tensorboard同时记录训练集和测试集的loss

1 为train和test分别写好loss

cost = tf.losses.mean_squared_error(y, predict)
loss = tf.sqrt(tf.reduce_mean(cost))#rmse
tf.summary.scalar('train_loss',loss)#画loss

#画测试集的loss 但是需要用测试集的数据传入feed
cost1 = tf.losses.mean_squared_error(y, predict)
loss1 = tf.sqrt(tf.reduce_mean(cost1))#rmse
tf.summary.scalar('test_loss', loss1)

2 画图

merged=tf.summary.merge_all()
train_writer=tf.summary.FileWriter('./logs/train',sess.graph)
test_writer=tf.summary.FileWriter('./logs/test')

3 run

#传入train的feed 获得train_loss 记录
_,train_graph=sess.run([loss,merged],feed)
train_writer.add_summary(train_graph,step)#画train_rmse

#重新传入test的feed 获得test_loss 记录
test_loss,test_graph=sess.run([loss1,merged],feed)
test_writer.add_summary(test_graph,step)

使用tensorboard --logdir=logs --host=127.0.0.1运用 得到这样的结果 不够完善 但勉强能用

参考文章：
https://blog.csdn.net/z2539329562/article/details/80554432
https://www.jianshu.com/p/976eb0f62458