Skip to content

Launch Jupyter

  1. Navigate to the Central GPU Cluster OpenOnDemand website
  2. Click on "Interactive Apps" in the header
  3. Click "Jupyter Notebook" in the dropdown

interactive apps juypyter notebook location

Jupyter Options

On the juptyer page, you can set the following options:

Use JupyterLab instead of Jupter Notebook? (checkbox)

Checking this will laucn Jupyter Lab instead of Jupyter Notebook. Jupyter Lab is the successor to Jupyter Notebook, and very similar. Recommended to use Jupyter Lab unless you have a reason not to.

Queue Name

Choose the partition you want to run the Jupyter server on. "General: cpu only" does not have GPU access and so is not useful for GPU jobs. The queue options have GPU access (h100 XL, a100 XL, L40s XL).

Number of hours

Enter the number of hours to allocate your job for. This time starts once your job is finished queueing and starts. If the time runs out, your job will be killed. Note, the more time you request, the longer your job may take to queue.

Number of cores per task

Number of CPU cores to allocate to your task. The total number of CPU cores allocated will be cores per task * tasks per node

Number of tasks per node

Number of tasks to allocate to your node.

Number of gpus per task

Number of GPUs to allocate per task. If you are using a GPU partition, set this field to at least one, so you have access to a GPU.

Version of the Jupyter stack container

Select a pre-defined stack for Juptyer lab. This includes common packages required for the purpose. E.g. if you want to run machine learning with pytorch, select the "pytorch" version.

Once you have filled out the form, click "Launch" to submit the job. Allocating resources for the job will likely take a few minutes. When the job is ready, click Connect to Jupyter from the sessions page

Use a custom conda environment

If you already have a conda environment, you are able to run Jupyter Python code from within that environment.

For example, if you want to run a custom pytorch stack, you can use this feature.

  1. Click on the "Kernel" tab on the top bar
  2. Click "Change Kernel" from the dropdown
  3. Select the kernel corresponding to your preferred conda environment

NOTE: If you do not see the environment listed, make sure you have run python3 -m ipykernel install --user --name YOUR_NAME_HERE from a login shell where the environment has been activated. See more info on the using-conda page