[tf]|[tensorboard]tensorflow学习笔记（3）

参考：莫烦Python: morvanzhou.github.io

with tf.name_scope():

• 为变量取别名（用于tensorboard上查看）

with tf.name_scope('Weights'):
             Weights = tf.Variable(tf.random_normal([in_size, out_size]),name = 'W')

• 上式别名为：W

with tf.name_scope('Weights_1'):
             Weights = tf.Variable(tf.random_normal([in_size, out_size]))

• 上式别名为：Weights_1

tf.summary.FileWriter()

sess = tf.Session()
writer = tf.summary.FileWriter("logs/",sess.graph)

• log是事件文件所在的目录，这里是工程目录下的log目录。第二个参数是事件文件要记录的图，也就是tensorflow默认的图。

完整代码

"""
Please note, this code is only for python 3+. If you are using python 2+, please modify the code accordingly.
"""
from __future__ import print_function
import tensorflow as tf
import numpy as np

def add_layer(inputs, in_size, out_size, activation_function=None):
#activation_function=None 默认为线性函数
    # add one more layer and return the output of this layer
    with tf.name_scope('layer'):
        with tf.name_scope('Weights'):
             Weights = tf.Variable(tf.random_normal([in_size, out_size]),name = 'W')
    #随机生成行为in_size，列为out_size的矩阵
        with tf.name_scope('biases'):
            biases = tf.Variable(tf.zeros([1, out_size]) + 0.1, name = 'b')
    #（一组向量）给予biases初值
        with tf.name_scope('Wx_plus_b'):
            Wx_plus_b = tf.matmul(inputs, Weights) + biases
    # Wx_plus_b
        if activation_function is None:
            #** Wx_plus_b
            outputs = Wx_plus_b
    #保持线性关系（不改变 Wx_plus_b）
        else:
            outputs = activation_function(Wx_plus_b)
    #relu  Wx_plus_b
        return outputs

# Make up some real data
#给出input的值x_data与y_data
x_data = np.linspace(-1,1,300)[:, np.newaxis]
noise = np.random.normal(0, 0.05, x_data.shape)
#给出噪点（使其变得更像真实数据）变为非连续数据的一组数据
y_data = np.square(x_data) - 0.5 + noise

# define placeholder for inputs to network
with tf.name_scope('inputs'):
    xs = tf.placeholder(tf.float32, [None, 1],name = 'x_input')
    #行不定，列为1，可以解为列向量
    ys = tf.placeholder(tf.float32, [None, 1],name = 'y_input')
# add hidden layer
l1 = add_layer(xs, 1, 10, activation_function=tf.nn.relu)
#输入层的神经元为1；隐藏层的神经元为10；输出层的神经元为1
# add output layer
prediction = add_layer(l1, 10, 1, activation_function=None)

# the error between prediction and real data
with tf.name_scope('loss'):
    loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction),
                     reduction_indices=[1]))
#每个例子的平方和再求平均值，最后把矩阵向列方向压缩成向量再向行方向压缩成数
with tf.name_scope('train'):
    train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
#学习率为0.1的梯度下降法减小loss的误差
sess = tf.Session()
writer = tf.summary.FileWriter("logs/",sess.graph)
# important step
# tf.initialize_all_variables() no long valid from
# 2017-03-02 if using tensorflow >= 0.12
if int((tf.__version__).split('.')[1]) < 12:
    init = tf.initialize_all_variables()
else:
    init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)

for i in range(1000):
    #迭代1000次
    # training
    sess.run(train_step, feed_dict={xs: x_data, ys: y_data})
    if i % 50 == 0:
        # to see the step improvement
        print(sess.run(loss, feed_dict={xs: x_data, ys: y_data}))

tensorboard使用步骤（Graph可视化）

• 操作系统：win10
• tensorboard版本：1.12.1
• tensorflow版本：1.12.0
• 更新日期：2018/12/30

1. 先编译完成上述代码
2. 在生成的logs文件目录下可找到生成文件，如图：

3. 打开cmd ，cd到logs文件夹的父文件夹，如图：

4. 执行命令

tensorboard --logdir=logs
#注意logs别加引号

​ ，如图：

5. 在Chroms地址栏输入网址：http://localhost:6006回车即可

http://localhost:6006

• pay attention：运行http://localhost:6006过程中不可退出此状态

summary用法

1. tf.summary.scalar()
   • 用来显示标量信息，其格式为：

tf.summary.scalar(tags, values, collections=None, name=None)

​ 例如：tf.summary.scalar(‘mean’, mean)

​ 一般在画loss,accuary时会用到这个函数。

2. tf.summary.histogram()
   • 用来显示直方图信息，其格式为：

tf.summary.histogram(tags, values, collections=None, name=None) 

​ 例如： tf.summary.histogram(‘histogram’, var)

​ 一般用来显示训练过程中变量的分布情况

3. tf.summary.merge_all()
   • merge_all 可以将所有summary全部保存到磁盘，以便tensorboard显示。如果没有特殊要求，一般用这一句就可一显示训练时的各种信息了。
   • 格式：

tf.summaries.merge_all(key='summaries')

4. tf.summary.FileWriter
   • 指定一个文件用来保存图。
   • 格式：

tf.summary.FileWriter(path,sess.graph)

• 参考：http://www.cnblogs.com/lyc-seu/p/8647792.html

tensorboard（histograms）

• 完整代码

"""
Please note, this code is only for python 3+. If you are using python 2+, please modify the code accordingly.
"""
from __future__ import print_function
import tensorflow as tf
import numpy as np

def add_layer(inputs, in_size, out_size,n_layer, activation_function=None):
#activation_function=None 默认为线性函数
    # add one more layer and return the output of this layer
    layer_name = 'layer%s' % n_layer
    with tf.name_scope(layer_name):
        with tf.name_scope('Weights'):
             Weights = tf.Variable(tf.random_normal([in_size, out_size]),name = 'W')
             tf.summary.histogram('value',Weights)
            #tag = 'value' ; value = Weights
    #随机生成行为in_size，列为out_size的矩阵
        with tf.name_scope('biases'):
            biases = tf.Variable(tf.zeros([1, out_size]) + 0.1, name = 'b')
            tf.summary.histogram('value', biases)
            #tag = 'value' ; value = biases
    #（一组向量）给予biases初值
        with tf.name_scope('Wx_plus_b'):
            Wx_plus_b = tf.matmul(inputs, Weights) + biases
    # Wx_plus_b
        if activation_function is None:
            #** Wx_plus_b
            outputs = Wx_plus_b
    #保持线性关系（不改变 Wx_plus_b）
        else:
            outputs = activation_function(Wx_plus_b)
        tf.summary.histogram('value', outputs)
    #relu  Wx_plus_b
        return outputs

# Make up some real data
#给出input的值x_data与y_data
x_data = np.linspace(-1,1,300)[:, np.newaxis]
noise = np.random.normal(0, 0.05, x_data.shape)
#给出噪点（使其变得更像真实数据）变为非连续数据的一组数据
y_data = np.square(x_data) - 0.5 + noise

# define placeholder for inputs to network
with tf.name_scope('inputs'):
    xs = tf.placeholder(tf.float32, [None, 1],name = 'x_input')
    #行不定，列为1，可以解为列向量
    ys = tf.placeholder(tf.float32, [None, 1],name = 'y_input')
# add hidden layer
l1 = add_layer(xs, 1, 10,n_layer = 1, activation_function=tf.nn.relu)
#输入层的神经元为1；隐藏层的神经元为10；输出层的神经元为1
# add output layer
prediction = add_layer(l1, 10, 1,n_layer = 2, activation_function=None)

# the error between prediction and real data
with tf.name_scope('loss'):
    loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction),
                     reduction_indices=[1]))
    tf.summary.scalar('value',loss)
    #tag = 'value' ; value = loss
#每个例子的平方和再求平均值，最后把矩阵向列方向压缩成向量再向行方向压缩成数
with tf.name_scope('train'):
    train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
#学习率为0.1的梯度下降法减小loss的误差
sess = tf.Session()
merged = tf.summary.merge_all()
#存盘所有信息
writer = tf.summary.FileWriter("logs/",sess.graph)
#指定logs文件夹保存图
# important step
# tf.initialize_all_variables() no long valid from
# 2017-03-02 if using tensorflow >= 0.12
if int((tf.__version__).split('.')[1]) < 12:
    init = tf.initialize_all_variables()
else:
    init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)

for i in range(1000):
    #迭代1000次
    # training
    sess.run(train_step, feed_dict={xs: x_data, ys: y_data})
    if i % 50 == 0:
        # to see the step improvement
        result = sess.run(merged,
                          feed_dict={xs: x_data, ys: y_data})
        #重要：run merged
        writer.add_summary(result,i)
        print(sess.run(loss, feed_dict={xs: x_data, ys: y_data}))