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

[2/n]decouple quantization implementation from vLLM dependency (#8112)

Browse Source 
Co-authored-by: walker-ai <yiyun.wyt@antgroup.com>
Co-authored-by: leoneo <1320612015@qq.com>
This commit is contained in:
Peng Zhang
2025-08-14 18:19:03 +08:00
committed by GitHub
parent 4dbf43601d
commit 5aa1ebd242
32 changed files with 6506 additions and 202 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
sgl-kernel/csrc/moe/marlin_moe_wna16/marlin_template.h
View File

@@ -18,13 +18,12 @@
/*
* Adapted from https://github.com/IST-DASLab/marlin
*/
#ifndef MARLIN_NAMESPACE_NAME
#define MARLIN_NAMESPACE_NAME marlin_moe_wna16
#endif
#include "gptq_marlin/marlin.cuh"
#include "gptq_marlin/marlin_dtypes.cuh"
#include "gemm/marlin/marlin.cuh"
#include "gemm/marlin/marlin_dtypes.cuh"
#include "scalar_type.hpp"
#define STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t) \