I see some pytorch training code using something like this:
for batch_idx, (inputs, targets) in enumerate(trainloader):
inputs, targets = torch.autograd.Variable(inputs), torch.autograd.Variable(targets)
...
pred = model(inputs)
...
why? becuase I think (inputs, targets) are already tensors? thus we can directly do like:
for batch_idx, (inputs, targets) in enumerate(trainloader):
pred = model(inputs)
...
Is it because I misunderstand something or the first usage is just some old version of pytorch?
really thanks