一、对于损失函数的理解

• 计算实际输出和目标之间的差距
• 为我们更新输出提供一定的依据

二、头文件

nn.L1Loss

大概含义：

代码：

import torch
from torch.nn import L1Loss

output = torch.tensor([1,2,3],dtype=float)
target = torch.tensor([1,2,5],dtype=float)

output = torch.reshape(output,(1,1,1,3))
target = torch.reshape(target,(1,1,1,3))

loss = L1Loss()
result = loss(output,target)
print(result)

输出：

tensor(0.6667, dtype=torch.float64)

nn.MSELoss

大概含义：

代码：

import torch
from torch.nn import L1Loss, MSELoss

output = torch.tensor([1,2,3],dtype=float)
target = torch.tensor([1,2,5],dtype=float)

output = torch.reshape(output,(1,1,1,3))
target = torch.reshape(target,(1,1,1,3))

loss = L1Loss()
result = loss(output,target)

loss2 = MSELoss()
result2 = loss2(output,target)
print(result2)

输出：

tensor(1.3333, dtype=torch.float64)

nn.CrossEntropyLoss

import torch
from torch.nn import L1Loss, MSELoss, CrossEntropyLoss

output = torch.tensor([1,2,3],dtype=float)
target = torch.tensor([1,2,5],dtype=float)

output = torch.reshape(output,(1,1,1,3))
target = torch.reshape(target,(1,1,1,3))

loss = L1Loss()
result = loss(output,target)

loss2 = MSELoss()
result2 = loss2(output,target)


x = torch.tensor([0.1,0.2,0.3])
y = torch.tensor(1)  #正确的类别
loss3 = CrossEntropyLoss()
result3 = loss3(x,y)
print(result3)

输出：

tensor(1.1019)

包含 batch_size 的代码：

import torch
import torch.nn as nn

# 假设有 3 个样本，每个样本有 5 个类别
x = torch.tensor([[0.1, 0.2, 0.3, 0.4, 0.5],   # 第一个样本的 logits
                  [0.3, 0.4, 0.2, 0.5, 0.1],   # 第二个样本的 logits
                  [0.5, 0.1, 0.2, 0.4, 0.3]])  # 第三个样本的 logits

# 每个样本的真实类别标签
y = torch.tensor([1, 2, 0])  # 第一个样本属于类别 1，第二个样本属于类别 2，第三个样本属于类别 0

# 定义损失函数
loss_fn = nn.CrossEntropyLoss()

# 计算损失
loss = loss_fn(x, y)

print(f"Loss: {loss.item()}")

三、搭建模型与损失函数的结合

搭建模型
代码：

import torchvision
from torch import nn
from torch.nn import MaxPool2d, Conv2d, Flatten, Linear
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10("datasets",train=False,transform=torchvision.transforms.ToTensor(),download=True)
dataloader = DataLoader(dataset,batch_size=1)

class Mary(nn.Module):
    def __init__(self):
        super(Mary,self).__init__()
        self.model1 = nn.Sequential(
            Conv2d(3, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 64, 5, padding=2),
            MaxPool2d(2),
            Flatten(),
            Linear(1024, 64),
            Linear(64, 10)
        )
    def forward(self,x):
        x = self.model1(x)
        return x
Yorelee = Mary()

for data in dataloader:
    img,target = data
    output = Yorelee(img)
    print(output)
    print(target)
    loss = nn.CrossEntropyLoss()
    result_loss = loss(output,target)
    print(result_loss)
    print("***********************")
    result_loss.backward()

输出：

tensor([[ 0.0956, -0.0318, -0.0674, -0.0565,  0.1168,  0.0389,  0.0496, -0.0039,
         -0.0221,  0.1028]], grad_fn=<AddmmBackward0>)
tensor([3])
tensor(2.3833, grad_fn=<NllLossBackward0>)

打个断点，单步执行前，注意此时还没有 grad：

单步执行后，此时就有了 grad，为后续优化器的选择做了铺垫：