tensorboard用一张图显示多条线

先上代码

from torch.utils import tensorboard
import numpy as np

writer = {
    'loss': tensorboard.SummaryWriter("./drive/MyDrive/logs/loss"), #必须要不同的writer
    'acc': tensorboard.SummaryWriter("./drive/MyDrive/logs/acc"),
    'lr': tensorboard.SummaryWriter("./drive/MyDrive/logs/lr")
}

data = np.random.random((3, 10)) #生成模拟数据

loss_data = data[0]
acc_data = data[1]
lr_data = data[2]

for i in range(10):
    writer['loss'].add_scalar("data", loss_data[i], i) #要想显示在一张图 表格名字要一样！！
    writer['acc'].add_scalar("data", acc_data[i], i)
    writer['lr'].add_scalar("data", lr_data[i], i)

writer['loss'].close()
writer['acc'].close()
writer['lr'].close()
exit()

终端命令：

tensorboard --logdir="drive/mydrive/logs"

运行结果：

 

 

要点：

1、每条线一个单独的文件夹

2、每条线一个单独的writer

3、表格名必须相同

屁大点事搞了好久，到处查，都不对！！三句话就能说清楚的，要么搞一大篇，要么代码乱七八糟