def test_get_payoffs(self): game = Game() game.init_game() while not game.is_over(): actions = game.get_legal_actions() action = np.random.choice(actions) state, _ = game.step(action) payoffs = game.get_payoffs() total = 0 for payoff in payoffs: total += payoff self.assertEqual(total, 0)
def test_get_payoffs(self): game = Game() game.init_game() while not game.is_over(): actions = game.get_legal_actions() action = np.random.choice(actions) state, _ = game.step(action) total_cards = len(state['hand']) + len(state['others_hand']) + len( state['played_cards']) + len(game.round.dealer.deck) self.assertEqual(total_cards, 108) payoffs = game.get_payoffs() total = 0 for payoff in payoffs: total += payoff self.assertEqual(total, 0)
class UnoEnv(Env): def __init__(self, config): self.game = Game() super().__init__(config) self.state_shape = [7, 4, 15] def _load_model(self): ''' Load pretrained/rule model Returns: model (Model): A Model object ''' return models.load('uno-rule-v1') def _extract_state(self, state): obs = np.zeros((7, 4, 15), dtype=int) encode_hand(obs[:3], state['hand']) encode_target(obs[3], state['target']) encode_hand(obs[4:], state['others_hand']) legal_action_id = self._get_legal_actions() extracted_state = {'obs': obs, 'legal_actions': legal_action_id} if self.allow_raw_data: extracted_state['raw_obs'] = state extracted_state['raw_legal_actions'] = [ a for a in state['legal_actions'] ] if self.record_action: extracted_state['action_record'] = self.action_recorder return extracted_state def get_payoffs(self): return self.game.get_payoffs() def _decode_action(self, action_id): legal_ids = self._get_legal_actions() if action_id in legal_ids: return ACTION_LIST[action_id] #if (len(self.game.dealer.deck) + len(self.game.round.played_cards)) > 17: # return ACTION_LIST[60] return ACTION_LIST[np.random.choice(legal_ids)] def _get_legal_actions(self): legal_actions = self.game.get_legal_actions() legal_ids = [ACTION_SPACE[action] for action in legal_actions] return legal_ids