sync the upstream updates of flashinfer (#3051)

This commit is contained in:
Yineng Zhang
2025-01-22 20:33:13 +08:00
committed by GitHub
parent 9f8f2c7f74
commit bcda0c9ee6
3 changed files with 8 additions and 1 deletions

View File

@@ -40,6 +40,7 @@ jobs:
run: |
pip3 install torch==2.5.1 && pip3 install pytest && pip3 install vllm
pip3 uninstall sgl-kernel -y || true
find . -name index.lock -delete
cd sgl-kernel
git submodule deinit --all --force && git submodule sync --recursive && git submodule update --init --force --recursive
pip3 install .

View File

@@ -47,6 +47,7 @@ include_dirs = [
cutlass.resolve() / "tools" / "util" / "include",
root / "src" / "sgl-kernel" / "csrc",
flashinfer.resolve() / "include",
flashinfer.resolve() / "include" / "gemm",
flashinfer.resolve() / "csrc",
]
nvcc_flags = [
@@ -91,7 +92,12 @@ ext_modules = [
"src/sgl-kernel/csrc/sampling_scaling_penalties.cu",
"src/sgl-kernel/csrc/sgl_kernel_ops.cu",
"src/sgl-kernel/csrc/rotary_embedding.cu",
"3rdparty/flashinfer/csrc/activation.cu",
"3rdparty/flashinfer/csrc/bmm_fp8.cu",
"3rdparty/flashinfer/csrc/group_gemm.cu",
"3rdparty/flashinfer/csrc/group_gemm_sm90.cu",
"3rdparty/flashinfer/csrc/norm.cu",
"3rdparty/flashinfer/csrc/sampling.cu",
],
include_dirs=include_dirs,
extra_compile_args={