I have a question: When I use topk=1, how does it differ from using greedy sampling? Or are they essentially the same?