### 实现ResNet模型并在GPU上训练 要在GPU上使用PyTorch和`timm`库实现ResNet模型并绘制训练过程中的准确度曲线，可以通过以下方式完成： #### 安装依赖库 首先需要确保已安装必要的Python包。如果尚未安装`timm`库，则可通过pip命令安装： ```bash pip install timm ``` #### 加载ResNet模型 利用`timm`库加载预训练的ResNet模型，并将其移至GPU设备以便高效计算[^1]。 ```python import torch import timm # 创建ResNet模型实例（例如resnet50） model = timm.create_model('resnet50', pretrained=True) # 将模型移到GPU上 train_on_gpu = torch.cuda.is_available() if train_on_gpu: model = model.cuda() print(model) ``` #### 数据准备 为了训练模型，需要准备好数据集并将它们转换为适合PyTorch处理的形式。这里假设使用的是CIFAR-10或其他常见分类数据集。 ```python from torchvision import datasets, transforms from torch.utils.data import DataLoader transform_train = transforms.Compose([ transforms.RandomCrop(32, padding=4), transforms.RandomHorizontalFlip(), transforms.ToTensor(), transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)), ]) transform_test = transforms.Compose([ transforms.ToTensor(), transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)), ]) trainset = datasets.CIFAR10(root='./data', train=True, download=True, transform=transform_train) testset = datasets.CIFAR10(root='./data', train=False, download=True, transform=transform_test) trainloader = DataLoader(trainset, batch_size=128, shuffle=True, num_workers=2) testloader = DataLoader(testset, batch_size=100, shuffle=False, num_workers=2) ``` #### 训练与验证函数 编写训练循环以及评估模型性能的功能模块。 ```python def train_model(model, criterion, optimizer, scheduler, num_epochs=25): device = 'cuda' if torch.cuda.is_available() else 'cpu' accuracy_history = [] loss_history = [] for epoch in range(num_epochs): running_loss = 0.0 correct_predictions = 0 total_samples = 0 # 设置模型为训练模式 model.train() for inputs, labels in trainloader: inputs, labels = inputs.to(device), labels.to(device) # 清零梯度 optimizer.zero_grad() # 前向传播 + 反向传播 + 更新权重 outputs = model(inputs) _, preds = torch.max(outputs, 1) loss = criterion(outputs, labels) loss.backward() optimizer.step() # 统计损失和预测正确数 running_loss += loss.item() * inputs.size(0) correct_predictions += torch.sum(preds == labels.data).item() total_samples += inputs.size(0) # 调整学习率 scheduler.step() epoch_loss = running_loss / total_samples epoch_acc = correct_predictions / total_samples print(f'Epoch {epoch}/{num_epochs - 1} Loss: {epoch_loss:.4f}, Acc: {epoch_acc:.4f}') # 存储历史记录 accuracy_history.append(epoch_acc) loss_history.append(epoch_loss) return accuracy_history, loss_history def evaluate_model(model, dataloader): device = 'cuda' if torch.cuda.is_available() else 'cpu' model.eval() correct_predictions = 0 total_samples = 0 with torch.no_grad(): for inputs, labels in dataloader: inputs, labels = inputs.to(device), labels.to(device) outputs = model(inputs) _, preds = torch.max(outputs, 1) correct_predictions += torch.sum(preds == labels.data).item() total_samples += inputs.size(0) acc = correct_predictions / total_samples return acc ``` #### 配置优化器和损失函数 设置合适的优化算法及其参数配置来提升收敛速度和最终效果。 ```python criterion = torch.nn.CrossEntropyLoss() optimizer = torch.optim.SGD(model.parameters(), lr=0.1, momentum=0.9, weight_decay=5e-4) scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1) ``` #### 启动训练流程 执行上述定义好的逻辑链路以启动整个实验周期。 ```python accuracy_hist, loss_hist = train_model(model, criterion, optimizer, scheduler, num_epochs=20) final_accuracy = evaluate_model(model, testloader) print(f'Test Accuracy after training: {final_accuracy*100:.2f}%') ``` #### 绘制准确度变化趋势图 最后借助matplotlib工具箱展示随时间推移而改变的学习成果表现情况。 ```python import matplotlib.pyplot as plt plt.figure(figsize=(10,5)) plt.plot(range(len(accuracy_hist)), accuracy_hist, label='Training Accuracy') plt.title('Accuracy over Epochs') plt.xlabel('Epoch') plt.ylabel('Accuracy') plt.legend() plt.grid(True) plt.show() ``` 以上即是在GPU支持下运用PyTorch框架配合timm库构建ResNet架构并通过可视化手段监控其改进历程的方法概述[^1][^2]。