Hi, i really enjoyed your paper on text summarization using Bert model.

I have a question in this code.

On your paper, you said that you used beam search algorithm (beam size is 5) during decoding.
But I think, beam size is fixed on 1 in this code. (fast beam search)

Wouldn't using beam size of 1 in Abstractive summarization get worse results?
Wasn't there any affect on performance?

Thanks in advance