tensorboard可视化

Posted 2020-11-22 lazycat1206

 1 import tensorflow as tf
 2 import numpy as np
 3 import matplotlib.pyplot as plt
 4 def add_layer(inputs, in_size, out_size,n_layer,activation_function=None):
 5     layer_name=‘layer%s‘%n_layer
 6     with tf.name_scope(‘layer‘):
 7         with tf.name_scope(‘weights‘):
 8             Weights = tf.Variable(tf.random_normal([in_size, out_size]),name=‘W‘)
 9             tf.summary.histogram(layer_name+‘/weights‘,Weights)
               #tensorflow中提供了tf.summary.histogram()方法,用来绘制图片, 第一个参数是图表的名称, 第二个参数是图表要记录的变量 

10         with tf.name_scope(‘biases‘): 
11             biases = tf.Variable(tf.zeros([1, out_size]) + 0.1,name=‘b‘) 
12             tf.summary.histogram(layer_name+‘/biases‘,biases) 
13         with tf.name_scope(‘Wx_plus_b‘): 
14             Wx_plus_b = tf.matmul(inputs, Weights) + biases 
15         if activation_function is None: 
16             outputs = Wx_plus_b 
17         else: 
18             outputs = activation_function(Wx_plus_b) 
19             tf.summary.histogram(layer_name+‘/outputs‘,outputs) 
20     return outputs 
22 x_data=np.linspace(-1,1,300,dtype=np.float32)[:,np.newaxis] 
23 noise = np.random.normal(0, 0.05, x_data.shape).astype(np.float32) 
24 y_data=np.square(x_data)-0.5+noise 25 with tf.name_scope(‘inputs‘): 
26 xs=tf.placeholder(tf.float32,[None,1],name=‘x_input‘) 
27 ys=tf.placeholder(tf.float32,[None,1],name=‘y_input‘) 
29 l1=add_layer(xs,1,10,n_layer=1,activation_function=tf.nn.relu) #隐藏层 
30 prediction=add_layer(l1,10,1,n_layer=2,activation_function=None) #输出层 
31 with tf.name_scope(‘loss‘): 
32 loss=tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction), 
33 reduction_indices=[1])) 
34 tf.summary.scalar(‘loss‘,loss)

#Loss的变化图和之前设置的方法略有不同.loss是在tesnorBorad 的scalars下面的, 这是由于我们使用的是tf.summary.scalar()方法.

35 with tf.name_scope(‘train‘): 
36     train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss) 
37 init = tf.global_variables_initializer() 
38 sess = tf.Session() 
39 merged=tf.summary.merge_all() 
   #tf.summary.merge_all()方法会对我们所有的summary合并到一起. 

40 writer=tf.summary.FileWriter(‘logs/‘,sess.graph) 
41 sess.run(init) 
42 for i in range(1000): 
43     sess.run(train_step,feed_dict={xs:x_data,ys:y_data}) 
44     if i%50==0: 
45         result = sess.run(merged,feed_dict={xs:x_data,ys:y_data}) 
46         writer.add_summary(result,i)