TensorBoard食用指南

TensorBoard 数据形式

数据类型

1: 标量Scalars

2: 图片Images

3: 音频Audio

4: 计算图Graph

5: 数据分布Distribution

6: 直方图Histogtams

7: 嵌入向量Embeddings

添加数据代码

import tensorflow as tf

tf.summary.scalar(name, scalar)
tf.summary.image(name, image)
tf.summary.histogram(name, histogram)
tf.summary.audio(name, audio)
tf.summary.distribution(name, distribution)
tf.summary.graph(name, graph)
tf.summary.embeddings(name, embedding)

TensorBoard 可视化过程

1. 建立计算图graph, 你可以从改图中得到你想要的数据信息
2. 在计算图的结点防止summary operation用来记录信息
3. 运行summary operation, 只是建立了summary operation而不run, operation并不会真正执行. 有时, 我们有很多summary operation需要运行, 但是一个一个run会很麻烦. 这时, 可以利用tf.summary.merge_all()将所有结点合并成一个结点, 运行这个结点, 就可以产生所有设置的summary Data.
4. 将数据保存得到本地磁盘, tf.summary.FileWriter(logs_path, graph=my_graph)
5. 运行整个程序，并在命令行输入运行tensorboard的指令，之后打开web端可查看可视化的结果.

Python代码

该demo使用了tensorflow提供的手写体识别数据

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

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

logs_path = "five_layers_log"
batch_size = 100
learning_rate = 0.003
training_epochs = 10

mnist = input_data.read_data_sets("../data", one_hot=True)
x = tf.placeholder(tf.float32, [None, 784])
y = tf.placeholder(tf.float32, [None, 10])

# five layers neural network
first_layer = 200
second_layer = 100
third_layer = 60
fourth_layer = 30
fifth_layer = 10

# weight and bias
W1 = tf.Variable(tf.truncated_normal([784, first_layer], stddev=0.1))
B1 = tf.Variable(tf.zeros([first_layer]))

W2 = tf.Variable(tf.truncated_normal([first_layer, second_layer], stddev=0.1))
B2 = tf.Variable(tf.zeros([second_layer]))

W3 = tf.Variable(tf.truncated_normal([second_layer, third_layer], stddev=0.1))
B3 = tf.Variable(tf.zeros([third_layer]))

W4 = tf.Variable(tf.truncated_normal([third_layer, fourth_layer], stddev=0.1))
B4 = tf.Variable(tf.zeros([fourth_layer]))

W5 = tf.Variable(tf.truncated_normal([fourth_layer, fifth_layer], stddev=0.1))
B5 = tf.Variable(tf.zeros([fifth_layer]))


XX = tf.reshape(x, [-1, 784])

Y1 = tf.nn.sigmoid(tf.matmul(x, W1) + B1)
Y2 = tf.nn.sigmoid(tf.matmul(Y1, W2) + B2)
Y3 = tf.nn.sigmoid(tf.matmul(Y2, W3) + B3)
Y4 = tf.nn.sigmoid(tf.matmul(Y3, W4) + B4)
Ylogits = tf.matmul(Y4, W5) + B5
Y = tf.nn.softmax(Ylogits)

cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits, labels=y)
cross_entropy = tf.reduce_mean(cross_entropy)*100
correct_prediction = tf.equal(tf.argmax(Y, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
train_step = tf.train.AdamOptimizer(learning_rate).minimize(cross_entropy)

tf.summary.scalar("cost", cross_entropy)
tf.summary.scalar("accuracy", accuracy)
tf.summary.image("input_image", tf.reshape(x, [-1, 28, 28, 1]), 10)
summary_op = tf.summary.merge_all()


init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    writer = tf.summary.FileWriter(logs_path, graph=tf.get_default_graph())
    for epoch in range(training_epochs):
        batch_count = int(mnist.train.num_examples/batch_size)
        for i in range(batch_count):
            batch_x, batch_y = mnist.train.next_batch(batch_size)
            _, summary = sess.run([train_step, summary_op], feed_dict={x: batch_x, y: batch_y})
            writer.add_summary(summary, epoch*batch_count+1)
        print("Epoch", epoch)
    print("Accuracy", accuracy.eval(feed_dict={x: mnist.test.images, y: mnist.test.labels}))
    print("Done")

查看结果

# 运行Python程序
$ python3 hw_recognition.py # python 文件名

# 查看tensorboard
$ tensorboard --logdir='five_layers_log' # 程序中保存的数据路径

Web端TensorBoard

用浏览器打开网址 http://127.0.1.1:6006 查看. 下面是部分截图.

1. scalar

2. image

3. graph

END