Fine tune seq2seq with multiple output

Hey guys,

I have a dataset where i have a question and answers to that question. An answer can appear more than once. I want to do a transfer learning to a transformer (was thinking of QuestionAnswering or maybe use DialogPT) so when i put the question as input it will generate me x suggestion for reply to that question based on the frequency of that answer in the dataset.
Is it possible to that with Question Answering transformer or should i choose another approach?