“flash attention and implement it in vidrial, then on the…” — Latent Space: The AI Engineer Podcast at 12:05
“flash attention and implement it in vidrial, then on the problems where flash attention is really well optimized, we end up basically matching performance 1% to 2% faster sometimes. But on the problems where flash attention is not well optimized, which is a huge number of problem shapes, just because there's so many…”
0:00
32:59