TensorFlow 笔记4--使用tensorboard

使用tensorboard

---

1. 保存数据

---

定义图中：

#保存单个数据。例如：损失函数值，准确率
tf.summary.scalar("loss", cross_entropy) 

#保存多个数据。例如：权重
tf.summary.histogram("W1",W)

#把所有的tf.summary.xx融合成一个操作  
merged_summary_op =  tf.summary.merge_all() 

#声明保存的地方和图  
summary_writer = tf.summary.FileWriter('/tmp/mnist_logs', graph_def=sess.graph_def)   

#为变量名划定范围，在范围内的op在可视化中显示成可以展开的一整块。
with tf.name_scope('hidden') as scope:
	W = tf.Variable(tf.random_uniform([1, 2], -1.0, 1.0), name='weights')
	b = tf.Variable(tf.zeros([1]), name='biases')
#结果是得到了操作名: hidden/weights   hidden/biases

运行图中：

for i in range(1000):
    total_step += 1
    batch_xs, batch_ys = mnist.train.next_batch(100)
    sess.run(train_step, feed_dict={x: batch_xs, y_: batch_ys})
    if total_step % 100 == 0:
	    //运行op
        summary_str = sess.run(merged_summary_op,feed_dict={x: batch_xs, y_: batch_ys})
        //保存
        summary_writer.add_summary(summary_str, total_step)

---

2. 打开tensorboard

---

• 在cmd输入：tensorboard --logdir=D://tmp//mnist_logs（保存结果的路径）
• 复制生成的链接，然后在Google浏览器中打开（别的浏览器可能出错）。
  

---

3. 同时显示训练和测试曲线

---

设置两个writer，一个用于写训练的数据，一个用于写测试数据

import tensorflow as tf

......
tf.summary.scalar("loss", loss)
merged_summary_op = tf.summary.merge_all()

with tf.Session() as sess:
    summary_writer1 = tf.summary.FileWriter('../log/view/train', sess.graph)
    summary_writer2 = tf.summary.FileWriter('../log/view/test')

	summary_str = sess.run(merged_summary_op, feed_dict={x: train_data, y_: train_label})
	summary_writer1.add_summary(summary_str, step)
	# summary_writer1.flush()
	summary_str = sess.run(merged_summary_op, feed_dict={x: val_data, y_: val_label})
	summary_writer2.add_summary(summary_str, step)

• 在cmd输入：tensorboard --logdir=…\log\view

---