class ObSelfPlay:
def __init__(self, config: Config):
self.config = config
self.env = CChessEnv()
self.model = None
self.pipe = None
self.ai = None
self.chessmans = None
def load_model(self):
self.model = CChessModel(self.config)
if self.config.opts.new or not load_best_model_weight(self.model):
self.model.build()
def start(self):
self.env.reset()
self.load_model()
self.pipe = self.model.get_pipes()
self.ai = CChessPlayer(self.config,
search_tree=defaultdict(VisitState),
pipes=self.pipe,
enable_resign=True,
debugging=False)
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
self.env.board.print_to_cl()
history = [self.env.get_state()]
while not self.env.board.is_end():
no_act = None
state = self.env.get_state()
if state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
no_act.append(history[i + 1])
action, _ = self.ai.action(state, self.env.num_halfmoves, no_act)
history.append(action)
if action is None:
print("AI投降了!")
break
move = self.env.board.make_single_record(int(action[0]),
int(action[1]),
int(action[2]),
int(action[3]))
if not self.env.red_to_move:
action = flip_move(action)
self.env.step(action)
history.append(self.env.get_state())
print(f"AI选择移动 {move}")
self.env.board.print_to_cl()
sleep(1)
self.ai.close()
print(f"胜者是 is {self.env.board.winner} !!!")
self.env.board.print_record()
class ObSelfPlay:
def __init__(self, config: Config):
self.config = config
self.env = CChessEnv()
self.model = None
self.pipe = None
self.ai = None
self.chessmans = None
def load_model(self):
self.model = CChessModel(self.config)
if self.config.opts.new or not load_best_model_weight(self.model):
self.model.build()
def start(self):
self.env.reset()
self.load_model()
self.pipe = self.model.get_pipes()
self.ai = CChessPlayer(self.config, search_tree=defaultdict(VisitState), pipes=self.pipe,
enable_resign=True, debugging=False)
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
self.env.board.print_to_cl()
history = [self.env.get_state()]
while not self.env.board.is_end():
no_act = None
state = self.env.get_state()
if state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
no_act.append(history[i + 1])
action, _ = self.ai.action(state, self.env.num_halfmoves, no_act)
history.append(action)
if action is None:
print("AI投降了!")
break
move = self.env.board.make_single_record(int(action[0]), int(action[1]), int(action[2]), int(action[3]))
if not self.env.red_to_move:
action = flip_move(action)
self.env.step(action)
history.append(self.env.get_state())
print(f"AI选择移动 {move}")
self.env.board.print_to_cl()
sleep(1)
self.ai.close()
print(f"胜者是 is {self.env.board.winner} !!!")
self.env.board.print_record()
class PlayWithHuman:
def __init__(self, config: Config):
self.config = config
self.env = CChessEnv()
self.model = None
self.pipe = None
self.ai = None
self.chessmans = None
self.human_move_first = True
def load_model(self):
self.model = CChessModel(self.config)
if self.config.opts.new or not load_best_model_weight(self.model):
self.model.build()
def start(self, human_first=True):
self.env.reset()
self.load_model()
self.pipe = self.model.get_pipes()
self.ai = CChessPlayer(self.config,
search_tree=defaultdict(VisitState),
pipes=self.pipe,
enable_resign=True,
debugging=False)
self.human_move_first = human_first
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
self.env.board.print_to_cl()
while not self.env.board.is_end():
if human_first == self.env.red_to_move:
self.env.board.calc_chessmans_moving_list()
is_correct_chessman = False
is_correct_position = False
chessman = None
while not is_correct_chessman:
title = "请输入棋子位置: "
input_chessman_pos = input(title)
x, y = int(input_chessman_pos[0]), int(
input_chessman_pos[1])
chessman = self.env.board.chessmans[x][y]
if chessman != None and chessman.is_red == self.env.board.is_red_turn:
is_correct_chessman = True
print(f"当前棋子为{chessman.name_cn}，可以落子的位置有：")
for point in chessman.moving_list:
print(point.x, point.y)
else:
print("没有找到此名字的棋子或未轮到此方走子")
while not is_correct_position:
title = "请输入落子的位置: "
input_chessman_pos = input(title)
x, y = int(input_chessman_pos[0]), int(
input_chessman_pos[1])
is_correct_position = chessman.move(x, y)
if is_correct_position:
self.env.board.print_to_cl()
self.env.board.clear_chessmans_moving_list()
else:
action, policy = self.ai.action(self.env.get_state(),
self.env.num_halfmoves)
if not self.env.red_to_move:
action = flip_move(action)
if action is None:
print("AI投降了!")
break
self.env.step(action)
print(f"AI选择移动 {action}")
self.env.board.print_to_cl()
self.ai.close()
print(f"胜者是 is {self.env.board.winner} !!!")
self.env.board.print_record()
class PlayWithHuman:
def __init__(self, config: Config):
self.config = config
self.env = CChessEnv()
self.model = None
self.pipe = None
self.ai = None
self.chessmans = None
self.human_move_first = True
def load_model(self):
self.model = CChessModel(self.config)
if self.config.opts.new or not load_best_model_weight(self.model):
self.model.build()
def start(self, human_first=True):
self.env.reset()
self.load_model()
self.pipe = self.model.get_pipes()
self.ai = CChessPlayer(self.config, search_tree=defaultdict(VisitState), pipes=self.pipe,
enable_resign=True, debugging=False)
self.human_move_first = human_first
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
self.env.board.print_to_cl()
while not self.env.board.is_end():
if human_first == self.env.red_to_move:
self.env.board.calc_chessmans_moving_list()
is_correct_chessman = False
is_correct_position = False
chessman = None
while not is_correct_chessman:
title = "请输入棋子位置: "
input_chessman_pos = input(title)
x, y = int(input_chessman_pos[0]), int(input_chessman_pos[1])
chessman = self.env.board.chessmans[x][y]
if chessman != None and chessman.is_red == self.env.board.is_red_turn:
is_correct_chessman = True
print(f"当前棋子为{chessman.name_cn}，可以落子的位置有：")
for point in chessman.moving_list:
print(point.x, point.y)
else:
print("没有找到此名字的棋子或未轮到此方走子")
while not is_correct_position:
title = "请输入落子的位置: "
input_chessman_pos = input(title)
x, y = int(input_chessman_pos[0]), int(input_chessman_pos[1])
is_correct_position = chessman.move(x, y)
if is_correct_position:
self.env.board.print_to_cl()
self.env.board.clear_chessmans_moving_list()
else:
action, policy = self.ai.action(self.env.get_state(), self.env.num_halfmoves)
if not self.env.red_to_move:
action = flip_move(action)
if action is None:
print("AI投降了!")
break
self.env.step(action)
print(f"AI选择移动 {action}")
self.env.board.print_to_cl()
self.ai.close()
print(f"胜者是 is {self.env.board.winner} !!!")
self.env.board.print_record()
class PlayWithHuman:
def __init__(self, config: Config):
self.config = config
self.env = CChessEnv()
self.model = None
self.pipe = None
self.ai = None
self.chessmans = None
self.human_move_first = True
def load_model(self):
self.model = CChessModel(self.config)
if self.config.opts.new or not load_best_model_weight(self.model):
self.model.build()
def start(self, human_first=True):
self.env.reset()
self.load_model()
self.pipe = self.model.get_pipes()
self.ai = CChessPlayer(self.config,
search_tree=defaultdict(VisitState),
pipes=self.pipe,
enable_resign=True,
debugging=False)
self.human_move_first = human_first
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
self.env.board.print_to_cl()
while not self.env.board.is_end():
if human_first == self.env.red_to_move:
self.env.board.calc_chessmans_moving_list()
is_correct_chessman = False
is_correct_position = False
chessman = None
while not is_correct_chessman:
title = "Please enter the chess piece position: "
input_chessman_pos = input(title)
print(input_chessman_pos)
x, y = int(input_chessman_pos[0]), int(
input_chessman_pos[1])
chessman = self.env.board.chessmans[x][y]
if chessman != None and chessman.is_red == self.env.board.is_red_turn:
is_correct_chessman = True
print(
f"The current chess piece is {chessman.name}，places where you can play："
)
for point in chessman.moving_list:
print(point.x, point.y)
else:
print(
"No chess piece with this name was found or it was not his turn to walk"
)
while not is_correct_position:
title = "Please enter the location of the child: "
input_chessman_pos = input(title)
x, y = int(input_chessman_pos[0]), int(
input_chessman_pos[1])
is_correct_position = chessman.move(x, y)
if is_correct_position:
self.env.board.print_to_cl()
self.env.board.clear_chessmans_moving_list()
else:
action, policy = self.ai.action(self.env.get_state(),
self.env.num_halfmoves)
if not self.env.red_to_move:
action = flip_move(action)
if action is None:
print("AI surrendered!")
break
self.env.step(action)
print(f"AI chooses to move {action}")
self.env.board.print_to_cl()
self.ai.close()
print(f"The winner is is {self.env.board.winner} !!!")
self.env.board.print_record()
class ObSelfPlayUCCI:
def __init__(self, config: Config, ai_move_first=True):
self.config = config
self.env = CChessEnv()
self.model = None
self.pipe = None
self.ai = None
self.chessmans = None
self.ai_move_first = ai_move_first
def load_model(self):
self.model = CChessModel(self.config)
if self.config.opts.new or not load_best_model_weight(self.model):
self.model.build()
def start(self):
self.env.reset()
self.load_model()
self.pipe = self.model.get_pipes()
self.ai = CChessPlayer(self.config,
search_tree=defaultdict(VisitState),
pipes=self.pipe,
enable_resign=True,
debugging=False)
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
self.env.board.print_to_cl()
history = [self.env.get_state()]
turns = 0
game_over = False
final_move = None
while not game_over:
if (self.ai_move_first
and turns % 2 == 0) or (not self.ai_move_first
and turns % 2 == 1):
start_time = time()
no_act = None
state = self.env.get_state()
if state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
act = history[i + 1]
if not self.env.red_to_move:
act = flip_move(act)
no_act.append(act)
action, _ = self.ai.action(state, self.env.num_halfmoves,
no_act)
end_time = time()
if action is None:
print("AlphaZero 投降了!")
break
move = self.env.board.make_single_record(
int(action[0]), int(action[1]), int(action[2]),
int(action[3]))
print(
f"AlphaZero 选择移动 {move}, 消耗时间 {(end_time - start_time):.2f}s"
)
if not self.env.red_to_move:
action = flip_move(action)
else:
state = self.env.get_state()
print(state)
fen = senv.state_to_fen(state, turns)
action = self.get_ucci_move(fen)
if action is None:
print("Eleeye 投降了!")
break
print(action)
if not self.env.red_to_move:
rec_action = flip_move(action)
else:
rec_action = action
move = self.env.board.make_single_record(
int(rec_action[0]), int(rec_action[1]), int(rec_action[2]),
int(rec_action[3]))
print(f"Eleeye 选择移动 {move}")
history.append(action)
self.env.step(action)
history.append(self.env.get_state())
self.env.board.print_to_cl()
turns += 1
sleep(1)
game_over, final_move = self.env.board.is_end_final_move()
print(game_over, final_move)
if final_move:
move = self.env.board.make_single_record(int(final_move[0]),
int(final_move[1]),
int(final_move[2]),
int(final_move[3]))
print(f"Final Move {move}")
if not self.env.red_to_move:
final_move = flip_move(final_move)
self.env.step(final_move)
self.env.board.print_to_cl()
self.ai.close()
print(f"胜者是 is {self.env.board.winner} !!!")
self.env.board.print_record()
def get_ucci_move(self, fen, time=3):
p = subprocess.Popen(self.config.resource.eleeye_path,
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
universal_newlines=True)
setfen = f'position fen {fen}\n'
setrandom = 'setoption randomness small\n'
cmd = 'ucci\n' + setrandom + setfen + f'go time {time * 1000}\n'
try:
out, err = p.communicate(cmd, timeout=time + 0.5)
except:
p.kill()
try:
out, err = p.communicate()
except Exception as e:
logger.error(f"{e}, cmd = {cmd}")
return self.get_ucci_move(fen, time + 1)
print(out)
lines = out.split('\n')
if lines[-2] == 'nobestmove':
return None
move = lines[-2].split(' ')[1]
if move == 'depth':
move = lines[-1].split(' ')[6]
return senv.parse_ucci_move(move)
class UCI:
def __init__(self, config: Config):
self.config = config
self.args = None
self.state = None
self.is_red_turn = None
self.player = None
self.model = None
self.pipe = None
self.is_ready = False
self.search_tree = defaultdict(VisitState)
self.remain_time = None
self.history = None
self.turns = 0
self.start_time = None
self.end_time = None
self.t = None
self.use_history = False
def main(self):
while True:
cmd = input()
logger.debug(f"CMD: {cmd}")
cmds = cmd.split(' ')
self.args = cmds[1:]
method = getattr(self, 'cmd_' + cmds[0], None)
if method != None:
method()
else:
logger.error(f"Error command: {cmd}")
def cmd_uci(self):
print('id name CCZero')
print('id author https://cczero.org')
print('id version 2.4')
print('option name gpu spin default 0 min 0 max 7')
print('option name Threads spin default 10 min 0 max 1024')
print('uciok')
sys.stdout.flush()
set_session_config(per_process_gpu_memory_fraction=1,
allow_growth=True,
device_list=self.config.opts.device_list)
self.use_history = self.load_model()
self.is_ready = True
self.turns = 0
self.remain_time = None
self.state = senv.INIT_STATE
self.history = [self.state]
self.is_red_turn = True
def cmd_ucinewgame(self):
self.state = senv.INIT_STATE
self.history = [self.state]
self.is_ready = True
self.is_red_turn = True
self.search_tree = defaultdict(VisitState)
def cmd_setoption(self):
'''
setoption name <id> [value <x>]
'''
if len(self.args) > 3:
id = self.args[1]
if id == 'gpu':
value = int(self.args[3])
self.config.opts.device_list = value
set_session_config(per_process_gpu_memory_fraction=1,
allow_growth=True,
device_list=self.config.opts.device_list)
if id == 'Threads':
value = int(self.args[3])
self.config.play.search_threads = value
def cmd_isready(self):
if self.is_ready == True:
print('readyok')
sys.stdout.flush()
logger.debug(f"is_ready = {self.is_ready}")
def cmd_position(self):
'''
position {fen <fenstring> | startpos } [moves <move1> .... <moven>]
'''
if not self.is_ready:
return
move_idx = -1
if len(self.args) > 0:
if self.args[0] == 'fen':
# init with fen string
fen = self.args[1]
try:
self.state = senv.fen_to_state(fen)
except Exception as e:
logger.error(f"cmd position error! cmd = {self.args}, {e}")
return
self.history = [self.state]
turn = self.args[2]
if turn == 'b':
self.state = senv.fliped_state(self.state)
self.is_red_turn = False
self.turns = (int(self.args[6]) - 1) * 2 + 1
else:
self.is_red_turn = True
self.turns = (int(self.args[6]) - 1) * 2
if len(self.args) > 7 and self.args[7] == 'moves':
move_idx = 8
elif self.args[0] == 'startpos':
self.state = senv.INIT_STATE
self.is_red_turn = True
self.history = [self.state]
self.turns = 0
if len(self.args) > 1 and self.args[1] == 'moves':
move_idx = 2
elif self.args[0] == 'moves':
move_idx = 1
else:
self.state = senv.INIT_STATE
self.is_red_turn = True
self.history = [self.state]
self.turns = 0
logger.debug(f"state = {self.state}")
# senv.render(self.state)
# execute moves
if move_idx != -1:
for i in range(move_idx, len(self.args)):
action = senv.parse_ucci_move(self.args[i])
if not self.is_red_turn:
action = flip_move(action)
self.history.append(action)
self.state = senv.step(self.state, action)
self.is_red_turn = not self.is_red_turn
self.turns += 1
self.history.append(self.state)
logger.debug(f"state = {self.state}")
# senv.render(self.state)
def cmd_fen(self):
self.args.insert(0, 'fen')
self.cmd_position()
def cmd_go(self):
'''
go ...
让引擎根据内置棋盘的设置和设定的搜索方式来思考，有以下搜索方式可供选择(可以多选，直接跟在go后面)：
❌(1) searchmoves <move1> .... <moven>，只让引擎在这几步中选择一步；
✅(2) wtime <x>，白方剩余时间(单位是毫秒)；
btime <x>，黑方剩余时间；
❌winc <x>，白方每步增加的时间(适用于Fischer制)；
❌binc <x>，黑方每步增加的时间；
❌movestogo <x>，还有多少回合进入下一时段(适用于时段制)；
这些选项用来设定时钟，它决定了引擎的思考时间；
❌(3) ponder，让引擎进行后台思考(即对手在用时，引擎的时钟不起作用)；
✅(4) depth <x>，指定搜索深度；
❌(5) nodes <x>，指定搜索的节点数(即分析的局面数，一般它和时间成正比)；
❌(6) mate <x>，在指定步数内只搜索杀棋；
✅(7) movetime <x>，只花规定的时间搜索；
✅(8) infinite，无限制搜索，直到杀棋。
'''
if not self.is_ready:
return
self.start_time = time()
self.t = None
depth = None
infinite = True
self.remain_time = None
self.model.close_pipes()
self.pipe = self.model.get_pipes(need_reload=False)
self.search_tree = defaultdict(VisitState)
self.player = CChessPlayer(self.config,
search_tree=self.search_tree,
pipes=self.pipe,
enable_resign=False,
debugging=True,
uci=True,
use_history=self.use_history,
side=self.turns % 2)
for i in range(len(self.args)):
if self.args[i] == 'depth':
depth = int(self.args[i + 1]) * 100
infinite = False
if self.args[i] == 'movetime' or self.args[i] == 'time':
self.remain_time = int(self.args[i + 1]) / 1000
if self.args[i] == 'infinite':
infinite = True
if self.args[i] == 'wtime':
if self.is_red_turn:
self.remain_time = int(self.args[i + 1]) / 1000
depth = 3000
infinite = False
if self.args[i] == 'btime':
if not self.is_red_turn:
self.remain_time = int(self.args[i + 1]) / 1000
depth = 3000
infinite = False
logger.debug(
f"depth = {depth}, infinite = {infinite}, remain_time = {self.remain_time}"
)
search_worker = Thread(target=self.search_action,
args=(depth, infinite))
search_worker.daemon = True
search_worker.start()
if self.remain_time:
self.t = Timer(self.remain_time - 0.01, self.cmd_stop)
self.t.start()
def cmd_stop(self):
if not self.is_ready:
return
if self.player:
no_act = None
if self.state in self.history[:-1]:
no_act = []
for i in range(len(self.history) - 1):
if self.history[i] == self.state:
no_act.append(self.history[i + 1])
action, value, depth = self.player.close_and_return_action(
self.state, self.turns, no_act)
self.player = None
self.model.close_pipes()
self.info_best_move(action, value, depth)
else:
logger.error(f"bestmove none")
def cmd_quit(self):
sys.exit()
def load_model(self, config_file=None):
use_history = True
self.model = CChessModel(self.config)
weight_path = self.config.resource.model_best_weight_path
if not config_file:
config_path = config.resource.model_best_path
use_history = False
else:
config_path = os.path.join(config.resource.model_dir, config_file)
try:
if not load_model_weight(self.model, config_path, weight_path):
self.model.build()
use_history = True
except Exception as e:
logger.info(f"Exception {e}, 重新加载权重")
return self.load_model(config_file='model_128_l1_config.json')
logger.info(f"use_history = {use_history}")
return use_history
def search_action(self, depth, infinite):
no_act = None
_, _, _, check = senv.done(self.state, need_check=True)
logger.debug(f"Check = {check}, state = {self.state}")
if not check and self.state in self.history[:-1]:
no_act = []
for i in range(len(self.history) - 1):
if self.history[i] == self.state:
if senv.will_check_or_catch(self.state,
self.history[i + 1]):
no_act.append(self.history[i + 1])
logger.debug(f"Foul: no act = {no_act}")
action, _ = self.player.action(self.state,
self.turns,
no_act=no_act,
depth=depth,
infinite=infinite,
hist=self.history)
if self.t:
self.t.cancel()
_, value = self.player.debug[self.state]
depth = self.player.done_tasks // 100
self.player.close(wait=False)
self.player = None
self.model.close_pipes()
self.info_best_move(action, value, depth)
def info_best_move(self, action, value, depth):
self.end_time = time()
if not self.is_red_turn:
value = -value
score = int(value * 1000)
duration = self.end_time - self.start_time
nps = int(depth * 100 / duration) * 1000
print(
f"info depth {depth} score {score} time {int(duration * 1000)} nps {nps}"
)
logger.debug(
f"info depth {depth} score {score} time {int((self.end_time - self.start_time) * 1000)}"
)
sys.stdout.flush()
# get ponder
state = senv.step(self.state, action)
ponder = None
if state in self.search_tree:
node = self.search_tree[state]
cnt = 0
for mov, action_state in node.a.items():
if action_state.n > cnt:
ponder = mov
cnt = action_state.n
if not self.is_red_turn:
action = flip_move(action)
action = senv.to_uci_move(action)
output = f"bestmove {action}"
if ponder:
if self.is_red_turn:
ponder = flip_move(ponder)
ponder = senv.to_uci_move(ponder)
output += f" ponder {ponder}"
print(output)
logger.debug(output)
sys.stdout.flush()
class SelfPlayWorker:
def __init__(self,
config: Config,
pipes=None,
pid=None,
use_history=False):
self.config = config
self.player = None
self.cur_pipes = pipes
self.id = pid
self.buffer = []
self.pid = os.getpid()
self.use_history = use_history
def start(self):
# logger.info("pengge pengge pengge222")
# logger.error("pengge pengge pengge111")
# logger.debug("pengge pengge pengge")
self.pid = os.getpid()
ran = self.config.play.max_processes if self.config.play.max_processes > 5 else self.config.play.max_processes * 2
sleep((self.pid % ran) * 10)
logger.error(
f"Selfplay#Start Process index = {self.id}, pid = {self.pid}")
idx = 1
self.buffer = []
search_tree = defaultdict(VisitState)
while True:
start_time = time()
search_tree = defaultdict(VisitState)
value, turns, state, store = self.start_game(idx, search_tree)
end_time = time()
logger.error(
f"Process {self.pid}-{self.id} play game {idx} time={(end_time - start_time):.1f} sec, "
f"turn={turns / 2}, winner = {value:.2f} (1 = red, -1 = black, 0 draw)"
)
if turns <= 10:
senv.render(state)
if store:
idx += 1
sleep(random())
def start_game(self, idx, search_tree):
pipes = self.cur_pipes.pop()
if not self.config.play.share_mtcs_info_in_self_play or \
idx % self.config.play.reset_mtcs_info_per_game == 0:
search_tree = defaultdict(VisitState)
if random() > self.config.play.enable_resign_rate:
enable_resign = True
else:
enable_resign = False
self.player = CChessPlayer(self.config,
search_tree=search_tree,
pipes=pipes,
enable_resign=enable_resign,
debugging=False,
use_history=self.use_history)
state = senv.INIT_STATE
history = [state]
# policys = []
value = 0
turns = 0 # even == red; odd == black
game_over = False
final_move = None
no_eat_count = 0
check = False
no_act = []
increase_temp = False
while not game_over:
start_time = time()
action, policy = self.player.action(state,
turns,
no_act,
increase_temp=increase_temp)
end_time = time()
if action is None:
logger.error(f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
# if self.config.opts.log_move:
# logger.info(f"Process{self.pid} Playing: {turns % 2}, action: {action}, time: {(end_time - start_time):.1f}s")
# logger.info(f"Process{self.pid} Playing: {turns % 2}, action: {action}, time: {(end_time - start_time):.1f}s")
history.append(action)
# policys.append(policy)
try:
state, no_eat = senv.new_step(state, action)
except Exception as e:
logger.error(f"{e}, no_act = {no_act}, policy = {policy}")
game_over = True
value = 0
break
turns += 1
if no_eat:
no_eat_count += 1
else:
no_eat_count = 0
history.append(state)
if no_eat_count >= 120 or turns / 2 >= self.config.play.max_game_length:
game_over = True
value = 0
else:
game_over, value, final_move, check = senv.done(
state, need_check=True)
if not game_over:
if not senv.has_attack_chessman(state):
logger.error(f"双方无进攻子力，作和。state = {state}")
game_over = True
value = 0
increase_temp = False
no_act = []
if not game_over and not check and state in history[:-1]:
free_move = defaultdict(int)
for i in range(len(history) - 1):
if history[i] == state:
if senv.will_check_or_catch(state, history[i + 1]):
no_act.append(history[i + 1])
elif not senv.be_catched(state, history[i + 1]):
increase_temp = True
free_move[state] += 1
if free_move[state] >= 3:
# 作和棋处理
game_over = True
value = 0
logger.error("闲着循环三次，作和棋处理")
break
if final_move:
# policy = self.build_policy(final_move, False)
history.append(final_move)
# policys.append(policy)
state = senv.step(state, final_move)
turns += 1
value = -value
history.append(state)
self.player.close()
del search_tree
del self.player
gc.collect()
if turns % 2 == 1: # balck turn
value = -value
v = value
if turns < 10:
if random() > 0.9:
store = True
else:
store = False
else:
store = True
if store:
data = [history[0]]
for i in range(turns):
k = i * 2
data.append([history[k + 1], value])
value = -value
self.save_play_data(idx, data)
self.cur_pipes.append(pipes)
self.remove_play_data()
return v, turns, state, store
def save_play_data(self, idx, data):
self.buffer += data
if not idx % self.config.play_data.nb_game_in_file == 0:
return
rc = self.config.resource
utc_dt = datetime.utcnow().replace(tzinfo=timezone.utc)
bj_dt = utc_dt.astimezone(timezone(timedelta(hours=8)))
game_id = bj_dt.strftime("%Y%m%d-%H%M%S.%f")
filename = rc.play_data_filename_tmpl % game_id
path = os.path.join(rc.play_data_dir, filename)
logger.error(f"Process {self.pid} save play data to {path}")
write_game_data_to_file(path, self.buffer)
if self.config.internet.distributed:
upload_worker = Thread(target=self.upload_play_data,
args=(path, filename),
name="upload_worker")
upload_worker.daemon = True
upload_worker.start()
self.buffer = []
def upload_play_data(self, path, filename):
digest = CChessModel.fetch_digest(
self.config.resource.model_best_weight_path)
data = {
'digest': digest,
'username': self.config.internet.username,
'version': '2.4'
}
response = upload_file(self.config.internet.upload_url,
path,
filename,
data,
rm=False)
if response is not None and response['status'] == 0:
logger.error(f"Upload play data {filename} finished.")
else:
logger.error(
f'Upload play data {filename} failed. {response.msg if response is not None else None}'
)
def remove_play_data(self):
files = get_game_data_filenames(self.config.resource)
if len(files) < self.config.play_data.max_file_num:
return
try:
for i in range(len(files) - self.config.play_data.max_file_num):
os.remove(files[i])
except:
pass
def build_policy(self, action, flip):
labels_n = len(ActionLabelsRed)
move_lookup = {
move: i
for move, i in zip(ActionLabelsRed, range(labels_n))
}
policy = np.zeros(labels_n)
policy[move_lookup[action]] = 1
if flip:
policy = flip_policy(policy)
return list(policy)
class EvaluateWorker:
def __init__(self, config: Config, pipes1=None, pipes2=None, pid=None, data=None, hist_base=True, hist_ng=True):
self.config = config
self.player_bt = None
self.player_ng = None
self.pid = pid
self.pipes_bt = pipes1
self.pipes_ng = pipes2
self.data = data
self.hist_base = hist_base
self.hist_ng = hist_ng
def start(self):
sleep((self.pid % self.config.play.max_processes) * 10)
logger.debug(f"Evaluate#Start Process index = {self.pid}, pid = {os.getpid()}")
need_evaluate = True
self.config.opts.evaluate = True
while need_evaluate:
idx = 0 if random() > 0.5 else 1
start_time = time()
value, turns, data = self.start_game(idx)
end_time = time()
if (value == 1 and idx == 0) or (value == -1 and idx == 1):
result = '基准模型胜'
elif (value == 1 and idx == 1) or (value == -1 and idx == 0):
result = '待评测模型胜'
else:
result = '和棋'
if value == -1: # loss
score = 0
elif value == 1: # win
score = 1
else:
score = 0.5
if idx == 0:
score = 1 - score
else:
score = score
logger.info(f"进程{self.pid}评测完毕 用时{(end_time - start_time):.1f}秒, "
f"{turns / 2}回合, {result}, 得分：{score}, value = {value}, idx = {idx}")
response = self.save_play_data(idx, data, value, score)
if response and int(response['status']) == 0:
logger.info('评测结果上传成功！')
else:
logger.info(f"评测结果上传失败，服务器返回{response}")
response = http_request(self.config.internet.get_evaluate_model_url)
if int(response['status']) == 0 and response['data']['base']['digest'] == self.data['base']['digest']\
and response['data']['unchecked']['digest'] == self.data['unchecked']['digest']:
need_evaluate = True
logger.info(f"进程{self.pid}继续评测")
else:
need_evaluate = False
logger.info(f"进程{self.pid}终止评测")
def start_game(self, idx):
sleep(random())
playouts = randint(8, 12) * 100
self.config.play.simulation_num_per_move = playouts
logger.info(f"Set playouts = {self.config.play.simulation_num_per_move}")
pipe1 = self.pipes_bt.pop()
pipe2 = self.pipes_ng.pop()
search_tree1 = defaultdict(VisitState)
search_tree2 = defaultdict(VisitState)
self.player1 = CChessPlayer(self.config, search_tree=search_tree1, pipes=pipe1,
debugging=False, enable_resign=False, use_history=self.hist_base)
self.player2 = CChessPlayer(self.config, search_tree=search_tree2, pipes=pipe2,
debugging=False, enable_resign=False, use_history=self.hist_ng)
# even: bst = red, ng = black; odd: bst = black, ng = red
if idx % 2 == 0:
red = self.player1
black = self.player2
logger.info(f"进程id = {self.pid} 基准模型执红，待评测模型执黑")
else:
red = self.player2
black = self.player1
logger.info(f"进程id = {self.pid} 待评测模型执红，基准模型执黑")
state = senv.INIT_STATE
history = [state]
value = 0 # best model's value
turns = 0 # even == red; odd == black
game_over = False
no_eat_count = 0
check = False
increase_temp = False
no_act = []
while not game_over:
start_time = time()
if turns % 2 == 0:
action, _ = red.action(state, turns, no_act=no_act, increase_temp=increase_temp)
else:
action, _ = black.action(state, turns, no_act=no_act, increase_temp=increase_temp)
end_time = time()
if self.config.opts.log_move:
logger.debug(f"进程id = {self.pid}, action = {action}, turns = {turns}, time = {(end_time-start_time):.1f}")
if action is None:
logger.debug(f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
history.append(action)
state, no_eat = senv.new_step(state, action)
turns += 1
if no_eat:
no_eat_count += 1
else:
no_eat_count = 0
history.append(state)
if no_eat_count >= 120 or turns / 2 >= self.config.play.max_game_length:
game_over = True
value = 0
else:
game_over, value, final_move, check = senv.done(state, need_check=True)
no_act = []
increase_temp = False
if not game_over:
if not senv.has_attack_chessman(state):
logger.info(f"双方无进攻子力，作和。state = {state}")
game_over = True
value = 0
if not game_over and not check and state in history[:-1]:
free_move = defaultdict(int)
for i in range(len(history) - 1):
if history[i] == state:
if senv.will_check_or_catch(state, history[i+1]):
no_act.append(history[i + 1])
elif not senv.be_catched(state, history[i+1]):
increase_temp = True
free_move[state] += 1
if free_move[state] >= 3:
# 作和棋处理
game_over = True
value = 0
logger.info("闲着循环三次，作和棋处理")
break
if final_move:
history.append(final_move)
state = senv.step(state, final_move)
turns += 1
value = - value
history.append(state)
data = []
if idx % 2 == 0:
data = [self.data['base']['digest'], self.data['unchecked']['digest']]
else:
data = [self.data['unchecked']['digest'], self.data['base']['digest']]
self.player1.close()
self.player2.close()
if turns % 2 == 1: # black turn
value = -value
v = value
data.append(history[0])
for i in range(turns):
k = i * 2
data.append([history[k + 1], v])
v = -v
self.pipes_bt.append(pipe1)
self.pipes_ng.append(pipe2)
return value, turns, data
def save_play_data(self, idx, data, value, score):
rc = self.config.resource
game_id = datetime.now().strftime("%Y%m%d-%H%M%S.%f")
filename = rc.play_data_filename_tmpl % game_id
path = os.path.join(rc.play_data_dir, filename)
logger.info(f"Process {self.pid} save play data to {path}")
write_game_data_to_file(path, data)
logger.info(f"Uploading play data {filename} ...")
red, black = data[0], data[1]
return self.upload_eval_data(path, filename, red, black, value, score)
def upload_eval_data(self, path, filename, red, black, result, score):
hash = self.fetch_digest(path)
data = {'digest': self.data['unchecked']['digest'], 'red_digest': red, 'black_digest': black,
'result': result, 'score': score, 'hash': hash}
response = upload_file(self.config.internet.upload_eval_url, path, filename, data, rm=False)
return response
def fetch_digest(self, file_path):
if os.path.exists(file_path):
m = hashlib.sha256()
with open(file_path, "rb") as f:
m.update(f.read())
return m.hexdigest()
return None
class SelfPlayWorker:
def __init__(self, config: Config, pipes=None, pid=None):
self.config = config
self.player = None
self.cur_pipes = pipes
self.id = pid
self.buffer = []
self.pid = os.getpid()
def start(self):
logger.debug(f"Selfplay#Start Process index = {self.id}, pid = {self.pid}")
idx = 1
self.buffer = []
search_tree = defaultdict(VisitState)
while True:
start_time = time()
value, turns, state, search_tree, store = self.start_game(idx, search_tree)
end_time = time()
logger.debug(f"Process {self.pid}-{self.id} play game {idx} time={(end_time - start_time):.1f} sec, "
f"turn={turns / 2}, winner = {value:.2f} (1 = red, -1 = black, 0 draw)")
if turns <= 10:
senv.render(state)
if store:
idx += 1
def start_game(self, idx, search_tree):
pipes = self.cur_pipes.pop()
if not self.config.play.share_mtcs_info_in_self_play or \
idx % self.config.play.reset_mtcs_info_per_game == 0:
search_tree = defaultdict(VisitState)
if random() > self.config.play.enable_resign_rate:
enable_resign = True
else:
enable_resign = False
self.player = CChessPlayer(self.config, search_tree=search_tree, pipes=pipes, enable_resign=enable_resign, debugging=False)
state = senv.INIT_STATE
history = [state]
policys = []
value = 0
turns = 0 # even == red; odd == black
game_over = False
final_move = None
while not game_over:
no_act = None
if state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
no_act.append(history[i + 1])
start_time = time()
action, policy = self.player.action(state, turns, no_act)
end_time = time()
if action is None:
logger.debug(f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
# logger.debug(f"Process{self.pid} Playing: {turns % 2}, action: {action}, time: {(end_time - start_time):.1f}s")
# for move, action_state in self.player.search_results.items():
# if action_state[0] >= 20:
# logger.info(f"move: {move}, prob: {action_state[0]}, Q_value: {action_state[1]:.2f}, Prior: {action_state[2]:.3f}")
# self.player.search_results = {}
history.append(action)
policys.append(policy)
state = senv.step(state, action)
turns += 1
history.append(state)
if turns / 2 >= self.config.play.max_game_length:
game_over = True
value = senv.evaluate(state)
else:
game_over, value, final_move = senv.done(state)
if final_move:
policy = self.build_policy(final_move, False)
history.append(final_move)
policys.append(policy)
state = senv.step(state, final_move)
history.append(state)
self.player.close()
if turns % 2 == 1: # balck turn
value = -value
v = value
if v == 0:
if random() > 0.5:
store = True
else:
store = False
else:
store = True
if store:
data = []
for i in range(turns):
k = i * 2
data.append([history[k], policys[i], value])
value = -value
self.save_play_data(idx, data)
self.cur_pipes.append(pipes)
self.remove_play_data()
return v, turns, state, search_tree, store
def save_play_data(self, idx, data):
self.buffer += data
if not idx % self.config.play_data.nb_game_in_file == 0:
return
rc = self.config.resource
game_id = datetime.now().strftime("%Y%m%d-%H%M%S.%f")
filename = rc.play_data_filename_tmpl % game_id
path = os.path.join(rc.play_data_dir, filename)
logger.info(f"Process {self.pid} save play data to {path}")
write_game_data_to_file(path, self.buffer)
if self.config.internet.distributed:
upload_worker = Thread(target=self.upload_play_data, args=(path, filename), name="upload_worker")
upload_worker.daemon = True
upload_worker.start()
self.buffer = []
def upload_play_data(self, path, filename):
digest = CChessModel.fetch_digest(self.config.resource.model_best_weight_path)
data = {'digest': digest, 'username': self.config.internet.username}
response = upload_file(self.config.internet.upload_url, path, filename, data, rm=False)
if response is not None and response['status'] == 0:
logger.info(f"Upload play data {filename} finished.")
else:
logger.error(f'Upload play data {filename} failed. {response.msg if response is not None else None}')
def remove_play_data(self):
files = get_game_data_filenames(self.config.resource)
if len(files) < self.config.play_data.max_file_num:
return
try:
for i in range(len(files) - self.config.play_data.max_file_num):
os.remove(files[i])
except:
pass
def build_policy(self, action, flip):
labels_n = len(ActionLabelsRed)
move_lookup = {move: i for move, i in zip(ActionLabelsRed, range(labels_n))}
policy = np.zeros(labels_n)
policy[move_lookup[action]] = 1
if flip:
policy = flip_policy(policy)
return list(policy)
class SelfPlayWorker:
def __init__(self, config: Config, pipes=None, pid=None):
self.config = config
self.player = None
self.cur_pipes = pipes
self.id = pid
self.buffer = []
self.pid = os.getpid()
def start(self):
self.pid = os.getpid()
logger.debug(
f"Selfplay#Start Process index = {self.id}, pid = {self.pid}")
idx = 1
self.buffer = []
while True:
search_tree = defaultdict(VisitState)
start_time = time()
value, turns, state, store = self.start_game(idx, search_tree)
end_time = time()
if value != 1 and value != -1:
winner = 'Draw'
elif idx % 2 == 0 and value == 1 or idx % 2 == 1 and value == -1:
winner = 'AlphaHe'
else:
winner = 'Eleeye'
logger.debug(
f"Process {self.pid}-{self.id} play game {idx} time={(end_time - start_time):.1f} sec, "
f"turn={turns / 2}, value = {value:.2f}, winner is {winner}")
if turns <= 10 and store:
senv.render(state)
if store:
idx += 1
def start_game(self, idx, search_tree):
pipes = self.cur_pipes.pop()
if not self.config.play.share_mtcs_info_in_self_play or \
idx % self.config.play.reset_mtcs_info_per_game == 0:
search_tree = defaultdict(VisitState)
if random() > self.config.play.enable_resign_rate:
enable_resign = True
else:
enable_resign = False
self.player = CChessPlayer(self.config,
search_tree=search_tree,
pipes=pipes,
enable_resign=enable_resign,
debugging=False)
state = senv.INIT_STATE
history = [state]
value = 0
turns = 0 # even == red; odd == black
game_over = False
is_alpha_red = True if idx % 2 == 0 else False
final_move = None
check = False
while not game_over:
if (is_alpha_red and turns % 2 == 0) or (not is_alpha_red
and turns % 2 == 1):
no_act = None
if not check and state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
no_act.append(history[i + 1])
action, _ = self.player.action(state, turns, no_act)
if action is None:
logger.debug(
f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
else:
fen = senv.state_to_fen(state, turns)
action = self.get_ucci_move(fen)
if action is None:
logger.debug(
f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
if turns % 2 == 1:
action = flip_move(action)
history.append(action)
state = senv.step(state, action)
turns += 1
history.append(state)
if turns / 2 >= self.config.play.max_game_length:
game_over = True
value = 0
else:
game_over, value, final_move, check = senv.done(
state, need_check=True)
if final_move:
history.append(final_move)
state = senv.step(state, final_move)
history.append(state)
turns += 1
value = -value
self.player.close()
del search_tree
del self.player
gc.collect()
if turns % 2 == 1: # balck turn
value = -value
v = value
if turns <= 10:
if random() > 0.7:
store = True
else:
store = False
else:
store = True
if store:
data = [history[0]]
for i in range(turns):
k = i * 2
data.append([history[k + 1], value])
value = -value
self.save_play_data(idx, data)
self.cur_pipes.append(pipes)
self.remove_play_data()
return v, turns, state, store
def get_ucci_move(self, fen, time=3):
p = subprocess.Popen(self.config.resource.eleeye_path,
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
universal_newlines=True)
setfen = f'position fen {fen}\n'
setrandom = f'setoption randomness {self.config.opts.random}\n'
cmd = 'ucci\n' + setrandom + setfen + f'go time {time * 1000}\n'
try:
out, err = p.communicate(cmd, timeout=time + 0.5)
except subprocess.TimeoutExpired:
p.kill()
try:
out, err = p.communicate()
except Exception as e:
logger.error(f"{e}, cmd = {cmd}")
return self.get_ucci_move(fen, time + 1)
lines = out.split('\n')
if lines[-2] == 'nobestmove':
return None
move = lines[-2].split(' ')[1]
if move == 'depth':
move = lines[-1].split(' ')[6]
return senv.parse_ucci_move(move)
def save_play_data(self, idx, data):
self.buffer += data
if not idx % self.config.play_data.nb_game_in_file == 0:
return
rc = self.config.resource
game_id = datetime.now().strftime("%Y%m%d-%H%M%S.%f")
path = os.path.join(rc.play_data_dir,
rc.play_data_filename_tmpl % game_id)
logger.info(f"Process {self.pid} save play data to {path}")
write_game_data_to_file(path, self.buffer)
self.buffer = []
def remove_play_data(self):
files = get_game_data_filenames(self.config.resource)
if len(files) < self.config.play_data.max_file_num:
return
try:
for i in range(len(files) - self.config.play_data.max_file_num):
os.remove(files[i])
except:
pass
def build_policy(self, action, flip):
labels_n = len(ActionLabelsRed)
move_lookup = {
move: i
for move, i in zip(ActionLabelsRed, range(labels_n))
}
policy = np.zeros(labels_n)
policy[move_lookup[action]] = 1
if flip:
policy = flip_policy(policy)
return list(policy)
class ObSelfPlayUCCI:
def __init__(self, config: Config, ai_move_first=True):
self.config = config
self.env = CChessEnv()
self.model = None
self.pipe = None
self.ai = None
self.chessmans = None
self.ai_move_first = ai_move_first
def load_model(self):
self.model = CChessModel(self.config)
if self.config.opts.new or not load_best_model_weight(self.model):
self.model.build()
def start(self):
self.env.reset()
self.load_model()
self.pipe = self.model.get_pipes()
self.ai = CChessPlayer(self.config, search_tree=defaultdict(VisitState), pipes=self.pipe,
enable_resign=True, debugging=False)
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
self.env.board.print_to_cl()
history = [self.env.get_state()]
turns = 0
game_over = False
final_move = None
while not game_over:
if (self.ai_move_first and turns % 2 == 0) or (not self.ai_move_first and turns % 2 == 1):
start_time = time()
no_act = None
state = self.env.get_state()
if state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
act = history[i + 1]
if not self.env.red_to_move:
act = flip_move(act)
no_act.append(act)
action, _ = self.ai.action(state, self.env.num_halfmoves, no_act)
end_time = time()
if action is None:
print("AlphaZero 投降了!")
break
move = self.env.board.make_single_record(int(action[0]), int(action[1]), int(action[2]), int(action[3]))
print(f"AlphaZero 选择移动 {move}, 消耗时间 {(end_time - start_time):.2f}s")
if not self.env.red_to_move:
action = flip_move(action)
else:
state = self.env.get_state()
print(state)
fen = senv.state_to_fen(state, turns)
action = self.get_ucci_move(fen)
if action is None:
print("Eleeye 投降了!")
break
print(action)
if not self.env.red_to_move:
rec_action = flip_move(action)
else:
rec_action = action
move = self.env.board.make_single_record(int(rec_action[0]), int(rec_action[1]), int(rec_action[2]), int(rec_action[3]))
print(f"Eleeye 选择移动 {move}")
history.append(action)
self.env.step(action)
history.append(self.env.get_state())
self.env.board.print_to_cl()
turns += 1
sleep(1)
game_over, final_move = self.env.board.is_end_final_move()
print(game_over, final_move)
if final_move:
move = self.env.board.make_single_record(int(final_move[0]), int(final_move[1]), int(final_move[2]), int(final_move[3]))
print(f"Final Move {move}")
if not self.env.red_to_move:
final_move = flip_move(final_move)
self.env.step(final_move)
self.env.board.print_to_cl()
self.ai.close()
print(f"胜者是 is {self.env.board.winner} !!!")
self.env.board.print_record()
def get_ucci_move(self, fen, time=3):
p = subprocess.Popen(self.config.resource.eleeye_path,
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
universal_newlines=True)
setfen = f'position fen {fen}\n'
setrandom = 'setoption randomness small\n'
cmd = 'ucci\n' + setrandom + setfen + f'go time {time * 1000}\n'
try:
out, err = p.communicate(cmd, timeout=time+0.5)
except:
p.kill()
try:
out, err = p.communicate()
except Exception as e:
logger.error(f"{e}, cmd = {cmd}")
return self.get_ucci_move(fen, time+1)
print(out)
lines = out.split('\n')
if lines[-2] == 'nobestmove':
return None
move = lines[-2].split(' ')[1]
if move == 'depth':
move = lines[-1].split(' ')[6]
return senv.parse_ucci_move(move)
class EvaluateWorker:
def __init__(self, config: Config, pipes1=None, pipes2=None, pid=None):
self.config = config
self.player_bt = None
self.player_ng = None
self.pid = pid
self.pipes_bt = pipes1
self.pipes_ng = pipes2
def start(self):
logger.debug(
f"Evaluate#Start Process index = {self.pid}, pid = {os.getpid()}")
score1 = 0
score2 = 0
results = []
for idx in range(self.config.eval.game_num):
start_time = time()
score, turns = self.start_game(idx)
end_time = time()
if score < 0:
score2 += 1
elif score > 0:
score1 += 1
else:
score2 += 0.5
score1 += 0.5
results.append((idx, score))
logger.debug(
f"Process{self.pid} play game {idx} time={(end_time - start_time):.1f} sec, "
f"turn={turns / 2}, model1 {score1} - {score2} model2")
return results
def start_game(self, idx):
pipe1 = self.pipes_bt.pop()
pipe2 = self.pipes_ng.pop()
search_tree1 = defaultdict(VisitState)
search_tree2 = defaultdict(VisitState)
self.player1 = CChessPlayer(self.config,
search_tree=search_tree1,
pipes=pipe1,
debugging=False,
enable_resign=True)
self.player2 = CChessPlayer(self.config,
search_tree=search_tree2,
pipes=pipe2,
debugging=False,
enable_resign=True)
# even: bst = red, ng = black; odd: bst = black, ng = red
if idx % 2 == 0:
red = self.player1
black = self.player2
logger.debug(f"best model is red, ng is black")
else:
red = self.player2
black = self.player1
logger.debug(f"best model is black, ng is red")
state = senv.INIT_STATE
value = 0 # best model's value
turns = 0 # even == red; odd == black
game_over = False
while not game_over:
start_time = time()
if turns % 2 == 0:
action, _ = red.action(state, turns)
else:
action, _ = black.action(state, turns)
end_time = time()
# logger.debug(f"pid = {self.pid}, idx = {idx}, action = {action}, turns = {turns}, time = {(end_time-start_time):.1f}")
if action is None:
logger.debug(f"{turn % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
state = senv.step(state, action)
turns += 1
if turns / 2 >= self.config.play.max_game_length:
game_over = True
value = 0
else:
game_over, value, final_move = senv.done(state)
self.player1.close()
self.player2.close()
if turns % 2 == 1: # black turn
value = -value
if idx % 2 == 1: # return player1' value
value = -value
self.pipes_bt.append(pipe1)
self.pipes_ng.append(pipe2)
return value, turns
def self_play_buffer(config, cur) -> (tuple, list):
pipe = cur.pop() # borrow
if random() > config.play.enable_resign_rate:
enable_resign = True
else:
enable_resign = False
player = CChessPlayer(config,
search_tree=defaultdict(VisitState),
pipes=pipe,
enable_resign=enable_resign,
debugging=False)
state = senv.INIT_STATE
history = [state]
policys = []
value = 0
turns = 0
game_over = False
final_move = None
while not game_over:
no_act = None
if state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
no_act.append(history[i + 1])
start_time = time()
action, policy = player.action(state, turns, no_act)
end_time = time()
if action is None:
logger.debug(f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
# logger.debug(f"Playing: {turns % 2}, action: {action}, time: {(end_time - start_time):.1f}s")
policys.append(policy)
history.append(action)
state = senv.step(state, action)
turns += 1
history.append(state)
if turns / 2 >= config.play.max_game_length:
game_over = True
value = senv.evaluate(state)
else:
game_over, value, final_move = senv.done(state)
if final_move:
policy = build_policy(final_move, False)
history.append(final_move)
policys.append(policy)
state = senv.step(state, final_move)
history.append(state)
player.close()
if turns % 2 == 1: # balck turn
value = -value
v = value
data = []
for i in range(turns):
k = i * 2
data.append([history[k], policys[i], value])
value = -value
cur.append(pipe)
return (turns, v), data
def self_play_buffer(config, pipes_bt, pipes_ng, idx, res_data, hist_base,
hist_ng) -> (tuple, list):
sleep(random())
playouts = randint(8, 12) * 100
config.play.simulation_num_per_move = playouts
logger.info(f"Set playouts = {config.play.simulation_num_per_move}")
pipe1 = pipes_bt.pop() # borrow
pipe2 = pipes_ng.pop()
player1 = CChessPlayer(config,
search_tree=defaultdict(VisitState),
pipes=pipe1,
enable_resign=False,
debugging=False,
use_history=hist_base)
player2 = CChessPlayer(config,
search_tree=defaultdict(VisitState),
pipes=pipe2,
enable_resign=False,
debugging=False,
use_history=hist_ng)
# even: bst = red, ng = black; odd: bst = black, ng = red
if idx % 2 == 0:
red = player1
black = player2
print(f"基准模型执红，待评测模型执黑")
else:
red = player2
black = player1
print(f"待评测模型执红，基准模型执黑")
state = senv.INIT_STATE
history = [state]
# policys = []
value = 0
turns = 0
game_over = False
final_move = None
no_eat_count = 0
check = False
increase_temp = False
no_act = []
while not game_over:
start_time = time()
if turns % 2 == 0:
action, _ = red.action(state,
turns,
no_act=no_act,
increase_temp=increase_temp)
else:
action, _ = black.action(state,
turns,
no_act=no_act,
increase_temp=increase_temp)
end_time = time()
if action is None:
print(f"{turns % 2} (0 = 红; 1 = 黑) 投降了!")
value = -1
break
print(
f"博弈中: 回合{turns / 2 + 1} {'红方走棋' if turns % 2 == 0 else '黑方走棋'}, 着法: {action}, 用时: {(end_time - start_time):.1f}s"
)
# policys.append(policy)
history.append(action)
try:
state, no_eat = senv.new_step(state, action)
except Exception as e:
logger.error(f"{e}, no_act = {no_act}, policy = {policy}")
game_over = True
value = 0
break
turns += 1
if no_eat:
no_eat_count += 1
else:
no_eat_count = 0
history.append(state)
if no_eat_count >= 120 or turns / 2 >= config.play.max_game_length:
game_over = True
value = 0
else:
game_over, value, final_move, check = senv.done(state,
need_check=True)
no_act = []
increase_temp = False
if not game_over:
if not senv.has_attack_chessman(state):
logger.info(f"双方无进攻子力，作和。state = {state}")
game_over = True
value = 0
if not game_over and not check and state in history[:-1]:
free_move = defaultdict(int)
for i in range(len(history) - 1):
if history[i] == state:
if senv.will_check_or_catch(state, history[i + 1]):
no_act.append(history[i + 1])
elif not senv.be_catched(state, history[i + 1]):
increase_temp = True
free_move[state] += 1
if free_move[state] >= 3:
# 作和棋处理
game_over = True
value = 0
logger.info("闲着循环三次，作和棋处理")
break
if final_move:
history.append(final_move)
state = senv.step(state, final_move)
turns += 1
value = -value
history.append(state)
data = []
if idx % 2 == 0:
data = [res_data['base']['digest'], res_data['unchecked']['digest']]
else:
data = [res_data['unchecked']['digest'], res_data['base']['digest']]
player1.close()
player2.close()
del player1, player2
gc.collect()
if turns % 2 == 1: # balck turn
value = -value
v = value
data.append(history[0])
for i in range(turns):
k = i * 2
data.append([history[k + 1], value])
value = -value
pipes_bt.append(pipe1)
pipes_ng.append(pipe2)
return (turns, v, idx), data
class EvaluateWorker:
def __init__(self, config: Config, pipes1=None, pipes2=None, pid=None):
self.config = config
self.player_bt = None
self.player_ng = None
self.pid = pid
self.pipes_bt = pipes1
self.pipes_ng = pipes2
def start(self):
logger.debug(f"Evaluate#Start Process index = {self.pid}, pid = {os.getpid()}")
score1 = 0
score2 = 0
for idx in range(self.config.eval.game_num):
start_time = time()
score, turns = self.start_game(idx)
end_time = time()
if score < 0:
score2 += 1
elif score > 0:
score1 += 1
else:
score2 += 0.5
score1 += 0.5
logger.debug(f"Process{self.pid} play game {idx} time={(end_time - start_time):.1f} sec, "
f"turn={turns / 2}, best model {score1} - {score2} next generation model")
return score2 # return next generation model's score
def start_game(self, idx):
pipe1 = self.pipes_bt.pop()
pipe2 = self.pipes_ng.pop()
search_tree1 = defaultdict(VisitState)
search_tree2 = defaultdict(VisitState)
self.player1 = CChessPlayer(self.config, search_tree=search_tree1, pipes=pipe1,
debugging=False, enable_resign=True)
self.player2 = CChessPlayer(self.config, search_tree=search_tree2, pipes=pipe2,
debugging=False, enable_resign=True)
# even: bst = red, ng = black; odd: bst = black, ng = red
if idx % 2 == 0:
red = self.player1
black = self.player2
logger.debug(f"best model is red, ng is black")
else:
red = self.player2
black = self.player1
logger.debug(f"best model is black, ng is red")
state = senv.INIT_STATE
value = 0 # best model's value
turns = 0 # even == red; odd == black
game_over = False
while not game_over:
start_time = time()
if turns % 2 == 0:
action, _ = red.action(state, turns)
else:
action, _ = black.action(state, turns)
end_time = time()
# logger.debug(f"pid = {self.pid}, idx = {idx}, action = {action}, turns = {turns}, time = {(end_time-start_time):.1f}")
if action is None:
logger.debug(f"{turn % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
state = senv.step(state, action)
turns += 1
if turns / 2 >= self.config.play.max_game_length:
game_over = True
value = 0
else:
game_over, value, final_move = senv.done(state)
self.player1.close()
self.player2.close()
if turns % 2 == 1: # black turn
value = -value
if idx % 2 == 1:
value = -value
self.pipes_bt.append(pipe1)
self.pipes_ng.append(pipe2)
return value, turns
class SelfPlayWorker:
def __init__(self, config: Config, pipes=None, pid=None):
self.config = config
self.player = None
self.cur_pipes = pipes
self.id = pid
self.buffer = []
self.pid = os.getpid()
def start(self):
logger.debug(
f"Selfplay#Start Process index = {self.id}, pid = {self.pid}")
idx = 1
self.buffer = []
search_tree = defaultdict(VisitState)
while True:
start_time = time()
value, turns, state, search_tree, store = self.start_game(
idx, search_tree)
end_time = time()
logger.debug(
f"Process {self.pid}-{self.id} play game {idx} time={(end_time - start_time):.1f} sec, "
f"turn={turns / 2}, winner = {value:.2f} (1 = red, -1 = black, 0 draw)"
)
if turns <= 10:
senv.render(state)
if store:
idx += 1
def start_game(self, idx, search_tree):
pipes = self.cur_pipes.pop()
if not self.config.play.share_mtcs_info_in_self_play or \
idx % self.config.play.reset_mtcs_info_per_game == 0:
search_tree = defaultdict(VisitState)
if random() > self.config.play.enable_resign_rate:
enable_resign = True
else:
enable_resign = False
self.player = CChessPlayer(self.config,
search_tree=search_tree,
pipes=pipes,
enable_resign=enable_resign,
debugging=False)
state = senv.INIT_STATE
history = [state]
policys = []
value = 0
turns = 0 # even == red; odd == black
game_over = False
final_move = None
while not game_over:
no_act = None
if state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
no_act.append(history[i + 1])
start_time = time()
action, policy = self.player.action(state, turns, no_act)
end_time = time()
if action is None:
logger.debug(f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
# logger.debug(f"Process{self.pid} Playing: {turns % 2}, action: {action}, time: {(end_time - start_time):.1f}s")
# for move, action_state in self.player.search_results.items():
# if action_state[0] >= 20:
# logger.info(f"move: {move}, prob: {action_state[0]}, Q_value: {action_state[1]:.2f}, Prior: {action_state[2]:.3f}")
# self.player.search_results = {}
history.append(action)
policys.append(policy)
state = senv.step(state, action)
turns += 1
history.append(state)
if turns / 2 >= self.config.play.max_game_length:
game_over = True
value = senv.evaluate(state)
else:
game_over, value, final_move = senv.done(state)
if final_move:
policy = self.build_policy(final_move, False)
history.append(final_move)
policys.append(policy)
state = senv.step(state, final_move)
history.append(state)
self.player.close()
if turns % 2 == 1: # balck turn
value = -value
v = value
if v == 0:
if random() > 0.5:
store = True
else:
store = False
else:
store = True
if store:
data = []
for i in range(turns):
k = i * 2
data.append([history[k], policys[i], value])
value = -value
self.save_play_data(idx, data)
self.cur_pipes.append(pipes)
self.remove_play_data()
return v, turns, state, search_tree, store
def save_play_data(self, idx, data):
self.buffer += data
if not idx % self.config.play_data.nb_game_in_file == 0:
return
rc = self.config.resource
game_id = datetime.now().strftime("%Y%m%d-%H%M%S.%f")
filename = rc.play_data_filename_tmpl % game_id
path = os.path.join(rc.play_data_dir, filename)
logger.info(f"Process {self.pid} save play data to {path}")
write_game_data_to_file(path, self.buffer)
if self.config.internet.distributed:
upload_worker = Thread(target=self.upload_play_data,
args=(path, filename),
name="upload_worker")
upload_worker.daemon = True
upload_worker.start()
self.buffer = []
def upload_play_data(self, path, filename):
digest = CChessModel.fetch_digest(
self.config.resource.model_best_weight_path)
data = {'digest': digest, 'username': self.config.internet.username}
response = upload_file(self.config.internet.upload_url,
path,
filename,
data,
rm=False)
if response is not None and response['status'] == 0:
logger.info(f"Upload play data {filename} finished.")
else:
logger.error(
f'Upload play data {filename} failed. {response.msg if response is not None else None}'
)
def remove_play_data(self):
files = get_game_data_filenames(self.config.resource)
if len(files) < self.config.play_data.max_file_num:
return
try:
for i in range(len(files) - self.config.play_data.max_file_num):
os.remove(files[i])
except:
pass
def build_policy(self, action, flip):
labels_n = len(ActionLabelsRed)
move_lookup = {
move: i
for move, i in zip(ActionLabelsRed, range(labels_n))
}
policy = np.zeros(labels_n)
policy[move_lookup[action]] = 1
if flip:
policy = flip_policy(policy)
return list(policy)
def self_play_buffer(config, cur, use_history=False) -> (tuple, list):
pipe = cur.pop() # borrow
if random() > config.play.enable_resign_rate:
enable_resign = True
else:
enable_resign = False
player = CChessPlayer(config, search_tree=defaultdict(VisitState), pipes=pipe,
enable_resign=enable_resign, debugging=False, use_history=use_history)
state = senv.INIT_STATE
history = [state]
# policys = []
value = 0
turns = 0
game_over = False
final_move = None
no_eat_count = 0
check = False
no_act = None
increase_temp = False
while not game_over:
start_time = time()
action, policy = player.action(state, turns, no_act, increase_temp=increase_temp)
end_time = time()
if action is None:
print(f"{turns % 2} (0 = 红; 1 = 黑) 投降了!")
value = -1
break
print(f"博弈中: 回合{turns / 2 + 1} {'红方走棋' if turns % 2 == 0 else '黑方走棋'}, 着法: {action}, 用时: {(end_time - start_time):.1f}s")
# policys.append(policy)
history.append(action)
try:
state, no_eat = senv.new_step(state, action)
except Exception as e:
logger.error(f"{e}, no_act = {no_act}, policy = {policy}")
game_over = True
value = 0
break
turns += 1
if no_eat:
no_eat_count += 1
else:
no_eat_count = 0
history.append(state)
if no_eat_count >= 120 or turns / 2 >= config.play.max_game_length:
game_over = True
value = 0
else:
game_over, value, final_move, check = senv.done(state, need_check=True)
no_act = []
increase_temp = False
if not game_over:
if not senv.has_attack_chessman(state):
logger.info(f"双方无进攻子力，作和。state = {state}")
game_over = True
value = 0
if not game_over and not check and state in history[:-1]:
free_move = defaultdict(int)
for i in range(len(history) - 1):
if history[i] == state:
if senv.will_check_or_catch(state, history[i+1]):
no_act.append(history[i + 1])
elif not senv.be_catched(state, history[i+1]):
increase_temp = True
free_move[state] += 1
if free_move[state] >= 3:
# 作和棋处理
game_over = True
value = 0
logger.info("闲着循环三次，作和棋处理")
break
if final_move:
# policy = build_policy(final_move, False)
history.append(final_move)
# policys.append(policy)
state = senv.step(state, final_move)
turns += 1
value = -value
history.append(state)
player.close()
del player
gc.collect()
if turns % 2 == 1: # balck turn
value = -value
v = value
data = [history[0]]
for i in range(turns):
k = i * 2
data.append([history[k + 1], value])
value = -value
cur.append(pipe)
return (turns, v), data
class EvaluateWorker:
def __init__(self, config: Config, pipes1=None, pipes2=None, pid=None):
self.config = config
self.player_bt = None
self.player_ng = None
self.pid = pid
self.pipes_bt = pipes1
self.pipes_ng = pipes2
def start(self):
ran = self.config.play.max_processes * 2
sleep((self.pid % ran) * 10)
logger.debug(
f"Evaluate#Start Process index = {self.pid}, pid = {os.getpid()}")
score = 0
total_score = 0
red_new_win = 0
red_new_fail = 0
red_new_draw = 0
black_new_win = 0
black_new_fail = 0
black_new_draw = 0
for idx in range(self.config.eval.game_num):
start_time = time()
value, turns = self.start_game(idx)
end_time = time()
if (value == 1 and idx % 2 == 0) or (value == -1 and idx % 2 == 1):
if idx % 2 == 0:
black_new_fail += 1
else:
red_new_fail += 1
result = '基准模型胜'
elif (value == 1 and idx % 2 == 1) or (value == -1
and idx % 2 == 0):
if idx % 2 == 0:
black_new_win += 1
else:
red_new_win += 1
result = '待评测模型胜'
else:
if idx % 2 == 0:
black_new_draw += 1
else:
red_new_draw += 1
result = '和棋'
if value == -1: # loss
score = 0
elif value == 1: # win
score = 1
else:
score = 0.5
if idx % 2 == 0:
score = 1 - score
else:
score = score
logger.info(
f"进程{self.pid}评测完毕 用时{(end_time - start_time):.1f}秒, "
f"{turns / 2}回合, {result}, 得分：{score}, value = {value}, idx = {idx}"
)
total_score += score
return (total_score, red_new_win, red_new_draw, red_new_fail,
black_new_win, black_new_draw, black_new_fail)
def start_game(self, idx):
pipe1 = self.pipes_bt.pop()
pipe2 = self.pipes_ng.pop()
search_tree1 = defaultdict(VisitState)
search_tree2 = defaultdict(VisitState)
playouts = randint(8, 12) * 100
self.config.play.simulation_num_per_move = playouts
logger.info(
f"Set playouts = {self.config.play.simulation_num_per_move}")
self.player1 = CChessPlayer(self.config,
search_tree=search_tree1,
pipes=pipe1,
debugging=False,
enable_resign=False)
self.player2 = CChessPlayer(self.config,
search_tree=search_tree2,
pipes=pipe2,
debugging=False,
enable_resign=False)
# even: bst = red, ng = black; odd: bst = black, ng = red
if idx % 2 == 0:
red = self.player1
black = self.player2
logger.debug(f"best model is red, ng is black")
else:
red = self.player2
black = self.player1
logger.debug(f"best model is black, ng is red")
state = senv.INIT_STATE
history = [state]
value = 0 # best model's value
turns = 0 # even == red; odd == black
game_over = False
no_eat_count = 0
check = False
while not game_over:
start_time = time()
no_act = None
increase_temp = False
if not check and state in history[:-1]:
no_act = []
increase_temp = True
free_move = defaultdict(int)
for i in range(len(history) - 1):
if history[i] == state:
# 如果走了下一步是将军或捉：禁止走那步
if senv.will_check_or_catch(state, history[i + 1]):
no_act.append(history[i + 1])
# 否则当作闲着处理
else:
free_move[state] += 1
if free_move[state] >= 3:
# 作和棋处理
game_over = True
value = 0
logger.info("闲着循环三次，作和棋处理")
break
if game_over:
break
if turns % 2 == 0:
action, _ = red.action(state,
turns,
no_act=no_act,
increase_temp=increase_temp)
else:
action, _ = black.action(state,
turns,
no_act=no_act,
increase_temp=increase_temp)
end_time = time()
if self.config.opts.log_move:
logger.debug(
f"进程id = {self.pid}, action = {action}, turns = {turns}, time = {(end_time-start_time):.1f}"
)
if action is None:
logger.debug(f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
history.append(action)
state, no_eat = senv.new_step(state, action)
turns += 1
if no_eat:
no_eat_count += 1
else:
no_eat_count = 0
history.append(state)
if no_eat_count >= 120 or turns / 2 >= self.config.play.max_game_length:
game_over = True
value = 0
else:
game_over, value, final_move, check = senv.done(
state, need_check=True)
if not game_over:
if not senv.has_attack_chessman(state):
logger.info(f"双方无进攻子力，作和。state = {state}")
game_over = True
value = 0
if final_move:
history.append(final_move)
state = senv.step(state, final_move)
turns += 1
value = -value
history.append(state)
self.player1.close()
self.player2.close()
if turns % 2 == 1: # black turn
value = -value
self.pipes_bt.append(pipe1)
self.pipes_ng.append(pipe2)
return value, turns
class PlayWithHuman:
def __init__(self, config: Config):
self.config = config
self.env = CChessEnv()
self.model = None
self.pipe = None
self.ai = None
self.winstyle = 0
self.chessmans = None
self.human_move_first = True
self.screen_width = 720
self.height = 577
self.width = 521
self.chessman_w = 57
self.chessman_h = 57
self.disp_record_num = 15
self.rec_labels = [None] * self.disp_record_num
self.nn_value = 0
self.mcts_moves = {}
self.history = []
if self.config.opts.bg_style == 'WOOD':
self.chessman_w += 1
self.chessman_h += 1
def load_model(self):
self.model = CChessModel(self.config)
if self.config.opts.new or not load_best_model_weight(self.model):
self.model.build()
def init_screen(self):
bestdepth = pygame.display.mode_ok([self.screen_width, self.height],
self.winstyle, 32)
screen = pygame.display.set_mode([self.screen_width, self.height],
self.winstyle, bestdepth)
pygame.display.set_caption("中国象棋Zero")
# create the background, tile the bgd image
bgdtile = load_image(f'{self.config.opts.bg_style}.GIF')
bgdtile = pygame.transform.scale(bgdtile, (self.width, self.height))
board_background = pygame.Surface([self.width, self.height])
board_background.blit(bgdtile, (0, 0))
widget_background = pygame.Surface(
[self.screen_width - self.width, self.height])
white_rect = Rect(0, 0, self.screen_width - self.width, self.height)
widget_background.fill((255, 255, 255), white_rect)
#create text label
font_file = self.config.resource.font_path
font = pygame.font.Font(font_file, 16)
font_color = (0, 0, 0)
font_background = (255, 255, 255)
t = font.render("着法记录", True, font_color, font_background)
t_rect = t.get_rect()
t_rect.x = 10
t_rect.y = 10
widget_background.blit(t, t_rect)
screen.blit(board_background, (0, 0))
screen.blit(widget_background, (self.width, 0))
pygame.display.flip()
self.chessmans = pygame.sprite.Group()
creat_sprite_group(self.chessmans, self.env.board.chessmans_hash,
self.chessman_w, self.chessman_h)
return screen, board_background, widget_background
def start(self, human_first=True):
self.env.reset()
self.load_model()
self.pipe = self.model.get_pipes()
self.ai = CChessPlayer(self.config,
search_tree=defaultdict(VisitState),
pipes=self.pipe,
enable_resign=True,
debugging=True)
self.human_move_first = human_first
pygame.init()
screen, board_background, widget_background = self.init_screen()
framerate = pygame.time.Clock()
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
current_chessman = None
if human_first:
self.env.board.calc_chessmans_moving_list()
ai_worker = Thread(target=self.ai_move, name="ai_worker")
ai_worker.daemon = True
ai_worker.start()
while not self.env.board.is_end():
for event in pygame.event.get():
if event.type == pygame.QUIT:
self.env.board.print_record()
self.ai.close(wait=False)
game_id = datetime.now().strftime("%Y%m%d-%H%M%S")
path = os.path.join(
self.config.resource.play_record_dir,
self.config.resource.play_record_filename_tmpl %
game_id)
self.env.board.save_record(path)
sys.exit()
elif event.type == VIDEORESIZE:
pass
elif event.type == MOUSEBUTTONDOWN:
if human_first == self.env.red_to_move:
pressed_array = pygame.mouse.get_pressed()
for index in range(len(pressed_array)):
if index == 0 and pressed_array[index]:
mouse_x, mouse_y = pygame.mouse.get_pos()
col_num, row_num = translate_hit_area(
mouse_x, mouse_y, self.chessman_w,
self.chessman_h)
chessman_sprite = select_sprite_from_group(
self.chessmans, col_num, row_num)
if current_chessman is None and chessman_sprite != None:
if chessman_sprite.chessman.is_red == self.env.red_to_move:
current_chessman = chessman_sprite
chessman_sprite.is_selected = True
elif current_chessman != None and chessman_sprite != None:
if chessman_sprite.chessman.is_red == self.env.red_to_move:
current_chessman.is_selected = False
current_chessman = chessman_sprite
chessman_sprite.is_selected = True
else:
move = str(current_chessman.chessman.col_num) + str(current_chessman.chessman.row_num) +\
str(col_num) + str(row_num)
success = current_chessman.move(
col_num, row_num, self.chessman_w,
self.chessman_h)
self.history.append(move)
if success:
self.chessmans.remove(
chessman_sprite)
chessman_sprite.kill()
current_chessman.is_selected = False
current_chessman = None
self.history.append(
self.env.get_state())
elif current_chessman != None and chessman_sprite is None:
move = str(current_chessman.chessman.col_num) + str(current_chessman.chessman.row_num) +\
str(col_num) + str(row_num)
success = current_chessman.move(
col_num, row_num, self.chessman_w,
self.chessman_h)
self.history.append(move)
if success:
current_chessman.is_selected = False
current_chessman = None
self.history.append(
self.env.get_state())
self.draw_widget(screen, widget_background)
framerate.tick(20)
# clear/erase the last drawn sprites
self.chessmans.clear(screen, board_background)
# update all the sprites
self.chessmans.update()
self.chessmans.draw(screen)
pygame.display.update()
self.ai.close(wait=False)
logger.info(f"Winner is {self.env.board.winner} !!!")
self.env.board.print_record()
game_id = datetime.now().strftime("%Y%m%d-%H%M%S")
path = os.path.join(
self.config.resource.play_record_dir,
self.config.resource.play_record_filename_tmpl % game_id)
self.env.board.save_record(path)
sleep(3)
def ai_move(self):
ai_move_first = not self.human_move_first
self.history = [self.env.get_state()]
no_act = None
while not self.env.done:
if ai_move_first == self.env.red_to_move:
labels = ActionLabelsRed
labels_n = len(ActionLabelsRed)
self.ai.search_results = {}
state = self.env.get_state()
logger.info(f"state = {state}")
_, _, _, check = senv.done(state, need_check=True)
if not check and state in self.history[:-1]:
no_act = []
free_move = defaultdict(int)
for i in range(len(self.history) - 1):
if self.history[i] == state:
# 如果走了下一步是将军或捉：禁止走那步
if senv.will_check_or_catch(
state, self.history[i + 1]):
no_act.append(self.history[i + 1])
# 否则当作闲着处理
else:
free_move[state] += 1
if free_move[state] >= 2:
# 作和棋处理
self.env.winner = Winner.draw
self.env.board.winner = Winner.draw
break
if no_act:
logger.debug(f"no_act = {no_act}")
action, policy = self.ai.action(state, self.env.num_halfmoves,
no_act)
if action is None:
logger.info("AI has resigned!")
return
self.history.append(action)
if not self.env.red_to_move:
action = flip_move(action)
key = self.env.get_state()
p, v = self.ai.debug[key]
logger.info(f"check = {check}, NN value = {v:.3f}")
self.nn_value = v
logger.info("MCTS results:")
self.mcts_moves = {}
for move, action_state in self.ai.search_results.items():
move_cn = self.env.board.make_single_record(
int(move[0]), int(move[1]), int(move[2]), int(move[3]))
logger.info(
f"move: {move_cn}-{move}, visit count: {action_state[0]}, Q_value: {action_state[1]:.3f}, Prior: {action_state[2]:.3f}"
)
self.mcts_moves[move_cn] = action_state
x0, y0, x1, y1 = int(action[0]), int(action[1]), int(
action[2]), int(action[3])
chessman_sprite = select_sprite_from_group(
self.chessmans, x0, y0)
sprite_dest = select_sprite_from_group(self.chessmans, x1, y1)
if sprite_dest:
self.chessmans.remove(sprite_dest)
sprite_dest.kill()
chessman_sprite.move(x1, y1, self.chessman_w, self.chessman_h)
self.history.append(self.env.get_state())
def draw_widget(self, screen, widget_background):
white_rect = Rect(0, 0, self.screen_width - self.width, self.height)
widget_background.fill((255, 255, 255), white_rect)
pygame.draw.line(widget_background, (255, 0, 0), (10, 285),
(self.screen_width - self.width - 10, 285))
screen.blit(widget_background, (self.width, 0))
self.draw_records(screen, widget_background)
self.draw_evaluation(screen, widget_background)
def draw_records(self, screen, widget_background):
text = '着法记录'
self.draw_label(screen, widget_background, text, 10, 16, 10)
records = self.env.board.record.split('\n')
font_file = self.config.resource.font_path
font = pygame.font.Font(font_file, 12)
i = 0
for record in records[-self.disp_record_num:]:
self.rec_labels[i] = font.render(record, True, (0, 0, 0),
(255, 255, 255))
t_rect = self.rec_labels[i].get_rect()
# t_rect.centerx = (self.screen_width - self.width) / 2
t_rect.y = 35 + i * 15
t_rect.x = 10
t_rect.width = self.screen_width - self.width
widget_background.blit(self.rec_labels[i], t_rect)
i += 1
screen.blit(widget_background, (self.width, 0))
def draw_evaluation(self, screen, widget_background):
title_label = 'CC-Zero信息'
self.draw_label(screen, widget_background, title_label, 300, 16, 10)
info_label = f'MCTS搜索次数：{self.config.play.simulation_num_per_move}'
self.draw_label(screen, widget_background, info_label, 335, 14, 10)
eval_label = f"当前局势评估: {self.nn_value:.3f}"
self.draw_label(screen, widget_background, eval_label, 360, 14, 10)
label = f"MCTS搜索结果:"
self.draw_label(screen, widget_background, label, 395, 14, 10)
label = f"着法 访问计数 动作价值 先验概率"
self.draw_label(screen, widget_background, label, 415, 12, 10)
i = 0
tmp = copy.deepcopy(self.mcts_moves)
for mov, action_state in tmp.items():
label = f"{mov}"
self.draw_label(screen, widget_background, label, 435 + i * 20, 12,
10)
label = f"{action_state[0]}"
self.draw_label(screen, widget_background, label, 435 + i * 20, 12,
70)
label = f"{action_state[1]:.2f}"
self.draw_label(screen, widget_background, label, 435 + i * 20, 12,
100)
label = f"{action_state[2]:.3f}"
self.draw_label(screen, widget_background, label, 435 + i * 20, 12,
150)
i += 1
def draw_label(self,
screen,
widget_background,
text,
y,
font_size,
x=None):
font_file = self.config.resource.font_path
font = pygame.font.Font(font_file, font_size)
label = font.render(text, True, (0, 0, 0), (255, 255, 255))
t_rect = label.get_rect()
t_rect.y = y
if x != None:
t_rect.x = x
else:
t_rect.centerx = (self.screen_width - self.width) / 2
widget_background.blit(label, t_rect)
screen.blit(widget_background, (self.width, 0))
class SelfPlayWorker:
def __init__(self, config: Config, pipes=None, pid=None):
self.config = config
self.player = None
self.cur_pipes = pipes
self.id = pid
self.buffer = []
self.pid = os.getpid()
def start(self):
self.pid = os.getpid()
logger.debug(f"Selfplay#Start Process index = {self.id}, pid = {self.pid}")
idx = 1
self.buffer = []
search_tree = defaultdict(VisitState)
while True:
start_time = time()
value, turns, state, search_tree, store = self.start_game(idx, search_tree)
end_time = time()
if value != 1 and value != -1:
winner = 'Draw'
elif idx % 2 == 0 and value == 1 or idx % 2 == 1 and value == -1:
winner = 'AlphaHe'
else:
winner = 'Eleeye'
logger.debug(f"Process {self.pid}-{self.id} play game {idx} time={(end_time - start_time):.1f} sec, "
f"turn={turns / 2}, value = {value:.2f}, winner is {winner}")
if turns <= 10 and store:
senv.render(state)
if store:
idx += 1
def start_game(self, idx, search_tree):
pipes = self.cur_pipes.pop()
if not self.config.play.share_mtcs_info_in_self_play or \
idx % self.config.play.reset_mtcs_info_per_game == 0:
search_tree = defaultdict(VisitState)
if random() > self.config.play.enable_resign_rate:
enable_resign = True
else:
enable_resign = False
self.player = CChessPlayer(self.config, search_tree=search_tree, pipes=pipes, enable_resign=enable_resign, debugging=False)
state = senv.INIT_STATE
history = [state]
policys = []
value = 0
turns = 0 # even == red; odd == black
game_over = False
is_alpha_red = True if idx % 2 == 0 else False
final_move = None
while not game_over:
if (is_alpha_red and turns % 2 == 0) or (not is_alpha_red and turns % 2 == 1):
no_act = None
if state in history[:-1]:
no_act = []
for i in range(len(history) - 1):
if history[i] == state:
no_act.append(history[i + 1])
action, policy = self.player.action(state, turns, no_act)
if action is None:
logger.debug(f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
else:
fen = senv.state_to_fen(state, turns)
action = self.get_ucci_move(fen)
if action is None:
logger.debug(f"{turns % 2} (0 = red; 1 = black) has resigned!")
value = -1
break
if turns % 2 == 1:
action = flip_move(action)
try:
policy = self.build_policy(action, False)
except Exception as e:
logger.error(f"Build policy error {e}, action = {action}, state = {state}, fen = {fen}")
value = 0
break
history.append(action)
policys.append(policy)
state = senv.step(state, action)
turns += 1
history.append(state)
if turns / 2 >= self.config.play.max_game_length:
game_over = True
value = senv.evaluate(state)
else:
game_over, value, final_move = senv.done(state)
if final_move:
policy = self.build_policy(final_move, False)
history.append(final_move)
policys.append(policy)
state = senv.step(state, final_move)
history.append(state)
self.player.close()
if turns % 2 == 1: # balck turn
value = -value
v = value
if v == 0 or turns <= 10:
if random() > 0.7:
store = True
else:
store = False
else:
store = True
if store:
data = []
for i in range(turns):
k = i * 2
data.append([history[k], policys[i], value])
value = -value
self.save_play_data(idx, data)
self.cur_pipes.append(pipes)
self.remove_play_data()
return v, turns, state, search_tree, store
def get_ucci_move(self, fen, time=3):
p = subprocess.Popen(self.config.resource.eleeye_path,
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
universal_newlines=True)
setfen = f'position fen {fen}\n'
setrandom = f'setoption randomness {self.config.opts.random}\n'
cmd = 'ucci\n' + setrandom + setfen + f'go time {time * 1000}\n'
try:
out, err = p.communicate(cmd, timeout=time+0.5)
except subprocess.TimeoutExpired:
p.kill()
try:
out, err = p.communicate()
except Exception as e:
logger.error(f"{e}, cmd = {cmd}")
return self.get_ucci_move(fen, time+1)
lines = out.split('\n')
if lines[-2] == 'nobestmove':
return None
move = lines[-2].split(' ')[1]
if move == 'depth':
move = lines[-1].split(' ')[6]
return senv.parse_ucci_move(move)
def save_play_data(self, idx, data):
self.buffer += data
if not idx % self.config.play_data.nb_game_in_file == 0:
return
rc = self.config.resource
game_id = datetime.now().strftime("%Y%m%d-%H%M%S.%f")
path = os.path.join(rc.play_data_dir, rc.play_data_filename_tmpl % game_id)
logger.info(f"Process {self.pid} save play data to {path}")
write_game_data_to_file(path, self.buffer)
self.buffer = []
def remove_play_data(self):
files = get_game_data_filenames(self.config.resource)
if len(files) < self.config.play_data.max_file_num:
return
try:
for i in range(len(files) - self.config.play_data.max_file_num):
os.remove(files[i])
except:
pass
def build_policy(self, action, flip):
labels_n = len(ActionLabelsRed)
move_lookup = {move: i for move, i in zip(ActionLabelsRed, range(labels_n))}
policy = np.zeros(labels_n)
policy[move_lookup[action]] = 1
if flip:
policy = flip_policy(policy)
return list(policy)