Minst训练模型源码:
import tensorflow as tf #导入tensorflow库
from tensorflow.examples.tutorials.mnist import input_data
from tensorflow.python.framework import graph_util
import pylab
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
tf.reset_default_graph()
# tf Graph Input
x = tf.placeholder(tf.float32, [None, 784]) # mnist data维度 28*28=784
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 数字=> 10 classes
# Set model weights
W = tf.Variable(tf.random_normal([784, 10]))
b = tf.Variable(tf.zeros([10]))
# 构建模型
pred = tf.nn.softmax(tf.matmul(x, W) + b) # Softmax分类
# Minimize error using cross entropy
cost = tf.reduce_mean(-tf.reduce_sum(y*tf.log(pred), reduction_indices=1))
#参数设置
learning_rate = 0.01
# 使用梯度下降优化器
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)
training_epochs = 25
batch_size = 100
display_step = 1
saver = tf.train.Saver()
model_ckpt_path = "model/521model.ckpt"
model_pb_path = "model/521model.pb"
# 启动session
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())# Initializing OP
# 启动循环开始训练
for epoch in range(training_epochs):
avg_cost = 0.
total_batch = int(mnist.train.num_examples/batch_size)
# 遍历全部数据集
for i in range(total_batch):
batch_xs, batch_ys = mnist.train.next_batch(batch_size)
# Run optimization op (backprop) and cost op (to get loss value)
_, c = sess.run([optimizer, cost], feed_dict={x: batch_xs,
y: batch_ys})
# Compute average loss
avg_cost += c / total_batch
# 显示训练中的详细信息
if (epoch+1) % display_step == 0:
print ("Epoch:", '%04d' % (epoch+1), "cost=", "{:.9f}".format(avg_cost))
# print(sess.run(W))
# print(sess.run(b))
print( " Finished!")
# 测试 model
correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
# 计算准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
print ("Accuracy:", accuracy.eval({x: mnist.test.images, y: mnist.test.labels}))
# Save model weights to disk
save_path = saver.save(sess, model_ckpt_path)
print("Model saved in file: %s" % save_path)
graph_def = tf.get_default_graph().as_graph_def()
# 保存对应计算节点的名称。
output_graph_def = graph_util.convert_variables_to_constants(sess, graph_def, ['add'])
# 将导出的模型存入文件中
with tf.gfile.GFile(model_pb_path, "wb") as f:
f.write(output_graph_def.SerializeToString())
#读取模型
print("Starting 2nd session...")
with tf.Session() as sess:
# Initialize variables
sess.run(tf.global_variables_initializer())
# Restore model weights from previously saved model
saver.restore(sess, model_ckpt_path)
# 测试 model
correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
# 计算准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
print ("Accuracy:", accuracy.eval({x: mnist.test.images, y: mnist.test.labels}))
output = tf.argmax(pred, 1)
batch_xs, batch_ys = mnist.train.next_batch(2)
outputval,predv = sess.run([output,pred], feed_dict={x: batch_xs})
print(outputval,predv,batch_ys)
im = batch_xs[0]
print(im.shape)
im = im.reshape(-1,28)
print(im.shape)
pylab.imshow(im)
pylab.show()
im = batch_xs[1]
print(im.shape)
im = im.reshape(-1,28)
print(im.shape)
pylab.imshow(im)
pylab.show()
运行结果
执行完后,可以看到,保存的模型
保存模型的方法:
保存ckpt格式的模型:
# Save model weights to disk
save_path = saver.save(sess, model_ckpt_path)
print("Model saved in file: %s" % save_path)
保存pb格式的模型
graph_def = tf.get_default_graph().as_graph_def()
# 保存对应计算节点的名称。
output_graph_def = graph_util.convert_variables_to_constants(sess, graph_def, ['add'])
# 将导出的模型存入文件中
with tf.gfile.GFile(model_pb_path, "wb") as f:
f.write(output_graph_def.SerializeToString())