def play_episode(encoder: CardPhaseStateActionEncoder, agent: AllPhaseAgent = None):
if agent is None:
agent = AllPhaseAgent()
environment = FrenchTarotEnvironment()
observation = environment.reset()
done = False
reagent_entries_per_player = []
actions_in_round = []
observations_in_round = []
while not done:
action = agent.get_action(observation)
new_observation, rewards, done, _ = environment.step(action)
if isinstance(observation, CardPhaseObservation):
actions_in_round.append(action)
observations_in_round.append(observation)
if rewards is not None:
iterator = zip(observations_in_round, actions_in_round, rewards)
for observation, action, reward in iterator:
entry = encoder.encode(observation.player.position_towards_taker, observation, action, reward)
reagent_entries_per_player.append(entry)
actions_in_round = []
observations_in_round = []
observation = new_observation
return reagent_entries_per_player
def test_deal_petit_sec():
card_list = copy.copy(CARDS)
card_list[0] = Card.TRUMP_1
card_list[-21] = Card.SPADES_1
environment = FrenchTarotEnvironment()
with pytest.raises(FrenchTarotException):
environment.reset(card_list)
def setup_environment_for_card_phase(taker=0,
shuffled_deck=None,
chelem=False,
poignee=False):
environment = FrenchTarotEnvironment()
observation = environment.reset(shuffled_card_deck=shuffled_deck)
good = False
for i in range(environment.n_players):
if i == taker:
observation = environment.step(Bid.GARDE_SANS)[0]
good = True
else:
observation = environment.step(Bid.PASS)[0]
if not good:
raise FrenchTarotException("No taking player")
announcements = []
if chelem:
announcements.append(ChelemAnnouncement())
if poignee:
announcements.append(
PoigneeAnnouncement.largest_possible_poignee_factory(
observation.player.hand[-11:-1]))
environment.step(announcements)
environment.step([])
environment.step([])
observation = environment.step([])[0]
return environment, observation
def setup_environment() -> FrenchTarotEnvironment:
environment = FrenchTarotEnvironment()
environment.reset()
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.GARDE_SANS)
return environment
def setup_environment():
environment = FrenchTarotEnvironment()
environment.reset()
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PETITE)
return environment
def test_make_dog_without_trump():
environment = FrenchTarotEnvironment()
environment.reset(CARDS)
environment.step(Bid.PETITE)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
observation = environment.step(Bid.PASS)[0]
dog = list(observation.player.hand[:6]) # taking player is always player 0
observation, reward, done, _ = environment.step(dog)
assert np.isclose(reward, 3.0)
def prepare_environment_sorted_deck():
environment = FrenchTarotEnvironment()
environment.reset(CARDS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
observation = environment.step(Bid.PETITE)[0]
return environment, observation
def test_announce_simple_poignee_valid():
environment = FrenchTarotEnvironment()
environment.reset(CARDS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
observation = environment.step(Bid.GARDE_SANS)[0]
poignee = PoigneeAnnouncement.largest_possible_poignee_factory(
observation.player.hand)
observation, reward, done, _ = environment.step([poignee])
assert reward == 0
assert not done
def test_make_dog():
environment = FrenchTarotEnvironment()
environment.reset()
environment.step(Bid.PETITE)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
observation = environment.step(Bid.PASS)[0]
taker_hand = observation.player.hand
dog = taker_hand[2:8]
observation, reward, done, _ = environment.step(dog)
assert not done
assert reward > 0
assert isinstance(observation, AnnouncementsPhaseObservation)
assert np.all([card not in observation.player.hand for card in dog])
def test_announce_chelem_player0():
environment = FrenchTarotEnvironment()
environment.reset(CARDS)
environment.step(Bid.GARDE_SANS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
announcements = [ChelemAnnouncement()]
observation, reward, done, _ = environment.step(announcements)
assert reward == 0
assert not done
def test_announce_simple_poignee_no_such_cards_in_hand():
environment = FrenchTarotEnvironment()
environment.reset(CARDS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.GARDE_SANS)
environment.step([])
environment.step([])
environment.step([])
with pytest.raises(FrenchTarotException):
card_list = [
Card.TRUMP_1, Card.TRUMP_2, Card.TRUMP_3, Card.TRUMP_4,
Card.TRUMP_5, Card.TRUMP_6, Card.TRUMP_7, Card.TRUMP_8,
Card.TRUMP_9, Card.TRUMP_21
]
poignee = PoigneeAnnouncement.largest_possible_poignee_factory(
card_list)
environment.step([poignee])
def test_announce_simple_poignee_no_trump():
environment = FrenchTarotEnvironment()
environment.reset(CARDS)
environment.step(Bid.GARDE_SANS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step([])
environment.step([])
environment.step([])
with pytest.raises(FrenchTarotException):
environment.step([SimplePoigneeAnnouncement(CARDS[:11])])
def test_announce_simple_poignee_excuse_accepted():
environment = FrenchTarotEnvironment()
deck = copy.copy(CARDS)
deck[69] = Card.EXCUSE
deck[70] = Card.SPADES_1
deck[0] = Card.TRUMP_15
deck[77] = Card.TRUMP_14
environment.reset(deck)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.GARDE_SANS)
card_list = [
Card.TRUMP_1, Card.TRUMP_2, Card.TRUMP_3, Card.TRUMP_4, Card.TRUMP_5,
Card.TRUMP_6, Card.TRUMP_7, Card.TRUMP_8, Card.TRUMP_9, Card.TRUMP_10,
Card.TRUMP_11, Card.TRUMP_12, Card.TRUMP_13, Card.TRUMP_16, Card.EXCUSE
]
environment.step(
[PoigneeAnnouncement.largest_possible_poignee_factory(card_list)])
def test_announce_simple_poignee_excuse_refused():
environment = FrenchTarotEnvironment()
deck = copy.copy(CARDS)
deck[77] = Card.TRUMP_16
deck[71] = Card.EXCUSE
environment.reset(CARDS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.GARDE_SANS)
environment.step([])
environment.step([])
observation = environment.step([])[0]
with pytest.raises(FrenchTarotException):
environment.step([observation.player.hand])
def test_announce_chelem_wrong_string():
environment = FrenchTarotEnvironment()
environment.reset(CARDS)
environment.step(Bid.GARDE_SANS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
environment.step(Bid.PASS)
with pytest.raises(FrenchTarotException):
environment.step(["test"])
def test_dog_has_wrong_number_of_cards():
environment = FrenchTarotEnvironment()
environment.reset(CARDS)
environment.step(Bid.PASS)
environment.step(Bid.PETITE)
environment.step(Bid.PASS)
observation = environment.step(Bid.PASS)[0]
dog = list(observation.player.hand[:5]) # taking player is always player 0
with pytest.raises(FrenchTarotException):
environment.step(dog)
def environment():
return FrenchTarotEnvironment()
def test_randomness_when_bidding():
random_agent = RandomAgent()
environment = FrenchTarotEnvironment()
observation = environment.reset()
actions = [random_agent.get_action(observation) for _ in range(10)]
assert len(np.unique(actions)) > 1
def test_instantiate_random_player():
random_agent = RandomAgent()
environment = FrenchTarotEnvironment()
observation = environment.reset()
action = random_agent.get_action(observation)
assert isinstance(action, Bid)
def environment():
yield FrenchTarotEnvironment()