hf_text-generation-inference/server/custom_kernels/setup.py

20 lines
657 B
Python

from setuptools import setup
from torch.utils.cpp_extension import BuildExtension, CUDAExtension
setup(
name="custom_kernels",
ext_modules=[
CUDAExtension(
name="custom_kernels.fused_bloom_attention_cuda",
sources=["custom_kernels/fused_bloom_attention_cuda.cu"],
extra_compile_args=["-arch=compute_80", "-std=c++17"],
),
CUDAExtension(
name="custom_kernels.fused_attention_cuda",
sources=["custom_kernels/fused_attention_cuda.cu"],
extra_compile_args=["-arch=compute_80", "-std=c++17"],
),
],
cmdclass={"build_ext": BuildExtension},
)