david/flash-attention

mirrorاز https://github.com/Dao-AILab/flash-attention

نویسنده	SHA1 پیام	تاریخ
jayhshah	c92ca63268 FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173)	5 ماه پیش
jayhshah	5018ac6ac5 Fp8 kernel with "in-kernel" transpose of V in producer (#1100)	6 ماه پیش
Tri Dao	3aae9c18c1 Revert "Changes For FP8 (#1075)"	6 ماه پیش
ganeshcolfax	1899c970c8 Changes For FP8 (#1075)	6 ماه پیش
Tri Dao	7f67966cc7 FA3 initial code release	7 ماه پیش