def get_move(self, board: Board, eps=0.0):
self.moves += 1
legal_moves = board.legal_moves()
if len(legal_moves) == 0:
return None
if random() < eps:
self.rand += 1
return sample(legal_moves, 1)[0]
max_q = -2**62
max_action = None
for move in legal_moves:
encoded = TTTQEncoding()(move, board, None, self.playerID)
q_val = self.net(encoded)[0][0]
if q_val > max_q:
max_q = q_val
max_action = move
return max_action
def get_move(self, board: Board, eps=0):
legal_moves = board.legal_moves()
if len(legal_moves) == 0:
return None
if random() < eps:
return sample(legal_moves, 1)[0]
max_v = -2**62
max_action = None
for move in legal_moves:
# simulate move:
sim_board = deepcopy(board)
sim_board.do(move, self.playerID)
encoded = TTTVEncoding()(None, sim_board, None, self.playerID)
v_val = self.net(encoded)[0][0]
if v_val > max_v:
max_v = v_val
max_action = move
return max_action
def get_move(self, board: Board):
legal = board.legal_moves(self.playerID)
if len(legal) == 0:
return None
return sample(legal, 1)[0]