Tensorflow模型保存与调用

一、模型保存

saver = tf.train.Saver()与saver.save(sess,'net/my_net.ckpt') 函数

import tensorflow as tf

saver = tf.train.Saver()

with tf.Session() as sess:
    sess.run(init)
    for epoch in range(11):
        for batch in range(n_batch):
            batch_xs,batch_ys =  mnist.train.next_batch(batch_size)
            sess.run(train_step,feed_dict={x:batch_xs,y:batch_ys})
        
        acc = sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels})
        print("Iter " + str(epoch) + ",Testing Accuracy " + str(acc))
    #保存模型
    saver.save(sess,'net/my_net.ckpt')

二、模型利用（用于断点续训）

原图代码

saver = tf.train.Saver()

with tf.Session() as sess:
    saver.restore(sess,'net/my_net.ckpt')
    print(sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels}))

更详细参考博客：https://blog.csdn.net/sinat_34474705/article/details/78995196

 

三、模型预测

import tensorflow as tf
import numpy as np

sess=tf.Session() 
saver = tf.train.import_meta_graph('my_net.ckpt.meta') #导入完整的Tensorflow图；比如所有的变量、运算、集合等
saver.restore(sess,tf.train.latest_checkpoint('./'))   #导入权值、偏置值、梯度等
    
graph = tf.get_default_graph()
w1 = graph.get_tensor_by_name("x:0")  #得到占位符

prediction = graph.get_tensor_by_name("pre:0")#激活函数的节点

sess.run(prediction,{w1:np.ones((1,784))})

结果：
array([[6.3214763e-03, 3.0254640e-15, 3.4949583e-01, 7.1336553e-02,
        1.4892056e-10, 3.7446508e-01, 1.0758874e-04, 2.3885693e-07,
        1.9827317e-01, 8.0683282e-09]], dtype=float32) #每一种分类的概率