Skip to content

Latest commit

 

History

History
27 lines (17 loc) · 1.46 KB

cuda_guide.md

File metadata and controls

27 lines (17 loc) · 1.46 KB
Raw

  1. Uncomment the "runArgs" section in .devcontainer/devcontainer.json.

  2. Uncomment the nvidia channel and pytorch-cuda package lines in env.yml.

  3. Install the Nvidia Container Toolkit and configure Docker by following the instructions in this guide. Focus on the sections "Installing with Apt" and "Configuring Docker", and do not follow the "Rootless mode" section.

    ⚠️ Windows Users: First, verify that you can run nvidia-smi in the Ubuntu terminal. Next, install and configure the Nvidia Container Toolkit in the Ubuntu terminal.

    ⚠️ Important: When following the "Configuring Docker" step, use sudo service docker restart instead of sudo systemctl restart docker.

  4. After installing the devcontainer, verify GPU access by running the following command in the Visual Studio Code (devcontainer) terminal:

    nvidia-smi

    You should see GPU statistics displayed.

  5. Verify that PyTorch CUDA runtime is installed and configured correctly by running the following code in a Python console:

    import torch
print(torch.cuda.is_available())  # This should print 'True' if CUDA support is enabled

    If the output is True, it means that PyTorch is correctly set up to use CUDA. If it prints False, CUDA support may not be properly configured.