MADT: Offline Pre-trained Multi-Agent Decision Transformer

A link to our paper can be found on Arxiv.

Overview

Official codebase for Offline Pre-trained Multi-Agent Decision Transformer. Contains scripts to reproduce experiments.

image info

Instructions

It may be necessary to add the respective directories to your PYTHONPATH.

The offline smac dataset for this repo is available at here.

## password is vp1ulk

How to run experiments

  1. setup python environment with pip install -r requirements.txt
  2. to install StarCraft II & SMAC, you could run bash install_sc2.sh. Or you could install them manually to other path you like, following the official link: https://github.com/oxwhirl/smac.
  3. enter the ./sc2/ folder.
  4. set hyper-parameters in run_madt_sc2.py line 19-49 according to appendix.
  5. select a maps to test in envs/config.py line 142
  6. run the run_madt_sc2.py script