Sentences in Abstractive Summarization


I am trying to use the summarization pipeline to summarize some text. I use BART and Pegasus, and often end up with sentences abruptly cut in half in the summaries. Is there a way to prevent this?

Also I would like to know if there is a way to set the ‘temperature’ in generating sentences. That is, how much the model sticks to its input instead of trying to generate original sentences.

Any help is appreciated, thank you!

@demegire please view the discussion here Summarization on long documents - #30 by marcoabrate

I have not come across the concept of temperature on BART and Pegasus summarization yet, other experts perhaps have an opinion about it. If so I am also interested in it.