JupyterLab in Open OnDemand

Run Python notebooks on a Phoebe compute node from your browser. First log in to Open OnDemand.

How Python is provided

Python itself comes from Lmod modules. The packages you need are installed in your own Python virtual environment (venv), which also contains JupyterLab. The portal creates it for you in ~/.venvs/.

Start a session

On the portal, select VENV-based JupyterLab and fill in the form:

• Slurm partition: cpu, or gpu if you need GPU acceleration, or small for light work. The session runs in the matching interactive partition: cpu_int, gpu_int or small_int.
• Preloaded moduleset: keep empty.
• Python version: the version you need, typically the latest.
• Session duration: how long the session may run, from 8 hours to 14 days.
• Instance size: the number of CPU cores, from 8 to 128.
• GPU count: leave at 0 unless you need GPUs (up to 4).
• Venv name to be created in ~/.venvs/: a name for your new virtual environment (default venv_default).
• Existing venv path to use: keep empty; this option is not ready yet.

Click Launch to submit the session as a job to Slurm.

Wait for the job to start

The session is first shown as Queued. The first start takes longer, because the portal creates your virtual environment.

Connect to the session

Once the session is Running, click Connect to Jupyter. JupyterLab opens in a new browser tab.

To come back to the session later or end it, see reconnect and end a session.