Source code for aac_metrics.classes.bleu

#!/usr/bin/env python
# -*- coding: utf-8 -*-

from typing import Callable, Union, get_args

from torch import Tensor

from aac_metrics.classes.base import AACMetric
from aac_metrics.functional.bleu import (
    BleuOption,
    BLEUOuts,
    _bleu_compute,
    _bleu_update,
)




class BLEU(AACMetric[Union[BLEUOuts, Tensor]]):
    """BiLingual Evaluation Understudy metric class.

    - Paper: https://www.aclweb.org/anthology/P02-1040.pdf

    For more information, see :func:`~aac_metrics.functional.bleu.bleu`.
    """

    full_state_update = False
    higher_is_better = True
    is_differentiable = False

    min_value = 0.0
    max_value = 1.0

    def __init__(
        self,
        return_all_scores: bool = True,
        *,
        n: int = 4,
        option: BleuOption = "closest",
        verbose: int = 0,
        tokenizer: Callable[[str], list[str]] = str.split,
    ) -> None:
        if option not in get_args(BleuOption):
            msg = f"Invalid option {option=}. (expected one of {get_args(BleuOption)})"
            raise ValueError(msg)

        super().__init__()
        self._return_all_scores = return_all_scores
        self._n = n
        self._option: BleuOption = option
        self._verbose = verbose
        self._tokenizer = tokenizer

        self._cooked_cands = []
        self._cooked_mrefs = []



    def compute(self) -> Union[BLEUOuts, Tensor]:
        return _bleu_compute(
            cooked_cands=self._cooked_cands,
            cooked_mrefs=self._cooked_mrefs,
            return_all_scores=self._return_all_scores,
            n=self._n,
            option=self._option,
            verbose=self._verbose,
            return_1_to_n=False,
        )




    def extra_repr(self) -> str:
        hparams = {"n": self._n}
        repr_ = ", ".join(f"{k}={v}" for k, v in hparams.items())
        return repr_




    def get_output_names(self) -> tuple[str, ...]:
        return (f"bleu_{self._n}",)




    def reset(self) -> None:
        self._cooked_cands = []
        self._cooked_mrefs = []
        return super().reset()




    def update(
        self,
        candidates: list[str],
        mult_references: list[list[str]],
    ) -> None:
        self._cooked_cands, self._cooked_mrefs = _bleu_update(
            candidates=candidates,
            mult_references=mult_references,
            n=self._n,
            tokenizer=self._tokenizer,
            prev_cooked_cands=self._cooked_cands,
            prev_cooked_mrefs=self._cooked_mrefs,
        )






class BLEU1(BLEU):
    def __init__(
        self,
        return_all_scores: bool = True,
        option: BleuOption = "closest",
        verbose: int = 0,
        tokenizer: Callable[[str], list[str]] = str.split,
    ) -> None:
        super().__init__(
            return_all_scores=return_all_scores,
            n=1,
            option=option,
            verbose=verbose,
            tokenizer=tokenizer,
        )





class BLEU2(BLEU):
    def __init__(
        self,
        return_all_scores: bool = True,
        option: BleuOption = "closest",
        verbose: int = 0,
        tokenizer: Callable[[str], list[str]] = str.split,
    ) -> None:
        super().__init__(
            return_all_scores=return_all_scores,
            n=2,
            option=option,
            verbose=verbose,
            tokenizer=tokenizer,
        )





class BLEU3(BLEU):
    def __init__(
        self,
        return_all_scores: bool = True,
        option: BleuOption = "closest",
        verbose: int = 0,
        tokenizer: Callable[[str], list[str]] = str.split,
    ) -> None:
        super().__init__(
            return_all_scores=return_all_scores,
            n=3,
            option=option,
            verbose=verbose,
            tokenizer=tokenizer,
        )





class BLEU4(BLEU):
    def __init__(
        self,
        return_all_scores: bool = True,
        option: BleuOption = "closest",
        verbose: int = 0,
        tokenizer: Callable[[str], list[str]] = str.split,
    ) -> None:
        super().__init__(
            return_all_scores=return_all_scores,
            n=4,
            option=option,
            verbose=verbose,
            tokenizer=tokenizer,
        )