Skip to content
This repository has been archived by the owner on Sep 25, 2022. It is now read-only.
/ amca Public archive

An RL-trained Backgammon agent

License

Notifications You must be signed in to change notification settings

ardabbour/amca

Repository files navigation

Amca

Status: Under construction.

Amca is an RL-based Backgammon agent.

Dependencies

Dependency Version Tested On
Ubuntu 16.04
Python 3.6.8
numpy 1.15.4
gym 0.10.9
Stable Baselines 2.4.0a

About

This project aims to design Backgammon as a reinforcement learning problem, and gauge the performance of common deep reinforcement learning algorithms. This is done by training and gauging the performance of three popular and powerful RL algorithms:

The testing is done with the default parameters and implementations provided by the Stable Baselines library for all the 3 deep RL algorithms. A custom implementation heavily modified from this repo is used for SARSA, and the hyperparameters are given in the SarsaAgent object.

Usage

  • play.py: to launch a game against a deep RL trained model. For example, python play.py ppo amca/models/amca.pkl will launch the model called amca.pkl that was trained using the PPO algorithm.
  • train.py: to train an deep RL model (with default hyperparameters) to play. For example, python train.py -n terminator.pkl -a sac -t 1000000 will train an agent called terminator.pkl using the SAC algorithm for 1000000 steps.
  • sarsa_play.py: to launch a game against a SARSA trained model. python sarsa_play.py r2d2.pkl will launch the model called r2d2.pkl that was trained using the SARSA algorithm.
  • sarsa_train.py: to train a model using SARSA. For example, python sarsa_train.py jarvis.pkl -g 10000 will train an agent called jarvis.pkl using the SARSA algorithm for 10000 games.

License

GNU General Public License v3.0