Card games

This repository contain codes to train to play card games through reinforcement learning.

Folder structure

• algorithm contains reinforcement learning algorithms. Currently, only PPO algorithm are implemented and customised for big two.
• benchmark pitching different bots against each other to see who is the best
• bigtwo code to run the big two card game
• bigtwo_client a simple GUI application to play bigtwo against the bots
• gamerunner training the bots using self implemented PPO algorithm
• rayrunner: training the bots using Ray RLlib
• playingcards: classes that represent playing cards

Installing dependencies

First poetry, then run poetry install

Run BigTwo GUI client

cd big_client
PYTHONPATH=.. python client.py

Run BigTwo PPO training

Training configuration is stored in `gamerunner/config``

cd gamerunner
PYTHONPATH=.. python ppo_runner.py

Run tests

make run_test