This guide will help you set up your Compute Canada environment with VS Code for GPU programming and job allocation.

• SSH Extension in VS Code
• Jupyter Notebook extension in VS Code
• An account on Compute Canada with access to GPU resources

• Open VS Code.
• Click on the green bottom-left button (Remote - SSH).
• Enter your Compute Canada credentials ([email protected]).

• Once connected, open a terminal in VS Code.
• Create a new virtual environment (replace my_venv with your desired environment name): python -m venv my_venv
• Activate the environment: source my_venv/bin/activate
• Install necessary packages (like torch): pip install pytorch

• Run the following command: nvidia-smi.
  • If you get an error saying NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running., this means you are not on a GPU node.

• Run the following command to allocate GPU resources: salloc --time=1:00:00 --gres=gpu:1 --cpus-per-task=1 --account=def-<YourComputeCanadaAccount> --mem=8G.
  • Replace <YourComputeCanadaAccount> with your Compute Canada account name.
  • Adjust time, GPU, CPU, and memory allocation as per your needs.

• Once you have allocated resources and are on a GPU node, run the following command to start a Jupyter notebook server: jupyter notebook --no-browser --ip=0.0.0.0 --port=8888.

• In VS Code, open the Command Palette (F1 or Ctrl+Shift+P).
• Enter Jupyter: Create New Blank Notebook.
• Click on the "Select Kernel" button in the top-right of the notebook interface.
• Click on the "Add kernel" button at the bottom of the kernel selection list.
• You'll be asked for the URL of the Jupyter server. Enter the URL provided by the server when you started it.
• The new kernel should now appear in the kernel list. Select it to use it for your notebook.

• In a new Jupyter notebook cell, run the following Python code to check if GPU is available:

import torch
torch.cuda.is_available()

• If it returns True, it means GPU is available in your Jupyter notebook.