Update README. Fine-grained layer-wise lr decay working for tiny_vit and both efficientvits. Minor fixes.

Author: rwightman

README.md

@@ -35,6 +35,10 @@ And a big thanks to all GitHub sponsors who helped with some of my costs before
 * The Hugging Face Hub (https://huggingface.co/timm) is now the primary source for `timm` weights. Model cards include link to papers, original source, license. 
 * Previous 0.6.x can be cloned from [0.6.x](https://github.com/rwightman/pytorch-image-models/tree/0.6.x) branch or installed via pip with version.
 
+### Sep 1, 2023
+* TinyViT added by [SeeFun](https://github.com/seefun)
+* Fix EfficientViT (MIT) to use torch.autocast so it works back to PT 1.10
+
 ### Aug 28, 2023
 * Add dynamic img size support to models in `vision_transformer.py`, `vision_transformer_hybrid.py`, `deit.py`, and `eva.py` w/o breaking backward compat.
   * Add `dynamic_img_size=True` to args at model creation time to allow changing the grid size (interpolate abs and/or ROPE pos embed each forward pass).

timm/models/efficientvit_mit.py

@@ -508,11 +508,10 @@ def __init__(
 
         # stages
         self.feature_info = []
-        stages = []
-        stage_idx = 0
+        self.stages = nn.Sequential()
         in_channels = widths[0]
         for i, (w, d) in enumerate(zip(widths[1:], depths[1:])):
-            stages.append(EfficientVitStage(
+            self.stages.append(EfficientVitStage(
                 in_channels,
                 w,
                 depth=d,
@@ -524,10 +523,8 @@ def __init__(
             ))
             stride *= 2
             in_channels = w
-            self.feature_info += [dict(num_chs=in_channels, reduction=stride, module=f'stages.{stage_idx}')]
-            stage_idx += 1
+            self.feature_info += [dict(num_chs=in_channels, reduction=stride, module=f'stages.{i}')]
 
-        self.stages = nn.Sequential(*stages)
         self.num_features = in_channels
         self.head_widths = head_widths
         self.head_dropout = drop_rate
@@ -548,8 +545,11 @@ def __init__(
     @torch.jit.ignore
     def group_matcher(self, coarse=False):
         matcher = dict(
-            stem=r'^stem',  # stem and embed
-            blocks=[(r'^stages\.(\d+)', None)]
+            stem=r'^stem',
+            blocks=r'^stages\.(\d+)' if coarse else [
+                (r'^stages\.(\d+).downsample', (0,)),
+                (r'^stages\.(\d+)\.\w+\.(\d+)', None),
+            ]
         )
         return matcher
 

timm/models/efficientvit_msra.py

@@ -441,11 +441,18 @@ def __init__(
         self.head = NormLinear(
             self.num_features, num_classes, drop=self.drop_rate) if num_classes > 0 else torch.nn.Identity()
 
+    @torch.jit.ignore
+    def no_weight_decay(self):
+        return {x for x in self.state_dict().keys() if 'attention_biases' in x}
+
     @torch.jit.ignore
     def group_matcher(self, coarse=False):
         matcher = dict(
             stem=r'^patch_embed',
-            blocks=[(r'^stages\.(\d+)', None)]
+            blocks=r'^stages\.(\d+)' if coarse else [
+                (r'^stages\.(\d+).downsample', (0,)),
+                (r'^stages\.(\d+)\.\w+\.(\d+)', None),
+            ]
         )
         return matcher
 
@@ -455,7 +462,7 @@ def set_grad_checkpointing(self, enable=True):
 
     @torch.jit.ignore
     def get_classifier(self):
-        return self.head
+        return self.head.linear
 
     def reset_classifier(self, num_classes, global_pool=None):
         self.num_classes = num_classes

timm/models/tiny_vit.py

@@ -509,11 +509,18 @@ def _init_weights(self, m):
     def no_weight_decay_keywords(self):
         return {'attention_biases'}
 
+    @torch.jit.ignore
+    def no_weight_decay(self):
+        return {x for x in self.state_dict().keys() if 'attention_biases' in x}
+
     @torch.jit.ignore
     def group_matcher(self, coarse=False):
         matcher = dict(
             stem=r'^patch_embed',
-            blocks=[(r'^stages\.(\d+)', None)]
+            blocks=r'^stages\.(\d+)' if coarse else [
+                (r'^stages\.(\d+).downsample', (0,)),
+                (r'^stages\.(\d+)\.\w+\.(\d+)', None),
+            ]
         )
         return matcher
 
@@ -523,7 +530,7 @@ def set_grad_checkpointing(self, enable=True):
 
     @torch.jit.ignore
     def get_classifier(self):
-        return self.head
+        return self.head.fc
 
     def reset_classifier(self, num_classes, global_pool=None):
         self.num_classes = num_classes