make devices consistent in cache reordering #43177
Open
+6
−2
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
Contributor
|
View the CircleCI Test Summary for this PR: https://huggingface.co/spaces/transformers-community/circle-ci-viz?pr=43177&sha=2f1a9a |
Author
|
@vasqu kindly review when you get a moment, thanks! |
vasqu
reviewed
Jan 9, 2026
Comment on lines
-80
to
+82
| self.keys = self.keys.index_select(0, beam_idx.to(self.keys.device)) | ||
| self.values = self.values.index_select(0, beam_idx.to(self.values.device)) | ||
| # Get device for each tensor and then move beam_idx to that device | ||
| keys_device = self.keys.device | ||
| self.keys = self.keys.index_select(0, beam_idx.to(keys_device)) |
Contributor
There was a problem hiding this comment.
Isn't that exactly the same as what's already happening
self.keys = self.keys.index_select(0, beam_idx.to(self.keys.device))vs
keys_device = self.keys.device
self.keys = self.keys.index_select(0, beam_idx.to(keys_device))Do you have a reproducer to check what's going on?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.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.
What does this PR do?
This PR fixes cache reordering causing runtime error when keys and values are on different devices
While working with beam search, with the cache operations, i got this error attached
this was due to beam_idx was being moved to different devices without considering that keys and values can be on different devices when during memory offloading.
changes i made are just to get the device info from keys and values separately solving the mismatch, hope this helps!
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.