def run(start_player=0, is_shown=1):
# run a gomoku game with AI
# you can set
# human vs AI or AI vs AI
n = 5
width, height = 15, 15
# model_file = 'model_15_15_5/best_policy.model'
# width, height = 6, 6
# model_file = 'model/best_policy.model'
# width, height = 11, 11
# model_file = 'model/best_policy.model'
model_file = 'training/model_best/policy.model'
# model_file = 'training/best_policy.model'
p = os.getcwd()
model_file = path.join(p, model_file)
board = Board(width=width, height=height, n_in_row=n)
game = Game(board)
mcts_player = MCTS_pure(5, 8000)
best_policy = PolicyValueNet(board_width=width,
board_height=height,
block=19,
init_model=model_file,
cuda=True)
# alpha_zero vs alpha_zero
# best_policy.save_numpy(best_policy.network_all_params)
# best_policy.load_numpy(best_policy.network_oppo_all_params)
alpha_zero_player = MCTSPlayer(
best_policy,
model_file,
policy_value_function=best_policy.policy_value_fn_random,
action_fc=best_policy.action_fc_test,
evaluation_fc=best_policy.evaluation_fc2_test,
c_puct=5,
n_playout=400,
is_selfplay=False)
# alpha_zero_player_oppo = MCTSPlayer(policy_value_function=best_policy.policy_value_fn_random,
# action_fc=best_policy.action_fc_test_oppo,
# evaluation_fc=best_policy.evaluation_fc2_test_oppo,
# c_puct=5,
# n_playout=400,
# is_selfplay=False)
# human player, input your move in the format: 2,3
# set start_player=0 for human first
# play in termianl without GUI
# human = Human()
# win = game.start_play(human, alpha_zero_player, start_player=start_player, is_shown=is_shown,print_prob=True)
# return win
# play in GUI
game.start_play_with_UI(alpha_zero_player) # Play with alpha zero
def __init__(self):
# def run(start_player=0,is_shown=1):
self.start_player = 0
self.is_shown = 1
menu = Gui_menu()
# run a gomoku game with AI
# you can set
# human vs AI or AI vs AI
self.n = 5 # Rule 5 목
if menu.rule == 11:
width, height = 11, 11 # 바둑판의 폭과 높이
model_file = 'model_11_11_5/best_policy.model'
elif menu.rule == 15:
width, height = 15, 15 # 바둑판의 폭과 높이
model_file = 'model_15_15_5/best_policy.model'
p = os.getcwd() # 현재 작업 경로를 얻음
model_file = path.join(p, model_file) # 파일 경로 + model file name
board = Board(width=width, height=height, n_in_row=self.n) # 게임 판
game = Game(board)
mcts_player = MCTS_pure(5, 400)
best_policy = PolicyValueNet(board_width=width,
board_height=height,
block=19,
init_model=model_file,
cuda=True)
# alpha_zero vs alpha_zero
# best_policy.save_numpy(best_policy.network_all_params)
# best_policy.load_numpy(best_policy.network_oppo_all_params)
alpha_zero_player = MCTSPlayer(
policy_value_function=best_policy.policy_value_fn_random,
action_fc=best_policy.action_fc_test,
evaluation_fc=best_policy.evaluation_fc2_test,
c_puct=5,
n_playout=400,
is_selfplay=False)
# alpha_zero_player_oppo = MCTSPlayer(policy_value_function=best_policy.policy_value_fn_random,
# action_fc=best_policy.action_fc_test_oppo,
# evaluation_fc=best_policy.evaluation_fc2_test_oppo,
# c_puct=5,
# n_playout=400,
# is_selfplay=False)
# human player, input your move in the format: 2,3
# set start_player=0 for human first
# play in termianl without GUI
# human = Human()
# win = game.start_play(human, alpha_zero_player, start_player=start_player, is_shown=is_shown,print_prob=True)
# return win
# play in GUI
game.start_play_with_UI(alpha_zero_player)