Commit efab335

authored

Optimize flash attention (#2334)

* optimize flash attention

1 parent 984e0ce commit efab335Copy full SHA for efab335

2 files changed

+426

-119

lines changed

csrc/cpu/aten/kernels
- FlashAttentionKrnl.cpp
- MultiHeadAttentionKrnl.cpp

2 files changed

+426

-119

lines changed

Comments

(0)