Does HuggingFace have a model, and Colab tutorial, for how to train a BERT model for extractive text summarization (not abstractive), such as with something like BertSUM? So, I would provide a new dataset with a text summary and some sentences within that summary as labels, and that BERT model would be trained to learn from that dataset that those labels are the the important sentences.
It would repeat the code from this paper more or less built within the HuggingFace framework, GitHub - nlpyang/BertSum: Code for paper Fine-tune BERT for Extractive Summarization.