TensorFlow - 可视化工具TensorBoard

TensorBoard

TensorFlow自带的可视化工具,能够以直观的流程图的方式,清楚展示出整个神经网络的结构和框架,便于理解模型和发现问题。

启动TensorBoard

  • 使用命令“tensorboard --logdir=path/to/log-directory”(或者“python -m tensorboard.main”);
  • 参数logdir指向FileWriter将数据序列化的目录,建议在logdir上一级目录执行此命令;
  • TensorBoard运行后,在浏览器输入“localhost:6006”即可查看TensorBoard;

帮助信息

  • 使用“tensorboard --help”查看tensorboard的详细参数 

示例

程序代码

 1 # coding=utf-8
 2 from __future__ import print_function
 3 import tensorflow as tf
 4 import numpy as np
 5 import matplotlib.pyplot as plt
 6 import os
 7 
 8 os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
 9 
10 
11 # ### 添加神经层
12 
13 
14 def add_layer(inputs, in_size, out_size, n_layer, activation_function=None):  # 参数n_layer用来标识层数
15     layer_name = 'layer{}'.format(n_layer)
16     with tf.name_scope(layer_name):  # 使用with tf.name_scope定义图层,并指定在可视化图层中的显示名称
17         with tf.name_scope('weights'):  # 定义图层并指定名称,注意这里是上一图层的子图层
18             Weights = tf.Variable(tf.random_normal([in_size, out_size]), name='W')  # 参数name指定名称
19             tf.summary.histogram(layer_name + '/weights', Weights)  # 生成直方图summary,指定图表名称和记录的变量
20         with tf.name_scope('biases'):  # 定义图层并指定名称
21             biases = tf.Variable(tf.zeros([1, out_size]) + 0.1, name='b')  # 参数name指定名称
22             tf.summary.histogram(layer_name + '/biases', biases)  # 生成直方图summary
23         with tf.name_scope('Wx_plus_b'):  # 定义图层并指定名称
24             Wx_plus_b = tf.matmul(inputs, Weights) + biases
25         if activation_function is None:
26             outputs = Wx_plus_b
27         else:
28             outputs = activation_function(Wx_plus_b)
29         tf.summary.histogram(layer_name + '/outputs', outputs)  # 生成直方图summary
30         return outputs
31 
32 
33 # ### 构建数据
34 x_data = np.linspace(-1, 1, 300, dtype=np.float32)[:, np.newaxis]
35 noise = np.random.normal(0, 0.05, x_data.shape).astype(np.float32)
36 y_data = np.square(x_data) - 0.5 + noise
37 
38 # ### 搭建网络
39 with tf.name_scope('inputs'):  # 定义图层并指定名称
40     xs = tf.placeholder(tf.float32, [None, 1], name='x_input')  # 指定名称为x_input,也就是在可视化图层中的显示名称
41     ys = tf.placeholder(tf.float32, [None, 1], name='y_input')  # 指定名称为y_input
42 
43 h1 = add_layer(xs, 1, 10, n_layer=1, activation_function=tf.nn.relu)  # 隐藏层
44 prediction = add_layer(h1, 10, 1, n_layer=2, activation_function=None)  # 输出层
45 
46 with tf.name_scope('loss'):  # 定义图层并指定名称
47     loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction),
48                                         reduction_indices=[1]))
49     tf.summary.scalar('loss', loss)  # 用于标量的summary,loss在TensorBoard的event栏
50 
51 with tf.name_scope('train'):  # 定义图层并指定名称
52     train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
53 
54 sess = tf.Session()
55 merged = tf.summary.merge_all()  # 合并之前定义的所有summary操作
56 writer = tf.summary.FileWriter("logs/", sess.graph)  # 创建FileWriter对象和event文件,指定event文件的存放目录
57 init = tf.global_variables_initializer()
58 sess.run(init)
59 
60 # ### 结果可视化
61 fig = plt.figure()
62 ax = fig.add_subplot(1, 1, 1)
63 ax.scatter(x_data, y_data)
64 plt.ion()
65 plt.show()
66 
67 # ### 训练
68 for i in range(1001):
69     sess.run(train_step, feed_dict={xs: x_data, ys: y_data})
70     if i % 50 == 0:
71         result = sess.run(loss, feed_dict={xs: x_data, ys: y_data})
72         print("Steps:{}  Loss:{}".format(i, result))
73         rs = sess.run(merged, feed_dict={xs: x_data, ys: y_data})  # 在sess.run中运行
74         writer.add_summary(rs, i)
75         try:
76             ax.lines.remove(lines[0])
77         except Exception:
78             pass
79         prediction_value = sess.run(prediction, feed_dict={xs: x_data})
80         lines = ax.plot(x_data, prediction_value, 'r-', lw=5)
81         plt.pause(0.2)
82 
83 # ### TensorBoard
84 # TensorFlow自带的可视化工具,能够以直观的流程图的方式,清楚展示出整个神经网络的结构和框架,便于理解模型和发现问题;
85 #   - 可视化学习:https://www.tensorflow.org/guide/summaries_and_tensorboard
86 #   - 图的直观展示:https://www.tensorflow.org/guide/graph_viz;
87 #   - 直方图信息中心:https://www.tensorflow.org/guide/tensorboard_histograms
88 #
89 # ### 启动TensorBoard
90 # 使用命令“tensorboard --logdir=path/to/log-directory”(或者“python -m tensorboard.main”);
91 # 参数logdir指向FileWriter将数据序列化的目录,建议在logdir上一级目录执行此命令;
92 # TensorBoard运行后,在浏览器输入“localhost:6006”即可查看TensorBoard;

 

程序运行结果

运行过程中显示的图形:

 

某一次运行的命令行输出:

Steps:0  Loss:0.19870562851428986
Steps:50  Loss:0.006314810831099749
Steps:100  Loss:0.0050856382586061954
Steps:150  Loss:0.0048223137855529785
Steps:200  Loss:0.004617161583155394
Steps:250  Loss:0.004429362714290619
Steps:300  Loss:0.004260621033608913
Steps:350  Loss:0.004093690309673548
Steps:400  Loss:0.003932977095246315
Steps:450  Loss:0.0038178395479917526
Steps:500  Loss:0.003722294932231307
Steps:550  Loss:0.003660505171865225
Steps:600  Loss:0.0036110866349190474
Steps:650  Loss:0.0035716891288757324
Steps:700  Loss:0.0035362064372748137
Steps:750  Loss:0.0034975067246705294
Steps:800  Loss:0.003465239657089114
Steps:850  Loss:0.003431882942095399
Steps:900  Loss:0.00339301535859704
Steps:950  Loss:0.0033665322698652744
Steps:1000  Loss:0.003349516075104475

 

生成的TensorBoard文件:

(mlcc) D:\Anliven\Anliven-Code\PycharmProjects\TempTest>dir logs
 驱动器 D 中的卷是 Files
 卷的序列号是 ACF9-2E0E

 D:\Anliven\Anliven-Code\PycharmProjects\TempTest\logs 的目录

2019/02/24  23:41    <DIR>          .
2019/02/24  23:41    <DIR>          ..
2019/02/24  23:41           137,221 events.out.tfevents.1551022894.DESKTOP-68OFQFP
               1 个文件        137,221 字节
               2 个目录 219,401,887,744 可用字节

(mlcc) D:\Anliven\Anliven-Code\PycharmProjects\TempTest>

 

启动与TensorBoard

执行下面的启动命令,然后在浏览器中输入“http://localhost:6006/”查看。

(mlcc) D:\Anliven\Anliven-Code\PycharmProjects\TempTest>tensorboard --logdir=logs
TensorBoard 1.12.0 at http://DESKTOP-68OFQFP:6006 (Press CTRL+C to quit)

 

栏目Scalars

 

栏目Graphs

  • 通过鼠标滑轮可以改变显示大小和位置 
  • 鼠标双击“+”标识可以查看进一步的信息
  • 可以将指定图层从主图层移出,单独显示

 

栏目Distributions

 

栏目histograms

 

posted @ 2019-02-23 23:38  Anliven  阅读(1925)  评论(0编辑  收藏  举报