深度学习|PyTorch学习笔记（四）（PyTorch基础实战） pytorch|深度学习|人工智能|Pyt

PyTorch实战：以FashionMNIST时装分类为例：往期学习资料推荐： 1.Pytorch实战笔记_GoAI的博客-CSDN博客
2.Pytorch入门教程_GoAI的博客-CSDN博客
本系列目录： PyTorch学习笔记（一）：PyTorch环境安装
PyTorch学习笔记（二）：简介与基础知识
PyTorch学习笔记（三）：PyTorch主要组成模块
PyTorch学习笔记（四）：PyTorch基础实战
PyTorch学习笔记（五）：模型定义、修改、保存
后续继续更新！！！！
1.导入必要的包

import os import numpy as np import pandas as pd import torch import torch.nn as nn import torch.optim as optim from torch.utils.data import Dataset, DataLoader

2 配置训练环境和超参数

# 配置GPU，这里有两种方式 ## 方案一：使用os.environ os.environ['CUDA_VISIBLE_DEVICES'] = '0' # 方案二：使用“device”，后续对要使用GPU的变量用.to(device)即可 #device = torch.device("cuda:1" if torch.cuda.is_available() else "cpu")## 配置其他超参数，如batch_size, num_workers, learning rate, 以及总的epochs batch_size = 256 num_workers = 4# 对于Windows用户，这里应设置为0，否则会出现多线程错误 lr = 1e-4 epochs = 20

3 数据读取和加载这里同时展示两种方式:

下载并使用PyTorch提供的内置数据集
从网站下载以csv格式存储的数据，读入并转成预期的格式
第一种数据读入方式只适用于常见的数据集，如MNIST，CIFAR10等，PyTorch官方提供了数据下载。这种方式往往适用于快速测试方法（比如测试下某个idea在MNIST数据集上是否有效）
第二种数据读入方式需要自己构建Dataset，这对于PyTorch应用于自己的工作中十分重要

同时，还需要对数据进行必要的变换，比如说需要将图片统一为一致的大小，以便后续能够输入网络训练；需要将数据格式转为Tensor类，等等。

from torchvision import transforms# 设置数据变换 image_size = 28 data_transform = transforms.Compose([ transforms.ToPILImage(),# 这一步取决于后续的数据读取方式，如果使用内置数据集则不需要 transforms.Resize(image_size), transforms.ToTensor() ])

## 读取方式一：使用torchvision自带数据集，下载可能需要一段时间 from torchvision import datasetstrain_data = https://www.it610.com/article/datasets.FashionMNIST(root='./', train=True, download=True, transform=data_transform) test_data = https://www.it610.com/article/datasets.FashionMNIST(root='./', train=False, download=True, transform=data_transform)

## 读取方式二：读入csv格式的数据，自行构建Dataset类，即自定义数据集 # csv数据下载链接：https://www.kaggle.com/zalando-research/fashionmnist class FMDataset(Dataset): def __init__(self, df, transform=None): self.df = df self.transform = transform self.images = df.iloc[:,1:].values.astype(np.uint8) self.labels = df.iloc[:, 0].valuesdef __len__(self): return len(self.images)def __getitem__(self, idx): image = self.images[idx].reshape(28,28,1) label = int(self.labels[idx]) if self.transform is not None: image = self.transform(image) else: image = torch.tensor(image/255., dtype=torch.float) label = torch.tensor(label, dtype=torch.long) return image, labeltrain_df = pd.read_csv("./FashionMNIST/fashion-mnist_train.csv") test_df = pd.read_csv("./FashionMNIST/fashion-mnist_test.csv") train_data = https://www.it610.com/article/FMDataset(train_df, data_transform) test_data = FMDataset(test_df, data_transform)

在构建训练和测试数据集完成后，需要定义DataLoader类，以便在训练和测试时加载数据

train_loader = DataLoader(train_data, batch_size=batch_size, shuffle=True, num_workers=num_workers, drop_last=True) test_loader = DataLoader(test_data, batch_size=batch_size, shuffle=False, num_workers=num_workers)

读入后，我们可以做一些数据可视化操作，主要是验证我们读入的数据是否正确

import matplotlib.pyplot as plt image, label = next(iter(train_loader)) print(image.shape, label.shape) plt.imshow(image[0][0], cmap="gray")

torch.Size([256, 1, 28, 28]) torch.Size([256])

文章图片
???
4.模型设计

# 使用CNN class Net(nn.Module): def __init__(self): super(Net, self).__init__() self.conv = nn.Sequential( nn.Conv2d(1, 32, 5), nn.ReLU(), nn.MaxPool2d(2, stride=2), nn.Dropout(0.3), nn.Conv2d(32, 64, 5), nn.ReLU(), nn.MaxPool2d(2, stride=2), nn.Dropout(0.3) ) self.fc = nn.Sequential( nn.Linear(64*4*4, 512), nn.ReLU(), nn.Linear(512, 10) )def forward(self, x): x = self.conv(x) x = x.view(-1, 64*4*4) x = self.fc(x) # x = nn.functional.normalize(x) return x

model = Net() model = model.cuda()

5 设置损失函数和优化器

# 使用交叉熵损失函数 criterion = nn.CrossEntropyLoss()

# 使用Adam优化器 optimizer = optim.Adam(model.parameters(), lr=0.001)

6.训练与测试各自封装成函数，方便后续调用
关注两者的主要区别：

模型状态设置
是否需要初始化优化器
是否需要将loss传回到网络
是否需要每步更新optimizer

【深度学习|PyTorch学习笔记（四）（PyTorch基础实战）】此外，对于测试或验证过程，可以计算分类准确率

def train(epoch): # 设置训练状态 model.train() train_loss = 0 # 循环读取DataLoader中的全部数据 for data, label in train_loader: # 将数据放到GPU用于后续计算 data, label = data.cuda(), label.cuda() # 将优化器的梯度清0 optimizer.zero_grad() # 将数据输入给模型 output = model(data) # 设置损失函数 loss = criterion(output, label) # 将loss反向传播给网络 loss.backward() # 使用优化器更新模型参数 optimizer.step() # 累加训练损失 train_loss += loss.item() * data.size(0) train_loss = train_loss/len(train_loader.dataset) print('Epoch: {} \tTraining Loss: {:.6f}'.format(epoch, train_loss))

def val(epoch): # 设置验证状态 model.eval() val_loss = 0 gt_labels = [] pred_labels = [] # 不设置梯度 with torch.no_grad(): for data, label in test_loader: data, label = data.cuda(), label.cuda() output = model(data) preds = torch.argmax(output, 1) gt_labels.append(label.cpu().data.numpy()) pred_labels.append(preds.cpu().data.numpy()) loss = criterion(output, label) val_loss += loss.item()*data.size(0) # 计算验证集的平均损失 val_loss = val_loss/len(test_loader.dataset) gt_labels, pred_labels = np.concatenate(gt_labels), np.concatenate(pred_labels) # 计算准确率 acc = np.sum(gt_labels==pred_labels)/len(pred_labels) print('Epoch: {} \tValidation Loss: {:.6f}, Accuracy: {:6f}'.format(epoch, val_loss, acc))

for epoch in range(1, epochs+1): train(epoch) val(epoch)

Epoch: 1Training Loss: 0.664049 Epoch: 1Validation Loss: 0.421500, Accuracy: 0.852400 Epoch: 2Training Loss: 0.417311 Epoch: 2Validation Loss: 0.349790, Accuracy: 0.871200 Epoch: 3Training Loss: 0.355448 Epoch: 3Validation Loss: 0.318987, Accuracy: 0.879500 Epoch: 4Training Loss: 0.323644 Epoch: 4Validation Loss: 0.290521, Accuracy: 0.893800 Epoch: 5Training Loss: 0.301900 Epoch: 5Validation Loss: 0.266420, Accuracy: 0.901300 Epoch: 6Training Loss: 0.286696 Epoch: 6Validation Loss: 0.246448, Accuracy: 0.909700 Epoch: 7Training Loss: 0.271441 Epoch: 7Validation Loss: 0.241845, Accuracy: 0.911200 Epoch: 8Training Loss: 0.260185 Epoch: 8Validation Loss: 0.243311, Accuracy: 0.910800 Epoch: 9Training Loss: 0.247986 Epoch: 9Validation Loss: 0.225896, Accuracy: 0.916200 Epoch: 10Training Loss: 0.240718 Epoch: 10Validation Loss: 0.227848, Accuracy: 0.914700 Epoch: 11Training Loss: 0.232358 Epoch: 11Validation Loss: 0.220180, Accuracy: 0.917500 Epoch: 12Training Loss: 0.223933 Epoch: 12Validation Loss: 0.215308, Accuracy: 0.919400 Epoch: 13Training Loss: 0.218354 Epoch: 13Validation Loss: 0.211890, Accuracy: 0.919300 Epoch: 14Training Loss: 0.210027 Epoch: 14Validation Loss: 0.209707, Accuracy: 0.922700 Epoch: 15Training Loss: 0.203024 Epoch: 15Validation Loss: 0.208233, Accuracy: 0.925600 Epoch: 16Training Loss: 0.196965 Epoch: 16Validation Loss: 0.208209, Accuracy: 0.921900 Epoch: 17Training Loss: 0.193155 Epoch: 17Validation Loss: 0.200000, Accuracy: 0.926100 Epoch: 18Training Loss: 0.184376 Epoch: 18Validation Loss: 0.197259, Accuracy: 0.926200 Epoch: 19Training Loss: 0.184272 Epoch: 19Validation Loss: 0.200259, Accuracy: 0.926000 Epoch: 20Training Loss: 0.172641 Epoch: 20Validation Loss: 0.200177, Accuracy: 0.927100

7.模型保存