Skip to content

[None][fix] xqa precision for fp16/bf16 kv cache #6573

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 4, 2025
Merged

[None][fix] xqa precision for fp16/bf16 kv cache #6573

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 cpp/kernels/xqa/mha_sm90.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2734,7 +2734,7 @@ __device__ inline void saveTransposedOutput(uint32_t threadRank, uint32_t warpRa
reinterpret_cast<Vec<InputElem, 4>&>(f16Core)
= convert<InputElem>(reinterpret_cast<Vec<float, 4> const&>(core));
auto const dst = idxMat < 2
? &swizzleBuf.template at<true>(idxRow, 2 * (gmmaWarpsPerGrp * m + warpRank) + idxMat)
? &swizzleBuf.template at<true>(8 * n + idxRow, 2 * (gmmaWarpsPerGrp * m + warpRank) + idxMat)
: nullptr;
stmatrix<true, 2>(dst, f16Core);
#elif CACHE_ELEM_ENUM == 2
Expand Down