headsup-poker

Heads-Up poker openai gym environment for fun. DeepCFR training and evaluation.

Visualization and demo

python visualize.py

DeepCFR training

cd deepcfr; python train_deepcfr.py

RLGames exploitability evaluation

Add gym environment registration to rl_games/common/env_configurations.py

gym.register(id="HeadsUpPokerRLGames-v0", entry_point="rl_games_env:HeadsUpPokerRLGames")

Training

python exploitability.py -f rl_config/poker_env.yaml

Evaluation (current model: 500 mbb/g)

python exploitability.py -f rl_config/poker_env.yaml -p -c runs/<exp_folder>/<checkpoint name>.pth

Model conversion to onnx

python rl_games_onnx.py -f rl_config/poker_env.yaml -m runs/<exp_folder>/<checkpoint name>.pth

Play against rl-games model

Uncomment player = ONNXRLGamesPlayer() in visualize.py, use python visualize.py to play against bot