Fix CI issue for vlm_gemma_3n model #4278
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Liu, Kaixuan <[email protected]>
Signed-off-by: Liu, Kaixuan <[email protected]>
|
@kashif , pls help review, thx very much. |
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
albertvillanova
approved these changes
Oct 28, 2025
There was a problem hiding this comment.
Thanks for the catch and the fix.
I confirm that this PR fixes the test:
PASSED tests/test_sft_trainer.py::TestSFTTrainer::test_train_vlm_gemma_3nMaybe we could add the reason why vision/audio towers do not update, something like they are frozen during training?
qgallouedec
reviewed
Oct 28, 2025
tests/test_sft_trainer.py
Outdated
| per_device_train_batch_size=1, | ||
| gradient_checkpointing=True, | ||
| model_init_kwargs={"dtype": "bfloat16"}, | ||
| model_init_kwargs={"dtype": "float16"}, |
There was a problem hiding this comment.
Suggested change
| model_init_kwargs={"dtype": "float16"}, | |
| model_init_kwargs={"dtype": "bfloat16"}, |
| # Initialize the trainer | ||
| training_args = SFTConfig( | ||
| output_dir=self.tmp_dir, | ||
| max_length=None, |
There was a problem hiding this comment.
Suggested change
| max_length=None, | |
| learning_rate=0.1, # increase lr to ensure updates are not lost due to bf16 rounding | |
| max_length=None, |
Co-authored-by: Albert Villanova del Moral <[email protected]>
qgallouedec
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When we run test case
pytest -rA tests/test_sft_trainer.py::TestSFTTrainer::test_train_vlm_gemma_3n, it will fail both on CUDA and Intel XPU. Further investigation shows there are 2 reasons:This PR fixes this bug.