I can create a dataset with values with the same key but different answers

[lauraBaakman]

For example I have created a dataset with the following two items: [{question: sorry, answer: scusa, question: sorry, answer: scusi}]. Is this expected behaviour? Do you handle this correctly during training?