MaskFormer - enable return_dict in order to compile (#25052)

* Enable return_dict in order to compile

* Update tests

src/transformers/models/maskformer/modeling_maskformer.py

@@ -1254,11 +1254,16 @@ class MaskFormerPixelDecoder(nn.Module):
         self.fpn = MaskFormerFPNModel(*args, feature_size=feature_size, **kwargs)
         self.mask_projection = nn.Conv2d(feature_size, mask_feature_size, kernel_size=3, padding=1)
 
-    def forward(self, features: List[Tensor], output_hidden_states: bool = False) -> MaskFormerPixelDecoderOutput:
+    def forward(
+        self, features: List[Tensor], output_hidden_states: bool = False, return_dict: bool = True
+    ) -> MaskFormerPixelDecoderOutput:
         fpn_features = self.fpn(features)
         # we use the last feature map
         last_feature_projected = self.mask_projection(fpn_features[-1])
 
+        if not return_dict:
+            return (last_feature_projected, tuple(fpn_features)) if output_hidden_states else (last_feature_projected,)
+
         return MaskFormerPixelDecoderOutput(
             last_hidden_state=last_feature_projected, hidden_states=tuple(fpn_features) if output_hidden_states else ()
         )
@@ -1387,9 +1392,20 @@ class MaskFormerPixelLevelModule(nn.Module):
             lateral_widths=feature_channels[:-1],
         )
 
-    def forward(self, pixel_values: Tensor, output_hidden_states: bool = False) -> MaskFormerPixelLevelModuleOutput:
+    def forward(
+        self, pixel_values: Tensor, output_hidden_states: bool = False, return_dict: bool = True
+    ) -> MaskFormerPixelLevelModuleOutput:
         features = self.encoder(pixel_values).feature_maps
-        decoder_output = self.decoder(features, output_hidden_states)
+        decoder_output = self.decoder(features, output_hidden_states, return_dict=return_dict)
+
+        if not return_dict:
+            last_hidden_state = decoder_output[0]
+            outputs = (features[-1], last_hidden_state)
+            if output_hidden_states:
+                hidden_states = decoder_output[1]
+                outputs = outputs + (tuple(features),) + (hidden_states,)
+            return outputs
+
         return MaskFormerPixelLevelModuleOutput(
             # the last feature is actually the output from the last layer
             encoder_last_hidden_state=features[-1],
@@ -1414,7 +1430,11 @@ class MaskFormerTransformerModule(nn.Module):
         self.decoder = DetrDecoder(config=config.decoder_config)
 
     def forward(
-        self, image_features: Tensor, output_hidden_states: bool = False, output_attentions: bool = False
+        self,
+        image_features: Tensor,
+        output_hidden_states: bool = False,
+        output_attentions: bool = False,
+        return_dict: Optional[bool] = None,
     ) -> DetrDecoderOutput:
         if self.input_projection is not None:
             image_features = self.input_projection(image_features)
@@ -1438,7 +1458,7 @@ class MaskFormerTransformerModule(nn.Module):
             query_position_embeddings=queries_embeddings,
             output_attentions=output_attentions,
             output_hidden_states=output_hidden_states,
-            return_dict=None,
+            return_dict=return_dict,
         )
         return decoder_output
 
@@ -1593,9 +1613,11 @@ class MaskFormerModel(MaskFormerPreTrainedModel):
         if pixel_mask is None:
             pixel_mask = torch.ones((batch_size, height, width), device=pixel_values.device)
 
-        pixel_level_module_output = self.pixel_level_module(pixel_values, output_hidden_states)
+        pixel_level_module_output = self.pixel_level_module(
-        image_features = pixel_level_module_output.encoder_last_hidden_state
+            pixel_values, output_hidden_states, return_dict=return_dict
-        pixel_embeddings = pixel_level_module_output.decoder_last_hidden_state
+        )
+        image_features = pixel_level_module_output[0]
+        pixel_embeddings = pixel_level_module_output[1]
 
         transformer_module_output = self.transformer_module(image_features, output_hidden_states, output_attentions)
         queries = transformer_module_output.last_hidden_state
@@ -1606,9 +1628,9 @@ class MaskFormerModel(MaskFormerPreTrainedModel):
         hidden_states = None
 
         if output_hidden_states:
-            encoder_hidden_states = pixel_level_module_output.encoder_hidden_states
+            encoder_hidden_states = pixel_level_module_output[2]
-            pixel_decoder_hidden_states = pixel_level_module_output.decoder_hidden_states
+            pixel_decoder_hidden_states = pixel_level_module_output[3]
-            transformer_decoder_hidden_states = transformer_module_output.hidden_states
+            transformer_decoder_hidden_states = transformer_module_output[1]
             hidden_states = encoder_hidden_states + pixel_decoder_hidden_states + transformer_decoder_hidden_states
 
         output = MaskFormerModelOutput(
@@ -1803,13 +1825,25 @@ class MaskFormerForInstanceSegmentation(MaskFormerPreTrainedModel):
         )
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
-        outputs: MaskFormerModelOutput = self.model(
+        raw_outputs = self.model(
             pixel_values,
             pixel_mask,
             output_hidden_states=output_hidden_states or self.config.use_auxiliary_loss,
-            return_dict=True,
+            return_dict=return_dict,
             output_attentions=output_attentions,
         )
+        # We need to have raw_outputs optionally be returned as a dict to use torch.compile. For backwards
+        # compatibility we convert to a dataclass for the rest of the model logic
+        outputs = MaskFormerModelOutput(
+            encoder_last_hidden_state=raw_outputs[0],
+            pixel_decoder_last_hidden_state=raw_outputs[1],
+            transformer_decoder_last_hidden_state=raw_outputs[2],
+            encoder_hidden_states=raw_outputs[3] if output_hidden_states else None,
+            pixel_decoder_hidden_states=raw_outputs[4] if output_hidden_states else None,
+            transformer_decoder_hidden_states=raw_outputs[5] if output_hidden_states else None,
+            hidden_states=raw_outputs[6] if output_hidden_states else None,
+            attentions=raw_outputs[-1] if output_attentions else None,
+        )
 
         loss, loss_dict, auxiliary_logits = None, None, None
 
@@ -1827,16 +1861,18 @@ class MaskFormerForInstanceSegmentation(MaskFormerPreTrainedModel):
         if not output_auxiliary_logits:
             auxiliary_logits = None
 
-        output = MaskFormerForInstanceSegmentationOutput(
+        if not return_dict:
+            output = tuple(
+                v
+                for v in (loss, class_queries_logits, masks_queries_logits, auxiliary_logits, *outputs.values())
+                if v is not None
+            )
+            return output
+
+        return MaskFormerForInstanceSegmentationOutput(
             loss=loss,
             **outputs,
             class_queries_logits=class_queries_logits,
             masks_queries_logits=masks_queries_logits,
             auxiliary_logits=auxiliary_logits,
         )
-
-        if not return_dict:
-            output = tuple(v for v in output.values())
-            if loss is not None:
-                output = ((loss)) + output
-        return output

tests/models/maskformer/test_modeling_maskformer.py

@@ -14,6 +14,7 @@
 # limitations under the License.
 """ Testing suite for the PyTorch MaskFormer model. """
 
+import copy
 import inspect
 import unittest
 
@@ -54,6 +55,8 @@ class MaskFormerModelTester:
         max_size=32 * 6,
         num_labels=4,
         mask_feature_size=32,
+        num_hidden_layers=2,
+        num_attention_heads=2,
     ):
         self.parent = parent
         self.batch_size = batch_size
@@ -65,6 +68,9 @@ class MaskFormerModelTester:
         self.max_size = max_size
         self.num_labels = num_labels
         self.mask_feature_size = mask_feature_size
+        # This is passed to the decoder config. We add it to the model tester here for testing
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
 
     def prepare_config_and_inputs(self):
         pixel_values = floats_tensor([self.batch_size, self.num_channels, self.min_size, self.max_size]).to(
@@ -91,11 +97,12 @@ class MaskFormerModelTester:
             ),
             decoder_config=DetrConfig(
                 decoder_ffn_dim=64,
-                decoder_layers=2,
+                decoder_layers=self.num_hidden_layers,
+                decoder_attention_heads=self.num_attention_heads,
                 encoder_ffn_dim=64,
-                encoder_layers=2,
+                encoder_layers=self.num_hidden_layers,
+                encoder_attention_heads=self.num_attention_heads,
                 num_queries=self.num_queries,
-                decoder_attention_heads=2,
                 d_model=self.mask_feature_size,
             ),
             mask_feature_size=self.mask_feature_size,
@@ -196,6 +203,27 @@ class MaskFormerModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCa
         self.model_tester = MaskFormerModelTester(self)
         self.config_tester = ConfigTester(self, config_class=MaskFormerConfig, has_text_modality=False)
 
+    def _prepare_for_class(self, inputs_dict, model_class, return_labels=False):
+        inputs_dict = copy.deepcopy(inputs_dict)
+
+        if return_labels:
+            if model_class in [MaskFormerForInstanceSegmentation]:
+                inputs_dict["mask_labels"] = torch.zeros(
+                    (
+                        self.model_tester.batch_size,
+                        self.model_tester.num_labels,
+                        self.model_tester.min_size,
+                        self.model_tester.max_size,
+                    ),
+                    dtype=torch.float32,
+                    device=torch_device,
+                )
+                inputs_dict["class_labels"] = torch.zeros(
+                    (self.model_tester.batch_size, self.model_tester.num_labels), dtype=torch.long, device=torch_device
+                )
+
+        return inputs_dict
+
     def test_config(self):
         self.config_tester.run_common_tests()
 
@@ -265,26 +293,47 @@ class MaskFormerModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCa
         self.model_tester.create_and_check_maskformer_model(config, **inputs, output_hidden_states=True)
 
     def test_attention_outputs(self):
-        config, inputs = self.model_tester.prepare_config_and_inputs_for_common()
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+        config.return_dict = True
 
         for model_class in self.all_model_classes:
-            model = model_class(config).to(torch_device)
+            inputs_dict["output_attentions"] = True
-            outputs = model(**inputs, output_attentions=True)
+            inputs_dict["output_hidden_states"] = False
-            self.assertTrue(outputs.attentions is not None)
+            config.return_dict = True
-
-    def test_training(self):
-        if not self.model_tester.is_training:
-            return
-        # only MaskFormerForInstanceSegmentation has the loss
-        model_class = self.all_model_classes[1]
-        config, pixel_values, pixel_mask, mask_labels, class_labels = self.model_tester.prepare_config_and_inputs()
-
             model = model_class(config)
             model.to(torch_device)
-        model.train()
+            model.eval()
+            with torch.no_grad():
+                outputs = model(**self._prepare_for_class(inputs_dict, model_class))
+            attentions = outputs.attentions
+            self.assertEqual(len(attentions), self.model_tester.num_hidden_layers)
 
-        loss = model(pixel_values, mask_labels=mask_labels, class_labels=class_labels).loss
+            # Check that output_attentions also work using config
-        loss.backward()
+            del inputs_dict["output_attentions"]
+            config.output_attentions = True
+            model = model_class(config)
+            model.to(torch_device)
+            model.eval()
+            with torch.no_grad():
+                outputs = model(**self._prepare_for_class(inputs_dict, model_class))
+            attentions = outputs.attentions
+            self.assertEqual(len(attentions), self.model_tester.num_hidden_layers)
+            out_len = len(outputs)
+
+            # Check attention is always last and order is fine
+            inputs_dict["output_attentions"] = True
+            inputs_dict["output_hidden_states"] = True
+            model = model_class(config)
+            model.to(torch_device)
+            model.eval()
+            with torch.no_grad():
+                outputs = model(**self._prepare_for_class(inputs_dict, model_class))
+            # encoder_hidden_states, pixel_decoder_hidden_states, transformer_decoder_hidden_states, hidden_states
+            added_hidden_states = 4
+            self.assertEqual(out_len + added_hidden_states, len(outputs))
+
+            self_attentions = outputs.attentions
+            self.assertEqual(len(self_attentions), self.model_tester.num_hidden_layers)
 
     def test_retain_grad_hidden_states_attentions(self):
         # only MaskFormerForInstanceSegmentation has the loss