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

Fix kernel type annotations for fp8 quant and logging (#16994)

Browse Source
This commit is contained in:
Lianmin Zheng
2026-01-13 18:14:32 -08:00
committed by GitHub
parent afe285f7bd
commit a4825ed588
7 changed files with 797 additions and 684 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sgl-kernel/csrc/common_extension.cc
+4 -4
View File
@@ -142,19 +142,19 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
m.impl("fp8_blockwise_scaled_mm", torch::kCUDA, &fp8_blockwise_scaled_mm);
m.def(
"sgl_per_token_group_quant_8bit(Tensor input, Tensor output_q, Tensor output_s, int group_size,"
"sgl_per_token_group_quant_8bit(Tensor input, Tensor! output_q, Tensor! output_s, int group_size,"
" float eps, float fp8_min, float fp8_max, bool scale_ue8m0) -> ()");
m.impl("sgl_per_token_group_quant_8bit", torch::kCUDA, &sgl_per_token_group_quant_8bit);
m.def(
"sgl_per_token_group_quant_8bit_v2(Tensor input, Tensor output_q, Tensor output_s, int group_size,"
"sgl_per_token_group_quant_8bit_v2(Tensor input, Tensor! output_q, Tensor! output_s, int group_size,"
" float eps, float fp8_min, float fp8_max, bool scale_ue8m0, bool fuse_silu_and_mul, Tensor? masked_m) -> ()");
m.impl("sgl_per_token_group_quant_8bit_v2", torch::kCUDA, &sgl_per_token_group_quant_8bit_v2);
m.def("sgl_per_tensor_quant_fp8(Tensor input, Tensor output_q, Tensor output_s, bool is_static) -> ()");
m.def("sgl_per_tensor_quant_fp8(Tensor input, Tensor! output_q, Tensor! output_s, bool is_static) -> ()");
m.impl("sgl_per_tensor_quant_fp8", torch::kCUDA, &sgl_per_tensor_quant_fp8);
m.def("sgl_per_token_quant_fp8(Tensor input, Tensor output_q, Tensor output_s) -> ()");
m.def("sgl_per_token_quant_fp8(Tensor input, Tensor! output_q, Tensor! output_s) -> ()");
m.impl("sgl_per_token_quant_fp8", torch::kCUDA, &sgl_per_token_quant_fp8);
m.def(