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

[Perf][Kernel] Fused SiLU+Mul+Quant kernel for NVFP4 cutlass_moe (#31832)

Browse Source 
Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
Michael Goin
2026-01-09 09:40:33 -05:00
committed by GitHub
parent 8e27663b6a
commit 34cd32fe30
8 changed files with 278 additions and 95 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
csrc/torch_bindings.cpp
View File

@@ -558,6 +558,15 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
"Tensor output_scale_offset_by_experts) -> ()");
ops.impl("scaled_fp4_experts_quant", torch::kCUDA, &scaled_fp4_experts_quant);
// Fused SiLU+Mul+NVFP4 experts quantization.
ops.def(
"silu_and_mul_scaled_fp4_experts_quant(Tensor! output, Tensor! "
"output_scale,"
"Tensor input, Tensor input_global_scale, Tensor input_offset_by_experts,"
"Tensor output_scale_offset_by_experts) -> ()");
ops.impl("silu_and_mul_scaled_fp4_experts_quant", torch::kCUDA,
&silu_and_mul_scaled_fp4_experts_quant);
// Check if cutlass_scaled_mm_fp4 is supported for CUDA devices
// of the given capability
ops.def("cutlass_scaled_mm_supports_fp4(int cuda_device_capability) -> bool");