def play_with_MCTS(game, black_strategy, white_strategy):
game.restart()
strategy = lambda who: black_strategy if who == utils.BLACK else white_strategy
while game.player is not None:
game_strategy = strategy(game.player)
oppo_strategy = strategy(Othello.opponent(game.player))
if isinstance(game_strategy, MCTS):
move = game_strategy.suggest_move(game)
game_strategy.shift_root(move)
else:
move = Othello.get_move(game_strategy, game.player, game.board)
if isinstance(oppo_strategy, MCTS):
oppo_strategy.shift_root(move)
game.make_valid_move(move)
# print move
# print Othello.print_board(game.board)
# game.player = Othello.next_player(game.board, game.player)
# raw_input()
# Store the value for value network training
# if isinstance(black_strategy, MCTS):
# black_strategy.store_value()
# if isinstance(white_strategy, MCTS):
# white_strategy.store_value()
return game.board, Othello.score(utils.BLACK, game.board)
def evaluate(self, game):
# player = utils.PIECES.index(player) # black-1, white-0
data = np.empty((1, 10, 8, 8), dtype="int8")
player = game.player
board = game.board
validmoves = Othello.legal_moves(player, board)
arr = np.zeros((10, 8, 8), dtype="int8")
arr[9] = np.ones((8, 8), dtype="int8")
for i in xrange(8):
for j in xrange(8):
square = i + j * 10 + 11
if square in validmoves:
arr[4][i][j] = 1
if board[square] == utils.EMPTY:
arr[3][i][j] = 1
else:
if board[square] == player:
arr[1][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[2][i][j] = 1
if self.check_adjacent(square, board):
if board[square] == player:
arr[5][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[7][i][j] = 1
else:
if board[square] == player:
arr[6][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[8][i][j] = 1
data[0] = arr
prob = self.model.predict(data, batch_size=1)[0]
prob = np.reshape(prob, 64)
# softmax temperature manipulate
prob_sum = 0
for i in validmoves:
prob[LABEL_MAP.index(i)] = math.exp(prob[LABEL_MAP.index(i)] / T)
prob_sum += prob[LABEL_MAP.index(i)]
move_probs = list(
(i, prob[LABEL_MAP.index(i)] / prob_sum) for i in validmoves)
return move_probs
def evaluate(self, game):
player = game.player
board = game.board
validmoves = Othello.legal_moves(player, board)
if len(validmoves) == 0:
return []
prob = 1.0 / len(validmoves)
random.shuffle(validmoves)
move_probs = list((i, prob) for i in validmoves)
return move_probs
def get_move(player, board):
data = np.empty((1, 10, 8, 8), dtype="int8")
validmoves = Othello.legal_moves(player, board)
arr = np.zeros((10, 8, 8), dtype="int8")
arr[9] = np.ones((8, 8), dtype="int8")
for i in xrange(8):
for j in xrange(8):
square = i + j * 10 + 11
if square in validmoves:
arr[4][i][j] = 1
if board[square] == utils.EMPTY:
arr[3][i][j] = 1
else:
if board[square] == player:
arr[1][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[2][i][j] = 1
if check_adjacent(square, board):
if board[square] == player:
arr[5][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[7][i][j] = 1
else:
if board[square] == player:
arr[6][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[8][i][j] = 1
data[0] = arr
# print utils.print_board(board)
# print arr
# raw_input()
prob = model.predict(data, batch_size=1)[0]
# prob = np.reshape(prob, 64)
validpred = np.asarray(
list([i, prob[LABEL_MAP.index(i)]] for i in validmoves))
move = validpred[np.argmax(validpred[:, 1])][0]
return int(move)
def evaluate(self, game):
# player = utils.PIECES.index(player) # black-1, white-0
data = np.empty((1, 10, 8, 8), dtype="int8")
player = game.player
board = game.board
validmoves = Othello.legal_moves(player, board)
arr = np.zeros((10, 8, 8), dtype="int8")
arr[9] = np.ones((8, 8), dtype="int8")
for i in xrange(8):
for j in xrange(8):
square = i + j*10 + 11
if square in validmoves:
arr[4][i][j] = 1
if board[square] == utils.EMPTY:
arr[3][i][j] = 1
else:
if board[square] == player:
arr[1][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[2][i][j] = 1
if self.check_adjacent(square, board):
if board[square] == player:
arr[5][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[7][i][j] = 1
else:
if board[square] == player:
arr[6][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[8][i][j] = 1
data[0] = arr
score = self.model.predict(data, batch_size=1)[0]
return score
def store_value(self, dest=DEST_DIR):
filename = time.strftime('mcts_result_%Y%m%d%H%M%S.txt')
try:
with open(dest + filename, 'w') as f:
expansion_list = []
for i in self.root.children:
if self.root.children[i].is_expanded():
expansion_list.append(self.root.children[i])
while len(expansion_list) > 0:
temp_node = expansion_list.pop(0)
if temp_node.game.player is None:
continue
for i in temp_node.children:
if temp_node.children[i].is_expanded():
expansion_list.append(temp_node.children[i])
f.write(
Othello.board_to_str(temp_node.game.board,
temp_node.game.player) + ' ' +
str(temp_node.Q) + ' ' + str(temp_node.N) + '\n')
except:
print "Error! Cannot log file: " + filename
def get_generate_data(self):
print("start")
files = os.listdir(self.dirname)
length = 0
for filename in files:
length += sum(1
for line in open(self.dirname + "/" + filename, "r"))
length *= 4
print length
data = np.empty((length, 10, 8, 8), dtype="int8")
label = np.empty(length, dtype="float32, int32")
data_id = 0
for filename in files:
with open(self.dirname + "/" + filename, "r") as f:
for line in f:
# Read the board, player and Q/N from the file
board, player = Othello.str_to_board(line[0:102])
values = line[103:].split(' ')
Q = float(values[0])
N = int(values[1])
# Process the board into training data
validmoves = Othello.legal_moves(player, board)
arr = np.zeros((10, 8, 8), dtype="int8")
arr[9] = np.ones((8, 8), dtype="int8")
for i in xrange(8):
for j in xrange(8):
square = i + j * 10 + 11
if square in validmoves:
arr[4][i][j] = 1
if board[square] == utils.EMPTY:
arr[3][i][j] = 1
else:
if board[square] == player:
arr[1][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[2][i][j] = 1
if check_adjacent(square, board):
if board[square] == player:
arr[5][i][j] = 1
elif board[square] == Othello.opponent(
player):
arr[7][i][j] = 1
else:
if board[square] == player:
arr[6][i][j] = 1
elif board[square] == Othello.opponent(
player):
arr[8][i][j] = 1
data[data_id, :, :, :] = arr
data[data_id + 1, :, :, :] = np.array(
list(m.transpose() for m in arr))
arr = arr[:, ::-1, ::-1]
data[data_id + 2, :, :, :] = arr
data[data_id + 3, :, :, :] = np.array(
list(m.transpose() for m in arr))
label[data_id:data_id + 4] = [(Q, N)] * 4
# progress report
percentage = int((float(data_id + 1) / (length)) * 30)
progress_bar = '[' + '=' * percentage + '>' + '.' * (
29 - percentage) + ']'
sys.stdout.write(' ' + str(data_id + 1) + '/' +
str(length) + ' ' + progress_bar + '\r')
sys.stdout.flush()
data_id += 4
#Map
print("\nMap:")
bin_board_dict = {}
for i in xrange(length):
key = int(
''.join(
list(
str(num) for own_stone_list in data[i][1].tolist()
for num in own_stone_list)) + ''.join(
list(
str(num)
for oppo_stone_list in data[i][2].tolist()
for num in oppo_stone_list)), 2)
if label[i][1] > 10:
if key not in bin_board_dict:
bin_board_dict[key] = []
bin_board_dict[key].append(i)
# progress report
percentage = int((float(i + 1) / (length)) * 30)
progress_bar = '[' + '=' * percentage + '>' + '.' * (
29 - percentage) + ']'
sys.stdout.write(' ' + str(i + 1) + '/' + str(length) + ' ' +
progress_bar + '\r')
sys.stdout.flush()
#Reduce
print("\nReduce")
uni_length = len(bin_board_dict)
uni_data = np.empty((uni_length, 10, 8, 8), dtype="int8")
uni_label = np.zeros(uni_length, dtype="float32")
uni_data_id = 0
print(uni_length)
for board in bin_board_dict:
uni_data[uni_data_id] = data[bin_board_dict[board][0]]
uni_label[uni_data_id] = sum(
inner_product(label[i]) for i in bin_board_dict[board]) / sum(
label[i][1] for i in bin_board_dict[board])
# progress report
percentage = int((float(uni_data_id + 1) / (uni_length)) * 30)
progress_bar = '[' + '=' * percentage + '>' + '.' * (
29 - percentage) + ']'
sys.stdout.write(' ' + str(uni_data_id + 1) + '/' +
str(uni_length) + ' ' + progress_bar + '\r')
sys.stdout.flush()
uni_data_id += 1
print("\nOver!")
print("Total sample count: " + str(len(uni_data)))
return uni_data, uni_label
def __init__(self, heuristic):
self.heuristic = heuristic
self.engine = Othello()
def get_generate_data(self):
print ("start")
files = os.listdir(self.dirname)
length = 0
for filename in files:
# if sum(1 for line in open(self.dirname + "/" + filename, "r")) < self.omit_lines:
# print "ERROR\n",filename
length += sum(1 for line in open(self.dirname + "/" + filename, "r")) - self.omit_lines
length*=4
print length
data = np.empty((length, 10, 8, 8), dtype="int8")
label = np.empty((length), dtype="int8")
data_id = 0
for filename in files:
with open(self.dirname + "/" + filename, "r") as f:
self.game.restart()
line_count = 0
for line in f:
line_count += 1
currentplayer = utils.PIECES[int(line[0])]
x = int(line[2])
y = int(line[4])
# score = float(line[6:])
if self.game.player == currentplayer:
if line_count > self.omit_lines:
player = self.game.player
board = self.game.board
validmoves = Othello.legal_moves(player, board)
arr = np.zeros((10, 8, 8), dtype="int8")
arr[9] = np.ones((8, 8), dtype="int8")
for i in xrange(8):
for j in xrange(8):
square = i + j*10 + 11
if square in validmoves:
arr[4][i][j] = 1
if board[square] == utils.EMPTY:
arr[3][i][j] = 1
else:
if board[square] == player:
arr[1][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[2][i][j] = 1
if check_adjacent(square, board):
if board[square] == player:
arr[5][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[7][i][j] = 1
else:
if board[square] == player:
arr[6][i][j] = 1
elif board[square] == Othello.opponent(player):
arr[8][i][j] = 1
data[data_id, :, :, :] = arr
data[data_id+1, :, :, :] = np.array(list(m.transpose() for m in arr))
arr = arr[:,::-1,::-1]
data[data_id+2, :, :, :] = arr
data[data_id+3, :, :, :] = np.array(list(m.transpose() for m in arr))
label[data_id] = LABEL_MAP[x][y]
label[data_id+1] = LABEL_MAP[y][x]
label[data_id+2] = LABEL_MAP[7-x][7-y]
label[data_id+3] = LABEL_MAP[7-y][7-x]
# progress report
percentage = int((float(data_id+1) / (length)) * 30)
progress_bar = '[' + '='*percentage + '>' + '.'*(29-percentage) + ']'
sys.stdout.write(' ' + str(data_id+1) + '/' + str(length) + ' ' + progress_bar + '\r')
sys.stdout.flush()
data_id += 4
pos = x + y*10 + 11
self.game.make_valid_move(pos)
#Map
print ("\nMap:")
bin_board_dict = {}
for i in xrange(length):
key = int(
''.join(list(str(num) for own_stone_list in data[i][1].tolist() for num in own_stone_list))
+ ''.join(list(str(num) for oppo_stone_list in data[i][2].tolist() for num in oppo_stone_list))
+ bin(label[i])[2:], 2)
if key not in bin_board_dict:
bin_board_dict[key] = i
# progress report
percentage = int((float(i+1) / (length)) * 30)
progress_bar = '[' + '='*percentage + '>' + '.'*(29-percentage) + ']'
sys.stdout.write(' ' + str(i+1) + '/' + str(length) + ' ' + progress_bar + '\r')
sys.stdout.flush()
#Reduce
print ("\nReduce")
uni_length = len(bin_board_dict)
uni_data = np.empty((uni_length, 10, 8, 8), dtype="int8")
uni_label = np.zeros(uni_length, dtype="int8")
uni_data_id = 0
print(uni_length)
for board in bin_board_dict:
uni_data[uni_data_id] = data[bin_board_dict[board]]
uni_label[uni_data_id] = label[bin_board_dict[board]]
# progress report
percentage = int((float(uni_data_id+1) / (uni_length)) * 30)
progress_bar = '[' + '='*percentage + '>' + '.'*(29-percentage) + ']'
sys.stdout.write(' ' + str(uni_data_id+1) + '/' + str(uni_length) + ' ' + progress_bar + '\r')
sys.stdout.flush()
uni_data_id += 1
print("\nOver!")
print("Total sample count: " + str(len(uni_data)))
return uni_data, uni_label
def __init__(self, dirname, omit_lines=8):
self.dirname = dirname
self.game = Othello()
self.omit_lines = omit_lines
def createGrid(self):
def btn_restart_game():
if self.playing:
self.packAndSend({
'to': self.opponent['client_id'],
'action': 'restart',
'value': 'apply'
})
else:
QMessageBox.information(self, '系统消息', "没有加入任何游戏")
self.restartGame()
def btn_leave_game():
if self.playing:
self.packAndSend({
'to': self.opponent['client_id'],
'action': 'leave',
'value': ''
})
self.restartGame()
else:
QMessageBox.information(self, '系统消息', "没有加入任何游戏")
#设置主窗口的大小与位置
screen = QDesktopWidget().screenGeometry()
self.setGeometry((screen.width() - UI_WIDTH) / 2,
(screen.height() - UI_HEIGHT) / 2, UI_WIDTH,
UI_HEIGHT)
self.setFixedSize(UI_WIDTH, UI_HEIGHT)
#创建布局
main = QWidget(self)
main.setObjectName('main')
self.setCentralWidget(main)
self.othello = Othello()
self.board = Board(main, self.othello)
self.hall = Hall(self)
#创建应该下子的图片标志，初始化为黑子
pic = QLabel(self)
pic.setGeometry(UI_WIDTH * 0.212, UI_HEIGHT * 0.12, 50, 50)
pic.setObjectName('label_for_turn_image')
#创建功能按钮
buttons = QWidget(self)
buttons.setGeometry(UI_WIDTH * 0.77, UI_HEIGHT * 0.4, UI_WIDTH * 0.2,
UI_HEIGHT * 0.12)
buttons.setObjectName('buttons')
btn_restart = QPushButton('重新开始', self)
btn_list = QPushButton('玩家列表', self)
btn_leave = QPushButton('离开', self)
btn_hall = QPushButton('大厅', self)
grid = QGridLayout()
grid.addWidget(btn_restart, 0, 0)
grid.addWidget(btn_list, 0, 1)
grid.addWidget(btn_leave, 1, 0)
grid.addWidget(btn_hall, 1, 1)
buttons.setLayout(grid)
btn_restart.clicked.connect(btn_restart_game)
btn_list.clicked.connect(lambda: self.packAndSend({
'to': 'admin',
'action': 'list',
'value': '玩家列表'
}))
btn_hall.clicked.connect(lambda: self.packAndSend({
'to': 'admin',
'action': 'games',
'value': '列出大厅'
}))
btn_leave.clicked.connect(btn_leave_game)
#创建分数显示屏,包括黑棋得分，黑棋昵称，白棋得分，白棋昵称
self.score_black = QLabel(self)
self.score_black.setObjectName('score_black')
self.score_black.setGeometry(UI_WIDTH * 0.08, UI_HEIGHT * 0.25, 100,
60)
self.nickname_black = QLabel("<p style='text-align:center'>黑棋</p>",
self)
self.nickname_black.setObjectName('nickname_black')
self.nickname_black.setGeometry(UI_WIDTH * 0.03, UI_HEIGHT * 0.36, 120,
60)
self.score_white = QLabel(self)
self.score_white.setObjectName('score_white')
self.score_white.setGeometry(UI_WIDTH * 0.08, UI_HEIGHT * 0.57, 100,
60)
self.nickname_white = QLabel("<p style='text-align:center'>白棋</p>",
self)
self.nickname_white.setObjectName('nickname_white')
self.nickname_white.setGeometry(UI_WIDTH * 0.03, UI_HEIGHT * 0.68, 120,
60)
self.move_record = QTextEdit("当前下子情况", self)
self.move_record.setReadOnly(True)
#显示走棋的位置
self.move_record = QTextEdit("下子情况", self)
self.move_record.setObjectName("move_record")
self.move_record.setReadOnly(True)
self.move_record.setGeometry(UI_WIDTH * 0.765, UI_HEIGHT * 0.24,
UI_WIDTH * 0.215, UI_HEIGHT * 0.15)
#创建输入框
chat = QWidget(self)
chat.setObjectName('chat')
chat.setGeometry(UI_WIDTH * 0.755, UI_HEIGHT * 0.54, UI_WIDTH * 0.235,
UI_HEIGHT * 0.45)
grid = QGridLayout()
chat.setLayout(grid)
# 聊天窗口
self.messages = QTextEdit("欢饮来到呆尐兔兔五子棋,撒花~~~", self)
self.messages.setReadOnly(True)
self.edit = QLineEdit(self)
btn_send_msg = QPushButton("Enter/发送", self)
btn_send_msg.clicked.connect(self.chat)
grid.addWidget(self.messages, 0, 0, 10, 10)
grid.addWidget(self.edit, 10, 0, 2, 8)
grid.addWidget(btn_send_msg, 10, 8, 2, 2)
def use_random(self, game):
game.make_valid_move(
random.choice(Othello.legal_moves(game.player, game.board)))
return game