Support huggingface

image_classification/models/uniformer.py

@@ -8,6 +8,7 @@
 from timm.models.vision_transformer import _cfg
 from timm.models.registry import register_model
 from timm.models.layers import trunc_normal_, DropPath, to_2tuple
+from huggingface_hub import PyTorchModelHubMixin
 
 layer_scale = False
 init_value = 1e-6
@@ -196,7 +197,10 @@ def forward(self, x):
         return x
 
 
-class UniFormer(nn.Module):
+class UniFormer(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

image_classification/models/uniformer_light.py

@@ -8,6 +8,7 @@
 from timm.models.vision_transformer import _cfg
 from timm.models.registry import register_model
 from timm.models.layers import trunc_normal_, DropPath, to_2tuple
+from huggingface_hub import PyTorchModelHubMixin
 
 
 layer_scale = False
@@ -330,7 +331,10 @@ def forward(self, x):
         return x
 
 
-class UniFormer_Light(nn.Module):
+class UniFormer_Light(nn.Module,
+                    PyTorchModelHubMixin,
+                    library_name = "UniFormer",
+                    repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

image_classification/requirements.txt

@@ -1,3 +1,4 @@
 torch==1.7.0
 torchvision==0.8.1
-timm==0.4.12
+timm==0.4.12
+huggingface_hub>=0.22.0

image_classification/token_labeling/tlt/models/uniformer.py

@@ -15,6 +15,7 @@
 from timm.models.vision_transformer import _cfg
 from timm.models.registry import register_model
 from timm.models.layers import trunc_normal_, DropPath, to_2tuple
+from huggingface_hub import PyTorchModelHubMixin
 
 layer_scale = False
 init_value = 1e-6
@@ -230,7 +231,10 @@ def forward(self, x):
         return x
 
 
-class UniFormer(nn.Module):
+class UniFormer(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

object_detection/mmdet/models/backbones/uniformer.py

@@ -15,6 +15,7 @@
 import torch.utils.checkpoint as checkpoint
 import numpy as np
 from timm.models.layers import DropPath, to_2tuple, trunc_normal_
+from huggingface_hub import PyTorchModelHubMixin
 
 from mmcv_custom import load_checkpoint
 from mmdet.utils import get_root_logger
@@ -240,7 +241,10 @@ def forward(self, x):
 
 
 @BACKBONES.register_module()   
-class UniFormer(nn.Module):
+class UniFormer(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

object_detection/mmdet/models/backbones/uniformer_light.py

@@ -6,6 +6,7 @@
 from timm.models.vision_transformer import _cfg
 from timm.models.layers import trunc_normal_, DropPath
 from torch.nn.modules.batchnorm import _BatchNorm
+from huggingface_hub import PyTorchModelHubMixin
 
 from mmcv_custom import load_checkpoint
 from mmdet.utils import get_root_logger
@@ -333,7 +334,10 @@ def forward(self, x):
 
 
 @BACKBONES.register_module() 
-class UniFormer_Light(nn.Module):
+class UniFormer_Light(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

pose_estimation/mmpose/models/backbones/uniformer.py

@@ -15,6 +15,7 @@
 import torch.utils.checkpoint as checkpoint
 import numpy as np
 from timm.models.layers import DropPath, to_2tuple, trunc_normal_
+from huggingface_hub import PyTorchModelHubMixin
 
 from mmcv_custom import load_checkpoint
 from mmpose.utils import get_root_logger
@@ -240,7 +241,10 @@ def forward(self, x):
 
 
 @BACKBONES.register_module()   
-class UniFormer(nn.Module):
+class UniFormer(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

semantic_segmentation/fpn_seg/uniformer.py

@@ -15,6 +15,7 @@
 import torch.utils.checkpoint as checkpoint
 import numpy as np
 from timm.models.layers import DropPath, to_2tuple, trunc_normal_
+from huggingface_hub import PyTorchModelHubMixin
 
 from mmcv_custom import load_checkpoint
 from mmseg.utils import get_root_logger
@@ -240,7 +241,10 @@ def forward(self, x):
 
 
 @BACKBONES.register_module()   
-class UniFormer(nn.Module):
+class UniFormer(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

semantic_segmentation/fpn_seg/uniformer_light.py

@@ -5,6 +5,7 @@
 from timm.models.vision_transformer import _cfg
 from timm.models.layers import trunc_normal_, DropPath
 from torch.nn.modules.batchnorm import _BatchNorm
+from huggingface_hub import PyTorchModelHubMixin
 
 from mmcv_custom import load_checkpoint
 from mmseg.utils import get_root_logger
@@ -332,7 +333,10 @@ def forward(self, x):
 
 
 @BACKBONES.register_module() 
-class UniFormer_Light(nn.Module):
+class UniFormer_Light(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

semantic_segmentation/mmseg/models/backbones/uniformer.py

@@ -15,6 +15,7 @@
 import torch.utils.checkpoint as checkpoint
 import numpy as np
 from timm.models.layers import DropPath, to_2tuple, trunc_normal_
+from huggingface_hub import PyTorchModelHubMixin
 
 from mmcv_custom import load_checkpoint
 from mmseg.utils import get_root_logger
@@ -240,7 +241,10 @@ def forward(self, x):
 
 
 @BACKBONES.register_module()   
-class UniFormer(nn.Module):
+class UniFormer(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

semantic_segmentation/mmseg/models/backbones/uniformer_light.py

@@ -5,6 +5,7 @@
 from timm.models.vision_transformer import _cfg
 from timm.models.layers import trunc_normal_, DropPath
 from torch.nn.modules.batchnorm import _BatchNorm
+from huggingface_hub import PyTorchModelHubMixin
 
 from mmcv_custom import load_checkpoint
 from mmseg.utils import get_root_logger
@@ -332,7 +333,10 @@ def forward(self, x):
 
 
 @BACKBONES.register_module() 
-class UniFormer_Light(nn.Module):
+class UniFormer_Light(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

video_classification/setup.cfg

@@ -3,7 +3,7 @@ line_length=100
 multi_line_output=4
 known_standard_library=numpy,setuptools
 known_myself=slowfast
-known_third_party=fvcore,iopath,av,torch,pycocotools,yacs,termcolor,scipy,simplejson,matplotlib,detectron2,torchvision,yaml,tqdm,psutil,opencv-python,pandas,tensorboard,moviepy,sklearn,cv2,PIL
+known_third_party=fvcore,iopath,av,torch,pycocotools,yacs,termcolor,scipy,simplejson,matplotlib,detectron2,torchvision,yaml,tqdm,psutil,opencv-python,pandas,tensorboard,moviepy,scikit-learn,cv2,PIL
 no_lines_before=STDLIB,THIRDPARTY
 sections=FUTURE,STDLIB,THIRDPARTY,myself,FIRSTPARTY,LOCALFOLDER
 default_section=FIRSTPARTY

video_classification/setup.py

@@ -24,9 +24,10 @@
         "pandas",
         "torchvision>=0.4.2",
         "pillow",
-        "sklearn",
+        "scikit-learn",
         "tensorboard",
-        "timm"
+        "timm",
+        "huggingface_hub>=0.22.0"
     ],
     extras_require={"tensorboard_video_visualization": ["moviepy"]},
     packages=find_packages(exclude=("configs", "tests")),

video_classification/slowfast/models/uniformer.py

@@ -8,6 +8,7 @@
 from timm.models.layers import trunc_normal_, DropPath, to_2tuple
 from .build import MODEL_REGISTRY
 import os
+from huggingface_hub import PyTorchModelHubMixin
 
 import slowfast.utils.logging as logging
 
@@ -258,7 +259,10 @@ def forward(self, x):
 
 
 @MODEL_REGISTRY.register()
-class Uniformer(nn.Module):
+class Uniformer(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

video_classification/slowfast/models/uniformer_light.py

@@ -8,6 +8,7 @@
 from timm.models.layers import trunc_normal_, DropPath, to_2tuple
 from .build import MODEL_REGISTRY
 import os
+from huggingface_hub import PyTorchModelHubMixin
 
 import slowfast.utils.logging as logging
 
@@ -396,7 +397,10 @@ def forward(self, x):
 
 
 @MODEL_REGISTRY.register()
-class Uniformer_light(nn.Module):
+class Uniformer_light(nn.Module,
+                      PyTorchModelHubMixin,
+                      library_name = "UniFormer",
+                      repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

video_classification/slowfast/models/uniformer_light_fp32.py

@@ -8,6 +8,7 @@
 from timm.models.layers import trunc_normal_, DropPath, to_2tuple
 from .build import MODEL_REGISTRY
 import os
+from huggingface_hub import PyTorchModelHubMixin
 
 import slowfast.utils.logging as logging
 
@@ -408,7 +409,10 @@ def forward(self, x):
 
 
 @MODEL_REGISTRY.register()
-class Uniformer_light_fp32(nn.Module):
+class Uniformer_light_fp32(nn.Module,
+                           PyTorchModelHubMixin,
+                           library_name = "UniFormer",
+                           repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

video_classification/slowfast/models/uniformer_light_ls_fp32.py

@@ -8,6 +8,7 @@
 from timm.models.layers import trunc_normal_, DropPath, to_2tuple
 from .build import MODEL_REGISTRY
 import os
+from huggingface_hub import PyTorchModelHubMixin
 
 import slowfast.utils.logging as logging
 
@@ -420,7 +421,10 @@ def forward(self, x):
 
 
 @MODEL_REGISTRY.register()
-class Uniformer_light_ls_fp32(nn.Module):
+class Uniformer_light_ls_fp32(nn.Module,
+                              PyTorchModelHubMixin,
+                              library_name = "UniFormer",
+                              repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929

video_classification/vis/model.py

@@ -3,6 +3,7 @@
 import torch.nn as nn
 from functools import partial
 from timm.models.layers import trunc_normal_, DropPath, to_2tuple
+from huggingface_hub import PyTorchModelHubMixin
 
 
 def conv_3xnxn(inp, oup, kernel_size=3, stride=3, groups=1):
@@ -236,7 +237,10 @@ def forward(self, x):
         return x
 
 
-class Uniformer(nn.Module):
+class Uniformer(nn.Module,
+                PyTorchModelHubMixin,
+                library_name = "UniFormer",
+                repo_url = "https://github.com/Sense-X/UniFormer"):
     """ Vision Transformer
     A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`  -
         https://arxiv.org/abs/2010.11929