FlashAttention-2: 800% faster Transformers using SRAM (latent.space)

2 points by FanaHOVA 3y ago ↗ HN

0 comments

[ 7.3 ms ] story [ 122 ms ] thread

No comments yet.

A showcase of shard-db. Data from the Hacker News API. Refreshed every 15 minutes.

Source · Live DB stats