P21_神经网络——搭建小实战和Sequential的使用

21.1打开pytorch官网

1.神经网络-搭建小实战和Sequential的使用

（1）Sequential的使用：将网络结构放入其中即可，可以简化代码。

（2）一个对CIFAR10进行分类的模型

（3）模型的实现(下文)

21.2打开pycharm实现模型

1.第一步的Convolution

（1）在第一步的Convolution：in_channel是3，out_channel是32，kernel_size是5×5

（2）Conv2d的参数：

点击查看代码

class torch.nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', device=None, dtype=None)

其中in_channel、out_channel和kernel_size已知，至于stride|padding|dilation是否要设置还是默认，要根据计算得知；

（3）根据Conv2d的计算公式和上图的对CIFAR10进行分类的模型来计算：

①NCHW是指一种张量数据的存储格式，具体表示为N（批大小）-C（通道数）-H（高度）-W（宽度）

②在公式中可知：

Hin=32，Hout=32，dilation默认是1，stride默认是1，padding默认是0

③则32=（32+2*padding-dilation×(5-1)-1)/stride +1

设置dilation为默认值1，stride设置为默认值1，则代入得：32=(32+2×padding-1×(5-1)-1)/1+1即padding=2

④即：
self.conv1 = Conv2d(in_channels=3,out_channels=32,kernel_size=5,padding=2)#stride和dilation默认

2.第二步的Maxpooling

self.maxpool1 = MaxPool2d(kernel_size=2)

3.第三步的Convolution

self.conv2 = Conv2d(32,32,5,padding=2)

4.第四步的Maxpooling

self.maxpool2 = MaxPool2d(kernel_size=2)

5.第五步的Convolution

self.conv3 = Conv2d(32,64,5, padding=2)

6.第六步的Maxpooling

self.maxpool3 = MaxPool2d(kernel_size=2)

7.第七步的Flatten

self.flatten = Flatten()

8.第八步的linear

self.linear1 = Linear(1024,64)

9.第九步的linear

self.linear2 = Linear(64,10)

linear示意图：

10.定义forward

点击查看代码

    def forward(self,x):x = self.conv1(x)x = self.maxpool1(x)x = self.conv2(x)x = self.maxpool2(x)x = self.conv3(x)x = self.maxpool3(x)x = self.flatten(x)x = self.linear1(x)x = self.linear2(x)return x

11.搭建网络，实例化并print输出

点击查看代码

dyl_seq = DYL_seq()
print(dyl_seq)

得到：

点击查看代码

x DYL_seq(  (conv1): Conv2d(3, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))  (maxpool1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)  (conv2): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))  (maxpool2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)  (conv3): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))  (maxpool3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)  (flatten): Flatten(start_dim=1, end_dim=-1)  (linear1): Linear(in_features=1024, out_features=64, bias=True)  (linear2): Linear(in_features=64, out_features=10, bias=True))

12.创建输入，使用ones指定我们想要创建的形状大小

创建一个假想的输入：

torch.ones(64,3,32,32) #64个batch_size，3个通道，32×32的尺寸

点击查看代码

#torch.ones可以指定我们想要创建的数的形状大小
input = torch.ones(64,3,32,32) #64个batch_size，3个通道，32×32的尺寸
output = dyl_seq(input)
print(output.shape)

输出为： `torch.Size([64, 10])`

其中64表示64个batch_size，可以想象成64张图片

21.3使用Sequential实现模型

1.原始繁琐的代码：

点击查看代码

import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linearclass DYL_seq(nn.Module):def __init__(self):super(DYL_seq, self).__init__()self.conv1 = Conv2d(in_channels=3,out_channels=32,kernel_size=5,padding=2)#stride和dilation默认self.maxpool1 = MaxPool2d(kernel_size=2)self.conv2 = Conv2d(32,32,5,padding=2)self.maxpool2 = MaxPool2d(kernel_size=2)self.conv3 = Conv2d(32,64,5, padding=2)self.maxpool3 = MaxPool2d(kernel_size=2)self.flatten = Flatten()self.linear1 = Linear(1024,64)self.linear2 = Linear(64, 10)def forward(self,x):x = self.conv1(x)x = self.maxpool1(x)x = self.conv2(x)x = self.maxpool2(x)x = self.conv3(x)x = self.maxpool3(x)x = self.flatten(x)x = self.linear1(x)x = self.linear2(x)return xdyl_seq = DYL_seq()
print(dyl_seq)#创建一个假想的输入
#torch.ones可以指定我们想要创建的数的形状大小
input = torch.ones(64,3,32,32) #64个batch_size，3个通道，32×32的尺寸
output = dyl_seq(input)
print(output.shape)

输出结果如下：

点击查看代码

DYL_seq((conv1): Conv2d(3, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))(maxpool1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)(conv2): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))(maxpool2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)(conv3): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))(maxpool3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)(flatten): Flatten(start_dim=1, end_dim=-1)(linear1): Linear(in_features=1024, out_features=64, bias=True)(linear2): Linear(in_features=64, out_features=10, bias=True)
)
torch.Size([64, 10])

2.使用Sequential对模型重新进行实现

点击查看代码

import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequentialclass DYL_seq(nn.Module):def __init__(self):super(DYL_seq, self).__init__()self.model1 = Sequential(Conv2d(in_channels=3, out_channels=32, kernel_size=5, padding=2),MaxPool2d(kernel_size=2),Conv2d(32, 32, 5, padding=2),MaxPool2d(kernel_size=2),Conv2d(32, 64, 5, padding=2),MaxPool2d(kernel_size=2),Flatten(),Linear(1024, 64),Linear(64, 10))def forward(self,x):x = self.model1(x)return x
dyl_seq = DYL_seq()
print(dyl_seq)#创建一个假想的输入
#torch.ones可以指定我们想要创建的数的形状大小
input = torch.ones(64,3,32,32) #64个batch_size，3个通道，32×32的尺寸
output = dyl_seq(input)
print(output.shape)

实现了如上同样的效果：

点击查看代码

xxxxxxxxxx DYL_seq(  (model1): Sequential(    (0): Conv2d(3, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))    (1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)    (2): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)    (4): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))    (5): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)    (6): Flatten(start_dim=1, end_dim=-1)    (7): Linear(in_features=1024, out_features=64, bias=True)    (8): Linear(in_features=64, out_features=10, bias=True)  ))torch.Size([64, 10])

可见，代码变得简洁很多

21.4使用tensorboard进行可视化

tensorboard --logdir=dyl_seq
双击dyl_seq就会得到中间有一个Sequential，再双击Sequential就可以看到各网络层：

然后在网络层继续双击，就可以看到：
每个箭头上显示的我们送到该网络层的尺寸的大小，设置的各参数：