david/flash-attention

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

Autor	SHA1 Mensaxe	Data
Chirag Jain	50896ec574 Make nvcc threads configurable via environment variable (#885)	hai 9 meses
Tri Dao	dc08ea1c33 Support H100 for other CUDA extensions	hai 1 ano
Tri Dao	fa6d1ce44f Add fused_dense and dropout_add_layernorm CUDA extensions	%!s(int64=2) %!d(string=hai) anos