Skip to content

CUDA: fix negative KV_max values in FA #15321

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Aug 14, 2025
Merged

CUDA: fix negative KV_max values in FA #15321

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion ggml/src/ggml-cuda/fattn-common.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -539,11 +539,15 @@ static __global__ void flash_attn_mask_to_KV_max(
all_inf = warp_reduce_all(all_inf);

if (!all_inf) {
KV_max_sj += FATTN_KQ_STRIDE;
break;
}
}

// If the break in the loop was not triggered, KV_max_sj is now -FATTN_KQ_STRIDE.
// If the break was triggered it's the lower edge of the tile with the first non-masked values.
// In either case, walk back the decrementation by FATTN_KQ_STRIDE.
KV_max_sj += FATTN_KQ_STRIDE;

if (threadIdx.x != 0) {
return;
}
Expand Down
Loading