split

Split a dataset into configurable named splits.

Functions

parse_split(value)

Parse a name:percentage split specification.

split(source, output, splits[, seed])

Split a dataset into named partitions.

undertale.utils.datasets.split.parse_split(value: str) Tuple[str, float]

Parse a name:percentage split specification.

Parameters:

value – A string in "name:percentage" format.

Returns:

A tuple of (name, percentage).

Raises:

argparse.ArgumentTypeError – If the value is not in the expected format.

undertale.utils.datasets.split.split(source: str, output: str, splits: List[Tuple[str, float]], seed: int = 42) None

Split a dataset into named partitions.

Parameters:
  • source – Path to the input dataset.

  • output – Base path for output directories. Each split is written to "{output}-{name}".

  • splits – A list of (name, percentage) tuples. Percentages must sum to 100.

  • seed – Random seed for reproducibility.

Raises:

ValueError – If split percentages do not sum to 100.