| """Deterministic 5-fold split over the 200 training pairs.""" | |
| from __future__ import annotations | |
| import numpy as np | |
| from .config import N_FOLDS, N_TRAIN | |
| def fold_of(idx: int) -> int: | |
| return idx % N_FOLDS | |
| def split(fold: int) -> tuple[list[int], list[int]]: | |
| """Return (train_indices, val_indices) for a fold.""" | |
| rng = np.random.RandomState(0) | |
| order = rng.permutation(N_TRAIN) | |
| val = sorted(int(i) for k, i in enumerate(order) if k % N_FOLDS == fold) | |
| train = sorted(int(i) for i in range(N_TRAIN) if i not in set(val)) | |
| return train, val | |