Python atari_py.ALEInterface方法代码示例

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(self, args):
    self.device = args.device
    self.ale = atari_py.ALEInterface()
    self.ale.setInt('random_seed', args.seed)
    self.ale.setInt('max_num_frames_per_episode', args.max_episode_length)
    self.ale.setFloat('repeat_action_probability', 0)  # Disable sticky actions
    self.ale.setInt('frame_skip', 0)
    self.ale.setBool('color_averaging', False)
    self.ale.loadROM(atari_py.get_game_path(args.game))  # ROM loading must be done after setting options
    actions = self.ale.getMinimalActionSet()
    self.actions = dict([i, e] for i, e in zip(range(len(actions)), actions))
    self.lives = 0  # Life counter (used in DeepMind training)
    self.life_termination = False  # Used to check if resetting only from loss of life
    self.window = args.history_length  # Number of frames to concatenate
    self.state_buffer = deque([], maxlen=args.history_length)
    self.training = True  # Consistent with model training mode 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(self, args, process_ind=0, num_envs_per_process=1):
        super(AtariEnv, self).__init__(args, process_ind, num_envs_per_process)

        # env_params for this env
        assert self.num_envs_per_process == 1
        self.seed = self.seed + self.process_ind * self.num_envs_per_actor  # NOTE: check again

        # setup ale
        self.ale = atari_py.ALEInterface()
        self.ale.setInt('random_seed', self.seed)
        self.ale.setInt('max_num_frames', self.early_stop)
        self.ale.setFloat('repeat_action_probability', 0)   # Disable sticky actions
        self.ale.setInt('frame_skip', 0)
        self.ale.setBool('color_averaging', False)
        print(atari_py.get_game_path(self.game))
        self.ale.loadROM(atari_py.get_game_path(self.game)) # ROM loading must be done after setting options
        actions = self.ale.getMinimalActionSet()
        self.actions = dict([i, e] for i, e in zip(range(len(actions)), actions))

        self.lives = 0          # life counter (used in DeepMind training)
        self.just_died = False  # when lost one life, but game is still not over

        # setup
        self.exp_state1 = deque(maxlen=self.state_cha)
        self._reset_experience() 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(self, args):
        self.device = args.device
        self.ale = atari_py.ALEInterface()
        self.ale.setInt("random_seed", args.seed)
        self.ale.setInt("max_num_frames_per_episode", args.max_episode_length)
        self.ale.setFloat("repeat_action_probability", 0)  # Disable sticky actions
        self.ale.setInt("frame_skip", 0)
        self.ale.setBool("color_averaging", False)
        # ROM loading must be done after setting options
        self.ale.loadROM(atari_py.get_game_path(args.game))
        actions = self.ale.getMinimalActionSet()
        self.actions = dict([i, e] for i, e in zip(range(len(actions)), actions))
        self.action_space = spaces.Discrete(len(self.actions))
        self.lives = 0  # Life counter (used in DeepMind training)
        self.life_termination = False  # Used to check if resetting only from loss of life
        self.window = args.history_length  # Number of frames to concatenate
        self.state_buffer = deque([], maxlen=args.history_length)
        self.training = True  # Consistent with model training mode 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(self, game='pong', obs_type='ram', frameskip=(2, 5), repeat_action_probability=0.):
        """Frameskip should be either a tuple (indicating a random range to
        choose from, with the top value exclude), or an int."""

        utils.EzPickle.__init__(self, game, obs_type)
        assert obs_type in ('ram', 'image')

        self.game_path = atari_py.get_game_path(game)
        if not os.path.exists(self.game_path):
            raise IOError('You asked for game %s but path %s does not exist'%(game, self.game_path))
        self._obs_type = obs_type
        self.frameskip = frameskip
        self.ale = atari_py.ALEInterface()
        self.viewer = None

        # Tune (or disable) ALE's action repeat:
        # https://github.com/openai/gym/issues/349
        assert isinstance(repeat_action_probability, (float, int)), "Invalid repeat_action_probability: {!r}".format(repeat_action_probability)
        self.ale.setFloat('repeat_action_probability'.encode('utf-8'), repeat_action_probability)

        self.seed()

        self._action_set = self.ale.getMinimalActionSet()
        self.action_space = spaces.Discrete(len(self._action_set))

        (screen_width,screen_height) = self.ale.getScreenDims()
        if self._obs_type == 'ram':
            self.observation_space = spaces.Box(low=0, high=255, dtype=np.uint8, shape=(128,))
        elif self._obs_type == 'image':
            self.observation_space = spaces.Box(low=0, high=255, shape=(screen_height, screen_width, 3), dtype=np.uint8)
        else:
            raise error.Error('Unrecognized observation type: {}'.format(self._obs_type)) 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def get_num_actions(rom_path, rom_name):
    #import os
    #print os.path.abspath(atari_py.__file__)
    game_path = atari_py.get_game_path(rom_name)
    ale = atari_py.ALEInterface()
    ale.loadROM(game_path)
    return ale.getMinimalActionSet() 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(self, game='pong', obs_type='ram', frameskip=(2, 5), repeat_action_probability=0.):
        """Frameskip should be either a tuple (indicating a random range to
        choose from, with the top value exclude), or an int."""

        utils.EzPickle.__init__(self, game, obs_type, frameskip, repeat_action_probability)
        assert obs_type in ('ram', 'image')

        self.game_path = atari_py.get_game_path(game)
        if not os.path.exists(self.game_path):
            raise IOError('You asked for game %s but path %s does not exist'%(game, self.game_path))
        self._obs_type = obs_type
        self.frameskip = frameskip
        self.ale = atari_py.ALEInterface()
        self.viewer = None

        # Tune (or disable) ALE's action repeat:
        # https://github.com/openai/gym/issues/349
        assert isinstance(repeat_action_probability, (float, int)), "Invalid repeat_action_probability: {!r}".format(repeat_action_probability)
        self.ale.setFloat('repeat_action_probability'.encode('utf-8'), repeat_action_probability)

        self.seed()

        self._action_set = self.ale.getMinimalActionSet()
        self.action_space = spaces.Discrete(len(self._action_set))

        (screen_width,screen_height) = self.ale.getScreenDims()
        if self._obs_type == 'ram':
            self.observation_space = spaces.Box(low=0, high=255, dtype=np.uint8, shape=(128,))
        elif self._obs_type == 'image':
            self.observation_space = spaces.Box(low=0, high=255, shape=(screen_height, screen_width, 3), dtype=np.uint8)
        else:
            raise error.Error('Unrecognized observation type: {}'.format(self._obs_type)) 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(self,
                 game="pong",
                 frame_skip=4,  # Frames per step (>=1).
                 num_img_obs=4,  # Number of (past) frames in observation (>=1).
                 clip_reward=True,
                 episodic_lives=True,
                 fire_on_reset=False,
                 max_start_noops=30,
                 repeat_action_probability=0.,
                 horizon=27000,
                 ):
        save__init__args(locals(), underscore=True)
        # ALE
        game_path = atari_py.get_game_path(game)
        if not os.path.exists(game_path):
            raise IOError("You asked for game {} but path {} does not "
                " exist".format(game, game_path))
        self.ale = atari_py.ALEInterface()
        self.ale.setFloat(b'repeat_action_probability', repeat_action_probability)
        self.ale.loadROM(game_path)

        # Spaces
        self._action_set = self.ale.getMinimalActionSet()
        self._action_space = IntBox(low=0, high=len(self._action_set))
        obs_shape = (num_img_obs, H, W)
        self._observation_space = IntBox(low=0, high=255, shape=obs_shape,
            dtype="uint8")
        self._max_frame = self.ale.getScreenGrayscale()
        self._raw_frame_1 = self._max_frame.copy()
        self._raw_frame_2 = self._max_frame.copy()
        self._obs = np.zeros(shape=obs_shape, dtype="uint8")

        # Settings
        self._has_fire = "FIRE" in self.get_action_meanings()
        self._has_up = "UP" in self.get_action_meanings()
        self._horizon = int(horizon)
        self.reset() 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(self, game, seed=None, use_sdl=False, n_last_screens=4,
                 frame_skip=4, treat_life_lost_as_terminal=True,
                 crop_or_scale='scale', max_start_nullops=30,
                 record_screen_dir=None):
        assert crop_or_scale in ['crop', 'scale']
        assert frame_skip >= 1
        self.n_last_screens = n_last_screens
        self.treat_life_lost_as_terminal = treat_life_lost_as_terminal
        self.crop_or_scale = crop_or_scale
        self.max_start_nullops = max_start_nullops

        # atari_py is used only to provide rom files. atari_py has its own
        # ale_python_interface, but it is obsolete.
        if not atari_py_available:
            raise RuntimeError(
                'You need to install atari_py>=0.1.1 to use ALE.')
        game_path = atari_py.get_game_path(game)

        ale = atari_py.ALEInterface()
        if seed is not None:
            assert seed >= 0 and seed < 2 ** 31, \
                "ALE's random seed must be in [0, 2 ** 31)."
        else:
            # Use numpy's random state
            seed = np.random.randint(0, 2 ** 31)
        ale.setInt(b'random_seed', seed)
        ale.setFloat(b'repeat_action_probability', 0.0)
        ale.setBool(b'color_averaging', False)
        if record_screen_dir is not None:
            ale.setString(b'record_screen_dir',
                          str.encode(str(record_screen_dir)))
        self.frame_skip = frame_skip
        if use_sdl:
            if 'DISPLAY' not in os.environ:
                raise RuntimeError(
                    'Please set DISPLAY environment variable for use_sdl=True')
            # SDL settings below are from the ALE python example
            if sys.platform == 'darwin':
                import pygame
                pygame.init()
                ale.setBool(b'sound', False)  # Sound doesn't work on OSX
            elif sys.platform.startswith('linux'):
                ale.setBool(b'sound', True)
            ale.setBool(b'display_screen', True)

        ale.loadROM(str.encode(str(game_path)))

        assert ale.getFrameNumber() == 0

        self.ale = ale
        self.legal_actions = ale.getMinimalActionSet()
        self.initialize()

        self.action_space = spaces.Discrete(len(self.legal_actions))
        one_screen_observation_space = spaces.Box(
            low=0, high=255,
            shape=(84, 84), dtype=np.uint8,
        )
        self.observation_space = spaces.Tuple(
            [one_screen_observation_space] * n_last_screens) 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(self, rom_path, rom_name, visualize, actor_id, rseed, single_life_episode = False):
        
        self.ale = atari_py.ALEInterface()

        self.ale.setInt("random_seed", rseed * (actor_id +1))

        # For fuller control on explicit action repeat (>= ALE 0.5.0) 
        self.ale.setFloat("repeat_action_probability", 0.0)
        
        # See: http://is.gd/tYzVpj
        self.ale.setInt("frame_skip", 4)
        #self.ale.setBool("color_averaging", False)
        self.ale.loadROM(atari_py.get_game_path(rom_name))
        self.legal_actions = self.ale.getMinimalActionSet()        
        self.single_life_episode = single_life_episode
        self.initial_lives = self.ale.lives()
        
        # Processed frames that will be fed in to the network 
        # (i.e., four 84x84 images)
        self.processed_imgs = np.zeros((IMG_SIZE_X, IMG_SIZE_Y, 
            NR_IMAGES), dtype=np.uint8) 

        self.screen_width,self.screen_height = self.ale.getScreenDims()
        self.rgb_screen = np.zeros((self.screen_height,self.screen_width, 4), dtype=np.uint8)
        self.gray_screen = np.zeros((self.screen_height,self.screen_width,1), dtype=np.uint8)
        
        self.visualize = visualize
        self.visualize_processed = False
        rendering_imported = False
#         if self.visualize:
#             from gym.envs.classic_control import rendering
#             rendering_imported = True
#             logger.debug("Opening emulator window...")
#             self.viewer = rendering.SimpleImageViewer()
#             self.render()
#             logger.debug("Emulator window opened")
#             
#         if self.visualize_processed:
#             if not rendering_imported:
#                 from gym.envs.classic_control import rendering
#             logger.debug("Opening emulator window...")
#             self.viewer2 = rendering.SimpleImageViewer()
#             self.render()
#             logger.debug("Emulator window opened") 

# 需要导入模块: import atari_py [as 别名]
# 或者: from atari_py import ALEInterface [as 别名]
def __init__(
            self,
            game='pong',
            mode=None,
            difficulty=None,
            obs_type='ram',
            frameskip=(2, 5),
            repeat_action_probability=0.,
            full_action_space=False):
        """Frameskip should be either a tuple (indicating a random range to
        choose from, with the top value exclude), or an int."""

        utils.EzPickle.__init__(
                self,
                game,
                mode,
                difficulty,
                obs_type,
                frameskip,
                repeat_action_probability)
        assert obs_type in ('ram', 'image')

        self.game = game
        self.game_path = atari_py.get_game_path(game)
        self.game_mode = mode
        self.game_difficulty = difficulty

        if not os.path.exists(self.game_path):
            msg = 'You asked for game %s but path %s does not exist'
            raise IOError(msg % (game, self.game_path))
        self._obs_type = obs_type
        self.frameskip = frameskip
        self.ale = atari_py.ALEInterface()
        self.viewer = None

        # Tune (or disable) ALE's action repeat:
        # https://github.com/openai/gym/issues/349
        assert isinstance(repeat_action_probability, (float, int)), \
                "Invalid repeat_action_probability: {!r}".format(repeat_action_probability)
        self.ale.setFloat(
                'repeat_action_probability'.encode('utf-8'),
                repeat_action_probability)

        self.seed()

        self._action_set = (self.ale.getLegalActionSet() if full_action_space
                            else self.ale.getMinimalActionSet())
        self.action_space = spaces.Discrete(len(self._action_set))

        (screen_width, screen_height) = self.ale.getScreenDims()
        if self._obs_type == 'ram':
            self.observation_space = spaces.Box(low=0, high=255, dtype=np.uint8, shape=(128,))
        elif self._obs_type == 'image':
            self.observation_space = spaces.Box(low=0, high=255, shape=(screen_height, screen_width, 3), dtype=np.uint8)
        else:
            raise error.Error('Unrecognized observation type: {}'.format(self._obs_type))