Code for PTM
This codebase is built on top of the PyMARL framework and REFIL for multi-agent reinforcement learning algorithms.
- Docker
- NVIDIA-Docker (if you want to use GPUs)
Build the Dockerfile using
cd docker
./build.sh
Set up StarCraft II.
./install_sc2.sh
cd src
python main.py --env-config=sc2custom --config=mtmarl_exp with scenario=<ENV>
Possible ENVs are:
- case_study
- 6-11m_symmetric
- sym_asym