Repository navigation

Keep FlashAttention inside captured CUDA graphs - #28

Merged

ZhiningLiu1998 merged 3 commits into

mainfrom

perf/cuda-graph-flash-attention

Oct 15, 2026

Merged

Keep FlashAttention inside captured CUDA graphs#28
ZhiningLiu1998 merged 3 commits into
mainfrom
perf/cuda-graph-flash-attention

Commits on Oct 4, 2026
- Keep FlashAttention inside captured CUDA graphs
  
  BruceSheng1202
  and
  claude
  committed
  96cb307
Commits on Oct 5, 2026
- docs: describe CUDA-graph capture mask decisions
  
  BruceSheng1202
  and
  claude
  committed
  3b8fca2
Commits on Oct 15, 2026
- fix(inference): isolate CUDA graph mask overrides
  
  ZhiningLiu1998
  and
  BruceSheng1202
  committed
  c72a8fb