Learn how to build large-scale AI applications using Ray, a high-performance distributed execution framework from the RISELab at UC Berkeley. Simplify complex parallel systems with this easy-to-use Python* framework that comes with machine learning libraries to speed up AI applications.

This course provides you with practical knowledge of the following skills:

• Use remote functions, actors, and more with the Ray framework

• Quickly find the optimal variables for AI training with Ray Tune

• Distribute reinforcement learning algorithms across a cluster with Ray RLlib

• Deploy AI applications on large computer clusters and cloud resources

The course is structured around eight weeks of lectures and exercises. Each week requires approximately two hours to complete.

Ray framework official repository.

Course material compiled from the Ray tutorial repository.

Get an introduction to the Ray framework and data parallelism. Topics include how to:

• Run tasks in parallel using remote functions

• Make dependencies between remote tasks through object IDs

• Create nested tasks and remote functions within remote functions

Learn about Ray actors, which are remote functions that have states. Additional topics:

• How to implement Ray actors using Python* classes

• How to use different hardware resources for various AI tasks, such as training and inference

• The analytics ecosystem, which is made up of toolkits, libraries, solutions, and hardware

Understand how to optimize and speed up functions. Topics include how to:

• Avoid waiting for slow tasks using ray.wait()

• Process remote tasks in a specific order

• Speed up serialization by passing objects to ray.put()

Explore how to optimize functions, including:

• Accelerate Pandas* workflows by changing one line of code

• Implement a MapReduce system with Ray

• Use Tree Reduce to execute a tree of dependent function in parallel

Learn to access different hardware resources, including how to:

• Send remote tasks to different accelerators and processors

• Use custom resources for tasks that require complex hardware combinations

Get an introduction to training neural networks across multiple workers. Topics include:

• An example of how to pass the weights of a TensorFlow* model between workers and drivers

• How to implement a sharded parameter server for distributing parameters across multiple workers

Understand how to use Ray Tune, a scalable framework for searching for hyperparameters.

• Use Tune to reduce one of the most expensive parts of machine learning

• Search for the right parameters, such as learning rate and momentum, to train a neural network

• Combine HyperOpt and HyperBand to perform a more powerful search

Learn about RLlib, which is a scalable reinforcement learning library to train AI agents.

• Get an introduction to the Markov Decision Process and how to use it in Python*

• See an example of how to use the PPO algorithm to train a network to play a simple game with Gym* and visualize the results with TensorBoard*

• Learn to create a deep-Q network (DQN) to play Pong and play against it in a browser