Ying Zhang
|
496fdc4f6c
Add seqused_q in fwd / bwd and seqused_k in bwd.
|
há 5 meses atrás |
jayhshah
|
c92ca63268
FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173)
|
há 5 meses atrás |
Tri Dao
|
bafe253042
[FA3] Bwd
|
há 6 meses atrás |
Tri Dao
|
3aae9c18c1
Revert "Changes For FP8 (#1075)"
|
há 6 meses atrás |
ganeshcolfax
|
1899c970c8
Changes For FP8 (#1075)
|
há 6 meses atrás |
Ying Zhang
|
dfe1a59e4b
Add var-seq-len to FA3 fp16 / bf16 fwd (#1072)
|
há 6 meses atrás |
youkaichao
|
ef3e358a25
remove lambda (#1056)
|
há 6 meses atrás |
Tri Dao
|
7f67966cc7
FA3 initial code release
|
há 6 meses atrás |