鸡翅|时会_pytorch（网络模型）

篇首语：本文由编程笔记#小编为大家整理，主要介绍了pytorch（网络模型）相关的知识，希望对你有一定的参考价值。

上一篇

官网

import torch.nn as nn
import torch.nn.functional as F
class Model(nn.Module):
def __init__(self):
super().__init__()
self.conv1 &＃61; nn.Conv2d(1, 20, 5)
self.conv2 &＃61; nn.Conv2d(20, 20, 5)
def forward(self, x):
x &＃61; F.relu(self.conv1(x))# 卷积、非线性处理
return F.relu(self.conv2(x))


练习

import torch
import torch.nn as nn
import torch.nn.functional as F
class Dun(nn.Module):
def __init__(self):
super().__init__()
self.conv1 &＃61; nn.Conv2d(1, 20, 5)
self.conv2 &＃61; nn.Conv2d(20, 20, 5)
def forward(self, x):
return x&＃43;1
dun&＃61;Dun()
x&＃61;torch.tensor(1.0)# 转化类型
output&＃61;dun(x);# 调用forward
print(output)# 输出


卷积层

import torch
input&＃61;torch.tensor([[1,2,0,3,1],
[0,1,2,3,1],
[1,2,1,0,0],
[5,2,3,1,1],
[2,1,0,1,1]])
kernel&＃61;torch.tensor([[1,2,1],
[0,1,0],
[2,1,0]])
print(input.shape)# 输出尺寸
print(kernel.shape)
input&＃61;torch.reshape(input,(1,1,5,5))# 类型转换
kernel&＃61;torch.reshape(kernel,(1,1,3,3))# 类型转换
print(input)
print(kernel)
print(input.shape)
print(kernel.shape)
# 卷积操作
out&＃61; F.conv2d(input,kernel,stride&＃61;1)
print(out)
out&＃61; F.conv2d(input,kernel,stride&＃61;2)
print(out)
# 填充
out&＃61; F.conv2d(input,kernel,stride&＃61;1,padding&＃61;1)
print(out)


输出chanel是2时

import torch
import torchvision
from torch import nn
from torch.nn import Conv2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
dataset&＃61;torchvision.datasets.CIFAR10("./data_set_test",train&＃61;False,transform&＃61;torchvision.transforms.ToTensor(),download&＃61;True)
dataloader&＃61; DataLoader(dataset,batch_size&＃61;64)
# 卷积类
class Dun(nn.Module):
def __init__(self):
super().__init__()
self.conv1&＃61;Conv2d(in_channels&＃61;3,out_channels&＃61;6,kernel_size&＃61;3,stride&＃61;1,padding&＃61;0)
def forward(self,x):
return self.conv1(x)
dun&＃61;Dun()
print(dun)
writer&＃61; SummaryWriter("./logs")
step&＃61;0
# 卷积操作
for data in dataloader:
img,target&＃61;data
output&＃61;dun(img)
print(img.shape)
print(output.shape)
writer.add_images("input",img,step)
output&＃61;torch.reshape(output,(-1,3,30,30))# -1时会根据后面的值自动计算
writer.add_images("output",output,step)
step&＃43;&＃61;1
writer.close()


池化层

作用&＃xff1a;就像高清视频换成低清视频

import torch
from torch import nn
from torch.nn import MaxPool2d
input &＃61;torch.tensor([[1,2,0,3,1],
[0,1,2,3,1],
[1,2,1,0,0],
[5,2,3,1,1],
[2,1,0,1,1]],dtype&＃61;torch.float32)
input&＃61;torch.reshape(input,(-1,1,5,5))
class Dun(nn.Module):
def __init__(self):
super().__init__()
self.maxpool&＃61;MaxPool2d(kernel_size&＃61;3,ceil_mode&＃61;True)# ceil_model false和True的结果和预期的一致
def forward(self,inut):
return self.maxpool(input)
dun&＃61;Dun()
out&＃61;dun(input)
print(out)


图片处理

import torch
import torchvision
from torch import nn
from torch.nn import MaxPool2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
dataset&＃61;torchvision.datasets.CIFAR10("./data_set_test",train&＃61;False,transform&＃61;torchvision.transforms.ToTensor(),download&＃61;True)
dataloader&＃61; DataLoader(dataset,batch_size&＃61;64)
class Dun(nn.Module):
def __init__(self):
super().__init__()
self.maxpool&＃61;MaxPool2d(kernel_size&＃61;3,ceil_mode&＃61;True)# ceil_model false和True的结果和预期的一致
def forward(self,input):
return self.maxpool(input)
dun&＃61;Dun()
step&＃61;0
writer&＃61;SummaryWriter("logs")
for data in dataloader:
img,target&＃61;data
writer.add_images("input",img,step)
output&＃61;dun(img)
writer.add_images("output",output,step)
step&＃43;&＃61;1
writer.close()


非线性激活

非线性变换目的是引入非线性特征&＃xff0c;可以更好地处理信息

ReLU

import torch
from torch import nn
from torch.nn import ReLU
input&＃61; torch.tensor([[1,-0.5],[-1,3]])
input&＃61;torch.reshape(input,(-1,1,2,2))
print(input.shape)
class Dun(nn.Module):
def __init__(self):
super().__init__()
self.relu1&＃61;ReLU()
def forward(self,input):
return self.relu1(input)
dun&＃61;Dun()
output&＃61;dun(input)
print(output)


sigmoid

import torch
import torchvision
from torch import nn
from torch.nn import ReLU, Sigmoid
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
dataset&＃61;torchvision.datasets.CIFAR10("./data_set_test",train&＃61;False,download&＃61;True,transform&＃61;torchvision.transforms.ToTensor())
dataloader&＃61;DataLoader(dataset,batch_size&＃61;64)
class Dun(nn.Module):
def __init__(self):
super().__init__()
self.relu1&＃61;ReLU()
self.sigmoid&＃61;Sigmoid()
def forward(self,input):
return self.sigmoid(input)
dun&＃61;Dun()
writer&＃61;SummaryWriter("./logs")
step&＃61;0
for data in dataloader:
img,target&＃61;data
writer.add_images("input",img,global_step&＃61;step)
output&＃61;dun(img)
writer.add_images("output",output,global_step&＃61;step)
step&＃43;&＃61;1
writer.close()


线性层

import torch
import torchvision
from torch import nn
from torch.nn import Linear
from torch.utils.data import DataLoader
dataset&＃61;torchvision.datasets.CIFAR10("./data_set_test",train&＃61;False,transform&＃61;torchvision.transforms.ToTensor(),download&＃61;True)
dataloader&＃61;DataLoader(dataset,batch_size&＃61;64)
class Dun(nn.Module):
def __init__(self):
super().__init__()
self.linear&＃61;Linear(196608,10)
def forward(self,input):
return self.linear(input)
dun&＃61;Dun()
for data in dataloader:
img,target&＃61;data
print(img.shape)
# input&＃61;torch.reshape(img,(1,1,1,-1))
input&＃61; torch.flatten(img)# 将数据展平一行&＃xff0c;可以代替上面的一行
print(input.shape)
output&＃61;dun(input)
print(output.shape)


正则化层

加快神经网络地训练速度

# With Learnable Parameters
m &＃61; nn.BatchNorm2d(100)
# Without Learnable Parameters
m &＃61; nn.BatchNorm2d(100, affine&＃61;False)
input &＃61; torch.randn(20, 100, 35, 45)
output &＃61; m(input)


其他层有Recurrent Layers、Transformer Layers、Linear Layers等

简单的网络模型

import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.tensorboard import SummaryWriter
class Dun(nn.Module):
def __init__(self):
super().__init__()
# 2.
self.model1 &＃61; Sequential(Conv2d(3, 32, 5, padding&＃61;2),
MaxPool2d(2),
Conv2d(32, 32, 5, padding&＃61;2),
MaxPool2d(2),
Conv2d(32, 64, 5, padding&＃61;2),
MaxPool2d(2),
Flatten(),
Linear(1024, 64),
Linear(64, 10))
# 1.
# self.conv1&＃61;Conv2d(3,32,5,padding&＃61;2)
# self.maxpool1&＃61;MaxPool2d(2)
# self.conv2&＃61;Conv2d(32,32,5,padding&＃61;2)
# self.maxpool2&＃61;MaxPool2d(2)
# self.conv3&＃61;Conv2d(32,64,5,padding&＃61;2)
# self.maxpool3&＃61;MaxPool2d(2)
# self.flatten&＃61;Flatten()
# self.linear1&＃61;Linear(1024,64)
# self.linear2&＃61;Linear(64,10)
def forward(self,x):
x&＃61;self.model1(x)
return x
dun&＃61;Dun()
# 测试
input&＃61;torch.ones((64,3,32,32))
print(dun(input).shape)
writer&＃61;SummaryWriter("./logs")
writer.add_graph(dun,input)
writer.close()


loss function

L1Loss、MSELoss

import torch
from torch.nn import L1Loss
from torch import nn
input&＃61;torch.tensor([1,2,3],dtype&＃61;torch.float32)
targrt&＃61;torch.tensor([1,2,5],dtype&＃61;torch.float32)
loss&＃61;L1Loss(reduction&＃61;"sum")# 该参数有sum和mean两种&＃xff0c;默认是mean
print(loss(input,targrt))
loss_mse&＃61;nn.MSELoss()
print(loss_mse(input,targrt))


CROSSENTROPYLOSS

import torch
from torch.nn import L1Loss
from torch import nn
x&＃61;torch.tensor([0.1,0.2,0.3])
y&＃61;torch.tensor([1])
x&＃61;torch.reshape(x,(1,3))
loss_cross&＃61;nn.CrossEntropyLoss()
print(loss_cross(x,y))


使用

import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.data import DataLoader
dataset&＃61;torchvision.datasets.CIFAR10("./data_set_test",train&＃61;False,transform&＃61;torchvision.transforms.ToTensor(),download&＃61;True)
dataloader&＃61;DataLoader(dataset&＃61;dataset,batch_size&＃61;1)
# 分类神经网络
class Dun(nn.Module):
def __init__(self):
super().__init__()
# 2.