How to train on those datasets that have multi-characters


For text classification, if there are more than one sentence in the text, how to deal with these data and provide them to the Transformer?