Welcome to Ray

Ray is an open-source unified framework for scaling AI and Python applications, providing the compute layer for parallel processing without distributed systems expertise.

Ray minimizes the complexity of running distributed individual workflows and end-to-end machine learning workflows. The framework brings together three layers of capabilities:

• Scalable libraries for common machine learning tasks such as data preprocessing, distributed training, hyperparameter tuning, reinforcement learning, and model serving.
• Pythonic distributed computing primitives for parallelizing and scaling Python applications.
• Integrations and utilities for deploying a Ray cluster with existing tools and infrastructure such as Kubernetes, AWS, GCP, and Azure.

Who Ray is for

Data scientists and ML practitioners

Scale jobs without infrastructure expertise. Parallelize and distribute ML workloads across multiple nodes and GPUs, and leverage native integrations with the broader ML ecosystem.

ML platform builders and engineers

Build a scalable, robust ML platform on top of Ray’s compute abstractions. A unified ML API simplifies onboarding and reduces friction between development and production.

Distributed systems engineers

Ray automatically handles orchestration, scheduling, fault tolerance, and autoscaling for distributed applications.

What you can build

Batch inference

Run inference on large datasets across CPUs and GPUs.

Model serving

Deploy models behind autoscaling HTTP and gRPC endpoints.

Distributed training

Train large models across many GPUs and nodes.

Hyperparameter tuning

Run thousands of parallel trials with state-of-the-art search algorithms.

Reinforcement learning

Build and scale RL workloads with RLlib.

ML platforms

Compose Ray libraries into a production ML platform.

The Ray framework

Ray’s unified compute framework consists of three layers:

1. Ray AI Libraries

   A Python, domain-specific set of open-source libraries that equip ML engineers, data scientists, and researchers with a scalable toolkit for ML applications.

2. Ray Core

   A general-purpose, distributed computing library that lets ML engineers and Python developers scale Python applications and accelerate ML workloads.

3. Ray Clusters

   A set of worker nodes connected to a common Ray head node. Clusters can be fixed size, or autoscale up and down based on the resources requested by applications running on the cluster.

Get started

Install Ray

Install Ray with pip and verify your setup.

Quickstart

Run your first Ray task and actor in five minutes.

Use cases

Explore the workloads Ray powers today.

Examples

Browse end-to-end examples across data, training, tuning, serving, and RL.

Need help? Join the community on Slack, ask questions on the discussion forum, or open an issue on GitHub.