minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

support cutlass fp4 kernel in sm120 (#11737)

Browse Source
This commit is contained in:
AichenF
2025-10-29 12:25:16 -07:00
committed by GitHub
parent d717e73ea8
commit ed1044ac1b
6 changed files with 543 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sgl-kernel/csrc/gemm/nvfp4_quant_kernels.cu
+5 -2
View File
@@ -199,8 +199,11 @@ inline int getMultiProcessorCount() {
return multi_processor_count; // Return the cached value on subsequent calls
}
void scaled_fp4_quant_sm100a(
torch::Tensor& output, torch::Tensor const& input, torch::Tensor& output_sf, torch::Tensor const& input_sf) {
void scaled_fp4_quant_sm100a_sm120a(
torch::Tensor const& output,
torch::Tensor const& input,
torch::Tensor const& output_sf,
torch::Tensor const& input_sf) {
auto sm_version = getSMVersion();
TORCH_CHECK(sm_version >= 100, "fp4_quant is only supported on sm100+");