simple question about the input of the model

I have a simple question. How the input is organized? I saw the input is
> x = {
'cfg': cfg,
'template': torch.autograd.Variable(input[0]).cuda(),
'search': torch.autograd.Variable(input[1]).cuda(),
'label_cls': torch.autograd.Variable(input[2]).cuda(),
'label_loc': torch.autograd.Variable(input[3]).cuda(),
'label_loc_weight': torch.autograd.Variable(input[4]).cuda(),
'label_mask': torch.autograd.Variable(input[6]).cuda(),
'label_mask_weight': torch.autograd.Variable(input[7]).cuda(),
}
but I didn't find related code in 'siam_mask_dataaset.py' .