Enable more test_torchscript (#16679)

* update _create_and_check_torchscript

* Enable test_torchscript

* clear_class_registry

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

tests/beit/test_modeling_beit.py

@@ -192,7 +192,6 @@ class BeitModelTest(ModelTesterMixin, unittest.TestCase):
     )
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
 

tests/canine/test_modeling_canine.py

@@ -219,7 +219,6 @@ class CanineModelTest(ModelTesterMixin, unittest.TestCase):
         else ()
     )
 
-    test_torchscript = False
     test_mismatched_shapes = False
     test_resize_embeddings = False
     test_pruning = False

tests/clip/test_modeling_clip.py

@@ -151,7 +151,6 @@ class CLIPVisionModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (CLIPVisionModel,) if is_torch_available() else ()
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
 

tests/convnext/test_modeling_convnext.py

@@ -138,7 +138,6 @@ class ConvNextModelTest(ModelTesterMixin, unittest.TestCase):
     )
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
     has_attentions = False

tests/ctrl/test_modeling_ctrl.py

@@ -174,7 +174,6 @@ class CTRLModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase):
     all_model_classes = (CTRLModel, CTRLLMHeadModel, CTRLForSequenceClassification) if is_torch_available() else ()
     all_generative_model_classes = (CTRLLMHeadModel,) if is_torch_available() else ()
     test_pruning = True
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
 

tests/data2vec/test_modeling_data2vec_audio.py

@@ -372,7 +372,6 @@ class Data2VecAudioModelTest(ModelTesterMixin, unittest.TestCase):
     )
     test_pruning = False
     test_headmasking = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = Data2VecAudioModelTester(self)

tests/decision_transformer/test_modeling_decision_transformer.py

@@ -148,6 +148,7 @@ class DecisionTransformerModelTest(ModelTesterMixin, GenerationTesterMixin, unit
     test_inputs_embeds = False
     test_model_common_attributes = False
     test_gradient_checkpointing = False
+    test_torchscript = False
 
     def setUp(self):
         self.model_tester = DecisionTransformerModelTester(self)

tests/deit/test_modeling_deit.py

@@ -169,7 +169,6 @@ class DeiTModelTest(ModelTesterMixin, unittest.TestCase):
     )
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
 

tests/distilbert/test_modeling_distilbert.py

@@ -211,7 +211,6 @@ class DistilBertModelTest(ModelTesterMixin, unittest.TestCase):
     )
     fx_compatible = True
     test_pruning = True
-    test_torchscript = True
     test_resize_embeddings = True
     test_resize_position_embeddings = True
 

tests/dpt/test_modeling_dpt.py

@@ -154,7 +154,6 @@ class DPTModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (DPTModel, DPTForDepthEstimation, DPTForSemanticSegmentation) if is_torch_available() else ()
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
 

tests/fnet/test_modeling_fnet.py

@@ -284,7 +284,6 @@ class FNetModelTest(ModelTesterMixin, unittest.TestCase):
 
     # Skip Tests
     test_pruning = False
-    test_torchscript = False
     test_head_masking = False
     test_pruning = False
 

tests/glpn/test_modeling_glpn.py

@@ -150,7 +150,6 @@ class GLPNModelTest(ModelTesterMixin, unittest.TestCase):
     test_head_masking = False
     test_pruning = False
     test_resize_embeddings = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = GLPNModelTester(self)

tests/hubert/test_modeling_hubert.py

@@ -300,7 +300,6 @@ class HubertModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (HubertForCTC, HubertForSequenceClassification, HubertModel) if is_torch_available() else ()
     test_pruning = False
     test_headmasking = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = HubertModelTester(self)
@@ -445,7 +444,6 @@ class HubertRobustModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (HubertForCTC, HubertForSequenceClassification, HubertModel) if is_torch_available() else ()
     test_pruning = False
     test_headmasking = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = HubertModelTester(

tests/maskformer/test_modeling_maskformer.py

@@ -177,7 +177,6 @@ class MaskFormerModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (MaskFormerModel, MaskFormerForInstanceSegmentation) if is_torch_available() else ()
 
     is_encoder_decoder = False
-    test_torchscript = False
     test_pruning = False
     test_head_masking = False
     test_missing_keys = False

tests/mpnet/test_modeling_mpnet.py

@@ -205,7 +205,6 @@ class MPNetModelTest(ModelTesterMixin, unittest.TestCase):
         else ()
     )
     test_pruning = False
-    test_torchscript = True
     test_resize_embeddings = True
 
     def setUp(self):

tests/prophetnet/test_modeling_prophetnet.py

@@ -890,7 +890,6 @@ class ProphetNetModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.Test
     all_model_classes = (ProphetNetModel, ProphetNetForConditionalGeneration) if is_torch_available() else ()
     all_generative_model_classes = (ProphetNetForConditionalGeneration,) if is_torch_available() else ()
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     is_encoder_decoder = True
 
@@ -1100,7 +1099,7 @@ class ProphetNetStandaloneDecoderModelTest(ModelTesterMixin, GenerationTesterMix
     all_model_classes = (ProphetNetDecoder, ProphetNetForCausalLM) if is_torch_available() else ()
     all_generative_model_classes = (ProphetNetForCausalLM,) if is_torch_available() else ()
     test_pruning = False
-    test_torchscript = False
+
     test_resize_embeddings = False
     is_encoder_decoder = False
 
@@ -1128,7 +1127,7 @@ class ProphetNetStandaloneDecoderModelTest(ModelTesterMixin, GenerationTesterMix
 class ProphetNetStandaloneEncoderModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (ProphetNetEncoder,) if is_torch_available() else ()
     test_pruning = False
-    test_torchscript = False
+
     test_resize_embeddings = False
     is_encoder_decoder = False
 

tests/regnet/test_modeling_regnet.py

@@ -127,7 +127,6 @@ class RegNetModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (RegNetModel, RegNetForImageClassification) if is_torch_available() else ()
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
     has_attentions = False

tests/resnet/test_modeling_resnet.py

@@ -127,7 +127,6 @@ class ResNetModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (ResNetModel, ResNetForImageClassification) if is_torch_available() else ()
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
     has_attentions = False

tests/segformer/test_modeling_segformer.py

@@ -165,7 +165,6 @@ class SegformerModelTest(ModelTesterMixin, unittest.TestCase):
     test_head_masking = False
     test_pruning = False
     test_resize_embeddings = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = SegformerModelTester(self)

tests/sew/test_modeling_sew.py

@@ -303,7 +303,6 @@ class SEWModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (SEWForCTC, SEWModel, SEWForSequenceClassification) if is_torch_available() else ()
     test_pruning = False
     test_headmasking = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = SEWModelTester(self)

tests/speech_to_text/test_modeling_speech_to_text.py

@@ -273,7 +273,6 @@ class Speech2TextModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.Tes
     is_encoder_decoder = True
     test_pruning = False
     test_missing_keys = False
-    test_torchscript = True
 
     input_name = "input_features"
 

tests/squeezebert/test_modeling_squeezebert.py

@@ -229,7 +229,6 @@ class SqueezeBertModelTest(ModelTesterMixin, unittest.TestCase):
         else None
     )
     test_pruning = False
-    test_torchscript = True
     test_resize_embeddings = True
     test_head_masking = False
 

tests/swin/test_modeling_swin.py

@@ -177,7 +177,6 @@ class SwinModelTest(ModelTesterMixin, unittest.TestCase):
     )
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
 

tests/t5/test_modeling_t5.py

@@ -512,7 +512,6 @@ class T5ModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase):
     fx_compatible = True
     all_parallelizable_model_classes = (T5Model, T5ForConditionalGeneration) if is_torch_available() else ()
     test_pruning = False
-    test_torchscript = True
     test_resize_embeddings = True
     test_model_parallel = True
     is_encoder_decoder = True
@@ -777,7 +776,6 @@ class T5EncoderOnlyModelTester:
 class T5EncoderOnlyModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (T5EncoderModel,) if is_torch_available() else ()
     test_pruning = False
-    test_torchscript = True
     test_resize_embeddings = False
     test_model_parallel = True
     all_parallelizable_model_classes = (T5EncoderModel,) if is_torch_available() else ()

tests/tapas/test_modeling_tapas.py

@@ -422,7 +422,6 @@ class TapasModelTest(ModelTesterMixin, unittest.TestCase):
         else None
     )
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = True
     test_head_masking = False
 

tests/test_modeling_common.py

@@ -617,19 +617,21 @@ class ModelTesterMixin:
             model.eval()
             inputs = self._prepare_for_class(inputs_dict, model_class)
 
+            main_input_name = model_class.main_input_name
+
             try:
                 if model.config.is_encoder_decoder:
                     model.config.use_cache = False  # FSTM still requires this hack -> FSTM should probably be refactored similar to BART afterward
-                    input_ids = inputs["input_ids"]
+                    main_input = inputs[main_input_name]
                     attention_mask = inputs["attention_mask"]
                     decoder_input_ids = inputs["decoder_input_ids"]
                     decoder_attention_mask = inputs["decoder_attention_mask"]
                     traced_model = torch.jit.trace(
-                        model, (input_ids, attention_mask, decoder_input_ids, decoder_attention_mask)
+                        model, (main_input, attention_mask, decoder_input_ids, decoder_attention_mask)
                     )
                 else:
-                    input_ids = inputs["input_ids"]
+                    main_input = inputs[main_input_name]
-                    traced_model = torch.jit.trace(model, input_ids)
+                    traced_model = torch.jit.trace(model, main_input)
             except RuntimeError:
                 self.fail("Couldn't trace module.")
 

tests/transfo_xl/test_modeling_transfo_xl.py

@@ -238,7 +238,6 @@ class TransfoXLModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestC
     )
     all_generative_model_classes = (TransfoXLLMHeadModel,) if is_torch_available() else ()
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = True
     test_mismatched_shapes = False
 

tests/unispeech/test_modeling_unispeech.py

@@ -305,7 +305,6 @@ class UniSpeechRobustModelTest(ModelTesterMixin, unittest.TestCase):
     )
     test_pruning = False
     test_headmasking = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = UniSpeechModelTester(

tests/van/test_modeling_van.py

@@ -124,7 +124,6 @@ class VanModelTest(ModelTesterMixin, unittest.TestCase):
     all_model_classes = (VanModel, VanForImageClassification) if is_torch_available() else ()
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
     has_attentions = False

tests/vit/test_modeling_vit.py

@@ -158,7 +158,6 @@ class ViTModelTest(ModelTesterMixin, unittest.TestCase):
     )
 
     test_pruning = False
-    test_torchscript = False
     test_resize_embeddings = False
     test_head_masking = False
 

tests/wav2vec2/test_modeling_wav2vec2.py

@@ -413,7 +413,6 @@ class Wav2Vec2ModelTest(ModelTesterMixin, unittest.TestCase):
     )
     test_pruning = False
     test_headmasking = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = Wav2Vec2ModelTester(self)
@@ -652,7 +651,6 @@ class Wav2Vec2RobustModelTest(ModelTesterMixin, unittest.TestCase):
     )
     test_pruning = False
     test_headmasking = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = Wav2Vec2ModelTester(

tests/wavlm/test_modeling_wavlm.py

@@ -316,7 +316,6 @@ class WavLMModelTest(ModelTesterMixin, unittest.TestCase):
     )
     test_pruning = False
     test_headmasking = False
-    test_torchscript = False
 
     def setUp(self):
         self.model_tester = WavLMModelTester(self)