apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-24 09:44:10 +08:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

moe preprocess op support 160 experts and fused_moe triton kernel name add K (#3121)

Browse Source
This commit is contained in:
chen
2025-08-01 10:46:20 +08:00
committed by GitHub
parent 1d93565082
commit a2f5cc54f8
3 changed files with 47 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/gpu_ops/moe/tritonmoe_preprocess.cu
+2
View File
@@ -168,6 +168,8 @@ std::vector<paddle::Tensor> tritonmoe_preprocess_kernel(const paddle::Tensor& to
run_align_kernel(64);
} else if (num_experts == 128) {
run_align_kernel(128);
} else if (num_experts == 160) {
run_align_kernel(160);
} else {
PD_THROW("Not support num_experts: %d", num_experts);
}