PyTorch二分类器用nn.BCEWithLogitsLoss后准确率大于1的问题

我将原本使用nn.CrossEntropyLoss()的二分类模型更换为nn.BCEWithLogitsLoss()，调整训练函数适配该损失函数后，模型训练时准确率数值出现大于1的异常情况。以下是相关代码及训练输出，恳请协助排查：

数据处理与训练函数代码 #

# Data augmentation and normalization for training
# Just normalization for validation
data_transforms = {
    'train': transforms.Compose([
        transforms.RandomResizedCrop(224),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
    'val': transforms.Compose([
        transforms.Resize(256),
        transforms.CenterCrop(224),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
}

data_dir = '/kaggle/input/catsndogsorg/hymenoptera_data'
image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x),
                                          data_transforms[x])
                  for x in ['train', 'val']}
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=4,
                                             shuffle=True, num_workers=4)
              for x in ['train', 'val']}
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'val']}
class_names = image_datasets['train'].classes

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

#############

def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
    since = time.time()

    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0

    for epoch in range(num_epochs):
        print(f'Epoch {epoch}/{num_epochs - 1}')
        print('-' * 10)

        # Each epoch has a training and validation phase
        for phase in ['train', 'val']:
            if phase == 'train':
                model.train()  # Set model to training mode
            else:
                model.eval()   # Set model to evaluate mode

            running_loss = 0.0
            running_corrects = 0

            # Iterate over data.
            for inputs, labels in dataloaders[phase]:
                inputs = inputs.to(device)
                labels = labels.to(device).unsqueeze(1)

                # zero the parameter gradients
                optimizer.zero_grad()

                # forward
                # track history if only in train
                with torch.set_grad_enabled(phase == 'train'):
                    outputs = model(inputs)
                    _, preds = torch.max(outputs, 1)
                    #print(outputs, labels)
                    loss = criterion(outputs, labels.float())
                    print(loss)
                    # backward + optimize only if in training phase
                    if phase == 'train':
                        loss.backward()
                        optimizer.step()

                # statistics
                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(f'{phase} Loss: {epoch_loss:.4f} Acc: {epoch_acc:.4f}')

            # deep copy the model
            if phase == 'val' and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())

        print()

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

    # load best model weights
    model.load_state_dict(best_model_wts)
    return model

模型定义代码 #

model_ft = models.resnet18(weights='ResNet18_Weights.DEFAULT')
num_ftrs = model_ft.fc.in_features
# Here the size of each output sample is set to 2.
# Alternatively, it can be generalized to nn.Linear(num_ftrs, len(class_names)).
model_ft.fc = nn.Linear(num_ftrs, 1)

model_ft = model_ft.to(device)

criterion = nn.BCEWithLogitsLoss()

# Observe that all parameters are being optimized
optimizer_ft = optim.SGD(model_ft.parameters(), lr=0.001, momentum=0.9)

# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)

model_ft = train_model(model_ft, criterion, optimizer_ft, 
     exp_lr_scheduler,num_epochs=25)

训练输出 #

outputs shape:  torch.Size([4, 1])
labels shape:  torch.Size([4, 1])

logits: tensor(0.3511,grad_fn<BinaryCrossEntropyWithLogitsBackward0>)

train Loss: 1.0000 Acc: 2.0164
val Loss: 1.0000 Acc: 1.8105

问题根源与解决方案 #

问题分析 #

1. 预测逻辑错误：torch.max(outputs, 1)是针对多分类（输出维度等于类别数）的逻辑，但切换为BCEWithLogitsLoss后，模型输出为单维度[batch_size,1]（代表正类的对数几率），此时torch.max无法正确生成类别预测，反而返回全0的索引张量。
2. 维度不匹配导致广播错误：preds是[batch_size,]形状，而labels被unsqueeze(1)处理为[batch_size,1]形状，两者比较时触发PyTorch广播规则，生成[batch_size,batch_size]的布尔张量，torch.sum会累加所有元素，导致running_corrects数值远大于实际样本数，最终准确率超过1。

修正代码 #

修改训练函数中的预测与准确率计算逻辑，适配单输出二分类的场景：

def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
    since = time.time()

    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0

    for epoch in range(num_epochs):
        print(f'Epoch {epoch}/{num_epochs - 1}')
        print('-' * 10)

        for phase in ['train', 'val']:
            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)  # 保持原始[batch_size,]形状

                optimizer.zero_grad()

                with torch.set_grad_enabled(phase == 'train'):
                    outputs = model(inputs)
                    # 单输出二分类的预测逻辑：sigmoid后判断是否大于0.5，压缩维度匹配labels
                    preds = (torch.sigmoid(outputs) > 0.5).squeeze()
                    # 损失计算需要labels与outputs形状一致，所以给labels加一维
                    loss = criterion(outputs, labels.unsqueeze(1).float())
                    
                    if phase == 'train':
                        loss.backward()
                        optimizer.step()

                running_loss += loss.item() * inputs.size(0)
                # 此时preds与labels形状一致，sum统计正确匹配的样本数
                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(f'{phase} Loss: {epoch_loss:.4f} Acc: {epoch_acc:.4f}')

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

        print()

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

    model.load_state_dict(best_model_wts)
    return model

关键说明 #

• BCEWithLogitsLoss要求目标张量labels与模型输出outputs形状一致，因此损失计算时需将labels转为[batch_size,1]的float张量。
• 准确率计算时，将preds压缩为[batch_size,]形状，与原始labels形状匹配，避免广播错误，确保统计的是正确匹配的样本数量。

内容的提问来源于stack exchange，提问作者Alain Michael Janith Schroter