Update tiny model info. and pipeline testing (#25213)

* update tiny_model_summary.json

* update

* update

* update

---------

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

tests/models/falcon/test_modeling_falcon.py

@@ -278,9 +278,9 @@ class FalconModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     pipeline_model_mapping = (
         {
             "feature-extraction": FalconModel,
+            "question-answering": FalconForQuestionAnswering,
             "text-classification": FalconForSequenceClassification,
             "text-generation": FalconForCausalLM,
-            "question-answering": FalconForQuestionAnswering,
             "token-classification": FalconForTokenClassification,
             "zero-shot": FalconForSequenceClassification,
         }

tests/models/mpt/test_modeling_mpt.py

@@ -362,7 +362,16 @@ class MptModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
     test_torchscript = False
     test_head_masking = False
     pipeline_model_mapping = (
-        {"feature-extraction": MptModel, "text-generation": MptForCausalLM} if is_torch_available() else {}
+        {
+            "feature-extraction": MptModel,
+            "question-answering": MptForQuestionAnswering,
+            "text-classification": MptForSequenceClassification,
+            "text-generation": MptForCausalLM,
+            "token-classification": MptForTokenClassification,
+            "zero-shot": MptForSequenceClassification,
+        }
+        if is_torch_available()
+        else {}
     )
 
     def setUp(self):

tests/models/mra/test_modeling_mra.py

@@ -22,6 +22,7 @@ from transformers.testing_utils import require_torch, slow, torch_device
 
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor, random_attention_mask
+from ...test_pipeline_mixin import PipelineTesterMixin
 
 
 if is_torch_available():
@@ -280,7 +281,7 @@ class MraModelTester:
 
 
 @require_torch
-class MraModelTest(ModelTesterMixin, unittest.TestCase):
+class MraModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
     all_model_classes = (
         (
             MraModel,
@@ -299,6 +300,18 @@ class MraModelTest(ModelTesterMixin, unittest.TestCase):
     has_attentions = False
 
     all_generative_model_classes = ()
+    pipeline_model_mapping = (
+        {
+            "feature-extraction": MraModel,
+            "fill-mask": MraForMaskedLM,
+            "question-answering": MraForQuestionAnswering,
+            "text-classification": MraForSequenceClassification,
+            "token-classification": MraForTokenClassification,
+            "zero-shot": MraForSequenceClassification,
+        }
+        if is_torch_available()
+        else {}
+    )
 
     def setUp(self):
         self.model_tester = MraModelTester(self)

tests/models/pvt/test_modeling_pvt.py

@@ -30,6 +30,7 @@ from transformers.testing_utils import (
 
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
+from ...test_pipeline_mixin import PipelineTesterMixin
 
 
 if is_torch_available():
@@ -154,8 +155,13 @@ def prepare_img():
 
 
 @require_torch
-class PvtModelTest(ModelTesterMixin, unittest.TestCase):
+class PvtModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
     all_model_classes = (PvtModel, PvtForImageClassification) if is_torch_available() else ()
+    pipeline_model_mapping = (
+        {"feature-extraction": PvtModel, "image-classification": PvtForImageClassification}
+        if is_torch_available()
+        else {}
+    )
 
     test_head_masking = False
     test_pruning = False

tests/models/t5/test_modeling_t5.py

@@ -560,11 +560,11 @@ class T5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
         {
             "conversational": T5ForConditionalGeneration,
             "feature-extraction": T5Model,
+            "question-answering": T5ForQuestionAnswering,
             "summarization": T5ForConditionalGeneration,
+            "text-classification": T5ForSequenceClassification,
             "text2text-generation": T5ForConditionalGeneration,
             "translation": T5ForConditionalGeneration,
-            "question-answering": T5ForQuestionAnswering,
-            "text-classification": T5ForSequenceClassification,
             "zero-shot": T5ForSequenceClassification,
         }
         if is_torch_available()
@@ -583,6 +583,16 @@ class T5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
         self.model_tester = T5ModelTester(self)
         self.config_tester = ConfigTester(self, config_class=T5Config, d_model=37)
 
+    # `QAPipelineTests` is not working well with slow tokenizers (for some models) and we don't want to touch the file
+    # `src/transformers/data/processors/squad.py` (where this test fails for this model)
+    def is_pipeline_test_to_skip(
+        self, pipeline_test_casse_name, config_class, model_architecture, tokenizer_name, processor_name
+    ):
+        if pipeline_test_casse_name == "QAPipelineTests" and not tokenizer_name.endswith("Fast"):
+            return True
+
+        return False
+
     def _create_and_check_torch_fx_tracing(self, config, inputs_dict, output_loss=False):
         if not is_torch_fx_available() or not self.fx_compatible:
             return

tests/models/umt5/test_modeling_umt5.py

@@ -296,11 +296,11 @@ class UMT5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin
         {
             "conversational": UMT5ForConditionalGeneration,
             "feature-extraction": UMT5Model,
+            "question-answering": UMT5ForQuestionAnswering,
             "summarization": UMT5ForConditionalGeneration,
+            "text-classification": UMT5ForSequenceClassification,
             "text2text-generation": UMT5ForConditionalGeneration,
             "translation": UMT5ForConditionalGeneration,
-            "question-answering": UMT5ForQuestionAnswering,
-            "text-classification": UMT5ForSequenceClassification,
             "zero-shot": UMT5ForSequenceClassification,
         }
         if is_torch_available()
@@ -317,6 +317,16 @@ class UMT5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin
     def setUp(self):
         self.model_tester = UMT5ModelTester(self)
 
+    # `QAPipelineTests` is not working well with slow tokenizers (for some models) and we don't want to touch the file
+    # `src/transformers/data/processors/squad.py` (where this test fails for this model)
+    def is_pipeline_test_to_skip(
+        self, pipeline_test_casse_name, config_class, model_architecture, tokenizer_name, processor_name
+    ):
+        if pipeline_test_casse_name == "QAPipelineTests" and not tokenizer_name.endswith("Fast"):
+            return True
+
+        return False
+
     def _create_and_check_torch_fx_tracing(self, config, inputs_dict, output_loss=False):
         if not is_torch_fx_available() or not self.fx_compatible:
             return

tests/models/vivit/test_modeling_vivit.py

@@ -29,6 +29,7 @@ from transformers.utils import cached_property, is_torch_available, is_vision_av
 
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
+from ...test_pipeline_mixin import PipelineTesterMixin
 
 
 if is_torch_available():
@@ -153,13 +154,18 @@ class VivitModelTester:
 
 
 @require_torch
-class VivitModelTest(ModelTesterMixin, unittest.TestCase):
+class VivitModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
     """
     Here we also overwrite some of the tests of test_modeling_common.py, as Vivit does not use input_ids, inputs_embeds,
     attention_mask and seq_length.
     """
 
     all_model_classes = (VivitModel, VivitForVideoClassification) if is_torch_available() else ()
+    pipeline_model_mapping = (
+        {"feature-extraction": VivitModel, "video-classification": VivitForVideoClassification}
+        if is_torch_available()
+        else {}
+    )
 
     test_pruning = False
     test_torchscript = False