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

Remove QServe and FBGEMM FP8 quantization (#31109)

Browse Source
This commit is contained in:
Brayden Zhong
2026-07-17 17:10:34 -07:00
committed by GitHub
parent f926c30c57
commit 238b2b2c9c
18 changed files with 1 additions and 2566 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sgl-kernel/include/sgl_kernel_ops.h
-21
View File
@@ -726,27 +726,6 @@ void convert_vertical_slash_indexes_mergehead(
*/
void ApplyTokenBitmaskInplace(at::Tensor logits, at::Tensor bitmask, at::optional<at::Tensor> indices = at::nullopt);
/*
* From csrc/gemm (QServe)
*/
void qserve_w4a8_per_chn_gemm(
const torch::Tensor& _in_feats,
const torch::Tensor& _kernel,
const torch::Tensor& _wscales,
const torch::Tensor& _ascales,
const torch::Tensor& _w_szs,
const torch::Tensor& _a_ssums,
torch::Tensor& _out_feats);
void qserve_w4a8_per_group_gemm(
const torch::Tensor& _in_feats,
const torch::Tensor& _kernel,
const torch::Tensor& _zeros,
const torch::Tensor& _scales_i8,
const torch::Tensor& _wscales,
const torch::Tensor& _ascales,
torch::Tensor& _out_feats);
/*
* From csrc/quantization/gguf
*/