Skip to content

update and add Expectations for mistral3/internvl tests #42616

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Abdennacer-Badaoui wants to merge 4 commits into
base: main
Choose a base branch
from
+26 −8
Open

update and add Expectations for mistral3/internvl tests #42616

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
1a24eb4
fix and add expectations for cuda and rocm platforms
Abdennacer-Badaoui Dec 4, 2025
5d2bdde
differnet expectations for A10 vs A100
Abdennacer-Badaoui Dec 4, 2025
e017e82
fix rocm expectations for intervl
Abdennacer-Badaoui Dec 4, 2025
419344e
remove heeeere
Abdennacer-Badaoui Dec 4, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 21 additions & 6 deletions tests/models/internvl/test_modeling_internvl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -430,7 +430,14 @@ def test_qwen2_small_model_integration_batched_generate_multi_image(self):
# Check first output
decoded_output = processor.decode(output[0], skip_special_tokens=True)
# Batching seems to alter the output slightly, but it is also the case in the original implementation. This seems to be expected: https://github.com/huggingface/transformers/issues/23017#issuecomment-1649630232
expected_output = "user\n\nWrite a haiku for this image\nassistant\nSilky lake, \nWooden pier, \nNature's peace." # fmt: skip
expected_outputs = Expectations(
{
("xpu", 3): 'user\n\nWrite a haiku for this image\nassistant\nSilky lake, \nWooden pier, \nNature\'s peace.',
("cuda", 7): 'user\n\nWrite a haiku for this image\nassistant\nSilky lake, \nWooden pier, \nNature\'s peace.',
("rocm", (9, 4)): 'user\n\nWrite a haiku for this image\nassistant\nSilky lake, \nWooden pier, \nNature\'s embrace.',
}
) # fmt: skip
expected_output = expected_outputs.get_expectation()
self.assertEqual(
decoded_output,
expected_output,
Expand All @@ -443,6 +450,7 @@ def test_qwen2_small_model_integration_batched_generate_multi_image(self):
{
("xpu", 3): "user\n\nWhat are the differences between these two images?\nassistant\nThe images show the Statue of Liberty and the Golden Gate Bridge from different angles. Here are the differences:\n\n1. **Foreground",
("cuda", 7): "user\n\nWhat are the differences between these two images?\nassistant\nThe images show the Statue of Liberty and the Golden Gate Bridge from different angles. Here are the differences:\n\n1. **Foreground",
("rocm", (9, 4)): "user\n\nWhat are the differences between these two images?\nassistant\nThe images show the Statue of Liberty and the Golden Gate Bridge from different angles. Here are the main differences:\n\n1. **",
}
) # fmt: skip
expected_output = expected_outputs.get_expectation()
Expand Down Expand Up @@ -567,6 +575,7 @@ def test_qwen2_small_model_integration_interleaved_images_videos(self):
{
("xpu", 3): "user\n\n\nWhat are the differences between these two images?\nassistant\nThe images depict two distinct scenes:\n\n1. **Left Image:**\n - The Statue of Liberty is prominently featured on an",
("cuda", 7): 'user\n\n\nWhat are the differences between these two images?\nassistant\nThe images depict two distinct scenes:\n\n1. **Left Image:**\n - The Statue of Liberty is prominently featured on an',
("rocm", (9, 4)): 'user\n\n\nWhat are the differences between these two images?\nassistant\nThe images depict two distinct scenes:\n\n1. **Left Image:**\n - This image features the Statue of Liberty on Liberty',
}
) # fmt: skip
expected_output = expected_outputs.get_expectation()
Expand All @@ -582,6 +591,7 @@ def test_qwen2_small_model_integration_interleaved_images_videos(self):
{
("xpu", 3): "user\nFrame1: \nFrame2: \nFrame3: \nFrame4: \nFrame5: \nFrame6: \nFrame7: \nFrame8: \nWhat type of shot is the man performing?\nassistant\nA forehand shot",
("cuda", 7): 'user\nFrame1: \nFrame2: \nFrame3: \nFrame4: \nFrame5: \nFrame6: \nFrame7: \nFrame8: \nWhat type of shot is the man performing?\nassistant\nA forehand shot',
("rocm", (9, 4)): 'user\nFrame1: \nFrame2: \nFrame3: \nFrame4: \nFrame5: \nFrame6: \nFrame7: \nFrame8: \nWhat type of shot is the man performing?\nassistant\nA forehand shot',
}
) # fmt: skip
expected_output = expected_outputs.get_expectation()
Expand All @@ -593,9 +603,14 @@ def test_qwen2_small_model_integration_interleaved_images_videos(self):

# Check third output
decoded_output = processor.decode(output[2], skip_special_tokens=True)
expected_output = (
"user\n\nWrite a haiku for this image\nassistant\nSilky lake, \nWooden pier, \nNature's peace."
)
expected_outputs = Expectations(
{
("xpu", 3): 'user\n\nWrite a haiku for this image\nassistant\nSilky lake, \nWooden pier, \nNature\'s peace.',
("cuda", 7): 'user\n\nWrite a haiku for this image\nassistant\nSilky lake, \nWooden pier, \nNature\'s peace.',
("rocm", (9, 4)): 'user\n\nWrite a haiku for this image\nassistant\nSilky lake, \nWooden pier, \nNature\'s embrace.',
}
) # fmt: skip
expected_output = expected_outputs.get_expectation()
self.assertEqual(
decoded_output,
expected_output,
Expand Down Expand Up @@ -658,7 +673,7 @@ def test_llama_small_model_integration_forward(self):
("xpu", 3): [-9.8828, -0.4954, 1.4561, -10.3438, -10.3438],
("cuda", 7): [-9.8750, -0.4861, 1.4648, -10.3359, -10.3359],
("cuda", 8): [-9.8906, -0.4995, 1.4473, -10.3359, -10.3438],
("rocm", (9, 4)): [ -9.8828, -0.5005, 1.4697, -10.3438, -10.3438],
("rocm", (9, 4)): [ -9.8672, -0.4888, 1.4648, -10.3281, -10.3281],
("rocm", (9, 5)): [ -9.8906, -0.4976, 1.4502, -10.3359, -10.3438],
}
) # fmt: skip
Expand Down Expand Up @@ -934,7 +949,7 @@ def test_llama_small_model_integration_interleaved_images_videos(self):
("xpu", 3): "user\n\n\nWhat are the difference between these two images?\nassistant\nI apologize for the confusion in my previous response. Upon closer inspection, the differences between the two images are:\n\n1. **",
("cuda", 7): 'user\n\n\nWhat are the difference between these two images?\nassistant\nI apologize for the confusion in my previous response. Upon closer inspection, the differences between the two images are:\n\n1. **',
("cuda", 8): 'user\n\n\nWhat are the difference between these two images?\nassistant\nI apologize for the confusion in my previous response. After re-examining the images, I can see that there are no',
("rocm", (9, 4)): 'user\n\n\nWhat are the difference between these two images?\nassistant\nI apologize for the confusion in my previous response. Upon closer inspection, the differences between the two images are:\n\n1. **',
("rocm", (9, 4)): 'user\n\n\nWhat are the difference between these two images?\nassistant\nI apologize for the confusion in my previous response. After re-examining the images, I can see that there are no',
("rocm", (9, 5)): 'user\n\n\nWhat are the difference between these two images?\nassistant\nI apologize for the confusion in my previous response. After re-examining the images, I can see that there are no',
}
) # fmt: skip
Expand Down
7 changes: 5 additions & 2 deletions tests/models/mistral3/test_modeling_mistral3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -355,7 +355,8 @@ def test_mistral3_integration_batched_generate(self):
expected_outputs = Expectations(
{
("xpu", 3): "Calm lake's mirror gleams,\nWhispering pines stand in silence,\nPath to peace begins.",
("cuda", 8): "Wooden path to calm,\nReflections whisper secrets,\nNature's peace unfolds.",
("cuda", (8, 0)): "Wooden path to calm,\nReflections whisper secrets,\nNature's peace unfolds.",
("cuda", (8, 6)): "Calm waters reflect\nWooden path to distant shore\nSilence in the woods",
("rocm", (9, 5)): "Calm waters reflect\nWooden path to distant shore\nSilence in the scene"
}
) # fmt: skip
Expand Down Expand Up @@ -432,7 +433,8 @@ def test_mistral3_integration_batched_generate_multi_image(self):
decoded_output = processor.decode(gen_tokens[0], skip_special_tokens=True)
expected_outputs = Expectations(
{
("cuda", 8): 'Calm waters reflect\nWooden path to distant shore\nSilence in the scene',
("cuda", 8): "Calm waters reflect\nWooden path to distant shore\nPeace in nature's hold",
("rocm", (9, 4)): "Calm waters reflect\nWooden path to distant shore\nSilence in the pines"
}
) # fmt: skip
expected_output = expected_outputs.get_expectation()
Expand All @@ -448,6 +450,7 @@ def test_mistral3_integration_batched_generate_multi_image(self):
{
("xpu", 3): "Certainly! The images depict two iconic landmarks:\n\n1. The first image shows the Statue of Liberty in New York City.",
("cuda", 8): 'Certainly! The images depict two famous landmarks in the United States:\n\n1. The first image shows the Statue of Liberty,',
("rocm", (9, 4)): 'Certainly! The images depict two famous landmarks in the United States:\n\n1. The first image shows the Statue of Liberty,',
}
) # fmt: skip
expected_output = expected_outputs.get_expectation()
Expand Down