TensorboardX 学习

最新推荐文章于 2024-05-21 19:21:06 发布

原创最新推荐文章于 2024-05-21 19:21:06 发布 · 253 阅读

0 ·

CC 4.0 BY-SA版权

Pytorch练手专栏收录该内容

4 篇文章

订阅专栏

作者因论文学习pytorch，发现其中TensorboardX可视化很有趣，按教程写了几个Demo，介绍了在cmd或Terminal当前python文件目录下可显示的内容，如ResNet网络结构展示等，还给出了代码来源及可玩的Demo链接。

因为论文需要学习了pytorch，其中的TensorboardX可视化非常有意思，按照他人的教程写了几个Demo

以下代码在cmd或者Terminal的当前python文件目录下输入都可以显示

tensorboard --logdir=./logs

1.graph ，ResNet网络结构展示

from tensorboardX import SummaryWriter
import torch
from torch import  nn as nn


class LeNet(nn.Module):
    def __init__(self):
        super(LeNet, self).__init__()
        self.conv1 = nn.Sequential(  # input_size=(1*28*28)
            nn.Conv2d(1, 6, 5, 1, 2),
            nn.ReLU(),  # (6*28*28)
            nn.MaxPool2d(kernel_size=2, stride=2),  # output_size=(6*14*14)
        )
        self.conv2 = nn.Sequential(
            nn.Conv2d(6, 16, 5),
            nn.ReLU(),  # (16*10*10)
            nn.MaxPool2d(2, 2)  # output_size=(16*5*5)
        )
        self.fc1 = nn.Sequential(
            nn.Linear(16 * 5 * 5, 120),
            nn.ReLU()
        )
        self.fc2 = nn.Sequential(
            nn.Linear(120, 84),
            nn.ReLU()
        )
        self.fc3 = nn.Linear(84, 10)

    # 定义前向传播过程，输入为x
    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        # nn.Linear()的输入输出都是维度为一的值，所以要把多维度的tensor展平成一维
        x = x.view(x.size()[0], -1)
        x = self.fc1(x)
        x = self.fc2(x)
        x = self.fc3(x)
        return x

dummy_input  = torch.rand(13,1,28,28)
model = LeNet()
with SummaryWriter(comment='LeNet',log_dir='logs') as w:
    w.add_graph(model, (dummy_input,))

2.scalar、histogram、text

import torch as t
from tensorboardX import  SummaryWriter

with SummaryWriter(log_dir='logs') as w:
    x = t.FloatTensor([100])
    y = t.FloatTensor([500])
    for epoch in range(100):
        x = x /1.5
        y = y/1.5
        loss = y -x
        print(loss)
        w.add_histogram('zz/x', x, epoch)
        w.add_histogram('zz/y', y, epoch)
        w.add_scalar('data/x', x, epoch)
        w.add_scalar('data/y', y, epoch)
        w.add_scalar('data/loss', loss, epoch)
        w.add_scalars('data/scalar_group',
                      {'x': x, 'y':y,'loss':loss},
                      epoch)
        w.add_text('zz/text','zz:this is epoch' + str(epoch), epoch)
    w.export_scalars_to_json("./test.json")

代码来源以及重要参考：

https://blog.youkuaiyun.com/JNingWei/article/details/79740825

https://blog.youkuaiyun.com/xiaoxifei/article/details/82735355

可以闲着没事玩一玩的各种Demo：

https://www.cnblogs.com/kk17/p/10077335.html