Unity has released a new update to the Unity ML-Agents Toolkit that allows developers to train faster by launching multiple Unity simulations running on a single machine. The release also includes support for custom protocol buffer messages, render texture observations, 2D ray casting, and reorganizing the toolkit into multiple Python packages.
The blog post announcement includes a review of how Unity ML-Agents Toolkit was used by developer JamCity to train agents to play advanced levels of Snoopy Pop on a single machine 7.5 times faster than previously possible. Click here to read more.
