Models (Player 1) vs in-game AI (Player 2):
python3 model_vs_game.py --env=NHL941on1-Genesis --state=PenguinsVsSenators --model_1=./models/DefenseZone --model_2=./models/ScoreGoal --nn=MlpPolicy --rf=General
Play against the models
python3 player_vs_model.py --env=NHL941on1-Genesis --state=PenguinsVsSenators.2P --model_1=./models/DefenseZone --model_2=./models/ScoreGoal --nn=MlpPolicy --rf=General --num_players=2
Note that DefenseZone is expected to be on model_1 slot and ScoreGoal is expected to be model_2 slot ex: DefenseZone
python3 player_vs_model.py --env=NHL941on1-Genesis --state=PenguinsVsSenators.lostpuck.2P --nn=MlpPolicy --num_players=2 --rf="DefenseZone" --model_1=./models/DefenseZone
ex: ScoreGoal
python3 player_vs_model.py --env=NHL941on1-Genesis --state=PenguinsVsSenators.frontnet.2P --nn=MlpPolicy --num_players=2 --rf="ScoreGoal" --model_2=./models/ScoreGoal
python3 model_trainer.py --env=NHL941on1-Genesis --state=PenguinsVsSenators.FrontOfNet --num_env=12 --num_timesteps=100_000_000 --nn=MlpPolicy --play --num_players=1 --rf="ScoreGoal"
NHL94 Discord (with a subgroup dedicated for AI): https://discord.gg/SDnKEXujDs
Video of AI playing NHL94 (1 on 1) with explaination about the reward functions: https://www.youtube.com/watch?v=UBXXn2amGUU