rlskyjo.environment package
Submodules
rlskyjo.environment.skyjo_env module
- class rlskyjo.environment.skyjo_env.SimpleSkyjoEnv(num_players=2, score_penalty: float = 2.0, observe_other_player_indirect: bool = False, mean_reward: float = 1.0, reward_refunded: float = 0.0)
Bases:
pettingzoo.utils.env.AECEnv- action_space(agent)
part of the PettingZoo API action_space is Discrete(26):
0-11: place hand card to position 0-11 12-23: discard place hand card and reveal position 0-11 24: pick hand card from drawpile 25: pick hand card from discard pile
- Args:
agent ([type]): [description]
- Returns:
[type]: [description]
- close() None
part of the PettingZoo API
- metadata = {'is_parallelizable': False, 'name': 'skyjo', 'render.modes': ['human'], 'video.frames_per_second': 1}
- observation_space(agent)
- observations are:
(1,) lowest sum of players, calculated feature (1,) lowest number of unmasked cards of any player,
calculated feature
- (15,) counts of cards past discard pile cards & open player cards,
calculated feature
(1,) top discard pile card (1,) current hand_card total: (19,)
- if observe_other_player_indirect is True:
# constant for any num_players (12) own cards total: (31,)
- elif observe_other_player_indirect is False:
(num_players*4*3,) total: (19+12*num_players,)
- Args:
agent ([type]): agent string
- Returns:
gym.space: observation_space of agent
- observe(agent: str) Dict[str, numpy.ndarray]
get observation and action mask from environment part of the PettingZoo API]
- Args:
agent ([str]): agent string
- Returns:
dict: {“observations”: np.ndarray, “action_mask”: np.ndarray}
- render(mode='human') None
render board of the game to stdout
part of the PettingZoo API
- reset() None
reset the environment part of the PettingZoo API
- seed(seed: Optional[int] = None) None
seed the environment. does not affect global np.random.seed() experimental. only works with Numba installed part of the PettingZoo API
- Args:
seed (int, optional): [description]. Defaults to None.
- step(action: int) None
part of the PettingZoo API
- Args:
- action (int):
- action is number from 0-25:
0-11: place hand card to position 0-11 12-23: discard place hand card and reveal position 0-11 24: pick hand card from drawpile 25: pick hand card from discard pile
- Returns:
None:
- rlskyjo.environment.skyjo_env.env(**kwargs)
wrap SkyJoEnv in
rlskyjo.environment.vanilla_env_example module
- rlskyjo.environment.vanilla_env_example.simple_episode(config, verbose=0)
create a vanilla example