PyTorch-nn.Module

news2025/1/18 16:56:06

1. 基本操作

2. 卷积操作

2.1 torch.nn.functional — conv2d

2.2 torch.nn.Conv2d

3. 池化层

4. 非线性激活

4.1 使用ReLU非线性激活

4.2 使用Sigmoid非线性激活

5. 线性激活

6. PyTorch的一些图像模型

1. 基本操作

import torch
from torch import nn

class MyModel(nn.Module):
    def __init__(self):
        super(MyModel, self).__init__()

    def forward(self, input):
        output = input + 1
        return output

my_module1 = MyModel()
x = torch.tensor(1.0)
output = my_module1(x)
print(output)  # tensor(2.)

2. 卷积操作

2.1 torch.nn.functional -- conv2d

图片来源于：b站up主我是土堆

注意conv2d为二维卷积，conv1d为一维卷积。

nn_conv.py

import torch
import torch.nn.functional as F

# 输入图像
input = 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]
])
# 卷积核
kernel = torch.tensor([
    [1, 2, 1],
    [0, 1, 0],
    [2, 1, 0]
])
input = torch.reshape(input, (1, 1, 5, 5))
kernel = torch.reshape(kernel, (1, 1, 3, 3))
print(input.shape, kernel.shape)
output = F.conv2d(input, kernel, stride=1)
print(output)

torch.Size([1, 1, 5, 5]) torch.Size([1, 1, 3, 3])
tensor([[[[10, 12, 12],
[18, 16, 16],
[13, 9, 3]]]])

stride（步长）若为2：

output2 = F.conv2d(input, kernel, stride=2)
print(output2)

tensor([[[[10, 12],
[13, 3]]]])

padding为1时，相当于在外侧填充了值为0的两行和两列：

output3 = F.conv2d(input, kernel, stride=1, padding=1)
print(output3)

tensor([[[[ 1, 3, 4, 10, 8],
[ 5, 10, 12, 12, 6],
[ 7, 18, 16, 16, 8],
[11, 13, 9, 3, 4],
[14, 13, 9, 7, 4]]]])

2.2 torch.nn.Conv2d

介绍：

in_channels -- 输入的通道数

out_channels -- 根据卷积核数，如in_channels=1，卷积核有两个，则out_channels=2

nn_conv2d.py

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 = torchvision.datasets.CIFAR10('./dataset', train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=64)

class MyModule(nn.Module):
    def __init__(self):
        super(MyModule, self).__init__()
        # 我们的图像是三通道的(RGB)
        # 3×3大小的卷积核一共有6个,因为输出为6个，每个卷积核有3通道
        self.conv1 = Conv2d(in_channels=3, out_channels=6, kernel_size=3, stride=1, padding=0)

    def forward(self, x):
        x = self.conv1(x)
        return x

myModule1 = MyModule()

writer = SummaryWriter('./logs')
step = 0
for data in dataloader:
    imgs, targets = data
    output = myModule1(imgs)
    print(imgs.shape)  # torch.Size([64, 3, 32, 32])
    print(output.shape)  # torch.Size([64, 6, 30, 30])
    output = torch.reshape(output, (-1, 3, 30, 30))  # 不知道多少时填-1,会帮你自动计算,相当于batch_size多了
    writer.add_images('input', imgs, step)
    writer.add_images('output', output, step)
    step = step + 1
writer.close()

注意：output要进行reshape是因为tensorboard在放图片时，没有六通道的处理。于是我们相当于把六通道分成2个三通道，匀出来的分给了batch_size，这样batch_size的数量就扩大了两倍。在不知道具体数为多少时，我们可以用-1代替。

3. 池化层

图片来源于：b站up主我是土堆

说明：若没有设置步长，则默认为kernel_size大小。若Ceil_model设置为True，则不够3×3的部分也会保留进行计算；如果Ceil_model设置为False，则不够3×3的部分直接舍去，不进行计算。

N—batch_size C—Channels

nn_maxpool.py

import torch
from torch import nn
from torch.nn import MaxPool2d

input = 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=torch.float32)
# -1——batch_size数,1——通道数
input = torch.reshape(input, (-1, 1, 5, 5))

# print(input.shape)  # print(input.shape)#

class MyModule(nn.Module):
    def __init__(self):
        super(MyModule, self).__init__()
        self.maxpool1 = MaxPool2d(kernel_size=3, ceil_mode=True)

    def forward(self, x):
        output = self.maxpool1(x)
        return output

myModule1 = MyModule()
output = myModule1(input)
print(output)

Note：要将input转为浮点数。

tensor([[[[2., 3.],
[5., 1.]]]])

若ceil_model=False：

tensor([[[[2.]]]])

最大池化的目的：保留输入的特征，同时把数据量减小。

nn_maxpool.py

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 = torchvision.datasets.CIFAR10('./dataset', train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=64)

class MyModule(nn.Module):
    def __init__(self):
        super(MyModule, self).__init__()
        self.maxpool1 = MaxPool2d(kernel_size=3, ceil_mode=False)

    def forward(self, x):
        output = self.maxpool1(x)
        return output

myModule1 = MyModule()

writer = SummaryWriter('./logs')
step = 0
for data in dataloader:
    imgs, targets = data
    writer.add_images('input', imgs, step)
    output = myModule1(imgs)
    writer.add_images('output', output, step)
    step = step + 1
writer.close()

4. 非线性激活

4.1 使用ReLU非线性激活

Note：inplace

图片来源于：b站up主我是土堆

nn_relu.py

import torch
from torch import nn
from torch.nn import ReLU

input = torch.tensor([
    [1, -0.5],
    [-1, 3]
])
# batch_size,channel,HW
input = torch.reshape(input, (-1, 1, 2, 2))

class MyModule(nn.Module):
    def __init__(self):
        super(MyModule, self).__init__()
        self.relu1 = ReLU()

    def forward(self, x):
        output = self.relu1(x)
        return output

myModule1 = MyModule()
output = myModule1(input)
print(output)

tensor([[[[1., 0.],
[0., 3.]]]])

4.2 使用Sigmoid非线性激活

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 = torchvision.datasets.CIFAR10('./dataset', train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=64)

class MyModule(nn.Module):
    def __init__(self):
        super(MyModule, self).__init__()
        self.relu1 = ReLU()
        self.sigmoid1 = Sigmoid()

    def forward(self, x):
        output = self.sigmoid1(x)
        return output

myModule1 = MyModule()

writer = SummaryWriter('logs')
step = 0
for data in dataloader:
    imgs, targets = data
    writer.add_images('input', imgs, step)
    output = myModule1(imgs)
    writer.add_images('output', output, step)
    step = step + 1
writer.close()

5. 线性激活

nn_linear.py

import torch
import torchvision
from torch import nn
from torch.nn import Linear
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10('./dataset', train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=64)

class MyModule(nn.Module):
    def __init__(self):
        super(MyModule, self).__init__()
        self.linear1 = Linear(196608, 10)  # in_features,out_features

    def forward(self, x):
        output = self.linear1(x)
        return output

myModule1 = MyModule()

for data in dataloader:
    imgs, targets = data
    # print(imgs.shape)  # torch.Size([64, 3, 32, 32])
    # output = torch.reshape(imgs, (1, 1, 1, -1))
    # print(output.shape)  # torch.Size([1, 1, 1, 196608])
    # 可以使用torch.flatten来拉成一维向量
    output = torch.flatten(imgs)
    # print(output.shape)  # torch.Size([196608])
    output = myModule1(output)
    # print(output.shape)  # torch.Size([10])