生成模型的save和restore

1、save

saver = tf.train.Saver()

with tf.Session() as sess:
    ...
    # 保存模型
    saver.save(sess, 'save_path')

举例： 

# 载入数据集
mnist = input_data.read_data_sets("MNIST_data", one_hot=True)

# 每个批次100张照片
batch_size = 100
# 计算一共有多少个批次
n_batch = mnist.train.num_examples // batch_size

# 定义两个placeholder
x = tf.placeholder(tf.float32, [None, 784])
y = tf.placeholder(tf.float32, [None, 10])

# 创建一个简单的神经网络，输入层784个神经元，输出层10个神经元
W = tf.Variable(tf.zeros([784, 10]))
b = tf.Variable(tf.zeros([10]))
prediction = tf.nn.softmax(tf.matmul(x, W) + b)

# 二次代价函数
# loss = tf.reduce_mean(tf.square(y-prediction))
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=prediction))
# 使用梯度下降法
train_step = tf.train.GradientDescentOptimizer(0.2).minimize(loss)

# 初始化变量
init = tf.global_variables_initializer()

# 结果存放在一个布尔型列表中
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(prediction, 1))  # argmax返回一维张量中最大的值所在的位置
# 求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

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')

2、restore

saver = tf.train.Saver()

with tf.Session() as sess:
    ...
    saver.restore(sess,'net/my_net.ckpt')#载入训练的模型
    

举例：

#载入数据集
mnist = input_data.read_data_sets("MNIST_data",one_hot=True)

#每个批次100张照片
batch_size = 100
#计算一共有多少个批次
n_batch = mnist.train.num_examples // batch_size

#定义两个placeholder
x = tf.placeholder(tf.float32,[None,784])
y = tf.placeholder(tf.float32,[None,10])

#创建一个简单的神经网络，输入层784个神经元，输出层10个神经元
W = tf.Variable(tf.zeros([784,10]))
b = tf.Variable(tf.zeros([10]))
prediction = tf.nn.softmax(tf.matmul(x,W)+b)

#二次代价函数
# loss = tf.reduce_mean(tf.square(y-prediction))
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=prediction))
#使用梯度下降法
train_step = tf.train.GradientDescentOptimizer(0.2).minimize(loss)

#初始化变量
init = tf.global_variables_initializer()

#结果存放在一个布尔型列表中
correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(prediction,1))#argmax返回一维张量中最大的值所在的位置
#求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

saver = tf.train.Saver()

with tf.Session() as sess:
    sess.run(init)
    print(sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels}))
    saver.restore(sess,'net/my_net.ckpt')#载入训练的模型
    print(sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels}))