david/flash-attention

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

Autor	SHA1 Mensaxe	Data
Tri Dao	7f5d73a162 Add env var to disable specific hdim	hai 1 mes
Tri Dao	3e5d77a102 Group instantiations for different hdims together	hai 1 mes
Tri Dao	fc2fd95a18 Renable FP8 kernels	hai 2 meses
Tri Dao	64d92bce53 Split PagedKV into separate .cu files to speed up compilation	hai 2 meses
Tri Dao	586ba914bb Move fwd tile size to a separate file	hai 2 meses
Tri Dao	018b9af683 Move .cu files to instantiations, use generate_kernels.py	hai 2 meses