Ray
Distributed execution framework for scaling Python workloads.
Overview
Ray turns ordinary Python functions and classes into distributed tasks and actors with a scheduler that handles placement, failure and state. It solves the problem that scaling a training or inference job across machines otherwise means rewriting it around a cluster API. It exists as the substrate several training and serving systems are built on, which is why it appears beneath them rather than beside them.
Technology Graph
How Ray connects to the rest of the stack.
Applied to
- Inference concept Infrastructure Ray used for Inference High ✓ source
- Pretraining technique Techniques Ray used for Pretraining High ✓ source
Tools
Software that implements or integrates with Ray.