mit-ll/hanabi_anyplay

Overview: Any-Play Learning Augmentation for Zero-Shot Coordination

This library implements the Any-Play learning augmentation in Hanabi Learning Environment. Any-Play is an intrisictly-motivated, diversity-based augmentation for reinforcement learning algorithms (RL) that enables RL agents to effectively cooperate with novel, never-before-seen teammates on collaborative tasks; referred to as zero-shot coordination. This library is used to train and demonstrate such teaming in the cooperative card game Hanabi; although the Any-Play augmentation is environment agnostic, and therefore, could be applied to many other domains beyone Hanabi.

This library, "hanabi_AnyPlay", is a derivative of the hanabi_SAD library by FACEBOOK, used under the CC BY-NC license license. "hanabi_AnyPlay" is licensed under the same CC BY-NC license by MIT.

Citation

When using this library please cite the Any-Play paper as well as the orginal Other-Play and Simplified Action Decoder works on which this is based

Any-Play

@inproceedings{lucas2022any,
  title={Any-Play: An Intrinsic Augmentation for Zero-Shot Coordination},
  author={Lucas, Keane and Allen, Ross E},
  booktitle={Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems},
  pages={853--861},
  year={2022}
}

Other-Play

@incollection{icml2020_5369,
 author = {Hu, Hengyuan and Peysakhovich, Alexander and Lerer, Adam and Foerster, Jakob},
 booktitle = {Proceedings of Machine Learning and Systems 2020},
 pages = {9396--9407},
 title = {\textquotedblleft Other-Play\textquotedblright  for Zero-Shot Coordination},
 year = {2020}
}

Simplfied Action Decoder

@inproceedings{
Hu2020Simplified,
title={Simplified Action Decoder for Deep Multi-Agent Reinforcement Learning},
author={Hengyuan Hu and Jakob N Foerster},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=B1xm3RVtwB}
}

Other-Play & Simplified Action Decoder in Hanabi

Important Update, Mar-2021

We uploaded one off-belief-learning (OBL) model from our recent paper. To get this model, go to hanabi_SAD/models and run

wget https://dl.fbaipublicfiles.com/hanabi_op/obl.zip
unzip obl.zip

To use this model, go to hanabi_SAD/pyhanabi and run

python tools/eval_model.py --paper obl --num_game 5000

Important Update, Feb-2021

We uploaded the models from the Other-Play paper. To get those models, run the updated download.sh in the models folder. If you only need the Other-Play models, you can download them by running the following command from the models folder

wget https://dl.fbaipublicfiles.com/hanabi_op/op.zip
unzip op.zip

We also include the model evaluation data in models/op_raw_data.txt. The data in this file is used for Figure 4 and Table 1 in the paper.

We updated the evaluation script to allow both self-play and cross-play evaluation using the new other-play models.

# assume current work directory is pyhanabi
# method can be sad, sad-op, sad-aux, sad-aux-op
# idx1/idx2 ranges from [0, 11], corresponding to the 12 models.
python tools/eval_model.py --paper op --method sad-aux --idx1 0 --idx2 0

The evaluation script assumes that the models are saved in the $ROOT/models folder.

The model used for human evaluation in the paper was models/op/sad-aux-op/M1.pthw, which was the model with the highest cross-play score and trained with the best method.

Important Update, Sep-2020

The repo has been updated to include Other-Play, auxiliary task, as well as improved training infrastructure. The build process has also been significantly simplfied. It is no longer necessary to build pytorch from source (thanks to changes in pytorch1.5) and the code now works with newer version of pytorch and cuda. It also avoids the hanging problem that may appear in previous version of the codebase on certain hardware configuration.

Introduction

This repo contains code and models for "Other-Play" for Zero-Shot Coordination and Simplified Action Decoder for Deep Multi-Agent Reinforcement Learning.

To reference these works, please use:

Other-Play

@incollection{icml2020_5369,
 author = {Hu, Hengyuan and Peysakhovich, Alexander and Lerer, Adam and Foerster, Jakob},
 booktitle = {Proceedings of Machine Learning and Systems 2020},
 pages = {9396--9407},
 title = {\textquotedblleft Other-Play\textquotedblright  for Zero-Shot Coordination},
 year = {2020}
}

Simplfied Action Decoder

@inproceedings{
Hu2020Simplified,
title={Simplified Action Decoder for Deep Multi-Agent Reinforcement Learning},
author={Hengyuan Hu and Jakob N Foerster},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=B1xm3RVtwB}
}

Compile

We have been using pytorch-1.5.1, cuda-10.1, and cudnn-v7.6.5 in our development environment. Other settings may also work but we have not tested it extensively under different configurations. We also use conda/miniconda to manage environments.

# create new conda env
conda create -n hanabi python=3.7
conda activate hanabi

# install pytorch
pip install torch==1.5.1+cu101 torchvision==0.6.1+cu101 -f https://download.pytorch.org/whl/torch_stable.html

# install other dependencies
pip install numpy
pip install psutil

# if the current cmake version is < 3.15
conda install -c conda-forge cmake

Clone & Build this repo

For convenience, add the following lines to your .bashrc, after the line of conda activate xxx.

# set path
CONDA_PREFIX=${CONDA_PREFIX:-"$(dirname $(which conda))/../"}
export CPATH=${CONDA_PREFIX}/include:${CPATH}
export LIBRARY_PATH=${CONDA_PREFIX}/lib:${LIBRARY_PATH}
export LD_LIBRARY_PATH=${CONDA_PREFIX}/lib:${LD_LIBRARY_PATH}

# avoid tensor operation using all cpu cores
export OMP_NUM_THREADS=1

Clone & build.

git clone --recursive https://github.com/facebookresearch/hanabi.git

cd hanabi
mkdir build
cd build
cmake ..
make -j10

Run

hanabi/pyhanabi/tools contains some example scripts to launch training runs. dev.sh is a fast lauching script for debugging. It needs 2 gpus to run, 1 for training and 1 for simulation. Other scripts are examples for a more formal training run, they require 3 gpus, 1 for training and 2 for simulation.

The important flags are:

--sad 1 to enable "Simplified Action Decoder";

--pred_weight 0.25 to enable auxiliary task and multiply aux loss with 0.25;

--shuffle_color 1 to enable other-play.

cd pyhanabi
sh tools/dev.sh

Trained Models

Run the following command to download the trained models used to produce tables in the paper.

cd model
sh download.sh

To evaluate a model, simply run

cd pyhanabi
python tools/eval_model.py --weight ../models/sad_2p_10.pthw --num_player 2

Related Repos

The results on Hanabi can be further improved by running search on top of our agents. Please refer to the paper and code for details.

We also open-sourced a single agent implementation of R2D2 tested on Atari here.

Contribute

Python

Use black to format python code, run black *.py before pushing

C++

The root contains a .clang-format file that define the coding style of this repo, run the following command before submitting PR or push

clang-format -i *.h
clang-format -i *.cc

Disclaimer

DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited.

Copyright

hanabi_SAD library Copyright (c) Facebook, Inc. and its affiliates. All rights reserved.

MIT authored extensions in the "hanabi_AnyPlay" library © 2022 Massachusetts Institute of Technology Subject to FAR 52.227-11 – Patent Rights – Ownership by the Contractor (May 2014) SPDX-License-Identifier: CC-BY-NC-4.0

This material is based upon work supported by the Under Secretary of Defense for Research and Engineering under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Under Secretary of Defense for Research and Engineering.

The software/firmware is provided to you on an As-Is basis