Function _resample

espnet2/speechlm/dataloader/iterator.py:465–519

Resample a list by a given factor. For integer factors, simply repeats the list. For non-integer factors, repeats the whole list for the integer part and randomly samples the fractional part using a deterministic seed. Args: lst: List to resample. factor: Resampling

(lst: List[T], factor: float, seed: int = 42)

def _resample(lst: List[T], factor: float, seed: int = 42) -> List[T]:
    """Resample a list by a given factor.

    For integer factors, simply repeats the list. For non-integer
    factors, repeats the whole list for the integer part and randomly
    samples the fractional part using a deterministic seed.

    Args:
        lst: List to resample.
        factor: Resampling factor. Must be positive.
            - factor=1.0: returns original list
            - factor=2.0: returns list repeated twice
            - factor=1.5: returns list once + 50% random sample
        seed: Random seed for reproducible sampling. Default: 42.

    Returns:
        Resampled list. Returns empty list if input is empty.

    Raises:
        ValueError: If factor <= 0.

    Examples:
        >>> _resample([1, 2, 3], 2.0)
        [1, 2, 3, 1, 2, 3]
        >>> _resample([1, 2, 3], 1.5, seed=42)
        [1, 2, 3, 2, 1]  # Deterministic with same seed
    """
    if factor <= 0:
        raise ValueError(f"Resampling factor must be positive, got {factor}")

    if not lst:
        return []  # Empty list resampled is still empty

    # Integer factor: simple repetition
    if factor.is_integer():
        return lst * int(factor)

    # Non-integer factor: repeat whole + sample fractional part
    result = []

    # Add full copies for the integer part
    int_part = int(factor)
    for _ in range(int_part):
        result.extend(lst)

    # Add random sample for the fractional part (deterministic)
    frac_part = factor - int_part
    num_samples = int(frac_part * len(lst))
    if num_samples > 0:
        # Use local Random instance for reproducibility
        rng = random.Random(seed)
        residual = rng.sample(lst, num_samples)
        result.extend(residual)

    return result