Awesome

Transformer-based Working Memory for Multiagent Reinforcement Learning with Action Parsing (NeurIPS 2022)

1. To switch between ATM and RNN agents, please modify src/config/default.yaml to enable RNN Agent parameters or Memory Agent parameters.

2. To run experiments on SMAC, use command like "python src/main.py --config=qmix --env-config=sc2 with env_args.map_name=4m_vs_5m env_args.seed=1 > out_4m_vs_5m_atm_qmix_1.log 2>&1 &".

2.1 Map 4m_vs_5m in is "atm-smac/src/envs/4m_vs_5m.SC2Map" and should be copied to "atm-smac/3rdparty/StarCraftII/Maps/SMAC_Maps".

2.2 You can modify "smac/env/starcraft2/maps/smac_maps.py" to add 4m_vs_5m.

2.3 StarCraft II version is 4.6.2.

3. To run experiments on LBF, use command like "python src/main.py --config=maa2c --env-config=gymma with env_args.key="lbforaging:Foraging-7s-15x15-3p-5f-v2" t_max=5000000 > out_lbf_15x15_3p5f_atm_maa2c_1.log 2>&1 &".

4. Note that, on LBF, the parameter "num_foods" in src/config/default.yaml should correspond with each scenario. For example, "5f" means "num_foods" is 5.