pytorch_translate/beam_decode.py [481:509]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
                torch.sort(
                    logprobs[:, self.eos],
                    descending=True,
                    out=(eos_scores, eos_bbsz_idx),
                )
                num_remaining_sent -= finalize_hypos(step, eos_bbsz_idx, eos_scores)
                assert num_remaining_sent == 0
                break

            # cand_bbsz_idx contains beam indices for the top candidate
            # hypotheses, with a range of values: [0, bsz*beam_size),
            # and dimensions: [bsz, cand_size]
            cand_bbsz_idx = cand_beams.add_(bbsz_offsets)

            # finalize hypotheses that end in eos
            eos_mask = cand_indices.eq(self.eos)
            if step >= self.minlen:
                # only consider eos when it's among the top beam_size indices
                torch.masked_select(
                    cand_bbsz_idx[:, :beam_size],
                    mask=eos_mask[:, :beam_size],
                    out=eos_bbsz_idx,
                )
                if eos_bbsz_idx.numel() > 0:
                    torch.masked_select(
                        cand_scores[:, :beam_size],
                        mask=eos_mask[:, :beam_size],
                        out=eos_scores,
                    )
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



pytorch_translate/research/multisource/multisource_decode.py [404:432]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
                torch.sort(
                    logprobs[:, self.eos],
                    descending=True,
                    out=(eos_scores, eos_bbsz_idx),
                )
                num_remaining_sent -= finalize_hypos(step, eos_bbsz_idx, eos_scores)
                assert num_remaining_sent == 0
                break

            # cand_bbsz_idx contains beam indices for the top candidate
            # hypotheses, with a range of values: [0, bsz*beam_size),
            # and dimensions: [bsz, cand_size]
            cand_bbsz_idx = cand_beams.add_(bbsz_offsets)

            # finalize hypotheses that end in eos
            eos_mask = cand_indices.eq(self.eos)
            if step >= self.minlen:
                # only consider eos when it's among the top beam_size indices
                torch.masked_select(
                    cand_bbsz_idx[:, :beam_size],
                    mask=eos_mask[:, :beam_size],
                    out=eos_bbsz_idx,
                )
                if eos_bbsz_idx.numel() > 0:
                    torch.masked_select(
                        cand_scores[:, :beam_size],
                        mask=eos_mask[:, :beam_size],
                        out=eos_scores,
                    )
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -