【Tensorflow】添加Summary Scalar

import tensorflow as tf
import sys 
import numpy as np
def main():
    #(1)定义网络模型
    sess=tf.Session() 
    x=tf.placeholder(tf.float32,shape=[None],name="input")
    y=tf.placeholder(tf.float32,shape=[None],name="output")
    lr=tf.placeholder(tf.float32,name="learning_rate") 

    w=tf.Variable(0,dtype=tf.float32)
    b=tf.Variable(0,dtype=tf.float32)
    _y=x*w+b
      
    loss=tf.reduce_sum(tf.square(y-_y))
    step=tf.Variable(0,trainable=False)
    #(2) 定义summary，参数一是名字，参数二是Tensor
    loss_summary=tf.summary.scalar("Loss",loss)
    lr_summary=tf.summary.scalar("Learning Rate",lr)
    #(3) 合并所有summary，并定义tf.summmary.FileWriter，写入文件
    all_summary=tf.summary.merge_all() 
    writer=tf.summary.FileWriter("./tmp/",sess.graph)
    
    train_op = tf.train.AdamOptimizer(lr).minimize(loss, global_step=step) #每次执行train_op，step都会自动+1
    sess.run(tf.global_variables_initializer())
    for i in range(0,100):
        x_data=np.random.random([40])
        y_data=x_data*2-10
        feed_dict={ 
            x:x_data,
            y:y_data,
            lr:i/10000
            }
    #(4)执行run时，讲之前合并的all_summary放进去，得到其值summaryValue直接用writer.add_summary记录。第二个参数是横坐标，一般用step表示。 
        _,summaryValue,currentStep,wval,bval,loss_val=sess.run([train_op,all_summary,step,w,b,loss],feed_dict=feed_dict)
        print("loss=%s w=%s  b=%s summary=%s" %(loss,wval,bval,summaryValue))
        writer.add_summary(summaryValue,currentStep)
    return 

if __name__ == "__main__":
    sys.exit(int(main() or 0))

直接传递一个float值而不通过run(summary)的到的实现方法：

import tensorflow as tf
import math

def scalar_logger(tf_writer,name,value,step):
    v=tf.Summary.Value(tag=name,simple_value= value)
    s=tf.Summary(value=[v])
    tf_writer.add_summary(s,step)
    return 
   
if __name__ == "__main__":
   writer=tf.summary.FileWriter("./",tf.Session().graph)
   for x in range(0,1000):
     scalar_logger(writer,"arc",math.sqrt(x),x)