【PyTorch】土堆完整的项目实战

最新推荐文章于 2025-02-21 08:58:51 发布

东东记笔记

最新推荐文章于 2025-02-21 08:58:51 发布

阅读量198

点赞数

文章标签： pytorch python 深度学习

本文链接：https://blog.youkuaiyun.com/lidongfang6688/article/details/129031044

版权

【PyTorch深度学习快速入门教程（绝对通俗易懂！）【小土堆】】 https://www.bilibili.com/video/BV1hE411t7RN?p=18&vd_source=0f9b599f83ed1428365ed2868460a265

import torch.optim
from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torch.utils.data import DataLoader
from torchvision import datasets
from torchvision import transforms
import torch.nn as nn
#加载数据集
#训练数据集
dataset1 = datasets.CIFAR10("tourchvision_tranforms_dataset",train=True,transform=transforms.ToTensor(),download=True)
dataloader1 = DataLoader(dataset1,64,shuffle=True)
#测试数据集
test_dataset = datasets.CIFAR10("tourchvision_tranforms_dataset",train=False,transform=transforms.ToTensor(),download=True)
dataloader2 = DataLoader(test_dataset,64,shuffle=True)
#定义网络模型
class model1(nn.Module):
    def __init__(self):
        super(model1, self).__init__()
        self.seq1 = nn.Sequential(
            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),
            nn.Flatten(),
            nn.Linear(1024,64),
            nn.Linear(64,10)
        )
    def forward(self,data):
        out_data = self.seq1(data)
        return out_data

this_model = model1()
#训练轮数
epoch = 6
#定义优化器-随机梯度下降
this_optim1 = torch.optim.SGD(this_model.parameters(),lr=0.03)
#损失函数
this_loss1 = nn.CrossEntropyLoss()
writer1 = SummaryWriter("logs_2")
all_loss = 0
#训练
for i in range(epoch):
    all_loss = 0
    print("------第{}轮训练------".format(i+1))
    for loader1 in dataloader1:
        imgs,lables = loader1
        #将优化器参数清零
        this_optim1.zero_grad()
        out_val = this_model(imgs)
        #计算顺损失函数值
        loss_val = this_loss1(out_val,lables)
        #反向传播函数
        loss_val.backward()
        this_optim1.step()
        all_loss = all_loss + loss_val
    print("第{}轮训练结束.  all_loss={}".format((i+1),all_loss))
    writer1.add_scalar("train_all_loss",all_loss,i+1)
    #用测试数据集测试模型正确率
    #测试集不使用梯度下降
    test_all_loss = 0
    all_righ = 0
    with torch.no_grad():
        for loader in dataloader2:
            imgs,lables = loader
            print(lables)
            out = this_model(imgs)
            print(out)
            print("--------")
            test_loss = this_loss1(out,lables)
            #测试数据集损失函数值
            test_all_loss = test_all_loss + test_loss
            #测试数据集正确率
            right = (out.argmax(1) == lables).sum()
            all_righ = all_righ + right
        writer1.add_scalar("test_loss",test_all_loss,(i+1))
        print("测试集正确率：{}".format(all_righ / len(test_dataset)))
        writer1.add_scalar("right_rate", all_righ / len(test_dataset),(i+1))
#保存已经训练好的网络模型
torch.save(this_model,"final_modle.pth")
writer1.close()