skip failures

tests/extended/test_trainer_ext.py

@@ -37,6 +37,7 @@
     require_torch_non_multi_accelerator,
     slow,
     torch_device,
+    skipIfRocm
 )
 from transformers.trainer_callback import TrainerState
 from transformers.trainer_utils import set_seed
@@ -98,16 +99,19 @@ def test_run_seq2seq_no_dist(self):
 
     # verify that the trainer can handle non-distributed with n_gpu > 1
     @require_torch_multi_accelerator
+    @skipIfRocm
     def test_run_seq2seq_dp(self):
         self.run_seq2seq_quick(distributed=False)
 
     # verify that the trainer can handle distributed with n_gpu > 1
     @require_torch_multi_accelerator
+    @skipIfRocm
     def test_run_seq2seq_ddp(self):
         self.run_seq2seq_quick(distributed=True)
 
     @require_apex
     @require_torch_gpu
+    @skipIfRocm
     def test_run_seq2seq_apex(self):
         # XXX: apex breaks the trainer if it's run twice e.g. run_seq2seq.main() from the same
         # program and it breaks other tests that run from the same pytest worker, therefore until this is
@@ -124,6 +128,7 @@ def test_run_seq2seq_apex(self):
 
     @parameterized.expand(["base", "low", "high", "mixed"])
     @require_torch_multi_accelerator
+    @skipIfRocm
     def test_trainer_log_level_replica(self, experiment_id):
         # as each sub-test is slow-ish split into multiple sub-tests to avoid CI timeout
         experiments = {

tests/fsdp/test_fsdp.py

@@ -36,6 +36,7 @@
     require_torch_multi_accelerator,
     slow,
     torch_device,
+    skipIfRocm
 )
 from transformers.trainer_callback import TrainerState
 from transformers.trainer_utils import FSDPOption, set_seed
@@ -176,6 +177,7 @@ def test_fsdp_config(self, sharding_strategy, dtype):
             self.assertEqual(os.environ.get("ACCELERATE_USE_FSDP", "false"), "true")
 
     @parameterized.expand(params, name_func=_parameterized_custom_name_func)
+    @skipIfRocm
     def test_fsdp_config_transformers_auto_wrap(self, sharding_strategy, dtype):
         output_dir = self.get_auto_remove_tmp_dir()
         fsdp_config = deepcopy(self.fsdp_config)

tests/generation/test_utils.py

@@ -35,6 +35,7 @@
     require_torch_multi_gpu,
     slow,
     torch_device,
+    skipIfRocm
 )
 
 from ..test_modeling_common import floats_tensor, ids_tensor
@@ -666,6 +667,7 @@ def test_beam_search_generate_dict_outputs_use_cache(self):
 
     @require_accelerate
     @require_torch_multi_accelerator
+    @skipIfRocm
     def test_model_parallel_beam_search(self):
         for model_class in self.all_generative_model_classes:
             if "xpu" in torch_device:

tests/models/bigbird_pegasus/test_modeling_bigbird_pegasus.py

@@ -26,6 +26,7 @@
     require_torch_fp16,
     slow,
     torch_device,
+    skipIfRocm
 )
 
 from ...generation.test_utils import GenerationTesterMixin

tests/models/encoder_decoder/test_modeling_encoder_decoder.py

@@ -18,7 +18,7 @@
 import unittest
 
 from transformers import is_torch_available, logging
-from transformers.testing_utils import CaptureLogger, require_deterministic_for_xpu, require_torch, slow, torch_device
+from transformers.testing_utils import CaptureLogger, require_deterministic_for_xpu, require_torch, slow, torch_device, skipIfRocm
 
 from ...test_modeling_common import ids_tensor
 from ..bart.test_modeling_bart import BartStandaloneDecoderModelTester
@@ -585,6 +585,7 @@ def test_encoder_decoder_model_from_pretrained_using_model_paths(self):
         input_ids_dict = self.prepare_config_and_inputs()
         self.check_encoder_decoder_model_from_pretrained_using_model_paths(**input_ids_dict, return_dict=False)
 
+    @skipIfRocm
     def test_save_and_load_from_pretrained(self):
         input_ids_dict = self.prepare_config_and_inputs()
         self.check_save_and_load(**input_ids_dict)

tests/models/fastspeech2_conformer/test_modeling_fastspeech2_conformer.py

@@ -25,7 +25,7 @@
     FastSpeech2ConformerWithHifiGanConfig,
     is_torch_available,
 )
-from transformers.testing_utils import require_g2p_en, require_torch, require_torch_accelerator, slow, torch_device
+from transformers.testing_utils import require_g2p_en, require_torch, require_torch_accelerator, slow, torch_device, skipIfRocm
 
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, _config_zero_init, ids_tensor
@@ -127,6 +127,11 @@ class FastSpeech2ConformerModelTest(ModelTesterMixin, unittest.TestCase):
     test_resize_embeddings = False
     is_encoder_decoder = True
 
+
+    @skipIfRocm
+    def test_multi_gpu_data_parallel_forward(self):
+        super().test_multi_gpu_data_parallel_forward()
+
     def setUp(self):
         self.model_tester = FastSpeech2ConformerModelTester(self)
         self.config_tester = ConfigTester(self, config_class=FastSpeech2ConformerConfig)
@@ -550,6 +555,10 @@ class FastSpeech2ConformerWithHifiGanTest(ModelTesterMixin, unittest.TestCase):
     test_resize_embeddings = False
     is_encoder_decoder = True
 
+    @skipIfRocm
+    def test_multi_gpu_data_parallel_forward(self):
+        super().test_multi_gpu_data_parallel_forward()
+
     def setUp(self):
         self.model_tester = FastSpeech2ConformerWithHifiGanTester(self)
 

tests/models/informer/test_modeling_informer.py

@@ -22,7 +22,7 @@
 from huggingface_hub import hf_hub_download
 
 from transformers import is_torch_available
-from transformers.testing_utils import is_flaky, require_torch, slow, torch_device
+from transformers.testing_utils import is_flaky, require_torch, slow, torch_device, skipIfRocm
 
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
@@ -221,6 +221,7 @@ def test_save_load_strict(self):
                 model2, info = model_class.from_pretrained(tmpdirname, output_loading_info=True)
             self.assertEqual(info["missing_keys"], [])
 
+    @skipIfRocm
     def test_encoder_decoder_model_standalone(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs_for_common()
         self.model_tester.check_encoder_decoder_model_standalone(*config_and_inputs)

tests/models/mamba/test_modeling_mamba.py

@@ -22,7 +22,7 @@
 from parameterized import parameterized
 
 from transformers import AutoTokenizer, MambaConfig, is_torch_available
-from transformers.testing_utils import require_torch, require_torch_multi_gpu, slow, torch_device
+from transformers.testing_utils import require_torch, require_torch_multi_gpu, slow, torch_device, skipIfRocm
 
 from ...generation.test_utils import GenerationTesterMixin
 from ...test_configuration_common import ConfigTester
@@ -305,6 +305,7 @@ def test_config(self):
         self.config_tester.run_common_tests()
 
     @require_torch_multi_gpu
+    @skipIfRocm
     def test_multi_gpu_data_parallel_forward(self):
         config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
 
@@ -441,6 +442,7 @@ def setUp(self):
         self.tokenizer = AutoTokenizer.from_pretrained(self.model_id)
 
     @parameterized.expand([(torch_device,), ("cpu",)])
+    @skipIfRocm
     def test_simple_generate(self, device):
         tokenizer = AutoTokenizer.from_pretrained("state-spaces/mamba-130m-hf")
         tokenizer.pad_token = tokenizer.eos_token

tests/models/mra/test_modeling_mra.py

@@ -17,7 +17,7 @@
 import unittest
 
 from transformers import MraConfig, is_torch_available
-from transformers.testing_utils import require_torch, slow, torch_device
+from transformers.testing_utils import require_torch, slow, torch_device, skipIfRocm
 
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor, random_attention_mask
@@ -314,36 +314,101 @@ class MraModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
     def setUp(self):
         self.model_tester = MraModelTester(self)
         self.config_tester = ConfigTester(self, config_class=MraConfig, hidden_size=37)
-
+    
     def test_config(self):
         self.config_tester.run_common_tests()
 
+
+    @skipIfRocm
+    def test_determinism(self):
+        super().test_determinism()
+
+    @skipIfRocm
+    def test_feed_forward_chunking(self):
+        super().test_feed_forward_chunking()
+
+    @skipIfRocm
+    def test_hidden_states_output(self):
+        super().test_hidden_states_output()
+
+    @skipIfRocm
+    def test_inputs_embeds(self):
+        super().test_inputs_embeds()
+
+    @skipIfRocm
+    def test_inputs_embeds_matches_input_ids(self):
+        super().test_inputs_embeds_matches_input_ids()
+
+    @skipIfRocm
+    def test_load_with_mismatched_shapes(self):
+        super().test_load_with_mismatched_shapes()
+
+    @skipIfRocm
+    def test_model_outputs_equivalence(self):
+        super().test_model_outputs_equivalence()
+
+    @skipIfRocm
+    def test_multi_gpu_data_parallel_forward(self):
+        super().test_multi_gpu_data_parallel_forward()
+
+    @skipIfRocm
+    def test_problem_types(self):
+        super().test_problem_types()
+
+    @skipIfRocm
+    def test_resize_embeddings_untied(self):
+        super().test_resize_embeddings_untied()
+
+    @skipIfRocm
+    def test_resize_tokens_embeddings(self):
+        super().test_resize_tokens_embeddings()
+
+    @skipIfRocm
+    def test_retain_grad_hidden_states_attentions(self):
+        super().test_retain_grad_hidden_states_attentions()
+
+    @skipIfRocm
+    def test_save_load(self):
+        super().test_save_load()
+
+    @skipIfRocm
+    def test_training(self):
+        super().test_training()
+
+
+    @skipIfRocm
     def test_model(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_model(*config_and_inputs)
 
+    @skipIfRocm
     def test_model_various_embeddings(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         for type in ["absolute", "relative_key", "relative_key_query"]:
             config_and_inputs[0].position_embedding_type = type
             self.model_tester.create_and_check_model(*config_and_inputs)
 
+    @skipIfRocm
     def test_for_masked_lm(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_for_masked_lm(*config_and_inputs)
 
+    @skipIfRocm
     def test_for_multiple_choice(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_for_multiple_choice(*config_and_inputs)
 
+    @skipIfRocm
     def test_for_question_answering(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_for_question_answering(*config_and_inputs)
 
+    @skipIfRocm
     def test_for_sequence_classification(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_for_sequence_classification(*config_and_inputs)
 
+    @skipIfRocm
     def test_for_token_classification(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_for_token_classification(*config_and_inputs)

tests/models/musicgen/test_modeling_musicgen.py

@@ -42,6 +42,7 @@
     require_torch_sdpa,
     slow,
     torch_device,
+    skipIfRocm
 )
 from transformers.utils import cached_property, is_torch_bf16_available_on_device, is_torch_fp16_available_on_device
 
@@ -184,6 +185,10 @@ class MusicgenDecoderTest(ModelTesterMixin, GenerationTesterMixin, PipelineTeste
     test_pruning = False
     test_resize_embeddings = False
 
+    @skipIfRocm
+    def test_multi_gpu_data_parallel_forward(self):
+        super().test_multi_gpu_data_parallel_forward()
+
     def setUp(self):
         self.model_tester = MusicgenDecoderTester(self)
         self.config_tester = ConfigTester(self, config_class=MusicgenDecoderConfig, hidden_size=16)
@@ -1071,6 +1076,10 @@ class MusicgenTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
     # (and `torchscript` hates `None` values).
     test_torchscript = False
 
+    @skipIfRocm
+    def test_multi_gpu_data_parallel_forward(self):
+        super().test_multi_gpu_data_parallel_forward()
+
     def setUp(self):
         self.model_tester = MusicgenTester(self)
 

tests/models/musicgen_melody/test_modeling_musicgen_melody.py

@@ -43,6 +43,7 @@
     require_torchaudio,
     slow,
     torch_device,
+    skipIfRocm
 )
 from transformers.utils import cached_property, is_torch_bf16_available_on_device, is_torch_fp16_available_on_device
 
@@ -184,6 +185,10 @@ class MusicgenMelodyDecoderTest(ModelTesterMixin, GenerationTesterMixin, unittes
     )  # the model uses a custom generation method so we only run a specific subset of the generation tests
     test_pruning = False
     test_resize_embeddings = False
+
+    @skipIfRocm
+    def test_multi_gpu_data_parallel_forward(self):
+        super().test_multi_gpu_data_parallel_forward()
 
     def setUp(self):
         self.model_tester = MusicgenMelodyDecoderTester(self)
@@ -1075,6 +1080,10 @@ class MusicgenMelodyTest(ModelTesterMixin, GenerationTesterMixin, PipelineTester
     # (and `torchscript` hates `None` values).
     test_torchscript = False
 
+    @skipIfRocm
+    def test_multi_gpu_data_parallel_forward(self):
+        super().test_multi_gpu_data_parallel_forward()
+
     def setUp(self):
         self.model_tester = MusicgenMelodyTester(self)
 

tests/models/speecht5/test_modeling_speecht5.py

@@ -28,6 +28,7 @@
     require_torch,
     slow,
     torch_device,
+    skipIfRocm
 )
 from transformers.trainer_utils import set_seed
 from transformers.utils import cached_property
@@ -1097,6 +1098,7 @@ def test_generation(self):
         )
 
     @require_deterministic_for_xpu
+    @skipIfRocm
     def test_one_to_many_generation(self):
         model = self.default_model
         processor = self.default_processor
@@ -1205,6 +1207,7 @@ def test_one_to_many_generation(self):
                 "Mismatch in waveform between standalone and integrated vocoder for single instance generation.",
             )
 
+    @skipIfRocm
     def test_batch_generation(self):
         model = self.default_model
         processor = self.default_processor

tests/models/superpoint/test_modeling_superpoint.py

@@ -16,7 +16,7 @@
 from typing import List
 
 from transformers.models.superpoint.configuration_superpoint import SuperPointConfig
-from transformers.testing_utils import require_torch, require_vision, slow, torch_device
+from transformers.testing_utils import require_torch, require_vision, slow, torch_device, skipIfRocm
 from transformers.utils import cached_property, is_torch_available, is_vision_available
 
 from ...test_configuration_common import ConfigTester
@@ -122,6 +122,10 @@ class SuperPointModelTest(ModelTesterMixin, unittest.TestCase):
     has_attentions = False
     from_pretrained_id = "magic-leap-community/superpoint"
 
+    @skipIfRocm
+    def test_multi_gpu_data_parallel_forward(self):
+        super().test_multi_gpu_data_parallel_forward()
+
     def setUp(self):
         self.model_tester = SuperPointModelTester(self)
         self.config_tester = ConfigTester(