Ray - A System for High-performance, Distributed Python Applications

Overview

Explore Ray, an open-source framework for scaling Python applications from laptops to clusters, focusing on ML/AI performance challenges and its key features for distributed computing.

Syllabus

- Introduction to Ray -- Overview of Ray -- Key benefits of using Ray for distributed computing -- Comparison with other distributed computing frameworks - Setting Up Ray -- Installation and setup -- Running Ray locally on a laptop -- Configuring Ray for clusters - Core Concepts of Ray -- Ray’s architecture -- Tasks and Actors in Ray -- Managing Ray objects and object stores - Distributed Computing with Ray -- Scheduling and execution of tasks -- Combining tasks and actors -- Handling resource constraints and dependencies - Scaling Machine Learning Applications -- Using Ray with popular ML frameworks (TensorFlow, PyTorch) -- Hyperparameter tuning with Ray Tune -- Distributed data processing with Ray Datasets - Advanced Features of Ray -- Ray's fault-tolerance and failure recovery -- Monitoring and debugging Ray applications -- Ray Serve for scalable model serving - Real-world Applications and Use Cases -- Case studies of Ray in industry -- Best practices for deploying Ray in production - Ray Ecosystem and Tools -- Overview of Ray libraries (RLLib, Ray Tune, Ray Serve, etc.) -- Choosing the right Ray tool for your application - Hands-on Projects -- Building a distributed application with Ray -- Scaling a machine learning model using Ray -- Performance optimization and tuning with Ray - Conclusion and Future of Ray -- Emerging trends in distributed computing -- Future developments in Ray -- Resources for further learning - Final Q&A and Wrap-up Each topic will be accompanied by lectures, demonstrations, and practical exercises to provide a comprehensive understanding of Ray and its applications in high-performance distributed computing.

Taught by

Tags