SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Fix cache-related tests (#39676)

Browse Source 
* fix

* fix kyutai at last

* fix unrelated tests and copies

* update musicgen as well

* revert tensor

* fix old test failures

* why it wasn't added?
This commit is contained in:
Raushan Turganbay
2025-07-28 17:30:11 +02:00
committed by GitHub
parent fc2bd1eac0
commit 1c6b47451d
14 changed files with 89 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/models/llava_next/test_modeling_llava_next.py
View File

@@ -515,7 +515,7 @@ class LlavaNextForConditionalGenerationIntegrationTest(unittest.TestCase):
# test that changing `strategy` won't error out
model.vision_feature_select_strategy = "full"
inputs = self.processor(self.prompt, self.image, return_tensors="pt").to(model.device)
inputs = self.processor(text=self.prompt, images=self.image, return_tensors="pt").to(model.device)
# verify generation
output = model.generate(**inputs, max_new_tokens=30)
@@ -536,7 +536,7 @@ class LlavaNextForConditionalGenerationIntegrationTest(unittest.TestCase):
model = LlavaNextForConditionalGeneration.from_pretrained(granite_model_path)
self.processor = AutoProcessor.from_pretrained(granite_model_path)
prompt = "<|user|>\n<image>\nWhat is shown in this image?\n<|assistant|>\n"
inputs = self.processor(prompt, self.image, return_tensors="pt").to(model.device)
inputs = self.processor(text=prompt, images=self.image, return_tensors="pt").to(model.device)
# verify generation
output = model.generate(**inputs, max_new_tokens=30)

4
tests/models/llava_next_video/test_modeling_llava_next_video.py
View File

@@ -467,7 +467,9 @@ class LlavaNextVideoForConditionalGenerationIntegrationTest(unittest.TestCase):
padding=True,
).to(torch_device)
inputs_single = self.processor(self.prompt_video, videos=[self.video], return_tensors="pt").to(torch_device)
inputs_single = self.processor(text=self.prompt_video, videos=[self.video], return_tensors="pt").to(
torch_device
)
# verify generation
output_batched = model.generate(**inputs_batched, do_sample=False, max_new_tokens=50)

3
tests/models/qwen2_5_omni/test_modeling_qwen2_5_omni.py
View File

@@ -413,7 +413,6 @@ class Qwen2_5OmniThinkerForConditionalGenerationModelTest(ModelTesterMixin, Gene
logits_padded = res_padded.logits[inputs_dict["attention_mask"].bool()]
logits_padfree = res_padfree.logits[0]
torch.testing.assert_close(logits_padded.argmax(-1), logits_padfree.argmax(-1), rtol=0, atol=0)
# acceptable numerical instability
tol = torch.finfo(torch.bfloat16).eps
torch.testing.assert_close(logits_padded, logits_padfree, rtol=tol, atol=tol)
@@ -698,7 +697,7 @@ class Qwen2_5OmniModelIntegrationTest(unittest.TestCase):
)
text = self.processor.apply_chat_template(self.messages, tokenize=False, add_generation_prompt=True)
inputs = self.processor(
text=text * 2,
text=[text] * 2,
audio=[self.raw_audio, self.raw_audio],
images=[self.raw_image, self.raw_image],
return_tensors="pt",

1
tests/models/qwen2_5_vl/test_modeling_qwen2_5_vl.py
View File

@@ -403,7 +403,6 @@ class Qwen2_5_VLModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.Test
logits_padded = res_padded.logits[inputs_dict["attention_mask"].bool()]
logits_padfree = res_padfree.logits[0]
torch.testing.assert_close(logits_padded.argmax(-1), logits_padfree.argmax(-1), rtol=0, atol=0)
# acceptable numerical instability
tol = torch.finfo(torch.bfloat16).eps
torch.testing.assert_close(logits_padded, logits_padfree, rtol=tol, atol=tol)

1
tests/models/qwen2_vl/test_modeling_qwen2_vl.py
View File

@@ -362,7 +362,6 @@ class Qwen2VLModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCas
logits_padded = res_padded.logits[inputs_dict["attention_mask"].bool()]
logits_padfree = res_padfree.logits[0]
torch.testing.assert_close(logits_padded.argmax(-1), logits_padfree.argmax(-1), rtol=0, atol=0)
# acceptable numerical instability
tol = torch.finfo(torch.bfloat16).eps
torch.testing.assert_close(logits_padded, logits_padfree, rtol=tol, atol=tol)

8
tests/pipelines/test_pipelines_image_text_to_text.py
View File

@@ -119,7 +119,7 @@ class ImageTextToTextPipelineTests(unittest.TestCase):
},
{
"role": "assistant",
"content": "Hugging Face, a company of minds\nWith tools and services that make our lives easier\nFrom",
"content": "Hugging Face, a company of minds\nWith tools and services that make our lives easier\nFrom natural language processing\nTo machine learning and more, they've got it all\n\nThey've made it possible for us to be more\nInformed and efficient, with their tools and services\nFrom image and speech recognition\nTo text and language translation, they've got it all\n\nThey've made it possible for us to be more\nInformed and efficient, with their tools and services\nFrom image and speech recognition\nTo text and language translation, they've got it all\n\nThey've made it possible for us to be more\nInformed and efficient, with their tools and services\nFrom image and speech recognition\nTo text and language translation, they've got it all\n\nThey've made it possible for us to be more\nInformed and efficient, with their tools and services\nFrom image and speech recognition\nTo text and language translation, they've got it all\n\nThey've made it possible for us to be more\nInformed and efficient, with their tools and services\nFrom image and speech recognition\nTo text and language translation, they've got it all\n\nThey've made it possible for us to be more\nInformed and efficient, with their tools and",
},
],
}
@@ -150,7 +150,7 @@ class ImageTextToTextPipelineTests(unittest.TestCase):
[
{
"input_text": "<image> What this is? Assistant: This is",
"generated_text": "<image> What this is? Assistant: This is a photo of two cats lying on a pink blanket. The cats are sleeping and appear to be comfortable",
"generated_text": "<image> What this is? Assistant: This is a photo of two cats lying on a pink blanket. The cats are sleeping and appear to be comfortable. The photo captures a moment of tranquility and companionship between the two feline friends.",
}
],
)
@@ -161,11 +161,11 @@ class ImageTextToTextPipelineTests(unittest.TestCase):
[
{
"input_text": "<image> What this is? Assistant: This is",
"generated_text": "<image> What this is? Assistant: This is a photo of two cats lying on a pink blanket. The cats are facing the camera, and they",
"generated_text": "<image> What this is? Assistant: This is a photo of two cats lying on a pink blanket. The cats are facing the camera, and they appear to be sleeping or resting. The blanket is placed on a couch, and the cats are positioned in such a way that they are facing the camera. The image captures a peaceful moment between the two cats, and it's a great way to showcase their cuteness and relaxed demeanor.",
},
{
"input_text": "<image> What this is? Assistant: This is",
"generated_text": "<image> What this is? Assistant: This is a photo of two cats lying on a pink blanket. The cats are facing the camera, and they",
"generated_text": "<image> What this is? Assistant: This is a photo of two cats lying on a pink blanket. The cats are facing the camera, and they appear to be sleeping or resting. The blanket is placed on a couch, and the cats are positioned in such a way that they are facing the camera. The image captures a peaceful moment between the two cats, and it's a great way to showcase their cuteness and relaxed demeanor.",
},
],
)