fix(sparse-attn): correct block_size computation in backward kernel#1
Open
fix(sparse-attn): correct block_size computation in backward kernel#1
Conversation
fix(sparse-attn): correct block_size masking in backward kernel The backward kernel used an incorrect block_size when applying the mask, causing padding positions to be treated as valid tokens. As a result, p = tl.math.exp2(qk - m) was computed on invalid entries, leading to Inf and NaN values during gradient accumulation (especially in dQ). This patch fixes the block_size computation for split blocks so that the mask correctly excludes padded regions in all cases.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
fix(sparse-attn): correct block_size masking in backward kernel
The backward kernel used an incorrect block_size when applying the mask, causing padding positions to be treated as valid tokens. As a result, p = tl.math.exp2(qk - m) was computed on invalid entries, leading to Inf and NaN values during gradient accumulation (especially in dQ).
This patch fixes the block_size computation for split blocks so that the mask correctly excludes padded regions in all cases.