Is multi agent learning possible by simply adding agent with "gym.create_actor"?

[ameliacode]

Hi
I somehow manage to build humanoid task related environment, and it worked pretty well.
Based on my project, I am trying to implementing multi agent rl to it.
It worked well when I just simply add simple objects with gym.create_actor

However, when it comes to adding humanoid, it seems that I have to modify some lines from scratch I believe..
If I am right, what things do I have to modify?
I get the fact that I do have to modify obs, reward, action space
But I don't get the rest of it..

Plz help..:(