Skip to content

Hosted Ray Clusters (Beta)

Ray is an open-source framework for scaling Python workloads — from data processing and machine learning training to model serving and reinforcement learning. Heata provides shared Ray clusters running on our distributed, low-carbon infrastructure.

Submit jobs to a fully managed Ray cluster without provisioning or managing any infrastructure. Your code runs on dedicated compute while the heat generated provides free hot water to UK households.

Getting started

1. Get your access token

Sign up and generate a token in the Heata portal:

https://www.heata.co/ray-sign-up

2. Configure your environment

Set the environment variables that tell Ray where your cluster lives:

# Head-node URL of your shared Ray cluster
export RAY_ADDRESS='https://ray.heata.co/jobs'

# Authorization with your Heata access token
export RAY_JOB_HEADERS='{"Authorization": "Bearer YOUR_TOKEN_HERE"}'

The Ray CLI picks these up automatically each time you run it. If you prefer, you can pass them directly using --address and --headers instead.

3. Submit a job

ray job submit -- python my_script.py

That's it. Your job runs on Heata's Ray cluster and results are returned to your terminal.

Full documentation

Once you've signed up, the Heata portal has full instructions covering:

  • Job submission and monitoring
  • Working directory and dependency management
  • Resource requests and GPU access
  • Viewing logs and job history

Contact

For questions, custom cluster configurations, or dedicated capacity: