PTM

Code for PTM

This codebase is built on top of the PyMARL framework and REFIL for multi-agent reinforcement learning algorithms.

Dependencies

• Docker
• NVIDIA-Docker (if you want to use GPUs)

Setup instructions

Build the Dockerfile using

cd docker
./build.sh

Set up StarCraft II.

./install_sc2.sh

Run an experiment in SMAC

cd src
python main.py --env-config=sc2custom --config=mtmarl_exp with scenario=<ENV> 

Possible ENVs are:

• case_study
• 6-11m_symmetric
• sym_asym