【Tensorflow】Saver使用

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

mnist = input_data.read_data_sets('MNIST_DATA',one_hot=True)

def weights_variable(shape):
    return tf.Variable(tf.truncated_normal(shape,stddev=0.1))

def bias_variable(shape):
    return tf.constant(0.1,shape=shape)

def conv2d(x, Kernel):
    return tf.nn.conv2d(x, Kernel, strides=[1,1,1,1],padding='SAME')

def max_pool_2x2(x):
    return tf.nn.max_pool(x, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')

xs = tf.placeholder(tf.float32, [None, 784])
ys = tf.placeholder(tf.float32, [None, 10])
keep_prob = tf.placeholder(tf.float32)

x_image = tf.reshape(xs, [-1, 28, 28, 1])

# Build net
with tf.name_scope('Conv1'):
    w_conv1 = weights_variable([5,5,1,32])
    b_conv1 = bias_variable([32])
    h_conv1 = tf.nn.relu(conv2d(x_image, w_conv1) + b_conv1)
    h_pool1 = max_pool_2x2(h_conv1)

with tf.name_scope('Conv2'):
    w_conv2 = weights_variable([5,5,32,64])
    b_conv2 = bias_variable([64])
    h_conv2 = tf.nn.relu(conv2d(h_pool1, w_conv2) + b_conv2)
    h_pool2 = max_pool_2x2(h_conv2)

with tf.name_scope('Full1'):
    fc_input = tf.reshape(h_pool2, [-1, 7*7*64])
    w_fc1 = weights_variable([7*7*64, 1024])
    b_fc1 = bias_variable([1024])
    h_fc1 = tf.nn.relu(tf.matmul(fc_input, w_fc1) + b_fc1)
    h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob=keep_prob)

with tf.name_scope('Full2'):
    w_fc2 = weights_variable([1024, 10])
    b_fc2 = bias_variable([10])
    prediction = tf.nn.softmax(tf.matmul(h_fc1_drop, w_fc2) + b_fc2)

with tf.name_scope('Loss'):
    cross_entropy = tf.reduce_mean(-tf.reduce_sum(ys * tf.log(prediction)))
    tf.summary.scalar('Cross_entropy', cross_entropy)

with tf.name_scope('Accuracy'):
    correct = tf.equal(tf.argmax(ys, 1), tf.argmax(prediction, 1))
    accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))
    tf.summary.scalar('Accuracy',accuracy)

with tf.name_scope('Train'):
    train = tf.train.AdamOptimizer(0.001).minimize(cross_entropy)

init = tf.global_variables_initializer()
saver = tf.train.Saver()
with tf.Session() as sess:
    saver.restore(sess, 'model/cnn_saver.ckpt')
    merged = tf.summary.merge_all()
    writer = tf.summary.FileWriter('log/',sess.graph)
    sess.run(init)
    for step in range(1,1001):
        batch_xs, batch_ys = mnist.train.next_batch(64)
        sess.run(train,feed_dict={xs: batch_xs, ys:batch_ys, keep_prob:0.5})
        if step % 10 == 0:
            print('Step: ',step,' || Accuracy: ',sess.run(accuracy,feed_dict={xs: batch_xs,ys: batch_ys,keep_prob:1}))
            writer.add_summary(sess.run(merged, feed_dict={xs: batch_xs,ys: batch_ys,keep_prob:0.5}) ,step)
    saver.save(sess, 'model/cnn_saver.ckpt')

以cnn分类mnist为测试。首先构建一个tf.train.Saver()对象，然后使用save()保存网络参数，使用restore()加载网络参数。测试如下：

存储

代码

import tensorflow as tf

data = tf.Variable(tf.random_normal([2,2]))
saver = tf.train.Saver()
init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init)
    print(sess.run(data))
    saver.save(sess, 'model/test.ckpt')

输出

 [[ 1.20965123 -1.97269511]
 [-0.66441923 -1.37852442]]

加载

代码

import tensorflow as tf

data = tf.Variable(tf.random_normal([2,2]))
saver = tf.train.Saver()
init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init)
    saver.restore(sess, 'model/test.ckpt')
    print(sess.run(data))

输出

 [[ 1.20965123 -1.97269511]
 [-0.66441923 -1.37852442]]