apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-25 09:57:51 +08:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

init (#6642)

Browse Source 
Co-authored-by: gongweibao <gognweibao@baidu.com>
This commit is contained in:
gongweibao
2026-03-04 21:55:31 +08:00
committed by GitHub
parent 5c8f5184d9
commit ddb06ff83f
306 changed files with 40627 additions and 34418 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/gpu_ops/cutlass_extensions/gemm/dispatch_policy.hpp
+4 -2
View File
@@ -30,7 +30,8 @@ struct KernelTmaWarpSpecializedCooperativeFP8BlockScaledSubGroupMAccum
// n-buffer in smem (Hopper TMA), pipelined with Hopper GMMA and TMA, Warp
// specialized dynamic schedule For FP8 kernels with Block Scaling
template <int Stages_, class ClusterShape_ = Shape<_1, _1, _1>,
template <int Stages_,
class ClusterShape_ = Shape<_1, _1, _1>,
class KernelSchedule = KernelTmaWarpSpecialized,
int ScaleGranularityM =
0 // `ScaleGranularityM` specifies scaling granularity along M,
@@ -38,7 +39,8 @@ template <int Stages_, class ClusterShape_ = Shape<_1, _1, _1>,
// granularity is `size<0>(TileShape_MNK{})` along M.
>
struct MainloopSm90TmaGmmaWarpSpecializedBlockScalingSubGroupMFP8
: MainloopSm90TmaGmmaWarpSpecialized<Stages_, ClusterShape_,
: MainloopSm90TmaGmmaWarpSpecialized<Stages_,
ClusterShape_,
KernelSchedule> {
static_assert(
cute::is_same_v<