ec2

Setup

Clone this repo to run the examples on your local laptop:

git clone https://github.com/aws-samples/aws-samples-for-ray

cd aws-samples-for-ray/

Start your Ray cluster from your local laptop (substitute cluster-gpu.yaml everywhere to use GPUs):

ray up cluster.yaml

Install JupyterLab and MLflow

Attach to the head node of the Ray cluster

ray attach cluster.yaml

Install Jupyter Lab on the head node of the Ray cluster:

pip install jupyterlab

Run JupyterLab on the head node of the Ray cluster

From your local laptop, Attach to the head node of the Ray cluster

ray attach cluster.yaml

Run JupyterLab on the head node of the Ray cluster

nohup jupyter lab > jupyterlab.out &

Tunnel ports from local laptop to the head node of the Ray cluster

From your local laptop, tunnel port 8888 to the Ray cluster:

ray attach cluster.yaml -p 8888

From your local laptop, start the dashboard and tunnel port 8265 to the Ray cluster:

ray dashboard cluster.yaml # This implicitly tunnels port 8265

Navigate to the JupyterLab and MLflow UIs

From your local laptop, run this command to get the JupyterLab url (and ?token=)

ray exec cluster.yaml "jupyter server list"

Navigate your browser to the URL from above to start using JupyterLab:

http://127.0.0.1:8888?token=...