TFDS now supports the Croissant 🥐 format! Read the documentation to know more.

tfds.even_splits

View source on GitHub

Generates a list of non-overlapping sub-splits of same size.

tfds.even_splits(
 split: str, n: int, *, drop_remainder: bool = False
) -> List[tfds.typing.SplitArg ]

Example:

split0, split1, split2 = tfds.even_splits('train', n=3, drop_remainder=True)
# Load 1/3 of the train split.
ds = tfds.load('my_dataset', split=split0)

tfds.even_splits supports arbitrary sub-splits inputs, including other tfds.even_splits outputs.

Args

split Split (e.g. 'train', 'test[75%:]',...)
n Number of sub-splits to create
drop_remainder Drop examples if the number of examples in the datasets is not evenly divisible by n. If False, examples are distributed evenly across subsplits, starting by the first. For example, if there is 11 examples with n=3, splits will contain [4, 4, 3] examples respectivelly.

Returns

The list of subsplits. Those splits can be combined together (with +) or with other subsplits (e.g. split + 'test[75%:]').

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2024年04月26日 UTC.