Skip to content

Speed up CPU GroupQueryAttention decode with a grouped flash kernel - #33920

Draft
Pierluigi Urru (stwykd) wants to merge 7 commits into
microsoft:mainfrom
stwykd:gqa-cpu-decode
Draft

Pierluigi Urru (stwykd) wants to merge 7 commits into
microsoft:mainfrom
stwykd:gqa-cpu-decode

Keep ungrouped decode fast, and propagate NaN through the split merge

e15f58c
Select commit
Loading
Failed to load commit list.
Sign in for the full log view

Annotations

1 warning
Windows OpenVINO CI Pipeline
succeeded Nov 5, 2026 in 23m 27s