Hi folks!
Does someone know if there is an BicNet / CommNet or another multi agent algorithm implementation in RLlib where agents are direct communicate with each other and don’t have a central shared Critic, so that the number of agents is scalable?