PyTorch is a popular open-source machine learning framework for Python. You can use anaconda to install PyTorch in a virtual environment. First, SSH into a machine with GPU capabilities. Check the CUDA version using the following command, as this will determine how you install PyTorch:
$ nvidia-smi
Shared CS department machines with GPUs use CUDA version 12.2. To install PyTorch in a virtual anaconda environment on a department machine, first activate an anaconda environment. Once active, use the command:
$ conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia
It may take a few minutes, but once the installation is complete PyTorch will be available for use in the anaconda virtual environment. More information and PyTorch tutorials can be found at this link: PyTorch Tutorials