tensorboard可视化

代码：

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
from tensorflow.contrib.tensorboard.plugins import projector

 

#载入数据集
mnist=input_data.read_data_sets("G:\PythonN\Graduation\MNIST_data",one_hot=True)

#运行次数
max_steps=1001
#图片数量
image_num=3000
#文件路径
DIR="G:/PythonN/tensorflow"

#定义会话
sess=tf.Session()

#载入图片
embedding=tf.Variable(tf.stack(mnist.test.images[:image_num]),trainable=False,name='embedding')

#参数概要
def variable_summaries(var):
    with tf.name_scope('summaries'):
        mean=tf.reduce_mean(var)
        tf.summary.scalar('mean',mean)#平均值
        with tf.name_scope('stddev'):
            stddev=tf.sqrt(tf.reduce_mean(tf.square(var-mean)))
        tf.summary.scalar('stddev',stddev)#标准差
        tf.summary.scalar('max',tf.reduce_max(var))#最大值
        tf.summary.scalar('min',tf.reduce_min(var))#最小值
        tf.summary.histogram('histogram',var)#直方图
            
#命名空间
with tf.name_scope('input'):
    #定义placeholder
    x=tf.placeholder(tf.float32,[None,784],name='x-input')
    y=tf.placeholder(tf.float32,[None,10],name='y-input')
    
#显示图片
with tf.name_scope('input_reshape'):
    image_shaped_input=tf.reshape(x,[-1,28,28,1])#-1代表任意值；黑白图片所以=1，若是彩色图片=3
    tf.summary.image('input',image_shaped_input,10)
with tf.name_scope('layer'):
    #创建简单的神经网络
    with tf.name_scope('weihts'):
        W=tf.Variable(tf.zeros([784,10]),name='W')
        variable_summaries(W)
    with tf.name_scope('bias'):
        b=tf.Variable(tf.zeros([10]),name='b')
        variable_summaries(b)
    with tf.name_scope('wx_plus_b'):
        wx_plus_b=tf.matmul(x,W)+b
    with tf.name_scope('softmax'):
        prediction=tf.nn.softmax(wx_plus_b)
with tf.name_scope('loss'):
    #定义损失函数
    #二次代价函数
    #loss=tf.reduce_mean(tf.square(y-prediction))
    #交叉熵代价函数
    loss=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=prediction))
    tf.summary.scalar('loss',loss)
with tf.name_scope('train'):
    #梯度下降
    train=tf.train.GradientDescentOptimizer(0.2).minimize(loss)
#初始化变量
init=tf.global_variables_initializer()
sess.run(init)
with tf.name_scope('accuracy'):
    with tf.name_scope('cor_pre'):
        #存放结果
        cor_pre=tf.equal(tf.argmax(y,1),tf.arg_max(prediction,1))#1表示按行
    with tf.name_scope('arc'):
        #准确率
        arc=tf.reduce_mean(tf.cast(cor_pre,tf.float32))
        tf.summary.scalar('arc',arc)
#产生matedata文件
if(tf.gfile.Exists(DIR+"/projector/projector/metadata.tsv")):
    tf.gfile.DeleteRecursively(DIR+"/projector/projector/metadata.tsv")
with open(DIR + '/projector/projector/metadata.tsv','w') as f:
    labels=sess.run(tf.argmax(mnist.test.labels[:],1))
    for i in range(image_num):
        f.write(str(labels[i]) + '\n')
#合并所有的summary
merged=tf.summary.merge_all()
# 定义一个writer，写入图
projector_writer = tf.summary.FileWriter(DIR + '/projector/projector',sess.graph)
#用来保存网络模型
saver = tf.train.Saver()
# 定义配置文件
config = projector.ProjectorConfig()
embed = config.embeddings.add()
embed.tensor_name = embedding.name
embed.metadata_path = DIR + '/projector/projector/metadata.tsv'
embed.sprite.image_path = DIR + '/projector/data/mnist_10k_sprite.png'
# 切分图片
embed.sprite.single_image_dim.extend([28,28])
projector.visualize_embeddings(projector_writer, config)
#运行
for i in range(max_steps):
    #每个批次100个样本
    batch_xs, batch_ys = mnist.train.next_batch(100)
    run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
    run_metadata = tf.RunMetadata()
    summary, _ = sess.run([merged, train], feed_dict={x: batch_xs, y: batch_ys}, options=run_options,
                          run_metadata=run_metadata)
    projector_writer.add_run_metadata(run_metadata, 'step%03d' % i)
    projector_writer.add_summary(summary, i)

    if i % 100 == 0:
        acc = sess.run(arc, feed_dict={x: mnist.test.images, y: mnist.test.labels})
        print("Iter " + str(i) + ", Testing Accuracy= " + str(acc))

saver.save(sess, DIR + '/projector/projector/a_model.ckpt', global_step=max_steps)
projector_writer.close()
sess.close()

tensorboard的显示与原来一样

PS:视频课程所得