(Part of the codes in this repository can be still incomplete.)

Overview

This is a repository for training Baba-Is-You agents by RL, for visualizing the agents, and for debugging a simulator forked from https://github.com/utilForever/baba-is-auto.

Requirements

• python >= 3.7
• cmake >= 3.14.3
• g++ (developed in g++ 8.3.0)

Setup

pip install -r requirements.txt

git clone https://github.com/jack-and-rozz/baba-is-auto --recursive
cd baba-is-auto 
pip install --editable .
cd ..

How to run

# Manual playing
python manual_play.py --map_name [map_name]

# Training a RL agent
python REINFORCE.py [save_dir]

# Save replays of a trained agent as gif
python create_gif.py [save_dir] --map_name [map_name]

Thanks to

• Chris Oak