Skip to content

Latest commit

 

History

History
50 lines (39 loc) · 2.28 KB

mlx-setup.md

File metadata and controls

50 lines (39 loc) · 2.28 KB
Raw

Deploy MLX on an existing Kubernetes cluster

Prerequisites

  • An existing Kubernetes cluster. Version 1.17+
  • The minimum recommended capacity requirement for MLX is 8 vCPUs and 16GB RAM
  • If you are using IBM Cloud, follow the appropriate instructions for standing up your Kubernetes cluster using IBM Cloud Public
  • If you are using OpenShift on IBM Cloud, please follow the instructions for standing up your IBM Cloud Red Hat OpenShift cluster
  • kustomize v3.2.0 is installed
    • Kustomize v3.2.0 quick install:
    OS=$(uname) 
curl -L https://github.com/kubernetes-sigs/kustomize/releases/download/v3.2.0/kustomize_3.2.0_${OS}_amd64 --output kustomize
chmod +x kustomize
mv kustomize /usr/local/bin

Deploy

To deploy the MLX single-user mode on an existing Kubernetes Cluster, clone the MLX manifests and deploy it with Kustomize.

This MLX deployment includes:

git clone https://github.com/machine-learning-exchange/manifests -b mlx-single-user
cd manifests
# run the below command two times if the CRDs take too long to provision.
while ! kustomize build example | kubectl apply -f -; do echo "Retrying to apply resources"; sleep 10; done

Then access the MLX web page on http://<cluster_node_ip>:30380/mlx/

This MLX deployment doesn't include or support:

  • KFServing for model deployment
  • Multi-user mode
  • Istio mutual TLS

To get these features, please install the additional plugins by following the instructions for MLX deployment an existing Kubeflow cluster.

Delete the MLX deployment

To delete this MLX deployment, run the following commands in the same manifests folder.

kustomize build example | kubectl delete -f -