September 2020 Summaries

Ray 1.0 has been released with a stable API and new serverless features, making it easier to build scalable libraries and applications. The update brings significant improvements in autoscaling, resource management, and community library integrations. Ray 1.0 now supports Java and Windows platforms, and its automated memory management eliminates "object evicted" errors. The release is the result of a large community effort with contributions from over 67 contributors and 458 commits. The update also introduces new serverless features, including production serving and detached actor lifetimes, and is backed by Anyscale's open source development model.

Ray Summit 2020 is an upcoming event that offers live and on-demand streams of talks and tutorials on distributed computing, reinforcement learning, and more. Anyscale is hosting a full-day tutorial the day before the summit, covering core concepts of Ray and reinforcement learning. The tutorial includes certification quizzes and covers topics such as ray tasks, actors, multiprocessing, introduction to reinforcement learning, optimizing market investments with multi-armed bandits, and keystone lesson: RL for recommender systems. Registration is available with a nominal fee, while the main event is free to attend.