EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[1/2][resubmit again] sgl-kernel: Fuse routed scaling factor into moe_fused_gate (#9088)

Browse Source
This commit is contained in:
Trevor Morris
2025-08-12 20:12:38 -07:00
committed by GitHub
parent 8723b4f146
commit 13c48dcf88
4 changed files with 32 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
sgl-kernel/csrc/common_extension.cc
View File

@@ -175,7 +175,7 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
m.def(
"moe_fused_gate(Tensor input, Tensor bias, int num_expert_group, int topk_group, int topk, int "
"num_fused_shared_experts, float routed_scaling_factor) -> "
"num_fused_shared_experts, float routed_scaling_factor, bool apply_routed_scaling_factor_on_output) -> "
"(Tensor[])");
m.impl("moe_fused_gate", torch::kCUDA, &moe_fused_gate);
m.def(