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

[1/n] apply wna16marlin kernel in moe weight only quantization (#7683)

Browse Source 
Co-authored-by: 晟海 <huangtingwei.htw@antgroup.com>
Co-authored-by: yych0745 <1398089567@qq.com>
Co-authored-by: HandH1998 <1335248067@qq.com>
Co-authored-by: 弋云 <yiyun.wyt@antgroup.com>
Co-authored-by: walker-ai <2398833647@qq.com>
This commit is contained in:
AniZpZ
2025-07-02 14:21:25 +08:00
committed by GitHub
parent b3fa5dc3c8
commit 8e03b641ba
27 changed files with 6104 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

89
sgl-kernel/csrc/moe/marlin_moe_wna16/kernel_bf16_ku4.cu Normal file
View File

@@ -0,0 +1,89 @@
// auto generated by generate.py
// clang-format off
#include "kernel.h"
#include "marlin_template.h"
namespace MARLIN_NAMESPACE_NAME {
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 1, 8, 8, true, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 1, 8, 4, true, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 1, 8, 8, false, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 1, 8, 4, false, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 2, 16, 4, false, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 2, 8, 4, false, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 3, 16, 4, false, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 3, 8, 4, false, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 4, 16, 4, false, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 4, 8, 4, false, pipe_stages, false, true, -1, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 1, 8, 8, true, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 1, 8, 4, true, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 1, 8, 8, false, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 1, 8, 4, false, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 2, 16, 4, false, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 2, 8, 4, false, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 3, 16, 4, false, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 3, 8, 4, false, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 4, 16, 4, false, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 4, 8, 4, false, pipe_stages, false, true, 2, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 1, 8, 8, true, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 1, 8, 4, true, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 1, 8, 8, false, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 1, 8, 4, false, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 2, 16, 4, false, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 2, 8, 4, false, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 3, 16, 4, false, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 3, 8, 4, false, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 4, 16, 4, false, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 4, 8, 4, false, pipe_stages, false, true, 4, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 1, 8, 8, true, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 1, 8, 4, true, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 1, 8, 8, false, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 1, 8, 4, false, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 2, 16, 4, false, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 2, 8, 4, false, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 3, 16, 4, false, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 3, 8, 4, false, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 256, 4, 16, 4, false, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
template __global__ void Marlin<nv_bfloat16, sglang::kU4.id(), 128, 4, 8, 4, false, pipe_stages, false, true, 8, false>( MARLIN_KERNEL_PARAMS );
}