def test_encode_cards(self):
plane = np.zeros((5, 15), dtype=int)
plane[0] = np.ones(15, dtype=int)
cards = '333BR'
encode_cards(plane, cards)
self.assertEqual(plane[3][0], 1)
self.assertEqual(plane[1][13], 1)
self.assertEqual(plane[1][14], 1)
def extract_state(self, state):
''' Encode state
Args:
state (dict): dict of original state
Returns:
numpy array: 6*5*15 array
6 : current hand
the union of the other two players' hand
the recent three actions
the union of all played cards
'''
encoded_state = np.zeros((6, 5, 15), dtype=int)
for index in range(6):
encoded_state[index][0] = np.ones(15, dtype=int)
encode_cards(encoded_state[0], state['current_hand'])
encode_cards(encoded_state[1], state['others_hand'])
for i, action in enumerate(state['trace'][-3:]):
if action[1] != 'pass':
encode_cards(encoded_state[4 - i], action[1])
if state['played_cards'] is not None:
encode_cards(encoded_state[5], state['played_cards'])
return encoded_state
def _extract_state(self, state):
''' Encode state
Args:
state (dict): dict of original state
Returns:
numpy array: 6*5*15 array
6 : current hand
the union of the other two players' hand
the recent three actions
the union of all played cards
'''
obs = np.zeros((6, 5, 15), dtype=int)
for index in range(6):
obs[index][0] = np.ones(15, dtype=int)
encode_cards(obs[0], state['current_hand'])
encode_cards(obs[1], state['others_hand'])
for i, action in enumerate(state['trace'][-3:]):
if action[1] != 'pass':
encode_cards(obs[4 - i], action[1])
if state['played_cards'] is not None:
encode_cards(obs[5], state['played_cards'])
extracted_state = {
'obs': obs,
'legal_actions': self._get_legal_actions()
}
if self.allow_raw_data:
extracted_state['raw_obs'] = state
# TODO: state['actions'] can be None, may have bugs
if state['actions'] == None:
extracted_state['raw_legal_actions'] = []
else:
extracted_state['raw_legal_actions'] = [
a for a in state['actions']
]
if self.record_action:
extracted_state['action_record'] = self.action_recorder
return extracted_state