 Tri Dao
|
2c996ca25f
Use SeqlenInfo for bwd and epilogue
|
il y a 1 semaine |
|
Tri Dao
|
29cdfedd80
Use Bulk reduce instead of TMA for dQaccum, split across WGs
|
il y a 1 semaine |
|
Tri Dao
|
82dc825759
Don't use the unsafe convert_type function
|
il y a 2 semaines |
|
Tri Dao
|
df96486c31
Decode: varlen, paged KV, leftpad
|
il y a 1 mois |
|
Tri Dao
|
6e8b25e426
Refactor
|
il y a 2 mois |
 Ying Zhang
|
db80387343
Add seqused_q in fwd / bwd and seqused_k in bwd.
|
il y a 3 mois |
 Cameron Shinn
|
3cea2fb6ee
Add ArchTag to pre/postprocess bwd kernels (#1180)
|
il y a 3 mois |
|
Tri Dao
|
bafe253042
[FA3] Bwd
|
il y a 4 mois |
|
Tri Dao
|
7f67966cc7
FA3 initial code release
|
il y a 5 mois |
