david/flash-attention

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

Autor	SHA1 Mensaje	Fecha
Tri Dao	2c996ca25f Use SeqlenInfo for bwd and epilogue	hace 1 semana
Tri Dao	29cdfedd80 Use Bulk reduce instead of TMA for dQaccum, split across WGs	hace 1 semana
Tri Dao	82dc825759 Don't use the unsafe convert_type function	hace 2 semanas
Tri Dao	6e8b25e426 Refactor	hace 2 meses
Ying Zhang	db80387343 Add seqused_q in fwd / bwd and seqused_k in bwd.	hace 3 meses
Cameron Shinn	3cea2fb6ee Add ArchTag to pre/postprocess bwd kernels (#1180)	hace 3 meses
Tri Dao	bafe253042 [FA3] Bwd	hace 4 meses