This code is implemented based on EPyMARL, and the running instructions are similar to that in the original project. We incorporate Hallway, LBF and Traffic Junction in this code, which you can experiment with config instructions like env-config=hallway/lbf/traffic_junction. Besides, we incorporte an implementation of Full-Comm which you can try by setting config=full_comm in your instruction.

For example, you can run MASIA on traffic_junction (medium) by using:

python3 src/main.py --config=masia --env-config=traffic_junction

and you can run Full-Comm on LBF (11x11-6p-4f-1s) by using:

python3 src/main.py --config=full_comm --env-config=lbf with env_args.sight=1 env_args.players=6 env_args.field_size=11 env_args.max_food=4 env_args.force_coop=False

This code will use tensorboard and save model by default, which will be saved in ./results/