 Jay Shah
|
7c1473e0e5
remove Is_batch_dynamic from seqlen traits and handle fp8 perf regression using smem boolean
|
2 months ago |
|
Jay Shah
|
b0f067efdc
revert epi change for fp8 due to measured perf regression
|
2 months ago |
|
Jay Shah
|
eb9c0ee22a
add rmem -> gmem for fp8
|
2 months ago |
|
Jay Shah
|
d437d3dd5c
remove smem usage for when rmem -> gmem epilogue is used
|
2 months ago |
|
Jay Shah
|
5e3864f2ee
change default output type of fp8 kernel to bf16
|
2 months ago |
|
Jay Shah
|
2472e5e0b4
add 'in principle' fp8 kv cache support
|
2 months ago |
|
Jay Shah
|
68ff3f7bee
add 1 mma warpgroup option, enable splitkv for hdim 256
|
3 months ago |
|
Jay Shah
|
0a1a0c22b6
refactor for split kv
|
3 months ago |
|
Jay Shah
|
ccf5b9bb0e
modify Ktraits for decoding QO layouts
|
3 months ago |
|
jayhshah
|
c92ca63268
FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173)
|
3 months ago |
|
jayhshah
|
5018ac6ac5
Fp8 kernel with "in-kernel" transpose of V in producer (#1100)
|
4 months ago |
 Tri Dao
|
3aae9c18c1
Revert "Changes For FP8 (#1075)"
|
4 months ago |
 ganeshcolfax
|
1899c970c8
Changes For FP8 (#1075)
|
4 months ago |
|
Tri Dao
|
7f67966cc7
FA3 initial code release
|
5 months ago |
