Tensorboard可以记录与展示以下数据形式:
(1)标量Scalars
(2)图片Images
(3)音频Audio
(4)计算图Graph
(5)数据分布Distribution
(6)直方图Histograms
(7)嵌入向量Embeddings
使用tensorboard最好这是好name_scope或者Variable_scope
添加数据到直方图,events, image
tf.summary.histogram('记录的名字',记录的变量)
tf.summary.scalar('记录的名字',记录的变量)
with tf.name_scope('input_reshape'):
image_shaped_input = tf.reshape(x, [-1, 28, 28, 1])
tf.summary.image('input', image_shaped_input, 10)
合并所有的summary
merged = tf.summary.merge_all()
将summary写入指定目录;
train_summary = tf.summary.FileWriter(log_dir + 'logs', sess.graph)
建议:
- merged,在之后的每隔一定的训练次数就run(merged,feed_dict={})
- tensorboard --logdir=‘summary路径’
如果出现问题,最好参考下面的顺序,同时添加变量的时候,变量必须带有
tf.name_scope(’****’)
import tensorflow as tf
#tf.Summary示例
with tf.name_scope('layer1'):
v1 = tf.Variable(initial_value=tf.constant(10.0),name='v1')
tf.summary.scalar('v1', v1)
with tf.name_scope('layer2'):
v2 = tf.Variable(initial_value=tf.constant(10.0),name='v1')
tf.summary.scalar('v1', v2)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
merged = tf.summary.merge_all()
writer = tf.summary.FileWriter('log', sess.graph)
for i in range(1000):
merge_opt = sess.run(merged)#如果需要,就要加feed_dict
writer.add_summary(merge_opt, global_step=i)