train_cartpole.py 文件源码

def main():
    env = gym.make("CartPole-v0")
    model = deepq.models.mlp([64])
    act = deepq.learn(
        env,
        q_func=model,
        lr=1e-3,
        max_timesteps=100000,
        buffer_size=50000,
        exploration_fraction=0.1,
        exploration_final_eps=0.02,
        print_freq=10,
        callback=callback
    )
    print("Saving model to cartpole_model.pkl")
    act.save("cartpole_model.pkl")