Which scripts used to fine-tune t5 model after final alignment .jsonl files generated? about booksum

[jigsaw2212]

Hi @GenTxt. If you look closely at the experiment setup in section 4.2 of the paper, we follow a generate and rank approach, where we generate summaries for the paragraphs at the book and chapter level, and then rank them to get the top-k summary sentences. The value of k is chosen based on summary length statistics in the training set.