pytorch学习笔记—损失函数

注意事项：
1.CrossEntropyLoss()是来自nn模块下，和卷积，池化，全连接层一样

loss = nn.CrossEntropyLoss()

2.损失函数的input:是由特定要求的,第一个是batch,第二个是类别个数。不符合要求要进行reshape

x = torch.tensor([0.1,0.2,0.3])
print(x.shape)
targets = torch.tensor([1])
print(targets.shape)

input = torch.reshape(x,(1,3))#1表示batchsize, 3表示3个类别
print(input.shape)

 

 

 下面看完整代码

import torch
import torchvision
from torch import nn
from torch.nn import Sequential, Conv2d, Flatten
from torch.utils.data import DataLoader
from torchvision import transforms

datasets = torchvision.datasets.CIFAR10(root="datasets2",train=False,transform=transforms.ToTensor(),download=True)
dataloader = DataLoader(dataset=datasets,batch_size=64)


class nn_seq(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.model = Sequential(#使用sequential的好处不需要将方法一个一个拿出来
            nn.Conv2d(3,32,5,padding=2),
            nn.MaxPool2d(2),
            nn.Conv2d(32,32,5,padding=2),
            nn.MaxPool2d(2),
            nn.Conv2d(32,64,5,padding=2),
            nn.MaxPool2d(2),
            Flatten(),
            nn.Linear(1024,64),
            nn.Linear(64,10)
        )


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

nn_seq_1 = nn_seq()
loss = nn.CrossEntropyLoss()
# print(nn_seq_1)
# input = torch.ones((64,3,32,32))#使用内置的全1矩阵，第一个是batchsize,第二个是通道数，第三个是h,第四个是w
# output = nn_seq_1(input)
# print(output.shape)
for data in dataloader:#targets是目标类别
    imgs,targets = data
    output = nn_seq_1(imgs)
    print(output.shape)
    # print(output)
    # output1 = torch.reshape(output,(1,10))
    # print(output1)     没有必要进行这一步输出的结果是一样的
    real_loss = loss(output,targets)
    print(real_loss)