Can FairSeq's speech-to-task model predict punctuations (e.g., sentence boundaries)?

Question

I read their paper, checked out their repo but could not find the answer. Does FairSeq's speech-to-task model predict punctuations (e.g., sentence boundaries)? I just wanted to find out before I invest time and effort for implementing it.

Answer 1

I posted the same question on their github page. The short answer is yes. Here is the long answer: https://github.com/pytorch/fairseq/issues/3029