This code accompanies the paper "Models as Agents: Optimizing Multi-Step Predictions of Interactive Local Models in Model-Based Multi-Agent Reinforcement Learning".
The repository contains MAG implementation as well as fine-tuned hyperparameters in configs/dreamer/optimal folder.
python3 train.py --n_workers 2 --starcraft
The optimal parameters are contained in configs/dreamer/optimal/ folder.
The code for the environment can be found at https://github.com/oxwhirl/smac
agentcontains implementation of MAGcontrollerscontains logic for inferencelearnerscontains logic for learning the agentmemorycontains buffer implementationmodelscontains architecture of MAGoptimcontains logic for optimizing loss functionsrunnerscontains logic for running multiple workersutilscontains helper functionsworkerscontains logic for interacting with environment
envcontains environment logicnetworkscontains neural network architectures