We provide some running examples. We will update the examples if we achieve better results on large games such as Dou Dizhu, UNO and Mahjong.
blackjack_dqn.py: train DQN on Blackjack.blackjack_dqn_multi_process.py: train DQN on Blackjack with multiple processes.blackjack_random.py: run random agents on Blackjcak.doudizhu_dqn.py: train DQN on Dou Dizhu.doudizhu_nfsp.py: train NFSP on Dou Dizhu.doudizhu_random.py: run random agents on Dou Dizhu.doudizhu_random_multi_process.py: run random agents on Dou Dizhu with multiple processes.doudizhu_random_process_pool.py:run random agents on Dou Dizhu with multiple processes using process pool.leduc_holdem_cfr.py: train CFR on Leduc Hold'em.leduc_holdem_dqn.py: train DQN on Leduc Hold'em.leduc_holdem_human.py: play against re-trained model on Leduc Hold'em.leduc_holdem_nfsp.py: train NFSP on Leduc Hold'em.leduc_holdem_random.py: run random agents on Leduc Hold'em.leduc_holdem_single.py: train DQN on Leduc Hold'em as single-agent environment.limit_holdem_dqn.py: train DQN on Limit Texas Hold'em.limit_holdem_nfsp.py: train NFSP on Limit Texas Hold'em.limit_holdem_random.py: run random agents on Limit Texas Hold'em.mahjong_dqn.py: train DQN on Mahjong.mahjong_nfsp.py: train NFSP on Mahjong.mahjong_random.py: run random agents on Mahjong.nolimit_holdem_dqn.py: train DQN on No-Limit Texas Hold'em.nolimit_holdem_nfsp.py: train NFSP on No-Limit Texas Hold'em.nolimit_holdem_random.py: run random agents on No-Limit Gexas Hold'em.uno_dqn.py: train DQN on UNO.uno_human.py: play against rule-based model on UNO.uno_nfsp.py: train NFSP on UNO.uno_random.py: run random agents on UNO.uno_single.py: train DQN on UNO as single-agent environment.- Save models: refer to
leduc_holdem_nfsp_save_model.pyandleduc_holdem_cfr.pyfor CFR. - Load models: refer to rlcard/models/pretrained_models.py