Tensorboard可视化操作

在网络训练过程中，为了更直观的看到训练过程参数，可以利用tensorboard进行可视化。

在使用中主要有两个添加数据函数：

添加标量参数：tf.summary.scalar(tags, values, collections=None, name=None)
例如：tf.summary.scalar('loss', loss)

添加直方图参数：tf.summary.histogram(tags, values, collections=None, name=None)
例如：tf.summary.histogram( 'weight', w)

添加图像：tf.summary.image(tags, image)
例如：tf.summary.image('image', img)

其他函数：

自动管理函数：tf.summary.merge_all()
例如：merged = tf.summary.merge_all()

创建文件函数：tf.summary.FileWriter(path, session.graph)
例如：writer = tf.summary.FileWriter("./log", sess.graph)

数据写入函数：writer.add_summary(result, step)
例如：writer.add_summary(result, i)

关闭文件函数：writer.close()

打开可视化界面：
首先找到
1、 tensorboard.exe文件（一般在你安装对应环境下）
2、 在tensorboard.exe文件夹下打开终端窗口
3、 在终端窗口输入tensorboard --logdir=输入你生成的tensorboard数据文件夹路径（例如
tensorboard --logdir=D:\my_graph\log）

4、 在浏览器中输入生成的http://localhost:6006/便可以打开可视化窗口。


测试代码：

import tensorflow as tf
import numpy as np

# Make up some real data
x_data = np.linspace(-1, 1, 300)[:, np.newaxis]
noise = np.random.normal(0, 0.05, x_data.shape)
y_data = np.square(x_data) - 0.5 + noise

def fc_layer(x, in_kernels, out_kernels, layer_name):
    '''全连接层'''
    with tf.variable_scope(layer_name):
        w = tf.get_variable('weight', shape=[in_kernels, out_kernels],
                            initializer=tf.truncated_normal_initializer(stddev=0.1))
        tf.summary.histogram(layer_name + '/weight', w)  # 权重值加入到tensorboard
        b = tf.get_variable('bias', shape=[out_kernels],
                            initializer=tf.truncated_normal_initializer(stddev=0.1))
        tf.summary.histogram(layer_name + '/bias', b)  # 偏执值加入到tensorboard
        y = tf.nn.relu(tf.nn.bias_add(tf.matmul(x, w), b))
    return y


def out_layer(x, in_kernels, out_kernels, layer_name):
    '''输出层'''
    with tf.variable_scope(layer_name):
        w = tf.get_variable('weight', shape=[in_kernels, out_kernels],
                            initializer=tf.truncated_normal_initializer(stddev=0.1))
        tf.summary.histogram(layer_name + '/weight', w)  # 权重值加入到tensorboard
        b = tf.get_variable('bias', shape=[out_kernels],
                            initializer=tf.truncated_normal_initializer(stddev=0.1))
        tf.summary.histogram(layer_name + '/bias', b)  # 偏执值加入到tensorboard
        y = tf.nn.bias_add(tf.matmul(x, w), b)
    return y


# define placeholder for inputs to network
with tf.name_scope('inputs'):
    xs = tf.placeholder(tf.float32, [None, 1], name='x_input')
    ys = tf.placeholder(tf.float32, [None, 1], name='y_input')

# add hidden layer
fc_1 = fc_layer(xs, 1, 10, 'fc_1')
# add output layer
prediction = out_layer(fc_1, 10, 1, 'out')

# the error between prediciton 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('loss', loss)

with tf.name_scope('train'):
    train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)

sess = tf.Session()
#合并
merged = tf.summary.merge_all()

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

init = tf.global_variables_initializer()
sess.run(init)
# saver = tf.train.Saver()

for i in range(1000):
    sess.run(train_step, feed_dict={xs: x_data, ys: y_data})
    if i % 50 == 0:
        # saver.save(sess, './model/model.ckpt', global_step=i)
        result = sess.run(merged,feed_dict={xs: x_data, ys: y_data})
        print(sess.run(loss,feed_dict={xs: x_data, ys: y_data}))
        #i 就是记录的步数
        writer.add_summary(result, i)

writer.close()
sess.close()