Similar to this feature request (Accessing activations from the agent's brain? · Issue #3712 · Unity-Technologies/ml-agents · GitHub), I would like to be able to access and save the current activation levels of the networks that I train. So far my search for solutions hasn’t been successful.
The main reason why I want to do this is because I want to compare neural representations in humans with representation in artificial agents while completing the same task.
Any nudge into the right direction is welcome!