pytorch神经网络初学-优快云博客

本文链接：https://blog.youkuaiyun.com/dylan1998/article/details/111505508

本文介绍了一个使用PyTorch实现的卷积神经网络（CNN）模型，包括网络结构定义、前向传播过程、损失计算及参数更新等关键步骤，并通过随机生成的数据进行模型训练演示。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

import torch
import torch.nn as nn
import torch.nn.functional as F


class Net(nn.Module):  # 括号里的为父类

    def __init__(self):  # 构造方法
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 6, 3)  # 1个输入，6个输出特征，3x3的卷积核
        self.conv2 = nn.Conv2d(6, 16, 3)
        self.fc1 = nn.Linear(16 * 6 * 6, 120)  # 全连接层
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, x):
        x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))  # 方阵写一个数字就行
        x = F.max_pool2d(F.relu(self.conv2(x)), 2)
        x = x.view(-1, self.num_flat_features(x))  # -1不指定reshape成多少行，根据具体情况
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x

    def num_flat_features(self, x):
        size = x.size()[1:]
        num_features = 1
        for s in size:
            num_features *= s
        return num_features

net = Net()

input = torch.randn(1, 1, 32, 32) # nSamples x nChannels x Height x Width, channels为特征的数量
output = net(input)
target = torch.randn(10)
target = target.view(1, -1)
criterion = nn.MSELoss() # 损失函数

loss = criterion(output, target)
loss

tensor(0.9012, grad_fn=<MseLossBackward>)

net.zero_grad()
loss.backward()

learning_rate = 0.01
for f in net.parameters():
    f.data.sub_(f.grad.data * learning_rate)

import torch.optim as optim

optimizer = optim.SGD(net.parameters(), lr=0.01) # 自定义优化函数

optimizer.zero_grad()
output = net(input)
loss = criterion(output, target) # loss函数
loss.backward()
optimizer.step() # 对参数执行更新