import datasets import json RAW_METADATA_URL = r'https://huggingface.co/datasets/OleehyO/latex-formulas/resolve/main/raw_formulas.jsonl' FILTERED_METADATA_URL = r'https://huggingface.co/datasets/OleehyO/latex-formulas/resolve/main/formulas_finally.jsonl' FILTERED_IMG_URL = r'https://huggingface.co/datasets/OleehyO/latex-formulas/resolve/main/formulas_pic.tar.gz' class LatexFormulasConfig(datasets.BuilderConfig): def __init__(self, img_archive_url, metadata_url, **kwargs): super().__init__(**kwargs) self.img_archive_url = img_archive_url self.metadata_url = metadata_url class LatexFormulas(datasets.GeneratorBasedBuilder): BUILDER_CONFIGS = [ LatexFormulasConfig( name="raw_formulas", img_archive_url=None, metadata_url=RAW_METADATA_URL ), LatexFormulasConfig( name="filtered_formulas", img_archive_url=FILTERED_IMG_URL, metadata_url=FILTERED_METADATA_URL ) ] def _info(self): if self.config.name == "raw_formulas": return datasets.DatasetInfo( features=datasets.Features({ "latex_formula": datasets.Value("string") }) ) if self.config.name == "filtered_formulas": return datasets.DatasetInfo( features=datasets.Features({ "image": datasets.Image(), "latex_formula": datasets.Value("string") }) ) def _split_generators(self, dl_manager: datasets.DownloadManager): metadata_path = dl_manager.download(self.config.metadata_url) if self.config.name == "filtered_formulas": image_archive_path = self.config.img_archive_url images = dl_manager.iter_archive(image_archive_path) elif self.config.name == "raw_formulas": images = None return [ datasets.SplitGenerator( name=datasets.Split.TRAIN, gen_kwargs={ "images": images, "metadata_path": metadata_path } ) ] def _generate_examples(self, images, metadata_path): if images is not None: img_formula_pair = {} with open(metadata_path, 'r', encoding="utf-8") as f: for line in f: single_json = json.loads(line) img_formula_pair[single_json["id"]] = single_json["formula"] for img_path, img_obj in images: img_name = img_path.split('/')[-1] if img_name in img_formula_pair: yield img_path, { "image": {"path": img_path, "bytes": img_obj.read()}, "latex_formula": img_formula_pair[img_name] } else: with open(metadata_path, 'r', encoding="utf-8") as f: for idx, line in enumerate(f): yield idx, { "latex_formula": json.loads(line)["formula"] }