PyTorch训练VGG16模型后无法绘制训练集与验证集损失及准确率曲线

核心问题点 #

• 原训练函数未对每轮的损失、准确率数值做持久化存储，仅做了控制台打印，训练完成后无历史数据可用于绘图
• 绘图代码逻辑完全错误：
  • 未初始化loss_values就直接调用append方法
  • Pytorch模型对象不是字典，无法通过model['train','validation']读取损失数据
  • 未区分损失、准确率两类指标的绘图逻辑
  • 定义的ss_values变量无任何实际作用

修复步骤 #

1. 调整训练函数，新增历史数据记录逻辑 #

修改train_model函数，初始化存储指标的列表，每轮计算完指标后存入列表，最终随模型一同返回：

import time
import copy
import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
import matplotlib.pyplot as plt

def train_model(model, criterion, optimizer, scheduler, num_epochs, device, dataloaders, dataset_sizes):
    since = time.time()
    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0
    # 初始化存储历史指标的列表
    train_loss_history = []
    val_loss_history = []
    train_acc_history = []
    val_acc_history = []

    for epoch in range(num_epochs):
        print('Epoch {}/{} LR {:.6f}'.format(epoch, num_epochs - 1, scheduler.get_last_lr()[0]))
        print('-' * 10)

        for phase in ['train', 'validation']:
            if phase == 'train':
                model.train()
            else:
                model.eval()

            running_loss = 0.0
            running_corrects = 0

            for inputs, labels in dataloaders[phase]:
                inputs = inputs.to(device)
                labels = labels.to(device)
                optimizer.zero_grad()

                with torch.set_grad_enabled(phase == 'train'):
                    outputs = model(inputs)
                    _, preds = torch.max(outputs, 1)
                    loss = criterion(outputs, labels)

                    if phase == 'train':
                        loss.backward()
                        optimizer.step()

                running_loss += loss.item() * inputs.size(0)
                running_corrects += torch.sum(preds == labels.data)
            if phase == 'train':
                scheduler.step()

            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = running_corrects.double() / dataset_sizes[phase]
            print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))

            # 存入对应指标列表，acc转为普通数值避免后续绘图报错
            if phase == 'train':
                train_loss_history.append(epoch_loss)
                train_acc_history.append(epoch_acc.cpu().item())
            else:
                val_loss_history.append(epoch_loss)
                val_acc_history.append(epoch_acc.cpu().item())

            if phase == 'validation' and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())

        print()

    time_elapsed = time.time() - since
    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
    print('Best val Acc: {:4f}'.format(best_acc))

    model.load_state_dict(best_model_wts)
    # 返回历史指标
    return model, (train_loss_history, val_loss_history, train_acc_history, val_acc_history)

2. 修正训练调用代码 #

修复原代码中损失函数和优化器定义的语法错误，同时接收返回的历史指标：

# 修复原代码语法错误：拆分两行
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.0001)

exp_lr_scheduler = lr_scheduler.StepLR(optimizer, step_size=7, gamma=0.1)
# 接收返回的模型和历史指标
model, history = train_model(model, criterion, optimizer, exp_lr_scheduler, num_epochs=5, device=device, dataloaders=dataloaders, dataset_sizes=dataset_sizes)
model = model.to(device)
train_loss, val_loss, train_acc, val_acc = history

3. 替换为正确的绘图代码 #

分别绘制损失曲线和准确率曲线：

# 绘制损失曲线
plt.figure(figsize=(12,5))
plt.subplot(1,2,1)
plt.plot(train_loss, label='训练损失')
plt.plot(val_loss, label='验证损失')
plt.xlabel('轮次')
plt.ylabel('损失值')
plt.title('训练/验证损失变化')
plt.legend(loc='upper right')

# 绘制准确率曲线
plt.subplot(1,2,2)
plt.plot(train_acc, label='训练准确率')
plt.plot(val_acc, label='验证准确率')
plt.xlabel('轮次')
plt.ylabel('准确率')
plt.title('训练/验证准确率变化')
plt.legend(loc='lower right')

plt.tight_layout()
plt.show()

内容的提问来源于stack exchange，提问作者cliff