Skip to content
forked from hijkzzz/pymarl2

Fine-tuned MARL algorithms on SMAC (100% win rates on most scenarios)

License

Notifications You must be signed in to change notification settings

benellis3/pymarl2

 
 

Repository files navigation

Running Hyperparameter Tuning

  1. Build the Docker container by running docker build -t pymarl2:ben_smac -f docker/Dockerfile --build-arg UID=$UID . from the pymarl2 directory
  2. Set up StarCraft II by running ./install_sc2.sh. Make sure to install the 32x32_flat.SC2Map by copying it to the $SC2PATH/Maps directory from the smacv2 repo.
  3. Make sure to either set use_wandb=False or to change the project and entity in src/config/default.yaml.
  4. Run the run_exp.sh script.

Citation

This repository has very few changes from previous QMIX work, except that it runs on SMACv2. The original repository is here.

@article{hu2021revisiting,
      title={Revisiting the Monotonicity Constraint in Cooperative Multi-Agent Reinforcement Learning}, 
      author={Jian Hu and Siyang Jiang and Seth Austin Harding and Haibin Wu and Shih-wei Liao},
      year={2021},
      eprint={2102.03479},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

About

Fine-tuned MARL algorithms on SMAC (100% win rates on most scenarios)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 97.5%
  • Shell 2.4%
  • Dockerfile 0.1%