tspeterkim/flash-attention-minimal

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

230 best points · 39 comments · 1 Hacker News post · 1,203 GitHub stars · View on GitHub

Hacker News discussions

• Show HN: Flash Attention in ~100 lines of CUDA (230 points, 39 comments, by tspeterkim, 2024-03-16)