Skip to content

Python

Using Python virtual environments with uv

For Python specifically, you can set up a virtual environment, which acts similarly to a personal module but for using Python packages. In these virtual environments, you can install whichever Python packages you want, as well as manage different python versions. We prefer you to use uv for your virtual environments over venv, as its easier on storage limits and backup requirements.

$ module load python # (1)!
$ module load uv
$ uv init <project directory> # (2)!
$ uv add <my-python-package> # (3)!
$ uv run <script>.py # (4)!
  1. This will load the cluster default python version. If you need a specific version, you can instead use module load python/3.12.4.
  2. This creates a new project managed by uv. If you have a prexisting project directory, this command will let uv track it.
  3. If a virtual environment doesn't already exist, uv will create one named .venv and install packages here! (If you have a venv already, check uv's documentation for letting uv recognize it)
  4. With uv, you don't need to activate a virtual environment if it is named .venv, you can simply run your code!

Using Python virtual environments with conda

A similar process can be used with miniconda to create and activate a conda environment. Again, we prefer you to use uv, but there are some cases where you must use a conda environment.

$ module load miniconda3
$ conda create --name <my-env>
$ conda init bash # (1)!
$ source activate <my-env> # (2)!
$ conda install <my-python-package>
  1. Note that this modifies your .bashrc so that every time you log in conda will be sourced as your python distribution.
  2. run conda deactivate to exit your conda environment.

For more information, you can look at the conda env documentation

Jupyter Notebooks vs. .py Files

We recommend using Jupyter notebooks for exploration and data visualization. Notebooks provide an interactive environment that’s ideal for experimenting and analyzing data. For larger scripts or batch jobs, .py files are often a better fit, as they are easier to run on compute nodes in non-interactive sessions.

Running Jupyter Notebooks on a Compute Node

If you'd like to run a Jupyter notebook on a compute node, you can do so using the sjupyter command. This requires an active virtual environment that includes Jupyter:

$ sjupyter

Running Jupyter via Open OnDemand

If you are unfamiliar with the terminal and just want to use Turing resources with Jupyter, you can use the interactive app provided by Open Ondemand.

Other Jupyter Kernals

jupyter also supports matlab and maple kernels.

Matlab - install instructions

Maple - install instructions

R - install instructions Note, this also requires an older version of jupyter, and npm. So before following these instructions run 

$ module load npm uv
and create a virtual envionment with jupyter 3 uv add jupyterlab==3.*

This allows you to access compute resources within a notebook environment for more intensive interactive work.