Official whl file (torch=1.0.1 with CUDA 10.0) ships with wrong torch.version.cuda

1

I recently downgraded from PyTorch 1.1 to 1.0.1 to get an old project running.

Specs/Environment:
Ubuntu 18.04
GPU: Nvidia 2080
CUDA: V10.0.130 (from nvcc --version)

After downloading/installing PyTorch 1.0.1 from https://pytorch.org/get-started/previous-versions/ via

pip install torch==1.0.1 -f https://download.pytorch.org/whl/cu100/stable

and running torch.version.cuda, It returns 9.0.176.

I believe this is the reason why I’m getting CUDA errors in my project. How can I get torch.version.cuda to return the correct CUDA version (10.0)?

Thank you in advance :slight_smile:

2

I have solved the problem by downloading/installing a different .whl file via pip install https://download.pytorch.org/whl/cu100/torch-1.0.1-cp36-cp36m-linux_x86_64.whl.

This comment from the official PyTorch Github has led me to the solution.