Question For a trained model, how can I know the reward after each step

At present, I have trained the appropriate model through ML-Agent.

when verifying the model, I hope to know the reward or intrinsic reward of each decision or each step made by the Agent through this model.

So how can I get the kind of data? For example, I can print after each step, using the Delog command?

 

Have you taken a look at: https://github.com/Unity-Technologi...n-API.md#interacting-with-a-unity-environment

You get decision steps per "behavior" at every step and you can grab the reward per agent there I think.