Tensorboard使用

    with tf.name_scope('loss_value'):        
        loss=tf.reduce_mean(tf.losses.softmax_cross_entropy(onehot_labels=y_, logits=y))+tf.add_n(tf.get_collection('losses'))
#        loss=np.square(y-y_))
#   监视loss值     
    tf.summary.scalar('loss_value',loss)
        
#    指数衰减学习率
    learning_rate=tf.train.exponential_decay(Cfg.LEARNING_RATE_BASE,global_step,1000,
                                             Cfg.LEARNING_RATE_DECAY,staircase=True)
    tf.summary.scalar("learningRate",learning_rate)
#    优化损失函数
    train_step=tf.train.GradientDescentOptimizer(learning_rate).minimize(loss,global_step=global_step)
            
#    反向传播，更新神经网络中参数和每一个参数的滑动平均值             
    with tf.control_dependencies([train_step,variables_averages_op]):
        train_op=tf.no_op(name='train')  
    merged=tf.summary.merge_all()
    with tf.Session() as sess:
        tf.global_variables_initializer().run()
#        写入日志
        summary_writer=tf.summary.FileWriter('/path/to/log',sess.graph)

#   迭代地训练神经网络        
        for i in range(Cfg.TRAINING_STEPS):
            
#            产生这一轮使用的一个batch的训练数据，并运行训练    
            train_data , train_label = DATA.train_batch()
            xs=np.reshape(train_data,(Cfg.BATCH_SIZE,Cfg.IMAGE_SIZE,Cfg.IMAGE_SIZE,Cfg.NUM_CHANNELS))
            ys=np.reshape(train_label,[Cfg.BATCH_SIZE,Cfg.OUTPUT_NODE])
            
            run_options=tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
            run_metadata=tf.RunMetadata()
            
            _,loss_value,step,summary=sess.run([train_op,loss,global_step,merged],feed_dict={x:xs,y_:ys}
                                                        ,options=run_options,run_metadata=run_metadata)
            if i % 50 == 0:
                
#           向日志中写入信息 
                summary_writer.add_run_metadata(run_metadata,'step%03d'%i)
                summary_writer.add_summary(summary,i)

代码中配置好后，进入cmd，输入

tensorboard --logdir=/path/to/log ###注意该路径与代码中定义的保存日志的路径一致

配置好后打开浏览器，进入

http://desktop-g63nqii:6006  ###前面为自己计算机的名称