Tensorflow study notes-save and reload training results

Save the trained weights and biases:

# 定义saver，用于保存或重载训练好的权重和偏置
saver = tf.train.Saver()
with tf.Session() as sess:
	# 保存训练好的权重和偏置
	saver.save(sess, 'net/net.ckpt')

Reload trained weights and biases:

# 定义saver，用于保存或重载训练好的权重和偏置
saver = tf.train.Saver()
with tf.Session() as sess:
	# 重载训练好的权重和偏置
	saver.restore(sess, 'net/net.ckpt')

The content of train.py is as follows:

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

# 获取数据集
# one_hot设置为True，将标签数据转化为0/1，如[1,0,0,0,0,0,0,0,0,0]
mnist=input_data.read_data_sets('MNIST_data',one_hot=True)

# 定义一个批次的大小
batch_size=100
n_batch=mnist.train.num_examples//batch_size

# 变量分析
def variable_summaries(var):
	with tf.name_scope('summaries'):
		mean=tf.reduce_mean(var)
		stddev=tf.sqrt(tf.reduce_mean(tf.square(var-mean)))
		tf.summary.scalar('mean',mean)
		tf.summary.scalar('stddev',stddev)
		tf.summary.scalar('max',tf.reduce_max(var))
		tf.summary.scalar('min',tf.reduce_min(var))
		tf.summary.histogram('histogram',var) #直方图

# 定义三个placeholder
# 行数值为None，None可以取任意数，本例中将取值100，即取决于pitch_size
# 列数值为784，因为输入图像尺寸已由28*28转换为1*784
with tf.name_scope('input'):
	x=tf.placeholder(tf.float32,[None,784],name='x_input')
	y=tf.placeholder(tf.float32,[None,10],name='y_input')

with tf.name_scope('keep_prob'):
	keep_prob=tf.placeholder(tf.float32)

# 定义学习率
with tf.name_scope('lr'):
	lr=tf.Variable(0.001,dtype=tf.float32)

# 定义一个神经网络
with tf.name_scope('l1'):
	# 权重初始值为0不是最优的，应该设置为满足截断正态分布的随机数，收敛速度更快
	w1=tf.Variable(tf.truncated_normal([784,1000],stddev=0.1),name='w1')
	# 权重分析
	variable_summaries(w1)
	# 偏置初始值为0不是最优的，可以设置为0.1，收敛速度更快
	b1=tf.Variable(tf.zeros([1000])+0.1,name='b1')
	# 偏置分析
	variable_summaries(b1)
	# 引入激活函数
	l1=tf.nn.tanh(tf.matmul(x,w1)+b1,name='l1')
	# 引入dropout
	l1_drop=tf.nn.dropout(l1,keep_prob,name='l1_drop')

with tf.name_scope('l2'):
	w2=tf.Variable(tf.truncated_normal([1000,100],stddev=0.1),name='w2')
	variable_summaries(w2)
	b2=tf.Variable(tf.zeros([100])+0.1,name='b2')
	variable_summaries(b2)
	l2=tf.nn.tanh(tf.matmul(l1_drop,w2)+b2,name='l2')
	l2_drop=tf.nn.dropout(l2,keep_prob,name='l2_drop')

with tf.name_scope('output'):
	w3=tf.Variable(tf.truncated_normal([100,10],stddev=0.1),name='w3')
	variable_summaries(w3)
	b3=tf.Variable(tf.zeros([10])+0.1,name='b3')
	variable_summaries(b3)
	# softmax的作用是将tf.matmul(l2_drop,w3)+b3的结果转换为概率值
	# 假设tf.matmul(l2_drop,w3)+b3的结果为[1,5,3]，转换为概率值为[0.016,0.867,0.117]
	prediction=tf.nn.softmax(tf.matmul(l2_drop,w3)+b3)

# 定义损失函数
with tf.name_scope('loss'):
	# 由于输出神经元为softmax，交叉熵损失函数比均方误差损失函数收敛速度更快
	# loss=tf.reduce_mean(tf.square(y-prediction))
	loss=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=prediction))
	# loss分析
	tf.summary.scalar('loss',loss)

# 定义训练方式
with tf.name_scope('train'):
	# 优化器通过调整loss里的参数，使loss不断减小
	# AdamOptimizer比GradientDescentOptimizer收敛速度更快
	# train=tf.train.GradientDescentOptimizer(0.2).minimize(loss)
	train=tf.train.AdamOptimizer(lr).minimize(loss)

# 计算准确率
with tf.name_scope('accuracy'):
	# tf.argmax返回第一个参数中最大值的下标
	# tf.equal比较两个参数是否相等，返回True或False
	correct_prediction=tf.equal(tf.argmax(y,1),tf.argmax(prediction,1))
	# tf.cast将布尔类型转换为浮点类型
	accuracy=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
	# accuracy分析
	tf.summary.scalar('accuracy',accuracy)

# 合并所有summary
merged=tf.summary.merge_all()

# 定义saver，用于保存或重载训练好的权重和偏置
saver = tf.train.Saver()

with tf.Session() as sess:
	# 变量初始化
	sess.run(tf.global_variables_initializer())
	# 生成计算图
	writer=tf.summary.FileWriter('logs',sess.graph)
	# epoch为周期数，所有批次训练完为一个周期
	for epoch in range(1):
		# 调整学习率
		sess.run(tf.assign(lr,0.001*(0.95**epoch)))
		for batch in range(n_batch):
			# 每次取出batch_size条数据进行训练
			batch_xs, batch_ys = mnist.train.next_batch(batch_size)
			summary, _ = sess.run([merged, train], feed_dict={
    
    x:batch_xs, y:batch_ys, keep_prob:0.9})
		# 将summary添加到计算图中
		writer.add_summary(summary,epoch)
		learning_rate=sess.run(lr)
		test_acc = sess.run(accuracy,feed_dict={
    
    x:mnist.test.images,y:mnist.test.labels,keep_prob:0.9})
		train_acc = sess.run(accuracy,feed_dict={
    
    x:mnist.train.images,y:mnist.train.labels,keep_prob:0.9})
		print('epoch=',epoch,' ','learning_rate=%.7f' % learning_rate,' ','test_acc=',test_acc,' ','train_acc=',train_acc)
	# 保存训练好的权重和偏置
	saver.save(sess, 'net/net.ckpt')

The content of test.py is as follows:

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

# 获取数据集
# one_hot设置为True，将标签数据转化为0/1，如[1,0,0,0,0,0,0,0,0,0]
mnist=input_data.read_data_sets('MNIST_data',one_hot=True)

# 定义一个批次的大小
batch_size=100
n_batch=mnist.train.num_examples//batch_size

# 变量分析
def variable_summaries(var):
	with tf.name_scope('summaries'):
		mean=tf.reduce_mean(var)
		stddev=tf.sqrt(tf.reduce_mean(tf.square(var-mean)))
		tf.summary.scalar('mean',mean)
		tf.summary.scalar('stddev',stddev)
		tf.summary.scalar('max',tf.reduce_max(var))
		tf.summary.scalar('min',tf.reduce_min(var))
		tf.summary.histogram('histogram',var) #直方图

# 定义三个placeholder
# 行数值为None，None可以取任意数，本例中将取值100，即取决于pitch_size
# 列数值为784，因为输入图像尺寸已由28*28转换为1*784
with tf.name_scope('input'):
	x=tf.placeholder(tf.float32,[None,784],name='x_input')
	y=tf.placeholder(tf.float32,[None,10],name='y_input')

with tf.name_scope('keep_prob'):
	keep_prob=tf.placeholder(tf.float32)

# 定义学习率
with tf.name_scope('lr'):
	lr=tf.Variable(0.001,dtype=tf.float32)

# 定义一个神经网络
with tf.name_scope('l1'):
	# 权重初始值为0不是最优的，应该设置为满足截断正态分布的随机数，收敛速度更快
	w1=tf.Variable(tf.truncated_normal([784,1000],stddev=0.1),name='w1')
	# 权重分析
	variable_summaries(w1)
	# 偏置初始值为0不是最优的，可以设置为0.1，收敛速度更快
	b1=tf.Variable(tf.zeros([1000])+0.1,name='b1')
	# 偏置分析
	variable_summaries(b1)
	# 引入激活函数
	l1=tf.nn.tanh(tf.matmul(x,w1)+b1,name='l1')
	# 引入dropout
	l1_drop=tf.nn.dropout(l1,keep_prob,name='l1_drop')

with tf.name_scope('l2'):
	w2=tf.Variable(tf.truncated_normal([1000,100],stddev=0.1),name='w2')
	variable_summaries(w2)
	b2=tf.Variable(tf.zeros([100])+0.1,name='b2')
	variable_summaries(b2)
	l2=tf.nn.tanh(tf.matmul(l1_drop,w2)+b2,name='l2')
	l2_drop=tf.nn.dropout(l2,keep_prob,name='l2_drop')

with tf.name_scope('output'):
	w3=tf.Variable(tf.truncated_normal([100,10],stddev=0.1),name='w3')
	variable_summaries(w3)
	b3=tf.Variable(tf.zeros([10])+0.1,name='b3')
	variable_summaries(b3)
	# softmax的作用是将tf.matmul(l2_drop,w3)+b3的结果转换为概率值
	# 假设tf.matmul(l2_drop,w3)+b3的结果为[1,5,3]，转换为概率值为[0.016,0.867,0.117]
	prediction=tf.nn.softmax(tf.matmul(l2_drop,w3)+b3)

# 定义损失函数
with tf.name_scope('loss'):
	# 由于输出神经元为softmax，交叉熵损失函数比均方误差损失函数收敛速度更快
	# loss=tf.reduce_mean(tf.square(y-prediction))
	loss=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=prediction))
	# loss分析
	tf.summary.scalar('loss',loss)

# 定义训练方式
with tf.name_scope('train'):
	# 优化器通过调整loss里的参数，使loss不断减小
	# AdamOptimizer比GradientDescentOptimizer收敛速度更快
	# train=tf.train.GradientDescentOptimizer(0.2).minimize(loss)
	train=tf.train.AdamOptimizer(lr).minimize(loss)

# 计算准确率
with tf.name_scope('accuracy'):
	# tf.argmax返回第一个参数中最大值的下标
	# tf.equal比较两个参数是否相等，返回True或False
	correct_prediction=tf.equal(tf.argmax(y,1),tf.argmax(prediction,1))
	# tf.cast将布尔类型转换为浮点类型
	accuracy=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
	# accuracy分析
	tf.summary.scalar('accuracy',accuracy)

# 合并所有summary
merged=tf.summary.merge_all()

# 定义saver，用于保存或重载训练好的权重和偏置
saver = tf.train.Saver()

with tf.Session() as sess:
	# 变量初始化
	sess.run(tf.global_variables_initializer())
	# 重载训练好的权重和偏置
	saver.restore(sess, 'net/net.ckpt')
	# 打印测试结果
	print(sess.run(accuracy,feed_dict={
    
    x:mnist.test.images,y:mnist.test.labels,keep_prob:0.9}))

operation result: