Automatic tensor type conversion

I found a problem when I use the cuda tensor as the default one.
Even when I set:
torch.set_default_tensor_type('torch.cuda.FloatTensor')
input x (numpy.float)  y (torch.cuda.tensor)
output x*y is torch.float.tensor

Is this a feature or a bug. It might be more convenient if the programme behaves as the default setting.
Many thanks