深度学习模型训练可视化（TensorBoard可视化）

你是否有这样的疑惑？在我们完成深度学习项目工程中，很多时候我们自己构建模型结构，或者在学习深度学习算法的时候由于算法结构，很难充分理解模型你是否想看到自己的模型训练指标与参数的变化情况很多时候，我们自己构建完成训练后，需要在通过修改参数构建...

姜君泽

10197人浏览 · 2021-08-25 16:10:45

姜君泽 · 2021-08-25 16:10:45 发布

你是否有这样的疑惑？

在我们完成深度学习项目工程中，很多时候我们自己构建模型结构，或者在学习深度学习算法的时候，由于算法结构过于复杂，很难充分理解模型结构层次

你是否想看到自己的模型训练指标与参数的变化情况

很多时候，我们自己构建完成训练后，需要在通过修改参数来评估模型的好坏，同时还需要对相应的指标进行可视化，这样是不是有些麻烦。

它来了！！！

TensorBoard是一款神奇的可视化辅助工具，它能够很好的和pytorch进行配合，本文将从如下几个方面展开：

TensorBoard可视化流程
可视化方法
TensorBoard界面交互

1. TensorBoard可视化流程

首先在pytorch项目中指定一个目录创建一个torch.utils.tensorboard.SummaryWriter日志写入器。
然后根据需要的可视化信息，利用写入器将相关的模型结构、指标，图像写入到指定目录下
关闭写入器
最后通过终端输入命令，在命令中传入目录参数，启动TensorBoard。

1.1 在这里补充一点终端如何启动Tensorboard

这里以我的项目目录为例

\torchproject
  \data
    \tensorboard
      \events.out.tfevents....
  \ten.py

写入器目录为data/tensorboard，项目代码和Tensorboard代码都在ten.py文件中，写入器将相应信息写入到目录中，会生成一个events.out.tfevents…文件。

生成完events文件，就可以打开终端，进入到data目录下运行如下命令

python3 -m  tensorboard.main --logdir=“自己的目录”

在运行的响应的内容中点击http://localhost:6006链接。即可进行TensorBoard界面中。

2. 可视化方法

2.1 可视化模型结构 writer.add_graph

import torch
from torch import nn
from torch.utils.tensorboard import SummaryWriter
class Net(nn.Module):
	def __init__(self):
		super(Net, self).__init__()
		self.conv1 = nn.Conv2d(in_channels=3,out_channels=32,kernel_size = 3)
		self.pool = nn.MaxPool2d(kernel_size = 2,stride = 2)
		self.conv2 = nn.Conv2d(in_channels=32,out_channels=64,kernel_size = 5)
		self.dropout = nn.Dropout2d(p = 0.1)
		self.adaptive_pool = nn.AdaptiveMaxPool2d((1,1))
		self.flatten = nn.Flatten()
		self.linear1 = nn.Linear(64,32)
		self.relu = nn.ReLU()
		self.linear2 = nn.Linear(32,1)
		self.sigmoid = nn.Sigmoid()
	def forward(self,x):
		x = self.conv1(x)
		x = self.pool(x)
		x = self.conv2(x)
		x = self.pool(x)
		x = self.dropout(x)
		x = self.adaptive_pool(x)
		x = self.flatten(x)
		x = self.linear1(x)
		x = self.relu(x)
		x = self.linear2(x)
		y = self.sigmoid(x)
		return y
net = Net()
writer = SummaryWriter('./data/tensorboard')
writer.add_graph(net,input_to_model = torch.rand(1,3,32,32))
writer.close()

在这里插入图片描述

2.2 可视化指标变化 writer.add_scalar

import numpy as np
import torch
from torch.utils.tensorboard import SummaryWriter
# f(x) = a*x**2 + b*x + c的最小值
x = torch.tensor(0.0,requires_grad = True) # x需要被求导
a = torch.tensor(1.0)
b = torch.tensor(-2.0)
c = torch.tensor(1.0)
optimizer = torch.optim.SGD(params=[x],lr = 0.01)
def f(x):
	result = a*torch.pow(x,2) + b*x + c
	return(result)
writer = SummaryWriter('./data/tensorboard')
for i in range(500):
	optimizer.zero_grad()
	y = f(x)
	y.backward()
	optimizer.step()
	writer.add_scalar("x",x.item(),i) #日志中记录x在第step i 的值
	writer.add_scalar("y",y.item(),i) #日志中记录y在第step i 的值
writer.close()

在这里插入图片描述

2.3 可视化参数分布 writer.add_histogram

import numpy as np
import torch
from torch.utils.tensorboard import SummaryWriter
# 创建正态分布的张量模拟参数矩阵
def norm(mean,std):
	t = std*torch.randn((100,20))+mean
	return t
writer = SummaryWriter('./data/tensorboard')
for step,mean in enumerate(range(-10,10,1)):
	w = norm(mean,1)
	writer.add_histogram("w",w, step)
	writer.flush()
writer.close()

在这里插入图片描述

2.4 可视化原始图像 writer.add_image

import torch
import torchvision
from torch import nn
from torch.utils.data import Dataset,DataLoader
from torchvision import transforms,datasets
transform_train = transforms.Compose([transforms.ToTensor()])
transform_valid = transforms.Compose([transforms.ToTensor()])
ds_train = datasets.ImageFolder("./data/cifar2/train/",transform = transform_train,target_transform= lambda t:torch.tensor([t]).float())
ds_valid = datasets.ImageFolder("./data/cifar2/test/",transform = transform_train,target_transform= lambda t:torch.tensor([t]).float())

dl_train = DataLoader(ds_train,batch_size = 50,shuffle = True,num_workers=3)
dl_valid = DataLoader(ds_valid,batch_size = 50,shuffle = True,num_workers=3)
dl_train_iter = iter(dl_train)
images, labels = dl_train_iter.next()
# 仅查看一张图片
writer = SummaryWriter('./data/tensorboard')
writer.add_image('images[0]', images[0])
writer.close()
# 将多张图片拼接成一张图片，中间用黑色网格分割
writer = SummaryWriter('./data/tensorboard')
# create grid of images
img_grid = torchvision.utils.make_grid(images)
writer.add_image('image_grid', img_grid)
writer.close()
# 将多张图片直接写入
writer = SummaryWriter('./data/tensorboard')
writer.add_images("images",images,global_step = 0)
writer.close()

在这里插入图片描述

2.5 可视化人工绘图 writer.add_figure

import torch
import torchvision
from torch import nn
from torch.utils.data import Dataset,DataLoader
from torchvision import transforms,datasets
transform_train = transforms.Compose([transforms.ToTensor()])
transform_valid = transforms.Compose([transforms.ToTensor()])
ds_train = datasets.ImageFolder("./data/cifar2/train/",transform = transform_train,target_transform= lambda t:torch.tensor([t]).float())
ds_valid = datasets.ImageFolder("./data/cifar2/test/",transform = transform_train,target_transform= lambda
t:torch.tensor([t]).float())

from matplotlib import pyplot as plt
figure = plt.figure(figsize=(8,8))
for i in range(9):
	img,label = ds_train[i]
	img = img.permute(1,2,0)
	ax=plt.subplot(3,3,i+1)
	ax.imshow(img.numpy())
	ax.set_title("label = %d"%label.item())
	ax.set_xticks([])
	ax.set_yticks([])
writer = SummaryWriter('./data/tensorboard')
writer.add_figure('figure',figure,global_step=0)
writer.close()

3. TensorBoard界面交互

简单实现一个实例

import numpy as np
# SummaryWriter用来创建一个writer，可以记录我们想要可视化的数据。
from torch.utils.tensorboard import SummaryWriter

writer = SummaryWriter('data/tensorboard')

for x in range(100):
    # 曲线名称，y轴，x轴
    writer.add_scalar('y=2x', x * 2, x)
    writer.add_scalar('y=pow(2, x)', 2 ** x, x)

    writer.add_scalars('data/scalar_group', {"xsinx": x * np.sin(x),
                                             "xcosx": x * np.cos(x),
                                             "arctanx": np.arctan(x)}, x)
writer.close()