PyTorch 中 fc.bias 和 fc.weight 的大小不匹配

Posted 2023-03-12

【中文标题】PyTorch 中 fc.bias 和 fc.weight 的大小不匹配

【英文标题】：Size mismatch for fc.bias and fc.weight in PyTorch

【发布时间】：2019-05-05 20:57:46

【问题描述】：

我使用迁移学习方法来训练模型并保存检测到的最佳权重。在另一个脚本中，我尝试使用保存的权重进行预测。但我收到如下错误。我使用 ResNet 对网络进行微调，并有 4 个类。

RuntimeError: Error(s) in loading state_dict for ResNet:
size mismatch for fc.bias: copying a param of torch.Size([1000]) from 
checkpoint, where the shape is torch.Size([4]) in current model.
size mismatch for fc.weight: copying a param of torch.Size([1000, 
512]) from checkpoint, where the shape is torch.Size([4, 512]) in 
current model.

我正在使用以下代码来预测输出：

checkpoint = torch.load("./models/custom_model13.model")
model = resnet18(pretrained=True)

model.load_state_dict(checkpoint)
model.eval()

def predict_image(image_path):
    transformation = transforms.Compose([
        transforms.RandomResizedCrop(224),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
        ])
    image_tensor = transformation(image).float()
    image_tensor = image_tensor.unsqueeze_(0)

    if torch.cuda.is_available():
        image_tensor.cuda()

    input = Variable(image_tensor)
    output = model(input)

    index = output.data.numpy().argmax()
    return index

if __name__ == "main":
    imagefile = "image.png"
    imagepath = os.path.join(os.getcwd(),imagefile)
    prediction = predict_image(imagepath)
    print("Predicted Class: ",prediction)

以及下面的代码来训练和保存模型：

Data_dir = 'Dataset'
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")
print (device)

def save_models(epochs, model):
    torch.save(model.state_dict(), "custom_model.model".format(epochs))
    print("Checkpoint Saved")

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('Epoch /'.format(epoch, num_epochs - 1))
        print('-' * 10)

        # Each epoch has a training and validation phase
        for phase in ['train', 'val']:
            if phase == 'train':
                scheduler.step()
                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)

                # 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)
                    loss = criterion(outputs, labels)

                    # 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)

            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))

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

        print()

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

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

def visualize_model(model, num_images=6):
    was_training = model.training
    model.eval()
    images_so_far = 0
    fig = plt.figure()

    with torch.no_grad():
        for i, (inputs, labels) in enumerate(dataloaders['val']):
            inputs = inputs.to(device)
            labels = labels.to(device)

            outputs = model(inputs)
            _, preds = torch.max(outputs, 1)

            for j in range(inputs.size()[0]):
                images_so_far += 1
                ax = plt.subplot(num_images//2, 2, images_so_far)
                ax.axis('off')
                ax.set_title('predicted: '.format(class_names[preds[j]]))
                imshow(inputs.cpu().data[j])

                if images_so_far == num_images:
                    model.train(mode=was_training)
                    return
        model.train(mode=was_training)

model_ft = models.resnet18(pretrained=True)
num_ftrs = model_ft.fc.in_features
model_ft.fc = nn.Linear(num_ftrs, 4)

model_ft = model_ft.to(device)

criterion = nn.CrossEntropyLoss()


optimizer_ft = optim.SGD(model_ft.parameters(), lr=0.001, momentum=0.9)

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)

【参考方案1】：

原因：

您以这种方式训练了从resnet18 派生的模型：

model_ft = models.resnet18(pretrained=True)
num_ftrs = model_ft.fc.in_features
model_ft.fc = nn.Linear(num_ftrs, 4)

也就是说，您更改了最后一个 nn.Linear 层以输出 4 暗度预测而不是默认的 1000。 当您尝试加载模型进行预测时，您的代码是：

model = resnet18(pretrained=True)    
model.load_state_dict(checkpoint)

您确实没有将最后一个nn.Linear 层的相同更改应用于model，因此您尝试加载的checkpoint 不适合。

修复：

(1) 在加载 checkpoint 之前应用相同的更改：

model = resnet18(pretrained=True)    
num_ftrs = model_ft.fc.in_features
model_ft.fc = nn.Linear(num_ftrs, 4)  # make the change
model.load_state_dict(checkpoint)  # load

(2) 更好的是，使用num_classes 参数来构造resnet，并以所需的输出数量开头：

model = resnet18(pretrained=True, num_classes=4)  
model.load_state_dict(checkpoint)  # load

【讨论】：

感谢您的回复，我发现了错误，我使用了第二次修复，它就像一个魅力！

现在我使用单独的方法来自动使用 num_classes

@AmritDas 我可以在没有 FC 层的情况下从 resnet 中提取特征吗？