我正在研究一些代码，以了解如何在tensorflow中保存和恢复检查点。为此，我实现了一个简单的神经网络，它使用MNIST数字并保存.ckpt文件，如下所示：
<html>
<head>
<title>children</title>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
</head>
<body bgcolor="#FFFFb0" leftmargin="0" topmargin="0" marginwidth="0" marginheight="0" onLoad="MM_preloadImages('1-b.jpg','2-b.jpg','3-b.jpg','4-b.jpg','5-b.jpg','6-b.jpg','7-b','8-b','9-b')">
<center>
<div class="wrapper">
<h1> Some title </h1>
<div class="pics-wrapper">
<div class="left-pics">
<p>
<img src="https://media-cdn.tripadvisor.com/media/photo-s/0e/9a/e3/1d/freedom-tower.jpg" alt="" width="100" height="100" onMouseOver="MM_swapImage('biggest_one','','https://media-cdn.tripadvisor.com/media/photo-s/0e/9a/e3/1d/freedom-tower.jpg',1)" onMouseOut="MM_swapImgRestore()"
/>
<img src="https://upload.wikimedia.org/wikipedia/commons/thumb/b/b0/Empire_State_Building_%28HDR%29.jpg/150px-Empire_State_Building_%28HDR%29.jpg" alt="" width="100" height="100" onMouseOver="MM_swapImage('biggest_one','','https://upload.wikimedia.org/wikipedia/commons/thumb/b/b0/Empire_State_Building_%28HDR%29.jpg/150px-Empire_State_Building_%28HDR%29.jpg',1)"
onMouseOut="MM_swapImgRestore()" />
<img src="https://www.nycgo.com/images/uploads/homepage/Empire-State-Building-Observatory-Tom-Perry-2618.jpg" alt="" width="100" height="100" onMouseOver="MM_swapImage('biggest_one','','https://www.nycgo.com/images/uploads/homepage/Empire-State-Building-Observatory-Tom-Perry-2618.jpg',1)"
onMouseOut="MM_swapImgRestore()" />
</p>
</div>
<table id="Table_01" border="0" width="851" cellspacing="0" cellpadding="0">
<tr>
<td colspan="2"> </td>
<td colspan="7" rowspan="9"><img id="biggest_one" style="display: block; margin-left: auto; margin-right: auto;" src="https://www.nycgo.com/images/uploads/homepage/Empire-State-Building-Observatory-Tom-Perry-2618.jpg" alt="" width="480" height="640" name="biggest_one"
/></td>
<td colspan="2" rowspan="10"> </td>
<td><img src="photo_over/spacer.gif" alt="" width="1" height="3" /></td>
</tr>
</table>
<div class="right-pics">
<p>
<img src="https://media.timeout.com/images/103678916/image.jpg" alt="" width="100" height="100" onMouseOver="MM_swapImage('biggest_one','','https://media.timeout.com/images/103678916/image.jpg',1)" onMouseOut="MM_swapImgRestore()" />
<img src="https://lonelyplanetimages.imgix.net/mastheads/GettyImages-538096543_medium.jpg?sharp=10&vib=20&w=1200" alt="" width="100" height="100" onMouseOver="MM_swapImage('biggest_one','','https://lonelyplanetimages.imgix.net/mastheads/GettyImages-538096543_medium.jpg?sharp=10&vib=20&w=1200',1)"
onMouseOut="MM_swapImgRestore()" />
<img src="https://media.timeout.com/images/103444978/image.jpg" alt="" width="100" height="100" onMouseOver="MM_swapImage('biggest_one','','https://media.timeout.com/images/103444978/image.jpg',1)" onMouseOut="MM_swapImgRestore()" />
</p>
</div>
</div>
<div class="bottom-pic">
<img src="https://res.cloudinary.com/simpleview/image/upload/c_fill,f_auto,h_510,q_75,w_1280/v1/clients/newyorkstate/dumbo_brooklyn_bridge_park_julienne_schaer_0732_fdff1de5-9486-480c-a1c7-4135d784c75f.jpg" alt="" width="100" height="100" onMouseOver="MM_swapImage('biggest_one','','https://res.cloudinary.com/simpleview/image/upload/c_fill,f_auto,h_510,q_75,w_1280/v1/clients/newyorkstate/dumbo_brooklyn_bridge_park_julienne_schaer_0732_fdff1de5-9486-480c-a1c7-4135d784c75f.jpg',1)" onMouseOut="MM_swapImgRestore()" />
</div>
</div>
</center>
</body>
</html>这部分效果很好，我将.ckpt文件保存在正确的目录中。当我尝试恢复模型以尝试再次处理它时，问题就出现了。我使用以下代码恢复模型：
from tensorflow.examples.tutorials.mnist import input_data
import numpy as np
learning_rate = 0.001
n_input = 784 # MNIST data input (img shape = 28*28)
n_classes = 10 # MNIST total classes 0-9
#import MNIST data
mnist = input_data.read_data_sets('.', one_hot = True)
#Features and Labels
features = tf.placeholder(tf.float32, [None, n_input])
labels = tf.placeholder(tf.float32, [None, n_classes])
#Weights and biases
weights = tf.Variable(tf.random_normal([n_input, n_classes]))
bias = tf.Variable(tf.random_normal([n_classes]))
#logits = xW + b
logits = tf.add(tf.matmul(features, weights), bias)
#Define loss and optimizer
cost = tf.reduce_mean(\
tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=labels))
optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)\
.minimize(cost)
# Calculate accuracy
correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(labels, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
import math
save_file = './train_model.ckpt'
batch_size = 128
n_epochs = 100
saver = tf.train.Saver()
# Launch the graph
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
# Training cycle
for epoch in range(n_epochs):
total_batch = math.ceil(mnist.train.num_examples / batch_size)
# Loop over all batches
for i in range(total_batch):
batch_features, batch_labels = mnist.train.next_batch(batch_size)
sess.run(
optimizer,
feed_dict={features: batch_features, labels: batch_labels})
# Print status for every 10 epochs
if epoch % 10 == 0:
valid_accuracy = sess.run(
accuracy,
feed_dict={
features: mnist.validation.images,
labels: mnist.validation.labels})
print('Epoch {:<3} - Validation Accuracy: {}'.format(
epoch,
valid_accuracy))
# Save the model
saver.save(sess, save_file)
print('Trained Model Saved.')
并最终得到错误：saver = tf.train.Saver()
with tf.Session() as sess:
saver.restore(sess, 'train_model.ckpt.meta')
print('model restored')
不太确定，这里的错误是什么。任何帮助表示赞赏。提前致谢
答案 0 :(得分：1)
您应该告诉Saver哪个Variables要恢复，默认Saver会从默认图表中获取所有Variables。
与您的情况一样，您应该在saver = tf.train.Saver()
答案 1 :(得分：0)
Graph与Session不同。图是连接张量的一组操作，每个张量是一组值的符号表示。 Session为Variable张量分配特定值，并允许您在该图表中进行run次操作。
chkpt文件保存变量值 - 即保存在权重和偏差中的值 - 但不保存图形本身。
解决方案很简单：重新运行图形构造（Session之前的所有内容，然后启动会话并从chkpt文件加载值。