Fix divide-by-zero in GroupNorm two-pass kernel for large batch sizes#1984
Merged
Conversation
When batch size N is large enough (e.g., N=512 with C=640), the heuristic `blocks_per_act_slice = 256 / params.n` truncates to 0 via integer division, causing a subsequent `div_up(params.hw, blocks_per_act_slice)` to divide by zero. Fix by clamping blocks_per_act_slice to at least 1 in both forward and backward two-pass setup functions. Add regression test covering the exact repro case and all three heuristic branches. Signed-off-by: Tailing Yuan <[email protected]>
Contributor
Author
|
Hi @crcrpar , could you help review this PR? |
Collaborator
|
Looks good |
crcrpar
approved these changes
Mar 5, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
When batch size N is large enough (e.g., N=512 with C=640), the heuristic
blocks_per_act_slice = 256 / params.ntruncates to 0 via integer division, causing a subsequentdiv_up(params.hw, blocks_per_act_slice)to divide by zero. Fix by clamping blocks_per_act_slice to at least 1 in both forward and backward two-pass setup functions.Add regression test covering the exact repro case and all three heuristic branches.