in eval.py RuntimeError: legacy constructor expects device type: cpu but device type: cuda was passed

[kwonseungchan]

channge the code in utils.py 235 line

original_ind = torch.LongTensor(range(true_class_boxes.size(0)), device='cuda')[true_class_images == this_image][ind]
==>
original_ind = torch.tensor(range(true_class_boxes.size(0)), dtype=torch.long, device='cuda')[true_class_images == this_image][ind]

[Chadster766]

This change is required for eval.py to not error.

I would be good to have in Utils.py > calculate_mAP():
for c in tqdm(range(1, n_classes), 'Calculating mAP'):

To have a progress bar for this since it takes a while and I initially though it was a hung process.