Tri Dao
|
82c1aa3514
Move PackGQA epilogue code to pack_gqa.h
|
2 mesiacov pred |
Tri Dao
|
df96486c31
Decode: varlen, paged KV, leftpad
|
2 mesiacov pred |
Ying Zhang
|
3669b25206
bwd benchmark + small fixes (#1129)
|
6 mesiacov pred |
Ying Zhang
|
c7f20a2d31
add cudnn benchmark for var-len
|
6 mesiacov pred |
Ying Zhang
|
dfe1a59e4b
Add var-seq-len to FA3 fp16 / bf16 fwd (#1072)
|
6 mesiacov pred |