mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-20 06:12:07 +00:00
e4201f44cf
* All integration tests back everywhere (too many failed CI). * Upgrade integration tests after 12.4 * Attempt to remove the specifed compute cap. * Common arch list. * Punica uses raw ASM which is not valid on 9.0 apparently.
22 lines
662 B
Python
22 lines
662 B
Python
from setuptools import setup
|
|
from torch.utils.cpp_extension import BuildExtension, CUDAExtension
|
|
|
|
extra_compile_args = ["-std=c++17"]
|
|
|
|
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=extra_compile_args,
|
|
),
|
|
CUDAExtension(
|
|
name="custom_kernels.fused_attention_cuda",
|
|
sources=["custom_kernels/fused_attention_cuda.cu"],
|
|
extra_compile_args=extra_compile_args,
|
|
),
|
|
],
|
|
cmdclass={"build_ext": BuildExtension},
|
|
)
|