将训练好的t2t模型加载到内存并开启api服务

• python2, tensorflow, tensor2tensor, flask-cors
• 安装配置tensorflow-serving
• 将t2t模型导出，与训练目录下的/self_script和/self_data上传服务器指定目录

  t2t-exporter \
  --t2t_usr_dir=self_script \
  --problems=my_problem --data_dir=./self_data \ 
  --model=lstm_seq2seq_attention \
  --hparams_set=lstm_attention \
  --output_dir=./train
  

• 开启tensorflow-server

  tensorflow_model_server \
  --port=9000 \
  --model_name=yourname \
  --model_base_path=~/self_t2t/train/export/Servo
  

  OR tensorflow_model_server –model_config_file=models.json –port=9000
• 安装tensorflow-serving-api install tensorflow-serving-api

• 下载flask_tensor_api到指定目录

• 编辑conf.py文件

• 启动服务 gunicorn -c run.py app:app -p ./log/web.pid -D

• 查看服务 cat ./log/web.pid

https://blog.csdn.net/csa121/article/details/79605215

https://github.com/tensorflow/tensor2tensor/blob/82726e2708f7dd88abdd7c484c5cb7bb86cf7ede/tensor2tensor/serving/README.md

https://spacewander.github.io/explore-flask-zh/5-configuration.html

http://www.pythondoc.com/exploreflask/deployment.html