[Backbones] Improve out features (#20675)

* Improve ResNet backbone

* Improve Bit backbone

* Improve docstrings

* Fix default stage

* Apply suggestions from code review

Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>

src/transformers/models/bit/configuration_bit.py

@@ -63,7 +63,7 @@ class BitConfig(PretrainedConfig):
             The width factor for the model.
         out_features (`List[str]`, *optional*):
             If used as backbone, list of features to output. Can be any of `"stem"`, `"stage1"`, `"stage2"`, etc.
-            (depending on how many stages the model has).
+            (depending on how many stages the model has). Will default to the last stage if unset.
 
     Example:
     ```python

src/transformers/models/bit/modeling_bit.py

@@ -851,7 +851,7 @@ class BitBackbone(BitPreTrainedModel, BackboneMixin):
         self.stage_names = config.stage_names
         self.bit = BitModel(config)
 
-        self.out_features = config.out_features
+        self.out_features = config.out_features if config.out_features is not None else [self.stage_names[-1]]
 
         out_feature_channels = {}
         out_feature_channels["stem"] = config.embedding_size

src/transformers/models/maskformer/configuration_maskformer_swin.py

@@ -69,7 +69,8 @@ class MaskFormerSwinConfig(PretrainedConfig):
         layer_norm_eps (`float`, *optional*, defaults to 1e-12):
             The epsilon used by the layer normalization layers.
         out_features (`List[str]`, *optional*):
-            If used as a backbone, list of feature names to output, e.g. `["stage1", "stage2"]`.
+            If used as backbone, list of features to output. Can be any of `"stem"`, `"stage1"`, `"stage2"`, etc.
+            (depending on how many stages the model has). Will default to the last stage if unset.
 
     Example:
 

src/transformers/models/maskformer/modeling_maskformer_swin.py

@@ -855,7 +855,7 @@ class MaskFormerSwinBackbone(MaskFormerSwinPreTrainedModel, BackboneMixin):
         self.stage_names = config.stage_names
         self.model = MaskFormerSwinModel(config)
 
-        self.out_features = config.out_features
+        self.out_features = config.out_features if config.out_features is not None else [self.stage_names[-1]]
         if "stem" in self.out_features:
             raise ValueError("This backbone does not support 'stem' in the `out_features`.")
 

src/transformers/models/resnet/configuration_resnet.py

@@ -59,8 +59,8 @@ class ResNetConfig(PretrainedConfig):
         downsample_in_first_stage (`bool`, *optional*, defaults to `False`):
             If `True`, the first stage will downsample the inputs using a `stride` of 2.
         out_features (`List[str]`, *optional*):
-            If used as backbone, list of features to output. Can be any of `"stem"`, `"stage1"`, `"stage2"`,
-            `"stage3"`, `"stage4"`.
+            If used as backbone, list of features to output. Can be any of `"stem"`, `"stage1"`, `"stage2"`, etc.
+            (depending on how many stages the model has). Will default to the last stage if unset.
 
     Example:
     ```python

src/transformers/models/resnet/modeling_resnet.py

@@ -267,7 +267,7 @@ class ResNetPreTrainedModel(PreTrainedModel):
             nn.init.constant_(module.bias, 0)
 
     def _set_gradient_checkpointing(self, module, value=False):
-        if isinstance(module, (ResNetModel, ResNetBackbone)):
+        if isinstance(module, ResNetEncoder):
             module.gradient_checkpointing = value
 
 
@@ -439,7 +439,7 @@ class ResNetBackbone(ResNetPreTrainedModel, BackboneMixin):
         self.embedder = ResNetEmbeddings(config)
         self.encoder = ResNetEncoder(config)
 
-        self.out_features = config.out_features
+        self.out_features = config.out_features if config.out_features is not None else [self.stage_names[-1]]
 
         out_feature_channels = {}
         out_feature_channels["stem"] = config.embedding_size

tests/models/bit/test_modeling_bit.py

@@ -119,7 +119,7 @@ class BitModelTester:
         model.eval()
         result = model(pixel_values)
 
-        # verify hidden states
+        # verify feature maps
         self.parent.assertEqual(len(result.feature_maps), len(config.out_features))
         self.parent.assertListEqual(list(result.feature_maps[0].shape), [self.batch_size, self.hidden_sizes[1], 4, 4])
 
@@ -127,6 +127,21 @@ class BitModelTester:
         self.parent.assertEqual(len(model.channels), len(config.out_features))
         self.parent.assertListEqual(model.channels, config.hidden_sizes[1:])
 
+        # verify backbone works with out_features=None
+        config.out_features = None
+        model = BitBackbone(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(pixel_values)
+
+        # verify feature maps
+        self.parent.assertEqual(len(result.feature_maps), 1)
+        self.parent.assertListEqual(list(result.feature_maps[0].shape), [self.batch_size, self.hidden_sizes[-1], 1, 1])
+
+        # verify channels
+        self.parent.assertEqual(len(model.channels), 1)
+        self.parent.assertListEqual(model.channels, [config.hidden_sizes[-1]])
+
     def prepare_config_and_inputs_for_common(self):
         config_and_inputs = self.prepare_config_and_inputs()
         config, pixel_values, labels = config_and_inputs

tests/models/resnet/test_modeling_resnet.py

@@ -119,7 +119,7 @@ class ResNetModelTester:
         model.eval()
         result = model(pixel_values)
 
-        # verify hidden states
+        # verify feature maps
         self.parent.assertEqual(len(result.feature_maps), len(config.out_features))
         self.parent.assertListEqual(list(result.feature_maps[0].shape), [self.batch_size, self.hidden_sizes[1], 4, 4])
 
@@ -127,6 +127,21 @@ class ResNetModelTester:
         self.parent.assertEqual(len(model.channels), len(config.out_features))
         self.parent.assertListEqual(model.channels, config.hidden_sizes[1:])
 
+        # verify backbone works with out_features=None
+        config.out_features = None
+        model = ResNetBackbone(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(pixel_values)
+
+        # verify feature maps
+        self.parent.assertEqual(len(result.feature_maps), 1)
+        self.parent.assertListEqual(list(result.feature_maps[0].shape), [self.batch_size, self.hidden_sizes[-1], 1, 1])
+
+        # verify channels
+        self.parent.assertEqual(len(model.channels), 1)
+        self.parent.assertListEqual(model.channels, [config.hidden_sizes[-1]])
+
     def prepare_config_and_inputs_for_common(self):
         config_and_inputs = self.prepare_config_and_inputs()
         config, pixel_values, labels = config_and_inputs