tanadata.py

import json
import datasets

# You can update these with more detailed information.
_DESCRIPTION = """
TanaData is a custom dataset for instruction-response tasks.
"""

_CITATION = """
@misc{tanadata2025,
  title={TanaData Dataset},
  year={2025},
  note={Custom dataset hosted on Hugging Face}
}
"""

class TanaData(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")
    
    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features({
                "instruction": datasets.Value("string"),
                "input": datasets.Value("string"),
                "output": datasets.Value("string"),
            }),
            supervised_keys=None,
            homepage="https://huggingface.co/mdevoz/tanadata",
            citation=_CITATION,
        )
    
    def _split_generators(self, dl_manager):
        # This URL points to your JSON file in the repository.
        file_path = dl_manager.download_and_extract(
            "https://huggingface.co/mdevoz/tanadata/resolve/main/tanadata.json"
        )
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN, 
                gen_kwargs={"filepath": file_path}
            )
        ]
    
    def _generate_examples(self, filepath):
        # Adjust this logic based on your JSON file structure.
        with open(filepath, encoding="utf-8") as f:
            # If your file is a JSON array of examples:
            data = json.load(f)
            for idx, example in enumerate(data):
                yield idx, example

# For testing, you can uncomment the following lines locally:
# if __name__ == "__main__":
#     from datasets import load_dataset
#     dataset = load_dataset(__file__, name="tanadata")
#     print(dataset)