Red Hat OpenShift Artificial Intelligence - Accelerator Boot Camp

Welcome to the Red Hat OpenShift Artificial Intelligence (RHOAI) accelerator boot camp!

This module is a short, intensive and rigorous course of training covering the following topics:

  • Rapidly provisioning RHOAI onto your new OpenShift cluster

  • Working with Kustomize to tailor your GitOps based installation

  • Creating and deploying custom data science container images

Once the RHOAI environment is up and running on your cluster, we will dive deeper into some common use cases:

  • Training a ML model using a customized development container image

  • Using Amazon S3 based storage to save the newly trained model

  • Serving the model using a custom inference engine (runtime)

  • Using data science pipelines

  • Distributed model training, utilizing multiple compute nodes

  • Using Large Language Models (LLMs) in RHOAI

At completion we hope that you’ll have a greater understanding of the features in RHOAI, and the ability to rapidly implement your own AI/ML based project.

What is the RHOAI Accelerator Project?

Building an OpenShift cluster to support containerized AL/ML workloads can be quite complex. There are a number of additional operators, features and configuration that must be provisioned in order to expose features such as GPU compute enabled nodes, monitoring, ML training and deployment lifecycle and so on.

It’s possible to install all the various components by hand, making notes such as an installation "cheat sheet" while doing so. However this process will be quite tedious and take a fair amount of time, which is a problem for a number of reasons:

Repetition: For short lived clusters (e.g. demo.redhat.com) the cluster has a very short lifespan, measured in days. It’s important that we have a way to rapidly re-provision a new cluster through the use of automation.

Customizable Framework: The intent of the Accelerator project is to be cloned, copied and altered to match your exact requirements. This could be a customized cluster for experimentation on specific components, features or architecture. Or it could be a great starting point for building a GitOps cluster design for real world customer facing implementations.

RHOAI Accelerator - Open Source

The source code for the AI Accelerator can be found at: https://github.com/redhat-ai-services/ai-accelerator

The accelerator was created and is currently maintained by the Red Hat AI Services organization, however contributions from anywhere are always greatly appreciated!

Questions for Further Consideration

Additional questions that could be discussed for this topic:

  1. When would I want to manually install RHOAI and associated components, instead of using the AI Accelerator framework?

  2. Who maintains the AI Accelerator framework project?

  3. What if I find a bug, or have a new feature to contribute?

  4. Can I add my own additional components into the AI Accelerator?