部署模型、开启服务的方法，见此处。

一、
  h5_2_pd.py这是将h5模型转化为saved模型的脚本。

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Date    : 2019-07-26 17:19:07
# @Author  : 
# @Email   : None
# @Version : py3

from keras.models import load_model
import tensorflow as tf
import os 
from os import *
import os.path as osp
from keras import backend as K
import tensorflow.keras.backend as K
from tensorflow.keras.losses import categorical_crossentropy
from tensorflow.keras.optimizers import Adadelta
# from mymodel import captcha_model as model

def export_model(model,
                 export_model_dir,
                 model_version
                 ):
    """
    :param export_model_dir: type string, save dir for exported model
    :param model_version: type int best
    :return:no return
    """
    with tf.get_default_graph().as_default():
        # prediction_signature
        tensor_info_input = tf.saved_model.utils.build_tensor_info(model.input)
        tensor_info_input = tf.saved_model.utils.build_tensor_info(model.output)
        prediction_signature = (
            tf.saved_model.signature_def_utils.build_signature_def(
                inputs={'images': tensor_info_input}, # Tensorflow.TensorInfo
                outputs={'result': tensor_info_input},
                method_name=tf.saved_model.signature_constants.PREDICT_METHOD_NAME))
        print('step1 => prediction_signature created successfully')
        # set-up a builder
        export_path_base = export_model_dir
        export_path = os.path.join(
            tf.compat.as_bytes(export_path_base),
            tf.compat.as_bytes(str(model_version)))
        builder = tf.saved_model.builder.SavedModelBuilder(export_path)
        builder.add_meta_graph_and_variables(
            # tags:SERVING,TRAINING,EVAL,GPU,TPU
            sess=K.get_session(),
            tags=[tf.saved_model.tag_constants.SERVING],
            signature_def_map={'prediction_signature': prediction_signature,},
            )
        print('step2 => Export path(%s) ready to export trained model' % export_path, '\n starting to export model...')
        builder.save(as_text=True)
        print('Done exporting!')


def export_model2(model,
                 export_model_dir,
                 model_version
                 ):
    """
    :param export_model_dir: type string, save dir for exported model    url
    :param model_version: type int best
    :return:no return
    """
    with tf.get_default_graph().as_default():
        # prediction_signature
        tensor_info_input = tf.saved_model.utils.build_tensor_info(model.input)
        tensor_info_output = tf.saved_model.utils.build_tensor_info(model.output)
        print(model.output.shape, '**', tensor_info_output)
        prediction_signature = (
            tf.saved_model.signature_def_utils.build_signature_def(
                inputs={'images': tensor_info_input}, # Tensorflow.TensorInfo
                outputs={'result': tensor_info_output},
                #method_name=tf.saved_model.signature_constants.PREDICT_METHOD_NAME)
                 method_name= "tensorflow/serving/predict")
               
        )
        print('step1 => prediction_signature created successfully')
        # set-up a builder
        mkdir(export_model_dir)
        export_path_base = export_model_dir
        export_path = os.path.join(
            tf.compat.as_bytes(export_path_base),
            tf.compat.as_bytes(str(model_version)))
        builder = tf.saved_model.builder.SavedModelBuilder(export_path)
        builder.add_meta_graph_and_variables(
            # tags:SERVING,TRAINING,EVAL,GPU,TPU
            sess=K.get_session(),
            tags=[tf.saved_model.tag_constants.SERVING],
            signature_def_map={
                'predict':
                    prediction_signature,
                   tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY:
              prediction_signature,

            },
            )
        print('step2 => Export path(%s) ready to export trained model' % export_path, '\n starting to export model...')
        #builder.save(as_text=True)
        builder.save()
        print('Done exporting!')

if __name__ == '__main__':
    # 指定路径和模型版本号
    model = tf.keras.models.load_model('..../model.h5')
    export_model2(
        model,
        './export_modeltest',
        1
    )

二、
  client脚本。这是对模型进行restful API访问的脚本。个人觉得这里面的关键点是搞清楚你的模型输入输出数据格式。所谓输出就是神经网络的input tendor的形状、数据格式。输出就是神经网络输出的数据格式。可以通过tensorflow的命令saved_model_cli 查看，使用格式如下，这个命令要在python环境安装tensorflow下才能使用。

saved_model_cli show --dir  ...../resnet/1 --all

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Date    : 2019-07-26 17:19:07
# @Author  : 
# @Email   : None
# @Version : py3

from __future__ import print_function
import base64
import requests
import cv2
import numpy as np
import json
import pandas as pd
# The server URL specifies the endpoint of your server running the ResNet
# model with the name "resnet" and using the predict interface.
SERVER_URL = 'http://localhost:8501/v1/models/resnet:predict'

# get_inputs函数处理模型预测数据，要和训练模型时的数据预处过程相同
def get_inputs(src=[]):
    pre_x = []
    for s in src:
        input = cv2.imread(s)
        input = cv2.resize(input, (224, 224))
        input = cv2.cvtColor(input, cv2.COLOR_BGR2GRAY)
        pre_x.append(input)  # input一张图片
    pre_x = np.asarray(pre_x) / 255.0
    return pre_x


class NumpyEncoder(json.JSONEncoder):
    def default(self, obj):
        if isinstance(obj, np.ndarray):
            return obj.tolist()
        return json.JSONEncoder.default(self, obj)


def main():
  # Compose a JSON Predict request (send JPG image in base64).
  #jpg_bytes =      base64.b64encode(dl_request.content).decode('utf-8')
  #predict_request = '{"instances" : [%s]}' % (jpeg_bytes)
  images = [imglist]
  pre_x = get_inputs(images)
  pre_x = np.reshape(pre_x, ( pre_x.shape[1], pre_x.shape[2], 1))

  input_data = np.expand_dims(pre_x,axis =0)
  
  # change the format of input data suit the model
  p = {'inputs': input_data}
  predict_request = {"instances":p}

  # to json
  param = json.dumps(p, cls=NumpyEncoder)
  
  # Send few requests to warm-up the model.
  for _ in range(3):
    response=requests.post(SERVER_URL, data=param)
    response.raise_for_status()

  # Send few actual requests
  response = requests.post(SERVER_URL, data=param)
  response.raise_for_status()
  prediction = response.json()['outputs'][0]
  prediction = np.array(prediction)
  
  # 输出最终预测结果，就是得分最高的类别
  pre_result_index = np.argsort(-prediction)[0]

if __name__ == '__main__':
  main()