def main(): env = make_atari("BreakoutNoFrameskip-v0") env = WarpFrame(env) env = FrameStack(env, k=4) act = deepq.load("breakout_model.pkl") while True: obs, done = env.reset(), False episode_rew = 0 while not done: env.render() obs, rew, done, _ = env.step(act(obs[None])[0]) episode_rew += rew print("Episode reward", episode_rew)
keys_csv = csv.DictWriter( keys_file, fieldnames=['step', 'keys', 'action', 'r', 'x', 'y', 'rings']) keys_csv.writeheader() rew_constant = 0.5 prev_action = -19 print('stepping movie') transitions = [] debug = int(sys.argv[1]) while movie.step(): if debug: env.render() time.sleep(0.005) keys = [] key_string = '_' for i in range(num_buttons): keys.append(movie.get_key(i)) if movie.get_key(i): key_string += button_dict[i] + "_" game_a = game_dict[game_get_dict_key(keys)] trans = {} trans['episode_id'] = 0 trans['obs'] = list(obs_list) #defined by env.reset() above for first one _obs, _rew, _done, _info = env.step(keys) num_steps += 1