Import
from pettingzoo.classic import hanabi_v4
from pettingzoo.classic import hanabi_v5
Hanabi takes in a number of arguments defining the size and complexity of the game. Default is a full 2 player hanabi game.
-hanabi_v4.env(colors=5, rank=5, players=2, hand_size=5, max_information_tokens=8,
-max_life_tokens=3, observation_type=1)
+hanabi_v5.env(colors=5, rank=5, players=2, hand_size=5, max_information_tokens=8,
+max_life_tokens=3, observation_type="minimal")
colors: Number of colors the cards can take (affects size of deck)
@@ -445,7 +445,11 @@ Environment argumentshand_size: Size of player’s hands. Standard game is (4 if players >= 4 else 5)
max_information_tokens: Maximum number of information tokens (more tokens makes the game easier by allowing more information to be revealed)
max_life_tokens: Maximum number of life tokens (more tokens makes the game easier by allowing more information to be revealed)
-observation_type: 0: Minimal observation. 1: First-order common knowledge observation (default).
+observation_type:
+“minimal”: Minimal observation (what a human sees).
+“card_knowledge”: includes per-card knowledge of past hints, as well as simple inferred knowledge of the form
+“this card is not red, because it was not revealed as red in a past”.
+“seer” shows all cards, including the player’s own cards, regardless of what hints have been given.
Observation Space#
@@ -692,6 +696,7 @@ Rewards
Version History#
+v5: Switched environment to depend on OpenSpiel (using Shimmy) for future compatibility (1.23.0)
v4: Fixed bug in arbitrary calls to observe() (1.8.0)
v3: Legal action mask in observation replaced illegal move list in infos (1.5.0)
v2: Fixed default parameters (1.4.2)
API#
-
-class pettingzoo.classic.hanabi.hanabi.raw_env(colors: int = 5, ranks: int = 5, players: int = 2, hand_size: int = 5, max_information_tokens: int = 8, max_life_tokens: int = 3, observation_type: int = 1, random_start_player: bool = False, render_mode: str | None = None)[source]#
-This class capsules endpoints provided within deepmind/hanabi-learning-environment/rl_env.py.
-Initializes the raw_env class.
+class pettingzoo.classic.hanabi.hanabi.raw_env(colors: int = 5, ranks: int = 5, players: int = 2, hand_size: int = 5, max_information_tokens: int = 8, max_life_tokens: int = 3, observation_type: str = 'card_knowledge', random_start_player: bool = False, render_mode: str | None = None)[source]#
+Initializes the raw_env class.
- Parameter descriptions :
colors: int, Number of colors in [2,5].
API#
max_information_tokens: int, Number of information tokens (>=0).
max_life_tokens: int, Number of life tokens (>=1).
-- observation_type: int.
0: Minimal observation.
-1: First-order common knowledge observation.
+- observation_type: str.
“minimal””: Minimal observation.
+“card_knowledge”: First-order common knowledge observation.
+“seer”: Full information of all cards.
API#
“max_information_tokens”: 8,
“max_life_tokens”: 3,
“hand_size”: (4 if players >= 4 else 5)
-“observation_type”: 1,
-“hand_size”: 2
+“observation_type”: “card_knowledge”,
+“hand_size”: 2,
}
- Hanabi-Small{
“colors”: 2,
“ranks”: 5,
“players”: 2,
-“max_information_tokens”: 3
+“max_information_tokens”: 3,
“hand_size”: 2,
-“max_life_tokens”: 1
-“observation_type”: 1}
+“max_life_tokens”: 1,
+“observation_type”: “card_knowledge”,
+}
- Hanabi-Very-Small{
“colors”: 1,
“ranks”: 5,
“players”: 2,
-“max_information_tokens”: 3
+“max_information_tokens”: 3,
“hand_size”: 2,
-“max_life_tokens”: 1
-“observation_type”: 1}
+“max_life_tokens”: 1,
+“observation_type”: “card_knowledge”,
+}