Update linter/formatter (#576)

Author: JenniferWang

.flake8

@@ -8,8 +8,9 @@ max-line-length = 120
 # N817 ignored because importing using acronyms is convention (DistributedDataParallel as DDP)
 # E731 allow usage of assigning lambda expressions
 # N803,N806 allow caps and mixed case in function params. This is to work with Triton kernel coding style.
+# E704 ignored to allow black's formatting of Protocol stub methods (def method(self) -> None: ...)
 ignore =
-    E203,E305,E402,E501,E721,E741,F405,F821,F841,F999,W503,W504,C408,E302,W291,E303,N812,N817,E731,N803,N806
+    E203,E305,E402,E501,E704,E721,E741,F405,F821,F841,F999,W503,W504,C408,E302,W291,E303,N812,N817,E731,N803,N806
     # shebang has extra meaning in fbcode lints, so I think it's not worth trying
     # to line this up with executable bit
     EXE001,

.meta/mast/hydrate_cache.py

@@ -14,6 +14,7 @@
 python .meta/mast/hydrate_cache.py --model-id Qwen/Qwen3-32B
 
 """
+
 import argparse
 import os
 import sys

.pre-commit-config.yaml

@@ -39,8 +39,8 @@ repos:
     hooks:
     -   id: ufmt
         additional_dependencies:
-          - black == 22.12.0
-          - usort == 1.0.5
+          - black == 24.4.2
+          - usort == 1.0.8.post1
 
 - repo: https://github.com/jsh9/pydoclint
   rev: 0.5.12

apps/grpo/main.py

@@ -140,7 +140,6 @@ def simple_grpo_loss(
 
 @dataclass
 class RewardActor(ForgeActor):
-
     reward_functions: list[Callable]
 
     @endpoint

apps/sft/main.py

@@ -287,7 +287,6 @@ def __repr__(self) -> str:
 
 
 async def run(cfg: DictConfig) -> None:
-
     logging.info("Spawning recipe...")
     process_cfg = cfg.pop("processes")
 

pyproject.toml

@@ -100,4 +100,9 @@ prerelease = "allow"
 environments = [
     "sys_platform == 'linux'",
 ]
-# override-dependencies = ["torch>2.7.1", "torchaudio>=2.7.1", "torchvision>=0.22.0"]
+
+[tool.black]
+target-version = ["py310"] # match the minium supported python version
+
+[tool.usort]
+first_party_detection = false

src/forge/actors/coder.py

@@ -10,10 +10,10 @@
 import tempfile
 from pathlib import Path
 
-from monarch.actor import endpoint
-
 from forge.controller import ForgeActor
 
+from monarch.actor import endpoint
+
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.DEBUG)
 

src/forge/actors/generator.py

@@ -17,6 +17,28 @@
 
 import torch
 import torchstore as ts
+
+from forge.actors._torchstore_utils import (
+    extract_param_name,
+    get_dcp_whole_state_dict_key,
+    get_param_key,
+    get_param_prefix,
+    load_tensor_from_dcp,
+    rdma_available,
+)
+
+from forge.controller import (
+    ForgeActor,
+    get_proc_mesh,
+    host_mesh_from_proc,
+    stop_proc_mesh,
+)
+from forge.data_models.completion import Completion
+from forge.data_models.prompt import to_prompt
+from forge.observability.metrics import record_metric, Reduce
+from forge.observability.perf_tracker import Tracer
+from forge.types import ProcessConfig
+from forge.util._shared_tensor import SharedTensor, SharedTensorHandle
 from monarch.actor import current_rank, endpoint, ProcMesh, this_host
 
 from vllm.config import VllmConfig
@@ -42,28 +64,6 @@
 from vllm.v1.structured_output import StructuredOutputManager
 from vllm.worker.worker_base import WorkerWrapperBase
 
-from forge.actors._torchstore_utils import (
-    extract_param_name,
-    get_dcp_whole_state_dict_key,
-    get_param_key,
-    get_param_prefix,
-    load_tensor_from_dcp,
-    rdma_available,
-)
-
-from forge.controller import (
-    ForgeActor,
-    get_proc_mesh,
-    host_mesh_from_proc,
-    stop_proc_mesh,
-)
-from forge.data_models.completion import Completion
-from forge.data_models.prompt import to_prompt
-from forge.observability.metrics import record_metric, Reduce
-from forge.observability.perf_tracker import Tracer
-from forge.types import ProcessConfig
-from forge.util._shared_tensor import SharedTensor, SharedTensorHandle
-
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.INFO)
 

src/forge/actors/reference_model.py

@@ -13,6 +13,11 @@
 from dataclasses import dataclass, field, fields
 
 import torch
+
+from forge.controller import ForgeActor
+from forge.observability.metrics import record_metric, Reduce
+from forge.observability.perf_tracker import Tracer
+from forge.util.ops import compute_logprobs
 from monarch.actor import current_rank, current_size, endpoint
 from torch.distributed.tensor import DTensor
 
@@ -27,11 +32,6 @@
 from torchtitan.experiments.forge.engine import ForgeEngine
 from torchtitan.experiments.forge.job_config import ForgeJobConfig
 
-from forge.controller import ForgeActor
-from forge.observability.metrics import record_metric, Reduce
-from forge.observability.perf_tracker import Tracer
-from forge.util.ops import compute_logprobs
-
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.INFO)
 

src/forge/actors/replay_buffer.py

@@ -11,12 +11,12 @@
 from operator import itemgetter
 from typing import Any, Callable
 
-from monarch.actor import endpoint
-
 from forge.controller import ForgeActor
 from forge.observability.metrics import record_metric, Reduce
 from forge.observability.perf_tracker import trace
 
+from monarch.actor import endpoint
+
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.INFO)