I’ve created a new learning environment. The tutorial is great, but for me, most interesting part was to train the agent myself.
Unfortunately Testing the Environment explains how to test it in Unity play mode using Player Brain. I want more. So I’ve tried to train External Brain for my own. And I’ve made it!
I renamed Brain in Hierarchy to RollerBrain
Then I copied Unity-Technologies/ml-agents \python\trainer_config.yaml SmallWallBrain and create RollerBrain with parameters
Here, you can get working trainer_config.yaml file.
And voilà, as you can see in the video, the agent works perfectly.
Now, it’s time to understand more deeply how is this all working! ;p