 Jay Shah
|
bb169e2a54
revert to single tile scheduler
|
1 mese fa |
|
jayhshah
|
a5a75274bc
FA3 kvcache + split kv + gqa parallelization (#1236)
|
2 mesi fa |
|
jayhshah
|
c92ca63268
FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173)
|
3 mesi fa |
|
jayhshah
|
5018ac6ac5
Fp8 kernel with "in-kernel" transpose of V in producer (#1100)
|
4 mesi fa |
 Tri Dao
|
3aae9c18c1
Revert "Changes For FP8 (#1075)"
|
4 mesi fa |
 ganeshcolfax
|
1899c970c8
Changes For FP8 (#1075)
|
4 mesi fa |
|
Tri Dao
|
7f67966cc7
FA3 initial code release
|
5 mesi fa |
