The MLBatch setup consists of a cluster setup to be done once and a team setup to be repeated for each team that will be using the cluster.
Batch users should only be permitted to create AppWrappers or other
workload Kinds that are natively supported by Kueue. The cluster setup
defines a mlbatch-edit
role which enforces these restrictions and
will be used in the setup process for each team of MLBatch users that
is onboarded.
This setup has been developed on Red Hat OpenShift 4.14 and Kubernetes 1.27 and is intended to support Red Hat OpenShift 4.14 and up and/or Kubernetes 1.27 and up.
To start with, recursively clone and enter this repository:
git clone --recursive https://github.com/project-codeflare/mlbatch.git
cd mlbatch
Detailed instructions and configuration files can be found in subfolders, one for each base platform.
We recommend using the most recent stable release of Red Hat OpenShift AI as the base platform for MLBatch. Please see Red Hat OpenShift AI Self-Managed Life Cycle for the life cycle dates of currently supported stable and fast releases.
Instructions are provided for the following Red Hat OpenShift AI stable releases:
- Red Hat OpenShift AI 2.13
- Red Hat OpenShift AI 2.10
Instructions are provided for the following Red Hat OpenShift AI fast releases:
- Red Hat OpenShift AI 2.14
- Red Hat OpenShift AI 2.11
On Kubernetes version 1.30 and later, an enhanced user experience is available by using ValidatingAdmissionPolicies to streamline quota enforcement. Follow these instructions when installing on 1.30+ clusters:
MLBatch can be installed on any Kubernetes cluster version 1.27 or later by following these instructions: