flash-attention-minimal

★ 8 updated 2y ago ⑂ fork

Flash Attention in ~100 lines of CUDA (forward pass only)

No plain-English explanation yet — one is being written right now. Check back in a minute.