Question Does ML-Agents have an API for interprocess communication?

Suppose my model to be trained is in Python, and I have a simulation in Unity, and I wish to use ML-Agents to train it. Will the command "mlagnets-learn" include the required training data from Unity to be sent to the model in Python automatically? Or do I need to do this some other way like using some socket connections etc?

 

If by "training data from Unity" you are referring to observations and rewards, then yes, Unity will automatically send the observations and rewards for each step and receive the actions through the Python API. All you need to do is run the command "mlagents-learn" and provide the path to the config file, along with specifying a training ID. For example:

Code (CSharp):

1. mlagents-learn /Users/username/GameProject/config/AgentPPO.yaml --run-id=Training1

Hope I answered your question.