Repository navigation

[DSA] Support flashmla_sparse_q8 prefill backend on CUDA - #44217

Open

yzk1989 wants to merge 4 commits into

sgl-project:mainfrom

yzk1989:glm-flashmla-sparse-q8

Open

[DSA] Support flashmla_sparse_q8 prefill backend on CUDA#44217
yzk1989 wants to merge 4 commits into
sgl-project:mainfrom
yzk1989:glm-flashmla-sparse-q8

Commits on Oct 19, 2026
- support flashmla_sparse_q8 for dsa-prefill-backend in GLM5.3-Flash
  yangzhenkun.yzk
  committed
  0944fe9
- [Fix] dsa: Address raw FP8 Q8 prefill review feedback
  yangzhenkun.yzk
  committed
  06404b9
Commits on Oct 21, 2026
- [Refactor] dsa: Address Q8 prefill review feedback
  yangzhenkun.yzk
  committed
  efb7b86
Commits on Oct 29, 2026
- Merge upstream main and resolve Q8 prefill conflicts
  yangzhenkun.yzk
  committed
  6b1ac36