| import datasets | |
| import tarfile | |
| from datasets import Features, Value | |
| from datasets.download.download_manager import DownloadManager | |
| from typing import List, NamedTuple, TypedDict, Generator | |
| from PIL import Image | |
| Example = TypedDict('Example', { | |
| 'index': int, | |
| 'tar': str, | |
| 'tar_path': str, | |
| 'img': Image.Image, | |
| }) | |
| class KeyedExample(NamedTuple): | |
| key: int | |
| example: Example | |
| tar_count = 2 | |
| files = [f'./{ix:05d}.tar' for ix in range(tar_count)] | |
| class MyWebdataset(datasets.GeneratorBasedBuilder): | |
| VERSION = '1.0.0' | |
| def _info(self) -> datasets.DatasetInfo: | |
| print(__file__) | |
| return datasets.DatasetInfo( | |
| description="My webdataset", | |
| features=Features({ | |
| 'index': Value('uint32'), | |
| 'tar': Value('string'), | |
| 'tar_path': Value('string'), | |
| 'img': datasets.Image(), | |
| }), | |
| ) | |
| def _split_generators(self, dl_manager: DownloadManager) -> List[datasets.SplitGenerator]: | |
| return [ | |
| datasets.SplitGenerator( | |
| name=datasets.Split.TRAIN, | |
| gen_kwargs={"filepaths": dl_manager.download(files)}, | |
| ) | |
| ] | |
| def _generate_examples(self, filepaths: List[str]) -> Generator[KeyedExample, None, None]: | |
| index = 0 | |
| for filepath in filepaths: | |
| with tarfile.open(filepath, "r:") as tar: | |
| for member in tar.getmembers(): | |
| with tar.extractfile(member.name) as f: | |
| pil: Image.Image = Image.open(f) | |
| yield KeyedExample(index, Example( | |
| index=index, | |
| tar=filepath, | |
| tar_path=member.path, | |
| img=pil, | |
| )) | |
| index += 1 |