bugfix: torch.export failure caused by _make_causal_mask
#35291
Merged
Conversation
This file contains 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
Recent changes in torch dynamo prevent mutations on tensors converted with aten::_to_copy. To address this, we can clone such tensor before performing in-place operation `masked_fill_` only when the code is being compiled by torch dynamo. (relevant issue: pytorch/pytorch#127571)
|
This seems legit to me, but cc @ArthurZucker for core maintainer review |
qubvel
approved these changes
Dec 16, 2024
There was a problem hiding this comment.
Hi @jiwoong-choi, thanks for the update! I faced a similar issue with vision models and can confirm that this should fix torch.export. Alternatively, we could use a non-inplace operation for masked_fill, but your solution seems better because it does not change the original behavior.
ArthurZucker
approved these changes
Dec 20, 2024
Comment on lines
+172
to
176
| # Recent changes in PyTorch prevent mutations on tensors converted with aten::_to_copy | ||
| # See https://github.com/pytorch/pytorch/issues/127571 | ||
| if is_torchdynamo_compiling(): | ||
| mask = mask.clone() | ||
| mask.masked_fill_(context_mask, torch.finfo(dtype).min) |
There was a problem hiding this comment.
I agree with @qubvel, tho I don't mind modifying to have an inplace operation!
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
What does this PR do?
Fix the
torch.exportfailure caused byAttentionMaskConverter._make_causal_mask.Recent changes in torch dynamo prevent mutations on tensors converted with aten::_to_copy. To address this, we can clone such tensor before performing in-place operation
masked_fill_only when the code is being compiled by torch dynamo. (relevant issue on PyTorch: pytorch/pytorch#127571)Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
PyTorch: @gante @Rocketknight1