I have a translation model (TM), which synthesizes its hypotheses using beam-search. For analysis purposes, I would like to study all hypotheses in each beam emitted by the
