How to use the sacrebleu.NGRAM_ORDER function in sacrebleu

with ProcessPoolExecutor(max_workers=self.n_workers) as executor:
                futures = [
                    executor.submit(
                        sb.corpus_bleu, b[0], b[1], force=True,
                        tokenize=tokenizer
                    )
                    for b in batches
                ]
                progress = as_completed(futures)
                if self.verbose:
                    progress = tqdm(progress)
                for future in progress:
                    s = future.result()
                    ref_len += s.ref_len
                    sys_len += s.sys_len
                    for n in range(sb.NGRAM_ORDER):
                        correct[n] += s.counts[n]
                        total[n] += s.totals[n]
            corpus_score = sb.compute_bleu(
                correct, total, sys_len, ref_len, smooth_method='exp'
            ).bp
        return corpus_score

def score_corpus_multiprocess(
            self, hypothesis: List[str], references: List[List[str]]
    ) -> float:
        tokenizer = get_optional_dict(self.extra_args, 'bleu_tokenizer', 'none')
        if self.n_workers == 1:
            corpus_score = sb.corpus_bleu(
                hypothesis, references, force=True, tokenize=tokenizer
            ).score
        else:
            batches = list(
                self._batch(hypothesis, references, n_batches=self.n_workers)
            )
            ref_len, sys_len = 0, 0
            correct = [0 for _ in range(sb.NGRAM_ORDER)]
            total = [0 for _ in range(sb.NGRAM_ORDER)]
            with ProcessPoolExecutor(max_workers=self.n_workers) as executor:
                futures = [
                    executor.submit(
                        sb.corpus_bleu, b[0], b[1], force=True,
                        tokenize=tokenizer
                    )
                    for b in batches
                ]
                progress = as_completed(futures)
                if self.verbose:
                    progress = tqdm(progress)
                for future in progress:
                    s = future.result()
                    ref_len += s.ref_len
                    sys_len += s.sys_len
                    for n in range(sb.NGRAM_ORDER):