url = img_file.replace(IMG_PATH, IMG_URL)
return {
'code': 0,
'msg': 'success',
'data': {
'url': url,
'objects': res,
'caption': caption,
'tts_caption': tts_url
}
}
if __name__ == '__main__':
"""服务入口"""
port = 8026
# log init
log_file = ApiObjectDetect.__name__.lower() # + '-' + str(os.getpid())
utils.init_logging(log_file=log_file, log_path=APP_PATH)
print("log_file: {}".format(log_file))
# 路由
app = tornado.web.Application(handlers=[(r'/piglab/image/object_detect',
ApiObjectDetect)])
# 启动服务
http_server = tornado.httpserver.HTTPServer(app, xheaders=True)
http_server.listen(port)
tornado.ioloop.IOLoop.instance().start()
if (i + 1) % self.check_freq == 0:
logging.info("TRAIN Current self-play batch: {}".format(i + 1))
# 策略胜率评估：模型与纯MCTS玩家对战n局看胜率
win_ratio = self.policy_evaluate(self.policy_evaluate_size)
self.policy_value_net.save_model(CUR_PATH + '/model/current_policy_{}x{}.model'.format(self.board_width, self.board_height))
if win_ratio > self.best_win_ratio: # 胜率超过历史最优模型
logging.info("TRAIN New best policy!!!!!!!!batch:{} win_ratio:{}->{} pure_mcts_playout_num:{}".format(i + 1, self.best_win_ratio, win_ratio, self.pure_mcts_playout_num))
self.best_win_ratio = win_ratio
# 保存当前模型为最优模型best_policy
self.policy_value_net.save_model(CUR_PATH + '/model/best_policy_{}x{}.model'.format(self.board_width, self.board_height))
# 如果胜率=100%，则增加纯MCT的模拟数 (<6000的限制视mem情况)
if self.best_win_ratio == 1.0: # and self.pure_mcts_playout_num < 6000:
self.pure_mcts_playout_num += 1000
self.best_win_ratio = 0.0
except KeyboardInterrupt:
logging.info('\n\rquit')
if __name__ == '__main__':
# log init
utils.init_logging(log_file='train', log_path=CUR_PATH)
# train
size = 16 # 棋盘大小
model_file = '{}/model/current_policy_{}x{}.model'.format(CUR_PATH, size, size)
print("model file exists: {}".format(os.path.exists(model_file)))
model_file = '' if os.path.exists(model_file) is False else model_file
#training_pipeline = GomokuTrainPipeline(size=size)
training_pipeline = GomokuTrainPipeline(init_model=model_file, size=size)
training_pipeline.run()
# -*- coding: utf-8 -*- """ File: da.py Desc: 需求识别基类 Author:yanjingang([email protected]) Date: 2019/2/21 23:34 """ import logging from dp import utils from dp.da import Da import constants if __name__ == '__main__': """test""" utils.init_logging(log_file='da', log_path=constants.APP_PATH) da = Da(dict_path=constants.APP_PATH + "/data/da/") # 意图分类 #res = da.get_trigger("宝马3系报价") res = da.get_trigger("哪个恐龙跑的最快？") print("trigger: {}".format(res)) # 意图分类下的query解析 #res = da.get_parser(res[0]['type'], "宝马3系报价") res = da.get_parser(res[0]['type'], "哪个恐龙跑的最快？") #res = da.get_parser(res[0]['type'], "跑的最快的恐龙") print("parser: {}".format(res)) ''' # tree test tree = Node(name='root') ctree = tree.add_child(Node('car'))