
FlashAttention was first published by Tri Dao in May 2022 and it had a deep impact in the large language models space. Most open models you’ve heard of (RedPajama, MPT, LLaMA, Falcon, etc) all leverage it for faster inference. Tri came on the podcast t...