[Feature] Apply Cublas Grouped Gemm kernel (#3629)

This commit is contained in:
Baizhou Zhang
2025-02-17 23:18:31 -08:00
committed by GitHub
parent 07ab4d4a2d
commit 67fc595bb8
9 changed files with 513 additions and 1 deletions

View File

@@ -102,6 +102,7 @@ sources = [
"src/sgl-kernel/csrc/eagle_utils.cu",
"src/sgl-kernel/csrc/speculative_sampling.cu",
"src/sgl-kernel/csrc/per_token_group_quant_fp8.cu",
"src/sgl-kernel/csrc/cublas_grouped_gemm.cu",
"3rdparty/flashinfer/csrc/activation.cu",
"3rdparty/flashinfer/csrc/bmm_fp8.cu",
"3rdparty/flashinfer/csrc/norm.cu",