Generate tests: modality-agnostic input preparation (#33685)

tests/models/seamless_m4t/test_modeling_seamless_m4t.py

@@ -360,8 +360,6 @@ class SeamlessM4TModelWithSpeechInputTest(ModelTesterMixin, unittest.TestCase):
     )
     all_generative_model_classes = (SeamlessM4TForSpeechToText,) if is_torch_available() else ()
 
-    input_name = "input_features"
-
     def setUp(self):
         self.model_tester = SeamlessM4TModelTester(self, input_modality="speech")
         self.config_tester = ConfigTester(self, config_class=SeamlessM4TConfig)
@@ -379,26 +377,6 @@ class SeamlessM4TModelWithSpeechInputTest(ModelTesterMixin, unittest.TestCase):
         model = SeamlessM4TModel.from_pretrained(model_name)
         self.assertIsNotNone(model)
 
-    def _get_input_ids_and_config(self, batch_size=2):
-        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
-        input_ids = inputs_dict[self.input_name]
-
-        # cut to half length & take max batch_size 3
-        sequence_length = input_ids.shape[-1] // 2
-        input_ids = input_ids[:batch_size, :sequence_length]
-
-        # generate max 3 tokens
-        max_length = input_ids.shape[-1] + 3
-        if config.eos_token_id is not None and config.pad_token_id is None:
-            # hack to allow generate for models such as GPT2 as is done in `generate()`
-            if isinstance(config.eos_token_id, int):
-                config.eos_token_id = [config.eos_token_id]
-            config.pad_token_id = config.eos_token_id[0]
-
-        attention_mask = torch.ones(input_ids.shape[:2], dtype=torch.long)[:batch_size, :sequence_length]
-
-        return config, input_ids.float(), attention_mask, max_length
-
     def test_initialization(self):
         config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()