vatsj / gamer-bot

framework for training RL agents to play (simple, finite) games

Gamer-bot

framework for training RL agents to play (simple, finite) games

Capabilities

Provides gamer package, which consists of the following sub-packages:

• games: finite n-player games
• players: players of a given gamer.games game, with policy deciding moves given the game state
• strategists: agents designed to optimize strategy for a given game; trains by producing players and having them play games

How to run

The main file is run from the project's root directory via the following command

python src/main.py

Dependencies

A incomplete list is below:

• Python 3.3 and up
• reasonably up-to-date npm
• Python packages: numpy, scipy, chess (can be installed with npm i numpy scipy chess)