install-mlx-on-kubeflow.md

Deploy MLX on an Existing Kubeflow Cluster

This is a complete deployment of MLX which includes all of the following components:

• Istio (with mutual TLS)
• Kubeflow Pipelines, in Multi-User mode
• Tekton Pipelines
• Datashim to provide access to S3 and NFS Datasets within pods
• KFServing for model deployment
• MLX API and UI

To deploy MLX on a Kubernetes cluster which has Kubeflow 1.3.0 already installed, we use the kustomize plugin that comes with kubectl client v1.17+. Clone the MLX repo and run the following commands based on your Kubeflow setup.

git clone https://github.com/machine-learning-exchange/mlx
cd mlx

• Deploy MLX on Kubeflow (With OIDC and Istio Mutual Auth)

kubectl apply -k manifests/istio-auth

Then access the MLX page using http://<Kubeflow_Endpoint>/mlx/

Replacing Kubeflow Central Dashboard with MLX Dashboard

To deploy MLX on Kubeflow (With OIDC and Istio Mutual Auth) and replace Kubeflow Central Dashboard with MLX Dashboard, run the following commands:

git clone https://github.com/machine-learning-exchange/mlx
cd mlx

• Deploy MLX on Kubeflow (With OIDC and Istio Mutual Auth)

kubectl apply -k manifests/prod-multi-user

Then access the MLX page using http://<Kubeflow_Endpoint>/