Training an agent to learn a Unity environment is not unlike much of the training we have already done. There are a few slight changes to the way we interact and set up the environment but overall it is much the same, which makes it a further plus for us because now we can go back and train several different agents/algorithms on completely new environments that we can even design. Furthermore, we now can use other DRL frameworks to train agents with Python— outside the ML-Agents agents, that is. We will cover more on using other frameworks in Chapter 12, DRL Frameworks.
In the next exercise, we see how to convert one of our latest and most state-of-the-art samples, Chapter_10_Rainbow.py, and turn it into Chapter_11_Unity_Rainbow.py. Open Chapter_11_Unity_Rainbow.py and follow the next exercise:
- We first need to copy the output...