david/flash-attention

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

Autor	SHA1 Mensaxe	Data
Tri Dao	88173a1aaf [FusedDense] Support relu, rename FusedDenseGeluDense -> FusedMLP	hai 1 ano
Tri Dao	43798966cf [Docs] Fix formatting	hai 1 ano
Tri Dao	3c7cbfc195 [Docs] Mention that dropout_layer_norm supports all dims up to 6k	hai 1 ano
Tri Dao	4a6eaa9f27 Update configs, add results	%!s(int64=2) %!d(string=hai) anos
Tri Dao	0bf5e50038 Release training code	%!s(int64=2) %!d(string=hai) anos