[Kernel] Factor out epilogues from cutlass kernels (#5391)

Co-authored-by: Michael Goin <michael@neuralmagic.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: zifeitong <zifei.tong@parasail.io>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-neuralmagic@users.noreply.github.com>
This commit is contained in:
Tyler Michael Smith
2024-06-13 11:22:19 -07:00
committed by GitHub
co-authored by Michael Goin youkaichao zifeitong Robert Shaw
parent 0ce7b952f8
commit 85657b5607
12 changed files with 274 additions and 232 deletions
+4 -4
View File
@@ -136,10 +136,10 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
// CUTLASS w8a8 GEMM, supporting symmetric per-tensor or per-row/column
// quantization.
ops.def(
"cutlass_scaled_mm_dq(Tensor! out, Tensor a,"
" Tensor b, Tensor a_scales,"
" Tensor b_scales) -> ()");
ops.impl("cutlass_scaled_mm_dq", torch::kCUDA, &cutlass_scaled_mm_dq);
"cutlass_scaled_mm(Tensor! out, Tensor a,"
" Tensor b, Tensor a_scales,"
" Tensor b_scales) -> ()");
ops.impl("cutlass_scaled_mm", torch::kCUDA, &cutlass_scaled_mm);
#endif
// Quantized GEMM for GPTQ.