 Tri Dao
|
0890032358
Implement backward pass for Sm80
|
1 ماه پیش |
|
Tri Dao
|
14894c5717
Make BwdPostprocessKernel work with Sm80
|
2 ماه پیش |
|
Tri Dao
|
2c996ca25f
Use SeqlenInfo for bwd and epilogue
|
2 ماه پیش |
|
Tri Dao
|
29cdfedd80
Use Bulk reduce instead of TMA for dQaccum, split across WGs
|
3 ماه پیش |
|
Tri Dao
|
82dc825759
Don't use the unsafe convert_type function
|
3 ماه پیش |
|
Tri Dao
|
6e8b25e426
Refactor
|
5 ماه پیش |
 Ying Zhang
|
db80387343
Add seqused_q in fwd / bwd and seqused_k in bwd.
|
6 ماه پیش |
 Cameron Shinn
|
3cea2fb6ee
Add ArchTag to pre/postprocess bwd kernels (#1180)
|
6 ماه پیش |
|
Tri Dao
|
bafe253042
[FA3] Bwd
|
7 ماه پیش |
