sles skips

tests/models/bert/test_modeling_bert.py

@@ -26,6 +26,7 @@
     require_torch_accelerator,
     slow,
     torch_device,
+    skipIfRocm
 )
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -482,6 +483,11 @@ def _prepare_for_class(self, inputs_dict, model_class, return_labels=False):
                 )
         return inputs_dict
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def setUp(self):
         self.model_tester = BertModelTester(self)
         self.config_tester = ConfigTester(self, config_class=BertConfig, hidden_size=37)

tests/models/biogpt/test_modeling_biogpt.py

@@ -18,7 +18,7 @@
 import unittest
 
 from transformers import BioGptConfig, is_sacremoses_available, 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 ...generation.test_utils import GenerationTesterMixin
 from ...test_configuration_common import ConfigTester
@@ -298,6 +298,11 @@ class BioGptModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     )
     test_pruning = False
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def setUp(self):
         self.model_tester = BioGptModelTester(self)
         self.config_tester = ConfigTester(self, config_class=BioGptConfig, hidden_size=37)

tests/models/chameleon/test_modeling_chameleon.py

@@ -26,6 +26,7 @@
     require_torch,
     slow,
     torch_device,
+    skipIfRocm
 )
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -285,6 +286,11 @@ class ChameleonModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTester
     test_pruning = False
     fx_compatible = False
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def setUp(self):
         self.model_tester = ChameleonModelTester(self)
         self.config_tester = ConfigTester(self, config_class=ChameleonConfig, hidden_size=37)

tests/models/cohere/test_modeling_cohere.py

@@ -24,6 +24,7 @@
     require_torch_sdpa,
     slow,
     torch_device,
+    skipIfRocm,
 )
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -287,6 +288,11 @@ class CohereModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     # This is because we are hitting edge cases with the causal_mask buffer
     model_split_percents = [0.5, 0.7, 0.8]
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def setUp(self):
         self.model_tester = CohereModelTester(self)
         self.config_tester = ConfigTester(self, config_class=CohereConfig, hidden_size=37)

tests/models/dbrx/test_modeling_dbrx.py

@@ -327,12 +327,22 @@ class DbrxModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin
     test_headmasking = False
     test_pruning = False
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(os_name='sles')
+    def test_beam_search_low_memory(self):
+        super().test_beam_search_low_memory()
+        pass
+
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_with_static_cache(self):
         super().test_generate_with_static_cache()
         pass
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_from_inputs_embeds_with_static_cache(self):
         super().test_generate_from_inputs_embeds_with_static_cache()
         pass

tests/models/falcon/test_modeling_falcon.py

@@ -31,6 +31,7 @@
     require_torch_sdpa,
     slow,
     torch_device,
+    skipIfRocm,
 )
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -321,6 +322,11 @@ def is_pipeline_test_to_skip(
     ):
         return True
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def setUp(self):
         self.model_tester = FalconModelTester(self)
         self.config_tester = ConfigTester(self, config_class=FalconConfig, hidden_size=37)

tests/models/gemma/test_modeling_gemma.py

@@ -33,6 +33,7 @@
     require_torch_sdpa,
     slow,
     torch_device,
+    skipIfRocm,
 )
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -335,6 +336,11 @@ def is_pipeline_test_to_skip(
     ):
         return True
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def setUp(self):
         self.model_tester = GemmaModelTester(self)
         self.config_tester = ConfigTester(self, config_class=GemmaConfig, hidden_size=37)

tests/models/glm/test_modeling_glm.py

@@ -27,6 +27,7 @@
     require_torch_sdpa,
     slow,
     torch_device,
+    skipIfRocm,
 )
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -300,6 +301,16 @@ class GlmModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
     test_headmasking = False
     test_pruning = False
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
+    @skipIfRocm(os_name='sles')
+    def test_beam_search_low_memory(self):
+        super().test_beam_search_low_memory()
+        pass
+
     def setUp(self):
         self.model_tester = GlmModelTester(self)
         self.config_tester = ConfigTester(self, config_class=GlmConfig, hidden_size=37)

tests/models/gpt2/test_modeling_gpt2.py

@@ -28,6 +28,7 @@
     require_torch_gpu,
     slow,
     torch_device,
+    skipIfRocm,
 )
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -534,6 +535,11 @@ def _prepare_for_class(self, inputs_dict, model_class, return_labels=False):
                 )
         return inputs_dict
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def setUp(self):
         self.model_tester = GPT2ModelTester(self)
         self.config_tester = ConfigTester(self, config_class=GPT2Config, n_embd=37)

tests/models/gpt_bigcode/test_modeling_gpt_bigcode.py

@@ -18,7 +18,7 @@
 from parameterized import parameterized
 
 from transformers import GPTBigCodeConfig, is_torch_available
-from transformers.testing_utils import cleanup, require_torch, slow, torch_device
+from transformers.testing_utils import cleanup, require_torch, slow, torch_device, skipIfRocm
 
 from ...generation.test_utils import GenerationTesterMixin
 from ...test_configuration_common import ConfigTester
@@ -453,6 +453,11 @@ def test_disk_offload(self):
     def test_past_key_values_format(self):
         pass
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def test_gpt_bigcode_model(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_gpt_bigcode_model(*config_and_inputs)
@@ -503,6 +508,11 @@ class GPTBigCodeMHAModelTest(GPTBigCodeModelTest):
     # `parameterized_class` breaks with mixins, so we use inheritance instead
     multi_query = False
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
 
 @unittest.skipIf(
     not is_torch_greater_or_equal_than_1_12,

tests/models/gpt_neox/test_modeling_gpt_neox.py

@@ -287,11 +287,21 @@ class GPTNeoXModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMi
     test_model_parallel = False
     test_head_masking = False
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_with_static_cache():
         super().test_generate_with_static_cache()
         pass
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
+    @skipIfRocm(os_name='sles')
+    def test_beam_search_low_memory(self):
+        super().test_beam_search_low_memory()
+        pass
+
     def setUp(self):
         self.model_tester = GPTNeoXModelTester(self)
         self.config_tester = ConfigTester(self, config_class=GPTNeoXConfig, hidden_size=64, num_attention_heads=8)

tests/models/granite/test_modeling_granite.py

@@ -301,16 +301,26 @@ class GraniteModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMi
     # This is because we are hitting edge cases with the causal_mask buffer
     model_split_percents = [0.5, 0.7, 0.8]
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_from_inputs_embeds_with_static_cache(self):
         super().test_generate_from_inputs_embeds_with_static_cache()
         pass
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_with_static_cache(self):
         super().test_generate_with_static_cache()
         pass
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
+    @skipIfRocm(os_name='sles')
+    def test_beam_search_low_memory(self):
+        super().test_beam_search_low_memory()
+        pass
+
     def setUp(self):
         self.model_tester = GraniteModelTester(self)
         self.config_tester = ConfigTester(self, config_class=GraniteConfig, hidden_size=37)

tests/models/granitemoe/test_modeling_granitemoe.py

@@ -300,16 +300,26 @@ class GraniteMoeModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.Test
     # This is because we are hitting edge cases with the causal_mask buffer
     model_split_percents = [0.5, 0.7, 0.8]
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_with_static_cache(self):
         super().test_generate_with_static_cache()
         pass
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_from_inputs_embeds_with_static_cache(self):
         super().test_generate_from_inputs_embeds_with_static_cache()
         pass
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
+    @skipIfRocm(os_name='sles')
+    def test_beam_search_low_memory(self):
+        super().test_beam_search_low_memory()
+        pass 
+
     def setUp(self):
         self.model_tester = GraniteMoeModelTester(self)
         self.config_tester = ConfigTester(self, config_class=GraniteMoeConfig, hidden_size=37)

tests/models/hubert/test_modeling_hubert.py

@@ -23,7 +23,7 @@
 import pytest
 
 from transformers import HubertConfig, is_torch_available
-from transformers.testing_utils import require_soundfile, require_torch, slow, torch_device
+from transformers.testing_utils import require_soundfile, require_torch, slow, torch_device, skipIfRocm
 from transformers.utils import is_torch_fx_available
 
 from ...test_configuration_common import ConfigTester
@@ -590,6 +590,7 @@ 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(os_name='sles')
     def test_batched_inference(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_batch_inference(*config_and_inputs)

tests/models/idefics/test_modeling_idefics.py

@@ -602,12 +602,12 @@ class IdeficsForVisionText2TextTest(IdeficsModelTest, GenerationTesterMixin, uni
     all_generative_model_classes = (IdeficsForVisionText2Text,) if is_torch_available() else ()
 
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_from_inputs_embeds_with_static_cache(self):
         super().test_generate_from_inputs_embeds_with_static_cache()
         pass
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch='gfx1201', os_name='sles')
     def test_generate_with_static_cache(self):
         super().test_generate_with_static_cache()
         pass
@@ -626,6 +626,7 @@ def test_eager_matches_sdpa_inference(self, torch_dtype):
         pass
 
     @pytest.mark.generate
+    @skipIfRocm(os_name='sles')
     def test_left_padding_compatibility(self):
         """Overwrite because IDEFICS needs image attention mask to be also padded"""
         # NOTE: left-padding results in small numerical differences. This is expected.

tests/models/jamba/test_modeling_jamba.py

@@ -343,6 +343,11 @@ class JambaModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixi
     test_headmasking = False
     test_pruning = False
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     def setUp(self):
         self.model_tester = JambaModelTester(self)
         self.config_tester = JambaConfigTester(self, config_class=JambaConfig, hidden_size=37)

tests/models/jetmoe/test_modeling_jetmoe.py

@@ -28,6 +28,7 @@
     require_torch_gpu,
     slow,
     torch_device,
+    skipIfRocm
 )
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -299,6 +300,11 @@ class JetMoeModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     test_disk_offload_bin = False
     test_disk_offload_safetensors = False
 
+    @skipIfRocm(os_name='sles')
+    def test_left_padding_compatibility(self):
+        super().test_left_padding_compatibility()
+        pass
+
     @parameterized.expand([(1, False), (1, True), (4, False)])
     def test_new_cache_format(self, num_beams, do_sample):
         pass