Swin main layer (#17693)

* Swin models call TFSwinMainLayer

* Tidy up

src/transformers/models/swin/modeling_tf_swin.py

@@ -24,7 +24,13 @@ from typing import Any, Callable, Dict, Iterable, List, Optional, Tuple, Union
 import tensorflow as tf
 
 from ...activations_tf import ACT2FN
-from ...modeling_tf_utils import TFPreTrainedModel, TFSequenceClassificationLoss, get_initializer, unpack_inputs
+from ...modeling_tf_utils import (
+    TFPreTrainedModel,
+    TFSequenceClassificationLoss,
+    get_initializer,
+    keras_serializable,
+    unpack_inputs,
+)
 from ...tf_utils import shape_list
 from ...utils import (
     ModelOutput,
@@ -1069,15 +1075,14 @@ class AdaptiveAveragePooling1D(tf.keras.layers.Layer):
         return {**base_config, **config}
 
 
-@add_start_docstrings(
-    "The bare Swin Model transformer outputting raw hidden-states without any specific head on top.",
-    SWIN_START_DOCSTRING,
-)
-class TFSwinModel(TFSwinPreTrainedModel):
+@keras_serializable
+class TFSwinMainLayer(tf.keras.layers.Layer):
+    config_class = SwinConfig
+
     def __init__(
         self, config: SwinConfig, add_pooling_layer: bool = True, use_mask_token: bool = False, **kwargs
     ) -> None:
-        super().__init__(config, **kwargs)
+        super().__init__(**kwargs)
         self.config = config
         self.num_layers = len(config.depths)
         self.num_features = int(config.embed_dim * 2 ** (self.num_layers - 1))
@@ -1104,15 +1109,6 @@ class TFSwinModel(TFSwinPreTrainedModel):
             raise NotImplementedError
         return [None] * len(self.config.depths)
 
-    @add_start_docstrings_to_model_forward(SWIN_INPUTS_DOCSTRING)
-    @add_code_sample_docstrings(
-        processor_class=_FEAT_EXTRACTOR_FOR_DOC,
-        checkpoint=_CHECKPOINT_FOR_DOC,
-        output_type=TFSwinModelOutput,
-        config_class=_CONFIG_FOR_DOC,
-        modality="vision",
-        expected_output=_EXPECTED_OUTPUT_SHAPE,
-    )
     @unpack_inputs
     def call(
         self,
@@ -1175,6 +1171,60 @@ class TFSwinModel(TFSwinPreTrainedModel):
         )
 
 
+@add_start_docstrings(
+    "The bare Swin Model transformer outputting raw hidden-states without any specific head on top.",
+    SWIN_START_DOCSTRING,
+)
+class TFSwinModel(TFSwinPreTrainedModel):
+    def __init__(
+        self, config: SwinConfig, add_pooling_layer: bool = True, use_mask_token: bool = False, **kwargs
+    ) -> None:
+        super().__init__(config, **kwargs)
+        self.config = config
+        self.swin = TFSwinMainLayer(config, name="swin")
+
+    @add_start_docstrings_to_model_forward(SWIN_INPUTS_DOCSTRING)
+    @add_code_sample_docstrings(
+        processor_class=_FEAT_EXTRACTOR_FOR_DOC,
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=TFSwinModelOutput,
+        config_class=_CONFIG_FOR_DOC,
+        modality="vision",
+        expected_output=_EXPECTED_OUTPUT_SHAPE,
+    )
+    @unpack_inputs
+    def call(
+        self,
+        pixel_values: Optional[tf.Tensor] = None,
+        bool_masked_pos: Optional[tf.Tensor] = None,
+        head_mask: Optional[tf.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        training: bool = False,
+    ) -> Union[TFSwinModelOutput, Tuple[tf.Tensor, ...]]:
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if pixel_values is None:
+            raise ValueError("You have to specify pixel_values")
+
+        swin_outputs = self.swin(
+            pixel_values=pixel_values,
+            bool_masked_pos=bool_masked_pos,
+            head_mask=head_mask,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            training=training,
+        )
+
+        return swin_outputs
+
+
 class PixelShuffle(tf.keras.layers.Layer):
     """TF layer implementation of torch.nn.PixelShuffle"""
 
@@ -1238,7 +1288,7 @@ class TFSwinForMaskedImageModeling(TFSwinPreTrainedModel):
     def __init__(self, config: SwinConfig):
         super().__init__(config)
 
-        self.swin = TFSwinModel(config, add_pooling_layer=False, use_mask_token=True, name="swin")
+        self.swin = TFSwinMainLayer(config, add_pooling_layer=False, use_mask_token=True, name="swin")
 
         self.decoder = TFSwinDecoder(config, name="decoder")
 
@@ -1350,7 +1400,7 @@ class TFSwinForImageClassification(TFSwinPreTrainedModel, TFSequenceClassificati
         super().__init__(config)
 
         self.num_labels = config.num_labels
-        self.swin = TFSwinModel(config, name="swin")
+        self.swin = TFSwinMainLayer(config, name="swin")
 
         # Classifier head
         self.classifier = (