Skip to content

guidellm.data.tokenizers.tokenizer

TokenizerRegistry

Bases: RegistryMixin[type[DataTokenizer]]

Source code in src/guidellm/data/tokenizers/tokenizer.py
class TokenizerRegistry(RegistryMixin[type[DataTokenizer]]):
    @classmethod
    def create(cls, config: DataTokenizerArgs) -> DataTokenizer:
        """
        Factory method to create a DatasetTokenizer instance based on configuration.

        :param config: A DataTokenizerArgs object containing the configuration.
        """
        kind = config.kind
        tokenizer_cls = cls.get_registered_object(kind)

        if tokenizer_cls is None:
            raise ValueError(
                f"DataTokenizer type '{kind}' is not registered."
                f"Available types: {list(cls.registry.keys()) if cls.registry else []}"
            )

        return tokenizer_cls(config)

create(config) classmethod

Factory method to create a DatasetTokenizer instance based on configuration.

Parameters:

Name Type Description Default
config DataTokenizerArgs

A DataTokenizerArgs object containing the configuration.

required
Source code in src/guidellm/data/tokenizers/tokenizer.py
@classmethod
def create(cls, config: DataTokenizerArgs) -> DataTokenizer:
    """
    Factory method to create a DatasetTokenizer instance based on configuration.

    :param config: A DataTokenizerArgs object containing the configuration.
    """
    kind = config.kind
    tokenizer_cls = cls.get_registered_object(kind)

    if tokenizer_cls is None:
        raise ValueError(
            f"DataTokenizer type '{kind}' is not registered."
            f"Available types: {list(cls.registry.keys()) if cls.registry else []}"
        )

    return tokenizer_cls(config)