david/flash-attention

espejo de https://github.com/Dao-AILab/flash-attention

Autor	SHA1 Mensaje	Fecha
jayhshah	c92ca63268 FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173)	hace 3 meses
jayhshah	5018ac6ac5 Fp8 kernel with "in-kernel" transpose of V in producer (#1100)	hace 4 meses
Tri Dao	3aae9c18c1 Revert "Changes For FP8 (#1075)"	hace 4 meses
ganeshcolfax	1899c970c8 Changes For FP8 (#1075)	hace 4 meses