chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 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

9
sgl-kernel/include/sgl_kernel_ops.h
View File

@@ -936,15 +936,6 @@ void es_sm100_mxfp8_blockscaled_grouped_quant(
torch::Tensor& quant_output,
torch::Tensor& scale_factor);
/*
* From fast-hadamard-transform
*/
torch::Tensor fast_hadamard_transform(torch::Tensor& x, double scale);
torch::Tensor fast_hadamard_transform_12N(torch::Tensor& x, double scale);
torch::Tensor fast_hadamard_transform_20N(torch::Tensor& x, double scale);
torch::Tensor fast_hadamard_transform_28N(torch::Tensor& x, double scale);
torch::Tensor fast_hadamard_transform_40N(torch::Tensor& x, double scale);
/*
* From flashmla
*/