[tests] remove overwrites of removed test

tests/models/deepseek_v3/test_modeling_deepseek_v3.py

@@ -263,10 +263,6 @@ def test_prompt_lookup_decoding_matches_greedy_search(self, assistant_type):
     def test_assisted_decoding_sample(self):
         pass
 
-    @unittest.skip("DeepseekV3 doesn't support contrastive generation")
-    def test_contrastive_generate_dict_outputs_use_cache(self):
-        pass
-
     @unittest.skip("Deepseek-V3 uses MLA so it is not compatible with the standard cache format")
     def test_beam_search_generate_dict_outputs_use_cache(self):
         pass

tests/models/gemma3n/test_modeling_gemma3n.py

@@ -405,27 +405,6 @@ def test_eager_matches_sdpa_inference(
             self, name, dtype, padding_side, use_attention_mask, output_attentions, enable_kernels, atols=atols
         )
 
-    @pytest.mark.generate
-    @unittest.skip(
-        "Gemma3n has a special shape for hidden states (due to per-layer projs) which is not compatible with contrastive decoding"
-    )
-    def test_contrastive_generate(self):
-        pass
-
-    @pytest.mark.generate
-    @unittest.skip(
-        "Gemma3n has a special shape for hidden states (due to per-layer projs) which is not compatible with contrastive decoding"
-    )
-    def test_contrastive_generate_dict_outputs_use_cache(self):
-        pass
-
-    @pytest.mark.generate
-    @unittest.skip(
-        "Gemma3n has a special shape for hidden states (due to per-layer projs) which is not compatible with contrastive decoding"
-    )
-    def test_contrastive_generate_low_memory(self):
-        pass
-
     @pytest.mark.generate
     @unittest.skip("Gemma3n does not support QuantizedCache as it performs cache manipulation in the forward pass")
     def test_generate_with_quant_cache(self):

tests/models/git/test_modeling_git.py

@@ -455,18 +455,6 @@ def test_model_from_pretrained(self):
     def test_beam_search_generate_dict_outputs_use_cache(self):
         pass
 
-    @unittest.skip(reason="GIT has pixel values as additional input")
-    def test_contrastive_generate(self):
-        pass
-
-    @unittest.skip(reason="GIT has pixel values as additional input")
-    def test_contrastive_generate_dict_outputs_use_cache(self):
-        pass
-
-    @unittest.skip(reason="GIT has pixel values as additional input")
-    def test_contrastive_generate_low_memory(self):
-        pass
-
     @unittest.skip(reason="GIT has pixel values as additional input")
     def test_greedy_generate_dict_outputs_use_cache(self):
         pass

tests/models/idefics/test_modeling_idefics.py

@@ -496,21 +496,6 @@ def test_generate_without_input_ids(self):
     def test_generate_continue_from_inputs_embeds(self):
         pass
 
-    @pytest.mark.generate
-    @unittest.skip(reason="""IDEFICS cannot do contrastive generation yet and it is not worth fixing""")
-    def test_contrastive_generate(self):
-        pass
-
-    @pytest.mark.generate
-    @unittest.skip(reason="""IDEFICS cannot do contrastive generation yet and it is not worth fixing""")
-    def test_contrastive_generate_low_memory(self):
-        pass
-
-    @pytest.mark.generate
-    @unittest.skip(reason="""IDEFICS cannot do contrastive generation yet and it is not worth fixing""")
-    def test_contrastive_generate_dict_outputs_use_cache(self):
-        pass
-
     def test_attention_outputs(self):
         config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
         config.return_dict = True

tests/models/minimax/test_modeling_minimax.py

@@ -209,10 +209,6 @@ def test_past_key_values_format(self, custom_all_cache_shapes=None):
     def test_prompt_lookup_decoding_matches_greedy_search(self):
         pass
 
-    @unittest.skip(reason="MiniMaxCache does not support `crop()` method")
-    def test_contrastive_generate_low_memory(self):
-        pass
-
     @unittest.skip(reason="MiniMaxCache does not support `crop()` method")
     def test_assisted_decoding_sample(self):
         pass
@@ -225,10 +221,6 @@ def test_assisted_decoding_matches_greedy_search_0_random(self):
     def test_assisted_decoding_matches_greedy_search_1_same(self):
         pass
 
-    @unittest.skip(reason="MiniMaxCache does not support `crop()` method")
-    def test_contrastive_generate_dict_outputs_use_cache(self):
-        pass
-
     @unittest.skip("Model needs refactor")
     def test_attention_outputs(self):
         pass

tests/models/mllama/test_modeling_mllama.py

@@ -368,16 +368,6 @@ def test_sdpa_can_compile_dynamic(self):
     def test_model_parallelism(self):
         pass
 
-    @unittest.skip(
-        reason="Mllama cache type doesn't allow correct check on output `past_key_values` due to `Cache.crop()`"
-    )
-    def test_contrastive_generate_dict_outputs_use_cache(self, assistant_type):
-        pass
-
-    @unittest.skip(reason="Mllama can't do low memory due to `Cache.crop()`")
-    def test_contrastive_generate_low_memory(self, assistant_type):
-        pass
-
     @unittest.skip(reason="Mllama can't assisted decoding due to cache format and `Cache.crop()`")
     def test_assisted_decoding_with_num_logits_to_keep(self):
         pass

tests/models/moshi/test_modeling_moshi.py

@@ -603,18 +603,6 @@ def test_initialization(self):
     def test_generate_continue_from_past_key_values(self):
         pass
 
-    @unittest.skip("Moshi doesn't support contrastive generation yet.")
-    def test_contrastive_generate(self):
-        pass
-
-    @unittest.skip("Moshi doesn't support contrastive generation yet.")
-    def test_contrastive_generate_dict_outputs_use_cache(self):
-        pass
-
-    @unittest.skip("Moshi doesn't support contrastive generation yet.")
-    def test_contrastive_generate_low_memory(self):
-        pass
-
     @unittest.skip(
         "Moshi either needs default generation config or fix for fullgraph compile because it hardcodes SlidingWindowCache in custom generation loop."
     )

tests/models/qwen2_5_omni/test_modeling_qwen2_5_omni.py

@@ -418,18 +418,6 @@ def attention_mask_padding_matches_padding_free_with_position_ids(
                 tol = torch.finfo(torch.bfloat16).eps
                 torch.testing.assert_close(logits_padded, logits_padfree, rtol=tol, atol=tol)
 
-    @unittest.skip("Cannot do contrastive generation, has custom `generate()`")
-    def test_contrastive_generate(self):
-        pass
-
-    @unittest.skip("Cannot do contrastive generation, has custom `generate()`")
-    def test_contrastive_generate_dict_outputs_use_cache(self):
-        pass
-
-    @unittest.skip("Cannot do contrastive generation, has custom `generate()`")
-    def test_contrastive_generate_low_memory(self):
-        pass
-
     @unittest.skip("Cannot generate from inputs embeds")
     def test_generate_from_inputs_embeds_with_static_cache(self):
         pass

tests/models/recurrent_gemma/test_modeling_recurrent_gemma.py

@@ -66,18 +66,6 @@ class RecurrentGemmaModelTest(CausalLMModelTest, unittest.TestCase):
     def test_eager_matches_sdpa_generate(self):
         pass
 
-    @unittest.skip(reason="RecurrentGemma does not return the cache")
-    def test_contrastive_generate_low_memory(self):
-        pass
-
-    @unittest.skip(reason="RecurrentGemma does not return the cache")
-    def test_contrastive_generate_dict_outputs_use_cache(self):
-        pass
-
-    @unittest.skip(reason="RecurrentGemma does not return the cache")
-    def test_contrastive_generate(self):
-        pass
-
     @unittest.skip(reason="SQRBound is known to have issues with gc")
     def test_training_gradient_checkpointing_use_reentrant_false(self):
         pass