dataset.py

import datasets

class TTIDataset(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")

    def _info(self):
        return datasets.DatasetInfo(
            description="Images generated using different TTI (Text-to-Image) models: FLUX, DreamShaper, Juggernaut XL, and Pony Diffusion V6 XL.",
            features=datasets.Features({
                "id": datasets.Value("int64"),
                "createdAt": datasets.Value("string"),
                "url": datasets.Value("string"),
                "positivePrompt": datasets.Value("string"),
                "negativePrompt": datasets.Value("string"),
                "nsfw": datasets.Value("bool"),
                "browsingLevel": datasets.Value("int32"),
                "statsSummary": datasets.Value("int32"),
                "usernameHash": datasets.Value("string"),
                "Model": datasets.Value("string"),
                "cfgScale": datasets.Value("float32"),
                "sampler": datasets.Value("string"),
                "Size": datasets.Value("string"),
                "seed": datasets.Value("float64"),
                "VAE": datasets.Value("string"),
                "generationSystem": datasets.Value("string"),
                "resourceIDs": datasets.Value("string"),  # stored as stringified list
            }),
            supervised_keys=None,
            homepage="",
            citation="",
        )

    def _split_generators(self, dl_manager):
        data_files = dl_manager.download_and_extract({
            "flux": "flux_images.csv",
            "dreamshaper": "dreamshaper_images.csv",
            "juggernaut": "juggernaut_images.csv",
            "pony": "pony_diffusion_images.csv"
        })

        return [
            datasets.SplitGenerator(name="flux", gen_kwargs={"filepath": data_files["flux"]}),
            datasets.SplitGenerator(name="dreamshaper", gen_kwargs={"filepath": data_files["dreamshaper"]}),
            datasets.SplitGenerator(name="juggernaut", gen_kwargs={"filepath": data_files["juggernaut"]}),
            datasets.SplitGenerator(name="pony", gen_kwargs={"filepath": data_files["pony"]}),
        ]

    def _generate_examples(self, filepath):
        import csv

        with open(filepath, encoding="utf-8") as f:
            reader = csv.DictReader(f)
            for idx, row in enumerate(reader):
                yield idx, row