 Tri Dao
|
7a802796e1
Big refactor and update
|
hai 2 semanas |
 Kai Londenberg
|
284e2c6e5b
Make FA3 paged attention ready for upgrade to Cutlass 3.6 (#1331)
|
hai 2 meses |
 Kai Londenberg
|
b443207c1f
Paged Attention support for FA3 (#1268)
|
hai 2 meses |
 jayhshah
|
a5a75274bc
FA3 kvcache + split kv + gqa parallelization (#1236)
|
hai 3 meses |
 Ying Zhang
|
dff976a84a
fixes
|
hai 4 meses |
|
Ying Zhang
|
7b4e68e04f
hopper local attention
|
hai 4 meses |
|
Ying Zhang
|
db80387343
Add seqused_q in fwd / bwd and seqused_k in bwd.
|
hai 4 meses |
 Charlene Yang
|
bdf733be55
Add q, k, v descales to FA3 interface (#1210)
|
hai 4 meses |
|
jayhshah
|
c92ca63268
FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173)
|
hai 4 meses |
|
Tri Dao
|
bafe253042
[FA3] Bwd
|
hai 5 meses |
|
Tri Dao
|
3aae9c18c1
Revert "Changes For FP8 (#1075)"
|
hai 6 meses |
 ganeshcolfax
|
1899c970c8
Changes For FP8 (#1075)
|
hai 6 meses |
 Ying Zhang
|
dfe1a59e4b
Add var-seq-len to FA3 fp16 / bf16 fwd (#1072)
|
hai 6 meses |
 youkaichao
|
ef3e358a25
remove lambda (#1056)
|
hai 6 meses |
|
Tri Dao
|
7f67966cc7
FA3 initial code release
|
hai 6 meses |
