Launch NGC Notebooks on Charmed Kubeflow

[nohaihab]

DISCLAIMER: This feature is currently only in latest/edge, and will be available in stable in the Charmed Kubeflow 1.9 release

With Charmed Kubeflow, you can create Jupyter Notebook servers with Nvidia NGC images by following this guide.

Contents:

• Deploy the dependencies
• Create a Notebook with NGC

Prerequisites

An active Charmed Kubeflow deployment. For installation instructions, follow the Get started tutorial.

Deploy the dependencies

Deploy the charms required for Notebooks integration with NGC images.

juju deploy ngc-integrator --channel=latest/edge --trust

juju deploy resource-dispatcher --channel=latest/edge --trust

Add the required relation between the deployed charms

juju relate ngc-integrator:pod-defaults resource-dispatcher:pod-defaults

Wait until the charms are in active status, you can watch the status with:

juju status --watch 5s

Create a Notebook with NGC

From the Notebooks UI:

1. select new Notebook to go to the Notebook creation page

2. Click on Custom Notebook then choose Advanced Options.

3. Select Custom Image and type in the full name of the NGC image. In the example below, the image is nvcr.io/nvidia/pytorch:23.09-py3.

   

4. Scroll down to the bottom of the Notebook creation page and click on Advanced Options.

5. From the Configurations dropdown, select Enable Nvidia NGC JupyterLab Notebook as shown below

   

Finally, launch the Notebook server and connect to it. Now you have a Jupyter Notebook server with an NGC image!