VLMs: enable generation tests (#33533)

* add tests

* fix whisper

* update

* nit

* add qwen2-vl

* more updates!

* better this way

* fix this one

* fix more tests

* fix final tests, hope so

* fix led

* Update tests/generation/test_utils.py

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* pr comments

* not pass pixels and extra for low-mem tests, very flaky because of visio tower

---------

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

tests/models/vipllava/test_modeling_vipllava.py

@@ -28,6 +28,7 @@ from transformers import (
 )
 from transformers.testing_utils import require_bitsandbytes, require_torch, require_torch_gpu, slow, torch_device
 
+from ...generation.test_utils import GenerationTesterMixin
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
 
@@ -73,7 +74,7 @@ class VipLlavaVisionText2TextModelTester:
             "initializer_range": 0.02,
             "num_labels": 3,
             "num_choices": 4,
-            "pad_token_id": 0,
+            "pad_token_id": 1,
         },
         is_training=True,
         vision_config={
@@ -99,7 +100,7 @@ class VipLlavaVisionText2TextModelTester:
         self.vision_feature_layers = vision_feature_layers
         self.text_config = text_config
         self.vision_config = vision_config
-        self.seq_length = seq_length
+        self.pad_token_id = text_config["pad_token_id"]
 
         self.num_hidden_layers = text_config["num_hidden_layers"]
         self.vocab_size = text_config["vocab_size"]
@@ -111,6 +112,8 @@ class VipLlavaVisionText2TextModelTester:
         self.num_channels = 3
         self.image_size = 336
         self.encoder_seq_length = 231
+        self.num_image_tokens = 224
+        self.seq_length = seq_length + self.num_image_tokens
 
     def get_config(self):
         return VipLlavaConfig(
@@ -120,6 +123,7 @@ class VipLlavaVisionText2TextModelTester:
             image_token_index=self.image_token_index,
             projector_hidden_act=self.projector_hidden_act,
             vision_feature_layers=self.vision_feature_layers,
+            image_seq_length=self.num_image_tokens,
         )
 
     def prepare_config_and_inputs(self):
@@ -140,8 +144,9 @@ class VipLlavaVisionText2TextModelTester:
         config, pixel_values = config_and_inputs
         input_ids = ids_tensor([self.batch_size, self.seq_length], config.text_config.vocab_size - 1) + 1
         attention_mask = input_ids.ne(1).to(torch_device)
-        # we are giving 3 images let's make sure we pass in 3 image tokens
-        input_ids[:, 1] = config.image_token_index
+
+        input_ids[input_ids == config.image_token_index] = self.pad_token_id
+        input_ids[:, : self.num_image_tokens] = config.image_token_index
         inputs_dict = {
             "pixel_values": pixel_values,
             "input_ids": input_ids,
@@ -152,12 +157,13 @@ class VipLlavaVisionText2TextModelTester:
 
 @require_torch
 # Copied from transformers.tests.models.llava.test_modeling_llava.LlavaForConditionalGenerationModelTest with Llava->VipLlava
-class VipLlavaForConditionalGenerationModelTest(ModelTesterMixin, unittest.TestCase):
+class VipLlavaForConditionalGenerationModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase):
     """
     Model tester for `VipLlavaForConditionalGeneration`.
     """
 
     all_model_classes = (VipLlavaForConditionalGeneration,) if is_torch_available() else ()
+    all_generative_model_classes = (VipLlavaForConditionalGeneration,) if is_torch_available() else ()
     fx_compatible = False
     test_pruning = False
     test_resize_embeddings = True