arena

Run:AI

Overview

Run:AI CLI is a command-line interface for the data scientists to run and monitor machine learning jobs on top of the Run:AI software and Kubernestes.

Prerequisites

• Kubernetes 1.15+
• Kubectl installed and configured to access your cluster. Please refer to https://kubernetes.io/docs/tasks/tools/install-kubectl/
• Install Helm. See https://v2.helm.sh/docs/using_helm/#quickstart . Run:AI currently supports Helm 2 only. Helm 3 (the default) is not supported
• Run:AI software installed on your Kubernetes cluster. Please refer to https://support.run.ai/hc/en-us/articles/360010280179-Installing-Run-AI-on-an-on-premise-Kubernetes-Cluster for installation, if you haven't done so already.

Setup

• Download latest release from the releases page. https://github.com/run-ai/arena/releases.
• Unarchive the downloaded file.
• Install by running:

sudo ./install-runai.sh

To verify installation:

runai --help

Quickstart

For help on Run:AI CLI run

runai --help

To verify the status of your cluster, use the top command.

runai top job
runai top node

These commands will give you valuable information about your cluster's and jobs' GPUs allocation status.

To run a sample job using runai sample training container please run:

runai submit -g 1 --name runai-test --project {your_project} -i gcr.io/run-ai-lab/quickstart -g 1

This will run a job using the Run:AI scheduler on top of the kubernetes cluster using Run:AI quickstart image and requesting 1 GPU for the job. To see the status of the running job please run:

runai list

Once the job in running, you can view its logs by running:

runai logs runai-test

At last, to delete the job prior to its completion you can run:

runai delete runai-test

Updating the Run:AI CLI

To update the CLI to the latest version run:

sudo runai update