OSError: CUDA_HOME environment variable is not set during bash compilation

[Arka161]

I am following the Install.md file as given, and I install the CudaToolKit along with PyTorch (same line of code):

conda install pytorch=1.3.1 torchvision cudatoolkit=10.0 -c pytorch -y

However, when I try compiling the bash script after cloning the repo, I get the error:

bash compile.sh

This gives:

OSError: CUDA_HOME environment variable is not set

I am in a Conda environment called Redet, and these steps pretty much reproduce the same error in all my machines. Is there anything wrong with the install steps?

I did try to set CUDA_HOME manually, but it would not work with the torch_cpp APIs. Any solution?

Thanks for all your great work.

[csuhan]

https://github.com/facebookresearch/votenet/issues/2

Hope it works for you.

[Arka161]

@csuhan Sorry to come again, I spent several days on this issue, I am not able to run the compile.sh successfully. Is there any version issue/conflict with Conda? I have tried this on both Linux on Windows, and following the instructions 1:1 always returns an error.

[Arka161]

This is the error I get even if I fix the CUDA_HOME issue.