4>8{_OE6RSO1?Rta{<gEyShe4yM_%`jDW9j*)j5PVf3o^h)ZbUR^881RJpSGJ
zX{`GA(Qn+e@!3n#Sk>0|)U%E=U-_=Pe9Tc>wtqHtt*+a%eWN4WOk6oNiA7!Uc;w@j
z7axAd;)I%wgR6ds&2Fjv{bJYWqps{<aOG!vG<daFqN}dQ+ie$~-@Q-siAk*Licf5R
z=UMbu_c^p}j;D5c{obYP4^B<hmOoRbj2r*L=W~)!HS70OLe017%w3lszQg<^R$Vdj
zzAs0AvMBMbw#K(;XUj)wCcM*OL_&4TLFDc#ob`S1*z?KWGqc6j)!(%c{nLF8U@O1q
zhbpZ5J=NIr%Xqv~_shfne#MN$x88Ta;k8d~vn=(c<A1Z~LtlNeER9`%dr;d;hKyL6
z%2d|Rp~Vk-{`{xw?6D$=U28tOGTH6!nj}``d0Mvfw7i^5=b!2GwTbUe-0A%TmnK<t
z)(2a)-!@&6P}j^MYyWlYGwkH5NBgCwnzbJ0D{TUAoY$f4^1EM5Y^yJ7FZAd)XLk1=
z6I*)3W<y?R_t5&4Ds|PqSl;W4_1wePcfeiO^Ju4u8@D@pTw<!O#&1@;r!Bp9uQt<X
zCPnM>CvS4*Hhb21`+M)U-h&YB{HjC07y2JIKXt8H`&T#Lh?j0X=9Np9C%URzkFWf?
z-^D<FUH{sSd+U|=M=VTY)x7sRp_D)A^Bb-@w^dD&S9SZ}c79iXx1#sr3kFPR|7}ug
zYU}r_w6pf_L)m&RYkR(&`s<g+Rn}d%(ze?6llDm9e+lfo^LuU1zWIlg>XyU*GS~UP
z?%$h1epMY8FX?*Ui939ohN{EsEa=nYg&8R|8wcIL3!oYwrs=kSS5fbjZSFsH_M+66
zzIX0l$BsH^K^nXE82Q&G=ifXpja4r`=cXqf?z|v%tsiZ)%?*o|PD@HnP3JGmkpKzU
zBarp`xuNAd>kmg>aM=8XscW@Iy0k21{kx;K^O=5!QY`1XH)MbB)K$FLo=|_^kkzkP
ze?I`fM~{E>k&{}^O+ry|e2C#iJAK`2y^klS<*FhN-LUK5i|G1eE5GjV#(d|;^GH+8
zd%h$--*z3owMp#{mke2+`qsBLJ*4+V{eDc#U7xLe@iCijTB|xXE%$yiZNIH`RM*a{
zy-xe8SL1=HZT(=yYv(uLWo8<?+U63gJ>79tD)}$`N5upEA6c2?)l%W%svX}Pa^V+;
zp0+g2tFHW`T@r|az&T&n%-{Bm1qpTS99;FKFYd1VFIBtZxooWAy!%I0|Em73UMT74
zzq~w(E{}HD)~tU|vQ+Zr{kvzXzay%f{<ke|yldcLQ`4eUHP3yeoBmzWqRZj?&_#Wp
zH*dYZ`q$O(BW&B<w(6Dry1YCgEn0o`i*7w{&Dd_xI%m&MV%I}v)_UoIFJ>pP>ap_=
z`)*;oSxKzA>4x<-{Y$6mNvWvs-)HI0TUq|t{+@}czq0o0#^3bH-}~)4Z&?z%+IsIN
zE58?aX?a%Ls(O{awbeFP_nLa+dX=GTR*qu6(<gm#@}*lWOo~>~FVk?|>p>{LYc9%<
z+uwNYq|v`wmc*vE)}v7DJT{`sv%TuqBw6P{?EC3;>U=zM$9Xdn+nSZ1t$etjk5$Ju
z>bQODcc-SY>h9}r-T8)>r>C)M?B6*ac4h5xzkhdJ8mnG+W7}Ky7&Io0RS(<olcqh7
z{W`U+w${hc?Nt44&gL_B`zpz*y6yR~^$!c4e7Yu8sjl5NGdA0-N5eU(Z=F51U#%1G
zn489?^}gNxrgeMHPGi-)=k~hgvwml`=zLFnTD8G<Lu!)P75o3#H1_+DyC=O|@56t6
zm-trQ^*+@6hYwh1=~IV(pTw?3uVc3B*Zsa@X!&<~x$E_Fnk-L3*S!4}>b%zVYvsa|
zf8Sy9w+Xcv4!C}a{bu`n<EHjRzjVWS&&PG^6Z5<JJDRcC6}f~Kw*9^%?QQea>yt*L
zDh`V;-~8~?mJCf}*JJ;9^>;5%`6SJ&s(JDD=_fqg`Q?SFZ&mqan*Prd7ruA#jG8o7
zZSm#U(Fff9T^g(UJ_od2U##EJgWs{8hsyljXxU@S2lSbl#-_gKt!kZ|(si#|gYJ4Q
zja3iq-h1ppb^e;jRM!5`-jDVDcQ9?&=lXtH){mjZ<Gagw)z;r_+48*Z8#P!yKaE9w
zuaCCt>pQjLxnFgglg6&k%szDH^;^tNW7Al_w_^HTt<cIXojfObuKssvDB7;!ei!XG
z@5eN&_N#Z-^y<?ps!K&pzsD3yxzp<YrC*nx71gF1+WtZ1&nYL&ytC;pi&NJMKaOhG
z{;#h+Jt-}9HT`?Xrscrbm-f15=(A^UFf@r>i~c*KdHbvA?=9G>|MSlG^xLu1TWM^%
zbNkt6-_-uS)R)@gE!w&JsVDEebhqj>Hofus3uf$EF(i#uW8+<vzj@b5l#lg(PW2p;
z>Divw#>NTbwrItBHsz{C^Jo3#K_Kt<;Ie)nOQh+0T@>q=g{$uWXR~kHjY?E1dtN8T
zo3>NjFN<NB|D_(^rS2IQOln!blB&mH`$k8$nYeOl0u{%T7|yybWS#$c^QrtP^>NSk
zkw8oYM*eZLS;P1KGKo!9>rCEs)BS*~9L7XmzgYh6Gar3nVyepl{AF6IfA34yUf=QJ
ze%mxGn_s5!;(m%@H-D!HTuu8O<$RA@S$(qaD9ZBNiZ>nSMVHfEuS3nB_x`Bp@}Y9g
z>Q~X@IB)$~+ga}UY^|5HvvvXxH2ZgY&#QX2uiCZM4y|@<RqM{4ZF{z={nNiy4XQS+
z+Pqc89X+eI{3*P9t6HsUow%Yym!4I<?>cg9QfKYfnMwjAKmsH{0wh2JBtQZrKmsH{
z0wh2JBtQZrKmsH{0wh2JBtQZrKmsH{0wh2JBtQZrKmsH{0wh2JBtQZrKmsH{0wh2J
zBtQZrKmsH{0wh2JBtQZrKmuhz;QUdSU)y2MyfkzzgRWo`Nq_`MfCNZ@1W14cNPq-L
zAPNGZ-#-@Z-^Igc?pQN^acWEFt-HxfGk!fSrN+%6ipfnUNq_`MfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@
z1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W3Rmf#sdf8anKi;i;&-KKl5Q&-VQ>iB%`=@#pKdt1~;%
z)sgS)e9nP?{xT^yl?}PK@9f%S%+2p6u2fbp<4J%7en~)Fmvq-R-}!&p0>+R436KB@
ztck!*E1R@l@3QqOspJsSI9StIWnAeHXn*d*>o#b$ETNJE#{uU7)`8GgI9KOTwCnbq
zk9^CUZ)ocX{aHJKS$|(S^Vqv<S5o8Vu=ZYNDhZGP36MbS1kN9I`L!M9%uCB%WA8`$
zPXZ)B0wh2JBv3j8#P@4u;aJW|W&Sh%=l7{gXXJ+2l6oIMu<bK9)UDLjdFyWS(u`kE
zOQ~)+ggIKWd$rq&j(u8RxG15LLn)4f>@$HmLnI)+{~t;?L{{{vcmkz<KTtejdQeIP
zLccF@b-jt@+#{=>rM~X5eQW;Re5hPXX^muC;vrD(?<27NrS)FQ#lvUrSTlZc?(<`g
zi_#hw>`(S5-@oJghoP+l@$3|f@pSv=maN$8$yZaT*nha6!?}!@E{2icRI{EhgqqK^
z-7IJ4JV6|f<sQbo4MlPPP-<H4<1DuJ7jv~`tt1d9fxPoO&Q{ZF5^zHR^GV$I=iR6A
zyLoH>Zl0Uoq*W3i0TLhq5+DH*Ac1lr5bARX`c>ZF8x-#=8MgC$i5$ssO@A-U^|&aP
z>uKoibluOk?B_j4bX~qw_9w16*|x75U$OQ}=<5dkBLNa10TPIV0RH|*xu~n(HJjSy
z_lM&6{!rHbjbql+OA;Uf5+H$C352?D!25nw{!X?Oe<vI3dO^QQfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@
z1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LpcDz%{=3Je$cSws0TLjAm<W{1-_tGk
zcE&U&=_?74011!)36KB@kN^o3O`zQGAK3nf*KN>hSrR&jqUQ<gCjk;50TLhq5+DH*
zAOR8}fxHApzO(Z=2mblX#J1-3i{(jx1W14cNPq<XOJI4YvxW|PWq2C8{?}vvj|517
z1W14cNPq-LfCS2qfbH`&U-rtcr`Tc=AOR8}0TLhq5+DH*AOR8}0TLhq67WkPl;^yD
zTg}=?fCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@
z1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-L
zfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@
z1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-LfCNZ@1W14cNPq-L
zfCNaO^a(8Qbk@*euMAJiMR!}#u}|v@7bWGQr9Xn%LJ}YW5+DH*AOR8}0TLhq5+DH*
zAOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}
z0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TTEnfd`uXJFe$dJ=<69+G@vEziw6Q&Yo?1
zwyOQpzf}#YHm%ycRmB}WtG4_pynCx!t!kaPqC=OSRlV;za%@uPmlh+5IkN3PyVUfV
zl={+h4;X&L%_I6JR5K2d3@f_ni9qMQFFUqR-}$L6_2f3IApsJ|LBRBS&Fik&bywt;
zqw9hwy4PV9XI@k>KB0!qA@5+>@QAHCp1tMFR9ExX%kp6mSaQ^Uolf0naB5q_XdS&E
z0TLhq5)cV+Ul;biPV67C{}8ryuA{Pp`luv80wf?KVEaCz)a}VQGM)rTz?{G-ZMLeo
z?!i^5shZdPb4hXkR<7Z^978E@`aFf@{JuY8n|CPVkA4(Opz;3my1z8HPPHp7s&%^0
zeQf8CeSiO#=7Ue!=9n2tQCifvVZ9_!4g|JJx=xw4;j*OMRo2%#HGB4?g;h04xv7jt
zQSs$IKGFW7__N+(3B<mSr2lT-Ln2o*uEow~)=mN>KmsH{0whq%1aRNxN`15I&|=;m
zAOC5CN?rBsC+jOu0+$@rc6`s17pJCLD(^Amy)U%=98A|<yVhN={qdf8Nx7)6*;*>&
z8}0DLjrEW~hy*YXN?k)_UrIie`uJh{*8Y1ijAh+@8uH(=*B<Lk-@R6)Ol9|r&&}Gg
zPOq0YsZ{CdCyjXTy@m}cU8&4J#9!L~#>s6OuU9Fmti7su*Y8u=eos_p5B*=Rig`o*
znLytA&REN%|BLcJw)v;}J+7~G^cyuB0oT_p*Y;KWz-$zliv&o31l$qGd;OE`*BKw@
z@Ace`G+K5^AnWrhmp+uE?TuPpcJ-LPD--I^9LjMFu+1bu0wh2J?g-?)PeFOnL&VEe
z{2qDHVbOnvNq_`MfCNZ@1S%?udM?U(&3+#n8aLO^1!<q(dEz=<JOSNxr+AO)0SS-*
z36KB@ghQa{^U!jyUv_S-8z0Vyq9<!2Q10VAwD#M64_@`Yu<dly?-rn3-tUy!ZckD6
z)-+dr#&Mlrown(`G<{!n_1YOCf$}7vyAIjTU!K&<r`4wGvskWth_R(4KmsHX7lFL@
zl_;<Kd{4|D7d3hd0=D;+uI<6^qE#IKzQbADFW%o{e|OEx51)9}b#*FjYwYjqgm!M4
zwV#X5Cyp~!xF~;P9iOhoWl{a0s@L~=s`@UNZaVf=)MEZlTWIT#?eVOdKTIcqQX=5{
zd8NoswvK20jzrdczT3(A*8crf@DaxY!zA#(1XTNt(9=V^zliNVE${yBf5(!I|Kq(Q
zTbi|fHhi~f1M7EDYO^i0RU84==i099>(2k;2xqG~>pEnbFPj(4MFMsR_}b^#@s?Id
zfCPRbkas_+qUdW~a9t0cH&cuGzOn6d2i3S`x*Gzz`*`Hny{|CcH|v(uT@SL3W7)dw
zrdRA+)$OmW`E@-t9X~REPHD4M#dQy^N@LZX``&nb^G9n{=T+T)2t7Z>iK*Ip|I9U~
ztk3V5&+gxcmOYoq+uyS5gepF7y(~`xBtQZrKmsla$lg~9J>J#z-jucOeQNCaWt?1}
z6WX?KdOsp-KjZBXkoAl1yb;$4c$f7(eXN5hM-793?EJ_yUK}T}g>}zWnZGo?s~%fF
z=3PB}w>Pi9EboRu*5^FB`BdKO+KVN>?!Kz%{I0Lhs&P>2>8|;@wqNx4Wc@$C$Ktx5
zFYSNh<Tj1htMsj=`w-jfhHJjM_HDOMX0P%7^SZw@w@!7bsOfmpE$548sjPppbEu-@
zRnMPGC0%!XhnBxo=0Pm?F`}QY#)E7fE!B8e9BG&5kf-i5IORCtIEY~!6t%t=z26k;
z*L}_=mM=OS`@YOKKi&S$n%{RjY}KcGE+Xc4wT^6+be%G7!(~a)s(b%g^!49WfB3dr
zbiezmAMfSps;mBp<$B_({j|Gw0=C~<^lfkLtqU*J_w~dz0axpb%5PWoy0$B8m$(m$
zHJp{7j059IAY=l**YDWti}m;N@cZ1O4!HErnq_IZsca5rJwFL0pRIG1qU?mi(bSL7
z=Z$Qf%i?vvpP_pWfc!<T`@YUgP>=4poUQz_e%6iWcQaQ1ZU*{Cm!oNZ)EAo;HEv{_
zH)?g+)noduOhm=HBvajUA6NIcrq?Ocav~?+;iATm?Rs^`Pto~ZuQRgw$@pRk6umFY
zTYs_iSbL}hie67$*B>f(`c+;8O!t|(<wUAD&otG%*B#&GeV;pJeeR{3&(^u4tNSgP
zUET2-TYgvL3-eZYoG^be-=`;cyWC~#Ni6ZM&jnoDFO~7E%PE%eRc?O9J|1K5uh{v=
z?tjta2m4W7wRPX(+fGruU7vr&ZeMr&#Xj%p|Ihao;&&4m-m2r-Th2^Njl-W0tpDJc
zVRb8QtF6D!qiaX@o<zCE<2|ofwW#Z-?sX6OLth_s@6ToVi`uWDUhsCUMU6jO^~(Cs
zcD(C;5bejhPs8|)T|-~zvicY5^Dl9{+nTSg#+@zyRCa9nFMBR7a`zpU`Bkp*zWK90
z5(tTa?dwxWZKE%80;=m^-gJ4*Od$ah2$evo?E`K9o=LQ4>wY1xoxJO{?ed|G@7Tw6
ztmjeiJL`E#sQGN~lVh_N+I)sTw&%6&buh0zw3A8#Bv8HtLcK1U`jvOT%ksmIxOqhX
zq}({<y`Hjs#hDjXj8F1%a5e7AcZFZQ<)-VZt8#hAi>mzUtuH<l>-%i*TRf+VEvy<Z
zy6MG_D0)BwBoN!Vmur7<J{qgqdXF?VJF=M^YCp>SWjqO#0s&X|Jw@4deciEb-}ZBu
zqV2gl-=N*{JhzOc-;25y=-$s~&7ZeDq3-AM`W4G~%-hbQ%g4f5=4Vm+AJm(7A7H!x
z^R6eZ%iC_Ru07Z7qy6$AkaymPQeL*MmTP>ejowhkSzbR(&tF6BPwXFJRXoRyU3K>l
zvi!EjnasZLcw6nE9TKofpj7rHw$^<W+q~CZS$S2xv8CsI-XM-kl|9vZp7s2&)boWh
zj>P_DIIQO+p^OvyK>{Q|0<jX%-A}}R9aQbZWa+W?ugY)PexLFCbvkvQ!D&_wvT-BE
z>&~~(@{8@U6;_Q?ReD}OWcF>{w~KaU^)bG*37G!w$@F<hDCM%Qqo(=t&VQD-`}g&H
z@iZMT<yuZ$U0wB)SP#P_U_!vv__S@;q-`u?kAUs|q`f%z1EKVry^(EZDHIN_{V=1$
zQY1hEBtQZpfpd>K;L<y5mZj8UIf(rthRgN*Le_4PPtoDnpOfbGzi9rfp9DyN1Y#%P
z`u;hq{n$Cu|6&Mi)atUU$Mjv9P|JfuF{7-aLUvzJ?(uo|@hGo)4-NTZ)7ZzyPR*V@
zX<=1O61%F#mn>a1-m>m*LeD3@KNwp$biOkGd~=pvKUL#AmUL5|zTRV$@pm16$w6($
z_dI!V>RVm6i}p#tgn;aNpEcfe9?7l)w&P8hlw+Bq_O<26nr$vU0>1W5rB+Ygkzo6J
zp|Yp?{)lP1>-E?4_fRa?_1YirnU}cYkazsYRzCJ~K=>caIRN~%t>U;dJ#SEz6YKFE
z&fCwv{bhY5KmsH{0;NSjbzc&Cy6ydk%wB2rdik{`>$x-Yt^W7i%WtHx#Uwxi<xZgJ
z>wt>CY@gyfUgcNTbk}}nwV%~r<|6?TAORVHyyv+nZ~7b-<+G~k^)2-C09*c<^6_0x
zHXdd1s^=oQ>9TrM@wWO&*N)1sSkkk$JL^6ulzc_=-?{IN$2WhpR<)^y!m-@^FnwQ?
z<%)T4o!sqm7q8pmI+~Zat@o6&?oVyy%X*!Z<r6tFOadf80!0xpJs$|YoUQdfubra$
zRaLL+_qb_)?Z3l!{dcof{m*m~AOR8}0TOUYKs-l?J?#3sWw!TCw$52>+0i{mwmn{4
zj`eb~Ej!;rd^xdTTZ<=9)c5I&Z#g|6f$$0BJx4PwA3hl~53_#1!PWaiW?r)t3B*mn
z)qHYg*L2-1m2wzue?GAOgJXu(t<0;g&+Tcy>y&95E=$P4j046yb<G59ony)DthrXk
zl7JrqrtjtHp37#<@2AzQrCbPv{@#(wKiPG|G`?K=-B&wJ@56lg$9lph5c_>Ip5Mi)
zrt>$Ha;Db-+vUQZW#M>dy5A0`1@t5Y0$J~;vgWh>o|5hLMP@IAz9`a%Sg*UW`<-=O
z<;_>**z!?#UOu6f_ra_b>d5*&La6ylp$~N05X$|It{<xNwb0XTpAYEt`I_?DdcS&2
zwi!nPBtQa15HLNLaeW^X+WxlO{BhmCuI<~pUKeF2)ct=}zg+jPYx}zU&mz`t*6ELc
z@Acnb%UDw>5s*DE2rVAWU9aP&_TitXruKBpp<dtXj_vx;4k`)cA)tHy8A^WJ=d5`c
zvMdQ?C7^r%Q*QZjoy^L!)boY5zob8I|IU4^d|l5&Ti1Kne$u{M0$I;Vb@SQw+SfkW
zw%;;)MV;5%s#kYBgqA<{eW0)T;Hn<%ON&z3yo@#87spungywgw`%?PN`;{N>nU`=J
zl-4*1ZGEIaaS<?mKab_?{++i{;~Cd1ptmGI0wh2Jp%V!0Jkpgvw$6Kf+3|Ib<*T09
z&zWQQU-x}R)BIW2J7422Z#^mwvh^S<2j(LI5+DH*Ac1luAiI7teyfgWZ#gps2Ulam
z*86WNJ9+nEw#&QXNxLNQ69L=fH5PkC-zU2seX;aIsr#wAFL#~pdYsAZV?Pk9y6zX+
z|7Vw)K9f>%(BmMRZ?VRkt`D)cmwuA~2^3F2bq<g>J#Pz5_r-bZWqH2@T>V~0*KX1G
z!$sAv%L(~)*N0p7+GCyRyVt6;t?<WGb=$%GgHPG!m>Ef_aWEY}S<CrdRasl81Y&vL
z-SoaRlyc>It`h3Fr(YyM0wh2JYbH=?`)Rhv?|o9a?sw4cHP0%4u|=KV6kV@B!mP<9
z0aw50%4#>1`{JFNJ$ur^s+vSsU0v^7_IWAV{_}zL9~?8RZly|X|9#?evFGc0Q;K6V
zgf{u=KU?+ao+n~ISn8_FC2xLl?B=!Ss=Vl@?eMur9dPNLHOo?8s{0;<?mi^z^Q+ME
z>2g7S+v9=uvJn9Pa?xy5n9DYS*suGx_Dx0EDe5|5dmPc8Ujprip7ZAx)%7c>ZaLWM
zRbTseSM{ju+TuhzehBzJf789siZ#EV-V47LU;7&0<CyhbcJ-LPE0fqbn9dl}`y<`@
zRM*#XPuJzi{3JjE_6THsFTq}ov{WR4t&*-&rfs+^DK{;WIqN0?k$~xQQ_)V*;k?hW
zbp6j;AItC8>C}A&r=f7r?SJMkmOv=uhJIw5$Hk5o`?Z(*c}c0a-~Nc9#UcsRJ>m7Y
zdhWkbwMtFb^CH{Ey3Gmr`kjlphifls`#IR!YqOtP)O%5^clZ%EkLaKH=3svWg=x_h
z*Sz~KSG3CQ9GC62dw#F``%YPYSL2&@OPfG!@0*DJmUi#3bwv}feV>unUeP>Qe^>-$
z`+?Zvi>8|OeZ;)?At-Np-d1$Etelun@9&9Z<>l)B+LztD*FV$pMXxWcf6W9;`!lq1
zS?8}T-<s)^eyog#ZhYyJVhc%t1W14cNPq-LfCP#p;Ol;w^{n}QjiS%NiX40YTX)eM
z%C-Lg?SoSOKhCRF8p?T0sC`i?>v-(*J(O{3>PM-JBeX;JoG)+wP}hfY@hi6H(xSf@
zw|O;df46Jh_1YirnV0z1tly>3&1bsq$;xHrDCRq*%~loHJ-8|@T1CH1!+C$tq1%t8
zmOs{gP2THaXytLAVLA_DE9Yt+>2d&fQ;l{0!0(;=-gtcTM{8Bbs;2!GO1Zq(ucFJF
zeorQpa<0}PUv^#X<8;?iS$?cDrJ}NS=8X?+KO*y|RQi$aSo?Qe@{VVg_d`H-KEU{j
zGcT$bpM=W6&*)$+BtQZo6W~5%?fVe&$;Jo!^^nJU%)Z6?_hW4P9eN*zy5B2xzw$m8
zF17Nic~aE(gj9Z+ru%+QZ@WJ6eFui0J!xT8P2!7#?f&#_&-Qh;D0{xS#af@~I$3Jv
zWb1?Yy{e+Z*S!4jiDz9`r!rRk^MUmr95bwLr7OjFjAj3ocORX1y@`dB%+I|1iYi~!
z_0V>`v93e(yV&bUhZghp`1ns7REDCVygya6AHLr6iLD;&E1<G{Vkq&p*Lj&e@Su`F
zNCb+0&X$)&QTtL?^~%<d(BfUSzbLzT?+=SAAKJK}KZ8%%=9n3YZ4P<YNtP!85+DH*
zAc2qx__`knnZf`1mX(L=b0XDw&Hwrw!~b3Ve$KYtto?7APxidXbv({*ZU2rH@52_Y
zG0Xw_NCG4fHUX@Yu_$P|-jBGp9}92ziR(Z3%hdICQC4rc#5dl5UiX*g)~OCnZJ$Td
zo||)Ca1CX?U|kW_wSTt16LZU6d#p2k_ga<IY#fSu4ri-g*Vj4M_HB=UnLTk{+X}np
z<J!LIxr8o1+v})nF0tBwpxM7k_q?iS`>I`A{kqjot!mxbvu)2-wSW4zszKGJRhze}
zxT9y)mOq7eZ&j;RtrJ&t=+d*Q_gzPhP3lx!@>a+5b{sq~ajkd!-^2qNyf7!VH837)
ze*3#uKmJwXOS=qtw*7bWW+kPjvlkq&!>Q*?Ph6|5emAv)@h4MLyP=f3Xs_Cnx-1`)
zX4RtoUe=&Vv&I8^rM9%__V}*<;=3Mc-s09#NmljkAL|Q?z(qgo-|v|>hNeX+XcSdH
z>;1X^(GQ`kY(Mh+#*f@|>5EI!QqfHg+WpW`jTfY)rg_KL@3-$hsmX*|mAaa@URn7@
zRsF6Rxa+E<D1~23myO>3kgJ!ars`^(!9QE->VDLgf3&l90$K0N*KQ@A)L8?o7G1sF
zH>s}1(?U0%%C7Ui;@z~Wc&mBW&*EFKx(B|-gR6S(`|RGfA0+dWQq$FY>A#_@=Ms7I
z{kKN5|G7RdyXO8woA2@9(!{r#(fDsETj$*WwW#+$SNl}ob`Kl6qR)1H>r|^$b}n6X
ze7l|6PkL<nqLf-*9Q0;u_S_GSJgaT<mbI(%YW5nLlLW#cQ1rfh(mO*Ibz65)Qtm1n
z5qbN^cX?a;G~afzo)5*E&-FNvjo+;Cdp%YAfipi_oKVd;WaaESpR4l<*mdo{%)YJn
zLQL)CJ(noDyzli~tWVajuHs#<w<`OwUZ*h6MRnCrzU=<xf%?rKd3=Lvnacc#<(#AF
zbD_NT+q(ad^`jWCn!ll?o4&6kwp$i%GpDBC#I9ebuC?gvVBY%kj&qck^@Fc?(|u4V
z<xJ;8(dAsvFJJbH=4h*a-S;z!&Y$)9Wa#;9ts}nHOV$$x0blEz?Rr9=k6Hb*HE)Zy
zled3e&j!<ds?4vf{mXoMznc?znf6=g<y_BK*|-fQzTDgIyC1~!$JoN5ogc;KPps$X
z^n3O1FvK=*MSpesQ@oD5x^C$D6?%T(*B8`h>$yYdcF^9~6zAo>)i)>cd3mXZZE-c7
zZ?@_e?btexh8@v9!bq1X^69F09~)~p^y^$+|5W=4^t<VPUu8e^baDSBhGksDe4&KJ
z@#4y_Sbv9TiwEXS(Q12snYJgE_nvh9RDCA~yj0HzTwNbk^@f)2`}dQw`p|Ds)wqz^
zLA>gFDynfLONafes%nRro^^Z_{T@T!`c?A}<G@s9`)l3!to@;D*R)->%lW=97wfa#
zo}%pKoe!$=S;wI&pYAx-&2MWzC)z2hU$fRL`irpaekGRpQa#5qZCBRu;3}W^e8W~)
z^jG|y#n|4LO@IBc^{Zui&ua%9_3hR5s-aDv8`1N#aqCt?#qVXzpERIWn;8|=GVL+^
z`1T{$`7wnaJ-xwzdWZdxLdAWPxF1ukr(!zRO_76Wx2Q1IlX6#E<6YN|>3SiS6UUdY
zuxh`pn=bC-T!n4TH&f2#KHo~c{ifrB<@^xPz5h_<_v7_{YcYM!SyZ|Idj9kORM&;j
z)(6plRe0L_4=t*_Y29jDD%;P-60bTp4L#k~exWEks`<wBeDC*X<rmBOjLc8n`x`O8
zYTfjeE_)6Wd%QRfeesly7smVfyLqv57T1YV4cmT?Bd<O2Jt9|OU+a)qkLmf5XvcS0
zJl`uiY^(jc<I7gPvDpdb{x@`PMcq%y>UGUacm5QeKlXhI{4aVw<{ei>`)|8`-Th%+
zexa3Dokug>?cXuYyYEK%qQ;dtK5YA8d%hQCujuQi?fQ%2XscelPZO(R-pf=xzZ1i%
zcB}4p#rK<Bg>|3j$oA8+c-!lhu02&hi0QiaUFVmLKht=zy$ECfEK~71lr=2!$y&}<
zKC~}3RkdH_>N>3Qhv{qH$D&`v>x1tw{GhH*K(^lbj$hp?{g5qx^=n>x3ETT(+FSkm
zOr_3a?W4<ms`xtySs1^2km)Y>Jlo~gy7j9~RaP(JWxvbF8ZYx(6)*m-Lk#DAuTJ;x
zO5*n-D%G`Xn&0*awRL~4doC;Hx7{vNd!k>lg>7B8v)U>8`Lu5RrsKkPxuU<Xg7&D+
zjl}y3-1pjAH$*$S{TW*R(C%B&&!MaC-<d3`Jz3Yq(DJz&mtsHYp8q1h?el2g_c8z6
z;9oZ_@3T&&FYUL-;;zSbuBcS$HTNIde2)j0CMq?(9*O%S+uTE2-{FsV-Etkay`Es+
z%KR4lHES4lP1W={gl;)q|L~oFa#dH?4{LV!TKn&Kbo)E&c$ei9>y!OnFY9>1byB6O
za|)zqtyktZ;$bhZX7wj)KHK|=y!K@4Y`MjY^FkKJeWov!`Qa<x_VwMjJy-o?%HP^`
z<%cythqY}#Z-1F`l9f~Sr|y1B{63N_Ec2(Rc-8rs>-3`99c#Vic0G07UfFwivUsdx
zy6c0O-_`gN&$DfZUHJ{Wrsp&)cW>JdlKF`#2b}xpD(;h9sp)fM-EtUzp{TfC#1?kF
zZp-ZF?LXOhN2$c;<&(9%tbek`tLnjZICj-tZ$r!P`@3uC@2vHz#+zzg)t&dMaVtyD
z+Mdw!**-UbJ=JqJ-F`xT-T8#^=UR3B%9~#{E<=mYdLALWe?vUR71Xqy7)MakelXp)
zhf>bge9-L&aJQwZ>xFK*>3FbRPM520eqAoi|48!|w~k6&anS8QF~4eGBuf|VFiZkP
z5>V|MOD)}WJt?w})>OA?J6O)`b){7Js@@k3HN8<)ziS5Wx+)QMO>;;mPIvzf+qU<q
zq1!{dMU8D<m_8R2+s&|D0<w9id+sN`-vTdj-l0<Li%?GHm+N@=K`pP}v1je?(61Mz
z=3mx+2{oUsb3$D^SVv7&x4yjj#p_4WVcB_;?ReAv@m0>%yi)m}cRYn!KGgfXqWp3_
z4rJq3Rj;pfnSYEgX8#%6eYG!tLwW9-)eqbAH?-@C?e@v+W#t`9KHYN-)BG}?w(rkE
z8<(d3=zgyxwEU*+4y9b|`wd^$LDsY8=Y6iuWo7HA7$3^`75&l;$2z`T^}jE>vFu-D
zeui=#DcTR)`>U+>Wc?CLd?@QD{Lo!DZRMB!j=rdP)pOHdq$|GLWBMH4v>f^~RxNrx
zvt7Th>#41J^0wdBx**!g>zA$ay4M%e^9)<%M2=!O>-CoTo^Ai#yjh8B4yNN&>{nZ1
z+s|KJ*|XJNUv_lY5#9W%>w#*0$$H+W8n3#3`_AvHzirj?i~Xmo?|c1wUn}$5*7rc3
z-}sT6E`4!HYFlM~BR=ms<oY@wvk(78jrAN$^jo!VAzf_0t8h{Gp{%!R(be01ljh~%
zYFvbFH}CgsRpnLR@wJui?Rw|vI;z?qm1??de#CYzrP?Qlo}P6dA8I~ve-K(&9OvCW
z*yM}@T7RC(6#LHEC$4PweY5eYN=?^4)T67|U&W@f{j%$LvHwlOru|&1<uGpR*6y_Z
z4tGyVbv5gEgpkklc}vmd@Vz77DjVmrbwjmo+ujGM>XW70_9JwAMgQ){R{gSZ7izq%
zcFXL@<|pHeCGh+0yH9E|p;o1-uH0<jeKt$#R2HSK=!JeCRkVNN_dr4kV?5M5XOn9u
zHXNVWR^QKqWc8t)R1(l3fcqF*D%-c(j=yl|lQ(odWMK-$eNtXM?6FI)tUGmD8oD0T
zw@=R_r+u5MblvXL9+^30VJfQbbBnxOW#wi2L&n$io7na1v?va;{>mGVwJEl9cKDn3
zyiRug_8lMUdAiE4qW8IZ>o4lN3M#IlrrYM_d!D(f57&jP*Ky=87tMQ}@>M?X-@Prm
zyzTemXix8bL)CjQp{Lt=Pt(+n>baflxmD56)lBQpdcCln5Buk=YWut(bbDfdiqC~a
z|Fd#b<qQ4!UsnG_Ua^G5evLJ3>-;cwJF@xqW|Ki(-??RBYFka$Y1wrxYrJW_VmWc$
z%N=&~o~-gcCda?CmW`{d@v8N(sC5VDfx4P|oH*_;Tld*T^$Y6N?HAMhzV5eO)#K|v
zJn!{Uw|`XY2Gcv9x8vY}i7O76hq0?GK6&lR_B-M}MK_$+ucFH<_d7ZFS-R_x?fkLN
zC)@tp=9|}^t^JcLJG$#Z==p8U_q_dHRC(L?6GgSdRlTbF%&g~hrukg8%l0}d+ROXC
zi0|^MevT#G)_UgKj;rgHFT1YpuSL7A)&tlrYX5DzJ{Dch_c|r3uV@Zp{TNT7sIPT^
z^{n~*-Mr^_MVBvXUCdjrt^Iu{^H}u5_WdQkk8||&1_SCH_CqSFcs-VdZLI@hdz%g1
z;?-l$+PG4t<KBF=PLqKRDt&3!e_h-DoaOZ^vueG=E|_reCJic8I{k}@4X^wAhLyV7
zs-oj(lk3&1%&NM-*ULH|O@BwKTdv<8i@P4%xuVjQ4!WV!n8jzUNOU#p?_hP?tJ`0?
zTul8D=c(<m?RBnbd$!icqU`v-KXz4LQT$BnwY8rq%8qHfLo4TuU(V-iuI7no*Y^4@
z+LPVqo5qXx=f1+?xv(rOw!i3b(f6Bq>(9DR4K<(b@u?b5p{=j<XU*qGp>mPEPva_H
zHJ(-Jrq_|Y<xGD^9Lw`JT|Z5Egi<cl-%q>tOE$mqUbkJ9m-%ZN@5;|m?dtB6WcyrM
zysmvQzj%KlhQ<7<ux>x7?1|~R_Eq_Jx#!s~x7Mv+Ez`d|P`~*jk8e;d(|cY!;HYn}
zu2*fU;&TF782bUdH;=IR`|sZKuDpDQam!Pv_<QCDN1oNTdCS_>qKdx<K5XcUKHK%J
zQ~kjoHrcbqL$&^+Vm^dnPv$@3MSo4hXunEj&sk*g;A^U(?lZIcRrL8T>;L)hq{zl+
zsPU%zfxP8(pD$(2uUc1JrHj{JhWGFH%o{^fT@G2-yHNARz7B@UHLG8y*MC{Lyz89n
z^3I;iWc=N*YM<^ZU3OnpRJ^Np7G+n((e}6&?-v<%`+Hm4d}Y`9qQ<Z7@d<mjpHszV
z4;(_%qQ;r6dS&w`y7-Ly2GzdB*7(Z%{K{5&TkR3;WZf5+T0YhIQKiSuLmUUDVbwlQ
zOt&>3LbapfXnWrTd#ZY4OSgRvV*0!R<z)BwvUrs9rT86+Yqi}kruOoFFFfz}(sh4V
zAm-2ddk0m%tmoRk^Vwc+qT9zR-b=>19jiv)AC~SjRQp6%>9+10i?S2_`1$wzcFoV#
z_zBgn>3+cW^+sk-mz&H^)_C3DBNm-swO<K6J(TxkP5p3nUz~Nk%JPZl0J>qBU03Jd
zvif|-yWSsWZI5n$g_2+Q{d(8DL$z;u-OXAql>J_){B`{bWqy^LAFi%9uq&>|uI?|y
zdVGgX-<wCdrq7M&`PsO2t9>ipcaL51UVdJ6ecmkEmz}q(;$_z#T)$+`^RmuA*paEs
zzUsYhSvlYFdDjQs`{ttZtNJ7KblEr$JzjU8o;SbleoL0$bY9?kADWt;JIKnJ^08em
z>pEU8`E2v0Jrc-5!1a13^TGChN!;hT+OLT1DYf;<)p@mRyJEkHVb}Hhwx9QX8}P}y
zZ&Q^=Jt~#$SIQ+`<fG!s^xNOP`th$4a1i<FhFxC=VPEA}DEG4}KVwUm-7l)*UA0qH
zFVoF_FNgJ@XmvGiY}t+dIg#)FwZ&a!M?7CQ4d;E1U|L>v?jK9Kuipu%#)GSL>_=>?
z>v@^?zDHHw6>rgQtm|*+&n=YxMgE>S&b3VUhpKUBYB#iUuGb;g_VfN8s;Kg|uO}*d
z;<}1(Xyeh;AJu%7rGtZO#d_ph#q)kKEIUV0?LS33V%Rm$SnZph(}vbwvA<2jzT1`e
z`Gag7QN^3CYr5rp^Ng)N+s{*V?Zw86{*u6&2-G`glWQk79G{ems@88?>1$%C6X*JV
zO0;jQ-J%`U{Yu_+u|C^j<@t%?I|}7mUL5P8y%!|<mpAPDe$QEbWj>z8ao}p)iuL#o
zo8HHvoNs@{`c%&?LQl_oe_C|;yw~lb%NNbhw0_(Bo1*NwT33D9)jh9~<rmjA-Eh(K
zC~y6tuhT03T>o7@*w-Bw$S)grw&PKscwbpm82&)<yKGUXz5mdn+MCv`mMQYXj;R*K
zA@+Jrd5h)pu2Z3vH{Cz5+_UY!n>Q=b<q-Qkg8#bh&YEAGZ?T5+?jLQn+u8X}#`nF@
z4%t2-FDKLT@GGzCKF@QNU-kZl?7dsl_n=Wu#ZUIWh^u&&KY7zl-`_5}oUH$2iFY+#
zeA#t1?`3xLj&oW0tn01m^D5bOLKdIZpS<~G^{e9Z)|<DyX+PwxSGT-uyvgEC+a)W9
zdSvrS7B92sIv)O-s;l!_U4OFvonc?=w5@t@pH!6Q?O)k=k0oBUE{B?4^!=Kx`gPa;
z(DD~GZ}ZlhcitCWUYwWW^#so`Of~B{lq#R^bKJcBB63uP^NvGX<yC&VN_V|~%W6Ms
z`)sY-u#>f3Tl*#0$y%>zKC~~Zn${C~IdPrM8b-Us??TU?G@w?S85Px8wa4(|+mBr5
z$F$Tml=C$7Q)r6!{$kZw?tf%{W?lD8^O=sftmgo(@@4fqwEGlYe{Ai`WaEqRHT@=b
z{W>Lw^5Bql-^6@o-<`|4E{pk0uV1?5eD$lYU%L5i@5f?a|8)7O@?(73y8gPhqiT<7
zIyjiBYQJM@S5+?Sd5!OUuFfG<Jj8U}b<h<jT|e^XchwH1-Hh+wna)$OobBra?1|$j
zZy4u@MX73iiY?vsb;h-QTkE6lenvTebN)UW^T)Slou978xmdq$zZ9KcTvxM(#de1l
z&f1Q0%V(>f%8gs>?M1&gs_J*mz+G1*dG+_(cc0W`Laj<u&0Bxo@}}==gi_A+KE<_t
zTl*lH9gKJCnhChx53ae^;$!o^w^Vd_*?12<-t~GS+BY3(uI|%Ce?))thHdx1=&$el
zI$3?9Us=Pk@83lKUER<6vTOUkH|sdi%?CeCxrp`Qd^hiP#MXS5eFtXbt5fT|^X7Mn
zN<}|h?<;KENB_pErsFbeIdS}E4QI6<YChZJ%(uNzxyI&K*70sDAFdCvs_s747B`h0
z-}A3j>QiyGoo;)55ckb_!@lOFimNX_veu*WS2sOtz07C!9;1v~-usfQ<#oqH-g)c2
ze9q_T!CXK7{az3HOEzv0FXJrUC(6RG6RTz&|DonXKW9}}=cT^v+TL&YvKP$yV7SiN
zK5xl7US;`g`GfD!6|IL2UD0Q|zICc&SMhhfv4vyj5C3E1F8W*a_YO_#&-(jySw7i5
zs^dQ%{@t8?`lm8gUB`6Kc|<>Azkb30aeJ?K-MhW|CRtU+P4#^jr2nJd4>#TU(3B*v
z%KV5W9{fYyN2~mj-6us8uh{pA^F6dM#*b{BLpoHpZ*&zO%RJQe)73c0+YhpF;3__^
zzpC=K?q^EfPBiQ7zt7=(or~nH2jf$=FO|i+`ksnt7ygOrYQBqh#d>XrW#_B5<8Aw`
zYfrV$3O!x-`k6PsE?-~y*V+2VCU+jcIE_tZ@2Q%`+v1&foXE=C@=s+aw(GL!ujzS@
z>vEIc8M3I`x{H4CUZ?HfHHqU#_MWL2k8=)D#e8B|%%3+5|HW|{%JoF_%XJt$VpVZ|
z#u7HYj)YRqH|JR9fyz(c+-%iXD(3~_{hF_^YFwz&#qa&f?z2tfW#30u`DH8J);KN7
zPAK~aQ$HTR;N+R7O#dW_Rb8Dw`?4#0Uh6B~);tcyj<4&Ls=c<)<y7`m{vaLqi?QjT
z8#;|yeAbGjR5b7Xtnc#T{dsI*)9Yg><zkz!vHFX3(wF+$XSl8h?^P77V!x@bo3_)P
zbN2jQR$up3uFuJB+xK;TfO^Dn?K|vhzfshBVe7h%e(-hQ;i?|n>!5FYwx4$uWiRh{
z5Pg@&_;;nI=bA;A6W>pgg^PZ!BlC;#yB|8L@q)A{4n>a(+x5HN7rVCaYX9TQuCM#?
ztk;29-aAnBk8B^Am4j|RTjN}}UR(KH?O%P{^>ttDsvh6GY}aS|`c#y?qV^Yg>$TmV
zDto5ukIHVS>3M&z5lVSm=R30NY`Mgnj(=4-oG-fGkA>DxDlV$@Ywka^`5q50O?;_p
zAF7+KYG2lLj2~)~W{n5-O2I9Z_1Cu_uJ)^0?Sj86wRJr!%1*nT+E03H`l6JYkAti6
zrLr5^byU|M-_M0?)fe0QPNn9r>-~jl-o=(~dQRxOov!<}X!}K7=XLAVUB_klb<fd4
zJ%7>l%hfp3osTL%bnWTp*X5?0U)LWozv;QBXvbAp)qlF_uKWz$?(kFZefYbJMx>>#
zy5qrDe$)L&DCM%AH^iDxcKr?|K9q5g*ALg@Bkww_d;Kld{CPPRUA|QNh3%^8H?iy2
zY0(^X$9L9sQ?zUAyv~&!-REqk`CUJcb!A`d&r%D!8dtvThO)mi^~2Zwi|=}5{hxRK
z=$-?Hp5Jubu-qffTiiM-(dCf!Iv#pH({WvNyG;G|UC#D?BJa8<D=%AT^Tw-wUtzir
z45eJG&y#JBOV~rdAnZ!>j!Rj2+5ABKL$mu_{^F<WRr=E8>tsNW-)vB+(glm_?(oq?
z>sG3?RYk|oCfBQ1DO26|Hna9`-s^?#^*L*P**M4=ubTgu4>DENqiTnWzsf$+F}_1l
z_~lw%w^!Ehp~ma-$U3iG_hZraRr6FgUF-+dIj5{2RO>_DbkR@SVc+|8*Y){6Ke1Jx
zDHq%0LA2-lJ$~EuiS4l!maPY|#H;Qb#dKTTU?=aqft{?X`+i_3`Ay#&3#D8r98LXD
z?F(F`XC2qN`E0LGdFQcddENb<@BHHW={hWaH%zQgOph%ryZ^8?p7ZY0W#z^BDgIpr
zUt!S?RanM5mU!L!e;E&3<4f0$@AXPHeq)PQU5{n`99q2V_cv+3_M|S$$E4&C1_#;v
zQ>}xh>yxZLS-k1_Ou3YkaTDXkeW`9(w8wCH5>P!CG))(2nyxRZa=QCdRsO8|a<P53
z!n*UgsQjkmg5~aQ`#~~4DVl?=@oH+v^c>N4IrKwp=QyIjS=X(+`Aqv;RZg|e%hFYL
zeW%O(%o?xjuWDb<bhG!Sb-C%DOUd%<_P_1?s{N5^x-QRH@|!+)VY#2bx0sbf*7Z}B
zPn<7V>o?73dM+mVDGSTi5yp2sZ^yv{lh`@P`aN&F>U^K+c7C^Q+E1p}Khyd`DQCLQ
z#$L|X`L?TmHoeZt%K7fkqR!E5*K4{z52ak*^NG;Qo386w%VB<J&2O4dwa-?i!%y29
z%KMsm{fKqn>U-U`U7xk<aHN0d+1I|`_Pj3@d#3AgDCKnf0ncBu>Yy7sjahuwio}<S
z?FogmsUNEAa_H&4)~BrXgudUj<)7(wR-8Y&Vb|kP*B@K?b-!QWJO9zs8w{v-*bk|x
z=H35h{r#e?_W@+*5U%61t~0WHrSf-xWbF$*-u3t>Rr|KqQC&N(cxK(dmP@{(&l64e
zD=4Qsu9%+$bO_kGuIbddw!E&+8%_7`;`@zR!=~$NR{Of=4vnh%T{CdkRf%oY^`qY&
zi@P4%xuP<ws=m*bH(mB!H(7jEzmd<>zpVXadOwzRoY~H|wz;1t?ZJ^}wQb(AcC~BG
z<6}9NwRK-yw4J=yNmbvwy6)MwD_%#lp8sM$8;XkUuocdG-V|H;y!WxT$3xzEnzy`a
zp8HC-J@0+l6UTqn^Vd-F`RZShlkafW^LX3y+ti+~e$lPR*SvCFkFD!xQFd(K=g?m5
z2hRL#abn8B_qna^yp`oQU7unrCu_fHyy-XzrJSwxH7h@B?+5;{f4^tm7@B6a=W3s2
z`~E&FM=@X4`&rldM88bKdAY{6F45ngf6totzC5(@m~UC<f$#l;?fPuJFDKf`$`ASS
zuJ?J*Q*_JQ9=E#nGryhRtCC|F9Am$px*A7VA4^?z`(Kv_uInn*wOiCVQmNK!>;Bf2
z9o6`;ovu4CipuYbw{N?q^UAlMrB<Kqd6U(i?s^|;epmYs-*#=^R~2n9>o^QGU)J;5
z!t>?){dwN=hfvDLdR;KR?uJ$l^FX{#6ul0ie$g-8aNc!Fw>~j{DD$jnKSVB}g)tAP
zde1%c@+x{<hE~67UgUjlX<A<8muY%Y_y4x)jdlM5zj2+7Rnd-EHShi;l=7kNYqR=e
zx*nRAQ>~M>)}^BD$j)KVe%bHQdJI3l{m6BGOp8|E?~U88&-K0r{TPepy}sprpVIXA
zsG*e0x}JuXPkhhERygl?&@Hd~d-Blo+y0KSYkRieTPw<5DEmYBVLJbc;^(T~SnpR<
ze&^kHiu;MY_FR`&@vxOHaw*lYtNEs~TNKZ{`_iJz+Zx}dcJlU{>-E96{i4T%ZvC=z
z0>=A!&sn#>Ld)-ae#Kf}-gOe?ZJ$&5vgdm}F+Cq<xvE81Z}&}Vivz}O(HiTzVaxBL
z`Lq7=AmHo#S){3Zo|`wnSidZs^_*1qTuyaAEvBpdkojNg@wV32qU`vZZ(==SdsX8r
zYkJ;!X{&sxk9)K)@B4$Xmj?$^HT{k<%4PkYa@Ksd#=of@({n^sIotQCzWy$ZcpbJK
zR&mo!$2<_%cZPrdevNAm;4f3#^FXvGtG8U@ZS}A1>!K~cLboICGh+>hGGAotzia<O
zA7?WET)(#o`_FIu$W52NxFmI@w$5*hvg7-C0qa}+d8fE9h%M}Ee{XvowY_gB6?+df
z`#0&HSM_XPwQH-LTJ7Ab)}1}u_H0%Ar+=#&RBc+dd8>*$dRA@uQ+W4QwOZ9WaYctN
zJ*#@(b>!Hjj_rPu+8gtmi)Np9!<<B#K{Mf^d*?5n(Q``TOIz>K{jGJkot%`KJ~8W^
zJ3jp3iNv*Pb}zGQ64>{t=dSPG`ims1YMQs6y?_6uqqcAHMQTf}I~cgZKC`d5e16kK
zNmk9;oW0*4-sH?r`X}Y0S-HfXugevK1~=MwQCgIO-)*B`Ib+xl-=@0ykHZ^2^jxPo
zX{>tDdL8cC|MmH4(TaWuU1j4Vbie7J9RgYRF=9SDwg1w}nfH#qsP_}Ks$J<lPaOZd
z9vg32EmL@ucc1>)Z&tm3^qWIc+Y0}vb_sO;zS%Q3HD0GWHf?!L)5!zQUY4rVF0JCM
zneLa1^4N4AW!v8izV6oe@o_8CqSSVKLa}#yo%Z$j`gBNQs-d)le#A*&^EU^6Ua@#V
z;#%|4Y4r47ZT5ZqiBzSg_3YbVWS=S1m!!Vc)L)b{y*`NLU{6;U_8PK#izRatDmmB~
z2TMnOS2600&y!fy);W1rJG%3g`Lhu?f4lX2+~2!@YFo3_&Rj(k$b0^q*Z=)?*s;On
zKh-3$>SO1;c=?lm{U*t)GLBiFr-hO)*7p&*%>L~e=brR=N;Tskj$hs9dL1q~weg%=
zzez*aMo(RJL7hu#($G}oUsQPP3zKhp`H|(RD=n&Btd|6I2o!xCvHd*Fmpwjb*BNEZ
z8w-JP{r8#OcbBC}tor_`!(aUD`NfH@hSoo^-k-p4Ti2D)`i1_GfDQrtej|3h^VUYq
z8oo6@iA7`Q4gYOD$I}_J%o{%e*Y6LwvLDL%a9%%L@rli@>zQPF4(r;lqU_tcUbwas
z+c?hZujzY(S<A`BRo?X~mh!Upmur01amjps-m8p_bFA+(l$zhMjhnpw7WKP@*y}Co
zIw!uLT&m%s_~xzG)q4yoyLtUCm-4#r8QJ1;$M>z94c>TqqN}#*7wzc!p~^4X&l=uz
z*zw0L_;PuotDjul=8r4uE={O$bHF&+?!t!kI;Ec`v1#L%I)3!^$JJ>p`hDkak8bwX
zt7+`I$KH3`vSi`YNx3TK4fSUN7hcn0Nt+`lB;}^E{(rpTkS&{zo1BEIdA|b^dwE&E
zxr#S^J`+m0j*rgXx5bEelUQ|n%eQuJxBq>KOkMT6E4vH!JofvM`+t{aRabuK+V%Cl
zRarf{{)U>r{U+5tPWfs;;z}F-@%Ra&P9C1-)lm7x?w4$zB^&p##2?>!$BTZu)3Sv6
zGl$s6sV)CQ-ABv(s(<px*PeKPVxm%8d~Dmv`}^pk%FEWp*y8ctYhJZIF0$H-?Hm#Q
zs=havH9adQ<|6?TAOR8}0TLhq5+DH*Ac0aL5c_+k^uO5etCN2*{8jI}+Db2#b&l;I
z0UrcHd5-6!t*j$Ef!MxF9^3O6nZM%S5Az+)K0?-<Gq!PK+ut=)i8t0ZAJ_JLeWx<E
zdWsr>Vt?EAqp0@+Snt}ud$0QYMNH53cb-f+*`8<RZckjtLJgODKe7EJKmsH{0wh2J
zBtQZrKmsH{0wh2JBtQZr5EFr=hi<X^&I=c$v8k))W48M4%Hd}oI{SzQ)wcB8fw!;U
zZF24EShc3x>nDBLd_o$V>i#{lQ1h$){*SG3Pdlp<=-+?osO?*Pk-FB^Yh{Q8RwwZG
z)0d6DYyRBSR9Cmg5DAa~36KB@lqZ45$M3YxqGuMTDlJbsY&8jx011!)36KB@kN^pg
z011!)2*`e)8(X}q-}i$Pl>|zWfa&k<LMdnKIlgY+=Ka33>+-(8ry#3Omn-v^H-Rp*
ze|yHcCw-pkYI##*>q&qFNPq-LAT9!}cj^Aty4y}pvMO5l-p~#E-d=lFQZ9;ii+Vzx
zC12HFe|j2=Zr=3Z3tDeJIn`CM-3*g}O9G*tf5!fIgX#Y}cYOH66A3w(aky>tD`yP*
z;oFp|#{p}k=|1V@+6x>1s_WuJSF`^9=&^o{?!Dxa#R+ta1-<7r?J_edH$^{%s<*FL
zuXg=^exKS_ReNRU%CX1K`(oD(KYDjg5{quq;mp?#Za6J*rK)i-=E^$f-n`+oRHeOp
z58Cmvso$qrRaLLA^wW>td%boa{ws-9-)#2!|Lyka=V__vxGr~(xqR`QG&bF#f6b;%
z2QNuo>E7=TZ*t}*{S)7Mz<zVzxch_giLTn7zrNe8s?T=&WcF@pIrog4Pg#{x^*G3Q
z+KxYTRqp|Nx89&ScI|#n_v!5(tXCbYe!AVoweR}ff+SX5((sP&THZ1%iB+ffIQ5=4
zhYd|)*HGTC()GjD_a(Bfmz!Qcb<aIl%t?yYtp3XKop@E}hBtgXBcb}`(0-HZ9;bXY
zAc<9dpN~HC<)je@9X>Xt{><UTrcd5~$@mYFSatI^2Yz0$ctH}1%GO8M@h5&bX`dE-
zYm!j(zM0n?{OW*ZNmlJM`-;owH(iv3qL0j)|NNsXXD6|0tmkr~-@kie;5NO64^Ly$
zQ*PPy$j9Cvl$xsR^&&R=zd!Yq=P#~VnHHrgUVA@xe#@~_mnNa>OCR<4?Zbn=PfA_?
zeAIhex4Cs$qN}QYHcdag&xVz6zx`1fx|+7HsB-DL#~r%rn2KswT2#AOuUi7X?-$&*
zrlR66KMovz^1<U%+v?iyCOc1<w97i*q`p<wuA}c5bmO(Bj7nnHL;C&wp7mZ|nW|LA
z!FK#DUk_@w-+Pl1svQpXPagT&6Yoz<e5vj6KzlzEaP=Js(e817J9yi=H+`GNs$zS}
zB`ljKS$}Wb^x+5lH0d@a)zwxF)_?noA!8F;8rts@bp7da#h}5B_Fa_5s-gehN;b}I
z#b<rLoB4*n_}TM|6W<(c^)u~|011!)36KB@kN^p|A+Y{s!%rUc+i_`Zy7;j=?@dT&
zrlF~uUZPbJAOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TLhq5+DH*AOV{MOuu96
ztDMcY(iRDj011!)36KB@lmh|Z-!IdxZ>uHaZ~geBDQT(bKMrsB&~u&Uq<J-z?@a0X
zVf*j!mf9TMZ-*TlO#V|%B2!)7y7^u4bbUV<+ux_?aLK8S=iK^D8oTO#uQrtYq5qy#
z<=@*+UpD%#`Eyg(>S{fJ-4hOZcu~uP=O?l1wNGCD#AfrRCBAfe%eQuJxBq>KZN24|
z?tgm!wr>;LdgHcDs<yjsa*|amXTE#h|7|@#(N&D6P;|F<nl8Ncpsx~Btxkqq+4XPJ
zlcLntcN2=TbI#nI-u=TTUnH?=tlz`7b)Oc>JQn?Uuf?rLjQI7^)RyMG4@7z0`)XN!
zvAwGBU7sA@KN&G9G1W_IKi~BDH@`|^)ps_3>xa50k4|FO#~TjWvgx?VNv!%{_2dJ#
z955k?P4_*f^BX&CF(uJe)x6A`UeoRMlfG;|AuSbkeV^poeyRK&<JiX21z&e-{P?&P
zX{;LS_r65Gmma#s?mI7BklNOL8;tBTW%`mdbj`b-tICIRE|=Gj+v~Kizt^Wj5|yfW
z_)fR|T*SA%JHBt-Z1Bd@lUQ}>$nPpfo$+~MOI711R!;D{^)B7tT6f#Y3AH>psK&D@
z{jqaiy!^?(ev`zay3b8x&F}j<k+;4(Z*A19;al^Qa#Pv-n%?WhM^~(~adl`a^EYq&
z+`3n8He<jSX>2;=o~s6IG;w%BEr!DnORqR&@T`{;TdMp0?a=dMUl)t+{r>PKXMWN@
zp^^i~LA>)I?|#6vyy|?&G~Lu6lnebjVd`Hj?>Va8y9+%%w&yFNzq0odVu|nHf9a_0
zTYQn)QmhN1D&9XSO7lK1FfA{R+oHppH$C`*)|*dGOGQ=l$M$?!*lWn{EtbqpV%Ksz
z7ejkx=gPLn3GI;p36KB@kN^pg011#lDG<2hu6p%b_Wd$-t);-Kl-gi=&h5IKZN~dN
zaAt!C2F*-j*Oo7hUi4$PX-Vuly~nBdyg6)W603fG%swywz2k#vtUBh(I_KWJ;j}bX
z?fB^IeOrupH=(YbgYG)ovHNo$?!D*YG*&hJosKRi)BFooZnoZUTildpRnvM<?)Nv;
z`^}zX-cDlCz5^cIWZ9SxlUTLkACI3f>g3^RtlE3tUP<$hrl+<Q<Iz;}K9`Zz>*{_-
zRv+WzBB1+wpxM>C{d&#=3zBlvCLcb%)oa5>q^|ULPYm3q*YM$K(Ry5ydcW&&^yoBp
zZE?xnEe758QX0A*^uUX6-S^C=X{l)AmpXp*^~cqzD-~DQx4-<uRl6KmQEgkrdVGgJ
zp5N?)Mgu-fW7Aq6%s*_zYeUo6)z!YqwcUFL>@{utreCMJ>U(^%J`yN`KrFxi5&blM
zFWt19?tOtQe-VAmI!S;8N{c|#hac?Iq}!C#mAd+!rEFc4#TV6Uw%cD+J4}D?z;b)F
z+4u1$Qd=B~>VMWt0wh2JBoH<M-M@c;{H|K{e^YMQc+-)ClhAZkyQYW!>$A~GtU9%3
z#Hu-`jZBJGSL4Q(-5#SB3>-LSenQR0!4`LwomjsY55Hr5|5Wtb^zSd|mQ#%f*Xes6
z)c({h=FUrF(c}Jh@V0et`Zjf?u6ffw36Ow$0!`0YwRgjc*-5PW&o4S2@M5E>No;!b
z9fNMX_LNbHFLghXSi{-~wC~gTk})^UPJHXyd}102lpBHGFPxt4I%Y&-s^8u*G^rZC
zDACp1kJ-QLA@9#lTx+@YC)-N`B;cRGiRVAp@$;+ar&K)-{(CCS8V`T5-7TA+x-g-Z
z7l$xMqdTv@>i6)8-#49<#;&pa9Z*-C-Hi}hwn^a9Q@S@g@3U!1ta{PC^B2$PIVC9-
zwe{YjFFX4__1yK{TYr(nrm?KgDnEVov#WX@>(}VsOD<WQ<kh@uVv`pJG`)VG4XR_+
zO+I;Q-=2HauZ~s!{^)u;?{<HKYS)@~oRoU`yHA<=-dV4FoI*ou?`<ts+;VKY;R#ib
zLuma-e@K7?N|(T|m$qH}+C4R?sg^F8)mv8o<dLsE@&3fbx32Dmd1#-}i*JAV)X8bl
zYTg2tBmoj2f$}0yZ_h4g4{K19WYzU9n0CZTqZcGpHyp}q^s%)du-%0X>vc*$O;j4&
zdB-Ljv|skl!L!p?^}V4R_PxFKtRz<5bicPATQPZQ8k@%YcMe3qpIlPsj@tb{OJdaq
z%NqTyf4@m-Z2JAc-aWtgd}<P_j(nuin0m*JN};$;*wSr=4SZ(#X}!~OQ(b;v{c7q}
zf9^gtDOz>?ww3?*Rqvc|RWds%TC=uin~%C(`~2}^6I+_KUfcVFto54aYk5u6$pg+_
zmS$DcdO|PPW%h5+IQOK_Q(X<cT}AuX@ZB38JLc|bY0+9V2i8vlBoH2f&6^&4LF>&Y
zr+GE&-vwhny}zSIUgV|oclSztyx3krtGn%U-zirtN@COfcG$7O<UiFU)Hpd{9IM`E
z4<)^(+v_KN*?dB3TfreVb^Uk5P3`yZzjW00Ext%?Yi#{fZvKkjRdgNxVd)iz44(CJ
zVq43t-(qX8?eksV_KNylWNiH$tG|tIIDg&^`_?3}XwAx%oz8gh>qJ*W@1K7h-teL4
zI?YL`t8=i`Z@09Zd&bSDtV(0i>3^-=@{dh_Ok>r~-#2^arpD`3o2spThaJ=3k+598
z-{Fv7HhvhVR1zS8;t8bd9(U-fV=AhBscDN{-JhBI6G}N(Ty*W4^0;Wd4tMSU`uwz1
z^oa{*);YZG*J<opck6=|ebQw~npaK7Mb>h<@1-GsR{N!rukT@PD*BwfD2YYKf4|}S
zli!$|#HvSadE@U^F8nTarKK`1*$xsQf$}9#-FwA>s|JlrsCpdAcf_#eBtQZrKmtAq
zjQPz)v(LL>P7<p=cFv2JKl#^hl33MeUx!fF@~xNM{l)WNrLk$B*FHJ(m0B~?Shd%+
zmtD8dp|et3ns+RnP-n?k_1B-C=GDA@u>9%-V*8Fg{f+bQ50q-Ztv;I=A^{R00TLjA
zVhEh~_w##PnT|<qYcZ_}x7Lk^-8-`N;)-hO+BwwvVE$nvUK^UmqOttmI^3B-&qyE~
z0(X4hy4m23rzgHNoR-s*7zyaUuWOor>B#RYMxF6_Lbc1m*O*)I$qCyH*x{ihR{hh?
z$8K=c(2o;eyh5+X_wVFw*LVJQ>-V_7cmI^?mxJ$f_0Z=L{UZUN1Wdo<BrDh9l2aSc
zx%Hb=RG)2PT_ivPBw&NU-tP}@a^@%f6W6NpbzJ{_X7}A?Y2r&&{)C!-;WZtWv^jD@
zqSA%EhV0&A$=sw=)b{<5YkPOz+NfE>x8^5Qawz3-AX_K*Z7{OWl<7;-teW+Gaa-$|
z>^lgJp1SISI+xU>uJy6sta|_GH;1IU`u5Y8jlOIC+|-uJ`p;K<)_yn57s~gW>Bl>F
zeE7o?i7yVm#;>d%)8{^=UJkta%)a9C`Arukv1%;up~(Dv|J30xe)jz0BzA3e*I!OZ
zXD&%%QH+188G%sy+1L9kzUm2;bGi6s>pq8e%+A4l?ZdLJf~NOBp_P-3$E+N5^XcAS
z%kt~iYb(F&b9vg&_WYyK(|fhq_wgsvSQX!6q?-Mm<-FWv=Z&u7ZLL4C*zt9)<hmYP
ze>W`e-;=dn{_*iUt+VKv#i=iKHEw-9&kJqe=8C5)yRv<pX}m2zeT^Si^{C!E&`r0+
zm3G1;;Oo4~RXwHh`x{s5K~}p(jZW9^`M9<p>+`X!<0902zRv$*t*5B@?5bYV-|2|{
zbh%>C;70o{O0%l&-}BaeFKOwaTkO8`!UbtomAzMFdS9x$AI$sRQ{8=l>~|ijcw6^p
z;`mX87j2%rci$E()6Hvc?6q6nTGiCmIjH!F>EiEr#ITs(cG&hgi)b&ezqgHk<&0rJ
ze4FZO-g>i^Z~E|qeVTNelG@U&^)ep`kN^pg013oIKzBbb%O6vl=_?78I)PB%<Mwra
zZtH$LuN~d<yiy;RvA0k5zG5u#y60h`=Fhs0<Q;FN_I)hzJS+5Y-oHOnbospR35HVM
z^gWsPTHJcXh+i*FseU=+UH4d?1W2It2o!Z+DEr-UEb+0vcLTrkK1WCS&}M}0`wyo1
zUAO!8WA^WQ$osRCSkx5{-RH{Xp5IkJ#Aes@{v_0Lv7S4K@0(-|yB=>BeBG__<KtGO
zv8ro6v~Nbh^gU6OGpqHNrLaE>Rbx3Hm-)H*n*%?uSiB&KRb}mR9skRb$`}$L0apa_
zKChOQcjd2ZyRq)~;P)*p=bmx%DXY?~>Y69*lK=^j014P7koA2QF`w@H8?L@DR4R6}
z_UaS6H#ua>BWqWiYSwy~ue1qd-Ph^nD{X4-T4(xv(04iA@eoS>{n||&TkkJ*s$<tr
zPv7-lFI`wsZL01@a5l?XzenIYUp8-<%QXSKC!(va{W7&L`)-9QUbntbzgLshuUOu{
zQu%4?JDj@xA4~pFf5(y6FW1knXy5F6`7&;{=ZmZFt=aaI_DFyP;vpbgzf|$@w2+>X
z011#lF$7G%9~pYNVra5f5+H&9C7}CUJe2&te((RktoZ-`5^!~%pt5W0z4HII#LfTC
zym$0Py`QL6?Mua0g!=o2tbVD^u|rQU>i6Tadd2Y}hE>mdN<CfVp&ORH4{j@7*B@K?
zi~cT<s(#h?=2hvc>y0Yi7H8Ta0TLhq5+DH*AOR8}0TLhq5+DH*AOR8}0TLhq5+DH*
zAOR8}0TLhq5+H$)2>k2o`a9IUe^Fvv$A7=!`jg+7o7hrQW}|P~;*jYdk4cJFQ~$+s
zCk#Ghiy2GnR-3A5k6{uZ0TLhq5+DH*Ac4{*p!;1fT^j#5yx~L7b()i^RM+mh!&?tp
zK4@S{B?lMdK==D#Mdi=>dmQGwX!dzG%t>r>u=RVkqU;nkZgtoDQ1h$4lfrbn-_<G?
zp0?L1SN8J$J}%UGO}|QmKq%i)DGi!z1qqM<36KB@ghk-N>d6OeIbcFUEr!GQ1AF)U
z;`6BqwLCbeW`9k$*H8Mg`Gho9omw+u)tu8tCNfpEKa_OU_ftYoANfe5G4+lcmAcYU
z`iXwHC2&)6<}rU7G9ZmrxA~~swa*_vHldOO$3gM)pw`WwZ1&OD)k#(@wf7YMZ^{iD
zZ#r^tQmT61-_P%HWjZD`Rkv3&T91XmI}csetI4acr@k~6F7(qbf$y()>y`K3{xYE!
z!{OH7ytsMGQBx8sIhc$CTYuNF)a}?^HEE3mNPq-LfCNZ@1W14cNPq-LfCNZ@1W14c
zNPq-LfCNZ@1W14cNPq<5CGbGAfBW9^s-Equc5StDt6f^vy0d57o~>&C^lw#zs!gjl
zZ&h(e&#End3h&;kR;yYkuISLEXI1aJjvSlRsW`ps{0VKxjZZ!~_^Lq<)t>qvMZBo<
zzMpY#z1ow~hc27<kA+)I`;Q)c-o;D4e&O2`igYoIc$xmG$^7m9`q0ee#ZBgYwc8h;
zr{jP3&555MIqyG;a!5!1-+y0w>1p*|Nko5bg;DQIPfq+`=V#wZVHfox9)4w2*oC4T
z?4ll|Z`}K`Rlh%Q@qc)Td{JJduqP^bqaIOV7v+#2@yL&OF(2YZ{qmtE&z)3zX^Qfq
z?*756zw6bsqPlPQ*{2=*$@2fSA9j%cuqR(`d(I^b|C27OAMxM^yD}B+qP-_SefNaL
z1M63d3V)#cwp?)f%ridw&$vc;v7gc3Dn&W?CGtVNXb(6+QC?R?zq6KyA1cK>5*78L
zztBFY$P0df4;1Z(isJxb)Q2$2BfRvzD{A-np>{R6LuKtleehRQ*hfAn^1<$3x;|Sg
zd1z&du$Yc8<`e9Q3XW(G6z$HcVmVwN#PaA5v7f{+@`?V5_GI<IA5(pQ(0=>v{@iBO
zG8O$m7=9ycTVY@1HRO$b``-Og-D=m0`elmiuuOHgSm-~D<4vl&@4e-NwW=}i#QB0S
z%7H(4VSXYV?Z7&NFv^MXNS7(rcPQ$W<wrc~Ls(SN4#HwRh(~$YMZBnDKIBKdEI-m^
z+F-{^uIjwc+yp8c52o?xKi&S3`5SsX*7MjD{e*QK{^L4?uuR4I40}*<{YQR`V{v>T
z9{z}S5QabSQ&i-GJ=r)E=MVD39_&EHaVWM=OozQtROE(sA-?WsJ^H?O?IzWx3T{}(
zWs3T+KN1z!A#j7A2<s~PALU?ARQM17#B_v_4`ER;eo>z|E->zJKQE@s!lFGfES7_x
z@JI9$=}?SkF(2yDRnZ@mNB{m|gPIGPO<9m2AGo_x>`O$IjZ?&<zo76}RJ2bPkNp7B
zF)qOy_OmL+kC-1^p}5W?3`IQbU_FMPqFq`0&>o~iF`h(4Iuz~4I26Y##tZy|Ut$>X
z@JFU7hxUs+5Qf4pj7#_{D*QB6(Vi-dcA)>n_QHQC(q)`r59umJe)J>84f+q|;UDrN
z9$`^2FJK4$`%?I$st@BqmJfC?&Jc!zgQ&18UKbIM^95PG=x3yh_EA3+`QRrM=}^&5
z)-cA0YCIzy^9^B9v7U(a!%wtFRHQ@2ba8(Q-cYeVgkcYgcu~P2G(~@lJj5{OCE~?#
zhIpBR2ih;ohj{n}#W+B^s91kckC>h{EcPSX2^H%>ewD&6RX#Bt<xsy&(LdmXc_QY=
zdW?AFN56>mBP{bnj7L~h)CYd>7x7S8Ii!mUzhMvMMZd+cSPtU@;|=-YFYH2*59LHf
zd0j>SpdQgK>J|G9?GaVR3;D!!)GJfuN4gkBJlZYRgE0J8t@lXB^$_bZ$`ASX`6|Ll
zmyLh;i~I<~FDT;0d>Cg)M;Q6geo>Jh@nSzC9^4RyisKODKpY>kFzU~%@CSCl3o7zQ
zJoq4tenmeaEVdK%#iD2j_@F-!7CC?i+J`XC)jK}-eBJBXZCEW+%!^PJ<3U%!0sO!R
ze8C6p23JurzW%S@cc<UleZ6Xgaesk%h<S%F>Xj+xJ^X}!P}o5j{SU={6YUZG1~1r0
zIfPLTiguwK?4g`Ykso%T-~&5Q)DJ~F5Qc&e%AtOwqaLJZRoH=j*h6`6KsxwCQC=)3
zhG7qW>MF()6!nYaBI|kwzfoS~j&!sG^@;V1@yIVK+7CrK!r+T|)Gt%iE2_*tF`sBx
zOh*{`vf4#CQAIllV?BU<DEbBIS+CD1k8)7NLlG}3{Dyz%4=DOeOvn7jd_jIF%0b0?
z#W3=r9{4S)j0^Iiyh_n-DDoqmb$nv}p}ZJIJQViDFyf)$C{yr~r6V5oh+)J-W$B1V
zeNdDW72FVy_F(-(JQV5RhVqDqA|3vqJXFkw`fz<hIh@Nwas5X=jAt<(;~jRv0r4mY
ze^4L7P{c!F2VvMj80n}Neu?`R#EXh@um?_<e^_@gf1pT5Sf-}u)v|K%SL6l1p-2~b
zp`EaUenmft=?H@}6!B2lfg&I3M;P^r^}!C};Xk;dT?mVc{7_uK5r(23F%16@kABCz
zLOkk)KTz0(B3@LKhr%ut`4I+BDENtrb8A&Tv{RK&Y&ZN-Dg0L{>K9dPmntmQ1OLER
zOc%q*hxsfj(xDg^xXvP8RMZ21ux=wC>VZAv6BYSH744ut--`ahxR<HuziAll1~-I7
zd*C3(>xO0a5f2sXgMX+GVR76b9SVEMkFc1Icqr^5pKHxpAN+$q2+PXjdhzc5*B^KC
zAGWIoSCoSyUZz;T#B|Irtb@phc;pwyG18&PCo0lqD%wRD{vfO>kNi+^9%5aEJ*10@
zde9E!M?WBpe4>iwkdAWj5AnzkMZ8MUPbde)cz|M^MLJZRPY5F&D)U3t4)}-i2!jLS
z#W3PgKNRs$ltViahRWIldnhNWSPu43FY?0<+J!LEVFzK%cPQdzs>&y>7w``%mKXCQ
z9V+^v8!wiZg~17Y5SA(IAU_oGG8O$tKY<U*A)iW-9}1qZ2metI>>(cc5D!H<>>|D>
zg&*)2{$U<qJwZC+kq=y82k}r<yTtOaqbeunLl}zsp@_$LlG#T-F&*v3ItUfx#rtbq
zhoHFrARo>d5D!Ius4O0SK;aMaLtzK`ptw&&7>fRe!Y=%VUE~)P=}<8p_GI-VKh80b
zAMv7M9-=<v7yD5a#(0DxKgNwXp2RTnp?=ICguzo((GJ3*!VXllBZiR=Tx8=w<cToK
z!!F`s5AnK+^BviEf<4ra{zW{vAPhx*nTr17yhY5f+W(>b;E6EWqf(43tT!q}ewm^?
zxQGfqP^5z^!mtAs$20iB4pfYXJ*0yV<^%GJiun#je%OP8I~4UG9bv35$Ok{r-^eG{
z1HVLj=vQ!q9nmhz!#?aFKjNXVi*gu$Q2hQ4>1Yr72mOU~*hRevqkfEg*g^Y|4~p_A
zC(8#OVtyGH#3R309`UGGY^Uf4@<UM$?SLX3VdMj6=*zEee0#kn>sEsgxFWx(NQYff
zkzb6Lh0&i-_zjhn%NjpKc&`a|W%Z$dMEfWw_K&zup*$3J#NSKkhQ)ehVfd+1_y?8Q
zgP*vLfg2R<%d4;lRr!zfy!GZSkMS;3a0^wz(NtCZuwKa2v|f~hU97XX4k8}?f-n^E
zP^^bCg?%X2ON4P<!}A&3?;u`Ol!w1aN4zV={KC8kN5n%B50%AZJw!ez@<S01MZBo6
zC&p(Di|d$JPE5}mhJB19DEbxS4f_zJi;8?0PslIEV_ae0!4Ad?>P2~|Xb17A59v_Y
zwXLWR{z4HC73)3ixXSTAR;^!+bklyrxCcKd#wAp&A6!%_mQ#h%5AXy2!4H{Y+(E^9
zzzK?U-0wrTIpW9p=MLQBKi```K2<wa>8Mwh59=(_kzY0*O#7SVe*S&1YYrHPS;s5#
zp`1)H9!38#UNC;3$PYz5m|sxT2fK*Je1v`6e<2^{C+v$V=0`aw>P0!!3qQqt$S*4L
z%i0b5D2M!r7t0}xc4HnQUMwf(Lw;Sw^$LpmU>{+zez9F*80`gLlt&nfd{D$gQ4WfA
z3t=eYaejgQ9MZ*h;M_-SCpaKpRM?Tl!*3}3`t1A8&mVo}ivN6f4&`AF`BjQ?$S1Z>
z?0=-o6!n7#xWNwU0~e^6j(o6(`p|yFLy;eL#Cla>_yNDMFM~ZO>PLN0q{E+IKYrYu
z-~FvY^|>wEUpDvtO{)(&vFU<N+wNSA^&ob|{1n@Pc<e`H>Bxt93KjDsjQmi<%M|uZ
zRc06U!*1658}LARD8>uM5ftMc^Ah<GM*X7VdIdWem*Tj_eJ$pNN?{KQKSYI}$Or$x
z3HcF^{Ae$@i@cByyO^&C!w&3=ige_Mf*0%~AHq=BL%%`cC(@y4ANCjM2PpXCJWj<Q
z<<M?~W$P5$2SvM}s2Al?KlV%L2b^<(gSh^|A8<tbVGs6^PgMAa_8<&D5Qg8P!XDz`
zC&I7`g*_<ZMTP%Rq@x@Z^&<>L{RpEyh?gn&p&d|FdGw1)VGnxN28Ta$){2EG6zxTQ
z$OlC_+9j?lqW@?o_@O)$;{)Rn`7oXk4@G^LKgbVzlizMLcH{Qz{O9|o@Dqys2#fy0
zKH3Zaq3}ag)Q|Syca`u5{vsdB!yn|s^P3_6e$IjX|Igl?NBee`g&jvGO+g3(h7uV>
zhENA0k)_h=i8YiVrDdpAvm7cF3{qDm1*;(9fPhk2U5W!02P()EC=rK7k$YsBdMHyt
zK@lWMF$)JoQV`oz=yLUb(#2Pw?)^K%o$~vK?|SCvecpG!d!KW2FY@Xe^-0#Jb1AO%
z>6fTK`N*aE)lagz?A9sIT8hV|IQN<R?M$Kf5!x^3!}H9(MEirzD>|QKbpNP}&ZR!+
zaoRWg>U^N*R9=32$DbSh^yfGJrF(#$>^)BI-01nCb?QqW`a=Dhv-;xDeB|X<@@t$q
zqxtcdhn*}>sekL2&w3iKj`^6Qdyn4t64I-$Ui)<J%#r`phx^cb#`DtiB@cb)Z(ZuN
zFM8`?NA}9}Xx-$A=4HO<p7R|}kL=<rTfaG7`pAFtfTup}(|6B#<%7?9);Hhzdj8Dk
zUUKI*{_34}C;E3Zoge!kzj-4)(yyican4HT1>gVr%bs@r)wj9+_viGbe$aZz`V+@a
zuHE;--tqg3-F(FJa}Mpt6212}rE^z0C-gm!&NDlCzf{k2ByY*jKB&i!jLwJtJ6CpO
zC!_sBdU^H3PM+vKZ65jDcjSrIY5n@IfBRzZd99y(_4B^tFQ2~kh4NT0nuF(-UcM=g
zy%g_$L+1v~3+YSt6V09*b8t`SbGqNn9ov87<(d5W*G^Vv;^gQ2*f;Uk$!?$d(R;ts
zyC>z*cc~v-ve!>N=W8unXRU1gs2{Z6{Zd}*l82r=ksrDT$Wxqi&QBbF>tolSI5Pjn
z(~DnX{mJ@3`icBDrSsf4diB<>W67T#o7cT#&g}e4_GurwuFCS!`yQajJ~#Z?antXi
z{rbC)mgs&mSMyq;_pkRQ+LxueC(7<UGVPCjTB3M<_S1Rrb58v#^Y8OhKXpoerGBt+
zmBml<VQ)Tqti5%}{L3l6e(lrwl)n@|<+l%$pZrVgdG{R1i~5`T;n#CP7LV;qAJtFK
zwLbWJ9_2wl4|>1x6UUD%FZ%N;_R_lf%Qvy{Wb4$Yxyg_8=2*(3zWi$`P9NBN+<Hs(
z+t(z^uP#4%$!HFx=K<w2FaA?r`e{z=@;OK9kjt*G{+>7I)O%@(lfUnL@s;&6^*QPF
zUGle%NfyVxM14+dJ^Q}&bAx-BU(XfUIUqYf=sd8?i}Ekg{Zh)K9$7we$shaN+6Qx8
zyAS4p@{mjO;pbkc-z1AK`I|dCeW^aOlS}iG*L>AmOYzR7IC|8N?|U-(J&xzy?=$y~
z|K5`xeGl7T>+A2A=<UDfNgT4bU;VM8yrug3bT9euzN}MS{~egQnWJ+pzxC=*A1I&u
zke_+W<Ij=Z@5o;t)yuEmcfI@#_xb2Y{@;J6FTXf)`)(b6){FK@KjQR(^|xOB=GlE@
z7mxH^hd$-kSM!>`{>+P?f0x=^=+&)UJv*}NN8i@7MEAfAo$kNV`SEijdOxA_(fzP5
z6V(&v{Xs_gyYJ4IKGhM&j<q`<&aFK3s88#y?7dE}PxU-E)wAnYJU#!~^@ZX}`$gaW
z>GjW!lbzn&Ja_c^QeXepV}1O{{N$M^u6BK^JNY>erT)aBI_5;4Xgz3MNH2fS9hu)m
z@%pfSGJDzi$gjUrzKQB${aU|v>(y87ji2Pk?N@dV+&jK!oO92!_q2JLqj-M$rAKkz
zht4B^R2Qv_z4xYh%73!hKFQ9>VX@EUl(&BRF~`dE6YaNold*C1(|+_^On&>-n~QZW
zQNL4P?C5-AzpJi#;w>NX=wp9=Y|Deq!G5?e+)t(NaPMidIQN@;?lV89@k8<AN_Ic@
ziCdz&`Y7F_^!%+${=MTrFQa+z^K-R6#aV~^?o)E@^vEtB_W7S=&n4dXqPw1Z%~#!i
z_gh~5#w&jOckggux-Zlxmz}fqO|<Tb=1{wQ=3GDfD~`YZOZN}D*Z8@g$t8b&WaM9&
zzBE_&VC|LNlU=X*Siij1S(zS}?EIGKd6}rrQPX_Nsjuqwx3pe%b)3(M{3hB*)JN&Q
z+LT@AB)bPEdhRB=*C$SS>EE{Z@JlYg=(f8!&5Pc1QQ8OVY`;@~)w^#!*X}=@ejefP
zoFISl#6JII{fMLY{5wD9WBolx;_1oqPqZ#^{OIK&>(9FD&ra|A+dbQS{E)vqQyu+?
zL-|Yj=*g&#yk+Cu^Ue)>&#nCO@uOd&bz%MG!`dtJ>wZ{I{nhQdPPXj%K<Af??hVgR
z_lqApy1&TSeAVyezgy?udx*aA)wA!H-Oo*Bb9HXWCA)bdKQj9L2btf=K3At|eREn%
z&+SpsIkIogt^Vjs=R`cN?bAIfu5^#f#~<rYu3g{O<KFPy;rzNM__KRn#mQs+?qBCk
z9&;2=_TEQ&e)^#=`IY>py!tM!7un^z{|g>_#pBPr)#`ue!~DB{(?0TN$F+KXxD;Qz
zICPGdX#dU2evr*`V)NI}dmN{E+9zxt{phcBUP|YXzx?u{IjBo6tydh1L;8|kKb7hA
z>3e(1BOcptW&JxR6ZxS$<fS<K%&&1>k9AM}jVCWr-{w>5!<_wo&i9df5bcXR`o`M%
zd0tCDU(j1WPI~L+hxDC4d-djH4*Df~U+4qnE7eE&#CgxS54>-XU)etO=lNC#*~QV5
zt&3mn?UUU;PjT{Fzx_2Qdh$eh)J5_76HiY*YU*#QYu)y7>c=|d!H2)-```KD*WPls
zaq`mpd&_10__rT&{hdeDS1G>F6+QO6*iSUi&V#;eT<z_XERMWH>q2v!;_7Gp{Pn4i
z%Jk;to}lM%PUPmNFZq@HTBovp>R&%H@<)34$f!TEd64Y`*?UKv|IU!!`^%rJ$>$!F
z$NSHJFZr&wJnH-}y2;&kFT3SSu6f%N?!Nn!tKR*L&wB2?cJKO-vw!2-@4Uw@`vZUR
zSzrA_FFbeGceejN(SL7;p6iDmd)@_i|N6hT`;H4={-DQPaqp4lRhnD%Yh~*;FZs-+
zdEEQhc<0_6Q6BrH9<n37eB#LL^qvFt*sX^wKY#B{=fk>DzKOn<*!`SYJv*A4e#ouc
zIkO{szxUnOs4h-*>CMZ%g_GVsAb<JHgKX~IANePWuU*~R=}YH`KkCPvCVIY0b>u_)
z)qLIeB|rH}`&@QTr>g9{pgET8OY9uTeNM>!oX`6K*~M>4^LF3ZH+rNmm*N`Fzj}G;
z&6SLko*m^w`X%yT+oyb<SF$`u#kF(LPbnVR$)$7U`7hlY?x)(l2b$NMJ16~APgakN
z;@I^==7;p+^zGhbxBe}+6Zq>1ILFSB^Fu~@^j>i8>G_w=AJ#5@x_{VFJlT0ib$tK$
z?)3cnPU4TAGtZ59ao$tjTS#9&b>&fqpFZ_dvRg0mlV3h{#G$%m>!TM>M*d`eSbg(~
zNBxVVN9SsZ&h3&PJAbrqsBiP^cS_eg<rRNaG$%Ai{kKnc<Yyn`!Rp0f?PT=xygB#x
zT=eMqoyf0w_)m8J_6hymsNd#s|Cl41PbnVFiJv<1>>dB_BitA4{P~fcL;jv`&p&_c
z{>rZ}<fjh5we<TH@#gB!G4!oZb?sN{n4kBFy!0izyrul|k>wFzvZJ{6?L0YW>^(o$
z@4su-r+9xK*!=Xvj{33B{JgK|-8=01QrA4><EKyY_L1zKB+Dm{I{Iaom)(5Km7TwH
z$zF<Am%ZzvSGUw>^Oy&{`O2pbx&G|ptgG}Ki?4t4^V1hQo-SXr`Y)}+Iwx72bAbC@
zzjJ`j3Aya^t{-vwA=hqw>}2b8e#Fu9b53fnAAk2k&!2c?uPkrRtNf*TnS*)J9~R9I
zo0lHh+mAlkQC!JBk^f|G{rZV7-MiisWcy3@=P<q(Cc4j2eD{;=d|>bG`c)REUv}rK
zbPkcd=bv3ZtUrCpPk*&fa{a9fYbQ_C2k!SiJ80b0pLLlx>X%$D-DlqSt$WaQ>^HuB
zcrK9LJSy{ZZaNS8#+xJZBbV$`eSNe}_3Z8yvb<9re*XO%dA%ptr}+9g2mHI<%AM<y
z?4ECa@_uPfXbxm~%>(JJ+s{klklzw}-aU6q^q%Q^$h@q_pX0ygs`Jl3@1?ifl~<qp
zr8?rRQ(v{yBmdIhx9Yih@B_c(0pIwM+wS`BcjWK-_bazAdFXNJyiRuOk@v9q#xH;5
z>z;J>`*-ARzx3wqJUb6$bLHoIK|cP@fA?P;J<9Jp2c6S?ABbaD*L=<MXJ32ko4?@J
zpSoMR7saFRCf`@`=$juqvYR8i55?(wqB*dW#j7{b{;-ReZ}0f;seG@pqxG7Dxm%As
z{H<FYe|oHc_43+B_0ajk>U&R;)sdg9-}Xg+qG&w_rSoAfevjwe|MKl``RGSobl1hd
zx4_?jUxl8>k6nJ=)i;0pU3UFGn7?~@FaK@<*?t_9_V1{v&x7vEQoQr$yjhp$h`n^4
z_e*^_FXV~t)wQ&q(tE)=k={D#t&g3I{7QA#_RCM4d}Q=J=XonVC+^?U?_aylsb2l$
z<6m}8`qjU6ikt3R=R%xwp}zBJ{q{q@-Y@dYgZflQKl0F*`sb&w`t4Vq@~Wdx>%gwl
z`%}KoiN2ijt5bjTrdLP5Wc#+1UmbqtjrLvLvhkJmsXux7AwANU`lKgMRG+^-%Ke_7
zDX#gKWb5XqU-{XsliqhNiu1hrUN;B#k#i!S{OZ$-_j99j<>xMW*-@SS()#vG_mO#c
zUY%!p>$48D{!J<Ge!1WE9#x$4a8&t^DzBf5?7Q<#-~G{t^&&mem#yPmIX|8o=g;$r
z{7}8M)Q@wI;;ow>nu~e(4ld={@<Wfl;5{F_{`V@|d2HvgpY!1PJLvh5$Ma5}sGfVr
zeMFB_ynDjEfz^}Ui~Vl&douL>qOU10{Y3p^=R%L{WTaPzzV|3SKkGI>cH~E{UuAwv
zl-IoFL4CJBcKONt<gK2JtxHDv`(CLmU+dDB@=oM0pLLZ}eD&riulKd@Y0s1INc8ie
zbJaYpvz1S+_lkY9?<Z@2oa2f5TT6K+fA*5UbBvQ;^@lC%XQF!gocyX6hqWJ6mKWvI
zC%t>reR8tS-+t>_2fBZ)x9ocPWAjumpFI509LXpT(xdMadF>OKA5MAL#p}1^*F5!`
zWbx=dh5r8iOYZ-;4}JBwf7ajl^CESv&v%@kXW9L{s}6nR^vRE%JhA!s$;a+}jOI&L
zujiOw^Vn~CR3GJMM|$L6d-eL`$By(k>E-cV(muq|qr9bgh^w6}52`N?`FkFnBYKp_
z`Qwl5SiL;@;HRE*p)YjKoipczU0(I*`IYimhk1%;mq&gw>YtrFtwWsr6XjdWcR%^z
zxBA{U{MxZojpk{;eSe$Nv_5k}eq`T)$ZqcDO0PayefzIZWOpvfOO$`2{wF)VI`yY_
zUhCg=inlJ5M?6+9pFHMT)=q9-`RK{~^=BXXohfwQ{QErSINeX?!fzr!b0eFZxvC=%
zfBR04&Ka4X{HP8)nZ2~{^43n4haVYR&pPR?Z!P<KkjhVa(+ki2!7u*99n>f46X~&i
z>r>yTU-PFIkMbUr`YnC8(34B{lAr#_$iL^3y>#DKU%7eJNAo4OK0V5(K0UTRS={8u
zUW!N03)1hGtvAW?*G`Y@WTY?6gWg<E7W<rgzC54p*ZK~N&f|XbxhI?teY-cvXnlU~
z@8?i?eD{bai)$Tm{Q0Se-ha>8``8t)yx^X@>P|Fw^C6p)yr}+KiYps$J!l@*V;((6
zr;BW!ryAW4rE}n1u+x{;Re$$ByZeVM4?lJ6I~m21d#=dN5z_ZLTB`5oV*h=Qf49-|
z<QzMP@|F5PcIzP5PJh^J9rKyUf3i>d#*sIrbA|TNIYIY^ebE=WWZy5XL%e&6TskKw
zik_D~cl7(E=heCFzL*QWx$~dcIDYg<-{(g?e(Lfo<>@|&o9z6ZJERv+Mtzm+D1M^z
zU%Pm}2XasGXRk~@<*R;LuX+>3$;Xf0^Go0Rm&|UR^zxATS)aL+ae8j`$xr^$bJX)9
zUR{)ry|U*^UeB-m*f?^jjyU=2XAbOSah@N2EwOVV?+t&SX05(`Scmgtz4S=04-`i(
zoulgIE2rnn{-e2)aniHPZ-21+VNUYZKFQ+H^JM+jYpzqCqtfeZZT|XqUhL*%ee|X0
zac%u2Kl$vhdiwF-p)b*Qh5nrr`Q-PWbgq$IU44ruPjqiM7xIzitt_88kk?XP)UWzo
zZ}al6zB0dw^*=1xy6mg-Ykf#>zWmAT*f@IrD35c3?gMm>vZMQ>6o=}N{oF-%p7n+H
z!<@+GPiAi&eiOyBJLk?TvX_moeUj}*?dEF#tV<v2nGZWQPJZ^1zr5_`Kwp|O|Jvn6
z{*~=Ndu4j_QqOs0m)BhAQCw+0rF`|Lw;$%jPab{nCnLQ$q(}Z}J?zNdeQ-bO3+qR=
zF8@9A61|tu??sp5mh8U&OY0PG9{&Eg=Vv+}{{4RE%so_^Q>orW{mJiMG#7FFoNKh7
z_HDmx-&^_r{`a`f`I#~g&+DnW9{oAr&cAcs=YZYwL+-x0FX^#!py#hH((}XG`JsEC
zKRwdRgVsAy+_VmMRImBXTb}w)dULP8{Pbk^Q`aMpI;Hu^BQNqNI|uaWKKJi>xNp55
zya&`l_03P7z2m>X@ZPa5_2u{PjL4@S{i1xRPx-yC%xk|?|5T&7nWK5ppKARcRzLQ?
z`J6Z8M|Li&m&bj|56x8`r1zYX<wO4DiTaxChox7?ddW*{{rX!!TDSS|r^o8a*t%rr
z-gE8TtKXk{=}~+s&qVc*A9-Ty=mXiU50~a(ANiL{@%7)zr`F#u+wRA9KmPjfk-IOC
z`aGzE?vKj7clhDdr}rxIM|$@nx@X;!<leWw``rKHQ622M$fdd^KlN(Yr~3TK{HJ(!
z{y6EU{QTL`9OT8N{quVv`}GA6{oD_J%ei;oZ5{UJQ`h%F?UniKn_c|Ww|+|g^2&$x
zr$;~kkk`^aul1k&{JGX-r?<YY*YnG6uH=$Gzq0k%?Z0`_%TF(UqByi3q?dQfQ@wT3
z>sKCh-tT+SUC+Jdt8V`{{ywDm=A%C>`uWzmKWJX-(>K!l{qx#7eva?G>>kA0)mvL<
zQ$P9buX9uO9Fd(<{iA)5M}Be0k1QYR&$`Gc-rU61Uw#x{S^w-4)e*1W66F)`zN8l~
zuleyOBmYy4)`NaOsPCoqS#N2d?Gx6IepAYi;^pr;tfzYZ@|06vaqQlU^y-<Py0!D0
zDBk{%vCk{NiQ?GF>bF17Eq@e;{K)3gd}LJDe8~2}^P(=f_dI=RKJqpXd*jIUrx%Cq
z#~jqPzUsx(qd4rm=u3WU*>k%#ZpqL7_gr*+=F)uDMNj6xl$XE#Qx~h}&puI{b3o=#
z)+gCKP`xGEAAe4)F3Mk-9`%RLUG4H9JAb6FA3dsPo%Rjsv2o&2eex3JHFxu*M|s$n
z=(|$Chs{mbssHvtKhZtwIdg8!%X~`T$^JaecfH?Vu%rKu!v7x+8TFwqSw3~x>E-9I
zPwSyacI)(A*Sz+hKfC#oJ(ud0=C02rs%L%r#_2ty&h$LeucdsY^_TiZ{Z947PxeVK
zZi)6m|NP~h<h6Qvu=_^8mg=H7vipFH{MJ&ObL6@8T$<NJ>#SW~wEz5_3$pv)b3`w`
zR8JhzPqfbe^pQ{6d)hDl(c-^v%ip}!(I2WucHYX`$<}R7=0h(JnV)!^`VfcBM=uZi
zME<plFTEePJlqL*F6=wmd8O}tVBfs=x}V*@%HD^ib15J8_v!TdKze!U$@<j~ySnuJ
z*yUwMcJFO@_<Ihk_nknG>grb=_lA1(WN~Es);RW3eSPVl9_dT@^i{H#^6=NUdi?oy
zo$QTwj!|9|C!ctF@)G&Wi{i}R^FvQA&0k*r){oBB5}nJ|Io0K_x$^R7oO9tFci;7v
zN1gvgH@Vx!G#B$TPv;8x`FX(nJ%9Yrd9tteLmm6?{*<3x{6zk~>*ZJ9I`qe0^0yu`
z`aV;KUVXB>)w3ge>3fjh#6E}SHTid4>gu;~)mP>}<?$TxYkhk0s1AR6bLsr$LGdV_
z9jkYq{r}9ctLOj!ip(w_((4!X!w;QLdHGpCKdfDR&$m9X@s;H*ovTva`qQ@``pLif
zo$o159&!9id62y_Jx+6@@A~vZFAx2G*?Z4>sCxJ4ME?3h_p&&Cm6!DF_QUy7hhARm
z;OF@>-@TQ8UtJ#52ew}QymuzOJf*zI?)Pccv#W#j{`+(KlApY_ll}b6pT07E{rQ#Z
zBD*|fc6$5HK2aPy8GDZ8o!E0NkNl`l=U}SS`=WK_qj!$wQ^$FCAN0Q9FHh@|*-P>K
z)FJcdN9ITN{-NiO^!&-rllXoISid~>1*f0Z&6gkAXMNBkyLkQC4|=jV`K*g<j{L;Q
zgSD&6F7B`?zj{mg`8kKyiM~thn{!QO_kF}}Zsu<u@@`7;C_h;~GMYy@`HNdic}ns0
zWal2OgS=nLZ(rn*pWgYk-|U`OacF<Vk;U;Vt-It$uP^t6yvUB$X+P!RFF*3vKk_Hb
zGf_VF%IfHg%zvUh`XNvL^7He2tH(}%qIk0E?st9q_1<qD_m+K_{OKECJsIo2R(4-E
zkNsbpr}5$@J^Ms)`r}7tr!V<ce^7a;uKfHanlC$9{K=+sj8i}QN9*W0pW-LIy7Eqb
zhox5so7cMZRr^|5{<87x?&VTH{G9j7-3Q;p^xjv^Pv%GVUh#exPfr%--qSC#lTrWJ
zxsdfE4_O}ik{w&e{?^V<KC*f4m)0X6(wFQLorA{dyZb|Ly?(z)R%epwCyFom(~~Fi
zvo3zl-QMxv@A!SIpC6nH?0Kx7-CX=0$y|J|`njZfb$x&O&Xlix%cF00l;86vpLLVV
z+R4^CQ9r)t`KixNu3eqR)AO%PkLKmPc70^~;m=ddi=J%0exFwITTAh+LpG<QqB`iD
znv40C`Y!b;e!o;tzVg|}{_LH1dEz<O|GhlVbJ^!dKh|lVtRI~l_c*!dij4Z^@A-4@
z$k+2I9-Vvn^o{1>-yy%^RkwQD%|H7Sum5-c)i2epUZ2+Qxm2G&J%9b0bKe`*?LK_&
zE8qW>-~EH1wEJxj{hsGu_F1>t)lc_LUUkXshaL5$u6fhTQ;OFIe`F_@?CLpJ_Q}4f
zE57?oZ=T{kXX5pFP?~?|wUpOB=->LQXQxN=@Sbz-tdDGt{CdvKhks?yx$i3L#GY5{
zwvYVm%UYU;^@yV<Z_1;Jmv^E*&BL6K9X;>l&Y^mC_1WzMxpsPFNAdbb{*~!BW%H2H
zx$HXGOZCOQ=^6j_MYsNed+y%v)lWX_r(gG(yXU;}!Dl_|o9}!*#alPZ<DBzDerP|-
zC42pkTGl^WC$>Jn(mAdFsV2AY_P1oO|B~#v;phCAFRDBFvDd$Pc6m+~t+VT2>StPy
z=f?UDo6WC}C0ZXpeUWi#-RzV9q&MfKczx<moV@h%VC|LVt-tSt!=nDwSC_2*61_L%
zL-P<%&;LHJ`o;@xeTzHo{?gOF{3*}6_FmV&FZdsnues{{^Ur(f?Z$sMqWRme^XvP~
z_nP;cd13QeukT8<es(l>Kgam4^qtC%@^}yXE~oeJK&tEa1Lk6V-ZS!u7pK1b)~iq7
z>GJ81pZxy+PWiv}^PhLw+pfLiZYfS4bCn;>P5$b;4zm1Y)R%SpIl#KjhaLHuQ}gM+
zbl&9Azj%6m(IdU}ljSiVGK$ytq<8N7a}K?EVCO_%e}3vCKkL>fvNxV=z4q6-=$$9B
zI`sC39qCULoeO=H`mjE-bvOQGtwUcYntcw<$sEWhy05kM^yj(OH?m{(WORO<4`i3e
zbLJjlFZt8UPoBuH_02_`K0KdfY<+R^SFT<>J?f*h57m>c2i4ba*Fkpgq5IN(hV;Ha
z`sV<C|HkgU@4j>&crN5a{T|kL<)(SeQJ+iu!GEHCkY|ZY``bG9%iO&m&A0nIodbFJ
z<CLfN##;w=AG!{*xs$PTGgtP~z2W)teaKHf=S~0oOMdi?w@!ZQmtCK_@|i1&mtTGR
z$WD*;jUT=K%wbdNbBePs{7UQ4mwxE=$xbfWty7(zv(sIE*ux%i?sdO#x9fjywBC1I
z`0@um=8Ajoe)I>fyzZRWe#Z6oX<v>?-}9`FJ|{cBB|E>hbPh`K&f9cO+4-L+>UXN=
z`CzZVIqKiN(l~bahCIlQ;yr)j%n$jk%_G0O`t_Z_?z@1VjJ`9()AL7m_bSqhBctDY
z@b`Y@kKGq@K>I`HPj8Ow*1tsl*!5Zmw%^M7D&4o{K(>EKUy9@B9O+ws{7@co=zd{0
zCw6-IYp*O0+0pqX%Zv0~mp+=8T*`y&Q=a-&Zr|*io;UlkMEk;zY=6<u6a1{tdF*q|
zPdvZM@_KGP5B&Ahdh+;Ate^U&yyBP0zg&uAx4x1c^+{IGT-eb($l}P(HCev0&x7;p
z+^IufuJvo2yx9F~|5JSRmGxKlIcuEy{Mhv?AHDv~n|>lc`S__zF2!}fJqPS4Ke==s
z<iq+^ZeI1pqqx?gU!wcK{UQ&2W%=~ied;_bi|fA2gZ$9Gk@X?JxRPBQiYJ$~ld<!s
z$0?t<?x(n}b19EF`y@X-(wDW%GwGcZ=cDwTwocc}uFeuK|Ak{W`M{muxMM%3eRkhs
zzuV|h9(kQh`cfQD`Q=;7wRM*K%+<WjmA>(0>mlo#jPxbD^Qta7C;X0z)A_T0&zHW;
zN#DpW4*Azk-jwpHi}JJ{Jz6hWKJr9<p7%|^pZJUSb0EFC?DW=0W_Lc2zIn*{NBP*v
z?DQo+^~_VBOKiR+x&7E*eKg*EtG_;XWaoO4J#XyT=bj((o9Mo*on9XFa~64tQ-9`S
zo@joRCq2KZkJg{o)%Z=<+4xOm=L5y-zwfC@_WPH~PA~5g>rXZx&zpU)Po6vLL3ZcO
zI+0#H(j$GzZe9A+H@$jf<k!6H^kj6dtP{=0y3u{h&%LRC_lmjFyJzV2gZ!*dUR0ky
zy?mwq)D_Q<jMcA|<@e`)<o2gd=c7;c#8od&9{Xcn=ErVM;@n&8>P=LiKk~QE>ebbs
z_XE8?QUA`NIC;hMM{%ALeTi2OCp|lQF61Tahn_4>UpVR6%@@tnI#3?6y4F*fKfV0U
zm3g8(;!wPOsa`+s1OD{tA-#UYv0FdAdi3Het81<kotN76qaOR@PwyO5c3#-g{cN7<
z(2J9g9_f)@9MYpY`Vh~*oc!bwkNi77{&>(o|FxbIMRPe(^_J?Gqq&oj-hSwx9r=0g
z>3c5e(S2c`OV1&_=anAWE9-||Jkp~)6ZNbA`Yq`vyZY0;u~bLia@sF`*3<m-`YOeX
z<B#lR{hb?hUi6L5ar4<<dHLx>J~UtZ-+SNpjDA{=%&tE_XDpFF>USc)iBlZ?6n9X)
zIiPyhPe%2~*n5Q^@>?SR?hCzmaw(1<xqj?u{bl1T%S(^uL7x2BC+f#@;Q63$9DOOi
z`o@#_O)|Z8b-nC|Mg4gG_nTK9G*9;;y`PJE&-ngyzqoh!xySsx=kIs-o)YK#*59)s
z>qlO)y7cN<hyK);PyHzm`$50Mmf}5c&I`J4$j%l2iOzNZynJnZ^NPdTD_ie=&*R!S
zb8$Y=zLVvnHy_W9{daDVzdT4^Hm>$5UYtI}Rqs6V!|FXprMxJ<ln>=;KjNB?j8mL=
z`@r9G>|E(%dT!<MoKN?ZJpB0~y*`{{{Mvta|2u!jwSTg6*mOQoef^Tzo%<!(dGp-z
zr|-Vdqj@xr{Y2AwI4YXcQRP2qUi*d4h3CgUq5W~+Imh<3{qPs>oT-ESuz4z*tGwdm
zr$^^rpQt`sm;CZg{^Ha}dVR~IAM@o`>chRo-u(3T&pLg-@Na#6AV2F?pUjTxlI77a
zeJLMKbCrkRMEUj8e3kWy;>hSbNq*Fq?>_Ucz4yO9>^pz;E9--PT7Ui2kr(+n7yWx`
z-h<XhMtbY@{wCw3*N6VPF8x_2JBlYe&t&AMj=tIH^^5fUP(Cz|(p>qK^6AH%$l^=)
z2l{u=&~q=py4H16G_QklsXyyN&kxe0^W!}AoY3>vzqytC+=r9Bdj8^<Xg%We-St$K
zSABW-A-}G_vOMM@PG9WyjbF)+ULNuk&re_ctpn?4-Qw7Nzx%oWZF>*D<noJdyUV^r
z`OvyNr&GN3c|JYw-Yd8iKiR9--+uYv2Y$%|zVRcs-7VFz?uqsZ`MKYgXg>VZ?YT2g
z`CE@a){npZ&JEf5Li1#omn{E8=fL}xpFSp8U*gr%k3P%`<>lv`k*y2q)g!kr_1Xu2
zdi!HP^<V1O?~^8b^H^W$eJqdlp!N2<jK98XHwSt6(TlepY##rASiT$mI{@w_aq^od
znzMbde%}T9px0mH_;s%I`ap9fFR^*Z_E}%%*?j7uKE>@F|97^%AI-<Q-OJ6(Uh<P)
zU(Sp9l99jq^2o1$?;q>uXD-f3{l&@GdOZjH&^Z!^>{z`xWY@1a>n`<0&tE<=T0a@>
zqvvHW|NnXP=EiQm?D|0aiT1nuLzZ7&lwVzBH?K`8zH}bwJ=fN8s#Cw7=cRr<XZ)O>
zm!JKCtMB#pw;Z#d$iK39?0)DAonQA0dXLI;qG&zNDO&G-={d7rviX{a{X+Bee2PQy
z{E(fDp0l3sK6mQ#a~}Ai-$za~Pkol`rMa)Axh_#1&$au|dd&&VQQzjEj(p_$O?kx0
zht3x}x%8YkHzm6~`lFwoCv`l3Q(k^kp2q2;da`{${%EdbacF+*hfU|edaZvVKdj%R
zuU}>Bt-iAShfQ@3E3bUzQhsxAKe>;@IluDK9~GUy{pP>fn}6sPKmToKANs$e<Ntr7
z+;5+2<IH`c{bMJ)pWIjUI6c?$=#M=4tHZw(-@0%7@<+bzNoT+RZ~VPj{!^aXE9+Ce
z<{^6@l22D!?-I>}pL5ao0Xu!!ICgRB*cW>DzV|r2d}M51?=$i8O%y+|as1RTotG(Y
ziuZdG>olL{7e7&5_R7^ybLt#@zm@W#dMLkj@T2Ef@|!3gCqH%3e972&`nCS{i@$vG
zVApA${H>>cWPX+H7d<wgc=}Qt^4BMQ??LOs{(M4jKA!uD{5?-({h&VBy=UYl^XF&3
z$o!GL-#O~Zht@?lC$c(a=d@p0UniU9cB1=G7wu#FsJ`E=?4^Bcetz^ze*96Kb2{nQ
zo~N#t>>SFk9{ZLbdh`YF`QXkCbq<~b=h6EFoon=Spx@&-pXj|Z&C~gl&$*oZPZ#}B
zt;74}sOE53ecMlTp4>krzqNF3<RzQS(mK>}PSH8{zIT4vQM`5dJp@@EdU5pRiA(Q)
z>(hsG;X6qms1Gt$uOE5QTuXhF?h*d%<Y|5NBQH^(*!5M<Zf@31=8w+VM020~olo)n
zCi26{kKTGFncjI`qI~9p)ssv9(|NC6eCJ=;I{0J#)aT#*;g9m5xIQ23)$>Pw`ap9h
zn}eULyhpuP-9xpD!znMly!x8f%`Oi0i}jPAfBPk)buIa`qxG7zb?B?IJ|~J}CtHtx
z)t@Mios7+I9qi>&-TKRq>||7z%&&cu%kCSyJZOK=J%;j_v$(SRX)fYh-=Ej=6K}oz
z$?AEYksjr-9`RT`8TGTYUiC`rRj>A`Z|g;UpnQi-{gmpofAW&uI_fWO`mQ*rUVe4E
zF6Y`g;g8Ni$=-bQ>S5PUU#g=Y&%Hj8pYJQ*XIMRXswc1WrLH>YzlUJAUhiqo3G$Oy
zef@NP>Y%<Rxq9Dg=B+<@)Fq2IH{ZYP;>D5Cy7^C3hn<Yo>j#~O`in#U{E*%{P#iz3
zojm2CUrTf5FFzjixo^BWheh}LVddLz9`iaZdVZXX()qxp^UqHneTrj8_k-t)9_g33
z<j-EpTRqu*rcdm7n`Cn+m+JC2Kj(!!)oWfd%4>ZkKl;k_@~owP-P7jky+Q9eBct~T
zfAZe({|?vRKO@V7Q=Z<l)`!#j*q7M(sYj2_71FCu_VXq`?^AaA_GSJ4-iJKsTw%`}
zx#W-foyc#A;+Oo``Iqd-Ub%X9G+*=T941*jJ6XNz-N*KC@@pRQ#MUAAy-bg-Klv}!
zwVuh3UOlA8>h+22>S3P~`zdcJAJ)Dki$nF?N8;)EP2^`iUFQ^cs_V_cddQwL_nE)n
zz|Z$Y*C(DI8SB5PERVjrAKr7`Z{AZA`MIC@$<I#iIjG+IQ9kixc6I3GVV`96k)M3_
z4e7mi_^Ycwde5mjP85&wRJI<!Z}R-9$Dbe8&d)kpkA2hgw5h+jqw|N(ePw#*SG;xl
zE)lmW<uB#&ysA%+`ayp3nXf*}P3L%6{`xY<k{u_z=fwGR4*mXxUc7v09_Hmc0_o9y
zknKY$-n_j3#n~sMUrTZ7sz*<jud+CFeyp#0^Fs4PdU@!Pojm#34}LiLv-cd>clxsT
z8oT?~ePs^qi>xo??>?qScKhO-PIP`IPWL>$xubcZ`<SdxGV+`3{94C)+0p!vAF5Nb
zm-SPp{m{E7>^na)KkFu!`d7d9N#FCSAA08ytM}hgvm-y>ebuwy?fDOS#`}NZ*1P+q
z^NaSw9LvsyT|Z=g=5|oZYagfQsCw%`eYiKA6SR)K<G)u?w|P%D+5UR3tnFW%_lbC9
zpU4lrzsNYfSDV*9(u>FH)uYEruMhc9oVn3s_2g3j)mPRBee1|WKT+T8&Bu=F*1!7N
z<=rpkE1Qqq`t?oEzxtkE>v11=&-dPvM?LhuP?sOl^QWK4f1>j$U;U=M`eb*X`A+G2
z%|jhBKl+LKGGF^Kkw2<qKlF{QtH0^rkz$v3sSkBZb>%~P{$%8*ez|1lZ{Nr$9_h`G
zT-H9t(_{OPhaUNx15Wzd`C<E4XJY-QKE<o=o|w+be)Y}=I`1XBe#jH2c>eAUe)QzE
zv~JIJxi(Mz?639OH*)VU{iA&R>Fp!E?>papzSH@)j=p>kIwvR}KXpp^^;3Vc`qu9|
zS3dF9Tj~q#7xLGSyc7AM{q}PR`ggSOHCLT~{&_FG-Sz*TpY^DZ)=k!je$kxep`R#^
zzR`2Qu0GN)QNGr<4rC`!b*zuQ_Y}Rp<ejLVKIL<M==V$cv3<}xFYNSWociFuMDgZR
zS~vaVPmkhIe93=_;_PSXe3$&_$@Z1(xkvR%_QuPP{K))KU8LtPzI5+a@4lm#2kH5#
zi}d_aU;OA%-f90lpZw8W-P`q(w-k^1lCSdQ$6wz)r}Un)CGwYtycDMo6i;sa^gQzu
zr#_ksn!EX!GrK(G(zy`NPR9Dt^H-<r{bzlpc~n2i-ecw>{;=shpQ^OqhfQ;}zxLn$
zcK`XYqdfM7T(UbaWSq`H_2Q@e@;N_bdC2VaHx!zi^^?uT94gcAm+GvIKkWKbUFWHG
z^^tw@XYX_2xv_raU&`OQWbyWoULNu83-vwEe$FHNPT|-7`OzbP??HX3PeywEqC92$
zcR&07;qU#Z4|?(Km)#uX<&W0w=V&sT8~-U!?ffRTU$S+&r_p-!MXxS9@*_`F_o%79
zzFLpHJS;zX%#A$Jd3JvJ(UT`yhyAo}vOefb=bzp_lTjV{$P?A~98LZ6lc)JxpC5W|
z<Tp2Y51W2}W51^S`Y}(kzQtJ|KY1tK_oBO=d(Btfe%J4f<dcUyQ5|#ESJ`-TRR`@a
z`&xE=?sM-QRLA|QfBli21OCo6KYHslf7G|WD$}Dp{QCa19(Hlo#a^=WpD3>R`L`dk
zdFY>v^0Alf@~KB(nZA_Q`LPf7Q$NU09`WQ-oPJ9EBfE2BADlb+)pedq@$Lz-eymSk
zf1W76KGY{8zuM{b*XKc9&wcx5*B9#F{@90}KXuW1^?6t{hf-hWz)sc|eQ7TIkzeDT
zo6ga?Q5^Y1QQcC%^h^ElE5%KEcIQW3dh>HG%z5IHpZzNJ=e)|Zmim;pdE}R`{flcJ
ze&UdRQ<|%O$nv7R)}cRfWPYBj+UrN2C?C7LWcDSRBR}yI^^N@Gp+|P};^+QjM}4fN
zeCmmluVmLx$!<T%%~P4*MEOwN`qQIxH?j8s8K*e<u7hlzIQe%Ut%tw(CGwZg`o&E@
z=UeAQ-wo#Oy{w*o)j@OMM=u`P{l0)7J$nDkYkm8rIq5^*iTqKYNRRYo=j=UCZ@m-w
z>)ZRR``|v4r~c}d^(XTqPt+g(%Jll_eDrM}*yU+m^JbSH>ra-KADNvV+4p<iZs>9L
z(>-!SnTK_IK5^6adrx~l%KqG@u5&xlbBN+locno+?j!MJ&yjoFbB6rUoYcqq`?-?6
z@$%3&o?X0sp`T>yc>n7!d)oO|-)5J;I^xcBI+yO@wd)mUZsMl<;jriG96LYGCp~&D
z%Gy1T`sCm9-#YC4<#oQu>_<)YN_~5N)>6Oq_r9i=zZ6eje|F@D^w#4$nBM-g)0@A)
zcW9rQmyGh0ty5lPCztH{LVb}Nr#`#)f;jWjm*>{~hy1L&b)55wp6iKIT=nLl4{~We
z`bO(P>*X(x{MIQR`H{uTN57WxEm3~=i*<BP^w>U@>eOx?)gM$gU(_$QZ*wlaFXTCD
z>Tkd7y2##xohv)4%g=i6_v$B~_0zBW%w77oe*W_=d)u{lyk4dk?>r%W+4EL^vVGBK
zDR1>;_USzHW9Ltg?wRTP!FgGtI%}zZ{mqYF-)MehdB~-F)i=*nPn<mb&^a_mb0RNM
zynLm2>n!W9Uvcc@CF&oS{A=%=^rMf`dgQ5{%)UhXDsH0ZTYc0oif>)d3w>pJocdx%
z^-6Pff9c=%3cYopb)b94drG`~;@Ig+e)6{8NuJI-y>*K3+*(&({PZIaJH5Q}Ao~*e
z%TtQyhwX!(zWE`4w7*D?jpL^d@*|`3=ebAw!f(H{&Lwu9M<qKS@>>^qQ<`5X552jP
z(fy$>{n*DPdXD|P;#|n1kIM9=dh%lL1$y<w>4RMIm#=yJ{KYOG`tucW^nNatADw%$
zIOp5%qv+9oq4R6s(EQ97yPx#>L3;g>%}rf(rn$(6@*q9Zn*&)ra;XnAFa0+^**wIR
z=DC*kWv&0@Cl6Y$z7Lz`+q~-P(|o!=-sAk6hipCap}gYB*m{kd<kr)NI!KS|%8Twr
zd6B<-<j&Q3>%2Ta&MWE%8|QiCPxkK=siz<7Ad4gGU*DB`&*?}1WO?<C@-4CP`qTef
z+MhFpU3b^jIg!m<yyuA>*~{9UtJbe<AD8^qYdm?$&QBfdr9V;BXQ{rq*FMSO*t-v8
z&lg(1=XiQ9*&AoykbP6?&%I<F=D@y`SG<1ZnP_gCQa<yue%}-59@LkAH;x_o`5vS%
z`N^lQx$vjQ_DPS{gXTf@K0|ht-#&?BC)-alJJQQX7AHUQ>-s(4)`$E{@#fFp^I+cm
zoG&szvc5}wW9?-2Qs4CakskF!M)MR`@>7q!GCy{6NA;0EJBoJ>(9cu-PSJ;aNRRf>
zpX<@1`u3MU+GqPK4)wi6bvg&n>-1fuuhM$WX^H%Csb7BW+Z@n(%uAo*>CMZ}x$4uS
zIV@4#qo(txAM0^XSf~Eo&-_aBL3v95zaUFAPn_oA9M^6i^=TbTG#7vVEZ&?-b3*>@
zr*-+^lD~N8OTCHwv3}EYRlmL$)o(ogS~lLfX}q~qUs*ouMdzc>JALzzt=s#gb5YOv
zRL}Xaug(+MI_1TlSN`(pOWnhw`J#K<dezm3zN~|+AL}p|_Q_w|erccdsebRB_Q7s{
z$nH6ExoMy6PoJf8V9xeK-_93(=Sh#wJHMN~`G;Qd^WS#%^?#mjU0-wM<<B_h!aMG6
z`SEkube|9E?>xvu#_1f`Z~KhSvFEw8|MX<fgLwM-se|lfbZ*@zOSB$w?rH01m%r3U
z`?n6{hx~m{tB0*a7LS|i^@HNwV<`U;#kX&9?Cyhp=ZKrw`I}qm=bDN9^=17?zn0!N
z)`#M)hn|e|Nbh-&-?>EdaIWdaTR%T~`|5c|dHCJ`1&_Vr@#o!Y*E%MOo9G@<2kDXi
zMA16U-TL%ne$I>Y!C$=jqJ1zA>+%0zvE}JG0sHQp^gOWBPxO7{eE9C-?_8D6*F^6t
zcIU8U-<0|hM_!`!p*c0Kc5~*Bo=bAsyzJ&nu3u&Mu0HmT|Nmeh|MuND@zv|6?0q91
zy*I5xJh^@8kDqvcrSn3sF4;NbhklNz?B@`6=UyJ?iQe-;FHU{y<d5oBRv+E({K)L;
z)lOztPrQ4cAN^WcK5RaEdC^?-MQ?uQCBD?(ME>md2iZ|w`|JGalfC|Ab@)w`x6hS2
z*6sZAubrRggCD)`C2{SCo`2)%^+CqQ)0ft%Px;9F<y%X6#Wydx<d5vl*FMazdNRAO
z1Tykdm#i)s#dROa_6zw>bWSIG>+)Y><H+qxo~b^)`(wXU*K;hdpG)+C?jO&u`@wyL
z`|Zd1Qb*mTeEjP-^)vZ5zOsDCE}!>H?fj749Lek_o1Q!K$Ei<xJlXy4xBltYzxRTB
z(er5iU5|5$_Vu9W(7$W=vCGf9`sQ!H%PxNCx4!3+U;Bx>?yjZ2OY6|5zTH>k(mbjs
zi}#)&v*T2kzkA$s;~pc6w~nr_GXK`6SJ%%;exC6AMKU|R^Fr32e))@^$luQy@*;a>
zb3*Hx=0d-g@|l}Csb{_Dx#8Dy;JqjxfAQk<$4?)nxb{Ov=V8jLj(qyU&Y%B8>yWSY
z+Xp}O#4qKq-FoFCH(%xEF=u^x?&(pSe8|2j<wNJC>!_apVNu`KJ>_Me{Mq%(uld-~
zdeK}<c5(9ZQ@?ut^{c-)di{%McfKd`6YqVge|@0m&HbUiJkBM`%T7l9`aV~Oy>VpZ
zCl9?mWaP(g9o4fVKQt%nNBz`ped4`O$o8jkoumAyKk@cWT))%l(Of3V@80644tvR7
zngf6N$o$Fd)=Rct>tLt1ZZbcty?*Tc)wR!4fBHAC%JiG^u<~qL*SfKN>>d9-vhN??
zRpwLj^Ihh<i(dZeoY-gmSdTblM}4)P|K8g3=zeQI)*(N6iTa(WzfI|$nBvr{UB64~
za39K3dJgE-kzYKyG-rKa{U(|J-tqINzPjvF9d`MhBV=D2&kvnndC<Hj>ZktnrS)3B
zy4Il|`LKD&@}T_erF`{QpB>xJl#gEC%Jj&t4w?rU+XuOM=yAFi)M>o_kiGuw`s0Uv
z57GO6aIcxW^Tlt<ORv5-&$s=j=cg{JgZl7$Z}W4{s>^R8KXv6V*-;!nGSV-R{}Qdw
zzFLpCwRP%ezVcxGoDZD7XXuf?{U)o6=8DdRIC^Y<lb#>DbEXfZM|z}3dYts^-lzQZ
z!%k+WFWF0elU{!0w-m?EK9I|<Pu`vb^|5*CFHfnSJnZs0Kh855#q&q^$)*&K&Ry&4
zQy-pV>!3G>Qaq|h#_kurzSz}cFU_fXeXZ5|d8cvBZw}2%uW#$d=50K`QeKo_zs*B#
z9<n}~kIXKQdgM~x>L<B%)RRX&_io=S&V%=ieB!%b@|WgUJz4+FM~~_|w`6@<54}E@
zDBt8KAL<+RN5&~%_2%yRuUx(FLiFd#WbAjEyzU`+`IFiCm;BIr%*Q-Sc}w4E`c;pg
zeN_k5BiFBXekH$&*3olDZw_SiT(w^1_U$>KcfL@(e#q*f^{|`kMC(BFM|z}ROZ~V{
z%&BZXamb%+eWm&E!}`f*F4o6SoW987$o%b_JnY!M>^Hl-^3jv+FTdLPv6uYq1DPLL
z|6RBI$RF!34?DSj?B>J1L~;6GqI~?w^5C>S{fnRIUaP-zR(iheul;F%&X+mqS08@g
zNv_>}@4l5E{rRIl_{*bi`xDQv>n2+_e_ZNUJ$~X!{pw#oWd8D?{Pnkw;?Vi9AM#8;
zFYwm~|4nKB`ji)~pWOWVNB(|q?f2C7$$5~cb?qyC^U5zD@|(yHomcr#U;03C^4mvx
ze%8k>FFo0JmN;~7)Zs_A-{hr!<<qA)^(ymYNBYvb==qy3y*`khKiPWpRXP_~yL{?2
zA3HX_^R1mMuRLURtFO!-r@G?QE!Cs1f9?F#MehqgZ;+85>6a+K{c?{Wed#<{k38*v
zt?YSJ$Gzuyln2%G^QLv`d*b9rkMiqBKK0DGtX;lRK2!(kvGHVm@*{gs(fjTsn`3ET
zQ9k=&zp?SXZ|tA_?Y$$vJo2J^WPN#V=;dc`U4E!u^RSotq!*9+B+Dy~9_j6e`MPJE
z1N+R6Y(M4GSILh0G$%4Uee;+nJBrgc+4_*)`;y)|Jx9*1{QB5C{^zmgfaXz}4?pu&
z5A`9QpE&iAzj*r2k-jvCiRP$|b2QcC$1bma)HiqEeS62}`CTu6!+k#bk+XLFee=$p
z+~<p49_K@TcKwuZ+k5yWmtS<--5Y=LyZ_7Y-Rln5|NZIK=|0Gd)^Bd+;2gPc__34q
ztv~0${OHY*UZ3iS)35VK?|fIL*H`Ui?-jB-`m-K#`|W({N0vuF^v<EV%g>)5ySb9h
zt^UnZyY;HyJp9$+r#?Cd;`KXGU!EH>s)O>#uP=7<bie3VK6ZKd$y2-i<d4pYb7g+=
zsH-o}tNHkMzxs2nxFuT8<fpG2I_;->;&hvb`^PyXJ8$Jd*XLZzyOfWgz9!1U-*Y$l
z^Yc8}M|sHX6Xh3ovT0q`PsUCAJJrwie%ECVjqiSv>*wBOcRrjOv=3x-zmnxqr{5#=
z{Ok)^9^`MG?L)s<yFBvCM^Ba?Cq4UGs*Cbgre9*~@~2-~7r#;-oW4`c+51^N-x=ci
z9w+Y||IR1N<IfGMpYn<0kNoxHd^+c3l&AAJD%rfu`AnnlKIhwd-2-IjXrle`-DY3y
zqxZae^3bb8mQP>up?LZDlb7~MUU|sk$R+zh&zp6sZ{B2VeRIP0p+EV|!#<dk`t*Jt
zWtZQ6%HzInU+RftC+n;4kEwt0;>hi@vVQriCy#o*2g&XgbFp6R`sl5%RG*%V^4Vu}
z4~i@K(aUF@_D_H2VSS}MOYF~cTX}yIupjO*v@h&r`-z^{%Hq6N#nU?vrG0mfDtm6&
zk-qdiRbQE3>AB=jt{*$rPG(2?(z@&i|B@eeUe&Xg{K}<z^{;Gy-4pH^_37Pn?sIne
z$o!EW`A>8&v3Gry`QehC-&!`G`QXyJ#PdV-*`2S>jg0n79+ZzCx%oXGlb!!$*C)Ty
zTpDj*ksX_d9&4Xu^O@q@gY3?=?*eoWF44TzCCiWQVef1H_OF!3^M>^D%EL~N=G1*7
zPi(w>lg}K@hpa!;AJVU-xKh39$=0P`@?p_D-2eJ958U5d|ArG~znllpfqAl<r+L@k
zb1dHZbsyE=x!^}O`PS7>sV`&~Us)fhZ+<AB^&&gULoVf^CpVvdvHpz{kL(l8buIO=
zmiBRpo`3sYF2!%^-}5Mbsx$eE)BmYX{aR<AL+>l=b&km;|K_C^kLs4};yi!Ot#j|(
ziFe<(zIEwWy!X94>dGe%@)J)k*?mX!=QlscA-mtdS|{oot*3pGk-v3Kl+XOc^KTw@
z`jTB<vj4urziYQCJx}uKSABXiyFR)eekk64tLJ;1%n#Kei%0p9AGy@0c>Rk*ab*5T
z&#!v+wUmcH8JG0xo8uDARetg6knInDb`(z*SAFNqzSQShyZV#AzWAw+&IK;%-AnSS
zYfk1ykNnlq57MLi4&7Volijyu@n}8jlf}tLubz7r{d_1+es#T1=&ieX>Gfm%=I6Y4
zPt{L;eIFIoX@B}Dt#hJ0=smTT=CLW&(=XZFtW)3i**QS_BM#*iKT+Pbbe<+ob(Zwj
z<J>OMc|BF>JXn``*;lk*rRUomrk`u3^-uR5e|G1@`D0(p##xvBu1t^OC$_)(v9G0g
zeVMy>c5=yI{90OX{n!5dIr-_Y`$>=XbCMfhSsb#<i{|INdk)00qj{K*b1sg3qWuu>
zx$)e|hwgQ9{XK8uvGvuf-n!}K^M3a8kN>WgexiK(c2DceI@F=($8JvglCN`Ck3T>8
z<<qbCkNtEG_DkmkH_ao?{vf?|p!HZUe`M!xzLl#Nhvs8n#82~OM|sH34|$2sIY0T3
z-oBa#z5eOtBlFW2I&YniIw%i{V@Lbae&qL^I?;OdZ~l|sx{x2)e(Kxv<J@{a^sP_p
zBcpivtdm|IWJh}I7bkz&KGem({IN$|{D;qe-!Xal%TJH(gI=7xWO+(?+)v{4Pe%F3
z>Y{u|kM!zQPcHSXzC86~m!BWHpO8N`A6Y*8OpoT}yqXX4L-nj{(l=h5dhGpf@w}Dp
zi#}iS$cyd+a-V1K+5Vo9zr08<uX$RR`5`}jt1lnQ%kJMBV7FefynDz0yH)p#^{6LK
zKjQUUvX}L<-{|kf`TzO!_vNe${eF|IFXvM}_2|vnx_wvj*RSWo^TLkqTa;fP^3bb`
z^6HN)KN<Cd^w@i!o<HihcKULyU*neK_Os-Fvh4N;yPx!E|C}%9!Sl$D;>;Q8`O#PI
zzO$paa>=f~{?YU1e9J3e$*xa%k-vOote<@JC|-Z=uf5~<8avt_{$%T{A3cAhcR#92
z_TKScAoCNC^!m0w^q%zmn1_D(xo7D0TdIrBX(`T}<nKDwE5&=icrTd;>Z5cX)K|BZ
zhhANCm6x83`mSAl^JSm><*T1~acJJ&xB5l*j=bg|US8C1>AX~5nIE!Orbm5}`J?lT
z^r(N7hip##`y9}t_>x^-_R9S9i<2LJ`J6}l#E$&T&3U139eRH9lF|A1oYI#)PnFdd
zCoj_LLp+L~$gkh+TlwQB(C6LU>B;uRb7H<^=Rq93_)<K7vi-*D$rC*n-Us?ded<S@
zlE3p+JH2}7UM0JC)zdc_J1>67Z|cLkJjd=Yq(}F}^!;RgW#jvEj=ZRTW#8TOOXR<n
z`daIM*nS7CZyx67e3(Cd{hSZ`=KSzG=sMMxzxnjVujf)f-gmy&e8*ZRd*eL^d&j@O
zQC+m&iTn<mQ@x|oukGJ`X1?wx^K@U@AO7~a=Zan)dF5*!@u;5okAExO=tlDJM|$}u
z_FPV~zMWTeFQWSF`k+VqOh)Iq@lzgtQ{UCQZ+af|gT4plNBdc}Kk@45gN*8t>$g^}
zpM3U3KkAj%SG{=oYHwXKnm4(5$mLYGe)^$ThaRmD+n0Lm&Y^s$j=Uv5dimt1mxn*<
ziyxUk_Pp`O#?gDu<Y`^~I^X2dd7`&I`P8c)|B1eH#p{EfjMhVLUhl=ut8wa;?)T>9
zSG{vnyZ+Rj`eEmf)<Hff?bDVYdh`YF`QXk`&!OkR^Wr&VFWGVGv**eAaGpy3$WET>
ztktXUyTSK|I@s@-%4_TKlb5`foxA?&-DCCv*XCdAr`}qBeUy!t*SgUD_|EKo<~*S9
zQ0)Dvj(u=%H7{A8?O%WLpg8;CTzNlO55M+9@A*^5`sAZW`T3!_iK9pTu%o$|WBu68
zPaozWk2)oL<HVyl{?^GZ9_h`odUpEyRp!@ylBfRdmn;w2dXOERA9av__m$qhpnRn{
z(pw)HoqOvn`7N>i>u*|z^-PpcU4439>Gew%=iIWR?<4spnwK~-fBVO-J{kGdE)PF_
z^T*cZzbWN+J|}iRoFnl_FAu$Sl968D_JNGnO}5`S&7*qrE#15P)j{>y8*e@COS1KK
ze(dtAkMgTmvZFlaCyzLN^!#^S{Ouo_GpdX9rM{{^QL^>u1E+QKw=Qz$NsmXh9{Kfc
zAKMQ*|FZqDTc@APr#$rPm;RijdUN(YQM>c1A9YH3N`B&zzGTPtLl#Hw^D7>$1L@^i
zOY!w@K7H}4oxbckDmxE7U-C~BcTnngsza|oRF~Zx$xGy~{@(FFpXj+}pYqukbY8`g
z+1;<+2lVfM{bf%(|LWW9qIh(V^VbiVAF})Y@;tIzf9EKUzGO%F$ffw|$?Ow*kJZmT
z#;<ng#hf~K`(qvE%+L4DV_*8Rn?32PZ?S9tJV$;1ci#Sd*?-4Mmbc%<>RL}}JzXb1
zcI&Zj`!CMA&5z#_<z4DiKd3IbWS_{Ny)yko{y6z9>CLNt=3f0IJO9XUNuK=lh5RO}
zqi=Jm-hR`QogcDu<6L`==$#ib@{`B&#h;!(yXQ|mvN#mSE*|;GL$5F7H&GqrXKv<;
z)`j+g|3v=sl<eL2`c>vX)n!NV<Z{ZRKlVwc_nxXh`&!yZdF+pMpmj{Le$;JV`Pp&O
zv#+J|#lJHBT5cL|9p)t79L&{ov}EUBwtsWql-|3gzSL=b{d+&NyZ6ZEAs^X&Lsn;^
zdg9ziWOmdSyYCM4ol$>%>z9A?O?}eKD<Az5%>(rz4_X(}n+v)1`1_7;pW^r-Kl4EL
z%JkwUwqJ4j@Z9*0^xswaUfVnVbBX<ACtD|f_A6d>tEb)kvp;dS@&12?Cd!NA^)3Gr
z8~5_FKXCQE-u{+j<|H3lKiPcvvs>@OU-bR&{P1gUx$C=JJ*0QeoD+GipTECfMz3G~
zIO*B-kNi;|&IQ?b8`=AZ?7p=gbRV&o*5lqZPxaLiN1j-}%8j$`vVLUiLizPiFW=I4
zlQ?rkd7LNjhp7+#<}FXDFM4t*554n4ULt?{sK544Hb?TQ%C4h*nwxvm{#&29$-}>S
z*v(PAdtM&%^?pQg{5<c-?jBW7KV;{VtdGj=Utj##D?5i+KXar%(X?JPH*;y<{7UN<
zZ_eV;e(@{i;m@v~`%@nN&LMwxar|m`o}4@Hapy=Lb>*>7?D{2_`tY97uRJCHiPoVn
z_4I-C6XiK7PW{uHC+fRoN9W7)!{1yzpX!j!3&oiaz5OR+{mhNObL3p{*DqO|yyC@?
z`FS4sJE!W=%dejG*G~3bhw6GSt7GowtuOV^JWKV;_Fubyhrsu){>(`q-QV_Oee$!L
zlRU_;?7HMB<)Lp}W&Y+n#p$>4WbtJCHj%&j>Xzn?_RU<Kmy#c<BM;g?{g^xdQeA$?
zPhNG&6UDQW#aowi$4|fJHPvSq=ee>^;_NSf=bpd6|D&(H<3FFuTYqsVt`uh<YBvvk
z>W?gr%-_#HWZz@@(SO&~`b%=}Idio?Wc}&CeY5lDM^Em3ERS`H7uP=6QC@Osf011r
znuEF1>kFM@c74igebc?cujEfJzxj~Wt1O;g9`AMENt0jw`H3?h=iNH##p_pm_m94P
zibv<1zdFdzJnWCW^!(_Nzx&X7P+e5lJp6e#JF183Sq~ZciLYKi>LY)0DKDA_KYIDC
zi(P)CUrYHWPI;VHb*K1)>is-%&^*&R4x0CLkGFr$8Pa3VJ3o4KewXN;pXhxkujf+U
z)^~rQy!_oG@=cT%t>1a~9Mbc1zNWnN{K@WN>(#&fD2_axJARXWDZX)@Kdjwc-OuLk
z-etE=c6zi=CHq>6SI3;ivt#wv;roHTv|s92pYIa!?BeAo+i!i6?IXQ@QGVZ>>ga=9
z$~RHnCF)Zi&xiAco)^zYDNa2<A6TdLn3sC$)BD~byD!X#f63oG)l~=S?VrB<+$0Y_
zb;MgA|FX~3rn3CFY5e5hzo$l?{7*H#bLgBof6f^h`P(=9?R=Q8{YLTbNqL+Ha;YzR
zb=7gsI#>SgA$i<Co`=RARJKk$Xuc`_sPn7aI-APwH~S{<QBfUzI&bvO3mM%<`gOjX
zBj=cZ`&X~}O=Z6a@jN-d?!&#~zwdE>`*U;s=@<1ymdA5v-JAAVe{-JZr@lD;WcTI`
zmF`>T4(*?FM>c1&d6MmK^VH5yKKo(6#PcKTrxaIz`Po~CUVhJM_3YKl&u*SmfAsn!
zi<jSf!`vpC{}eADdhXPvZ#?~}O8dCw-~R7MOy_0Oc{yJ=#jD3&|JwPjrFdK$H~Goq
zd{2IdrB|={*4B3)G@i`vxu!3tKKRw%eXL*o*>Q?ryKedUljXN=cBJQr?T3ubFAi(3
z%&(Ncdhy7wY}`R*^-lD6&)UAdANku?viW##^}XgfU`KlNT`KQH^AcD34)Ffx-@0UM
z9`W?%fa2swaegi~4{`h^PU}9=dUHIiy4E}8V=wtvPi8+=s4w@P=R<$&o+IQ>cJ7?B
z()sdppPv`yFXi?7PI0KeQ<a?$d8yydfxHyAWM8VU4ld<8)$P`S?RV1GZ@==CpT4v|
z&OLd*^|x_<`_DJ#V*cdQm1oL2%-?#v&rg^8UjJME|DU_VnvZib<zq+ps^8P~`-JTK
z_dkBhpWOf5FZuLc^nHW;$>_V;x%J)SJEr6(4&~F&TKcXkTX(;*KYyL5Z!~xH{P_TT
z*IU`oC+39aN|v9@j`Z@AQ9o#ZOLq5D?PPvu8c+1k{o8scX9CV~>G^T4-Am+Byg1}1
zo*pMXJGxKY>+&K$q{pTA$7FY})$V;W>8H5*ua(_<?zzLxzqXGhKliw~+Yfqu`?-g`
z_pJX8g8iUW_cd3YfBtzdz1{8-?JsuS^yv2~WVFt%Pe1&cf33U}zhqykBfgyCx-M~(
z|B2S?$N5C}5ZOKBUbBz-<>y>>-OhtJb>-t<){p;0&y{|hGxXil_0jA9TVDOfD}MZU
z@34#3$?m`FP)|Ox`ef_cbUj_C_`{;PSl6bsF8gavo6>qVjoZ{;zve1_lKFQ|egCpM
z-{K}ZAMDPb_nLnAk(W5_pZS*4`sq(qT8H^{?zQt<qUY7~Jk4{`tE(UThwAeym-csS
zzwuvw+0B1$XPU#>pBn7LVe@T!55MH{i*CCsujktD4ZKI~yL;o=$Nub{cX{GDyC<6F
z=zLD}T=1W0ee#GSdu}{m-W%+tIhiMlH(z;C9{$+A>u*2Q^&E?99{%j&=#gDKKeB%L
zq5Axn?qlx{`OrGmLv~cxzLC2Qbvpmb>iWGL8Pz4@+Vw8Y<EZTB()D&d)`8}3U$K4G
zUmU8_ebN`Zd9J1BiofTXJkj@%bHUGi^n>gZtpmj&Juc0GU)g--i?w?W=*fN$VSf6t
zK6B`tIxq8Xetowu`LX%r5wBk|dubi~uzqBA>n2ZZpXP&&Yn^Ew=2K437k!_%>eXvr
zvi`b%)?t5;pY_r&(R}%-+y2y@Xn*A?YgcDt<0n6Q>o7lL$Ekn*W&PLoB~CsvR!=@`
z+DB}D{gg}d`r!+Xe;GgL3D0@drT4t}9v9#9;!k+Sa~|=Wi*Nkb|9k1JFa4BDZ-4QP
zzV|top7Yo8c^7}e#h>ud>%QUX&$;xczV9FY&dyENephe(7AA1kr7yqZPrc$M7w(sj
z_>o_F!S8(OU+$c#To~Whyq&;y0^12}C$OEsb^_Z8Y$vdtz;*)L32Y~@oxpYi+X-wZ
zu${nm0^12}C$OEsb^_Z8Y$vdtz;*)L32Y}YOyHg8f8CY0_*egHXUo5R4jtPiUvak=
z+~FU6U}uSY*M0K8f8n`*u~WYP(Lev5$G`fcJLO;f#7$pz!KJ^sbJ)Dav2Xep@BP_p
zj-4rd^tV6rFYo^Q*B&eX^#{KBZWn*f?;o3Zrw{FZ-$Op~$H(wYna9#PdVaU^b^<4J
z0-nG9@?`eDt-YPVb^;$ifoDDL-Y@=|t8Q`OC%^7dcRY5VkL=)O|MSuBy2mg5`OaE?
zeAZK6TKZmo&}Y2)Nk8}DKi)arc=;pW`Vao@Q{KCCrt#E2QQHo;6WC5*JApH60?+x%
zbDnt4o8NbA%k3Oa<T*IAo}%q)wiDP+U^{{B1deI~+wYT8^ZR7l?<>CH)vtWhv#;7Y
z)A`{`-uY=?`A?pIto+G`zUH%k=kuPwbJ)DYb)Wg_o4)JHolQCI(^kKoz;*)L32Y~@
zoxpYi+X-wZu${nm0{fr9eeQj`SH1FqA2_xtr+;tg3%~VS@BWHk{oS1<{^X0Fe9i41
z@LM~lD}Vi&pZa-^f5lsNZm7KZ%f9DNKj(+9I<}U-{Di-M?dyK!w~w7Fe8WqgeeU_U
zeAltlg}?q?zkI<}@4NEYmZ$d|_J31vBhRD>?0(0mKj}yA@xdMZ$@@O`8JB+XukDn7
z`XgU>&#Qm*eaF`FPWO4pl|TKYcOBc5&wTzXzi#&le{}3r<ySrRXRdqRxBbDflg%^f
z3~tx5oxpYi>n8B^FM9gzKlRoBb?0>By49U}UJv}q-@4NSuJ}6_o-E$@+#CImhdt#3
zJDc)EPx87Sy7dRY;_7R6CicG<zx3%(`MHl>`rmiTPrKW7KYiJkzHet!UUK#wFZhig
z{*|3APpvte=yQ0o>V5a!fBoVszT>^ePFME#57XZ_t-kaA?6*Ghn;!R1Z+YSGJ?0%3
ze(nFh(SdotKYQ}S|M3g2+1Zq9|9(#W`g@K3edeQ<*RJzzZ+X;Be*8Xf+L?IlnfG|i
zQ{VW$os-4G`upBJUng3A$&Y>QSA4~b|7z!``GYt9=l6N*+i!N^-9GPgUjMCU-}rid
z!zX;tm;K<~Z*pPT=j);OyYUPE^J{N<;dDM<_NL#y`ek2m(+k(~M9+C!Z##kQ1hx~{
zPT<5&;8$<*nP+{<_kGfZ<$W(W>zkhS-#)fe-sPje@TK?o&g*u{{@nK;?)=o-zUs4n
ze`kr)pToU)=Y4zs@^Qa;>~!O2TzTGk*Zi~h?3}Lbe~){*$^G|pP8WIUIXPYIr|Y@b
zcki*v(>|QoTYFn=JAv&4wiDP+U^{{B1hx~HCouhf_=c{ZC$_bmj0qgppBHTN9L>MC
zdsK5h8E0)<b31|U1a4Rpc=BI7<Q*@+(|dNd{J(Q}@lC(>xu5*LKRCAKb`FPe4sKYt
z$#%6jlnH$QHJ@|w2fy>WW1I4@{vG(~&qufV&Hf(e4duL^DeJu8y-)wq|HIyS$JucG
z|6eaFYD5>&qpu){aIQr{c2|oYLiDI1h!!=1L^j0gBI*(?L|@{pzUpFGZOK+!SleyG
z{=T0_KJ)p-B-iwrbMAfp<7w`k@;=WwGxzSynRCWfcuK9Jqj<Dx9x<ct7YOA#%JD%v
zV1P#h<g>xk#wRBevnibatWg+$EvDIic&|_Ib0K44xu!FEgi$fY@4Fa1&Q0%jp*$FX
z0pkn^*9-CcZo=nkDK1agmRj#n>wezGQOk|jxE_F#VIW5iAFmPXzOEEUL3nVce&8tQ
zeol{h?;h`!LG_}ectJa000v+{F9VG~pPlkxKt3Dq_n)-Jua{48Il}js+UH^7^6YJk
z=V>x}+fmlzReL|$-f_U^FaQHE00S@p12Dj<fp%r@`rR6pNqOhi1DQ38jXz64zCEyd
zo**t5fB_hQ0d)rO{MFd!uj-Bp`?=(fn^D^m@i-uR>#fUOpUm0QsXY$J_i6*;@vdfX
z*LkBh$!c<ig64R5b#K647=Qs7fC04z0&7jay1HK?1z+}%oiD%V4<1nzG}gndd3E}n
zbxEWG)%Ihx`8K`3^Ge;U=Y(}C{hiyRZ8O(FjUP~ZRO`!8^NpM`z2?y&k0>j2RL@qm
z%c${Rysncbj@wb^VVQL~cfL{I=Cldy)2qtnG9SbfrBpjEPAZ?)`D=xBO{^M!D0iJV
z*3<cC6<)lh{8Pdk51#cMN-ic6vuM@z8}-6~90S3j55f}<B~UT9cep>NRZ24Dm6d;M
zG^Ws{M5>gI`n@jIE_;1Nye_KV+xF_k=P&>R|1=<ePHgo2HCoI6XAAHN48Q;kz`(y5
zkUMS(=|&HT@Oen#!5k0&)>rYF;}~dAtK_~tAI4J<9?%bP9thU~)~;|$pX(#NS9#ah
zf+;+}gO&Y&wd)Atd9?wn){j8F$M#(DIz#mG@am(Cy)Xa+FaQI*8z>ss^Tw*fw}`Wo
z_v488VE_hT00v+H1{}x0`X!UbMy*PrQVebR)98vr5-97ueEIYTWA`Ld@ZUW+PS0ht
zCF|b#eE#rOCx}vF9ui~cAz`z($d3xo$?%-a%I9Q8_doe(0#dSQU_$OsmglftBdn1{
zSBS}a19I2%6w+yZzuWBm)*D6yj$~l<n=@bES2u*PO4n1Kg}ee2sNS5}zTmGfS|?IR
zVd=C>efEz_q98o5&<`Bx8i}^>V_==<@p`u#Zy+E%I6^-V-iO!v-l2FtF>If9-#a6M
z!UH^@A2|MbAY4c7(dhf&&Qd-<zZ9*BRNN25>$;4M|DuX}{Sevn3<&oz!D#<MBLgWs
zz=M_jKs;X-w&y;szw8Xv+e4YGI>+(^A0Z$g9<2GcMXfNvrva&X$))!5*=j@$1I8E-
z)^!!%N0pkV82ei!oj-bus`X&Z(L|{*00S`K_y+X;yP5U!9RG--g)jgEY7Gd#<G~T@
zzOEEU9o<8Xal3x$9T!j4GVa>Gsx_a5QKQ+q=Yw{(i3ceN59kN{p9gBcpO8;q12vX~
zUs$;$m9Wkx=N_Fq()BW7mD8&QeocSAM|tP?j3L3*4<!&@_%!X$<b;eoHhXNcIV-Wo
znH)CWD4gZK&3`}nWwVD=V9G?-Q~qsFP!Jx_4{#pfI`Hc2fN&jkc1~QpaoN_WM5-6x
zY|vu<PoG3mR=KEA+XuZ4<*>=+t}-18pUSvSj3Vw^@NwUw!$<w|v}=EZO3~S`Culzm
zzyJ)ufUyQbH}!hMcIOF|%ekA6y;D0rff~&_pFMmw`NIeb{=0`YO-@A~eVjr;K0REj
zv|(y+i2%wQKUunV!+`BE1cV3l1Dpp|Uk8l+b;NJ!m&fFNd}l!{1>phxz<JLDey?DN
z90p(j2J|uTWvhrYJx+xYrPQZ(lmY`V00S@p126ysFaQI88JN8HUghsy(<ms<1NsdN
zC^4}0e8ieFbEBwU?6l})>~r5J${WY*e_Ac~g-FUP7v%WjSogIFlvmz<l$bMSMm*(}
zZEt*X%V$6|<(26j*EfGRDT;#1JdB8+*{!$dEegKcgOcwgXHHu?YQyT70}&L22POT$
zna>%tonHgOJin~Z+ve!;?HzV(&lRsT#A>#$8Zh<u)L0@#%!|**y!eiGocO)cB61jj
z0T_S*7=Qs7fB_hQ0T_S*7=Qs7c-cT+I{l}lGxw?9e7P2E*)zcEJiENaWzQ=<hXFGU
zXdMUFvgg<nD@KG<tEktwz<smbf2;7|AvK$Gs-1S<J0pTxO<^6A_)_l_?o}X_vce(L
z7o=AUNurKIv%D(3hK%-)zi>7)jwodnw}dNSAKg!RWsyAhOXv4aBw`+`GA3cf<V3;>
z114u~*<U=12#jt!t%ui_8ALA6wST93`Ej{yTB)>OAw3MpG0>n^$$fi1jHe(x{JkGc
zJ$m)tkRwq<V4XK=ldL9JsH0NtI?q!19BrHnukDd@E&NX!M<358Zl7E)TGNTwpEqk;
z9QQKsCFA=y3a4s0a?13YM~6J3QWV#VZ5V(77=Qs7fC2su+`16G?s^@63M%s;_4krd
z`>Pb#@jSagFkyv``gb(i_3Ha1{9pOd0vK={1E**EHaZi1gz81!%{;&T!`+3CA_!=W
zhl&k8I<kIhFy)o6Wk&U@TPT!*#&{5~do-Ze>*HN&-X=;pZSSS{>BSR>(Nw$6iPr0=
zFs=q1vF__ianw;gXzeFf=Wp*fZct*!bJQ&A{hZ%wf2bb@U;qYS00v+H24Fy$fuC~q
zY;Zj&uMLC;mivLSUTu78@i;oE?Z|pfznQo125}Z{-Z-gXSov*)H734Q^V~a2B8ky7
zz6YwSA!A{=rZakkQP!xgHOOU_fh(2fmh!uCm4fhqe!$;(P-#o6{jP1UQP5fsW?ipQ
z4<81Mnty<g)+0(7fB_hQ0T|%J0589v1)EF0?t6XkPpQ-%-#FH_->AarR4b=G?p`U-
zJCRDU;V%Vt+loa{y*N9jL&wj{#Zs*le1R|k12EvM27F&MJyYk#BMQFkf&O+NzW7_$
zDJyKVD)i2$&*G?6bk_4j+-|*o$Ji@JTpza8894jy3YxVpjq*lytwk;vfB_hQ0i_1K
zhL*|E_j(91i%R1|dKiEK7=Qs7fB_hQ0T_S*;|yFLllSqR1+mm>8rMQ=OO|?XM(_RB
zwidNJyMcEa6&(Hj)oha5=U#>09XC)ZihUi7%2mB#qZgDF-u!59-x)qnshF;(JPUaR
zCQzjmw-?(m00S@p126ysFaQHE00S@p126ysFaQHE00S@p126ys`WyH>EXV7gzIBF(
zc_h}e%cPSRs69rM?d$fy?=H1RrR#jztB3c`mCncBrf6io=izs#nCTtYH-9%Nii(+?
zFTdvx9#K?a{wCvonB#SwYNg)KWvwc&?1<9`qaOrOUg_Pt^u*Eu(bOJ0eAGWryY@FI
zZyZ&-%)vk611RfEURdk%MHBB(F?;&ft&{xhIu)q@l^G?%01U8dpw*2L1rN;plY;6z
zuzFk(moWyE{yl*7Et1Y3y+zITV9e1+sm@}cPwMGKJ%71Ktzzq*58Bx#9;Cdo^5oU!
z`)%7ytfK9Q1wU09=tCw2-)S}dwSfOo{P+8a_?g{$d)}gI={neJPx_7+D$wVA^5wvd
z59kJu^mzw=2&4Oto!L3qE%bl%>Rn@N^(TMYT{pGN?;Xd@9TGzAF|=C{sZsqV)tk<8
zUZb5b;0y-5k4;!~d*UUcm1VN(9Lp1Ygs7!6xJDbTH(=Jf!Pd9*uLv@GElVyoEuA<T
zqpP}I*j_t@fbd{tKVaoN=PeHLdS!cu9ouuo>kP4qMtv_-Oz~UzQCciDn$LgEv!l`Q
zKw^)D=;#3vJ`XAAm>#xwxId>=N-{N?@ul7=+^aw;1>xb}`T=j>Lx>j!U;qYSz*+<T
zS%nvGDgTrRJb2c3D7lzOKzMNEeqilchuUEP24DaNV89pyRgb*=_QP&r<k}yTe~hdc
z_dklq*NC@`i3X*@01UtY48VY67`S<DVD2>yt`I3YMr+l#A~^Ixc;cZ1sy9;>&-=Vt
zi@l_1V9y(?4&Sm%wd3Mg%_kg*m^|5GU0M{f@y4#p4{g5PCY`9I(0&jGU;qYSz&Q*I
znZ6*sT1XNF;Q{^N|DOlWarUE?Fkq~Kc4hDS-5QliL47=EJzC;<sI-=c{4f9mFaQHE
z00S@p126ysFaQHE00S@p126ysFaQHEz@mXVZ`3APO|DS0Y1I5IQswG3AMrL4?LTN_
zAmx?f_cNS*o5gbzF~NYd8pv2!uIY>(Vbmzj?)qZjoeQBv;QBQqYn`kTN%Z1ZEpDeg
zZl6WeGT^5crOVb!AZF27`x)BJwgK^Y4Q#)9JvCPRu-&T=9Sj&_fR*zPaTt4@G<yD5
zR?9c$tU;+}88B-;X5sy1@%ygg`qz0LuXnrg2I6RBeO_j>`Wxzj0T^K4fZp>cpgd#d
zr+`;;y|RDiB0?B|0T|%Zz}WaNs<_t=p?pS<s9^vGU;qYS00v+H24KKB4Vd-3x$<v~
z#uU1gNR`q#j|y51126ysFaQHE00S_f)_`~%6KmUQ+l72E00S`Ks0LQQIrH^>bwh}l
z!Z_3_1$I2oE)YyurP|{%r`l=vy)z;xYgF4F<bweifB_g#XCQfDt<M)tyhFtty+3cy
z;xWP09>><s?6P6+H7cgmd?Ax>_P@KYz<Fvmwa#O+IH26d0|#$Vwftn}@AY=9yiKj<
z^@svxCw~z{dE>1M(d(|)@u$49r*GXl$<MA+v#IW?fLul!_;}=z`BB6Cs8;Sjc4p^b
zw@_*o->;B*s#Crgs+LB3M)@!R126ysFaQHE00S@p126ysFaQHE;HU-;?wpc(_qS(+
zHIArxyyuA4&xn}6lr9;X{^4HAI!~=FdbjbP=Ls(y>mRc6@Vy|Sm3@kpYjkeeQ3CSm
z!BMYsXcr8?0DlJTjVmzwbN<XBVi<q{7<k#h?3fN6KQ9+c)v`z1X0C%8KcKv`<okUl
zC9i!%dF7oNqq1GYW2m#R^SSq<+f@pqtng{a-$u4A{wL*)+w$~ZRX947>cy9@F4%$r
z7=Qs7Fw4N;i!N7Mc|Rl0!X{6qb@ABpkT?s?>PM&t24DaNU;qZJG$5?|WEIu6y~C#+
zCTI4JB;a2jc7A;Q{jy(25LS8ctnW~AF_D1qU}Zm$y59IsXpwaO=q+kCrQ#Crm;2uI
zHWKYWXk;L@$J9ReD)jESfwD@YuNOwoBXjMK$v;L`j3Z{#S^i#^YNy%WQQyn`y?xby
zslTVj5-IBSyG4AR#5N4T01UtY48Q;kIJ$w1h2@&g=n+Qc(#LDWy00t6QQpbgyqT=V
z<LLb#?Slaru--tUh;LVntMHWj?-irJ7lrk7_5QvPzW*zx_$~Y>Etc}iLUi<i2%m=(
zw8n$c{ohf_7p}PRrQRvrt3WDM%gKB1RsP;JjjCmE=!5XYLkSdw2Q&JC(C<MQfB_iL
z%YeP{3TAyS{4STv9UHrqXw>ilvB!0u$Lrm0yn)E2-s3Tj&2ZhhxotWH;XzJ65YA&^
z-8|v%Nxdoe{8E}uYP}1k>ro@Ux^=v*_9paW5C&iX2Ati1+WSgY<<t9kz486!ve^xv
z5Uq3z?K|+%pb&C;L8->$c3+cCx}f^dnU|vp$hU{#QLQ8D6)a*CbLFQlN1o5kYqQ7Q
z)Aw#}T{MqPFa8)>anq9@a@fR_d*0ldSEtWemqaRXeeoF!#?*?YQrvnzV$GSkQIt34
zZ!+$OIbPSPR{Bk->vwc*2(_9`zJK$*?_ZpxAUvQS=zku_y^bQCHUrBV{Wrh&1n+^v
zYq-~Nuj!t1zIRt|_gw$^cX-j^C5FG@o@1W(@Y4U;e%n2#drr@+Vc&WW|7D@uZIWAC
z?M{@trAyB4ZI|Ts6L@^_@UgRMWmAD2&XiocefdeEmM#~6>tD3nR^q61qOO9b!T=1w
z01UtY48Q;kzyJ)u01UtY48Q;kzyJ)u01UtY48Q;kzyJ)ufUXAK&QpJ6mtmm<{C5vx
z+0AdCo?9eG3YDVpykAUVT!5HA-yhIsQmaQ)Z(eB9sKm3^f{58HIx~6C_rLF;db96F
zj~QO+u|&+n#Lcz&u25=^;yB!QezNxB${_0K6!%|j!+@0r?46%-KesY=)C~hL@Unr&
z@zbUk&B*DOyKq3`$^SR8jzF&xBUAny^MuOf%cZivCH4Ey>y0m7tjh?dyfJ^958q2~
zdX+jF+3!(^5C&iX1{}-4qJkG{AK8*lKzJ~-A2`-ogx0_S48Q;k7-ir^P>%-bZ$BX*
zTORBkZ%4a7VXr-ff7-bA@oHZxQ2ZQA?`^fu3GE#xd=3LJ00S@p126ysFaQHEpx(gI
z(QcQk`9~3@+&S#46AL@UQZdI=@(NxsHl9fF$va1H|28j<@Jf#tduo2rJ(f5t`%Lf%
z-FQEO$|Y;#B8{H!YT>>idphhsML~G@dp}TrrAAIiGO+mCsI56>Bv65!TJ}8Gzg7Sh
z^Z0<#y9zaqBXU`x?x1HIa>o*-bXmOL_e^{$VU-EC>J4kwOe9v*k@_Io0s}B0&%h7)
z1M<3FyhM1Tyw<Uo%0HL;h(U3Q)GC%4@Kecxr$dNRnr(9Cf^L&84T`7s*!RDGrwg<<
zetuE&Gmq=7iKR;UUH7?9FIGK6K(;)X-4Cs*&*;D7jV?c6MauzQdIb=Ha-WxreRr$b
zt?(}YblLY~r0&}&y&ou~SNq;~_f|LVRSC)>0?mHUMeaBqsd2}0%P}fli}<!(zbTwX
zv{Kx^%-$CIv(X$gDSWy6lIQ<>&rEpVY;)^7k)Z=JshGlkfvKN-dc0Id7X2=%L*F(l
zQ~yC>zfg}9&GwA?U;qYS00zu8Abc-~sdT;;@$)6&ccc`B-&tZx{k~y!zuNV0Sjl&x
zS+fd9S40qADV{%vm+iv8d^jOsbTHwaYX9zHC7;x|b9-@`d(}p5FaQHE00S@p12ACU
zfZ6x2d8yC7G3>FzfZXRzNN3Hww?=ulqOYM&7=QuiGT=FQFWuGsA_f2Qkn8x+A1kj*
zB_KTL(GQ&KYJ%1}i-Gc&JjeV0kl#-!v$pm4>9r!!R4K(?jT)cjY~_l*Vw-IP!u;!2
zk!?HjqqFKfH~P7|d;>@c12CYs0jtKTv|oR~KeBBSVU702%NhONhNG3QcR#f^UPtUr
zJbqE{ZN1xt@?Zc4j58oLZdvVqd)F25^+>#}_4`&_o}+9V*8@;84CKghb)IX&o}CE<
zga>Ep2aa;?FSz-1)7=3%{Pd!uctJa000v+{F9XLH4<9?LR<_-B>gBG3&3ih%e2U9?
z)G2D`)o+U`ehz*7jn?mHm(FVwmuGKVYCaad_wzDdTq;hr<74d|2Ye0#FaQHE00S@p
z1FRZQ`@4s=eCJ2h&wXUrNec4qfz|T_alrr#zyJ)WGhlBWVSN7Kd+ANDQmr1;9ToO-
z$$kGuZA--CQ1evll>14`Od_V%@jK3vU+r-~zE>Nty8oyhFNb_D&&w~@$E$k-_QC)R
zzyJ)WHE@A0Sedl@1_fXC@VG~@bFFILBcQP!zFZoz?b4n!B2aBVR-3O*%L`8*SHDA9
z=Y_Hto88GwA#yn-yWN1z1Cxn8s`cfl`KG+Nug$mFVT2Vrs%NX(rT2I*p}g1|bV{d*
zsf2e5>y0@Y#p{6x$4x2a{=I5;I$`}uv+EV=bCmb*?7jDn&j+-b)anrxJXm%8M!hf~
z$AG>2!@RBoB*d+D{Yh5VV=_Aqdws>|IPBGn&tU)t{%OE{=O=3~t_-3AS$}WppDn;A
zFaQHE00aMK!06vwD8J;|?aNORcu?ly-})*(a~uQ0?@93G-`DT~AAkRQ%h~z^Ygag}
z<K2;8pZ5mf5gwJif5O^z1o6Du0PFV^5uclD^If5m^KQNRC}S@SzyJ)u0PhCyefZb+
zA@9c#@xuTNzyJ)u01P;e0rx)&%<ecnzfFn-Lf`#<SyWCN>-6c>dw<U{&k6YN9vr9V
zve}Y?ea;ma(=MDUrPO@h*#E_gbs51_tp_%Hi~Oh@<^3$Q<G21ryKN=#@UjQ_X97~P
zXh3+ckTtUC3Ncx4KzJWdE_oZ5YrSDa;7A74{_Zv_pWO3UX*x%nN2BjXj~QO+u>^z%
z7W#oBT_e#Jehd_ynY`!w-*-?D9vq<`G{1d%Zjl@*)ZV-)azA&p-#ciJ+vRHhQ3M{~
z0sTPwJdhf%$lHA&R^#R8m!dV1#{5Bc+!>Zc1+LuWdwp|20u?CCdnvr{#_DlKT*~G%
zfG_9pwQ%2%JsozRqO8!_-XBMM*_=PYnr~aw3Ilu^kotQ?YCoT?M$|B1jDc0Y?`<e|
zKbgFpr~b$;!$SW@JSQ;rIe{@p6Q#lc48VZn8z{UhdQ_{mm#DMQ@sB842m>&n)&QP&
zJJ<8>$GzXU_Iz&)QA<D9=dI|PL~1k(7V+rN;{GWD!UOsN|L1|)?<eHb*MP9@H0$i%
z>c+h)L0N=Vx-8!BdnP`W@Xo?NZCv|!wJ+s`GZ%E5bZJmLwMVJvIIQj$&hljoUIawA
z=dlTFGWXBbJ3q-HAUvQS;5@)};MLaw;X3N<{N%jPq~K~%L@&C;WepyA_6A{<TkbY1
zI3!nYKe>F(Gby^;-e_tRFSNPgGP1>43bO4X!B)Lt&6<fsiq3vLLHl6<24DaNj5V<E
zWXIK4Hii?qobu+rHs5B45u>@NeytZ@r}-1`-#v`0m)<x>@niz>>7mQ`+$jr}ClTJ5
zzll%lg$4Zx2oLB7I1jA84jB9Eh~LtiZ99Krl}i`_;Q{@?dCvoWuV9EA24DaN^f6H6
zz|1ZU4p6F;`qYk6U;qYS00v+H24DaNVBjwUn7>5X{3Yl&FrdW1g(i(kJbNvO=*DKv
zDjZ!AL6ojnV|9A*yh*%mw_kt2KeBBS;g!PtRp4Lqc$w{i*M4Jl+^oL8##&rTzLT6e
zZNQ3_1G@AIARs&_=?Bhy&Y<o58W6sxSeZ}WS>s^kK8vtEh!pYti0?-w-;a9r7k;m_
zh#UrB00v+H24DaNU;qYS00v+H24DaNUN&IVd#f+kf-QRnjG70{S<1EN6`#X^nFcx~
z-pgO_L_AT;D|rg$>2@oQSjASJw!IM*<B3)tA251Xp~i8<Z00&X^vBBUQi;{fT{xie
zWY;wM^ZfyBCbfFR3|V>Zr&k<$J?D`sN9i@h|6$l-+r6t)DLXYdcqJk80pXPwg9;2A
zek_WLDfN8U?ETGepPpMJM+z0#XX%mC-%L%Ra@lJ0vXNWKbE=g}`xVl|fE)vQU0+d-
zviEUV_to5ePU)0J$y8wBUp|}=FglnxDqjTkXpsK)6QVbDdw)jn_i>!8JUQ3G|D<uY
z<0aRN)^t|A$0d&Yhx`F~T`yiDYWexp(s3>Oh7l=>>%}$<zyJ)u01UtY{{}7}U(xHA
zZ!Qv0nTM<MTod-}Odw`6f13~AOK*CWvO-7wI~whJ^?efluY70$3^<Mfz3)r#Zbo?T
z7IbrMzAKcP?O{>D3$>4INhcsY{98ZhyU}BYS9&ZFsP=xfS-*qTKIhVU9X&eQ?Q%8$
zC;~dR2d(|Y>ilZY54HL9o)^n%f2bb@U;qYS00v+H24Fy$ftWwYjyuDWC<qTM_XB0U
z+W6E;?+ZHH`Q>u)xBf-DZ6&Nxc#p$q8s7ue)nN8>ce(YcYYlRlWq`MJ4-jwtkzIy`
z5_o`z|MUaB@B5-Wb?d5_b-hMCd>AllUU5EJk0@aP24DaNV1N$;9nO?oyM6gd!a4<;
zQuiH|-p90;eq*(7>rJVZ*H4P6pL}|}R7Muni^4cKtrUELFaQHE;H(Dl{?@<0zm>JE
z$4{>niKeX3>gQ_CdVYx8y}9+B$j||q)F_JU!?rpDtp8rO7oWP;A{Pw601Uu@QUiLw
z4=h!l()f@b24DaNU;qYS00v+H24KKA1FXy+VO$HXEm`gSC2I3o+gjA_><0D>Y+fw2
zZcaa`b%ccdVqaE2_Yv1`Zypl0&&kB?#WoDU01UtY48Q;kzyJ)u01UtY48Q;kzyJ)u
z01UtY48VZ?2F6SZU+%u-ITh1==O=3~t_-5~Xz#gz@VW4OLoOZly(Fpeq2m4a-V;#!
zJXCEy@pB+vw^@0<uQm>!UcLAC9P^ymqu%S8n^hh!^Gn%#?_7M|>VIWMi7)^ItQz3u
z`o(ITtR7dyWsCvw_a(NQRXDmLg0SYnn4^zUoyCAz>kithr_ThB(2e&aD6jna{(v@<
zT0NpxQMlhD9^WbUt@VC@__`<DCz85fA*5r4&T?L(oiN}G2E5+*;>Eg*V5*hEbyO{#
z!8O`wy@A*pbV{d*sl;en-_pM#u<Cb}(e<jm&LH2bpL4>GGC!=G=e)%syf-M&tHj8Z
zKgT?wR+0C0g_2tp%D4MOK0kZ3_naM{^D+Oq)NwNFIX3F)(IEZpCj=h;-2-pmLx>j!
zU;qYSz*+<1=P!EQKfvF!%Kn~(FLQlad)A?L7=Qs7fB_gV#=v(;9s0Ienfj0St)!ke
z8xsvmg#j3V0T_S*$1osw9E)SLR(&f%K2CXaTKSwdz4<s~h-Z`YcO=(QbhP^_!jaJH
zKAqX+sC_?K$S1TPgaH_U0T^%&1FXJZjJOW?N48BO@L-Jx=Q#V(N*FNK0Po|FS&Lum
z(GrjI)LI_$!vGAx01UtY48Q;kzyJ)u01UtY48Q;kzyJ)u01UtY48Q;kzyJ)u01Uj^
zfYdz7TKB(qtaF)lf20tB!Z={PSmeOWE)5P)s+Jep+;AD$;w&|buRa>s3j@{|5YMxY
zZ8q245FZz7dJAfW0c#DU&vQ-KvonE!j_u*<e~+`X(b}^MwQDsH^9R{+XIK(7n!-FG
zQuBgnU5AI&{Ni!OtZi%UN5~HYFyJ@_<jy;Rbmpu#1J7)DcAPUDEr9_TfC07)Y;Jug
zGIT&DWh*K~0|PJs126ysFaQHE00S`aDg(mv9;11fxVbjp6-t%zRk33q48Q;kzyJ)u
z01UtY48Q;kzyJ)u0N)1sOz;TZct3)QDU3@Mf6m&rTKgT&lw7-g`ANbW)wT!uU;qYS
z00z_<__Y6bJ1hD=r($-9uQ=xYTglWO?+&OGIqlp-DrWvRAHJ8~^eVN-h5Lr=>9G40
zHJg1mdd%=jk0olk%40~Guw7w9EsuDwJu^NfoLJ4@0=raizV1)L8+T>ys5t+^J;E!$
zUz4(7)}$z6Hq~7fkjrQTwHnu|ceZab)ynD1YAugUi=kHW(bpsBqK|J8wKUo@%7+0M
zfB_hQ0T_S*7=Qs7fB_hQ0T_S*M>SBm*3X$|ir%BFvBT9p-}QMff{K}=*NpPrUnCIL
z`N?^oNx{{k2rn%5tnB!oHzX3ROt4jNShHp#0r~XcsMk5P3kG0-KLf1H&&Z!SL<|Ek
z00S=@(E9s6)8*R_`^Q8O-uZT(`Xjpx3njdgm%l6ZigV%g`%lVtK0#O^ul;7qoBP^)
zn;k~<;>%YTY{38wzyJ)GWx(0yc~(0vpVxe5^&`{+126ysFaQHq8Zi5PKXH9(+rI8#
zKKHjv#8dDu4@GAt@A>}s9h6myk2|)ldEeIAy-Qthl+IJCS9+<q#QR&sx9$2(;WQ$}
zMFlU^KC&g9*kk+k2mB-3CJ|O?^z}@<4zk|c0V`S#=+Y~In9U1K8kKnVS`cw|vNoTN
zv&AE~|LOI7M|_>cHVnW348Q;kzyJ(5x`D#Kd^jOsbTE<2tMgnF_Uud`ypz}Gjb_K~
z==~q<g8>+@-hgnw@#tu`%hmj&2>9<Fg!gUqzCS3G%lr4b-uGjjrMwqGJsPCH{e)=c
zD&O}ul)Ilyz*l<^uDIP>-MCjJD2rG{z3!u;9KPRIG3)pV{T_q?7=Qu249I<MXjHl}
zlfsv~FL_Soa>$2MdXs8}{OnQbI=f2OwUHVpVb=En>T!(U4Z?Z+#@ENAYn{pE=f3lk
zwHH?gQN1bEZ`JO%x6Xy|IqpaN`~3(<^CI+P5C&iX2AthM!9M2-jA<86d7<9Nt3a3N
zcZLn!L$&g9YNc`kVV7yg&iVHF6~98{a`O3JyFCtvP>^pAty-pf)eg<;C+6&TfBwBf
z|6G3dxaj!LM|LjE;ing?*WVfA`_VHhrrh)9%?ph_?DTdV6)1PzquXUScUy8gk@CiZ
zMLasRxPOXh<){%YPYr2#hgeNs*Ih&0e6O45EXPmobrk8e8Ccfnzxll<cn=(2!@Z{a
z+wM8%dw2DA&-I^whZh}QV)z^GIp%o}Fa4kG|J&)F(=%(>x8B2lS?G3~<knU@a>D=&
zzyJ)u01Pl~pz_ka*JHC2sa0fJ4uXRL7+}gkjoSG-T=R*iO373s5DW~!01UtY48Q;k
zzyJ)u01Uu@=>`(=el{)G?JNc5c`*F|p+Xpd0T_S*7=Qs7fB_h=&_Go8{Wa@O+eATi
z9xOa`s2T<w-oVw1gIoEgBoSw)!yjnQ8yJ3|aKIv$BZODJU;Ufrk$WE!t>pYLB77Kt
z0T_S*7=Qs7fB_hQ0T_S*7=Qs7&}^W5x79~~jY*(tS$5~Ga-F8eQP5fsbzYQf9GP;3
zItw*-d=vl!FaQHE00SxvxNhxUXwbX}0y?J$6<vm@EO;M7-uoEvq^~EYP67mDxq&C=
zPOLuenMTYe#+T|dzV!1itNk`}UNE&skHRhcjhmNF)iQKn#epCAM^L3ydOu3({9N|h
zEB;*RJUT*pR=#(n&J)WoAPNImKC?rUJm2TA$)&>LksJnKzybqr)~b0wq*oYGOQqj0
ztffD7$S<$k*boZ-<-vk3$Yho7U!ChZt?q5g3*|nC6w4-kIy4?JFQ3h7MlRSs_v@xl
z32zK?nch9iC57n43yVB!PI-_>%w}f01b^Gq45F3d{)KHAfB_hQ0T|F?K(D_W#4#wn
z-(_`ras1eZ0T_S*7=Qs7fB_hQ0T_S*7=Qs7fB_hQ0T_S*7=Qs7;MYK{>;5ryN+nZ}
z4G+AG3$hm{zh?p>|GR;#-=-e_sB#RI%fIJ!rvD4?2k1?)!?=bXi~Um`Qc#-*u~%%v
z01P;<0ein=lz#5K2GMdDfB_g_+d!jfBc4tQ45U_*ZA*v_2J|(+`}wAH9s(i#i1Hoh
zZLNHig64S8x7VUH7~sQz(!Wz#ORv}aHz-HlygX9g%svZZ`Ze9M-{%(5i)P1<`e49H
z15)!t>AhcWTgCHmnYB%REYcvp*Hg+0n|4ksz1uC7T1DaS6Qd}%KOmiR8jw4lz}eEN
z?F~bIDSCQj(>UTRRO<=(U;qYSz*!6!ec#<#JffZQ4E%h2fJeIxIc-K!ulqq#<;ZI-
zQo#TW$T6Vx`KMGJj`DX;-9X!0otmB@pga#o-M`YST&aGG{l*wj`+fxG>E>&mZn^8K
zN<H5c?^oL^!w(b=SmbhqXeIC8XL65=R64cCfz^Cke+TK6-)KMi`*`O(l6{YWY<MvG
zc%yt6fB_hQ0T_S*7%<9!)$`+v>o=;MW|ymW-U_qhM13#-1I8IBaQe-(lRjAlga<kO
zfcN{9R>iON43@fYs`q}a@gl!t8L;ZT1*7ZL>;Af<l;c>v7OjB+7=QuCF<|ugZbvQO
zae4q+;tU3^H;Hd_dQBMRm1e!?ZFD_)|6OfXc}D#`;tc2B-`n_A#a2migW?GY59kLt
z51jQn@b_z?ng4gG|LfyRo{y+ns(mlOUOw@70I6-Y&)JR2*X~O74_wElQC?~E{a|tV
zQpZ>AeqMjqYn?~RsB6dX(|nc`h{$8(jjmg}7aBA#f>=ePj)&RhYVDW2<X1aCl2OMa
zA@66?g5Azi)@fAxQ7#OyWZ=oU6RVGVrV+EqQrq>6<^Ag4G>_c-kf<fE<Barlh_YY+
z29y~np8MS*dH#4tKzSZY>`34JY26S4^5wzaS;6|>5#soGy^kZr&HMZ(MnC`6TE4x0
zj?c|CVAZ%Lb6bhpV1R7{K`zs~XSt*hF@Jw!*rd2s>xeyypT|jkZwcRT_Nex{E>$n~
z|23Z&Jh1J7?W=_IM#uZ#vCezs(DGLqxc1rI_Io@7iQbf&*W0Z9uWCE?!2k?6mI3kn
zvAk}xwk|B<IoEYs-P?w(A1<7QW_>TB9(~ufg->lhdH?%L@88d6muJ*_X=ay8`gCYK
zVqQL*T#kGF@y&*ni`blnc@BRX_+45t8|&Qp{*P17UCwRemBVk`IG*5A*apIbntpIH
z;|Be%Sbm#T6n>973(b1}TU?K`_%*AY_MR&`>*sjh$M*9+XFopTaeI1gTeaTJpB<Kt
zdcSfi0pUSQKQQb2((HOlyLnb`7x<iLWlFv~d$R|n6RT*pH`E6MFaQI*8!-F6w%+v_
zJs*M9$JeUwQ&#J>`govz7_ibn-9UNkzspT$Wm`};48Q;kzyJ)u01UtY48VY68_@gy
zPx)@EkNz5yK(*4fcu0=XbyKKTI`+{+>tFx|<QPyp4?C;*)XoFJN<ORRwb8p?R(#5h
zLp&b|Z`*j!?yvXk?0JxLEp&!7_WU^N=gx3U(MA}60T^&*12z1neS9uEn}C0L;C&qV
zg+-n<r#wg`&PHcGhtYNzV9kJ|{as|#I)p}*>uB$}$-N)NYC1<bAIzR##M$e!>bkDi
zd!gcTw6-VwK;eKzE=P!?QSJJvX6IA8?jkGsv_Aj0I=|d;->jz7`+HF9^~C7>_U7?7
z>vN;qhw`1nfcSfy*KJn+KCmh-N9(<g^7n$;aTKf`Pu}Mn<^8#U5Wil>fz^J7xM0AU
z4X`>dwqAe#OC4Li|GpH;ld2c{Uu{6@zDcd?{xNk*B~xeTt1X(hSLpZPtH;M$<wx(H
zpx5;o<xD;PQRNs45Ac9~!2WsQy0v?uLGvPrT<Z1rvsLA&z0auk_YEuMiRXuBecP(v
z8+z4i_Va39>SN`6KpZf@qJd!-GjI2dxlDMa(eHh;c+3zJ48Q;kzyJ)ufUyR8R}Nhk
z<MN1r@W4VpFt)d&bQpjE7=Qs*4CuYSH7o5x9Bdk3{W*=<@v(N4tXdD0m3qa;W9R!n
zPCa)yw+)o%LF^CPFaQJ2XF#uU2U^Q<KJ#cX48Q;kzyJ)u01UtY48Q;k7-PWbb>8ij
zZ%k`YDhya_z^eC)<koAg9n=m3FaQHE;3x)oeJ{wYc|cLmut{;N))BcL9Ob-lwsurr
zy7zi)b|Ph+*PFyQI=v>0>P_+SHfkOcp&qsK324oyb)HWleh>y=00v+H2GkfZ>igBK
za*bMlN+?%Yhf0lG<YC8v*70-BnxCB`h)7@n24KJ(1H$*dRg`L_Ro_Qaamn2;cfLxa
z)5(n&>0kf`V8EV%vJ*F#>G)nr8|WAw-r9e(!H=Eau(3{i{n6g%q5CQh{J=kg>P4ZR
zAeZUgvs_Y$TwYk@S#!#RL}E6DczLJX@dB)_qiQcMtB;FW_bpM6zUwP!eV&T^9p`PW
ze3ZfipB}{XwTQ2yTE{CmYksjuY{LKyzyJ)ufSCpwO&jraQeYtEm2$@|vzpH6b!pA)
zOIFlXuF_4vYNc}8oSjy<N1ZSL127=RfW7;&_}unZr>19!*u%BY?zZ3K8A$YE_<_O!
zi(HNnR;cv%rr!67<n&pjkv4F`^W9?2{1b>hN|%5XFrc3Sy{{8;%hRs~C<z7}*8uP1
zOIbTIa@RGJJFY`%I@Ww39vFaue;P3QJ~Qj(3r6%B7s$$bDu}~7=aKAt1Rh@YAoO<-
z24DaNU;qXj$$(zZMXW04&000@hx7^~deM=NE!qMDyc@9UKBr#wCglBWTCm$$$}7LB
z*eWS*P(0z46=x5A|J#@s#A@<>9wL4i;K_hq?@tTm=-qa+?z2h7&D#5fQt@Ix48Q;k
zzyJ)u01UtY48Q;kz<{v^l)h(TR{C1k{bTBsN~XM%_4S18#V6b!;g!NV0=!c1^_|Qv
zPw6;m*3v7Tf7;lqG)jj77=Qt31H8;{DYS&O=M+LbLVO?$z<~7zc>Nu(_w`t>Bk_Lq
zZ<<H$eMt1CUh$zE7=Qs7fC0xgVAgmkp&p}czWsCVcptaiHmx)|PLvM=j%c8U-?Wd<
zWoHv!>AC6mVZ$Co6VMtDR=w}0SG{IEpS4$yBVJR`CVmY3e0+dMyA3&QAUvQS;5@MU
zI>67B6OqCI48Q;kzyJ)u01Uu@asztJ*Ti}`yp4<3dmM_@<(1uet6Zn4ag<l`)_<&u
zSNRo)?HqE~%MhnyyH$t|24DaNU;qYS00v+H24Fx>1FWvkZggDYbzjZi*7F$tTUo5E
zH|y+iM0MX^v+lG_)GGd4-@s?m23(7W<QQEyh4MzL`k7w!>NStLqn0DxbNNbPZ=N8j
z&-rR4qJ;q%fB_hQftL-$HS}2QpYo78I$2%MMz8mytSZOZ|Gq~13kNK6IYL<X@bXm;
zTQC3vFaQHE00S@p1I}&$^O9IIFA2OmhIgy4AFTiVX;gexdq@5H7*M({125_8yeQW=
zGUW>8mHHfAlmY|#7?3)T^cr`fPqZk-*$lA$elxG}iTjb-?|-9yhYID2<HR-$z<~7z
z^!i=EOF7mXM+7hc126ysFaQHE00S@p126ys&TK&Kd#cXt3~h%27;r2Da-YLER;$q(
z7=Qs7fB_hQ0T_S*7=Qs7fB_hQ0V@os9nWM%tf&(P^j;UJ)^-1wI;E1SS+r`NBGh}g
z{T|OiB6u)s-6Nc_FaQHE00S@p126ysFaQHEpw&Q3zouLE``jYDQ>zE9`AfTbR&N*h
zoJjG{4ogS9UpbX%<%LC_HK#mCB)svfimj632E`NJSz<@}?oaE65Z-w8;^0=kDM^G^
zTIE~o^#S>fF`#xFIBWT|-bXaXF8igr)c^JICC^7xE7@-;B7^}LfB_hQ0T_S*7~t7J
z)^Agfe^fb!It$g#vnS-!`uCAIztQWGOO?+`-xd7wM)A6`HEQSUaLp&4@<LXQztM3t
z?VMJ6w_7TeqSSo6X6-lHhuZQ-l<zoiYvrSqcb4zA`slAQ36xjz{(a5r`kiXy66VV_
zi)znb<kQE1*899x=hy4^iCmjfzeA1MuU9+8<>(!+T#w>(YTtL!JD)f%wcFR5#5X#<
zCXA}3*6|Fy=jZM3Cm~+7$D7rB!u3JzdB9sfqy8>Z>Ot%8;zrX(Je?F6NO`5u?|7y7
z{upoDtc~YT8jo;2=9O0euGU(=QT<)%brk8f8Q|r2<gkmGw|mB1CXPn4-hbw$9&NoD
zxnTeXU;qYS00v%Vz^MBmD3|Ybr<ESR_p5)?JaX?t@~Xaued-MqIQ?eYNuMm@>{M^i
zoSdGUejhgMK{Ns3!ODIh)!VJQe<Br^+Wkh|hcv5PUY~ymar658$x-6gdwhgd<*6Od
zW_3QZ`;pq~B=VVSz^w6=yw)Qe4PjnTM<@BQNQ3xZPpMk+^8ChLoV@;Bl-Vx9-!?Ua
z@XlQG@`ub`oJl}<P|^>C`+KZV?|I01Esxdlq*liz{H_ql`gPl^dEEpbAPlf%fVJy|
zQa4J+t+Jk;_xEKHzhk`jjN{1Gaa4L;M0yy20T_S*7~si(S<gRtaUyp-pwho<l>WYx
zdOpO9PsGWFfx3aVw>mXFLmZWCSVj~u00TT5FnWBG+}{^+)A8*3@3EVGUtF(obboJy
zGyFfYafe1%XTD}rOT8T_{T+_<`)k&nwu!>S%O3RZm%NpS{_t{tFx$uHUsn5V=Dc9a
zE6tv-SMT3xym}Ynmiv2=)pU9vPkZI*eO<Qd{W8|;UDoKoroAV44;)^@{cZPJ?m6ds
zclCD9^`C!-7ad+=_#5sy=6Me<{h#f(-E+F<^voLet@rR>7P{Rgx%ECycFP-5e$Lhk
zzy3+>CQ!KJ-i(mKJx9$WX0hwG{!_*u>P#Jt`%bKntFY_>HJS<!ukOjWEFCs{<IStY
z9(AwK<N_yV2CRs#aGuJg$>IH@x|SuE#{Mw=7SW6Us6gU_dSB$rzmw#!@yhUylXJ&y
z&uf!Q2_6629h-O~hj+awY8GeZ=u&z@zXWO(|JQJX|I_29|IRC$o)NXw8eZj16}=BU
z387X|h(G_=$(Ov_CsL&h+C406gHIymm2G!4sOR4GF}0e)abTS_)?Yi+`-^ZQChKt{
zK79<tkNtUb*sUO{l={^E&r-zwQ19)vD}vt2JNpE&M<F7u*N<VdChdGyF_!90AwE!F
z1I61#RQ29;gYw4CuN|1;Up|ORQD0lul@|Eyh3A1>4~bG*6|q;>ah<x3j3r8`cf5_u
z)!Ec;>~(4t^==O<<;6c)P_5Z-F+?p{X@?L;y{2O_OLe|YdF6l+Phw(*hEYeOaD2qv
zex|6;OxrzbG{yBQ-Od;B>8(<CW2qFCmMf%R`uUaUH@`_BW>ctNEqy!mKHs}e646Sv
z^>@ASje9%)bmFY+y{mP~mr+T?YCbDJW>)E$X@qx9-d?ostg)GdS60v8^5ngo&xu)7
zd)^}7s|^%BSAJHzj0mbXUtKHqvSOgY+JV_mPlr*p)Y^`aS`oi`Ur#2yGPT*>ThaZW
z6Rou8@w)3Ttv-I1NO<AIbe|PVFFqi=(RX9x<xdCRC63A-WBP6MC=pJKqPYK_kB(d1
zcKCBDmr3gr4_Es*m9oyB?LFg{j!&noQS9HU?aO7|wtO4@h!{ny+NHK$y`F1G-49@W
zzt-nOTJx(terofbnCsu_s;e)tikDZ`s~+roh44nH=Z$KQ2l5$f;B?bIfyHYi5ND@7
z%gdV7PoMQ&1mTtWTxYgiQ1&WON_#%6`aI9<rPsed=u3I!mF$md7oQzYwbCm8s22wG
zFrf805G(n4`+a8C`z=P-bNAD&lbUCoBwDFQUq%`H7|?s1rCH_uTHN)u#F-(4cXpUq
zrfH$~Ly5D~=rd;GFBdO7UHX(LrO|O&RsM#7uXh{JCyB7m^2B3gu+JUhXk50x?Bt-&
z&Jm+%uU}1@v*63>`x1!NeD`+dgp=h{iBar5zhUd6A4XHNxMXCzCw}YWsZ!cIPI`a7
zZ|K^$>b#atc;%-0Uk*As;346ic`km}Y}@8Y!W%1BDOj@5*+2qX>w(w%twP*Nf3LgW
z4sWr%|6T$r^Pu<lhurc4r^Q}&x$93HopOD!md+#U+{RMBz944PY!9uT%xQY_`ZK~C
zUEbLDs8{z4BA0r3js87h_o!ZJR4MhUUtCVVt9dV8E#OD=qPTpy+iIVyZJN~j{KDrx
zgm?CcA6F&qy{ptwDfQm3&`%n#DphsO2a$v~^7cLKGPv_J-<$g>@ASI4JD^MCUdkKw
z`kgJDH{f3eW^Ao+`Rng<*_@5`j=$Hnz5!d^E)!O1)qBLO%ztR_xEb|%r+3cOoqaTc
z@=B$@n^{Xgx6rc5&zq%Cdo=2}8&$4P$i1}Pd6KAH8Wo4qa!0*r)!E(u4z-F(<7GAd
z$1MwbxvWT`MzLPM8M^}Flc-kS|7zMB+v?vYMp3WdbK-JXJ?=tW_P$5wCneUXdiD_I
zl|AFySJ@bIhJb&05RSK1Jbte7=xsL>soqrDp46^=Jd3AYrMz%_Xjb}^s6Z-3r5?;m
zUua;>X+yrbOL?PJ^V8~GZ$yzYy@nKtqgrWJe-YQi>fgOmarqCO7xkjkexj9PA9A-@
z8Rw2T^qId(>``t%Ksp#uV<7PB*RS-i6G)Aw)N>`d`_;4&dH%lv-k)0<H9pv^a)ot?
zc%{_eC%o(*GpJGbIxQn8uRMEb)%09(4=Ascnx|W8|Np<DS^K{(5jWbDJd@AnXv|xU
z&RI7pw@si~miKJC)NyzJn-qix^aK0nfoo^CRrBUurmWDc^VO{Dkhq?zW53>1IyHqV
zrMMikx6?n*`?&VKdxRD0)&5JXdsJ`u+8rvFdezHYIli0wEqC94np#cXyouv=wD$uJ
zxlH`(ogQI?7k2)BeMZIPXu>;N{I)J~*Yn53DvJBH((O@S*S@;`m&e3Wc(26#V*W?s
zh+Hc5#d`V<wbHMydUlWU!staIS^EwLQ}AUER`oOSanL&stLJ4v{r^5c#+?sp<{Y{&
zgYe41R|=OYHuVMJm0HiUwAxck_gkOS=4e#9PJ`glsCo0H$}QRV$CJC}|4HOhh?f<X
z_Z#r&+kl5uFS61uv*O_G{YfF-m6zj6wObKP?9nWLLOt_G4a>PdxwwsYcG$9f$INNP
zY-UmL2eN5E>N%)c`vp7uPOOisu<Qa=O0DA~h5SMstWvl>^3Jwtot7S3a*gmtv*$x&
zwI3obz73?umK}97cMLU}dRzKz%A_V&zFJPSa#qnMGg5s1plYdiyVRCv)O(C(m3v}d
zVyo(NZV<C5ey@`CZMFSE?|dKSEVCwZ%WF2PsI@(f%hlP`ZtQicH|LNhpRai*gz`$W
ze&5TjN9%DmD}Tj8dDB;XnA;|oX2m70r)IakU7}qdP~JE{!N10p3HPbq6#G%T?a{|0
z^4?dOR4?Y)GId^jo-E2MT^2O0|7NvR${WWvSUUKJHtCdi?s_`N_r%c*YL9CBx!Qc?
z9)7Woj7=nZQEePX=WAK&SZvs-6VzxbZO6pSfECdd&Qp+057`|bdGB`jqxPuucx(N8
zO=*3+rq_F31gpw3dLE2fz3WC4TNX{6opSs6v+`qRm7bYK%%a@5knZ0Ntmv3)MLdn6
zdh_3Flm1le`s8Nk*Sapfv=M7l$el;n+9>~7`_9^VcJx0GNv+~PBf=-IGO+*7TZ0nr
zCR5g!l)v98pNJ=vH!ke{N1fd<QB*Epb!@EM=kE3R=9F;{snImLjoBV2y?bZgB%rw-
zd^a{;{&e77qLfDW2dVN)-rwZbYtLh<l~VPXy}u*1#U6c}LU`pL-F~hep6@9!n_7=k
z{Mes2husRItaD81*W-^oze1H#s~@@f*RBY9EAQ+RL@lN2k^6oe>*=BkU7A*KaXjIb
zYR9*!eQ#E(Ul_e#DsG|vweAOfm$rUP<g!eS$(KftJ58i`_s^qG!z+gnR%z6{eS%lL
zY1RE$z3bJwj<VifQQpf2q}H_`=RN=P(T9@=uPkvoU{lQ0D58}wTa(`U<@uYPW<{n^
zR=KR`o9+B#GpN;+HmNOz*8979=a+jwr>XnQL-S)3D6ib|;fJ{zE`LPD)OIv;<Y>|3
zRGH#uQ>apEtyi3XQi;xmj}1(sN+~Xf*X_V(FFX(AdPoHF+77GYZdr0^><{B_5oh6m
zm^+`7RyT=K8hw1^&Wk3b6VKz}Xxk0e4$OXfI*jth;PfAtb?q5Rd8bwDM;cYH*7H~C
zctoK-qsJBH|2p}Scl$)j3WauydC70+?U)_Olo!ffcTt>9+-_|1WkBxzNFiMnuX63L
z$Lyl^X!LyVti6ZA%X$KG<5c^6hLwC*vOlU_e0DfBigNEGI7>RE^K7t|-l*pjQsoNY
z5=wt(3F+0&hnamjsAQh1;Z%w(mu(%<Y)u3eSZ3=Yk20%6iL+5SPOQ?X>#kILmmPGQ
zTxZ`k>S)yaeNm&!n|6PZt@Qh+s8N*a2Xg0aPaj$T@yxnssa#6cYu5g&7q>V6a()`+
zm0HKav6f%$_ny`Hj9&lSs{4j&-`AEpf0e$ESSYkamhUHVgcVA)Pp|#k&lL5UX}d>3
zWghJLzu6-Em*88eL@x^WO?ao)`Le9e?-e*N;Mvk!lvhscJK1m4>klcf<aHfOaonR`
zwCe2ce~0qMJhPWx|Nfva)tmONqvFp4deunVbvTPyO>taGw`b+(QhGwa1PYq#p-NEn
zu2tVYp}aDpNSR(kio{W~Xw?00Udt_fuKcWa84;9M>a~7>UgapgZt9huwfhm`ah6is
zL2m>;uG;JoWrcc=vo*UsVSE_xoS&3fqw3j1ls9(W)_=<QL!GHw>UCWZm!tIl5i9Af
z{(U3%DBd1$>iofP*JTpcdC@a$S+)CVgcs(1=c!xmce9C7s$Jhqd|dU~ejauC$0_q3
z5!R_!d(A3G?|G-i<(btE)MKuJcW-A-I9WcGm_?&om)f<DXYsVFRBsv`_tZ&aTZBz~
zLd>SUcJ0}AspIbcH!1JzI_}pQ+ZKmVDcbX(_vdowb4NOP^WTqbdaG~9p`13esCR$R
zTAtkTHLRzzH(p%t&-J>$WmP$~j%?}Sza^SF8||GZQuoue?$>+XQ{InT*Ub}FE_(Y3
z1&#4wub=DnxiDT^icNoyJyCFaEY(V*-d{AjT&u=W=yjYYru(c|dhr3_g+{N>Cam*p
z^muEb{Kd_t-Uu!bMtNgw;*A{M^`fZR)T`eM<+u#)Jk9s!ek#Qt@#Ctby?2$eMyswr
zTI=O)K4*LJYMmd=?EFTLUl-0RkSzlnrVsLH&@+dPSF&YOc{DxSd&VyvpH6w>(;ZF9
z?XI6eS)*_S#t(>^9B}+0(VM;VRXkCw(mi4|g?554V5Nb+hv#*Fe)1_Xn&LKwrnWiq
z^mY_c%cs}cT^@QRobbX*f0Qg*r%Mb`%QLxs!@jH=L98aP?<qB2RjTTm4<d=9(yQyZ
zPF+XF61^#$w?^NeVXb_l-j`Qfu3qyfnEgDTwfS6>#>4u$9#Zk~_8yuz-g-^PWR~iD
zoAOSj@84TJU#8S}8NK)O^4?Ish)-{ox*JP*W&E;TJ=c7dNO>i%-i5fO=G9_-|E!`-
zW~BK1L9L>_^~|L{cfTFpVtN0)gf}W3f3mJ2C2#&2PPJ0_-4C*5U~SUfxTghEDDS*`
zWJjAib5bd9{J3R7FP9Z5ly~yFeu-4vdX2XrpXHyrdP6$Zn@Q^v4_Es*mGa6CoeF%@
zb?sBiJEi&&>-(*mH(u?x;m)~#e)a6}bHXa?6d(VmtwKqgR#rW5wbbq4qBg6j_V_5x
zCpCXD>-!sY^Ip~9?TyrEYJFa#^tc?37<_ul!8FPnt-7A+UGIROkL14cBA)Wj;_V`;
zdT+WxS)<(ZK|JrMZ`BT^HhdpN?NRGEcdPU3HSdkR`B~(8H!9t)lh>5`vUnx||MFmz
zXSL<Bvfh%tI3`XxTDn!rBkCyBYdz>6zAbJ`82ps*#@1)EW=t)WO!T7Ke$Q&Yj&~M^
zeHW9<W{*|o<b9spHkZvP8s+71OWTrCPvVJDoI{#?zUG|}$_veYPsgbG^tx|nb~%YR
zmi&6sBb4&WA?4?6t?=ug)My%YJrK&})xTA78+G5~kjuoc-surW^k&<%PD_t1xkltt
z?R<|ye`1CGuI9aXwSXT{%T1G7pI`Xghgijg;mvk7TD_3+O1;KiT2+qP^T6tS;(7Je
zZnJV8i{lWFbJV*1*}EMx+($p9(Z46`9@Q&NHT4=-Aa1YJw$|~TtUqs%iqEWh-;J(^
z_jM0=y+4#5TXxjV+%Z%srOr#O``;dSD<Mb0ZPZby_Wcc3@=48y!OMQ$`V(*CPQ~%I
zO|7wXnC%eZokqQ9#`^q<LVQ}U=TiCa%?K&nbJRS-JLS%&BIU#E{q+v^o;s@LBf=|t
zjOn-0qeM92mA)N%pYL5KiReYC{-gJOHnYq7dFeNAeez%@0j=@iyRq@|rvvX2vnbV{
zrS`M(J6diWJ|XwgcIQc=j?R@)%NLEVpFnvb@5kdr^J};Fxy4hZw0C~#J<pO}<=Kl@
z>*u_!qb0<P?@4RECk5|%uTw2Ge*<s({VRCQ4*M*I@<P_;eMLNM&c}=6!uD~|`g=fa
zexsi2tBps<r*@t>*7J3CSv2L#3l|CNT)QIZt-P~O5MC+Fm(MHBzK^AMeVxBwpHVS6
zn&?I0y1*(E7mRDurSK(c6(f8bZ>5)V*vw}1?HmKEr-u?&$m@M~dvVWQS?t|rBhraI
z>U}<Z+<8gV!-qL-ymQR4B31k+JR{D=l&(3i9_m}jCZ^u~L}__jM)y1r)8r}9i{&1E
zv5t&QB)l`v#qXMJ+Z;)hQmKEl(kq>p&))lvR(-B@y*)?EuUEhE=;IN2@2gB|HMO=!
z>G(vc^m@JbE>%vIpy*wzzI{UVrqSbZjVgaqiOz+O4NRh1skJ`|`MYlGKV|%(&Xjj5
zo$t|Jdcli5${pXNG@ZTrc>R3Bk@Ba`EY4%oi^6j}N2S*HIOOKvvUJ$+jW@3n-pSiI
zW_$6nI^I+7>rSf@BVwCX{)6z&6LbArU3K*(yffS5q<8Pkn}k<Zu2QgMp|gR6SI#;9
zR!l;{K*B4hwAozw#r+V%J4>7n*c3B0iWo(u*HPZTm(~70Q|d`=`Wox69qRo>IFU<x
z^|Whyy2^RKBq~L(jAe5UG)khpQ0aX%*3$Dnf8^ZpIp)s15J&Z<y?!kB^CctOJ@H!~
zPu0?1J%!JepVclSg6d6s^_l&-(f5<(w#Vvqk<`|&^n4y(w$R3@ixQ|(@-qJ3UYxa#
zZ0X^@C7K$=di`eX3W!glR#WZuhqZij3oV=cyjcomog=Gt%AIF;3Uw6jzw_3hguBVq
zD2lJgQtK<J-7l_3YTMp*!`VKsG`P8&SJ_0$J0rI4^bDV#OtrG><_Rkoz5RrmMYX?c
zrTRDa8*4!Bdz)6LvwHqDt@Rr{pQusgZ$DGiXQu5QHH-dk>l&}gnL@Nu?ODUx`VK-o
zdXEEE>qqZ=TGxYdwETMewO3w;ZQ-5c1D_CQr9Cg=&yAiprch{!EZ<M!h~8}R+q%eI
z&mR-Hy!+?Tr{R@Dh*1=K6}B6ftFx)y*y~g;3gw#3F{NLRKl1zv1+DcUb=<}CxJhl_
z{dDW3<{2l6QVQ?K%ca`);<UbhFZfFB+Q+ka+Er>4g?gl@b)9D3^Nao1Tb~P`|NHu7
zVt>4DzuPFg_5RK0sMX}vpS`%%-p8<4p4QKwRH^kw(6&^<8cXKhpIWSQ8sU|rzOH?B
z{V$J+)zsQQjLsi8E%vg@U4OzW<*t9CS30AQhfu!K-<zzZpH;NUj1-?gDDTwz`&w>(
z;ka8x*7}cBJSF@7cyibLKZ#K^>wdRh>sFdwj?(cvtflAuIk6Cb&>Mk|t2TQ?m9jkX
z7#Zwyhu9;pzX!x|OWogLeLt`33yd7I@tb2_E2&wu_q|zYV9jYmzPU?zC+p+yb7j}A
z^})JCVm9p^cYF>5S`6qlu2f45YVxj_P~xL){Zoiq+M5?j>htYwv#Q@L{e&1zHBOu{
z&$Pai{YJh1keW?rbd5GUo&nz91Gg%Etvz9T?-%0^SJ*&!XLz4x$Adht6JDAAdEUph
z@7*JyG7sYZ&)WP;;`p>~U&;QccJbNaly~;4<FfMPk|fGHje4&~Yd_#6|I8mQ-PrAw
zLe$dgc9PHXPhGtso$^ZF*O?N>Z`6CaMwMG;>mrXbt3wHI)N9^%tIMhPMZWwyNe&yY
zY&H8(rw8#RZCa`II+{MR{^OZ-&r)8g^?fp{^XvUSlD+cO_LK3US?N=v0;$my;*%P$
zAhloZ^QoP+^X%w<Ad(6c;t`l<_R{O$AM~YaDU_p@R^5l-y<WAx#rb&sp0OAA{>R&n
zU8<ea<|yoO(|_lcP0t9cj3T2(ZC#i}c;~*MYu~E#S~}62;`5EyZLe#61Gc(dCTgiy
zyVZ_!&?}!*JFJ=)Qz|aG`^_4UBUO*wc<rT|m7`1P3H=hNQraueQ9n2OdAw2O|61Jj
zwZxergjY_TG`2<9#3zJT>is^tRCz|7fA>nvFXn$Fj!04M{V=8TPaJM(TT<#tJW)!e
z?=h*J_n5x*ym`CAu>^dz2c`YhcXPkx?)y(uy(rYLbp5l~(vkZ-E+-PRDa5OmM$N0B
zG#<I{G4Qr-*IM_3zDrv_CcJXmoCRN2-<Lpm<)`074(@+Gmhi^1cY6+f)-ay%PO0Aq
ztna_`LCu^)_hk^?xMpUL9SvKg6JFW%!Z+^i{L_iRc>l@cW){t+0>7_Zu|=_jEUK15
zIa--{aLTl%F3+hw&TdqpVXxHZR4>jib?45vCDW-|D*awAQ)BX_(c?}NtEn_@z0$W_
zwsl0aH4)Tm%5BfGgKm@S?7K$wqTG6u^7lLC6Y+%VMY;8=O(*r<g?PW(dRWh=bUkAs
zy<Y7K@;*LlO3??z(a6gD40;_Wz00{=)@{qT;g1NfES5gjd;WxE!Yk#@U#T{o+WsPU
zT~4FYsg0Yrd=(b?Wk%o3q*__OsMqO%83k-|*<s+~`}H5^wt>#+;j7dAt92Sxz~*SY
zJ>aX1#(VPHKw~`UeSPJ19Z|XSUa3u|b{<E2`P7b&)GJ@Vt9dV8E#ODgQt$Oeg?_Bp
zyxc~YGe5z<#+3>8sooSk@J6e~DM{6<_Bp#$KHiRx5U)||w%9Az?0MpZ`pkZQqIUd)
zR6gOj>BXe=iHECwoJy@C>+2LMjnC@)-&*T8>VAP#xq9_yl=Er>ysftr|71b6X1~P{
zrF^vo>-VbtyTj4*>2+lM^5vre+e0ZYoI{#?zUG|}$}6>cR=fU;y?k1~2d&P(qGPTV
z@id0=%HH!Ewm$k{H06~>&s(S0b<R=C5uTqLMZw#Kfv<NP&?kwoLbdb4u#)d|(>{U4
zYa|ffIDSCX<bdN3iCGl-pHj|$KV}ixA4c@1(l}U4&&oV^YQ5NdF2L*OuASXh&6{_b
z^3M3NKW`4Z6-2dCZNHG4FS^jBX$2R@6Q#6QzSj9GjGp&IxUZ{}Bdc}Foo9Fo<%J)&
zEa>I3B8Bowx$i5fO=qwFN`21j-`PUkTL0d*I=|HYVAl8R_1s-+Ia2*x@BLEGL0RA5
zcvY#YYd(l1a;djJd*w|`_gS&@;se4f&3?Wr)MwA9QJ?F*UX$K&wk&llHtf_1YBZ(d
zRyx0c(dnh)Rl2|FlIykGUoT|ijX^2C-3NNTX5*CyzIUBEH=>};Xe#XwM$JQGRJs0c
z>l&}gnL>DHxrbk@BV!YZ(&MEzy;1Wbn^kV%bLD5X%ZQ-7Qtff$HQ)X_Zw*Sgn@o8l
z>+2Xu#i#dkORfDcdQnK$zQe&(D_NNj%<ZR7yn-u45#DLk`p;&StJk{pD5ulXV@s|P
zwH^i!+gR8uI)XSGYj)e)CEE1?WsQ2j$6-}@v57Zwc-M=fyfOQ7P{}-1!zr&U-Y%l5
z_of?^S4z#F%liJ6m*YycTM<lnrP6uW#J?|3%u8%lea;P{mG+La(skwD9(OAtN5O4W
zE)#|~+u3OKLMoS?T^3FG^1?-;l_lC7t@?kHX%V?R<TCNAcY1^oDT<Ge(rvTHeJhPm
zY5KE=R!z?p_ke=0_F(TlI#xfo`nj^)`jz%qA^r4wgKQ^bZxf>_)F-A<^U#XR71xVx
zbq0*?Cq|cV_IzW){d}bieD=cgK(2?xD7GpwBDPuOKL~Hsd;Mf}XTd*n=^d9)Uagfs
z(0tDd+KgtMJRc3ZQ!%fNS2ntQ@b%`lM8X>X*;9=9WXZ^OPyE)$Q(jpmD0<hbZ=X=H
zo$Y6e`pmT5qgGQKm(km<sVy%peU?dh<&q`cH%zf*63|=^zrNV|dr8;aHnV8-dGPk1
z2MW$^mrS%$Do$SiZnt{fTD|Hw`uz%f<xBMorTf?I*-pPa7)119w#P~D-kCRvm`dYP
zn_lkyN><aUU4N04e7xT87vgT(FsOC^kM9#+S-DEVl7-F&604~8yk#|C=_2VpR@ccQ
zMpLT&A+;iY^}e1=%%W60!hY7y6RCC9h5f9>tu-FA?-OdRkJoj^5I5iJ#F>5kgzKc3
zTVLN=eCgV3Dwlb-Oq~~>CyPokFzCtZX9KgSR(|>3tuv)6J)>gQo|kuM?VZo46ick5
z39k2^QMK$mWbp?>K7K)!@_cmM;<m$|Q@toYZf~u~(|7#d7t}1e4DLM5_vU_T6a&A0
z{Yw8jfz;8st|29F{uxfKrrdFrX1{MO?$>5-XLo$$z1!W7f-idzjsq+6>8OoE?l?16
z)A2fPR2+BWjU~UH^a!QAQ0V_g@#OU(R~qh3rM&S(^J};Fxy4f{UdjHbcJbNa)GFFr
zS8;gRLK~+pN}!IyU!K3&X;x$!<%P?NzS+({HiLp}cxX8MWRcz%pHg1wU%_j3*k>`+
zXv+N_FnYcXv&y$OpR!yJO4G^x`$Q_8(t7nuZ}s1IdevWPaC0}WvWb*e+WX$e=lZTU
zE#;Hf@pL8cZ*uFk=P@;lYOg>0POOisu<QcmjjWD$P#V{}jj~(s-+Yd0rC#HJ^eV^b
z`R>K#tL^_%`OLchH2ppHM8WB?lod*OVQs(O_XqT9pYP^=%iZ^%ro6IeT>B~;W6lu0
zD0q#&onv72^iU#~<Gkm8KKgJH;hprY=gr#{jwM!6@NX15Qd{iN$0>wYrZ(GqE4u%4
z!aIANTk_Z~B8SZ?8r8p~$~F4_jnU<s{oGWq`j#DZn_Oq#HOecu_kJ<%aD@$oR|?~O
zjiS_f$J_ol6ZdEPf0Rm;Qrymm8D%maZph&$mwR(=EAPJIIh9Mf<EeynliHQ9xjr+S
z%B8)yJ^FY=-uo((D&?jEgZ30WkwxWF>+^!>-xGF^>Xk<2^345T8ysDfL6x$|(rG_#
z%}%FUDegZ~+roMzygcu=7pGqLmBr;q`4jJ-)w^y)v1QT39>wYP-d1{lNNIYl@s^oA
zeO3R=r&KAm)+;x^-s@M&t=HMo89fh(S>-QlQa^pxcM(J>eK$5<{&e77ViZSxUHj_#
zUmg=)StztbmhUHV#3;5tn>Ax<sbm8FyNA%!Hb<V`jv~C%==r0yUbhn$jBC@S@FmJC
z3k|F}ZOAuwsa9(BXm$Q!vnK6)Rxy@>@L)tgFuQ-JcJ1R?JnbrFjat{AlbgRmH}6#q
z-rh*9rqXeAQvJkz|FZ?}6i+0)Qfl5>*7vir?z!4HR!k`I(YF36gm;#`+jHo%hVev-
zQs<M@{z`w8ELx{a3~>~iwa$lL^_X=Zz^Hn7U0+ii_uiU^ujFnTO_Y-N=j6P`FLhmG
zZGWeC&eWZKG=VxAy)u@~InXGH@=9LULlNQ@o{#BG*3KWf@wj$&TQzUqWokC3-CtxY
z{r)M+E4AJqvpT<C$6IbcGb^3m=e51^jBd}Q5}gYl8<<3Sp;12Uy$^`bgWk$J`veht
zFzPt)R_?tL^NaZ(i6f3eUY~2G7pq;#Yf~~&N~8PR<n2Z4&KjFZc%@N(8keiHsomJ?
zlvhs73|J9e;XIW~y*+BZKcH7U>9J);-OL?BS>=`wKg`u|`6J3JKW<sj%Vk9h<&`U=
zmM<DzKY^M}z52JfoEAM!l_`EUg_=#hkN+jVp|@jpBvV%D?)CWQlyMKK(G=E~adbX7
z9lv>hzYHp-xPPnNHfr5xaXebzM^~F)9Ea4l)cg^w?+<z-@Nv~<kEmSQ^U*tB#S_IU
z-6LjG>%7@Y^Vd7rd+Ml~j|lISnzz~Bac1rFr1go1t9_hGxo3WJ=e5!Mx)ii~Sl9-i
zL<$<~LGAgsvv!^x{SQP^r7W{`kw=-;p+qj_&Zpq4>Es?it@DS=tq18G+kjW#ynts*
zZ&AHzmNoJFmRh&v&W~+W`|8GKjBu%(Mp$S2;JK}TxSUCNV|1ZQ(+Vz*C%kj~fT+m<
z#~%{4G^+o~E%*Lc)85!t|2DCj!usYt;>T4<d+#dcm1ezXA@}$yP1pJR^%)hDqlr=~
zt;ekNYUi0z8sFuzZd<+$e?)kt)O`kC_W$-tFS@}sg=nSH<F9vmqu$Fn-Ly|&@fr!l
z(YU>BR`r{upAgVk56`16|2SpdBchdhpD)*4e`)pcvqYjd^^Oze!NA`QRC$&@plgi`
z0vhY#@6MI{f8(+TquVBB6Qz`@gZ1^_j9PcmQOj+)Z0m?-Ya*yzO7#!5`}b_S)NyzJ
zn^Z5F{ocIzeM4Ti=f5Abi0ls|ywdHbPrQOFL=mI;^U`nL`sBe*;%HQR{wmG4+-soQ
zSABEY)KY1Ea?{`J-e~2H8@X*DJgDghwO0N>^F1qQ)5?)!Hhy!=YbE84O6QAFn_lku
z>}=^qy=c|h-Tw|Xn+?_u%zk=0jB2IP$E!oF^sB3$-J@nv+#ap>pRCT$`~6p?^PwA^
zUfe&#+xs7HJ9epdPMfn*>_h4H-9L{$4X+$RjH1$b^h)n_t#80qx66c8iq|z)x~=s&
zy4LR@rTO)qw_9#`TIXe%TWHzj=gm?m@8o`bHmBpvmCmEDS9-m!qe{yeJZxiOujmM(
zHy!1@|H~`uRS)*PLU^av`=3htm7}Ep?A?wT?xUYlt#lM0Mztef#HY7P-HoMMX;geD
z7Y2ARVDx&2td#GUFCPuq9!k~HsAKtSao5)pXNC}?X;l1<TCUdj!&u3G_Ry;7x#AvB
zwRF_ui*~(ipk}wdU7}qdP-mgm@k|3^?tD&K-6VR^>UC*f?hV#$sr`PoI-l^~WN`YA
z%ewZAq`c9rBR-_{PknB-&!ntV?tCpqr!(ugD6OYy!=TpvKfX_RXST;l@7|d=39p=U
z{H>UTf`NorcHP#0%J@T_siRQsyjW`UDLrrPrT6MOu2a{MvBc47ul-6tH*5Ti(e=n3
z_b!!Au0N~hMSk~o=7f{wQ;FWZ*&_Ux;9IFgFUs{OPPg)MT&Z>|f{E1>$E9{#>-tl?
zzJH)s+`_yJypi|krsDWpmRuV9!}wc7?s>rLdf`Ic<sN>qj*LwtdQ+}Hz0y@!<d+$J
zGn4YlXI@DsgFTZeuZ$>Crq_@nag<lqI<lpQ|CVUVE7d+PU?tzIqD^L``20b6rC#@y
z#pURAePq2HrSH`!ZIAEfe#_nWpQch&TCQ*(*=$PveaGtl;bjYLoVqB1+T-zal}B&8
znMhe>#MYgj;nS0;S@d^X*LY3N6v8@tjOn-0qeM92oxIL#Wc2&vMwf5Z{P9Y!_pGL`
zT%};iLT3XB?^L?}0PE?OHL0II>$?cTD@&XX*c3B0im=Yct4dW}^Fbt0OIFWorEzUP
zQ`Bds?H=WgQoqZ2**|sC*cM?EpZw$gVc~P-XSK_SpjK09e-fwfI_}pQ+ZKmVwG@}9
zbX)1)iH@3HsW-FJ`wyKL^`g^$!W-MB)>t~sc8Ksw-ku-Xi&wnvvf6E_ey?`kYOVRC
z;!?Yxm2vE9<KXRgi#Xmw18Ytj^37e!E7gu`VkO_c6YJwDEW1E?r~B>j7R&qZCA`t>
zae-3T$%WG{KNz|&iSSCPaZ2{uZ}xmzQuSH=caPTk$A@O6Pl*boM$z7JWbN}axqZXF
ztQ$eJQt7&T_R=@~J@!Pw>9Ld-^12?o5H~NLrQ$SteQl%4*L&SMv&s{XYv6U;s`1eF
z>Q#FD#DA~wzRlZvdrIS-Lz;ZP=A96#maP4Lkc#J!%fzqV=@CYZrVz)JHk&KIxF13s
zotq}LKELp}4>5~^U$e-{{RgdaBrX`&rc2>V)GBH{UcBD-634A{K228AtNor(o6o3m
zhE~rHptk-zvzK1~{-7`AmAuX0Z&kcnuaiRlN`I6rTBl14(MrAELz7!hdEzlL*yj#0
zi}RBbYg9dZi0Vb7=GQj6T(f@93H1p5)@n}73|J9e;XHL#TJ0b8JBNXOSMy%HTELH}
zrQCU1#5T>Ihf=72eyKZmzAc$f?Xh0J8M^}Flc*GPtbVuJ&{}C!ikJL`-j3OkOedA-
zT=>|)B-uP1G5GY9gK5+rh4wfbYplO^sP`A)1XSihm`8AAwNAP73{R2lxADt%^<48+
zB9&rO_nC+0$0kstIpF6bxv#v4r}pS<=K<Oa11uZxh&s2i)UPjyUQ91myOP(YWMVY!
zy|=2@=Slhdo$`rzLiM812lm!Hd?k0&Xu>P)jT2_|^ShsJozy(z<Uig65srtzwMlp5
zo)%1@Mp3B8Sy{g9`-iLM&n8;AW@e8a4O^rWz4*tNxGKNr&SP_Q&RtpT-DV@wiPfAo
zXTg`%_x(Ti?mXV^>FOVN5+a5WV`C^qiJBU!=HYuAQ;C?XC5D(o(Hg2%sUf6>swzP#
zMO!hHlFDH0H!)ACAt90wnPnoAgorf1UtgYgo|pYNz2}@eopbN{$H&@x?KOP%+UvW|
za8G7u<zl6rpF+R0*3j{7z2jM*eeK|P@@m7Ye(=<mf1mb2UhP<4_lu^oGj!)S-g%_U
zth`!rufKKPy6d4c^J>NQhV+>6>({2_)r#eQ9$V_VhVG+b{|g3g-0pvv=zCt@6MA;)
zx_Q?vy0*Te@1cFWw)xM0dv@r#de61Gw(8TjXUG4@-|E`BYwK={PW^S?o_Ac;<(+Jq
z;IT*sdR%z(2IF^_oWtSozINq|Z71aL!PTC6{QB)C=H=ok{TFw<q5J!JbHyTQ4z*w)
z7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZvfqCQZ
z?6dNqNqOK`^N`Th>hkxF`uoprnV;2$!`8ZX@PNUS^AwL7@PlsmKRGo|@xaqo``!_K
zr)Kqy2b_1-%zF+Ql{Ypnyl27@du=m5uW$6diWJwFe0J~7yL^xr!e8%m|ID4YpOnLv
zQtxT~h8zquB?G;tAAiB!58a-Xi%l7O_znhwfiE-Ax!WxrdptcLU$oNw!_Qp*qYT#k
z{P~UYe@w`W#V=c}d};DAzr1$NMX%<C@X&vEocPEc*XPxS=Vq5a-s=DV;YoAD%CD%1
zFc=61f`R76z`n=cbkI7duT~QnJTzt<G_O+>*#rZ@Krj#tG!z5**8QIO!>Dy@0xOG$
zLi@eW1CJbb%7BS^bH@QE_Py`k;}&M{^gr%7bk?|G+0(rqJZ*m4ck3O?{hYASKE-%2
z5DWwZ!9Xw&3<LwgKrj#tl+VD*8*J16(|mefD*kNib7!>Qaax|@vpcUp@T4t%lhuan
zd~@}h{qG!_S1Uf;_RnA2;nEp-OU0}IdfUYhtT8t$7U%r@kbU=CcR^MuIHujGx4Ygl
zBWoyJ<@lfMyZSW~v--j{ztw;3{gxY-1y(l?<#*~>*_;^oc-Qf3{iF4y94>s~`H$B9
z-TQfpllQ)$?<q^(%8SK2_kQc$m#-X`mx^zHc-_vU)*P2tDn2)Fr6WeonU*&-Hm4el
zT!MjMpg0EhSaRQPufOqOUTs)hc9o~K<KBOJ^s4pm&zcHP`QQ)7oqO_xyi{y#CXYCI
zo8wRT=J>o2p1aD>VefaJl*5{VTO6~_X*2T_k3a0#b#{4ec3vuOu;`ua@7-qPftAHW
zW7lC*`78AI64f%E{@vQPAK$UXUoLz(uMJbaPa1Vq|HIarSN7k*rE&ar<(W_Y^r3Ni
zsp#)vOU3ki7k6H^^V*wc)APo{PWN@+;Gn}s=e6OKL)V{p$IuyhA#Ccu1HSa>d-{L<
zig8(8F|Du0mP@Ula`pY)_cgs%oigLdu{nHX-)|1s>qqb9Df-=gXurR+<NHIN+;@1M
zqPkw`wsj9^Ke=C48#Xr2;V&2n27-ZLpm7;UpDW}-pTl}CrDz>o^4|9kUU2xByf!TK
zJ1W)sce?3wx7uX)kZQ$3bzAA<<=z`BZQUgo_grt8CqBrl6$|wv<Awfxs>bcw;Vl>l
z27-ZLAQ%V+f`P&eRO{=ghK{>1V=>bd40s<q@XP<MZgc0ODQYcz1_QxB%VJ=MjW#`f
zonb?>z%O_>?55e>F6;PK9vD2-M;)}RZW7tH7zVz7#1+p?o}IU9EKG=f{=fsK{A%}U
zd0^dmNN5eQ!Wn3P#;bEr+;2qQR9NBk6dhhueg7c+9&dfiMH3K;!9Xxj?+mPW*Ae%w
z`N^AEW8*1ZM$PKje|%nPxc1;xmkd7p**wL5`(HS%`(sn`z~XqQ??Nr~`!DJD694eR
z;d8Uz|H>N*>pQQo7Yqag4adOe^V-~V%WF^Od%gAOfNg(2JA?Dz+H>?1pAOH;#X|pX
zRJTrV-oM+6v+_b%*XKvt@0yF;a1{`igMn79`u(BfmM?U9Jr4{XT2dV}m$N_R&c`-c
zsq0&LePMIqCGrRcf`MS5z8LWLQ5`ysIkMOE_BAaP>x<9SPM0%IJO9l2Z)AtI9{u8B
z-&n4uSoGXW?G=(6`|k;x>hr1k_WR36JMI5!&5XQMEVK_X9t;En!9Xw&3<LwgK&3Kp
z(bdmAfA}w6%&Q#_`o{Ig-?8NfSzvYZP^sr6;tB?WfnXpQD35_rtMAbJ_9NcS0~>~i
z@^+>2R~7ra(dB0H;_rXc>*f_EWbl;@XPmIy&z{dy>_4&lj_dt+NFMm#c_?=uV%>jd
zpwRcu8f&~-e^0Vn<Noi}5VOHRFc1s`1LZRC(`_ftm_59GP2iV#*yR3oIv(}?@mXNQ
z^6=Tn9VcCM)1nNPTgT<D>)h>@jy;|pkXJgs`gpg^Ry|`%UM%i=;)A<Be%6@05SGhV
zb6t1mAII&r@R}#GO2OvJb5rHjP=6=1%fpu+eC1A4@@mJ!K3t>Mxm{-FaM)Vc4jwRg
za$fD|`<(EW75iR!<bfmeQnA$U!!?$EyB~h^{?NTY%7^yZbmQf=IH}RG(x02f##d@R
zREmCM|E^3^_3!W-3>3q_)qlP1;s@54o258u+OWTl*>rJMtvGU*r`J1n)!A98IC$IB
zj=Af-SF)Cdg=)p$!#5Nb@*Cs9Krrxs8A$(K=HVZoKQO=ek*rej|7H*)!9Xw&3<LvT
z#XzBdXE4SWcf6td`+4xtG(3D&t;SgMVc?{0JN)sqv!2cagNLYtI0yPbrFFP!{r9CK
z?&|*c!DIjGwttn@M<Sm8&OrKiEFpjO-}h8HFaLd)V>TEF27-ZLpxPN|sL$uC&2F{t
zW5gc}1Ovf9Fc1s`1I>qlGbVoPq+8!>UsD^Nf7|jOU32HVS*7EjdoBFe6@MO|1%5RT
z&8KoJGn4k;S+Mw>+m@{<6;FTm{C&=BvwTfp!|+g<)lz&^lb^^Rzw+6RHG#oHbF72n
zPeN!`A_F&HbI^vr`($8VX;_I|MNCc4z@<;$)Bo#NjLV9}19thtCttsNg_>%?re`=J
zXif}N`oA~&%kRB-%+$l*%jz31Ty*ld$?Yd)%?0}mx@P4eBc^77!9yj~L38SnkxTV3
zu*#LYPkrIw*;!!l&>ZSunHR5Jc<AVtvVQwtIIa6*Q}S~0^gr%7bk?|GSzz!Gbr9#E
zrS*Z!TQ7IckJ_(NQwV#Xx6)7hFMc}<tQ!y2b5m@ZNJpH0<z=^T->L>4yZZOXef^IM
zGU)fizV)YnpR@SJX|?{I*z&h-eg2r$KF9)thp2<5KL>}l9{u8B-&n4uQn1*6m#W(I
z(%&Tu`Hn6Aa^cH)rSi};PuuVs3{)orH*R>~`=7t~XkIQ>XVwvQFwigz?6~jtt3P|&
zC)u}8uG#Cmt4>*p#s2<$=#}^H@HEVBhSgvo7zhTMKLhjM+H>?1pAOGj3O4^8jSPc<
zU?3O>28v>!-Aaqj8vI(`s!aUnH#+Tb+Q5-{L*t^``uuI~Ij?7d!9&zR^`C>HdQ&Jh
zI0NTg)#}#4kB-VK9k=}1O*ig(^5c1>;<G!iKk%e2ev?%@Uhwm$&bj!yfqAvysMUAq
zeftsb=7sQqTjrhh;$wfystq0d`}gRv_mbsi<?y%-e|YU~GbZPO!9&zRoP+2C|Gf|B
zqb(iJ_(eADrMJf9^@aD}duFeH-TZM@skq!vZft+k4S7~9Hgp~{9`11d>US>80xO$`
z{YF0U&yT-7HLnd@dOwN$gMnZm7zhR$mVsSXdT!iSmru@%#m4$OgX>r5_4Lh~tyB~E
z)ja(6JHKCU`#JCCfz`=F>kUp`VfWWM)C2|(Q3r7jq7RhU2O9QvRK3=3+-d$wZ+vZf
z78pE49kkqYP`w>2A`b?FfncD17`XMxPxkrf*$?wlv3|UVm0%zk2nK?IU?3O>27-ZL
z;0p#C>+{s+;wS1R7$}8-tvhY7>T4&xmDd+8*Xou2Z>&2quQu$p&7Pf~*l1i{t@x+)
zez5bt+rE`oE4F&@$`jw}@j+g#IAzVp?%8{#33;{Rvj5y~&1d$UoClW1!(nr_8N1nu
zBl5ui&O<5Bq?W9E#a+{TjlE=I9vD27QU@*h%tY?h%Yg6S{A|h<!}eTZY!=uwJhVUK
z)ww6`HzIFp+-_Qzix2tdw7fPv_}BqAE<5stEHHRz%sQyvP8*R21HnKr5DWwZ!9Xw&
z3<LwgKrj#t1Ovf9Fz{stHhb>XH6A%?K~~@RWiKHWVjzF3-x2>f;O#uHJRS=18sou0
zV=^#ot!oDl7(6*I7yIph;k53LP05=I*O+{E@6Nk?kk=J!`#h(i=7(Id;#&JYGc#{$
zT;|1V7altLrR<b0qh@vNKR&M%9K7vm$K3VaD_L{F2UmOQ@$0vnm^U<*+C#px<NHIN
z+;@0hDjsm&Su^iBWK>?QSm@vDe({1y2R-w|!+E9P^sAQ)JN5NvGI;T@Wsf;)uQ&2y
zaq7N9c06#|F<D)))Vd1&V4xTVs<nSa+zs3RJB~VH?3-82%HfVTzHwORPIL0+ioblc
z)Bdm4%*g8-oBDnk{=c~A7t;&>qx0*1?w`5y_LK6Kg2nPuSG^6-oxjQL-yWAYHl}fZ
zcdOoO{rKt$S-Dv3Jw5dsaxf4K1Ovf9Fc1t>KLZ1Idw%(Ox6aH0OXFeNOXk)5cH>1^
zW8<3N>c94W%Z<w_1-<Vd_{HZREro~b?|hL#FwlG$c>cfxru=I6X?cBNwKH@3$D_6#
zH2k?NuxWT$^4|9kUU2xByjrojd|#ka^XPHm%^Qs0VR8<)A3J&Yl2)U#Qn9h>tK9D=
z)wPc{_1~+9|Hk><fV$R6Q?Ea1)K&ctTWem{SlF%8oA>Yb;;g*Bu&MbA|G_{o5DWwZ
z!9Xw&3<LwEF|hIm+w}i5pPmN>50zX8rB!vqYxh2HrJwd+{B~9>KD+b!15euGH(5);
zc0c^+{h@n*lvf&_)NO}9o_5yLc|&8vS73P_D)n~}%gZKKH5LPhT>k9t5A|D|1qKgM
z2i12D8teV#@bll_&u^@L9e%2Vfrk2?MRjB!Q3eCSKrj#t1OwH<z#k4@qvxuF-^nW-
zv02;iB=q}3$1Pvz@_JrZytPx?i=O(yysS2Sy4QoJ&2Rf|R$sW+TaOOd_V=^1x+1;;
zgMnZm7-*>sY<%IJ!!|s!O-<mJd02MN!GC>ax3O8J;5V)q_2jv~9h)~5w$$^H=Dp>b
zbC-NJZ*kU8nEDI3JO=i8aI5vs9{FxwZCGBmu__n{27-ZLpi~APf9=NUyIi{<Yb-1^
zzR(W_f`MQl7zhS}fnXpQ2nK?IhGU@18K<3p=KMFZrpAWLuxTwXf9uxgk6G=5tg*0Z
zvkmVpoq^8XZt2+L=>d7C`?_y%&|#x%#gwmw_e?lquWiQXm4fre-Pvd5L6h<n|F~p_
z_uBn_N>(bS`Gy<}1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1NF~9
zpF!8GJY>YwEXD7&IXe5(NuOthu=Vg6fBoyt)3ZW&O8><jZ|MGhUMx2DzZ=_F{)+wg
zWa&M!?>7hR^`rOl6br?h{?1)}{rhLruU<0j)YqTMsvW1?`Pe2ab$u(ZR{Z{-U;D{7
z)}5IZ!aFb9dF@TJ>3Ox`gR4FD`1RXO%qty-zx&#iGq#<Or?}1oj~sT&fQdP*|IQp1
zgMnb6QW==@^F#LCZ`}o1V0k=L>b^!?4Z}d8`)iEP-{kghkIO5Chlbhxu-Z}>=&;7{
zw)eCdoi!B}`+h^E>TNfz%f*NMb6Vb1IP=s+^S}4p8%zKFN&gOI`ueBX@3$7xOaJ{u
z?{};*`Rv}EcljXy<)fYUf3;@DQk-(=`V;RMI-?*j<_lp%z0PbY=Q{EX23i6Gr~h%!
zp|i#f%j$}U{=4JENA9>jFBe+^*O6n>Gf?PxS>NMLpXnD74DGY&#>;JSQr_I~l8bw;
zx6BhC<blCMQ`SMHobzgnW5S?wFZ}y+gL3%M(SO>1cDwOeQ{ji}95>+YU8m;tjs5n&
za9a1rrsNHcwY@L@-K~1B_2a82WP#0-hx)$16tY*T?;}NArG5XXq4=t{=TO8O3<Lwg
zKrqm>46JhePxf8?nu!^#>-&+`AK7N^wQKG3W*!(kRAL=8?Wqgz!9Xw&3<LwgK*KOF
zY^`et4;VZ-?|t4%KkdKx?F@dh#|>*=(syoF2pcAvuo?^m1HnKr5DYXA28w+jzj<U^
zeiq(Ws&CA$Jni?d?*3_Z(bdmAfA}w6%nRYD)pzK9`w{Qv%?0g5`uEt?BD<HCi}DW)
z27-ZLAQ)&F3{>0iPDH%b`FkPNwr?%t^hcJ#K*KWdbgu_bo8R``EU<ofsOxU|J6%{;
zd$B$k2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3Q%L<YV$=C^;EddIA+
zvGMkgM{PT3_;Xof<0)N6&Fa{Hd|p?)rq`-dW*j*-uU7Q$wbVCO%I`pz8b>9b&4?)&
zXsHa;{QUWi@_$Uo8w#iW^6is{Zm>=bT)*uGo9;2`wXD8y<qfv!|7kuwFBcE{aE)H)
zcA1$s7Pizni@YnFfg3kG@cqwUd^8WNE*>hosv^2zpkWx;X8tkX{QFPG=Yh?ehb8ZQ
z|KJ6OkI7paHq5CBtBu9L!h0qhvDY@^^M*#Rdv<yF@`JD3X-W>$*E8wAlha(}9&eBQ
zd!OEu@=C?Vs<-eH3<Lv}&%jXwe$egyC#U9t!9&!+|DJ=&Ke-WMFc1s`1J%jE!mqWv
zI3GPWug>Tr>R=!k2nK?IU?3O>27-ZLpk*>}&Q+~$9sKC1yrJ>->3t9AwEOtHRBV~M
z5?KcW!9Xw&3<LwgKrj#tl*>SU|2y*aZLi$CVqGv03<Lwsm4Qnx?z!GFPkfN4_~2?!
zJ%0Um6Z3T*c;v8C22?JVyD#m}cy;cH`;Eve4a?0Z)&&E>Krj#tl*ho;Km76So2@+|
zPjRK&);*y8<bGKpe0Jyc2cERWZ?Y7RzVFk!ezas{RtR@~am!yF{p-P5V`E+ay|TK-
z+jGZWdp$qz^{lCJ`>~UUFKIO@t2SJ-<{zt%=rbj+R=mIaZny4t+oZg)vAnJjs~Vbt
zyPp5W-}hK)Vpdn2e)W=Jr@sD7)>OFe8$Y^jiv{oJ<zhqg9QK2OU?3O>27-ZLAQ%V+
zf`MQl7zhS}fnXpQXs!&ry`W>KV+VbhRT^G0`rg0Yv)upx+&#g+{@|Uy51#)*Ug@~=
ztcSMSdbz$?wctOmKXu3-E*X~B6>nZ`#2X8yzm^A9ClAfF&qZFrKrm2!3_SkcU;l8_
z*vWZQWA!l?5eEaoKrrxS2GZZpEmi*P?0q|&)M;#9?Re<FJ5GG$j_dPk#e?5H^p!rx
zeU!Hpyl~OU<0iMClvN7O>(gU?&(kO8)rK_#w>V~<(`M%Ng<tM2Aq4}$Krj#tG!_H>
z_P=mi_s6E>Ed@)xuU5N$W7Sdk2?m0JU?3O>2AYzAdE@Tvv+|%xc~fDzxj+8w<h!3e
z?VUXE3m*JFLZ#v<T}I97*nfN;7(6s(9n{uu`cLe><9a_Hk~cQi7FRmI-d#uByXGfv
zX0>7O^H%z4|HW@-h49Wlj@xVDHBV%fivC@~hQfw^z1r!%?i(C**yy~maoAed4jwRg
za^BLh*Yx8rxci~ov%sd|VY_KvE<WU+)AHKz*ZbT*bLZ_R<#olhPlg-}1Ovf9Fc1s`
z1HnLZXJC(2em?QDuYH^qi-lenR@?ZUm+idvrrGp7@W1oW-0MH`3kHILrf1;evyQwZ
zf2>W-MkAlR>-qIQ%mTlfhy1O6NBrY}xAX3w^qu(3mE*Je!u~&bs9W3B=VbTaduFeH
z-TZOYVzu2vDka`OF4^I|cE6vJ)fGQF`cM1MZZ|#){O>&IxOM&A$Yox<cHyC;U&_kG
zrux0f@Y}S%_t{kZtNXxUAQ%V+>WhJWJMMC5P0zV`Q{ng89G(5?q|dWr@z@y`WT(D8
zEh~hj-k%P<?)~vA9XUTM7KirPbmQf=I4LiLE#>bOBG3By`xNI`{a<_bBfH#s+lN_w
z<J23EzTmOnPRbe^(|=zo<gfmBx2l~NbsQKB1Ovf9OJ|_c|NX=?zWUy;`uaYse-C(<
zhc7?)%AKa<#bSN=4?9i!{pQ>H-}=Yf+P#;hxb4SxZ1I;1U(Oo~|M0@$bF<$6%2O=%
z+@!BdZ~u7Iwu6R0m(_;rx7}dVJtn=DRU7uW@a7H1?=U&9D<0Wq?zL;}^Jd=EIQ-q$
zuAH&$ggh{Kh&rhMb5Lv_4ZXTCaN~CWv%c?neNX7wsp}S9x9r;bioS>T?b_x)|LxhK
z=juJz>e{ML-<}=+BY&%F>#nW4EjsnreS6+<RhM_NW$Na=1+AX*^GEJI`r(iA!18!#
zL8TI@1_Kqxz=BEh$E<YG?7Ud)FmUmRksT)FrQ#3I-SM7&V;;(j#kMa@JL<MYx8_ZS
z6}M3lP%sb-1OqLcfpYKJEu6MUKNtuGf`MQl7zhS}fnXpQ2nK?I7SBMruP0mTx)xtB
zK|nAN3<Lue&cNrxyIehG?df@7%i<v|%To8hTJALqHofqcyH1*!R~n}Lha3z911*<<
zQu}YE>$lvUjLd_9U?3O>27-ZLAQ%V+f`MS5sTlb9z>XVTeD#NUVDQkCb<kAx-%|a~
zTlxAEr~PG49@sp2C{)+yY}{kBcXpVYH+TF=*U$ex@aTDYrDLIeit%8eJ{ef}`PkK-
z-D5^xSFE(p2_v4${QS1Q_NBQ$ry451`s^KHtL_ZU-r?Ibo_}d>UM?=Vzh9db=1tA3
z73<DmSO^A!fnXpQD20K3H(dA5;oUpb1O^X{SqG)mXy`V72F_df%pIStIwvm{zju4*
zU-#Ye-&tL8?4S18@{p}x%u2=8e${hcx0heeTPimHii`}KGXulg-8Fy2&*$XDVsqvq
zatj86fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+
zYGYvAgN_+<!|NYsm4eHR8G6dT%Pv#X)L0vDi%*_*$9uyDWQDM{^)Vj|1Ovf9Fc1s`
z1HnKr5DWwZ!9Xw&3<LwgKrj#t1Os(vVA&(jn%C`~DS4^b_sM+gzMo9WipBj$@3hUx
zp>O0V);)u;5DWwZ!9Xz35DZkxzcbSixrNPOAQ)&E209E}JYr;r33*^;^UyH6U-?%5
zyv<H~zrEq2tTw!G>)q!bbHJ>uSUhanlj9Hi{m3lv3mz)J5+cH2pz#^_=+1+-+2!q7
zd3|ABfA>0F-}sD#$6z2Bs1ydO<=^*iE^$`MzDFFvKrj#tG&csW*|G1lTb(vFt8d)%
z{gXRfdfwQqRJ>#DgF0-`V_a4$UiOck-JcshBZJLtpBg%!`g*+;cFKGG+R*zO_Je_7
zAQ%V+f`MQl7zhTclY#Gc8hY5E`{!kWU+}PC()=+iT{Jr{7fY>zhWdB-*1LI+wJ%w+
zI4cz&K5@%CdY?EyD;96M;hX&iy)iK>7K?rTes{aup8E8&g;|QF@*DcWKrj#tV4%<Q
z^Zxv!v)j}Zi#smf<kfdBpPLnn8|}T^3eP>VC@U7zztd<expnUqFWhq22YI<TWRr)c
zytV1jyja}thMzvROq<p<DcYC7U?3O>27-ZLAQ%V+f`MQl7zhUHn}O`2`wr}O<lA|D
z<G>fc(e{9+p35r*ue|+->s{Ejbxm`@4<3B^-IqRpDyvp3_I=NC^~!zUVA^MIwI6fi
zGg-NK?A`}FGiu?RS#8+%g=t6Kw&>QpT)e8sHV3TuyJ=amSnT`PrRo*yzmQ&Cf8V&&
z`tP*cs?Sf@{hq8?tnYps_JV<6AQ%V+f`MQl7zhSh1_MW2`I}!%-E>wSSQ-y4qdFqX
zU?3O>27-ZLAQ%V+f`MS5x*6F1>IEmvX*V^m6dd%kZ{5-6XCLI1f~#(E_HUQ%Ju@#B
z)8EOeHhIwKRess`yR-Aars3h76DLnPbN!F<a<RI*QAA%X182T7Z0~=)|4ttG)jSMs
z+wa0}o;D{h6@RtnzuUImVfmU;vDkeJy<i|12nK?IU?3O>27-acWuV$#k2iI^jk`Pb
z@9nDVe%|BnJ!a)C4eOuDuow)qJO;*I_Op|}`_Q<&HoW1%=ZAkVa%>*hxIDDHifr7B
zk6n52_=5(_$O9{jhsLdf@D>aN1HnKr5DWwZ!9Xw&3<LwgKrj#t1Os2iz$G`mv&XtU
zrsjdcL)1Z>gXjaL^?|SIsWBD|1Ovf9Fc1t>I|GyZJ-kZ4x2EPT1*?6(BK}|?7zhS}
zfnXpQ2nK?IhGF2BQ&&Ckt-npr%f*I?Dy#+r!9Xw&3^W%8&i(a`JFL<7?YyqoXRkXu
zAGhX|yi}aC%#wZnxyz8eQt^&eTX)%M=D+ghf@AjD{*#GMjmYbY&87Y$k6@ra7-;Ih
zPcGGex&Ll(rLAi}_LiduZa*=vHtaBP@raQfCgjzMjm39;oXxNi3>3@2_S>%U!cSHm
zpOuQm+7G>8AQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z2nJFH7HoRqEq9$XGY<?Nq7LF5L?0-x52SrH<X|8e2nK?IV4xTV8tZ$o;is~`=Uq%~
zgibIJ3<LwgKrj#t1Ovf9Fc1s`1HnMkF)-q`BX=A5=&U?2c=&I1&~z0Vo*RRKRkt|%
zx6AgPnFj_BQ3nlo4jSVuhmT+&7zhS}fnXpQ2nHI0fun}L|8T3G6Y{{|A?l#;IcSKp
z7dC@|U?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>2AZCMN2eaK{Pa0<vSRVx
zu}ijJ@2yX>YQ>vw_-4OBZ%oXp6_4HffM-T6d^0N-51aPn_=A2wGAkGPZTiZM2pWQc
z&xd!pddk|<^T6PtKI)($YBX$?&%pkpciLv;&^PkH;GsFzLHQ>mRt5vr!$7CqR(*cL
z?)PLZ6&q`(n(E(?Zmj&mPcYCD7`U<Zudd6#{#ahA*b??2atsE7fu?0(_72~k@%&43
z^T5)0Xxd#1@4-N2G0<nPJ3Ak@=9IitocL0Y9$gQ3Bd>3)EOsKAU?3O>27-ZLAQ%V+
zf`MQl7-$Rzjy`MZ{)4|UBC8D>BbM+{I|IL(*tXM(ZC0r%759E<gGKND`NgbMyms^o
zT^=3$X;v&Aci*LNuKB}HvWCWv$3A}hA?q$vQ!4)M{1M;z)lnm|V)49br!0GZ>vlE8
z;%)CP`{ZfQf0~tw1D?O^JFWk;LQT0?>i2X1^vNSH-?MA0nqu*q-`zTC&h8y*O2sSJ
zTlU^dKbn+R3cmcu4%hZ-KPjtkto`JMMss4|J8NI?T#pau<c*EZiNnYx7zhS}fnXpQ
z2nHILfmMI?$`6)&eYKiWan!(hyH6SO@2oc5b-&L)-|US=S&EIDWq1n)f`MQl7zhR$
zmw{*AUE$-s_ne&v){Te8-L>!*473aeR&2M!r0cs+%3BJyj9rZ^gMnZm7zhS}fnXpQ
z2nK?IU?3Q%I|K7pzW&5%f0>iFRIEFL<yrXnz>XVTeD#NUxp@B>*M0Np!(YxT4gD@=
zePekn$EskUei&Hx$g}2kyJt#XD%Owpuo4Ue1HnKr5DYX7104n~9x<}RguGbHF1qi)
zZb!bIHxzb0@6cWs?e<<?2<!WPS?TspnzzP)4gUUmRxUP7)r8exAQ%V+f`MQl7zhS}
zfl?V5_~JL(9`Mw2d8xS1lkLAZezP_;#bTxY-fw-yRU7Z?-1qR1?T&vvFBUIcY3Lol
z+<15nSHI@?Zl6sXmzRpQ#Z~V7KK*ytcJoUv%Nhzx&69q)`NX>N7-;Cv?aGTPRs{of
zV_@gk#~-u&-S6a0jcFF=Y}{kBcXpVY2i6Y{DZdRZe{|<T+wAi8tURzd9vXT-!hSFi
z476MZE*-P%#6P$CFslu}yYaebJ~MlCUMiOQx|;riO)tFVu9Ifw<zl&?(^TrZmRq5b
zc`y(RR0{*=EqvyV&sLq2mx^<K{>Z&YKm1W%SF9GEBF<nS7-;?sESNNZ%t{x{&I5yo
zsDn5M(Fe-w1I_<Q5g7&p!9Xz3vKeUT?^CefvRRJotBrxppTGJyJ74*JUMv>+yPX*S
z?|;`+ZFLdx1_NKkz{1bRuJ-I6GxBQ1CHMDhv%<Wo`B&vu&10qhenD;eg+4F&e0Z0u
zr>s3aFBSJ6z0)=$hrW@Qi=}>FaL6VPO?hk6p?SGjYMiD1o#~M`U;bv_z2C}8Mc-p+
zYAp40xVq|B^Vy8pgMnZm7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z_$mf|u*TX$o_J(Z-q?8Eot;;k)@yuT8=kcB*4vD1_jq0|uC_^^AHMqL@Vr>u>h_P8
zIr@aLd8zo*hqnFa?8ipt<>Fzdbm-Lf^QZH2@zUvwcROgqXR>1Pl*>C_wcyr&Wz7Y*
zc<Zvslkb_FRV()V_Jh0jnmjKn7JIMI?%vk#&CN>1LoXdSW20YA&x*xm4t?g7D;AB<
zYQu9rIOLOk9vYi9G=5bNi?QlwVA^MIwI6fiGg*pfbzEh@25(Hti^c7)UU0&kc2o0m
z@sgX~*<;-vQ}a@>$4~CLYwxd*%M0NKv(Nw5_d7q37mHi&-eb%f`I}j>`29tTu5a^f
z|GZTE%i#6<4f=RwUI;%r@wY!&ZtN3zsrc%io4>pJK{w~cVr~B}(a2WYUjEY|4`k)y
z;qA}dZhpIQS+RKUv_tM2vCEXKSX}k9%`fQj@RY2va9H12KkB{eYguiWo~L5v2l53k
z_gwp>tXM3zPOtdGb6fWvF(j)EulnqvUq5o{-C41C>)+bHd)D>;&d$H~tnSYZexYQn
z{tg-$)P;fFFTe5fKFiO_TPoJZ-L_o^eSGKF-pxzJKleTJ@YODwl{Yj#yZ-O`kJ)@?
zUMW}`pD`Z{R1yQNZ}|Qu>rYsmm5T!=&VIS)t{-N_;%(oXanqtc<Fit6>jyqKX8GQe
zvZlgH+Ovoy7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>
z27-b5WMJIf&p-IqIdA5bigzB<{+#{#49JVclYhA1Ge@p8KPwm49lhz9&)olXRxJK)
z$iR(0S#xq$E}nDZ<Vk0)|508pKGJQ%u;>2TrlzUz-LB&<yZ^M8@^bMP-#uotAANsB
zUM_BU<%h>Cf83C)p>X4af7I)ub*5#7us+XX*lNBEl>2@5SXZ~d2MGSk<FEOi#K<fd
z2nK?IU?3O>27-ZLAQ%V+f`MQl7^r>*>iYMu2W~NV-*sM`m(>?u)nl6jR{Y(xtWxo)
zq3=K3s^^3}uwi+q{*x6M1Ovf9Fi;r`Y}NLag;!lLBM%H78nX^Y+;-${BOjfWmx{Ih
zUQnU=$~bKiMKBNyR5Amrbld3t`#v3!*Ed#j&JkNM5DWwZ!9Xw&3<LwgKrj#t1Ovf9
zFc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr&{Pccow;0>onD%rHx+(9
zyvx;7)}Ed>H7?ln!dvb-X=dI~xa9tRZC02!H7^%C3|u^7WQPfPvDo>%LwjAc+k1Ik
zaqOS=+47LBU(AZdn{N1Kzd>(I%u2;O);_4i20g}Q#p0gtpWNZn^TuYy;<g7JGv<cZ
zKhBE9QvW@z;avvr_rkNUX7z=|=2hu>P1V=JZ!i!H1Ovf9Fc1s`1HnKr5Dc_j27b27
ze#cz*$;`ac@#j;IS^dNHZ_7%>?4tV)>~`ead8OcCr*!Dl_VcIn`o@-9!I60|5DWwZ
zUuNKikLPaw?`3A>O^siUDx_c_7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}
zf!Y}8ztz519dyauyrp4nyu^ItGEnZniyZ6f_U}qJ6n~9dJ>e}F2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?I=Fh;O(X0Hj?{{bCfx$!6L7ao=1LgIB)vr0e+h^0p
z<>lg$H(&l{-@V_;3gOImCaw1DdLy$+$2CV@u-OwkkIe$V%tQ0PNkoRhKrm2g44i-M
zS>2x-{6f~;@uL%e`;+CyK9ScId+oNv0V|&ITviD8>A%Can_qHSRvSLP?Z{Jq-uCZ#
zbHN|1vG$NB9+{K}1`m}~2bI=uBc5O&7zhS}fnXpQXgCJ8nX%ccKRWmItTr68*Y=-G
zd}>5qEFSZ(^Uq!K^!s_S*lOKx_gOq}bY2h}j)Jh<j2XD@&d#e%>oq=aZrF^sEK&;w
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLpduN#al8KstM7SzPw3gH>y}-&>e~8>zK8bh
z+U7t1?b)H{>OI%$+Nw|Ao*n-qf2(WjuC2Q*I`!9md){$Xmv^#dTAkZ*^*#6Oy<yGx
z0T14E;__dwd2g2;@2P3uvF6xOYo54Y*R^Xdet%lGHAk;g^Sy3^PQB~V^_J?9|9s+2
zTb^*#hBb#B_~~m`oxM@bn-8tK=B3}+u;!%e_IdeN&wZn&%@e&Z{r1i4*ZgJUUp_p0
z%K9~b`r6t1-nZ4W`Br-`|Hkt@2jp+;wP@tiFTa<;3B%g|q{FB4@)Ren_~Mie&tH%~
zanDvu-udLie1&~)e&yL!j?EuC{p>4VxL`*9!;A0ked{VC^4^bi`R-W*p2_MP*@urj
zl5Kdyl8H;>IOFT*b-#T4#H?7f9`5$sjJ^v_pOc-p>hZh0JnF-Y{6wc!Mt!*D?4{#-
zw7IAK^3&#Jdp-8fxx1aXDC^Q|{m-|U{OM9OpRPM+_}x!G+I>d$aJ!W*S$mTXH4CpA
zy~W54SF173-%}^Pbnue<<}US@-+B3qtA1;K2I=jy;D#C9#xKsE`{AI@d;j_KrT+hV
z%RT)c>fEZvJQSz6*_+h!rVihItyRs`3GJ60@k-m8XVzUbc)LwL%T8{0-YdJU(x&Ex
zzGvTa#D+_<vySXM{Kn&^Wa6bKPkGm5CtSYHsSC2repWO2rJ2jrZ2pf$M~^vuxtjYA
z9<}zQ``Rsy%X)mv<EJ-G-~XQXrY_w_^Hc6V<cht!EweN}dCCj&&z^kM_pp~QpL)l?
z-dSqjI6tX>^+fOL2RpAZ;DKS;O9!vAu-BWzGxn@67M-KPTMz0mXwlTX+Oe+PSwGTO
zoO80i_3_nzyZ_){JhHSN(>`OL?2r9Q$L*W_W!E_W{OS|^Ry?-TtWEy4w10MAxZXKW
ztvD(_?%eB-AGh4trTy>p_aB{j&WYnQ^;B1WVtvyxpbpcz(vQ`zdQ?BwsUvj)&8yFW
zvp>7}ufLj-t4DnywMSn((0Gc)vN!I!Lq1u2!JHhbd;QA2z&yGC`Ip~&>80ayamd5G
zb0Q!4vyTUQ?2!4#1KIh;Z)%?m`6Jsu_Ut2+7k#pL({;|hdHn*)8>ZvplaGAyAz$`A
zcRh#Yqpx{xTTgb+D1@G4p3{(BeT_YG;1~CorslzYNIgLJp|o!FEpZQhFRR(O-zS;+
zpigGU{m?y9-MELTOUS-@<kvZH?wkX4=N_xx_;((hCw-G2n8w8(Ufd`7bAQC6x_9pM
z0e#`o-(0@aONTEr#Ot9S_ZYdtv7MLX@^G)P-uZ^+<s%RLK)lJzxYuXLckcYia*xej
z`npVi5=ScYXCC5H{~_!9P(JJ$FNB^$o(tkm<>Tg_yX*X2W-ommN)LLzh|fNFUXY>q
z;E88U-(u|@=l>u3_~R$l6Q}XokU#sZZtx}#yo(>-#?=*j5YJ@#WbsYe_p)QZ^YN$o
zsZ)BayX$tJEya`XyZg_Z5BwyjCvTYYVVr&%zx}Ni)EB(2B|B%%z4PhZL4D1B>SOBC
zzS~!_eS&zj-u+D<AeYOle%H`=?;0E2_2!msm%i@zKESDab>8#9YnH9iPrWZtZa%Rt
z7zhShwQ`=_WAvjl7xlQm_j2uP+?Vt>_c;AaUm??Xuc!_6InOJoufcSjtRL#1`mTOU
z_WbiacRwKOpX?du-*ek|`W(gsfB5t~@Vd@@i7amUq<rvW9er`g1L~LNp**ZF7UPUm
zCIe}`s9WdI>yQ-PN7SqHsUEz3Qg7znBh;-rH=iOq)}{00LUxQp@sQK?=H&(Ho3{?q
zgU0DwpXP0xekvQ6zkby3?GrX1`e2(HeZe~Wm!f@v?9nsMAJh-@J@;*W%so6sdFqGk
zL;Xo#(VxV{zr3u2cwTY#Z^le+vqjBkXKZxJCA)53(|+rH5Bubjt(Nk_KD+pH54Jz*
zU4JuPZRmBF{gEdgjC);1-##1ndW%f2u~5AFlRVT#ePd;I=fY8`SLi<Fyf`2FwDagb
zVI8FJeCWIEvD4hJFvDLxGxhR4-rn<^MN2;?P0x+j&HAN&qTfOFte>TQ*m#Qe#dx}|
zHrcvj(fc%o^yQKIEu>%UxV{7R9Wulx)GzdpLVEUvU1;Cf@%%Ece|o>l`&rO)kbQ`!
z`o^!W(*GUhUmkezyj8FGqpvU6SExVe3wUDJ`qUmj^zclkZ@(bkoiqN(o>TG@m;5~U
z^)2Vey@_A@!fzUvx}`5qy!alA`+_*!n?0}S>Bssm8Sm`yo9fvI`H}fC&JNzh3+;>7
z<*7gN|5ntO{|}G&fc%OJzhwKMpTg-w7k2x>?8QsJ-z9GO`@F?Iup>_P<!2lp^v&ao
zULmB<Zz>xP{Nnp}cyKTAIi7v6uQ06-{Mw%s?UQq(KCELGI{)g``%BI#d+LX5eIXRL
zeAD`32RawdIW#U$`tre>aqE1(i3k4Ec%4gqEk*O}vuFSG2RxRFU+x-pmwZktPWy&G
z_29lMFZ&L?Z^@qT8|sVr$Mb2IPJjHk76149gY?bAO?T>Z=A&n|u3^{b)xJMSrtk9#
za{BoMJNBLIT=^a+KJ1U@k2=Q(J9r{{J~=1y@Hw<N`LT{Zp80bwjHmmgUd@X~{7@ZE
z|Hg~w_uOy!8vNl&eR|(nT*jR*>x^5Ecl?MWo%j6${M-LZL-$noC;Np5yz5_hmbd+|
zU*??$deA)n>_47ozrFCi+n(uAbM$RT&e(C-S~c|f5f^*L*&&OY?7Zub`dX@Io*%OH
z?5S7t{NpLr!-sxr-gt`YSiLznX`Q*}IFHbJ-*=&BJ-+4TKFbf@+0h5>8$IVlob1wb
z&iJL5%K8s`{IWxKUinYwjq{t@V}~Al>WK{HV}DZ~_@|Gz)Sh$2uQ-13<c*{E8rWut
zdNR%q8M2$Aee`^GPu_CYpb;~tF2D49?BZpQAAEUEh|_+lAAGPQZhDY?e6b6iJL{bX
zJoArNe#OauTJP4Gw_d%;0~+^yhw3rShkfJh%TwN|f9uo%ed|-ypVEECkNqaYRNwxJ
z2an=6jz_XQ@Z%ie1y2y4Wa}V%5Fg@T-#DINnirWpd8O;|DsK6?ucSy%Jm%eJ<mX)N
za^y=_b|11@&EH=>X!&;i+SlOWqg(bk<u`k;y!3k|>3zgKN&fP*ub!v!p$F6b5;s3S
z_i+#HeA?E}c9^_!jeH*XZQtwWZ?<BMJjm9`*XO(LGxB$@k+0A1J#XA|{GATF=B*cx
zxb3UIzY!NXjf-7z!qksC#2;B4{A|6|IS2gvl$Dn5qxj@+-|%1@&uPBw{rI1Ef9LEg
zSE%8KU3TOrE^$KjsQ($qKbbvx&Ii<g<)Lnz6YG87$v&#LLj0%u;Pnf8c#<F9@c^Aq
z{^Vty_#r!Fh%foEFCOuVlT43Y`s}f99RK))b${jBFc$EuueiUa_fOB|LibhAYwNuZ
zrVmp)#_97*57JAKKELie^t_)7>ATOf@1CyTmWtN%&mZ)@H9h0>J>SXZ-NU_aO)o|J
z=lL#w_Mm%#=L<XZ-9zw+FYD=(y?>4`^X$o&K0Wm!Ugw$&^?x!f_B7CEM}GF-b3uP`
zew}~k9$(_dqkPDazP#u`{R8r=uhoXd=AY{I+M)f+r}l3*WYB>x4H^7)t0DHq>jS*V
zW5*Bs^%}BPn<3|4Jo<#~c5F4o`vYF5*k^h9Jj=S&f5?@=KzbfL$L+hmW54x*6rCsM
z-g=*B8i)G4*L&*5=c4+abL-ygzCzaT)UQ6SPm=W??`N5ppL^EKqc8jDf)`e-NqKQT
z#RbLTyyze91MV5_U0#pKOFr&dp2PUVvwlO?Z~1rr$l{kD-uVms;`a|Kg$Lb0UD*G$
zPVHyl7r)<ux4?#BKtD?Rk2-a~(Z}5v^&!s%=gd8b4Eb^2G48pb@9{&1`aHQ1%7dS@
zPwUt2VP5CD2a1Os_d=-e;z7K|@#1-59FOAGuTymHJ;%(G^$X+tv+KEmFZ29+j_^aa
zp1$=d;>CJ#6q3blo<GlX;|){Y<y*x^ntxNv>eTx*-gi=;-j8zMftc{zsV`IyUiaEp
zbwMtK>H*SI$Lz9~BK_LXI(q8PJ<qv!Z?(_r0OHL&ednAV{ei5XS`YOJc2XYrN#m!7
ze|hOY>_hYJ$=1uOv9XYU`7~BMb@gX|ijCJk)3>kq5D&d{zw9GS*E=uiJU!=%KlRU_
zb;ijldQKRp&n{k^3+IhKe^B0dVNbmD@B!)bm&#tRiI<)@_z}1DP~5&pY8~104W`d$
zyr=s@AMyjO6Cas<<EdRd%2!_aBGZF-k+*SZJVpI2U2ngv*N4=DzGuDhRL|>X_e}aQ
z)i<vHnioeJXPQqMmwS`vr{}ECHN2k&`Ge{uMScs7n=gbuUs6A%qI*PRqy15z_Qg1~
ze`K#0ypQX182jLT7oWp;-^l9@vgfdR^1hANE%el#IK*i`)RA+g-r0A~oKy4af_-t5
z#Ul=J@<R{e#rd%gzvAIvJo<@#V_bZY9yHD$9{6)!_{E?2txu6Ye%W^(ofC0*E|}Mk
z@lIcSg=F^d!@qTfWcjA^^0Xd`i$5qXeLO{Z(og-dt3K_2it0r@;UCZVw~jufpCbO*
zfraj&*3(PTI^*gojhh|gslM0$&b2)B1ARn4H|{**0qQsGd0)Z(f_>xsd%wXs6|enL
zM|kv{Sv+D|pP$ZNen_>TKCM30t@AHm^JID3Pj<*D+GpeH1OMjj8?-NE$c}idcdndY
z=hS}jR~uR{Z~F^VKlJb-&a^JYt#65+ztnE3XFdDSxP61{lOaCHsU8`!PZqDd$xz(<
z*av-<tUp>Oe}4E)QQYi6e$Cqt>(z<+qh}pE;*eh<bbs?4v)=PSexB3pr1=(-ji)F+
z=apZ`pLJyGjHj}3bxx+oUz&&U^j?zYFOF0nU-*Q6mmt+=$2xgK@!D_uZ@<kO*T3`y
z`H`XLfcSj=D(_-ZeV2>!Dt0|ypzr_ro}up*dM?s~__m+WJj5%xHdJ5G_Y+S0=GU)3
zp#SH&bLn$lpMyi+A9Np35BA4?q~qeUf3>0WP0#Cj^`oDt6ZNG3*(c-bNZ#sAed=50
zjq_(7KkTRT?q$YP**JUR6({@lwU8h6$UZrZo6H}3^5+lIV^@555f{7St_{_LI#ws@
z!nnQzjl)zw9Ty+HG;Z_8q32|wd3ty&g!I@WL-rwl*`t@{MaHB2@W@{}Zr=Gx&ndg+
z>C<x#$?RK4hVqh!yy&OsxmPN(&o5rYsh;`Ak8$gveK1eoyn19G@?QwkeA!_aU)J+W
zw%#~>vi!uKo^So#_mA`m&w2LJ=S?~<UZ^jp=RWl>K8QD{zS!fJ4C&)Vyy8jo*T=+1
z_PoG@ehl?JNDtC0^ge)j$bW5!ANlZaf9$vX3!!>(9vU0<5Alg#Uh;(Mk{)CST8~%z
zZCstGpHklsz_)$2UOlt#{C=|jqZ{mYe(R<0JEV2&_ZP%NmJfTzt(T8=Wbr`zN{`Gg
zJ*W>jNA&TnZv4(QJ?o);6&KlhGJR;D#eqNZrr)P!4?jLP@H;xj?T2_$<kz@($Pgdt
zynQj=(CA#KZ~FuB>p7(lrsL`skB}aecRG(Z{L&YPe&oD}$2?x;LpBf7_{r*oAO6i-
z&puha`h@lJQ)f@MU9{QEgO?wo&gJVKBwpvixvg#g=|T0A+GXE8S|0X|tPYG9dXC69
zwZ{&BWOnUaI#16yeP~`>g;0H%7azN1dC@b^4?W}d)wt(7JM`2AWZ&}}sxRy8pL$|f
z9kD0AbX=VpXAj@x)GvMG^w~3>qPU^>?3eY>xmW-EipzM4*3mc5kNMQUJn#kOA#QqP
z`^+BuX`Jkc$9#(B>6_;dKjIK4R0s5pL-hs4%^u$5K^BL&tTQfdNRMCl7qa;5OFEv$
zN8dQ!#ckes5Es2P4>CUG1I5FiIH7syeA72>Ka9gdcI2UNh*LgLoawxA{L3#z^ZXTy
z>FZ#8dyd$D_a@IR&rA0tayn0TZzQ`{rS?-<JP=?0UWL6>rU%(g(dUakKlDDI=bP7W
z`nG-axewVs;a$C`BXv^<*-7Wq`tp1e4>?8WL|w2WKI8m}A74-&;=>DMmkik@LvgaJ
zkC=zndtcb+wARZ%eV^6e#reJl{^|Mr*XL^Nr*+_SNpYwPdg4suF>f6|>9}~*m;PZL
zee)@%=T<+^Kb%)}${yqg+DB-Aibd<#Q!nx|59vYU_RF|^#<O_XSI>~06!GPp`~2VM
z3cjCI8`=l`$hqO4%zqlMxbb4%`V`GO&+HrLC$&qTp69A@e%LckHm{yi^!mcRB3;L?
zasI^%#Z9*F#@R7XPWdxl2=zPhq$ppQ+DZAbK8=gaPP&flyo!rHWZ!%tv@fZ?acF%i
zTbKGJr)VExI-lC(SH9`I^NlzAp33PsyY@+5>XE-RK7F92qyB|A`+`?kXdJKf(tTjJ
zTG4aT^IM&I4tf5oGxh4Z>z?UeTpQ{O?9j(=L!o^vw9ove@!4<Z5-;vao|7rMkI=`f
z`%t>xz0W%O?sx8Q(7y3&9sBH=cOP^wNbRr(t)~zD-69#%=Z9R#kNFhYXUF$A=((@5
zYaUPZ#0A-P?{P1pk3ak6`3#){NYA<9M_laTlWd*(Pmw?9o(TCb7x`1y&P9r~%{wRQ
zI_Jy0I_6j0@-faI*?DtL@WvlI(CaI_u}7aQuJm3dPUnYT`S1hv3$GvbK{9lIB&X>4
z#1Hg*=STlFFAw<{hvu!LXPjU2sUPE>YwTJFtus$wo@AJg%hNc%=(7v?lMl4Ng=BeK
z4^z9TOiw(prQfR>Du3rtKZp9L^R3@H_xiB@>-;*e`Um^aIAqT{{`4#Jo~zdB8{#z1
z52TmoW!`gwe<&Yz*)b03vq!d0KJu`REHC!0rw1Eq|4X&&eAb4}wZ2A9QJ>I93ZZ`E
zIpVoxyb$8w^Ns9#p#C0C9YOUC?XP`PpY|)YZ{NJ06}R)k9zAx|qkEEjlX3C6m!)Vw
z`NIp|@tcl&zCilWI><hmJ+itMci`PmKiYjp7CbZ!56(r3>Q+Bd@5YNobzi9{ze4@S
zdG#F9kMt>hQ2!!B&pUka<KCd}n%6I_(?|K&U&&r~LUDNA%8$OpUm?T?KjKQ!`t&(g
zXrI|l?cm8hM1Pkbp6!D^57WB!evr>o-IJks^>g?AV$t)RK7U>>c>R?keSXDb9O5s{
ztEpvisZaZ;4%Df-HBN^1)BZY7>^UFwAbs=h59*44_ROpM)W1B~<%eFmXkXO{-rVQp
z3H4R;cr!14GPI7L6w`V3>4~$c>#Di<*9U4teW|wZ8Tvh0-{YX4_8I$ZKgsmepLz8o
zPJM?gKKl9<yT1RCuCty${@J7NeHM1;sZ0IYI`+ivxnw*&r|Mk1X&&m$`G;x$DAW)1
zAM@6w=o~}y`V%xxPd_VGmcQqw^`5K7p?iWnJV&j2^1CNI`LkcGRiiI@?z*3g*FLZV
z*)h+)_4sDjbDv&{;^3Ekd7C%x_v`(hp?z`>^7_{NV~@{x_0UUKt>HJ-KXtFpdmebr
zvNiPmI}P~tI@9m3h|B)kM|q1Anupelqc+4#itHKZ&pC2_jMs+ju<N;$u4A7bnVxYn
z^t=|A^Gk-#w{u<So~55a{YQTxvqR53|0(iM&pODSdHoY#@}O^?e(H~GUCI-FjpGyA
zH>jSRSF-cy9FS9VUZHcqj{cIObLxGY)UMYn=F{I9aekpVz24DZ$o%6+y!6B^E_|hE
zy>a_!ytaMwyn*Syt3%_SkH(>S{?wUqb*WB_(^F^0>626B*K^%DBd6d0G@hd8yyrPP
z<U)SLW8OOG?=~6=ojdW9)40gceMa0V+7J6oUtXzxI*vDpZ+gbbFx69!P(4HaL_Mj0
zc^ap$E}?nodw<a9;Lg2sZQSn?LC;yAn|scu^X@tNgLBW1zCechjs9U=T>7VR&oSuv
zuD{?#U-VqZqx*q4^%e1w*<qI-{mXbN^Q*s!lV9`3$({@Pvv}FT1AowcFI{h5p3pww
zn;t*-f@z-OGtRF4N@e=cb4DEMlB^ExH#Bdb#7*WG;<pgmU+8?<e|^V3I8V?#wBGsU
z*S@gBpLJyO_8I?Ve%ZCo-zDg0^wfp*>{(AXZybt;J@rrDyuPD<%8S0bPkAt}j`Rn0
zWPjBuR7a^@`J{C$&J^)yTs-NxdHuRnv|e4|LEg?kJ@LCYnAeBpNnii8Z`M06jg6)9
zQ*HWbeYj7kf9Tvm`w6WhyPsIEpFsBsdhQj@pZikk&v_Jgiu7wk{+&PPQC+J){o4IN
zo$E)|m%H!sfzAQ{{6P7V({pZ|9jMQl$B%h-=%qa9PxPF7`Rj9d5<mNp9dV?x{vc0r
zii;gGKlH`zclsL|`4<;C<;OVG*TkEyrzbD<rB0yr_JzLn=E?R8e`!B8?|EfD9k<^7
z#JSMV?PI#0J~WSy)DKzyfx*Ck7)alDaDMa)^PVgEJ=t?9Mb9ti%XwngeVFX|Wu5Wb
zkX?Q}7uoY1G)|^3Kl9Exq*ohS@7&TCKN;d#zai@f?C9_E)PJD<NQUC)AM!(n_Cq}U
znJ2SLHgCOgC>}C>aTtg8O+C2}8HedU;+6dr)A-C=2aStIoc0CUAE-}T2koDA^sGz!
zJw0eYp!NE>=L<XZ?2~-b{L{LpXC43ckBpCW+`Msm5D#Q}&W(NcyO_J4y8amxS8F#!
z9oTPv*ijGip{I|eeM%hGd0rY%dGP)nJN7kwT`eAY;9Fe$xu-$?%!>mr&bNJG*Xw)t
zdwKIOKJ(DL^*e8I!pCpiu<X*`<q-!=;}@^^og4M69>r}v#FPCe<5iyK8yn@5)}7}b
z{_vyDj9Z5%dD%DP*4t<N@J|o={VjUN*~cS4kUm~heKLLh1Mk+cZ=ch7cGC6q+)wo>
zuLr$Pm&#rTd4GeRai4cMmt^Nze-xLvjeB27JnXwqJBMU-Ex%M|$GH+WlplM}2Y%HN
zzxo_~{OSM3@zK(6v(6D7okQo+xk}MFbe^08_N<4V2lVN&LpBeM^RJ&cSM=0_x)qOe
zWId#3Jw0}ve|5#4?=g8USf9$`Hcy7e`8AGzdD<5|rhKI`J5V0{$)60<@syuJdhGKj
zAM4o_r#$3ooPGR5`-*?|p!MVw)6YZnIjFzs58~%v-y(|>(ogHeynaMa|ANNd3&f*O
z8K*C9C~oW6!<)E`%LkwK!Mt_+@}uv|$GCn@_B;>|ee3Cap3oz+<GEm79^xZI`PhH;
zs6Nwibwkg*x;0)%kG_4j4!T#e<2;v(X&z*GL;ToJ`z<c<Li}0hc}O;&t|!yupIwM2
z@sd+yk6vmo)weIk`Ge_t<Ip<$qtB)?JNA=bnD#SvQ{LqV#Zd_PEi|8w<4yebjci^X
z;-r__v#;WX^vsjlk#{;z52p6SCoXZ@7knFM$9%fpJWS{DtKK31#UlUWO0kfgR4=t7
zU;UXr-cpngfAUUc<NT4?r_aB6`^hgo@zvJX=%tvh6Bqx5;!4*QE2sXQU$6Vqd9rw*
z@!I^l_tM9o^8)e7KHk&1rEkAee#q?LQ=a@o`^q2t&M_H!u6VAK-4~3Lt#eO-p7UgO
ztoPi9^r85~t*^1eul2_92HA11$BTLPp?N$T7k9B}U8P`caqEvM;sL62=SdynKh4)T
zUhM~2eYpR+XWLgYJL--d<7wTfdv)TRIET)ix{{aoO<;Qd=@}>UCti8+Paj|SBtvnT
zPx)jo9ml7-hWN3L+|a0w<e@*|OTKu4wawFm_)hD=e#l$>lhr}WV|~f;NKxK}#_gN^
z6en4H#_=mJ`&vkk|GFZ-DT>ScZ{DBtx{&PkulE7zdmU*Wrt7?(OwqYZQ9n@!&RNQX
zexR=8gJ0|MEHC|sAO6Yu78&v*5Bp|)iu9~k_w;;URGd(}@?b|?=Gi4feypdL+Q$d`
z^vPcL``#+_y*%HWPRD)khaODz$m&V`hzsBH5}$tIoH>W;JFSCM-@X+ZXO};`LHmH$
z6wQmDJ$_T!xVYt`?&SsXAs&8GG!OaVpB+d~eagf0UY_iTj||zDhrIA<9<rOp;ki^B
z`dp$A>J$8{Z+SxXhiCT?d0MXzk=ZwHJ@h`D_doG0{}erMyk7DA^jzi#;@3UTdj92+
z`oXvS%*)65kSCcQzsBu9WZ%4a$oc~Q%@;!VJm~Ku@nD_0qn9Fn(s4X_{egeza|P?{
z3w=nBoT7Q?b5Vbf?ejPLfFFGwfA+<>cTVZ)-_94kRAvWn_7$J@6K`a7V4eNuU;mI-
zA!G;EcHZl2-si3UeV_WW8+L+$V4(6C(7*Im{YF32|LJ+IK)>g#|9jr*BgQ>1JRhw0
z9AHO(C0oa?d2#8l?1~feZ@uRWzt)MDKYGT=DY8>&oIS`Nbble^B}Mxq|8zc;`7<s)
zeAr(+n@>@Gsb8{v@O)&KEN<h_`r1x}=O}%4)v5jW{AVB1W5;-E&wVeA*Zsgc`Km8=
z@NC?ELgSDf>mfU2`p~@l3H#7JLmt+NU!MM*BKxZjQdAG(;Ya-R*(0a*hky5c@j&~^
zK0Ex2+qk&xuXxlqJI+1c=;6nD=TJO&GcG>k_@GZVZ`?X4E_(bMhvwBcd;CCk&%XR&
zYR^94j~+YnHEuor=<x@|2gPTd{gbc!__Hs@?OVCXKYz{(KE(~iEkClj+510no_qcA
z<CYtno2M@ihzE7YJ`|txAx`ml-C*A98t)6)C+Bd`fiDdi{C2A$?(wPY-;Fg-w%#~2
z?{%5+6rE4cP4>KgvtE8F;-OfyzL0*Qd3AwLGGxzs?~mf$>qX=C0Ty~c#X5fNtJkxT
zA84Fie%K>>U&Z@2#tX^Tr+M%P`Gxe&L;2A2dYgWlH~n;+zWCC4GUTri;sLUcuR`n6
z_~mI|*u$^)ZP{br`~U2jXD7{5UdGiSG@j}e;!{0~m!7!955*^6{vf|lT>L=eDY{qD
zbM8E!Qq(75y3Tsz>_hpce9JFgPfn5l6zR#648_G?%A0(R(-*&cU25O+#q-?r(sPFF
z?{A<!P1cw7Y4<Mn)r~mWch95ec?!*YKGE}BOYNo4QF-G9e^6ZNRekdBejy+B!~;E_
z#g)p|;m16F&5Hv+;x_L7Aa0oU1^a1S-O5{@#iDba)-72a`jyW$^#?LLolo2P*$$Ie
zt|=AOFJAOv{Kz-W%e?d8e58EHBeiQhMfsG9@(`DEZr}MAr}`kfCz9<0Ju*9Zf&9wH
zzCq(Ds#|`<&7X54KK`6b`8hYvmvfz>^G}aVj~(YoeDululdU7OCqCos(5EL3<Ed;M
zT9?ZB;*Ts2=iNAz55zaD?R}ASo*!|D7dkigovi-I&LuRTo=4}7eTW~3SMy$1rfC1I
z`#<vkyw1XpJX6F&TG!<C+~7l=_9K<W>p9~5i&viFR(ImTZ;E)vuk$WW^Ww+5=cam+
z2YvQD&%|rJ{P>}7-nl2UoBH!yVn@HR-Z+#WbT7jT{_U%C1nJ=e5BeAXWJn(m^5LHz
zUf7i{WXC!(d-TMKfBf=eTs%-csW<mD^_9-6M`&Gb=zhVz^@XsIzWb1MsUG_&(zDJy
zd-62z{RFbtS-vj=ecyqs9(=FJ=i`3Yu+ZnP>A3jBk;W+>yz6)BSU<3j=E?eoyxFsk
zzBtT7>+q94AF4%mj+{g2JUL&^5q|Uy{T82)9v+S3%{b&IMduxgBh}N7(|Prf#;XpD
z(-RN${G(S0(|FQ;LzYKvnC3xd$G++7`aNV{9eQ1%?^$P`JwM#bt@pgaul4M?|N6Z$
zeL$S%)s=Ph`6HXxZ^XqvJ5U~;51yam#~b~0-0LiT%X5RBG%oSsO`iPVOWk;`nJ*NF
zx>9%6LC<4)@=yo-;!Pgnq^B>-S3c^bu}~h+`6EMl<4?bV&K-UH={Lsd^OGXI6jQy_
zK0o>z85&QKo%H!-oIX2vu#f8AKGN5B#mycW;*G3cQ=a+7i_aJEna-1)3w3I}=RBEz
z@rx^sTfb&koGHpf9QaE6hj}O;`{B7RKI7tWF7-p_8uDx2Idm?~<3V2XXWw|5SDLr{
z__L0EcGLLuIdNG}PhT<4k9{=`({ZwRtY;t6gY21y`aC|#^zkAu>)4ei{>5cId)7H;
zcreaCyX@IF>%<{0NDuluhV*$UuGEilh*$ZD)4HZc^`{;oKAb<}&K<N~9?q>kp?{Fs
z(?5;VOZ~9VpLu9LMdwmn@}ZZmH!r@_pLOnW>b1~*vS*&Y`cxmaq54&?^vU9(m!1dy
z3!!@jKJ0^i!H4G;9@MjYihsY=zhC3;Je(i-i$fmHhx6s$>HP7BPxU2F>r*|ifBhYb
z`;~azU-@&Nbnm2R-1pJN$-g|*v-s>!%BSZQ^xPuj-G1Po9s7U}{_%<jagdz{<Lt9T
z4~oM%Fm7Mc{OPf)&ee&y*=5f<d5DJ}diuXQ=g0ZB9`C8_J_g;#?W=mR@6MGxApiV|
zLmp|K?kmRav;D+JiuBp74O6@6`y=T*|MbLZ-1{ls=YirBkNUJ;-1g1>TgSe5*daSd
zWOh^8IJ+>lZybt`?EJ}>e|pXf*?v1uWcK(^Wp>34<(2CD92$yG9M-X8y?D&0{Ms-3
z4#kZR<AsnuzhvwAhvI<hgDfxTxdfd%sPCtJPyb*S?@+wZ{^Cu&>L2=?=a+H*$>zn!
z4jIZD@+)p=K1F$nAEtWjq$uBVQM~pEug0NxQvdoUeymp~?8=M2c-0v>^{2k+u>;jh
zZ79Cl^eSZ@k9fg9e(>t~=y|Gs<*Sb6rEcv5ef6K})8mI7<79TAadC)WpH)ZntYg<a
z*?hX*I6a6Tb)p{iUC%$`;$%;sD1`XN7c4ZM&g0X#*Rk#??lolhTIh2gdU$lt!UIhC
zH0~Z`UL4}$&%K%KUZrk~`y9&Oze4X<L-#Xiy*h#7QuqA1@5>tx?t$#O-y4?){+$PP
zr9RY!=P|^Ox^hm%m-4Tk^f9RK^Dke0PCcq8d3%2Qej9&y!i&7+sgC%A`0%>Q>pW<`
z#9=?(v)T80(LOxz+rHP$-)zMiaqwe)^M5Qldd%s|)v#~9*K79AI(*};T(nQ+uD4I@
z6hiy%oI(4Q`j?-6Plm?R`Y{jLR}U$Q-#8h{N4(AjyU;u|u5OLv!8m*Tr94<C4*5bn
zST7G~T_LoNAM40aUokI_bUw|?I>?`MEq?n&rsten&p*V2{5<ExWt_fo`$LuwKgRiY
zuIcF)`he$yaWZ{+;*lRY#gqs7wIMsj>hWV;Ayk+6CBs7NN*zz_*@rZrQtcMfug$Ky
zRIgtDi8Dp|Dbh2}9vRY0<26oC9OOdic?><rJ=dXl@Zr3$%RdyKdFVW<EB?$w{vf;j
zv(KJ*)B_n>4^unrLG#v=+1CdOA-xp!73bc1e$4Cpc(M=Trw{pIk8Iw!{sYCKkGbFK
zNBWd|oOSx0`@Z{<`wE#JH17WDzGIzndg-`%_Z|13sT0~SIpUSJHQs->PCY^QQgx+Y
z6+5q9)r0$1$_xI|yycZ*%A-8-W1b)L_Ju$DY#;3(e|Qm>c}S0}e(>Xb7+0TU@rYAA
z_@ZZAUdHvIrSkuL-&dWo>mDI*aa(7ftb^i{AD)~S=Sy68;s@g2diL3&r$6D1zH<(p
zKfJKRuDqfAtrNF;wlCtq4}J9y`NJQ3_+r;MOznz~9Vl<S+8_0Z2lejtve!+}=V^FV
zSMu_Dh#lwCez7YrpNo0Dt1igadEKS&sEc&J<g0(<RXpP5M_r_dFZ<69{)|KXL2;1n
z3p@5NU9YbA_x`iE*~K?KJg7JG{E0)}^dLUH9<L46gV*6+XWJijlIn?HT;hiEH_slV
zr;ga?-}_9SOL$jD__7}IXFt>rf6lXUeA*YX{2;rjEKkTU8Csvp{HAv4L-X|cF%MI{
z^gO6n=y^nj{8*1S^Z0}O;<->f?5b<}@=?c7eD<Fo`eb&={HqiDs{itz+O^*EQ@({z
z95A)3E}cvJ&z`vPs~_v@;-&}jBVP9KNe_RJUC*l&*@OIuA7A>XdHskU9@K|*@?jru
z&^mP|KC-;hJmiNzh;O{82eSCd>P8%7@v<jgdC_A>oZ=TRRA2Vpdh5hl$i6u60`X8A
z`krQOdgacmNBb==h)-x-9%Q`W8^7tg+Tvu_x^hum@(Uc)VbG$fIUdxddz$CEdx(DE
zeAx&4tX}N5dG`?Kntk_6_ey@;OQC)R*@yZUlm`?aeRY`XJGc6Z^Q$jfN46ds7bj#V
zMfXZPrE#X?_)3wU{n3}~hxp_#Px<&<)B6IxZ=&Cc3!m0Q^Xz-y&3O|CJ@u9%`=0Y@
zUyzqN5<lb*AN2VVhk0ln{>V_=Y2BxG?Jqz4vm;LP{HG`meCU71*(I|N#R2V0DjQEx
z9rG_f_VHm{Ug@~k#a`d4L;7TTULQc~#7&Q1af?$u8n<6$=UQLE>o1<Xar9mT+YC|f
z_SJJ%oaw$<&%S-J9$(_6kEe9rc&cxm^FU7HvEDqM`N5NUapIT0d3AtC{EE*yaqz>g
z&z+rDe2Ozg{Sx2eVUJ(-%tQ0^q5a{Ho_L)j{1}J&G-S7-(LO-&kmXSb?TdN&v*Z1A
zb!*)HmLF(*sowv*FPX~xrQ`hL85Yua-sQobx-_pYz0c)!TRJZf$ew)^M;a%6Jdwr4
zk8$?YrTSvu{l|L##G}6TE9ahndi;>tx8Ax`Pn`6`X&lN6vd<r+Zyw59e(5+lU2h$J
z#3_$7FS2;(r}qf!AwBDAL-B|=-4FSsc8pt3hWIgWz3~+38Bb+)oiF>VkEi48q)5+r
z>c>9uM^F65Qxu1|)OmVteV<7GwvIi1%zK>y-5=bS)v^A;zxsjn$nvtzb4tJA2fAP2
zhrJZtH`I~3QlIQW@6*?Yp1XdRfeiW8&&km1fE3;9>3M#T&3oSj@`F!t`5jI8=x;9H
z>7~P$8M5Pt{dx^qtId#KE<E_qBerWhWZ_Zg-qh{mWtPguQ}leUHpGj4hx!v#_fUP2
zod+^>Zc|j((0O3jxiap)$B+JC-ncq8&u{974Cz@<77zRM$@IiwJv5IG>r-CY;}2R#
zrpHbpzw}bH&bavSZJeHQviatQ_N%_oJxibVy3)Akg3n7mH;j88k)d&RpgzqHyWU6h
zK9_Nyr|IWp_WAR^7@0kvx24~oX2-fTF7e41@@t+fUUAAp9MI=}DXLTbLp^&v=zMwJ
zn1|MR?izRQ%zHk|N1XSrvB6z$ZrQd5dVQ(Sdj48xKlzh4{_)~-SI<@F)AP;qu{M+k
zO!L!6@nqiX9&wW0r;N+T{X;$A%Q)FOdJxa%@hCp=(}V79=EW}`<L*sqyy{V$WctR_
z_|1#oI{knj<JMb8HeU!+`($XHIGsyKk3V{3C@yFp@Jug7_Usq_t;0VwPqu%?$?DB`
zx}WrnJ0H+Ge#mJ((f3>+L+6(t==C3&9%Rq`)bm|mU>{mXPrUjYdxh?w=G{A>@e~Wq
zi!aqT?%wTQ;eLq^@zOWXp79jrZG9nq_UNT(Khklsygd*2!9P7{UVXC<?Vma;gz8S-
zcud!)av^(##+_ew<l#Bt^%P9U)vI}Nl8Z(ARGU8Dy}onq_0jZs?>^x9=(+2AsqQD_
zDR&=o#opbPsj+{auli<s&tV^WKGXBuC+mmqG2#-p?<w)ee`?2mZuYa9$uG@Zrp7w=
zFYi}a$BsVXxvajNPkDPDcrJ-spEr+pNKakyD}VXA-{4K0>AeOIFr6nu{^;9Z&mD2n
zcP^p$#fKkt4%K;z&cAi))wv?$n_seXklM9wY2TZB?ymE9nVs8D`p&U9pmq8Rd-j_h
zaf<`8<GiT@d6{QVUY@&n<sXlp`{GmAo)7HVS9z*G<JR$;#zjt%9lVK;U$VTR{P}fG
zoGW~)Lwy17c$Po^*5eb}2kShS)Qx!LCr<Iphd=vb-g@WOdgs#q$;Udfb7frq;t@Zz
z-_ECfG>#|vvBSQ2<iVf&t9-;~TwUR@5WnK2PqyE9)K~DNujq67y!z1Zj9bU9xcGy$
z#pygj=Y&1{*(Y}CLF0um-5>TnkD>dLdG^TZbKX2Gq-PyHnBGf_L-yQ<?GrxKle)3L
z&W-vqPe0XjZmfgW(@&8-_G?4+!k_iV`Gd|0eSL;Z53*ATt;c7%c>lqp)}C}<yQP0;
z?LOl^Muz_G+PwSR$?eX2Ww%w@EWHQ0Z@3p(ZynTk%(F*!pK@-USEzqMbp_oc_|XTT
zz68}H8S2aG33`6<m&PR@@r%Rrh(7-GEqUY7eUcseWOl@1J-sbw4H_|X>hd-4v`eQy
ze%y-fYWSOZ^kpAi@WP5U{%#Ea;?-x>tLLTXJ=EtsHwx8nq4CrXnLYN?arX2_^WwCg
z42`Gy?8!Gp=b4?fpI8s+JKxYenIHDdL(gCO;!o|<X9wz2kiFEeaqHAOyY$I;Ad3$g
zXU}@`;!lyDb^0Dz9%ONn`Gw{7$ygT*1Os(vz<t1T$#c!~+q~zd`-bP%^l!X)e$V}u
zuleym@BYr&SFW)1a~%39TIcze&Wpoy6nZYo*nQglI=!#E_ow$=uN%CcBG-ocRk5i5
zLj5~kZ{9fB^Uw3udPqN&jkDwR8a;WkYd+N{TgUGIWADDBuBxuJfzyj1Ma70KwiqHY
ziQPoczDkS+C9%ZB5=A761$&FJ2Q_Lm79@5o(Rl3zJMp}hXt0Bdhy?))0Sh2V6S(7!
z&)>=To;&t9FJLf6{^1#Ot+{4jbFID4c@I$h@HzL;(0ckjcgTx8@z*+f(Bv2=&v+s}
zSSH7KqIKlb=RO0iGtWQc(76hI4(W4NC_m(ro9Cx^JD<gmz0PU(VRFcoKkFdBVHqEO
z{;<P3_X76;=sw`SkI#DXA_wB5kDtcLwVwa@*vo(M;Geid^YoMF{w=QbqdiBkhkon4
zfAN0XJwEkWXI}orP5$wzJN85Ndw!M&`pEY@<@t)8^x@+#d#v;RhX3Mi-a7d+FJEZL
zKK}7XzW9^&T2F4)6*>6WCBOLjhmYNf^r$QRkp282AHRHf&Q1C3lW+RSBVV4`O%9sh
z)`=r~tcUolGfqEy=rJ!Z&TY@ZsfT`gtS2AMesQ8tzSO1tW}SP&e+uUQ_P!gcC;4hD
z$^(9NE5Gt1FHj!j2QB`_)vx@>qr5?VLSD&%^g?#3D|F7I$-xhevlp@#8s}#j&3=5D
z7x@xz_aONdck@u5_`_fOpFYSw_NQIRCmzNjd(e;_;+IICaqHNHW}k8O2KmolaWW3Y
z8Pdys@q^^yM?>S}us5-upZu~84fzl4518ZVTu(m4m&o45-4}gMk?6TCG53Y}T<`hM
zxO~f_Iz>a{(75Lzc~>u<H`&2Hb&7^Qe?eQvPUGs?JWRR9t<y)KarTqTKl_9n$eu*=
z^^o3NZ=7B<B-cE9&F8pzaj`$;$GCOkVjRjd8j^>WfBh92w~umMyrF%JhSuxLX!}|n
zL+c<r@H=Pie|n8W@qzqJEQ?dwxcra{Q!cyNDSrGnj!*yMr+BfW48@6D@$j7H`Nron
z?58*78Yg#g_fPuLF81Vj)-(U?3;g!a!M$H?yY~k3tM(;-<p-Vlv|gR!BVQl%p2z=#
zM-D#qf@Y__>-~uSpx@*=zn4Hx#)qBig1zDk>4EGQM{z1c>!5Y)<zLPxAATs_;s-O|
z=(0G7OCo;jp*Z0eM|2|oGPIvPFWR4;=k5FK2lU8aqIqaNOndN^jhjzQetxhU@{@k`
z^`C<3kAG-&tX|zq)Q$N>bxw}BlVjhgOZ+JZjUSDl{G`dX5Aea{$CvBaYhBX#(?9a@
z<I6s5opJt&Gnzbo29nEueDtzA<410y_4v_gPs+!a^1dpxbI3X8?~RN@ad%$uFP~qM
zAD{K&CvWnERu_rp$zca{zM&!g@<a|)NAga-d3M<M(0KMyv^cUud?7t$$ZmeQN0h}|
z+><Z)%_p+gxVq2v^uRptoM+Ax@vzUu4SIj(ym5}Y=c1iM;%}Y4ZQlES&kJdf`!|$l
z@m42Um-zV0ethg?7yI4Mtt&(R!}N=P;>bUI*5O0*6D=Rsr#$O(9X*i$P@JK7L41&W
zcCZVIr_Y_#6aA2!vbd@Tc}cX+`&)YPC2d?hAwKa#n>UU&55<8!&~qMCr~Frc>Q?;d
zlMm~lar-Irlyu5N;}d6c^?C8ar{2^7Iq466{;^xW^=;$gNFRCRi@Uh8lYI8jYaLX7
z<j|u&<wqUKr#?Whb@<3j`|**V>+wT!%IvhBosb^$>?N1~iS!v4Z>SE@_{dHF@rfIL
za`<JO{zUa*9l22bnV0uOa>)}HpA(Ch`0)#E-oD_kxRaCTL$1dU)dN0u($5e3I^!jd
z(0cI^f8&sRdeQte4vo{perO(A2l;6pZJu5BF+Kd{kNmOMI{V!Ihx#bgU&`)j#;s4(
zr|6R(d}xSIee%P+b&#Eqy^TfoyYG<;=@)nPEbi=NpZpq6#0SmGpK&Pv^rw9G$vb2p
zIrQ*Loa@ESdT3u+pNNk>{f3?O5I;GPJT(9Cqw&Eqed2`Adi?Z2<M^y22ijj~>!AIJ
zhUD^xeDh`U^sD5<PcC^7e;J+Q<mEc!;smWntGmSHL(`+)V6IQvIw(Hking!xC;LD>
z(}#xSl*N_(Wqzv<^6=@;Xy+wA@I!p!&mS~B<nWsw=ec=ylY<XBfAD!<Ab#RuJ^RE_
ze2pjVzN~MX=ZC(>4t9x;{wY3-ir!oBi#_(Q`F|HY_LR|mHr%vHgWua-T<E<C{k|7#
z9LlSB$t$$4)R#D`Cvx)rjB!YQqIzJb-{JQAc>0LA(PQ3wAitLf-5<Qyg5F!DKKq7!
z?68hJdacKgPn_@THLT(Okxd$s-@Lex<NvAUpY=YUGjDyy-#+DsyxJe$i{Udazxu&3
zyFT~Ou1(e|<Xe8|^SwjAU!hOg@9snT81(mv>gtp+|JZT4(JK}At~GSU*9NauwAg%)
zqwo6tn#HG6Z$9={m9>lB%fC2!k6|m<{?2ljUNc`f`-pE8r>%0s-`4r%D#e^9FS%~Q
z)yo$C|6G43vFs^rk8C+&jpF(>Pq=gPd#e^>j+?sL&gZODj2&{@p#9pcQX3~{n{$pB
zbN42z6%DtKy5-W}tX$JOzi59^yDL^Kt~z+T?S~$+e6iP+yHDEmxh^&AzW$Lr%-g?R
zp?`}Pv`_e3hWaiPZzwP3$ro>Z5I@8Z<&PcYrv3ICv>u&EpL)vrQ(w*neCm@Q<f>0~
zrJmT$f8*q;GjthR2dyJN$N49p&a*tv#j#8td#n>D=QKTu;)hS%#7kc?4()gH(E1gc
z9r~zw>&Q2*ZulcE?2$itl3#s|UUJanv%~Wo`^@8K7yZ8H4gXU<2mZ4Buun6u`i**&
zXLTx1*2`bkrE}Q4=X`ZxJ^kuYKQfP2XZ9C6>BXON?0bCX*+CvV$f1w_kUV{Y9Q^i;
zdGgI$Zyf4}>fJa$*(VQ){Gz9^$X<5QkA`LP#D%@`WjvAH?9KTcCr_Nw(E2ZHAAQ~Q
zszdR&zm3bEb3z}<`oWLSzOzrEexsil*H4@i_#k=ehF$C-$2@!ZV|`-!O%A>6^nE?Q
z`|Wwa`R99s5B<Gz=jahF8|d*Iz)p5$oQ%V-TOBU(cS!`kiV@JS^qoB4?R)2>ev_#0
z=xh3`et@P&|3F(uAKLn)jq6kF!$)rFv5ubfgB<hZTSq>$9&H`|r16n!-Z=a!*3#F(
zX3v%0zd2u>|IS%_o*&WBdgnAg`w^N?TKyQe|BSOI*W))1=|{7Vz2@B)>Y;nQe&SxN
zzG3bk<51s#?)B!$wQtQQde6cx@~nrRFRV9?HotkdKGTL@+@itXDabpy^m#wva|<+i
zXrE`G?Irn;w?y{ggUN@kH(z$&)Ti_{$j+}jW?bz*^&(#O6SU9t5pvX(_t?%c{N~l2
z{V31&tG-}7_YK-UEJJmzZlU*(<RqVdqz~AC)<f~Jk5f+4)`_q8750PqMD`lT&u?}b
z_x{2Dg!Z?6?fs$O&G5SkXy0%0drUshKqq>CPOkU7^qZ$2vdjA({z7q5S7`DfdGf`6
z>--L)-%<2CC}{rh-{&~i@q=7w92zh4m!0g=KgHiXyUfeC-@)>E4mAGJWn1lk?vIu(
z{LYnezvD;`d-+YD-`(=L5hOR|`5i6tp>^!%*PX4~ox0-M9c#Zg>36)uS=_`QT2CH6
zG~|~$$of#9)}wu|j~?SmTW1_yriWeRsSo*QFFE>yb@sP$eq}#MT0hXAay>chqMsjV
z`!o5i69;_OL+j9KhjpnBzxgt>jy?G3(MNJzpR}*U3mO+c{oJ^`n0KB)=L6cgLcVnn
zA3f}Ij*<uQvj=*PbpJu)N8_g-4e`+j&BHSL_~AU~KfA=idZ-`yU0S~nY}~$)cYf*H
zhi$gXnga)PDD<J*pZ;l^);G2<^euhP@A?_n7xgoGPFU}O_r~nAO!4~M-)^wM`O6i1
z?sN4~>;81v;>2GcwfB@amn-xqzaI#V>uc;uG|$h}FAw%n+OMBLeGJVH`@z0)j%MG`
zcf`Z~w$3?Co;;9azZs{WKI2(;{8CrsLF?^1deo6|cBwyTKjA0WJbBiuS9OWD5A+jp
zv5p<`&tH1gF@AorgI#H_^<Q=5Clq(}UFM&<fcVsrJ^<xOzmdN(lxOSdAs^b0_{j6T
zqTkV<&!^;CPo8nK_2e6e`1$AgjUU#foE#@-VNu@ng{)h0?F)4JA-?9J{b0Y)XW!X}
z`W8O*p}xprU!ry7>!ajYkES>89riQ-@F(&Qnzt{YxbcJC>|w9GLUo`X)raS7sBYDX
zx<o^AecvwaWtVaGLGhJ0{gFQLc3$Y4&L8(=ezAvtkUucrGk9MC@g<svId5OnE56Bx
zCa1BO>(G$@;=vB%X#D6hB!^%8k$?3F<yHOCLqFPkB4}LwWq#-*M}Jc%jYaSGd>_>3
z$KFGEp9STGea6|DwDF9maq%<HpY+c-zw`|>6nFb8(f*Sc`wb2C6Y}usFZhh>bL?QB
zeoe3S`jI|pJv6Tm&<C9(_I=u6J$d9<ho7J9AYc6XqkfIczk0xDf5<CZeagFh%a3@W
z&41A-4&vhe?t8huPyFzV8+-Jfzf948PVbvfZ`Hooea-Q6mK)xw_B-mHm%LZ<9ESG1
z<i75C1)ukX<a-a~J&)%z?}3sw?l~&+Api1|b*vucRi4x<e)~@T)Q9uj`6HgrEB<*N
z9J*1Tp~sxmrSSa5K6KVGTE0?`_3Sk7T!j4eoJX$bNxye*Kd^g)1G_)b@!h3s|No2U
z5q{F|`B)yV_|AF5Z~kc6V*HeT9gaD3wPM!Uqr1JjM(4u*l7H`2{eMDpoA0#p>{XU8
zu6_5B&BpZWTJ+iF+?@~k@u~%R_O1Bp+xDY8iHCd|N84xW!u~?rw@Kr3?%D6DKWX|Q
zIpjlf&68su;zN_GJ|RAG$aO!GKmA5s+YjO-Z^qH~i8!cJ_iJ|P_w@Ta4d-q8Vcpji
z#Z^AUH{($@j?cL!Z@veoPe6R+sTZ_-LFbqHQg6`yVF%<_%BRPA`&7Lc=LdOEoswhR
zJ<Q)9u|Ms$j(l{=MY9Xi2aPB4tIU6T<sIs``nP>%U+M4q9va%;P#*Ye9P*Pr$!{DV
zIcWUsAs6O2+B$sXnpdahp?T}k5TAG&Pg=Zk+<N-?&pz{L^H5wMJI$*H^@BDKop%r)
zdE~2e`vxERiR#k0I@2HABb*z?^?COo=zf9ryl$Ow>&c}TCZBrG{eaK?MSK<&osad<
zy`&yr&RYk~^8<=o8R`el<ubk{<NI#$^lKkh8XE_3)epXEs4xAeqyDE)E^dGATlL`_
zP-pg~zNX*#9MbbQ+VhR)D)Tw+IjZcr8b7)0f$V~5zw;o^2jj5b`GT(u%htPRmaVHd
zUM5$a<a+#&-}sHgtUKo~%=?16BroNYXPtYBaqIG2&w2Kd@4m|}a`3?%N9Q{FA;0uX
z`vIDFo<aV(ccGp0>X96L)|<y~-2DJH_8wq8eQ0|4Xa7KY+-K}__aJ;wJoE|qGcFGN
zh3uh^on`*XpL$9?Nz>1+q{)+a`$v75&-LV}qpY`_N3%~npnQmj{HLGvu?wbtei(<w
zt>Y(Tk9?>Pb>Lo#U!RpX_dfRXhac=0Pj;ZKV@JjdU&=8q&eo&pA&<S{4%wOO%%@!P
z*h3E*Ke^(8Pu-#M(=Q&mU&J~2<;D9a&+GaontkHN4)g3)H~82ue(YoiKI7~bKkNBr
z9TW%lC7t!lK790B&ko~|9XW5Eys{gg{UHA0gkKz;^X#LS9D2mtIDO)QPQ(xCMT?WX
zvqPNHZtKiLcAA$@_K`zh8S>M7e(vf08a}9=(bn@zKGdW0Nq+S)G`snOmLL8>e2LbH
zlX-a}4-LhOJ?e;m_HBM%obySGllX|Q&yUdJtUk%{cM9GwLUogMpx*h(fAPqAG>+fA
zJen6*@=~vH{N^FM%+m+U?iGtVE}u|dp*TZ%VlSE>;x0aD`OJLFtNJmIR!`75ZofeL
z$-c3#jHCIfkE`!8ls9=m%eQgyAP1j$d4=?|N4$-@pO}Z_X1wJA(kuSP>1QWCbxm&S
z;ji`VWS96FCl^0E=t+Cjm%OV3NItvqsWbNKf8?t>{AI{q>(J~@`;tyQXvlB%PA<FY
z6EFP6@xw*k_qksE@*lDfs)GeZc~c)~^^+*y#_Q=LFZqnC3-S{2C8j**pm=;$QGV?Q
z^_+-L-N799`yD>t^ZBOpI_s^BmUsORt?#1~v+p|>Jm;6`<BvX){+l;$|LRNP2l2z)
z&-$x<Z5|ETOAe$rQNPH%SO@tDts{>;{7qyRB-c8~4s!Xe&&Z2?$PV)G8)t9sKk>px
zzWvS~`<Om_(0m!czJQN^iyOs3oY)JE!+({1<ja#h+Xvz*Kd@~7u+#cP=csXgr3~?v
zp}HoQ95m#Y`=@b!Sm!y#d?LOwO!+CVY+czn{pQi)!7p)#@~WQj@z;9t(r)AQny-iC
zL;UQrzjObwmp*8m{l>|cfBuL!KKjwHY}~wie99%Ce*44yfuHsX`_WKe(~sPz$k89!
zfzLQT`xdfST=W^^>@hAMX!FpxI^wVTpvO64o_+GnetnxC<g!D2tVf$?mv#J-5BXuA
zd_v>Nm-XQNgLTk)Xq`ADvacSWx)H}R6c2LiAv>Y<P+ZaK2rUoxpL#G4jn}isyg2b&
z-MCM<H>wk#?^?$m^5jb$kSp%|R5#*5UixjG{4$zc-_!N?`#xv(eO&ig-`h6MUvb9g
zd%-@Z_W76lGgO!UuEO^}-P_&cjf<0U>->F2&KviB4zh>6Xm;he_cuPrf%y4r9h7(c
z#;r51zd-#|evGRN^{O7!pL$jo<Usvcy^yCK$<_DJ>exDV(N8Z_5BeSX>f8B24tvq$
z;Nu@8A0Ims`Ns~Zp77ge=8Y%PCl2iM9GLU+%rAC}FU)a%($6lKezC`XQ|IDtUY+8j
zAF4<7qF&QKewFq4CF4sXup|Oan&@}VU+0W{WxqKmAU@x-Mf?6%qW$cA_WdsV-TwA|
zihke2B|nip_PcwH^GM&yK9zL#D}4;=oBrO*_t)IFom-GU^svKxu5(X;;x4}Y;=lO0
z7a7ku>x<%<_afus2<exP3FF$&-|d-}wZDs6<Huu99Qm!43;$1s`<TDa8hlgpN1vOx
zVqqP<{!Yt1&i4WRJ=uc~O?r0c(>fLDpLOD(ZhvsVs)vm4+`7TvWAR6Pp*UENc5j0G
z5J%@OIhjB5#0l-(MML(<GrhT<K6T@KcHdE#&Q0}3F8|cEdHY13(bn;gUF?ATARoGa
z*(dDHdjgu>Xy<wAw|~Tk9pYl2(<d(GA^-71`q_`qIC*8r9`WKoIrNavKjZAi55)rw
z<<q?h4a@GS`e^3MK9Udd<`0^FeStsf9C}{z{8NUW`yl(p**N<=Cqd6u;_bPL|Ku4L
zFY=*t58_Y#<UxAar%xESo?K{NJ+q5m?@g>{Ct4o$M}0r%^<jP1x;#(h*?G+_Xk6bi
z4%L&oupgkhQ1|MC-|9;qvku9zU!ZYz7-tWCIgc;VepA=_2K&(vKYjYAc#1Qc97s+X
zii1AHE_U0GXun$m#Z|p#{Ta{m#JWUvsD7<um;CXU-R$EhyRC!CC$8>`_{5z)bwED5
zjMK+oeEd@<@<K0UmpXF(*$?)Wd#rQSet_zMUh!cMWUn|v{<EhZvX@`-M?V^hr~HZ+
ze)gqa>%?2#T9>r-^m-4VFX->a^)-2A7dz!aoZW}z!Tm;l-0#$taq)uc0?j}9e`??n
zf7xw~`4zu+=YG1dn0`7>_$`m(?cA}i)eSl7m%aQDfBuM%d47v)j<dr)RLATxpZQlm
z(7NP9vx7YSl|AVnK4_jk_NqI2V2-<&i4!@Q4|a+_ntt|@BmVrcztu0=Jb%QOUF76=
z`eD6z^CM~U+$Y#UKL6AO`}m1pz0qr*vdjL3>{(P~$Kpc!45|zDBu?&6`jPX(Iw(%!
z%nx-dKKK*;?l|=R*gaWa)o1Brul(x=DIblGJnQv&_gAQoq4g{GKjV4-RUhO+eOo-$
z9lNZvpY(Tm(Ep+L683}mLvrnF`&eG^*(dCjFXQBsXI?yVo?Lw9(a^mh^_Vw~mM>_Y
z9R0<9O|-AsV?DpokY4*ke$*cn2j{*zHLgC?vw7o5t1orSFZ$IbKIbVk&VO=@<0FqB
zkX`m8yM1nGp4}NAdf1uz*p>Rx@|%3-7Z<8yb#K4wgU&~Gsuyt-2Q*~Ayg+fvdQ3ai
zJG-1)XlUQFlN_{u0Qr%~Zg!E6Rv!<XJ7BY6yEbpoU+8C_bHRFc=-c|9{)pdxRTuUr
z`Dpt@9*pzHdhr+U%&Yefo_ioZ_Y8I#*O&Ece(U4TRe2Bx`-h#7{p`jkZsf58vcr28
z^~!Gcu+RJA{7gwctdl2wRG!6~o$`P-53T3F__0Slx~HMbP#y4}-Q+@XQ&(tp%wF+_
z{GmVBi>LUQmml+)2lhgC*l+f^xaGWkp`O^uZuL!`aeX__EBiu!hxRwxxd5GW`kj83
zdeQW-&pP(Y1Ao!bI!M2J9X-bR>-+0y@yl~cJ=uTkvj5m`z4uo5*(u&I?~C+`Ya)K~
z_{R=@vhTiL!y4`%*`$F!>(JuNPWG!aae--1)=S3MI^+1!?Bo~!t%K?+`_aEU?LMym
z+wb~-{-CevtNQQ13gunh=KeGeb6)=B3$6ap*2%AYLi35_sM9jE&c0`ladwc0HgDWI
z^JVtp$CrBTLve%F(JTJO6H~r*^obwD4|AVdpL&e**E!`JbpB?&I-i_l>f1T1j+>nP
zRNGZo>RR|+QQtGrhssbK^(B5mek7lBPajh^$#0!;`M{rcIdAbp>(KH;e&#{H%W-)$
zp8C<^0r`o>2l<o84w&=S^TWJ)PUJsSS7>puj=kdQyfIEcG!EIF_TgtQG@eKvJ0N@1
z6LddKG;bX_#_bdL2=9G7N0_&7oCn^=``u^1_w2nqe&{^0KkZxc=(kUOzTo{oIglLt
z+j(Z39P(42dp6{+bCA9Ijd{;0^oS3A_~n6q&pGTz^V2vx+2NeU&klN^_@UXw4|yfu
zy!gr|zsO^kerp~3%v&c8^gwo_q4?|X`Txy{x9=_K)BF*aw1-{dFCXe2EnZ0*7YEOe
z`0OM1e)~1|wZ5Vs=RDeaXk1^=SI9FDjl<-p-#pY0p}uHdAIo`q&7%`D4_P1TO1zig
z#Gm}-%?Xvo6$kqd%8&O4>Rug|)u+6xAN7addlYi*L-X`IryxH2-2T%qtY>GUdG*Xb
z_3VCwF1yF^k6mbeKz!UM$fKYC{@;VNhdgr7$;W>2O+EBk&mVrGtxJ8z@mW{)JgZLi
zS$Q;$Hg6otqjSo*zF^*Z=Zy0MKU$qQH=Q5&Ao*#ZaePqSvInY9n0(}?f6hbSA7ih6
zX&<7UZ~TPhix0buqwR}4&&Xk)_>$*6jdAtQUUqqZlJBk5kLOs=5$wr%^{o!ovARxF
zzt-6={8Y#K2AUk}=!5>>3-a(m_fF3_<g-UTh`)OiesSa<J!tWP>M-lmJ;*x9PWHH`
z80RMx5Bzzqo7eC7F=N#SR{P<xO&i2ppTmbHmmGCxzu+?u$)OJ)Bv*W(br~=7_J=yR
zAI-B%UZHvMq~Cnj7eDB;&Nx5V>7J&)^FzJR1KGn5d0@XfR7chshxpC2(|jVkAb*m!
zj{KCHcHy(mJU=0SlQs^`!yLCi>9r4`aVS3QQ^)d)rUwo6b@u4H_|f={L-T0zj9bS(
z$bR*q-pH{Y;!~gOh2lnjS$(PpaZ@MqY95V`T*xoykaN-g^EoJf<IsMhAL^I#M=l!j
zTfg)<ll^WS?H&l}(O2;&iibYHANz(JeOzBuuja|&HyS@XtP^*9=8dZZ^32N{x#Zzv
zk9ez-)Wc7Ht51GgZymkRyn1%85Px<-a*|f}{Ij0jX%D^Zx6jlIv|c^phxVO#i8uMy
zJE!ai$gZ-u^Vh!RAA86*Z=ca;-Z($mZ63;tdy>3ZFTd$0|M{)I@k`zF2fAO02lQTx
z9qLq`@v+CaIOqrbMAOGla&o==x*zjj{NzpC*g?PkV;>r~-u+zM#KY&j{=XM}#<;)d
zL%WyzJ0hRonlD592eJdLzKqMCK3OJrVaGS`)@R!Ai(52U|Iptnca9#>vLXF9Z-46-
z>@*M6BfpG)QAn=*(FfUu5AsWWiMRO$MeFT1{i{qr`^1qw^4wVc;>iE9b?maQ*ah`9
z@*sO)nSAFDOg{4L5Au8uAfJEDt9S81n`gIiXkXF?jgyD=c`5tN(<5K}GB1ACL45iY
zyX1jfXx#HYqzB^9x-nliZXG|=YZ<biJob^pPyQQ+*2@nXKY5TH>LmHa(fTs9&OMV{
z_Od7U3wihx#SI^t9Q<hW>^9$6w4UAKYTS8DzBsZ^9Li|*f-mjLdG!l(o?ZNBk9f1+
zy(RskpFR8|kDvM%q_+&yfA$zh)60%T_Rxc7AA4YqyN6hBz6|;6p2XgipX2nn7m;rr
zWCuOT@BZrk=l<os=6$jEm_ENY@1BLveT+Zi%^s*u)K{W)Ij%0f*K_~%IjwsqbgxG9
z(>m{I`DLDbc9tRg*oDtHG|n&eD6jGa<w?Bc+rAQaXk6UggWXTuyXB1@_0JCWLHifY
z59`tCkA0)==z-R&Tm0-d&hJEi^OJx0j6>s|AJjb>A0$8P$GEuK7vgIF<Z}?Y?36F^
z*+H&xs9xzo+i%9p_>I%gfAOXtA3xYlPg!2rt!~LNPEOMHnR$94JJI}PFS;JeoA2TH
z|GE9{i|1g!-%^J1s-F01pUNv5AA9v7<M`!)ANr7a_T%&ayrc2^JkK~i_P6}oZ_vK9
z&!PQoy>WI~XFlHp$S-^NFCNhMR@~p+N9ogFJ%>W}LHB9%>~l{+vxnXK7k%c{9XrkI
z>u5+0xyGS!s6V3FsXvyXb^O7HhU(D%l5hLiygEVKr{X{!J&+tUOn>-Aj{3n*kN#oY
z{xYu*xo_By>}3x*{2{OGbNaGz>lYUJX+4@h{Il<K|L_Z{NBh`0kaF<ToAHoucIYqU
zuv7oFUi{@(T-6mhkbZR9C12KOUDK0je|xS+>jUZy+OOG1jN7034!(MbpB!=1hv--L
zX)oIQDeL%4F8^Sz^Iik;2cN#`c^M7mk)8a)?;gk=@~z`H+B`qJ|6vb5p!WgC54~h`
zk9(Wd-becTaYzok`O9u~M6UO{n|^rXi49$vHTXQ<dUEmW-|7hR8=rjfhriYv_r91N
z=tOqlgZPrpI9Ts{pxxI$a)){Qw=38qFY+N?XlUF%RKMo=3B`r{MC-_vAAP}kab_?5
zFvr!KzM_xmC-muO>_Y2j>~lU^uOFlN%TCCjCD>;8ll_KN7c~y{S3T4x)lZ`Q>ks55
zZ5*Z?{-}>KRG<2aeyNYrn|;H)e!(y6^(+06ed^KYH|CA2Pkn+sc6tv#YmcJ_?KrJ@
zgLTIB8*=I8hj_8qJ~U5`dHo5}r@zrtTX)n&`@Y@kt*SiQkM?c$56^G>7FYh0@BU;x
zKjm4Uzz^B&cN64U-o(eadxQQZKKhA!kT{Bm`f|^5zty*VKi0haAAS5!fAD+H<^IPG
ze(3k^m*n6_`@0tJ-SEo~Kl!t`QJmyCQQsFYeycbB${&Bs%l8sofBvW$A6Dzd!M=jx
z2s8fj1NANSr_P}KpzSwx=YHVcpkMgADBm;o9>u=XFQ9!$E;OE)&vTwz>G3{<z8v>{
z+V{7_OW%O@zdj;f;w?V855z0S(e?p9Gp=Z89ojl}L-xS5v)(v83yS=K>`dA?`}H^D
z_@Qx_<Lb|NqCT$9)P;H|L+{t<Wv4pS-}&Jjh3XSsw$6G;uKHxZ_mAxNeq4Q<r-!_Z
z2Y(Xr)9<|;I_K$u?119Ue{#fyJa(8TkA5_C-r_Gq@*zFqp>B+$&Fg1rhk6lj_Tb|$
zWOqH}hrT0D_8~u@arUyGKTzG`%k}2Tvkt#;=Yu}u{BVxzGxbpaf%J*DzR6B@;Db4i
zwhkY;5FcbOd-&<x<exa;vp&boJ1-KQAMQu>&UfbzKF=H0qZ8e$-P1iE8F!!d9!lLo
z&l~h6dJbR*6tDDKT*XZtS(lh`GS3b)Onqo|3h8C1acEo~javuH^x{`P=;SvJ*~{-7
zH*Y*K=jrDUyCFSj`v6V8akRL|qy1X8zw8faJ-KL@_Tr--;)CqQ&pvXYd3KYhJ~NK^
zA${bb$w`z4a`B_nZhF`+uGY~jo{8k?gNf$Z$qtxxYW%B$`ryC2z24vG?}_?cj_b?%
zwD}x2zp%(Yc0h4eCpn(;?2{L?x<)Uoo!0rB(Ydeg-OG%-@1XHP_bB@hpM7Q=ok&jV
zA&1}Q@taS1Wi)>FLiU+=ucU`Od@%3p?1SV%_gT*)P+vv62g?I};*fUhZ|sFWXY$^#
z4C%LSVUd5HXZQm>f5Ch%uufbP&C3scW#~CWJkjJDx4yA{%S&VV7nDzbSkQG}7QcFT
zJ~}VmmyFv->ec<hdiCu*$1mRcBy_&JFF@~)?Hl^6W0!cTC$xIV`Y^7Z%TQgbbI-lr
zkB|%1HF@~GpCiY6H}3}%y+8Cl3BOlFk36WaL~+s&GLPabKjQ0L^L%3+eQ4|GkzaWy
z4<;XeXdb^lt6t<4pLNE|P@O=2K;z<J9Bn>P-QgFPw2z%-Xy2$ubqnR&docY@9jOoF
z-kY$CJ><~ixmO=`p7@@Ga|zm?;=q6MjjJE`8qeGADc0eKJ}<M*INEc(d)cIQkL<Vp
z)NV!3>Is`4bX~V%?MHuc$gFi%FRWWol)qdz=TG11u+NdL8uIVk#LK=D&-~ofy!}MJ
zadduONxr``)CY}=CmN<7?w!WfMPkmE(Z-?uPG0r}wEDNd)su1iT|Mff=E<>MU$D+T
z!pAQA*gE}C+@XA^U-3lKXFYy?_&nXdrC(oShk5&oe>pA=?8fhWWe@q*+u!VDhyH<&
zT=7HO&+?gi#5wCqJnYv*^^cDp<My|@w%&U>s88rCxj*b9>(wdw<al1RZ`onJ?+dWg
z`HUaZqp#Vg@=Bigs)MX6_K*khdw(fD@~R*53*tl5hi0es>Po!S8@rq@xnI<Qx-$>S
zbAK_e9@VLFc0%>(K1H8-C{E_>OXnUv=JE40`P9ES+h^j%FNn|nkZ<#lT=l_k`9deM
zmwa_ejyUOmiPo_L=6KTdB+^Sx>T4`b4&;Y=k{76M<UyR}QNMBzk~iz{(Px}KsQ<cO
z*&pQL6BqMozkQRao>LxLy^^PIvtND47klxE3;W0=Pd%9z4|${Cc%pgo)IYuI{L4c2
z@*6FF^2J``X#8k>M!fVFeD;C7n1|vc9^z-cJfi6#*SbXg%zjXp<T#h;=ZE~0FYngj
zV@KwhT=UjJ{Niu_y4Trvo@4OYxAvWRwDq~4_)QNv<Qh-fc*?Us6MZiMdcMcUKK9s`
z^q9xb9{lXa&u{aPT(o%d!#FffE@Vf}vx~iGXuWl5kNu?%>@W4APqU95^zoZ~b$}0g
z@2!p@Jy2cR&q=cr;)BK$%|mkW8BaU#L45iGI_Jr;U-gZo$+1t(m!Y`nSNe=TVVpjQ
z4^0j<k52vU%le>?e(|(k9AMeF_&FEFMSRtvdG)By(dt`WtADih)<OF4vC}x3ooMmW
zx5XD9G@tR|pSa>@2Y!4><2Q~bk6!B`yZL7vT1PH>#LIeo*6|;WPn~$K%yIPr$&pw5
zP(N8vBws!8Uq00_J!tWCj@u97!(M(H7cctsN#p!em(CS5Bo8egX!glJzu04(|7i2%
zqscRlU%tsfCm()@A6jQV5nnx&H}!?Kj$HcqmvLi1lplG^b^0Ye^g{JxJ`rEqC9lp^
z^`Vd9x1Jt;@{7IvQ4j1ulV=?A8;S?{X!FJs(++mh3)P$T_>4pIP`vmlF8r`Bjh7*P
z)`_dQyYE1Gai3!cIgnjtD1PR}E0MqKvkuA!JD|_M+`|&xi($_DT+QcV=tS$p$Nun~
z<vGQ;zG~h$KJhm%Z#j<Em#t6v#=SpGe)IC6F6G~Ign9W<m+l?biD%Y_{;odkL$vkA
zA-(wMFGF(F3$z|Bu24M14-NHM>*YuO%sU@F@51CKhkX0hIIL$kK6=?9ujE7gkpJH2
zimQ3Fe9*%m^XzA@aq^9mlW{gL--+atE6?P!H<3Q$;=o^e$W0_4pM2Lta^wT8pOMEt
zean8**R12OI3>ylKOsJT8&4XadHV3NQ#|nV1HbWlNDq0|(*w;XvP)f<S67f+{7_u&
zPyFmMpMDr8m;dZFu0QH~_^gNQr^mQ+78+Mi>JQCsa>T=U+4BN9xv$j|)KALL^9_5%
zSzPH!dGwmM4jNA+2bS4o9A6^2;#CjrQ_nMgZ@}*=L45w+&hPyAf2{oeQto%OKH=|7
zeLj`)l6L+&x7|zpT`0Z%=k&h$^j7T)|DT25Pgrr6X;)qFy-o%HKDunH{m=c;(uL0#
z@sqE<eIIDTxc2jRd!}U}4*Z4U<$gvkKdtxuBY(Fme)!o}hWe*_4Evyc1;tT(Qa?U^
z`yDEF8+RX+2k!;hNe)_HaIX|k^WvnR7|%K-NB!^{O}_Q!)hRx8=l*6Kzy7Ar=zIE%
zKHz-SUz0YD-@ZiKxArGJ`U@J;57WNn(_irAI^+B`&mQ(B%8PaErH>r)(ar&UkUWSF
ziff`cvX4CaoG0o=-eHcbTl)&)SD(3#-SqPtil=$;vMy=kW%+d8@Kc<~w?5~^gIxCE
zV+T3pk_WAW)<fgsM6Y@F(*rY)#v%Q%%wGJS1IhJ#h=%r|d03{0{G2B*^^&g+ApS(_
ztxIH|`ZsTX7-uJakUsNhc|k+=STFAMnuqi!l4soaU|_CiAHC{FJ(0(5eAdAnXTN!Q
z#78a~KQzu>_dD_u$=5IK59oZ-XT0~}uk|T6{T3JYk#8TflYjbvb@bAMwvQ9ZX9u}Z
zeL(Ru&rURZ$YV#wjX&a&c33BVxgLMoV_(TD6gP6w?1%KA@w1mb;w_&1HV@SUzC3@(
zp)Zjfc9<7$G$c=4jib$rA3F189W3ih>|&3&LUQGS9a&G}OCEjn<44Q0adD#0JUQ&s
zxA2jN4~@_F1?k}jn!WWfpKIMy-Cx~%jl2K4XO|(r8jI|Qja|2(eB$r>cu<_tP`uGF
z?G$&&E_pL<9Xk2cCp51P^%Z>`pZbOP=*LIDzMv1l><8AD#liPS>7zeU9i}{SC*OX-
zuOH}7>^5Jf$2fgx^2@NET<gSByzC3luh72n{$cXSS#MtXc*nwXVV;Bbo#zGbAIN*~
zp-Io~d|IdC*7F8lF#VfL{mW22Kz4d=VV~!Ac6wg-o&k*?pWn$w^N;`HDSkfJ#Alp6
z<anRpx!>nc=IOm`%JAJUJG^t@y^8lT;ydf?(cNBMqjLehSMgqgKk~>9`p6}ZKJxqy
zn)$;vTV>6G13DDYdmiZh3j5{XJoKK5z23+8{c`KYBkh-G{$>8fU7qCs;74xw{t>@#
zThMPmHx~6J{PwH;Xk6d4|J;l81^d&!ht{PYa-n(qTK^$mpJP`dyY+2y7Zsgb<dDxV
zXdb_DexbAP=Dma8;%uJ(;>urg*_Cl2&w8}_vyL7#yXiwi_eAsR)p;Qv_9wmai_dv#
z9If81SNH7pK9n8e0r`W5;^SOr7g}A@tFG-&_VJh9?64lnx4IA)<K8Di<M^EO#@&OG
zUmeJgJVN@#1wX_mpJlZ1FB&t>8DHm*d$0O%Pj>D&hm5Nm`B8WBMvuBOZyc%zcA&|D
z*6Z_V=sdL^I<M82arzVaQ-<QHf1=q@w$Aw@ZsKg;=u7;yZ^+B~QEz4SqA%q-@~xwv
zeaSB`Ul!)`+!A`>BA_4J=ju@X%fGtO*Yz>|53TR%r)d3Co!M{pEmYT#ed=*xQT;CL
z`UPD_pZF~3`Z9iTbS`;6f_84v4~;|X{Qac$(0HQt&Pn{xxV+;xkJe`r{e5eh|BHJ(
z`{|<Uw}13?^x{T-q+a!6o}I?kv$)`sM|~jqQ;+j8$I;}XVd_ESr=Q)B-TIPs&Qtcf
zFL*Au4u3tw&kjgFWG~EkqOE6l?yJ;?<~MtD+&-j7eCe}}y{R|*0@^<Jd6a%&9J(iY
zuJHVR)xq0sKlG603+IIX?>ta9#?j6v{uqbGp}3PnewkkLP+VZjbAEb`A=kO*{_EUz
z?_-bqsde&04_vrY|Le0#{@wHCPdv?|^#inh<aadeL-S=QkB}abCnxufar?yoKd}V=
z-~Yo=76<(bilcs35A9?5)-TC3kGB8pbNkoxu;(K8XSDk@+CF#QkcW?4bw;mxGVc9{
zdeb+ocfV#2J&Ep@;+K8g_W`|!^*-5qTJK5e_xuaRS3bm3ymKBOyY%nO2O5%-G(F_v
z6G!sV&MD`MIA{FP^^iP#&I9qHpWV>7{=^>jfDf(Di95R?K7O(Xnopz$AAiW@55&iQ
z^2k9arhhq)CLfY3&-}v=)g8JF*&*)qq+I^tCyzbJPd~Z#E!ukPQf`i;t)sV|yj+K$
zpY};T6mNPL6vcy|`j2t<Rd&!P?#XW)KRM1@_L)bMYuvgrw9Y)eIqp2yZ}4ZnjN@Ym
zG|yk-XnOoVOy;diyNpBrqw%}{=REt>3ls-_keA3lXdX>JIrO3N^B>Jl>+LUn(Yfb5
z%zL==*}3ffw$DYq|2FP?&GXRr!+kF9`}6JzK1U_T`AI(-vX@<GakqawcjbE?@lhXs
z?@c`Q1Njq2^7VUkqI-76+xZ6N0ph1$p811@@>+)cG*3>-XMgIqPCv9Bl7kknv`<{w
zV}FuwT)m68b;jvSJ*kf$&^jo85T7`tJoZ8MK=wj*Lw?c^#g%?_KtFxt=Kdy!9yEQ4
z_{fE2<L24vK7&@r_Mv;0{fLiW`1zN9<5M5dxmS;NKgK6N`nLE%e&b^ozuAMvXFX&e
z%yIh7%O@nCJ?O@wzNr7|>+%BeLH$-=_PvwFA~}hk?-v&NEgl(nG(KotywT?M4QO7z
z<ik7^KhKfQN%f)b)pJ?@$7dhpIqDpAE_y#kzvpD<slWH~J)9*tXRp@}YE}DQ(iatv
z?4!;*{Z`*V=Q)D54u7J4iO>0@@9L{iU8`I5Y@FTZ*@sS4zv`46<B8_U<tNN>`vvmX
zKC_P8GJnX;_!;*cP98ntYTkPG7$=9n;sNbn@~uZhebK(wH`oK&K`%Y@lb1*@KKa(q
z$Q4ie%tLj64|?zAd69kWkT-hJ*0U?uu|vGccW#%V`#if6t;3fzyPPY|WA=#$KIe#W
zenR%6)t5Ms3zOge7Ef}lht}gmL+jA&faYPD9Q8yFK4?9}#}E609H@U~U9-=5>3ns6
zf!0HE&5NsgOMBUej~wIBIQi^F)0g)e``UR8eeT6S_K91@(Kvgo;}84ve`uV4^w4i!
ze#~1h?)>n5ZhebC&~rKT9>C|A?kCo}=j3|<@5{@^&HG%`=hFPqU)^Kff6dDmzoC3W
z<Kmb7Nxam9ycj3PIK621Ip4>2zxH?f?%%#IZam{_y?c%CVf$Rpy!e~fzx*A(c$0(n
zJ$9dK_#SyZ#1G{S`h5d>64^;VyC8ouZsLRw4Sg?OeX0{Q6n`{6^-I5f0olvXw1d6Y
zmtoeceGAp6I<s%pnZBW)s9$}=I`wTmxllc$$!DK&a-ebcrX2RM%e?j8!?7P)hmRh}
zE_On8z|?C!ezdrVmvOZH@7!`eqsil+zHR@xZ@6#i59IO(4aspY5f}Q=;zob^ZCpIv
z7a{qOUV5Ap;(<>6;z$p>=uwx6?)&29{^h++&Wk%FAF@ll-3Q5M7vwjjA8j4^5Fa_l
z*{d#~yo-<c%D4E6k38^){4&4g*}9ZNpLzPw*3*}I#KS$BJ;u@WpvA{Hw9n{Qw`g%<
z7k}`f@v|EZ`2qP&E+h|az5Ey_hy76flTV(!8@HYx^kiJc%{>1g`R3V2POg^^>(noM
z60L{yvRmH8$2>c%OPU?{(d6NmfAvBhJNaFP<mA153H|>^Ks|_~yofVe{N+nt(77Hz
zIcSI<4cUVxAF7j_*LR`*Z=G>8`yqKyl-LRF8|%ok?=oKG(T~r$;(T)6xi8>P`Dp%;
zpJ<+)?6-d*d+ir|&^)A{T=DSzb@TYy#UFl)13T-Xd$s35-xKrvYTSEs@5%FXXU}EO
z{oHedao_jyoZw#RJV)#Q`n>hf{f=GMxkr)54$ps{Lp<Ml?)4m!wC8Z+J~zNmKD!b<
z2dM+~A<ocq5<UEJ{_%?+^jas6^5eP9IKRoaK52P{;;4T(Kg#M@eexHdJX)XgnQ#6l
zA9>I``{=iip#75Q{ic0m-hM+vdLe!6;2$*JSR|J|ajUmpT{#!<>$i}-Q2(%w9QxTK
zzL_WX;d5`nFAn75gX#;P_2$`$<`<eh@<pC;e&~bxBy{ef6UE2+GJfZab0^Vzd1#(L
z`hDs*?i{mD-5R%!ec}P}TMyX{?F0Vu-#weX@=G6-FY&kk$S>1tK9PSpo^serAA7_P
zt*_{BD_?L&m+S6Wrhwl2LGr{`Jk+y#)5qx}U;oADz1D)FJj#!Fu*W<<&5Miwm(hBB
z=Fy4#wvTc>esb{po){Y9C&#!v8&^mC(x31{{AhiFJn@0*7_E-b^r&<4=)vz^CeHMW
zllzUl8y6@0ho9o|?~3jV{7k>mu(5o}FK@=t@*pqbX&=%D<x!lSE5^myJ(s*PREOEW
z)q`>R7Zll}PwP|ajy(MZT@S@keOu2Tdc~#8f8)++aaF(O^-J>AJN?G#wI8jA{FgWA
z9#e+$0mVmL>4)ww{IBs%J?aP3x0qb@oWV}#nz%yx<V&3Bhy13;I`xCb|3zc!kr&7x
zbz;2C4)ajH#g{z(qRZsvxOk}xc_Lr^=?n5yhWO-%Jo5D!ddXMU)~9^?!8|$8IK9U8
zFM5na>r$V2^$1gr_4beXj30l+p`Kmz<A>ymm;FGVb?jgtI@hDgmoMY=pdo&FA>VVq
zdGgGwWAo(EgHJs9BYx_e9OsvD@)GTP`$yc|qkKPyT>F^bl!s<FI{ji-?z_g))<gEu
zBfj*b#nCwAkM+jup*%qHA$e%^h=%OLUxxByKJ6pNxIEy8=IsNhZnIzTk9_?6F-|@-
zPM_~N@Y{HqK6=cPE8pb7<A+`{y2riE8cvw?vz>R_uw?^2b*`Qv`@HXRFLVByM|<z*
ze(m|odi~Hm<OliA|Lm{UL+hYC$dhq$%GTk7_~e0Hh%ft%dHYm95KrUcUe;IaTl4bc
zyhe*Zd5|7_#wR{HYsa5Yo?9g!pZyKlt<QO`BF}orF8uh^Fa7Q>o_opl{RsNSiyZc#
z_3sG-wi>z3k#nk$9`pF^3;R)A`ENe;um@kFeB+}ZZC)MGuRhf!J~Tf4sVvX*@yB}l
z*kNA2pg8eMe#jL+`6W-EwH_b6?6*%6`Co5bT<Bw0qWsYhv;M@vJo$<2%5k*y;tj16
zCw3XvSHwr$<)g7kzBpKiHc!9!85alq^2i>1;s?o7muT^T_#uAt;t16@{zU8Xs~_i{
zaqHO6Pw}RoKI`y{2iknfC5Qj=s}7)Z%=zdXgx2X_^zfJ6?4S?g$44)INRN5v78;-P
zn0@%vEu>!@tXF63!UxO7*-s98<rgjQN#o;BzK6gsA86nAl^>t0ia+#uUK#rQ(dU3L
z`Q3-im&HT867^yI)_z6jygJs`@mXiycs;D;_w)BK*>}v7%Rli>y=Z*akt6=}K>IY&
zdUc?l^Z|OrL*3~I^ocuu_gbI#dR~=Zb%alT)f2R@pwFe%CB5~~zDu-U)iZm2Ki%ie
zKL3X7b^lhU{DIjY^aFaK{aA*XAN66}I`13#L;s?pK0>ZKR6kICsRRD08}sT(eOiyU
zFVTtgJBQ2W>7iF$qZ7%^aqG<6Z|s%_b!va8OYwj?ZXe~m{V2ckNRNE0KjUTlOWmqZ
z`GnS^6YUS<>KU>RZJ(-F<KhL)^Anx>RXm`1a*V4t^YTGHKK80pddW-LI3!0Nj9X{l
zq~GF#AF=~YK6#1ckk3DQlOL_V)tmZLmx<yiuGT^EwT|7^nOB$i>?3lZb>!eTuKx62
z{;OZK_2Meu=Jj*=6CY?jy^x-?%eZy)l20#WkMC#MpZ2r;MvwT&7c>vqi57qBjc1)%
zm;Upo4An)}9li9S*};A^{p6?*e2|=!Yy9hqo{PLM@?J>2<$X2JO|)~(xu;)<t9+HA
z^^hF2b@FGt4Dr`P_S461=XPT1SKsQ;K1j50a$G;Q@9`zt_t_VW>u*q>(-)xr;GXK9
z4Sg@tyzlY2NBe#d^mlmvZV&o9L4S8?zh}I}^WP2CLz!LrOwRK&$I<Euy`ZSh7Ia<m
zJNM-y=NFY$=P5^>8%L*K_JzK#&egT`iyPU=F7e|BJB>rn9r__Ti`#yAPrc4beMkQI
zh2OcrFXM~qPrgrDLjP+c;5=~ulvjP*{&U`_8|RBUht3uA-q)ewovquQy5ia$3-4>3
z58m_m{=Rbv|LBH>`&aF>OyU1U%X_cBz#jgfJx8F`H^0#4q4=Q56`!29f5Z)K9Xj>e
zXVAR3(U+KZ*@yZQea4~jMEY}_{W&jg&OP$!&p7BC?AAa2wbi~v`l3c4>p}kHSH0S2
zP#ws#^UuBAJ<2^@{irwhaqs2a%dA&-<U?}&?u5RgFV*`VI{k}^&Q*50|D_-NPe0i`
zcF1jm_G`0Bp<mWReL_6y<zJtXXZ@Ak@=LD#i;w<qT%3*TlMo-hX!Rj~;)j-Z=p0dp
z`jI}R{`h5{9nKB*lIMMAj`Kg|u}2;uzr;^H<0lWFb6MQP$9_t^*71itXq-Ij%|msk
zJ|TP2;wj$p!fx{Dh4|$KAHS?uZ{j0A=J7j6)VFw}#asO7%W><}8M)%bUVfXWhyVEL
zQ)m2^C-Ttd@r$><Xn)aXUjJYZzcN0?$xm8*z1O4%l5bxo$`^j|6c6jgMI6~<9a^5m
z72-p)L*9(jhqeylhslSA<as~l|IHu|@*ApOXuJ&N)%OVL<!8ptJUb!z@|^3$xeUqW
zS3RtEu6$YZ;=q2id>Gf)py%|m_ny9gM~;0e&h~Z2UtH^<yx@~pST>*h_4v$}p?VZI
zG=A~&{uGJ}I?ok;*ni~Xmlx~si>G)S=Xav-J?fwM^%wmKk}r?!;SW3QM|MKv(Dy*u
zC2saNy*aO+LUubh*hd~Y5FfeZK;!I&^2>go3*d+54_s@Ni*~r{v#Ry%#P`6t12!AB
zYx9P*)4cqeCzl*@*_GqTkB0JW9ldCA#ILTzNuJR7*bDUw^^A}GklsXfl}JCi>~~*K
zukKgsG?83(v6~&_<KwUSME)eE9P-Ngf39yVO>SfP@nu|+cArug)`^?=iw`<ceC<c5
z50%l@xxdM4%D2De-#X_`>T&Plk9_irzbS_v^310_#v%FQL@zWC$xA!<Ctm!puf@qY
zWOvTnH$LYQU+DRW9DdVZMzhB{deH9I<R_+n^Va9QJVSi!%W-jFr#$h?I6c0{=6OrJ
zJP*@HE`6Sxts|En_R$ZG<Adr)U8q<4AD=qGCw`E9?DAfNKjz89XT5sjw>YueJUhe%
zAG!QRL;4}P^ue+|`gIu>A9>bK#9h4@N6U{ml7mm3$`5|^=3Fx_Kk`Sv_eA98`GRJb
z`-VPf9a^2^XCK4|*{h#I^WugN(gX3M)iL{W9X;e*hab&uw0Z0JDZcWQ`4RW83VKd*
zzwlh?UMrsBEFS8OpXy9J#7Uph&%D?6KF9BI=+8c9vma8fey<O*OFoRtr|*UMT`>KR
zed>e1?#t#OKiTK=Kj)FYoiuyt<rh22rN=xPiZ>b`zs>U-pZMXEKY6lF-QyD%NDe>9
zQ8%7vJ(qe9K`uGI2d&>gcC(8f<MhcVJMo)Wuk3~NL;D%hLmreLevnU&J^{(&m;S78
zu@fJ8`1D=!a$daILl0EP5TEbOkt<K;6Zywp>)45(ef+`CU-M{u&^Y8@&Hm5-&V+yB
zgHN65TlO#etV2U`%<~VjM}LIuvJT(kzUQ!0ywxu>u21B>Uw^QVlFt1{F1|$j1s}{h
zLPPt@K842VL9^F$3;*z$XFu9JyX4b*9(klE`Tcz|yCHe%+`2?{LoX~F&w2G`oZae~
z9gyGR3C$<+M;sFICm))>^r|;#T%ADvL;I4S&^#;~H=oEKeTUu9xH#)`@@l;}@TUyb
zrFr!z9_~f%FKFxVi>rC}2DE#Xb&34o7k~7RtRMAn9$Jq-<r-I)&^jnDX!6Auk_Yjj
ztw%$Akez6JiR#Jy#QSl2*v&pj{uhnxSybewI&v<^v$~fr=cxMuI+0z*#Rr-v2c7fw
zJ3GZuo!US87kkMg-?@*5{I$>J&3bi_^Y%HtXz1Py*(DzTDX1>oxA?oT{)w}FwXkx&
zs_XQZv|s(^_!mXn4>0AjE9d3kez8xDryO+Iejvwsb%mxMpSbwEDUrS}3h9URqal50
zeL+9bZ}kuTHW8nGXT3fJ&7<+zPxO<Qe5qGol_&QK>&bIJG45XC{_Q^F9zw2h{u`&y
zJ%yd*(+A1L&kudw=e%fr#rsb0WvoMcFN4qfr~7&hYq)=8lLqfIe6L2H*(>huvHBup
zKNL4Kzt|x@=F?7m-ed8XKK{uYI#HaVdHI(wenEQpZykHBr{8-T^7v=I%zkpeD!90x
z`<LkZ_lkghr0>`t>bzc`)(`CmeFEBl<k=VY6Vz|@Nwj@Qp8bjs4V`O^Mdzb_=zakC
z1;q#Q2Q7|`#m3tCW%2p{6aC$(zZdd%snFl6LVwrl^9K2GP8s(ez`39v)McJ;{%$ru
zcfm)l^9HIT^3{)ffc5ON4zgbz;&aYH{Ai!2IDgcU&yVnXu7o-6J_+%uJNwi<*zW+U
zPbjYTrFfXP-o9{eAP;TdyVsQ=JO5S49{bhvIP^T8^XwHreBvn1_OWsGDvt7`U+Hu5
zEw17W)gR3Ilppsj^Jw<7i#~Pg-okIlpG5xiOP$ioPy0pO#DiXO;3s?S8~)mFIc^<3
z^3d!OXMLMJ;(_K5JIJ9|KQRyGjUIBuIcf2wmmK=o<z9}?JfO+ppZw6HzR}h}cFUjr
zsxDyqiOzZB>N4l$O&y@|ix)X%ei=_Bhg|U$NB%+mMtsoX=()XK9K{nqdHMsyrytP=
zjYFSj>Z_j9j6?I#Jpc`j>)-T1?<d&FuEwG`C+b7uYJbX~I+p*e8+lcy>PMYerw<rc
z-{e8_>Ll%8zdXu6KI7_$9`_1;%Q(G|UdTUw7`L8YcClZc%zJ9cb>1ggM<09SCy{({
z)!+3)dgW7}CttqYbLiuTI2mW3dLV}$^4Q5P^NHd>E>!30-o8jwuk6vM*(+cCM$?=1
zfo2D}_?;W-TU^asN1yNKhztM8MO)8qeM%hhL-kG{JLH2r>*<kK{AeiNP`vf0%!_f~
zf3S}I^g{7xr#Q-+eG9FlM?ckP(>{FSgvO`OL;i>d`_Pb`(7w}m$%E$6;*!WNe##p?
z;v^pGQ@+%{c<XQK4|+~<-imh_k}p2wJC{6HuwVYgo1gqQ?(b#LkR1H(SKg~a`4@M3
zj6?F+<#SZX4yc~UV~0GF!(M!7`sjz`k|Un-DewFd_j*XLK7eM2akTugBlCw3TF+kb
z&`;RMZgS~?#`z1ahdFNl+JE}4{fTxC8`lq=gU%o0_7mEE(|5_SZ>+aZ%v%TLk=@Wd
zKkQ%giQ<uQHV(_;Wu9Div4h|IfEh=A!<;8qTs(*2gXD|5ycvi1$<fEWPx8F#y_EcV
zZ)H8?r+Cw292&Qdeb6{Q@fCmjq+XqiKcrvX$rn9n`mD=w^Eocw?j7>PANjLRy<0~v
zWG_1FQ#{-wbDm$$ANh37U9#clH{*dv5A4%#-Ja|B?B28KSp#<-*t6OH{@br(zh(M$
z>DlCrf&EtaUwXZsO?x)oan1o}4(vDh?5#&tnm3s^YU1rr?EX%*_Zk~L^T@+vtC#M5
zUhDnmkE+hu=)k>pU1xZWUUP*WlMcS>^=j8Ae)-zNhrC_g`o<0Wz4Xq+>a8!#e*TCb
zyj9KlW7;-9ec;nA8s={Q%VS@><R7*99eZxR$tOb|s;#pglH2~QQQzFQ>ATfK?&|u=
zb+^7y!)N!N*z>j1KderE|CY0l+-FL)!;K@?>b>vC>U~cfc4L<rZ&t`zP@FNY*ER2)
z@oc3Gk3Q?7JJ)aXQRToM?^GvzZ${;?U2eF1(5CNHuGwnYzn%A!w=2i3zVyJ;uYa(@
z{^!q{@vRF-JXP)9{=vsO?*4k^-OE3^?}E$5R(|)yGJUQ)>K~QKeg1su_3uxsRFB)_
z&|6=gS;O=H@b_)*ee|=+iTCbw*0lbgRYpv>e&Yu^&8tiwbm8rrKQOQIXy@*~nK5*3
z<)HuC@tV_4Y+9WE!~3pY@6r~<)lc5_%Hzj1E6)3Pi<O$M-K_Z5DxZ%3?LG4=hn)ZS
zb^bEBd9lKePWk4;`^>L&n=$OBizke$+5OZSAMJ9(V-qWvZ?N;kF+D!1%xbsS$ydJe
zUd_K-S`EALnoB>az%l1`e`DOm(<<#x8up``I)6}Ux<%KY_x-{2%5kUlJ!0>xrdHIC
zx`HphG^FbfubE!Gve(bvoAu{u)nx`<)@)j<snrKQJaNEt+fJzdaooTa=AHaf<(d=v
zE;Vwu+bd^$xO%b2<1f|fO+K@p<o}`N&*^i<pvP);weEp?|K;@`-dVl=kYByF<8v2P
z$RBdYseica%y(<~l5crSlt*}0vyOX=-0*`c`^s>Gji0>cshi%d!gDV;VbAl&JW^5r
z<ZUtfnM2<{{@&`inSG~R+2OhB4JUT|>!VAJuldbB{>|yyp--n}|5V$b<U{*5?Gg`q
zT6TMD!<lctSgFbTT&{SkZ*>h<x}xWt=BtkUKl_9n@)OygsQw_miRyOF8a*#vaoaJq
z{bXOM$30s2?QqnSb1LUf9lqH?ug$2~2kO&48h^ytQKPnhsxqSI^6gq3)x22mlztBm
z>(a8={_WAXuCsmf;+^WsQ;+XEuhQz!HV4mcH@~*;>@)l7+TJ(4-Jy5u!oILCU+i^T
zxBlzRs_nNfxBPyU!;hO?+t>WP{hI1slTMvktH+^lR`%QXgHLMqlZ$_+*Z#ceExkXi
z;Im)#A@waD=;t<nq|=o9C)euHK0WHfYcG0w#(UMCKb^GWvm-vMZg=mrtxuacqq^aN
z_uTiZ-+Wv(?|e{au=?4S`}}r?>D6^xw|;v5)-!7Rdarf+EOp+xbE{)IZT#F_?WWY~
zS$+Lt>e16a{rM-gb3tFf<=hwlbXcd!wZ1M->g)PlnvH4u@TiJ9QrG5<C)!u)7wWT6
zJ(Z#T`$ChWJ3qJdV-=`x$paKOsDCB0rwr4dq-**=@1F}wm-%a4zROUa%FqkEdZ@q3
zCwlD7-yZeZC(~>7sK4nm>e+d-!qaE2c+RVHs`5pS`f)DkgYwg2_my@!`HD|#_a5hy
z`;9tvep$y}eciaeDQ@Cve>&InJM{$R(Y*mq{p|AVKYC$S1$H{_idLg<oLoCMog0bb
zzs2!KU)b)H&#HG;|9V@?v9qds{NdpF|FzZ3+CIK*ou{sQXy&x)i2k>%d)Dt~*Y@?$
zK0n^|yL~5DZ@%-h5x;o-gW7qnzle)<;wgU{_c-gf$J{=(mdEK`FB{zEyK}0E@|7sx
z;w|su>K-Ug#^qTaoyU1zEWhWm$Bo~ySut%^^8shA+@e_b^!cN{`^@Z0kI&}+V3q5p
zRKCB?*{AH;Yf3Hdx!$=0x4U?sZ#H?QNdegh@s*)EPjt>C-o5GSSG{mg>jFCW;KtQc
z&be`uPiyzN2gmonZi8QZTp7GyrOB~f=GMmD2i-Tf9rWq82b}Y1<rn??H(&3qPiy|G
zKj*Bt>$hcSUqI&@bbdnmp?=NIt@>?x_~5lCR5tkb{f~U>!4GQ49(8a;zkx%Rx#^9H
za}mAo&>QE@n_R7(d+Lt;*$3=r{Y71=NBdTv&iv{B_5(d-Xy2wBeIoC-{B=(-Z{M(s
z-Q?<@_&vY$y?yYr#~v`G_WU9b`kg$ekE?#O$<MBC-lQ0`@u}as{f>_+?o*zh)RjIc
z&#zy&XV<&teO!T_TgI<>-cwIL*}O<Ram6S0+WN(&h5CKG!=T+SXx*l^f9+!^j`aTI
z%m>$*v%!d3{oCKIhM!PucW8@3{kVtmQ~jxb``dnWucTKW7Jv6Pbt^9T$YT$5uW)aK
zKY!`inQu&=Qo9!#x1Z4VDWvy~tJb?{z%$K?wI{TB{@ro^e4cyk(Uo65=;D^O=STcd
z9~-=E_dWkHsYQ{f4xxSt<pZj_dg#1MB*$|pf8@=65NCSi#d+*KhkJrL_Z*_G?LU0>
zy}poWzrlSsd#J^AV>=c<oczkEuk5r`(dOVIdS1L++hWJ3cGz;z4NDa(+%)C0HjlL^
zW?Ve`{1MIC{vUjL?~bb;d1$BFzEFSmr~KO=_MPV#^zH|Z9N(~B)7rkV&)zw#&5j>z
z(W-#%Kb{NpFXQ^+{K0GgWTjmuSKRmYo1Mnrv(olId#@_qzq$16-#+rz^_9*K|82`R
zhJRGmXXK%{ut%?796h3X`t#rDQS3La>fGM_!EVz>-10>2xyd;qpYlt;zUKKD+UNGG
zeSPJy?WT9@KC=StKbUogX7>R{t^VerAG}dzSKd4AW2oPip}J1Y`YNOC)5fCr67Efq
zKT!X&zx6fehI_4f?^m2R>H@moq16R6Za+FN)syFA`S(0!U!v88b67olj`g03z3N(h
zx#zm~s(<$c<LqV+JJg?iS<fH+!M&M1_}GPy9QPXc5a*NUWJo_05Axl|`E8%m2idD%
z(n}9May;LA{_vd8<+~RwUHSOa>Iv8M>3{s(Pa2KnTDQi`AAP*jmJ_SqAOCvSZ;ajb
z50h%|A-s2xclPTK?tT1#>g$x3fBL%)&(E!rYrT97+kVdXeznXi)h&<QcIPdZ8(Vcw
z*cbML^AnDHrSsO;?%cG{kDS-;P0;&c@00L9di3C4KU%$c?LDw_&pD{Cm7#N>44u~t
ziuQ~9iTZSoLh;cjFKl}8vMY}H=lgZz&RJ+3Ii6=m{_4<{UBB~QMIT~Ei<dk1zGJP~
zl{>e%Z|Eim&aIH^y#+q!jB`xiq=!GwSL^H}eMbKI&2F^5sGpww%h$WFK4og<v@2Q;
zn7GypmB&A5vQ@{aZ&&nR&nxsgcld2w{=d;{na3}mJ-IS`-sNY!J+Dc@Uv;Ri`**u^
z=v(g(tL$*Y{om+v!T1XM#gQF8uOOfOJ|DqH4!^|3y?gH&>wehkkXI|tf9O0BfAZFu
zJhnQx>5Wx=U*B@x>VHt574I7T`TIlusULBZXY<z4tFJk~%&Q0eLcXlSZ#?DbWBAk~
z8ozVOyn1wx$a6+rdLJd<DHk6U-`f55^SFtxI+EAg_~)OupcCzD>(mcaFY?CkMEx4-
zx6T_hJ?<6kp=aXruRpu)L6d6lljWz!3&pnQZuGa>J@uEx?MwgZ@TY5^L%NqbX9iXN
z*lvfx?^m6lz5n{ynoF&)Z|(CO@pSH3C+~^!)mU^6xpz1R*+ag#ibK{XK6N5~^-vxb
z72R*;0ov#4)_X<wFz0sqEAC||5BWKUdz;TM#Mij<-hI#g0}a{D&U&biB#OWIq#XAu
znDg#EF!RJe$Zn{w!K_Pl1Njf>g=wF-LUus=2I>d;w&%gdqJ83f1XEADZs1>k{Atza
z29FNf=%6Pznq3_{VcRJ^&YM+xPp<BoR@dBi!+WOG#@&xSrw-WmcN;FXNt1@9?|$UM
ziTgBZP+!)o7wg0Y4bSWR;zk2DpH)TkUtGmoJ*pG@YxnO|9kR#N+I`qM_j~vD33t5t
zV*8~=S6<%g@U~5#f4TN~x%-B`DqcR%7f1ErxzYK_9^d0&FSKvvnLM9E@?XAlAB!8k
z&MWs8_Z;V%ey5+SZ|Ask0lz-z{=pw{vya@L^eguk_dn;Jdylw!UiDt8vFQB{^c(@5
zr|vWEhwg3qi~FWHU9#%zkDorfNrUGy`(oR!^ETb2*(bH{-KYa7|4((ByYV)6kFTA3
z&LMTHj^uyJw?{S{ea@KL_mAu!eMY=@X?Ay){qB6V_Pq;vRA<fw_Y&`Q-GALXy<c&!
zcTbgX_fPXaHxWOy_nf{T;C}6M9Pg9Ndv5djlIJ(}$XC~gwx6-;N$*$eN6!hhI{p0n
zM(Rtw)b4$s&%4*ySMDp~Ef4ijUs3Ps;N$kAp4g+?N44i_{m^--Kj?Sfs~o)FxQClv
zG_#s$AG_bu3%$RD-cO=^4&d`T`@s7;_Z{^C<y#!(*S*^N6X%Tg=K6&Gpnj}VKlVQ)
zZ`0k6y=2O2b86pX;fMX{^9Ju_J(u&_eP18h|K<;eT($kwiaN9J$@O`l_qg6Kkl%ax
zb^fu&v;X}34DG#xdQ#Uuml5a0P2W1{%!WZTs?d4we&QVR9>x1oeaHEvzZiEus0-(Z
zd&BEj95lDrM-yt_D^uU{<ozLja?m~p?>B7Ho%jFuC)E?T*!}p&-<?_Yc{u&fSN5rY
zeD()9>YYA)(f!lC)p<=G|IF)~o^R}L=aA3k*(u)S;1_>=Tc3rioxO3dTkdJn;JqWe
z>~H<nearc-Z$tOhFB+Zm^6Pn+UVfU_KcR7bho9n+s6L?kh&Z@E&?}C{^=EebTv5NU
zU-c31shnGo9PhQ{SD#DN_aHs&a9`0+_*D<RH(`(X^Ot;mlYZmwakcn;er}0}`v_!T
zqI20j%KJ_I+4-uk*F*a2p*+KS@ydDqD6yX2oELZKocBFR=Y#vFy3^0S&;8`q-_P6k
z#-_!4Lw`4-^Brx9ZY%Bc#B%f6);<r^@7;U6ul70O=DUqK=9~jt7N_lS*O8;HX;&=O
zvfplRZPlUp>Au@+H}<@a#jB?@|HhTuEM2&-`drg`=eK**$#*|=z<Cd}D&*BV=cjYh
zxmY&td&_w*V~=&>;C$b7=!AnRr!^_a=ePH*K4)bo`+Uxvc<@`JpWblg&uZWM_B|D!
zFY?>>CdhT(TJL`I{E+)rU$dB25eIqoy@8W{T^V;kj~0b{hWiA6o&W9&_{b%PKJ_6U
z<oUk2`tUshadf}npM3KNpZxPz9LP8R&F_D>;h(Q<R{K7c&p+I2+_&9(e4c*KfXmmM
z{>`OopRc@h>4g`3`~8;1^|P;@wEK4LYu{T~arz;re`k}{wa?jKI=szRzaG)KIQ_Eq
zXU<)&L-F8Qm-fDE<xYkCxL0j;&a>yPu|>yPz3g%SP6xhsW%EKj><f0-U-ro=Z@jh4
z;EOsGFTdM<`JMJ`Q(Sl3TKjMGa@%6UZYR9C*PoWI?L*%unbm#DWvfnFx@LdwGj^$C
zwC_2K+pa4-+2r{4%hvX5-)kS;blDF&6+Tb$_XUR?v&DTY?AWgMy|Vm1g6}Kref8tB
zcAK?St&fy_uR&h^cGDmFzW#Ch+V>yC<)XFL+u*Wk?P_v-PvWF4@B8Dm8+0nv<M!YE
zaHqSEoKqRQ&f4#HKXF!N-uR<-9&%vkS{%hyow?822Y)>Kdu#V<-@1VEBwq5~V!-@|
z)|}n67AJClwbOP-fA5Soh0oc1zubM_`P6!~Pyg0`T&rTANn4$AV*jNJ-*?hiJQset
z?I~?<JHJ)2!wJLhKm2=bYTrk{aOAE3eblLK3w^f#>)YJ&Y~NP3I<(Gy=8x~Et2^~|
z#AEmVV9Q7USwHGm9(}K5rL8Bl-u33Tg>w}ibNpu0UaE91R{rCD=ly22_QlM%k6CW!
zvF!?ZuwU(K`$2!PukF*Hbl>g!J^$LWwh#4*b>BVW_LI9eDLmh(OXK>J_ZZ^u+}HO{
z=(^{L|9!@ciazZgpbvV^_WbRAj(dvFx7=r(OYS53ulFkKQlIS8H+>INJj6kNl}EHb
zZ$GOaaoenI-^qQaH!INkq;uN7x8DBu9)jJz=Za4sR^RH`-zBkc$nY*Vt$bzsf?jz?
zyU#iY+)u<+-2Gh?|EzPKvO_!|JNYBd%XGe|^E!W=Ua`N-i<5H?AN}-_1HDher!Li(
z`-XY<3U#P1JTL1<kbUY#9XLn%tIoX-(1)Bi?ib{`XYkWLW{2}k9oiS{fa0Zp+Q<C;
zANKA&*zT&j1NeO<Ave4(Ad#Rl1Y?58L`0)d%lQdWArK5cA{c6dla7F(v>2FRs}{oG
zqeKM?)S6gmeG-C2p@@WYh!G5r0n&uQ_#hY+i>NS!5n>+wqxW-|`OfV5os$PTEq|@~
z{eJtg_IiBR+H3D~uYUZKeAqkO_WK#_tlN0v9R5>o?9ROOFApI%>(t(OzO!?BB3F3?
zs(zVuEOmT!E7X3nMD06K`$`n<*w5OpC$+C;U+fefsVk~;!8!Xj`x5&KbuoOX+o1SS
zkHDvWpZz4NE}`yWKKnlVWaqQ$7V0X-#jpLKx-a~-uh-6gvs3E)z<kz=Z}kBDs>9HS
zcKB9jvQGPT?eOWl_TF2Mez{Ln|Dg~4jpuyBd-_nXF`ju)df*RmS3HF)=JlRlt)Km|
zqonMXJsOAoiqqmF+*1e9PyArt?q9h-!%nQry)*SY)Ohqu9>(|H`8oMnH@($Sa<Lxs
zI$vV1=0nxr)qCKPI+pMBjeqv0{zpI7?OYO0@oUt5H}l#5@)Q0CH^_yY@F8!4NBXf>
z{m4^)@*pR59plh1{U<)CcfmL9ohv%;m9Lvmol1Z6sn6k89nAN&^Vd9j-^0Bv@^Mb(
z9E3ls8ySy%n%_FLV^{R@(_2q}?(`GaHt}g(R2&kY;Fs~})BN;@>aTuee&=fDp)dGC
z9^|PVIlv9;q<?nnJ-kRe@Lv9CJoPX5V1DbvKR%uJuowPFkL-p1#W8hm_teQp97OrK
zeAPV0mq*FBP<d8T^%&GRN%37r?SIGx&X6}d(9XV&e9h~boaK}Jj9+@cqbdIDD1PfG
zzeY>#`JFfhXXOJZyhh~}sPhzfEdSxR{G+Xx;rn%-`Ds%AU83||qVk(<rR0IaEtLF9
zv}C{RUmjngaHmAsb&0ay5{08BS|>ODQlg#uNy$Ftd8pqv!ZY;?_YCAYzN^o=CoMk{
zw^8Q^?jyLT!QS2LcP~W#!4B9hJK)FG1$TWf*{k>BQP*_1&Lj@$FCK|&>O%0}eJS~y
z^|;649N2lU^J#gi`xxv+Jd?ko)<-Vvl3v6Y?c^Wg9X-Ky=lk~E>K*LS{sRAgw}n5r
zLO$Z1IBHyYmee@-OKScSbw7-rN>tv7PxUQwBriB6PcmNGDZcBde56xKp5}!s;)w6~
zL*0uIN1W5Uci{I!@?HKY9~EcZdvM>&y$$y=__Mg>*}R_RXP(_-AV=TXL*AR_2mIdt
z#{NZ|q&Io6ajjRKhJW%G_wV2r{F86PpSB+Uw&%wO?f-a#--}D~Pq<|Ng<tPW{rIPQ
zitxq$+B5vdzdRZxKlTW}-D{ISsBe&mXa42BIz5rMb&<1s@o-7LYkv62pNu0ez)kCc
zH*k@C@^AX2_q0EDf|3il;nO@QJ>x?=zdK31WH<0rdvb;+^u|uK&-us!E|I_A!HC!L
zQ1_(x6MN?`?7;mW^7oxTiod=a7rv06c+Gyu7r*LD+QT{Pg6H-}+OrSi8js)L!~FQP
z4sldG(vMwuW{>I*`lF+c`^n_J-tzMXJ=@>d7s`*^r%}gNhwxn;T)j%YSp2ik^Sdee
zs`l~)ex(k;F5!|mKprR@B1iTj?<IG7ft&0h{axJAPMjf6xCIwb{%c>7ls%x<`KPA#
zZ{+M8L_OVkqx&B6@-HpB_~awT4Yej5e8<iEO&w}^{+nszrape*P-{xho6cQ()llp9
z51sqqt#>^CeEk)h_Sty$P%EkS4;=mYr5}6AUM+O#@^{~N)|Gp-PX1iq0pECfxAyy!
zmri(MR%5T$SMLA#{MUT^c`6@-H{y_IehRPHzkE=><eZ3o(WiWsU9l5>=NZnr2Mo95
zlW>lmJ6B_m{L|khbAR$Z&y3mU{bNSA;46DHKRoi@I*rR-;g5dWTPHt)S8&{W?a5a^
ze33IMpD>R6L!C;U2>$pxD(WlFL*?1d^WX~{Qn!ag_RG!>ozHm>r=0&9$GGlIpzeR;
zS3BouN%4)g+YjG=HXo{9gu+V{zo`AYa|$%;XZh?r+`9NTxv0n47sCPj8}dcT)pyk2
zPif=H@Zaxrzrj2xIXL$~Gv32*`L}aKaZr2^PuV@3cJEmJul~<Z<!$1vJkh-TQGSLG
z<BC_t!Jl^WIq?O*{=OK$w2$V8?g=>0SD)}+oJQG+e2PEGlj+U5FMPL8bbc&e%Lmw_
ze$K7z!^L0uf;a<*$dg{=De^P=M#)S5LSFQ1T<yu(dwMbtN-o;52m3SgunY6a8`&Ga
zKYRX<4nE?f4NW+2KT5C0_1@o^#IHP-9oeU{H_!OBUUEVC6Ff!XwY-$Ryhp{2r0~Xl
z9(YE-D0%xm06E!5rQYq2G7nTI;t%p(d82r&KfC5H@;USIH}d7r^h-bVEZ&)y-aWG;
z>yjrLpPu-!yn$TlK^|y5@`L1Ge!)(h$KXSpmG6kt`uXm?@vT!H*zP}kFWs~L;<$b#
zYW?&N7u8izb}PQh53El+`)ztqccg#zgR%qiQKwb+VDH^hc8)LexqmJXW4H8*Z@5Ap
z_O0wuJ9RAhK|b)qvpAJ@%>VK|`z9Crr<9}hkt2OrmwF0&;(yj<AIomkCG7W6cE^s4
zYd(5M@tJ;25A@03;Jo#k*Ld{74*8$`8u^On)`MUACMVz78G9xl@zJw9SiWT+^OfQH
zoA@rjh8xDWPZ8hM)$p%=1E<t8tP5_D5Bbq6`yfyK)c3SUtxvpAFW`Ud2QJw!@n`vm
zI;MLXsD5x0|K_I;cn-hh$M6fKFZ(X>#dsxZJXCzKuS3bNMAe_I-@1~juaOu1+Apv_
z<In^C<ooW$+Q%d%U-rT;=#3x1De=gAc{Y5K56H8{1MSs2eK()HO5WqW{U-Y`j`p7U
zHGRsL*gyHgA9^N7&*HUy+S^CNJNgy(%#UhE9{Bg39UBMV_z|z^f!)F}{Gi4sU;7Xg
zK9dW7DN%8+TMEzAi`k`m3;Dt~xW<1x!$Wup@7M#J)z34^uGt5<^ILMGucYb^aNGRw
z%{sleF8f;4y6uzY)B5vAeyct~KGw(o)mh+xyq=udiT)`3BTw>0$%h@f|73o+ikdIa
z2gDKI*(-gJKf7d)+QB>R;5&bzH~zs-`M3G)@5sw}hj`D=<Z;F|4k|C=hxjg0`I~b{
zc4c2{T`x3+r{tZuj6e9n&*TB{!+Pv{$jLdI{DZvNqrA+x-jgrtJzSLM@nd-)D(;9U
z@@jdr{Tus{_oz$QufcI~kUhxD*$er}^F5oNT*-;t@Gl?dXU6p`p7THQl?U=a@*y|x
z;SE1v-~3-Z<5%gYzH2WovVZv?eUlTva{eM;=GXimp5*-efu7)<aqxv-c&R-c+&uC7
z?|a85`&!0_f8@gc>`%y1zC~Z=Ns1r(FrIaa7v_gQ<O#RcN!78%0rfmooKRo0Zgn{E
zL>}!~J8?%{$nQvuk6-6@epll6OzPZzSK{|1e!t@PQR=+j`@M^JDc{DIxaqs!TjAH=
z6(K+SC3X$3_z(LhUwUJ|_9x`b|Jkp7l(=s^<MT`T5S)>R!W(r#_^+PlykCCn{)zl6
zaRVOtyBz9=;*R~kc!O$JM>8+Y{1N|2@ej|El7si~73E*>o8QB^r2LJ)%2(ht{DTAX
z0{a#7$Zznczj%NTe5=#w?-`DXZ}zF|35A#VLXF3+*eUtLSMgO`vF}3hOD^iCp5di<
zDo&Y~KGav>Ih+#5)lcmI;gEQ)ek0C`d&Y-@<RPAVhQIt!e}0c2`HOYpo4n+E#>Kz!
zQF*%c!V_|l|F9GOCNCAI*#jK%Y<=udo?ssGV2||4f911qh27}Ke#nFUiu2?{@B9@m
z@q2c{-^@b};+psD-go1Z7e4TXvQzEZB|XtMxw3ct$yeUQPw-=2)O^NKZ-KMsMd`u#
zC>%iHep3Ex-&v&b|ElCAb(Fm8D1SkdPw!FwhRU}})ZbMRzv;(*((i(uGs?&KDL<x9
zc4uAkUFVMSF?kDr7N5my^-|;G(|6Q9l^x1M<t6!@-NSh}$8PusJ7m9b5@jE#_NZ~t
zlKj1g14;Qk+-1k`o4omn_<|4Ay2WXEEo%Lq^~*TyJAc&AvwfZS<cEL!q2{qp?e#aG
za|AT~!1~(q8UBB;;)8im|KHF3V|f}omS=e{?+_2<d2kE<@*nX9{=;ebFTRNj+L;%A
znvcBTmh)!M_%2cM!WVwKrS1LJ@ZZyBzw&c=ggBfyC{CmB#54cp_i#Zw&+y4U!~CA@
zH}I<;{>)>aBoE-v;)l2<U-ey{Wc}<E4#8tMsGr}T;?w;2wEu&X+TjDA)<G`(g<Q;M
zzhk}P7<@Iqd5o_eO1_>^b|vqIx9UssCEw*U{E#2O2RKrq<lQZ0hvb05o1I0?%bwYT
z`Nco`S^9!E`dcr&xBrx<!993VqV)A5{a*?2SNRf3&g?a*IHx^*v1^pR*|YgjxRO*I
zqFc(}P<cTeg<t5|n;*V&`URWYe=pg3jDvsisf`=M`LMbQ`JiyXzLnjIoAy)uQ5<w0
zDUKS)K1}{#T>C~ir0!zBAkNCqtVi7${@ef3KR@KBa1ai|arrbnR_F4a|HFCj#VPoW
zlCL-kw|%EKd5}2Ie#qIn$k(&+jca}S!)f{;XZZ}dqU3E}@3oij(3AbF{6*eH{^Y1`
zO5Wmp%G<Mj2OJeQ*_r&G+{A79HM?M6&N1CL_j@Pt5nt@zc>LSE@&|d6`VzTXx4HoS
z<nQ(g@<RKuZKZjygPz0@_=P`qL(lZ<JG+5H_~&o%Kz>fnsQL7_Ui`2Ra%Hdl0fk%q
z7S710*cJO%M|R(vztJ12J$pp)LBIAZ^enDezjYaxow74_#7^Z`&Sm5c_NDleZ@5<^
z?_{rVhThySQ-7vMI6~gu(|c05%8%`n)NSmSzP90<C7=CiPs{ke!vXbI?Y*bBPO13I
zuEa0>@z2ka%B$h4@yG#wqc7%FJ68hm+rHO%lXD<`PQP#f?ufgd#Rq-^$Ke7z=P%+k
zKHSrGPaodvuOGWncS7L{`@|o=hHvtWokj7DT3@%6zsN68I0Ik!H~;r6|B_GZkIE<N
zsNXv{k9GfszJ2$*E$62G&Y$n}>O9vusdH8D-J@rRo{i&o8L0PupMl~Bzv6^^#&`C?
z9@vfiOuU8L;wBsuhu9r?(v$Y=R{Sy_J@Y?t+_U*hv<@Z2QT`{M@fUsycf~jU=A7Ag
z@+3FovWF6dA4&1oEyYi#)Vh<((>tZ`s8dRh+e+Ol;6Lm)DZO<|x0SE#CvPiJ=M+i#
z?~bN*ej)Ep%J1u_yak@BBg<FiL-uF(-S$KBQ+5weJZmRUb^gPC?KAi(KTv;k&gA!u
z?8x8U<@fA}-LPx8D=y39*&&=Yk9w=$LHeB~y|N4Qpzgz%$GsTi>+koN)`4&9w@&y&
z4&ttLiDTlkJV)LF$Kf)0vs?X*Z{4UkhVK%tkehXA2UpX6?Q7&??85rsaN?_H`y=%i
zc5C0O4y4XwpDu2Ck3Vt~FTKYnJI1H=<+}dV&-O7TO8?tS*=5qi(R^lSCCXk)lwBv~
z|LnSs(l>jOzhqvbPJr?=lwYC7MRPoM)hVUNZtV&lzt9wJ>?~@3CNGt*c1q>3@?ZN5
z@zFj%>rC_@uE7!ctvIP(=NusQ&yM&X`%}k}PaD^Kp8d{3UB@|*c&R>w4}8Jd<V(GX
zy}HlqJ^XY3RXgz;E~-b-gZoD0WPbhNto7kbJNB-=<DQk@U#n-zE8si3pl9>@4p;aq
zeUq>CYe!$&Iag3`wI5<1+PNQLe({K3*zdWI0<YvH@&Iv0{PlNDd{+nXY(E2+<u~#b
zd53xELA%TcjOXuc_<dPY`3Sssjtr;m&)^<ClN&r`$Ls=r85f^^ZzIq0-T3qi*NuZZ
zUyv{GcYcL$`&|AY{+mY~RXp=|E#yt)Y&`2?7yR8g^h6HsH}MbtMeq3Fm*fav;H7!^
zGrjAN`uh*YBR}@VuGq8s40*y)<MS(WW6#!ST=~EIzHo)yJfq~uj?~Sxw@x_WyYmzB
zQTMmMfy3$q^ud1RrSepEufD^-{T&<jATM^_M~)~zQ^&zKyiwP5-sb#+|F~CbU#6XL
z#36pk4{|-^2LF8bcT?qA?45nM&u=~O3UwX__uw5rAwTwETy}`UW9u(bae{qn=lnsO
zAQ$llWf$(B=+91^U%-F!;m>}A9{4MJWykaj51e<vFXv7uJi<S}vo81dM?SsBsVgt~
zW7Gc|b)U`sHgVN`5zi>OIq!3BXCCq}zj#X?>{WZ;jf={|Jb$aX?D`RpKG*afm8YS`
z)eprF9MWHXLqF}T-~DgvMdyv~zwnB?2U_}hZ+-M^KF`{FrdRFQ7kOz%uJq3yQF`_N
z9_Tyu>^*(TpV+&7b5d~tj-vR3qwJR*!b9<fotG#(7AN2_ToTXWuklfNGhEgmzs_Ue
zm+|11xC3|j7o64~PO0zcXFm8ZZ-R6B8&7}F`oly0jgKF=2v_wpj&^?k?{`^n)&Hv!
z*X2vb6A#GQcqlt#_x#WAl=zqU;~7r+KNRerAF_Lt-@_U9zq2U+>y)xP@@H4<RooXJ
z*%f;ezvzu!(G&Y{e;pNv$<;na{v)o)Yvcp`od5Al^<g+8KJjDvl}C_2y@^xymBz&f
zf8?KV0>ARpjPJhFhk4-6H5dNnrq9jW(uBL#X}<+ee7An<)Gn#~bj52AnK$(EtxfV^
z@77H&_6f!@Pf~b|vVVGn$L!vH2J@I-oDxs04`1RcoVQ+b<4@Y7#@|-Tzftz%-ne<$
z5k8&Y!VCRT_G!ElWhY7HU+fmXkR$!EFY}`0OHWDhV?FxOb5i>@_G(^n*LVKIpTuGN
zCeJ9pLe+iv760=dPQfj4#y(74#y(X4q~bL^EYXyMXZVPUGtOtkCGi^7zgud*40p&w
zJ&U~UkJ*R)zWoUOu}^Y?AMjm&_D5cD(Rkj|pYiZXU-<LhIP@p(h%=~tr#OT^_M~no
zJ{Vu!j(pS`$xS_i{^%cmqt0jK73|JB=|!Fdzl{sW=mFp4r(VE5jbk5bUi)cwMqc)%
zp7}lhgd6k?PvJb=<o|HT{u$rwS>A;Y`jrpJ|E%A4=Z0DL^WHk-iKx6n-Y4IH5Bv)4
z)~{Y{|9@Ko;vV}~&zARy-{PQl;(O*9+PTMNe7Gv_f}i3#9EAtoi^J--&O7+0_V)Ge
zvC5;|Z*cES{l@PW)c56M@Ibu=|M<pF-g7lCyvHYb!a?KX+j!=Ir`q|h?jgS<KX?Vt
z<QK++Tjr++dV`1bChx!>d=j6?-8i0or#JBd#SebXo76tmy$JUv<PC{;-mAa+4rk#Z
z{DjNmahw0)-!t0hix>9!;;a1|Do*;p$$l?wKJ{UIq2dH;zB=mMO8lWO@tYp)6YMwq
z9eZ^!c{|F^_@O+4-ca!nB|lVrTr=w0!&aTMwW$u~y?Ic1vHVJYpuU63Kb)h{lmG82
z4?x)?Dt|!91(k20=0oKjsQ8VtCluf8!r$kSA35Jb;RY&yLgj;~b)soE;yo%KC{gWD
z_Kxy1lpmqifyzgc!Z-LL&cG>l!4BC^;s%_7gW7q9pX}6kxaV1WlpVuU^$36GN1X=$
z>NNIw>ZaDouI1tOUE&LSU}x+HzKVPHZSrAxyLs^8-X6bp&Z*AiS-#3X*}Z)QJHVH9
zpw__;-JfQU;*{^+^FLHvN=iTe&WAdbc%lA)s*8)0{F?s#PS(8ii_$+kWN+Tn2mQlG
zd}lv;q#tsF+x%QU1-FgEe%T`&<Nwydeq^cQUZ+&t;OD6Na*6u?9{jyT#edW~QS#@P
zs5*{z^ej(fA1GWSPja)K5+zS^*U#^^?3>{cy@*@%OdtG3z1@8o_NyKIkYDj%_9f1H
z7H8NY%C7KZJ*YZ1e|LYAzl&$`7IIW~@H<ES>CwF8gF1iYxAJWJ0sogzUQXZg75J6)
zM0h5jhM)X`zU1BZ<@^)w@jrQnI*5A?{92x&A3f8T@r~y`f&802@qcwEa+42v&ksse
zzKK7653lK&y%|TGWH0ax#gF#h^CSKRue2Ap#cTTa-81SQE<aL-LfIv1eZE_t^^-Gu
zm2c6T_xu1~{Exo)mGv2iUFLl31z+alN9u9rfd}>>{D@t#BlzU+E{dP_r}#nP2g)z(
zZ;h)zf0F<6Oa9{i1DrKKyhGu#I7^@GB%k@a{_KYS__hA1ct%hB6;8n?eo4;kMt}LK
zd`6s-k2sh0e{lTYFyqU!<X`N>{?5Hb^Rst$<M(9tk@9qW+5fE{f9YL!-q~NR{^hRw
zo-}KhYTBVQUw8bn5pC-4cRK&&C*m;vUv=TuhhKZo#)fsd&m)ez7wGSQy6@-z2Yz|)
zLudW>Zw9Jg-FDHcp&NFo21lGe^JhoxUXA?5Q@{O(Q}(P@%o=*+rjHF)H!PZT*@Aoe
zDtuWdxftL7ow7df?endJT==7W)%xU-<jCLbN65waXaD$^>+YR9viikGcA5Ufkt3^t
z1zT?&zp=McCm<jC_Io~Z^#8T!$Nmi;{8`?szKehD*%AA2FHD`CJ(3%Fq28+-(4+6_
z3+x@0zt&OrOyGh0ByfN~%Dd&$;+J?07t{&RFaP>&Pffk_xu*IUoPrb1N9F5qI;r{n
zT@n<(?rFgr^*;HCIsyLi2j}E{a8uqO-!YGM!ZY&q4Bzy}xBC^^Td(+u;t$TNW730m
zaNNAwc^1D>aY8;I&g0WFT$XnxB`5174|Xozpz;p+n)U1F|0A*^c_RB$x1v|};=O12
zq_{+0@=(vem^yy_+kdjLi9hT1ERUo&-|0nu<sJb$b}xfm*b%<?mwFEWF|K{R_u?{n
z(IdOC9{NPtfqvFOE~vV`aoH7rvCm>R##`{Viw+;0($lg|{zD$O-1FG!Cmhz>lFzeK
z@sK~+4>?DWpID#y<k9$)Ps<zRQR0F1(<i$kFZhV6pVA9lmmhd9pYz^&^cPRWb#X#I
zV!y>t#69_7*2#?H?|z#{9wQ%?kD%hP{l9ro>w|yf!q3UqvvKXSe1{Y8fS-7VEB5c^
z<1fase=?tb?90B4y{NOH=0nK^<xks7Gah+nN9KpW@SXj`DfZkgC2#p9dmwi>OJDd&
zTw=%S3-V{wJoZn{<KdJ3_^=PMkL54&Vs%g7;iqx!x8M(d*3bS;JL{ug_M+};eD=kz
z*o$`dan@;E>tpZY0o-Db-W!h`$Q`c1v5W)o8cxDJ<B6Br!7=y^=ZtGUcyGT%pZG%k
zUckAg@$~1n<YE6KZzV4{$R5}w{m>ix%Ke=68V??@5Bg*0?23NrBkLF93CjPiPhH)(
z<jjBhjc4<)Q+CHr#2a`je&ZAFnGe2s#y`EVH~KO^KR3U+Aw02OeB;AB-s6)TjEfKQ
zkhg1Z9P2VZdoe%C{!!z~Kb)f**L!v?PxQ=B<Wtto{^<=>*C2oS3^}Vy(FaQI{E*$q
zL(ESu>;gaTLyH6ApK}U%mOR(^<Rq`+@8n?p)@R-9LtJ7X<O%QK7Jr4a{8E0x&-okw
zClB^XUhD_X^ILw+PT7rqp7{~`!%w$VUI>3sa>XAx(K9~819-u1^e0#E*)zSfFYVa{
z{cBHN{1M;UcT45#>;r}C?#HkPe#`Fg!=8=DK8+)vLH#{W<FZ%d@^A5r+}R(y6`#d#
z`3e2WpVfujyAbD$OOEO->ap^3^(1v9`9(W!o%L5gnttD94R%N0>VV>%_Tr~?qx{k{
z`TISEeJT944<uLmhl}i#U3yOHoRfW`<PGmod5d+CC!F(LUgvvKd2rHf_u}(sK5+4s
z%bPFC;s3WB#0T|I@d5twe|Z3Wfa~x@yb)i;dAMoc1`pvHzvlPZPP~_ot9z(Nt0Snt
zdoSLrQ@|(jMBd=NyiweTw@KA^jE^7jT3y{f!@A@%@QYmROUU7lD_?)+ov+{66wf@X
zkE=h*=is!uGAdsB?sq@pr}%H*N3WjYx40o+%=)wO*o*v-UCC$2nLOExc%M{0B5oK@
zK4)I>*0|<lFXn-B;)(do4$Pw;J0L&yo^d6g$zPqF9>iVa*w5InS-&`Jz1ov6zWJs6
zhW%=1J=*bWc0-QF7vITGfAa92yvfyfRD5Uu-pi}`P3gON<mK$ty2*t+oinpv>-CIJ
z?d+HATjZbGnMXdD?c^)$h@Z+k;fp*`p2VKiG0f-ufZojSyY-U0{_<4$Dctnzeu(zI
z`@M$WcfcR_uhe<n15h`H=kiy1sBz)jw~oB|wkc<9Zpe4tn|3~^zw=N1P<ZXxJp?%4
z{`+<>U2*Es5A;@W6Q9mK-A94X=2H(>mv&yN{*Ai#z2Uk31(%E)sEnshB!2lj*4s;s
z@0`~;wf~dJFT@S;MnC%q@x^=Vv`^5Fy!2zQ?8P|VyZ7UFK+f--1EcJmpPPq0v18}e
z_EF}AOZHXl!8rCw@Bv@qEPF5xoFPB!aSnm94}L&?@(JhY&cE3SdAf({J^7IXKG~=B
zSU<bsPwLOcB{y}DHf}FlK7MfQX-)U;-1p-@{GA`L2l~J-xuS3Zo}=P~XX~RU{oofp
zd4`*wGcOQ#)PuAqPq;<T>dL6P3Ths5K*g2j#8<97;JTg)bq|G|!!L3|`6v5eN6yRG
zojeA0&l#l$xQM!sVLs2O`z3G+PPh-_eiFUl%d@{{#n0HK`!DQ8zV17IjYpm+e1)t0
zia#0OIitKl9F{MrW1-@%^`ir;hvwXH(8h-RAFkk&{P1Jl^hLk)jpEP!7V`Fd+`6?(
zCLGsa8OPsYT{`3XU1z_yr-I}3126GIPT~f6!$JEDIKdyZ=Qrf)?^wYP<2i?QzX)IK
z!@Vqdq<mQZLqGU%pTK+kqHqdd@DYFhel9A$vWs?nzv-k)Hk`ew0msZQj+q}7ckDmp
zQSe7zXdkN`DsPiF>Tmx8Z{=h5WAZlgW54EMzs9#8uphJEGOzFU5$+RdXI%a9V;^q4
z_+=;L1h46fe(ay%mOMj0xQ@bGan`u<UH#y+eUA401|RH(ec>aicI?5t;)D16$NJb2
zDjuV74sP%txWR9=7eDwhy=yN%%ERFmevD5(){#_y>q5yN)sMV9(<8qT-{}EAsQf7T
z!iVwr6FnQ>IP}4;@PqPK{m2Vn>`2~bohUn$kI4U=v)NCpM;Tv!;XD7}Hz>Zf<Dd98
z4*&Mudw5_S?1dl07yEY4?k}n5^Ka|aP94TQC-)n(ULpUHf8gIUyH@Y99`nE(a<QJI
zSqG_mwl4T7&tspcc!ZLp_9%U^3--!i__w+dxzn5ViU;gTJN?Lszmu2u@JYRioXLy7
zuygt_AAit}eX`$fsrBi{9@sUxlArlYRNcz`NO6Z8;DUM!J0%}+P+Ucg>zvNLS-*Gm
z-ni^QJ;(g&HO{xxNz%^bH@@31uygt$FSH~V@do8r-BNlgQF`i>(sQ>|T&ttjr#@pJ
zAWpzTH1`qwQap4YA9emu9{faH;9th!|NK)t^vth}$6iYm->CS5s;i>xCn?;A@9M*T
z*9*7VyY}vHzyW?#N98f};~YgDj-LGe8uAmj?Gxx9j(S#S^ml9R&)}r`Ecwgp=mEZ?
zb@AfQ^}Btr|JPxEjmkIWyW#+OYDW$zf1*cq2K}~|%46-fQ2MeTmxsV#ILq(EkHkfJ
z7r$W_;?dgqYaYGt;og?-sBz^x{1hIG>+(~$0T0Df6z<D6yoZPU7&RY%W_SG2e$01N
zJR%?QOgth-?ac4~m-AP05ZBm?akcl}IL4(9bwca0e&fSCe(M=-nGYZKZ|>7s7khOt
z8b9zEUMH0Ys9Wmq-i`IhquHPPJMhJO`2stW&yl-(d)C7)tk=GQJm`bGt%qO8J5hG>
zLea#3=X&y2`vLkQ2l68~`7M2GPrmG4zGpvxlDB^J#a`U=lP8)VH9vcD{wd$`%s!2)
zzM(&Rfgk!gm!og*=?%`jY4qGD=RVU{sY9v1(>wc3oNMPBNA3NCOQ)@CsC%Js7FDO5
zx@qfQyzJVi8pf3ui3j+A^Ui5X^NSPw5nu2Ce)?{F@d>~3HaIKps7u3JaWAPnLY&8!
zx`K6}`s2s5x~}|1d?iOzUc_JNpI^g4&+MFD<z4K6-5Hm^!!@{Kf9E~?5tpnV{*pU+
zi$nH{*5keL>Dzd1z3#vJ2lm|is;2w9{9asu^Tt8N0rsFhN>1uD-s`U)oF)(J(B3-a
z+v2P7eP^GZ=|LXsy?Q4(kh{3gpHX_$AAj;2dP3O^xoO9q=>dQAD*m%4R6ZrH=+Dpa
zgMa=;PD!nYp54b`zmu-nyn5Opo0{$uh_CDz4!Tc7PU_9{2*>H0T;Q&L@DZ+)gE+&^
zj7MMk%MZ;%PoCui?2SCgi=E*M{;SKg6Yb;~^vb?b>lDBA^WOO^|8pLrevL1b-HU7Z
zRA;b0`@lLXuXxKF54!6$vwB<q^zx0rzx&fYE$d|u@Zj^;-PySRz%5PZ2*#g0XUX7M
zqk3C@e`=gfYu~=+;9YxK+N0XBuiUS)H~Fr1`r}K!B2Qs=?#~z}pS9E9{KjGD@^bkA
zd(_Xk@)iBW1N$@a7fz{P@jHIVUipoDNIlCu?2^B;XE?*Iy=R~Bg1y;yi^uW;@x*sj
z9ZK9q?eF=A`~!twsCqs-_e@^+CpUlpK)xW(FPcAP#^U~-%K3=*@(mPUsJw_?#R2+5
z=^5^bPxK{@!(;vE%lP~iPU$Cqln;vw#@D`%vKw(kdpHeu$lrR{v;DJv_(IvO{_KvP
zj1Q;SCqIC9)<rI!@ogS_lc#nie)QK*UJv(C>l3%kM;_+Uj{f<PxXymuU*j+QfWOF7
z$%*~(3;u)3E43$Ae6eTk`H%U$=Qr{R)b9?}f#}gZ_+n4w4OhsCJjltuNqhXFzQZ;B
z%%4<TO3JV3m)+uvov{b=vPXVye)dn#>Y~Oqk31Y+iy!croo4*-9koBjFL|gVSReoJ
z-Fy6syXHm7KdJe)mx>SUj9(>Xm;3>)$&17xc!J8?b3FdV?>uXdy8nbv`7eIti{hbn
z_EE+cANh^Aq96YeH|Ytk%hT8ooMC7Df*r9NekT7`FJTY(r(bcC{EUlFevlLnkPk{8
z`onYk3-icZ;S_n{2adoi-%)l!zS`+WU-%+V_-h=v$DZJ;@#t0FV&9L_w{^im_GEqV
zfd7*Vy;?`+`S`USxQb75fY;=#Zejh_VZH8?u|M@6|K~fYeWUv#?$5b*D(`ZC3qR~$
z9YGxh-FMb`=dYjMSGm7tUip}I<|B9KMEJKZekE?o_l!r6#&hq~{X5UrWjy_T_ue?<
zZ9aNQxs!|W@Q-?L{9Kp!){p9k;=e@M32GkHI#GH+=>g>jN#QKNR;P15M^5es!w>$X
zE-vn<gTo*GMh^DV?k9^g&Kd1*;jR4;d_>g|=+ECVgd^e*oG_2|lcV*BU*y4`$%S2d
z&+l_Q?TzO<KIAWOL_C1|@)ndl!~^f)cgo2);teWqR>zhf%iq-*;GMh*F4DL8^z*E}
zdlvp)jC{!N&%`NqBX2UE_=!LD0p||#BlEEj@<H)YN3GZV^y8eGU9vO!fZOcY{aJag
z^B?g=-jI}k*guMo_8s&ro_H3&=+!u=al|9kIP8SK$j8V-{PVs<jYF^aLFFq+<?HOn
zyyQ<GzGwbR|Jtzw<BH?#K|B7VJ$r*I<}*IN*`0RsKI59#d?^3Hzj?j)Y(C?o@<;8=
zhfg?y!eM;l!+6H=Y#i(JY&_#wAN|lT{9=#r#k?rK`MY`8tMT{`zwu1o^ok$+u|NIV
zelmRC!T!nJc&PsPVgKe~pZK<3>(k#n^2OxOIPB0o;tHIl5A*O_IAnb9`LXqTPY&{F
zaZo-jE{T`=q52!wv+wkSYM1<w18SY{(0kOn=|g*VVtjm%BR=rynVlM+-q<A^Gmdqr
zgV3w_*#$r07sjV&^O&FCv19hg|LNWM?9+Et-GslXFTxq~zz@&jv2`0CpW>x>WPJIS
zb9wR2eMtI{?}(dzr^f!oO}`&v&+<2Ulf0LH{oNk<CHoMk{r=S7dl!$5Yd;Gg$xD8u
zUhf==-O;<aZ#?J5zI#^ZAwTk_H+c`cAz$`J59|>i;;(-4BIh6SE&QO?XFPVu@8!S7
z)sJ29BYeR>@?)=X-Fwf@SB#He_UQk7vs;v&=%1e99y@_!D8DOFxQoJVl$=oKjVQZD
z^+WNCI!{3De^B^D-s}V}{OV^f{oQ3Nwl>}WRKIlo<~=>I1Mk%}>4SW|SBFBMm~q15
zg|F#rndfU8&RO!=pZ2u;-Bfi+?eO7z4Rucs|N8s8$MDZQ&K>b-zrcRUhuy1}%4f)5
zJ2=LUj7vW5dBaiu>RcKQ7>}RAQ`ETR!k&{l&nEv)DZgRQ{Et7nhYo-E6$<CYE&eY~
z@f&t$z2YGM7vGZd6Z;i7pA^12XZGyxC$m?7SBiaFCn_$Ii}&&x@5u|E@kjV!UGfV3
z<v|(u;XgT&6Zw&sXY;@*@?fXrhQcFuk#>zw;~P)>gd_MCAH`+g#S_n{_+xzcuJu>f
z5I3w(yf+^?8izdbt9|m#F4dcSHxBAOezf<Ds`F(%P94WR4C`SJp7HHH`ynUqtw%jc
z{Y<^feCkU0!#BCW7xq9N^rN5ki`%}l56|}h_{Bdt`kr}~cKFiHdf7kxpm*)f&;Qt)
z`~^SwP)B1=_CYAWD)ncdN#Q>HBUgO!UleYj<d0fM9TnG-%1h~;-1xh^NL;}ue})_0
z(*t?a6aL^dT=8r@?1^2nU*`ns?al?*iNB+u?(Dv$e(;Vy*(JW&jeUT2{6&4-@74W1
z7<F>?@9)SsC&|BysQ&DC{ruUu+WWgR&IkP7Tpita?fJKJNP0bM;SVQ#?aBU1JM%b~
zu`YU2Uq{VH4$dL`T@imD%(%`q=+Qn4uGz2AFMnb${E0l-1v_A;@C}~A0ps{iul~M`
zIADD7gx#X_YTYO~qvCiSWoOod;uoHv@`<F0C*3~FE7?EX!w0{Hw<x}}Lw)Bj_~GyJ
zAp3^IpL~Xk>Ys2(e^fk#@Ag6Np|dOgMbG%P?`9{t@6pfy?_p=Fj^B68$aDLvr<Z(c
z%V}Tjt=xaaxA>=yz>nB3yLWB{Z}^$N*W&N795M5ed7n9BMCE>tcwyc6cU}q)@$2tl
z@K^p0C;c4_e(hXNT|+z|cX}pY6kp;2T(|#Y@9a)|McILSGV*cMI4HlipRoTBU%eN%
z?1${nJgW<!)+;`_U&kK(Zp1j|LG1(8TiCySjD3W4xld|-<C2Gap4LG=_MhT2e&y%<
zMmzlD(|h;Q&1*dST>M(Mdm;9#&cn>BZmPb-9_R;O-s78`)Mbo^U;WJQ-Z%M>qkX>b
z-sd{mv;OphZ}x#d@2!u0TZer)3NP`CfA;I#kX(|&X%sGt2k;M0xwjw>f+KK6{v)4&
z8{(m7<G>d<57+oFe78>iEq^vH|JEM={1<K*H}OS0HV(cr50JmelkAJdH+dKQFVW?9
zj9Y!?<s&M26CD1-<r^2S>Konu`$+OTd6au7rw-iryD>u}t83<beckSJMpgdr89erP
zt^6L`?~?t0IRB?keh8;M``r*cS5LN|^Y@VsJaMn@e{tiJ4gJjTo{YZ}<^GOw{oOEe
z!{5Pk-)7;PZ(cm<>VeAo{XIkZw*QkSZt8D*e~0SlYle=#=AFHj_(mS)LFr#!D1T%x
z><3QRCpi!B9=^g|_G$h6#r}fdh;RH&KlbPM{HT2if8mex!H%`(*Zf#MV0={E#-DrX
z@=*M?_50B;pZL}T`YY?o{8yZmx8ei7!DD%iXSm1j?NjLk4#01?N-pjz!5O#&f3%ZF
z%4bmf4RVrK^K<+qg=28ndhOTLE9se@;1=8#7vziZj2ze(yMuH1)J|Pi9JYV)o*c<f
zyfPpD$lZQM9OpMl*(G}-XY!H%l<=|io_xrQU#hdSGjbCT)x+5ryA+qjU3^);`o4Pq
zywUv^UUBz8YxBhKzwaHN>}y?o`IcKIf3vT3)tU3=J-PXLx@6M(mmj!vu(hM8`Y^rG
zzw-mXH*&t9uDsigw}0XB**z`!iTuns@Jc=@A7oeLK(6}B6Zx%tF{!!;e4$tN!N1fw
z{9O(E8hB;j0O!OZ@yWQTI)VL{^CvjwJ$pvs0{@2x;-CKb@!h_OAK+7-0GI89#7WQo
z{-HQQ4)Q4Sv;UG0z&H0^;j{RMKX?e&=|y`~JO1k2kzCX#tj9i%p2(lS@mF}k&fqrt
zLfIKVgjf7ooRo*bCFk(wk>|2!IAL7R@Efk9>;W|%DlWlWR9=a~DHPu*xuNn-6dorf
z*EavJKjYMQPk69N4)zoL*txrVj;Q@N{*BA8?K`wn-(VN!lMgvRq-XjOXXwNJ#d~sh
zj;7urUa~9e5|`M8dGI4IBlq?^!*Y_}wd2w7Jt=mG>Q|!t-M*E)QSpZS#aUFIz+cGS
zvwfX-<(!rsS}%VUr~EF}IU_rP&*GZ6;diIr<DdT3C-@Kf)KPfl-iWwm9{Xo_U_T{3
zz+3lT=o9|17w24X6W;S*`H_7Nyo00aZ17ip>;F*0P3vKQ>fQD^_9OBh_>3=kJ$u)W
zed{N#$yedBzo%^;aYQ`CkNueU>h<K|-md+D^@+RUhBzmlsVAnq@h3ir3+j;O5zp8I
zJa>+ut}5Q>Z(jL=@AN?~;yOIVr+mP90r{&Vh%@v+4(y0Mis$$yU;lsH{?t79hv#t9
z{)d0z8}7kp<Io#>XU}lke!{x7=Rf9$Bi2bz_6yp{H`qC<PE?}!hQIbd;w8Kgf8iV)
zN6DMN$gji|cqZP<<IV3rh<GC&vuF4TAMtDb_KWZi&Wc<13;M%<`SF@j*B-X&oUKiD
zLh;)^%zj1u!LR&Be1Lc4=kJz^E940;)c?tiJ@9vS;k&pYzxAG7bV~V&{^Y10Abzkv
zc>;Zluk@gO@&z~KA-=;G>ol)=96x8*`mt;N@BEP5+;j8Z_||D2ewzMn9C~E8@)bB_
zew07sn>^vGIN{urpW_2gnqNQT@-O&^vIqVmzrjDg6Mx_dJ<8MAGs@mQ8`pfk%g^ky
z?Pukg<}t5vyvHa0QQze|aF0LncX}arI8T0XSG^F0&+<6^@RRGt54jr`zvL`0XFvGm
z|KwtP{o$kh3a)4mw=&<uk9`Vzw@%~hFOK0)J9;sm_xR_3@R;1mnf;QVxFN1shxzG?
zeX|eqvNP|k11{o|ALjbW$vpBiIL3a-8Nc3>hk4<;XY!L@B%kcbcYbF5+Oq@eF%MjT
zEATMewe>Lk|1^_7^RYkcWY3=UGrslHhxYhI@y8#MAAZ4K#T9nJug#B=7ya087)Soa
zPPAtS^hp24HLrD&GdpId)?@u}!n*l`^=n6e-X|aAqdoZ>*Ltl-JR^Vn(yQ^T4_~=H
z_K!dPQ2fvrzOp^LFpu{5fq&l1r_?Xq-&W@|KKYV6+;>iE9C;!>P~&Q6J>((}miMFn
ze~I6lJC}0q<@}02+sFBTDW3hm5BHty)0~Idzsl3>!}Qm$j>?PeSMC2$_AS5Jed`fp
z_gvlAQV(Bw?fWO}c5!cuzSt)`BTxA+obo#w@rgXQl{zn!kK!L*!gc#`e26djmWQ$%
z>u1;ULi-K)!Cv5^xM2SxF44RFtNprjRd%FaVSh`H-m7D<Gw;<$?5E{->HyANoqxKA
zCa;Ef@=N)$XZvIFRWFy9$~W1o_xQ7Z_M|Srezh}BQhHXWF%LWD59;FDvn#k@zvdi8
z+>oEJSNREifg}2tD13z%@PmBKBi^u2_JryOZ^aeu*{8T<A0V%>F7qUXt0;d&Q~u<P
z!VOejf#!VbUa0d3)H+f78&sS?#ch-yqx>A@4=6dJ^n=PnQSwF2gR*CopQCU9Rp&*;
zJ5>Bh$}ZKBwU=)gj~%=B#9rOMa!%@8P@Ncl%OC99-Ph8O9_$m~uYEi`g~Rd<c?db<
zn;yjndXs;M@4kyK@*UsR+w@mIH=p`|yvR9+db7GDJi)*7e9!6+__ls=MSiNzNAGYH
zAMUw{JK~7()QNnDE7l87zIf7OBTjqYV2eNFlRwJC_@(#c0!PF<cmtQ&J^YrJ!U?zl
zf2>zK_jR;i_>Vt4=B7I~HQ=;3Ag-v(^C#z<@PHqSN9+WC@=NQ5%W#!^w1Ye9X{fk`
zvTJ&x=hP2eAy4urZ+6QstOK9o6+iTz-q`_sa6WIp>X}}x&+mrm$NbhuF3#t@m%qwe
zd_Vl=ckTAFAMI1|7x6%xhJW-cPnLhkXXLf+t;wIo3;Be6$^7gMU)p)L9(G{;{M`E0
zQ>}wNJ9khY=D+Ha-s_K&FFmm{_`v_+C%b@e_%lBL<G1oM{sr%>6QBH;A2}ECUVP&}
z@G5Z{fB0sP?A-k9);Rc+PqIhP)<d4Y<BPqQD8KT1GJYiv$<O(PdcNOVI;R!)-~-D4
zoCjEsctPIeMIQ8wFY_AL{y<#uUR;CY@-ffWYkl$oalrYCcGd+K%_}bAPd>??@Q*+G
z6feam@5NpA;hFtgAN`OU+;qO;nSHS*{p36JN-oxk@}u;B@7V+YV_)JM|7G{^6D3FE
zvkUt_>$47&9N3BVra!Y6`)m5ZCpma_Z^FFvP5%0O#;^6!zj&s-{ER)J#_>CE`INc}
z`Km{=PkQE8`k`<{o@t%F!!7n@Ty|oAPTt<5{K-1_k#(~t@~0o`aZX`;^JiVrcero8
z{F|N913k*C<qh_6;yk_LhkWp9eEA7I@k{Z9AK=qIc}G+0CU^G4pXrso(S!ACCq9vv
zXL93L?2W$32`-x7eOv8N{A%YtieJy_2<%>ezhm*;xv=wNe~*Fx*w5kL{PuG>KKr#k
z_933(lia;02friAcIL;Y`bYYseW7|Rc^KFF*{^+${S1EK8~YWP$i=hv#$hM?-8kfi
z4|sz=d}+^r{mu`Lii_|;yyxfcbKqMXHZH&O9sk<NKggS$_<{c9Y`xY;KJ>}o&99$%
zwKE@kOZy;iIA|X8;=?+;XZOZqSK85&_U3nwgC6jSe>m*feGvBUJW_xB;?sNbB`5gH
zF8n@?-S9VlKyL8ad4l|ioxmgax8apKE56_^KHxfe$q%&Wf8=Lf+Iz+q`C2#r?F-#k
zkw>u`_NtxzF1*Dj{_!balE*j~=P%?g9<ncT7LVj1@=WWH*XWNQ)OY?PPY_4=ll)EI
zgKu%se$YPFeGl*L3%$pWdV&3w{^}L#@Th%{by_znURxJ=+h5Al@uhB}j$nKg-|}62
zv1je<)9uIXr^&}SD7*K2WaFapV|4`Y@nIhNpf`D-e%hPYINGZdXlH*(zwq0+iv2J>
z@l)q2>Qwgs>J;9)$8S9KDss|J9fDr$FYM>x6+2=d<PMMdxBV7BV|Vt=<mUb)duK1~
zgj}5$S%>fB!XB(oJN&|RI6#m1K;<cLMxF|vtWP_5A+M9~@NeV5TYdt+;0wPYH|^x{
z#)l8ui~sP~{vZCqHT7Tq3NP6`dmu;iiI2uNFFwcx-ixalkJRnuVe0hoTYJysD{jDF
z^2Qf_<jLgD-*TTIj)<ei!=L}NXnt{5To-qJH=exN|8Mf%c;cG#4&UXi@)G$1IkGqR
z)A4J5axkv=o76bsz5LRA{+=AZoX3!p^^=Eujs3DGIHoSfulP56gJW<Ve$Zo5ahLsy
zYwTP-Lp>3m!C|<so(qTJ#j%&qp4hXdr&6!?yQfK?IQ_dTUoxU{j*|8W2idcIE_;>t
z$q(4K_w-C3<ZC}5|JIIP=#AXy+k1XtJod-`*n{`RN6pXP$e*6!qWcE&OL0<OCGSy(
zgKO*pKI7Lunti|__$g1)p8adb9zDYY`z6oT2|w{GK8Zu@(>m}ij%X*p5jTw^{)>O^
zNz0$yn{Xe%c>z3@r|}=};T!qkL%W<e`O#iJMXvl8<;Tt&=+FG*lk3KZaoj5xH^d7#
z#&772z1errH+yG)?meq(%k%8t#Vz}B{oxsWcJ9nC$b)_1oBqvD9w_^QOYDZ<7?<9(
zCufv>!7uIQEvP(#pOY84@(1hlot%?z{8+E=&c{&iwP(+$@#zhPKlE+itRFuo2Y75=
z`4zu_^X`?&7s)OCfFH_p`GL5^ZpjxOdA2V8gv#gJ{yBU<)q3QC{Kh)qHF>iy?aA3V
z>;ffMb|8MUTl#0$>;nFw)&cLe^Ng}*a$uLJ@8pJBpZ$%xR2?Nx=lkM<{9pXRk9vFh
zCp>`z<b~qXcyN(jkPkh>JNEB8|0aL^#RvL?589`_vlHLR*Szr9z7KUz$vr#I_J!n$
z`nxUmOMXws-mQzD@F%!VPWETo(+7KI@9dUb$PYi{;JfkVb;iRVY9Gn3$(dfs&wh#>
z@elR^hpa=rRUEM{_C-(n%Ol+5Bro&vYxfkyPy0USb^O|V=|6B>-m9JcHhHq25+zUb
zk*9q>Iq1({JmbSNxzH=Qvln_bzW(Hcnn!<jK)?Fgf5Izr(BHnwcm3!cUg6*T=CdC7
zo8udgozbsn{n<aiCvW=+`yDu|KWbk5!lBd;yVqVE^qxG8!ykR;=lqb~*f}|=`>4OL
z8{=B1aqxqBZ+>yteJkUm#&^HUK7(A@z5EaU(zCq7-~VP8@+SUG|NM=-$Opy0@9>j-
zidTsz<jcOu4-W8KxFK%v7x#|vZ(l<`>_L0;&>#Lvl;6-3`LIL#O#4{&isDOtM~?JF
z&h+7#9QBv4_>RJrq}uB*ui{_&!;7T$BPjdvY&=xm0QL81{r=s(Z~C=<`eaA^g<Y^0
z@?>Z9!9Gy+J(L{zm3S_1;3w{F@o#ZaUhQ5SJ4WqS)uq*s?OQMX{gM0izp}6T$h;$O
z9y@(ZCExb@5qXJue78>VNS#OBQk?T%oD?_h1L2)@`dz&Jfw)U=`11F}{H}!_-S6{v
z@cceO{>87zmtVm#`+DnezQ*6_%Q}oF4;6RV3;$#v+F3ujs^hBvlQX%nGknXh#1-c^
z^i5vYC$B`!t38~CW9%GdmnG`{qk5e<!0tTLEB`^+LFUEyr9XckAu0K)6TlB~8x>#R
z9bAEv>VV=qoJz|7;5YvjkI0RF;61z09)*vd`Iq;`<!{=tOMd3r{HXS*eV6>ed;L*%
zArFLy@@4mCQNK&`oxV}_0LPrC@oVpQ6s2e90P5hVcq2}r;wbyD?`9|J_xA6;vmfzN
z9Y;Iyjvo0Z`B*nxLD>&~fkW~(_Q8)(`GkJ#HL36BNA*X|BTqoxw`WiC58v^b`L}lT
zz)#h=$r~kK`HX#n_v!`oo_g_KfBJ^o^lm&j0$=cFUG@+3Yd-r3>o5--@t%FiU*MB@
z@T0$e<fWf+&7b|m4d3YzzR(XGFhBhAUVJdGxI{jv@lkR@;TtO6p>P5fceFPT|MVWk
zCqJ-XktdNex$rORl)uT#`60jXUi}jFonMPv)<K@;)ee97E>Zp>j@UoRxA5;f`TxiA
zaX(o9>Y<7s^MCv%H4a=*ulL^g?pe7%C9c@#@E`J%x5$6xS;iv=>!eS7z)g1Ny=U=G
zzD=+Cq2ieS^hdAw)t?^a&z|*D@0M4~W5|_0)yK1+^G*8?`60QW{6PLD55X__q2z$-
zZ+~FlBmT-mQU2)JI^>Pk14pb|-bOC^nNL4;PI5A?`HZK%b;&1v=SSp6PsZ_`{nH=H
zPT?qg6R%OYtG)OGpT!OMglexp925`D=Y8%QytgmqAMD8Za6hT|X@2`;@xZ+X&*txx
z@+W$fFTy+L@o>cc$vHdxa1QA20Jx`-IAeYCF#XzbdD-&ugJVx?!YlVha{p!Dgp!Z=
zP0r-+{KNe^^=7zdpNrBDN**Y@LfJPeZlL^tKe225QTUMlAWxHTiiiBwJn#YKm;99<
z<HLHiH;?_GcIsii<IlLL{RF;HeDH72_=aPCe~iKb{LzQHivDQ+-xPe+QR`EOaE}EJ
z*eAndb<u9A^&hux?UD({J^y#p)eFpH-S&s5{T(}~qw?kw<?r;Q{(!=Vr23)kUfu$K
z*ggD)|KuQ_5(mXC<H~pRr+;$LkN(-E_v9m<l_;D|%0I*_{NfYES5kQNLQyzkAHZ(h
z8^Q;?cfKpXaIWUNJe(ZTPSx+!``qVJ=i_(&9|b>OG=IvB#r-`MeZe#Tui@|iy7BW5
zP93P!73tGG1one(bwGBj9_aqAe3V_OL(x~_96J#o%#V^AJG6iHOkVQTq~wc=SMuz%
zD|H#|<-5k^cjBh}H~j~H){%a#KML>gV}A3Y?tj`(tE0H@D8GV3_QRRac(1?w20!uu
z&nS75i@bpT<hk;5=h)-~$K_EdK2UiLia+%L_niD)4f&BgPoB!p?6=&HlPAJY`3uT_
z<)7VB`a<nbO4Ph1DsM{)hveTSD!(hy(!AcwBTKYf`Aj=>=If4MHllq$Rop?@848C`
zeuApQqW<m$>i_tl=J$7*k~$wjFFR<`<bU2Urb6KsDqf+fCw!y)r$pTcM(uy()$aeN
zzSXJZQ~W{xug>fDUf!#dWPZVK`86uu$UERVoU?DFA9#iuw~oRyc8Ky@`%C$R_izP;
z2Pk|-#VgdgR*5=?f{XAEzR;U`D?Ehr@=x|I@0K5mm-2k^U%pD-@-cA{ZqchcoqCe}
z7&+1>Tp|x~+_U(`A5r?t{@TMo?ND({yS5)(Fn92wnM)h|M;#x(eh<PA+<#Rk(_XzA
z#jm)gE^42wJ|Uk&$(LT}i5^h-R#N^1AJx~a54Ap&9#Q%&QTe_+oZrbm?1SkQU+SVs
z)f3>0`*89xeh&ZSP4ugdAupCEs6(<B^*Qwq_bE~9_sm}SDay|I6UyK68UOmjC32FF
zlcRCu(d>fUJ^Q}^>?Nso<R-6Hk8qzrzA1kqKk-Gsw%;7J_YW?ewyt4a{LcBIeTnz_
znJ48#4){{H_4g9(<J~V158xI)^=BW}XP-xY;)VB_pKFI|XI_3tZ|vLr_@@tgG%tTg
z;X&q0{_oR*w_SAj;FO-0zfZ_d{GX3W*KA%r?T}5)&tG?E<NgD;G_})TKI<IaJn~!l
zy>{D6#X;**#}hB@54@K@$j8-*Q2aP2tfO!?srA=Ue4;r{K9i^4;aP{@ANhTfeS~#s
zXFc?hR6F|)zbli6$dmEm8P(rB_G6y4^LH%V^Yh)`v9LbBzw&=xj887UTfb*~`yG~i
zik;~1S^ntvv+Ulu^k`gm=e=ino!`xqFS{ZS^0Qv!*x#CmzvyQkRD1eLJ^8&GK2Y=T
zefY#fzxwaHw&X)e#WNKC$qU6n@yzeX#7X&!{83&b&o{381-@ncl8=hRzLzK*-Bv1o
zY%7J=sCCMd>ZrKwJVBl%ud*M48}fg43-92c@AggFv44KUPuP)uaD>0gGvTRm`K{l<
z`o9a_^Izvb`tuj#IB!wU;FqX;U!G4-@&M;d@=A8iUgeAW$-nd`XZ`5QckfYlnE431
zm9Lu5_}a4%RGpMR$QwM%bIdExkx$tl$h-0F{sa5)OyBO&$+z^k|DZ4OHxGI8W9zif
zWT)E0TfaAChyJe@`?fEVzrYXA<j5}Czuz1FKW)~nUdCQL!!h`!Kfj}Q`%L_6$3EGs
z{hMd}v1@+ApZO=blE1nIoFsqpu-|~I^rQ}KU!fm)8P9t0>AUselV0!%H~6vNALsh)
zTa3%!?Z@DXb<>}D#5H)NKfASF^2LX7-Saa(`!f$Y7>_=zn_jgick{|0Q2Ty-lastg
zd{^&6%_q*|*SM(m<Y9gMmfTSJ0e(?-V}06_n|AoekM{hBo$&+x$PtAva8g~@clgX6
zwKKl{+IePID12pi_G`|G?T747Z#?BYx6D|%rD@;d{~h|>_vATC2G1JR+wyl9#IK~z
z!{NBPq50J<{GD3o-|8v!>F=_^J^n;a@=(-w`7J)#34Pelk(+Y|_`)ybm-1Kr`4@W<
zC;2~osH5UZ9Tmq))c&@Pio2asd9iizJAT0*_zUVgxuu^QM|=LxPU$!0443q0=i2i_
z`7qp03YSV${RdSSqG$OCYCcpQhrO4me4viXAK;Yu3m42M&oQ27`H=dE_MX`lJCsk!
zx7A^sw>S^hUVrV4tNx>Y?HL}}H<6Ea_%}ZJ`<<Hh>a)(N)h*SD>~lP;bE(TYPel1Q
zT!$y{!0%+)CHrH?@(%XsImgw0XH)#)1BH9=2|n`Y^fUe@j+2KtNFMymdW}n-<l@;l
z_(5}i{ZQkYkKeNw&v1^Nq3p@CIO^Ph{joFt;5}R*U-=}Oa<X3hz&UtfT>OiV_!Otv
zEs9_MS4a6hzd-2$g~xCl|Kw#JIP^kMc<o+{d%|$TeoLJN&N#=g?_tm4n!kG`Z%~hu
zKg$E`-{l+f9{(R&Jx={sT|ge9-lZ;u56|}D@_KoYcJe{?E#9m9voHB13b(}za#9a4
zp7=(7?orUUeVqO4!S}8|>i3@>Rrxzo?gz>L<PYTH*?RahKlW@M&+;tioa|Aa&HnhY
zIAp!#qCTLG!2iTo?Yw6%#xXCv<HzcG&Vk9tcX_+`A)h68cqZQ1huL4#BiwPm;l3uj
zVIO{X#=rRsKjGi>3un~H$(!G>3-Q;z9yr8~=s`brz)snT{`m3UJv8^+_>uL~55G{K
zW6$z<`$kkfgMG-O)MfEc-*BAX#5wcvf6wYQ?q&J^s`Lq0={e~QizZ#R;GVupo+Ynw
zUl@*w3-(RcEx)60?e*8r`1Ij?j2_)T5O3fHyK|q#IQEhDkH!@@@nL-YvkUQyUc6Ty
zHV?b9KDci@@e{=tDqltAgXl9?eqremw`^&k=0V91Ww$82LGg={AF3TnUP;9}l;4U=
z{1;xbGx#cw$*bV0{EU6VBX)0{?B9F%g6faTA5ra4c!<g`Q1hVpM6D0y$K<oERJ{n5
zza{lu-bc>XYo9AG5I4QoU;Rt|Ew9iI9*HyN%lJdD<VDW(DUPv2I7<HFPvV~TzN7ky
z*PhMKUVQhAvOj#PAC}tVgPuKS9ol%PdMf_BXE)^OJ^LbG{E{R7P`E`u)<qxY@vI+y
z%%{I+`oW*~#>bENVZQ)JjAvceA@4E2@$m(Jtsiw?$2!zI@Z-JpdXJB6&(Gnse)22h
z8An{h4}W5h@)EcTKgrGf)@NU-U;F#;xrk@{vp4eMul!1T_KA`|zUj$2<y-uYo>1p7
z)^A<(EUt>j*$%&O+<N6t=4B7o;~AcrR~{wrQrC0-sh%ny(%w0~d=g*mckTQ&kKXri
zZ%h3JmG`oflsh?*ukZeUuKBEs-@sYxWRLV_pC|7WZ{ZvJF)q2v6SU{Y?7+IS9?!mv
zgD>OrPxIr`GaS-Sp345^C+ZjUPag6tc@Tc;s5q-XJK=}c!#}JCzwA#x@8x^!nOyDH
ztOE}C?is~jr<9+Nk8$YH?=R^SZjmEAVwXwn5A`=6{aOe6fD`Nn4x^=Y`Mr<4U*2tB
ze#)5-t#3{0t?a+hFReWGnO7hBYy&Qt-}=ejzRbRX{`qO@pMK@_@^*M=pJN>Px_yN9
z-rJYB|7YJ}-JXqW9P_D%iwE>UU;IeCOiJ(K1DuinqQ;>=IAtC<1AmMw{^3i02`{x*
zN0Q&eHTz$6W%rrb89t0tqWI{P!keVlpYQn#eo%UWA8?M`*d@x3l9tw0_a43`P5wRC
zjpw_#P@>u;6>pQWhdRn0QTq>Zf_~*&byRzleoK^`N|atplwWj9$;<u?fBeFC&+;97
zvPb?$FV@Ze*r{>oLqC+gcS`Y#!eO*dPri@+`cs>BSw2v~Yj`N$c!tmLPF@Mu;i-JW
zcX_D(;*2_;c7B)TcQoP&YJ3#l%LlYW<rS#>0;QJ{6=&OY`0rhlZ~Ohb`W_kmy@wn0
zoiyW$x;Z>jS7u-IsSYeoSO>kh@9(@udvyf&6x224Q|42LcHT)Zp0!sGLrZzZ_P)1H
zaZ}!e8h`aKcis1-S-Vs>zVU!X_kVj-HTJEanE1WfTO00$sQ39l#{Tbx?|!G?K8d_i
zKFL1hdHh^_MCJMVrQfqdcADp6^eB#_@&)6=8~EmYME)tCVL!gZBX}duyFU*1_^W-R
ze2hL&<6BqqX@7-Z6yM}#U+6u2A?L(h<G@dOAU~vUb^~WdK6pWO?1u*`{me^G;t9DM
z&p4>{%QNVce~8!OrFQI!{jw|l<fqm{4sgLZ;sJf)mwn2Ij0>0Gn(?fIy?PJNt;4*X
zQhvFksQVvq-uVJMw@-5K%kMRutH^tus~$1)k$ImvV?;&o^d!$=ck*L$KIXywR-J#$
z$QC<i|8SFE(T975?vIKC^dP@N)#t@8{<FskH+|_1$Bt;Rdo=gk?B(xXy5iKMALy<A
zaQViCtNKQ_aS=XRpLKg?muWZhT6h2-%xAyI@0=U*i&ek=w>R9f=avS1Mg9HY_Wp3j
z^}Eh~Z%?)G%{MQeboD^x?@u_FbbjUU?ZR<?=fe99&-E|3WZXbyeDd&j|IH5<I<-^&
z&u;hw+)95TSNR;d!fCWLKO9ikGavg&f447pzRNG<7x2b8fjofy$rsqOesBv-e#r$c
zrGM}T_QlW0QGfPKPAHu6T%zomACf<Rkw?jM<wNvF4xUltIbXLP_r0w{d-{eC&TUY9
zq3{L82Z~?R@5NE=Q1w;uU0u~aU!BrD-K)-=H}A>KeJ#I3lrK0}5+~Ik{0<7<qVi0*
zC7;mG?^4`bac@Aq*F4&}SL40&b8>|1aMQgU-|=Pr>K^#hPCj6KeEaSm3VGs7-5S3+
zpZDUCJPRMjF)ulppWWf#cm4eS+V9)wMSuDrKmEy_yv&c%hq^j{Mfo>9TDSiAB{%lS
zp78IPy_wH=?Abc?BUh9>^(P18d!|ounS98vM9B%2N05Vc<BxsG*NyADar9?L+R-09
zq56BKS9*bS+L4Dm8!l<j9_ZV;=oNnXJ0Sc7?!Y_$&&sp!?u{DH{PgSD@0ZL&Zsdqh
z`m%oOFdlyKfs&_Z^YSm>jZd!PoPDnI4u6-<{#cy0Pg1{-C)pp#v*l6hB=P`pSN^B2
zF7C?DQTLzB;~tJY%)KjlvwF9Eum18gd9`}C{K7p!`v`f4-)VTZPU}F)u|(}R$icI?
zFW!om@&)+dyZ7=8?>&>h`!B|qPnaKNAN-2lm={0B5ufe9P~YW=_L(TY<xlcEfA^5x
zsmIBKoxiX@`K)?|XZ9c;r$_e64(0RsNq>T)D7m24gR)!HJayFi**Q7N-{1p1z<>UW
zZ|&d~`J(boc^7+@Z<CYynD+8w@BQ5#_OJdXzqWpLH2JyuUD_jik<Y4sz*Bj;d=%ao
z2hOoi)c(PK4$b$uPi43KfqYSN_Ka`13diw<S{J<45ALXwqJF30?;`noFxs;#{F4uR
zCO7woyth7fz<%jZo=^VTxo>U#{6ZW@#d-I+*{itgy}ZEh=;SNTk^P?6c;Y2Fii`3h
z{-V9SjJ)wJzQQ;2%2)OGT|e<x{^orN=l@*a%a6p1I;uSi_qUbu&u%IIs-yhv&yDi4
z9Yxs}dGJs6Z6AipkNJ&y;Q<eQ>%ShD*xzzLNnGJ){MY_M{x8p!|DyJ%evi(+;FaIc
z!Xt5wzoFK}A3ej@d=F3Im3k)r*NnRMuvO=5ZHkZVLtMx_%<m}ZOFZEB_<;-LPJZ5#
zH@qZ2xMe>io|spi#dy}IPEW4dk&8MHT;sR+bq^XJ^d;}$*T$iDaY6or`W=V$;ln!E
zCHWbbp5Zfj!3B9Nyp^}2)}zix@9fri+KKz*Wu4-uXZ9!`!k@V1_cZvk4s~zi$ZN<)
zo<(o+GI=__<X!BJpV&u<|MY5{SA6;nCtmmMo)$fd-{zw)dU1cyckx)glpf8ezG}Vr
zME!jYdZ2H9CjVxa=12Lnb;&d7pB#+OFTA%t_CXKkN97gdUZQXzDfuQPr=;wWAF;0z
zg(FGHr_|4T>qg-pN=`}DQRtoh$*Vle@9kUcAD!pPU-^f5wC5M}%>MYH@ze#9-u29=
z$9v}P(f+%$?8^Sa{?R_){?YHB-MbL4$W6SG$HGbWt&Zk?AskbWaUVgP)6RF#>{)(p
zeE!}3yW_+6EzCn+&U4_NeWUY!=Y;g6KYO4b^5H-5(RtD{4_<KC9V7Ovu9@@ob-T|Q
zRr&kCSHJ32XTRsMq3YH9&0ew3meG~-Eay}9ea_?jzXA7K+_!ml-MoFCJ#S34V%E?j
zH+^idQrB^Q_0a1+^)Kh$KUmrC8&7}dntm5%eE!OA$<6QB#83N2cE^6$AN}xW`q5r|
zfH&lqlz!ke+(`Y|CyN*Q@gsF7btZC#U-(A(JH3ks@(6xM4&o1cu}|dpN##*+7=?H2
zT>NEc@+W#A7k=!1Ec^9ddvYg7`r${YesITlsJe#V$<UkMWy%NHlX>ZZ{``F{^T0Q8
z*7)Kae4q#VWuNMi?sL1(tDiiBUE|xj&4*v(iKpz#d;RfEPu52csP(cR^Qg1>?)SF*
z#ytEIAL>f%P~PD0gsDq7cXyu-&WjK9O+WO^PT{b+*9Y&t@PSv%9^Dcb_>sRW4QI)P
zAJG@;yLG_>c;*?tp!^%<Cnd_C>uB3g;=Z2qDSj!hl_wil+>`I~`&_R)!8q*MegY1N
ztM)nKt-L@SW>4lb9$ez@@>G5)FJ*`9id^{vduEsH8D8=?dXpz;PmkiWam97~@Jse9
zj(Bf9a1U<i4=2nA@4e?Qxo&<ZF0cdl?EF6xaR$y>4|!P+KQXR4uR0_=F`n;m3LnOS
zW9s_OYw%}X@Phx+6Th)8_DHY%5Dth7_$Oa_U{CZw9>$S(SdaC|pV)~yxbr4>WF7c-
z|4}>ZR_7&m<H^U>J^2y863<b73y1U*Z|$4lFkFOdsQl*aA0Knwy>mxaDElA}a(2!l
z?}XFN5uL}O^lu*glDjyF(l`Cn7k<S(-{A&4XD8b8H~jNcbhjIC|H9+5ds@F;bk{-O
z`PYFK+*ChO$5t=1?>>6L;uGF^e}BvG_1pvScTd<8KG}tQ!+8|@BA@m;_TT*jd+vQz
zQ-AA#6WLzBPAUID=_9Fq^tMv=TcYAZiPB?I@q=H(XV1QqpExAXwQkS)u{UvvJ+WKk
zlD}v9pL~knu}^lL^J+)0@Jc`R7kQF(kOx19H~ORO!TjFyBl@;p{JBp>F7N<umZ)}h
z)Ox$6{8L>CpU!{m=lLT%gn#ba=ughp>F;>TbM%uR!71{hZ}vey#z*yMkMx?9pUB(P
zh17%OVZLjRPjQf4$S>u6_%2cVx{|;9UZUhyqWCCL{FW#`DA7_s@70@2R9(JA$+eF1
z&+Vn^?BWodaWBcb*g3u7pWWb>y~?B5zxQxKKE__f{W>ZiQXiD}xi{_EeokFV{wP0^
zcly5>>eTQ;9mwxt<)!is_on4#_P_E@^W)2Xp``LjxJgdp4LQo^#07q+pL2P9$gALv
zxMMtgt23%!z;k({`%CuC{{M=0@*evoRDL2~mrokU_~Z`Hjq7~F_}0T->=)>pU04_U
zk{6L1d-N<n65q_@nZ3%7l8SrymUrQUUl`YSRGcF><KvHi)0=qiy}X~_;2Xt9iHgtW
zmv7Jqzad9@weR4E?oD}?x7(-L-?%4d9C||Om0!4rA|AmTaVyVd-NV!lZi(a8h1##8
z{HH{%3zb*019Boy{zZTMRr|~j#dmoUdlygi^B#o<DF3m~7mw)|j>8#s3HFu!#0B{0
z?;Wr^_vPTFang?EDb|T!l$^-X|3h*fz>nyU-|0^tz9(g8#zD2yAOGw}yi0w_6XZ$i
zF{pDo`BT!&&+@sH$K-qbq2^DjF2yg^CH?+~zuR9s2l8xRF22DpcrKo*Gs)BB5#-Ds
z#U*&gpWw83=zf87VRbw8L3z2lhIlBCBro?T)I;R;a7O+s-xi<PqxageW6$)*&gq*y
z`OePB!+mAH*Mrw^lYfdw>I>q9I1NAfIsWNGJWWb({F}bhzxg?w7iZMb-KT_qo~;*k
zuU0#LpVau`lzf7G*bBd8@8r&|P<ALTd#|3)?#x%B^k6>wK6Ygue6UOUha3ErJ-f%F
zKfA`CaqtbFtOvCYanN~)Iv%?>zIY%{W7pQlp7Wj-zWffsy6|To;ogq*;!j-B9<@&T
z<@fxBoyfQOgL(>kFc1HP7xv-!hBNBmsC-pC?UcgDq;Nod$1gmTuk#aBehaVVPx3!F
zVE@9e;4?Y<9hUXN3G$R@pxTLV?9};={7XI{zjM#nxd$A>pZ9It8m>pumvzfS;Ft3_
z_+wu-=@X}acjZe)RPc(N;XS_-2iU*$uv2lsITCwRmmx>;ay}&gfKT)Rr<@DHUB7>k
zM|$u6LsIkR`479NM}Cgq5@mn-@hkBd|MCR(DPNLD**_&^cj6{W4#qdH`~lvOCz|8a
zFaGTZ#0z;R%AfJCo%~Ur-YJ#G>)$+o%7>S%ezuY0^Fug*AN{O@y~8_shIq^WwPSzu
z;63?zh7;nj_QsXZ+5h-X-o_z6_9QOKFW3z{woZEI2l(;#wB-ThgW`jIzzz0+vQP8k
zi@e}#&QJgHTk`PSDTR+ot#e?()?3GK?5$Au1>OJj{L$Mk+4SL=eJ%S}=Rn%|9fP{F
z_u4x@k~etvce3q&J;QCkQ!uXga2Q_ty|C~0<M7is-h0+x9%3AMIb24?Y4JwC%yYDd
zOZ<^tzytY)@!+|6<SW*zKmUU-`muBRHox}bru;B*R$hWH?d4m>W4C_CCSIz)d3G);
ze(-bltlsFnUY#bXdZ&G$zyDx7@yb42-12u}+}9ER#b^5o^5D<*7yJyr#`E_k%m=67
z5}f7dzN7eEb^N|#MxNVWp~i*Fa7JDwkI`S9SG~)<K=%suBTsfjkMsiP?epL~{aOck
z!$EaIcm>C_hp+4kPRb+YX>gIhvN!c4e!!2sXaC;Ab^aiZkt4Z^+uGxk|KOLL*t_w>
zH|Ja0i%0y=_@3bn{*0@g@yN%1$v%x-;DdGJ3$<Um`0_2cO#WtH%RF!g4w9?9+qmY%
zFFep5j^oFBe8Va0#1FlZD}Crk58}PNoL#aj`MW&MchA}xUp{Ai_J-0sJQTOp4UFfR
zpR+5tp+3d0<ol@Kaf`pk<5#GC6aLAKpRr%|MqlJiU)E!s<kS4}GI=gO@nQY=#)oy<
zSCB9NAa{5OH|UETQSuW<oS(pB{k&f^f69!-{XG?&cAjNj>b?HnkbBhr?hwD!52aUe
zlbygD&)VUOyx5QYu|)YVT%<4l<Gc9enOw=)d7*PS_Ca5s>BsnRf&Zz8!W(uZZjcA+
zJ<85d_Ew_ohu-ZY#S!ZwANoT11^K{Hc?&*J>qp@h3hz+3fZ_`!e-tjF@<4H$JWCWV
zqWm(cyc3>VpLOeRUE-{H;RXt)#eM6M7m%0rkhl5h0aYKvN7@TJ6L07fzi<qNZ*YP>
z;0ixC5C1ejsvrOH%)a=SXMUo-LoVzLzx;@N;Ul@4M|=Fa563^n)ASd4qxIQuke_FE
z;XQqkAHOyazqL;IMz8F{`n02Gcx9i>UbIK`!v|_T=0&}iSE1hLxcYhDro(kR`poxm
z&wG5D4<#q^Bya2RTu1TY@7~&f*srh~_=56(c+TGS6L-WR<A^(+^|vpuKhsV;<KO&?
z{pI+^vk%Z8zo_}eW&0-YlZxl~6HoDF9`fWraFE>O!_JSy8C1Sye}vj!YA6434@I6N
zPgUn-|JEaZ8JC>lj5-6n6n9Ym%<lNF_`=V<*N(kfKir~Ec29r&*t)#8esMbO4IknQ
zTvtbdbNs^Z@}2XDH~v1Ib5wk*ciM-#_u#wpVs&NuU?0Z!-T4stiZ9mTJ-IrM(GGsY
zM|fnN<OnzE8LqHj_KHux+Z31RDdT`~)f4H3-HS^l>OMt@nzuy7-<?IBYmhU!IS=!m
zeWUD)p6HXF*rhrfKhfSfn|-l5g!b$~dwhx;Z8{up_*YVTmnVoHxen{WHyq$s_Ur5$
zU!LV1<d^ek&raY2J7!Px4A0mfyA*feKMLp2wtR;FuabG-jksWch>}-Q`$2JmUC4*|
z9csNv>Dzgm`!H~se7dFNDGtGB@rr-RtKooo$cr9P>txrN*UB?de9%jY+MguVUj1|5
zS?8U<etKU;@AN>=^eAq_Y3mYityf$XXXx3palF@_{h|EKxYpg4^96GUADX$eVH|Ro
z$MO^T9!kE(V^8{tlkC|1;y(W9o&NL}FZda~q56{_DvpR3;+t`NN6k;4>{5I3mXFe-
zx;wq0sej|if7mBEqSi%@@^<~;pzr$8FS&WPZ}hAmedW095e|so{L8a>@ry6t(Y8Ge
z+n2aP&-n9RKmJWl#wU*w)em3R=YD|tiTh<GD&F~>DtU;T?isijqb_aV=y!P9;U6E?
zO|I?*c<;UrIg6*pwJvhRpZJ5CPdoAzukj_X6VJs-?fE(TF(3Z<k9pV|dH9Zsx8jp^
z!Xx(S9Ev~r4(IqQe>5Jdy?quuP&Zc(=hx=34snSb{XGPD$bQI&JlLnd4<w$uZ$duw
zM4xa`{;IwEQ_k_(hxhpKo}P@)AMnAi=$T*ee>ft4W(WMsckA}uI?Rh2m%Qm0KKSl_
z5K3?Kh}!3%@-fst%((0vf9w#o9u!~V1AfIte26#X#IBq_>M#EJ?peGL7u6BOP0!|I
z@A?^!|F8%97kYR8pibxS_~HY<{LQ$N5BbmA5B}~G4f!p7*w6Dj{HP;nrygK^#$n&=
zkzMmo?Ztn8m%+IBh6AX513QJ+@&xvm?X>4NaF`#nZ_oHf*&RLb3;1Dt{oybFW1rS*
zKK2S%*e6O({Fyzo7jj}Z&iCbw<RXuOOZ?_nw_UVq=!RV?axp%>%wv7lZ(ZasUU()4
zeujT?v9B`@KVp~GWn6eoALP%T;X1#vf5Jb0@L^tZ_Fa4YpyVRYlP@M^SM+T@@6E%W
z=)ro)%Rcw`cl>7hy{k4gM?SsBsVgt~W7EEvT-dpMhWyo+ytkiIe{!!Fwf}R!*|U0)
z{_;3_v_5v>@73VXeKc~D7ohaW?@@XXU)ZtvQ1R0<J7wpd`5pNiS3OsI_GLWr8m@}N
z;(>LsTlnMuZNod%IP6ZIA&*MG!-sjufqmg$KkGyJ1wVnK_O0sZ{3R(|FuwUqR2)V5
zBRhvv{Li@7NniBinIFPUe!{PegBqWF;Rm}DH^~Xr9zX0x{>=W(tDk4%>Tf)Ju`lwG
zkFzWD@gs3lJY|p0o7~SPXSl*X*(JFrWtaR8&f$YSTaSDQ|Llt#__OuWqxa<DJ-;)L
zJi|QVpJ(f|K7Qmo+_0}_cjTguqt284@I?;f!Y|0ry7>V)+LwIjXKU|1eB7pnIBZ<&
z<d4?DkJ+_$)@wc9TbJ+hR^y@Ol`py<VSVg_+}zW_xBQRX$RqhDALEge{^TWIdXLIq
z;hb@-m;TJpZseQn#JbIoPwPSXsrUqc**)B0ALc7j_*6$r{X2b64oS(OM8)G06;DbO
zzWMvo;=l7Dd4l~3dCEWFE`KE_IKeKxhc9r^K3knnKlb20oO4EY?0%?c@z1`C->^61
zIiFV7b*|>`LfQB5d-~vy-m_=(v1flLh+O3h&Uu`N`a7iZI_DtbhP+E1ihSTTI~GUC
zS3Q7S)t}V6#94j<FXWr-#C&ke?_J18oU*@>Pv8$<en$kS=v~~?4_}_O7gxo7zYAlZ
z?9O-gLq67rkM?-WmX9ACds<Wcf*009j^Y|zrgw7DPrUH^dG!r;==W>Jfot%eJn#>n
zj8A_hYF~hF`r^O*n!l=x(HETe`v&`yUrZgp{_Q{6*fhR#Z2JcGiZA;aa)T3jj?Uij
zWxvhtjmJ;CXSe*td7<+z{rCfWBTsrVj=Y_}vOjsHx*@w~NBoQZ@N@YR`)8NVUCF^d
z%()!-!g1r_OB~^s>dy3Mo%o?=|6f3R_Dw(dW2fv}-oifNsC+^Em0yVe@&R}$Pq8ma
zytH0ad^Zl<#s|H^WpNQM^Dll*9=_v8KA@d-zzy<XH{?PN>dyA}zH5(P)H>uve#hWF
zxsW?5-okNy!an(dxZ<2aJJh=140~ela3?AIgU9TMT;VZ)67R{w`o%~72M_6yAMj`R
zAYXu|^rfHg?1Fy9C)9lOj$i%733(<vvJUT!PjB?E|NpRe-{H1a)!Fb59N^GIG!z9D
z^%WHfcA}zYZLCPML{UHm1Qjt#P%Ob1q9lq$G_fm2B%mf@1BoJ*wE+VH_L5LU5s|7$
zIdp;Vd#{(@{;v1t8sB)<+It@``p>?vG3%IPjyC7>tY?c0#>E#q<sa;ae&lX_=Fy+r
z*sFCI$GFB(ZXM#fe&(gG^nsmNhjGyC6Y_KTH7*o~=o!CbK6XYv__2=U*Er$<{l+)>
zLGmLXDBqyR)~P*tTZeM|vP*HCexs*FawZ@0$a4IUyY;g-{tUnB*$sc4NS^Kw{XS9L
z^1U*06Hn|PzXZiK-z$LPwfj(cg?mi>J;xQ-jpx2bej^{0Pmmvfg+Jd};1`mAdd1)R
zzJxrTpEWK$k{5{2<imezM}P6h53v*eTRU=)=jJ^4^_>!anxAL?%H@;RubiIY3k{9q
zUY33Eb7<p<1N4sF;fp`#U+JYh%l((<f%I8_a*&6S4?f77AF*HLOMl2u-Yj3TZfG6s
zPJBu~V;$^{UtquUMeCziXnHi%@6a#n!>@7hP44nJd5LoE;5xTHZo_qdyz;Dbu>aTW
zchP<i?f292V83r>@17&cV||ywb8>Nz{i5lC@A1e_?7#b0`2c&6Z?j+j&yHV8zvF#8
z_s95<$MU22r*EF~pxL+f`a4${pZ}qEkUui6an$R_&l*pA^W(?7{3zPE<Sd_n)~TKN
znDx#<@>X$0IZVCK9^cld+<QFwJD;NEm1z5?o&NL=swcOk$(Nk)WBuey4&+K+^v1d%
zJ+Lq4!3WxU%qPEs<~JTbvRwaafAMKt_Xqm%qvEW2?34NRH;;b&ne~zbxmvGt0)3WG
z7#GSH%;O#bKM&jeGi(0kvXy3Ex5m=9zVGx^W}kliKfdJR!|!rCG#`EOonk1D;U~0{
zKk;kwCiV2q{>fk1kM+xM<Rjz?-7D%xPTG^NeUa~(Up%Cb<joJ$XU}iMKmODGiE{vd
zYJPgdud0_<8%JCe&%`l$YF{h#{RF?O5ceUz^~aw)0)JKhBK~{7Ngg6@7+-#5oyz4~
z<OJ25-#xp1VHfx`pZrE#qA&K}y%m3>UcMv`R!{%f2YuoP={G%+-^wG{89PsOzB0af
z=`Y&;q#t0<_{0}KB<>rB+>9f?mfzYR`J49Et(^YIljKM85PpT-;D@}}H@P{_8qf1F
z`L25fwDTQ%p<i=(^KJhhMmu^ckA~(`Zyox{kL1DlwZHOO^Wum8${*>4a{;-c*>#TV
zeBqqn9LwJHhx`t{JU>VCE858ieZNJ%>ODF4JN$=o-$8ZnXI$^mImdX8WL)Pla>56s
zuk4XMvm5r}dp`7D{6LF8;-L2QjNf2q=H+kf7k<f~oR!Nn#RL7x!~T;qJCbknfAVqm
zNY6Ys;YZjPz2w)7XP@b>xJG}?Gq?VK-gV<+Zu#wqU9k(lv!j>BV;}UJKXC35r?htu
z?VRLZklvu7b0#?}=MV5_UGCfSUYtJDTXCG9BX{F@&Pgux(?0Nv_JRNN{dVs`$ScHc
z{5u~iw@>VXJmeqZ7ro~{<qa7R#7F$nU-6E=w_o(xIQ9$w<ffe8AwPORj>^dm&93ML
ze@1>uE9ZCU75_gevI~Bd9e94s{^cF?ULNC~g<g;&yQ7EvvGv#o^P}wxe<1##y%&rR
z^5aj~qxX{4+gJIU^`hk$<V(-w1?rvS<d^&{JEjN5Q9p;b{T`D2;>-Tif8)`A_CX%v
zF@3Ur>s1fUgD>N<Gxp_NW}oP(bs0~8`aie6+kVeDSAN@i#TD~wk6-q}?#xRc$eaC%
zd;FH?&h`oN5A+MNYv&>QN-yNS_%uKH=&#&3&^q`-d}qBpM?HPBF7p}3eB{O7krU(>
z64|kQj6BGjJ|sHdSQq`pr}pZNhtE6*nxDPWuha)|lwZqufe&_pCTIFaUp+515Bbn*
z`%i!IPk-5|dGHVIhxX>>ul3^>ts6~_>=hs6ZytG*Jd*y<7ktxS{c?Tg=SR$sCU5xz
zKLq6?_>=cJZ^+N+o%~8Wd5ic#Ui`cD;X~ZzpU9Ow=n4NRzb0Sx#^*P^-{twF@6gZ}
z`)(Zaux{^>c^^jJr9C|pFO5$=+VR8IjUVgqoiE=5^Zj}J_%4KY@-BI|@3@+eyzws&
zl^>dy{*bGA$X$DJ*SyvveycYgKY<^5<@>z&Ay;zY|K+9ZggnJj_C{aq1HS2j^OSn!
z>8H^6qL1<>{Fsj%tjj*}KkR{h=;xfsf3XvOPTnLxw|;u6zx<P)I`8re<f9)u^j;9Z
z0-ewJ8GcrLvY*o;e=-!EYkkMxcT?PVyH9n_RqpwLd{rJOuFG$seRu!myrbMblXxW#
zL;l!)rXRK+;-`Bm<Jot5Lr-(x?W6q?NA#md8Sl{ImbhSiam_l^vp4r1?9_WsP`Nz7
z`}X{RdH4bR@*nJiUCYnx7kNN=AN@3*dgJnc{5Aggeg6;I_h@$B=f8eA`{0%4zW2ah
z$PMyO&X?{%=%449-kbEi)Aydyo@cQ;c`BsGo;SEZbie5RY<808<h*}$e{5ZzBjDTf
zFwZ&Jk8<}*dp+`<|M{cWuQH3aAHMJJ{Z98nXzkJLz<G>Z{GZl6f3@z+=T=^MmfSqw
z;RlZ0`MBesb=oas=y@{#?RT)^CI2i=(L4Ii-uX9n!;iY35?}Za`sW^A{v%$%%n#%V
zzP|;<aqAGT%!eO&9J>*3?60`TuK6LfcI?x-=GJ-JcOLi$=LvR6pVae9{4M)X4&CFq
zw_(TWPsKHPt9AN6ypX?u{53tJpX`C(BnRg-_oVFE`qZ-r`-?CBo*#tbv3N^<?9DmW
z`uPF;iWlT&zr}y!%Kz|5Pw26E`L9I$(_?XrKgEZ<fM3sX@NFD&hWK+1&h^_LaxlL2
z(+~a^UugD6KgmTua>l>+k=Q}5+c}3H;4j2$@mpMGNAe|dH81|~P5+_%LwnDO$cbF-
zJ3EpG+IQy)@s?k6KI)Jku>bVUIP@RVOZvf1#3k{={Tg)N!EZSSlCOA8FZdbrkcau~
z52P3Dz`E><yuiNmx6UEPrML7Es>g?YLu*H$?c<z1E?ai`txvjcEDzz2=r29u*Z2+b
zGRGw!==^M5@*?BYUw+kk?2CEr<Hb82aPA8~c$XP^KyKa##UFbyE`1aS?Z0!Sb>h?a
zxaFNg&kI#MaW9d6(5FOxqe5~@v@eicq5Xl@4|6{H0__{S8;bt#gx`T*^MQZ<+~bbF
zW#m0NNN@a}?&{ARe9Io6TxrJtm5|5K1Lt7*mAqSiV_)5Cuxos4$NsIK9s16N^EkUT
zo_h@U9G>T~C(l>;Gx@*Y?K;29v+yB~%Om~Y7;<#4B%f4HzVzF@igOXU`YtRx;ji({
zpWAnSUVCUC$eI1g{~*6e{^FMYP`!9&{o;n-NwOz;4!tMDZ#c)WH+~SN{kV5_elU*w
z!nuhbf${`#)A_=A%KZGl`HffkkE)#fJLJdc75hT-i}X`oY2Vlb{*|NU75oCf?3^I}
ziAVa$%lIApYd!jjfA$GK_Md(lPq}zVpYa9L-sqM4M}Ecc#K?_ZDt8Y7#T&HWSNXj#
zefGOG<M_Wa{2;rfCs4fe{+@HK^SSo`wRbLce)ay}fj@rcLF;_?=CS(%=XvLSD1V~Y
zeh(oJ@V*DT6NlBa19_C^zRox720FK~6Zs5(P0!@l@@M7V&k*07gVeKY=QHO4?dcmm
z<iGiKcEWBTeU;bqcl<NG5--??^R{;El^>KhiBtTnc*NgoCl1j&@y2-itEboUQ~I0j
z{I1=8nAds2{;*GZ9=nA6y8Xs4JBIuOJ+LqQ2>x>2*2zxoL$1?!;-dUPd;9D@oZq8&
z&S&;fd~|PVedOd^A<oHD?4x{xeDot<H2o%D{@Ohf`S5e}(m3K3yA<#ESMBLB|7joi
zY5cG&@?<yUV%_8k`8WG-9mcgjNPn!4oRTJAd7F8x%Q*ZaeWOp>$%p8@dPtx6Z}CDu
z_5krqKGw^g$b}xNr;qw8w;%Qm&HtkHmuD&$CzTu5y4+Kc2Ya+X`tcj$oA;jiUG|87
z`amDa(|w!g2<{g=|B`>{=RG3NU!0%hd*Yk8C~m122i$+q2m1xtF+HT$^j+SGW)I}7
zzk2zOa&omUs6NNX7d>L<_5pt#W?pH(>5KQO<caFdqdf6%k3He#haSJ$jC~bX#aZj;
zca*aO>!EjOcFxb4mp>BE#4pc<$sOPHQ8~M%|MbuEPkM+S{+C|KkLZ(gHb0B@J^*BY
z@@Gij`6c$kZ$bCu+R1nM1A6T~**=I{+RL{*zp+2!gZz>GxhE7C+@G;`<5>q9pXBE`
zl6=s;y>YCQJ-Aor-}Sd%>$I=tV+ZVs+~g_Fq1NMG(LJaB?Ap5M2YX~U(76wv?A*BW
z2y*4m=#71l-};V;^CEvH@0QQXv+cKW_(%Ha`J3}C`9bo=zjKEDVkhKmT=HNy^o%~D
z`BUcr<@m*i{pAPEZ{6ng-mdoMCvW>k&h&xasi&Xplm5~h?eJ?~=?^<*w@^Hw&rm)6
z;^)xTN3YF`f8*Lm`%GTO5r3^qIU4`uZ(jL?b+H@y3Oj|q>j~*MeG{+gsr?rZ?1%eb
z`)$42+h6OXU-SxE7yihdo>;Gax1ZX}>-D4WiR2~klppa2?43OupMSxpd7ydH9>^*6
z!Fh#Wv+wdc_Z-eg@>lnoo>#kP^`1R@Cui|i{Dk5@|Hls{zx*71&|V&<y*SI>`8j+T
zpI(p$f1q4@dT3nl_lV!d!v{SyF1@ke^p>8x_n?>dPoAS6G>`W5k9}yzuHCno7k}<+
z%meYwkK@ny;t4spKO!&tME|VQJrcd<x7?Gs=b&fuFZLndgY=(WK=~hg<_Fma|3H7l
zDer^uL*#6Kt<Qcc=Ld2={CS_wds^(vI`GB*l*=FGnc^fr%1_8|oCDdJ`vT`#_9}kx
zBgQd)_UA{*RXhB8&SD+-HxD`JhktoFe@ZX;1^Q<FxeoU@&^Z&zGsyvpU-+?)_D_6-
z?92H{Jma^VbKDQR_kq^OuIQ=#O_Xmy<3RQW@dw!zv=7iZD9>|$q^FYr{c{iNeu{n3
zbN0cW_yPJvALUu@h4@2u=Kn)_50HO^`~hTd_MaWoC;mwuKp)wyag@_{G`qkzd$oS&
zYyH?KbRSM0&PnW-e{v2M7qXpv+dZT>$X^@JI^EOrWA@2C27l@v-Fqec6My1<fPW${
z>(2a0o~NDtuuk!W{nHzK+c$oU-1#f_$>N$k-18rCSX`Dz$Xn&f&h7k_JjnS*9G9=~
zZ~UFOORwpZyh%R79{4N$#9w&~R4<;Q$;UpE8~?7I`N-Ki=%an%PuY+AcKSi@-5bf5
zJ;$RT^nzb7zV8Anr`Pn?`q_>BU4FyTzkmBitIV($>#+{cC&-I_@%!|WU!muT<YE2R
zN8c;tXRJHNWpC_Pd-=C>seDXcD(|P)+RMxA7rQ{yEB*nkoqUj<&}Z^vhxFRL7yEGk
zV?1_cU)Yy?T3*AB=q0;$-m~BQEc;>4_LUvD52qK_&wkPFBk<?G!@T$~KNP3*GajUW
z+FPIT@ej#?pSGX!Fze($*ef}ci+y4L%C$2N{=KIHr;GZ3IK9LdanyaV^F95ikM7sm
zx&4&i^UKZ|{D^Y{zhoczOGxtkf%B;E9C}{kd<U(A{?Z5jO+Edz9(>S4{)+$SFWh_a
z5A@kQ<R|{|6a0vA*aN?y9`a-A#Zh^jdsypq|AA&N>=Tk}qWwvf_YOtzEs?*nzw$@&
z<*(R3|6)J*H+m&rkf(DR|HMzoC+(;Fj=!?++S7A#m#^U$$`Ac+P&}vK?1G(GhrA6R
z;w^N)=DiPc7th&~`#0^~-#P!=N6)GCbFPrT$RoUW=sPm<SO2fV`-9etKjr@4oB53I
zeZ!BQw#_vU+2Z<<_Yw1c5r4|%)82!zUU`)L_I(EWMIP2~|Gi&D?$&D`&7UZKR7f8Z
z`9b#2&(M!V{*xbbpU2<0=ad)GfAN(+;E(tLdd>d%7x|IABKeosy0>#qweR!;pX?Bd
z%N@#B6WL{lDaVTL#!=7DPKxqB`|h0N{?vPN;)4AVN9;3w5%1jp*yrR+oRHT#*VzYr
zITxDWxkLVIKIncFU*Z+oxY`>}IsHr&AC;r&3B6R#F3gXH?7?$s_t+4>6_OJ<kfZk^
zA%3C!**OM(#!+q^_E9<X`!2s5KK_i0_TK&xE6qZ`*Yw_sd3=Aw{XR6V_dy~3vrlMx
z1-1A8H2j{8JjvPp0)F@(>y;0sewzPN%m3;#|9I|_*<6=%2EXOp$1dnMe(1A(+~u9G
zymjk8zkcjIM~|&1^B#2izls*u62+Ge#or1~-2d<$pSsD?nMChTRhaEN+Pxxt+)h7#
z-65}7b!I5~zYrbDM_|=nelrw5cFHk7Ir$AM&m?+Zx<h#l6yGZJUUr4@o(ijRhnD*t
zI6HL?VsFkf;-9>FPLFT<{-QYNe868hPl;QQpVdwtV;p%oduPA=0{`hfa(<58PK*2j
ze@!mpfpY@+kc&7ocRso8_r%tTf9=rXnsV*Qojl1+oW>WwCw`#i3;Z^kyp4+w>rQlD
zvJdhLh=2Y3Uj07r*mB3Czqn-9dB^>d`xJ8I$K{XYAYMDSk+*v=afUvapWmGp$=&+s
zBYERH<%U1<AZL1HU-k38f%gUU*A7|-Ju|=Y@heWT8|OlHrJZtlxbuzu#h<)_T*-r;
zId7+((NA%d{%J=K>8bgQ>-|RM)`frTAqRZ%N6vTdV?39!FY;pj=#PCO596CZ^%8&N
z%}$iN7jVvZZr2`dU(HM2^uhjHe}&F*6<RNJ9)$csqUVxT{PcXmy4bDz{|@;}_YuZ{
z?*GjL=_`5gkC5LXSNDR@Ip6s1Q=Av=5B?xN@xcy^Cm(U2;6A`TA#^X}IR`nJhyJir
z=pI3P`sLi5h)?aw1LkvIa=?fCjtZ@ty&1Pc@^U}oUd25V`LJ6^PUOJe(D;y__`Rup
zOq3tV-{lj|Dei~)FXvI`Zg%5-lwb4Sl78X_e~)&b!rsJ1{z<;UZ+U)i|Gl?t9C6Bi
zyRYOA>51p%{H5~{KJ*iR<b&kuIgkCepWeT6e)0TXJ$-kMA%FU6f87Jo5BF;F1kX$H
zlm5oN6TR^Jdh2Hg#$jjd8rlzf<#&?&1^Xs<_vhx3w|GuzUh)>l?5F!l_nyX2yAux}
zy>%XyZ_{h<(|_e2dtScZCzj6AS9(al<W=IJ_$<G6J{Mn%Bd&{E#vwm`THJx+s(p2j
zXg~RJ<BQYsNcj)D7I*2Vdl}>LEBqRLvw!Z->5cnI<FSA8b#JRbd8lWX<N@V3?#<GV
zc+cMbr~eluAD}n#M$ez;?Dd}x*!5jM|HjC=J#TZ*%^uk&{`r0W!}`TnNbc<4IP_ZF
zh3tZVanE8rabCPIzy0T@^d|@Fh3wvV^o5?$bNOi^eYbD)6d%UX&-na4yU`z?>;`SW
z<h$%9QG6p8=OcM_hxD*R{E?6MCh*IS@XcQN0ngX?6LQ19cp}f`59y72ICj9V`yMFd
z?-J8L9=+OWpZxfTt{%-7*+qrofcQ|M=U32kG05MG?|#?lKFf14_cZ#MU;LEs$mhgi
zaq9=SeB>X_yJg9Y{MYwTKL0Nde$U^ZvGR=XbiC@ECm;Roy;hr1p19$^UHQA!Pg-T>
zhudwi)@rX^I^$d*kHcq$zWatRaW3yIt>eK5?){b1)?al7|KkTe;IvJSUNS>YiPq)5
zjGW{h74lQ!HvcNFuxsaHevo}=&)(Ppzb}sQW8R08U$Hymxd+#ue}VLp-(tr<S@y4w
z_~BpOI^Oy*|LfF0-F2l|-&=AYN$%cD6i@6QdGVv-oP3kL^8@^o=hX5`@j#v>-_wtL
zAU-@lmzRm}&~sP(kq27-g>Uo8zr2ShzKI*|Ma(ZBf%eVw8T{&JesPWevrp+i^)o&>
z@{{<I?|9!-x$*JqzQX?SyZpHLWuKg@jE{eL1AWA|`;W8}>oX4d%cJ;n?>*WF--(cS
z$x9M_ufVz5cZlQ#*2xblhxoBxdW>&+BA=mO<SHMbkN7w4P|UdXNBaMkfc=pduy=W+
z{i3JxK6w<ovro>Oo?pt_AivIT*av?K={<ihPcjbsaQ;H`$G1Dj|Lo0goBhCbW8*@3
zCbVu+5<i>^*e5>aU+f)<XU4N`@`9dgLh($VOYhl>^PTq2N#qaZb>f)wsB@0rf7vI`
z5B&bvb6nr)Rqp&G-qHhhOYY*SxNTp^hhH~dhxW(1$Y1_0p3xV23g_bDZU0Zi{PKI}
zWpRr?vVZi>yzZOO^uYYqDc-Oj{?2;rFa5~&<m=olPO(>hP`UWTZ<CAh*|YxI^V`<J
zKE)9<{;fyc!iRbA<J@mu^j_RGKiYiw#;5&Jo=BeN=P$@lJ(O3XA-_Ox<T?1p2R})k
zP&qrKKk^oFjXdUy`0kLM&`<VAPRj9Xzwu43P`P%A#$&JMXGf48@wYJZio7SGZ}J`J
z{2?Evf7;7KoQK_?8b9+q`)HqxD{jiK<k@~t==qIvmU8b;I=70K?pu{xkLM!Zvm_V!
z7dykZePCDO1^+BgiVO5xym4O+#Xs$gORoG<o=2T~V7_PJoU9%#KXoor?m30>L~_Ot
zBu_L{4#|}rL;lYGL*@Lpdro<@d6a9%Zj7THq=&}QUpYRj_S!+?n4kYqZ#?^_J^f=h
z{F?SqJN{c7F<<smP7eC91J57a&wDO_FXv?KjISL#q@T*|GduHq!n)}byVZUux-Sx!
z=rMljIeyH`zM*>_??JndvR?NG+Ov;D_bukLF86lEp-1+Ue%MFttw%Yz&=1I;i<k7A
zA7xke(|)OUPk=A$mY12wxb`2qUl89tH<ahff93Pgy=vAwUl_;xFW&R?-ex|p@m#`v
zlX&ZX$@2*D5sI(w6PyR|<++poo_FHQ`NVfGjH?_U@;v#vJYAkI|MxsA`8K}irs9Nr
zSiBcEl+z34=*%P4%fsYl{1*L@udq9Mz(4Z?^hUXU;w8Duqs#~CJCp~~U+7*`9#bK{
z#8YTo>n9iSc~W%0?fpsj(EJa(N0Tc*z+XGhkiUJCuS0&xJp3&`<hcUB$j*%m?U(o>
z-x3e)gL3<>KiW9jS)ct<FCKYrXZ_@ZfBwiitk?RD=e`fRZ<B8tALcmnJ9#R<X?*R;
zLp$|;=O^EV*3S-{ujG-=z4VbkHy;1*Uex^FAFyBiF!Vbt`eYpZl7s!T5AHYc&o1z3
zJpHW4@8adf%FP3_+&J#f$-#W~0bgkH!#5h=`0zV-{p3IH!JL~spW~<Kr+aYs7vxC)
zjDtV>Y+UCq@r8dCU&UeaV>kAhKeYd7{*zy@e&gAHXgu;Uo_&$8xc9P8_FH>;uiW_f
zrtkF7d5v7;5&Q|{ADs6*$EUCOcP?|^B9E+4e|Ba6?Q6zOb|qfhKflkgPiXQN$HZmv
z(SD1^^vm;f_2$73^m_v5IdQ{tNBr7H`$Asyf<1_5_~7@<2jw~9A${i$te2jPqx`n_
z0gWd=_FR`9;G4WXcX5A+e|*q;_e{pOzxcqvap^PtX5Zd-5oheD=T+ose5jmU$;bVs
zJV&1Eeir}uQtrMO|Kk6HkGT2A+iw25F}~#0o)?)<ehTH|Xz%&aPkAZChxaWqf0d7V
zPsjgrleeDsuP@&8F8{F7EdOnt?#<~r{qcT``!IaTtMO%h@^a(xyX?@o<it<9hfz-N
z@k7q^h+f!Fc5FQJh%5XE|HdEk_v*zdd4ziW(0BSTUdh+km3u8H-z7Ilzo7L%dJNea
z#5c5V$e%&+5sE{QKZ4Fb(7A>`f%Jpla*qb>8-Ji5<fowWMDk9gKjiBCX5I9VUb0`$
zYv>dGqj&BJ6U|32>79KdU-`D*1Gzth;t4(=IeDH!e#RY&@=o#8`t@fw?3X{FKdJxx
zt^VZYxt{jqz;5JG{E7MO5C32s_3Vv&$&38(VZZou`oZ4uX`k)Ce&k|Y{lyJ=E;}=i
zal|X@hUUdTyEUJ2>;w5}Z{IUMu@mj-r+Mioy|<s_X+DTgXy5ob=QVzUe{}wFK1d{A
za<^Xljz9T2KY}lOJ8zORx!Wi6up4@8T=Vm5@(lS6`<SEOve$ku{+Fvp@>}mi`%Z)R
z#^lrF0e^Y_Lr*>Bl%;e3&sJUtSNp*MGuym+mAUuG-u%+h`7eLUDl;!!<vW+Jb+^@K
z<jsBu>z-5|CO>okFW(f0+#}$}J*eja@+flEUw$ucxv#K(?cIBlBXnP_UVHPpzxF)J
zxb9(`59|ZJ@hd;YulMrIXFte^Je+sL6Xzaie)?-3=MwTJSA5w|?XAnYjqCTjX#KQL
zxwtPgkN4!9!|;h8aYnwYp8w|um6NA=l|%dfz+dh6xxd|F$qf0Tp?2PvG#<X`r*isb
z9r7h|q386-`1IJmv%}OY^FaDUP8He*Xn)AHLir`+NBB{Gi9en5XQOwowf*52&Ha8q
z<AHOh_E5YNANW7^$8Ml~=TG&=pSYpiJlZ=akRz0zCbBzv$WQb0_)^b5^UL;`-WuPy
z%H0ziM>#u(%IO`XcgC&I`1}Ap5C{2b`%d2K=`p#nJATP?B>RYO{sxkh_)cDl>=%+J
zwEhZH4mmzPAbz0rL-`J5&xz!eeAy?+J`?rhA1ai0$Rp)RkX_hMc>_JP4)#pH<u&wM
zd*@{+K53Uot}y+TyoSD*M_hCMQce#cKQR~I$M=8QqrQ0Mn7^@J=X(Cj^D_Cr{o;r5
zOYY(yIkFq$%k!*<9E`(XKyo&Y^RhTX{_aJb%k-BQ`hS1+pWmYw<N%FhKJ6g6xR+7h
zq46v1-uqK;eHGe2NX|pCT1Wbcr1{MX>F<1zeu)RxWn6lX5Ai@d_3T<a6!+*W|3;3`
zd6xX?A>^0XqjMzxXk2#Syk~rI9zST~vwO64K<%9u(d342&sB`$cgOVJ_mO;e$h|Fo
z-G6!x?|V+3&(kA*MZff;?2%pI8^7#Ze{v%q{@cEZ+sfTfqs2vfL64qv^^d;s&rex*
z#{G-;!>o&b`JI9H8t8-XT6qt``x4#{@cw}Hspp^U1HVP??uXoO*&prP(~&oSt)2HS
z+)sP1Ab)ip@|?_mLgUa6dgHwh_u2TcPWRIG)&AK}epX!ISLvC!hHrXA&y9z_MDMeZ
zulsiQ<lgVF&(>pK+@G7*J)?Q?ZJpYaw|%ES^wqrzeX{T5;(p(~x_#!)_+9<^7v~9b
zq-XBctQ)`PrN8{3`z7~s_M3dITl{n1>z)kX<Zk_bXC^Pghw<qfIk<QJ^|QAB*t`C8
z<yrI4H*zr^f8qWL|JLU|(!C=-a~|W{Kl@LA_&4u4*%$lEkMnEJRs1kNME>*{|MDJw
zj=vPQ<U{hN3Y|L^3!P^meT1G%TPL(G`L}(s?|whXp6H$ZhU}ZaV4ud}hqPx`p66+&
zKf9*a{FA(qUD#LX{y6uY-r0ZpOW)}Od!`5UmVC^sy|{y4_K08lK_9K3U!%|D%1`Le
zp6H!={+(a89`^^(IQ&k9<Ru@0`a}96jzImP`60gPopTVBhe2{7S7<!rkhguO|Ljm+
z;k{b+$G*i~`(xidpOmlgr|il8+XsGLIlsXU)bls|hw<dY_{NWVs2{y@&mg~*Z|Tn;
zk{ABa_K#fC-pnhX!=H7a<!$a+@*WYt-Xn0Yp+C7p=UZrf_<-~h+AoNIC@+KffO+1h
zpB0i9ejquMf9AR5r=9iLH+H0)f3yGOf^T~6K9;{n(|h+g^k!1TANkTZ>xATufBI&h
z>^nb0Zq8+i=F9y_n*JryzY58xLVQ-}-f$?g_eA<oq5H}T#m_|N7WYCGy3eUF@BNat
z&mHFdBU=20_^kL*PS4%%bx8l_i|iWmgOg(AU%K*K=cMyydw!)tcItVQ`*+{ncmK$b
zi-Vp=`F%&)8UMxanTMW>$MnIyj_(SmU3#ub55;@W-^G1-m3sb!pXFcam-p8B5q4^R
z`3t+ow|$(;Pj358uJ-($^&1yo{Aso~u6W|S#Si0)-?lFMBCm0erM<jLKm2Jo6wQMl
zdSQQ|d{2G=o%5hP3W~q@kw>dHE|f1r`44nHfZ`YZR8DUyv|rFVp?!qp4(X3`xc8Tw
zqud`r=U=~{cb?bZy^VV#@^x<WJ0kK^E`NpM3b{L<pw&b74(4~S1NAdLlvktigFku<
z?KiZ3=-dM7bs{~5@})%2w-for%71!)+<6Fke*t=)Ut#)f{t)5=W<G>J@AL4hP`+i|
z;<J2D+%yh9&A(+_N;>_&_Y|OYRLCy*7ygp}7LTEF@-#nrXWr($>_mA<#aEV-*H9$?
z3dL6_E+txDh2lmceRf{<d!o$qyf5NAWBw0-_wfB*^q5PJ-0;(*RcBWD`&&=A|BWln
z_+FQ|?frc35&7;|qW5<b$$@^;L;qjsfp5L_%!i+G<4AsD9o{by&#(T>!ME)3$(3fj
zS7!ajV@G~Z>pQBx*XH;0-WM<)KCIL4XYpa*%p*^7ex7^(>}wbAGkVLC8FKR-HowEQ
zPWwWh`0@S{WC!Q}<Ze$o;yEkL{U3-8c76RDe!0WSGbf$6(U13e@2WGqeEHe0xbUM(
zXV{6niu}k0?R!jU^3v}4uX@$e2mkEak$v<0+WUHsU1P5+_PS=}8S<iE)??lDgMQfu
z@-{CyZ}`*$AN1ATZXH=aeI!3}H4fC?zM<h(NA7ad=g(a>bI&JTyzIH3USsCMuO0Nm
zyM2E38Re_La=-1~yVDvoN9_J@=dH8*vKjj3|88o3>|-|hm+yUX+1&ew+FLI@U}xmD
z?G`&da=WjuIs?D>_aC|cH4j>I?mK+^vEK`cm;9K0;Q#n(znf6cFL)o5{@CX^{_eKL
zUtjmNn@8-N9JRNv|MJ&=_N<RBT{2^T$x}P}#y-sB_W|-F_Ch}(`y>CiY=6va&iwdV
zGx)^E8pr?o`^TTY){O5Q(kI^;g!u6uq5V1hPqx_glFL?^F`sdLcaT5v|5>e*T-Yr;
z^Zu!MyqC(adT*6oTd)4)hadIqCH?0g8ZAF`Z|}K-eA7L%=M3^>`K#xN@>1yhfp)%t
z@>cg>o_m?s`0{4?y!_p`_>niu+uiTTlan9y#_{~md+naD;S1m7r5(BXofJMi57*xP
zDtT!~-p-fiF|OzR#<3311tEL%{(Jf<d5`>+zi^J@2l>4_JG%e%eB3$FdB!=^{eg1N
z^E~$<U(er+=U&8pfPC8dh@9O=c+Tw{;9To@y8UHO_E|gotKRc9&*QYSZuc$nRrC9u
zt^Fex&num;oagAb`x1H~PP!-X{2hPhb?$PH;XZ&o<@w$h5ohU}^N2W$Z@)K&?ltX;
z`Nco`>-`ez^gaxG<WIZ@^Zt9^_}kOpv1B&K(T{)Om)LRo73XTtZ}sy&i*rAHa}TGS
z-|{@vd51pPN9(gc<m-8^{UcZYfqw8i@*n-(lksE9%}?I+(LQ_rYn|G&Gx|f{$W=Vy
zXWcis*AW-I&p=-8SJ<KZB>V6E2>V1|*qiek{@E|&AD#Ez^U2fAD}Q&6hU}a@S%-e?
zn0}g<eC@mQoAa}KSN36l<URW1-~Jn)eUcaZThaNB-<`YnaZbGT@CR&r&%d~0M6Z3H
zoLt2-$RCiGbCdgb^3$JO$kY7tY2Pz~zAM4c^K<0yJ0ad%){orL&auwh<jxNom%h+T
z=sC2wKPl2<dIyaQl_%npe>5IBL;3$R_jvL)zc}=+GkYCz;$!~g*Xzz4dE>_}+j*O{
z=f0PI%P(L1oa0_`x0!#x`J<a{@c6sW{Ke_pzUet1T6Yee@0?TK{f<XGV8^wW&7D)6
z^L)?2dBJlO{J1xxSL*47_X_EU?|{=U=LY-cd;-}Uze#WTA#$Q0;uv}R9)$O4wD+DP
ze=44PehuXn;<9)ouU0Qlpm+2PAN+y&_$hqQbAFJ%dY{j@)?;4JljIltm;G>lmVYL)
zXZJqNYxc+g3$UO3h5fgm{Eg@OzH>yc$=AIvdzbH7KYek}s-68H2XYjrJ<kw_=!^Nq
zBYFtMQRg7z+b{l{f9Ds_X~+DY{Tqtzi-w}OW`21%%=>Bir1O_NK>lmJ-iNeK`YX>6
zXXX3!Q@$lHz@L23xXPjN<sFHBhs|E-g?)9uk^3vp^?e(0T;9gd@>lc#(l_?SuIPhu
z{ze=Vx9G8c^q5@d&O5h#f7!a^kNh4x$3Nr;tOH;4*E-n=e`z1&!D#zrf8|&5U3z6a
zcE``rBkQ9N@@(gS?;qkvJLvZ<{I&DBeAs!?`O*7#@?3g}UveTx@=brOopIO?KVp6K
zlN_wedgS%)E!ds)pv@zWkdt+hzx6qnurGNw`w^eaqrG@%U&&cN`^X;YgY$-Q$ldt(
zHIDNSxswO~PG00kU-4}`_Kt6U-8l43y>pK9OCtL-4!JwGi096op7*;i(*AX8EPd<y
zPG4o#d7EB3C+X+h>m2S}=043ic<!EH_s^{Pm&;a~Rqj1`{E&-sa!aJI;sJk2&msS2
zKj<6#qF?ld9<wX`+>g*3`bwYak^SNq?IZsAD}0NW@_YN|{NVn^{KiYvuS37Lhn_Qa
zXgwYJy?mnk_zJUqMU#ts2GU>tmOjx_e#?Hjr-k&>`x4&Y@PF;;qjBi3{@xF9Ut{0o
z9nd{VqWLStcZcS!kUTnMhmhTOxT4Sff1K}*k$`+OQG64p#B+X>-$ILP;;QqE_{GmV
ze~72r@waGsy?#(W&ku_W@@4t6^EMRM-7|_4@*eZbr{t^d5sb@!vtRkIcr31v15|E&
z@q--jBOXC<0RP&FYx?6~9xJbrx0xTxkL)Wx-7Cq5)En1$@(|}Ta<RYKLw*z5Psra}
zxBLZ)3y{1!<mdH+{C$Pi2farK)9+gs%zAN5KJWKOzHdmbkX~N+vwK|ro^RedmN)y|
zvU2gu{-vJC<NUr?T$Jy54uNma8_c&*_=wLw=%4=iO-pAX`>|gBgWjpH;*|W3Tp>Bo
zBYMdXp>tjKOF6wX4n36bp&{=rU*%7=mq*(V$nPb}gZZn(Y`>yudRC$P^bXw@RLGw!
z7Roz^;uc?g-uHI9aGkmP=Knr$>HR+X<h#tsr~O|~=>F09V24w;f8qx|vF6<SM-RE@
zZr@z*mNjO)2Zj&NZQO6l>%Dj5Js$64c@NEdKkmuhyF>3$dEdo*T<Y;>T<=+VFGjm}
zt+&PYFF9!4x%b?(%Xz&I<b5Ld^)LR|$v5nD!<uvd2lWY8{N(R`^6WKdysu;3?(@9|
zmWUtypmllA3tETw)0{&*zk}Xm^&T(zyWe+@Oiu2T-7~w_bFVCpd){OJ>=%@GI7fP~
zOdf(C_uTI5yf-3_dQK|-x^I&wK=a6x#FL@ud?HRlaaEkFP##htKU^VyooGFY?ia;_
z3bm7m4n=-@DAK<~dQzddBJRLk?@&6QuMDM~bM1GYN9m*cQF)8}BjddO@;dp5JVpMZ
z-uD%JH`@CZ54-JtO1>-a6OX)yuUy=Qp8t}A{6CR>$sgox_#+qbKJOuY$HO@CZ~41(
z7Gwu#Xr1y)>+oKe=QC*jNM4cWIQaxUa~{N}JWJjtKI7l>GUK?fb$*feu`~DEp0_!7
zdd}!RTYkvCw1fPz_b(Fh$<NBu_)~G*^Duhi{W<)4K50Jh8R)Mc`_s>R32E=f^Im}W
z9>fvv$$Bpvf5tc8P!wOG^;IZNIB$Ax%YNN2RY+ghf&WL!KG-39*UtFr<<a<X4yIS)
z4}0Wi%;R1U?cCsg+`P)so-^UcJkAI78ETjP@QH8FXSDNtO}u8;>_?nr_u>~j_58zl
z_;8NLxBbH}f8%)xJJS!J_7VT?WAG!da4$n|_)U4fyoi6*jy|(XcH+K=zhY;Z4=HDt
z-k0~j2mSWD0JQhqp?mAO_wx5Ye8;D5^8fv=L*8dxe#Ut6IerB{?A$uppL*+cZjdL+
zJN-^bd*gW@#eCu%zKkoMus_Zl{1V#y)`R9>j6?qFwXdE_IM?9Ia|CiUF8;jN|Kne+
zy}`40U46#5#;K58DkN9u3+EQU2O@8Ie<Ho|yM;vi3h9IA?)05JjW3RnM~C`*?(aPu
z`^oMS*}Z<^6MrJUDi<%^pYRXjz5JMc%P-^y^wPf4e@O4FpI?B+A-C#Y2;c92)NL8@
z`}Tj|OZI-?c3W)s(Ct=w`}>~rzU@~2-+%YM>)v<Yd;RT}{KNb9-td3Z58iI2?N-|1
zrkB6%eS07NjwhWtTDs)LTV1{56IZ!#{N6);@a#3d_s!ekhjuyioPD>rV4Qg4Ub`K;
z!EQerclaNh9&+~UU-EBbD1m?8&EGj}pZ(4s!MXbN_Fw%|7o9L(QsBo<`qEQ=d*-jl
zD+<Xa=X?0k=RI=mi_RWZc*-F+{_eLgzGYOQ^;Wp+ORrny(tBMts&MVi|LqeOfBmbY
z3M;;pKYop`Jm7`<Ts^MvecN96itV;pWp-M8=$424@)eg|F@_bN&s+Jj7d?0LOU4zZ
z-YphQFB9o!VzqC|lh0}&?MH>rU1PT!zW?EsXDU2xm$UwU=XYK;s_?6C`}C7vewQmo
zi8t=D=|i7<+4o0@xz41IJNbgmcX{a-#*<>sQ_(qJMW?({-t4x*lh-@yvpekh<1wuC
zFUza(D*c|cyc)kN&-Evbk6cgE-8@}=&YN`3m-I8vz38PoUUJR2!fc<k{TzzPU+UFR
zdQP7Q#>I@A{6>Xy;}>>2l+UVNHLvn+-{}{eG+vsg!paXQ@A&KL@!esy4_&`X?pa>x
z?LzUN^Hp@}XT{fwDyJ72pOcnvB=Qs8K6UloK4g84M_*wlzbbEd&@mf)YG&{2$BET`
ze(IsCpR(Or*Nj#ae(dWXe(S26TsW$*<Kx@wy!vtXJpbBJB6(Ef+V}M9NvD2SG`*|P
zK2})qsl4i!<?ng%d&gHTTXm+wXFq+@U7zrR)n+QJ&LPTEZ<DU_OXVrwiq3HtiZ1sj
z&KH$^bmeKsNq6h$#;^ESUg>97o_drtzRZ(Y#k(vw&ZL<04W(1w_<6(g-f-$Me|5na
zR(d$JJm;B|X6F^A9(DAj@?GfmPh9Jb|L}uLemsU<{CeB>cYo+k&;H&xai|<S{hqcS
z-{Q%}@BaE%J~X~~L@#n5$Wb{uao6R~JAIW8FCQlkU2o>i=*)A_-MB;TDcj|^yFd7z
zvkzPM^ttmg`9X3^#D9g^zM`xC$>+4R{IEiKMTO)r6svKRPs@k?iK(Ya^PfYJoO55Q
zbNfSHdi1VKzVW;9P&9sr>;P7Bn{Rn_PrCWfo;-T(Lzd1y_Hk!_^r|P^I^O6@fBMy1
z9&r8Kz3Q&J9r@cWF8$@4zg+p?_n+{Z<%f^Sd8=hNo_yh@zZiAM?<7|1A`e)#8(N<G
zJe2P2p|badF8{*wpL^Qpj`;bg!YV#j<@T?`w6n#cGp|cpoJ-7h9i94z?&P3dg?Y}a
z=-iLI*G_t<yi+cd(#h9C(cONiujHz{^P}qRBRVnjgh}ZxU&C*Q=^r{jLrxtI)t7~C
zH&l*$9e(PW&);RG8TTjjl%LQ2nlC+co!!3T<ACEY+j#$@&Kiq@iB+6ktn#7pZ>Suy
zT_=~@)fe~s>@vqsJ3!~XVA7p^Wxf4_?tK;t@jWdLwI6v)t}p4KdQs)OIbPCTd|=lV
zy0@DaJAaw=sdq`!PuQ(*(*3e;F!@89ze9T5p>Zqh`cGOvRPXRJslBAXG0tLPxBhP4
zZhiCZ-;H0@SNfH9k#v`TOj@6Q1)cLwOLuuiCx7*o9F=P)ze!YIy%(mu+86m!hn2sr
z^q3!We@OlnR^uz5bpNw`o`2BUZd%%S6_QgW=PoYfz9&szD(w7cH^2R?kRDc;{x#`J
zFO+xulW&E2@0aw@JbGIF$bS4sh1EK$a&mw91*cwc-qQabCz2!7K2d#z)&485_*1T3
zg`FN&d@0X-KIzInD!a_`%D$D$!z%3bxboA=Q;(7+=Q|swy;XFj7pbQmo#ULd=kFeJ
z%6-?Ih1L8+%jYZKj{hz%$@L}89xCkYjJ{Qv^G!-0b>5E8_~GhHXD98aeJB5nFG*K^
zKzSuM<=y=Ftk66acKT8EQ(n!xqRKlz*R8Ls@5aylPMVyi#i4RM?Ij<7_7Ok--5BO`
zFnPzfA9>(eYu@F`aXuGgf8HxlZ$09Sa_#Wv9#OgZr^U`Mb@lkIaN7OQk6bFO=2gzV
zr^QY$7ptB>cV4RSnLpq1gKyt@?U_73RJ3u2B7L0}<xv&V>k2a;O1it3?bbQ1UJUJ5
zt!rpGJE*XWTU~wUm#cY|SNh-0tG?<#w7lB~d{)R$RM^R<+ehoHu<~olyYV~zCatgR
zrSd<@kN?scU%mO~H<iE7s>W5GcAa!3zfL}#9#!Ki@5b-yjZ<OQzw^7*xC>R@@z>Qa
z6d%<%oqkvSlz0B4TVGXgeD^_>|LMxR_3;xGcKy5gt8tZg<L7-%MW?+_O7kO=qWt#i
zXP&nEeXl*WeDBjbJAJY56&fdT=siXDo0Ptz{Yf`}RbT08+FvF2t~{S_e`)8d*1K_?
zPmf@=@AL<DdRvXt_3P^At3SE^q&xXl`=flQo@!rVr5{zfeSxKaXuh{FRKDg}Xzcc1
zeU-;%xqAwjc7!I^>_1c<+KCs5+yC?5Y_!hk#&L(0{un>8(w{8v^t!9>-t(^PU3sS`
z+D(g<+?3nr3bm`SlTTM)&70-b`srt4Cs*~(FEGccX!f>PSo*2kes4PMyw*Jwt9h#Z
z$nt9cE4eDqagx@q!fc;(xBm1~NvD2hyi7X%Cp!I6MR$HC{a?4f<Ui?Z-?BXA($UUC
z-sijDU%qC?_Z+;|%ol%g%`Yy0^aeAZ`{8bHc+Nh5ay!(o!|L9%Du3I`k6HHThg>(R
zu(QKX&fh(5;}bW!@5!SI`JD<od3N=eZ#BC9fqO0=Re0z*U%TmzPdjr|;dw7x_sfsI
z;L1^j`8>U%(;m>NcNKl)8+ZGMdtbHs>=E}_|L7;4y4oDR=ZeGL{mOf<Jp1A8-*w7e
z_giar+ckIp`Wrv6!NmCV=dHKJ*REZEt{?iDAOF@zZ{2DAS+w!dFx#P_enU~eXFYfJ
zmYrX6_t_)X{i_rAK4+a-a`3!W-o&4<PkxdAuJl-awpR|j{@TH`iyS{`{h;~KuqszS
zEgE+ycJoyB{PFkQ{l%vpcIv3Z5B=q`&(B=(?Qw@koxahfCqD9oQHAGkwDU=8o^#Q-
zLce$IFym*^mE5j=?TgmB<njZ?3ypXG^fu=`cC*XJ6;}RM`T6TVVafgO_lptC{jBm`
z<@9%2JmobvjUIUJ`QsIZ;^w5^x!ZAk^bTA8*@m-;&%f+%)_(Gr*PBh;@wG>O{@q)z
zH(O#cp7C?MispBZ+4-(F{q5!J&mO$V4M(1G<$ANn-euW~&bs05vyXYxw;u3-f4|%8
zz3>0R8(w?DI<u#L_F>yCJ#p>XyB_@d(d^%>KKqKl`01)Yc+4uZ_c`;i_qg@fH;=bE
z;<VjA^%pnH{hoC$uHN=L70)l9xBTE^H<-C$^xyyT;ER6s?j<vD<G)_#KQB1&mXUgR
z@*|%1?+^R<wWBYr_3)jx+~JzJ@6#P~@b~s#a?Yi5wDIA6p7ERQcl_*CBl`y5xz6&}
zjz9bPapK!w`28op=$*eAz40fjfBsDyTrv9b*S>x8Gp@gGbm<14x#D-9x_0!z`)_;N
z`d_|k4z-7${hQ6c_N|v)IfDAZXPmUdT{qkA;?cDD?bG)8ua_)aK1#$FKG698kA45~
zm!E(BH6w^0_{DYKz1hi^T{?pJBL`?*i-qPNiq<(4#S3}>#rqD2`kA|bV)OevbKi?b
ziCui)M=KoaCr`R?-SJkJESZ5r<x{nL+AaHz&UnWyqYBfGlHT=2TkrnOFTZb8VW&^m
z9kk9pc6io*jfNsVU$^DMkN??5OJ)+QJif{|vOM#$q`Q5p=2uQHC&fJXcKW8i)BB<I
z6~CG1BrTul@YCmi^nkVYxM7s&x%p7cb1J?oyHmbU`Iq@%w{H1Tg&kkj_h5&X=X>)>
zcluGCN0rmVMDuihSGx+$(_tmYYTPV$eoV}DO-iS|OiFkA)9GL5&$@c{GZgb44qcs7
zs`Bc*miv?R?w8!>Lz``V&A7s$>&SNOeJ(GK@4u%1uIM~xCB4NC@A&wmUiIq{%=)CY
zgXv!@`Z=FpZ<8xuy=11sjJsRk^E0D;KDW}$o{u?my(6Ezti=7lGqZfNRaTpUuiEq9
zfAocKt}>Ij^IFSxdjCml%p^YNh<&%b*Fme#Bxbvc_C9EZc@LWFOgi(x)Qg3pSNq6T
zKfdSNR+&kZ_d5^#^ze7geDcthXW)a5dFZB(xa6i0%sfy&^~$GxdWWmeT{;7IxMA-{
z?fk@*W^VfR>ATE4bk&)^`|jw8Z(aSaC0@P8I`_EXLpP2RjR$j_Yu>)YyZ7IB%^7&<
z+2^i**oJG(Jm8`?UH-)fE}Ow0Tw|}_togxhZWux3_&np#kKVN84Xe*QY5N<#a^SC4
zn#uDNIXNGZUxoR8TotF4S9U#VIltcdU-e!5GhT=D9cRA%@mKlLq2-mlhxX5Q_=5cX
zq*(E#Jm*ci%Ma*7i4m7_{CU&OcS}0`=+-y1zFdF9pY^k6s60`9h4dvc{Yph=zoa{P
z4z15|htkGRxp;2|(r=h~*&R&}U2oSe?+cR7{4D9}If3#VFKO*6%=wbe`;ero@3ATG
z&gnzXrB%D+C+S>I(w+ZEc_v-SRe8?0(Dbx^4S$vXvf~P?=hj)??LR%KFn_O<^6S=L
z@vl6`OS<A~((>wD)$yPFB%S&=DNPS5?Chhf&wMEDGilFr62*xM?dP;u#p^8Z_C5K|
z^(CGAnRF-LtWSH+I9$=4+^T*#Uqx5_l}|gr{war~E55Sbv~=!!($)ORyLqj%!%mLV
z)_43_$F$h3YiNBnf5(3}el_mUa`&1QR^xPhtMB+%U&V*?b7=i0#q|5=YW_RA9Dk5p
z6Z!8dZi`R;&qRftJiGHxrANxU{$0IwRM_!Pe<nrpUuaA|Qs1Wai~rO5rQB0a9j$+L
ze^J@Zq~(>r?8+;BR^IhbJ+fZ;b0wcnPmBY*_Ub!+r>!TC3adOg{pF-IIab(>*XdE$
zUwyUiEYEy>T6t%`JVz&;dN?WF`EUBQSeWan=#&e(<DVTR>Yqq&vi-C)J(?C%&yseJ
zTA|+y48{581IZsc;{bii`LbQNzg>ML7v=J#p*X31Rqg%$phCaTsL=nDsjwP1%PYN3
zf5t8<%=SrF`d^K=(B;+p3Pa2HIQP)4PQJ&gGZj{TYoW^75v=0wwB=R4qP)7dRK8H_
zT&!`r`{%r0PMW`(7M&Lp$+yC^qoi}Zq&t2q|D?RKTjlc|zw_TYe#Q&*P=8U$Im=Vd
zNoV_trq8*Kq=)j~waa{FQW{_Ow?lChR^O@V$~!$!|HYmD<6ED<@Ujsk_o3^|{GwYQ
zJAj@1=UYE?9rJA`j=(ApsOGUB6_VF{v6^2weXUTtg~o1Om0XmQ`=of|mD|pK;)Lbn
z3cEa(omMzhPMLS6T_xSEGuwA`jwcROSdB;T^EpjL^EVZCdQr(;IsNF+da84sb#++T
zjq=JrR_o95z3#iy6SsWl2giw>oN|9Vx*M;m@8tYHt-g}OwB?n2y78)hi(TH?Lp6V8
z2U+g_*G-C*zfqp~=cF`yt}xdzDLpi9RP8Fhly~}6jjO!t->nxP6;^tq93K^G*J0{!
zbw8fx*NU#<PBpIbPF|gU8VBaO?r6H(hpxWUH}%yxd5%xI>NmK&c;C6w?<}XsD{7y*
z_}|r&_oVz*c4NH7Lh`N9yxo3lmzd}Aych21!;X9NA>Vxas&oI>-2Yv8;IVJH>W1S^
z8dv&~>)}V_jr;+9cW;+*5PiaZzIx!X*IqM%`d4V43e5|<^L(Wr^DTF8)M3Y8>U$*@
z<r}U0cN<>*=#^)DXDm^>3hhgWRsYnRr0rXlr(U5`?~~SUzL@+Z-Px)7`SLMse|A_Q
zxhLX#p)u!4I`?r>y5qNtC-!%>O}_t=(|-D+abjne)jfssZvM_cclGK2(Our=-l4)y
zuJX<KA~{Wp72nD`zUZ0%Ckk7+-Ta-y_Sx_JQPr-~mj$n%^7nq$Um-cds-5yqU()WA
z?&MPWvy4wwyza)&`f7itEhncA`J<sYX`JZxX=r`t*A}b3yBDwYWZH7_9l9S~JN8sz
zryuS96z9Q7^*ikoO&*Eey46q1e`k02OU!j2u<XO%{M9FZGODo3XO(9j{)bB2|0-^`
z``pB>&JL>jd=Dt~p`t5(>1~IVe<7z1yYUyQzT+ePQs-}n)=z6Ex$a8N%C(<w{4rjI
zmEJ3#cK%9EmA)%K@@=c^d)2v1W)sN+raw&Dc!`yLD9`rQI;!$Y&dQU|q&vM%{wAfX
z^Mw8Guv&K&*Q~q3YP_y|+V=|WbA{q#h3?@htk$VK_dDr42i)1{D()!n_^tFrdDp+Y
zf6RJ%RQa*W9+dmevj5Y&q7fezcJ@lH6?XontMB|zCHJA_)%fB-hn1Zx&paXNp?r4j
z*neWzudA=*qdfh~#`n3`fnR+5Rbza}d*~6m!cN|m-&Ef9$5)5sVjn9chYCCWCYK7U
zag}%Lt@s{VZk`Ie`Mdf`J`2TPCD*Pz?KJ5wj?l-%2OYZVvS&W=mQkL+lCHk{)0O*f
zV|Blm<&~VP{Z?M}tIF9^V#=qYJA3WwJ3W+NOpBEsFI0Kz-<c0T^?@fIa^dZsJ6G$>
z{&yCdK6U%2KINJJ?<(U_(%t@6@>7oAp*XFa(@xO%ug(XHRi1Xa{DpfyZ0kL~HG)}R
z<%7yIk4`%MRMMHR(vJ=+zmKm9$-P4RQ(@&7mD8^dtNE2z`(LfADtE4#6s;%akajzi
zPWwe$FFTyp&Z>Tsu3P(j4()xO3WvsXaXqoxKjqcDL(6l%6-`%qlKO?FA1fNkyF$Mk
zh3=Qq@1eW=a=!InIO@G;yy=*;$BFE>!pdHhSN>IbwVw-BZk}n89^Fyc$zfW)s<@!M
zo41Qg{?A|{zX&_K%zf+Dn|fNU<LYOgw)=grJ#}1R)lWJ86Fa$d`N(|h@z-G|ck#5s
z`Sv$%Col25!fKsWd*$<;*ErTuVJD}mzT<DP>r<|&=abS$|NN?bKYQvm;||4*3MaMy
z^au8Bp)k)K9e>@tL-#f3MW?^)X#HTvU!||gyZ)(HNq6h-#;@`%<@l_yo4>Mi<<+>#
z^W2IScNYrfjgY;5`M9-rJ$U!CM-X3$*3)6`AA043-+#humLEQ@Fz*MFUMM+CI}Ur8
z6g{s?zk?>9X)*UH=}v#U`p&*OzAL|-<>WCfc6_p{M0SvJ7@BuwzoB$De(odM?~I0`
z_t+CVJ-%`I^3#_-_xe%Z13crrfBvh_y!D0=JbR57Y`W{F*Nr~%sQa9~#<!M_GTsf9
z$F%ME@e0Xpv9Qzs#jfw<*3CcP{yBd2+;-aX9RJQrS8=hshsijXbl1Pr>&h;acjL<!
z?yT79?|kdKxYCW^?N3!d?LDgdLzw!MwEwqT;n4rfm;VQ}qdPm!{9sbL^EcK1%T(U=
zS1*4^94fbLSM5WVA9~K$ZhGU>&Ky<P>E(RuQ$H*Jr@YgLF7NC9U#xki#V#M~_|Lp2
z>G{gL>)-Lu-&fd;U)867WEb$&U*F-Zjc2bLcbLz`Iy&vGqtnikuKw?FSDx!nI{)t}
zy4r`XJmYm`|H`X<Q{LIteD%tDD!asIh28vFpX;gOYv*r<*5`al&-ed!cX2Y;U(s(m
z{xSdgqNUf5A9C2skG%CU=ghqqRPBrW;oJ{A;$dfe>E@B|9-yIi;#1~{^ws}W&ij}1
zUbgO+AAP}<qdb@MAFx_ir8mkm->K;Qf5l0szeu|C6Iq}Bs=Jr%>bv|X^>U%<YJZgH
z`jgf!(f%(srk&;fpqIY-SDWvB-(QZ&yUPnIIp=&8o$(~;s-N<dXVRsAyzTd|**@vY
ze<;uPNv9qpZJav_JNa~cb^W_IBTiRX>6h}VpYkQIJ9F#*wa(8*iQ+cgYT1n^UwG*+
zMv3XSlTJUGw0x+-)R#LeE#6Fu-FiCx@t%-4TcLL1ZH3yUpP82K{9{*7{}v1LUi`bq
zZG7TJ_dR(uDW=~_I`uy3Y2~Y5r$=d*X!rY>7o{GHCl%tO!jwnSmELB#JTfucC2hSG
z=6aLv<lEV4S6}g^yz8&N(p&Wv=KfDg=Y0pdns3tbJU1*fUCB*(>O<10&q;Uk%<(3r
zD|sXz6)hj=P~MxE`zc@U@|&){@;8%~SAI!(weBh}QQqxS7vIuPC4KR->sPt{zrQ|$
zogS&rcqy(|nDMBi`K3hjL-*;aFY;!5vNM?ez2nC|PK&9(Nmu%zT)v&y>9>0NG%a@V
zX}<ONtdJZktmaQSCoPV^&YrsZ^EcZ0q&3gEXk1~}zw<kt{#E{WX!(5oO~+rietx~e
z)c>UUg9?Yrp=#HyFZoHj>rW38{oeAplP}nOmzREFT#eJ|Pqi-Pm3=9ncHZRc#x2j?
z>cFR8IEK}J4K2rCg_VC$UfF9`p8MO;=F9K@pnZ=6rXDO7-QEAH@9cNl`XwiA_>?Q|
zebyLu>+I&~_N|IDclPqkbI_f<rmd&X@~sNxTNP&9os?!b9sbGA`){`C%WfW5nEosE
zsPZ?{%CGCkuBv#K<(*znTA%Ar8h;h0Ur##o>J2yhi(}sP-Ydt6Id9V4z7MUJ|5w=M
zKb8JFXH@9<Ql$^6mr18QlCJhgd8ZfrNQGVhYF)~^{_3B4_5b;QYbzQ%zn}Wr(Ulyl
za{Hb3zS#5$Px|&IN8Rs=iS4m_zdQXv((=s;D>*3_mnx)p6_WQtW5tj1ZoZUL(pCTb
zeh)g|mslve@{@VLl=QQozUi({c)@BjFz<`1ai=Z!9!cIKCG9&RFyqO~AN9IJpR?+Z
zMzGS`q2)P#(nH6o+EwQu<yF5GRnCqRtGv+foaC2PyM-#R;{4EZd{)RGRA`^3#cEvd
zQw_x`pBY+SJvUH3bezR%mv&dt=`YaMQDOSONomhd7Yftf<fRq1_SAeIyo$fdyS%H4
zcgj2a$n{Q2<8vqu^$%UUh1zf9(APT(yY+VUliE$T@A7Epq=mvxuNS+1sJ;$u*ZP5h
z|IeZur}CqVRo?NlSoN9r4y9AC7mFU6A9U@8?pyk$&0loEyN=)Isu9d}erWe!T(SA?
zmyHrn_~NT}dB-O%9361(qwf0HZ7!Jm?tt<{{quLmNguGyTOPf~hp(RF$M2D~Oa9Q6
z->LG#J6n0azt+*!?`o?2VrcneA9waguX@6*<4Li)2kz$W@``EwLX}@;d7k^x-8r~>
zzF*yYDj$0O%X{6VtGFU>t8i#stoZEqjU4^oRJ4BR4m-am@2W8K9p@Y6{3gH1pP)N5
z-XA(vdiclkL(kcFiwnk+%3-K|R_!{w>gua|Xa2a;ckQOdE<Vrq{J_pTtmLb_+t*5t
z%DecezFP03=j@^FtNE07`m@mW;@D9B7rUJ}HZ68>rPE7st3vS&%8x61%>VzF_NTt;
zuRPZ^^t`2A_CII8r(bf{CtNw|<dOB{V*SwgnOAh*XU=-#L*v7gU(#tG^eD@xrRBNi
zbxwmRFMLmm)}QeJ-{zkbH-7imzw)8+#iNR!?)e(Npna#OklfJ_|LPNyFY|PW5B$^n
zMD;NDtD8qXIp+SL?F%}Q9+5|)`oxTjH?O<JN8Yu^#bfa!>yz&Mw{c*$&+(GBpUN|y
zi#N*A{OF`;o`pvIEfy{`zs6$+zi*gvW4?6S+nt3@JMU=wdS^rO^xX@%$GL}Yb@Dw{
zovBc}3cGzzf1dPnj@Wn0dmXg;Od`LX@g?c%zC(G%ca|q#D~dkg{Ac~(ek)%#-gUPl
zf4jw{zZ_q^%K!O&l;5Sh_Z9!p?t!6l=pLJ2$^4NV>4kVNem(XD-#y^Q&t5e$ukT|f
z{^NDG{`e)2ynb}TeZG3&vDaQRA~*faFK(j~^;;~&&wR17|HZ1eKZ}K@z2xK1KH}%U
z8@o^c_9G8GYt6e{ITlZsthfK_pStLTvH9go^e9n1{gQ9%PtTw{1S<C(XXC$lpS#@i
z-_E*pE*~~e^&XXR67iM%7*BmgE5|Ri|Hjo0|Hg&Mm;4GE*Ld<g^FZx7G;WtC+IRW@
zvmaV}XnocXtyjBA(R|%=A@%N;pnD|ah|_5NkSkg{@>7q7+Er+t3afs`?~r~ZdXBZr
zJ72civ)+2+m^`eHetBNW?!+VYzIVQ4&-)#7k4sJ&!z@oa+vT|CV<-6Vyc8;j;v_Uq
zqWTWaBVLK4Xs8_OhcEJi)=MssUtMjJ@BieqpZ;jh-^vHzR%>tgw(lQx@fe<P{2}*x
z?QXvr!>=Fwilct|&R@*^zD+qlj!vB0|K`0a?JE7V&I*UxJ^g1N=+3XH$LCO-^!(bj
zf6o<%z5A8-UU{}ceyPIz|Bgws+Yb5P>pu7Vr=4-}g(LET_K9Ccm)PiseX6h=w|S3;
zy~%Usncj~$<Csl0d)5w1X3l@WFZX-uYAemW;f??Lnd@G8>)iL;wfBE$<*&wt@@C_j
zPq}f_4@L5x7QI)0XTf_u_XSTo>&UN+GQQA9cFP~7zo_g-zZHe^JqJ|%7wh~|jVqs-
zFN(V>3i+Ld!fE}>wEf&q3`O~3V%m{?6!+x`na{{4pmk(^BOj5U$XEEaME;N-<h*F}
zLGwWK84oJ2kUX>AxQXnrx`#4eB7Ze4%4ZVEV_GEl3a1@+MfDph&$NS-W4_;ome=I{
z(PGiwQ>##(nctg4SN)V*SB3Pu!fszXx#aJ9lg@og+Bz!C_DTEBa)sS@*0X-vJS+7t
z=}yksKIx2e^aiF}DmuqOCx3pwh7bNE{otgu_%sx&btoS?&a~|oDxaiR)%j+j%I#-`
z<W}K!PyO%BmM*zs40C_0akHGhROtCiV%2_VxjbQ7tn%c}530}gB&}VADaWM6uL`R?
zW@vfE*U)l$Ug6Mjrfpa4dlmn(JoP>4O76<5`LjI7MR$5$$u-OIlbHH6Dc$W$%JZ2&
z-|~ZR-+JvCnD&q~Ijks5d3S!e<16E6w$J*cb03qQ?>){?KcHRSza(A75#`xF=`Jom
z<J^m0y5l9+j4PC%OpD!q%daZz?58`=bnBlle`)_or~W0K?UVLgufn{aNV;49eCu<*
zq$|A|THf*3?O&%KUH+NtNjm4R?oXAcKBqpUKT7|Zbhb}A$Itbq-Q{|!`Yf;L&Tk}t
zNq7CNmwaK@uZy!)o~k_gOgi->>Gab{SMpSzdXRLEmvm<ro&0jVq{XKSt)s#mFX^hE
za(Yo=#h3DwXVThLnEEg&o%)h=+Hum=x|LUaEARYnSD*Z#Qx5r_U*=z#zp_7=`Cr<#
zb}-xLzGnPQI@>3m>mzSC-+KE~Avsr={ypimv-A_$KIvRvj@Qx2cj{T%LDJ+5b3Xn7
zCjaU8w41}`dANCDJARCp^3L@qo$JqYaO!8~AF1ET@9pFl{N7GpQJ?EcdQPsly}yv_
z%kgu)sV_Nx@}G3tQO?)tbIzCY>(-Ow=l*0p{>UxgQ%QLwo%|(jT-eDc`=>mdb1USR
zD@=Vhj{gUYhUx!Pzmrb>E4n(bRD3H>e&$Q3{wB@uRG9XYw0*BI^&#nMUgg<-Qo7oo
zq2(#BN$E;1%DeS+^L72ZdUCF?n}65WUVG1N&ssW@d0E;`($=4t{s;Yy$GrbH-#YDz
zxqGMNEBVj*q|;9(&2LtiemCh(?wwup2Nh=joIlr_^+WYvyUYtZI{k3c$xlb;e4Tt#
zACk_z9_{}qK>Amq_pmF>_%><$&3vGvQ!gsIlb`yOdqr3BQC`_m?n~14rNWAzEKmPa
z(K&z88P_u&CY}2~EuDIkbf*{Hyj}lXSJL)5G38g${LD>TZg#=mpIm;s_nWg{`s<|W
zZ(@#9(Yd~)t<&%QD$IUKCx4UD`Ft<wYJEe?Q?5yO{H7fy-T8fadWDsLQ=aoBUHJj!
zot(P$r@WHR{YttUzw)cfGY^=QKJERFUApz=Ys|p(!&QEwJmsDDbL0$DKB*sBpLEWb
z`)3~5<s~^@*FN*$lvj?I_4tChKIJg=CEMEvnDR_I_b2IGU(T27NqOb^lm8q)Y3qR5
zpWVZbpX{Ifb^K?0`vTb^On$V3xgROtq{#tVH%$Jsf7a_)VJEMYN76Z8=Z{hz=?}Ai
z)~Eg7?!0;PcMjWUzw<}rpX*CH?IG7w#VOAjR}^NRnza07DAF5w+OEr=cls(HUOuic
z^PHq}pB9U*^f>h(>9o_NyL_3wPm8J7<WEn3^YP=>{HIr)GpD!8)7~mt`#TCNzJ``}
ze0Opu&kDQ#U480H(jEVuJ)~bux*NZ%Pkxh5znU~ZQDOSk>))~K@2)uV#xZoSxbs@e
zc6$FwYs~Eax{tknot4*`+4C`Hu6N{<m(8Ji?=x!W`#1bG{_S`2iyo?fRlCK~i*8)?
z6<^8^IR3JY_dn{aF-*T)?U!<LO6>fhdizje>T%NPr;|><nKVA8MRKSx`AwSsRhV|4
z{c}A@+vf^%{G_XWQJ(vrblP>&nJ*=s{3K1!D@=V(+Bzz%##LVJoATs$S~}lrL8m;?
z{{LZzxnD`=@5PeNd&Cu$&;D_qKT6<_5?FW%RR8Dg_qBZCCG^J;|0sb!O5hJo0-1ML
zd7Qj3v1-3i<;mZqbn+{I&3q8;e!0TxyfL)Ax~E&H@+#j~KJC7cPs)F&z2rV5ZT$J7
zdyvFw*PZ?7&q8CKZ<EgRV@2n4By^`=IbYJ<zT|$Qo!ihb?=u#fc8;vDyBAcS_i#h$
zw3DRCA+eJ)zB=s2sr=i}^2*=wGaXj_v%DIIURBu1C--AodZ@gzUG@A-dCDQ_ZhzBn
zCEfL}>Qi5nuKG<{p8hB4&i^J~=%IXc<IlH#T74TDZ@YHnk(he2Sai1!-Mn4@YJaM@
zsl4i^yxX^wXVT(a$A5M2uDt8t)u%m8OXH`)?zwx_uPZ0V3OoPb$!EUx6<^A`^>uPH
zPK6!+oqcusP|f@O|GwLgK6>#fqY7_X^Vs*k?3yo+h9bLzCqC<?uYBUAKN@%IOS?<D
zv*VP{+rGd1Lw9=i_r?`=^2+g&?)d5IGe1bWvj=%pC)cXJ)8~9X)zO_i)u$Xg`uMYd
zy6bP(ymn51=#}qt=Jz|r1*jhFyQIFW?7O7Ojho*$MXQJSsjypLSKs-Stk3;c-|1sj
z&kpl<B*yKqTi<-^^Zl`fraS(~XIiYzOUkQ#QO>R^Onq2ry7ME`$|d_%@oPnwS8`Tv
zUplPh*6Ex2%HL#pcaNmL;wSeZ>CXP=Ti@wxt~cqfzkQn)?Z>n@?Yz}~O<LZ~Kkd5o
z>*U|n+fVB<ZiSscFiwS?Jk(e7r+p+{^;14y{wsNkCtY5tp4=zJO0LRN9!Yn8b0|O7
zkvOzJc_q5<tB^mfF!TILX?bjgx!$BZ`Kr%;=xSYAUg_ab`Ll-#^S-Q$GuD-u<0M_p
zKWTX-59O1NH)(tJS0O$UJAb4;?KtU_JDNYOuv=%f9_6{7icWr$#&3lc|H`{{RQ1(-
z^wc^NJA26XNl)sRvVAvySD)i0-S$7}OHcXjnZKU<|JUv>x_(*j|E5)#dXjXdZ_3$E
zh4iPwq5gEC+Tk0f-prT2=T*mk@N1tuVpQq9@|5$WH2Z<`<-6*Suc}^owZB8l%~PRy
zCdHxi4{cY?&ps=x{G;-!pYpV8`dguKE6nkduErf&K3{tsI*)etrQ63!9@Ra&a`SZP
z_di21<&LiSoV2{+x04e-77M#|RDNLEa(;IxcJs_vzvtVZ--ipePwc3|e6OUVox3U=
zDu+&AI(_c?S9a3%Q(yHb$0dLD)dL^*+*3y|e<w%Z$jP{g-8$8q7gqV1@```Y=_;)5
z5tLWweC5@+L(4O-?)qi>PJg@lZa&|~tnfKU?7QW?4qAPt!fyQJFWb|nX)(u7+PW$`
z&ig0&S?L>n?U3Hi7c(AI>&SLb+~I@UY_ikUqvx)%+YR6U@X9kV*K54$eo1|Y=?Bos
z7dqcV^}B7T9nAd@R}wpU<Ez3hPIcq+la;)zr@~xk()?0|=C81tS9zz`^R3VIOiJUk
z!fw2(pYm#b2Q2&WH-GhspNwE7-$~2)8+u1?{2v#;yUln!l-577)8nq5ygD4pXVvb)
zogcFG0bg8y_M(gK{o}Vh<!-Zk>~ZCNU;d;uXaDp)k3Zr&TP~SB^WmpH@Wewd9Q!|s
zOZL3qG55ISlyUw)U2^SEKmV^cY2`yve|XE`&)Mmmvp+FH8wU;X>HqY4KPFMVeLz>^
zsL%108{hjk-tS4|f4tWNwKI-!&4(}RwqNMwN3VP4u`fC9cJfiLf1>fti+{9wxO|Vl
zyYe9$|9qsr;?w^>|MnvfJZsInTshw8OMm*+TOM%z_~@TsweM$7y=MIQ=bd=cO1oV<
zN1t%~A@_RiZoe6Cwf2T@`~FcE-wusOUn{*=E>Ep6<7m?KqQZ>(=*$y3nm$xm^{@C+
zUd8E2%PYS=w48rV?B0j6j}^K<>u@Na*)I7_x|=uKC!O)G8n0Tv@~Yp^a&ff6<h!CX
zP9}}t#4cV=s;|?wPrfFlE4g&#nTID`_2-W(%-=00o%0OUm!WbQy8p)O_O0WulGCK+
z=?9Xo`0)Mn3Oj#S*`M;{V^Z2U6;|VQ@~g&Eo_;RR8CgFqZQM?e=v#%=dMo)TH-CrK
z^V%%Wb#&u)a_joruL{$its7QyQ(ld$yo&>s-6=OuBL0RVIaT*+>N~9ZE3fi8<+=Y!
z=Y3$(^1uo^{)g7*y(_xYNBdr3w_jDig(|m>3Ufd1Z1l&!wCP0$e_`!e=c&Vvd-EaR
zeEX`iu=0P()83Qr_NTMwt{xv1cKxTV|Hys5eeb{f<OSmj@4x)U7r$oHi^diDU0LF`
zqh)XZuT8HVPm4!>_ep>A*EiohuCU^F((-e+zv~Z=IrBT?#NGD!^nJJA@8ofXcmM1E
z+U8rAo;0p-hgA>y!h<gP$!J>q;Ti9E%S8{kcr+>g$BAqI?0`$pFMmhq_l2K2VWo#Z
za*wmd6>jnOH~w(b?_4pSFFx^ukKg1~ulUtyTC^T`+FGx8|G#|ts!@l_KXTL|GjF|d
z)ZuEEzWJ-GUiOReP)t2vZ2DPOZoSue+g~{viobm7pB;S5xyOtvtnOVtdeb+S?eO+%
zNAQ%#A99yRzxkFCeB{2HyzXARpEIg(mm`ip>8v9!9d%gsS6=b0{EfeQ$y;9ekfk%#
zI$r$xC12X$(yK=mc5>|Ws;d9=-@SR~gC20hs6y-O@XUXD&_zFb`Zc2poqwHYr}Z!V
zK!sOdbM~{3y8PXx{~Bs9=~vMwzw;Y!TJo^-MimaVtIB_whny;`@|Z53R(a;6<>mbG
ze^2c@oOGUSU(K84*$+K5U#{BaJd@I0zFmE9$aDVco;=I*J%Oa1UlM&ks>8~UReAWN
z<$2GKCWi_=Ur%(N>u^#&tM)1Hq;tQ}*)D0%FDk6~QC{`Sa`uwgo!ctAQJ(siG(I~t
z9<0te_x#s`e!KpDSC11P@u*Yw+Ii#Oj)!99*Q-1_%lY*R#jQ!P`>savku<;DVK-03
zcNM3VckxPnbv}|`O^f7_akg4dRbJ^wRc_u6t&5(+w3~|V>?rL%=^U5;#CN9`>a8PD
zUYYW*=)6CxXmYBsi`!K{<((hw>MMVx+&mSsi+N*{ukj0&-YeIx!z%Ap-u3V5>D#o}
zolEHJwAk_0%|B^<r62rnqWcOsssEx66?W_AS1PQ=Ro=-*yJ@lFOL@vOY3(LOd0^+y
zht_9aH!0oOq58^yb^cNRMCXp~ezCL9YX37nEi`T4AA0E0E7pAVS4KlI`R&%%)#C$J
zenR;|$)UP8!AFNf?{Boj|Lw4edztqn-Pu`{zf4-*<tb?&=)7m><XY8>%hO_&w=1un
z&vfNqz4|j}?Ya3`;|g!L-+S$L>;}92Y@Fp)d{h2{Z~o_*YoC1mxWb1VxAAsgJN$R!
zX|a1w(aAm6lQcW%u$t%3Z+XMYKi}=w<HT-%)K}xqx4g?Es``}AQ2N8$zw4B{?zh%#
z;+uZ`sEbD{-EDT;HFy8|8$YnY9PT(idWS9lY{S{Ne((DCes|jq=1@7*@9EE5Z;P*8
zyZ&5%w03BypL#URd7t_5Z+-OEoz|a)m;LrV_g!iK^=1z`=!$C|ywjh|9{Am(Uiz0`
zy8G-iA9UE=F5UiabL%{3!{zV1@1HK4J?gw2pYg-hm(HR0_bcq=TKP}s%L=RWo^p9o
zhkh?qVb{OnUwOs9a_uUVk7c~@94XQNl}wb!cUa}k)pPx-ynBDH>R087la`ZL=U0c;
zAGYj0Kf1@0KR2qd^Q-bEe8DQ-R(`9SM}2j_DBgCMdFD{M8&|#cS15m}&~x4n^Ev7c
z_uKQQuixb6QHOb6OZvj~FMRBpzq)QzVfCC<dG(xCdGeifr60<>yuYjO&Q)D~=V!Wl
z_A)KzekI+FEAN{YbG}LGq4To$M0r<*-TFIw@7{;$)=N)e<_TS1Soychf3n)9&)w}e
zqYkV06O?!Uq0|3-e?MvW7ZrB<qu#l2`3v`a*w%Y|YgCP=Ts~1@o)<j_xO&^qeC*Gj
z@PqLazwy|ERyq5uQH52UuKc)iaW(ZIX?at{zws)p_{(!WIhe0P?K-UZv40hI<I;mf
zc2!}<lcdvr{f@7~+^3|g^<{Y_C(lEmd&4t6yX1FI*kIWV+-LWr9=7G7OJ-o^2k9sI
znQGoFmmgPo%VL+OygEAnzi~xZ`FK~J`@Gn6#s8t_eC?(;KJCoWeDRdm+%$UNx#y3k
z#f94ct6%$~wJy2*fbnACcaPin#EtHI@@TQJ<9o5@HEw4Y3$;(bZ{y_eZT?7K_M+!*
ze#v-6$zi_wve^H(y0Y8OuXcILoxPslSSZZ+A#a%!J3n@3t-t%Hw>j^zn_WKsea9-_
zSg7*q9HV@-e|YM}kG$x{afOS$|M@$zie4yr80SxR-hZ=AUv~3&u`uoPj-spQCd#Yx
zs`AQT|Gt#p+2g_=3V+r8#)>K*niu6gP15t_H_!Vkn)Z9&NpWaiIBEWIXKi0SFI`dP
zL-loNyM@X-yKxrler8fXx!86^?%lcN&R(DXZN7AuPx(H_Vqx{YovvKIv)Cx_s&My9
z?(?C|w!UUuVU-^%f6y@-d}?Oz>&JH#R_BO?D(~)#)lZvmFLpfY;z#GV_?<-WB`r1%
z%>x#yT^DcEFZRB@_rG4c&BNEYadiIrPgrul`~70Uu(I=oE}yUc&DWoF{+`_~G<uGg
zxY+j^^Bvzk#k6?-MmwLh<~bLQ|4=dgW76IGgO&Yt`HlKckHzZ>=WB0W{99~)+40}W
zXTJ3->Um2wPbc4L?Yinmzkgru<$T}ye+n+txwnfO-T7p`c~|E*7Q4RVzgvG--|>G(
z%fDNHRnPv1V&+3hSMv-lul&@|{$0DFey-b3^$R7pl;eEq&mQ&gr)>V()n@PPdEADZ
z{lziwdheCvX|c1*N}rTZdOw)$=X+0wj}?WLzB&(A*x40(m=?QpUe?by&%2}hr#z?C
zfBmZSz(SR;DE+MFsqQ_McjvN)y!7Z@mwe;@{T=KgA-l?bot9qc`=Z~s@w)rY#jZc^
zW$S+V(HC4fs&I=P-tqBAz3SJa3f=o9cKQB7#p|3m>C3koUH`y6myf1JzuSOS+!$J3
z-J5jf^R@f#zTwVNKdm2`v|sleVZM3kV)vgm|Cldd-T5)=J3p|Z>&4OeVt2k;DF0?0
zaW}Ek<2$N8^Vp>M(?q{_n-=rja!1kgmHW9Lc*Mib_|naz#4aDm_*c>0I9+}BUD`?O
zyM3z8p9_7ynbbdaeqd<*wEp~#>c<aN@~X}^Szh&DQRUV9JXOBC(B<8EWoTYmwHvB`
zRlBr1&+jVi&Yg>`uc==PO?P>6cTe~GUO(S^md;<xcjk-y6YS#4V$~1LZx`Dxy74DH
zKdtEY-Tn-f^B;P<>izum^OMqkFEuSz@oS;VtN8YZR?Z*J7rXnk>OL{c{XdESX_)K$
zBmIBhGgp2n%kw^IDBYbO7CMh#(c=yE&z1ii%5U|)bT@xD{?PlKydOg6ebS`#-9NGU
zeV)1RMWYJ4`;STeXE$Efcjv;ce!lll-S}<&9sGXnr2C`&LZ26hj`yA`4tw`2@4fQu
zofU_!<M+K?7dICB{N9~&x^q`|&g=5ss(&|cSI@pC#V*b)RDJT*J#Ve*r~TfB_gfOH
zdp70Wb7S>6?qbm?$D}K{DZjJ(k513Kdi!=~LHjeQ9r06l)P7CM?_#&_;zH+lt9tu0
zEq3e9ymL|-U!C6HQT0R5i~m#GvC|5x-{&Z=o{uS?RA0LG)7~Fe{W|^c>T}#X%HQR@
zxu5KC=(#WNS<v0RUUgm`THf6Y;A63HzR#7qdp!R?cCm22ermpQ?d+wiPrY4inq4IF
zUyFs69aZwj-yIbv=R0rb*XLb-(wCm{+cSSXp7fq?MYW%=U#a>}Y9G~o@Bg&&?*1|3
z<(-Yrykep0O7A;6Us2D;<(GFheD&9NIBVnC>&A)Idn=XQE_QkM`}wZkean1N+^R6|
zQ>*($<wN~lj+?acD(v#8953mvf3*(dRQi<r(e2}W>+zFv>eg4)5A|b1&nb6QyJ~%V
z9e(PW&);RGnGQWSJ!ikCUvk$cTsg}8r^*9|mUnW_{hXA(9ltC8|9_{Q9>FU9DDU*P
zlK(=LXPm3zQ1Ur+esbtA{XoS>mg`5JiQRje#>@2$rOgksU-Gk1^rU)RwNHMkc2&7`
zCyIl{Pr0Prwfooiy#L!TSnU^M`lVcZnEWV*`u#uby?4A-)wTYuh++d&Kt&UY1!IdH
zjCE~E6cjNQ1bYLc(L^9dV-y9WqOnHBf+cn_lGq!zg^H1=K~O=OO0S9nN>S0@dp~c!
zC-*+wamQL~?|nE2#y>oaG258W9COUF)?NohPrZuPI)z@_Q`zbt?P_{nS?8(BdF!+C
z!%NHgPu}{{L-!{9J+Hj$5BCZ12U5R<#9P)23%$7S*^50EdiPeWYaUvEx0UyIEQALO
z@he5o9@4FP(R=;3>_$&Jbp`Q9;*y0ruT{4E1%1|ii<M8?%PV7tV%B;ryU=UBDtq(J
zn)g;bQhnBb6wQbKT4>pi9-b`p_)$JprX35txahra$UL#o<JH^edG=Z9R)1A*mG|r>
zo$|y%#jNp+dMwm>G#+WckgfXi>ea7Q_QsKPcvJNJT=QG9?sID7V_Wa|vpFLY+BtmB
zzP(QWb(4~X)_g>7`8)cWcDtYS<c24GoaU8vf2^{`IhA#tS2^$a=grS}Td4MwE938=
z@}aV27kbOT(0jkv^wPceRlV}wcNo0iTPhzatG}tN_iZYB_Ic-QZBJ#dyw`uVS7on$
z&tJXy&iHS9=)Q{%YVuvW*SZ_^8}Y!*!l?G+#~^-dp^m@WPdy;%$kfaHLl2?{(Sztg
z>I2oE&?~-t<*y!?aLuG;;;r`2l676PWa1(AW<~mIq32g#y0s5Q&-!H{>z9SzIS*be
zWc{_!;u}5p$NXJ#WRQMZ$nWK>e&{|+W$S#7-s(4cZ$GW=sqB^4afKHU-V_-RYuwT6
z{#IrBX(8=c*m<p!mW&wCprlCq9y0D~7qarBvd6EE2Qu^7>Nk3gS1RM@7J6~U;|Dt|
zq&#+6cAzI82oIq0TiY^r6h$xI!h^M*czVy@JiZt&3%&Ne^i`VwZq0!+UZ}ZWXWge9
z{2+6GZt?B;udegDkJEjQ?$hwYqKN(UhjaD|PyOcDKRq}-VZ3$zpa*q4h(93vzvI3<
z>aB(YzO3Pw^6-RAytMp*dO>ZUxTyMS%WALsg^m;bw$QUz$Hl4#z1j!QAo(n0K7iUE
zm02$=wA!)sj5qff+JBYR9+e;O+WDW4-up>nq2^Z^9xT-SDiaUcuPEwx(LbI)<wfR~
zg~T^S;)91459oD1ShBXSvhs+m_G$gt0aCB#^JH&cdf&ApK3S;yL1b&(RIh%MSGLXz
z=)L}X?J<vwBJ;*V&ps<2sb1^x`enr})m!6E{bTMu<CKS*EKOLCv!dEhya1JNwab&S
zuP7>?Dy#h}<9|ibvfG>Qq_h80yHwUVj;!+(S$VKz@>}SQpH&V$anwSue(nV<^vc7t
zg<g5fZuIK^DwEGb%YV>Y^BO%oTj=$#XgYjiCrDfXSr<Va$0P3T(dGx2u2}-Hhw>gC
z-oEv5`y9PS341{JU_7v|2Iaj%#^;naPhEG|g>NnM-XwNeNO@56VGpS3d1d+u(yoPG
z{aTLk0%-@-{2E`hJ<bCl^?+KRme+L3ftsFGW`BqM)aRj|qgCb{Y$3c_X!*To7wKOA
z*{53QmG{!wU*PW+GR{_ipIh~i6OQURDy2R}`kNOm-qCygE1GWckX4VLX2o}Q?K5`m
z#`97|crO>T+MP8Y{bGLuTI1{0TXdYPay9G6(tq7?X0ORfR=Hd|iBlHhw-#FE&|7gT
zt6rZkt8DR^Rd2;9^p=01S9>hkvZvhfv&IFz)h>G0O$*uQSm^mF?O5pb--_Sp>1R=7
z94++fC9YV=Jh0I67xbLtEVSxHZ}khk*T1sU<^3-8Gs`c~d+pN?(8^DGT|=E$oWDT)
z+(JFiTlw<pc^=RF%*z|&XQ8z}AfBk7sBG1%dgWJT>wX41EY$O+%2v6&dL3Vtv#z6A
z+qL3YUVUDFv-0bBs%-gDxq8drv6J|m71>XC$a7r_wf`#XJW*Nmt8B&3to^olL$B>+
zl_{@Ce6rB0U*mwv#NVu_aRhsm|GaY6`OSK-Xw3u94$DvJm!ggfvS+{AYspqUmYtQ-
zV-NjNwD?7j-Rg&0k0&eN$h2dj-XHPa2Jdrdy2|W3v&K2{VP95MevoM&eyo04aSOdw
z4!vh5`HCWbWTDr-w|;B<s%_b`55KUG^<{aXH?H3J;GcQX@(0y3A1w6R_4=1L-QvTm
z$0~>3>K}S9u2YYNMg5C7UMVClWkvcATKz*$J`0sc^*8(p`$%Vek+pvEgV+PI@6b4a
zJ>&yv2h{ek6SU$G=O!(Wp7R^;X*_$({N`Ubd8FodY4jV!9@+)bgXr@je1Ph=+7D!S
zLsrCo)l*K>kxAEnkPd3S=t1_=@D7p=di7{J>z(rC$@nSHf8g6fZ~j~5&?|qnEn|m;
zmOaXMQQ4ZG=!p*&GR_`SpQ6=2^!PdN#Ss@k?;ganD{DIQ0QCIWOSjs!{2;3y-irD|
z-g3m%a*=tjNPmi=)~9mOdMSrLSjhTiq2&jj|7g0^K6>m`wB~hQz1maPWUD`(zgzcy
z=!s7jdOR>b7FzY9$9}6k>pA&7q&^F+b~LUdtG`*Y$CG6bdY#u<WpDi;-SVFrz4QG7
z?6lB}?_PYh#sxj&q3H3h`7_J%-ksM!&koWp|5dv@zBK)@?=S2ATA$`63)yebzBP{M
z;XNzTKM!f&L*+wd&n|18qNksVI`5Qcm9@OenqOt?v9R2JYJXH#d-BTV+UNO!m+rOi
zmG}IWbZ`D!>7IS+FDhGpfnM{goOL|495Vg4Q2j||uYG>U#W>ZhN6vd5+~b4ZIa<>x
z2Wq<BCn4)NV>bwI7JBVy{hFV2{0cw7KJ0);WYWpczFp%g`9P~5=zA`HrOlWpW~F0x
zesk+rubQ5&(rxo6HlFp5^nykMhE98MRPx-@>mGOI$>VB}{mAwwOug>;Gv_C)OSiqf
z<Az5)@pba!Ygcddi+#UK-hQIhwug0Er8MWxmxeFC@IT)ZCY|32wYl>1`G4;FUBde%
z<R>5XfavwR2tPe~_yu48eSU&nx9_&uxV46TnRv+k(*+;x*7T<xrq=XxLBquj7k~J2
z^4;A>bQwSH!DQcc&cA%*tPhs?UI}`5V_ZSThjuJvpKPHue^rm)WJUao`z{c@q7@JD
zClBF8k#bi3mY(@gE~-72tbCU%dv>cG8c$S)N04~Kb6!Q-CtfI0-ts5><+zjX|NJMj
zrz|r+m=_JMy!nOSetCDwywdoB%<uQX+fJVJ#r7x9NtidxZ;<cEB4<UbKdOgEkhp81
zb<dJlZ~43W1G3uf%Eo!o^Cx)7it0})vwmBs_T-f<9`owqIV<kf=<C;~&z_WktS|Vp
zh4?-Bt#M>s?z!OJbLK9ZT+<HaLE0xjh#sW9zQei?>U-qanstzL=C7i~2j%D|GKd|l
zk0;JJX^*F0ow7{3<bwy5(Nix-|7jnYbddT$);r=B>pPft9((!OPgw{*7Fuyg_4F$*
z!mou{*FVdC)oWbQ^N7lM=P~tb9AcdVHO`>7{2e{(DfumAJ+sg%ho1JzMQdEWdh_~`
zm7nnd>0efaZwskcF>C$mpLzXF^S^vz!_RlzX+)Y8wLCKWt^<Cz=%aa~HX3r^HlHVl
zOlrOi-n(AUSI-#VWXS#JpLF>T=dU@W|H0FL)v-gvA@}!d*kk@js|?}$rOXHXp7#NX
zBNo3|_1JA8?SodmS>ui8zh1iKub$qsPvci!*|S^IJ$v$|=e5_`*XcRLlGP8&<=^r@
z^!Q&<B+g`xFFBu6P_GrY(PM|AH{Rv)tMy+n`LL0jH<**cC;q^D#mI~^{>!}*_d+1+
zC3cb@S=%RmS*Ycy$3o4ItnFGd`4!1eTvx<Di=wuN?A51s)D_vY+e*iNUDqv{^-vLi
zwvc&R6tz8-89xiXcA0k;YTU47{MJI|byl?cW9hAVs`kPk<L}vH`44)nA3aEWie9^#
zZuu8_&98Fa^*V2U`fDNUQC3tx(*CGy#dTM2{Jmuz*Sxajryj2sf9SQ}S!K%}^cGKE
z{8%Ra-`}agUeFtNtNpBcYyP9Rc*;Aj^vA-i^?80^rCa^Ts<-S`z10uRr!xLj6g@vF
zTCY_O`&ke0XZ)rpDt}fzUVk-R>qTZA0Cl~gUlvjisQ!drk$!1@OSZ;A^_HFJ88=Yv
z@?>pKW!kaOtDkh8$5wrn(p$Wvx9*oow~+pLs5~NPjpMrRtL(L}e5;H-7JB1g@uPa{
z9HDwGr?Ry#SG~pwWag)Z#63leXNx!0YkMBAq-+0?wSG(1xTiApS!ng!(yM<hBRlUu
zS$5>r69*OH)kCj6<uR*l@rGXKL#1WfQDlBtsBuSS=8MIf$G6&}vgP0Cm+>>>IX-sQ
zcl7+t<GZe;>-<E{YM;e}##>L$>u;8yqu2SNvS&|Lf5vXwU0!~T9eI)YTU)4pWt|W3
zD=+?f{k5Lo&~8!0zN~2Riyy)ZvWM!=DtqO<bdOg}$Nv?HCyMk}`PY4<rdzT#4mInk
zu^w6T+!`;{dpv4>mA&#_x{gC$*{Wa1OJ$31)oVGG)n6(tYkSB#4wkI_LFT<s3-Rye
zh1R)R^;Uee^jZ5;bR0eZ)i{?|#^3P+4=sC$L-f;nKi*ouEPu1~RzEHK&|CGXp8T6`
z_Jcl8?L9wT^wjUxeCEE-QuOp=k3BZs=c1QCtNDF}eoxMldEXHK<9Rtq{Z>Eo>S?E1
zh`rT99iRFp<2Tj9`o>3X+gmN(srTe<S`OK1y|>c5SgrGs7sty@r{9W`TkNyS=9^DS
zDut{k)k15(TPZ#JqiSKL?5yp2*mqP5t2M8*UXQm89_cXgr`yg=tA$?sMbmj75VYo>
z_uQ!FylLFCd*kn|6T~r%L*D+#+o$O7R#m253$1>!zq8QVFXq)NAIRQ#R4U!t|D*T(
z#v7k%r+;~Tx1IJsWmsZix%+6Z|6cu;f2qFc{O5dPA@BWK=-Kb3Yd=)B<`3V0vryxL
z7hg!X;vahQS;+jekl*!N$UTGdfQ&zR^TOgy^;SKW{Z<@&VB(<%y|?CSHNR`Q;mW_g
z{p=cR*2w(cW&4kRmYlQQYC{w~|E2$V(XzMFddn`0-@N*)@mlvS-hQ;wJZOE^JW;*H
zA9~9ltF5=<B6@Eet#Kosc*uR5BI67{S<!3H<I`)$>aVGHo=<4J^6W1+-CM7`@y0JK
z)cKKDW?l2_tX8^?M^@S6tz3Ie|N8fNq}=+w^4`4AbA`$tkGeiyJN~6_&OdNWGW_XH
zU)|(KuP=jn@252Xf^$w-XV%;cQc&ad=~LgFI&rlRQwy~|m9gJK`g_yp(>m<D*4K%m
z`ja<4n$A9|C|c(<=BJ|CtFmW@=U?z;p^l%*Uj3TC(sH@&E8i-6@qu)!Us|8aRyp+L
z_AhHaUOQGg^;Zh5c04`fpy=&Os-0dgFP!U6p0(L++wT8WYN2<3NIbSs*8?jbdiCSH
zGWCGo{!8PYCsUs{zPdlQWa}Qo^UHG6z4%VLu2aY!Po$GyQTZ#^pDE9{gXoK*+G*`C
z&=+loa=QP>JI>T&p;w>IH!pwQbmotR%C{w3{qyuzeX7@ZuQKt|LgtU6&V!<|w?8Lc
z+eaqdLh@N?ohQ*3?U$BQ-j|V`^SZVFELX4ZN2;86UTA*p=gpVjaKjG{oRwHeytU98
zNAx=HJz3|Om)}d*`c(GHTk!<FSDtho2bI-7RkrR=v_I97)n1i5uXWOr5d#{OEIfSA
zzP(QWb(4~!<yY0#6Hh>%<5_4ue?f1>JM>xmLw*oH@XF(77UHiKTKz`Pe6~>IiptdE
zp=H0<f6FfPx(?@+z4q}V3-un%llktDV)wh>9MkpKRZBYmJo~)<dgG#YEF<6Bqs<R4
zU9)t=zftdv@MjDE@!M~btv{Na=EZW?+pO_Y@Ab-+CvS4hYip06k$PyI@2vB4*YB_S
z_LDCbq>6m^3bgu-Uuime3-vs%<@3tS+pOrlpQrYz%sC6x^bH=|ezRZnn2~_YH_gYn
z+lphH!#GER=q<!93$1afmOk%z!kdLW=d;j@UtW6AdB*(n%3I^&*`?{$KF^DD*y*9x
zQ*Io~>j$2FR=UN%#XEYdoa&VqOE%Bt#&?}~FTv`URUdk_v#4y%EA;TCsQWmTf70ns
z_w4qYPZJM`mlj%a552aFtoEwx`KM)%rMLL^=8?Cr@$9tL6ZBqr`0~)RQ~T-FhyPnx
zt@X*PSNp+yvrzp?W&G4aJ$GsQ<;rT0%J_#B*Hmx$IeO{?)qeIh7Q%<NYstFLQd!%l
zeo)t0?5-_T`&8EPQd#paZ@FCFbUak{>c<Z))cTMue?(6{3(038{$|A=(k-<1e_8cb
z{6cTFTibeTJy1RC8>s7z##@Vb^g3@mKg*j=I~IENY5#Owyz=FyYx^oIe=1w^0=?$1
zw5)vPm33WFS>tk1+3L4fpH*(jsRwSoVbgU>r+jqnUaxPwdda$HLvP(vqStX(+467n
z#4FAhdB;KX^Y<bw)ch*bFAJ^xK6;HyDs!)9A@SEjYrR3Q{ZZMQ_g1@E^^B{9%7@D8
zKPp@GqSyYaOg;-eezM*(7R{&qRXOW8SoxIiyfX8(C^Anh^!T^#M`}y2&)HS>`1I`b
z`fr_k&=(ySEvNHFW#Wy6nqOs)C#zocS|7Zzj)3<(+j_>8kFPt#!aJV6=&GSttUkm-
z>-#t8tvHU}T8Gfz_{AO9pY_8ALoB2n3#rFKs~mc(9D4p9lZDh{Av{=Um9xg#8h6IO
zD0<H|IagSycB$;mdn;a}x7I84Ry;-T@dDpP(Od7Z(?VV6RJO{Y_xeXZ3%&9>o+^91
zdG&kctDWxI=f!K*9}AT)m5DbNTJ@sW`c$^Y9lgh|#}D;b=#}@{FE`z5U+0_3<(Aj|
zh05jDPkHLq@mJa7Q~R&7_g)U~V_C@iU=|X$E%f58RSvzij-j{gu9O}-iXwLAMfQat
z>#K#HpJYuZPAM*3XT32G44jl$n74m<^K;JuAD|V-&}%<b_IT0!DtqO<e(HQxnSNO4
z%`f7Oh1%Y|o3$9U@0)83QT*wi<L;cl&-bZ?{GC$^Ex$!y?z-)bgVlfZ#A6FR`!sIn
zl|B2tbox;)T7HJUTsx@8LY-gB$i{c7@MAAdd;C~-q4(AmZ$5hIUcB)5_x#Q)@A2c!
z1FyW-KgQ8S=XWgD{6$Z^v5<V#!e#tf`eWIZRqxsDjX!>Fp;zAXKhHj|pI-YOzn1@c
zew#Jj^Kb9|hVK=FjDyFEIbZ%AFTKB$?!~pD=^ysKc+)vIeUW(fTm15z)hqA$MN#`H
zr|<u(?5#VVUy#l|w<xL|_?tKHym?DH&vTKB;=b#gfBDE+AEb)bID7NjDyMp{o~-E}
zpH}*!N4p;Q_RQ%C==p8A>G+i*@0qCHlC$!n`G{YYM&4ua;vVVDV@1*}^zOs6rqlnH
z^Pjx+rHAfKJS;k|DChZ&6>qcZb)3C&)^in4Pk$5{N1jJn^9{YnhsI5nz5ZGLgI?`N
ze)-B@Juu;#Nhzp*p9A|XwB{>%s~mcbkH}g5+8aM>zl9z<EcEQK+PCf@&|A;*RIks?
zk>SZh`l*N=9(wf;IOMv2H0`y{5D*@*6aKz#^U+>?+D}jWjW~J7E{DIHE@-&8;o=Wp
zPM&*u-Q%u2d0etrqpx3|K6_FPzVOs<j{Vbv)05jyp7h1`C(o%t-nT+;<s;of=C6g!
zw{sS?YBP4>s1!u6yr5@Zv2Ro)y(rR-g=&||Ryoys@k;$vW&A)<^W~MbKb4YKY5u!4
z2hMmQwNUvWzFDaKwq)-<rf9m_>&ZI)dHqv)$Xj0Xt3AXM)_qM^+3T;B&bq7isBDcV
zdiGnC+iTs8`i*#CX6m8xs512yMfL+G{`)@2tKW*7=)LEDUi!peuJ_#TQ|BcX_G_?5
zqy4U0lvp@-uVnGo-M&mL^!kUN6-BFE^xnKy|Kz-3p*f!azJF0a@?@TG<wcJ#?U%|{
zd8^;(2YoWX$KT)lDg~{2vg*(JW^mJQ|GswVxN{b^Jb3JC%iuEyO?&YEj_Z{y<le(V
z&;OJcm5D1Bdi~@c#X_rI^j`hed-d86OV)jp${v6Cg@yE^C|Y)_-ukXYRz2&9h2A)L
z{quPA(!DrlrF-WNZ(nG=|CCkFxx_-<PpGWptg^=&>!gKVd{%o@_SOT<pI7$$ihWd3
zw0J?!InY8Y?xOeX^7!}0*W=$Sul}TR-tp4>TA#{Z|M@OUr4hgK;u?Nnp=FoyV9DP2
zkxqGUJT=`~chOt%9z8r+NIe#M-zBhq2aMh;Z?#*l{^s5Oy(8`ayZ)!1v318CJMPqR
z)hqfR(Z6GZ|N8fgre|z$MvIQCT;BhTjs8pCrsJv|SMBu8>HYejard9vjY(F^yPu8x
z!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`
z1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9F!28|(C@GZe|l@f$C7gKudTk>WwU2LO#)YJ
z9?p5_2Y=e7_2)@_!(ZN2db7=*3zBj%f<`d#-x=uLZ|z^cJ$O`NVbW*Qt9tF$sFW2y
zXn6Z$uROaX@o@MXQy0Cs;hZEe9}oY1?Bf5yKrj#t)F%eIPW;h5KRM>Zq*^%iy^rRt
zap8MOxmce#4O@eOU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+
zf`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZL
zAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>
z27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?I
zU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ
z2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}
zfnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl
z7zhS}fnXpQ2nK?IU?3O>27-ZLV8vvh%VmFfe9g1|l;p+H*I)X}wmbYQ$%;Sf{KT&w
zSoB^}TUg&m>(>)5+T^*WAEh3)y?f~Tr~PYU8W=o;AK3GtzQ=>GKNtuGf`MQl7zhS}
zfnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`Qs%
z;MyN<cwA|fkCRH_*KK~Y!w1*AofO4|&mMl`ewSR81_lq|2em&BYAfzUJHbFO5De5;
z2A=Ew+D;wky`3u7*Zzdv!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<Lwg
zKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&
z3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1Jz+*<TZn5
zcOO3~4GbP?!w;(C#ZhZ85DWwZ!9Xw&3<LwgKrj#t1Oq+;ZRa)^zsZ&#q(#whA#wx*
z!9Xw&3<LwgKrm3Z8F=%A!*A?-%}1$)-?x6`+2=kSlY02)Kb}6ZbI<W<xmdT^3hRP_
zy3N4BL*E!R>+I*!a`DhpUwCKqlqb``;3112)RkE=`|6_)*#B=ICRwrN_iNtu^#==*
z+Qz!VSJ)8@1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ
z!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr
z5DWwZ!9Xw&3<Lwgz>3bmoTrCOxNz5x({k}2O+MW4<B^}FbrlbP`n+xKd46JQVO_Pm
zT)S3u|Be0y1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<Lwg
zKrj#t)DH%}-eSUE_IPP%QrEG5uo+ec1HnKr5DWwZwadWz(;KgQ<{l%{YU5|gRtLT_
z`n|N;Si9^+Yr#M;5DWwZ!9Xw&3<LwgKrm337}%l7wWl^XdU6^VJgiK9P?zRleYEAx
z0ekh`@w^Yx<%Rtko;zy9lF5mOt<RmmwCO%AN_p|*Qkw(5?a`tX7(9d@toZZb*nt=A
zcvR=HX<l6KxJT}5dDDc{!#1n-KjXOt<I>v3yuZV)k84qF*u2Zz-yJ`E&&!hfjvbbs
z(f+k_-bm{!R{Qx#wd-GA`=v^^+o#j%U(SE}!?eDk7pE(oUK_EgHrw#llXAbuvZ9r*
z&9PaT+NyS(w(4)%zf<%1w~kAe7wQ-njq7>$xpgJK_CMrcAQ%V+f`MQl7zhS}fnXpQ
z2nK?I6^jAy-l?vquUMl|`@N|v?<duM&!fd)AQ%V+f`MQl7zhS}fx5^*UD=;nb|jN7
zIpFXsC#BWKx;QprQ!o$=1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrpb}8F=F67yCUq
zEM2A2xZFDy1_T4aKrj#t1Ovf9Fc1s`1HnL12A0?F<BQrG<${4=AQ%V+f`MQl7zhS}
zft87Yyzkp){T<`V#9{O`7zhS}f#u0S-oO90Y0r12o&V(ZsfW*Xe{H9Z^WIML;_~Do
z+71SSfnXpQsJ{%{zu>SJ{@h}8S}mM9yY<?S&v+osi}jb$usj$D1}p~J^>}Q%b1r%{
z@$l;{Cj4cOmxd;J(P}sJ!9X<`sPyyrtoMJB@52p09{EYSqVRC|)8}n-&+`-0YGXC$
zLew1$1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&
z3<LwgKrj#t1Ovf9Fc1s`1HnKr5DWwZ!9Xw&3<LwgKrj#t1Ovf9Fc1s`1H~Ay{yubl
z)%WRi`j_*c{xG#rzi(mXQ++XZqf{^u3<LwgKz(JPzOI8|_qC@sIC^p#JOmHr_(6S*
zTVZ$oU?A`Jg4R3kk^5TSG$E}t)(^JA%3vTE2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}
zfnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl7zhS}fnXpQ2nK?IU?3O>27-ZLAQ%V+f`MQl
z7zhS}fnXpQ2nK?IU?3O>27-ZLAQ-4k25#Q%->vokUH?<h*t+9R9e3`y>J|Nu=-;uy
zfBkz#(=#?WqeaJ6F7JQFM*k&m({a^~t9JV4^nU%%xcg7-#w4rNrmfLdFi<fDc6jdL
zqo%YRohpudcAYDy+%Yk=@S^Sm&pW=`lvJ@|h8Y!E_AD>GVdhGwUb@41m$aYIury)J
zflpoE=O2lMi?-OH|DY2_C!pqg=;x2m>-ze$3HX=!$KKv_{>h1j4Tt^yvcZQ;PihNw
zU^{I2*eQd?&q+Y~1)@JNIp^*}pZOvIhy1!U@XePeCKl%H&u{O)sKcP!ze_;I!@?I{
zKYsPgCVrWK?|wDum|frhCIL^~WWvp*%cmqD{jpGaQCa7wHJ<2u-@NwXwogBkdZ_k1
z`R2H12M-^SDAK;#Ro~@S8~p6FQQJ*VLHMJ;I=@t=oTBnt+j6fBChj`w;ql9?FRVA}
z7rK5S!$YOf@*Czm_>&F?o;7%#C5eaRw@}M7&Y+GTGV@u{vcFP#E3OsQ=Z%M1^Iv`Y
zBfU#MUzpTYtkif!J^woc$MqR^<M8GUN{Z?avp#Kd-J87^Bp%KgaqwYV-9I-`-238#
zdfoNu_(W0XgIcNUSYDYpNjz0-zv}PWe!u9AG%u>1AqNA&K;;-9&SagJ#F1{lxuD&f
zgXg5>BKs%Bs~)~)(+!?jrKDK7p08LnS^E&(hi4sU&8OpI$y!cjs~mdWe_FEkOJ(XS
z7j+%e`Yhhj(|;}J$&@P>XRS4C@zu?yCZOf@J9OD*?W@mUtpqa9E%fZS(yjI@rH8Ld
z)fe>y1HnKrP}dpYeg@2XezW3xrS(PmQ#-QuOY><xDszt3a(Yju`;4M;Q9dZA`(l-?
zeN(0N*aNcu;)i*W@)pugwNS@f<-GUP*xByNzn{40S!<U-y$AB-tmDT1O;PLDepO4h
zo(t(cF|yT9cnk~%>H-6LZ_D#mQ0J3nvlZt&ebIG?a>w^tyx`k`i;}#^JugUoVAgoA
z>ujZD>psfjU-edb^m?vE4>E4ZApBGtwLc*T1Jz&vKlf1g*?DE+Y2g2SuSfY{j2Vv|
z=54b5uiMW^XCBw<{CQW6$c)re*YMT!<0B89KQj%i-#p--%L}dbx3=uj_~*?t-EUZO
zm+S89z4KkO)7nDb`&m)Z8y72`=RQSI<9c5Cp>vy^vupENNl~=s2l}k`aG#bJi}FUf
ztoJkYM^WczZOeLYwB*{hD{Fh|ANu?><p2MD)0Kq>onOdV_oZ3y6RdpJ{IT>_{P6r=
z-}}}5gjKKR%PU*`&Z}o!@?u_l>wEsZ{N&BAeq#9@>s?+fx-Zaj)gDJ{erDBM`?Jso
z1HnKrpmAB_imso*STGO_1Ovf9Wf`#UGd+FL{ij;b?~b#gwI5&6^~&3dD&u#$Kd82>
z?`NrOy%)vr@D;81u8Qh)f8@#5e9(K6qVkI3e;4lCaoFK^%}Fh^{7%QaQu2zz>+<S{
z_a0NF?BYGrqFAoK==mtGtna7Rw)~U3dktD+#JHp=YJAQr>-h8ine}%&>+b!UOiMw&
zKf&`=^b3FYVzc!}O->(Lb*EuJ8vLK{age^(i1}YM=svS1o%cYVnz+Vw&yAX4zCY37
zvA3Qb(05h}(k@7Q)T8xcZ-;Y_p0wVPV^Y#V>{489lg%5p8v0&(@kQ@<I`+)B(rZRs
z|4M_&A1(7eDD)T4dv5VrZ%nB9UI+EFj=%#vv7VED+ZJ6LTy^KPRPCUjH8^I(^@l9|
zpyqogHG26D3VQ0V>$t-S1I8|RbiGnm)O|MNtO)Olj4#OeDKah=>byaIZMWCIJE_y0
zq$vJ=r-%D&H)%;y6xaO6^Mn3!)S9KDNI$@v_kZ_*Uv1a0R1{zO_K;R<9Qt-z6n|2B
zdyV@qxi2k>eCMfJ2v4<*{63*5di!NJ{oiqw_?j1W9_N*HzUGx%{p#!oxA=Tf0{-Zr
zn<uQh`W?x<llq?2ZSc$#Y}D?zpB&tNdV2mZe*M~`?PsMk&c5RA9`k3V8#Uc_&}D;X
zrGNdr@$lE1%u4TAukD&YnLi^XKk^axynf!b?PsRn9`=V;UHkrLy&HS|V`ImRnpv~H
zvCegVvsLTc8c(c=^Q>>W-syXX#6OMyw?B6JWdpwZxMn>g9(LPp&EM>I&A9aGCA*&X
z%!J`7&+XP<viP6ROng7(I}F_?o%q?9=5uSlE2ift{jLl3D0X<I*>0~6UccmF+lAv-
z>$pwxk|OqaSakoHb^f0I+R?}Ec+2{wqIluK_pSdy-*rnx@v<MBI`zggS1;v7`lrux
zk@N1C$<OaDJXAl)E7P8Z*11&m@K7%Hdg7!Ry&J7nDvCNkJh@zZy>p<Kp7s40&6m~g
zJw88k_sR3tC>2F7F62$mYIkkr%ey{m{?l4rcg=M}Hz+MH)c5f6%I<zdzDKJ0i{3L<
zs$5?GBY&m*tK9n)>%O<B-s>0V=oK9;9$InUFNhbt_};UB{qa{xQM_U6QBS2;%}a{n
z)^~QD_soeOCq;468jtR`_kA;xqIhK2E+=ffTZ_`lj+TFYw^{$m=l1wIDT>y-C$4X{
zc;8hzb!b{rB(8xPr;%T3xAeA!Q(Kf2`%Rg;_x4{mEm1xzK5|ay+c)~*IwkHmLEd}j
zo{aMm_dX9EKjV(w8^4x-?3agMepR~}jowczw0O^|_x6KD)3fqbZoa8)4(jyeu8m7z
zQG1tHxkmrm_QFQjv?vut?s*lnj=$!!c+mG?t$C@xd!fGZ+kLU?R|_X69&!(m6^R3i
z8*TXNhU>JPU-<Vx^!MI(b{aXi?b4*usOQD3vfhgz>o{hW*>~35-~9VIL*FC%_5S|(
zX(_m8kF9>wqvt2-D<i+}H0aLh>7*O~dDFJ-r_~_ODZt;J^XRi@+&MK>WS{lU;~zF_
zH+V+Py%f(SNC%0JnvP65_g>V?{S<MQe%4paI-Z)(^Q#^czW!o^b-qlxG`Ql&?G}H!
z3=Utr&9Gme*Q}%nkJh{>8qc&`vy;9V_0{loN|i$Weo0>0x)-wa%umqDmshXxuBfcv
z&+=sUO+^tNiemo`r55j<ynd<Nb0+&TMelx*^{i5O_oP22+m35qsuVIW%f)i{Wt7+V
z1C;M_<-GG(<6~WsS+A;voI8g8{pDlkwV#uMfBWdu_kY=aPCDVD>pr<~`#Ckgqvtsk
z>E!>a(@CeT(R5DQzwr+aozQnq&GRdsV<}qvcG+jiw{NxnII+-*57xYUy7Qc${&VNq
ziK5oSzV4tCHs53SwLVI8->lCI^nTOxSMIm1c&>3;`(u?OANFxR1g&!D=_mGqq{9nH
zdlr%pq#g^s_l%WiOQwFtLH9Rmua3XUdS20VWOxQ?Us3Hrrk<>*exfqx8%4&|LhP}y
z`My1V)a2nb@zCps@`~*FLD6*lg>^^KibJ0Mo^k8<9CCl7Qf;HPzp}>7(_8%{E?CI(
ze%6oWg}i@)J)o|amA22~Evw!e7t*uV7x^m70Pnl-Ud+3HdA8-WCd+<Li2DZiMJL?w
zQrk<K{O3Iz$}6gWa<8H9)o@QidEUp-e570W%(}M@J^Pj>B@gjC3t8vbzw5e6InH6@
z8}Bh}=ac8x{QX7BYaBah<m!_T=>Bd3k`H`h|1sS<zHm=U{pBL{QV%lc25<eQe<1r0
zkab7Vi$hkt#Xb+|Pi-OnEs9ou(Lc5RnN4rp`h{iwP8NC(d5;>@@lZK0Z#te@pC?x~
zYWcM>zby3p%S-qAo0nhmd;3Z+-Lu<Dr+v`tAL-=Ji~9ajQMu@Rq@3?b<>k<MgRJ*#
zw8HzYOHRGL<<r|uOXpnr+L1qOGPMTLA2fRF(;hf*QrhHzYX;o${N%LV`|A!ma@5qC
z=M(z7vQPi<vx_=!{B&~0AD>zHyHV59(f5vczfIriDbEXTdjGHIA9>`sbhXYmjU6~|
zN;=|=4TgRG!KC#0ul8?u+Pq0=^UV)_a8kD!HNT(WJ>)gt{o^Ue{`Hge?$*6t8rOKT
ziM*%F`j7oQPhlUze&G*ye4F&&@!{muPS1Yd?#OSG8*bWs!NL9JB=gVzTia8HyqhY%
zIORtj)@(mD1$iI%($Z;vysyd3RB`t&cc1ps^Hb8so9#U3&-4GY-{E~>3!fc6?WSY5
zpI!6Z<`+*+?Kykiw3_<4|LHXQuiK7VqicF#%Rv*z@Bc)4#ZmXPJ#4$lHSZ_$yk_55
zMlO7C@a**3T~2Iw%c$uoc5~0fd(SuTe*U-951f*o)wBP%8*M)`-SWp5x4iVgiRl)r
zY;o54eP`B;3v%nZtMuA*^z54aYwvvcw9oFGS;G(g!w=vCp1$Aj=+<A4nw8$rYTpif
zZa*uX+V|lB@7y_~=J$$Qw>Y8I)^|=$w_R<Y*8iM0wFb36$+Vw5_;{bm=>f0Kd3@n^
zGt!+uS+aA}yJpq+6ZdPh%RckuK6mW1Rrguu@0N65m{n$;)fVbGJgcnx#=LUgc<ik!
zn$GV4n9s~xMQgn|`^0k^yx4bkO}x75XMOfr&}VAO`D1Rcdyg5?ZF)+4As!ui_=cZ#
z`EEk`+3N?7Z9V?Abl#t*Ox$<v7t^J8u65Twr$1B^A4s2j%cyPcxa{vq-tlIA$cuVT
zuw>RxkaI>+%-cWq^F@*N)GtCV4Se(EiAh!-be)D*Q0Kk+leVMzRkq$+$*Z^GnDyLE
z^Lut_{POHm9<s{HXI7c<$%@+lO37CH>@z*g8{afP=ht$Pc|bh$p2KHNx8kX_E}{4Q
z$9le?dd5fnS!Lx_Wq2uy*rDeROD=jI(|*;Ktnma{+xPhM;!NIj?%6=bRqZM&d;P(m
zvtn&M=fNMdBD{fKzq8hB<;xn+uqP|xmqpQwOPZc_+&7r?YOArgj7h4E`n`|5^C>IO
zdGjf6p8WTQ2mP_vc2jD;+rhofTfbQA&yT+Gahewo*m$FPKl^-f!h0r1b>4a4z#d~$
zekZ*5qC@X)yxqi<a}hG{r*e-%`i%z;SopJf6KlQ)()F3|w%fbg<dpjg?jIDb`A}5v
z&C|T;S?7&4epY+LU#py@FLxhj)r&n@vFP|LuX5bGF>kmZ<UOn<Zw_4i-f^pzxOe3K
zi2EVX;>Q|i)n|?KMe|wqq5s(hZ>Oh!_kB_nTh3|S_}mtaO0R$a%duV3H`7&q`Skk@
z@0^t0`{X72&F(QiJ#z5povuFg-IRNi?)^u9F|yB;n)Ek5Xu0OMji=Pa1@4iFEBtPU
zc=PcCv#;$pZ*t8&G55s_`b}tjcAtqg-|ykQbnYwfZ+6&g7j>IjbDzjPH1_a4hR1F=
zrr{;mOh`fQw;6BlS3jS2dBgW!nVEdK_Lk#by|zJ#c>nnPA79zL@3fTX$J|G8-^qP9
z_qg~Mc<w3JzO>H#>1pdDn;f=Rqei6{yRN>V<$cqWmwN8<)<%E*sOBD?{5%Kuka@;;
zAh3`4-lX+uA9U_=ecI#lcOQS_z{xfDt=P-GF6TM!-|=VkJV&SB$kfX{DfhnU$qzE_
z)_W%8qhIA>(fQ?-gC9lg&x*`fMJs+6)n{GbG@telS@(bF6}^3!_xnZf`wZ5(0>1^-
z4&HlG-jRnq^knbD=8R0TBEM_Qi`Ko6=ReFhMdBG~)x-Q%BtN`?%v(^;_r#5Ap^jgr
zW!-NPuZgFuf2?2dM7(3Y;&~J66zd+p*W&kGJl|qnQKTH}Szgq*iNBYNTCd7F|5eue
zA5B-eXuQKNeZLo(d0rGXPO7Z?F48^zt>-3`w^04plCAnxZ?#h`y=8A!J>#7f)z2*1
ziqEt|e5^L|yItZB=Xv52&pkQk5dV2z&$)=_^}KJ%{*`rveK7Hh^Ar1Rp4W4}=iJYG
z2E13ozMlOo_ZmFk=X`L;(ACdsf9I5x_e9QYyTy;MUphLyZ|p~(e%a>zw84U*M>YR$
zT>8!DkKFP4H6N#+t@r+@9eRG0QXhKu`@F}(bKOTaJhS)k4<@F(hrs&|<Og}rg!e@#
zPk%kc9|Hex2CV0JMfF+Z5mphG_4g3`-d5j-oPF&-xBl$5S8DF#crO9}$1m6qFi)6A
z%tPiE^ND%NJY*fxeKqgVvJR4ueHiB`WQ{}pz8v}FOKr!dAoB*huoJ|7koysk{f^pM
zRAyWPgMlmz@c!D0f>xi4&ZkE%c;i=nkKDMVsK5Vb9}HS{czWxeg!i~WjY}$1kLro5
zAp10A;@--F@U*<p^9Rzs?@g1hC>EU`<(4aIZ<JfH8Q}L-oWs~(f$aZyF2!?I?%%nG
zBd&4pkA9EI&E}lcWO@p+5957izFSRx?whzrLI&>`{ravC3>}?<-sqG&Z}OHe`W@)J
z<yil!g<0<}$j3h4L+@U~8&5C&-PMLJUaho7skTw?p^D197p0ij53T&(xs!8RQS`<q
zYkIkUt@r$voVTBzf0J(UUnxE9)KyeJ%R25>z6Ce!*7b{1S1nZwwO#(+0;uOU?nC<R
z`130d=svsV?=i5Bfvl_i4v2e8?k#wpJ?yo`jqmO@t>*7JaF4<I$oki^?d8W@-)B<I
zclKC!dEbEbnfCd;+O9YK@|11g7+3TAkxzE{rs*w5PDo$5_1g2d=rJng@6zZwrd%0b
z)gL_B^J|^AJjb#y>$!vXCOqUk2U_cz>i^Yo<KLV(Zc+kLUQx?gaRYuqD<67o2ifYc
z^6bgfr-*%(M()S+;-9vBamm>&ze@HWaQmpY2EL!3(DmR`7Ec_N&e*-hThoV*Nk<;B
z<-y;#9iPs;c}k}Xp8wD9K6&4V=L{h8l;@-DkD0eTXW$-_`vsmi@SOIPhZfD=al1(=
z_x%@~H>~I2#*<Ro(fQ5uSL)^cl`%U`I_iMQpQK+l`|SFQCXHPNxqoeY#FVwaKl;xJ
z&oMywT)5Xk`&~J3Wcq`xx}UV`s3|5=pQ1JIytrbO%d7X|4C(ka&uMrbbllE!pZsi-
zk5iui@E$8XzcziZ2FE-<G2La2E}yjQHmTxh&6m79Xnv3XYNzA>7IH4liq`uZR$TM+
z8ow-Ad9`H96NfF-@|085e8}WOt`=JRY4jGqo?lu1j-LKmNIq}>P;R=e&&XE2*kz&R
z4@LFFv)aa@_EIjdz2*9M-uBAPkDqu5A0Ar%lvVH9QD4*Z{=Fe<-dg&i_d8k3W&J&8
z(R_z&bIQ?6HrcdPU$NZ#A+NmtZndavy;sY7Bg|9gndNWjH6G}DAD*oJ*7cab&uF2Z
zueoQa7FzK(Z+qNRgVyh$Jw4-QA@wWjxFK`?D2kRHs>jbNjo$jJ{)NoF6^LEdzER)5
ztu5J#*VwB(S6XJD%e>R~1xV-I4S(1d7z_jh!9Xxji~-^+>lVnm1rmQj^t^WqQXV8<
zrI7m<knhM;3b9*}bqT(Hd)ws8fA;dA1hn>(mD1C0rP11dR!g6EoH!S-?t|<v6**^E
zsQL5Cx-VlsWJUEWZ+}@dz36_su9o9m$9VDmKnvMVDlYupi_O*_H96()zw!6mEF`M(
zz8Ki;Hy5;fbMTxLBtC#VSF^BE^B7(f^>>&$pRpU%{K)(cSp9->9$M!MORv1>IaB+=
zJXgKSyl<rXqO$Vg*{%IRCcb4w`de-Ec(KNhzt;mF8t-%;K-}cMLuK?@PW9|RK;|j&
z6B*?9WX$inimZ1Q!n1|iFK@m^`u~jU-hM^>JFCoi>i8g2PkkQ`eotp%m+S89z4KlF
z`5w0?^S$7r$oE)4{rkMu-w(*D&-(tJ#&5<O)Od#;B;AwAS83$$pn=#?E>_x3>SI4s
zTUhCP(%P<%zb~$+`yOQP`w66%%fH0~akE@x{|>4@qX*fqc&P6IVW(A&@9-;fo?-u@
z$i4`~|3UPiwog8ge6&N{<^5j1Tfp~K6!{JU`z~bi>A8#jT2{oq+Cpoe&b|}G9{LTR
zpz^LV=T!377BbHCLlM8Hf1sXo(U*(tlRVUW8TiWTN1Bh{_j{=AYJVBWtf=vZ_b@^4
zIladd>1wCSI_}7t5B~=7ACPj2>hH+bdyeS!K9x8c7z|W`0qfp_eU75F9$0!UkIelY
z_nVYgd&m#!IYMRjTjU3+mv$^9zt)Rh%V|Dj@`IGaj%p+ATJ7LFtQ%QT<25qt39=&N
zjGp|8<g*YSJ@mfUsOkFsM)>p4+MiqYu^uTh9+g7x9;d#iTmH(r3bKA{e)OC#NLM{F
z&s~to&prpgW?#U*Uoq?W>b#{r57n<KEfe30V%GjgzW<#8>)v}s*Z=QqEUT^5{qeFj
zS5`y3D2lpnTGlBatgGnp8)OfOCz?;=6*Bc($T~@WMeHq#T3=TAh1ZW?{j!N)CW_eW
zp^k^QkH~75m(Q}xDp%D1w4C*PRC!Pte=3T+7nQX<_c{0@<=_oj=NB^NK)nw^#x4u_
z?xT(i^?IoMTluj&D_ZrSx9WknqR2js{Zq9NKUF_Ork|kNq4zq-*w6lqeImc})_N)>
z(~gDMRc%y%RGE1PlFmLIL=S3y^dQeq;gjdC=<!$XYiS20AISHyK*kefyp?b4QKTN)
z<DQUn0kR_PSO_l`@^?2tYhGe6{%)a`xAJAx6ZaS&kn-5&A^9m!y^6$l@-ZHu*AD5l
zhpb4tg=&|T554ABS@%VjKhST`%1689BD`4j;7?XMc7obZWUIXDiSHJ&ZU^={f8JFi
z(iM#dYuqcPx9$P!i$3dnJXXF+@mRDT*8Td5dY-P7T$E3*T(gtD8THlhbxQRW^*x{Z
zCfAp^va;C8J|!=D^DNTq76UwAuIp&AX3a;wlVGLi{ho1M&tL8D30K=c_&rTlwEUsk
z`k`lDGW>7<Xi%yxtag8*{$OBbU|@r*Zu#Wu_ZBAg73Vy4!?csu{VoX%9_peWcroOj
z$G189)+5$01uh>R+74^d@5AkzmI8x^@PoQL54<>9H2s)Ajd}a1(Tz*>6)Sar9QFLq
z-^T|Z!N-694$g|=4@Ki-lnVxefnXpQ2nK?IU_cpY-+rsdKfi9hQsDp1gYuSF&iWj%
zw)5rfN74LCMqfPijmu{w)xx57t?cDaUAo2gz30wPDvfo8*AC6c-2D6#tC#94Zo5&#
z{hnT@S*gBa(Z82dSITvGrP*$;4qm^sym5KWr<n&|w(7};HY%;GxV(5@(c0zxwQ?~l
zFGat95aoV<+2BK_Co3usy!V*ZkCu17oA-bBfM0Fbuv9Iie|L<2eb)zuj!r9ut>>=N
zYtzxQ)0GY36SRJ(#qaMF`8xpoewE+Nc6jWqX9x71l_K-IQ+}t4tmvhaAA1$`eR-=t
z*6-+x>aBLr>+c6WnQ|WTJ7N!Y+=|McJytruTV7sh{XSLoe1E@E$nS(fZ``eP^=p;+
z{VGU%AbJmVK3TG^x1K-g?~GMm{hlqmUOVQ##KMn;G+pbdHZLX?T7HLK+f&*5E~w^L
zncp8<sQEov{m+{>mfxV~cj*>-<EQ?nvKN2&9zvy1{j%JBI_1kf$9wb3OSk+dtDf;L
zidoyS@_GI9(ye;4>b?5CbmC=EEcZT~@4Od9{SIYbneV6;Men;*R=Un-m9<>f@$urU
z=MQ<)J-f7DS?i}>&5zzgiwErS;=h+(G~Se3PS>Hlvb7%8mfjkdta|l_+Lp`x-J^aF
zFRz?;yfy#NE_gdV{k!jzqR9TTC}urxvOlR5R@zUp_QT7Ub)6@lw{Q2-bzQM!+VhaV
z@1fZCh$(A-fApV|thnLbBX8Ps?iY!L{5=B;`MW|E>i9D+_^z40hpO?A@4t~x*9Y!}
zvZA+sm7C7_#2Vjn^~z_Z<#PM)mDlmBmdtl5tBuw?Eq8xTdA<Lrw9LHf-=Wmvy_453
zSy=P;RaV_;*pCKJPq`n(-}w$8_p^*E>55vOc~3p$=U&%CEzft)sfTox(SxL0{qXw5
zK3CE5lh7Y~d(-(RC&5ED9@MWaxzcfh`dGiSqPHH9?(Nq!J?pupQu!*q&gQKTp0Z-z
z`n~*KyL_M6;@diZTmD-qz4cu^*1bw0>jr48SFAINtb44B$ULXyUY>hx>L)))K9KSt
z>zX3#nuR*w^?4TKiyf6h?7?1mgJ%oLr${{@=L`5$ghxxS>F})i(JRtF3;8=OpyuNo
zrF_#q{Lo*;qVr7cs4dyEcb!Maynn!t=O=lw+~+Y~e5`i5ch1@W`ptH{@7|3{7G8Mp
zed~YFcimD^^wtkAz3Be0Qr|1{?DNuV&PP9gd|ubrpG}I&*7{jj^}PSf_;C+IJmTJn
zxI;W9J|UA%IhBc<q*I>n#8VG(i1ia0#6D!~23g0}yS#a;2QGg#B_Hb_GI5o4P-XIg
zS{{3~JbJAUJqSP4j~@GJkNaZM6_pR#rQZ62JN@FHhdnyurbKbii#q?j_f8WNMQ=XA
z8>sUcy~mHX!}x$!+@3Sy;KR1Me{Q14_<;I%R#n#e^&F+|f8aNu_FHA<4Y=ASn>TDV
z^u5$VZ6DdXKgN!{=<O4Vrn3*@eu+3kJW?cHQLZTBFZi3zW76>p5PPsg-<KpE)O4ON
zXgcQ*kaUoEUEh#(9)9T$=S0%E$LF~Z^PIR$TxI>E9@6n25Iy-AFZ>I8$fx70<&Z(f
zfqEHN#+z{jHJyCaM?Ul_qX&r(^h1$2pmx$8Xw4sZ;XDLCApJ(BKH9@hkaQ3|_h9$|
z^G@d_c2J)7ufF|}-ld-}toa=Q{boJEU+@c%^4P(-6Qq6YQ6wFtANZ~2M^AqE#SVCc
zKlJo>#@Scg-DCc&6h9(<!wdOA>Z3e*5WV^h?K4l1$<N=hLVxkR=N6y!#)K54e;_<y
zZ+%DRU0%%EZ+NsYZ@bGoKl_3mx@@!d)#tBP!Y|+hSrNa)4?*S)^O$@feviy~fcnvc
z=<#dHbB_go<iig3ZPY_M$mAyuQ7=e-ctEenyaSoP<bw}I+NT|4?7>dTVF&q0rycaj
zq@$-EO{abArXPy<$%=|ccI|S)#=EsBtt_~rj$`y|#bCht`xaUC?B^_OyKww!9k*#-
zvhbd9>-QXTf1{G(iotI5Vr60AnRRa+diE_%N-G<dJHG5I^CInX?zND66p(TjvX2AV
z57riP9|Ur4RJ8aaAN}Kfzr1Mueun$0teE%tq*rg=epvb8$wPPy3<iRMU?3P+xftL+
zUhi{M=6(mgBKJgr1J66Y+mzJeL7y9!D_i>{^p+jgdA6?Ty?x8_O0TWxP+R-)Xy@Uh
zrnDTLu53K;98>$vb4JdGAn~93f8q-1)YI&!hu)e!@2v#c!rGcgD|<VP56CzwlFoYq
z$X1*sy<F6NgeR};<5}tcXI=00o}toqM18?PU1GqCYux)0A2~-UYM*#6ugHEqFVddQ
zV=InozjzN^k#$@1t4#lE8}ssH<%hp&q1CSHmDjv7<5g+QJFfhm(L?VZoOI@yBKuc(
z)O3|~e$<vsKNLNFil%2BFO5sc%9|%=^*1XX<<;k%r}&437GGKCpO#nqls~Ut%g(I&
zynd~IX+0Oy-$PjTC?B4z?W?T&PE98c;J>wnYDZQXdqH^QJwuRj>^bbPao-;HWKvg9
z`L+DT`dv!so1L?3^I1uydC+kQxw;HwUH3Gf7Z2-OdUaW@z1oZN&hJn><okuheG9X$
zBjjV90M*~}%G6U7^<FWrtotMEP(AxBOHVrGJycI)9R|JfR=V0}`5E@I9{|Zu{fsC2
z<%P-@vifzoe`kbx>l^C6zuIzr<0tI>zZtOR_5WRJ<w-rT<)De<_kSYwkaz^2#MQh=
zJ$Z5E8IkB|Fkmylezj6q?e(<M_1nxuN-$791J>_D^6I_k4qBge|7ZQ304<Mf&9_Q@
zr;K_U{cGC`8(q_)lo$0L+LFs3$*3V12nK?IU?3P+{tS5MWbP$B<ovBj54qQ@Eu`JL
zj{HuxzM=YqwZF6U>L-?*b^JU(w(8HSxAwuRxB9F4to^D~K5yUerB}+1O4q}<SGwM6
z)o0x=;z!kLw_5dPeZQ+(?MJ=AKz(F@^+4BAWS$piMdAlY{gp!NJ|wHYQg&!P`kp;9
zf5)8f3-W!?c_;NfsoUV0Dc`-+?_#$aGvfL~mVS`(zK5n4m7myuOt+3N+><KuU247)
ztBC#0H$V8nN!@0or0@I6$b}CMo}Hf6v;VgnZ9lUH`99yJrPKa+Uz3@shfP|a_Ce<^
z*QZ(0y5HyftS#GKe$4fKCZ&pek9BUZdyg5?ZF<^o$Ddz$K=;`-{rTHRpT7Ug?sL+q
zeIFk1&Yd%A#)on8aLs=_Kj<$<ty#*7RzE7G*YA(jml$Q;>lW1$k1Wg@-y+`z{Rf>e
zI;rnG=)6E?erS9zSEl}~XzkBEJ@!`$FWk4|u*2_~lX^(ptS{KV@edB2(05L%==IY|
z*S`Z%X}P}7x4L3~&*w)xJ8j<$OSO&qe&&LkcI*1ZsjHUs-iPlZD0Y6cRqNXtPfS6r
zk9`lw@9P!W=Yf1*0i<5tukrU1D7Ry?E?qX>{$B~m_pQOBI`2GiV2`n>BK2fFuV>B2
zc?DEn^UC@=Bz?bFdDZcPPf*iUW}mM}c~I}I)h_ZW!XtL`ee!EYT>nag$sg71b5D6_
z(d-?!o0O8S-_^Ec<(F~aVwEk<I=}DCRP99O9HV@2e!>nA{{-nL{Ue{^qFWa3_ww_@
z62;%%Hu>_Oy*wzv9!2F_W$yiW?w>c#S?f9VfxN$v71?K~z1qLpl9`w6ORRfdJ;$qG
z@%>Sd`sym`d|pvx{L4cf&*de1`}cd_*=gk5wo8+8@%Ubg7koQ#QDUK<KlQt8`g=Xi
zk4#*!FmD{nn*ZIeCLOcu``;v4(XubE-inh&_1^u;cboN}d~T1glf0Of$D;Z2&Y!v(
z7wei`y3SNv)^(YEjE8lNH_IOQW8JsVng>??ta|ok7HZs8S^ITdpK&)1Z{DDk74yzh
z;tcItcEhWO$`kwLyofzu-ttBBbI#y*k;ow52lVdaIWJR=`+vonzr(cZPQ!jQczVit
zLi5p{?whd(<oO5f>bxf1L+nt*4(y{Io-cv4k4!%LOTC=i$q!Oak$NoD{=$o*o<HfA
z=EGhPyFs2$DUxm>?ZBgj_<!Ka#Xx;~9}au7j*_l3UcA=0XUSRHU)l2M`OcF)emK{1
z?+fzxo!MV8uB5Yn;GRzJ-*jAw=elpDUPap7;e-KW7d*ON3AFA%)DMyApT=+Q2SM(o
zLA3{&zZ>u2zdCOGn-j-PN-Tsw?4o}7K@XB{A^iZ!$2d?gdPTjj*8Lwc<**-qIS-RB
zFEXx-yCUaw@-e;^>iq)cK<+`X6P}RiS6)=Rv&vUJe9fjCJh4g%%v!E!zNNd(IPCEa
zHYkCNXHit1^nR9oxPHGAxx=|fPg?KDF)7GC4P<;ky@#iMMdgS0KUA+W`ItAL*8lX9
zT~B*v!tfL%ABcbAcg!D<bZy^~>A#263tIV!%bKqClCJk|mcLl`t6ur6YqI*a%BQuu
z?wad{Zcy^@k#joVzR?fYDS;ZFknx+kitHz>dwa%@a~S@`I%Xkp7sPMt3UVJq|A`Nv
z#yd;a`IA+q-N0ZV3j>c_@W!wD9=UO;zGA;AQ}^Ef>!zg@1+$D<<Ov3Xf%?h7i?_Y{
z&1VzVDlIRppR9$|!9Xw&3{;<i?wgK$?UZ((B=r@$y?4dt(`Pp-1y+WK>U&(Y5De5l
z1LeMdTIupd8F=%)-QV71=~qd8!J_?)aw|6jM|SOU!p6I`C<RuRhn3sQqW8f-Fc1s`
z12AAczYhI*m$aYIuoOJhZyxaTz+fO42nK?IV4yxRaL#w99npGHvr=F-9;~r#^{cZV
z+~V^^$?`(WuFwYq!9Xw&3<LwgKrj#t1OqD)1Lgib@#S6qiNC&Z_T6V~P^z!kzeB0T
zdnd17s;gMve7)1=UtIa;sf|l3D;6EEtiSK8tNE<oF|MfkqT?Lpf`MQl7|6=N-VGm`
zd41zmOZ62$8q##Fr`o)j)K|Rt)`xrdK6ay0;EK&d*7>-y=HvT6f&cS+NWn+&@!!9X
zwX*v|wd3PG<JRvv<o-sbz+ya9dwin)V4xBVSl^3V(e!XN_Rcl7JZ|t>rMiaL7g)C$
z7<>Qk=fD2>hNZyD@KCMM=KHtx6{og2sMC|XHZCo1)bV+Gk6nj1ynRVhS5WQqWWF02
z*yOr5doM_qHxJ%;Y5G?U=Xbo}>y1h)D=Hr=n*8o+Ll>`BTBB57aroM8hW+}yW~KUy
zwf%Q=SM>H)wD}qR2?ka~2I}iM0lWG88TA$Q?-0~CxxT-DRaf74^X#(z9sz%^fxqiv
zq26;^^%d3Y^ArAVjv{|&Aun3|P>*Fde{Wk+|Lzd7+N-=FlixypkD75}O~Gytt$tbK
zTxmUjN5jHOw^!-<EW6QLe-|XHp1)gXq1PY&UL}a1C|dS(x$eH+JKr@sEsECPjjWX3
z>%V1prP`++%MNSY&|CIp)l*Ma)W27%vR4m&w4$N<KXDY)`C^TCR=vhut)IV31+p$!
zNPSj1dTl4GoV8z8KJ2p)`znQu+w#J!`);eh{Jld9b-%&icU0u>Rw>TD_McmScH1kd
zg|w^ViJtWhM6alJB5S{uucy{Ov+0dnzmQsJ`H|LxZ29?~7j^!5@0}(l+E0D|_lf<-
zbnE!SJt_F$@iXq&z42=a$lnVEIZyI;04s%A{TZH%qPK6%o34Lf4w-rwr>w}|@ym<W
zK0mA8dd|bRFdx9%w&>d6synBppw3TE=I?NW)QhZ0x`hjW_hPg4M@>#aEw6tkob?CH
zo1ed{3-Wg`!HefTxA?3#CZr(kfI4q=o~qnx%!um`S^7Z=YP!xtcmS~jR36C(DjypE
zkUc;B$=$sMtubO;0@8m)s~(M~$U`1_viD(gMkb)nFa3MJw?B6JWdpwZI0dzRmFXXG
zhPbTbX34|>Me4DTxB=2%=BHIIt6q6hd7}+q-Ef_j^Aiibzh8}=An~KwN#Bh6YWO-O
zMfe5P53U(;{VNS7f0SCN{OdUCJkk0!UFB0AS~Ppd?Ixw5?x$7e@84Ru!@Nzl|8@Hr
zDM)#pZ&_uX$J7hzc}DYFcB1F+f`S@H)y~yA-!yjMyeTP2|14xaf?A)Bho;j{P}8-%
z`iK4<boD3VEl6Ad`Mc{NdQj_A{?%T}fts%UVcvk)QG+rbtKYL_*K5bzmw=2D*nHm}
zKWg%Dnt<9q@fOrLtLu>RJ7Dii$6j&&J1MyL#Rv7e>(lWG_|pLojhww?gA(}IphmY|
ze&lxv*zU@|pSb5)YnMR%`xnZucka)dPCSEeFl+oJo-1k`Ryk|^>L2)#h3cQ$E`KK*
zWL;ui)Ao>=@7hmf?w35&^BuBve~g}SvXFQNzWbMFTTW{-I|V(zXuR3ZbN)Pkc52~Q
zolZJ!jiz%_aKc5`eRARUb5gMN+*NvQI(l|$;mLjO*k`Nmvr=&9PnPW5^sZSc*l)+5
zUwJ_H+38=tYu;o;`(<(Moe!V(*_|^}Q2S%Gi=KE2(jKVyYbxuwvYvn%H)sdcxWhPs
z<O6k`!~e3Pwuc{sv;%5>m9<^USxEc^b)Ds00m37QonYrTTeZHe@x&C=bx&o=6-AA^
zDq}aO=^BUB4o%nc%D<*_?g5owZI5#usC=nC^b7RzV+ZGR?_S)B-=3a$tEl`~a@KW+
zbs;ZupHwZ>=aR^}&gpn6Uz(1bx4%|?<(YFhh`)e3Uh2P^uKq~9AbtyKdHM_1>@PRj
zykV=M@1@G8%KAPy_miNtPNVnEoun%dSKt0f@6yi~Cd?D;A|LUS_<>&eLQg*SwdmoI
zxR2gL?D3FuEBsrCAHDn4q+@n{|C<C{xYt4ZT{&=M3Ua>-a=#0zJ=zc5ckBGs^E$kN
z>R;$B)ceHNU)?t0-Zl#oQ0>8Qj_<X2!M6h!nfDpiJ;zlKU$f~3Ppnd6p9@;^7oHX8
zj5zqPt?r+jP~O6<?`3hX;^7OgAHVu#6TeJAwHtp@<Q&F$5-&lGi=M3QW|h_7%9XA8
zQf|G>H$~!&SHH!dWv`VVz0O;e)y}Lk^?S&9*utHD@z28^opDoQq1vmm=Rf$Vg;syF
z&OfVO)iZw-Ilo${^{A}<S3mLkVfmx-Li~h(Eyp?mvaWy{-?hBPb7adeG)}3^e6vt_
zQd!p><wM6y&#%bvrS+@5JpU!XBKK>crfYrUyPvw|#an%lfcpIVppmOjKA`)%iK5nv
zyxCs2-ni=Be@#K1_u9VpSNBiU3v%BJ(hi6}fZP{?Jg)}nH>mcg|6vEnz8BQ~F}|S3
zhqkADsGq98Y5Uq9_i>=Mhkt?0=h{MzQ!48|Nc*R<zCR6LjI-v~`HF1ypM85?)OJ<I
zP8}a)Z=9)5{Ri1ghaV7okjq8dr(VymH6Q%|;l)DQ%Zl2+tmVjuT_EG7>l-rXLQwlb
zJ)owmOueAjAM)vXhOFgP=Kc}HPLO_qYOl)7I}5G(NL&WFC$o@!7W*IkOp$R_w9Zo+
zr{EzgVt-cT-kf_!5Wc}K`z-nPt=1nWAn#Xz+CP=Gzd8;&j+UG?UaFr}N@m<G<h~5F
z?A3jCUYUC)Q2nk_vNaCW&$wEAV4sD@^%-~L@a7Gcx%bd{pfc?-|15-8?DUX%L;cz=
z>57_9?`62>z)qfzVUM0)*?*89d$`{LX@~ue&O7W<Ju>ZsjH@E)Ank$VM+RBHD5q%o
z1^UxkU3blOLpLZ{2oDz0PYdCvC~ALHX8bJF-(#t)ex@?_Z5Hm(Wt+9HK7X~6h4@vq
zvHPSGKO570ZUS;n@Nj(NJ%;Uk^8BPITH}s?DlWQZ;eIbaKP<_MI_~Al#4SbUS+&se
zqr7@u-|LF(@kZRqin>268b`cx*8Eie@MQIOOJ?3!>6YHIGxU$0GHCpq1Rkn2AD&tF
z)}d$L(xk+>4CMVgQ1^AZKjxkbtT_ibzhj(n_7!*cm_MuLy{2b}PrK>Z?PsTFpLkA#
z7yHgmd2fev#}W6ue%`h1XQtSRo_xG#gWf{&S;)L+{bXOlxr%-ISv~uIyV3SDQ`+JE
zERgpmXa_raPY6BvIS11}+F}0CK6Y_#MeoIhN~P=gV=t)wqVji}^`CrhkFS%WNdNO9
z&rL1ld{J%8x?U{reAe^lta@)>Q`8?>Z!FYxjro%oi354jva7cA#H&i9@|soV-c|dr
z^A~%0&T;LcIlpfF@Tm0d^G@4u@YSPJo^SA+B`+@B;k--QPiR=uc9Au%Av6BOXA7x^
z=SJM?a8Jhjj=YaU{q&dolqYVI&hsY9Wkq<`{Hz-w{F5I&<w#fiH16|$4gIDc@JPKN
z<w5wh_C0y$G2`x`?$43oqiDRc%2B_f##h?6kbPoN)Oz8)(((J<2M@jcth47Q7w+3}
z*x`50N%1G%6W@CL?1Sz)dt`dSdBb`RZagX7d(okHH{Nby`r7op8XWWd#I)-(-)*;d
zx5+iXce>)Jd)gkh-Q;waHM)G#vfHF|LB9!&&+aoZ9ePg3%g)$jT*~`%#2?-ZykgTG
zPkHm(5oyb|mmhO|pGoOEkAK*#-QXGN{sV3w_13`m({s;maO!WrJty7ureB`2?Hl82
zekV!%x%`z+dwzY}yrkm;`)~7ln=g~0PapNh?*}eOHoSY}O?%G$BDu8v($BUz`-^1h
zoon5-&*={(*B?3HH|++$ll1!2)V+V&_4f1+cYK@l-|^w(zOf&D`emE<(~U+wJL;RO
z$E5uJnfDCZy}$0DBS%e5FTUvgPRE}4R=VV-rMF*q;dAMcgFo+d^`Y;k=Up)PrnQ#*
zD`otK{{7`+=Cz+w^ZRZ3$$Jk6-TCzOo3<R0KHLACS4J)!livH}CHu|pF+M%?@C`rf
z^4)}#_(H$=T^aN8`LFhGciOy3Df|1mz3x3`NVn<fmOs9@<)sHsOu47rs`SVgeLKCG
zj=p!q`)&G8PuE)bansX!e3-uRLCZD2Z9FCY!B*W*+I7^Fn)Vnc?BC@#Z%iHf)OiW<
z;^=Ni4H`9PRPy83pSpX<R-=>7dp8{M*E1$3+n)I4;t2-~Np`<*r{@m2@RQ`Mspk*e
zw%_bzcDL`=8!+(88hx8HTK;)>v*}5*H8&eR<@gcFlg+z5HKg0yHRJQuH_uGF=hlyt
z7B}yB*n(GPB_I5F^!X25KD`D9+_LxI?^rY=IsAiGciuAU?V9?(Sa<T16E2&W@cS$L
z?v-1wJ%5WHqiX!}{Xg_syw=k9Qjqy&A>&CLEEkEFS`X`$g~a_zqt%|L#~&=z{j!zM
z)5BM#u(rN~&F{{#qIJ(~^_z9PC|dsR>AmsFn$CKo$T_=Qw0P70vCm_F#65r__XQyL
z1jwM4Q<?h-?jtA%65m0M6O1?a7Vt!V?m@I&;=1aUk4ni!_di9;@!YMhBI}4X&y{y;
zpK9qfUh6uc^HS51@h=d&LA8@|dC?kY{CRnycP`QAe4cELcct`JoU4?c`!~klL-h}p
zE&p9!di=b$Q2DTAkLTsZm)EYo58<6#i~1?N={_^7tag?wXWi$KuiSOYil3IA=Svpq
z`cZAU=(v03*as=*U7x)Co*jA9dCw>-a$i>zz4)T(%(qIT?n^A0{jP^<H!|Z%eBiwZ
z?qfjmE9&=3ROVjGL*^Ct_11h<J@#9u_G-SbuDo=IZ|AO8x^|Zn+ubs1db;G)+gm=p
z-L$mt5&e39|NOL???rH~;rZ@HO}8C%+2C0<-;db$m5~b{96USa`wlm>+PA}=+s~@`
ze#F@89~(Pn)XbFU$Bo+k_LGC#Pfw|j^9Sc3+Q%-QL!a@-XBPf$)U=v*kiR|b53RcP
zosoXO-_fnV9yKfF_xXG;W6q_o9r?p1Q`6>~AN=5?ZZm4`6S?Q)UJT^BbhNw2<YseD
zYVsd@u?ys0jQ+eb^7~GM?wp?X*|_<@^RE6NWqvTeJQoJN`KtR6&!6(9W1ph8uIEix
z|5aJr_hc(Bs2)3+XNunb-b>f}imY<p{e||cT5{HPO8pE!<$eOWQt0g)tCi0C2$kvw
z^;q_LeqZf$-6v(0d4IXmNZhO~%<BJEK7Lo17j-`6mGP^hsLwg`%9cOp)$2YbuS~qJ
zG?qK>+AbWwTE}ggm$IVn-#pnmCsb>lSl;z|d}z9kpLgEY{fWwX{kq!ui~76rKt0-D
zWbFrfMK4}!{M7NvDyuyz(;r3cw<l+fhuV(H+HaNBewEdai^^)3%0=t-%JJOELdsk5
z%+q`K?^e3^9N7B1q3ZXR?D?0b4?Xje;eY!_gHpL@-S6eq=e6UP8~^a-*4M3GDvH*4
zcyXmt>AjvfX-4lxYn43Yc?gLAW<{Ofx_{IA0ZrHQtnQbx#?f-~@x3f?YTt(kymRM_
zl=~p=f4C3gzUAB-zu0$ryTLWzJALeiV;Wv^&4l#Q(rJIZugT2x+DjX~_RS%$r`)%k
zaKlS&FKIF(rGAip^FG>)-CMjhedw5)_n%(sxyxG{{q>`o_n>%BX}y2GciQLwXjtO+
zYrOZhb&C^PZGGqDn(v2mzs37eycah5<*yDsb?7UJhu*nJ{Qz0V8`&E#Z+~XpKm6ph
z??+v9->f97|LVF?Et&b{)vs|BS^MMN>r^Ve-;}9)Z~t}EQc=YI<%MJK|NZ>eKi{zA
zVYTej@BSB+b^MFU?AN^crRhc6^X7TpbhQiFtA}{xp*KIRbk=Rpj=bsWm#mkT9Yyu{
zMWxZ&Pg;6-%!|~AKjcNvzw_Ederuh|t7rYiPS#P-^DFX!R({SW>~D%9erh3pX89}6
zm$G8f^^Wi6*A~)m4?W(!@_EyX@>f^O<?TQD`*$d{c<<!(OXZ>$f0)Nbk?$jB#jNdc
z4ykR_bxvg~Uae?);&W|dR{tO$&lMHD@nD|iMc#9veMM^>sg$1cTeVT|jo3$kUVqr%
zl3$T@u=}PXUpuAUCrMVc;=Q(mT<LX{eO_Mlcq3i?%9Hc@vF2wSilP@Mh~wV<zqL<8
zukje!oA0E1cG7>3@1p6Inzvb>e`-D*_q;MZmWw@~AMxz8eK#!SMdn9QBz~8RFST2G
z+rp_WN_o-q|E%d*-zTy174_e&<%*6^T`6bH3(r1lUU+e*w$eTOz4W%P{o>*A0~aRz
z&YJHa@V?sJt$V#RuJPoW=M4P*kH5Rc^Ge=V<vS7lE{XR>$q&MhKF9H7)*aCEU)C8#
z)&&dooJD(C(b`8?{)E0#>q*{vyzd$@ULMkq+QuD!IOvou*KAU<ko%{snDuvO>_-(X
zzf!+w^sj9%Y;;YFlA^_*H=i}V+;y>NdG>L&jb1#+o1VA-8fR3_+rF1y{mGJb-<Vg<
zI&bpk<GZ$er-b!{^+VScWc_^$dVU|y?=$%Q1nJ0%9?y{;3<LwgKrj#t1Ovf9Fi;)?
zS<hE_^JP7sTlYbh-aAk0xhtz|ofEU_IiGkuWxa=_J`25mS^Qag%Wge4A$#p>x;{tr
zWb65S*730N>HR=f*&8>#r$V;u@buQ-(a5VWFFGv$()v3?@Au2zy_lvmPFc~q4?@5E
z84C-7ft8N|UDt>k++UHd^S`L9&;PU@{@yF7>3qKh)cSb-1M)lv)cnZwn|PEJ^?8q$
zQ~NdDi=X;Ftlp2ZuK=xdtp`uX9_|yle?$iL{*mWvd69A8d&ZP!9Qgh*`AFycFUZ`p
zQjWh*iA;X{m~!aJ2Qp6NQzRcqenl;Z4A0H??eU`~52uOR!FQlR>{ir!W5%6x1MR7u
zq=Sq<<<*a?E$e%{@B<QeK+RvStnX{F@1{MF`NTax^A+9{c^{p87LreqaS03tf`MSb
zVu0_Qc}N@sHD0Mq90YZJD_16-7Df7Dq4vk(FZAUxF!0VEo9{brV#0c@sN;al{)6wa
zm+MdLb3m<6?aC`-Z&6fxR4#86qBIPg|BGM0_GtTADevpC59WP6^z5I}gXHJD!#Rlj
zAo<ysa}FULWPeY+_$TKm^qgDJgV;g+An71_>_x^-51EIG`rT8NbsiNxFJ%24Bl~^Y
zCH`~I&w52Z_TA)1k8GjN&!RGZ85j%%1HnKr5De5u23Ti{qOR|(!xk1j7gCOUJ<c2K
z3lzPv(C6FiquBR>?BhUfN9C+MTE7ovUt^)Rr?OQJJ-mU|KB-cAeV>8-9BA2veWY8c
zcEZ2D=Z#(wzXFv{m9@RRvbIloea}SwQ_CUa=e++zeIVyr_V4U>@oz=u0p}0SJ1V0G
zN!R*RCLez%6g%k;h#q@E`a?Y+=Pyv(rCe6j^2n5@9Sik&Tvi!Bq&=-49&|j>D{6V_
z^^kE;r2kn_=Pm6rZuk#~zETLE;Oh5m+4b5n_a#}8`K4&x%Q7yy{xL49$ImTP{xu$H
zy7FzwS`Xs_s+}tHzLDy!erP=NWX4NT{SLk@RDO`P9%RxjRNmA+l{G*7C~wGCKhTrU
zvQPE&D=TXGO3BuIsV{nK9w|S%@6i1*?+38IQN%w$Ew8d=mo?t_r#_dqWd1Ha$iA&8
zT6Uq=drH3V3StMS{8_T@ud>QkeMR+*tHy<_GI4-)o%d{5UqE=|+`{`d<R{L<JIMRI
zMf>N)b@dzk5!7`y<nkD>;zQ_<y}jxDlat_K<>sNh+4Y~&Pw%no@P@Z9Nh*c^sri5Z
z-{iCZ@m0(F8<eVryIio>RolNZC8-wvvgi9pyf%K7Qnm2dRfZk<<5ioMs)hQ!x=PEB
z?L7ARZRV_1s;`*$`we;XpVhj}<4cD$E>#N`+_YQQFHT*xR4sh-@eR-X^jE8ts)fCV
zw0)|_E}tjW!oRHk_dj$QI6tWtp1O34?R(FipOlM4PyKzE?r(Mf&HV=ZWZw5Us`JhR
z2lg17vd>mzzpco=8)Uz&SgHL8`)vBpzCQ54vj(rTB&lyaFpgPK{kdFO?>VY1FMNCY
zktYuLvgV$a_meo+a-L;=^4tPh?~_#~KS;eimt+3WPv$A-cji4Z_Hj>vo^$O%Cv3jQ
z?rVLN&<^i~@p};T*bmZA>g{mu(UaCYa!gJ9A)g}cGVa(zeb`SqeXoo-#d~Vh!*`HK
z*L!;9nfl2`T!eQJ`#}9WoYVu-56XkIqewbPf2mLP+7HsH*Xjr9)Q=46dwk>r$%hPL
z5AQi*H|f|1Vjt=F7s&XNA4ISGQV&S`AnhT8^oRP9LG%_{^-{kg_It=Y0`WW0Dv#fT
z<cC-6fj^M>j!e6X+Aj58e9`-zjy?0O6eJ$dKJkbA)O%X1>#n(O=mw=~;gpl7cInV|
zaZ)XO<>M~5{-kfCQnipcUoFgguF0F9@h^%zPpWIkJn^v0b@%n&`L5Y1NPM-Be2N->
zn8(B)klztAPtlW~xCT-$`N;=jC;8DMQy!W8tOuG-K1J#Uv0IUJo-dG2+^1gRnf8nP
z*hRV`@d+e;DYE}!pF}<QJNvQ&TMn8ye*Y&@_6?LH9&+BJ9_+`z$PdyEGVM_x>B!WB
z4AMV%tMAD9w_12=y5(y->@qB^7S8(FUc;O3KPssf4!C9Szu&QFMp9|~`m@8I>Di-s
zsoKc<mDNJdb=5+C$6YPF><6b#z46S|OZ62m8gupLXTLHnsTR&!YuMten@vrsg?nDq
z`RBcNnwV4z+aJHf+RYA`l$49y(=d;+#(maJ^0S^IgS`I$@;zZi)@?=h4<P#?+C#7Q
zus=YL>>>HNN2Px57u0Uj6<N2j7o;Be0cjUGFTw}qkh3D^w4%7vFaCMhqcd(wilV-E
zioaw<_Dxw4zX!E_PiCJ5W^I>z`aO5<trV%>L-zaskG;1E)ppsogZ8#zG++++L$ISA
zIV1>jO-$T3fA{9*287Ss5wvL_T+IesqP9$IB{6oib57$l(iKf&!&b24X~ecD1`$-!
zup!qg*b<Z=ShK2T)%+XvRK4%>ynn0K&U$OkF~=Bl)_mTzzO}!7a@V_^y#IZE%?<Fs
z7on5?{blw<KC}32AH?H7Uu-}7^=s=#|2>WW{SKWx55J3e4nA!^=JRR$+wUW8{y+2I
z{Eh$qzy3SF<o}%~|KG#ppJ)HmfA1grcmB{n{yV?M|DMcq^WO_{&j0&%{`XJYm+y1y
zY41Io+c|E{XZu+m@8{Oj{yB1M&i{RR^|bG&Tl3Yv=zU^+>|XhO%%^=G_WASwzLihQ
z=RfE9wDnta`*XulZvXegJkRQB|M%iZZ~y<%t>$O_{Wf~P+y8xq&F%L&&iQ{Y%>UgL
zo&0^!o@gF~qxbPFZr`uZ&UufsCx2b7p7!hdtQ>uvq37`%f9=2c*MIfj`n|u7&VT#A
z{B!^NU;I7)@UPFF|J48UC%^lxf9WT`j!ye^&cBBao&4uM|2;FG_W9WF+xGjY&H4Y|
z2s-)a13vlBeLngBo+_XJ^Kbw8f9OB>Xa4T5p*<hrcK%1@w|;+rRo&5^QT=cIum9)&
z@NfS1@BZ@M|Nm3$&+kwExt9Mvh5vgK=X~<tf1tl#v43yFIsZKm`|#i2@cF;|rT_Tv
z{rmsy@BR{<Jny4U{_kM+<UR4<3-KNp=bTUezQEsid4H^D|EB-j|JlF#zx|W{<1hJr
z%X-e4&nN5nWSmdN?TNmRbH95}+wWO9zrU|KAN79ZKIpk2oZkohep@~HeaG)xd-B&a
zpTF|={@-8u$zT7SU-HRc-~ab-{kebRAN<Mh{*u2B{QjT(!N2*J|Eu5m<zM~he)F&W
z+28&@8qRh8{=@ly{(t==|Lx!TeZTWdKDo{(-^=EV^T~BSEzb9tJz0-Ve&6$Ru)h3#
z0-g5jkp1}N?<;&-{=GT(VIQ88alRM-Jk0++ocWyde-~%I{lDXGf9_%ZKm9-bcmK)1
z{Qv#k_Wh85&o9^Qz1g4tcUoNM`S|4bHUIkp=X~0qYnW$$ZsmWE;gfl2KAwxeAMxLd
zb6-AL&nMptoqR7o*&m(u-=}PjbUYtF|65P=@4JliKGA9CMLGZPVfnN_*Yo=loy_O&
zzv$%ezr6q1)9%my_d3q)_ZR*?VsVr+pTF<%*RkdEbLZ#GUzhy#VC%d$`#$9yo#=i1
z^S}PX_j$)Z`*(l#=YRa^w_o}8&wu;=Km6Ih^0ROM0KNYBlOO-lAODGO-~E-J{rLMS
z{{!E?|MvZF{r$i2SAX{7zxE&f#^3)7-+lM}%*TK=D1HC^usGBBVvd`y0uHZpAberW
z2@6kghyxZT*wlj!z~!O>!r}WL3JxZ_yKEcVe7?aA${Qr}^K2&mcDomU;=AM1m&y0z
z^Q6_CY<JRBz~+g=KCyMYj9UjHo&2mRf0xR()w>?OzXMtcR~ls)NWF}QqK6gTL=up3
zp2z{$&iUYW@G@!+l|YFoEDMBoDm<;hz-UmKn(-XcXo^woj_y;oQ=~hFxKkRB7o*e)
zRU|HBH6AQX*($5E6Fi#kG|PiG!Uf%Ekii*XVPN)grO72s(q3Xc*Uh@>e8i9H9qW-_
zYp&5tBekGHJr^WzBoL_(@N&5B0~7#W9ta7p_UDHq`^Xs0bukV{Km~-mIU!LL;!s-&
zjxHQzNM~@9X442e=r&Wn$abdn>u?U25EEi_^;(B6n1CxQLoB*1()}dHz@(o4cF4hv
z4$9$mZ_$KC2!stl0l?lK5CP18A~_$#Z$v{dDWWyJ1zuPXgg0ReG*f0WZghC%&fqB4
zb5)!}qAz+xcwG`scVr@;eHxF{nlE^y+vsSXsdaPNv3tyj(b*rH<!q5tge47tA>P{{
zWoNu?w0bqbkKhP<)SkFEpSo|1CJrb7IGc`Yy6fH9X|0OU*e<Mw;NAB+$>0IQ-PPhQ
z@zRbS*J)|2w;;dl?p}5<daQ0Z2fW3@uR}h0_$Rl2@RLHfox|r?$5%o=$4bROHV9YZ
zM<y>f(B$Rcb#DJ1a`M7w((v+N1s33ipaK@|x#Fa0KvR%P?wQAEPxf86TUVGHtfwn2
zHCJccIYxt@Ian|}ksFCdHEBA_lbT1wC(m`BUFZ~o_$g?*&K|McyLwzY?u5CcIKW-+
znBgQh0@rs!9T1)_f=UR7f*Y4|ri^A@o*M`p>;Rq|7MSpelZW8*5j}6}fd|Bun;IR7
zM(toNEQDatewG(VKMndat2pDLXLvmM4fFw5c<Kg^_IFgz!%vZpuQ{+8l$tN%EZ`wE
z1XI_Bx1f#k`~c4+D?RQJ&t4FLi;y{D2@ncpoh>546k#Ih%%;Mjg+|eEA0Q^ey+cM8
z>A+<MjyuJu@N0w?h$4juH?)?X%tv+r58)7O1Lt6Zq6ILEojmFhu<C$_MqZ#Iiy>Yw
zz0hWXTFuiDi5%vDu0J7)-DA{EjA#fi^P(!i0UR!`E(A$V*w>8~!9)QR?WfUNU`DtS
zN8V<Hny#^(M=X!gBDOV~yB6mHpj`k5P#$F!=lO5d#ouz+^^{YMrjl+C;Hl}X*5=&h
zYMlG@({*$}=k_R*FcVk0#mhu*HSB}7W_Mz9HAY;FQ4*rcDz1iDNl>HnCIcM+yfVAs
zdN+><dJE@RjV^;bq|ye&$Q|MEc~yU;12}+T7sLm*lki9y)*4uXmP%8KUBLlXac@3s
z5EKI_kAQG0Mzsh#G|Ghiva|uX=p~o<AxaC|$L5(Eg2H7`M<s!aa4@f8iQXt&<s%=$
z&-5Dgc~A{q2jPA!!4#q5t>Q|&n4n(SCKZQu-3w9YaKzF*WVWu5>d9C&QOTsirl}1)
ze1)Q+2-Qp{CDLBTci5j~BPY1PQrfS@W*C!pFIec5tD-AfU*S3URz2TthN(F^YkyoY
z%x8N#4k#4>H@~Jw0W!c2P@!8)(zNIkvZaeJk&m}?vfWX`$##>pqi(0v=VqB>oiY&q
zLc7~O-{O&1xqISM4ZiNpPg)~b;2NC&d<7CY=LAL4pBS%By2^)FyUqwXBixU8!~}H!
zdw)h^L5wC1EU@1PoTuIKG^1##mzqZnACKuP=B&Tex1#2RGfq%QH%SXyyefF#otoMQ
zuX^Ff%+>aR+tptg^oPFv;r%Ov@Rm%)?X?09uR4Sq+fh1U;VBMrz^LV}t!)4<7Znf=
zzr{nr+Mw%f+t}vwjyEW8kj&4snfND)z4#N~9iP5Tz8{|_t?p#IlcoYTPaO7%t>a~!
zF%a@t#qqbecy^D<jxzR4h}<ZHext&$*RVKpy{-1(c5p~Fhf1JC6qW@-I~AVRU|=*T
zP0e@?X*9*Cc1QOq+bPl=L)<Bi$BR*Fg(?!4u^JB+rfikf*$Ey^cbetF8{vX(G|1o#
zurM(DxYFbjCTTCRp6g~^bw1)p^^W!E-nGs-&9RPWS@!`703S~<$++5|zj?_%GM*gs
zY#JN^7G@`a2BHv$x`W{8!a;^~1~+Lojj)4mGv$kHXIj4w=Wq!zAx2lPb?AZ#xUw?D
zqRS%PPht#A>iKVn9Ng%j9A5VpO=yHb*Z>p&?Ck*&!2BnY^FjPZGz60(TEko5g#|%)
z6ShDzWhUcBhga^5<tT1cZG6!q%N-X5(;b<}XYb>Y6Yd91d*3^vWA~U5qxj$6@raRA
zge3<6L%cUd%FcM(X!UA<58$V*h3DZ@4-%t^0}24nrlXqfdUtkOt70^^3#%b`_q|Rs
zc))OXwYW>Xw4=v$S{myu$S=FQmmQ2As~gS%Z}IT!kWU`|$?YHfq|j~W@Hy7;m5|S|
z@;w<}?Y>WcO0TeRJOix2g8!CZR}2-faL==uG!1A9Qpr8@812cv>vro3bA$DCrKRTT
zj626@@G}Psh9`0((WoX(XL(Zdi1_5W&a(@hLJ&U%P1o5YmU~x^OUIoscN7QQb-nw<
z*Uy4FAUxd%l@JaE_wD8sHO;&{HxM}30X#A+FyRp=55eandfwCn4~Q!_H98WF+QC{_
z2*IHJEH99L8uVpWamGc@@ObhY=mW0s)D0f(@2H-KpCTP!b6_(l9sABRmfAFO3+5ZW
z0n^>e0pJpbXHp;{STIGX3x-E&^r7_;gE7QJIP0Ev#)HM^%8;uLC-adVz(Y6$^$sQ|
zS^%@y$)g^TZ**ZF5OX!a0|-%X!s}EEZ5F82eD*-jv-&YFc8}5IqzgLhjSyZKs0wfZ
zn^AgmH>D~)DG4qSKxH^xXju!)6|TgQw;7?PYi#Ec%VV^NZO!Jc#ks(IJH%YYdH!2<
z_6wN(PCV6UD(OuCJT;xw+Prs(Bg98|>@tN;S!NPu;>y$rk0rg;un$^|-HFZB7;!O1
zNr)<|xEf+5L5<Fv40HhS%It#c-8>@bJN$$Vlt&C7v@1B^>?I5v1jPW#BO#oMQ7ys_
zjWS`sENuWTddVezh|<FLv3cf(pl}(~QE3$q-;H7f7orA^I02mLHPP;B3c3!${XBvx
zLd9Ffm3T2hy|PUz4(qxXqR!!nrF+P1T_M$zv1+1{NrO#O8+iB%MMDv)nNCWiy^Qa$
zKgmW;aDk<?UyIE!ChcCZ&?#3%SG2yubMUQtzTFH{b9C1JxL}yi_H-OjDgbVNO^*U(
zfE}Phx0s}9(I;fjy!eBAPqv$+J>=of%`(S2XW;94W#{Yc&$<3}E?@WNC+!q0a1G9X
zz5<DybAlr2Zxo+)v+InIGs6A$kC>niVDHaJEQryhfd%&afb+CFo@Nv+^-}Yw;o~uV
z#hmq*`c~AOaK;G==_YAmi&q8jyHiv9;8idDn7P_MaJ%{|gZ}Whe_{X1AiO10aeJ+R
z!>bPA#&(oWSa^y<958CRYik>T%S8o*!*B6Wur}yA+cviOyyFeZ8zl4dY$pDRVlV#0
zcgLqMlkdmpNvk{A?xd-J%@c=xV(WMrXAFcqR&o3-E}q?^vZIVW6CyXtpx>x4>@_To
zTyLvAxE&l)&7l$~5rt)e&`yP?H5eETN>ekQLmEvns@>6j%65u$#}Ic)<MCpYTA_->
zWvs@7g(+KQb#{VB)178{@J6_x8x1lz11t>8KCU#mgh|><tmnE}SDlaeQN3e5x_7N}
zPIIi|S=N1k0>H-;Ofs(a=WkxJkBleBJevkbfQ8u!pn)jFq3$3!x^R#ooxx3-O(X1}
z+f4Z)+nLs{!#P|+Oo-9dYaO~^0<NqKvFNf$_mda{lY0K!AqO`)D2La*MH3n!5H<h>
z0DF5t1Tg=J<a`jn5e>nlh}Q5Hcws>#?vW->fn@4TS6CCHT)pWWTKb|#ZtA)qn(oL%
zK6?+3)S54Nq}%9do~d<n*|B@fh*ACCx_tuR4q)^4j@lw+#C<bt^=g0*;HNEv=iyTi
z5~GO&3INWgqnhq|cXnE<Vl=i3t08#zy-qTCz;Ji9xJ$gWqsMhx8tW~{FT1;!9gH5U
z8_of5@$l=APagis?H~N4&~4}NIo9!&kk7I5JsDr^zE6Kjudr}D1FXP;|CV1@3>C0&
z&$F2{4QL8d$vyKJ?a99DcIygrgY|T!rRM64JI847GY1QXCvqdvs3uKkc~bL;_~g0H
zvkRR<5I+S?*V!YMdsmN3$DJ^D6bIaOz5B%1&w@H2JlzMC5Do?R?dB9U&AdD}5IEQY
zJTfdW;Snbf!RI4--qZsRh$}ZWIuecA!CF`d!Jz#tFOYs3^kr6Y#zoKYc=8+Q1FrDY
z4Ib_9sGf(PA{}3IU^6Hk`_40#+B9+t<{P~M)7{De;1Y&sQXnE&Fh!^fhDT}iq4g1i
zF~meT>z;PTgT?5|kgE<S^N}6ELpTKW4kjpC0JGT1qaKlObYUM5b2Y#N2vKjs>r@ME
z7O2&H_CU_F`Y|tdkJ03$3p(qK5MCIl3UC0MQF?PXr7Ars2`&*pWjI}ESqsb+uEdeI
z8KI_YZ08ZnW3-5E&E~Gfxxjoo#9YOB{#$kS3z+>*Jk@9_=}iDUHJ#PkymyHs#7B7S
z5~0q)0Z@gRxDuH?8GEZ?AG9mG6Pv3s;$n=F5LH%jHN;AS8l5*8=m6lA*#+0Tc|_1#
zILB&q8QdY2HXugs2#3$B`Xe2{0Svn!KDeEPN7As?z!J1nno{fv4zP-Q^I?Oa7(jUh
zgi|r9McAQHChV7`4ZuY&xx^1qTG&1|&)g6cE=P4i<#1WDDp=+7V9|M|SLFX-3|$A|
zemubxq2jILO1zk$UfCuUhjrZxQRi^P(miChu8``<ST#|}q`{`C4Lp2>qM-=YOeZDM
zUdDIWpJXE^xWH1{uf=8<lXfpy=#;CXD_URSIrvsR-)@GfIXY{9TrkXMdpZs%6#zHC
zrbhuXzz$HMTTIfl=o7MMUi_K8@w0oSspe*xW8oXGRIcllov*WBx#TmrU-#xGZ5u3b
z4bFeQ0*Rb+f+Fc}6rXmp>x_^y!u|G-n4k_|@6Sjqh|#2h1@`-Z^RzpjW)v;;QuC<c
z<1u~3ob{LbR@9tu#t91PCTU@dR|W68Q&aokRWJOQx!OK(JNlJDKmWxK-+xCRKmIiR
z=Ab{ae|HdGm#Me~SHR&_hj3#%N+&El#UTzDwcNF}4Z!810>a_<dMH>Mbe(M*+kD>f
z2IUQs`FS=I|7funf8x92)0fHj<MX7|oosi~RKVtm!#=Tfyo@sjLLRF){uUR{?oruM
z#-0h08)eXMR2cRe7DukP)gIgq4yoo)36zM!vOs93!qXZIj0UBt8P6e&rWn=k=ssmT
zMY>~%JEie>F-omaMdC76<H5p|t+F~h!K3L;vpjesT+odM8JqzY24){unq0yp?IqT8
z-K?w5NBpSXu^!#K);Xs+*6}RsK0pEB;|V4iSNrofFWE=NlVhGugCoGg>;%w26yi{K
z5FA}N$dJz9Ce5Z1cF=96e39);>(}8NE+Hnw=<2l&T`&PxR)$z~S)}_(jDbl#|Lu^2
z8y%Fx>)xUXjSvVMfC7NMJs<*@|3q>=h~J2YU{XYDcniF+AQJaTlczv3b*3w<iBYcJ
zbPg?j(IYo?T@X!oWFnuvhevA77d+B!bTrS@y1DGwJ!ZtHesA4A0dNPfd3#4~kuu`G
z83wOb3;7Tv5z3naSW+Gu^E`a&<}sQ$pa9@(I;!cecW0-yDn?_wuo{AQ-|HlU2Ml*t
zi@U^2J9=EFrLo?E{Ia`y*}>?sy5St~77xD;`Q+iB-2TB&3f*=NpJN?g3Hcl=-;?px
z?)&to^a=~dGr$Tg_;2}j#ZUnY_dJ_P(}1QRmE1Fr(VpzPZnv&5H&{<sT57J&xO0pK
zKXb5Pcp^6vjcU?#mM1ljh)<sDJiE{-1o2bQbe%n7xp(!rbleGZM{&Si*Sk-A{Vb>h
z!qa_F3E@z1-)>G()6C0r1A&7bz$3!~6CQE$5PUwO=S@BEfVgr~qa)F%9jt|g5DePS
z@&f6nL0@JSXI%6Qk0-x@KHv&Z-Qdywj_P^%Dbn#Z2R4J!vF|)%sZArdV7}2CFx{;j
z04`y8CIupb1yh8&V0e^9A6g$V7(+~iv+ikUJXnmb47uuXG9TFiJcL6~?_h$W1u%=9
zJn9koMi=%0F;@dTfDrX2yiT>yW`SDGXAk5&s~_`X_ZUr1x}dY(2;qf+ssIPD8KpOO
zQ>xOFlHd{nREE=qmbJiK;Yu8Nn-OZd#&#aDJVuMy)@<%toD0miL(Em2=f72Fzku2A
z#8Zu?lHLTsQ`1?k&3l(PLVSeBE)nV+8~|0Ai7S!Wld-oN_CdR{JF&SMBQC}$2~lMg
zS3|5MsL^?oferv(nO$(bn@0q_g>$S%m%$xUX#-;9j&S(Asz1^J9Kf&(;)B~scq9#L
z4J<)Rr76X(-~g+*Hy<_#iUE{IKsXhnT7(@MWx{@0+5lYil1uy$rG@Qd^UMuF;WDVB
z(kdRl8^s7NL=7Bq0yxuaqTSULbRC5Ic?460inod@@nV8{Wt&tS)^#sLox>4J_mJ7T
zLaHZY)kGze2Aiff@bDFih9Xomos>v>8Q)=ll8v0;0!wMX7Mo#A+Pz?*Q?81xXnlp}
z;9K>4yBVhD=&b#5!7!ig={TTN0NnhV9tFq%J3xhQF-g;+PspBm@dx*wY&S`J$ittT
zWsY^uz}NN4&ez$WbN%aFzV6LW+9_Dz8l3-p1rj;u1Vz%{C_e3G*BK#ag!}CuF+m-`
z-k*_J5Ti*03+(p+=V^C5%_v&xrRGt?$7A}6IqNU=t*ANSj1v^nP13>^uL|CGr>6G7
zt6umqbG3cocJ#Z0(yt8qBj5h$CBHK0yY_aSm<3O{D|-Wb123|H{gr&qgE-hDl>r67
z=Zy7Gm(RJ$L)Ug^v%U2@7kGob?mQhiLXQC9+_BU>OCmk)5$hjy<{$h@`#MB4ON2dY
z!l=*MJeF<q+<`!`4D>BMnq~?Mn{rQb_^vT((2iEfjdDgW8uoc+(~BaeHYrap^r|Y2
z|4u=^OTAIOE4UPZ^>)P70Uy9-%f<(a1n=vGmu%${(<eQf<GUO)o-}&Uw$0jtfOG;h
zZEuJsJz<i03)9=A<ddKPo$v64na|qWv-;ljsF!{gC^wU@+!Njz7FhYUPq-df_~3n=
zbmx<7@S<%4n`7m>RVHpWLFu=)WSeJZ0mD*vs?i+F$wVI;YHt3wCJk}cN@4~M?RQni
zS=&xf0C1DtAEXWkO%BRwgk#42b83gSa}B>m_V{fL*YaZHlv!VLD74168Wvq*8ULEM
zUS5+D<ei0!3qhq_ORC=)wr6;(4S&qOW~SnLjd#CO<S#H0c1@0O#SovBUN>O-IqJF=
z=PZrpJ9S3-N$0zK{ZnnHZtFXGq!|rLjq|MFakHf-^|)rwTIVbFy>IKR<}=?=C9inr
zK2tbq1V-54vP4Ic{Z8-uKaS8AkCku4n7QN(C%Q7w0XTBl*!yf?Ki1Qq)gxEhd-X_N
z2*86sY?Qnut3FLw0wg!IiS#_DiDt``_U})#N_U+MICR!-^Y#Wlbp!lG!6)Cd<t-=`
zlcpcA9Rj}%hChv0d2?w&>t~RJ^AGR7!sFk5p5V#j#Y2b&B8<5qx}b7Y2KWpcd<AfF
z0fuJMv?!0f`1XzmK4$1EB0j>kx@7H3Mu}8%(XMq5okf#wY`lR7ZLqW%Hw&r(q^@&R
zS7e-=WP=-k$`krNp(VKdhTmxeU-y0Ub#MMo@6X4~8vNVSV;b)+djop|@4bP?{6X@V
zF7N#$UTpfedt~ij|Jx$P6+~G64nP4-F&8MUO6K&R$`VgtZ)NQvf2IG`)YN-Jkn+#B
zBDb3bB>-FXtLGnbjlAadyR0i-{dw;B_69zD1M-Vj{7lgPJD03qvWm+a+%)xFQ{5%T
z<0{+yLx@7Oa8T03hi0!i)k9Yxo9%VTZ9Fhmi&x15huE!NX#?q9F#5~+_kWYn1=GU&
zA-B)hvuERQp#dcTaQpGW&==C?+=42$`GV)Be$5R8Z<z!IC;&S()V$?V-{&SgDW7iL
zsDSWd;Hw_VOaQwv-VONFd(AutaCZO;$A}SOLULOenr4t*qN>DD45VsAsZ|A#Wd5oM
zczr%x76sP@3opBw(NE27Xh9NTE9YSH>W$U{ZDTTK+0)S8KjjPZO-fF$Ju0zfvSRU+
z4^mk6j*EWpLl?IVhi>({qLzp30holO?Rt*tU)6GIEE?PbCOCi#&pzv~4Em$r{>8Ol
z8Kf@MlOEB<bf&I4C=GSMb7&_T&+>#M2%Sp>9A32u*Vc)JiMY~CJjEdl@Y%E!grH)?
z$yhh8HpfXC@i7%u5MH{6RSC+t4eaLtv`P%B0W_FsbaUt=#=R%B{x*-iV@`|?o~hq6
zH3K%$92y%4ZOP_rgz!?gRhw5A08_P=LPiL7&CpzGWy}q$xF3qt;WnX!ats6j#y()8
zbssn35CHxFUL3!>^hi$=;vracgHk3O)hD_q3Uo0APWjaC10y6TxcqP9coy3Xd0L*#
zhniurxp<aBL?Xhb&H*f>YOYm+KKf%KJbl`7Mk`2tL`W%r063Lh#R2S5xl|V%5q6>^
zFv>bUn@w@ZuO;;Aif5}9O9eOKvym7;rzEzx1xY32z$BFs1L1*7g`9!_1?Hq8sz447
zD7GE$QLdP@u$7Nus6^44^3uTe=Vl5Q356nMx;YW!EF6xo+5j5)Ya(lN^}}@N<BJ~Y
z9Y77hq#KkDZ9C!EDee=agR!9C@FKA?Mn~6?3K^?qzycf-DG0ZptyK}l(ioKwJTZvW
zh%LzeIx{U>D)kbi*f`YGpjM=E$PEesq`YoTvM;@)PZ?*w3>3VE_129Uv}LAUP^UIV
zhh)2oBGyopilj?F7Yexul^EfQRM%oKbU9^NlaBR>?0QB~&gzRia^no-C`wsf2#I@0
zw~TIGBjb?Kj;Sao!Am2xpmy!Ko7Bxw0IBd4RQ0$=$jzgK6x7#pe*otiehE|oMEU~D
z_b|HxL2wCHglDypZZZxus0?rbdt*tIv0|cv#>>74<IiTIi4EpWXF!MwU5Nrn*fay*
zx|b|J<*J2Dd!a|i`hbC;p)C^4;A3N0g@u`(h6B6=Ttk5vi<-uS7>ymeQBkGW9H_s#
zAd)xr2)95LgtJbchz)QJ>x_&&qT5-*fFGAE$~m91U4)H?1PTeuIJ&T@%X!v%ccUH|
zH&i$aw@-`PTYsA~Z&uHI<b!b5**=JmaO9KlBJo0x6cdfhOvbIDotz4eSO5n{)KE_#
ze|Q&<FxQ=yN_oaw;3N^CY?M@0`X&s_T@Sr8!HXjuVe?EwRzj7CRtWEum=U9RTEu5X
zt<#Y{0UN-8F4;0(vnOTw6R|{jG1X4Hwv_=G^@a3iI1E_MXFygZm2`3h3rvJvb3)X8
zdz5WC<x}@!w49tJjtHd!z^W4^fX6VOjo1O%4cS<{sas$M(zqC~6nt4u5mrMmsjkAS
zFR(XoNew8k)Z)fzK`Ye&6ab7D(0r3ph($<@mbRejY9I?7)5)>sM`*)Zpfx2!MTEl<
zz+{+@?!r@w2}C$nDIyykoRKENiVd#J)rmzOz(>L|F6J!mnTTRKY6S6KD2C0p1-Yll
z3=iH77lfI(&@vy{kY}&aU#{e=LpkTPoZ__>1fuo1P)Tz90Me5*GGMB)8#xD15f@@q
zZbjRRb;;0>YTC^?swbyL24L7o^L5s(5#AX(iuneusX?U?Dj~dUhYVCKxPL8=(&OES
zBL3_{5TrTGs%Sn+%sf7YOAZk(8fXa=NSYvtkQgS-q6iC#$lw+{cXnOOC?-uBj}#+Y
zfjn5?XE@!08!t@NC2$O1@W_Nsp|F(CDl}<A3*vFTjG*2};i!`+;!kejS(U_ve6j~R
z^;k+8*);<DHJ}Dp<7(d0Y^9~<>Wr@vqjv4g4sl^2!Y)C}0^|bMBH#pz5&e)zsTv=$
z4IT3g@z|NHvYqy<Svu6uO_Z~sd5fUR1TtWUWcyex$>sz9aY@g3R%XWQsrk<(y<j`z
z3VOEko=)um+i&3NsIn6mj~g&E=oE+pC>tQWxl@2h2){)#@RI!{O~y0xB0j>oFV&AY
zsy`(_;9v)^Q&cb*rRlRCEF7K2pi@DC;d*k~bqCS1;J^=-aig%JL6M12Jwl1YA*djU
zGPX5kVu`t^K$R%1>L@EnauSYg=I22UE4b$BjJHT=tF%C_Y7k@oS_QIUK@8<M5#X?D
z;BW%cXFW2@aawaw0dU`QGbjP#7eTF-11H^6aXL&<E%i_UlIh?KI5bNv$~m9Xw>P$^
z98K3w@*x|x;!`(j<pKjz)5P@{ZK64AGl@~}^wl;HKJ)-IkV)xQ0km>iE!SD~v7pU+
zJ5*_S30gWcz9b)a8aPHq5q9@V`;;FO^*!|nUk-CnCB&ouWv*`;O5s;U%=WGud5;aC
z=Lv8C2XB|gYymIC0Y`<i1ki^Ec<pmA{-EUVv_$suyOCjkR7CjQ(>&7vT?*g;4!-!V
zy;Ps0N7r^-cXyLHAo?*SlLBV#A6H^DvM)A>KXFMXWq?oNp&^;cSO_SLOEJn)Q5B(f
zGX<7{erOIjVDrO_Jb@j<=F$huh9#&nkB3YGbwOu+W`rC0v0&9G<cFZWN}r_<>4hGt
zI#p1KFt$AdUH~bBloSdi!a|c;L$>sIUqw(`_d}zt;(C}+9fDOC)*4uXmg-)?sw~sy
zWH5`_W;u_<jTjUI=+Gz9L>M$eWQ30{5bui=hi-xtaTTT#A*2v|I7#KL;z=eXA~bz*
z#Sn7}*|lKnHXobRj-ja3P)Xn-R4X8ZTx%~a$av8;i~<00NuFgq^;*;lDhVv4a;_Z`
zswwf(C!!-9Dgr?O)#mL$6pQOzXUk?R05+-ID?Q38lX@&7a6h$gMN|pXLogK^-h!gg
zsm|C_$_19gWd#&i=b95=X;cD48J;GNOsJ-eMAE|^6)&4-Gz6W0F&Mo%ANV(sj#+IH
zu5W6xKGLMT3#1bin>`D47z!T_>ExI+G7dfFz*}q%G&h#JtVhF(YFfXp!c}6oCW<0t
zrvNG?Pyld$Qh)$p9>bl6AEpNEGZ;6^<FdyI@o43nf=Stw&ZxwgIytcl#?8rT+yOL3
z0})<6x~-jq8KR*e=c>)SL`9P{rP;~h2F7dUDxcypBIiz34D&zSRJE2^ZO$#&*dl-&
zR9MB~#gUO1jzUD!xXz@=RIp*MILX{dSXL^i?r;Y9EbV*NBR$x&a6LueBjI8u5t_=#
zfNNT<v&soU)~D0fZO&ALbDdAfg9SIj(28yNWO6$}0l;|yqgwN&01_3;BtQgU+(QXe
zq8g&g)i@g5_w`7<02R`wt2;pO)!kjfN_$!&*9|UU5hjw(x+=zFG_<KysDPc7TO^OL
zKDdm=OjKRP(Grmlx;f$_T;8<DfUV)4^{89q2o?(Oplr<UOieSBl6i&-E?<Xz1Dnh@
zxcCXpREF;8U4cd2y+?&D@*egmkz^ub5vsm1p3%6Z)GcY;7N|iQkAqIDhywr<@BpmW
z!IpF)!AUr>5hKvA4Ep&men|iS-qOd9KTW?m=!do69VEPPI2dG5!u!FU9$Rt4SHR(Q
zEQBviJz?P~4spQ31e<!W0k~XLK=|_ahl9cHFWX6NKHuR6<qeYgc{US&yWNXF@!j$1
z%jEm<dD7}mwmWGmVDrRbpV&HH#u)=4k5wFhi;HLXsO%_X&wL4FIO4E){`e1W2QQ%J
zPzjWX!m>bUr^3@342%Y)sTt29jiwmY?&v;cJ4L!<h&!e6cri+?P(|W0R^!3Kl&!Kl
zJHeyrPP06CBV5pp1{s_I76xV?SDIYHB<&^EbKR_~&PV*H-mxCtyVg0UIo9zk>pnmM
z;NuA<8CU!BH!s;o#*<^7O@kxA!t4alKosIocMu$1ILMIB;3mzc5q8jRrhJj@OzYR-
z94;Xy#OUg^4qY$-S5}5tbXlbPNsNI>J^$^HgBu-`!|UFn35^g48-N0Uy*(fTnEym_
zK8W9lhG0@eYj_L1upkI;!WL+z%w*i?@XDRB9L0^QjW2p+x#OZ>x+4?$?0r0P!u_CW
z?|Vmd>>e{>6#v^h9x-x?u;c(>i1(&Q*%@yetzHfA0sOSJ@H~9#L1Hv<KmownbX3z_
z@6JwZRgA`VVKoHrzSl_x4;b#Q7I%r4cJ#PTOJltS`DJ(avV+lMb;CK}EgpUy^2x(L
zx&4Em6uRvkKF2z~67o4#z9-|W-S_EF=@k}^XMh!0@Za+5ilG7)?s+zorU6YsD!FGK
zqdnPo-ELiBZm^!NwA5UkapxEfe&%4o@I-DT8r7ufEKh145uZHQd3K>w2;!%p={kGF
za$mQ{#4a4g0Z+W%^|PQ32*>Y2{5@je^~Bxg)-)O%4;29JU<dGRVSx#cIC%&@AJOxs
z9(X`pxv9~SXw(kY!a@iJ?Pqy`^wXd(vx+k=dWOf7-#{O5g{N-tl>MDzGWZjASUPik
zOBZS67MwZC)4klS8~{c%Jd*+u!GbA5T`)XKqYtf*7>pq%!ddsUGaf8PSB6}5IGK;^
z03O02sCO_y(E^yoP9F7$e4`8dfS9WR9zckC6JDoUXtO}A=CcQKp4E?ev3rarCtc84
zZ-nr|KvjSP*o@MfyD3%aNl9>t04l@jLd#lUu5cxeyv+zTU1K|sSRSKAY-=`mEzSkz
z+acyE&hy`@vtPjMcjBo=Q%P?E;Hl}X*5<uS93ei!W0weZ4i11S%*2(*?8(?$4f~*7
z*`3&2jS&}Pl!U0VimM@364dCt$v_7HugosE-pwO|-oiOnqs!n9sk8wxaz{9PUezDz
z01jZ-1@Xb{Bs`LawFZ`;rP7pQS8#w;+?x*@1jPW#BOsiLQ7ys_jWS`sENuWTddVez
zh|<FLvAN{JcpyThRXlt*!huo&H5h47?MyG!<AFH34#NFtf+<49Tg8=lF+shuO)3uS
zx)-9(;fSSs$ZTC9)swMmqLN92O;a2A;1$ZN$ynWgEjh_XHod@7+ONfC7?XA{Sm>0i
zqAOZo;W_wLJ>PDIsX01pe_SxkXL~vhC=~!Vzoth4GQbW{p<7JSwCEGEXI}ily(imE
z(jM~g=VqB>oip%ty|VLl_UByxI+w3|^OJT87PtoIKVN}F&N)Gm^f!u6yV-R{$Qj{&
z`$tSr2e9{NBo@SI(!c`yeZYCz9ZxfgmU^jq)bR0`zGBY$OMNS9PB`NPg>;j&u*Iu_
z_uZ+feekLme#~5LAGjU;?x6H5gMRq-s~7yrAl!CW_6GI_zTgHXz9oIZt?VoB4eSl<
z4eSl54WK_PR`JjLmnOBUg#>MH!D3VQ2KEL%cmw-;#0NjzeYL%Ty@9=fy@9=fy@9=f
zy@9=fudsprFEwA`DepV(4eSl<4eSl<4eSl<4eSlvv4JQ5$o%B?cP#D}DZ1?(j##%$
z@|84uZ(r$Y@4H^Jfy1w&L&7z4J=OIq-ubBuZ_W4S+bVXjH?TLbH?TLbH?TLbH?TLb
zH?TMGg*Ncie;0ac^Dp#}_f_`>UVH=lTfvJzpgrE+z}~>#z}~>#z}~>#z}~>#z}~>#
zz}~>#z&mZ=&Hkd{&BlGFcfy`^Z(wg=Z(wg=Z{P(t@XSB?AKZJg-6ZWH&vI^-Io3l4
zzN1OL#b;dQjxB8Its4Muoi%v&|4b`zy><4j8*ISs+5rP`T)N*O_TA{=22NZ3X-0=5
ze+@#~r&wtT1@JYN@W5NP-L~L?&ELFL{gpw#`t6VHUm5h~kJ7*UB_iWDzvv!+Z{TG&
zu)ogcAI;v?{H`{0Z{Ynmu=W+d#(w4Pq?^3VWxyZ+ULG9!yz=wev+WJ+4cxPV{bhg8
z_O|w&H_+evJ<|69IDnV`%+<|ywl}ahus5(bu)mScy$9f*LvwP(y2uu_drurL2^P+}
zCujiRv${H;NB4WZ0o+6z_`D6|w|RZ9_qTkH^o>_>t=`$*z}~>#z}~>#z}~>c8$f?l
z%eFHd$=~tbz~Tn>TWoP*o4z-&H?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH}FL^
zu>a=li#+Ik%{SS=#8>15<xMWT$9(z*(3=V2Pak*>us87J4XCdTl{{oH)Dhues6{$(
z+3?=LWg9r{Eqa<!zUmEb>ymZFWsl)r<CUwdaBrRK5k2Z`7qq&-^43M3q?^A{WzGvu
znsD2^?gr4C0pYKElReqj-T->XBK&J_%iP{9$)twd(n(t5e~F+krAJ@l;$Q2Mum5mg
zf5NZzj=BD}`ibkT{oj!@aGwek0Nnh#fsBMogpx7%zp(%nlX^|t+1|jXY+&*$!(8_^
zl`+}r-oOiQV86#-_~Go)_6DxsK>HTZf?>}&*3e=XAGm@2ydJph3tG{y4Ep&me)#@7
z`uOpu={E=c@wML_q}~8M=@AXBGj)wYX{ZC9Lp#xUmM0`Z=v*q`@Tx_)woWWe#Fb{^
zDGp(P&!(jy1QjDr#=3E}IZo1ukEy7F@X|f3N>IjaU_S?-c_MpUFW^r1o_123NB%J<
zMhDL{ULUZD=Fr$cXiGL{BZQZ_t=hc00GO(^6f#1vYlh}hD`Re0#r;sE4z~#<lw%+O
zF!lixt^2qMhXC*g@Z$L0rAK<25D&qk8<aBPs6Np>QJ{+{aLT849~dD)!M~FOAnJ54
zM-;%&d{}HIo~00xh_I=101K&_Yn7mn{+I|)pSGOQ3Q`{tQpz6y=CV8o)Lfz$!KtA@
z$jqb+5zL>BRk-=JgkD{7vgt*s$QAJ-1p@Y=B(}H(i7#U>r!rz7JaDOyQxKrQoK!>=
z$iV@{w!=Nj6_XbBjF0tg%1eW(aRrG&k%HL}<18GGu-X6``D-F;umT7VeSFa)-5JyX
zOu9ko(6$qfo#H+*Iv5KI4lfcbV{~*KsgSW+1}wlak%Dmh*;*md-26$+&e>7B>1@yv
z5Zzm=@*usn=mF#gg#hYCcSPBjUec$Gvt<VK8Vl5zLFX!R3+mLy=#XqzQN$XGQjv7&
z=RzSDp%Nopk-$BGR>3vtSdXYh&nU`SeUV2VCIdNYsfI+BLXi>QqyWm!{1QlnO|PaX
zC&5c2wV*omJRFTj5T{fKxQ3|eagUINM+qsYujQry&NciJr~-)e1(xq&b_Igq608W%
zY9rlb9B5D(-~jf<k|<-vL<NnPeG$f=%|sI$%$v@D5EZ%-1(2|52EKJKS$@h@3z_yp
zkB;>L13^PuB$~m;#;^(tGd&FlcnP?M0x=dfjR`RtJ9ML>O0PLke|145Z|V_lfhq`R
zojws8;2PE$8GA&xvxEUZE?JawK4rTI8xIK-5|(jvVN;j$to80jJu+^na29T#7Pq(l
zHfNsGS0duC;;7GAKFPmWc|(tsVKgo?8MndRhFZ%`fiv)Qf=XfBsFHlqBgU_%l1edu
z_MmeIk+jiLRp~b|aP1s+SfDz+*E&iEBw5CA{fg9mL=-b(v@b^(<Ph;$QR{T1PrwE+
zpi8!l7wnUal;uyv63N?W*Ic0&0HXq=H^X7Tay|pHDygKCBUoS}?3xpz?%Shm%PF6_
z7o+9mEOA6A6#!P9C;>c%`E0}v$Zp8S;!WKGGmyr`fTiHea*D7Tf=P81UVVYRflF#Y
zd8HONP77M82A}|7ynyDLltL^*VzjgcO;-b1;FwO1H9tZd)&i|587d+ijsPaZd~_F{
zQcNJiu}Try=-`Yr5msz)Wv)&v@&G;(mT@s>anD2)(@`Uc_d+pjwk^m#MP_*LZnz-K
z#D$jm$c8+7js9{aXC2Bpr{xr{wIC3!&xJ~o;|GwQq>%wrjorvOfQq;fqjD?SUaU)o
zhE!8;UJ+jy&JO5ifH_t?sYGKRgVxlb$2Nn_Hu1m|%Ztdr7KO)q6N>n=&5$9@VOB-+
zQDWxtDO_@haPt7W3<^4>OjC2iHz}SWA`<&7yo_hvOm5Oadpnp{mWFWtUWQVvU6(+G
zFL<QRI$THQu$0e!$!;NthDH?ysl+(6w38^}Pvdmp%O2>|^47Wrtg^VBq1Tb(EL_c7
zOjoQmS7&^Q7+t!uun-|PXs4?=pim23i+~d>ruD(ID^=s87S;#zVqdFo4ND>0Y0sLa
zL;c)Dc@o#Eu&H5K2tn+k={gHrTZ$NKP!g;YL({OQ=0BJ8f^Ci~c<sI7)DE!yy4F=9
z!1*~KQWeZoJp#uGsJ;^-Av`&t62hT?rV%0$N?6Y<NumPaLm$#1Ei3?mgB`#FM+JjX
znm+5n!qI6AIu#Tct|zBmcMvTL4*XylHwr5n6qyLsqm?)uf(nu-V_Q=umY9nQREg56
zj<SLzC*jCuejenof@`kMc#DL#N(<zw1~KNZRUjJ{#88go0C`0rq|k8*NT2n{Y{$tc
zy?_dU5gp>CP~FOblQtCFMN<e&`E_xL1Wy1C%{7a1&gY@G=wP{&pX5U}Y{jQ(r(vBL
zsxZ?PqcUmb#;7o20{;ldxu9@3C^4noy8$&o+5iurzE9{Isx14S{DuZtA}c-a5f54r
zfs2qi;c^Vmq$G7v0R${U&mR`cXqpmqEHv-!@nR(nWz2kqse~E+)~y0dkx_(=E?B!X
z83%mw6gIyVXPrHff90Vvz(ZJpJ?RnNljP{hQ6<F1|1y{Be3EO9`P-dtvoxCYls0GK
z)pNmaf!YV|q9{5Tg34XiO)(tNV}}^c&QwaDnSp2sFGHd#zyTawrZHPk8p1u{f^#@>
z0s@B9$>EWcAvReH*r1+Q;>bI-4oU%ZFr&IkEFYn^X7gqb3&cQ0gde#Ex&X>9E4bv!
zxQ$b;f=ZQ!$^Zv2+?m-XkOSC=&Bee4kk{3?oUX*KqXVJ~W>UZe*9EPH?5J%O93O5p
z2KW>f8j_ieg@D4i6r(H^RS{}8Q(!6Rhvt9-Hb2bB6WB3qE`7jkSb{3^c*rDB7j)KV
zM!1n53s#LnehAvD^jZ3lUg(jkQw5a>W7{*}1&}gGNufX@EHtS#WJ{0tRRpzlKQ!7Z
zu7?TLAy{=`t$`(IsqPi5$}(+E2D6xLmh(v5h(R%c4t*j`gh3-jM)>Fg@xDlL=q5-J
zS79m<LJGl$lT_X+o@7EILem#l3^AvWT?@8u^RY?o7>Y^_l>{zAwE{B8wf542j2B(Q
zC;$+b<XOg3uSKn(lE6YL=h`8mni4O4B09pMA`k>nZQc$<vAE84wrs`%V3W$d(xa?0
zsmCG$_fz{;M3pc-1XHo$Ehq||>Wn?5Twpm|RzQJut~v3QMkPR$;c4Q?glfu2Bt7g=
z@v?bFL(usbgVC$=5q}e@b&grf2-i0?S+|~)cY$<*VzXzV4nyI?A)OqPM#iDX9C(Y(
zf#$|?m-T3PQBCXDRk%v*)<jW+>=Zzy1PTD|PYMtK%wxFI@Wa%AeFo!Zd0dY17)=^e
zFe#hT8I>4QCnr|HxH&nEJAlS$Ai~Q>x3zOHLo^iRT(x<ZsA!T_G)rXF=Hldr!JtGI
zWZbS$|I?GN2}FcV*8{HdDIQ}d>8Mi`j{Y;4A*r;ynSh~ULJlgd;_%|gNR0>8=F7#H
ziByEe*#WH~SNLoN6M#B^lekn+-Qf)IS=#rkM|!Yl;d%<buASOEF`G)0=pq`^S=ELh
z)oQ`kZO$Vgoa=l-9xU)A?9k3o!62WF%>gI?IPYOpW7~W}qGFi@hyV<?p$kaeo~s$+
zdR))f+xtm9QrT2UpRVoz!S9Wth?VxVT&^2jz#>c}opn`=$7pC%sZaqsE4N4<VSR8J
zjhU#rilZeWA9QoXN4UIcj{#f5J?l}o$Pp|Q-a*-z-I<zZCMEL>6<ogF)YsUHpU_O@
zz<rG^-$hm_@*ehx?_cK;1re%YV?3jANvT`XxGg~M?IP*4YT*F3HU*Oc7Q=Ft9is^N
zoP{HMF$DeYp!6$){`j{)v43TdSj#wcd#!-Os|~_+m&C$ETxlkr;t&S-Y+4FJP%+|U
ztQ%LG<0Osvn2IV0FWtkc1ZCU?_HzK5C$h)&0`7G0X(zRL<R5clbnr~$^#Pk`4vh_j
zwq$cQLU^g$s?DnlfT>zbAtMC4W@s+8GUkR=+z&<SaGOv<IR*j%V;?Zlx{sT12mpTo
zFOJ_`dZecb@enM!K`9fC>J!})1-h65r+jMnfe{iET>P{<n8h|jo|Y%`p=MZYE}o?j
zk%+LVa{vpenroGykG>kh)2A(Gw1U(}gp~3JfVnKs0X3KCMQ}tYTmqDJd^VfH&95bt
z_QkW}Icc#}h}N293pQ_VsnFsUNYE-<F4+y7Qc5<(&WQj8=A<I35XuRZ21a4h!d8L?
zQDEsdSL6#4)KB3ep-`k?BE&cgha;>ufJXkB$QrBw!b2Zl^hkFGH2{-tP&%~jgkz_;
zPmB)6f`Y?~#L5^QT}LWptd;=_a7?5i+<vxJMHEY8R6g*;AW|c?Ap7ggv}~!=OORsY
zP*a0ik;)-AC<Ku5x;4qZ^pZYhoc%IT@EX=zH)hb5nRY>)+87;@?J9~`Ls2S{F8y36
z<RVmJgey{Ai^0(4lx0mi)+4g(8AUm(FY?HZGmxVwWpyDW?jhYWx^<0=Lq<EMqMQUT
zjnsnLwdZb9H%9@a!c$Py;~pV5j}lT)U(5XgoNM?cPz4a_3oPHm><R?IC0G%j)keC>
zIMARnzya)yB~iwTi3%Do`yz}#n~5eim^YmPAu4nw3Ls(A41DWeviy{*7BcOH9v$lg
z27-pRNHl|wjbRlQW_lV9@Dgwh1!62}8WUnPcIZY$m0ok8{_28A-qa)90#y*sI(;HG
zz%{HhGWLjWX9)v-T(T(Ve9CqaHXafvBrN0T!lo|gS?k@6dSu*!LWTl7D_3#*w79+X
zu{rZ*^~^^;2xpz`gZKzXJ_#=pFZ4(;(YVZH+#1@+so;nOaBxHo^#t;Vcku{w-D#<m
zXRHNI5&_CaNmZq9!ob}1&^r^nIMNX|&opEuREcPX@J@*tF^Z=}d{)#t9qALW0SxGp
zE#n3IBqL?{6R|{jG1aafeODy{FzO5G);G>s&S$_<C6#n?1Pe@rU2{UzeS4H`IptIL
zVziu`C5{NC0>G*hC4k2;pN-f7*$vrPys2AY2GY0~uoQe*P7zi^FsZJ>t1qxOa7hg)
zuhinkX+bO102Ba>7tnl@Qiw%JjFz^b>1rSg9Mj3M=0|A5TA(#0Lq&wc5x``akM6=#
ziU~wGRw*JI9h{LS!io*9%+-lS9>7P!GA`yU?wN>UI%=eGscrv0SHM=%e~HuJbd?-S
zKuo1V!MOrQ_SCD_XxyB2DCeA(Q@qxKK(sy=DoI`|fb=Ad447){M$Q3L#Dy4@ThaDn
zT{1MJntJn!_`+~@KsN)-vEoT38v7WurUpH>8Em$R2c}qFME<oXJl>m7#Gh@33~3It
zDw>ZHGmlT<l0$@>2iRp$&?#k_nj5}J@eC1>*k|EoJnLq1lMdS3!Mw6Gg!A_@lw$3=
z1S)*NBX!o{Ix>f)eD+Ir3qdqAswhY$#-XL1L=k@)rvqR1K&O_s)-_<2#qA8ejudC%
zYTjbHVy(G4<4eTo(v^jU2)RK!UCjZ7THsm)oM17n51w7A8XvW=KA0E#T77F+3fWG3
z)+`<B=O)UNxK@Qt4Z}hRVi!%<S=icA#8`upV4WD6hCOXRUYBOQV4LF#UVE=NwF7Lw
zu630NaDEPmR0Z=?kHB#Ps_%qI2u}{Egm5UJX@p3G64o<IlBfXq(1&zL3kyKtU<dHP
zQNdu8rq6n?aC91jP6Y*q>&a=?9Yo8513y^CjlxO@k~u8E&ZaOjfVHq7KoOdDgOQQw
zBplhm6*_?f>fKgw&DEJ-)nHp8dC14jqDqA2p0I!>BK#&18<IGE)+3|X4_DakzBBvR
zCgq3{(4$gSRba|bky0XfwschvTcQHM0la=CWW)=I>jpJx)y60rr;`c%BOK>~!r`FA
zlxEj}8X#?e2T&gq`i3gYz9+w-0hY*0k9)*}7DV78WKOso!!s#K9aI1Ti_r6j#WI?v
z1RV>_dwaZCNkbVkUtua?hQD>Iz*1xsVWSJyE=|S(pFD-lZ^c<>Pvl>Ds0{EBR$xzh
zg!d#ldU8|=aq%B@xXvfJ=9s_T={8HFNl$5W7G6CU>=vkf;4X@ygCVHgW!)6R5j}Q@
z(d<m6^qComhVU{ZssbFq!DSk=1*IX}6D~N1BPSqWIGr3GIT>P;wSW!kc_og#Q|q7<
zKnF9btHkmVYHK!c_OL(<R7CiZYoH6D?6QJOu8i9_<tnIDX{Zcv0K=V`Z2~!fjo4fa
zTmX4pjmzmu>^eFix?m;+OmJP$YRHb-R>AS%Mq_|aVWA<J$yf*|j7u@fQc)G5b~6Q*
zf_`WYIAHU`j68uI!{*Wl%!VbXGLMH$0(C)WeP)Ck`LSTtDCCEry-J^@59x&-sXA3q
zi7>W316}|rgOn5sB*H?IT0^$<cwa?OTlYhwt>Su^P#uC*7uFhBf|lxD!Ky6N=43F7
z*=9M9#Elpf1L)8v(nJ_ELS%%GE)egF6o+nt6mb=%5+S4zd^kzvt>Q^0BqB6@am5gG
z3fZ+_>oy;o)Q+L3)KE#_B2+6NgIsGbEy#G$HH-oPaY>$KJoQ@C3MvUKq;jqu5~?Zj
z(kG%L94Z1q0M+L0KopDXTxZK>EC4pC+$%lGDwBFFB5*&oZ$(rI(?c*78{UGV(5cSY
zQ_2OF!({~&Sm&A(UujeVL>ZnYj!dYgj6~AI9u+T}XEX$ze=!)nIv?>jky_`NwTy6m
zQ<HV;NqHAYCnz?17V0n*J{;1?F==ERddz{h*c@nXEO%Lth8NYeeqDvD#BNO#MaWJ8
zR7#)#;Qpil0l++lI}JZf4cKQeZkEU87?07UF$I%W(~L$X#?;9P^5%LyId!v)Zv^Rb
z<`VY~Y-u@`VRNJ7bdpx+OJwfwi1UWQphOmA-0o5T)03|WM1)P(1FrIU^hQX^PVF`s
z1_i}3NtVp$AO{sz@xi^!%yBx`&#*))!qUJ2tuI&jYy}g5I)Ib7R8ZaF4Dean_pC>H
zuxH_V3cjwL+P=Thv`IH8)m|YZ1W8Z}wr=z4&2&!TiSz>lBnfc_`E-MT0)X=#Mm6OR
ziHc<sAOf%vqZF%A<!T%~VBgmx^_EmfpRVoz!AA=!I4kXGRa`f?fJK-{I_s(!kI~Sk
zQlSENR&J3z!usGc8Z%LK6-P@%KIrC%k8pX@9s{<9d)A|Fkt0|ryo0hayE8S-OxnyP
zF5^bf0iqaO{125(MccV-jV<3rRw?ox_9%6hiHJq0#*6Wc#wDe0N#nLa4bpfVbXr9m
z0GNOWV7(5uq!S5F!jX*_q5jICKk@BPuKmg&VX?!#fxUsffxUsffxUsffxUsffxUsf
zfxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsf
zfxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsf
zfxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsffxUsf
zfxUsffxUsffxUsffxUsffxUsffxUsffxUs_4bZO)`uQ(@`2IWk`0=OdHwXQx{kwyX
z*R~mZ1A7B|1CQ9i#4mn(#2CBB-oW0#>ulhhZwtT$z_;GM*yY!goo}7!R^6MAf9tik
z?)^3}c&iRd?|09$Ck~cVAzV3ohV8dEus5(baKZ+}H!TrnW1H7*0KRh;<*Jf-V<bcP
zyfb88XbpRQfdX7BEL^k{*bf#i8f4p=4eSSC7P8a5fxUsffv>Xx_(y;!)KDN_XFvO<
zkKBNKr|q}fBRBt)Ui%S0W$j<I>3+A#Z?E6)b$rpUgHO9Qy^-~s*(0BJXLTQac~#oU
zMH}cl=ZMdX8V+oq*Lg4$EcI!71j)Tn$V-wgzB=Nu^vN0OepE-gjqeRiY+%2=CYH1*
zdjop|djpqj;N`yq^MHKzf061)xS4};sJRt<WUpIi{50L(`KLY6`|p$A6XlJu-yUY2
z=P%@ZM>F!>>FZB;_Z7Fn-oW0#)f?b%Bz#`oZ~N~JyvGLiSHycfrz@uXxM%;M3)MCc
z>UPDG-=2E|uf2i%#kIe}W*?H-lu46|M`o=zZ*O34V0i=j3$A1uTejSSy@9=fy@6NW
z!2T9@+SAFa-1|IjpWA3}U~gb=;N%VPcW6FOw%>LS+yH!ru<*d)PH$B%b?en`)t!Eg
zZTNm0=x?4Lz27bN#CrpK1JBq%{jQ;pXAA;*=E2Hz@)fXO1{Pz#%uH&KvdjX>Sdn%z
zzXAH)LFrcp{i$!iwtr>N{MI&eZ(wg=Z(wg=Z(wg=Z{UF&IOnHc-~!-VuUG8y>&ecy
zPIRm8&Bwp>+FQ3b@P#)}zv}lF{}+A_?5pn$><!$$0rkyOCAWKS<=(*Fz=v*Ne~<Xk
z2m21!+HbyhcuIT5sSSMeo4H=@qwhHNlsD<(4aj%e`@P-XZ{mv|=NGUq&wM}mH+a-n
zKcuVu?_aT*(4>+7cbQX&{k2DtW@E=2NE^&cESo&$Z^r5d>Sp(XR~NVWGaKmdD~|Y_
zQJ=%%j&zoo5zf!)bFPb>vFGRZd4?O`rJS6hNsj7BxADD!i4E+x*Tj-GWp7|_;LC4-
zzE__A{qlUrlOGfh$Y)<dM^0|JZudf?tPlUDioNqUeQ<m1r)=Q-ub1;3pR$<UbZ=m9
zU~gb=U~gb=U~gb=V0Huh2f*1)X<psi%dSrirV_e`1N7x~wm0ws8_4g{{Z^fQ5@u5-
zO)eGjh1m|9_qrS4r%ETsuRHD6m`p#1l4<N~>|x(>Z(wg=Z(wg=bp!cNi~UE&>e?3g
zd$|kC+SI*)y@9=fPu{@S`U9cd_G{hwli$5>dAIiV+waD=e13cAy@9=fy@AJUp#5o)
z+R6Aa-9NBP{gpw#_U+f#er1qaeouNtyV9B3-k>zp0neeGXgtdkk|1<06>xafB3xT1
z7AE3KGw~FMFu-TiQV@cQ5hr8axY`^iX~f4=R6%&@9#$nN<2JCL1JEils0PqrqS4Kv
zlNk4&(E8gv@{Tz%I(Vjj&(sXqM003tAhac$vk}5e-BxX0T>#8I(it*B(2--7L}eUz
za;8V4VS#EUY+z$hSoe|5j$_~tVDN;fO{C~pkMu|(9)d+TC}qM?eWH7!Ko?WsluzwG
zFhYWYe<ue()ahQ1D1f2)u-HsIOCcf=VN>S-7E(3WDnTFpF%h0VZ8@VAq&^~~ls^ER
z%C6!7_NZK{3yugoQ4$zs9iPpnIONw7dUeIKRg0y9oAB9444_jITik-Al5t>?%7}sR
zz@<V?L4X2tQV~@k2L}|}4)-WmOj_8=M=?~QXia%(VEc13g^Pqjkuu$!h;bGUM_6qD
zjr=u{wYmCXI`r{HkMs_p24K<+N{6<caO@QKiP6DWP;hvWSQ(?E>qv!+)iPiKj)@e6
z+t1djh+=7s$_JhpL~6tqWPhERmMxWf2~unvYHCm`QaR)Xg#c1scO}-G^Lo;up`0;f
zS~s`_hS0pmIFxIdqY;pPE)=*3l@{TO)UU2!pjB{9I@Tld>KR2jt1t4%T{DoQmTE|3
zDHIv;O$wmw%rAjN*z{_OauU2WQVXg>&%@Dp1aV4*fNO}V9`^`Yc$AQW`dV%Z;9SEm
zfhvGVUtswjW>+8xF2RcMtTxh3#(@Tv0S;hqEQvB!OjOW#|34N?xT>i{T8n}Z6;#yN
zPC?R#;#{N`&+<qS8@#7cR3z%P(-b_B+^#$@9~;9eEF83p%D_v&H55qX6Fo`d9m>2}
zk`xsO>aQ+hQ``ns5bo!RI~E>r(Xh@&cCERQn6%MS5SJ{<IiJT4$eEOpi-8Trq|k!r
za(phh-ksT_MlJy=oQ1oYYJcl*^EM_Z02sA{IO?<Nk9-ndth}K|iiyT$CgavHx|$;v
zz`<?0?v|s7*eJf}kw&(Gq*BbEJ+O)}mo{3eD!lxDb~i&D6(F^XYBbMG0!fy!D##X!
zBnvH$88NB{Xcj3ze6G|w9qALW0SxGpt$$#Z6)~iT6GCA56R|{@Uzzm~7P4w#bS>cG
z<uSr?J_D92sic!5SYRUTniHbrTdhq8PWjZm7%eAfi6cU(0C1+rnv8{>VLn@*4hVLu
zDvT0CZ)pp%qGvr~tBRDvmzhRqSPel$J^`nG=@Yaya7hiA4E#sqw7_ye0l;v=Zkv=s
zTti0q6qE)P3z>3EC&#@JMG@Mt7HCb$P*LG337df#au2N*L%@QSDWXb*8-=5qp|&*A
zL|75v%3PgT<N<snEaPI%;+}~prlUp>?}cL6Y+I0fip=of-Ecvei3=_BkqvqF8vWUx
zvkv8)({hT}S`diV=RzgP@dHRt(#U|R#%|;sKt)`LQMnauFV-bPL#nAauZS-UX9sjM
zz#J={RHCtuL2GKzW1GQdn|NS~<wfLQi^Ai*2}S(bW=IhIu$q5hO)upvtjeXMK7kyz
zOnB6g>#l+d2v-JzBtoK?G^-*kBqD=b@Z8yTF{79?X*^PlYz6XQp$&u?umTHkUa0T|
zkMNc4r8{0Dmhx%f&MyQ*qq2fj(#1OJB#QXcI34)12RgM}_qUy7mHi|}@zuC$P@1i@
z)LfnMRbq7MYQjQ<+@PJV=72&ia4iB(u$a~d&#qLBk6KtC%!_@kzBMd`Y^ObImJaoE
z6Xi)<tHP#+VIc&ui>B)=Y;7rGtU*bzP7F=Mp5{Med!t_HScz~2uf12C+5xs-*Sbms
zI6ntOs)Bi{N8mUC)ptT9geM17LO2xAG(sdo3G109NmKxQ=tDZBg#{pRumgDDs9-Qk
z(`P+cI693%r-A~*_2jhc4x(kjfgdd6Mqx#RA`_u{v=WCyP(c!9Y-`HI5_3_3Dp6Y1
zQC5)TBplhy&x0IRaLv^jZ;{YeX@OkTAjbT)3S`5A7|L-Xz+u(E;RK}5dSsU4wC11!
z;J)W(Py)m+f?6#HPP(V!beN)A>Y)H6)4>^VXqH%%b3UbSZ){OHny#JXLpE&1r*72B
z1qP(1iR%V6X%)w4bC!b^>Oe3kHABQ%z(Z;XrmhWdK^x`d;D!cRA}c-a5zk%_fs2qi
zVhIomWDY8TfJK;$b!JjiEOS`U=5X3pXo?cF)Q^)TvTg=sOGRO@oTwDx5$0B;?I~+A
z4)|oM&2Pn7XHVo`d8iEV5LRGMdL-6h#*-XfLX{Awf2pq@9PvqTj?KTZGQU5ehLk`W
z)KUQD%31h&o7K){+%E6Y!4On-<e~v^M31R4nj2CneP#xtA$)l{5+E~=H>3rnA@*1>
zuU5`_j59r*94^k(WG!HWac9?$Wh*|2o4kTj03FQ8Gvzauk5F5)d9#NFVxS_zk6Z&?
z0A-gITykaH#wk}prAk9(fCCur%xn|L0c^zPV&DSE>uOw1S7O)E0nr6BDPV%@f>uLz
z)V2za4>uYEd<qK<$xOyVKw(^pQI?9T2(_CjuoUz|bHD+cA7<nU>=-teK43O1L6vzt
zWD=+gI_ond+{ljwt41L|1npJ&EPY5X^hnjIf=Yz3?HTX_NExK0P#_T&n$#MyrN{d!
zg4((t8f_KV!-VP(th%t)z!J1n_X<{JnKmbbS<E)ec_ePcpcp`hK9MHEpb;V?d~|_$
zU!*v66Qqc%FqH@)h2X<UDsL4}G9eM6>5D6dm{Z8E1zWfI*raw0MWu#H0vDlL0U6|4
zduc(&i>_f50EkQSEaR!yqE=8zU?G)r?T}DSiI+YR9pO+B2m+`!ZwI1ST<1DlHe&&>
zN#$PYQC6AMV-bP-seLP=N|+vkso3xq6opQ8#-36xupBNcpujrUocKzk5+KU(G;w4?
zHDx4{9`>ks**v2m==_Vp=+*g%zlqd3$E;<9>zkUaTTjZnKsrIO*|SiGq4435PL4?<
z<IrOcyv61~b7Q&7dNjPKruFM8TqSmEq9{Uk3ZPN~1pxOa1qcA<G2ChRVQRoWgK@Jw
zE_<91k5;}Zn3PTFj7p5DlM}08+?<@o9YA9=5aH#c+uAvpAsPyDuG+jyZjY|;ynPy>
zykRgXkp&sIcJ)6!`I<mP*mOPMDxcypc9M=dRpIDAgBg-a%bN)pDkkKh!YU3gj*QfJ
zP;I_kjG0J9SezZu8ghltRxkmm12~CG1=StS0H39O&w8W>dls&z;Op9{%@eb!G>I;v
zF`ZRy2vV&UY~AKO0>Zh@C*;8bPr?rE3>6IW+1MO_0)X=#Mm4t0CnPGCNq`8za2vXS
z)a|*NA+E>ue7(J&)FYKmh4ks_4iNm_D2iBVPs`=H!38YBMABJT#dwT{HkAq$u(NWD
z<Pp{fm(iGss;f9!BJx2uM|_0KoAwy6HQcivb&DLqLg5{hjoF>4X=YL~&rreT>#%QN
zllcZ0KcSh*&>g)iu&BHDsIW!e!yYA)Ohha~)i=g78kdy1C5_txHAv%e&}kKM0AK<h
zfb}}ql1?Nz2}d?!1p1XhKmWxK>Hptb`uOpu={E=c>5ux|L1+aC7pn<EtFk#iP={A9
zgzHv`g^9S*OgzOQ4Di{s6ojB+#K~9}R~_nrxEG<AiYf>%J;JI4Wh}<Bu+1A46%Y<@
zxGZQR+!@%$HjnIMPK?4ElW`=8uxJQ}4TQF2bM``bsoSc}s|$dsT1z1#1iNNvj<npc
ziu-YvVA%&uwC>aIG7|xR0JjI_ts_OpdZd*h9)d+TC}qM?eWH7!Ko?WsluzwGFhYWY
ziywOjv)E?H)AD3K)C`Nw#j_M55)n3a4qzcwbFC8e(N{xw`n2VYR*?FLkW&5tFqh>y
zpym?22u=+JLS`mqh+zI~tisK&CG_fwlT9y5MXrb!DG;y^C9%aVNPHQ4Ih7Fu;ektq
zoPq!a=A<I3Kn@NlwjJ(Ku9&p2XMC)0Q(hWOjVnkLiWJO-7-!*dgw+Pn$X^p#gB3t{
z=;MnX>CT`AVA2grhqj$?>=gHj(ZN_yaCnhe8Ka}?NQI2mGGGCYi51FtK0$^ZPZ1>z
zXhCkGv)0$c9;swgsHs7laTGvqPza#Nav96M^pZYhoQ*P|*I1y&4Ej}(*t{h*hAC)V
zXNKBSM<XEpTqtl6c3ozv6eXcN1kIf*r&D`Gy?RDb&gzRi@>ChfQA;%>vJ{Gp_$CEV
zcIKBrB5Zm!ML7vx8mR@<q37XfJc2l-LclddRgZgwEIdj`L47SZ1#qt6mp~Ojq%W|1
z53?%}1eah%cvc(fCgVVZ$^ZwjH<m;hD<&#vyzGlG{%j_i*kImt285{4l_-FOO*8PV
zd&%-su3E^n7kYH84;Tm<+9J^mJ~oC`SeWT)IKWH5H57=ksA)`y(b%CI6;*o8f%>Zp
zB6(Afa0^sHIP3I@*Z|kC&dAs!x}7Bq_;JaiobxH$Mc8;qppdYPqYIn5oM)|fH|mjb
zLxr<&`?R>d^|v|moW2qfe-%f4&hknA#mXCcqzt2RnaQ{f?l#m~b_$$<rxR2P<3^R_
ziykq4J(X07`LhR|JBXx>ma0m>iGge9u)_k?>AlubIv~k1hU-_P?jxd@5u<%M!XSr;
z&x%^7BYgrkfB{{yWxQaYWTY&AB9=(rM!V(;y#N>$AiWt51D5j{kX1<~ogBde6Jgh!
z5Ov=kWm``9)V&xjCufNxLa6|->O=|PG0bNpc0hJRHWqK{7MOuFE(R<GUzSsZ)euao
ztMKXz><wH}1IjD4xN%z0N;Lol0OJKT-=q{`5fY=NEoizL$O6Z7a;*6g+OQUAP03IZ
z;cx^n8RnzA@RVW#5sp=g$VLZeq=~R%gDZ1&Vvz^%k+6)5Ig5KHqL_{vLA)1=VY6*P
z?kO_EgLlIPVJ0rL%ttol*=zKdD>>^>&N(foc&!D2XnihJk{myP^dyZ8m}=}s&H+@!
zg&37v(e`3pGBl)`dh?3-!f<v#Hv`PE;z=bM`xvyQ20gYJY_^F9rdVD?{<SDP-kVUw
zpKXQ=X%4e0nvW7Qk5A!}Lxh_L*kw@ADP@|P8@@^L3=xsoXW?Z$>t=G34%*wnys|We
z^Y=28V(q#FDty5sb=KiJGKZyn_DgmPK{Pa~C`cv7p{1Qf5q}z|17G$)r<S+YHDHy+
z?F_w+6ldXT-eS69t+_hmOT_5Xm4$@}xj{Q!%>jj4;93NnU@@%^o?WRLAGNSPm>2t6
zeQQ_>*-m@bEFJ3SCd!kzR)tLs!$JsR7fsh$*xFLWSc8&aofw*iJvINiq!(;+T)}Ja
z6{mK9?bo%g5&_Q70g<X;p6U@ePC)gY5DDSQ0hJIA1vHHiiBQ6NW=Rqi03Z604ryTl
z2psGH9ylr(jMDU34;GG2W6-Ifz;Hb|?YbxuED=V`^Dx{ftY}bVB2<r7;=`#WQN|>)
zc{2(N3RFqc<p7l~#*sa|r4u-y-fIQdT%GZD2yK-X)MiHq=C4&C8y3W5j^hA%MIofn
zaS2GD^~h|;$tb;m3V;zE;-ygC%7K$M6x&5p2u%5Paft*^01nMHi*nBAp||K@xs)TH
zh&!N@4O{VPEHtb$LltJaVpJxr+!z%`OyD2kI2RNS2PLModpDp4NE_e*)b|N}LzQLU
zli$z)OJt?TJ>o$NB5)BhCtQx<nUtgsDu94R==sB98BJ4yj)msEJzlJ&p^TZYFqJUF
z-?~*`DKd(%(FJRlCgXrlp2Ft0;;geL@~=Ep26zZ7uqQpjdy*VIIjV%X_+RF7olkPj
zF@L+$ZI(upp3>$lym~I!El~TwT@*zJLr}TPx+#Vudh8IR*_le|Gcynk;bllv1vr3%
z%QR*SN<+9OTyPFYPC&qLIypRYGQ=ip0UOlwN*sBo)<G$N4rWwWiRB~I)@<JFVSyN^
zi0~uVKo>ySWd)a98MkrDRZywYP#NF=hC4Ic1abfyvAGzy0P?yTm(!Klb#y><!AuI6
z;JTpIkR7$Hg5$%D#sHtfLPIi>u@F!gmtvHqqAEh|W(q6?{m>k6z~+Y;c>+6z&7}{R
z4NFjE9uJuW>VnSt%m_E~W5KFX$PYn#l|D-!(hEIOb*i8eVQhN_yZ}-LDJc|4goP%x
zhHUBazKWo>?uSNO#q}_uIs~gOtTnI%E!Dk(RavIZ$zT?<&2k=z8!;#b(4kMHi7;q{
z$Os=@Al?@#4&4MP;wnrfLP#O_aFWVf#gj}(L}>criXr9{vTMQCZ9X=s9Yay6p_0Hw
zs8&D*xz=7<kny5x7zF_0l03_J>b0m9R1#Q7<y<=?R8!)mPeeyJR0M(ms?FPhC>Gbb
z&X&zs0BlmZS9+9HCiPfE;C^b~il`E%hhQo;yah#}Q=PG=lnX3}%L*v4&NU~#(x?Q8
zGCWNjnNUp`iKK@;Dqc3vXb3w0VlaAjKH_g8wazhX8R7b-ChOLd@-C20P;B-r)L|%m
zIHZ$f(#SaUm;-OIIndl#?y?>YFRE$%x(Zi`-I^$hkevdklt2N%{Ye1=fO!me8h)4>
zu+L!JERV}E9-~QP3MOS!I-?R}>g2>K7&j-UaR<;C4MceP=(ctaW{8G@oU1nP5*1C-
zie`zd+FYF6Fc_4`f{fc0>VJCjHGzn*>3YCbKE-3~Bpr3C!qI;QGbELkHxn>aOvpin
zRUBR%8L9E0+I+bfGm(n0I6I&<<O-jyU;<DFa1xgasymzkK1=(a^+*r)EL=~)*R@ld
zCuUP=5?w@NI;+|cq*^W5y3Kh6gmaxw$b$u*gdN%$Dj4Lmu{i(*0Ovi7YHXWNNK`D7
z01<%UHgo~0+jBKTT#xJddV4>qM=F~N>C@F6Ao#se6tU8tmdka63s{7Sq_eJy@fZzl
zDitbVXXO^jBdiZDqcIazS8=pN<b!UG_z0Ib?J;0$xMw};7CC~2!aFD%vpZAM%%o(V
zp@PfTVc)<e^9`PU#|zuEfPvw;#+L6Qs}y+;dz7Y-iHJq0Mvd`|#wDe0N#nLa4bpfV
zbXr9m0GNOWV7(5uq!S5F!jX*_fqr*T`jtU{`rB`;{mLL=vBSNAy@9=fy@9=f*WJLx
zpM^IMf2H@$Lv7*Sz}~>#z}~>#z}~>#z}~>#z}~>#z}~>#z}~>#z}~>#z}~>#!0j7&
z(qDU?*!D!L+t;|2FR_9BUU-Seu!q<i*c;dz*c;dzSlYn;^LuGso4hx$H?TLbH?TLb
zH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLbH?TLb
zH?TLbH?TMGW*b2ND=~z>+4c6gdjop|djop|@3VpZN6-5_t3B!7z}~>#z}~>#z}~>#
zz}~>#z}~>#z}~>#z}~>#z}~=h8`ytOcHM4Qoa?U)`i*aYX6;u7rC(g7KGLs0;*Zjv
z9xx;5tb~|HzfdXvb}Tyj2Vs;|kW%{1ODR9@*IhWOE}Z6*e2QNVYMk#wcdFa9dE}&Q
zX=1Xafdg`O)`F5zf&M9knkix>RCiUXN$!^&qG~54_w*6YmkV@*{J|GUXPGEZS6&K3
zzS4k2#=4qHQjH5Ezm{t=HBCi>wU<f>EEXU+Pw^LGRV2Itb&xzvn&~MYy=|YMW6?<d
zg7qG`AhZPcj}#N3KIr@|%TFeZ3RlXz(IA=-<RNoU=`$ge^cLqEwX8ha+q`a5={Hlp
zn;0p3(#pkzDP*{yg5PLWkW1yIyhroHt3V@92I|c)sR1-|pTzP=mBIqBz8bz^o+9aA
zE9tg+mrO9T({nm(CkX+)m!VmwnEvDRW?6ne(~BRMpz=|w6w{YFz0gSqePlIJ_>+k>
zFB=wh=J83wVpU#Mk{e`ea_}D!mRZEEGv!8ITIS2Ih1p9tMR0_h+@NTw4c(lstv`_W
zk~K1^c=jso>$@Z&c@`-w9dkv(3vqFf0*|yRmuQXz=9gAn57JgcQQdY}dlX2srRnI!
z4GG4c7%gut@%sA-h<Xd(o%BFvU|w)<N0A|2mEPYdoV45^mB_#M=#jn;ku-90rrk@H
zo7O?oVCf%PJHJp)2^hrSnOr2|x<H#?Ha0d&kSEFWz$p}em`Dq1PUZzGQA>r3sNgU^
zx$aKRY+i!2mFIlanQ?P5SPDfJN%|FYRqm@O@rr^<)S>v6ijb7IVFHAvZdM>Q9N(k5
zY8g3;XGWZsmnT`DCk>#9RH&cKA$j^>qHc9-mi$WZY)~@gRhr6I9xY>a8pbH4xVZ9H
zE$=y3Qcb$)G|y7yYK)mAH=Q90PMK5yCA`JTnM?z%TME*4>Ie<(J^@sVxstV=(<3Gc
zB)+&H^sXp!Wbs_fGe*i4oJ<|Ry@z;P`I`(eB{|2Dy!ug{>`6vYiKT;U8~}d0fFz*8
z=U#W5!t?Hsh8ka-c0|LPRv9)JG0i{3u$aoNZxLAIQJ(A~lR!eue3KmH=bneF)a>GK
z8W>3dF89icX;%#+b>+4Ig%0&=#p0XVl>Ip|4RA~EHBDw8N!l*NkX1tmmZ!_nKlOux
z;!ptL_KD%aB@fDT8!>N^?Ai8RI$ZHOGaNtb8E*U@qR9``NA&}FD<dCXUL%?S=?pDQ
zBQ1To)c@CZ!WiN9LqH--ES*_d%L}cMIpDML!T_=Iz<~neg^c_Ht176Ehq5dlI}~<x
zn|W`gw_o}vf8_67Eg~@`*(++A2_#Om+K^^sNry^Ah3JDcm64CiDNsRVnc~)tf|e9k
z{3<6j3^S9d9;UXv&3QSD^Lc1|MCe8>8r<^8E!FcPCU~<HdS}Rwnrn~<)L0T%Vkwj{
z*^<xbQj(wF#F<=&>BSx$40e2rk(E@BcxnjRsrd(J0bJZ#u?4D2PiR9uqe$kdWU4OL
z;#bP-<hRTQJSuD+RC{o?6L4XgET=fXX}Q+V{9NS2=_M0MKT|q4t<%&s&^u{rzM~-y
z=BE|WKlz<o5_&Bk-aEVH7hfA7tl+az9ndO^K{YD(nAPS>&yp+eaN<KkF*<y>OvsNS
z9<n?obPG0!wYQ_?9ooJ=%iCwgF2CK51a!9}&Smwy*xTLL<mQ?tg;UJ(puy7$d2)K&
zxR3Cm=WV7J$6f$f%x97U$MWu8eig<GW8F@Dc)$DhUoQHfq4^C1p78BUf37m#9-(0+
z1Rr>VgEtH+BOFZr=v&%;lvm7_826JmW)A2jedfcwWpmx@50NQdxale7u)lZkNHyrU
z)fApAom6mtRA@kJhqIw#u}HSunTaLm4z|jyJY{*wP_;{L`6FTM&=g0(vY;8#a`+zt
zjZ;t4Y)zDCG7<PmVRnyOw9dS&uw~cs<>*WwvLKZfr<^)ePHxtcS}Nsa%mrykF%e2J
zpCv6;NMbE=UdfS$uNBpbS7iv}#4SV6lg}gO@k*>PswT4fhi;_)$+@u4A<sk3WK!~`
z%f40<<(gPs^TDRovlDy7hE>LCKRvbAaKxPjgDy!Q0xB0zP{#8itM|ra>^*3N){+_*
zT0TAd^^r7w`w(HnQlMa!&?XK)@`maQKQc<u2d%O!HGdV5q{xsUz>Rx$tS<pOK}OI>
z3Qq}{5WC{icyOg=H|Qie_rz-AVl+7!k?CgPKl$Efu7Fq<!uv6TFf9Y~LPJUb2G<N&
z4LYJ9)teCq`J$ZklkeXUv3sf&e5M<7o0AC0D0fP)`Mp~V2`qSxazZbqP}6Lb!CK0q
zTqHa5Q60;>f>gmr;v|$_)juIO%k%K{g_Mpd01*bQZ-HcYa3gJ7?w2pA3ogsn9i_Bq
z9BD8SG{7{y%!XbY&?v#k3rz!c0Od=E%i&fe3WWsMf576{g3fNyBUzUWjgPa`I`~L#
z>4U~AAN8zg1^dCYgp9jAzRCmfnI}jIjoSWSOADOu9gO;z+lfi3p-9U}4hSZlhN@Y8
zALi#sYk7f>G=!N$r#h1Zm-l6Pi1j^Mo@;sDk_uw!UEU~JN@lSlfHSwqMECp^W%KrB
z7TW}9nF38iRVYHWGFdV1ui7TRP3TUYP@(|hf~9hvJo<KuF_&&W%|d8Q0oAGDL!53)
z?}Ick`%UXYBHQFQMPm@!{7X<@bMMAG46>>^WpW>0qK7_HgmxFtPR-8m=i3Nmxt$XT
zfF5}Kk>9%F;T2v*)1Avd^hV54w~3<Gq8A%P0e}_9CyQ?xZ*U$`E|7OY`zW6@A0?no
z;ZYGWxYSH~P`sOScA!qfj!d`xNSAB9&?6QNPHI~89C&hI7(iY(g`~Yq@gQj|A6J1X
zUWF1Q?`$9U&}hxFbKALcqeN>=d2%YF8l}1&XXSmEx}|_V=*yrS<s3O)R-*y#{|G%o
zQ1xnz5}Y<l^R3-nPV2YbY_29fC6Qy432IM0jHI2?_g)ho=#~wYHM4`5B!F6VxzWh1
zxvF4Yq2%<nd*27?%bVanAE{oQDSrzhgSp#`?dLq?xaY_GRh3U^ACDfRU1#ey@64qr
z+st{TCuFT1K!gUiz{-X>?JE<iT)Lo`4P~9Na`My2Z?DvVn$Z6f{}5JMV;gO%v#e6m
zl28x*ksDF}(oPp`X2rD*V5?rvPx3v<hQ~K{Ev=g@w3%(P;caK@dz<`76}(e~<cjNs
z7Yac!$Z_Z|lD`HBR{xa8cl_!)#v{h46CVI8F7=t@V?`%M=&~Nw^6_F4Ls*qCfE-6O
z$lnI369wo4SCFFP*_qvkJ$Q<G79lA!&>c}E%*((Pp!Zz;XaPYv=?HVgnHraG)BVgh
zVVs=Hba_OsWp)+PAwM%wv@l_IONxV!%m`Rj?kmTZM;-em@}>DFwR)e_k(%;n4YOw*
zCLdz%BqWv_%mLZt{ukeA&a;j`LuM+UMw3^SUrd40Qp^YQ%68~I7JSq-C2lX4SV6rQ
z+vK{i5d4*}O`)vlB2yo#Rklz*RCYkoAUL&ErZ*+k!#nY%%2SixmCN0|RFY*p&y=c#
zIM2*{?#3tM_B^Uv{xDBw<=bU_nq`ZzWZZaMjU7)y@f+);q3*It?>qEQzCW4gPkIn?
zBPuCWdOVVHYfJ;xj}|bXPbv>6pafuHjTnT}CDU1tEb8<gB@Wz1u&3foqAbg2$$DP9
ziaBW<8l{}NpfjRGANd+|qZZ^aPthPMrjJN(uo;e#tfkjBpiAbfmY4hh*5)(pG;Us8
z9=kkLdSnC}rcH<S>QD^(D=@8>ksat^(qQGjHKf#m%*(}0$f5#so^^Q+b|^;0ur`tX
zYdpN9<rnLO((L&kVw5NwaY(28=unSXoQ3HnRq`j{G<5wZcdUl8PEavpEs5&?%VLUi
zZpUeNv4z*m^QeOtJ+WA>F1SoEBX`GtRX`6vHD!LEk*>ad#VV3sXVowS<7Q+#{OI)*
z<?-w5Q1ueU!@G^A%wLT3k}u3Jbe6d7zON`5=C9z!=|R@y{oVSK9N@sVVsk6Zw<^=N
z%j8*U{`?Oqyi1;)6hV!9)G-q!>4~X7iaJ~9Vj*T1-<ZwanVUqTT8upNB&9>?U580W
zE;6I=8imSrU^?JOX#vZqkbC=LZ&b(0GZ1e$gg2T-ddF!=ou*Ax7TMk%G<o(4)<%W=
z`TckFD}#RiiysR0$DgL(9Q2!OzdMNfyzK`)&;7ACus5(bus5(b@cA3if8y!v^KZXz
zxHqsjus5(bus5(bus5(bus5(bus5(bus5(bus5(bus3kDf%11|`)|&U_O$DJ1A7Bc
z+d%#Xb>VlZ3$35Fv)y=aU~k}o8~B)SA`iaEgWEmu9kE;O4eSkkkqzu`V_)P!?`!T2
z><#P<><#P<><zrr2A=sN+B17U?ioGpmQQOuyW5u;@MQ*>-Q4Ew4eSkk$OiUzpAUJE
zS6ySjL9W{D)xPD)$>-PO9$h`fhxw=f(*|@Sda2=uU2tD#Z(wg=Z(wg=Z(wg=Z(wiW
z{tcv`7RwLZ<(0GgQzZGTG9{JS@d_v=WA{(E9rgzH2KEN_2KEN_2KEN_2Cm$|{<-bS
z?QZYAfxUsffxUsffxUsffxUsffxUs3-N04<GIW*i%Pw*cwm0z94eYPDr>_6An$zzN
zO20DbH^2Sa3w~u#{XIzOKc;eY{qm?+{_nc<W%9+pR6hIq<uTXaft8*$eb6Isw=Tcx
zE9VWn{tftc9rBt^o|>KlJ&b&oCsUtB`SMJ7DdoO7Y=x6Yd0D%<7W7?;R+K(Uxo)Vu
zm-FiAT2$`;yzE5>gYB=&U^%xfWvHSz(`lUoqYWKxAg!OC03XrQAPa=EeSVDj+d@UM
zpnMZxb8@LLIUZTEwK|c=;3^eK{>D@<Q{ZrnR1zW@z@%!X<z(5c45?%)X%l4BluTN=
zgyzRl7tmQfBC5zlSWRS=p?t~*tqP@QZr$u^&<5eW2y>QRTABT74@zsTR;s>asYK#L
zq0bUm-ra|4H19SeLmawGxeDF6hIBEy8C`SmPr3hqk@wT<+0S6TstU4eo`t4a_*kSb
zINwek&NM87o(q~uYNAat!ASG+JNpFvonwAS@f$e_=)Gn>9AN76>9kW!jM375RGS}g
zmg%}VfSq803F53=I?Q8GE-Df7gD`&wOU_ANKg;(OA)}H=XF~dJxT=})o#b$9J||^|
zv}M0#!Qlvsfb0b>!Y&QLUOtGlT4x8+uzdJVku6GmC<z;7P?}-Z`dfbFbx)1w<?|k|
zQ2YP0_ihQ2EX#2iuo<oa8<0rxl9bk<FDyVyAwIBMg})!+mX%SD?w+|_GnL_PW^V2g
znN>aaoC6FrhQw%d2#)3>cxF6VOc-!IkluA@U)rFI;DhD|(xG9iZt3M9n|82&C$v84
z+>$Qq-DA)1_T0WX?Dmuf(&>)5l|Sy!ZcVnTsTG;X4`Ry*P|l-PyHcX~mAozN8#R1$
z83}g|IsZ%_ey(#p<D#&cq7p=IXaR3|C2|(DfXaX$D?<C8i01o}*9ZE%ee%}~U#ouq
zP2hQc+E8$t`l*SJn@sMTpkTf2F>)Ll?gZED@O|WYgH%Zyz2B0Qhw?M?KU0m5@b|Ee
z?&xq_{KInLbjB3dUqAe5O#4Y;I}*$^;cxnT0>3Bldjel|0_T4=c>iRFUyl5z8LE+h
ztJhBs`Tev1?82vAx67N*pM3JukAB|KPe9tBEEHODbo{lImy%h25kXCFzgSKaofO@6
z51~)|T)>Y+xJ{94T()t)2{_mL2BUJC&$0Rb(_JGkTKy&FI-@SBVbOXKq5MnTJf^#>
zzg5Gjqb&4H9$87vB8?X9Qj}*;u$oY}M-I8_l{lS}Y)c+-8K4Z2l|EhtKa^eX??X-R
z+NR;|mi!eudPen?LWBOKRqh~9%%5fc!9$-x{%x!#hSN7k`FNz~fJ0%$q;o6(lMd*s
zXFWb0zBk*HpFCsomb<83(!yc<WLc?hhzQfslg2Ohk<hQ(R(^qxo`Bwzr+cg$$2T_J
z;0IoE_x*8rpnfB5Tea1!J@-JEmM-uPvbh@I<?iBv_iG5Z6l`b0_u1$To^tci0I5L=
zk@_`!YhZw_ui5{>)z2J&e;@m&TRyf8w^JdKe<-Y8)J_-v_J5|m`MztFvGvCeCm_x+
z=`6QZjQVx4yyaBTB{kQ5rK8bs8QB5yswWx7i%wPd&xiZpm;Wli`2_0oQZkuLZm1CU
z$k3Bfq%P{aSp+FN{6x+sxhc({4{?HRL|vUREw6%Vu`RqQ^W@K&XsHjq_$oXl28$O1
z>hQ$oa*6iJkj8iXNh2?dY&D$LhEZCeNF<6}bqT1I+$FIoEW(U)O}PZSqIi55&S~hf
zc|wz`qaGM3zy=%L&gL4HCnGc(nYpb5bDRMOxhOFb7QM12DqKDbjVo^gLd_w9+Ff~<
zkuDO}MjpF2iLRE4es^eimoIuj1FE<(bx}_Sg~`0e)>zg6BO<0|UL$ZNbCzTA2*e^=
z-87mdDurztVG8e7J4%SXO*X4^j6_P9uWb}Fj1H+<NCv7!?APm)c7>bCwuR(sKW+R;
ztko+EAfnbei)x&zy(Yep9I~LaGmzHYI>$y#nAHont{fm`!J}b=)f1wLh;ZAmQy$o?
zU`Ez;vF07N(`s~<SuCq%JVQz37!u@R4|;9~=(yb|+h`Dm*jgA*JgBk`Yu;fSCswne
zmDR0erG(JOLYUwq(qabaCi8DAkh&!H&;xC4GZ602iiu>6IY_K^i}Ge>Wo05u*G%Nt
zCAtC16Q9BaOWVa|+MTKeOR~;l-KwRVR}d0K4mu5ei!vu}7T|&gG%zJiW2+ggyg-!L
zn>57x-F2pvsEX9<RVmAC6^&jK=%EIl7wJNopo<0RS#m9NZ4x29QnkE@oGdLWuqSa2
z3(C?`m<cX5F%e9v!i2V(_+-`pT~a0+>qAIdx~1z^dhth?po|xVi3QA8V|?WXo>9=9
zRz8mT$ANs9YiimKzv%!eCc?076_N2Klo~p#g=jr_J@FC~HR3)JJF?v!x<jiG<J`U}
z2%~Cw8)>MGXj>95+Qh|6O(w-ElukSlAEl_Jp^lnxA8i$oW|{J;xI<I3ufVQPVxK_f
zh9^V>ZPBhO<a9UtMgEc#@Y^f_*p;aO{iEU2&&iM83=g+<L|>cVJ*a+D(04`}CC4-;
z{r-a2nnBNm`?bEBx@HYzG7h*pVLC8tN=y#AD-Sd$TrmnL<~P+_uC+<V#R9XCr4$k2
z^%7@OiD?=$lcE=|-eLvZR2#f#`_6JcOZ!5ibsbYotI=vmlPediC?oe>lE?EeE~X1N
zm9EdbJzq)3AtRjb#`Z6JunEeckfpx)l%-R`DAWP?T+QVVSl3>7ue9<WPS1}|X58qh
z6YC!r@j>&{a(P@aPA4Gh`I6~HzrTR`1CNHE@XsFoM^0s#UX4EVBuGBb6x4T?XDS~|
z)@1xmF@1LB^(Mn#n;&@sc9-+&&vFJe{*jyb>`=oejeiE)$CrK{%g<K&JX2=*nv06*
z&4dJ;{a(KE1o+{V`#|@mzU&6^>i0|TrQc)2b>(mFX5{ypo(>jpFR{>v-~*ecCiiS_
z%I=x%Dzg7@i@oW$AZwTWAwO=Oj>o?x5~hB@p5ft$cY|%N`oL4RB)ViV!@;Dotjf<l
z`52evju|(ubOy%<$=3J$sp(S6hjpAl!8|!x>svm^P7=O;5gMx>rOPZZO><#dI8(C6
z)qKuTMV`PNL~-j_yhZ}|PC@A+yS3dU)k|TUBJHa6V_|;e2;gb>esrIkc{d!&*7o!~
z?U8u2;Uw=;q1IJ3okzLXofZPaeFE(u2jEr1d8p!W1Mb)YZ(w)A*CsqodA>Idx%GaB
z@9eQqC7i$iwXn^}fLW3cnPxveemnAXFP4x$igMu!V|~E6+jpY+R|fqb|M)-8{>mUf
zY8sxCxcGh&Bzn49lt?eoKqLkA%7Js!t%9573b!fJImvZ4wkRt@w|&Ijw&@+O!BSx%
zW|oNz;A6P$R%Cz`NJB-nJ~TFuA*|*kF5I0P7R*OTTdD%iCRt!0YL!XV7;jj1rXpGB
zszCB-APbUG3%oApArn;dl2|7Jt=Y;9ur&@`O6(3Z?;<lW+ed2@u4uFJ>X5}?MrO{i
zzRsE%^__PU%5#auyh>t_Dp#tA6Anqhc)L&$LumlbJG|*qNt>3Ktz6^`Z(K$<^1`ik
z8B0yG$n|qc%=BlOIxA>GS_LAR{lr@kXS5PcR{Ca<;=)!X$<<O8&J&`!7>Zh1sT4k6
zH%sd9?91C%0|4`6#L1hFg-(E`r<a=eY5^;V6kuJIF@deDNt}f18W=^It(8X*XHEn%
zLZ`|MY3*e;mfMQ(Wm{gB%387vDZz{48>Bef)42Xhp<@~ZMG<H1BEkjAZ}lY+3;7zE
z+S(!#%l6eRny{oPNntgE#F8<m7h2Zajz=FiPAOKL#adY)A@Y^o&92TgW@bH*Ownb%
z<*Bo)ep6T~9~aiCO)%Gy`|{Xwv0ck#(m!0dq7N(5jt+SidhEd}$nj8rp}>`h*>D0f
zC{KeTnc&GbK;{iNqD?PUzH(=;Sb;u@V!E!*Ir7;c`b9(?OXdkZAT+W(`3`f=nL@PK
z+uS%`7sF}gj4%a_6O~>?XlkjankcV{nWQ=r8_Y(2$sz|OXjDs^_<nlnvy(fbyImcL
zW8OrPC~RO7_NR@?lT0_UBT8f~7V!fjkBL7m^2|{C{P4I31(wLlR~}a)60Ar_R(oi}
zm33xklnR~kK<wgtA}Wnaaq_Y9CRY2_#YJ>Z^~IC1oKIuxprvA%Z3EFrt+^n5kS$z|
zNXX(6P=Nt@2$nj&f`u7OA-`l{1FfW3U!>-`ROS{O6s0Aya&d)T&UozQVvil={>yNo
z?OD{7r=d%F8@3LDYxP7{++VEh;L0mrvm$3oFSM*TXC$Kb_M1jnjBu`CE6NygSTiSB
zMi)_-jly2$G7bvNB{qQh#;T4BaExObdO<=H#W#pCl~PP1qWCHc$Gl|{@F|2wNG@Cw
zR%we=ERIyoa%Lo=R%|H}_N5`{MJxB?%mK?3NU4Q1B^8pnIL=pAg+@}ZNCqm@>6$Um
zNEB7tdqr*}Ly(8=iDgFHY7Vr<i&#~xg*|CiR93pm%Oekr$-GV>1F8}fSnq0sGjpRB
zOO@WI4VvE4wNgk6Lwu2WQmCy3_X@7Ga+b)HV#i|jT#H5|ORohBVJlUcT0%m!+(H&p
zohaEsI1?7gC|jeySWpq)Lk>)pR;DGFIm3R6v@$50Dl9V>=!!OC1Co2<Qb@Z|ql8kJ
z-`1&M1`c!Bl*!T{Ud^X~DVgeajFEMb-#!Mx^nCIfrke7HgW~>Dj2Guuz~39XS4zFk
zo(%5B$s+lKK^oR^GffzbCn7Y)FGRo)Ibf9fy1&(~m9%aV`U#^pQos!)GbFMZseCce
z_PwR%d^D??yXy#)c@+svBzKU{k9mspO(HoWE+v9(2vvIyBw=occ?#E)R9{}oiUk@_
z#0J{Hlox>!hNqS0HSJn0q=4gfHH=n!d$0iKJy=3VDKf%f0rW~@mCM5FfQpQ@hLF(2
zLhyBfp2FR-8MYM39QR-qa%~t0;*W}+6w6#0>b~2~h?}=tfhm|F_w|dgq?G2YWVO=d
z!ns_zfGGh(PeN7Nn)VYO>|hEm>KAopn7_f&#UInu%x18OVkdl#4AaGZJ&?|puGK;%
zB;{*kL_k+1Dj)P7T`8HPiv=Dt7MbqR+9yJb&A=2kr{`@j);EmO6I`_E%I8lO=8`xy
zH^y3XzAn5$OT-|u#quM}>oQ-zyj2>S3sNn+l2$5Y<vmh2v7j1nq)E?7kg*i2k(e+}
z^BldTO7=rT(W^O?v?S4Kf)f^DL{a>F#HUDSyk_PLSy2mXV{2}qn{`#vq_9?awka48
zRB!Q0D3>-onY7Z=B`XU3w}v#oV>%=G=<e8$B;sgd&8Uze$;CXmkqy6z=RN|kKV7yL
z$@7bI19L*Ly>kJ@s+T2+0uYhvc(zn0yQMAj$gJ`W=uWMNZpu@#*<XBoc3Eq@Rw1{^
zNSrmX^jj9rSid^xR1$HLAma(*0^SRUY%VDI3ML}2@)~U>7K>Il<tZabx@$&AdKo9u
zOz4ieoEq3KQgXv3rfPfhNeD@(!L+hQFNZkST8$-gMr#OqaW7t^xy~&d59%Xuz<PF~
z$#^Qo)0j-<i%A|~!iU<U6+}#tWW0fDss~M*^a@f$qU8((T3vFM;#r=&lXitbo(NI6
z&I9&IS<@wl+7T1Utdtz$NKT%?@#*su#;`v}rjRvVl}L!VDPEH_zwrs8aDWF$ix7kG
z$$jq04er=LjgFXFnp9UyO{gn?lTR(E`kI9j;i0I}tXEVPel%z-`Mv6uW+&57@=WXx
zo>XURGvq$yf*MH$W`nPN?5mwV6cWo<Q{K8wW0kg+!Tb<9dO-|EJCu+VQkRx;;|Uvu
z$lwsCe(S9Zk6;%&QfbfGI$J@xhU)m(Q#l#M&Qhiw1QAOwQ!eVnLRzxneo><XLYp2`
z<Q_7kKUO|_*kyH~9vN+`5-nY*AUj^?A9!!&L5TRz9Oc5DNU)AsnzA}d)Wra_*0Cg0
z?OIBeCG0Foh6+N7l8z^gIKu^&ENbFbu~I})CcCo}37l=hOQLmEXstmi>KZ*UQwVK^
zu^*H9nk77ue9nr!i%XCy1<$)CL<I|`-|3|rVctb7S7>QjBP2Z}^4m2+#N)%=4Aj<|
z`WzGx#`wB8zc6_@;sN22M5G0_=!99DDpy$%P+ok^M%S^;2x6BiuUQ|H+jdEs>4HkC
zH-@cYq;dnjQxFbmfSGhVVbNB-0loCXb(1b&^V@)_O0ZqJBYOk&8>D+DB&s-NhS1|J
zsRzs{?mAWpBVkxrE;g1~1IlZ>tmPZUzFK4@841_hAC_7uRQLs}5iRXa%?#w2Thpg{
z{kAu0>=#p%iJGeEMF1~mwy&(yi&hy__G{qf>TSEKIU8*-tTSO=mzcIqFAif*f)WMZ
zW0jh0H46~E`ccFE<&sPp$SdChwGv{uiZXjlF-^3TQg4e3753bNktm=f1+wZ9T;zvw
zWPiV}J8<CHeS^M9xsi4*vMyV4c6ScShc38~x;&HB_>fV@FgCXxF(%ec(*gM=h%cAS
z=a?;7Oy5NEB5c!m@4|2r7Z*ja%Y`d=$hEVwvfw1@iY8!+7DX(-*N1Vv3g%)G_)986
zDfRADw5!N+Tir(kfCD)EUi2Y<>tSoNtZ>bmbXoD28No{!+K9L!Qi?3|Qq~L?6|UQ(
z&mrF_HPBD8X{elIb7#LPA<W()l0k}=hzc|9-wN4bG2+JsQbpxrNUTTP@9ssa?x^o8
zX)bo;tJc+cP3sa0JSidF5;EdzlCXKY2xS0edKw0)zo!(`A_o%eVR95~UDHfd%eY@$
zMUDI==t^B@QzX5L5<+rH?a0JWb#87d1**8HDB<d65h+FXR;?FNkT6O08V&`H#7#xX
zLu!#BB<48ddOLSbLi#P&u4*%6%)~_wkaBy+Hd*DkWF(YRRy}=xEX~ST5E1wGqG+Hp
zrZ-KX6y@s7jJSSke3Gb_j&zmDvz5f7VHmJTHR+~^ueu2xkjndK#7&crjJ#%x!xOf+
zr3Tuvbgmd>Ho?vaseH&H(&_@R4Izx?@wJoSEX#=MGrq@y&~@*roZR0<TOV5;$mLFI
zwJeM##B#72oGLg>xQ32+U^N@?h<I0l`K*;fO?gRA7sOYEmz7$XUwk+g)&N?ef7k0+
z`Y?&JUd6RQmLzJ0oF&Lqk9|T(1E^1Oss~^Mec|bA)evg$$wSXVDeA~RYNm^0E6;Rt
zj#)X1Zwe_;+WeAwfYc#GS(9T^$XAnXTG1t>-~3w9IxM-3Y*1Y!6I$lYys~1u66H)n
zE5`$Frz_OD%5SnFNMKtz6UYcvP^wT}%!pQ%us~FKL%Cnams+Z{ta>dn;l(EsQFx*F
z(KSjVvQbGMudpzTksU6yCB9aDSnw0MRKmyAa50E#-SnpSMrJht;o_v(iC-32%&R3_
zw;tWKI=%wV83<zmMW%9=ky=YcEF&mKkQk+_4k<{Y3%*$Sy;eCq*3(4;qos$lOmn6V
z8ZR4ri3Q>LZ3*j@NQ=?D(#3Qs42|Aztq(aDNvU?1a>u!XcZ?nx+Hbt1FmjI&af;ip
zZmhxPL-Q_BewVf1L6ev$^3-I3^ULqz3b-Yeb0iE^siSz%RU%$Agc?;x6&J+C3cbtb
zZl!H?i;SC_KIA;1TzqP;vqlVPBC%ght1B%Aaj+83=}Lj+X<cZ2scwOMgDYu%9hIC)
zx0izwc-n;<L4dtUyyBd4JpuWZ1f3FDMw8e*gkM+ry>LfX6!9b645h*{StpE{iI^}F
zBV1Y_0jff{yHTK`Nn0>Z1j*Op)sj-5xKli)BokD9#iZsQLLVhM-*YyCpgdOV8^Pxl
zDFYl5w_bA9g_86@t~5jm_c~^!$gDh*QcGF6Lz%Xi+}lg48%QZ6jC&|x)BN*b?Q<1z
zofdP-U#<cu%#}+>sl5dgfMx5#D?t(%?6}v2G(jAy1uasZ0IL!xjl=47;gxJsY#6V4
zXBDGGyL9X7wM1s)!2;+k`I1%+jUlm|W*Csb`?4FS`67ahNUVj+AztRs2qzSAV=j`$
zyd4^5GbJRW9NUFQRfS<@t*|+}3F}%L*BU;Fj4Nvzf8nc4D&4T9RLfA{ULg>>(+M$7
z_J%9O0`p-m6Dl^`izV1$nMbSK+P0Aa8_Ark##YgK2fEdR&4$(TXo&FGQD`Mf6)C(>
zb!yMur`&}yf=l-56XY2it+7&GQzBuYLzE;w?84-?Ey|Eqi4-tM6d#IeD0K~!%nEE=
zto)=XDe9yY;{sSRXo&)F;6PAdPU7Ob<KBYvH=1AMdycw`k6n|;Jevw=ks4Yw;?+{Z
z1UHt4$(Kg7RF#CTBj-gR))68Fa^+?TSRrMJ^q2zUZ0#6zvRdxSiMA|+>r}z&lS}*N
zadQo9n`G%6@)yA_+Y#jO#rPA0!IDhm*tay_zC|ZBhtq{!>|W`2z|u)6U8RV6@GpT@
zNIxBUAqfS#5WYhFt<Sg`>&R@WF&k;EceAdnQ8g{3qZeNjPK5j>%0fcPUI1Ryf-pAP
zFBM-xe%adeRlLXz&;rLf@J!g0rf}ob$`+H-R`gy_lzz!(1zGTV(M9V@?um$07U+sN
zk#eQYPZ`xHNnPa?w3!I9t`u<MhyagpzZ)(bj8r7nGRk<(0*(d9u(^-*zR}Fq>YAx;
z1c_6k^F6nA2shqVsDp$}Vh+BJ6$aYsY6}1XU_OV9Gr-p6&0)UTiL=(E5}g$gBN*Y%
z2~1bU#7eFr%38&hs4BLo7n4_14Li8<cj=K4`<Esq$rBN=-ps3Li&dX+Rc0fPjaw`K
z_K&nHBrqL@tD<I6HRCh7s&;~ZtT`b(Urkz)VzKsCL@^ad#$lKg^0vqaNkzcNdnYM}
zuIE6r6jL~Tu0$C^zja(7T_hBx3?Q*eIN8~0R8_&@>sBU*n5dhi9JEP_H-^?MpFp{z
zLfFBsM~_~Q20(&_s}rI2T(Y9c_HGXatt9jji6rF_M#PhaauLkg+S*WBYMtjJn<>{s
zakh^l^017LOLM}4<BYB3Hd*eq1f*9>jRCN`n9f&J@FY%UxirV#vj+562K~o>{u})X
z^k4qv-~Y?{n}h!A+20*h?4n^rB&ZoEDX=6Lb?WH~4R~=7cZQNg>;rYO=eE%eS1W|q
zPWM!`CTF=`;test)?EHL22}u7znL}ERy#p1a&R6A(}+R1_>ZiDyo`eV8OtThd8>L1
z*nT&q5{`Rd6%W=^eYflNJvV*Eb8P%@DniZTi4*0hAHitN(S=4SS9oyWVkvu7&(#dT
zt<{#^DMK2BZzEe*SYU!zSJ+@CW^@e!iL(KL4S)~ZLKmNLRE(gMG^(X_yh_$=bi1C4
z*vSOL?zaWNNBt4I*%aWJs>>QcHn^p0qzzGE!)<&>m>Cev@Qj+3^^6tjLt@cm>7-~_
zOU>I>M4#freUKK&Xd564J@#P9b8OXVT{Kc4JXN=FtJnZ8m{B+glAjLVeR|$Db2Al}
zbTwOB-Dd-oDR@b|^ddk7@4`<{8R{{y-FY<o#W9>ydy^drSHL;!fp1fhAi^x7Yc_Q5
znjj5liy2)dh(I{5uu<m98MT^e%@sMGv;&4p5N;M1yi!fkQuBc)>jDix{SNQQwL@=o
zk?+=u>_l>bBp?8s9G?t<@q&c`ywQ8tT2QlHtq77W+$dYS2f~|fl#O7t%oZVyDD!jG
zQXu(2Kq)mE$N{P&4O!GgG{_8sqD<u23itJ%s@LsRJ3=)xDsF%SaG#*%L)8FL5rCf|
z+6<b41^~Vq36zZR1ZK=*53a|xbL#S3B^+N}=0O~2f>w3S)Q%QNBuxtKUoOBp_mjX=
zD_B5a$=Jf3WMqe>AmJ(8ICC%SR#4bTu5|+|P$5bJkD#gUjzBss%?=W0By?fTQ4=-x
zVH%yX5yd(0S+=dZU^1TLP~-?HI<9$M>|ZiwyN7Y}ZD}r_sg82qLA3#JJK@JYOF=VD
zse}p0&K5tQO}5rhMwDm@5pf$=A=0A;UEgyb6B4c{0g_j|X2n1gq(HcP>;gd$?u8!g
z%Ild1{Q!~21W(u~^zyx8N30t$g)IrGo_k>%!0hmj=qNj}BWUVJO66sNSLA~5n1&T9
z3w1=Hq$u!+DxUT6pG<Uh&9!C(OVeVW68Y~WY&_so;GA%rf$Mn&r4-D>jHcnQLoPq?
z0MZE-jIR@YJ=Z;U`Kop45I@E3&Q&`WM{SVB<sr<1T$1CcwrFP^@gQ3&7+geVDeEb&
z|DoI}E0rawl=Z6;Lz_>)v&rJ``p}JBTKT<Ii3{>dV?@Y?<$Gf;k0&T<W$q(A6?;MQ
zTo}Plk&a#gMAV2Zsxb>UwV`h4fiZbSWWh_4X3j<kMkBA<YRhKpP7!MF(m)fh@*q?L
z6l9N7Txuq)4C-nRsx5$vh%`W#;T%W(jNWq$mwf>DO;2ssw~1X|oCHl=ad3++=($xL
z;z&+*!TC>!fViKfw%QQ3B0O1}R8~+hzyqXKh(S0Q2FVegZJ@?>8-j?K<5aNBXq=&t
zn`v5E=xPs=f<lBhnWwYfQ44sr#EWw<lQWvf6vbm*?)AbhK_mXc<&XTj=*k^Q4{FLp
zpH7Au2FRvx%7cq-WQ`fndWuVJuaex+6=+P0y;D176C%D=rde82MM6Cs=zt<}y#$cZ
zId*<W)N&_wWRn(2N>Z37-r*hDdMu!*N|M|Ik0{ZT?@m8bU0K&Lvp1|l#Jlieq!Vk`
za^P9s+R#DhS>5jCCaqQ4mX`I_EIKk!hT%kDC4nYP#JjqnqEh80P~GI4E`W&Kl>z`K
zqAF@utge!=Yy|O4RbI0+Z;a6;!QgEtO$slZLX|}+mM{TS(bCx@&<9azF(gw%IDz+i
ztZ?ZRT$VHG3J=(avWhr-+{X$AKWr-EqY^Sh1<`0zWMIXT(lrxYFBbqKx%*sLx+b-`
zY_`p?0p>Ij6K49?R&n0^4baun>t0!v<Lj!|*uy?hjtGQrQx8IiJm9L7r7)hT0&aZ_
zK>vUtd`xL8_MMp-efX)dcD`;t8=%q?ylPj*fTFZ$h^KI74tjm4&KRh!M9r+(hHk)8
zP8b;=923^|q8gc=;xdFj#PSc5)Iye=4BppJvB;YmH^%D0aEWL17)Lgz{>X=L<7{QX
z6Q|`g0+&41GcfQ?vK0Q00NsOvxim?-pywCTM0tqH4{#*)9yxHa1fFKiL3~kvL0vS6
zs)&|6ClN;^j+Wcn3cYRO9jn2h*dPHErll%sc1=octEeS~w8T~$zU4X1vB}Fi3ucrH
zFQ{Tn$*i5xPzN7TwSE^~T_Y?0C1K5uNGZTaIDsjQLs*Ulrw;amN4Sk1DyWKHgEoQw
zo#?U^XzY*}>2*fEC~FU-Ty~(((i~6ijbCdDNy#Z)DsGvYwUp5f+W;-51#keK<g;`}
zw}wcD@HF`lf+_$Hn{lNF)@1`85wO4t`}&|@(WJ09MX*AMb>-hK^Psi%588weMFb5n
z{~@MuD-qp(bqk@ib4kiHm`So(G^0}1xt=HUix5>+Tr*yByo}>sW;S=!0)mPp;aNW2
zpLLsaANv1mM0cJX!bO4NRg=)(!;314BPxQG6h@P}q@jg~l8Bm95D`h|p4&X1#;s4R
zTgMG(r^BTzO3D-x-%{;1qe{ME5r4jVs}&SL3Fp(Wf=$dYTb41+z=oE}YgVL4bBcpo
zkmmuG40mqj9SW<NnNg}j%r%}_jboum-iq;bIc7jF1<y&2^O%O)YI~48P;W9o#Er0m
zo$LATy(>0EfcOe#lTGE;H5haQ)Ge)L_3{ikkN|NAhXvXV)d9S~DM_Jk<7(U@iW$Xb
zylRKc*=ZEYozowmRG?JN7TAHFQ{7!md+f_(z$zs&pX2P5@f>GOj-h5ECXCDw?u{M_
zTzMJ{&IKw7nFkP2ZTc=ewJLpUw>(j1PN%SLs4jbhSW4;2(Ygpy<O9U~DITV+vZby&
z#aAIJ!Gv{#l_t2iVjI8=@BsDZNYe41kLR$7wA(4Z9e)@^6gw$cNkSu0CUYFS$!LzV
zlj9!j?%Ni^pjEv<K&SBBBCV0-w!v-T)~^A`DDa4+wC9YD;1g%Q`J|ugb@ENAG{W03
zNQUruBI_8%78Vy(H`w#RXMpy)bL1J@RFgvQJz-9Iu&Dxwh}1RndQ6mplBD7u4!QyA
zRkw2Y@t~FgxQbHN^eq)(2YYUIXE;)uo2?hRi88FO<5p&obzZ9p5=0?@Aui>qDuvbh
zoVZzEW!pd#{XPutidqch06iK8mSn<WTiH~9C<wM+dBjq%jbf~=<DP0&NDB^sKxNff
zJA%1t#ZZHYQyjKHMs+}_szlB5t{MZ!NoTQ@)RD4#W^kD>h(|c;^W*S*2o&cHrC38l
zY}IN@X|>p+1sEo)Wm+uD2oePd03&(nbzuwp;YsF}Xg(hV2ms7;U2*Pvxw)ch32+DS
zyz8wws>bJxD&B>o!FT^1krPvVr+1_jwDV4>7w$n)ouAqLy>c7xUYf3CCj{AXtHjLY
zW>vJ-%9>@}CaokaL<GXyK;EV5Q<g#(UknT6xirPD_JB#0RebVESPEfX(!4FSX1ZFM
z{27vqu!APaIlLhu9sN6YedtO>N!ai!F3~k28&CTuB86qR8E15+O0_4h^#FYqq$Ff&
zu>oe?h?WAC!Tl@Z>+tA=BKo_7)?XR)U;pvn&i=|E`m2V@i=NBV*E2PzJwC73qdjOe
z?0cE4&-H=FXQVO#f5{v*NRy{9EDW?Ym>#L(TF6J<D__pZ_0;5<n$sS$K8>2fS$mE0
zX};)j-Qh)62K?wzf8+LJcRqv%FCN@ber|pKp%~gC=zJXH^z-mxZv(W?%@4#5Um>@A
zuMB=FcW-0&RI<mSElEBDNMzA@{OV`vLu2)gyxs;nK%&oZH-Gg=Z}d5u*~Rkn>eca<
z3!d<b!&P6xykRu|!FWC}o!P~?O@`i$gGuoRVc@AjN3#BB@^@?wowN`0hKFd}mmQBc
znp7&HAt(9>d(we=U%qsRC+&~6a4OchaakR@X1FStT5)+yjSP?bQR&8h_iIhL$}Eg^
zK?IAXT{#tyuaxtD<t;k-$65jsVQ*u}ovfnE>yf+duGFf&uG^`07FDAm)3SYx*{`C{
zRQb-U)ep!%Fj=vD=CtauJaAaIx9XqFkPkgR&61!2YbsAkTM9VY99(}C^f`Jljr!WF
zJHDJh8krj0(Y!p!HyCXDdedW2xAGCw+JdufH-qMRXzhZYTws83Ab%Eq)CLcDr(^kK
zdckX+Zu8TFM@ReW{N109wH)|Il2@m`S5KqgQ+`gR$M?Agzs@h5z)N@M(X;i4f93}`
zdZ2Y9eR<(%4*qm;vC%rvE8~27?Zqx1&&KEQSoVSfj7N=MaJ??Q@NP|4%17|QVJ=5)
z<6LULjl4L{Z=31IH-7J(&2IY4!V}0^B)X$PwXfn|&>OeF&oZt-8@38tUAF<w90q2D
zrydQhY$QkUykfLjU#2r2s0rw6H_-dhb5-~Kv+{Jn$Fn`V>pa!-@smfc4Qck%6M25t
zKeKW?pYh!IG~a)q?{yrv^T9il`=y;^_>JA^Y5tJ)8>ciYdtm;u&!F6ZKjt<(IDJGT
z+-~mQy{3=yD;xb|9+F#|e6X8$J~0ox9!%19W8Xs!{wZmD*3R~Jx`|uvbYQ2(FX%72
zJu${n$&vVlEe@r3THn!h>)SO&s5iV|-T5Sco6)VN@9lQzb|}BKp|2Hv;SS&z3w`H|
zbBy4@`OK~d1`nmLPWZ6Fvx0YQ9o&^R65l;^2b=bfPsjVsCy=+6RQJx%ic_IMA79_`
zq{NUHJMagEJMO|A3EQ_sb%3{6l1ab)<1K&hDZg!YHll%FJN(<#bXNQD1GI|*Ih04?
ztI<6DtFC`({g>aRFK_sxZrtVWFUfaqgFYegdZ+VEFT7LC+_e6n>?hCm!EGN~Pd_{^
zP1dt`$$umHqb55(rK^G;wX@SZOy6FoHNW}~oPgg44?fW3qn6jt(KQPOd&{%=Oy_56
zz#e|#_nm;>775<+csxIyPxHLRbtP_4zP7RK+Wup#4fM?Yk*_~<xI5yh{;uy2)6s2z
z=1|c#PLJ(WkNOJ!UbpLBm#iG?I}i4<-E-u<(;8F*^xOog>b;Ml8Q*)hRrTDqP!+;$
zv<K^p<qzEda9TcIx{1$!%07QDT8EOB<UKcTv>d_r^s6p%_ikKk*~&gS;+wtMuKa6(
z+90xsbiX%FW_so~pY8YzxUIY`G-tTJ&k45p3y8<3#LvdQQ+s4K=CR{e@(2h^i7dCZ
z)iWHI8~;ZG&EVB^`mEgW1lze$2_$fisndi3Ux#4>SG+v{W3I>1{x9w9$sE_m^BnGp
zx;!?0sJP=XYKHEk?%@P~n}BcsjZIyj)-`*>{%OgZ9;}Q0-?z6q)ZIH(tNyTzH@Pm~
zHTFk4+vd>vU7P;4zHhnB@4aPDj6}Ed+m7&kjb$5u)n?vzIG=N(1HXE`q4Q+_2HUT(
zsKY<g^UkO9-~)WeE{z%UC-A3+hkN_{HsbWoqwDDo?)bIQO}6|ar<*mmWak@^im9$*
zl<~3Onj?wa#)CwDFvRWFHa^(*8?5!O4Ek^X_`lBn%Ai{xqg(%s;Fi-P(OD;b8nfx8
zK_`3u=pTk``rP?#@vB|D-h;P|_93G_XwR+fDQ<P&wW4{Sp66A=SL>f1d@qzy#(_p~
zcX)uj$K^qjU;3Zz1Rg(Dj}>QUmi=<P{s=P|P1qoVkk@+vr6<ID-A+3HP%StR0Ds>m
z=7DDPbjov=^NPnaeZuo`*HZ;eStqTT`vKVbKRr0z->LpjKc7Ehet1(L!P_#{)^V?w
z<XG{gAR6x$o<u49()oXEGk747cwsB*`3dV6raCKqe)#7rhY8<mHu%2;wc%QN+dAqZ
zx>tOvudZJ0LFGBS;s*K8gx<P|!EAtj=0UVSCB~3o2O~U!VfB&danGvYtU_rnNF+@P
z59O)y>9EZYCxFD$onI@Q=9#e9y0q$#H;{g<O-kobr%dF{xwV3qR^3uE9H83gp`gki
zWzDXokE+l3kDBkJNV^A3AJED5hVq6a8RGC=K3j*67=Of6M&GRk$@q@sZD(c1ul;xB
z5mKAxEMkoUuns$7%%|3`@A#gbyf*i1)$iE>x$qn^3?2zF2shufQm^M;7fntLt5wZT
zz5Z6cor<k>pN87%>;?z}U?ww{WB0rhJY9Il{0lUnn*Rykqmm=>I>%FeIr^81i_@dz
z!<|O^5qvag;gJKvOYt9=<bCD1ZN(^CD?Ou20|)>kd2RMwhFZd&rv*R{;I@f-s#f%y
z{;4N`Zg<5kv$gJaMp=WXQ@l3(K!yhTK&!V}{;6}rFT1y%&ac@IJArS1T)%yQ$2a5;
z81eBYt`+yUI4VbwJ65J<*CTx?`}x5>+5wtF;x9YhFWt`fHGALfJa;;;xPMNc^gjE8
z%||D>I*@N`d^J?&e$0SBW{`Y3KQl&0b%Ix|WwyRy-`MJmfoe#c4*xcz)5cHh?@tVr
z0DgKPSomd)-~{)G!#NM-_ht~)2dz)W_$sY$*MsY0Xfhx3UAP;*D?o$2%l|DtZyVB$
zJb24|ub9Xrd7g+PsvQtsRlk{>06s0u)U(@l8Od?2yU3m@y6{WXU_C&o0Ef@kh2!fE
z;dLkCewR%z$vZ7Z4RQrIfQSFth?L_LUkm==(Z2Q|KHBZePWj6kf#X8`7mlcKdaMZT
zc~sm>5?%nlGklo*xM<te2K=~@f8;=a-?9DR9gU-n;E#M(-Za#m4XdU|OL8&GU7cGT
z7kH%vs2<^uer%vRfCCtILHf3<-}ECV@O%0nIo3bYRGtO8a<ndj{E?ga9s18Yfxq-o
zd-Jy5Z1;{w<aeZZoZ`DZe&Ahs_ZENs;sma4_Z1foKRB8;(6OF><c}`a#MU2tW`D$J
zc8<2bYv|EW!JXTYcg^tupFe8Y5A6Tf^3Oeiw?DgYH~blA`<HI(XB_T#OaX2vfQJe1
zD~IB*glte-k9t2<bkyNn#eezszV>YSFyo?s+<)H@ufJDjZ74PP{fD934UW@}XcPSc
z>jdbp4Em4%{5ScFkpBI@tiL(vzn}fxLFIrWiHMH5X64`$l|q&5NYu{Uj7F73cXg(p
z*Ud$Nz$LQW*4fR_r81$dMg<ed7TqW^iLJ$B1dC!$uIsK@cf2VhJ&-OKFc6pSV^s!o
z>;v!JfJ}u0t57f&+1dwp92052H-^#p;>cRRUQ;WbmB7<xLg~Z}*Z|!$Q<)mLj|a^L
z;Uc@~a$(XPa{E-4!hS}D^74%4csyXLDk8U;2vvdt0G~AA?yw7NkcWO80~-Jzjvp=&
z*+_!y_~J)U>Vo1ZQnhVBN+*4yZ7g09UMKq(Pm?}{%gQoatxL}<9>&SSBydHZu;7vT
zJ;Pm9kd&mb&IAdYPz80Vr@OdLbG3rhFol%b3&4xsm<=!w)e<inzN%ao)3%ggg7RC_
z8)haU7mmf9uM!HAt^iI74YSZ=50)s$dpXfaf$+g)7=|ITC^noZQV5c70DC0330f-Z
z;-d;#y1eIJ|Dgc&D|kvMOevTu6cLfMW>I{Kf8J;x7R?yDxEEde?BtHbXd{PDISQ{Q
z1-7;5S~RmWdNCF(fDeETIakgo)gZ?46-exew4&4@4}?pM0lJ1TBBChkZ!IyiT43ai
zS`Dd|!koFNR@BT~xx$m7EOE_2(}pJf9zyLv4A5?P8wod#=(659W-A8jF$(0$q2uC`
zXb=pCEGj9pikca^t(%O1_2<I1NMz$|m2i%x+&U0Hf(=$pNF%Bigfpj?ypXbYb3Kzy
zut=_VuBS~Ox)e1BW}>UGnsYK-M6@bB&?xfErRFtU@?23#$`HHbJ2KakPPB6cY)N+b
zHJA+$)ob&C9UyLqa4?O^b5)LqDnth!k?&ye=}@l9anT{7;{5MG`0NNTTuHKZB2tl{
zi8H$bWgm)}q>HK6Kr771SQ4NWhDDA-=TQ_mbQL6d3Q+}+1RPPa=wQujc2Sn7;}A4S
zu!aeln>0Z#Yn0z~Xe0&e%0c-@7gW>zb|e%L)ouu{D$9@>;1#YJS?-86(zOY4nqz3y
zMBcFD)NaDAa9ELzay<2DJoU>jOm`;oh_cg)5%De@HR%35_b~whVAKxcsL!cC^6?|y
z4R=ILDxR|(d&9alk!(T$`*h*P(N>@No!pVEFL9PqZ0`t)93|PNmKG+S?*?n<(f&ZX
z)$g^NB!Dc=alNCVhj>Vejl(k<hvPyGW<8<tXCq4F?P?9`vJLHWZ6-@brHPG~nN+gF
z3n5T`B1*)1RqB!~GTUM6-f`g4Whm$Jb-A!H30nZ&1Sc$oXppDy<*U}EL;M-U?p$?e
zvDc<j4?Vcr^v2aH?`g2D^UyB0Ed`)a)HC3z2NfQCR9QYWTN?qYsftE$ai7A;cfEmq
zgC%XC4sHuQ8S>whF#^Sa05F`;+V)+jkJ;IpHtLXA<kVtXEsjPMO{ilY)EdbUS@4pK
z&A<%YqgJOO24X#^hT7V#9rqwC2PjAoO*IoH<(Rq0lZrfmPlj_W=2je8ie@=#1o0rm
z&}}{7o+)cOxHn8Bv$UaOJ+dLs;OM)Mvw%3aRMvD=A|bV&2^Hn>2dGZsVxTNzyO;x5
zF%vVIPo>(Y<H}(ZX&UAh>1t^!12Ak<^)(A?gomO=v%aC4HZ&U{3gJ<EsX(OQ`&Dk`
zj)xCL_KqRwkJIvt7XI?3&j(|yX9K=)HC0NMJTyf3c|rD0t|=r%OSx#mlCaV_Lve1D
zHQcRSI@|J=uF@Sbzsl6iG?xY}T6M4#Gv_6+VkdSawsXa$%{Ey}UAW_Q%L5;ySY#sp
zTyVW0hKsIbB_ZAbw7S5Wmer~Zj65RaWc@T==9OkAEvvR>wvoY=p;;YAZ1rnR(iA3v
zgd4X3S*Ik><envF3skf#Ne<PG!MY}tZCp8%728!$&DK)>+(fw*xOWP)EFl-_QQQxq
zi+X<He@)dH-<7rTB9-qFZxe(=8ej(9+g0dA9+>S07~$(<g-chksW%|2YrIQ$WN)Az
zUD|V&jo@Mso~fp3sZNm#0`e#orLeSjoB~hWfdc>|x%*sLx~6tIm;ntnPSwOZ*yVh6
zux1b>APm;k(x`h79Tz<CgEVe5G8NQ}mQ0`+4nf3KQRa$?5RJB!S4&EnB#8Pjj_k_`
z<Z-tG1Z=pvWZYs~;ZZCBX<X+)9_=E=`ZETqA(4jKG!w94YQT6VDPC_~Se5D2SA-Ri
z3>BQ>a$~eg7M#Q&MH4VZN=2A}tlVG*Hk3;`YPqf1HyU>|k5;)i`H+oS+3J9FbePkc
z)tM-p`G&SsU1l`3FoFFM&hjL&@PaZ47!_?{#K$%xxVk#-fsgX$;E4vM1SUa5WU~vB
z@D!RR9Rh@gvJDa-!6{tEHEU@jWsVN^e4n-}JeR;zAIA|6JR|3lc&iy@8c`EIVW_3|
zP6Dg9)igbS7Pq>-B7fB(8sH;bz(^uW$I!;hno1#N#QlHN^Z&eOBiF5fWBa<RZH`8B
z&*t$iTx}2R4x}-26iv~CA&8E$FvajBpB-kDorzj+Rv;O|Wk|#V9Kge7T)6^|xo6E}
zj7#EO1?IC>3hhmiQE3_pg;-bq?J^Hq0ss4zhEb~j-eqIRXYz`VZj?+qWd~}2usE7n
zG?!nYovV4uQfS>`MI^$#5(ox(fY{{#U$=*!87^57<twH%TomvL_aL;D12}+F%W9K{
zSqvx#Fw|$F*Y@1z`7~~QV%<7!K;DPdv1m{#KQl8r*^3R*+YlG)R%(E)skTBn%W*28
zaHeL|N)wyH>edJ=1^v(*a6tFN3Z9^k(evyBX2T_jxsQ)Z0xMCom>J=%{Bgml(WoE6
zzLl6IhTIExMC<5F*9)P_6;d_m1}II&p-7+*5^2EMvZEITDn-2j4&ZhwF$U6yf}&A1
zf?79n4ocu@?KO1k%ATVUCtYvG`6kRjj<tG#iqFK~#fp*kgBTH$aju&Ms0!he<D@NE
zNp8(L4|}o8twju)s1ydz5oVh3TCFzSt>@SzDJe{8uHo-M2I?i@5^8<3;nXi4a5s@U
zMgf3~%4;2mGpRlytR<{8ftu<7NM3Qt@^u@n7BD?Py$ne@Uh|r{CxEx-!3vPxZXW5H
zrlssuu=ND-|KPD6WnD@adq9ekSjh-xaTC<EX6Bj86UJe#$niYgl~Z@toJ?;SZ}p{W
zV-J90l#&P|65F0LI)cPNLd^Pglc(>Zs@;@IBfL$5WC(BNX_i!x!X)7Eg6h(8XjVt7
z#Z-{v-QYIzjLn0ZC+R4Q=yIb~7CSC@N!qF9N2}?cX*^i)P>y>LvtIV8GS8aLm^KhT
zEnR0#cAStMRed8`YNoVKlwoDHq*oZ;R!ip&pfMVVaCwDopTiWXVG=Xj^C&6NO_|gj
ziA$RgCr=CpOEQsT--7;uV&@W)La&R!SX<L=3{j0jT5$Y_n<VSCoOdwkJ%k!WoZ@if
z$jC>N0Ldy|v;0|V3>;^5GihJjM6PSNl{MIn5x?OE*eZsRMO0gT7hb1`b0j>?l)}bb
z85mcqYo?kJOszWK^SPCw`-<xGrdt&%8F2$`ZC>SYGk2xG;<CrZ9*V3cz#YK2M=2;t
z>4%t4<7n``?~dpOh*<B9-+-hqd*2eP>}wYq-|z&daEWRbRWm-Lqm4>o1?*hhrt%4k
z!E<!2MC>U(AL7LicF+L8R|l0ht1)n?aAXlRFGz^)0o}}AOk1;-k~u?yudDoi)Kui6
zLNQxA>A@pw=FimmDe9Fek1!%0GDb>73WIO*AXtq`LKcy0J-B{YL#7@ZVBU>r>8Q$@
z9JgnKY74yqiA|yP<W+H6vLxb&N}c@OLF=y!`tSev$Jt*Q)NF7HJL8gH=2Itd_hww_
zuDDOl{cXOy*?+1Xf7?<2C5_+qu)NI(9PFq4LEB&W_XPe)Pk=sg7ymQt-l>3(ZH9Av
z&-4STd59T}_K)~b{($J-u%xng{p0x$*uq;!xOX0JZTYO(A2jN-Tl~xGzTyskMUQ@)
zzN2TK;Ya28Krr*M4SYk(kL~53*tp%smi`kT%((TJj^{@44aqpy3*a}jd|}!bN_C^W
z{P~-HaR*ZV@>#q%#S3?Hm40jU*GB(}>iu}~e;0W^!R~kN@kdnfM$>Ln_(#nDT|<7s
zIekHkr%vx;t_$j^R&ObK`JcVLWs3VIKk^XokN<}$|KQW+A93<?uI{2{@3<lA_%>rU
z0&W2Q<e!0H%a_jXGp>E4sY}9A*eWs8gnwV^e>nUBXK*<Jk(YwgdqY1A(_a~+<!@~P
zN7}pWh!<Z_uKBLu>t<UIDmAnBcWw9sef10<>`YZ3=znf`HjAkFdRBNo*lQI(d%D;5
z_FMgX0^kIGAA4{x&%t(_wm#?bv8o?8mJIZjW4*Q4Tg`6Vb?cW`iY=F5We(gh9VIUt
znCm^Gs;Zee{jPud3H;vIKm88=komvmLHd@aZ$0pj?%<>5PaMNMkD@dAM3<ja@|9cq
z%3eR`gnuV_{|UU~NqUFJ>pSBqbQD=|vUOAGhAZ%ZMrAwxHlyA8<H#$szf$}0jeSi6
zbOZ?hnyG$jm)}!)d}n^o=<yl-sfyow`={Qy-?@KJ;P(W+^aT2sWzWBKnBQjKaRPkH
z|2gjaKgTZHLnNE_|AyNC%ef}IO26XV?)Pvd1(xW_OI%DqNLCD+ReEOt0pPv<-N@^W
z#o>m}4avKA_rv|Z`9vPKf2{c7C;U6!?+N^#z*|oMekVZUsQpVWFBKn+`4REY8~^7`
z^pS&arfJtH^gr}9^&gs+o&ESd{HST)KFnx1f<OL7{SgDcaI{yYSKJ?QF26(m0VnXc
zJ!R8=ChQNmHIGL8OSkoC+rRwp3A}s)`)6lIUT*Sh@Kq;pb@Q&c$Md~b{n-Cy#W<jV
zBlU8_YVZ|3YRBK~&Ix?{zJPyO_wl}WHvhdEKLQWzU+}$ouJC`e@t52E@Bi<A|G(?6
z4Eo1E{>KY{Wl$VPM6{EdmAze1MYkn$;f!v@d7`@|yGeEtR=toV!mIT?RjtX{UY`+;
z4X`zrKaN2afYoniH*^E=I0sT}l9<sigE_va12V7J_zf^AA`otMaDBQ^9eaJxt#QV4
zY&`r*n8i|7Ims#pri`u;jfFwK9`y!2*A4)Fy+1tS{aeeLp;-`abuX{5!A#8PngSAM
z69O9mAGU=qKI5nuK`CieOWpQn@M5`lIrh0OxMsa;3!bTZL(OinAqs4G+W<~M<qXg0
zGKU0i+?dA>lVZY}aY~W;6o>moMrj5KvoOb-5CE$Nl_4BBP4H|nqi_%;KOG$UOq2g&
z-*^Jwbf><d?HgLo7Bjl;DIyS#E9?Q8D`(VdrZrdOc+w6SDna;Q0if3%fblQ<djjVt
z03OE)4lJVD5a0lQ{AEzx;kQkb-B00Q_bvVq<v-6(|BxH<<e2b2LgL9rpbWKXcuU!A
z!a$c3_(QfQBYdM)cmMDZX2FcUxSic}M;^5}t2k5Jj}fl&|J%Sv6J@^jYw3kjsa~Ly
z<U)_T)(C{dAF8<^$IGy#17KZc-Sm|nGgyO&cj5c)=1rh}sD9k#0+CO}Yc_cy1;X89
z7YKrIFZ5tnDg_7tyI%y8a>C##vejO(Bhj-(2hvtqQfvd59o`Y`uoF8{hDphNWh$bu
z6qY0j5grm&s4UbGg_5FE+^SKX@xSZ4P5_)VfZx^i*XQ>Heox@`3H*K<Zy(hroX&2-
zq|g};sv&@jh%`W#;T%W(jNWq$7dQYWYU7k`VFGf;u<nAMTh$?s<YX6||97PUaX(9K
zwIOUpc(OLBte{|k2S}|DgK#hmk|R9ZK#lDd3IGA%+|#Uit@@gU5aFS?6U{C&s(v#$
zqj`jD1z-W3eM=nrh>Bn(Gc%*T_Vf8L#{k(JBQ=6><LE#zF3o$FmbdhxeQ&f^>TV%1
z52RJU%bQQ&!EJcd?Xnz_#{?=k#fR;jSX6Dz%z8%q$HLSjo-9%rC5MrCJaEc%j=9$c
z*aRanf`bp%fp2vLkB3MNFFM!`US2_u3fJuAo^LS76QwwX(PRXPhD6l<ebK(@9f2n^
zNcR)EC8}kp1mSQ%ww<eHw9fJI3=IKr0N)(*wd$KqvdkLlYH61IjQCxXy~}4j6=*&Z
zBRKlbAgV_=`uYTV03(+HM-Hd>y#0kvMiT}NehP7GXe2`TPuU!pL&pBX;knh^Z${?_
zjJ5;x0EYTgpX{GwaKgB+70f0LZK>AIs2Q)d3wnMbO_YbI`~XK%?~wx+OW<kN9K;v(
z7t}=qP?h0`BN9i;ZCz*zqw5F;J(rypHGNHMR$AXlP&vY%_q=etptgRqc1Gi-1s_qh
zeivR{BP;$TVa<+6DZoc~1Y-h<kObklVzvq&i)>BZGZf>^=*0k0`l0Axq629RI~8r%
z21NaQ3-E~zRs*#BXLKWw4B=_=Ap}(b9%e+!p_f5VlvwG}I`kPJv8;q=9<;`v##=i}
zhqeQxYRs73T4&LWN?GT6p3I*JtD9yS%yIG`-~^=*Z5iX%rI;AF)`cTVRftA-GS4BF
z7p5dp58#X8qaq^l>$!FLG;V!j-8ybSXd_|~R?f_fhUu%3`193Ut)Ku(IKTDGOpZ6<
z1Y8$fv*L#|r}(2!)anK-8SdQ5I~4n$nHi-j#N5Y`vRNY;!dv;H9parJqp|S_-tpcG
zcSQG4h?2x6w`kA}5SJx9g`4980F@#f9kaDTbpQu2?1J>xURGC0FkV%YIeP#K<<9<}
zZ#klY$^-5|&q;X~3)jCTns)w0dmCC8s3bH~01?$$--V}ErEl$)C(6v}6xI#ZW$&LA
zw~1Tov!3Fs5S3uUy1_~l+*`2?U<P=AdUGV{c+bak*hJdx6z@ZRBu7I<Go*0b1F30g
z7}e;etQL*gGfr9~%WZ?(!g0<7BBQ_~lCqvNI)YDpYJpE0_)&-*3R9BgiSY6BmFhTz
z#TO^hJt!Z125720#h$T!rsj06(7)F0wd$DS9TAuXz_4S}w^YhH*mE;K!;#wDY`xG;
zlwr1-&A8qYjg`-Usap!4ar)D0F1I_71N3Oiumlp19Dn&g;{?(TPNB@c0lp>7xa2pu
zaRT`qQ=GyZCfAbDYXr#$5CBH<Z6Z{EVSq|Td<3^C&VWyNo=xzL>eJ1i>i4Z72ZX~%
z2Nvu-uMuQ~@N2hMSm}uSVg`Pq&&RoHa_n|$cB)4Mo%y)h1LVv2J=LYf=gBr?OR0ZA
z`cLu$c;k2<F+H2_Y=krQ*(CYYKiubdu%BB%e`U~r{O7;@>%Y)z{oO(T;~)R??C%cZ
z!gN{BzP)7EbT14+s|scveVX`FL`%E~qf9|c>0`I@$Nj>^B6YEpSMm`*4m9Td(4O?B
z_Iz?Gwlp!>(!hh5ooj()M9?o0XjX(OQ1?P<!u{A`Vnejt*K3@Q3EBa_Yy)eRiFi7>
zh$3HUKq<32O|wX2QSeo^ZPGOr4dxe>3>F7ao=5ysaUltJKnLMr)=b~>incL9VADwZ
z26Z1yq*{XGwPF(JjpnC}@5~s9HkXIdMRcQphvc5w=SE?cd(1oXOm6LSUe8t4hl%%=
zMq*Do@-*QJ4Hh)%U8V(0$wi(~zIh6`_-0TZhMP4&*7mbhJEEnK0M?&|pO{BV;<K`L
zTU|`hW~cj9g-t+d77X&Bp;<@<dkj5Y#?Lcd_HhZ4m#`GC9rA@%+bvg<#NH;<oKD!e
zDJC*+tt51+QclbMnlM(8S_;J>N@YI2iZZCDA~?cRZm{U64egu`uD+1>ku@?=oV}#I
zJ|zp0`=2VMBQps-3B@5-?1-~Q>7z!^#w)LNs$;pH_Q-t|NVBc!_|kiVvAb-@X}|pT
z2Wk)z6fL$RG7nIhVLM>&#z9tmFpe!e+hp518POK_5P5M4>L%J}C%94P6R=h?ZmNT>
z!RX^ci(dGlqIiBaoXe#oJr#2)p=R7{N&!#8^S~*J{xZoHx<j3T+UTn}7<{?<U6x`x
zK%3bPv)O)CI&Ug<Plvz4lqHRa`2ZKl=1NNC<tcU2ZF`_oB)vYErN5qRFP3|;qlT!f
zrLrzC-!sy9uG@?{=2RjvNnEeOKx(UJv*ov5a*Z{)d0!87EXj2~!`O)!cC}BfoinM3
zlf|w%<ylDb%<Gvd(CQhhXGWG9@5X^<(_}8(dTHfaLKBh7vt^FMZ7T_Ck2bO0QM!Q~
z?=|NL8EsYm48?YrNcx7D3#W3G5Ks0rE<q9x?ZR@1_UQ&uf<$k<URaao%OMSQy>U7t
z8n(4!xM8Ghzev%sl6PMz=*Cz1vYSkX5>oO`IN<xv!>N>A_MHw!xd79_oR~({WmLB8
z2T;(<`ZH2|R~tFMt(XRQruSJ5s~?s4S)?Jgh8A4DHb=koMS(a7klg;p@WCb@(DzQH
zJtXYu=PeyRalJE~-RnEr`5B_gA5=f$ACx<deC+UC*aB47WMewj(#J)A`<WO<cz-1*
zg-cblUgo*TTx@}@t{VfS;(><>jEflg4OUv9w}&z=zIG6`y4`uU*5kK+@r%E^JfdPs
z*lTLa0xDiIZAdd(Nk9@2F}<ip6}+fMAVD%svA3hZ6H(3AJYiF`nHu{r?d^MRhhw;{
z4~-uQ!pNhGT|RPKJRigj-CYSi8TqAo1xY}Sv+yL&g*i5lt4&os&A{rXhk0CBJKBu0
zKCZ5M&iK=}&N=-8Jpep-UP1e<dNB5T+Fp(HF76$Z0uP1LP0BZH+W>GeU8YwIo^ZY|
z4{nY)W<TrkkyWSBV_0*q+27FUfd?HC{i1$r<xo#;YR_(bIU3C>Z1v&|aON;jBiYB@
zTHf?svFhG6@pQUn6cf;vKvSk5jjBqQClYx=<*~6V7;tvy$8Dwph1&|r=ORAX=k3qR
z+u5}gPH|TU4Ze-w$>}NDUgJ&Y(_EKLeE?Xot*aC`%F}&(O~wafJum(8ey^RsZ1jtU
zwg(2>;pY$idn@PDBXnd&@C$eF;DJFj!olQUz02N@a>Y7j_&&KSIiSmW^U@w!=J)!C
z$gR5Zq;FMT_TL?LL=9rMRbrE^m5A*}MFnc@<!p!)i)Cw_O(?l_=&Y%DO1UtkcFT=l
z3o{3=coZBLxFH>m{UWeAMY?A5q9w~o!rv6K``p4i+hL6@qt-S@>w0A(S9Y9o>Yy5K
z=7~~5FJmLJA;lyRl2=K^3YByV&Q(0s*tRB4yi}2lmsW<LJI^EXcoG+k*d;E%dd?2X
z-MZRJt*C}x?j)xolFMrqii?<;sd|PmB6HR<OJm)o!Qsd{Ck1ViUJ2qeUKd^YbTj>8
zOvbwfjo>YLB9`^h_QD^EtiAXvg^sPDz$KxRc=(bNlQYv>b$MEGsQfhZZM`fQCTL@c
zr5X?Nw#TF>Mk|p){K?41cg|ut7su4;UE$z=b26HMb~gT6JKZK6gnA$+yA)q_-<xb*
z1-LD|CqJ~42Vp3({rlAE*oeC~4XI4)PAMQG?!>BPD!v2?t`R465kgZoVo)hAVsdDz
zj`CDc)$m%Hh0?YB8nxD)=cfx*Zps9taA^Gu<gF!mY$5yOO+FB_V`wOi%%Tb|K?O`!
z57W`l1~ker@<G#~dVqM-<Kb8<DjEw3)?eV@%!1aih$!<}QSs|+wH{u}ZM|qbdFiUC
zg0YxvA=ll$p0$B&<q1|oqxx@81?GJZqkhfq(xlWdr83e643pl5)J&g4zK?V)Kj5{7
zXmixGW_e(FF6AcbGb;C0?st}gID5;3!lh(3R}3&~r7U&lPgKwSV>a6is7!&TAr=-P
zP9_uMxK<l}OlVIbkZ6F6=u*p<N1Tor8};-$97uw6YIqZGx25M`O>BO*>qep3@I&DW
zLfu~iy>f5YJq)-?omd>hXY`@ZErIW4SFdL8@3*%}$Z@-^AOUp4=NCV^>E?+}(Y0s!
zMGreiVKYVEB1nfQ0C3Ul$>Afj8@vyZ4e(jedB~gQqXf1gnqF8TrEU^|cnfU~<o2MW
z`FSA+RA+%Qk47p6wh!)Jm`Dc00Sf(5w9e&7zlpEQui4E^fQ$Q{E0#Cg&pj%t89V!#
zXETm#O`~s4)vB)2+RidsS1)>C`NatV`o41f7)Q-te?95pWfHt?6!+DgZMoinvhlZc
zFOEIQW5@*BladgpFu?DOY@j^k-FakZ=8&oe$l2pYS7z>7g7b<}PM_V^UQ{m+!D~Lb
z`gA7x?&11-yScW%=aI*Ke=G;8*K9wJKBH0B`JRXJC}Nv?Ui9Oly*7Y&2R&$2*D0MC
zE3`bjpx7KUZ<wrhH~1-A4X6qIPQM6~*4U<+){Lu4sU!vB5^lsZ>|FI+8IS;YD?bI^
z<ohNYo4(n3swXKjxp`LdiPigl!klh(Z;<m(Y%lx!o<`8wJHfsY-dX;%aV)p*@pbDt
zk2s@S@dMz*MehkOCv|CrD2qtx#|N_%$x6Zj+BB&Fe+^P63eXEPs3LH7HuuXOHifQI
zNR<qPBZ`u_4D5i<?*ToKoCGr1uEyNn4v7+t8DEL5%qZs~<w;}j6<)O@dcB5O3dD**
zCDlXy02RZBiEn_mO#9?FU2PT{$5yk{oEOtBPV2V8Ho$ydT&3T+N#D+egGp^|Qa&oa
zG=fqgc`==w5gqsN(qmfMASEtPkh2Z*$s*}bLZ6~o(ME2)N-MS~K16ST)F61ZEv8Q;
z>aja5g7+nNZSc14d+8u%=)Mv;JiujO6<If1QSBoh)s6TvlXKI%DUZe|MYmhUZyjqM
z0HOH7decxZ+0cUte({FLoZoaK@)k=XQo21^@@z~4>B|EHj@H+zhgNL3F7%sEPBhUc
zB0AKR6|>=-NXm{<Mn~Yk7RNEIlP6=Smo`Q;Z-W|pz##$LOktvtMc&#SP1YaBpKXwL
z%ket+mtL!!4~}j4CN4W#9~xgaRtGVRD==LX@DOPybTbL7cbT9LnvS&mmLh?<K&qJ4
z&=;d<P2D{lL1$aAJj8`&=U*g{h>a|y_1;S>cv%-RHuDwOFKS}1%o8Sz)l!lE=3<ID
zw{ly3F9!VFl}Ej})XC$dD!yk)Il1ozKNP9M-x_jlU8?B$*`AMdFyS_WLFb4b(aTrx
zE8;=)dRg@&$cKBoo{~R|bm0x#g;t66<7^*mHRC(vnh3at&+pmGa)1MOrsutoceOxX
zX-J9EwO2^av&#|80`ipRd=Qkh{Yo|U)VJ{>WZOsMr`|HV-;2aFj6AuObcp^7rQbxF
zLUb+>h3zD_sIj!D*o}<a?mBXlmyRP7E{G>J@{MbBJTawdd?GnRKYE&+y>!XGMDY7x
z|AqeUp!HV<{m*~=ud}~0i0u7EL&z&4%4?>QNm!`MI^fwm_1qdHJcWy`naOqSWY4X7
z!$EzsKT^%Lgznul!m$CiVzY?>RRC6pncdJ0m<A9GOA&+c3%^{F<1|)>d!D2mZu|yV
z1tJh`b#Q&UP#t@H&#iIBb8P(dkTN%mrL1z2RZQ(iFj_xs)Y>6Cgtu~S4AcPDBDmZs
zLmGq+UmWvR;8f2NhZ7yjkEa6u0EW-<+!wmId{m5}lr*ZPb-YT}Y;?PxirC2n!|t~Q
zz(@TNyV(@rnX1bgKsLCgYorZPV8d;ENSGNA%<znwmGz7j>O*4DW9g)5SWC^@Rz#oT
z!+nqz$jAopH48oVV99dqRS_&2DG(mTJ=!WZfGcJc4ua&TgLj{v_uYI|Ugjvtoe5CE
zyYSOfhI&kFcOK1t=R9s}_9i<LW`F}Q)m=~zZ9`xYcAJ?Qy%-A?Jlx1!Iit3nY0VWm
zp0opoN)T=q7ratU(Ngnycupn+^*g*HY5@sr@o?Q86xoU71xY{vI5|EU0%M0g=>CYP
zc+I+p=SY_D6nYs22rnWa1GN(zSVXl2zyUl>1f^h~m-;%KXHgT8lY0O~naJ^OVaPp|
zscqe>c7(JuDlUKnuyqFOP&GhQ4B%&oc^1$Bz*i%Ik`bQ3jODn@jR?g7&9HDOEXLrZ
zf#JpZSJ#q}88U2mOW9_^+<zw-*<q<jcnUYp?Q^gdWHQCqx`7pFAlwvi0LG{7hIT))
zLylcIf>!OS!$3^wn(cEehC5=75R>s7TU}j|qT`z9W&I1H>UAe;B`wVkGu2VfJE%4Q
zz8D)+WH`sGPfG_N@980(#NixKqA5hgyKvN?`}f?(goG<ffaDdgSuxNADG=@+yFd_x
zd!Yxr@_MF0KS1O$!4ozLy?n3O5#LM?4y3&_=hy}?JG>*>U?+AY&ywRuO2`+YB*`A(
zAz_8eLLE^kDLTciy4KaqG>7KdME^4lb;^vJ<JqO}P&7bf2Vg`%c^d5^JS_+M2~JoF
zFJ^g#Y8@=KwOSi$tFs#*41k%;T#nt-gPnE6gU!;bS5Q^KoZ@QK%W7UTZ8m~$oUHq6
z-r+J0Amey0B3KIu0OMPe&n6Y8uuO;W!&^VuIOaj=tO+6uUJ^7~6{TM11h+tCj(ZTj
z1}I1nO*IoH<<59eZ2??Fqyf4N=Q!$T^qyn5zyUB(8>egw6OcQGbr<y9st$1^C%fSM
zXL~^0&r(}$2wM@JtW7E_C>Y=YQY*wD91Mfx2+uZ9W4na{Kma)RG;3a~zGfjrcqnQ#
z>rYViJ1YrwM0GI901M#Bfw9m>)WtJ1qrLX?`7p--*&HJ^f~(?PxHRu>EpO>X`##uS
zsk?>5JWy!~YayY>wevdy=er>-l9Z%S?RnYZ9ht^~s%(BQE><PU_2&dtpI4!Zcj3cG
zC)QP4Gdr#s#fLHI6O=-0;OiCVq7{fv)u<@rIcB2`F#C}h!NCXXhTAmM+YoC7Rsc?g
zzv1r*9G<}bQ0)kMAR%6682p$=_#zR7a9D8B=+q185pFXP|474Q<_jp;;0-X7iI_0c
zzqSHrgLSp^ey{V2``%f3*aylHf$%&7$5%qgjl>9U{qUoS>Jg5dfF8ieWx$cc30}39
z=}L=1c5pp69IaS?#y~YB(nOnP1D1Hg$OGY+NS%=gnV#a}uT`?(B+muoG(qv}1xY6q
z0~^ZRjN+pOdH}!8hy1OFtqgeLblS3n<2C9s-O!e*az>L)7xer>nkWxZ`2mik-XjMt
zmcY}jIfyUnFQ|(Kpen-=M<kAx+q%#cM%NJxdM-OHYWkYiOj~!b=P@B-5FV8{nrmm2
zT`s7t%dDNzxM{&hRIT5ISJ%jje@R%gBT@?R5gx&qfFdM8IIftj!p9<8Q}+zTcr!W<
za3E#>Q@sv#P<9}_-tlI%0Z~8Sg7L)AG0g5Wx;!dTvzQCvp_n321>iOc8M@K~YqJ55
z2v}f+eSJ_c&;ul<mGI1i*81JJ`>jCY{klfTpiU%E8;?`WB=szsQ7P+O&y)E@h^i{C
zSq5{Q{0BHeDYQdOwHg-DPREnpxK;O{6wwH84J^g-!j!~nFJ=V#T+t;xw=SQ?txv35
z#|_B)usRkEO66x}M$Pi|Rs8wttyWL~C7fSfb6m5^lmR!2fepQ`DXb*O;S_JX8ag4x
zxRC{-2@rFtM?Mn{y(>|(I0J;kJ)#1HBcoA2g43>b1FrYLceW$FhH|KK9|;EC0I@dV
z5+<O!LSo4fhj3V+Em0l70Svn!J=n0w$E!MJ&Q7CH?(F~hmLq7(TJO4^6LA-FlO9Y4
ztWhHKInF*A&vDk|7-}YB!pIEa-sqvgm8ZerT%eMWc>odBrtiX2tJ1f2%M)ehbPDT+
z>azF4_`(EH2!Ghi1~(xxI>nbCD#65~A&kGd!U50(L%{&`?nu({o{v*(BJFmHUmSm*
zf6GxX7G^TX<8Eef`FN!VGDS^mmgamP0i9B}9%_v&w+(I!d%K<^qrfAQ+MY8yg6Ynj
zx7CXKKi93GxAMHqSQ3{!F>yS77sZ!U>aO81+yL!$=g2d*sV0Tqd%~RbU{eJU5vgnD
z^_VCHB}v8K5rSC&3@2>*mP%O%dv4}uI8vLNtrxn9GOVxTR%VfPUaJWbL?M77F6F5z
zh1L3;xLIFi+dvcjJ`C=PS`6d>JsJj<WFp6Is6P~TE+HxOx(JN5b=*_PDGF)9;SZ>+
z8f!-|SK+2`8Z;%uq0^n|U~wm!SxQqVTi*ab5o2t1q6L5l@G2Gt@d$@UiVD69Z^u*S
zPBW#DDq3nL1W#1OYNoX#NE{#ljO5xS7F2&>fJ#Pu1f%`&zGyD!*@>R#zAy$w0KPd4
zs>Yi^74O2);Jg2h$k{2r(>qcM;K|2G?%+Wwno)d~VO(qu;mKbnLGw%%F#)b3O=W8q
z9voefkV6E*+dw`-)u$|lE?xnEA)ad^b{7wrLs`Yc=22Hzmn?pG_(yIbhK>lIGt$gB
zqcb(PaIFW|k4?zbZv(7PBU%bn2A_YFfIr*~(#HfS(!Vn3fBobC{{1V1&>3yVz8$v$
z8?KHBuMO|1YE91e`iyXFfUUXwaSW;etbQ}Qp&Nk5!BWH^TzZUb26LRo>Tu7Kl*5hR
z0INU*!mSRjPZz3VukX1v&UlWEhu?&=Sjs9VS;fGV(KVv6Fz6Mz!h=F5LBl2ORXx`t
z0FRalyi<lW2#+17Mijz=rh7Vt4Q66S-!~}m2k_zfKo>u~Q89v2(x{f!@hVxf(d~LF
zVkZ*}yWbW7AN5D<W>bJ?sxE5)+2EG0kv2qu4Y%<jVP-%u!!v4D)-zV94~a#OrIVsz
zEj4dj5q*je_d!}9BOAciEcDoeCCjl_MX+e3KzJ1QXsg%&u9#6c2$G);-hF!Bck@+w
znWH3kCO`%6!cR{b>M^n1c{KZ-^SG_qo9sxK0S>@acR@Y04S_}2ZDwZlVk}tja3gc&
zjM{dlHCN<#(he9ZLAY65@JcmBOU>uuIhhdD@9>VO1thG+!*zF1WG9jrBmn{7<oIL=
zj2-r%`y-;_HR~RpBU!>z=w%cjyoi7d)J||<5!Ds|2k<lzl!AR;>g#ZxMNLFb?g12K
zBFDRhA@@|Kwso)C5z@}6xBw2o))}lr)c{d3fS)1eSwI5-UyTGxMtA}<mg6!vA`}NS
z!@{Mo7=x1rh8O2wT}wu0$gtrpWt$0e|D9xHhovInDcm@>&%suZ$rNAf23DYfa8tkm
z7@xKq+Wp85Id<U)TD7YV12Ls*w$HH`?ua!)OvZC;b#+CGj%%Km-@h_w9c4eK9k_Si
zRYh4iN7Ph^co&{#D4n~1&s_`&SCjzBD_*l=pb1hS+&y-IAPDzD4|e7COoM)a$YX*h
zY!rI=Ua=#-nI0TSduh(G4PbV7N3_9C>`0y^$B&edFGNX_J;Fo63YCRAqEJ$Fid%K9
ztC?vI&9h1U&x14L=6H7LI}{BN*#Q_4P@YD+2v5s_eu5L0!i!m6p;`w^ZLQXZ+Uo2E
z2m@dyGnZrc^k8Ql@nEy`>J?N~FsHa0^|G4ROq-418z<}jns>NN1IRd@iwM>N0>Jpz
z<g-b|DJ;_={P5OKHja5vI%|T+f|mr1Rz<1TIl(PZnd2TruK@}YL{rU#Nx3r~R9gTS
z5ov%f!#R%n8NKHiE^q)$)W#{>!UW`wVci8ix2i)N$;mD_|Jfc8_p{Vi8^TtECu@_+
z3JL~zfYb^x2nWL;Il{9I)YxvJ01yDqJ<Xces;^lH5gv*f&H58m{mx249Z?-jGQa|O
za$qd<5q0s*%xJIud_K%EKsLunjo_+y7cR}aTgzK|(Y_D1SL$vdF%MK)!dgh^aqaw$
z!1-=SizFo}RC``_ct@sjpemc+i;GoBa{W0$)#p{H;$8SK(usA|*36D;M)6?``UIuW
z8u)s}xo8EVQ#C5ec#hd<1I&ITMsV=Ky5Tkr^)|#>ffayL;TVo(BY3xUFYY<^p#cXq
z!RU2Mo<%8^Fagx0rO_7>A=#tSVo0Wjj~hK!?2{zqpnR9^2t2Yunx@bYi4eZZ5QT79
zz$466Gm4udJ;H4!;vZ>v%zOa_8@vH#G7%GI`qx(AY_P7D-tTo@ao;;D5BoqlA`qTu
z;P^@ixse#btsj0gQ9Z(u6VL+~xePdRIKivdGF@pg$PTXOhNBhh&lsqNM4D*RY`_vv
z7<nKZ6R9&2A=6V_{IyCJoaDJ+oF*uKy&&m?Vqim=n^AnUKo8)z`H;W$u$2K%oK9Po
zaJ)ucrW@K)RnBO#>4KhLNE78DDnG!H)O+N>#S(a$H3#uU{RMT=090i-;)ul2a$6Ug
z!st4JLC<BUMNMDRnrZ6}_B<v;48o%lM|16rvdaavb(ys@8aFNYh^qCw@ah^_@h=H$
zc0@`6KEfjy6HtUC2*(w(RrpwBYwDh%7;i?$0S=_>f2!A^4$2Os*E`;fHX!QfTQHs&
zI)>SOMwdq=Y8G=LJQPy|ssP+ZAwySsU~M+w5djOVu&)ma26}+Rv=W|q&|1G6cfS=#
zykFM{8PtgcYU6Q=nWUaYGb&}B>v=N22vJqVHOpX*lm7rGD1~;2saC@x+Ua=G8@K8n
zlp-48t%0RjUYL?t?Zu2hpDVhg=ho%Zxb=y3>$m}VA6CbrL8<)A%&1wuzKTCzz10c|
zpoH_QYmRGHnKIx;F|eW6HHDP~Ih^8cS3@VH7&o#&Gy!5x^~h(!p?4)}7H5EPxJOig
zaAY*<M{wG;Zou^(_|A5u*H8{s?jymV8z9yuT*3rYS4b=w;t&oCv?ZzoIDlanqz4-o
z`FK^Q%-Lxa%ANf`-*N<PS?gWbb0Y3yZqkFvfHg{FKF8T7<2lZn97D}SOc<FV+#5X<
zxbid@oC{PEG7li4+Vov`YE}ByZh4~2oK9igP+j(Z7+;tm3gHiX+2AHbMyL4lLnWAa
zG=%XtS2zHgU?>=%-W^Fg-t%#aO{CpU@r&c{^KUuo#llSHc-+klE+4P-K&Gf^&C;Cj
zBcN02)<dn4<+j0XVQ<%SWE6NrQrmMzM=;%)^R`-X|L3|D^j4mi8B5}lCnk=k@1ppU
zO5HUah8v*0?i_i>Hr1ridrz2?9&D-rA|iFoydD#!pd_jIJ3=rEfZ>Eq-%=^-V9(9`
z3`c5nv-LtZQHJ$(+{!Gn&TBP6f+z$q#HBn{rLbC`6F2LtY#V5z--p3nQHy~bphv^N
zl1${-4fTh@&Lt#;UKfF}wvKxWIYl8YIQ#*XRb%Z4<|^D2PJ^a|ICQ!*9W3rdGfQa-
zW$PQ@Ct{4PPP72<0A9tyARgiHNKwIe;q7?J+-as1QbkM6gy4y)Sk1I{1c?I#fRS9g
z#DeNC3{c64k6^Sv-WSaUJv-6!+!w~62*5XoLDhIOsN!8X8hrQP5ji`>cX~%k0X+E_
z$sIf>MKg-eGK`DOAw2oZBxs(gA|}99q^WGp!h@qL5^{(@cpJz^sQQ$p(8VhNFvN3h
z#O~q&b118L*gWbA>ypLWLTje0rOBTmxd=OGlAOaE64J^4FE8=V7B|at$s6M68W9f}
z-ib(I@J${Bt5HeFB66(<*N;ue)MEq8yAdq~Dud5IO28lP2I*r06wzN9^dJBEZ~yu)
z^!oSzvi|0v|L6Db4nl{uo&R?33T(JKBD^-dr>ZqM+v_vJu>rQ`^2af#0<ikc?1pXt
z9tTSigK+6FvKh>A8mq%SPf`vyegmum5eT<BxISH|j=jF;);QxiHXeQ@&SEL6oMaUP
zQ%2W_#=@Xi<O&Z8odgY+v{&_9ivT=YCh$%f(jYu`oElLG3!3if6gHTN8GYZNz#qVe
z=L22*^hU)9N=c(yTF0wo%|^HDsfe9SFzkL?0DRORv71c+o~gR50c3+)x<=X%1vcEq
zhlH5{!3@u+Sy|6mp*|!QJ(f<2hPBkZZAJ7cKHLXsfsAYbU$f9-50)&)UKPQjkpkgS
z+@q~x1Gr*F;UGwUI(YZ#dEd=f<z<eN+?fCsybC`)WvIu*cIVOTch2LsW^b}1VFow=
zQ{4si&^81XVYiu?(TlNQ!NZNrl{0GFnbus9<4HSUs086=altFq6fHHMhv#HMP`|@F
zq85;_77y3mL6MzEUXTO?fRp2sAux8>gYJ)rir1`rc#dQVPobAlfbb#$GEh6gfkjkX
z035*6L{JL$d8x0%c@{MhIk^W=l!+Yg7KYqYncCL9YDY*rqv8TM09$9U4pjq0#Q=VW
zm}dbE0DLtPC>h}i%vg@g+=x&d&<qQg!eR_g8W>)je|0SxnIXf5x0G!r%>8$gksX$b
zgr{)h+&%|eK_*jtts7W@2Et7N2Vi{KZfN%-JLK4fBWTsGIt;{=uGv1vVz?vL2r(JY
zvDMWTDLSrsUVi`Xpmmh}oOa;ec~=!>;T%y@A>v(lnxS;={yldwBwSGfB(HeQih(9b
zfpGWO1%e>l3q9DC*E0?J0V0nHp0H8q<$J}B_-1-=Anm0&$2Nf3;T_QiJFz2qmK;A)
zLcS0sN%ja22`f|<>WD%~(J5}#wXSBSIW*5E^}iF&jGN=xrSDKQKx7ACL_m2O?IJua
z2l@$4SPCy@d4*~nEVZ>-8)~bw8z2mTnao^{-P41eb;N_s(yLccRl%I%YShbWUNdbr
zf^VFx`)l6eG7TW(crGGX3kU$?Ta(Ww6{oOFhw#H&KiN3uLFudsA`4y;G+Gs<Ugre2
zKxK}55WNN{NDxgm6DH-(cu;KtTtuV+x(w$y>Sy$xW4OQpFi{()Yzq^RJBD=^^xUcr
zaU>_Z;QVKMK-|w#TWtti5uU6~Dk~@$-~m!A#2_3DgX9R$Hc(@`g#thTIQKMbUaP)l
zAw+m6YBcLlQ1v@233WtuFv$Q5;K_lp&_~q8Gc%*T_Vf8L#{k(JBQ=7n;$65j?`|z`
z=|%fK*j}l-g~U8iX$fl~p~toJI|ApsAuW=Wq)_d7+2I|T#(}DAelIRoCCT;Y1XZ6`
zp^A6m!$>FARa-MVt{KILG3XPNLTljb73ZQAh)&h0DC0S1qYW_okr=_j2kVC0G}PM=
zYXw#SPK9GQmW|-u*1fpr*oOuj)C8l~EqNBDSi%HQla@wbNQ7jMN{b<x8a{6HSg}u%
zkc0AFx+Cz&25Fi?LnK1@Dnk^)VF8aYSIsDHj`Rq(nTUU+;W6_C6m0MYn8`#;nCV|z
zfwRH7T6(|NdBuJ2tUT-k<%mFdo`K^lA>>A41h;<p(M0tKM@~QwVB|93$l(O9TFZ2$
z#UMMlo*Rx<tUqI*8WL%uO|t<@JYnR4a7?7mNQ6vJaq-tGS#Xl)f^nLl`1OLM6N-Ti
zWo}0C(E>ey-{wR9*27i?JaIa0S;Fxeb(wBxOI10e$)*c>ej!bihp7AjM^f*R0~br+
zY1SOX7xfp^MFUWk;fNy=N6T$pXbPk22nIcuofb8HO>3sDJJ|D>5HSdkN*v9#Gs-R(
z)YfIz&S>1U;3KNm@4~BVWW~QEtl1GM1^5V$U`#*}k{}#c%vRxJk*%qFhGM)K9S1m&
zvj3@GhdL-bkY4Y2GunWtpKrl<V(1uV_ZeLtm8e<Fh44^J5vT%i8-)yA>4CM`fJX!@
zu)@ARC>ZDg64Od}=0R)yZruG=An|@(BV<q~5~z*GDQ1#-7R{)Xb*|^h{31kE71u0-
zIZplqoS+oiA*Nali)g3gNpIY$dr*pKgtrEkVtHXoVzn1D0)4LNlAc?aPvh1n)~(|P
z<b7Bjiw33gGc%)R`T8pUeDzi<D1Z{qudX?+S!K$A8^yqeUe^>>66A1-w_OdLkYe1(
z0?`DBIn^Vd35VX5s9Br=!r>lK0m6~ds2{;;*SZ1Md*D0UkzPYNRJo4?gKmIWn{Wve
zP+cLhWQapJEYOyy4&VTWU63AZSmfhXoib;qQ7Cuz|9s04v}LV#UC)WQi@8Y;CIi+e
zk@*~FpN!`?YjO-V6ER_AhH!86P~ghbU~n!_Nyt2ah-%Y!;i*;WTf60nGIKhGbwhR8
z`(b=xf+&PP>}7+S5E-4~%MX=c;?WSs-(2AUXo8_&fO>Z%>3Gk_DK?RIJH;=K|Nm$2
z-4Y}Ta^pbqGCTwCFh8r6R$6JL)qe6q{NQcZ@pQl)7U3QlnU&QY*<Im)!C-hqWLEXe
zkTv!{`KT5XV=&s~G&R_CJkt&7BD>iOecl&<N}<|lYK|Ps0?Wj)+~^~rj(1CPN1xFW
z40mR~tY)l#U$=y@l>4d2<T&L*$NlL$NIndO>S`9v70{fwk6e)rIl=VW73M)VCRPW!
zMeOW(J`SXW<b>p(FhMT>h8+(4!IiWT9DNht!x7um>^$IAA;ZSA9Hk~%<#rk&Kx6_K
z;*uV+f|;fF1B>~YHZ@d*@5A6!RH6tL&{;DuB?hAHg8EHijs+x`p4SbF+F8!j>11_k
z!r?cl$SQM7FlOPRup87R#G%qkcQAPj3}V3zri?xTz7dPq>4CujaR9euiJ&;b;Yv}$
zPvOIQQs1GcU{c25nh3#FC}Y{v+!CZ12mlz-wMsuw{(%5WDe)5Q<+taezM$q982wlm
z+8_zQi_M^HJPgXX3rCI5`ZFTCQ+y83NJ@a4?=8848!4$q@h-b?lG%hSedz@CQ)ToH
za29cLG<#ixt)n1}p#Z{%hP;K$4_+{x+zbFiJmz-WRovhh@-p6K&Z@$^AaR>%#q=~c
z=rbf2i58qdXLEr<yytfZjf->NH|WOkW7BCWF##q$1O^wd4Bme$0ev_P;`a`aMZY_!
z{mP*K;~)R&g<lzj#2@_Gz_Wp81J4GY4Llq8$qiip0pTaRerit}kQ+JQ&WLooTa(ZO
zJR5j6@ND4Oz_Wp81J4GY4LlooHt=lV*}$`bX9Ld$Ufn?Zj~>u}`k*)p_sZvt&b`OF
z^ox@2DgCm_=eKp(@|U$Ey?$j=XpO(BSv15p{&!Bjy{44FJ1ZAgf+=^;QN254@6gz5
z{+#`uo<#LM+I^-<KVZS^Ic(v`K>R89eGU5m-2Hte?pYf3cWaOIlg>Z!_$F)T*zjF=
z*PRt-#MuWS=lqoMx!&5HqGI;-Qv?5fbzi&1mTf2RFFD`6(X!PFtgwS&i4Ie~dnN1L
z9S<nc$cgrHl}ky>>OVo87qj%J{CSJ`iRC{vpA9@4csB5C;Mu^lfoB8H2A&N(8~Dry
ze(}FxesSxcxwqd?<+(}Ua2Gz;!{2$o|IVpD_f&yO&)UIa9(ZR1&ttT+t%v(;;Mu^l
zfoB8H2A&N(8+bPGY~b0zvw>#=&jy|iJR5j6@ND4Oz_Wp81J4GY4LlooHt=lV*}$`b
zX9M51f#d()`#Ab-vpoGg8+bPGY~b0zvw>#=f7}N2R|fs3fBc`f{K}yGvrYEW?hZnU
zx9Hp~w^-F_TK?rOOu$Yruwq<F3c|1E2dSShFSma=@oiInT}NX>qrbk1>zV5{j+h7a
zHRKF820eriw70k`wO2T)Ih26#l)T{Fr2|X7y#TEI%>ld>9JxM^Y1%m{S-?&sjK1N;
zxj`+mUx99`{IZY=(9yowQ=6TL=l`!v>Hixp=e@F!)Bi6%LxsWB(lvV87fM6;so*;a
z)|8#^td+Dk3qm>!q-wv*2{Vxy+}}0FOWocz^<6rbhWXLrmpZ<fv^(f7Kee@=+%vjG
zlID$PDpxs{P{jmO#yJs7>_qOYbEnqM47YH-c{i@fE~b|>`u}&^s|K@!%|du00UF2!
zuqx(g_qsKd2JXzC5s>;KJl1J`hr0D*mA!OScE)FfD$Qs|x|MnYaN;aENvJ#B8dY#)
zWbXR!Q$Fk@mW_5wf&su<<ELbCG*OPE5}<?!5yoS%e+Wf1WE9U{jIwR1b|yjzCPs1s
zJ(!8|CpciX8J>}0{F+dsLuz(gie#B_O+!UdGQx8&7rmFo2Z6zjUDcr)jd|ThT!h>0
z=?ZGaglM<$;1)nuoA@3cBi-+olf@_k1=N#J5aIK5ux>9BEhrpGUu4hy3WD&t^~?-j
zFk@B$Pmll;9A%<?NSuQ+62;c+Jq^qrBIerJo0$=f!9dTwY^`Y{7Q%ZwHKRuH2Myv|
zhQ{+@L2CwyAl&j6b!M4{5IFtXi_Fa4SgbbzKDfs8kg#IDn0{0uNC)#M<AvBz!CX7D
zaz=yarW>4?J*;)a%UPE^B)^O6lZ4U^dP6C1#)13*5AKCAdJ512aL+rkLUJ%E!(BH<
z9TceX3U$nlu}11-2p%MX5P-w5Ly5kH7;qUk1_j66=;;O2gCDV8?BK=pJ4VDU#7MNc
z>I4+MAkIxVlA%s|%q=&HPCpCbPEZ8lVa_-Lq9Yv0M9FT?B8tobP69BjiTJ&Pw^`?<
z5d-19)pJJY-eBQ!b)-hA5fnlAenK3#P%R+KC_^Q<F^J{cX~pt+9L?M<4+_J;f;`NH
zf&JS!OU=katD6+5QB+D$a!>^$P5+=n2?cH%Md<fKG8`Ha;eGC!QJ@SJI7=0HfP&cz
zU$RsQ2}V5H7iz7hukv<U$eP((*14FrtdGA7ZAg3r_0poQBwfk*h6PPMylW7oi}0zj
z2rMKyg$#G12!s*c6AmPulPhiX0D5&{7{@iHUakXHT~Bk@1@|-7wIg%0PNoMhnE1Jf
zg?H>Ip7P2-0+W++?{729f8Im0B7CdETST@hzJpv%QDIaY;qT~cPm?<1?zBZYk^=;l
zQ$?rH1XNi=CxBwGN!n@n$+<~CQJcf0Z#C?6K<Ez7**}aNbaoK(rV(cscfsFmprQNY
zLDO|sw`^)bX58ZNypRI5ByhjtNx>HOnr@&RCY-WR_BLKus?i|s1UEwP6XrYXs`gt-
zAW?)L?ea=arH6Fx!mFAcLsdFs(dt5#wII^d5*fh#J)PKvg9Hj{1zk%ASpY7ELCmVf
z4`bR|>@eD?8xR0+u?JDRi=TQsbWp{^q_bcxx`Ni6vw9nGPyoOi!X-7zMmv%b-Z|%t
z&b?&4T%Z&vLGBMl-i^=8K^13aM6P8;g>SY0+{s?q_~BOSV2x4}D6)>1OajZmD}#S&
zF3-6!Y@#OT_5n}SkHSZ>vd@zQ6h?Si<JE_1vl!^%Hn?i1!j=n8$rF&JG%QEpnrUaw
znmx#afm!1mDaA^#X(9Z)%r`Ucnrc^Pm%qBhQFrw1Hct3i0Zv%VBHAJ~m929G<!eot
ziz#o7F06P>O_76yCSn4vWhII3AhBJ%D=t|`{x~^hCFeK~N3%%wa8?~j*~zK-%V^R{
z&a%XNiLdqaTK0v>V8I;7-i11wDR<YqTW@U~uKHSOl)a0?hLO;m5SYO<UI9r61}rn5
zoAjJqcz5d+B`$vNs+2mm+&2@Q?M7E33&1dd_z2I{jy$LU16gg6mXjXuYy-z@wWyu&
zx#?v(wPVdn|Boe8<iKGqAeho12L*QVg~3xpqqA0TU<EgrB0T|K(IR$wU@$-&z%3a^
zNO_7HVCPVNPEf*6@KMdwb4e;eNjnH7-A(6BvYTD(({Jm0M`(xZBW)(F(J>1}-i@Po
zAgb&8$ws43855u>2~Un`8aSR7x6|V0>j~hZ#UXLgo%LsAkUAw0zamxGp#74I^buZn
zS+*{H0U-b%2)#7sFG@z6=_4>hXz|(!-73W*810tU|G(6I@tuf02RoR30W?$c%we<+
z86}?uw}5ianm5RFw_8U-h_!?VQZQ3F#mDgIryOzUR|ft2fBb8Es@m({{-yoqp#Qx0
zyMt1V?sSXlWG}UG5IW2Qnul_s@G38)6y~I&z=Ufq!n1T^VIq#W$V{0K1MKu%GD<Kp
zYUeobIO>XvG~#0_iXdF7m(d7D+Z&$y05l~Q(E>W?=xB1N6xw~o9Q8-P(ylo*dT>wY
zJyS6tQ}vOtgTyg3`X+>MiI36IR}}zLR6`}B1bfb*c+8Eq71ZQ@BvOaH!w9hr5&(>K
zKu7yJQ-_BD&<F5l`(345xtkI%LCHG^G2tw}P`y#0$`CNu&a8eQgc1Z#|G^}pF4vlf
z0XP&t47$k7N=R^m>Dg?+K#Our3Hj)^iSYI{h8b0m>I9QqdjYtGr^p1DhH@z$IKk}1
zNHEm$>~tAX;H#K(w&E=1cFeFMb~?!bbVj1vZnUgu7a}c;7zi&^6iONN2ATe}e=ORc
zxo%4DRXW@)jt;J8bi7+Fn4H4*{5Kh6eJhjj;E)+yqx0GEZY9D(<{JO~LamQsh{4T`
zZg__}b8tp_Ef51R<sF1hWjo<qsjgF_2W`QG(?ROY8a)~(W<)zJ0|sE5D1-3v%S%a8
zom!*uhBpS0oahGYXKyI!Zp@D$*@J?*13D%pA99O80VS=g4(rmTeMOt~7!X`do!Xg2
zS!0e1>QrZRD3((cu?~}3kn-BMw~!vpl#K91(zOg28ZJqy)3I)mT=&RgPcQM{!VIwG
zZia%_%-|7!Z~?^1b}3SVOpi8;o#NsoH&mhfWoy<#IAcbD)kM)5pAnLnQ9=ZrV_X!#
z9K)9)tAHq9;P?!iDo7}Xu#EClnWz@+A_s*54&Yfh5~H1%D3Ed47h(40BAMv0ZJL3Q
zB=nS2KtaX@a%uW8J0n9*DiOQsfo=`C<Q@hKeuu7Sc8$_H7KTw6D2p6}Kr6s%2BcQ3
z2{l?PDB4ma%M5-rV#+K!BgRJ&gj-%d5gXtc&K?QRNG>-L4)}FwM9sG|maDMiCP7Ao
zqg@R=#A82uwYyNa1~nA83m@;6l^^v--*~I}#z#5`x4f|q;v*dCG+a{V#2FEhj7Klp
zUU2N30uxby39hJRK7sa{Rc6Gv(;2B`+gCTHlnRJ4IW<N4CI(n5SJLzH3ZzUt60H(q
zPtk_;5~)%InJ8}$CwBkYNQZ48PysNYN4EXIDp-t!nP<2m1j=_rv6!z+J%U7xjxZVv
zsLbg#!s+dDAd^DL>5>&3uweF_8?2_Q(y2oK8N}{fuEk`}4XMG5831NVj72--8*FFY
zbWiBKj*&Is?z+MB?sdy_6iIIyiw)<T+z2K3Mx5%kFHm-X#T~$Ds`hO3hXRs80Kl-r
zUcF_8WHk-3Gf*57iRhY5r`f#_SrzJ-8#zZj6ij$>%3`1gu934t6GfoiDMRfXr2`(c
z#5{^kWCg063R|myGkbmF0A7Tnos+v4teTDrLA(b<&}H4=n!y`*a5fwWVqu_dK9V8L
z9-|Kf_BzDAxpIc91_ILf%~M>?9PKNb@rdZOrlf8-Pz2%LRM2Q9q9Z&eK#rB@1he!s
z*NA!oIMz%VnV-E*M0jUptL6`=x&s=8Pzd2Y_mF{tg3njE#WUV@$g(eLLLOoI)+m4I
zP|TvlX92#5sVbx}zJxWR%|`)*Cy77`CQ$~Lvns44B7rM7SDva2tqg9<nGq2hBajCJ
z-VkQM2n@h}Va6Po5gU0Ny5nkDs+|Me`9c_I6i!fbdUCcpi7fjXlumR_1D!cN*B>jz
zsOd$F?04g-fE3#)nYr1sQEGH+v>*{oE8yo8n*gB-tU<sj7UOkr?<vdqsKls)d9FK9
z2G=VxpM#m9UOUpyP2{&+jdm5Y%0LNXl}yiGS$d>cGq^-x+-wfLbeF%!_F_HIF|lw2
zuf0d6><Jj__1q~^0h8VYR9S+})ER;86x6vAA|bpfc*qUZtd<*isfiE^X0MUR&55KC
z12Cey4&|X-7yt<qJOS2l6&Nsb+m{}ecytZ|oeBa5>uGM!JF&JCPV`_H7b+te3|=tP
zt&KQrf&!5k?J<@_Mj{6jn4FZGD#i#(xd}%y>-`{w9XxZhXImuHBW}c19oU$^7lCXT
z2u)3#3Yahzm~aBxmmbWTI9?n?0NnTN0%9P(BxqC)>~zh<=`dLu>Sh2UbAlO|P)-?9
z^X-hfz3_;})^hA79g;CAI~|aU4zruR+!J|qd_V_RRcbVpFoOLEXMUiVc@Q!RXcdi6
z;$@i<Y*ih1!&|x8xKJRaK&J<~WxWd$@L(DzECE7A+5`y@;9xf4><#V^ls-mq^kKIi
z@C-5Bo!jA*7#|um$wV<wP86!}3Y%7x<;n9jGwd|#=wHPxZ#<E{@=zGyC7i(SbStc3
zXfJYd2t`Vq{!uS4w>qNL?c>gvb~@1jJqx^mn0yzGE2A>Il?R=k-mQ}+C_IrR2f!7*
zrq(DIq)_`}2BIN+yE{riVl(cD8&VK!Y>2DWBz=q<y`5y*7g-jv9{rR+yYgSBx#2`H
z{Eb%OMG`>;-2d~qpi}WQT!ab;lS7q5QY|Xz7#n@B0RzN9L4=>F7I_84DjRsa&Hb5a
z7Wplm6uOE6-r;O?ZjJ#Qz@cQ8iOsYKNC)r%zh`tyDv!SP^KRVg^mD2>0o8y^DwwKU
zM{2Z^7a7Fw%|mNSHNei0nW5-KJ0y@<hicSPLo}FKZ3tK@sE7K14Z0p?a0g=>qaSrZ
zZ#V@-*71@_U=C!j_l)o%{XAh;sN|R6I7;uOH#sNHh~}klw~*DPIj!IoQ0OdVK_U#a
zWE4X<OwgWC9>4(%s~~>fNXaU}qpm$c40m&_Vud4f^hsb98;h}DB|EVQ7SO3ql~Z9*
z2$2zfG(f#C$~aXEq>7uc<b*^r!D~uU-Y(wKLP{`)x;Qcr*@bKkjPcROB7sR}aAP0o
zGcW`K9?VIGRsG@ySCfj4kpUnJ#WVC`revEJP8et$j-3)zjCkz}(Gd<aii83Rldy>>
zBcA>2xh8;LnsLw^+<Auf1oeOLcxP&~Y!`5YV#%?r5{%@5klpN=W=szlo7)*MVN_P1
z_$enjAc^5+;*ki&l2CBE*)7o;eMU=g(iefF>!g!DfMVvv)F8t1q=VOm7t?z}RD#LF
zo<nwN1|2i$G@EjwU23d>S7c6dE{1zrw+;u*v_5XalQV2BK3W6sO5s9+n{wQZ-s@qV
zOmlCLDcu2tPs_-&A~R0Nj4FQ!46cdNS%nNsr#WMW;lpWeT>%tE4G}J{E<5&Nj8HI;
zIXe11<+?R%3c^<n7RN*ifKG-G31`ise<c(V2*H#Xh_)-~4-|7OAi?y!ZdlaLa2q=*
zTb-J4{KHM4jk)ZPVBot6IViA;!@-e|w<ZRnWjuTFy);B%JFS{g?HoFB7sLVV!YDrB
z2G}WuyVosmiH)Daa~FM&gomDjIYf>GES9sgr<@WDrN;Q^*QE@dPm~{r9<oqMi5qC=
zpp}*zV@Ld>SkyR}A<?n})B%iZq=e+8dMNU495p`Ioe>>?0_|&xPe9<O*(b*`$8#iF
ze82-7%tp#y&#LhnEoGDnGhpSUR;5>1Z#-JZOcdS4(GrmknjG;F9w+TKU}v~{-Rcr-
z!GQ2i!h_h0?oe!SNt_|U<Mps^A;bIu4?m#|kwbO#RLn`HYUiBUBjMQzV9p4EV5Yh?
z&S)J<+3q;kjnJR7y3uKhO@MKSz~Gjw$ceU*>3lPe<U<SeyMx-V4EoRi_%D0EGAKyI
zF>=N1fs_QFJGEww&T4@d%$awLOsNV?xXK_rS6O4_oaoK@5n<Z^J7fB}4YB~NdUbR{
zCjhszLB%2=QtCx8+6Q@n<{49b0*s0R2)8_VUQSfSF@N-}aKxi+@%T~Ev1Zj|q2!o&
z8B?heOl6J>4b*Gy*u>~>^m?rUxJ%Fp-U&>&*w~Z?3)F!cZK^OZRXhZM41hPwLKSbR
zWR#$!457hw*+t{Ubg#0lbDVG-^{Rt$CF_QqSz<yInDAi$?1Ic1uF)nA1Gw-IIajC{
z6*dS57qNG7I4=?kJ%|~J(VmC^XgMei;XtW^Yl#|#jUf7N<4~vS{K0;313z@9exU3P
zC2NTqZTA!f5RNnK0jV=<)N-mbW<<N<78puFcp(9x*A;;913w$s-vD?VE4X9b$^`)q
z;PVfItPVeI2HLs{8(Ww7CCtBP(qD2zF188pBMe+D1kzBM0Uulzt1wVy2Y!?6P76QC
z)#*={P!nqO#c^iQ896Jlm$AxLUOd8o&3M*Rj}P?*xAV|9xG{nK5mXrfAABza84-?l
zt24L-5NA3KXY$D$bMcdeso4`gDlqDdzE1%I&PWELXFPjJKpn(DxNFP=K@jeVZp_T%
zo*>^9&8V0ahHjeZ%o#Ic6{mB9I^_lc%YeuX&xq>Jff)&{By*ja(uOxRQbBmvVusQ>
zyIY+})-G<@D9_^0dDjNOrU86c)ko*qz_Wqn4LtA0<yB3@YIhYTi5_ty3j#Qihy!>L
zj&{$l(KW@c0vo`ATsU}>8Gu-^OOJ%nx2#PZ(Mhhy{QrQkLtM{XJFN*j5gw$CDkn%N
z-~v)96hSy>2GJ27Wgy3<g#-`)aI9(4d}e<3IuYTWv0~LKHL7}butsAG#|#h!aMVq9
z=XWa+no~z=bj*FfU-VHx)<+pCK{z;C2*#!OSY_}B_n`b(XwKAWVW4hEv!3JT4V>JD
zvucOwFt{B+3A=bP&W=RpX3wN+bbMPFyTpS;Fnh|TB^oD6F&u5KH38N^87RR*2lK$E
zI)ck3Qo(}?Ho?m$sFC69y<GDKf!I;7gV~FeAcbL|=l*%tJoJpf%@ovmLrbKbhEfm?
z8)Vr!t44DlZ}(6W00;15o3CXr7Kt<~sHeG+@*VbFo!zBVZUu^$ff6irYY^ol9DRHO
zK7f%*fg^=oydJ-yl~9F2fo};cHH}CJ|CZrEAJX;*o99w>y&A0_Fd7H&0Sx)OI#EB`
zV8ghsCDbAVI=Gr!qbA(WQ$pu%P<zswE$_*VHy7NL2W<iPT)a#y!K76%3Mi$VI5|Cg
zC#iE7r6YV+n2|cSMmhT+G^y9z8m-=Ah&dJdYM`8=1oIe(g79jCs#vC!U~U0PqEv*B
zsz2)giYD*g5E>fc$Mr!LfQ#>*h<Q-<Ldt>K-xS)~+!DDa)#xWS{hrE6H8~qIp*)0l
zX~aNy&rH@PJ>_0;dcAI4(RpXi)Y$}7)tQ8ub)-fs`D>Wy^VLf=BLN!l_*5xIqCF8S
z;5gwpDtd^six=N$s)i9(lq(?3#*te7bv{(cf$a5W5Z)Psg%S`xq`SMrhg=W1aju`J
z$?@FTD0l_Lc?h3Hj}WAS@Uw>;<pCVPunOYm5uGJQc~wq~*}peZ$d&znzU7D-O7HLp
zj6N|}v99{3Sk>eonyo0EpyW_b6R_MU%0zfB1zDNq-5@@~O$NqY`P+=4<H7ZAckz~p
zQgFbk!Av8pomd7$3b=sQiqEV`Y64nLg^|=QK9_$3f|}~35Ch>o^NTHQUXJBSmWgMR
z?+cX~C5RuiH$H)&leUKxNlsW2!gF>=h$c)PipQMFE_4d0%KQZOMXj8U`%S%qUH}Y7
z8~B4OX(KrLCccLwwyD{9z^g)r*=aJ*`E20Xz@NB*@J<h=OgaI+03313V=QbS?lIWG
zTo5^?lv*W7Iv@aGM4viB`3C|hrNm2c>S7Oghvr@fKgix4ewROwh7=GEA03!5*E~j$
z5W<huUWrOYTo*^+J9@vJqaxdCyJEY1)X-fzNdLi~4Llq8&<6BZ2K|?R{9k*&GU%Vi
z*FQVZJ9h{ep3^>J+kfLlEQjd70b`(jlcF=d>>HQ<*JS*OB>b715|<nrK>k1Gu_k*w
z6lJra7jil<YolfyCml%>#j${(cI~AM;d8EgfTT=vn5!sHr<Vhb&gg?^Z!L?kP^H=A
zs+^rzT^R+ji;6zhxEP@VC^ZFoCu~(YjgQ5ugjcZC+_6xm(!e&%kY{i@11F{ij>-k%
zT_zW~OKj4~dzr!&peYwJvs&DqMqbtw9a$M&>20ZAF5W$RT$$#V7PmmBu38J3H%+?Q
zQSrTGBbA!f&8%b<vlURGL^TSSp)s`3&zKUWVVy|2WLL`mX)`D`uFXLwJN(4@cN0zU
zsM0InLQ#@zRSkU+m|m7T1rlq+AwlUwl7@O4O84Y17QEc%dEz-^wMX<YRr#+~-O6^d
z-1}EPacGi*N0AFu)83Aa?~{cBQwra%o+!;LYa$bm81m{_s}|h^zY%LmR!mE7W?PMx
z&6V3hf8PH}ZOccTtWMiVan#UfS}m5r>l$!th8Bq$9#STE&Z8URsFihK5@>*uXn|FD
z3%jI1l}}#U#3lKy0rtt-IZhl#Lqwuk%q`Bf=bWR&VsznRa(;@gmvKB88f=FZ2*JTb
zLe%5*VntOI?RI6u=A$(JvVg@s5C{#EB5q6-Hv!kvym<yHm0yP?j9QZ{dT{b~<~C}*
zj2>s{s5ynU{-v*-&cZ(DX}VCi8fzYWMrN^yobX#y%$DY2;+}0fV;1RVjDQiN=P9$g
zM-ft?w1&z%le~==r^_X0ovDPxi)OBOgNIuV9o|qUNHpQ#eaH;cbhFnstRu(P#t0kK
zvZl4B=lCWb!Y{9u^>QOpg5NeCXj2iae(@f3=8|DT^yb6~7P50>_J;I5Zk>~A2(~<x
z5CUIQihnLp$jT_$+~d=;0<FSA7??0?2#_1P1!@zFrV`w?GNo+-L$BPZR(|8L*CaKv
zmm%z{c~@$o99<#`-;48;cCX~V8gE_44#&O_%m}K{gnSUNT>H!cGzgzPlU3Ym`U%y`
zKuare`!sM_HEQx23Mv)Jka=f2Mco%6{@!DoJJ*6!o~QEXKe}}*=?k2yy@t^(K-qPm
zMzaQU`0naxw9815-L~oqwo@O5$oiP=eO#$1Gf4Ya@|SZCZdF^~AT#<y_^)yA-9uDs
zHr;f%NgU}ZE1b^0WWf?JaH&k0gi7$6rOcQgWI!XBm)~!}RBZwdv<^ciWalBd2YJlJ
zi)3vU=%ohh;IUMuRigHW<nBuE-4E75R$$#@mJJG6gbpXthE_O7y;Iqv#U14ye`*|N
za*%JzIjv?)`#J||q3=egsi@uZKkeXcUcQxZ3tNiT10`|i+9`SDfAYP?6P4?u1-|X|
zM(LY-?l)^$*oLzcC4T%qheo6~*dYuAiG8#0lC5<CdZ!${P`Jj2HiCOp;|;NCC~BqG
zNlu5CtylB4Q0Qp6^bZghr#+b_?*G*xPP9%s8LGGO_%*=a;<nymE6It0TWnoLE-A_$
z`t-8$0@(9jHixFXos<Q#w{j>=KZn?Bsez?FTYHUXx+tZ`*;*>!NNIM}PS~1Iol>J~
zibHF%18Fym+5qz8SQTF3v1%u_NoW`@MIpW+>htm1sJ+k!t)b<?wMGEUVI1>_Y>GG5
zFn^3qeLk7$FnMx^9lByjebRc<>wU^EH+)BHYnyxfmS5h*jRNzdq~oU=fwi_n%MOVK
zz5Tr7N$E`<OyCa+SDcA6(q(V4as#(Wk_gKNweptVyX8>r29p|oZTZXDw5z$}0-8x3
z*`)Wx%NB2ldat@y)#a}0Z-=6Opntjfw<o@-=}p-kdb(FvAAY@K?_O`}6&E|A9j(Tm
zyMD?C8vD3hkMzXJ>ZH$D)Bvo6BQT{3^dVccYr@J09d-Gx)3MK4eWaykKiAk?sV~Y~
z<?cr&@pP;)^z%M7M(t52Y^u?Ri%-^@`ql05aiLu7b?ne@%@_c_XwfFZp9ZED$)TMo
zS*7MS2+15!6~SHu5k8FT7I+(BC8!JD6ADiFyuXQ30oD`6)NX6L_tkV={$AxzsU2_R
zI0|}P+%mmW=sE5$*CV(El;AM=<jA$dW?kymlJ7bDFvz;6&aEo6R$Z;>ma9Y~o#9y&
z-92Ny`79^9vnDP;d-6)%a_C15YE9m87dn%{b`y8`$NRZg?YP!fRNfw2DTn0UL3jCw
z#r5ZwW6t}dZ6;Z3ct^dz>VAc_-A}CeL9t&nWF%$UScu(Zu8At^T78s-aDFz<$AjtT
zshf<?RoZOq`0ebEiyOQ0%l!9mXs8ff9xnn*ZPVMScxi4aSAt8$Hc3TaH8knIv&DC6
zPlM2pGxqA1BK0x`)Ux|I>GaTGsg!(|YS$s$T^#A0HS0)8+F8^2Jd3*1@8?>a0bd?P
zLsw*50&(<9ek2II(u_i)Gt#J=fkhOCt4P5eCAa*hH%gjs=b*=MzwkZ=%7P2TdnP7S
z`|5en!h$g6H~{;oDplCAJU@0!frZWl_T!*UdJ^rCJUY({f&wU!ilNvBE*fzL9;aPo
zEb$7#qWT^E6qWX%<+a$|+LI4#FfQu`nwgn?)@Ej##z)SgHSVUsr%%YG1{Y>4o~|Y7
z4HDm{YNid1Uq$U}r#JW1I`3mbmjOwi?pwX!8eOVIsnk5xTS_=2iN*~=hgI$mTHm}4
zT7|r5E3Xxk)xM^YvoAEH!(^S;z2)8Xmf2d)%Ps0Xp~D06UG??b<bvl$D^4B6F3ys?
z;qW<;#yAuB(KvN_^tz$BO8y4f&q(CvuhjJC;~=>|^9Ljet;hwD-hTk!X0~dzHVh;>
z%1)%<OvQNG_U@%Ou-OwS2&if7n@?&4y>J9&485*?Q`a8Chto-c7@oT9o28JfI3}#v
zEmQtZ<Xmp35bi3tPw~cwMqM?hP1RopywI$45Sq!q0HI$Q^zZ-iuQJxZ{Y(4JLH~8{
zcL(Krap2io+0Fy%6M8oAY~brRu=Y-7R=dbw8*0c6uJYQq7UO>A{BXH26nz^F?r2gB
z!~r}i`~v^andHW0Ut7vy7XQTNEgn*x^%!?<KzH5_w$Lp=xV0kjnH7<4cgylWb@FGM
zl-DMzUMlQU9aeSr`We}y&lLy^%RnFLRX0mmS=4Hhg8nm)A@Z$*@w*QTNw-Y%3#wp3
zLX_`j@;TE>y2x)c+R*lw=NG4%g2P2%6G!yC<xqzbU{c`yedb%X@)pq_x;N*uoHKr?
z^;lM$^j_VFr2Fh;`GGjN_cgSjG}Ink5@=9?9!hh{ZqlLRrC|8?lMZ1$*?CEJJ?TSb
zriOm1?$pQ%u=BHka%H|eW%fd07zmud*o5nWfj@XzH_aSn;*a*XZ7vod(b2!H(2ol(
zOtwJjV~pg{XQqH<s4J<p#+uSXozcwP@n2yf>Y>nsnd7Ib;UQg25CHI$ejR8PP!YvZ
zxYs;qbnZQt@BALQkE(orTZiL)Sv%6}S2l&#_^X;lLu})J=hWM4N(sEPa&aY?a`zn7
zyF>O4jlJg2+3)E|RNtfBXR7oA7R;W*7LE+WpK{;Vp#L28z7qE=jrzN_NBT*p{STGy
zmR*{59%M#b=pf{rpE5qzTf0-AEB3i{e*a9bb!gdk@@A9sxm{nWRJK}y6?QNz(NW4D
z^Q89yyVY-kU@;Qy&O5O#C7o#gY<OM_d{kniR~G^D030iRpZzPF{^%N?4c)%(Pp-7r
z>Q)*E!0;2~E`DkPe{UOT3odGKA@u_k|M$~st!W$vCUjNq;XNDJ+`#U$w#(ma2h**v
zoyK(E4T%hH`w*HC_%T?1VrlxU;`Qc`{qR0Vto`ff4Vu<o-9#u5%$i%G8(9<806WV9
ze*?I<fPpr+ZBaAw<=Z=N_?(ePM0|wjZ1MaplaiA2rd{V6*{eEzu;UGUXoFJbxL6S_
zpft`|Jdv?;lMF5ZB0o^~2TFqBKMB`U6Y3E*-0S!!+q_h(_YBi#%8bBd32Y}&+NGqE
zMm`??BZm0+*6u90a|d>~`#svfp5`hInZf-6PsX!>U$+5$_UZYT_rSqB!3oXqe??ec
zF*%YV9K=evfhAs)d>8bY`k$!+b~F|reM=j0{$DZ}*d7b=n@fnHmii;7H{r&eZ5)dp
zL=QrD_BUyTH?GD)C&7IE4o+fIFMlTx?m$}6-?@#gsrj>JE93W)m!6G4q7U7pPuxI_
zO961C{093E8G~27Kg)i`Z@8yVkIx2d1H%)J+yS@j__>HE&i#}T8wo!Xc&{=3mY&{T
z;8NEwb@fs*Za2yD4<QN3i3g!f{HXVxU43)~vDmYYxQ!3mY4OSVU=x36sb9Lz);9m+
z1>BWiI#rm425fFWLTvhWKaD#9k_}+ji#7IaS9U}5M}JTDVgH32C?1&x6(|5ZwaodI
z8~r_#Sf_SQ>qY^Di@+S65oQ8-YUBHW<Iazx-#5*j9f8>a94c3g3JWRM%u<v=x>V7K
zAseKo7GrJ-0EhXzNkHq*hh<T4T`=&qn>qTa*_#feV0xI1#nGMSMwS^oV~yuP`~9vB
zXp@Lf*KUbDGA(2Cj5jh!_KKT&zlR>CEt~eZuE_n8eE_yZ%I*1P>3>vf?jYf?8%CG_
z3;gs+zdNY?%Ao)HkN@_DUl~N(vdQiI>4Bg6&jy|iyt09{cQUiuMgCg7Avd_n>mL}F
z^5!$N@a9UReETMKRGeZU4&brCFYy1INp4*BwWS<pwMQ+)gLz}ykJQlybm#403*7>Q
zo3zAdRz$kpEzAGZ$)9afUYn?TsjyFVSk>9<XJn5)S0FGf1AU}d-7H~cQL9M`{f)0w
zKXNL2bjk4SlPJFuQNEkW=S(kY)L(bD+3YWmDNZ#8hl{`_j_A|vuu_x&lZ^CNo^RR8
zTSR~8_MFdh&iJ9$V_9v|dvzm{?z5NW2jbw~*U*B}P<wDmpg{$CD9tImNr#G;g7O`Q
zQ*{XI$<9l%>q#FfGd1*6b*DyFfSsQOlq>V)DYF*}!$9Et#U@-24E({%x@qPp6MwY7
zZF8{ziH`nlg??OUVX_5EA7dnsJ~IU@LtRO&HP)0C>WpUQ^7E4+^E?!KFmwD=HQaLj
z1OWhh+UrKEfQl%N!oBu6qjT@EeCPMbeN^T1+d3TY%i57%zp^Q`#$VMe8e$v&JEz`W
zQ%c~Sm5VFEl)LAs-W{@cXzVqA&VEl%qWT`~K2xP1uweEaws2%1{*?Q^2L0!#_m#M3
zY1H4XJ<?A)&#w%6@#frmM-F>w59;TN9d>eS@cU<atwYPUlb4sATi1EJSlMa?R@lL?
zL`Nxq%#+>+>{h=+g2hO*JMYB0lysu`v*CF$@KK44UR?yp18}VPefF<x`lD-nHgx;C
zKe^Igt6OOx0K-p^yZEUI{Jm|UEx4$`h13sF{NGQjwWe_xn9x<Zhxcq?a|652+Ae>y
z9Za{vb{f-tHzYE+?L%ln;KyM3iKXeYir1S%_QU%evG%W@H)vXWbrYdLFl%m!Ze&eV
z1MDmd{0-pZ0tVXPwnfdzmv8U9;d4eF5%Ce8v&HkbOiD`5n|7URWUuP<!HzfZp$$rz
z<6=d$fYLZ;@kGYXO)|Iui2Oj^A1Dci|0G;bO{hoMaIfQ^Z1YmB-ZM;}DKi3-C9s`9
zX_t~t8u@tmj~L?PTf4L1&K=m{?)PZ_dYY>=WCr&OJQ>dhe%%K2*{A1U-UA2k1Sd4Z
z{}o|*#pFnea1blu29|hH@?Fqp>VKvR*wI*g^et_~`G3h^V0$dcZ!RH*TI!FS-h>->
zws9<a5IqRp+25oU-nbeIodomsJ2;6=z5Ja(xC3cLf9E!~rsmI@t&HDGUV1hHi9U3X
zK5+vzE(O4m@*C_wWDH*Q{w(_$zu}%fJw6-wgEtW0wBct0{|t}ux2)mq1uk{{Qdch}
z<93rQ{}7UpoOlq*#E*L4+0{o^5Q{zQh}-y}ofe;*4>s|X`lb!Er-1+Q0($!W4+#y}
zhU-V1)W#c6=UBi2F#zm8J`D5?*OlGS<k8>Ie5ilt28u_fK?MrHPAzkO<x+poB-W{&
z)4EXr;UX|cXM~vmp4#|6;JEYS==V)?XGdUm0EfyIqryVUHM10DkS<j;V#o%msl}L^
z0>EMZZW7S?^I=&OTo(*{?PiXCYWAiBDVQE+V{vq+xshcC&sgI*(0;#b1KK3w)3sY-
zk4($hJmZZFlD*=l-tVD@Y0IWPt}AlCWFLSnk#c*!S^6K<nmb51?1m90zyd#g>aPs?
zZ~yqe-|#DgetNif)&I`B`JG+8>mK~cy3d{bhP(TXci}BPAyI^X!_9em__7Vy!(!)h
zd)o=i%HK$Q*R=1_y}ZMrp~1znL^QIVa)fKbgOuU{F%w6O@wU@S+c_3O_{cl(YY9cC
z-~@<rhr7j*!9D2X_k_Ys!Tde_Kb<}sc;^Q8-xT{9C;cx8CQ@=v)PFeqi&~vG_2K4G
zImJ<U?)w7>Y6ju!ExoR=R;#EpK<v`RaW@p|{#wm5*1Fz0^Q`8N#qS*<szv*Q<HfSk
zPA#B7#<Q0MRvneNro>q()g3p6Mi3rohm^x$5P8hg4Q#A(gL+!_65*Z(HdQw1r-L2n
z)u9?at@Aa6AMGC%M|l7Ko%aX%nHIkPkAW}sZ7@4+g6>=6Ga`|U2QQeWL1zIM5MV~3
zZi9Rw%)uGyQtz>rD;m+x_q4ODy4J=Tt14P&)@b?UuqfhJ6(M$u2vjcutsvEu|Mi4e
z%=&mEMQ;YZv6mnnOe_48Cu|1>aJMlf+BoBK1j+B>`T>p7q`xv+f{l;x-BI^+_Sm79
z^TcyTkXtl^W-uvva2NIVXkNpOz5JhMqqY%F{T{_WGtbX=@R>P(Ta|ua+;Zc`XThDI
z2*SZT0MQW+Wa>rDB8tp`THt%p^9c|G;XgJjZm5pbD7B9w2;WbLODa?g$TG@M30@ck
z=}7V8s5N-ugu*bePz2_}!2WH-QVR{fNGWP3C^>7jC}lrwKO=VYT6R==i!JS;CK29S
zsTnnTW%<3*jn<d9cjKhjrO=M9EB@}xd-rig?L6n4@fG*rm$dm6yZX8&K6mpzSLr8~
z45kC{8#*j2B``YCZp{qt8`}G@BpN8<(cWtBa^hB%cay6rDvbI-_`7>N*T3G`L1qYv
z8Gyr<n72k=V@=gArcF8sApLjOK(&7wM|1Du;~Xe|<b9Qon~srgcgw~QWX3HH&kHF~
zO9J;Ro)m21^_t9Fr=OR;!Z@w(1UIgf`3rN<z7YP26`|U9;h&hH)u`H?unG83b5EHY
zH&EQf0+NCt0O0Z=BB8tZH*VpIR6RzGUeSLc2Q*Q@1@uDQFG$4$;mCX1M)lr}-!%>D
z^Dg~I=ess=c)#mljnY^svWpKZLg@qTao@XvliP3<K8lr1ZzqTfNCPl@Udxsy5UEbY
zUkV(~R|E@)Cr9`%UCW=-_D|iIchx=6JqX==KdI8))wcAMyp?nF4&JbW`#XFTZ{5HT
zKX2!=e0_uGBRq^dH}IYZqN{d`y3XFh9NP&Cnh)jm4J_XmCOX@V+yQ{$^Md#Y&()4R
zr~m_5ZIPCf9`9@e$7{8yo$$HoWjeKE&5FOiG6jwt)&hbl4{}go7hf1WH8eVF^#)dO
zgDKJz;1w-mrw0ZD!~xurafFnor~!5k<>v$?{1kueZ}Q8=N>Ea(P=!r}{!HMu(RYM)
zxW4vFNuCW1EQ%uU#?d<v)dho9@RGzK$%1x{RR9j);^S!)T=QrB85xvL3B<2R6*g$U
z<RX29*ImBU!jqgffH!Y6W(X}_E&R$EiU;!*#W&=}&Y}%UWlHuyO7)a4CPKyKz-}SL
z&v^H`m0Q}x5j}{~uMGP4|M=JTYkS-4-~Ofj=Ai%n!tW011-nJXvKPu9hK_@^=7^5I
z)ed+tk6WGl$*Dbi!nG{Itt&a#SUD$pbACkFHo(r9er|&-0IOadozMxu?O-a3AY5vc
z(FjI6v^Bz`AE<0j@d>aD6hOG;!SiyWDvtT1Z-pZsZHv!uB~YLzj42{pikDziYW6wD
z8E$Mzj9H_vVgMdWP3}q>;vgLE1SU*vT?Al(I#8pnHVo|b1!MrcSr)2zk0ql7C1nT=
zZf-9JvsbDecSX@k1jDX389+z%QFM|?z?H1S91tAf!EG%~hyoLC?ZbeH0YVMes7YDP
zenNhym~~FA7y~xArfnw*@8ZRMkP=A98t^j`-FBnVqU~ATXko-acuyX!DPsaSqefvP
zh`!tSs5AQGG@tTQAIVwi043ao?`|3LabUXhsP~n9EOX`}Gt$LC48W9k5IU9Z1hcSO
z9jVcSwqU~PAa!Pq+Bnr2GooE_3k;<o+$0X-8X6dbYdWWUG9t*I;Tcg07_btj=d7T}
z90*!42m$~OiWf~F?64c7J_=+!d!vT?h!*f*dKwuBPb5GCxfR^8ZsigH2XNI9q=GpQ
z`MEjwdQP`II2(`@1JRx(3|V6t+s-*_Mu=OZq5@(7w#wivvI>Zj0emG;_W}w4_^2h2
zQo<D&v1q5hQ9yAJW>{juOu}H(K=5GySv7eih6D$Ea9O6pTz?KSGQ*M)@L*0cmygDj
z5QD)!<`q$aihzrP7=ZD%UC{1FREV|{OVF~NwP}bbU88)oMQ}!}5Q;=R+Lo73#29h)
z<8l3e5!PE6%k)UvONYk<_D4|Z00zpG1TrEV?VKTzA+D|XaT0c3>;>kaz+L#tKB!|<
z4+p}W!9+S4&t6D(6EP5ekSHm0;*5w0!qF-UegTno1y@)}^z=DnMtmKku|XYEqmN}k
zWQJ!%3(SETiM>Soj0xi@kenco@J?Zd(mK0aok`X%ZrPmY$!=I!P5^_Vk^N860el4O
zP)XsipMrVBhx9x=@^;e!zk&l6Ow`DW_)x)IJF~i>c6xLIBm!U###ppn(~UXmh#Qln
zXE!6WdSn+@p)oD{+0$Yrcw=XyzWO7ax&gtyo&yW!f&hT=t%<t{$-zwBA$;@JS2B*d
zkt(Z$f(cI!YAusOt?fYn-D^R`O_y`SL#VY*D6B%}K*!l`WJv%g5^(@8!qM*eHM*wQ
zRbT@+kP8QIG6N7RcIlBY`j)kcBRa|Tn165Y5Z5!;PHVzWga>J($_Ww*xPVj&MGy{}
zL3D&i8OX6|Apryc9BZ01pP8S%PDFTTWUJ=ypsKf*(%Ie0jfn@M0In1m>-=tw<T_HL
zWA6L?qK^WyKFUxDwv4-QDL$$V{@@;z9~+x9by^sx8%iZ%E)3|p=J<@j{uZc-5}aVF
zJRdVWBUL+)Rf`|dQ!F{PRz#Z(Gl&O++X0lYix(~JNH~`b_tH*k6fdLT2Pl}SO1WM@
z7FdITlNK{^Cct_r10~2AIg-6bm;ftdCnR{ljiR2ZoSs=k=_S}=0MBst%A}DJx;zL4
zmYYlTND8)?0U*`E^+Gt%DYKPI6i!tTcHr{o45wDcM*2*h5xALxI=6#J2yY4~gm9R^
z-PBnTjW*IsoG>NAO^4!VoFFGANSNRWFr_L`Va9*$1j+{MY3_Qh{fu?($lR<0=_r8k
z*aO=~L>MarB{<c?k0#1TI8p+903($GM+!T5=9;G`9R}IrM&EFBV*XwPvSA=}G;t=N
zkt>We5RL<(G9n?-U0n1vOAhSBzMvgCNP1&J;7N#q3FTOg;?)8^fN#^G|ESk$*vWu9
z9u6aqiwS<u3*ewJy2`9mgUm=6{R3(Uxrx#@a76VQ(QvXD?ncgreUN_;Iw%0LBHZn6
ziK6*-HZ+*MY6<pmowl1a@UxpeEj@yx-#bJRg!fEr)wwmwDhHvdO3kg&e$#rqTUq)p
zJgasu=_iLdGa{(~FX0mG9gu}62*(+<RM#=f&QLu<y}cSO8`z+-{w`me+DO}=9#1?8
zO+e4DZ$Y^+w1!%}Mw><sWUu#y@Xn|#kOknO6%ur&8&+ll&H^yO636o-p$Hd{h~|KM
zZVaXG#$B%&V$avsLIkxVfrj?j#gV|SNi`~IV{G(;_<@PCGOk_(qaE}MI6%R)C5Buz
z6KM|jCw+*uypvQEM)**`Tue`q<XG-OkANRDM#|`0m3QM-r=L^B35au89+L{D#_LFp
zn&j)J==0S}H6sBU@OY~l?d;`9YH$iMFrmjeDJl+9*u{sbib{x44$?r>0gCMMkxqp}
z?HtHn-vPqmj3@!&NT}qOU^TT4;CwfHvKbk3$cHS~5u@N05OV`AW&pA(3^W>wAsi-X
zNt6e00K+PX4+a)`c~zwtvu&i1EBpU^%MmoP<~winDX@yU$Y@jo%+WxMM?307Jlc^H
zZOEwt6-HtR_d=%xXYLw>a)OdW;vMK#E_xRpN~OBFTI?uBcDt}@C@*TC+6N*iw2Sv1
zhRnfr-Y(t}Q3?)NHJE9HwG+#LNC6knSR7GCeDupbOeC$gi%;wC0^RD9BrG|dBUTJX
zyPT#5n~-O^F{=mGQ4=XP$n_WsAiQVJmgdN@EU-+R3uXifb-Y`WIr@y2;Dwz}I_uxp
zb*@Rl=S`91IORge{pmYMJ`9EGcm~w(oV<lxkqtS)^s*J^K{qB=2f9V<?0G&8q=e*z
z<k2j61vF+Y`L5$eP6Kcz1#jRFt`J9X^i6ILM{HBG^MF@{3>(jKl$vCf+i8RVkqKak
zOM1u(W|rO$Eaqq0)KC?^4}()ti6U4)XU)JA7&udm@naE-+F8yN<sbomgNoU`Oh8L8
z#L0vl6xhXK2_%#ogsdFMUYu%(z;?nerj*=LRF4syCJ_`zII8n?cw7S6aiCxab7Tbt
zVhv1=W-NP}UV;<@0RSVqR_O=IKM+7EC0>HP{PtYb7t|aBqaW)+8zcdEu^E(&hd~*4
z;i&Ohe@0|?iqGL0NeOWCy(L$0BPG=+-eng~GMjLvFP)%%s*K(N&LU2ZX0L0obrgg#
z6hQdUkhhTe!3(C7n*m^m$J}naiW?k5UdFr3Syh-9ByKaUn4adUKDi9ZMWS*&>|lXH
z`R#YSqynb7m~Ql&&R(@!)}EFKfncWkHO^=qO4;r>*Nx-HrqkwDU6hT1Ik?6J@cvr~
z3GL$67(VHD2en@r^xyySKVJBiK}f#*-#sJpgFhR1Ht-8JfIftB>&$8w`D?KuH@M2{
z|A!Km^5!$N@a9UReEU9rRGeZU4&brCFYy1INp4*BwWS<pwMQ+)gLz}ykJQlybm#40
z3*7>Qo3zAdRz$kpEzAGZ$)9afUYn?TsjyFVSk>9<XJn5)S0FGf1AU}d-7H~cQL9M`
z{f)0wKXNL2bjk4SlPJFuQNEkW=S(kY)L(bD+3YWmDNZ#8hl{`_j_A|vuu_x&lZ^CN
zo^RR8TSR~8_MFdh&iJ9$W7)CE(U064>}C0ZIJh@-vMw>y9$eCCP=Ov+;FR5@L&Zx$
z`HsV>I)wFP=Ox+oqz{#u8v3caQzI+D&d&nMmHG0N*$ahXAaMR-6Rrma{@`WZG;@@R
zKic26xmbWiNB_1$KQ6Q|*#f1HF_K50nF5xfuB6r)Yf1}sMl*By`N@!Z9tu5}Iew}d
zZn=Jf0DwL1b)!{4MHENjUi+NUx%XJU^Lykzs`B}59gg>9?MSa**%VsiuWA+zv5o(o
zQ*W;+CGgJ5#g$;n-E&m$4%s_2_L@Iuzo#creUEmZsnQQvFnbPLI5H4_%6(sh{&Up(
zO5C$F>hIPb=_j4=KP9`T%SDxWkQs5QgOGE6%J^Jw?LqxqvCpmZPn;=k&*rm}W66o^
z@3YHRE3m>2h9x>m`D32+K47={Efy?BqTP8X)}^Eq&7Td=i-C_yZ1n0PuI9zs-(vqs
zv%F`g&xUT_>nB&*YjrCP1Yr0Hau+`}fxov6v;`M6xRCk*ivRm*wbnEa0~5L`_wb$#
zY;IupS=;4rwu9+b*iK`*?}kJMw|xjr2>ci<Ke04@R`Ggs$bNXABi8=)^9D_8uWlj~
z2xiSK(T%K$YJi<(fxiJ<T);pZ+_tD0`SR_ZH+;^>BO*S+bGCT?mPtv;dDE_QjqFvO
zKG^XFKD0q8b6l*57El`JES|{Nxk&~W0FfW4`vWDx@SlY1sR{K68}4=dlWktA)q95N
zGi63#vIMphDD6_xNh2Q*{}Drcd~0_W+_?ig-2EQyUr%$DhRoo8fhXhHz^~hYKKu0i
z%X{G9o#2FK_`f17ub3Q35e{M{+`tkqO1=yFO#RPP0XrItkG`diIR7sh3~Y}D`OPK7
zP)q%h)0=SP&Nhxk526R5JNuio!W&m(p_5>~eg`M9sh7VK2zMZ@=<nRd*3|r2vz76C
z$xF{hAkl~J(I;-8#-#u_QhtN|hm66i-k)VZ<2T&Xr^jalwt?XZNA7@IcKlpK6z6`*
zh>e7w3B1=Be@jnqFL0^rm%4f>8Mm8c`G=5%<ivweCVtfW&aOVXf>`WXN8H8-?X>vh
ze6Wc>wA3$MXKS1P@dED3FP$n(LjyLqA0ak<yPw9L0Lcch>%|&-wkx}#`J=xl`>_AQ
z4HS<|g9;RYom%Gn%8mY>Nvu;lr*)$M!bM<?&ImIBJhkzCz;Wlt(eIn)&W^zB01lNa
zMumlxYi23RAYH0x#E=bAQ;RV-1%Si+-6Wv(=fkonxGos@+RYsO)a*?MQZPNt#^UHs
zb0f<Pp0UPrp#6T=2DC}Ur)#&w9+{T0dBz(#NDl6%@YmDN#`CzY$o<{PS|a84e6#dF
zsx^1eb=VCfOn?P``qW<;^gsUb|Gec_2L1GSb$z)TdQXy5^M42Wwa?h^Sl<_Q`3K+0
zmq$BzyS}Xbjz{1tn|jBfzhCo*Z`F(C57Q5Wx7<0e;KseOx2OdXYtzmU`i<t-jUf$r
zTWYxf;T``}-4BoXlzTStY+!z~&q&)+2|n9<Ht^03=tC#RFMR^Pr0sXK!I~oc4*BgT
zt2fcDZIPQ3zNeF$2aDekbjW<1*86i;b2|FZ5zd#M8NpLij&Kc>pPNf6hpzPb9GWvX
z5`Vhsr@n8%ZchV(so@#1JQRtz+6M8NoLDgjtQkYC>~p@ffoGjBt?$jr^W8znj7w_>
zM~(0DOPv)F<*ZQ48V?vhD=%#0TQ<*qys?=#N?+Lelk%k-cpj6VzcJ4}c>6v0qgL$0
zsNsHCV}(f=M?(5NW?iaATSX86FrwelIx2aG_M`D^;Mu^14Lonk3rl+?<yrVE+xxA#
zKYrJLyn)|(=Wgx$xeK?h^AY<KHz02k^j0~;EpF$P4ZV_krC!YW$=b1xZF{TnE%J6N
z2MIq`_@Jj0`kDr{k+10|ZRziHuHB1#BgIQfiICz(i(q2M`(WgVU9xWzx>Wsplj7$8
z^c{WUeBap9_dc!P+S%q^G`gJT@o>qxcicBh-`nT~wRw|-%xIMUllKwle1om4JKnF=
zHP*cEh(y~mPtQw9r>aaSqzh<C>uTPT_*+Qav9_I$cm9`o?-*t1{v9P=)YJ}NRC&^m
zH}HWiONkE@I-C^MM|czKp=POLn4Fb25WOkmI`A0J2A&Oo4GixVb`6tLy$ss>)X-TS
zMFiJqa1ei1K1P~OP;FclrXvO>I7BrT5X-gN%m@-14{-ed>1J71PV40;H#MTfmAoU(
z)tJ?hv+873>ES&acsB4u8?d)=ET7)e>BF#p9FBSBk0<FG7HeFxZxgyy{Ym<xHee5v
zo%M6tqkrV^Hw~p*`le-lQ!h7d`H^`x@ND3#H!yzcj6VM?WMDqc%Rg`limtQ^&-?1d
zJk5bSrMGXsi!<`YJN#A6e(@fk);_<1+n?Rr1wX%&r{c4L^9}r(2i+Q_)9f@}2`0w3
zIfos4kPLPlndZQeiTVIb8wd-R<=(jB8>N<Jc}Rco2IyA?{ri9XYx}jm?e%Z}(tdN$
z|J?iCL7W&bB1K4FTZ~tB&#9sSWBFL6*NYa&W;pV8>{?S5>R^c$b2pJT6R~$1Vs2UD
zB4{1JbgEbc16KF1Ie|xra_wYSn6=9NsP@GNsYidozS_`&^cf|z8qAPY_ZA4ge`Zpk
zPU&b$-lh-P>uzs=to6JTX4FEexpe8BR5=X?QRAX+EWUmQaW&U-m3iaF=i2v8I&1$l
zGVgwEm7KTN{8371Ba4Q>s+k40z11>-M>T?bgBiPqj~@)gY_;c}ik5}D_5c($tP~F2
z8QGb>u1gACg*Ie68SZACzJb2F(uGoekR7rOy^wQ2;mx7=2WCg5&+r>sqXqKotkxXQ
zkKY_G5Gxi+Zr%4n5<#I9qP&Q=3Y9_$He&lviYofMG2TN_#G9$7C^9x7TaOp2@|8qQ
zR^ct|q5{=Ed}$LbkBD21-bdDUuwdRv_Hb5@<P0=@g?Ej0Yt~uMHD=Q=v!Ao@%Lt2;
zO0V*|<!ZC)-0*QZ<=3QH3?Oy-2|71_lF-6F`8oBY4BE?j!LY@nq$If^4>V5{wk7~2
z2)5xyis9$ym}x%kBXy}#7GF7I^~68fj*k@u_Uu&d4(F_8s<Fa(p_jaY7le*dbCKEO
z-1KHSQrn5WLsa{xHvHzevD4f#w@pDs0UzOjeic3|gBx6_s7sZ@6JFDFnV&a!uKd{-
znmSu;pk6Z5a$B+Nf7ZJ+z(Epqe-QpiNH_SATEDdj?}#2f9YA(APD5d3j-ku`D|g4+
zMb&K_6c67cFlC6KqB_WU4B1=mdHW`Jz|0L<!tXe9?`ZRDv#{10^&$)01S%XAlM@41
z>_Iz940W2zE6B~ng!kG*JMS%aRsCR{S0(@ayy?zPz|Ib0a<l1~h3PobQ&u>geXq4q
zo2JzF=fN?zhfDb4rUl}hSzi##u+w=H;E_$kguLsud`nwn6Kg-C)}FRH(X0~P&nXxH
zir23|dc~~83GI*ErvC9nB*yZas`t@pw(BqI{gwRdm48oWxu0@xyv_R;dat52`LQ$H
z=`cyT)j+*!Hr;ngUu_BJ;d_8{J{`$DE>)Ks@ZsB#``jsh@(GB-H9oWv+@;n18d4kV
zj$UsOxucO=G)|_uXJyjqiPlLc!?gC4GHi~Y!_VtOuH;nUZP@Uv?aj8^g^zt^><Sy7
z?i_0yFr8<WQ+&!k1ycIQoYS@@!j;x0kqKz369#}p)U@-afY=Bp@iBP2e{Z!-wN3i9
z;n9t`pr%qGGfuGn`~-n@6UuW}l}>D~il1P8aVNy<;TJo+Uid5%F8eGo#M}OYhfLox
zBK%2swGdfAoHDT1c4)~V(V(}VOE-DB3p{~8C$f^6I3vqd*NCBmwS+Ym%Lb~<7XUKn
zEmPV+`XQGZd99V@?0V?sbXm|Q{l2YI_wRn)*X<80ef%!HRFH1#=EC$z*v#6d+XL7C
zgZ{GD!|r>qTicDSBv`X@u4G*2R5z@`w8EPwrz&}R^K3%%+i8%U0u%uwJST6!HWya8
zVBWI4+i#A|a!YrwicC-Bncz{vS50;{$QY;RX~Z7?2R7jM!Hv&1K@Y?7<1jro6OaZt
zjwWh%mYeA*YRM=I*C~^MOWoE|7|q?Ke>CL~@UZf4jnJS+y?aoMWAc&EEiv!r7PUeI
z7f6f3v_kU}kTrDW{>T$x3aj1j@>g}gj7PWq%2N9nqTtsaShDV2^Sx?I1Jop2M;-k!
zRBpB1@3n0etO9y@B-+@i$O?x<Txl&qS)O&;2iM?!@qh5Qr#6**F@4heqkwk9^q#e`
z7M`0IhfnuHYR3j%RHyV*VQ`1o4b{eSBKd|o+G|+ZmN||3z7~g_PnB;6`Y7GnSyyUE
zKZ+Pz*>FleaHiG{xW}Pam1L^p8ebEghMfS=>N$`1FlXCMpCYxQ(HG81v442z2d|Z|
zbI4%0UvB!i2E~Tj>|Xr-uvlST9Wtv*6;=%wF<2my?zKV5_FRGgEI&VS=goiuD3Rn*
zYy)4>P)@E`RA15BXsuE6&z!Q+J%xcS66rcHUUpf$PzhT+7{qpHXCd@vh?+YraUAtj
zOv4dR{)mC@)a}K?LeG}_hT1&lkL^6ne*0{vpS-9d-33}pj3*}szR8w*xq7|WdEQGB
zS*;v-Ngz9o=Jzl}D7$_5^EJS~&Fo{4HN4(6%)1_AYX4n5->W+;dU9{5HfjEv<+R~*
z^H;X9@+53LEGu%M&E9_izt!BF?cOo^rf#%giyL<wYU~8^KMr;>Pq_PoNJf>HG4l9U
zoOCdSP3F^&lhvIXqA_OhHzV}BgW9hQ`k(*!U$6bjps*Kt*R8BdcUiBrDYgJDJjDs4
zg)2xd{WR75cG-SaDWoHMY^W`ra4mV3xwGz9O1T}CrG^VKE9C_7D8SS4gz6=t5}1|(
z?r^<nl^zy@`g7tC>kPNpkR0lazr>hwsX?LZsf_F?MTm5zH0g-#RwGo=Rwe23HFhBS
zK`I3acR(AVvB~_AS1;Q;=vdX#)GRm8C|jP>JhypjCCVK+yf%8)3PJ~FtOsbK>yMJM
z77|Onu}RXorgbeVRO7+)5~pf6VbEsR@69*@<!S<l<f8Y3#1zmP%ZC?S&Va;Wp2nLI
zO@;v|{@%caek6tNqWo#o=3N?L=v;2p0FG|#XOJao5A$6gvoocNLaZx8N)@2zkM=-+
z^A;MLAbClYY$eku^oL)ZO%(ewqWZGo5I3Hklo-29jx<74QkY*A#w=oMPh6;{vQp54
z5<TV6dNu!0ya%_+(B!%_o>VSfEL&UBId$?_OUmn0(hzwz0#_<>B%nK{D8!5z@u-FL
zQKM($l~>zlnXaoYW1j-4x7Hn9rY1q#3$2k9f9Z%&sR!ojs<Bc|WxN_23yF77`UqFm
zrqM}=M&Kp#;S^L&xK}$^$ny@ED;^isMq6X;4>fHUexwMfUjxqRlF}6Fl!%(}5tTdP
zepVEtdSOQiDo)~piREs{4H0bmU6w>rHgPfdtrS#0W@^v4URZuJgIA@*1-5wHmt^7<
z6_Kg}W7<VXp{I&pAxAeG5H*kIP}Q8)z<AGy<GzmB^TN#)9(Uq=C3b~e>Z;fD+Doqe
z3~FB24eDIF7sGr#<p#2<y=!gHn&rtN8C~v$WUsugWPwswq^`6qIbN+9ZM6zI+Jy^9
zgB;=9&gqts0#Xki2)!z@CbHa5zRFESnL9YiO?-a<ZCkWBs@c9ene8fFDHJDpN+U?@
zsL-{r6$)ZfhzQ?my>JH2mrZKwcyQVgH9M4Ic)*A`{2|7OC4bbFgi-j&Z*rALkwQ$K
zgblv$+?-NbWuCMUiU~N}BRir!>(ELg`vMYrn7<bk-Bw2SHziX8S8DI&F#9aUcM+P@
z96E4%8;<_yiwsc^AgKL`=E;yJ;kAtn4+(4fzNNxDt}D&iy<Xwg&k%Kfl6@9GN%t_)
zvE?~&Fi_q=E8QxW-Y)vrcVZaf<10YHY^>~!Y3>Wn$!6H;ywX4{Zn((6xR8-vA*Bg=
zd&uMDV+CQW+RC%l9>4a7U;NqS7A2FzT2)adP;#SbLF&;`IwTeasu$%bgAe5hBuEUC
z?d8aDM_BbWb~rGGo`&eDJC+}PTaMv&o(i84x{zB3Q=X}{I4@#>w@RTWBfiwnAQ7l=
zB<|QTG1??cI!8lAK0m}v)@FLJTPKa3PjMs_vyXVG3EJuK2jl`+)I8`$s?r_WQ1{4+
zehUrJ$<N8(<c*zQ=?!QoJlrV1abqQ*Vzd};Hb1l$>nGorcyoGau%g}&y4TfibPe<*
zE%kTS#7Y0OBKpHmt~hj+*R0KIe3@ehh-KL6)Cus&rlCSI$7(JQy-HTzX~(Ao5jwrO
zM&MgfHz{`sO<@9c?sZh2P=9~Q<Fg~jkK0)S&349lTbvhrU41Wo7;|u87ppud@M8h)
zoF1F$Ro-;oX1h%64ZxD^Y^gv|p6=~y5Z-9(I`!lAUfaJ6`k|)nfdO~;`qF<kv)&$|
zWuyc@IKjySgTe?0oqu}Du^#z~jUmJ5&Yj5yUD_8P!z0V_IsX({$}2a$lzCWxR+teL
z=+kD3b=FP_Oy4RpkaG`vLqU;fvR2uM;#-9=YAo(jE)1z$Q{z{|Ou;kGjO_#$q~+Kj
z0tcn;w%$Cc(V`RZn@m=(OL%2lR#~#=+HiEXR}93=5htH2D2I!=qag`5VGBe-vWXx;
zKE)L&lrl=NujDGn#;RhXrHmllxY7jOd2W&06FXrPZRGTaCQ5(U5AtkUKcq||;(H$J
zjyaW|iK@zniK=ENcFP1c&D?&@)E>i@RSpaql3oFl!5hRlA2Ir>O~SJVwcsVGu+aFN
z+24<}_Wc!1$C{Dg<j^KgU%Vl`*exR$y=axPRDKswilm{00w2^_Q7;9%K|_L*65bpV
zA-3XC_+TWLJE#;nuh^+$5nAkw$TV5`hvy#pDu}ue+s}|ty3(*Mw4@YZF=oMX=!||A
zcOf%)P;>fe^E(oDPcgw)su|anf<Z!Dsa^Tm6+r?8SBo9GNI+c{Vo*bz#6gl9UaBZh
z1!Wbl#%U;B)1Q*d>D;_tD05;4AecqzD^RQsCd#rh9}k%WlgH8>CAWKKWv~%4V6?nU
zN8f8uD@Ds2MT_hL;-SmsSSd;>6A9KoU~wivXP4+!j7LVskG)p9cr~^5qW0var=kk>
z!)OgTuJ-w;HDo7u&>U*j|GFzM&t0_ondw5OR4};G(gYNpE<-A&&mrGOTFV!_S`$Mb
z+09;TSe{F{iTaGneU<wislblf@}O`kiPaee%v?ze)%hJY`u;YnWeQZfK;2LjCZU*g
zCc^!iW%x0nnL0tN0<yqJHNR=}?TE02E}z3fsLcRr*YGATm!;>Rb!>Xec_q<g_@QtH
zp{_56Ub(jOP7TgdB_{XgBYLW{MDST=b}M#&y}3+4w#!Wk0iYXRzxdG&H&1wqwwdJ*
zJ&0^|nac7KJ?Ib_08X6Q*?eSXf%^~{fOkRrkT2?w98gAh6hsUQxkxue+m%^?Djild
z`dE+hIMxH*vZRBPib}p2Gr15fpgC^|X?t1XLDF&haTJK+RiuQb&GL1PjH<`VzVoC;
zPR<#^i(L&_M{X`>o;-)GS~KWHZv#2xJW_mHj2d|TRce%As^ttZZfnJLwc6y<`p0Uc
zpE|w8iET&+S~E8jX}8pURm2*)vLRWsDTGc6$fMhZj>KFw8T%P2o!+~zy(nKEg4c9n
z_HIx75kvxO(;UmMX~cH#x9wfUXOyp7uhE`2#z((1w=9-f(`t8Ul%4=0IOxWhcIeZ-
zn4sxV1=%JT<A%v;cY`0VR)C7o-|&YpDUD^aX!bZOrHVr}@{0>meJPiNqc<jN9Kd7t
zwttiEO)}R0Am^@bQfMpNG{I}9*S(g0mQ_3{A>|$8l>?PvXyn|S2a)$c8zKFX+joAQ
zD%LI5=#+c{?5OBH;bSK+w9wPKrQ!3%G)9mTvw$W}D8L_sREZ4q!V#3waaK0<!y49w
zo{~tJ40J_g1#xLu0ea^2<px3Qbi_E}Y=z}xx}Q9W;iN3xaf=pXwkFaMznB<G3?{3!
zjFT5L3YMnNk#otjihT$k+I~{gd0Is(YF}!yrZ&Sz=({BnYieu)SmgD~Txm^H+g~9Y
z)y^S@SH+hhK&}M&FrI9Op0VPkYiisc6gxpZSj%ucNCf_l>0Kl#8f58}O0h)Jq3{Gq
z4uV_RBznnFjakWtlDihQm21`K)JTr@Ix}W%*xa-6t(u*NkLOu@<RAJ?Og=94Zps$x
zl;gt3(OU5)6hBxuHT9BAd+yL5o}a||i*7{RNOFWqw<l7r+SHJ~+%O=Y6h5GU5P*X;
zVh~P4wo?yI>hx|k4%|kuCsC#(PUBO&?rYW1C$&SZ$e9D#BTDt+F=!$;*yJu8qF{Sf
z`2)7X8B+G#_ZFZbc~rUZ4cO~%V7GR8aoo1tRJvsZ56sa$jH`kO_A3z8VPXZknberf
zt0uV<Xgm%wfs+I_pYym6QxG8$)FQFI+J{4qe35P__0E5YAr=dnN$2$HP`50Z6WgUl
z_=|9ATK{lG6^wC$M9gw2N&j*(+3b()_$*`K`+RQo;F3EI%guqw3(QJ=;(sfUn_udZ
zpEJtS$2Y8^<!4U?ORz3RvWKr8Zx*+&*Tbx@2%gS%+$Fym>B0lsg-)@3_kBf8kl(@1
z=?3TU{;pn{0&IAUjQ*I&Q%Scz4EIX?^B>BXRk(M`C}LJa+iaAUPt5$1Y?jDDA+pL`
z*u<xki&U+S7-@1RQbB5KGwtA@F@)z73Twl3z?YPOVHDunbFdfEcDM)P9X2rwEhD|+
za*B529myj7>7dS8t5_Qa`0JnlNq=R~|N6)OyZbAHnlAs<Q+7jD{Y6EZbM!@AVp%qE
zDIuP$GKermHXsR_#e=D4RuV~3wrY$G%F6TBjT>oAIixz(F^dymd817ciup^q%(H-;
zL?lZXa`m)cp^>OrYd8kPHj(LMpk2HO8SNDcTTEc7!mL1*43SZuQ(T+^{&JKJm7-{y
zL|)vvtFXtd5Qj%?F$x=nPgYP+fOh5GP$VP}NFhfx$UfuN2UCnHe9Mc<U0b-8=3Hx5
zB!pxbkqEaWJE@z9E1^M#+S!0$>J)Mi-Z~XRgcKH02@pgUU+fU!b`?Y7%;*=(Nzb71
z?d-~eINl8Tt0g@^8QKwJItAI5VO%>Y)Pl{Wi95=G1U*;66j2@>1Tv(UYX~k998;j+
z#hl4zz&cUK`klywIJq^)Bi^7Pe72;%4MBiVETBX>cD|?u<44;tI_Z&|UW7}koiW91
zi_xhsxq*!!L#}C1u4&piVD?~4JmZh2^wUrd*eQiZZso`NX4_ftdE5dI{kG^pXls{-
zEI=;Xb`Eicb+>f1KXPzhq?1nc$##IO5NDj??d3eBFi~&|X{Kip9=}W`VrCOj{%Q|e
zk#gG0X$E($$BffYwdhb*k10fl?Au62olj$oBBDRNSuO+ItC#og)1c;Xlp1q>i^#pP
zke$NfL?$<!LiSa@;e7_lAeYs91yV6sqFP6#rGTVUgv!yy%Bs(YhGR%@dPB(HA(6xg
zVw!Lb5~in_5lCdlFOf=EdMBF1ct#DAm$O}yL5=`?vME@<GNDc6NHgfK_+)!TM%6Ik
zd?)BG2*zBwDIB3VhfIqF7*kP6bccA+VP6nQ&0(f06w{dNK^{Omhy}!MSB~*GmLmdu
zWfob?dhr}|XcH#uhBVRx%3;9B)#(`)Bt^1S*<gkm7DCBZy4Q=o+mb=zCZA7==YsBg
zUFERvLTtnf3F%i-q_pa*=h`Bf*5PhZx`@bOa@tsw_ZCE~F(ui1U{_(X?aTNaw{q;X
zi;?0<Ao=apEpZI=yspr!Hm*&a^i)m7W=bHF>EZ!b39jF4rWMUd+^#X7u;xby+e(x&
zPUDXf^7v&2-51hT-#j}lreW!T1t>#Ve3EUx0fr6d;97ZuC*J~g^7G71C_)F-bb@p;
zU?I*4O46jB%gT0XFq8Sx6^wo5?7hkZmsPE>om$wt7%FnuIFch?yrnTtHsADS7>B>t
z6c;9N0wu{89ny>-ETf3otgC$CnnRwkOep$ZM#Qxg_1pfYFU_7{1rn}vHjLo@Qct*Y
zO+|c;j_R`~nWC#IouhyWN<G_orZH8C?HFA0hZ?>4*|IfJgq7OCWkpBOSVANS(G3Z}
zSFLMwA-OK6u;=|5gR;#YPIBqhPBX<bWZ*}ZB{Ej<R8!_<a+yJ(WvflDW9)>0aod>m
z8kmw=p$C{GiPAvx0u(i#akxnXsYP#^3=S|#sh+He9~SW2)<sN$#7NB)i7I~tlNd+a
zMZA0($Q&!(G&P5^QU#_+xRNu>mAD#<8D#ZIHwKz**>+lr`SGNf`N4EZ3|U6A9q*<=
zKA!_oJLFP)6e5UTmfJmt#*q0Y$7f?le8IMKuFcto#ANBGchzDvFH=L?F@<9zp!cf5
zA;L@o7^6qIy<D4$8hNyhWN|xtXWgQhznWv*>U5qOF*ovCSVMX54IOBuZ|TOhvN=0O
z&9<{=^nqBlNg$3^ZDC@tXViF7)O@nkb<ARR(mR|ueDCP~@mEhIUP?CKP9~7kUM5O{
zg2<H}E<w1DNbu2@NgQ!B=7Wx0)8La?p(d04q7=y>)Lxnv5;da5Xl2tNJZa-8$4^Q%
zEOHe0x^4u%<yowqY}&<qP4PO46Rm%d)m#HO)Li;YsNut@S(K~A#f*GOVWyG>5l%V7
zh%|7hvk6l%qS7m-sM|o!Gm7(=%pT>%%tTd&sD%irdKQs5CLqg2F2#l5R!?u1Ib)RZ
z5I5d<2JIwaMTs9?e2;t8Q7y2CP;+xo;<7#|pnD<6s6x_kI}EpunooZ@Ghl35uH^_<
zMaCdy7{-$78%`%NjDt>L^(!ZlOD{@_FZ1oBSRW;#9P!i?NzPb8gNfq~8*L=;$~C>X
zjnT4AiB56V;)oy?F-VenhWiXjE210#GKJ_z=*LWEnFCLjZPDS>htJzd@#I)Db_T6p
zwBJU<m>#%Vza9L*(=+UU+@ZU}#g)bRv?PA!#n3cnvM48wx0P7>GA$QrfJTYKJiYM2
zX*z~Yp@S%@Ewa}c-Ae2Pi-15?pmn3k;cPw94DuCglbTJ5&cufzq-6R?iSV-A3^LL#
z^jD*Lba9`EC|bQouUcG@YLVt3l2#!_sjz)X7Nt5wA!5ug!TBn5H4Tvr@>IhDdg1`$
zS9rtb)*3=p=K!@rD*8nSy$mv&LB(58H)*J`#x<D=I=RjqWOa>E$67|tKxe?BZ8qz#
zrJAj+`am5Q(dHB6^A_<bru-?7h?Z}5*Os9;N$U~Y_||9lwXwkH>LY8{?_H6Fb)`e;
zO=_4eH)W_hxVN-^)LyIS{hV5H%Jr<<m~COzAM1*qa{41a8E(@C+%xp0(+bf|%!O-N
zYn<93Mr_Bl7>ka)Z#!Kw+1Dz1Oaa3=V9s!o%e{Lqf-o8yWLoeCnN}k9>|$kU(UOdK
zrL2`d*LZc-gig;Fb?Yx1yQR@h_wgbj;@xh1dT9>v$)=ZK*Uz@J_$np6v}~9;+`(E|
z$|`kFahtK_PxUzV@e8-{@v2(|dEK@#&7`*Fi|(a)6f}D{`Id(;QHg6bHbR{67*}4C
z;Rr?B3uGtQoOY7wmw-deIE`SYYbap8#YKw>>8k~dlSDMZ^89q8<wrX@5ht5oBbI%;
ziArnk7(Y<Zn`aqepk~gmsTE4_2S8RiaY&qAy-2qs7CD?_iJD6tNhV2khFNs7O+xKu
zzMYn3emlwJSHODSu*gQ#lz}Y^P^}^6Ib$yJ4X0BoHj8a7;Db3xM*E7sLX1V+fDjL+
zMMLGT?7=aC*_y(&#Evv=PME${asw;mCcJS1i*34OOrF(_3N7-YopU9Nt`V1J?Me~m
z8D%JCa66ewC+W4m9HD@KQz51joy$lD)48Zc1J0^J+w&|yrRgsVkZu>@DirZA4hKKR
zj%`vkhTa2JWn4p4TzK<m%6*9S(NF0iG}O$5k*8q^48Fty7!xcr3b0KVdf6$a(G=&P
ztj{6}c}o$um|{JY@4%F6WAy~1%(WYbXzgm&oW&GnnNmMFn*_pFbA->rWmV@>eq)4m
zOEriXL^nUiDKg?rlpLn^i$^Fbf-Xy4nvxr~6%AP26(vQbj7-_aG~3P*H=oXt6Jk80
ziAHY38g8Lv1Rci<Oww6_Xl&3$5I)>hBl=47)MoUs?#Y1W`b3h)*B+LtrO#3p@e&Hq
z;aIR4=SZ_mL35mMe#qk(NqWo-Iw`8Y6BJ5`T9PsvPE^X`m)UZvM^XCnmjc{4ffjUJ
zAchqiIqVyuon{)&zAICg)2mS(voty>)xO*?K+kr_UpWaisxrb@7SWWLXn7U8(W7ou
zaoh5HQR(9hCjraJrxevN;83^W32~d1?4q)?WSg!TW+yX%%-|eqM3Xs-MzR+rh?a!{
zA`t<MMRa?4qyeKn`j!LENC%V>j(N={H$pL|<FpQbb*K{?R>2dWNcLsfoY=FwHB&t|
zl1ztAghx)Y%LD=~V5eoN&&L=ujwd_dMoI~em;vY=P;Lx#o2>4+9U+bzF9oVxB}AU(
z4)j!4E2p&XbSr2WNT=Ci2Co6~GR#iHtCKkGi>YdR1@kSh@-0T^5GQwfl_0Y;B5#~1
zHKvR)^GBYmTA-a~_JO5C$0aGdnL;u9qKI+r6m}q4Ku+_VriOoDxF7{aHu?6YiGeh-
zouS4UH>Gx($;{|XeOZuBf0<(^2aM?~HI!srIkQJhLvn8^qJm0p36^806D+1D$BksC
zZy^I+)#Y7QjK=4bo|kd*C}v;g%W0^cF={r&(92>mzi|s!x(qs&8lW+cL&meyK&xT6
zofMBY^Wnu*j#5_Cn4?73EUFqWp3ZNE*>;Kqd3`GJW?#5|HJSp0%P3(HkG8jBf<_K!
z;%BL@tfY2WW~x~>5^It@6zUNF%xLVA=tVm#K~TE$N6Yf<lz101tABpELJM(J&(6#<
zuHd8@d0t}-P%ZjO1a~4FZ6gFs%m4@{0%Ql@O*bSy&{9q)+-NZ|XeYRT-)6drhg0T{
zNexk#593;KIvXZ2`q!V<&Zl@869rb+z9`<5<Y-ssvnQ~c!m^q|H!gM1Ud`n{CZf75
z$hL~J6!FC{l^%^56b(Kn1>a*VCW<WEE=C*tN}|F%-*iq+(~z@W4Y22%{n!jK2sEPH
z1w(dD(%9iIbM-EXzTRJ|74J9K=KF85@{eV?8aAB`aEu?G=0_~tLE-U;+>qx9hukr@
z;z>P?$il3JK~V<?9cnjHiN|j`+LyQ#YA=F0BpcC~9NehTh&lZ+OisGW7b%)OoeiLp
zBCApM)OH!I*I@bU^6IirGM1<(9SLiWQOM}Er2=C^ifIs?<>)Go_yu1noUme(gwEJ<
zq$>ghj3EjeKHGwXE9t6SZnJE?C2}KjN`ke|n6zdPwc=EXw;UbC1H@)pPwQDLSgf+d
zmvX6rDo@)Qc#TP9N5$l=v&liSc(1~y4eL$nnM90V?sWae5RS(vL0_Ut4uerjZN9M`
z=Prb9AmuVE%Cw*fu;t_)AUY><%<i+tX>^&NXwD4{ScUSW1bVS}I2jDS#T80?;@TZZ
zficO0RL_Fcyu29Xam1plP8U%R%2s52Qt*6Zv=aQ+OIi2SvA#4$2frd$vSm2Il_uR2
zzX6JIH=+?yvn5~HgDnCIiX+OZmPChKuFhDyv=wL~IocUv_NAu#hKF-7yNJ%rHE7uy
zVXb~Kyq%H|fTOZyj6jq>#ry%_&g5Zeq~Q-eC-xF;^OnxIOX`gKlP=jM({Ya~XbkR&
zGKp$m%@|)Q&rH!Ggk1e=HVLCI4>SsXi6V@%2$jxttPQIzG}^fEsAO|wP@dF7MB!I6
zI=vy1?I80+gy@pS2_)?=ma?Q4)--`7DU^vdq@YlGBKh%)gy>HBOM{u??R0}TAT9~Y
zUzR2*bzQX^Bup<pV^**!W?yD#Sfyo9;jm3+Px9n|Q8eJqlpIy;$vsC1e5(_iGBu`T
z(<0Qqj3vx~Gp99U#5@3b{g2)p)(}3ky@&?*Xu9+gnjP#b%VQq&fFoou8WW9Xr?Hhw
zFo(2!qUDZ=VI_=a>G5RVbTLL1h-y^Axbe)7I;n7$AFA7u8y*swr4R?&gKyb^CVIs4
z8z_!(qvmLvPQ(c#Kb~w8t}lx<H<+$MEN_<EZY<GIL$dliTP>AWu6UC78M~~~8|ape
zQzi)5u7Mv+53@mm7z{Q)PZsxcdQd2xqjaH{`Yy{s!3|^Th?^fthJ{m1x-ya7cJY%9
zjS4PzH_*i-4!<<p<r?TBonjz`_+=bD8J9N!g@>pkOym5S!#R<lS)MF;QJRqGkGF?!
zVrMj}wHUG(9d)ul*O9_Og{#noFnZbPmc|I5zA|7<ij2nE!EbAFy$(C7M+A>)fS~wC
zDXc)arm9cb^fH_c^&M<M7LO#e<LL=O(Be|WJUdC)?D32NibdaS#w}KfHKMO@EYN(e
zUf`xLRr#ZSte+aUKKL|8GDyn)QZtq(r*ga=q3sX<<s-~ZbIMC9SRFcxFDF@yrAdHF
z$Z2_j!tn`behQ_UmOtv6j-P5|zKOSB3|N~KX1raDL1+6zu*AeC*I!vncoLL6%k(oP
za4}5^HgWvLj<(W3vg}2cd_~l-d{T~OS{~~(M}2y3+;QO%`5B*E)yv$a^kj1sXCJZ%
z$}I+}z7tzkCGevdH3pjRxQ}tlR5eqzlk!}S?S_~b`Sc2qh48sWmRei`EB1!&;^<Im
z&_!*c@vHOs4c6TjZ4Uia%uX(+F}5i#AfKH_KZUgbzx_o^Xq+HPnj0mEIrt{;IK`R6
zD$X|8Pwz$e!)HjNDN~y?mrjaA-)zm`m~rfE9GW@5Oon0<E>#oAa+)T6XmJX7t!C3p
z$TWa>auUw96sOu+gDr*~qEK{1N%@vXV0(35w<I!zl}Rjmp~0hy+MH70G7CpZaQNaP
zbeTgVwfSk@B0kBla-@T#PpR5Fn`hcW`saV9Um5i8|M9O)tG)j1U)paD`rmuMJ4kzZ
zj%NeU2A&N(8+bPGY~b0zvw?SSz<z?0pZcy=bCoOr;7nTcNJ`}RdZY4M{d-xl{b=7?
z<xzV!@ND4Oz_Wp81J4GY4LlooHt=lV*}$`bX9Ld$o(()3csB5C;Mu^lfoB8H2A&N(
z8+bPG>o=hPb($P&|J@q1)~LUJJ<kBo2A&N(8(7&u|41~;&Xw%HhVv}tuesIFpw9-L
z4LlooHt=lV*}xyZf#>JCKl=th_xJ4n(uR0TA77OhbmOeoy|~D}Nqn!Bhy85e*}$`b
zX9Ld$o(=rk4fxLzZiG<?tMEX1%x4472L7xK{IzfSG<b6UtQ-6rr=i>N1-E2ufqezX
z*k{)qn;1G7T-+BVqQEYWZXwbcTP5Bqcg!vQL2Ph)wK-7AJ>aj}fZbI(&E6T{O3se{
z-^k576)$6d<5r%TKfQtHnfvs{i+ayZD3<be=+}PehQ|M}Cnt^*?YBP*WdZKsJ#BQ=
zZk0WpfiI??$v^RD1J4GY4LlooHt_u$z~2lhV!zzr`{y|7P-n!W9wmQSOosR)M)|VM
zKJ7dkcsB5C;Mu^lfmb#_zdNY?%Ao)KkN@v0eq~VmQGtG17=B_H+?9Ft_cqawgoPlZ
zjnYGATLRS(n~QV!pp6!{-xM{n2rtqZ>NS{@2J1TA;)WAz)6rg}np|B|8s)h+RbQgZ
zZQ$*(j&^mfsk89Y(ri9?nB=IYtXr!v!7}9L=oUQUQO^I$i#&0D`(ls!aOq0OOM}^P
z18DbPfH_w&6Jobr%-7<K%(Y@`S8X4{I}s&!wH5U#XeQ0oqO>pid$oh(JeD<WJ5N$<
z;WKSbV9uif``}_GEphmo@RhY#LUl&P6mO3C)~<6sAja$Wrs=piRWn^SNL~nuqa-^l
zO;Sz5qXe6bQR!sDki+;-Uv%o9ZezW^F45l`;@fK1bvo0f1R|;SC4ZGA@(Jqs+3Jr`
z#?L(|qC9iZX>wznKgZ~^tiNU`Bku7-YQpi6<(XQg>dBrOaFsD)ey`w`h0f1*v-)Js
z%ASQnr%jS?ut`%*vx`$9KH9}jPmz|DwSd-4ADoSmv+JAMn3sDs)_9ItXa4`s-jx7q
zF}45c-Zm5>BGDpAs8skYIaji#B(h6Nb<-jhz7Qc2SxU=SvV6A?f0FnT#krwHl1jEh
zw|(Dk`=b7H=FIxOGw*$unR(C5{BF(s&Uwyrp7WgFdCqy~op+Yl`sTc;E+buw7i9pH
zaZm<`4~SMzt0U)4Nu{zSAFFQ^$|<!a77v;kmS8&Crq*X!`A-FwQlm(<p@oo&in2vt
zlQ2|l%RBa;121KAYl=%0_pAk~Ct3|1^@Q4rW>D2<S<R*)GTAqTdQCN<SyoKQHMBoo
z%u3}=`!5$V4aFMoqLSufS&fzQLUpgY#p`fm%oeY!=x`0*fL=*XaKpTL)i>^vYEsK6
zS!S94sn!hv6Dv+J5qFMj#@W%t2~l=k`(xeE{uJBlPL`=fyc<ei_)aM@q=xS?{;DpR
zJHuVqI$_o4>bxv2t*FX7wF<Xm<cy}&lmWbw)<wNnszBB=6|qh+efiH)V5lTW8BJPJ
zy2o6~v+7=H%<O^MGHcwVuJ%?ptCs0XdVy1dRbWOGLrUj(O0l37a3#fr-T=#4GCZe=
zI897PNojR8OI*#AGX$DdMZ(2uTZ(6<X5_qT!BnDES6WneyjYW%3~S~&nhIhnYI?F*
zRn@!2#>MofHNsG}Cxe=jQC3w-6bZ{~7R?onv}Re?<Mp@#x1^>u&3IITRu^kIrJ61^
zCAfm_f~i5>S3SYBR0U>1uBv32NZpt~F|HWYbme0GPf=AW^2&;$VOOoEtLq6nJ+)O?
zM@caixk<HEz0i;<CB{xOsnt*`>I!O}tLcs;nb3`wlrUOhDMlavZ`+Pi&U&jTm1H`l
za>EasYHFbjtZDU6&+*}u_lf>Y6%93W6?59q$Rwo%hQ^~7gR1}5)+{-5hNqQ%yjb^5
zFOm~1t>pEDuBMe}sHo}6Jg1eJD_LZ^nxerYLF;2Y8fJ{f_(&*ub!UJ&N=zVc{J*7n
zW>8U=A>p4=uIOlXRHMv-YC|``W$IovuQ#Kq8};QB|1X1@V3USPSytY&jHbj3Sc&qd
zpq610Ijv?5sgW}3a8-S;*HSdpoTZ*6qiUIz)n?6VPHA&(TOC=bc3a^<)@R66wLCYW
z=rOrgsOoT$uEP`Ro?B5h)QqMOlVt<0Y+AYKsD~i;n`b#WtRu4f@}_jk)DtFF|4d0)
zqLQkBT369i<yK~_RW*YQO-`vUFau_TidTwEmNM6sYKjf6%ImR0lTfL^HMv+VQ?x9N
z>KR2NX2<FpW@8F%vWC7=O}8y4n6|vrt?6aljMi;cjVbCHyfW9b$W@UH+^m*2<&319
z(*GM3jB=$(X5LU^`br%x)m&0NQ+3rEx_i8o7n-fBwip$<SgFg*^4bb;do0OQ43&(M
z&`V_@*Wp#&*mILgYs@9aKl6c!72rE2Q^iK3Oo>%c3K%LGmXrUnB1P56Fa<@V?kpK|
zft==ku)X<@D;XB`N{LMAN~T0I`n+LNC6;6rwM}tSbB{@t0u?o7uFp!8lunMKZPHaL
zD=MlPMN{)Z>9T5Gv84(Wkx@%&Lm3v9lK*(Qtjzu^e_YN=6cbuj7Rb6>iziieMNdwv
zWvZ5(<0cf#iYZNpHN*-PO;(TRm^!b*Olg%_Qpt1std^2ew3(WcP%zUSw-#^39l1MJ
z;=MB@PAwQ32~tj#Ww|m=j#bSSozi+`vw|-4KH=1?lRQ(8Qnj0^ey0A#)^}<@Re7vk
z<XvXKX+Uh7DM`7cavP(GsyLOoCD+_>vZs|NjaEa}T+&9&iuNWfQX^xe%qC1}S%LjG
z{qaKe(@@i7cDOBNZ|s-?FHr1g`np&xQ?v{{SyyR5Rp)6|T++rgO4iX`G7@UrN@d+S
zwLv%Yib<t0%}w4UZ=V;aUDp*%ZZKEmKdx;cml}DkTvOEYNuj1>m@}n#p{Amswlj)Y
zif`(1l5(_Uc3esG3X1Qhq~>#++)}_YdL=y}*HMq-OaaS$PztzA4`wD~1+M*>tnnmo
zSFz8MMo619ZA!UcHU~U%HD*Kpr>5kj(H&l-X&7>@qR5r1W@>bUs=CfatcQyGy00vu
zX_%3v&3TiaCXOrD1x(raGkck8+OT3u$qw-UGH?cYLKQRrID>4@(^?IdxBh7!sZu$s
z^Q{#zneCrR!J(d6L$j-)xy(LotEmaKuG@RYrHbF21ADb3+_rbkklGqng$}$Ws*+)u
z)>$)VXwtrVl3ml(G=H>kG>j@iDi`x$mE}sBmGiphnySdC9LSU!J6Z=z=mTegDXC28
zM42ku$xOwr@@9m-oH93(Oq!22H=*gLYDCdbwUX9Ya>%T!y_1t_TWo1m#Y)Fjv1U=v
z>Ir6qm1+VFb&*jjr+9lN9i}EXX_9D3j>=x&88SJgN}0nf$;%k>oV-_(dQQn|T5<*U
zr)FhE<v$mi_lgy>z}>Xoe#&OeHBBSMy3C7I1+9xr&i`u?O_7zzOt3=55O+nZ&Px-y
zr)wrsN}?<oJBki-llz>6RX0=&&y18CcI|Ai7L<)A<h}X7T7#9ze=6`y`1blJvwkMg
zR|gd(ROI>`8BJz;N+p|`HX@B-`yyFtU*IIx%(;bZ@^TN1o@x>UdH2~==QEz?m2$;^
zDrZh<#@O1REDo%Sr9jCU5<ROWxVolb<h1gV(Wt{x)|@%7T2d4YEwg=Dsuq~b&1tiC
zTslTkrdVN_?Tmz+QpB>dN)@xClq&|gF5fHer~*d1`p5NIZDvUqsur0SvP?CmTV|??
zOy4O2Ze5p|5;5AQ5^hK@*Afi1w36VqWs#*sQ?jUX>n1(D)NIu3gMHduO<~7gjTu!N
zlR>GZW?6|+N88J&Bu(`3BGoddUoohtDnlYyVMCzQVNxZ-HCemN059V@s-6t2B*VW%
zCX|Mh=96i}u%vWK@t^t4-Q|1MIi)YMvZBv+YEKnimezaAGu#H#V!E1^l9n^<zao`E
z{gZW=P?7Sq)~2GaG$8NPe=Mcg;Dt=Ah!hPurx;>;>z`7EiI^F#W>rvxMh{{Om3pdm
z1>zLa5;&Q1lBd;(Yn8<GDslIme0Gv2w&rrFqZ5t=ZaCnnaW@=v*G9y;>k4^EbC$HC
zD{e*Jh!fqZ%HxT2vJj)`q~;`#sV+j+$<<WluJy&KVb>Z+)lgO)O5Awj(&RgumK~=`
zHgX%0;+&JbB+t{W$Lu=QO;ui!k3N;B>b6VzvKjux!{F2OokRWlz=7?Vy28DZ)a07b
zKt`IXtMJDw8ReQ0P03uY*42uX2DOY;&pxL#$lKQ%WM8Ma2?k0vwXWHOmR1xrD>|}L
zZi9iG<Np=QMqah9l`73Lnfy;NpqpdWH7z5hDygbk4Tju^in2g8ngS{<vTFK&&4w)D
zUu@M98Pq&?k%_g<w&@sm%)Mr|JpIQ#QyJoZs_yc3b>A4b$|ftZ-Y_lp7w?`bQ*`1>
z(~;P<;%G++0@XXsPbF)}xaW%Vy8n8IbV+;*+z;Ms-gP$e@}Cj`47p8Nk0qHr;XkgX
zDl!LnHN%QhTb432x>QzE{;L9B!i6l&3)PfTk4t!+7`bkW$+e84qqfQGT6$uraksQS
zs{K`JXfn0BTqV7K%q!zhv%x&IA%zK*BHk08miNqIdB=NX?U&(}T2pI6!5Eoph=F32
znNn0#fmW_GVHi^NSrw%oL-|iJ!S+glBG6L2REZ}=i_4X^*iNp(o8V3=8Fgo7K=Dqg
z#O<jqnj6%-h(p5@)`nVNMLDk)o3%6<Q?jHLfoaDo@rvBE(SRmp3UXTcugiGCWXCQk
zv%q?-@0Fx!r&u=SO-W`>YlFGO(ng+@T12|QnozZv&}4uoRci(qDN{8gsg5!8k=c_i
z%E<n+tjTk&+@!CSGBu{mYV*BPUo)gwQ{-Gj(NQa?ikc<&&~Tl5r8Uj^rM9gUXdZLN
zd6^-zr1T7PTJ5MTQ~!(xOht;Yq)urTWYmo)GHZ^jDNL(+Jb+{^OP7?q$);+S8#9}<
zr1gxR(7iE9G_y=(D^?4XGF8FQ<0&m~<TMGlZpz6?BNv;}@)~Ns@hXZ`-ZAo-gyp%+
zU{K+&q|q^#8^NQP(poe!F)7xRCQ^|D+iL~7Tih|ls9wgKa4uoxs<Nq$DQT3)>S;=`
zsOjm*eNgsXSKn(@O&VHKD`Kc_iipcKsZz*Ul0{s~H7u35D_XItz+19pR4z;w99vgv
z>y>#z%gU%F4K+=oO*)R;302D+DspkmA5&sRxOFw9=F}plqWmcXuC%D^bg@<(BZ$?s
zR5A=|wM@N}HJP&6fIb{d92<GuvUzV-(G2r)#V(^sF(mIau_BRoJg@FKXf7%_wll;=
zN>kDb)e>DnPgv?}+FEU<W!BW>I`YC86*<c!s#Hs=SxwAMXc1x}lqvB4stvWGR^PC!
zn_@)@b4o&OSxxhHOqH}+nmrAZm1}ysVV2O1SfqMK<l;CDO%O8$)Cl0EYCRqxW^3kV
zn5v@A>*z(gTrZU+iYWz|o=pwCx;er1<hrH`N>a;f$ZT5w%w~`-nE{<ib<la9?bQ-Z
zK@WJPz)+A={J&mB&suC4QaP3QkGmo_sMjg6N@9cQl=86BYBFD}|FqyVq^!H@Kjygl
z$1P|%d*sT>UaiOKYi%p2nbw4Ip8u(utXzaY#S89?*(J?6hFUr)cUct37399Mnk=XP
zW0ttR6fG*mY5Z{(^*tAJFD=HEHgq*Pp|qV;9|EhS@v8C6JEHep7HetF4j)@in>spj
zhFf7a3~f%IW|!AcjVT$HPV{FTJw9B_J4+o!U}|40WOC+>f=W`Zr&LwtY8eB~8^eqy
zR3xhByq;o27AuCibzZJCBqvq7+^8Z{Dk^fhjB6-+rlQ%Algb}M?vf%=#agMVFK4xG
z%Bb34^>|%wDVe=tHyP~=b*zn<bD5;gPGw@pzOFhHjBSNqmRERInpJXU2U1KMZ5k<N
z!Zf(b@LKXl#;T{u?WJ0&8JVWM*;dfp;K69FWl5VlZ0<rTX}x{LmQqvQF_1S$A1$8n
zOo`eihb>8^NmY_W+N@&41NT?AYgRI+b%iALjk-pUm2uJ7F~mwr5$m+7g-WBEB?EcN
z$TKy$s;VuclGMr-v0BR08SR-_Zdz+aMn%Y?Qv0XK^c<6#{`B@03z|9Iu40YpsKt6B
zHqQ!T|EadQDl1g}l!R7YLsrPr3^Y`$YqhM%)wR85o9VFXd~Z%EM$HZ~75-1rH(Su^
z^8$U(&1qT1O(mnojEPjMrmQ3j%*E!kQpR`OYpnrp+mbV9cpI7*Y`ocj<Bw-7BBPL}
z91G0_EG@gI?zD=`VJ21ftaC~gBX5%HLPaIkhnPICr@k_%eT#cDcx4bLNH(ZAz>xnL
za!sORf}&U}dL1QUUolo1o7ZfZ<ff!tYRzftD#jMZ=C#^!a)<gc=2OzLugj$It(zrE
zS`%pKY|O62p_YSKwP0TwCyf)>IlyJks!k<N@&xwGRm^FVTe0HUyh&9_#+Anjv>HwY
zPI9fDBc@$ZwHRAJHm@~omYdUBxm}PdW%kOvnzfAQtTNTSt<Y$?#D-#;b28^7kEyOU
z?4a!+Rx3F|GS!@Poa8nOhkaKZy-K$rMUPTcPuVL6`wn8qShjE1(Tc8at6?kTRbq-_
zvR3!3h1Q&{wZvYrQadBgV%=U`!oiMe->Edtj75-6t%P+P3S-1ES=psHi7JSbYwkN1
zXvQ2A?8Wv{&6L)rfsXoyik2!_)lMTu7L(O%Yv@#*s@!HAcoR;BGzWAvE9z-IA*-3w
zdMzU%7dvS<$&Jdfskj<(0;Pt%)X@<q^;8{?wG*3<?Tue~iyxLuG9EZpa7&&l%Z~ca
z1#u(q<a;uFB&`$QNW8=_=1}G!X7v&jm1x#b&{EFU>`UyW3EXn5P+~{yYuZb-?%1KZ
zhB<AgYa?^2;w0BhBq&#lVl-m1nw=7P(i9!)$BW}7stLyet%(vb-hwKLFEH71EOaPv
z5SuKfnRGC$i}ggTeWwDe+?tEEq34~{Omb6Fuf`Hia*L9KnnPVXu^}^3)+QX}4q`ix
z9913qnNZ*AoSrl47%9ClKH;cjE-<I#O=yzjv@~rE<7>rBn6AB)TX86J5Nj<vpiM0s
z5v%MVPN%@$7l#HM#ED#TsA4O&6Pjh_w8d|y0#)u%tcZD;DQPk6z_f$G1pHgAT5`;o
zlff8yOja>tC)4f67b=o?F|T2go04&5x_NUVu7bJHluoQ?Q_4j)1-em(G6%6U+qyw*
zJj@o%X<gA)sMXfBQcQ58Y;QJiPIH6t60ISp60_W#*6LZ&Nh4OS=FPx=ak82#F&ASu
zCTp&4Ulc1<^07kIjDy(1xVc5^SeQ!8hRkVMB_?YlFxj!ww-%dnhGipVDrQNi0=MK&
z4aC}y&FgMD7U&w5gp;~eZp~Q?s2L{>C%JATUV%%T<a$Lt5wDrpCmVem5wl?@wv*W^
z>eb_G#7p$f#3oWujjd|VJDYJRi5JI9)CS@U;w5HZ%;}Oe9m_2RmW;`~liW~IG*mT1
z+ejIjES0KUt8J}k&1pI*8aEq}uRbvSSu%#)L9BMlrqo8ntT~9C3LMo9&l06lt?r<2
zCpO7U$wba%P!`jImetT|S(9q0m&OTriGx@vG?$svx=lS1U#d&=M4U;zww$nP86~O8
zRYS4Wta)orZOm5aqGz%8)qJej;*CS0K~I8|V_k;=2eI?en7J`kHYF|IImuHTdOI_^
zOix%^Rx?h87P%$k<eF8<)K=;RN)t}ibh(}|+OiQDg$^mZI(9N%T`$oSR<jObixSgu
z&@+q8X`{ZqG&zfg6-Ni03#<*Ba&bwMz}%u?I#wYzuhmLX?xfDl8D;usyCG8A)3O?B
zPH5;PH<{K{IXYCZ5!;AVQ%)sLatCLe)ScvJGmbi1ZygmJ3LM1dHZ{`<TJ^0tL(xfY
zXjoJB`l(1&d#P11w2hS2F`jcMbr9QGP-KR_k&0EeFR_=#*7ZJb<A;q%^TW2tRw&nS
zOU@gQwHA|)$tGw%*0K1SYHuA2>>aW+=~&mXK&30bJYM4HUlOVo6;sBPRLv_8FOHYk
zx#Os1HlAGCV#<;+*H68G)lR*ZQw_J|HZG-N+Nx{KX>{33oyoA#H;GJ1tpObqmFY@y
zg6pZ;PL<4Zb6PPJC(z8>7T5}H>^SS$43kZL%?TMDOgV^66>XEY)og{yHXU0zHt%eQ
zIcTHJ#7=)nshgEdX`VC*SR*l6qg}@WLnS_CX~HIB*I>yFm8=o(iAj<|^?22Ii6dn;
zswOYgq?3}9T(0I+<Ro{jptxXP%U<f-xRbt<T=R*Qm!zY(Y2G<hh!e*NG$VEeb}~&r
zGjy=5iB;6H$+wkyjpS{cHB)TIQQfJ)NpAK*NheZtEVC`J6~?va9GXt`o#alI9P}N;
zPF8H>P8L#;TXnUZiVQT7MqF%OlRKfaerhF-+KvS#kK%OV1hTT7jM;FKI~SN*c2aUO
znyQ@JS8H)_&Y1mjOq)Eg*o&=T(&I&nAWo{3*-M?Nb<lSZyJm;k*V@of5$GAc*wkg7
zwA~x^%*8R%n5?R4FEyEoRfx?eP&F0lNvG;5$emR6Hp~fKnJ4tDSto^bg8B)H5;WjY
z%Pcn1Mv7b1@*2ugQ_^VFR%jI4S4*{0jJ`E%&8bbt3S;wzhfXC<a-*hm%D%3>RI!qv
zD6SQIU3;l9Xhyk_vYK{FZnbLUEcPuKMJHaut2)SS#kNAj1D0}9agsaNus3Dc(NY?k
zS~rUf_w1#%4y2%}*iThobt;8wDX1pyah#zzfl=Rn2y7X1qQTy}y;QbgEwJVczntWT
zhF!{`mV-FP6=qP^)Dv1Wae|Tx94lxQW3hw^HpMm~Zq!<!=k!EO({9Ukdvb!O<M*-)
zW})g47sa@zWSvV)+V)EJQeG{#EH-a4o*?N|%PLoMW}!K4vTq|ciA+h2L{nWgX;F$V
zHX=QyOs-c#Ri|Pnc}xRZ)>++BV##QA6Zq($Z>|xm5}Vgr@mj96V_;m1aRPHcw0eq?
zjiy>|Ut%xSOlFSUzOvVwk`oqzTA-K9t(w!;oVCQ_U|f+Rh?6F+V=3hsCkv^{V|r!N
zl9PsQM`WRs2Tmoa>O0x7lRMhB7Fcrz2TpRwJ~~x!k~<l))wdPe^;+$KuCFKbp2bUI
z>#)3|mTre9Z1ikIimpk{lO}=Hs$+qy=2Ye+cdVc`=~T}_?jUw-#Yx|x!(82VuUayf
zX>3@Ebj$kXTe`%l%W8dHFE-B#l9lO3ViR@-6dkp|Ty9RA?U?1}v{}cTcGR&eu#+)k
zPIA4W5*2jKR1-FqEh0;X88-w*%G#(~a#K^5cCD?MJu;W1>V?*tQ$4fo64MEcI2ENT
zH++bx8I#rZQkBaUovJt&XstQc)D)6KXEP2ZHewr5Y~xNUvGUlwEe~-@YHO*=<LEaG
z7%8JC7MUeu=<t-e(44mIjlEDdq}S9F@zzaRiJY*M#h5e7oC=)eM$Pz?T8Bf@29j2c
zZ_p}nGGxe&6t5YZSERPWG&qm5ll<qp+OIeZdVeh?mW-~FhQ~(qC8(UR$ke)XADsM+
zubh%39kZr7rlj7iolGy(6LDIGz(_G|Yfi84TwtkS$(Sm~CT;X=M6tD!%IkLYgteN3
z*s5g8#puUmxnis^Hm{jWP41+vS<^6qo?Ve9i$&Y6*ehu*%cv=sL@enbc9iq#rdp<?
z)pY6wN!GWj#W`eG?4HcASF8z}LK{&@&Nx;v7nsw!LG?o+zRV;sCG~nnLMzqKwoisS
zOBt#bx#I_vS<AMrt<YS{xlu=NoV;|DJ6cM{JIBhg1u@1<SyPgCJx-uAqLpZ<7sMnS
z${oaVWnSV`?j(1l)Tu(Y<T3PE{5JVyO0pU;S<MI&+Z9=5)|^_`R>;a63LL~v>S`+n
zIMq*Jz^RI@+*W8bpK`yLUye3o3u5!;b{*3Wl^n!LI^$H^xid~R+>$#v9&bNh!nsrB
z@x}2HO9e;f4VQEadcso4Uh2%HVc23OMX{ZI?unB>b~OwSoNF;1YpFG7*icf53LVRw
z3Y_FI4(Tpf^-VcbQdgD}4q`{Sxvr7ct>_7*%qeG;8)iys)v=z*glkEyF%`9}$-W}u
zJ6>v|V^UX=NvJSA<Lxn4PLix1+iHe+qg5lNIBD256_}Dnd5pSIplCZ)G1QEduB#`k
z6d9U4WvHnoaRRMIOhHW6K}!>-K9wBoJNCv&-|)aG#a<aSqJjE0q5L}Xn}a$e`R<^%
z1~>bi<VvFllpl55b!&-VIP{#ScAk3OOd2$t++gCY`!~*`!1Q>SUFN1~y$|mtQN@eq
zHTb7&Q85(;M~}^`blL=ZQAn`Curjyw?Z1Ws?Rlub<wfbY@p~xH%0s(;Z`XM6sij18
ztgx|N{TFh#Q6PB8Bo2nXyl{*(dM^zs2JN^ps{W3t{q&|?wRRJrJ3NT~z2n23cdhP!
z%mN}hRy*yJayLx<ivq*RL;1to&z;<FB?$^HB>g*->9Cpv1;e(UyZ!w)ET8vb8)x*#
zo#PPEc)WMUYd`#7qxHmVjOuyIO#6L#hoV*oyKW;u@PIf7&O8wPJn@051JBQ?@+$>~
zi-$ws&3>+Wl|95STu(bqsQ2(>5){Pwe|OIRFKaWQdBKg-C=fgZF%IhGRF@7{-AI7h
z;i2&OYP4#r4K(a{<hX9n+|Z_&c#VkT?8LEGyyN$kVE?`%>T%I|Ki*yA`LTmZrYLIv
z;8$wLqoV&V?zexXDo#7^jyXFiP%IuEd~y7oPxox2K=6P#z&yY@fOP=t0M-Gl1Ho7a
zUK~5-%?WG&prT>Zhx@duUGy^nf`=f*f#}zXd+pz^KXsh?=yyd_boBbXKkWPS`wDml
z&wl)l7qvL^`}}TL+Qaq0-d+!U|L7j`?!0Od1t!PC8{Hoo)o}3+3Urr;Y=1A^yXe&y
zW_MagMaOi%2jKVO)P64x8((3+{tW$##&w2%kMsOJgFl};={FKK6z%>@(f#r2`bXIH
z_IlPEM>MKC<FqXl2p*yk2kCx42wT0NKHo8@p9ewxzI5T5A$uqGpF{k@usrt#{c!jB
z)IoU;RjPg#3_fyWzcM+6QO7TzIlt>W8@CW3c<>?)ZrDC`(dkzmra*srC~o#alQy}V
z2@pJ(<G`=uNwnkd@w*e>qFwJRnm?DV|E_iGN~H>eif<jeb<>X@uAo5hfH=TBz&eoL
zI)MAl{k`8j)8}8HzB|tm0gvF(jpxw>HI7bPc~6}Ve-@Ic;yXEY)}PUCI{}Kx!|BJa
z%YCowK?-!2hp64><JSG4LA?$DetUa88~9C+-zC>PIi=32Qemoi$4__tbkkG&X{s2u
z>xI#ua9#(ljVsP)rFT9Hyn#3T9^n7)0eN3{-q&vEauO5_`}J=XPA+)nl2KBjUx@3R
zyt&RPs^{l|)^ES&SxN1T9#DSNZP%?OLC2t;S1-x$RIk_eO5d&XVz3j8_os_KPF|mv
z_G%u%Zm{0l?$vmO_5LedM<2}V=!16rC;I1;U;dSJd|y&I&VL5u{AbX<-zMvCrth6c
z{hOxsKD?J?%0o%tLs5?FAH#V4qtpFhL6LiPKUT8qceeHLo^*HKlO7ZvJAWk2duIAB
z>KEdE7k}?}5uNW$kF&y3U(|70v~u_71sCrmK(Tnp_HjZ`+V|>s&TqS3oi`7vUAM3M
zPWSx<qCck>{yn{ESFPP7+dPzX-Z<Ivp#83(`HT0dB=<g*sPGh)-!(;ht`Pi5?|rii
zN&gOII;<u@dmh^Ld%MPiPc0>)V^GdB!%u&oQ<v=VRkrlp{yqbH{JvN`zrfw+7sT?}
zu~WsG*B-T(h>l*(>#)5iBB=erbHu%Qj(AwUr=cIoT}PX(_l|1!-DewdT<-62Irt7X
z-w(bxe$J<Rw$W5k^y5YN7mVY2j8kUg<oD-&RL@IXG_S!wWs8bwRPoSvv!AP8We*V*
zgZBNvtMeJLcGJDiCH6cK^4^|@1Ha%G&+84}^NNFZUNbD$W2f4WsGi4IPdiPh_wZx_
zj5ZIc&KI}zalKP|*E@kX@HS!XA5`=>2>bg%yngN$?00+ShBn0{T^=qNcxjt?^>+~<
zct9Lr9$+0vZygBQbFoDIodSM|{W}uI2je61@0@<eXO|lvtyy!#DhdP-hy%=nbk~8P
zKBpqAey4jr1@$sJALTcHZv9RJJMKJJHrqVjIHFPA8K-Tbz{EWCyyBBn`(3qz_=VZN
zZzk3Dx9*=ZxJsL4Bxu-RSeaY;_FqH8hPZx^Uq2YM>#g9gH}}1k#H-);om@0}Y+j|)
zCQzU|JY@TMaA-fh=knb<<Fz0DuhDt}1P@`3gT0GhePMQ|brcv)9-?|ZxnKP*JiZ#O
z+G+y@ip2x&BZ&5W1Q>s@$3NBce9#~K`P@mr5wCgh>vPkqcEb8QDf*S(`>D0(#)ln)
z_S^~ho7{8syyi*t`?j3?!}*M8p3lIz2YcKnH<&o<{*Ch}5Ii6bFb}W}q_+;>e6v{R
zn=u|35Bv@kyx+k-+<DjP{>LmJK=6>wI0)+dQs8&TsgHhFM8N}iNE8PLst!Cqr^>Gs
z2p$jz#yk-HJ)r6O7nbMqJN^Cwd3K)rnCa(7RNf1Ye&RjxcV4xKmJ|nJzb-`7{^5G2
z^sZ;}iZ|R3=+6CsQRi{!IZy38^}3lf>R9cxPs-ge^)Ko+M)kdCLA_7Y>+APmUySQ0
zF%JHYgQ&-~@T>f&+pb$nzys$YsGnQ#8_z`u^K%i>jqj)&f2Ye2>S5i$x)I&y3Pojo
zL_e!bhpTQR-~l`!4x&2`aNf(^^Ilj_v7UOlo?^bbJ72+PcKOWqdkoNiINnEraZGO<
zgSP(RcPacn7v|q5asD0W-*Nsuy64|9|1tkD|1tlu4qzSdavkvdJay3G*>8T(9)36Q
z@^=Hom%H(`yMC4P>%Fv(`i<xPcz2EG#||bycX<ffc~JP9-QTU!<v-o$k%(F^T^>*`
znDOiNxZyc$-aLmbTfdK{`#$lg+=t)g#z$+`9I=W5GvmSQbqV(IdDSL&GeI064iE>J
z2T@-K3I-p!v0s^-LSSMZyn5bOrgnQ?@yV(EuG&F>;33FyknMHY@9)*aUS2px8oif>
z9XocaSo7MW_L8t-NzXM!xj)apeJ7{R`ZL;XCqVFkIB=c^M~>_E%nfad2@pIW4&0ju
zH*BA}==7@&Q($^LWcocA^~HLQdDO?F*B5I~?et=)!c6hRmG{)?@Mj?b+Vg<-y^8t1
zSG-R;z4u9{%3I#oo%gjHx|{$x4_+UC`Sm-A*Zt7!2km~2sQ8QDb5i>~$L;rfIN6Iz
z{AJqD#kZe3_2nmX3W4eIkSYIu)z9?%jbHV>{@f1A@z;9VX+ph+ClereNF@$3ecvhS
zi}k*39ABn(d<nji;|u30?axzYix<DIYeIYJJ--pWrN^6J-*eCop5x=?b9@kAiQ~(!
z>-w_wy*OLv5AeB@-sg@JZ`nG|^Sa$-Z6-7?xN#Z{3TA5_C);jVp8GSV_S2ho)!I$M
zhV{3+DE&5m4+VmUAjbjDyQX*E_3{oytqykGMu6ZUigDm{{evj-3LD$ie<61p1%d~2
z9L#NbRp*Z;?WSI1=Z~a$&rIJ%gNhHnIDXEjd$v&^ct9Lr9$+0vZyi{)a`)#27w;rM
z@Q^4DhMx1(&Qq_ONkv7`o<p*b^zTrn!)oFdhVA@hQ2O)Mv0FF&_~8l)1P_P<ap%E>
zYliHd*nbX*I)>$ZlHcz$Nw&SPJqPT@5sm83IBg5{8?*hqyx8{fULtqjO9VcH$!EIj
z8{Qj`*?R-PUwZsy`#mim?!0St|6>-AprEMNZ3X4JJg@!q`#22kW#{-8yo2{_eD8`%
z+%*_h=9a$w*HB<)Jf!>Hv~1TaTd(!A<M!ngdX1u7PZsM(w$1})>vy=O5BF(RyXa>k
z3gWqr*5^9n_p@O9einOP6nVU7&X4!Z1?Bunte;W;y)o7PKY#OkJ9=EOi2}g`;sEo&
zt92kK*O$T1aNd9I*0|n3cSQ9cH|G{c9i5(6;g%fVbKLo!6MK9dl)PK>RdhRUU5}n^
zJM;Gp{(SDF--ug?`{Z$-ya@NnUo?7bUZv9}(6C{)-y<89-#ugd>DBc<sh;<Neroq8
z#QvU#{3qpZnEDqDI}eyAsm&Ac6>PqKIeSBgzcy~CLBX&+=T?;KCE$lx*RS|B9%1{Q
z<9C0ueg6a6?>P0*?}{jRNR9`;)|FJ-@%s7GqS?p&X6fB;CMs`XJ1#|k62DIZ{8;(f
zyXe&yW_Magf!^?d^DJVYX9>#h>|yh>{Ne5APVTpo1O>hNJqUL1uJQcX!6a-veEH1z
zUEkTbg#d%W1D@j=jpw+E`aKkWh3mY6+vAz+`^CK0o7`aHtot|4qd@S0IKVu}_BxR4
z^GBgq<@|at?W5oUJRlA*4`SDW(F4kly6w8P1SlpC+5VhOwS7E4EVJi_xy2uTfAZ${
zC((Z|b?g4VbnS?KKPUVP=ko)+e$FuLPiS6n<1`vp9`N4u^xm5;3UArIzdmT~i~3%|
zcKzP2@!(TS2@pIW4vcvaw(G#qpJ2TQCDZW=>is@m|E>=EohH<KcrpPG;31oF;P!Pp
zQL*>0{k_{v?UP$53gSJA!F-P*#tq|kxavlNI1nQaaQ-FO=U+-yoOa$Fb9T~fp;yl-
zhuvtu2O;dwvqS%)@!Ud;bGXJi-Rosgk3W|j*Q<GRy&8B0uM_*vA&3Ki;~;3)gLvhy
z==T*2nsn})3-11h0>J~~0P{ffb)e@JpPbt7svQIvO&+5DUhugsuj>5Kq}|kQ49fG!
z-1al-&&8~;v0eQaa<@^SyF9oZr?8jWa}#l1(cSkQf=}@2=KC`dPlzWG<EimcHE*u*
zYnj4q;RORPZ8NX_E|P7;^-saR{wdqz7uNfvv(>L*FE1P;jowS$!l1vu`i<@njcT}f
z2lX3MeNT`m`l3FMyVn<MPwn(#sX`!lKpbEmU>!(r9Vk4$8m-!D0|jQr!;53byg6a*
z9~1~45C`$|z^nVz4}CZLx$0H+5TKO@(XStK@-J-fH9~*F`M#>4jB9bT51O>e-AsVs
z0de5}JP67<5jH=)y6z>_=ek25?|lm6d!L++-ys+CeHa)QcgE%Zxg)CoxH-2lY8dtF
zon!ki>T{r@=2uYGwT_)C*1Yzpy#xpz?BXD5^ETD~igG<n(Eaf0dvUsUy}ExMcKdfI
z(_u9M5834*-Fbm}sol2)zJkq{=<hjrC#TN(Gumw@*~S}3G^#t}v@H~9&x2^skH>X&
znO#@s#9y|K!*M*7-Qy|nFE;;o{B+k(H$AnV`h}f8lIA@#eHR6~!$Z`clY{<m-Sp#!
zD`?Pp5cR$_xBS9$N&S5;Dfo7mZ_&T+`{iHMt|Q2_zh2!(lCIrM`y*QYba_C%aJ+}o
z@8=@Kws+C!v3ZqFn?Ql!A&ha5_jTue?S?KVq9g97Pw#&Eu<;hQ`$D4DAHRRE#C?Xi
z&oI$_hIoD_p5H0r^E>^H%dmdG@ta4qhxG*O$!{cf9-O%Ho;n@=EF?gC9vTcQb4%a;
zYbemlL(sn;1^svHSpRWeJsjuNy&A`CJ#RiJ?PmLV<|Vbib^nyXRoW~gZedC71*P1r
z>txck<Mw;fV9%TP&w*F)n$CL-!}h&scl|2o*L!In1%`u%px)0Y%JZ*-dc9`0?rTc7
z-PrG|$cN+ld5j~*(arlnf_`1VqLsTpFSvLo2?`!LuG=#=v?(UQ?DF9E^*3lQJI`qd
z%6T>TiT86p_|#H@IB+)(Vt;>+dOqFnlWx}w`tRBBAMcNOe(Yd^IPfM8mR$4XlsczM
zg}~%^@awrQQER8#X`ht4Vd`HLm>Ca2|6Na%>%xQj`xE>Q_H#g;#t+Z;%Ix`G%i2t6
zUU1_y3Iq>`1Iz=g1L>^;^nrq|{U16&fmR;8x?Uja=TXsrvF>ja{kbjhFBs3)!8l=@
zg76$D#3|x58*y5H%Zt))<M&V?cyJ>QM1AjOg^lg%zmU6)W(&vEetOfcTDu7_Gahh$
zI<xcB;4d2diE@7!{224*gtdQA@BkhV2bc$LtpibiKbv3g4M95{r#|{!5%r3Ppx!^8
zuHTazOq_N9#(5M79uNnZ2UrKvTL<twDm;%W(etPdR2_JJPL*FN(8>euTle<9^`PGu
z5mo<1xgP|6;C*aid><R)IvjC5dO-P6w_UfE0Ko&|0Q11_bzpXxo2K<Xyq5r@$wRu&
znHRO*{iA!#yYs3=G+X%m&F}5#als}U7Q}I8IF2*@8pq2!6tz0obsGsQ;yigY&XWiI
zxxZ2I|AK*+wwYId7YQ1=_1+}darZq`qVidId^K9N)drd^49fGvH9y_H|J<+RC$xj(
z?QH&Df%~a4yPpdDfj``ri2E|Lb(|Kpc1N_8_}}op!#m{HY0;rY#}=gq4R15NMGo8N
zACrGv{s}FlXNTukVfnf(O0_7}@<5m8hUZrslGF8H!v{al#ku)4TD59*-C_Ap*|m^l
zdxM+(PI9GEY70P4gcA=cO#m>E76VRk1v#fPQy$V9ln7PIlwZ_$Cr;DlF+!n=u7Uwa
z2GX5BsK+VwiwDO6L_t~%_%&{5$KQDZzTM?Jt;vN-Niz_(cul(F5mOA?IHNy2fWbq0
zJVb2zqq|@L3`Cd#d`|Ge2M^$ZF(67D@X^2?41fVJ00zK7I2aJ+^A&#Zc*>NAa71Nv
zdy(n*dsW}95%J1D*mY<9NR~&p;sJKR02lxRU;qq&0WbgtzyKHk17H9QfB`TNRR)qh
z4+A|OA>aW#Fa{6@hy%od^E}|ufjt-i17N^A27>ZAoGt@F=^y;$BkVR0>BbG}fq^J9
z;PyNe#a@)hG@ZVSGH&pLj~;jc4~zlC0pb90fO&v*0P6tO0jvX92e1xc9l$z(bpY!C
z)&Z;oSO>5UU>(3ZfOP=t0M-Gl16T*J4qzR?I)HTm>wur@0B0L}Q5nE`z(>Yy9z-2)
zul#bGKiKmoF2O5!Wek)Q2b`tU_a)gyIp-8Sh`~eZ5sn&Q01SYE>@X0t&yDP`<X@{n
z8xQ!)V*@;Z2gX3UalqLN{yyFQqaGjol6VMygrEmuVxXjWMR{`b8a#m~#sK0V+Ht^p
zkA2V>$aLNYt-tU$TnuCyzo^gS0X%>Q#sK0VwK(9~!QFc`-oa%Ao`;zMui_ncdDMUh
z@W2=dY8>#<!(KE7f*Mcon@19O$czWK=VQ9#;kF;J7fuG!jjyoO3+Je!S5apmEb|ln
z;E@a-g3UwJ*CX`bo&nM4wY}GnX3Bu*;|~A$2!jXkz!*q34mf+*yUjql<AHj7B)|iB
zU<@D*5C@0@V;=Bmz#a^MfowA1_vcbJ`<l(R{pK0%@zDkk;DIrKI6xdA4loabwhr)F
zi#-?s17H9QfB`T72EYIq00UqE41fVJ00zJS7ytuc01SWuFaQR?02l}t13~-UF6{n>
zE4t7dFaQR?02lxR(PSX($1CU`k8AMY4iA2>5*Rmk#tl4zN5%l+0C6C49PrV>9t?m1
zFBri3&si0X2QS7NHo*WG00Uq^Tn0q@9D+YQR=|U|Jcv6Z;TaeJ17H9QfB`T72EYIq
z00WsafbUbBLGS<`7z2m{#6dc7z*)l{3}lx9tY4ge@Bkhd17V8;KC0M*0WbgtzyKHk
z17H9QfB`T72EYIq00UqE41|q=u${+3e|Us~2k^icKpY?r5C@nCSO=o94)9)L4+g*h
z7ytuc01SWuFaQRk$pF6h@NowZ;DIrKI6xdA4*Z)3Jle1a17IK<4B+#Qvko4>17jfD
zall6bdoTb7zyKHk17H9QfB`TNYzDG@T%BtB!DbuYg8?uA2EahF45a$`0DaCZcmNNK
z0mK2~0C6DhJmAsk{~qfC)&;B!;;svrpO~MRpO~Lm2e1xc9dLgg;87*s9`haZ9rHc&
z`R@13#=OP6#k|dY-eMkM9$_9OKaY5{c)!QI#Jt43OnzQQeZF|VBB2E^00zRwK-AYw
zQTrdZI75Ge!GPEE!7moP?g#AikpvImfiaM59B{U<2LoUL3>XX~yM92Aa|RxQ#e)&)
zNQHp`%)emGKa3Z~i!qRD9B|gK2Ls_`Al3N*eI5tk0X#4U5C>w!0rvv?pfeEG`4qHq
z3F|nZUpx}M<{@bD4u8P_7ytuc01UXnfY+a|u+K*cJh;n)8?g@?U;qq&0WbgtN@4(?
z-<%=vkeG*(;t1td1`tn);|ctLAI1RU0C9jgz&yY@5Uq89_ZE9F00u;2Al-FXq&~ox
z=rE9Oyt-8{I!t@pQ@6$gcKA4f2k^icKpY?r5C@nCQCkQ2toME&HC{{VulJ1Gw@{K_
zDCfL^2k^icKpY?r5C@nCSO>5UU>(3Z5Vm!I&uZ+!K(HAI+q^`7csz!chhWdP@OvND
zxI@3fy{<%Me8ai|MP>ZZPpm_X0mK2~0C9kMfOP=t0M-Gl16T)wu?}!AvCotN#CtH}
z9pi*?VhkV-f*l9kd+fnLG#SA9#m5~yfCt6^;s9}gIKVu}bRFQc2>YNgkm-1%J|AiD
z03H|vhy%m{;sEmi>j2gPtOHmFunu4yz&arAI>4h7dr=v{`pHKIJb(wr0O9~~fH=TB
zz&e0+z|D1lvx$9B8NhnLM-DuI2gU&60C9jgz&yY@fOP=t0M-Gl16T*VSO++(*n<Hu
z00zK7SQx<Z0v}2603H|vhy%m{;sEmi>j2gPtOHmFunu4yz&e0+0P6tO0jvX92e1xc
z9q?lv;4EVwRR*xm@R0`(;DIrKI6xdA4loa}4qzR?I)HTm>j2gPtOHmFunu4yz&e0+
z0P6tO0jvX92e1x=dmZ4t!yXKP0WbgtzyKHk17H9QfB`T72EYIq00UqE41fVJ00zJS
z7ytuc01SWuFaQR?02lxRU;qq&0WbgtzyKHk17H9QfPpA7;P>_7QREnX1_NLK41fVJ
z00x51fZxYUL2pmA(GEJFvHpw3FZ=-mU;qq&0e=~Y{aiqv^B*-H{Ea%a1_r<Y7ytuc
z01SWuFaQR?02lxRU;qq&0WbgtzyKHk17H9QfPwTG2;1*C=`(?vA~F!R_(Xqrlz|8E
zz!*RrAP%Aw2Yj@HyARrY4KCC091MT~FaQR?02lxRU;qq&0WbgtL|`Cj--CnVuLxrs
z1|Nbl-cEjoVGPg*FaQR`V!&y=ft-&McmNNK0mK2~0C9kMfOP=tfZaO4S-~C*fB`T7
z2EYIq@PYvxe{oj9gF8HU5u<L{#Q3^1zS-u{jaYyUFc5YIvOQ1HK94Z)03H|vhy%od
zn{mL|jP4%uDID`Dx)F??g8?uA2EYIq00V9_fa62XXe<wI^9*}nAPfw|&J*PMsDcOZ
zz!*RrAPzVK*n@%SF@W`tk2!b%4~zlC0pdW!IN+m#Js3!i0jy)3C-4w#9+Hc8Xn}$B
z7{L0#c>@pNfid8A9B@{#2LoUL41fVJ00zK-e+;;NJOg`t>|%Ks(O4bR;eCg9$gk6)
zV~hJ*lo~X=&F~gEY@dHj{&D#yw2+=1o?nIK>$WJ>BB$knF3%0me`|2FNl%uQa_cld
zc2c$Ps9$(~-@dD+)L%z}h7TS;v*pO5VhT);hlke|G~GUR3ymuFt5<l$fmw?Q&>J2O
z&)#*(j$Lah(4L2j|NN@!_idy+pp}Ql6GxqVF!w7WItK0Y68?&Lob%fANB{l#<{}zY
zyn5e~T935bO@Lza(5Pj>FMl_kMuG0|AbPw9<+wJg-#?<#?+c!}xA?b%yJ%4GoN=$W
zI-%Ka8Wf~A&t5UL+j0^Xbo-QOSk4dObK9NIZSd$VkA9uEe!Xt{?+>3^M1iUC5Y+FP
z@H^P&2{3+{ji2bpcb_)yv+!WeEd<DUIP{<S3x}`RM}m%iA1}Fm{O`3rzxfGkdtv#W
zhJK`XJwniVi^};tQTvJG%gl~1WBCip?+5TRx$AdCzn%jACBE)O)Z;$4KDR~fAN)$~
zd~r~IH}lI+#7}DR;}>6Uty{1Y&F^~{cZ@skKf(PsSO>5UU>(3ZpsfRbeXl?}!MaXU
z^zn*4|BJl4=hNNd^QnomF08cVO9BKBhy%<6tOM3{VEOU+uPvUukpNTULG<I3p!+BK
z^_!yk=hpEO>{yRu{rY{yshuHROV`?au$YLB7cKf`YM-`?X{w0pQIfkJC9FI-#XaO=
z{tk?B5o26Fe68l1&$e$NVM9@`mlf5oR6lQ_pWbnb+q^yUSh?pKkC;w?;DN<~+wmou
zJ<-3v_~l<vKd-&MFVHLdH(xlu{zoeolb~U*x62&WrQ%lN7l!rwCHj@w`Gv4P2XpA{
zYx~^3q?kk<!#cj`ml)4)?Q`(>4?A|+O@RLLfX}_mKKH`P-=|GpdhW40%Ly<VJdFF}
zP|kfbw-cZ}4+|#lYFlU0R1#Fg`w*<}LjWJ(1MdmJds0$e2k;y|f1gL#`isu@AN+Y2
z0fGm_0p>xr)`6&>CtTU=m46z|+DCz6@{p~05%zY^IPSs+w=7#sqk^<$yH<@yPNtb+
z)PFBS|M9!2o4*^Udp}~W&(@7S?~8Q=2p+;52T}QaMn9u<JSN(B&nb7wueW`$fdVt*
zA!z3lx|jQFbp1nHXjswj-*toj{u)vJzrF9coOz|!5TMvR_&wjk-rf}h?>=o_m*+{?
zP?XO}zuv!zcA6hMsoHlGJS65J)AO?>)t|ldr1{d=A_4>tUc^CI&qw}#|FA=E{j-1o
zgT=$Om9Aeh@RMaUD5yObIqI0{`>hs!^GMISOZQWt*gS-FJq`Q!(xCKjP`{cBMpoZJ
zfZ!oYaS)W_9Vb7%I^Wgvnr`LlzO;ghj$yt36#cTlFWK+UQ+yuc^U(Zx=y&{y+McNQ
zf#JEUVmx;>b{s^FW6}JYQ*Y~;$G)+M1Re1l1?zJZ+<sp`y61yLyPgjIh<QDyTk9ge
z$Jl?50WZPgWmv~&7k$@pI}JMGzN~QEm*v*_99BCg)jn&(^SL>BqNCG!UC4uVJ%Zcg
zCHlM!`g`+4^&i(8dV9TLQ0|NHYW^Rdz3Y-4yVg=*c6ku(y(3}q2j8oM^}RaN@ygcy
zG3lOvb-UgxC*5AQeT%~c2p$jzCC`JVm1q5QsNGtUDh{4-&BePKEvG>6fH=tBJn;MT
z9`ik#^Ii1crGxf<M6b?g#M({w{7%@P<B9&^`QdneIG!Kw^*Ye|r*j^fS>qoX6pZTq
z!?D-rMkT*;bLG}^XKkUtX!DS*c^{Q|hJK299|p$7+i^Lj;!T%)wWyee6)UcNYGTXw
zUs9l$Jm9z{+Q&5*e~drA-(>#%ZrF@x&Z@d@Ed_!H!~y03)`4iO1DRfri2BjG{uJYl
z@jftXF^P?Xpk1F4c7L;Vz8~$*Y&o*1nEJ)T!)ptgZlAh^MioW9uQ6MGx&7Rou-fzc
zdW}aOEB9RE5z|RfFsS#(h4ptDzpn!qwY{4!9AE#V6^jWFJS2((zt=Csqr35l`%r>;
zA4=?b0pxMM$nJbm(BmSi-?!0ke!d0IE%NrcML}O5!s>sv&TFCF*IS*?Y&QiD;2}{Q
z;Cgp&uXj(EUwZTG6;rz{CqczS|CzsV_=<f52p(7*ymiDa74yz54n6wyJ1*Ms_jhxz
z`3~xF+PgPh*|VWkxey2*5C_4T2VuQV1N}Ps@6R_EQSbmB5C`7R15vMgfnR-BO{u?*
zfCun^I1qmx;C&nZyl*4j^U+bMSMceb&5AGBN`T-2aS*<F(0JmglMm*8MS$P|abV1Y
zu%0gp%k_tT-=Bc?zM0ym?PBW0L#D^WPIe(r%sco2KZ)ig;sfy!<oLjSY{9&bO|)^#
z^!Ow!^-sHH%$&Z%<`K6ss^<@adY#&!el-`2tiFQ)!2{yJy?O9y<30-y*4#pX;6da#
z7<%=9USGG{Oo8A5aS;4@5S8;hqODgwujy8<?n^5u5Ii6b#GMCW{XG}`!hH?6ufhBK
z8ZgiJJUipK3m@FFY%vubgZi8ZQT-0f?;q%gn9p~?^WeOF9vt}fm*1fNE(pKze6&Q*
zM+-`PESR{fZJkL|Nl@_BChhV&w_Ztr;3150fbY*ZPlfYTI8TLjAhUJA>*wzIZ9gi%
zTce-24<pF;VFV?9{k}e=)@SR+p7+H%5)|~CAG9aN^@VRP>U8LnGykDL@Q^4Dvi05-
zd@soCdw~;wQN1q<{r2~HHk&8iy6E4}<Q9enah-se*9io5T!wTlU2E^bVghuRhipA3
zGu`8)4_~Xf=CkcvNZ1kYIdbPcM|fT{p4W`$HH-heX3SH}Q_R!g&(p9Um$}uy*RN>t
z@SxMSQ_(P}-w)w8&NuwsbQ;y-;Oc!#YCY0!H^~-myyltTI)9#9=od;6`NR1ichCQT
zPkTPoJwJtdg-0BiwU~g1Xz<|I`QvPTj-%b|9!Eu$f81}H+5M(|txIUfpZCl~ooCvz
zU8}|;C(}&v;oJ#t?fI&h0_}MS%5}(5^Yf!`JCqrA#5M}_mIu+FOXRnIULS{Lx}GOn
z-|K#8_LKDodhUFlC(C2hJ})xuuV2r-*i@_Se|HVoM}diXxVF;uO9p<ji~`f+At>>c
z?0Ob<J*WFoM3am9yhZpmd&Sgl%ZaEwoVj)PQ570iEd+*x2kp8GAg;qPUx$Nu2v<D#
z{d^m=H?!r)qGAdj{NX{=<0bed_VFynCs^a-*L9ECx(+T|$6Y~rzHw0e4C{Ts*^19}
z+wHq}!1<#;SVcs|uzdcwb$^v;c7n3rz)!rtAkq5^MEm_ADE{EOC2y}=0>5JOJE&jH
z1tY8PAfjT}e@{mL(!1X3&T(_Q*E?w^^%{fnd}t>>7k=|d&$>(ZQy_Ri9AF+`9Y}8-
z@awu$v=glN+XZdBM7vH`yFbA##PjO1^SnCI;tKwxcRf0I3pa0~p6Bt~uiQG#kDXNQ
zI~sJ%^nNO@>bL%)^Zf^Z-bKQK**eY>?fg03vxN67C3??NRN~|Q4b>WV>-;N86-ArR
z@JH<5=iC|}zmKnC+Y8!rZAASZl&$MlMQb-AkJUvt!2lQl1D-M9)%X9X*`4+K@!j{2
zJV=6uo&{EnVNv5DTj&3>^&Zy`N5630`ir*{Q822{5q9c#)Ssssl>Xy)=Is2=9ChBq
zaz7dRk@&vsu-;#ae!27hT<{1Uhp*U2GL3_W*A_J0K6MKPf`=f-LC)Y$PTx)Uk*MR)
z+t>EFeMvF#3Pt-JD5(Bq>-<lq?MCHuBHibah*mGC^8<cI<2{Axj`L-8b6)!8!BsSD
z=yYD#FL}D-5%hXqomb4(b;D6<*Q;@ZU9r#akNe|L&V4ht6CikqLL3DB^EK%H2jzEZ
z_!-Xk-um@i<Ye2q?wU*2ANTnx8Wa?DUf?)W?BmRAjgP3GmzkbF4qE+ejX&D;W?hKN
z?-%GNe$Ti2JwMayMNmKSbw|k_KSQr{t-S|}iQhaReiO&9SNw?fyLd_d;P}g%<FByt
z>h%0&$iJD|r|n`Y8V|A0r$Aop_W-0jKBAsSfnT`3>A<YT#3>H`@81_(bnf<lNVf5*
ziL)-OwB$<y1P_P<%!6#N10{WLLwUH~Ymf27c>djV8butq8wcIX{WZG&p)C|BCJ&;$
zC$}VjoW93E?(g?p(ay7l#UIf>f3nT9=={JxcaI-L;WH@jnU1QTqF?VPN*u$F^wuXl
zM=bGk#A5jobw1*Ig1_Grz;`hC4$Ad1vFi--sjbUl<twPiHDMhO^b7NHYPaPCaez3;
zBo0Jd*Ry^86YYz6oiD}(<09hq=c2|Xjthc)Tu{>Q^j??ayizdED+O(wL_Kc}zeb&W
zF!w709>4?Q0P`T#bpXek!9L!^_+k7qIgc2&_=)OuUa|f5`?)Tuw&&OL^wG|eT935b
zO~3<qKpdn$4<0;zX3LR9#S|D#9z^}Uz%Rd2{d`YWKWOJc-SRhV$K9#+2gf7c9FGL$
zdeyM{ne4g;z3e=H6ug7?Y&`cCafi4|6nEJ^9`gFSlw|FvdK?LT{QhD7{lly0XGhhp
z+w&{c-={^>&(`O-XgnnAPqshT(7re48Ne%e#qUY!|DF_<`4fBH0rE%x{rTo1nko<f
zyZYmoe{Qmo0>J~~AbuWr^*%knem@ObJE^W;qUekIo?-YE&i8_ceO$d-cVQRrN40xD
zD&oQ#7s;MygPyzh|9Q2pBfhQi4L+jDhiK0)f<Ng!{};Ta##>a^yP%!_@Y-Lut|Nz?
zXgu#ll=ExwBfZ~uyyi{9UqS}3PP(&BdL759+Q;!{YR8}8D_DGabsh?K@q5&cU2ADr
z;z0CuE!EF8uk=OdDck;uc7Dd|^AJ(BkM9HSd>;Uh;PJq$#l)>R@caInsIGrr_1mlC
z7T6W@ykl6$<(zS^w>qKOZW>k;?Kl_yMDscjjC*F|?)LrJVYeq*yoHVLupSSfU&$SR
zdiA{`S-ZFn+TH7*!KYY!2Icv~VY$B@{mAS&7h(PW9F>0g{rhmXfB!@KVqYI4+V?m3
zliB$T@RuHcVLdKietiCGizja+K=6P#$lg4_`L4w0yYRVX|GDM&_XW4xL)<5hd+-x3
zez1<&uVYc=B`oth+v9h}aTh+gW!Yj11P_RV@XdqSJ5QP~jV&T!LAQ^iVK2M)7mmBB
z>Bp7st63N{to7NtvFClUj$|8yI<If<J1%El=`|!M*!qjk_aFRu7Xg9?!~y1k(>f6J
z<B3lj_gQ$b<`xnZ%=YIm+ULhXcppeK-Ukx&^$q@K_xm3B4>teVK3^?b`^nz#;k8~^
zeh)!E!u7i?##4;(#OH2mpS$>8=gs+T(LTR}>W|;gA9A}rQ9c)fa-SmnoY`_@Q85iW
z5BQu;?Q=SouS5Ttzi{}9eFO*|5C@nCSO?Nu2g3HeAGiAB_Hi)mm9Dk-U@-v?V)3x>
zn@4)qUAmtF!2{v|^8o7r)&Z;o?ym!)oCk#;V*E~*={Sh`dq%dOPv_QqAEIt2?C0On
zzi9l<;`KO-dOjm4*Y(5CVBLp-@xpio>As;%<2CBfUk&Pg$MD<wJcv!T+WvRffPFM7
zxM1S0wsj^=B|z|iIKVu>I*{Hv;Pw4Xu>Z}}K5ZA%pz@IE@pV}1<2ieH&K{n#7yq2S
zLH%kj7+HM>0fGm_0p<bL0sD1eU8CMN3?H|T0>ML+;^6$geOFDXzm5R2%R|(!e~Ehk
zMZK;eUB7bcG(UDyweM)QFkQbzt#|8);+*@BSWbc9A(=S%qV7A3j&J@C2`jF?dgWJT
zTQ4WTaPtuR90KGsdya!w{JC`>u-|rqcHD{YY1#drmhI;rh<beL<d@s)O;CTgh2O#c
z-QDl;b9>&vo|wl?7#EDoj$Lah;=r3Y$oBDCRNB9A%|Cs%)%$_^h2?L5zw+w?e<MM|
zk8&QL^2U-P0!)vG^<R#nmEYJ(MMVjJfo9;#j+a$iaPckzwCACC<mvl=d;G{epp}Q|
z1zXx*@KP}m9Roc%W#{|;a&d0+daDUAyF3Ivj{a#|sZyzvwi2M&JY*}*%AG!a+ygZ>
z(6C|a=I2(d_V6|$D$dSnU*qSh?-SAS&U^Bku5G`H0>$7VTkC`<?LPQ!?PpG2zkvn?
z$9FojbK|BnXjstgIPhwHcH7UA?Dc;0lv8RyUQB`DAyFKZ6h~nx|GC^_m+u+3lLiGv
z;r)f@mpwH3j$#S~4~T={%!3APes6pAj$eqVII%gMePpvADUkDU=ltIXo>J~G2|D`y
z`=BVFlkme|23pR)yJ*BiMHC1gf*1$GFB{tKwLLp2FgrXf?6b7ripuM0*wNor8Lfqd
zf$p7~&HnoOy#yFc9#;K##=KWwIx;U)bUU7ptFr0#Z)WZ!nPQo(TdP0&^s#xsVDsR1
zd<EU!@M~VqJ$Xnk3KWZnpvM*b2LoUL41fVJ00zJS7!ZSjOUAV5xvud_3Iq>9ii4r$
zM;1K&=?^5T_^tHc&#O;eKt)F}RylY92EYIq5RZZS4X=Okn3v~JU}`-4+N#NxHXS!m
zpf^0+T>ZJHhJ3f30F&e4&>%W|=FAxc=q?Z0Ua>1Zc169Z-)|$L<0JIIcL&H$k}9@+
z=)K1pwAe*~(d40iUi*Dhc5f#@@PIf-ejYUYVdkF)%KuKohRY8A+4Hj3c2Zz+Jlxph
z#Omwc-9UiB;vwgnjtvekTt|W6fyKer@5+Am@i`kwSn-4(9)Gg*xKeph$LoiE^})Yy
zTTg<9?=JiJtRXGFAwk2>|1MjQ(`OG28s1Ur)8FSd{D%02r`9{I)_<<pMuFhL90#v|
zef5hSThAw9!`gk<{&nfUHj${~2TQxXdVZ;m1PC6yh=b)d&g(Ygqbhkoe|e~X`LJI9
zm$#Mz!Gk#tin|>B{ZY@XBEV?y&~DaG)#{8{LxA=?yfx~&yB2KNMuUo%KQVUX-uo)%
z0l@>}0P_IrKzi%I1AkW9*}UcN1n3VBTOX)$Ua4&d2@pIW4lob0wGOnWD`uCw_9GG$
zbo+YLvYY2OYCe4)1&YbTxT*EZPMBCsfbR0})m>fA{;5hajS7zMJ#InXMa48zJaOfp
z12$g2jsU>};vn;R(B$H2PkwOfVwx)cXZRKO{?YU!nku?|`~Z95WMJsg*LLaKeKpNC
z&e=pNo%2Nz0fGm_fp_yDD#yWI_w%G%=bvzDhqW{)_{3}duj$sUn1&UvZ2iwI-!E8A
z{lY#Y{(htX1@kD-T^=Tlnkn7#_*McGn};7ZKlpr&+}$+W_{5Z}E`9gBLK<|G_MKGc
z?w;#N*zkgC8&{q3$~u~DBy(TuRE`!?Ab1cV4!k~_(5mdm<bxjTC=fh&5eIdvY%TuJ
z%d1J)(68?gXvf{}YhJIXZrdL*qr>&n#xA2#$6*EQ>JL7!ji!q^mk+t@(8OOTP;4H$
zHT>!DAu@x64e#nTvDP!k|4e|{<>98~<vy!AeLV$&hbYBC(9e@rI8^zhqMjS6+vxT0
zKW^E_eY?@RZ}-rocPdUl_Xpw^zBKfd@%wKpod*ODEDj#7{cMFbFRi9g#}zNXdeh0b
zeoI8df9<ZcV$^Fp2~aE^{<r$CVYwss6JTN<QoX)aRQ(-q-BY*5W19$&^WgUR&kJes
zzDHIsr>SDgJKy}~whwkwAb3a=2QPQKyv<1`ZlalD_ntqWUS|6W8Z^A`kAB-eDB4PZ
z;DN=#H!BuY8+z7i8gy)Z*Mf^$?p#WN!RDdguy6l7W5W^xwC7>zZR1vKE?!MU$B*yn
z``^n4t|dV5V2*<mKAR=YIBE+O6`#Mh!Fku0-$jA$^6=qrg<p(6=6wqEhKJAh_nNfv
z@1LmXnC|sFCDpt8l}Sy?ZrVYC+2ujh^L9b^t74O!!AJf4I{`+6hhqlSzOCkuld0eM
zY3ZWz&-GkNvyDqH{`u;A?ps3AMYqpOyJauk{-9oF@ep?Y(v8pS&c5Y`i>j}sK=6P#
z$lg4-XxNvP3SQYu{K9`aHJ<xY`^7YBcxKU0NA@c!rcuR)Z@zHn4I4I7QSpk+H@?@n
z=S&h5^y@hXE$FUcttJ-JRPpN5XZGCj=s^Mm4~PTI1FQq-tpgidT{!5ha}Lm;;QK8P
zHveSM4jMH)c*T2%+a8!gqlP#A>$WOS4%$M2(dMD*E7c!5?~bk1Z|uF|>LHi(_>Ka>
z1LC0MdGM|D-_NU0T|k3^AKY;F8C_1?K*NfAcf3_@+2Y>_P)r^s-1$+HHH$V=Ab3C=
zU>;x{NN*k3w&M0tZ?&68gMzbbK6US&$5v1vct9Lr9$+0vZyhLi`t)%R)Yw3Q-~n-f
zd64cpa7U?6f1lg%8v+!Ahv#n_^~7sGt)Reg^RS`q$nK9lw~a;>qy9YX*#1A=_VA<o
zIxL~U#5`16(|6#rUo0Sg;RVP1eP`|Kzb8S#J~u3`{=ZY#(4gTZXTQ>a$k+oUY*=UM
zEt~dVRZOCeO%{#(WMs{62@pIaii33TdrG%n^WTPzJGs`c1n3SAv!DF%xGht1@_@nO
z;XgHZT)c1c777FpVUB|*FKkeB|Kcqa7)>6IIe+@j);DjXKyP?hR(j2^jit>LC>9UT
z^!#=D(l2&UAb3C=h&K;XeLi|n^mn{`@)ZN;{6&L`L4W>myIDU~t21T|0fGk>2OFC=
zyLA7KnW4ucXMH>Mhlf{EAb3b64qh2^>!N12t|39gs9r~aen<N{lAw?ODKjgqx$Emc
z2@pIW4loa}4y3mZ;P;$p{+@$z$GGG7T=#y@-udxU%Rl^fAq9HNL(tz};kEyxs=spm
zvDFt7Ab3C=gl`@^+q!J6mS^uJK=6P#Fy=x2Yma_BykI#E8-8_Puh+gCxPb;8^DaE|
zzAI-)c|b9E$n<?$SGNA=mhTs=ra<t3I9N?Lj63DF?{EB?0R82m_w>iVZacSt0>$9r
zvQnS6JhR7Y0t644!~ve;!k^>vZQYKg%T1X?-Nr)&d(QrH?P3x&Y&o*ul9~;sk)Yv*
z^&b)P_j&>Z4~PTz=E1s8?kcLcW;p?Z2a)5TbIWee{<&)l1%d~}LGb6n?40&Bey;jH
z2|G@G>gdngeZP&0f?FS`a$c!z2MG{7AP&Tx2MhZw?YE-xdI|&&hy%<6tOMz-1L;2B
z(eHYfjA_wxUE`G$=nfC#JDu6Nanl(T2p+tMgR7oA{<j^KiU<%qgdq+lwSV`d3+5lB
zK=6P#@P8h>{`;M~uPHx=1_ht~?C#=b4R#VBcnD)0T>E^hDSyq`Oo8A5ae#S%bs)WU
z!0+>DUbold+i%|e^xW+f2p$jzm<O4z1Fgz_Og`wbj{1d-R*inP+^8ZN6g>6YR$u<-
z?7;*WEgs7K`~T{kcK2Qi1P_Vg;JwRQ_n$HTAn_Z6_MGRS_&f2eK@Cr=v4jG_gO_p8
zq4#l>y1%!Xh6QhW;q%s0wymN-@Zd%q+|%!Hrx*6VOM&k4FtKjiFKYg}llqPKm-^z#
z#fN?+sbcRpPdTOb<Ha=Wc-5e4Wc?F+2{1Drt{y%7^?r~2Nr2!XQ5<|*YUaxNwdN8a
zct9Lr9$+0vZyi`T^#3Y9IBqcoCg$OdEAJ@%@x{}LXn51|a-UV5zMcZXLy+TO@1s-4
zzWV)Uk}W*(h$+uJaLqyL7Y-lt{<geIMKtXA@jZS2d-=e%B&>+{M2PX8h!Hb7Tt98>
zG71C_hy&+&;MeEP^<Pyvu<x506bK%o5C<31;(d>-UQU7O@qqj6ai6^x_t|GFt~N;T
zt)F-83YsbU^?4oCcG}JQsal;eYY0#*9#)^x>Y=+16j7kJJe+>n{_d3)|4o6MhZ!R_
zl=}VlWz;L|H?{WS{nfTpukne3)gvZ1DxzNFhd+0^pwH+v6et!CXS_eD*6ocJP+)RA
zJaXswon<cFOT5O{t{OAz`TI6gU}7G&&l`2@?ghIjFg+fszEb_6^X}M6MaR2enbf50
zrX5stoSZxH++nqMQ?Kxy?gy?Zb>}uJIv#aThh4>2e?fuZA(c2-KH`%lb>H4Yf#Klc
zj5@dG6}8<#fZ)N+IPm)WD(t6rUZC4q7ydP6{C)}q4`Gah`X~OYM!jF>6Cij9Lmc2f
z2K)OMp8U3Hi;L)1;x%41^r)VbKU_`x#(9@FDyVp11@#*TG}?d3h+FnhpgTO&%sJ)n
zQQZ!au;Jrx9=~b)yML3Q;4y=0-&XU-$y8MQvc>TST0T3Q1Ran6d}+g5Dy<@6!=jeY
zUe|HJ9ts2xL5_o0rmsJ@Y3^1EOpS+G$DLex(q}s<5IjUN4sJg7v?f>nvyzC48(%-Q
zV8wxL1PC6?aZs^I&fued{+)P@!wS~bAADdN4Jz*IF!qEtmq>*`@PIhLJit1T-a4?X
z^qOBAOPeVWJS2*P+p7-w<Ebl;$`cjG?q6FtpvNyn6m<Ld+&k~dZ@RYqDhl+MhuoEa
ztgSq74*`M)#DTc;0MF^>&*{G9ng?FHeb+`3be!9G-&ceFy_<-J<6o}0?5ua@(y-vZ
z!r8N~dANx9jRmKE)^gavl~fcwzDLEIJ6*q*0>ML+<6!R#OWK}#+|M*sy#A?2X8!rk
z5()$le#SxP<_+I1zHB+k76vWeTOE4mx6h<)R5UC);hgC|O<qc(jvLyJ?Ectu+b9q`
zAPz7OunwfR4jd}jbM}{O7ZYG+Jk-x?zi-O!?IdU@`Skb3*&P}TZ?=JmhW-C2{luc8
zV&XTh|8f+q{Ki)5H@^PW>G#~cW-%27D~x_|!LAzn36S$J>G|L8dbGk+0u-AEd|qbu
zc?tf~<L~LVhacV7VF?WyHd;0M*>a<bC@?u5-fvsxqjz_1q(Jb1IKVvcY8_a$?6MZ)
zTWz60@DSuUc)Z1{&$at`K8-qh{eFb2>(##b%3rn<U}ikjm^A44JKC<KLBsXGce!KS
z*-L5Iao;(04>w-BlLQ?fe`DT)hYAKzx3JcZL1(QjJCy>#LpI|etmhFf8PlTYy2dMM
zwy;&%kI4r;*3qcp?61xmTI%0xsN2})%YzpcJiLGaz2V{Fn+`s8%iWX!qs7Cc>sI|b
zr`Hw&1P@V&gIfn4^Zmoet)xN0u-#|Y`mO~RwcNRs0>MKz;$Xx2j|lmDJrM<i_I`#B
ze=GcA{4wuS(QtKf>HqDx`C|$M59T;%f7WC3rX8F`MZ@|HuYd8Fm*>!^V3TvtoBR3Y
zzY$<Kc(}atr6Vd$+f9MsA&PPE#+x-;F08nd0KMTMDxZr%?dPI`?fWX+vXTT9Prr2J
z6Ls@ek)UC=e$Ps`-Moj!uej~$RTKywl8J+$oezn+ze{f$w_<bgY61*54;z~|yLA7K
znFI(Pf)oc=zkAw^-EaMaW*ZL;qQhs-oI!x#0datNknMG#cHgysUHY$01PC4w2bc#~
z2hv*yUjO~h-Pe?#LxE!QFyx$*b8-(Xph3aAdQGhL%<(@HV0L-<v{sMGKh(|11G>w@
zO@AC}amEeXsi=72nt%FitM>!V7S?`rOhNtaGilKAg+F#T+AMuefWhG5%GUqf^8JF<
z6euPSqTdJb<hM;*Ttv4LU@&<YGwPczPkwrc0F&ck#)u82e!qPg6%9Y1dhT0KJhGAi
z!2{yJm<PRA6|{bFdod9m$9FojbK|BnC=fiD<KX)Lp0Sr!-$R0i-@m$j!r8U{Awcke
zI0)}Nm^5mpbj#yg2{1Jt(!H+Qt9t9Ddhfk()-(zP4~PTI1FQq-tpgWSUh#OHTH7cv
zm^@tZ=TYB`nzn{y3!j*B)ur#AS4e^2!Ob`j^>=spCH8d@7$1y}m)Gq)F}Bgd!hW*|
zFg+fY)Ev0~$L2dI5Ii6bFb{&h4$RJJU*qSh?~|zGnd9G_H#PS?A}Zo~uV7y9g>l2U
zMe%w~JbyAA&!0?p97Vs5+H3!=@4NY!4d4Go+{RbN+`6dQt!qfsF#pB%ZCV}LO`?i8
zj_~F<;`Ohuez9Zg`2?699?EW>->CWYc@zj95C@nCSO>5UB(o0GKk;8R>is&OWD5@!
z>^b|(wTnr%v2)9A&;Ge<3k8CQbmAat-y^;5Z*JZlvybSsjk<+{Umm#i>&6R6Q1PyU
zH|~C<;<0&t<Da>o-9x7AAwk2T<wq7g{pk-xG`#eKf^tiGmC6Hx2gCv90oH-^)`6YZ
z^?K*Q%MKA>W;|T`e5)yc&Dl(W-~n+EKM!i{7<AUkvQw$sxajU__h0hq+Z1T!A*k2S
z{jvR$+>gIlNrB)2agg~uSlDN2zZI3&Qy_Ri9QZ#E8jgHw$L;-hQJ}v(c=h|!*HwPH
z{J4{H@_^tW%5gC3fl7yOYIlGF!9$edU`?ODkNe`UeKcx__XJtr6ZG{-U%xc{r9}jo
zT^{N;y#B>wUY<im$CtZZ-sYqeH&GyXu#1C51>5&kxMd{)+Vg<-jJxxm@saEQyL8o;
zYp5vr+>VMDzW>yA0t63XjDvlJvu9oNa1j9plZW2ZAOE`T+yV;pmWMaqtl4s5#hpY{
z6y<)*W;gt>Z1Hp32+*E~qGtDvf2aH|A_@+vw&$9MCvPM`cX{a0`?yNo-`h<6#<{hJ
z_R9NY9RY#|#6kQ#@cMqHSo<wU_FGc3!88JNhll=;zFT&1<v%D8JOnuonjO*ngC3Q(
zQLpifbE-F)G-L(!8?P_Zf6o6--$8)jA&7A>v0=NKRj$}fqmGSMjefS=s3HnXj)(Ct
zS6p`1J9DWhIILh@{lN#e(Xilit5^TuC96K8z-aPN>OUXfbbpP}BwKj?gG;*1{b4r+
zf`>51!8Z->{c`hBhp6a?`{~oWpFV87z1i%}KWa`~L4m>K;gyd6+EDBL9n>#;y6xdd
z_jOo8f#AU|4o(?#V#P~0@1{WT5al=s>;0g(&MlbNxnbPGHE!n?U)|-1zt#|-Jr5f%
zoV9S%o}C0}<)Oytm%Gkezkvi5y?&35-}Wb-@<*%6OTMK*@L(4Q_#Tz`d(<)KPv6=4
z=516|9MX7OaqelWs9Wgw`s7u6v%flPXsLg%q1nRkcV5%4W}STm2p$jz;hP7a@9#Bf
z<KI8guppj09IodMk8XX$w9|L4BSAw^o?p;(Tlf4Izx<U34O<uea%r93J4sORhVy6A
z#S>Q$Ab3C=U>-QF1IJ7{a$%M8r_-R~<zqj-YvJr4X;ARNW)D7B@yO8>m|Y%D`KA6_
z>&vd7z;N>*;V;q*yu0k<vxc<zh5)0<!x`^Ss&#v#1r*46$hoFtgTo8g(QM<hM?Se}
z_5b!zAb3C=U>?M-1LaPiKJI}U8z?Y29-3S{?a2>LT}*-Q@^JSnlbV#>w1WaO<00tR
zNnLj4kk<c~zmH}czni}2<cjsT6CijHArA07g3O*r(5>N5hYyh%1PC4w2bc#~2hv*y
z{xhfBh2uYZl>)6ic=g=w=U$o8`P2Tp2oOAo5C?~5>}%Sza2ExF2gCv90oDPm16T*#
zUk7#!=yvbZPtT@6@Zd%qh<cyT#Wx*%>Xy4H$re_9{+fXeFZ`1NGvnb<!Jf0fT)UV6
z!2{yJc^)hr_EX+~it~wR_(ipU9zDJJUjzsq5C@nCSO?Nu2QC@YqUXBCD``}*?X>!z
zj4a$mfWhYBv&+Yq{dCt-5*1wW@~bzUeCxLa2p$jzm<Lz~(pv}SY$BD;`J#vb!Gm2K
ztpDwb`om{@LV;EuJ{Wk)xeL!<LxA4!(Bh8nE$(Z-oB)HxL)hLYg#O^YQcHT3%0nC=
z4&01`T4f#?S$^1d3Iq>PiUY5nzYV*GJH4>?U79Qpcz;f~-k+1|c#3vCL#F-&^*N2A
z`n|u_^pob@^&<s>hh*a5>vI2|P-E#z8dhvNzs*T`zilMIaP!dn%~MXP{dh42ip|6L
zPG@#*+;j#F8WxnDHm>y0jU*^|-j8oD?RWM%0?ZB%+5TPofj=wlY~J#B5)^d%xjVxO
z+F$&y?+y_lcnEVG;J#qo7mWLY-M=pw^D0{NYH0b91y6ta0|9!=!-!8VDW$x?VR+x+
z9rEk6xUWU$7NrIaZ!^3_4%_z~e)sUY`5p5+^_QVb{&D#yw2+=1o?nHP)ooF#MQ1tF
zbFi#4q+FlA_x8J|>wUwAl<wQLN2lRKa_{ZlLD~0G_I=sDkG#+6AQzVD(6!^ey=C1F
zUAuQ3KBPQ@d+zHhOL9B(lx^g8c&MkW+VNgF+3{XkBd23GS-WGm9<t4xj@?<hcPFK^
zH?x0V2e$9rMXpr3bJqtN$(b(QACmX^U08*D*>KseJ$v8NOLm|Xv()|Gj&ib0_kKNk
z<;!Y4?q&6Rbd=Mj<$Xsb^RTSYqf_5}W}!zHdDl~JuyoH(53!mLH0pD&jHP>ZWUb{e
zE2VqoKhQ_6*{i!;klRb~qE}z0)Vot3X1@>fvUJ~0Jy@o1PiE$!-d%e#vL5Lmw|QL4
zu6_D+?b+qpp7+W0^nIZBJ?Gg8<@Vw=8`}w+*a^?K6JB5^ywFbgFFWBy!-pIn?|egB
zsoW<UUAa>>Qn^<)Qn_0;Qn_C?Qn_O`Qn_a~Qn_n3Qro^Yvh7<V+rBlj?OP+;zBRJ#
zTO-@PHL~qnBip_;vh7<V+rBln?OS8pzBRV(TVvb4HMZ?rW81zpw(VPE+rBln?OS8p
zzBRG!TNB&9HL>km6WhKuvF%$E+rBli?OPMuzBQ5imV0j(wnX&l^ly0;%kR{+OMYKD
z@d!_JV(a!ptsZ(#UXkU6`cbw}Kh!G!@fK2k|CR^3<Ui4(lv377QFtbQVE(iD&*cy5
zKRkc1yuJ8*pNBd;G$i-od%F*3HFNXJp7oTHIwHS9{xR_jeyNt<o&4PJ{OW_79ot2c
zre2-nxXpSrN5U3v4t{(#M@qDv_*#x6`ZoFF94RbY%`&-CShkC+=SpsFI~(RoqHTMw
z%$5A!uD>T&@?-l!ZpZCy?c!VsTRaE5&nhip;~ZRlLuo15?YHMkOCoPA2A7tywLMX<
zyyV7q$0g;ZOt&{%mY2NT%Dhot!X_>Uy*?~2`LXRj`UuId?Zz%gNU3g<UphkadMo<x
z5t3iqFY}L(GTD0nb%f-`_VK18By8Y7<QVuwj?q-N&NYveMBbWScBJITc5Ig;C8w?P
z%Q8m++1_~dNGa8A&Hs*+oVKeQ93>^YjgfV+fdi3aVCAJpN!Y-FTQShB%~29Ia3FFF
zta$t=$;+*7*$NUia3FFFoF&I-7`JolRFFj2Dz&K~Wn-Iia|KD1?Z-DONM3L4zpo&P
zww<`Sg5<_ldshVs8#u@$21XpLAi24HR;8j8wC(b06{U2x!h0%8iMAHwDoWVI;$ZoT
zic(m%$xj?DVe^iI!*3rgMSbi1zoR7)w&_!jmay4#@X*|&r9|63i;tF~xV^mgXelh)
z!|RWhGTjbdeyoH|EDp}T>sSe!I2`<S_py?f+gjOvGTV-J$4c13%fWRWj+KJG<#jw(
z3j4NY#<7wc+cw!ws@s@rD@kr_Z(LSc@_K7^S7j;PZR~TEB`>!>rd5_i+uo>gqLj&Y
z<KrhvQQNNi;6y1%+x-8XD7m?PIJ%02EtniU@>vxrnXT!URV2T*{@+%Syxxx3P(@00
zTX@Myk{jDUFPtP{;~YHt=1G!Y+s}JWl0?{gN+(NRZSAU^EJbM>@~^5=wzk1Ns!F15
zyLwlZ{MfF2tg3_!9EcnPzr0md!WK*pYEGys`LVsXzpCW*_UMVHNN#WMjXy>5VmslT
zQzUHQK!g~W_`xX>HgPyO`^!_LRJRv?J4Fi1)+V=_gw1^p{yDpvl<jT9Rn;U>wuv`Z
zll<N$w5cXx0|z3<K!u}Dlfu63C#OltZe2b(O%iSExvsjD>h?gTnv$2>iO1EH!oDr*
zT~qR6`)fr_$;<8H%C)3ax1;KuE_t<8kZp%?+jq|CQc$<s`kpR{vUMw4Tk>M-UcI)I
z>9*<nGbFdR6K<>{1$C?b!<mwsTaVk%k+8YXLEA&;NI~4z-%wu?X*)2Wz9i!I@`U=5
zAKN4U)R(Y<1Gi(KZvO^Swzli#w!Pe>0S%-~w(Wx(NTP3>n>UoOWtW2<w4s!(?e~QZ
zr7&#|9(A6S&8_pAMv~WCgCiSDLD<xns7dmh(M&H<li#CrZhph-rry=IVk`FIwEXiO
z->Y-^;AS=EEAM0DySJw{yed~t<~?}tkg_rsoi%LH5i$-~@y$sUWvuZ!yX{Jn3U8|Y
z&hav8f2r1MU!m>B6Itb%&z@IRhPV4qdE`_X^DbEXRSg-7wp1xyTgCzJU31-;GS+C{
z@6~!T7M}FMujk6xe&Ld8=gBC&a8$=8GQ8dF()TZvG4DY4tryE!^ywStUnb*#?(-kJ
zT*ew_9G=l!#=^fEl)pyC_9NT3Y9XU^!_eWaWO%#WH%o4mF|V+=Mq3$+{(X9v+hiPY
z-pwD~DPxTd19!BOv9REO7j=-a{he!`>?EU9xyr138Qz|Eb%pLS<~{rGHV?>H)b#!T
z^p<hJ?mzy1NX8l;A9+SU84J5y((Q2>+t=v+$&)fle~jJpv<z<#{juqDGUi=(=$XMX
z7UeXUJ4D6-)9yL;RT*paef5^tWGp;q#@LZEwqLz{#b_C&H*1~sh751ta?=BE%b0h}
z!2gYtv1s;Z2i}!&z<|}4e;{Lxi%xw0BN+>~Up0S{jO{;or1EDnN*yQM@wp6dpYq4s
zU&@%b_=vS%$yhYx;&W%nIN-`|eZG^i#-Z20m@Q-BlzE5e%Gmyaf38&Grf&Tq3+4RV
zMeP<XmN9SiD<}OfW6_QO&)%89`%wM=f9|z(tyedD;$w{>R7j=ykR+cd5~b4ZR{BI6
z5mC7%TZrnTkmXinDP;*E&PT4DYu}^mTC-pK?tgB~@BcL!pIOeCnKPf4$7Alx%$((Y
zp6AT@oH^TsTW_VBoI7*$cB;Plr4Q_&8t&Gt`fjR;_x9+#kE(QN>YM{qQpW7cc7&?G
zb)iPbsHPWgFz6)J<dyGjI78LfXKaZJRKxyFNtdZ6o{bwHoAn{VDYcK}pqkRQZH-)1
z{bj%HmX~V!)@AeJsV0Xm+;j_7U*jrG3R4Z|XdNm>HF56eTS`!sIuoCdYD(3kZOT&h
z@AFNpKs7z3>G4Wblb`*hb``3=gsI)DQ4KHMx!@kEi9HJDtwmLO@R4TsQ%yP1X-FNa
z{)s<s_ft)Obz`~uRFg}_wR@DRZ{vNF8&M4peEoD|s)-GU)oDsqidoj9In|Vz=a;mk
z>hDmwz;jg7?|7j_YpTgR2Mm9Ss&Dw*j8~|Jn;)qdq?(w|_jWs~(!wXFy+t+UqmR$G
zr|PdYrEVvx=|?hpccGd*?v}J}RDCbiFW8-GxLBuFpHNL)GbW`cRjJ>GJ$<tD>Oa%Z
zu~q-52mY-GEaw9_a;Z;|FYf%Eb9=tH-9ERpkD`T>*U~9UdPK1k?<wwhPeFNc=kLAK
z@3p+j1+Q?UrXS@hK-2H9g1$Got=zc99=F(Gx?eI%%QN~_(XYy+U)5vJkMQ-G#yE>z
z>9Q*x-s4nWNr!s8&GB?yHNY+HN5^?!TovQ0bX*YQs@BF;Qy#vlN`K1Hf1%d91F5F}
z7M%Gd)#NvaULH)<S7GTRL#c+hpX)cAYT`F{t{FvDdh+?AKTu7H8_@bks{VO%MvteO
z{{G?pKT}P<yHxc_RDB1Y=scBb_=n^<)2SvtKRMehs#2kh$L3H?S&=X(Ox54(;SKYt
zra#oN#3HK6r$;9(rRtlse%uPG;b6`qtEeWHu2o|lRcXs>-8NE9`Etm-EmZx1B{yxS
znx5@k;~i9!|F|Qxi>j~Fb6fUO4Oi*!J3uvY&mV0LQ<X*?`Z<$oO3RYRPf+z2c%t?h
zs_BcncRx=x`Qu3oE>ZQ}zavj<EcJiv=4LslCjL}!NE}t^m3O!2p_)=+bh&t{{`KqH
zCQwZukYjQos>uy%o-Rt&cezcS5>&%8zUfhlYT`SKmz1F@RXQ7AAxp3R<6W=Y6%X(B
zh5C2dws|l%<&$0f%$5hU{xF~F#`CG{;%YV?v;HFGJFjq~re7KJ^DsY;wo{y*pNDp2
zbhIPwYM)@9Gv+x*<2+}t)}v_WVZI91VbFG(;Oj8FT0fz`(5CflSAPh2_jjV5iuHI{
zk4MKHIIYLS?*QKSv96i6=X9=X#{18;??2JTBihodcOL?;aPdkyc#p5Ujg#B#!TuoF
z9|Ze@(0Nn__XiREeZ=kcklXEZ%RT{C_ffK&v*q}%27vxN?5~CWwdg!xjs3MOKWBoj
z|8@I%N#x75{W91eP_X?0@%xMYEtN`)D%xRJ+F=S3@IbrNNW0W>xw#qu%3XAnJImz-
z?Rrn|hl=un_Y2H>jXdua&k3FrJSW=EiSW<I)da9=7YolrW4|fe_A|h`VXPZ=a@{cF
z`T*bMG51{_)T7>LceyPe?Xt(Nc=WC(a+^J7=My@W4yWQ}_Pw^N31HRE_3H1M<N1ty
zii&Yl$IF?=_bGKdU$@)mm3gc9{wKcw`M=-)Wc}Q+4$m#^DzBu=u5{QHPqe-7@m}|5
zbzOvLzb|epr)K*B)aAmf=`uU+A}SA2?N0nUu}L7yc#NO5nn5w7+`2hs7FGYcYxj%H
zLX`C{+WMicq&oW2ZFHTD*Eo4CJ=p*D+WIW{k#_H<(oFVU`gZ0bs^RI&FE6E<_%^;r
z$K-o-G0S>1cXB>JzYTl(WW|r~(L}}fX!O2cvwCka+rGzoNWs=ab|yYwRy|VnXq&QB
z{rh|qMc7B&YwZ`Wq(_u=h+?O&{N;J|6=SJJ+xw}}_q1s1H%)IEZoQS3i{#vyqqkG_
z%`bgm2i0)5X4Q97O}w{9=Y3SALsKz-H|uw9=J_Qpvwr6~R4(uw)$}`FXwjN#^3DOn
zU!v+8J~!ik$K?|9`QC0vuS*M`oc0#gl#lTJNH={yGOK83{X@MN9qWhDkFj&XJ+vN5
z>`^eUo_>rgPY3_!!>)Mn9&GG81;`g~tXI$)xAIy)FUsqc@(K>ue{Q3NOSGiZxBt`Q
zt!ZNQ?th<NXZ(;3)P3@)&D%Rs#rr<q_Z!jnaCqSBryFOriyIEB)0C=&@wvz5P!1_8
z5(b5-`g=XRVLsLLhdP#6L^b*J=%l4oeUsLYTR}A(%z0!L)x^@ZYOJFwZF#NRMye@a
z4w>hs_x<z%IVKm1rS><ddAcZ7-{m%SN>B~Y_@+lGs)_F`UQ&jtROxJdg)F`L5B&pT
z_Yd&OM_%EkAGYcHfHuwRvx}2m@o?)$xS9aeH>hul)3T8A;q=b8QRTgU2G32;>UwyK
z%1Knaqjml{^6UEjH@x59!TtK&MdRWYd);D3O1%fi*P_NK^E=&YSNhZuH6EhId3Bw>
z+qt{lKChOG#NC-0L$lzIEK_L&6}{g%4rew<n?&6wPUu|lw=DlJkGtVIHH`O_8*@t*
zue?tmD4fL$`@tL94?YW7)<5)5T)%(f3LWrCS$=&!>Tx%(@$y=Fur3noB1KrwgL&5G
z#$BVez7js&R}<h^JIv|$J+jZE@a52^@ng)>^7cHfXnhW%HQl1D%Z&Q`xeCCsb_S#0
z`KZswg4Z9{GdZ3b>KIBIT;a*AiJp@5VJO(B$-BKnNx=$%m_TeGrgx}WC=i?VHy9sG
z2o?%RJww4FS?3k0gP6vt#d?K;E&A3=Ev94(l+HN*-+urC0w7?G!2do^S86J2;lVcu
zfPgy*;8}NPzQ8IFu$uttf{4U{C?X;_009utK>&4$4oKHH0kLav;KWV>-YtCIO_`nL
z!<%uTY;eritkEDk_7fY(_P=94y+bdADh6K;zWO1R*Zw;Wl#LDwz126%P5WSn;JdG8
z1=6WWr(RjZMlT1uX8G#WB=|u<3VujOjDjBpV*b7N>c0+p1V0V-4E754rh`v?>Cvyx
zyPcAP{ew9xeH_a2nJbvL;td9$O7*Q*;5AA5{n1#>V`0}=35T~Ctmqjl-RLO%RjlOp
zW6H0wl2?w>IkHJ!Iqol&O=5YRsFqD~+j07lY?9r_<1c2D>^PRucvv5M*JYD%{FlM0
zO4%hGR~S6{M0QE<v9V8f$<0TgZ+6M4<D)y`Bo;@j2jV2F$H>NU63-*Yh&Ty{dl|g*
zbDU(yaVmc<$*!a6Yq=z&$CLrNB;H5*cey0Hj`eA|Br8XUZMh^C$FB!-NjSj3&1GOR
zEu%({Hl=e*Za(Tflv}dnDDYZtN$Y6yIb|fsF=9||$>_0XT5d_}c(iIBN$)X?y2Swo
zZY~2mAIu}+00UNKpk1>(5)LqMa~a6^AdkfJsF*XagaZuRTm~x9GU~;nN`<_V8;+Zs
z<&~T`rZ>+kx#d_fGOxt@NL-Rva@$e(P+p0}QSM}32?rQhDFcHq<ds++zZJ<RiFVX2
zkxw#ngkQ`j-RKDXkWa$lE(SX?@=0DfrhJrN!oiQh<&^wVv>$Dz<(J%W{4qVhgyZ@Q
zIxWpF-RO96U4AJVkI(n!m%MVk_fLMw>T#iN0SSk@7~J`M0SSkD7;Jl?fW-6IOT*W5
z9B*Af!r@&8jgksTq96X33rJo+4$mtfu{e&>aEu<q9=lm$ag2DVpv3!V@_a$b>@mDo
zL5b(Fd3Hg`ZN~^-VadwT^n=1uv>gp56qZCf(xw%bSROwQEh6C%lR@X-ib#5nI<tyM
zb{!wiFCy_ia_ui789i1%P*h@ZT>Px4gyRZ>t|N;|b{%U^7nR&_yd@Qrcpa@v6qBOq
z=zm{v$*H67o5dx!9Va^!m+Uwm>sDOC0S0a^1M5c>mvD&5p!B%nk{!p`bHydzN7ur)
zO018uqi>aX9EHZ*D&YVFH<W>2Cfq9Fa1Vn!XWc3pJwDrbtK^lVS+)`q4)zQ#-dRF&
z`q<x~gyfdvm!>5oyN_|rN=P`sz|CbKZ=RBp*N<~@NlEYV+T@aw+m5&Pm6D7eZ{J*6
z;&~KKC@p#Y_`5@CiN~=mqqM~HxW8Z-$>@=%LRpE|k(Y+;#p7({vXZFBGabuHZaLcJ
zEGO|e-Y8X0vU(g`QeI+p6lz*Q67?vx>~@Le@#eFYB^>M-JbkIMB;xVU6L(2&I?gBG
zCAsnVeB51<9Y^PjcS$(FfOQ$D_+eGaspD}PH_t;#t}0nMGW%AQ+<qLYUroZ{EQ2>m
zHOZ-C)9PxHmyY&%s!L8DZT8$P@jj~NzDE+_xcVU<DG>9&AM&}ucXUSet>;Vo@0W#k
zpS!R@o}57<>3@B)y;7&-a|breTqTR9(_ZL%_?9tSiKu#4&*#VGoAnP7RqJi)Jn!YB
zyNT%f)$8{!U;X4R!duxswbnx$et!M>b7$7dqG;1TE0SKB^b4`7k;vtL?#aKqG!GN2
zTC_{KnRhhaBt!Lah?Fl{_wQ+PN0l<4pOr<`zdk>6v|-Xg*{!N@kJL_4-~Xrb?LR$6
zh7n%+V(~gpRBC%dR_t?1r~9WBO!|2$al7*3Q?tmK$u*NU%WhF#=~Tk~Xw?IK<C?A@
zcBvAt+qK1=Uu~ILA%obhwRhZ~XV;S(iKyByY2%mG&mJVA>V%k|`qjyQm>AXKD@J7`
z_CG9(s$*OJcHioOJ7l-3Xnf7qb5r8-FWC0#Zla}T!&Uq>E?4clVk`F((Y5C7C2O=S
zo*^66)pr;Fyu-5XveWv{{H4z|?|DqNOO^aLtE_(y_1d9skF!44YX{ufLm=(-$8)tl
zeMuHY^Y`sqe|D}@vM9Q$(b{G|{C!*&O<zoykgr(&eKKq1RjMmjY*J@Y$<wmk$}2sQ
z!i~0m=SV!%_v`2AP_pKC3uU`h9lzVo&Gy$@O-HT%WXHstcM!8`HGVDsi48u?bN~K{
zvQgCv&*=4bt&<L=9Xu#|O<SEvd@=1@rYxHFxmaV*mYBn`DEejgp{2S#yIb~(HYT5+
z_;cd{SroOqoxwW3?lTKCtkM0X?35a<FZH@Vy!ueDkJA5^71irc?QZtz8MPaaKP`Jj
z^MAZI_NR&e$VPSCA5ZVTy>~h>svkWcyY2AKHL}}OozHGNZ{Ik*{+7C9WYM%ma{lZ`
zU)@BEs@wRx-Jb9E&7WE2@rr)8YtbKLziyJNoZl<@LE|2|3KuQy7e(J0cFAAxr80ie
zw8W{#bI&X;<L9NTA1r&z&7T(a^VU~-J^Id|k_G*u>Zr-m{&NKi`Q5HJKahQMtzFv*
zFFpU|n_vI*)<Gh=maRQ{am%f{h$uR@`NGj()!$E`&T{BjD&xZWuA7M{>b80PyxNmJ
zjwH;U^y5*o?fB3Uqn<qgwTXi$^`}?t<n_CbaBtjha@+4NuX<D~+)hi|tX=THPQpsH
z;!&;-+BovhapO{nS9R@$=l7PH^%oIMMQQiwwbN)nuG8*UkKHlnpRRj|sG8V0Fzf9c
zdx_}kmiB~FUpl33;b7JdyVu?7&8O1qVoSvA+AWK!dgqbo<)L=RdLFJ_;i=QX`aJ!Q
z*UJ1><-9DKvgW`#yiwDR{nWeaKO~Z><4vycNX_AERuNIuuJ)*!6Yq0n{)3kCivQs+
zL*E>&u#9*`?V3NzJD#T&W>$H*>P}e{?eO%dADZ;vDvPRaU)Qa+FV~`_l}mO`mF-qu
z_0Wk@<LBN{;wZ7IX6H?t^_P9vs+EC-JBg@jc3lFma^O_BMRz|l_KTF`gqOZE{PcIN
z-Z?=;)9S6CoK&yDG2#_H`$*4UdiFa;L{nbnQXQ^S&!5$On7r=aYQAJPm^S_7z`X?O
zP7Y4@+w+cJU4BKMpT)|4tL<H<{F~h$LtPGVek}pL`x}|<56=9yTKb(Uw-LQ+v|bpw
zplH=`dsh>^T6N3D)eoQjn~18s_PvV~?zN>R`p!ge$Cw=tH0#f)5~9_gEz0#rx86Nx
zeV^_!FLf$koQ^lrIF(!Pd*wazk5>-f{@aB$L=<()d?KUm6t|XRy><mQK05RGfulrJ
zEq-_1firp>BBCj)d8~GYqr4B=^}E7r-w$5)pWF0gxa55sWwYwF{pGFh@9!4>Imh?9
zaN_QvGArdhpDs51+VKxdt&qL0toA9_TOUMPpJH^M1*7B1-3ynR+O7CL!b{DL6SMNC
zd<Q@@UD2oP134T2DT}IKH2UDN!)3xUE7kiv>2=q4d-;XaHXb8pRc}0syT|k2Jo2$S
zGKo=D{5LGKu=naNr)9Ou!EW2%zU`RIO7q|HbnVO4=>9<(s@MZ{hQR$#4`}n}?xV6(
zs+DTH=acD;Z$bW@U47kF61!E4lluBI@9axnIq?W#t?GD;o=0kz;Qo!*)!N0!Xud{r
zfARaTiY>3Oo`|XopR2U}p5Z$PD~-e}(*07ypIiRHh#4|3)ry~0``M{*Mm)32UpI70
z7FCDd9oKYExf8Nzx_iXoc4rG6mA$5Q-`!9!-`PDfD_v5nN&9|-GYBhHp5tb5{c_!1
zC&ABKn~ocFXaD$Ie%`92>;9E>H|=h|Rko^Qa}OW)<yS{!yH!aC6a+v31QY_nPYdLJ
zy4K&stE!}vm42NuC~4Byn+b2VYHz3Jbr<J)`4q8R-#Y#6@n@zTCtg!l^PZJ(ocex%
z*>PA_{*-dP_3M6jKRfq~EQ&rjEF<rxH!jI`slrLEZW+HEe4@kfR^R<4>s56+Sf9^6
zQtr%yjStA8>P;_?Y@RsmxGbtJdib4U+4@|NMb$nzmfU{eg>y3084lA|Z++}o?US;o
zYS(w9Sl63g3D4<u5s|)cuIz&hy|N#)Qykj9ck7hP(lKIGz4Dy6H62~Py7!qa@yBJO
zYBoI!KMu#VIk{A}OU>HpR)4*IPnQ>TO_!nIaJ_QCYF@BY;S_mq%L4_seL|qlaOn4N
zu>wt&EtcJ;M&}v3Wxl@J4!ib~<Q<QCe>l<3XAT>mmNWl)SyWYjH`8o;#ppUmr{4GU
z&fAPucSiI77V^Q2n$<IjsA_cH>y~U`U-;&A#BHi~-x9aEGrC?#?cQj4<?Y_7d8@qB
z#VUMdy{%o!%5GNS7_D#gx@Toyq;S;jP*KmFYJaebcS6ih{p#dDOhi|``#HGHo%(l3
zv~#EC)$@v%)%Q3f)qlM2tKY|ro_qDXZbs9^JD=<tci<Hd@AYBXjvt%aX<QyZZ{?lO
zX2aLpAH=(VjaT?~wcmNgv*@X>hnMxAB&^hIIWg<c>N+B`@f!8FIOU^HA1ix=uvVjX
zxXpe0#_9F9)Ey)9Qn!6qS5$je-(CFk4$HR7yp?zP<kgN4JzOpBynZ*0hU?V#T(!9H
zz8(l1XtDLBoPMLq+b)EEeFA3pbB`8(*QbW7Y67hGy>nYQV_W`q-|B%o<Y-ma-{(}9
z8|B>SK80|fKl{;FH<3sj6uzRU-t}y(+=;@)teuh8MHs!W*N%@?J<vC<=?da@tyA*3
z0~=<plHIPNe2>?to%uiB8~f8lx?dy|98eA-J_ma1U$wjD)0a)C*LjO9s;bi?%DFPX
zL(5)P`x{3JM=O5ix~SjTP}_aE-Ag0W&z&H=RN-Y+w;dlkV$`z-h^T6`-RM>KtmcCm
z4acePZz}OxE&q1=SL$I!&5ld+vRkz8f-CmA?fdM!@(1BwA@JnIvg4aPcbxFny%`TD
z#E-Zrv(oM}3pA|J{iH0aev!BGOZ}#0%4RjxF_biT$?%web1_&UP(5&WAf|VyStt;j
z^*0zFtP-r&==bNJ&e!DM^SZ%=V4;B2GZZY6bzU(L6NqV?TC7(nShsIIU)ujX4k^D^
zYmkke``f?NKPT1nN?)$YMKw9&&!YLL`o2Bhx&YPiQ{~3oLRI-^)L&xWrw<g)^1PsD
z_2N`gKAhh9Hmd%bd*}G5rXMbny&To#u?-vDPSuzA{-8UkhKo+vP=#vZ>MbRzQ<Xl;
z^YXn^QtCZ8z7|#g`8ST-Pc?n&h#C)3O@1S-TV1NY@|Wk;ryAZ?HLfAm#IF;Z1gJ_+
zd=Y9wCFQ0+w=|{d4<GYAO*Oqs+1H+-nq0m4#OJ8`_J4Z3HP!I<(`qMCO?+-o_gAS(
z1&b^UQb}3<Xx=xd`g?Y1_BPe@I^&0QpqhMY^Y%_ueZS@||3205YY(*hfNG-ejmh1q
zN}GqD{)9@(7c1)YqUvvSsYhR`>9JLp_NSUW`^5r-sQNk%YVj4-aOHW!zoweFJ2PV_
zRcU0IN+YPGw0P?66srFCPpADrHGR=<=YOP{oV>g41ggH;h5JmT8qRExHi>HDgw6$j
zqbj{TuGI`GDYtG;nMKvVF4x{URMY$4Uu7QE<VV`Qn@ZJp>D!r$sD`I6zr2)c;@cM=
z`M2CuysQ6edY+Qrdd)hj{-FblZlaprG~9YC)#TinqqkG_%`bgm2i0)5X4Q97O}w{9
z=Y3SALsRD*ppr6XSGFTm{jCc%Iz~0UaDzc7sV1*{Z^Id?zCL41T%a2EZ%VpMHSuiR
z_}HxSKc)7O98^=<wylwis=w^F-SSdR-@0sGJk{jTg_~}n>T6u3Nnxtt9IZpes3y++
zd`k(cQfK1xQBA3Ov`ty6{(ZiQ6{x1CG(BF4YVxz6)UHC+moT+^HLBsII~UwTHL*v*
zytSxG4?fcDeyS-aIt{5q)j#pa?S881uWl??pK5Z+xOR_H^=-UwawDqYfv=x#Of|9L
zusThtN-@iNG^d&}^Zb&QRQ(+)7kG|p`W-K{XiYVF=YZiaQS}X<oAC<OaPuP-gH#jq
z`QC0vRa*Guw700HeDv}8_Ei0~rqu03HT_6N?=Do6$K8_FjjHdZ`USgF4HxUw>JzGo
zYsRGXq$>5>u%}O!Uj1kKIkxH_^}xUNfaQDuM=td#^2MFMb8gQUx7+8I_EEHO@>)7Y
zNslOY;yuM3?<puR?)<%X`n{G{x!@IU)byiV1!(&HRnYe)x0M^W*y9#EO!rGhX?aG!
zD*9EK^s9R8`4PT8(->#5D_wTQ!+V^{E9p>=w>h4!s|L8G{pdIkjH_Z?m5vKyT-DmR
zYRbbmRq0PT`Y+UacOcdD--0v0q?-KZ(946V`YJ4aWGL0}_H+HR=5r)|bLW~-RHY}M
zFZu)3l(+${f28W4H)r&Cs_E|^-v2Yz<hx5%pG4Jn;EB#tsfK?@o->_l;`5WU&7vw5
z%6M!J)sz(pgThq(y&m2$pKAI;9ZM{tntXb6(o(9vN$bb0pc)S5JhF;vV(D5n)=`zV
zyw+_a)s!!X%-cfMA6RnJcB<*w&NbdaHTjP_Lc6H?Iz6{#FV%3B{=NfL6Zibl<}g)h
z)S;g<siw3ndHe)be}N}zpP`z*xO?~WRFglRwBQm|-~BuC#Kuzp$8K(xgKFYW^@hYz
zm0o#wdmgGOB}SKvr|Mt7u5AL<^Z_|07owWnpyuhKRDGA*)G0wVJmZ@lrKl#pvv^4v
zs#2x1@fEW4>ObD~x?S<`ZeOT>mu;H|V^coa#m{VcFzXNVsct-<$}X;E<1y<mQoi#F
zH){HoF+UIU^JqK8>G^qRM@C0G(ysOi<~d`Yb2QF#_G&$fb{^)dU>yc+rwP6e!>jcZ
z`U`DZ&vx~PfOmf<+NoHNhxK@L+=0`2Jp2yeeIM(ZX?srRx@NrpT>JhLZ9Jkay?XZ{
z@Cp~Nq=WbPy4yIp%^vIzg8f0TKM0*iWpIBG(ceehUJtq5KDX=>V09lQyE$8q?`i<(
z&%^#&*k6mz1J>AI%kpz3==xu`ua`u=T-z^${Q(8r9}vI4*xypA#HgYjcBLJrAOR1w
zOO3QkEti|C0ifJPN4c|HUeK=h^nR!)A9%mOyw}L{Uh$mZIl*(H{hSE@d|XWct9G&Q
zJT&&3vTZ*DtQ*F<VJFuOGp-NtT^@7a<v~5_jdqvY^3g7P?21S4dLp;kV|G5FQ|WLj
zUS{8GyP5!2?Od<^t~s91$fu|nH+8(6d3>KzxAS$oeO{Tjitm5o`=9^&{ZH1<9qaJi
z(ysDKy6j4aUGYTQ`yTIge^%E;i1z#9wsLB=A3$9$yqYev<1V7|5Y_I)uM?XDvW&<0
zS*sZoL&~k2Q)W^1ue)}?$Sg!z|DvrQ>Po7kFWpAh*?5hU*V2RiZ?CP-k{@aJZYs@W
z-=%M7E}|NqzWnl1s)=vodvr{`M;EiKM{_6V1N7Ulr%zV=_#RDEe2+%&`!%cg7PIYp
ztcMhAJ!EI%^JUc|RgbnQOVz*6H&KLr#J$#j@k)9`NrxzQ`pRFPS6?xfYP7wd8huZT
zwtmy}rs39GX}L(wojH0tRp0#52X;^mcWYLCH`T;@dvxAMRXQ{k^LMj;=VqQ?(lYCJ
zu0!Pl&rwak<AoNjsV46nF#IK|zTtB-{&!q1F`w`4cJ#Wm@X2X!QBC;>-;Z?D_an24
zcGf@Ci_x)u82uPK7u-YZp~M~q^Xloxxbk%He?IJr2k*hgzEgmF@y2=ut#K=__4A^<
zUMa8OaQ){tTDU|@I(_>;J>HroR`34z>2<~r=|J5lpW3{=6IHzL<9)voZ4ZYBzJ9uK
zR=c?2usThtN*JGeY!2m+vLaznn5w_m!yD#PO@F9kiA7YCPmfMoO4T=M{kRoW!@-<K
zR#8nXU8}}As?wI%x^1MI^5u|uZhGHOACO~mp;&5vgPNy{QuSSKQ>O&g@QiPIl%ksW
z&f+Cys7jU2##hMFtN+kHAa?%%uYBYcZu()Hz7J^Aygs`)*%c4BeuS$DKz)PyrZ_DN
zDIZSnd>d8X>u2!X^sKIjx2T*%wL4nppCiAn-+#mV{T<w|&s{VwZn4)bcBIsMaC|Ll
zj55E|t#+kP9Z};UYMfWs>ARh~+wJpexk%idnK3jA{>U<wMo`iFo#Sw3gS1K1ed2`9
z1%J!(|MIvSu2aK!U%4^2Wbw-T^nt=zys#g<q5a^qkY)Wt|HSqCC$7)|pOod-=c68X
z^BOO&r3dRGu`W`C^*oqoZEoB(TI(y}(|t7oj<v&_j^88uJPKb9Z5ltuJS}g}(~8#T
zAX?Kc%DT*`&!4LR9BXGV`kjyZd@OkVaXpjcsiBUcqz^;EMor%B9ZCvT2*d<p12Mfr
z%|e0LtiQqdU_!7^K<XI^7RfrVNFBs9PA%3e6dcvJo-a+w4JZhJ00@8p2!H?xNK(!v
zgS*Vly^pw6kw-TskT#-3V*EXS5x47&Q-Ql6ASQtw4;6T-cE`VEs8}3w%2irEyD8Q$
zsy-ilbLE$fPRMT8b$tW3J(89wi=t+0p-9ia*>DwqA0_4)ljnKKAFVnswa;g<J=V)k
zYw};q^8G$#mu#0R`EOQtH5mAD`odE(E7fa9{lT%Hep7al%uDs!0e8j-6iJ@ctYO#V
z1nMph<?Ec7zO%y_0=>e)SgD3<5C8!XU{AnoJCKznqgt=Y{`IMo#A`b7*21q>KXQ&h
z`Et1TT&-KXx4THZqO&(WyJGp?%S03vy*=mjdjhL^R2@!%UYq}VHrpYhsQNz6?IGZM
z>ei7De-|chQ+rZxJ{Dg6F1*7%ou^sBvI#qhQ8iC0I(-iM?(>v$7mpIRtJm6pUTJ5!
zJsrH;e|G&IIo-aBmiT$sgI;+*kCyU)bh>N&Bii$WcPmwS&xKz5-p`AN(f1p-yLbEi
zQ`qOsJYlz$Ti*4k)8X6oUV?bs*)NHFLB6=Lzuax*(&>5{@$(_z)^%)p?b7?cVU`~6
z@f%*@n@z7?f4uvPd4;bxeNMUCy)>=pz3;?^BdJsNv5MEO-#1qF>J8^;wza#vk@K>a
z>J7)JJEzxqAbxkOw?RILl@Gs`9sJFvx@(BtYP5dQ>z-A;t~Wlo0|7e;xUGLSTJ6#M
ze&{xLdf!*Ay0i0{KwKaI0?|p}`>k{2e0Rr$&1$7te`mJ(C0iIO7Ki98?d(%6%04f!
zlh3|9>`RnbA(~we>=u80hsJk!>ZgBXyVR-NHT(PS7Jt0Plaz3s$``~70w7?VfZhG=
zh|k(Mhv@76MYG@d6J^c^uXSYRtcmOk^WvSE7Z2{>Ze`#0$WJz0*g*jONPPN{?B;4m
zSrIXgNc&nFz5jao$@kp#@L5?)MP%e9o2$i*_~waR!(^vai{+}ki5V-t%;|TVYL#Q&
z*R?pQuk-pofI1$Wg8&HFPk`5Y8zo%(Q-l~n00ejuuxnh`u6PtqtoqZOnqRA=0}28l
zpq~J%^*MHhqd#SE2?8KsmVniDL}uglia%E0i?b^nj2B#6@`ko}YQy0l4$FG=THox?
zoa#?z6|d4CBC4`_KXNLZeq(3ck+=7LBC7J57ld&2eaFDA{e#)poCpyFKmY`sC1AE4
zW!7rm<LBz|&D!Z!e`puzYZpW-C%f81i09hxXgD>#6fN-@t&iQZ?pAGw+232Y`BSG$
zv~#`vXKr<;9@o-(K3K2k(uxPyLBP!fjLv6=`vOnZ?)bONi-ViX3TzXHfL-G@yw=6r
zMcS_MTEsK(<Mf56WGxPMk7pykYu7Pi-*K__9oL&*cGaU=@z^ylnpHf!+I3dL)#8im
zAmAnfyw+dx4%h1Q={5g$mnp0Lq;}h@#K&s9M+pyleF9qFW5V?tf3Hf-$F6>X>*pe_
zIz@n2KaD8iI>r6obG2^m-tHnXs!rL9c%y^B=eIqPYi#ozexn*4bdXBc1l%@H!>m25
z+fS@HBRmiQ0c`>mYNoWhDaRR^m1@U?J0Ji8AOHd&00JNY0$wIy^?Scw@$b}r2~Mx~
zG25Tult1-43A5+Co@rL|dF%>@b-xg+@Ob49!o5O3Yd$OM>)_3nK&WFVY4D`sG5>}d
ztPqF|WDCUf4mAq}Vzd4R<AVvoLIJ5~C|D%xyka0G5YsrdSg%lURNs2OG`7YdG!Os*
z5C8!X00FTHJWxIT!aqCL61S?@bopEF`>21y!EOhM+m$~F!~g;y00R05eBzs({PUZc
zL{z<L>4*v6-f^6`UE4p?IPKl^V?-3yS9<JswdK2STgP<WL84Xr-Bqftf7wprwARlb
z?EJ#l{}8*>YDrs_yTHjiRvad*RIfdSC&kX2Jw8Zysa`wa&KQBEL!Y@bd$rXv)Lk5S
zJx9F5J$~Vb-G9|tKv=7>QVrK200JPuo`BhYH&&KB_Dw>Wy2&wquW7+wTISfCC$}HU
zm&5j?p{d(P7w~&UYn6EN#kBtMeo^%0K}om$btsQtRBiS1ovHHJynd&YRXwT>=VZTQ
zCC{usB`d1Fk8^tnWdGr=0Yj@VCvH=FQg1$%p0gk9f7htNL^O@x6w|J1pWVc$nkN;V
zK4bU%(=6qqJbt%p!BzJcczR5X-z)mSXA8HrZhux5MU~$_(e$2LF`p%$J}L9khd&+j
z<l+}L65i_8c{zIR%GRyivHDL;C2rRPb6)<tVvP-imwt7+^jm%Nt|!9l&f&fnzYUYG
zpx}UVVB2%RyOmnx&P8+c#r{K}_HZ!ze&crcF9ts9bF}780tJWvEeG#p##Nhm$A01#
z<z0_D9sX-0AO0(C(jEc@2b2Rm2dD?GUk^B4Pa}Rl1js8r8^5)AFA-Jswx7+?+j$~!
z%HPL#5z*9adiDDIbwJ6)FYZ4fi>7+h=ajpC%iB!5+;NK>Nu9EfRlK8bfBV6$_Z=p@
zm6g4E!->=%t~VT`?!NkXSE%G8dkGXAP!7a-4o(!Pbk9w{ZYFlC(fUQN`?Uve>z*-w
zoorS0#s_yGU?+hY<sW%;)~bCn)L9M%mc&)-_uf%iGz|@YJ)`0D12S*5^O-<gAOHf<
zNnk>$aXX|Ix&3DKl_`%@e5Q9czv$Ym+hg6fEtLGxqS0B}*{3@8$$~Y<r%slk(aC{*
zi4rTs->Ob|d($UrL{xnxbmxpeGiMOH)T!LPBz^l~vwt#WQMGH!-=BK9@==+WI+ZVo
z7X(1SIDxg&56ko3x<rP81IodFp9AAhRCM?dwVevP#eUB~{#nbv=%}6CpUOX56)ped
z?SINpu{pTEq`^WE;7_1Mo#Kn1-+YufrCKSpdp?=o_-^X?V3DTzGl<=)#Yuhr!l6kM
zyS{XUuvT?EI0pd`u%7_$^?mlI2r+^H2=F5C^U?ONm+SgB@ro*(I_&-Srh#2jWvg1Y
zW2yZsmMoO*RwW%!5C8%F1h($EkQw;;7!gJFrwlGZ00hhuF#BB&#h=-Dz2Z-_{WCt<
z`s|^Xnx+%G)GPUHS32Cj&v+#J{@7b6RC{NfUsP><W^S$K$1cgDYQKy1&X@9?mPOST
zW6tjE)aAS^suq%$<o;$;4!@{6D0`=^|1^p7^H#g+Q^W%TAOHe(5zxCYo!XsQ`*?p>
zLmj?ZJKgHf?)eCa&zX53X1RN1{0Hgq)_9)P{B&!547)(U%>-C~m&(m)g>4|9Md1CP
z63R4db&5dU!9gpZa2*7U5U^`J^Te<N&ko;vfUs60X|>|EOZ~JS{YD=mP-i)Wmn}JP
z@4%x33Jxd-?4AQFbt4PE56u3)Z1)m-h$uRI-S~YM=I$qMQx<uFFhBqVK){Uz%#Q0Q
z{+vov{cZiK*Ew)OwyI9qFG{>_`JOK9(D!{~QOZ@c*!lRzwxfUCxP>^aPQ3>>mCuDI
z#m<{OK1htJQ}!a>=pdl9n~Z97&_OC$6L6~DXB{u=ej(PJ5grJDfHr|n_fK!#efM0#
zO0{Fc9S{Hk5C8!X009sH0WTA1Uiha8t<skeXmoN|ll0s(joW9)(WbnfY0<t@w{@ED
z=Q|JVmffbj@(1BwA<*xxQg!{yb`stS-fRhkI);*h6$04;*#j}XL(M{g*sQ<7_+Ub?
zP(bP#3Kq#auNa64#57JV)+-bo)wiB6jjas`4Fo^{1V8`;KtOB)sSTg1oxge-ajS|=
zm%sJ-*AzSW>BQy4?aH47VgLaU00Dgjs*UVACZqT&BC5ttt=XV$wN=FJTH>GVi+}uk
zB@spSl^*+DZCkd#WBRWvNVKY3p40b9|6bqxwvyPTR!ds<lS$K$Y*|5Asa|{9bXeB)
zw&&9bFV$-Y+!-UV<fBTjw`sYYctr<&v-^=VjTaNqwdb9+_dFW2oQS4RmCyb557|}_
z)@rO&!!-zi00^)rV75MHWy#DxE*DrU|4F>2y{E4ldRM*q1j?5~i%H*=`K*2_@ru@|
zl6&ur1*t?7ee0gut)HJipNOj28Z4au*dJlyl(MQv)#3c|<WrN&9A7{bRo}<CJp^jJ
zF+AJ*tydDasXeJTA4~h~owuwQx0HycPPeDb6Bo;8=~7!yU&@h2+^$~fhf>mU`_q$e
zZ&`Uc5k-e|IhQ%_*%d@I{b|AMiw{gmBfPZr7t_<8dU6Tjtp($b-nwn=QX;DQH#K~@
z(of5X+m-kDWP_ZyFX_{36@h{S$^o7OqxAsq)}>qLe|ps}(b`jy?0sY0cN02&vYbG{
z;eX44+vY3ru1B2?-|g@7uuqJB7vvN2Nre7Ev*pz3dK&TbArPsZ>h}A3=jocI$9f)~
zUE!HcuU>x*hfZ%0OqfSRQ@!bP%3a0>-<9nZH=RUMr|e@DZ{s8P$boHXgtxM?S8q6x
z=I`qb$EdpzzpN@<Jb4*`f&<EdIM0Do<CI407rpMktXJ;O*Y8<Ltg7Dl;0^@rBp~X%
zQ?J>bGc&t?_IHbjXj<%{g@GF978BlT=QDx0KmY`ylfcHo1t00tcO@~ayyubG75<BT
z=VpH{$0`yn8l9z`eX8BIP6PJXyMBg!i4rSBtNjE{`#)Or=j2v#D~Vm|RPK%|{^7?<
zmoFlsD(`h&PUQ>Y1pyE+PQa`4q>(On@6Rks{uqCvqQi&V=R?6hG1uQB-}vTRbkt7n
zPo>#@V)(Px&(4~k`%4-u1Ofg8?3&N56^~X5&93v{?Jw8xhySSDJB`?_TAb9^jjl7-
z>s}oX&OrbK>?gpxezQMCh!F%pfENK?`=L1%uEL4cIw7<BPg?b-qyq{9AfTTB@Aa0f
z!`GiOxC8+ZFiXJd_u0(G>lJ@awdc&n%X&N@Qg~j;XS>po5i@V}eH)e$QI+@lHs0Ys
zlE2hfAFp0Qpw4jMy-zsr@I{>`t=Ro-wlBww8<t9Vt6lXe;sF5=00FxQm>u6XYoFcU
zM=@)sTm6mx>zCf|KeviN!Qp!4K<_*cxWo5KMEYKdTkB)k1p;m+z<M5$o6`#0KtPK?
zwUJ%NWE5XTpzh$Hl~1@10!9ej|9yvhR)4yJ@YYECWt+WkB&}B5#!ju-pl!8P1PTr)
z2Y3#=Ru5RI8(H|R+^=`bUz;r@qUbZvrH?#cX&G^wvd9aB0Rp@UjJd1pnPs2%C8Fso
z_1^sO?6}E9RIN7o$7XM}olRIN?{cCJ-|TyW;!ho(UFSEyFnLtF!}Ez%waZ>n<B67i
znR%zvu6{mI%b8d0KL6=6>A9a?Mx0is_Q`T8U(L=(amrrA8yy79)&tQ&2dQLD;Ei$L
zP3ZK=aw592o~Oy0Gr|J_5YQ&@WbRTeJ0D#^SgCeQxB~(p00JNY0w4eaAmC*Jc73N3
z@nm28<KHWZ+c<Q}Y+Es={VEb|%Ild<oqx~o&%OLR5k=o0wR=w90%77d<&{4O_X+{K
zzF!I6Yzc%qhLQ$P8Xog+sKE+>?13DCnBJjgp+Icb-(Y+&Ay_CN^$Z1zWSv(G!~|j*
zrxxoK3XbYq&zHv57=#7_AOHd&00JN&Hi3n^Z(rhnaxrnMicOcl^_K=0FIsluLgIGi
zPXaN300@A9J^~rP6{`8ep+!Vgz0@Py4>|fRB5v1=O%MCFS6@g(QGKPyepj^$EWT;t
z4~s~&YV#E%FK3oqOq|wthWCG?`5%jjU23(YU2cD{%D&)Y!b<hpb8Fo#!y4sWOn9kY
zJK)Y3f#RJ@o%{Zc1;i`5BY)rSHHXY6qO03~=e6t`(5~y|>CY}AtkqblhHDT20T5tM
zAY)<WpThC<J<_1EvgDoctQG^Mg~V&xf8e<JnIjeuC|?e5pSnD4Z=3nVGrDH=vY*Q=
zAa4tzPi5QHx>3&sL{#0ezt!OLPc0x$DXV%^9nOvm?Q<7<FO?{&zK?Tz2=MAZuo|vC
zsW%_Xtxsjf2P-ZjqG|KLr*(|4yNDQ7^Q5BF=YlGqj+nkFmAGANwtb+fzsY>!6-|}D
zO;~qsJ`qLTUOu$!EB<O+&z_AI5?=aN<wpwnSA+>~y)?67Z0pQaBC3k=UV3ZQ`Q@wZ
zqT?0N8*;es(xfj3e!Q4K!2#vKzUP2<D|NuLFTXOh)?xy+hl8l|j<oDH`o0GD=H7Ef
z<Hl;<39E2;*P~8{zwq=!54G;Fh(N&s<p9qC>VfOm15Ve|h@TGut@f!X*Y)<Rn5Fkt
zT#-9-^!<~Fre@Qt*Wa|xo11-l=Z8cz)tf%2++E!E-oYU?=95V3lzpt?z3<@A+d90s
zgz#2Y_Ua9%`Rv}Gt-C<iU1+G@aE!VWZQL66iaFl^`NlTi`hWOElXcw|61&xC{i4@B
zt9=pl#s_yGU?+iE_nyk!yLSPBI?F-Sewy(Wo31{;ZV}<Fc0Ln`3j{zQItgTae<|E^
z*<xZ=wce9OyB=-xK%%p>vrl#V&pMs1Iczb3MkfdMB}%Li^^V`T)t%kXty8)CRH5^0
zUmKfBL{;AJi%#VW;spT^FiwEiIwyq7cO4n{+QV1tiHbD7WAr_k*S>&S_FKI!>b?`O
zTkL&9kbl<l&+2mz|7-}Te?N@=UbwX%fR|ln`_<L{CKSH^`R!8{6IRNmG$0fZ5Ssw&
zeOir%r$x`T^Yl3HbFH;cl@=%U^#PAQexzHWC4{xA<H0!yfPno3c<)zZe~J(z2!H@D
z0vDIP*ty#+sl+R)aH`oL-)lK?FCbRcYx~$Le4rrUMFPCWTX==*#XLZoK)`JTtgb(F
zTYBua$LKhQ)9zXA52iO9yYoZz_-dpiPa5)Dn24%>x7hvE&%Q7bRhzxga@@A6^N6S_
zDtE>1w~F7DQs$?5L{vTXRcwo|vj0Iu)eXZdwfRz-LU^m_^)&1U0T2KIH375x=c&Wd
zd*0}=itVetI)m7ydhHU$9q(~LgfI5@LRilSL3n)T*?_AtuA<b>uoDE_L|{k$zTImM
znNL`$o6-hbKtL}6yZ1jpd^g@FMQ^^r9SGcr0Pp<?mGHIZ>s!5kqqMQS$7|ge!Zr52
zNLK6G5DwdQg5W7OPnLD1nXpEb`QW{{+>?nYdW-z$rdcn{A#PJM?BNFl>?W`$N6hqf
zlZY&u9&cA<?THbSiKzPeD^&w;OeBPr+Ffqc`$`x+x7%JNzE=}FOo{6^l~`3JJm~ca
zG+#0Da%M^TT_&g{4tBNYub+#!>J)*uPhFn2x6ORwv^w?O6{onHU03Cly@)qD2rS6e
zVrsY0LSj^-gAP*3n!u&8)d%=jE+E3|%!aY8GgHYWM#`Eq!UF*i&?aDYTvF{{J0{!#
z0T2KI5C8!X009v2G6BqMz45$Ouk#V@d#9}I<@HS8{L|y#w7cVXB8sNIyyQZD`A_0D
z<&{4O_X+{K_JIO#wgf^QLrK93fgFLHftcQ*W}!fA*56=!Fd<kdAoUCdi)5Ww48#Ot
z8mAWP6$*~(ThEuq&W2n~OPyUeeiGrOg;)PIb^P`fG87z84%j^h>}mjn1Ogxc0w4ea
zAOHd&00QC?h`;%x7IOWKM0Dl#KF&JaX1f=JetT;fajWtw$ra!IWO>P7FUYL*jhYP;
z#!tT_dtD>ti(Sv}7Zn@VB7LXHc4?$^<30$000=}IflCjT{?Y%)9@#4zZTW}vf&d7B
z00@8p2!McH1ZIEpc=a!X>BOkoWgp@J0T2KI5C8!X5Qo6w3aw^juezH+#pIxrv`6Qz
zT0Y*tl0d-$<p9rtTj~L&eu9F48woThUhI#>ukDnf;Na$R;Kpaij7>X!YqWWMCLyeJ
z(LKHVGfHnEqUw%O)557!7ZB0(r9$oQT@zYGMAr>n!e?goTtY<E%wD<n)c#<TESla<
ziWK_nP=*XO#=%UT3qK$L0wCZ<0yocFe`xdmGei{ad^T-DO7BxdQ8zM$P0>T3`xoDJ
z{k3?e9M|Z+Ij_Hb7zz$32Le3@Zm+2(3=ZVV-r^*I^5vkEgXWKPn*Y!2!-Tgg@j*cV
z1V8`;KmY{9A;5bdil3V$ygxAe4jFna2iE(FSPf5{5(`T~00ck)1VF(11lF|b7Wd~9
z8%TxNreiKWbo&4F%{7}d20wd_R4uVFRP@!e|EJ1$lA?k8TK^m#v4<#{_mk&ZKRkXn
zQ8dT?KA+!Pcq>tBq;}zorpv1R+HwU^>jMwnlCVC{_e8JOekJC@ypii=QB@f?QMAO6
zGL5^FJu=i@4n}K&_ljq4B!9I*hJr&hlmnyBhu7UFEiX2!QmySW)EN%HPMJ40wP~vC
z7R@i0NlR;)DeG0`_a1s>-iX~Cyk0*ceINh={0Z2#kFnl(_@~(0F)SSYa>LoZ_7PFl
z+qr|(fdB}A00@8p2!Mb#f$pE){r%j8(=sd7jtO@_00ck)1VkqA!Iq@4<EzY(MN^TP
z!*=fy_^MEgKO2U3l4w)!<|tAI0w4eaAOHd&00JNY0w7?NK>ChSmt&Ugmi4O9@ZlZ=
zKmY_l00ck)1VBJc0&ZE?0z0l<XC!86g4G}Z0w4eayb0`&E3_!@*&VXi)OTiex8L$E
zm7(C^<#O=t6ZI}HnK4-wRWD4uG_n4Kbuus2TYuj7{?`S^{`EJ3x|4(6e1tm?009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009t)HUhV_uhO&Q7yD(XH4aL-
zY5l;nCCTlhWwZL`@%*3M`|1{1RK06h*-rV^WXg)pd35yR%JIi!Ev-1F_ABGI9hRZq
z;GjGQ(WbLMd@Fg<GY1JPRlZZES9#A@nw$5FE<YdNMikZBXZ)i@?e4zkl><btKK<fT
z9hWW3AiOo&%A;4)d$>ZY8QH7uCRR26=8sy)^*54e*HFh$(%?zMWBx5K!3u$#ftvy`
zy+h4Hf!M6S!T4Z8uuwqi844E3I<FXr3B)u`E!Ha(9M!j;FYUj@EN*|yTU%~RC6V;M
zKG|NWK4SNKZHw%cwe;7x2R>+(zF2mfYS}Bw^*#?&DOi2RL0MG2^}*J+zj*$*EV>?f
zW7ei_cV^1GmF-h&JhTpb@9%i*?o&imjZ{zSy>FZxe1HHy0&e?VRokPayg|`#%E!Na
zoaoiwPq%z`>7Lz$w<`8Ir491LzE^MADcSAn6d$j}yXDe|%|CSBA={<A`LRmJlAfR3
zJaf%?Vz+KTKC-|Uy$%qlSRAHQAJo2Ull3H$>TTEA?XKF4_zl?}O($;GNPL~Xug*8K
z^<AWVFdMGoZ*QfH625$Ke$n;Z4-GfHmLtY*RCCVzxy!6yH_A?{x_!#~T*-g4I=t(y
z`=s1o$*feb9lZOM^x6^0-K~!u-*D$wM~U4U$u6hvd#xSAs@yvrPNd(LXuHqr`QY6T
zs}(-4^h64m_wQ+>_}zYA-9F@f?sgt#+v`@-Wp!N5YW!OM&9?uv!ZUi^Z2WHb$NM=$
z_*cgNoo?5#j$gEVb^CJ|$v)QkXEZ#c^`&0-YCH7$Q@b<!d+D@4b^5KItGVbsXEi>+
zIvl<BJLS%5d&w#Pk?I?*`;j;g>Nl+Wvv0=wt!kurwC?ljU($*%(sifn7rpWG9>2Fc
ze9`)8L~BpkWv|}yYt@}-<wb2T>)%~o;i>Irea^d|&93ly@#ht;U0e{)jmKw2o2Sit
z{L*Onk=k=s?;EuT?!{g|5pDS=TK(e{d$q<pow{zdo;2#8b-N}~{AT;<L}|a+^?t`X
z9#Qyu)y`=9y+e9$+&+s`kMh1R`g3OXcgCzgz4;;f{6)RLJ9>Gj-I?X@v_Ew^Sf4wc
zKC|(&=D<3<5!1K-xjyzRiKOay^*#q~ai=vd$@;pQ6Yuj#_VbE=#Q4W<$<^+H40RWW
z4qL{gmjCvwEQ%UkXXG~bOS&9->EQjBWl^-ky9X--)@SqEt-R_XR^^=acw(gRtd2X2
z>YrCR@M^eD&GWJvFYA6)glBGF39I?2*S}NGU!;0A+V5M<7q`5>!H(<i2N~)8Pn7#w
z{o(5JYjs>w^nP8%eo>zXv*l4;4xBnS+YiXvpV{=Ywgcg@-M@o3PqXF5EB-{SHx;|>
ze(%?t?|R=8^tu!MeNS&X&AQW@FOl3?eLsp6KI{9e%Rle>RtaD9@~qgev_C~rv+YgR
z{*?M#uPWt6R28LsXxYiSo>9V6o+DnWl^(ZVSKj+s|4wLy7ipe@+51}YS-ozxoUp3L
z^wwLv-I;AK!QYL);}VIF)qXd#{_S38VfA@a{9AqQ75|a)#qIYM9!@FRPd8ro$*Wz!
z`gcHi9$2Z>{XUG^VfH<H`K+Sv9DM(~h~EKY#an5oYN^q5!~L}ctnO!`lv}IeDE^I>
zn`=|^|If^pZ{GflrjM2T|4-(||Ks)hbmLef-{oZ&!eu*N6gel2xlrE6p;y+YMd}ZT
z^u4@D>z$N(*sN;h?mZ)`wO!Qq9MMv(bQMT=b)!5}3PY`NIMKX9;_&Y;5$LrX?Ee0f
z5+Cb)(0cByUXN5?u)44H{IkBU@Nnw=pH;j{I-no`0{RKCdY`f@9Q`S~;?nEyGe&tt
za?ff!DpEL+?1*$<>1VN9S^xfeH9WiO19d!F{i1eVk5q1P9|S-E1V8`;KmY_l00dYP
zFx$>$X#>LgF9Eaj!2TPOzAI+Sfmwh0tP;l+tA1&%aLkqmqyBh}6G!Ubv3g&Nn_I5)
znwMu+xT5xp+Qo-=JVu{myWERYL&H)K@Gb#X^9YQFqwY_~`L+A{7%g$4{?00P8=d#Y
zd%m#Q@Xfy0i0;p5d4+or00Hv^c)f4i9WL+pAB3-O-JjKd6sP>VrJu&E9bPM+tk!Ak
ztr@l2jSY%^Q$GIX<HW3Lh3nSqyq-U+<x_7yD(>|D{wVI1_Xp_p*ONtBm&EG6k{@nU
zz3*|N_CM%NpIhCro>yyk`LkMXjM~fUd1X}|jN0!u_q@MbWcT+~yc!>?c4ee+c$ZIB
z&mHS<^!7jUYG-P3XVu=eT2Hg)tv5U^F1)YnwNLGiw_OPT`UK4OKe)vo?|BE;r-rK_
z00JNY0;~z>?Z;!~j<x-63y*dGrrz|a+jnj~cS}1ak{!I;IZlOdc3)rko3*LiotZK?
zL>C9W^{?6;>)!#p!&9e6lyilXmc5bs6<OWaieI_TyFc5m@L7G&TPqyB?M@}%q1PW@
z_6irP^Bm0jkJMh|eP5|J)aurDJ}bLek2@>j8Qo_|YkWh!o<a#flB(xBsoksZyY*aM
z9&yf=0ITOHQaEg_mL&CaA=<gsamY}|P*Si$;HE%aAf|VyStt;j^*0zFOb8YVNIgTr
zB3b7Z12KV^#;L`6g@U8{*7K#Y&IN=A0w4eaAOHd&AToi;q42z64b~F3s>p2Zu)X^o
zUq16ej-ABqY6lC%0RkWZ0(uAx-_hjovx~P7QPr)#*J`^u{&lRydwEt8QB+Uaal(zL
z<<x6-m(Exa%yntDY?oRsYpnU}wMShKus+vo2izGWQ2(2^2IMWjgFxNIAtCPTPiyzx
zPDE2)<;Ym6hHDT20T5tMU_`wQ7f;_5CPtN&C9K<bZV6Ac-#Jn3b$h=t>~nYjKCkwj
z-f&sfqv~+f&$IeI&g~(fw;b5*&YskpkHsy&<3{aJKX-c1&6A2wpQ7}~8!bO>cdxfR
zyTzT|^(w3Qc=NlhMU7r^j|`c&y1hI{vX6H@unIp?I&lBWczK1{H$Bt!ri}y&4k!oq
zJqNs7sjTWzgrjeqj#WFyE#V;F_2oM_Im?N6J?eD$Xs3!juM_!){1bUzvMA-(>3SOR
z^C94t_OMwy^p1C%rN??6u3h1oO|M>mX7kHwe|ppBl)D!X4E(-*-pM4AI%OZLc<uV#
zV`Z=2a3a00>J7)JJG<%?#N+Pu-^d@a^2e!p1xD)^z3z?9>(z3vH$J!n0Xqpe-F`*<
zSJs)?)qZ3hPo(xCtNV676Nn20Kp;8^Sgo(T@_WYmcf#%AMQ3SepK7i5A6z%LPE<5*
z>`RnbA?h8!fx9c~8lB2rtH1M3`{&)xa4KIAF9?8uaRR*S1BCCc{h*K^#`44X6BQjk
zFup3*_$u-X`IWU_e{|GN?oXxL`!!*ozV!s|FKMt41o#tpDWS<!Dcc6wDb-4$TiZQO
z_0MZ@QeWp)|EuG{IS7D&{RC#tnO64T?p=hJ+MgoC2m&C$i-1$@6<+ZwoJ8*rSJD9m
z0T9qnK(yyxf6Cwz1VF$n0ju-t%*N{#f0ycHrq54aAd8}QuRHZhKHHU!P~5IZ_We}J
zFRDKG)dOw6eW{{fRQ+tw>E6#jb-Q0wy)9$-uW<*;`?YlX+xJaB_e*)dTXpvvC)UnR
zDd!heo7Z0Y?cJj)_(j!*KO`OQF{!Mdx7t;oA|4O`0T8f@fZ6_Av-a7&zQ(MbZuMt3
zKg8$kzSc(hTkRi#e|LYEF<SG-t@SbN0s%J@_<O*%&u7lb<!7aCPAhB!0WAV{k0T(y
zV=dmxvyy0W(8?!V2LU4lM4Km~w%15nt+>T^9o_L=N3;EH@E0BXY$2b`<+GK#k%eDg
z>#?FOTo!qOFhGDeffey7dp|F?PZmv=kH~(@_jz;pMb*-qTJ0TB{+!H8d6yG)_*Uz4
zb$E82*L?iN11CG?^jlTC>=iX0_4^^tjqL|!H!o4k8SDlDmIRE>Q&8Nqv;kp(fSQ2W
zeou8cI0pee1eAJBbT#|jX!+B_4sJjI1lSN@UH`DL3ZZ}i2!H?xfB*=900@XiAjiaE
z)la3S6KHgD=<#vAf6DaSL8480J*RyJzH-NmzKaOd8V6qagK)19(CXJ?eI2~n5(sq+
zB@Lc5Jm%j}gB1dCfn0%@-l1lpKy23EV0<tkSSTR%3<Zm1omULR1Y#Pe7V8xXj_O;_
zm&Vo@ga!g200JNY0w4eaAmAK<*Squ``b^Gk#O>-FA;b&<AOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w8cb0z>Z1Gvd;?
z9RvyvZY~GcE4#P`0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&00K@Bz`R?`yLI!tTc^vX6VD{#1OX5L0T2KI5C8!X00FlX
z7(MjDnW?ij5T{g&kJIlDZi`=w6Rv{*2!H?xfB*=900@8p2!McB3ApXOERsE5%?G3l
z1V8`;KmY`;M_}XYDI<%ozr!!8*1P!jO~rqz=oeM1F6+{CU3?|KsQN+4oozb3SjMlV
zy=H8$TeM0Uzgu<ClUqv6K3Cc=s)}0f6}xBrwzyA&>ZSd>^?G$5u7LmufB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zuq9x2zZ<qzAT$sVi-6hZM$4aAsfV>7AU1)HzU+fve`W_^rDD^7^&kKOAOHd&00JNY
z0w4eaAmBWK>I;h&i@Ugyh@xpN)-0~EaUF4+I-irU00i7eAk;CG6s!=)704Zk=^bho
z3dCmp4aNr(f`tN7&rq;P)_KK1OdzIlYO!9S;HbX!d};1W2P^^s5C8!X009sHfk*^8
zls^4VyJcr&y;|^0!sz>_9hZ6O8zWx$<(7vI$&u{FeGmWv5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5a31NP{B*TAG~89
zfkr0>UgZkmf&d7B00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?x+?YW6hOs?==$lSN(SZkw9h_ZoC$UR!Ob>U}1b*q(?#{=*
zm`g;}vDJSV{oAV5L{xqG({>LPzPOb@o#7x#J*s7==<7KilKMAWH)=QGt?GIY=O6$A
zARsz{-#a#V`(l^5vQZU1-LM}7KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)
z1fqq&{JcGLKjl9q^VVqLiFASh2zZA8@A*>RA?0jJwC209zV2))5jzNg00@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@9Uv=9h&3?&VoG(6_t^ad*gatHDR
zVtR*~g#xize}nPCgkYh7)H4(;l677&5EF=LoLa0`C^)KbJzv^?G4=jxYeK&7GDsx-
zuTQpDzCE;d-r<c$WzqEZuNv*1Ghv@Bs!r=av`zbWcgUjZ(4psF&Yf{Y7F}yqP2MqY
z%rTj_vVCfehnCj*i=%7m9wDM?%H7}182I3EVpWZkgAWkkN1#&irz?KEb^~#%Dk-n_
z^_Bq-2L6`y>UrNc>q^8dCcIU#&nYeSrZ08Y<}Zoc)hRw+i#J^O-eyxD+(YbA-uwn{
zy>D{m9}f_#T0XD;`*R;JmF?Eb72E!DQ~9YxRP9rI&(o`xgh{k%_RJIC{P9*MiKKVm
zwtLm(cMlP_>GRK|x9^pEAAz#vV79*dtm()F`~TV_YpL0A6@O=k_W$D9P5%(l)oeYv
zziHLuwb!PTXj3Ks&1%{8z3)zcYZqaqdhOud9@T3{BzH+8Ccbw2+gpg;8p$rF?oZzQ
zV%3o^UXs0{qW2Ff_7@y7HGbI_7i7`&_52f>Zkn4ZvsPZuhf}{VywVdXTv6)jNa^5p
zAHNINk~RB&eQhYl&%X8L9J?a-c4AhI$EeSTQa>A2eW6EoWmi9t-SI>kmv-ttt9G<o
z!m&GFdB<mWeQh<qH*atF)a&=HBv5l4BG-`ST&T;bxlkm@?C+J+{#e&fX8D~?_@v3#
zziyS?re@P)mp`|T=SH&2Y=3qnu4db<YJZX9(Ynv;y;3W_NY_P~_jNe_*q9>myJfHF
z)EooL{?cf^EQ&^YUfp_M{rk=O+-yB%)SpxDIjrJ!x*el8{z&7%R_|M#H|Q4sYi?e4
z`;Si_BT#TKQV!fUe<G4SHCC03Tig36af?Q3&sn|imUKpo9iq?UP}?8rd2;%`+FqR7
zGoOf6JCap6R_jTl{#o~1MT+0-eD@<?U)Z;$)>`6rW!*1nS9nI>``qqcwEENSJR{l1
zy8ptd@FJztuKSVdQQr6MN++*)?0)|B<_E9z+7+(ebufB)sNI=uA2{t#oetLLyywC3
z3ZFFx*5OUb_IT?-TXOm%sXAU!#zz!8wbp&GzOLrP`#e%T&ntd;?Jtv>Ry-|xMc2Jk
zIcD&?CuGr;_dXS@!xz2$h*n;@w7BW$>G|hmyOmcx#A<w-*ZdEyaP8{Xa9cdQ%7ItI
zoxko>_Ak2Zmqk;(`+$hz?(x1|2L15YDp@qGy>sf2=I{I<^H!&xzmG4?er$e)-(~3a
zIatjX%zHOC@7*qLZXXZQYoAs-NnL)ewzuDzvdI^$drB5vMP2`G_We;^4xBolP@{Tc
z{%vPvr`2rw&F=33e`4<=rR1O4a^n?$yxYON!&marF4ddwtkbRZ+l#8a_iNLeFOl3?
zZKp;GpY?s#{aLK$twjn)t9>ilb-n8~;O^S}Fht!~+pOK9&lBghenxaItnF9oP1e5?
zTH*D~yL#ngUDnC0RVzNL*WL2_V09l2R{nXntLzG&6%V_@v5Sj39=-jCtlY60PqzBp
zEB;xf%c*db`rRoNwH@k~df%)athp=C11s%*t_JDd;IeF1joNXb*XR*7-^=Ycs!BZS
zdQmxNJ?~cuS4$Oppymh|-G@-^UYW1TYQGhu^BJ^K?ezLlt@!O)7od(ui<9~~uX#P{
zc#NJKEf;X_%={pu`5MW++24Dw`cpsG>hnnJ^wjRv=aJ^`N4l?-pOM=8TJJfL{MmJ%
z*Lobg!WFeX<h}2l(tfd96AS-P?v(}yWmc;7+*!SD*Y|2Kw0xp|t%8ST-m36m^*pfN
zuR(8kN;;q*00R06X!Rqw^}7C)U2&=P-szR=MtMYXuhs5keLa#Lk?ymeXQDT}NbeI~
zxo>qmfOY?}-thG1BQJMKxivd4BU=0^oZPBL`%8GIL$vRfsqGbgJ%`zGD7W}i=Z9$L
z`yT7D;<XM1{i5pSVZBG>x#;tYs=wb-zUj&L3;TI1?|KyBgMgO_L>iCKy6<HsZb+Ni
zeq*z?e_kth#_+OxiCt>ePPh8wHSTYBxU9xg)Zv&d&tCCo_xKp=_*jki*d2~t<HNk;
zv1^>(Xgo&u9m{v(Q~v|;XJogj*?kP$=5Ox6O7TZ;*(8gqtk)ge9iDo9vsQbB_4lRp
z?qkf_o!)--XmzKR|614E+`A&?!TuYGsG2gPeBF%Ow-UYjdxso7%FfzB+@|9u{`lMO
zr2Qn?RI6M%b)9v6r8hjM(rYx{tp%T3RqoUo*)G*<pIz+_*7FRM@bsqBsyn;Zd-9IQ
z==i7G+<RpnhI&4_DCeT{7lpH4J6W~Y_12HP>L;Z>7e$Tszu4toZ$2vS)cwC&=Zbr+
zaVAl&^PY#b^Y^}^e(Q2k=B-M8xJ~u$YbMHbuC~+a{7I+%v+ifJyZl+T->AK;*6FYs
zA2Vvd+uZYhpAfx$>K6O$s;79DPrKS(M&r>tPn}nJ(BjT&+}CP7t+yS)+MN~`-q+3U
zE5h3@gnxYktjnF<;aPn?Bl*8RHF2&Qt<TN6kHj(B?z8&+aXOqx-w7F;uh-b$3rl{z
zs`ccs-WL|(m5f{4`zX=k?bhq8`hoR^<5mu=?9$u+uGgJjyLi7JvkKp??_gMshxdF>
zr^1i6b^+3B&-Wy~%2{vyt9EC0yvQx>D0O;7Iaj!A*=zQF%cwuC_?7FTjC)41lXrjb
z2Wvjdd3SOSe<UsR=ctF~<jvvts;c)Dj@EOh)?<ACMeG+JwK^#~rAob_R;PGGir20C
zg(-F!t<Tlt`EEHknhw_P)#VZAYzZjkmAA6BT9SB;KiU<p)%l8{j-jMrg+QJ_-at(6
zP_s}VHtTONK9~?J6p(s`f<>~<D+Xc$F^yA;^$G<?^{wYi`)@A1{_y>2-52j9k@UYl
z*<N|RW#gfr9Q#|=(p5PY{@PO7B)d%?{_UZKmGiHXMb$pd+s<ooY^N-$KJ?_s>^t@z
zmqphcXJ2~%RQhR|x3Ya|jfYkx_Jz$)lqN(}ePh8}Gqc?nmaVFBa_|8H{0NK=#SMA?
zz$>y_RZ01{H;>%))4YpBum19A&VDD?94EY0vCk>}tI4!d-)uV~yIq~)<F$C7xLBp;
z8wd8vb}4Uu)ymBw7dx&bR@JF?MWpv9x8HYby|33U-tUEa?XcTj)q@x74W74`xLtYw
z{#p&c!MmeZwcU_O6g6AlMal=W;VS-?eOLZ)o-OBP(beqxy4CkCR{rgN&XxQ(tE|$k
zgrnCE-u+5??TF;gYP-X%|44Q@bzi9my`s|Kc{3)=IzXTv<G|?wQR+Fnl~Zye2Fv^X
zB%%Q>ci<m7prdcWd19Bk9EW7K-@$6StiI=2jbF>Z*>;Xrct)=~)$hSO-n(lYnzC->
z-^6Wdw!P<8e{SPz)gIpEl6ClFcD}k~{LdSSsA~87usS}Y^`&0->ip2_Pwmd`b|kC#
zSl3r-E_%;d+rcV4z4klhPW_ySa&ET%kHnePa}g;Xt^3<j@0>Mw@ma!Jwc?9(-D<zN
zUH-csU2-}1^3%jCy0A{q%HgR;2$U}ev-ODE{E6Og$?Er(cR1Yp!93Cr&6^YJ@YU^j
zoQwUwX;=QSYA>=1CsKRP>V4kzE5c_p&&{nIS@(;v4v*Dyh;WRpgK%4UwOUUa_0M`;
zPsI35wlhU}A06@H19=Z@jtz^ZqQA$9Zok{!>y6sO`uEVO@Z7RqHex->*@4mgcbj{=
zf7kV9hh6h@c*Uc4-lbk1YIklcM|$m1r-S*q$##?Dzl+R4unjQMx(MaII#yBIql%rZ
z`z^J?Q*+{dzPj6kjanSYl(m#s{BB>*8m;!3UBBa$zd}D|PWX7&0a;YFI`7qP|7Pb0
zTlL4Q9%9vhWA}THQ}1_9$IGi6cr{$7^UG}fc0W&u@A~_lY3<v_`nprkpQ!ByyX>}_
zFLv39c*Jf8Tg^YM`bb@V&Hf%K{_L8c!aE+V{s*^TH_Js`4xBpwuAD^L?K&wtt!C42
zmOuP?V_u%wa^n?$%I|=vs`Wd`>$={2XI($oJ$}hLKHl@t^yW(>cV^eeEB+#d&-%XA
zen_={yZXJYwvXN7-|qfutMR#IU8UZ61ByG**VEnKZ*$YcZBkfNz4&aews$VbA7-t*
zd8~fyjn}Jf$`%$ym36ADe<!rU<Mmu=#b@=p)pi7{b}sAok=gL9{vNo+ztVmZMOlrr
z8V$$pb`PuaVfDFJ{Ijyhsc@{;UsnB#@_ec7w3=?O`B$C?W@>dlsX;@`{theAX!SR%
z-)F6!N3v6~2WpOh(%v<z2_=G$B(FTlY`+y{ey?8DO10H;=QaOYoYdEYE9T$x@wT%v
zYgNZ%^c?ePu0Nl~ZCs4zYb5vIHu@*2^xDfZ^a_Va^<t#;F_FgOmGuc`RVzPBjIYsY
z$kRE(TAJ<N5eqwQ$R1W}t&(YFXRpf<c3RbTiFPhZ|C`!Q-s`sW)@%Atjk%YJ=&JSH
zS-q~cj)(PiUgO2;a1|br>PffXSJD9m0T2KIg@9B2s&>Vzq?47}H6G8(UWDUL0%qTr
zME9rUhS~8`r~KJfUn%h@oZPBL$7gt_!>)RnS3F9-i>kc1*cGnZ-`^tH=TyDODqg$t
znO8iv?bE}%B1QNh;9&wr+bxQF4<m6!($xK~de6-oZ})f3&D!Z!f2`L_sl$sj?#SwX
zq;{3r`@Gu&tkTab9Ny{Y9e$*7bGP1)H2+PM`(~dz-u^`47cF+WZ9bDxdqnN;QS4UQ
z$BJsTS6I)>kQRM0WYvmXVY`)g{UUn!TKTVa-RSe;HurkpQxtb@Z8xaf-_dfeRW6;n
z&Z~biQn*f~*J!+UmrK3&@m?3LgwOi>zDjs{(`nV6UHet=j)(XD+@~ro@3!OcZkd;Q
zZM}+U>9KoVFzfhumw&s%XSIGwZ~dseFR9gPf1Elzx1aOsr?C2-pf?{CclBF0yZyVB
z`TdbpanJkriFf$C>szzyOO5(d^22Q!slBe;=k@&X4%hB>JnQ&a^Rc`9*<C+cjZg8<
zYQB-(yjjPm@ZhDa+C$MEj(T4dUgxaKmpVLF&mHS<Sl8=D!&7ro&Xw^It7^6VZZ&?h
z{uK^(>$o?M-1XDEi^QnfZ71TpLO^R@AW^QHeLnR16TO`1O{Z0N%KH#BZpixwUeWxX
zChK+hyu(w*O|{f$dqV4Zvids<|8LFAc3)UlbI{`L*6YgtWOk`rIk2)z>$x?1U9Vld
z-m_ST%ex<yRrp^0UFFq(f^fyye-8O%ET8n&ziM~)z1Mr(*BdU%MwRzCmfrBy=@I2z
z`JL6W*RFb9D}LoVtKTW7!%=wi(wvQR%!o_B#UDv;=~#N{bF1_D6`ghHfuB;668v7#
z$?tDyck7-Qf3)idWrn3!`#Xo<Zv7?ivO4c=%jtJYm3l+1tlQ_T!eiB+$*aCLT3>0k
zo7DEH`vbJb&DG;B>hRV6)c2Khb^LaptIH$K*%DClnYXgFT9QOx*Q0(ujh=@(hLQ$P
z8Xog6*I<P}-ax)UOz%*$P#`wzZ!kWX5G)jsdWM2Uvd$|8VgfOZQ;YQq1xNL*=SyRq
z9S9EuKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;
zKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l
z00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)
z1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;
zKmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l
z00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY{H69{z-
zB?T)4@&)n-VtR*~g#xize}nPCgkYh7)H4(;l677&5EF=LoLa0`C^)KbJztvn_}~)+
zKmY_l00ck)1VF$Z0`uxtzcbIQ9mFeY4>iOB0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
z;u6@tHs7=DCrf_OR9xb)90WiB1V8`;KmY_l00ck)1V8`;KmY_lz)b`$#Qr#bVR$Ej
zMi&P+l|k470w4eaAOHd&00JNY0w4eaAi$o$eVf<L_+-&8qNV2#9W1smyo=~n5}Wgb
zT(fo&*2+HR2oVH800cmQC4qm+kGc1r*E5J-9db_~HY4{=qE|ml-}C;u@;iysIwiGL
zj|=N|5~CXI75~EB|Li1IRUVKrKiB=c2rFe-1`rkqI7Q%(8KH-IKAKL<>dMgIsYOTa
zAWo@M+z>AafB*=900@8p2zZsisLu*MpZ3ZeqF37%JNDX{)K$c&K01BuW1G%xCcIV2
zr(AWLe!q6iE@D*&we8WS)sS6;mmVJ0=ZoLB?jmYc(g6hl5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009u-N8qOdk9Y2rw2Sc4
z<xjWGUS!lR0tE+@19r~=zZwA{g8&GC00@8p2!H?xfB*=%l|ZrOTbj(h=bUU+r@!_>
zt}_Go$)c%S^96Q+00@8p2!O!<5*WEA<)h$-I|-C6hyP`a|APPsfB*=900@8p2!H?x
zfB*=900@8p2!H?xc#}Y=V<>6xq~S6DrY=|^kUtO~i0K__77D~>{SC$k6M}^TQqNGZ
zNY;79KujQ}acZ$%q2Q>#^?YgGgo0Fo00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H@<0;6g+p3rbrI)Q=%%E5o11Fh$6!O6XUzPd7lu+n3PUn$%2`)!1^YT1wL
zAOHd&00QnHV6@&BwY|pb_tGulxTEyI3J|bHAn~K2e-`<Av20dH?izS>@6_qUDYce*
z_yz$G009sH0T2KI5Kt4CwLU(7`zD8Ez1s5DpW78Foh}>Imwqqv<B5m=l6k9=zSbLl
zsXp@c9mJ}Zc(1_B<FOfpmu@Z8!T;Fs45C&g9Z(Pe0T2KI5C8!X009sH0T2KI5C8!X
z009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0q+qA
zbqpm1D+J;L1p+a>L(M{g*sQ<7_+Ub?P(bP#3Kq#auNa64#57JV)+-bo)wiB6?Y|tp
z_^eOgVCS83B>k^XwpW%9s=lh>)-JM^{xhu2<c%NwN!+GW4pqFh)cCcssOmPJM(x?z
z@!M6m_56!K^>JYP)EW=1;m8kq)-8TWj&?Op4n9DDAA!m*ja^qiV-0buDk+cDPEp@K
z{KB;E4=*|?^H#+^r&O!m<hJWh@$p)`cDHwU^J7&`+s52h<V;|jY`2Qup5a|T@(%y6
zq$zXHRy`vd)lc*1+W+c(mq{dDey-r*{)0~vx9RqiMSk5-`8a{H<zTkHi&W0dhO7Ad
zI#=GGeqDKih^{#{J&?3+`axn;PfmUPq5gSOWv8`LTHRf*?cPi5QYHV*s=8fl^<1wV
zjZ+_ObN-f{gqP~IBa%C(`tKsy<<xzn&n4WOtIwi^8?W~jUe5<Fdn1L*D?O3I<^BE=
zDSp@AfA)9(fY!%l$u4y{5VQS+R+D8mUd5l)__h2SeeYD<YlUa@y3uhnQQSM#E<?Ox
zk830UqAUN5)|YzStNW$(`cu0TwLHEN`sAkd4fn~cRLw>2Iji_thojeir`(D1JFm9W
zbUQl&=ue7vT|FsxT-d5ch(_}^uW>2O2qWB3+Qn{HUj2toh3j<xwbA$!n$M2ec=D($
znp*AWj^uw#`|;gMw7o3z)){a7{9v1yi!zii2j2a$k^28cxzC#u>+tOwr$an99v5}$
z_ttE@ZhIahH0LbeHp>bA%#Cxqt$bOv$L-|+^Q73$lhVuc|FQSxfi_kD|Np&aQl?Cm
zkTUg-Qc@}V&_FH`%@HYeMN*U|Dq~1V*HETV8A{1eNm-JJBJ(VnhlG0#*X;Z2U61qm
zy<ff8d+&YD-shZst$#eN>9t<#xnAqF#<Q>2^Le!LqIQ%n1;}&6Xv22Z+r2~6r3Yny
z=k+;WtmN74Cr!Gu-=jv#Q?&D2yvry0IbG27v-6{Vr@K9eJKOU+?T=XbbHD0w+Rsk;
zkolrrf5fZ2Y=`IJSiSG^x*pGo@K#{`noXOx*_{-!Ls@I2^L)I@#cEtgp8O8m`6IvT
z&uZtXe%=4ZT5i0?SETjEYFzqNkKOo7wDd^(DcW*=`5~|NHGbKLPOeva2ThJ&c92!&
z2CaY5_rQS*I+U4vKxGw7wBMII)hAZ`$7}pWJ1=GR^3nE7P~S%zy|c#e45a<a_sDL;
zy7gGl;uHNoU#Fkj{vyA$gLs7##ZDq^e^#IGO#A-#J#D0RlU2(3{d@YXCf{kld+irv
zeQxS?eO~h+<vHyayL{2M+fKgM>HI$5D~$N}Y%cE!qvbhK$_H&<l_=-$Zr??||LFsh
zCYSwL1r>Gc|DeSCwT|R`)Gv(Hd7oGHFfP2FCv@`UJnB_gS6`&@yl3=Ezi8zo%KQ^6
ze;zGgUYGZ3yhSPpdyF;4`kjeaemT-{FUuv$xQkSec&`^@<uAOh$FKEAQ02;T5p6iy
za~ikxM`|C`qx-&qSNmY6a?r<MWDG<*f8tLgwI^8)c}`K+>O}P`k<MGumecVh<Nf+x
zysXD<*r`33eDS?!ciXRa`Mp~IW)(ShzoWL>pGbe}b=a;y-KIak^oQq1%swX+Z~S!5
z;+Ze@$`J}G#_C=nX!ST<XL_A);p3m3`fTWN%PVv{uAJtN)jCS|odK`oX?L{iDaw28
zKb^{>9R>{ph5_dcFm8i_&Rh0Z&S>}ZiC*3<A9nd;b^mLphh4hY^Pp3CLHW*4w0cJS
zeTHc3kF>nk_j+#SVYi9yH>7?}*I8cWgQrAA-nBbLzuOV%y>OzGb32X_Eq|=$MN`jz
z{~oCHS3wIWnmxPKCsuyMZ$E0}8wRp615WobGGA8aBoW#c>AW7P?y<h#i&W1<EyufO
zzc5mJa9hsq{xi~aw|cnE?-%c0$&Hk^L@DRCKN2Os)BSj&=Zm#}c3PiX-V*hFQLOH1
z{jSHW^EM-=O|3Wb&$Zn0Am1VD)+=6*1H13N`jsA(d%$?<sq24TytJR7g(JO(Nwjj&
z-e(21yv{D;6(5vw8!vuhb<RpXPTLtP`}fN~I&EjP`Ffm}cTel<&ZwZH)9+VCn(yF-
z>qhr2y<ho-iLzgfR*ytEFL0}eSLZD8l7oLV_@CbicIqE>^Ssk@2wgt6b{Oyck@`ow
za^rIpXm{+f<0?{g+K=S3`<)Hx<k$6IP~zR*kGstuX}`HGM}H(5Iz5MRdmfo6`GaET
zIz7F9UlVKli&lTP`g*lLlK1kFhTZC)X!+gxQ@8n<uM9@_ca3A^Z~eAU-Fh|N@ow$f
z>-;+2ozCBO`MlOQR`PVb_#GeVe#URTjQs!3K&<T1uX=RrL$BihJ8J=ob@`p)NWTYA
zq<;Mi-YAzRWYF;NzwAIaa-!W2nexMyl<K)bNgi~3Cu+P`e%q;>MCHLNy}aHZIMvJR
z@2~hJ&u;ynDCu7JW1`4Q6n>2!k-a}Oyv3F`r~X&wv%9ZL^mN&tpoBSp)ae_m_XE26
ziI?j}6aMe(X{WqIdJbmF)!z2__MIvyJVg6`D$???@*BE(@Ea0!`QsgrAC4CNI>+>C
z9VOTGQm)hAfkax)FMgcLmG)@D-ZJ3Sexl9it=36NUcG;amz-$dou_q5Yc*){@Z^8<
znp!@kU`nBs<lbpb(o(Yh`EP20)WWGnQ<8e7r567)Trnj%CAs0;lD*PWNA;;wX7N9}
zV}7}xcI5f?H)V(aEa`2`6=l0G>(*wTrNhv;O?fV@^^=uoxFyfIcYjoHoC+$QD0p?f
zF0V~eLB*9hUR-~}nz1VA_^-OJ|G59lp_bq1?WlDxt!>SXEq__ES_Ku~ZMFX4p1prj
z(V~0iOo>EhVDRYYx<4^vu?i|~Tvhz}d#+lc5*6wDQoXObbo};o%PGccy?}o4!u3U`
zzFYH=<touJUi@UWdSC4}^4WeJH>y~n-~4&ijxR5$T6avXZ&j=@_3g?dzFGa93MzKL
z@PdX5wk=RW#rvnfTDi)}Wy&qu?R!c0%0AClIq(=1jMU#X=(&2sIq6-L4kMLIxiu3i
zl$!YVb`^A7xob|Xx4z$|+~UJWKmDQWV=Gm>vFX)cK6$joQWYzt|0BhWQfuERykf2L
z3Z3e)qvUNx#tmDg{6eRC*yW3tU$)aLUg@7+_;brLWebI}3SM<dg*s0j$rlPbKK;p%
z-qU6l3Iz?<4&T3^W07-1UZY>*qr<l!mCE_#S`};bt37sd{hr6{>QD6ay~7GD*-&GR
ziWMe0A8nhw_0sAO=P0)rZM&YcI;^`P?|Ky}Mq9tGe6I~%FW)w0p>m5ld2Zv6jv8}&
zy$<VDR>LmeT|cD5hZ|I)VZ7Fhe$|`R?55kz+3)(O_OB`^=vTkSN^ZRE$F09~nqTtg
zw4BV>vFkY}OV?bdf`+YU6#3xYYgQ?*Q1aq5oH^=)3p;iE+)6Zbs(-xlN&C&}VLP6^
zYQJ4Qx^%DVmvVIV*~MRfS+NHDMlY}u9Sd8nR+gwc(8?;<?78iEZeO@u1qEOK;pc)I
zHf~Zu!<1WFU2|2&W)*boJ8kf}BktX-f{F{bRsEq{zs)MqQTkOAj_o@W!*|x^JJ#pv
zS*<?1^<1>+iSD1Xntr)jzc+gEbDLCF!P~wc+WExYzo|sUL_0@5Tl7h^cVE%^Nw?*_
zzF)Pgf5-r9$e^pHE4SDy{k4q;ny0CR#|ZD++}_W5o!{+!X`<)%njgRAB|4vHi+*nX
zl3)4l{HWjQe%rg>^1Ytd?BqG^kD%}SoW5^%%7@J7*Ld<<u55?bVW+$&T0XCN@LJxq
zAMX9~mRo<b>`>O*<d%K+FWR}z3JT6zu;<FjUA9_&qvXl&@S+dX#@4!ji_)QA^{2l0
ze79-YR;#RnUj6<e-g1(+F4;D8-9{A@+`VpM`gd(stDxfPvh8wp&A(9v1>?2;^2-l-
zbuSg`bwj-RA>Q@+Wd~VRZcy)4g7zN8PTwVu9eeHm3R_jsP`57hIzC?G@1^3smpA`p
zo-+8~d5E@O?8b#jkMH}t7m`+fb@%G^D(I;Dor1LA9S!cdVb#WyR-{;UcG3QmUOHn1
z9c$0*`R$(C*+N0X%WE}Q(s|P<%PmSfh*!AOC(U2~;-Z;WyfM=Dd$nIOa)S9DFjBk8
zD&_pX2k~2etnZ<n_Pg8fd)eh*@y_lIIp;4`LB)q^)Zh8Y&=txrblNX=`J(+Us-1kV
z)88sRY2G!Z7O9}&(q%nP?>V|e1s#KSPM;{>&AlVn{K5@pZ&6P1?$-zW@>;WPDyTU4
zt4h7|4A`cEio<R=cyq0P?NDCh@(K^N>~wgm@(Y9Z{Xy*8>v=*a?~!X)<xhI^56f%R
z)fa928_%Bj{maX?s7NuW_Xj$?dmX&0YPUYWsi0%fe)k%${r=YbHed6`qLnIEc%o(Q
zijUOVprXZS$35l!+CRifZoJ<I#mjD^tvAtrmlJP&7!O`ywCA^O^|*b{xgT^IzsPcn
z)T4VZZuc(RZMxn068ad7jDg|roC?kUqOpn;d*s{N@=&wM$}1dj-gE0(44SH(qOR4^
z&X<0bkJb-IT3*MKjQ^nho~!RVzi7y7l=Zj`k2*GP$^6CdDT9g4gWG<!%is8!CU0$g
zdx<jm7Y}ytTkO_F=M-J=^2Ax2l@9I7+okLJvq|oDZGXNaTS$lR)W}(`X}4@4iMsr;
zj$gEWMX3);CWnHG6E~K8_=)FFSwX`WU%h)$nfs1ee&OUNXD_&?Y^D`!e5KKl8X46x
zEU!>E?xKxf(Yoij!xwF^{Kn&*zrH-V=oZUwgpYE+-c_U8neVM=v3#|!XWg5wtXL!M
zFlZPs3^-@NuYH|ga-FyAubhc;zwDPDZuzjwAC$k-(Z0x``!hOzuVRgTMr_S{=GC86
zR>!_?d{V0NvCmbc80oo8(91Etg7&@o_9tswdq%ENu|jxCRD62Heb=2^XRdOJez)Vc
zPj)mr`092msJQflG7Hw-vEB+Q+I@c`(bI#<J9_sX_`|X~BMw<X#Xg5t_0CozDHK#Z
zweQg%xBZlE`Hg=2Q6t|lkd+yT^g9_@nUh3lTcqczk?J1n-vLCbXQGz7a?8%iBWldI
z{K5twKhy24=N4Oj<K&@_HyirlQp;;>)NXaliMRY>`GudX&epZd>Lpg9;?1eI?5vgl
zXDg_<;OYSjKfGyz6*P?WJSx%3Wjy)(-0q+Bwh|55Z#l&Y#YT>5m3N7iXxO0V>J8_l
zcTqvdu4{h({P`E>tDs`Ezh8}a`FC5bf4FDwUlgcYue2NR#hm*V?NMG~yuJgcd&g|z
z|N1-Sy8hS22c0*!@6#jQfBId{Z9B3`ew|&$D}MD|cg!p^Xqk#N#;d(<>zz<}V3k~7
z9<pMEPW8+2L5)G1KAvXzjedQ9!D&0A&F8ltGx9(Eq3dHSl^qX3e@75>{q6cE(bI#@
zzhC-0^^g1VzrVQ9?Gu&6Hy*0Dt7M~}RixPc!V4NM*tS3g1z&r&@ql;g{iwXgXwL(j
z_9OZFRcTmo&ZZn8J0yQlzYB)me*33yo41&D@h2-Szmfh(G<52poW8$Gw0!p*y0Af!
zD$7(*vDtIm^W460x$+8QZGX4-ICXciOE;@X(XGCr%B9}_cH34fT70P9Ys;@I^Q#pt
zy463?^6S1Y<W+pn9I4g5E3io!jE)Dp-@kg5?zep=>U(Xc-{<k``%60By*k(Sd!6N%
zd>t=-$6tPX+XD;c?zh~c-+CGO|DAz^|9jvzf4AK3dqdHt|994c6C11FflVITx^KHN
zGnK)=c<^dp5^p(obno-hl#=T#9o9>pJ?rHzYb=R%UOAFnck?PsqHg2_{qN5ts{T5@
z6E)uHxr<l%5|sz9^y)ip@VO)I-K@Mqr+Rs{5B6J5qK;dsSGSw9-}O=LUsYDaF5g{0
zq{D|BltIITu><>Y;MD)he36bXujbik<;eB~CCvGvPG7&yf23Sp{lo|Dd-6Jcy;|q!
z<T$-2fBv3J4%hkh2L&eD{a>{D`1KAhXmS@9DpO`j<s~ZC=+*glq;l{Z5__z<=*mH-
z3xpC4_rLY>%JTKjSka=}@2@+}@Ae#aY}&Gd$xnQxBz`;mX!bShW+_mf#}O~F^^zZE
zjn4a>0_E?~<oCT#6ui1#m)9mKiCt&j_sz$Hp0(nI(jHCNTLvP17w5K|w^}D9?Q7QQ
z)6AS}RZx+4Uy)*3r?gh7<x>i!6i!L*oz^5RCEK6>rWQyooLV#`sb^Yh@jt^AQ<77X
z8_q4+D=l?YpE_k0d)*60o?*Z+U>GnA7zPXjh5^HXVZbn87%&VN1`Gp+0mFb{z%XDK
zFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+
z0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn8
z7%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXj
zh5^HXVZbn87%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKe
zU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs
z1BL;^fMLKeU>GnA7zPXjS($-bU%uwDE{*3XgTdw@E8D1Pn_<8(U>GnA7zPXjh5^HX
zVZbn87%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^K=2tz>y*}N
z(B$FC|87ZY`IN#bMN*P`r!`4S$@b^JsRdFCrxr~~>Y0{W{LgU3l;o7;hI32yN=qHp
zr%sv0NlBxpH*9gB<W_6Ys6)rLq;6EdEUCMy+&SCT?CVDte>D37b$<KfO&+^$@;|U&
zYUd@NHCSmqc|+x<SKhQpwZ87H_kVh8b{Gr%(sbYP-xpb6yEaeXa?6GTVL8;>q{gL(
z+E4t$N*cauNU4kiR{s(A*X=rUi8}E`i&d}p|J};0UbxGuYqwiF+VrVgZ}vv3=EyDG
z>K|UA$}XAp-n~nIQ|LWl_uNXY#;&)@J$I<z?YHi*x^&K!ealz-!tL5MvhA#5Ro5!?
z{_xsiUybRo+aezQtEE+I-E?hhYx$jns;+wYcU8Z7pW69O{bqsa^I)S>b;j*Y3SGY8
ztB(dGWmxc)YtdIV2iN+}LjKQ(tG0in;x_Bo$M>aeExa_`4(danr=H5c@bt%DsfGnd
zPguBVvqBFLxziTa?%eA6A1uP#r;L88Mv+m9c2H0CT~nqksJc8{@7~fgiZAH7NrB`?
z@9Wzn<$t-?+OU2PFJ3sM%9I5b{Q%M~iO)Cg{?$$UwkU9Xp{ZpaX}U+zU$g^zqW?kK
z1Hb4)JpGRx+DZH!U#xoXo2EO%yudH=u`?ag4-&CU;_n^Ou+ND>pNIRK_JHUCq6gyu
zz9$WDv3&RChr@g$7dhAscIOvH>jFDG)qde`9g{<hyPNhjY+YsJCX0Te|E<zjZaA5;
zK0IErbH+RECrmnagkM4qc1C^V-&X44){m}Prry4I-1}D+Uacm4xvS9K^Y@3xIr?B{
z<eM_IMN(ScuPyut_RlyXo_@fthP=ElW%37y!u^O{(ti4#{Ma>!ebYYL#W+P@<iV3f
z>=Su}sfY24J;MtKALIw=4;{jX9ZG#=m~rA1@&9fSUYsI)!aw5|ehAY~Abes+(k_ro
zywnG~!5`qS$=BlIYg1FdIb$KeTeB{6pDT4JJbyATfj5j=p1GuKQi%G{pYkB>q#v+L
z`s=|FyUzQf^)ZVu?V$hh$J9eU^agX3x-|WldYKmE0>nR~ANdKx>(~LU=TAPlH9X(I
zPx(9ZAIe*CXV}j$kH6aF;<jUk?GMjKr{C^e>-G+N73uUN>xsJy-`K2b|9uL70^$c4
zPt1ex%(^35zk0Tmue|L3hTnAkLme+zdHdj!I~4tb-+y`Ol0pY=*&XgD((inGU$!pS
z?o#9<p8f<WPk)l1da-XQk9h0}#P0Ar_-p1L!V;bKC+h-<j4K^7{&a|a>rf|m%dVRH
z2DMpjNrXRmki1C#na?sNwqDriq2p@K?&_^7AInhm4}A5^cjaYQex0cvSlF=Gw;vx-
z>@VOS|B8PC3twMi!;l4Y!}C7#I=s`rrGBrwc22t!s@~<5cU6Dsq{{u%syq8^Kdx?>
zxZ%z2mmX2WDv$o8P3@!V#Gti>UV9*EwnX%!9U$_M4<euRhf;<;^;EV~YE9m8yB~S=
zpsL!pZSMLb4y%2YN<2Lx_n|N^?3X4jn7i)4(-{iBnTPR%?QYC}(@ld9S|y)2{z}1O
zhr|0e_+<W|oa}%2!%q`t{R-dA3ycf=HGU9%sfT*04_=99zM);%5$Smv_v+fQe9jPd
zi9WQe*T#jVDs?(;p$~o-dqQ98XC9$mkaDyKWE?TSvChPfX(#(H(pk5&u44TNPprRa
z7xNnBSVxl|{TP3&3&_tr3LnTP9zRmGd9I|J?>%Y}X1xHf6*6mYKQU~th2G3#vfP>t
zwYN08;YfJDzVGCeHiN%AVqu5OgWrF%<biTi4_l;DFZGcRp5c%7(QhF8W9E7IV;s<K
z+QB}V{)ERpFSRbX?TG{7a{}~Q@?4ksMPJ!tu@9hMsh@QL_QiSvKgarjd4Mqc6~-6#
z3(|h<K<Z07$wxcTx9B%@OW*rKx`jQke!_m}KX^h8Vfur04g6rY@XokoJRG~c^QH4{
z8DkNT-69{o>5tD#P5RgHZYRR~bH*e2D2LwgKz*dsevtJj_DVhwp0FzreHq8h5A+lM
zgMP$6$$1X@zz=1w-{<J6HxGr^rS$6yN4}a+wbd>QyGI`NAqT&NU&Kz(5Bp<Y1L2eQ
z(x3NSG;DhI0tdt6`huGVJ)d_;rbRq{V)&_=AGSJi++uv++NE>vmw!EK;a56*d;j45
zdyj?3GrWWF#C#1Ov<sv?gvm!d^9b`j{9v!_D=0_!^IyJe**|ag5PV3)j_eS<>=1pt
zM(oiJ@jG_N_=YFeMa*OHK|Fk8uW}qR-tF}BI-T_*;|65Cg1=(CfwUXH!uXPizmNzY
z2LC$uj1>>eV-m4P9bymIHORUK{jm$iIerhszfc~;Z?b-3eq|oRpW-*)eEY>5uZ%x0
zL_GO77Wm+)q!Kwoz24e%=LIK{L&(KWLE`b}(to1|_07NO;%O`Mo(%hU);Gw*zVTb|
zLOSaxJEXis#vLf-;1}ozc$4Fi`JHrKztS%{WIcv_>Sw=L?eF^@<{$QV*aiGzFWmc}
zKl>xX$R`XBl&3v9#I9*CDCaZg#~~A%HvMu>dicDHcF<n@B<ofDDEUD2mPoop_(4Aq
zy+GE@I%GY?IupL=f7WUE0oFI{7qDma#vYMF82-_Nc?>zMXIUSzuAzTvpB=(G>oK<|
z=X3Tu*fHl=tdr<}_W!IKSuf(p=r8Pn_G71nCCYv$j6T>CdcrUKk)L*9hv-Q>Jfko6
zLzs0h`qST}7rFlS4X<BOD15$6zoQrSLs;VUpIY?4yY!jxJt}&m2mMa_;0?bCzr^EL
zh^IV$3j4-hNT)pZOZ^~x!6SAcd7&Sv5Bo<S5Z)O_gyDzs=tcXnYvP$l@N1N(pO8a8
zvLC#)>KikkY@HN>2kHlD7ktt#+C%-M!y94hM{oGU{*Vun4nMSqd0Vy%J>ZWp?UebE
z3m@=EKEkvMc_4NIKjfo4`qOUgm~{A~y)vG3&OzXb_D~MFG97u?D`EIYZ;*WKi{XcM
zQjRd|DEb*Ww3qsrFR`n8Zcq9)-`xko`*Ow)@@W@Hdub;KKlmN$$^84B3jNBwMSGZ6
zNN2tyOnX=_(H<#x^MLdpr?k%(V!wl5BR~BHat?-F;dihb{5SJ3?Su!$5$#4#;*ksD
z4;e44r-_$H|DYFk&bb-wqW?K(fIs4C4@fzXe3ZvdSO?NR<kByaN5(lwKFT##x4e>X
zLQ*KBYSGN=N972koTr(uLHNVp(C^qcJaH~byU0g>Vb2n2AL~ox(I1qDKhEE&hw|vd
zxTXE%1Bs_U;hTOyZ;<qAf3L?GpAz8*d}VFd*Tyd`5<(915XksMKJoC)_@rI%fPC6R
zJ3#6KsYfDmIlq_4c(X(F(ILEoj8pss{VvZ*IXC8<2!04-57;~P(@*R#u_q8eMn2?V
z7vz(OUBMguf&CFi9_8sr;weu*!!PORMLOj{$s_x3>;XNnE6RcJgdTD{qBneyPJYrs
z;z8EmApGbMzX!rINc|vufy{Rx?F1PIAmwyOJM55tf^Qw7mk#N7kb03ve?^MSqu2*X
zJ=hcWLcQ=qI{iUE5)aSV3-PoEp0F3jKkZ>WGmaVegc+}l2mAo%AIPJA(m~n_q7V6*
zXQ+>JE$Zbx6Zu#Vl1@D99n$e@#N&@Z?z2G3;m6^H^0bFA`ol9wzk;+El>I|F^fma`
zyK}>X;o;BU9~e7`%nq0rxK99?UqJRBP7!|vGB1Fv6C~nSCF0+8h#%D<?X*MYKZ(pY
zcF6oB%ggr2Fudr{&c5@Ee7b(OQKu}4?2DPlb;v&3DN21<Z#YHP10Z~H55>72_gfpA
zzg4rq+b6AyXKh(Jv;67syJhyt+><fibI&sTv&L(icFPuG{zne;IQ7tO^dX&d2zX<@
z2jLgJneU0u{QF#o^#JxqJ=hcJl2;jKKg&9m@yR(Z>qE}fs85H;ljBf^S!dEdiRh!l
zg}*O3|E-VCgx}$E&S;0|A?uT2)?FZe2c#cC`bVOaOPGGrq1?9<hBt}2{vaNEb&HH6
z9r8|8BIAbf%RT}-MIQRWEB=b}O5|`}NjdJvcqhVn7wPy-_$D1YX5Y!V9`8>Gvu{9e
z<|*`LJ&nC%cf>Pp*oR@q)X#Y?;~u*to$|~-GK_xQ7t&th(UbDjPdtddvQHqMenU^>
zq6hrJH{*o*Xb*Nu|KeXrmwLfB^^lGn^r0T&d2dCSeCSO){e)fMPia5%DE+|uF{v--
zq_jsO>Bym7<fDD06EDN`1NxH>UhvcCkNr^}zb}9f<Wr7x;<0D!fO7E2IG{X;e)J3Z
z&=)!Ih<>s>{eoT5F6@VR{0QwOO#dN=d60696A-)2=#un(y%MLwzpG__W_%zI|3p31
zi+<FHeA+F`aqdbv?16fz55KQN%IgsQ{X*<BQq<{9Jp4wA^pg(h7ad~HPLXi{V(-`i
z;}-kpcS4Li{1eDMA@@Rz7m#@uxzxkE48}G2`TYjxwAdZvi20Op$2`mUp&w*^=2zxX
zkaI59!;%;DhbP+2?^W=l$j8pgubNrrP-^mQF!S$qG3{hs#qU~3*WvWW1-`ag<_&@H
z0`fk<4jHHT0s4)31UrOJi8?zZ9>1VNc+?^OPKTVYN~GT-MSf3*-ss1>Akt+R`+u#^
zmY=#jlCJ1q#tZX|j3=FV$|GNn3+loCK*kaC82yVK(a+MqvaXZsZTg$>Q|<5jMR<W<
z^uu20f671mY}>|N&e>wsK3S>C!rmF-?^3W{Mh^1{<>mbwa>&oTh2Gc!=hxD1*iVuV
zgeN;>e9%trx6y}n2<vj}hH;6%f^YgmBKv9C4#FV)$bOx8<|7dOkPE^G_96Kqp7km`
z(|*#CPx)P=nq58O!5ksN>^nK1;e3L7L4F^BJc;b5Bw|0jYixK`h4#w|pBqA-t9L(B
zp~cZcArSpRerE+%``f-j_$2@PSB|WfYi*toIDdb&UGoNH4}sVd{lUJUeHG)9`I&cq
z*ggD!v;%qA5%$ty&#Vzgs%0wP^)PQS?=U{3JjOHg1LK)^><XlQ#xr_=oGWpjgneUQ
z@JG9u{|TcfJW>w(r5_kq_yK+&g<gcQH+aIH$OqDH>O)TuIUw;M{RYAxi2fkFfYc8%
z59koSS&y+kfk*6%e3_G5HQswTUx@m#1M;)}VjM}tj^KxV74k`kAI3X+T>Wd)`wve#
zZNUfamdN}CuZ%zJk8@6m=tF&s7x-bl#=epEN@U)R7Fi#I^dtSmx=<qeGcG{p8;SIv
zTZC5~%6=hCe)y(d<}>soA4osY|BO2j`yrjM-8?MQGyh&sgZN?Wk@~O)kl)GEPV5}y
zcU&Oxj1$V~kZ}#77l_{g@q_qr!jxxSgkS4)U}m{@7UmAmv$R)^N5(&XnsLRx1HW+f
zfYkmwuE`Z5pOnLVg`Y<+9ip!t(te59ndFys6^Nc7_6H&dWIY4I8)&DmT{?2?5IfKz
z?U9IG=up=mGM;gS-vn6~fW%A0j_gp@%X|)Ehmj(3B(g5FLmiL0cxjI^On)+;f$*fW
z3x4m<`WpMiKQfPi?7Q&G<bwzN9(Ir2k{>_9Jt6xS=6B{v@{`U!7UcUJ_<PFZckt_s
zSLQ$b40^HOWq*L))KB~97s|6fg<sAu$OmFy=!2ew(HH*t?gl*2e(FIkh`zKFyJtR^
z^~wGvztn?q2T$yq=s(iY1G(ty6q(m_h`rGs9a2C224XL?liv%%FY@O-(S7EIM#t2K
zzK7}*TYXIN`=UZ$)PMKu%}KMH6dPIf@y^+1v%ebk@xGiF4Lz-Ll^?jW)X1dSS3TBg
z+kj7!!$`gxkNj)=6$g@MUwB*g=J#HfJe%*&fSuar&m7<9lp+jb*Q^`Rm-PetJM`u}
z5WNV~Uh;vo6FsSy@8?hs<a;};8{viW@Ql9bK|0^fp`NSHnbGp(!4u){=x~n7{*rYp
z{X==`VZ2FXelchmFbsId0P73pDdr2#KO}lrvyl`823Vi4F5sM+^K{lbtVdWk5N3VJ
z{KEWAm~{?vSYI%&vrfY<S)Z{kpgc%E>=}K~A7q`)x|VZK!mN*wLwj_H-sGn|?EsNa
zKB+hBFv{yNQEUf3`8^!{LO-*9mGSf|@gVW=1;Q8g(XX7NkWQHTk;}a-^5{qMvHqlg
zup`nX(k|qnH}!$k$9e`msSmk?LDq?cIoD*}1ai*EcmmOn^(FC?LtpBr-Q=e{?V~&>
z$2IZj3!)!k@*x+bUe<xE8)-M?$VV8Ye+g3`NPB26NdFLK9K#R#(O&!%@lFwYf@j{Z
zF>a`z-~Vzyf!@4Zz%I~-c-oJhp+D!U#7ljV1Hu<be)@%bVR)h(dJ>PG=s`Z@!Xx?U
zSH=x~k^GE%>Y+UK!5{q(-_%cj+C#q+FT?1IT;w1hWITcRAM94{cVw9MA&>g76V6F9
z|2~(N?Ij+6$hidPLd+lVC;6rx^hG{l^aDw!9rPFagY*;qK>xxk^5|#wq3DO-KtB6o
z+JzkQ^SwXz-N<FVfIRYJ&&UVKhaBF~&@On9$b2mOpZOg70_hKsc>$E?_MFq;H|THL
z$M^p^f9GDF^DovrtmpXsG3B|Jz#q}iAmfH{0KfDD{Y$^XGxt}rJbWXcb`uZZlw(}b
zZuEdZ&P6$&#ZKU#Fv#BvV0<xd*bl+K{2n9u`Hnm9wehd$DZ|7&McHo3!52vT8RzK3
zJsR<hNBV_%f$vh#UWw>IdFC7X6+P&8`h|GrG5i4Wa=s#r9|f^5koI!F!+2v{;m^s3
z-7_8tV_%$;GcG{REwLBu!zog~MBcxmC-#mX2GI|MUl9L8zrYXug<R~0a_}$1<fA@0
zzKN$FsTX<FOFVW)Ipk9>av1m6DdQX7h|m0c{$rdI&-e!655!KK_8azueCkCWh~LES
zXczUJQj5M`wIxUR_foVIdC12euz&g$J!lv2RIpF<z)taJgylV-4C5!IzvFj-j2G_h
zxCdf>VI0vO`i1$1_8}kppd9(RFS0}IjClq>#CjQfr5^kc$UPB$55z9{T?hH`-^_=|
z1+j0=-SJl-`xo9{66PG8FzXuDBa|ne@l@^a_hgJ4`XB#II`7&TckqM0)X%$5;;D!A
z59@s1yAaQJFZsR>=Yz}}^gBEfrvHe?P8sLu%Xo5%*q20j(jh#9^auTbKLvR|&bpUz
zN`8=j<Gw>8JTi{4C)$r(^ujOTU&u#)aj!@{q$3ymhIf$jF8F60gZLkbtaByujzc2-
zd`bE(En4KrAHqM}_3rRzs()WN1mfpF-lKrD2ZSGx@lStY_wbB;pbz7Rcl!7@5dVZf
z#~)t2Ao=~vp8mh@8{*d`@;llVOX?55|K{AGJQD_#*t@%6DF4#st1s(vP8jjGI{bQR
zvmzN^6bR{X*U`ro6)jybB$57MeB&?S2R}+Wc86VIN4!s$c8Fg<4)=J-V|_^f;7>s6
zV>}Vhy*+ZMpZuiL4%U~1>38%aOuHBl$Y)$3AH+}6KE^e?P!9cRrwrqlsUN#zTw%AA
z$6w(mK-$N61+iD`9K_Bq?vR;t(xjXr>XUwz@s6J+A3RYn?LdEc0I3gT+<=U09r7I<
zkog+KZ-Mv$9Wvj5tSdoTj&;67)&=lEIqU@g%X%Jr!9K_jk`JUmkPpAoKCuJ%h6nWE
zoC13%o^eQd@-dDWC)gkLvOXXm@@OBtVi)Mcx{CEG_JsYTAM0N9V_nEPm3G1#cEi0B
z<DUB{<gv~to_ZNq=!tyx$;7iR#%|?$3O~<2n{h*bGA>y^q9^vic!EFlfPeDQZv2A|
zsSlnR&oZ9&VQ*4z{08+B#vVcBg2+KGNPXCcEYCPZUs(=5(Fb|>JHqftJ@CnRLOymt
zzfvD|L3zs2F7&26$oxt>=pWAK+#>R*pK-$YWL#1|c0xYpNq9kje#c0=(U<nKe;}Um
zgCCUVj>yH&lb`cM)>-TmS?}??7|vUe&$^2JA9hDQtc&PJ+Cje%W}gZopM5%V(3g1D
z!RSFe^F8Z8cn9Gb`@>!dgYXYxZ}<U;*nvd+4)wZ4^mB{YHRF(W;MW*$_%FiPEAs~R
zvQJ_@!VV=e4rn)sJP`dPqK6&Q9}@AWI+XP}4PzJ4qSRN0;VV*<{UO8jBkKX|7=Oh2
zfqrJ(FfZb7S>F+lT@z;hg<s@gkJ1j&gLMn*5$uWjsE7F<`HVZll%v1Ui}LtS+Jl`_
z9{+?NWnBcHtZQ0Sy8GH?xr&6pzmET7z9T>HEYSxZ;F<c#Pd|}PIs7N<F2)o3V)wKY
z`=XuTPaR&Z{ld<?A&`1?m??f=qYLZCE&XMOoQLqP7XQRK59fC5^GFA=Bi8rG!`|_4
zgy9c)Ap1e)6Y2-aN0@x{7l<4kvL28KkLU-=aYTQDvR|+h5c>pKKY+~RP7%2v>kE*1
z8N@C*pG9x%f^h&cPO*E&0dg4Uj9ZZL12PWamGK86m;A_wC-`8Tk&d1crT@mCfz*dR
z5)UuP$1XU>L4WDb;1NVVc1Sz1BkYlMc%+@|rx+)UE82;C^g&<Z@k6Ae7ySt0Z{V5s
zbM8fdg49p@S=V4cAo1i!KjeV43w{`1=nenmBTPP!_L2@Vu8@yk!e7BJ{R<D+1?Axt
zJ!lW{*c*HjhG%${`#;h_%7ct+#xM0TzFDthH{>HsI`T;8-2?j$;z8E?+}ooEJQAiK
z;fr+ogZ_s%>=6Dy`~&kn{lNJIywMK)0P6zyLq7Wt`j7ndiwsj9eOZsuF6I~RrPw!8
z4j!;8>=(a6y|kBf%E24;(@y#UK9NKJ5(bHf2jY>3eC&e#8+JuH{s?^;Pw0vMAm!*M
zS)Ovpqa5Q`hOr0u#on+B!n7~*@AVPqSIk@35B9)082!(A6zgTew1e{aGwQ*v&=0;r
z*74X6{meQVzGx@>Ql5Btq#k&LH{^iSBjvz5h&=LRAMnZk0K1U(M?R2t(hmG7{fa!!
zh0!0sKzZ3d@=2r~$}wNT7xf~Kc-qT6%sCNZ=4IlUAE*bpGK_r2JN3XH{SIPJ$OFmm
z6tNQ>(jIumufYq5zO+l$k3C^0Amb544#+rjipaG?>I2aaBwnJlBN>J##*-al|2ow5
z3x3)Tndjx-GiIJ<{$rkId^4{jpZxeW<kEinn{mav1j@7DCX7D?NtejI1p3H+g*WtK
zULs8YQ7`#3|MvIjPdxeYZ}5*i>_g^fK9T*69TU$u2boXskH}#i!+L;zp?;8cA?eh|
zJjQyOc=9v9(Qly6|G*3TEA}1uH~4^8+5?ZoGw(4@kV|>^MlNCe407o=<b#|eksrT^
zUxEkLU&PaX^3xB5sgHJ34x~QLztEp_<_Fp>5&6iWybjS<hsei%;2%AxkMT;Fa>S#*
zv@h%#ePw?7jq;%66S=fQ=9Br+6J#7wP9o!kc9Ea*$fthnf_TcICw2xRhje%$9{%A2
zp2!D()K5O-ksqZ0n3pM!zSN7|Fkhnw?PnZNAMKIjll2yU7d?@Sz2jHVgM8FSJo?cu
zAnlV#e-e)z;%Nu#L)LNhBmBS{NWXFqF69!Bze7*z#ShRA#G?=GB#a&)^}`QHe)@xc
zq#Q`U(N6Nobu;ay|LHf$7wt!H>Y-imFA@3Vr#$kA$DZMl_QC`4w4d@IdeCm_VZFfo
zOg+dYA9|w?h&&K^Aa>6^EAfmIzIVrb!FopO!~Dj4N4jixrr3w_eH6;U2RzX(#vAR&
ze$b0^0Q4rFbszV)+{?1wBFuN-Xg}r1$L}>rr#{LvZ?g^~4DaX<FRc4O^kv;Ck@XIC
z1>(Qp4a6?61L`N8^BUgAabG|=_F?Sv7zZHoIOn9Dl*c|n>c!6KSH>+!I{C35^r79@
z0Vvbak9#icj^87(-j~R{<rdLHhqOb7@Fx*F2eAY8$1=Z8K6VT;UUg`<A7vakMai?Y
zZ>M429fFJ>#v^hVe~d@B$UcF42F5@8KKu;*fxUs)5B3UT*NjWX3-S0p;z800gUAJ`
zU&^7M(F^^tSMpOHJ0c$W*c<W)Gmnsu@-iR!*`J{Y^?<B%@CVHQ^b3go_!G**7y3{y
z_KzQC-hmg8{K#iKU~hz(5Aol$oAFBf;YY^Hdgxc$VTb7J7WrMQ4t4di9@HWI<u@`9
zF>kP5C7yL4>qU@phknFkPxJ?N!~BImffwXpXYfxs!pOtkun)>J-Xvmo(w|X3_3*uZ
z<iP{u0Xt=!!58yCdg7n)lhlLWl&3w|DLi85@Wl7eSa%_p@j^c_Zr~C9;DdHDPN|P{
z>LnlglFt1N>ond&5avFEcZGAmF8y@wSr>-Lhn@2~aDJx&KiCcPVCLWJTll2?+$Zuo
z7XEG;-}ij!O!+sve3=}olr(w4?ddr~%<u4w9by;UyU@SPQ}h$RKPmc6-O~5IkZ$3(
z_`59B%lk-vZ^e3_b-Bzh^(Q~SN8x;ubr-+uB8)!R1NFc^zkde#dxwv0nx5;HE0aU;
z339&6Jk0t|j(>QAFL>g7621A(JN-dD*cEc<XZYdo-of+yn=YQVGVe)?cB3clK~H{P
zNjmKy%(*sl(1US^9=r>NSL}g(4RU!WOPFzkJ+p7de$kKfTHf38dkW?=`U`u-&ruKK
zl<`LU(3kxd>Bz;78DIEA>_xu6Wgo!4i2lakv7h1mgLMUdka31TB`nV&829*F^5HM=
zXCUeD1CoxvCLdw?hkTsZP!E2TFzM(8axYJQ{46}M59IfV$l-k&>x9qi7rtifk4F{t
z!5{vT`apgc0-xjush57CpQs;w*`JaRen9Mnbmj}}kb44nm&ka*4<HY_WBjqtU>snN
z_(l8$<v3Si-AQ@I4|4H4)X)5c{ZdZqg<R?frQao<`ImZ`9~n=`(;@uQFW3?5DD0H}
zW*%ld;y;OJo}-^Y?m41G#t-MWAby|v7|)D9>;pfGz98)*AM)^%*fI9ax`puuPmBYQ
z`r(T({4%exKO&5L^g%xQg6Pfo1Q~CPYvkcy`MoUTobSa_5Bf)ow1@W7PUa=%Dfpnh
z@CxtnhTkS1ejdMp{SwB{fY=A)n)bmLd{7U5hjb8L(GSG0(hq$9miNWH$Kzcw-@!#L
z<p?ugm~WY9@h6Ob(h0*e>v7rvkF1wie=v^Whkm1furvCRFn*8vIR`)w_A`7Z2|k*A
zbkD);GqcT}TK>X9dsgR|O_+QV;g#=x^ZjA^pLG@e$~uK~+QGOX9zL)q+J_&*U+`Xv
zb~3-<ckvgjvsee=??L2(ltT~th5o>fv1{f%>S5i9-)CN7{m*>HIT!tfUqcV<gLLFG
ze~}M+VZ5R*ewlWoH|>CTc;uWJdtsfZvpf7b>GTuws0TigiyhK$65)w?m~lgYNQ94e
zH|D?Trojip`!M_n{lU7GeI2|ZAAaz+tgEDc*dODPenDUOr~l~h%)j@o;M=8#7X7kT
zwh(`BgY^L8it&s-$Y=e?cwrvl{EPDNPdwv|am6^n?_tN(i=6R;TQzK$mSZ-EAB0cr
z3_GAa_JIEM8}@)+q%)rI$JjG`BcJ(zeCUb)r+o)<mc3(Jv834#r3`!OscfedJfR2c
zEcC@5Sa&f#kcXXM7t&s62mT$s8CTdf{f2#_H}fm=4f@gFyypa2_mB_%xKF3OglRYJ
zq2IA9`V0GIeU1MBv3t&;7|-O#ui^g~AC!l0<^}fi%&Uot%wJzjD0$ac<%)(_w_(rt
zZ^i@jQJ%)Vx^^s|GyM0rurtcx5AYML1G#@;oZ`1Y>>oc&KF*1l9~h_DFM6Q|c7&Z`
zC)g+Bkbb0{%n$f2;<111i1Q8X27gL9;%Ps~_yX~#@QWR@-@rexPJ|cQLpt(k7vq+6
z#sTF(#u4q(*(v2|Cw>=yz<9&|;ivHn@IiU<Q=W2!$uH#)2JvTvS^pppBp*nA^dlaB
zgTG>4z<**__$Ta%{P;cOkj_2?KS6(^2lB~J|In}WC+XCKe(XD#H}HeGe_C~CpY6vL
z$nP~l{(c7g3gk)s@w?PRKKLL%cFMS<y|jmP<Piq3SI(ViKmCK<VUO?*FX#cnJLgg8
z0YC7CeEbjXqa6EI>VdaD{knABQt-5*KK>31?IWIXNxw19;fHo&2aE&wCO_%qr#+m5
zGLF#`e!N1)6M7(*^$UE#D+mvygXm9tke~5->9?n}Dj1?2$VVQ0P>%jWALb{@kxu`j
z56C$<$hw~XK`-WekaYrxpN|yrkH`nv59pA72br%x)-i4oJG4XUwL|<q?Zp4$2eAkG
z9p0HI>2Lalaf>|el~@PJ`dHt=E69EpemVC<F7+dqb|R1d$8Km3?L$v^q+ievd!}8q
zll2Soh^Kz$5#*yE^-xYCew2EdU&u!~<B59kU)0a|A)WTpZ}10C%<Je&z34$a`Iyh)
z3HyKt@)J)O{ooV%=u5lU_mGah=*9Yo_Q5OZ)QdjI#sA|cX%BV;PxyDr;ir&?|6+cI
zf6_tPg<jMPf8?WoX%BpX#MA%8qc`%=gLa}f>muqSAMGO^IoK=pqBp#<-bD}k3I1pg
zaw*4nhIh)7PbUYzLppXu`=lKBg@4Mk{-Hk7?GWBU<bm);JCRSA`3-x)p0Oj^0dk)Y
za<9+*I`0iwU)Uk~=`i!}^CINXPwcBnCyX8-{m#A+L=N_ZKc@eYi@&E|C9=<VitHb0
zk3{AZ9lD*5$&X(HnNL9c6i7Mj7kxOd0P)ixyn*<2<Y8Zg8ISM|zmz9GdeWb)m*_|O
zk9hn)a?ykG^gDXf&(w$i!p|^XC=V~>BTRn$0saZzS!YVUXdk?w7d(KhgBdUA!+wo@
z2J2qtEBpxiFXj>UO~m6L@E_#k{Ed5g{4;(5In;w3{1Sc{9`SdCnKzjqs2^mWpd9@P
zzwpla6#c=u5B)(r{ZBujCwxk~<a`DH&G{4cvcF{gMLoztF8VOeI4{C(B~QpjKk6eN
zh+Oh>j;-T^{WH7}Mt}MN-brVF%>I;m=?BWupWF*_uFHO%c2GZZv0K^!fA9tB&Yj?$
zeAo%&1$(1CAUvWs`e4WCqw5Fig+J<(dY}(>hh1<^iG7g|en8o7>V;3<hw%=OcUa6b
z%#)nMb1#H{!T&I?Ghg9f@wdd|FG1{>cF<0EpuO}1NI$?g{*w05eh|5o$4;<6`U|=A
zC;ft5Qy#>w$&VhS!!z{|4>BJ}|AM~g2_M8$4t_xRpkDYOo^>1TLN5N9d4zUwPfZv;
zbx42ekaYowe)vb)!TJch)*=21ghvoN19ftujfY3p$M_}oC-@1F@k@K*fqucR@E^=$
z<j0=yAM^+D*faMQ^au9`j5p*HMn3Z>@~}5}q8#=q<#IoZy>efLKgVC_Fs)NstJLx-
zMN-aBN$#E2BrPS|pZ}&7NG+UNG$pBLT59n>!xdAKQ<59bE!itAbyS}^Wfn_sXTpX7
z!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+0mFb{z%XDK
zFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+
z0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn8
z7%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXj
zh5^HXVZbn87%&W+O$=n5$`~@>{oPj3@NDWM(^kWPVZbn87%&VN1`Gp+0mFb{z%XDK
zFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn87%&VN1`Gp+
z0mFb{z%XDKFbo(53<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjh5^HXVZbn8
z7%&VN1`Gp+0mFb{z%XDKFbo(53<HJ%!@$|bKw77?R)Z!FPyTniQ_H8EpHeg>xp!KV
zw3KXr{+n7LwQy?Dl%$?%sm1>cS4>GxNp3i|WUsW;QGM!^SsXoG*Im`L(3|bg2_-5%
zUM;2XS6vTSZn4ejhdbO^?6?J9+xo`K53foN)%dCI$nVDFN;pJ6+PVE-MG7qJ{)37X
zkq07o(7Q_qbRE9lHJ^)5_Z|JgReLQRE*knul{Kw*S|Ih<A^8&ppR4la@?z;56?oGl
z8I3L-uuCN>M(+WP1jE4D#{mA&U}W>FTi)b65EP3`RKK4ndIhB)jh?||;ORSleE#9b
z_9}yh2PbwA>3+$Si*CQ^6@PVZ-aK*b1?!a8=$H5Ns@+mfx$O1&yt{6;<u`iGvs-zx
z-7;*q&XoE64!doiF8_*d>%PBm&mm>-KY8$LoEo`?0mDGzGa%0if*#JQ{ykQfwEDj0
z+v>cuM+F^Y#aFEBiTC)7b$jjV<G#({U-zws2gAdk_wL3Hj2*bK1GoEl-g&r1JAZD{
z|2G3!ZGHQ{TdPkE`ua3c_180=wNswz_W8gc7jIDp4G+c+VmA)_zHc$|V{@+s&slqx
zFs(BT7zTpOfZhGJNsrC@RKv63Ig;NQ8M{dgy9x3f@Tcy&-{0BUm)mnvr}@o$Tl3!5
z$$ML4C&o?^$xebk51eiK=Y9U73Oi?PQwIOyLHAxbtHz&g_U^R3tOw<Nw%vOxr|Ubv
z^6|TQ@VxuZ={Ee9Rm$N1=E135n0%2jfIS-g&+l6pJ`5lK|L>h7nmt4|+D)mf#(>>9
znAhpi@|V@##hLc}=kF8^AKvg0?Km^#{b9h^w>S1}c#IE^@|=Y8lte+l&R>k&|NM@L
zKi<(uH4GRA3<HJ%!+>GHFkl!k3>XFs1BL;^fMLKeU>GnA7zPXjhJko9kZ9-OvEoDb
zy)F4pB3Asws@<j@!+>GHFkl!k3>XIV48&^QH}(AIdj!MB+0BR7@4@sV+@u%=3<HJ%
z!$2$;knc_1h7;xQd3vQsEI2oH7zW%j;C5U_nm^ILXW^Eopyl^!UG4Y&jPKIM3h76`
z+hydR-TPY8|EB*JcK<<TRXgx%eu=l7pnR_{-aMK5<MX>s$vf+pXi>_ISJ>;{pVs)P
z?#S=P<O)TLy7wYFeWUFUSw3F<k!bZwUhKk&roZ3%*zK!>`kll?|2t1kc@4_?0zK{Q
z$)qR7_gQuGkS;!}>_eQ6yGZjH`%Vn|rXM1;Th<AF&%cqLhuf9IAA5y<@!*wyUd!>T
zy+&?g?1OdvfZuS6Iyte9_uBpwMV`}sF!}!Xyc@6njCH?6I^Uaef8Dou^<E`XIkCFO
zwW~+BzA|<a-MyTSSK?#E@1mivR9Vw{r=`PK>5-`Qu>Ub=7%&VN1`Gp+0VfP_Z_52A
z^Q-*3OS&-Y039-4M!LSNwcym80}uY8-XD@<*sn#>m6JA^Jca?ofMLKeU>GnA1e*cY
z4+f(<$8pPFP_0I{4z-JS%a6$)TYI$QTi3txdtdPi2X#Cdy$u6~fwPAJ-96jLn!8?p
zs(ap$#3vqoa%-tAr!9lQ;DPhKpkTb_Yg2DZ-><qJuo8ubv&RFQmKp{Uj{$yXkSPAZ
zD?N1kMz`^H`%}NtqkXp;uXlz)tyhjrX(xX5bG+p;{tY%Q^k(~WLUuej^-sugnqQYs
z+NCZWt-nar^1Aj0EgpN2e#L~B4(K|3z2%mNv`%TQQp=|lO(~X=+&isFT1vJ*|4l8B
zS~#_6N>b0X)Z%}JE2bo;BsZK}vR7K_s6KVdEdHm}Y=1o8`j!WGD?9vWNpEASx2oD_
zZ=TIchaZmrX8V+%r>I23dG*?k8q#u?3M!UJPpvh1;9eC}TtB7q{8CTuRzb&cm41F?
z@uD5dZ}fK5x|fzC*LR(dja+R76<^PHMa7&ox2tH;J#(f+EExE_`S)vnYq#8r7LPXm
zy2nqazfexGPMs>BS3WYu(&1gNEuK4l*cN4n^PV0!__j^oS#Gg@=8o;%S}j*#_5KxR
zoxFRQa*B52`RE60C-+_WopOqmE?e2S`z!BSAbr2^i)RM!X`il~;_;;mvh6&xNBNED
z7ca~)Wpv67%QvV*$9VA*uX^{_-BNDay!A@r&ViNoKls2*3*3Kep&d<*es6&#I=njS
zm5<k0;D}M%dJIcjt-#CI)SWovm!%4n{odl?kIv2e)Cwh$^)qsuqTl>Uej*J=J1<9C
zenh9ejr!Hyp@ND-d%XJb=H`bLxa3s*sq1#^QbEBVa?hx@_nAY=Emq%Jyke;Xa}~(^
zCy{ZZLz!QOM~t36=VU9T+~Ta<-%hCA{xwU7WnZ{2X~@>^RifhhjhCFf;_G#)<4>)t
zzxC2eTa^CFE`0f<M$b;@_M5Up=^tfy-h!sJeysbm0weWzIeJehR=L;jN{5lkrJU3|
zsNp&XHqH6@=qBYB?d%~|>9d~b*KxqJi>+88{SUew?_QrPyk7mX=0#s$rNEA5?-})3
z(H|8!;_!-nGpj6AA0~C}e@*pCVJurNb?|eyFID9}s@LRTi9b|{iyrPcd(%d>WZVO{
zf8Ah*HE#L(W0S7lX329vo&SK3{Uet3U(s^00$=ZQOP>`@ms%1Bwz+6og$i3OiLEc1
zas0*4a)qMByYj#AY4aoJhIA;`2k4(@_{}^0`rWl|r^+gby#*a3T^Gl@T*vpHOnRlt
zHsv+mmM62(nRZ7k9k#vsxsKy2@3ZW1#KkAKKQU>m1?ISF#wU6A?y~SpcIfoHh5eom
z<K+h$yf}aKnXC3IzcAMA^4iV@s{A+B`)<366~>w`xAoQkdEew`if&PEQP$ITPX861
zZ(eA{3T6FXg(F>u_qeV4{I>@lv%oKZd8hK`t4>-@v19!%gC9!UZ0(<StVj1vJ8kjV
zqj&fIeeym_hiBT}dQ*o+$1EKd?Ywwak2*O*I^169f~Q0G@3ibtw~v<b_*LDxxQy?5
zrrE^N7i_U~Xt!RM>rP#`(c7)c<e!qRoT6O!$#9!NBf4F9-eyaOIz7CKce>8|{MU~=
z7rJYwlGw3y&xh}*yTsDr;d7Ub?s)GR%MM?vR{GqH8QDV7;-<sb_i0%r-I93K$Ql)2
zXnEMu;hY{FYhC))NlS-Kr{}-6cxvvD9bVYwP>%&a=ML$x{|Dn+z1jDm1u~B9u*TQL
z3*O)FcT0z>uNzP|<D%q{4)d3Mu-xnQ^M&m2=`Pl>ciUwP+2QH)*Uqi<U553-4ZCm1
z)72jB{DRZ;-*$9<xJ!wGA#g?LfoDsNKVUh<od*Y%eesH%Av>%Xn)}AfsVA%=t8%nH
z^5AhR()ogO{h(o+-}~>aIex413M1t&()rcza*?*jslTl^VfKXnO_M@8)Xn3>%kvMy
zb@rXtvu))cEvJ|?xm%;`OFvL<G1C16=Qvu_+^gC3hdhspG;CLoE<Jtxv(>+ynlogF
zYu-AkuKPH5NFwKD5_eB5Ib!UzbjvNu^;4wb<1eipI<NIE%P+j=(^<s|FI{ixaR0F6
z#vKQ)vYaC4mf(nk**|}v<uXf$b#6JcuixBtmJZ#%7fBnO|I!}Kezqj)<_EbhmSN6Y
z?Qql84_|ue>ys>r^X7L;di;eImJaRC={fh+;rgrURQ#joN=t`wo#r;I^JARn_=N{s
z9Gkr3hKzrlC(H9?!gCfcDc!DMQb>oJ|9I-b>SL2b5*J+l$U~VG^M!PH-=p)o7kxBG
zNQaBMPrmxI!AT)IeBi?RXKMX^z_P=I{a$={UD*Sc9lm+u{PinJ9JdaA(Yn$V3y#|2
zjZKT+_|1;Pmc(+6@7uMk>mEyo?R(~}@Iz+0rNh^!-8gI8`&%qKe6`z6k6pZHqvaQd
ziWeAB^{ve+Ug-AzS?1@Ru<kyFc<yoR(C*zi>GLvPtloLvb}Lf6qF0$=x9$B^fp+)B
ztds0e`e7NCb{uWk&JP!UW$50|79F&lqTPJqb$b1Vs~e3y`Gc~<62qQf@z&AZmJVa}
z-d>&~JFVC4I!t$NBI9-IO0VKq-*Dv}-7k4n8H^1N_#?kiw=VK4-s%2O?q6aRp4ah_
z72_tYRaqUcuUX;EPH!m74xRQ_wE0S2Fra#iLi4OdL7l&}Tc_&I=Vg4}vX|XCb>%S&
zY+Ye+zaP(JS`y_xOorDtyP^51Wjid1I{k^)o|}kwGBTf6_X&Q>nK1p{i92&=3k40m
zI!8?uIe!1H!fW~WR2?_H&R6?YP_ahAJTJa***@hK<$Ybe!->K_{xV*8-i>`KKbe28
z3M$S{ey7{GT?bUGkbcs=uO>d|zXz`K!_NNSe7sNDq0{%y^8KIO*UE6$yBnTAXWBNU
zL#OLOyWcC?ooC2&`rj$a^9#Fu&lj`KEBC?Lp-P9k`vKzZ?mz6_5xPxhJuUgrz4vzN
z_nrD@r@zCLc2Kjz*rPo^JfI}zS@_kOc6ASi&(q!VL4LdQCD!FSZ1~xZr<QHZ9&(F~
zzg_i6or6a#r)YN`txK2v?RVJjcffwtZ}%QXrr&k>uMZ4L{>{=M??rUDc-=*x9KJMn
z$SGFcIrWo9Yx9I8%6%ST_C>n&yPbd1rME2n>4^7R6bb3De4kLErkl?TfsdE|x%r4r
zc|wumwt^#DUC?@~k|^&BWZ3P!p3J|zXL_rL9y~v!!_8w0zxYS})0V_%7d3CP{*jHA
z4t4j+GX9bkPnAFKhJqm-s^_}bpEf3M2<)}BWZ%A1i-hd(=)^A`ZTEc1kPdS{F?wa6
zJI)Q|DpI}q&9iga;(0rU&TjC`ZYxs!<Jiodv%bs|lK8;mYj@PBlq;k|?ss)|>h$*)
zk)DUzy<gya0y=D4qHC?88#h=I`?f0j+0)<dw?Mmhq`LHG3obeEQ|{9iIBb07@Aeis
zCuE0q?+tb76?Wfy{<Z54SrVCN?NIvHNW=V|!fQ<1d$RurM}D;=`ZXWv<hnhl;2ss!
z$&FR~6%ST??AlR(sGy<K-|M$-^5*O{4UVZtPn4}Ua8|}YP~O|xg?0B8y5DINFXs~(
zmh-6$>--k+SFE$<=lp7|rNb82e!Avp*{w>4=&3{9e8PSMdzQ$%06Vn%J`Cf*PL9{<
zLHQ0sy!5m?r{JBP4zVAnDEE^xY<K?S^!=^fz9-UjyZW8}&dh23(Vj0l?H8HPZXd4m
zqcYyEKb@xQ=8J)Qay)(Cd8-x3yXDi?(qq@3cSz~5{p93kD{ncdBtCFkoogChm!WRm
z&@KOh%hThIo61z1)#0TK6;!nI=hE(U`)j*+<`?juwO^!Ma`7LQ9qP_$h}X@>x_G<q
zJIM8}E?nxG=aUXLJ!ILT?%gpwxJB&WE7a{9<2}Bd`c>{FbSTf~;}z!n9MPg%dvTlJ
z?*230cA?YH?w-u4Jvm(;M4C^o-~A5T^*`TB(EYwZ#@n3(k<L3{xAx9=qU=!T$K`w#
zX&C*zMt;{X`$cyjXgA(i_se!ZIO5Y!2F*LF?2zviesj~Xve$&RTd&-+;F4YAcUd_m
zFFAK@jvuU-SKKuCqp6d_zfWV`nYQQh#y3~|(vt7}<oa9MH}{UZ`_<3$e(*`z(QE$i
zd#Qhaznp%cy`X$&CFO?d<X&cNSf(V(^}U^)^1U?>eL(rU1Lhf>pOCx}mg7zC+o=z9
zd!DQN-KOr`LdHw|yb9~i`QXJ4b>l}DPe1BV_NSeF$nht~0b$9f4D0TFb?0kt<IzKh
z{4PW{&&KK;)2_em(w)xtGN0XdrSI-mfBE{ic36HP>lk@|?=<Z6cNFB4eo)%649juB
zeqPRJeud?`6R*Nv>&raJJYa{@8%%BgXsZKCqMXlkVcoqL@$8qKqP*9#3+sMQB;!AS
zedf98Gqzcv{GF~Y%<n-Z>c%U6D*ue}?GN0zMCnlX9?34Ab&5p#T_U{m-5njm^XJbb
z@4dTZh9!~Txgigf@8o4zp3mv#8F@a)@0M@a+ImMuxeqP3xFfaK?p{0puyiQ>OQd1e
zgK|BhJAbi@m*?IxeC0FWZ>v-KfU?8=7wy?M_Juu4BJ->r>iS6+uUpR$@AlkCr=RZJ
z$Syw5;mPNANjs!KT|0=^twRz$UU#k?lz6A#-*TEyu5+D+b^9lsKiAo-jMvTgy8AXc
zew~Jwyxi*iiDy=XzkeyemoRSp*102gEVCr?&QAI>`Thrg$+?2OUzYQ!uH8Y4m+SUe
zg?VQqzYD1Qeu<24^w<->pLq0u^~LiQro1=ya2R?2WQPqdU0q|-ki8aI?eN(5)xHx}
zou@L(4KK9!A1L>&x_)LomS`y7&&MjP<4MM||M41SJ#NEx@8wOpzu#4I%CFmgwaf2x
zUOij#ZQ0#-?x>B0LOQhjovYn@HK)JV(dN^1DCh1v)O{E7`dcg1Yp^t1DBc+9_hb^K
z9QPtYL(T*ILe`nFMtLp~ZP@PnNp8~_#+*CZq3&LSb6Godd+!rze%dA1k%Z+rmtC0q
zKd(@J*FhJi9R~k519G3N3+wh9#OvNO6L0tXh)C0Q=LN*;zH_IGm-{>!9y<BZzT%^f
zTVA2=TsR!>@H;#5oIr-_ws_>nTTh&{?9gr<%Xzg9U!7EaSd|+4EjyI@$*|pdJn6C=
zVcmDP!o1r2eOKN3O`fmFu-$j=oTkgad*?LFXxDvrebAU3C&u;39de2n_uKaT^;?cx
zI^=uac37<A?h5-~K0l;G^o|teeMii~;_r8HFOk*I>A9~iAAhF-Ubxrfo<hFs;T;U`
zju?;J<1mgH=iCe97kE!r|Jax1wp}?h{C5qJ$2lwdus&hk#CrpIzDihk9xdasXWn1v
z&a<4ZZ=LSr(KAx)G<x@WA6~M{0(I~2h}Zp{5M8|7Prws);TG-QwPQb^?p>5F9{Y5<
zKcqd_r5&=qc8bN@zuxWY@0Td2h`l>S#*bTM{MjLV@;iPz#2+|Cr~O1edEaXn#y_}4
zJ3A%a?fj+NU+UtW+L_E}*RQ18{XRsdV-HS|dBf?t2)*1Q_1K|&7h)IQJ}_Isr$g(N
zQ*`YQBm6soy5G_0;^qCOE^N22km>jx-p3h??f1b9-&uq2NPox7`sHjz`Q01Bb6=iN
z@Yui34Y@`3DWLpbfz$J^Nc)5JmDBl>e7f&?>EbzmiB!)-E$8&yCenOP*FDkZi}dd_
z*_D&)#Ir5TIYLko{}40`D&JCXr{_Ac&L_V+C&O}|!@L37?W1Kn?;E{Fuig*q?i-1>
z``t&>>9-cz(d6j&7U!T&k?$JWq3-<4t9X8&;ucv~M2l|gwac&TPrjp~`<<hVm%p=;
zVVR$EOuO;N@4Z33a|_DfcW|ByHo5QBeHUh*pg=jV%X3ZnI~1qq7VxD*ewPw0I{j`h
zziYQcez)Khy?!?uYkBhbYIb2cUvU2tZ`A#cEUU-!Zk~HCuTl417G1p4-&K&$t9e{^
zFO_s|fgz`>XAjw--F<hY-}jW|+@4oE&9D0|xo-c>dM=FO@6U0M;u>xKUM$}|h!pvr
zyhL3)y^go@k3mWI`*&=1-#v*oJ*%8&yZt?9qU2xEw#mdR4yUW2qI|Cq?{KvHF}rmz
z>jXP=>MynDmYTmmuiID2c=jnlLEb?{i}GGaes_tm)AbVlX!l-7zIT*i`S;tAE5E-i
z!@76J#M_NGyZ2LapDV-Cu4I_^VLG%sm(#t&AiiQ~?i(+sp0Ff7{$q>n_v}ax+2L;u
zK1v?<O{Qgs^hcthY_~2f&pTvTX9x0}BGzGf4x;<rxs2DHcj($F|GuRR>)yTF#dqKT
z;~gnE4q71JkF!JFJ)zroQ=DH(l;;b&aE+QtpY?fWlhUDX-^9I_+wsEh528iRm7_(s
z`bV3e`Nt{BaV*2~+>x-|@2*I1JE#AO&NnZ#bVxqGaZhruqI>Ebwd_!y)9dCP;@jT*
zT*vX1_gPNS&YtAoJ095PqGc5-Yz_ZCF<qGVrPANog>~ak7ay&C+LhOxSH(M?cjR^$
zt>4j=*Zp1>JB$`}zjJmP?{;4CD!-jQ`IYYU-a+Q`OP{PFSH92kDy%ziw>#go+h4>g
zUH1-J-d7TKyD!)Id%JjkXX+H?zSU{i&c4~t`HfEXJzMhm_5Fib%a!{-8Mgc0o=n%-
zt32N@;VtE+&0DXc<w3`{jNe}9f~Q0G@3cU>eRS*Z9{+vhg9ZQb_YL{3BFNv3bc*u3
zWKP4pKe0pk9z^nx)xuuCTeCZ-C*7^xQVx{ATa{tm_etEoKRI{%MN`Um-LBlCT)*0d
zb@$s&<K2E|Jt+B0wdhbi`{={Upy9#Tf&Dm$bbgZeb56rv&mWP>i?rSGF2}x0r-$x+
zsn_v#`_-*CT+k=Stl3J3%*#4te{F|O-wEh`Z*F_<nx0x~^1!`Hcg`Q}c{=wm{2o|`
z@^>q|7tAUs_jPA$SpIG%R$<+Fc&y@e{><(AJ68Ew4@diMZ$yhvZ!9(}({hWtdkW&^
z{g&=MfZcoWHpP0CZE$#n0>kfPGk0w7)@r%pdma+;%bfcX2IW19?%rDW4!~)=Zhq0l
z%X=rCU7C1r7>Lz)^+wzOlVZ32N%Zt+?^Dt`rL`I~d3f@_8A&alQY@u-N^<YCCTS_z
z{`@z!Kx*OCqA5u|(^8B78LpU;oRZvdZpmJ0siXSTDYN+hh~$62oYSxSz4!04?2!C^
z;SIZgp7vIiomS9rNb!n&dTgC&1r=YazBPZ_YiFyVVo-UPdUvSRWcD{H>ny*~ukjnH
z+`Uf}Z<ph$Lsn4m#~E8XuiksiiWcqoYF7AE`XA?ITG8T|TKTFT8+F(+=!^%ue$k}|
zZQhdl_9(h8{XMnaN>qgRU7P+W_d|zO$|?3OJiTS*N46-x5IxW@UU+f6(N&VBEVU9H
zB|rT#KECyVz4NVTvF56g(_hQCNhJ!(c6uFtq;}^!uNc3`iWPdTzhCm=wT?{`yNgs0
z{8&)2-treNE%fP6$}N_>Zs_b*d-hs(SiJqAwx4g_VnvD{%sX$#{OrG3UgOlUzg^$^
z$~jh|qu=`)ujNPTpMI$_qw3IqZBROlR4(NrwZB9wmt#n~8<s!0+H#A>-sn|5_o(EM
zQ@rN7Lz%Cp<qLU@D~_BQ{MtR)La{>nKT@nV<kATR=44o2p;JBl-cLBy!{mz%15Ynr
zwdm5`nJOsQsMMIXnRQMnzc5Gsn={^-bWC}Lo1Xoq>3OY>tDxb)4fjo7-*&I^8vU|r
zx9do+^83{uJGoWcHy-;-zod{~Xjgx%(kp!1dB^PZgG%D20Z&Yr-tw^0VYUkCb!$#I
zq`bx%MF+NAT_jyej1?b|){}nC%-PegKW(|iNb93qwCh;pfL^Un9)2%*!Z*)sw%j6m
zM2n@XcdVMxf49nNSa-*`a;2ATRY60)&YS&`JGFoR33(>(P+0|6)S5G*`k~)cP%vor
zE%l8z&(Zo%U3od)bouP!|7g^vSo7+~tgM1Be}C1E3!0`|Zc*MlIt@Sa=!d7CeCc;9
z(a^5n<DGunqN>@hJUmSW4I|wr#+&b;)&JS%Rlm#`{fgxk7JU1X5*gE$D6dh@H$e~A
zZ+g1NZ8scLexcX<)IpCnZ#VImBPwX<_c>YSH>z`w5xXqE(e8dS(bMI7$3zWBI`23w
zx9;N-XPPxTpn{HaU7YA)r~PB{`EyPWuhE|0#k#!Te%-Bnzt)%BU%Y%>o{{SZE>{L)
z!$Z*aWkJ_B(cejBRsFo)4@N7`FF&f2i+}YBb@dSM*S<1Q<odm@^Gp7z8Sm{VQhU3~
zDj2kLn4sTpJ<@ymgir5StAdV~x2U$V$cZs3-iW`97bfaCh*Q0$PpI51b@x|Rtk7vc
zO&T!eh8Obhv4V<De<$FS51DVn_8PZ+RB59XG?eY|IvnqN>`3ePng_4tIX&-`_2y1b
z8@~7R(<)MIvE}>EIuuNr9V;Ai`)9?nuQ;LnM#+=kVY}ZA_*H+G@!bb>y!VKeRj|g(
z&oxb*muUqZqx~*Htjh=G`w}`m6YbrLojz}D>T+9yANO0aM!)=!SNmM+&?Q^HSv}H{
z=+}9>d{-W{aHQ{2{IY|1$?eeN(-O5Zeo%g)SN0uiIZpSfLCF`?bDc={EpE%jYy6q-
zC;jj5Cq>&Yg_1|+>-)k%%V1*ikX7ET>&CIPU*4-kin`wsI*s@G{br={oa&os`69jN
zk#=Ch|H**U?+Zn`PIdeJA-C%#xAzo7_xGB%`>hpLP|@k{YrM|)pH_4Bg(BUzo~`BL
zWg$-QB@#8C{EmUw;du3nte1C(@_iidz~sCn*CX<KT$BUlI|RNXX!m_Z-S?~9-m7}`
zt~p-*Lw>(A+Vd9p4+^?nxA>L6&BL#@J9Y6c<rM2pPI-RGx<6D<F(}`e(&_os?jGfS
z?v|l~hLP@z<5dp(mh*~VVYk2Q@cX@wuARr9UbCsi*jypIdCo3f&dWg$`*j~D<wm+s
zc3MuZd+fq?`&`ybexYCUVxq})+JAQWbmLma+qK{AI*awVQ<Up-ufvg^_jy%Lu786P
zwj1w>p6+&^WtU%$n`pys&nIPmS-;)>KGJpM$rbf(oO$DK+P?!O=Pkng?z_x`^V*I5
zcAWxQZ-IQ5j`iQ^?zL+EaQl9xL)snmcYh78Yg~0yt)nW|2!D}ctlw?P`gD3X%e-Uw
zZADI6iH7mIAJy?B<NZ1plJ&R^+pTX*y1(zfobv0oU+waHb>0~-If?T3Cg9m`|3f_z
zgZ7?O>MQRFf)>`DD<*n;r7;Tzlss}&Wi|Ag_j|ASxbC2Q8J1Vb_;q_e8EO7l?H_5E
zSLcp)=be6~(+-1%0mFcE2Asan;QYZUa=zgf?e1Tlx9qQ+PVdvbUYEv7o?AZb^2h4@
z)lLt)^vb<2toh9GQ_5?!TUWWQ&+gnYQPRC$zsKwM2&3=+s>l1o4p{L*?AGnyGl_Tp
z#kF@<`Ka+WD^>_k65W0m->>|3zuTATvK?NBBmIs;qLfP%KNqPUk^T-OQOd>YJ05=3
zlTrH06`QWjHan;o>HT7&lykf8i8Mcc)9&8I?RNm(?nB6*_R2YP?))iT=`dE`b+mhL
z!F{n)bh?itpWWXtu>0<#o!`~@NvG|R`RwW=J<@fr-9D9cyZdYO@CqZfN5-8&e;D}X
zoPOQ!y?>wO79-Ur*565rRL?{$=k#36<jWd=m&0v8>ed56iFeyxliyoAPn31N)WhxH
zC64tx<Fvj=?_CqET)fUvy{fmy!#|#yz3`+JRFvn7k*-(Ve%~uz^N8EuGn{St?an1+
zy4&CT@IKrQ?S7Y{OSd~mkm+vi*6w#w@k(bu9yI*=_||+^mN=p!#Yp{?U*&ofU6=lz
z+HS=f{c4Zba%1gJz1BZkc}{=N>o#AsdO0mWvgoyyX0*v33ObI<^J({q<Ih+@#o9C8
zd8Y7{Ng;#r;o+ll&DP%e!Er07IDf@}MO}9tva$+#_1$=<e$cOcZtwkMe(tlJqAYJ0
zmiJh0-?iK6<#hdOC&#b;L9Xt5p+SlF`~6s~-h0s>iH5rUsf?HV>7ayVJAxK=`uj$w
z`C@H<(VjEJN?*Hko=DSm`sZxhaqK1QlV?lpG_~UMUHY9=Zc$fHwDEqA^PtQ_iLR&L
z{e*mZPh}T&`wm8~+X&mOqwLc8euv+PzLBEtepfe+BlSyOm(%I(cf5`lzvJ!pAG&nE
z^)mARI|E*SA3WCb{Qf(z|D81xD;aS6E+Z)Ub$_?-Y>D^!_wdqg2uHggGUaDf9r~{g
zO7Z}|LBX^Ae$nfG^7{9ZPW6lR{duJI1?AnB-`|0Xm;6LOKQQ|J@4pXd_=zn)PW`XU
z7xZ@qvOPfw^PMc6z9p|4I=j`Ly_VmotDktczxztm{C=N5z*n^B_T6ow<(J>HK3l`O
zzZ2zmJboo!7_I%sy1cwkY&hiA->ztvG}|rurMKI^UzsTRW4&LutIx0gF>?Q6Kz>)q
zukiW{)6x%>S#EiSc7Got;nVXxma*gH9<?DtOzV`^Dz$t{@stuN$-UE>q@`s0^WW40
zsfAOErX=-DOD+CqxME6jN^--wC3~f%j_OmV%;Kb^6Cd8Zrr)XER?|09Di55x#d`0v
zS6{e&#$GF9(Ca@ndTP5hwM*-roqn7jPA4Du?Z5?t3ijF-POtE8-Q>6K+hMi-<hz^m
zE&1Jgd(;o@&b@q>b@%p<j$XNZmG#=1iDj>;vfO%bNb^DM3;$xB^XQY;ztna1Kf>fA
zo$}}_%TGOL`RXp?-?YG+j%-_6xYkG7`q2NEDb{DNkNLx*9n^#TQ>RCMRJqiz)`xpb
zzNhLPvIvt;qSO~Tl1JJ{I&wkslTLop(HFi!^dTSR(MuwHg>mffwdZ&FP?2w5+MZ(B
z+e``by1VD1Sw%MgW>FvYP%nIvk9Np><VOzuL74nff6AdB7}o!9JE)X%$H`SqW+>vx
zhdnOcRO<UVV-JM;_4*m#UXc6Q{b4(%e&T7DMEU`{thr-YmD<l`gyo<wSSe%7FIPOj
zE^P0#4@6JGAiT)>qYXzp?_K=$l*YwNoKQWkTRy$>*yHMz$<Kavf3u^i{-iHnxS(UE
z`t76Z&Kv&zVYT7n^|`y=e=v;s-@CEKs1xbx=?h+)eaoc1YRo&cekxPxfLdSX%;n#|
zvQI55cuVQt>AThH!CjJD7EV_mk9}cHvo8;+XG%05ex_G?IA84%C))kGEkoUQ+5H_)
zUVb>74|yZb?YX;dp55x|+_}3hsJ=6dbrzRhFs9lDHKR_8&mY{fSiLu`(uEUttq9|q
z`D2@3+H##beDIB$cg)(X$WOj`<F2^nlbe^SLX{rR-fHOla61+x^(mb4$o6o#%9$xo
z_aC=K<(N6HSB06Y!gA0D{aPH)H}2;FTh)!l=k(1v_ji^1{OotnTel~SpR5>lP1&5k
zhIxSx_}yM*@cCcw%nb8CZR$1Khu(Tj?f&B61#7AtS4Agg-n_ogk#K)d?z#8B9#d-Z
zL3LBnk&n$DaX@`kDs-g#;6K7VQVu@JNB@Dyp?_!>d=M^p>8yvdwb`%ew{aI<w5(Q#
zjBr0w5B*L)<kQdSK{|GWJ;5*O=!G33k96W`5Apav#xLWM`Gx+!>7us(YJKqgu>D{k
zAnC~K-S^>HPe1vS1>z@*FZ<UQ(;r-F%^TBWOoQ`xTNn3v<4}`@o5S-b`SAy&<A*4B
z{!53?Y1ZnfHMK{zNhL2nVcl}%o|;{9WLmp=r*FA0_sMV=|3-QI6Y~vvQ7=5yc&Jvz
za#asls~&%8!EbvqEabfYR=$=?hi?t*P5AMgr3z2jzbV`f<l|>S=2`qG?dh=lfk$6$
zaWK4|D3QHs;TI<yx4>?Bmv%Y&(n;%`MbisEp^`#ZG#WH1?={Jx69sa<bbtHg(2%=}
zlzcr`_E4^E`}$s5CU@xC7LC8YbxzLEdnM~!RJ?MIFt%+~;6RU$bA=YBjZL}k=WOBf
z^YY(z>jgcNLO(9)RByug<Pba#dV5>ZV_T9#)mCRrt?*Lz5c#6@t7l93Tw6CBP41s=
zF@Iu5tV2Q4neVQgknOymcI*kSSFw|3!*kX6|L#P?>l4D1XI%>tpW}<u+vfM%7oNAV
z$9ImFsrgC$4dL}H>snckcA&q+-+!&xe@fEIuzbcFau`Rn55(VqtY^CY{>L2~^Gs3r
zlW_Xq>viH8w~RN&H~t7DKka4xhMx2Th<{|hB8;DXsbYx<HH#hz`*q@(cUUj5PGFv5
z-fC3+{8z90`k2B`PyT=G-Fdv7)4f0NbCQS<L=7S4c~(Qss?Wx4sZ&9WRV^aosOhS>
zsF{j+>`_`X*L)lEkP6R1D59tcK@1@=#S}wGNb-BR&3pUZrH?(Fz4zJsWc}k~t!EA2
z^;zF(t>-yAtH1E-?`KTO$9et4r^VLYd7G)ve3{Be#7B12{_#KN=riW^eEe^l{`k1w
zzn_qbOX5ZA+>hSTYw34VaZ4To#bt3)+))p$SDX^p@Fh;^AC`?{9{uqwZmL(t7oL^z
zghyrl!gy={?8ApY*<sl!hku>!c>ke8cWM4QwNJ)ZHm*ESy#B#*f7$H*Ek4WVC&txp
z_@)<3y8EoJ^5ftC@JExM8$CUR!v_E9=Z|mI)NX%w`{~nX4cU4|3d#5LcMhM>I-;q)
z?~N00yLU=c`<wgke0;BUn)Ch6-uj$}?pSALKEJ^`6yL;G=-j|w`2lfIJNDu{^U_CS
zpPsSK=lOWcPw+SV*nxXb|MkE{Cnuf{IOmZ5FT6NC>3zT^Gnanu)5JXHWrxP$FT_dn
z<_?~^-k$e=mK=WQy}!Nuy3Z5Ompt><MKAvN>!i=87tQgL#iu0tTc7oc-~6F`gC9vJ
ztuW}8YsV(?7XAYd%6R3sRzGv_O~0SNDL10+gW)^>&&<R+w3lD;Py8ppYW;X0_Vshu
zef9j8`E_VV4*l^hAF*EY7-!=8`@Hm<%|@lr`uRockslpA?2VCM-#$Fg1AqM5f9^m3
zk;{)vtVer(+<6bGhsME+^;kC^=!gBaukyaD295aRRn4E|`Jrb=O<nlV5g&Y!$e-*-
zm^W&NAMNj%=e#>M+;@5+?|Hw^cJKaI{~1XZ>Mzf;AJ!r7qrb*kcES0UzI(%&$(Jq9
zAMwOCO*trUUwhcix2)E8bn4tsFU3`MWL|vjb<T>f51;z29?B>2M;^S}fBIlw=mWH$
z>{@$yx%&tD$6t_}JsL;+F}}D4`C)MniUa&If2=-?2U+8YJH}Op{3kiZJ?A~~#rWbk
z|1bWFXZl0+)?qyL5Fd?-^ewbw`WW$F@hb8VdH5!83H5+~7H9cO82WMkTmB&~ikI>y
zaYKB7VO$gs#Vh64j@aP&C64+cVc-0}`NRQvi*cP#_*dr==L_eD8Qr(u`ix(^l{(jG
zr=NEGwY;A{R?m;>XIy0{FFX6e4=&p7`ZH6=kIU1Y^Ner4wHF+D>&3fv%gwxK;E6}v
z)jhXw-;>{ZA=f=$R=?gQ%ieSM<vnuyy>#2Vg9r7<DHn_OP2TK2Ak1Uwhx4+0?v7dS
zL-y};$j+_(+w=LL^Rj#cAKJ<LJeT(>)6WpeK~5;1;ZNQZ?w5TI(Y>a)>0IgDD{j*p
z_ulTW`@X#7l~=s*TIxR2bLhGIP-XYpSs2bs-{n2{#~Xf?q5bfjKFXt^=k7zUz3TS|
ztkgVzE;*$4-%s3r?ws|xFR%_gIhT_YU-)JpP<!iizhJx&t%DvwdP2`aImtr~`$j(R
zE#%3eKH|%{4nO9#59A|<_Zs%WIL<Bk{=xfM<JlMIBm3ih2fe{xcz=RldT8IA+vubB
zJl0Q6>mWZ~4}NgPPp&<CZ2rE;`1T90)=y8g(_cL_zWJ;luj<K1ZttV)GyO7;_fyuV
zY(DKmy}&QN`Frd1oL<=negqHJjSutdZ#?k{AL{ud@{z-Q<~OhVDd!RA8s`J(e8c}a
z_qfLj;~+VlJItq@`Jj6(^OIA2HlOy&;-vU^TKn1~E`0Cn{JoU>OmZ4u{U?tOTX~`T
zXQX&AuKCGt-Qury<W_bbGoJS3#v^&$$1T6b{JXBd=k#>tCl31cj@Qjh)w_?A2dH;H
z=Uy&6|KUMBJs0=LktL_~h?nk3oUg-sD)(H@-R_C_L3a5aKQZgO2=mA*%_o192gpC1
z>&c_N`w#J6KI*-Oe8oM2yfyD<XZcrtk3WUtsrjtax{br{=x09b7mw8AN!~4g5<l>*
zUO(gTXW{|>#ozHC{EYU>{6|*3_#ke$kB~QImDv%bhvtL)w|FYQci$?Hci-wf;CukB
zM}82-vCxn6*VbkI)=NG-%4hjyeiqVmXumz@-}!Cxi%;UVvV6mQ^wRzs2fy}NJ8|2*
z;sAcMH;?DW!-sR4dq-vWe(pKl2Zs8Y&tGP}S7S%?7xDuk+7Igpk)L4C{DX6#{qVk$
zU6GUjV!xi#t5eVa;UD@wJR@=M=swu}s`HU^G`&^EoA=#6Kkm-WFKL;Y(Cck`F0l3R
z3%<(VAG;TJZ%g0RYw!Jrdv|gwYfn!1sN^9h{@j;o@7|Xmb<TF~_FO+auwU(sZ(e2R
z`8Q|&aoi^DQxo?D&hg>_Ir(?{=y@ogeYalY(^GufN8{f#<cKR@Id@9_dk=afp6IWi
zeSr4UzT@9|#4G%|PZ!t375(T5y)nN2<Ot{UJ%)WSj{f+7{Hl4>(?4>1--B<w;E7(4
zt5o!!%slMVeuw&H-RkjT|M4#Vlat=^SIYRZFZg2j^wqrD^E>vDe&YjQ+KczvvlITt
zI`osb@Jr_N-p4rBBVO4z{tdr)Wk2#2aag`#K6!_DCSDlFe#wK3%dh)h$o<p60Xx09
z^^)E4-}&%c{Dbw|SNnq}`)Gd)VW_9#m;JF{=8gET_wA8~$ip|^U37Ilke@ow@z2hk
z@<Q!G<R9g=;yC}yFFHT)kNl=G|E%mh82VxJkk`2KdH&x$k#iAru5mx)e4<`{t-o=d
zL!3X9okO(K-}%eD&^b_9du8$)-*ZS_agqO)hl&sUd#JzSuzXK^H;(7}s~2a*Q+bB6
z_!PDy4>Z1ghWd+x<Pv9{@5O27JZ1U==>>fikMX9Rd_tZ@f8_z8p3+NkReSB^CB~O;
z+aJ8)35Ifr)ABL)L{9Uu1ITWzS3f)|vvc+A96yl#dLH&;kNC=Jr=C5BNDt^8Iq^tN
z{97L$v@;H#@T#4B81MAgbL)fZ=@B{XyYa$vyZdtfB1HZs3;B~$(S38NIChN@zq#P5
zxpMM+{votSdg<KfoTELvbw2k!eWCZY&3l~xhn_F?$UV8+v;F`6QTP0NTH~^JywV@<
zJKXp12lUc=VdIipKCiv~F^>DFu-?4%hkohD{`enyBu<*&_ip}9fE@1Wp!0_JBHnBJ
z`vm6)yo(e1LHBL=qPOlH>6thrKB^bb#5?1L^Jy<mX|JBXyVtTV>$ShmJ^JGj@9y95
zkMHpQ(D}|f>5=_XhV(+&I>}+Zo-0G^%tAcs@19&44|u_&@$HW?x$Lj{aDM%u{iWaJ
z(oZ{O_o41jwWm+~BR%Ku#U*}}|K>;4i#Pf~<FaStu^;y}^xr)W|Ksm>_#yWU?g`nC
ze)P+_?1%d~@f*+h$43|kvdZ)<L~*5BNYAQ;_A3kVnuXr?Wnrb|X5ajd`*MCB(mTjs
zvtR3xw}*ORe0Jv?z@C)Z8UM)sjK_b5_GujT&^XFaz446Ck6Mqso1U@<`=Bgds^?cc
z*PlNQdB>0TkexXfX~%yX$9$gK2W9?LJ!GHmo#pH9jpak~ASjOWAN&gBU-&71=jZSL
z#Bc45BY$<D>AWN!y04dqc`x8z*1fH~EIj`mcj943ytLl5Bt+}CF7efRjpMz7^Bm+C
zv&#IceA_z1cp%P0@j<@tJi;#d8Rrl2N4@=p^e@ya`b3ZDD|?`S;<5Vy`G$C<y}XtF
z;1BQi&po$%Ok8At<lzUz8FF}k%dW}oUKmf-$8X`6p82lA_ZhzD@ON0gs~CUnO2-^E
zZ&SPWzQcHD^6jUrbY%DTJ3jn$ySLBj)_(NLJs0VBQd7I{JA7{feUG86o$p5cJ*>Yc
zac}<g63qj4S@7GxyLG=V53uk2I6mDk&`)|qAMHno=D%ju>8tE`Z?D{@Z~wJ@pSgPF
z*ta}`oy#{azv_zR)<3jo?#g|iT5)9CoVi{%EP2iyTlUK1%^Tml{<%M!H|Jbr9Ck!s
z*uVP&_c8Jj_UYciy@dBy`1;+jOQx>VckcY(Da)74V|?}e0RQAY89w+g=PKXR(qndO
zKcM(&U-&af9`f4{a@aR^&VPjc_;LH;+{Dk)ul+B7{r%5Rog?S(w!P=%SLp{{=nwnS
z&U13pM}KETU-&_G$8U=__~n1C-+J(%9r^7u`Jj3D8D)A#Z^R4ngS^hCVf?b+J(g)d
z^}?^`$jRUFLT~v$JQ&A3>ct=ak$%Xp_}4HFXy<*Ib(`Nj)?ps!bbgkf43QnMGkQi}
z#by4K9G;WMy_R}$RK0aOhtLQ2f8u3`><a%-K1MG5DTjy;any5u)cIK)k>}xyy(`Nb
z#aqZ;*p+_H@%ERW2+u**A<pIV)>+@%DvOs;y!TxEmj8<5^etQm{h`0?lRZLqPw(iB
zGCPBwL-vk;yo5XyD)W!*oxk93_$zuwK7PXZp6kcH*&X@pH@$%JZu%oXqF3UX_~Raz
zoq29v^3fmm2H6$6V?TJ{f7ywB#+P-erw{Cj+~N?s<xkyPi%a;!FMq4OefQitjANbF
zBadMx?B8?qvNz`+=PCAsXZPgzHJ)|ipFH__XMHbfJ@f!y_McpkzjbaS4}RIFd>gV~
z_Ut~0Uh#|cP@ZeQ%}Xxmdq3zLZa#T-#DDz`RjqiS_xy$X3g6v2_s|#n%P#3HJy#Zg
z*bzU&Z}Y$E`2~9CoF#6!M<AE?F7%6EW0&F#q^JBn|0REsUl@nqaK2Q}Z-`6sSoPM+
zU&%Z0MLzb(uJOem$`c@d_y=~*pXl$oarhDQo6oxKmwxt<oqKM4cJBN^4&#St9(>ww
z`I)#uE^z@*@|JLZa>!5kIewTwQxCPL&-^vy@9Ymh=eaV!PcQi~WvE^rM6cDmhlKQ0
zJKsyQLw=K8_%5D3IPZt(??b|Rf1l_+z~7rd{qQDV5!d+-dSZW-*_HjLFX4XDH+slU
z=>t8WcjlF+xOa*e8Hfy&&wzL#euzK(pE$(-iLbEynPbmJWMH0?)*aA)srhraKDh9Y
z|9W<x+`_pHu9~-BpPV=;Zn)17-`szQ%i<jrAC<*Vaa(*B@72>oc`DRjTz0>&pK(I;
z9#I@7kNJ(~y`cP5JQY91UG>_D=jM?I;6<D_PM8PC*F&rqU2<RV+%2D$FT3xMH_H#?
z8S*fBe#npf8ehete*9gfk>1hcETlgn(lh-EtqVVpow;97CYN@`HNW}H57pyeJ*3b6
zj#-&r>Idny`N;#xqkniVB#-sT+rxVm_DxRXDw79>^N>$J<t!vWzR6Vx`Acz?-!u+%
zU+cc#`AR-4Up8Op4?}%$kHqhhACk*@wS#!X8^2C|?R+<7KiqF<=YGk1*icS#$Ul^!
zap(h-$LNO_eg&U+H?MQ7@%0P+1bveyLGtAJnDy@*7#|u}dvVfp`=e}Mv^SsiD3i;&
z=#P5nxqCr8Ie%$q-S{#O`N^j}-aU7}WMA-V9C{n}!=K-YfWBMu-W>Yf1;0mn(qR`Y
z^XU8C+n4<O(EIk?yGQ=_M*R-E-!t_)2J!;GU*LBN#BcH1bM=15f#1^(ig(V<?l-)j
zaIfqBh5k7g;|1Cu{hW`*P3L6iaPzwdcc10noE+Xidw=VFi}#}Zje95eHr|KxL(cK!
zqM!7^{?k8t?zw%ZcgCY9){hT*r@eOiLGt0py@d66j(7F$8MKG?$@)Th>`z!v@7X2$
zf!0q?tj9W`c|Et^;)D0F(0f+zrRarywNCkwdU+Xn@XW658~gS?%DGWJ9{3UbS+8|z
zXCCd9VYpB9k)0U_PsYI)q#uxe(QA4D>5u2^u2_^O@jLvMvfrKYoe+QKyCA<u<M)~Q
z9q(88L+y;m@9|&!UpP-F2l<TayCvhP=MTw6ZomI*9PNxlK63b79%bhr{!|$nM}86V
z!9Tk%h4MajXa6Dnh4Mi8r~HRp@>;09JV@R{PIlt_Cy!F^T<P5B{D?olcck7uuzt|_
zSN>o<^12YMi@fBwZv0uN{lbI19lxGi2mM#pkN%1a_Jx1RLi~}_`uJad%f2eJ|FA4>
zd7g#*2|q)=&_9HJA}o79%%6qmIXTG7uT~qaOWdZ{^hTN9*niK(IdU3D9*aNi_;>qH
zFW5EI-~9B*bK}u_>o6{Trl;gJPOZ^>h`4N@#U1*qEZ(RWx2)Sb)zhD_Jzm0b_4i!a
z`0^n6LDs#Ke3L!WXZf~uvJ3Ob3&^h>`%tePe<Xf6FZdlYaYlS{uO}XAFJ90O=Tdna
zy)z$wqdoo3Lhp}#mj~tDP+l%ybC0GhFPGPOukQOc@2A5&pT4uJVllJ_c5Xa;sK>A8
zA@bW{y>TGF0@a88vdY?<PuY0lqy1A4wGaD;_1YT;hUXjcogP8;?w7@1<3M)9@9|sm
zUw(wWhZx$Cd|95%z94@gUl08?|De5g?i1K2d)3}Jkl#0s_MVeZ**N@>_<~Q!|KkbI
z@-TMlzRYv&*>{+i*l)i_>GvPyrScyBNS-VYhWwIz*Zl^+Bu|&`u~+$?dE`~(;-BpY
z-terRKZf|HC-g)AkavC&k9en_^jR6wZ}PI=EI$+0;~gKNzhR%o5A$pJr#zW`8_#p&
zLhZ?mfB7e*KgMCd_<`aJ{etwF{_-P`-rEoRAbuK0JNBpy^`{T^ncedr#&=JkUVHu3
z(;I%teS-W3`ffxXLI0GY@3!c*{>F8GMn3t5e&q1|2|p=65pR{vYdrTN-dC{~>%@aR
z=s({(aLU!6^vcaY_j$V=xbFfvf8SCn9=y^|K3(wXIdgFIfB$B;A-m6++wA6X4}Y@8
zTzT|&7~$W;n9uJI=y&Z0tFQ6K%X8(_lh-^OFMs^~&rR%^|9c$cK<yy8^i%fS-{<Jh
zpVKRP>wN${r=R=_yV9ON;kVg2`{I9;_1E70H+x`rkR65ZG0f|pnBVtZ6@Sf-^I!Y}
z|IJVFE9^Z)dgXk>-#h<{XXN4^&0~J~4mt28epyc;6xZCpSigMRc;cS?8XwB!v>x;N
z4oy38(Y(<3`n#7BSHp5xZ`?5cY6qPg@JUYcLGeJm;Q#e=e`4RnLGi@?8V9e&x8ATU
zPk`1%eth909G~9ek)Dv7yq?>)5c$Ut`B8r-;P31BVgA>93f~KRu06l(yz0J1+=2YE
zxZu1juT*yL&+nT@e6bGpXMX+6@B9j_SG>~CbNBVWgLS?$uX7$=jALGL4L{^`|F55R
z)<s_9h>POAI3DJS{I$3$PAlWpxZ<+>-}?f2CBJT5?cA%`U;dZ>CKvy&tiOH5t9^y+
ziTy(L?pM`^7`D@%zfy+$pnmL*-N-lDJ-;Cz)!uXCkt-~#50QNAnLOH&mwhW2JHByX
zIKFyF9_2zvu0rkDLud!=Lmo`;=#hA@EZ=Z04D$x@R$ic;xGk@T#uuNV{3M*$_~LYE
zXU?VM<S*R^@gu&MXHPJ!XV3Cj_DOE@K=qJ)l1n}Lt;gTJ-n8o2L;qvz1#<F#^0QO^
zkiU?hg>pGp$z#M(_fXD%>b+;o^F8bDtmuXD`73@!KS+PVe){84d-KU_tp`u}JapFY
znBy6*A&PVKnxChCc!%_j9vff2ZGQ5|ON{6H6i9yK@Y8tV$3ygfMVzq?af$x0C-#G9
z`XF98&l!gvi?{L({D=#_M`SPVO~py}VLtL&C;9M+XNX7V8}0E;U)|@~U*q9D<d;72
z2l%xg^h`az<wN*kXZ~J^UCJBqDnAfs>96w>J9B<vZ_3a)$9nCnajci!;oCZ_+d9dC
z=Y0JV2duI0rw^xiW54WDKl^|$dQJcABQ(DD^xXL5r7!XjeA_Q_7|;0Tr6<bx(vH95
z&*TgIE2LlAL-EFV^eyxc>Wv5KgFIGViVr9*>SrC!74$^ey0xb-<kH``>e&tbv5)ji
zJwE9%dyuEf$LKTppgc_;;vA>F^HG>D$+zTh#?d}R_XgTIx7!c?hn>(PdWfHJzwzum
zN*?dK%%`1o(X+5$_?*7spPt!I$nNY*sE5|6-g@*WAN`OglM`RoLr(I0u04795xm)V
z>k93RT%K!B9`i!;kzafMk^YgJob=oJ#S{K5jML`Fk8vO#%tJ2o(^sgS=gP)|_)&)J
zko^3gb;;-10r~BNe(LdVec`<L_B_->{NtS+WUb4(@rJMPK9m2Fmx#OW0hGl9`60cN
z-_S>P%}(&hF7OkU`ALW`{?d5tR9rQ$JW!moU;GU}%1+6F7yE^0{Mje%_)+If82Uf^
zsUN>fU+58jtkeAV)jsog%Fwv<HbnN|TnpX9InP>$dj065?`r+tmUE7KFTYbIE{Uh|
zQTe|7Rs7Xn{^9R$<<;_9_fGO+dMU2U`@(aX{Lt@Gxu23h8lS$)e~f2dfB)@#E-!a)
z<UNaa<iu0f_pH{hzr5bxA^Ck4-&@!Raw|Ir=!Z}1F%P+|3xD_~N64e^F??_BoMOH7
zKs#k}S(ooW=!^9k*Y9un{j*D7`Ol@V+jfqe?@#EFbAY^G|8O4jS-*X@zwTX(Yrn`%
zKELlnpU7ig?eT|S`U2_S^wB>$?w#|y<*e85koo;Azn`X^f4>NC^w)ZneNW>ah2G%_
z5Au5JcfRC@_)GrPIYItRzs-*~`>$-h<WeRF`S41g!}ahx^n%=eH`Dh={0pSV<O$J!
zu@`#4PTcR1-}g!MjGq%%#3^}>IK^M_d;A7FqK|mtH_1WYL;2XRah1jEFs_GX_uty%
zf!~1oZGHTsqwjxpzTEY<?Y`EbGw06veYsFSz1PJX#HW2{H|p_9&*{ZQJ1^Mpw4UAb
z?{)wA`dT|r+@NR9?{E1XXZCJBeuy6Mo9qM6;rf-WKSXwbchBuByRcvMmHzP?%H*^E
z`gvb3U!Yg?PF^kFQ6C~bl&`UWa`XSnP<wfldkFH#AKiyQ_8@<dZ$ak|`H?(JekA|$
z-q7!idM_vslczYB$m8Uv*6E%N;s<Z?M*ZZI*2%7|OWvjo<+b`5$8+`Jc%IAqjicT;
z`aynLJN-O|=7st}`$)f_d6k{xjSKOX*N<6$KPXR?mpZ>d`)>cB^P%$qq+ixSKk2vo
z1p7+=<l~Uuc@D`H=HKM;T%05aeZVI^oQK#?Aq@S5yn!F%ANUP%j^E)g*t7R0;d6eT
zJ%|HPym#;9+@~y_@q^-y`NUOu1%GQD{3HMAeYbHTdDt<z$zk7>LpilKj&|1PK8?OO
z*Yhj<k8_dd&P%>e@ZEy%G3Xh+g8Umjr-$r>-tliCnwNZ@JJ<QHB}C83?>YHH<WJZ)
zR39S0YdysxzW}WtALOzR%FsHjt6X%B#;<XdA-$HjLGco**B*K<pQ4XYToRA?5B&2l
z{0|gQ?ThxIA5zv1(#NnYUgL)yK=%^9kMi%c({uL=?8D#J`FGpxD`Y>$vtH*AdgXf|
z-y`$O^qjozwaM%63*FbSAMNB9<ggy~ewW-j{64vVryh!H;xT>o_mJ`#^ZQ;xy?LO0
zfOv4fh(Eu(4)JUpWvCqt_sQ?MFM7~IcWt#;w|473Zqj;--EeJlJ5;YfyTm(v@Vyp)
z!LIC^_WD~FdsJ_KeBVY7>=VB2w|aV`-Z=i=Q(3%`U$bBLE8>;Bh<@^0)@42NI(Y)S
zu`b^q%5#jzzU9Nl)807bkjI7kPcHeAJS5x)^6)485Phbv#uL}%N80gkVceBZT9^L(
zKfWRRq4#*ihyM7cXHYx$s_Y(r;xzlCU-Z~|<$3tvr>#faqOZmgr}-EDgr6V>-t{vM
zeWmC8seey_ocs!V43T}aXJzgB-Z=5Ld#5zz|9u922%6t~?2A6KUwq+%z1t6Zsl9qU
z*$3_Dw{h*C{l=?uh~fzUfN%4P8}xwO-kZ=%`$2B<YEMr3MlSnhfAxp-n|_)vM9=BF
zdCAR=*&)8db&}Ki#1Z53KgKl<UdgRY5Aeg!D%&sel9PWSFW#(AJ}&;NCztk6J;a~q
z)~TMKGA{JJY)6qlmT&PN{GsndLO<zwe*JCl_3gg@?FoNv9K7&9=5wDV|5SFL#cw$$
zLA=mw_Q%iTkG*(~H+iFW^uu{wJQaWBr}9qv$o`!xtpi`;r+)Guc2q8smp_B@0eK#O
z$luDVJg5KsrE#na&+Gt_+xp}i&imdc;Eny@$9+8B=#~2b=X`o%-<_ZBpZyY_$wlAv
zr`Pn4p2>@hPagH|2b9HA`a%Ec9epE@_RxJ)sCU)_@#Z`gp11t_EbgV<yStxoKTmGw
zIQJgVJ+}K9=<kW_1ADYy^4LfBKJHDzdm;DO?l;_H7!RNLcc1K@nSRl$@V?x89PROK
zeD^2z-FnF5IiBp3{h;UaIq`uVvJ3gF{S`0hqwfUB<^GTU`(6OQ_Md(#^NZ|38LHRM
zc;+)dKc$~M7Jtgd;}7v?|JgnJ_Wp-ny5GaE^+N64zqu!*PxMFG`$2kU-}U#LKGP3n
z?@7JC^WFxAde0Ap>m#T4INk@TCkLdz#<za=hmd{phvsz;iEr!iKFNDU<KSC=^VvUo
zDh~;f-{x2Ie(9P?oB#dTYg2Kao`mDm5BIt3kDcQcAN&A4W1sw(GJT}q`td8)OYfBF
zkA3p}hUff__jKOhhVQ%gd+D}!2M_9z|98mz`^?_E(<?|nLch&_lgIl-$j@nK9OFWH
zE0piZ8~L$tAMGdPU&KlKi*J6zKJY8{ucLWrwfoy&@>&PK$ZqJRdU~ZFq{sB0zUj~Z
zi7)I9%1iLePS`#9*o}JQ(|_?vzO7zdlJ~MxaZJ9Vo%dn%#l4QWrrvzy7w71W_#yw4
zcPh(MmBZ)ii$&);`D}PE;W_j?OW*LtF09i&;ZZ$3VGs7rJ{t$`cyun{hn*{&W8Fu{
zGyGktb;#G9U!nE6M^M(@x!8H%Ia|Ij4mgj>chraS@;BNUUtVln=QZbd=M44Mscb$x
zxmOF3oaUjA?g#Xz=i-_?U71|k=}#XZxgh@R8y?7|opSixx+kvrhm$VasayME>uh=N
z^P6>Vckc!9P7df?show@g&(|VubhSW$$Fkuk6+KjaXb&}wJ#O%QEI=iJ%7q>=^g(E
z#Z7rMf52bxhv7X**iU=y%+G(=KZrN`Y2BfoD3cGGhrHsLa{)gm4^S3gtgE9auYlyT
ze*V&W*)9I%HMK_jQY_LR=R9S4=X~hAXkY1{cI*UC_|Q&&_apXCymBu>|HOIy`73-w
z{3<(N$V=o;^vJp0eT4Iw{6&BHyZp!b#&h{U`yj9NK<D=S`evO|l<`8JLZsL7y3h{f
zed4>kPTZ%zp6d_UiG8ELc*n1OlaHyVFZib~_TN6>n|$oZxOlK$`hj0%yj!2=kbLx+
z9pF{oNIvVg9`Y+|Z-0%;@0iy<8xN|NcUhnD$VDFf<HvmT*t#J-rBB9JZ{NvDF3-tj
zpNwNa$j?5UgXxQPXm34uBaijS$D#JtFOJYF<68%P<|o7nd|0P-;FCUEr*((tBlne_
z=l6M*-Qt5@S-16Bw{?+2KkL>{zUlqHdvX8HyZ4;V9o|1WpLp-C{_L&KdFYOHW~Sbg
zx<_{of#&soOI|J?m(R=Z<%8bKd9ND2&vcG5PdJXeA5VC}pM6zN&*&|EP){G(3FLo0
zr&s)x{jneHUVL*;ArF<WC>x(%$;*wao%!U~;-s>;=3GJ^`oX`;kN81xiJsDb&*>dK
z<<Hm|`RO-1#k=S1&c3LJ<}t5*QlBr+`n?79_S-nh*3JIyn>@pLRGA*&%{a!jF7xsS
z_Q^iDr*h6yAHH|CPxg&o*k|(62m9?gx#<Tqj_1nqXZ?+DKb7q(#Gf*~Ha;}JdFZ|Q
zq4}*_du8KjXWkIWftPR`ayd6U$K&0;D2L})@{^0cXb+7O?zi^PeH0$_Gmies{3IlY
z{>Fu#hjQaZKXO<v{~gMsj6Xcn7yDx$VAj2p`vdcm$N0+T#e;dYvrg^wgP!Blyvpt~
z=$X8ce`D|LT%2%EDsSXR`MGf2zUvcT@y`E|%lm3^*?oxbirklo!}QvBLcRwykNEx0
z<l9eK>B#Qw#?jt)M*7{l+9SW%?SSU|eT@F{g)lz*UQa!_<sIZCkGQVAxUHUFRPUZo
zzMu@{1IqFZ`Gk8v`3JOLA=+<wsW`>2izCYPhQ2`iLjU+pNT2Dk{Wq?D@+j@d$3L?N
z`pzEdl{jWxh%db2(LT7><ZrYO+i5Q!RHhg9KkU!X_zui-@v>ONC*H+js2<{j{W||T
zf7^HH{iOG4-m7@e>3ysBTHaInyI}QM=)ITc-hUa_`!Vl*y+8Iom>k}ZLhbbPUfKIv
z?`ORagl$9G23~mUrOp5AzgkynS+u2JOY^C11KL`;ee>VeKCKJ1F5J>|N?Yq9-;`Hr
zX>MuWc4WVyZLL?FzQr<6g>J8L%DeSC_PEEUypYx#{$ZX64j=K8SJFnsdgHUR<e9T+
z%F+-1<B7Dh@RlzQTj#~i+S6)d<5+x@G%^qwhzvvqA_I|ukbwi|US{0^Z@-z=3d8wh
zIWkaB4E%bz%Z{CP?AWwaJn5|VHGgsDTWQ3|!~d@b<?8l;gZJEUr4g^ESy=A4vERQK
z*yXh2H`r~<@#*Y_9o_#*^VLy#V*l@Fpi%D=quh;sAKYmB_x)BBpGF2E1CfF1Gf=MI
ztiEKk=M3fUN9-5#mqvb{KfCJh?CIqiZ`I}M=h{7feCl(<o=xiw8*c}q#F2q-87S8t
zW51XuHBz2bt@q@e^}S-8w>sYgf9o0J|3wBO1CfEqKx7~?5E+OJL<Z`QfolJ~P#nKD
zf0q?`jyxCh_kZ!;t+MxSo$b-%ewMbMoxNXOW&TFhyK3`a>F=a#HD0xOiQ^Z?K-}-*
zcq+D4wcgvraclecqml2(cP;*2xmdqaYP(A1OQp^m+eZd!!9euS(LdM1KX+8mYGDg;
zj@mQORqwyw*(W#brBkm-8XYg%VukhYfAfJP#A{Y<JMGctCz3|N8H=8|=?hywn}ir@
zd`Eu=u+jIM{Ecdt<^G*Os0XoJ9s`v+FU4_6dp}Yye3hp)u~%duG7uSv3`7PZ1LZKV
z!`@GCc=nO6rx7C$Q4hXr2j$pQIeq%i`N3*u?=UW@G?pVvF@0M8w#13+?)`ikG4k+z
zda&J2$wr^8_j($!K6og`wqu*fKt~xkY@KJ;Uh?({X;;UN@)Y);+;`ujH@W?`v|J3w
zD^%`w{KO-=J$*-7DjN5Leb!&`!mIzDRvWh%I_{*W28>CwkUWi!Ge+ID?61?u(^_Lk
zd0X-?XKcP&-?p@~u%q(F{{NSOJx8SLjo9Ihw9@#0XNjfAKyeK0yUQy_|K_Y0(?-Qg
z$H(GWi>)F9k%7oSWS|fO_3|A=sd5xr*X;g0{_b6Me)q0kd7a($J+4185E<w^25R;9
z-1Rn3=P}-~nM=*LI@-Qt(#GDAf!Z=~<o$o|{m4DnrHzUmwb#z>-_gH+jQtzq?{sTx
z)p54SKx7~?P<;lj=(GNPiw*fWDHY3=?1+|+*V=C5iD|8{TsaH%Tm8MoPrG{MXVXe!
zp?M!)@V{=KZ^Y+Gqv2iWkGXrHML$g%9Y?oqb=Hn|zM0f3uDaxt>yA3_g``ojQGRzI
zlr#MOfkrJau=uLSZ+_*Zq~5TQUyRQq1C78yweE#$wf2twj!Nu5W$6e1@kCl_9%{A!
zah}LPWFRsyyD@O{UOk@M`{xVgDvh%nk8!<`fyh9|8R+cyp7l0=#~F<yL<S-QwPN7d
z-(C5`#d^GzHYzUo^2*mtojN0_7QWl}om02n?X9HJ=yzaC#Y*2-lsjIzzXO}qkG|Ji
zJj;@Yyt50I`*#eD%Fjt>wXgY$Gv7)hHU<xks%KH|$UtOZ_GO@2?;*R&xU;Wj#Fa+|
znwlzY$C1;>Kx7~?5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgcs4N3*L)r$m
zuGG@6WwDm#Q`-i#wRHRDzpZ^*7ie9$rRkKm)<wQ4uhP=o(!A}+enZ<@uQ+{+WuE%3
z4C5Xgb<os_6VpN%_U!xWF>_5ga%@s5e16#zws`UV@ktiW-DmE#FIfNewBGR6YZrTV
zz8xP<8x@BQIs4|l{xm6VRQz)3ZT>l9^{Ht`v9tH@(T_h_eD$@TNUDvY9oM=%aMe?G
zzk1GTNu%PZHTGWl&>@#39mPVtS6dI_`2XJC3hhf*eO_;IHIxVMo$aTy<|kigVU}F=
zR^M6tbhUZQ-PfZ}?f$^^+kTLgiq+;XwD)q$S?7mp*H`Ks)LG+Ho1c0ezgWI+eQxs;
za+kfG)Egdq?|Bn;o_kDM@3`tgr~UDwqd!ec#X|QIm3khYAItT3h4Sih$JMW1?Wk0_
z%C)2AFMH$IIgg*5bT*bNf4Tkaf4Nw0zL3?gSUIZgmx`66P`g5QTdn7X<m#;FpKtfy
z_Z_^}7fDya$GQ!kI_9;nlSaqR_Iu^dU+Q_I*!I=3k8^*v|GJ}goRU@wi;a)fT2G;I
zv;2Ie>I=<Z@6RihXBIn7z4K9O9<O~z*LB{1J-o!GyMC6GiyfUmtAG0+um5bHYd%g(
z#jNp4tuJ-nLhUPM_Z=Ol+<TyE_uHrU<=eh;$5Uxn!D_$%E%u(JQthghug7u!{d==Y
z?OVCyl<F_FFXSh}_q_U*isk0bo!zg}=cQ74)4tq$lWO-PU%g_XIA84ZTKm0n^}f(P
zR7xKT&6D-K*!;iLc3JZksxNh(&T3z&bGUKJjWd<%S8IEkUFQqqccsg<x*u)S@8!td
z(Q{y}`n#v77FPRxNF4v)zYi~z=f~%@`M$o<^s`d=Vx`6@l+ShadA09>X4gC<Yu#PV
z?_{mBUg~R=M|5<aa^G*%YrkUO=hmuSNBv}}{VTmMh2}5UuT<)MTkUa1J=?p_;Bhn3
za<Q}Wkj}b4srPw1>Mx7!U+jHnvF)nGYu5b5en+BRXX$yNd5hhzYPSphXtn)TIDWPC
zr_shO_FgXBudH&d@`!59Q|vh`YyE}l8%;mRTWDP!eZJT(Q=b3DOHJ*y!tn17$}M+v
z-PP`2u0EIh9<ZzTtM-0{;}?p<Vf#vzD|K&HFXM#qzfsGTKF?Nryhis2-lH`Np0HJq
z12(#|TdvkvDZdkrQ>i@9xvW-LY+l{f+J*ZU%iqsHt@N?d^E7Jw%#yp({M2f^O8Lh^
z<78=Zq5QO?&&$0}pI!Q8?Mt!s9o?VUzq-FenB}h;tv<^ihI(2r=f%#JqtW)4yjlBP
zs6Nz#&MH^y`@>SlEqDF1OWcd={P*7tma8{irC+0;rwirbm406AzVq)9|MmOZ)y@x!
z-Iqe`vc6|1w%+}4xbMC*2;*8fZunj;d><A1apwu+XJI&wvcEHep*+<po4?i=^4rn%
zh5K3Ua=8DcmTS#-)_kS%)abuoP^<U!)tcwTwO%;<gxkBfR~oAw7krP=C@3E*6sIcn
zyjtJ8S4uwP(CdFZH2Au~KkL?BE{6WU+;Uc&&8n}K{)zw9#xS3$bh%VL%;H&n$cOq$
zVX5(}Qtd0vd#%P}#~sB&_hR+_T%O%oSS>$P?QuJrr+0S$O7C|^?I`Tu(ftd@D^xCb
zKb$ABe$SxN`D7t~QfQs=`S&x>bGx}tZhm_1_HxnRRml%KilxTstksp8&vf<nSv-aH
zmCDb;al&~@Ef?|w@%i`vyNI>or__Bd)IQXs*}YsSz81QVER<)In-}PpwV#FVt=zkW
z*y!J5*NgliuZ>bJq&FRXE`O;tb~H~4`+M(IDP+gBx}T_(K2~~Nh5Tu?o?A!6$UtPE
zcm~3I>3T00zp{U|td$?_ta(c1qfq-wf1i{!ULiROJ@2f)=iA`UA)5?Xrbn);<A7<8
z+;GL9Ub%9y()Y{7j#sPvENdQmTj<<bZ_jIe--oAq#jO3!s;~As>gC*At{ho$pi$~;
zZ6B4MuhH)b>pd?mBwwZfJ&;<DS1bLgwjYh-M+Ukw1KvB7id~tLdRSYz`F^>w`!1zc
zSgxG)+OOL0hijGhR%;zu=b%c}muoMv-@pI6(vX)z`D)!he{S;=a+kfG6ngJp@6T&x
z|E11TYP=}--n&}w0pqx%4w^b~Vp=aeR4e|}YTWQ1CM#bl*ABA!6?@-Z>2{Th-(6*#
z&>v>)e^z}Mce{GIR`*14o`3)EwS~N7>1Aitcb5Kl)_O~wx7KmHSovxdZ$tg9RJqu7
zcC=j>FB-L6>iuD<?JKpfS>t5oRgG3(x99L{zyEg=!@RlC%AFObsx@z=;!3gpG3&db
zcbAzs*MVovkt-K#6^AM{Pqpr$I%-eF?xS{d?6<*@!=ITeR|vIlG(THwc`Kz~g~qWT
z^@hd%T}G+*m-W)VqxRX^{VRQcQ)~Up$_tCFFIC=3<!O~VFO({Oy|u5j{)OY!+TVB8
z`D*2FtF51vvfoONQ>;IzRDK_hlf`?XyrZi>&*G)n`eOGrY**?%i}uCJ728#ofoh%q
zYBg@P?jwv_S+;9GTDia1SZY6Z+|{wz-yJCT_crUTU!nK+#qyzDmIZY7d9tg{pT&2*
z*4Ns9)GH5_lB?MJgkswjD_5AO*K4`f|Bh+7^VRAcTkCmhl_$n|y2f|Ov->_4>wm*`
z<=%JH%e`Q@o<=E)ds*^^zYA7uIlOl%RL+`TeX;Mm>b)JltA)NBxb*CuUp(c@p1CZ1
zq2HHxEpuX@9MrF?U{^gaHOjg=+Yb4S5VL+~ytC`ejelXkYWbC7-^&zxPSegkW~r!u
z7G~uIoozp%KE`ro8K_nMU0JD{nzHn`QR++mo&B(V+mN<_7v6el^Z)YNx>CzxEsM7_
zpV~H{t)<&H|84Elx<Ko~ElsDiwJ!2ad6kytmga3o_8Z#Pdd2BmEb~;Qw<C@h8Hfx-
z1|kEIfyh8)p!y7)GxN!F`t0*qT5nkWmB$$(1CfEqKxCkj44i(>Umxu8_47%@$ix4y
z2bI*ZI96mJG7uSv3`7PZ1C?c9nR%}GQ@1~kNxKSGmYq0S$iNBf@AA~{C%=`3_}c|e
zTkPxACZzR(q1>??8Hfx-1|kEIfyhAd41D%Tx3zxy#xrT9u=tr{%g8`vATm(j4BWf$
zXN#Oa>DaVV`0E!YpY`Jjw<ZxI4^a=I9W<IB7<Bktw=Z?+xHMwt@bFsGRqapCcq5G%
zd5C)OT|21nY%)qWyE8D)R-3JI#w~BB5hD-vtOs4)PLIC$g?*3x?A0`4<RR)ow1eme
z{=FaQ>Tw{hFES7rn0**%^zR_Uom%&SZI9jM5054h|1S?APq7>shzvvqA_I|u$UxU)
zV4>YVzGc`|ccz8#uluaF$2~v&TT&~m)OVF}oXUQu+Ew}NdX^nm6d8yN)Efig-v^dj
zu9v?fjB=Frcb)a7hf$u$Kx7~?5E+OJ%+3tlwBf|9yS?*rI`z0C4j%a2Z3!GR<HaBK
zyyU}V$?JFdUB3t3$>V=_U-i$UTP7!Pi3jg`;@!7CNZ{RDEjIm^i@un^(Hs6~=DyD!
zl!iF?S2z8#|6wnrS-AespS*gVg<eTRJm8ewKK$dK#;3*N1y}v-!QLy3OB)rR>pymf
zUv2qBn#WbH-hThnUQeHkSW_NK^${Np{kV7k$6iSr4NqG9i^I>@eS8`*@=%H%lxiPc
zz5TJr9Xan`N4=PixS;#vzZ>~_#2WLk-A3cC{b;c<X}#b%gC8A!%8$pV#p0aj?YZB$
zFJ4L`b`B3+-9N?kMFt`R-!d@br~6NNa>0M3Sy=A(pWn_9|F0`B(9z#7=<L52=?Z%g
z*AN+q3`7RHG6RRsd)Q;?4_{A;#pz4!zT8K<k539=a@xqvdJlRhi5PkKMh_b8ccdyM
z@8vu8S^L=6CM1<YzUYuk4ta3D6Ru0U3R+jOSa1J+AbE<FKemeuL<YWPp#OkPXQW%c
zlr{=}D|!6C$UtNuGEm<PbX#NMl%BmWPhgnehQA*@=&e<bU$y%KX{p%z-DlpoW|Prr
zxwzNxwO1Z+`@?CeDE}%I>wDLtbdiC`Kx7~?5E+OJL<S-Qk%5jeu+vAw2VeH=__S2)
z*sQU4WFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtPEYcddjzpb?2Z;O0IzP?^<
zLRv@<x~8{?tBMRn2I`T4a=&Mb{W|tNV5Rw^TaUf+>^uiw_j+3INZzi7)!Nra<Ehm5
z@UeY)-!D{(w?@;it}gFIyRC52@R#37XIJd%`#Za>uhhR^i0z*|r_Vl*r5)vAcGI)C
z-pD{?ATkgchz!gQ4E$>AAMCzW%d<(TxZzp{ZL{~J7t)B4ho}e94r=8GW(S`f*BBXy
z4AeIR^K7-*DremCb{cW^<e|Q`vwG5%dQK?RzN7b(vH$<Q=dIq($B`ogjmtpWkhX!X
zE43`%vP4Vssci$=TDpDn-_|~@3$!lW(sW8&>muKjS7~W(Y2J2ZzoBidSDe1ZGEdd-
zdY5jRHf`{c6VgV*4Ih2ui4T{4Ic-$jbJ1&(r7wRXZB%Sj-aAX~dh-DPJ?Uq^ns3a9
zX`^7J_O0IbA<8p*GjQ2$>#x05kMU`}qTShP<bGd1@aiXNu~;d4%^C+^<m)WFZN&1c
zp7599X}x2$_^}W5g4J4Q9Jf9hSZ3_)^Y%G*avHI*c-Zm%VH5klG%;;dtQUV*NRE3K
zy8gnI+Fnf?1?$txG&SwD@*c0BwC-E!?21e6x%q$nzWZBAXXELIzrXJWD}0#N3hPxL
z3&~aOe6-N~@p)t*G7uSv417NWM}9qh`-2aDC#e>G|LpN;WFRsS8Hfx-1|kEIfyh8)
zATkgchzvvqA_I|u$UtNuG7uSv3`7PZ1Jz`p*zYQ1yVAZxjeJGEO8IU%>PvO?rJ6o8
zjvX0@3`7PZ1CfEqKx7~?5E+OJL<S-Qk%7oS=P<C~WpgZe%?gv#M#Xp^RNQ?~=V({`
z&N^|~f#c6!eq1uUVg2q%eM;7-zyB2FZtUOhDs}(r(|V#*k%7oSWZ+u{8s+!wx~iPz
zp7Xxl!T5jwj)C8exxM#ex6McjVRFfw4_vqOr%9>!=Z!{Q{FgI6Pm0A!8}2dR1*gnN
zip6*TGWy8nF8d-W7XLEzZ(sGkZ)Vb{xadYx)_(41UnjN3f2TXMQ{8^Q=rioSRX$4^
z73bWm{mL)r_&li<c9w25%I{HS$$3GK-Bx>g=BTt<*jf7$=Zy?R21;k(N2~6>{_z_=
zmllhKezz|^ukYVQj{8&E{*-P}<@H$ZvXjnxey>SsrLjC|Vz0<RWFRsS8K@Qm9}oR7
zciv`SC5?txulMRli*ERXT%%&!v19hwdi%L^mBLEvM@N7Ewp4kt{+)_K^`*|!(e_vM
zf8deJ-+w#J!j6s~`*)0iiwA9b_#q2ClXeznNk3`9pU!_z-#62U|KcHwr&u2uhzvvq
zA_HBUfeR-+{>n=if0i~Hj#*%hho=mfl1BV~9xDCsi*#*kXlk0&w8sh0fBst9sMyu_
zC$29t5E+OJL<S-Qk%9VWpwjQJt36)1-+h<cFKY+u<=+$C|MrXSy8Og(X|b5Ko=Vl1
z%X7W-YxKXDX!O4$U%F}9w82MCNE;Qa?f<GhexveE?po=8rTk@A8)wo|Pd#|qQh!f6
z3t!prwex=S^9e~7F8lT4z2BSkeiE^BdAQ&gH$Jn*dy|t=v8&ljy{(5IEfvdM$MALg
zz3{<^N0M4$xpKySk%2~JpwZqNhVs_Sxo~!sqh8O$%ihp?i~C!jPKw2w-hX4YOZFI_
z)*J4%!NMoL_2@s7M#nLm9CmB}{_m!ZipOn|9Pq@b57I`(zn*?T``kZ$Cxzw4s)K&>
z;-o`oe3%rA<<1lP)fWSI4qNJ+-)%Z2X;jQQ?`PGIKYW>+|90KGX}w|A_ww~#zvT9(
zjp=*HOKG`SU)IpnwAdXplk10$PHT<z#zU4o;!LeDOODQ}@3TVN(evFrF{w1xYCT&X
zx7pj<@7$bgbewShhcEtS-EO&voy$XyVRv``?FKWGM#H23JoK$scKIl2RGjkHjfZY9
zZ9>vjaLAsA-#6ow@kvLqR(>>Vo{s(=_ZxqBv1O@E$0V~mX7O98`jJC-TB`ps)00NU
zQ@2=U;J7d5%2f-i{k?(lE481UHBP<WlNOSz)c1V#*1poek5p^@>8Sj*lJ|s%_P^$#
z7e7lXjkS`$TJu!<z0k>fezMWJ4}FzH>=+N?OtrCEdKSkmkAYHupSoJ@J1f5|kG0rq
zc4MHc+#ghH9kZK0l)K(~J0Da^o<hH`)zzO@I{z$Jzhb}0KxCkF1|Hl0jIF-Bt#__c
zSh@uD-J{pat6VbUmiN+#k%y=U(GKd}50vXor*8Am;@7<~DQQ$Jcio-cZ}7>dfB2(q
ze~`<PW6O=s+p%x2X-Ty)YyMat87PN=FZ+Ko_wt9lna=L`yVG}C=EAK$PG>hPM`!A(
z&s|3>)pXCzFQlbnwboGV@3_Qv-~YQe)!LspZe*Y{8R+QuB5gz323~mUrOp4lhSrr@
zmS|bBrTNsh0c|bazWHx!pVkFh7j9`frLA?5Z_2B*G`BQwJF?%<w$>|7-(s1knwn1D
za>~rV4H}g^y6cP2f41Vg>1B&PcJk<BC+6|BjemIRRn4!bA+Gbt=wBYW(8y$i<u={m
z$)~SP?jHTeVGsY{x)iS2df>4u{POk`w)Gp=XP2v{q`i(xUV7-pH`C*rE<EPkAC69@
z-T&C(pRV+3veM<B&wI&zZzP`IkX-llt<OK7A9wd26aVMWBmSQ3{>9!W%rWh~^vW}O
zwyx3Vla$=%(a!k8-x_w@Qult44teqJwQnCYJ};knFPQx9hErA;nf}M!t1SH5++*_i
z(VyO1weM5!r^bWE?R(!@|2$yCv}E29SN{6L*Cr)@TCe54Wq&m;HQog?4jgpG+pp)>
zcl=xDEWh84!&B=z=<|(Ro9-B!Z~yh23!n7<rtjp}g9mc|ZSAklSmYmnPkRl0sOPQ+
zwkOJ->u(+8Sp0}Z9=d<vKl9__i#(y+)~Vk=kGp5NXODX+`QYfIkK1OUFB0nt(SAU@
z+pn$uH1_(%?>{S{U*<WePwv50Ho85bU+T5DK5}UH<{|UEy3ygI@_O_?C(XIzxOtvV
ztlv2FQ8^3o^&OnGkKulqU%hpa^ZwmF7<k0br=^}7SNWW?et5z<yL^^dzwzulyCi?L
zu%piJ&P!j|bl0BEIreKk^W3uJI=e6VbV|RhkN(-$Tl!vkz*8HHPWh3sJKy%eN3TAg
zw?pmMxp(_DC+zTc+I-<dKls6^FQwY?L*}vn^w)mVCuRD=o|Mo1^$tt-`0BOf$R!rN
z^OfH3<?VZm1IP7x>+|=LeYah5&H0CYoNROW@w+ax{N$w9PW^WJ*%ecg18)A{S94ze
zVZOa_wtJ=L-Xr#TD{nXa4m3agvEI>-bzf`LA19<Yo&L$oH%@vzwH|sDqV;KKo%AML
z&wZO6chAk|PDmg*_&12xEYz+b&hmf!PKe=p3YGCxDu(ms%d_H9xQ@KPp5>=J$EPyA
z$NN>kT6NBucYKhKYxDr$+C%n3FN|wn_zm?_-@o?xWls7e@ti-Q$Mh7QHTJhhzj9Y|
zjvV~A_T*NE_FZ|oHSSq;i`Awi#^LY8LG@6-aDB@9!G#{#@_&|?_mjl9_#p2FU)^?V
zpVK}{R@`c(G4oECl!|-&!+F<Vw)*rw@1^rhI`)CZH~lz|&u_BDf-9^zIn_>CJmH7<
z-~K~4TxG4ZKFZr4JH$8n$!ERvqN9FwcIp3<RgXU9Z|NI(|F&Q2Sz}+`@2x!V^bhij
z_Dz5Bgdeq!_KlyQAKDq$c<PO}+H<1^t-IE%`8cjUeG%6R`Kx00bFIt9pSbA7lamnH
zM;6k*EX@143_mQsg~*SEDBfh*0l&s>`7L>&_)RYD@hxv0_T-Lxj#>8IL_Wm6$YcDE
zKNvpnnrq)o_UOOP@C*JqK96`e&wT^`d$U#kGC3W3d+P;j|KXGLfR~;avEI~A(vvqh
zZ1hT3f0&Zbc?JGv|0@Um=FBhBus=Uyob^{b_XnH3`DuO}endOZ=_P$rwtoIz8Q<2=
z@5n>=Z@l2!{-_r>j6=`FCGn1*^_)L;4ltg0FyWYU=l;ir@8t6)<Jlj2tG)5*t2lMf
zsATkj8DAtrZhG<QKm2}r^4dbZnpgbdlZ4#X9gagDanSk9xcc9;`@%;bbKchpG!G;v
zx$%lm@y33rm+z6oyz~-}P(3_h#rwL=v+>7?{`i6R3(qg~-uLz?kH4R^9PrpN`#d`}
zk0&nPW0haM^JRX0&tCJ&4gF95B1xuQw!+ziW+a~5N9)~f+$Ya%oclaE?9TIUT7T4M
ziSp}n@37L4>%L0x2|b@{n}J7N*nejJx&1RQp3O^NtrLHf_L^t(al8LBk<at5{5XFs
zzn53g2Yw4b^!}8yx0vV7qbBBgm0wu*$RGc9^;Hl2H0ihIsJljP`}TM4Gx!bm$uGGd
z;5VFiLVM=t#0mL0WN&!j=Qexohethl)7WI8E0<j9%5Gzl^z4EA9dO$4yk8r<RPM<)
z28>JSf%6l)gVt$$`AaxHJ;#Ije!j=1r%xSucK+Oj5B|wM*$4CR7vx}n;w(Pc5&zF0
zkW*YWzWs&#fPFQubr{e3@T)(0Avy4CJaXX0c=U@tT9@(o5%RES`e`2LTK(l2c&1<M
zKwN|5%g;OOUW@$nOgyzO%KQ($$f=&+QuZ8@*E-}|?A3GjLl4-Iyxjef{_F(b&bQjr
zAN~j*A(EpIlCKcryVA&i6vCND?Kt?S|M)B|gzSjk;Cu1{ufMbOtskfI3;SdL@EY2?
z_VVajp>@imAfD_id7*WM`a_QjAw7%uulu)J@xb4Q$j@eBm@m+Kevp5qAMAl2lb0Ar
zdv-1l(~dtRr+h*l^yM8x2MyV{M@~Ly9?zkA@dEPO<dx@mE{^Cg4mt<%=lp<m@T=_H
zbMd1PY9Asw*%kRAd*%PxtvJVC$qDU;{VW#U6WK?8-TeF+y&jeB_4=iQXXNjR@IX#H
zTc3X9w?6I3Z+_)0q}TexFrK;Ja_+Pq`W*5ju0nFcu)VV9AzHt6L;D2rlZEbY=#6|A
z@6H8bzAE3P$L5#6(Mx4~%PXu0zwVvL!G7qYJh%{&n_ZTQ<b(FXy_bErkM1#UnS9j5
z9oCqVitG3`j{Tx1c)=4ML$r?}+Bf<~PW#{-uAUsuOY{_K7tW&`+O77+-Q}9m1JC}V
zXO8^r&AOd~?01NZZ@tOh=gm1i|DM~v8{fG4u~XyH3%trVl%Y68&*+QswX;u<zT$x%
z;FDbX+gEai^WX=HC-{Q=w)L3DI*ji;B`(r~taXWpklgGP|HjGlHS7M;yw>SCIgMi-
z<i;btGk&ONAurzVh2xr^Joqx6{UZlCt=GPgPk+yiubpvV{`o9k*gg4;i%;=h*>n43
zJoWV7I`mVfkLJZE`!ufho*Umf=pTN_>A7*7L+PJ&YHwWe(tf$uC4VTddKj)td-K>|
z^VlEXWsr}bRPVk-fAZtq{(Ejd?YwW_|MsyHo;#*J^_+ahhx*|G+J_y!Jba$hmi#(@
zU(KJe4}2Jhec>nEFL^M30Q3EP$1@A*x%&v~7w_m1dEKYc3-==A$ot{o`x?*bq4oG~
z2+AwSWnZm}o<h&rzx>=i>U(cqe(B<)7tc+&>aO#CzQyOM_X#1|NAuz#T)%y`FL<?Y
z@?B-fF34#fyzq<0b+1Mb$w4pVANGNtfb@^PsW-m$LhU^Fo+L!_;(eQSZ=Cy{(|hOG
zk9~sd)41|6JdmG$Ss#BxUyN^F`U=exqW<n3Lv)W(2=nXCcz<XAp!L#o_m=YMu$<S2
zS>HieuYI>a_DelJ=$HBErFQr=j&>n>zlJyZXP+K_e3ARMz321%cXN0#KKZn_E^>IT
zURitc85a-q!8*yMtUtM}pInd~d}p3n-$fcP<i&h2ymzzTSs1qGN1$<wMeBst3;Azo
z-Ozp%!u&qW;wO~P{NXsxa}eJU&yc?;7x_beJJcKgM!j{}XS~@zdda?w%ikE!eUy71
z=MnE)p!a>s?x~EYo*d$~dVW)0BX5UzW(VZAUUC}8xb9Pwtpk$7y0n)ElHa+*djP1P
z=j62xaa_A_Ui!lR?U#Lmq5jd!5c!>m|N8DX@(_9W=KJrc2j%Gjf6Kqa(C>tKbfL2R
z9g0^_-VmbaA(AT#!?+%n`Qbt+UWUl87eaA1>@SXGVc0$_i#J&)o`p!BLiQoP7K`FD
z4EZipmM1{@t?z#9v+t9{UGJyFQ|->$vCn*`%-b_(z3TB{9(jViKt3Q3AcwN{_D$Xt
zBHr|u7noOGDBiO><62K4^c|sldwHXJ=TG+o>{xw>?AU#97Mh3rAr?A+`g<1l{UJJc
zhR8pJ^Mz&m4&~+0e#6kdl;vk37P9;B+^%e&Lo8&E^t@E0PgzJWLZqMW<(!kf-}GIF
z{h?R*rMJGn3-?EU>wd|3()}jCY(DMCslD@``&jklqTiuC&?EKqUVGzy(reZE4%u^R
z{_g|Luf6jJJ29T`I<3=uzWcK-_k{KbPx@JheZVv1PqZUPSmuYw5uUHuf%QZ8C-jvZ
z##hFd`OU)*Xz%_|Jka0!J!SqyJN@_x_9)&MpWe_{>m(m}ZhiZqxvzS8M*g1KbMy00
zz7sOPcF=SE$ft~VWr!zoS_iwLU&c2sq-WcATl4LW4*oO^+nZNCeKDSSjbnZEZk6GK
zh7Ugc%apw4wXfu75BRVS`Ydnb4?LH**e~%@KkfNh&-rDj-u;7q;kf*gdU9xQ-OAeG
zg<Sl!=hjD0J*S`ieimw<h3vapXdUHZmfY&ak#bS|&O-a2h3ueMWH0QBp8FnRu>q4e
z-F&ytQuf16*uhbo|74Th`+kwCpLhJDZ~peZFH`=MU-8`gGWnbN*`;~Kp%>OZbECta
zo0|VF!{395o95y7p!ZzPU+TqG_Un5O=sOwvN6*R0pW=z!S3kVsOJjF!YF8Hb_~~0$
z`}3|dhV*EM;)`+UJHA8x<UgRlM}X`d%Et>KJLO-<i!c6}eTPWzvoKsoSjIyZiqlz`
zwLW~7jkA6yfgX_$KjHrJr}k4G5$X#)lfThJc^rNA-2K1$a6J2D9Px|&=l6HkdoO-n
zJb>ar$d~*~9AppTD}7)e^iQ0?o4n8WY38GM`ayQ9zww}Uz6<a^PW+UoS`U92&TD==
z6+-Lw-W<>LI@~Y&L2vl2EIG*I+)cmTzdM(|`p<Rl{QIxx&GA#t+w25y5D)GJpnl?w
zdwk<~?(dMS6Y58=$ot@F*R3>WR8x+;?jig=l)o3_&%7`9_aMfF_Af+#zhmBIKfUOl
z3&+oyGj7ktr;Y#fF1>OP572zjzCeFJhnM5tzvR3<KW@%Jyg=)~$AC?L@ym&KH@AoR
zJ%3@J?Ys9b(E8oeus?dv?yTQ<{4pMUCmq_mam`OI>xKUQhW+si(7NTN^wPLcduTrE
zeD4<vzTJO`ru_cdZ}(s9-n!_i`RJK_bIziN#)J0BI`Iho-5n$+q$gR(P9QlTULb#w
zh4>1QALgI%w9i5_{<hD)-E#Jg9(c}w;W0$dt&jfkXXeEh{c`S7FMe2Oi1gULnn#(x
z2vJ;gZXvJpo4ixLsa=Tj*+MA4%R>1p)J|Ste?qk1c!KUtLJa#?svPc%vi!n2<RM``
zBOihIl;2pF{iR37$3vkw96qOaP~0Lvf8!hj{k;PJ!A_uj1+oL<^B-Br4@33<$qUWv
zIln=#m95M7jQkmUVaMwE3w~6+b;vKoXK}$e{2l+w&$It5G*2PS>Q5i=1L+~;PhmKo
z{fFWt4CM>^hh=$H7K+b>(BH3TAw39@9%f;dTwy&sqMr~iFiXGbMIofOkbc2%ePNk?
zmWuwqtPmEGKYT7<$U^qPuR!gsJ4AVu^94UiZ`lJsO8-4)=j_^kYfoRaXD{-k&<~5_
z;-z>iu0r?A;yAgC$IlzjIf;FcQ`{tv^_rL5?2(@EhxFGykM+_!`ygJD56{NckG$5e
zo%4hC5YO5{^?0yP{F!yh%gCpm{LpvW^wRuzV6Vn07R8S&bdIqfS!h1|Lf>)YL4IVP
zjql!X%k#UPJFw}?)P11)MEZ913EQ20^aIoK&)qLx*k{G{uibuTD!%LY`wdUs^TtPJ
zrcM8O>6vZEOi$fkLha}wv>y6xU#&+zB(HD}N*;FNeotOxeEs|#6<(C_65cDh|2XiO
z1>ZRNhObh*$+z%I9z427C6D=xZ$9hwevcl9_AS2-QGAIQ8Hfy2jsbB`yc0*n^AIaH
za~!D=7zpz!_XqN<@H{EsQ7=A=AL5_9#W>EZ%Jfj4k}uEt9boM}hw91UJZ2s8e)*d+
zJCN^qu0E94Iw1M<m-l#o0LdZmvtIIt_qFn}5bH&E?6W)uPvX104lnjce(8LHZ+^&g
zsGWNR=-i{7b0%JmXP@vWKc*+-H&537XwMFmwb$Qz_)mXt=G=xi{m4gd@@Yptyh3(G
zf5@Y3T=m-956{Udzcw$_E|kN5%j=9!KKY-1p6eIZYwx_|IXU!G#*=f4dh<i_;{}HH
zNRJ@ijB7sQ6^rKKPw20|1Ei<kSDMFrKIb?4<GnFGgyf}H%I+Qgy)rZpKaFpQSNq|d
zf)D+SV}G1u$f2G;;FrzgIX{UH_Mp9eqrb+7{0qMMVW^*Z?6b0Yt;cxKx*_=?IjxgE
zD?|4H_@k$xeEbD{As-Cm7dz$``627HF7uI>Ug4XZ+L70Nl6rA3v;*^)*Lc2zux@(o
zJV|eybM>Qt_SyW-yUy>*_Q$?EKYQ-nY#gY5RJRS+oHnz2yL$uklPg5=n7ujgiZ}Gu
zbM?yh6WVw8<nA54r>5WTx1Jty_z!kGzbWT^y!VWM9=Y-0Ume|?^ZTH_^ZDwH55GEO
zov)G=mTTJX{1;~=mwtNH^j!}7GXMMd>;Z~v=5^0(Kj;lRQMM1(Wj`SOGp@4!c+;Q%
zR3;amtY03$4!o~`cymwc?<4T${ib~|uknn}zU+61^rl=ip8otkz41OB59IV5pX||n
zj_<KN$6tu{0UC!t@I6i#e|$&dx%}rl`Z4SG{P_ib2$~P#8<L+NG7o*SPv@R=#I+~P
zF+H)b*0atZSG#_{t)_kFcfP&9q7U?=SoC|I>{Xfm*)M*;dZ7J-<P~S!n~9tF!2`dG
zA9ko-Kl5v6K4@NL@g<x`o?w5)17+ir!#MPn9Q+ME^xU}!>TjQ*dMM9PE`;Q@9_z-h
zby*MpBu<In^aS79;m!Esse4rUf_d?x?C(PK&yVk1JnMHO@R)`CHa^&k?_KZ?wWGh}
z;)jjL-mI5C(_8h{!w=inEY#ok7|t7b^W6I&`$W(AJ9!7xPM*!bD)Vdn5j3y7g`bA*
z!F>;{4CN2}y1YaFqn>>HDgCkU{2e{l&U$@6A-^%M??Tm^hyIBZ_;7CXU6sER^gHD2
zPoBz-=!baEKIALx$oTX^|3YZq5Y3x~;XcR<*)e&b_a62Mf8O)hcl#>e43YleiQeL$
zUh?nk#lEvMd9!)7=U?&6PT3*!UfKD<?<_+7LX_u*D6WRcugF8$_p%57c$>{Oo+k%G
zKNapLejq;p@oN9*ulJk$7d^zQ@ASzDt(!i|`zCF%$2^yh>Xp+!3y*(e!d=6zm^=5%
z%kP|X<ggw&<wK7?bfeQx>Y0P=Ckt;_c1o{jwwf=Og;Q^Q`t*n2nI{*b{R#Dhe^Ku}
z9Y2LX=({X-pzM90ap*b!VP8Dgp8qi)yK}FgpLyuDdh&X1-<7RTdu92m_@G`t?a9xt
zE3<q3jORJNtVg~0!>;)=dV^Q(wKI-+*fo8G{78uOT^_=3$mh(%KN=63*Sh3C`jZ>7
z8)!Z#US**;0NFX@@1T9pLU9io7m9mOJ{}@{F^~1q2k+JSar!}T@d(Wa?St|0Ods?U
zSJdlIZ+&-Qz4S-D_!Y)0_Guh(jh&ef8kc{i-+01<?+NHRJu<#^vOjt#4j32Tp0fk<
z%g>x2=^MR~r)j5N{%9U~j<_jqn9sV@n<vb-#A$q($2#q&@vINOc%l!5kiFuE{9!%2
zC12==#5?lH6O03`+kD1z9<i_N6o2M34m%{T{`Lny>_a{KG`{`w9DnM`Yn*UC^ODp4
zs5dUXh2kqY!hZS}i^ikZ^o4!;9>zTU485V}<iRuh!-MtP4|c;ZyHA3~vCn>&)4#K%
zo%=xc>l`TlIOp9s=*gYe9RF$l-|v(U`nwnRZN|}GJfk=AIOv|n{hz!Rn#cK7y*!wF
z@;vchJNgLm4C#G{_=TSH-yzb25Un3Cr6Qh6#ZW$WCC;!b`bvL&H>sW-$@9b!e1>Rz
zdJ6S}_zaOf7ean9MEahE)?F%w^ORe*4_R1f-G!dpuPl_u%Ny)7za&3`;+T6C_CS6-
z(gS62TbW*oQ;^^1pZP`lC>|S6yodCh-I<R(<W|;?f3%PMfp+{Vze}IAXODMXu*}}q
z-8EPKck_)apQiurEzCo1`(R%E@q%aZh#cnE&phO&C;9r|cW*b{{`_h0F4R2-v)ZYb
zXB0x~3sGKK2z{UH`(637eErl(r}lW|$)@%VmKnRjSr>er|J?!YoKO8dzVC$bFD@(7
zbLU`aJmnC{^VUT(?z(vL^we{CIy6oe+JDGC*#SABdiHAnq5XpN35NcbK12CNu_&*A
z@=qxKL3+uLD3gnxyBB4D@>b8~NBW2QhEMygKc38|pY`Jxf999JvM>F~5h6Q<;-Wl|
zUt{O|A3tY&>mV1u>7JOM<oEb%>!BwQAI_EZ3NP|Qc_BY#T=S@x-||!BvrpQ2Ze94&
zj(+fK-j}#1w-4cS|Nfu$&_3y}o*md<`lF0D_2MYGJy*64{jCR*S9|`*e$od#X^%&|
zklQ%MSI^$8TR-*mkRF>ySwD6HjUx{?pL`s;$MXJ9UN1j4zV_~ky~htxf8RYgACOCX
z>!QE*hdlH~J^A?;d>WU&h;!t2-|V|D{Q2DgJVNVsE|Cw&|M9B5dFT;;;XX=S5U=P<
zzW=QFu06f=U7GL6ytib>^c+8U;Q#HHd7yJ5{jx6g`sddb{4PFya!zDN<`2vK2YLBz
z>$G3w3;W}hKG=75WPIZq&-(1AeRKZw+&)_`zWGz{SJWHdcLvJ&e<x?wdtvW&oS!}S
ze%JiwwI26?{D}2x&kpR9_2b8Ucq5<jLw&Jto-2n)PWEG;%*)QiLF2I>{X(SA{DAy|
ze$X59ksmrISg(0Q#G`)VFh135XFlWNSv+?B5|_nO`GEb?PyTFO<_qsP&1WC&AAN-6
zhW3Y^7DD4^p>>kq`HS2T&v>H`<RO2!@B9v=&(J*(bWfFq^sErFcW6FH-Vm)9S}&w8
zS$u}|c*;Wd9-{bJ2;J}b&X6DEf5cz>vOoOUFMgB!;xm6sKjg#4(~tbd%lr9Reln}<
z_hUnOjYsdH{Dpk<L>bTSPuUqBtdHMQ_S`<xSN*lO4*rT<^hmvYTzn^&JlA;gd;R!D
zh=2J8f1|zkAkG=~*K=hU-t!orUOUgY2b9OKf9tg#d?{O(et0!Lzi1w1>oSl2%FsQM
zeIdX2!fwSU>!!ctr8m}NJpNEWXncNOJMRbVH{L@uZ#bTDL&QrKS`Xxh@d5P@=f@M?
zjYDpH>#uBFe$+l$m$G)y{KkXUtv|W6XP?R;8lU{e)sEcsk9^vj-*fZfQ-6BGFVSoK
z(VuXB>(O32JX<&YA&+yXexbbPW&g%8uJyBna2?`2|3of+$2#=a&bXc{+XwPnH`I<C
z<P}e>N4;~B{EfW!50V3a{2+hHKDE<N9w!gCPIm1)XFd8^r}3;$J4hb;W?!rm55}=>
zsGo7{v-Y0j$$f?A&XMLdA3Z0x^@L~~>k0cA*EsZ4{AWk{iKpTNdxFN(pS=9EJcAz5
zS7q{>k6hw1ls`azOg^Jte&qMe<R{M0+KGGmlT&~Bf_XeAk8>)1tV?@5;1fT_F+RD}
zo8P(C@65>;<(0l4v`+Qzchraf?uB;7lV`h+zz2WHpOVkL0`%`mk~c*893G(iAbNy9
z`e=XXGkcPEIoJ7}5BjdX^Pl$?>@Gz2#q3v}75X=R-E;nipW+|LVLj}bzAMvnWzV7g
zaGrD@ranaNLd0(tvg<6Q$63g3Lk#D0zek_w7i2dfhTnnG&n#r`S;%gS#lc$+9<b-H
zrzX(-S{8=-uYCRLe|__Wi|5Eezc1$=*1fHBAUiG<?UVa^|L!ZE*|GBsdt;aO$3Cz_
zdKvDkeb>(OuxuV^9&+)c{Gxh#&Mzv{S9WGS<~NUZS+{wNuU<Rz(=YQvc1VBl1jRFX
zOz02DYhLn-4|u_!cJxh|eBzO^_-0?&0l)9LxI;hjLJsRSu6S-fa^f-M&v;pA{8G_(
zNLfh#L$nWBXkFzZy(xsw=e|Sr{#xED50<xzD^T1K59yn@CXR_$<WX-N`KNirP5OyH
zeAsXO&8r_&PhaVeJkj_|oH_rn53cOiZXS8C{_+xb!2ejce(aDwlY?H#gUCT2=>h%M
zAHV!?*pEEy3cvW_cjQ;v<DcJRhwM~;e#Gx!liz-l*Wac2JLW%h|5@KlUs@<fpYW-l
ze~;RJI~S6locJ-1zi$$!#I3D=yw}A2Uh0-RYwT~2e&w#_+@N!>-nRGsJ#+FI=N|LZ
zPyWt(LV2M1`8WILp3yz&mv;;uG-Tr*Ip}?Qm?!%?Zhx0$9ojwAy=C|@%g>Q}=C==Q
zaCL63{O_sLJNqx*$Yc3C`YjLs@X(Rh{Q2RT$?nf?I_<QNrYCC-dZ^EV+jY<RyDoh4
z!}ggy7}q@&eQ}PaSM)?bJjoNB3;aDhzs#=mmv87Nzwvj${yx|D-1eKk&|l-pXRQlA
z_*1W)_aExzt@0Lp(I@)OUg#5^_yzfn_kZk7nZA+3{?S){N<M*S`Kx_UCYNzTyTXrq
zZtvUi&yUz=d4qZU?x4RHV%OHoj`%a<;Kg{_TOa>VFPwMy&k*&4`q>BVm9--mzvMl;
zc;_4!#>McvSnZ(uALAGo$`4Mx;kJ8@|It^e@7nO^e5k!~q5gPt&NDCh!uHl7ju@Zb
zhWBCgpZ}vL`~iPVk3-a+-uqsx*G~O*`q>pz63Fj}m-0D&N}Tb#8phS1p0nG{dh9%N
zv%5Y|!tdAl5qdAKTA#Sc&$4r9K7NhgArJc@uXULx3+<QvVNdiv%)8tR&|mkA^jtm?
zqW2sfMf$>j<5PU(AH-Ab{l0{}jXr1(_2b9P<9>ymxbNUE_!W5|#M42aZ`|5+$Jo?+
zfBwjGW&RhR{D$$#2etG539>)(8iyY6v)cKdlsx3`DDtz`Zyn?o$LJCNDIeo6JP-Tx
z^VX-m{`?@M5A25D@*E%T$M_Zg(s#D(#rJjU@yM_Hewg2c##1J@JcnHD#d{3vCBJy=
zIX#wN@-M#QVBevA^5^yo5A+*q=Q+EgzxI<H*Zky(15dc6NBe;bzW2lNn>V*R-&{E0
zs}sAu*P|V-dc<oNA27IEJO1zneLw2^=MDE6cJjNccWbAg+Z=xUt_v+cIsf}C<OubN
zUgY&}(y<RLzUjxQb$M<*(78(9!;i>Q_#tu2_c8Py;up$`ti!&DtMrUM@jK?X|MWo~
z>b}A`AdG|7!yl24Uo^gP=$C%vu>W|oe)<>2b?X&JjLXjCAI6h^@=NT){Px3slE?o0
zzKg$v{H*WY=n1)@I0)sl@^O4?Z$3P+SMwQ%zS(E_kp0)6zr&;a$9@@?z0pH_Lhn2L
zj*<PPU-%`TdF6+pfAyT-vmX1vU*LsYd#=3rf9!KibGv!0-#Gjs6nBlwZsaB8rpM$k
zF8=vH{oFG-N9ac%$*rB=jbn%61;6d@Kl~jDJ;sB$Os~nM9r-=S4}VP_`wsbc`+*Pp
zuB^ZDydMhJLm%V~P&^FF#>F4Gq4kIRsGohIkMb@Uu9F^V?;cA2?p(kw>5cmmdAzt<
zuV~%--`aiHmtV}E|9uegh=0R7`*#k|pMTJg|HI=W>)x^bk&Azw$Orj*{h@q?{fK++
zzvv&ot}GtdXL`q<@#FkDz2RT2i$1##asQ%Tng1oX_0tz}Ij=$bqU@eVKk=OZ<)7t~
z+SwP+>7R9ouf_}Gk9Pcn^IVpm+E@L|M=tRZPhot~jvvuZzYw*@t93&4_=ogBf9N^X
z-+1cz3;tGqtezjGPy8r<#Sh7|=xMlL<iZF0694205TC|1pZLJPif`fH@$GeW`uyC_
z&C%{0A-|O;>n{!&*Y`g7kRQt<v=eu%A3x%P{EFR$XkF}H959Y^nepf&y<r#h6S~L1
zW9WCB=jj2xW`F#cdVb${P(6L%x5y=bw%?w+$5qzey{+|WXFc+8_c-p6$!lH4s}$mi
zzS0Zlv@E^A3q7G%^c}zC^nQRJHNN?bryu<xm-{UGsD0R9y?N~;|7^Yd3jU$I0NTfL
zk$#5#-4|vd{{W5CQ8Zp5<j)JC{bC3BrRVkovSa(h{_VH^#<TD84*j70Wk2@KJtlh+
zmqY%H%RksB_DWwN{Z@w7i&y@T{_^kQ4&Lamb!!j#BlAM@K<&(HT)Z1kT)?;d8sGAG
z&&5&mD2un`v5(dV&95JSrayW419FnzxZ<n!ko^3VIH^CmmH8|4h&#s9jy(9{chuW=
z>%xP3DDpt_dX8u3RpW+89{tTn9{gHQi1yL?w5NaS$tnJmQ<?sfJIou5!w$@AJ>=Jp
zob&{e&%FA}v+=H7$b;wR(Vsl@&N|3pee}%y_?IVYhp$3tUC_AXGQKjjF8VAFv_AVo
z|DpHe-dlTr<hwuLH}H>zkbGGvKJdT%vG@Jzy+4rGK=#e=%5#ikzxZwXFMgX}IYfTF
zSR@B|L!^gUC@z(X)(Pb!P(3t1JGMS~CFE}*{f7J_d5w=p<Es~^@UFkIami!fjOV?W
zedIUIL%)p!<qe@6^oPCcZ(Qxwhkk=z>o4CBPa*#B#Lro`=j;?u@(X?*Iu}_t9^L2h
zSJuH#X>VQ1&i&S>Kfk4n2lBfAaIfGV)BZT8xqr7V<KV~rs&@2K+5L*|VDLgu^7212
z`j<y8G&12g^mE?Q&px=<;V=B%IlFNmDF2t|7zY{;@<Z<Z_yIiOfn4<4_;_`1Lk{gd
zhwc;Tm3?&I<=z3G^n$*~W9S9>%+F4YXP>RtIQCON^0<d{&I{jX;2&n)JJDBmjZf<%
z2mK<qb!&$&`bnS3iwEP-XYI^y{p9Aa=r2_7T%bL@h2gr{o%rt@;@;GF&MEvhKS`hP
zZC>wV=&5xm(|>vZ-5cT^FZi`T@(*(Jlh(uD_2-`;eR1ETo$<`WKPlT^{@8rRF+RP}
zo<C(5=1~v%8~%l!&<99Q*fl?)Y#+=+PU|F}=gQ{8r+wB>9%p{-vyflPLg!&fkCo}S
zJP<EgNUtFMg}w`fS>trHUOr22=&${eFR&l`Cl26~ALJMLKl)7{=#_Y6T=U2S*cbi4
zlljT3zwzl0WY6@#{N$6TYKISb1b)ru{Z1CP4QU(Lx>C!MElagDpV~H{t)<&H|84El
zx<Ko~ElsDiwJ!2ad6kytmga3o_8Z#Pdd2BmEb~<8?V6g#o;bR9ztvw&3gLHqR<`E)
z{#W)Hx!RPZ(Qxe#PaD3`8lNSNim#k}@nsjD@Ojdxcx=BLPW<G#DM_Pa&*A5-zTfko
zCDq2tT5ZQiTmJvCcjob4P5b{prwJJ%B^gUJNMx$6i+yCiACb&mM5ZG{hD?cwQzCIW
zAw?8rR+6Exjuc5Mng?l~htoXQ@7HyozK>S6J$&}spS?ck{ZG%<TJQB<@AX>49zJ{T
zQ@7j0Z%>)*lr(xNH@ijc=EJA-Q0a2xuyPMv*?oW#`qRN{T-?f!I&tbn8y{StY~|$L
zhIDeq&Q(bzy3bg#r2JW{zHt09-Sy*D|HgZsc-cAiD@A;n;?&LcKWlf#C@1MOrBBk<
zovD7}H|}*xyGiE}ci!Ch-Y*t8akInZ%ZDCt<3h()j&FSKj4vxJaBSt&#}7QAPsKS-
zP>v~m|I=glTH@HscHIl^tatrVC+W29`Z(-9chuXfj@sjp9~?J^^-H}+F8}n*79FNJ
zwsX$*`vxwlKhX)wq@UlIijVOR%VB#Cx%S8prl_dsH;y`k+Kukte7rJ6zj26NJPO!$
zpMmQ)TbO_I=}XIfJ3`sUPb%bG{bK9yRMa#zs~?%(LlxATt>UJs{&@K{J^#JhPkQ;u
z^_uT}#Bs$=TuizivxnZg^3f+&IzgFkztX*~Pq%%y`X_zZaMRU8<~c#><}Yae(0{#j
z-?~Rj7YUQ0!+DDzJ^Z%Ea*J%?^7dUzj~HL7C|hWY?!<ZUppO6PesrlKTiNl|r$_d0
zU#cim47y+XweLTs!<8K-<Q74{^>jP0-;r!94}1L)_Wa4~`J_yh>*JiX^u5<^tTM3r
zVr3g!&Ft|~tJ0HHP$o_Pq{riSpE_yzet++`#m=_-WTx(GoAw*^{U+U?>G<(~-n$KJ
zJox)q$5tk7KXHpE=<iWM?HNzo8MNP-wl8VVmnOBpT`u2v>!Mp{I<_!r>(tcw<e^`;
zYuC2oXm#{G@AsMg@El=GIz;t*y!eUs`3BK5T0Mha$D6h{wU4^_jrzW>TRY_6ad_=v
zKOVMR1?9A||2(8g%T+4r+`jqx%TFG(R@ugr?yB3Uc<+@eTUh3wC%$*s7b}%89v!Z{
zx!FCFuUn&p(dfW_nJGTr<6k-NeYRTJ%DDGmzvGm&_kn`O<M;g@zk29)du_PBebXt$
zj&0QY5p_Cg*ZY3siCTxK{e|sEo5LUY@YIPpMYiyQT8|&L;+hT0G@~9LFMnRo!FtsX
zTYgoI$ImU_VYQPj4EnreQteHDry^?q*)o4@=|`r%Clhv`@v8Smvp43yQ*N_Lmtmiq
z3R}<gbN#0M^SeY)dX1-T<)*)Hl&SS9?e6^jt=zV1(j((GIJPmaam>`Zb$eb!^8a|>
zJ>7Bn?VK*tYLA<#gvr#w^l?5e|8Bo4>hss@de^J`!PiFK*7L~u%68hi9$U2Bxcud<
zGgMIO^FY!&E8zD$JnZ#~U;o$z-8ZNw{Ns5qQ#$NCYf4@!=yMlIvoGp=;LO`cE>&?c
zQ!}JfmD&g7+%ng(m41I$32T?BamjSMVfD&%JzL)o(c`x5y85Z{moGW~sc}xG8T31Q
zUp{;H4IkE9<=D!&?mL-k=j&l_HCV963dc5fxbmS1(<ZHQY~j(zb!+zYpk+>6#}!lW
zJDG}qOy_ILtsn4<3d^9^*-5uIbzKy<->}!=ZuQ*mpGkhO?-jV|8#LeVe&99l1{cg2
z|K;l+IDQ#4PW`><Y)fbA`y=VLYwL9{ai`mVy}sY|<k#-43qS4FZREF3So&=TyRg?y
z)OY(X#o-O7)|_$abQKl7zIQ}C!ptb(H9n@x-FoddKQo<g>U`hJubWPiFA9hPqJStM
z3Wx&sDp33Q=5x#KxmHC*l!)tn)41MKre9H=mxjHLO1B>U5Q+k#fGB_hZ&nyHWNNKV
zPPUSMM#Z4d1%|axH{QJmHGHS$y(?7QjO+LxcRNw*=ym_aRC%WVz8~}2?{_GYVi))C
zkm#f{DS8Cm*Ytiwoi_D%0ORuO_xI7H+GYAadAjXq`uWOq+kfHx!izq-;aSI4CUsu>
zjVrF_Go#)&v&El$Us%4c?C1N^VV^_HmUWWUeuVw|O=0^J_xq<|<4kJ3py&T$|HRAA
zsb8s*q65#-*~+B;or{tZS5mKIqW*qdru^CRy*IBoZGFFtxNUjQ$t_O(y}+>ND_-vr
z=;iIY7u;Fz`lU|XwC#PmxZ@2v{;yx!`k(t8lT&0X_c-O=Hb*_4Q{<O-9s16dllpH~
zN#n>)ue51<^(JK-)BSE{(D`cGUVm>SuC(cXMdd&4`59H*QTZ9)_}m#^R#@Of&8Xt@
z@{{iGG)bu+_Pwj5)QjqQ#!UOM<?k}<ai;q`XQt~XWgSKpM^y9JYyUG{e(BtgS66Gl
z-U-WWeGV{azf6x)Qtxm1jVr44;^wEMpZEG52V3v2$2GpD+KKw_Oq<?sbpPo-&#3z|
zmA>wVioYx!l9LxTO~o6R-+wIK{las;ndO*bN%@`0bU#h)Ke}Jj^Um+_D5-X)p5t}X
zJ?`gAlGdJ^E~fLt>R>u9@&DHfB;|fs*f^p-57hnrwQ7>9D=@t5y0t^wZgsMaLGN2+
zn|;3v->5L`=ygt1)aMQ9YHN1iQsMq3N*JvU=xYl*Oul^R0XHsm;-W3OgpD)mzccGM
zj*~uYxasO4^PHqpf38W=X8(SSu5VPoABmrG4_w)OfD#?DTL-`UuWlFjbBIy5r}xK}
zwAbG!(4XHmmG;sr-S?l9Qs4CV$Gvp)TORg4SyJlB`@Vm?@0+Q=?{@FC8><YgzF1|7
ze&=~o?5=sc#`oXeP^{pzvj04!Ny}9#C{6Xp?{^N8W~ZOeXF6?ae8le`3OqP<V!b_n
z-Q<|&usw%dd*laGlrS0{3fmO6?AE_$?te!yzh>&N)ZwYzPKC`1Q>AwNn_nTnVt(aR
zPC;RQl^yBYsoYd<lVMdm7UuWr+_>84U8nzpGX|c$t<HGIi@Vm0Hs_0T8h2fA>MUg&
zpC8)m#{SQ)QnvDgXRBSZy5BNoD=VIUeVNPpFI2YkyMyvho3&rDGR<gLZTj9O)SGn8
zgR7T0w($H0Uo0s9*nB5ldMUT`(N;~LxNn$~F0Xj~oce2C7^j5(bnqG%xAOAK%eH^v
z>B-7gzPIhUqgtQ)ok}Xv{p_yg->!4VI>#@+Y1LzBpN7SbDb|=#rNY$B3uF7krRbou
z^x$u%9-6yJC7pWQQKi2+ZTQdiJN%}S!l?S?=BGxPcRp-9>mDaAy2X*S{B)0(Ve8wv
z?-Tcuc|~=PJZ6cql~=dg{!IHl)~cZVqM**fFVEcS*ve4@9_sXQ?=4Q!`QuAxRX_3A
zoIG0?c3*q{^wj0g+_%zkV_3h`OZxHMZ+!oFDfhR6ljf_SeCE(a%hZjdolNtDmV+DZ
z_x^e(E;9aM8TEOB-#AQPclwP(?BY@2{MYvR?)?ToD_dym`HpRI2i@oW+MAlyrtibL
z`8U-cFTdQ!Fh#HSlU_bz{)Sg>e`=N!7n80>*m;s}ztYt|-S*w;XG#y^Ny>e%sN?AT
z%@O<F)c+G@8%>}0M(v;bK-t2m^)c<Q#qu`?efGs1Wh*oN+y?t#)YQi#tp6<q4>Wl9
zfhA71FzkKu!y2?2wPE1~Cn&?VXS!b4`4Cq<+pZIdS6_F2({bsVE$zHgar2^k2Cj0F
z&Y=56((TiIekktxrk<ZL?KkTC-nzdh=fC$z^OdU|)4Y3d^Zgnhx7M+hanIMV?b-eu
zsl=Zh&jV-cIEKAXmZ^G48wXS8o%FNaOBZc;`KQ%N7>^EpmaiB);-ZyG7>^EquOr1S
z+WV1izjtOT-_$<p=J)*5KR*1HbJwY;=++MTexHxm?Mz>fMdd%!&wqt&-}H52-2U(C
zdS%f+%gl9b;fzO4m~%?i8IG;g??2d@PRe^m?3bD1-fGfK&+awIv6XS3C;1(xq|C3d
zad^FtPq|+`bUR!2YducAA5o`m-A|*|A!>hq?_2BbzVOnzvejQMRb&gfk7Am-PNvgd
z&vlt<KdS46r1%N@yF}9J*>=8S>-j!$+jd;?`}}^?cBb~B?ax+j*R_24qI7BcxS;#D
zZQeu`uc>*EDZhTtpTgSN`dp$d-wpB`=bq&cx^8C8O^$7Jd*3<g{A^u+Vb^U^>ZLm_
zb&tKbuxIrtjxd=zB<1=k-8kH?qxAV(;jPm?dG)qEio)`#$3~B-)Sz^c?ab8gv+1_$
z_d682ojwmFy|V)P{aClO-}4>aF6eV!LH7f<c-`{d_6y1X`|n`xJZnl`Dyr+)O!+bW
z-jwb?Q!|9$(QKvZ?>1*@Tr%CR+wmaj`L^E2(c`w|`$0*IQ`gZeo%H$a6<^TvzogmQ
z_B$H!+IIfS_e0`3u9%u<rs7ZcK4Q@8wxrvq`<<br)aUo(sA&2;IqY-Qw$+Pje)!!F
zyq?nxs$ZtR_oBBS_BqR_>Lq<YbBizO-+l0kPrrYdZrbm9@@tpw^?~1ZunXIN>bva&
z>Gngf@3roTN0=D}+~z@C`EI?|?;9p19rbe#e*L-Wr01K`NpB};n(sXcll}KMf{v?~
zeN@*Yner3$b1GhXd6k>~`z3De_dj`bxAB9wDqH#59*@@fYQr{VE7N^WB2)FbAD3;6
zdS8h8d%Ub;{rcHd+Uq<&s&e|}b)Um;Irq8aCjAhK0-}H@fC6!U*CX!s=x0<k{k?Yl
zryFln$G@O{{2m8wvy1w9Ia9ymdif1|A2+Ic%xl}u&*Q!yfKED1olgh#8}xTZy&qAh
zqrN_h+h5$*1Jd5k-$(L#9xUnQanIMN+w=N8HtFU1_YO^^quw{O#b2h^r`PjKQLh*A
z7w&!5u+Nuf%Q{JFKf?Zw9JW7kUw?#+GpY5iZngcH_Is>VLeb%`=0RM~^O$NU=zTA-
zf9|N(=YFT6(jlpTw;<?!k4&GJy0!22+^%i;UgzCOFE_P+#pPF@Ut!Nz_}#-Vz1}~R
z^342Q*r4Md^>Zk3`wM#CJgN3k|GpKMKmGatpy{CV)wF%w^OyEa_bV#@2h88++y>um
zP;oJ;xTEr8>bMqFd|rOC<#~MDo?p=8HT`{;xco=`o#(jxvG3Z-xO9rE9sT>(Vbfmw
zpXu^+pRY6ZyQOV@O^;Jt*XO3%33{F#)PBQnKd%1a#?6jtnjYt{{{7wu(eJ<5k`DWO
zUsCD?y$|eVA9kLEtrzzAM!m4@2Ce6%TiEiTaU{*&^nE1#dO_0R{%&#5<6c;M)AQ5z
z>vmJWC%WmLl<PXLzprMh+)WqL`AJ<rru%KHe^c`>F2APFM{OO~xZ_UuJCjMRZ_D+Z
zsW|;!2m7^io0nYI2(K@5S^tG9DjnSPPHKMK&kH40hp6I;`ujnC;|lwnf86zKc@Hn_
zd3Dn3#r<Bdw0FXsQ>xBzL<c){@Vo!&cB2M7)am2iTO3=d_s5nr&kMQ5OSx&fwNJh+
z$3d?+O+T-IUg^@*^;%qh!~X7*ZS_pwe{%Dm>GzM*o&Ryy*Yyhf{;8?Bz4o=FmM86Z
zNmJt^e*aJ)uJ_ZTy8k79{<xo5*ru>$e$CY3scNa*PKC`1Q>AwNn_nTnVt(aRPC;RQ
zl^yBYsoYd<lVMdm7UuWr+_>84U1$BcmqzXOYs+zt7k8~0ZBC0<XRqDos%?&Ke7NzO
zzYaPgH_ukqI(yKmN57quXDi2_eaaO-6y)UD&bDXOKjPV8IeDfT?W#@R+jG5tx;X!}
z{>m2KbkCSN6$bWj(xsPjN48wj;?b60sdTyDbNyE5ee$*=^rwT@xVV+i*t6cO;uGg7
zp&1>}y}^E$4VuzuuJX$>tNv$ogQD5W6nh-@<qwAspQVT|Q@m`#tNZW!+-fK3G^Nkv
z{ck<|!Xa~1QfR85_}zQs8=JoEy->wP+D$soEk148s%sWIar3%aYl;tUzS6OkeMaY(
zUp8WmV=JE<x=+1XTh}>3IlD{U+~J)+Rkrfp-1eP6dGrpIbpFuytS7tNJ=U>>VfVRt
z9ok-2apXkh#;|^=H%Dz6di%rc9oxBe)x}rbGINm=lr63vIDf_c-#D4(>Ke0p>^^d)
zii?bYSoUxE^wqC@FkMAOzi|xgaoO)Z4jrmY(Qh1L7morBU+6WtQR$}~TbTc!w@zGi
z&3MNaZhH3CO{ESS?g;<T!PKlawO@JpHPs(4zfm0*-28a8pY-y$uP2kP$7LHHs@rtl
zLMJHG?bqS^H*Pw8>l`O6({10a{<wcvT0Hic8BS2T`3su=pXK9^epi(#5+*~3^_v@=
zwjx!k$QGWpw%kdh3Q8Ac3vJOoTjI?0ya~Er`nA``$8SB|&h5O&*8H<yIjHZT-+om=
z8TL9W=y5eJdwrafmbP_%d0i*Dl_yRAq{m~+{P=G3r#Ch@YN5&&+M+vg2HoeAV(+$3
zM3s*|!oUB0Au7F1`w4qr!Ee2w_e+v$|MYJI$_*Q}%&~=j$4BhayT9h8U)b~iOx5%1
zPu%6E&L{aj$(G;kvL#M*7sm6urs?ws@&DJ)dn9dso7zX+{C-jKlwPg+Z&y*#tsU|^
z)GIY?-H1&pC=adn`hfnYZdFNV(0(PwKIys!jc>szH%~nEvJFZoIs};qP0ky+u*0mi
zD$``Y%oH=-|8(2`<g#NokGNyCii>{7$@cri8)hzb-rH-L@}gfobi1V9SMrKW??=?>
zq@AbwjVEdyqW0(a{<z+5y8BbM*4K41o%R|(Q|){02T~sHIm~o*>-POIqn<lsons4a
zeJ_)^ZPA%HZ+<<e+t)p}I&ss~J`~k`J-^5EsO+MOJ5zqbK1X0`9%RaIy8E?PefAmK
zXv^=<e&h5Sr;=WtsrBb|{#(+^(;XMv@7GBDc0Zq~&)?1Co}V${=haSF#`XNU>2_>=
z&LZsdLi#+A^v()Ad%z=|hn_Y^dGVfdW%u6wxcSO7gMM!>==Z_=erHbXcK$Bk&a<ZE
zrK;TU!rm1&6)RhL+jWojdG&#nDlTSfhMZDk@Tiyen&sHamN}o@`r+*h98=8HxMaHB
zbiccssru8OFWYeB{=<~*^!grs;>tsQ{~_x0_gCgEJFEJoOP!L!*|Voi{ot3Sj;;J+
z{vpH5v?z9LW8Tc+cQv|bnPUrYnOE`ok*^gyaUEA4dHAOJHFHL&sA($x8Dragw_@uG
z6_!gT_s?xouv}#e=N<g^HSaf`t!!u1_oahgKL@q9U2nhfzf*H<*+(vzsnVt2{ou13
zZfUmZ=?@(@2L1l=p`BKDYVi7Jj$ht&_azNqe0s2Bicj}_yVD)Did3ffQ_mMCPZ{uu
zV;kM(gDv^#{v8&*zTfra*RD#>HPZ&)@~ab;e%rw=-RC7#)1%IwIeCoI>GnPAu=%>~
zdYb+SGo!#^HxKXj#L}Nsrs&q|sL!XP@|QH7bp7%#Us`!#t>2xfskaj}{ppOE9s5mM
z=?IgpL(p;cvOhTgllmP8OjEY9!h=VDm{&AeMMW>YyvpCdc;BlJI(m&VMYs0%uXNV<
zx9;AqY~|{W?=GwIUa2BmY3uiX$Ne_x#f!)8R+MebANl<Y1=sFb6cx8$xMk(r{nn_c
z$U5G)MVmXX`)s8$%~j_V4Lq;fI%S&lYwI)ZTI5_c!AX}__1PoU_xNv}xJf^RqJStM
z3ZTGVb9evu^V5r+Y$g4Siu(N|Q|WZ$z30ZBjn5xatm3Av&!u_At*&YHkD6;&s;HQ$
z_uiI1^YK3CU;m=Y6q(nipWltz|6}`{k-F)#xhgKA(=NR7luv)^JwxSWm_hHa?9z{j
zGyMKOnp8Ve$C0@FX6pCRsM^icbCprGyY{Zir(d@2M`bJdd-t|7>gQqH{6)NNW{tbB
z&zD72@0}&5E`0caCCZIw&X_#m<foS@VP<uxy+_pt+6|bef--9sN<upl_WKZFyA$_u
zCTyHZtrz$A>7_lhe?LL=-C5tL-=7hG!TyeIQs-mP=ZV4|*G$**``w_V+YOn$?WoHS
zUg_A%52p@Yu;Im3j;+++FAICV67)VurrxWbQ2Fw~O&^=%2>)7#pyPj8+4lWA7UmR{
z6xKRsL)Wi%+pLntyAOT(+*1c{RkrcSBfo0e;;}6%EQ8Kh)Apvmx8UX1bibnV|JC(R
zO&;H9rHYGD#T}KObe}(Ss~=T-UVh{{DA;w7=p}lMJ!XcJw0RKrJ9yb%&ujlPU7qgu
zSx%VT`GjddEO$yeO^;Jl&n;xikLmZ(Tih|VOzY34D$_JQ&SCxg{Z9WE9WQ;R%B2$>
z+qnF;E;m2(&uNaWeB#YR`uDm07bhx)ohM=I-J9FK^Cyqqp~5n3yFu&y`-Q%r5BO!f
z6E%azku-bL-^c2;_NH&AwOr%~)6=2ip5-dGzki-%nx^OH>DQn7&6mgj=Ge-cuAI_;
z@zT*w)O6E5s`t0k^%M3xD(Tj9)5Uatw`LV*&VKE86_lpq690d#z!Nv$_(Jan(;ZuB
z`uz2;Rg+v@fv|s1BkA?(e>vjI^9OBkN-BSP|Hgi$>uh$SqCRg(d;Q%BDfjpHIdq-S
z*A~Y8+?g%9gpISt6IGiXU+D)YD*BDX?RU<k<xhEG)i0wak5fr!RNu2pnxAy9bEJOT
z%PNkXs6>a7p@ZN3SGU{mxqhqjK6%@*m3n_{Nsl@7g$*C>Hrw^@6P&Uz_v3C8mpG>B
zrB_tv2buO0_5DTNpO=n)%WZ!T^~G^d)w}Vr6{@5&s_)SR^>gFZ*Ogg%;wqIX`km)V
zvAg7gg`JK+YOR9Jo0h9JGk=W=N>lx@_3yF9^?eh+?U))L@%x7Y-M&9&)N^O7b4)Yp
z^^mFWeMQx7x_T70DQwxTf6v_ij$(ezRJBy~RBorjW`(IzJO0hDkY6#saw?~wFu%%<
zbnR4bD!0k7svQgSdv$JHZFKt6zt-kkdtZ3-YDXwK$ULxb9i+blhzC(X6c7bO0Z~8{
z5Cud5QNVr$R_wO#$McVvtAwJ%j(K4Jj4qiyXnFmsyH!YSbWC%4oqO9JQhJFK7fa@7
zO0Pu$Q9u+B1w;W+Kok%KL;+Di6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KL;+Di6c7bO
z0Z~8{5Cud5Q9u+B1w;W+Kok%KL;+Di6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KL;+Di
z6c7bO0Z~8{5Cud5Q9u+B1w;W+Ad?F4UA7y?-}TTfOIIj(+t`KG*GyfjHr;>ik3W4f
z)|J>nFMG<dcO!N>iO(<f`jY<p@f3ajTA!e6RKDXc-|;U=-|-K+KC<1O<Ek(--#wM%
z*k2#VL>GU$Wc&Uk{SyU50rLv%KC;`v@3bqM7nC)&*1K@`*1PBF)W3_o_@%82&waX7
zUXLF}-ah8I?T#1C_gGpG1w;W+Kok%KL;)`aPD^dRr&6DN^Ms;<%!ByWLDcu>z1F#u
zivps6C?E=m0-}H@APR^AqJStM3Wx%tfG8jehytR3C?E=m0-}H@APR^AqJStM3Wx%t
zfG8jehytR3C?E=m0-}H@APR^AqJStM3Wx%tfG8jehytR3C?E=m0-}H@APR^AqJStM
z3Wx%tfG8jehytR3C?E=m0-}H@APR^AqJStM3Wx%tfG8jehytR3C?E=$QJ}C*VaxoQ
zsp_dCQn{T9n-!)??f5spLVm^k%Bh@!!u%>b(zR2$soW;Rs&*{Q@71|+wb8rA(zw}M
z9oO{!O?gqTdHu;_&v@+DuN|T2kPY*oeS_~WIq01EPPS8D$C3`Fzyrs$KDYNr#VTC}
zYgMd80Z~8{5CuemKNN`jyxSJtueq?xL))KO;Rr<sKl32n^SDgcx8-_e?^NsiCVjNk
zNh<aI!j}C`;>?T!aj*N#^io<C1w;W+Kok%KO11)Wyf4|~FTKx<0&(BxlJ?~H>}>cw
zyQIzI%<NA7TfMzev#Es_uT`dK>b}~6PM6tp_uS}2MgNMGEK9HFufmo`HEwbHse4+^
z)va=-IzgHAS#wF#19RHGyiG;Tu>Fquy2jM+5$n#XfB&V&ey^gUe!oBNv~9;{Q^zgS
ze#7pYe(PO-*2#BtTC-7^X1eP%)AKFtx~*~IyBDp>U*p8ZDub_^UOTT`ktv4NLFz@j
z|BIfYXO!m?am}Z&=XrkX9r@8C=QnsQx5yOtJmbJyTU1%7{4&$WEoncy?gyrO9;Mg!
zd!887F0S*PlG={n{m@psxQ<7eYDZs(l8)xRq-@tQ=>5c`*~fjKD%;z$b$^$*MFCMD
zdIe11$JPBu^}Aa1N{GiW3Ix5*3u_-Hwsh)vo#&_H`|s*z+x?epKcABRnR)Nh)bHx1
z{ATLD(%<bRQYH$-t$^SCFx|LJKS$x_|M9wY%FL^sn<o?<;+O|+>m=LG6K(5{+qg)+
zC{Pj=@cVp&*adqY#I#-|aju%^bJ+V@QqSM_c$2P^i8&=cMS-OK9-rxWyza}VTb^w*
z&hL2HcKt=Xrq7S!@}KU!jk`YjXPQCJ3o><|UfTWRy(Uxd8JX7C?>?FoJJv;9$1_v=
zfT{S?-PfafZZxQFroMY%>UYt&{l<M9Ot-zb^`X6V;|^M%cGG24`$1IeH&cGt-!sLi
z#zFjKmjXfe1>HWV8g9SSd*#!P+w%%hJ*SEvw|2-U9$`2I^yiENr~mi&-58H_8T2~9
zt=IbcDCuc8og`lr5Cuw(0_k4Ad)0SS^3ao3yzuez(nYp1=y54&_L;igNw;0o*NH*z
zBSmHJwT^<8yREx)^G)j?w}15Ty3R>@IsFj+wE{`oe<hx9-)Z%$+h5ND$<0;;(mk(;
zyMDHggY@eU1^n()QQKwuzLt!Cb|?_^IS|vmj_Nu~{Dk{1W6<kQv5)rs6x-)*y1xTu
zx_(^eho<*Y@h=L90-`|N3WUAR2^(kJ$IGyBO1+Y$K;Y+TV){EVLF2Oh{66mAnWQ~C
zzB3*6I2E*By3cQW)z9?xlC*E;dw!YL)$M2<_1`6o`}~deyw+7oE7#|n+xSbqC?E=m
z0_iFc*L_yniTb#eu0MUhjVhh)xRkW|rjPSMuNQ*an~py&|CxTCp`^BN`~4j~{;0+=
z?tPH!5?iT{LsV&w-(eZ|I7)k>Ko|vVeLqU#-ueBuFk>%uL;+I@1ihXxZEx%OO<ZxC
zng>bu8}|Gzs(PmN3HrOcZl9_564TY?$d4X5zrky{MM3HJz60Njf$7F8^+f?uU{?j2
zz5d$QU!J+n5sD6B=0U#?r}n#I>VA1qan~6tIikS-DG=3pp<h4$*RuRCN(KDtpJ_YO
z*UeG+kLx-%?slTmNBo%mUWe$L41J^Aqs5nb1^8Xtd`s!I;P<{w((TaO6dxS$YLjc*
zmd*>xZ2LRPQ9bu2e(Zb?A?Wu<GHoB#d9A5+oN0Z6wx87FKzc_(sqLjeT-WPfaY%VE
z1q#~~w(QowXYPM%<=0Fdk*bl(?Nr#TFjZ>DzxfsNE9O^D<rEa=SJ{!Soytw+HW^m6
zV_|-;&W)>$-nFrgBi?PfY{g>7i@Vm0Hs`j^JsX_aVZ37-m#jRw+NIacb8My8xF@~b
zR^9cu-)QjUko(^J-7(E*S8Z@_t?HKgt>dOi%2w**7B(GRIjJEEm{VY5g?c9*I<$vM
zD(QKvEhqJV=d&M`Up_ykYqRg1V#hRzFH=nVahmutrBAlC>v#S~WoKHysLls&|Kipg
zs((M!iJP`vx9IV*K5gZ1gGa1>ZnxD=P;S^b;qX)L+o-%)(0%&J6+T$1lE!OlwEXRj
zd8<{@`TjmVAHCzlbt<V`JT!mxeTOViFzkNkbv+Q)FZJ|wThh`=IX`W>bWQ*1U1q7c
z$oPk4)W<`=aYS|g;pWG09A0*;`?%>9SEkC7emn^p-@c7%pE~}~mCAM|_4w%(m#On6
zQ|rU@eONdDruyUM*Dao?^1a$mdU;ej`1O-?J;GkEr`xZv?NTq@_TB35pWpw%p*6QT
zLFwi%Xns=8i=&Ri?YvL&gWU)B(ktrggG~F2dLEkgXZv-$-{YDoyP*4}UweJr{npd%
zO!eEeU$6UFLCbYr{HBw-KK#a?DZ9AtPet91zV7^{y>ve9$IZPD`KWYJy7X$ttvqU-
z@aNW^S3dQV&baT(nQo8k5nGt)^SQ9?``tIf+L?Z?1ph&gvq`fb-tPWMYp>nxB%ML;
zD+jet`abJ7KGXZ7_}}^W=xou$>pn!%%1vKi1@-TBK9_E}E$1Uy6KBBZS(COtc8Mq9
z&-eR1zKLD5&$lG)_?7NH8MJ=b>le3rUhVg<v$o5*59H(pW!=7|x?MeHt&?<yy<U@g
z$$rnn^!T`~CtLDO>l2s%r0DInkD4w|>hBtU;|hBHlCHhq@krYFM$mZtUf=lDL$~wm
zKWTP)KcY@Q_@7f7)OmS>V;iH^A!>hqpNrDlP0De?Zya&07hNaQX|H~nYCo>?k+|9k
z`hJ?&hkLJ2*NxxNY-Q4Z&-02WQ^!wJ`;g!BVK2K~?w4lU{ojQ2$J9K?R=-}y_oSAa
zdao!`e!bqKN@{sr@6}~${rNrrFR6Cvj!RtU-_lOF`~P0&HMW$yoqy@`H{Eq?iymHa
zy6Khl{IKg>p9h&rr#t_msvoxhwth!Q;`aAFU@x7*&YNth=cP~5%f0UJL|yK8+@spp
z@Dp`E;`T>BOwsSSc-eWKXLyzWdH+GXzeK<F`ugw8`nY(lKkgsve)Y85{p6tS>GomM
zdcQN3PWSoAq@Q2z5??}|dkuOY=GP|=m^AI)70OlyeXb&D_F=D6yy~U9{=DjE>i%fD
z?Rp)jqAEXoUWISVowh|q&8XtT597>nG%UU5qiyB8?C;unM`Y!Oj~n{=idUNX;FsyH
zN4>t+d6}v5$i~y>Jc(*Ou#VGZrp^<iYBwr9!}<xkZ)Uol*M5-n^04FORxjOkq}NY(
zU(x+}rNg$z^{(4{Ka|HsuYOAT-z$*ra{_*k|9<WMyno~Ke8;VKN!^b!wU4{$r04q`
zr=WIvJ3-S)9Z!jC=fB4obX>jc-HyK{H9zQgWORG4`&vQQwQlb<o^ItY{&N4XTh7`q
zPv?n44;XmPwfp2{ns1dK_2{PFW%6ugT<2p^w`2PG@SyL#M`iE#Jtx0*rtTkx{e3BD
zJyZ8<bieFBQO#4g-}OxA(+{C2APV?bz_#;7|2_M|GVFaFTj~YX!^>Vje>Ii%io>fs
zQ`ebZ*9%hq$M+Y5p6^6uZ)(2!^~=1qZC{Q1J{&qFl~LVyjjDgP%_lux)93%D{b%ZU
z6IHwVb)Ri%)8lK}zv=rnrv1ltJ{;A4DSkwOlBht~=YmS2R&n+@?D=)r7~|TnX(w!)
zNv#+3eE_kK_U}aN&$mi?`pFeOSgNAZ!SDSHv5WS5?&z7+`-DM{Q(@0HOxKI*exmp}
zx!R@I&2z%)V0xV6_V0GS!g)|sjOzFJpnl@k)3)~9#@{X9w*Kky>i2PNP3!$CDQUNP
znW_9t&7Vy5H)y;1^QorOe&Y-JdstZesIG5xKVJQe>UqI*-ycazeO<q((~UmrUD$Z;
zB4vtU=U>w5MRlFXb!S|RD(+1AiAyKj+VR>)DSvV7`7Otl*{*b^JD#SGhf(`CwV!X;
zIN|V9?%Sxsl5z4&zvs*Ni|V<BO!@KK->`NskGr;6gMXDOGEK%KX$*Rw(Kh?2?gM4Y
zPt?byxaVKG*O@`%O}9Pj2fbhCW*;>Eq}iK(J}vD1Y^j&t?*U=;Gd(|(dVKM_|FeHY
zO*h@6Ixq9vzogl@>0&xR?DdoBxWxZoD-hK>$&{a{_Y>XUU#n(M?$%1j9b3sMooAZ!
zCQje4?Mr*+2}Os$n+I+yCsX&e{oWr5d%o{>{o|HDyxski)?T~WiHdG|CpACr<496<
zh$^nA_4FH8*!2>3y`)}O>2c}jC6W&J_s&E&(e2QoJ(R8U!0-O6+lBp|FDdVh=>4%J
z%{+69H?H$uxAw_T+Wo7f#>0B>`W-&(db6co)W5&k>M!j5&UEW#S}*nuopJ4FLEDMy
zI*9!|Y6ji!+{S+wKl*pM-Tsb+W~ck#MQ;~BdcC;Q`h1jhv<h(iGtFqD&dJfQr)^97
zy-(?8SJ<YoWq!?6jnt8;+)jnf3R9(a{F`4PzhZvnR8B!*ew7{R+Ns=BZj)hEI~L~m
z>fE^6=%_~_I)4T0FFf)4edal~@yXIH_o_R3gkvlFEFN4iY)c=<Rvz=s*m_t0{Iz2{
z_wQG?>2YuNQKlK)ijBrorw;p{xZfAWjxB7}wRr9P!r4x`j7B~2A__#Kz=cP>@WpNi
z&vcSXdbxk!HaC^NeU0+VR=eNQ`J_R`$`qfN@JaKZwk%S_mnq(O$R02Lr`Oji=`^KJ
z(&nwHev02@DzN)abI-VF!b(SImkzJ>*z=M)7tV8R<(xIYZd-cUI49}6^T>0{pY`KN
z#}+1Ym&wUlQdH}RrQKFKC6y;#@m{aZUky;1=Gjke>va34E1bBP)cNQY*P7=BT{NS?
zG-Zlj?Mt~RAPR^AqQKuPU}}HH@87pBrJ{f+APR^AqJStM3Wx%tfG8jehytR3C?E=m
z0-}H@APR^AqJSu1Qi0w}?!K$|_|=Xu9vxQo7*YGob;}*0=#b6xpz`NKf5}<9(n$(U
z?xEsa6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KGNVAMbNdU_!gY=?9v!B9ee9Btj@sbF
z&CJZM<hGjh?*W*KFS#R<%*afQgZQ=MyBxNRQ!-bNL?sG{0;0hGDPYTfnyqo({NOF?
znl_&11m*vYp#1Nz6nJxLjWVm>S*yIb;nV|aAGvC!3d%Doj+%1#<;BV`t9`wu;_=_F
zQ+|2;zm6}s`TaG@FRvO|;jVkzZ&bGO%CdJhTYuJi6*vEC?v$jw7oIx1%^987DqC5$
zYNORZcU-69Vy5QC!&h|J(D2+Tj;-8w&GDy>8Z*H$#Y~Nhv?~gT0>KqHuJ3yfzi{D1
z$1nLli_q+M%0%B}=o@@Rg{xuecgAqhNG(x76c7bO0aFUpJ964@N0nKiY~xw4r7qli
zQQ0C}dHgG<40-yj+@h!$_54Wpyf0`xVgC+Q(00<b_xk%e>H3%Y=_#;sL7NALoi)|T
z6y2gff7{>-+Bi!b;U7A<=_L80fG8jehyvNIfT{aTrv1kKT)3_6JzW3fJ}q7;cA{c-
zuZEnQ=~ta`#ri8|Ikqy}$4B}n3Wx%tfG8jehyux1AZdSB*DId5zb}^j5lmKu`}eEc
z^td<sC|lWQ@!*1CTlzS*a?%N{mfX;8sbeb()?awy_xsFqY^B%lUP&+CFm74S_~ySj
zw(^bZKD_JgrxrM-nXFwmInj*!xavdAfB)FY6qlF1r_%gyCn`5qdhD=oh7OyrgqhXh
zi3y)H|7pu26_m-D>Dd`c+|QTfKK65^A7&LfaWUIQ&DP&NB<^^=pJeNG0C9WW50LT~
zzS!;HnNCU5A=~zG>5nKN3Wx%kQee%Bg_WANUZHH|$l?>f+yC2P$}jW#Y^Z<qz*SDt
zSa0qb`Fq^A!m*7VPWZ6$zsApY!ZK6yUfK<(z}IbSUbgQgbDX#sE=sB8roc_Jo;&o-
zDVvq;>{GkmDODPmDiUT^hstXn=~1g}Zjo*L*MrB;{I<nLWh*=0J^cIqk6W#>g`=;k
zIqKt^bBfZX+rAf9e#a&MY<~Skzo@Xx*6%h3?U(8CI<CUF-m~YeQ?~Mf^(Vgb=%Hnb
zOwqLd+2TL$=hC8T&(`N`i2JJ__x$L|Qj49a825Y(`u$qnKCXCIS9ZR7{(!A2YQ`16
zt?lgJx6MtZZ(pN?cIaU1Jj>R&gFatm%HHdH+NR30bzG%iqJStM3Wx%tfG8jehys7F
zz|~zRwCdF-w<s$9eQWlW7G8Vu|NrvbT1O~4$UKmBko0{Z><%0D_sFEw3)}BZ*Q?g;
zj3=9n->t|kjxYPQp3u2?xiU?+_9b5w2&2GBSG?D2^H&2@N$0fPo8R}FGu<g^3^Qkv
zROi5TQ#*{jXtoMUQ+*gV=*fnK2hMVY>FHo<e8jIPV6y^+Z3<g<>)$i?zkSHBnL09c
zR4TVqVY9+isU833SIDoJUpbXiP?%q3N4j<@H<jCDSk;b&`Mo+ft~NR+=gxY?hdlMl
zLg%V3#nYNinYs%pKki>EuUgoAj00!9Qt!QsN^MXd?z`gmS`&X!ue4tA`l<uw{+Vvy
z<=tH8m91*U`uSbUmHJKX)2sEiD;lm<BR)Isqo;o;Tcor0cLOi)aePiu_x{VS>hQwu
zMIXJe<g#;r-lYC<!)@zkeXvMPy?jBFUmCCYlla&DV7K9`yGy;erPQD89P#FPot8aY
zp=jTt7mlg&a&8gv;D`1G4BY3*a`krmv;CQ0Hl4gr%QAoZIpwWe2i#qIpFhiK2S1Rw
z^ms1r_w=;2L-#5|$1B#C+HhI9PaNXGo^td+Z#NPb<A5L1#HqJKKI23@AO3G#@k2iL
zj2HbxSIW^7JN)36xQGWkh#!bQ#)C9^V|V_!XPn(?{r*MheCvDfmu-8^o`23O;vjDN
zO+Lgwe%+|sQ;+`X<>c%0ntbY?_V$2g*WJGA&-qS%)J%8tWBiDp@rC3=@^#W*?4WK(
znsVl)PR13TNwd%4$BX!NvtxYG0sD{Np8nmD?+*TRKcs!#5B<2Z{F<9ioW0V4?7Lp1
zKiHAh=~b@VOZt!FA>*pAJN7BcnFq`_(#&hdpM1#p>-8w79)7T+UFuU#J>tcVa*iL^
zQBE3vI%yX>>gl9isF&+$;?eV2N09beSFDquWF6?#=R4(Y`IPG<o^<!sl2RXi8E4Xr
zGv`61DaRg?5A}R~-zAM5=M~I*%2|J;u_qq#A@y}Kp5F|wU%&Qym5SUr!sYzfE1&a3
zovcIFE&BlHGvs5ZlXBK6q<`q4j}ztSf*<@dPWt*G?N*<BzeIP&3lbO9<Dp-~^YcI2
z41cB6oL#Puu)_~&FQNzg9{nL7QqH&$C;sUV_4PQ|H?gNaq(1FI{810Tj5GDIgT%}6
zk?|%?JlNx3_ltk@$BsBi6F>E_XCH?6!;XI92V#dG(zHW8NdNS>@JBs;AJKJiD|hpY
zf4!V|nXlNPH|>yy*y+R{^jogm>v|HOZm0X#)AW~mka30di}DK9Z*TL!vCAEFhpa2c
zk$sW*3W<a9B(0P7$tPaYkhuEo`}S*FzMA>xbprN~evqa<`H*sW`wfE{?fw1KKkvVg
zuiNYK;vZtqd}n==hUB}ogCD(IZwEWb@rHKDhxo$}^Ncjq?Z}6e;}??e*1ldJd&ZY?
z$oj+1jm$H>9Gy9?LHxVXEgtMCXFkv_<vOv4dbwA9U0>|o{E_cP^w3FrUex<xdzyJG
z-1+{o=#iazu>b319W(Ci1MIV$50DQz{;=<2r~6_5g6vbIIp5)Yl;Z$tot%&8`&hQ5
z84q1wbf^F5jb4=drLL=<=Dbj+ZpVBgpYdfLFh4k-WZa3H{*vz&Kjk{p9e3h|#N$TY
zPLGFrtRK!B$k%@l;=EcXx<cAPPxM7E$|3qv4k;%d{6YMtt83KtiI3|J(ySBW9`V2X
zM7kaJ<m<nWv99okole#r?Lp?Z-adZy`%k2?XZ&f8dc;XPj3a5rQz!ND3(2RS=z>4)
z_d((#4(2cQ-TV;`ej)kHYrQ|%v7fk+xX_FKk=DzJ6aQZEqYK0@^&#!D-pJQU`^+Eu
z&3;0C^kM%Xt&{bOALbu+_@f+BPJF~gKd|HXe)><`)FX{O)cxymQI0*N9sHAq#K-(2
zf5iXRJLS~Jjx=Q4na}v4oHWFq@y9RiLCPWdtRwPC6F2QJ4%neT_PS2k6E`|Ty`1{&
z%k+!(X_vUDPq|L~F%KD6;-MYv^!bRL8>#0;w|aWMUY~iZ+tGiwIPs(Fp!>m&c1WWS
zY3h-N<YN!n4;UB58GAi0($v?<xUv60&R0l7?4h2o`)7a0p7{`z=z@RZBdwRC6YUZw
z<amM~@*#fRh+jA2M=!@8`EGXP>(uKr?vVLLztCS#ldt<HjXi1NrF}1bsR!N08-H$e
ztCwy*{fEp4(vbMk10DQQ?-%pbjqE#6w<DkSi4VHv>*e$hQtp+;&W+UPdXejTH~Xma
zY1fO4vrg*i`FecVLA{)`S2=c!m)AHmkFdi(^(lws>+Mhu@kbiE<>NOfbzSu|{c+2u
z+>7)#Ea{h9ef9y$-QuAil%uDce=onZrxX7=u_K=}BtExwN}TNT)Z=`DH2d(3hcDiz
zVAQutC_2bIkaduheSrOn>pHI6xt`|woa-sb`80OqbAHWr7spG89qo`$J>r41136A`
zo<~0Rl;e*y?LfDBl<OpJ>fsksPF&R2?a`V0DjY{RzLBOr#6SLYa-RbKl%orz9DT@#
z)Psy8Y3w+@k;X4^q62Z!U+l|{dGVu9AKJHwdgMcNpdEA|4v7AGntVuo{L&8bQXe}#
zUi@&p$By_${O|r6c9hdU^k$r?M?B<XM}5k%huD*boX=1X5;vqB>bco5j>JR%uqTZl
zbj2P!;-#E==uSM;!%ipm)W@DQ<4L=E{LDw{VTUf*(++mThkxb=<(%jFCGENqzd^}3
z1ZAJ&n;-V`*M}By-pf3K_#^F?^iQwvmDc@p9Af-*zsxW4{gQcu&WtnTe_omDYnL>e
zukc4ccGP43&@b%qPd@XRe7!zth#$!BVQ%HvLFO~|kbRvvAmzkK+}x*NT<}jhY3#_y
zFXi}WyjV}VuB5RiO<bfYrycqQ=|APz(+@~I#7RFO`8w$*cBCQpq-lq=9vAUZk9nr|
zn{~?dDgA)d$Bum3C9PBUr`Ol*DaSACfc6<D_DkNce7eHOeVbRPQMCPo+P{~Zcz6+Z
z)JF%1f9%jh*N^dbt4CbC*NOc@O`hm}&oTQIF^=?$`AIv(LAkD%-~AtdI@#w;k^Y+^
z<Ksr&Gj=2Un;RMTsK|U~Jed!q_3@!y`o}zj%nwL?^kzM?573SqsYiQ|_E=Y>DTlOA
z8h;0!vAy{Ep$8P{Wc^`>9@K~UbtC7!lv6L$<a!mdUKv+(gy;ZueaLqkKk}Im%r~9r
z1X=&+$Ufvo>~wvJ&#iy-Q;(0hy@($8@gni-agwJ0kbFo#bn1GNW_+Prf0*Zbf5>Ou
z>Tz+NK|GXGPuBxK^v8|(r5wKyzl<;A$MK8)(;xO5;=&I^7wUPD_VEvChqxg2q`mxO
zN59;3(Cx4%ACgahi3ef_v142r56Za?Vcn6&-ff=amw0sQx}Ym|dbwK~doNNSJNiey
zKA!X!-LPZ+=%n0@x_!_zaWn6!PkY3V9clVWKBOMy^pkOb%twele%#XJL;SNY$=69-
z^iL<}q2v<>`Cj$)a?<$0FQgn|M;bCNltadWG<NuRquw9d!Jk__T~B@7^m09of1P^V
zdOdx<Fy8t;KtCY5&>m@>_<<Z>XqWcLr{DTMMVfM`=hKc(+R=#}^Bcc9*)Q2&NYg&&
zxsY;*o$gmp6A$^MA^AFW-SxCSZp5RL>(2YX`guT)sRtJ2Uvlr+?VDEL<$F?`uY@K2
zL3h%O563gEvvl35hrM2oK6?N3eA<Ed8KFIYPCoS^dNI!QLvL3v*XKL_=#Ng_AMH?&
zH1+VWlQ?vLq|pI?ZloRjaee{O3%}SyeIL^2D{0y%%{(Tbc|bq0gV>RV_+frQ;-VgQ
z<P*P6uYHjH4nGim7)RZ2*!4v{+DVt_CfxZug6OJ`n_HTB#dwp(UeDL(4QcH4`g%Or
z>-l<`c3DUG)k(YhJRl!?{A0&D!j5%IyLvt9V-KlEn*Nd2iC;*4(%92);=ztMpxz#S
z^nC0o$1m-XpDD6mF)nWBrKa*}7jmA;d?7w`^dh=K{OHvE(!Op-K5>zT*t_XNIdM}D
zdybp>{?9yt%uDP!K2uJ8<|{-8<`?smcp%3Kp(r2<gjaz3;jB}hAJWObpfh~SQqx8S
z*q8Nvoc$b9jy?OjPWE%i{L+ti`f-4??vM8HN4Z;?eA*@LM%_PlUiCS?LHy7UH{Em{
zsi!k3bLr>HFZr?UzRUjnUM}YiI(5AmCrJO$1+q^;uGdLJ@^wFy(+<Qx_UH~N#~w1C
zlv5w-^*Im1o;3F#$fq8}4!^WZIr-Elt(TLA`uwF|j1Os@%v<vD!}%g<&I7R{4T+O-
zo%kVMFMI5$hfca*w|qS={Glr(4$|B|CGAChJ(x~&J>b<2`>vZE^8$U{?7Z@sKg<{6
zBMn&>_#>Y*aX`u;cBHXm{PcN2S|2~^>E-033++I?U)bsGlTSVLp*`B6KW@Yi^BX(r
zyZPhyAL?UI9MG+teEiUF+M_=Ckb3l+G;x!!r^$zm7xR~P(E+_#&pIim9z+MyBmQ@P
zo%^ml@6uH*pLY0tRf-_}pgw7c9{7dCp_6imANot0_Rs}C<m>IC1II;lgydrfDaSu?
zLE0lO;)LY;rCv`@>+R8h{Oje+KkRf;Pp4i_PveL7&<mZ&hk8F~k8#2d(jIo`t@|eq
zozx?rbrqD<hi>&bZ_&%C$9@5^qg*HRj^i}sIKg~HC&)PH{>Vp1bVeuY>-C5;Eb$k%
zKJmHvanqgtxV4K8x?kEuAIiN*JzZz=A%5_$6Fc%r>tx;N`J}1OdeO=J;`)L2>DcGV
zXP!a&M?2_<f6~lX?5U4m{eBE-h}{U6-+3vgJ^D*rq`gSokhr0~-q4Ny==~#~e&827
zw|$Vfb-K;pOy#3D{c_V6JH4EINI7Zxk6x6!^_TTTIce%a_5=1`+R^LLFXG|6inJT`
zdb*unPMUtYk@=>V6BmB`lKF>Dtb24pFUrvglF$A_9K^?Z;qPBTboL_tbvxRppXf_I
z{US{{{UIO!<U`8c=yu<U`qzH>L8mHt2Nm&L&bgmGbIpryA6ztH#Kp}&IBWkR+QmO<
z<|S!2Vn==G_8tU&s7G2Sey|(yzx|ec>`8mkO{bvwv>TLqe2fqIr1kkhKBSy^PJQM(
z^ALN=$!9*|2a=Cpy*_D(KR4>}V8?nSjX&Zb4T+!lNK+2U*GWD6L)s?|$wvqB=?`hj
zX`ky+_C=kX7eJ1~UiYmyZc~r*N_5mo{KQFo5PQ<}gL;tj4APK#Iw_|f?dWOZ!X9Er
znsH?R*Vh~6ZloQ(oc%=iN1FEZeV%gsk%r_$J>RQc<|(@3hj<|6_;ZU>&(HQWdc$n*
zuk>%{zYEDei7w28!Ik&AspVPKih`1Th<!}o&q;G$Ng5K5PW?E-KE*!F`i7K;CH>Ir
z<KHi-j~@8d)8u0hX-7}%`IO@ak`LJ**$1)nBJKGl=ZE-ZoQT7(AKi|55|qquFEanV
zh>lRV=ePsi+NB<T-0aB5o-`yMKTVr9J9zV;BmTTz;k?N<(jQY~J~B_;h#l(|a$d=N
z^s29)uQ9*zPksFBY4T}LC*`a^=0{L6pY(d<>*K+GqT8EFd#x+IUE(I6_{oRl>!h6e
zq@f-U`MMu{ycu`&)9q=`jkKdvx2IkF>*ds^AK2;C{SZI%l;bpZ5If58OTQ`Cj}!X(
zVt&&e^+-e7B~3k@`ufrBiC3?O9i*Hze)W9fC7(FShve&eP!4IwjqEp+>*rOZiHG<}
zL-KXvS0~rq_@|tF{Ll~5=uQ8jZiip&Dc95Fdy(@|{6g+C=sI$rm44|&H}Xl-PpIe9
z4*9fCnsVs2KCp+h7nJBpJ#-`u-P$2O${Bxs|0(Hd_Cu)ahJKW@F1Y`}{!MvUJ=y=m
z*7sYFc>UUAM|{hEy64IIzwh_w_vt7X?tFbOdSs^_?7MDcAMi`oBj;mLk$sBe25Ft_
zV|qUElTVs>NxKm{{1b<8#qYHy{^I&~qD7y-)`$57^?AqnI_nZ0^)&gA^Eu9sxKBfx
z^Ha_zIWHs)IZuGpKjExu2hYE{a*<Bf6>(6HH0QyjIi9eds1NC<8#$lE-ix%0KXf8Z
zzlkg9)aMIn=95mZa@}4}>v~Ym{!ZGBdVM|3c`D<ik0<dnUZk-jpZ0bCZfWfC!@PB~
zC!h02($ps%l*B>5pza4dh#hI2dVM`jeLdfAn)bZx+0R(-)YH?tuK2~?jk+CaFS@ly
zJ?1w$F+Yi$ai^SqQa@XXZs<eWZ-2e)@$2S~dEizab-q_TwCC3jd*%<vGv*Qb%rlNx
z9H-c)baI@<4q``|{g(3q%BfG9eU*8yr^zRN%DqT`XovdbbG|`2`Cfi0*X@bZjns3a
zTRrmidbCfuPV5;E()1TS7$3$3U8sjWq+IuhJ@X2(UlAYeQeP*2XotAm&Lf%UVM(0C
zfv&Vm9P}SM<~Q+D4za_(TYT&@P+y<aM@PzaVn;sfntb$!^ouwl@!LZ5*W;((=mD{V
z{QfBPrvUQ;-68qt%6es;Fz<9?Pd@XOe8v$nj*xunxuwZx{~<2cxf|&p@26m=_m_C_
zPk+dV<U{=Eq#RPtK0^Q9&Ii~}u!rR17hUiN$%o`a^u`~goO+Oah&?2qezFgf5AlZ`
z;{z#oBYxb7o_e{iA9m!sk^Xs|kGj<-pLj^S5nY&Xkor*92RpsKZimjCH{g%@I+<rs
zw`U$fxB9Frh#z!7XFVU?$%oVnO1(bqa(u?0Zb!Zs@#kenIqg8|Gd~$W^0Cv)$;U6G
z9`@+3`z5XKkIWZ{9qmE#buw=t{vq>_ddwf}$=B0*KNvskDTm~{5&w{WF|Wzj>yr<u
zN1FQNGafoA$3HqypR`Weq5q`ugJ0TX9+1zt>gSV`Q;#(Mq27*Lnz*TlKkVrzX`Q+~
z{pC1=J?%i+gZQC6_Djl{FXS^%Ip4qz;zwU^q#^A>bY*|j)6^rM_UI>mu*){;^Aw%v
zzh18AGmgvy#uGBGq?t$POnuVm4bhvpNb6+$pk7Xza!9^T)~g;r{iGbC3&f6jOFsP~
zO}|M)^3jKU{83Im>l9KB(O*x~KjOl_KJKKc4{6^mO}-n^iE=$n|0##M9r@7TRhqXI
zdZLd`z5lciDMwe*5Puw>87KNjnt28pH}*?Ny!gkD-ahf^#GZWoK+0*4_{gUo?U9c?
z@l%g@v8R4m5~o*v>~!k&h!?%-4|b&K57+Z<$9e1^^_d6g!2YD?lcpTf5Bw4z`J|~2
zX-~Jqo_;e<5PQh@QV-oIXC0I8McN@w`iXyt9qo+x-}}+*SNgtAKEyBeA^m1u>fa&7
zj`&Cu563y?0sSS-xT6dGp<Tui>iOu6U+hTZAJPuTQTmNO#1AR=OZ@2e=mW8X#7+P8
z_{rBvzwk>r{wOD(@gNP+g>j?*l#@n({E|<5Uex2HKK|TD9FX-&ntoBAd`LdT-YZ`0
zA@M@yAN|4~?Lu8w>NCzd@q-=yNi$xoOZ?Lw<Ah)0#4r6JjbGB#V}6<@<4GL!gE;jx
z<3~Ah>uLO<pH9}PZl|ZQgS10jkb0Cu^08z5=m&98PCdqrv`*@I5kIs?J9=E?>+R@i
z>M`%=55yjS(ECg?vlgry^n%PU)&<0#v>Q1e!Vmh8rXI&dH@dZlAL1SHzvI9DJ_mlu
z#~<W(3+(Vqd_2FWlls_^rr$a#XFWmo8Rj$dOg~;@4~dihlZN;WO1Ig_e#N{dUfn<O
zGv2yBdVTz`u1Mntd!6{lj(YT;cCpj*84r#p>=VSH`^7(G9Q=~_i3k7W>ttUc9!Nbm
z;)nXAA?@o#A4oZAh+n;ZxA@Qp-E`{Z_;rhy`oyEpKhpTs*AeB+XZ+D$($se&b`U$#
zkpAf9v`@PbdoQ97q&~X1jfa<CuX5Ufv=3QVI$4i8i5GvAlMg8;4QUr*r&I42_V~e`
zeD+mzBp))q`nXb_E{O-Ci+-FUe)ca&e%wSCzhqo=o#{8!_0;Y01G&y;y+g(e>iK#a
zyPzZv$iB%qGA<l{y!674Zl~`f=)^o=9@8Eq4*cPta)|!elaGJWtar|P^nCo#9;98;
zkbRr+AWi@D@xTsJPTYE$c68z&GS9H1AC%)4>g_WRsju77FY4*k?I~y6iI4BE5f|e_
zy!^dQ$|3djaiCq|=KPuQMOWgXAJij0%5~DNPCX9%a{f)Z?vH$(#G~7j_A18?>i&rz
zy&?KSy&k%<pVL3xo_52MabW*}wBvTZA2y$U`K7K)Txqxd1#RDNKTX*YFYA+inBxlL
zO#94N@~KbSFGpzC+j>3x(k^7*V4ZNBa63*>&OQcNxAcehAo<us;vkKGeO>7NA&r0P
z>%<@V_$5t!;$vOYAJUNH2Wern?}4*FqCew>?#vf-$BuDiyxGUGgOuZkd8DWD2WbaV
zk2G=P2NE}S_|r*!{8Aq~<|pOQt(@}<<~jDHby5!9_G5IkjqGpe#klDG@ai}7ir*m_
zC&)Ov`KLY3+nA4#_}G_qQlGzP??vkKz3sk3E0lYy!U4X&@A1g+8J&*lUaRQcQ*K*x
zO}k@!*CWoT$^3-q#(YLU%E{+Aj6LPpW5+rpjU8!-9p#Ka>lnYx1Lj4#Ojl2@_TA!5
zx811fqbsC8j4R{AxNzP@KJ_5;LMQf6_fNhXIc^b$+x(zhC-G4ay2Vl0rm$sx&D2q;
znyK7Qh0O|6rFQ(AUm?F@e&tk7L1BKC9qHPs+*EFpVO2X8=J)E{xY}s9-7a{3ha-oj
z20S`fnPTe|cO8*BcBT@F4l)m-TL-~cy3`N_L;+F2tOA8sRk^Rqn1#xXW?L8UqJStM
z3Wx$FRe`~6{<-SMA(NEtbUS_~Ex*#K&F^2`bBRhSO|645*YDS~-txWjqUL!u#<cxp
z?P4d}=`}8OD<AD_K6$h<&3BxIBW}2DoQjKH{g!f3Kok%KN}2+l3tCnEZe~tiwy>m)
zpY&H05Cud5Q9u+B1w?_k6sZ33iFfWfYNZOwxZ;y`L;+Di6c7bO0a3sn1#JKQu%yS&
zOzYdV)$5NOzDx;42bl-54wAADus;b!fg~z$W514-5BO@NBNQEyJP(q%cEZFocw58a
zb;mDJQE}e2KW)tUXsxo9U9YZj&g177E8BR;C6CSh{jgQab{@K>sPn(pFIKj4UezZ1
zj(u~EvW@ju?)g^F9)*rDm=0m~T&W`phytQO5*3&}YsQ?FpBJm7(ChbgJ<cR*DlwHD
z1q#1v(`E8~qnrlk3~S%FLl2?oAoIYYb&&L(%GCFg#P81EqhcP!_1-J(2}J=>Kok%K
zL;+F29t8$|(Ra;-Z&b=N#d7aIvGmw(!<BGn9S+=F_p36~zjTs9duFUeDhh}KqJStM
z3S_?moHq@)<*WXqyAF5Y*4N6se`5YPXKnLohg@~xo63#){W1N%(8kUOHGK5)*-Gbe
zbt(<)f9?tul)4|UwC(R*=)dFbsqVOLZr}Mz7_SaNcLJ~9XQe#cd*Y&7GIa~OKC<0D
z=y6T#!<`S<qT7UDT9)o{+-S!yO@C)o{Qv#;OtXD|lKzPTqJVh?-YL_f``!y@sh~Xk
z?bAQ4f95Z0+g)$mQFq<z4qWqm!L5z{IlzGz-*e*9k8k_l@%vtd`5tF?3o8ydVecz<
zU#o0oc8`wqP81LYL;?F0_%!#dm%3fJS_wr5nFr?A!Kin;9$bH~=_)GPx8q2}qJStM
z3Wx%J6j;^ny;}b{?+3>=`sth4hyrmd@OqPP>P<fBO{b(WZk?sQWGQgr#pf*>bonwR
z6dhz9$U2aHAlN>TtleCq5d}m6Q6M@6g8tp>=vo#}qJStM3Wx%tfV%<@6ntKO#<y=N
zzdW=1tSQ%SS*C=+bg<>UhHQ<q`PyM!uITlX3QG5t6Rf1D%`rtodd^gCwEex-Y>hwI
zD2ufyAPR^AqCiPeAgbpLljeu-hzNK74x8waU3%F5yo%rV&%@gJ-N%!*e@Q$$KM(JB
z9K}u)5Cud5Q9u+B1w;W+Kok%KL;+Di6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KM1hi~
zz&#spnts&h#f~sN9hf(-9@c*Ufyc~r!m{;>yN*a5J5$+8{qLaYe|Jew%kvF8Ki^=x
z9;}0s#;BjCdvL(C>d(|)uiU8roeb0S=B9Q}Z8&n~Dg_T+Q`Gri>ldqZ*|%$}*B?22
znKI3iHXpOKzrCtdnziBF0V-WqZ8hw(g)=6qlFq_5g)O`F@0t7G`I28VRWns9mD{PX
zSz)Twj(_tj<X6nEoXROE%&)Q|T|1SV%55^NYRAI-UY#3P8y(#=ySn!Uk9OU#*fGT+
zcQpR_$Po)1q39s<Ai8xB-3}n0L;+Di6o^Ivuj70)t%w&<Kok%KL;+D?X9dn5Ua9AM
zFAa5U=h%Y!#r;1hc9PEJCk&bValtYtsoZ&XM3>t4z^=dTJ-1jz%{$InIDTS}B~G@}
zYb475^l^&{m-cf^vv!$dZ@c)GSx#K^>bI1O0-}H@APR^AqJStM3Wx%tfG8jehytR(
zt_qy@#-I`N&RXRJ<*uG2M-&hRL;+Di6c7bUx&js2ukQcMC+{nvojNdow@tkKxDB<J
zJ3`Sx=7Fq(r0fIiPeM^3i3)W2q|IIHo|x(gMTaEMgCwq<FfkS99d&%IS}UEXIQ^x)
zULJS%M#omZ+2yHG-yFNfv5h0E{?us7rZtZ3+~?s{1$m>EJGS!fdK*s4*>AOD8&7$>
z^^d6;^BiF?9m4FnQb!aJ1w?@)Dv<i%-#veNWui(7cRTakQ#LdltNb#Fno3M1M}Z5r
z^{;i#Ws4L0UQ$xumy)=W{XSbs8K0!@RQXQh`v3TFmJ*uPfq7v1JwW`^o=_AJ1w;W+
zKok%K>``EB&XqMz_@;c3DSCb1s-%`D?Kw$%W~@Xi3Wx%tfG8jeWWNI3XSuZ2rF(xk
zVuOO;JfGkCu{zuTOq2igs`IZt^Ody<>fd!6dGd)Bx9m01fqVBlx6PP4raL-=z8{Kx
z(DS*ZJYO6(j&4tF8#6amCJ$bA`uJ%_Ox)}U<JBSPPO$Et8C|-R+wKTOhmtT4g07Ej
zx0n4n-u+qEyUQLUkN)!Q4Ng)R^zXSW9<Ls{=)Sd%Ff%%2`~D>T69q&8^9uayl@S9U
z8Z^%d%DvWXzOQu6XI0nkL++~5arB?uyUuHyhF6%T-~|P_Z#{DRROR=*K=VD$?iRQY
zVJowHbfkBpfG8je*r&jr3syE=*8c}bC_2bIFuxA^UU}80SKPVPiHi2^I1;faAPR^A
zqJSR-cuvAL`sth4hyrmd@Wub$*DGm^TW4u6SqemTy&-<0y?zrtMb9Yj2grOfGoO;R
zn@cpJfG8jeM5jQ|zb_wM%i>8C5Cud5Q9u-MSD<U<ySBZ)ak}zL-UAfw{9dG&9{T%|
zlK$)WNKNVDcYf=3?kgu)N$oPn-gfaVvm7^$Y}DwGAurE$gvrt&*eHv&C?E=m0-``k
zQ6T939b4>8orj6vc+Y!9@4wgEw&%a%j@R!#9(P>Q-p4H}T-wj^tApQh6gyEs6c7bO
z0Z~8{5Cud5Q9u+B1w;W+Kok%KL;+Di6c7bO0Z~8{5Cud5Q9u+B1xlI%yPbLNDH|G&
zRl@XiVBVxY_;=5rUYV%E(w66{+~VZ#0l_c6AMnX_7v<!+@t-TsYWC#b>m6Zs=)gKC
zX>8u+f?2(Po2#N?_ra&UdC+gG9XIOxi{IyG!rJNUeay?JU3c-7V;r6O-(B-cyS+ye
zb$&^kkJ;MaeM+5K>cKfPRJ!as`M26D3%+(rIt$wrw#=`Ys+Fpp%I#FxtT0t-$G`a%
z@+;<7PURF7=2zK~{{Mxk+$O`Sb}Y>A)wyxC(b+n?q+g<dC?E=m0-}H@APR^AqJStM
z3Wx%tz~3v->fj3t`c)sIY~|*U+kW$}>Wfs;S^n5}myT+(MA^c>pEd5K%Z|U``4Ma8
zD>u5=loFd1_~rVk=k;$g&WVdQX(O?U0-}H@APR^A*`mOApBI&F+<K;pi`k-^^g|R7
z1w;W+Kok%K>{B4>^BwyXkcdTr=oCmj@I$kQkD08ZX4nWHIP~;Y%WhcVB%N;e({0VK
z^XrI3bq`prY+=o#%bovA-xbO>{(AC?6NmMftxWN-I<<bAdF5=!G>gk#|M2w2-z!@g
z*E&u2_r<XLsMHe$L;+Di6c7bO0Z~8{5Cud5Q9u+B1w;W+Kok%KM1kZhP<PuwC(T<k
zTG__rD=QI-0-`|F&r^uMc%RG7l-^Z8m~_xZZI>xqm?@oPhY<ymq`<6!hqgRt#9ZaY
zB*iSThytR3C?E=m0-``T1?u;^Y|k4C<~l;VbO<-vQcG9h)_afqxP8HNB`ldb=<`9+
zqCgZ1d{X(A`<Hcp-HD2G2JJnjOZhJyTe(~F!rU*nPjgH$iZM%)kLJ(Z_RQ)5n^d;a
z*7rjs?r7fc3HqMX!I!=|<lRNZN|+fPOuZL%;=`@h-9K)eGDX{-kKp*2q+?N%VmW*C
zQRmcey<UZ7(0wQA_DTPqs^9pI8JAb>qK3nrsAziKFSzW}TE{%xN5##!kAG2pZ#b$w
zB&v8a<tOg>qPG{-IEbI@QXt*)ub^s#-JiYM@p^tCX!)XR8ovE$-${-ez2cGbzgHmL
z?{x58>x%Dh_~_bWW;*FI=yi@;@6ugQ*;e08C&?EDL;+F2v;sl*^`zOmoqyPpZ~D7a
zru;|s`<<H~zxPL?wu}1teADY!{EGrMDR9eAn<ihe=x0Y*GIX$MZc418fG8jehytQO
zQWOY!o@6@iu)j|vrC!kAIg)0d?)Q;1UElWi2}X8lQtP1^D;-<;!Ku%+{o&`*dA9QR
z^IyFE)aq*;(=@%0ihof+6c7bO0Z~8{5Cud5Q9u;<KLwWm@a?PLG+FQXWzhR1aoO8`
zy-WO2|Gm4o{n`H9pnf0f|5nP*|BL!P%$>bUN)!+UL;+DC+Z8zVs3-q1Z@_#fE@r!)
z(!X#D#5Mm+wUh39H%X}<^?7An{%rj&5OGKK-gmb73AdX_Em1%e5CwuOpx=Lyw7>7-
zCrzi|BQG^X0doq(^*e;$c5J;*&hHeaX>KNn=S(UP*Y#-B?WC)j{v9OK>2%|j`l5g+
zAPR^A(JJ8f9fq*^(MBiUvP}Wg=jpotY|}3({oy-vNvAD;$0_W7oAi1~pC1z6)bsi`
z8Rv)&$<QGw`@6)ET?+WU51e$nsPCic{-S!n+xGobKaVb{X`aW4o0)!YXZK-;tsmNV
zv!ip<*~KIB+HQ1`#;DGNoAmkQmRn{lbfTud&v>Py>Q7idanE<Z?Pb?dIeEQ_YoAYE
z&+U$=uuWmhZvA`a{<jzTHB+@ybyB&V3Y!(CO6~YJze0Y+{K~1Eg2MbNJJPjNxvAVH
z!>V>H%<t8?akbIAcK*GyHXk*i%5ukxyVi_0r|Y<aF<V}oq-^7g)~~%ad(uQ@D;KXE
zpHo<GfwGlF<;tls@6A%SvtGkHHyzM)vNFwRS8e*<o_J;728~*6aBSh%m%euNoO0`&
zbm^ts`E@po?Y?E2N|&~->rBP%H7;)D?WWz3H+1o2$5tNfTyWd;zEhl}65Vg<+p)!<
zuh%$!x%`Mal_$Ql)-lCn7H@1-zWREH_%g*CZ|S{v*M)PPq|=l>M|3^tyk8f6sFFfc
z{lssj(x0x~x9NH(F4At&dH8!>23?rDS6<xw$JVZ+4_Z?$&sHAMq3qVxXO+scm8DNu
z*ZAlL+nu0n`pg>}4_&>`v6YWiIQh{veZN*o=S>H9=y%Awzbji9cAxunO0^q`r><6R
z4C|MABW^BOu<fEx9oyOG{xg2;ytS_rl*RdtJ~;8!&MMRFI^m-+7tb4@;v(Z8mS=pw
z`O&rqtahTJ-#BWtexy&8wQH0q`i(>E;!&X8pi7?JBel)3g$K>Q>vgB~N=Il`hYy-o
zy7s`!ik+xwYF4km=IrCoXjCCDZkp<km*3mBo!&GxVu~_FulAE(Ui<Dz#W{PtqvB%H
z^|)u&<_!-m7^i|V-F`j%TbH~lr+%lxGTrvw>Nosk@wa=lTdac8&0o;`YBlbj@Kp5!
z^0JK;*EBis+QA3p*~Ya$&0DduUFAGmc;oer>sNjGfV^yB(N_(db-A-ro-MR(e(Uj`
z`;V!+eRx&nyi76Ze(BeKa-T<jC~zv}LBI8MyM<k5ZXWo>bd@dKu(;~&6G|^qK{@sJ
zA?xN;pXmhUdsl6{`-?-~bdpAWoRgL=?Ag6k=|i?Tw(<3+U)^iS0b87)Oq%{lkEg-p
zO>KrAyw*uNhc#?IZoitvPPVXI`PNU5d1$R;3uo4u`)=W{s~y{D>bVd6#`QbI=jHEv
z;DpO(JGSzM9*wuvon7oCoeTFI_3$3guW(Fp&sh~W5B~Qu$9C5Keb63{pS{|#g{H<+
z{03VuZSQaW&)N4acWmK3*DZK(%s2C$q|ns)<mm1HI(omy=cv=V^=s9r$6R4dIy8Hy
z#Y6Xe-c$+W(V>3R#@!1FepbSGb%^Uc*>5|h_E9&#uOF|R!+xKmqM}<n<X^Dhhu#Og
zyUhv8yFQw9{i%a;@{-QC@3>;nqLbG-wsFciwQ_GBxY5ZL4!pT%wYqz4aBShY8(Ll2
zuyDI$E9VYAsh=8^lV>~Ie0lJ9!=Epcmua$JW{N@YZzs+E;Hsy7QDe95PF(amPU*g%
znd$n6e_g3U-HR7FUi7PnZWr`-yQJ9b{fIifvfdr{w|H`+V;iH^A!>i!4{F)>;_tUA
zFY4`jeU~C><v&+#{C3AVOB~bGbuyjar*G*+Ph3<k&yA+q5BfY^((R+V{>!wV5r=er
zeD?K=lr6mPp&?aHU$aEn%5Q$Ce#bi#Rw`S#rR%Dp)0?bTansa3)br(u)yh8pt&=X}
zesAx%o-Mbp8#Y_n$_~!SxAr(;rV2_^^B_}xpSUOIiS{FxDO0@ewJ%GL-nK&7#-tzb
zgT{B~4JTe!_q4gnHpX>6lBxBV)bHXXC9ZVG<@NsWse3=0=m?Xk!<^=gzM0tMSI2f%
zJab45r`bqHr#^qv{r=eW4bOFJcEvnpJ8gX)Wc$w(E0w7+M+K!m4<x;_0{3ki^YX$&
zepFr@{MI!ur9K^|g7WeecOS9-l(EV$gFeTx{lX=MAEk<vU;2HXAgOlUKCSxT#Or@>
z;^xk?rsSpGe({$4-V??;w(_(}FD`!eo=Hwz%+w62IBZOxMQs;2wzAUI&kg;s)9;Qc
zW@=nA-R`>gPd@0QV>c?>_{H2ChduP!dSyFvUTtvX&2`o*TUh?7hQ}Uu<XV+!Ui;*{
z8(aLi*zwD5Cm*}vv8U%cwlXRAKf=ag`gdYe{_A~qRomNZEpcq&)&)O5_fWx-*uK+z
zV81RG9Qf)iCu*8HF1|OsLf`gz(^OdYe`nN%AGMgEvV~{8e_Xv(v!%**_S$;bMUTx}
zucG4p6`St;Rm)Y%6f57~sL{m3R;z3$?^~ryzx%<4+m2n@{+RET8ylRs{ftlF`a${S
zomKW<@#WUh$}fX{CoHJ_z;EAJc1X98DpO4Qd(Cd~jcEM(Nj-AcIJWZP4o}^C)WBjV
zD*9bde(loTcm1}5-EBu+RJu(58Woh(Kkm}P?#I>ct#n3x|8VNH{qFzA#-WZIiKo-5
zl{IU8|GBc2^uspZ)^O6*Ge6JG%M{)8a@)r;o!{xyMO~avXFGAzO(#9S+h>g*tUK@*
z6*cvCf~J%H_i7}*KmM*z&~f#$zhc>WNAzAa+_9Cn-P!xeue!dXqN0~xUgcrGhwHbV
zTl-1*drM*CIO(9eea>$EtFo21Ja+#FTMk{UY~z1!J9~N28!J?{F~91bz1wyvR#7qP
z=iyk#Vej(=tryqxPxR}}miu3NUGr5=y3Ew|F#QmU0-}H@fC6*(Iq=Gx_F1N~mGm<z
zUNg4rPh)p4a!fPbc+<T<7`Fa_kGJdIV$(X6ZG2_HOI_|bwpc~Q>mR!R<k1hj<it(C
zzr&caV_sj^YWbo?H_ukKFz)LhbkdoW@4Nn5zt`!PY+32}rQVOI(-kTfb-DcR+&o*k
z?whagefrUIdA9QRGf(P$!<gJWTgl&Bx0SYh_tq^=+u#2R`W|^$`>6NLZ1MN~=iS?t
zS~f#vigmtx`|=TQFH~+UH}~2}2TWb8gqhW0zq9AoDC{#{1!cCZlce?|?DLRe`!nv5
z{YST`J5I&LuyH1}Ue&2r+`Q%Z6^>AJ_^WyF@xQNBZ`NPq*vi`<IeP3Hy_Pz{%;*sG
z@A$=KKQOP$9rxV6O!;M0_Y0%;GqLixr8OEYQMU8^Ivd7z-!e_v%DDdSx2bmY@6Lri
zUwJI&;wo*fDOP?ts_5p)i{4+Y;%2t|omEr)2s-}L9-X=O6}J~SwzACQFaG+>XM>er
zy1gfp>3qN6MM;|7^Nre$pT7HS6_!Egt7-d8-3K+@uc-XT{XLDS;*QGCO$T@AcgVZH
ztEd@Od|rN{e(o?+{%n1|oVe5d-f&XtC-wVSUU7MS=Pc>v;~rl8bNQ(woNQs(?`-Jx
zy!JoS<>}6wu)j-YTRqd`ROY9r8@=<^JjYgc{pIR0{Tl!5nBvqg-#@c%hqcNyO^<U}
z|AWuJ>c^|<u5!G1#KeUkcJH`c*+$#H!*NaB@~;g2&ngua!_Jei_0m0m3)^ncdZz9-
z1dStU_79!;R`qSa&2&uh#al-&JnYG0N0^=taUI`H&(EJ<sdv=$!J{2pIlT1X+4mkZ
z*NK{Lx_|J*?+<KQy4A@vuln$eHTN!9<-|=lT}<b{_SWJuPw!toFDOmNCI0_ffr%g8
zGW*TGE0wL>`-g)cJA30?Wr}~Tn&j#V^tu0xA3JaD>tq|_{_aoG+ACMT{b^5C9qU9z
zecq7%plPLR54^0{5k{*6`r5*F({9Kcx_Gh^7j4lcY@DZ8Y_+&i^{<_%=r@k2zei@u
zU#5QV58AG+&*yl>{o0F9R{Qsi4;-QBAoIX`9r)dUbvylcQrpsce{4x#w`j_(&gm0e
z|31Nr)~~%ad(uQ@nqGSO{kv&Nw=?}c4=)}4mdEw|G*j)Ey5C}(-%P#loay&1{Lb^F
z*y(?FAnf&oss8wVem!Y+NxR=?YJ9}+9}09T9#M02o8Odap5ErUardt*n->+|Y<I&2
z$6b+=XDjb6>~!lLU#xP%van5I%lw+DI;o>mxt$7|6{bq<_&2{oe#QLCshons{3<)r
zwNtsN+$O`Sb}Y>A)wyxC(b3KN{pa1?;kZ*4C{z5oa@Ws4{d}GhiViXlqFV>i?EvCQ
z6c7bOfoK%C*Li$!PSeTCjnT9sUPJ*=Kok%KM1h?Z81|1t2i<t|Gmh<?cTL-_D<2)D
zlFshs3!l2`!bvKr+<A6Hm-^<;)DZ(N{9Q%Oanr{<wdDBOD%<Hb5>?K;v2yPN*E*(I
zsrG;a&o4LAiHlzSmU2-*6c7bO0Z~8{5Cud5Q9u+B1w;W+Kor<jfmgp;f6nq@%T!SA
z>Pd1$0Z~8{5Cud5QJ|zNaP_+z-n?@ABu8kc4$R+Y_g?z-T_4X>LeW9yfvkh1>;vpi
zLQx=z3Jhrdag(wYS2{w`A<6R~iEAfJOe>DKx#~|}@0%ADD^w}I<=-`CJGS!PeZKwq
z_kYh-wy~b`TdA9-%vQFu^zJ{de&~bw%2xjVXv6Q0E105e<KSWQ-X3=AWF-uyLzq2R
z>WBiOfGChe1!}(g@53&-WU5LEgMPmuiJD4GB}ai{A35W6_1seRZNX7Tlso!Hq39s<
zz@l}K^quP5DyKBd%U`90W_4g5#C0A-dqPn_6c7bO0Z~8{ut$N%+UFl!ddLE0ibsDn
z{(?H2O63W6*5Q=?Pd0t3(R`H@+B0J%Qc*w@5Cud5Q6T#jxaH?N_g>#}vvWej|Hs~a
z2kbfC{o}tg2tll>5yY-h<*_NwtF}fI9V(O<eMHn&yOhSJEsa^NnNZZIJ<l5{V#JO}
ztPBxJMD~vSzP`M!{+^HT@toIn-*+N7|M7OL&+*yE=RD5yy7OM+qdUKr!zsNd|9<IT
zU7s)az!GyFIHqqE<KMx$?hjiow`Mjfhwb(r``!~fj?H1}J=KW4RvLEg6QlDKF1GT9
zcTdYl<q^A`dd12^Hv1s270=n})-hXdIz0cTVAOxvo;QD=t=>$q_|mT}aC7THS&5|{
zz72j*Z+v{y=l}P1Z+)`t-9z#czZyL}d&z~TU9s@cY<|Idf9ItBD)I9=bYAP>o1ULa
z`AQ9x8kqYUc;?NIr(d<_;H+M}d)U64cRai?-}yV|{p8p;UaR71qYgZ|_nm*s;EsEZ
zx!{Eh-^n{1c+}2IT{J8!!nrT;H+zSU*Z<;{?@WC)`?}&c+d8Fmr3Ok3lp6T@Y2cIx
z*ZT1bmC<>Lr5;K@nEP=M-zzn@{q@T@CC8-(N)416C^b-v2I{@PpI_$JB3n79)WE#f
zz;4GMIp*d)-^#yjIIs0w?l*sFpz+-H%k2Ern7qVN52YWJaZu)g`trd1mCZ|TN)416
zC^gV>8rY!ARm*mJ^4`2wY`L9FN~H!$4U`%vHBf3Gt$`Ceopt()+YZfZ#o?>{_R-3u
zd-4+N)5FwZd!09Umx)=4UyUCAx#VLHU%K(Iyk1O4PJJUqzi+LNS~<JaK&gRJ1EmH^
z4U`)ADmC!dQk$Liw_m)Sf77u1I~4Qg?@*L_{%ZC7b&uOsj+?y3-+u6ZKDV&e+*<xS
zhV%EoV_2&l%Xy^+N)416C^b-OpwvLAfl>pd21*T-8YneTYM|6Wsew`hr3Ok3lo}{C
zP->vmK&gRJ1EmH^4U`)AwrOC_JG<;L<!6KQ5}T(7e)C4<iUpSG^hwrK-0nArop8kD
zSF*1wwt8*!O*dUJDucWK`Mn{F>@zt_;iwDdym;u#Ph};3GxQ)1zHJ!)J+kJu_t^j5
z%i6yAOqRm<@76Z=??e`@U-i=qwma?p=^5Os)u3m8+G)W`#7$0|d2*X4+f{1Clzc7S
z|84WfZ>qd|_UyaOoqc9!&Ba+0Ti<-uOONK?cI<V0uY)eT>zY>o*I&A?)p^~{>ve8*
zZm+$1b#DFHf4eW<eW~utbZ$7iSNCN<Yp>n8Rp(Y+`YwM?ukP2LzuhYRzo~9k%2jHh
z)Ih0$QUj$1N)416C^b-OpwvLAfl>pd2L5*qOkC{JgFm`)d|qOyhyTAH{O?{<j($}d
znDyFr7mQeLRQ_$luS(zLzF&<7)_MAZ3%5M)y}ZP)Mi1>aeD(07&-@^-6@NTviw)m=
z;+4E!{O3i_tkv!ODOs)f(Tq119emkSS-m*t);_Hlo;WgZE}nD#3$J~6@2IR^Jacf{
zVK+2<oV66*df&l!Y;)(-?5l~+a@SG=r3Ok3lo}{CP->vmK&gRJ1EmIP(Lhu0ZL-t<
zv3r+&2IMV`t<GwH>UuAJm^T$`k-3~xYG8iVK-+zeoY>{&5qXKF9=^$bu*qsG9P!k+
zL3vAKt@nfFy!!4_zbbvruX(!UtJFZLfl>pd21*T-8YneTYM|6Wsew`hr3Ok3lo}{C
zP->vmK&gRJ1EmH^4U`%vHBf4x)Ih0$QUj$1N)6Paf#at?b?uBL#^)uLdYGU6pcY>)
z=ad>KHBf4xIT~p0@2)pEey`I`-};@rUc6_kKP}sPn@L%{c*BNU>^NYRnORHY*-uZr
z;g`p>uGEW7<)dD?Q}a7svF&odpEEnF7n|cV<=X#f;G`x0`NqFq8J(B-?bO46JfZx*
zQUj$1N)416C^b-OU~X&Ril=%!-1Wwf@)Aoulz#B#IGEe<_jRTJ_l19P-!YXzd5NVS
z{x?79efTRcZ2Zfyd9Aqjahu=VXYieQt=RLEbyuAJYM!NVzavNf_<>HZXSL$*hOV*P
z-)6m%Hx=*ef68e?K72eY!t-a`dConD_Rea>ImcxW^;oli)>QoYv(rys<4+HifB)|5
znsC2nd|$fEjs2GSWOV*b$CfrrbAQKPFF#G)&py*<y}M-JZAa&G8>g&!>Q&btGA2(U
zeo8DgP->vmK&gRJ1EmH^4U`%vHBf4x`5LJAyCn1L{k#A3YJSwn*z10>;Z_4a$-izG
z<tW?#Qv(~_bLpu5>wlD&`0dcc|FnMDDm74Qpwz&;)4)eVXB^n|sB!tX4d<PPTXv`B
z{`*$-u5VfHOKPPC=B@^MJ({oIYwGZ<#8MAU@q_!G?X<vdkNrE3Snt1k_)VWbccZwZ
zUTUD!K&gRJ1M^w~?f$jqdxyQ-s!}gr^1;6M9kbGitfg^Y%U|wSYM|6Wsew`hr3Ok3
z%zX_^+I`to&$!^@yrpsObN=}bTW#_9Pd9t?gS^D~OAqtP{^dTU21*T-8YneTYM|6W
zsew`hUyTM@`uXAOPVcL+Lb+?Hfl>oa(ZD}`{lVokZ+RpCx?)pgF4vSAC^b-OpwvLA
zfl>pd21*T-8YneTYM|6Wse$=T0}~g!^x%(f9G|xo&TsOS{FE9fHBf4x)Ih0$QUj$1
zN)416C^b-OpwvLAfl>pd2BHR*|IQy?nf=oEya*qk@zaHRv>BeyFSz;aHC`Clb3#@t
z9(T@Vn=RdIK%T;e6@GB`?E41gwc?u%|JC#8#@n-6@rCWzd3&3mT#?s{(OzY{)Ih0$
zhK63p_d2NiTAkPH{Qb_Y&h52Vug<MM`)~KfyD!y!na&Mo_v*gvXYI8+x9Z%gOW)<s
z>DB$(^S4{2f73Phn-!1Wx95;y*^&Ef@Y-bujL2TuVdk4pKKD*`^yQ;|eDgaav&RnZ
zx5u1+Ovrk?_u#N&4<1tOkMqzu>!Iy7Ir848vOewiU+a@K`sc^K*89l^_Z^+>eMaxI
zuI})9Zl9a3y7}%qH)yDszwpjiJm0s&{<-!A|J-2ddvAZSx^KTdI!r(9)z>oPlP<XJ
z);&(TC);VtUq*KsGCSM)-=lj?=`p^_-<n&TeZ+%DwyH!r=Ama_h^~FerT_2WJ!a>B
ze7|pwZi%1Usno+)tB0m%fd9Gjfa5Mb@cjkf&H1^u_`Umtp0}KN=8#<fzSBD%`&j$8
za^u#CCwU_FzxZ6wk=OPgn8^>?;!YciTlmOY9p3tQy>ZoeKYob=w|efak2CQqf9a2H
z{Se*Q)(^LBee<FVeKfI}Z_G#ke`>RA%`X~YpMURt{{}nWFspjbIS1q~e$0R6eSR3v
zul%~6&n~g{<ZM)@XHV$9>p~Ut__cG6U-F|{mR|Y#SJzvh(r3Uy>pbz#))nnXruNz4
zkU2|Lj+oi)*0UB|v~tFKA9eZ8pyexLuiEy%FWhml3V$;npZF&a6o1;*i=StD9dq+9
zR+*E1@3)l?`i*H*F%Mc#fA&4>%ux>?(7s*8x-t9xVeHAzw6EOp+)oy2*JshnI{%uv
z-{2k#R;;JDi1b4baEFHm9=G+uGc)7p(Dz;K7Ww^hYb;(FIdsuEzj%Iu3jNaC?HBj@
z^W^Wfuj+L~?>*1mu}hoEdY|0?`k0XmRoB_4k{#4@`9r5<zixN@^ILalUD@umm1eE`
z!L;(<_p330<?Tmq{{EbSQ$Mbr56+8+dK`1^GbfME-~aQTQ)iz%J~!`y_3vN&{+T1I
z?E@wp_ru9sjjOhYzPRX7i%*-FqvPB`=bn4ARsJ$(Sk7PLKITF3Ev}*SNgLgWHL9bW
zHkUNOANY;>Uta(Ijtjl=boYC!^@}|G>womVVV(P?=5WN$f7pMcNz-!qQ=5Ni-#2c?
z6*v8EM9vQy?-+R7l_#%I&2#+H^G@8KUq10pR#~mj)Xcbj#f^UTh375f^1HV6{1zX^
z<x6y}kL#?rE{<ETA00?;`-@L>)raQei(b&Bcl*&hzU(7D&6B5x-22bV4_;`+3beoT
zIMUaTA8mTqcCI+LjO&N`p?*kS`;m)0sdf6TgXkL9whz6UZy)348`qyg^w@!ZA^kQN
z?Sn6R;z!~hAKKQD6Y?X-Kb=1i9e%7W?)9T<Kk}oiAK&yz-~0$a>OuD`<U&u}&?`QS
z<Ii~9PrrE(9ej}2e&(?QzR)*LfB13E#(sK&=s@f7Yo50K?TdeTAkH@rU2XKlxAVh#
zdQeALM-TXAFLk5*?%wH?wI3Yu-T_Nh#toSM&NWXhUfE~0VKcw;$`Td!w=N!&_QMA~
zrqFzK6O4AX9$ly(I$zmwc>8BZta8GX{J;%1xM#tKCgcbIX4JaNZ9chL$C+=v{k7vb
zy2iE9Cr{MN*R3sImH7GnP^pJf51-u=*6IiHw)`OfJAa*1`sHuy<ay{^HZQKjSDYX7
zzWvR&A9~IuXkYo(xdQb=`46I_okIDZzT{W?%q{es<eZ54-oErszvjvNMaT}&IUL*W
zsi3;cJQ&xjqhY=C%!lN)4|I;<&wltcPko|}<nM95{oLoF6VZOqJ)pMd<k*jnIyHsS
z@A#j7b%8e25A~-|oER5>>M75IkX-mCpEkW|n-8PDtv9Y6(Y>O3NcWDn?(x=g!~fB)
zGQ9ici*NE$Lq+}Ro>Tj&6=puL@ftI#?>F^Z=bqHPrh7~Gv*gily?%1JzqP-%`xN!8
zadgO`-#+erq4khl>Sf4Y=$lVIa=GuN2Yjsm{bP6g^uHIW9J<2>+wXo++v>eKdJw<v
zz3CAh^4m|_JZK!B?2RtOC;iwTKhQkm+V)`&{q{9q8-MoGZyv<2{Uh2BJ@I87yXeOk
zJDSHH_C-&;@I!6>3H3w$@)nH!{7S#|-XHKod>fZ<<yCsH4qeZG@*_SVdh&#RZO?&m
zy>`q?_Om}a=<26`=Lo*ZfiAsUPfw9A{Ftu|t*0;ZAimABPvjrp^hO@%B|7|+|A`Ow
z<ah2-%}1ZV(u;oksIQz~+UgVam~$wi`Ox_k>qY%gT@>jXkMz_R)}gO$oqqElem1*n
z(5nkHs!O4|R@;2{E3w}=qz8OL=Q{nR>OSk8>(F}gsB_JS#_daw<TsArh~!hp^F#hB
zZnTX@v=2TZIq<8Wz3>Z-)4zD(*KwZt<i`j4`l0psvJXFq^U#OZu@kwfbieqHM89}4
zA05vJ_>KEoPtWp`d?L@Nm)Vuvk*@v}$^#MY$Nmw~!+%R7J<u0@j_kSZfbEZ+nU8v6
zdY7M%o1L=@{fakwalUEeoBfT82mR)YCw^kSHuSs#Js)~rL>E2z!9M8XgWTwwr)?g(
zP&^waZ>r5-=oO+z4*iF%zT&03-90O3AL}7I-~%1|L-Is)-)cYe$sOmX+WZ5*{8PMo
zFG@c9phs^hBo|bFm>=okKcyG@LVg4JIsM2h<Wi5+i}&x|uknEa4UKU2>mQwc^tgo@
z4;|5Mg&%G_qxv40JcoCCbK(6rn3Lm+JmL<DSAIf%`nOIzpr1nVA->F`Z~C_nf27aY
zw%$5?@dxv@$wlAx)fUhC`GtMRi9hq~V}JZ=v$r-nke*{(9q1f!o~kQje%E&H^CNXH
zJr!YUUv%-K4t2hl`1yC1r5;K>eD=NVH{B29EvVjcUO@R({!7g>Zk;*=eP}(p^0xC7
zqT}3^{~<cssd?y_ANh#u<Gi@uKIUmhIqXADa%np+=*c-0_d|z%Bbtwna}9lnj`{eo
zueNdXp?>qAIwJ0iZbavo`Q$+_wyn3nb>{2GC%ZuN=qL6^{lxvyhv>1p{<uFm*gg7(
zb@;I!ef|7Xy{8>fon#;NwRWU$J^qaAkL%2X=BtnJZ5$nK`XDz%2kIw3y878M_D8?K
zKfSOQ#Fu%IU*oagzW9yL?O(UHx>J2BK2xYJbuZ#R1Rdk*R;VAUL#@+xUmEFKXTEks
z_cHdypEgv7TW^1D^5esOulF42;71!q`sS&_-3Oa*Uu}Hj!~HaV%rg%4+mAl*<9&$z
zpywpdKc17&HLmS>5I^yG%RF?@bzb1pJoMr|?!EX0ey#I73gdeF>Mv?XIqZi&^2j&n
zLFb2Y^2RoP$YH%aXTEXglspRUXPx|{AL@tbYeRIb(>87$x#$c35y?q!+7anDB6{Z2
z2Q;7k_46<G(}wzEoLP@9`NR)?$OFG_dH>tauk#(M3)CIa?^E$0F4P0s;#1u68{_C1
z*S3z|@%uQgjb23jlLwk-AAC8NoHO#Rb0o%%w)lb0QR~zd=!-k+wXM?+oiFHVr^e9}
z$L6CC`M>9U`NMqsLGO8eSKvDbdBQsK;)C4Q`@X|>I_e7hIIki4_|vZr)^ELgd-XrR
zb}vY8);VAG;~Uz~cvBI*rmjz|<9E&_$iC2XKt%cx7f@V^8+ihXcXo*DtdHj;dx|?K
zp2cm%xQ}sh4DB1~YDYu|eSDC|e0s&dykwl5*2lJfbW<ok*<1WLfB7rq_vVXx{%t)!
zF%I?HS6h8<ALHUY_RE*@i8gfK4#lH!^t8!CzEu3^XAkv=xWzBI$dN+(;fMb0W88Y<
z+W4kle6R=qK}Q?^=;(*mqX)?+@A4CRWDoX?<Lsedp2i0{<cIVb_tkH``Sgho`|^YM
z{hob|qemX|(PIbvJ6Ei?Kf3g9KDy@XXHSSOe$4Y-8#=!K!l&nE^C5oFqc8fgFaMB-
zAUU)l|FaI?^u@1ISZltDc~|_&tH$L+ZFyQA#J9T7xV#(@eSA8PO8orq0ko7J_$R-F
z{EPqaGxl>XvMW2qJjTEIv3jMcNbXpt#D4dh@%}T_AL}OfzUD*wXrphvb<n*weyvl#
z`VPW4`r7DfqpKf%bYs6ZKCOqwtyeFaZ#*J?tiwmD-+t=%NKczy&_iFpdB*i?+s8co
z&?~vDi}cO69-0U32lZQTpSUmnj9X_P;}E@6+q#J2Rh)`X`CD9<`1#-0n^$^}FQXl_
zqutDRj<TckQ$OTy&QJa5IH!#}ucM#H^Z2pe{>JfbJ-*DdPejk1k*;~<z`yz+MOQyW
zHzN7jn|$~;Upvy%hLJyQ{EJ_5p-=BoU&f(*@I&ADkH{|MwB9}tpZJ2-X`2W2%Qx~@
ziJ$*nrLRg4o)gtso;TEAp4+t5cl=-72i0+Lo%Kzv-^z8R21*V5M+4$X9v8>r(EW>f
z?uFogD3$;BZP0+a0LtG`zjI!_pngz~IH$DLW1j!?t7D8i|DpQAzWkOyJ4cM`cmBlr
z_#}^dMw>k3i+ti2f9e+dqi=uo#Fu__@TcFt5$6}*BPY8=y=gnw<s0>a`_1MeeWnn9
z^Bb}|J!(VG`;b48Q(ktjMvwBidw6vkKT>z2%P-wixHp07Z)lx(aKCWE8G{bmc=LuY
z{XW%xk-qyVaZ5hWE95lJpJJS#Z=9X|ZYqW1-Fk6{5Br)YUunm6)<={ViV)xS=NH=S
zXC8Xy>4&L)<LIJOgzRtL<{$Stt#O&wjoL*R`C%`7usft@bn%T2v>ryf&L7A=;tbk{
zUA2wVKmMS3_%R=fBmVDvbI)L%II<31h%fn+f8ql=_uS(_>&QW0*5Swg^bP6NdTse5
zwcb4T6i3cYNKS~JHof5wTF<`bX`6>{?TGf#j}AVx$pfu}=F=<vLH16e_lCZY@xIS{
zK=1p!4}`v>@m)-<XrGAQ8~TpPzS`dZnOEZHe|Nmp!`E96>Zn*(=`TX(i};tH<Zo@~
zsk|glL-|@;-4)Tf67xCw>NM-L<tceh-ov-^$b9FGeeh*}>+KuS{>}m8@}}nkdD6bd
zoip@@Pi^}{{pQ&h#&P+WeC&cBd_eP!n-BG)=e*F5FLbogH{W?~opETr@gl?zf7G^4
zo1V<Gk8%53hhOp=M@QTGTG77N+Xtd!o_@%0j7P+O?AJyg@>9=Mo>QQ4&neKj{Gi`?
zfltqM{(b@U_aDTo_!CF!DtwzK9>g;}s!Qbo@xXrcVjTaFU-}*@=0*L!r;<O&%|7&s
zUc@-heEKpjU#g$z)%Qy38}gHve~LeH)0=tpK)>=6y_*NkgX&T1%wwl0SLBoZ=+kpP
zI{1!!>L(95*_j`+7yH^@-cH>+k<UKb&_3)0$qTLLC;UL$I`hyakA7|Y>o;FMBcE}8
z5c$!r6;pbeSN;5fd=dEp<bT%jXMUugKDDin^sQqD^X!vqN4n;T7pUKS{V?{68}=uk
zd60bg=NHD&=ZE~kI`qwl_M<0s^lPV(y!PS0aew3DMjId65I@j7XkX)yUFpd@>&&A+
z827c_xOMbsAN^22OzBO(b<Rus;@3QKKzu>_S!W(Z2g>)gV&uy{`q9PDyvir#KBWdq
z4SZPxu?}<pRxirq^13{!-+2S|L;dP%^;>MKkM+mr5d9Dx^CFrD&9{%bGLEa;ol}q;
z^kAMm=v<J`<zxL(ALyWqo_Z4Eo8Iw5Ui+m`{)h53zM=ZZctjp%K6;)9@kd_c&Ijx4
zV_xh}okz|`h<*z3AIJ6E$NuUt<JKA153P@NlKUondOm^Hc`k<LYeW6^)ke=eh+l2|
zXp;*c_}4biKF~Ofbo58WH;n7Z0nNh~d9?9^PDIbu-UDby^!~x$dG>eZVf_2zp3gn6
zyHAiupuFpO6?#8m+&ut#-jnDzABsPEpr@$s*iOxppTwv9p?(w3@{M@)94c?gpYnjb
z<(^5tHEx~!1?5Bg%9k+Jk6-*%{pE|l6CC9yuXv|Veq=tn#<lT*j&>Z6e4A$<aTe``
ze|ge;{q_|v=$L2ReEn59zW8@g+28*3VV!<xT$`TkUn@pB?hWNB@ogU%*V!M^hxstY
zmv!bvG;UvdN$JykdBM3IzweNj#0`1T6>s85Tl~0Jv)(@52gxJ!!S2SzpLOV~KVp3A
z_kEOg=+G}Z_%@E7crb3haqT#6fAX6b^NMxi%sg%R#eVX&e)iWk9{HB9=pP^UvrgN%
zb^4tH^rSzwy~n0^a+)8}di!aU1JX-G^kTpK2FW?7vGIYGmRPFd-$}tIJ<7|Gp7qX0
zaiMKKJ@b2dgU(CmGmO7erbqD^_xD_6Uvffv5eL>g=d}4LetqXA{_yL&34f;xM!xBT
z9@te}MEmj&^YNvPe{lfG$8Orjz2|}KOipq`^3sp}BN`{Se~;e3b4h-3lb4<3Wqk8*
z$p6TJKYY*^dG*tq{rL}m>Crm;#dhpR-@f=E7e4V1>6<^<AD<9^_|_IL>H~D?$N3xi
zGfy5PAO53!{1W0D-N+~T(Xo#<q!(@b`};cljh*<1bx=G<f58_!7}t+~a?t~R^wXF9
z<2-zrFMq1<Q^=krj_x(3$M~$&L#c<B@B?+9`Wt#ak}sgRj&UBx%@gPHihQ9h&uGgJ
z+HoIkc|~4=sr*sY?|kqbr#uYhY5UmU`6AD&ui|`i8+Wha{zJdGCO<vW3w_Z8G>_dN
zyFl~eeTQ>UKSUp*1JP55K<BJBzlzWGo*Pr>`3*n()&A&Pk3Pgl<dZxQe-YW0o!QAe
ze!#xk{2rgio!jhef5^V(;hR60M{e}t^sdKU-eppoMsk~P9O_5kzWBx;zYrJd--zhK
zIL;s05t4&^_<`u*Pn=qB9zJ87um?KU**C67$2@30ztT3&AFPMQ?N1N(HLlHW)<ffv
z{i3|~p|9w_aX!DKe{}7mpI^W@ZeR21ogeFu`^C1p-a7omdHU_^IgGsK<5xb8{pdh^
z;-9{ZM>G#wPyhS`nn!>7tz&O~U_a~3gT~S0U(PB0Fw)UakLW@BK<6Q(SI-Ib4?VBR
zn@~RW+`=yS)~0{^<I6aCpn2?$p8VxKi1!@+-P`@H*=O|TFD_GAz4h@Q_PA=*igEW6
z&S`wRXW8bp?4;8l{cdHIorjz@V4W2z+dqGC*EMckrFtLa-@irghnM$kIAYuHR6OTz
zwNJnE_HA6ULND>(-_-WLMts<RRHtW8=)UVhm4_Z4{lZ>nE>Us+dDIpwuGXt(hYI=0
zP2cv5^Zk3u<YZs#$>F|;p4lVX%l_h%9o%=I=ROE}j`e(N9r@Yg@W*>^bj*1RSJ?5e
zGe<pqK>K#pfB)Hilk-TtL4Gf8Vm!#t<crAfQfNNEq@U===BaP+7xRGi=F1oKYaea#
z$6x3}Tis+o_x<|GZy);kwjuv0vF9yko;f5h(u26-Kk6oa1-++%-lr6y=K}Nit$y_R
zwK#(O7k}bW9FtGK^UQpCPrO0<kykw9%X<7l@4?8UpIqY4zW8(wJAcu~HzYUy%#%;-
zhhOz9xy-W;zw~GydNUtg^t2&<=*PU+Zy){mq#x_hg+<T(O&wRCI!~Q*?BV?6H_mZ!
z0iDC-lh^6JUc?7KL61GqV@H0<ZuAS$Ge3p+hy2yP+T@OyqN_iyXK!`3dG;qKza%Gr
zVHc=A@IIb@p|2g0pV1dTwjaLiqmDu+&d0xTbhIPdKlalPd5!DWj)+fa|I|L#7tu8j
z%J)S`e~>-cm){!aCq;B?9hZ;e{Ah3EQ9f;Y#4oz!Ko6bfqW4+qW%jY3ynsLYvo5v2
zc$WWD^Vq?B^72RWLiqq6Df;v$PPEOBeoF7=M}F~T9n|l6g#PfC(vyDjkPkijhWIhA
z9nt=Aoq5)a5B<>jAM*`+n3qEPLwYBlc9fgm^+SBpr+H9+mG3XUXX97)Gf&%meBswT
z{o3?lU;G(ImmfP%ApbJodVH9NZ{y|{A%DatdB_1H9emjj@*i~Ldi0H>gFZgAp}eB*
zh;+>}&i?jOFL|zK59`o(E|EvSHhrV7jSjlxus_rf+08n1-A~}#y>#S*Jmgll^C$f4
z*Tzq+sP2SKt!v{t`GDWhi@H_6{G$!|k33@@xzH0Y+Vt%n!+E9drZ4j%8fSlc)z6RE
z&p5l-&;6I@EA?=kZy#tJeRS+^9)ICi{IscvUR=)(^qv}z{n}Oge36IoOYh{>rblu@
z^3$)j_15u|_&a!c$v%g7dvoFaH<**pyy6EP4mftfM*R@~)<xu}>IZh_AL2<J<#~@^
zvafjX9iRLlFPM+6b>6d?ryn2a#eLle`%d4z9{+^&&R+D)PxOl`Xg_jkdwxNe{n!nC
z{m}e~>|~v`yo*o$F^^hTE8-{mU#d+X;@3QUKz^oP&~^^;J8eil>*$BP`XPJK4>XP*
zI_7KB6FnFg-}sg%pnj-dUcm>x(X|i0?2n(g4t;37dFa|lKf2oXgHhl5Bcc!Si4T6x
z?)stm)|-!C{rDs&`)Q*?zsB)vo;K8PKXkOAerP?c>MQ(Sf*<o2^YKS6d~2Hr#gqKX
ze)LEVbhWLAv0odK10DPrhxCFjJ6VSgKA}8|Z~D}ZXxw`I*dH2??UdiiOZailkO!TJ
z<RT}#>WB0c>ES=h759mC<^0xWC;48RUimA%>UXYN2id{Ai0TRR!~xU~?TZhHu661N
z>!5k)Xye!WSaiv6-1lng6Lkyv>I8D(2R;4fJ3r(TsQ#Bf_?z(}L>Kbs*pB|J&2Llw
z8RJ14{fPF5_JMJJz3tTg_M<;YK5+~6i(_>0sm(4C%}eobzIjl;b&+rLjC+pM{;HAw
z<WKQ}Z_gLxD&m)a7tM>~>;vr&@#$Q1j_?Z@&mD3ZhtXfOAwPoj;yaFdQQp8Gy82-%
zKgoA3h4etb^n@>ZR{!A}pY##=F)m-myegmChrS~}_=;$rb<lSr@;W^_Pw=5X&T|gx
zPvy_D|NpH4cBe;r)(>O9`ObCd{y^J$$p84Ie4q{ar+OfwdHe|KPoeq%J?%IjJ^qeQ
zek|_fy_grBFXBKRhWf3Of8}lGP2|J=5P$f`r@RZzw~u|@cR}~A5xrl+pZU;yZS(LS
zksj!ezVNAEo4(M|CWrbBk`o^gUHa9wzjY9O{F)c}isR_w6W{p54}H<E`#@+OI_Tmn
z_OrMBVB}YSM06vfYyYYqzIaZ@C!|07qGugC`mKvdK749hhd%v7_N8AaUd+=*KgvlC
ze4(q2Zbb2bPjq5i+{#Dtp7T~b>o*S?Cx>%ie$sCrNdLyw4eUd&^kQ7wI_)?=_Txvt
zeInvZzc#zG8#{>;Xg;(ryU{-+7yI}gN838~)powogZ)xS@Afkff9R0QJa(pE{rm%6
z>y4vhJ>*Bwx$NAA>OgYF`LQ2A`r~^2_R~fe>NgKv=d^kJo?PaehYoq5d5}Mr*!!%j
zJG`E!^dMe}P@Jad@KeZs?8={E9A|g)*&AA?&SdB4561b0_4@5&fAy|5y5{S*j(qrm
z=4qoVZ+nh058uYo*ESCw>(Phe2;cMv|KCps4Gl&1esqDE-y5^XLXGmKzq=z(%X9n!
zJ^iV+{uDYN(3S5X|BH5zALY|1H+qnM=}Eu5tsT+6*7>~)zU&{{=#qn;Ai1^8i*(Q@
zFaPjd&9C@F^nYz~@FQ(>wfUua^oB3~stv6(A760lN)N8Q^$D#Sy>EB^S|9zIJoqv%
zh4cjJ)qM7V>U8p(XC1!r1;wTN4EIphqpwXL^v2HQSD%P8@*4M?A%4t<<mI>gN?nT%
zxzMxD`CcpH13mj{L-SIIPwS!e{7FB(h(q=f5BmA7`RqVmsGmLTqo029k1uGxcF{a~
zXBXJ(_+AHHcGoqn{%`*3zE<b&cV549t8;tp)vI&s&;Hwe@$O4?U#4@z*}b|i`&oPK
z&aFDP>e6@lb9!~Z_WbQu>EF=M`?SR$d!qfY{F>eO7%}~(N%^=BZ#)0k1IAa|drvri
z(v+W1%p+d8|HbY8d(Oao@Lzk(-hJ%Ie91@tz0KRNjIH7e`;Y4U)>$9taB|11u3c!u
znN_-Hf843k^{Ll#c>7B;`XBSdck`|n9X{tr+fK_n^ck>vw*v>{FK+qx65}5jmm6QJ
zUAyPzY&SJW@2>o@e=aa&dVX%}#k&sJcuW=Xwaa~zwmWmw)O^z&&OP<FcTC8;t+L7U
zk3TXbfB3r>9CY^OPv+#Xk8$+z0asgghtY@sZE!w)k^N4;=8X~Auw#Y{9JuQ<8T$0l
zrSq1XO}zKvymHA7yYGGctNAC34(!`?-Ld&`mt248b5lp;J9YeY^UYdMtm>!jKFjZU
z`d_Ey7ccqpCOx;DUga0vR{3f>ZG6x8{MSEQb;!>y8eP>3yS)6`y&bo{ZgQT_+U9^$
zcbl9)deF{4+whQaRir0~?%@XyeWdlJ)AHLceEEu_mKc|h8aV2XzWWTXw#f}=@A}IV
zUcT(3JfeN!Ki7QbhKHw3%Grw_7~f#ir>}eFmJf2c)a6&Tp0wbE>i+D(Kk;S#gCC#P
z^TnM;R@bv5|K%^+_1U=lYr{uX{fQhA&5L;5DTlNl@QXXXWZ%d)IoTcJ(>OWt%b&!7
z@%NV4^u@au8J|afTDlzzpL*MyKRU8Uc4kAR|G87wx#ZPns(yk0jSsu${$HK^N#3#V
zFW-9PH}6;Ti+*wX%(}BT-Tq(0s`=!s{;Td#Ie2Jp9=+Up*Q{6f9y}($Jp0okyIuQX
zzRj^a?X}H!hE)C4IQ#IQnVTH8-_O>3CI8dVUdw)Z`fC|~I{C`;PksE|8#CkV_J_Vp
z9dyHSA7$Hi-~8xn*BF`I-}k6K%S?Tz+Sb3siGN*iP@i$xy?6fpqN~pNpgL}!W4iac
zcd1+6&Tjn8uionN;-%TljvWqKV(BY#$Pf6txLo6iCGvZ_PWVzjl|LeiM|Aj=y!cYX
z@k_qA*pnF~FO>K21I1Gc*|iASA%*y@6;HYDpd(J2Ju$Br=_8`}i+EzE#d@7}<3kyH
zF0k&q8&7-x<*GkuuXWh|HxJ(WuAF}P%leBx^yK3E49<@FY458q-+XxX!PR5$|MOKt
zGWm(#rVicxy5VQPkjXdfGIZJ@<FD!PVl~gnTl(3F|Ji5xzjmJ4W~Bic+%!Mqf<Bu~
z$?$`3am(+Wi~Q3*#&7&->-QFZWMKBpDr0}z`^iyNy^z;_&^q$br+x7QNA(~4!dg8>
zSI<>?r&s$rFXb8a3;%b1JKrXpc|)(qcAio_SM@uW7Cm?1_jYRaK{ekxXV8P}#y<Ro
z-nu>Z<ZA7gJut5~kA7R{?>X$A<#*j<MAg6PS-s@EGtYUzZsv*SR9ihV<ks~D&YC#6
zdJduQJWvOr@0_u}wmilke%kt9%N%^yo7H-jU%*=J%g-aq8z~fDDa2<AckBJYs>}R&
zSQfGOu?>S)?>H)p=zKi?mxI^uG=4(X|K9sMtyY<oozijN)BbVx)NF(G9(a76hsIR<
zyM1@fk5+kMWOm3M*S>JVq4#DpSLuA^ayz|O%_EEcuHTNgEIB57YD~kni%)nz+kfg>
zA20ZuH?y{{FWK$R)n{b(nK0|>8^^vhJ$rfgf3ADP+S9Y2zjng3w<nL!_CDf>Hfz5-
zrMf>p=pEg69_~K9`|gAC@#_s}JpO|Bs?P)J2>DLk#@_r(f3&}PMcpCq$#b3q%%g`3
z|9r~6e|+!p?8oOkwC2nWUag}1`TDj;FWjpC>(zY7{^|*OlF#K)=v)|l$-W)${`8aV
zy$jcWd+1*#WtVUI`*rTPb8O~0MqWop|A}WFa?JjJn@~Ns9(^Y}_`op}t967rh#lBb
z8^3$Lb<^gzb^Ew#H}j0EqpVX`t}$|<{vQmTT;<m}gAa81BmY?Dw)0jz`otL-{>e{n
zc<sSi!wJjGuIfd+dahF!LC;z0Oy}<oFI9HBaFc)M;*b5rF}>2W^OYUVKVsc0AAR@3
zewlgb@dNd$b5Wc*7wHjhwC4F|?7RE;jJ^l_>c|B;{$NBk?)V2gigV|%_4I6?>$Y9`
z-xvS*{ml5BYuEVk@;@C{-A8+eS#y54;tk`g_G`E6+Z!zPqoGy5JM_3t?;JVi?OeQw
z6ZzNks`^uYRwp`l4}7%G>`%w#)%spssW<4Cy!4=6+UKElC%<}I-%NcaABcDLIDXWf
zp6fiHJ10D+J6Gfj{6hKIyj7;{_{pBzjH*5d%NObs`9wdxIXBTG2meH0-Xovq2>Nu6
znNQ#3)picUtp+`R<cBBSms`)@)wA*uyUM4|U;K(c<DP5CL2r=#_#=IYU-_I}=#8Gl
zr*-t9F1L?<{wdz&ar>E%E{t(3eyw+Is4L|KdO_E7hI0+FA35YJdee`Ok$1lH@}lhr
zWoQ0%fu5sQf2n%zhy&*(yIs=iFH0}?-tf#h?ztBJvU0~_R~-FZPA~SyFFELyoyC`a
zdLR!!B0qmJP9FNFkBIg!LjGBV{IaEyK8nyiV-fNvbuqf)PCU~)yNM5Z>o<3IeB*}O
zU#i|;qBpM)AD;6d`Nbu9A^XS!@@GW(CZgxe5<kD6o>zL1&m+p4DU_%AH<Tx0UgiJl
zA9apABY&ya)Sc={^^H16{dwAbd!9S#h*_C=>LK*iBkHuPH`r^`dpk_b@CofJ{?z;O
zp}e>5)PA@8W#rHd>L;)BoPNZ&JQw?oL-`rnKO#Ex>fC|G#hJXrPU6)(`^#JC){5xL
zYv`(j<Q4n4H&SoPFZAc!QCFhtJVQ@@p)co-b_)3gdN7`6^hX|kst%MN`42slQ$J*%
zh~$VUKH_}mHB8B=Uwz@;!E>4CJkQ1I6mj7>NL-L#-ls?YPmk=yj_P{v7d$txBl*-r
zMJPXb&aM~v6(pB?#eTca{I6dv^+xWwNgeOGL4C_U<fadL^_)%Lo{J*NBN5q`9<{CS
z+U}4J$2~nW)2|Ljhu_H`=8@NY_GSlppIqq4qn?}jfxJFxdYhi-uF$p;&l7FWG5pbU
z1%Ai{>Fti|^1Da>ep+UK_YLkX*uT|Q9e;kv4<}deKgeesAMP=nv-sf;^gw?7^1gkY
zJM8Ryat~lX&-2bD_ddq?Gr8=8o^y*n)>(1Nuv71uQ++>To^kRx$JAx)YF~0`^8@z9
z2YvBxc0*4;<S+aVKk^&Du^+p7&ap51i6`^PiBEpczuk9PNB_pjV;(u;JnQ)*KIJ>~
z%+rsqaeSj=A9;lxQtQMAIuPIdPd|O*i(lD?e(__Tesc0x{y}f-?7Xy|z1;htM=taD
z2fq|&{DOb6t9{68ALoR8EPu0udO|!q&-e|0bY9^DpUwgGl5u=NbjaiVhkf^3>!9`*
zuQj{+T&G_09@_rsv7h@Obq&3<GkfZHPDB(}>Sg}IuiWd2S9JNW^V)p|{jnqavXA=>
z`$j+DNBm1XU(;`w$?L8^Et3zS{D<#&@5j#ZJ*D~#l81c!j9uBmcSGz)Prd_Eud0vr
zvmZT>n;rQFe(kFca*w3mRA;ik{UE;+H~2>v|KxS<s87|m5&4g}X9x8TyF>A5ANkGs
zNDuVwIfEb53%_7z>!E!j(t~^!ksgbX-csl}4)PQC#_S}X#DVxgm!8BqeakcAP`@_+
zATN94haUMu3OyI_SNmwA$A8#~y~%@r{w}`w4LitBP=4hX@(I16%iqv7kN?`oJbcKn
z_Ct@}oDb{(=@+sWv_6I68rlc44|I+}`i5~I{V7BTrsl<d_N3>C^h2MJ-eFv4p12ps
zO+|T1e(~JF548CgyVE~@qJ8ND|M*46Ii=2)pWH)TIN_Bodb~9?8=kK);iN-8%{;Go
zf4<Y@Q+7J^f+?Bzpy+!}Fn-t$!}F2<H8q3ziJKjH`|Hy(=s8^6I5)M`dCp68J+Ful
zaySpz!}F7M_|Qg29Rc0D-L_!AKG$72F|)sY(ATeR-1#cc$afK)i%{MYZ%}-d`1yBj
zr5;K>eD-}~=?6{o1Luo81)YmfT~aGLmtd^dJQqZab%D0?FNN}D3Z0KdD1S#(M--v+
zG@|n<h0d#{qWU$WdZ{QroKLl4w5PheR#fjo&lhW*HD}<4&(Fx5*X}pfCG<~Db&PrH
z9rd~A2!5-c;P>cxKH%^8QirIY(1-42toNP-UHZqb`pol?_e9$CkV5a5p16FI9#3yF
zBlkU#?}+?v$2@diJ>s^t&OL8N9?|m<K2qo&!1G5$_dO|8zZ5;E#&Q0{ekqLgRP=Xt
zfb0=5`n9%tJB9JNUE6(QM1GaRwLUoJj?>p~sML$<?-a(ortP`ow5|3z>8TH<Rewk3
zzR$gs?={%peTU!Q@PF?^+*|p+&Ao!}<Gc48eb)p3n3$tup5O5;``kIpUGV12+<P|f
zsrV_rtmkL=_Io(@Wa7p95#RB9{t`F*-F)#yPyF=yv$mP_r{8{B{T-h8rU%d0`khbw
zRoi^|%5#ADh4PGhx4s*$_uC~`np6E<0=sx_MMppS{E;2VLofEDmv~OokNhEjiGS<K
z#V+`eKaG2TieGK04sd?jmwwp^zx)C}Bi>zZ?4^C@RKMG|kNwf-H~P&(m)`ZeuQZ>X
z`6I*!eLIId_u}9F<YpJ+{1_kXpl<VfRr1i2dHfyU{Li_-KK3D(dtv;sCwq{G-^sJ;
z32|e-`bfQ_E|U-4m&&)+vA22rlD~-$a`AV3IX|p(zOZMMiyYo(sRQh%UT{vx`zchH
zq>y}b3+bV$s4hyOdyuB0^C*R>b4@>c)QZlR;XAB0Z0EOURDW+QuP!)o@%{S!VP<Zg
z_g(6*BW4eLy2~lEbMu`C&XH?J?lgXfi>FtAZ*HF7lR5u3fA@FiG+sWnTF-5D((#up
za>lG``@$37T4&2W|MPoj`Br|ASNy(!f4jdI@6KO&K%7J0L5d&o`$oGP{_EAF=Tv{6
zB>%`u&PV<7hP>Bvr$21mezS%~C|;a{_!FP<ACx~Jdqed$tku88sde%-{^bMbRYdtA
zh4NPlQ_r*dJ@=$Al}F^yrlRviU8l{jd_Uy<wBJRsqk2u9#*X~exkTU4`QyB02kZD3
z^j*H+fm<g(voC+cFS|M)w9SLg1#*zjKJ-C8{_7m_Jv)B*r8>blxvf{{(!bx4(TDn4
zJx;FpJi~72ifjBsdMd)0AL!5THt@|}>M-Z5`0@Ru?+_w-FUx;@52nt5;#PdABh)`o
zp4jCt2Yj#LjfM*Jo<baX&j6i6-g~I8_=h?MzuqHEzWkCqwtVo@>hHpg`@U(P9s`Co
zuHCBgYUl4Rvfy!TDu=&2=(&wom{t9~qxTIk75}~~@E&F2?dN@ZV4qeM=P&*teL>$*
zL3)LE{%!3`dp_T~;{A^ILeP65`1hype&eYVTUW&ApKp6Qdui2AtG|<!C)vfl4gTEw
zKy{w?F`nzaUttI5ly&Ut{SN)93)q{#u#fx7&G$Lsvhgd=%%JBb@ynm=%U{&9?w9bP
z&Q~|7XZfxD>6;w-+1Yn0&Ov^|KHuH<gcFB#X<b=r$xD}9^G}nrwa@7OP`_mssQ5jj
z_g;`*`6W6KKT!P(#WQ3_=p2Lm2J(L>Z>Es{&>wrzGy537zS}MDcj(rx;+~K{h!1w<
zhY{s%d0xKbckC>l=)rvFhvz2zMU*eehdw#vnTXbjH~XsR<SX;=6VdZw5xO5pp*j_+
zGd(}XJg@EiWlt!+BGPlcSoB;)-w{0@tN&wtug<4;sNQ9N_fhI~=dZd#y<Bt-$8q%p
z6hHLC-s%U~`6r{hyzuI*Tzv97=(~AHe|~4Io`Cj;@&ptwP<;Zelb6*E>O=eU4|!KV
zdF5Gk5C79oKh6Vjr+&h(xOVP1Hy}DGbdDAwxg(NW-hu21V?Ln|NMDe>VB}vsrcho?
zA%7}D=U)nA+-vif6tZIxiX(Q0>IJ9{jaXz?<Iz6O9T@4xwmKw*>Vb%VJ9)D`{@$jc
zQiRUyT2VcmLiL0?20pRm`X^6)batlhi1^B<SD*d<rwtWxFK)FxCy7V#uFYThC4bj1
zuc<%ThhI5IotN^we)k;loRrU<cb*sVFW*{cT)#MV&xjv<I8VfDlt-S^kG_0?4|=h$
z_>o8Gg`W7WebJMrw4r(EIM>XF=8JpsyVr6qLVV-L_mu2!J$m8*AGM-$Acg#oKc>(;
z@@sp)<-WjkJp0M_p8MnP;>piGP#vZIqeuMFyZZv;?rqou(l3;coD+VZV1MUmq~~6S
zUx-s}{*F)jz=!)wdED>${JjYA>pqDe<CA~Tx43Zs5aY!?3wzNQzUjmB969`5gNV+T
z5=#w~8fZBUcs>)S@`rOGV$1DZQu;b*z&Yri!hHmEFD<`22b{m^E$I9(?i^7^IbZo<
zY&$27$9lwhXPvs0pF8)VadiXw&OP;`bKbb~3OX;*@jlLZi;i<N@{PWFiQZE<zxa-N
z-o1?XRo=(4yZcu9@H`0poilcF&q#0X8$4IKH*jCcUg*hN?Cd?8`)2Qp@WY?bWe4?&
z`W`>(WcMrhb<gg85kLN(o8N)(1N`$t`#^repZNp7)5Zt=;1eDF<ixM_>U?p?&+ShS
zkbdb&+^Zj*C!Ra-Z`{7(M%#ROLLAwLz1W+4^oK9|+mGDnLHE@3;U2;J0(zrQbm^y7
zL|>iFpUJ@=<S&2MP#sSW`O^1*^kqK3@I9^P8GpBsejvTLFZG-Zjr)$my{+e9^VrXG
zx#xFsv8Qu{-Q+#=ypLiZaYinF#xL0&+RytValwC#^CRonm!FawvWs&IdM<_L%X{vr
z<wr<Q^dgUv3*Y#`KYj2U^E|JcPab+^Z}Db-^WD4fZ*6|Tp5BkMyYr45P#&Rg`k){B
zvY&nU13mfP(0uFpo9}?uzJB)WKVENE_3yc}FF)!qXUk=mc)fKa|3#nPAwK6k@zAfY
z?%uA^J(c$l)<<+6(=&eU@422`oI}nh=N`1qdBmT+KM>#I+WoM7<QwOU_k{A=s=F`s
z)YUV_RsU|Uyyy4G{E8m6<t@M4=ZE5weW3dnek<Of^Ta;lT)yK+{EYs@Bmbf|=Po|&
zBYx<IJkWiub5K2EUw)52sNVVQ!fnp#^<3M=i08h4^0q%;d4a}yal4*>-)*m#+B8Dz
z{rzly7V9eCAK;hV?1NwD2K~`XME0WxZQtFwr;=C1rN5)0&T`Kq-=N1n=!!$g&QLy!
zNIr4HuHu5fJMY;~zjGdXzwN!4xTzI8KD_JHm45eO_3!J7NB3IV^u+(6ya2_Myz3lQ
z574K0qzCq7AO67p;>14Y(~tObE{PLyOV8}iPV!%*i=KQ&5A4Yw)Xm;+steTD_U-z>
zo`;RNrcL$t9pr)H1l=OUAAOTco1f{YXXh6E`@8DoS0}nJ6*uU)&t_lGEA$TO9Y5Om
zkZ0K;h2oK(@GXDPlm6(J*4ZCF-uL2{A3^%%f9Uhe6p}-op#F0IfIj`wpZ$DSC(mfJ
zqjm1l#gl#1r_lG3<WToGx1hKa58_q*rT#&O9`PkEyuV;K_7Jb~xbxB9<I(SXL;H}Q
zU-DD@=(o?B<A&tdw(66=I`YO%A6(+oY{4TJI()_=bE<!bNBwDka@*g&{M0$EO@I7`
z9*Xd^<@)Zo-OH^i5!o+=m#sA8ljqKFXngSb54-R7U#%L)?6KS5|M>mZjowpwzJk6-
zSNEOq`cMAQq5tfB(8+HMd2wK?M)Yo;boA{PzSGcn{G;d2I`S8-zl8MbToFh7SKNtX
z`ePsVc8>CI@nKv&=ld-6CZvDwo$&+l5mB8RF}{D+R=@HO`F!XbOKx}lUacw+Uy!^h
z<WEqZrw92}UXF6mlln=W0M(QH-?^iHAWsUPIR4t*ckI-r;@?m3oof6YB0Hq;h{4<6
z_QD|xR#Ld&E2mU;IkIgfBKvrLabFA75Au-b9{ElE<#!IwC2h}#==uGIIF_f)H&4I&
zL3I4y!nohx;z!<u=(%^cPP}?wD(^z?C7rkU#fQ92etFpQB7Mnkkl*2t-$8WbRs1;r
z=@*iRUyCE>EmS|ny3~CB3-Kv0I&a09{7YW)*-sna(7hH^2SEI!P~8pDh4LYk_fp9I
zke(pA(76@SdvkePJwQM1MdYFAulxw|pXhh&O&;U?B;F(OSNDGC(vR;b<9VgM%!Yqn
zEc;|i?m5o>es@4`@`C+*zv(%Y+~(5{z2aZIi3f7>Cv^_~q4^Q5gZ5XKp&#W^|Ir)!
zdR`W9_BUVL;L|+`e^GBbuho(0%V+dKPx$9I+WbvENFhGOr*jUU;@WxVTvK=9Umb!E
zajajSaNc?Tl2_F6@|%AAXyX$f#__{$(GL8{e0j<|<NOgH^uR9ik-9R{^E*8C7<$eb
z=zJCb;@kXs5xsb>@fYs__>FUnziWF=AgAwq{9SK$BR6|^US)UB`TWN{4u7eB$MnU&
zALhOTeb4vaclmoH{@$6k@rdqU)Te$Aqz=<|PvCb=zU%b64BtyyXCL)e+((<=m>=Vh
zUD#ROzz?BuD88H@5&1zx{)?}9h3XdZf`4)6IfNg|U*?Nv`PKa+{n1-Q=a7A%^9Ran
z5#^U6q}Pb#PNDOlUKDrDMR{OuVUb+XUhZM|hk8Vv?|y~9$S3^7`RUwr?(!>k;1}Y_
zx$nH>f8teMRPV}<?opuo5pk!EN5^x8Iz~QrPB`cBM<3qPppQ@blb_`==O}-6Zt9Qc
z9=$|AP-lCeAs<)s`WOFhoV@6{+c_+5)XVsB{zLf@;x~oPmm+lTd<nnc+dV=G-NVfO
z-cwzF^841+`x*B=od(?X+RNW-)7bBZ{nuRV>{gBDL+=y4_j5nweWd4A_df1}j6?Od
zdFVm>;M4m|d_eRex==oZ`qeY|f$q7W=M5O^ZRZ(`<KiBwOKU~>0mk!9Jp<(@=L`SG
zpZiVU8LMBtKjl~Y)r;a8eeq8(^0s_po%2aPqX+S-&a$64lLzn}QG7%Eu&esP^P%T7
z&l~cV?>GET4IT5)b1%e?wEh0kJjg#iXW3t#BR4zIBYOC9f5ac%x4OSlU(+l4(GS@{
zevyadHz>b3_x%o!U7+vd`H_Bd@fUWpzc#t?tDpZk580RB#kk=|#(mE~kM!=`6+i4E
zZqVhYReF5}tlsUw0lD`E&Kvok9?30!oNwxW`_cn6-?@YiJJ`qbh38J^qUR0vGS4~W
zoO7PJ|8_6tdBJmn^Gv-@KK5b{`_Mmm=nt~5@2%WF^B?sN{?X%4-n&8Pn(sHrAN9np
z@(TL=((@|+aG&FR5dZYeKg=hWdW?PfjXFsj>PIKqS^lJd{=o0Zr(U5a^t91sZ_jh&
zmY2<g;!mB(-rDreudHJy`)lJ%Jj+k=I=hft{l%}?SG-k!57e>4K}#%sMfKnDH12s;
z9g9EqV;}#%ulV98=#Ued2jwOAeC{Ra$-hG?9_ZURJw<=SpK<<Tebg)Y@NGSL#jW-D
zk+0c7eZcR~i^z|ioBUorFfP8;{hnLcU%k&SB6{ATXZx_Lyn?Rv=9A0$>^|4u*Yfu#
zotw@9>pU0nf6wRq*}sRYP0##Rp5j-=q4PkyxkwImi{~rx4Ed8hz|Zi_{^Alw{+w4(
zJpsi{b5TBk{2SUoB7cPBh2lv36(RphA-zR(J{O_iE1%MF-_!na_SDRKR{kTN*nz#-
z5C7(qPrjlD{-MtFKFK_Kh5DiXh|VMLi|sG3%Rlm;^BTX-6YIp4{3FlFU+9pZU5(2_
z^x_<LUh8+y#%}gg7g#3`kVF1L$KTac$9TTN55JKQ*+E{jk9q@=gS^mubm)tn$?5q3
zpXSMv^eUf<Bla~<9w0yZ_NT|##=m`_dnEd?-uWO;S#Ljn>wZf-;J1qH+C4XCyQ#VP
zP@Z>gJ0HBKQ2()aMEsDG9h!>H8~GWs6T7f0d)S}fd(Mq@DLv9J{qRTehL5Oc{p>=1
z{1`{yJbEJ!|G{rW>)3}r(I>Zk$**pK_TyJ^U-HI$h97!~^vHt`<K*^S3CY1O;);Ip
zZyq$xzuAvpT4!H*il3-k)MxHn<r(+E?1*1_wx2k)5Bty)KcWwDqz<;u^S3ts$t%yH
z?_484bf0J)I{rSY`xtr>&+0w$(Qo93-1Nc!p!L>~R~!HG3?vUEulsTD+4+fe&P{cq
z`xob>dW~N>zxca(@}qg4C;310yrAxQXxI@qY_Rty)qfY)`JtY(4>T@*^ov7$;7i?s
zZ}Q;BIQi9o&~q?!Zu{MZeyF}sr;yM6hV}M^_`x@R$m9F;$UnKPr+=tU#*cFqee~jd
z-<A3f(LIj8gX2EOcX)o+;_oK<ou&I8_XFM|d)`Key~#@+e&oAzXr1@W5syFj(!0LD
z(AbPWlh5-Kz43p3g@5rOp1comPwL)Ap74F0c=g`Gdierf{~mz+XuWyz0s8WgaVRf%
z4s&ifM?AMiR9DL9@~iJu>7CtD&lT?bB8txxsykAMuN10JBRV&rbJ%x6><pbh5u?0O
zzS!o!DHMOT;<$ruy7I;kCuNWwQYbznx@U#H2ln?f<t6z}f4wN4yf>1M)i3S^<s;_+
zKcg4<kRF}?@|1P*g!%Grq#yG#yUA<5gOR_gdF0WNJ+d<!D)c2T<WKU+E86@Gf8@Zo
zyo+!7UcO@o=OgsKMBcL=sz=pD>T~=;`)T7xTU<c%@lRg%CMW*I9eeOc@kBmySZ}^I
zyG6T;Q}eXRk%|xf=+=w)Pa%CpOyw{A{H3YL{zd40v-;J0aOarw#5trsRo`3ZJ5}eJ
z=YHo9zteVpsjt<a_S3KKbsmU&^|Eu*zScz)C*lSl;)9;;gJ0(*KXxvok3abg9f&{o
zMC>6>@Z%hpPt-x`LjEr9(C6pk&^Y?!Lyvs;$Co_pT&8Dje2PzYk{7L4*KIxSjo-hq
zXZs32P`9v?Jc~~1Innz%^%*&0zP4}wd++bGT4hq!_Vp#Z-MRXV%)b27^P1;5df<<q
zk31K-_oQdf73zHZ`o4xf)xXX!`f+}$L*+f+L;C%RI?V67)G6NQ(u?;3{KI=CbbVL4
zX@159eKwntZLr=0kFWF4n2i49KXsnxcjuP46!*Rl^c-uRHonB0-|s?ofcV3|`Mz^?
z{`+o~Kk^fC;=WVf=dbAVfAPzH-tUNCa?k@ivcI^M4}2Hq`!{~WkKFIEmpG<hb|pVQ
zRTsFoKu><bulp4GfSzmc<=j^f^DFYn7vkG{Id=BkVqd>o5Krzoe0N74>-hzHxOX5I
zeef^(l@Hw);a~oh|6zQ-(Y6j<=N7pje#zmyGcJ#_v-5%7=o=mLj6?L;m%oyOeDnpq
zKco-$`1}HY^a$Mx(X+O?-#MY~k;nWV&fnGde8221<7X}OWZTNcpKP^!n>AX03Hg<L
zL@sq4fAgM3zwZn1B@Xzbc=6s7eevqMG3f7Z4EWWN3v~R!h)n#Emw(6`>NE8sL<c|g
zBah%;9<`7AgA|G*@grY|5B6eT=y{#o@~QVd^bft~XfBe=eUp3$)g#8~lRcb6&M|i2
z5Bk+7&Rg+HFY>lLV;+4#{PGueU^o5F8~0MqA@iL(&LhY_#H+ZHXPi6cL3Hsc4?2H5
zZ#5U?d2!`jc0Q1s9`F_IiEnY}JX4RMM{a&6?jgN<4itatFZ1L%`?^;XC-R!_gZ;ZJ
z?%~|GdoB@I>?5DrpTCiVoZ`j#A>Z;h_L9H(fpdd>=|vrke|A^@Iycm(&Ijj=_bk4<
z^qh!3Iq1=M%k=W2%~yZs`}?+Pq%Y_`(mwct@{8|D<t^WVs$<=k_@0iP@uxl!x9HQW
zdWK%1ydqEg`z)RV)YamYKamIe-Ug!2Zt5BS=DFDSvz`~&pWXSXe5THn@6_AuC9kUw
z@GGy=xAR!rzUGT(b%*;G_r3J$Ifh@@-+J+iALHzVZ*=TW-}sXk<Xz`8yG8UJoj4Jn
z;?2Ggzv7dB!FZ0yf9%BG;$K^yHqY~fadn&gC~sItUV1bSdd^bspo<Ut(yO>w`TR|H
zzwMreMt0=K>O6euSD(up_+$5v|Mtx4_g>ew(YQKPp7gykH0~VY5A?(C?4!<SzvvI-
z^m`-r)=w||jGXi#Z@NElPIx|r@~`#k272Ic@{+v6zvNN*mOSP|e#maV3ltaDv5)*m
zuk1h`_7l(S9{bV3FTHtAwNAYV`I|gOAN&;G>?W?_xVAb;-Ru5|zpMY$OP(*CtMe=R
zz1i~HjcoU`TNkeQcj-KLiGOwx5AtvHXY=Tnz2rmZhx}z9$Unq|=MDEt{FhzOWe<5>
zeW8AKo_KDM_wnN#6aUUP&r{+A-~3h{<Uis?T|j@(J&yY_=ezS$UFJCgU3$hp|L51v
zb@F;Hk+-z@E4kUx`{<OPl82tiO;64Rb`THz!1-uj_Ctr>Bk~{mk0{>h-8g;GAN|t@
ze^75(C%>CdzmcB&M^1FvM?by0KlNTty`!Fs`t$q^`62q^!}%h=dTvm!`h68tC;Pph
zzvrk9v`*fJ;?jE(|E{uqp?vH7a<AolagK{q=-iS&$mMxh+&gFJ!M(comfFz2*Xe%P
zeTMHR{JWOUN%<zuQ-6zFaqK+sUZC0z?@gTt`1HQn-(?!M+QrX2{dAj(ef>M8-WzKB
zZrQ!Le(2wKbbsYLj+R34E^cCcv5)wnSI@)Z-nlEE(7W$#=*xJdBYurL2ke7SZE++I
znGfC9L+?4<5BQx6`$POf`!yBi8F>@Za|+cvP~L**){E-uB4qa>L`R-=u8S}7%7@Mi
zd6M7p2Y$m3%wr#R6Yt`Sy?oc;{#U=}d~u2ny7G&-@A0E9*PVG<tBN{?9-Pbc?m56X
zJ;;;nCx5yx;qUUWw(}2N81?L&h4koo16}!9{L1U*MSk#UoIK8L{E<r=T4#UHRrbd(
zKQ`|7#h!ccZ6E9Ps}G!So}Zk9_GcI4<g<@+h@R;~8-0Ab2g3({P;WSQ?S~KNxce`9
za^Ghh;?q3)(FgtMCl9`@)3y%sPjY$AgZ>_YbDSLX<DL)y)i~eeu>F3v<}0~#A0OmF
z&%XRI+MgcOE9zS7$i)xnpS`VPhv=X52FZuMwsG$v$zLnFmxJVj=+L8kO8WBm7$ARi
zzsDb;{HreY{>SsKanB7!sQyo(yx@FhC&(|L{14ehUT_|=C;Q5O{G6TTVdqCg^lC->
z;xD2+okGt6^`d(iDDOh?3iXR;eyje0=E2A}eX$pP(gXW?j$;qw?4qqs^1Ngnw4b~w
zPl#uT4|3TD8s`W2;!ostZlFVd@&bDF$dC222fF-}UZ8yJIY+$tJ7n&)tUv#kgV*mg
zenOU?xY?1nzdkKf$GETYeZ9IBJ^yZi=L7%V`;lws8%`LwQ1$P$s*Bz8`QAfa=DEW6
zJou3h)Jy7R|4zN{d_2GUZqV<w)lvQ(TkHHC-bvHj^gMTkwiVAaZF|mnc(oa=D&*XC
z!;km;*ReA*^Sw`BcE>Zm_s9WrvWd5!_vwLsT2=h}vrj&;!s|W$vOs0Owl9udWXT09
zTmIs`-(GOo0u}SNSon}}hfHo=`R=|aoH(pY>x%b0_E!(OFQE^AhnjvppZPw^bA|7c
zJO`4)Jo@mvRCcz`^QC)6-wS!J636Od_YU4a&^x=aFMIml%{=*k?<0<Av-Z1FGWE57
z?|aB`O}||xue<)V>fhl;hn;qPf6(n~?)sm<L+tMnr|zBjzk3w-W9}in?<Kdk_Ydad
zLw@#qSm^nX|D!K{Jdb+r$G7^Qz2P4(dUpH!o|u*KC(jr345QrY5Ot9E^ycGV-XXX9
zK>5x6raZ%5-V^DUkD&U+cR}v8yk}GAsLRa9w|dY$5IN*adC~oqd~P4~%~RKq$9<6Z
z0_LFu<zeT8^Fv<OZytG!L+6sb=A7_+>U{M4$v*P0I>tQuz$gCck^I{DQ1?2woQuvY
z@ozkZ{5OT<gubto-}&RKoxi)tg2%O~c&<*N`~u|(DBnQ&t))@kx3BZe^S?O9r#!;0
z{EeS_zW~(-^llzG)YtAI-Rp@%^#lK6SAMGBJn?2<@h@IHAMq#klsK0^_@nyBdr$tI
z!d}PsI;i_vo!9UDgU+qa?X_30&aFTDZ}-K!FV%gS&JAbx>b~q}?X^3%>fEYJ-{sHg
z)&1J@w_BxuL&M*nynlyJ29C+=MS1edt``qJYW?AP#Gkhw{i~m?_fa142b*3n?6A?}
z@_I4S*Dm7wsqftM&n^eQo~Q6H7re6P!Y}pBBc|rZ{uzfH@UQP4@NS;MI6u`UPgC)}
zH7^)+!a=v>5r01W#lQAAW@wheqWvPhRC~kQ-rs1^4+mutQ}XIxbiG@8k9_QrY;I$b
zoJ}1k$GpaR`Azw!IG)&cTH9%FK9)i8K(5sKA1?gR8vFH_n5VGFu8*8^>-L8(HZH3d
zqo38=?$`CLp8H%hA&Z!blT?4vzO~wEn?+7K`<0Jg&r>*~>zY5<e)P*(Q?ZE8B0JSO
z-c&sm@mK43t$Y^E<Nx2g`@1(+x=+ht(Rov3?|2@@c9ER;N@0;)&E>y{e$ly3k4;7S
zD9WG8Px@1Ssei9Kztd^_kqwn1EQ&w;$*U1l=Vz_`_-!f{*&qKYjCzmlBL9u!vzJ)^
z>OUShGEZTgAKNj1YuBoude0s6U+8e;mdg*FlBMv$`=5Vg_Iookn2MX|Pv}RCdj6`~
z{4}-yZkL?4LWgWdUW6$>jQKFOQ~Sn#`i}9}ZSCb7H(27iJcSnzestzZI}OZIICPn<
z8&+L)NS4BA&)AOs8`~+qYVGI$DNOOLzc6kR&vnsGDf{xD6c))@biPJEi0!{WcGX!6
zbQ+PR@QsU(y5xp?@63m{zxBm!TTiLtu1|JZe4WlCvlPBEVfklX*tAV$yWOTgbjAm*
zlK8LQ18?8xvW7~;8~!r-(2GxBu(IDTf3nf(Th7i>cxvxI@3!2aIa#eZy>W8a?LTU$
zK>XpiUVQ%IYi2f_xKJe`x{y5R-*?o7!|vRBa`r;M9s>?)H>rwOtoGx6%RDe4gIm6R
zeq-0e-p^8)(*MGTeBAzuU8ZL#oVaj@<%VB5BTM0Ry_VX)-(R1}Qn-GX^RL_Zh%s3T
zQ~q4kAM=xTQ5+PF7wy~s-i}YKu*y?e3RCC5{#1UP*Zo6R*yqXigFnnu7|+qzPT5m`
z%pa+CO75rs&%3a--Wi&=G#1%6#ixFCaSGYFUPM2Isr_22U;IQ&@e}*wIhJat?mL>>
zU!?EIZ@ulJ^D}iW>M!!Uc<$G$k5~_B7x7)>KSg>qu3oPf@mYjL_QFRBV?C)&KlSQ6
z&M&fWiZA^|@)=j3*Na8`(L)OHnZhFaMdL+&82KuipUOYA_NU}b^%wDDe2XmyT{B_p
zIe7|qedohp@BY!`JcUKihjIMIbJ`v?;s5-*+4W+>!K>cA+(%F6;$&`NN}sVm<v)*|
zci2sDUo_-P@1ev?JkMg=ydo^hPjNiOum0xhGj(4W&yl9K`D+U4G1{rAZTd^?SJaOm
zxYzYZZMxzg7OJF>9g48tJZpY){b}D{Pi;GD#nD+29{A41TRpYh0+mrC`~K$dN4KpM
z-9r@VvDR^VP5Ci<7U9i5Sm^1_i%iZUrq-#aQdp$-)V*WuPsyb}#h?BnK2qcHoNI17
zwNH_qo4RkU=WBE473G=c#&xay)vKo>zEk#!{q^o!>-^b^EHUt=n}=jY$j|G=BKeE>
zEV66Sc+_KT7x7n=kBjne(fV4)qnt&0D;h77JMP!gZRd7TUUI&|)chiOjW<{BxwSqe
zSL`p^zp3NSxfI5@X=<C@*`rpxblhM6boffG8)2#*)?bv*J#W?O*D1Xgt*=!+{++_;
z2eF;vuh#xZuikc3`7YY0R{b^AAB)x%*|)jlMe?WgT5o^RIhBh0RDZ0W=XN{3kBIGB
z&(V74#eGuk_&lSXdcUkc#ZR<Xs_u&OVmtNRq`%hqO7U6kKW@tMKYpuCW#2E5d{JL9
zUQ+of&a2hmQ+g`uuhoAeU)rfSDT;67waRB++%J{KQu<BxH<eCmU8<jdz!X1C?T`Bw
z(TV!3w_Q|k6wxbMU&McuH`R{sF|>>B4U70Q&YvR|<(W9Y)^<wH*e_4Td=%S7dM+9-
zvUAaRksXT0i};({<MLs|TKA3XVw<0(uvWh*npZ2oabEPhRJ+IzqF-tkt&94K?IOA7
z_wlIjrnXc0s3^XQ##4T&KeexZ^HS*ij~MNoY8UA%<*)iZzeh~%SJaPgQ?bZjQ}Vhe
zfJOTk#hrS3Zeerxw?4&RQGJ}!qyAIg*?93Sb{U>Ul(*t@Nvd7sFXV_=WIueSuxNcs
zZ~9a6=#Tuyw&%r`M*KAwW4?)Pa;8xIRWI_(lzk!}u^siCY8UxkN-q5={!;!H`%~-n
z7x7m#o|3!P{>X34|FIqI5Zle=w`hIbuikc)H@1uRm-kXwsP}^3OY@6b@!<zASn;?6
zKK|10yZJASa%vaJn;I`VpN-dQAN-`SRysxb(7i;&qII>7*P93Hoqx!+&kWqVYpV*Z
zH$GE(<R>YVuTz++Yn$6&YaGOR_T`5uztW%b^Oow5{Ka<4F0tS9L<-&eG!-LX^~Ooc
zfAp)%n~L7=MT~Kf(vN<AkwWvDiufsduH=6yZ0`CLUHwJ$#ZgnSNPk6s8vQ@Ei_Tke
zrZ8oH{Y7%ec!}-QzWR&w)70^(r&K$BXB^x3Okq>`vyPrqnA*42{-XRH?G@Wa{*}_}
zyy`EKPadN0h(+;V>v&Ndr^eM2^?uhL=SR7k+K%yFtDiaFBF1|Y?Rx9JqWScg!pN8Q
zy1%|{--gF$<tdEu5ZkHv;CC?UHMVQjbDWpT8~SVY-_*R6-|A1T)8AA-v<{u7B0f`?
zvR~{^<(W9I-ZnW?$c`zD{t(+KzH99-l&9c%KDAFuo_hPmSCl)pi_W7sKekhRkuPOG
z{i*f(<9^!4=M~2MR%;u-FvV|?9q2QKDZ2XY-%^;8KgFN^A~{oZi~9Lf3hAK;Q~TgQ
zh4I{~x1HLzh~GG#;xA?Q)ckt;>y<Y(Kb2Qg^^5)@ep7T?sz0UQmhLZ-vuHf}OH=uC
zzQB|{Q-A*<_7}YeQ<v6@hYo*c`1@_%%v%ame_v$Tb$)vHAOA8wFT$dFZK+##d4G#_
z#^w=If5$_AD(~qpvUBS1oz&Xj)O)*H*LC}7<tx|u(d0aZ&9!f;4o>yocftc}@6)eU
zV{=jb6k+PzEb=G*k-}R4{z}SDDZ2Ic&n>&A{GwL5n>(*4?uy3a{xSa)<(rh>=uhcg
zf2uCgAJ3E6PWeH~pY%sLTdF<rxL+;u(b!XRNY3V>_>JdUY!}(vcuH?Y{aY=%>^U#r
zG&1-19Zq?0bjRawdaK%QecojcA3N#V98TKr;0=Cs;IrAo_Z-pn{N5w7E4vLDw_dl=
z)xSGDXyIS3x$l}&tADT2-+#G!<CO<3{OqLay!C!H{ErXzm{4uU_u%N77uTB~`LGUJ
zuf6%^9k>47V;@)d)sN0r?PfeRV$_hTTyfkyh>sIby5aVzhtJ8rcku%4ezMJ{+4FaI
zT4#?g4V8$auX(ipv{PGERzGi_<FCK9b@lfUeiw1-QWvh;W$D%x^U#Cn(JOp-<wwr=
z*ObZG7T30K*y!~STO5o2p6lEiFRJrW<Mh%}==UG-J!Wk4hZL%>QkbfXYwb_zLw|}d
z{YCe~MfG&idiyjLqrU5Hr}W-Z{qa1DZRcQ9(erBxYyG~Ysq<3$cAm8q9{E_8w>maX
zO+FvRbF#VZR9&p@Nnum(1B&8Ayr!^deab)Nv52YpwaOjm73rlYj})ypUhBSbo^~qk
zQ~sde`V_{vueDw4d6kMo{q&PUbz=&P@>9{hQEEKp&-&xODLO^{=tTKywF5iDqVw0h
z6sG*F*8ccCMQleoW4qRUQu89c*iOl<-*Z~M=sYR1xABy}>o>2d80C&_ewV^jp4A`o
za%`KI!sb5TvsX)F^w*+x>K?6LzM?<-`yi=2P((k{D{A|$8hYN2?R^ip@sssO4a<$k
zw)rn?bjDll7aN~zn-|-j<DqjZCAa?Q&+}{BJwz(b#5+8t-Bo=*=>0~P!dmO$1xNkv
zkp)H%$s;ax@_=2Q9ra|M!eKjZcHXD&+?~%YTy4^f{f}Mq!#sr(ClA{5rm=s@V&3By
zEsd%Cq`$fQrR=VsKc_J2A-0`6MbD{4a^NS0ao?hLQNAgP*OWf=r})(0T>k8n#DagH
zt7w0GHW!QZm>RElU-_x%c}U!(u*eTn{MOo^wAUBkE!8V`t@CT;FEuZf?^FFn@m2I3
zP~@M{PF?qGcV72lZ5jvN*|y)|%PiKo>_M%6*z3wA8}~kQ?S?t`E#CO%uP)eT*lM%$
zLst34Hk)2_MgF@Ui!Fc4TcfJ~4kdczSHGyA{Jw$x@W<agFM9rp_2#WZuQ_Fr1E(kd
zeJ1OQ&_431`*Zj0M^C)|xUo}*R{uR2-?!q^JpTW{y^HR9UbjKje`j&|pZ#*Dw{CnW
zqX%-<iu4$fpQ2kU#(mijqQfpJjQ$wsp@Yx3ueN#e4Lw47PNDlI^XVzlv%aXFG7m<&
z=<AR28Mi*}hhP52|MV9<f7Cm!UV#1%c|`G9gzT2W=ohg~Un%4dDLin~d;Yk`sn1p8
zDK)RCKl0Vm?V|mS^A}hzALzi;`hEVk_*xwXF4*Yr`I{HxHnvlCYHmM&tQEy`5qh48
zShQ}?Gsi8t^wLjeDZGE}U7r~9{zrKc-u!T@4L3UQtvq7XYiy^U&uZ<D&wb5pKe5NJ
z58r0gYdMVa#I||O#ajKLrRK%w-dg#O{*`KPbX2F`uGnpGmdvXK|6ZMQA%#WtP0@IY
zu6$I4sXUN6A5#4>9%5U*N@1L@oz(l2_wVq@z%jY;qWfQb$9s$z2eEx)|LsmYdG_qQ
zrSQ~2s~#}6)1r+jte3x^pR&u)H+FjVOZP_d$JHypu+s1Q49`>e&TfmpdiTH=^Ar~4
z%Z4@I|M~LIPRg5#+kSdt>-P>Alc#X$WB-2OI$O4FtQCEy#(z@iy#1!&*sZc@YaKi(
zs}+mR)8>wgx2B@+;#wN(t-t5@`SBd>x5i4_bY1H8ti)0e^RpkM>W;Gil|2`JsqdHm
zow1_tZCiRgRi}MZ`<q$^s5erW`tE;z=}$c$&_{E>fBUM}H?{s~>N@v|^9rN=np?Lu
zb$yh#*19d$FR6B<Kd;-V=ZM&!;$Q#Vj)&+EO>N`1saW(p-csXDjni7!wY0tB?<rfV
zojQM7s^2|J3hS*uykCLwJxjgqmhz_*ou=kR>xzCiR`h*c9G}<ErA79g+jXJu_P;7@
z?(gzaf4@k-_v%f>qI}leJkVVEi`Li5-`svaEcy;(ZrAPV?yoD>`aAu(HLt1l&iuY^
z{VwNUxA75UvRbj;^V59wUg{idsea#0)QWRE9{xY}-UQsztf&&Lq5u&;3Mi@=wA;B6
zWt35w?zNb`%*Y_~q^4*KC?bu3v=2(972c<XpNKOmK6xrmK-to_v_a6PPztoj)KH9q
zil7403~$HYJNAx<9qXUYf9^dubDzvS5wT*$irDd=bMA0%W{!QH$NNF{ykEQFm)_@z
zAKcrPwVyvseH}s{7skGR8aoGnUV|*^%lNim$Co}{_~+U-wdj3W#_jc7D}ML4U*nax
zeBa0R#w^dV)Q&y=_;m!?YkliezV<(T`)3cA+Of|^x0SbF?~lb}uPZh`_4_7cAFsyd
z8q2Tsbt%$8kBhO_h5Oy@KRxXB54rOGQQNV{JIb&1yhb{x?^^iS_1H8&_4EF1@y5Px
z9s4*kR=-x>zD`Vy`Qt6Jn?6p)_A~Z6vuS?p^Wm}2bGDrydwz`V@3`}qUvT@gzTzw1
zv$xcaoqtT1KOWleJ5K%ljK9Bv?AQG6wO_h_>QjfiV#mJ!Gq%sM=M(?jIP1Lf?0Y`q
zflqqNA+pRv)_y%Swny}F-{MnWH~DeFfA5CRf6--+zR82`_r^nx+4A3O<@50CBFFrE
z<L^uJ_k$3Rb!|PzY;8Tqd`{Nk^Kc#aXPo(*=Xu#~`gtMiBFl3k%XKIZar5`{_;HFX
z?*m!h2eQ0>YcKxh4}9mr-LH5lpQAhuKKoZMe!`P4FMmI3DgMdde(+uYzYl$2&sy|#
zG~+xEvi2Mt^ZD)b)W&=st3USj;g`JN)tCIt;fMCD<-XSP<9FMZUx(1+!{WRzWchiB
ztks(v^WKrZ#mAoKV|ngp?O2_?o;K#^D_>tbw~gO=v)5h!PoDRI!_Qvh#y|V~$9$;#
zKXGTB|CFn|^n-tNc(wPx>RI=`^BWHNb(<}Jyx`B@#`ewgjM*_Alt+&n#(B^1_0Rk}
z{Zqd$x;Ag@`tv#FTHgDZKJ(b}{@B{S%|6Y?5q|w-JJ#R6F7m%m$(H9}Yv)0+eV(8g
zJ-!e=_CBzEzHw|nAD;U8G+!5Fc}|pPoO#yr^|bRbeq0t^SA^Tg89!c@T7I2-`@8P;
z_%HbQ;U2PR9>zC)|7L7{o|7$KZ?=qEYxiZJhc@Q(Aj_Y(@aHx5@r~*jAA4RgZ&N$A
zPn)-Uv$5Tyjjf)I(d!c9R?o&``>_3NZ1rqBra$%dfxoXX*N%OCzG<HK!Jpq)%m0p_
z{kU(xAI|^(DqH^d7x>>@c+49=bl-ov(be`}db^i==XYLk&Hej*<FlXq9p8QK{^M@*
z*7IL-r!Uz5mM?tNzhCyEEA0Q-^FIBVk3RBq<=<WAzV`XWA796;eY|qa&r@sp@7Y_+
zbFdxDTU{ITdH8k4=50K7PS!z|&y6gqW1Qzj7U?o>pI0_!9<r8i_k&_vH#Z)8{-b>`
zZXYK$o;wfo`SU8YueEW0ofzB4Sbo#F`20)lSbzSwgDiTSq4&$iaDILvi~2E+-lrP7
z4r}Mf-v73H;Lp#H<;NScV;^_X^<#Xj-ab!l%+GJOd_630Zj8>$xc&NNZjAQL_}Isp
zP4i2iw^Qpj{d|b`ur15i5!tn$mq)F~|6UBTn_eGVKlVJHo45LF$D2NX)~dIk=W{&u
zd1iU&;}GLxA0Ni@W6#I2dd~CXVeIn&&hz}pZtLfYkuHCnnQHmGb1nLL%*l^G^ZjM+
zk6!mIKGtXK<Enjp+IZ~a=Gf!Set*Ztwx5mpeLri_>n7uNPc|NVT#WtuKCH)|Ct17c
z#}lh(WB$4ZS<AQa*nIXn@Z*f@HvN9(SUo=vkR7YHk24#O)sH={(cT#!(;xdjru{xU
zf81fq@1t5f)^}=t?0L^0@A>gyEr0*VTK@VCJ^%Q5#rBuK^n#b&;dNKO)LP!pOD?+N
z6F>gWvo5{Q*Sz?eAHL@`FTL9X&%f~>f8v5mzxk?<e)ZFS`r4P?`|xG=y49_(1MOX&
z^2v98>@C;1)LO1*oOM|L-rx6<-+93|oqy?le(Ci;_xT51aOqud@X@_{zUTZ)5B|r4
zzwAxFe636G{_1l+_N1#_<I<nJ{1d<S(_eqpOJDp4&wSo%e&kB!-wEW$<Jk4!JtK>r
zAB>|tGtOU^Ew%jb*W1Sn&&%gz%h#2yeVrML`T2z`>mqCCw6T4iw6W!**9ptBv3<VS
z7@g0~vsP?>-jqMDvUBmr0c6L1{&np7PW`#Vx$h4!pU;IX^N>Z4D~n(Jl#3tx!k7Qt
z+41%I)b$T;a&Xgwvw!5``(J!;1-d_V^;74ay5Qif=U#m3IuzgF;Ov95@AaDNU3BrO
zU;MEPzx)mSGT&|AK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$
zK;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;J;$K;OWbxPg!S!KLr`
z!JEJ9u-j>Sc-@O``|O)M_05Ofo~%8biO=nR=DvZxfxdyhfxdyhfxdyhfxdyhfxdyE
z4Lo}PR(JW#6EE1GYR~_L|MH?o{rp$#kJ)Ga#3i4<?0dg_f2zIdh5z`4FL=|J?4NG-
z{SQ94?pJ@?dHYN4(6#HizJb1hzJb1hzJb1hzJb1hz5(9At<Jvq)GPnxZHH&N<$ZM9
zH_$iGH_$iGH_$iGH_$iGH_$iGH_$iGH_$h5=4{|b=l;ZFuket|4^OH+bFNfBY2QHK
zK;J;$K;J;$K;J;$K;OXT8@Tu>7eDsY^$%`(aI=H6f8^r(Uwm){x<7UGQ|Fz!;NYz1
zUVQ306yM<B?1QuK^_uHlbn&TQ{ILta{0+G8JKlQN-+#dm{n;L}xUy&Z2Kok0`UcSR
z-r{5DeB6J2`h~B%<RgdO4)4M4bu7N?-uM0Dz5no^_qyHg9>yNuQ}cKKs^59c%ijKx
zy|wJ$|HHY@f88_wWpA#%^t<<8`np%Yd+(&%seA9dw1Kh5$=%L=%B?^D$=4npwLPlq
zbKD<)&^giT!6(1+yo+D<i2t?6cI>*Kj~7;dD$my07@cR=@GJe&S3USA-uf$hOD)>h
zNx{b+H)Hut&-v?je)-29{6Al@x75Dv6<_qD?|t#R50_eg{!Fzude9B7`rn>*&OY1U
zfA^K2|H`*~@NlZV?PVYT=PNw^Lx;#-^A_Lri+4Qt{Y`$V{yE?BQ-Al;kG=O0*$=(&
zz4v<htKI|iyvHv)?@r(LmaqH3;h05_57uG(``3KJ)qnE(e{{%}=dku?f9E@%^P<na
z>yRziv;BjAzxlWQ;XB@Nc%zSf@eRNGZXcX(U-=skeDC%C*Wrz_z39E~yWI^xef520
z?c;&#klps{ntgqp8*lpg+>P&k*YCRIi~fD@xGZ{{jp6pXJmZB|{nV}h;13U%+86!w
z?SAXUpZwt7Qv3F=`{$eg+b1sHTWTNh;0Hb8y0^Gsf6Ur*+jwkmzwq&g{_NF1{^&hq
z$Li<i?eppy&prRke)YM3a>({sAA9AezU#H`IAm)-?$~<N-|pdbkI^1Femq~#*uC4w
z@mldwecxy7>&B(|wf10j=04x;>z}=THn#n2e(J|<TZdxXZ!EUwLGjY-Hn;BgUiY?F
zT=Is?_N+y7t&Q{J;c>UP%74D&kH75@S^M#2EdIo8?{UW;ed%8xvK`}d{+W+|=~Ey6
zxc45;wS10lx3hsa{@Z)L_2zH-KYMen{W^kUzTR`~*z?=wH~qZd)-QejS)HSf=e{m&
zi*K*bw&STEZ%#Uozdx{T%U}Oq@%2COoJ;QcfxV^HK9A7j;IlV+>VLi4JFd7tX6L>x
zF@Nd&w(h_E>~B6}|IVM-n`$5SLvQ=ZH=KWkePnshY{wp-b6-#RqXgRfQrymo`iwnJ
zpM2iCZ~m6wc>AHXV;?W<<BDTchprcYoLb8+{d(kkU-HbmzwBo}*F4YoTuZI}eA-?&
z`}ruxbN6w(XWiq;KmQMJ-($;<ho#oeJr?u#TdhTpZ*;yNyu-ok@AvZ`Ioy`D*OBK#
zk6Sb+<6J+spRva!zy2U==SK0;>wVmHW8eRunzyfGHXeID9LrCA{pZKWTGqa9OpQ^0
ze%y{(`|)?v_~B3f)W3Y@rJvl}mi@Lbd;X6<|260BZ)*ATmbEQ<eltGyI<je=UpFqg
z$(?`nE|354;s5*2e|XckKmH?!uXw}zf9jz({P5vruRZVeAAR$C4!`G--*<)o{_GDI
zd$ljU*+>5Gk5~R5Kk4iL;yVA&I}gvl!xO*!+}}9dd&e`M|G$6mh95cnwTFN3#kaiL
zpYHwdum3++|LF(3<B;t+U;GtU{Ib{l&Eaoe`w7o@!<*lI_>ybg_fMbsxc432@?UTH
z{y%=zdk?R5_C=3<#AiRb_x8s>_A1Z%>Nhv`3;ybBfA9-_{DS>o_}uyDUUs(+AF_Sp
zFTCP<??3m$hu44X`@iJIU;mGL@A}mb|HMTfyZrDY|JR>B;6I-GzQcRI=1Sjnx1W3O
z;nSb``X^m-wfB_wZ(n%Jr+wxFcYW{S!*2FZzxR({{H`+BU%vcruXW)iA3S{Y554&I
zU;X&MZ|t-0f6+5O|El*NzWzV{?t6akm6z_Z<&RhV{{B7w`bwAI<?(Oad+nRAca_gv
z?Y)N|d(^WZ^|(tuRP?U<ne(1=<@;ZD_)oXF>2q##>$e;}@8vi7-2M;z@ZSFCzV#Wu
zebJvE-t<0ye&65z;7{%G`>W{V(__E*9WUH}?tAyx^7Z}D!=C)Kk3H|QL$>^R1zY|+
zfh~HSSPLI}UFE!eeB0Ms=CS4T+WpwqeVgZ)b+%=H{rj)=eK$Y<@Aug9`FL)&jL)@8
z=jQXXMLPE9CHQeYbxwX>upRq2&imlGxR0G<?Bmv09-Wu_&b9V&y6Z7I|5BV^f7sgn
z^L4d*;p;oqvJP82uhp^F(e_90&)Yg1+de3^b(_Yv&s;rwU8cr!`yjsU$FB=x*4EkB
zJ`Omx`(oVA$2|VL+Uj%6^=z#UzrI<V`D|^T`&-L=`+2whxXjN7Wcl+XYyaf%T@SeT
z$3Jw)7X7>!dcTV2v(}!&#{7ASwf5t-&2Jj>_j%az&v_xspD!Ji9s4>QeIIiz{Ij?C
z;C)|y-g*0M?c>Q_N1Gp;e=L9T-S;o}n@3-Bf7^ELuj{Oy{W!CBJobJ1-+cOOzw_A-
zxb8l(WA!#~uk)rcf4#e@Mem>4$0>^Kb+hrUKk@c+4sLLMd0zzeW8Bs+jrsKnS@ivM
zd)=P*8$bRf=l$+k`_}Tj+p=iBO>zEt=(Vif3&-2q3+s;UVeEAhJubF|+s9Rp_4}pv
zdhvB0v-b1AP2+1l>i@ju-`?;=``fbo`LVV3<AL4BnHJm6$JY9M=6$!i$Gv{<=2zQ4
zD!b|PeCc(!Ur($Zqt6ErzO?_lzVCh?eaYRvW}oeoUUk_k-hYQ{?~hsjyyK1MyyjEq
z>|eQj{_%v5e#eh}&wu!;{eQmQkA3ME|J_$Z%ik|&%je?0zxbXX-uv3Ooxi`fwXX{(
z9{YTF)6WaJe(B?HTXmbhzMr&u`@FGV-yc<M=NOC8_oK$@P#)=>Zur>ewQJ><eq6Gj
zr>qs*@5kAA>AYigXrK1=)qelb#{7E6-{(Zuo@;6Rk;}gMzF+;4f7+XC?YV3`bsj#S
ztsnb*eQF-f!Fxupd;D>FExYOK3(t$}T60d-v5#9DFa18PeH|G4d9ATLe;;8j%b(A5
zd%S-?80~-V^8<aJU|)|=Y#;YFUi<q%R^L8fP>fy&$8hwzh4#VkJ6p@w1<kh>Zu@dP
z^>uA+p8dKD?c4Hf%s(GwExHc&ae`w0_`9h^=R)(XjoZg5Ixpks=N@nTZJ&S3Gk@jp
z_pIf|Z?|Xm2JFWR{``?~`|nEe^UC5=WAwb*6@2QWVeIQC^m@SmE)23`uh(`@w$Ic&
zUq`m~<CWFrf7get{do?1{cZlZV>=(mJQv5-+UEhsY<VB(^EZU^-q}ul+;!fv4U9cs
zH_fB(moU!PneE(D*vI8s@ne4dlOMjx*S)9wJ6Zg5;cWTq%<J6dnQwjhjeqOVTKl^3
zWpBOg1^4{yn-8tE|4tTvznOoojh{~(PqkcUE$g%8y(7z?pYrRk)w41B`A)`JXUy_^
z*3Nx=tj)jnOMm!R-}bs|?04Juu*;8wrH_02x^>dxwclUudfR#5`ndPH^!YKjj{lzC
z+IH?7h|gcop!~M*wSGS5xO8sw(I?#U@{e74zuUftU3eU9>v_t$==-QA1z+pum&bH?
zU)!?i`zCAQWAC$WnzuivxmL`7uYN6Szs?@}cl(xpJ^RGxzVYR+z3b(p|4$CTpJBh=
zwlRObY;ish<NW=(UqAb%umAE(>c0ohk6Y%kwdc40PB-iD_YFDEG3V{yd$TduvE{#q
zJhorX^Y<C-ykjw+b6b}8x+~WH+y?*q%WU~^#Fqa(3AXn8^jx>L<$c+Hc7DF@Y{#x6
zKVI1KIoGnAe%;J-O|`tIU9nU5Xy2FNc<Q+=4}Bk#Uq6vW^Rk}pGZyprP1(+UJdWvG
zJv+ax=jR2o{B<B(tH<%w#}_~T`FbL2=j8WUt+nT!8e4rEvyR1=#{76;%lcET-7jAc
z)|+d`&cWA}t<|^J&Cb{3zJb1hzJaj~*w5ck{JY=u*2f+E(JviNwf5gH8=Ir&PsRr9
z$1VPRf-V2Nt>tmdUpHDi_2V$}?B^jz74z5YY<Zs3)!OT1<D+`iqR)>RxBngve}0eb
zC!X@(uXWA0Tz<&*?%)5A_gv!vZ!P~1+EJZrtv>wxW6RICO)cx8Ij8<Tprt&1J(+6l
z*O6#$#?jxq;*WFa-$R;;^Zu4v{`?<V`+QtG-t>J0`*q6H*nYi(V!n>u_6_t6)D57|
z<2$bZzte5)W9)s6o?rI=Zk;K6v+tvvwD_pL9$VU{zEXB?mo~80&mA0>4*Ght<G=Hm
zm%aTXd%N^lI|;K~@|o}W>?^+My8GL<Ct(+T#`QMvv@2fisTaTV-2Ibkf8u%v*ZBRr
zUuXZM+8aIMMSu4%zkQAUlWI5pdBfwbf8nRE^fiyX-rpQ{dpvts@0F}8$qoGFcRb<K
zfBCED@1InAo%{Udxp#c}1^Z{Z9s7Cqwes?@)BWAuz%T#hP5=6zZvEN4ZjWaVC-w8p
zqndl|XAXK_cGvN>_ki^5&pYgTjLy;V(`y4i^Iv}W!oPj=C-+XOefAGO@Z1Yu`T4z*
zYTt0xTfP3fpL(tRlWLFa?~V7qNAv%UP2IEod9kTKUx|KBh;j6DR8#S(-?y?qhsLoz
z2mc%?vSV{y^WQ)9=YRO_Hyuv3Jja;jzu&Ud^3Qo4*GD(L-ed2Z^Z5{u|NUBI?fZIs
zZnpgY9Uryjf5(mIU~B&#C_49CeC~C&kL%n1^s~R@>9_vdL$>yPeV(5!|J*Ly=iTXt
zZvQVo^u9y3_VZ!%b3WY1&Tp^l*w4}O<H%b3?*(%_wqJhyPPO*m%N?8d9*_8n+uZUC
z-?_IdcIp2QXzT2CMDb0o^`Hk`_lAG9hb&*mu|4w7_44D0E&uz0{P*nH^5385_^5xL
z$-WNRk4HAP&r=)oe#iF2dEVEUwLizqeeCB2Q{zqly-co0KZnlevzF&xYWZ<E)sFpt
zclP;Wb*H}GY>UVL&lj?G-nC=<dC^>b{`fRzZJ)Jbo|m7;9FINUmgc8E|5)EXKR8Cu
ztEu>=|IW^)KQA_Q4|92AuN&y=2F5r2cX#jpJAd(t7d`v$50T|QY{#zG*z<z(n_iEx
zy`7Z&g+KaRkNn2(y6XN^`$zxt;a7UjyZ?1>sh#`yT${giZvOwh`E}3QlXAVs=Cq%G
z*_gjSVZZ)aJGPI<zx%l>f7Qb-`NI9F7JYnY+<yIq;wS&^qdxeIFZjZJYsbES%b(A6
ztN;Jzsrwjvoc4U*z)9MG{r`)W#`fb5$4m9LRcF8N!2j+bf8XNAFMGlNxWeV{EdS2}
z=h@niJN)s5Ki^zyZq}P>?OZ74>%x{l??iTLf5hXjf7l`(ns+MB=Rp?fpn2xvbH6Wu
z_KkGr&ck0{j#;bA{Vl$&7(FiR^J>$%(Bon%K6ZZIGoObaZ)0^R&w6WF{`W$xMg6yh
zk9}-h>-BR}9X>yL-0}Mv%(Is3*p8js<~N;x?Rq;O$M*ZG9FIM2`TKKIE%%#h`TYpC
zW8a_QJU@StwO=Rj_W|tnAB*|>p3LK~D>+6M&BZwSdXe#|`$N26f5=0x@;%S|)ZSEk
zmp5Ga>8G80mHp$goBrJ7b*}v<*Z!jGeSGh@>@D8>q+j`l`~U0SQv1g9Ui_L*eCI#x
z?W+9`Z~w$yUVZ5`_L2SbA9&H9Jm=r9y1%yN|NjI%&vp%8YhLuc{@$Ot{JD2}^~?6Q
zWzV_fmtOT}UwPL4wk-O5dM*4}&wSIp9{Pm8INY{H*XOu#{x~;gKlh~<U;h11{Pdo+
zV|D!fZ~lB~%<|9qAUoD?+j;(f1o-<6{63Mj{Cx+uHqSpd!Pe&a=P4fa{9k#-&3^vF
zht^(nlRN+DT^|48L*{YJmglsV=Vr_Ok!5|J5AiI{KcDn7U-{rGJn!e;e~9eGPr3N9
zFMR7cXUFHksp}uy?BMGU&i;{$?|<>Z73lud)lZ#w>Vku_o_q1B>ri}ygR>9LzSnE6
zchSYC9`s`uzUF#ooppA<E#1M=x}_Um*bkMtxC`>YGAVt|OWWidAhHc=LMb@p<1&~5
z2>=f@KeC*i&BkJofs|ln7nA*rkDh%t!b6*ZR67qL6AJ(hz%yE51wZ4X>KKZm0x_41
zOHlfvZt})d%}N2B$<@bfoN<LbaGu0Gw%2dBF?2qOOCOwO0);CFsRK!qG9jyP+v2u3
zv-wQ3wLt-ZHy?0g1)hv;BM1`>a}Zcb2KGSu2#-Q2iE!9`^~ACXmQ@X`Vgtd|S285H
z%{i-f^@e8AIj3eCq&`8>n%j?@5x3v8=`b)wNfoE5-X?H7UDb)Ej^OirR5oZCEt2@V
zhRajzg0J_ZBRt<!9#)wBg~$AsPK5)d5m90gN6Dd6&&qAR0iH&+R)MHNxS(ivZoyo>
z?xrGN%12e<Dgo#NVEEw`@B!>kKAAk3II+*L-f-4$F@mlu3Dsa?WV}t&$JKY!xp%OB
zbd`o+LVCVhD4f!PT+C<ASd-UiLOHz&6yE!}`N1vvdp{G_B`FIx4lch`-U%Na&&6HP
zJH{BX$u~e$HK+-_#vX>tU<M=r>@_`d%+6+GG0;Ftu(FHEe#S=!U-(c-JIv@zss!i)
zSe2H*CA(lO@iRWEULq7l1!WYlLrIptr~{RXv9$os<aWetoN<LbaGu0Gw%2dBF?2qO
zOCOwO0);CFsRK!qG9jyP+v2u3v-wQ3wLt-ZHy?0g1)hv;BM1`>a}Zcb2KGSu2#-Q2
ziE!9`^~CW{Fu0)A&%2@Ej;zZ$g)`YSNPU9h2yYc~M%-#YUwmMSk}8fbu_g3)x~juW
z9l_`MsA|vzsAkvjX7FPM^nP@NN1MvS3dj0|$Iw_aa%*g1u-(=h;O?rm3PcUU1x35F
z3v~G=k>A=!H|ew0S}-nv{ZS^9Cle?38P*%l`YlG#btR!1OpJ`TY5KVOZaVi4)^EDX
z>GeL-M<*|8VR?S<=N1RI?C<?d-gP1_-1xD6-8AEf=y)#ff?l&RVv}!xsA^CXdW}5{
zm%$830N87K<d~h!#$up>lwf5Sll_d34t^1UO4?yYXHq3V7r?5t1TNVHV~L;fQS}m`
zC@LtUfE`M*^hF)0RE(_!a3;4SX5)-2<bm@f=CQqgyN#jqNnHBiG!rOXIY=Eynv@Ay
zecKkd#hJ}#nyn2A0KEBt8!PZ+Y#TwCXqbb*N;0qq(noj{LP><f?yDz`e}cgUt$yAO
z1$Sg!&MBP9ra|fx6i0ZgkTc>|`}yJnQ<PM3e2Fcg$J13EX6gt&&qq~*CO|d2hBt#B
zGobgQBRtwv9#%NkFFc0Envq*$3xn;p-T-%3tyLgu5H2X%on4^IH;Me#KDtRCzZUcX
zF#PZe_yC?YN=u%lI-z`6Z#dh^Pe9g{glaG`VZ==+$JKX}nRm2y(^XEd_nAIAd07j~
z^LszHJh)YV?`QI^6LI0j;S2lB_S>eT<GHvCdJ~Qjn|uRARfC$)YwTgT3}!$Az+TfM
z$Lwr276T2W1S`9k>}Py*@QVOc(hf5^lPUqa09K_XaLFzhOZ<$Fs+R~wQ9&66>`;=W
zFX}+0Vr(sdGr1ix8)sZ051c14kL~r_Z48}H;?f7FnLy#nLFz!#q)f=_+qSqZ&TKx@
zY;8~g;LQizSb-;F+X%u$!yE)wl7T&tKEk6AN+KL~Up;X|6HuGf%5UhStDkmb!L!sU
znyE8T@ZxOTcvQe?H4Rdqpp3Rznc$4La)&8$15=b#ai$2a#yOs@YW%4q_&gs~4VnPe
z>>Az-e$0U0kB;zYQ+Zf{U)qNlyPjKP3xj9edIQ{DwN`<sLAao3cXoj;-z4%|`{*Wp
z{94cl!0^K>-~)KpC@p!G>V)!Pz2R&tKLJ@+5~{((gb_EP99Q2>X5P`-O;<U+-e>yg
z<Yg@^&+q-*>fqK3-}{k$r9XM_PTbKAz^mydT(F)@q=j4a(o+N#NY<|UCvM=~Tx4Hk
zxh{)WvAD3bJ`v8*9<z3<o?a8c0gS&21_c1tGwgEI9)OS*-SiB*^uH3$JKa3?rhD4a
zdbtGYgK;NNK;>x!hlf1<&cEVwD(oe#1}%u+o0z!<BsWuLF@0|TmY<LoUp&&$7A_xr
zdF9cbaOb+69*aA=Qyg8VoqMR9l!5vsTH#8ap-UV%_CRX$eIxM(Z-C8vcZI%zzJZ<C
z!1QM~Guhhl)0aRfYuU5oDomdKdMieWP-(L}h=QV09FbIRe(&ej2e+B~-cKV<Km)o^
zoWP+D;+}Rr%51e-y3})9CeTfBpTQ)PpvBcCQ=BAX!lFEa$F#$7Apas4VsS!i>6b~!
z$%!W40I|xTCQQOEm{R#3GhhZJeAq{MgyHPMGdQ9avN6vStdnw}Trf9-_z0Kt08Rrp
zfH<6R0ZqWdjdeD5cLhKoKyWMDkQGG*k(%yYT!PXUb!Y-eSG6MW3w)HzhO^9Ki`ibr
z!&u3&V|1O*kp{D>3!*hsjbGrShz(`(P`Hn)IHft_Oz0>m>}EAz)JJ2fNtuw<gM}GC
z;iE_fu?1gKbR@={;{p)?;~n4vG5>_YhICLltceNEvTamA0>JP=fcQ<ng#glb@CsE=
zcXe?W6z5fW6-aSE(Vz)v;H)$OmlWyOSv0C6W--cE9Cu`tmKh6*P%77<j)>7BULE93
zEpJ=pN_uO%ic~L6F?7$|`sO$zLa+#wTtQ^t0(!KP9Cyaci6Q@F%?vg54Iuc=IFc^?
z+_b4s92L|ZYPgZu5-`+tC|AdESq3y=EhdkXY;)Fxx~4}PZ(zZGb%~DfJd!8<a^XX%
z#HG)w9!5EU0~qqT&5>9$oCya;K+J?YSqB=B-*O5LBNND1vq_Rj02n<+LZz+<5d&<6
zx~6MFiZoVBB`*ESs9wwi5`fFlRV8_vU!vkzKF5+Eea^cU5}ibZG+rWh*t*T&k$pC!
zNE6{)x4U-w<c4X&bi(Q20DoJ{%EP%cIz5Tka2tRNWg`0eEf=OGE6F;OBSSDDk%l0V
zOK!NQaf*qK+jtU~L>nrA%c%LipW7UK!`%0NnpS5TfL1%UhoyC6x+~|=m@e*u-auo-
zCf@*2)u1Ny8haQngBg$ju-Ej+F*}=$#Xtio!OAWs`xzf~Up2WID&c%%Cl&x2fSWE6
zaLF!+>E&hEw2(5=M|>ZbmC{$uU_!CB63lCj**G(K?0FLN*pr9Y86rBL#H9}gpFrWt
zLFz!#q)f=_+qSqZ&TKx@Y;8~g;LQizSb-;F+X%u$!yE)wl7T&tKEk6AN+KL~Up;aB
z6AUhB_495hxFhRwPT@>84e~$Va5`)Qaz@-nn-R@{DN3q1BU<aGkEg3T!_*Oco{y>q
zO@L~a;9iY~(GYX-X7FPM^!z%)qfO;u1-!YTn%&B+v4z1CZM^~Ru3D==)F50?v^%>%
zmv0jJt$lQpK7K9e1K_!@DoDniWfs|0*b2CXvDTG@YA~^tJyp@;>RTr`ZM;oaIlbOz
z`sn2KE-cUQ{d~j0H}>~_ChvMP;=(JnS<_zKH_$iGH*h*`pueD=j#ssxsc)cfU>7&g
zUvGEusw|kvxsVRDK)9D2;|9*iONb7=1ovt@jIQxq{1_h>y>;I}-$38M8M6WO+Pv}t
zj;KvD^>OuW2I;``4fq$$N2jYUW9IjMzVYC;{k@-M8`M?lD#~jqea-Dr-$36$-$36$
z-$36$-$38MF>S#46wQH-X}I3DZ=i3WZ{T#?fd4A%qr>M%HFI&)Wv&k5BmLvTPs*If
zwV&R%Z(w)>C*c{iXvRfF!@KI)zJb1hUD-f>@8`A$x10OkPxI11Km)o^JiJ34#69i0
z#o21LbgAdIOrV?MK7&anL5r(PrZ`E&ghhD-k7<YFK>kH8#Nve3(l3*clM_w80b-Rw
zO_+pTFs1T6X21+c_^^-i2*cThXK+L<WMiHuSSRH`xnOPv@ewZP0h|VI0C70s0-At@
z8|!TB?h1fFfZ$fPAuEasA~oH)xCEsy>d*v`u4+Z#7x*Zb4QH9f7PGyKhq01l$LKnr
zBMoL%7es5O8o$6t5gW?np>Q8paY}Q<nb1*C*v)FbsE@`{lQJQz2MaTP!bgz|Vhg^e
z=tzt?#|0t)#yh|TV*Uw(4e6kASQ8VRW!tEL1c2d#0P&lC3jw6>;1#N#?&{($D9)?$
zDv;uSqCpeTz*%VmE-BKlvuIRD%wm+SIPS<OEi)Drp;WFz9TB5NygJC6THdze%3#@*
zUYcU)o~!!*_fxmTJ89P~mR#u?l*QswXjl|))g9_Ol&eFzECZUb7L&(GwmECU%JZu=
zDx*fIbd)UeTV`S$7jEZUHAaUTXrfDWgv%sP`f}Y8mp-d{807#CV3(h`PK5)daZq9q
zugU5~ITNN9?mc2A+{rqSD_p=S1rA3+Dk#eC>Qh!0AYbS}C0Q%^2)9Ck$5ggbre*`G
z2`SQ8EtR<RE2DZb4@dwuvly)smg?e^lyM|M`i7=MNOTer(s+s3VXM1vG@NCYwWtc=
zvR-jiAX&wS3`&$_aZ24xXS#r4nlQDU4i4~Vo?tRP59iM4h!C&gHUJmOMD+Xr_hVN>
z@BO^+t>>J57Tr!=KfDEW`~D^nFKMEc53^E|KsUvl!6cKQ#nmNKoFrnxqCA4fw8L>A
z=gAqm6k>5CSjNasCP{A0G=PV?Bo|a|F#~2m!iRm7M;Oi~);Py7Oh_sgvN6x&P<<#D
zU>!f>qwRd&JfOCL8$gH&7tjPO+*oI0Fm_Ahu71Wx)g@CD6|i)$E-pdoi#ky0>!~^u
z`1(G|rNdcfvBhjJ<6*4i*fF}!=SYKD)dkU-sm3qxQN)Hac_`e+Rh-fsaVB&W6n3+k
zFY2SQ)TB(v>cPT{pYTy6gV=(vDLN8k&T)YVfbkA+ftY{7U_&~n9M;4HXW2F?AOT?b
zAVB=4-$DTCJ9vevr@Oki3ySlqyb7eapJ>noG;mg$fJ=(>>ns}85wjR&D~>xdO3RD|
zMJSc)P)EdQ5w8yNrk1y@xH4FFrI)4{y639?CQ#iH@1$L~SaPLnP!@|zp<z*cZ0@v)
zrm-d*!U&mAY)WZ0G?k^GWwb~t+BLj6%9;Q=>_87)q9a@;dD52`M&i<ERS%;azya*?
z6W6J5pfnCj4B|Cey(nkG)WW?-%!E5x2XciAIHkbhC`biG*<F3g$^zsI9jGL0B_H8d
zDDardR?5_DKs6yn8mpxemwshbFXjOWz-AVsRl-tToRTt*BuL-TbO?z~B0?H35j$*k
z7mkLr%(50$AzaogjtV5J_>e(~k}OWCo9Rp!FiaDsw$s4@{>&3hrsv_@866SgHQWZ^
zLYatue-p^ACjSMX+aKIv?t4G1)NS8D-$36$-$36$-$36$-$36$-$36$-$36$-$36$
z-$36$-$36$-$36$-$36$-@vYHAiwu>hl4xz_kOm2P2`)A+wayp^bPb4^bPb4^bM@O
zf&QUv^_6x(vw{8?GYdJRLc03t5pdEU2l5onSu#-Y;%x6?a|^cH>vZ}mr`_cZ>P42Q
zIOUOi`tGBjtZ$%ipl_hrKz|WzR&quF<+>N=OgQ!#b=WiEf}X}_%<ui&@nEmN_j4Lw
z=6<xkfxdyhfxdyhfxdyhfxdyhfyEm*qOWgjo9Zo|`B)3n{PJFmlMh$USRBDw+2s~&
zSC&5`<kIf)2K6FKRGhF8>7Nk`>WAzb=o{!8=o{!8=o{!8=o{Fz4dnNJ_73*vzW397
zGD|=Mx=>t#Lmk9D?Yhm`YPEE!=eA6so8mr$NhU#yt4pRhNyLOjc?6GXhvPv0MJ~kR
zgx1n8laP}WO}+tQl|fCIgk3PD@;zq23`qE}kMan^*@b6tL@i`vo+nr*<v_V$ZU*rY
zF6RN925taxIN<`CfQ1|DZ0zm|fIxuYR<<E4iV7k%-MP2~r7!Bx1dy(3Mc^0sD3=Xq
znZ*{fy^M#kl4Hl{I-esAW>ptNYo;2%z()}q%H*MNA6IcobHtg@QBc^;YQCtC#!{0q
zA*%-qGk(HHkqlxBzNY9%j5)^zA^^razy)Id34;ykpmJCf6P#t+sDK23;e!D2n|=!c
zr0?Jrs-EuZ;w~u8tMV$4;(nq*6VSj}X#y@O(yy~<R7c#0QMTf^OsRtUp`1$OJu!<=
zD$}8k9C|8hWvndoTEUd@aE);kEW6U<2H0}dY4QL4<YV%g%BdRw5fd)y(hpe^wtT}V
z{Nm)+M`MP%4&~}}F3W%>ti|MUl5NhKu=4zBjmoGIDjg+@{Fa#*$A#PZR*li22Ab#+
z9pN&`lfGQH#HG)w9!5EU1K8y!u2bPaX&jUo#A~v8QO<;^g?o>f33sv%<O&yXN`b>s
zkP3>jyZV%s1;`gVP)XKGKEka~;4zi0l&RT(YC?)MR!b!={mQ6b%mWgD%`8T%gr&MT
zC1o5*kiMbm5E7k4gfw0vcG&7J91UlgWi6^gxU5$k6-d@he`B5L0)}bAH1~9HfIo8u
zlj(UlcSc8qcn!A!xKJjd-~YcKyPEvo&;G$T_4j^wH^$0GkxaRG0^JnX4knodEv_z^
z;v^9h7UdB<rX7w0InT+^r4WlF!7@g6GD&h{rU5+ECApw#iy1Hj5<cvsJi>4`vBo)u
zVM0=|kd1jBhw4MQ0PFY}A8qIJ<^i=0+yFvMxPT^L;l?@}gRxs0cl9$qsxFzLsDPz|
zb#VzwU(|t0Ur*JUz}NRtE*;J?i!Ek*84qJ6$Bxl;K1UkNsxFAuOf`Oik0Lgd$wT2j
zuHuyDh%=$1ps<_Od{G~br6y%URu2|t{DhAp8N?QRP0^7UbB+r{0E~Bl3&i{r1{=~r
z<*+6uILo$C0SN%Z2La+Y{T2d9-@z+XJ>AvCT~M4?<y9cX{X~N%pn<c}1YA<2UuV&%
zj+n(LTXEcxQCem!C_<@RhdLrgi+FXAH?_QN#g)ObE4?(u&^=e37VrJgVbmt_$@!3W
z5dfF4^pm<3G5gj9B`Qu^zC~oH>rk$a<gyHC!dgroC)wt#2`kU9)~Jjcq0&*Z$Zwg6
zaa_2aZ`BwbYM_ZO(Gf0_Jn74IOI-S_>S2@vIDlP#;yM)$l*U1cLA)lb7v)TtTDbR!
znQ$lTK(25BrxZ9G1*xDYyQ@!GS%7?@1C?a0<Rjb)1s+q`N|~Ass3xRHW3^P`(yxr_
z#XKMZ*vw+IN?59kQ&Pr}1nC=^4k6J=L`dT$Vu!8n!qIS+S=OQ|gv)xxQGsL?A2KKr
z{-iu)6`c`hx`1JtFtwQu4)8}NUYyRuxidPV#A~<>z=bjq{r=vMT}^)P=bH{LocrDn
zD|Op9&^ORG&^ORG&^ORG&^ORG&^ORG&^ORG&^ORG&^ORG&^ORGa7J%Hekb~jUPM29
z-$38MZfqdG_jBRFo#wswa~9pg`*7hd`~tVJM0&e47HewQO%kVEO*tzAoh{57apEi*
z-$cxB3GzFN`8ArF7pm0~1Bn!O5KtyMQ@4;TGkzwgnaa?3L_D|=9tgeD4A4hYQ}xDp
z&Dq+5u&%Hhd6yO_KrwF0<@@wlBgt4ED->ZcNJXK#ohdf7F!i7sSr8kD@z@PG4OJy#
zWdx&0O>jlDBp2(*8V(@_s<k*%IotBoP%20rJ;=c#IpX1N;5@Z7lN6UXNXw^%HEYaU
zP(@%fRLdIKPe?-Zm?^_otxo5QI(Y?=T9&619Q!Q}6PKAc5Dfh;e>mo`MK127*3wz%
zbly$EM9VkwVwn47vDAas6uk61iBk~&WJ9?Hw?#z@`lyn80v?~j;)`RIBN8^@c<L-E
z2oiaBERZ^JmuypmffiX>bx=nNK^#&tl3ql@oy?R{I${iXP~uf)^>Rf66s1_jD#hv}
zsfZi1WQgOIX;Fkxo}h>(6EW&oYf=egxsvMQFT&ef?2zu#ahV=$X;d7>2%uk)#C{Aa
zmZk}mKqS<?0l0EK0x73vah^)no>!_F6h*P4x?7`RdjvX#MrG)oO|#dJV$W8V#*(A8
zjD*l5YS}H-h4x%t90oSDRXGcpS*f2qy-DNlp^)kZ2yW6V84z*gW7nX~N?hHxqfF#Q
zHc8Rx(k~WH)mbZ&qcDpzZJrIHJXI>pU1^T=OL#?EmlSxmGkf;9xll#7+1)}xETV()
zBPmN|BPFRpTtT>_Zt-{WlN)RJ5@gEIP(!)WICWGlxsGi-L<}OigdC+}j*4>Q&?y>l
z7J$kTK1+OMayV_hsi4Bi3q*M}a~<z9Jrv^ZXwmT@QuzcB!?{JagG0zw32Y<J^O4QN
zLCKUb0we5sX*9H>Lc5!J3k9==59IheUX{F^{Cpw2?2E2hjq_X)oHHVEBY(Ikwu<6q
zqBJC71PzrB<z_B(x9Hx%1~yUXW?i(tbaRha%J@<-3s%VrCfj0xvkS`T0RmH8t8J!(
zs(G+d2BnHa0k1~dmee2>a2o}+%9!U37Y=8sQ#lB3E3Q&z$16wAPx=*a>8VOC*6}m5
z3yRFbN}5rYW<Sne=NAm($80d<rHDl6SGX0YjCQO}niY;fCzyLf6?q$J4QV%N8#F__
zP3#o7`MsYz9o%`|dq25RH+=(r1APO11APO11APO11APO11APO11APN$?gsMrD$d*m
z^pp1u^bPb4^bPb4^bPb4^bPFd2J(A9cRskwviE-IeN6f-Kiwri!t@3mCyNd5OT~BZ
z(mPwFm_Y*6jS}HzVa3^6klsX_Bu`YISn)upflpom1!Ev(Ku?z`>^<jc_8<{W^)rxj
zgcGT~Ra6ryo+UylVFizYinMr1Y54deJP^+4X{YjJ)p*S~L+Tx)p9Y;d?kIhfF!{vm
zDNSmLUsooT?-=Qbm$&}u<I7@sjiTekBZWJ*<VZgTgcY2!Imj-D(sG@(MQ8oP<-rGq
zY2uA;af9*0O7Y?c|9H-glx^|Gqgufuis=fsHy3PaM~-Z3E`Ct7Hat?|Lx(Lf)_?RI
z`67G6z5%%b_yt8G(I*D+Mju+X^(2(@?4X|xPWN#W=~W@V6wpEv#wOV4l1(ocm6Kjj
z;(g7<oiaBbJ|fXOad&w?Y>sArwXYgf506ne#R**LQ&{5SB*7ArQ-YM!KulVOjdO{B
zg{$>)56&mc5f4yGazf`z<Yh~;Xs{zP+@}$!oZ<zi_*5r?><Wa16|I8BqY}5wha=VJ
zd#Hz%*3>x>!oAOwsPv^q>oYUFNW^4~RUh{;NnGY<<q#S~sCt=PSIdLf@EV|2rPKg{
zsFcnWX0lC6B_f9dYegzmRVv3+>y*i~@|jCudHv3ek}FOrJ@HQd5EjlXZzuj(8yz5Q
zB0P5JgLVFswo@KY)5Oz0>A_StOA_%xM0XXQ<W1{(^ej4GeXh;xA9vA|_-I4h9l*PO
z>GIN}DE*3@e9gn;d`8;clpAlE=%roUJOK$5jffr_^F3MWm1}eVSjg>bKpwm$mD&pu
z<KO#v;aks1Jx^UfyajaExo-kBlhh$q--eg7bMd?vrCo#BYPA%4MKd{cAH9szMWkE7
z$*WMzd3TB8Wgr%X<AWwNh<)i&n_EGOlL#g(RXl$IvXl9#O~5EFSaDTVnkJFry{OR@
zSx8mKASar<1yWqmxQ}obZ+X^`O}|qcBG&@Z+6Cn_f}34XB4Bc3NJCLoy~`TeL|j8n
z26MQuFp^V3&X=Rd>lx5G6mj`ao-&p*R)R{4P?qL2(~|@x1HvGa0wCO!&q%?eZW%KQ
z#avR1S#!tf0ylupP&)rf%*+=?xTQ8%QEB>3nM@Q;22&!PgM5y;HpbzObJAwvNFhJm
z)b&!_$VzwYw)rGdr!{d@=W3pnEGtzfRu<SO7K!NOCRrv^RfZT1z-6wLZ0Tnm4Gn8Y
zt}0Lgw<j=DgDFuNJm#3p)&)xmvxM~ve6;Aa=@t<^$VCK$Q)r+f6!Cn=yn;n1Q|rS=
z(PB0r>38lE8L=RK7DOCVL?IX|rDDEvba{yAcQQ;$1LlbNHCZOe4l$BjJXy=yF`0>}
zp!CU3WhkMjO^V@`DJjT6z#7;=Voee*71<#a{3)xUyitu%l*M;aRyUDi%t{#R2Fr;E
zF-~Y$S<XeP2H=DOLy&N~kGJS#ZUBl`a6#jOm`gZAqEaNHoQw;-NMt7%@?w;s!X>9-
z>7<F6(BunFB~xAcO`A%@{6z$XX@C5AQ6yxjCP^Uu%5x-@I$hQeq)dqPR2D%dl(YmG
z6qbaN7=gjhK{9d5+#=;8r@GZEtBRF$#}cI^sS!@UD@7qGm6GJEi3pB2UU2FX;Ybe2
z#G(7*C5N;M8dc<7=13^|6)!3z!*F%!K=hls&`T#X8pT?iQ%o*4o;lV#A`ayB@@UQ|
z2pTUZhWx{aNpMr&0D_-gX(V0xNxiL&oN&ihXfZ0b6j_3T-J(a1D63)hNTo!LLx4&L
zfU$<0xiPPbC=qWd#Zgi!k#U%tUwrsQsaQT)cmS4>7$ZydNrMR~<_9^|F-5qwre<S$
zgj2<`LNcmK=r~!Skh6lHr;VJg-lc&@TLfBW(mrL8+^b3$O;savA|YH?O6GyY1h^c<
za_iAVqvDcNg6c;I;nab1acyZ^1cfm5v-i1#%Tj%e5>yf-E^eG{@(m!0f-4_ttU-)Y
z7M$qQodk(Qw}ho*6o83rW1^5b5>^tCFX-G>K@`YZnQU=J$g&?j4%kW=L8TLy!v`-#
zR5alaQAAg#MHY!fk&azsX~<+w(u&LsGNDjnGDT4=$3j@Zlw_goS7NM8Q$#lpk>HBd
z0`d~BuvAc?5Em9ZY%5E8!&%9!%H?29aRAcqP?(d+<bd>qYXa4c14U<$EG0^^IFd2o
z=iQS@rCA`FVX>MO<Z>-jHgi;<2iCJ?z6Q4yso}<(5M9m30Na!e5d&oD7GD*4!3om$
zOw_1|48ep%WQZ+6>}N|RPLW<+vP&jpG?yb%9=`!LfJ>|V7l7`1aJN<7`*G$Ro4}SJ
zei_*o<k+uMXbFk~v=ofUE}lS{l77@}fYhzP#ZgmmFr4)opo$SZrkTSw=nI!;0wAYl
zh4eWeZB8U3Si%*2gPKqfr?`H#M5|vtrR>8#%7ug3kfeO%Bx1Vf;$jV4!{QYu*Vwp7
z!OQ4Z_$aCQ86QRLI8`Yrju<w%IipR(G60YRuqn=AC07reoGJuK_!1RuBrJVVDcDB6
zWMrlE3xOOz;iIZh2Y!K%=F0M+Q7fTnP^&4L#|B%ds#-2b8q8!DL~EuRzraTk8_MLN
za35E3N^``S&{0s>&1x*lHl=mMCDEFc2?Z5q{DhCX%L+{pTVw)7u!>upG^z>3{MnTR
z=?f-+0|R6L93_WxqUUTb*Fyc6;H<S=z2rotHaW6SYSZ4Ek`VGhZN^0q9pO;~B@v#A
z!q*Y$i-~wo-0&xrl)~bqVdIw`11$}`kxMG&zM<y~z<pSwDbiT;l~oy_I;uhjRj?kO
z&sEEnC(9GHis(|!)Hxw(X`J%#ijkM2^c7^j-syvB<*{0_1uss-bu<Iw5o8xfOs2<t
zRNmqSF=6R56G!x*S5ws!pkh$_!YU4-M|O){Q|mlHokQdX=Ba!^5+J!L6kw%RuzE3T
zLKdl&$s4CNjJk6p)-pklT(=!QsV3#DUw|CPtYk1D+f0kb`Z+Ew@=2zdT1<Q$VTGwY
ztblvX)pqy^f(d6ThZ|X=xZ4U<$;PrU`njmVWf9z3F4o+^6G$qqPArs1ID8z<EkJyP
z!~N7mD&>2&Id64s%uAEF^f@nW%mWgD=j~66gh&H4i$tU^=WKjMOhLKl9NSpTL1Af0
zPHU8-;ViSPMO6s5^(7uC7)Zr(l~+MgpW#dwFiaDsmeauj{*38oQQhIN0@0h22H<)(
z;ewtltFSeNCS;be@@orDfl2^#)reKqHK?AhXFR|6bGL)L&wKAD{RmI^!sCc<`V`^8
zu=_VVvKtQ^W8jef@qKilgO^%CWP!w3NwbT#z+G<u!y!>>Roq+V0A>zAqj+pW{Y*NS
zS0gc4D@nWItMjX1{(PmpI5RcU^*EO8rAOt=c@e!ntdi_;-J|&?;Y%;#BymH^PdM6J
z+7%na+u1NqO|Kc^Qd^&5{*zC+*?DMdi*FkhFYdZ_p?3-oo6|R4%Q+Pbx#{wWN6Z9w
zVJoLhtqDJpa7JrRn0*~eZo8qF9)3dP^qWw%Atf|z6$o<RmbfWWn^bR8O;GPNhzkdF
z2$PPZmH08?Z8W|rL(R<+cp7&S+04EAk}%tc!^~DXf&v;gJ?l1X=In9vMGRiOrSePb
zHeAr@RxZ~#c%4qSgP&&Sk=JpPqfOAE7n)64%&J}GRc%(sog%x7cEQ21(-z@{!-BbR
zP<%!bH<}?LsjCZ0@j<_?w0cd!?D2wDBeF3tRdUY7=0lfHAsI+EeKs<0@~LCaQ$3T}
zbS;L?Kb&<d2ld>F0LI4zN;Y`>$KS%g`F8h%d(3<9Cq49<`}~=)#2qNjC`q1g3vKMx
zoc`GMw~#xm*JzH?MD9eIc{t3in#-qlOKQxRD~IzZk`>a|gU!t#KEidkktmC-qBYVg
zr?=WlhzobxV4bzq4IJ<FfZbt&mw^Zf=Q>~~lue)jaScz`na<Q5wia!3plxa|S;)XV
zFclJhE~1=b+e`%#0B*_|n9Jw7Gt?gWy8MPdI`f*WA~Lh5pigYCX&kYQdMXedhLVFd
zRzt<hE0v50kRrlG>CU+u%)CMA&aDBxwvP>Z6gxmY*22fmxGQ;mULA=UO~Qms1hD@&
zKxBXe*w^NP*0yK}gs<KInDln<c)K@Ix-@7mi3>lw0%>HL-1uW)xMVmxuD(-@T{z4x
z%&=1zr4Ms5Hjw|`&pi&ldG32Z`5|8)`Q+889N&RrZWOl!8<!5`prpc8pvsp7u37@!
z6!#cRG6~vT?T8epI7!5WMR^2|X@}!LW&J8wg5reM(l3*clM_w80b-RwO_+pTFs1T6
zX21+c_;5w5R|oL0noZs~$1qGtDi*47^>W_^Xsa{R=;}fpafOe<4n!{{#bL2z8@K^9
zpQA!6PRxi%xHgqVDx-RJVBsacAY}nEKB{0vQNfswBuih^p$Q;e)r!C`@KG+4S?P0L
zTgi~Ve+13QN-h|4IFm|sK1UifJxtPsnM|f+ilfk)kJLmGRCd{V$`Nx28n1-v3ANx#
zT>6?-B2CdzAtfg#eU6XDLX$F~XaX~S!bkN6sybo|zNYA?3eEC>2!Mk$$XHm$(jOac
z4wtoHO-yi>Emsu_5&%XEfOzS%p)+HiO0Aym?&1u_zhV;3lYbE${!k<?Ka;a=%2J>c
z9I?%GrZsSpD@5f8zHm)Xu8<6F$hmC!Z(vk~R3u7;t3@ia9$qs-6IE>@nnCCDR1OY!
z0Cr_HH%GL9ny6l#1bNIYE37)45~U7Q`IYC<YIf8cFDHillO;3M)Hi_OJJU$I^mEfx
zAriusmw|XCRh(3PME65ohjMj1mt{Z`)?)HF$u?(ASb2W6MrG6pm5!1{e#=aZ<HGHH
ztH$V115I>^j&PaeNnfs8;?ieT52GBw0qpV<H&@6)>GK@XqQoFxlhqqkt1Y>JZOtSv
z;ZD|pY^{!`6gV6OsX)jhTwOy+lD^P!FCyg;AK|tl@R-V0wJq7ESEBnxRdi@ZZWJ|u
z5P;=aEEBEfm#8?3ECNZ8zM!-&Bsz%*X}m=2ur-}B46=r7LsW%uS+6)MkSy|J5;7=J
zlEtwaQ{(9ZhH1joZ#p=@pE~+nULMY!(a|DaQv!=+B7UALKhwABl5BG^HUtwAWe5_v
z=%&xy<3Mzm9k=l$Q0PlOfXk@)y`OJBIGFq1kL=QY-T>M+=TDA3eNJ+;J6~sSL1i7G
zqdoje-A=;M>x|V~>G&(j%$;j4f2A=umO1jdbHgL=yYZ%WLjdm$XLe)qW1k`}NN-6;
z$3Eej1Mtz<uBm-`YK5;V=y(1VKZOX(OZ!7xj9+@pUk8u_@O=HTiueWH7)j*Y!4lHq
zqG;z_ns(!fS%B@%opR1V=@9XL^Ye@-h_rCsX~*(z(JJ8ab)G+dd8MLK9z0tP=d1RL
z$s6cPp1i6~+;{_L?4xb@!F8pMyz4Q_7$tbB3$X3=*`D%(WA1iIURB@N@MYTaeyV)+
zh6g&5Y<}<O;NYGczxM;q1>o|To;@bw!-6c}*ZHXJxhQs}6DU*Ck9``zbRWwb`e?XN
z50$yN3wo1{5u1DiEK&7pjvbqHvSXjJ570NTr47t`0?cF29C@qaGiT}a<KF;!Jovbe
z4xD0kfv2^Kh$p;r*xZxO;74laC)3`y`^W`t!bcDt;ZX%85$?C%2AU=~!a0tx)fV;z
zrU$V5Wsc_RC{OHyBd2j{9B4%3q@}?L0jy6qiR4~Et<$5l0e_zjIu5{{H>G*?$AzOS
zZ|9-dVyL#@vkPpG1p=k{TnF_+IMR3c34#en$>B!UD2@nD4-Z1+33sv%M7jVT%+z^s
zb>g5r!r?e`3lJaS1B>7iw)OM_9wJLz`cMjx06cGhNJI=HDbg2|)}2gjO9^Sp8^Eud
zg|4PKAT*GQvySCw9cTu5O5Z7*87<eP&K!>DGfSq<J1RR$@H@3-E#g;h6XhB5dq4L)
zxYxY*e&YWe6s|@3DZ}Rf5lR634~t5IuBKBQP96qq63#*zjt{U_5e-0mAjLh<U0O)G
zkIf6UQDiZd@W6yQQ@4=P(m^JN(Rj{m32|;d>JSGYS9pLwmYS+J#%s>j7Vx#CCFQFc
zH&|miY)tVTJtp#>{1D*G6ib&Q@RES+k$pQV(Wyoj#6Dha)Q#kn&GS($xmKH7rg@U8
z;e;Ag7E}${8EI>fYOMl3dea(?%jJ-ummh@{msS_o5giIxQV*eubgD%0QpJmvy>#*T
z5~e;Gqn`0rs(r_qMnQ4jaJMTr+IJo&d-PZf)A1j2$fs>EmTlZg%!N#`?q1{$qE=!>
z)pQp%!yWyOe$>nW^-rl}a+Jt!!$D9^Ge}%BDP|@&-sqroO9$3|%8gE-&XR(>IU!Ix
zaHn8XK3H_i8B7R;dFCo1^g^LqTP_<9O1#RfURG@Y(Tzs1zr+ZXmgPApSa9n8GnpVh
zSGqQ-AM5aCB9t)dFQ*{9y(OinEXh=yQ(}^W5kE3=Qn5=xnm0}pT=xdx%Jm5Gq;b4-
zSJ}$*N);gnX>CVzPaQXH)gFOFH7Y}|pJuNg#h$G!jU`8G8401s*0Niw3+=hQI1FrP
zt8x}Hvr<2KdXvW8Lm|}-5Zt6!G9co}$F4z}mAJZXN14crY?7kWrC%(Zs<T!kM`0Fc
z+B_RXd8$;HyV4x#m+*?TE-CPAXZGxIbD@fEv%7_YSVRZoM^cu`MoLnHxPov;-Qw@$
zCpXscCCHSap@wp$aq6gAavj@vh!{k22{}r|92Mopp;I*AEC7`we3tmi<Z#-0Q$dB3
z7l`s|<~rVIdML!*(W2u+r1A+MhI5N-2ZxZW64*wb=OdekgOVv>1V-5N(r9Q$g?2ad
z77At!AIR}{yefG+`T0V4*%w{28t1tpIA=uSM*eV7Y!$`JL}^IE2pTFM%FSHnZqdDi
z4Q!&&&AMoP>E<4<l<}ov7OavLOt!@WXBU*w0|chHR@+PmRr6q_3`!M;0$z=@EvZ2&
z;5G_sm9eHbE?W+Q+X|7)j#rMJpY$u<(o>aOtm9{97ZjO=l{BL)&3>G{&Mz3mkJ(_z
zOA(3CuW&0)8SPk|G%Fl|PB8a|D)Khc8q#jkHfV-;o7gFE^Lsz{I=J_|_kPL+$sL2q
zSEvHX12CdP94ChQsO}{wjvdyE5!uBPC{xmp@(o~`4(ErnUISDyf{&#2(1_g31Bp4&
z<Qt%o41iD4BjU)fo`Zq`d<h<8IGb4G9K$dnsaVLyJdZ=oDZj!;sl?CtC}PLfrKC7{
zqH1zRo0|4LAP3+X8d$;4_^4W-qNqU3<>C^QzNkanhjdjd0>8jVbD95nozT?opq8eF
zXdZi%uH|y1!90Qs4$b2i_~?KRDkz&l1P6DZC?j#{mmb=gK;eQ+^(%a|nJ5652{V4e
zM^OXB7JN<75idc|E2KSHD1E^KaA1H8fTQG4PV}73<yxp86P#tsRmFk?fJ>QhA~o}q
zY3~I+ySboE_z0pSJgT51!kyND3SjgAY3K=P;5@)g+^1_)M|@C5*)m=&%+y5n(DB`4
zQCTXL_lYVc39)fvA=t&W1V_QLE4{2g!PM$Po(2{A8Rm%X2-hyZq>8gMYnnEZ%QBz|
z<q2!5dNFT8yEoPtr!@@mL|~O<TjufNdb%3fIVrYSWdm<^f$cF>U@8yH8DXS>aHQq%
z69f~ElEa-gq}vLqwT%(em8Eh}0AO68M#0sIhVlqUQ$idiPy>X+{nP|5RV$f(@G-%d
z2P6Qab@(_E5$BN<=?hBhLLy3NTt`8^0aVN3tfGnlRUurIYImSuAQd%KUIn_vv}n43
zVVW>?n+^`}rwu-z<qn5sh~AVm0N1+-7xZLVg{>(xA+wC7H)ZUtMIr;hwi?8$x(3zL
zb&Tite(rs6pZ?yD-cX5nRRze7iziT~q#r9Azz7fLhqGP-R55~&r1j8<+{^=sInm@B
zppguKPtqge$giG*f&qL99%MM1SmPYSFd?Z}$i_U6L(M6_!bhpZ&-f@}$JV8!IC-LK
zaz>k)_B|j6;29cN!O!@pTA-q+K+NUh5|qBEL)(XRRVxC&z(;dgfA1$Q!rzgPqRSP8
zP(isyh~VH3)QpNarQd{+DmSb7qCVP86adVG89(8pr~zUNzNYA?SwjV^Ckv%7cmNIz
zkO6R%9LkBFv$<Rg^<#pwY`LmfkN|Kg6HcUNelqR7pl3H1v<V+Ubc9D0ltj4G8c+d@
z9v}@p0S%l7n2Gyzjp~RG$|zgLtA&}Gs2)1Ldn_tTrSd*er6eIXPAmkwxR&53SazkC
z^(UBGUC7fQ(RZk_<R&;y<5<<=1~Fl2H4{hlpjT7X6QE*H`@$*?p*wM#Hj&FRpb6#a
zYN~oMZ$hytSzh<Y=!Ck4l52Mj=ObBb=cL$Twx)?^7uX(C1*Y<v2?c;O5H3oU>+ll<
z6RJD~4mT!7akmw!Y8xY_D@)~|0Km9Fje@Ha4doGzri3_3pauws`>6?Bs#Y@n;A4U@
z4@dw;>+o?TBF-Zz(ifE0g+!FlxQ>E+1E`k6Sw$5AszSIZ)$TySKq_jeyb5%SY0-26
z!!%*)HXR(`PaAwb%N-8O5WOjB0IqivF6ha!3R_cXLS`9z65jiH;aks1f5qt3^}}00
z_w8>2nQO=)y5cULK$(($lxzTQl<jbSIO{b)6(jgaS`UrL%{-8p6HUGW8p#0oBt0UI
z{OUO<7{Hg{L58!5HO?^%6OxLBY|QgG)SU7we3VN3jE^FAY+Xu<lP9VsXSAtl-ve?0
zo}qyi{EUyP1uBXP#9S^eLFtP+w0%fdwIc8fd^DH!H-X|J{2loyx?DjB6_jg)2oCN*
z&8Ub|`b`+Aa<iH*>Z8p>0l-X{@e@9Z8X&gdYl@DVHB`WQvQYYh2jIW}830Ggp`7SB
zo6EIOKPEWKmaB>d2>_Qe;Y4cYC)3^wdUkU`oA41tM|f00NrXGC0TsaL0n*SD(7<_s
znYd5asE+ucjIw3CT9~Pc>Y?Mi$D*=SD(@3jN)lq@#6qx(YYC2mWmkGxe}bvig**)s
zeTOPbZi3@9j#Vvg5EGVGGjT)@dNox&0V)QyFRbDax)Zl)6S*t{noypurm7e7CKQ{J
z<#m6IPN-`rxpvoZK9aR|PKqsNYnpg=f$cF>U@EVfPyk2+;i6Qz4nILKp~_R>aART=
zcUz&VwlQM5vQ!QV0E`RND7ZS&P#)oEN{FKbYJhOKpPIm>Y9-STJ|-CRfCON)4j)G%
z;yjWfeL-nmNJI&Z>nO-KfND9MRa6n6Dujzt?G6+Sq@sq(t3bDy7EKp0OcSPV)4>7$
zw87`I+~Kec(VLP6;CeUVf}Sj^ur-AyWR|ff;Z30M-p_px?l<?npLEb^TQE!M;Yrv4
zytFW}(y5fb3RY^+W%>sC2Kom22Kom22Kok0pAFEziA`T}q+jt%rSJ(<`Wy;<CPmx@
zw?#z@`k<1z{ab!QS{!xUF+2~RuJ?&O#b!%A;$t6iyUljftNRA}2Kom22Kom22Kom2
z2Koln2J%1tcE5xB&wcNQzw}UYYX+-)qP>d9ny1=SYr=shP$o=)R9xHxJ#0c>Ls>_~
zh@P16h)=(`-QvQDS3>FwU%2B!f-GEF2VkPKsftuj&;%G;*f{m|Q6XbGGoNizDI^Wc
zNcvoK+*Zz%e2Hu}T~(@{TI(kALb8z=$~9yRAQ^<Ob~ZfH?2MjOM{~pmDox6Sq6y5v
zgeb1A&-T$;wXvrfsGvOcH||jcjDEq7&)N1w+t#14xjIVqbUDHu7N7zcJ(L=H0vcEk
z&_seST=2Rj8Dn!5ZgNV9>cV7|Q$2~s;bp2xDiM8FU8i#kBQ42Mu<S~2ya!j+zj@1u
z5FD2|CPtYx^*3P|Y9=n}(hqtj9BzyFWv*ez?KNDoT6X>3M$v@J2XBR1B&qKjzWmJF
zsnq+?(JCFGBuW4r!0wERn=53Y^i?HF4B|Cey{KwJ)hPwz45w&oke)_I)U-uLMv(=`
zR}zw>5`KW`AOm1|GDPtwn2ByxeBLbWC@m5d0qk}g;&P6y5y)^ffMP0E3fF43mQFg>
zVl}sLI=Er?m9EK<%)GNmcJ}cPxk>WS2q@>*7qmG1QYwHbX=Bs};0w$1dq4L-c)-H<
ze(E?hfir~z#hSUer>&dUsCuYNWi0kw=>)ndZZw!=612FwWQvnSOjwjh@R)Wu4&+|<
zdc9<Y^vfjV<V2HifLLWv6DDC7OsRa2888D9KJ24B!f<xs85~gy*_h`E)=4=~E|{A^
ze1yw+0H=W)KpalEfF@w!#yT6jy8<8(Ah?xn$cmzZNKJPxE<x#wIy3>Kt6CBG1wP7U
z!&zpr#cVI*VXWlXF}lv@NP}6`1<{(R#xL+u#D+3?DBQ<YoYEX|CUg`OcC(r<>Z7sL
zq)f=_!NQE6@KGd#*n+PqIuc{fae)Yc@eXi-n18}xLprD&*2Dy7*)}R50buwbK>ViP
zLICMIc!jE`ySg}o`tSY3@9l?F`l5$9?HV}4cBEmV`DFZ%5q#mc!^Gy20VK;C7lm6K
zo?ihnNWQ$M63J*nlWg)NkZzV~V?aaYglmZ$(E@6sdMt!3bIS^=4yQ!1166+Id9<1x
z^~TGIA^!-;05|mwkof?Yu=I1&R4fw0warVaIAYZ8yF*=va&<D7Wk3_wV)8i2HfK#(
zd49D<Wz-0jj*>-w%S?>p!tH#k#^_K3O>~KlaGB&uU#?r?(q~l<qa45i?D7-Wsc@h)
z4oVE-HCeqVXTsFNy+_Q1J6Q*Eg$p>Pz~LxJ1!Cg{FFn;YlqBg_cJNA~I5DlGT7hsO
z#Pv*NTSdqdsBlX{W3^P`(yxr_#XKMZSdLYNf~WZ<DvlzHKoX=cD6I>LP9j1YFA+Oz
z+fIZGXZ;#nHAPKvRDiloz>Q7}GnR}YG?HWaT40zaOuXsf0Kc|e*UQ7XGdi7!*Kiwv
z3uPjI9V8*9C5tneX9y-F(hwwa$qo1AUuGyyZneM&eOV3QGHQPB=K%-bGWWe7*(Lg*
z6@beGRtNE6K^E}qeAM<_6uZ(1lqu=QJ`G^HkL3-0G+d~M%3Rz9y~)OiO}+t^sCqTW
zj!in*u}|3t=o{G52If5h=CNmvyjAg;vvm6LZvZ_WeB4I|PBFW{(^^Hu6W%#&?n!6x
zBQ^7rY46*8<bpQgBZ!XhsDhFR_gilRO%oj997ot{3;P1o1K9mCM{{+QCw9S+(>OH_
zG$L}+(%^&u)~B09a<8D)=~3E%zfT_>-WjUdH5{G7&PlPwP;J3y7uX(C1*Y=KWR8P0
z5bj7sXf7Y=91D(yPq>qHAQ~LN6VJ4?8rGo%!qNPQkMMy-a0%N<t3hOhLs3uwV6+Y&
zM<QUDnRK8s4mW)oU|gd!1~xEdpHnykEw-DQIULbvmQ0<0RCbi$cWTR8#IHOGG{5)r
zEe8*r_ukK0@fQAZ@p-p!_0TO`SCyiFxZKj)ji{0tfl>k&WQN=jRkcFFfIujTo2Vv|
zX*XhHc~wz=K<p?}V!Ekq$4c@ktDG*#+l|-LUep>_4_%E-@H(6I22=}`I3*BHyxJ@v
z@aQ@<Os&=y?QE>huY&m_hw{7uRXbgks*6UFQc7D4Htx2qkCbV!W%nEHi!R&rOSe(0
z{IDS-T#8Q0Enc56dmf*|Xz7lR#x^C!PNx@8#Ii%}Cfuld9E<Mx=U$>eT3qytGg=}E
z%c`Wpm><t(NpLMWp~kA@1oM)uXnHQF(n?_wwke|{=efka>Z3+KJW5_Twdtmw5DhL=
z!x>eg)wUUhwqe{&JMi!WCud5uUOz?HIPex7y0*FmIdDtd6sb+Bx2YzmcN)Zv13H9B
z$I(jsnD90lpLS4lR-(CLU>g5)dbr5RcliN7YnGNQhta+0IkkhgbozyZQ8As1Ot@vO
zY{oPb&E98`q6t)%?y{|G5zzAJ;kwzQ?>*EpT(fkU_>vRg-0Zbo7CXWb@_Qn7$Gr`>
zm(3#*=3Tzb^->Oolvl@7?>KU!J?qR(^@#x7Bf8+=aq1OxX;?59x>aK!)f$eCO?rN~
zy5Qsy>PoBE6we$lDm5Zo1?q(57?b)jyL<{+0rL2x$|S8!s-w)QT9eCXNb`rIRUmu$
zTpAZ6#Ds{C*QIHOCTP>duT@Sp)T1i8HT@FoBpPN0&d%@sJn-N_^WOX6PU^a~*Rase
z0Oxc&Jv@(_aKU=jMOwHuFFi$2fn@Eff8qw-%|-S#mg};36^jc?>l5J|?J;Y&>ghEB
z9KhkDwJt9azyVy!cHU_ZKuC*jdJzAM?@d*g)z3TKJod7QcD_n3LHc0u2^3Iy8o}Wq
zPrvi8_?!xRNvlB%^1k+(uo5IUp>!3>-2N>;AuYbjc-t4rgQx2~>lCxd+ds%A9gg-W
zN84?ay$sbJ-6^)*Y0L6MJ3MW2r{fOOnI@iWBCT9-%ct04g?xHD1IHdnZN3vE-rx<e
zdGD^!H_$h*6C0TR>}DogJAV2S2xTpMc3g$Y(_e4JC=n`cb_Y>VREi^#%FXZnJm}!T
z{k<QxnEAozuIEc0YXv-yn{dH8zeo$W=B1|yDv+%1pVk}b*S%cc#j99cSX!S5=V*^v
zyH!uG3E%+6UpayT0P7icIcg6;NQ-WIhW_5qrW-ugdij8mJ{W2O1yr6!aDVRyUd;CK
zk&qUL61$bigZKA-j_$IY*vE=m(b1h^%bm6?ua<VY2B+f=)0rloY$B~(f6J%XVugHq
zI|IiaNNv6oB;Mc+uzByU&^ORGuoD}Y{_JKZTRZ+4{oc=m4<54cy&t>A9-k%~z|Rxs
zfQTHagGonBK^qk-rIQ$(*{2`9Vn?^0wo{9Qaa}MT!_P3vHlJ9A(uew!NTciQgua0l
zH=rJrE9x$hg0sMc+r?-5s4K7x+<Jp$V;xo1wvWGU{TX|zqf}3qBYZk*nu6AoQL3M-
zBKX1uAJHU3ww69~*vee;VWuXk*L?7nsU{g3=zQ&t@XDC&X-N@Pw|e91%&l*ZGXf7a
zE_z&yGP)UHCNAmH52_|Kebx<+ve&wkAEjEmr4~u^>>7TQTRLNVrHk%ddI#FzVFJ}o
z+TvF~2ogj%dWl7RgzxO)cWB^tlopA51n}$~kLZk~XSX}@f^`qDSj{b*4yLua{3GA!
zjD~PvnJ9j7Ob3J)njx5wUcM|8N()pR#sm;s*tqna<}57FfA8lZ2j4pPy`QFG1Ej$L
zTsLFRhxo7{*ZXw}EkSWiZ#P6_7f+x}Nk8g0fayA%AI^FWP{jy7lGZ~bax)Jk=0uZk
zfJQO^K1q*=BfokM3I^~cc#z?2VvTbQ!-S+_Ash2N4mGFz3Lm8sKjWi_9b1=@;^c{{
z$r){G+V_AQfM;l61wZ4XYJrNP0x_41OHlfv4s9RORjmm80w2v~^*c{O2fKq>P1!uQ
zT6m)s04<jz4dyZzv;wk)U*MwyI;fzuLIekQpeQ48>6aebnLy!!O!X^#w3#RXm<cm}
z!bed9#1?!_(Gf2}&?}@pStxzM0&rk}41lBLP)_um&E;CC9}}Eq%T>jK1b|DKa3VGH
zlWFe-J-fM}P520+BRs00B*LB6fC^yr0BPt6Xy81+Ox&kyR7ZSJM%gl6EzHzJ_0aL%
zV^LWumG_A%B?+-{Vj<YYwFF1OvYWjc<|f0&p-+QE-_pUUEV&5|DsORvn6R{(i6eT@
ztEuV<P%)@|VHJnaow!Y#$YmMOgz|JXRlS%uq1coxulr+kLR~}2wY!G%k*u|IQfx6>
z)5Nn2Y>%k|Q+dsV0zeuF7p2N|_z8juRh|Ne8xy0r+X_{+jS<t8rE*XJU|gU^!PSX|
z@(4#$LL4Pf1BAo<)C4Y7E17=qF~OJzBmkpz_&5>~=aCfY3rg!kB1&jnM?t;;RLkM4
zqKW`jAzYMdcc5S(6*W{|1-iwwXu5!5nlN>n4i50A4L+ac4u@rk-jp-|*SiT9^ki9u
zttm7ivy9dMxLZnsN&p){6sqbPR8JQ&PVfD^@Ed+1{ntUKt{>h4dg$CYfwC|wQYd}h
z_e_XRx6{M(xCs}mmtCZVTl3OW1QkeD_fP8$^y^+O@8VS~E-bB2gmbjVtlg@o*933?
z;}4sl0Kj^NU5?rV5YnQXo?+&@jGML_te<zfdF;W)CpRgVAboJ62^3Iy8o~50fzj6<
z=~sMCg}tQJpal_k!EI5|f<CBZZvU2_kQPTBcMQ*ir|UiI6tl=Xc4{-u(H`Y!yP2b|
z79QOxw%lpU@~bU;mxWHp9i}r)JlRBAxrUZ+<3c?Boq?kdq&D9R5^wMZ*t~aF=o{!8
z*oh5He|9sItsOso352qiJv*+#<ms=sVw4D#HoJo;C@RGfN#*9h0QAs<hb?^XN2&GT
z^x44ZLuV^T-D=ml;)$QDoFHsKC21oW*vfLcDBM7E9kzZUwtAt3xNza=d!lckZ{Q5u
z!0G?Oz$eUB8@k#WVLHQ;9}lH)a>W@<+>v#yGDZ@q_tm3I93ux>?eZt>Q8`WHh;1xa
z!Kh#-Y`THGZ+Go3IIZCZrPCT5DQnqjmQ~F=*jhb;&RSCQ4sb;5e7Mv4jMG<%)`0No
z%AGlD1G~ELU7c&zayxm}2KqIhwXQRe93LU=!z3YeUp@AZVoUkGpNAbhe9?P9XHC3q
z?<f-oXg5DT(p2ff3m4A0jW-Gmr;edX^@_xmo7ka=Ch9Jb7%OQO)fT8th+xShK$K5*
zCoU(@gk&iqtOUX#)XyX#L3LeN>#~m(ROeU0{IN=Tab{|y>yfsfuh&e4#q`u}ak(z6
z5378NV>02#9G2fUsh!&u?PspevGLIXhVc!@-RNazO0j(igwt#|^4jIzKoJ{Wyv*v6
zIHS?+L}{7aahhFk616%SW!eChrAMjwmc86|!!COG36axpLe++p(6m*E*Q$lrM`~~3
z(H=@GXDRDIp+giZohCYiyNPp)>8qwZQ$<(t>avX7cu!}QMq_RWhrvSPGC1>W6$(Qo
zeLP*tj2tV7ScL|#gweRw6eUaLm)42aXS_=fnb{x9(&KE&J>GX#!-Pp0I*Ute&<1eJ
zP7*lhM1$9E+T&sM)rgstcES}Tf{+!_Y3RDgb3t(UxIb}CT^bh5m&L7G0HQ|JrpgJo
zg1$F{>I*DJ^JbK3#Ac7D_Rz>J1i%z&^rBf#Qjh_skO!Rs0Vf5<beP8l$p@*PA*?|I
zI}y>v85(3b>sAixxfKD74-%AY@U!!KKMy~6#Ju-@qEb9l_)MUY&A?b9eiD&Lno7GY
z3#+MkHhqU$8=N(yD7A=xs*}Rhyiko)SaE{{zylMe2L_QRnXz0!3&{g#ONevxQHMAH
zxxxedvD8$(F<x`Fwjl4QUfH_8v_VyplgXs3NcqU5HP5)EqEM|G)Xi9!dZbHOGqGJQ
zDPF*7C_zl!s98K|VJe5@VjbC9965v-sMg|4<!sASL#ZHj^dJX|<cNp6f%DYTOj2Cl
zAT2*Fy7J`$RFO;wpDqDwWIrJZ&10qvUp2p56?O6oBDE|}Cph+7940O^Zy*@@UH)*)
zV~bqeNv)-`(CNIJgo&1K<i#-e%VMbqttoivcM_){{>g@N3vP>w7W7di`2;*Zg~b=g
zDn}%2!tvBuQV=Ba?pPpo;x5^y1_Ld!wCbRa6oNRUWF)<aggcojrF6s?@Sw!2%<AQe
z1}I9gidBl$M^X_tX2}r8Ez_b1qdY+oO(tU0vDTy##&RXq#b1QCx7Z=wrQ<R^*wUyt
zj1fS;B8mMNR4h#sD1k_*djoLgdIVBV&Eh<jtUa$(GboB;M|HPG!}bVt3XRIpJDX;&
zAH|-nER7{cYZ(clN7S-gstfJ8yf_SOXsdD-GP6=Yd3uw^-9sVO4G`R<S27^t$j7ch
zo0Yh_ZAY2Ni)@mj)1_Z5oT{@{Bu8NuXWBd)M0u)In7h&(>6h?|v@R*|Y-jfDadV-H
zZnL|Ef>=Zc<401K%0^03gSdilN8RG@<R>@Q@FmEUp`nIyrE%)0T5=uRc!(H8atS#~
z#T*so#-UR*;4A=@BYc+l%H(j`dQ(A#lNX5cYUVoLXL=~a-O-}sL!|NvAck{`YzK#s
zs}k5op64T*hl7$SVFX6l^U`Q&M}>Ab^A-wb4Ijwycf2ZjJNfxSc-a?Svl{2QA~<J6
z;zs^(QEU~(%S35N!U!5FAIi;K=5EowgAHt=(9ODNed*>NuaxnnViv5D6->6p0%sSL
z(E|jgxK`Ut2UYW6r3^|HhXP)Wv@NMYD&RH>YLzk18!jBqQm1kd+*Vwr%#K%%o}ct9
z-qKT*T&&|~W)~Ehg_Sg;EX{tLz0NNf#E;ov$V(B4(64YSP8sc3oir;Pfle^@hAQ$l
z(i+lk(l%&@c$?TMaPxaVk2rYb%=dn{Q@k?e;)IJ*E>{FOjtQy>0AozxIAyXx)uAe`
zi6uD*hb(x6Dvd%()tOvh8AL7g<Pbtj)g6Z7K(!^IzG}og-G!UgQAivh-Z314sAI|k
z7Jnx}PLeqL*1EkBKaWo7^d|vv7D9L?daxxQwWhXdD7LA*Vv*n`sddSwNkUlb^CXp$
zv#BbT#(ORyCi}_8G+`1>^|`r7#9deoTP*llG$UlPa&ei=iFn8w<cfGsE{?TXnevgX
z1alM0R6KYbkd=P|jl+b8hyc3FE`<z5Cj%7^9tYW#vGPM8sIUibTXL;WQyMv<M18ig
zTS1LN!HFGiGR-n#T&6Qk@e7!AIOH1Ek8uo4J1*e9nZrF;N7~@0<3%SI5+R_J;Czv!
zTtexd;Y_T#lnnPtD<!I*>NsXgWY^Da>Z6PsivlhrOShaVf}H0><Yka^p0#|B6EK2H
zui8GsB@<U#X*wD3tV1}tG0#B}(e;Is`I)Xp2^ho0$ykYleDR5_qf!p=Tu~k|MIdSx
zVRjNdYejC=<yOY#68d;iA5?;4!88gJs7F*_AO?(DkYg{;IdiRuO1G$6I2quW2^@Qt
z6c|rxAtr!}b2dV06`U=FT%S0hFfFI($~^Lk9P49wk@m1iD|h7NKXGpMY5sC61BI+8
zD1JmmLbxhLf?0wcqOu=dCa^|M$6kpb_<69*iCSD2Y>Z<K2RWHdBru6>)+)sbm)SBd
zfiV&83PJh;X4z!Z+*D)!a7eUp6Xz>T2IF&U*<BJ(4XPje#gL{*H~E-|6A;WpT1yFu
zWetVNJp0Tmuc!&AYD^S1CF8wbX6PnzY9K0ihbeHXq&YH|&QGrQ$EtDcl05;!@~xa9
zVId4ot|8CdVw&?xA{XY;sKGv2y2ZFOEC^?WEgiB%FxeZmBugab2ajFP3U+o=d8wz3
zEV*R?UCvr;U16Fd-|{)cF!^BeD8wysS}y@qOShC{#b}8g^U1_&8Rs;aa7&W0CIRw3
z_e+Yo7X{fhHcXR~G7;u1aMDgwM`{)lWs$;h=v3XPLd^-vx8w?dNyP@DOmiXIp&WB&
z+i9Yjr6b!iXK9nk)$|a_t*GjTDYC3qAPML~u|}DYepFF_460u}b!iArsXB$A7Ovz`
zOXgHO4pQ(GB_4CU8pevI1#Ha1(n?Mdl7#9fDi^S=MHZo0V=)SMoROCMM2;wAPR_@4
z<`i`s=f(*Z_vAWUl9nlo!!0wpKtK&!a+udB3O(jqj*X=j`SNBp!V@Y&o<==i%$66a
z`oPV#?uP&YmXmXr{4`EN1q}Vf^E<#Q6HbwaC5}kO)1^AWHz<XoM$ASi*J(<wiAwpd
z5c4c_fb=WMC%?`@df}F<;%Vbu=@_U&vXH4HmBv0rp^zyZ2(_9P$4aSEr|T4orb<#>
z%*RQ%CN+~QBcHBSYMS#jO#r-MEjC7rd7hyxt4x*$l_wFi=Mk8cWD=NDvM~E&6Gau4
zWElb(V_(LxDBp2S%2$@eD_Scv#x&I?tx2X*U8)K9IThm^GKOp}sp(wT;GtD|43svt
zCgvbcQt=iwlYV+Akq>X7Bo<gp1vHAEB~+A=n{T;JrI{^D%6mVLJb2W+_kOrjw|xVB
z1APO11APO11APO11APO11APO11APO11APO11APO>wE^)Bv4Ih$u?7TsYG*f)u5G**
zu0L*n9Qb=#i1u|ZT-=#cmLDShk#bCzJpa9)M;$zR_Ip3aea%XLa!O$KSn%YO(9nj%
zT9v&D3>Shdl@Cccxd4@S0Hfn@xErU~GR~IqVH>*8?|O>^M3mrog#6BA$|SisT%MI1
zk{xbV^P|K_$nCQ(_2gf&r{nV&P6|SzMz~^R<tCZBhFFsC$kdn{280taSt7z?Cc(^~
zWOGR{)r;L(OTmT3?An9*36YWXV^#t2AwqKT3^A4S5u<_>(><@t$V6t!M{<nG%UIEf
zkJVV5v=lMnR1ynIj)j7yVTq6**Ext4{#{Z_g0Yfl5iA*o1K=m3a-kQeZJwTk^sJ*>
zwt>cO!s98jn@nPa?)0nP@-z|7IY|h942hP@a!lJHH)i4_BpShlm4ezVdWn+Ap}c5B
ztsF(sL`l`mDxyUtN+dX9X%PX4H!`sY3lJxbo@Y&l8rDvw3xmN{2Kg+<HIP^nB*1^9
zToDdV^dpI8Od%=cfF~y?-Xh<+0Tg)gCFV0o$YE8edXXr@G=#mUHRW4zVyvUe3dxkp
zge;J*CFSA@lqNBa9Ouanq+-_OLzsW3M7&RKD^_sPNMPjZkdGr1nNi5}faQF4a#>Id
zKzw9~hUA7@sUepFy0MjB%1pecez};mq+awB5XVZ{%8-?e#>prMf70(%6%ht#w8TmG
zqAR7A-QpvtJa|ahy@`%I0?iz+37LA6<rNKHJQ!>s%Zi`a^`5KQ24WPZK2csxLTH?g
zX!%ccihgp0n#pV)B5i~(@8~HS%!H*E!V6&|2dN`mT_Q;mes$w@=gS)|cYH*|D_<^u
zjl|<2DdYmxvL2SS>o3L`p$C<mBjqc?e1V5#&S#GdAxSeU(t|cWu5%m(_-KU1Ke3ej
zc&npOb*2<qW*w`Iv!;Q|*z*TUyU82ia|#(=gx{h;0}$6uf^ls^irpxgD(Rj(P`aYg
zFA51ydG;%x>vB0IwtJH!OzDL92uGTTk8q>|a3q6pATt23rv$GxYOxkRaJs3fxuX<a
zR9Fb3&Rm6g{3+}yoY7-w0+L({*WCu9{8C1J<sS-CiqQ~Px4e86<&=z7sglUz7V`C1
z#w(CK(UbuC9seXj_JwH9G7|auE`7x3Oteuml@O_TF%aq@F0bMYN7Yb~G%6wCwN1&F
z1}R75NLKI(U~I%vTL0mwC{%IdCFyiu8!{>5ayU*E8+)+;q4JOs4|kDNN{-7ZQ>KYx
zSg^41cuO`NBn+Oxa;~ro!OI6Dk)oNLiYSGqUCeT~Y*R(tdRESArE{2pxVdR`(;b&`
zAGYZdWMTjbuQ~E#l;CnkQgf&zBa3A`73Uhk&$IK+%TZ{EbYsiK8CJtc4(Fwf`EW-z
z2}h42L5n@<?D1C`g$QLR;>Pm)-p`{C9y9m7AK5RqZh$lpfYrQ-In0Lzx!CWc&=M3!
zHby=oyLbX+O8T)y1DM9c`QfbB09B0OBWXP}A~*9uVoo&q252M$;FI)-IP$CKpkM%B
zf(IGSCe}E|Fic1)7P2wV<4|+TukcYS@iRV(*s*mfDNdfKnw-(5rhN~{0eFT6R`4@E
zsurjyDiCwIxCEsy>d^KfUDb-fFYwV^Rz5BYnE`iDtErpERts;m0-)t`q`_R~f>uDb
z@C$r&KnE3+R*2x>4)p)CcOGC-9BmvQz=932qKJA5*4PkBY~P-}Tri3yu?Df7-B@B*
zVkcrWdUjM|i;dWgea_yo#0nO~h7|#g1<N-VxF~q%-R*6e*?W1OcYoaOPW#Whvu$Sf
z-)KHv|LOn#rIGd@D|RskNZr+R2-9mn%Mn>&eowPoOm8r~)?5PS_Wor5w7aDJWfb9)
z-9N5M``@PlKmGrcz09m8dzjf`X7#^Oerek$Y1+K6F|676|J`8RYNRpLe>Ahf%<BJw
zee%!0yQc1+g3p(q?4J5>YJkr-m|kOUcXR9ijm6wXbL&kcZDv<9tIavcXS=7Y_-u7*
z25ek!diS(z)BZ9R)VS72<NA*@yXE7CkN=q~zR6-4f~5rgbeXd2r%@W#r2n0=m$}s@
z*Qbrb_>0i^@2CHz#%~2ahW>A7>sEb|?w3OTx6O)wOeFkosL$7@Ta)hRe@WT6-rVja
zzK@$e{`bksG<z75|8KAV{xPhj;?KGqncOSg%5*<JE#1dWAOHJzd%B;WDD`RZkNO$c
zS`qHET~fo+w%%Bz^h}>m!|(L#(ueu%x6J|4r(tab*6#ON4nJ98E)H|+KM9)VUlaSM
ztTMYHP4JW-CieYzgT?hGlQI|T^MuXqX);<va5DkStTrDYO;*f@PT9@ec5~~^<<jhW
za~sShU~auxeLik5A2f|5))(ATg#T_gvy++C=HxKF-drS~l;_i3KK=ibT|WQU?2hI(
zq+Or(7m5C}6={<E^42uFe)7XyD0AyS37YobF9i6)rhnr|vpvm^v`JX^%lZh??V0Xp
zN@~`x{KEd$W!(0?OlHn}G=|+j3-Z|tV|bGrOs+H)tuYoO(*;cT)3{eUro4PgyV6WZ
zGpo~vHTx?~=#M`>{+BkSaeeB$pDvmpE8qPr4XLM!M9AAc&gG9@Z?Ub+y1dVXZvzrJ
zAm{*o!0vJI<FpmWw!D0X`Evd9j(5l8Jj(&?)4{9Arp({&xQ>Bz=}`S%=&L`+9OHm=
z>5#qPvI({3Aw8|U8L+y+vX;9s@XI<RdrqwC>~RSzok}}D!gW7f_bc(be?hS8yziUc
z=YXKYr+!d3D5m+_Ue7q8`CGZ4pBHX+7z6p~aHe<k&VO!Y&^sDyEC^HQDzX~WLR#8O
zTJHy}&6iV--M;OQJ;y-M0sO%B<3MTm+qAUrLgyp$WyoJAHeTk$X72_4PIYK>9<$D)
z8C&fh^x+k!G$z(tTP48d2`3k6(=-3=A4&#Y({n)30sH`t184{6ZU@Rb7eyaDA8|l_
zI=FsFD8G5cZB8lVd!4Mb`xaWtBb8tAJ^ugq;82Sext#PuQ)%xTp*;5WF9Q8QKj^1t
zdq2g_Bf?cb{v6E#K?m>yI1XrQ2jqHg3hB!G+#2);{pEabqBQ?w{r+gBbIAuKIwx>I
z&;k4ajss{1pdE08b^yjprX4SN{`NnAEvxyQa}-MTeuuSm_`Z*#E!^V1L+4exkc5F>
z*5Of$!C${=`5x0kTE+uFK42Wzr{lQ59~{dcd{1{$O$`g-Xl0$B1~>PK`I`fZtAo_f
z_pBZ7Wo|#{_epoZPpBU{>IdfKp=(|qV|^&?{ScJLvEH}ZIv*zN&u<=4+I6FCd?4gU
znd4xUUavh{jx1Je&K=B`@IKSA-e*dEzYghHH$Fn>`LFoKXS~k<(s};}^+HFzz`Sz&
z=9L5897DIR-E$Ocn(Y<GD)o81uxIm!SSZ4EkagEVpcCjs<9J%J=Tn3Dt(&(l`=ibk
zesOt*fo$v0W&iAx>ncCs9F4TS?^4S1QMvlpveth&eSW`wGbmukU>7~`%Q`Fz$@*wS
zMrXZTbhLKTQOir)JP1n5PgIj}+e;QkdRnR8*W-2P-)?b0&;k5Fy5m6T@e9=^)($vT
z;RUBO_OHLG@Q6l99P1Q%UPyEC(RP0Y=M^}wm^!a0b-cGw?{Ab653BtLD$dH-cF34}
z7^qAg_|EUmdN@{(N2&8o+jDb!^OOELEc{m1_X?msp+1%L{S8|ES?co;q(?{pTc`(l
z>p^M#c!hrcD_4B1+#g>$({;zlLot{Y!aOqK&Lac*g1#_c6MOSjvHl(q@~@2dxwO{f
z)Yg9_^{e(6i%q4zzhk{zaGxsgeJbd0oBob6zqzGyg5Q|-8%G(R$;xvH$R}IRk)Xby
zzF2zhX|KP+I9;~K=^E=#ta;v*(yvnNa!55l0PEwuq|yn}v2MNwzB;k?KImq^>ITbN
z?#8S#>+(Jmz70s^#3t+gbs_&@>Rd%uW7>6qaZd7#bBcenJ;Wp6CI?Kb!{&;A9_jt)
z6lR65&d>biX!kfQH{TyRI=QIFOO7_e^I_A^hsD<A0q(xK>`$9HTDjCc{_*Mcu^0$C
zu;2%@_Ul(_zQlTtrj&Rue)TZu`lAF4v`z<m$05=d9_|~`-Zy|Apa<Mf)B8LjY~hZ_
zw|ZvKI~rL(KM&k_zi{_yPI|fMlw-GV`(w{B&^{feg)LmN`u<f61RWgN50X76R(1Bc
zgdL5vzlU=4{5jfnQTp;K<@=hB?#DuxUpGV-H0PH{^zWaxq$WtyRpT^87ogKP8<yIf
z+T7ZF8Z=3#^)UP{t#Q&g`5!7cS=W11282TDsZxWcL}1&}DRp;XiV&a!=wML5(flB#
z)?sO{0*>Z4td!T@dWG<gsQ@efEq6X0(>vHNwcLJVZTjph4JD9L0c-sV_DiW1&;fKX
zC;)x{egJ*|#{skh&<@Di4y35BoFyDja6Bo;n?Em!-2VM}j^te<w;qswO8tWlpo2jH
z@B{FJFZn@=I<Vxc032U@m4k5BC?J)8S(6u2`-qi4Q)$6EPyiGF1waAo6%ea^hImuz
z0CWHy3<`iBfFFP#z;OWW0JH;gv;!%sv%7@j296s!jvMeh@H_B3xNd`XfR1(`B^R&+
z1(cxxv@0p~Xh{d!YTHtNVY_wp3Oa#K1_i(mzz@I=;5dMG0NMd)2cR8*b^zJ|Xa}Gj
zfObIKb|58oSb_qe04M+ofC8WZC;$q80-yjW01AKtpa3WU3V;Hj04M+oC`|!)zMfK}
zpaa`Fuzp;E|5@jMa@RA|lXdk3I)P3G1;7u$55N!LIFPFyNICjo2@24qfLzA|q?_Ug
zpabY&PyqY@`~ds_jss{1pdEmA0NMd)2cR8*b^zJ|Xa}GjfOY`d0cZ!H9e{QK+5u<>
zpdEmA0NMd)2iR-}QgSW+68wWr|A2afdNU{hegJ*|egMY-v;)u%Ksx~K0JH<p4nR8q
z?Etg`&<;R50PTQM+JTg#8J3^`C;$q80-yjW018;IfKsl5pd2Z64m!xIgY`!q><J2h
z0-yjW01AKtpa3WU3V;HjfV>L8{bP#$pabY&PyqY@`~ds_jsv0Ffs~^PmO?3@8=?!+
zmeh3BbkjIZ(FN!<&W5Enr#81Xp9W3RX*~?TOKY4o&i;oAPS$CchSXDsIiVWo59}D+
z<^r~r>t`3cw>9A@2Lv6!57<2p>}`KO9DPwM>BW5vv`>fk^CQpcVqbGWx^##*7?g2V
zmS-49mkzV1y50EeNerfymHn4pyisR227(R-KcLl5`Q{6^J{r5D&n=F2mRr84dA@r8
zU?AvVjUU*09}N5Rdw#&0Zi{Ah#JexN=7gr-pYsa7b=k`S`RPz~$dEhRE5vb*#-q;*
zM53@5juswTo9QplspmMkdG^m2ZAxz3!LiCMeuK}wE%uZHf)3yZwjT%Q4h-vGI{FP(
z3X5-z3-jF>#{v22(D-#!@nN0+;aF#f<tO@<E$|#GjdI@y{Fvvk|J;~FObgo$S?+u)
z!#xfNI)ER*aRBWA-R(e$+|_@4dh-GX^3y?S&ux;DmhGQ+*hOtCmz?kZep9J;7$~d`
zQavBzdmSs*b1GV{XCWVS-Csa`Kz-13ze~G6LjPvy-wgem+3nw~^y9bs)Q!JSuX`5*
zK?m>yI1Zp4;MWeY{`_ucuC@6$bc)76&;k4ajss{1pdE08cEH~8t4a&+XyeUTzityN
z|9r=F(H8DX?^*AczB|hSzpTS=?d}h57nR7-!VzWlS(1-j;DDe5_yHUTa<v2XCw3}w
zs^?UWR>C+RX~y}?tkPNAqs47bD%#tx2gZBw8}9+SIhJlS3be}<sQJJF>C!>z@7<w%
z%I+7`VrP%uZ~uP50UcWh+WRL${-8fE3;lWRz5nMsz5>FrIX@H3hs19_B(Zh->He}h
z2|J!*xoK~E0^#Akgd@C{aJ2paP+pjC*Oqyge{xy9^yS_}%q#^yAJf+V1M)-nJm*k9
zbk&dC*X59Y%Ked8{rVtYe)G)m)lF&h(^y-c+n#N?3tjqhv`{J6_4(FUuX`o0vs!BQ
z&3t`9h4)V}t7LutPO0*b{W~}Jo4XkUK?gDYV9JPMXXh48;(+4lAl1BOeA8)~>zk;F
zn{Hwt=m35o-EkoF_qLS!-A<)Emy+sv3T=A1lx*O>(d!Tff)1bhfzrln8_NUV{Z0Gb
zUslIO!Tpdp_d~SkD%Eow+R}l3l-Bj5wD$QrEBiq|Ip`<n2>s;1&%w{d^m7<5sO<5A
zP``B6uhQCerCf)z^4yt~>vUs2LwjW?{_1Ta2Lv6!58ya}b|9r4P}+3>ln3rtg}h&t
z>Uzw0odO(kXM2S>jusu@{*})ASEv`$^&(Zj7~0bT|1j+zLidxSH9n<`i-B^$xP4p3
z?SsF7zX<6sw6^b%ucNU$`rP8I)j_Q5AzJR!XwQe(^N$(hm+E<iSn0rXS!KMRp{<{$
zrTnno&s?hOK4Uurq;nohrPPa5{tW5B{gO5JORUW&PD_6%T6KftUY_Hg@4O09g@f-}
zIQDlfSgZfv8V3*D*ZBzsN}~g9^8wJ7A6n1%wB}2!=V)@pTW<NH=K1RVgMpv}_yHUT
zw6_Da+$T$&4`%1}PyH@|+|M~_y`Eq#U#!k&1NoNb`{a(IGcEJ@Kt7Z`Pm`_nD^|ZD
zR^nC4crvN4FO{C2aehKu=`-a1_1a%Q=769B_yN1efvv}V`u*>B?AK}Znj{WrpAL^2
z)T{j_=`aVRONWh9FLxZ7eFq1mONT}K8c*xmB9fz(>Dx*8ZCZg}B73YWcs7w!I!%`z
z){3n_kt0W1F5{kKAn0I?9|YFizv;UTcQFuj_|y-EyXI*SHT4=NG;_|{dh_|ae=v}r
z4(tE!UTyG#d)U#Kw@a}hPkh&LwD8V5w|kYZU&V5B$Iotmbw2(Kv&sS2ay@V0e}w~r
z4&Vp29|ujY*QpVj@hw&gPx$;cP`xRRQyRs-?=!|<=y3LX9rs=1Sm(%&m-@V!^aLx7
zHOuDD68K^pCl~8wX|T3mu4kAQZk%@*wH+PL0YL}w12_(#9iY1%=+U6ks>J`@;(+{g
zsOI(PPN5lbSSft<OH`Y}x&Owt@_e_!<reuTW1z4)G|#kwTiY)Vi_ISXeI{Ld@f6$2
zrhhljdhqg7EHvlr5Bg?z*i8%s9l#IZIDmG5Upo*SRqe%*vT+y)I)ER*aX?!;a7lOM
z;<5N`7zjFmAHZ<{?Et@a;HMm+LEjgD!m-NC-g}l0P(8#z&;k4ajss{1pdE08cA)IU
zEc1>BpTocu9UN^OuCbpNFs%a4bwmI6{?Ch;bymlTuZP{eg@Iqzp+(2DHT<7PV_Mj^
z|KHvj6JjwCbO1kq<3O%<;O?`DmDEnxIa+xva==fE%U<Mw;_7g2aP)}SybrKcoV6^j
zLP+v6%$E<FHtkdKOfUxo9l#IF9|vjgdf*o=3e<nn@(;iMhdChVz@i^C_CDM{Y2;fD
z=-4`3JMuXGW|`X<n4*K{jG{w*+s9)d=wOQ<*n3^FtngO<S#NJ+pkwH8{(jRVn<vL|
zK)Q5rEwrW9$Snz2ZvN&vciH$YCpjSK0DfTpIGFnVys@*-#9|=m0DkcKaqwrEDV6f8
zqA;u6T6aaJEia-up?Gv}nctj--@`!A0sMfy<6zL9(fit*ScZY}>QLxV+)n@22QiRs
z9kkKq)>lo`akTKy2KZ>!dZ)42r0shrH#6M3<cFSM);ayRLMz^UpFu4(=O61+eMQ7f
z%qpQBwXPjKHGh1@K}YUmxmaRa%#;($?_i*^bvVEFO&8~b$sAA|9e(nel>OfHyO>rw
z+V>WqypH*u4ch9xZZ7}aUB|{@p}3-ZsUCiH_G6_n@kspC?TJa4b&i?aGW)dlN!Zai
zUsu1&)jAOv2s(V~2OEAE@_XHh_c^7p;I`LC@+{7*2Bzqc^~k{;Cx5z)S);A<|HA(K
z=1rDMw?<RExCdijagN3xMttc0%d1rw_+=g5FMIjbuKO1_r7)_@^q_;A9$_Hp0Db_+
z0ki}B+JWfrm&ZOyyvqSW2k-;wj)T>+CyXrcE{RhLX`6qXZ+@hjhr`}<u3Wj%`|DkL
zjur0ec(dBrm=7EfbO1kq;{e)$ly*QV&tsq*%6h&=OMOZ;o|l&LLO#+SN1czH@hZ9S
z9tQ*+Snz{d3pM@Y{BC1M=Qp$Vt{wjU0}hy02c`9AGgqEpb9g)(SNs_V1RcN+;5dMG
zfL}W>zRib`uY;m6kZm2rx*qJ2-~X}a45S8rS%<M<mCxx%9O1-fr6s>~`1auw%qm~?
z8uZum{5LU>Z5>#9j?|z=hKO3-pKz>lL;mwcZ@!<3#pa}};~wAL_7Vd@2QmF%@seWx
z=W;yffYR!4=%;~qR_=I)foXN<=ltlOV@=+2tZ>!M4P`<#2RONzw@a}hPkh&Lw6Q{^
zZ>Ht^^%w?9t3&amDP?YF+JR|hYxg2m8os;D0YL{A{DAfE*s_-Ym-QEH9atzC15N8t
zv(?+5+jhbn&^jHO1~+UT*Wm?L3bVJ$u;|xyuQ3pG06&1^fcAEP?>HUS!<D)}f1}QO
z7so359XzmVOZfyW7pL}p<ufJn00!EpLtxIHT`n~^%>h9NNA?5So{vhEpW17eWsVx?
zqz2Nd!*-uvFZYUj$vG;O@;%ol?(fI%n~c;zVRV?YU|!0nTXjQpLE4g<Zkiq%rzyGs
zoyOU)9HQ%_E3NIS?LLN}r#81Xp9W3RX*~?vN^6`nJ;;y#A!MK4b21>*d#+(I&#;(p
zSS&Cs78(|d42v+sVzFVd#IRUuScDrEzZ(|I42$K4#R|h>rD3tku=vBUSZ!FWF)Y>^
z7V8X)^@c@+VezM7vB9v|Xjp7AEH)b!TMUb>hQ&6+V!L7SmtnEPu-Iu>>@qBN8y0&E
zi@k=$KEnbV7J9>CzhQB}usCQ~95O5p8x}{%A~d&uufc<R_3zoVf45#;y9^mH=m%e&
ziItU2t@JasvWlsdRZXp|W@=@1Q!BqVwX%jTG|%UmuViv9DU*qPNuf-vC8aX4mK4jx
zT2d|(Ye~UOtR*Egv6d9g#9EVOt8B7tl}(ndvdOYlHd(gHCd*dYWZ5d4EL&xhWvgtm
zY?V!x&Cg`n{7jb3&t%#BOqR{hWZC>omd(#(+5AkF&Cg`n{7jauipjE7F<G`MCd*dE
zWZ9~iEL#<mWvgPcY*kE_t;*NB-Y*Re#4E$%wPEpwEV>)o*rCmbPA07<X;eoV8r7lA
zwWBpiJI4P|LG4(L)4$ug|GUEk?L_S)?PTqgF*<Dsv1wBW4-Fa`nqf%y!9#S0pf1`>
zWq$tmzbxA9+MJ)a&QAWFd?xF(MMLTh`3<4NWt^$Bk!_rjP^JrVMnc%_Va`YhD>&L2
ziEW-F4QuV}G-o8XRh{dMShx74&WJS|`@1s|%lfW!MnYMp_0CAiHjR|eTDyMH87aZ;
zWOhMHv14DmATVh)aH-{jXtw&_xFDfzTUQq(mo?V9AR6q$G#3OWt_G!6yC6Px{jduX
z)2^O%K}xlWo*9t6R&H<xM5}cq&(x%_9@SitquRnIuE-u+>hFrIwf3`JkvR*Eb4B*r
zbP|Wu_Q*90qQQn%$%0t7pGX<NKm(e6U>NaHF!41Q+Jv-c3bn}OEQqh=>X{W;YrDr}
zL-yMF>DiDat2!$ivemNe%ZAKZ*%R53P&W5%b|j`9bIO6hQZ$%fC<hYDPA<uTgtX<m
zav-JI!Am(1AM0D(4asT!Nct>Uw}Ea*2y<TPhQzecwQh(;3t#7kz^v2YccK*-Xh5?M
zbl>KN<g)0UZV1di4Py4VAuxG0*u2*bDaFDMk#n0Jw)CVM0!ynwyGS>rB#S-khQMss
zpyx$51ZG}?Z?3o@A<dt}M~h7&c@fjTxlOcV$c8>}LsHr1XXN}0X0-->h24=<HdEz}
zXt2O??g$JtpxFm9R&hsQ!fJ54i8}(bUxWSu?#Nne-_9L@(XYXYAKVcw_A=NVfq@28
z`@rv{e!xHjihTfWaz|ia)*y77J91=;-|vp3G%nH|fq@1z`@q{mc@T~Ea&sQ!sCG9p
z4-(4mk+{s+S7-7fFwlTzA84B?Kcdx=D&$9eZQSMj2rNHpl`M{JEa8FVwsnm?5STO?
z931C?z&_Vt;Smo61{%=p14kcwAbV`t3l9Y5feOE7$D&IVKwzMO*gi1fs{%-Fn^Cy{
zqSbzwQUHO01~mJ?s*r*ROg;^kpDT!1wKLZXBB`uHHBTglWo+n)<gjUNJ&_m|(%BP%
zfd(}DK%>5%NGfYF!V_6C-(XJ!W?qArOFfa&EMz5VBfzNAV8M1zB&Yp-$P<Ze^PLJI
zmhEMmLWm|i7FY;@iK{`OJ%x}EHu`)aM61RBT?ol(IT8yYIjlu)FC?TTczGeQY-lMj
zM1yVo#tVU^Xizq|2m+H{gP4s)5SVEVX5%6V?8_R|I8+3I6+t_ki!x*D1B)U#4eN>`
zC0NM1qDU&sajqx=b4(2qx2llT_I9rd*=x^+7ehi>iJ8R^7-&GV4-5<|hQzR?4T>XH
zY=8UWhz0YTS{zwxRb~`NwAq^K-pG>esOybbv|mPfBPng+6>r3{ZCm7nXtO69OCVb<
z^Q{s{D4Uk0Bmx5sX!e0tvr8f|tkBL<ND8YutQ-O}uR-#k<&hLtxpM`?s!d;A0V&12
zA}Sy-+coHax&orfB1nF~Km(e6U`c_BNN6iJwIUMB0_ImlVA5!?B!c(`m?#=NZcqt<
z$*;ka`;`!wJQ^%~SP98twVzZ%O0Y~Nd=VIEK(h}tPO6N=uu56|kiFK>!w)IJc9L?)
zX}Tf0AZ<wv(l}|Hrsx868fU{&n^Rj}TdCQOj%{)@|Mz!QZEkHo4Vt9WdKi9}Htg=-
z$!D@oTQ{U$TyZD#-(UESN{w8|FZGXILNgJ@l$mut3t@2VfdV-QRg023lj%>9zLn3)
zJcNPS+dRrohz?Kk^&|)%vwc`$LUpzKr27F0|I{Oc4`J{hO`4S=R0WQhS(Z>=VC|*y
zgn{RxynG4KFWI_QAqcNmdrfshb#mYMnuIZ17FYcrVQ}BbQFRGbB|hwLK&Zc7(X|O-
z;M~CG8bb8#Z@T6L;h7IawIo#Q6IE>pV}^V8{GKq_w{=7ZLe;~GNu3Gx%eK`BA`I+s
zZ(Mgml-J{kmLU98<LrG2)suz?3?PiDvu6Gv!r*_d-WW=#+K{!x2twl@ZF<p`FUXHL
zelC`-V#O<UJD`+uNY$R{d)DeRnur#t>bq&2LA<MBFUJ#v|8~00WI}bb_Y*=0W1K6V
z2qg^O{e$~VLe-F&ZD$ke%k2-FOBi@JA$9>FT2j2sVgpnEXlo~FJ+JVs2Uc~aHQ%)6
zOKE;<sr?9V+iE~KsSI__iQASF#vI+6yoxY5_-@^`gsK|(r$!L!6B<QsBn(_TEbmrA
z<Dby&j#%+ZeST#9ykhhVs1MujTlk(|Y&}k)Ug)eBtJ}}6tK0uYDlxp!)$m<}>I+%!
z?<I^0tyz9QVQ|AfgAWm^-Y?pDlu*C@)Q1y<fdk$(j3h**D}<gS2#@J}{sN(T!Hfb|
z2xD66J4X`+XH8gfgHUy_*rQv7`cVPCcL@WlOc)kNh@NcO`-mWX<sFxJLUreS%@PS?
z@;93Kf-v~ZPnTX1s;2(o^_Ebtz7q6-Fz~hO8fU}vjLkLTGZKdP?p-wtq1t=VsO*F>
z(I@xkBn;NQ&76l&<zK#eenNf5&boqxf%|@q@*+gTuu4T3Uh!EEZ$fpPDxxG|OnA$r
zGK9hH#@8rMX#5lUJmWCugH!=&*UwSw*-^`F?ztf70p~R(oY$oGW2w^<{QW{|0cbx@
z(ei#os{0H}*R!<v9WD7_<^7t`f0U-1(#j+Ddsd?#2>o8d`25J`=ZZY<i)F9()C>rH
z{^DCMzTt%KZ$L{v#BP@zrJUCG?*KjMxQ~N+pran7J|8;Tb7EEQE#oSY=f(OHw-5Uf
z22RMEy}IG~aqWfyH3`F?51n6!P#v-AhMF*@`(>X7guz8z`!pt0UH*DgGeZ50UN8L#
z1Dh<Y(~=N5ot)5`Fnnk7@$U%LgUY#gAdD&7scmP%;M>!~x)7?u_QiH1)VGc=ttAZ1
zp&HPK5FKf;tv_M-*q@VsBve=5Tz4p8OnmH=;e^4f^F)p!RCQ^PcO0R<;L!Ht2?Nis
z44*`Zrd_%>l`y=1=JKJ0>bKPg&mfH1s@*w@Fu31>4|51rB~LV5K&Zd*W?C3w;JmNS
zFC|1RIu=+?816c~^D091{=F;K5XOvn`fxpAaOI-D8wpj9z8$uOP`^BQ?{>n#j+<O|
z5u$vtP4^OppUyo~PpF<;|I$IinEws&I!YM)az)SyLe<8Lt4|Z^wVC425(XBlUiAVY
zit0J)GGX|v`TL^?)tcj(ZxF_~yz#$97`&&f?hc`9Xosl#g!=NKsz-!@_xAL7N{E&|
zS)V`{{#}uz=Y(qaZ)?0Fj5#)LTry$s&l?ZFCsftEmCf0igip*BkdZKOUA_6Pgs9u#
z8`%iMy;k_-BvfC#&^tF_%=C<#@(~6%s`jz~q3T1AI$nhOzvfLSN*Fls*zw|osLbn}
zB@Imd6Z-njVctUs)ZYo(QDNS<fj_f(T_9JziIuKc@!Hz13Y8C`;<41vpKm&R!wLOd
zApQL~t@*OLKVWJB;QGM2@f}Ke?r7;gi&ej-J>RtFkGAntavg6{*OS!g3H~13-f?!R
z0>F7lobwQE{zY4UY#vt%{;G&^u~uIv+uOd`IF4AZ^PbspGd$mKTe{A)x8FDO?Fjf4
zj3*%P0W6ItfcHf3o(SF(k@sB!&+`H24>sQ?{5PJL((4ji$D@HCf*&ej{D{5RF>oBv
zc^t@IFR2-T=Y!%rAGCJf9@_h{$nK|%%~vP&b0(-CI@@<k<K<Yr-UfdIf3xeln)2%o
z@Nd)p&Acv@`yK23M40yvGw+R*S}w8XYnJ-@*W&zylt0f@s(&f9Tzrok(8Kg~g}Hg(
zp&em!o<4Z54)4|By}I!4)y>V5XX*PH(2mN}j>7pK&i8gbClh#HUa|a9taxo5x0!w%
zA@oDGa{hkG?~iGje_E>RS~#9epEp1k)4IU@v$E#7m#UqC`HNuwBKZC!nFq?k_b1{0
zS)Ti6d%qU~?I=8lgy)d(9Fq3&w{TzM;P*AD4xrS2_~73T?%xjbeLA?_gzHVX-lXSx
z6V5x1dETM5-udQ>Z#c2%X@&Qz@P1Xu@B6~_6I?&R^%FhUPsaBia-DzVN;mLal$Wm|
zaTs@AW-*b0miY--^9R=QtG`~+wF${|;M~CG8bjvIjEA#z{0r#dnD=2~&r@pad|=_d
zh525NI!+Z)lFTn@{8Q?4`(WSJ5gkY&t^Lj;^rM(j;0n>p-2A)XFVGIsG#)_hy=YW+
z(vHPMpWF}k@#On|@cl*j{vv6lga2sKtd!xr7dT>OS;Ei0>yg@!!+9U(&AX853kHBc
z2<s0<X+c+Eb&YYZcp{XDtCaaERR@bbx<$&P9|irC;d|BO`4y$#tA_S0<^B(J;CCMk
zy7AL3P}O(SID>fb-SNGThz#K?@3_PhsypXvmPjb~bJUpdmPuvEbBN$}<7<>Bgnj_f
z4*>cB(AN(D+R+Ka;z)am1fG{g^5?LIGLS#Gua@?{Iz@Hm>CRfYSj(@y{W4f9FKhX=
zw|?}?w$%tCg@*b4-ZhLgR6dLs3#du%gTkK=onMDg9kJ?$nlPsOWuFFw!9`sAG$vGC
z{(4h0Lj8<hFX6c#z0dvB?%&pUWvIVn<HjW$>hI@`hu<6OujZ|6aGzuQeU4D)2chB#
zAG3W}VG^0T+WoyMLYTMELEbN=IuP7PNqZj^csC(-fkFEv#jWa}o?1<Czk7=AdZK-v
z7NtKoclhHZRRGr8bBC|TyfVKhREQsiXFd?sl9WZQPgJ!bv~}E1xS8_}_@~VNY2|ej
zoR7;l&u@6XlCiU{AYtIXU!%MT(J-u18JPM<%lU$~{HT-r#@8g7qJ5qShu4iVKNj0>
zXs>VD^JnfmrL5*#=<&|R>mJYpu5(O{_XEEGzcA$&=Ee=PKCS`$9sHe-zk}a^-|+F9
zz$$6`X@VbtAARaa;1A;XgTw2ZDgZ6d1I2!hLVLeOrTU{%%O%wNrQm<A-WW<mQf<gu
zVg#Xa9!nU%lJ>et>HQg$)<4!!j(b|lNlQM`f7hDUe9``%588id|63Z~+dG|{(3&uO
zXY%py2;n=p@Eu$hzJm+xK0LQ*K%9W~@rGjgsaWwUZM=?9ItdjIjE_29XEG5}-R%8@
z5JGeF?<rkxrI(NT`O;G>0G>~Oc^s5fekxTuX6G?B^L-my^99cr;Cpj0Pd%AG0OqNO
zdFn|cZE>D@Gwr*%c2sHYzS7Dg^f)|tPnxje25}Oz^M|K8kW#Lf;dl_|`JuV{d40M4
zVRMb56Dq}zl+q3vs|M=j%lE_3k4Bj9OMt(9x%~xwzN`;iuUWWWV|`xc+6`geM?=31
z*Xf;C8Tw`H-@9TBVa$l957!e0S1#(ikx=#M+hJP>^~;0zZYQMmdoM8WW9>E7NuzGA
zA3Ce;J*)W^Yn(fL_xH=+{bgnRF8DQEe|<VmPQ8E)zXrbrzh%*H)6XXsUh!EEZ*o1W
zj#EXHB;@;DEI1zEcyPqy!O{HQ(aH<sjV*ka-p2R+wD%*o)L+*4e$CeT;9*|VFVAZ#
z)O84aw+5bH=5{96vC}hd%10R7sM^Z{gsKlc>Ua_A|C%?UC}H5hW5<gVqB5^@mNYQ+
zkCpebwB(nTe54<*Y_7kV($7B*ULSMbZ%<J`H$)dSMyGAoeDD-qkhY}8N#m?>nxYHP
zX`Bs9ZBA`&Z9WZ}q|<sBewQXYIQe(-nXJ>c4XM}3$0(z9bo%t~-yol^Aa~VY`m<rz
zFGT~}Xj1?xmbMy!{D1;hDFBDARWiX&pa3WU3V;Hj04M+ofC8WZC;$rBr~uq;*(ez7
z4hnz*pny;cKm#n4l3$1iME-&}u;~*8z*9aELx_L>6V$HA{)@%%k6kf+64+9tmbQNq
z75obdfC8WZC_udehTch}Q<AgB<<m|{Q*@nlrL|qP-Nz90{Jevb3+b)YKg6)lkJ>@n
z!QBm!3~fGivY|gySM5kcsG-fZqcuo7hID1pj@3B*yRG}bJ510{)K1b))=nY4qC$vG
zn>u)CkTz7Csm#whL&z-J?4`fZE2?Nny&=CLbhwN&l{T`CGZM;lLC#1ByFJVq31J0C
zJ0r2plcZs-ot@^4#I~w)oe}F6ztkDAW@CSMMq*jtb<RjA%e3AZDcPox@>y%wFFGS7
z*qzKSNGW#gYZnA2tp+Z&ToBDx{~H%1v~BC^g5<KsS{FoveVFEgz{J&{)M^*R$F3iC
zL1Nm~vo1)fHqkQ!ve(KD&VXpOj$1PzDXd2|SLCR+u!$?O$CmoLB5SSvY*%E?LgQSK
zy*8c1A+<en&4OsKp;fXV*6k-!1~AZoW*-<vd=yN44Td%$?U_O?GC2$4Yq@%6Mb_Hx
zG1-v4c7A#`WXY<|%7$#UEc>z{b5{05b|jR|eVZMLX~&#$Ag~k-<`>F=#Ilo1av&jX
z`K}yDX?E~Z4#da$7I#B(T0fFLOV(|m8xq2tSGplFEp)9LqS3<Fxgjv?H29rp1qK?>
z>;v7mxgohMdZ!x#vrmJVJ#Gk0UJW+ybwf(A@I&O>W``|3>4w14YS1py4JpZD&$=Nn
z+coHU(G7u_*WjBgZb(S;C-Ko@lSp2~v~O+`?HID558RMcw)q)3KZ9AVfnQ;FB$drn
zxg#1Zu$(&r0}W{Qfs9q$5ty(V+-~BI!0gwce}Fr(*4npoM_}}8aN-AdM2o!)c1K{K
z0o6Y6JE<Qq(12neK%3kV*q1d3-R6!Q+2Z%RBPorGbVp#I0nI+}woo2KqrKdm2RW+U
zjm(3DvU?;hbN1Dlya)_5pxFo7X3CFfwWJF95nmg3IX?o+k6I;*V;f6&Ah~T_V-Ex-
zjRptDc_6UQHCTAW1A&1CH2c8O#~#QYTlT^Ofq9_9ui3Ha5(N+#Xdt!^O!%q*lG|of
zE`VsYAEp#QV4wlbKCmjJAOe$5gXQN6B3A9p^@2z$>rl-TiD4NVdLlV&T3b&fhJ|$Y
zL|~u+%|6hmuP2hqT8!{SmdrQU6M>o6;N?<Jq%;dzN!kc7sx(-z-4n@ae;@KhV%vPD
zLWpI1*`^Sp$&LjULSW)*P-ssfB!rDVUkK4^@qZUWa$1hWLP!p4k=qLiX$f9lNGuy#
z$_vq88^7^FU@01u4K9Mfq}L#3V-W;qT7%iR2m<@E1~m>9L10DD4(FoG*!sYtNKV7L
zqDToAvaTqS%5t15iohIGgT$>WB(=TWt3vkLv*E>%P*!4QF$4x0(Ch;P!-^p>Y-xky
zh!xx4zBppR{H7L1)>@Sr#Sv|`rn)z>WIO75BNpwKQQk;OTX@AAv25EG`5@Zt$;J}M
zR?B>=1QN=oWhsfkKm(e6pw;Y>NDM2qvlNoTstzlMz|3oq{AYP2g;nlc0kLY+S64tv
zF|UXU2+Vd3`k$_VXtD^BA285>W*=Bmpdu35%1y0^#Ik_-6%m*;8Z3z*z5ynR29Fz5
zLSXW1@Z^3a1SXFL3m;ZOa#-yrm5>rFQwd)L1{%=p1C5g^BQdN}RzGB~HT3X9O0b=z
z9CDg&h%QK5QiC*38mB3`0G-C!u+-+%=GNxZph-Hdhv9c=!)E_ZK9hCYr6Kj=iaQ~p
z|F)~Q<qcZD1uK>I>h)W*(M_Kg*K<PC-h9y>e&~rFe;=-O8}lV?<;%07ZI%2ZUU9b4
zTsbb+7;`kY_a%-m&BZJ9`e{LzXZgN<jK!kR`85~cwE6|zo1Z+!0Zr>*PB&ZEyWg9-
z!D(e2W}Sh<G`j|5ypM&Zt$Lc<fAyX=cS}u)!BWv&K45){0)^`J%zdTSCCn;^6pekG
zJn|}*ngxdY>@4elg|oyrS6b+ce0`a-mSX#5+hDg0&$?g1tTCa{w(T!w+{aSW-ujq%
zdC7s*s}EtIusXDV9{5A#n*`2Y3VnYhSA73_?{;@i(=!~a>^h@TSO(Wsm{rc(x9vsd
znNe71+S~qE3cq^mf#n`wJ;$um->1^=1<yRh>1nC__HGX@*|1vJ`Lozw+M5rd!h4=x
z8XFk;l#_}=<y+`@a@ERkwfdbKSZWGgj`Z;z3m-eNsdqdE%A*6{=Zdt18+*1^h-=H!
z*b>wG9pB$IZq>|or^RJ#iG1U?H{5U8%XOKyErGL__T~e^%d0@+=*oL9P94X|MSDws
zDqo|m`twIHYdq|Bd!%aA8O)dV>Lyfpp{_H8%8$8tl4tdsyRQFZ45Uj3bM=*eeTIc+
z&uEU^#zNDWPJyY-YHKpy;)J3xULYu7jRJhn144(hCa<s$e+BA}>eaSzn<UN>`KMrA
z$ky9WuB|sLiW8f=<9%M$_J70)#T7LIqW8DngIVQ)!n!i63%uc2<%NM8>b-CN563F+
z&goG0!Oq(pt9<%km1e{9=N#*7SR!Z7;6xX-*tD)SMCkPHKa<~GL~6c#aPUNj){_!B
zxjCf!;gSOy+~I`cfY5VAYVJtpgyPhU$4fox^oCO!SA}J{c`tuFCpNA1XM6XX^L$WP
z``@o|jzX8J6}?|X-r>Yz`uZq#!P(2{&RtGy_URn-B=YfNPH2{0Q088-d>PboGrg|x
zn+*l3hvHiixo%;(DU@2OYjZl~QYUh(Qm+0#LZvJ8^M9jX1U4Fzt>(iBjujg7?WnA?
z)^+A9pO=_bo*FSIbnJ$gSZ-SDKOy<z*JjI~gzcrZ>4~+U$4^nU@?Ia0<zo7BHC^!X
z?98zTFkiCTFW#8%EwdZV>9PGfClu4`YwhnrZI&)?9vaOFO>5H^YCqQV&sT>c$!>X1
zefN%&n|$+cZ#bolzq7ZWs$AiPy1q3RkG=kpKD@Qh30~x!P%CS<mz>ayI5FHkAod++
zq}cBno}PO$d2Gy0PARlBu8a2VthQHKj-Du2WY{5&Ri2*eT`QCS0nSmGz8zG`@0L?f
z2X)){CpJgc%2{Dj$JIG@J;1C{DeqCldT%baA0GN=itqImu^g)u>iw(K@nn6|=1D%}
zq6S*01K;C7?Dx)6#ecN^z=AGQ-*U8(Z@mZ|uKq0Dyk{j(Vr#jl+v8GO-=4>|(pld#
zX-v`6oVB#IUA6Umxcyvo)5IDtIM!%syqVD3h4#1f@xgo$_RR;8UQe-q&)a_P$EN|#
zYF631$BPO#o4n>&<&mHDPCZ_n;N+s%{#CQh^d1N9KI2&B<Q*~G;gK;M>-@U!;OM@;
z$8$ncEdSY)`Np1pU7lfkDOUdMjdxuBqFY9J#$mb0H~&#CdX+xc1uQrD=70F?D&I%t
zOyXGONuQTNO$H`#Vw11FjuOt+`x#pI7rI{@E%8Y;4`=#x())k<?Na|1YW~Rd<&x|7
zW>>ev*Sp`vax-C3rI@eIT*rJlwxjd)r?<~@d|5QC!>Q4W|Kcps+W9&8?kChZB3j}x
z&L2R_yhN7r(e|s^Q9Db<VteUbW5u!WqVHg-Xw;ci3e~PUdOUKSM}$sS=;yQ$pLO%G
zNu`s}{vvgJtUkZudwoPpII-@Br1qcm=|!#Dv~9#s37B=JpH~-to1Sl2EL~XZr)6t?
zjmOQE%sr#GY|W^)#JV3U4=XXjNo_6L4=)>hsWeiXW2wot`VYD8thSb;%RYWKW2m#*
z5{G-dA0AuBNzIqdtF-m0IpuH6mw5|6c(?z>D~>hxes;#&d*EY^Res1^<aA`YlN_rQ
zdp;|%<6rG}^<?RgSdLY?IWGxW-eD$Yl@H35TGtdW#I&&P;7fUPyuQa-OH0q6t^H1z
zrRP%Sw%0h9ck03UPqDS!*ZRnc*$p0IN1;&j<;oQg>-97tKKJg8tDa$2skz*EXVRl*
zm{rQ9yYV_*Z0=~?^Q+`c?rK(;xnB08J#sjyS?BtW_ZqG0{*DuyM*sOI`E<+i`4X_b
z9O5_c-h|`vSSifqSu9J{pPpe>*tGKFtcjJLVQXouf2p|U(X^@kdOgSXvc-}1hhBLl
zVM~;%T$a*lb?ne}_25gGHEJiiXKhpcCT5Lsmx}eCxAry`n-%})*8R%pi(GmZtzL6O
z&PfRzUq0Af-LpZNdmQV$sP%Gsn)rZYm46;>JgMT)yBw?Bv9jNy8kZk%_A-6@rfwPe
zUA=12m{mr-=<7dbaU>Rt>Gh)JcY%DTLoVI+RI|!(JpUpJ%&w-DV%4)#scYYEZXb{N
zQmlLl74N*^)y60G%B5zNgS<ojiZ7f^%_{%+ZtK_QX58XfrQGw>+nfGRn<L!jlzPvx
zO26i1%NPIkHODF|C;K!Wxz1fJ7Hw@OeLJ_<Sg%ncc68dRn^4d5ZB5tQ{;YmS-I(vd
zy%p}nMxNtXWt}1&9$ngbma~`U>eslwz}@fv_$@pM(?Vmu`W|#ySmMqj?5H&AAQamM
zyJdLR{R(E4*OL}3UfS*vRtnGkacl2`k@v8p(ULz~+b_bnUqYj8+h5GMkAYv-VcD<u
zw(QxRfLWs@e=)Z`cO`Z9jD;RzYuRIQskJADJiwON{`NK0tKc1MD+lH-9Ff22Ei5;4
z9f+vcZpj18Iy1U<UGTEr9n2bCPYw*d`aTM?&Wl%m9yx5wb!>^WUT^H0dQ{iZ<veDc
zos-u-SW|WvwnTI7B&-Jol%PQLq|6tRu06!861SXF@W;yu*j`Fimn(9-i#}a2hGU)f
zj_0Q}e375qFIycKi)rJ8#QKf?&hr2(g_GA#?>j!zBg{HS6-W$f*ZCx-jZ(E|w4}5D
ziMH;RH+r0&wB(QWe00K2I(_cD*hqWxJ3McZO2;?d!Aj$k+xK>kz7~yH;e<X5uWf2~
z0|P+^G5mnm<9pqZrP(fzh{bZ#TszwHc=H!~n#{vOaZT|y{SN*?`q^7ak3VM|s1=li
zX`$HTlEsey>#pPXd3Zg=tnu>fgLO{*?->@0<F>Te?%eYlww9$5o;=MIc@MM7ECHVX
zyxDOV%T06bj<xH1aRH~-97w>l(c1j;-Ou>_UZKc0A3}wbx__ro`KCRdGT%M<Y1z{^
zm^Iq_opZUvSD#oeeDSvDm{lgv>Nj^?|Hqhh3jIB9sp303(`n)cpImCz`DmM`*MTw~
zYFcP-dp#?|f}*3^-Q{Sbv0V20r`+>_|61$h!XA@eVAfe@t?SHJJ})tA9D3mI-?rWS
z2lJ)T4}p%YfVJOEg#B#$&ZMn+U9K_aXm0OI9FU(5j(rphUwXdZ$FMh?@Fhtb<BA#q
z(feEP!AjxZZ#&*Ey6YMiis|POxAi-@O5Z=NH?jZg?whxlxJ|K`R?79gK&jJBf82@v
zo3H1JHNMB%{LEQ?`?b1|lbUr3_5FBjbu+i0Q12zu$0OBue#HLHpuOXMg$}>1Q^?ai
zE2FVctQd9qT>aNKu~0PUXJV~?w`_L7Qzb8Adns1_#E#dkLyMckTkEmV6e{0RpGSpC
zZ+5Smr_QywgIVQ~4rOlmh26qJ)828?tiGp`>N|mAr!({Nk^`$(AHuBB(sw__+AdT(
zXl*>S%>Q9;KGr&|n^wo|DQ1n|4~z>w6A_Q?rM2~K?|wTMu8GX#c9CO^4KfZYQtkI6
z9IKS-`-0Z`hppe~qHTYnuG{2}XX&=YbyF%nz+%yoKP;&1+VY;yb51OlQ|(UPe&8X;
zI@2G&mUQL2-N!fAl*Z#;VcN;Je71%YyWhN6@n>2y`R(({3plB0tFBV*Z=COmFYAu#
z)wXb(B#u=|-Tzsr^mfPlysGX0h+~!ZzE_qz{OPIQwKDl1;8>?W{`L8R%a1r)nZA9J
zs^6g0{rb|EQ>x$TpMQv*F5h{M#m<+|@&8dY$AfR0ZsAzvk)QQWJzkvPfYR#FbivEB
zGshmlKzVe?a`Z&GBEt@GtTJ=!Co8YKJ%-Iu=yp!*__m*mZkkx*1t%3*{f>Y!-(L^O
zRCvt0WX=+Wt~Yz<b++`qW#jf5w`yj))8aDb%l+kdtv_})5%Z<F?;;rWY#d#A@5QO(
zIH72a*H%6+x@KQluV+{(JQhB7VpH#UY>u|-BlZ4L^=o1+9jWVqZ+d*cW572(Yr|bW
zT()nOU*BS&X&tQ9G5vn5_w%>)zB7IK&Hc8wpT4<x?Ooq$;<kOaU9+nB(%$qTJR1rK
z-LDPe6Q{o*=*p(9Y_tL&bGfd-+70Uyc5^~gu6~jqOD6AsMY2_$_c-T2n0B&$e=O8@
z2aWMq>opC3Z`f$~p%<L3R9&0XDVI8tlZw`s$JYIR^p9x}Rd*%FD#gCPu~pA``?kHv
zJTnTjN}<L<i5-uv^<wY-)kE>Eh+Ma@Qkkn(hO5=@+`z2T-f=mW!rSXVtc1^lNB>pL
zHv!v9WBF{I&(PlS)YF14&+>i!7_&xu$5+T5-dW!>X-v`69P4EDyUgkHE%f+Tsp4bx
zI!UPU=jr1ywxeRv-uBiQUa0&@{hm*#e4C3Wu+f-oH6KQBKzVg2zt`o}Z)gW*mFFsZ
z{yOO74bEDc^GD-)=!YjyKRl!E=K8TY`ug9Rjc)q1xSlgdqkgQD*6%Y~n=e}DN3%9x
z{yvp{FL>r5rk%$2S}OMbsnF&RC=uIBTkrdY-v4N;xnbVw7noJ9-g;oU$5+p>qf%&p
zm+QTywdo37UZut-)%XQ#^*fcX(N_KWBN!-;4#6cGRtr0S7PHRu;}zj|hf|{$|HaXw
zLydEJryiXD6a#J3!P<7z(tc8n*PK?rpnLO^$2hqu)O@&n<9S>D)PvT;o^Zev9W41L
zt=kW~-5#kLbq3qZIrURZs2idS(w5ZtYAS1-rsx868fU{&n^Rj}TdCQOj%{)@|Mz!Y
zZEkHo4Vt9WdKi9});MXL{5$zf)@kd8)Qc<b^ii+y9hDlnke})wyM$&Uj43ngd=|ps
z*aHP}5ULg>b#^DzxAIw;hcGaEn@9Nx(cwwHo&@1zwht>zsIGQ@uZl3{sYeDM!r(ue
zG%H1@3LG)BETO)@+DqjL1J6Zy`4XaEvUROO5MHnLn(Bn=<i7DW31hY_uKGX1;J%Tg
z>Jq9-eAwTBP=CFmYZJo2xq;0!gy`Gfbj=CEGara*NvPH*s@f374EOH&Jz=nK>xd47
zs)rMkIuq)bZL1MP7}(+7xbB1~ug4KBLHMc0+4~ZzCk+o6Kp0bJ&HO=x!T(&nF_ciX
zA!~^dgvLMG^r9_akRNgUTr6G1idX7(Kq=*rsy);9tkq{U5iL;Fchfk7cvr(-jwcBJ
z?R1^Vgz9GRCxj5jI9EIoN*KKR2lttTsv$Gm&L-5C+aETUFz{|d>;gixq<ERd2B!Ye
z)=tuTUg290tm;f_zG=;u()`v^`w`x@)qrqP8S0!9w=E}(Il47@6=87j-MVWDRW<TY
zjUdz~G>Y6v7`S#=-mQejKcU+lvEr5b{K)!w#poAMAGY1M@IAlSdYnSN&{;25x1U{C
zxBrV&VtApe;kyXc7qZ;nOBfScv;2O-;D&t$A0kw}U$pZmp?>?R4<`r%2fS+-Nr*~U
z2t7v-9@F{!1w!?L83nEo#<bLTjwTGwny}&qq3U3<N4E&|qXK;I5(ZY8Ff5J`J=wDN
z5kdIMJ1+5r>dyI^B@)KuZ#44-VepxsF1;dDP5r~`Eumh0CFlcT;A_`4&W7h1n`^{p
zBn<D}yJ{9fwfCY?*$HE!Pwvl27_57nIS--AzkKujg!+t~bp;6n_x&2>MTmxBm5MOD
z;<Fy!gz7j|L`lM!@Rmts2!q>=uTh@R_$Tyv#$nC}sRGijpQF~Zqn6v;b3xDp&TC3I
zuSxC4Ql}^Q`-Rj3(0-nx<^70M_ZgP1XKC>}TJpil`!%EgC`~t|l}GINtVTZ&`n`nl
z`H{`f6?xtl%U<uP84&vX#kX91!wKErfR=oS-7Y&yIj!yA0eaAJ9|!e7M?FY=K6JF_
z#H!p|##JKEi}fdNANC^*oRBwrb;I-H+6@D05{5q?I=>E~I%3rgHDOHm%RUVVgNwNK
zX-ufP{Pm`0g!&o1UiuRTHd$DwB_VP;IiWRS_|D|x-w~<@m2>Yv7*n=W+s=f+x2K17
zAykF!i|t0JZyjG+OBk3#HJ}e6I?`fWf5PywKPUf4sII=b?oh&*_}D4K34>SXi5x|!
z>e3+ZI6{5Fq3y>L2A*FTK8X-byL4|VVR-$_<wFV8Z>tZUK^U`DyK@#{aK8l~<`Ak%
zo@ls$P=Djiv@pWJd0(AhN{Cu?EU=s~+;w{ARfOvOdsnO>j2ZFt;d;X0%0+!Q5~?13
zJ8TP~etGcT?Sz3HH@WO0MEPQy?j;OAoqML9P(8W+rGtbq{~O|UlrZ?^il7sOs*M*{
zpC;66GsT}J3@lc?>IFg+)pOKk!th!1_eT+`HODjGAdGQ&<9~}Vcu!f~9YWR64pH|B
z_2olVj|c<r?dkE95G{SOK7la&yCO-?3Dxf3)_6r2b8OtWWWwN|Hy(aZsH%A@o3k?s
zpO`BkBVpjWdh=ZgQMbW2vJr-Rt?<c7sJ?iicW%O%=@~cWBMfd-?PUQ%)rTH+ya@Gw
z&6`k^FmT|p<HZS4nb$c>8kqVg^!1&?yoV5|zZ0~h!n|(-e`fQ#K(2ZdD_ybTwY6Op
zDj!0{W2v7%-*otf6Z*M8`ulNO^JR5^z|;c3^?`NcJCyR=(b9butA0&;zG=@NZR4rr
zI^LwNC#lmD{5`n6<Lpuefb)<z=ONnsi?;mOJgyY{RT1N2t-emSw|%p59I;&IJ+tFx
zc)s7Zbe(B$zi;N-5%4P*Pe9%SSQ<|N?}^|&5xgfN@4Ez^=L60kY`#zUZ#*xh*Cn=&
zM*}|uKUBi_5qqy=;5eZ3IFP+wQZoS02gP|lXzjc`wD)6?-A@^tuTJXcOi({`w(pk4
z%dvXB4gLoHX4iE!<<}kH-=_VWd0i;?JJ$P&Fz+8`-Ww^kTw=}FEcNxT#rX*-f1asS
z|59qX_#QW)hw1AIbMwAKJHqBXeehl#-mAlVb>ZKuo0})k()Tl<9hIjYh4Vd}@9la{
zCh)wxV)>(3@!C3WGyOP1=!b0O{QZ>QAJa1bv{cu%a6FklZ-6eQb%Fb5WzBOhRXYRo
z7s32R@cl_L50r)PPs07PJonG`elG;tQFsms&mrMCB<<sG;l9Sf?`u*WK&k!k!M`2c
zza8ZJba1^1*PC#?Nze5roOc}ayhCff^UW9EaAMEX3h!6p{i=}P_l4^xxPF4`Cwi`*
zjPE_<I{(O(Zs55nFJD9AFz&p}Vj=@A^AoV<53J=^f4!n>6O!q`xq;0!hRmB84`=K6
z7tp~m@598Nr_|Q@z`}bA^SvB(oGPLunP1ZQr_|^6!M?2{I*>wI`<+MVM=_(o6{459
z`FFuzpdF-XJb>DJ(WvaC9gB%RxgYN1$@l-@`-|}XMbbzI|IwsbDZ_a$aKy~Agr9xa
zBefxi^FGX*cOlgm3;=%+)*p=0g08~q8sl8?L?{tgDf3gR4i<ZKi<Cz{3i>I-_o~VB
zD@wmt4eeRV{U7MS?>-oG<ELAos_&+82Jzs#<9i<w8NyfIafv5Xch1)=kx=gEs4?L!
zlgg0i5W(%n*C<a2{Q#gJ0Q3W(uO9%kqZ5Y3k@gY^JTHsn&tVN^Ab)UQE$w}Eit5VK
zowahYmS21OWw2IW*79p_{pgo%s}V#B4fFfGYZz&$d>AhlP?OvTg+CuUzYd`~V$}^b
zVNCbSJ`D(ii@5e_OsKm2^`>To`Wd}m!gD`*pZlrZzpe4gP=Ck9jY~Gv-_IKlzc<uh
z&0E>vKF9X^9HGt+Ld6q4X8W+hBr<ii`+HS{FmIuQykAOnAh?f`_C6}`ZbIw=gZ4{`
zTh%{3wVL35_Y~drMEg7~N`G$d@W)B20Iav?4quOXWqwbn5I+jfd?2bNDT`X4sA@xK
z>$snAGv^!dPnrGG%IhXLAD3^Q-|&1TV`p7K!oYpMMtKpUVOXUyF!hg?^961BQ789}
zuSqgR`#cd2uN!54EVkd!Uf;Cm&)j!PS<Sc5<DHGyJ)j3%=a?Gr2Yvy5VahMejT>ft
zTm$$!_&Xne2fqQo;o~=fRnqp;1U~{l`qYoWAH?wohu1Y#09u|0iv1jg_I`^>^+%<a
zOQ`ot!T(&nF_eg;+K{!x2twmLmN0%L?RAmT`!gu5f2^Y%_q3FgmVBiDt~IUsqWwJ|
zwExilw=}%BcRD$tHDUP9<m2BF!gp}tJGd--2N&9Xcy7^vI05bB4aM?PvEo(Qcpag1
z5-J`TA9cFUWFn@z+4~70gy!boQ@Y+tFCX>urKeT^JfHsZI4G(7RH}5$&SPxm`!=-Z
z3!X2)_vT=pdNO|i%u^5Z)RRWq;ym?c+IMs9sM6YfrIknMad_~aG-1UJ;v{D04^MR<
zrCcw=@gUCgLv#1@`f~fj<{CvOREi%dr5!R>4b;n*?}woujWFMr0Dt>(`wRMfSs%Dw
zvv9q}`n=4w8^XMghJG2Y(>t#+^vl@4cf}gQm=RAOt|ttxT-0|Xq3Y4M!?qCWmk00N
zPDtzbUSQtG+H0zlM%`RLbXMDYR`V^^ICuE&@0Y*(%gXp&@N2mK`gEL}dI1}L4SoxL
z%c9?=pHD2j;<Fy!<a$;er-~>^$oIQga6G{A;E2bAqxrp~l^4bvTlg-$jqm$u??-N_
zzpU~7nyvG}!@Q<np4U{U>k#;E4LrZh?M$v?r)S)hk1)7VwU-45RUdlP@gmg!HE%*u
z!oY#Yju$6HWnSkjX<+IfEAM4#$uBMWNIzcLTz@mApMM;@KIYU;6+kyc7c@qvZPtA7
z6kU+Eq{d0(tZ|y63(#qt4NGlKZEkHo4Vt9WdKi9}CObI!ck-F6(=H9E7gwCz3B#YH
z3cSn~kk>o+Z7ej+6(ZB|Im7p5jl)vYTz>ej@2Kr}W>&N39AENHSFCXEldoQD=kf&G
zN^|*9wJNlv{_lFsm*(OXdi|~cxwon@<R&K;h0ZU;C$9ok-c&C%uFpHnIwK0@@8ui+
z1`AE0YKL#T$~T-)`G$DtRA52dR+pYGf67TskKl~~O-3eROBCw7OQ-sfD@zL4TK}-W
zb=RwwbcgMr04QKtfd+G&C(iRe%CSy!&p9kB0DFJ}j-$W^?>Y0X)IE<`CF|`J-{)qm
zg%j$Cvi5pdsPrHnP=JpD7eX|J8%8E@a*<D-#SF9m(9~a?CtT!cr<fW*-0UeJ_4`Yq
z(z8}+v96!2O&|7?Mgggxhf7l`NCgz2O95-|Phdamp6j!s)6&l8GtB+x1qOl+j^GDU
zwYtIAPUhIL`3@&HX@B0v>hoJm`4;Ma$WlB~`Lm;?Blh`QE<G*zE#K{OJugh3u3Wl8
zy7nm$x!5Z(r{6U!H|;CM$Ix=0_w7wTWI@hBt=2u@Sf$X<>y0&Q>3mGvc4NMQpnyFJ
z^eVB*uR{IToK&<YuSyI{Te~84euTa*5Xwg?xh|gfZI*?cy%Z`x*2W|BdlhTbh5h7J
zfbaMt+QJF#ze<TuUhn#ty4cT8#EO6B@4tq>dG`z}g;JlV`KD)S{26Q8`RV}S<e7gV
zz3v}dGSmOjc0Lp;Kc-7d?^=7Gb3u6f-czQ3{w~ySu|1CUdbjs}Q?Br=K2Me_-+Ye)
zrCh(#S`NPTB6m2Y+~*2)o{>5psoPz-{SDFw1waAEQ9$VTG!P%ZelcQQuR4w=LrImO
zfLQZpN)@kE*Xv5CYbXOKAh!Zi-%pB_p4>XpmcCfmQ$oip_IXL@ezkJPXKubVTGxx!
z??LQ%S$lu4l>CZaueQcdTe}YV0R=z-PyiGF1wa8%Kqv)leQ!bg_gd`z4vSE_LOh@V
zC;$pLsse|;^%`4wp_5uD3jO?1>iBFu&%*xv`V)X|pc@VIvk2vfQk_5e&Zi_-IHk-3
zS)$JK?Y9TL!eZ0tM^ep;0qICP4~Zqcjr9)%1wa8%02BZPKmkyIRt3cFA7*d-wCV==
z5?2AS`mstCuT=f?#Y%_u`4CvoKkf7M%bh=QTOWv>Rt1E5-V5=_`~G5h?DKiYYrW@K
zA+5&)<O>uKQUTW5MWM#!2&thVF4nI<rOLm(&j;iR&(~jQ8%HG8d|FC*Zw%#N>pO^I
zzuy--ezA^6spCyQf3CUTeA`h7CyoMAo%d-?M;ukzja8`U%d9G3uHAz5bj-^R^#Jv-
zFKZl@+7DQ5&-mus-1h|d#s}d*0Z;%G00lq+PyiGF1wa976cFlpguU@tlULa1a|M1a
z{B4#$T4z)X#m`fac8yTu?yZeS+S19Fnq1>u?M+we_bfuEXK(rC4$oTu6?6HOdmdcA
z>C4?OSg7=c>Q^LGJo2@~(xhx}KOSjvsr*!=8XqHcI_92FTf1JWnw5G!B&DS%c0JMB
zFTv6B#rOIH!cG0fdBR0b?Bm4V_=Uc2NgtoR`E!);)}9|(+0WAaJhpC6-|jY;<2-Sm
z_fd{@3iaJtbH|CX-?Z5MWo?Z=eLKP0Z`#HWC?!AU{Mgd^JB?=i9CPF!PH3i2*W5Ts
zOYK?ucH7?f9>(zY-mj+*FVy{DdfnhRC}4{MmX0S|x7(67>}$UQV!eN`KW7kwoC*ki
zem2G@XPF>vPyiGF1wa8%02BZPKmkwy6aWQ40Z;%G00lq+PyiGF1wa8%02BZPKmo^9
zKxzHJ&6P*!8fK+`ll$bW*V?%}!9r7LJt00&02BZPEGZy#J0^8}VxKoG=?B~8QGmAo
z<Wj4^`nU+lzcR+RD5V~S8m9yCES~pmmW3Q&9i(nY?M*NJ_zKp4+p7nJ2L(U@PyiIL
zO#xQU8|F%te*7c+?ssNZv*#S>V4n`={0-KF0-yjW01AKtLMR~hdkenl2~k#v0~7!S
zKmky|aTT!i{?s_H7*O8tGTS=u8tngk`|Uxmu&p{cZc6|q2L(U@PyiGF1wa8%02BZP
zKmkwy6p%xKIew|B>W1iov?Vpp8W)Yz6kUK$<7`-Jb82&I^J&l|oz}zfyR^nh<K*AT
zXR=PaG^Ac!aVI(Q?9gcj^(vHHvgdYAXqpo>!}1E#7x<pSQqx?1_^yAx@KL+4`;RbR
z@=aH)aJr)9vfOxjowJqZ^7EnLTld8oZ*zQUE?%M654yi7^Nit#II$>nejz@26_^lz
zqh{!VI~?o$;RlyXo~!S3LQ|;Pd9=6d&C*pbVW}uoz9Aku6<E~l{{EhmuVAV9sBy_m
z`})OUOYFI%{<X-B_pr63Q)S2%Uj_EJ3C(@tXd(v`RtLWI0pUOaP~f8i!y80Sn6}~x
zW}O2QTq{MCeTdWZV@csZPyiIzaDH4;blbZetDLZQ%pY}EC2_3s#k9mCuP*+>DV;(c
zY12;i*!E#a6laM-r3dkV0(=x`oqyfh_A$q?T;!8y2m=a$0%;W(`u*86c}Hbcv&OWg
zF#pRhQ;7zF-(2N{ruh_MiyR6(cQ50YdBasK73C-+qzMYht$?NXKd{}p_hhW;RBQJ#
z?{QxDFc5Tb1V50fovnPsb4|A1XR+K|v3PyHM(7%*mD4H|t*btH8!L@cwWm_~v!kUm
zKDynLE-hoQSd>f8ze+T0yz6WnCo~ry-Wc(H#0ySt%B3r$Yo7uWM*QBTSn?%KZrWFh
zkD;Rn;sUoKpL4A8asFod#Rrl(R=KKXyX;%^&p4rItl1AmUR59W;8%_n8uJYV1?*9v
z{y>$>lE(+IRJ13rN(>toSD{*|O7R?DKJKutRosoo9IIS4Zr9Rx*KTn_Q79id6j6D!
zTDJ$=OQG^(Z9JJL)%IxV7K??Vwduls@+u(o`%<~$%fHci|L!$f3{0zoyx#RObxZx+
zsNG&~#jm|<$8pmHrr+qG!YA|G=cFdz>jkO9oe%2pz5kD~91wJ%!4E#RWTyX<dYq<I
z>6tDqy=#BajGg+d+HK4lXa9UhSES*6%sRz-&n33U34MN-I=(xzR+n75IGSUPwc9<M
z_rD|8IG}Yp@I4M#9T&w)zWLUR+~HX3pJpz<xZ5QE{#Wl~)>yXvh@3eZKEqN`ZhwRH
zK><*}aTGYvzCei<r4Dec(kHCUq~hx_$2zNcuKn$oy{9>$={TMYB~^j~wDx<pHeXvS
zZXR%Dd?Xf|N~min11KQ30_o>%V)eJ&I?|T@c-8LU_5Bhs?aX{)@!CaYldxE%z5KT3
z?@yPGHQqWUVpjO}$;OD0UE(ooJe$@1McLC&F{@-vZ)1KNxM_!Vn)?9LLfYDO$PXw0
z3V;Hj04M+ofC54(uyRU<x5L}N<CMa%<)@cj%k!S&OQCdyct8PA02FXk1-{zy!?T~y
zU*v@1&o#4u-{C_z$11ITj|ThEJr5Mr57dv8^OQ+-{*Y?kTA|WuUAEGzi7jtn*7>&U
z)5&*BU&mt8=tqz5_8U3I<0c1mEFFyX4+I530Z;%G00lq+P=HniyoVHhc>4Gej#kpD
z8{|t|1z5k|{N{4+6{{{sVOGidcqi8L5Az;<dEO&&TOWv>Rt2KD+voM(S2!TsI<PiB
z6RpPs<O>uKQUO};1I5mlkQzc<pny3AX5S7|{jf-nt);c!3!Ap0qvyI==di7`^j+_6
zy088^knk6_mDZL|Y5TEy-#O_1qRcafAL3Z&q}{bcD=eLZS>>R=^Y7}G`7Fmeiw7<Y
zTmJod&JxYFldxVM1sW88a%f8Cr<|iw9@Xhh_0N58R&Kf*gY6}~CH!*ERAWPPKOXfM
z13?Gy12_(h?ZB5?4drb$H*b;hN(v<c1wa8%02BZPKmkwy6rf1~d!IuH6`m$dSjvmg
z?{VcWyRGlV?A>4P@=BHdgy|(WX!@SySYz0QHF0nEJ;bc?g;S+hL5rVbpu9R9FEWG6
zljS*PjVnqGs2%<FQ_MQW`i>#rc%^E0Asxq7V8puI#RJ;^gIQy-yJt6U_I-i*QmR_#
zd;CMVW!~ew?qRX(NvQinh{wKu1LLFHJ?YXi1_Sx&Aog(r@q+@O04QKdfegzlOkd!8
z3Ijn0n*G3%AH#M~02ENZ0#Uy&>g+V~78Z)iSB$O_iZ!mBu2Mp7Y*nD)(DnIzzDdMF
z)7E^z{-6LT01AKtpa3WU3V;Hj04M+ofC8WZC;$q80-yjW01AKtpa3WU3V;Hj04N}z
z0>uW*zTCL?OOAGCUa-I0sB!l=p(&s8kTNI$3OK3)wR~gVKDhliW|c0Zd=eXcy9J9y
zNA(NZ%e~?JxTNT|cR5-pwO&1!)V~(F@g4?(4&Vp&9|v!rY>XJ$B_0Fi)j{fZ)ZX+K
zHM_sR=j1C`DYRD)2oDN?0-yjW;8+UaZj=1`U%iikpo5Tp;8;fylok{K1wa8%Kxqo(
zd^~wayVnmmd#N;Cpgf=eC;$q80(2?xbIt7EclZ#_0p-=<Ly=e2$36I!1B$ByT}L0}
z1{44VKmkwy6aWQ40Z;%G00lq+P(bMl%<)Stp>BvSNLy0lqRF6fnxYHPX`Bs9ZBA`&
zZ9WZ}q|<sBewWrbX`K8!`ApVnmxk1fEAFIpeW)wnvjXdb+q<5@LepG9`>y_d?a=Y3
zu+%h{AHM5fG#|flbA{8GFZre`R=B?L8$$;ACvvvZTz(3^t`nCj^cBaK=HeB4eM~}@
zflH6>!(vhB{6c*4D&QaQ$7!xdV%9kVhfbci^$Zr8Le);uAz`DYtBzu+C{(^79y%2`
z)TC7nH{VDsH7hL{xiZ%CEVe|i2N{m7x_Jg$OFC7CT=7+4+>4;?@lmHSP*@%K)(3<G
z1wet13LGC>rdN;ePGQ!0{p7<TF&j?f^!!*-_zx5S1r8nb+tH!x4$LZ742_?A_tqbn
zRd%b_Z_Mk}o3PR;)RAWG^J}5fgLpszJ_^*>((bJD(lMM|<dbIz0}6lwX%)CUCvU^4
zA!jgaOj`=`zr=b!Yd%HTB8LJ(zyGlmpB!a`G(iEm6*#l?!~A_?PGF#8>5#2d^ng6a
z4r3tb;0S&oRXcm0oA=GiefzQ8e08;8w*><ZVOlx5nNQ8JFOFcP5$&$v-gVS*EIOrf
zXa`HD+_Wj*Z3sAl#iCqtuF87k8{d7qu+XICx=}7&Azk|vkh<TSx%BKS#mCT5`0J<V
zTD;;|Wd~efQgHKlj#WCXe}A;qxmZpp8f*54kbN0CWIT&mp)ub;P{1Aq)*ascuGZ#A
zEEVm^s}jT7+pld?@^9sJS59G8S##MSzjG5$W1%RNj|er+MyUK)8;{WAR;*1I_LEnE
zsV%?1m8sxf%sR`)G>NX$?;r-I)j?kG`k4CrwZo#1_uYe8<Ea%*x|JBP7Yj|f`l&`O
z3LUoR?M6&HA9dP(=G(lxu~O)Dywit5we%RsrVbxlGSmNM39G-S+>}VHG@33gy=$%Q
zS54b~LftpW9gkSg1H|^Yyspm*UH|(8W|jT@YIN>+@E~T5eZ9w*Ik<H@23n^B-{Zh}
zOSMM5E1ktaadhBYFLH-7_uh-O^{ifZ&8p{fpqJknEEVPUH%K2800kUJf%Nkhu=e}A
zy<YzdS#X78ou>}$njCmJh7+2O<H=A`B`8p}Z<x!7&Sx;I6zlm)`goO4*H8vfKyC%x
z&KG<<G2#?vopS3)TlzoV7#4MU^$AQn#U5|+OP8Ga!%LmON@1%KSqCh4IgME%Yxh%1
z$#44WC02gZ_TH1Wb{+Bq3V;Hj04M+ofC8X^PztP5hm8)ovjZ!Iv**okcHm|%HD3y)
zE5ri|fC8X^qbksB%la`r_Uz<@;`eWc9KT&|B4(9d-R?Geo@*Bdf)3yZa2#0M4oG$W
zSaI#oDI+?Z#Xx21VDuyCXD?1ad)jq1);|yw00lq+PyiGF1wa8>6`*~bAFaAUzQk34
zcE7ZgKUT-18uQIce>TXcwDWM%qPMuM55!KZ0&>r91L^y5n(L9+UL9C_{!Q!g0Qmw1
zgj8VXL-pWqUbv_ojY4V&ae)Hn6zFhoSU_y=1K3(_nOx@9$byHkCHkx#wC--B!`N2#
zdKXk7!{Wo(9RF<5YE$u~!`N2NIlX&miy=p_wak=|LB0F;qnLGGczE$x{uReCYuxUu
z?ezHl3Cubd|9Q5UXYfhPD!=a#vn099Nz6K})!$q@3G3xiV0qHb*ct7UI7g*Cs?(cl
zk3lC#7Va2{?Ipb>{Bq7%{XP!lo9*u#LH*HPe_w7jl(*H~yb;PPDU=8l00lq+P=I9x
zq?%71(vf!ldeA$q-Yg%Zkax#ZVA5AL?lpEgg@vNr3g*vWrpS&NFF2_vcX_2sfB)jV
zd&4wGF>Bnp=Y#v-+yTrg?}mQ=eNm5tm{sQLv}tLdrw1|XwD!F@ZTs2U|A5y0h5jD8
zz41xa?m{|_t-uc<`!aOMcowt9N=rtrjP*Q=`BJJ{k9FF=;_FK%F_51QV%HNqUr76W
zK`uQ;;LyqQww}R2emaPKoIw1b04M+oSW<wN@uH9q#~deYZM|FaW7rM~fC9=_pjgr9
zm+v!1Vxg#f#po(w?C9UUC%bOPLX)miLT+qTz|wp@e7D=057-|R00lq+PyiGF1wa8%
z02BZPKmkwy6aWQ40Z;%G00lq+PyiGF1wa8%02B~HfhpV1E~rrNAXYks(CX7TUNj%S
zadU;!n0E4ge`zV)r&_@Opa3WU3P_<q&1Hl9&P_axS*4|R)!yw=)B%lY*!ulnsmt*9
z(ZqWR4bNg)D0TVkb&bjI{`dd}f)3yZ_8$j(dj4;EgA&IvP+lFRZb$7+&))BI(H7ob
zJs><N01AKtpnzj3K+Am?<b$sJVtebuv5q1rEhqpAfC8X^(i9jU{HE8XLWi-vRGKbO
z9#8-j00lq+x)dl-=yosnu(KE_uMW1phll;EHR@gIEVfk#x{f}`4JZH#fC8WZC;$q8
z0-yjW01AKtpn%d9nB$jPLfsHukhY{IgC?WKX^JjDr*SqcwK=u9wfQt?l1}Sk_+47#
zq;c}^<TF{PT^dp^uDFxZ^`WkO1NFPw&Ok@7&@@-j>h8xfM=U>zrKY+3@LfN>UDs|4
zsvpCA$v0iG!d+=Rwc>aE4|BHCTz)Pz8r?jn=M|1G&BZJ9`tnP(PhGQb$6`_F{6c*4
zDsZUl@waZ|!>z1yz&{_h-S2S_3r(SFr`qU%Bb^2wz*14Dd_z2RD)8aQ$K~U+$FS5~
zo4;nq8=H?|p-HF8kSo3l#9pbdy5V~S1BKOrZ+$>GPyiJ8sKE3-Bd>06cNDYEuitN+
zHRI=FI6Xg>6#fGRK!NApyWfw^w+*w(>Geil^M1Dmv&snvF9z=FumdZdLLF(bTk}<0
z{^TgOM4{4yct8O@3f%Mi`!FsM#>quKd4@2c04R`F0lx43<PMj%6y|?bFMf0MiNi;*
z&@`VSY>`6&sr^xoGD4c5fZPfcYL$>vTly#lI+hNr2S%*?vC|<81RWf~52R{mwd<!9
zjjg%|%gvYukvD#QugA2qU_`fe$>$GXrSaCo_H&Y39mZl)Dt~sgbl(5rJ6(O|Fcyn)
z>3O8~n~VF3&EkaS`w1?~Gfq2-<)&P^Lb~=TVClVxRNL(<#mCSew!SYH(&{nCD)FWN
zvv=n4S`PpJKc`I_ry|)ZD#;q6D7tRR)-AMHqC~_Up+1W2iAqA2yGRm|lt_glHK%=(
zPWxV`b548GzR>df_Tb&;*L;j?o0)6oKCgey>&#p;?|IL2mTRuLmV4&soBPjpY}Fed
z7<u{FPVX~WN-eAV`OL88k>5}d2t%Oh?2iv`uP~kI(lEZJV&1UlztgW-_F;n})0wS0
zb?rGbVz*6avQ$<d8FKSWV{4u|m4#JV{&=;Mv$AZBnhm~pWU1G3kuF*SHLpMZ;ljnI
zFx&OCrz`$3u<aBEJ(feXz3acFH$JD#O<iY9Vz%kzMkBj#KVuS;t&=)tx4+>22~4*h
zIPvHB_I{bnbWOGW-zEj~?rqR@3WM6o;lDH4KmXVM^^SXMemI?_n*MWb&5dpM(Hm`>
z{ydFYr{8?Grv0k_`>u)Ys))bp9(P*nqFuW-n96L`!O7*lhJ8Ab*`_@g7y97gsuLO1
z8;79dK=*T#`K`R5?Gkmq)2mN9`@+2c+Xu8<!`>H(D!=XD6^pw6qI?hl0don+{!XKA
zdqYks-*H`Xqhq^%|H{xx-72qhWUIM751xU*{}PaOzc?y8y7q_ummFe100jIJNR|6N
ze(Rm;>1BV1R+T-~_jw0@v7^w;NfTJAsH*onRM{O;J8}J~d;jlwGt1QW@ABEYza93u
zQQ7%L=ZUcE>!|i~QEoKftyaxl)K3u)0w4eaAOHgYM&RLH>v!Hcehf<$J#_2fxif~G
z9Sf@e#tO$k00ck)1l$R{`uQ_G7FVC)$kIY%AG~*1?E#LhdU@+MbvIo8IfH@&`T@p)
z*LlEwrj?W8aLa{bW?nyo*`l`YbMt;*v88YM|FoRJR4eroUEjq(IeNcC;>MfW9|{5>
z00JNY0w4earV`No{+ZwMOyvf@)Dlqjyqqq(Pt2P;YgnZz%vQyBw2u9bmUbSNe~&`-
zcQn+_KCo^o0qef&qv(4$wsRPIa_#1;Mowq8sHx)teu02Y0_`tdJN2u%zd89$Wio^<
z5D*bKtzW~umETTcp0!8KCo<of%uaLBH+p?JX?y=E%&+Dwx%SLemrP|MJ=FWOQ!0Nw
zmHAb_@1srKTet4L|Cz>Y*CEv!&HZ!1GzN`^gYEst?>I9xea_SizRJv)#@tjiPa-}V
z0&^c6UZ_ZJo|9iS8tUejnltcFX4hHMSy(l<hX3xFsq-uR3brl<H*gDbJ;V2#f1fo{
z@2Kdx$JDDSY6Jox00JOjGXY)C`JkLj-i*ze!oVRK4mOX`AihtXU#e;4i)OH>HHh!X
zla{9I`*FHh=T693xVimSN0;iVSJdSshV`tNbNSEAHf{R(#|2tW9Lj9f_}PDS+kNJb
z%vQa4)64hd{4k!`uFF4JcW8H|@ys^u`TC-7`*!_;*{)YC{I>UjDqk^EH0t>n<>wCq
zr#IQ2k*GF}*`}iXX|M6RTG#D)2ps?S=SEbwQ`Fa$DF3hP`!{d6`S<6_&S21B98`}J
z*ara+00B1w{Bsis<M|sI&)>-IM{fNX$3XxDQlG%7bKkwTX1NJWmZm;2vue1m#{4Pl
z?2jDTYE~`bjb8$)zL((F52ObH5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI83ev;_w}ImEhe*6R~fASe;cRd4jI$r%1mauHoo%Q@dGzcWU}@DXMz7g00cll
z2Lb-~du<CJdZplJg?mn8s#HfCm{`W)_FFF8keJTAs_r_N_IoC}>tFMcI*nqtOkz-Q
zKtBjS4*dRJCelwo?(5H{G1(k+&!b_>6aAitt?^-bAU_Cz00@9U{vzP@IuYsoU03IK
zUioVjp|&6Z0w4easYamGltuUKx%vYZR#S}&>H`8G00JOj76E)8(~RH0X#aZ0y){3a
z&Y)^Jm^J#~4G4e$2!H?xfB*=900@8p2!KGU6X;#z?;0jMBwHn_#tX-b#M8Sc8z<uh
zj{HqzB+4Yp#?!hc6XlMCtHsmf=?z;{>6T0k=uvM&rSw$igZhC02!H?xfB*=900@8p
z2!H?xXd-Z8`wP2vcxRK7y{6*f^jT}6;E)piK-2I+F(3c}AOHd&00JNY0+tcTDE;Cc
zPW{C!sGdA`#i${xW;^*!Eo)}D3j+U2U|8ca&-5y^#&J{HFOj7^e(e9)hb>OX%%0is
z+S(I}O4EO>EDnMI2!H?xfB*=900<}}P<q0`-<EFL$D-Eyx8**0^N52iYJGbDuR}lS
zzK7Yajn_^|e)sq+mfuv_K!IVc1a$S=ZQJ{tcmE$HGI{;IM91azil=1;RjsfhEUf(e
zy1mlRe2U4^2&<yC==;x@SgF_B2OQH?D-;w40-6Xsw))1g-R|GTOjAwOM=_QY_@U*w
zZ;z?8(J@sY?{dkr7pHA;Wa%?Qs@AJ>(FW&}E@u=SeEw=_dJZdB_VwR#oUppS<w+H)
z-?GAyrM*^Ode#%qZgF&}<--B)g8&GC00@8p2!H?xfB*=900?*_(0|Lov$iI;JF4{2
z+p11_ed-QJl~#W8qOS{lzS{|_FTHx{!0~1FI#4q?c<jzd1_B@e0w4eaAOHd&5RyQj
z`<?T$ifw19s;$1tJFVyWhnQati51yF00cllJAvnywHiF+!f_7iwLSMt`sJ6oj+;KV
z^Mo#!wO;19>Ej<|oL{C?wqvWFeR8?B!|JVcc(t`Yy8PDOku??A;ismj7r#E2k-k^<
zj)^yB=Cb^%(`s~!jXrA`Q>7)l+;i`>L)J2pjxYI6r(ru+GOt>A(`#>CcJgxORYm2_
znRQpGso&)>uUc-%ZB=hsyoT}W)sLvsdFARg-L;?eTrr~f3TCQ~ec|!W-`=%_@!Eas
zeScj2?s^ur#;>?{SZ4oq%vK%I?#?#dYv(cOu^cp=zEBJZfB*=900@8p2xuelMAxgI
zXi%hJrkh^)ebGj9_7utF^_(xSYc?^raHdFKoBv#e=N~JODbiD3KG443&;prmdfRzd
zO)v3AflSg0Z?QICjw_fc(!pn)@Y{^lMKj&>j`QALf6n{GGI`b304N9qKmY_l00ck)
z1V8`;KmY_l00ck)1VG>z0<ylF?`G$<^t+z;?w%iyK7U2eVUdpC`)A8(-_Li33@M#8
zqS^A$dd3Hp8otEVIZ$vgp&uMG=>8$l`I`Nu8%&tzK>y<K4~P&20x3_R{v#8rTr^=j
z^QxabH|5k#)ps(l+Pi61qt)qqnO9v^rCG11&iI4zTKd=eQ)lkj!NTe(z5dg&?X=$*
zY2M)nf8U?Bk%iR@S2i!)rSdwaT4g`q)#kmHO%9h^&va?Qdmid@#d8~3s;KPtShl0#
zlB}F;2K{#qYrbo+bjWtbRB6gj{3%w!GxyDX?z@sR9Vj>)=?5v+y-^(y009sH0T7Tv
z;EfBH_gFCh59U{|y!MyomEYUVylSg4O?Ez4V;>W#S39C~A60s6Y5P^1nOEK4@Ya_y
zTC8EVYS?`W+w#l0KO@VItoOXDUtRLD(e0KxP_-PSOqMVL0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0x3%1-a5@Huya0lNa?wHk$!r9`4O)!%5#FMuHR!dy_^*7PS#cHm8}2!
zENSyRC%>q54Frcl00ck)1VBI`0scGDD_(Aw^~($M7%Bgq=FWG#bzb$uD;N|Uj`RbC
zT@$8300ck)1V8`;KmY_l00ck)1V8`;KmY_l00b;0u)SNcbyq&W#7Q-M>a;=)oc9(v
zsjjl#JIpV3EFGqB8w5ZA1V8`;KmY_l00g8E2>X2yuTG3QF;?x8rGGk5Zyegzys^xX
z9}YQCa4@kScpdjh2Ld1f0w4eaAdor)KIw8s(ZT1hcD(9=uMd4)f9L{7mcCQ1VJ+6<
zdneV@lpZ3houam53Qza}0w5rtz{GA9?wWMlTF0--`pz!wWSn27RJP-ngZ!QVizXBB
z`ulHuI+N>^3clUEyUH0i-@L;S=^Ytgm3ZhsyBu9Q@40UW-*?X*$4$2_8ga{jg?pT^
z`qDd}K0Exh?T#!x_2w(r?(MkWfr5ic{U8<QD%1f4KmY_l00ck)1V8`;R1?TvQ}J;6
zthJ6vn_ky((#Ek{S>v@+lHWZ(3knYC2N(yK2mYQ1RL{?{?A!YOtE~FN4hVn%2!KF1
z0&dTvtu52N`H<uKI=1Wc%UTT{a^W}!3J(9)55kQm|GA$TQng;4i#9lJ>i2meq!0F-
zRet5>zvm){Ndzj~J9^)jFTLW}swT07FVPdA--)PMsPXJK8s#w3+gmp2Fl$y0gMtJ4
z!T*kf=tr(U560c_`Sc4ib~sV1Kd+G*1k@7fUE^>2$qva@iK_7;@#Es@-II-z@d8Kw
zCNdIb5@q9QU6YA&N5a+O>GAZ2Evj@&CI<AVx1mzHbnm*~e&>@j*R5r?>hPj<p1(VL
zx+BuCz>024e|%Z_$xOD2{JMB_#-f5ZWHa3=@+W9~kqN^({WN^B6I6qit1931r@Vi5
z#sfQ<UlsYYYUjtL-s!fP1yzw<+40K`^<Vng(`y`6D%)R?y`cRtzvL5f^BXUkhi!}h
z=-Iq0(u%+5$X35RMd?Rg@^<@YZ&}E6smKq+|3yG{KhbT^^zrGppR4Ydve$#U>|N4-
z?Ng7GTkq)B|129c`RsoCouE3S({J~0seRai%H<$yoV-@&!jmo;y~VLzb&m_#<*8by
z>$3ZM+n<&^cm8%qw#w?~B0IX~Z`I`lT?gpiui(q0>u6E?=#CG{F=#&B_48Vu-*ua;
z^s3hh{fFFiVzEaLIJT<n`Bk?)-PcD^mFIPwxux?uuejM!b)28igN9Z4=hbdh{lxZi
ziEq&BLs|LAkDw}hzgYJ5fyiFi^?guYvh#VJ=iJhre$~C_wET0Yqe^A<S5xh%UVo{w
z@3lXwTJKQ#@4j@~q>a^zWs0=_5343FEU=%6^z@r{E;|3cZH$!cLj4iYy`SaJJ9Vi=
z=Ly~M>T2lEa@g~u?fL!D@}FAuz82Y+y-p^(wyN7Y$n@j3k1O5tsqFGpt%Fk4uGg87
z)@`<{-{<Ar($hY_EDgHPWO_bb=OeHE8R`5HkUdZH_IPh}_2;?M@1A|q_ClE=-PN?}
zeIE@x#Qf@k_Ya-EFY~xek<RGarSy;U{$#w0uAjZe^YakTHNS@ItSMTxa*u+UZYpZ$
z|E#awVaKI=m|qRb+mwB@p#Jmubj|;&%8|W~uG^mMeIsgL+1IbC?TO}5ukRaqUH|g)
zu-7ngRi&!NM}D*G^|=sP{aICi@M@P|ujvl+eumv=(LMjlE>HG-K%Q^Vu&n*Lpydb6
zC(CcL6Vwmg`NEDPS^0I>o63{rf7EuU-`S?J&W|EH)Xq_>s`Zhm-m>CV^=sYh2P#kY
zd03S_@;6mf_VuT3d#c*m7Q0c;pQ_4tn<rJ3tIDpb^Neb{ZtW2ExUO-gYW=0EzmVNj
zy}xaGy=;A7Cn~?Gzxq{G;}UlN`reh__An*A*L``U3-^5}Q~0Uwcc$F8wbc*R?Zev-
zyRJn3aNni!dOuZFI=(;14(oc3NmV(f%-do8@OppWZ#voih3v_?-X?px)a`j$I_{UV
z?DRo-n6lmv>c6UaLYG}#`#3ybS^Y2UJQg-TUoSrF^?r}(`zn6@OO^X9R1d%PmzAFH
zPrBv>Q8}ibM}zt$YaR+~$LsTNqI5y)?>C>Qd|B~f=VMWRQGZq)AGTlU&M%9Hs2x<r
z>*gf8JXKs&+m$u$W!d4+m$rs&-3RkKf0%Op$NM2`e2dyKXutG(zDD|Ef8Rya9=62u
z{Zv*sXn&NI&+oipYI?8N`=+JSwGX7a99iQbYCEF!5REs{`_-cOhrb=xp-8hLnQrR#
zKCi0dqW8MJ#z);=QI+ra{b|$Ei{2j%8?WmACU2MSzd(=u{WPAFs<7(oTh;buU1x~w
z_+2OY9d~a1UgVdoc(?My9uM38rp>pu`4g6Z)cHm0INkC2bsr$Abs4SGRjaIdRF@sU
z=NC#}s!Z#}&W~xCrl_j*3~!hEkF0*I`uba!eO>)kb~%3MeQJMQ>UX^uHodNTiOTW1
z52veNyOlG%QtRbAGv+z2>hinkvgQ-ABU)dKDn7T>N!@>Ucy+6%=s1<9T1DlDji={{
zysF>#vc1}+bbjmOcf9lWRov!3zwPSRf8HNaSk#~0;#K)WeiqAGe8M$bmoZ-X`OPm!
z)6!FYRVmNQl(66NWm<Z-@e%cLRqgCHuelvp<(DkGVfz_>U86f3)_>&BSXA(aY!;M*
z?CU7q_JY>ilzeXUkLkxn>i}7Pd+kr;hbiqE)DKzfJ-_D{;ulu^mMcmh^?D?1`E(tZ
z-_@-@$?EU2>`;FUdw<gQ{KP?)+ID}F_gDA2Sao?(-+%R6zO40xto}@PRJW6C`?BUq
zzjnOt=bM&}|J@p|VZZl7MZd?w$J2dRzsk>Nd^l)7^E-}HCB56a#?<4cjjy2o_?>@{
z{;%)X$>!j-pVRr$ugdb*Z@E#W_qyMKbay^EbKP1Nl!NT{4eEy}`Bbg1W!X(tJ7M>k
zOwI52e!H&pskR>%bp2#nK3V5^*>+@~Crq~|nrC(SE$jPCQQ6^tuPiJ#zt5T3l3v$*
zpWn+#wey{A^|8JGnZjE%j|Sbh6y@`~eh7M>UsXP@`_r=0i8uz0=l7)%{_FSH{mvt%
zt?&K%W6C;4H1BwgzrI9)r|Qkx<Cvyi{YN?w009sH0T2KI^9h)GKR^~azt?B7(|f(&
zZF)LW`-z)hey=0=^s@RP?0^6WfB*=900@8p2!H?xfB*>OZvwjR^W<;J;cG8@Unsms
z{w&CPe!yR;NbQvXzwd-_|D~Tjy~a_^LH4*Ndv5!ZZpWkE52kwLS9NP2Q;(}2FCzP1
z=Q%!|+xtu+&TjeWxaj$GS@E*Yt8RAujzhQkL&Q~fyxVhVe1A|CmR&E{JNEbZqUNFN
z`#FA}FQs|NtGaP?`yPTU4sP?K+kEAA+$#^abRs`h%_Cmz`kk*$P0yc4yw=}ux}eX0
z>(1x5UP06I+!4;ZBCYs)4$mPdU(@q7s`qxsiv|^m*{bW$%{k-D<BP_i;NYhpyioeG
z$)!6Mi`lM0`#nD|`CV_SPVe_SVxshXJ#~fo^|R`7`SQGmz4|5EkI^0PcAanfakuL(
zQ;(a%m0y>64U4YdY>9WfK6N|pcAc#HxM+TM^V^i;ZhXAXqo$<O-S17^ckt?$*LhiX
z`+A+<qE4sk{2-gB$e!1JpsJox+2y|r#D_(^Oo{h8F7sPDQ`*PPkD%Xi5!sD;9yFzV
zuk%^ZbRzz4@usZjbot@NgN_Gn2g*lwqMqmYbu#&-OaHTM(B!lG?RTJhIoNvt0Dk}b
z{hzmce)v+e#|p%t(Q?o=f66N7m!D2)x8<2nnQV3a9dob0ckVNtEH1DE0{=!J=(?0&
zcc~8lTOGn23wk|hO1>GLe!G85?Zb{L^~=_j>(NyAL)0#+_Su8Z6H(>!J0JP=Gbm@j
zewmV9)ZbDqKIroW#35+E2%0bI{g<HSdwpKS)O4wSezC27v@dC!2AywGH6P8Rs#I40
zkmaB5aVE=eksVX6i}Jf4y1u_*N;#tbNb#nuql5Yp)$<y%%9S<mrkWk<pCa{pU21B2
z-RC*k<;fZcursXFPs0~GA`YT)pekPEhi&oseZ6CP{oS4qk=?(8?kDNWC%e8Pd$x^Z
z-e1$&Bk20omVC0lGoi|ksr#XR{ZhpTcHcic<AI$lEC;{siS!?rdZ*iFCOQxKjn_5r
z^Y33o73R--y2G~ar>oiz@!MZ*PtV`S>|NvU8YMd<TP3Q-kBb+Lr*}^_PR0uy`J2c{
zlu4A0r*%yx${h(;i>Jra8@8y@Etwe5quz!}>9+H*&i^t+`@DK(k5x>zia^wV|M53V
zp1p?YR*^qJ<HPzNv|LsBTEFpO{KFeMGrubGr~liHE1tZ2h!a#rc4f!=t*_tovi*g<
zXbIf0@vcV0AINiT*V!4LtbeP@21mBaYMl-%cHHq)i8)M{%JLg_%p{<zAMxe<IPUw(
zTh`_=H}$*zGLtgA3QFLm^WMGd`NFvjDwjjh_CP)m0D=D!_-M^7TboT^#%x#q{7m8h
z))fDN00{K2x9H&B<1(48T6OGkcRhRDVrHw_KF@mfU)D(T+Rn1dgB=hELcnW3Rh2G?
z&&UG;AaE=JzxRPmNq=lD#38@?kzUh_OT>{V2)x$o!ShD<n8tKz6m>+IAP_YHuh&CJ
zmm2$w)Q*?;{N<-#KilCz!685N16}iM*yrz%Kbq$-ZR4)1Kj*h{RNq&Kis$`hZW)zU
zaFHWhX+06OM#UB7h9eMlzoGKO)#AUoH*GFb{;YcQ99uPKV(GDS24*|9s;cLPsh`-Q
z|Lu|Z4+Me}(7i4W?rCb~3_IV9NM=s>@cQ}8Rt;N!KEJG9BJ2Dv%O9_H&dQmRD0XP9
zBTK!Oi*(Tv;Ku>NskNUUZSVST>9+PORp&=t_pkYKRMkheU0eH$$ZvXXAFB5~Sy{aQ
zJCptMf2y8;jN0x$*Vf!vug`TMotp2csOqn(dz>o2b=lRuA0w+gj6;9pFz7hY)erM)
zIYHYc>U{k50m6EpFP61F^=c>T{)X~F00i<60o%_fEjRZX*uLZlX1mI|ufyBPKRp>W
zO$7q@UX{1+Ri#2(qYfYtH39y7X=^xYj;59G^?Ez%bh7Wu^7c&mPJk>wY+vWw;=k?J
zg|_-1^!xUz@|iZT!yga;0T2KI5C8!XkVPQ-aHj{pSh&SW6|K^|+Y_rMrN@G*EUvHv
z0w4ea`IUg|{T@}<`MT^zz2A!R?R>uj?UujWP1pIuw&$Zo@3SHPpWhRveuVcL{JqzZ
zD!i!up&$SPAOHd&00JOjDgjyV@v5?8DmVD0mVmDPLS^Mx{(F?No~MMJ591%+(3$z=
zVC(aRy8Ko<`@p)X1pJO8qz^X#sODzdKDepl0e*piOal4!zK%?Wumu7l0;=wFsoH<h
zWmgp!-F9{Hl3k8&PO8eY{re88xahX)b$u_IClMbFf#Wy4{OD61c02i1qoHnYsj{An
zG`EKT?wRg+6XoU4=f=?PcD4Jz&l;(BRMbCHucoLG2!H?xfB*=900@A9Nd!deP~Gt+
zF|o-P+4q;DuKP{T4XU}W>0U>dMqOWB<yX0<VDptzPL0{7Q}63}<~=#5$86Q-SDdu@
z^<OK;Y}Io+6>2ehX|<T`TDw83jbH7$AZDAMIJn%0O{P?j*{;o7AAimtbuNjSqPpf?
zl#{;+$X=i6vgh@A15r9%tuFhyo&07`bvyCjw?{a0!iU$-XSzA)dXFeo%TYZ}U>^iP
z00i6!{5bCW%3IdvF(^2g+z;IPF^+=(2&6s%+1~?6ePU+SP}P2uS+#^WehHvo_$7w)
zAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaCKK55`b)>ZQed`|>MFYp
z#(aJL`>)@V7BgKhT=1U@A338?Ot#AA340&_0wCZ<z_$0kd4GrOo3d{7l9OVp)QumG
zM?*l?_floq(M<u*`Qtf%lb`cP|3v=`(m!?2qhZTSwf9=X@<4tN009sHf&4{4)q5YX
zd&|~l)0eUQ${~M^BGeWHKmY_lAk_%8ZQXZIujem_h1FE!g8F~}2!H?xm_^{(`D>;v
z?3%}*(Q?3jG(GpxOc{N$zl#KWjVqqKdx!%L|IERZeh)uD00ck)1V8`;KmY_lAU_i5
zUE}XsCp#otC91}Y#*4+%yC)kb;{}fVO=Kj>B+ACqx+W9lj)bel)8pw4TU6<mObqBz
zZ$qW@{KyP-1_2NN0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T3{dz><N}IyB#Yz=488e&`1V4nepC0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JOj0D(50di1@k$V!%KY5-ce00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&ppn4meGlv#_0u8-1qbv4j03;(fX1l`MS%bafB*=900@A9
zRsyx|DO;}TTfaHbzc^@xkHSCz1V8`;KmY_l00ck)1VBI`fe+3v(f`nwYZx?taZuQ$
zU>XEK00ck)1V8`;KmY_%mVn>qh)hdw`*%QA{`ACM8Bb&}C^(qd4^npGL&ZP<1V8`;
zKp+AFwy#@*`X2#m6bS+#00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w7>0fzP_kotM_<tQZs=&<`*UWX%JHPIqt%1V8`;KmY_l00ck)1V8`;juOE4
zukzRTuWW7aqa7CifB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfPh8<9iI55)c9X6h=o;+E~6+A009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KINd$V=_}fsjL$Xz(YP?vycs#v(vT-tA;K<)ZMxsojY&@-NGEweGxLQ0tp5CxU
zm2SzzfFAWWR7#iR2F!o}2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfPg;&znyew%$D;uu&A`x=2N$fJt>#z(r=!7^!%IO8PA~LfPRpF$AQ0D5~)D|1V8`;
zK)?(F<$h|}b=`C6F{m02X0$TA009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5I92M
z+XlT)8d*BmfvVwvUIqdn00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00OcI)J=P%@qz(snJ$&(H|(e(@Sn5xwtIBe
zKF3zo^}DL7%IR{}hh6WhaYD>?z3JY;S2X&)aLhKv?~CF0#njBKW?M~_esAl0Z10tN
zarki^XRx4ZcJG3BAOHd&00JNY0w4eastAmGZSxI(j@;y!s%;D0ym;y#yPQ<l)60#&
zdUe;mjxJTzpI{dRKmY_l00ck)1V8`;KmY_l00ck)1V8`;@)H4F&q;<YCqK0b>IwoN
z00JNY0w55KK->N;zB#e(Z;oFLRvL1F00@8p2!Mb$0;=A(GR>|x8b}2KAOHd&00JNY
z0w4eaAOHd&00OB&pv3WK%sF-T*G#s4bZ)zfg@4$@bZcs~3TgoYAOHd&00JNY0w4ea
zAOHd&00JNY0wAD@zyq&7TKD9G%bBg}_c<F?9AOs(KmY_l00ck)1V8`;bP^bR$!DK`
zmo=BER-JW1Ss(xcAOHd&00JOjB7xpD{`M-_A=xTXHC{YkBA(to**F<5aO7_yBT*(%
zHlEfsnJ9N8TrHj+PjA?wO1ETUK#zJGDy2)W)7d3s1OIq=remx2JG02(*~M2ek%k4<
zd-)Y3F8E?IldU4Zwr1=YR4!w!qgzG(1dZR{_l<X&E!oI|YS403<r~_ve6O^-3dH=X
z$e+#KKJQ=isp7GqDzYm({?hVmvkpJ`8&jpS{T0~@+7I(fJ`p#+@t<7!>F8cp?PRv=
z<yjSq&hNgB$yUESMd{By?e=Of&tJrJsmKq+|3zSLnJ%~V8NS@HRr~!mf5O(q+Z^3m
z@v$=}|Fra7=B9t1`Tl}4(x)@8s=8nPdha{i=3VnUvrT2MW4+p|ys-PF*^lS3pxXSd
z?DpM?Z(vZl9Au4?%#$x`JnE`-%yt!BN7@p<sZsTvSI*C7wyLW2wyb_GvZKmx-F8LS
zdB68LuwhxP`OL4j>i_k_=RC2D=~7Yq=#JNYp3+@EujNH`KIs15@-oMNn!{A7*L9Dn
z>11Em!k*dt1x0SVqy1NJ>~!4J>$sMkF6eaupHH@5e(gmy-`QR+agMqk<VR4w@8r*_
zJb3*UN0#c^k8zt9+>Qt3B|BfEU$5&{|C(GUtg7C}>soT|oSQl=a8gyT%pLVX({`UZ
zwyD?tctN#CzB%sfQB1bdb-UZT)$RI%pGUpc3+av}@Y-qXZqJ&L$#m(lrAHUKxX-dT
zuZpjBqSojL99yz&^O$UZkEJ0$q<f!BR(ZDF|Ka@=?f2Od&tL!97JlrU8=JJbVh*!a
zZMiSv=BF<HUT03X^`F~uT1Q5$vgTQ`<F@Ygd)$p9j{o~Uy4U`Ubp8nV?Z<x8^WNxs
z{=41CC1)nC+2e?`(;dfc+dJtG$FI7rcSX<1g&p7g{T-fbehu?nL}7kCO5sA=cfJ4f
z<J*~E4a!^B{b{#yg8I+r8~9FvcV4`Ew_~fy+CPuV&Zxq@Du2^pk|Rs0echguH0`)(
z{ps~O&28OI$NfH+Lg|UCDz)W0+H^nZyr4@}^#`wZW#3ouYmfIc?7X6D|6El$FCQ9v
z)1<0n99hcq4H|Y^Pk22ZwEUp?ygtY8HC<3Ybmz+*(__GQoenxttL}PJdCvTAzHL-(
zuVcHGX;<{lj<X8HY}3Z?_iFvs#^NzmO6?rAR;l>puG?=tEhbxcIyb%Q)I2?=N`JUL
z_qrvg7mLZ(ju$nrHSl0sOt<=dUplJvVb?dN<saYc+>C7}E^%ztNelKAdZ*JSC%<ac
z^QWrvkKgrbtWL&0N2IFCRb|)jb(JaUUvKpBluawvIkxM8J8!LD>!huYY*p1?$gXMU
z3%~x{+@sUmpRLGpY}4=G==)InC)YSoZycz<`c=RE3F*Ck2Z1>GZ4Xn@%eo%|J96(Y
zsp6@+-_dx0`kOz9m+JQ6?Vt3)sEk?f9B^z^zt5$ZmVU(2meoIQdB}-M`Tig~{LP5(
zp1QGbp_r{&cH)4#<4P8e>C#hAs#Ue{Eyu-V>zQL(?@nfwh{;yj{e|qwdVihl=~8+w
z)^rWZ!|QV*s?*u-e^5UBdxWaOJYQM&@##6AsFbf4ANKp)ZmOhL)sLdG8`b>F_b0dK
zpj982HO}~Yg$)Po7oz7Eb;r|ljzLxQ{JYn9ukS~xzP};s`77RzuJ<4Ka=bp@WJ<cA
z>j*xds`*khF1<bnXKFfC?JsMcC9>mn-q4jU?0LkL{6WtHQRnk}9#%C^gte=B9Uiow
zb*@rr>oXIMi$$%fo-b5wH|pnd!<Mh=etgjWC@Y`p>lR)1Rj-d#*^m0XA*=kb@B5mP
zU$jo8@#gpYbzXmODyrvZ!ak=Kc3(oo+w^!@<3_d}uh%o8>x!`Ps^&@FF5PDjs{B3$
zg;hPT!Q1sak9bY5YaR*v9ne(Iueu%j_WkpL;yJ%NBK7)wzt{5<pU&@knBSjE^)Nr)
zZ42}5ZCg0(x<*!h(fD+Wm;HTexAJ5km+ik-d$dkhtwS%r<h~PD?ssg_pzj5n_MMvr
z?Jh{~{p2o3wNn3)_5BWA-`7xGj_7++rpNo;AE)-$rO(fMy~w4V3&(V+u6l{eQQc47
z?ECGn(|&4o*0pRCb5p<fKm7W^r}ulm#P2u=n%=Kpl-{kpp!<Y=_mlkIH}$)3;&<P|
z?>-y9U*hJU*L@>X)A?;Lzv)H&ho9eW|8B&<J>7Get$J~(Mg3kqVI5PYJTFtil@EPe
z=(JhenXP*NDd(<Q@X8Kms;X+|u<LVGe#x>M_WgU+*GX>nZG9h&{0{rv0P=_XE*rIj
zty-&k=QR@s?Qm>W(LO!hFAJ-p^*sMvtJ`&fEyu&2uPDFY^NZK#(cRMdEmxG@@AZsn
z>HS_mq-uJ%@1ULYUW2vchHh|7QR<JTy+5woAL1ZOW!*1}%FgvquY76E&_A72QPb|*
zxcTGt`4+cyvidXGx#80p-^@R|P|OyM>N$g`#%<VggN}Dq^Y!wR?#ZgMXP+ZmgWmu3
z>qpRXOv`8M^J%)Cw^v<`>gVTF+gHs;mwmVMEgh%xrC*hGUFWylsM71&2g>jJj<VZV
z^!~Fg@uGb;zwta*x^FK_ZQ<$FPruKD*pgn>cfDYzLecr%w=oe1Tl#~WpZs%6riA(V
zEY-uJ-~H58Z_)iETjKfqBVoDOa^K3YpZxth)5Egg!_;L@^jw-P@wzyvD#z<O(A0FY
zuD`?D5zV7rOU|8hQ>O)vEOonI$sd<ht}HuYpSSRv-|utEZt3~wl5GjAe(u@r`E34p
z*ztz^2R1CLHJ^#*9k21K`$bUuUi}Z6&hNgFTl%1W$jY~(?hS`eZl500rLz1s-Ht8u
zJMX7$@4NH<+WNdK`K|l+R8^Jt<KSm6JhMfKm~Hy)6MJ56y|;ACc0KjnudjZw=ZP^>
zH0*qRai3*xUKL;Mn5I-O)AxDY{4@1B-LGH#KBX;T{yN8RnEIh=wPnBBR6kX>&wVF<
zR^`F#w>Z#fIfQ+GF~8?mJx;>ff2Q-!3Y+Q{jH%MF{_*+U_90Xq=lyUC^Vf53VOyV9
z75OdueIZrPYe#L@?fq-7$3^|IS)K8h)oGg%Gfh>0uU@r%TYpEB{2qLKx4~mhPK%kQ
zqJHca?{-}n^f+HGe}9p}{Chk6^WFS?HNV$IqV%$_3;jMPB1+G5GbMcdu2*AqGWI#D
zRm4SByxa47x{u3lU+3oJA*aa~jx3e!pPN0J7s9GrIi?<`_ex}|tojc6<o%Z(Ihez2
z)1O{CY0SqJH!@qb&3hLuKX2$pCR_bJ7sIFL&)>Skrri(F?GJzd)|RknU8Xud=sYNT
zPr)r-^xV1Mc(-!gzWYkw&oNC^zYig+zCrtq-+a2)pD6E*cbYBP$aHbI;Rk2x%@34{
zx#_t5rOPHhK0X#y{q9Tq{mzHmb0hRTi%7j*N4mW??DqQzZue2#elNyt|Jmzt(YnCz
z{2)8MtoK)=w!@!CWOE{WqW)}JJo&3y{jOv3YkJk~5S9H@{l1^+^|Jjv237Z)b=eJi
zU&OS0{QVWLVSc{xyWUitUe<TD_<HIJ^ZPie%XM4Ni;jEsOV<4x*tx9nsH@g7+c@aD
zzD7C!yq^*F`Yh`H7S?~$^Lw2~O-UE_`=fsIyY1tce%$N#L}j<HE%TwO?^MX<DYEDF
zx+Cg&RCf7p_nkz%Oo{jVeo=l&Z%X^P`4M!VMr1d?w69w|Rn5m_SH#~fo`3$tw(!k8
zI=%hbiY&*gy78dnUhRp}ksZ;xOxAaN-0bk<nf%hF{Cy;ZuNZN`7n_-C4yNsk!k_M!
zW<Q?C!g8?vK1HhduWSC4RgNm|{P{&!SQeMCcHGu|rW{xO9H?shvY&(Y+U~Ni?^W6J
zS|7jZRIgK1+gCmQ{PLvhHCxr|zFgGlMC~HFUgP8a&NuvfmTqCy^QD`8zu$xO`+jH8
z{%y<mrg=X_{VmnvgU+YK!SD4?*z={|@=Qr@`u$Zm|5ROX=(fv0H)LzrwtZdO{dGJ4
z2R%;ns4A7UZj<Go>gxmEzAV2*c7mQik+1sBE9Yl3Ssc7RmuqS|zt<<C{z&nn=TQ;g
zyw<>jX))CtMl5Yv{o|I099vb@`&!gLMQZAIN&WiecRZVxKJ52pqRuaRpVYK?(Kt{Q
zFYA08wH;aO+NkW<^4w`qKlybyAC}#}{qAS#PA|K@B73^)CA++!`^<jxiTLsHQOys!
z%GGs$KWI6o?HdKHhui%d9A9^P){IOR6$h{1DM7j&gUV&Bb%+B$Uj_}ky(b~-xGDF8
zg3cqd^7*}AY)X1n_iKXoTU+z-_xpO+_`4>_4#`%Ds__!>lJWHJ$;Qcefg^tt8HqB9
zvhlR8$wav$;cD^pczVMYRk|e;1A5fkP$^wHhtHEmKX_*IgO07*cFN;F7Q7<U5ouUp
zQ?KrNLFTC6nQRsLC8~$&c#%Ir<6mp~$b>s~<T^n$Xt}ELefik$7kxG1AoHstf2wae
zSg-el^(?50?8=VMDm-=gooP!PRVv$Gk-b|FURbm3=JgKrpE-!Q`Hdfx+j_>3VT&2`
zpE>yDDN3*Ue8Afm`GNSq2+V!%ss?v#-tO3{QO%1b>owYV`S`=kP5rLFRQF3;=eIW3
z*8cUS>A6g|UR-4yJJ?|k3#u0uXtDIxvlxTQ<sfUE1YO6Pme1C4ohp9w>ke6AS^ZpO
z$JYH5uYRZXo-lgoPn(%Yzqz4E&)xHOvZ!?B<nw>(J$*kDwU6%j#Oz5mUS7H1(WScU
z=eazWek1BnVdGWpSCHMH>ljn=HNUfA-!3~=J5aMZcphgi3BAtCE@qS+mUZ2%%t4Qa
zqq>f^y<EZ@^nQV?eB?(^l|8@fTGxm5J18&N`JUZ2<8bj`_Oh@lt3QeCr20O_rI-IU
za>2)|nQiK|Kgzl;<?USm<mXwBEZWY>oRPS$^PJs*mESM*S}&wKmO$OxJ}$Io{Z2=h
z9$R{Jp|W3WANq3MW+!Tmj=-@cr}}*DR=;Cu$Pek-2lo5?hFkiu^K{htz3%@Zoto$5
zbgl1Im6Iy{LKPq0uGg6}?7YCQ|Dp=JomZkhuKPZ?uJUwUcX{p4Nav40*!>u{{Jb~(
zb*QSan_ofut=oRjqaBM6AGAC@lh?U7wQY0pl7g8*mF_!|@?7(4*sVS2xM+PH^!yi;
zchuLpRDMwZ`Fy(0GpfpQn<q>^E_?o@_BFlUynmwp!LN%&VXx~$zFe<i;;Kr$`j2$~
z{#>Ki`<tTsEq>!w^#`wZb?t|V%HjPKjbpFzLDzMunvdrjG%RXAzwtrK51NnGJF+yW
zAG-5-Juj(Fr@P)%Ug7=sKeH!$uVcH?bAYy~>it@3=crZH_)*Qz>-p4gx>P-n%Bq*>
z`oq?ESzMyB<F|jOTKcHxPgUjjYp}7^S@RD&B2`tcD!WmQkFe#cIzQ^RD|?@f?5XN6
zWLMU8PgHhnTlbQ`e)oa$dwS}xewBDb)qFq4Z+n=MK2`R)={#YZ((~oIRCT}8^;~wU
zmZQ3Tc>A`$Z)*GbEb8lIzCXwgU;AC%sTZ7(<v_i0ux))SyT6b<S?dzAr%P4uN6WGw
zl!xy91W|cG{kL`hnEci?Kk$61{ID8!pDb*CzFvIT>-dWLyp8hz{kwTltvC4o6t=&H
z%`dB;@bwBCp7rV3TT3ka)A6gHR!L)z=IwT(*0A$n*y}E@`^>7(7pnGKqPngN+pfCv
z%NpOR+R^Lz)V6f8u1|P7s@lO+yJ62GrsR)mz6x8e*YgOaQ++;(+P-KW3fj-2de4_H
zH`e#f4rfmJli8+xJ!}h$_AkQ52knot^6~F&+8Xxy9*b$|Xg;=GWqoHdYCEFqQ5tW4
z&ud=yFX+8p-5U0~BW%A7yY7i9zpnfIe#?payc)JYdCed8yql`|O}#$!>zCho*0l6~
z-^22I-gE1}{C<Qe9CZHlI!-B_-~BYoXRG>sPt>&Ze$P9mr8kYIUw>rx`>^+){N{I?
zpHv;Eb-HQ|d*4TQe!ufCrB~HXLG6n6^<>9W|M7bs7p3=mem6C}?tUw)Jiq%s)c(4Z
z#+xqHRWDIFUe}#|uN%Ck_qs1b>0TT6QORA;e8b$-Yx%O$dF2>39dYogqU(*Y@j<zT
zE!Qo7P`|w9^Sdw4&zo-NN3VWRIyZa#aj)w~UF!nB<Duo>zwgQ`Kfn34o8Qv&yi5tp
zzTVPp&(!rI`K79z$$n5(<(DkGe)kRd^kIMRF>L;*??;3!U-vkcU7p*z#q{I+I>NTF
z>3rS%6J3vbUC+6t6Sb$CU$!3iyAL5s@Av!?bY0Eo^INVcz3uDXu+Ke2oj>S$iO(l`
zUWVdLd%vIhqwW1KtpCJ8mh$UP+rn<=Dcg?Qc7CLNrLZciKa-uP=V`y?i|#+Ej`!=Q
zDd|P)Teo<xzxz(<ygnZh^!HiZu5*LV$6o!WbhgctB7f<3hip^Z-U|`=8+06)l27(N
zPE_{X=0Q46=S#mT>-xZNxlyI}djB+RI{vynzlCMDZ`kXtsPl{V+5E=yT>a*EOCNUq
zFUlX){7LJgR8v#Vv%DX=-=9)l9$!CO!>0Csem+a}u;{v5SH0b?Q%yh4UoVE`X4^iU
zSATg<rib0uC8i#adR>{{%QyAB=G8CJ`9ODk*m)HBK>!5oArN+dOIH4{#7sT^g{_xc
z{^u)wn0@i;!ZBIuR-WqPIS(8+<G#ECF<q*<KBm~G`vbNq-M6(>3trZLYR5%|GNI9O
zn0r&(HWx1`m}#3PiXQHqHor)w?RwdWiMz}GQasZX4LvXSo-lgoPn%iT_cEmC|GQ$M
z`!8POWv%OA$J_7Jn8qpWzC-Bo<<Wqt>wa7O@`yQbGSm9CUw^#Lm!_xlTW_R4cHRy=
z05oL``dq2ke6p(%)qN<x_2A#LaSMk%Z<>~0_PoZQ7iHUz%ATzGQq(_1?;l1Tuljuk
z)%K(27q)y;?}v-}v0FU<y&PF#zFg6Gb(=p_J-_Sqc@D4X-0I==IM2<LaM*F<H9vp8
zk{vc}{K{@$+3OkE{<+yp%$`)^<(2!Lu<BNhsmE14uPv*-s@Dgq?9+SswyLc6g?Kw*
z=U35rT~~b2`CRrq8I?U-`*T!&2b~Aq)-R%d=ye>p^&hwL-HyAh>qY)kyjy-c?)Ud7
z+~yy*b+cQ3ugA%Mkp}HIe)D;~-axt*OK&}kG21wJUElZ}M|^s}?@RId-NIql=PPdc
zzT*2Ue|OxJU$6KLyUnX^<@al_vDI1g4?AI1)b4)g2ifUW?XT#z%bzD?b0T}T%s1q(
zYV~@4&F|?{w?kC+y{>PO?%4a=VR_KH+*TE>vu%k_mHEu=x>0nTpKqdizCd-kd_8rA
z`TY^q<$8TCg-_?zFWu`wQF&qK-Kg{1GC%TuM$L`N=l${;ru?>Q*y|Qk&v&-82Y-Fa
zhrQ0DrljNjR~;6uLu`o`?f>xcvfKB|$9}))s|g30>FRgC(rbFzJVo}r?uYu_zcf8P
z-+%eAh?gnxs^=|R?VHj*ZhpA!-;0im?3S$8XyfJM53{gp+xL@sfBpXMqTlo){%-N6
z&xdaQxxLT8A9v$%^my3!WB)cJ+D}pST$-ERqxIADk7(Zj@#EOR4s)0)4yImT!mp{r
z?@U|bfP>u}G)<q<ib=KpCJjtTLvGK*m~!0fzHEL;=e5~I*E??UZqIwkKJK+Xe$(mt
zj+v@*;#ahJWpj3+m@4(l)Ao7Tmiq&ucJb=J-+d;(-=*{VU0qSSu-6TK`=Q_YD(G`G
zs_sAd_1|xLQGZLdc)#O<(h~>2{Z6!h;x#_(IEgBM*!#;-=g%+eNL%XZcifqh-tT$U
zl=Q0R4Vp*ui>g`=i0rE7O6No9<3AR>BGZYAgR1eV%dTH;Zv9#`9&Cws+mFF<Z|?^N
z<-|W<Ei3HSZ-e@4T0UFGnaEF5_PK-lVOoC>`QtVoWFNQncMITmk)FHf?PRuaaN94G
z-LIv8w?6FiLVSMd75-;~UiZUIP3LzV95nqu)BE38(fKh|;$^MtRM|0YKQ8Ecgnw=k
z;j0?lv3a{=3WuQcNPf@vzcXgA7`OF;tmCG%SM8ol8{M}uEi)<&`}}@X`D1<G>~Q9k
zKbdR|)@$v}#gC`nHU6$kvO}^}qH4Tk{P=i!_hjQ_yugvaiHt;<MA>*+*JPsHk#Mzm
zdOW>hiz?lci2*(8ZK#xPF9~=K0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&knafeuJL!3lO2++5>?~J$4`i-
zcTYA>#tR(zo5)C%NtBJJbxkJ99SK*9r^nM9wy4rAnHbQc-iAu)`3@K=3<4kk0w4ea
zsYu|{6R-Sy(x`2YU;U+O%d7@3u6A^3DzZnNKmY_l00ck)1V8`;KmY_l00ck)1V8`;
z^b=V2<*#p4c{PtkrTS}(5<vh2KmY_lz#IaPOsQS+vW?j+YBh%lJOKd^009sH0T2KI
z5CDPHA&}ec)-l8OtYKmG#z!Cc{heWJm`GEH7-|9nAOHe-36vZ1_)lwI&t<l#UOFfZ
z1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;KmY_l00ck)1V8`;
zKmY_l00ck)1V8`;Kp@o#JW;zsn;-AVWkGfJBbAfo2IMj*ILPP+sXpFNKM()`5C8!+
z1n!@g*66FJ@)-0l4r(aFDhPl82!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*>S
zC(v{EfMN^J&SB8h<`DJow1h2x>xZwsGiG@ngMx#J{XqZBjS@is1V8`;KmY_l00cll
zJ%O7#pYhVW!?Ky1*8Ao7Cri{`!~AOB8w<UgQ#^-3!2$gM<3RU3Kr`s4rGL5bkr%R=
zSMA$k+*!HB*RZfk<wHRL1oROoa>vSB`n<fD$<p)IwLSIci&rq&I_`>huD`YDD&|#_
z&Vyf9ZMKTl%xKwVdfDJQW4BZ5-3!(*C^(=Wm^Tje&6Fq!1V8`;KmY_lASwa{Z#lQ)
z#b5rwyy{CE=gs}#`l$>G4l4UWY+%18#cSs>uR3z_-LD*2kjtRqI7BrMp<ECE0domF
z*?7merYHUE*s6otwpdqr&I-p?z5BGQI$ZVgcE@&oZ0}VY=PcRbK*8ZiKQMR3gJ&QB
z0w4eaAOHd&00JNY0w4eaAOHd&00JNY0=5z;m;C;`HLvHf{I0gPAN&RZ5C8!X009sH
zf&5P3(pziII?yziK~sZ6e(%qyKL~&T2!H?xfB*=900@8p2!H?xfPlFKey+Uahg*uS
zX1es=o%4Ebo4kspx|-YK@C*b%00ck)1V8`;KmY_l00ck)1VF$qfr>3&n3_14!=T{c
zr62h1Uq}xEAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY
z0w7=~fibu2eWJ?^YgyE4Cp~zbsswu1_`7Dw4#`%Ds__%z8S(V)$;Qcefg^tt8HqB9
zvhlR8$wav$;cD^pczVMYRk|e;1A5fkP$@lCxu9Mk00JNY0w4eaAOHd&00JNY0w4ea
zAOHd&00JNY0w4eaAOHd>N?`93s|u%Iy_`j*DT)=<0s#;J0T2KI5C8!X009sH0T2KI
z5C8!X$mayUzhm(^7xd3&VfB=64%g}M>1rm@e2y8_2LTWO0T2KI5C8!X009sH0T2KI
z5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X
z009sPKw!%5^%Gmq$zeftK+T3jZ_LVOP;ijZ4+3;yWB~yX009sH0T2KI5C8!X0D+Vs
z@cr&@UKsFY?U-Nvs#>8nPi54MdDRg^UYK;|&>AtXI{LZX371@ZNel`O=m!`FUgv?7
zn3Pco5C8!X009sH0T2KI5C8!X00CnNtk}YOpLEuA22E`auUG!`?qB8|cA(&ZejtAw
z7&D>65fA_Y5C8!X009sH0T7T#V8hmtcPH+c#N70a-%544__e9buU_`phMohPPGeAT
zKtI4Z&^-@GoGM`o1VF$N0<!0s<2y|5_2uBXOt!8VS#xBqCl@fU`pm_r@4dbFGFI)9
zwz=)wFA1#KeKuFxSu>kK!2$iiym4U3Gz&LD00ck)1V8`;LJ+v{P{ZC6OU-b+YM0^<
zPVTeeTL%gbD*Hi*E{&`p00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&
z00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0w4eaAOHd&00JNY0{=qbz_vR#3|+j0
zx#=z2iX3>W-7@A?r+#x=wYQ6|V$fqb)ZcnytvYp9GblKqAEeGW_?Icx$V1;X?Q{OL
zQNKI7bjH}x*LJ>Pfs^WLWMjiA5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH
z0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!X009sH0T2KI5C8!w1h!v(VB7_5
za+qH&a!0p1rE2CdSt_M6FaiRS2=uP;x1nT*WUEBgct*TbJiU9eaWY=u$lpXpqD-P}
zJgsXoQSL~%T0A|T-mpcLZpp-e9`!a<N|)ruf6eS3)u!$nB_}(!>N~H$)+6zBt|QX_
z)->3Es<?=}c_h7jxqb1Kj&2qC6Er@o|3S-Dm2Yc{?_auh+HS|Miu}3Y{=p0C=geb4
zRb*Fo{4Z}`l6B+qPL3*-?XSpQZijwZ<-4w6(0}G2;^sI0#*d%xdd+FSIkxM@4<8z_
z_{E7#w)*8MN*{FI2%1mi2jc%C&^FoSxP>=vaBS6}{UU5Wue|)uC#w6U?0G?#Jyqv%
zH@jiigCl>dc6!Aj^Bt&M4zk9H-|=Z`dbfUT`th*yan$)$&4aS~xyX(xzjfOc?RU8O
zA<CyKp3X0#_R$@$d!Eo;Kd<F^UEj%0=heTU=`MaF{j!aP3S`=<t9F+gzUYFT%vRmD
z?X-4hPTs~+U4!;dKA+qCO~?6sy27%~6T18rl_RSj`OOaTq<$Eck{?0!%R(7#_doOp
zlchfvd*!6JveGhDYoB5zzZyR001K)?dCATf_4SL_@>6|$M>T%Kmg}`Ys%k&Fk6ote
z*!0D_w>h%ZYrT-}SOQ`Dli&QumL6S5_w}o&yyyrVTXKHCZ+t8b`5|5V!D-FL#NNDj
zrvv?qgR1kGF1w<61MzB~+abQX+F5luy5@JUGpFC@%=q*xT8$jPzxqBW)s(Mi*l?<z
z2V`A0xY_aApOMZV0oD83yf<|JOtrervvgkZt8VK@(Krt~fAHgj=bB%`Uhn6L&QC$x
zB`9yd*XgR#2lb!NH~56kP3m7+Fw<5Ix({tizMl@4zVO;Zdzmby_VqfhOi4%knL*X-
z^L|0kTf|kBx}BGB+}ryCUe^O&`-@vTRsF%M-KhHo?`P2SG37I5J~G7*o^Q~wtaV7x
z@`L6JdOi^43+jjNd<`ex@X((p=CY_&cfF}R)%$q5?8_Q=y6w1~_re~hc8*$gbCp$|
zs{R|cU4!OVRUg@QgI+(FnlEU3nUYU-``TtN>iJVu`EKo?s$5leMe9-1<K5aJ>~UG=
zLpM8K_XYf>Q`KL{?zz8SllSo2^~_d{`uZhk`GuF%X|>>oEez_71NB$GYU;WuRs8bX
z9;T${@0TJh_c?Djo}&BU%HkDwJiq64q(An#K*i1TKjy=(E0I6k_Zq#vpQI|?|8xk+
z5pLHDrXSb6|6<#ClH6U6nv-?E^{7Ej|IqzaH?8^EGn4mr`Q3?1gW_;o&*(ZH)PK|W
zflTr5fA~szPe=EB%9r;)_2MGD&cCU)ZuMGkThc|<K3sQ7<NKfPu&L)!S<hicWhd-<
zJnT3!<+&+U=X+WG)~}tJ4-Y>${DVInTXb~WVOKqre$WZ44eB&$(Pm>>OtwxKH~0C;
z<qO1ktrffNql%3S#%xvDoOIa}z26a4Jb%8lHSG2tm8r-5u18F*hph9V=)4@XUq-bK
z<;(Tk&ZeaId;ZQZ>4Wx1S@~40`*hpo#|gs6-j7nvBUSGIyUmwhPis-T_3+t_NJaa8
zZs#?(_VM~$4*%WRU;a30>l?jyGm(0|?iPKIS9G5B8ZY8ydi<dFL!LY1k;9H@>h<~~
z{ni_r)qA6OOtyYA@z=4>b}JN9t!H1?aQVWY3de$~=z3MP-(=rwJNZq=U!SWAo7V6B
z`V)4a#Vx<zeJ8*3o!fcJ?R-MV{gzAV{jTra?xXU@!><3`&bK)J@4u@>?QE->(!Orv
zRM&Cc?HJYdfUN6cH#=U}x2B{E`Wz}hFGn5LwZBd6FKZq#{rNbr{+YUO?B<u(`cXP+
ze_g8UJV52>s+Xu7ulL=lJ!GrjdD-jv)9<>~t6fn#zvTrzFH^tamF7p@50Sra@owiS
zzt<hQ(!14*KhE=V3;Vsk<?Xp~<<q-``TDqpy^c3GJF<>@^@q}l<|}@Fb2|^2eq0t`
zksY3wDPdjfN>$|qo!52cQ`OF1?~91isq#yf-LUOL^H)?Vng?u&_q&g4T6(|h2Gi23
zx__a|u4z2|`XgF5%Gw9w?^jcp|BeKG&qFki+Y(RrUBas0^9x<4g;l@hiqhNCULrqb
z??>pi7q%Zzez)Jx6g@W*RXp`aetyYsVd5Z5MdQSlc+vH}t?@KIZBe)VCa=e3^=Go9
z`uAdF+o$!DZWTRWXnMTYe#NKryZ()O90uJ-;P?C8!eQ67rsUVX9+OobT_@SDrtg!x
z`DaQ$4(o^4eo5y`zbb2e=(pUc(tG6@HeJ|uw>`h?_O)fdTI46cehC`pxyl;9WXEk@
z4SL-6{%o56)Ly2k?CUAmJFVTBlee+3986u0^7C1$heh*<u6l=E*F>G4=NFcnt^12^
ze!J}xn07p>=Oe<FYufqOuRr{FvMn68{i4qAHjZ^27tW)SzE?@lG5CG{LG+$*RPlcI
zox`U0%gZgjs^=e7+g1Hsm)~&~^g7s-e17YzD!p*t@p!=0=LS5wA2^xpb$DP$BxQW{
zgkqaoUvZe}Qb}$bZ$`8qX?ncu{;k`d>G$2;{OeP!<X6Mz9AHpz2-6P+pU}BU{VNM*
znxbJRY*Vj~kJc;X^;*==$)rm8s@BV@?S{QiaI_yP{wL~v8{5lQ+&pQfuP@#FOO<)Z
z_I?_*|EkAHSo?1MIPCGT{_**BeGh^!hxfxR?DzQ`zC7E*vgbek`DfGaKL+(j*8C*u
zkE-|8gW9M2Iku|b@oifAp#4>KK2bk*i|6MDSz*3hkzK$2SXO_KJ<rP8=j6GW68^GK
zM%(=l{lQeLtoF6Vj_mf8)ep&zZ2#Qs$-1xTR*tF1Wvy#?JF@EQ^>-6YOBZ&3E$aNL
z-fN|Cp5N5e`w4#ivbEi!dLBpCJZPJpcIBTg_3)F$Geugm&z~c1UsNPhr29YIx2=Rz
zEK|3N{P7wu>;9*%_Kx~IgYvJ*dwA`7riuf<E)#{r?icud{#A5+E{b>aPgF0z@nOqx
zJD-Z`Cp%u`hpu@=RE`^8TaNSXC8}@O`24!BAmSGGK37!v{mw&S(~J1o9&g%pzF&W2
z?KkPRqiR1umtDW@WLkQ+=Xg{7I6rThcHdprdrN#hb%pu&NmZ8{^!Y}wewngQ6||nF
z<cn(GENr=ae~u~~^twJOe`S>$w*5@aFN=?AJ8tub*Ll>GbgFsr_D$`7Uj6dAu9n@t
zvY&HNWlz;UfNY*3dtT2gUi(X`q!aNnC7$0mM7Vspeeso!DedFtN6>L1vMa0oRn3#S
z?5diN$*zdMTYS*=HYK0meL^=Lbey;bRk9=Er0YHqzmG(I=~7+e3*~e<ZsCm^9O7Wg
z{t*0lXvE?dCo)+aysnFpPVI9iy5>(=<%Iq2t|)(0_hn>p32Vn~AIg;DqUU97iI+8h
zk{z$@9<=?a99`<QK7P}Q#=R}^UiS%urt`~F*7a~ycKGWFQ^TTm@%mjpzt2bbU2pOG
z@qWX8{o~UI?eAXm`MoZu^j_`zO-J@ssi?oDT71y+8FBD?-tarFQYF1;-^iBusLoF)
z_pkd6JcqE)+aiCPx^I-6>_p9h=22BD>-;FoKh@ty;O)!uTV%(U`8(D9w0)i6cO50_
zj}-6Mo@wc$+J}@?E`NPyTiEaQnAh~wKSiqRKC0?+=)Rq;%C9$V3rD@K^IE=W9H@%7
z<^4=IKW!aX`NeOy{x7?Kn?66uuCK_Rtn0m~?D*X`51L-YkB@ij$F?7Ln`fgw?)N(#
zrlt@2JzL}(IevfjeU54lquUO<>Y?<54%8n9e!dJE*0nAQ`uvCNd|t1YRi%sS{2sR4
zpyzwHd_nuIt@-%z*1N{vHA!|zwn|iumx`B;r*}^_PR0uy`J2c{lu4A0r*%yx${h(;
zi>Jra8@8y@Etwe5quz!}>C!p$ZMFRVuHEM_TXo~Iq1z|4S<FNl7Fg#lA9w8e<Vq%6
zMSi{iw)1}POTK5iRpd|5`1|*LGv&pH7P6okv|LsBuKcn|Zuc8=9KR~^=hn81?;Bcw
zjuTWxc4f!EFyYGg+YMU5RH<x#MfPS7N`6+iSq_8#GY1hjzww>!sI<G@swvENy|-3_
z^$&cofXP<BJVoidt$6YLypc1RE*1HK_`e8ToU{Jh{<Wq#w(7Ns`s?eK{Ke6&s`{_$
zez~PzM%Id2OPFoi?EB8=U36d`)2%ILjelg`)O9SVF5ED&Zm&ss3@Vp{tZ~wA+!y5w
zKem+FuClN5blaO&dDgn3<FXm_SPruKxyVl8_YSiTm9m-bTJZDBCyn}Y72|cl{=r$(
ze*KV%^z9R#TGq1nD8}n&$$6#H9~r|`YyUFEn$DX20~586?s(nfUU&Vxme>5LwMB25
zIiKm)n_fM5`^i<-FjZ>%I4|1t!H;`w%VxG}x1E<%@6sci*{V-D?UvOpww$@C*Ky{S
z?%nLtgKphCo4Ki*9aYB@&vxqaU4g|ctg7<Qt6kgIzft*5e4}2sk{>~}as92$X0Ko7
z$kLlO9shZUFK0Wd_1ya^?0UH06ep+#<t01cqm6q#w|LM*7FKhL9)J56r3N!qy0zpd
zzm~pfC`(oC*n8Z`)k;lawyD?t`1%{e%AdD?1e2|F-QYG4QaX|H_g%f#3+av}p!>Oq
zV@r=NbpF#>C$wvq<wUK~5jeJF+xGco`+F=6`5|4`q3O9<c^@oW$)JC6n3-37O68tw
zn60X--=Lg--Y>GX-|&8m)={$Kz0RCr=clfP5|>WMSk0o=_;sx=yJW{2CQJGI5@EwT
z$}fGR%z~vX)%3L~zm!_Bay3(}Ui&lB`6D1}yol`Z-u!vran&BEyMn3Kn=ktI)N{+P
zcSL&quyxIMc3$iF)h0DYKen&sTqaV{b4L7ivfnVjU&?dMuVJqod43f3nx9V>l(*`B
z@6~=#|M`5w78iK&g~eHpt-7#Q<@%+c8s^xlOXrPVP-9t7CQGS(Mf01f@nP4`G#|LB
zXnuE#C$6eAZ2gh{ulKL^4H$USjqR2*H{II2!#8^iFJ)m>Re$hmSJnGAUhVRJir%mD
z8vo>rB@R@qvy9oQvab($zCpvX=KY}M2hI0P+QKonFJ8`MX;43O=R0)P<4cBQjB=t@
z-SwvOw)|RdZ;59QIksz&<A0dgcWA+wZCdxqOM7&9tZ+<~QaeYjU)}w1m%>d-#boQ%
z$8{dCYig;ODy=<o_RwYn3dLmWMc>rEWa~reG2QxH&O@!T=M{>Xt^@K09oOjWEl#Rv
zu`gcg@ZFSyj;-3F#NoVJ-3!O^t42M4sw&^@IYw3Gs<QiNot5{^$y&^8(;oeEiyiJf
zi>cB#?!9rtski>ZRO_q7bMM>L{97hlRrMFL8}`0s)cH%l|L&F9cjq$Ol)vwf@V`I*
zL;cmSs(Nl3cK>=V-EVuCl0H@1kAKhAmay*c@~G~2{QX~q{e5>$b^Gx4+g*K5nXKRT
zIJW9NLw>8?spKDyty<uLMoS;K=T9dp<@<x|@Z!0ps&wDF(XmzU-Z{Aa@XDJVUHWcz
znYY$%+2F|5>~aSiT-AApBU@$n7qWNgthPm7EV+_}RkEi`U;nPp@%`^y!`yUm&qJ?Y
z_gOZJN`vw+<^BGk{(F7y%hYr$7nSRD$0ZAxYUTMZFFUGlod@Q#u-dEQ*HbSmKbu9R
ze7*Rv*YA@={T#pS@{6?o@Y7$PTFp{L-{|<~=Fw;8Fx|@cr>L)g+@90(dYrFU*l_Et
zDdRsaHi7w7-QVFBmG{!!-#`A%$un3`-8<m((>pvl!U?K3cU`=q#P%1Mo95K{VchfI
z40lxN<<C9v+VU^wJF?Xo|4!u#PF&>ZR$KOs$WK}4(Zx+~8GB~$c}%vx-1oM(-@9=-
z<5l;*m+bPcyY-u~Q>zVdWa|Ytyf*#!(H}UfRP?>TsN%z(M@-2d^?5#M`Kj_ed(isx
z`9$lXp#ALN6%~u0k(=X0t>1V1xp&JG*EqK6sr{;cv?e*hu~k*wUklnFW#yB#?ugpX
zE2p2<a7Nw?W}E8%4w9<!bbt3=Re7R*OylkTV(UgU*}BT{s$So>o?i0QU(z<Lb#&{N
z)8C#vc+74msPgAszhPPL*}%?U@9F#HZ)<wd`p#>-t@X|?eyhHo@!C&xr5o3x<#!c^
zEp)tU%?C$3Hh=zfC#+Vz^r!K6ZfA~P{ph`evCJQCV_x;ev3LFaRqu|DSM_Sw)O23k
z(bROZzOx*aooV;)xbXG7B~H}(-IO~=*S{{?@vHyoJ^t{x(%FtI-MM?)<Fh(tJGyk`
zm!s-k-+h&%TWOuHTFVb>@!Ah(jCE|$pQ|6dt#;ur9NSeFZ|Xm8-;2iaV}B1SRd`VQ
z>(ZdV4-hn;u6l{eiMpTMQe;}0n=^BmS5@`<h+fN=l}=SXU{~(%NAf)V{=O%l-<Gi7
zdPbd|pWjsPTd20LdS2#vnG#lAZ`Jly_lvOhRkd^2{W(>B$+COb<>?joKflchs?~Z<
z?tfy1-A+_`(cy2udH3r>j;*?)*qH;mKKX}Zt3ES(_On0t+2+`)<q8a4ete}(j;*Td
z`9EED@2%Bf{R1B?V7$t@|1SET=(<y{KXGusMU0f@dAHy7H2wJF2d;58eY}K)RlnyK
zS>FfsTdpX5iLx`EA3A;wvsGsgN`6+iSq`&Ry}lpimaagjO_|-_&xncip()P1YU@kI
zOi}8O<(rQ<)U;-WSbkOFAWN_PblO#)_S@;$s`ICAFL-&eeNL)q+8?iMs5v6nu|<cR
z{PBQR#j>2R>eint<_>83?Nz76yz1!ET`pR9^BFN)bmNAFPv3s#@iE<6x60TD7Myot
z%r@P0`7PINzqe4#uYOVgw&Az7C>Zmqtxp=gclx;nV{UrQ*skL?T~i?DS9czMq-Os)
z1!5xYy{}FB<F^%z`PJM}uf8^RVu6@fb*rB($A@m-Q}V2?yB*uKbm!~}OS*1wY*$<N
zE6DGg9*fsG;nsDI?K*H!@pX^x+~T;Y+j&=X{L-He4ar<K&k3s{`=-Uun0W8&GqQF$
zs`dQ3H)lP)`k<3vRd&1E{@nF~_cpF5JZ&xut321+UweDb%gytcEbTMq-0nr+&SSDQ
z>!JNGH#;Yn=~mmm<LlMma;G-mQ+972gR12qYaH~ez4O|-7p!AZsjT@ywjF*xOZBkm
zyHL97ZTowA{5l{kx0CMLTkg+lc`Vgb_xHNOmZy7NAgesD=WA2b$>I=|9oyGGrt}NZ
zJX*YJdgHP;{@}>c(qGOzA)Ytd@v6F>v+!EZ=+mm-FtF}8rb|WN%?ldee!*Q=9A9rT
zvrQju-0Qi;gC;Un>i4}SQ{Q{_>eqSSZcDFp`*<g670o+d;~zh9SF;^YPGY91SO0^i
z^V?tD(g*cJR=yk4Gsi#w!V*W9%JSQEJ4@${UQlCMPiC98p4s5-$A4b!*s3MIYkuaV
z^L9J7YSVM7JXx>ZKF4<bD8152hj*mKY||P;%6>5YK#`d3I_X!ZW%EZ$#7xn!^R=n(
z?^C_Pz9-ae#f#_XjhxA%(xBHhrsY$0oub-q*mkFWs9J5iel^Ws)$LR0^IE-EfAyVX
ztCl-2IjeK$A069O*1lbSwWE5Rgteb|=e+(uFP^|uX;}aG{LP<QTlA)x^O<ht{csEW
zT~D;ixIepFolKT0D*JaH_<hDI7oBt6mN7p&wyUgt6;Xf8Zga|6_uszSF-^by;Nj*U
z44UiMstapXu3!48VUDe8>O3Rr$8PbW_mIQJ^W}=plRf(97CYQ|79+jwKaG2>d?$;M
z^1s(7>-T*&E*rXiLYu`*mWKVEPs-18GbOz4oc4WdH!c}dt)lPH%Zh)lT=)4chD~Lr
zsO<KY{hXO>|J>~P{a&mny<0h^9+&-`C)tx#U;cX-wuD>%c;e@MFI&WHRr(%?Yz?~(
zi~2eu=zC^Go^Lp)ZNs&W?JE2C`*hh0`u*0h`GU@a*H8N3=|6AJW+IJW*XpuMcC2B%
zy8Vti9WVUeVb-BiHWR7a?<IG+wqVmQu35!IDzZ!QqUYl&UQ};Bp4PFtR8(Jz-#1|3
zO*gh%&fL`P_tJy*8^8I&ejf?><$PyK6~_bF-Osp)?RVTX?DH!VUcWVCRy@z~s=X?H
zGkj)+4USiJ`#ng}x}4%~FWlsc4c{+!+|=!NLNY5%e)q~}7dawz`(5Yi`}@usRC1mp
z((|9rI-y;=EGMkq+II1ML+j6Rc=bC!$WBl5r0q)QQBv8Q$ljobo|xQi!f%f0O8%-=
z{=G0;!t`B5U8=erqO#xjnYUV3zIQFNRp*~`?|mJ1=Q7*0@wD@|KlD~EvsHW5yl>{b
zojJ@_jq1HS+2y`kJomm`&A(-Sm7j0c^__a$f6mzDn4;&sR;0}EJzJfy%GXm@xQx?k
z;VCz~>yWB0x5alQZ=1P&v*TC2`en-dmTup5HT`(EotISa(j%MMrmBAj!mE9LzshTv
z_Q7mb+UE<ZUOA~sC;Hwrzu#(V_=H6l-ZSNm941n)^QbB5hM)h);$MFm=;+dpcT`@r
zWOcS<t9sopklntrp9fH7Pu0GGY@Q-}eLj8al}CR3mAUDQOE!<_^z1a2D*A4A=|Q({
zp3Oum;$=#_*L6dFOJ_>^xcO1=^UEiV`f?TXt0KF)_5<^KIU@dU@voeITEiK6GnlP9
z^yH6b*DUfA^Qvw<=y=%s?PRBKtL681?LLR`I(Ocyhc@lYVycz@9V7Bfm)d$C%=YtB
z(Cbpu@)d3R;K#kTWi#7URsW&;vAT3-UiB%Jd#+)&sp{VY(=~s}Do57u?D2NOeg`~g
zepy_?+Og$%?d<oTzoY%(1<Y2J{hqDYcHh!3BWp#iCCqg7S|7jZ`1PZ0;gyTZb-Lq{
z1x%It<+*#+)sJ2{a2c~zRi9U7*%!4-PMsgdJ^#&c$4$L{*Ok)c6g~d-FG>w&UiHPX
zcm4cT?~aZw>i2nuaUbm~xMbOQW{S%G4ghaY)ZbDqKInRfIOw`=p>k|}9%|u+iFJET
z%44>wt<TZ%e!shS@|V9&$Yr)Gy|-$c^6g<uSk<_tc~q6kx;~WUpRDzuS39!&7TMvy
zzk%?mzI7g$&vbK;{oc23dw#zsBkGS7fAX>k<EBo{bZpg&bJl;`zt%LzR`vS39j2yJ
zbsbCnQ>5Ltzr1o=izKsM-#+1~Wi5M;Vz%nbeQ$gFy&I=9Ta~{*Z)#XH4pharm^J>9
zc~jRh+m-eMY*kaA!<OZj+xlE~|5kNBK$czE^%dE3`woif$8|jyBI3u#Klx&b0~PBm
zW2vgL@59Qr7j%DGRlcC#M^p7)0PK!_Y+uW{EG!4v{WibY<LAquVO{SHn({oPs`q)k
z+6_97MV(L9_lS8rLHn((`5vva^1eA)iy3L}8h_U$*&*2~Q8ivVUM8O2J=r)JFL2~<
zA|p{IQ8u2|HJK=PBwQ_?9#3!BqDr@9VnC028!DyS$^m|Z00@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=9
z00@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*;>M&Oc3kA70-*#a@YYFIVk4hVn%2!H?xfB*=900@8p2!H?xfB*=900@8p
z2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?xfB*=900@8p2!H?x
zfB*=900@8p2!Md$1bWx_+reaqWUEBgc$xT#@$~M=#>serBYzVai86_@@wBeVM7bm3
zYVq`Vdczh~x+N0>deqxcDLrk;ifeM3*WTl-dU$Ixx6EAU&i5Pk__Ozj|Is3|$JFUl
zcY@RUie8`0>Xpep{k~z#ZKviQO}F>*GA}Huxz_2KcsgtN%?q6kA3r+&<myXVosV<6
zG}yg>mAUei74OfO%<lhn=E#di&TuF{=}C|O(fFY&zGbB6uNd>i+i5!-QnE|hrTGh!
zpDMr0A*Fo04yxI2%)oXV9XI{+hn9<8dvmjM-3w25e}2_Ir`yVz<<IT#hjYshi$46h
z$@-)E%$_krs<+y3G<@}@hOIAJx!QT*=^_Pl#%w+6Pur_XUVU|+olc?hRkqK$aqm&R
zvQ1ja=esR+cx_y(%C5(U{^pD*|MdkQb@|gd?dsm0ie>F}rd?g(yM@IL9QA9`&sld@
z9<|${cBA^c_0!IcD?B=|-|vi+I5oJu$2G+dW*=?;-UD{+&q=I2+HO@FmE2XV+Ise8
z|7JJ5I5vx&`p6$Il{$I%QKjRg?+<CxWb~%(?9$|=t(z=u?POm0X}Jdr&S0dJ|Gat4
z8vgv^Z;VgB_1sh1_djkIJ1cAahj-4+W^=M`SX8;hTK4?htkT=+Z)9!Tw69-(#kQlz
zsa)z;)PJcSyi)%mf2bX(ew}W6Wz5xEa~Sm(YQL{PYB_)7`MaGv&gl1j!Hk{Gj_PIi
zUOAvpjE)me>Mt~&C_RNqX*^N?r}C*DG|ouLU&_C(O^+LH%v|YIZ&4)e+B^3-)UI@#
z_*4I&aY#yBXq-}iBCWBiNweKwA9AQ)ZhEeDsk}$F9j!lcB>OZ@%l>>r#XFzbesnx^
z?D*(yFO<k}e%W39@dv8qIayD=Fs<xMn~u&CWS7Pjai(#Y_R*sE%kS8Bbo|ozpmgL1
zDb<IRIFM5RAkNedbewoon9>t}3g?xt(E7n8^Vnmx&uen=)w3CuNBM~_g~=Y}+fwnz
zax=QFW)!A$ly82c-qjb~znOjZcI#>19>_g9{%CyAc%b$aDfvrUutAS~i{9GBNXZT<
zjXN5jG~UP`-am@xmF82L7pY%Sd(*h1@lEppjb|DU)IJn`;)#+o4}CR;5idHv|NN(`
zOszlAA<m>Uzwt`rhw_u&(|^MWKeXQOkW#yol3h~j_oOubNNIeKQokdm`GJ(|@JjQ<
z_qS#I<lI**MoQ_Z-S~DOzI^{6p46Vill&%o6i@w!w?kp_pT;FAjZ0o>ywiM1{ek9F
z>OZ7(J{Bq8zM?SA(;{^{kK(x51ZW&n`_p-c`q|#U*Qw|1y=Lv7{kAxy6~^bx9b5I#
z(e)RlA2NDF>D#CM?oh2O%t-&_?DnN&)DP)=M)L{HPt?CC+_2QALuRgC!y3J_wZIEC
zHymAG(YWZ=Yjva3{!EX}duj7m4=gAcd%e_!U;WakXpH**tahERd1igF7}+JId4$Fd
zDful@THlD2{PL?bf6#m%RO!4gQkwUuJZf)JzTPx{lJdv-FtsNs&xzW_t5SawDb0VR
ze0k(IDb<sd;z{}X=nB($gU;i8f2Va1oi}LRK}zc#%1`H~cPb42<l-`m8J*wgyhdRv
zm(~&dI*HQp<<fee??==QPQ{B?9EdMJdL2T3{ZZ<?wS7)q$H+hOpUR<nP&<<Sksq{7
zOBVa-==CwJBPbosPo#AI6)ByUMN0mY5+_pfpOn@Gq@r>up6rm)I*C`Buc*GHR8LZh
zC#CU1{g<~(>uyr&|D;q8Qp!*5PWePi=LcTtx`<br56EviKTtf?llm9!AH2PB^0}=(
zTkg<&LrUp#Hq_bLyUk*U=0hs?{pYvFzx{ldL+dS?7peTM2S&E(GhvTI*JZ?!*5Ra-
zp8TcrCiP1S6Ib$^`T;4Oe~CA7rS_-sH)76&``#>8Fh<vNR1fl#<~33}4-pS4hd7b{
zR6iOIlwQ<7DIK*7aV0;=f4aV&xbgB<=k4Edbl-ydQ~jYU-g>3Vo}=-!-$fi~-lOZO
zFH4U)^MNLt*vma~X1(yh_M_K<BN~*s?6c|n7%ADM^#fgR(z=5DBObhxe}1+4>yIAS
zp#N@mdzq`6SATQs(ft?VK=Ulkd&HNN=0TdLNhv>FUy78tk<xf2rTI#hQvVexU0+hW
z(ses2`AK22vv%aCPt-bbz|s9M;zIq4<~eE~+8-m1w7w(WRDV+H7u3(m4qcB?JdG2Y
zKj?f!<D13}%^ws`<BH}#3e$W=VH!tdhvo@hX*`qCyh!{=$qp&`P3vw_n)gVl-lTLL
zPfF#I^6S$|+v~S%S*T=;l;$Ve@8gyDl0EX5l>8*6{W?-wuaolQm+B=_suwAhOG@RE
z_B-#>VQsG2$4IF^Qvc%TH=2K_Us67rALuxZYf48-eo*^ebKu;IW_I6kbbp)tBF?n$
zP4TqfP3JM<MEt1Ul#be+_RVSEh0ZTzpZ4d;j!3D0()g$I5bbNy_$0l2!^n{{s?R@q
zUxemqnh$8))4WCeX}r^XPD=4Ke^NY+6Y4K?oX#^e4^TNYUr>HJPV)fu<BT~?w*Iks
zwnIws)V{=n<_GHc6#jqg-FLiQRkc3+1VWYGMFA;FlP;j3tZh(GA_TFbaKQi)kq$~x
zkX|ne7$Ay(2vSAqN|6;oiWH?v?~nutkU|K7K;EC9H{aZQhfdbo`|MNp$^0izW*_sJ
zV~#myS^MC}yf+^BvCb+dT{-QEkBE1?@GBznYi@si-LbcwJbQzD+SeYx*t*Kk;@_}`
z#!vc+KYZiFFMfV>0i9=Q2ihmZd1ZfM-tGPJO=rEi?v!%>@6R{oqT}B!mw9FLM{nN!
zotfctE`4o|5089rre7exd8Qug=@(XCZS49#nl|&kgx^8mp#2PV&Jgv5-e?E^<b?VS
z<%j&MD%75RpRjR?NFI7dAIuL(-;JYjh1O&91~2$f551sI=A(5sqWa`h&O8uiA0WBN
ztsZjXk>8>J+M(z8Q%-&Aw|+zQG>hy?3bmg?e5X*qQiwN557{IAgY0H^Aw9L;7zchE
zIpxF4)%Se%fMEyTS{QHTq?6w`$`?_(`G600mY(wCwZqQCSg-R?Xn&|)<s!;oh5QQs
z2K&iwv3u;I{B<#<U+mN)^IZ7SPKQmJx&JZ$=?VQdZscTtZ`<w28=e?CwPdI9YQBBz
z&MlYt^QCXjOxGXf$N}|>+|izsi`|#cJdsX+*fI7;*mwy`r$6*gJrVK29!sZZ^qpK0
z`Qy&R*&ljLE@&LA<K(k1iCCp~J(X@e>9z6UKTer(PXXyUy)+Muv-<cO<}1INzR*YI
z@M>Hjc_6!CJl#L?3+XRCH(%%h`vTPijid7Vr+xL2Te|wi(*x%g_&s{wwLiM{^;2f<
zd+>udewX~>-#g`iO}@DIxH5%b|Kfl@&-=CUC4WwN$S&}YAijk0Xq?y)`P3i7?1S>`
zi*=u$gJ(P%XY<-T7f&yYhk2+Se4B6bsZZE=+m{IA-+7AuN;mH81N)&L+9i*ElZ$;|
zkI185dZT^wMtu-p@++@C@%l|Z^O)TFOD_EKOZWryMmcDF_z~LEU-Hrid^mqHu6Uy#
z_DR-la_X0SQ)bvB@~fX-=m&Z2huB|!s(dikuRJ-mr~l3cgvn*UMqlU=UgJHK@xzby
zl&3f1*&AW}(+B#?z8Xj8yW-_z@8}nQ!2Eaq%if#+@~QvH?SHoMf~Wky_knBu_YHgO
zGN9oc!@C>f>->XXX@9RD_7C)l9+3y1P`^|8***4(U!y<b)i11F?ciOQU6v0@$G3XP
zp?!G#iH98e+(PfoJeM`@^1pN1qYEGYtpTI>Q<weiuaAx2XX4ELw0y>uTzJuraU+lM
zC$IeUK!5Q{Z(~0BQ}O08r0?dnu<?iVQkXrX_rmOj`HDYw!MtT3)GIxORsI(_Qpiu?
zN2ZY6>~KVOtE(966~+g|XS<Le-&07>=_914`f1$pAk4oqkKB{9&(=479lJh#iIu;<
z-W79>GGELy>p4BwUX+7=8s|ly*<|oSyUo!sPo<j=P<mZ7KOp_9LUQ9td3tW1LHYC>
zT7TUm@C)@vKk1)!*}AHJ^+EZRtD3*YhrNff968CwE<)o-zwn}d^0HUjp_g^h`0zLQ
zb<UxUk9M?c|H@8D$CrLdS3iBH|N5z3>B_rb;Kw>IbsvIP^MM~o4*gTVdm8c93+O(D
zyyS!GsX}r<^+k;J%)Rw7hwk^WIU3EPdCGpUi}WWVxgft{n<dU(Yu5v28)e+-U0+B2
zhZ8T^<<6o1_4dqrhV?#h-Hy{Hy;Jgcg!wV@3-h1sU!{NSpErEujo|}FEqv`3zqtO{
z0i*a)aQqz0Z}q|w17@Z>*N|WM&R6bT>}MBDpLzdbf6mVHYs?>E>Gl`qIlIVy(Elkj
z_BHM)yNc#p3ajcTX9`pPjCg#cklje3d0d6;bF*mPRUvy>h0$Kr4X5TSJICMWf7@q3
z`xwX{QI4H)9%Y@yD}6Q2>?-|)_6gQC<zhYTzxwDs`z3$M&ar3uPe0U$FZ)K{kEqXj
z0r3Rwr}+)mBkNZjXX~8$$&XL#oc`e#50GDJ-Bu63hF_%r)?NM3U-p8Z$iA>2&O6)(
z%cp$or}=4pldd0l7v^7?H{Qvu_tL&A@9@^b4f-QpKVx}v%CG-;HU9WAF7zSZ^WZ<C
zdgviNrVql#QGV}cofjH^dLy5B=N96%r#v3XZ~j31K<lG-C42vQ+tXhkH?{D6z}|c9
zdFM0x4QzNv;(XmXjQ1(-hn?4n-{IaZezonUbIm-bQC>Pb4!wU+AODQqHGbrEKBivp
zEY&A{?z8vV;)rwJo%x-U_O+*eVRGxgFtnd=?xY;K@jy=Rd3@*h#rOXD%$n;@FE%>o
zfhqScKcE5SCqFsyk9Vlw@~wCB@RLS=V$O#C^2_x@{oaW{_4y9c_ZYu=cjMm<d*|(9
z>^a-5ch<KDHNM(7^^qH=&)LAUcGRPN?P=dR9iD#mhpiv|$`P~8?6+~!pB<LnVEFJG
zrx(gUx#0PG4OnVW126K$c4E124}Ev=h2sV^v?q)Q^%#G4f?cuiq<8F~F#VKY|J18I
zyNL&OhMi>}*>Ure9f<RnorUZ<J7)g}ou9Ht&PCY?$e!EpkxM*(L%+-ecAdRtuk~Ab
z>x6XgET!WQzxu-t$9`(hxR8h3?4k9|I;g)fUjM8=?5zG-FUcXlern&i&<o=nc_)wl
z;Rkw$XMW>DIbrQ6hgbVze6m;CBR?MKoqk(i$bnbuH@^iR<iV4ElS`PMJLlqe(sO#R
ze)h(E!?$rHr}0*g{OU8V`ay2}7O!4m`UTYowFmhlc#X(^FmBd&XkVva+95yxRR7o$
z@%kmdd4(VPpr7j3zA#>lxBU25FMYG$(+(ck9peu1MSgmM7y4pd)h_<bOYImR@jpG~
z;6XpRXo1Ei2d)0fI~y<35U+pEajk#!9$J^IGxjOQ2hVTa`_9C}j#{X(<vJ_xdH*8w
zH`Eu=y%j$Ah0uI}&WG`8d?5Y6hxH5UCw`3!y^wC4^+$d9H%|B?uP}L(qld~FpV$t0
ztS9QDpX!NnX~#HehklToT*i&u@*79>&~JP}a*|(IJM4t=kRH32WOw7bWgfA|#)sY4
zfAhdR(GEHFOFfVrcx5-#$398tN0CE4P<vCtyB%TQ``F*u4;xQ>(_i~z`x|`5ai*{M
zS57<nZ`|pf_Ts+JeoVW{kzZIl>aov;=4;IFy^MZB{+M>juRi^-F49}&t&7^Fm-chw
z)k7|*J?A*a-*+zhNx%6m;*Gy@#!LIgjoj>={un=ctzPm;=XdCbc=dYc<eU}X<P%no
ze2`r9Vy?UX`Q_tZe0%11YTA!>LYRG~*X)?~Q|F>NzHczV-m#DDTX!Qrg<nNq%wu8m
zNW60FvUT44u+EuJ>;yeGZ{&~jk9}2M`}CV!!qy-2!TLRA#=2mgVu#HS{)~OMc>OaT
z5o7y!vA*Dwf2n`$TpU++K|k?=2XcvL=kUXh2+OCu{CE-1&KZ{|mvG;hC0UAgF<||6
zUQ6%j54{l2KcH9iNq+h*9ZEM(`FrB&z4E>vjL6QIm*z9Sfd9x2Kyt{(&eCi3IS*5h
zb2;aH=D&WJ|K9g0j~Dg+dhN5<o%7MT8{`$wUm=hB#j8g=p2V|b`e%P<T-{sAZyd>u
zPkxqm*+X)suu2d4ecHzxx!D`%ZO(((Z}v<-tQX$>$;WS2zkbqd_89-pgXkN*aL$NN
zVR|aRbq$ZszxZd;$<3cr5C2tp{=NR%m$A>}bB@PfR-bnLeuQ}W=?}dyuhnZjoEMrG
z<mXq~CwPycAI1@{`s1DE>?8O8>x*;EKPqC>pQvZ{eZ7TIK4E%>H+->c;_0Drq(A%w
z`sm(5`EM@$@?B#-_`#XyL*kwP2(vfPd^UgiGyF2^zWV^}(~lI=6KKBCCwl803jNdX
zi1bBy^TquDJ4fCrVPDUlNM{%DEkAw27nC287m^F=Ukb@DUA<7f<W;Y9<?uwmAU@b5
z{ZpTG`X!xST3@W++9#KH6y~>g6z0Ei@s5c7U>72qC+ui{LH+vf)7Q-TxwQv2=%IN>
z-^3di<H#P;7wcmjfA(SB8NcIeT^3KzBU%sHAMG1&>GaI_lUKTO<_mk`Tri@1^p3q@
zm-Zij!-_w@bFPMaQ}?>^vqSi!uRDEt<m3<ha{h*}`&V{Id3sOZ)hA!%$#~F9^%zI}
zXQzbqOMi`r^LX_{bRUgpe5ePi51Nn8jjX%oCH)che$ae1FTE3l;=Ly{|J^^}Oa0=>
z6YUoLB}YX1h;MxIH~4@2Zu*LU<4Z2>N3<^S5A{p?&gYD?ewmN-B_bZkp+5al-ngkB
znpcp1u;1n*TzAI2#}lNtkR8SgIgP(@P%r({5Az?7#xbIH@ZuaoKh$TxrQht9{@{oF
z+K)(H@A&8)x%3O~;-UVKi~Pn*I(;#o;>kk~p>*w#1L~LlY7g?;wP#%MBA+lG)k|O0
zAH#SfuYCHW9w<MQk6iMRBZb<D^3yk{9r@_3{z7sXH~mpwSiZ<7J&gTOUOVJakM{7Y
zzrs*|p?G{ja%oq4(xG&6L|*hmJKBYDp2-LCANyl|kV8J@@TeUq9m?l^aLNpS-Fi%Z
z_Cq=C3)|m7>tYJoIpc|M>l;0hZr(!nIMyS73fWn9$^2&@jHmSlzw8w`t#8KLIg|JJ
zzC$Iadi4vh($$MUeh_<2KJp9WLzw@sf93`GAvwqo=@Ge%uXgEydhwuMdSpJ}k9_h$
z>HJ6i(VlsSKXQ?uy!fQ&_Or^X&pt|f^jtsHC+zzPdLqnUv(C{&-+u^ech?OT{Nu0R
zn{AYI-)S83^}{xK?C9A?`922fpM0Nw^Wzt-F@E4EVREQfI{ouKiTvc{UrF~p3ch?t
zGUvxmeQ24p2aGbl%JVyYPaV-dN&oq+P&&D^E1jO;O}ci4jfZmhP`~o*3H^!rr0b{h
z^b=q7Qo8<$S6=>z<i#WVqCCAej`C{<>PNF^o~6+KnBJKO><B)|seSXu{9|{`FZJoC
z`RM)9#N(Db<;r0THLQaX*+2Es4@gdEorL6s>{ncuwCnuVc|5zUpD|1>`)c_n9x~vT
zFD&(Lv32>mpUivlfQI$bepvgfZhQSsH($MQ!#>J>oqe?q7&q_1oe%omIr}g9m)(2H
z9G6`-$IN!g>0V^ir|w_!Yr8Jq5D%?m-k*7|?K}Wa_-1dlryl)PAGz$q@dUkVkNK6;
zUx){MLF*?ZFMUWMULso0BBtsQALkRj@opu~C-)5Y8RnV$XX`U`KcZgq&V7P=6ZO(3
z?~lw&`LxUKk<Yw!j=?TD&w%#p>@<62-+tDdmmGij)`Ld*F2s3|^CY|*SKl{ycWi#q
zAK{oz57Z;S@l!8-hV;rj5>Fq=OV53upq%?ocH>9WKlal@{y1>v`^}gxpLpl`<kmm(
zvm5&7zKA{ZUYfppui;)mee}ot)L;EHpXE27@$9~lT<S4T-G``8zoaXNH|=ZBeXM#`
zT==PhZ=Cx7?Bfi?1G_;U^~1;`yRJR=e)7Bj(tqzf@XH^R5C8a4uX{CmPVev(?I!(~
zF28lp_fy_|`tBskZ+)Owc)^$TkDPu_SpR*eV;;&U-MrU7>#uPTZ(Q+44)%$h?4<oC
zz11$e!#^`F;@KxW;75AwFZ*Y_d?#f*_@ngG{AXX)N1xeQeETkFrXJ7udw%4hKX}8B
z@r-tN%8Y%uc{Edw8Q)LQ4|4J+ya&b$zYKqP=ZBdW+Gpp<ZJofMFgs#=_{H`^{6_t?
zK3KP`L%xT#Uc~x@tt;x`Z`<$EPky(2_6zFgcgwH8>Ji3=a_+_H3qI+y{Vv|rN6r-D
zNq+n~-^3GpBwoA5K|DS2e$Dt>->ujB@At^bC!Sn*#H(>O|M2O(F&^j_{`5=zc#8VL
z?}64a{gj{H7#I1quYU6mALKNC{3HF>AMY>pLppx(3gt5{{5D~7@&Ck=OL_Cdcss8!
zuk;JbuRMA1Wt`|UJ)u|ll&&51L2}Z6yc!?v!W2)|0r~i!`pM3yUwi7qoASnyJtmL!
zNIv!AMc8^kt`u75lw*I@M_%<=cVl|YM?UfTYrTo-{5j)7o@m$kgVtYqV%>@TB&Yi5
z3B4tUcMR5B>mB`o+BGiPn=)fvQ=fH8IXoLj--AK?@*C{)^xr!)VRC3+{p7}ncz&4q
z&o9w$_nz`Wys>lEVeQ9$;@kYAPk3j~q|*oe=D+b<*lY7cf7HY7$?v>@eQ=IOuk?#Q
z%8tut-LO8e8_vD|Ysi^<?KRuGrTriKX&m$ivg7O?{Z%jf<h;?iIj>~Tlvf`<^%I}Y
z;nb&J*0*c!+To<HEjPVXKR<%}>cxZpSwF0o<e?u>zwr&tH)#CWIpyh(e31U7>Q_#_
zy2viE)Am91L^?U~WFANRrW`$G-|=c)hU|lW>K}f!CtiKzF^}0n`OO#eN;~Y9_MmZs
z>WSlmXZ<6Gde}ep;LmzUUi~vp_Ql%MU;e*#r_8Xk?4b6@8~aTT`C|R}R9-vskss=B
z6_P_gjE{QgnXqw!c)~L|t^dO6=a<EP$fsY%3Gd?BA^QaV=hw++-rAqZC!QVR$MK7`
zix+l9IF`3xRUe+zLw@tfI9qSnQR8lXWyi$ZH;|WJ*`Kkm)*<^={il!dJVAZ(S#QW^
zT-j50nq2DTk6736i5K$H7krxU#*N&5S3;QIWgN+;KIbOn!J~5ei6?TA6B;k$BtN}l
zAJ|EHXMC)e;_dtB4L?9X@Sr|&v47-Nzw%MP*eUIhpWa!|t<&;Zr^!iw@uA;%AYYW<
zyrvh?Zm5@Cu>SDp=mEPxUj7@slHa-}pK*7dV!fkxc&5+x2lR)Ym>=YgeyH~Nm*y|I
z^jkjl@jso5%C9{6h4q^r>kr<<L;cW>{3+DF`m~21e!O&Y;Zr}Q)1!4~yuULJ<dWa`
z;vbUJ`MP?v3#BWk|HhB}5!n%P;hSI0zL{_QX7WmRe{KHoOVwk&FrWCR>@@$M-|JmC
ze^`AH)h{1EnSU(I{=+9ucw_Jq4^N%>_an$*eKF6?Z}pJNIgk3Z2dxMAQNR4wX?8<<
zc+#%(#h6Y`^+D|!Ct-3!?<kBL9-a58j~vFIeB^=F2lkJhh4eyy*=@WSKmFA|`N<Q<
zpB*MY{V`tp5A|0VsvnX=x^~G)UyZYVK)jHPzWDxLIsEE}_Vpj{>Q$exe99Rg@%U4}
z{^%dX3;obv>B`YtVf{4D^nl&dKlP|rJh}DL{K2zw+LJ%>pg#S?gZ#qs;YFB#qCM}I
zgynOtpdEV6-a_jnWXB<SAiECPS9sCs!?xUV#Gp}7y-<Ik@qpTe_=D^|G>%X?$e)4s
zF)*$(sj&L#t9=%^=mmYZZWsrCpM3l}<B&q{AW~@m!k^}Mi+7)B9p{JfGsMepAL0I0
zy!x9(zZb^8_dTQYLw<mL3$)&o-~L>F^@#VrfFHsC(I3CtE&bYyKls%@t+;r@{V=~m
zedOT(k>~s?UODkkYb`kQduI9EW9p}Nq5i5**m;C7zM=D3_{GZ~UF;(_&e_lp_2I=l
z<FA=l#@RV6|Cb)v|LH$H5?0Q<a-PheGau!rC%$94{=V%Nop|knjoo&feaAPCnXBQu
z@Mh7tL*G?JbT16$kBH}p#v2-^i1>(T+#x$0?VkOV^BCh1=PAFPzb?$LHty`4dX2mM
z_BHA;uJp*bkwd=NPHay;`(FLA&$GX@k8=)cpDo@wg7qZwD&Bf#KTrSGow0w>zImb^
z>lpcsi~87kc3e5%xml;Bdxy!tP+mL6!}*wa^HP7U3+l5T$VWcopdR{ZU9x_uPdWYd
z&O*ENk>A1}Rjvv@x$w24N1ixeV}S{~AAJ1>7H;_7kG+N7$2@!TggdU>bKZt`tL%cm
z>&G5=C$`B)mjCjR2hQ2Ri}x+;4!woq^-DZImYp=N_!7pGdf0dNL4NXBpCG$RAK4${
z#y+z5^om^CH*d919{o3d#@~KM{qk9tyg&0kjJ)J@PC||p(qrfx3z7q}3y{BLUu!<`
z*ZJf0mtLD!+JWkeXnxpV^5@kD&0Ash2$N5Jkli3>M1Ht@Rpq3^C<l8Jk(|mw`N(g7
z0Pz9Me<+`MW8Wo=*VvBs%|HGxex&n<@nk%mqsp&;^h18*Cd^Jb7m%)=xW3p&Y1e+h
zIV}I2{grM!$xkooyZqWy4j;yY-10}>jlX#PiF_MB^GP~8?)<{NvUuOo7-#iJ*H7!H
zad0j`KK*iD#%?*MHs0)-diCG9@elX~?3DFW{rKX4vUkp(+1uy`C`Z4HAA5se`?{!a
z{0{9ZXFS<s{No32coCLPf3%M$ew6;uQ>cHC-v-%R?UGM^<3MkHH$#5oN<RHnj(q$p
z>jiz8GUNVKynUAOwBC}(dSc#Kx8*l}(E61^JV15;AM{QA<i@Y@!3#O*fw1wX$9Th|
zbrt{YA%8%7=9T`D3m@Xy6Y^<ay!z;o@gtw{&|h*HkBH=`Li<zw^1d72>NU>fqJMbR
zj(kvle30L`ipP`vj&k-1>gRv4AM%?w+U4)zO+U0le&bFq_2_?`58AIn{+sg*exUi`
z9!S5)OU^2!$M$LZW!<4S^vb?;;#!MO`P4as8c_eqfp2<fpC=vS8DH#g%r753p||RZ
z@$yTTUq7Tn^M@YEXMBb6qhI({4xiHHR}aK1jPdkVdGbZnE|g!t)hk{;az=fnKjyJ<
zu^%;0=rj8&pK^HBo_GJ^$tT@BX4izRH~7a39^}(c`Gu7euRMKaKlvlpVfD%v$4NbS
zB@h2anE#_5ey8&Ig!~-k$OZY~zB{&_@VEFi>M@U%kNtJu!`~Gqr*ix&-x-=u><~S|
zGybg4=4aJ8Hoe53b4hy1{?R*Oeu#ZK{j`6E^cb?Q^8231zFazVexg3-&d#~@*Ez_A
z#XM(?zjd~m-?3RI-PcFdf9H(OaqUa^ujFTE@Wmb(ALYz5^AnOMBE4=FjdQz@oq%{Z
z{?0X=>(T@B$+*%B<>;&anU9eE;uT-)mw92m!n5_!JaDhzJl6OaH)y^>`HUkz*;%L`
z^0QO=&z|9rercB<FJ0Jt*FWv?SDbU<g&uoXjc56#laoKiemHNozcSyot37&(SN@y!
zjE`}n56ZjOA_pF{V_uoZ&e6m}@;Oh}A3SQ`_>qf0=(}+KTa@2?RbGAKA^!EpI-?xD
z(JthlS+C^NuQ)HXWB*E@$&EjLs(FAP^3Yd$2<aRC*meCPpZTdD#$UbWEi@nSqTl#b
zkAA41y!3;d;?<8QeujGKwR-7?demzi@yBn`Z{_E^?}3LuddLT6{w}I?^NoK+U)f3X
zNj>B+F6zO9{yR6q8#`eAalb9U-%WAOWIwAsdkFER-|UrfreExreE8+B;D5@D^9D$d
zh1nN+tl#nptDn6fhw}pSHI5s7HLmnbf0U=E_@wvr0-yHJ%Bzn*#10upVe^?Cryurr
z!p<egB}~8BYvV#rJVhSpyZLWD#JBMyKi<8kR36{-)!$pxu5rN+eP(Z>T*e8HkRQs<
zJ8$GCK=XpU<g%_9Px2W@<@5`W^ig^B;zv348c%3^*dO-IyozD*^h&z<XMXZ$<aaJ?
z|6rd%t`r*g6w*WUk>2S)`)?e@%TEv8Tbmc=o&4&>JG*6k@F~ph%12K1so%Vy*N{Hh
zC(v8_1?cxP{Jsr;guQm3$8JOVLf`pS=7D+Ty#aebzap9!_KoHRd%zAF4=CNf#CXta
zJVEn9d-M*{6Z=ib50*|(?Qh*bx$jZV`9VbT&<FO~dma4ZEh2uPc@516dIZhqX3>1r
zfAdnjuyL^t(%Wbs_@8*P&v#BO-h47IjGK1Vt9|Em?2`3be)06lIEoi$m)tYpPyfwV
zdP6^iwWl7a9_{F-`yGA`z0yDDFXV~&^dBlGteh}8@kK7<#~(0`<{5o8F6^T9UA-w(
zFMc3i$r1Tv7vzVr9d;Jqc!qdUjvnZzc>Y|f9r8pu>9u~ss4wO%`;I?hJdlTccoK&C
z0o6mVV|?tF_MrYj{lrVmC!KuC$%l9CiigT8Ctn=Lm@c1q<4oVBn@^D5t6x6$g8u={
z8z|m*>$iPR3hh_%VP313{j*+37tjB(-_c+4L-F)SKiN@o;9ov+M8vOtJGX`GJbj~o
z<iWczy9e1b^~G@@r~36ze*H<Id{9383TszBVe=>YU;H2Qncu-LlRx@9{37<$`ma9q
z2=fP=Hz{v_tvr8?-^U-|Z>Yz<fn59x^713?JEZd~jDz}&4?mDx+E=gg{4nwI=?A%-
zr|^sT-SnA10L??=6;U~;eHis#*t{{{p?V^kzj3~r2lSWT;Lmuo>(Km2Aw4FCa>mEH
zj^~K@i<q)w;_;9|ddwa}{l+U4Prvksy@c$D@nye+_5WAzZv5L}@4Q{;mv->YuHp}x
z_rmni{z5(KC9iWa^G!N_jD!60nb+<O@Ss2Pxeub}>eY^N_=&uzN4?6^2lL*1Fi+Wc
zcEh+t#Jm1!Up&OWeB_eu97#Gk$-@prWIxO!<Ag8$#<wsz*;VaDjQusg=%sxXIg}?4
zd!nCJ$j?^4_O(MF=qbCRUVfEw`b(~ec%Vo0KtAInpRjz^3t{ExC;Knn_{h(Hht7HU
z9qg%i{sVhUKa4m3o4z?mvfr}r6plR6KmL#Pl|OHM%r|`CAKE`b{IV<jCH*!J@k9>e
zX}_Qz`XH<x=OW}V&)G5kkj_5oukj%V)L;Eqz6zC3p?MFjm(Vy?A$@~UFYv^#!ME{n
zo~NJmUAy=;-pXr-Unia38%J_M<BC7?H0BqNcRaa&)UOzCpC09-2ljQwAIeXk^wT=P
zt~qC?U*esgYhQW(F}d*x)uTNA$f3ORMs}Fp6?VU%UHvrQ^xwD|Fa2hhv`-%WvHs{U
zUihteAqTl2{Zn3;9-4oUT<XWulo|I;+Q$PulrBtv%tQUQepsj2Pxe$k@+fbf$R|vn
z`48gpN*;(`>Dr^u`VZAFOfK_^{N@+CNlx}ydG(n`^pibMzk1kL{L)`&{<0s|7hyaL
z<Hvd@-MG+G^^jLseqsF-R===zjidU^fBYLC<;j63{iToekDiOCx8|dCegh<zd1oBV
zduUwbgZ7`$e#QROJs&y6d%q*ye6;_hm*m$k{v`h~h5DaDc7XnBkA6bq0mT~^;{@ps
zdj+*)U4i^f`ygTUMZ^R91I-Kk>bHJE^6L*fumAMZe4=-dzC!u~^$Ti`{#BuRV|(PK
zC;I9Bj6Rx2P`}6l<)^petvAkfq-z(?5%t%2Y2W%M-FhIr?u_qN*=<N)*kj{NKJ!s|
z?W#A1^$(AD(J%GjMLy#o4COOF)i2)ouvglPywL~s(o^4Ak`K~<<0DMpA$tSa2|SY1
zxrK7#^-DZ^$-dd2sMmV}??Txr{n2mj(_4rK`o<0$mpBgmDS9H^{4tNA`AiP$lJfW$
z#zREqBI1kvOQHR<b6@*Ozl-C%T|VnGy9lkf<a4g9ob}oK!<T)n_MA&W^7yXKcaO?R
zkNL&R@19ye?T6Lldpq&+LH$;5YJIQ{#eDL!2gZy2!3Q4Y<F~Wt><Av|5&wnXCm;Wh
z9<hVwDZREX(i`@HpQwJQf4*156LhW+_20a}C%)(ho`m&VKh0mrZ)V?=BOkrc&dm29
zGtNiZ4fEXl4gPKX{S9{2K7~CpudNUKWcETi@1yu}>S52+r#~rVH;f;-p?b_m<7%C<
zKa<YhS?~0pUK&UFv@2{qh^Spi-`Tf_c!KzY<c0JR8fR!+pm_izzw|z$cSRA^%f33F
zvyRYH{<U{=^qL(opN*UGrBC>S-XqEH9h3a@+B~Os#$8yt`ivhx6vlC)-_r45J{m9e
z887;$UHvg1)JNagLG$0ZvTO32N09w<-;HnUmGe`2tbgibC+WYiezIfurq}wdK4Ivc
z6uabouJ^juA?emVzeBD)c9|Ub^1TZG2VdS*(hEF5<?*b%-wz`{^m}XkNqnnshkLj9
z)wY|?)!+x1AM&yP^w+w}KV-*!Ulr}Sbn;lo&0FgbzmFW=M@#qjBaJV<jDvRY64T|=
zZ^#}&<swSg4``gVFP>ebZ&PNRf3auIuj6`7kNLaig?`a{?U-NmoqaOztq;QFwSM4J
zd3L}!8h`QZAU>3*m)05e$)_HuKI=ApV1MbWc>To#{iHAWW`~5y!EUi{>h-%h#)my%
zpX6f)r5hh|LgTA{))o4sJbNIYb7FRvJko{PLGxaiJmx9AvW{6t<hQ=ahj-(_&w}ic
z@rB~q1O3&meICEbI*wocHIDAP@M#>a3-nk!cvGKt@PiL_)%vMCKTJQ>Zyu75pCCW}
zw6DL;(dirgv3{Dr#?}0wr{<^mq+jZ@Zqp<3&`<iWUWj+&Wt{oh+NTG8M}u7!uYL01
z)p;bo>6I`$#$H($?VE&^Gye1pfAp9<##8<FCG4wy(Ia}ALOjto_S1NgPkDObJVd;D
z^xybEdP?t%W7JFfBwl~DqkeiwFX)YQJkneBiPukjiI?B_I;WxU()Gi9H-5$o#(qX#
z%{%V~%s>6MZn1~-5U=ce<Uzm8M|x}@Za>dXtB0MA^+~6f)+O^+{JM$X*HfSMlU(wX
zS36LCewy_rg?OUR#zi~o6K{OjLGx1mcr#AM!@Qsu?3nQ(zj2~pzULsPd?_TSae?%I
zrk*}|=Q6AA|8^mszOw`7hj9|dhkWFW^GLjT2-o@U{x6*TyMd$lul}wN|CK-P_oSpd
zpOMe+Pz&34@U#7{f%4m3@byWv-!R)KzrXXL&1T>J8y_7w%DEbSq3`@S@7?$hc%Xmu
z+PJdc?5*|1?@Q~Sb0z+}F#q3qC3LQc2jgO&?tIexV(0k_-W`%lz0f*iU&nu>-{iy-
z{bd*F9l7```eT2g9w;B4?K`4>O@939FZ-xp<^_GW-rHvwSMO2uUwQj!>!bYSg6su(
z$ge&74C8@U`hZ92#!LP5%D#yl>@VK%z+S0We)Sqp^^>1G($#N#lqUyNUOD4HZb*(O
zmw4rkxAX0opPcNVa^#_B`lo&Q$Y&i^zx?)H>SH&xi)a0&@A5(Q(gR_<s^7Sq7vzNa
zCztYY%H_kh9yfX3Mqfky)<1UBegmK8i}@4LxTw#%Yuyn~kMKlaAwJ2YUh_qH`Hhcp
zwyx7B>x1+-zT~10!qyA&K;wxQ^(rS!4)y4d^#qcW{ZZb!O)mB6FFDk2{l=59^8B$B
zlAj&*PLX`%$ESUt_Q*p%<IC@IU#)%fQ9OT`KV^O9hd_EIpL+C{-5|I6`EB%DfB5nI
za`oX`*#5>o)4tQb&3hF{KJr*+`I*)|@#Mz?Ka^h$`EC4D{Nq{Jyc1S0x%A&S;)mYx
zGv(t)*$=5#I=v;I{#n<JyL9!~kK2dQL;Nal|71PDvv}_Sp!o{fFZ#(|v*-NEi2B72
zIUi>q%@_I(>7RC@-SV!19W#%W$AfY7dsN0px_Y%oUg_eM(~fyS&*g{mvzOY5^PGL5
z59*`$(&?{y>4W}5?HDKR<3<0Bt8o!;KC#bGy8PM^FD$=)k{@rv`k}wlp?MLJ|8M_b
zeE2K)*I)ftuYQ}ycsC!6t9_P!>5ul*8}-w;Xh;3pr;p@QFT|&Q7{}N@@#HXW`r$rl
z%8Yvkywf9g*?E=mfW}3Cp>r%evFFBJIsJ4`roZeFJEouZWBTKF@9`-g`P5@Srak?|
zzj*seyqRClFO0MGMLzu^m+{0WyU32Q+xB1D$BTIVBB%9%zQ^@QKKnvwKHA^01LVQG
zeV#BmloxMbsJ!{6Uh|IsYFy|oImBy+zTv^Ti68yNAD;Oi)@%L?eV3pAK|lE$&P%Ox
z)@l07PjYTePWnzy<l{#<SJIAoN3ZFjF#kk(<)HOEo}c=?NbQjyALP;wKSfx(>eX-Z
z$<H1d7yO!Mc!&H8>#%zReB(p^w9DS;7x~Bw^;bUjY@>4?m~!v(17`l+A?NJ+ryPIC
z?@E$Wys+PybnnMc_|f!_{q&GO4s3Xz!EeF`{bUE+qe&-+->VVUAN^KNJ?20CGmrSQ
z=8Jhn4t9)QsGmNYU&^yT>>D(nj2k~qebU!EdH6}AKQU*+-yIQ8ukZk^JJ38Z-t@va
z^J|TV{QAQlvr~~Tb_Q?uV|XU7bYb$D2jcZh`{X9CczQ2C-XhYc6sq6v9yu2u`sGmz
z?RxGa4R-GK&n$cOL-R~4$SuEprg@4-{yn*&c`Zy{?V4ZomwpJ7hd&h2`d}Ti-Wo^i
zt#AshSL!i8#Pg$!KfTtTa^lHFf0R=%ec=!O{D%{6U3C6IqrC6o-`Lmi)6}nB`RS{4
z^;=KPe|`>m%wK*|lv{oJrylKFw~dSS-aUeR(($i9c(opCPdn@iw0@We<_A5YujUuK
zC7*iCL-paCU&3EAzx9j$yT`+qcs$~r9~t|r-}HmL#>KglcTmo~`TOjrd2XF&@9@FC
zv6pz`7n+aw#|L@LNAn$;@8qx!<JtL*cIYE{)DL~{5Bcf-J{8n1ls|=dfvJ4#E7UJM
zLiP^g6|!@!A~~Byd{klVKYfGr(mZAV%uD&qYiNAsGcV*9Rxe~%>;uG`7wotBr#(nM
z>FkWKd8i%pN<ZViM7z@MhsdE_>FTA&>;QYrzUr5Hgz=^y(0<2#j{Zm&ww~xW|I@s)
ze(0Bc?s?pUI`?+pt6g?WzvXB5$sueUl@rF3cJUw{(i7z&yRE$W$j>4#e&`)EpYgzs
z(pPp~`_du57P7niBJ<AQ)3Gn~JM-2d_1S;2m-Hd7OYAZGOwWwF^@SZb9`>#Di@jpc
z<!2A9w<#o_`k{G{LVQByB9b#AeTDe1i{wrry{e1G-Mu|IgpDV?(r>&v*NI3z_K7~y
z7j{HD+J*F*f2Tc29{Jau@m`1?nSc6c-azx;yfxn9p>(KU`ayrm1Ic0js*nAM@yh8p
zz0r>R_*cJvir0?(QEv6iXTHbzD4+gmN7y*f5A#}>ol}lH?1%oEH{=w?vw9+L>Qjy%
zreEZvAJz|cf<Dq)<)VIA2gI9K#z#Md>5+arFXCTYM}@7|_*M`5D&Bn2ALB2qzslps
zxCrABZ}^oC<N6cdl~}iwS8qge;YqspL&h5~=COLoBOkpme&Urgj>;Ja^Fvtu;*Fbl
zNM7SEY#vyzj01ZG)eFf7=@rBmH2<M~rqDhQs+Ybv593GTk6$I8-zOdK_LIW&1oAJT
z`izHlPnchbZ}H@2m+;R&(jSOlNZ;@%pYd0}e9(MT4$thGeUWoF`$zUgIr>2_)NfxY
zpZeG(`+fGz-zgThk8<wCz6j%2JU?;DjQcHoLhU;5B@a2RpZ2f%iw~$A`60c9@<Tj9
z?Ly-Y**Ef6A^EEiUn#_Uv&c_{>No!E1bxH{{vf@^i*pf3zaalznB3Z@4|u1i<T75`
zWv|Tp7|$+RC*)&4%)5x%iKyQZ&BrR#&xquXh@UDnPR0!yx4NiY6-Hj1pEZltxrpqx
zb&Q>fNY40ua1Lh%G6Mq!I7hNR*;jZ+?!7(#hQFnpc?gx0&${khU;W;dMgK`V%JH+j
zqt*^PC_npVe_)@&9`FnK|H`{3@P2}wCztlc+xO@ne_s9eJM6#nAM&|}<FDa^A7<Ur
z56BO&PO*pTvCff`{L-zP>@xq!J_{e}{pMeuUT3eb&)-m9IqR%^5v|MmE8TdKlYCG<
zej3ytNKYa8pnj&1T>26Fr5$=EOwWu*^b?Gmc=eLYI+Q}=-zxGmBl4%5H+l!-9S6Sw
zAN&sbU>xMbD}O;)K6=0IjB_pL;M(W!(?5PNe_#Lk<?1Cjf8V*Sb6V%n{8e%oZ}O;5
zKI5g|5!sIv>Q7zNPx3+jF1;qNaWSs)n>WS@Z}e2Zq4A2S-W1w@rZCEhUueJSJ}>g*
z{Q=ZpNKYcNGxR5gv0Mz(gA^LCi2IMfVa1={Iaecv^tZdvdSHB=kI*0YzV1Q%9dhIB
zotgJ#-WQpt-i`g_u>HrZwZPmB`mew4Rq$rK?em<^vKvtS&^{1q2Y&bIYv%mi+5;QV
zyCkSR$ll4X-iY=M;{9G7{jeTQnQ{JL9bmulZvABM@oQgcy@TetFg-VJ(0o&`eXer+
z4e}^w9<USS*DpM=1Jd<hJF!0X%8xhW?ERZ@GtcR%d7vJ4fnLZ5wL@-pS^eHC$%hyH
z*1vUUtV8tM`LuCUzju205mpag$l*NCyrVD1D<ZkQr^bWvQ;%}O<kK&99(o@pAJo5y
zanIvC`-|`W^_ex-pI+!E#IJR{3h7N1((j1ob;Q_@X2a%X3XMYwjaNi|Nfp*@m%g$S
z_7@S|KN=r;3h7I1mwu?vII(B+So`ATgLpE2))W2+q+iO>Z}OeC-VBW#vHyr+!#_4;
z(IJZs8F=i7ZAJ{4?O*>6UvT&$!xtYi;OG&<m;6_Fl_3L%4BT@2m5v!P{G8(lk6mG4
z)aM+|3}gl}1DS!$KxQB_kQvAfWCk(=nSsneEd~b8cHr?-K0T?-n0e^WdQfXJ<V=}?
z%s^%!Gmsg`3}gl}1DS!$KxQB_kQvAfWCk(=nSsneW*{?=8ORJ|1~LPgfy_W=ATy8|
z$P8o#G6R``%s^%!Gmsg`3}gl}1DS!$KxQB_kQvAfWCk(=nSsneW*{?=8ORJ|1~LPg
zfevTjp<<IqzBlaUvMx?}`nkn^Fz>jsSzPVjM_)hsnAgf?@f*+o>G^Bdd9AF%ozL8R
zm7knGzO0M)U;LrDS6ch^GUNO6(BbNx#m@|61~LPgfy_W=ATy8|$P8o#G6R``%s^%!
zGmsg`3}gl}1DS!$KxQB_kQvAfWCk(={g#2J7P;W(je+l#t>QfkAO8E%qsEub;{E3x
zd)aw!O)gu-emfw!W0`@>KxQB_kQvAfWCk(=nSsneW*{?=8ORJ|1~LPgfy_W=ATy8|
z$P8o#G6R``4rU<#o$CB|s(a(_RCn-{$)aZlG6R``%s^%!Gmsg`3}gl}1DS!$KxQB_
zkQvAfWCk(=nSr)2@X^;^+V|WKk1H}}9<m;^ZyvN|a^^}h1DS!$KxQB_kQvAfWCk(=
znSsneW*{?=8ORJ|1~LPgfy_W=ATy8|$P8o#G6R``%s^%!Gmsg`3}gl}1DS!$KxQB_
zkQvAfWCmtg2A(@_{5^;N^3k%paNLJx|IqI?m{Ml^zdX#c3YR;c8ORJ|1~LPgfy_W=
zATy8|$P8o#G6R``%s^%!Gmsg`3}gl}1DS!$KxQB_kQvAfWCk(=nSsnezhU5r0biJX
z!U9hgeFZPy{>MxI?dtJGf5(2KK)EBCfy_W=ATy8|$P8o#G6R``%s^%!Gmsg`3}gl}
z1DS!$KxQB_kQvAfWCk(=nSsneW*{?=8ORJ|1~LPgfy_W=ATy8|$P8o#G6R``%s^%!
zGmshR2Mp|X;BgD?u<N9<xAC;~W{^2z{}IE6e{9HNLlz%0@YoUCj2JT8zy2M*;P6F;
zFFs_z(IbX0`LFOQLk12RxaIaM9W!G1ImZtkyTZVJFciJpg<aPja>Abu9bdGHr~Tr-
zlMnsX<3+?5KfdUKcOH3bSr>bkh1_~(pl>iR&w(FadCYY$lo@-MhrXc=S&Ga+W*{@r
zbp}4U&Q})x>7E~GWb7L}7||cD`Ns#AnDAto!e5^;=**`UosdjVmCNz(&ufYw`<Q!1
z{$$NZb{SJZ@tuNu9dX5}cl`aivcKRyTitQscfLEW%-9|t*q@A<fy_W=ATy8|$PD~D
z0}meY*exG=`HiwuQI7_1`s^i-F8x@U!cOg9f0aAlo2)qd7mokMW21_m#&+*_V!!%Z
zm_C|!{RQ_}WTO}Nm}g>{an|C&dExio`_A59UgGT{b#BsK@#a&zvD4>ee>vBw>PM%K
zQ<lGfGSFT2uBYqWV5ysTe*Dve8WGw36b>Kx^>>fjVseo}=jJKwZa;RzyEZ@jFTa_y
zQH9pEh*j^Q+U*yo>g#X*aqdrZ@8bG9U+hbF&9}Pz?LX|-uO71PgJl)=w0+L)wRVr&
zTfFs8ZzuPvyBJ{q+KuhrFSlA>yLe0SQ{~@nu*;`cpZw2(jk?$_dHZTTy=50$ZMU!T
znB~e0w1a_G?P_<GYe%1Q9o^4Bca49l-tO;F-)_PFG@cXH-CyQ>?|<%*(!1V@@7vm#
zrOOOt1~LPgfy_W=ATy8|=&uYo_jJx#_Z&0lYwp}P^Y#9GeP*YLLsvR^Y?(3h@W1td
zed_DjuJ1;>d!6ZRI~3m`?Y8qC&oA@bdqrw|txGV<@7xTg>gj5HcaKX|{q54PPOHZ|
zzNj}j{GWdptxAt_dS)OqkQta&8EEyLNKcn*_xjeV|1+!VN$z)MAT!W;48->nDLdNk
z_Y&3z^$KU~toQX*<JEaHB1@bZ$P8o#G6R``%s^%!Gmsg`49uzwRPAs2dU|iY*U#;C
z_U}}7*Z9q<cBHD`?i;+<@IE7j@tn>%+N_HA?f1#uZ<{o4qrde(zwgWM`~LU)zN`=L
zuMg2K3img=ko)ug->+tVx`!X{3w!I{u&d+B9z^VFz5T8HeWR{i_nDLCZftnyJr7?v
z_dGLkx7kOYwA7+=H>3+g`K80NHy-=ry^j2!dIoKF$k?q%&eb^UxM7<fzsy_>;bt-Q
zdzLx=|NgzK-r}Qb-;>ib1D(J?f3xqoKkx4!IgjDzWWL^?ugp*8=guQ<EnC%tsx_e1
zdi!g+D!y`hW}p)ou)apL{#K#=X6k!a_bJl<4QK3ctJ<kb*G>vk_c(neKAwjNr}`7e
zt17I%6xyHm6+Xz3lDEIaJI{;jr2a{Vb6xoHC2v0e)Rx!z*iN%ya`ZQBmWP<X6Ks8!
z@qZZT>+goLd_DUe&i`me{$D3CV4Z}~F1I_J=Vi~%%UPB0c@F&W%44p3q0E?hsG|qn
zZL_QP->GzVxh}TbUg!E|t#k4I`abX1XAC;?sYNH08M~K<?zW@3{>(sTATy8|$P8o#
zG6R``%s^jYz<q1Py8HJ2o-dyx%v$FNdA##@xALB{FWB-dLuR127;q2Z-CDc+f^3i9
z-yZkYc=xq-yX$vT>ek!W_~>qV>hjW8^07~?V!P~_dRmpEr_24@yY0I+Kl?AgnX{4O
z-=7ETV5eiV>u;y`H_QM2fB$gsrq5pT=+cjseI55Y;)+x6`1^BZU2NBS$-culfBHIa
zDf#kywEQ0JKfOomw7=6w-mdzk(f{e~b)%>K!k%ic>if1<)8qK{m2g*iZC5|KDu?pz
zT34&;F|Yd?R>|E{>0PyV#uL`HchUdp>2S)Q>9lyfcNII;kE<H5S@nC7-0%MNT~b%=
zbS|HN|NTQRzftzK9`qGItf$typ7P^bZKv*d^tRr__Up>M$08fOxW_yb%PNfZcQxEo
z`lTJ?-&MJD`Tn&|r1k|HEOqnFkAHelquFzkyVu?Sh7bJmxx#ru>O8yI{ajD`tG)~D
z>i2W{+0*jp_WrYf&kknH3}gl}1DS!$KxQB_kQvAf%!&+D-H$jwNTK}=v~TKe)Lz8y
zYB#nc++FSEdf)%=#$|qbm!DbDR^~qcI|HeC$Ue3kQ|)y%zS-~CyV{=eT|Hk+m9IMg
z)8E~8-sAaYo_nu|IDF*S-#u!J$wg0N)EB>B?)SrEdXzh+=kWV~2S^VhI=6@J34~K<
zoa$n`$H%<tH2inV+P7oq1HCccc(1ZQT}^Md-t?9F`u3=2=`sWDWnkvt0}6k)HpaKN
z)m&SrF_5y))}Lmv)B2y~%?$J;11bN$ZoIz_)>Z7uKF^A6b=42=uJZO1buqO*#Q0dB
z^zQ0MuJ=E`!^*t$EHAU7ZOeVm3}gl}1DS!$KxQB_kQvAf)Mdaqhy7s{HhccqY<}`|
zDpv8A)8GG|zNh%1Up>`cZl`aIdv}dPU3=Tx`RIM8Vt0*SuD5^B&%Qdoy=@P2`<a1O
z80f3>Da)1Zcq@OOsH=LD@~2|G-f_qHP6>Mt+f(vt55M_7wP){ByPNM$`<*54zEha`
zezb0Us@_>WKBXtk`Yx$@a(w^zy{~4^n|msM*027dU;T9)o3*38ou6OW+t}MY<n}WI
z|C@o+)|(L;vHyr+!#_4;@gYkL8F=i7ZAJ{4?O*>6UvT&$!xtYi;OG&<m;6_Fl_3L%
z4BT@2m5v!P{G8(lk6mHl|E{Yo|G)OU<;Gkx{>>78>EPSXoxaZNWvAe`HXHnjp`V;s
zrttAASG#?gJzgqX#Q~QeF?9C(&M&KQ@ZdGCSnbU#i@Ip=PTTydSMFMW;RgK8^xqC!
z=kP@vDSY|NZ~S1?*$XwQaIW3XedzeD=5N$R^0W&F9eeQ-s~z)7*($b6j&{|vz^9L2
zZIAiJmCa(R{Z@HVE>&;4<2PRUytj`2?fA00uwC-LxbV{ZG`1a6r0|4ci~nJtKfhjd
z71x;KlhYQT<CP+X?=AM|UswIh#G(qnfAdKX9<b@1MOX2r-A>%^<RARKNa3U%rd~dJ
zqlrZowrd>D+JD^86E`1Sb~+w6?;5MjvGI##yK&@Ehh6&Y_LGzARcalJ@f$8Q_rV8!
z^tH05uwDAm?s`sqb)_*o&pub9E}r_m58uD?%Fh>7$j;Wq-(Ga%i&x%#Wf}3T$q&AE
z!<u)NRru=E8^+x9$GIC-Sl7R}?e-%E?0V3HMP2;V#w-7H|NkzFh;`dr{D8mTbkyKM
zjVk=oN)K=N^cmv{Sa*Ki`12D_{OV(4%DUKYJ&X0-G}oUm{_LKI7oCO|u6p-Qf4^sR
z+0%H%Q<omF@K?u_t>WAF{cW}}(_bz#wwH$uHy-eZ<M(~Kykeg7f4BU(kIdXBt#Qmp
z&fIL=Q|0<6t$fq8uiak`y7Js18$IyEOk981TE~x@``HrS_P4KWa>wR(mvHd-r*FRd
zt*1+PYB}GhzwnnAOSr=ZzklSd<)1I%i@VPI_>PlaDj|P!sR_e3x%|{|C6r!;TWtC8
z(DSC<TEgSj|LFFMANS8vzu*>kzjpb)i$7Pw%a1Ll-+AjZWyGj=LstCus@I)2wv1TU
zKH)p!8jlsX&+(^6i&pWi^B=wZgggIIMqGZy;iv!bi}#f&th-Lgx9Ue9+WN?!{Gv$V
z>z`kE^fqtZQZ$Qao%6y`XO4Kdg!H2d_xtr81AemR3uOwU{cd+S^4+On^Eu+@-dz6H
zsi!R3c<YjDci#QL4>k_}<mVq){*XBvBZiDT;ydpQXuP=1-M7E8**gUseEG!(Z$5o`
zF>>4Eet6$PZ_GsNe!KDeoBnXWi&lHB&#cGM5A5!6yY;zU^|9MMje~C;bJC5MzE*Y@
zTAw-<dwM+D-CkUe`)jyr9^Uqa@1JwVj$_LztgEM0?R8iBes}G%{?C5&P}x;%R=$)!
z)N1_08?C+7kt-Zlwu-NB{@jXJ-}ZP}h5Xt6g2t`t-f6z;F4}Yar|&K6;(p5={hh5g
zzqhE1<Y*P!eeN0ku<i=i-M6IjwaQn!%RRl`cUD^IxXX&J;=S)~an)m^$CllNDIWVv
ze6#mteI@^Pe_LVvU30xxRAJS6+iLpFf4_IR-4`0rIP8eee0SNGpDeDuX@M2be{|%`
z^O=pF-(~c~6@OdIx8N@xS?J-L%ij(-^rTO$f7#4)7yHf^Pn~U{eHxFKu-)@)R;e$Z
z&vbV<^18t2k)yt}?@L7$R{3|g?YGl|7hd*gSr^yd_m&q1FE_qSVU=B^hgEpo@t-^X
z!LQs`q%iW)Tj8ZX{GI(qJ^f^v!VfNe^7=b0d|T1o*y{OtSIf2P&sLq=Nw4cSrS6NI
zD_7xz-{0nvH6~9eQ@F*{Kd$(~+BeL!w>y02g8Ap!>%lUGUp@Tqf4%0+=gMZ$xpceH
zIam_IyMy?yDc=9Z@D@`>zj(y32g+8l>b~stv+h6X!5#lzMs!{u(S1V|R_&9<oVx1k
z-~QGOMR#N5v+h2or}CxjTk5{>d#{`{ZsUudFZ&9*$7vQ9`N_y}f4FB-k;1zET-AJj
zdEhaNzdZP@BI0cm4_^O(+g>bS>OD{0_{Ne8p0(D`o-M1eYCp-}Oktd-F?`m!*W5g9
zSXoA7XH%G(&u?vg{ZV(lcWqgPk3YG>v^B1OtgONv{&@NUtIl(KnL_?|#N`kA=72Mv
zzol#yqaAB^xK;a{n*XWzDu1j>|LUgK-51w=uhUoaIsfXedkgi(_2j0TE;(oDXRj+-
z#Y4Zg<zhEq{X$uV`=7YhtMeQ&xk%yO>pi#RHj7O!VBK@O9iI8~o{t>+OaY_)Fy8z)
z_tIxgn*EAz?RbB|58~&#uah5N?!RGN&#UG`tLateBiDTCsy%leHolBFYT(b0xNz+F
zGKEo|@11tm5g%IM*`f+#IbrVscHD5aG0QG7ra12EPv7*yub-QVaXi_TxDLekdEV{D
z{iOZY*B`xh$O_B8THLnJsi$xD^k0&g+8?ZV!56lfwA8C*cjKYcx4Y!rrJpNOh)0<6
zM`C=uH%a+x;#>V5Lb;b7TKN8}22Ls?KEBT2RbD;%p`xoeXr1eRH_y1SMa19TviaHH
zSY~ojg_GyM|J?n)F}+CPwFj<z<uA{hT%_=kpG_-2dFfn@D*XEIZ(3rJ`yMYMruI2|
z9zD;>w@#T*RN)od-}1;ayFXAs`IGy)8Sfur`ry|VIQsjm-BF~lSwA85z9tp#-77!S
z{+&Jg{1Zpa_qWw2l&fDh^amFo@l5I56E1b;fe-%dt(S_&x7%p@ckdrvym9R=tFH04
zxBd+`yne}rwmtEc;>k^}oqxj%2Q{j2+V)@j;((3jY?x==h23pWW|jK?alum?jrits
zWffXap!bK(VypV_^hegcY40CREK^w5-%af!`6X3oezl6R|K5RG-%~hZ+_aO<fAgs#
z;)+{;bL{<xzFgdS;=$i}Z29p?TzB$$zb;>Ur-*p)DPOy4tNErDRakY-!@o;ms$bq?
zet)S$HyQcl6GatPt$%U<np!WG+WW9`esbv}MGC((^{LtZvBl#>3SU2S;MV({`p=>Y
z<9Ih4u6nN!)2sFgsdV>U5u24ab&q46N@3k|-qe1+>fS`Ubuxu1e#G<BQW*V}RJdv%
z&}_PSH{#{hx88J>KNcyB@6Cl%{(CAuz6;}TbQRb5_A1}q;G9>>6b?OQ*hl8OazYs~
zb)S}scOH<!_lo<berka?%804@&c5rLcii{TxUvdU{SyEBhhKi}yn|jYBbtZt98lQ1
z?i5y?^YABAXuhP*f2F68ADY6C54!Qy*B2U7rZDbLn+>1(@}Pabxxm9^3h@|`zmr1e
z8?efMNckVl#z!9A_eHcXO<~>p!B+Ft^~>Tpws7ivJjSQ&o%lg>{^fwFXFps-jQjgk
zSpQSs$5h?ZI1j7Bc)lj=JSK&yby2+Y$<(@)@{`1O^?PdTCyahttKq8q_qY#dN2@UE
zNetT;*Twi<yYq*3VbqK0Kj`nB*Y11z%X>~JBUbS!J!J>QZ#MUoq0@JNvzTZ3kAC^8
zho-iS@m{DZ9Qz-`(O#s&ah#hC+s`$Nb@$Qca|+GZu3}sV+ZCpV-G%pF_^mHq|E1AI
zr=tCEr(xau^LFJM{JH6`o%q+siWFAu%cZBz8{#=jDqM9>o${ya>VMn^RNYrayPgWC
z_Te$U>U~DaAL?p+)w|1>-mdf7y5rK<@=e`kxdT?eb#f8WdYr;<?|bb5M;|n$Na4ya
zePgi=9($vx!o8mO!8so|=B=Wu80T5Lzh_O|Z!NgbMQ;^<8qjDJ2k$jy#a}Hvu1KMC
z+7!Ni+tj0ue0oe#h3;=#)xVUzjPcF^qaRfjKJxWDwi+^MQc)LM^?$4ULFw_Fs&3eO
zhgR|MCnw$V`BfKcq;S{=FF50yFAr+83*-HK)qBg-{cMbn>yPkn78riQVrRTjv<u_;
zuk+gIr#lyjxaA*4?{)u^vo~PX@8w8e?b(w?&pULEMipMQ+lDu-_S4xKDcp6rySG1U
z|G69R`rH3D?V1bbYou`D<qm)8?g!^@r0}rI)<1L24PQ^5<Hr4XtM;I7Ip+|airy*p
z6n54BNtLfUXKFP)>cI)8|7p)<e=wi{zr5I{PaJpnyG0fL<(t#re)We_lJEMx`-&Lt
zP7J^Oqen;Hw#(QeBL6#{JF|zWeem3qFQ5OA5pR?*zLWA^IEA~ceENxZOrMf`*Bsvw
zrNTdW>ublpa`|gT#FX8L@y0E6u9u2u$9o&e-z-M^UpHJeKT`J_U5$@C_m~|D-z(ZD
zrPl2q9=S{T)SGiOQh3^L2EB9P@pCnvI&Fnx4q5*LNsRL}hCgxo9&c~?&>W2_O!;e7
z`r^F43Y}YaH-3Gs3$Gdcr`OB6Shep?rKje5SL0Lnz^V9nKNZ7u^}pS}Z&J6us9!5D
zw)Gb`Tlvkhw{V@mt+oB}m%Ukb6}``{i|&n5*y?#+l^t$(`hHs+a{7Obc(#Zb-@V0i
zrK<b7kB?k<*eCuuq1gG}RTn*Cp9zKcmChI43-ePVu6SSL*xA4PR2h-~=bafo9$#<K
z?{4+gdrIhjv#S{2+o!_(D(`&a`=e+_-S<U%<vz*2&pVWe&EBKPzv4%Cyz1t49xbAu
zDeQND$p@Vqz^eOv>8bmh{g(gwIe%FEjuNKsIpco4>YO)~o|@P396IWuaD3-yJly{n
zuZZ3q#P7l)FZhSiuZr&l;`(iTBSv|wOY{uJ_tY2Mzr!oX7SEJ0u2WI|_?>_19xmQH
zr}BwcZ=65Eaa>a2R)1fqtL3&^^sxID-Em-}E{;2Uw@04aX^}=17Qeafn1^4QuhA@4
z`IV{k#%%W;fBkb4iz@WpKnkDv(gg!Aym3lVg#+ea@QnA?pS_X7n~&OWla>B7tw^Ew
ztL#+@xBl}#cN#qD%>u4-_J*q*wBzLDcO1U&P=5;JdA)GeIf3;uVmx1tc1!q*<8GLL
z!gVhcDZKyqe@>nBkwFdpfYGmu;gin4XxL9rm{dfxe@~t7i;wRvVmR*eh3l^0zL!fO
zIZ_z+Va<j|oIc-5M{Rd^k;2qFZ}$mxG3tSE^xsn9s`v2m`=_dKv-k0-dmiz1^)c#a
z-S9aB4m|z*!v{91(C;Euq3`Oe?E6j2EqVVQi_P9h;Y*i%@T~K$o=~KaU4p(-N+G|}
z{x+`H!gb$^?)AgXr$77bLX8xr?hnPM&KqMqKd0{gsp=h)?|Q4S>iY@l(GRK$d%v7Q
z{#Oc9;~V2s@0(+M^bhNXQ|~_yS?tU|FLC?$a>;GC*z2YxCeFn8zT(q|F1XL#>rI*Y
zyP@<l=8NB{@}E-}{b1oZF7yg+aNU^$*E?fOnZlGG8ROaQc>iPH)GGRJq0^9E>1!DE
zAr-ctYd5kV-Hj<fw{Cpu-9fx>?yj);)NXX%)ZLi!Kid@_&k@=kj{Ebrhr{2$i02$J
z9KSEG3jg=#PTAqoQ|51^aHqW&S?HVtKhUVcR{zdbYMo5Qr_QBP@y^vc6+3M`i}Lbk
zI~C)5fZhtb2WS@K`<||bo87;5HGk?|b$7?d{qZ0F@$=Q5U1wSWtM<8d(+Avg#WmMH
zIlc0H@4?2mez43%<BKYczvq(**F7hS`BHvSmH+3yrwYAGg{gD7RD50izIedhL+1V7
zJw;t~9$6Rt{uDh*y+3F+-Z@35V|>TbDPiwP>!R^bq5HDBnA(Sl@2dY9_vKaLRKI#E
z-aDi!jQ8_(!&U3O``8pZUxD%bBsKq=-KTWb9;ED``^1RxzP@fa<=0ym;_uJ#A5!-O
zsrUtUJ$#2lwtccpp?8|*O<ipE`}(@)vhrCs>f#DRpLq0lCyXxZqI=@H`1aP%?sV-^
z4;Qd%oa1?4RoFdTU5xvSR5*2iR23i3f2+cE&s}_<vC`wae&n{l4s1k>_OfoheKlVd
zZ(U7~eD#&^&hIS$g;Q^uSZ3@V9=dCszjDod8!S0|u12faU2;8i#%-_ew84}jh4DTx
z72f8XhrRZp*)J|zMe9?<)Os$y%8!Zr#?<~Ybq*c<k{GV*r+epBg?0V(nD1+kZ1kCF
ztG-mG(EVPsnEE{)@m24Xy$7wrlw4K!byfSi)bC)4PrVlwUnMs`$NR(H!qmJ>z2g+`
z{3YdgnrH38ude>ow@w@QOaW8p>M{Pp2Ogh#*?KP&Dg4%lwwS)jE8~l&Cx7br9oL*N
z6XW|R=UdO7x5&Y-Zt-4G7w>v*^rIiS{LLbwcWO}mb$@TCujaew$jjIJ!|4MWUBzbC
zt*ZG}H@(?*Q~CYwaHnF`dz`+W-mY`_s&lKpw$GuD@%PndwJ;vqg>&t8?nB3KHGiX1
z(f0`b1*`nWzLuVPH`r`^yYyK-@q4ho9`@Z>Pos0$PDA&O@tuAu>>fY$cOO#msqd|0
zyuY_n7wt3R{yv7iV~g=s;paD6`uIUNjVkM6T|YABt2%F%9{r7~bM2UJysOZ=WBZO~
zvF<uTpPEJcqKK*Ak7{;56wg6Y;YmA8y?pdW6N|bS&(Blg_+8bqC%$mtr=EMesKUDU
zLGrcw-COGS(!|H_hU@;0)Qz7xbNNp+UM#AxZhi9A-H*%XyPCM~Plc_Q%_6(hRcyB1
z-pcQMsJk)R$yqJjtlwCbKYm}_Y2muRJ5rS|&YQa5Rn^UxT4%Z%&))Ym`hL5=Vcq+g
zp3c|P`=0J@&wWs*W9s)RdOJRThmi`W-g{NWSKYJqbb6|P(XMw_IDYroYB+UYmHM3m
z@m1e>SEa{y*;PEH{$AZfH{Z0v{ztx5#_wSd+U423mbz<FnZmd}H+zm(_1-e3$G>}9
z752M2RrklztNgB*p1RMjicfu4D?W}x-Ecg|s2i?&Us3m6X!K`VU2o&Kr^3yycizvY
z@WJnIbIBT$CzR0pfx1YK>SD9|psMo$-&w;dJxrz7Juhl@-x1%{#4x#A#b)J+^I16B
zQNRD{caT%)T(B;BuN;4GD;4(l5USpj$Mn>Ds2Jbs`d4*6Bt8DVPF2|VJFxC|Gj+fB
zc0TFvj@L!!QFSr(`*o@Ks{NVYyMuN0E!AF(kN$nteRN&>n966Ju0p?$oWj&Q{;K#Y
zd!6!|#K-rCRblfuVzfKWhW$QcU&Gjsy8cYve38elhDRQC*rm^IKe_BJJov^f$Nty0
zZ<jFs9;fir>wRaXm5#ft=qkqXNrnCHSi7+594)mT#lMrvZlvBvnh#ZWER}Aa*Tu_^
zEvDaj>oX;+>p!RR#eG%X-$_XM%`x6P*j6$6{V^Qx;rTsP7}uq`;RQyI9QCDrUn<&#
zsdwX5@qO+0BBEc|)8W+o@;lb4??n0K@Y?g18*|C{H%l1x!h7x1zBYbm>G%3mzsnW(
zDXoU9&gWu!^p_U7<mBZ?Ej_u6@3`3q=~e!WbiZTRD#m%;SHkRA>U~uF-Q869_H!Oy
z@~PKeEAeZ88P{8X-_d@8ziEGOT<y1=^9cK1M7(uCIVgWb=ho!3f7gBzr~Yxp7uLR^
zOpROYXRG0=cRevZB~L0oHE&b#_Dxk-_nt83b1s|mqr}(sZ>#dvwd1Mt_84Eae`(k6
zYQ=hzdOBtFi$@H5piKFLG2Z!I3Zou-mr{lCyYjl>Du3DUnMSPo-B{^W`~TE=Qr&p-
zlU~=w)bFas_`2`^<9Suxu<!p<_cSrS)$b)z_NMBbt8TjA>#d7*`Kg<a-HMp{dz`6v
ztKz*Et-`8(qIBb27gOK;r{df7eVpGZ=;}E|tL3A<SB2}&Px<&imH6`NTW`9`ACrGq
zFr^o%zaQIO@%9H*$WB(_SATcg)LX|sP{Iqp`qz<9E&WJY7gM}c#i#t?R6M`4E;?sP
z;iMBLy!hs^lgfzE4~zCi*xzZ2=NDCB{&8Jwc3#Ev^19*ZccsGdysIi)*B_LxE-&&`
z-A}~y&p&a*e1BVgLfI_Z&s5=Sw;uAt?O%DmY!;inXS6@7i}bcxw0~_k?(*U5-ahyb
zbI$xb6n=-$y_feU{{Axm+<BgHpr`!Kg?7DYli~aQxj1N}mo}RF(1|nAecjA`+m;Ux
zJ#X5rGtV!~6ZiOaf4{ow@0dt;9vt^)sj&NX{e{hrUr*()%U4(P&3)&e9@%~KNyQp-
zd~(|2bG$MW?LX?G_xusvr^oLsgsc9}LrkyhPs{h})Ema!^vAgyuxh{EY<gXLQ8!;t
z>6dm|ea92aRh{=sue%?SFZvI4!*zc@%6W3t`9w^wsvo~yMdyW`hEYE{Eu8XmyBZ(w
zEv&b7agC#fZvWykw-*s-uEQ1QzRsf~I{&W1)ckWUox-Z`kgC4RvfiiAIelG>-_No)
zDa5b8yZLYU@hzvlJM^yNumo0}J2m_FvHablh-WQ3?Zw?Ld%lRc!A_T+vB<kG6e;v>
zxn1b*$hB(!;@`1};rKpTINqzJ!d3ovyhn`Ts{KyYd``W)7Vkc^3RC?{#i!=4cz$~d
z?Gsa2wSSdvzmdXL^`_Z!RrW5Y|NXDmocUbYsXWlPs&_RheHL#Yn?mO;RcPOtS`VWA
z<u^yaA%;`$-o;nFYv2bv=S^X}KXg71lYa63<A$EN`RKCh{0%=<$X=yzz6F2r$U+a_
zTt;lxKD0Z3)i`uYdbHcE@>F$x(d;>%{8ir%wL87l{x@5$+4G4$lYgG`f4BU(kCdtT
z+V1#fdF_<^V>TMT`C}7)UDQRtt63LQ`=4gx<2okX?C&sEy<dy(15@F;?;+%i_x+Er
zx9E4b`szJp#H#x#>2d!M!>N6<c<-Pi4n1YqN9Mb7LYcy4cG!FU+0UI^w)*{V%HK)F
zSB+OpPyL;q82?^z-_%bn@J0!%_FpmmGnX8B^=v<wR#ahJ*J3#S4pOt>=pR;{dq|J=
z$$47U{xFpu`&AW=`^mT;_wKZ-82i~%;i#A0-R?>M`1KiQKm117EJnFfcG>-%_b63Z
z_x(%V_X6_8@42ELrNXKAeW`f+oGN6G>*5Yy8vgdrZ=PJjs(s0gKR@xruRb=Wtc$*X
ztcz9eRHUcgJy*rYdycx{l%E-SuCg=IozKN{>Z-7NlB)A&>CK+!JD;mU>lBRdVp8E|
z_a#;58ZkY-FHeP?2Q-V(uc-<*`yJ1s{~WZ-o8NxBsKR{?DW?8n_t_dLjCQoIhy5O&
z-?{dF1IGQ6a~j{>2uHuxd!Bf{XFZ7L5pn+Ac+-)ic3%0>68gIU(EG!2zZtsMfve8e
z*yg`}{pWih9niR8<!7glJZECz_b=@ap!0M7?~4mBy-#D?F@^U8%6Sj!yBYmq=lDt1
z4e#Nc6SRBW>)OlS&S!mYH@2J4RDDN1_~;_reCU*-3ZuWB3U593?<2Qqyft%w!EV!^
z6sGDEU*(T9n;!3FqCb-2OZ>&xeB+WabIsPMLiyW`RqMg5n%-`nyQ@BSwONev?0NaX
z?H>Bvq_VF6l{()LZ@$8~??{EMpLMb7yWm#S<NDj~aI1e`eVI=jIAGFO-YDR~WB;=H
zX}_FaRH5J1slrq}mo0tu5@%ofPSI7|e21Ox`20K53g}%z3ZtH;!twXMn+>biekWq;
z+@{^}@w;tuMXdU}Sp3Uo(K!XIvMc<yx@aG6J&suSchu_UOYNt`^M~sqKRAWHw`vws
z<5{)etILCYRrh-_z1evb^V=^}Vb$Mxm7e<U<_i~(y7;883~00)-S?+3b>1gFzJIS;
zk9-%<E{xyb#_-_1rmXm@g~t_j(S9z4QGdD`w(m`0>`y8j?dPl#ZuOq2s$7*G|66(=
zA-t>l*i+?`?{?04;ixl5JY4qm^se4J#qxF4=SwR+yy4Slj4PVOp8oxyy8UT3pLbuK
zirqa9{Js>%bEW8?royRnyVO05__)uAVecKQFwPfY`;%C249DMFa~@Fl{=<20MDM_>
zFz&DUP4*!$rPuD6QsWcjoja#6{;sL>2<QDU-c!eZ(x=pUM{GaEw|c#wt-^TEB%Jc2
zydN;`RY>mCeR0&U7>?(p!uq-MJOBCb(N~>MRm~IW&DzyY%^%<0)D83R<9Rs#Qs~|%
zV%2*G>8n4o`6bu=<GwP9mks^F#Ya3-(wB(rUc{8VF`ivcVZ476PQA|-AAg5uz%5r?
zbM2GU3x5agsLj^D{ghLu7SQ>i?~qb|x6bdUI9Gv@7h(78(0=LEmj~_h%>^DVBYGFl
z{=#T~yqk&lA*uI&t2F-l-#;GyP63@ex<~eR<=iVm_SSds5!nw&58PKm@3>$*SBURF
z@D<VTkNB=X;@5wF(-Mo^_jm!_r$Xx-G;iWLighpED_dWbi+uY|E}p-5m!dzZb3*4X
zFn;eB@2{<2)&=t@>V>d+p>c@*RO)=u{dd)VHTos=0`doAyTayGv#5M`VcqZ8tMa}6
z`GrSs^VTgz6-GUZ;U7QvnbR-0>hdy$>`4l}A4s9!%YfazF30aVb9lp54%%^Y(bGJn
z^rchcty^`mO0U`VPQ}!HN2kQcb-2^Q{DpR7mEQ70J00Dpbt=a5mQD-z)H!@t+lhW>
z4D*+p#nkzXbD37L>K%0|eY5pmKWf42r<Bf_sxW>x6~p#5seY#7<2h#Ca9w+p+K0q=
zdNbSWCv5)7c_);0vDtMbzUQwSw*RV&_S3McovL-9*>wBkzKYHEr`7y-f9fmS?7GRj
zMaK8%p{`$?@;BqYwkq82d#m{S<jsbiU)BA)Y4SOzN@3N0OFebbc~8W6j#U+Q-qcn6
zx8IRE;xpe}_T?vws(x>H{gMl9d*UkvjKB9(8J_&$Yd5TUXXzZg3afs%!*|m?h0YV}
z;%&DdF<{q&9xPIL@F`!rYODFC7VSdktts4lz2}zPX0ho-3a{AymPel1{ec1=`T8AO
z4H-14i0EBz#Ad&<XK&)~%2b8@{l_Z&%>u(uSnP~9iYlyHZ|myUQoo#b>-PqYE@73Q
z<M&~!FusG?@170z8~?>WmpzUDc3%I{Pd)hXlGl~X2Xudu!m9Jvy6Nt#Bc}9Ue9Hcb
z|F?Htmme{7_WRB+V|=_Ps2h&*3ZFG;_A9=%<NZYyx^Ih^^0(seD5t`4KKuJa?MCB}
z!n!;*2G4bW@x=WlJZ-%h42;-+#IWHX8?wZZC5H?=cEmO#hRpV_e}^wPe39Xc4;gUu
zh~Z2AE4<2(fkOsvx&2DVj2M2-@q@>%Ffhem%kfigEDt#9tg%IZ#g>VgQScvYUboiB
zk3Uhi8@no3UoYQPed=rd+w7jTW?Osm)n&7o>c_HQICsqfSKe2&im7_r9siFHoOt9n
zwtcPWE^N2FsrtUL#-*E|Ht)nD<A3HM#apxSKU)9!`M$rws50Y!=Al_WQ~A%oZGq9Z
z&;Cr&Q<#z?$N&Ef{P>o2Mlb!92}P%3ciH!<dh6=X(<kqC@ijY6E>rmKb-x_F=x1Ln
zVcmLL&A0bazh7#TiMN)WhGU<8b;Vov{%6_K*ecJ;Rq0VqZ_Q6z_l%V;Uhw9LMOU#?
z?Ov4ovXige=%D#tE;<#z^UHx3-?;CyMaKWj13S`PxcTS{S6uG?=gUsVt4=!U;<+y$
zT}JFK-qicwPam?`m+pAI=qjqOxAE_v9kb>ITfb2B6!vz1Q|<TE{v@`Ol1Kb%$9`zz
zecyh!Y!$!s$*tEqXOp+eW^u$%K5+Aw@0w7yihI3u)QWc>F|q6_Zt<1(j@oRaNoA+v
z7cL)i;JuH%T6QYd^}p$B#Jc6>J7JM)c3f?z=gKOK?etXm!nuEc<RuqPFS?37C2y*o
zCHH*e`&T^sT#>Ok9`M=Q{u4P`#ZEg9An%VqIrrI*pZY-ADz?g3yUVppf7&fqtZ(m4
zwi$KfkjILQeT#=UPO0#No7{ZX`fof_wi~r~_t~FVVeJX;741UxX3PxKVW6*^f7OvS
z=V_gRc+Qgwx86d%l0CH!M*pBye@D5E|9SQ`htK<T*{R6?>r_nXYp29_>iGq^`^!6z
z?|tX6-QQhyY}wsd&l(OGF#5Bbe)q3$4{WrGpS|wq`!2HffJSfQhTpm4-BEW=E82za
z@`u&4@W8X@J>j59MSsJccK+?Vd;I8?qPwxKK6f==tM4Q`wOn_<Q;g%$Q}+NHy>$20
zKU;H5QH9sP{rtRZFZE_w7nfae{s;ei%z#F3W8^i5JA;AlUia(P-!8VQ=2u@&@9Osh
zy<NWD`&oMN#N6MVIC|R4Wv3vyXH`7)iEm!>`1i+^8N0?qUHxg-ysz6{cjfCTe=OBb
z?ALXRpSt72i%uw;#a7?B^tEy+o?4BM{MHRGa@S{f*!YNdi*{q(_S=>3f)%%W^vmbJ
zU9=0^C0BRXbI3(Up1j{T-Yj|>yStyM`ZxUX55F?}wkbuY;|D)Ea@4Sw#}%E1@ttJ5
z!>Ms^HNNh8<2<@9#(SVt*!M)RyYCeqyL74P+if?#>@Qe%{Q7FXX5aVrmHbct;Kr{k
zu*=(J(^&DHUF!S0cE{H}_v^3uTD6yby<F9OL|>Ob%GXohceJ~m{_=g;x!eBfs9&Es
zzGyebac+0GQ_k;Fa`v_Td0lyyTYtZ`CZ6<CnXw)Zt?GSOzvJsG<=b_h*zS5-9hY{O
zoA8MnMqTr%_lmlB$92yg@bI{)MY9;c)2<3v`57_2+3$5aZQoDccE4Azs;}Mp7VGnU
zTgKLY*CX=T-Qh1haL`H%|NgGBQ?b+a%)8~@!tU0uP9L96eMcVUUg(q+mYR6=y=A+x
zyYwge1FeR;OTMnw`|8l|EwTOjQ;XihR^`9u;?=er{?ea{-o{qt@APy0o;sIrmET#V
zT&wn|ua)c6_t;f(AGP>n6Mk8~SoAbT{qOB}omD*5O^@Hb_+F!@vF^D>cjv3iTPok`
zJFNTAqVN2sOkvgUo9}e&0~h~(t4GT!bdTQb{^DDcR(Sr#htDn|#`A@$bKBH;b-V8;
z+g0D7`8J*S^UJ1|hkfFXi3_Yev29qzQ>*E%zITh|etYLur;ONZLK(5U{H57ep0N4G
z7r#-~#j5$x)9KB=D`@ri!P?!gIDdOP9M{WE4X5mUj_=yvRp=}HwW^o<U$x4ChnzUA
zY!*-3dHdZj`Nx}OyRcb1-t;@AiuX;ees7WTADWG?`wp?)`MPVr-l{%Um8<&>y6Sfe
zdn>)!-^pq99nb9#o_x=tW8Wxx3*C=)`tPid-EGO;2d(o)(O+?q(XU_q+#zoj&%AWN
zgndtYZzjgy)uc!9_dBZev+5l~-S_0?&B)h}x%#D_Jy`koh3c-$_MZ`({ku&uf8F-`
zTE5;s4@$M4I){kyCr-Td^WPeGd(o+Q)p^f+>e#h!E;<!Em3MNVvE(YpANBktMW<qK
z-+RUOtuJ-4>YZ`bdLq59pBVG~;n{!w<K}aJQe^zkJhc0MzNhQ!?(ea>yZ*ZGmiubH
zcJE(%y1rJ=3A$RYYG2&0^sfG{V^7~7`_8h{apgZBFnZL4No98<J+YtaDpsx6ot9p;
zpY3buPrSI{W*4mbWKkE5e+ui)qq=rC?yFl3_w?^!rTD2EU)SI5Yx$h>^i@p#?q<8c
z=X1^!&u5%B?DW|KzP!do_sslz48pTIHtWx~>;9qYok`vFc#hX;;dbp?qn+#S-xuxf
z`n&4qRF#i>^_6h5_mF)pe>_j>YvJx1=dS+VQFqDJu6n!soTR(!zwznyuQ~3+@06X2
z&pbG8&{?O>Hmcp&Q|C_Y@*8qJv*zzKw)@>;tgl%++0*A{eYL%&d7H5h>*~I%*>+R;
zySi@ocKLW8)@k8(@s}E3@qO)FlU#Mr<@#$rJhThr`A?^Xd;9%-yW4N|JCsf@SEbMK
zdzq?mfBhZAcK569_dk6tU#s`Ht$qhoRj%3h{qpyfdE4FR2|ZPRSKlG@R{8$ouhg{@
z?fR~!-Szaf^Ve48se4aSweIHh%s?kIaQI0NZFKjy-zhp3Q||$qjqj;3jOT=%8XmgL
zZyud8_}#LrSl15ASNHu;tNy}!M{ct5tzUe-tc%UwAGCVET34Q`e9i91Ta~w}Tz9{7
ztg4?rbrtLCV^8PnDSOse+DW~mBcI>1Z5DmUT7|wJY8Ly-cV6**6y93JzV=SGUGmg@
zXWLiu^%ecfa&-20ovHEbtG~<K>iK88%cbmRyT323^0#7oU;EwTR?l(zO1Vzm$9d1*
zUHJHiMof6)BM%g?%HQrTIeMzzX8phJ%Ac}3b>r)vQ}=Yfy7IT0uid}L+NvDwD%bAs
zVyfzEm)w25p6wQScfzX+Oes1I`|5XHb>&L^K1`>?*WHIU+n;9NKlT;Bi5$MStBX~B
zL`?S`S6%;J{&r!z_et&cH=F(a<(R*#_PVFuKQ!Bpel`2|$723=@AsOmukQELeLY{Z
z<JVX6w|XC?96ii9YctTU^(fb~)@*A}zPfCchwj?9&noq{YZkYAzg|^eyT1FV>VLE6
zv(4V?w<~{DKf0Q3f6!mB+26;g${+pkzT)q6RgQM;Z@Q{~eXV?7c@LkGqw3vFe@U-;
zXVvcXR`syEzTd1XPrKI3-ma%ze_y^XZ>fCo9yb-XZ>)>$`rU@AzjM{y=jBzu(-YIH
z{JpNGw|o9o{T=&e(^GoZUGdGnzf0w>E8hz1toQ%1ci-`L6;=QEb5rRhROuZ;uK|g$
z4<O~?Qj}r=kq$utK?34~bO>D#0w}!*s35(Flzr(XQUjq#4G9E7O9-JSzb`NE`}kuX
zx2NsXZss38%<P#pYd*8qnl-ccxjAm>B|c4B1*<)8mg-k7AEnyW+du0iKg&+qrM_O@
z3)cF3^#j)5rt8km7mI)AuwBZP-jBCx`+C_$z2){jci(kK?e|>LXe<@q@_t_LzgyTS
z`FZ8Hjh0jH`}9)n^0tdCd8NLeY1QZLvX5rzS!(_BdY-lJVSTyZA+&0{QumVW^1R;r
zez@*ox!&=z((i~`wSAbcggoX|zH|M(_L^_=K3%1-)ZaS^pI7>wY^i&cyyja<PF{ac
zt<iGc?f!Qg8-=BKY}WJeJGe^A;lBgfZsl^nD{ZxQtDn2npJ%^!Qqr!N^?TfQssG1C
zYc8?s#HmTU<9l20Gw)7oeU?;;VZ566)?0tM=aC&<A(nc-Q0aRGfA=d3EB$*5rP_z{
zY*cyX=O5|2`9>WDSnqMGy|=5kUU|z8zlUhIa^ByGM|r<qYR+p%rloiY|6S{LD_8ox
zSgW;f_wTz(-FM_2=jtu5(SNU|)$FlSeaiiw(z;ce7iG22%9FzSL6<#p#<!+?nub^^
zKk~fN`-`xBt=~h1^HIiU-uLF@R*J3q?<!QQ*Ap|Ja_kR({C<*!mGV<*`yrz*y5;DZ
zUQMgTto!r4^TLw{-m=T`yN^lB#ahSD`g<(p)>pa?&1%ojE5$5*%l&;@slOwh_1#@o
zeOCO=s;@L&hWE{-%9Zk2sXkf!m#Q!K_rI;$uGHU&3!j(b!}GqgY&7YtzkZfhi{#{u
zrOs8Qp69Jkl-HUJwA%f8)aQ)Hju<~Vi9GZS9$M4uv$d+%``|{s?+wSvo@ryBl$+l+
zyIpuc&~D|tf5%p?m27bPS-%{(YOg}2*y#7J<<6(lc9qV%)!Gjnxxvmuj`=F77gow=
zt-o(-cK=pA2Znl8x-PZ0Z};~pJzf8LzhA7?e%hsfwRU2DmA>by^xiCNU+UjwY}My^
z=R1v-m-oMe)+~9I-ov(Pd+!z74NKjtx9an(_Y$?%i=*|z@ZTk=ww#xKjsAT^UV3NA
zsrC24vHztj_qy?^q*fj({ksjV*1po;sb<-u`bvMN5ZjM_;@(S-O!DHP*?(`RRmVT@
z@zIYhb;2iU7KYzfwQITE^0H9hX8k+Wd5zmH_u=AoqcE@hy43Gh>ixV`&-vx*_0{d?
zy*bwg<I?FB^S1X!|DBXd$IE-ZJyl-#UbtT6a`~$?FOTi7-t_F5CM9|CQ0uv{-2T3^
ztM+$(rTR7dJzpIE#fj&o<B}{M%KiOGY<KU4YfXMF#Y4Gqv{buB{Z7Bq-zC+neY4Le
zrM}Of-kxXKXSw-#rQd^P{rzwC<({7^egCRG|Al9-(ape!VIy`Mx=iPMo%45goG@aW
z5uLri{_oJ)hYlP%PiL>=M+}|s>+*7)9i1Ip?zGhK5koINX~XeLbo9FE@(CL}bodqN
zxV3hF@u%yJO!vQM=}A|d_I&!_RV#h>nvI@HH++2YM=o3ar6PX((WyhWy8HEF*>nAs
zA^9`!vER23JEAMu@!Xxx-hbc|iSmOZ-kG=0J~t*uKmV7-*Iap2GHB9ePpow4i-~^g
zE%fsh_x$L-)bqEGf9&qLUVklF_79tOoc^OX6aDWx^O3XX7(O=H_SDYh&%9|&^2HS!
zJ%7qqZztQ2yYkajM~yDxO*4;r?(jeTw)mW!uclsf<?;)^Qf&9ss|(+E@8usR<UVm@
za@~oSy_X#D_6~a-wev^GkT=g?d4oeHB`<#P{1a1G`#9-)>(sFio$`J{Kl<Ku_?~YK
zd-$OQo^bN6cOSUl$Yi}KW1kwh{*}e~(lec~)E?Je@q7Z$`gE}&*KGb$GU1iwPP?(!
zvq`akH!s%7yvYeM)W>u44tXHYbN#KGdh29eq4kIC=8;|}4eZ=~TsmOxKHER|;ydYO
zXWcz=u|XfE&z*C4$1VdWrmtPO&TJQ4@^N~|C4awc;m1EJmjAifv=#0<^8I4FS<d^;
z+82K@F@@~byyy=nJ$UW1bKf;JUG~xacG~vJX{quef4}j2^UeQ7dh<<dj{E(_pB0fF
z<QxCM1&;3e?qA<cLo_e*gw}_D4C?phm!r42J!RMI@w}S{E%@j0&lKzNXg>OBuRp{y
z9^lTO&vey4&UmA!FFheW**E;|lphQ^yzeK;w{O2|fdL19P_*m$o?h|RTXuXZdG&yO
z<~Vf8`-}ce-#G{E-DkJoKU9nl^Zs+;OD=xy)wJuw$NJ6w>}%=rpL8C0+EruHH3keg
z;*qu9Dk44bwg0ta?*H<}+mltVz52T4=N*-j!*1}e9-0?aj~|Ff>jK%g`S4ry?4G|s
zyonqA-+1ZMj?*R<<Ai$gLs`E)ueoQVnVy_bjBEU@sBaLb=}%9{PL$aX)GtKy)6P7_
zGdw7Vs6E8X?=F5|-&t1}pFs8ek{<Lfh1wUfo0np_JHCfy`-FKE%iZnGKjbYen;)LC
zP(Mh%@yUVsAx}GqZ|&ht$F2U?Z!Pjgai6kZ*q^i`ckg}+5B&27uO{r1T>Bhetj~$3
z&$P=^7rc?MPx~|bHJ*I{?z`((JMA|A#N^kX&2!+L?|o9-AL$|9KRw5H7d>I(+i7vV
zy7x<ZK>m;wcZ}Qjj;n?*d&lTv9wY9E17X||C$4$s;x!lV_ez?TP{#W9U_d?rH@NVN
zeP0<nshCfMd4+u*t~bv`dv#s-VY>NWZ~gTG2YpgRv$x+q|Ip<-Ew#&w#r=z28P`6>
zeiu3E)XCdC`g$?G^S6Cw|Hbx0x9Up82k}oodEkSW9e3FlbH828BiZ?dvmXEUrZc>q
zDi`yJ?);Kp*|+7z^3U5h-1&wb$2^|ex5IsxykjQ3x7Eut+?c{*`|k6C{nLI;AM+81
z#1;N*e}l&3kMtJ@p?y;vvagUU?rL9@(;e@{vDrU5x8M2;PD$RFcGgSlc72x2-)Gx_
z|NH09i*ZZ6cqt#kqw%kJ_gm+sn@uk6_wuLPW;x@gwfFq0C~wRD12?<<%rA=Lkgvb<
z1ij4D{P0MB^E7Y!2%Z~_;xIq4{_5GaIL<E3L)m%9eEEy{DC5PxM!xY*yXU=gx4!C>
zlsw2T_?<Ey?Tf~zU(t@b^A7%NzWjh+6`%9R?tb*dLpVPFq@R7$y3-5Vrzf3#;>fLk
zF|L?z;YmAkvXCD1GJbLX-TN~hL!R(QPx*oI)f->gy3+qA7xZ8I|NU|D;ykP~-a@;g
zr*-$-`kNR1$fJk((8KuJnJ1JtSa0iW-JQQe6yMZ~m&)X^GxIWUXg#2L*e~!+UuF9c
zdx7Ga`9kYKJ|5}CU+G~SNFU>hpXAFA<*VW>eXIw0p&t0cpLKCQv>&U7>g5AhU;gXo
ze{$R8)HziDyYBm7;xnVZO5GD|d)I-(f6(z|vAp$Kb04+N-#<%F>{|8Cv)}o5{gKxn
zeBF<K@@et@V$LPbTkoz7C#U09K4<Q~zx7Fa{R+>$^1oB3r1Fz17cNXW_MG>M`9^pj
z5$d_g(my`v%N0H@-hZ$o<9H5@N6u9XF7)*J2TUs73*h5B^WA%LzxyYo__sdxA@$bJ
z{%79_>pj;$Tu=7I-b*37gZlI95ZMbq5O3HM4D}K>oqOmdu7u;Wf9Sa~G+tPyw|esm
z<tRh?h<o~52lEcsP1(9AhxZfKFGT%}gTJthf2f_Zc|-m2jxQ*lm?t^<LF-3PXrAP0
zhi`I2#5X%PF4R7hi+}SrURIf&=7AS>t86@WslEE5UC4L4?;Xv<J|cdG817T<5yj6C
zokK!&?sjjDAL~dTdgIG_IFC4YI9G_z?q$U@@l6~Q*OkRNahIGdbbnx6=Q{h1bBlf7
zbLHYXbmx5|zIx=Png4feDlZ|&JqZ1rL+NFn%G%>wz57w?XME$?-|QFGoxbeSe6722
zruNxvvy-=Yv$${I@ufxnIpByNe^fjVTL;gLqij9cq5Mlb>n=a>oSy7P-W{TR!h0Eh
zDxcxUP=3OX)QfxKhP-g}-iJN%+B=hq_hRA|`TW&Am-_`c&tE>-=$|M3d;e#@3#`8V
z=AS+JX)3R{eaM*;ZawvjV!m<QI-_p+)#ab3@(}&(^LS?;?8tqF@u7PQ>l^yDvilD9
z3fTwyGcLK1e0JqM0Qn(056c(vAC~cM-LzAN*3<eaYj544IK&^sBjxk1JL=c}bNjmq
z`+)4uJ_^kbU*d-Nh`(RXztaWNhX4D$7W=`Y_+XvI1%7WFs69I&Uwq)t<hyq<F8%nO
zajc{GXgupou6)w`p!VV+J7rh&6Q|UR56;KVsq!Ds!#F4Y*@vvBIAR~o8&^1GufP27
z9K8y!{%Y{(g?{}>;=E-2`IY%XyyFR<><7>I<ac2ly!OjyuKZ&3r-||ZGVE8EynOnU
z;(NW&pWLI^r`a<<TlLSqXB+wEJBj;H`!*D3AvsXLQpis27wlcxej|>lhwPF)hln?F
z#m^--oa2XU?fg~xqm7?9b=YQ8i*eX~#V_Vq>h4{8@AE;*zMc2zE${ID*f`4c7RSYX
z=QL&Q?FYu8FAVilk7s@>UO;vgK4*7$WUu_nbI2Z{coNDNkDzle`z4<~p5p<J<XaE&
z%!?fR48%vX5pU+9eCZx{o^@%*q@sV${PT6EE&9wyiRb*^{k#0B7zexW=fpKAUlI4j
zOZ~-n_gLb)c%+?v;x9eKXL*|T;sCwj+qZtS!1F7AUd(IUv+xJceXnuYvi;}#)-k=h
z{%8EbV~5_^tLysZAK2nYKkw+8f4{kJ*>r*4U2p8O{gE@Q*3os-`yXxn)^B_NADs8k
z<8JudlwMur@elVb<X^kjIsfy>roFn9#e*yq5AE}FANKLgKiTcmM4S~*@nxNy<K*Ys
z;m>pHhhO`f^N>9A#Sh=vb@2ygDoi_fx1)b`OP|7d-`a4Y`Ihfnke7HLD}P!0r|U1b
z<SHK)?QFMgwtHi<g?kqkKIQuPfB)f^$u__J#a7R~@Ky5Yh|4?I`A6@<C3`-;^!+0S
z6y!xv9wZ)$v-GF0GJA*iIrG!se(nB=e)KUf^T)6L<4>44=0^j*Ojg?Z?7z;qW=CPw
zx$eLI_do1ius=Gd$uGn=dWmEFPW)iM;^NSm9=vIWMaLJvV-m0A8TdA?GQaSi2v6dd
zJi-3!z0}#i8#~WuV>$}*6#e88^kg6QWp?L&A(SUiF`jtA4px6|(oM@A)~E39>|4Km
z$W~Jm==%fwvLDaoAJ)NnP9DxK3;Tbz=FWFbE`DDp?&0Tuj<|V)^E*FF^gH;#zy0i@
zD?d)WC%bdRd$(_Y&9sC*?vcGu)BcOu`yKY<{W}WZUh>Gho*CD_z@GXp)^)<!U-m0_
zKS$2F+t2l_6Z`imJoVZt+m9REzi`o|7ca5af&Gj2bivC9EPCaZ{R+-2_8s<PA7d}_
zUG`@E?0@`-9jMn|y}ZLYhJQIXi@#xfX5aQx=R)_=;-5Uk`4h@Z<VQzcxBJ3tKl1N*
zyX66+-WlGfPnUQ;e&UVCFLP+0uG<dz{tx<%?%Vb5uje@RyJ^3!#Seeto@4*iziZLk
zFS&d2as9eZ_~L_Q#;(!7_*_{#<%K^!@Qyuq{kI%)jT7Q6H!iX5AvgBv3hR~S4dRA8
z%lCoMddlnUPxdQly_8Se?6>#cwA!@PesaWwH5d5P6&+oj*ZlFVPhS1^{$(H2UfF)8
zoYl{B?V<JXe!@PfUO)Bvm5bt^eAfMhJRD!_&Amq!mdeY-ZTkg3QWl@(okQOGYuD~G
z&QRc2gLnSk-#5OuUqQaZU*BK<`-@LG_{-vZ9QJM<t+#q*eq-N0Y`OXWJY=3J#qV3q
zJ3Oz*WAPJ~<@v@5(Rm*K=IdNzT=Q1u|9PYMD(;Ghi_bONqPMO+EfL3$IQpPpy}0J*
ziF)x?9G6cVGviZ-e|p%LiTy<Uw_n&F_^)}|f6b3yi?jAY=TY}R#+6stzvxR(`pEap
zlm6}-$+OSPQ^aZY@&S3Cd=T=hT5-+u7r*V+i~1J!x#+cbKR%&f;qpOWEw}C8W+?1A
z@3C(_SLjpdKlc5Tms)H<aX*x2$h*9^pg;cHuel${@?Y^2%Bz&ae)hGHPkE>NTKdRy
z<c;`qo{=x9*S-|GACWiFM?HJ7-uQs*M?NZ_V&C#AdBmrkqkr}2=DiA@%YS`G#Lk>E
z<To(PqwwW?VI3j+qpz~{)ef@T;yE(-eM7AO=I=dPl@I%gea3zq#u@uki0--6+h53W
zt~QRc{7Sw3Gt`5e5VP#h`O5i6oEK;8i_pG6fA(D~x*rMSjrcAvcV4Bp{6#(@uaWo5
zYdn{~xSuf}a_o1Z9?H&#zR$v!=TN+qcZsjXd365{$vGb1@?i1ZxkcIYFmEwW<2k?J
z!#L_89z2Kog-AZsKh#HAo*}<*u7ma&=XUF3{qQ0$fzE^MRy{j49{xgf{|V_CwlfdB
zJ3neKkL3UEljW`QO@1zKRgVX8L4Qa;&z%p9Yu(web%EAL8M04)1npz|S=?af@&@-2
z;d6c=KhRztXMa@oJp*~(Bl1IfhUmP`-^?>?=ealn<p<VRd|-F(x5(vx>`^^EmBlB1
zgKz#uKYWMioD1E%I)^$}yKjKLv$G#4+XvXEdh(pp)Z1^^DLdz1crah~hd=jscn(qg
z6Yu#yp75a_Z+P}xe;CHs;yB&k@vsB+^iq~@uvd01zhkd3v}5@q{|TRy=N=A*c4D3(
z(!Yq^`6GSIr_}Q-J<Z>9d9iy)=M<<NKQ}+~$D4NeQ1)DVsGa*%^C`|x-fUgFzYo%n
zK6o*Xdh+<8GJ9r+#?vlb5B2(Mk4NLFhxAlt&*ag=c-oO`Ja(_2_Uy?y_4yU={OOsO
z&RD#Eb${zV*8OdW+PQylZethb=UnBS6{7nJzt?zTp^j~SFjv2VbG5SZ-LI4D{s6KI
z_a)(Y&gI7SKF7U;=j1sr?(*BEU%UOYUIp_tet2JDeDa)Q@#Fb}vuv{Q0>}3%m=8Uj
zhspQ;Ks)pD+&KE%x6H$RgZp^z6UfuveAVmcUY{QHhImxg-}^FrxgTL4{EuC-7yRJM
zJn_g*$Ps_&0oe`x#UJDFd-qi286Tf`Ay<3rB;L^zy1yb9Kjc^kJb6y<{g+w)qq&|K
zP=M^rxLJDOC)D4%g#GYp-t^U;eHfoyeqlb4KbfEPF+bzrU%YoOML+8g**$v~f7mtN
z@uQ#lLG!adFyzhrv}Zr^Ui*>!%Dun$AKq(wUjv=b<y9eiKO`S>@9n+1^Nf7EnAddw
zU6F9VuwOy>l6}Ovz&Y3ZPWhUA*0{zC(K$sw<AwW?e2O2+=h>Ha_Pb1WqMh@AJd!-)
zD2Mx=bGiFE_l^9U{kZ>QFYKRPS%3Pm5Bj@5b<SZY`a$ziCSTsiPsB;>$(4sfc?16m
zzn}13g87)I_~pH{arh;>=b!YnF6Qk$1wQD{zdYAofA^~7tH%RB##4ySA0awVI}h_G
z=(%$XJ9mCj&yJmAJcsfr{>1*Zr<Z#N<7fx9R}ST}XZ7x3^m8wv-Z=E6pZ=l08ke3>
zd(Z9H<a@5HU5KGx+T)p?%AV6xy>)Z%fnV?2-8-r`5A~2;gmSbu4n53A8HV-d>3nP)
z<C-V==0^|D>7lGY9@J|WzUPVM)?z^16xZwv;;T3=-a~O%Tol*6xAh)TTn{m{6ZTRm
zmXcHMbL}8K!m|2Os2@L+&xm*8m9lsR`D=*$vefvcp7U4nqE!1zpPPRv4Cg~1*jnBa
zwW*Z>-wQc!xVH`OakHK~hd6hH`<Q!V<Jf=o^IW~W%f1+*->G=70m&oBcXIBhowLGo
zl(P2DQQ^4ZdsE|?xBmDemmGZheo0wBdgEC;`pbXKKfI6Ej~se9Px@ZaJfLytLyofb
zQ4X<`{rR4L)>RK1@zlvP72I1mPuSm_D|}xh&v#C-zq!A3uJoS7dCmT5A9Vimy^`-i
z+)w!a!*^-EGj-q6@sr;j*Z+mS1^0UHJDg9P2d$6qQ0VU*<{a)G*1XLtd~c;bMDufA
zat?M5rH}7j+=G~}di%e7H|H_(tcQB%GG+60Ui4hPp`CHO_Z;=(bkIKs&ea9kkA2lX
z=)J4`%D$vsqmX|5&ODUqWq)*UqMdx*eAt2agWgZMkAUo2JLh)aIl33oPF~}FM%<Cl
zc;Adq<9JRl_UWDk(nFr)eWExgzKSQ}2o#UPvik5l6mds9^B!C~{X7rbi_7AqxDK6@
zJh$)IzvNfW%g$lYx_WN?@vFZw-q^Q#@<Ki6M-Kh0YlzxIa+P6Lz31jdzH#+)UZy|3
z^oQo@K9F6uJBkP4d5k|gpNaeKAtAqm;rlo7!2K-L&i$l&9_YD!5%OpKv=cY@H+{?t
znuoXp)obUT)&5K#`OyC4y^;HL`@Fos`PaT^zqIezf1$j={^<U%s7LqvUEh1)nO=A<
z^4a}<SiN@iAjf?DUJvi~ZTmq{Zuk3U&)JRqMck5Kif`hb_$6NOH-2ef@;$b=FOG;;
zz8lh>9LV3pGJiLoaeV(`oGi32(nDOQ54rYLaYj5*HXamzwWF7Lh->r@=jA!Qd?!T?
z`SK2SA+EdsRkrT@k6l21!0xpR?L<7`XCbm5_2k&s>8sv7x%|=nyL=h4b7g)hkMtgo
z|B^$lJlXp_?f4O-7vxV+Udx}&7ti|9&vQJQ2h>hJ#Se`mZ<hy)C)&}&bK}v+yo?9w
z>AAQB$rGRHkAM0qldDWm$gZrnd67eJdO`Xqvn&0wkl%*L-{cceULn5mQ+Dg#EiB_D
z^tWPt_rGU=AM><+>e&JQ+@o3_{n(*-lN0iTC;ixoa;UHV_+~%QeCQ<((3f0#saK}A
zcE%&ebM^FRH_Gf$yu(j8A7yc0d-BbPKIUaT=%s8P%GQg0L-tLcd;$OXWk<dXmJfyd
zhxcUSruM!Q4)+1=!u#Yf&$TZZ4^Q?p<Epps@F#x6Pvq6ked^^6VSYe=a`>Tkp+A!2
zyypHcEc0LIFGz3uC^SF&J^kf9^szrXH=Cb&`kDtmAh~$Aznj1L$*auE{T4mQ_Z*Vr
zx%nv5lU?u|c_+JM_mE%uj-4OKyXDzofA#Wu&!O@8r8vVc_^&d(Lll?E3FU`n^EEH|
zk9Oii7RsyXVSMYWtQ~)32mH|e1wZ45ke&D*Uj7OBDf`tg%oo`yIr_=N<-g>T2R+x1
zzWf&t>aCacz&CpkzkG)dtv7oq+F$oMN&d=C#Bb}S?74BoY2UXv$BE<W#cTEAwLHcC
z<Nnt9&V8fvlJ8L5GehS>d8zwI_rLOQ&&jo~kjww*izog`59eX|uX)NV-GjOxmG9ZN
z=}iy)=wqMrJ&yT#P96;PcfaS{%HQyB9>xvVgMP*nkBmcadV2rt`ziYRyDaoH4*sma
zJS*he{*Hh8%Ujrqdh+Q_zVYy)eK?*nIrP@wbMZ|(b|dcc7x@qV_$j}^7rE}6d>`rW
zVmM#1C%=pK{iN?5AAWM!W`F)uN5Of`I$0lb-N($f$*$YKyuqA>U+(tKI)@CIqrhK$
z=csHw-TQlwz|P5~r~c4${n;HJtfzUg8~!5hXy^S7zvl1!#k|>vGJCbI?1Mk>d+SO+
z{Ao`Q&*>@tDT_DW%UBoZWBTI7Jn${Pgzsmx!@qgai(S!=-pc%#f9a>*`jDgEyx1W>
za!+SHte<(Q5ABA1v2)*(vm0{hZQbz4uAuSo4gH=$oT0b<O`am|u^aFCtb_6C#oxsz
z`xyI~;ZF<y_NI+yC>S69^4jx1TW#K%3g7)?=wlb3J4@k}LpK|m%ra~7_ZFl6_|qQ^
z+hIWQ@8<Y>P&4hj#qT!Ud&YwP-m}6D)>!w^Ifl(#IAri=)3)AemcpEWd;Yu+o|>gt
z-sFyRpIrHlSqs1ZeAzy)9yM#>iuYDnam+twEa*>Pf5*dj^8PN5@3_AjwZ;+Sf7iR<
zeW`mQ_HG^U#Xh0+_MI(z#yk7tm)3<Jl5c%NJI24bs?09Qk@v7`NN@Tm(@Q)1KD+k5
z)j0e|KX$|K`4Rcz2fH*c^4Ym@_^tNrJ6s3->@U_q{u!Q+oQHg`;GE~YB)@h3QSbW$
z=RfZW)r&*Ui_VkUdB5m;1;1OAFFR-H=kLgP4@nR24a0M+?<>sDxzc$+9B>|&H#=v@
zw>>uxJUAc7=fws4B7cJX$Ug4xRcJ51hv@H%vVZntUt~w(oP3DgvlI5s9@vd|M;`wX
z-<+@Pm+X{%%a@!JoXgb9kDO=JlcS$9KclyL@mHRupL|Te;=NY)zc=MP>3n3JoU`;d
zj&rBHPkVOZ{3pNk{?ECOUh+Tlz$e6mbu$jW<nzWcFX*02**qaV%|~9v?v?pB`xYng
z1J!GX59e0%w$9}HJ&AVKfxkih%*(k~J-c#GM=$pQ=Bdoj@jzd4=s~{z^fbPGOCG`>
z<!LbVFZ&p~WM}Mwy@`{~!=BsEpmQC&fc(O9cBG$s0P)N^>8Bm!U;2@2|5UaQv8S*<
z`?jyk>!5v7ytD7wC+&Oc<qc4tX+6X-_0WDujyPxEHlB5srz(3cZdzCU*#)%UD~t2&
zG2A!ociz|9r|etM`(gW!dhNY$<)8Lt`@MedH=%a!Y3;Y3L;dM5-r4``liJZs{}8o<
z?$_`kE-CBpzL#H{FTFfh#ydU9(@xnvoOb$a2d$Gb{@iElk01Kk5BQgTfZv2=^&z@n
z&>r&p&_DU9{Y-mu>7yNgR@N>H%@<$z7N?<hcoB!hUvXJnckk=Ddh*42<C4p6>{r%T
z{HG5&FqCiKvM<;V=!HjokxyUyi2lZ<pL=HOroDaBKBz4Jr-yyR`WZ)jv)=T<w|qi;
z6wl<5>d6sT<vI2p`vX16wSVD7K4Sj(D;Mn-@>+SP{X-m<&x+H~{$u~ell@TsgjYOi
zXI<C_yYoDp7k#1prf47C|GtHLDthD9deaZT)+f}*y3j|yj(>Jw-?Z;rFY@tfKPFE<
z_chK3^b~L82inVH@y5U91>wB_yE8xWf}dD_`>V3~^K<b45A1|q&QJ6b=d1_6GB0tP
zeDdUbcyn*Y-t?oN_~|~u`OG}T6aGeD_8~vi-#$RD{`fVn^98?D7Eky|n8%tYeY9iO
z(EPQhpZO`X8{^~4{N0~e2jh~fopHq{^JeFGf%s8o|Ll{$vpeyUeQT$mIN`Z|{2THg
zC=RkaaYX#^T)nb$3VSv_y|p73+SmAzcFIuR2JKJE+Sy;_Ir19gnWu5_Ag}VAJpN>#
zwtnK4xQP$r$t&f*P@d~OzC0GX-*RuHte^bZy_5ONJE1&P-ii<X%wOD+FN#b0X%GFq
z8^2HS_e#ZM{>l&d6F=df{7bwLhuF1v?_A2C$g@u3qJ2qRG`@aNy?7;liR;SVcW7^1
zam)JiH~EeDCGX&W_=fT}_73@3i012l4WG`7ejnuC+kKt(^v5%P#Rc-|VI4v=pRm9A
zki##ae$aaHC;J$5kH=2ffj9}Z!#la+k@2B={wgkrTgv949Y5hm;tzju&f-U5+4(E<
zOMb?$eaAzN=gz(Skl%}U^kDzy8=gPuZyZQ3&&|&~#1&=jmCa9?{^Y4AH{=!X)`uO6
z&+JfsrtG=(;a}#hEPjjUc(cy@nH|};<lWji58#*o(3c<3k6nd&lc$|H#y{+z@@r-K
z;E7!NXh)vBM*ibIFZ4U{%5!;-{M&f)4E~QN>uEmnZ1XW+?d-4CUHlZEvd}uRFTcOF
zzj`lZKUHu4WViO`upK)J(R->al&{<W>~qTaRAy(+d;Hcqh4J0{Mf;NRj3fV4wr?pz
z_0axnT=rtW5P!v4?afDCD~}haojdHi&V}@}&)SFSBOh0mZ<D9Let7Vle$ENn)5AQ;
z#S8wmli!oCK9oma{rQjm$$n;E;y3()--;vrgkSLkJd?-IJr@_qC6E8g3w+PwI|Sc<
zxPO+%xv$V(JNbro%FfAnVz1T--|W}=D_aNs@yS2TTbz)0dG3Bn9CwZ~f9>%TBKhK=
zGJeTt$MQgB`kN>FF@OBJ7w|p+kDlWV|Kb39R?jZ=Cl{}t^9%lrS9U56iA(&N{_Ndz
za;y_SHGkuow|hHf>rFm=$Q6IY9X!*^dea~O`m+z~CvMqC*+CJze`hCtu^0YkpHwy;
z|Ay=%)WdWBYF~u>AKGWxXE-1AP(H$b*n{&7dtyiIgB@uf+K=|uH4E9L_Usz!$Bx8b
z>t{W!H~!@r@|fa!b^m^mpKA~KE4#L??1w#QN3QkN&$?+Jmgx)i*WS9(E3_l=Nj)Cu
zOAel#<M@|-&OIJKqE9$q<3am_yxuuOo+K~w-HYGbL;s$Ye~;1Mm-Jj2`uC;$dsF@{
zxPNDA+)BqCbMAJ%yZn7h^~%uStvu$SGZ%aC-T(XVdWG$@59`UN5B=!x`yk){xUZ34
z$}91PN9QtmCUib>Kjb{*d=%D`qn;e7zvs?1p*-h8sJ(Nfvggi4?jO`^@0@CUNRE3n
z=Pr3Zl;4P7&Tr6pfE;>2?Z~IEd|f>`#*<I+<1k+m2gG6hiuK*U>!UX`Z#-Ea_0Ac1
zVn6K3Js&^w{h05<_+w~q+LPzGdjfW8-RNT-Ld2u<6&|evKAqnnIr`&)J;Lxj?jArt
z<Edx=cw^Vbv)=lX&rU)lUwiWCr(Qe#&D;4N($jP8VJJs^7Mc&tY8Te~cQ57V?g62{
zy9a&W=X*c*x7z!A-0pXMU#p%RWq&W*_qZqgZsp_Vy>DvbJzRLspyyeS-7@fJX|IC3
zMLXYFJ9oIpl%M$C#yIr%`vL#X=s$k+^i~gj(5K)#M!!d(zwtdM2QTW)!*gZ4)7x`;
zY3KJ3c*mRHr5lfac=TL*^P`u3A@V=|uFS9bljr=5J@W&oo}92hf7DL=6Nj{S&*=Lg
zW$o>s+KZc>i(^oJ;{Heb@b{FyyA>xr7cccU4msi{bl!pTH2aBs&pAMz<{T#94pDwC
ze{;TYj&$xaj&qH?kiUc|f24;zoqTfCJEzIp*)@OVht3Q1cAj&-GoJefe!*_}1-n)j
z2YlD&yEJl)&u{sadhx-zM}ObP(bso!*1@^Ods_35Z--^`#hdkYzR@1f*44Sty0Rna
zoa8+1oaVW6lQMf?ht4hfhiH9`Ltp0vda*ZhJtyCD=W^#4^M%ec>=3{Dn+H9~*H77f
zBL3**xqA9Ihe3X69UwmOKtFb59j&)^?k(`;yhU&4F?y*t4|?l|AMGJt@ej2R<tT^a
zS#S3$;zoFn?VMe_C+YqjF8(}skLg^i-n!7+{H#MbzIx|TaWm|1KI)C9Ob*_(&sqm@
z50B>Oe#ZTuam<4~m`6Asd5~V_r`~ws{eyFc`(a35dO5dy&*J?L`=T#<AXk6+tN7u*
z$+?Pr`LsM%KA?;T>u!Ghn;zD~^Kf1GBR=V^EWVPfz4*Z|$!AB-aqOMlvTyUZ4tTb1
z?pNIJ^FMKho{+xE<j_Zb$cr)_*$Xsp^Duwwt&CUW;3F)PLq2;9`QYF9A&;J+T>j`@
zP8s6gc>D`b<kG|V)<t{rp#J1)5A|dJ=5HM9ragagKLCqzy8pe)@P5+$f_nq^1I8(~
z@Ba5sl-VPDCC9xkp2Kwr=dWEjKkLSh*$cms|JdjFH9NOYS|8&;c0@nxhDZLdJ+yAc
z_QAi0X1&~#i_iGdpWWk`JoW&^HStgUq&K^@F8YZx_%7PRWyf8%#oTWv@&I{^{MUKf
z`Ai-R<?HfaWq+U8cUH!=Zt`R6=RLUn!nwvd#<|8l1v?POt)KZjcgz3z2cG=BV&_r6
z8*#3d*LyBc_gp=c&pX$fm$>iTU>x@9y`%NP3m%<Q<ni);=NtB|zr2pUi~r8^@+j*-
zZ|jFI>+8HhUv|J=!v5;njq#mt%mYv2KVG!MxAByXM?YwOc((q|3+6{J^Ra%AeDd^z
z&Sly`c_@3W7xI%V`Of+7A>0eNCsD7z`wi%R#Pdqg_|W?T<EmG75APm@p2}f6?V-Ga
zJ<CJHQ}I=NP!<pDE1vTw{>z_?V}EfjXGhM_`te74Df16V9=n3Y_}snE7Weh;^NfAj
z{-kW568FtlImF^P-R}kL8{|Rl>|g4Y`JH<X_Z9qEJQH8E6W_8BAD$~`p?1(ZDMS6O
zpLMhUL3spZ2ka8M&*N{({K|gj-p+du??=M2dKljOd7tOIYR}2>KF06#>FJ)-`vT}a
zf_r)SoAs3sI#1~b#SLZmspQ$W<t@<u4YhZUGp;;G-o;PlRrY<zKAj7-lfUzO`GCAo
z{w4nuCycA!zQTT>^R<0n9wiRg57`s>Q2gS@^pQV_1M?IHUpnizW-n~`)`>H2|A#pX
z8yr9JgwqDhS)eC=$YIaPFJ^w|=<oge-;eNjM%>qj>*riv>H8dZ654_Fb&t<3yoa;i
z>;tj~?eUKnd4>7oiT?6t`@i>r@+tSW;stqlq#rx=J*<6(Jm+NR9`Tvp^q`mdhG@QF
zz5c$>HI8!FURgh=9vUx1^R#a4z`8*5!;><^k8&0okKM3G@x^-;?`h-<>cjXe-igQF
z+lY_eQ+toydCec+`sCG);=ilm-`9ZrR$kzKo4xR3c4NO~r|c545BsZj`Z*6dM~Z9u
zD?3k;6Yk^UpL+4s`kNQ~rI)<Ox!yj3PwPsacq6Wgx8#_YxTb7eW&Ol+@*)3WN9MsE
z`31X?H?t@Afch)5JAU9fzc(&D_?@zKfZ_gQJ;QavuYT4kYkl#8M|M+`-~D}=`yO)W
zQ*76rPijYwyipz?UgJl6H*fmT8yZ(0!48xmy{tcE=j;)ds`nh~N3Rg+$$m;99>hgu
z_QHP1QLpT|cq|`f7oooN!#8_ThW0Uet~|%{aNhg|(#QDnV*Ra`GW##uQ}_Lsb#{+s
z{`^n;#+x!ejRWz*&WtY}g?_GVzcO#~@h>it59w>Yl<jZ&^B3_>nO}<A;u)UQn+L?d
z^%ifSd9Z6`^U#idkR7lO@t&X4*Sg7vmE~3THTCQSZ^oxTyHch%9>o3dy@axRKKw%G
zF!?h5yyth`l!rTSlP4edoc}q;tEY#&-tPm9^ZA^EF1~2QEL~Y>-w5{?cF(@#FY*<B
zq90_xcoql3eM3C;oF7>SaouxvO^$l@V_bd}>aV|e0ObkpecUHG56j=&?;3}EJV4{)
zLx29Go&M%+K1F-&-XHmodVT`=hw)kk-T#X7{JK^2%UciUZgEEZ6i2KR^!<jo%dW*a
zd8#-je#qZFw||N2o*M^>Yhip6XS6d86lcUU^Hg?UuPoltgPz8R@;P}h<R|=}zlqc8
z&BK1AOfJ9TkD*-qk$6+|v+jJ8znPEy$b4bQkNrY^2(7E<c!B&mM0&F)_F$e6ug0-I
zl5ZaL5pT>p?9V^MSA4J&?{D05_|A);drxFvavo)G?2EqSvM=rQcMt9RA@)UY^Md>W
z-^RgL(NDYIGl%mdhrRJ*-#LbVcho(obGzR)_&cWlT~=k!-GAD5+?%?Gb?*1veXae<
zJ+=F7yjur(gZn@BYW?u+J(v3;_RfB+Pl)W*x>;Ae<5B(p$<Mm;aq9@JlYF|kUQ@<C
zHE{ha6ZReIqu%-(-}=*6KkdxhJtTX@7yE<Sd2b>fgy^~W$ZnPK3_XYRvJU*sx~a#P
zIH8O`@sPiZE6VtQcvFU9y)rZ|xsZQ(t_;I=VO;i{eVDhn7MAe|osZ<Z@-HYabB`cz
zgW{ocpgh?A;=X~LV*VSvx03hTx7;U+$Ks9s39sycAK3>y#|wVt75IhfwNtM@KFPHX
zS%_cl$+2GgnO|sk+S%WoH~2gKLL^5&WoSI(tB0XK&?5`O@y!#OKfR6XoMB$<k3Bm7
zE3-3psLZa|n{tTGDc+~CU*m@P6n*h!{^B=#<6q*Zcq;EC$3BT)_N+{Pi1^jsx@w0{
za`=b%MlbQ7{_Na3Wnpn1-TRSz(mcu4j$F^_LB9~~=g@ml`@6hB9w84<ubuwvLtfx_
z8qTfmlktaFaR<N3@;vqI6?(2LZt*8}<(zFi{m7N)D~EQ2PwU6-#5?QYd?_E4&xH0$
zo__92?62Zi_})`H`*gV9lMC6OJj%ZBe$xGpJWPJ({fs<Kz5(?YmyJ`5H{JG14}6)Q
zvi|Cop?0BM_9L(KeVOm0%*(v#=f24Or#Po9uCY_!)A(*Cv}?c5WdHm}SsqP}dmYcU
z7a!%%{Dyz|PC-BZC*Sv+p5%$Q%HlMA=x;p!ZCvL8<Am=4@g(n{KRwM)*}tn`JmcsG
zLp`)tFCKXh>37@CG1`%@-tWBWV?KBgr^4?;e5V#7|8UOW&*F!8&0fQ}BW}sx?91Yp
z{X|?M&wZ@C+wUHsdsgwt_ma+K?s<KOr{4FQ`a{q8m;T=S``rva-1n02_m93~@?E0e
znfm=B{pn#`@yYjz#y74qefuufb;8+S_AB^~);PXj_4`WSN&1eFe#WJ@?<~y^>Q8^a
zhc$2CLqhhQh2o1iq@B0{y|3rj>{frMp8q(fhjEbqvNQH)KV`S}PtTnv=u0lW#2@mV
zYr^*CB`zuJFRqyvUgVL+h2i&;=B;0d?oY%6<C1TDaRQ(Ch4NYZ2_!E>yh3@3xZwPz
ztevv_9D1%S{|V*L6Iz$BpZu0x7*~EKzjS^Rub_NXekxy*@5-<B7vI#2qssDl@e~>_
z)YE+ELB2fB{ek-yc?SKAhX?0md9-@x9s8?1n4Mb(?di#nl*z}J_3_+!NIPg<v}5n~
zXZk_wO+Vw%Q<**4m)H@MKbbE*eBWXo<_*OWW%Gj86+hNb9xY$AZrY1G=E48$ujZr7
zzwu(;^fzxj>u=u1G4Bwqhj|v~)BSxiJ7RbI2|xVBc|d;Qyuv@&E1p8Tr8lIf`FhV2
z&X>Qqcf>z*UU1%VZjf)dmsBrb@O`s<!uvS)l-k?(opanb`99P+!8wCJvk%Xmdz`1_
zYuY;>$fwA6FU*duhxuC<^Mu}q;{|{8q#rrf)%&;NxWV^G^d(0eBj0(+x|tXKJ*PK*
zv}1?nX&mz0<7&tLigw+7uc)5=vsZRWAIN?oy{wz_ZfGCg@5smGgY1cY`hHcO=l3u2
zN$bY${C<bsSr_-)?8o}yQ<)t3xBL(K&Y!;Y56AVMi9b1aTW@H5eBqD2%Iq#Ihy1cz
zJcejIc4PmA;+OT8FNibZq5RT*B%az&#AW-cyiWcrPn6Gz=j7-w-jf@m=f;6q`KJBB
zc-qMW=|^AtkUZG_fgff03V(s-V}Imd)-k;2ur5$M5%<Y)pF@swnSDVVv|f0^A0Ei1
zuYApZVxP9(nU8sh1LBN0z@G4AJo+gcABNBIYX4J~$B+-Lmvwb6F@N%vA^G@{@5@V+
z*)4nUoIOEy7}~MCjo)~#-ukdZ=NIFP-^SrT;<)v}k8_pw^fF#Je%M|=^Yxs6kcUU@
z$%S}__$F67`dep+7ya<3-hAX?{8BwR`mwjL-hAk3UgXiwbMlnQGat{{JAKJD9$w}B
z#<BkFOndeK@y`CWw@#ju=bXWQ*fn|jk)vKa`jV@?=jLHN{wU5Gmw&2vKKDEvSNm2$
z_qpyz&BOhv`$+eh>hXau^VHtGhWk+Cki)*%1KzZQ__to-8@ntQ**CjX4*O@-Yv(yR
z{LuXgKhsVe<5&DY>_<Mi+Uw8nLb=*Qaw<i6p8H36q5D*MnEOokj_xVln|ki|Y3@5c
z-}}WI-rj4Ez6JNK-(K>_yPg@>zaUSR2axaY66o)_-*>wQ^&ZA^_pQ+Vto+j7(R05l
zAC}iTFUr&9^Xlc<5P$M$`Gs?FF%RkfUPoEJA?`!@hx-L+U*$jI41WxfKl4+5B)*A5
z;;wjTU*n&~V}I;iJI~2yxBA;J*d>3lzp1y+@DukR@_2cYvUnhWz?1hh-Xr;Y!2WLV
z^~*o7#gBg8(RI{yyDz-<BfY!)z1~G{zvS-8$Mx%i&b{&$=NRWxd9HTyNb4`Z49nJ?
zeaKtYD|-%|gUwUD^No7_LzD-TqknPS?(aRxV~5s5+_DbV$9YP<JYQMAupK?{r=M~8
ziFufx=j!nrwr3yEIOeNN4m)z55Qph+K9C(cchVov&OiLxJ%IQpPKtBh<M@6NilfTT
z^W=E`^WA3u@VKk{7xrIf{g38)VnE^9p{MQG=ZG1K<%=%8c!{+R>|eOx<pUPIa?5@N
z<BN~pXNj-&1MjDd=lvGFp*Srr58nBEf8Y4teg*S-e#JX~dgi4w7A_z3)pFbZZH5AU
z{H__&@2S^T*?!#Mf6r&gTYv4^ea0Dz^1N@;j~>vt&^-lz=CAzOJ(c?l_Y)8=;=Owj
z_l54?{JzxwfLG`}s`IaV4fjFebN4>#-4D5!qObcR=j#yZ<9>(!_BHE_C-;)}HE6z&
ze&*vIPkrcb<_qa>T;H#_H+BE&p36L~oA)^4iLy8r#tm_TzWh@h5U1>K_BHW9zU7{Q
zzuUjm+b`{J&~tH2d=Y=_!^X8g+i&QPXK~1V414w64ql;sOnd$yj^fe&Ab#S>`1UvJ
zz%Ka_-nBCxyQ8Q3WaG0J<Kx$TH9s(a{j|fQxU7sP_J&XEPG54&S3A7mL%wJp><Pbk
zun*7^njbzxeT@?$zRAHaIiB+`b|9~p&x#MeTe2?v%e>hQ-te!zxJ9mU<lA^67cb(2
zb+sSk&-mJlui_s4>0uq%q50F>b8?((#I<mq^tG-~9HJLHq%XvaI7wf8;72?6p7aw3
ztsi~NM|>vFyz!?WJ?S6H(H=kMM_!2d&O+@&yA18!bM|Un@reJa*S~1@-TA6`EZ&RL
z{F-0!U;9446qoqFe#($ukUruS)E{c6A7r2SvOd-;l&2gbdHO-;Am>2)nS9>7tOLHq
zOW)bJM{^HhAH%zS4?5RrZy!|eJV*~{-9q%-K1na}oW9oCzA7JQFL-lqQ7>*P<3W4R
zq53dx8J}MA2xa$N&c)_we%gicp1%0SpY^m3&f(Sv@1EP&_=)>9`G$K?W%hwzaSjja
z$+eG*JK~Z!#m?;?;u!x3{e>NSe`_54STFe=eaLaoAm3u=`~`;bhd$Of3+ZRR{8v07
zNBq(rfArK&Je8l5hj;qh7qlneeAp3x5%;VozQjNK0lUVB@$C=vhxR#TdAfdhpg*}#
zzwo(v;N5e4(NjGnmtEkCJnfXBe&iPWckeI!n*Z@9_Ql`i>HLyE@+bZU#Y6JM1##0o
z8vo}n##PVn*im>t!C&YDL%$ODpy%R6F|KxhZ=uZZ#9Myqx&HJ}W{>#RkH4CSGI_=^
zfA-GK$tlVY{(b;E<u_q_a@h-B>EV6|PvW=rp&xzOk9PJi>u#Tb;=3}tBFDUx@oFCI
z*f`o7&$_ZB_2xyc`Ff5Qd~2_aSMfxg;pf_kOQAgc;LrSxuOEKAPx8LVJ+JTcy|2>F
zceLJDxfgeDt_<C;E4w$>pI<wtL;kNlKQ<q~n~<ORJ+yO^`z60K@!gQ`yWO+>^4jx1
zTW#K%3hu$&%Xxp|zR`W0_bl!UwR5khz4s{EyB~BP2<byV_ebvE4tQtZDVI%}q2PSu
zyyHCNJ0<5Y=cA$@b^pC<`2w^bX-5uy>46XTNYFXcIY8Ndqilb$&&gkmYyY&*ktZ(j
zNA@h9v1b^@Kk;1NWE}HWc7Aq_P?nF#5BMLy=7;7<Z}Cbz*N)!ommi7`;*I@GTrh9?
zinq?0>e&PS@g^SNQ`xxcjc>ig75TM&n|_{?Q;eV8@8|4K@&|c`xC!wmzRK&=TQ72q
zLq49hSC+TqTR%u|>tKDz)y_OUx1RK651xCkC;yNq&<EmEd+TDq(VtxNupabQ5BY<2
z)!(`q*S<ynaG&7^_91r1&b2oVJ|O#Jzs3puRD0{r4wbDdyEMLfe#=gj#h0+`-YdMX
z6PLtSank&q!$QBKKl^4^=8Y$I#g6gGU**r_use3huGtGaVh8p|Wp*I%v7hSCj>xrM
z?2tXN6ZPa8$2>zh?2a7$)ti@b%@<G31CSpnn?Jei3;)LP9IwW+ZcyH9y|gob{NdC0
zQ~XMPBcG8ESr_%@0mVW3;*;LyrJla#DPGXuI^a`(D6W`~^)xT@@SNVxRqC^lUC^Js
zLHA(tI^Rz~d|GGv;m7>M3I3#P|HZ3$vnzHF%@5K~9x6}u-3UF^JBQ+py_<*i!3R4u
z5BrJrG*A8^4<HA>?8-RG@_Oruck5^UmFZ7!_Ql@#0i?g@^fV8sy?I)PR>ko5!Lb|}
zhz#`V<vu_hwjYY0;u60WZ^R|%FmXeC6Q|`p?uWFO|EZV%$>+p<d7km@tKytIL4Igl
z<M?jT`JDdpRQpI67sXrqhJ2D-dA5B=S^nw!7Uy>7H}@CLozAc1>+e0Wet0oY`KtEv
zTjyH&r+a4kw7g%v`zU!Yy?h7bzCpclv^O5ToTJ?%7zg6n{-Hgjk9AkZyLqt>WqRl*
zuQy-%vuk#0KVt{%n0)r>90b{q{hI&SH~EJ=gB{to?faoV<k`RZBcz}4_@{G|`9R-O
zu{-E@o}Rm}Vt4rRo<*4*T7QUl>&)J?XHVp?Blmdj+kC%AFM7J4^ZPvCB`Amb8_#^{
z>7I&yklrEEmmJ@L8J~apt{aLs+MAbnC{D@a<&RKa#ZHu=xF_zg8};lDviA_pKMVOq
z7V=}?WB6X&eWUwQ_nq!ToxjCr>!rQ^>_;5(Jw5)7Cw{O)-$nZOP{?!7D&G2@g#Fl;
zoZGeM2fhb&pX<3e#V<p@wl3DIRy3~sT-@RJ*0ad_r;81_X7iU4eEEGh`*knE|NVZK
z9kYM)_gyOa{FR;i{Xi&}zx&;QeNun-?Y?hiFTQKNWY5Q!zJJ7k!l*y~^hd*X7*Np9
zJlKhSm3(@+w<p&<xcEX2J=ux)VP9c4#<L&uZ|7Ti1beVP{9OFDpNLQV%5&#8c4I%V
z?(z`#8}#58_CfY7j)*_v0zJcV)YIF3%<knO@&Iv<-{48yXTSKyt9Z)}d~YV7V3+nm
z^P!(O<-S0^!JpZqb)}cMp)7ARu6c+%zAvy&__Ge;G4x%uaqwk47}_nr<A>sovfm$t
zeCwy)`z!Agvho|eSx4(-AEqCD#S8P%-uh}M-t$NNtGCY9M}K~1T<?*>xXwR~$BwKY
z`TFw<D3A1984uP?dnjMv2ke-?spr>WKk=I#@hg0qSBUf@4~qNV%PT|r@EiT)kL-ef
zLHa>)1BUTL{IKtdgUbBn>@~Vi;Uk8P*lp-Co%45otFz;T5!;OD?EUqBht589;Lv$G
zdmTSw=zL$7m+S24?AUUrrG}3fdhtmcj$fi9p5X%q=9uj_-}_?eUWF_Sn#W>fpehEY
z&VR~~=QevfsT8XkEOv?vv?&9F7XHWTBhPp)X?NUu@mZHopMNRI8{2eQquP;yx-jsA
z6OSDHqm5oq^TxU;8-+#&dKd$F`Q5%VEw=hLZ@rPWJ03U7-3Q;f{>y2^p2b5CbK5wb
z$UtPE1_s)FpNjhTjC`ZiI@Vy~v1??YZ5ViDu0w8|?aq&rW@FnB78Q*QL<Xv6VEpcd
zYhOI!@+8En1;&n%fyh8)ATm%D1EtPEc|DK!x;^V&H?C(~&vtrWQ*Lm~E8ayp)5kqw
zv(~v>dyefQ1CfEqK&vv)DtTn9@)Pxn3`7PhXJFO7-#Ou#YoAWr6~p`SW|b>vKK7`E
zfxNy8?J08d&JV23k7hXhh=-><mo^*e6EQLn8HfzD3Io;dAI-WCi{r-o)}HV?m}tk%
zwc}Qazfq6>&Or2&=K4v!`2X*#9-l@AA_M=8fok*i*sr|5hw$IHZ{=rkeal<lp2AZr
z+hWu$G7uSv3`7PZ1CfEqKs^{}_dKp%_bZ{?dT<|QMh2!g26la9>ovDOdraEyIK8b~
zoNr?pm}AYsi!A-p$TVVQJnXc@g@=EB`15JI;jZ7g>o+U*dm(LCTyv414DUa9RN8LX
z?C%=G@f#au;%Je9#xl@u@geHl9&x>~R@`$(J8IH}hn#%dOKHT&Lp|((=j?LQY0I4O
z@heH*`1+WuPI~))#wCr$1-IVk@GGu;H>no!QZ25(@14h9F!GtS-SMQ;MxA!v=m(Q_
z#b*8fFt2f!-ulf0&Ya<IX}jP*zqon1$Bus>ZC6}po!=b4&RwI@cEwi9kE@NJS^e_<
zU1gLP{~lO7{XMXDx7T)mzuheTv*KZ_j|^19!1s53V(rroeJ!mOo8=F2+<$!+{<|F(
zIcS?_UP;TvYHX#JK0_9LYo8@HyfbYS)}m|d+jAM%`Ob9BJHJ0BjTm`|c2L?6dTu0$
zGm8vF1|kE^WnlMpPhDg4lP*Xj)|ZF723~c>kLP$ji5Pikq#ZOjx=cT#|6;!DSDUc>
z*tAw`b;jEDJLGoNyVdO>>K7S^3`7PZ1CfDdGEnZnQxN{HrrdJ5^6TBMTzzA^*7*DM
z?Y<7}{+&-x(Z89oHI5w_hzvvqdNKp?JFe;L?|a7eZp?c3<e2fF<`({aFy;T$Hr7Q3
zA_I|u$UtPESqxk}`&xG_Fzmx5V&tJG+d;eE!?vsb&5CPr-2cqLb)Oux+6(J-r9B=0
z^Xy|S2F~4ZsWtC<<L)G4d-JgM;<GNFKL1kEC>*r#KUN=k#&b!l;N^oxocLhJqe;Zb
z!`F5ozV%eB_1?PP-=~z4U#tGD`aXX^uPKY(^Vp+ltK*?}Jv8$Jw_TZ5i#^q^;=CdQ
zk%8%#fm-*acI`j?u6SiLpF003L!R60>7-q;-oHDojLq1(-5F@rJiJ`JN}W^U^Y5JS
z%(YLaArI|tw^9FZGEnU~uwMNd_3v7JlZ9A{3`7Q6iGj}^`(&1@_dO-46t~-co8NDE
zR9Dh0Z1?x72mNuqUr%2ApK0FMO12esi3~&rA_I|u$UtNuG7uSv3`7PZ1CfEqKx7~?
z5E+OJL<S-Qk%7oSWFRsS8E75@dHws<<;rRG@Avc4FYgsCm*1>*<-Rwqb^hg!Q)#=r
z`OK1c%&A=~ZE?z*X~b{xki}Q4)cZS75f@v1o44Lb@z5&kQ*S-)e)@`==lt2^Bw~Ge
zxMkuq4}W-n-vX@nx<>htfyh8)ATkgchzvvqA_M<oAf8XZc|L6yKBE0b2HKGU_X!cd
zdA}0!(T-~yb?*5LgmyZ;lzaZl#u-KiA_I|u`Z7?jzekX_-1@Q*rAG!L1CfEqKx7~?
z5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIf&3X5`r)+I7u)5Hw9%OVOrpekF!1BEXCFLq
z`0HuC@cX+yvG!?)zLqu{&t9XOy%EDk>^5|n&Tn-t(Ajaqh;2r6_Wt_6LuVg4aOgap
zy^bF-biS|4%XM~ic5J!RQo~0Kz4)XJ$1l;5JNFNMIBCHZR(&*SSDb0~nN~e>t=G~l
ztd+2r?isn?XXD1Dd1IDdgKr+T?lz0{Eo9+gC-k}dsx9A3^2RKEs;wWn;;zdcIPCnS
zTC8^5di6VflYLenx$Ts+Qq0ol@V}igcKey0O{>K$xp~)ry6EOlO<m!$v|d<B-z>kb
zv_5bDZM3{W`<-~!xrcp}R*L1?YpHc8_xZf9{AR|nhd!UQJ6=2Nw5xYHXF{4chCEgq
z9~*6dq296lO$M&NVu7b;+xEV+U9r-5npgYLiyyV&3-7+3l)_5$oo}vA{-rYxxo@E%
zKb)Mj8(#m+21gYxd@E@<*61JYzMr(J{^jO%rS`#gjeDWqdBrtyO7Yfi&%4gLYVz)v
zf0&lSYR@&*=0o~r<wg1L0N<3@uJ<RQ-l0Dzms-y(-b4BID$5VPxhl;ovBySX^Je<r
zM`k~b#&>4;+uS#w+`lUek6z`ugO9qgPgkRH@CBQ0eBL7cx~j!;_iOF4crW)Jp;h;%
zQt_zN^DJJ<tuJ+6$?N%l_pb7vma9+s67y}DSALwYR+URz^1?BP4>)pS60tHKFx_n4
z=us_}8$Zfz*Qoegu0C1qs^u&1ex>4mxzF>8M_F<je*e<wd!HKbd$R1T-u30$W4U%z
zX*>2(>YSuq7KZ2Za_1G>eUpK#b8Wrqzd3sIN{!lgN{!Pz2EI9RrGDJg+JBRc{7d!n
zr%-Mf7qiYqA8-A~JN94YlcW@uiqoZ@XN}*e`cnF}`tx?rqq6qzcBv2jt6j^Dir3{v
z&RW;MRNQLU=erF$euv8je4gZm-W&8Ztkr(X?O(1v$99!5P%1Aj_qq4bA%@yCYM<Zc
z^9TNP@g6f2vanjeF7+Kw*1bWk^A7XLkk{$8oW*xopJm5kyK>9b@?Gz|rQZ67^KMi=
zQ?GFv_1?T*<K%5m*0<fxJK?^gTrNN5o`b7>Z-m!+q4`Gq@4wS3#ZS5WbYAD0df7v{
za?71>z1oF()*2^Et#7T*>$MM&TW%aGw_V6@wdJh*snYst$FH_u-gaB99*yo->YUQ3
z=Z&7X@ftP1dXLl7?u+u8k9BS~mV1BIYVB&>XS4d3d;ZM3o&B@bG4K7T+45@5r<&cr
z+59zo{AS0`dW~P|yYqUDTQ9kl=Gm>%zSKI@`aHZBuU9#1zpJ&rUiLz6t@<^)e_r`w
zqvf=E9@1+1vDZp5?>sWf`{v)5sB}GArG2&c^sUzK;of_F_Ux-)Chdle=CR&(*KYi`
zn!fd3pHO~Y?`N7Nr`-GfR&CcT-i#aCL*C_;UfSgMGyZK_(yo~GcOToWzTe)1554`#
z84G!1-u6OXUgvW18inQN#mzofRLe)JeD6}V{%!g@NOskzd(e>IX5T;OHGVj+uw3pv
zTeJN)@A1o>cWn3H|L$vOPtAUxnOD3B=T&RD+V27KiUYOEnLO<$J9Ip9U0N+>eQz%>
zFLhoi*RI?zRLgg%e)P@4YVXDEca28h!R8(3$juw`dXKaF%olC>@=cGYd1Kc4Rl0xG
z>aSV-tL3j!{mX5i#aCYSdE0fZ@@m~jTBUzh{A>04w-*~X<G!c%E0l{zZQA$e>)hPC
z&}w+cs4F)e^_`guwPIGBsMgN1<Tbj!(l}RcysCG*yzYT|nw)CwF|U3L{9wxHEne)^
zRW9ar4#`_irGC*W?eoe9nk}c)Ij3EoXPtArUp(d)qaXgDb83=>+Ch24V=ta{{vw-o
z6td9yE(^;&Cx+ks&?5`acxS~Wb{^fgPztm11b+{=T5J_RtW=LywdZo@QLntEQM}Zu
ze^$KA^8a?J&wF3Zl2<GLrTUi}@2YK=HLtwtv+OIYzS4P=%GV#ecdsk=+N5_^DJ(Z$
zgzZX=A3m=%Z?E+}qLh8r`+2478QbUk@7sm;*Q|22c2=psv|9VT@5jrPm)E(tl$=K8
zIeCo}o(sw?=QVy_a;o*GO8Z&=YX5Fl*sol^TfJRY{?h98mF9KUzfo9gzEf%cM$Id0
zoO1oKUf-#eJHB?WpRxX?or7LUN@1mbP-|Y2HD1<!X+D)=*7?`?S@~<$eOFdJd0AMg
zA8Vhbe^z~^{M5RRrTSOu-{E~-*8QLQto6&P&uc#q<yfyuF>4*O>a*sRRiEWgVSKAs
zylRwO_to{nkgrxLSL$D_);`NVTcy5U>s)D^2;0|7ue{`D#jkSf!}!|ha;x1-RH{#2
zejeVRHmZF3Ci|>Da@#3s#OCm@>u1X!uzKGw(^6QeT{fzHmfm^UQ&wIR#%be?e0t{D
z_j+k++OAmXxgx85rTa@>c|fJ(X0@;Odx6+*^QWe+@L5_b4|(}d);V2$*7q#Us%MX-
zeh<;ppW`7ftk-$5QF0rd7uS2dtb2@BtuJ-|Rqpe=?3tWv7drd(C%2lb&@NbSyULQ^
ztax7QxLNsU$@&U^e_n3BUhR3al>Ab1v+~SRa(n7?`AfTDxp6YKJ7etjGd-I|9{xKI
z<>m|a%SzGvr&_Vnahuh?*59?o{^k8XFz@wkw|yYgH?KGs@>=eDh;rN2n$KnRZ`OBJ
z;XP(gDL3l<MWy4oC+;bjm6x|$eHf2gr5whyo?gy+574N2H|jpGUgL!K0=4p1?(eQE
zZI{=6U#*;0@$Vk5$1@uoRk-l2B(%$RDYtt*RO@@Vy!vP5y`}1#l^2KOX2sFG^PEcK
zS*zudmFB~>wl8&WT<P<?*RxUb>TL%hkEPzv|6=?*k3TT$8)>yzsUPIszTEqOdbi7p
zJLT4g>sD*IR)5vMT>YzUSMI%Vx$UaeBd>nt{@t{^+U2EptINrHU)<{T<=SnlwySsC
zthfAj{dXopy-Vfat@^yDeE(Ljd!TCN)*7!{^}K9etr`!@)vMM#x7=~+-LBR5*DU>N
z-B;LQy)nGMh~++aU$y0XNx3}KdwuHdU-g#X<L|92{C;wJ<)hwyQmg!W`$4Vxh5hTT
zZ<hRe=lNOtX}i{ke?O_#a=rO)mi)ZpV(2HzrT%_ov!2)cK0eC-@4uJM`i`X8_3XA%
ztk=0Z)FUfD4*fVh2ZjBry?+S%h2z#*4)w0K9P-vG<?w#5UgdiGQC6PPtoppyx7K@{
zuz%kAm6DfNJSsJg=XrgHl9wFsG4n?AtrhcHpGM1Rm+!(d^{msqL9OScOnn>meO{e-
zsWl&HRR2bO2bA|X_1@>R<X0R2O3lC4?_WywFO^T0^5atWSZaQy#`nC`IN|eBc~{sz
zESKtE?(<M?SgzJ?D(#2Ih`TO(;IQ+Pp2oxH*DW>hwlQ-SN@1<PcMZ?GrOJbE9=7f_
zi}fv(ix-@>{R_u`+P{#6rM?#{cbrnck1ll%&H8)G>PzX7*Y_aC3$fDpkazp6eX`x^
z!}C(!JdxM*e7DTJn!feMMxW#I8n0BollSv_uS2usH|zb+^geE>e81J7w|ia}o)6l!
z9NzbqTMpw*`0p+>yPQ{i2<O?Ta+qh-d)!iaaMts3^(xn{V!QwTcd7EuOByAw)bIKl
z^}O0R9iA`Rtz0X=VLWP;a=U!5QtJI+xpA~r+I6mQ$H?D2{B%+-)@m=|cOZG~?~Rhv
zZ|}i}-u~o_g;JRJ`sQ_BuD6_8zo*NJZ+Y{XC9l-|s^_hKE~`|(&@aPs7GJGW-&6mN
zZ>{rcmGvp5N4f9!!gg75FFts=v5!p|&{Zpz8_xz`u<6F<Ez+;6-kA08*)*#@i~myf
z;X1Zkxzc;DQhaz`>-!pVvhuWAc`i492*;sUqp(r=U>s-6;RBAGnBbw>b9wByz|*sB
zdtchBJha<ANl(?cQSmdcaq{}^B9!AC)@aO&ms;i2`g_7g_0P&f+NC}`2iICIciru$
zm13iLEq6YZwyQM1ZngIHejigQpQY}tTm5;~y<olS!}CkI<*aon_uN#e9as9EwOl>2
z+Ese57`88$N9`)*L%&L~(l}dg`%3kww0)Mn^|bnG<0&30wWC_kVU_x2tF&+Tb5o`M
z)2i)jy$5WS{<Zp1v-;Qi-lbXn!*g<8WpT7#80HDpmb3DfO6wbKclF|x+*XZq)0<wc
zvd=e45B8UZrG76PpPzWvxrcp}){BQ$Klhi?uh#pxto~W{s=iV8r}ZAE)bBD|{duGB
z|MMO%%l=xep4~SJ-Rp(l1y>peE8XWR%@a$tuT<YgwJ-NRq*DIct$nHYC3)L>qt6wU
zj+gg*v*fi}e5+KSMx9%-_zUYx#i4e8-YC9%$~e`ocTbsrPn}1*?XRt(Zx}DDEm!(}
zq*d<W+f|QNv7buyC^x^Zv|Y7!+phh>_r7PZ(LIw9!$#~jbeYZtIv4EhIAO##BRYG3
z{okRp4;?skp3Yv!j~F`N*X89pJ32eI+-a%dBZgjl(uU)g=xA3SqTXdPu=XnZZNBS-
z=h8-Fnd&vX+01W0Ip*?pN2SfiS9@LB^^d7z(n_)6gv9|P1CfEqKx7~?5E+OJL<S-Q
zk%7oSWFRsS8Hfx-1|kEIfyh8)ATrQG3_Seu)3fY<<!foPu@ohh-v3W8?Mv|%pGO8F
z1Em>QY?h19>3!C-X(_BUPL>`ywulTw1|kEIfyh8)ATkgchz!)9flBk(*uFLHNvuQt
zEhLW6lNjjy;SF!>f55;(h*NKPzw5P)XDC#Q)pE4eHbbVS8@-t3h1Kd&ZofU=T<*w0
zeeO>4!g9xr?IHt_fyh8)ATrPl2IQNK!e-2|mB-90KZ<hxm2WnCt;6?A&DJN59~p=Y
zL<S-QJ&S>#4LosyFP9&iMvOc}J1E}|dR7#Ovxy8u1|kEIf$4>TTMl{ojeoBFRuZv3
zJWMZk6z3TkhzvvqA_I|uau}HWK<`yH9`j@xvAH~~e)pD#|KPX>lf1DU8;ETp1CfEq
zKszw-+iCy!?d-choJNd1RAUG2VB=B8+8J13&nYKeGxM8CUO4nyTl{j{LnbG1&bM#f
zVfX<Pk}Ujga@o|sE%<CwE^c=HJtOZs=E1blxaTtCuV3Pf;Yqt;-rvPt|IF?;yn4tR
zX{+KNZaDj^qZWBRZB@Ks(>0I()tJ}PcEw*kx$NqvANqP4@tZu<9*Lub$UtNuG7uSv
z3`7PZ1Fglttuy}en8%iwt#IW^mrvezyIC`FmBn{_{_$xu6&C3~Y{RYAo~wvM4%lF!
ztG_j0Aq(HS=A1p(*lNB)yJ2gsQq(3g5E+OJv>yXEZMy6|L%*7w)C=d{aJkW2-t<P&
zE_mn0`%L-AT#u#giral~`(775{dyX)IXvuf@QgPua=}Y!#OaHNMn(R^e*fH_hfjVb
zjo34JXp|kraUuhefyh8)ATkgchz#^}270eJZc6`|FG^YsXW4YU<xalttu$hNc<5<%
zec-GIJ)ExmS`zWU^RUb9t1i3EjrXMyBM;FIq925Dpr`E%aXyiO>4||x{oUZa#+jbh
zCeAf75E+OJL<aI_;NvBKF>Hk&j!VnMqt^Pt<2xMtW?BlvzaL(1dF@s9+kDpv&!t&d
z?eE|!{eGpV{JY$x)~QvW=anZ#Ij!~Y!L|B&<R4+9#K=HoATkgchzvvqA_I|u$UtPE
zTn2t|{x9y?`=~3^h>?eA2hk5&Jq~2WwXFNkYVR{+zkL5)_Q+>Ge1`YzrT$&FR=*Ey
z7yYv0POOg%L<Xi`1|I#{$HN}l{*xqP<e?TjSZB}|b3Jn1fI`H`!`F7u(<5rud9B>~
zp3YC4Ut}OM5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)pdJjg>)%n%Tkm>s8)ZfY
zA_I|u$UtNuG7uSv3`7PZ1CfEqKx7~?5E+OJL<S-Qk%7oSWFRsS8HfzDF9W51ml>a*
zxWJdok4>xPp?xhQY91Mg3`7PZ1CfEqKx7~?kS7D--<he`-xtYK=l{Q~tiQ`%ZvFpX
zR{a0SKx7~?5E+OJL<S-Qk%7oSWFRsS8R*q3JSUbam;3K`#dekbduQ#!SM-_k7?}C(
zC&ygA?x?g}>?!{aT%5<HUH_OmCXGBq9-<vo?+23~=)KCuW1dV~9n0GdVyi3$_IPu-
zBM0@lJB=85$ioh@>@3zt1|kEIfyh8)ATkgchzvvqA_I|u$UtNuG7uS<-WjO&_h)(c
zYxX+C_cq^rZ`17cX!iJLuhFf+h+!jk8@f#Af}IO>cAPL`n-QJ8zy9yg*@q4sI!|Y>
z<3|jg@9Xk%ogJMWTkf>f@DW2VK54`8OLXL}|CGOed-&kb$0j`$b0;hcEQ^8ulXpM&
zkH_{eREuQ|(!fSbTzJeGqj#B*R*G3F{(ObSkNxsbcO}(gmfXDSryg|0^C#~)F{u~k
zt#6dqTn6rWd*q3)oHZqBckDgq%KLx5<)=yBnAhrTw9XmZ&O5{BNuw|?y?d&hV|Sjk
z&g%DmkwmPFhedmR{`p>qj!jG94R<a%=&18<Ps_!gYG>2yyh_<ed|sX(g*;8K_7vwG
z8OW1?VLKmu$JFCrPud-`&ZBv9H9g4s<AhCiobB27(?;R+z-cSZaib&Fd+U4KcN8Kv
zhljq`-rDD9gQq1u9b3uf>#0lM7gu_1{kvaCYsGr%7-iR=ft#+swciaLlhS5ktuy@K
zjvxGR)&Y+v?S?zPy}*OZJo{QwE!N6g>>n9uCkD=*@W{(A{(f?rH?|W8QP-Z!z@=N=
z{lG=<zLho`doIIqW|4u&Kx7~?5E+OJL<S-Qk%8th5Pla>uX6Kd7)Oo_L<S-Qk%7oS
z;}|&f%dft-|Ni5XcE!BEf3H?vxhs~}_gCePH@&o*>%x9>UB1-Bbb7_@?%(^<yI+}<
z=7rVTO;5?oFTQAtOSXFB<uos>md{4_J8jyv^X&KOhiS91(epOm{_EcR-Z`g#l(svz
zYaG01!tT8m-eFQ&FSO2i;e|J>wdyK;Ur$S6v)@zZHGaF=RhHg&+&}*vhg>*5ZC9LO
z+2zikdfhu|yJ5ZZght8z+uT>3z08PL(&-iRil40}XTyWKp4@7w*ODx3HGNCzGojZ{
zjvD>(t7*GoR=ke&|2pTETK8I?kC<!Dk*kk*D{WVNXxam}jQMzc+HQF6^?%sqvfsR)
zMl6qqYVEw--|4q{yZd)rZ?S`B9h0^iR=Y0E?svu^C$4qh)vu-X#!Al{rOub(^Vbhv
z?CQT?`F7fF_?LIbEPwt_UQZi^^|rS^toxHWUijX`G-7o;)O&qG`FY)wG)vBQ{g(c6
zgR`zm+Z7x2_oA}K3GE@4BLl6#z~Wy%GV?nV-c4%7R+!6w(xKUZ-}OJ~R8HNdJ8iW1
z5mTQ_nvLb?`k&gg+PR@peeS#Gxbuhq{*9#G_@B6ob^m7I+lOrT;rfHdrR|O<zwz$T
zf8A(g($lco-_6(B@81^kumAgMwJrJ`Hs~)&v+&F<{`};M?@UZ;#c*D+92tlVL<S-Q
zk%7oSuU;!1@!_-A4eMQq*fV*EHXa#>3`7PZ1CfEqK<y0F`n#}Zy>F~_+&BOJ@!k`^
zeN@t}xW=Fz);e<SXVP}X@Vk+AE4TXhS*7&bdH&0jMK5|ZZCA|td;HC+uQVQ&`g_iY
z?f1#3WbrgDh3jr{@F{=VdrDF%KK|uu=S=$Wjig+hY4JT~UF*2L({i!*@9!MF=QA%Q
z5hD-P*umzXkG%E0H(yC2PG3CKp3#;kq0)HW?DmcNeOXT#hZSUfZ&zvkXPe)<;FV)0
zBs~S&?cX7-^zWls|3+b>&tZ+4SE+ICeB#o5ce>&qNyNtR(5UlTrT?Dy^wz%F`9Q7X
zx7&S7sBf+F*S}PJi_hDGfkwr}dW}<QKHBK^?O_E`%g8`vATkgchzvvqA_I|u$UtNu
zG7uSv3`7PZ1OH}V!dx5AaofCaq~+o!H*`$+%hBV~QdsTZZ#sSNCr0c%-vep0aj$u1
zUu=~>Je#&Vp1j@nzkkdum!<87<B#5W#gzwlCGCzYU-m}tmw$Ro(ymzQ`^k2BzG>Ir
zPnm7rr;<|m@5mVcFEY^c7-;u+I`(((zwp!5dlzcOMy>2ho9}YY;60yA+8wXk`P3bU
z9s60*)9~th`#jk9{7Ffx;p*3ZZ<%X8`7-G#`1$XKOqe#XqtNa+!|-3sJ8;OmNvmO_
z;#sSVGwPt3Z{6&VA0;8Sihk3Z9^-#`>M4(2IyQ+Id5Crp{a|{H16gr@g&|jc{PdG!
z)4cG1{`%AxD}DT4QZ7C;X3^`{z5co+V&tKoc93QNu|6`;iVWm^j%)TD8~yu}X3w`#
z<3IAquU9(!{2P;Y!$#Xf-s82)`=u;B>-9T{M$656T^l8@(Z5HXUdF36?zZaRsW-0+
zI_{t8)z3$#)#6e2?sNIQFHBCFg|+@YvsUR}>wCUt_21|4A?x-(eU?JC$POb;-wfRH
zt3`U<cJ*^<#OCs_&yDBaa?#N*r4g&+Vfyys{AWM)rF*8WeBa|~qcQ)Pd{g2kC;k1a
zZH7-udkXHd>`z`hX7xAIyzrLc&t0-W;f(}taq-Q||8V>>X}e>o?=s7M{!MFY|5B^}
zUEV{#{OWuA?>|22>DcJ|mTE2cm{azdXTav8)4Z`<eX{Ps+O0mX_X6eemeuZ|X%E~o
z=Hu~c-uT%vS1&yBpbyh}<88~HvGjmjUQg?Vt@itzTK|36u>UTD=Q;Y#2cJsv!Yn(h
zSN+-#j(PO`MgN)BiuKa3-g4jCebhU9&H6%`H|C}H2^%iA+e@F!SZFu=@tnKebi|V%
zCGCp0cV4yRm?ghT^2YG*!&O>tbe@yfcs=F4%Ul=so9ptWCZ_F%&HDES@`{7y%ISOU
zt$ltrcv{l#c>nP27TM}IUnT8^{U`5!?jMisU#Jz!UGG}k<rUu=EvMc8zIUVketWI<
z5&KVH4AgoKD%HQ$?+eS-FU%+EWhL$M_y6lP|4?qD_nExLE7z{_Y8Ud>D&;Z9z0kSH
zkY|%>al8BX{`BrwCZ+9)f1I$%j<Y@ce%fx>Zs)yF-@N>|S#nD4tL^f<()%9m!}!&x
zb77;;(cwHBRSw^O)qC7(zpt*hUU|!(^4D(<AN={)q+PLD@ost_ciiH;y?)w~FD31Y
z;rXcD%I*H|WR>DO?>JX)dHx-sM&YvqSG;e^HdE4e#d6O_d9PcTM^svFmpnF0PkiPb
zx5=#(+x?zA)W6;1+Vq~+WX-=)yKYwdQudeiyjl9jacg6s(f@v1-s2tihv9F&{DTjY
za<N>tYJGoM?l`esWS|}lxL1t0)~c)YeLamlL>}_AgL?REJ!R(g`?pq+Q%|lck)8KB
zxL)#l%J;c>&7<5kuJ%24qwevHQ|^2!ZP)I<D?E1RN$aeB?-xnEuu?u_`)@Mvy`ke?
z{A%P9g}m{bqtw4NX`N4BUS!sZNyO=shx%Js&lusb{XQ9$ES{zjBM;FIq94RKP+1)4
z8N0yrJBzIMgR%bdWYLQrO&i5SyZr7XOV3JwUr@^9%6)$p+g0|yG4fR#U#09eK93AU
z2I|K^rQbKk_Dc>r>b%?2$V0w7)X&1B<aTBtulIqa<g_zKQTO&^ApCuAEVtI*<*N3(
zfa$$o`*~v2I5H3!2pKqgjqaKe!$#~jbeYbDIv4KjIAO##BRYG3{okRp4;?skp3Yv!
zj~F`N*X89pJ32eI+-a%dBZgjl(uU)g=;-yMmnJ>)?ZU^&p$pA_^UE{8lWhHR|6T9g
z{mo+enqvlR`_e2=BuBpa+a>q@^X=)G6W?BY%2M~I+CO!~ZA(0T#EZ$LI}RE!?WXsN
z{r8)_aL;n<+?dQT{NDaQ+_x+F$yTe~e&$oJCudIGcaM|bdM)|+)$i>%;G*}7NIqQV
zqK{`i|Bf*Uykos1ZoB&Qu?bvhk9%j|ck4a{IQWYjuAlwH50Vf+d-sabiybp5DTQNy
zbNWn=tvx<D;@$Jl*!jbb0$xISb6>LXGMDszHaUH#?=RZ-i&u;D!^f+0&(N{-CzFz+
zmcFC+EZ_bxnfrl_u32c-NkznGi25(F;%$RATz*PX-x;?1)^?j*JSADL@AlKyyXE8L
z=mq;Ox8wLv6Z&m_@R2{9d&!AOXc%F6wZDCDrLo_CD}{ebu6*M9(W4XV2X7s9*0uM|
zd1Z>{(d%wF*HTMQOwU=f^Nz)Ld@0@GlU<MM_xU?TWDoGcAHFbp^3rdo<g4HGoae7v
z@UBynpTGF%oojwJDINC0t*hSn?Ca?)6Ar(7!66@}?8Wm}zrFGamv+3CLV8~~|HH?Q
zI(%Gl9!0(Sj=Sa$cRrMa;~Cey%#(h`E#k-D9@^`s=hGv5o&B5BR~}Wg*RVYA9j89^
z^E*FJX1(*0o!)<SLbB(U3$HZCwI3$Q=NB$@+8$Gjeqdd#8(!HJIrJol-NPNa=cgU}
zS>?jlkDmYc6N|i>r}@2i*kMO(KKCa@Uht_M|GfV@KU-nYz8@v*@8%1h8}h^c9R>Or
zhrQ89z5eEf7xO*-v<+vu`PbtTXk2kXJN?+VxZ*j!=p|l3@ngYX&3ErzyFHcC(|oNX
zUd5x&?|S&)kCu5Q!K-%W<2hdSo3P6)qmJ0&ss9rPzI*F(^SnCKv&HerO;6ix@1LJ|
zdtx0|8oKY{Oa1u9VmvMO>-Mw1oHXg5f1dDKI?HtnueQU&Po(_u*ue*lTIRC%(>H!G
z%PZ>~JTfI8;ujxqtKn-ex5}v>q%ibn{<L4W9~g)K@h^O4p*Sh-TOV=OIG3E<f9Q&{
zPf9;H=bb_GKK^bo9vIL3@F2d7dbrOjBhMO}nit+e#6yVIKh*!Xr>=a|YCo8qKz0O;
z2iXTK#*yy*kNrY+1dRjPRTkcI=3={EdCQCiD4yZPdWAT{uag(={qq|sq(3{PH@>aE
zcKlEL4UxZw`z*O3@^kS+oXo;ue|h1F<No)nBn$Cd3i(GChWlnLw-*EKm0gS1{GT22
zFL8j~zO%(#Z*8?uufnk#O_}!hJw_JyaV(s=?d-E1KWqPj{Z78eZuMhdgO?aG_#aPP
zk%(X79Dfy8>}TZ37sz>SrtdE^?8+~aFus|O_yfh4EW}6NC?4_aYSDg?h1N~H+ULgS
z$3F9$uK)KB6aEXWJ7gEo`onM>`y7-HK<iQp@hcC>LV7{=n}zg(VLN&@8tv=s-F{<T
zAiLdr=&8^DbMH4(`^od$-*EQ_uZ&KYe`eGktFQ7(DnH^!#dAY<J|v%I2lAt}mOp*Y
z_1^fXcut^){oej|-Mp9l{PDq~Qh6yq<j;83-uybwzIV<!Pk6C-9uDXG|FL)9@pe>I
z|M-*8LT@4}y$T|VB1I7AidgVI2u9%%1cA^3f+9t#h#(ylG1vf65R@WS5U|V@g3<&O
zkbn?@P(lqM0TM#_{dhfjUtjL8<C$~L%$%8f_doe$&+NVS+Mm7F+H3EbGnfASPd=o7
z<8t5nsej0yAbntHw_zF2kp5Z7{-F4jg~kK<TZr`Mx6rtu>(F)akUv580QvJE&N;8(
zC-f3W@tB47-68J(>r*dW@75WE`+EBrd5wHYzI(}SpWX4Em7W~jPue#)M}YKZ7rQUt
zciSUl9vIxGlLzlzaJMU7y7`#|ntx?TAM)@E@hP6!r?}4l%riduAH*;FVmIWltNmVD
z>ehAcU!d<dw<dRudHv<&!tIti?C?`xP5Qrd)8V`PZPws9Psl5MjfcME@LO>fpZr&S
z@q%9Xo_Xc5yB_etx%=p?KlQEK?1vumZ9JF<a^(f$!sgp=e8Q{=rwqn7_8~vG_VG)O
zSpJ%4(^uzP>ZoTno|10z&-Wbi#|tMW#>HOPQC8k-KH9}grycUxXWso%YJbMR>=(sr
zJcNiR<ALNTLvkVhv(S9Nu>W?;>>vxp>+9B;ed5x8`Sakpl=&3@>HFa|4?E_%^pRAV
zKGq?~kCgG@T#Ubj*ZJ49B_5ix!e!5={L*!Ht9=-M*%5sE+P7XcVT&gQ^Ih%vQ)tiZ
zmtBdU`XSeS^4y2`#)EP48|Yjrv~&4x7Ru|Pbr4!#pnWxz=i7HWe}ecmUi!i?|HUU1
z$D!+aF_e#27~bd4ZyfU3-9Pr18H498uD@f<w2eRV`5EcWH!uC$Eq?RjAbxSrpYHm_
zlU`2Y8rSZ6?d}K8Ovk_dfkn6Z*-PoMSByJtoo~G`crMJ3pMUn+vFDyQIoWjY2Sz{s
z;!}xnI!Cmw;?q1^@5O)T=-NT^qaHi3KC^52)aW_q>B`pYETp&iO@HV6{K7h*Kk=FU
z%fs2Zcpx5T#WB~#5q{2o`H%HY{;ZvN;Ji#+QeQkH&pQ9U8}Hio;I~Xj=xsiY*L8dt
zC;i-KfBc_cunT@^U6v=CpAhv!Uv?g%_41F$pTGJ&w~pvDKk@<h%>%y7oBe>eVBVZ(
zvJ2zluVH+`d+2ZCySxGNFK8e0`_Df+_MpvQNy)L#F)!9*>#pZ0%1}MX4&eNs8hgm;
zpME2e#~3$sF3TS22dx88zYt%okUs30pRnf;>8oFMXq}SJLVjc3*@1Q5Is~on+KDsl
zO8=p~DjUCXv0v*4#IO413E%8Sf7TuRkT0JwZ^q%i>)MeA_17-yHw($nLVl2i_{c)M
zL3V&|<A?Yx3!QI>H^wDzGY<B}9`U}~Csw~_;YXiNjVI(yee)Y4yVegLT-Oi(#REHs
zp<L&okYBN5aXLixTcNxH(oejTpB(qg%YXBYb!VsIU0w`%G;VQMeq+Am`Qo`aBVN2P
zdd)??vg545d0OU^8|{4Re6OYS)}MCrCF7&FviR;gG=6%i$B*&BZp<(HVK4Mx58{aV
z;y2oxSM6O7(KzX&zU%BzIYjHHJWzY_p4~$8W?sZKc}9q>?5?W&pP6~!LdP%nYWmUd
zj{45n5i?Ws={z#zb<D>9v+ZM7j~qDX$SqGf><|BM9`N`NUcc&x(_Tw|JND<BtnuS%
zgU|WoYoWjB&-|Mg=steggScrw<eWf$!XCqT!LFg_o$@PrBKu>X@<Q{-PU%fA=gRU<
z_R3DgVdIm38mGL9J*ualEM({5c=;*6#lP{pFV5)K`fvTjKRqFTR!=|9rRDK>!wb8@
z8-C@l{7s&0zKtJG^uV)r@^Jm&QQ5fVb@DUgF@E#L@9EFK^%ssCUtxXoq#wMILofSa
zc1#aQj`6t8ZXw>o{E%LFvhTG1<JZ25|C%3h5$~bi)*JWvk$L1l`nS)ar}5xjo=zY6
z0N&V-_Uen{#uw(x>eGXK*Y!hR{whv{yx^PO?9=(FdB+2P<1hHPKZfe_Z|$r%<nm*2
zoxihB_GEtO$zH{E@j?Eu)u^?f*zn8K(?dS;;f=O?>&xl-D^6Vh<kMaqJa@5wcK+@B
z!Fh$e)4tX?_0K=$3+#^{7&p5&{?Ja0!#K5v#`WmWt~zz|O<x?$*T~0{`t-uDGWq6L
zzx3h{?&I6IAimkJ_(spmp16L2pZ(+YR9+`unD<a$;}D<pOCS2-P5gl3F=YSz1Bx$M
zs2+cSu8Tk7JOAW={EGkbH+H5h4tZ`9%H!|WdGhs_g{|^0<H$n17)KW38UG=&i*7NL
ztIYqPcwH2lkE)O#vUkXD?VpUF9oj#MOZ?mXTED~v`7L?!UU5Rd&M%GM`LK1wb^W~W
zd+++i9-nw689nEGm>&FA{Db^IFVYWMKU*O`5y#o1^KpI!jYm9#+F763zx(oM>n+~p
zk<N+bXL!)w{=b#~vXhXvu$?>rKjJntFXYin{I+hhKXKf==0(p}@q!O=mHpC_eT#el
zkAKg3ubiLLlN>xl`3br9ht?T>fDd^KJHRWy;UDB#kLYhcUAK-wcH}<(<ypql3a!r&
zPa)D1dY`>rly`W(g+Kk{2U^de>+%BQ$9vt;J{|I3aaUY|r~cxswO(8Fsljsue#GAR
zuln}a{A}Te|1^28r)H-5gXY0G8-M1v{L(zJFQ{MhCNEQVuEw6#7a#P;|HMId-7WIB
zEX>MBTyNz!;!YN_%PeH~S%}A0D8FkLv-A$H^Y>OLF1L#z&-_;$lW(v`=k>0$Pi4QS
zVAt%O-LWfqz4cX`eqg7~UcGah5q;tx|AgW;yC%mx+VAPd{IhTT7&l(&!Tz+f{>szj
z)$$ShQu=rfPo6jc@oGHyme<Hj<Y)G)<Qtzn#kmo5UmoH;H|rt!)>+SQ?PHABe1!R;
z_p`i*;Q7COk@pSY1-st1_1X_lAN<{;dBLN8*u6Z%{g8M1Sogdy@%D3XPcOdYt$p@o
zo;S)L^y|5`^B(o_f+u!?KhNXYnR6(7vBMD65A997)(?Ak9a?X!zwE_za;-<+mlL=6
zoqUb`>lf0K-COs(Z@>FTcOSFk4zm(`JNITc@;vqJoB4xt8TtB;CVX)BFB~?q5AUA0
zhw|{No_S<9>`vZqK0o=PogUlc{!xAO@OwmdqaXSx@A8({uiIs}QGMEzN6sAw9Ju&r
zzdNGO`29`-;+LO@6YSRS8qS>lv)8uSf9BvmYxJD`){XDKa;t-vc{SPYZyUdI{M6S5
z-v@o}s}s)s(GCCq`v`M>SL=BJ-n}nsJkWi==QTdRYoxcb-w{LC@$vMg+b{H!TSxb8
zbM9SZPkDLXzE8dXf(3qa!h(HoIclFB)BP73ysqAzd;MYkbNc4*gYVsL!S#Oo`htBA
z{$t9w558pXKKhWq()1tv=Jbc>>(ej0U_Zw3)x*BJ;R#30+vh!1^Q#?nE@R%bXQ%pQ
zkL-Xv_RfykpLk{6XQv^uTlQld+KX@Omp|Hn@OSZnfAW8ELOstDp!jD!a8BTSHbm=;
z^~*j>T%#ws_;BAk1g&%8Exz1$J$yggxrBAf`UvH%<j|Ww{7`*-;uoK;lcQ|?5jVvf
zyvReuP50Huk2ruI^64ipF<<m@-FWC@JjP+2Gr#2OkDWOOl2<_T%!7PGJM)5H<F?;n
zckI%71hq3?t{WfT@MyiVKayYBS6P?*4w#?wJN~3!`r=Ff><iz<OAqz%X?*l1-+bab
zL_Ar)2lbfqyIAd^_UugF4ebx@1L=uh`m-1IqhI{EE<WLjf6~YN+CLfxKJ-hj_#nQr
zFL{)BrM>a!2hvB`I&Yqg$NDTEBOed=gyQfo*80`tBY!wg-@MDb^6-zg8Qq6R`M+_o
zEB)ffx?-H_(}#XPzV4`#E_~MlgX<mnA<|1+*B;tW(Z@LyIrxI?+dQfV@nPP~2YX=W
z?4P~6ZeBvf1Nqt+2fNlEJERA_*)Lv{2j_Xt_f)hOH}pqu`WP4ftV`PA9kK`Ufn4K<
z_|=bbuov~{59wolpqK0B$@q;^e|XXkFY**|)%rj`yg>F0`LBM>7x|E$?$f6g+JC8M
z|0z#)-FWe(oqeZq8z;W`pYgFnamf6LN7@;;dV~GV`Ob~}z`8?@`EWkXAI!Tl`_bNY
z`4wb;_QCXLhw>rvwR0UW<g+9B2)^*G?0LCy(%0`2?3<h~n16PI54_uVsK-w5#U7oz
z8YlbGA9>`<N60l^_G%w&|La@|ANu$Fnq2Dz6d%}w_0~Mw&&l`jNMHTq$9VLgbuJpO
z*MkA;y1babm5uD!`Cf?DP4-D2@ryn4KlQ{F`GI&Muau{{&Q8T){wpr<U)MwA5A+k4
z#dqfu?%Nkzznn`rCm;_$A(BgPJV5b9J^pK*H_y&d<SW|YfqeSm!8&3*=1G11kSk6Y
zm-?Q|h&ST2`GxiHlBh*d46qC9uXV?I>%M(^n4dTgWQY8XeL-=7AFyL~Cf`zTo}X^L
z<qNOR)fb}s{GLBs=bRtPKOukQXYcvv?&mD|i~0IKbnAMje0T5p`kc3s=X_3o&IiQ>
z<CQnackIu|6<?g|Iaeg#d9JeW+c=+eU%tVA#VfoyKg7R!;wpaBr<ZYio`SEFC+>CF
zgIA3le6B%mh^_3<zTJK&TnFs$<umL~9MaBv`sBIqxu5(iTyH!#v>u6z%KDXGv3K!M
zTxCD@`QF!5-#Hk$^z$C7af|EnLiL?5Id9Uh=PKfT7_W_6K5IR(zM6mX=q*olA9~&-
zuQv|)EV+1B&v}ac);xJ{_FXR=z1EU{n0Fw=(0<q-<d40=P(NkkmY47g>%08Pee)-7
z@K56uNBH}KANs<k$9;E+KI<+&6&J+0B~M;;qkSjL)hF(-Q*n!*i?jNdXFA_%h4h8y
zIee~s=my8_c)^Wt41RY({t;rB_wfgD9dGtU)(t3*LE|$H*X`StA-^RbZ^kE1xUSp^
z$>neA>xcgM5Z|EogY|>&=U4~WMW`S7_)yP!VgG8J;*s}u_^r5Op1gl$ey!`y2Sem1
z`WY%{T<l7Ijt|c<!tvr++!gon%l=&#FUW`F@DFwt+70`%UMPqCSkFQ^^k+xn8~@5e
zafiLhJLEr}cglCzFS{1!oL_iuD~`$E*frkqA`i1qBiH&C+Q0t!O?V$)+8d92%Q!tJ
z7tbLc<-7b^9%vj;{sOJ5e&3=juW%l(%s;*N<M*iEn}nXfdM+FKum0s{`jr>4;}AnR
z%Hp8-qKtR>CqCJw@u_bf#7q7I<x#h$J3Vyf?yskN-?`)?NB?s6;CFh@xnZ^YC;sn<
z0qdaO`4}fZ2@${aaG!tClYaaWvLndu__6Vj$FA8A|8!sfc+%cD%#(bL9?qZH8+{=;
z@>OMaYriKS6Bqay|1po^g#4JF(gz>Vc#N0+#>L)TSKs`LpZF0+<P~@i<14+5(>yDy
zAI=wkAiENWw1dv!<ss%7>W^RG+kNpE-||Lf_Cim3gyUDwI3Ydt2l1GN#$z7wV7}cq
zuljM{c-5nSXz#|se#q0#{)K(JZr+XCd@CD2`PL6{UK!%ue#^Yz3+k7ANH2(o@V<G)
zkA95Led~?up`P^Rm&(>_^0nvJA)0@3+$YC6Ek0N;!~V1jQGIBil7;q{)+_NqT(@q8
z`R&Ln_nSCvsksJRSC1cCC$mtz7iYD%zKNf%E89=gS9@{IIh%g1(_uYw#BKeuBhT^p
z2Y(Vb?dQpHo&8(ytl#cuwKE=Z*gPonV{-VvcA*~p!hDD;+Cy<3zvSYRzV2)1x-vQX
zF>mrH?X5%Jn-V{u{>6FmT>j?0bn`1e^FBM?{eDAPJKv!gvG18DjX(UgRQ`6{)5p#I
z;B6xYl;wf!)b}Cyh2P)O>#|E{OxR|*5d-vMcjWoqv-<3jeJIlxf6D9=;uHTNs%QQ~
z`%tEjvUxR+`k{yUqc8dFg*<ZTp{zX~mGNvG>hlBgmFY{qyoJ8nhjRIg`_@5mz_`pm
zdHRFmLc7S1v(Wz7xai9s@P&8xt$W5{T=EF@@u4igbZ(xt-dfkh6+A$G3&lZb-k|e6
za-nrpKk_f_+-INSHNA~jd+VUzGsr8*wa$B=Ond7syVK4(U>#S_d7^cleCsYA=q2Cy
z_U4~o_wL^<(ihs1``(xK9YFaMJ&m7y`+N84OP+HO`WY9bzkI~JWTEF-@<E6nNWOe4
z3)y3c^a|0sk%j!pa{=p$^Eq*Vzv0V${^q#~KV#?O6+dK`>{~tSqxIbRuee1nJNMj$
ze$Eg1FTK2HtG@Sdtyk75>(C#*yU;zWoHlyE^EKaz{l*@BtIYSwdHX!Ka*pKO={rXq
zbl#Ekyf*mvyJkP~-0ZKsZ}#B#uARGw_53b^A2_%1dokxauKT?TJ9gdqJbj$wL;85H
zm|Xq%9f30W>O<G<6YNj;xj4grJn!}V7xFW9z}~&D#ZJt-`;b3}{nE#CH}Q-8i;v<&
z=tuMt58PL_pYVGX`w`zehx{3m>--iv=M{gA55L-rBgPHA|Au$#CV$ee`sDF*aZA7a
z8~@^hIA#9G;lFs+pYhO7o+^&0FV7{<edDE<ew^=`2lHuut=q;cUf_rR<mwm73-Dk(
z>`;H~PJi_F-lDunzTsTVzDYc^u7!3_fA1;kU*61q@IW5D&8zV^=V8yrD_+X$#9L@R
zw%?Y|7#|+#fp77~ef@{|20QmV4)f%FNOogg!!teDC%gB&njZRhZfo6A&-~$ueDlPQ
zJcqPCkz;-H+}nCX4xW_Bu@1?f>~oaq>$y9B4CT0Q9b^Z_%}&@8Ir?!vflp|h^jF_H
z%0A2wzVYfhyM_GEJX;@>!+9`X=P>kh9eQ8DJ`A7i!#IsgKjM>of&Z%KceDIh|E`ly
zfBtM9=#58t4SU6__#lrm|Ky5)?2es@L)JUvBxkUF@LiJ7-`Kl#!T8uscwar^pr3K_
zD`=f3PyAsQ{6~FuqMdm$5A03c7pLX*@@)Bl>*5)^bDoQTX#Jrl`FJ*Ner8?36Mprl
ze|kBG^E(OsupjO4MLxTQ;ypgdF%ROV{6#zycj@oC`7mDf+^3&<c%Y~GA(tO$Z$GAN
z9=z{jT<nQ_yc?(M>Weq(k&9>Z$=|(?g;#lqe)$ji<SH9KeXMW%kz9IM-#w?okABqS
zudcHXW%u!Hp0tPjf}W7Qh~MOzKl3Aw!!RCMN6asK!z(+)hjTCa27TyjT@=6U$LQs{
z^F#5OoD*LkeaMx&E!HRghyCf-`sKd;pYuoYojtHKhzH|0PUE*8*pITy@Vv~rCH`B-
z$&pt$hovt)jT2Ayv)*Ggzw|Lqd7k#+`7>U{;SkBCKl?Wh_p{`%7xl?kpFH)<M_x2<
z;)}S#@A<d&i$A+=zSQUU=GnYk$02?7j|co)xAp6~b=Gz39(nwgT=AD4?z0>7q71cX
zS9m8!9Kf4-VE5$Uhu-2Dzhd|N-n`*Myi&H#@ndm~Ug{YqJ8_?0^rb(&wTJQu<IrC#
z6c^bE`L2t%_%KfBJVZWX9<3ANoN<zCUg)cy_UxDb>{UMOxemLqPr#%3vCgv(*VUIF
ziywGrN6z0}5BsxjkteTX=j4d5=F_;@hcaHpMZD32zS`M0)8GC;Ugms5{>R_R7cb41
z`sSTJ;xRNXc5Yv8e(8&Ea`0um{Ffc;-|tN2ckGNj{o=tq8@K+97ccl@51t>|xAJ#!
z4L{ad@lBj3hdrv#9_S0nA)mdnYyIPc9PP-ruW+Ayaf*KIjvcwKKlV%?d698hM?F_H
zul8x==|{dJj%J~L_&Hv+mv4z{;=*8jn)7$T)w3S5D{({I#IrcZZtz0B>()Q=>8XG5
zkbHJ5ek+r!A9<m&@riHbyH4LwuIt9DzB0W-#EX8EwWpUnUO(~%<&Y2dLQnn|)_0$u
zLh%G&c(pD-^_{bdH};?6ONjQ1^r0Vn4&$hNl-<ZfH`wKpqn})B)Brz~|H|LRVgAWa
z_|KXL{qu)k+I`f3c;h_=?f7|!;*_}UKK~RK_@nj8b@ie7;-Bn5JRp}}^H1_Z)Xw_G
zzVI!6<K29kKl)lXA-V3G5BiIH#x2f=sJ?NsL*-Ck_RSyom2(yAC1iil^}%t@iC4y<
zfA_7o>|XrGr~Em*&Tk;z@S+UKH{S3%`FIZHXony3C(kz?dJoE<^ZgwDs~%pRi?bKJ
z<4IY4NDt@1{9b+in`iv92jhk08J9A<^}e-y$2o@i#;@@kFM0M4)@}OGkKTA74;o)+
zXX3Sdhko?3UTLpi`J8pv`iBqWCP#dR^oI0=p&y!0$j;!hCoOi#vp*d*0Lg*k7Ze{L
zJ)k%S=?&R8WZ#gUkUc?iS|K}<f6Bj{Yl#2iDt*-t<1l&7WwX$G5Lqa{HoxqO9f{lQ
z(slL}+O_At)?0ZwIqjnN5j@|sKOl$Rt{ad1Anae9b*^c=%AWW6J)-A&-Yei&pa0Ra
z$3JlXqJ5qR^FMl!M?Sf}vq*o}y?3L%@j`LKc$KwNk6h?|0O)-PJn(P(DS05Yu82qM
z+`PMPe+k7sDBfEqtfTs+2m2DIZa89U|Eh=0-{-qr8*la9SDt*!Tz%Ki)qm~btIX5q
z{k6Qvo*+I#bRT*@Bt&-J3h^JJafiq*p!GtYV}0R=p3C6Rc?=%$tiG~&uzwO)LV4ng
zd8aRbuwQ4t?9cprUO^vdy<yk-(T*PMj=kBB*he~dah)7;$-|F$rhjq6ee-D?(7H%}
z^8l@@=84|of%}j>SYPOcXMSq^8^k&JlzEU3d2h#iSK8B8oZ{E!Pnmq<B%l9!9}92h
zLw)P8b|I1jy?^<kkKez-2PZ7pm-U{P@Bi`pZ#?qoRhOK-Kp*sd3GZWy*WPRM{a$g?
zeBuY<$NcaU`$RmNC++Y6`GL3$*&%({MQDHOi9h(FpZeavQ=h$=AM2p|=G%M?%DM0A
zb*{MQ;S@jCP4*=3wZ8j1=J*iT@s)-0CMfTN><}6UWMB5<;-d8mPlNOG^3F%j^U71H
z{S$knw|JL@_VM=X^rQzp`MZ6EGCNXVS$+!5M~L!J^27=Cl<DERe&mgi{&+JFp&r_c
zYxIZY(2E?X9^Uwma}4tJYn<lII_3O|U7IKC0=ux@$+!5Sc&6-}O_{y0SNiBjzx;t;
zyYBhB=it^keq?{BY<;mlkqh~wb%Fo#dwS8^x{FWqibuQ}m-%#l<y?wC*;m=Ov43Uf
zPkzT?e%<$8kbdZ4zvH}1efDGjWgSy?p3Q%)_r}BT&4)Zj-lPo8zql<P$Q#6I_r*=}
z>Blbh?|jU;FWy`y-#LZxh39Yh;77{lTRZ+`eh1^gjdyK(@LMJ%;+yq_-jF=VF7crr
z9`r*m_C^oCqjFz-V2}2(>;iB4<tOai`iy_;ka6;B@<P9mKe0b?5%S|Kq!$#|#TV~2
z4C*r{A7hv5@4e(x?|JwmBL~DA`rsEY{7!q<*^}#@(~xI8_pWo&hWp<=Yw+(Lsc$~a
z7k&Ama|i1O{n)#4@K=1O7up9qf##_dihuls9U7<c(4Sq=k9^~R?2z9m(;x46WY6q}
z|7oY5Jc^yFPd~il5pQ_Wk8zj>?dfOy{2AZ+Z-s+?Gw1J6I4_XDS+6{og7$^2kbm0u
z)7v^>9*jdh>l?%$z1Rc)wB9=>w(eQ??bocw_;O!du@2i8l4spjkDnW#c{Y#qvR+tk
zl*v(_KE|oNcJwk180rzqQPvOpkk1$={UQD27wYjlcICSI^u(94`GVTvk6z*)zi^%Z
z@(*_7KD*V9pNI?W&UN+VhxEXU@v%$y!}Soq>X|?C^>19(Av~HF^N4?C{L#lc$S&nA
z{1PA9Ti=~S8XtY+J5aoFzC6hDoPCmf#kwj^8K-s6x}?lM&7(L@5B_TZh;Qc!#z(IF
z34i7#>l|A>^G|Q{zz*<7p8Lw|mOarEhV#$f*$02(zwSfx70#bHXdLQ|p7VEx@JEg~
z7h-shDSn4}rhc58v0HX+KM##hJLr5OMEtdj_$><g1H_+sH($=nT&K5vhxuYB>|A~z
zUz1OGt`p{K@&fr3JNKS|`QoSiN*v%H=8J!659KNJ(=WfyLhmt&!}f>rM{@Z+yYjw_
zGWqsz_DjZRJoGnC`y=lM@;i1&j{8u2GtcB{$L~GYQ67x%bACsHFL@>T>_A*LKI5eq
zf0E~jYsSlu%pd*u9iHW@^l_ct`P~{j72n(!|DgRhInX{#d}qJ<XE*wR<g#z;i1G0U
zd98NN8Q2lM_0K-|nRreQ_3&@KVt>k^T<z(_zQcOXsXT|Jzp`<%596XQ`S>$V{o6<A
zM?Gl1$)g{;CJ(A_Jmv*&?A+f`W3S>IKatN_|Kxk@pB(MjuXx0M=_CHx2RLsM@4Rm%
z&iSs8_o<-!5WmXoUH``Ky7MV^&Q8n&Kd|1A!{41l@>Bk8-o+h!us^(tZ{j8S^iUS}
z%%k}gSBwYW-q&!SUl@<|*>(N9ZoKj|<KS1?sjoddFn)T3dKovq=nwIQ2mP~i^CO;^
z4?NL>ojOmYmpCG?#;ftD57nnP{vrQ`@=)&wK>Ka^obf>V(u<!PCw{{E)+zpJ-=LoP
zRAzs8Hhvh!d49`(#5ccBavkD9nO~B_@5MQC@R^13g)9`m@I}sGyqmL+F;3^j%IuXt
z)8Ba1ClAtxe~V}AH00BD{sj36xsW}vo6&Rfe$O#|4}*QV?|G^HvF|t7FUVWvQS4tH
z?0iHX>^wp{Jcr2M_&0y#r`8Gf4YlL%;xfDCfBF+2`L+AvlX+uz>_fb=Ub}DI@P4m(
zW53EFI<N6v6YHt(Ht>J@*$~YS4EJfiXQCe%*2`NqPWH{8JqP5E;tf8n2l&t)|MmlZ
zzh|9xp1?oEEAy>?d6egXP(SuX=8Ij)gXCkb^DBOBeCA*LV}I65<7B_=P@KSvdhCo}
z>qq{pf8)oWynApw!8r@LkeqN_%GOKr=xP4M7k<Le#5HB}WBic6DciS*mw0ktKX}C-
zeX`o=hu<5Act~IIojm<P{h23r>OQ&J<41jZ&{KQ%qRjr}gM;=r=kFTgL)_B8>*PUl
zAU`2DyiO1EK|VAuP<`{Q9)B=?`e~<(clEV*UwiG;Gaq<&9ly%jsb{|37su%V$ycAh
z@e6(_4;EKKf6>nRnT6H?>xeie4!AFl=|`M$U%YjnUi1{N*$+Fl4wGjd)Z;JWhy4|M
zcb$BQM}CW6`%rO9JTeY(jXw0pPgr){xXIC88M>~2?dXL!d@8$7KjRZ$#5;Mj^BL=G
zn3p?WaSjW$mnT_|y%(yTJWc-QyZ!Pw=cV!i?{SfDUVWcTfA$UPu?P8_b7%RS_{NUx
z^YtfBp|^g-BYB*C2cGnU7vpe_`qdGCIq%_jjq0O^^-COu&KJb15ZN&_pOD-T@e;~o
z7a_`D*q`U)?9e)o5B}nL4m*YNm@H)P<eDeXag^C-i0+5zJ;*HNk6FmR>~Em`gYmNm
z*R5N4#SawM&Aa%;&ditZUFwJ3x-Wi-GthXH*(X0Ihg|yl&Zc>EpPtI<tH*!j{dmw$
zJ@2DhH^m`x@rXC`i5GH|wdbeof}fc`&xQVW?t{Pl^a(E~=0QLFNj_qH=EZeromP)t
z?(5%udeX<adx+wX@2Z(6{TVmDmFela{z44LOFwp^AMr$cagM!NUs|E|A>ZB$@cm8e
zmj3Z5uQEUEiQEv`3;VMU7>{wNXWrDa9+^Mw*r#>Ib@QXWa|S4mii67HoxE?5|2gkX
z$wTO=ti60od=~G;1?PhN46pJ!JbK?no`g4c0-ci>H?-cFC;WIW<Q!4{F7BBJda?(+
z<Ij7f>_L3-9Lc$!b&WshhhNYa8khQbYlZ9}3qyaFSHo8Fv+m=;e2SC!7x%>Z{eOMx
zh3nlqBMEsR-+Kq_UwiRMUgW-c6)%k6d9nF--o?Jmk9zo4#+P$G<6;NCi|KuP`ib}K
z)jmX?Xg_5BLw(piUX4%x?2mlcjaz%?h1OLlF5pZ5{D2(&^9OOnIKBTx4m+~G8>jeU
zyyhL>_R0Lnc+5L~>4AUxnJ44Yo<GsoeCtQtQRb)Qh%5LpPW-S}agg8IhlzXE2l=IQ
zHTfFEn|+`-YM+Eh_T;|$)<1S3UpGGf$sg$DI{OUgNf|Hra2=20p6mR`JZYyM|FIvD
z2eDu6#NQC{ZXVbl`%*S8>%6=Ge|R-t<^j5IAM3hvOZh=4&%CfJdMKM0c18{~zR-@v
zKYqZz`3t%HHRO$aJgLtQ#YyK;^o8Vl|4qE64}TP=#KW+jctk&bC2tbn?a!?rS%{Y`
zBoE?Cynx1sSJ&AYd*|=wmEQcsytxjwXK%{p!@P@+c%=`$m7P1tpYRR&qqxA{jLZD%
zkDk!HK=L5Hp?>Jy3fY<d#R)t@a>Mly&&Fk)P+xm!-0tI%U6D@@aX&=om(B&`%i@mv
z=GT0SyUOB-_~HCj{A3T}v^XZ-lOs<Tujp%DAbIYyYe=5@_@%cvKo4d7vM2swULm{E
z&UJoAKS)pa&9gEz@7533)n~8b13x04Js1Z&U>E#JJMyhxP<wLquU{xXcD~3C*&QD7
z?|kyFlQ&vsnMYRc6L-vu`LI4In^$q){JO4ezRWwh@(t@IIp&YOwL)^(t$n8VAl>Ir
z`eE1dN%@2QrTmEh7`Ob#`JntueJCIEyBzO3c`t&V`cpOzzpwLqIy{l1U+>e=2Ve9s
zPI8UI{K(t*kNWOIcE{euJNcCRc!cbX-C3Xcg?Pk1`KS8fy6SxSecyZ6FZTGvE6M0N
z?@90zenl^Fz&^zMn;&*<p5<5gpg;d-FZ5&A+VP+8`IT|8Z+!7%?b(6%*2I1L73ag^
zh<G7Bm``Q=;laEZ4}UbT#zRlzG;j2_Zp+KWN%>nA;?w*ahxu|oLSMY`JN>wBp6Dyz
z6nCtjcs73aYaAi!pWT>Ob{eAoq56=%kY7P|2<Z>Wf#NzePmq4zD+-ZZd6NC2^@zR6
zuh|QG<VXC$x?>*A3;Fz<9kEO8%nN%kAKH=Ye1QGf-;ry7fM@f^Zmm!Bk|&ur=Z@x|
z9OruUWXI(4JNCwp_?3Ca7ds92Mecj<q%59W$LOK|P#(XgkGOCA>W9~jTYd@Uoz@ZS
zkLTX>v;K%<)-UsbUwY%od6+zgUe24uGv^@Uwf5$ZUHKlP@whLp@(<Vfo%O|h;R_Gy
zk&6%dY3J`~TTiXu+8ZCf%$NIk(64o#-^nA`r*T-<jE|l3ck=?p6KK2;&mpRBK9$*<
z{L4B>9zB%Vw|V97#zlYjB2M8?oX|g>^}}xYA$x(=1><mTfOmc#=4s*^yT^}q=969M
z&-{hRZmpZ_OM7vh9OGp#<nc@Q>1F)%5_j}toa`SjuHzXRzj;w#8G4V!ed{+p^p7|1
zQTTgptFAm^%kSRzdIJ5OInS-mzijkp?&}-b=e)vo?`e{wo$)>Q)d}bR=!Q2E*R_M>
zIsfpSQ9E+ggW9`KZ+d!fR9XM*&ickLtON2|`5xZnOZ4=2^sKA&p}*&?(EaWHw(%>+
zPkk*h4)3F~N9(`!+5XbH?7HVL;<o3{;y-(42l6`mUw+R%*^%+8$4>NTUX4>b^27^B
zU-`7W0ng%=I4m!+PI=zPzUiyIycmD(n_uguI7u)5Y8}Ck@$ftGjbGtGKIB~4x*<-9
z3*s*x<dGrDv#b~54_?iy`8Gc9p_^~~iF4uzG_UTnb9uTlUdd%g+Nn=(ddlD2#~c44
zN7=mM3$OUWo9}p$FArrm_NC@c|K#CK|LmOp=82tJAGNdIY6taa9M;h=@3StjC;5sz
z)9)U{Yp5N&lz-rxJwx}6M}NlYKEBD<o<F#5y`YagO1&&(uf~ZN=b-dApW+^#<q7(C
zuIjpdfcElq>$kiNZ^j?`0Xt!*tq?EJb4mI;FK4&v(Zjs3A9@-mUg*WX$$|V5`Yw>~
z;2iqKHE-W>zL5jBj@n@DSKgRwz<qz8-FJUo_q}!H&9?mdm-m=9a`5l9yHCD))Bkkt
zv-@u~*8sUYzH^z)b~$|H;NNMNCpvdwXYxh%VEp!p-ZSQ3>dU|6vGnn~M*9T$xqXQD
zTb$oHFBG@^eItHrKSdtBjK}Y{<*Tl1Cyuj!WjwG`zuVG}ID!xBnf{Gq?>m=#<mg|{
zPU){-=sMn<&l)#<T_<0eebS4*;;DStJVEindaEDnFn#1(?nCpdp7<zFkgth5=9fNr
z7O&WWby>S`-q@x3?z@f`<5Q-$e(A5CGX2Rl@7hD{=+6$=5yYSR@)_qw#_7A`>_U6%
zg83Dnv@<^X>c_gFES`y*+PP0JcCTOkg!RPdaGk;<Bv;wE<V}7TO+I7?^dtx0-ZQTo
z;>)^c-LOvZ3wFdGLc{~RmhZ7MeBhNl_6_k(4!PuL&+pl*`sNAm#>emVOCRG1?MHv=
z(Mwr>>`y<+<jFhOv-k<s579XELr?alY@FszzsmG-JqwM?`3&A&7l*}7`4aj3UEH*e
z(T{x@m%Ph5E#H#Q$$#X_&N1X;@<8jDdfJ;W_wjH2!JoVbU+Nnd{p?TV`SK%W`wBSN
z&z$c$TK}zwcyL{u!~?&xK0<k*@zWo_A>zZn3$jb=t~g7svU>P;F5y02*)Lv<A8&Yp
z;rl!G{pP`bQ@-mtx4hNwU7VXJL;lOo*egA)lhz^Uu_5w5c5l6tm&+H$6>*9Eh-=~&
z|8-uZo%K(BW$Tst<hoBk<K?&fj-8QXzRaiV_6Ov!EB%n?KL62<eQL)q<k{?kowFaj
zK=Wf<;*W8YPY?B7H!k|n+kJ9eH-2#uKhXY;T<apcmRA@jG(PenyK}ySUv_|3^Jv`?
z&mjHyCw|R`{TUv~rH8V-6#sb9zk2w_qj{kZR9}1J;%COIp7}Fh_!TGcX}tIrU(_Rq
ze)Kip@alzcoW8}vi}aZv@j_g2jwOE)Px0dSM$T{8xp-wBmYnbA+m>E-bf0rr{_EUU
z+!BwqcOGSY#xIWH>CV0Wu>LuH^Y{6E0X?jb?ACdv=fTd4jnlg6oKk#e-`blO{%hU9
zk8$f4I*+rS8xIsO<qiBpKFw~}sr&5J^Ah<6dGZ+FI}q2!Z*f@sCQo~I%g))AeUQAs
z`pU1=gXRmrc(cwy^TwW?XIuZA!{ZU(&MUPe&wgKj)+=_)?$|wapMU6If6l}8M^D$;
zi9AI=o-^PHidXU;cCXCd*r~F5_~0M>RJ_oRUhI<o{Eoi#bPkRW_Ntxu;5xpbcIxqe
z=gH24^+S&S<wu^A@JnTJ%X2dEM0|1Ia|PobjGuG<o}Ij&J?kG|%H|)M2RxHw9d#}t
z&ReI%3-`^J{@Jho`Hg<q3%??tpQ>j*{chX%`K37M?~>4iT@J4I)1KVmrBx@S@+fhW
zztWSRxX-Tm6??TViAVB1>yv$ycqHB^Tj%T#l=(Y7<-7XBm;DAi@O}{e*sFQSLi}VQ
zKVk3im*;=xYim3|e;>ZY2mM?B?34JB=c3jn&++hV9{7j(@f?dC^Fz=7l<CVZ<um4m
z-QX9>XCS`$fqKg1xz1i)R}b&(J*@9O<d4vG_IJ(=tKC2Ge@6^BZ)bPbJNX(v(q6yT
zQ+{fnEWRqU2kVl!CNA1v;uCNB!H4-|7y1?F*|9SJWjBL7zWBj$Bd(m7vVU=bp4LZp
zCBEq&f9ylP#cs)CN6Ol<bJy{0zU6Ir6erE2^-Ubar@X^FS%>k*zqRKN;uF2_BEFap
z`v&U}|Fy2*Aslxo-}vxLFZ0R{@UA}O&*G9e&YrZxgYlRz^9*NRdF-wSd~ohQ<C32*
zdGe|o?K@$vKIl9G`n!KwXuhHEinT)P5FVlTI3PZt_cuC4^Vu%SGg@INmw!X{Z~YQ~
z#7}5_hwR0D_R8PcEqUU&_O6Ha?z(+}`_Of0Ut+!1kN7DL>DP7Rwm!?7$aT(S9kR}o
zCqI!FkSjhKhkXtC;-mY}J_|aRFh2V$<8y8;o|#X-cfgnDiJpskpVRl|Jm+;^Kh`Dn
z<w^QgR^NWmbv*i=g>kqJ@kS0bKgJ`z$gk`}$zi|vRiAybOT1*+gMO`R;=Fj`xsiMg
z+Mjz*EzB3hS!kZcN%gH8)^TNaVqFqP@xboXr?2$|AM!QrvygwYD{{s6yok>b=@p{+
zf%3|{*h(++)C!H)`fr_AhSpJjNpE&%-dq<~oF|H#>a$b!C$BOt^P;}8cx1l9`^KX`
z^Fgk74e1N{ONj0pXXuam6;JdV_Jcp?J>-fH>O*pkQ+wkR2gC*aD(i=Q?eVXkJWC!b
z9;j#j$#b7S%a8aOzWKj?jMw*j$j2)^v&wjPpFYa;Fn@#m&G~MO@yUDW<@+wyIq}E3
zMm{@bx8j$1(4YCSj>?Cv!`54I-h7*9a>YCKoM)I1_4z43F)!hKE1L&=hddgu^^RRx
z-{c|IG3~?oAxA&>6(8_RANhmzLO<^7$Gn***R|6hJD|ULa-Dql$t4e8#zDU8=D|F%
zd-DhB1IdT_h2{^6chLHph3pU-xBC2-J>gyaVn@b<Fa4W8^0kBb;aA=Z_5K9j)x(SW
z_;%lZRJ_n{s3-rXr+U^6{F-m_T-TrL<U{*7`$Oxx{hvHld;3Soj>$1j^5ipkx8Jfa
zvwyQsbxy0DyvICvjxE0HH_Qvvb6*}MPTB|h9+~}^deHtA572x-`*vu*4nzISBgDJ*
zMIm0Z(C-GI@wP&Lj~;5zF8R55V*Xthcf@=4!tS8^5dYRAyxK4G7x4kV{DT}Q4v6#g
z=fCWmU6ady=n*3O3DLS1BK=ySd>}-27-C*KqK|l5Hi}Qs{<B@o;>+_)^J#n`%J0J8
z2gh<`ATlswgmWYDPCmg;y-zNFI=6EE<DA1emHP5*W$*dRtGo}-kL6kV(cW_vdP4D<
zeDPj>DlZcc#Chjl_W8z3j(p#^ymuk*r>FOv?Tg64qxfn)v>y_WtUvUWkK@NUAbIvP
z;d-Qh<0N1Fv`$;!tfSUx_1yn)GHK?hyXG1EUXJ__&(_rttxxKkC;2n|j00LH!gVi{
zuYVZwU_4o<oxA~>C*!BDeJ+0SN?+rH?uR_!BMY<i&3j#cA<D1ipPqwx-Yc&%PWH(U
z<!$l|cEqog<=yf?ejtzcJWSrFA9D1gEPr>tsJ-V=>hTMCqkRFr*oS#`-+3{62=h2)
zev*aq)4Yfec5YuA>d79RZ_?lXoZpx?`mtNjMT`$0A>uC!^_zvpL0|hi$Zqgso}u@9
zVCWy(hiDw^APe=Ah3qN|*<XmQo?B-1SJZX;H}k^Z_@DO~z3=4thkD*0V>iw#=qEn-
zUcLFWueTqB-Xr1P{7n3T;XLx&EOed#?c*VRq3>p8^-m8x(U*Pld+QFnvMz|P)-k+=
z^WZ$gx?sIv-`exfFwV;t^vl2G1LT=c?VU%6pW?pxG(K@B9Df)G$RSsp$D{B6TSwG)
zpI_0#{NvlXh4xTA=SAWSWG}{pPwR^FYv<m^rM-Ez9?}~h#>-E<e}YHv>6jPusvIIe
z^c@Q0^8S|j2oXR04?pm#r4PUT^&RHv;~x-zA%=S%Jn<L2Lg#C~>(L6ulU8UxLd=S*
z;W*kYo7XHf-&ttBLbRT@!dCsVKfLj4b`~N#^qwrggx0IDU-mQ@w_e|S*At(3D6yWh
zTk*todU;=+zUEJV+KEr>OMCXM{^{$^vB>cUj2}C0_5S7hm+v2W-1sr$`$xU`@3=+B
zEirDX{t?HFAGh?I<u&?8_K)0RhgFXqKkn=kHl4Ea$k3j<F3*}U??Ipc(dFq-#cm0@
zfA%_8AHL6RNzvG?UUhGO#iAp&|K3+_NIS*4jql*IZu!!gt3IB#i&^7nS58<zOK#EE
z@Bil8-hb{AlasQrXnnKf9k=uvZymA6qe;a7%tID$?Oy--HoyMpjz|3?X*wQ%_cl+z
z@A$itqOn~*v+Ccv*`GJR`0KNhsxV8Bc>VtvxZ{K!=6(6{zot#ay&gK_`+LoGZ(21@
zn)>tto4s&N(h7Io@8};SANpOI7t6NG6MneBj(=EuL|@bJj3t*p{QbWj*;h4oJ0GWA
zG=9rVb~z=9SS$}~UUm5Gmu+`<+AS{k`@;`Ebcx&3rsKHXKm5$Rf4nJeDz31>{i{#<
z+yiOEf9Ijw_|vL=S^J}^<*u^U>4%+n-^3)uR^#vXeo^D>R*$0ESH1r!O7EiOgmzR_
zx#)3(^1grf|E#h3HS_dE><tg2uG?zoofe+Auc`R0Z9ll+J&VlU*9t?s>$IHre7{rq
z-M0I{uQqsP``<p6bc@~U5!?5cfvFokcl7u>Url<2-Hv|p8wXzb^g4e}nugsz-|CjP
zqS_agcNL{imYnc;PAqr!e648SvgRwi-tBowR(t&vwI3pf+=!8ZJPhQ`7t5}fN7q=V
za|ZH0uj;&qd@ZXwH!j*<tJ1ryemZ?Wr!MV^p3kb~<$a!6*LqbyFD|RUvivKQn=cxU
z7*VwymepU;&(Vs0{#(|#do8!Bc#+qhyRG-lGd?xq!mFpJ!v$}=>B)7s`OB2FEDY!4
zf8P1c<)`j2Gc6m7I-d#U95r#9>)*QTl%&@f@*2y{U|^LOjz4v|(X-NCV^M4sy-r4X
z?Y(zZbe_sC9{l*p|5)R$htj4Xy@xAad&*aL`PJGlBu&LfuHSRo5l2r;dX0JQXSIDF
zUwrq~UroBjy#19`uh;vBa9rKSv#RzZW%cv(PyhP#({K4>QWS>#<}GJEH|q9!-u{cK
z_xlMyU;l~urzS;VQF@gvCp-`8Rk`eOX31Y|-TSwkb?_5u({cYV&RY7)m7YqQhAYk(
z{pzA$dpa!(v*x|i>v^B=Pn_?!$0c_@nih@k+V7%sA31Dt3d^4Fg|FJ+m#gnTH61G0
ztN2z_zfIX^g?e6h%%r=o{mI2iQ*pTeex=iRS@(9W&S&ecf2iL+E37f$=H;JDBld=e
zrsOf9-TY&ojhEeG><ejAv8nm{P|2&R8b@fidCQONxW;xj-gidYbS&z;r)W8O=b`L+
z?S3ESgU^iq^vD13aMCWWcHRrC&A;rQlf3xJ3pQBg;D6qk<i&P>uhH(f>r%hl-+8p#
zzHIwj<9*vbI`4unrV%^iq3rny<^OcdPapr~ytC7$V%hJ<lr8`ByN}s#;!ZQtUSVht
zMVG(+z!~W=?|dX}Dwg%WSJiU!p4+xtuV{NA?{kk{I(@}&yqt84<kb!LKVzQzSGeQ#
zq*vG}AKR|-%}*?F)bym+*s1<yKQ9<6@;l{sxU|drJFU9ctIO|(iq>Pog1`OZsu$mv
zR*j+kPyFUx3%-4=NlCA;sPF5B<LXuU;gQdL;l<k*80a;2`(E-Jvv=NRy{leKJH@>H
zPygMiqyFcLS!r3=ss43shu5b3uBE8w6^Czs$i)L!zMNKtcy20|eJ)Uz{9c_iWcAyt
zb)@Kif4AU`D=mEZd;<}OCJ#m1OV#p<{@rg?^1}D1npQ4)UqfEi_7nE^rAdFE`?+7e
zkw)wd4+pGp(62u9_Ynh4$DwXN{N}-<e|XmEN4}V}!oz=b%$L{R`I#gy`kq)(*y?@2
zqVDH?4`8U(>-F!*y6wNI?=KG(z3cMcN~e0rUz&<Vzc<#cymp@>w|YJ`aKhUj+WVGg
z(pK1Rzv{R9e7W8CD)rM;Y<IkMtKYQ!S*UMW@2^)Sx0T=4{eG|PS)N`tHtoAD?Y>`H
z*ZNiQ&VIW6y9wHd-_>e*Iq&=8O|57CtHO5Q+s~?>_kH`S#?`8y9mmdk?gy_;N}GmZ
zzR{F&RlhIG%G<krkEh*zN337Y-+yT8d*Jjg$}YnA*6sI#dfmS4^PG^cx|GX$uc|7!
z`!4s(&)<95?6j^});xFncMr?bt0=iUoVUuK{<zEYX;bl{J2(CIVGB)9n~F_+uZ`YC
zt#iGSQ}lPIqP+IL=jHocWnocqGRlbzL<S-Qk%7oSWFRsS8Hfy&$-oEpU2DVV7oL{p
z#k}tgw0f^6>wd4Ew}<@>mG3n+^}J=M@E_V$R=KF(O%^RD>-ksHuIII<GGlF&9U15~
z164gYFRP!T&ONJ|ho-#`lBI7|zpt#ipSpg3wyeKX)GN7F&udxz)zx0Aif47{$NMiK
z)<wU%>k<C`tXqAv_BBnvUiaT8ME&ypJ}mN73{Q34_tz!PWW~{<e>dLl?=HH1zpUu?
z?do5*`n~#oVz>SJ9Zy+!_oknibnT)~B*O*E;;~)1S#c-t^|dbj&2C$+IWwsn=GDK`
zdPVQ&I@P17cDwxS?s@jQbM!z}Sd`v%m-B<a{{F0|zOZOt)9`CYJoo-FqZaRLDz5R>
zeeZbi-HZ1%70<r%@qtx7zgS<>@w~qtx7wa-E!o#J-1_`+J6w9yVtq}=4Q_wrxP|6g
ztgkNEZM?4Qcj%34>PF8UJ^s#DlTPtn?>&9YiB~<F^a@x0^6uyFf6bFg#NP1G?f!=y
zMjYM@6#YI=l-Jqc)s25Yu(Q7(h_~VG#l??)!YBW{;&-=wJ*^sxAIZN<9GL%OSHEMC
zdHRM5&NpkNt2SS3R9{gTp8pp8T%lL;!f|z4{&#C?e(B_GX6^apo)0B;#iG9RTz5HD
zKY!}fa?9c|uRh&=KejIIyZtV;_IY{Bs#n(c3d;H&d?>f<?;Mu>ou78)|9-;H*MDOE
zsYy{-^!ZViot1UHsPFTarC-@{%gU#U(z~hOwXDlJQ8zv2-gw;$m%nCO+9~G!9cbO^
zbvoX<wQI`Xc?|XJ)pxOqii3IO)a5<HZuQ9fJMOyG%lh4TyVvv1cenNGx^Hi#SJro#
zn{qwx_oKV5m)AbpeV?yx^@qxH<f8T?S#nz0Lsq+@uDAMoZcVxW$n|?pJL2eR32en{
zr}rmKeR_e-UbrTyJ9aw0cH14d^crs+vB#rH)3DWgR`mULd9A8`r+)`+sI&`z*BHx_
z?!NXX7bj)&Q1&@*+4Af9cY@m0tJCwwFg}&dH~CecpM~SBs(i;>Upw*6g`Q4e+5L9<
zd)Kn{Dq4Q{yeXD1I`@&oCZ~~y|IWh-8{EJ8q|ZH&HWiEd4pNlU-go5j)#~qL6&>fw
zmUqXo6F>aT)#e|l3K!Vm(I@}3>H-5z#RHd}vht)u78q!SMbDeOsZ;Fs@0R6l->H7B
z+TpY4?^u!BE;e=j$kMyj@3PwEr|SCMp63tM_EpbAr~Q3?_c8lT+-XMIE_T{q-ge9V
z{_w*OUE=n%shBrzcUmtmk72#=`)y4txBH#JaQs>2Zl6bF$?f)fw}02M+x}zwW-w6n
z_oAY_(+@lEzKKcXA@cC19W)~f{7cJD^NPB*``5^N`@i+p-frQQ-}`)0cl^btf4b;|
zcfFJ}6^rIKuhn!K@7sND*lqp1<7v0vLNj;%##Z+|pEMP{zf=~s%ctj$O~tys7n!9`
zyYtd*{kndaw<x{J`g<Z($t~-5oaA<EKTT=h>AAo0wflZdyTAVo>vy}Z_NslW=KyuT
z->IF{tzE0<VNJQ;)Zdx<9e>ksuZPa~{$6w4n--1YXj!=UZ%+E+kw-4s*DV&MU)8^N
zDoXDxIo*DjFt#6F-?z`2@2q`K*7YoVbiL^But#}oY<J^*XQX*~XzKc0w%&LAY0sN>
z+3}?`3*VT%^ET^U^<vs|Z0g^Mj(V^1r$6rUd|DL`;&sHxKx7~?Fx(l)%8!e_KHPbZ
z<BtqP1|kD>VW6t-0k`{It5|>5(pRqZR65josEZ%8t4C3ObNae-dLDnk__5<w?_a)u
zh5nJpjUO|<f7F}*j$3rx662QYA92k1aZA5hUZa0x|Hv(NSoPTP<IX-|(<v*DY*+7=
z^%r<<o^PCX;>2{QVoL&ARJic<@2vKfPu!3ejV-!u{Ox_e{rQ27=YBNl6zfL7Ny|O?
zY`W|1NxPUej{p71$Jd#0+Ffa<m?gL9>+O!W-TFoAn<ejuOU!<9zbjuzBK~I{vUqFv
z`ZGuDwbjuFPDmpDXCB(+Gpqi$9>4JG+fAFERE1f3#OwdZz{2UhyPdMoRY_BE$LD@_
z)z#NbPOHWnpZor#&nz=ypcS^ePL^$#uO4&I`WGDZbka1O^!4{o-stU<lB%)W`MCI$
z6+b)u7tf{<i{+uHb$XS#fBCCTAKovC*jpZ&zK?G=&gZS)ZTxA~{)68i`QR}hcsy-7
z-tdmTO&(tS`832<<L~zV&EMZ_rFGvkqOV)*R*$0ESDhCWrFYSCmiX|_b8mgwl(cFr
zdK{s=)eqiy=ka}``y%#+hvnDb>$aIwW+qL=RVFRH=>4;&B&{&CyH3k_&tW^2-)+0I
zKX=;^i!3;*uUqU^kJ!Gq47~O6MLx0q4KveTVYj2-eEn;lSm4OVlcr&}=K<aFR#f|L
z$Dw^u`eeyD?fWmyv(gVn4MeOL4_WgSUN7q$N<T%PqmUahGLVOXYd*f|x1adX!)dQE
zkFK#!=M3b1Uf+2S`C3->+$`^S^VVDIJ^PP8bEW%|rs7wx+ivkq-}_?HRBS5m^e*c8
zORwad@Q0Nb+wHO|(xzio@h4w295I4@bc%V;d&;hN)8jLLxcBi7Cq>~sYu~%W!PB2j
zio&Aws+*ju;z3?}?zUdl@1^v*zp~fka6YPve|PS@{HSLSe>E){Tb-j<bwA`amYcyq
zUC!Z(Vymk6Ld)vs4||<){ka=FpEMQA`n||p*ZtxC<wm}eHU*2a=cdZ3YTX&?{p6kh
z^R~V3k|UOXBk30N_E%QDqVuD6<%Q!qYyKnF{N2pg(@yc}#|G9~_1#aWO~t(L8Ro4Q
z@|(B3`B{tHaMAjINsGq3{TEfQ=y+9>US-Sq-6l8wWu2oQOnZ%Gk26bt_+DvK%jZ5f
z_0F5NoR&5Xx8LiHlfE(ki)mSyHSe8XKlS>vcfIVC7t&T(*88R5bN<7>zvavy+&Voi
z3n!oQ<uA{Z%ua_2_WB&H+y2X%=d$IN<quWKZOZ<(EIm7omvwLF{kklyyZ)hmMd#sF
z$t(K(vR=vS^!aG3cHul1U2baLOYf#W?{E6=^QzXrQ@gJ!Pbj;es`<{_U#Ig@bh~zc
zZ`|qn7Vg7Zm0Rr>!~5b|UMxF*FH3&e{u0VB+TKEWWu0RT7r7yyWtTgBeo&Woz4~2R
zQT^tPZ)Mdh+rEl^?q0RLUOyM8d;ewGPf@?uEK9#`<#zi#yxaDj>fdUgUe*0h^{=}>
zwbHMu-+y(=?{H~X^*U5me+N&x@?)>Bd3Q3@uqr<Z`w8v8ta-~TH}CJ}TGb23)vIz{
zz8Bo>c|f=CDTL=GS>???eCY~%eB!0FEbR7qZr$43XSG^~>vBJD9B5UqTYGM|J)WD2
z@x96JzL)R4%wDY{MfaQa{5)Ro&G(Iq=DlcpsaoDeAMW4gqywiVO~a~wH`}y+s<xl7
zzi#u_y0p*xe7A1(>cU^QdekldFIpddlXpETI<G9tZnO5Ad9Qan&b<B9tzNso`>I?0
zqMm#9N={kt2UjJxTfgab`!$!?@ybsh^K#l$4Bw~eb-CU5y;_|Eh3DB><#zj3zuo79
zS^Z~SZ#BNU-_IIXQ?9o=zjdqcTq<HQ=X>q)C{9$3^7OK?SHH{XwBM$ki)86rRorJk
zMSUlyZt<~Oee&*u>r$`X@tXgtu&BRll;t<!^{U3zs-LR-x#)h{T@UM0zpQhlP;S=q
z;7->s?Xu*?>;L)h%az4TINwbxxB8t^-uq#EtJ^tcyZV%Ut`zcBmvYm8m(;6s*u4Fg
zHP79CuevO~ijrIQcf`w<U$^gfx2sR5zpJfFyQ225QBJeJn_JiUYd20_u-A-Vzhm!_
zeZw8w<*lf5n0EWks^9K;Ox@}?^*b%}F6wz_*8I9&_4$9R{^(Woc~H~j;h`w}*tZur
zZ_RyPN}7r@AG~?`pLd;|3>B=)@2jfTqm@0k`h8q@KRh>&<-N8#`oIZE$U|LzN7b$#
zt?VW1e!IVuYPBwPd%xXrmsP)M>tL3?O*_w~Z_)2P_ex&g^Y?cBu&Vm<?51Ph`Ok`z
z@p@!nC@|2}eNEANS2dU2=0kO9U$=9sqV!pQ{k?9RIb~*2H!S+^-}G8u+0W;yl3&z!
z;EI-0RXnRpKUw)`UG%HF9`;$VTYdB9*G;Qe)boX^<TQ=nsPE6Ny84>QX;XOc{%YN@
zF7G90#nGbkuXgj_@SMBX<+|BVyZW^I{#9M-mz^iK`<`}Fen*<6XWsMOvg>8VoxIn(
z{eE*<?eps2X}zM~Z|qc$qS{sc{&!J&*IiE0>vvW1s(x>w=>8U5;2k?XwE8PaQ?Y11
zdL^%{_+GZ$Uhy^5`|UPf_j>=4)xU9l^}6jA-}JpNCY@qc@ja`by6iW))uX8Qy{nef
z?*2Ti&kiFFZwBh}{CBwPQPuhVQ0-@UdvWoj@73QsEPf>aE-@=FYwGo){dwW^-rY`F
z=&A%}z4zO+>qX7)zgtuDOGTeg7A-I9J&d7tz1`<S-PZ5aa?ARA3VHSE_B$4JX<yXy
z<h;CP)hp|}#O*$(>$HB^&l|0Sb-`}^K5Sof9L=(`vaV;HPqll!Ed9!s+o|8QYFCur
zP0JJMTNi)tRF9(a-B3=u-$@JWm;IhXS@O&N-dk1j%i^bOxvk<=S@(;IgL&mVIsg7+
z{`U(D_B9<>KX~Jv$M=oyYZ{*Yx!aCdWWiB=-D2MP?zUdl&z*|uFI-n*`9FUzBCmav
z^__yM<jT96j%B}R-^w1s=e$Li>vDe5tsde2uobVJ-VeW%Q`d5*<7>BF(e_xCyn`oQ
z`LWm6ygR81+vT;Y`r&@A-ExfYz4<*>RqJ?J{S<Xx-YYpp<;}g8)9L$mVSFmfPC~ie
z<_Yb#564+mIXu@dyWdXFd&<_UX!&KGD^(@8s&mEI&%j!%zWeDE4@JKRQMJ6H-z%zG
zUZ?S?F713bBMXa;b7jlB==zWE_rYyP4pfC(ukoIjCtp6#z);7c=dIiPGVgbXx_xis
z{bR4&aPf;?PCCV`a~OPv`Bb-Ma@)mT?N`d`x7B<8?ebG~{Z6095A}9c&qJsEZT{g)
zSJ>kdFQx5br~Tz^cg-<tF7tzjUQ3#ao&Jt7Z@auahV{bpfToq(eJ&b~KdapB_nouk
zc6)vDDPR8bJjv{|Q|z|?*uEJIl>MAF%J0o{ipYCCcyC4&_?MR5?$7Gh{$C^O?f*J`
z-cxnErhgBm+jEz&eO{~S^t<Fme=on4yu9OSw_aWU-e+FD@}9SKTCZI`^X}*BTCY={
zs&3crymVW?>UbXeoB6}NkAFDH%R{&4>8;v#Yd=kC-|72(#@FuqBHcdMef5}w*1zDO
zr<0~(*7t6ka=mW*g--P;`uD%Rk{9N`O)2+^uWtKo_r2e|^~KS$u+!(oVY{OAt9svB
zl-^l#I*miI-7kN&>BIXaMe&d~ZyYN1vh30Is(vRD``Lf|nJe9wggoRu&lqm?T0Mt7
z_qnNe-n8YkwCULN-~Du(FZH^8QR_<Ca>VPZ@U4$8@`?3tn3+cGEe~(IZnr7t-!^(+
zxMI=v-sy9ls@fGj-k~b5s`p0o@>i98#D41eccY7%m+<=wvAp!6_s^b^)P;wl=6Sfu
z$s0#o)hj#CCciFuZ`HpKjQ#CA_tuw9NwaupWuJAu|NRgD{F9?D9M#t?p1$s!5sp7#
z{Md1;_pi{uV*kkF#*Z1_KkChY$1OT;iE&Hyk2q%hxTW7LuhBoUf8-WBta|MDac7^f
z>6Dd67S7Ld$4x(O{hwT&3{~uwuyNo0`MOW;aCcHPcB@z2+fQHR;WI|x@=)3-)@^(<
zA9^OaaG__@b}?%l6PMg`g-`!>a@r|o$u0W&1FPM>{(|3mC@Bkz);CMu)w}F5=9LqF
zl|=l{JY@0K?)CW(pL_DhJ~c6EI_~!Q`!0LxncI`1v0Xm1>hJN`5~tpB*7HeKn59R&
z{(lU->-clloPGN}X;bl$lh+z~;|ae`tHv$=diVi5_s>XL;RdIjJ#oXij!*Jp*>-vB
zMZ5lDt~F;TO~czivg|c)+jL}K)!6NPjN9g!rKY^|pGm}Gd01xL73&;(`9IQb@v^fv
zJ@hA!{55SlrdzG{v4<w!kv0`iKjoI!Cf{{;+EhIB>tnvT=8gBJMPs+|r&ar%9-i_1
zGTTi{nvT~zzU+;2@A!NYVyp3Ydw<7$?w|WBSKps@i{0u`RQtSnN>TNS(z|Fm8%-H`
z>FIxeI;k3q9!Dr|_Fbn=y5}#`lZd_H;jbUPXujtskLYVE-n8*oZa!`MQGKm2w7X8r
zdGotY<#*ez+qkcNw|d0(y=7p=J6^f!$S;oW>lJo8`l8nJs^y&V<E0<p;e?;1onp7V
z71jQXgKxfI#|3Uqy2YaO$&yp_^Ta5xnCG-v^A%oy<&-^N8MXY4NztgE6_(s|s|Q}2
zoD_xRMvM&PVW4Qc?o?hLU1Ocj8MtihiQAub@dIfq?7WA3EkE|dcbtFKgMUk#hOKz3
z`~7y`XQ^BLrrArV@0Dk4{m(O&yFP6y*4@7I>KD#SS>=4uaKwmi`_DV>STEoD*EF8G
zwV%B0^V)N_^_Je{w1pRa<@t2D;1e6|{`U`UJu@u}!}&OB?NJk6y6Kg)Y+UEso37pb
zs+mdASkykbTR9=GvD^#>PI>FsPq==Q$!V{#D7Jb%j&A#J+Ik!6n|J-pTaR5d?VOk1
z!xc|_@bY;l-1ck|u{$2}+RxTA4_Rxg$EGCRV&4ABs@H3K3CGoKUevU>Rkpq%zj@1L
zy<eNR|Dx*Mv(z@%ZFS`2q$n&(ud?MlbJC@YY`*jN(q3cP<IIwO-)S%W^|pz3rA^18
z-%qYeUfJ)pXU%)3*Ylo>m-V}=&mDW!w`b1(rxcbIf60Au-=#jh^-DLWLj{}ud)uad
z7oq9DlWO|!hr89k(|B3;c6Tgy+RR_g_ehe3b=N=C@9Bw`jojjx=hCL(e!HHz>E^$E
zK5aTK@P%#exO~4C(xNe(=c3C?PTTc&yDsxo8nId)7M}IfSN?d;lr&;K9y+!AEx!5g
z-@Iqbdy}SP);iVe>z&R^(e1vq;`w7&KJnosFU~#JEqARn_3@-#96RFbJ%6?L6G>id
z_x$vp`_DRa%TGO*<i&*^9sPw<A9y}#8ZNQmyi0uJYmX&WVbS&$%DZa%!K>eQ){|+(
z&Ugs<EW7-(e@yt&ALn~6Z7P0r(!_h#JMr1Hsn~1Y%eJq_mN{>Si#B~EZ7O#9J;1uP
z>-4*it=g4kKU?4P<0TjS&h(^PEK9#`<yQ5bxU7CU)qmRu|NfbM-ugn)YwT42HRgZ&
zL5E-dSlU!Pa*>bp?{@X0Y1!B*zr&?n+0X5&lK<FIhi<*ZEq_mk8s>dYS#-V7{)>8_
zqEk6Vor{Fy>Q#A#yH`2y%sUnv=ryK0?DfIbU!0b9ir>5I__cREa%$QsF8se6t~1wr
zr>Cv3Q~mO`yJGA2%)V)#>FH3xc|ZG(?|$y->4iVPDEfH_d9CI@@BJHgJO7}g4}U(b
z3NQN1V%xvtT{F_A;v1iSb=rBqn2|>8j)z{YBSrUn&7ZGWWAO(T9EdnHc_`Xms+RYq
zmrg!)ixr<tn}$u@r_+1JIeUHjxF_yRi^i($C+zRAH6LB|lgG?Xn~saVZS4M^z3TO}
zX}H{qKe_y}551N)9h=5`yZZ8*G51|_#UK9mLedI1`2FLbpE>j8Brmo)hi&(Hc&oqr
z;QqI+{`Hpkynj~GDK0Skl+!2dF(v5~+kKy)F7>~D&w)4X{;{b^Q?V{QcB;n`Q&xZF
z!z(_YG!@T#cFm)Y_{UR8Q?aV|T*~UF({tFe+TA(Vv~#|9@#AS;{N>XhnsV4%pH8zd
zubfWnO}g`d8>YTCHO-5|?LDF|KDX7F=YRNA+Ei@%`_ox|5wHLExKR{8z51S^JiTn(
zf8;k#xZ;R?(x%{@KfLywyWe?3(sZoqcZ2L_m80K2=6!GZbJ8t-XqPp&{oPBqB|`<P
z{(FYq_IK#M_incI*bCCUX#T6hZO@&*|NQs;G3geM{@JpZoxAs6lU7*OxLWnI*$!im
z|Mf2(Oq+&z|L#lO>V@CaDyy9J-G$S~Eb-7IXZ<tn6pKE8kMcTuu95fqYIUtw)ZcLo
z<6En^Tlf1*pE=_n*X{AQq$~{is!Ms-?RT2{nRQ=Cn~Hhg59(FDvgWzlys#|2ijtf3
zefz@IwtG2kDz5i|5B%lr-<g#*6(4-hLbKL6@s+fx*!1_>w%cmncRn<3MjEj<JY2i&
z|NQu%%U?*#!gk~2@_Sr%@$WZzJ{|7ZE^nRw{lTzZyZvU>FUy{)lH1hhI`m$D@vX1A
zZ^K8^ZZT_qU9a1BO6Yak8jt++8;d`b3|GWMQTW8vNjpwHa_)hq;y<4_XQg9j%`-4m
zaKooJKK}>X&oeOGu$4WZc>a0E?eoroG%tSru-zVCeZ);^#Q)Ai+28dpOMWYR$+~~W
z^4I=(g&QAE^J3ojUk-WwrrR(4#S>|}IPa5d95LVcr_)~JP|3@i`u@uevwywo`m0Sz
zn~r(Ei_mTUT2}j_-lOTYoV;-;D^AAik%6JWKv{n;uqwGlJwK>wE{Do@Om^Jo{<*($
z_5JBk!(OjTS^ekD$D2~G)py&Pa=)zijJlOuRXnRpKUsNoUG%HF9!KxK>uKk#|3Ct}
z)pw|VFTJdNRatVE8TsIbE4=H8G-BkT3_Ix6?_pZ?+w1QZ>pv@w7X4kn-9CTcWlueG
zdy*GlJL&KpM{IjfQWw1I3vYSq&Ua5rnu=!}eDei6E^u?wG<@z$C+~jJ#{Wpl!ZUCA
z)R#Vd%!^50{Pm=Vj{U{G(~_p*xwns7;MtpIB)!J0xRdw#L2uh@)f-1mPU?nv_3yOa
zt!I93;(eE2kraiU>QPj?zyEXh$!qL4-#}GZl-_lhv(n5L$Nb<Ai}W=OZ<?|F?Z<p!
z(Y~hQy(4~e(RWv0xUZ?W`s?R>WVxR&+}CuR@WNk?xa8VJ`<jLa&3fawzkXwpzNX_h
z-?;c||GV45eRaWZ<8|GBPf8!-YV}>TR`*v}a?`CIcx`eLG4jxf9k9cQ!<&IG?R3Ps
zKR#$$8nL%LEVknTU*75JX=%jHco^PZT>R*(`h8H?Pw^uOB(8SjN3S|=-o6n-1ZTYC
zm8*{Y;^@AjFg!1x|M0mdf9z8elcuBR(A{ESO^q+M`*)D5u3y%7P3tB%^t-alof_`q
z-}%M#zAMg4ipIS943+mCi{dXYZ&~%i-z^tae&E!3Hn?TU32E6__IHy$@r47Qz5J2C
zq;<h-9$)sxxp#a%DGJ;DUVf|kS$0;|^`hcrS^AYNcj{A1y?udgpH7O#qV%r&cf7Ln
z+xM%V-QasCUY9f#i&|eoIg=J#`{C!;eJt%1_nUW~g-5^l@w92^dk{^>OD2!FaQW{%
zlQs?W{w}!adPT*-ymB^v_LdjF{QBGjO~?LOtKN0!&7%gIhC3|2&R*+GeIxA_^Uim-
z^~(Caf7x;$Snc-p7yQmcNmH?`?|z1I^V&yK|Gp2s|M9n-zxcxSQ<J9SukOG8iJ9Mf
zAt?)6*+bdy19bX#sOr}4;@5um#)eNlk-%2Gc6$Fi=iK_4ac5nY)*U+?U%Ty!+OJnF
zr|k7Gl;19|Rn_k{K9}{ofU4x4^|c2U+2LdNB~8PubB?B5@AmJ?^t$~sCtbS8<~x5c
zZ8~=Pd?1WZe?EV+Z=L+rJCm~UUB{oZ=Iq<=NvpzeoK=<k9=dVg4YzzYfo1o*$x45D
z-}p~FoVJT)>s7S;vY%g-EkE6AwU0eC@s2cNwLE0~U7x00uPR=a)z9*~-+21b&)k|e
z6<a-TE;`PYEpL-)Q^qg1>D&WV;d9epef_%=<{D@!eq{PVkItAp*FY;QdfvMIPEb?6
zr<wJ=1bh~~E|c3XmK7g+CAZc0q}%1E>iTv0zD3>isCphc?Jw_lTElvs_LsL^um0Xx
zQT?Vn?DfIbU!0bfg?V`l>n(clyMEO-FfnZ^w)?(oIR32i*)Q()$A64{GU*hv<aT?#
z+wX&P+kb4|3<egMeah(*_L!1Hj66g;DBcg65e5FG<tcCd`U%%>GC3_8yWJ0W+Wuc7
z>+S!xo_WYxTRk=<sXNx~{gQU|DVpEBR@3S4XF9d(PTS=jPrLQH%{P19zG%A%<+aOa
z)4o?3>U+(xpF3&k&;32AJGMJ7-PT{=3)|dr`F<~?Llv9;-Hlzh-)Zh=)_ox@3%j+S
zrnJB8tW6L7$s>PF+ePDR_wOG*a`IXuZ#?1GX{XrK=bQVF{Kg4a9I;Q@R9toJ(>oq9
z_IGL1G5mdBQ_F|$d+%mTkG&vm8rJ>2@GSi%>~qM-Z;!Y&DH_GmvhdR9&wKoV0~YV=
z7K_ra)AuPmZC8}uS#rYtLoD~^{A8n5uDk8~3q71P72ExNN0vRh9^Zeg_Wj3@?_Rxc
zQuls8I%(oP>z(*)nuXuH>-e>IJ#uQ=bd3A!a`)%0_Wj{^7VBCTudBifMt$kyODr*R
zpj*6ViH*+Q>#{e}Ug2^p{^at@KJ;4JRNUf|w@f+e;TO~4ibdOd(a%APmRIz6hpN1)
z_D^~F8?OEt$348`^KhI0E4F^m?3?zPo<?kshrDsLRlTym<3fJc?_jFF-syWGvE7N=
zpLOvAY0*5ivd_BSFN%lL*PS!M@du0_J8t#<75i7}A9>vPG2{D3z4`CBMaL~MZmIqe
z$BZAh^qb{1`bYMU++v4Sj~zem>=QPfvhv8f@)Pxn3`7PZ1CfEqKx7~?5E+OJL<S-Q
zk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtNuG7uSv3`7PZ1CfEqKx7~?
z5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtNuG7uSv3`7PZ
z1CfEqKqDA<V9osxU3RXgld|xYuYK|6Td%k?Eeng%d+g3Xo^R3qSxLlt^3Vupib_TX
zA_I|u$UtNuG7uRU0t|fV&M7O+xc|j;sN!AMt-8|&`%g;R#UU`YaR`xt$iOgTVAs3%
zJ!Fq(rl#<WKi{(V9q)Z3X@z;?e_7`|RmsImh*RfV?9#_qel(5vpLrOnZglN&C$D|)
zk-tq~Rd!g{enP*fd%0a6v;4bVIa$19$*;@xEWL;7^{jcV%XssyyLGKs7ar@T$534_
z<G3ONk%7oSWFRsS85q6{$g5kupOclBwK|V~=j=Vt`SkA7lWviou$8`}Z#jLLUmiYB
zUn_K85aRH)h&b}dKx7~?5E+OJL<S-Qk%7oSeg<0Q<?(*GzvF1dZ~mE#H6sI&fylt{
zWWae<*7;G<zdvggx8wcZuA9S+|KaKNanzB4$UtNuG7uSv3`7PZ1CfEqKx7~?5E=N-
z3~bUrd%w4wyI5b{al0eFcisV~FWDEdnLPYw>x)+-1CfEqKx7~?5E+OJL<S-Qk%7oS
zWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtPEDh7P7WVm2eyvBYa1CfEqKx7~?
z5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtNuG7uSv3`7PZ
z1CfEqKx7~?5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvqA_I|u$UtNu
zG7uSv3`7PZ1CfEqKx7~?5E+OJL<S-Qk%7oSWFRsS8Hfx-1|kEIfyh8)ATkgchzvvq
zA_I|u$UtNuG7uSv3`7PZ1CfEqKx7~?5E+OJL<S-Qk%7oSWFRsS8EBt@)7PETX8Zx;
z$BtXQf2IDF`$rx(e$4p(QE&b`Zqac|j9aRI#4+Q?E&XPBjsB7SBe&RL)nmtxJNtx9
zr>s1({T^eBt{7PGBX8St$EByFMPpY=#m1F0u=8f??Q`6WS!ugixuOMkc+k7Q@`-aU
zc|I)*3syIZYZwE|ePHTomp(VwKwd1%4nw({j`{hqGtPN1EgBm(JyG494D9gflMi|0
zq^U{7_ISuiZ>%=V83^NZEEkiP6=hbdtb4(hFN~X!G#wZ2|H^^)Uh-PfRD5Z(b-wnw
z_xvf1`0qRvW#_|H&d%%I`H>_4_WPt=tU4d6%luUBe{~u6aM7b(KZx}^iyv9M6@4CC
zwY<FXVYv8H9A{);C@}Di_5bwQcYOTuq$pf#lM&nRHRArHY|M(Yd9T-H?uy!fR-L!T
z{@Ob?EQ_yR+p*^jMW1hl<I3wNVY^t43=BsG@~(qJwO-kCF7B0$dE;K*dfn>VYkTQb
z|FYKKcIB3x7ndzR?>RtS>-E}xo7R8cI9k;Haj3~@>T|iO^)5SZk)PL|W4-nm`0Q41
z8*%6lUQfG4^Auux+7ws2EI;V<{66HR-Ex>e701-jlvH*eGc;Lie^gCh_uJhE4E6df
zpwl=PwyTP-c7HDs)_+@a@*|HfI4fy|oyHZsmW4&1;|^DOdFQ97{b1hyhD*IL&xz&k
zejiel-4rzsMd_1QPN(rYt6koHdtI-n_|&bOUh~lO{=4OWsJ5@myf&5JqT+d!)86_V
z?mL=Z4)=RaD-U=3txFt@di-|=n!3(Jz5n~~5&rwEcYinX|KItK{M3V=PA$D@`9{<9
z?Q~vx-7f3=CSL#d-zyIlK6-6$vHyB8P}ch&QEoH7TTw4-jM|jNKw0k>l%?-bm)q;}
z%ue~OOJ3fo{;^$Tpa=$r+Ie_9-znyNC$z1h!fQCLZp)qSr|Yt>Xf^)5Hym^S_|LqU
zW?{GMf2aHv-EOGG!Lst?vg8&OUy8PiR`S~YT}N5<i{h^-a$3zx)%RP)(WcyQ>U-cr
zEq<3Zo=)wsYJbY=uc^O3Dr?=WN^V)acWRHZU1Xq63^XNgYNcn^x`XSo^4DI;?KZAu
z?Td%o^;ULO)%`5{9ctHm<u66`o0Z?i>qGN>omTT|eaVYO#gk6sLc4y@YWzd(eqElg
zMm@^;JELXuT9kheH#vFZXRH?)sFH!9;)k8)3%zdFZhldvZM@l6yZUuozv}UY{q@=&
zy6wNI{@ktIwA()K_n!0CYq$ScznuMa+4*l%<PWvD*6n$4tM(yZu^bubhJkLMGj!8;
z6>XdPJ_^0N&9h?r_P)0h`HFnaIOoB%C_7m7lCN*R{#hdj%EBsQXzZiD47B?kDb_FM
z?^d?T)AQc1uVqEes$`()=PXS*Z^)}>Rqw^tt)D7wKK3<y7zq8Y)!$7B?|*ZxHRpb8
zjgftD{|y%U!M(T6)0c%~Z{BJDj}9!<*K2I|cbZw_iP!)8d`Eu%ou9n<XI<+Jw|&a+
zSqb7uha&?`ecoTR-c6nRFn_D6JXF_#s`!ijH1qxCaNax*tQ$5po^`5s);V3L@?yL0
z7+AdTZ>yd2-O&S`VqMplqMkoiC8ut2DX%_ZoNZdUs`)LepXEL<^|VW$n`@vI?(_Sv
zKeW|yqx<THMeh$fm6sJoi=MZp`A5`uXym`G=CRxN7@O8UD-W&udRE*SD%V@tU0v=E
z6~5y*h6V$9^Y6OWtLt;Iyn1z-N7k)fUH9k1MX$W}R#m;B5goHf(QW>A@Z~?=;){pQ
z)z>Sm$_~ozr*3huDt)poGu|0a3{+)rv7i6`_cn??zZ-7y>h7~;=aF?Ee_iye>prwo
zy~6jWyDgWUZ}m!kQT~@@$90XjMUA^xa_Y)!UcEZ?>$2Nr*>S7uW#w^Uzgcmjl^uq^
zvnop8EPZ-?J>;z}<)V0RC8sXqje7KkfuiCh?TeCEm7KiagKM`QJ|ad2A_I|u$UtNu
zG7uSv3`7PZ1H+erR_8BGxgVY@4Br`zBaaMJ$3S>K8_Opj^2SM1lgLBlVJO={7-6dV
zJw;XLu=)wd8OzOPAn$YDy4Gvzy3&+=M3$c6{vwwD`}>bY^OH3%O}XBRkETV{Zr@9b
z?N5H>(FJEEArDQngQ#zN477>^O}pPJPil{ySUWP%TL$uePomZLsq_Ahuif`}s;b}0
zufzN8?n`<*%dz*!Kx7~?5E&?&fnL8)AN&8)XW#Mh$CI+>xhe8{bzamIexjbmFmU-<
z=luGa`EN;z!eX?FQX&KGG0^RIZF=3lDd%bJ%}V#R^PY347w-4lEmytnwma{=w$rfx
zvfroZ-h!fp|HJ@)=@dgdXiB-=cu=?c?Z%6`)~_3{!%d&M+C?k9%8E1pDMk+EYh|z7
zq5PrDWgJ^%ATkgc7)lIO_58l<e%gKBU$^>Ae=bz@c}l1KRqeZ-?z>~VG53%E%!_GJ
zJalR=L#18Oc3;%IlqIL=@kV*oFwpdMFzWw*?{^f%L#Oe#Zuwm+eTw3%DRNrz(X0D)
z=dY+J-0r&D>A88k{ghST@27i(;W=qj%AL-8t9C`%aZ}}-zU~~&#~(0$?6}qYSMFb>
zf8=rF$BgeE_2$3h79F?5xTX3>95a60(r=d6=pWfXa*G{SJ$C%KvrpJ`%E}`PANStZ
z?77QIf4V+tDn31Jy%j&V(_={%c1ze^uYTaNUtWE4S~O<qb?P189&<zTd{Q-L>C@@;
zFP(7a`X4>&?xa)fwBNGYt#$sW)3?9l#iU)#(r2%=Mm>1!EBB_IVwT*ZuXj3sdD|7O
zZ<f5G{dcIz%i^uw>s80Cvif`K`f<|_T=t2yXl$3ytol{kUzQ&6`u{PIH{Qj1|DG4*
z%{$w*_p<G>({;Po?eg+nbiH=-rm+5l*WEPzJM&FXA{NU-QGOE6S5wRF#+lmH?{Z(!
z>-ba_pGDbumYkmRR?xg><tLry!N>gQf%O)>_?fgS%$lF7ug_X^?RA&C{*1ILta^NT
z`^&PYvajEE<j3zFx97qGo#Mo^zw+2so6J9ug&%ut(Lc_+$*XC*xY&kkeCY1wrlp-?
zQGV2E|FK<n^VcqadHt+ay}a?em7KEf7q#w}rC(mTMem!2i#+`Hdfz~gPVu5e&OBnD
zU%iwTh2Oh#^Rai${YF|g&YX463L70bEA13J<*Vp+MdcMgUH@CZJ@9A$Ov=JzZ@OUY
zvs0f+BG!wCy!Kd>orQ9&(_1|=<FW@%emKp-R_kTBZsZ+TtoQE>WW85V*7bkyy?3S1
zE_+q(9RvT~b5;BJy7u$$Y!qKAYn=_{7X7@q>Fal^b)@e1>*_Cg_3BkzEvnzL*1@Xf
z=AEx@>*bA{-HyMg`R!Itx8n%ghw;4IawvbOl)L36Z~MIV9P72mKvC!XS#m;cy7lwC
z?X&i?u6OzzHQeup<*c}!SMRKPVLu_S!?o-jqZMY(cZcuYcglC(cvp74PW2DlZF2MM
zbC$hiu7Rdv+2`G5%im#@u@C<AvFT}1XnxDKzoO-5`ghLvRkGHv@Or2H;I%9)n$O`X
zFKhh{uNQS*QI<XAm0R>VRlD*+J#RexYxlnR#o0-x*!1^yi+(O0^4n=S%s)FVciMla
z?TX%KcdAF#?K)kjif&i+xb;``y!T38)_$<=*URS1Jog&Qnuku~Yu<KcJx?lo-m~Pl
zn|Ji8e$(@lQ2$=>SX94R_B33s_d0%-)qkt`E34nK<dzk$%a%J-&JkMKQ>XXKvKMmm
z<^!=_vEL^uD=!?Xa@lLUSa$px?(*B6kGj+!D(g;D{k`h-q-?tz?s1^o@s~Y6q5Mws
zvAVQts^8PQ=yT+D<%RZ8bh%x9hI@VQT@{T*J%<hD6m2i$@sDniys9v7K2`O)ANJQN
zAK|%Xugab3AHJW^v~phi>b73UZ&B-gXoppo!*PY>y!KoaA9az_ZQfe9_PydYw5MLj
zmvFuAv>cAB+j6Jxdv^N%><|7w_TD?diX!{pCW|N{AYxV!5J3bM6%+P|xGE!ptO=A=
zF^w2l6ctRE5HVmHRK$#8RxlzeH6ZA$N=_0b=P=|UD8JWtecZ2SDLdTmd%N$Pe|R`m
zr%s*vRCRUVzI}Tp|FivZt;-bnrLDiH=hf{xc*ao`Z~t=O2aX##4@9-krh0v5Yadm<
zZgJ5oDGY1ho&52_vco@L?SxI+`PA?8d$u^4D)*beSN|SQo_ySq&lWpTv3lN+9;530
z>coYqo=4)+-tBiaajo}zJrMSNiLm0i?dx9aZRg#f>&$k3)AB#Rcig5&Sm(EFixYG`
zhSe@E=Z{qL@VYO=&Cl<?9asOTKSyOcKIfD2Wo|w6u!rXu8|8VB{mQm4QhxlLt|Jet
zzs#{Q=(-I{H^wiS>gOx;NgBg?9yjbbUOz{ry5HK`2mKvey4ywdxvZqb^}5dD>KC_p
z@_V0K?!(gNyyC{7`>x;fk>B%#-}ur$tn*&bI5MB4+44FjX|~_>7L|Wo<~^$M+2RJ>
zS5j>k*5_ikI7vGmvSlB$t+%ktng6hk-=O-(tzFXAeb9WF|8!AazeR1QYh1Q|!aiR8
z#`F8RyZM-ggWevr&T*sP&lRbDKC*ouX^Wfc`Bk>(^M%yY@8=o(ha9=-(KBu<m1Cn{
zeew5uf5Fzjt(<<m(suvER_-@{nTKsXP;OgCwsJY&ZnoFs(=AR|^{h9qoN!*h=T~=`
zWF+PO7WKXzb^NfdcWiN_UU9MIzNU7#e?KOwc~13sh$^qJ;>PuP6R&xB-S^|_$N4d-
zw4J|`QZBD!l4d7GXIq}O>!YxaSFfMX|9I}f<NIw>JugJnJ}lk*o?kN6uPx80^i5Y>
z+i{VUa=)J^ZT^28mzMX+d%aFbmmljfZueEwoew&M#jxJTo~bzLT7T*4$M*R<uI1cE
zOqF4+XX42B;m6(fdOa1kU%Jk_VfPdCK9Aqy$m_ZedLD3#A2hDtc45y?#^PD7XWXuL
zGF>0`ajo(G`tL>C^zpktB;B9$rC)}1eEaP;Q~rMUQ}u7u?BG!Yjb9zYp5L(6X{O@1
zT_1R@m%7^aKP?Y--bjjWdmmr_<S7TQGjZd)d%bY?4exI@*}~HoR~cEM#SW8nX6n9<
zTfcHY%XZs#{z=!klYX5m&y%U@mUKOA`AePRZfE*)d(i%iw@y9u#9OwSq%!EZwcRm8
z-ump{CRH8cazER)kJ`R>3|ek`e%Drh;h|5>X*8>JPF!evP9v&v+w;gl%cIIO>bOxq
zcU0lBy?f5CI@5t}&k?x2FEH-)VLgwMY8>0X<F(xHeW`T$d(A(npY3x*(DI;pgw@XN
zb7I)_VdWk7IB~!3&({2c9*1#hmz4XUZh59_zYn{gR6ke5rF~pJ-;hVD`*pgWAM@J}
z<4W~<EbMjcH(s`!58U>pxYqk!chaAA7&UG8J+<81-xZB|ev<lot!&FX=<}7@?mxf3
zjrurAHGb6RkyP`O?K8DM%RJodRIj6w-ahQ>Fu6|S^7$myJi^k+wjOg!T|0hWg$g-w
z<Li}vynEYg<#LkBxO|@!G>@piZwVUTwvOG(qdG6her@lsk>j=bQSSErkk|UK_7CEO
zb^lQ678g6L@wvr``uy$pzDU&mVb8CdzJBY&&d)7gSkKkE^+SEs^XoUhyss0M&3IE~
zSox$oPO8_NnQkwy<HNH3uFtsnbN=<qsLwO$iXYYWThwvWrH||<)&0-bzR%!>$96xu
zT7l%p4d?86@10c(Ql*S5ZMXP8-t+hAdDm7hu(9qVZ$Efn`8^A4w6&Avz54HY+zYQ(
zD)36#UfMss+;#D%Ep{)kvFSUL^2g8Et-!`dJ~_74@>?ntxY71^aw)gXd)Vb+?duvR
zp1(u$y04~sp7r`U7(c(~Pd9(p>m`rxGxMsK&Ce|lTRY0-`6H-p%OmXl!*4vdc}z+@
z>lpgg!K;6&$7Q<OdmT@ze%zN!8d)dlVyZgFwLSY}Qkky#*M5#Y?7YhsJLr(q^+MiB
zIWM~98T5Ka?nmi5uX*h!=(uD%InH$1@;-gk{RaKK=rymnJlDkWtMvA~VpuxaKL4b<
z-0S|Ke!<>96_sw*b>i_pg4|EjZM*HSUiZ<s`04j|{@z2t?_k^*_WUO8b9vbLhOMV9
zURdYJq{Wff%hI;#6_;{8k1GHC`8+Q3<CaIzd9%fp&m#u4gT5yuF72Xz?}tr)xALUz
zk1}8Db*$yR+irSLpXzy&IJWClzj;f4se{eV^!X#(eut*xDctb*ZW|ocAu~DzJwI!^
zKVOffst5afQt0+MIcR-Y&zJj+gYItcQ}X)0W9$7Z%XNf2AFJ)}{!OO!V*R)=)qNse
z?cK%^v_4(=C$%5Hy4tSOlTz;X{dZjIZS!J#u347bo{zSb7tI-2y4{je1zyQ>&_U_;
z{B_v%QXe-v=(wJp{Ll8swJuW-HiPElwViFfWm~!RI?3gGLv--__cYwD@7(H>`aW~F
zamBqp=sMAMW#&A%ZMq3vhopXvaoJZw*D2=xv5!VYzx%Vz-|Kado8O`l&uu7Et$abc
z7?rNXjm!0MNzWrL&rb%;!|&&1)}M_@-$#?ao}+rsFI{n?x}RrTuau{2d{Or!zo$)_
z9ko7D$4|=rH`aZ+81#7^ZC9rMCF3_aVd;?S^C)p|?{+;C_Ii(cKaug}^@-d59`<-`
z@odj0C8a!F&&}KPljTXjj&{pmuIsqjZpT%+>)rM>TYb9o)czj$@|IVBcD(8k^mBVj
zZzuQ5ptjpS6;{1GZ)eKR^!HeC|Gp{R`2}6KNo^<3^ZlFsK9c<I`JMCX<sP0}C%1H;
z{Py{|Wq$iNuS#CEyc)TsdgtfW{LQYPTROLNn@)8e%+GtdZ_8!1ONXvQjix7Dbxw_W
zCR-W0U>eYq0?S7|u>0fl$`k~pr%ux7)aSnH&8iM7a{SVkV)5kFrKWV8=me!LZrJ73
z2Y0`!?DZQ>R17;`jh9&kqFN{Ej{9J@G2d2iyupOcux97p)@wT6^N(Lmx)@g8*&3&D
zmC3IySiIRJm9Ne`yxfMjHkovD`ro>&yyDJPjvIS#Ij7mTdo6anGF#VKNgdbvfra%?
z+q}k6hO0wKT~9i0RUk|SUOJ`fbMM@`$|Rk(`#@o8RT5&kUC)GFUlKHp(-^Os|M<bL
z&e(30nbDz1r`bPM=vB5L+Zm@dA7w7bHd*u8!{5wwQe~7mYV7D0Xte0myxnhE>tu?l
zj&RV5;k`%i7~~|5S1l`hO8fa{P(Y@to3>X4lB7Ux-(F{%Wt*L_nIsJ~*X&i`%7b=)
z`pK8SHkoGjDy$<@1yli5Kow90Q~^~$6;K5-r$D-%tH|64wP#g86;K6K0aYM<3UsM>
zP2XcaEiy^vPw&mUd*}<>Os45|#;&>kyn|QWx74Icul<!2Kfk|w$~OPFe9zORU(u!|
z`3sz|neKIb$A$M5H$P*B6Bg5*f6)GF?f&ZJAMLZ=$uxuJpK7~P-tTm8jq}z!NoRvY
zUupf=iL0HYG3w`PQe7Xiz1RB^U;TTJ4<0D9!U>yU@3U_4uI_Qo=5{S-IoZNNb=&r<
z_04i8shm}@<1KTX*-p~<YSqG;O=rw=lFB27H(h_<@jp8ym2w}<7JK)u6V4rT(hOr`
z(#HAply##<O)qki#=l+MYuy#ctap^V)*<S3SmX0i6H9Gc=OmR;Uq?n2zv5MkPHj7N
zoul-oL(t>kwk^%xy8GaA1xe>s*X;lBKfYM)#En7cCDZLXJ=){`9TycjQS<d_*B<+R
z^@WaC4xf5RpXDRRI$jyK_24z1YyN%CPhC3BbJE4A*V{|CeEROP!CM?<Fdd?v9~u9F
zvYq#AIIhr1Dvvy>_r06fzv(2E#V7Auzs~w~CR3E_L2Xqa4hlRnvf-kc)#jU28Hcg#
zDu=amKW%tRp>Z<BUFGDZ?D+og4}0RRaVFFBlIyN~uAMXWoQqp6caq9Rqu<=`lOHCT
zsJSccwCukMn11j7P^!f$C+VF0;r=~4{`{enZQS_yr)}pRH_N1(|6L*f`Cs1~_9z}Z
zV4jmH?p>+#*5N0tG^tXKOKnv^6;K6K0aZX1C{?OV-oa<LJnk2x%uXG2jjIBxfGVI0
zr~;}$>I&H2Ywfi>?B~kU6)))Dmx@a}x98=xKAPWyp=<Rw8@6lw>L8O;zVf$UN1Xn7
zp-C#g{<?U|p{GqVap9IxFXvC2xX|&+b_34dcx{V0j*Sf`3}3h3BTF1No^#~cecB9J
zZ~W5l_f~%2+k4cA35_lp`?aG~9fGWb9d{f(WA)jSjIw0ukopy^w+OFX_Wr-#cxb7U
z6ng#pZgKbH`YCF%3c~t3Bw_!(%S^{>-J|Kek(X{YN#TYX&R0+D`-|~P=09Cb_jPEx
z#^u&err(2I()y2j-z0uizrPFn?|Wq`Ub^m6Mb!`UoK%K2?{52>)+a-Orq4h7TBFAo
zn56RlsULOxVaIY4S0-Z>XqKvgDxeCe0;+&2pbDr0s(>n>3aA3AfGVI0{8s_D=W1mA
z9yz_w9DM9r$BhqtRn)HKSxX(iY~KFi$9qp5<7ApuFFtG5lyfJTq_fJ_>W_UrWR6K1
zx9(eT&%)M=Owu{C`5SldIDLspD$g#v`R6@%Uu=@jHdnrL?ZjJ(j2r*EGwT0Tf$URY
zaqD4qIvw+blT=pe_2Sm9PwX_QGF>w}`GxA!E1WsSB%OzSy|G%;Z#SE4<EhntoN?#{
zMJ8^n)@0Z#+vgRVY~i%y>I{Fo>Q0k%uAg=JOPBUuV&cYh?Pqc6=ZekGUbJY-N+TJU
z{7PCLcRQE-{l>zPMyWdJI?&^wr0xT@eLw8qqxgNk@0bhUd3pO+Lyb~(h;kj+*1wjk
z0;)jt3V8i{m|^)vKcaZ_VD`W8_@5Vl>?Dnit4w}v!Q#y(seEwdNgXe5JJBSSwNC!~
z%Re8sZ$Z*IW83mirXIUnLDG0w|5G1mGiA2|Ww<(|x+@>>-QzvqJ?keYZcKGtwC_^!
zn!d+;T4XZChR2n^;pRbm6(pT2uI*c=R->u~N#o~JS}tqYtVTglvW}FhfGVI0r~;}$
zNl@VCR$uMk?!K)~wy@eA3mY`pzRC%UC1DNd82_uly{*@DyyqXknz(YrlaJiC`O9(z
z%5Zh~?+VcWm1G4vzjoX8)9+sCgv~B<FR1=Ok0K{(Mm=|N`+d$mdsl7IcGOZ67QZ>;
z<h}PBHq(T~sOLk*59>YgGHz1rDxaNN`GbNLPS^}<{6`+MdhesVmMusc<NAImuX)Yt
zw7qVN)f-IMO!Yl>>HhnqVfB~p_l`!@U!VJ`H>*0V$VnQ#{+*Pt{leM@y~auPcSmt)
z@Adt`ar0X~>Ve%KpI4?JReGK8RQ<wU?;7uftInx0&$#K3uIEKkT_4)MWGT>i@H=-r
zb<Q*=Dkk;s7`f#g)hc=El&a6YbL%P-cSbdS88=<~X4w5ibv<s2D|L&Do$CFT5$8>K
zfBd_foTM^opPR_{wUyoeRn&BYlT=1^K1)~JRPUQ4y}jJOzIe0i)V@V)Ow_DCxcgOQ
zuit3WMLEuN+2_=KYS?ZE&U0YY{SK<%=){^W%M|!!)Om#+-|cf?-0QFIan0s-EoV8(
z^y(1SedkQYv0axYr9ABG|G32qd!LIsUQ~Zy$rd-M*SX~F_3wg(J#VRAXWHf?Ddp%H
z_VG?!zs%I{d!p*s?R@4}pLF{tWgVwG&veg2(Efs6uLtc<+htaPpWd5y_s|!%nWWP1
z{SmKfr}};dul=Z>DiDPNsjgpbUs(UN&1)RhL3P;GI*8(!)tIV)DxeC)Ndd3V)5q2C
ztJCf~byDqO6BXk$SDH)a6iE8tNz(jN`+HZJTY1UqxzT`q=g#c7-6&NDT?cv`=sw_Y
zA4t}2uh~ikQhh#L+pjJ)rQ<{=RUOjxylYhb_<hbyu0v`2{T@yI!+kzGuJ=zeotIpv
zYO4aOfGVI0r~;~hDxeCe0;+&2pbDr0s(>n>3ZzGY{>^^xC%=1s=e&Bkb#m+GmhO|^
zK0mk2Z~x|1$*Y!EBezuV{JfgK+4XZv=az2Msm_D>c`x^ExvX~SQt$7*_U9v}O*DRa
z;g&MR=U4d1p&fRIM~h~)pE1+Hwo$hG@QvBu)OmZj$rNpQ%JL7MT6z8%W4D>Y&l-R5
z@OR}4p6j~yrS<zR`XlePTbA^zP-e~_ddl|Lj30mEP>+uIx$&Y(8?S1%V4Xv|8ygiJ
z^w4w5oGP`h?|$3SYaDFq=|_@!^1)`@E#4T~{<7=Wo0%6ZZ8Ndu+W+BRFL(X*!seTd
zEiU7ro_?fFIsR^>y^V}tlDM>EypVE8xn$$k>pCp!IR1}yFYBdEzcL@j?UyH9aoy(m
z?=3g9BR=yrV94qY_dQjyz&39(5A3lgbRY9vnT3C>A9SD}Y)HS79M`hk)=rkwA39PG
zDVOoFX)kp`57sxF`&^@U+LkFctmopdx|Ml)N0|cb`rmwf!N8`a3MlXO=-Z{8=YBCz
z_D5XWOPg{?|McVL&pb08bitOi=|KMw{UjM5amf#o4}MUtE9Q-Ssh7l$a{T4`q~0s#
zzR7Xy#;Jwpc6_f@^@76R&jU7+HzaRtNIdrI5o4M(ssGMC1?9&-{lO<a_b#BFB<*E6
z<7eKm85ix)51V<S9i(2i$DeZYz+aMjbcZqz+Trg;)}b5wnKiQxy`^S>jp?~;(_W%d
zfOW$<z?SQna@)G19<okZ?~rmyj$``8A5u=9*ifD?Ild^zCO^g@ZOSFdgX>T7NR{+M
z9zn@|Wg~gZdZ`O~LdK0vzMKy!my~hoSI)b%$y3I&+2p}_ROU-OSx!8z8yHunh(3^Y
z!1^M8X)_MSiH#rECD;%@;zGucjUS{Q+WaVoGC#^~BtGr1A?>J_`N{QwKg1u}uJ>7A
zw1em<Nj~JsJVMIZmmqPt4x}7^^2Fx&!iM-kSx)}M$4?SJDAy6TB>hRs<DC6omb3mL
z^)f#hA3e}n5`W@gqXV{$ws<yww{r4;=nN@`=mCk}v(5bjZaS)JfsOd3i|9n2jE8dO
zoqQ<AUzU5d(MkFfSCTx!BI{I=e7us+!;pOeq9^+dHpd}0dSJ`_koLq4o6IZwC%P~X
z*pPBb;xJFtOIyaF9lCRz(;k0FJp5@-J>+v2`wgTWq(9=&FSaCcX%F#}?Xe;GVN2TT
zZSnn<%Y11Msb^i14<ujw{gU>O_SiPs{3s_sN$QEq{80{B9~_5L2W;X%*$&-^i%nd4
z{IVZVjvvP(He^4*hU7<ojFUViS@*IWo4B_A=m$Tphp-{zV0@7FHqu_!Q*I;vvYvh+
z^)`R%@#pggwyc-)C~exwa&#lUyf4c605U%ieOZT$NA^SitV_9GXeWsu?I3=#UamWN
zJlbs9QO`b4e~|v5EXRh@k8;R3AoGfiZrIdIQcr(25)aCGXa1Q#`eB~2=?~iaqn!1@
zxGA@_$6xx92b6yJL;AB(`mql1r=ER?a!5TyPlzA;9ez;ePd_$YWE}h;_1N^oyfN?8
zQ;v<U<n2b{(=YK@M^TaE6tW*-M@{-;{OE*^=)iGAxg>hCzjFVUd6ID`m!w_PBv0;x
zNJia%+~PCuaNZZ~e%h;ZxdJ|yF@GE%><jEOtOwQ+Htq0dolwsF6IZTB;>rDm_VmN~
z-z}bOM|>EUd0`$HH{)bpAi6=uA2#V9qN8m-W&ik7&iR3INPNc6yh!3lIsP`{$NIvL
zICA|kF7`e0BaV&i*L?3U+jAb_xS}8OAg-Ku@`d<8$|1)k^)}+qeke))<j44kNB(kN
zr9bWHkGQmh%maEu{2_kW^haKj=n08Szu2-J`AW)lM7?Z}O~3frh`%j9?cD0|^Go`t
zoz3R>mGh6kB=bW%@~0e|bprYPLOlE>sV8srW`5}xGT+3J<(#irUy{V7KaOkGCHc?~
zHvWtkn|P4^<nc&+;?nQE@Bi5N*DbsM@%$G13;m)e{X+D`#?Osj{cV2eZtDkK{N~5J
z)1Lkz<&bjnz=oVx(VO*7{^*OJESLR|AO6%s@~{y-sK+K>H_{KpA6wF`Uiza4L^swm
z`atO?<4N1?`i$cT(l7HyKGbuZV9R=0&OQO<@rk{6!Ba=oe75u-*U#hu3x9tu4BL(5
z!SN2c&hSgxL*g_3REgeD#+UPt&3t1+^1)`^a6Dnt9vgp2jvqIYw<O0!rYXlI$Bzz_
zL&|B7zMNk;|4IK;$7?%&<vNji%XZ{Jev*tYYRWuVk1`+WM_%$e9X}gcuhe4`kNL!g
zoNvevQZMuMYLf?)?dXSk`oqS*(B(N<t^d#ONz$bQ>k2YoUP-&CNFFji>x*^8x~JT>
z9;t`451Z_3@_f(vo9hOyBiNV8i}tcy9{1Ri#Icd<0UN!pKU;gwD~zA(FUlD=Bn~!|
z<;0Qo_{n<w&|Q*wXMbkB*=X~p+}4iu%Y2{<`Ei}e`4K<JdZwSCw8bHBTU_#>o;a*K
z>g9UHhK!Rqka*+~l;q);w*K7W$npEN88`jN_Ocv*bb}|pe(lh1_4oSY_aXSPk8r)m
zJkSsA=?8y^KQ?6jLXJO4`XN63LGq*?%6b_G8?ygn6NmLC$3fnZ<Bs;yUzSUod?D)z
z+eY>UTfNtE#_Q&f-ZnqmK1iN&T-XrZAUctsv?(VaY#Z@o9Qa{V52+^~<{45iNqfIU
z2ijqC+(>=#^Ll@Warz~?lZTBQ59}A%_+!g@$|3%Y526S2fxnHMA8Ch9tVc<yFZHx1
z54Sk9lSD`2K;khkvYoW?Cof6jG9GLw^=181583~)C2iv&p3DoIIOM^6QjQ;SC}%vj
zdi;pTaf;2lBX1d>xWth*^FTfGg)QS@6GygZJTf2J5g%KUej)u}L*h^mnO|&4+C%EG
z$p_ok591D-<ReKQ@;D)1>CZf*njifVhkmdn(T(d5NIC8K9WOSN@fZj7P{zZS?Xl4f
zn>ZXt?6-U#!{+#s#~tgN^@dG6=7n;IKlPCOrA>LNWW2;d2OH6gcH||=xEMG4AM=hs
zHtUJF=!z~f9yWP1AH*X*{o_x$8}WzaXKP0uvR>vxJSg*{9HJZJ!<OyPOV;D(mGmq9
ziOYOo<41n9hqm!couqABf2<E1-P+meS=ZDvK3hMO<1gD&58eEz$DjQTT`0E^y;-Nk
zgEAg*sV`|n5A?Jhzvzixay{GDJ@Ke#J#qek96#6`U(Bcco=5F?DZo5I=2MdSm-{s3
zP+n*7xlU5XXWdZFdV<8KfBdjn-;@)N^^PBL@Mql8roAjDp6rJ><O}7v>7RV?qn!KN
zHgaE^`){^9$VcW+Imfq+>?@G{27hegQ*NX5!<LkK#p`I%Y*JN#^~(BZJyH%Sm*lv_
zkK@PfJjL;4BXK3=`lMghIsU|@p1knm`h??&^D>nE(2n@j6Q6p@vC#`#)>CdH<FK`p
z<>Vtt9P)wqVcUqGq}%#q9q_#?`z3MElldV}DEA+^PRW;Y>Cbcgyx)ud>5uax?WiyO
z{dtL@AMf=*g+{v-u)i}7>iJy@e&kI)<i|W=<3}F&W0ME{GLEy(IQ8@{TdEf@9_q=T
z@vQ#grU#n*T=kFlqha%XRqa>L?6dl@Dt~-$Wy_Dako=fG=9h7?jybNdU#qtA#JO+R
zDv)$rKXIvd+ZQ<hLdMTN11S$m;xk_|FZ71^yU}kx%s1;m+OnMWPTr8~H%Q))yjXXT
z_K^N*M?C5!nUA=Vag!gmB;$pw2W&~!fh@<C@ukgu30W>}sdu*6>_^Ot@;~oA4yps|
zhW6O3V~%U+bsfp}>>I>ke}e2!Q2Mc-aGa2bw8=jx(IKgH8#j7Gu6wxNL4WGeo%vuM
zxgOzqhk0OsV}7{4q&?;MlP`L+ZivJ64cA#*NAX@*eiuZXYu|ad{Oz6gD4-qJhmwrX
zmKWvHk8;LgBYw0eA9Q9s=#TEiqrD`0<1h7<{@Cb)&eEn_5<l|7X8xEj>gf+sF8#1s
zU&Mp-17%+L$#QAS_SllL9qW&N<TzwGwk;0*5eI+DWjk!n_t;)Ze2C8IPQS#*#vgj^
z&#NE0N`Lx?<R$$wWs_IX{%P+<beELthWVr%oBhT{xBl6e<UYgkEJ^>AV|yjzl>V$!
zN%~2Z^oMTPkbbeflJP*+JGx^N4?juSKlQe9>LKx=EXO7vY~quz?KmVZezHCF5S^Jf
z%AxF+crriwgZL4La><~_t*ss7pdXtb<@k}WB=wT`F%Q%e*NuMtWjy+o<u+UPM|)W=
zZQ?OrNVyw{XKP0u#KE6+M?1=~p)6;Ar5?(4NPm*7AL2l+>mb*S*!W3u9Z5OY>9mu?
zpK`7vA#oTF<C8=`N&JW-$$sFM<imV0j-(QuY-GGaiB58Uux<TP&N_vZvu?<Xa%@S~
zaZnPUyls@nnVZeHWWLyAFZjp5Z{K_MAAkQx?jtfj_2f%E^B_sRJkQY{e`%ASB;$e9
z%l7n}F4BM0{h=3>{ZY>PK_B)fbdzMC!e7#BJo-To=AFE3`BIJ_WW8bI7dFF=j}GXF
z9@sXLhoo)1($B3Q>g9Y>PF}1Jn@zbS<EB05TWtLCgOsz*WIojUC2=IfiX-D<GcVYZ
ztZU|#;{-q2L&_oLl8g&KDE-h`mg7$xSubt;WjS_K-)E2?`YBWUo+MrRvCf%4$bQ7S
z@tP0oSRPNbr=GZw{@8!mr?AP3^)2@=>KPB$i;#BY0qLK9h|6^&*ZIUJ9)7gP=5rAF
zQ%`&RsK<u*k+0N|c;qk1_+`1Y>7Vg1PHg6byde46@}QplvAyOc^TZ||{X*&?<@AS5
zy<gIvc(g|+$oj!%9biMsSzpY9Y>&<NH$0Cj<H&e??^O8v^Gtj{1$nLrJ(+iG=1USk
z@|FJNMc%Z>pY_D?PkV?zwk;lhlC<+0AAcz0vECUs<HBY<w1>1~+_t#*v7X$>`lDaw
zo%r+*iO>3##838%ANvlr^ru{sJSoQ}PsljXmHDQg{IMZ<VB6*!KgfI#7t$UQSCTlC
z(;l1gU=vTaLl?@ulDN{Ja`J(cOUm}lkJobPXR}#X)H7~L>LKm$qntdcXMIA-!zTTb
zkBm>br1Z1dvK=<#w#^^q%s25VCk`YY#2-@5@rbUnJ;y2MFZze{4~Z*n%H?rFxf?m2
zrJtNH@@JejN<VBU+fgnl^O3gHfpJid{^%*^LFPd{{mAhXhdlA4oPMQEIh6I-HZl$<
z<H&Zb12;d~;Rh*)loOY6V^a^wll9HKaGs?cKl;Judu-`PJ^qsPFXxARIj<03=0ksO
z<hq4?h)e(EWs5^Q=|?$nAb!+iL-K*LoN@A80N1VTJJdt+<~o`C9LkcV0PBEt!uo=&
z7uE?j>x}0NIo?>mkadbJNjzCjJ4ih?<oJP<<Im?o_A$u*0r6*?lrz4<-#;JXN8ZHe
zxksLN#1G2zP+V-r8CE{%%ea{jbb_3RgZ9h$hV{#QunzhBjvsljP9WzEh#uGwf5`Z3
z#E*W-gZ#*w_SpE-p8jPy$1CHe9)H?PQV*Fg+B1F|Wjk!?PdOwH@@2e?ACd=p(GF5C
z{jq6>jX&ks<U!u#iOu;QfAl7<B>6$&V9WT_L#{)mA2x9y<&wm~Pv*%u7(ep>sSkQw
z(qGW_NpHt|Fc0Y(Pyepp|8mL08;p(YJCOOek@diQLaqzg$GOg=J}5cPShtY={4!`B
zskXD}8C5ygE0U~3N%nhLj^6A?lK4|DZPvNlx|i*U$NaNT;|FbdQ=ctFM~**iH=?7B
zvc2@<dQ6sMGq2c?`Qo_YIOV#Ta>gUq8F>*8%K4PX5Ah)N^arUYADbWL__4o{FC;GO
zRbD4hZX<D6&$jmXL8%AjHrnEn7wxc_H_9RPa@^EYZX<DtBimuK{v=rstP>m2hj`=q
z-1cUb$LbZhk$IpUWW5uYJS3?PO2)&uW&DzAbKF9XPv!+(SXby6H*#EX9FRA8GA{fm
zw^6pY`B5(Y-Sm?F=*{}39@|Fz$b<EbO}$sjcGzCYIOvad(x%)mWgKki77u-F`w=>`
zPMAl^*|#9|96!XxhTO-)wvo7!Qdh=HUXrpsdGWcE{_!Io<h&+HKa3kv4vB-_*s`8-
z@?)IRk8)do_*2jI5@bHH88_`%H{=7QpNuc}8}y<-{D}i8hg=Uq%58BdhkV{V_sxfA
zZ5p-bAMewlo_?{Jf7(%x9@ylAO?yf9SJoHr_2K>?*Ll1*i2Jw1CmuHAXCBa%JjkDR
zkol#ayr_rrb6nUq{kjqTCCOis{AIbcnMd@4l%p^4u+f=vN%Uo0v?DIZF>#nL`VUIt
z*hn5xlYIg*ZpLM^DYub$LCO9IgT{%f9qWbl6IDNMahWHtBwyyutsVMHKhA^XCvCTK
z>SbKYp{&QI-Y*#s$2IY!9&$YRLEcm1c6@St<0tnK>Ltlf<|WH%had4F^$<TvbYXpv
zuZ`@ZTrXhT{AK^>NWbKRAN7<=%65!behW_=xlSmT<o$n9QT7*=&3r)4r?&aAm9rkW
z-e<mT<a&qxY~(r$e;aLivH#lIc`Zkue}3^^pPD%b6fhrhy#%${M_9kC+pzobnjhD#
zUj2jm$?-`WU5Jk^*!ZIxHh$PpwxirDX-{6XqZ}KO7x}PnQ4VD}?TJf&*wEIVa?S%X
zzO1J|NIiLDQ!hz9c|qxiO+BO>QZDmn9PHy9_xQ{9=tjAXw1+l7#tG5K=1;y5KWySq
z4sGM6o_3Pt5AnyAl<jS{Z1=ZThaI@&k{Sif1D_*hf7rxj+{B|CQqR1ye{+7IoPM!s
z$9ahTntICV2ck1J`Lb?^Lws!FG5?(B@N*;m;zwM@i7m%ZJ)~ad&p5~ry&?73v_~)U
zl7864#U>8AVMF{N<G0mQE{QJWDeG-E<0EhU$Vb}rL!J=bi6@C}l+%uQl*6O0nBMJ?
z+m|}@AC!y>k_QYL7k$VVTgF2d#!DWwBM(__i^n>mz3h*A+7lmAFWX_$&dm>h$)N3I
zoUm=i&-ge_Bw3F(k{9uq2S|Ku;xoUHebGj)Gug+u9%ntta_T9U<bEReV{PNJ#pnI<
z^g|r_v6WNL=O6SyKS;lvKV*LBL_GW@$xF7UUr0aL_>&LiHnI*_@65X-@yH(?WPjL<
z2R)fLIc{v;o6qlQ=?Bt3>rs+%kT-cyPJ2kXG8ykp(0S5%YPMiq%GXg*tt0jw$bL$H
zl>4QvJ>}4ET$?}fpxkG%*<U%YP|p4e@gp9T{i7G3`)ChkeBwg4^923LIM_DQ4&p~%
zkUY?XabR;k<a}+5M||`^FRr)fkLwu9xh^L@{X+5~9{pnzSL#SSnLqxL_{sCW%!7RC
zpM1DpL{F{{AlDuAi(c5ofz;C<anXZvsW<V+8=JUpWLyw`Y#aGIM?6QseHZ+_#$jEM
zH#*QhD4Ad45g$FVsh31g)(19z%m+5{ApVf~B;Mt%J9l04WVHgWBRNi3KlriE@T1)s
zpFPs)={F86nAH4*o2ITmpn&fK=#OziuG8glLSE>FjXyTm#T<9ow1c!`J&+H?k2sXm
z4@3u!qr%^RmxwOZL)sCKerQjB(hq;i<$O|3JX`;=oPPP;2<rmUFQgwnU&!@<-t@z9
z*R#$218zF1Y60t_@OPb9H~7<@zk|;2IML5W@`L1$p2UZY1JX{CII<j@dPun>>lr`R
zJM##shqiKA&weJ4clKfSC*tAnM%uCe(Vp?4FY_q%Bu`2F(1m$*%O}-(@|STaXS}Ry
z^2Q%O?4U$HNpz9NF*;Mv`U%><tsUcs^h=)f56P2!@S_|u&$1mh?a&v0Y~}$Qay+w-
z5f73-e)K~;Z2YlhJ>}>^J$X=W)74fkk1H83Xk1%=loJ=qa_m&e{*!7PZ6EIa3f3>{
zn&X7yK$7E8wxd1eP?k$u`mqk>`e*%6E-Bkl@3vlO=jP|No_@WO<DNVq<E5T>)XRK$
zj+=7nkIuBG9#YSBgX~xO;ZHgHJ~s7ozVV}8(k(yfkIgzIKK)|LyeP*H5)U0E@uNR!
zb9_+G@x*zD^PnWxJ%ztt56d{%^v87;aVUrM59uHBxrg&5?HE6LQ4Yz&jm)DYaUuON
zALQ$o=#VPqeA{d}E}LJv%h@+%ew1@uqc`Qu2R7{~XP(#}D2J5eM?cgPk9t{-Eosva
zKN*iasHZ(P{`ip>#2+#aY>2-k^^ks8mmEhjALc{);ZNT9aeT0Ea6C(!c#!=lRWfeI
zfeqPL&;dWjfgj`J{tNpW_4q^f9okVYiB9<0h>nmr_+!ibu_1os$NdZTTgd!C`lUbS
zh53T`$^NJ(p0v@^mJfMB;y~g-@?gEu4l<9jKWP)6cG!~iOF8RG#>0lhg?x^;)#F$A
z`}r3?`ePkSvQ8<-mZYBjm3f5hGuV(gwtC`1?z>{AO8SSy!zM5GIm*c&(k>`DzR*v`
zp<nW$9)I#6jwJg9elnh|96uZ7?<KJ5U+QLy<5tdi-RKs_Ydv|R3*$k5^duk3@uwf+
zFi*t6mj1+}T(+m4{;=tfc9dh|&wj%`M?UP&_|q@-ltbd-M>+kH59bSP{A?r-@{^<<
zQcs?gW6OTXn|z@3!^WR@Q1*i@^I@H!1ND%2oNp+H#Kn)iY&QDi2k~b-l+%xmtQYd<
zJR<!$e^E}mpp^514v>5p7esIRB_HZ(&-fwrP_AqIhzIe<hV%m&5BU*~_FPwBL*g+W
z#!WeKu<>I)u&E~x)(JLo<haR0Qntg!kN7sS|Jmv(C$5e7*+@I;(F3BBB<tLbwBz_?
z{OE}dSr5deo$b1q_>ee|aycHZ?}*QJ9;AQP9X6z#IF!?$v?+)5Ltc<^;>S49i+S=(
zjw`?Ig8H+r7!NkI%`@frF@9`xfXp9y@p-54_v?7B`_Y$qqaSFCPe05jao8_t56M@O
z_~^;;O*<&#Vnh5O{m6RqM1SfjXP<=B6BirO4xK5N#1B$0^Q4@0iEV2~Ui2^hXwP_=
zFZ^kbAGDPdAL36uY{ms;Ipd=r;$h>@I-_0^e_4)A9Q?2$<xq};aybtCsF&-7_>g?i
zUy}9Daf%%mvR|Z{4?1$(Lv-{?bj2V2u_5bB5}hS^9tRfwzF)eL`Q&*zerHZQ)(d`=
zvo6?wS?}nA&H42Gz1RMH#I%Wq@l%gK`f|O_eo8r{J@L4H!=HI34)G!Nv?C7rQ13?K
zLHx;sa`MKOBu~Z#$&2<j;s+TA{>&5ok|$(dvB@8s_UyyNk=HwvL(b#a#KVS^lRu;$
zvfjxDotQ_;$y1JB){`IWn|k7shdlnsm%OmijdJFdddeZ?<jFXwXCBE1e@H#;$rpdw
zKl=>vi9_D>52^P`;>dozwh!BnaYDug(Sv!93z=8)hLl6*oq9?9Bxxtf`mqsz8IOKh
zchsXV?V!||c98yQkIivN9Lj0OyizYI{i&xNc|+>m$oWtD(I5LH`wZhk2g)UBkG@bI
zU*yX;$b<Imhv-B-B(9C}xWrGEV^dFFkp3X~Qcu2+a!CHvvmU62^v`|-sfWa49wg;_
z$a-wz;Lm(h&v+s82g!$V*=*V~9`r*8>WPDmA8{c%kw5Jq<KX(hHXha&>zelTPe1g_
z^&r=q#OJz`@4F5-b!XA#qxLP>IOYD1BVRo14-%ieA^l-P^2Ub5p&rV5sjr+j#s$%v
ze6S(q^baW~57rZUVWT&BLUhCrQcpiN%K4M+7`Ge;`ryxX72`k`$hwkwkT3m^5Aj(?
z>@$8z-tzcCck;uA<SA`g?$#cEuOyyV%KXqxmeW7;BZ)q=^Gfn`^CvIJypk8|2-_`Q
zSoQP^=|3nr4rotYNpxc#I1aEO<3u;&QqFOMp2WqE{ek@io8yr<ltcQLWE{k2UxL&_
z`jPpOC*xuq<U>9E6Oa7?Tk1wR`OzL5(hvUVLww?5;}6LXGCs(-s3#6Okq0CXY}#Yf
zuN)8Uh|f6ir#<@8KjR}F?eUkS9~<!}4*fBHbiu}-c;t=EzRiA5I~fO?@wlyD`XL^q
zU-F_H9T^vT;ScE_f8yYe?$l%BM>+k<eFQ)HBQ7K!w#<k8@x!KHY}(Tw{;Wsx!w;f6
zc~edv)MHamIplK{WW2;BFUn;+;xf;)<M)chr9bX3N#aL&;qUiV=}*RIpQ9dKS)b&E
ze$-3dDVOmWuOxZ6k@vFm`(12)C(3m>`J-1<WFF~Hl5sMB=m)7M4!WVQB=z{q_LMUp
z*tBCmqMrF7A9<X}_UvQCll2^5vRsaXJY>1FS$A%v9sNS`bfe5i_Q$yC2jWLPq#QpR
zWnRn+^9u1J4z?U8^>QBYlO&Ff#Iuoh#Kn(tnTM=rypVN>E&b_-b&5V5U)Yd1<U=_m
z4sq#^c$Cu~oA%gpocKfb1=)`FlIQ?w=SJ>B;K%tv#+NqvMMc@4Uz__X5M4RW<aHYJ
zDEpD^I9?(3w1+Yd`a<T3b;G_&J#ip;(=WCp^^kgONIB$lF!j==+(zP%uO#~=dO@iZ
z<-~*dQ4eK5l;h8FMmaY1<OQk6U&bdN%H=w;wZq>lZE?t3*2}zQJ$W;K*plRDE0^QN
z-;Kn>51TxROF4ejV^c3ld;B2tip_o{uNyfYIsT-L?vnUJ+Hs!1hKxs&dfLf&)U!V`
z4s1xdB>N2I#G^mXd#n#h^n>WiIHbSY;r`A&RXy3y<#B`U7T@Maxs9wBxA;-l%lzn%
z@^s}%fAk-Bq6>LL%Gr0Jthd=g$8Foc*e}`VxPK9}A9=j^jl=#*zL0%do`1-fa>m2?
zhH@E~>qg?+NPJ1Z<KjAlcFZp}x{*KU2gtgS=MU-`FSaE0#G#&YNV!)sE}K7bZDhQV
zI5vM-&%8p$M_$;F_BQg~G3hV!qdon~eDP=fvA*!99#Rj9Pae!W@n|RGkq6f$_)!ii
zw-G;xA8}<l^<K$1A>#;2bYedx58|PxB<(mJ@h2}y*<R{_E|5HA9BleyKf<OS5{G>d
z(oayz@z`vxcZ2q0YsdH`(ZMTi@!iU8<8>=1FX_iTLAU<-U9eIWPzC<00M|uwU*@=j
z9A6x7kaCVsxzDq2<ByG>Y==!glH}``e&cYR=(iu6Km8LI+vblx=*H&;c|PKLMfwq!
z`GH)op)=*!kaFUZFC;(6d6o8*^Lro4A?u3s9sZF1vFVp~*yP7~jqitPM}OGp0$Gnz
z7ur!zzw`&m6VeVcKiK3!{*ZF%$8{Eze)!W5c|q#Q4^j>(hs+DQP%cS7=p*ZigCF(S
z)U&?HM-qQX97vwb7wu&|wymD{Q2NmxQqOToJM@rbzR8#N&^Eu+(=Q~xtjDGw;!3*x
zomk@ANFFw#hmGWCqb)AwjMGN9=OoxSY-C+Z+Vr5@7EhK>+cKhehZlcxZX0*;jH~NT
z`Gea!Ki2rN%cnUu^7)i~%tqE5_1I9Jk60g4Uu=k<)SY_Du_ZZ9Wx2G8BPsoCwrpp!
zX)o*9SE*+m%ls*Kql|+sDdSKt%cae6B*&4eP5-ty)Jw8X(2@3(yODKj^ONUA{Agz*
ze$eJ;E0^PxHgO@xpCs||lQ!kdGxJNi>{r_8O?&FG<$i??r629-pL+U9m5h&mh(~{r
za%d|@Pq%n-ec}&gKlBG_kBvV>PwL&)Eq>%bcvO`NuUFZ(fWL1`d+KeJaj?nDM&^NW
zNk94{9yWQ|);aa~_iy(5XVLub`JMCX<<`xums`3|e*65~GQa(sS0%4nUX9#Rz4P;G
z{$|(DEuCAsO{Y2!=I6cKx8<_hrT_a$R{y68r~;}$!WEc0v2)phjmJ64%<7QvWu)n=
z0;+&2@P`5~wHeXqr(<4ql=0BvkBQO$#7cp6haK_V-=19Tz~>L?RsG|m<~fIdQNR9@
z^A;#ohpbo!u{!898C5_PPz6*0RX`O`1yq69D6ogQ;j%@8mzbcuug>V)p?5DbC5^F}
zKutsy2tk3T%3RoZ(zNn9ZmjU`{i}}dKgNW`@gLtmwNn2bru5mpj~j8(*3cM&BpN~$
zPz6+hL@97umBkf)SvbQC?|#zye}6gskL$83lP{byVAynLzwHgbDmVL!|6$j)|EisP
z^gQF0`5kKBQgi$=6E%Zgm)hC|y-q!NV?p26TUHnwxAr}t`P~;RFqz`$dxtk0v+N5K
zHp4p4hZX0?xBDH~WY5{g#<0ilHD1FS4V>-Ajx*^Z$9LQr)^kp^E57^7&Wf|0pv=^9
znXUasz3<00exfc0G+C<xw)02W<rkgPXhn;!e>IXlhR&+>NRurlD6Ip_vWl>uXUZ~p
zti$Q^!O&IZddzp?PU`@)LKR4`0%4y=!itw*Nj3E}D)4g6edZURG1SDBm7be3?cDR`
zIAO8pantTD?!H^j;=F>7XC1f135zvvn{x9>rI(mg`R%&>rhWg^ViOjxer3<0ZHA3D
z>0%m}u6Csgr~;~hDxeCe0;+&2pbDr0s(>nxK?S^ij~>)-Q^#2cbvWl$6BTXG&1m_q
zpC?MHE*U&Tv};vB6;K6K0aZX1Pz6eq0&#ziHP!s)Ts<S_(jBF8{IWzX0v)U>pbBJ#
z0$<(T<h|Eh6&vXFd2X-IIVJV}0(lR;=Bpv&`i*fUQ$5dQd;6?7ICe3Jwfp|9@<kOl
z8900X+=c5tFES;S94ArnEoWKb71vERHv0V?m*3yF+3r(&-6!$0{hslvJ0B{pGjF|#
zD{a5e3|j8>cdME5vz^z1jyLRfN#AD%_v`Y~U5BnV%0KI{i&KpjsRF8iDxeCe0;+&2
zkZ1*Jec0^AJ$_tg-Z^W|eWz^PVCBP0j?P_o&vxUNZr8<pUUwt+!C<pR4?f%Npv}g{
z)AyfQF#Y;%#*M_2O!TB`@~VI;pbDr0s(>n>3aA3AK$;Y|pw;nBt~h3{QN~k;G%W+|
zMio#6RDsYH7=7b`ZH^rOh>4or%U=9Usho*UxhFTSo_G49|6%QxMdR-J@)L9S^WXJo
zd-a5rx#zZ}osQaLzEP?Uaaad#2kz+SFM6W)^JQ|B|IxusFRfPvvQdF1SN0oH{O)q&
zm44rs=J)-z`+j>y%`=`^=1AVw`}0b3N4{=SrF^b2KmV22Z)iW$q{_)h_L%T|r*$T3
zKJ@C4%L}?MH(~Lua+eOMyllQn7q4tysmjL5^GvpI(cDSZ9%`}Nghjo6^mhGt@<E%v
zX<K2riHg~{*JtyXc2?M}&0}XTHA>atw{?)sYhA~t3aA3AfGVI01XVyjFX`2GdmdEQ
z^L(m}ss29C7S~qJ?*@~~RP{<)d)x0jGgWT;9i!j!O#KeCr1cwgf7EuWfGVI0r~;~h
zDxeCe0;+&2pbDr0s(>n>3aA3AKuJ)bYvH`-hRvGhK)2_%%!i+@t$Nrb6O>W^E+8rK
zOTyYp!WcjKYRr3Aww!ND3X?D=|2N0|yNv#9+1zO+C@*>Xv99eOndF4UEiYD|b84Rz
z4*cqn8$LSxxI3J%dGDod?%1u&G?P>|S+~dQFAU6glEze@6OT*#4SkQi_R(8rn56RC
zqh2dtYv1W6sVp8n=;|pmCY!kM|L&>T^#5C$E<CYX!3vWq^B%nGqn&vR9r*kqy{dnF
z)I0~u_t;1E-foSX%)hsoO~+I=kL9e^mzF#I%qd2xI_NsE9|zgI2k6+`75MOtC-&^}
z>IUP+21gzCzyYH+8|AKbxbc(Cb;^Ff*eG|cgZoNU4^==F@T<V1wcbASgvFao*p%;A
z$Tr)r&g!oUr~<(i81vanKizlqN|Wsju8_9OW(7vyJG|MLWnY+VqwViMW?T8zz6UhF
z`+@~V=}m`hK9+TCs(>n>3aA3AfGVI0r~;~hn*ynR?sJPz{hX_3<XpO=RF2Y{4$oY%
zXXEu3ZgrHZLmbxu$F))wPz6*0RX`O`1xk(r!@Hlf{@-6tH(u%Y?{v8N*RJ^PFFPyF
zcCw9eSwY$IeNtTX)I3xHRUmu?s!YCc%79_h9lzYe+;G{V!Ap!XGdhI75;S6*74ZB1
zYW(9orfkUVrma`auKRhBW8=iHdpy4Q;MGpLXdbG2b+56@95;rQ&zXmhzx|_4E1a;%
z@3obG{{6b@ky$;mVXx9Lr~;~hDxeCe0;+&2pbDr0Q7a(--kh|{Ozib^!xwkYnL4p^
z*@2D6IZD+**MS}fx)1zkABcK?4lDjIFP3|^S>9wPX$))p8b=jS1yli5AT<Tto=5Zh
zoI+gvk6m-exo5wz#8Ijaf3*%$JDU8r=D7hkruw@#ZJ(X*jR|XB{jXeY5hn$%y7Qsp
zI`h_>q%uxgYA%^qU~s=KFWq(M>i>D(^}afzbBEr&$S73@T?hYr9Ay5;^_RiGx)tB)
z+xBZG)AX0MdaDA_D$u{#@A31y=XcJlms>CQh}_bB^4sU<mig`9yefIs@@nLk>Yblg
z^EbPGZt2|8Z93I?FhB3*zAcy4E*))l8eJ7o1yq4FDv<l$#V`Eu+%%)itPW|Ma_veL
zPz6+hNEPVV{p|U#%~@oW@z5dC$=2Y36<GeY^X`mm)*JZ!vCXRO*mIh5)|h=>c;}f>
zO4T7N)<NI{L>sCCs(>n>3aA3AfGVI0WVHg{AAkHoUp~9Y3Can*n-mRsugEEB%<4s<
zgDY_g+<5E4XL6tX#kld;X%{puuD`|!i^m=~XZzI`>{(Fbt`W<Q98)GVmbeMj0jmP4
zK#5o2`YtWLxvIqshx^*~%N>2)x!25hhB+;pobdj9)33(Aetl`{?Ehiqm(S=v{)X9(
zR|ehxzr3O4*gAzvolG(4b?S;Krf1t*HaRv1y?zd@-S6LHFz;cJ^%FI{uBUQ!+Lm3@
zVw+>*No!lRf1vUfC*2IXu7b8Z^Oa8v7q?pP#GRWi{^y8&w{AD7((Ss`t)B1uGR^I8
zUGUVNdo43bXQuY=O!r&jFEFx1fUwSwu5pGI4tRP~`GW5SxgHD3ESUz^VTAo$hHHme
z;xMk<?&Z##zS@a9vqWwkhAQwU1>$;r_opP)Pz5XsT=Lr4clRq^;l!1dR&T%Zx^JgA
zVezX5_iX>T?=~}b)1e29UobW_-rxMro%@_u?4-&vBd&hD)=Qh5uo(2dL(q1XWv0cd
zfGVI0r~;~hDxeCe0;+&2pbDr0S*AeTzo$-DeglS|`04?3HaN<7>X2oJoeoD8Pz6*0
zRX`O`1yq69Dv<5(Et*j3qQn0-qEdlh#&*IqF;zenPz5XsRGfd@BmI|_&VgR<cX<6B
zh}YjK#nsRD+_2a2msCH?a!Y)%e6A+6--V_4euqru5%v3#`+j`Ykh4Dc#@JY;*M=dF
zeDtn~inizeZTo`Xa<BVsru@SGJ)vKH((P|s&td<*A!xiz|E?<4{=;5R*%B}DCq^BB
zDxeCe0;+&2pbDr0LV@#+IIr^ig~g_9<?+Y#XtKr1*H4|Xc>ZPEjHFvT;@fz_PowYO
ztJFs0#uw*YcG|6{tTfQA-&JexxU<UE(gkk3>e<sq{<BlL0!bm0nyP>*pbDr0s(>n>
z3aA3AfGV&H1<soMY5x9WCK+Wsb=bv}X^|?R3aA2!P+(Pu8WVpRy4;DHk8~Te^zC~-
zH4R&Qzj;;N8UI6`i%Ru-SK3cnx!;v{oxRCXst$2j2X32z+x^#)R_~^l)~f>9sDSPH
zKU?{8gN9FfeSDGeO568Fw(=I0N}u2A$+88KVZFCp{?2B%)6O`txcLN=D!C4bnq0Sp
z#i*X|w8gcptE85P{TyY`=NC&_yQrW4&&Itzo5$3l)wzcry==XsR2_a>2id&Vb!@7D
zDxeCe0;)hz1%iHF@cTK<?Q?adA3nUa%6V@(LFxB*N@0C&&QzSZ{yl=I^2*fjGj09K
z-zlrzq-e-{MNV9G2)aLNJ5@jxPz6*0RX`O`1yli5Kow90Q~^~$6;K6K0ac(RC?N0O
zyFK5Q>T@|!-A7JZ+>)@ik}yWwbBx(qo`gC1zd81+cHPFGZ(L)7a>NsPUGF(^rwNPl
zxn<jP&Z(Bm=a`db^B$n2(d%>XaraA_?lRw`t#_HP?fu7@D*wNGYBv2ptmpGr{r>(k
z=r*7IYW(Zhm$uF}O4Z@Fb&yTRR5p(#Q}6rNevd!s%V!rkszWk#$mTsj$L6lU_-9*O
zcJ4*v9XI-YKDVU#%loTp|M~aW?kiC}Q~_1MuL4ospOT-8v(5Iav-+z7sz7iBO3L+M
zaD}vGHY<>=>pxS!pG<eZXTI`j;o?^79i=xNviVrnv8e*8fGVI0r~;~hDxeCe0&WVV
z`hB=teCqwa*QccWhxJ^U#%We<$DY$1)gfFRIIfkdfGVI0r~;~hDo}D1@cTR5u&=+e
z^*Oe<tf08QZ_#gFanV!rPz6+h@D*@-KHOG+(%M$-AE>;=QD#Pm@K=IHjI#nkf2SJf
zF=az;Ztv+2`?}pNp4aEm!g|iic76z2?)P(%`d`qvxc(YPb%<7nY}l)G461-CpbDr0
zs(>n>3aA3AK-3Bh`*!F@m$faGW8+&l*Z92V`rQkH-rvx6`u;9`kDI>7E?f73sQ2fv
z;wSZe%VCXQ<ER3vfGVI0q^3Z&K9?rPm95t+I<9a(kEM1r`ER{qis{+*mQ9WullJ#o
zG~cuD?pM6RNs12sSFW~*lLC+K{(XZl-Y9mG$~bAMxny1ee&-$Zxh8Gr?S0g$Q+DZ;
z`6Jg~2CF*MnE1=k<xZyQFKhKy1)^1;f3x4?=XcNVoL4XRh}`<QrTgT!&(AIM+rN2L
z@~Y+4$Su`7Kd<I*_WxIzTe?lBIuGXOz1+9uvf8Di%}%4M0;<3+75J#eHMg!F@RH+~
zy9}m<s(>o+|10p#=LO|kcAaNR3U%650aZX1PzC%ckbB#A?e941C!-8k2Y<7m-l~8q
zP!bjBa^N|=->mhev2pu{w~uIDYq`lZOXA$<nEzLSlJa>{$C=&t5B_&$Yl$kL3X~KD
zCj4;aDZ@9ecjC^HG9NllRUk?QxSp$Y^xLa`Y_rk?<$>ot|K;1uiwul1XN|22WV-^p
zb<8jQ#m-qyN#(US*ZHtl@7X3Qrt7}Z{_jmcpzZCeO<1hgpu$-b2d_0jdDP^b!`e0<
z<7A4fo1Xnd;l?E<Dsn!HiXCdr@4vj#N=K<WBw-zRJx@lp&uZLg&tq}XH>&+y#`U_M
z)lU^r1yli5AUz69I(BWd(E}Hlq%!E|scdb>s%i4(`FmW^dy$hA2EBiou6F3E%nk*v
z`C;2n7cL*`q{{42BfL>Wb^oTM#tmy`Y|rEQEe|UnuW`cuob0t<^-~4HP+-k>BcC79
zW{aczvkqZs8%B(z&P!C!`Lr4I(KXl1UFM{lSG`<i^mWZwJ4)5T&pPn?JNs<&4`UD1
z7||<GQlAIO_HjqOa{PWii<^J5JGyMXb@F7R%x)cm9tUB6zSnqver_S?yl6XBKow90
zQ~^~$6;K7DQefNZMTI%HZ*{_ERQYOLRX`O`1>&kew*9_k;ir|yKU(PvN12)qab2%*
z%PZ*p(^Q`$3#+}Gp4Hwh{@|ZS&2wx#<j5oD%)4lT6O;{qJ>Zzd8^;+pE<OK~BaXUb
zkco;>Jul`quBhshc3(3r{evElLE9z$JQH->nQoV<-!<F%b=ya^UKLOUO0EL>efF-u
z&n~$uLPxF&Bu|07DN{xiRjQohl}^XAj_K07LQY)S`iLQ~pHx^iCn!_>oRuwqcR;?s
zAyanH-*+B4sc?D2eK(k-koWH;mGeJ7xbxo&7n!6otoMAyB~GgQSgP%lcP5NB<FM|B
zX`JkSkGsyV&hP*IJBQKkN*Y}i2&O>Nzb7I8u)cRqcbw&&n!oYM;OUN62Ak#Jt<rse
zG<c@rv}8Rcl||(*zhidGZ;kS=bqM==(sUmuQT3Op`$J*%oAm4Mu<kpCvm0s@RX`O`
z1yli5Kow90;-f&+-%rJ7q?(B;pbDr0sz4A0lKvjTX_f9ge|x2kPPQ<r-_=AFH>|&>
z5M=HCRh!1M52&#0qH;OPzt-W4;a|UX*0%CFN#kG5{a@7`eD9SfKe+r?CoU}0d$)P_
zKC#WoG?V&%V9oo_@0-rF9$|lPl<s&*|9*u0y*}6YS9_(_sRF8iDxeCe0;+&2pbDr0
zs=$8~==jKWk8Bva&7_-A|6PxyJr^1E_ncA159_{Zrs8=0d>L22sOB|Yaie;^)o)ze
z`Jj2bEj2%Axygh@zwb{=*F0!H$^Jcbs`Kafd>qyHaOsL0m%npl+jp|1-1fe<xRqy%
z-a3x#P~eSUZoA^a&&%W_mAOCN)_K*?iaAMThYJQjapT-dIZ5Zrn=fl!^_j{!O4Z@F
zb&wsqqkR;q{;tj*S!+}QRX`Q+rhx5z{aXIlf6pLneI{N$^UNV_COgVtI(S>{ne&SJ
zK2GLFmaLw`K5jD|&+q+EH~)0MZzx^;`Tah_=5Jf=Uf-9LZa=sAj(dH&pGOY6Kb{Yc
z8*S^@t=x88q*^ZfwatUudza<&6k)wLE7k9<Wc#q5BaphOtqQ0Ds(>n>3aA3AfGVI0
zr~;~hDxeCe0;)i&3dHsI1%C6A?{`jB7j3T!r~>W^_&rY5zs`rfde1hhgX)mvbs%?!
zLykOR&b*5jI8ig*zu%DKR6E%UxP9Lk_j>t#Xi{wXeQDC{xIb6nmS51nj~thFaeLmv
zZ$77wtNZsRUAGt;lfA2I_Nss?;6;HZZ=S!$6}=ZZ*+RGHvAk%aKB_>rDNwILg|j9O
zUTealyuTf_9rblzw#~PWM-@;7Q~_1s&kFc`ADa4i*?jBd$tG+a{yZycsRF8iDxeCe
z0)MSQs_(VQbo=ce-aevrt>q?N4Ep`yU(Z7LwSMzV^?UbJe<$ttddTnhi$Q<C8P)f%
zQNNEg{1ug*5&f>aRNo)rc0XO~E02G)(icwHI{4iW<LV#vec^P)_xt_2`WKXM*>#=?
zs>8{3A2?z0&?QdT%+9ldeN<6D|0v_zGtnAVKow90RDn1tkS*_nO!fUxnQou1>)5FJ
z@q2$GQ~v2b_R@8qFY11xzW)<-{8aB3CcS;w>)!VMyttHyJzkBM9tGmEeoAT{ar+%`
z(0%u+m#d7vuK8*wDGdAmaopo&`g>Sg|4~0@oND~AcHy`k=TYZlo9}d$NA<mIQsaiT
zj*}Xvf3x2+&+ne!Ij>%B{oEsSOZUldpPyUiw}11h<W<Y7kz1;FeqPPr?E1N-b4$1B
zROi9`yqEj7Tvoeu=;N;6ZF*6u-QP0V%2WlrU~RYBTTYx{!e**@#l3y4+Pm+w$G4N6
zpp4u2-fK|q-Ag}RZ2ZzTj>~)AdhEE{o^gWG7B}qj7uL^Qy?OVICMt%VuPxrJyXRHD
z@cQ14a@RW8bn{#O<=VIEmVIiwNjirux^i^S&$gJb=~qvi|M0`l+V|B*mYQ_YmWP)A
zM}ZffopxCHdq+A+<xQ6@?J<7D5+~ifrRSG5PuaNAxG``4Jzl&0(IVrOQLoEtr+s<2
z)BbxWX&nCAG0$x7Jkv=xQ=N|v@1MHy=B4W$Ww<(A)@9A|<yS8-sj_&*{?|_}yWAw5
zYx}={;lSJ0nWXZbQs>s}-l@nWmE(4{Zcz2!g(hsKy8pPf?|<yk#rHhC$Vobv{Zy)R
z-QxL<q}%vYt#AK&_vaq1zs97>RPzX{{q?_|(k{1fhDjI0$~){h>mKTM+PRf~H0fs8
z<B;(N-FoW3`~EW6QD#PmN&`)wZ8w+6Nh;SByxK4C)v7sel<O|2?e+Kz8b8%`gO0xW
zicyy>aZ+Wfd1(90Dj=^r(ru?Y`s<oczhvI$rE`+T7s_98#^esAbAmEe-NI_$?vYov
z9=u_hlPbf?#}?<7ZR4L^cF@0#GG02^=1Z2}=Nx<Oz`xBiVUvD7y!ayXYLkslSR}4e
z74V`!zxLhQw!dzLlWBU%Regdhu;+<qoIbSC9yx9d-h;Q6Veg~1cm)+InDOt-H(@ci
z->RuSyDc$EWmr07Do*=uZ{Pja6Z1?`*>B<{<KJvK)=4UF{A#b-h23{JN#{{3wsxsh
zYl{;#z3qmjO10SM=@pONG~Efx{p$7j@rG`rolJB1>z$s-{baNWi-lMBUixd>Pn@v$
zPKRH+4Lj&nC)?Qnvz9#?emv1cO|R=YRlip|S3mEvj_aM0!l?ICIUk)GR(fvKlj}{?
zT>r|tVRino$%M@l7oIXLcT2I8X-ZwyPKE+SdB?xk?D@VX(+o?k7X9Dsa{RzWCaL^5
z=g}K(x_-4uDnFTV*cC-HH<+-<y14(jy^lM-OR<v_l6OgErH7gyeMp^6CaJuCkN!P=
zIew$bG`+5$#^>b^Xk7m_lPbOT7nR?lz4p84s=Zd5Oi_;OrKZL?<mZJZDAV;jg{b;@
z;`rOA%&xe=ghi>ZS39cf6R-V;<#%Jv>TAFFWt<6%VdWKdoc3>B+ViH*+nG!=>T%lQ
z-*HKgg6pd<F-hnBH<h}-*Y~SU(%9<V#*K2@uQE~5Ht#{pYxg_g%t_1NH*OsJT%)?P
zhmUi@=6mh-`R9J;7ddeMnHOHYqv={FDkk-N09)Qkf3LvWt}nj6VEQzZG`86Hs}&2r
z8EKNru<q|CHBQj|GVbl#?fGfX!ULxn8{?k8%rEHs8#3K4Ti%1v=(7&DUw_0(Cv3`j
z4%_bk$^Jd2UH+YuRF3Su?v%RsO>vUO7ji!v{`9G<9HloMg0B0sZrQK-9w*OtlFkkl
z4j#Gpp|hO0G3dO6-ELF+j%})LnP9x~ruo~8doG=9{Bn=ZN8EAM12c_R{`mX_b(`)v
z!+53bc|luws`r;O)xP;hD@N{q=L(Z9hP~coyrjK<BX0RhJ)^dpjs2|p2|4SWq_WG<
z?+Z>DQ0yd?^|zKU)2?OdoJ=w5_2u<<u5tBC>T}F8??%mU+_rDKpG??{YW-YzW}CH7
z-&AZ;WmNg48u#$7cMYn!X`{&$gXSOh{rsTug65y}=ZY^}eP->~mXC3wW>9@gO1o5l
z&m5Qb2Rdh8H+%35C)-$jZTVFVX3cWaO}YNPem;)M@2n$>Ck)sz%cM#<u1wii3~c#p
z!v#f7rkU#f`c%Ka^m@PG$S>Zwpzw`lCMq6r-q3TGHJNL&g**DKA2qwphQ$1SG_2?A
z4lKBA<%8pwm~?U3qw}_ZJZ!c}D$Cvb^rS~V8*Y-ul77D{s`J*c7CBGuyJoGU%uXGm
zS~uy6o78<)-p5Ls9oF;IGETbIjjf**>!y^-Z?x1To#}cXY}oxI{dsNj5BmF+q_pGs
z@%s6D=h$iclskN(@yaK!JAF;TD{GC7hhBL?%YOT>GhTVwi(ix-_v>2YMz4SO!RB}0
zh{Ki_?LOHF$|jpy{P^^aZBEiS^3wT_T)$welT^m-^Ml`fUK;X_x%sncPEtAah+*T(
zHeco>l~-Q#TIGJr7CA{}ug89z@<G3iPF#5BtSfRxt@z3D${}z4c+LlxPBk`qeLqdm
z_XqiXFPY!#{Tr`nc7DSai%e4a&X_8Ld*#eBNoCUi-i+7#>T&f8`|q@*J6_m-*VSu0
zuG6Dt(ysev{`}?Ug|+56NoTsg-)8-U^*o!`&y7j(3;RAu)bX0MYdN5I@3F=!ng4V#
ztmnrj{P4~RH}~0M+?cL$x%HFO-!Xe#M{)D>`g@X8{Z3hPU-PMxb~uvTj~&rt$sOf$
zf->oU2af!sdf!aC;wJU?mE_%NLiryiR4i~(rQF|A{X2V^ZtwN}a!^01S6u9#H`jjV
zri1s+Nh(jM^m*(1*4N01E2Em{RG+tuDzC8O&MtnufBOp;o1}8YceQSKYw9|aRPOck
z9gn}cv)Ck+H=j6t%)W>2Fm4=L`mF<VpZ&!s!_}eX^|`YSziWp{D&0OurdscJpB#CA
z;VBcE40lR8{puFh-_h~gugyQz=jJorzWBPMSM@sTTa#3V^?bN(ekng@z`lnM{QF`j
zE+j9n-;ZXiA3B7^u-@OBsW`Uhu98|F_1_zbYkar$-1vpgJ2#JBV7xM{-{oX`9KX+l
zB;Eg-dv1Mi$H6O%S9*OfaNEC)IqKdoi;Q2!<?ng;%{#9CLGN#7x?QIKo<vyvKUr?(
z!c(ejHf|g>>8f!}hi)=)qwRU=q?Pw6Dy%oI`yAs&xAU&ns{%<@phKUgr+#_HGLuw>
z^}S-c<}j@1t}-3x^j4#5|GQwlNtL(s{IcdL8&?{yeB+EW$~1jurEz1>>lMG>=cMa%
zbhdrY3hO>yrsI73`G7mite$J)#&qpxap~vT?$uhqb?HV&GA{YWJrDUgJJo#enDb~(
z&6B?|*~Y4;FFHJb&=ivtW}E&xo_HNsJI`5}|3Pkni5u^!a_rq32Y+ocP1`;i*8L>n
zY<~CH13ox<tMSUYqu0LvNvEwQF66ytNoAjXXUuqTpJJ0Vp4GPE1MOefY@%XP?<)k&
z`?4-;mM_10fyp#&`;OQ0xIQ1?HUFS~1AATYZqt*Bov;`*kFeT(a%lHyLq_eDlP-pp
zciiLLQ|jEB-8&VTq_N_V5lec^`_&|s4US*&kK1aNE=VdHf4JhJ$@`QlNIE}j_}E?>
zca|zh8eeL4#(-@{mo7*;<EDG6`}K=YKXBN^bKi5~M#dHNJz%u!H}!+@=Pn*<GR36)
z{@yLmev^BT-|_TxlQd4gtNOT08%{Ps$vRY)Yz2a@A8nV~@h9~=uG!90KY9H-qYS1)
z$v%q1AAPMs2j_pfW}Znm!yn0?BYORNtV!_;`|q5LDLZ_@%||VAV5KYD-7$RXCno7^
zwW;3I*W`Teq{=_9rsP|N2lcym!R5<MT$!}zfP>~MpO=W*4!Uwb9{I}ZgZh7O!lu`J
z+EseA=gr^$;3S=4z0b?5ZZ^NZ=YQSsgXK$2*!22)pH%&B**5;!We5G+M8#C!n-X?=
zufNCh>KFF&j<$6cRryWxw-@(ZI@zSksPc<C?y`@HKDhp(EhcP+mG_}v?^C7WxyzlT
zvGIBN0~*(V%_NmkJx`eG^Kz*kr<rPh%mEDto!)hlNjh8X`_+mC-;6X#WBa!*?RnGZ
z?M%8E);{PpPM`Jb>K*>==O*c#e8-Bhm1Y(?N#lO?di;1px6w|j^g7?E`bGVB(W8og
zO0_O4j<5B#lT_Yx+0q{4M=Wuorq}h6l)n#V%QNYJSC9O|dViQLj@x<&`@W2fm(=%o
z$-L$7^xbp~T7Pe|gPLBn=@%1s1|6T@cF*+d-14?d%H$-CXU@L=)W^ndbJ9h>x~9v2
z(dwNAgRa~wCn&?Z@0qDMN&DPHzDfT)k*WC(tKXo1KQ3<Vg8rQjxj&VA<M!Tb%WXGN
zv)W6hsL8AyCS8={Oqb2y%Yae$8}#>?QRfwQ{CheawP^4D%bcWgpGzJe^?sk99c6lT
z*ze6p&))C(`A$+f{Iz4A+1z=ilT`ZsJIQhLuifu}Gbb&7-`E)TJ{NVobRP%B*Op(^
zVAd=rsVuy@_tIb6e&V<>?0Ktq`^a*yPAE1(`TpiBhL8Mmt8rtg=Z~v%^^sdlP!6hi
z#gN@cZZ|=BcE__9&$?}!kwniH`#inkv74qlVUf6gnY7=7y5;Nke4Jl>((T``eYdvl
zuUp|L|Exp0=OJi+^<J6yd53>3bo?@Ce_q=)>U8(}<JYb?NoCZ3x5ulF((mma1NXhT
z>NY2-^!xX-<@jy(+f|$FxvtkdCn#-kQ!QV+=Ak;rf4JNU%2fN;_Q_Bn)9<U*{=1!a
zZsi}1>X4l}Bx4u&Z<g}+v|-tqdfxNDBg^dnE<Cf%+NW<SHgTuh-|dWheS_ZRuWs<(
z4igr=t{Q&l>6i02o%Y4%pO!l5qSx`1G{307m-4Hpd|oVRcBcNWp5K1`&P%HP=CEH*
zIKA=;CtLYWtL=5C)?DKxm7|+?oLseDkrNeDT|Y@_AN23XGd{n+8<_deH*Y;*>nJBE
zi}H?tui5i`P12a^@5nOUKC0hEq&x1=BZiGD+kBambe?s~e$DqddA^f0^1N2kI4ys~
z^*yW1b;2h5XjHt<IriFtf177gWmx$|_1*$o+_3Vt#Toltqq?()k8_ln)#3WP8ccZQ
z<yB5nS=jQWV+NMq=y;`VJyLEv9<)5#_otxm3(rnFto*$r9p#^O=zGW5xmWjH@7Q?K
zhc7LD|Cd!x(s`hB_I0xd&v25;xIG{0Hy?fn6nC;;r;Giz_P_7bX6sF=4Epy$!)ll6
z-%ZVQ`?&sn`*h_M_Ihvr(Tb6~-?_rXm0^!JTjS;awD6>7E*<JbO|SaN-_5ICSZ`eS
zIVN2?B<=4M*z!&FJ-pgJoc9^qbZpW8%`V3eTx68~bnv>5y7}FBMYHo8wpe7M=9UJx
z*X;C4>6}c{wojz1Jk#GN#jSt0^%>XtR_``$l-qulNtOMZ{XW9{?)jbb>g67p+aR}e
zpZxauxn+L)H?K-wwY(a+rF!S*)%?w_pIbV&bem3f9?Z{sxo^v5wM)lUAI(b@Pz6*0
z-wF(<l>bnNbEX?*W_9pAaoRu?Pz6+hm@05n`FF?MGN@vXG9Eg_beWX+w4VF!y!Fm~
z;|%<%<pZ-$U;lx#<drY#?2$XwC{>4SSO+EkKvw}&0aZX1Pz6*0RX`O`1$L=G&TZe;
zG9N84LD};9Yj3?_(?qAFahGChp(>D21;%u)*6g1Ll+AHtyMKPtzuffioUoWsxioE6
zKow90Vy8ftK6iC3x3+xFlJ*BLe0k(b15X_CK)Z*IU-E}dIqdk!`EB0$YMJrM2mk)g
zsuPY~Wc+gLn+q;z)nvZ$%ICk@^7l2PSDUbS`>=Ygf6Q5F{PLCiPuQ=nv%=UodG`2+
zRyJK=GR4O#ulo95AI>vjv;Se|{Jf)Zn2DMl%B=5K<L{G<8~6Ks@137_8*Oa#x~{_Z
z8{GPi<rjak#H5R19S>2*>EHCzQO$0znv*K`y1eTd)gE2w1m)&8Hr-Rb)2k-aJnZI8
zbzVGro=G~7t5vtl{$ocvC5^E=;cG&H6iC;3+P2>;?^yWA;j1d=1ZAK>C$Ldo^E+F8
zyJ3P87858zpp?75W*%+RXsL-i0}W0asRF8iDxeBProcrVYJ6S3+j0|EesS#^H4Zyu
zh6#&*yZ4{3+&pB8nR3llgTG(6-3g0pd+k{JW2-`wDl7Nz_u}k#3r$$eSyN}PK`#}U
zbTP7ps)1DjRX`O`1yli5Kow90Q~^~$6;K6Yr9jx<x662K>P>p*<ifQkDqcPGmzqys
z{IgMJw+^v7XfzpBKow90Q~^~$6;K7TSb=mskCmx@c+ZSqX7QBk&{P3cAWIcEugO2E
zJ+{|k2Nw2R)b^V%hZr}uJ@A&DD=Pi$xUuQRmm3Va<9m}-zPNEl)6XW4aU>hJ8Bp`G
zQJ)y)u64-L!=+TS<t=-iJ7T$krOv#d&zQwUrld0J-^Z+Yc<0z_CVb=A$nPb+vY=0;
z8fSd7%6R3@IcK)peBLZ0*=6*tf4`wj#T?HZc-qLd^Ts+|1f%}FUaIj=C~SDwB@ce?
zghkst#-;p{uS#uhw6s!=WL)<7y&h_@?7MbFPEsk?zuN!V7qazyspC=wQ~^~$6;K6K
zfs(7hbN_g&VuK@hI=3w8GxoII=URD4rNK*ndTN~`>Gr!~x9iBk&(&=BUCRl^jq-UV
zxA*MKsCL|aFJ3U)xbe|D_g=EH!(1a-a#w_oToq6SQ~^~$6;K6K0aZX1@UB4X`rFGK
zom*s-@zlZlOsc0UpbDr0@ls&x!?(@;wL{gMsL69td%SSK*?TS?>_{&E{)EDQugx=c
zH_ZH{{8zJ5=E&Z4YgcG+rBSL5aaad#M}5%0cayE{-1O3VRUjJ`IIYncEi1O3Z@jWW
z)tq0Bx_qti$|(mNGy3#pD~(s)JL=7$2lbifNKVapwtS64*E*^4E$5I^D(#r+z_6Y_
zllewv-`0NHvYDGp<%Gq0Z(n>!+keb7>Ebuf{V?X3!7EI*Fz3{UGv~hbvk8ko4sUbB
z-q$ZN%D>j(qbFZ`dV9fK6BV;@ug~T&J^lH)Cw9tTYLu$OZ|fkN*Sd~P6;K6K0aZX1
z2&%yNgHO2Z_yxP=_@(^4uA4oeQvO37&Y5m(O!aq;^7n&ocBbBc5mvu$&*NrWecawR
z==bk^+Wa@nd!={f&lVYFv^oUcAGMt-pbDr0s(>n>3aA3AfGVI0r~;~hDxeCe0;+&2
zP!beabID7si*}puz&C$twy5&g<#KEsy71!W2VPfTg3|3h5+$X+B&@9@jPW0<tKWHZ
z%T=bNFbQ+=e{<}1aKD~4KHD!RC?Bi5>g#`fIM0N|Mzu?Sa6qfo2B!M`rmg+JoyX1D
z?a3uhQu)fjd9AiS`lXXJ?*HD$|G4^-g(m5Izv|b+pK7+yB$bs9X)?6ftT0LC?ZfJ|
z{xN5zi3|Vlo|;YnU)8k2{CE0(;iSq_%lA5{+h?V6Aipn#^6#3sJ%^(8yMF(7w(62i
z$5b|t#Y`Q1!AECLGfLG#*Ma>w$mTsj$L6lU(C^pIdAn1Q<HmY79x!H7*|IsxUF%Rd
zeUFoFe|v+Y+_etwD^WdE0ad`S0<~(LeD7TktZ~BTJqH};Y;e|@Y^7hF)n64*1%fLu
zdE>30nqD6}+0Nh!Y0GR@pkKF-PkjD`)lRnY>XtLs)*L$D*l7E`Xwu5P{yj|f`_JDG
z&E{iS$EFIX0;+&2pbDr0s(>n>3b-k7?A*nZcb1yr1f^Si>a+E|-cgM!toQk7obeB>
z{D17-cf3~Bu{UrIs30muLnOf-#D3A(V%<-S9l@5w*lYBN;<XTatf)a_qKGvGOT^wH
zDwbp)m1t10U_(@-2?PP@y}Z{?_&)dE$-D2vfdd@K`p2JZ?OC&CX3fl+wf24v+YQf?
z9qQH&;yPinK(auxK(auxK(fHUj|IB!(6;##Jw|7Xigoq-H+<s1$G)<(>x&2dV$^j5
zezf|KY(>ZVV$UQ;vOuyx-CN+ewf7&o-tVSowZ)-3^c`{GdV}(W%f}9N?+J-xeYU`B
zoemoH%RT;)*B0xuPAh|KU4|dD(VeGG%8PKFFVDHE{jQ_)Wrd6Sd-b~dJ3TvXylLm}
z40t)KD?H(n+cy}t)0cU|WQY3X2P;FaN;)J9Bnu=9Bnu=9Bnu=9Bnu=9ENu&1)pd=t
zpPKhkR)oGQYT2^Q)lY2EvXn5{A^AaygOmsUnFp43ey%J2bJlqH{VwZI$W}D0EBzCn
zWPxOXWPxOXC9^=K_hc)}-=E!}Y41^kdQ8d^COiCJez0WGRQu6I|DHkYZ}-sA8ytN2
z;5=cn!_x4B+IwyqQJ*YO^gYaq(qEr!nPgf17C5WR@I5YU{nfwJYo~nD{i(xSP011_
zJ0w5&*Em@Iky|?nuDHCqO`{h_=gS*wCv~Du7FfC#xOMx5{Cjroc~W(|4qJBEszcRv
zJrC>Iq2V|Gt!`euc6F-`4X)`~z3w;d%{x?esOorZ+iQDP_v*gam`$pdu5^ievOuyx
zvOuyxvOuyxvOuyxvOuyxvOu!HN^60xvtI2ryTznDVY0)@><24t04K$g1(F4l1(F4l
z1(F4l1y)cCG~RaL8E0SlWxl*|1vOt1J6RxEAXy+;AXy+;U@2H&!Na{9u0CZ<zOvz>
z)>};a>z#Y{+xqk|SzY5ZEvNkU^>wFYOB*L$@brVnJUJn+E1WfVNXveEew8O&K6a?9
zbzEM2Cft4Sw!41xb)Inf+F@C(zdTwAmvo=0x~l0VTMZbUudFzI(@XBW=ag}IUE}VL
zjC*PQT_$FgaJz$#{q&Z1r)KpPJMA`YV$;t*$!d!`&ujbEpl*}1Ma4>foSr8O{7(zC
z?|Rs!m+k&}Uf1|P=ScsbERZaaERZaaEU>&SFuw1>`*pkdzAWLA*<pFFW12r%U};$3
zud@$t(z(&dd{Ob*p9~pv$QE<6CB?pDe%fU0q)8cE8oDK($pXm&$pXv60?X<?H_cPG
z@6(b!lRfM5{p^bBPs<}`ra6-Zk_D0lk_D0lk_D0lmWBm({PQzst-Z$TrP|`(J3id%
zySuGcsx8hxvfo1!<~A%<!Y$s|?%b{Snwc#swz&02e|&r5ye#4WZHJ}dLy2dyK(aux
zK(auxK(auxKz+8r!9PFbo%0ur%2!nU{?k3izP-&q@}-USc@2|n$pXm&$pXm&$pXm&
z$pXm&$pXm&$pZD!0{8ap^5d45eVr#vcBo4~sE@wBBFeP?(6*;GKWA*7Fxg>g_`!-8
z6N@7At!;n*y;qtI%oZ0HWsXFXERZaaERZaaERZaaERZaaERZaaEU+S2psxN7-pclO
zOjopj*Q~aGuW?!XW^LtLQu<}BUz%^lu)vD?yWS*s`hENI_<j3|5!Fe`WPxOXWPugW
z0$sLyYS_Uazmq3ix^~!XbjJy&T%YF&lO2|&AFOy8DM_3xkSvfakSvfakSvfakSvfa
zkSvfakSwq~EU={io?jYYH~;?b@>u)2oU^9w6Qe%d^yR#^c-boVbiSh9gltjq+U{??
zKf2G@Y-wY!)(1D=;Dn*szbn?IZzNvH0?XS1E86!~N#0j`&2BL%Pj;x^c8GDkxZis&
zZXG%I{k^Uq)p~4RTWs2Rr;$C59+}q`v&$ap)bYrV@`%w6shup4ERZaaERZa)JS}j`
z1-Eox<A|YoU18^{&I2A<b#R_A*`Yr9!6j?F-0R{^-p>;*TRSYz+?M7}7DyII7Fd}r
zaBI~sZ_3wuKL2;c0|)<L+mBy<FHe~4@Qoj=%(4EBZcF+*_r<+`t*dcM`a9BH-W_&q
zlluqd^$k}!@~kNx9~hO_ckHy=w24hW|0G*nT-oC*>6a{!ERZaaERZaaERZaaERZaa
zERZaaEU;`Wu%v%?ZBfThA35#QH7=W&|GVI_S-mt%vOwKfpx2Qfy!G_HbF;d}+W!8e
zw(k}GUEjl0(z|ZFB5_F;NES#INES#INES#INES#INES#INES#INES#INES#INES#I
zNES#INES#INETSpEpXPl&H8Nc-pG7e;i)}_ANbr}WAc?9>)J2&>Dh6|nzqC8#l`ym
zyNOEj)wMl~_`QA62DkjY$LKua|F%Psy;6U&Kz*>lkO}*pKY#tnS;D1jho^V_?Yj5x
zF*SqrVeOL)$pXm&$pXm&$pXm&{~i|T)uL*zAB=lDuPq*N-p&v7{`85w2rGS$Twkoa
z|IoIlHa}-<Uf*%8!`Esx{DJZLii%76_cY@8`r<!Hj%0z_Sm2(+$J}<w?NjoE|F<1#
zW79;NERZaaERZaaERZaaEKr{<(DiR0{^{gvhUZHQr#;ztSl7wJ@+HOkY{Mj5vOuyx
zvOuyxvOuyxvOuyxvOuyxvOuyxvOuyxvOuyxvOuyxvcUh(0`WVNW!3)w$&&i&p9Q+k
zTDt#EeSP^mzPE0_(DpsM_B^S&U5BkYY~7*ix}Jyi?9lL=|5i7zUc0(ghX&X5tX}t<
z_U0X`I#hK$w(Ye&t9y0dYs@B94UV|{{;Te4_CenHi@qbyJ$YF6-rjRQe&vnNvoqK0
zdUUs8BeJt^8++gbpO49Ue{|9oFF)3>^zV+3-q`b|X`3}I{prpJo_eBvgOc{v`!sp)
zwdy|kAHRLheh+QkFJFDX+u!Wh?6`8<^B3DS`(*r96SJ$H-)o1n`n;N<n|l2t4<B;j
z)7hx|54~=mj*Uv2{@|J}eb1^Yo%{a%uk=0otL$$_JllSsd*^1ex)0v+<BeLD`XAD4
z)};gA&y2%Q|A05Tjp_CFpzQj`+ON~+(E(*7U!=F*<lzU}&^US+$4`u_e<@+bkjVN$
zcDHVOHhf^sCjXd{vA;HY{;>PszUn!>^Q3b8@&oJTc_}-o+jmZzUiL@phmZLqdM<yE
zuXoeq+I`*j?OgxSv)Y{dm!_ZO$Dh~Z#G$JVDfeqXb$+)8ue@?t`T5gNXB*wJU_}1h
zHqSPHcFvew8^73Z{EH84y<pIc@#S$RpE-8R);mwj&2Kz3Kg1V5bjC+HjDwy>#8*GQ
z?8$EIiJ#~C#k1%5p%a(TIz#^9`KbM_KJNQdXXf$}L<fy0r+q>{?AqeQUEla*d<OXy
zKjWwTR$e1N{^X<ISx2^i|JLJ2W%Nc*Uj6J7`{Vkd$G-^Ck)M3l5At{GtZh8Gj3<xr
z^wl<wA3optoa<h{=&Q0Hn+Gndti~V6Wjz-c`SmjsXW#y(Uyk~hyuc328kdzl(>yDt
z1uj`*!N%P$-m)am*;ln+{yZyf(tAcGZ^)10NB-jf#>;b_8>ip+(T~o#aP{e%mWK6t
z?6#AetXFz^RD*k7-+1HF4?8}4+C6>Sls?~e=Lbq_u3N%?-|4q})p6LoY-Ph}Z~F>V
zkH-B<zwt4Dc`k2zj<5BRx7AD55wf@O_^#~89*~{bMSt{b{YC4=j^ctH)e-n5{AazC
z?67q0pl(s`sqf@nbz3}_s5g(FeBN%;cAc0Hz3I>y^Ddf{tE<#?Wx9puCT;bjcu^mT
zFLk6kSiaIvKKTubQ~GG*Pp*h@-u&)ve{Zq*w9I_QlgB=WkNMECqc%I!U)y;0h3q+e
zz?%QreCCwgIgq_09lq$v4adGbtK&~5&dWWAW4i76(T~oZQ$BB!n?B^TUXVN^oBn;&
zM(0n@$>lluPi)`*>D8W}T=oz4AQyYemw!3A{~-$=nw-@Ynvefzvlsu4enLL$>^b{*
zj$d3)b~8_G>yL<E<gaahcGIRGImvB*agId?@x>3~<2icqy7%*g&fERuIhlE4JVttL
z$iK)RksQ!IAWre|-2D2{kq?p|vXAGH4}R<c(M6;OyFvX>oJ2cBdi;%p#-ZmoMHv0e
zIA|O?{tES@(+}k%C?1SM7u)DPhx(y>C7)Oq>n8v62Yloqe9SNI$*C=mna_CRAp2V%
zenB3{4?Sl;h#!9-H@V_|N)L#RoJB}K_A#%1<Fv`4pS<SPj}AX=XuZhE-}tdQSlc}G
zM$gXdq@O;<YqPs~A^og#)W^Ep*C0KNM`yhCmRIP(F6w@D{^?haec||~jZ6G^i$Bjf
z=F`rrl_DC)PUa<#d8`M$%<DP(kk@nev|bP&`qGQM_^~(t_T0SqqqlBQzj4~ox@(I&
z^Rm17$!Y!Lb8YkCZ+vX)hvX<~>xb@7oYS4ZU;XZ!=eBP<JF~ufUAq3<Ex(-kFZUa-
z_dMr;pKi1u^PbLm+<P_qu=knh@b|vWIP<BmA|G;?cVzW#&3AaGK?zz{c3>}Ueie~l
zS-<!^`U$@1<pc8@Z@rC&*4a4zVBPW2Ccl1sw2Lr~CztgnKY5_%_AT|V=k!Mxk>198
zE-vaC$;FT9Iq!piTyyrsCN=QcPW@}1`=~(;KU24P4&^WWyvNZtAEaNjH~U&2a>@7n
zjvt`Y@Afv%BTuUjBN|r;<2vy}a(f=<r>}l=5v?<SVIT7tkDuod9lq%G^Go{=e?w;+
zjC976&vW+Fj^p)1cDB9|tt)%P_|@hI>=BWi_NPiHZ{SNV@<4RZ`WvTBPHpq!3)y3N
zqd1Cn1-sH8ic4+!#kO_ir%=E7U=bhoWmn_Tvvbk9us1ZHI?T8g9q9w<X?<4|<X_~C
zZT-s%qaC%KD~j;v58Z$77SGHm-<#Pl-Mgvx)#uPYT3f^~jx%5M3-scGo$U|uwz#FY
zaq=vb|K$zi^y4dTq4*W&@<>Gd#J@H^+E70<j=cEOUmHEtk1sx+M?@!|Xv5fVf3+X8
zlYaG-^|0?kbq+fFzI8AkInCp_`SjC|ozQu%pIstf_AEkv6yq0t<cojVKNh~P;Lo1d
zS7iT6NPqjBxQOG>#s1jFC-$fIKfjBuWQR(1neqHC`fY5Zvwoh(Hah$v`f?mBd>?0>
zJvSa-am$~ee1u-yYkMA1K8PsqppVbxjo4q;ZS%7mz2r~xv5Wo5b9vQs{m#LWUfy+I
z<9-M|eh@vn*dN>USl;OV>ES~@X+83x#wGVw?wQ;}dGG2T3jJ1JKl{PBF-=O*e)Q!(
z>~0+w7mMPzl7Gi><`*}{i$nXF_|vZq)py1lhrSXT2gMWl%p>mAbNt495z!eR?QcH&
zH9GQ|FShmLPfjQf@z)>Gb10r8U26aHcOEO69mIb`^{;#p^Qkt!6A%0<+5sK^^&CBa
zH7|bRfgBLMxZvm7(0q|TKGznnONud%rFQ+a0KecT@?2bh@AJ^B`|K;u<@^Ypaq>R?
zP(J1F&V#3oI^ny!Z9lvGyBFUX;S0&FEzV%k^P+gc$MYzk=k%tpcvi2m4|*8sjjL;v
zSIASuADw(=-N~yil6N4!F>l8E0CenV-D5pN4$q-}<F)aH@+A9OKlfhzk6sZykL#vQ
z9`;#O<j?GH9DipA&tt!K<b&RLc3~&;u#5HeK39FLUp?)8v*$5>*cZA75ueaKpnFJd
z&m*4xz4dw=^wiuu_WSP8^Vmio`D;URxQ}#is!fkdD9;tqu{(Q0{v$70H)x&3n{%4;
z3_n*NSSR1*InRkd<2<*{-m5;i&Ib1k+-lVly3Kx3`m+DHM&;i%J5NIM<AbmG$H#XB
z-uG+!Juf=^jQ2f(-;MA;^P#67efd2<cMc>U`{RqBb&v=6IkaBd(0F#lUtF1Y+%Z>A
zxajq!CHj$<eC%_^U+@3^Ilo%H<a-|HS^CpY-i7+<txeA8Px1#kZT_Y|B0jO7pX=wB
z{8`>44|@4Y{?hiGe)ws}dB{b6{E9HnLoWNc{4XDfbAABrC**<T*Ctn#(|SR9Nt^sp
zZ~gKP`Q>4D$H)3Y^3g+GW}Vo_c>Kj3dGO=co*PGh^xEt}5Av`(dVC@N;$B<+Lx->V
z@Mj<O0DZ-==W)E}?7+U{B)9d2<YOm(%s!suLtg#zzI;Ir{p_gUIQfWO@Q-@YgM8#d
zj~^r-KIE_-><^<{_|wPuC@;Aoe&&JXVh8k|+eeJ|TwX4cN4`a8oPC45^1b`Wi0I{c
zh(DwUdVJXl(uZC3(=Ym0TrcsiAM$I@#lPp`2kPes+WbZvKW+1fGbnzF5FLziSV!}F
zPCojW-*fVz(=RUM2mRU*zlg@GE1>xzqK8qQNN+xTW8H$Dy*!8d<w5+6Lx(?oAv*F#
zdi;$?ACX@`c}P5Hd(JNUt+#%8mp?%BM)~7B@(469#7{mjK0ZfBPVq`!^xEj<EpbdQ
z@r@6@`0!8lfVxEfGG4w>cgr{Q(J#M2&mlVF&7-aUMQ>jDN<TjIu^-5D5sicKxq8|8
zLR(%Y2RYT<@+~^+VSTNWc|B)W{q&=U^+b;@V*R`gOj5CcxU+8dDS1x(L2)bJpp%FA
zJO3#{@eS=`@;1IuzkDtq%G3DB*Z9dd#>*etp5sRk`r%7|bk0rAQS63}-sBKx_=+2H
z7>`anqVo(izw=$ZCo+H3i{12V+t<uXPj*MI?OZ26TQB;t17sh50{Njj67ti8D~AR6
zKmXzX{7(GBqJI7f`4#^7@<;RG%U{J2J`u&0arkMI4}Z_i?>RpB8y~TV9zXGipMHGw
zlOxKdAH6oYBH|ZO9T4*izUH+q`so+Zy2-ETBi1$w<Q05lzKL~^{0rqV`v5-nL3H@y
zr>)ME|I|(Fr*2bUsn3mzsQxq#;s@hA+V(s9J^t(j)xGq?SHJQ4)t6BH>UVbPQu7&y
zKO`SJLiR_;{?-Q?&#(QC3f9(t5`F!!fP5m~Sbu&2<r~k@>sRN<HxPgMPu-)<pU}&P
z@^PF`Ka|JtvG1s#$mMyF+>sx>%<s8<!E<$@Hhq16Ku>Z)dXh^&xz*F&KeC%T6kl}S
zN1}uFV|>t?&;00&gZ0Bpm!1^X$@*!tul0w~fB40sBL9Hw&#$y2dQL9Lk6;|9Z9U<P
z_4{tH`}tKh_(T*(=;L$tiguxwxYI@_pRgA?@vlu^^bntj)<wT|B{|WnD~uzbHjMnp
zrJugmLtIrt>ttM!-D7`IyJ$Y~@40oA4=SO0fPK_o>ML~&I(3G+RNbOpQCFyAwACZV
z(K8}F*j<|))p73kd{6N6#$Dgsy;;MO?-SIC){{S|#~?Xj%uM89N9)SY5j}^-L;B-K
zK7Jeh4jnt0Pro>`F3@^LIoLs6X<dCsbN=OnPuhOx2Ib%1&$<0SnjL@cYBf8L?|o@a
ztA;iB+b6B7en=m36d`@(iAr?QUXdQZ*sm?_$yY>Q)Q^w#gs~rg7@yPEIzsCO<xOpf
z-uy*~9vY{;tTE0X+v0=XQ7`nc(sOo);$?C5Hed8Z@|Y*0{3U+XZ{kRN<D+h|ud4@*
zQ#URy@^ALmj`2a?s1N!gjDBxFiT48dMn170Kk`BAiC$c3M>LOq`a$-<S3fKor$5eb
zJt6+kIP+<vkBClu6!|ee=7IExh`y-bbM~_y+Qy+Hr|0_7MY)UgLC+4__(i?Mk>_##
z*tY-iPi^rH=^LM`Z`=pj|LlwS#<qTZ%|kwV>W_5#A$=h~g!rO2zj)NIKJdH<#h-D}
zF6@9GxgfsA#Ws6+PJeokTYnKozL74<XFl_r$Gr5$M?ZRc^NVO7{Pde2;s@;;;zb<9
zbro;yD-P^8<|B7(n-8MXHlO+RLw2E`@$6ZI^k%mrI_pSo^JtTg9Om~N>W}l_Pae<X
zy6Q)duYP*qqwP7ljVBL2#(8d>er@9*InkL1z30$;+T`T7#-Y=X9-{LcJ$@16xT5j;
z@zaL*6d^lAJEA97<fqM#@QL)!iTvL@zBh4BbuM<UHa~me>%D_{(bM00ia&8}|1-{c
znOykeCqJkop?Y(dEl-;I-Yu(??~$C7$!i|{@tlo6L?89gMi)^$)5CN9+VUQ{h~mS#
zYMWm_`&b|QxqfzJ2je2We&Zo~X&VRin@1b!=O^UH*S!3MKN=tBG0yi_?gR9DFU5b+
zxhL^{O`DzE7eV~+aX%wY+#f;XAUVhfeXnJHZT!urpS+&Oap=iwe*NTSXMT^KT<j$N
z<qv$&X_JRvSug9TZT;98AATSn__sI^f8<85Z65vn#&h%XKW)$HXI!L1kH5BY+7MrS
zw2dzs=Q%#w?1vA&=%9Y<z`o)Q9lp`7_@iSF^wuTzqqknr{*o|RAXy+;AX#AXEFk~N
z>-IbOUS77Z$j_eJZ|(Q!(Wy_=8R+pdPQ9T%Fi!nopHUY;`%sj}JoXRt_!@^UqUY$z
z0nO(*dir>-&N097#>IZmE1`KH`Hj<0F8jQGZO@I5ZT<FddKs_ne$u%sp2OVJc+c)$
z-@U!}wCLQ==r_(ik9nNu^t%@{zkYPkb9~K*eo^`0XS{aQi@f-oAL^&S{y2`jkq&)a
z2jlgdhrC6IzvtS<;|tl#c~c&Ub)Yu-i0X889zN*k3H3YY$MN`}gZdX0`G0*!`J3Gq
z7wrp6I$nOL#K(S7iB8@rq9eCFhTgh)9<hiYdy+5m!xx=>ZBfy<IKRBRq9OY%s@(SV
zBKjhJ*0*R}QNR5$KDQ4?dCUXVG3fELpSu@vU*q?45P$q&9A_T)ALc=?jm~_Ke9-sQ
zp5tpgI&z@*Tw6T@<8!~a@!gR&I&zCA`B&Z+moYBTspt3wde61hvrry_u^(S?Ykp|o
z@Em<5<Y(4R{Y4JX@zD<#uc}X@R&)!fZ`51rF!iu{OuxDls>`(Tf!-6C2Z}@E)m2cO
zp)*c=6rq1-%Xf*|?ww;kkuN-#M<Si`Vp0B*@7zzzKlng-Puu)Z{=<(vMJV4Jhffjx
zvg)U&d_*6}-u5%{NAx=w=sPU+zy1qfx^?j0&or!YzGvU1h4kYO_8I%G=lljn`NRi5
zKjVatS3j=(f}CEkuJy1kb<Gz){OvoC9qc>yBlPw?^z4b=erg=_yb_|vKho<bPxMdz
z3q9ut##w)Dd?T7)ouLhlS4W8tc82l|{}NyN$w_V)`QpE#?`ui^|7`*1NB%F)w4pi_
zz4Idf;lKRTxQNC_G_QVjw0X7Tb8@)XkuT*>^@8!_F)m`{Z=C$%Ie9&&cM;kL%&%>{
zer;&JC}*X%aS{Lb>ZjI<VgddV*OTA#cXa$-JtQvF$NJF~q4<mbW?uc~MbCddC%1V#
zhx*AyPIW8(^rR2eZ$5mr@h|Exk~`9)GcUwHB0E9!kbLxYo}mwZ=%H~DqrBSk3zWa;
zgATH{`N#puM=$bw9_7*x@m*2kFUeUyEMR}-KT8YUQ!HzAu3Ogg%Ddjvxo?5a3-Y*i
zKre4;lizq~o|qrx4}740c}D*5+<g4NxyXFzJU2hS@|frJgX%u|yT@_=<(|qskp7E{
z_>fcHX1Ay}z4RL&+w>$C{zYx$J&#!QoIK=({Kh%hIn}w-Ia9wjyT^0C^~RT-tebO*
zdvEmC6B=KH&N=9v>)l)6uU^sb+^y|7q<7TU{hfKB`RPR;{Mak<V{d5Rw=R(1!$_xY
zVCP8Z?`hx%(LwR7&Y+Ly=CL09);RYf`tcFZ>_T7m<3H9D9X-vf4$>wcdTsJSe#buM
z!;if3ll+06{N%wG;&1+l>|j3mQX60W&^^9?x4}JsjBoP#9lHB{{l>`?_<9aKhknQI
z-;=iwYVUU1`e%-4*|^5<;63MWeQ)UYgZ@8mRQ_GO-${5bU(Y(~yxTfXZB(Nl;)f5p
z{eD+lUZbbqW%%C4JuN%Bmo^W3MmuPm$9Fd32Y>!VA9Th=<ZsYC{D5EZ1N!lM<09HO
z^z&P7a+uG4gntq82Ykq-A3x*S0X=((FZMP*BDo;FAinYrdqMdM>KD)Qj5w9Q$%%hN
z`mrl|>!wZ4i1^{-xi)>#u_OEHM;DQOA-kF<(#L*sn8)+j#z()l`K)j3_gp_Yp?-W}
zlmmbL{1JUb&(TBk#<qSK$Ky|qXgB@_`6)lqj(&u%ap=u&oagA-l|1}~{PK}DJL4Pq
zm<JvG;&^t8r%-Z>hbS-p;<E^i!%sV+@v%S7kIsIqO&*Akc_Sa~i0sHNko`P2j{Tv2
zNDgwL)5Z@Pht7II@>o~xh~zbn9q=`edE<C={M&Q<_^E#UDj_-8#XOb9ReH`I>ZT$z
zpZG$ju28S2i_}k^vzPJO_-cC&<GxKF_7wm0K(F6A*iY0+==hU(fb@gLnV+AK6MyTA
z-gqb<(VKk6X`_ey#W?eMZa%1gagknr@5|ogWN-Y*#h%7{4lDJOpMFLB$W_Eo9qhUF
zvcCA6kH1(Sdh(Nq#^J*r#(Pe0{n{}0%g>%iqz@z)I{fMjvI~2X6RJz)HF-|I^@7fu
zo*So*bq<g9ZEQO?qt|w>73a<!l@K4}%?F*6%tKzNE_W{w?@!#z6nz)v-`_iJ@Ag0H
z^y^hhz9;kVmqY#N-Ou<w)pr)Y%Rm?9(T@7ilitvM6MFX2PjB|nM$fM9Puy?ej~<<I
z=F!i8V!!A3L43@oO-|$J?YZ&jVn2EHNA%pi8GFM>uMUm-aJ--J{L-D)eB-pI8kg@U
z^o!RbL}$Ev0QVAp$7H;H%l>Ykfcl}h)3)!pufWfEXy1(OB6@W87pPxek&m?_#^?4K
z@|$1&(+}Cn{)3Jk<URD}lmE2oFMra*IC6WAue`=zq4D(LhsKdx+c<RWDX*c2?4r$|
z$;1B_6_4Cx#GG%v*P?WIr<w0`KWF2Tzwar(Y0Hn=@`*g;UVqH+H5dH!%~eb0mlxfi
zi$nL~@~Zr7{F9B>xT1C2<|X;zgm1q)_^AHhE%kb^(aQtIwl0mBRNd?s*KSan)qU`m
zA8*vM^yrN}Z<@AQ(-OI@8#&2iU9A&;;Sa_^&-saU#fM*L<E!mCKE}(Z{4vt;4}L)}
zZOE?dPhaaoZ+7QzF^}^jdDl36j5jZg{p2ABq(3{GpPlUo?4pehU;DXvA^T{Pj~?{W
zPcCiz>1$lH3qIsyPi^vwNBYrUd~18o5BQIM_7Df=f%@&U<YNaI?GyE9hu9z6_*e%h
ze$ksZqIvWi5Al&VqW!hmIr7schqm>N{p5&xkprspBC7KuqNA7JE9kE)B%gVr9`vUt
zL<fzJ$X>C(Qaj2eKSA@Vm&pam4ap%cwV{4>Rz&$&+=(~wBcAljzfgZf&*cH}rybFA
z^73csIV6|o5ygS}M_r))vu{Cmfa(eMazCIxb>4Jdwoe&{k9pN&#(TeruX<P8eFJ*)
zMLy2$=Fuh>dj0MZJdfvk^!V|MsE_@c-st3M<Js5#<~e$PV?DL`k9==_<LnpY#aBQ7
zvfto~9<mG6AJ-B8C`at~96Cp>Y-qi#546snvjb#TcH<}330i0SJmeShguDoiv;I&%
zkUz-FFVOMl_*{N5p8WKJ@{6`OAs4idJI~11tq-62hnu!*Q9dV-i(SY|j*B<%JL-$$
z*Dm2>|Fn<L+qyt<&?BPf?0^sZn~!|vV<&QYZoGbo9zT9#oHo>NKJ|tDlU?nP{K9kl
z>giXGec||~jZ5}p_BT!&vbTMa9OR}4xy-A+GQas^J*{pq4xh0b{{8&w1DcfFw?#hq
z8BebV*V*8nfm^LwQeUwXI^*zlPPE?@q4nn%{8B$Vh);2g&N%)B^~30o?2I3O;77*c
z$IsalJv#iwsW#M)k2Z8KW#5q3<TLHKpXi5iKKVnuuq(aUft{m&@+*2e|InWwm!Chd
z^@2e&#^>~?D`dCGFY3pi=@Io3KM}=;c!%s0*PUO8ThH~Y8=!R&C-}RcktfIxt&`{S
z4L`+)edw#-IP!`AZMHu5p!uh+QNqV~@oXP(PbB`Vzjc?t>`U&i$Zvd<qy6dI^lSd*
zTIKfu?z^l5e|B$Xz3fNi|J{Gw_};@4)+xEi(&o1@e(;6l7KhM!MY+ib)uYxuj*s)0
z52A<Wh2l!vIzj#s>Bxm|MDiLJ>lFR`1AU~^ACbJ#{`w=mb+xZ(ix+-rf04)7L7Tkn
z0_kIZ$S&HDp7<KC9qk+SA(wu9q5KKu8S!L)BOgBIhy2UFARjuX#`nM00~(K?e)^FY
zqSJPcG_SVvAaqW2{)}k<({G%%b1eS&*$>1MJD`K;oKLmQ1JT1M7kP@tdu}}Yv$OSP
zKhLc*JJQ2A&tY5#ZO`#9YV+^fqWI#E_P02X=j2j<kc*ziSwHKlE$+~>hxKAV@868)
z$Hwtz^N|mVSM%v72RrGvj`n|cpeH&=Zu1yVKJiLU^byHp9R2m9m-nn!C0tSO6VxI6
z%z9XVej*;BIEC^HG)^4i2k}8?{j@!oC-8;lMQ0rUg`Uep)*C%O<cw&3dCxeAzczmQ
zVNt(vklg6?lY{;Et9-!k*^wUT_2Wk_ZD<~C^2C16^+WS|ZhlA(>wsQAx`_0P{o43I
z@`!W%^+WN_p7=#%uf;|8$xyzH^JrUtzYlrg@7a@s2d-NFo&z29`xEyBzRUFQH0p=e
z$2#Ee?-$V%T}1M`HzzN<S_i1#eAdN&VjbvX9C~fz&{?<WAL7FPVPCS(cpecy`<4C1
zc>MI+7xdFd8=9B=^o8`(#)qEJ^C-7A{`?)ix>&xqf6{||<|U7LBU%^vIIb^$(a-O_
zN8vyEqyO+f^yVWMzS__@&*g3H$cJ9+g^zw{9DejRU(~}mcEFb$+W12K?17Kx<b;u*
zHo1(ig!tPh*&RPfAN2ToUQ`G1L-gAC@LT#r^y1Gv;?y|t@40d6N%qnY&97~|`SFzx
ztUo%>W50dVI>6{R>}fot7rD)+9py70J<JpRl3!V8^P?A+_{$6IO%C~nJmke6(mT#W
zPV$)_S{LJrkeuW~r;QIfxTyW)jL3i4#dCIHH~t~7M#RT>82j0my*-C<oPPRh<D+dH
zL?7wUdyYOL`Ov{gSKBr|&^!^@+d8l}`K_0J=sDES4v^gZM?ZS!DEZ3%s^9$P!H>VF
z7oh$)F3yLae)H>R5BBlgJlYY-tzR3`58~(k%{cwq<e{H7#2-EW`s4XL_N#ByXAqrs
z<m)*!Kh$rWeyBg{>$&x?pP<u5hmZLnxg#2f4%+|F>&Ks+jAwssbg>^F<MDTY0Id@`
zb*Sh1q4x#i0$=^!-+1ri{i^pS(7iW0?_K2!dBnWp*}j3!dl>I4$f0c>^SWR6-UXV+
zdzFaZJE_m<&CZ^CFBjz?5A^=Z`=)5`qIIFSbuurxJl9VTa(T~Xy!d8s{=^>SAfLQJ
ze|gWm_;?=qXhZzPL9~Z{^hIs@L^^ixoE{LpHhIJsqz^j%)?b_b(32NhC-RsNJ$az{
zv|&*{eaQulgHcZNTTl8JkKTCm(@PubHy(fe^oE|xv;2%4`n4hdmH+Iok)GTU(c>5M
zJ^til2l7QU9zSxB51LmyqVeXBh~9h=*&V9))!)W@ZXL|4A9}9MkLgbzb}~;y{(;VT
zejC>ZeSEH;KS6Ym-0VX?ZS&h->}&K<KgND_f_#Twf8?jGvws-}^+Wwoe?)oLe9ouj
zWq;>5<J~(Nr(fH=^pE|XL-V?yG@tV-|8ovDANia^A$t7L83*|>`$PTKBevPiIU7Io
zu%G$;9UpY~Lh&hoT5o(HdPrW+q2~}?5u!65#(sLUGyUkL4I>|P_>za-kscl7Pudaj
z!-u@+^z%1zdLGxgXg=%gzTJ8mPY!$`dH4-KEt<zT_a^S`#1mv!Xnn|8)^qZCyG`45
zVy<o8`hw)8hxL+g<9?{0y~UA!`GP&H7xeeEtc&*~>}ovw>$fiW#`*0N_Iq+ge)1fo
zul<M}Ap7teery~%@nnBuH*%8CbM*Y!bNnK*gYo={Uz>;C<fA9_`{sz|qc5bdHahZ<
z!}@x^rr$b~N56H6{Ng;GTX*9k9lk|z<+*X_^h5LUH~A&Xk4`^(=r`Uv<71uFWiZxP
z{6}0t`dgQ{UyB?5%5MCGA9&A5Z+y_le)G|TJp7j(=*!OZ<p<jM>t`==<D*}l$<F4(
z2R*seo9u|*a~Q|D-+=TLSH@}6AJUK9`nAas+xnsB_>rR$l8c;>9gLT6BAOSXhvLR_
z_9Lf$^!g)uPF`(DZu#83kvP$hj-7p%Y5hI-y{_*X#hr5p`#^HgkKWq!q?fk%A&>7z
z*_k}_aIc_EZs#D#|H%#U)kYWj;cGs6>o?vw_BB5`_dU*&+Uy$XjnfawX<lev`yBnW
z<9PBye4u?D7QIi6`p}E~+D9M%$Mf19yiN(d=lGL@oci78L39z#!w>0!j@=--h}M@J
z<b=lKFFt&iAJO;E=-ijmN85d}`%>Tc%csV>*JMXXZhndn{+^Q`o$=(+Mz0_8SI_Zb
zpK=~vSig(^a{qZfP8_=0kX+o0YyJxL%O??w=!)i(Z}72x=7IQ<j~}2HPvQ-|=W!l<
z%?tHI`P{tt=%<%FX&w2u`R()Sh;rSt(C+dp{?-M5{p3VXPUyLL*}-#d@r4f3gS_?!
zbqTrDPwGRc{y`7fkKE)W7kiR}eCX(JesuDYb!T^cjWe%t_&|D+6XIh$I^(R5`Pnxj
zdWf%i$i)ukqkps`IXou^e#YZxK60_Ieu$rO?57`+haAS6A6?OS&yA0C=11o_`q+=2
z95Em83*#c<Uxf0iapsB0zSc8hd>+T+&rjG>+q&wH{NjA<B~HALGv0jiDL*m}oq3`C
zh+Wa!-^gKqiuCweH+hqu_*oD9_+7M{Hu>mf9(?hWubdOn$zPtk2XGFB@*{g0kH7iI
z>z+tIdU_eBA6>*apSHYhe<44N`MQYCda!?#r_68RzC~_&LgUcWLmM4CYDXkz5k`LI
zEwBH=dIw#UH`05qpWOUHe?;~|ACX?<(#A)>`J=z+cb{NAi|YT>zjQ5N|FkaFPk+oq
zp2IjUw#6ZO`AxnrLUjmw@@PYO2Y>QH{Sont@|o9jZSz6T@zth}Hu>x?5MR&jThKfZ
z9sbaBh^`2Yi%1?ApVzh>=f_t+dF|)w9QxZw(L?rTM{>zO_7Qaa5uJIp(P=|;<hGuX
zA3q~sMC0*=abA4Q%P%2%aul^=zxj-d^02dWwl+HZf$`4a=+u+?tpmFl=UxF_oX@^!
zKXI-!KfX{M=Dr}-XP%o^o1MuK?d!R9u%7y%I)(o9ArC){`={S8;ZH7lMS5+|*^}Jl
z@f^K%7gv#<pG5oO2l2tTu8>{HMNaZW|DiX(6#vkA@=N;R&pzzFxOy0Gz4@IsJ=}{q
zr|`pge^(?2`<t(bu2MffMRe?HJ;+Hf_lEqyy6|r(ZsR=cZ=7|Zm-sQxen1}UO-}mQ
zC-6hZf6!ZZ{GfU0g${p+ujkM_^e^g1=Q%wivIlwOkxJw6;dkPJeZ)hI5Bj5r#zA!K
z#y`-rGdr;xq(8m%)7Sd(5A=|Jke@;Q>;%n&KRL}C+vKO0an5z}eG&Q&%y(P%Gw*fD
zWxRfJ(jTH{2l_hyk<WPaOZT*}@^Y8&?8owVCG;J-JQ>^ieLrs=`+_)w{8O8sieqsk
zZ|c{M@@tD*`xLq8M{a*#$#eboEp6|M=;8esdE_r``3YTYldlL@Ue8;(HRI3xSUsSh
zKl2A|C=TpD{8zia<I26dNxNi$#j}8Y!M>%=QO~IB#Jl~+euhr`$NEh@gpWF1J!hOe
zBOj@g@sqdE8HZjSMt=1h`Sfc?xyWZ6M2|l@81M1Phrf9te(Fbb?4jR$#<81u<!}7R
ziyyg+i*>8ISif=Pw+}>QZ|kBB$z#3rL(k1e&q`<<I(EUAKcLf34$sy7(cbK(pPkuH
zo80(VFUX!ye?)vBxsB5%7o=B|$NC!&)p6onJ*aLJ*Y<bkE@=JPuL${#^?~Ya^=kAR
z&wW27zo>`h0re_<$PLAVHhOZgTWyiP_?ln7WG8yai_Yuna^vueXuYGqM)|ebm)_PN
zpXgWY88ONc+vY1mcDBCOk-pIPZO&i((0cPv_GCZ((0q_SQSP`t#z%kR-+r%cetwD%
zJ<vIqp)0Zve>1N*TT(QiIG11K1%ALE$*Iks`5SsDzZ%ECte5%uJ^Mzqj@FAEoo{{L
z;rHzRZn?il?(Y#l{>H&g8uwna<nQB8xO2}_Z`fz862HRN`jJDMoZ`{<(vUptV%^DY
zJb&ZIzNe-KIoOB&_$PT{n?FRk(G{U_{LA{17m^#IBd<37@guJ|A)olQzU<0g>}h`d
zjbmSJe9`&$&(M+6dhvJuYP`JYx$hh7*WyL}W51U#`8|L4T?BgZY+c2h=jJyLlpom(
zA84F*^dI_JpSUhX{(%n0{ldJ^{N^)`{rQ1@@}P(OhrZg@(>m+NM}HBb$A^FEC#N<<
zk8kYPrdQO%IDGiI`5<}FLC+yNcIS8OkH2<A{z2ZNIAVX}t-pE24~+fjp*Z1h__G&%
z>7xz#kGys3_6yIgJ-hZisk&W<tvhVfq3XJxhxP2x@SFcuH?Lm1x>bh;*YvDj_nY?S
z9jZE1bv(B1wLPnQb>C~uCRGjo^YHucx^3r?*{n5=x$O5Jj>_)dbDh84c;L`%)h&l?
zKXd52*_<6tKjz1qzL#IH;aWYfy6^dHPMc;YwOHr&{Mmm0*}K(@&&za=-<?18)fF?d
zK2Kcn>pQRgGMh2H)7>Mld8_<<<4$Y8de`iU*-zd-f97Y?CS~8>wEN{pclvMKX3ghc
zX@2CV8NA@iQ_eZ=50kSZysQ0TW53vYN(R64yAE?2t^ZLLanI_V&+gTBOjZdG{p3$O
zKXStO?8#l9eQBrD=auE#V!i1jE_r%E2JblT#>S^E7?l64^!Qr0pZ!KYX{(QFx?cZ9
ze$gG>FZt_``?C8EZ}`RPFAmCXxas1ffAhs(vm>A1a@xR0`jv5$b@%@2j6ZyuM?B##
zuWZtE-4k>2ciHlemxhe)U)JZi8STcecIL;~{pU9L^m}c_me=XPYoFhAPWypbrR$5~
zGf$cH`(^`jXq{VJ@VnIp_nw%2dDnzzZ@P0>_UJpa-#%xzA^C`Nh7B5Y!0Q<~U-<U*
zCtY*fEBQ6!H!L0X>idOw@!cnNxp3Z;oc!i%clvP;e7gIS`CG@0?eNJ}<MaD3ZT^>+
zn~unjnA~;h^q)-0+w}X{Coi5kG@m&7k?!XnKPhkd;@|f8?7azj`yF4q`}GIE%um>8
zhd2NF;;{1b>l!vc>Z9$)<<^7U_?<T7ul$SO<4@nWubZ*O%>!RA%SR4KPV0Q%na6(8
zaX^o9{P2JCiYxwZ-rIJ(VvEP0n2_Bt`Mn=>{$xhkZ}j5>+1-5ji_=GLc=z`IY%wvX
zf5#5H>@=m%3;CMA?7z>0Eyrbh_1?bv{gGpH>tJ1e-|3K1v;H(SAM?cL@2=T&aQ5oB
z275N2JhU8F>^5hk4tKQP_lxX<J<ndf>g^A*XorLDe{SPeH;u^nDLQd*%HzG>I{&mj
z89)F0h3Z+=od)OpS3mv4>pKm)wj8k5D;a&R?zi?y_n!BUa(-b4$bZS#toKLXJK@BS
z^3%@S@bk0AeU@MOyH&f5{oXs}_+ekj4(w=s#=iaOTkSf3m2Ekr|AV)T8J=0k*EgDQ
z`2DYp&EyaCljq)b--LH&Wea+>J@7k+Pb}v#d5|9LX5IM<`NhfWTh80@M}HoXf4%0Q
zen)LNAs;(v?Bo589GQzt@e9eP4UL2NLvbMP&%61a9&gMTl|%6%e-xp0*!!K*zBlaf
zw{pI;p5pGtmhbM+ZI@a3s3EN<^ys%Bzv+?JvUfI~pYM8M*PGY4d~QBqm%p}{^xVY!
zrIYsi@wTT-Ea!`tZ`+~sYdg#?=ZCd#yR+fc)h6frwwb^4P7SA&kscAnQ|yoW?ElQv
zy{;TPBZuM$(i4&w@=I9G8w>Xdeh%fch{o@H<oUNw+F(uw`L{TZxXO{=IrOsMP0!f}
z;!ke&kmvXVv=18}QT+UNc+Yj`U-o_m`Ef*XZ=AerJ?_5vl%^m3^zjU;8;Vd|Rzi8C
z2rKQQ>3RLOfV=|L4c5DK^WL3~=sK`m&+<e2kGyPuk`HMukLtHS>X+Z_WAgK-zw11&
z(}Xd3+b%8gK1WT?^SOH*f5}19a{IP?ywS{kzCLu1vE};1b8$NFfS;ZJ?ydhQ=S6y&
z2cj=R<BCum))wuX;(u{bol=DKQcrGw`qR%Id(qc9wC+$ohVmVhXCS*nc7-vYRcaU6
zS%2U0mv$Sl@0c8t6WVW}e#n28H##SZpZ*8T+3wi2KP<=FrW1$dy{dZW&PD19avR5<
z>PL0%;SK-X>f|RsE}ys5BhIDP%eex7ek?xa1M=G6)Nkf{>GSO56UR*{*PF(PZ}l3!
z<T&xA-t*>9%=0y`8?@QJRbQ6tZuv)@X+P_Aa@OGdR`aqQ^UJUAz0>qE+7}`Jg!a!O
zBqx+VApe8v35c!;*&oV_5!F5FQdqTXo1dJr+q4XdYq;-i)Av2?`sw9!l6fHihvE~8
zf2cm7FFr-6u8*jWm-qRLb55*#`Ga_ncjPzwkGOD-!q+~<e|8=G=2@dJe5UMI_~WaN
z#b5s(=k9yh9^V@FFZU1XKIdxq^B-?K?8m?Us+?EZ**U*m<Hm0+*lR`(#fLn_?&cSV
z#~!+W*X^!pQgSYSx#cz&&3I{E_V^)>t$F+r4NBsQeWQPiTXn2^Abw{($>UyuU&yok
z9DUm{P2T=;__WNv;(TX5c}l!Hw~2G>F0R!R^ssK~2IuX!`#ihhCZ(yF^(B{krmolZ
zcyj9|6Uyi7kB&WS)i(bzs(c<5KN01P+M@HA^VwsYzJBiuGpFX(UHu}z%Dd3HLmWf)
z6MvAuKz;+QBjiUA-^E3KQH0{iJ;I?Qp8n1Ohm6YM7cF)i@YEWU%5frYtfzC7IHRXH
zV>f=se&~1H`?^bh@%ZE%S_g>Ey6_)<j=z0X+_1lWmVJ%4Z^&!*ovYs7c<<85!^`Jy
zejtwUH(ozKfX0~*k^|!7KDr3`10*lB4)D{Pk8bnS{I4_U-Ud2%+rQ)&$nN$>@<Vw7
zM!DrhC~rga7a_ld=7sDA`NItte6qzZhfd9%H{7qt@9HQ1<h(!Z;jISEoiZ)wpZtm+
z#`_oZx(~FU*@xv%`zd^WM3;NpAMtf2PudUMm&u#%H`z;F<+=IgU-PLy$Ww&)$qV?H
zmt5{|@Ry%GH=p$<mo|SA5Aqnf@G-A?gkOu3GoNbn;k}Q(Q;th{M!Y&7X^VgJ;p2Xo
zU&>476Tj+RZR3rj7dm+Xow(xP@(_I^dapnZaURirZ2zSXebr&|H@nN%_Hp;co~zTy
ziB3Br{i5FVhy2I+nf;8HugoXjV|?Jp?h*MDWFHvM>(Tz|Fz<~be{~Y%_b}2IwdIwn
ze3N~*e|l2+{!%-l{A1s>ukvg6=hoYMJANWg_=kK4tNVZT<Z};ADZhtefA%&XKaVKy
z$|Kg7--<7CnAbS*74M<x2l3&r#<3?qGA}yvL;FW1RG&foijaOqNd6+UK1E1=XkTV;
za>Al{>Z)IUmshMS|Fs_OHPlJ!es;1?TQ^7#dPd|w_I3V1KYn4K_x^xiLv+x6D5Rgb
zD2i+JmC!u&rx(4o)nC?kQBhun?8FZ08}?QIifixTy%!`0{lzJH^s7_sKk9jUiDUO*
z^z>Z)E-v{w{?-{E`>OrNJ}A%gPxTo6AU~16)h+USME+}k=0Ei0@8}|uga6P|yt4zG
z*5=M$O}3p=e!puz@<y!WcgyPe87B{av0;yy<?k8TS=@_5IQXIuhYc81Rr6uvd$${S
z_JZ<#3%lOB{_L;c=-!}49dEt)uXy6${6-zXFQ9cYUL5i({-NIIclKd^YhLjLhi%$+
z^fs5yD!<2c-g4if&Q_N=pJ`iH@ucq3@4iDl#jgCrIfFjzZ+~Wo#s_@Xe(fI)&x%l9
z;GdBF*_l1n8{)`0OFWVf%Gcr)iVJaKoyafl+i&&UU$=OET=}~z^_=^uR*(E@y|XWv
zogLNqlxF9>HmCd@ko!3HV`uX62X+>>?w_omJYc-I#rKFVT^emZWP16#GW6Dw{^p7K
zhTiT+oomQ(bJZ>DwEkjb#xB;?dHkH}o_*GS@Uu+5wGXIY)Vt(oZ~Bqbdjoci_BEgT
zMfckK-*WtB4L)p8^4$EZoz>)w*=zjwcW3yX`J2CQ`oP3SCH7!X>%rgnt9xYjc28=3
z-79+k;e5$|#O0;W9D3c<E^{+=KYOq<`S{Zlf7rZ7w>KM>e%pP!ww=~&TpIP#)dO2T
z(WKP*haK;}^3tl(@DE$=b<bf{CEwLeoz<w@b>Cg3^vBnp`ta2Y8kWu&GWd<{zdJWu
zcc05Qd-3=MnQ`xSdTh%soBvndcxg=M?8*kEm%H}rw9kf(O6)~n^$5SP&s!J%;JnO!
z_GRP6Gr!?)4?j6~;1QpW%h}t0FR#mc)`vZ{<qdhky$U|QZ{dgLa~~^j@k9RLeCnRk
zc~L&{{@Hg#_78cv!@*y5d~4v`^7lUUjroSZ;w%5lqxN-ed4rzx;_vJQ)j9Muzj4rb
z`BZ)+7rNWB->z{`uhF@4p!WgJ&+d8bTb`@C*^@og@#;Q)q7EaM{H5Q1YhL-oz6RYR
z$rDh#iDU04#V`Mk`=$E@`ON*1@em(*O8ooYNPNR~U!L>$W!p3;*|&`+-|2gg%*Sju
zBcqRbim+1sMgAhRfA}7^2>FNji>SU?QsgfY`7?A*s4L`O%L?U3c@y$OZF$Q3Zuhz3
zfW7kzb~@|P4`!C@5$`R<nY^x^k>AxJ(ci?++AVvu-1N88%JmRGvah=zk~id^=pXXh
zqN2P8<%dcr-qk7g5%FlhhV-|;LHnD0<a;Fkre7ZAm-^8=KgfHY!%FeXPw<Cvob`~;
zor9qLL7rwu&#fDI?QizGMMe2oJ*8gqdk1xu{g8gz?85K8@8);vS$6QA1YdE3ul~4B
z>I(Nr-ZQhS_mlip{^O7QK-+%qJ6v@QKf<TmzE^MGbf*S2_IJ-Kq4k3JMPxt7pKFWO
z-+OF+A>J?D^~eif{c>gw<r&Bi)qCm$a@BWSZA$ZF`d>XKhw`KRtd4~I)%i)iFCNr)
z`sHtVl0Ep9^S8Lw&%b@Y;(d#@dI?%baW4KLKeSHP*?Ot3J(nlM&yu2eD?)Wq5vmi5
zkX?(AU5ij2DME2x3DudkMf-Fmlou<Z{jj#^`~}4~|CJx*L3x$k<SqGuKk`p`gx@(|
z$lv@~{7k#;=EryGJHLDnBEF!wk@wj}9>mW*lJz%UJZl?gpK>0s4>>Q0Yx~vSa~JHk
z-o2AD=O5!QZgcdd4_)(LzhheXdno$`l-Je4?2Eti12muWEdAsKb-jBsa-g?gTOaoV
z&iCf!zv%g;`+eULum?NIpY9RG^QjMy@AHQPCYOKrgP;4!xc>4T`R)IZU%SWq;)ZQL
z8-B~w%>C)Z2ey7<w<o?;I<oUeBWgCUDqY&T-!Xf=TUBB|=PY%v^PGB^z15AzO+4vO
zcicC6YWaRv-C|z%FwS=o>8YNN{;FRdc8~0w?%d69+<&M$)idf+_Zs$hb&UM(9wg>>
z{w^-;BlfxOKl^l>4JZBg?@&+p_NM>X=&dp3-(k{&e^_tdDOgwEbvlp9@8U&X(XTB|
z<$HO`IageeLtRRK{%T(1)YtNlb(bgT;d~`t(0dNuYm3LoUmelrg#%`+|I5)Cf96->
zP#p0~`G!8m;b)vWhQFhCPU7$U-?_>?hWkq>&Q|%L<q1!GZ+7M$bIPN?ntx*Ns?t3t
zJUF!335`pa{rQE{ZtS~SX<&zMud&*BjY{q@&EsBj#3nbs{@QDe%J-e_QPH^v)fQj$
z64#!)pP?^0_FzZ%G5A|Q^V3uO(x1QJ$G_ww{$QWc=11}gKIDvfz<B$Dc$63HpHTeE
z3*I;R?gM%cCZDYP#<i`l|9DQm;YUxt|L$guYQ`OM&>w%Zb;BC()#M517T<fQTg9n;
z%YIBA`G`EehnC0bEiZ{{b`m%05Blq8A4pF7n|M=~$P2!6!PmLhy|?p^`wsOudz%lk
zyLGef>|j3k738<B@~=FHAA7r>mGAssMBXEp`F+o3zp($C4?X?8=drHjQ#X+-+EX2<
z?xAn=H-5xF@S~^s)FIZ(zO(r~^FBCmtGVTK8ve$+2b15#C3{#${zGr$(Xos5u>UzP
z@hf$0v}YX8uc0_l53?Wti^!hj^W1pP=|LWLi}6W+_K*kp6@MPs<&k#R{c3j3zt{`i
z<sTe;b+fnUm4E+B59>@0dDT9{p6XS4u#fz}j?np){pAhy06VY~zcrrx{7;^+Pup+p
zuj&?cf_(vhdE~~gcl}PIwhc@2xwtovHapul*rgB`{?6C>sW%`yh$H>tMm$CIoc$m>
z;z!RIN7hfje9bS7=Qs8R>tQ@P{$zgjv3iVrkiDW^@ROhDjSj!EUoNz-`S>kF2g%1S
z^r8<mulkmq=<j(%c9f^p#pI?xKU0tDM@L_JX_H4AMmdaA&+v1I4||Jy^UIs~n1>xA
zdf%=dRA>2aKwT$3#n12i?>}wJZDwW%Y&-tYKc71%lV3cC`o*>LEB@+c=TqkzzdJI{
zdDeGr@`3O1@$ufqKJR^u_Ym@d`$YA+bCL5iWJh%@KF;6qdwTvLZ;{V9buBx{bK2}i
zU-P=(r#Jc72VJbM=_UU;w^&c-4ROuBMSP7Xr}eUa*44QlKY3n$i1zpQNa!KHAU&Lm
z+?%L3=x1JjTeio-zZ0e2ryqVr_oAu4eprBiif8-5(nfhz{qH+2_pa(V`<D7myvxV@
z4ZZ!^ccSVQb*}GGj5kgjy}HXj8S7AWfamr(`-!&wMx2Tp@kJl`Q{EEK5skB7$)omf
z_h#&9U-I0(#s1`?H+$2IKJqU<p0gj+&tB-EbyUAPZ;(s90qaMUBq{330`iJ;tvGUz
z5dBp>=KO4bi}A%j(eo#Dpty?hV_#HvMKn%*YJa!ixqlVMP+k)6zJH11?2{w&?@qq(
zl=<cF9^KEG&pn%ZOukkJm{(m$fA@gp?<U{7xU~OStCa60-7`Y>r|K-{0d4chJN6Ix
zNFJ5XJf}B%(!;%``c1u4gq8fz`QG{a$^0*mtvYO0`S)bb<K7>*Pw@WJ_m<vIxYzJ~
zX0I{(PTKp%S-HGr9^<?RwBN{s&iU@I-hIDMo89l8R{lF@_AAfjNAlZ`oJ;(k7aw|)
zkKFDl-3u7+y8!P~*v0#DZTROS`}94e|NPwVnDDo6x7zlmwX(0L=k`~KA0(%?`>tqD
zdRrIwC)N#r{OtSW+T`p9hO}wXq$XnYALmyX_uFONj`C@{*Y@6pyyDaShx;J+DDtg5
z=>A#1{N#IY@hablOL@{cL|q_%%e%&@1LPs+2=<gOicp;Z#e;iec}D!<%fBP?V|>`l
zJq@H6{*avVo_i_hM)}&`Gm$5u^OkY!D*ssrd}6=0^A|dKmK^f8y3TyoL%x%j)P2^G
zzUb|H=AplR+B}e7%JcS9=STauI+fkko#uDH@9#sZAH{b>aYC+@1?lDcK6SSGhu`y0
zdb6WEz`pLu?FaT1bo@a5Z65rrgSPyGKYzlHyyg{u)*Ze4!GGi_aT8H}r9M*ksPFjg
zD%}>mu*vMIlJC~}sXQmILG_$^McpPop|}!v=*0=UIcLc$)(M?_NnXz(J;e_>y{D!x
zB$w~0#i9I5ekdP6bxW6dgI?=+@w|NUmG}1i%YM_#e+NijwO^}S?C17tZHVreb|*Ex
zvE95}-6D?piM*trT>Oh)LUlkyex)w+UdBDN`cpnKue!r{`l>tBanSzb_n3z@yX)%l
zk5|=@cl;k-|6ZTFSE+G6@&4I7^o+<}{EK|zNZk|TSUjr(pu8Y1?T6x-|L{}wf;cx0
zfA2-vm*0?w|GH1*Z|o}1XyZ>_-%0shncvO%P69vorM`y}=kgLc;(Ycu^P^MW@?&;l
ze|+sv>L~V=Kg6B=ntkOfb(B0OZsh^v`6qvpH_6W)&X4k%d}+UyH`vR3<aWMeZ}#=O
zJM|&^nvXvGo4)v4fBNW`|Kwxxq4&O|2-%4}<Wc(AM|?NP4(w?@^2pQt9^!*9yPD5<
z_M)ftrl<L=kABE+ymw`9`derFocBxKGnf}Yb~PV4(UG4W$!9&u&tL3s>RI)u^-|9o
z&mQ#BPcPqv+27Qye)q}m?7!mM`!w|Ip&vcFpu>;flApbxaYZQq*{|f4zHg4MK4`nD
z8o!T_ALUPI-;ht_%}Qt=l3(%X|Iq#*pUQ9Mx1Y*u>Ms0G{<1~s@rDg+<Ujj2dT}5R
z8mDbPp^tixzT}h7=%b#9@}Z}X`Ro(;+lSR%>;cPpdf~rsLSB5$i{84L2OVTjdDuCX
zUgT1j@Ehxo9v}JBdD*<|<{pRr#jW=X>@JVFmvN89U+4+Fe{rusPW1?To1b2OPfQPX
zf#hdTbnI&XwEpPTb@;Qd^-`a>|CFcc?fX0VPrh=W=>29yc~QOPUR9p7Z&)Y$rFo3E
zKhVQ^LHXD`QEvN`dW8S*1ANQ=yzt-WL2rKfp4{f;FX|3?k6oQ7?R)fRNB7m*^s-Lu
zLmtRpwMFt3q4P0676;;t|MDO4#&7r|yV#G|mw(dF`q=l_haI4Olb!fC<TsGLp}Y*~
zMLy_UCm)#4y4%<3Pk#Q&pS0Db{{0tv;G@11PyEP!FF%ln9KL6;{^p5wjB)g!r#AnS
zS0MTQJqdLL{^}3kgNO(5V!U-m&tBFE9rX9R)fwsvhz^Pmc7gmUVo@EUpPUiZMG@&y
zgyKnl7BBV#`AFM)Irs1G4SX*o&-$*{`QN>Qao$&pOMWSSea9<)@bw;%JW%|r3!vYN
zpc}dOCL{LyY<Avn+pT`la{UFl^ZSE89Jk$5hc>F&aoD$dZ+6gHH8XCR{hNUeS1<ov
zli%$A>>oG3W0msXA@Tc}Xm52fe}L{q#gqFTb-udUcVOyf=K<sJH?R1>-+c0v@1NWs
zshj=%6>(}_^N2(BGvs%^BU1lZZ}l7hw{G?Y`;G6u@G+m?X{qbvX>sdbhd=vnOxry2
zs(n%3Vi$H4SMJx1lLy7Q{KY;4$K1DL-xl*Tzx(*ZMLQkx$3_iGP@KzuzROZ)=(jG;
zG1iqo80Y?q{p4%sHhPICaq1q~zGA=fI{^KDCuu#!q4gA>>}#J?Ke}gg{!o{x525!k
z>RxrIdBv&kM){9CY2UFP){A`dyngF$pRq3X7kv3O`|&sa#ShWR>)QOC-0BH_Ee_>9
zcBGH*8_6q<ox{l~ANsu;J=m50<~5J{)BZ^g>&Nc+vmg1$;T&9EzlHz)oOoeB--D_n
zoagbQuYJQfe#oBoXYpeHaGy+m{$n5W`wzcoq!&L_AMroGm%-2P0NKfTM%}6I@&2Cu
zy{}Zi!#I!lp`Y`X`NXGuXT8mDf8=lUGEQD+XZNe((|M45<oA65KVdh|>Fc~HPaCJ7
z|C`r$#PSyZkPrER^AUU5@7b09^q{Zx;UDf7*wgtEKXJj&+0XB|`6K(vXYArRe)6z9
zV_n^=#Pt?G@~ilkZ@mB4j`Q<Vb%=OzpDoYnmk;DMZRZJbDbKO1{l>m5&x$YWEdTQZ
z>*G5wd7M7Z6Y2o|WxU^0^B4O&e_}uWEf336_H}vG_tx^Dd<@ya_wx2(`!D(VgLve>
z?CGA9Ui_lyyK{0or#Sbj56La>+P}5gS)B4q{wFTbsgwD;_;oIZ#`72F0?*BdzqpWx
z<wyC2pGSJ<T;B)yJ8bqN{v+<?J#^xmeEd>B|Fy5Ehv=bh<$uPBS9Ii%kIm0c_{d-E
zYF}Z0dXSSpp%X{s<>%<dTU>v3F)zEbH<V}OJ@KS&;6LcK%}YOiNPh9)Ils04TOabt
zOZ=X`{J{M)yV-Z-b$&zsc2_MJwC$U-Gk)zJ+kD2eoBcq&;NBFy`q_N|v`*|tKlPyV
zj5a^<-jqJ^ce&~jc9qA~IrcSjs?VYQTwL;3`nsRz*Zdt{`xrUc7m90bb`~GjS3f&y
z%MZqz$2fBGGtc?)qN4HQT>cV2@+*Ifbo^a?VBb-H(l_csKl6AGsUG5I`q|ra_HrKM
zf8>!D<TLB*x%HH1#Syfh^8@>=dFjc1>`xwkBOl2R{FNX2T|RrLQ}E?S?5IwWf7BPo
zL;c3{8~%kK`_Rk0^k4_}#HSL9XZ-Lnk9p-m{ZL+%N8-LgKYm3%{-mFt^wDp<_!l|l
zOYvnrwAqhb@`&gB3C8jI<q`a#IzT?>C(gO-!;i!{{oFr_M}Fem%b%@}x`Dsy_x%hx
z-4D`JUU1JsKHu&59@#qDzsN7o_&qtf>F<6<9${~E<Ryps#fLm)|4?tLADzGT<7c0O
z<X}JVfgt&<7x{g+%|Gm);@y60f0AF^`x@`ND)#3O{7Swuzr4vWb{_J}Yin+sk>hV1
zI`cSB%Zubh@40=Jp7Jq%_EY(RUpt@4zwVRS&$-X{`Rw36+If~gdS6RUc2=*GhrIgf
z<2m`P55Dwe5B|ZfaUH}3WFKvQh>!WmVSU)mdXiTgva|J<=hR8&x9@4=M=x!9;p@3N
zkbLBiD6jD+<Ji$W?lJfQ`P7N>gSwww?l<H$_H_QR&)HX?IFmQ*!}2ARkKB{WgYt`e
ziiq-w{fK<@^>=o}xw=+<l`r`}zqseBJ?8%Q7xQ!9oA6`u^IQJ9-D8{Gbn_*RYSho@
z$i=P^#V<egTpc0a?3eiQ6YIb)@nL`Vvflj0xlmn6UgP<b{o1+7e&(LQ@3*|yai5`W
z{fxJ7(g&J{9Q;y0ebK9D>~HQX-21qPF~5D%`xX0%{muIna`QuN-*@`H&^-<Q(0z^f
zWZsMMQ|DRtwa%UVA6hr|w!iXk`=b4p9{AJKI%(sBzkcJ$Pd|SL(!Y;DZhXC$qOW@_
z_fxx1{P2no4{cgvU;OFiJjy;09s7Bo=bnqc?g8;3m+|Th^LuYdKku!4hYwd<_&da3
z{yZCg+TN>_=tp1Qp^{6RJ;}!o;({FXgx;I_4$$u%y|3ecekUpJyzkrm^6KaNw_3GC
zU-c7v;p?8AACp@if$|R&e^6eB@;_96LiGVu??d|ow7)}n3CiD4o`Ci(NH0haNN$L(
z63Wl&8+D@JKl)wfQNKQ8!4LPGS^jsDJh!hx_j2-E5&90I2;CFA?{&}X-0U7o-g3|F
zy@vCG=lqeM@@ILnwkR(6r*(6m@3}Y<ck-ULJnkGJ51_-x`)cFFy?)=xh%<b>?-sZ8
z)b@Ve`$G9jK9vvYMLx*B=99nV1>@x($o`NV-n)o@`w2g?fAV|vmvg&1)_T~t)nm?G
z-jDHDaqPauzRcgW@nJ9DlgrEU+I#cwx@PG72Bk$s`5aoOi1H)ke-YU|B0q!HD<c1k
zXuYBP0Ow5ShjLx^msd7vy6%a&dd_<q>x94eJ@y~_G(S*RqE~l_b9Di~WH-Olvd{SU
zvFs!K&i6d*${y_JJ(cm~<VWfn{^7nszczWTvv^Pkq4!>tJoKX%dj7{x#S1yq`B42#
z9(4MhAL+;L<dV;<zc`a0JXfEoW2}q(q)k8d6u#mRU;8w_VMlSvKIUZ)e4urWXdWoO
zA$jFp==(?O80!-Cx4g|B@*X?$b0{A{@nL_L_pGaPhrGxx>L2%??8M%2etAUxXJ_}L
z_*!TF!H@VQJ=qhT{LTNIr;Mi$Kc+7}*42K)F7|)=xfhm?oF9wOIT$)WLiZWax<Gl!
zxm_HIV|9$W#rp>Ours~H0ev8P`<=LTuC~tF&Y}DxJ}0Ms`=jswt&8}F;*7spM`)b3
z_2vismp`HBujJqd+Q#E&J<U`0lZAgbU)y-|v5UIZdWZ}8L3~(0`HKGXg7vpgpi`gO
zuf&b@^gRW7{M1j*hvLdSzLSt|pt!Z~{pQzy|Isrq&n^GGL+4?6%Q$(;{^GnPU)yK=
zJ#OED%X{P^zjLMaat`6&(Z1>f@h%UDH}BiE`H%Q^PI8Vx=ibu&vF}{SWnXYV=04Op
zjbC_wDBg`1&*Bxj&#@j*es!P2U#yq>kDfn9`=Ymxc(2dD_?7%9j=V27k2vSg>}Gw$
zjl837B$v3ge&P)sf8gJ~gI166b7<Y6`a;`!@)PH4>#k04j&;73$DQZp8UH>b|B+9j
zyyE;pp16;xgWbcLm*4Y0^3sF6P+l)W`4+N^c<}uK{rR!{@1E7~<jH5>Lg${@@8t3G
ze$#zEyZU!Y+>b!x*~NJKtabDJl<)ugZ||SKN{!!pvWt1><KD#Ioo5g0hR*xRi0oep
z#fN)t{=g67{%PImX<uazb|$AbRNrdTi{IG4qJI3qK4E`BZ+`k(cl)_~CJw9%q!<5y
z_F?fL{wtw<#y;YH<_D{vGVzpY4NLY9`<3r5<-JO1e>2{GNpE!nIh_B*5%j*v`Ck5p
z&Z)-hm%rVo_`XmcbRKZOV;}I`_f_O4r|0sjethIr`?vYo!@T5Ecgo}Bx34?@st@oZ
zhy5HMerF$o>L~ky^8x<kisxc_*^kgc^z1-Re2q6hf4AT9Yk4N-IeAdN88xK!gdY7C
z<mzj2V?Xh}PTPD?oUj)=vIjlHG5OJJTYq%WJo1$Iw2zy|yymm+?CO4yAF#8w^Oo<5
z@kM8!w;tBdeF;AiU*5xdU+?{`-;Y@*`<pnn4)U=4X<gJk^0>UDUmfP2kR9C<tH0In
z>PPv5{}<h3lf(XH|D~_`ll<y0`pNV3l#lHT&~t6;&)>))&b86;6Z@6AR^3M)`-ywn
z$cLXrWN-FoZ}#xs-+iq8);>!==K$~f<32<W>k?63uPv&V>k8Qos_*%OdR{&!ul?J(
z)pttP+5WBGQSYj+{C-rPV@KZ&sE3_{#2dZQ8OI--r`5^okvKp9^`1;!LY|1m`_9LA
zA?j_<$;F@0`M$^b&U53n$uCa%yL{z7+j?0KdCI!@e!{xQ6TTB<FXu`8%!7`<n}?qE
zBhTdr=P!D?kN5qIc;Y9<^CPIPx4!bO`T3Fef!2$^iF0{~ABaoqZydhnf$C%HVP50#
z@f^m#ld2APUUVMueT;RsZ}S^@h5Ygsdj3XEai>m)`uQ1uAg?+<#)I*Y-0UD8)FbHB
zDdr`=`o?(c<s4?;F;BEJ`S_dtihr^LIryo5dTPt-?n}r4otOEO{Y0C8iXSMBJ!fzG
z9KW^R;zZo2H|XDN!7l5z_@H48JF~C#q!&NqzwD_VL2utTpY`Ey;)Z{)mwxw8?gzxP
zes=I&K30#4d-vm>>zBXW_p2AI7ymb}@38m<`>P||&+!-QXFMdgJZAkMKJ1`hT=RGH
z+rPBsHFlskKJqdB&BHGI9G!hyzE)p4SMY0joqqC&dpY-u=)5nm->`@ML>r1Xs19}R
zau4XdpdJ!W_9Oaeix>XO&g>wc)5|#ZqH~D&!(W>}S%3Z_e&}OewbcvqKSU>gh&Rvi
zweQP+=;XH|bS^1E@`+<~=9hQaksjnVuY9EL(jV#VoBrMhI)5MC_s{a2@1n_JpCq5}
zOU(=Ulk*z%eU5XT^EH2S{uT$m%i%x#U%bgH;!Rw#r~D=^`CUYHm~|De>K}e4AIZPy
z@Fx#{#n*HB-MZKZ%<Fq|>muJ9r=LGspAEY2*y)j{8`M}=`@HpzD2~04_Wg-_0{1P}
zC!+O)&btsFc>_A<*A~?m^j9Z{7y7ZI{ZHKZ_Z-pLcjj&O>QM(@(Xht-@g2`ioZaf&
zRm=Y_DSx-#{7^mQ_XPe9|A&q5-EQF73v%+=|J70M0r;zW()rbTi!1qtpYTWd-~MKQ
z;`j81>LPUnJMlO3Xv=%@jQT|$unyv#yz-fG{FJ}T8|oBo_Hyo!m-)SOhIo?i=x^VM
z=X`N!Klfalz1T&3n}?r3@r{l@$S3w+eA%BK5P$obx<a1NZ$9g*KCzzkm!FI?zkQg$
z(bIhDH2$Y9p^yAYUwqVo_IrH1Csb$X*LHtHA9CBztqVELi=N!Zi3fJ0H@mTidm_*6
zANcbN{pQ1m|7nv`{>K*`e~IhQuf?C|<a6&JUc{yNa$XQu{7D?iyG1B2+0Q)eCl1XU
z`5SK>di8=lCtvUn>!hCa`xEx$XX1i==*4c3d{Ex;yC!Y<UcN%_-&rAt^Be!>|NOvy
z;(SLwdD49Pp>-p-eT09oyZp`{JZD$-=C9<_7Ps<|eMZ|p4b@l972?yr@B0Aj1?}7F
z2ImF*jCU_7FS`fyen7qNJOKHd`UD?(h%<Q|AMYRNq2BPlfpY}EkL&Dw35|Cy^B$3$
z{FHw{=NbA&Bp);nq+dku7rcLi^or=cI{#^Z!lO;kY+O~NUb4@dNB(6W`!mF^2(6p*
zvUv8MlU?{pMEgX<Sa-&@cq~G8Q>Q?FLoRLc0mY;BjdAS#6@TJC{8e4;J&^N_xMN?4
zzkI{4z6Ud(^ArE}o)3TT>BwU~`tm>V8utrwvx7MIKA67ZlRW77k@|yw$|KtN>-XNz
z`(p1G$?1GyACmv<Bla8iu>V^RdCfR-m`7gm-d!GIXWtvKbHoFN4t{jg0~(dT-zTp+
zjh}f>uT5_MPLckI_ECNj{l@-F59@5du%9j}nm_uZbz1MUA#3;Q(X51@{Y^coj<cVM
zZ+!m+7k+;!Kd?9d760tTzSh-uhxRG!CJ&&Om-(xH@2R2s!2WJOS0}QsdXT@6)Ayg&
z!#>U**vCDfc=4XpeXRJDPrYX~kNZFR`rQ*0XXYU{Kj&9|M~bgHf?Vj_%Xpu!9<VR5
zr}Yzu?CRc9{E0_?>U_g*%m?LV_s#aNN@yL5&^;8CXQ2JE6521|BdhiAefJ$xGIrp1
z^1AOS#FfAQiXO^izRzWEzeizj?|-d>{AxeoU(Rdn@4i@ki4$>(9v^wreEgjpzJHP@
z)luZ-m+q~c>)Bg9;hY8SKm3Ous5_j)p>roa)p5r0JIFtr3$3I1`5FGs!}1$Dva@{Q
z{Gkq!PvmX!<s4|7c;NT;ado44%<H-LEaY^4mUrya^mcx=uUH51Zy(_Q?1J8NZTSu#
z=T`U4+UWdyO;G&`)t~O)oHPCV9{8eD7ov9$NFRAS?x)_<+Si=VoG0nY4#taXe#qYJ
zDt`Etw%_e*yXTfS#IfJOi5vBo`bgYCa#&|_`2N6sBmL;XZqB#v9qfb7xBA(|e(5<o
z@_WeN`GvZ{IC)=P$3LCpjTaB{qxiEA+xNwTJZ3(5Kt6-?;K$~%-r|y7o#UbPaelNf
z+rNFMqt3Cu>U(*VJobP2*f{y${)<l$;$MX9rH=ADdG)w?{Qg7U*RO6MXGHeqxAqV6
z#`^)|ts8r)@9B?D9N;e=)raPBUjVJE{e-``zu;f`?T?V(LG=~HAL?iCO6a{7{^X*s
zeLy_QPwd41AbRIE$Zy3ZyBeo0e~MG_5z#$^_bBeKeJ^95lUM8m^sqmm6Zi5Geo!8h
zztGE1_7QF8ZS#`D{f>O3pI!9FapX2HIgGPz>Ns@-|APEVo<gV1k0AOoZq(t9*86^u
z$s167+DF8be)9PplK57aigW(L|Db(Gy@@Y>GoD|nvz+gYgYGfZ$MU*y^c7F`3HDO&
zsTZ}aFFw|@Z2yJ-9+mnA%J=qF`z1fHKKNK4{Y6MG=Q?o?<sIvzo|UiU)wo~C1L(xJ
z_lNc)aYjCN<fryQacdtJZ}eps`QiVucOLLo6xSaIR4k|{MeO{ov0~T6wpWd)c@}IK
zC05X2iyAw|Sh1Hx(MaqSqtRHgyU*SgqbMLMz1jeKEdN*T_ujp~yXVdB-uoW!;rRK;
zZ_k-CXU?2Cb7ppS?>_8Co_5LS%tP#CUNet~3-r=|>?4lI_r&YbA8dE`>jOIGXrKHA
zBFB2e?@4-Wf5xw#IPm9y`xU-}!F%2OzJ&S2eGI=Vp&s)G|MI>%_4$4H1Do}{{;dZW
z%q@F)|4UC@anan?7rlP*|83qm$M=J7KIqPIiyyRLj_=$s4+`fSg9a?}Uf+8I&J*Y#
z_t&f|d|#jYM*a>De&cr}oEH#x<a^=`zcXJz%_EE(`y}q0SjRZ8z%R6e9>$sZ%Ka63
ziF5i1&3>ME<UWUU6y_KEYwn9^kGzOIoR?V0yh9)H$b4r1ihl9|`3QTkk9os=1M`J^
zhCg}U!~FvGGS3wmH<0^DXyzA)9UyrQ#J?c@0-4Vsc@xxq0`Uf72S^^`y&^@%f%#4x
zP>=B=UT6nD@cfl^f%Ogl60gJ)$aoOnj637U`h+}ujeG1yj{L|tV<+nnam9GB599oV
zeGU6B{LOtI>mKtFz4)7P0~uHLS>$uhXDDa=0BN7^fwE3=-_E`YKVl#Kp&zu%II-?9
zUigP~jC$k`+69>}_?`PU<{R^eyh0q{FZjd-`xMp%?kk|VUn39DK5<Dr(k}7G{)u@=
zenF4cXWv8|(hm1YtascGu|5!|tRL7*zU8}J<Y(+;-$nc|4~Z}KgY<{>i+PDY+K1-8
zjefFT@%<g_Bwj)M1JZAhaZ_Z!!FdnQNr@-k^CFHl9+6}JLYxub%m>a}SU0IhyZDiK
z#18gd+)tn%KIbgF|G@kpZqUzs;k*@n<Z<p>*q31+?J=Lw!#t)xAojEGXS`T%i68P9
z>jL?V`HbD<FXlDzLf+w=lXzvk+21hU#3S>9a^jYGgFH0I`6~Np;*4_gHvM6~v9ILs
zk+J{ece?D$IA`H}h5aS<kW-rH2;r|*eWmB!?Q;B%l6@=s_`Nab^0ar~=EI-d^za`8
z_NO4f2gZKd;T)gyA>`=~erBCxoXP9lFLDmX{9wN0AM9dZO~1GwW#7&I7MkZLJjcdv
z@+0jsKIp?Aw8Qxndf~HQW;~E1@AJEH@)-T4ebx)&k8@tm<=Ee&4`g1mE`#)g`xW%i
zZ`N0y_c1?+1L`sV$j?0QCvS6p#yZHjlQ+nZoHLUrXrJ|x=Z53~+GV}N&-jP@tmpjf
zqj^rrch>NiBI`Z*kGLdH^4yyD)IpvHlFx`2{K5JS60h`=^_F!NnsK9_<P+irdGZZ;
zkLLxfmpuO=|1iGfUDg%yH{~Gxqh0npw2z&{19IHoq6b>-<^31z<9wU_3-f{dJjRuN
zaesyVj3fI+)-U3gcw@e>Zh)+l$gwY{-`tOJe}w;7Z?KPb1G^Xx@;~bqc2S>s!FVFa
zz6qNBm4)=1_Spxq9<k052RuJ#{(`KZ*v&X_f699S#2tB>@ne5S98({8<_+`3LgI>b
z9Asa?I>GZZ<}Ld!&Z*dUG2Zxv{RDBLeq+A_G7cd79nQ1)-SeLPx-Z;ymHBIVZpr(0
zd`F${(){i27lXG}`YGfbn%^~$j|N}))1&L%)uEO=$M3HB9Rbg!XrKDn$Gjk~GQO<4
zApRr&G0xE3KjUxey}sq%>;L}fj<xJN+0XJ^j(Ng!SoV=Tf93v^=kDBB5O>Tw?!y=-
z#*=d{&eIux%2^MYSHvCrN#;HKW#%F6(hv5<ycbM7a6bzYcleopaDRXwIUisilLv@z
z{7e2}zG5Hs8CT8;@HhRYf8;g%MSYNZAa>#h)>G`lzswi>LH@yC_?7*?&NuQU^BMiv
z&v;`O`H=RR-^hXZllVuTa|+@ZBrlM!v7b0mq(5pGaYjGUPre~9Vh8pyj`SbD(ht@b
z$~pg|J^0i^FL{~v86WhKC-lA_yJ;VL(ZhIQ2lg}Gj321_(96Dn^{KTXeqx`>xRYme
zUf^Fv^nmQ=@dM+4zi5a1Y><46JpNZ?e8}s>0r#=w74{*-F?ykCk9o)bhkY$_^aFYN
zPrM)pQXgd8nTH_#!XK)ic*kz?0rp@Aex-fn$*1_sLhAF}nfXn7<XP51^m85xpY?$}
zMPB9{f%m=WuZ|!4Qq~Q|51RFZ{Ts-AH~zpr@&xNE$a=;4z&L`83-atEi7&>L^Bv+I
zq&{-&59vSa8TFB4|B5{82S^^Hf7nm{;l7eM!vDlI<3L<6Z}1cHtc&D3;*D`Y9zOAd
zKKw}jreEY0?u)qJVjs<Zf_7N1I7eXosmFeYa|aN+m=CNU#3lPr{%!;NQ$_4yyyz$Q
z&5RRr_?0-MUE&?P7;o$&&KXzM8{!GXfA|-_5|_+Z#(}tD9m3D}k9`;SC&)2A#2d){
zIyCh_?sGYZ;QeQib29FK=_iPPLH1qvl{hBO=`a3f-jUBK$6x4WT<|}BC*Pt6KM-g1
z8~+iXN)xBdOOQC=KAt$IJ?z8|`o;c^d5OPteMS%c!(QTn{Tb^KenOsg7(Lj_JVqY9
z#3hJ7(MLa+AB-RVp&!f#{KPyVeu;D96F&aKKg1{cDF;7Y^vA2_bYC*Z`6cJWoa^cN
zF6S`355V8$;d^+*G3U3$H|NltPhWrPrhRUhuU(G!AUOBqoRM=T>|s80K0tpt=e&Q$
ztv&7v7RYhl%Y36gd60f`Ud;13>X8o^Kjty#c=(%fBMw+ch(GFcpUXN<o+NHruR-=1
z+}|_bc~6{uIrq(!b6?H<How#7T$cNE?#nqx;JzGv#3^x)etzf6eun<AZ=k=N*YVt%
zIAwkjcf4nTKN(;A%y|>zO&oDf$+<TBbjFu?$ofdToKvBP_jAz4_b7Qkjr}@#4tt1i
z@;dF3_lRTmN9?=lAAIH!_vfs`>>t@@;b+!c@+A3)`*vu`Sr6D(qmOe6#({Z+pIHBR
zK7v2#7voD@bH2yCAs@1SvoB)Zzz^(C(Mul2Kg<ut4L>U~e;F^%)fqSDJLl7kKX&0)
z<`3VWV131}%unVy<4Ql-U$UO^yn;N>cof!;{QHIY3p>dF=qJzOXXXoh<QYHOW8Gn#
znGg5@|1kdek-X3PK|aD?ApHT^_o0vVhH)pZu#0-==ey$UL&-NF<AYy0uOUA%ZtQQE
zSKPlbzZf6#9OK3~u<qhlXzXS^R%Bhbko>?n@%)r{Wu0bxnHS_GX!b2UPhmdu+>1D&
z9iCH>C&_oj0qrxdXpg*0KN&A*`bj)PQ;s}-hObC|f{#AN2|oKi#tS|)^AtO{FJL}E
zv(EGTAilrD{W|tw2YSha%uDVIiEq|3)-T4Jc}9K%$*-&{%opq@{?N~O5eK}dNjv0U
zo;wpK#4iY+JWO6-{8<-i5Bu5Av;MK~V!y`oPVz0!l__UG$i9r{iHtLLVZZhd8pJN-
z@jJ-A5C8Dq5c^c-0qX|)LG~-$cd%b$9Pk(Y$4`t4^~r<GXAnP<M~N@un|(es`zjFs
z(T<KQ@yY!gagQIc3!L)gIsaYqKMUpX1MgY!_gX;m2*|$OLgD~KzYE#-E7C5=c!1=w
ztVq5~i^RPPnO`9L9FX}>e^|HSGhd-W;trZTK)xbQ*r#Z`5Qp5)unvO69rZz;zY)LW
zOI?4-XT%}<Q0|+sn|^cOfIRym;+l00f3Oclj(ku5X$R!~jQu6+I>>zwe&T+L`zY?4
zkYgQT9fHq3hWj=4Dd^!jGWw8*#=ks|q#d63qDSxNkXO5*xqrh>kaJV^SIF}|5A8Bu
z<R{M0S?^hw83*zx>nica{t`a>ZS;VQFL{&v$2!iwiTAoWSK~c2_Up9EeHZhK^@n(Z
zPy7)l)W<KZ&%`79SkCXsx6BvTeUSY&^VdTBhds<U=BtI+3lg7-^h1&P0J451Me>G)
zR)3X`|L_C%pX5#UQS1kJZc9ATKK?@wNPLhNnD?xY^p|}A<4T;6w}=C1)dNlc$VZG1
z^HRqF`*l8HKjRNge6aqrpTJ(qsfS;f&meN-Z}`-Mrr-3Ba`toh5&I}-T(OUG^x_BP
z$miU@Gk?&}y3M);&3Z^Z_HFFrShuK$Ke!K~UFHGffgFBPntsqfkocfI@*4NY*vI>;
zte1=f{=@IwcM~7vH}3b)Pkgf<fM%Sihke8)cF`W^G0Z>CMOdF1H}V|gL)<f;u#0iW
zulNNTzv55&i9c8;v77r%#+5vP9>#%oLB^Nok;rpjh#Yx<^#Xm!lV9;4_v4h~PiXQ0
zeq$b?2mOpY{iK|BC?`L`XT4-yMj!nJS@%Kq5g>e!bsNObApNwEbr<A(hWKPXXFtfg
z%D#a(WE_}>>OcIzK8O7n?<eayk)EgGSK<<T@EiL&#)JJY@j`v#mGMTNeGW+cvL2Af
z*=NyD#t~#3m|v`i!~^lhehEJzPoAV*@+IR<+>tkt=Y9^o(A?icqYq@C#{C-ha(_hL
z06CZE+@JhTJ=*7foczN&0n#3M9V8Eej1x$EApHP&P6XoDq=+6D;#Ui?55#XS<oO(k
zpLo7QTod=?6XF&meu!J*8AJ~VpZg5-GjGtte4-xrPqYW0`OUd8^MgFYeG~h2>M?(4
zPm%RRk@|}0cOiBuGT#;PvkR$5o&mLAE*ihMkaZYj9MYot5t{GWqmT8%Lj8TPr@aTf
z2imn`KCyo%Kl0p}b0CoUOnyd=a}b_SbKlK-k>m@WNAvyy@6EHHBQLSfV1L8!UfF+=
z?|2T#xeWVF>|!5HUSpjgZ_{7)_1tIBPxf=H`>aRoJCP?}vtMOD%JVqR$=Elt|I~Gj
z`wi}EK<pupbN@~`cCik#FJqs{zKMMw`#1JC+)p9L?@YPhrvLO8zbVol^4QC`;&=KF
z;s?eRWc)$ew~%%%BtBFh^6V?vhaiVNApODL$P;HeF7(er?n|;F<EO|znEQ3!6JXpa
zC!Xm)^AA507xYWp;l2)kQO<ss{WkM}_K6qt(?0QueLTP8{+0VK&Mk;{+F^c>cbFgC
zAF!XrUhaz&vCBg4dy*p0h43r$k$nsCgI{=#!M>Zk#`!z-IJaa!%X<#wCB{t=zgUPJ
z77|zF3GSaj;(+mBd_m5OK;pxN%vbG?(%jctsN=0PajHoFEVSaKT72#+@h|Tm5KpY<
z-2Za!MEoGnx=B1TE;^pvKk|M8=XX3`XS`T%nWxNuQ1ct@fvle({Q$`eAbAI*zx+;{
z{Gmv^(=OwVKgk=+ALb=A?@5w}Sl3twn4i3t#C<922Yl8?T}K%w%2{Wq$9)g>5ckXr
z{Kve5Pk)hTJcw`n!a7B~;&0*rKl2_3@kKrLH}OTDq&?1eIJX07k2s?~d5H63;*NbR
zH2W{kXQ@wrXrFT-wHF#W{D^+ew>eiQ4_e4~+m`Bg(#aEd>5%)&($D{4+Y#+^tZTf-
zLfkT6I=($9x8tAY&++~P?_YyUe|6N^y*6Dihknp{4@2*Hn9sZq!}<hrPQ-f_F64aA
zg^YtD@t}x3S&{dYEadrdTC~pHlutfXWIhn*<T3KEg~;=r5d6(|dO1&J{OC7v!Tcc3
znQydD9*56-<M}G{pZP$0<QMW5@kQQ1KkL_38|5Q%$O%IZ8Mu1Qx;4FO+MhpUk0CW3
z>i-?sd0>};-D}zn9x|{;J-tRv`<nK<?7!-HLk8Y>;kMIOZr|?8i{9RS;kTX-7hU_#
z+owJCL}1|!&kq0X=c7IeExcyGOFuYf-ASQ^>)$<Y!$qf#37QMP-*2^E+szyiC?;+1
z;ngl0bJ8JqhKi34f9;7MKmK0;YWdzjzH#jl$4v-a=(5A|yNi!r#xX1Y>i(sM<pzEm
zTDW(Infvdx_Vmz&#K$tTANa?UcibF+H#{};>;-yG2|!oBDA)1KN<Y`_wtIHz|859&
zxogL9zh3jR5Onp6a_ny|bd4YOy3iU&_`No{XyV~BzX-u)_WFJA(YHPpBt>f+bbcu9
z>L2A=&qXu;&Kq%1ufDs#6<Ub?tl00+Z|0w|{LH{Y>}V;}e!Q~Z=yUhIWOiU7^U6Z<
ziG|j<E1&Ul%@eB~_?DgUk+Tpv3zLqMrC<H8G=8v9{id{KH+)NPYv)`01dFeJO-ozz
zRrzf2UHU8auS>tHoOP_ZQ2nCx_X7@mV5tK?3N5VEE?3^L;skr}pKE?=JDNACPh4of
zn1>+kyAXK`Y1cxvQ)%s&(wb+K)_$g?ExVKQHQpFcknvEYT+xab_}c%hw3QcLU;IzM
zS!Tl5frZ-NthA1EQaWpYt$OrxSg(a%tm*VsaMy(E26a04I}?$A@uKg4`22vMf<6Az
zt>>&iwabBfEqvg2BZhV;(C`&q@nwyV#dq1Qan@RBYh0?uxAK;YpERyreoPwI*0?HP
z?bdRoR~+!ze4pGgF$B}*N0rz1mDcu^e&W2lw)^8^Uj!Dq;&}HRR_nLr?>pu|jT5!U
z+7DZN9WS-N(($fz-fMrMubA-l?N@KGNUepe9~QC?(D8;=zbdVMsuu0?H|r|-+cj^=
z+ls^!NFLWXb&X%ra+jTH%MbqW(#@Y+dV!p)p4vzJfI6;9Bd5qXDPL*Tr?lFq^f&kJ
z)BCk)Q_9W_=(p}Kt$B`KtK8yKPf^=d+R_J~_ysi&(mxlfKb6+;P}<55?0Z@Y$sZki
zju`pM*muIDXxZ=LlSec!STy;;!gY7K@TQH9{v@za<40-DJ1&|$U2W9(QCj0iX~r=t
z{&CVvlg7>WerVyG<M;10{j))#g>OEz+%wDm-*Z7$%xbsRBR?t<ci#`Vuy)q{U)G&d
z#N%zRE<59y4IdAh3%6MKu~q+Y#GJ5Nc+eWl)ULP0t6^50(0$8xE3Gs>u#miIp~hiz
z(Trc#d1v`Y<HSX?f2$OEe$-r;)K6AFTzT9oPugx$xu*`8|MOn!jR+N8<3zdUQSybZ
z11&|T^_wLRU+<$_^je4B-LKu#KNOyOTH}Cz3kz|Zf8NdfY%N%@FX3})%Wi9av{k<L
z(;8pqiG_O3rnD;#DYx`naha6Q{CA=HV{Xx!r<8WBtFHXzs&B3D@Lln4jU)A4sQZ8G
z{zvckbib~C(0*(DsGpTq{VrPb4)g1(!GHhz^1Y`Q&bzsv;T!~n&-+E2pZ;j=!)yE;
zzy}B1_|m9-_G_QBkaG&qlC$`<@4{<d>$TM0_kV9b$Kd{&@r2g#hOcPZr+lpkO}pBj
z(ynnOUs*^yit0a&UudnDmPSunRQsUSze?jj<Q0iiP}c{gRlkeYap-sa*US8F^n5vx
zxF-%Q)cLOY1KP@O%GY+ZA6gEr{)2YKrRKMj_dfZh`;VOzuC><AA1>XqV{Of(exI$j
zeD~TOhjiFt$TdsW?l5-B&3AmVWbO7ReDmR~kv(dc``c-K58HOB+M9;2HuZx07A;`!
zo0os>;F(L;?i${`%O;!lsNG@jZtFhzb@xI&%6C|J(aV?W(6x5wKks(dy7w$m3%|?u
zFHKo_!cw(oym{N9M-A^%d+)fR3%xpL>DpBuUi+KAYjmsay~gLk&f~k)?)u4y&2O(+
zx}f)WiyZXSDOWF5+x5GrcRc;mu7&#85B}!1H?IEMQ;QYO;oshD?V}#OanahLJ74x<
z=i@rn-n?S>ySI9&W9@C@cHQZ}3(Z%1#TVCJe&h-rYOfu5;wLL!*`ao$AO8DH?-SeC
z-g@5r$ISk)UG29Q_WR+BTfYxq-|&pli!3%PJpP-Wn~eT?WMQ979tVjF3$67Id5t?=
zm#p>HwV%;))ti>q_)^-X&vlNf<>VDm_hb0qg_i$Zd{=*}T~0hG>NrAc+$gQ%1Fd<g
zrD?TOY27y|oz(9dFV&*)V{;?pY@zl?;|AJgC(l1zezo!#eA=^6?M+L&`k`?Iz0-$N
z58ZY8S7EiWrSgaFZ<NOWiZ1)`KhO6pBt9&(#s&XZiaP(HwZ9haifb!gUH+pSd$OW6
zzu{+H-(2-5ryropUskzm9kJ@cch#rdil-sJn{s%~j`P<lvhQ;t=b;u_?|CHUGv6$9
z#g8@5lJXgE3mFFsiANV&^<8{d|EpDQ^&h^=PFJ}VpBA6Ew$N&i-|JY&@2y<Oymz71
zPNjV7`6s_WvXJp|q17)do?QOYaw}h0`r)e{rLBI$M~@42zhLDHmDBq%Er+(wC9HLo
z=dV`0SIW2EOHjV+cXE`|Pw=N3_vrh<nA-yjUHS3FKmRXS{lf{Ng;u=6XZ$V1ZVM+Y
zyYI7$d@wb%(0cC!K6Y5B@o3oz-_m3GCoSLF7h3UIZNA1`OVLT^k?TDjD__7Lci0k-
zY&>aJ;6mgrblIWhs>gb7WzPKH&j0QEPX(ZBep+#<ac9MC*7XQ|+K$rd2aC4i%#w%i
zikqbK!`io0$|n!GP}kF>w5$IuRql#YSGnaMm;EliuKLVNU0+=FwA{+u@U8tcd@CQo
z*L6y1%YSM4?fS2D|I*W6EIdD`H2#y;x9rsXbm=9X_grxMv7rk!4=JtVt+dK3?aC{R
zn}y`Br0DV^`_!bU^G4$-Ep6@JlJc$m58u_V&o&RfUA_O8K~mK7L6?3VC+q@Q4>P0j
z-52tc%RbFVO1tdRc(CZ#?(ex~pIzGJEads`n1z48TEEr43D(|!(q)exF}v`4F6e6x
z8vn)G2h9vNU-0|qrpy>0bhzNE$N&80jUoKUA3S=$3vZuN;P*T5*3Z|SF)>7Lt4^m3
zetY`61zP#^7e0P;A@X2adF)!f|GPU~@Zj)}`slNedgNc^s1L$Viqr#XM-l&mlw%M5
z(RPs|Ke2CBxpw*YVaxdW^#(nb9QBX`u@}@l?V?@t$|~1&Non$eg*rd0O|$Q?ka)1r
zn*Z>%ze@l5H)F1wz2(ngwb44if`88I6OKA^k559-HU8Bux6TJ$eD;ABy4F$horTO#
zMaDBL{^_Y*MlV<UP3S`1Un%X{Z&>}Wbo`ReJ1Y;s?{fRvEADp8)F3HZ<CK-3v>)ns
zrLBCSe9jk}3ti)?aa?WM<xeZGC*|vTwbE7|fM2P7R^Es2^1o}oYP?jNc8#~oe%Cm;
z=CR)YRXQG=<GK2!?WU!(=J}SYhaW6-^|z($wAyKI`z-xE*4*>{%dVUiT4>GJr0b1q
zK3V$VYg}9BJ@766!bi?R*5$OQ^H}YM*7Ja@^m0SSoObr9pNAHbk1bStmA3o?-`ZEf
z*K=si<5~{w+V37TeZ_0m+~%uL5&yPUOghf4^(ATln0JhqqMk1@KP=R|s_~+<-q%4}
zc}V$Ie&M|$%TH>D(wBbl<o{0TJu0wJ{bJFoS7~eB!Pohqw3UC@$F($8Yrk&A7ksPV
z&CREMjdK@GzOxX$7P|5R@nWLmJCw|$o&WxqWfu>=cfKF&%D<Fbd5C#lDO&pBTjRny
z;zE9Jrsw(4)3-lqtqxDjDSQtIKK1w=6EvuD`ujA>;iHH1Z;;=M>F=rFBS(GsDi5vY
z&`FWsmrXhGAGcg{>CXYbR{`li{Zgdg$YDQ-9+y4*-VHmbfAKdzzkA4r-v+b;svQ=M
zJobUeQ%-vz{sq;3^jOIKJLS3#ur6Bft;2W4`?23Xy~(Kir<XmSZP)FDg@#RiyzIOG
zw3`&I{6v4ybKgFDy>jo(9~b@(0(|htPD_lsZP@rixz<O{wH_e{y5wocg^xV6Xuk^%
z8dLW74$#kdX2s<$f9kPo!fA!S6M$Yt`UT=w=$1m(6-C#2t@|a5w)E(JR%tz-RNB%D
z-!-3H<?O#L<o?z|mwujGDC&KYMYE1tXxR<lwZ2>Zf^V(g@Y%OmXxVM?X;*R3f<vF5
zeEZ0<znh}@jCJ6_z7N0rpD8l~Q2k)NCro|^S2*PRTTc4b>;NQ>Tl3bMAMmaHC49bf
zhkp7;zSaDqw3QcC&Z1rU5PLGCu@AQFPRn=M#eTp-^^4{!7tJ_>R(|0An7{wQ_*?Tr
z`N$_l%a1w_m_H!#Kz;fJTKBhVmn8?Ec0lZcW<D@J@Y5oCEp)|?mFKK__zC}k^v8v~
zZ;|x(qNaWMd1tk!_T%SvCyv@?mBl}D=3VM1w$t%G6#9~)^}9EhKXhI}PrJSGcNMPs
zmOlr5xKZa{?>MRKK8O1g&=tST4-1Je3pc;{m4k-u`CSNVK2rO+PX%?J=y>Qnr(Ym>
z1Vpcev}gH+xCC{4;Df3MJCdT-*Z!*C*%yJj?@$^!Q2V3fqxRr03$=dIej}%P$zPzx
zgR4FERiLiJe3u7A4pjY&vxVv}wO4847F7GxA386!J+({Am8M+_@dL;>fI6>Kzxq$(
z*F|gp)lO|s<&{?dsC_P4=abUpb5Qdie9-DY-!EEj#pTa<aEsaDwGW+s(evjoP|J6Q
zcDw22>wfOrz3^S2K6};vwC^EZ3g0K9obMcgHBax;^PQW!7QSc1_lKwt&3BIA7wR2<
z>whlaruf|=^kNs^NunO*)Q5k`$QRDJaQ7~S?=juH#Ier|{9y6g9iN(S*0GB%Q48IE
zf!|zy={cPX-&MM?L$ATTZtq<9j?v#Q`1fnWJ1$iCz7fB3S3bXs=RAV@V}8HOI1x{r
z*YQ2N{+o^JcI$Qv<~Z-<ypg<tyq;Hb{s*cY@*wk{IL9958%Uf|4l*Cnhke*hf1v3H
z?Wtb)isWnZ0!aJx(?aZ1q(1HGe1*nN{6YUIhmRb2m;A}PYRy0D(=QPF7$<1@0S&?j
z;ZvV};!pAw{YZ-RPm%s7MeMN98c+DdxrNAq_{Bo}lN1>z^f9kN_B|GAK37`zOG*=0
zS&?<YLf1H2_2A<N3$6J8A3YW#XCd>?LXBUgUGqW5D=n?<K@%^o@z#FvTohElpdVDf
z>ba!0ukspq+Md#?U;TyMpvvPHP~%GFRlmxsecB)Ge_DC<m+DjdXrJ>lko|*&>?<t9
zUs=&=7e4zOMefTL*STVgo8Q=^T`kCXgX$NZhpd+%>lmo(3GoiHeuFxXblfx#(k`g>
zDXnpnHcnk}tNMu}5IaHmp!yp=sOMg+x2$`tgP^Pb+8>Qe?GNMTLi$lwU(V%qe05&B
zXsxfb+DTl3_!Y!&pxRHIf?AFqP|G#GwSQ{A*4O^1KAj&bPuzi8t~7F>mTUX?1yuR8
z_G$iL+(6o~P}`#)pxTFCkbOHyJ&^jW4<3_wK-Wi}BfC(a+gD1Hr!3U`n3lHm!guwT
zJYu2iIiap^N?Y$KCgp4Un$PGbm@GQlj?`U;FF$yTuzT)5Kc73`*)zHokn@+8LM#7S
zeD)0<oBlfzF8`4KL0xY={l>LN95<ot{kTf~!up#Obsy~Mp=U49bBgo5!j|%r?t88E
z4!+7Ot@~!Bb>CWT+7;jA3q`flMeBNwe=JmarPYrXopjz=a;|<^_264}Dc{wORc`qY
zzEuxCa#_(8kFNOAaV?|sf49!kgP$#=pP;p4v-+XqWYJ}I%ipu3eG4uBTm4qPWj}nK
zztDQ$3T^GjlyCVRzSUp&*r!N2X!*zDyT$|iEL8udrCsZh_P5%!)vu)OSoPtj^@qx5
z^$&6DLhN**r6;MMtoXC|=p~P+UtIRO{DwZOU+|Gr#6AnHe!*wHT1dYwL{D>}WtWSe
z)E>)k@LhUbdCOYAHrxC8z2EM&MD1Y<e0=e{J1<!~{+`dT*!tyewYR;t+`yCfU$(aA
z!2NFNbk1_MZ+^V)7h^wOrhxa~-ec&h6PK<%{<kM}+u+I`wG$t|?~Ffh?Z*4}?fk#L
ze%8JApC@+=UOuyX?Fru=ebU7@bgP~8r#r{LzEii_5j(8#(BHrAT85M__OGXZJ-B!G
z+LcD^bXe`s?zPYt^!)DD7iV>=ReWHv1@|BQU6<OwEgW{e?u@$reYe^4J0I>`yW)F$
z_IvY%PPMz=IO*1t)>*K2gMq<8Yb-op?Ru;4cKnG~{uFM%{BL(1-Sx}x(uMxG>hgPk
z65jKA$HUg`Iy(I0#AlCrcene(euIZT_u$qq1;1VA{V5B)Jt}zVug~u`px?OQ-ZQs2
zYpbpQ8}y!Y`k-|-doLI~VEsKmKH`G{B1bv%3R;o+)N8-p#}}@)%J_ml$`w`a|Ay`U
z?q+{~r%(@lAniWB{o$)#ckg4t&bM4K;>qyY5bSf_a<6~zzrP3c8>F7LkKT2*y!Fja
zn@kH<-09PucfDtJ0bdxp$<dRBe^L0}VgIRHZ~X2{p9OuMS!4X@6Fv>@zu>JC+I=vi
zaQ}vW`~|{Sq})QyUrO^H3&{Qkbmddd16{~-O+|eUmXszx(k|zf@WG_#yvSRq&(ok?
z`3k!%wAxcX`4QB0&>AP6e^{t`pz*7s)nELfXvw?y+OE?2T--%l>xSBE(Jp(fde*qZ
z*XJin>wHyO=M%Kc-sURTae}t?4=y{H2gE%H--XC4TJ|L6t3IWZ>d7jn?J4c@3+1l%
zt$72V@wL#Ef6?Q@sbgL{xKqbx0_F|qvY&oh=xR5s{d~WPd}^VNo6=T2_?i!tcFix9
zSK3vdaj?+xJNCO!&r_6EdzH5Ga#Fs_K3Du%^8&t0KXRKd+T*+tw@nK{{OTGX*L%0t
zz5%}aGcApsSuyGS#81}!pNntBuku~~c9pBXYSZ|c{w9rcm9zY=a!S*_3$>m_Gp?Xz
zm-1chTjd>Roc_cDGslN69P$2=?_4+S#XxbF1<tweSJOuZS#iIcAANT8tuq1(pS$Gg
z)7q^!F0k;YMIN~Fu&*8oEOhlt%Pl=t|17(b@~wVXi$8JuO)vQA!}~)E=X^D0_uD?Z
zJV=ZGeeJ@trmi|IaG}*Md{_O4pIP*yK{F-<7HT{yZRMM^{mm->;m)1^d;h36!b<U#
zPKRH0{av#I3#W9S^6Nz=e-kM7nYGP;aXZfnEc|R_$JO>a@YBFT#>s^%fBF3Ze_m~B
zsOakVup3sIf8tuRLkq2OSH8|;r8n5<#R2;+_j#ze^-m{v`1rt2LKj-|nYh?xrT$}2
z>iJ_}A##da{bP%sM;|^paG}<-=x6#IdgN9!-U-1CkLh{h@&|noxUk<HEA_hP$)5uY
zt^VscShQuYi*L1S@hyE{{`0nra%X=ZSm??_R(tR@?<nn>Kd$ki+?uBcoYZdkQZwcR
z)kZ7dz<<5_7vZdR+UNec;_EA%dB>zs(aOInr?k#rr8UoFr8Pe>Z(OMMDs7!lSDSCy
z>1wCTsqgLh@{DIgMQi^9-xbG{Yy4ZZl`mZP6?$%<wCg_A^}W)x<?Pp!qMmyy?Rvj@
zZkN06<I|QWeQ&$#z9+1?+F?J2Y0>)rjm3ZXq(|1+_m%dw7P`*KT<`bm{hQLR_~-k6
z7Fy3g;9K8~ZY_LQ|L1o3c^x)eaO;b{395}(zIW)3>o5F$(A-EIG#9$^lm7lf>D>KK
zp7!uR7R*_={euHWom#ti&cYMk$bGry{2g;yk>C5LMSjoN+_>AvoBZSaZ5GX0$nOBF
zjhwS7a(>!YG3z|g{_^+U_PgY;j(yJlK79Pqi<UqBz)uT*r;+c~u|MNGywD(W{M|qP
z4k0xA8|72)%RlsAZ<~W(4%#}p@<?kdx9)e6@|jmj*J0wth1PyHsa=+w=53{|{Ygvl
zn=3vTCq>tL8LhRv<KXS5Z8-J&u(^@%GqhE-{=U?_!msweTIYw0PI~{`R^(j$b=^l=
z<t^R+E55k)@*`L1Q235rrL>;cShV%ra&Ggjd19@T@U!MUt3A(ud-sLkdOoz~NwxWD
z-v?27>+f)R{+8`lT4{XH>iw|nzRxc5!PL-ozGs!Y_UXD_G}r!B^>RO{$oz4wH&*_J
zPyYL*Amf<!oO)i7XWSKA>b|d~+hM$uVx{vy@3UQWbKlpY-+HetD}Uck3--P0*@bh}
zLf8D#cuh+mWc*#%YVpa(Sy7)eRhxFjVWs7j&Ks+Kwf$h}OM0%CR!-j&hjzUW?Rvg}
z9t-)-xb?ju*K-i79(;b6XQ9<U-FI6wzXP<8@8``ew4S5ubBDAv{;d>U&zG#YflvQ<
zj&<-`)B1e2&SE+6pc8L6Y4Hx7a(oA#_NbQ>U3uJ<Pp$b3-*q4EI<IimclFbXZ}_hI
zj7PPR_FY)%b1K*PQqKO&${X;ld;?$e1^PkPe01FhVuyt;e`2SFuKPJxyy|_T(ysow
z+IPh(e$I;4J^;SuU)pocSIUvM(B)@Wxuq9Ac39|IpRM?TZ~3WOd{-Q~;-u1YD<5JP
z<Ac9J%ir)_cIkXl8a;ChEx*8bjX&|@Le^JB{0^q|1NA`uel+c9enJj(m8(7I1+`ph
z##{MHV-I{6jhyNyo~?01uOj+f$h^?{@QEAjKn}h58+rKjgS<^W<U#oOAA}F0N0D}D
z*M;b7Zlr&eqV|XWTF5wAXvGnHtDeQD9;oA^v@5@&$3oZs&2_)v8V~ii(n;e+`>nL4
z$K@Z@qqN#<(N_FecEPvwCFNV=S}nfSFZeG1y2@SeBU<rk`A6p!`IhI=yqBckBjUb{
zaiM?Bh0Ghi>!PUV$ojhpmB((@jk$&Ri~COt>A!{a$3m+fe8yYR(g)v)tE7Bayz9KQ
z;#=cbX~q}3LHrNG2k{SxTvD{`z#i?7tH0O<qDK*Xt@SAB__*Y-15|mnOKF!MRiDzd
z3#vZ#r?#)O)_29HOJ4OWt@X8B>9q11e@d%<Eyu5*whtd<{|vh1iBnL=H|_ocJ3!T^
zwA!cT7Oj4RR(bq@JuQvezohh_34PvPsl&{Wc=+bteR{t(ZA!p>`NM0@zW>r)zX~`f
z;&~D0pF9`goRWPD`Hp#oe*A}h_)%$<r=Eq>w-7s$qQ;}r#Ic2HCw86l(hGN;b<N~p
zv(1*<^7N-Z4cLd_zqR+DblIav)P3h+ztQLJdr1)~S0wMTPy6(yhZgzU(H{i-4g^|}
z@xdN#PigErcFxVmfA&rJ-^HLG7OLIQ=Z!e1SKr;=3W)<n^jZ5Qi%<VtXpNt1{%SlZ
z&A4PmSDX@07P{&)zbs@wL0+@`oR*JYb)1#%*mK0lSH`{*axSO+RC$ZG##iTOR+|27
z)#;SMZ%=<WBo6SKg_d3L(PN?d!=lLxF0|s*HGi0&F0{r6zTVeCTmFO3Iz&DNv1ibN
zL!X~~`^bPe*Kvk#p~^$^_sA`D%|G?G=38jm1NHB6@%{zKIt`-7LiAdw{UC26uQYr`
z?XS`*Uv1hI$7#!nTNg4f2Yt9v=U?wQDX@@yt4Ka@A@f1=vCF<n{p`|D|FMhtYN5tk
zR$Aw?#+!?_;(@%Pi2p(L3-bhjfMs^(|4xjKw`C7})*%;ad|5R9QU5ES_@RH;gB+;+
zQ5roK;;*Erez5dfd>y|+zvS=uFy0`3un>PJTKQS~p>`>){=v_xM`?{ir7b^ceCWKW
zHm!DBwEEXYTm6Htezj;F4`|Cy_*T3qpYa6IpB3p}TD1BPAAK&r&+T&TO^Q5USY*)R
z!}q!Tf8pUTpWEr19^ZsopZ64XUS}OAt!L$J<r61ak@&Pw?NXY&Xraq4>;4D6rC0fC
zSG8#!e`xFkdCq3})8!BJx{&os(N&Lf=C6g!1FKwrf0OpT|Fn4lyA|=b3+Eiaf2Zl6
z4GOD`u6`327HWT>$uEkWzba}TQkrwz=0deoX|*>kZTSzr&I6@g`xLd;qIrI<&sUXp
zwWoPpY5HZMwx_g}f0S?Ock$I;r8(cRQ0-M3KU?U^qb~bh<!X=8R==vv=f2uwya%lH
zsipavFUaE-y4F9vKU3P3$6fP7<5_9lUn%XHKQ8}kf0VA&Kj^dkUn$>ZpQ{}EnhRa~
z-?Zy*T6x#{=hEl$U!~)x^<DN<yIkj+((J!2wAMxVuKBO!x}Vqbw6y9|+La%y@7bxJ
zEZUWy@RNlaA4<FYq3!GbNNIh~QEC02heccWBQAS&U#PV9$D*|!wD#Ackq33%NlUx(
zkd|9{Gwr?(|G4~4xixNS`8v<4MeFzP7@w@)J)qx7k$6cv4q4@CPciL!ZT-&Q%3I2}
z&MT~OO&TwjU#$37e$wx`TkT}!Cyf^?KQb@1KhUoH;%c9HVWDf?v;4w(;A-DxzqQ`M
z$3GSh`Q4PmYj&K!)<T^xEluk>t+dPk`n*eNSAAE0aMgE>uN7C7@-4gc98GDfUQ#~u
z4&?n)&~=~d>Nk2V)bUkX<40-NeTvHu);xpn>Yu9|Jr=s^>v(6S)vwS_|LnI~ukB`z
zD7@e0nqOM((r@KUmtQP@!l!>0y7as90PR?)aY}oZJbc}MD9t*N6<zCr+U=sPc|EuI
zu5qIr`xqCHaj=m37P|ap=}pSF+D*!Lwc|Qx!A}+@wZoEg*=OYo_>8NC$Yn+Lhm13R
zaM?|{WmmQM*kvJhSXeDPEIpQ8N%@xEYVobOgzt)*O6{N>3$1a3Z`FhEvY&Xi(CVM{
zJzn^(b;l~V?5LEl@n)TO!ne*5D9?&KRnvVP<7J_1UrxC`7fVX(dvQu9)nm!&xu4Rm
zeWT?U-G5rNrQedb#^2&6{m!=4UQ)i*kEDFH|Ei7h5jo_9A%_fHy{1>qdNu9OAF{`g
znhy2<4(vRz%fRk6?FJ7S*rT3aqo#dL`(5^5^}HbiZ@h5ZX)CuUo;`XGcn^3Fcn^3F
zcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fcn^3Fv<VOF{?JKl-1O+U
zaBg9naFRFDd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Z
zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Z
zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Zd%%0Z
zd%%0Zd%%0Zd%%0Zd%)p=P1+9`_{f|wp~t5EF!Q0>v8QeLQJ5BQdE}3;4`2J^uu>ee
z!mOb$^%)m>^nR$11IM)X#ohzn1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP
z1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP1KtDP
z1KtDP1KtDP1KtDP1KtDP1KtDP1KtCz#sl}A@Z-YMk9jVr7H)TT?$OS>z8oaQi`F@3
zg#-F;5PI}}XwEp8c;LNnJlJV`=&@z}&}tH$H^h737wCat{8`)Ye!=+AV|D%TK<9s-
z`|i&V2W=Jqb=K@Z4qo_{psnMMKMvgC+2dabJ$gUX$H6Z!#rrOM4|orF4|orF4|orF
z4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|orF4|os!LOihV
zAC4L^;pla9mExjL{(16A|J@>&7KctcX}&HOZ;{K2bB>-m<j#A3otxXZ;1P@OaPoS;
z$$`I+@bTSjV;-1r&FC?U>@Xv6;cExI{pax5FN3tWOqWN``R4tpK}%s9bH6v&d%%0Z
zd!W^O;Dz~K{9^F1rE*|vHF$}Bcipngm1Dzd<9@?V@3r?{qeG9Sez<i`pU3~!{k_nm
z_k)jvd1oH9*8FbSwP5hvO?M8P7PJ%&KJ}uL@7cY5ZeB;r-=6P1;630y;630y;63om
z@jy%6=lya#>U-)v;630y;630y;630y;630y;630y;630y;630y;630y;630y;630y
z;630y;62b9J>b8aUES|ydw*8PpRIAOc$>Tjya&7oya&7oya&7oya&7oya&7oya&7o
zya&7oehD79YNLEFhnz6vkb$e$tXH#sP5bkQ>@lRKL;b%4I}hwKuzO9r!9xc2sHfMc
zX<yTRm;F~gZ^*zKFWh$8%I&M>qa9cI&%t+pFf&Ms^S^xZxsUBUCCG}?KUw^;m;N#*
zs1()S-of!V?)vZ-frU71hxR9I`tvt0gsla!%VVPlPTQ*H<=wBJ8G_y~={%tGs{hf0
zwteZbnPF?eAJ;kQtR2SAp97h%>3Zp<o7)5HJaXsW6PI}}Y%Qqydd%=n7fif%T<F5%
zx{TX!<eeXctqon{=UQL?aNM_Zei-&m05XqUc*E@d{@Ho<L4ga&BkFgT-PJ1B_1#5R
zt6i^W-t@pXD|Z=k_x|q%9_OwfT1#|Ybmt|@b-BM?4jenG_k)+8@nWF()>q^A-SD0N
z1&V*Z_4FO@{^w6YrN}<0T1bB^RD0LmvG<XWb^0!7E#oq;*x_2QYLChdd+DihL2E<e
zpZ!N`%a_<^A^x<==au-HSNd_$c4urk=Zs0AqTYAt{{5Noqdj+>JuS3wUhz-0`l0*%
z2VNR8{INGj1TJ*tv9#se4SFs7<4->ain_i+r`50W{mvM3#hcwa<P>jyeZ99o+w{i(
z)N-Y{A7cH4Cf_KMKhvV?{!PoRJO^Lz-z-|~S32oBK)Y#C+iNbmTKlY))}whKZCt6m
z6$kLOy|naAW0pL&_SOY+a|^L|zQ=BQ^u?9i<+9@3vY&R(n>cBUwZHG2%Zjf2?kd0Q
z_`UwwV}(w+ta$R&7oM2E?-IGB$hhtGQqR?{S!`T*`ha04{h|NF@UOET`S!>c#)g|;
zedcn5Hhn*wZ?7vy51UowQx0DI&%6GT`|a0(BKr#Tf)9@Cc-bQ(r-mSQWW^W8AG`Zs
zXU&((il^T3j}AX{`d^S0M_<=*m41&coXd(&bp2xV@3#LbNQ;)g(yj+8ultv@w65n#
zX{$e3`P*J_&jG(Va*-U!KF7kW>x}LblzwxOO<q3gq^`NN$i7E$(k46Y^1z0R<t)_m
zh&z72-$H*n{o6n>X`VZ@=Vg~%_I#IITGV}o_5-@_Ww-og{p)9hqc8u@=uso5g>T(3
ze3{9M&j=S;fA`6sUN<e=e8lW8It-W=9(nALA9}1kt#JRc<VHu{xb@<b3p9D^n89z|
z_d@RpA>|K!x8b1=k9{+=FzdLx>S^5Wc;fHJFR|yS5VW4BXuY)bU#?o@+f5&OICP<o
zkJ4$|t(H7-(Nf4htWspXvyganq2^;~jhD3cPV0BvDYGsb5rS6zYK^0wBeXPa^$))K
z(;7F6j~_wRr?l#Et*hT1^3P=+T5EE+(|fBQdRy=D;rJ_FxpK`NCWkZ6d3Mhg7M~Qp
ze&73Dw|#7Cc;D<_UC`~!Vjk4^x^#mBw!M1fl<>~CcDgvY^Sc0~-oHQm<ekF?%q-}8
z=fPQTT(Q=qu-i(%yM28Bso{RNez^Wek4`G&<8@{&yY^8#ObwIb);9-h{9MyLcj}MN
zZoAWnvBAKvhu(bN`qP39cY1%J-H*OM*yrJ&*4m=~gz(8@H+Z$jzdj7_d-UN4F5cta
zu+u*GT>0%G<HC<_n{nxRvnGV(^>wy;{;s#5ey*@z{rZF_x-L0#Lb%#PmtA|souk7`
zMqK`MhY24Ro>y<svG2#r^dBFtIp3D6y)tW3xZt*b{`hx0ObNGtX6AiguQfIN^YQl%
zA9&rQaPPg=A2#fm(c!we*E-&M)-B-=eNOGU&zV!hg+J}K-@le?mkajo(&yPRlLGLG
ziT}6bLOV<fU-;*+4UZr9b_hM`%sYp^e%;i<x^#Hw7f(5Nvya1ehduPp{MU^SU+Zwz
zmBZG5KHT=5Mep8o<fKA9>eC<g*YtaZt^T&@`eR=Whi-As=<{!VJ)F96_gAMpKdSJ&
z{fwo%4Lj-LcMI{s_%6QvB}1=$bV3N;(SE&`|L?IW;SKw~F@EB;F9dh|_P$yBJU2Yp
z@XEsuUh{=9g>`f4pU%JWnAy|ACx6(vXUAP%3;*}|UDrN$$wvhwzF&U!+r@6|KRMj(
zfl0^Cm^HZ&j|0zs`0Eo!PA-h!J@fB$_Hk=X3{O1x(kDC3o)XSJ<=>kgT>Dn|W!FzH
zKYje@LVrG8;+>J}9Q$E7rFQ+EBM*NuT;*4PUFMwL;|j?8N<B6^ebjyH-V=Vd*?Q-0
zw&}?5ktg0dYV30(!e^&!-eKQI$A!b+oP6beYtIOO-um~eeLr$~NSuR=%Y@wjb$kA;
zQH40XYpEmt^5(naLgMnqZLZ#I!95-io>}#i4Ub-Bf!w>ldhe9m22Tk(Ucc6Z2VDD1
zu<>c@J+kv@9|sTLyVCCOJ^y`xzFl`;>eIeIjSYJ*bN9Fpwz<E6<Uz&}zmiArOV2Lb
z9XfKYDdECre?MT)ZZCxE?e<Ead#7F*-2A{<ThBP=gTi_2{`>sy;BUu`4Ds7@KOZ)_
zZ}?Ino{7gRdJp}2yIDm)s$a>AiVrS%MCU73e>_xFIXw@ACJxxYk+=4^>A90G-C<gI
z{@yoj+Gni^g}7nAM?DL*eTzPN`yJMP>eP3F^Ox&&@Jm082%o>_seNbdFf&~8yKi54
zc-%W7`Hu5#^6v7h4mfbbkrNC1&#|Xp_VH<J&n(b8{^!-<L(eS675R{SLOF4WzSX<$
z-*ffrCKU1y<JqJ0ns=Z3*z}NcXyzyP%P+jU%tAjbHnEV8w|e^RuWx^BM)>M(3;*rj
z7d{GS9eCz}{U4hap0e9f&)&Pk^fDx_*=Jv~@;)6dx^8lz-oHLy?9JyEpI+FHGj316
z{qrt&UpFN@=!|Q(Ty?<o0<Qn@ck6e)etIGAuujo$`a^!e-yr%KXY$a>?N&bWlryJ=
zYyWNMVWal^sDP}0^pAcses^y3$S!N1_d<AU*FMjl-hWcz{nO1Rbe(zN;#0!aK5KW}
za_`LuAM0`Y(MyheDIhOiwecxGOzAx_MDDZWZr-=H|CBJ8yzcGyoiU-1M_3P8KRDOm
ze&x=Gul;uO@$ZL27F&MTv1iT<7ykOA&W9fJen|hYn{h;r_RienyubYX=;RP&{(|J)
zou6Oe$Q`bmSXjSU_sIjSd#pROM}G&M^zyNfEjGT8H^@t@x2(s+3+ol_;CI$l?4>^A
zs;K)QXxeXX<UGJa*S^fkGw`i*zO4Id^k&7Z?OOHln?6^XTeM|owfL_4e4bM&64#2@
zW!={?zZIzmy6j3?zE0mY7P<J8`ErWZIhpdY$A!dAQq*%a7p>=BNomfF$gkvS_Rp-Z
z?8jJ#$!n}*>}yzu&g=Et3$`CnJlA6VW8e4j2LJ1P^`2u3>)O^!{eAU|M?4g=?`9wI
z=vUipd+4n3A^U9B;f3Bh`{4cl^>LVVoadI@v5$@zK6$I9a@9iSjfIR;TGaC<S3J+H
z^36U!V&e(3I_4~_)UKrSGppPdf8FW0>p%G*$hw|otw%m3E}%6(E3NlEoZo`zQOs&*
zYpKWkiZ1+V#UT?;Jm&kLQcN4SuJxeOa-QR}->o)se+;JWe_DBcKHJu4?t`<UwO+XR
zoXb=jb^NqnnunlW`N8UstDaSEJvU>1R2wb(E9L9+t!mS*=iID27V^76_KzUvVIa>-
zIltkYNz2)%!Uxd{!bdO2xe#(7^4c!6qOOa?F?zBh`Ivr!oWr3PKD71&KJ^$MXwD;$
zLmzrT_{f3qL4E($qLaqE=8>c{_JP<3%{VIGDn}k<oVj0t&p9gNhoA8yaZUU&9?<Bc
z{<Nzm?lS1n@rCCW#4qiE*iAbw#4g%_W}fn#hVs^m#BWmMIT(J!PmDuWWLy<3zcF6e
z0b&p1Y$5WL(=IgSpru#qK_d^M7bH$9#ee?ikLM2k=_B)Z<>%xMJ^Pl$r-h)lNB^)3
zM2_=o<TTEavk-kC^PBNQA8}*(i+bq69t)9Ai<Z8mdDB|Y$an0cI7ilfBXWw&1N;PX
zJ_W+B7C!#m9~L-w+?Ro(o`)o*UH6Mw%d_s+taX&}0WCQfKdD_VJzAdi{*~XUHy2vt
zhabVDenF1=QT_b`>lS&6yrS_>Immnhu?zjoQ}pY6M2{l+kmo!Ly&&U>e)L%T1}&$5
zw1-~&#eEm|XPhgOpQuN^W&gx}le|p3oI~Iz(0ZPweDq*9h&|YS$BA8!->}>C09;~l
zx1P5Rex~q#&;EaZa?`UGpBiE>_20XB#5<duIi>LYo%dMiH+uSSw9wI8Uoxif{tx}c
z9t+v$;y2o5oUQjMkwY&u<F4m~$e~x!T4!4epZ6!yVr#KKs~!AKWNx8rJs_??@}wg3
zL6LgQ1InS9Z_wz4&-}OcDa;ECX&*hzLuf_xDBq&V)0ER5dUPH#f02W>5IyYokb{qY
z+J_G!??Uujm~<bTRgU-XJeGbxEvY}U<{#!mrAR!3#GfMN7OH$!+7<6}tDOCh3)$yc
z$i6KpvQJX9^t<@XbE_ZhM?vyFNWLQ9v!CQX8om`LtPjYcmpEoUVZVnS__Rko3)LRl
z136a!Imdyvkoq9@L1P~@NWZOe<d`?mmb~({ztAA^l_K#ouOjz>a|?BzE3NaEeS`~{
zw@HyWOpDgORQ<{QrJ{ANgkI~pn(A@U`hH%u>ACfOmA0F7A7agCjX!8t9O*bFUGI>y
z)-jDEOAbE$()%>#k>+pK6Y?+dk39Pi)+^*J<op3tzu`|s*L=nf`iVUH=?6$ZEu=pn
z{RY+V7OnEwr}AnC_JgdKE@VAZeM+l;t3?xEX_51~=0Yn^T71oaY5U{lEAznWhfH2>
z-3>ks+A7|A%DU^GGk9{)R&n^n|5|9NyV~bGdO!F$Xuf$+S(Z#*Tlb&Urum&-bD{3z
z+Y0Tv@2XaLwf1kSr^iyyEY)e5*=675&3aDj>$R@GgV9#UBYEa)l(h1l-iM{ttMZpF
zQ?tZ_JNC%6Rcx;JVp>Y?VN0*@)M}S4kn`9&{Gjv0qGw$C%Icq9^K{r&k>~nJk@E%4
zgZO<5&#igh#Q78FSe)bNIhLLaQ4i#Kk=8>W^*Oh5(a3@5h2|WJ=Pf)>0y!7MF51(2
z)Q69L_#ovLO+Aoyk%y)nK6>FR@>~u+dO-O2$I?5u*I~Y=WZ|!ly5PX$5Bf3yPr2l$
zuP%7`>d=K9Hh$~sRWANAP$cg8eK>g0eZ79loqTqnsOvQ6y`YvubAGDGxjHy##=n<)
zZQ#T}apxs3`{S+~d=|KHhd->dNY|am2W4pdoxcI&5BPLcx6cBQekkhw0sT_Gi`H|r
zpJqOG%JDak4?yw`sK4vwJGh|wQ)%o3neQNe2lalA_yBp{0HRmd6Uu3a{HFO6y~trd
zH1(jFzx<vCKJ96J#*y`aJWPG;K@a(#eu3yA4`CMwAA}E*e>5M_F8YzDpVR{>ryexl
zaRB+w4|$I7`B09Xlv_0IT1b6R<**OL|H#3ozm&sg{iU4uf+>e)-BLd7fQ&b^g^ar*
z_FyOVv4ij65nqfG_A~zIA<rX^KKQf)QV$wr9)ZjQMdms08B$)bGyiw~povS`g-^TK
zi5}$nZW;1CXT+a+zJcBFk#iyPw9kHye(L-B<W=O+uQY!LLix03A^o793uzzJJPJ+R
zR*H;=$7FvmvaR}weF*D0{b1e_N6?hR2Z>*3MSgF<yr-Nz4BcFaeOZxt=R)!dXw82Y
z|Al_t##}t&qaZ6L^`kW&j0?!R1Pu}wpo>pA<3K!7A4CqC^$t7Ws~j{)Uetb~kNp68
zsfYjIQy&_;;X^a7^bdX5OMTi0k;5MN{0<g6DYAdi_PC#CAI5$Ty~tCpH1(0Ukp60a
zR9^KX526ReAJ~nZApNGC_Mk!dpv9+Lk@3QQ_BY&L%(v`H^Y?uIzlHCzVn1;L(k{P?
z#(v~!hxVaC_LKMvK8StPS2@a&#~%EN9n1&nqmTBXu^)ZxbG034^y3feYkTmi2ht8S
z`%mlyu@6K)^%ZGP`K)8u4Km+ZAE-w;G)RA`hg~kjF6_tuAbb!$?Xusc9K;@e-=|0(
zomcVfGZ&iD@2#%6Y9Z^tBKv-j{g(^%Jv-JJka2}pw0`HT&)L*orCoNq%C$YUOUsp3
zdt9{IgP#=Dzgg*db$<AMc@KCGcn^3FWb=ToL)E6)Pqa4F?`x-}S*N+rt`ymyyZof1
zob>zKDp%<^>;9-(blP#?KBZcy`?}Ufb3X)XJa~HE@&M-^iq`L_J-_kqTzWrL*AJX8
zabBZH`Mfpzd~du5ya&7oya&7oya%El7_wuo`+pBwG6zP>+Tu0OM=M3$7yVM{)*1uV
zI;TKS+VhG^<y)&?y<Oe|-UHqP-UH3!0p54diaf7yp}wDyl;%5f7V3Kou72=478mk<
z6{zniaV`aNZqDx&c%O{-v3QS3QQueM{Ur8x$kV>k$WaeD+Qok4v|RbTzXehs+CqI!
z!ux$`QTsuEu>+(Z*pFS5^SdUH@ln+0Ym~$1_ZhT9xgz7JamC-`QCxBJe_YZ2#(#y1
ze?9%3UG_WT)ljkTmf!vG>S`YaAivjeq270>eM&Qrd7l#`&cQbaU$pJi1>Xt4V_)tw
z=7&BH1fc#tPy2&E`MV{wd&sh5ugVRc7(nxUk$x~Pij*t%>e{#O(rdpFfL-^#<JB3n
zUJXF_`u7mZ{BrfzTR-*a%fUB!q<!W)aZ6kfhak^kiBo9mp$|k4`gzY2r2hTejOt(W
z_)Q`64?gWE?zI0(n{B-ChXKfV(jI!iuQ%Um>kEdxUHE$p$bsY;+I^#D>7x%AGd{q7
z@R=WK5B2dAGzcFgZY<>eTSfl91nuK@X!KjOi*MP3pFrA$R>WV5`rQG|C(y)?BI5~a
zx!M6u{<M(%PQMf>S42O^`{W>eMb%@`^wSy__{f3UUZpg<Hu4#G*Ps^;KXcEeb0B$6
z-_t-oDL%M;x1-<dIDZa2_>OO8^jd3tXyLkt{yg&ZTc!t!<OPs%WgetO>bp?mGAsSo
ze9wP7U#?>=Epq;n6?NXO(|3(UE<R<xTmkdHBPHHk{-B(7jd=3tJx~n~5MON-d!F3q
zw_o<_kZY^B>{SQfb@~cjb8QtT9yF=%j+=fPv{g*HuC`XWo})%wzVnaohHVvD*V;N(
zBNuxe**&nu$6Fk|*7Zx|+A3~y`HH=6xw%KqW4s?+k-_&I+A8XOLtCX?<3u_4KW!ED
zxkFo{3-_aXf7(W9&P&=VG7mh?+aBOKLt91H@6G3S`Me!HzIWaO-UIa>xVQgm?;QAY
zw_Iz%w9ge<i~QW)_xWh{9`GLU9`GLU9%u;<@cw|lPo(b^@;i9mD+G~)1`FSd`_HSx
zhn_h-<b4R<mjHRMi1$rE<k6?^*C3}ze-!l`&3bQzcC{Qn-=|hYo_2We20JKMy!{VT
zuibdDX9Gpv6HJP{7sGpG>KEGOIUVnx%`1rCLHu7Bw~daxaqGn=hs-HGk956neaDGi
zkKeG{^Z?ZNA*)Thp7&Yrdn+G%75NTL+V8CJ{%NJ?@^@Ohsx6OyK;lN<1JO9z_JVs3
z_|1`v<iM^6{%_=0Z+6OodLFBPZ-n<lL5(YYe_zL&`E8-b74Ln5%pXwmoX#iYKrLro
z0eL?cBtL?@PY&{aHAp>B$K#R_mp|QM!iOQq`@x{T_p0@kCU1hQM<8~B)Ms5&B#ta3
zk8A#h*7(uC1J5`xo`;01JiE>YFBjgIXZ(2|mhsyB>NA%cwCVey`iD5eF6(`I;*{rd
zDyQ=uJFpl3f|M(&9*f2fkbWpyd4qXIee9|>T5*t;kDV@L99t`DK2<uY-Io8zGg*;+
zt0I0Uo{4|rh4z#tU%*$So}$Vrt#Q87+Y3zCe!yFSqRubXla|I0pzbg6OH#DvH+<Rw
zANp>?LmwXdW(YDKAaaW9-gw;I9Y>4_K=n7Y)`KSh@%MHWHD1UsNs<1d7urJhEf#8g
zD6QkZ-JsXPKmPP{plHoAi?8i4E}$ze7*~*ez9N6;jQM4u-dAeBSsyJ_dEx}r^0c(R
zw@AE!_zk4rAnQDM!vgo5+IQIO0PMZY-Qzyk=Kc^wFR1sQnty2*)O>;;zy)qvWuHq&
z-4cMr1MMklJZQYDJuVtM6!{$&{Q#};(tfi4qn?Etr%FGw>_2{WNblzZP}@<vHLs|>
zg>#q_e?NYSJx3LO2dDP&J3mF0gJyo~cqy&+ShUuIR(be}I_}u1sQz=&^piLRbzGFz
zIIWb%j+R30&+s=VU%B7fGeSk(pINkyPqk>qV{W6~heKQbfY0BVON*Kpm1aKz>Ucuy
zIIzC}lg2ImNsHJAQeW+YMh<lS-QA@7tGO-5c(_o<x!Sa@CvA<^b)mUvU6(4Qn=7wd
z=Ns^KeN0LdcbrRs?6cV~>-m@Nw|QP+p{~!Wk9IhZQ@m>9Q+}Azdt#`VcD(2x*j#Z{
ztvKL(-Zh?S`{7z=$<r48xY5);_jT!#vygSug_eHh6XzD{`ew<&$FD9wP|kXm6kX#*
zIrsHye_9$lU8v_q(A<ZU4-~5v&)lDJ?hk#^pvSIU_`5fX=iB?fF@EB;F9h6Ikw3^2
z<evdU-uvS1GbV=UgHK-II}6vWyibRVuA3YxQVwc4H0?u^KX@Lp|KFe7^sL3FhU7KY
zf4=X@`@@V6;|oo^Q%~>rXb=7L6V&s5_>7zD_d?iXA$BRw9Pzt8tKRxq;6h6;e5*eG
zv5@|O=yRcd_rTH*-_l39BI5}%-$9kPXz~MyK2XPlb%1hc3z2gn?StsU@2)&gzgr5E
z=IKi1Tz;b*KeaUK`vOX{4!Y3gC!J49Yrm~{*LGZai*oC`u&%gFTdwvgO}}(rLbLx+
zRDUtQnhS}`v}pO+;<N6xHe6@_p`+*AyF{+q$ajt`Bwn+krC0eHPf2NwOK9Sdc!9Ri
z%JaI8vYse1Uo1TqAN$z<f$-BJ^BGKAKdF4u@!|Vcx*ud;%6pvbOW7~NSEL-&`|cn5
zoZ54rGmHCL<eZrQ{*&hYd6m|60GjdF=jRrUJZQ}?_{8PhLhQ(joc~rE@xO&x+aoW5
zdY^33_@}kUzxu`UAN@aVtD2X0zkX&2T6q~het<>~baNr=fQ4z>r5~W`E#znU4xY+c
zH1=5KyazxY=G>L@5onO}T+T_5qdxhFa@G&}q3Z_pq%-dv`ucTKLw;8PAN{&6p%3I-
zh5kZozp<b7kMj+Xb&LM+_v_%JmvY+2PM-TSp4bB)nsXxb@jfc3cJqB~kaF6GR-~Ng
zHqh8hp1}_M1C6~}4_XnwFpeN{=vSmY#vP;{h}{;dUh08XInNEX9Dh@u7HQXo<h{9t
zt$jUdE%s^Lkryl^{w*ZFh<Eg8oN2pS{?-k{mzliyjKc4#$Rp^dKKWQ_=4~0ux`TZn
z^M&^5Cwh>x5PK}dFCh9Y#2yRD%dG|3w=@@8`!0)LZNF%ovOloKk2oQI6e$N8XX1l6
zLch|?2aQ+y%{~W2PV2ER&~n;^kNxn`L;bO*U-t27YtIZp<gN7rJ?KXteqdgMtluDV
z7HYfLlN8xcaE`=z9_KaWapFSvD`{!;TFAJjMaD<{QEeJOfPXmd+c`fB`zFBOj4$@U
z=l4vEEB9OKU-W`4y}uZ;*z&uMJ#%Krxh;R^g7~99?7yLzx6oPruJJ^>@bRBw(tP~W
zibE!xc+B^Kbq>pOfvomtt%sj2{P(1DgHNBE6IhtE|CU^}<BYggz1qGt&y&8-XVr(V
z>n42W7w2t?*va1?0ac$xYrm78(_vp$#BLBh+CQc72S|NI@^;ex(Y_+{*+P}iN~hIN
z+*S+cT)ELcUmVgt*Ic;AJ<Ha1UVGtObK&(v{_^gb?H0*37rwZ`{Ht%@cd=Y^;m>ov
z9rf{%opa5FXKa4_@0MR@sa$j6dwq{Pf4QX=%{3QZa!lvr@5*(`S*Y_Lntdav>oI&#
z=OOpIAn$vDD$n^Fh#biN8DxI~!Ux$`fanFaee{5wM}pknfb<KbT~PN4{N5X6e+cUL
zHFck$_Ue0K*a6}<kbZ-z&qZ^e1G0W&KS-Ro^l3S9XQ8eS>ff|9zstv7kaJm(_y&;!
zIVT0RU+4jKUDoID(5%DkKMq@Zg{M}#Y=Io*&mVEtoBw{XLr$NsLn8;e_9xV{5IvQm
zrN7#I?r$t~^#lKBMIC3QtF`~|dfJ8u^gR%bhqSc*9!>Kj`3cl=@)bya1K}q{&3mka
zpq677h+I<SJOX6h2I&XLJ^<8y!MD)0zOb%Z$a$!R+?SDWG#*$#K+c;$wNKko|0}Ki
zcKL_(9@PDa()7<l*ZM@c)@Sx>f7z$7UsL2959C~nzfa40#P^l(4}Rf!y6S}{E)}sC
zwDvpjIhRn>{US7e0yU1H=@0Rg6g6(DP3!n_UfEKp^GIpNzuHJVREoq6NPe}@HGaBo
zpjT1#lgBJn|6SFiUHgW<4>@7TAp=*hS-)n3n)c@p*<(mehx&g9b{^PeVE3AKgNF?4
zQBSW?)4rztF8i-~-jIQPFWh$jRok^|pX$F-@lq<H$Si1IEGb?ksu&S+UdF1fuN7Ai
z$8)w2D<GO0Na4lSYl>NE{ibwhEi-L03$nHpSI=rBEx%N^Rl8IwCFNG?mPpxTqj*zl
zn<{fQR#9cAb`hwsu?nk;F-?+XO_ed~s>kv2TAL_J*KScdqB3$>EYEOd>#<4_@GX%H
z_Qh36s~&k!b;j^xSoO6Gwkot}v$|PeX|ipJsES<H>k+b1bu6=?$S!HD+0?(O>lzgd
zw$T@p9p80ZJhp45EwR$EOjCtXV6bC#w=~nT&Ei{QP(8b?mMwnuT8`6HL94M<qs+*4
z<g(rhA@hdCa}8lji76h}d(X?eE6(+CyjJb7qQS+hQiCmisZ`5S@ixVC>ntlRD5YA~
z@OieiW^k&dGA-ZoxuwUF&%n>zMCOVajLcVwZ!(4(!zypJX`#l_hJW*d^$@u%5Q#PC
zD<_txwd57bB}+J_0VPsR5iSj`@}_jRNLi~lutmI3#YVC^)UIcjl$mXa8f<euS@Kqe
zwDM(kH!@5$hI}e>QY)`(G<BWS6(T)LUDvKqEU9~R5ox02GFlMxkRn<ft(TTBUaYHL
zZ&SpsqQc@;w809lvntG5FJfvHR7AA-OYZ^ifv5)(&aYUpSY}lDT)Va`v-k-Obyatj
zCu}BxSe06;l9~htH?xS<5sTl<=CpeCQNKj5TfO0l46N@_>RzR;kjNA-)HXbje<^9h
znvMKRN#;z5D~RJ|8*<UqKngFl;!5kMZZJUw(<ZYZfzrexRwHTor43rOOQlj$&7w4k
z4NW!NRGCwEMU|b}MWDjQDy%NXG)a~<RmP~R9>>dTZK5b$yG7}U%E)E0Jj0c($0|v{
zw?s177gr^%dgMXX8N-iZ)z>oEs?eg%>SlqZ$+jh;Dsow`N61FivCM`dyQHyZQ~#!}
zYg91UMqf;JeAjL9*shtj#7f69O%+Cg!H(74(oD-Xi*JoV_3XA<w)oX+IZjgrt;SZ3
zGNad}ZO50ybIlP+DPC%Ur}7>!yjWLim$Xvn6%8(4m6~etOQl+tinl4ATW48mK`GU;
zhR?IDHG@+vm1+5w&n-Qcd<K5zCNfvdU}U~ZOE%fqR>QyfX6qqxSs=>$lb6y0gKaBH
zT|rwE+SDg<HtEvfDz`PaP)Vydu!a0kwPvz9)UIcjlo@S^8f<27QZ?qpwDg+|l(m?R
zVYC~|j%7NUO<gB-1&0AGR@`O{Z_-0qdzH0b#F#3mh-mYd-UHqP89Wd>4P%)Zlvlrs
zWtqiKXo0J`t2|*d3B;<@QkB#sFnBJB%<~<rf?wxNAEoY9>IzkGy;sp%Jdl4WX``Br
z{Yy#aOo%Ip<7FFE(bPZ+FSX)I>!)roK?Tz$vmk-e#3EKBY5AoMTD421Qc}&LG>HvO
zHQZF0Q+Gv`o!Uj9!p17BF2*!TmNiwzsH+~w%WG|-C|$cn>4?h6WwAWNm957rNx-*6
zGT0YaC9QhoLDd<<k73o<GT5rnqRr}Nfu+f|C88>FS+7UPM%A&*h9bM9v1U{MrmkyL
zFxW<4Om=+NZSmNynYP4A$1+V7MuEYO)!oue%QlN|jY0M7wpzCM)oVFUQw6QYR*f>F
z*QIU8m&9|;5lJatYJsQn9x%LES8A8EQs)&7E?$+IYVk{@T9%5pDV|$rS!qEj)v|`q
zv#m9QQ!SNg`IgTuJ(he1e&!}JSIl5!zDi3r+1OUYzxihCA#zzD%KMX-(gK5RD@t8K
zTNK*VCvrCF(%>q$HMdYnt2Z!{AKGBAY?JMhGT#`Y2AkQNRE;??E&XN#Wi4i7812Tg
zW0{U-Q`bpd!C^p)6}MT#oAglDUS+KpF{TPCBHH|=_kj06)B_3USFBhpGpc;9U0aq}
z{1`*yMB{kTs;=^=Y-QI~YN<+UD(g#g)}H4(SOw?Jn?6e2tJD>$;Cip3wRj-^QqsmX
zoA{TK%$X2Z5XZ|lsG_NX6kck@mDW$)V1f#!O=dv?rHMtXM$+<28?<VdN~NTlMQIWn
znrgVIGN<l}Dm%4{K!uG}SY3>1k}PYgj8RuTj+fWkL{YkSi_#I5k;`IvhAUf-Rg!>j
ziDa-Zu1Z?<$b+ggh9ASKuVt`Rp+%e3%>ql4ZA(N|<g#9mkd3NinGHpDNn_2X{!Lxi
zs9>;-zL@OzuG`|VT{CTom5ybaDvSby9jm*gnU-x9-x`DJ*=@CK@vGNzoTds|jjbAG
zMz2fTjxUMlnj?}%99gi`oL@3!d&DkjrT!`!T)ZkZ)#8^*wJa5HQ#`lMveJT5s$~tI
zXIpCqr&=o0@-3fRdMx=2{LD>cu9(5de3h1Lvazj(fAh`OL*%kRl=mkur3D7tR+PGe
zwkWizPvmUUrNLEhYi^;ER&QV?Kg`8mWo_E7XP1=u#t=2w%-*DG%!z5~HybExF&o2Z
zH<lgCbTpf~PU;E{16r)O%^KdMhqCr6YrTjuRZtPp<}bYmya%ElNI1V@#bTLJ<#X-Y
zvdrSg7#b%U$BR~Vl}BYOyRK48RZ>$~Uz)S_Jm0}8IB(wcQR-f$u22Qndljw41NoPd
zHmTXPIbTX@Exq-2#Cag*k8j6`dEQ*|02vx=aWfhkF1GIDCHn{zyvPMuP+7*H#4JD{
zBbMl`BT8sXXCk45x~w)b@DrMqT63J<A}=x_3e3VqGJmGF$XA#<9w?4$X>=F(g^Rgk
z-!Gxnrd-@q+1&R{yT)wLV5DVf9p5TS*nxa2O-M%@sK0Kc3;)VRsiMZhdS3CWC{f2N
zW0x`V;ugQC&0jU)fre!#@9%n7Hn_8K?DLmRRQ5&Q1F;@x2pvl5L#(*FDJepvVr(9*
zDa>j#K>jUQ6-lNAUcK@nQM@YUs|iK9;+4uo=tM#h`SQ!kHI(a|_c(bkHA4NrxUI;E
z>oihvWknsu(t>>9qFCmyya(p42O`^sIzy@-qSe<pcm0*R^?C<KmXg%<eFg7<R^)*u
z=5u|KS!@=xqQ3nSOeq<M62k0#GAXmVFNpC#qfd*TEndY43oDJRRBb*oZ&b_LmZ7!!
zDy7vLr~<2SAyUofIj_0e9;hF~;^9TLwKucAVgNRCL(Q$eVHB)JnxU5sDw{#C)diE^
zd^Xetb@M4V`;#wdftK?xC2d->S=N`5EF+`*=(R2e5hs+2MG7Kg5GxbQtgmktM3r-G
z3da@1@lX_xRt@1dkix4|Sk7r^Mc9yFye08mvyN4el5dIEcdfoG4+Z1O<;yt*;l>LQ
zh?3FkXsxvTXx;cxu{2skSgNs(vJ@2g#j7U7i+ZBs(Q9ovX<5A-|ErEzx+%qZU9t5F
zOoxJkT>V9fY>_Bll}MM0M~bQ?AJYJ;!9uG=M%-0FjEkZWoMc#9uY{^6@{If?#b6Un
zG>FvwQz}|YMXOsS3Hb3Mb=-!6;y=;m)UeXB63SB0DAZ6|E?>tfFR^M^qLzkybB<z5
zt7%!2`XYsq%V-->w$*)+$#<k4tP+bKscK!?RgID;J9=#yS0WR!!r&Y1NX5uyD&a`I
z$fc>mC@|P{8tsxub7hxSt+f1l#mVYKq#~CQ!{WJ(+M6nisNm8NS4#_Gg$oqaVFNF&
zpiw48iHa&sV=<mYv;-mQ4T)SwE~TWN%>P#qDqJ+^h)_s2V1ysJjJ8@L9z#T{MT#5Q
zF@*(I;i9~?Vo^k6i7yiI>x2wVW<gr{Mn#Ra8vYdpkcuVt$6BM>(9lp5q7Jz%;cL&M
zLQP*cme&8<aFrJ>a*GY-iyGRk|3_&G67%zdMfu`YQ~K?)GKKLb6`CnrAO{)+aqKuA
zYD#@t&ux@$_&0LdSWw4k<VB?#c`^KGb*}4?>q2Xd43l3kSMpz8zJwPctYV9wui8{G
zWgx97-Gcl=({&eh2T=tQ6XJVSE>_NItm)*}3&aS;u*4)tOVZ-UHkZZ-Q7ojG)mYl_
zZ(gt-B9{e%ibc#z6&a0J2HRGYsDdff&}d0%38B`N(3WI`U%V_{;U9EGN??I8MKP?r
zv{_;n(2Sy1P#XCnzj#Gi0pp9}x#}8+l5{~)Y_jpHNYhLyr$`mAkZiyxyYV{8iHcf$
zv%k^Gznp8+O?58NGNlV%G71cKp_;#FdmhL%rdB13-;`U7ItDu$!-#;ilBbnTaa2Wt
zS9cLFU(c=kzo9UH-9SXI^Hrm)f_$`vyj%$@F;Swifrx3g%(O<q#y)6Mg|a3KOsPUG
zL&9JesueCQL4$9wP3Kzj#R}yT<(x)qOZibMdfgyf#>n^A;Fq!Uj55Bc(egT4SxuUw
z;<1(MGzeeFMx$bx)U^uBR5$9PT7gP_QEaJjRk#o>G5i=-L!}76a9MXzxULr}A!6tT
zMfN{h-IgpXsi109S42(ZGAdD(>%4BW39F!-U(RVXut|PnVf5eP&_%g<C3&b<9Jy@T
z5ePP7DOFaJVlA-vwi!l^O4&yJG!CI}mZ~DZctxE?tdMV2H=Su!u=o*cO%bcW;@9ge
zEi0uWZE|S|l=I6u=+H(R`Q`F?PI*C5k#&`}PKkU8QQv|xl+IgWu*+?1*FOJJ(q=Wk
z&iYbPxz=({sX?VwMA|HfbUaojmKjl;?9$W{R}jZb){IJtrUp`YTE$6MTHo0~1_9G1
zvmiqwW)-WEwEWTrtlFhgDXC^rn#6{t8g8o0sk@@ePVFL4VPh3m7h{?v%bF@<)K!n;
z<+V0Zl&;;PbVOz3vRIzs%GP6*B;Z>j8SIOzl2$$Ppz4g_$FS;a8EjQ((PnkCz|v&f
z5>XYotk)xCqv}{@Ly=w5ShJ~rQ`a>r7;K|2COf|Cws>sUOj}~5W0|H3qrhOt>TYSK
zWt+vf#-Mt3TP<7s>a`rFse)Ext45j8>(aL4OX9iah@=!RwZKz(4;WsoE452nsq=~k
z7q3c9wfLn{Elb7Q6wj@*thAt%YFWeQ+18rDsg}yLe9PyS9!ow0KXVhAD`qe<U!^6R
zY;3FH-+Z(65V<T6<^9P^X@SAE6{W7AEedVw6FHl7X>gU>np>!()f?DCeyCbASsiNE
zvrEd1Hbf0Jvp1<4b7ETh%?8R^%*HU<jb+C&9nGe$le&V#fEFumvxYb6p{%{iS}$Tu
z6;wpD`AhEs?}4ZX63(w!u~=qQ`CPlUEVK9t4Ruv_l_zW_fmoGVs*;)n2G1ptdA@^H
z@aw$kqtv}hU7^JFmDMg$<%zlAf%=z{uIgI<E4;e@2CiQBZqjd>^X;U$&_my(_kj0+
z_dse7_@yni%X|gzftL1wpQbJCao^rA$OC?QwSmF*gVY8F$u!ZAp6QxaRbY)%T7IkQ
zRO>aya%=0g*jwg3;62dN9`Jdhr9Ix(+Vc_8)<(?R>pkE-;630y;63om^nl;9{4&Sg
z_d4Q%{3nxsQ?t2$DJkL>f9XBoJ>WgyJ>Wg?3-mz4);O|1j$GDxqm=pux~I9i>>b-&
zp7naY2fPQo2fPR75fAvKavm8--wp2p?}1;O2P)6smWH%^>vq=CQ>CFY)f)u{yGqvl
zg4D!0GLBbpNt1;IDSzQTP<0RZ&@){+FS=%4?2|cwzCb(>8U4tm2zk<bAhQRou(J4>
z8}b!PJ>VT&>JJB%e<^A6nk}-vlw>(P%8y<*@ldl=ENXeNq<9r6k7d@En+1`|g-cU7
zt{{$A&=h-N)ewFIDZDy;<(!5_gbfMC8e%ZbI#xkSz9nAYwfeF=6s#+kFXt458!tp4
zN=C1vwbJsVb>m0H(r67~sm40WQc&a<ubL1q>WPX+ueIfbW$o(a_+NFz(oHGO>x!*c
zU^)~O<mxX<WQ#=cszka}JW^CG`IrV!4HjB0GUBcZVq6r3;3UJ+dL>jnk!R#DDF&Np
zqCuqYpHk6MDq7ttNx+X6spB>j6#t1fr-qf5l~9&~Mxln%a``$=d5KlS616nsn{yOf
zT20HE)E6m?Tt?f7vaRlmOui%aV3k<>NLB07u4<G-+0kpuxDuI&6$am6M=C}xQwc}v
zMJ`PhMuEYu(`c7Onk&1sYNh4ZD^6A?A{Due7#7cM)ZSECL<N_IxLR5eD_o$U4jXuJ
z1&uN(N>o&78jJBHq9q7XZ%E`iaw#SCWd6T`P~oCMM}$JM0VDj#Wwg~2@fadnEmGXb
zjwvj#3K!+A6^kMoOMH=tUngW}G7HkmH!5nZ)$p$<fK)88Kh_%6hK7ck5Ov6931532
z6>9prv9$i*hO4}Akt+>0wA%PjUNpi<$uDWEzHX_;QmV9>__8vER4NrOP=yQRK%*dz
z9mhjWsZZ;<jnWPOMlMSWN~9Wzd?V2+ao6>{GJYe&<kwqO@?T!Qgcl*KVvC=z+Eg)R
zAgw9gg8V|$br<!YwWeu7;R{(OT+V6K;N;f}#0bT(#3V>d(&EQ9m&OQDETov#SlaM!
zUa%e_mj!}~Ma)YT8I4y4+g6mQf+^I{Xh~@aq1KhqmSluqyewYfA9O`ZV1Y43F|53_
zSz;E@jG|Uh8u=o>ctu$O<BQ_C>KccVbU{&Uvhk`&(@ZI+NENS;Y``eH@jA+hiduZL
zzxm2$Jy*p9ja$0Y@-0&=J(hf($#GTSf%Hsi__0b1-Ii=YSKNiwu*J{R2o1+8jpa5n
zV~ZNJ6-(+?Li8&}3!>L~xfqt{E>&pB%TU*e)!P)Yt5~skRjRt4Dk+N+(QB<{m05i1
z$CbzN)Gq3m@~H_Tm8X?WacP$eT!US&rSAU*W%=s{B6^*#8f6vai~1WCmC$L35{(T+
zOtWRCH44TzT_{sCN(xE{m299534>j(!V)z220QPK<m!tR$~BgA8uga)qg3>|DX$b;
zEw$9DU&hZ@ga7~Ry?e}V?Q<6NYnv(oj3ND_k;tKlzZjwrQB<VuE(Hp7Lr<IvlmaEi
zg3&}FBqeDSON>ST0L?E1Le$id5ETOb1Be>Gl2mK3c&G_RiLnMbG!>-Cp`udnInUhp
zT<cluIlb?@_j>mCu9>;!nrr5swcckRzUSMYSyLPF3~i1x)7#>qFyT`u-CNLVQF4(M
zEi3bAzZ0F@b7xX%@CM?dgu6Q&_lZ=3CT1agnaeSA$=pzywUSJ7GsFEh--4rL!Zn;w
zm!9dKSPND}$eXw%^ED>SsEVi|;pF;Itz$&7HChTRJtq*3#{d<#!k~W4i!<|KG)4u1
zhgT(T#8Ejs0xCdwnif(}0^y`tb2Umm%|(P2UlG%on+m@26<HG@guJ6Qg<k^dFavX}
zQvi6K@l&6rA4>XJ@4xW;P!e2Qm*}u{!5Kjn6VBm;4Ve{T?FqKfZ0>)_hczg15Jz-6
z8s*gZm@9upoGi{n$YJ(#)lodx_ICv-qFY!MM}!0d4{w7WI+>pRx%1B4`JtqFaTLSl
ztqDWSx?BLA=QMMIWXL{S>q<m{0_s}Y>_!Wwj7z1~D3_zm4!oxfloN5u?dHL1u0%*c
zNSs~y16ebu0#T8Iw!4r<s&KMRy_8ID?vY`1{E4FnA%ijqPo>Dkr67~x?4#idCrtrx
z0H?y_9wMi@<U$FM0Ps-6nbdA^QPokLg!k1XR&vu|Ju2#3xg=pOpwE*!%^edMma#4&
zoCBGeI>*+m*`!g6wSWdmesZS_h5B-mJFJl+W3o+na&y;0Z-QKw(gHV0hum4aMJg$l
zUF^aK2`ItDQN9RC6<;R-3ujyAZ>O2nOo1yc7}nf^&nM5MZC|o<Wlzb0P7S%MT$GxS
zgLvshwN<f3wN0N}1g)sjJ#xo*W<jEZJIrM2P6-t@Uz1xCg(ZPD$t9gn*n$fHahj^U
zlwYtEpftj@bSrH+v({~npNM<w;f#Of>7k9#pK7J8h*Vz0iO`T5Xd1~v#6)=8?#K|6
z@DxsCNtUllB$HeZtT3#y0uT?tSZG%PByrP{Ex|lpi4im5;f81XF|Cv_$ZV5pAZ5c0
zOjw)Q&INO9{J_JroZCLtqw{Jmh+OedxZ+bCa9TXSP6zAYqMBm?5dcGF#76~3`D5~~
z#4xjVz&SR&6#?3a>sp<I8YBFwCf2wYZBYT+6stuhefY=vm<>{UH*(ra$R5R~icu~f
zNpc-`p!Cu<N+$I@=vA|dK-M?O4%^qlD$P!cGi+(Jj-F5G_~NAc9AsY=Gvs{lN)Dq1
z*&h@|5sE8$PO!i%_uy>p)<8lii4s%+jQ|6{s{|qf9Kdk7h#v(jV?~<HguqBH6$@~1
zTUvifX3mdzhPZJHHW|d5@F|qDwctoGj*?(rD--2I*XZP)JCjg@HxL&k+}+{0Poxqw
zF$>|#T#lJb=7!R&m1L5e8Sc0F791rLuHl5b^i21}TCgHQ-oz!DuQ6dpRYVO5C)bB+
z9V3dZ(NbXPIe~CI2B<g&&tq&I8z`?5hM<DL!?PN(H{z%qE)x|XJWUHJD1mU&thpMc
zp5`LLim!-i%uNMf`HHNG5JKM3n!+yub(n!U)+qqI&gdUX`oi~LbnS<dR@D<ZH*jv?
z+`zeka|7oF&JCO!I5%)^;M~BufpY`r2F?wf8#p&`Zs6R&xq)*7=LXIVoEtbdaBkq-
zz`22Q1Lp?L4V)V|H*jv?Bi?}iO-Wz${)^8KCGC3*^b=pZKlrsz^<<qJunl~4kJd+b
zV7t?=!us~zc@<-A5Al{}-9E{gJvVS};M~Bu0k8pkt2Ya1A0d2h<RiT4=aKKXft$X8
zP%*;qH}<J?*A4Iu)4TTSZk`PEQe#Em@bo{8^uM<&W$t=gPBF9r+4x=e{s2id{he@h
z{G<5tU;5r@zEkGhz{_j^eFwbEG#^sebF&}P-FyXSsSi|*Ucq?h2(M%V=acbDZp1mp
zxq**+15bD>(hntl@%wk@hmt<-hwVK5YuUgPzU^PjQO`MEegpovx%2WfofDiJI5%)^
z;M~Bu0sCpm?75K-e*=22#pLUUfAsT&&)dLFZz(qwKktlZr%M~)8>x5mys2edv5xT0
z4<(fqy+9!MBHfV}nBi=BZs6R&xq)*7=LTMI1N@WV?w=1|-A|cypc=j4R?pVw2F?wf
z8#p&`ZeZSk{v`FtpR4FYNuT+=@qa|7*EfB7d^YJzuKjcp8Z+bBU;l4(e*Q{xcsP*>
zFXda1@-UH#-G$^X(P6ICI;dj8Ih?Sew<6qRf-N-L4gx?}gAxaEM5m)sPK{5u@>j&k
z;%pF5>L^}2cGVrmRWMOC(XCE+Q7S=}@`O7hoBd2Ce`k*WV@cv}@r(8a;)V2)PBS;8
z!z-h~L)?qsG%#tiK2QLi=QMMIWXL{S>q<m{0_s}Y>_!Wwj7z1~D3_zm4!oxfloN5u
z?dHL1u0%*cNSs~y16ebu0#T8Iw!4r<s&KMRy_8ID?vY`1{E4FnA%ijqPo>Dkr67~x
z?4#idCrtrx0H?y_9wMi@<U$FM0Ps-6nbdA^QPokLg!k1XR&vu|Ju2#3xg=pOpwE*!
z%^edMma#4&oCBGeI>*+m*`!g6wSWdmesZS_h5B-mJFJl+W3o+na&y;0Z-QKw(gHV0
zhum4aMJg$lUF^aK2`ItDQNApfs_6G{+K@#?Z2gDBkj2AJFH0wwp@b*5$hG*;L3l}I
zlK63DPYHpBu%>+EIkv=ZT#3OCkF;uJ>SFtxqHD4u9=TJ+Ve5|HVIjxaL=iWkmP-k4
zKb|d3kx8xvxxoCO8nRI<lo-`Kglit95q_bJ(pG9#OB}b@?96)r?k$F+-{{sQTc-&)
zg#%m>#GG&{Y>FF53YRNlBD`&PWQa+43T-}x#qc#D$wlMz6Op<WTcfZjf$*XPNvvP8
zC5ZNohnNWuH@q^~kWHDnNt0?It%4btnA=%A7tHa>LyW=fQ#}gENaPqEEyas?Jr5y~
z56_rvGc0~aG8N&eN2a?E>n;_#!f=2HfDs+yD1r15A53a8X6<BiY&eo{`7P*5y@P5H
zjt1)0C$txBQ32Z&i$=zOO)@nbVUVh?%$&}}NAXE~l+#RUt`H0_Hx$3*a*+MG4CN4R
z^<BQh1|Pr~H;UP*%M4pOOE|XCF>}Hvb^B693(c;LD0A(LYfV6GT*>3eoNT!xXKU{U
z5<*FopbBUN7yw=+5E0-2UbnqHUV6k<#T5~;isvZ%%w2%(aTdA)3qHxr`4P|1<~TFG
zEglLJK7}$=3tBBoF4Cf9WghK!qLX{>OeziDKwOk?cZcIXkxJ0SEQBv}Ic6@I8%ncQ
zl1Xl6xZmbmaFk5Ah7;=2Gu;zw!HNiZ6PIMZ#)KJF5j7;7Tpy}+j3~B7OM#{51j6wc
zpyE~-)Q@>_W<HF@s37q0s>F>rDu+iv1qe^mLJCSCoHT2$MyaQ{h_K=-Vj6Q(!B@T_
zYa)b@ceJMPOF$iFV2*VP0IxHC>a+B(O!|`dKl}Vp5?ouC=&*Ic89@~j&f$a&nH6E}
z3AWH|?tjULH7IcqM|3(G<<$6?D}P0tEY3v8VfJ&?ZFpDRVcZcW@;t*wyuicTpodPT
zXMgG&1#_Zyss9$N;V#ft)2h9c3UEQ7R26On;T1MK^u72^1CuuE0|n4|PBSM+hU~Mo
zu0#|lpsuCOZnR*^xKwJ5ayiQEz<bI-IT4rKZXT@WN`wT2#MzZUkTruU5EUtCy9;Tg
z3MbptOUdNs9vMc*pE!CDGAM)aREk_&3NjhaJ{qoY(i8v(a4JmhA#$oqE|dTX01ri+
zN$nOFRUO4icwbFoB{vP$qoTf*OA_V+`aG%A+%bV+8S4_lIgpvDb8O9;O&YaW3uut!
zCwIzFs4pkE!x||vCfkH3H+L=cCdg$eEpU@`$ep!Yq>^IU#V&l1fD%j`<%^J1@pTfg
zaJFUscA8nu6u8oYVa+Z0eDX}%_9aVK_LLmx)R4Q%MX3onh?ibeTNP_m+w{3b(26SE
zBX^8v79={j!%U{`lu%*wHMuoWSQ2QHT+;c3Ew}&>r>V+I`2|Y>N+VoLx6+m~Yu)Df
ziMY2O&iGfJ9@+@~saD#GNaaPG2o0%$rjaZ}OoX@X<{_lGmzdBgoW@E*N+gq953De(
zvH}nfz*uNk0VHwLk}bhJU5ODh;o*j7`!TJQG01F_Y9M8M#+^A~vX|^$f#8L=X!ofe
zomb;xfPRP{8sbH4Z7LE`c*bO#VevDPsR)PjYcZ_5RJfx#(msl#QIHPekCD3)!_3;r
z=GgF71ZX3!YjqB4jPR?PSmR!_MFnhAtQMK{;UDW`Hc0K=$Z0Dfdla84M!9?>$#vX;
z(o5SYnbh;3SIsH{S>Gf(Y+nnjG&?EIu%*#DdOo4!i<9bekbPCmkn_DOIgA!$e^3-f
zD6Zr=!2+|~gR`|;0|}udN>Bwf0t^7J5{L+J0K?@XeiW>X6=^mT0wcLpEWp8SY5gUc
zIX~hV;>Io5WDsw{r%=w;f+NK^N`iT<Oq35@qmz5?OhOIbKwOk?cZcIXkxJ0SEVR-T
zUdjzeiPSE+$NPGp+^8xB#k@A(g6lv^L+m_;wPU*$tcVnB;*!kQxF;D^5j7;7T<L?Z
zMig74rN+{80^xWJP;n~^>c_k|Gap7{R1kQ0RpLe*mBS;T0)(e&Aq6E6PMS4Wqtw$}
zL|E|^F^##Y;45E|H4#F{J6cotC7=#7FvmItfY%xQLrFjT{g+<*p`=yyM9vMI8#p&`
zZs6R&xq)*7=LXIVoEtbdaBkq-z`22Q1Lp?L4V)V|H*jv?+`zeka|7oF&JCO!I5%);
z1H1nn=ONXTd~V>}z`22Q1Lp?L4ZPF_^lwV~()T~-*oTt%|NE9M<MUGXLq0l1Ir+cC
z$zK6)Tso*HxWqDUt;_*+5MFo}NrNDxav4oCy=Z>p2ve_&@~ON=nG#bWf60YTdc`YE
z<82Z~!%G+aW!V@=UW_k;6Im+s*z<5%d($6<*4(6MkRfhO&C=X5Boy?cxdW+FLJrTz
z;tY4Ufn-41hi+pkMj?ynh0nx-be=eqNJ=*A5NCbGfNKsQv`F4&&9C}4Lz+(8u|zFz
z6q&%sVI^IbElpKNsIRIQEm&bBD~%&wT&?RPQ&Y=lwZ6%6aa}}?vY+}4ffRknuC7WR
zdOr3zy6umifQXVu1bN26Zj`IlOj-(5gjenB=GH)Tel}QuI{A>=h}BqWy37jJ$lDW;
zYbfZx^n7BtLd#GG>r`xMTQWNW-rK_4Es2RZvve!KwRJ`0K*0U_Eo7C=vAA3qblv1O
zq6@zOmuEZD(%Mfg^3%G{r>5y;x|{aGQIgzIw`*(N%27L_VcOKf>n#cSyGcIEhNP4*
z!3OsSY4-{Eo|OBNySpwN<1)jZZaSa8id!`z<#;J|je9t>!@IrsU0KygmNt1<{CI_R
zJW|P{(s=ERJUl&)bjvQvP6R8~YVI-G^qmf`XJ^4Q)6S}swy}Y|E8WYpR`dk!sM$d&
z-*LP6Hp6SNa1nAElt%bV9!uUmHm}ot=Lcavih!36qBoo0uOOGD=v|42nk=lkDZ#CV
zLU|ZVuqjUqvLBno*OR1iw_ddNOB|~wBC+?qGPU^o${);{&e0rdIw+l;kMD*gn!0K5
z%Z#aD3bp=fnX3z@IW|LH&L+c?ytQe!o+PKuwV;<-c38e8G!DI-@T#57LGONsA*gGj
z{gfNWYm-$GMM9qLep66WdsRP2(>t`zWOcJml2eK@{2<UN_=h+XZC8l9>?+5kox(;l
z>w<iwx;@XiV^z8aU5lrab4kglqR3Yhx0Ki(<s6OtnAfIm-&SX_%@lj5IX9=;UWAUb
zUtxo}ks3p6XTg+vis};gRDfnP+}}@Pac1g7o4oY}6aZYi?IPK27zdbMDi+J!)|;|J
ztWL8LC1!GQfsqchATcZJgpcO7_?`ZC_rBr@=*35LtoX>CeDKJzb@uqJiYDY=JJ5<1
zp(mMt9#+FR&lU<@c-#1s_9Uij<jHtmpOrVvBTpr-In-U-#)rNA<r>4u`XlliskfVr
zcO>%q;u=|XbqAH!J*xO30Y#k9ka7=ut;1eMz!AAHFN}{-<i;`VP;$tU?E@*1vY8RN
zk>RDu?V>4CE?nE=EF>}Az4$>I4X?v84d$~N4ai3tmE_xEsHIG65akRaOD{F-PZ#}A
z($9JSbB}!}iQ1j74`BoJz?`ph1Lp?L4V)V|H*jv?ZEiq*9@n4Evnh$UdA4)va|7oF
zwl<*O@ba6)_#NR^$(cJhaBkp18?f)k2g#qZ=LTNg2F_>x)!mYFpmPJ~2F?v!-avly
z&i>`flW}g~+`zekvVrw2WJOBB8Jrt9H*jv?+`vb-f#qig{ZP`+egE?w^Pwbas9)TB
z8Q|`>;PD;qJ~9%#;wbcyHTlpQ5}k3#$K5G|vvY$FK8CC4_XO#>sOA1yfqpI#?9*6H
zdLz}a7f1#gBBVnfid+~Au!FMH+yd}q{!UXJ&Dq-!Ssu?Wn&<Bt#T0U$#i`al8+veR
z?=mD=l`bZ009%1QXNw;>D-i+Xk<RTelOIalrd8Cl)02QnRJomPwQmIP)bW^%d{0mf
z-CnLWyBvqJ3wi=~+yML;?b;qBNH6}mRoty>*QUMg0OUUWy_LF1+(`>YT0(C*@`$#n
zHwN+{^UP^U8w<!4FXrp0JD#H#8}V#>;|AJq>Y-Wi5>UehxN!<CPv0B<(AoP$mvhY7
zjMT_Uk8*S$w3G2Qx<r3EmtfwUlZ+=`DRSyjXGlluwwU@h3vIa!O2H$qxc8dbKHb5i
zt@0NU?(oQEY->+X(#3HhVSA6YjJSs$8Rl;vq|c5vFY}UL!C3Ehj25(t4SQ`JmQ+PC
zbhQ?Nbj8N%a-k7f=erZR(|e%}r1y8e?R0O(SN?{DxDex3Ypub#PJ9v;{cPpU!20Gq
z^QqzqFFUav{nnOs>>91wk}s?gSRC2$zTz<Njie$i!d5o5A-JS05NlCzs6o6UnE11-
zl|%s}H2MSqrop}KHG0$=uC2Ym9dDPA<WibKgUmYM_3U_B;3=%2PB(Kjp5Yh)v>uOe
z;xuki%mMQKL{(T%!Bul<j_Ya5Kq*)%7C?03$^>T(NBn?2H*YEup+;VV&ydFUezDWm
zHfQwW8=%`LUwj=GvUC~Wjr<KsE;(Y!yv7$z%Tv;|e>HdmS*_^hNVj0nWH+P=vtOQ4
zJo#c<zDr|zRQxCpcc>MAaodB9t(Z0R2@;>?_NR(HLeAHl+W>uU@q3s)CgVGGoFRW%
zKo~LE8bRn>FySgBM5oiE!ZIY4z7+Qyady0zzw%_p6@R_RI0)Z|W<q}B9{6-tKV!vZ
zzl2qsZrp)a%rVJCqIyOfmR*g>D5OY9R}GJhYG;^ZA!gKkv3wm1GSq}U>j_5&W%w1J
zl0va-PMUsL7YgX!o_;9l=e_^3Yd@5<E=Xpi;q3ryHHc@+({?DMTO2kr;YD^4;uJBR
z?ZJw;j;9<^Cndl)%v>>IDdMZU5rs#@h;*9_6wwvo^aNXIwn+4Wd_Rv7(hr}4AO9i7
z0E==UO6oGqM-WfQwt#4=*#OKV9eUB5S0eeUjhB;-6(>~_p2O_tnxnYaa$#3JVG)|g
z)&kcHhf<QRq=>?EYLvU@EO)lpILGBlL!KDEW8Q1qNySWbMM==rzRFE=RnHnCg%(*6
zqjoV%50^FwLzRNcGsk4dD8-0Sh^(*jggWMYqp?*@*H9-Um|$toa%)fNlrqkc<wP_l
zFCh1#T&dZVZ>&0vZc58X&R?mEwj3nsC)W{;YAVr-Y9&&wAf8FEH66m)x4Z>`v$n7}
zX-#BYx#Pl-v>v|%c_p*YL>Vwr1Zu*qX8o+nF&gtm5VNqTvUF<GV9S--hX5)T<Guqq
zp;|tOim9B7NUVj!aI7H($>pI-%$m-Tlo6s|LBQz5Df*@N$Yzmr9}*mH%8xoJrmTpt
zr$N%KFjIw3h4jj*f+(DV<eFZB$gMC%jEsv}r$|GG8RtkO#Z7>~({<#DVqY#koG4Wi
zl^T(g#*jdwd=&|0sc{7vq#nXJ)Mq@IBh)cCL6_julgqFO>1-Y(2eL}SI^}OY9)kO5
zU#nn@S_oipYRazSWD-R?8F_>w$s8jxnJ-$XzLZFi(=*PwQbUXx@K_wWD&!d!9ubYq
zd49uW9(l&sLO}?&6Okm9TRjzUm6s+!3x%r{!{pf5qAF@Uqg@7Jy!a}v_zChHIW!y%
zB!*e)Qdh3!oMa}Jq^+G=t#T<RiJEE-vb{)@No9(Ga5w7i3`Uq<pdf_06C}xrp+!d(
zBTX4e{khCy>QrBrPG~d|uuv=*)?8+#Ai0&46RR%>UD;DYpj9i)TL(!Al}L!V3toB=
zr!s29X@~e(;wUa$PoY9=EC_jd9w?srF(el$O~4F?wo%R2B1SbuM!t9iK{ZDS`G@Ms
zqX&$&=$c&43TL@8)r*uVD%{x+h7@Odz93vVD2;GjMipFTrNbI#8?v|%9n5?YvsAGZ
z<kyHfuU1m}SZE?|-~Fs%MFpec`dSgmU&MKisznj1+t0)beWpPhXdSI5>64tFXCX3@
zVLK|P_2%ePpwgP*qYq#Hu|b5R6g9*YoK3pp3Tu&aOqgwd$xM^(nF9@Sc$J*+aKlRq
zT$eIcC-?FUX2VRAU`X*bJtH0uhkT305-2yJ++StM!{AIh2L%FT!-y$JBB~4z#ALEi
zWEE9?j^xp#{@YKl*GN_HJ;4yk2%<bP?3zu0xso@Q5r-|W$akJlxp$-qjEKDPH8^xt
zT#PM}vtFeXmdK}K8DfoBkr7C)R&3aaxi9rMkFb;<E~og{#gTG~QB8^X{OdHj>Wc))
zWv$TAJ|dN_;!0M1E#inS4vU*`ccYm@R;7D(+~fpT)-c9b$HO5_5HC%ABMZlw+KEW~
z%4znM=>@^kv}6h`py6?l4^)Shbf}iB2&u$6pdeZy5_U)VZqgl<yQ&g-7A-cTtt6y6
z(Hrv*&w~M4it?q13*_QBJGoj)$~v^pYm&}z0i1E0Xi%9wS;F!cMI0;7a?DDY%t2Xt
zDT_n(4RgzmQ)gIoTr7o$T;5%tK<`k+g}?{(L=03L@|h|mkteL(fE`;4KnlR9Ug%yl
zoIu#K&@vS37aO`Llxqh^GJIvG5Iw<*%%xVLykxY83xp$&w)db3N-&sGEV%h3ka)@=
zWA(~RBz48CIBPJhYKf@x7Q^or^nNiq%0Jwj0FK3x$O}c^LvdPgq+|f%ke~uV)|2YA
zM3YxA4B2PBZgd+jx@Zf@cwQ${RfRKkKu)$PRgsHwo&YA(MXRJn0vbkPB`BWh%m_Rx
z7_*BIMxEp-%$+hG3_|Bfu<kmwXefyh*%ln>3D<B!kI7JR+$)K+(Fpj~k2JTSn8Z1@
zBtwb_aa^rXDzsVxOOA+3Nf5t?HC{tO`3zX?usIf3zuT-xXCeVsp^H$cSRitk&yJu>
z<NV+_uD<gn!2}tUuX3Sr_{9!g<}b}_CZZjli%;CZ!>g1Vaa7KStDPi51uagXCWeuJ
z0^tl@=r$ClatbaSmE$|cMRL((LM9gx3eLf}2pB2z(mVt?%^e>`(is;dPZe45C}%8r
zBE3aWrK!<IG`XpPX-J8x8=V#T!c9`BuIBPkLx?_<^qJ2a%JllCPmj+g{ru}cos?z;
zZW;nG7{;{7gqM_Bkn%8*irt0eF41AG)H<l5!wIW+jl97z*vpdR3htyNyC`uGNA&5S
zbC{-q3dU&&2C1FGdCMkr85bdwi#lAnaj;yD1-fRV62Kq<*>#2`%5~Xah*tnN=`F<t
zkT+JEslaK-N8?RnYiZV}383?wW=@a{F^e<AsM!*6AV|SY_zH`4P-X{?DBJU)svhsr
zWq;ICsqzbprfz{2zy^m2AzVn?Af^@;6!tKql2SMGvL6k|Lm7Z8ykr-kcG-$>DaZgA
zo(1s%4&eR>kPH(^M>xvNLG}}0T;(q~Jb%N?5Yjt}ljh-?zMw~CXpjT&@rm+kvX2_T
z4&6Ft5s%|1J?-fko529G_LYA7CITa2JgXymVBrO^76~KqRH)#Q4lgzA-2>EKzz-Kn
z;>~e6Th<EVf;OolKIhX?vzE_TL5)Vy_~G|T=A=JF(X!!KAG4EA12fd%T=_U$l<CTz
zHVQP3G9AH>_-vf|>6mV121DnG5DoID0@DjVdvPH`B#n;ImOKPPgPtyVqBc~lz{^%A
z&oX=`6t|EPrQjN%G{P^GQ5u!*zzd}W+V@6%RYA(BN#6(+?!p^V15Gnnh?oe+x4{&L
zm0~QhP!k}o)+V8XBtGl{5&+hrn`{G>8PhCxc2VaZ@6j=mC&b~aVhJ&tW~|IZ1LBOH
zawl@z9u;U2o9yRG#e<!b<_F0TvOSs^`~60%fl^D!uACK-ApoA?_*FzVcV?D6Ol&qD
z$qR+K37`UaRaPP!@#|IdK9pwJ%sHAX5|Nd8=h)d+B{aTv+^q!UC(T{%VWArfO1T6D
zszc*cWaL?`bs7b!O%vErI?M&y<P6R+u4bmty@(tS++Pt=%_zn8s(6ivvm!kPsFQ@L
znWc1k0r+r7mb{M`PyukWuKi41#9WZ$?5Dj=3u2Z~^EDL!9RNq65zIzLJb(kZ+192^
zI_6L$Q4tPrp2sPckOGS)i;@VR$}L3nd*OEMef_*EJL6)O%PmXReA8V4n6c7f%{9+r
zv#ybyNU}#4fwr$IA*8G)e!`P$yauw=etQ>@{p}iNnjpCi(N;R8P|<{IIH9415-YwU
zqJr?~k)Evts*B1Iz9<+`Y>gm!PFUHD<IzFItuUw`bK%S~v;e3e@bG}d1n=Mz1=3LB
zD4xdhQg<{t$5w-I%ZJ_1j2iaV!WEfHrm*n!&wPRORM~_@tjW(%;g^7zGcd=q8pli$
zCS6lFeElnve*XJkaO^`#^hcGCdhwrwK1zJiJ=E<0xxE)Vz%<~2Zc??cpbnAxJ<;7a
z=@SRKcPsi%yXJLe+4!!8-Ean$j6^&o6D!M!FI1%sXk!*4Wr0as!f0(?OS=e;>WiEY
zgP?obeHh#8kp+QW>XIiVYOrUT803)GOP7wO*HbsqeWSZXY3t}t=U}f{^`_pl*oP7y
zUCK89#yhlUsrIZMu|;?1j2CD}+E3gvjRqIGH6<e2B#uJXx=wji#rfM0h|}01Zf%>Q
zN_|XZXWvjia-I#DB{fdEU^fAGmOC)Zp`*Oan9u9__)d@SaNyp*Y3iW|ck`tDzNL&0
z^8TPhA$)UNFP@xlJm0>zd6Y-*@+02LfujdC{j{FkN8hCl{cEO(<7^0jq$#-1c4?43
zw$~PX*js(QM}FR9^G`k>_8q_BbdSCzTi=^t?0%#k6^SdFfZh*<@AXh_t>V%Fr1lnt
zOC2<SRg0^Nb`Erzw*43_vn{~o4ll-#Uwo(QjgxGz3W*=v*fBQhc@4gFUVv*<Uti6*
zi}!j@dVPo6{I%Z)dP861UAStdJ3W@^4eyEWRVTi+NS^4u&-DgR-xT)W0o&}YC%CG9
z;~>ZMHhxCa4<-GA_rLJihmveD*y4XPgKJ>su41SQYyuNV=@$H{&BmbK&z*UClsJmd
zx=)GGkWXoTRg>2>@#~sq&qB+@CD>E_42}(Dw(Gf->-eu1dqpQfqab|eQZJV$Msjz`
zKcA*EzJ*bDF9#e3z`JGdNxZ`>Py#0Id4LaH3oQ}hkKM;py6$WOXaJ61bk+?adS%@!
zdTuwF`mQ~bSaPY+6fUu^a3)&f7=Dbsj%QuT-b?GzI(g=+d>_!AMW)YmXooFwM}WHv
z9%=q6xra7#h<xYcZuZ_x-g%HWBF8Jyy^u++2X{J(O5aKCmC3xC!@Zh8?s|jbRN>vN
z!2=}gY&SQ!TA&uzh-O@oI{9spshc%5i_IE7Yy&yi=2GoTy7!n3X(0Ugd65#rk9PnS
zYl9U5rH_JjfO(ENw)P0O{Ate(oEtc@f%+X)BdB9Ewk2pvEZ?NIpexH<@=*ODvbXq%
zopxa@IF^L)yY27c4Zx(RFT#P$i^F-|?7#S6=K$vhnho6J8O<8m=XUy`q+j^{%dh=V
z(juwH$awbUU9uugNnJ+J!%K`A^w^z5CRg#Au2HHeiMBM6p<7fEYr;94u)$stmY84*
z&89ZUWhz}@O(eo2wH&dpfZ=Af^Asv7I+Pn<g;D}$uSke=1matb+8U0C0Yj6VL9<Pi
zq@Gc$$Qb)DF)<M@@bE@y39^(Y+}UF9&+*>`L7G`NH(vY;#0%*oon|ioRnM$%P0ZSm
zb9`^S*^;$1Yb_HsQsFtxoQ;%GTI+S}ctl(Z6j0aFW?Mx8nKCYwnxpK`yb!Le7Vjwo
z<wRVfn7qEKayUW)LS1DHF-tt;i$+N+G*>0q+$?HXwlEJ802axVOm3b$!=q*j7nM9I
zqHo|+c^9@5FcD-joc;bR;D~tUBH0a`u|wos7(SE$2>=g4oQZC$hpLX^q`a>tv67o6
z>rqkP$|VVN0ezm-Y3`W7u#9yH;T*`!)H$|h%_faHtp98-o=!>xB^lQ;5r=17FCy|t
z3uEOaJS9U#du^CpmS{;hp~H2)I{_IJ6i0;-#Uk9hHG~EtF#vBT^v?MWWD)s)Td3>0
zK)->lQjRzcVEkdtWnK!BTS+;w`hw7vJtYJhTAH%?>L88=KSAuy^(GK_cvMiMs@CTe
zU6U2@$ek(<TX*~p3pvgvins~2+_3Tw)spAYaBLOiTG07HO<8eOV#Qf9UzDGr?4vZo
zwRS6QIkVPoju(6)Org<jdmfv^d@ahg!}pMV{zkW|Y@H_H6b^7j5Ocz*uqkdJDO|3I
ziSV}Fks&7GDYW?%7Q@$skUPEvQQQi{f&_p~F6zzORRBq>QnDqmCap!xgohj6{6M74
z+@?u2kXFGAOw8>po(tx9<t0XPnK#E$+@teqtkHg4bvyaM#EaP4R3xJCjL9~`>6Z;7
zmor42V=k|WmePV;j5Drn4xj=U&WiY`;3$8L+m#q*c0SdNw<16raji#dmxjth_*G4;
zb1yof66ULwXB$QGtmd5iL2B<tPJ7L8j^b0rD5sgyTp<`<ZYX}q<skcW8Ol|>Mx?PK
zJqD<ggsGXOG`%#QVM#{bV|98yq2r5_>T{5N?U*6wmoAJJWPeZ;MJTT1Il%(6+=H{V
zTLTH9BuY>PGy)6&uM&s|Z~*H~*e@RitKy1?ZpCwyedaEJP6qG=R$##=nI_*r{!O9P
z*$CZ=(i=!7#j*r7ddr!U9M51>q71^LviG9SJ$5{zG70ik#R@AmNVvPhF`umkRyN0m
z1ACC3Rc<H+GM5zFd<%{S6RzQe#vmxM;wvI56_1EL3G+23%&7Qzctv7T<QXx`MU$z@
z$P=6!*Ayy16@Y&U_PjfFj1l3?LKOfN1Rh?MxDiL?@Cc{?;c43OrzFD4SIRfF3^lmR
z)m)976b>S+_=;?+tb9e*L<k}8XpQhu!&z;PbqWBlGy4C&?aSZ4zxG2(tLi1>@^}kl
z0WS48^XCTM=my$ny#-(BfIQ1<8t8@g@R+tQyV%iv8ej_b&^`K?jl6*B(c>NE$C0kD
z42@UgeNcya{#MWRm7O16$Kis~?n%L`7&3;C-@-{8DDPqlZ9f5rXSwBPk1e73XPOW>
zy*(R{Pb1|UxTorMa`K~;{g}2#`v)$nRuAc{%*)6<wJvx2Xc^6#8|aGl>{8i9x}y18
zDEYw8F301mo|H^-Jy>s+vgQPUXWALR)(yPc8#o4g+-z^}R}pvH#odJ3C;GAKZOaq<
zxHoV<av%5oKTrRp4V>R-Pg=CJu}>S!P0!mfZ{!una<iUt%bwWkriI_MjGKyY8v88%
zC^vBNfxAe!xHkQplHR}nMb~~P>Gm0XNGuhZ@Vx9^zgcgksY{pWuryKQqKXOUaKak3
zZRpneZT0K2a3+|T`|Kvpg(E&%dz2q7udh@MBmi&#$3S=S9@*uQ+IKnp2QBguZ66)A
z$9uGW7A{ej`~sLPObFpZ8o$kf!X9QcrII_wsF4j$7G(eqU?`JY0Ut@i*<b-k02s}T
z=<WEB3=@k?VM>s%+x(I(N+q)upsJ(zR*!=@)S6^K4#0>GaX2Yzu%)!r%)lJO?jJ(p
zqdM%>H_WzVb`k<Gd<q}O7iGG#r;P%I->TDzE{7_8nYe?|YxU@ms|k<XSvvNa4(+fc
z2hw}s#7S<_UJe#HRDkfPyyPJTgkRD-^B>~|<Z0onJ~+$svjDaykO1)f&2K6gpWZE^
z;BFoowa0sO3<DEw3bSDbCeS7h7i^sYHF-K7rxGvXr;oC?As!KD7HvQUFuawBj|w1v
z8PI{Q#4xi)o@2xBIF3se2R1IO6dW!{wY|Hch*Ob4|8jaX3sT=1P&Uij<P6TSw_g$o
zNNydyBv>kt@Vv*wsJKO4U55=mfHQ6sv(`l|fT1+xlO9ZhL=X2U6bJO_oQJ|Jnf)SC
z5+x`!3jhPaRuwx%M1TYM_!eEr9h$8KF|2qsS)+-Ik_ey5EkyKt;q!ra>Ke_gjq7$v
zEC8lz23L4#>s(tWVyHUd$!!O?7JQWM{&@~e&{A7)w4ZPdCp5-H34}-G>cJYAST(Lk
zp3>y{@oug>d_GhVczFA!JIddudJ_wdgc1n9X^L9>z!uj?UsF>L^w>y0l=O?<|KejG
zO8V6F8gG5}L*4}`KmPW2ei5QB@52*ZVi~v2bwnJ57urSAAjqg(Hnv;JM<CM*r(C&`
zPvtdKEqpDcVEK9A&@D{kc$<WAJMN;7))+uuWMv9HAP>(VUD6NX>osd{`YMp6#Frh6
z2D}Sy#6yjNprB{mfz&BMJ^7{h5a56Oe3~6*s@w>fT4nS+aVGA|r>nv>vL=VxOA!ye
zPQB{OQTg0D51pAr*6JQ3Dz8xtx$)&mQXMZ6j#tD|LJjz(6!E8`D4#r2Q>%wi#V}k|
zHFK2x)L_)7=tFjOQ+yBYV)un^liqEnIWEyi!580QuV&g{*8)=EXJDi~qtd;wvn0&|
zWk_wrJXV@6(Wf>j!nuZ5GJx6Y`NZ&EbUCO)?Wa*`+cGvlTHg9<!IhRaz9e@`OqJKx
zO)|Sf?$>W2Yiy3i<-(xrCbtn?_yxE;+mV*merl1Q)_p!TO)t~kv=@$&<c_*sTkBSi
z+7S)YrWRgrNoYI;>mfEIrGyDKxIaj{Pr&!2>1&g0^sk*ah%yYF&tJv$=|wR}#ww+r
zzK26QyxWWKUsk)TdKT{dLnpCEjqm)TMmwRM(+~P*`!x})SgXfa_RpaSw|id7PBUGi
zpZUuh*t^obJnP8sG!|gHOO~`$!Yo#Ng&CiIo=r}J(g>ebo*CQw$A*oY#e?yKFds$0
zOM~mp=JzYeWhr`B;*ll`t8PkgtD#UH#u9AG(}L{BCh_$oY22+Bt^E?m>WN6~eXmR{
z{=V`Dv!-)2hnfybXXoR)A&I7L8vHV2Dwsm8zg1>Zy>rOR*{8`{n|50&+guBJnPrFN
zTSDW|%L%XA*&OulXBdLICfZNAalAHJ6;UMQ>FzfLHMLjub2Pm}>r7TR+ax)qD8ml|
zoq~UeGtqX1xXZ3`Oxh`IG_x+qN2=TNoI6&fYtXfLIysk=j4FzJHE~Oc?NQFr$S?5f
z+v+T~nPTrW=MAUTarP^0FgH@;Zfs}4?b=gRm$;__G@If6eiDl_QzzQwtuLSe;M#2$
z$!^0q!1PkFSmw6glpSJqnvEzilZy+Cbf5)^Sy?B1G`Gd?^tZeB6|VsuD?V~3A3Snw
zojrc5q6zue4zyxL=t<_Eht)97vxPzz-ZuWEJ&EZWc`}~YXXOp^$WzH{4t3YI@nLU&
zxyEp^{)qfW>g{IZ9f`cYxJFi8-9e>wk1D=MKoKW2q};<^>#&y*a6~T53*%!Hxp53T
zlpL~T`#?&hY-U7mWO!+EyJ(7(3)l8I3rS3OFMg0l!|Sk2gZZpR1M-nZCHb}(YAMqi
zL^*@V(n}5d(}g~i^qJ2i&GGuCPmj+geZ{d)CsAK{QRhpqJ)EI|X^{ypHElu4!$c}}
z7m~X~hq+Sgpo$6SaKeV(ig1%-a5tALxHGd*;vkOb^U)x0CeCxhB@q#;t}<Kvs#7CT
z>ae~CK#>5z0St}|*<cuhb6xf=e~Z0IZz(Q71mV?DAB{Ipp0zaV0|n4|PBSM+hM2_}
zV$^JjQjlDey~<)8l-YqJ%Jzh)s>geD*&nqus@Zvr1z!NGfC(X7Nc$ir=N1(9Frz7@
zzY0;YlyG@*PzK-t#)p{NH6fCWTQn8H0+0YOni<ijDI|nsm`F0hQDzRZpZFJAq5~}o
z70~__9MMsn1P<5q1wASgf*gR4Pb6K|D{25cbnBQ!<~Z6e83T&7fa<azI-F@!4!Xlt
zH3}hs132o!J3x)13m(t}@WzgC3TOd%>pqOLaFvwOH9YSxkfoO6TY64BACKAD(*Xj7
zNAROQ`wY}i$DCUC=t6gzJCn{4eb;cOi1jF#zrv%`s?8Q@3Q%GPzJs!~yP~6Npl*~#
zxz~p4QBHIF2ISWARqyh0vlqa+f&_r|@oKUSBrRaVCb-q220h-RV+NRLQ<x1iFo71Z
zw;*`oEt-9*N0;JVK%><_Y0r{fIV&Rl0iI#;Gm=L~c$z8G-G_CT3SD71Km@>u4snz~
z`iKuEH5s#ZvN<*!$wR&u?sN_d9WJ;y$c7?LMW+5~MClf!0_UKkbePw)$r+quT%DN7
zt&e$0uv8%7d5?)vaf`aT4jX&`XWS@et&3U!Lutq-J(vWE9_~>nzS37j4vX+bIuo<6
zJ8RS+%S!+Qz+UFFAtJy5>~hhdIZaNwkUNwuL8h<C1Jn@OEDD-=h<-18KJZRmqp>Z)
z7@T6m>>Mj_F|$Z`#kSMbH4tkwcmvJUhSEUWLw4D_h;+icu{1$)+wwNwf}?!GHJs3B
zgAyygBBFxu=rPZfGRwT4+aX)~OM#V{b-hZSU_2qvfpY`?ZWs{9yf`x-Mq^YEcz8f!
zf_LzV0%<656raX=Z6P|&Ynv0yQ^qd=70kdK1JM=sLj5a~zT*8?o*zo`rOf;45*;?z
z)fiPwIENE9gjR&rkHK{yiCn>*nS~MuaYUbw26;1ao)a#Kh*))%+2U878i`Ve^*sQJ
z1ON_TaAe2^!x)_FvTylY>`i)0aRDL-ua5d?ym|7hrCA>+fX;K8IYBbSEY1+4W=oWU
z<eKbN7VDtQ4jfUoCqz{}-lNO@sHIWO&SNb20$2r12;oB72QfLfps<G-O{x0FfGOef
z+tYyx0OLbU?V1ot#x0tPU;#(~7|o36(-aayGE5{H;V3f)*-!inEzyA%g$ii@3XbR~
zP6CH(`hp&n2|*6P$0w35>lHPC9lCYQB6A#Vmy7|$T0nK#4;{|5DF@x*sv3n5zyTa}
z;T@nx(FG4^0(fIbI0du-ymcSOS-46{=^CE*7syh}@hv?ko{z`u?CAi3!Xx-mpM3`E
zr(;g7dvu{Y&7Dc-h`wvMQ^a}{%wOSAYSm_oGzBQJ1K&Ye+Fj96HBdK7qulGk(Ik2}
zGb?RHgu4rGNDXAoCyST}$G0H~OEH#Ms0k=;5>mBAbOB5NBmk`EYO)O^EnvbXxYeQt
zJ>H{Z2AF75m<=;9fflg0Ab8;|ntiHAm*QPOqt!rZ&yrm^D<b^?o?-Dbl1E2)nkm!W
zhjo_<U12yt1i*+6ag;#%hz}+;8MAh>IW`>0L%tX8bPfw0F1R?zh9XWyrv7O}=@z5{
z=b)o>nAfz)8JuHWotVk3k9kS3R3PDbkBL!ni@LfF8+-s~+$d(Pi&_9fX~-u%m;{L)
z?olYd(pN+di||D{6SJ>7Yt$giO8^7FUgonQBESLca?zkUO-{OyJCrR!rmx8Z)DYS%
z3YvL{elL7J@J?N$u`R(EoMOZ594l}!vq*Tww$s!#5NkAe1I^Ti(m>locG<g#bi%u_
zG(mFP@;2XsqkO_OoX}{45-YwUqJr?~G0&7T%e<c3AzS-Pft8tcy-J>7JR#75a|8Zv
z7!b$2I5QtcV^k1$ctB!;ckqb<X((|NPh)whJDQwht3kNs!|rEB4SQ?hicBR_2>CEE
z3FPo-P2rb-RA*q0bLUOp3-v=uU-|y4&JQK|Qs&ili4L3VYK$r-oWltlLMy`R$KX1U
zM6Te@%tDESIHJ!-gS?qI&k2`AM69~XZ1Jm3jYO%#`W^s90ssdvI5K2|VGPc7*|+>H
z_9nfhxBwA^S4Vv`-aL8M(yR{@K<7EloFExu7H5c2vn5JFa!vLsi*-<D2aYJ)6QZgf
z@6ly{)Y7PC=P?$10jvThgm5A4gP5FKP}sward0i7z?5+L?dd=Tfbk)wc1?&R;}%Ut
zumB_gjAlmkX$lD;877j9aFm&Y>?i((mgqo>LIt#c1xIugCxOE?eL;`Pgdhjt;}c1j
z^@<w64&6FtkvWdGOU8g=EugyWhYn}jl!NYYRgFRj-~f)g@D5O;=z<3{0lcvzoB~<^
z-ntLtEL<g}bPdn@3uLL~_?Dg%&&OkS_H=+i;Sv0(&pre7(=n&kJ-X1H=FX&ZMBg>s
zDPlbe=CAN5wQ93PngW#Af$yL!?XKvk8mJqkQSSBNXc9e~nU%I8!rg^8qz1C)lSNE~
z<J*vgr5H;r)C3ea38~s5x&S5s5&+h7HQ5G|7BFEG+-gyS9`DgH15C6j%!V15KnvJg
z5WMgf%|6wmOYttC(Q2TyXUVRd6_Ne`&#?Fz$)h7Y&6Mfx!@5g_t}q-R0$@akI7%RW
z#0Qg_j9EL`92<`0A>RvkI){Z07hD`<LlLJUQ~xxgbPG~}bI?&b%xl`@49+pGPR!)i
z$Gjw1Dv<EJ$Hb_(MO|Ho4L*P~ZWOcDMJ<4#G~|;WOoBuY_b3!!=_?|KMff6}iP_hk
zHENLMC4d27FZ0<D5#RuJxoFUwCMR9U9m<v<)7RtyY6xu>1<gD}zZX6qc&Dz>*p^@n
zPO)Kjjup6=StPt-+iB_=h&39#fo5t$X`t;PyX;*=I^o?|njpDtd7E#+Q9j`sPH412
zi4|WFQ9*e0m}g3vWnRzikgfftz{<?JUL{X3o)GB3xdDGS42WZ1oS6@!F)9c=JRmW_
zJNQI_G?X}sr?I@$9Zk-$)gavRVfQnmhP}0LMW&J|gnXEo1af$^rtnKZsxvUhx$~y)
zh5DhSuX_L0*M2Bzv(K44H*jv?+`zeka|7oF&JDc%4fwB|#!*^7gKz)1ukr+2E8fbz
z{9*LZ(|JfV>A+EH=K(hU)Ls>G1M(&I>*Un7k|Zp=U<_=57eMI<54F?V*Z}|kymjBS
zN?xbjdv!O8&(s|^!0W!F-m8(4{|CH{^T8g|4tUx}Y@mH?TJRCwPt9`!=LTMS1LwE<
zOW%RB|G9y41Lp?L4V)V|H*jv?+`zeka|7oFUf~AxLrGu#{%aojp`@FCOL{@zLi*6&
zMXM;5;E-4*-$x}YV*i?6um&X%j>m!-k{VwZ6fdNqIwg+QB0j=5o8KoHt;5?onX}xv
z(;TKp*8KZOBb6gv3O54t%vvjrXBZxAA*v9<txvK$s5>av@7~4T@!=Apt|4Zf+B*AK
zT8hXBPaU1&4kayMYe83)Fzhx0h<t!*5}*XY7pg3c?FHSoFx4d2w>)YxyoMq=aZ)Fg
zL3k=X?%qih$sqj5pm)%?gT#@E7w(JpKi(Wv0C+j=)c|gQGfT4_RHd{`0xiMT0&0}~
z+PKd+>JC@cD0CNGdH}c%ut`0ju$~bV19+|P=DZO@y}3C|`e?>5yG*kr6j*SsOlWfP
z(~?QqM-|RdM@+sh2k@h=nPZ6eK2~-IHtY-%9o%6t8p;B^!V!fR((KhlMB&wfgeSLf
z5MD@!gDq8*L-^t52Mg}m_n_86`o_VK62j3t!$ym%Q}+2yvf`KSG9n)IU|%GJ@VmXS
z!8u`PSRsk2ZVGd8lWHKXfEk#`t+Ic(z<1qLHqhwRqX^!{8?6RP8eKM@{|a9ukubvB
zMu*6)!Y&ZCIf|#f?&^DxZ{wJ?9&>Csl81aR-02(^I$V$jIb4eiB4h8>sgJoqYQIHu
z+DgbC#ixo<o_`K0f=GI+E#(fsqEUERq(?L3^q;o@csF3;F|$3n-Fk>JKnYgvZUpH7
z9KeU)7LMvnZzH>pcbJF<d*lGzlP>N|W6Mx-ol<Wry#EpIs(4qid}oj_!c(nKj96{S
zS!&cKHz{3ot<3CH0ASn=6`%^h2k<oA2zL|L(pQ7PGoZv#Je?Gk!bQS@Br5KH=2AKX
zTT)$a#S~c+)T&2qh$h=W#x^yjEX*00W7Pq0bO}#Cl=L<4f628UN<z&to_#xG%-~+7
zIXs-mgqQLyNO_n@#qL6Km*_B8Y8_ND;T%rb&|491GQk#_Z3h7$tU-x`IHJ?hD5u7!
zTlp*EWN{{fg=E#w5Z4$mDH(r;r?{)`u+3JJCTtzT;f>l7WGzp)v&G(@v;RULcg|Cg
zp9h$8m(#l#AzPY75MEHpdo<oOu$E?hpa440Y32mUkbSn+m52fb)U~wPjTTH9mrAWs
zE=QRicuyH9C*qRZ&4blkiI9MhIJ@!(vSv^Pq9O%tcOi{b;bfb7DVf~dBg5$U6Gsn1
z24xVQN|B38K_<i5N5d6PngZYePKC)mL{4?dg%ThE;Gu{!somnDs-rjw@2g3y<fg%T
zRMfX}Ny1z}pC@&iJ0>tJV_iZx2Qo8tj;)z*QVGKtm8^Yz7XQ31Q#c_sG*MWFh2T^u
zeXLtUPnBz{;IH1Q$sE)$=8D4bRDJ;j2XIL4p01uYm7x~6$EeU>ht*$>$-l8-J)#^w
znN`|wrogcU!<t+05g@!IGD-ZnvZsVVLs(NbU(t1N9wa5^dJ_mdJSwYE5`sCcSPLuS
zkvml!w(j^H7IK_T6mb)3xnboWswL0=1t2l#%Z&uxHK-{ou1c)9D5tCP6H;81Mz|Jl
zr7dUH`pxkZac@1GaVI>q5&BcDv=x!ci#QP)QUgsRS%{bjZ`&OiViKOhX)MX|Rf%Mh
z>wy)9RaOAv0T>JIDu5(zTCydWrz<gHCOq8mY(J)zG6tD#QVpbR&$u%uO!ku9D-gW!
z7VSROqw{K94A2kpLqoiXtxZKD3eT8qGn{_eP&g7~W=X@ug*tffqP{7DS|7#HC`bqK
z$FN<AVP@@sb8L7k0<;m=wK@kiM)*}ttZ^^eq5`%lR*Q`Pn#8(<4N`kIayp7y>eMr{
z$uCMHpqbKKkq&Q`SVf$%yT~ghu{s~KI?WCnWB_N}C}wRCjc3@>S;Dc6jywyC(WU!R
z#W}S$VwK$UfX}riAU3Y#ab!-m+=R2WT>}ZBBuY>PGy)6&uM&s|Z~(8{-X1SKVyohc
zh*-sQlzrwd!1g!`U4aFkWaj*cXJ~Vrncfx;g$bWR8L9=X79|&H(Xuj+_B+wZJ$ELR
z25%rPO1QhjF`umk)@6<j2lgO6tK3iuWM)N*85<Y(wU%-6-O+@Z3D<B!jl5fCe?yn%
zmeMz0qh&_L&xcaP<)CB4EEi3tG9yoLZd_BS0963~CD`-s)G<bcGYeG!R1kQ0RpLe*
zmBS;T0)(e&Aq6E6PMS40M>SsPG1Nkwgo8=m%I+{_b>u3yMn`K3zXaX53qb||a5tTP
zDCw8H|E2f-P}17%B7*0ROn6DI1yg>X3st&Ahi$h(6%)?kgl!8q*<0%y$(tcK6HLr4
zcN6Er5g)BR%8!=USE>dQ062hSpu2dF?D9zMyBz+57Ws&_kB-{oJ=#7Cm#9mA0ZbMq
zgm58^-{wGJ4>Oul$sJ?V$Ob2iG5`lKl*z4tk0jx2umB_gjAlmkc6>;NiAAO`CCJxp
ze#sW4lGzGS)lqz_$H5$GO)?+{U_^&FoD?<KQd(+eV2)w;4<YeU9ro%QW?M2l2>}>B
zg^%NlGF{ozMuEa_)#*f+LzTWv+`;IzdUVLugh%cy9s5j&c36@F={<1bBsXa<2a6mk
zKzLMM@{j_;FX^56k8uO?wD46QoaOmh0NWEt0C@i9Hx-Oe?-o&THxG^4<2^ctfr&PS
z*)RhWXcLDEw$6Z>JROfyi5KzHN4dS>f#Cra!0=WgJ}Q8GeE6-BIXB0y3K9VJvJb?;
z$HbXqNAGkx%vCp&&f4uHx#ZUJH4#X7mufEOffj)CcZyNW)?S%eGe`MU4QgPb=l5ux
zE=33FsS>(bGW(^VBuY?d761l-ttwWBhyVvLyhVr~1%*+97*@QRtkJ|pNrX@379#q+
z@VE`#32lm*<-0>;IKdSlbERqxu5i)TxwcNkP<6tS+YWFo_$b|L=XtR@v<QS>HRxT%
z&;(Cuf|@@8KYTt^5O{d|=0D2cr+TtJ?hVj~l0Nf!<1_yB`le5h&nEq{W1mjq!gA4t
z+Vd*^A#Y*q4P>f`%|DM0afQuJtSiDri4QV_$WEv5_S|Qwq!KA(jq5I#kHB+rAL`)$
zAtiTTs*RBnR28q3*yXOL?#p-;eQKbM${5KH#pCtGJ~^d@SH+xA;a=_4p>#v9c#T8D
zDx85OBLUZ#<=77<d%1K<X}Hyva%&Z%=|4%zGhnrSEL?v%^Y{*zXCTJT#WFGR$2Z)p
zuiu12(M0`j%v6a-S`<mV)b#ug^%i@bPGt1Yut{aiMx-%$)S9$NPGr$VTiRY^PlSWL
zrs{g+Il08LlLdYSz)d@(`So`5QI3jj^KZOEdzNd@>JeLXcg}c$cBK8pEz^u(q{r8c
zUW!HZ66<n)8qPvgNO^gH0Ao1_SF~h4v^898TXGZ>RakX-?Uxx)d^<5F+a~PIn$V4M
zH<a51yCz0`#yV&tZ0N2HF*41IaLZXiwAKRqwjwTxKJE4RR*&uQSoNpfhFy*L;n>Bz
zMB=ATNiVuVef+OA^_=M>$t@g&7t)IcU=?L1c|)ncLX<k9nP*B2*gKFWx8Gu{vaw~3
z2Lv$~H7^Y$Ze}V_DS3P_56K;;!LOR4o;@VeVeQ6_TP{C;1j{^S6SF)4mLFia?%m4?
z?fr@}P#w39E{A1p-o9T5vIFmL_<=LaJ;*`!eFy1raLgPo;KwKs-o1r8?KL&q()lRa
z->$lt-5m6eeJ&2N5acOe;D-R4TX;yvH!8Y35eyt&AKlw!0|Ykk!kd6qhzLFUrrtWQ
zS;n**4#6rq>Voktx#FfQLYIK6hP<li(j>d{+iRI;+h*@>`W&L(_AocLJ#hH6x8&~V
ze~i6Da5p77+rp@n{M{07f;i5T&SYn0=ye>GMv>nG=ne4qd(LPI-4%q>y?oI95#2|r
zKWOM}S>5zDDP_mu`Qd>}0Ey9uTgmrXdH(n*NG;eEzC9fp5mO+4HY6jr+*uEG=1HFt
z%T~c|mkFF#_N;rQJFw%+C__u=OP3iS-8;YTF4)Yce`V4yd;iOieJJUn57|xPbn8Bl
z*Hex{KVN(x`}#nSVYkQZ(~rrGqx;)=Uib~K{G>SrW0iU?GI2y%af*|xyr!U#7B$v%
ziaBPbV>86%gfp&l$^;`QF_PRx+Np(lT!HT#>#%aC8(8{=aL}E7`#e0+S@1F&i2GWf
zZpYnunfX43u3IbK%Dwz<?05V$37T}^D7Etd+GT^C8;~!tUni#uNs_Sef-$fKUI3*d
zJk(BaV*~utj_#XQ$?MeLuTt5$-i_tmHT5BFx}1d`vG~<ZN!@LMk5N4^o(wQueX8_k
z>KKVf5MIUS5rP-q%t`SsQsS@f?Fy^rhll8v#L(5&7D#t!V|BUEh^+J7iQMU(8+h3b
zq+d3SLGJ&EUZZF=W@)bQvs-qTC>aMZJ}fFgcy8vtlWFuazEhsy_!Lj*b#)6Ka?hlD
zI9E?`yfBTE+~YcTt$6e#8~T)jX_&a{=})ms8z}do1h8@+3znaWW&^U**lc(Wd4i5`
zr_#rTI@EJ#yP_MF)H047a%W;aZgZ#dy*EJjF#ho3-s%t5iwE$*jUFw}cL6ndz$sbs
z0lMf25P#eZ<Dq_>I6RLgp3qr8l=REr|BCZNNtbTN)ALWb|MXMyn|<Sc`VD!RhV~ek
zefkvKIJ&=`=jHrR(vD{(ZtNhBpv0(t7ip&!>Tw0WbF9P4oo-<18^S?%_U-fVL}$Uv
zY#{DyeYzcY=Vj*m7`kq)cq{kvyV2iG@Q`TIfuq#U16=++OKw2E#D1Ne+E$W;g%^y0
zE${*;9pRyNdK(+ypLTTLv`Sv5+<SF5iqF&?H^A$@quxtOm9x+z7QebFsk_bi7}W#g
z$pF*Ur%G?8j*)l-;Z=MdA$Z}<oD}aOCI0H(uCQu;c!+LE3|(z)fpmv9R+kHn$U5Jh
z$erG~ftTGt`enly<o=K7HHubamgWjSyJdHYl5qgz!=eI&=VtCZnMN<;JLL(EPw|9a
zSGV9H_e{EnbM+L*3)48sJ+5=tibqefp-(B8hKakL{uH~kfpQ;804w*gVELJ7HXu8V
z&4$;IC+G-wDt%n2Lp^u4E4oohE#t@`cP7^3Hg_uDdjoV2;}0+Ht^Q!WcmN;V=+W|g
z7f_Q2oRTFUpo@+G@yE?D9_q)5!}Dn337z#rNx$O#*S_ixB|Tvib~B#X%39vvzO_Tp
zy!W5^{)+d%Q&%az?7{pe`sa;tsZ-8pKRr|io3+K~82QG%lg({zk-V4AEi>Fwi{?jo
z>eW+vG?g8452ziW+6bToz*cp&y+#+*BsjS=*%V$o?44w4`#Xsu8HCp%wxfG!Y<Kvu
zvu|gAcv76X=H97VD@vdx*jhl1lAqiuL!rK$<PIllWaJ_rAcQmJ1$=af7Z|0sl)<|3
z44ZAOACVC}p}>MOp+_4Pl|y`7;d3kw)>^1<41gyCi4N|tSOdxeyuuNM7t-vZ#8G^b
z{kTK4yf?KKbO?l_cZP=FK<jF@No#gN7GUkM4?#i*U*DSd^*LErY+$+xIm~`2_b9w)
zfPRP{8sbG9V_ofSOCD-<h<x?HhgL^j;8-wm{cNf7vFfSOUia`ohYQjqhih>`WH$bl
zOpw|ep3`Yc_-r8lBtB^Ji8JTB1Eq8O)Bor;5D&<7beNcUcM6}}Zk@UeP=b}a8$mh%
z2Qd8Jy%&G?d2S`5fv5MeTif*Y(L6k2)~K(26X*wXkb~Ps=dO5-w1t3{s<z}THR_VP
zZ_pxy%8@F<kq(>>2myFK*>2)m`f3n(29!98r<0;mc;jF}5*2qpb19vHqf#qhku^cB
zdenw!vJGVHNo`@yz#OX%fTK%z`k|z+eg7-3{ZP`LW_vheCQOS=cyS0^*s@b$aUr=&
zbXZYa0#q^K98TEKS`lt=3|?l+!axePMk677uRi1<;W!m8y<8hA>>;U&*GMEfY!crN
zkpRE}4DK6bZ_iZrcdZ6$ag*LsT!09|t7AMGFPoFr*3zsG6hP-W&72?^VisqJQL`mV
zL2^y@DvNbcW(STa+Y_Ry9`Dg*f7H^bX6G>$d;zQiCWLSy?Sq({TTs};jHXomW5ASf
z`R(aI1%UA(rglw;B;yuMMX&%Q0E}iv^l1tSAsHr;jBu2hgX|~%g_h_*i$Vpoe+5T$
z6eoehHGM&k%7h>X;NufXm-UJozz*FyW|29LwoAr<VlAM$?1v6#+LVLta8->$2;cyY
zy6_HAqv(PMGy%M^Bb)+S0N%O}<1AbyrF0F?`wL{L<@lDK6VJzEcJ_3DK;aSmsLwtF
z_0ut@);+q=o#xJ@b41@Y+$mx`3g)lyD79*{MVbPX*n#h$EbXr7s2ZpnrBUuGyeVoR
zo@zgin0h!fD{V!Dy9;kf4P?zHi<k(<w;>5jF_u`U2@qFnlTbktA9eu=0PDG$Yy*`U
z(=2y(QRg1-(J_)I#Nn%A39;EVV_~u|0~4vZxO)qN7hVyfpilMaQoIXjv>GT4BH5L*
zBGMn=85Tbyd31!QnKIpdSa+$=6@~*u0F3AmM+u~l_+V0#F>5EAW5ba=<a^;x=djS>
zf{TM}DB@IP>YqlGZb2$=4mwJQc}<&~!8yj&iJ9E`n3n`g1rnb3m>3nesH^L+!3S{0
zjbhfis0A>ThJ4b4Ns#E_9);p7eMRK32w$W#G5fl+Mh&vO1TX;XWj-4s0vy0D7Y&-z
z<fIF^L)j8!`kFjI4WZ4VpqYp0_rm7`@6<IK+Y*eyDK^Z`u>uz}i-cEfJ55~!u||V8
z&`fP84YWOEm%WQfC%hX=6C}4SZ}TlU$|qdI32oaHa4lF7DL{CV+oGTx!jo+6F9lX+
zcA^7#-da?l0#pI`yJ0{a1K~_eTRc<{czC4{20Qpffi#pjil+g+)E!OEvDF~l@?rNg
zqlUe;a7Ct)DTI8Om;`cow5IS&K&mq^$GP*S?}hrIq+j{|S6%y|q|H7rko2RyK<_)Z
ze9BSAZ}Fbe=gzzNh~(_{n0@-ux$%p2zu~y&|GsU<^2&`JL5WfQF49gd)Z-Q5-Ejj;
zZ=epkvu~fT=&a``TN{Y`TAyym-Puw(bGK|@YsFi+TYBE7R{uPmheVSO9Hn*+vGJ$&
zs*oFyFR@=Ir?!<OVc`X1U<<qeN=JC8o!-U<_@^D+H?5M_DfeF8jp8$P#|`kh@2K~Z
zQspf4h{dmNO6qPiK1TJxcrw6r^{LXEsbeG_L3kCPM+jbcGbhEnNQu9?w=1lgA0DDx
z5<^#8TOi$`jn(BsBeKqSCvvBEZs27%kbc=P2D$$udX1vhn5DVG&u-aWqGTMv_^_w|
z;klXnPNvbz_)d9(<5N7L*VQd}$UT$p;aok%@xnAta*ylWwc^o}Z0J)8reWf)r$5Cm
zZJ^wT62QuRELeUfnhnTKW3%Bk<Ow>$ok|}U>QK*}?TT(xQp-4U$eoGxxXqo)_uc^A
z!}!CCd#gWKFCM@LH+r-@-v!j<0jFfi2k4?BK>TqtjEDMh;_y70ctU6WP|~k@|EoWg
z4<%h%>Y1+_c)~+g8=lej2F@F@?q_2sl1hZ+=3jUgICXe5-u(PwEzPEB+)X)lQ#^CJ
zlyeR`_tZMr`U&*ca5|a6CvoQq$3NQQHP3eREFTG9Z&^0tpZD?(PXJmX!tc=Xlz7+%
zUgEAjsppg0JnYtd48neo+*wZDXF9aQ7P%w9-35;{f0f)r8#zRNmB&4YdcX$m_E@Np
zn4(&O2OI%QKDDzp+cJ34H<^j`hIhAvJV5epv)w3k>1dbox3xEQv!-UTS;L2IAONPK
z(|vmzC^xzUPdZELDQ><#X@;}Sxq**%1L+%Z?4U+)DMm93iImty+ElnId8q9n@>S!!
zu_9dvnRYk(L8Nfcg|yS$v%v?hmG_tZ5B%8Y@!seL?(vLfjqLME=tD`L`NMyX|4NZh
zebcAMXOq6}*r$`E^3r;Rb>VtRMDrI^u1=3nRqlV)Sw~$RuBLB~z+hMy6xf(u;at*j
zuEu6@k9TYxP<$5GdX>%?^QyV7EE{(iw!;}%G7`S$xje_>3suDiv@r{jvVay<6Rpjw
z`MuGo-b;PsJl)goQ`lZN7ZfD%di9Z`#H2(OJ-?tLeEgb4qEC1{&~aI>$od!k()J>I
zBJ5hzRFlG+dMBL^%00T2ZT^jSXwPQ-xjW!w-BFf9=wF~6=sr`smKVCwTgF5&)lTgY
zW<p0TXtHm+QoPv`=-(BdcG_u=89Vr9>_%N(T^i#tW2OesJ#>_p8S{BvAKyuD{p&aI
zfq5;7Ll5re(0ntL@%SfTW0V`83aE?@g>WeJ@Wk~S*gsl+aO~D!*>``_!^|run#vBh
zy+gt4c&ZLe5!Cd{h2-90C)Y|`n4(R`ajxz9(7NdfZIAG<0bsXwX^=hUTMPJL51afI
z2(L%>p*B*^Ip!>nNdT+R{oAiS1Q}DH)QRT}#G2&B<=g>3?)u^#P!Vd3luLV5E4GqZ
z<mo<bgPO|aP7mdEr<E$bJ&#|+RYP7?bhoo?-=2B4ZT8)Ux{uK9-<5Ta>kdD!uxifH
zG4@=2OKjh@uJLlC_z11DZ!X9hcTwM;TJ@I$m&bJ*0E5DFcfTwI8YjHDf$jU^*-ecu
z>6M3_qIS6YE(K+TX9e=GZ#?q$(903=So)?gXqZznwzC30E7&{a#d$W4tAAzE*S-HW
z*L)}`I*)f-`@wI#Ji8r$5TnvY<9DvfHG<vCh|O~i13G(=GzipHdCj9Ra?TN^UK!=_
zw0QW!l$Z+nOD=TMD_-eP-X>u*x^y8+u3EwEa3hi*#;{oGxl%4!d((HpkfK3GduwWz
z=BB=s0WZ(E1BeUEx-`0GhUI^BylERtc@L;UoQyqI&KUL9Oz7AkfshviR(vDkBl)?U
zxa!*sl+lT_-C9kPxKwsVAr-OVX*ptqRVgd4T$Sadln{O?P5i!RRiem-Q_E$lmkk$I
z%^YPv^%x2%`jB1S72ij@*nOefB)3^uj&n%&8ou}rdo|MryB3fNKLaD}8I|sZoh4}&
zC_`!^=CRUri9WSK5zaNdk^#(C&nJfWqRT-YYCnxi+m^8b((=|<3$C=Z@g=!iVye8h
zZj#v@a=(5HSz~i7E*Az}H@S`I!Y{z(*^acd_EU@ewC?k%X?mINroC{KBzM&9+FG}A
z)Q)JFHns42OG4u*SP!uwDJ4v>!TmwneFDBGO<$X2qkrwZL6l+WeEuq~PcMo=GFB<|
z^gSHf;oV+*|5S}+X@`f!k5^d7BbD~3^)vGD^f=NjyC^#mtXQkZ3hbXNwsClUwv1~F
z&pNb$={6(U<>zIp<ae40Yr<Xa(|6o1e$(egA^064r$K3izvNlu-DC4Qy*=`SFds$0
zO9#=L&F@!`%Tn~N#6wLMR^61~RzsmYj3wBVrv=%MP2%fG(zshMTKgrA)f17}`(Bw^
z{C(vQW=-d44mBN=&d$eoLlRBhH27u4R4|2Ff3?ijh0`3HAunf>;Yr@wv|CS-)8<;x
z%Pc!A-x3;!UQT$`&gP(ZKf@5zHPL>`jpMb+s)!;XPj|m5sHwfGpQGs=T4%Dl*(S*;
zMHzk&=oI`zoQbw8#9elkW71AxqnUL<K2qJD=iIR>U4yR0)5*D{WK>b)tBG4mY>#q|
zMt;m|Q@3xcv)E>ez0;hVQ*AFo$Jwv2!Q4oVA-1z%$~{GOiF+zQvl;I1C$Ts)b)rq)
z`T`07uHAN#>^6)8OfMCSWp3+D*&$Y^*@zM|xwybc2U?Jrm36{Lb6fmQf4h5M@dWhZ
zBRW=m<W4?#<k&iU{8mL1@~<6e#fs3A%s&sSVVq|Rg)Y2p{7HKf(>3yBJg?8n8|IOx
zlGhySu5IJP-u`lp;bi?0`Hj@u&Bi+td3|w>th%~`O6wj~e35`6PH0HEhrQNeFC*ZH
zT$mTe$0%~+7<MQ*WXbk{lt|gkh}_8V(&Tp06e$<3?Qs^8nC@QuAdQCCVVMT=S&atd
zBaKS(Z86kRrZtFi29c$g8uq7)ekkeJy#M;6A4<{@w4V)BdXIq8d|`(MAz-|`jt&$w
z&CN(D;6QZ7HH4(r;%qr1Y3EyBMzSbdqUP8K4LS2<-b|iLcgJ+U>W|SRD`%fl6w^y)
zAgVDG_L!MUS2oq#D{aD=DwSguFmY?>UDz2HDOCZ_Mr?<a$piQK9U6}L+?B&ot{{8l
z^UEF0)9FTHe3=z7t)i^_@E#gQQ>s;U?IEFoZ0VA}N=;78K1mYBubd>n;>c#y%Q&sk
zFyeXE9TlWJV{}q`B`iSG)ube1j}wvD4!I)@N;m?F?TZ|R7?X;8HD+T88dY2`cTP^l
zI3>zUm@_QZtTR0&E*;ZyCsoii)f2*Kj3XYcOa9?>8D}$~Dy~#rItv4qa2qH?QZ}yg
zqH_JmjHL+IE^GrANso!7A~J@T8AWA;DD2^TnA7T0GUwUXT=u=y02OB=5xI&E%g?44
zOtZ;U%C{N<RgUweSQ2y{UutqYugE#2mYS63bh;<=O_gcteX;V@Dnf=|!*QR062$2B
zFhbIqC>S%@gwABrxNzQ-8Kgc*+I&X!V!WxI`w%6In$M+jLpiTJ*-Tp@U7}v8s;~84
zNnM3&7R<5vU3y*88_OYeLYbc-pVCf~ayGOOgPv5hsuZVyTzauJT*D?|bSvp%Q@Lcb
zMsVG8N2j|o)-p0_larW)<I1#3t!2u|RHEyP%jqHxvm{hda3LsESD$&NV{#g^#_;u(
zzAv&Ovp6T-DxwVh*p6?0dW}Bf_%+A;4iTADGqIer&tY9VBaI*j$cN#{B*nxWE1j1+
z;Rz!a9dLY!9XXCwl}C-{sd5=ebJUA++*4#35>;u#J-Dl^Lg-xKy{MFCqB`zHic+pI
z0hS`NOf*@;sdV`<DnxY2tlTFcab#J`6<68VT<CkG)7IAla9UPr*RdUTc}oTkUcxPZ
z%<AovbKh_e&>$CXE)5HROheHIAQ!6~$j~w_zGUY^Rc7C}Zc>%mvA*8TAW#dD4{XEJ
z!8RBrbx=LkN+Vf7dwi456kS;5GdPqBaf(ZB1Qe%&v9%E@X-y9%7o>_@U3@005;d_J
z#|f~2T)krir3tJI*-RHXma1<GcZ<>#ao-5!F{HChVL%+5NL&;tDWq6)?2>?D?2w<n
zcsWv{SLN#htvr?F@*-Ul@`g})jXdpc#+I0R6|J^8iCyBcl9EV~V*{lNFOf5Z2O;o*
z%9cjbYK?2w1S~e5^pLuNF`PN-J8pck%pRG{r(li^W2;FrOdJ>YhvUAb2_dlivK(oR
z7gL0u*;-1_qGAeC#cAOrz_h4d6hcR^xN;K?+0#%AUAMkMMLaDoaKe&nQ$^j}hca5Q
z)D^&3rE95P-(4220#(>pj<RDHL&<iSRr@gl5YGfSDnVi*oG5c{j+0EVD=8xOBJ|n;
z5vrMn;7oH0#{#oP+H0*&tVLWsg7S8?A*tW7gJBA0U1_7*ILXfCS{*QX7MUfI`DQaV
z`I+K%B0(wt8qV}H09VqP*&$hd5tZCz86eQUOPI=l5QEUfOb$`66-ZQMjJ=8z&wrZO
z9D|DJm5Lb}7y5}N!Ns3Sh&=Mfmr6s}+Kz}vFTIM1q<Ji;W97y?oA-&a$wRXmZUl>2
zYg}?{*nB?aFx$qDN;(A(K*yre^m0X&pmt799^q7tf3P;|i&SRznwcT~C7$JJno&WG
z(n~|TN!1RX%tZy}kp{n5ki3zTGmNXU7?d`IFtuo8J@uK$HKxYaYBj?RwF`Iii-8P`
zfMc1qKs*QMX<Q$fJUs`Plq9K+g+-s^g{}miSZah~O}T3{7Lg%K{mCRj=_!ac)}~D4
zs^CkpG%gdGrq*(gaUqyqYMC`*DKFDSm8nQQ7pqqGP27Y={YV>w9P$O=fI|UGmIb0x
z3gs%*rm|uE?kyG8in@1S)P2SYyZEA_@{$G4T3azUR*jiQuwmB5VzYv$8<%Q~Ls|0a
z6&TuAVlcgnCqd^8ha2(5+Hqf^NRG)gi8Hl>6pgDTsvEcql8OjX$@E3vZ!>mPtX)b6
zS?JVpmpYNGZb)*NAWMbiJz1y~svvcg%Q$V;_}5z2;|<2Nd$sZ`7iug$O32lhkMEcx
z%1iO6F_mGtLM*!uurxvPr5{TA`uD%~*oTsI!qc1^I5%+T4ahg_okdT%a|7oF&JCO!
zI5%)^;M~BufpY`h26p}8Q%TNsr&Bz!a|7oF&JCO!I5%)^;Q9^3PqN}ge7&YuK0wm7
z@8BFbE21J?{$%_PGoBLX2F?wf8)!G+?>_2SY}Shwz%w^EEUL;6X>h}AyGz@_)5+R=
zRo_9W1&@yEKs9=F?>DE<>p6Gy=I+}$@SE9yes`VUWN+q<eTakVhmwBn`(JnEhmz#^
zwWrwZ_?;fR9v|4<-WqN(k+toCZi6~$Fv8fHU}1vu`Q%fDobr4`w|HQH1Cq8RS8|7s
zeQ00y;l`=Gf9NscyrO88GMC&@L__VXRbG&}S_L>M4F0IGX$qUJ>=nmAv%+m=wuD)t
zWS>h6Fl#y%154BDuEo(hKu>)U!x|4NQIu9)iN=+)d?BBJ{EBiB6-8~3OBG_HYMI5t
z1F#aCtmsKLl@VKJQY%p&(S0t<qgUT~^wHJk)q;bEaF^gSjX{a&WSX0YcUtcdO<(4j
zHN3OjtWAsz{Sz;qg2vG=$rK7hJgy|0t5&m7rb`URef1<B>maFaPR2kVfvHqw)l_;0
zi=P8&&PqHC=PH3!J3Q0!ax(5C*4enBK(>QS@KJ_T<dbgi@;YRhV5f3ixNoi1>8!~G
zdP?Urs4pVK31GNfM1AV3#B}saFOqQKj!#MICnVzv=((H~KO<5qDw@4C){>V21fEVs
zv+&Dt0|eFf(-Cw)OZo{NAi|_8;kXmlj>rZEpAbq=HofLJrHU)nR!)!asB}SfW-*SP
zWEG@4K-Q2sV_c0HAez6j?@wKsa1Od*c<LfWbMPo^*(ae~oPBO$fU;H8g;>uAXHsm-
zaOzoKBOaCGRlCcv2Iw#GGvcWY2l0>M8!hgX$`c_!-JncKu-uM2jZ|bt!{`}a;UvaB
zyt0;PyG?qK=BFy9F1fiW^|NmldMag`_86PxMGUH?WlD<dhvY0GDth&q9$hqKkOjUT
z;N$B;<WzY(2mCyO93~Z;GFw^_zbcvqCV%I0rm7{RrVDsaK#m2jc3QcG{wR%!5+Z><
zMx3BB-dNOw25$(~v#@E-We{ycUg~LVu-U;8H-PX=!Z;Bh^M+z4dQ2p<^~PuDBlGb-
zJ>$-S)liMpd9^zSWL_K9&cId!GV*?^D8F^^ds_Ea$8R=TK$ZTYeQ7|JLtdh~!!Uzu
zxrfn>%nZ)BC_BRADTzw|$V(G)f{aRY?#E?5hqNLMrRbO=6cxG>##bH40;{27`VyS3
z<0Um_hLs<YVb+ZoZL+AmHhdZI5^(eYv)*;}N2`wV=P3`xPi=N+vWdJ7x*}Wx;7RVS
zhW4gJZ!z)qL~jWE4_)N_gp-PLPAcHJxr*OTsK}N`*1pK0cO-?}5r{*>NTB^eL+t&b
zy3~?u?C5-Ox)99Q;fxo>{0~(uFQrtRK>MPM6%K3Ns0StFhZ*K7#8rhrh*hyHPekJK
zMOS`>f>fAFw$VJJ?$ew)6<O*_SME<9x#<T%ipo$z78+-?L})T*y=nvIIg2zEaZma<
zu61lu9gPtaRuNNFb(dntrj@`Vq2>b3OG{12-Q~X8o%g6BI-(b9JE{$Nix}~Xl=5m5
zSMG*%5)ol&2bm^aWv~SBbR0{HwMWy`SA4q0He<Kkl@r`q>vUQjLky=3DS?87X}vnu
zR1#Jbs~Cs`ldLd}Z%LDvn6k!TtHCg1A9PW1J{QNtFjPc1rkuR8=i<h!Ry&Y@%9cD~
z^fIfihI;6fWHMhk$K%*oo+G~05NVu?139c6VXq}|ZeU%ckX45)i*pz1L>z%=0uE>r
znZ?Ju=&%ykOS5vdKBcpqSNnLo+F4*v>U^x7du3maE{ODvE1%BP$f`eT;uVuq+E_w~
zT=G<g7zyQ2TbR5oiTUbsMtY1|5|EkP$Xp{-l2M4O>TEO?dL%d|qP=eQxsMEg-IeC%
z<LFUoAVTEDU=j9}DoglAC+oH9^Qsa~oa<g`F76(K=_w|(=4CRvt^E(v7_*Hd_&LL&
zq4l^#5Eh@rjx3BEc?I+Gh3i$%<VL6;O8Ry0fBlsoN@B^}sPpTH9fL7Ui%fV)rUfYv
z6RFr;NbV9H=1Q%DDkhx62^)GV!cC6B-CVBV&dfrIgE*qkM}xeXIL`@}L`1B*%53qg
zPK`vV!}=ZoMFIc^FgP+~gJBHLb=kN4E%qk8rMLhQgjYv>G~PUU*3zsG6hP-W&72?^
zVisqJQL`mVL2^y@DvNbcW(STa+Y_Ry9`Dg*f7H^bX6G>$d;zQiCWLSy?Sq({TTs};
zjHZ<Q;=#01!XULM18@N2Lrm?O5J|=@nu=fnNB|hkjOf!85<)UeBpKl-GY8pE{0lA7
zffj`dX#Wb1=qOGChim$R9+e3}4#3AJk}m5NHGmzub<84j9Br430mWKCb=eOc&a^29
z-QlVlg%H329ChIxphnRJ4`>2-V@Egzv;e$yAI4d@N=oS(p7$5XQp@ozJtv-z$L#Fs
z0D;0I_)(vI2I{9{POW=%p*zi;N#}^ZYq(RydKAoG;ZbVUW{WfhD6s?IL0Q^e(NQ%}
zH%g=2>%q|^dN?yHZAFB;3vWmbWX&gwm<Y$WAqh({mRP6>C~gu`wMBFROaLSRtmkU7
z4J0jK!X~)Yq6R(Qqhkh`Xj7OCGcbV`u(u$1;Vqhdsz;aNT|lGNKxxmCT{$Zv{Q;g~
z@iUS~M|hek)7^)4mkM2BI6wrzhz@a-K>COeCN&wecCtA(9LYnz7w&Wp3mq=FILL+~
zPDQ5vX+-H3qyp!lqjZ?pw8<HqV_cn>$*qrhNw8EP;dzgVQE`j9x(*wB0B77NX03}_
z07GfWCq0-1i5~7zD8ABHL=KDaMLH9+uRCkhAj?Yt1HfM9vmqkD0qk<opgB!Wx{y1R
zEkUNQ$ph37+AIp1d5C^5d_M3_U8Aur!5Exk!|WU@a51w;c*VBU)HM)mG<XBe)P~YP
z+e3EQyNGncyRkGua@+DY--4rj!Zn=GXoC_fz9OQ6@aQqmlrqb_p4%Z?`%8h9nRUHN
zo?tv7(1CLU{%#l$$GkW*A4X$T5O{b%VuE+@i2`XTaTHHud8s>^oMWp&xaGs{XGRTs
zYvGDaB~u9bFfj?_@Mulpmw;4fV2*R=P2UUkLrK5>{Wt9VP}0BmZD0DAf6`}t>i7TT
zf9?PHn?LDqy*ur`J^aQ${HgEx*8l8pe)8JvYrB8{#E<%dU-}dO?B{;xYq9_0&-r6N
z?wh~t|9W@Yeh)wNy}$c+efKy1<xk$a{msAcU-+J1{bk?#$!TA94`2P6pZ<^hsUQDW
zKRNBWhweG}kN?Kk{e^G&`~SOlug(7V@4n?5KKC2`!n@PH%{~0nfAQb?p+E91|JA#D
z+duo?{)RvK+rR7kJ~{2{*u#(hfv^9mzw2}U+9#)dU3<9q_uZ-YnD?`Duea^>pZn_H
z{smv~ng9Kh(_XiSfAGiuvA_2#|K$Jj$!oVC>z^0=z*qg2f9CIe_22#Ew6cdk{focs
zH~!&2`nNtg?YW1y{2YAu=l+ua^k4qrAA0vJ`!m1ir~Q9_&-eZHPu{Bie}3%m{28D9
z`G5A4)4r`e{Iq}jbAS17{g%J@$xGYU_U~oh$~+(1?@#{VXMg(p|KYFy!FQ+qxc2bU
z|DO0q{{!Fttv~Dce(F2#ZU4<b@caMWKlGRWr%#UBXZ^b4>?6O<`zfFO6F&9BKlTUS
zomTd6{(Iq<{rAG>{?GlNpU*fypK`trobLnY`@s1=FuxBR{r$wrzwY;8_ukvz{jTr+
z*MGw2|Lu3rwm<Xt{eds}F+ctfeR6O6#ozy3U-eZ#=Ci)@rR}|bAN!C0nP2l0|BXNJ
zZSNj!ul@6_m)<k#|LA=^TK}(q>3{f7e$D6nm3OB-_mH21xBBPn===1w;9vCHe#=+<
z<nQ|*-ktV!@8S3Sqrc;C{J1awFFtveJ^%dm{B!6>{yF@A@~uDPH~qmM_^a<ud!Idg
zte^Y4_T1n9$v6I?pYq$j<K1b`J)F<M`F(J{51j7<=lj6xc^|mfpCi6Co$vkgf9p?t
z`#<|P-ktW`!%IE~*Z#Wttn2^2l5;O_b1!f0=Z|yV=lu6VzwclD5x@6qzx*$~JMFoL
zt>@sa{P!y#>zx11-~P5g@u&aj|M>2-=N_K%9GpKtKj-J?4|UJq@zegufBBn#;Cntf
z?c3VJTmSE|U;E#C^4913+28PeU-AF`roZ;?L$fda-?N|nU&r4o@TmLnBYyY?|FQ4*
z=0E@LL$l|fE1Z9>u=VE(=kswsALsLN{`u?C_kr_s-mm*}-sk<1`*VIC>-;>{Yx+Fa
z*MIZB{}cbE-|^?(-P@ku@8|pR`QKmN@4r7g|9g`Dzb85Wd#>}p=X&IS&viau=kxV3
zJYT>5PyIXp^?&r!zyIB7&pn*a!TEh~z7JgYK5+he;r#W&$MEZg^ErR)bN->f|DJyj
z*7^5f<-Z5(d>+o{;d~yB|NVgX{`WQ?y53j*;Gg{7f9T78_><FqTzmME|NJ-p-(UZG
z{<n7@iame6Jb%7Cf4)5a^X2({bAI3C@0<Vm7yK)K{L??~Q{Q>oV|#e1KmWP+-<N;t
zy^jBH!L|3{+rRBc{=heX?q`4JY2Vf!&d&#)pAY;HKOcC$552ATp_jg2kN*FTochoC
z|LvT6KKJ~L|F6^eob}JykNgk*`EUKouleimUdsN4Z~Xhd;1~X=A9{D%a}V?!{LuIQ
z?%(y@-}sk5IqiM-aP+?mKKbY0Cvg6K0_Wc+@LJyoe)FIH%uo0;f8}$&^JBOF)bISl
zulb&z`ZwR5_ShaC_4iMmx*zu6FLvJF^Zx$UZ~2R#oX>$g2mj~a`QzXA`Cs$DzB}#z
z-`<&kMRBBWd_aurn#6eGePATUYZWj?v#}73*<=EV8m|P81<?_+F)J8@S@YK@poSda
zcx*Ht;JGTI5*a*FK@_8LqbO1F1XK<Wlw%NG#HBOq^F;s6F+)#xRrl-j_`MH3FkSVl
zx4u5Ay9yk%=mXy8ndP<o{r<Vqvr)TERBdYO^BhY3T`HOFSJ&?v(Tei6E(GuMjHCSX
z`#3tErr-b$@PTsuK&|tLoj&ewnsqRP)>ivwN0i3S{)s@r0X|S~KTzvBN~jms;hO%i
zlt|?Rt*)Q=sNU-xUGqARD<Hh9RDQts*p^=&`3d^@O|wf0d6L7VZMtLg_L(UZ9C-G@
z)?(jlvm2futmu2keUA^Hd!ItB;}F?>ZgJZ;4^geE+UEg!vVN@J@8NZPP;KiM*0w&i
zy+4V+m*ze%YaG=wFBR&cu6f4n)t<Z_&roh<pI0jB>D;I5m-i?X9BShOd*wBJFQ3Q1
z%fmJejvl&~LZ#*K<aM^zerEN(L#>vJ_4_=wmXFux-%;)z_1-pJq~HJ!w)j9R$5W-=
z3wd(CO3g1+-F{gaZ$vrR8vjLmpgmZ69)Wc~@^=4b^3YiwJ1)6Oq0(|V>OExAwzv`s
zm4<`8=E39bJyxa4D^*@&J%6jMd^`UXn-X~A7*(s{xFzjz%UYfo@2J^$hg|PHD0O_O
zZohc$LD}aXz#sh8I8Tt)eEn;;>wXt+6i{ANZR;ggcOM;}E4z*Y>-?=fzh?dS|C+S;
z=Phn|1S&5FwdGZ{)Q9)?{84_s=4+_#x?Ag?Yh^x8)KgpYucE!t-g3@2-08X8Ivb~y
zQP!2$eu3*Jd9R<qTOQt6&l2ld^0=<)lh1AA=WkbS%1Ni-piK^PJ->kc*fu{TxY>tW
z9u3&k_2ZpHD~jiprFmXiDvlox+R-*-ND+a81AKsffIMJZ9_Z@WUiYxo4Z@4E@?6`K
z^K9d1@%hK);-Z!J^((9Al9zOrf5u>=pTSk{r}x)8gd2w$^^QU`G&D3eG}G%sjRt4o
z=Pr5&y@U7I=fjMFpE~Ju3%zn(9RBwO5Z(zqYDrd<cb-}+KP$eeZ8@~oZm2H^fB*=9
zfQ3M9k01DZJ~+bfAOHd&00Mjxz;V{nZr}h80s;5{K9Iu)7RvHP_!_>J!`H0(Oun9`
zu065Zo+uacnKk)L>0Zg2UzO&ewf(c!{wR;df#3iR0s;5{K7bF<50D3t2apGl2YAf`
zmfkCW<dql7jeIXhzL&p8VvitTKLW^`7PjC}I}Y}9D83m4<R^f9U|}v52l-o&?>$Pz
z9p59ZkHHf>1p@E^d;lM4w;x#Cj|c)F00JNY0w4ea>=3~9iKQ*Tp=u87I2I)V0T2KI
z5C8!X009sH0T2KI5C8%83E)11rH#M=90UUJ0eoPK4=mIWK>!3m00ck)1V8`;KmY_l
z00ck)1gau{`z{tb-~bK+0r&tufDh0Ql*$8^-iN4F0!sN#uKjAYT~Tik009sH0T2KI
z`3T5$UL)6j<RgbYfdB}AfI11tm9N#dA9cC`wE_VUkb{8Q@*(PDagEkF$kBJO4-fzW
z5C8!X009dDt)F+G{_uf7K<z%Tv@#+HfB*=9fLaKs{kUbX`l+S;Q4bIR0T57%fW7iF
zzSq*C-~bK+0r&tuV9^JbR?$vGKSV!du^+;h@FjeS^CaW}<N@RX<N<r-0ZZ>f1OX5L
zfyxQkEC0y-y_MGxpMU@efB*<cK|t=jj{R9$4IIEhAOIh*<pWD;5FZhMKiTpp+5_z&
z5P%Qh1NZ>_0C@m;0C_-rdBEa&M5zfNpIV550~;Kq_BSiv2Y<2QFSG;NK_CDhzz6UF
z`T_C)@&NJx@&NJx@&NJx@_<r#z|#8=L4Yj+O7#cqmn|Q#UK+Xl!a4_(4+KB}1VBI=
z1mrsIVLui>fde=Q1myC8g*GAxfB*=9fZPP+%JaPLNAC8({y+c(KmY_l00ck)1V8`;
z_#?pU^$yBy@ftXQgFpa2fDhmU^aJDp<N@RX<N@RX<N+S?fTfiXLEv!$TF+lg-2cbJ
z7;N-2xa$4&6ZH<^#$iUiqYw=Z4UG-W^tw=^!CCmZi{3%+=souNFr#7hLa!6AIqE*n
ziGH=fluw~G<KP~<tYul9{e(3Q>H1^+|D>c5R&?=#XD=n(%Al<1!^7rRip)6_D$aow
z-_*7otoTjq<>(U}e06uvTmrQ(huZcR{Qcv{&F(QPe42TcP$-`q+`OZIz4Yo{Dz!#T
zadIgwD<n`hI7r>T_&#a+4bHE{&#%EnU0iV9^!Ry`wLEy~5A@x`S*EF(L~Ckod#e5S
zGn*I|r|rB!py0r>59GeCW$iw>wdKL}9vjzt*0uvH_Z@LvY1?(>$VophZ%|T5)S_d8
zV;qk;WfQ2jIMi_+{9Ig{gH$O!5o;dd8fPL>YfO5%<HSP+#9IA(lSAH1GYTnG9uARX
za)QR3D<n`0hh6SLsbwAW2^1XQ1M~yr0ptPX0ptPX0XFhLH)E`4aZoOS^2MQjQTOkA
z-Y%hXYf({j(y~K!Jf&8tua}knzVMKaCziG^BKFp;i#uk=7vvLbRcStz?SIhxvF(mc
zUO;%&DF4ut2mK2u6dZo<gEtzzb#DHS%ak`g8yHo-`oRgps<JZfgmP5<+<nC9hQ`e=
z9wu_>!cBkaZ+6S1P;jt^4_Ldu|8C!PJ-2105ni?5$pfaK5or`!GY;0C@8kDvyG{Z(
zaI@w58oq&V;2Z1jOR1&bX!UuUsHAx>w#v<+P;gMv2cwo0ItJFyq)>2x56}--%L7Wk
zr=a%t2T{M;y)XQTpH|Ofd8zlzPwR!bFWE<+;2?(&<bIzO`{Q$7sLI+6<2yAQ-(igt
zYv<3_mIw3g)@;6AaC)2*FZ1T0oP5rA^+M+{Aw%+ip=wjD&&T^@{a4o?`HW|eLIMQ`
z_yGL?dBC<jfb-MG&riXFO&+}5KSnv_nHNTB&h~om$lCe+@cV48XM}cPqg|AKE`Zhj
z-*g$V);#APg@OZofPR2HU|Sx*d&*V6rwl&egLU+<j$n0p0P{*~YhG!4wKr?e`Jg=V
zKA%-xp5guTZOMLhqw3$GP;h_`&<~IYY|8_<KX2Xrd8NnOuzzL7Rn;<o#G`sV`5YmB
ze)dwrtqckd-~b<>AJjAtypuR7u<VqXK&|7T)Hs{Ho>#*6@;MJR^&aQrzH}$$MW4UY
zu--SvN~qQq{f*E5hIXsE-FV4EC?^}o4ZNqp=6f1w7qm-(DWAdzeE1+@ijzxeSs_tM
zAI*yfm1cf9-Z?7eL$&W0&mqbC91?hgH`d>=-}*!J-NRX?shLFU`e<AmYTf!MTECBs
z`eQr+<0%+Vs%;)PT-@Rp-O3CC1qUtq0OLW*j0aV=-aYTz%a{L>YRtMvcu}oAN6*{)
zW30|UgK|axdg;}@6db?-K48BesP#Bk)91ab`#s*r9Z`N|$Dc7?#>aSBuG{ptjWPEq
z6dd4#%KL%VpMO)!xGdHIQg$66t@oqq>W}j#KF*uaPF1%PYxld=dcBW&SvT(ucva1d
z_4_8T-#h0rb6g4GO{=@#rnS#`8+Su_G4D0zy~e!PQs)7+zo&%yDf>Pq_=A5{?*p@b
zzK`<BJC55uW`$2PuM!HCmV;WxPi(CRjzer6hpHQg;PrFd->drm9{7MyS)KhvZ9dS-
zd7{+!A8h?TR-bd{ZCn85<#Rj-?S^*aaU8OyzGXGPS5_`tuS@ZFRnNPF>vn0c+wFDU
zZR>fQ*8Wba`x=$yA#AtL`=3&^8>`=g`|Z-+Z?~2w-V3PNdjWFYXJNHH>~;Lc_u{^+
z8t==(*YLHNujM+9NVOlA_}rDf6LX1L6!V-ZGtZe?xStJ-DqsEJ1W}7({sU#^KVbDd
zALXjq^&mKbQ;|7`*476YXQ+Cd0q>hv{k}Q)fDhJZvETYE=wIkxHR)ftj#B123f%3>
zoz-z;l#7pbS(R$1UG71tWgYVg6dd4#e8T&BA}GIkgln7$?HFqvhu7kr_YV0wn`lMx
zTzA#ab<54?%`5X0TJ7%eBNQCq1M~yr0o(FGZO7T<{(DyL8_Qj9?62x^ajma!9{Dh%
zWsmiDDYRxB^3DwEX-wKdS=09%cAY)i>i}UzUFyzU`iH0j%8HH*I&$UQ_=A*pO$|C9
zlGQYiO09Ko1P*?`OBUr_gAW$B-8Azwg(}M->#J{uw;YvBp!VfpTIk&V-IwwxZ(6*4
zZsykHRLYv_@Q-T(FZ|<^{_`%BP^mRFa!b>{$D~mx8yvVcsfr7~;qvwuBlHId6dcs-
z0~L2+S*ljA)4rH#9j;T>)pN$WtkIuerCL*|?=SHFedXcQc<4|Pfr10iK3KW^`mA64
zOQ~EsKICD*qS6hNS1pge+ig!y8iDe~;gi+gQ7ux=P^ES3*Kr<C-7KO|>p1l9yXJ>s
zsRxK!biuHziEEo5Ay92`@ck&(WnPn0L@71ReKuj-kvmjs9hY*vjqA7B#9D0@pQ1}@
zkWZoVaA<zz!kim#<r1ic!@!l>eNzwrn?k_>K0rS}9zY&I9zY&I9$+I6eAmtIPe!ln
z1j-kO+~(OYH1jB;a%<_Lb0gX|K1HS0wq4t~EFW}-Dy>ot#tzS{7vE}M9<{ehRbE!U
z=h~h5y*wQ*5?=K{)`>QeFJ7ZiaQM9swvP{9J8*su;Z67ZPVRQxxsbA|tmG|}qiO>0
z_@3zJU3{L(rB#<k{hvMva6kIxgft4Z4~Gv%oeP;6dY!PM9w(RXSlTg}@T%bhSEbdR
zc%49N#v!WR8r`JLbv&Wq03V<qAP*o9AP=yU2jYjOP%<EsYE4gj&MR-*;tqv^gOWaQ
zd1g-egYrBI1qUtrASu5?_vf;6DU>Y^Kiwpax+i83rL_O>vbiJ2QUV2ss(cW-EzH5|
z=>iI6lSAv-Q%AQxA4Q<xAcqf1BX<mG*C(4o!NC?Elyx5Ztoy+N3gwGKz*qCrXKqcQ
zysCpwgM<Nb*D2IK9QL$L$Tl1*rfSoV1{|K*aBC@nf`dJL@W!0id~)A6Q?)6M?~fnf
zZw&VO^ZlYj1PTuD0r~;*fNgo8v+F3&j1ffy3Jxs$pmdgL<u^xe5hysY;Dd&7e@u1m
zvyV!xysz(q@~dN=9<{XNTXkpu^nIry0)3oA`HBv&Bp%EpP;h_`&<~IYY|8^!Cj{$+
zsCk_b_#6I~!{4{3IrnzBQ%0cRP#Yf{cAU`RRIAMds!a}t^WC@hDKrwbYD(jlP2=K=
zDHI&w1M~yr0o(FGaf82%@a<GYp*7>sdwJHWhELz5P;GHYF!fmd*~}|MEsFK8*;xM?
z?SOW`dh}_O*M2ay*YSttB#A(+<FI)|-EA9+Z&10l)o6zc(~5MSP;d}@;Ig)TO5?B0
zlovhF_s7*_QW4d<?rM;*Bs?;oLcu{zKIpY{Zqb<8W(oy|n)qNw_MF_y7m^7S9N+`=
z1LOhQ@_=cJtJ|#sH;GcZDXhh%|8v<uq2M6+V0OCGXD2fX2~=7R^-hItiC<hmq2K@?
zpdav_2c}0GhJVyJk!oE{Ee8B_y<0JXTF1f1e`8l4=i3wt4)6i`0rG%tc_8o1pq|F0
z9TW-<TJ(X}+dp4U3%f(0@^W}C{cjQP?7vCXro7El5*3}>bK!sPP_1Zr^xbZIa?%Jd
zdTjTwHM$X}C=?vv1M~yr0o(Gx`ppMS6LPK+wdth^9>inW9a>vWZ1Z4Nr@<Eq6dY{v
z!IzP(*WDlO<OyY)L*&DZmOa+rrBHBy4=V2m=>@OU*`9cXYE}DpZC-zZ`Fp~<#tx>3
z|LFB^3RQ-~S5pTLYtj5Bg^F|d^x?%hBQIoAC^+!!gT0Hxn%q8gm9U~(xt}d{J;%RP
z_w(4ZyQpS0<YB;~(hYRHMs(BSd9zwPC?c%tv<r=j2Ya6-yeT=lIG`Raq)-coxCLi?
zXFY8qP;mIY4|o}eMLGEzH)plIQm4MzcI8Lq1PTt;_yF@^YjZwqsoPy`<KL)H?Vf9B
z7N4R^YLHK%(sJ;d-nh{R8*fpqYLCL#C)dB2O`vRW_{KNgDb=T(s!fNdm+L#fT|lAW
z03T?lA9PRdopWi$O{#WH)~!uwZ#qGh*4E+8pZW8x63Uv!4_x=uF7tT;1qb*5{Q!Bu
zwmfia!F#Dwepo|z(c1@~ir?P-27yY;A$;Jfw7L_o6R0#C5>glSHGiB!rB*!8TC?X_
zrFvg3wspEs({srbDi4QxKe+jZo-8G5)x15+Xw&a*Q>Zc=zW#R9Nv~1a1X?wRPnxx*
zt5WV#duwXY`H-xpd6YH9dIng}peE}Xz$fqtd{X)PFO$5dhF&n|Q(iRssl72ykvE81
zlo#$OC!77`gHh)~W`<rTYSH4|qfc(jzDl9;a(H9TYd*Q}n<<nn4!ECKv-^qQBn>Cy
z{1(RhAD9VLoWsV_rw=&luTdyCzz66D$OE?Jf#U6RGq)zEQr0zoXbL3*GAVDmF6h_S
z`^-B-q2Qn<ADAw`-16?e3ls_tJo&(XZI{TUxj6(14)8$^`L0dy7p9Xtq2Qn?A1oMl
zHF0h8BSb4YxS!LuR|0MkC>tCSySMj<i^!r-aDWfc50D3J%LARi$Ou0gbCvR<QPC6n
zzkINqD6LZ#?a7NsG*c)W98Qm)bt=DEI?<{w`Bz=tZ~u(JMn8kAexiPo-XYvL%&2!1
zVz6<nv5Udq5HPz!CL0<Xn(1|+MuW5PSr@&7ep1B`A1<u;oxmkdI$hvWAuJQZav}Uv
z2;T}}g%DN>;X5I$62fXBtP#RmAw&xyMhNSKuwDoogs@Qvn}o1g2>%ko79ngE!Zsnq
z3Sqkt;)Jk62s?$aO9=5o_+ALRg^(ZwlMoVxutx}cg|JTu`-Sj>5U3D-6he{^4hSJx
z2nU65NC+tvA)>MOl%Sv~A5MPb!--S;{~A1P`uN^Pk&SL*8{Ne=dWdc8BewA+v5kGj
zHoh#jv7a%b$^S2WFVU?PbrSjAib{!Wt*DjA){1J0Y^|u5$kvLAiEOQ?naI|Ps)=kB
zt(%)@-P}a$<|bM<H_^JeiPp_cv~F&qb#oJ~o119e+(hf<E?PHt(Ym>d*3DhCZtkLW
za~G|fyJ+3qMeF7+S~qvmx_OA!%|o<q9-?*g5UrbsXx%(S>*gU^HxJRedAw{4yelM#
zQX$+E!u^U6AmrE?J~P59vS&q7oh2mI89s)8=yitK-eaFPe5QB!?X!U2eqo+rz9G~Q
WW(c2cG%T$6|HXo!8GbWn^!zWIX7f1!


From 843c7e8adc6c7936ecce699aba55610991e9e73c Mon Sep 17 00:00:00 2001
From: youssefmecky96 <Youssef.mecky96@gmail.com>
Date: Fri, 8 Mar 2024 14:02:41 +0100
Subject: [PATCH 133/142] cleaning up and formatting

---
 environment.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/environment.yml b/environment.yml
index b48befc7..d081b46d 100644
--- a/environment.yml
+++ b/environment.yml
@@ -28,8 +28,8 @@ dependencies:
   - pip=23.1
   - einops=0.6.1
   - hydra-core=1.3
-  - quantus=0.5.3
   - pip:
+      - quantus==0.5.3
       - recipies==0.1.3
       # Fixed version because of NumPy incompatibility and stale development status.
       - scikit-optimize-fix==0.9.1

From fe62ee281f5860fe106d9dd9d9818553270119c4 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <Youssef.mecky96@gmail.com>
Date: Fri, 8 Mar 2024 14:38:13 +0100
Subject: [PATCH 134/142] formatting

---
 icu_benchmarks/imputation/diffwave.py   |   2 +-
 icu_benchmarks/models/custom_metrics.py |  30 +-
 icu_benchmarks/models/utils.py          | 379 ++++++++++++++++-
 icu_benchmarks/models/wrappers.py       | 540 ++----------------------
 scripts/plotting/plotting.py            | 192 +++++++++
 5 files changed, 615 insertions(+), 528 deletions(-)

diff --git a/icu_benchmarks/imputation/diffwave.py b/icu_benchmarks/imputation/diffwave.py
index 458c2c5e..437303ed 100644
--- a/icu_benchmarks/imputation/diffwave.py
+++ b/icu_benchmarks/imputation/diffwave.py
@@ -330,7 +330,7 @@ def forward(self, input_data):
         cond = self.cond_conv(cond)
         h += cond
 
-        out = torch.tanh(h[:, :self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels:, :])
+        out = torch.tanh(h[:, : self.res_channels, :]) * torch.sigmoid(h[:, self.res_channels:, :])
 
         res = self.res_conv(out)
         assert x.shape == res.shape
diff --git a/icu_benchmarks/models/custom_metrics.py b/icu_benchmarks/models/custom_metrics.py
index fd88699f..0cff08c0 100644
--- a/icu_benchmarks/models/custom_metrics.py
+++ b/icu_benchmarks/models/custom_metrics.py
@@ -422,30 +422,30 @@ def norm_function(arr):
             x_original = dataloader.dataset.data["reals"].clone()
 
             dataloader.dataset.add_noise()
-            x_preturb = dataloader.dataset.data["reals"].clone()
-            y_pred_preturb = model.model.predict(dataloader)
+            x_perturb = dataloader.dataset.data["reals"].clone()
+            y_pred_perturb = model.model.predict(dataloader)
             Attention_weights = model.interpertations(dataloader)
-            att_preturb = Attention_weights["attention"]
+            att_perturb = Attention_weights["attention"]
             # Calculate the absolute difference
-            difference = torch.abs(y_pred_preturb - y_pred)
+            difference = torch.abs(y_pred_perturb - y_pred)
 
             # Find where the difference is less than or equal to a thershold
             close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
 
             RIS = relative_stability_objective(
                 x_original.detach(),
-                x_preturb.detach(),
+                x_perturb.detach(),
                 attribution,
-                att_preturb,
+                att_perturb,
                 close_indices=close_indices,
                 input=True,
                 attention=True,
             )
             ROS = relative_stability_objective(
                 y_pred,
-                y_pred_preturb,
+                y_pred_perturb,
                 attribution,
-                att_preturb,
+                att_perturb,
                 close_indices=close_indices,
                 input=False,
                 attention=True,
@@ -458,16 +458,16 @@ def norm_function(arr):
             with torch.no_grad():
                 noise = torch.randn_like(x["encoder_cont"]) * 0.01
                 x["encoder_cont"] += noise
-            y_pred_preturb = model(model.prep_data(x)).detach()
+            y_pred_perturb = model(model.prep_data(x)).detach()
             if explain_method == "Random":
-                att_preturb = np.random.normal(size=[64, 24, 53])
+                att_perturb = np.random.normal(size=[64, 24, 53])
 
             else:
-                att_preturb, features_attrs, timestep_attrs = model.explantation2(x, explain_method)
+                att_perturb, features_attrs, timestep_attrs = model.explantation2(x, explain_method)
 
                 #
             # Calculate the absolute difference
-            difference = torch.abs(y_pred_preturb - y_pred)
+            difference = torch.abs(y_pred_perturb - y_pred)
 
             # Find where the difference is less than or equal to a thershold
             close_indices = torch.nonzero(difference <= thershold).squeeze()[:, 0].to(device)
@@ -476,15 +476,15 @@ def norm_function(arr):
                 x_original.detach(),
                 x["encoder_cont"].detach(),
                 attribution,
-                att_preturb,
+                att_perturb,
                 close_indices=close_indices,
                 input=True,
             )
             ROS = relative_stability_objective(
                 y_pred,
-                y_pred_preturb,
+                y_pred_perturb,
                 attribution,
-                att_preturb,
+                att_perturb,
                 close_indices=close_indices,
                 input=False,
             )
diff --git a/icu_benchmarks/models/utils.py b/icu_benchmarks/models/utils.py
index 6c944ae7..34a5de0b 100644
--- a/icu_benchmarks/models/utils.py
+++ b/icu_benchmarks/models/utils.py
@@ -8,13 +8,14 @@
 import logging
 import numpy as np
 import torch
-
+from quantus.functions.similarity_func import correlation_spearman, cosine
 from pytorch_lightning.loggers.logger import Logger
 from pytorch_lightning.utilities import rank_zero_only
 from torch.nn import Module
 from torch.optim import Optimizer, Adam, SGD, RAdam
 from typing import Optional, Union
 from torch.optim.lr_scheduler import _LRScheduler, CosineAnnealingLR, MultiStepLR, ExponentialLR
+import captum
 
 
 def save_config_file(log_dir):
@@ -188,3 +189,379 @@ def version(self):
     @rank_zero_only
     def log_hyperparams(self, params):
         pass
+
+
+def Faithfulness_Correlation(
+    model,
+    x,
+    attribution,
+    similarity_func=None,
+    nr_runs=100,
+    pertrub=None,
+    subset_size=3,
+    feature=False,
+    time_step=False,
+    feature_timestep=False,
+):
+    """
+    Calculates faithfulness scores for captum attributions
+
+    Args:
+        - x:Batch input
+        -attribution: attribution generated by captum,
+        - similarity_func:function to determine similarity between sum of attributions and difference in prediction
+        - nr_runs: How many times to repeat the experiment,
+        - pertrub: What change to do to the input,
+        - subset_size: The size of the subset of featrues to alter ,
+        - feature: Determines if to calcualte faithfulness of feature attributions,
+        - time_step: Determines if to calcualte faithfulness of timesteps attributions,
+        - feature_timestep: Determines if to calcualte faithfulness of featrues per timesteps attributions,
+    Returns:
+        score: similarity score between sum of attributions and difference in prediction averaged over nr_runs
+
+    Implementation of faithfulness correlation by Bhatt et al., 2020.
+
+    The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
+    (or 'fidelity') with respect to the model behaviour.
+
+    Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
+    the average explanation attribution for only the subset of features are (linearly) correlated, taking the
+    average over multiple runs and test samples. The metric returns one float per input-attribution pair that
+    ranges between -1 and 1, where higher scores are better.
+
+    For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
+    or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
+    test point and the average explanation attribution for only the subset of features is calculated. Results is
+    average over multiple runs and several test samples.
+    This code is adapted from the quantus libray to suit our use case
+
+    References:
+        1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
+        explanations." IJCAI (2020): 3016-3022.
+        2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for
+        responsible evaluation of neural network explanations and beyond."
+        Journal of Machine Learning Research 24.34 (2023): 1-11.
+    """
+
+    attribution = torch.tensor(attribution).to(model.device)
+
+    # Other initializations
+    if similarity_func is None:
+        similarity_func = correlation_spearman
+    if pertrub is None:
+        pertrub = "baseline"
+    similarities = []
+
+    # Assuming this is a method to prepare your data
+
+    y_pred = model(model.prep_data(x)).detach()  # Keep on GPU
+    pred_deltas = []
+    att_sums = []
+
+    for i_ix in range(nr_runs):
+        if time_step:
+            timesteps_idx = np.random.choice(24, subset_size, replace=False)
+            patient_idx = np.random.choice(64, 1, replace=False)
+            a_ix = [patient_idx, timesteps_idx]
+
+        elif feature:
+            feature_idx = np.random.choice(53, subset_size, replace=False)
+            patient_idx = np.random.choice(64, 1, replace=False)
+            a_ix = [patient_idx, feature_idx]
+        elif feature_timestep:
+            timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
+            feature_idx = np.random.choice(53, subset_size[1], replace=False)
+            patient_idx = np.random.choice(64, 1, replace=False)
+            a_ix = [patient_idx, timesteps_idx, feature_idx]
+
+        # Apply perturbation
+        if pertrub == "Noise":
+            x = model.add_noise(x, a_ix, time_step, feature, feature_timestep)
+        elif pertrub == "baseline":
+            x = model.apply_baseline(x, a_ix, time_step, feature, feature_timestep)
+
+        # Predict on perturbed input and calculate deltas
+        y_pred_perturb = (model(model.prep_data(x))).detach()  # Keep on GPU
+
+        if time_step:
+            if attribution.size() == torch.Size([24]):
+                att_sums.append((attribution[timesteps_idx]).sum())
+            else:
+                att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :]).sum())
+        elif feature:
+            if len(attribution) == 53:
+                att_sums.append((attribution[feature_idx]).sum())
+            else:
+                att_sums.append((attribution[patient_idx, :, :][:, :, feature_idx]).sum())
+        elif feature_timestep:
+            att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :][:, :, feature_idx]).sum())
+
+        pred_deltas.append((y_pred - y_pred_perturb)[patient_idx].item())
+        # Convert to CPU for numpy operations
+
+    pred_deltas_cpu = torch.tensor(pred_deltas).cpu().numpy()
+    att_sums_cpu = torch.tensor(att_sums).cpu().numpy()
+
+    similarities.append(similarity_func(pred_deltas_cpu, att_sums_cpu))
+
+    score = np.nanmean(similarities)
+    return score
+
+
+def Data_Randomization(
+    model,
+    x,
+    attribution,
+    explain_method,
+    random_model,
+    similarity_func=cosine,
+    dataloader=None,
+    method_name="",
+    **kwargs,
+):
+    """
+
+    Args:
+        - x:Batch input
+        -attribution: attribution
+        - explain_method:function to generate explantations
+        - random_model: Reference to model trained on random labels
+        - similarity_func: Function to measure similiarity
+        - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+        - method_name: Name of the explantation
+
+    Returns:
+        score: similarity score between attributions of model trained on random data and model trained on real data
+
+    Implementation of the Random Logit Metric by Sixt et al., 2020.
+
+    The Random Logit Metric computes the distance between the original explanation and a reference explanation of
+    a randomly chosen non-target class.
+    This code is adapted from the quantus libray to suit our use case
+
+    References:
+        1) Leon Sixt et al.: "When Explanations Lie: Why Many Modified BP
+        Attributions Fail." ICML (2020): 9046-9057.
+        2)Hedström, Anna, et al. "Quantus: An explainable ai
+        toolkit for responsible evaluation of neural network explanations and beyond."
+            Journal of Machine Learning Research 24.34 (2023): 1-11.
+
+    """
+
+    if explain_method == "Attention":
+        Attention_weights = random_model.interpertations(dataloader)
+        attribution = attribution.cpu().numpy()
+        min_val = np.min(attribution)
+        max_val = np.max(attribution)
+
+        attribution = (attribution - min_val) / (max_val - min_val)
+        random_attr = Attention_weights["attention"].cpu().numpy()
+        min_val = np.min(random_attr)
+        max_val = np.max(random_attr)
+        random_attr = (random_attr - min_val) / (max_val - min_val)
+        score = similarity_func(random_attr, attribution)
+    elif explain_method == "Random":
+        score = similarity_func(np.random.normal(size=[64, 24, 53]).flatten(), attribution.flatten())
+    else:
+        data, baselines = model.prep_data_captum(x)
+
+        explantation = explain_method(random_model.forward_captum)
+        # Reformat attributions.
+        if explain_method is not captum.attr.Saliency:
+            attr = explantation.attribute(data, baselines=baselines, **kwargs)
+        else:
+            attr = explantation.attribute(data, **kwargs)
+
+        # Process and store the calculated attributions
+        random_attr = (
+            attr[1].cpu().detach().numpy()
+            if method_name in ["Lime", "FeatureAblation"]
+            else torch.stack(attr).cpu().detach().numpy()
+        )
+
+        attribution = attribution.flatten()
+        min_val = np.min(attribution)
+        max_val = np.max(attribution)
+        attribution = (attribution - min_val) / (max_val - min_val)
+        random_attr = random_attr.flatten()
+        min_val = np.min(random_attr)
+        max_val = np.max(random_attr)
+        random_attr = (random_attr - min_val) / (max_val - min_val)
+
+        score = similarity_func(random_attr, attribution)
+    return score
+
+
+def Relative_Stability(
+    model,
+    x,
+    attribution,
+    explain_method,
+    method_name,
+    dataloader=None,
+    threshold=0.5,
+    **kwargs,
+):
+    """
+    Args:
+            - x:Batch input
+            -attribution: attribution
+            - explain_method:function to generate explantations
+            - method_name: Name of the explantation
+            - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
+
+
+        Returns:
+            RIS : relative distance between the explantation and the input
+            ROS: relative distance between the explantation and the output
+
+
+    References:
+            1) `https://arxiv.org/pdf/2203.06877.pdf
+            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation
+            of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
+
+    """
+
+    def relative_stability_objective(x, xs, e_x, e_xs, eps_min=0.0001, input=False, device="cuda") -> torch.Tensor:
+        """
+        Computes relative input and output stabilities maximization objective
+        as defined here :ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
+
+        Args:
+
+            x: Input tensor
+            xs: perturbed tensor.
+            e_x: Explanations for x.
+            e_xs: Explanations for xs.
+            eps_min:Value to avoid division by zero if needed
+            input:Boolean to indicate if this is an input or an output
+            device: the device to keep the tensors on
+
+        Returns:
+
+            ris_obj: Tensor
+                RIS maximization objective.
+        """
+
+        # Function to convert inputs to tensors if they are numpy arrays
+        def to_tensor(input_array):
+            if isinstance(input_array, np.ndarray):
+                return torch.tensor(input_array).to(device)
+            return input_array.to(device)
+
+        # Convert all inputs to tensors and move to GPU
+        x, xs, e_x, e_xs = map(to_tensor, [x, xs, e_x, e_xs])
+
+        if input:
+            num_dim = x.ndim
+        else:
+            num_dim = e_x.ndim
+
+        if num_dim == 3:
+
+            def norm_function(arr):
+                return torch.norm(arr, dim=(-1, -2))
+
+        elif num_dim == 2:
+
+            def norm_function(arr):
+                return torch.norm(arr, dim=-1)
+
+        else:
+
+            def norm_function(arr):
+                return torch.norm(arr)
+
+        nominator = (e_x - e_xs) / (e_x + (e_x == 0) * eps_min)
+        nominator = norm_function(nominator)
+
+        if input:
+            denominator = x - xs
+            denominator /= x + (x == 0) * eps_min
+            denominator = norm_function(denominator)
+            denominator += (denominator == 0) * eps_min
+        else:
+            denominator = torch.squeeze(x) - torch.squeeze(xs)
+            denominator = torch.norm(denominator, dim=-1)
+            denominator += (denominator == 0) * eps_min
+
+        return nominator / denominator
+
+    attribution = torch.tensor(attribution).to(model.device)
+    if explain_method == "Attention":
+        y_pred = model.model.predict(dataloader)
+        x_original = dataloader.dataset.data["reals"].clone()
+
+        dataloader.dataset.add_noise()
+        x_perturb = dataloader.dataset.data["reals"].clone()
+        y_pred_perturb = model.model.predict(dataloader)
+        Attention_weights = model.interpertations(dataloader)
+        att_perturb = Attention_weights["attention"]
+        # Calculate the absolute difference
+        difference = torch.abs(y_pred_perturb - y_pred)
+
+        # Find where the difference is less than or equal to a threshold
+        close_indices = torch.nonzero(difference <= threshold).squeeze()
+        RIS = relative_stability_objective(
+            x_original[close_indices, :, :].detach(),
+            x_perturb[close_indices, :, :].detach(),
+            attribution,
+            att_perturb,
+            input=True,
+        )
+
+        ROS = relative_stability_objective(
+            y_pred[close_indices],
+            y_pred_perturb[close_indices],
+            attribution,
+            att_perturb,
+            input=False,
+        )
+
+    else:
+        y_pred = model(model.prep_data(x)).detach()
+        x_original = x["encoder_cont"].detach().clone()
+
+        with torch.no_grad():
+            noise = torch.randn_like(x["encoder_cont"]) * 0.01
+            x["encoder_cont"] += noise
+        y_pred_perturb = model(model.prep_data(x)).detach()
+        if explain_method == "Random":
+            att_perturb = np.random.normal(size=[64, 24, 53])
+            att_perturb = torch.tensor(att_perturb).to(model.device)
+        else:
+            data, baselines = model.prep_data_captum(x)
+
+            explantation = explain_method(model.forward_captum)
+            # Reformat attributions.
+            if explain_method is not captum.attr.Saliency:
+                att_perturb = explantation.attribute(data, baselines=baselines, **kwargs)
+            else:
+                att_perturb = explantation.attribute(data, **kwargs)
+
+            # Process and store the calculated attributions
+            att_perturb = (
+                att_perturb[1].detach() if method_name in ["Lime", "FeatureAblation"] else torch.stack(att_perturb).detach()
+            )
+        # Calculate the absolute difference
+        difference = torch.abs(y_pred_perturb - y_pred)
+
+        # Find where the difference is less than or equal to a threshold
+        close_indices = torch.nonzero(difference <= threshold).squeeze()
+        RIS = relative_stability_objective(
+            x_original[close_indices, :, :].detach(),
+            x["encoder_cont"][close_indices, :, :].detach(),
+            attribution[close_indices, :, :],
+            att_perturb[close_indices, :, :],
+            input=True,
+        )
+        ROS = relative_stability_objective(
+            y_pred[close_indices],
+            y_pred_perturb[close_indices],
+            attribution[close_indices, :, :],
+            att_perturb[close_indices, :, :],
+            input=False,
+        )
+
+    return np.max(RIS.cpu().numpy()).astype(np.float64), np.max(ROS.cpu().numpy()).astype(np.float64)
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index 79f13ef4..ae70406a 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -18,8 +18,7 @@
 from pytorch_lightning import LightningModule
 from icu_benchmarks.models.constants import MLMetrics, DLMetrics
 from icu_benchmarks.contants import RunMode
-import matplotlib.pyplot as plt
-from quantus.functions.similarity_func import correlation_spearman, cosine
+from icu_benchmarks.models.utils import Faithfulness_Correlation, Data_Randomization, Relative_Stability
 import captum
 from captum._utils.models.linear_model import SkLearnLasso
 
@@ -522,118 +521,6 @@ def prep_data_captum(self, x):
         )
         return data, baselines
 
-    def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir):
-        """
-        Plots the attribution values for features and timesteps.
-
-        Args:
-            - features_attrs: Array of feature attribution values.
-            - timestep_attrs: Array of timestep attribution values.
-            - method_name: Name of the attribution method.
-            - log_dir: Directory to save the plots.
-        Returns:
-            Nothing
-        """
-
-        # Plot for feature attributions
-        x_values = np.arange(1, len(features_attrs) + 1)
-        plt.figure(figsize=(8, 6))
-        plt.plot(
-            x_values,
-            features_attrs,
-            marker="o",
-            color="skyblue",
-            linestyle="-",
-            linewidth=2,
-            markersize=8,
-        )
-        plt.xlabel("Feature")
-        plt.ylabel("{} Attribution".format(method_name))
-        plt.title("{} Attribution Values".format(method_name))
-        plt.xticks(
-            x_values,
-            [
-                "height",
-                "weight",
-                "age",
-                "sex",
-                "time_idx",
-                "alb",
-                "alp",
-                "alt",
-                "ast",
-                "be",
-                "bicar",
-                "bili",
-                "bili_dir",
-                "bnd",
-                "bun",
-                "ca",
-                "cai",
-                "ck",
-                "ckmb",
-                "cl",
-                "crea",
-                "crp",
-                "dbp",
-                "fgn",
-                "fio2",
-                "glu",
-                "hgb",
-                "hr",
-                "inr_pt",
-                "k",
-                "lact",
-                "lymph",
-                "map",
-                "mch",
-                "mchc",
-                "mcv",
-                "methb",
-                "mg",
-                "na",
-                "neut",
-                "o2sat",
-                "pco2",
-                "ph",
-                "phos",
-                "plt",
-                "po2",
-                "ptt",
-                "resp",
-                "sbp",
-                "temp",
-                "tnt",
-                "urine",
-                "wbc",
-            ],
-            rotation=90,
-        )
-        plt.tight_layout()
-        plt.savefig(
-            log_dir / "{}_attribution_features_plot.png".format(method_name),
-            bbox_inches="tight",
-        )
-
-        # Plot for timestep attributions
-        x_values = np.arange(1, len(timestep_attrs) + 1)
-        plt.figure(figsize=(8, 6))
-        plt.plot(
-            x_values,
-            timestep_attrs,
-            marker="o",
-            color="skyblue",
-            linestyle="-",
-            linewidth=2,
-            markersize=8,
-        )
-        plt.xlabel("Time Step")
-        plt.ylabel("{} Attribution".format(method_name))
-        plt.title("{} Attribution Values".format(method_name))
-        plt.xticks(x_values)
-        plt.tight_layout()
-        plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")
-
     def explantation(
         self,
         dataloader,
@@ -677,7 +564,8 @@ def explantation(
                 timestep_attrs = Interpertations["attention"]
                 features_attrs = Interpertations["static_variables"].tolist()
                 features_attrs.extend(Interpertations["encoder_variables"].tolist())
-                r_score = self.Data_Randomization(
+                r_score = Data_Randomization(
+                    self,
                     x=None,
                     attribution=timestep_attrs,
                     explain_method=method,
@@ -685,7 +573,8 @@ def explantation(
                     dataloader=dataloader,
                     method_name=method_name,
                 )
-                st_i_score, st_o_score = self.Relative_Stability(
+                st_i_score, st_o_score = Relative_Stability(
+                    self,
                     x=None,
                     attribution=timestep_attrs,
                     explain_method=method,
@@ -706,7 +595,8 @@ def explantation(
 
                     if method_name == "Random":
                         f_ts_v_score.append(
-                            self.Faithfulness_Correlation(
+                            Faithfulness_Correlation(
+                                self,
                                 x,
                                 all_attrs,
                                 pertrub="baseline",
@@ -716,7 +606,8 @@ def explantation(
                             )
                         )
                         f_ts_score.append(
-                            self.Faithfulness_Correlation(
+                            Faithfulness_Correlation(
+                                self,
                                 x,
                                 all_attrs,
                                 pertrub="baseline",
@@ -726,7 +617,8 @@ def explantation(
                             )
                         )
                         f_v_score.append(
-                            self.Faithfulness_Correlation(
+                            Faithfulness_Correlation(
+                                self,
                                 x,
                                 all_attrs,
                                 pertrub="baseline",
@@ -737,7 +629,8 @@ def explantation(
                         )
 
                         r_score.append(
-                            self.Data_Randomization(
+                            Data_Randomization(
+                                self,
                                 x,
                                 attribution=all_attrs,
                                 explain_method=method,
@@ -745,7 +638,8 @@ def explantation(
                                 method_name=method_name,
                             )
                         )
-                        res1, res2 = self.Relative_Stability(
+                        res1, res2 = Relative_Stability(
+                            self,
                             x,
                             all_attrs,
                             explain_method=method,
@@ -757,7 +651,8 @@ def explantation(
                         st_o_score.append(res2)
                     else:
                         f_ts_score.append(
-                            self.Faithfulness_Correlation(
+                            Faithfulness_Correlation(
+                                self,
                                 x,
                                 timestep_attrs,
                                 pertrub="baseline",
@@ -767,7 +662,8 @@ def explantation(
                             )
                         )
                         f_v_score.append(
-                            self.Faithfulness_Correlation(
+                            Faithfulness_Correlation(
+                                self,
                                 x,
                                 features_attrs,
                                 pertrub="baseline",
@@ -829,7 +725,8 @@ def explantation(
             )
             if XAI_metric:
                 f_ts_v_score.append(
-                    self.Faithfulness_Correlation(
+                    Faithfulness_Correlation(
+                        self,
                         x,
                         stacked_attr,
                         pertrub="baseline",
@@ -840,7 +737,8 @@ def explantation(
                 )
 
                 f_ts_score.append(
-                    self.Faithfulness_Correlation(
+                    Faithfulness_Correlation(
+                        self,
                         x,
                         stacked_attr,
                         pertrub="baseline",
@@ -850,7 +748,8 @@ def explantation(
                     )
                 )
                 f_v_score.append(
-                    self.Faithfulness_Correlation(
+                    Faithfulness_Correlation(
+                        self,
                         x,
                         stacked_attr,
                         pertrub="baseline",
@@ -860,7 +759,8 @@ def explantation(
                     )
                 )
                 r_score.append(
-                    self.Data_Randomization(
+                    Data_Randomization(
+                        self,
                         x,
                         attribution=stacked_attr,
                         explain_method=method,
@@ -869,7 +769,8 @@ def explantation(
                     )
                 )
 
-                res1, res2 = self.Relative_Stability(
+                res1, res2 = Relative_Stability(
+                    self,
                     x,
                     stacked_attr,
                     explain_method=method,
@@ -901,14 +802,6 @@ def explantation(
         st_i_score = np.max(st_i_score)
         st_o_score = np.max(st_o_score)
 
-        """ if plot:
-            log_dir_plots = log_dir / "plots"
-            if not (log_dir_plots.exists()):
-                log_dir_plots.mkdir(parents=True)
-            # Plot attributions for features and timesteps
-
-            self.plot_attributions(features_attrs, timestep_attrs, method_name, log_dir_plots) """
-
         # Return computed attributions and metrics
         return (
             all_attrs,
@@ -994,381 +887,6 @@ def apply_baseline(self, x, indices, time_step, feature, feature_timestep):
             x["encoder_cont"] *= mask
         return x
 
-    def Faithfulness_Correlation(
-        self,
-        x,
-        attribution,
-        similarity_func=None,
-        nr_runs=100,
-        pertrub=None,
-        subset_size=3,
-        feature=False,
-        time_step=False,
-        feature_timestep=False,
-    ):
-        """
-        Calculates faithfulness scores for captum attributions
-
-        Args:
-            - x:Batch input
-            -attribution: attribution generated by captum,
-            - similarity_func:function to determine similarity between sum of attributions and difference in prediction
-            - nr_runs: How many times to repeat the experiment,
-            - pertrub: What change to do to the input,
-            - subset_size: The size of the subset of featrues to alter ,
-            - feature: Determines if to calcualte faithfulness of feature attributions,
-            - time_step: Determines if to calcualte faithfulness of timesteps attributions,
-            - feature_timestep: Determines if to calcualte faithfulness of featrues per timesteps attributions,
-        Returns:
-            score: similarity score between sum of attributions and difference in prediction averaged over nr_runs
-
-        Implementation of faithfulness correlation by Bhatt et al., 2020.
-
-        The Faithfulness Correlation metric intend to capture an explanation's relative faithfulness
-        (or 'fidelity') with respect to the model behaviour.
-
-        Faithfulness correlation scores shows to what extent the predicted logits of each modified test point and
-        the average explanation attribution for only the subset of features are (linearly) correlated, taking the
-        average over multiple runs and test samples. The metric returns one float per input-attribution pair that
-        ranges between -1 and 1, where higher scores are better.
-
-        For each test sample, |S| features are randomly selected and replace them with baseline values (zero baseline
-        or average of set). Thereafter, Pearson’s correlation coefficient between the predicted logits of each modified
-        test point and the average explanation attribution for only the subset of features is calculated. Results is
-        average over multiple runs and several test samples.
-        This code is adapted from the quantus libray to suit our use case
-
-        References:
-            1) Umang Bhatt et al.: "Evaluating and aggregating feature-based model
-            explanations." IJCAI (2020): 3016-3022.
-            2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for
-            responsible evaluation of neural network explanations and beyond."
-            Journal of Machine Learning Research 24.34 (2023): 1-11.
-        """
-
-        attribution = torch.tensor(attribution).to(self.device)
-
-        # Other initializations
-        if similarity_func is None:
-            similarity_func = correlation_spearman
-        if pertrub is None:
-            pertrub = "baseline"
-        similarities = []
-
-        # Assuming this is a method to prepare your data
-
-        y_pred = self(self.prep_data(x)).detach()  # Keep on GPU
-        pred_deltas = []
-        att_sums = []
-
-        for i_ix in range(nr_runs):
-            if time_step:
-                timesteps_idx = np.random.choice(24, subset_size, replace=False)
-                patient_idx = np.random.choice(64, 1, replace=False)
-                a_ix = [patient_idx, timesteps_idx]
-
-            elif feature:
-                feature_idx = np.random.choice(53, subset_size, replace=False)
-                patient_idx = np.random.choice(64, 1, replace=False)
-                a_ix = [patient_idx, feature_idx]
-            elif feature_timestep:
-                timesteps_idx = np.random.choice(24, subset_size[0], replace=False)
-                feature_idx = np.random.choice(53, subset_size[1], replace=False)
-                patient_idx = np.random.choice(64, 1, replace=False)
-                a_ix = [patient_idx, timesteps_idx, feature_idx]
-
-            # Apply perturbation
-            if pertrub == "Noise":
-                x = self.add_noise(x, a_ix, time_step, feature, feature_timestep)
-            elif pertrub == "baseline":
-                x = self.apply_baseline(x, a_ix, time_step, feature, feature_timestep)
-
-            # Predict on perturbed input and calculate deltas
-            y_pred_perturb = (self(self.prep_data(x))).detach()  # Keep on GPU
-
-            if time_step:
-                if attribution.size() == torch.Size([24]):
-                    att_sums.append((attribution[timesteps_idx]).sum())
-                else:
-                    att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :]).sum())
-            elif feature:
-                if len(attribution) == 53:
-                    att_sums.append((attribution[feature_idx]).sum())
-                else:
-                    att_sums.append((attribution[patient_idx, :, :][:, :, feature_idx]).sum())
-            elif feature_timestep:
-                att_sums.append((attribution[patient_idx, :, :][:, timesteps_idx, :][:, :, feature_idx]).sum())
-
-            pred_deltas.append((y_pred - y_pred_perturb)[patient_idx].item())
-            # Convert to CPU for numpy operations
-
-        pred_deltas_cpu = torch.tensor(pred_deltas).cpu().numpy()
-        att_sums_cpu = torch.tensor(att_sums).cpu().numpy()
-
-        similarities.append(similarity_func(pred_deltas_cpu, att_sums_cpu))
-
-        score = np.nanmean(similarities)
-        return score
-
-    def Data_Randomization(
-        self,
-        x,
-        attribution,
-        explain_method,
-        random_model,
-        similarity_func=cosine,
-        dataloader=None,
-        method_name="",
-        **kwargs,
-    ):
-        """
-
-        Args:
-            - x:Batch input
-            -attribution: attribution
-            - explain_method:function to generate explantations
-            - random_model: Reference to model trained on random labels
-            - similarity_func: Function to measure similiarity
-            - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
-            - method_name: Name of the explantation
-
-        Returns:
-            score: similarity score between attributions of model trained on random data and model trained on real data
-
-        Implementation of the Random Logit Metric by Sixt et al., 2020.
-
-        The Random Logit Metric computes the distance between the original explanation and a reference explanation of
-        a randomly chosen non-target class.
-        This code is adapted from the quantus libray to suit our use case
-
-        References:
-            1) Leon Sixt et al.: "When Explanations Lie: Why Many Modified BP
-            Attributions Fail." ICML (2020): 9046-9057.
-            2)Hedström, Anna, et al. "Quantus: An explainable ai
-            toolkit for responsible evaluation of neural network explanations and beyond."
-              Journal of Machine Learning Research 24.34 (2023): 1-11.
-
-        """
-
-        if explain_method == "Attention":
-            Attention_weights = random_model.interpertations(dataloader)
-            attribution = attribution.cpu().numpy()
-            min_val = np.min(attribution)
-            max_val = np.max(attribution)
-
-            attribution = (attribution - min_val) / (max_val - min_val)
-            random_attr = Attention_weights["attention"].cpu().numpy()
-            min_val = np.min(random_attr)
-            max_val = np.max(random_attr)
-            random_attr = (random_attr - min_val) / (max_val - min_val)
-            score = similarity_func(random_attr, attribution)
-        elif explain_method == "Random":
-            score = similarity_func(np.random.normal(size=[64, 24, 53]).flatten(), attribution.flatten())
-        else:
-            data, baselines = self.prep_data_captum(x)
-
-            explantation = explain_method(random_model.forward_captum)
-            # Reformat attributions.
-            if explain_method is not captum.attr.Saliency:
-                attr = explantation.attribute(data, baselines=baselines, **kwargs)
-            else:
-                attr = explantation.attribute(data, **kwargs)
-
-            # Process and store the calculated attributions
-            random_attr = (
-                attr[1].cpu().detach().numpy()
-                if method_name in ["Lime", "FeatureAblation"]
-                else torch.stack(attr).cpu().detach().numpy()
-            )
-
-            attribution = attribution.flatten()
-            min_val = np.min(attribution)
-            max_val = np.max(attribution)
-            attribution = (attribution - min_val) / (max_val - min_val)
-            random_attr = random_attr.flatten()
-            min_val = np.min(random_attr)
-            max_val = np.max(random_attr)
-            random_attr = (random_attr - min_val) / (max_val - min_val)
-
-            score = similarity_func(random_attr, attribution)
-        return score
-
-    def Relative_Stability(
-        self,
-        x,
-        attribution,
-        explain_method,
-        method_name,
-        dataloader=None,
-        thershold=0.5,
-        **kwargs,
-    ):
-        """
-        Args:
-               - x:Batch input
-               -attribution: attribution
-               - explain_method:function to generate explantations
-               - method_name: Name of the explantation
-               - dataloader:In case of using Attention as the explain method need to pass the dataloader instead of the batch ,
-
-
-           Returns:
-               RIS : relative distance between the explantation and the input
-               ROS: relative distance between the explantation and the output
-
-
-        References:
-               1) `https://arxiv.org/pdf/2203.06877.pdf
-               2)Hedström, Anna, et al. "Quantus: An explainable ai toolkit for responsible evaluation
-               of neural network explanations and beyond." Journal of Machine Learning Research 24.34 (2023): 1-11.
-
-        """
-
-        def relative_stability_objective(x, xs, e_x, e_xs, eps_min=0.0001, input=False, device="cuda") -> torch.Tensor:
-            """
-            Computes relative input and output stabilities maximization objective
-            as defined here :ref:`https://arxiv.org/pdf/2203.06877.pdf` by the authors.
-
-            Args:
-
-                x: Input tensor
-                xs: perturbed tensor.
-                e_x: Explanations for x.
-                e_xs: Explanations for xs.
-                eps_min:Value to avoid division by zero if needed
-                input:Boolean to indicate if this is an input or an output
-                device: the device to keep the tensors on
-
-            Returns:
-
-                ris_obj: Tensor
-                    RIS maximization objective.
-            """
-
-            # Function to convert inputs to tensors if they are numpy arrays
-            def to_tensor(input_array):
-                if isinstance(input_array, np.ndarray):
-                    return torch.tensor(input_array).to(device)
-                return input_array.to(device)
-
-            # Convert all inputs to tensors and move to GPU
-            x, xs, e_x, e_xs = map(to_tensor, [x, xs, e_x, e_xs])
-
-            if input:
-                num_dim = x.ndim
-            else:
-                num_dim = e_x.ndim
-
-            if num_dim == 3:
-
-                def norm_function(arr):
-                    return torch.norm(arr, dim=(-1, -2))
-
-            elif num_dim == 2:
-
-                def norm_function(arr):
-                    return torch.norm(arr, dim=-1)
-
-            else:
-
-                def norm_function(arr):
-                    return torch.norm(arr)
-
-            nominator = (e_x - e_xs) / (e_x + (e_x == 0) * eps_min)
-            nominator = norm_function(nominator)
-
-            if input:
-                denominator = x - xs
-                denominator /= x + (x == 0) * eps_min
-                denominator = norm_function(denominator)
-                denominator += (denominator == 0) * eps_min
-            else:
-                denominator = torch.squeeze(x) - torch.squeeze(xs)
-                denominator = torch.norm(denominator, dim=-1)
-                denominator += (denominator == 0) * eps_min
-
-            return nominator / denominator
-
-        attribution = torch.tensor(attribution).to(self.device)
-        if explain_method == "Attention":
-            y_pred = self.model.predict(dataloader)
-            x_original = dataloader.dataset.data["reals"].clone()
-
-            dataloader.dataset.add_noise()
-            x_preturb = dataloader.dataset.data["reals"].clone()
-            y_pred_preturb = self.model.predict(dataloader)
-            Attention_weights = self.interpertations(dataloader)
-            att_preturb = Attention_weights["attention"]
-            # Calculate the absolute difference
-            difference = torch.abs(y_pred_preturb - y_pred)
-
-            # Find where the difference is less than or equal to a thershold
-            close_indices = torch.nonzero(difference <= thershold).squeeze()
-            RIS = relative_stability_objective(
-                x_original[close_indices, :, :].detach(),
-                x_preturb[close_indices, :, :].detach(),
-                attribution,
-                att_preturb,
-                input=True,
-            )
-
-            ROS = relative_stability_objective(
-                y_pred[close_indices],
-                y_pred_preturb[close_indices],
-                attribution,
-                att_preturb,
-                input=False,
-            )
-
-        else:
-            y_pred = self(self.prep_data(x)).detach()
-            x_original = x["encoder_cont"].detach().clone()
-
-            with torch.no_grad():
-                noise = torch.randn_like(x["encoder_cont"]) * 0.01
-                x["encoder_cont"] += noise
-            y_pred_preturb = self(self.prep_data(x)).detach()
-            if explain_method == "Random":
-                att_preturb = np.random.normal(size=[64, 24, 53])
-                att_preturb = torch.tensor(att_preturb).to(self.device)
-            else:
-                data, baselines = self.prep_data_captum(x)
-
-                explantation = explain_method(self.forward_captum)
-                # Reformat attributions.
-                if explain_method is not captum.attr.Saliency:
-                    att_preturb = explantation.attribute(data, baselines=baselines, **kwargs)
-                else:
-                    att_preturb = explantation.attribute(data, **kwargs)
-
-                # Process and store the calculated attributions
-                att_preturb = (
-                    att_preturb[1].detach()
-                    if method_name in ["Lime", "FeatureAblation"]
-                    else torch.stack(att_preturb).detach()
-                )
-            # Calculate the absolute difference
-            difference = torch.abs(y_pred_preturb - y_pred)
-
-            # Find where the difference is less than or equal to a thershold
-            close_indices = torch.nonzero(difference <= thershold).squeeze()
-            RIS = relative_stability_objective(
-                x_original[close_indices, :, :].detach(),
-                x["encoder_cont"][close_indices, :, :].detach(),
-                attribution[close_indices, :, :],
-                att_preturb[close_indices, :, :],
-                input=True,
-            )
-            ROS = relative_stability_objective(
-                y_pred[close_indices],
-                y_pred_preturb[close_indices],
-                attribution[close_indices, :, :],
-                att_preturb[close_indices, :, :],
-                input=False,
-            )
-
-        return np.max(RIS.cpu().numpy()).astype(np.float64), np.max(ROS.cpu().numpy()).astype(np.float64)
-
 
 @gin.configurable("MLWrapper")
 class MLWrapper(BaseModule, ABC):
diff --git a/scripts/plotting/plotting.py b/scripts/plotting/plotting.py
index 779eea26..881d4e1b 100644
--- a/scripts/plotting/plotting.py
+++ b/scripts/plotting/plotting.py
@@ -1,4 +1,5 @@
 import matplotlib.pyplot as plt
+import numpy as np
 
 
 class Plotter:
@@ -49,3 +50,194 @@ def calibration_curve(self):
         plt.legend(loc="lower right")
         plt.savefig(self.save_dir / f"call_curve {self.specifier}.png")
         plt.clf()
+
+    def plot_XAI_Metrics(accumulated_metrics, log_dir_plots):
+        groups = {}
+        for key in accumulated_metrics["avg"]:
+            if key in ["loss", "MAE"]:
+                continue
+            suffix = key.split("_")[-1]
+            if suffix not in groups:
+                groups[suffix] = []
+            groups[suffix].append(key)
+
+        # Define a dictionary for legend labels
+        legend_labels = {
+            "IG": "Integrated Gradient",
+            "G": "Gradient",
+            "R": "Random",
+            "FA": "Feature Ablation",
+            "Att": "Attention",
+            "VSN": "Variable Selection Network",
+            "L": "Lime",
+        }
+        colors = [
+            "navy",
+            "skyblue",
+            "crimson",
+            "salmon",
+            "teal",
+            "orange",
+            "darkgreen",
+            "lightgreen",
+        ]
+
+        # Plotting
+        num_groups = len(groups)
+        fig, axs = plt.subplots(num_groups, 1, figsize=(10, num_groups * 5))
+
+        # Custom handles for the legend
+        # handles = [plt.Rectangle((0, 0), 1, 1, color="none",
+        # label=f"{key}: {value}") for key, value in legend_labels.items()]
+
+        for i, (suffix, keys) in enumerate(groups.items()):
+            ax = axs[i] if num_groups > 1 else axs
+            # Extract values and errors
+            avg_values = [accumulated_metrics["avg"][key] for key in keys]
+            ci_lower = [accumulated_metrics["CI_0.95"][key][0] for key in keys]
+            ci_upper = [accumulated_metrics["CI_0.95"][key][1] for key in keys]
+            ci_error = [np.abs([a - b, c - a]) for a, b, c in zip(avg_values, ci_lower, ci_upper)]
+
+            # Sort by absolute values of avg_values
+            sorted_indices = np.argsort([np.abs(val) for val in avg_values])[::-1]  # Indices to sort in descending order
+            sorted_keys = np.array(keys)[sorted_indices]
+            sorted_avg_values = np.array(avg_values)[sorted_indices]
+            sorted_ci_error = np.array(ci_error)[sorted_indices]
+
+            # Plot bars
+            bars = ax.bar(
+                sorted_keys,
+                np.abs(sorted_avg_values),
+                yerr=np.array(sorted_ci_error).T,
+                capsize=5,
+                color=colors,
+            )
+
+            # Set titles and labels
+            title_suffix = sorted_keys[0].split("_")[1]
+            ax.set_title(f'Metric: "{title_suffix}"')
+            ax.set_ylabel("Values")
+            ax.axhline(0, color="grey", linewidth=0.8)
+            ax.grid(axis="y")
+
+            # Set x-ticks
+            ax.set_xticks(sorted_keys)
+            ax.set_xticklabels([key.split("_")[0] for key in sorted_keys], rotation=45, ha="right")
+            # Create a custom legend for each subplot
+            custom_labels = [legend_labels[key.split("_")[0]] for key in sorted_keys]
+            ax.legend(bars, custom_labels, loc="upper right")
+
+        plt.tight_layout()
+        plt.savefig(log_dir_plots / "metrics_plot.png", bbox_inches="tight")
+
+        def plot_attributions(self, features_attrs, timestep_attrs, method_name, log_dir):
+            """
+            Plots the attribution values for features and timesteps.
+
+            Args:
+                - features_attrs: Array of feature attribution values.
+                - timestep_attrs: Array of timestep attribution values.
+                - method_name: Name of the attribution method.
+                - log_dir: Directory to save the plots.
+            Returns:
+                Nothing
+            """
+
+            # Plot for feature attributions
+            x_values = np.arange(1, len(features_attrs) + 1)
+            plt.figure(figsize=(8, 6))
+            plt.plot(
+                x_values,
+                features_attrs,
+                marker="o",
+                color="skyblue",
+                linestyle="-",
+                linewidth=2,
+                markersize=8,
+            )
+            plt.xlabel("Feature")
+            plt.ylabel("{} Attribution".format(method_name))
+            plt.title("{} Attribution Values".format(method_name))
+            plt.xticks(
+                x_values,
+                [
+                    "height",
+                    "weight",
+                    "age",
+                    "sex",
+                    "time_idx",
+                    "alb",
+                    "alp",
+                    "alt",
+                    "ast",
+                    "be",
+                    "bicar",
+                    "bili",
+                    "bili_dir",
+                    "bnd",
+                    "bun",
+                    "ca",
+                    "cai",
+                    "ck",
+                    "ckmb",
+                    "cl",
+                    "crea",
+                    "crp",
+                    "dbp",
+                    "fgn",
+                    "fio2",
+                    "glu",
+                    "hgb",
+                    "hr",
+                    "inr_pt",
+                    "k",
+                    "lact",
+                    "lymph",
+                    "map",
+                    "mch",
+                    "mchc",
+                    "mcv",
+                    "methb",
+                    "mg",
+                    "na",
+                    "neut",
+                    "o2sat",
+                    "pco2",
+                    "ph",
+                    "phos",
+                    "plt",
+                    "po2",
+                    "ptt",
+                    "resp",
+                    "sbp",
+                    "temp",
+                    "tnt",
+                    "urine",
+                    "wbc",
+                ],
+                rotation=90,
+            )
+            plt.tight_layout()
+            plt.savefig(
+                log_dir / "{}_attribution_features_plot.png".format(method_name),
+                bbox_inches="tight",
+            )
+
+            # Plot for timestep attributions
+            x_values = np.arange(1, len(timestep_attrs) + 1)
+            plt.figure(figsize=(8, 6))
+            plt.plot(
+                x_values,
+                timestep_attrs,
+                marker="o",
+                color="skyblue",
+                linestyle="-",
+                linewidth=2,
+                markersize=8,
+            )
+            plt.xlabel("Time Step")
+            plt.ylabel("{} Attribution".format(method_name))
+            plt.title("{} Attribution Values".format(method_name))
+            plt.xticks(x_values)
+            plt.tight_layout()
+            plt.savefig(log_dir / "{}_attribution_plot.png".format(method_name), bbox_inches="tight")

From 6782d4fefc92be69b598656a0ab178b67d84437b Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 29 Apr 2024 13:59:27 +0200
Subject: [PATCH 135/142] restored tft gin file

---
 icu_benchmarks/data/loader.py      | 4 ++--
 icu_benchmarks/models/dl_models.py | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 1707a2d3..f5ed350f 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -169,7 +169,7 @@ def to_tensor(self):
             return from_numpy(data), from_numpy(labels)
 
 
-"""
+
 @gin.configurable("PredictionDatasetTFT")
 class PredictionDatasetTFT(PredictionDataset):
     Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.
@@ -298,7 +298,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         tensors = (from_numpy(np.array(tensor)).to(float32) for tensor in tensors)
         tensors = [stack((x,), dim=-1) if x.numel() > 0 else empty(0) for x in tensors]
         return OrderedDict(zip(Features.FEAT_NAMES, tensors)), from_numpy(pad_mask)
-"""
+
 
 
 @gin.configurable("ImputationDataset")
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 98282bd4..59384af8 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -320,7 +320,7 @@ def forward(self, x):
         return pred
 
 
-'''
+
 @gin.configurable
 class TFT(DLPredictionWrapper):
     """
@@ -454,7 +454,7 @@ def forward(self, x: Dict[str, Tensor]) -> Tensor:
         )
         pred = self.logit(o)
         return pred
-'''
+
 
 
 @gin.configurable

From 3670bd6451e69e8b60db984ce0ecd25f51698fe1 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 13 May 2024 15:34:54 +0200
Subject: [PATCH 136/142] changes to get tft to work with yaib dataloader

---
 icu_benchmarks/data/loader.py      |  9 +++++--
 icu_benchmarks/models/dl_models.py | 40 ++++++++++++++++--------------
 icu_benchmarks/models/train.py     |  4 +--
 3 files changed, 31 insertions(+), 22 deletions(-)

diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index f5ed350f..809d3226 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -172,16 +172,19 @@ def to_tensor(self):
 
 @gin.configurable("PredictionDatasetTFT")
 class PredictionDatasetTFT(PredictionDataset):
+    """
     Subclass of prediction dataset for TFT as we need to define if variables are cont,static,known or observed.
     We also need to feed the model the variables in a specific order
     Args:
         ram_cache (bool, optional): Whether the complete dataset should be stored in ram. Defaults to True.
-
+    """
 
     def __init__(self, *args, ram_cache: bool = True, **kwargs):
         super().__init__(*args, ram_cache=True, **kwargs)
 
     def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
+
+        """  
         Function to sample from the data split of choice. Used for TFT.
         The data needs to be given to the model in the following order
         [static categorical,static contious,known catergorical,known continous,
@@ -189,7 +192,8 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         Args:
             idx: A specific row index to sample.
         Returns:
-            A sample from the data, consisting of data, labels and padding mask.
+            A sample from the data, consisting of data, labels and padding mask. 
+        """
 
         if self._cached_dataset is not None:
             return self._cached_dataset[idx]
@@ -199,6 +203,7 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         # We need to be sure that tensors are returned in the correct order to be processed correclty by tft
         tensors = [[] for _ in range(8)]
+        print(self.features_df.columns)
         for var in self.features_df.columns:
             if var == "sex":
                 tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 59384af8..9bff14f8 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -9,6 +9,15 @@
     LocalBlock,
     TemporalBlock,
     PositionalEncoding,
+    MaybeLayerNorm,
+    GLU,
+    GRN,
+    TFTEmbedding,
+    LazyEmbedding,
+    VariableSelectionNetwork,
+    StaticCovariateEncoder,
+    InterpretableMultiHeadAttention,
+    TFTBack
 )
 import matplotlib.pyplot as plt
 from icu_benchmarks.models.wrappers import (
@@ -340,16 +349,14 @@ def __init__(
         dropout_att,
         example_length,  # determines interval to predict
         *args,
-        quantiles=[0.1, 0.5, 0.9],  # quantiles to produce
-        static_categorical_inp_size=[2],  # number of catergories
-        temporal_known_categorical_inp_size=[],
-        temporal_observed_categorical_inp_size=[
-            48
-        ],  # number of categorical observed variables
-        static_continuous_inp_size=3,  # number of static coutinous variables
-        temporal_known_continuous_inp_size=0,
-        temporal_observed_continuous_inp_size=48,
-        temporal_target_size=1,  # number of target variables
+        quantiles :list = gin.REQUIRED,  # quantiles to produce
+        static_categorical_inp_size:list = gin.REQUIRED ,  # number of catergories
+        temporal_known_categorical_inp_size:list = gin.REQUIRED,
+        temporal_observed_categorical_inp_size:list = gin.REQUIRED,  # number of categorical observed variables
+        static_continuous_inp_size: int=gin.REQUIRED,  # number of static coutinous variables
+        temporal_known_continuous_inp_size:int=gin.REQUIRED,
+        temporal_observed_continuous_inp_size:int=gin.REQUIRED,
+        temporal_target_size:int=gin.REQUIRED,  # number of target variables
         **kwargs,
     ):
         # derived variables
@@ -379,13 +386,6 @@ def __init__(
             num_future_vars=num_future_vars,
             num_historic_vars=num_historic_vars,
             *args,
-            static_categorical_inp_size=1,
-            temporal_known_categorical_inp_size=0,
-            temporal_observed_categorical_inp_size=48,
-            static_continuous_inp_size=3,
-            temporal_known_continuous_inp_size=0,
-            temporal_observed_continuous_inp_size=48,
-            temporal_target_size=1,
             **kwargs,
         )
 
@@ -420,7 +420,11 @@ def __init__(
             len(quantiles), num_classes
         )  # Linear layer on top to output to the number of classes and allow modification by predictionwrapper
 
-    def forward(self, x: Dict[str, Tensor]) -> Tensor:
+    def forward(self, x) -> Tensor:
+        #Prep data to be in format model expects 
+
+
+
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
         # Static context
         cs, ce, ch, cc = self.static_encoder(s_inp)
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index a911a7ac..ea400cf2 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -107,8 +107,8 @@ def train_common(
     dataset_class = (
         ImputationDataset
         if mode == RunMode.imputation
-        else (PredictionDatasetpytorch if (pytorch_forecasting) else PredictionDataset)
-    )
+         else PredictionDataset)
+    
 
     logging.info(f"Logging to directory: {log_dir}.")
     save_config_file(log_dir)  # We save the operative config before and also after training

From 8e6db7150dbe2a78a76d6b1917d552010ad54391 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Wed, 15 May 2024 13:13:29 +0200
Subject: [PATCH 137/142] infered datatypes from dict supplied through gin

---
 icu_benchmarks/models/constants.py | 21 ++++++++++++++++-
 icu_benchmarks/models/dl_models.py | 37 ++++++++++++++++++++++++------
 2 files changed, 50 insertions(+), 8 deletions(-)

diff --git a/icu_benchmarks/models/constants.py b/icu_benchmarks/models/constants.py
index cd03bcd0..752fd8ba 100644
--- a/icu_benchmarks/models/constants.py
+++ b/icu_benchmarks/models/constants.py
@@ -24,7 +24,7 @@
     CalibrationError,
     F1Score,
 )
-from enum import Enum
+from enum import Enum,IntEnum
 from icu_benchmarks.models.custom_metrics import (
     CalibrationCurve,
     BalancedAccuracy,
@@ -32,6 +32,7 @@
     JSD,
     BinaryFairnessWrapper,
 )
+import gin
 
 
 class MLMetrics:
@@ -100,3 +101,21 @@ class ImputationInit(str, Enum):
     XAVIER = "xavier"
     KAIMING = "kaiming"
     ORTHOGONAL = "orthogonal"
+
+
+@gin.constants_from_enum
+class DataTypes(Enum):
+    """Defines numerical types of each column."""
+    CONTINUOUS = 0
+    CATEGORICAL = 1
+    DATE = 2
+    STR = 3
+@gin.constants_from_enum
+class InputTypes(IntEnum):
+    """Defines input types of each column."""
+    TARGET = 0
+    OBSERVED = 1
+    KNOWN = 2
+    STATIC = 3
+    ID = 4  # Single column used as an entity identifier
+    TIME = 5  # Single column exclusively used as a time index
\ No newline at end of file
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 9bff14f8..8b388b2c 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -27,6 +27,7 @@
 from torch import Tensor
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss
+from constants import InputTypes,DataTypes
 
 
 @gin.configurable
@@ -350,15 +351,37 @@ def __init__(
         example_length,  # determines interval to predict
         *args,
         quantiles :list = gin.REQUIRED,  # quantiles to produce
-        static_categorical_inp_size:list = gin.REQUIRED ,  # number of catergories
-        temporal_known_categorical_inp_size:list = gin.REQUIRED,
-        temporal_observed_categorical_inp_size:list = gin.REQUIRED,  # number of categorical observed variables
-        static_continuous_inp_size: int=gin.REQUIRED,  # number of static coutinous variables
-        temporal_known_continuous_inp_size:int=gin.REQUIRED,
-        temporal_observed_continuous_inp_size:int=gin.REQUIRED,
-        temporal_target_size:int=gin.REQUIRED,  # number of target variables
+         # number of target variables
+        vars_type:dict =gin.REQUIRED,
+        temporal_target_size:int=gin.REQUIRED,
         **kwargs,
     ):
+        
+        static_categorical_inp_size=0 # number of catergories
+        temporal_known_categorical_inp_size=0
+        temporal_observed_categorical_inp_size=0  # number of categorical observed variables
+        static_continuous_inp_size=0  # number of static coutinous variables
+        temporal_known_continuous_inp_size=0
+        temporal_observed_continuous_inp_size=0
+        for value in vars_type.values():
+            if value==[DataTypes.CONTINUOUS, InputTypes.OBSERVED]:
+                temporal_observed_continuous_inp_size+=1
+            elif value==[DataTypes.CATEGORICAL, InputTypes.OBSERVED]:
+                temporal_observed_categorical_inp_size+=1
+            elif value==[DataTypes.CONTINUOUS, InputTypes.STATIC]:
+                static_continuous_inp_size+=1
+            elif value==[DataTypes.CATEGORICAL, InputTypes.STATIC]:
+                static_categorical_inp_size+=1
+            elif value==[DataTypes.CONTINUOUS, InputTypes.KNOWN]:
+                 temporal_known_continuous_inp_size+=1
+            elif value==[DataTypes.CATEGORICAL, InputTypes.KNOWN]:
+                temporal_known_categorical_inp_size+=1
+            else:
+                print('incorrect datatype')
+            
+
+
+        
         # derived variables
         num_static_vars = len(static_categorical_inp_size) + static_continuous_inp_size
         num_future_vars = (

From 97d41f7b784f0545d8d895723249f4e20cc0ca3f Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Fri, 17 May 2024 15:33:51 +0200
Subject: [PATCH 138/142] adjusted forward method in tft to handle input as
 model expects

---
 icu_benchmarks/models/dl_models.py | 38 +++++++++++++++++++++++++-----
 icu_benchmarks/models/layers.py    |  4 ++--
 2 files changed, 34 insertions(+), 8 deletions(-)

diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 8b388b2c..868a2161 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -24,7 +24,7 @@
     DLPredictionWrapper,
     DLPredictionPytorchForecastingWrapper,
 )
-from torch import Tensor
+from torch import Tensor,cat,stack,empty
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss
 from constants import InputTypes,DataTypes
@@ -353,17 +353,19 @@ def __init__(
         quantiles :list = gin.REQUIRED,  # quantiles to produce
          # number of target variables
         vars_type:dict =gin.REQUIRED,
+        vars: Dict[str, str] = gin.REQUIRED,
         temporal_target_size:int=gin.REQUIRED,
         **kwargs,
     ):
-        
+        self.vars_type=vars_type
+        self.vars=vars
         static_categorical_inp_size=0 # number of catergories
         temporal_known_categorical_inp_size=0
         temporal_observed_categorical_inp_size=0  # number of categorical observed variables
         static_continuous_inp_size=0  # number of static coutinous variables
         temporal_known_continuous_inp_size=0
         temporal_observed_continuous_inp_size=0
-        for value in vars_type.values():
+        for value in self.vars_type.values():
             if value==[DataTypes.CONTINUOUS, InputTypes.OBSERVED]:
                 temporal_observed_continuous_inp_size+=1
             elif value==[DataTypes.CATEGORICAL, InputTypes.OBSERVED]:
@@ -445,10 +447,34 @@ def __init__(
 
     def forward(self, x) -> Tensor:
         #Prep data to be in format model expects 
+        tensors = [[] for _ in range(8)]
+        i=0
+        nan_array = np.full_like(x[:, 0], np.nan)
+        for var in vars:
+            
+            if self.vars_type[var] == [DataTypes.CATEGORICAL, InputTypes.STATIC]:
+                tensors[0].append(x[:, i].to_numpy())
+            elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.STATIC]:
+                tensors[1].append(x[:, i].to_numpy())
+            elif self.vars_type[var] == [DataTypes.CATEGORICAL, InputTypes.KNOWN]:
+                tensors[2].append(x[:, i].to_numpy())
+            elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.KNOWN]:
+                tensors[3].append(x[:, i].to_numpy())
+            elif self.vars_type[var] == [DataTypes.CATEGORICAL, InputTypes.OBSERVED]:
+                tensors[4].append(x[:, i].to_numpy())
+            elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.OBSERVED]:
+                tensors[5].append(x[:, i].to_numpy())
+            
+            i+=1
+        tensors[6].append(
+            nan_array#target needs to be there 
+            )
+        
+        
 
-
-
-        s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(x)
+        tensors = [stack(x, dim=-1) if x else empty(0) for x in tensors]
+        FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target']
+        s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(OrderedDict(zip(FEAT_NAMES, tensors)))
         # Static context
         cs, ce, ch, cc = self.static_encoder(s_inp)
         ch, cc = ch.unsqueeze(0), cc.unsqueeze(0)  # lstm initial states
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 944f4b89..e276efb2 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -8,7 +8,7 @@
 from torch.nn.parameter import UninitializedParameter
 from typing import Dict, Tuple, Optional
 from torch.nn import LayerNorm
-
+from collections import OrderedDict
 
 @gin.configurable("masking")
 def parallel_recomb(q_t, kv_t, att_type="all", local_context=3, bin_size=None):
@@ -546,7 +546,7 @@ def forward(self, x: Dict[str, Tensor]):
         t_cont_k_inp = x.get("k_cont", None)
         t_cat_o_inp = x.get("o_cat", None)
         t_cont_o_inp = x.get("o_cont", None)
-        t_tgt_obs = x["target"]  # Has to be present
+        t_tgt_obs = x.get("target")  # Has to be present
         # Static inputs are expected to be equal for all timesteps
         # For memory efficiency there is no assert statement
 

From afab7a44cdce9f4c98370ea950330ef59d41cfc1 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 23 May 2024 13:18:28 +0200
Subject: [PATCH 139/142] changes to get tft to work

---
 configs/prediction_models/common/DLCommon.gin |  2 +-
 icu_benchmarks/data/loader.py                 |  3 -
 icu_benchmarks/models/dl_models.py            | 56 +++++++++++--------
 icu_benchmarks/models/layers.py               |  8 +--
 4 files changed, 36 insertions(+), 33 deletions(-)

diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index bfc48ef5..0d44a936 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,7 +15,7 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-#train_common.batch_size = 64
+#train_common.batch_size = 2
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
diff --git a/icu_benchmarks/data/loader.py b/icu_benchmarks/data/loader.py
index 809d3226..a0eafc38 100644
--- a/icu_benchmarks/data/loader.py
+++ b/icu_benchmarks/data/loader.py
@@ -102,10 +102,8 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
         """
         if self._cached_dataset is not None:
             return self._cached_dataset[idx]
-
         pad_value = 0.0
         stay_id = self.outcome_df.index.unique()[idx]
-
         # slice to make sure to always return a DF
         window = self.features_df.loc[stay_id:stay_id].to_numpy()
         labels = self.outcome_df.loc[stay_id:stay_id][self.vars["LABEL"]].to_numpy(dtype=float)
@@ -203,7 +201,6 @@ def __getitem__(self, idx: int) -> Tuple[Tensor, Tensor, Tensor]:
 
         # We need to be sure that tensors are returned in the correct order to be processed correclty by tft
         tensors = [[] for _ in range(8)]
-        print(self.features_df.columns)
         for var in self.features_df.columns:
             if var == "sex":
                 tensors[0].append(self.features_df.loc[stay_id:stay_id][var].to_numpy())
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 868a2161..9e19a42c 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -4,6 +4,7 @@
 import torch.nn as nn
 from typing import Dict
 from icu_benchmarks.contants import RunMode
+from icu_benchmarks.models.constants import InputTypes,DataTypes
 from icu_benchmarks.models.layers import (
     TransformerBlock,
     LocalBlock,
@@ -24,10 +25,10 @@
     DLPredictionWrapper,
     DLPredictionPytorchForecastingWrapper,
 )
-from torch import Tensor,cat,stack,empty
+from torch import Tensor,cat,stack,empty,from_numpy
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss
-from constants import InputTypes,DataTypes
+from collections import OrderedDict
 
 
 @gin.configurable
@@ -359,25 +360,26 @@ def __init__(
     ):
         self.vars_type=vars_type
         self.vars=vars
-        static_categorical_inp_size=0 # number of catergories
-        temporal_known_categorical_inp_size=0
-        temporal_observed_categorical_inp_size=0  # number of categorical observed variables
+        static_categorical_inp_size=[] # number of catergories
+        temporal_known_categorical_inp_size=[]
+        temporal_observed_categorical_inp_size=[]  # number of categories in each category of observed variables
         static_continuous_inp_size=0  # number of static coutinous variables
         temporal_known_continuous_inp_size=0
         temporal_observed_continuous_inp_size=0
+        #Infering # of varaibles in each category based on the gin input
         for value in self.vars_type.values():
             if value==[DataTypes.CONTINUOUS, InputTypes.OBSERVED]:
                 temporal_observed_continuous_inp_size+=1
-            elif value==[DataTypes.CATEGORICAL, InputTypes.OBSERVED]:
-                temporal_observed_categorical_inp_size+=1
+            elif value[0:2]==[DataTypes.CATEGORICAL, InputTypes.OBSERVED]:#categoral variables need to define also # of categories 
+                temporal_observed_categorical_inp_size.append(value[2])
             elif value==[DataTypes.CONTINUOUS, InputTypes.STATIC]:
                 static_continuous_inp_size+=1
-            elif value==[DataTypes.CATEGORICAL, InputTypes.STATIC]:
-                static_categorical_inp_size+=1
+            elif value[0:2]==[DataTypes.CATEGORICAL, InputTypes.STATIC]:
+                static_categorical_inp_size.append(value[2])
             elif value==[DataTypes.CONTINUOUS, InputTypes.KNOWN]:
                  temporal_known_continuous_inp_size+=1
-            elif value==[DataTypes.CATEGORICAL, InputTypes.KNOWN]:
-                temporal_known_categorical_inp_size+=1
+            elif value[0:2]==[DataTypes.CATEGORICAL, InputTypes.KNOWN]:
+                temporal_known_categorical_inp_size.append(value[2])
             else:
                 print('incorrect datatype')
             
@@ -449,29 +451,35 @@ def forward(self, x) -> Tensor:
         #Prep data to be in format model expects 
         tensors = [[] for _ in range(8)]
         i=0
-        nan_array = np.full_like(x[:, 0], np.nan)
-        for var in vars:
-            
-            if self.vars_type[var] == [DataTypes.CATEGORICAL, InputTypes.STATIC]:
-                tensors[0].append(x[:, i].to_numpy())
+        nan_array = from_numpy(np.full_like(x[:, 0], np.nan))#target is nan in the input
+        print(x.size())
+        for var in self.vars:
+            print(var)
+            print(i)
+            print(self.vars_type[var])
+            print(x[:, i].size())
+
+
+            if self.vars_type[var][0:2] == [DataTypes.CATEGORICAL, InputTypes.STATIC]:
+                tensors[0].append(x[:, i])
             elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.STATIC]:
-                tensors[1].append(x[:, i].to_numpy())
-            elif self.vars_type[var] == [DataTypes.CATEGORICAL, InputTypes.KNOWN]:
-                tensors[2].append(x[:, i].to_numpy())
+                tensors[1].append(x[:, i])
+            elif self.vars_type[var][0:2] == [DataTypes.CATEGORICAL, InputTypes.KNOWN]:
+                tensors[2].append(x[:, i])
             elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.KNOWN]:
-                tensors[3].append(x[:, i].to_numpy())
-            elif self.vars_type[var] == [DataTypes.CATEGORICAL, InputTypes.OBSERVED]:
-                tensors[4].append(x[:, i].to_numpy())
+                tensors[3].append(x[:, i])
+            elif self.vars_type[var][0:2] == [DataTypes.CATEGORICAL, InputTypes.OBSERVED]:
+                tensors[4].append(x[:, i])
             elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.OBSERVED]:
-                tensors[5].append(x[:, i].to_numpy())
+                tensors[5].append(x[:, i])
             
             i+=1
+        
         tensors[6].append(
             nan_array#target needs to be there 
             )
         
         
-
         tensors = [stack(x, dim=-1) if x else empty(0) for x in tensors]
         FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target']
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(OrderedDict(zip(FEAT_NAMES, tensors)))
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index e276efb2..860a8339 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -445,7 +445,6 @@ def __init__(
         # 5. Temporal observed categorical
         # 6. Temporal observed continuous
         # 7. Temporal observed targets (time series obseved so far)
-
         self.s_cat_embed = (
             nn.ModuleList([nn.Embedding(n, self.hidden) for n in self.s_cat_inp_size]) if self.s_cat_inp_size else None
         )
@@ -552,7 +551,6 @@ def forward(self, x: Dict[str, Tensor]):
 
         s_cat_inp = s_cat_inp[:, 0, :] if s_cat_inp is not None else None
         s_cont_inp = s_cont_inp[:, 0, :] if s_cont_inp is not None else None
-
         s_inp = self._apply_embedding(
             s_cat_inp, s_cont_inp, self.s_cat_embed, self.s_cont_embedding_vectors, self.s_cont_embedding_bias
         )
@@ -627,15 +625,15 @@ def initialize_parameters(self, x):
             t_cont_k_inp = x.get("k_cont", None)
             t_cont_o_inp = x.get("o_cont", None)
             t_tgt_obs = x["target"]  # Has to be present
-            if (s_cont_inp is not None) and (s_cont_inp.size()[1] > 0):
+            if (s_cont_inp is not None) and (s_cont_inp.size()[0] > 0):
                 self.s_cont_embedding_vectors.materialize((s_cont_inp.shape[-1], self.hidden))
                 self.s_cont_embedding_bias.materialize((s_cont_inp.shape[-1], self.hidden))
 
-            if (t_cont_k_inp is not None) and (t_cont_k_inp.size()[1] > 0):
+            if (t_cont_k_inp is not None) and (t_cont_k_inp.size()[0] > 0):
                 self.t_cont_k_embedding_vectors.materialize((t_cont_k_inp.shape[-1], self.hidden))
                 self.t_cont_k_embedding_bias.materialize((t_cont_k_inp.shape[-1], self.hidden))
 
-            if (t_cont_o_inp) is not None and (t_cont_o_inp.size()[1] > 0):
+            if (t_cont_o_inp) is not None and (t_cont_o_inp.size()[0] > 0):
                 self.t_cont_o_embedding_vectors.materialize((t_cont_o_inp.shape[-1], self.hidden))
                 self.t_cont_o_embedding_bias.materialize((t_cont_o_inp.shape[-1], self.hidden))
 

From f6390533122157aae4662215c7a01a1ad7f6158b Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Thu, 23 May 2024 13:30:40 +0200
Subject: [PATCH 140/142] changed size check and sliced along the correct
 dimension

---
 configs/prediction_models/common/DLCommon.gin |  2 +-
 icu_benchmarks/models/dl_models.py            | 20 +++++++++----------
 icu_benchmarks/models/layers.py               |  5 +++--
 3 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 0d44a936..1544d07b 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,7 +15,7 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-#train_common.batch_size = 2
+train_common.batch_size = 10
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 9e19a42c..8c7a29df 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -449,29 +449,27 @@ def __init__(
 
     def forward(self, x) -> Tensor:
         #Prep data to be in format model expects 
+        
         tensors = [[] for _ in range(8)]
         i=0
         nan_array = from_numpy(np.full_like(x[:, 0], np.nan))#target is nan in the input
-        print(x.size())
+        
         for var in self.vars:
-            print(var)
-            print(i)
-            print(self.vars_type[var])
-            print(x[:, i].size())
+            
 
 
             if self.vars_type[var][0:2] == [DataTypes.CATEGORICAL, InputTypes.STATIC]:
-                tensors[0].append(x[:, i])
+                tensors[0].append(x[:, :, i])
             elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.STATIC]:
-                tensors[1].append(x[:, i])
+                tensors[1].append(x[:, :, i])
             elif self.vars_type[var][0:2] == [DataTypes.CATEGORICAL, InputTypes.KNOWN]:
-                tensors[2].append(x[:, i])
+                tensors[2].append(x[:, :, i])
             elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.KNOWN]:
-                tensors[3].append(x[:, i])
+                tensors[3].append(x[:, :, i])
             elif self.vars_type[var][0:2] == [DataTypes.CATEGORICAL, InputTypes.OBSERVED]:
-                tensors[4].append(x[:, i])
+                tensors[4].append(x[:, :, i])
             elif self.vars_type[var] == [DataTypes.CONTINUOUS, InputTypes.OBSERVED]:
-                tensors[5].append(x[:, i])
+                tensors[5].append(x[:, :, i])
             
             i+=1
         
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index 860a8339..b88d26f5 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -513,12 +513,13 @@ def _apply_embedding(
         cont_emb: Tensor,
         cont_bias: Tensor,
     ) -> Tuple[Optional[Tensor], Optional[Tensor]]:
+        
         e_cat = (
             torch.stack([embed(cat[..., i].int()) for i, embed in enumerate(cat_emb)], dim=-2)
-            if (cat is not None) and (cat.size()[1] > 0)
+            if (cat is not None) and (cat.size()[0] > 0)
             else None
         )
-        if (cont is not None) and (cont.size()[1] > 0):
+        if (cont is not None) and (cont.size()[0] > 0):
             # the line below is equivalent to following einsums
             # e_cont = torch.einsum('btf,fh->bthf', cont, cont_emb)
             # e_cont = torch.einsum('bf,fh->bhf', cont, cont_emb)

From 039377af74cb77850dcda7a63e19cd2ccaba3ec8 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 3 Jun 2024 15:20:24 +0200
Subject: [PATCH 141/142] added changes to get pytorch forecasting to wrok and
 gradient clipping

---
 configs/prediction_models/common/DLCommon.gin |  2 +-
 configs/tasks/Regression.gin                  |  2 +-
 icu_benchmarks/cross_validation.py            |  1 -
 icu_benchmarks/models/dl_models.py            |  7 ++--
 icu_benchmarks/models/layers.py               | 38 ++++++++++++++-----
 icu_benchmarks/models/train.py                |  7 ++--
 icu_benchmarks/models/wrappers.py             |  5 ++-
 7 files changed, 40 insertions(+), 22 deletions(-)

diff --git a/configs/prediction_models/common/DLCommon.gin b/configs/prediction_models/common/DLCommon.gin
index 1544d07b..6b76c60d 100644
--- a/configs/prediction_models/common/DLCommon.gin
+++ b/configs/prediction_models/common/DLCommon.gin
@@ -15,7 +15,7 @@ base_regression_preprocessor.generate_features = False
 # Train params
 train_common.optimizer = @Adam
 train_common.epochs = 1000
-train_common.batch_size = 10
+#train_common.batch_size = 10
 train_common.patience = 20
 train_common.min_delta = 1e-4
 
diff --git a/configs/tasks/Regression.gin b/configs/tasks/Regression.gin
index 6ce4c0f6..1234f707 100644
--- a/configs/tasks/Regression.gin
+++ b/configs/tasks/Regression.gin
@@ -16,7 +16,7 @@ train_common.ram_cache = True
 
 # LOSS FUNCTION
 DLPredictionWrapper.loss = @mse_loss
-DLPredictionPytorchForecastingWrapper.loss = @l1_loss
+DLPredictionPytorchForecastingWrapper.loss = @mse_loss
 MLWrapper.loss = @mean_squared_error
 
 # SELECTING PREPROCESSOR
diff --git a/icu_benchmarks/cross_validation.py b/icu_benchmarks/cross_validation.py
index c19076e5..ffa11b2a 100644
--- a/icu_benchmarks/cross_validation.py
+++ b/icu_benchmarks/cross_validation.py
@@ -132,7 +132,6 @@ def execute_repeated_cv(
             )
 
             train_time = datetime.now() - start_time
-
             log_full_line(
                 f"FINISHED FOLD {fold_index}| PREPROCESSING DURATION {preprocess_time}| PROCEDURE DURATION {train_time}",
                 level=logging.INFO,
diff --git a/icu_benchmarks/models/dl_models.py b/icu_benchmarks/models/dl_models.py
index 8c7a29df..68b6d2ae 100644
--- a/icu_benchmarks/models/dl_models.py
+++ b/icu_benchmarks/models/dl_models.py
@@ -25,7 +25,7 @@
     DLPredictionWrapper,
     DLPredictionPytorchForecastingWrapper,
 )
-from torch import Tensor,cat,stack,empty,from_numpy
+from torch import Tensor,cat,stack,from_numpy,zeros
 from pytorch_forecasting import TemporalFusionTransformer, RecurrentNetwork, DeepAR
 from pytorch_forecasting.metrics import QuantileLoss
 from collections import OrderedDict
@@ -452,7 +452,7 @@ def forward(self, x) -> Tensor:
         
         tensors = [[] for _ in range(8)]
         i=0
-        nan_array = from_numpy(np.full_like(x[:, 0], np.nan))#target is nan in the input
+        nan_array = from_numpy(np.full_like(x[:,:, 0].cpu(), -1)).to(x[:,:, 0].device)#target is nan in the input
         
         for var in self.vars:
             
@@ -478,11 +478,12 @@ def forward(self, x) -> Tensor:
             )
         
         
-        tensors = [stack(x, dim=-1) if x else empty(0) for x in tensors]
+        tensors = [stack(x, dim=-1) if x else zeros(0) for x in tensors]
         FEAT_NAMES = ['s_cat' , 's_cont' , 'k_cat' , 'k_cont' , 'o_cat' , 'o_cont' , 'target']
         s_inp, t_known_inp, t_observed_inp, t_observed_tgt = self.embedding(OrderedDict(zip(FEAT_NAMES, tensors)))
         # Static context
         cs, ce, ch, cc = self.static_encoder(s_inp)
+        
         ch, cc = ch.unsqueeze(0), cc.unsqueeze(0)  # lstm initial states
         # Temporal input
 
diff --git a/icu_benchmarks/models/layers.py b/icu_benchmarks/models/layers.py
index b88d26f5..0ea1bb02 100644
--- a/icu_benchmarks/models/layers.py
+++ b/icu_benchmarks/models/layers.py
@@ -421,7 +421,7 @@ def __init__(
         temporal_observed_continuous_inp_size,
         temporal_target_size,
         hidden,
-        initialize_cont_params=False,
+        initialize_cont_params=True,
     ):
         # initialize_cont_params=False prevents form initializing parameters inside this class
         # so they can be lazily initialized in LazyEmbedding module
@@ -513,22 +513,36 @@ def _apply_embedding(
         cont_emb: Tensor,
         cont_bias: Tensor,
     ) -> Tuple[Optional[Tensor], Optional[Tensor]]:
-        
+        """  print("Input cat:")
+        print(cat)
+        print("Shape:", cat.shape)
+        print("Contains NaNs:", torch.isnan(cat).any())
+        if cat is not None and cat.size(0) > 0:
+            e_cat = []
+            for i, embed in enumerate(cat_emb):
+                indices = cat[..., i].int()
+                embedded_values = embed(indices)
+                print(f"Output of embedding layer {i}:")
+                print(embedded_values)
+                print("Contains NaNs:", torch.isnan(embedded_values).any())
+                e_cat.append(embedded_values)
+            e_cat = torch.stack(e_cat, dim=-2)
+            print(1,e_cat) """
         e_cat = (
             torch.stack([embed(cat[..., i].int()) for i, embed in enumerate(cat_emb)], dim=-2)
             if (cat is not None) and (cat.size()[0] > 0)
             else None
-        )
+        ) 
+        #print(2,e_cat)
+        
         if (cont is not None) and (cont.size()[0] > 0):
-            # the line below is equivalent to following einsums
-            # e_cont = torch.einsum('btf,fh->bthf', cont, cont_emb)
-            # e_cont = torch.einsum('bf,fh->bhf', cont, cont_emb)
+            
+            
             e_cont = torch.mul(cont.unsqueeze(-1), cont_emb)
             e_cont = e_cont + cont_bias
-        # e_cont = fused_pointwise_linear_v1(cont, cont_emb, cont_bias)
+        
         else:
             e_cont = None
-
         if e_cat is not None and e_cont is not None:
             return torch.cat([e_cat, e_cont], dim=-2)
         elif e_cat is not None:
@@ -549,25 +563,28 @@ def forward(self, x: Dict[str, Tensor]):
         t_tgt_obs = x.get("target")  # Has to be present
         # Static inputs are expected to be equal for all timesteps
         # For memory efficiency there is no assert statement
+        
 
         s_cat_inp = s_cat_inp[:, 0, :] if s_cat_inp is not None else None
         s_cont_inp = s_cont_inp[:, 0, :] if s_cont_inp is not None else None
         s_inp = self._apply_embedding(
             s_cat_inp, s_cont_inp, self.s_cat_embed, self.s_cont_embedding_vectors, self.s_cont_embedding_bias
         )
+        
 
         t_known_inp = self._apply_embedding(
             t_cat_k_inp, t_cont_k_inp, self.t_cat_k_embed, self.t_cont_k_embedding_vectors, self.t_cont_k_embedding_bias
         )
-
+        
         t_observed_inp = self._apply_embedding(
             t_cat_o_inp, t_cont_o_inp, self.t_cat_o_embed, self.t_cont_o_embedding_vectors, self.t_cont_o_embedding_bias
         )
-
+        
         # Temporal observed targets
         t_observed_tgt = torch.matmul(t_tgt_obs.unsqueeze(3).unsqueeze(4), self.t_tgt_embedding_vectors.unsqueeze(1)).squeeze(
             3
         )
+       
         t_observed_tgt = t_observed_tgt + self.t_tgt_embedding_bias
 
         return s_inp, t_known_inp, t_observed_inp, t_observed_tgt
@@ -621,6 +638,7 @@ def __init__(
         self.t_tgt_embedding_bias = UninitializedParameter()
 
     def initialize_parameters(self, x):
+        
         if self.has_uninitialized_params():
             s_cont_inp = x.get("s_cont", None)
             t_cont_k_inp = x.get("k_cont", None)
diff --git a/icu_benchmarks/models/train.py b/icu_benchmarks/models/train.py
index ea400cf2..86677269 100644
--- a/icu_benchmarks/models/train.py
+++ b/icu_benchmarks/models/train.py
@@ -104,10 +104,8 @@ def train_common(
     logging.info(f"Training model: {model.__name__}.")
 
     # choose dataset_class based on the model
-    dataset_class = (
-        ImputationDataset
-        if mode == RunMode.imputation
-         else PredictionDataset)
+    dataset_class = ImputationDataset if mode == RunMode.imputation else (PredictionDatasetpytorch if pytorch_forecasting else PredictionDataset)
+
     
 
     logging.info(f"Logging to directory: {log_dir}.")
@@ -179,6 +177,7 @@ def train_common(
         logger=loggers,
         num_sanity_val_steps=-1,
         log_every_n_steps=5,
+        gradient_clip_val=gradient_clip_val
     )
     if not eval_only:
         if model.requires_backprop:
diff --git a/icu_benchmarks/models/wrappers.py b/icu_benchmarks/models/wrappers.py
index ae70406a..3ca1322a 100644
--- a/icu_benchmarks/models/wrappers.py
+++ b/icu_benchmarks/models/wrappers.py
@@ -319,7 +319,7 @@ def step_fn(self, element, step_prefix=""):
             element (object):
             step_prefix (str): Step type, by default: test, train, val.
         """
-
+       
         if len(element) == 2:
             data, labels = element[0], (element[1]).to(self.device)
             if isinstance(data, list):
@@ -342,8 +342,9 @@ def step_fn(self, element, step_prefix=""):
                 data = data.float().to(self.device)
         else:
             raise Exception("Loader should return either (data, label) or (data, label, mask)")
+        
         out = self(data)
-
+        
         # If aux_loss is present, it is returned as a tuple
         if len(out) == 2 and isinstance(out, tuple):
             out, aux_loss = out

From 2b509d2b59aec8880400aa572cdcd0d8b5f06346 Mon Sep 17 00:00:00 2001
From: youssefmecky96 <youssef.mecky@guest.hpi.de>
Date: Mon, 3 Jun 2024 15:32:50 +0200
Subject: [PATCH 142/142] config file for tft

---
 configs/prediction_models/TFT.gin | 166 ++++++++++++++++++++++++++++++
 1 file changed, 166 insertions(+)
 create mode 100644 configs/prediction_models/TFT.gin

diff --git a/configs/prediction_models/TFT.gin b/configs/prediction_models/TFT.gin
new file mode 100644
index 00000000..4a0a4e82
--- /dev/null
+++ b/configs/prediction_models/TFT.gin
@@ -0,0 +1,166 @@
+# Settings for TFT model.
+
+# Common settings for DL models
+include "configs/prediction_models/common/DLCommon.gin"
+
+# Optimizer params
+train_common.model = @TFT
+train_common/hyperparameter.class_to_tune = @train_common
+
+
+optimizer/hyperparameter.class_to_tune = @Adam
+optimizer/hyperparameter.weight_decay = 1e-6
+optimizer/hyperparameter.lr = (0.0001, 0.001,0.01)
+#(0.0001, 0.001,0.01)
+# Encoder params
+model/hyperparameter.class_to_tune = @TFT
+model/hyperparameter.encoder_length = 24
+model/hyperparameter.hidden = 32
+#(10,20,40,80,160,240,320)
+model/hyperparameter.num_classes = %NUM_CLASSES
+model/hyperparameter.dropout = (0.37, 0.38)
+model/hyperparameter.dropout_att = 0.0589
+model/hyperparameter.n_heads =2
+#(1,2,4)
+model/hyperparameter.example_length=25
+
+#TFT parameters 
+
+TFT.temporal_target_size=1
+TFT.quantiles=[0.1,0.5,0.9]
+
+train_common/hyperparameter.batch_size=64
+#(32,64,128,256,512)
+
+train_common/hyperparameter.gradient_clip_val=0.01
+#(0,0.01, 1.0, 100.0)
+
+
+#Vars types 
+
+TFT.vars_type = {
+    "alb": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "alp": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "alt": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "ast": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "be": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "bicar": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "bili": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "bili_dir": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "bnd": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "bun": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "ca": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "cai": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "ck": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "ckmb": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "cl": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "crea": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "crp": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "dbp": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "fgn": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "fio2": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "glu": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "hgb": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "hr": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "inr_pt": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "k": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "lact": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "lymph": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "map": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "mch": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "mchc": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "mcv": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "methb": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "mg": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "na": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "neut": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "o2sat": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "pco2": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "ph": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "phos": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "plt": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "po2": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "ptt": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "resp": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "sbp": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "temp": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "tnt": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "urine": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "wbc": [%DataTypes.CONTINUOUS, %InputTypes.OBSERVED],
+    "MissingIndicator_1": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_2": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_3": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_4": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_5": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_6": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_7": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_8": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_9": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_10": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_11": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_12": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_13": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_14": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_15": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_16": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_17": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_18": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_19": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_20": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_21": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_22": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_23": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_24": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_25": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_26": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_27": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_28": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_29": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_30": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_31": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_32": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_33": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_34": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_35": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_36": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_37": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_38": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_39": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_40": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_41": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_42": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_43": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_44": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_45": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_46": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_47": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "MissingIndicator_48": [%DataTypes.CATEGORICAL, %InputTypes.OBSERVED, 2],
+    "age": [%DataTypes.CONTINUOUS, %InputTypes.STATIC],
+    "sex": [%DataTypes.CATEGORICAL, %InputTypes.STATIC,2],
+    "height": [%DataTypes.CONTINUOUS, %InputTypes.STATIC],
+    "weight": [%DataTypes.CONTINUOUS, %InputTypes.STATIC]
+    }
+
+
+TFT.vars=  ['alb', 'alp', 'alt', 'ast', 'be', 'bicar', 'bili', 'bili_dir', 'bnd',
+       'bun', 'ca', 'cai', 'ck', 'ckmb', 'cl', 'crea', 'crp', 'dbp', 'fgn',
+       'fio2', 'glu', 'hgb', 'hr', 'inr_pt', 'k', 'lact', 'lymph', 'map',
+       'mch', 'mchc', 'mcv', 'methb', 'mg', 'na', 'neut', 'o2sat', 'pco2',
+       'ph', 'phos', 'plt', 'po2', 'ptt', 'resp', 'sbp', 'temp', 'tnt',
+       'urine', 'wbc', 'MissingIndicator_1', 'MissingIndicator_2',
+       'MissingIndicator_3', 'MissingIndicator_4', 'MissingIndicator_5',
+       'MissingIndicator_6', 'MissingIndicator_7', 'MissingIndicator_8',
+       'MissingIndicator_9', 'MissingIndicator_10', 'MissingIndicator_11',
+       'MissingIndicator_12', 'MissingIndicator_13', 'MissingIndicator_14',
+       'MissingIndicator_15', 'MissingIndicator_16', 'MissingIndicator_17',
+       'MissingIndicator_18', 'MissingIndicator_19', 'MissingIndicator_20',
+       'MissingIndicator_21', 'MissingIndicator_22', 'MissingIndicator_23',
+       'MissingIndicator_24', 'MissingIndicator_25', 'MissingIndicator_26',
+       'MissingIndicator_27', 'MissingIndicator_28', 'MissingIndicator_29',
+       'MissingIndicator_30', 'MissingIndicator_31', 'MissingIndicator_32',
+       'MissingIndicator_33', 'MissingIndicator_34', 'MissingIndicator_35',
+       'MissingIndicator_36', 'MissingIndicator_37', 'MissingIndicator_38',
+       'MissingIndicator_39', 'MissingIndicator_40', 'MissingIndicator_41',
+       'MissingIndicator_42', 'MissingIndicator_43', 'MissingIndicator_44',
+       'MissingIndicator_45', 'MissingIndicator_46', 'MissingIndicator_47',
+       'MissingIndicator_48', 'age', 'sex', 'height', 'weight']