学习pytorch14 损失函数与反向传播

B站小土堆pytorch视频学习

官网

https://pytorch.org/docs/stable/nn.html#loss-functions

损失函数

L1Loss MAE 平均

import torch

input = torch.tensor([1, 2, 3], dtype=float)
# target = torch.tensor([1, 2, 5], dtype=float)
target = torch.tensor([[[[1, 2, 5]]]], dtype=float) # shape [1, 1, 1, 3]
input = torch.reshape(input, (1,1,1,3))
# target = torch.reshape(target, (1,1,1,3))
print(input.shape)
print(target.shape)

loss1 = torch.nn.L1Loss()
loss2 = torch.nn.L1Loss(reduction="sum")
result1 = loss1(input, target)
print(result1) # tensor(0.6667, dtype=torch.float64)
result2 = loss2(input, target)
print(result2) # tensor(2., dtype=torch.float64)

MSELoss 平方差

import torch

input = torch.tensor([1, 2, 3], dtype=float)
# target = torch.tensor([1, 2, 5], dtype=float)
target = torch.tensor([[[[1, 2, 5]]]], dtype=float) # shape [1, 1, 1, 3]
input = torch.reshape(input, (1,1,1,3))
# target = torch.reshape(target, (1,1,1,3))
print(input.shape)
print(target.shape)

loss_mse = torch.nn.MSELoss(reduction='mean')
result_mse = loss_mse(input, target)
print(result_mse) # tensor(1.3333, dtype=torch.float64)
loss_mse2 = torch.nn.MSELoss(reduction='sum')
result_mse2 = loss_mse2(input, target)
print(result_mse2)   # tensor(4., dtype=torch.float64)

CROSSENTROPYLOSS 交叉熵损失

https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html#torch.nn.CrossEntropyLoss


在神经网络中，默认log是以e为底的，所以也可以写成ln

注意

1. 根据需求选择对应的loss函数
2. 注意loss函数的输入输出shape

code

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

test_set = torchvision.datasets.CIFAR10("./dataset", train=False, transform=torchvision.transforms.ToTensor(),
                                        download=True)

dataloader = DataLoader(test_set, batch_size=1)

class MySeq(nn.Module):
    def __init__(self):
        super(MySeq, self).__init__()
        self.model1 = Sequential(Conv2d(3, 32, kernel_size=5, stride=1, padding=2),
                                 MaxPool2d(2),
                                 Conv2d(32, 32, kernel_size=5, stride=1, padding=2),
                                 MaxPool2d(2),
                                 Conv2d(32, 64, kernel_size=5, stride=1, padding=2),
                                 MaxPool2d(2),
                                 Flatten(),
                                 Linear(1024, 64),
                                 Linear(64, 10)
                                 )

    def forward(self, x):
        x = self.model1(x)
        return x

loss = nn.CrossEntropyLoss()
myseq = MySeq()
print(myseq)
for data in dataloader:
    imgs, targets = data
    print(imgs.shape)
    output = myseq(imgs)
    result = loss(output, targets)
    print(result)

反向传播

在debug中的显示

显示在网络结构中，每一层的保护属性中，都有weight属性，梯度属性在weitht属性里面
先找模型结构 在找每一层 在找weight权重，梯度在weight权重里面

code

核心代码：result_loss.backward() # 要在最后获取 backward函数要挂在通过loss函数计算后的结果上。

# 模型定义、数据加载 同上个代码
for data in dataloader:
    imgs, targets = data
    print(imgs.shape)
    output = myseq(imgs)
    result_loss= loss(output, targets)
	result_loss.backward()  # 要在最后获取
    print(result_loss)
    print(result_loss.grad)