Fast transformer inference with Metal Performance Shaders (opens in new tab)

(explosion.ai)

3 pointsreichardt3y ago1 comments

1 comments

1 comments · 1 top-level

Seems like the Apple M1 Max is almost half as fast as a RTX 3090. Pretty cool!

j / k navigate · click thread line to collapse