This repo provides documentation and tools to run the Red Hat supplied container images for data science on Red Hat Linux workstations. The data science images are intended and built for use in Open Data Hub and OpenShift AI which are data science platforms built on Kubernetes. So these assets repurpose the notebook images for running on a local workstation.
The odh-workbench.sh script will run the workbench notebook as a standalone jupyter-lab server. Once this is running, you can grab the url and
connect to the server via a web browser OR you can plug that url into a vscode jupyter kernel. This script is setup to mount a Notebooks subdirectory
for persistent storage. This Notebooks directory should be set to chmod 775rights in order to allow the container's random user id to write into it.
Nvidia GPUs are configured in this launch script.
There is a .devcontainer.json configuration to run a workbench image.
Limitations: this only has ephemeral storage in the container. NVidia is not configured.
To use it, open this directory in vscode and allow the editor to open the devcontainer. Then open a shell and issue these commands:
- Open a shell in vscode
jupyterlab listto get the url and access token of the jupyterlab server. Copy this url.- From the command palette run the
Jupyter: create interactive window - Select a kernel
- Choose the "Existing Jupyter Server" option
- Paste the jupyter-lab URL here
- Voila: you have a running jupyter-lab server, built from the Red Hat ODH workbenches, running in VSCode.
- While the storage is ephemeral, git is configured so that assets can be pushed/pulled from a repository.
- Go to the Open Data Hub Workbench Images and find the image you want to work with.
- Update the .devcontainer.json file or the odh-workbench.sh file to pull from this image
- Save and reopen the directory in the devcontainer