A bug when running python DCTK.py

[wu-dd]

When I run the code python DCTK.py according to the instructions in the readme, I encounter some errors (as shown below). I would like to ask if you know the possible cause or if there are other steps that can help me circumvent this problem?

If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST'].
warnings.warn(
Traceback (most recent call last):
File "/home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/utils/cpp_extension.py", line 2107, in _run_ninja_build
subprocess.run(
File "/home/ddw/miniconda3/envs/env-wdd/lib/python3.9/subprocess.py", line 528, in run
raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
File "/home/ddw/Coding/Diffusion_Classifier/DCTK.py", line 2, in
from models.unets import get_edm_cifar_cond
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/init.py", line 2, in
from .score_sde import get_NCSNPP, get_NCSNPP_cached, DiffusionClassifierCached
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/init.py", line 1, in
from .unets import get_NCSNPP_cached, get_NCSNPP
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/unets.py", line 1, in
from .models.ncsnpp import NCSNpp
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/models/init.py", line 15, in
from . import ncsnpp
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/models/ncsnpp.py", line 18, in
from . import utils, layers, layerspp, normalization
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/models/layerspp.py", line 20, in
from . import up_or_down_sampling
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/models/up_or_down_sampling.py", line 18, in
from ..op import upfirdn2d
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/op/init.py", line 9, in
from .fused_act import FusedLeakyReLU, fused_leaky_relu
File "/home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/op/fused_act.py", line 19, in
fused = load(
File "/home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/utils/cpp_extension.py", line 1309, in load
return jit_compile(
File "/home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/utils/cpp_extension.py", line 1719, in jit_compile
write_ninja_file_and_build_library(
File "/home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/utils/cpp_extension.py", line 1832, in write_ninja_file_and_build_library
run_ninja_build(
File "/home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/utils/cpp_extension.py", line 2123, in run_ninja_build
raise RuntimeError(message) from e
RuntimeError: Error building extension 'fused': [1/3] /home/ddw/download/cuda-11.6/bin/nvcc --generate-dependencies-with-compile --dependency-output fused_bias_act_kernel.cuda.o.d -DTORCH_EXTENSION_NAME=fused -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE="gcc" -DPYBIND11_STDLIB="libstdcpp" -DPYBIND11_BUILD_ABI="cxxabi1011" -isystem /home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/include -isystem /home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/include/torch/csrc/api/include -isystem /home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/include/TH -isystem /home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/include/THC -isystem /home/ddw/download/cuda-11.6/include -isystem /home/ddw/miniconda3/envs/env-wdd/include/python3.9 -D_GLIBCXX_USE_CXX11_ABI=0 -D__CUDA_NO_HALF_OPERATORS -D__CUDA_NO_HALF_CONVERSIONS -D__CUDA_NO_BFLOAT16_CONVERSIONS -D__CUDA_NO_HALF2_OPERATORS --expt-relaxed-constexpr -gencode=arch=compute_89,code=compute_89 -gencode=arch=compute_89,code=sm_89 --compiler-options '-fPIC' -std=c++17 -c /home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/op/fused_bias_act_kernel.cu -o fused_bias_act_kernel.cuda.o
FAILED: fused_bias_act_kernel.cuda.o
/home/ddw/download/cuda-11.6/bin/nvcc --generate-dependencies-with-compile --dependency-output fused_bias_act_kernel.cuda.o.d -DTORCH_EXTENSION_NAME=fused -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE="gcc" -DPYBIND11_STDLIB="libstdcpp" -DPYBIND11_BUILD_ABI="cxxabi1011" -isystem /home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/include -isystem /home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/include/torch/csrc/api/include -isystem /home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/include/TH -isystem /home/ddw/miniconda3/envs/env-wdd/lib/python3.9/site-packages/torch/include/THC -isystem /home/ddw/download/cuda-11.6/include -isystem /home/ddw/miniconda3/envs/env-wdd/include/python3.9 -D_GLIBCXX_USE_CXX11_ABI=0 -D__CUDA_NO_HALF_OPERATORS -D__CUDA_NO_HALF_CONVERSIONS -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_89,code=compute_89 -gencode=arch=compute_89,code=sm_89 --compiler-options '-fPIC' -std=c++17 -c /home/ddw/Coding/Diffusion_Classifier/models/unets/score_sde/op/fused_bias_act_kernel.cu -o fused_bias_act_kernel.cuda.o
nvcc fatal : Unsupported gpu architecture 'compute_89'

[huanranchen]

你好！
非常感谢您对我的工作感兴趣！
看样子这个是在编译SongYang的NCSNpp网络时出现了错误。SongYang的这个网络需要ninja，可能确实不适配某些gpu。
您或许可以查阅songyang NCSNpp的GitHub issue。我更推荐另一种解决方案。Diffusion Classifier已经支持EDM了，您其实可以直接把songyang NCSNpp这部分代码删掉，就不会出现这个问题了。

[wu-dd]

谢谢您的回答！