minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[kernel slimming] Move fast_hadamard_transform to jit_kernel (#18475)

Browse Source
This commit is contained in:
Xiaoyu Zhang
2026-02-14 23:06:21 +08:00
committed by GitHub
parent ae95869292
commit c29394e3c8
17 changed files with 2170 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sgl-kernel/csrc/common_extension.cc
-18
View File
@@ -591,24 +591,6 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
"es_sm100_mxfp8_blockscaled_grouped_quant(Tensor input, Tensor problem_sizes, Tensor expert_offsets, Tensor "
"blockscale_offsets, Tensor quant_output, Tensor scale_factor) -> () ");
m.impl("es_sm100_mxfp8_blockscaled_grouped_quant", &es_sm100_mxfp8_blockscaled_grouped_quant);
/*
* From fast-hadamard-transform
*/
m.def("fast_hadamard_transform(Tensor x, float scale) -> Tensor");
m.impl("fast_hadamard_transform", torch::kCUDA, &fast_hadamard_transform);
m.def("fast_hadamard_transform_12N(Tensor x, float scale) -> Tensor");
m.impl("fast_hadamard_transform_12N", torch::kCUDA, &fast_hadamard_transform_12N);
m.def("fast_hadamard_transform_20N(Tensor x, float scale) -> Tensor");
m.impl("fast_hadamard_transform_20N", torch::kCUDA, &fast_hadamard_transform_20N);
m.def("fast_hadamard_transform_28N(Tensor x, float scale) -> Tensor");
m.impl("fast_hadamard_transform_28N", torch::kCUDA, &fast_hadamard_transform_28N);
m.def("fast_hadamard_transform_40N(Tensor x, float scale) -> Tensor");
m.impl("fast_hadamard_transform_40N", torch::kCUDA, &fast_hadamard_transform_40N);
}
REGISTER_EXTENSION(common_ops)