Skip to content

fix flashmaskv2 maxmin buffer padding #74881

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 26, 2025
Merged

fix flashmaskv2 maxmin buffer padding #74881

Show file tree
Hide file tree
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
2 changes: 1 addition & 1 deletion paddle/phi/kernels/gpu/flash_attn_v3_grad_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1190,7 +1190,7 @@ void FlashMaskV2GradBaseKernel(
// TODO(umiswing): refine this block constraint (kBlockN % 32), since some
// of kBlockN is not divisible by 32 flashmask_maxmin_shape[2] =
// (flashmask_maxmin_shape[2] + 31) / 32 * 8;
flashmask_maxmin_shape[2] = (flashmask_maxmin_shape[2] + 31) / 32;
flashmask_maxmin_shape[2] = ((flashmask_maxmin_shape[2] + 31) / 32 + 3) / 4 * 4;
flashmask_maxmin_shape[3] = 8;

flashmask_maxmin.set_type(phi::DataType::INT32);
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/gpu/flash_attn_v3_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1965,7 +1965,7 @@ void FlashMaskV2BaseKernel(
// TODO(umiswing): refine this block constraint (kBlockN % 32), since some
// of kBlockN is not divisible by 32 flashmask_maxmin_shape[2] =
// (flashmask_maxmin_shape[2] + 31) / 32 * 8;
flashmask_maxmin_shape[2] = ((flashmask_maxmin_shape[2] + 31) / 32 + 3) / 4;
flashmask_maxmin_shape[2] = ((flashmask_maxmin_shape[2] + 31) / 32 + 3) / 4 * 4;
flashmask_maxmin_shape[3] = 8;

flashmask_maxmin.set_type(phi::DataType::INT32);
Expand Down
Loading