 Jay Shah
|
c06cc0ba9f
change cu_seqlens_k to seqused_k for kv cache api
|
2 months ago |
|
Jay Shah
|
35f3542442
refactor names
|
2 months ago |
|
Jay Shah
|
bb230b8c54
separate gqa compilation
|
3 months ago |
 Ganesh Bikshandi
|
9c97808349
changes for correct lse write out for splits=1 and splits > 1 case.
|
3 months ago |
|
Jay Shah
|
0a1a0c22b6
refactor for split kv
|
4 months ago |
|
Jay Shah
|
535b8279bb
complete gqa parallel changes for non-causal
|
4 months ago |
|
Jay Shah
|
38ad0ac701
add new param to handle cache_batch_size
|
4 months ago |
 Ying Zhang
|
7b4e68e04f
hopper local attention
|
4 months ago |
|
Ying Zhang
|
db80387343
Add seqused_q in fwd / bwd and seqused_k in bwd.
|
4 months ago |
|
jayhshah
|
c92ca63268
FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173)
|
4 months ago |
 Tri Dao
|
bafe253042
[FA3] Bwd
|
5 months ago |
 Ying Zhang
|
dfe1a59e4b
Add var-seq-len to FA3 fp16 / bf16 fwd (#1072)
|
5 months ago |
 Cameron Shinn
|
cb516f855b
Remove torchlib dependency from cpp files (#1083)
|
5 months ago |
|
Tri Dao
|
74b0761ff7
[FA3] BF16 forward
|
5 months ago |
|
Tri Dao
|
7f67966cc7
FA3 initial code release
|
5 months ago |
