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

Fix the style of sgl kernel (#10398)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-09-12 22:20:21 -07:00
committed by GitHub
parent 99757cc3e6
commit c9ec4cae5b
5 changed files with 19 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
sgl-kernel/csrc/common_extension.cc
View File

@@ -99,6 +99,11 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
"mult, int offset, int cuda_stream) -> ()");
m.impl("downcast_fp8", torch::kCUDA, &downcast_fp8);
m.def("copy_to_gpu_no_ce(Tensor input, Tensor! output) -> ()");
m.impl("copy_to_gpu_no_ce", torch::kCUDA, &copy_to_gpu_no_ce);
m.def("concat_mla_k(Tensor! k, Tensor k_nope, Tensor k_rope) -> ()");
m.impl("concat_mla_k", torch::kCUDA, &concat_mla_k);
/*
* From csrc/gemm
*/
@@ -447,11 +452,6 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
"Tensor _ascales, Tensor! _out_feats) -> ()");
m.impl("qserve_w4a8_per_group_gemm", torch::kCUDA, &qserve_w4a8_per_group_gemm);
m.def("copy_to_gpu_no_ce(Tensor input, Tensor! output) -> ()");
m.impl("copy_to_gpu_no_ce", torch::kCUDA, &copy_to_gpu_no_ce);
m.def("concat_mla_k(Tensor! k, Tensor k_nope, Tensor k_rope) -> ()");
m.impl("concat_mla_k", torch::kCUDA, &concat_mla_k);
/*
* From csrc/mamba
*/