From 6ababeb7dba9277ce6a4819e97da28715daee274 Mon Sep 17 00:00:00 2001 From: Philip Turner Date: Sat, 15 Jul 2023 01:34:24 -0400 Subject: [PATCH] Update usage.md --- usage.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/usage.md b/usage.md index d4a019a..8986399 100644 --- a/usage.md +++ b/usage.md @@ -123,3 +123,5 @@ yields the fastest BERT training on cloud instances in MLPerf training 2.0 (June - [Jax](https://github.com/google/jax): an [implementation](https://github.com/lucidrains/flash-attention-jax) in Jax by [lucidrains](https://github.com/lucidrains/). + +- [Metal](https://developer.apple.com/metal/): an [implementation](https://github.com/philipturner/metal-flash-attention) by Philip Turner. This ports FlashAttention to mobile GPU architectures such as Apple silicon.