Multiple asynchronous agent training

The A3C algorithm we just looked at was developed by Google DeepMind as a way of training multiple asynchronous agents simultaneously into a global overseer network. Now the Hallway example is already set up for multiple asynchronous training, and we can turn it on relatively quickly. Open up Unity and go through the following exercise to enable multiple agent training:

  1. Open the Hallway example scene.
  1. Locate and select the Hallway(1) to Hallway(15) objects, as shown in the following screenshot:
Selecting the disabled Hallway objects
  1. Enable all these objects by checking the Enable box in the Inspector ...

Get Learn Unity ML-Agents - Fundamentals of Unity Machine Learning now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.