正在学习《Python机器学习》一书中第9章的“在Web应用中嵌入机器学习模型”，于是想用keras+flask做简单的尝试，代码直接运行没问题，但是flask上跑就是报错，如下：

ValueError: Tensor Tensor("dense_2/Softmax:0", shape=(?, 192), dtype=float32) is not an element of this graph.

网上一顿搜索操作猛如虎，参考了两位大神的解决方案，简直了，就是如下简单的四行代码：

import tensorflow as tf
graph = tf.get_default_graph()

global graph
with graph.as_default():
    # 执行预测函数

添加到app.py，问题解决

from flask import Flask, render_template, request
from wtforms import Form, TextAreaField, validators
import os
import tensorflow.keras as keras
import tensorflow as tf
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split



app = Flask(__name__)

graph = tf.get_default_graph()



cur_dir = os.path.dirname(__file__)

model = tf.keras.models.load_model(open(os.path.join(cur_dir,'demo_tensorflow.model'), 'rb'))


def classify(X_test):
    global graph
    with graph.as_default():
        predictions = model.predict(X_test)

    return predictions


'''
以下代码省略
'''

解决方案链接：

1. ValueError: Tensor Tensor xxx is not an element of of this graph的解决方案

2. Keras 的 Web 填坑记