ray-one

Local Development

Create a new environment:

Using micromamba virtual environments is recommended:

python3 -m venv venv
source venv/bin/activate

Install the dependencies for the project you're working on:

pip install -r {project-dir}/requirements.txt

Start you local ray cluster and the serve service:

pip install 'ray[default]'
ray start --head --port=6379

Install Ray Serve and initialise it

pip install 'ray[serve]'
serve start

Job submission

Set the RAY_USER, RAY_TOKEN, and RAY_URL environment variables to the values given to you by the Ray team.

export RAY_USER=your-username
export RAY_TOKEN=your-token
export RAY_URL=the-url

Each project is in its own directory and is submitted to ray by running the ray-job.py script in the root directory.

For example, runnign the translate-en-fr project:

./ray-submit.py submit translate-en-fr/translate-en-fr.py

The submission script will ensure the requirements are specified for Ray and submit the job.

Job status

You can check the status of your job by running the ray-job.py script with the status command.