Fix attention bias broadcast #24017
Open
Fix attention bias broadcast #24017
+127
−118
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
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Outdated suggestions cannot be applied.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.
Description
Motivation and Context
This feature was added in #21710.
There was bug when computing the offset when attention bias broadcast on dim 1 only in both CUDA and CPU kernel.
It can be triggered when attention bias shape is like [batch_size, 1, sequence_length, total_sequence_length] and batch_size > 1 when unfused kernel is selected. Note that cudnn flash attention and cutlass fused attention also supports attention bias, so the bug in unfused kernel was not discovered previously.