gfx1200-1 updates

tests/generation/test_utils.py

@@ -2569,7 +2569,7 @@ def test_transition_scores_greedy_search_normalized(self):
     def test_transition_scores_greedy_search(self):
         super().test_transition_scores_greedy_search()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_generate_input_features_as_encoder_kwarg(self):
         super().test_generate_input_features_as_encoder_kwarg()
 
@@ -2608,7 +2608,7 @@ def test_diverse_beam_search(self):
             ],
         )
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_max_length_if_input_embeds(self):
         # PT-only test: TF doesn't have StoppingCriteria
         article = "Today a dragon flew over Paris."

tests/models/aria/test_modeling_aria.py

@@ -195,11 +195,11 @@ class AriaForConditionalGenerationModelTest(ModelTesterMixin, GenerationTesterMi
     test_head_masking = False
     _is_composite = True
 
-    @skipIfRocm(arch=['gfx942','gfx90a'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_flex_attention_with_grads(self):
         super().test_flex_attention_with_grads()
 
-    @skipIfRocm(arch=['gfx942','gfx90a'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_generate_from_inputs_embeds_with_static_cache(self):
         super().test_generate_from_inputs_embeds_with_static_cache()
 

tests/models/emu3/test_modeling_emu3.py

@@ -319,123 +319,123 @@ class Emu3Vision2TextModelTest(ModelTesterMixin, GenerationTesterMixin, Pipeline
     test_pruning = False
     fx_compatible = False
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_generate_methods_with_logits_to_keep(self):
         super().test_generate_methods_with_logits_to_keep()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_generate_from_inputs_embeds_with_static_cache(self):
         super().test_generate_from_inputs_embeds_with_static_cache()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_generate_with_static_cache(self):
         super().test_generate_with_static_cache()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_attention_outputs(self):
         super().test_attention_outputs()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_batching_equivalence(self):
         super().test_batching_equivalence()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_beam_sample_generate(self):
         super().test_beam_sample_generate()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_beam_sample_generate_dict_output(self):
         super().test_beam_sample_generate_dict_output()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_beam_search_generate(self):
         super().test_beam_search_generate()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_generate_from_inputs_embeds_with_static_cache(self):
         super().test_generate_from_inputs_embeds_with_static_cache()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_beam_search_generate_dict_output(self):
         super().test_beam_search_generate_dict_output()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_beam_search_low_memory(self):
         super().test_beam_search_low_memory()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_constrained_beam_search_generate(self):
         super().test_constrained_beam_search_generate()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_constrained_beam_search_generate_dict_output(self):
         super().test_constrained_beam_search_generate_dict_output()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_determinism(self):
         super().test_determinism()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_group_beam_search_generate(self):
         super().test_group_beam_search_generate()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_group_beam_search_generate_dict_output(self):
         super().test_group_beam_search_generate_dict_output()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_hidden_states_output(self):
         super().test_hidden_states_output()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_model_outputs_equivalence(self):
         super().test_model_outputs_equivalence()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_offloaded_cache_implementation_0_offloaded(self):
         super().test_offloaded_cache_implementation_0_offloaded()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_resize_tokens_embeddings(self):
         super().test_resize_tokens_embeddings()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_retain_grad_hidden_states_attentions(self):
         super().test_retain_grad_hidden_states_attentions()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_feed_forward_chunking(self):
         super().test_feed_forward_chunking()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_flex_attention_with_grads(self):
         super().test_flex_attention_with_grads()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_forward_with_logits_to_keep(self):
         super().test_forward_with_logits_to_keep()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_greedy_generate(self):
         super().test_greedy_generate()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_greedy_generate_dict_outputs(self):
         super().test_greedy_generate_dict_outputs()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_group_beam_search_generate(self):
         super().test_group_beam_search_generate()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_sample_generate(self):
         super().test_sample_generate()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_sample_generate_dict_output(self):
         super().test_sample_generate_dict_output()
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_save_load(self):
         super().test_save_load()
 

tests/models/gpt_neox/test_modeling_gpt_neox.py

@@ -283,7 +283,7 @@ class GPTNeoXModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMi
     test_model_parallel = False
     test_head_masking = False
 
-    @skipIfRocm(arch=['gfx90a','gfx942'])
+    @skipIfRocm(arch=['gfx90a','gfx942','gfx1100','gfx1200','gfx1201'])
     def test_flex_attention_with_grads(self):
         super().test_flex_attention_with_grads()
 

tests/models/mixtral/test_modeling_mixtral.py

@@ -320,12 +320,12 @@ class MixtralModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMi
     fx_compatible = False  # Broken by attention refactor cc @Cyrilvallez
 
     def test_generate_from_inputs_embeds_with_static_cache(self):
-        if rocmUtils.is_rocm_skippable(arch=['gfx90a','gfx942']):
+        if rocmUtils.is_rocm_skippable(arch=['gfx90a','gfx942','gfx1201','gfx1200']):
             torch._dynamo.config.capture_dynamic_output_shape_ops = True
         super().test_generate_from_inputs_embeds_with_static_cache()
 
     def test_generate_with_static_cache(self):
-        if rocmUtils.is_rocm_skippable(arch=['gfx90a','gfx942']):
+        if rocmUtils.is_rocm_skippable(arch=['gfx90a','gfx942','gfx1201','gfx1200']):
             torch._dynamo.config.capture_dynamic_output_shape_ops = True
         super().test_generate_with_static_cache()
 

tests/models/moshi/test_modeling_moshi.py

@@ -617,7 +617,7 @@ def test_new_cache_format(self):
         super().test_new_cache_format_0()
         pass
 
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_offloaded_cache_implementation_0_offloaded(self):
         super().test_offloaded_cache_implementation_0_offloaded()
         pass

tests/models/phimoe/test_modeling_phimoe.py

@@ -354,7 +354,7 @@ class PhimoeModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     test_headmasking = False
     test_pruning = False
 
-    @skipIfRocm(arch=['gfx90a','gfx942'])
+    @skipIfRocm(arch=['gfx90a','gfx942','gfx1100','gfx1201','gfx1200'])
     def test_generate_with_static_cache(self):
         super().test_generate_with_static_cache()
 

tests/pipelines/test_pipelines_audio_classification.py

@@ -137,7 +137,7 @@ def test_small_model_pt(self):
         self.assertIn(nested_simplify(output, decimals=4), [EXPECTED_OUTPUT, EXPECTED_OUTPUT_PT_2])
 
     @require_torch
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_small_model_pt_fp16(self):
         model = "anton-l/wav2vec2-random-tiny-classifier"
 

tests/pipelines/test_pipelines_depth_estimation.py

@@ -139,7 +139,7 @@ def test_small_model_pt(self):
         self.skipTest(reason="There is not hf-internal-testing tiny model for either GLPN nor DPT")
 
     @require_torch
-    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100'])
+    @skipIfRocm(arch=['gfx942','gfx90a','gfx1100','gfx1201','gfx1200'])
     def test_multiprocess(self):
         depth_estimator = pipeline(
             model="hf-internal-testing/tiny-random-DepthAnythingForDepthEstimation",