Dimension out of range (expected to be in range of [-1, 0], but got 1)

vision
1

I’m getting dimension out of range (expected to be in range of [-1, 0], but got 1) for the following code snippet. The size of the output is [20,133] and that of the target is [20].

def train(n_epochs, loaders, model, optimizer, criterion, use_cuda, save_path):
    """returns trained model"""
    # initialize tracker for minimum validation loss
    valid_loss_min = np.Inf 
    global total_train_loss
    global total_valid_loss
    
    if use_cuda:
        model.cuda()
    for epoch in range(1, n_epochs+1):
        # initialize variables to monitor training and validation loss
        train_loss = 0.0
        valid_loss = 0.0
        
        ###################
        # train the model #
        ###################
        model.train()
        for batch_idx, (data, target) in enumerate(loaders['train']):
            # move to GPU
            if use_cuda:
                data = data.cuda()
                target =  target.cuda()
            ## find the loss and update the model parameters accordingly
            ## record the average training loss, using something like
            ## train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))
            optimizer.zero_grad()
            
            output = model(data)
            print(output[0].shape,target.shape)
            loss = criterion(output[0],target)
            optimizer.step()
            train_loss += ((1/(batch_idx+1))*(loss.data-train_loss))
            
        ######################    
        # validate the model #
        ######################
        model.eval()
        for batch_idx, (data, target) in enumerate(loaders['valid']):
            # move to GPU
            if use_cuda:
                data, target = data.cuda(), target.cuda()
            ## update the average validation loss
            output = model(data)
            
            loss = criterion(output[0],target)
            valid_loss += ((1/(batch_idx+1))*(loss.data-valid_loss))
            
            
        # print training/validation statistics 
        print('Epoch: {} \tTraining Loss: {:.6f} \tValidation Loss: {:.6f}'.format(
            epoch, 
            train_loss,
            valid_loss
            ))
        
        ## TODO: save the model if validation loss has decreased
        total_train_loss.append(train_loss)
        total_valid_loss.append(valid_loss)
        
        if valid_loss <= valid_loss_min:
            print("Saving the model.....")
            torch.save(model.state_dict(),save_path)
            valid_loss_min = valid_loss
            
            
    # return trained model
    return model

The code runs for few data points in the batch but then stops with this error message. Any help?

Edit:
I doubt there is some error with the data loader cause in the training loop the output[0].size is [20,133] while for validation loop is [133]. Following is the code for the data loader.

import os
from torchvision import datasets
from glob import glob

### TODO: Write data loaders for training, validation, and test sets
## Specify appropriate transforms, and batch_sizes
data_dir = "/data/dog_images/"
dog_files_train = os.path.join(data_dir, "train/")
dog_files_test = os.path.join(data_dir , "test/")
dog_files_valid = os.path.join(data_dir ,"valid/")

data_transform = transforms.Compose([transforms.CenterCrop(299),
                                     transforms.RandomHorizontalFlip(),transforms.RandomRotation(40),
                                     transforms.ToTensor()])


data_transform_test = transforms.Compose([transforms.Resize(299),transforms.CenterCrop(299),
                                      transforms.ToTensor()])

train_data = datasets.ImageFolder(dog_files_train, transform = data_transform)
test_data = datasets.ImageFolder(dog_files_train, transform = data_transform_test)
valid_data = datasets.ImageFolder(dog_files_valid, transform = data_transform)

batch_size = 20

train_loader = torch.utils.data.DataLoader(train_data, batch_size = batch_size, shuffle = True)
test_loader = torch.utils.data.DataLoader(test_data,batch_size = batch_size, shuffle = True)
valid_loader = torch.utils.data.DataLoader(valid_data, batch_size = batch_size, shuffle = True)
2

What kind of loss function are you using?
This line of code looks wrong:

loss = criterion(output[0],target)

since output[0] would return a tensor of [133], which would not be broadcastable with the target in [20].

4

I’m using Inception_v3 model which returns a tuple (output) because of which I’m using output[0]. I’ve also edited the question as the training loop works fine and it’s the validation loop with throws the error.

5

Found the answer, inn pytorch using Inception_v3 model return a tuple of logits and auxilary logits (aux_logits) while training but returns only logits while evaluating.

We can use aux_logits = False while loading the model.