# get the top beam_size active hypotheses, which are just the hypos
        # with the smallest values in active_mask
        active_hypos, _ignore = buffer('active_hypos'), buffer('_ignore')  # [b, k]
        torch.topk(
            active_mask, k=beam_size, dim=1, largest=False,
            out=(_ignore, active_hypos)
        )

        active_bbsz_idx = buffer('active_bbsz_idx')
        torch.gather(
            cand_bbsz_idx, dim=1, index=active_hypos,
            out=active_bbsz_idx,
        )
        active_scores = torch.gather(
            cand_scores, dim=1, index=active_hypos,
            out=scores[:, step].view(bsz, beam_size),
        )

        active_bbsz_idx = active_bbsz_idx.view(-1)
        active_scores = active_scores.view(-1)

        # copy tokens and scores for active hypotheses
        torch.index_select(
            tokens[:, :step + 1], dim=0, index=active_bbsz_idx,
            out=tokens_buf[:, :step + 1],
        )
        torch.gather(
            cand_indices, dim=1, index=active_hypos,
            out=tokens_buf.view(bsz, beam_size, -1)[:, :, step + 1],
        )
        if step > 0:
            torch.index_select(
                scores[:, :step], dim=0, index=active_bbsz_idx,
                out=scores_buf[:, :step],
            )
        torch.gather(
            cand_scores, dim=1, index=active_hypos,
            out=scores_buf.view(bsz, beam_size, -1)[:, :, step],`