Merge pull request #2209 from huggingface/fcossio-vit-maxpool

ViT pooling refactor

requirements.txt

@@ -2,4 +2,5 @@ torch>=1.7
 torchvision
 pyyaml
 huggingface_hub
-safetensors>=0.2
+safetensors>=0.2
+numpy<2.0

timm/models/efficientnet.py

@@ -156,7 +156,7 @@ class EfficientNet(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.classifier
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         self.global_pool, self.classifier = create_classifier(
             self.num_features, self.num_classes, pool_type=global_pool)

timm/models/ghostnet.py

@@ -273,7 +273,7 @@ class GhostNet(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.classifier
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         # cannot meaningfully change pooling of efficient head after creation
         self.global_pool = SelectAdaptivePool2d(pool_type=global_pool)

timm/models/hrnet.py

@@ -739,7 +739,7 @@ class HighResolutionNet(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.classifier
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         self.global_pool, self.classifier = create_classifier(
             self.num_features, self.num_classes, pool_type=global_pool)

timm/models/inception_v4.py

@@ -280,7 +280,7 @@ class InceptionV4(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.last_linear
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         self.global_pool, self.last_linear = create_classifier(
             self.num_features, self.num_classes, pool_type=global_pool)

timm/models/metaformer.py

@@ -26,9 +26,9 @@ Adapted from https://github.com/sail-sg/metaformer, original copyright below
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-
 from collections import OrderedDict
 from functools import partial
+from typing import Optional
 
 import torch
 import torch.nn as nn
@@ -548,7 +548,7 @@ class MetaFormer(nn.Module):
         # if using MlpHead, dropout is handled by MlpHead
         if num_classes > 0:
             if self.use_mlp_head:
-                # FIXME hidden size
+                # FIXME not actually returning mlp hidden state right now as pre-logits.
                 final = MlpHead(self.num_features, num_classes, drop_rate=self.drop_rate)
                 self.head_hidden_size = self.num_features
             else:
@@ -583,7 +583,7 @@ class MetaFormer(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.head.fc
 
-    def reset_classifier(self, num_classes=0, global_pool=None):
+    def reset_classifier(self, num_classes: int, global_pool: Optional[str] = None):
         if global_pool is not None:
             self.head.global_pool = SelectAdaptivePool2d(pool_type=global_pool)
             self.head.flatten = nn.Flatten(1) if global_pool else nn.Identity()

timm/models/nasnet.py

@@ -518,7 +518,7 @@ class NASNetALarge(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.last_linear
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         self.global_pool, self.last_linear = create_classifier(
             self.num_features, self.num_classes, pool_type=global_pool)

timm/models/pnasnet.py

@@ -307,7 +307,7 @@ class PNASNet5Large(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.last_linear
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         self.global_pool, self.last_linear = create_classifier(
             self.num_features, self.num_classes, pool_type=global_pool)

timm/models/regnet.py

@@ -514,7 +514,7 @@ class RegNet(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.head.fc
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: Optional[str] = None):
         self.head.reset(num_classes, pool_type=global_pool)
 
     def forward_intermediates(

timm/models/rexnet.py

@@ -12,6 +12,7 @@ Copyright 2020 Ross Wightman
 
 from functools import partial
 from math import ceil
+from typing import Optional
 
 import torch
 import torch.nn as nn
@@ -229,7 +230,7 @@ class RexNet(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.head.fc
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: Optional[str] = None):
         self.num_classes = num_classes
         self.head.reset(num_classes, global_pool)
 

timm/models/selecsls.py

@@ -161,7 +161,7 @@ class SelecSls(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.fc
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         self.global_pool, self.fc = create_classifier(self.num_features, self.num_classes, pool_type=global_pool)
 

timm/models/senet.py

@@ -337,7 +337,7 @@ class SENet(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.last_linear
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         self.global_pool, self.last_linear = create_classifier(
             self.num_features, self.num_classes, pool_type=global_pool)

timm/models/vision_transformer.py

@@ -386,6 +386,31 @@ class ParallelThingsBlock(nn.Module):
             return self._forward(x)
 
 
+def global_pool_nlc(
+        x: torch.Tensor,
+        pool_type: str = 'token',
+        num_prefix_tokens: int = 1,
+        reduce_include_prefix: bool = False,
+):
+    if not pool_type:
+        return x
+
+    if pool_type == 'token':
+        x = x[:, 0]  # class token
+    else:
+        x = x if reduce_include_prefix else x[:, num_prefix_tokens:]
+        if pool_type == 'avg':
+            x = x.mean(dim=1)
+        elif pool_type == 'avgmax':
+            x = 0.5 * (x.amax(dim=1) + x.mean(dim=1))
+        elif pool_type == 'max':
+            x = x.amax(dim=1)
+        else:
+            assert not pool_type, f'Unknown pool type {pool_type}'
+
+    return x
+
+
 class VisionTransformer(nn.Module):
     """ Vision Transformer
 
@@ -400,7 +425,7 @@ class VisionTransformer(nn.Module):
             patch_size: Union[int, Tuple[int, int]] = 16,
             in_chans: int = 3,
             num_classes: int = 1000,
-            global_pool: Literal['', 'avg', 'token', 'map'] = 'token',
+            global_pool: Literal['', 'avg', 'avgmax', 'max', 'token', 'map'] = 'token',
             embed_dim: int = 768,
             depth: int = 12,
             num_heads: int = 12,
@@ -459,10 +484,10 @@ class VisionTransformer(nn.Module):
             block_fn: Transformer block layer.
         """
         super().__init__()
-        assert global_pool in ('', 'avg', 'token', 'map')
+        assert global_pool in ('', 'avg', 'avgmax', 'max', 'token', 'map')
         assert class_token or global_pool != 'token'
         assert pos_embed in ('', 'none', 'learn')
-        use_fc_norm = global_pool == 'avg' if fc_norm is None else fc_norm
+        use_fc_norm = global_pool in ('avg', 'avgmax', 'max') if fc_norm is None else fc_norm
         norm_layer = get_norm_layer(norm_layer) or partial(nn.LayerNorm, eps=1e-6)
         act_layer = get_act_layer(act_layer) or nn.GELU
 
@@ -596,10 +621,10 @@ class VisionTransformer(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.head
 
-    def reset_classifier(self, num_classes: int, global_pool = None) -> None:
+    def reset_classifier(self, num_classes: int, global_pool: Optional[str] = None):
         self.num_classes = num_classes
         if global_pool is not None:
-            assert global_pool in ('', 'avg', 'token', 'map')
+            assert global_pool in ('', 'avg', 'avgmax', 'max', 'token', 'map')
             if global_pool == 'map' and self.attn_pool is None:
                 assert False, "Cannot currently add attention pooling in reset_classifier()."
             elif global_pool != 'map ' and self.attn_pool is not None:
@@ -756,13 +781,16 @@ class VisionTransformer(nn.Module):
         x = self.norm(x)
         return x
 
-    def forward_head(self, x: torch.Tensor, pre_logits: bool = False) -> torch.Tensor:
+    def pool(self, x: torch.Tensor, pool_type: Optional[str] = None) -> torch.Tensor:
         if self.attn_pool is not None:
             x = self.attn_pool(x)
-        elif self.global_pool == 'avg':
-            x = x[:, self.num_prefix_tokens:].mean(dim=1)
-        elif self.global_pool:
-            x = x[:, 0]  # class token
+            return x
+        pool_type = self.global_pool if pool_type is None else pool_type
+        x = global_pool_nlc(x, pool_type=pool_type, num_prefix_tokens=self.num_prefix_tokens)
+        return x
+
+    def forward_head(self, x: torch.Tensor, pre_logits: bool = False) -> torch.Tensor:
+        x = self.pool(x)
         x = self.fc_norm(x)
         x = self.head_drop(x)
         return x if pre_logits else self.head(x)

timm/models/vision_transformer_relpos.py

@@ -381,7 +381,7 @@ class VisionTransformerRelPos(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.head
 
-    def reset_classifier(self, num_classes: int, global_pool=None):
+    def reset_classifier(self, num_classes: int, global_pool: Optional[str] = None):
         self.num_classes = num_classes
         if global_pool is not None:
             assert global_pool in ('', 'avg', 'token')

timm/models/vision_transformer_sam.py

@@ -536,7 +536,7 @@ class VisionTransformerSAM(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.head
 
-    def reset_classifier(self, num_classes=0, global_pool=None):
+    def reset_classifier(self, num_classes: int, global_pool: Optional[str] = None):
         self.head.reset(num_classes, global_pool)
 
     def forward_intermediates(

timm/models/vovnet.py

@@ -11,7 +11,7 @@ for some reference, rewrote most of the code.
 Hacked together by / Copyright 2020 Ross Wightman
 """
 
-from typing import List
+from typing import List, Optional
 
 import torch
 import torch.nn as nn
@@ -134,9 +134,17 @@ class OsaStage(nn.Module):
             else:
                 drop_path = None
             blocks += [OsaBlock(
-                in_chs, mid_chs, out_chs, layer_per_block, residual=residual and i > 0, depthwise=depthwise,
-                attn=attn if last_block else '', norm_layer=norm_layer, act_layer=act_layer, drop_path=drop_path)
-            ]
+                in_chs,
+                mid_chs,
+                out_chs,
+                layer_per_block,
+                residual=residual and i > 0,
+                depthwise=depthwise,
+                attn=attn if last_block else '',
+                norm_layer=norm_layer,
+                act_layer=act_layer,
+                drop_path=drop_path
+            )]
             in_chs = out_chs
         self.blocks = nn.Sequential(*blocks)
 
@@ -252,8 +260,9 @@ class VovNet(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.head.fc
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
-        self.head = ClassifierHead(self.num_features, num_classes, pool_type=global_pool, drop_rate=self.drop_rate)
+    def reset_classifier(self, num_classes, global_pool: Optional[str] = None):
+        self.num_classes = num_classes
+        self.head.reset(num_classes, global_pool)
 
     def forward_features(self, x):
         x = self.stem(x)

timm/models/xception.py

@@ -174,7 +174,7 @@ class Xception(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.fc
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: str = 'avg'):
         self.num_classes = num_classes
         self.global_pool, self.fc = create_classifier(self.num_features, self.num_classes, pool_type=global_pool)
 

timm/models/xception_aligned.py

@@ -274,7 +274,7 @@ class XceptionAligned(nn.Module):
     def get_classifier(self) -> nn.Module:
         return self.head.fc
 
-    def reset_classifier(self, num_classes, global_pool='avg'):
+    def reset_classifier(self, num_classes: int, global_pool: Optional[str] = None):
         self.head.reset(num_classes, pool_type=global_pool)
 
     def forward_features(self, x):