['TORCH_CUDA_ARCH_LIST'].

[Yhc-777]

Description

when i run :"python main.py -O --image ./df_ep0100_0008_rgb_rgba.png --workspace trial_image --iters 5000 --default_polar 80", i met this problem:“/home/u/miniconda3/envs/sdf/lib/python3.8/site-packages/torch/utils/cpp_extension.py:1967: UserWarning: TORCH_CUDA_ARCH_LIST is not set, all archs for visible cards are included for compilation.
If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST'].” How can I solve it? Thx!!

Steps to Reproduce

Expected Behavior

Environment

[Duodecimus]

The environment variable TORCH_CUDA_ARCH_LIST in PyTorch should be set to specify the CUDA architectures that you want PyTorch to build for when compiling CUDA kernels. Each architecture is specified by its CUDA Compute Capability version (such as 3.5, 5.2, 6.0, etc.).

You probably only want to list the value of the one graphics card you have.
This is the same value as used for TCNN_CUDA_ARCHITECTURES or CUDA_ARCH

ypu can look your card up on https://developer.nvidia.com/cuda-gpus

[Hackinside]

Locate the cpp_extension.py on the line 1963 and replace the TORCH_CUDA_ARCH_LIST with one of the following strings:

'Kepler+Tesla', '3.7'
'Kepler', '3.5+PTX'
'Maxwell+Tegra', '5.3'
'Maxwell', '5.0;5.2+PTX'
'Pascal', '6.0;6.1+PTX'
'Volta+Tegra', '7.2'
'Volta', '7.0+PTX'
'Turing', '7.5+PTX'
'Ampere+Tegra', '8.7'
'Ampere', '8.0;8.6+PTX'
'Ada', '8.9+PTX'
'Hopper', '9.0+PTX'

See the bellow picture as example: