biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Bugfix][Misc] Fix silu_and_mul_nvfp4_quant issue and extract common utils for nvfp4 kernel source files (#23727)

Browse Source 
Signed-off-by: elvischenv <219235043+elvischenv@users.noreply.github.com>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
This commit is contained in:
elvischenv
2025-09-05 05:25:45 +08:00
committed by GitHub
parent 60b755cbcb
commit adc3ddb430
11 changed files with 382 additions and 718 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
csrc/quantization/fp4/nvfp4_quant_entry.cu
View File

@@ -32,6 +32,14 @@ void scaled_fp4_experts_quant_sm100a(
torch::Tensor const& output_scale_offset_by_experts);
#endif
#if (defined(ENABLE_NVFP4_SM100) && ENABLE_NVFP4_SM100) || \
(defined(ENABLE_NVFP4_SM120) && ENABLE_NVFP4_SM120)
void silu_and_mul_nvfp4_quant_sm1xxa(torch::Tensor& output,
torch::Tensor& output_sf,
torch::Tensor& input,
torch::Tensor& input_sf);
#endif
void scaled_fp4_quant(torch::Tensor& output, torch::Tensor const& input,
torch::Tensor& output_sf, torch::Tensor const& input_sf) {
#if (defined(ENABLE_NVFP4_SM100) && ENABLE_NVFP4_SM100) || \
@@ -54,3 +62,13 @@ void scaled_fp4_experts_quant(
TORCH_CHECK_NOT_IMPLEMENTED(false,
"No compiled nvfp4 experts quantization kernel");
}
void silu_and_mul_nvfp4_quant(torch::Tensor& output, torch::Tensor& output_sf,
torch::Tensor& input, torch::Tensor& input_sf) {
#if (defined(ENABLE_NVFP4_SM100) && ENABLE_NVFP4_SM100) || \
(defined(ENABLE_NVFP4_SM120) && ENABLE_NVFP4_SM120)
return silu_and_mul_nvfp4_quant_sm1xxa(output, output_sf, input, input_sf);
#endif
TORCH_CHECK_NOT_IMPLEMENTED(
false, "No compiled silu_and_mul nvfp4 quantization kernel");
}