Converting tuple to tensor

1

How can I convert tuple to tensor? I need it for the following code because the code is in the batch of size 4. Using append, did not work properly so I decided to use cat. for torch.cat I need tuple of tensors but im_path is a tuple itself and I want to convert it to tensor so I could pass it to torch.cat.

Basically at the end I want to get a flattened list of im_paths (this worked for class_probs --not showing the entire code)

###im_paths = []
im_paths = torch.Tensor()
class_probs = torch.Tensor()
with torch.no_grad():
    for i, (inputs, classes, im_path) in enumerate(dataloaders['test']):
        
        im_paths = im_paths.cuda()
        class_probs = class_probs.cuda()
        class_probs = torch.cat((class_probs, F.softmax(outputs, 1)))
        ###im_paths.append(im_path)
        im_paths = torch.cat((im_paths, im_path))

here is an example of one of im_path

('10folds/10fold_9/test/2/10316.jpg', '10folds/10fold_9/test/7/10392.jpg', '10folds/10fold_9/test/2/90007.jpg', '10folds/10fold_9/test/2/10214.jpg')
which is a tuple

The following code:

im_paths = torch.Tensor()
with torch.no_grad():
    for i, (inputs, classes, im_path) in enumerate(dataloaders['test']):
        im_paths = im_paths.cuda()
        im_paths = torch.cat((im_paths, im_path), )

Gives me the following error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-56-efb416e1f73a> in <module>()
     18     for i, (inputs, classes, im_path) in enumerate(dataloaders['test']):
     19         im_paths = im_paths.cuda()
---> 20         im_paths = torch.cat((im_paths, im_path), )
     21 
     22         inputs = inputs.to(device)

TypeError: expected Tensor as element 1 in argument 0, but got tuple
2

I decided to use an alternative method shown below:

###im_paths = torch.Tensor()
im_paths = []
with torch.no_grad():
    for i, (inputs, classes, im_path) in enumerate(dataloaders['test']):
       
        ###im_paths = torch.cat((im_paths, im_path), )
        im_paths.append(im_path)

flattened_im_paths = flattened = [item for sublist in im_paths for item in sublist]

now flattened_im_paths does the deed.

Last line is from https://stackoverflow.com/a/51291027/2414957