Revert CuEq bias expansion when chunking with BS=1#204
Merged
Conversation
jnwei
approved these changes
May 1, 2026
jnwei
left a comment
jnwei
left a comment
Contributor
There was a problem hiding this comment.
LGTM thank you @christinaflo and @borisfom
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adding in @borisfom's patch from #181. The other PR fixed input shapes for cueq with chunking and BS>1. This isn't needed for BS=1, reverting to the old behavior to avoid expanding the bias tensor as this will blow up mem usage.