The Simplest Path to Scaling Python
Develop on your laptop and then scale the same Python code elastically across hundreds of nodes or GPUs on any cloud—with no changes.
Learning Ray | eBook
Learning Ray is the first and only comprehensive book on Ray and its growing ecosystem.
Ray Core Tutorial | On Demand
New to Ray? Jump-start your learning with our "Getting Started" guide.
Ray Summit 2023 | On Demand
Watch all the keynotes and sessions from Ray Summit 2023 on demand.
Parallelize Python, with minimal code changes
Simple, flexible Python primitives
Ray translates existing Python concepts to the distributed setting, allowing any serial application to be easily parallelized with minimal code changes
Distributed libraries
Easily scale compute-heavy machine learning workloads like deep learning, model serving, and hyperparameter tuning with a strong ecosystem of distributed libraries.
Integrations
Scale existing workloads (for eg. Pytorch) on Ray with minimal effort by tapping into integrations.
Any AI/ML Workload
Ray is the AI compute engine for every AI workload and use case.
Parallel Processing
Ray is Python-native. Effortlessly scale from your computer to the cloud with one Python decorator. Plus, leverage and parallelize CPUs and GPUs in the same pipeline to increase utilization and decrease costs.
Building on top of Ray has allowed us to deliver a state-of-the-art low-code deep learning platform that lets our users focus on obtaining best-in-class machine learning models for their data, not distributed systems and infrastructure.
Travis Addair
CTO, Predibase and Maintainer, Horovod / Ludwig AI
Scalable machine learning libraries
Native Ray libraries — such as Ray Tune and Ray Serve — lower the effort to scale the most compute-intensive machine learning workloads, such as hyperparameter tuning, training deep learning models, and reinforcement learning. For example, get started with distributed hyperparameter tuning in just 10 lines of code.
Build and run distributed apps
Flawless distributed operations
Ray handles all aspects of distributed execution from scheduling and sequencing to scaling and fault tolerance.
Autoscaling
Ray dynamically provisions new nodes (or removes them) to handle variable workload needs.
Fault-tolerant
Ray gracefully handles machine failures to deliver uninterrupted execution.
My team at Afresh trains large time-series forecasters with a massive hyperparameter space. We googled Pytorch hyperparameter tuning, and found Ray Lightning. It took me 20 minutes to integrate into my code, and it worked beautifully. I was honestly shocked.
Philip Cerles
Senior Machine Learning Engineer
Scale on any cloud or infrastructure
Public cloud, private data centers, bare metal, Kubernetes cluster — Ray runs anywhere. Or choose Anyscale, and leave the infrastructure to us.
LLMs and GenAI
Large language models (LLMs) and Generative AI are rapidly changing industries, and compute demand at an astonishing pace. Ray provides a distributed compute framework for scaling these models, allowing developers to train and deploy models faster and more efficiently. With specialized libraries for data streaming, training, fine-tuning, hyperparameter tuning, and serving, Ray simplifies the process of developing and deploying large-scale AI models.
Trusted by leading AI and machine learning teams
From detection of geospatial anomalies to real-time recommendation, explore the stories of teams scaling machine learning on Ray.
Get started with Ray
From a dedicated Slack channels to in-person meetups, we have the resources you need to get started and be successful with Ray.
Documentation
Reference guides, tutorials, and examples to help you get started on and advance your Ray journey.
Discussion Forum
Join the forum to get technical help and share best practices and tips with the Ray community.
Sign up for product updates
By signing up, you agree to receive occasional emails from Anyscale. Your personal data will be processed in accordance with Anyscale’s Privacy Policy.