an issue when install chamfer

[zjuruizhechen]

/usr/bin/nvcc -I/home/ruizhe/miniconda3/envs/crzz/lib/python3.7/site-packages/torch/include -I/home/ruizhe/miniconda3/envs/crzz/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -I/home/ruizhe/miniconda3/envs/crzz/lib/python3.7/site-packages/torch/include/TH -I/home/ruizhe/miniconda3/envs/crzz/lib/python3.7/site-packages/torch/include/THC -I/home/ruizhe/miniconda3/envs/crzz/include/python3.7m -c chamfer.cu -o build/temp.linux-x86_64-3.7/chamfer.o -DCUDA_NO_HALF_OPERATORS -DCUDA_NO_HALF_CONVERSIONS -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr --compiler-options '-fPIC' -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE="_gcc" -DPYBIND11_STDLIB="_libstdcpp" -DPYBIND11_BUILD_ABI="_cxxabi1011" -DTORCH_EXTENSION_NAME=chamfer -D_GLIBCXX_USE_CXX11_ABI=0 -gencode=arch=compute_86,code=sm_86 -std=c++14 nvcc fatal : Unsupported gpu architecture 'compute_86' error: command '/usr/bin/nvcc' failed with exit status 1

Thanks a lot for your help!

[yuxumin]

Which version is your cuda? Could you provide more details about your environment? BTW, can this discussion solve your problem?

[zjuruizhechen]

Thanks for your reply! It's pytorch1.8.0+cuda111 on RTX3090. Seems KNN_cuda doesn't work, too.

[yuxumin]

Ok, i see. Make sure that the cudatoolkit version of your pytorch matches the cuda version in your system (nvcc -V).

[zjuruizhechen]

It's: Copyright (c) 2005-2019 NVIDIA Corporation Built on Sun_Jul_28_19:07:16_PDT_2019 Cuda compilation tools, release 10.1, V10.1.243

[yuxumin]

Two solutions for your situation: 1) Update the cuda to 11.1 for your server. 2) Create a new virtual env and install the pytorch with cudatoolkit 10.1. We recommend 2) for convenience.

[zjuruizhechen]

Sorry that i still cannot figure out the version problem...

But it seems to be solved by installing this version: https://github.com/otaheri/chamfer_distance

Thanks a lot for your help!