Why the cuda() function not take effect?

autograd
1

the code is

        inputs.cuda()
        targets.cuda()
        print(inputs.device)
        print(targets.device)

but the output is :
cpu
cpu

3

The .cuda() operation is not in place so you need to store the returned GPU tensor. Like so:

inputs = inputs.cuda()
targets = targets.cuda()

Notice that this operation is in place for models: e.g., model.cuda() will work as expected, you don’t have to do model = model.cuda()