Method __init__

scripts/eval/eval.py:1031–1096

Initialize the evaluation dataset. Sets up the appropriate dataset loader based on the evaluation set and task type. Handles both HuggingFace datasets and custom dataset loaders with automatic data downloading if needed. Args: task (Literal): Type of tra

(
        self,
        task: Literal["eng_transcribe", "long_form_transcribe"],
        eval_set: Literal[
            "librispeech_clean",
            "librispeech_other",
            "tedlium",
            "wsj",
            "callhome",
            "switchboard",
            "common_voice",
            "artie_bias_corpus",
            "coraal",
            "chime6",
            "ami_ihm",
            "ami_sdm",
            "voxpopuli",
            "fleurs",
            "meanwhile",
            "kincaid46",
            "rev16",
            "earnings21",
            "earnings22",
        ],
        hf_token: Optional[str] = None,
        eval_dir: str = "data/eval",
        n_mels: int = DEFAULT_N_MELS,
    )

    """

    def __init__(
        self,
        task: Literal["eng_transcribe", "long_form_transcribe"],
        eval_set: Literal[
            "librispeech_clean",
            "librispeech_other",
            "tedlium",
            "wsj",
            "callhome",
            "switchboard",
            "common_voice",
            "artie_bias_corpus",
            "coraal",
            "chime6",
            "ami_ihm",
            "ami_sdm",
            "voxpopuli",
            "fleurs",
            "meanwhile",
            "kincaid46",
            "rev16",
            "earnings21",
            "earnings22",
        ],
        hf_token: Optional[str] = None,
        eval_dir: str = "data/eval",
        n_mels: int = DEFAULT_N_MELS,
    ):
        """Initialize the evaluation dataset.

        Sets up the appropriate dataset loader based on the evaluation set and task type.
        Handles both HuggingFace datasets and custom dataset loaders with automatic
        data downloading if needed.

        Args:
            task (Literal): Type of transcription task - either "eng_transcribe" for
                short-form or "long_form_transcribe" for extended audio
            eval_set (Literal): Name of the evaluation dataset to load
            hf_token (Optional[str], optional): HuggingFace authentication token for
                private datasets. Defaults to None.
            eval_dir (str, optional): Directory for storing evaluation datasets.
                Defaults to "data/eval".
            n_mels (int, optional): Number of mel-spectrogram bins for short-form tasks.
                Defaults to DEFAULT_N_MELS.

        Raises:
            ValueError: If eval_set is not supported for the specified task
            FileNotFoundError: If required dataset files are not found and cannot be downloaded

        Example:
            >>> dataset = EvalDataset(
            ...     task="eng_transcribe",
            ...     eval_set="librispeech_clean",
            ...     eval_dir="/data/eval",
            ...     n_mels=80
            ... )
        """
        self.eval_set = eval_set