Skip to content

Custom dl_manager in load_dataset #7767

@ain-soph

Description

@ain-soph

Feature request

https://github.com/huggingface/datasets/blob/4.0.0/src/datasets/load.py#L1411-L1418


def load_dataset(
    ...
    dl_manager: Optional[DownloadManager] = None,  # add this new argument
    **config_kwargs,
) -> Union[DatasetDict, Dataset, IterableDatasetDict, IterableDataset]:

    ...

    # Create a dataset builder
    builder_instance = load_dataset_builder(
        path=path,
        name=name,
        data_dir=data_dir,
        data_files=data_files,
        cache_dir=cache_dir,
        features=features,
        download_config=download_config,
        download_mode=download_mode,
        revision=revision,
        token=token,
        storage_options=storage_options,
        **config_kwargs,
    )

    # Return iterable dataset in case of streaming
    if streaming:
        return builder_instance.as_streaming_dataset(split=split)

    # Note: This is the revised part
    if dl_manager is None:
        if download_config is None:
            download_config = DownloadConfig(
                cache_dir=builder_instance._cache_downloaded_dir,
                force_download=download_mode == DownloadMode.FORCE_REDOWNLOAD,
                force_extract=download_mode == DownloadMode.FORCE_REDOWNLOAD,
                use_etag=False,
                num_proc=num_proc,
                token=builder_instance.token,
                storage_options=builder_instance.storage_options,
            )  # We don't use etag for data files to speed up the process

        dl_manager = DownloadManager(
            dataset_name=builder_instance.dataset_name,
            download_config=download_config,
            data_dir=builder_instance.config.data_dir,
            record_checksums=(
                builder_instance._record_infos or verification_mode == VerificationMode.ALL_CHECKS
            ),
        )

    # Download and prepare data
    builder_instance.download_and_prepare(
        download_config=download_config,
        download_mode=download_mode,
        verification_mode=verification_mode,
        dl_manager=dl_manager,  # pass the new argument
        num_proc=num_proc,
        storage_options=storage_options,
    )

    ...

Motivation

In my case, I'm hoping to deal with the cache files downloading manually (not using hash filenames and save to another location, or using potential existing local files).

Your contribution

It's already implemented above. If maintainers think this should be considered, I'll open a PR.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions