squall/flash-attention
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
08c295c043
flash-attention/csrc/flash_attn
History
Tri Dao ee77b931b9 Swap seqlen_q and nheads for MQA to speed it up (h/t Daniel Haziza)
2023-09-10 22:56:33 -07:00
..
src Swap seqlen_q and nheads for MQA to speed it up (h/t Daniel Haziza) 2023-09-10 22:56:33 -07:00
flash_api.cpp Swap seqlen_q and nheads for MQA to speed it up (h/t Daniel Haziza) 2023-09-10 22:56:33 -07:00