openai-tests/ql/play.py

from . import *

def play(player_episodes=1):
    q_table = pickle.load( open( gym_name+".dat", "rb" ))

    total_epochs = 0
    episodes = player_episodes

    print(f"Evaluation: 0%")

    total_epochs, total_rewards = 0, 0

    try:
        for ep in range(episodes):
            state = env.reset()
            epochs, reward = 0, 0
            
            done = False

            i = 0
            while not done:
                action = np.argmax(q_table[state])
                state, reward, done, info = env.step(action)

                if i > max_iterations:
                    done = True

                ftext = ""
                if use_ansi:
                    ftext = env.render(mode="ansi")
                else:
                    ftext = str(info)

                i += 1
                print (u"{}[2J{}[;H".format(chr(27), chr(27)))
                print(f"Evaluation: {100 * ep / episodes}%")
                print(f"{ftext}")
                sleep(.1)

                epochs += 1

            total_epochs += epochs
            total_rewards += reward


            sleep(1)
    except KeyboardInterrupt:
        print(f"Results after {episodes} episodes:")
        print(f"Average timesteps per episode: {total_epochs / episodes}")
        print(f"Average rewards per episode: {total_rewards / episodes}")

        exit()

    print (u"{}[2J{}[;H".format(chr(27), chr(27)))
    print("Evaluation: finished.\n")

    print(f"Results after {episodes} episodes:")
    print(f"Average timesteps per episode: {total_epochs / episodes}")
    print(f"Average rewards per episode: {total_rewards / episodes}")