"""Build the cached feature files ONCE (optional but recommended), then commit cache/*.npz so the Space boots instantly and never depends on the dataset downloads again. pip install torch torchvision datasets opencv-python-headless pillow numpy python prepare_features.py --task heart --n 150 python prepare_features.py --task diabetes --n 150 Runs fine on a free Google Colab CPU runtime or on a laptop (~2-5 min per task). """ import argparse from fl import data ap = argparse.ArgumentParser() ap.add_argument("--task", choices=list(data.TASKS), required=True) ap.add_argument("--n", type=int, default=150, help="images per class") ap.add_argument("--seed", type=int, default=0) a = ap.parse_args() raw = data.load_raw(a.task, a.n, a.seed, allow_synthetic=False) print("OK:", a.task, raw["X"].shape, "positives:", int(raw["y"].sum()), "| groups:", raw["group_name"])