introduction in https://www.jianshu.com/p/5b74f1bc0178

run your server

docker run -t --rm -p 8501:8501
-v "$(pwd)/model/half_plus_ten:/models/half_plus_ten"
-e MODEL_NAME=half_plus_ten
tensorflow/serving

docker run -t --rm -p 8501:8501
-v "$(pwd)/pb_model:/models/half_plus_ten"
-e MODEL_NAME=half_plus_ten
tensorflow/serving

test your sever

curl -d '{"instances": [1.0, 2.0, 5.0]}' -X POST http://localhost:8501/v1/models/half_plus_ten:predict