Avoid using uncessary get_values(MODEL_MAPPING) (#29362)

* more fixes

* more fixes

---------

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

tests/models/beit/test_modeling_beit.py

@@ -21,7 +21,6 @@ from datasets import load_dataset
 from packaging import version
 
 from transformers import BeitConfig
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
 from transformers.utils import cached_property, is_torch_available, is_vision_available
 
@@ -36,14 +35,13 @@ if is_torch_available():
     from torch import nn
 
     from transformers import (
-        MODEL_FOR_BACKBONE_MAPPING,
-        MODEL_MAPPING,
         BeitBackbone,
         BeitForImageClassification,
         BeitForMaskedImageModeling,
         BeitForSemanticSegmentation,
         BeitModel,
     )
+    from transformers.models.auto.modeling_auto import MODEL_FOR_BACKBONE_MAPPING_NAMES, MODEL_MAPPING_NAMES
     from transformers.models.beit.modeling_beit import BEIT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -312,10 +310,10 @@ class BeitModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
 
         for model_class in self.all_model_classes:
             # we don't test BeitForMaskedImageModeling
-            if model_class in [
-                *get_values(MODEL_MAPPING),
-                *get_values(MODEL_FOR_BACKBONE_MAPPING),
-                BeitForMaskedImageModeling,
+            if model_class.__name__ in [
+                *MODEL_MAPPING_NAMES.values(),
+                *MODEL_FOR_BACKBONE_MAPPING_NAMES.values(),
+                "BeitForMaskedImageModeling",
             ]:
                 continue
 
@@ -337,8 +335,12 @@ class BeitModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
         for model_class in self.all_model_classes:
             # we don't test BeitForMaskedImageModeling
             if (
-                model_class
-                in [*get_values(MODEL_MAPPING), *get_values(MODEL_FOR_BACKBONE_MAPPING), BeitForMaskedImageModeling]
+                model_class.__name__
+                in [
+                    *MODEL_MAPPING_NAMES.values(),
+                    *MODEL_FOR_BACKBONE_MAPPING_NAMES.values(),
+                    "BeitForMaskedImageModeling",
+                ]
                 or not model_class.supports_gradient_checkpointing
             ):
                 continue

tests/models/clipseg/test_modeling_clipseg.py

@@ -24,8 +24,7 @@ import numpy as np
 import requests
 
 import transformers
-from transformers import MODEL_MAPPING, CLIPSegConfig, CLIPSegProcessor, CLIPSegTextConfig, CLIPSegVisionConfig
-from transformers.models.auto import get_values
+from transformers import CLIPSegConfig, CLIPSegProcessor, CLIPSegTextConfig, CLIPSegVisionConfig
 from transformers.testing_utils import (
     is_flax_available,
     is_pt_flax_cross_test,
@@ -52,6 +51,7 @@ if is_torch_available():
     from torch import nn
 
     from transformers import CLIPSegForImageSegmentation, CLIPSegModel, CLIPSegTextModel, CLIPSegVisionModel
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.clipseg.modeling_clipseg import CLIPSEG_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -751,7 +751,7 @@ class CLIPSegModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase)
             config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
             config.return_dict = True
 
-            if model_class in get_values(MODEL_MAPPING):
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values():
                 continue
 
             print("Model class:", model_class)

tests/models/data2vec/test_modeling_data2vec_vision.py

@@ -18,7 +18,6 @@
 import unittest
 
 from transformers import Data2VecVisionConfig
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
 from transformers.utils import cached_property, is_torch_available, is_vision_available
 
@@ -32,11 +31,11 @@ if is_torch_available():
     from torch import nn
 
     from transformers import (
-        MODEL_MAPPING,
         Data2VecVisionForImageClassification,
         Data2VecVisionForSemanticSegmentation,
         Data2VecVisionModel,
     )
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.data2vec.modeling_data2vec_vision import DATA2VEC_VISION_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -235,7 +234,7 @@ class Data2VecVisionModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.Te
         config.return_dict = True
 
         for model_class in self.all_model_classes:
-            if model_class in [*get_values(MODEL_MAPPING)]:
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values():
                 continue
 
             model = model_class(config)
@@ -254,7 +253,7 @@ class Data2VecVisionModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.Te
         config.return_dict = True
 
         for model_class in self.all_model_classes:
-            if model_class in [*get_values(MODEL_MAPPING)] or not model_class.supports_gradient_checkpointing:
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values() or not model_class.supports_gradient_checkpointing:
                 continue
             # TODO: remove the following 3 lines once we have a MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING
             # this can then be incorporated into _prepare_for_class in test_modeling_common.py

tests/models/deit/test_modeling_deit.py

@@ -19,7 +19,6 @@ import unittest
 import warnings
 
 from transformers import DeiTConfig
-from transformers.models.auto import get_values
 from transformers.testing_utils import (
     require_accelerate,
     require_torch,
@@ -41,14 +40,16 @@ if is_torch_available():
     from torch import nn
 
     from transformers import (
-        MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
-        MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
-        MODEL_MAPPING,
         DeiTForImageClassification,
         DeiTForImageClassificationWithTeacher,
         DeiTForMaskedImageModeling,
         DeiTModel,
     )
+    from transformers.models.auto.modeling_auto import (
+        MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES,
+        MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES,
+        MODEL_MAPPING_NAMES,
+    )
     from transformers.models.deit.modeling_deit import DEIT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -269,7 +270,7 @@ class DeiTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
         for model_class in self.all_model_classes:
             # DeiTForImageClassificationWithTeacher supports inference-only
             if (
-                model_class in get_values(MODEL_MAPPING)
+                model_class.__name__ in MODEL_MAPPING_NAMES.values()
                 or model_class.__name__ == "DeiTForImageClassificationWithTeacher"
             ):
                 continue
@@ -289,7 +290,7 @@ class DeiTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
         config.return_dict = True
 
         for model_class in self.all_model_classes:
-            if model_class in get_values(MODEL_MAPPING) or not model_class.supports_gradient_checkpointing:
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values() or not model_class.supports_gradient_checkpointing:
                 continue
             # DeiTForImageClassificationWithTeacher supports inference-only
             if model_class.__name__ == "DeiTForImageClassificationWithTeacher":
@@ -325,10 +326,10 @@ class DeiTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
 
         for model_class in self.all_model_classes:
             if (
-                model_class
+                model_class.__name__
                 not in [
-                    *get_values(MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING),
-                    *get_values(MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING),
+                    *MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES.values(),
+                    *MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES.values(),
                 ]
                 or model_class.__name__ == "DeiTForImageClassificationWithTeacher"
             ):

tests/models/dpt/test_modeling_dpt.py

@@ -19,7 +19,6 @@ import unittest
 
 from transformers import DPTConfig
 from transformers.file_utils import is_torch_available, is_vision_available
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_vision, slow, torch_device
 
 from ...test_configuration_common import ConfigTester
@@ -31,7 +30,8 @@ if is_torch_available():
     import torch
     from torch import nn
 
-    from transformers import MODEL_MAPPING, DPTForDepthEstimation, DPTForSemanticSegmentation, DPTModel
+    from transformers import DPTForDepthEstimation, DPTForSemanticSegmentation, DPTModel
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.dpt.modeling_dpt import DPT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -214,7 +214,7 @@ class DPTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
             config.return_dict = True
 
-            if model_class in get_values(MODEL_MAPPING):
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values():
                 continue
 
             model = model_class(config)
@@ -233,7 +233,7 @@ class DPTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             config.use_cache = False
             config.return_dict = True
 
-            if model_class in get_values(MODEL_MAPPING) or not model_class.supports_gradient_checkpointing:
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values() or not model_class.supports_gradient_checkpointing:
                 continue
             model = model_class(config)
             model.to(torch_device)

tests/models/dpt/test_modeling_dpt_auto_backbone.py

@@ -19,7 +19,6 @@ import unittest
 
 from transformers import Dinov2Config, DPTConfig
 from transformers.file_utils import is_torch_available, is_vision_available
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_vision, slow, torch_device
 
 from ...test_configuration_common import ConfigTester
@@ -30,7 +29,8 @@ from ...test_pipeline_mixin import PipelineTesterMixin
 if is_torch_available():
     import torch
 
-    from transformers import MODEL_MAPPING, DPTForDepthEstimation
+    from transformers import DPTForDepthEstimation
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.dpt.modeling_dpt import DPT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -166,7 +166,7 @@ class DPTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
             config.return_dict = True
 
-            if model_class in get_values(MODEL_MAPPING):
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values():
                 continue
 
             model = model_class(config)
@@ -185,7 +185,7 @@ class DPTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             config.use_cache = False
             config.return_dict = True
 
-            if model_class in get_values(MODEL_MAPPING) or not model_class.supports_gradient_checkpointing:
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values() or not model_class.supports_gradient_checkpointing:
                 continue
             model = model_class(config)
             model.to(torch_device)

tests/models/dpt/test_modeling_dpt_hybrid.py

@@ -19,7 +19,6 @@ import unittest
 
 from transformers import DPTConfig
 from transformers.file_utils import is_torch_available, is_vision_available
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_vision, slow, torch_device
 
 from ...test_configuration_common import ConfigTester
@@ -31,7 +30,8 @@ if is_torch_available():
     import torch
     from torch import nn
 
-    from transformers import MODEL_MAPPING, DPTForDepthEstimation, DPTForSemanticSegmentation, DPTModel
+    from transformers import DPTForDepthEstimation, DPTForSemanticSegmentation, DPTModel
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.dpt.modeling_dpt import DPT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -229,7 +229,7 @@ class DPTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
             config.return_dict = True
 
-            if model_class in get_values(MODEL_MAPPING):
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values():
                 continue
 
             model = model_class(config)
@@ -248,7 +248,7 @@ class DPTModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             config.use_cache = False
             config.return_dict = True
 
-            if model_class in get_values(MODEL_MAPPING) or not model_class.supports_gradient_checkpointing:
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values() or not model_class.supports_gradient_checkpointing:
                 continue
             model = model_class(config)
             model.to(torch_device)

tests/models/efficientformer/test_modeling_efficientformer.py

@@ -20,7 +20,6 @@ import warnings
 from typing import List
 
 from transformers import EfficientFormerConfig
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_vision, slow, torch_device
 from transformers.utils import cached_property, is_torch_available, is_vision_available
 
@@ -33,12 +32,14 @@ if is_torch_available():
     import torch
 
     from transformers import (
-        MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
-        MODEL_MAPPING,
         EfficientFormerForImageClassification,
         EfficientFormerForImageClassificationWithTeacher,
         EfficientFormerModel,
     )
+    from transformers.models.auto.modeling_auto import (
+        MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES,
+        MODEL_MAPPING_NAMES,
+    )
     from transformers.models.efficientformer.modeling_efficientformer import (
         EFFICIENTFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
     )
@@ -308,7 +309,7 @@ class EfficientFormerModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.T
         for model_class in self.all_model_classes:
             # EfficientFormerForImageClassificationWithTeacher supports inference-only
             if (
-                model_class in get_values(MODEL_MAPPING)
+                model_class.__name__ in MODEL_MAPPING_NAMES.values()
                 or model_class.__name__ == "EfficientFormerForImageClassificationWithTeacher"
             ):
                 continue
@@ -330,9 +331,9 @@ class EfficientFormerModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.T
 
         for model_class in self.all_model_classes:
             if (
-                model_class
+                model_class.__name__
                 not in [
-                    *get_values(MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING),
+                    *MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES.values(),
                 ]
                 or model_class.__name__ == "EfficientFormerForImageClassificationWithTeacher"
             ):

tests/models/glpn/test_modeling_glpn.py

@@ -18,7 +18,6 @@
 import unittest
 
 from transformers import is_torch_available, is_vision_available
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_vision, slow, torch_device
 
 from ...test_configuration_common import ConfigTester
@@ -29,7 +28,8 @@ from ...test_pipeline_mixin import PipelineTesterMixin
 if is_torch_available():
     import torch
 
-    from transformers import MODEL_MAPPING, GLPNConfig, GLPNForDepthEstimation, GLPNModel
+    from transformers import GLPNConfig, GLPNForDepthEstimation, GLPNModel
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.glpn.modeling_glpn import GLPN_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -291,7 +291,7 @@ class GLPNModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
         config.return_dict = True
 
         for model_class in self.all_model_classes:
-            if model_class in get_values(MODEL_MAPPING):
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values():
                 continue
             # TODO: remove the following 3 lines once we have a MODEL_FOR_DEPTH_ESTIMATION_MAPPING
             # this can then be incorporated into _prepare_for_class in test_modeling_common.py

tests/models/levit/test_modeling_levit.py

@@ -21,7 +21,6 @@ from math import ceil, floor
 
 from transformers import LevitConfig
 from transformers.file_utils import cached_property, is_torch_available, is_vision_available
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_vision, slow, torch_device
 
 from ...test_configuration_common import ConfigTester
@@ -33,12 +32,14 @@ if is_torch_available():
     import torch
 
     from transformers import (
-        MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
-        MODEL_MAPPING,
         LevitForImageClassification,
         LevitForImageClassificationWithTeacher,
         LevitModel,
     )
+    from transformers.models.auto.modeling_auto import (
+        MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES,
+        MODEL_MAPPING_NAMES,
+    )
     from transformers.models.levit.modeling_levit import LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -297,7 +298,7 @@ class LevitModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
         for model_class in self.all_model_classes:
             # LevitForImageClassificationWithTeacher supports inference-only
             if (
-                model_class in get_values(MODEL_MAPPING)
+                model_class.__name__ in MODEL_MAPPING_NAMES.values()
                 or model_class.__name__ == "LevitForImageClassificationWithTeacher"
             ):
                 continue
@@ -317,7 +318,7 @@ class LevitModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
         config.return_dict = True
 
         for model_class in self.all_model_classes:
-            if model_class in get_values(MODEL_MAPPING) or not model_class.supports_gradient_checkpointing:
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values() or not model_class.supports_gradient_checkpointing:
                 continue
             # LevitForImageClassificationWithTeacher supports inference-only
             if model_class.__name__ == "LevitForImageClassificationWithTeacher":
@@ -341,9 +342,9 @@ class LevitModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
 
         for model_class in self.all_model_classes:
             if (
-                model_class
+                model_class.__name__
                 not in [
-                    *get_values(MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING),
+                    *MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES.values(),
                 ]
                 or model_class.__name__ == "LevitForImageClassificationWithTeacher"
             ):

tests/models/perceiver/test_modeling_perceiver.py

@@ -26,7 +26,6 @@ import numpy as np
 from datasets import load_dataset
 
 from transformers import PerceiverConfig
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
 from transformers.utils import is_torch_available, is_vision_available
 
@@ -40,11 +39,6 @@ if is_torch_available():
     from torch import nn
 
     from transformers import (
-        MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
-        MODEL_FOR_MASKED_LM_MAPPING,
-        MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
-        MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
-        MODEL_MAPPING,
         PerceiverForImageClassificationConvProcessing,
         PerceiverForImageClassificationFourier,
         PerceiverForImageClassificationLearned,
@@ -55,6 +49,13 @@ if is_torch_available():
         PerceiverModel,
         PerceiverTokenizer,
     )
+    from transformers.models.auto.modeling_auto import (
+        MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES,
+        MODEL_FOR_MASKED_LM_MAPPING_NAMES,
+        MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES,
+        MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES,
+        MODEL_MAPPING_NAMES,
+    )
     from transformers.models.perceiver.modeling_perceiver import PERCEIVER_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -317,16 +318,19 @@ class PerceiverModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCas
             inputs_dict["subsampled_output_points"] = self.model_tester.subsampling
 
         if return_labels:
-            if model_class in [
-                *get_values(MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING),
-                *get_values(MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING),
+            if model_class.__name__ in [
+                *MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES.values(),
+                "PerceiverForImageClassificationLearned",
+                "PerceiverForImageClassificationFourier",
+                "PerceiverForImageClassificationConvProcessing",
+                *MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES.values(),
             ]:
                 inputs_dict["labels"] = torch.zeros(
                     self.model_tester.batch_size, dtype=torch.long, device=torch_device
                 )
-            elif model_class in [
-                *get_values(MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING),
-                *get_values(MODEL_FOR_MASKED_LM_MAPPING),
+            elif model_class.__name__ in [
+                *MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES.values(),
+                *MODEL_FOR_MASKED_LM_MAPPING_NAMES.values(),
             ]:
                 inputs_dict["labels"] = torch.zeros(
                     (self.model_tester.batch_size, self.model_tester.seq_length), dtype=torch.long, device=torch_device
@@ -380,10 +384,10 @@ class PerceiverModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCas
             return
 
         for model_class in self.all_model_classes:
-            if model_class in [
-                *get_values(MODEL_MAPPING),
-                PerceiverForOpticalFlow,
-                PerceiverForMultimodalAutoencoding,
+            if model_class.__name__ in [
+                *MODEL_MAPPING_NAMES.values(),
+                "PerceiverForOpticalFlow",
+                "PerceiverForMultimodalAutoencoding",
             ]:
                 continue
 
@@ -727,11 +731,14 @@ class PerceiverModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCas
 
         for model_class in self.all_model_classes:
             # most Perceiver models don't have a typical head like is the case with BERT
-            if model_class in [
-                PerceiverForOpticalFlow,
-                PerceiverForMultimodalAutoencoding,
-                *get_values(MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING),
-                *get_values(MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING),
+            if model_class.__name__ in [
+                "PerceiverForOpticalFlow",
+                "PerceiverForMultimodalAutoencoding",
+                *MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES.values(),
+                "PerceiverForImageClassificationLearned",
+                "PerceiverForImageClassificationFourier",
+                "PerceiverForImageClassificationConvProcessing",
+                *MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES.values(),
             ]:
                 continue
 
@@ -753,7 +760,7 @@ class PerceiverModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCas
         ]
 
         for model_class in self.all_model_classes:
-            if model_class not in get_values(MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING):
+            if model_class.__name__ not in MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES.values():
                 continue
 
             config, inputs, input_mask, _, _ = self.model_tester.prepare_config_and_inputs(model_class=model_class)

tests/models/pvt/test_modeling_pvt.py

@@ -18,7 +18,6 @@
 import unittest
 
 from transformers import is_torch_available, is_vision_available
-from transformers.models.auto import get_values
 from transformers.testing_utils import (
     require_accelerate,
     require_torch,
@@ -36,7 +35,8 @@ from ...test_pipeline_mixin import PipelineTesterMixin
 if is_torch_available():
     import torch
 
-    from transformers import MODEL_MAPPING, PvtConfig, PvtForImageClassification, PvtImageProcessor, PvtModel
+    from transformers import PvtConfig, PvtForImageClassification, PvtImageProcessor, PvtModel
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.pvt.modeling_pvt import PVT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -243,7 +243,7 @@ class PvtModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
         config.return_dict = True
 
         for model_class in self.all_model_classes:
-            if model_class in get_values(MODEL_MAPPING):
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values():
                 continue
             model = model_class(config)
             model.to(torch_device)

tests/models/segformer/test_modeling_segformer.py

@@ -18,7 +18,6 @@
 import unittest
 
 from transformers import SegformerConfig, is_torch_available, is_vision_available
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, slow, torch_device
 
 from ...test_configuration_common import ConfigTester
@@ -30,11 +29,11 @@ if is_torch_available():
     import torch
 
     from transformers import (
-        MODEL_MAPPING,
         SegformerForImageClassification,
         SegformerForSemanticSegmentation,
         SegformerModel,
     )
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.segformer.modeling_segformer import SEGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST
 
 
@@ -324,7 +323,7 @@ class SegformerModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCas
         config.return_dict = True
 
         for model_class in self.all_model_classes:
-            if model_class in get_values(MODEL_MAPPING):
+            if model_class.__name__ in MODEL_MAPPING_NAMES.values():
                 continue
 
             model = model_class(config)

tests/models/vilt/test_modeling_vilt.py

@@ -20,7 +20,6 @@ from datasets import load_dataset
 from packaging import version
 
 from transformers import ViltConfig, is_torch_available, is_vision_available
-from transformers.models.auto import get_values
 from transformers.testing_utils import require_torch, require_vision, slow, torch_device
 from transformers.utils import cached_property
 
@@ -33,7 +32,6 @@ if is_torch_available():
     import torch
 
     from transformers import (
-        MODEL_MAPPING,
         ViltForImageAndTextRetrieval,
         ViltForImagesAndTextClassification,
         ViltForMaskedLM,
@@ -41,6 +39,7 @@ if is_torch_available():
         ViltForTokenClassification,
         ViltModel,
     )
+    from transformers.models.auto.modeling_auto import MODEL_MAPPING_NAMES
     from transformers.models.vilt.modeling_vilt import VILT_PRETRAINED_MODEL_ARCHIVE_LIST
 
 if is_vision_available():
@@ -284,7 +283,7 @@ class ViltModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
                 config.modality_type_vocab_size = 3
 
             # ViltForImageAndTextRetrieval doesn't support training for now
-            if model_class in [*get_values(MODEL_MAPPING), ViltForImageAndTextRetrieval]:
+            if model_class.__name__ in [*MODEL_MAPPING_NAMES.values(), "ViltForImageAndTextRetrieval"]:
                 continue
 
             model = model_class(config)
@@ -307,7 +306,7 @@ class ViltModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
 
             # ViltForImageAndTextRetrieval doesn't support training for now
             if (
-                model_class in [*get_values(MODEL_MAPPING), ViltForImageAndTextRetrieval]
+                model_class.__name__ in [*MODEL_MAPPING_NAMES.values(), "ViltForImageAndTextRetrieval"]
                 or not model_class.supports_gradient_checkpointing
             ):
                 continue