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)
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 total_reward += _rew #saved_state = env.em.get_state() #doesn't work with WarpFrame and FrameStack wrappers current_x = _info['x'] current_y = _info['y'] keys_csv.writerow({ 'step': num_steps, 'keys': key_string, 'action': game_a, 'r': _rew, 'x': current_x, 'y': current_y, 'rings': _info['rings'] })