claification for using the gym wrapper for unity

I am looking into using Unity with the gym wrapper, and I was hoping someone could clarify the limitation of agents in the wrapper.

when it says that it can only use 1 agent, does that mean that it can only use 1 instance of an agent, or does that mean that it can only have 1 type of agent (but can still have many instances of that agent to collect observations).

thanks

 

As far as I know it is a limitation with regard to single- vs multi-agent scenarios, but I'm not sure if you're referring to ML-agents documentation or some info from OpenAI?

 

The gym interface only allows one type of agent (enforced here: https://github.com/Unity-Technologi.../gym-unity/gym_unity/envs/__init__.py#L67-L70) and one agent active at a time (enforced here: https://github.com/Unity-Technologi...e_3/gym-unity/gym_unity/envs/__init__.py#L101)