Sacrebleuscore

sacrebleuscore

Metric SacreBLEUScore from torchmetrics (torchmetrics.SacreBLEUScore)

When to invoke this skill

The user has predictions + ground truth and asks to evaluate with SacreBLEUScore, or mentions torchmetrics.SacreBLEUScore directly, or wants the standard torchmetrics implementation.

Reference signature

from torchmetrics import SacreBLEUScore

# _SacreBLEUScore(n_gram: int = 4, smooth: bool = False, tokenize: Literal['none', '13a', 'zh', 'intl', 'char'] = '13a', lowercase: bool = False, weights: Optional[collections.abc.Sequence[float]] = None, **kwargs: Any) -> None

Library docstring

Wrapper for deprecated import.

>>> preds = ['the cat is on the mat']
>>> target = [['there is a cat on the mat', 'a cat is on the mat']]
>>> sacre_bleu = _SacreBLEUScore()
>>> sacre_bleu(preds, target)
tensor(0.7598)

Quick recipe

import torchmetrics as _m
score = _m.SacreBLEUScore(y_true, y_pred)

Don'ts

• Don't reimplement when the library version handles edge cases (NaN, ties, empty inputs) better than a hand-rolled formula.
• Always check the library version's argument order — sklearn is (y_true, y_pred) while torchmetrics is (preds, target).