If requires_grad = False (not tracking gradients) for a tensor, is there any point in doing
some_tensor = Variable(torch.Tensor(some_array))
instead of just simply
some_tensor = torch.Tensor(some_array)
?
If requires_grad = False (not tracking gradients) for a tensor, is there any point in doing
some_tensor = Variable(torch.Tensor(some_array))
instead of just simply
some_tensor = torch.Tensor(some_array)
?
Since PyTorch0.4, Variable and Tensor have merged so they are the same thing now. You should just use torch.Tensor