Skip to content

Reduce default-flash RAM by reusing BF16 projection scratch - #1030

Open
Mikyx-1 wants to merge 1 commit into
google:devfrom
Mikyx-1:fix/flash-kv-projection-scratch
Open

Mikyx-1 wants to merge 1 commit into
google:devfrom
Mikyx-1:fix/flash-kv-projection-scratch

Reuse BF16 projection scratch in default flash attention

caa2e13
Select commit
Loading
Failed to load commit list.
Google CLA / cla/google succeeded Sep 12, 2026 in 7s

✅ All contributors are covered under a CLA with Google

See https://cla.developers.google.com/ for more info about Google's Contributor License Agreement (CLA).

ℹ️ Googlers: Go here to view more details and manage scans for this pull request.

Details

The following contributors were found for this pull request:

caa2e13 Author: @Mikyx-1 <leh*******t2k​@gmail.com>

(Only the first commit for a unique contributor is listed.)