import datasets

features = datasets.Features(
    {
        "source": datasets.Value(dtype="string", id=None),
        "id": datasets.Value(dtype="string", id=None),
        "text": datasets.Value(dtype="string", id=None),
        "added": datasets.Value(dtype="string", id=None),
        "timestamp": datasets.Value(dtype="timestamp[s]", id=None),
        "metadata": {"url": datasets.Value(dtype="string", id=None)},
    }
)
dataset = datasets.load_dataset(
    "json",
    data_files={
        "train": "fi_processed/c4-fi.*.json",
        "validation": "fi_processed/c4-fi-validation*.json",
    },
    features=features,
    cache_dir="/researchdisk/datasets_cache",
    num_proc=96,
)

dataset = dataset.flatten()
dataset = dataset.rename_column("metadata.url", "url")
dataset = dataset.remove_columns(["source", "id", "added"])
print(dataset)
dataset.save_to_disk("/researchdisk/mc4_3.1.0_fi", num_proc=96)