Method file_instructions

tensorflow_datasets/core/splits.py:202–235

Returns the list of dict(filename, take, skip). This allows for creating your own `tf.data.Dataset` using the low-level TFDS values. Example: ``` file_instructions = info.splits['train[75%:]'].file_instructions instruction_ds = tf.data.Dataset.from_generator( lambd

(self)

  @property
  def file_instructions(self) -> list[shard_utils.FileInstruction]:
    """Returns the list of dict(filename, take, skip).

    This allows for creating your own `tf.data.Dataset` using the low-level
    TFDS values.

    Example:

    ```
    file_instructions = info.splits['train[75%:]'].file_instructions
    instruction_ds = tf.data.Dataset.from_generator(
        lambda: file_instructions,
        output_types={
            'filename': tf.string,
            'take': tf.int64,
            'skip': tf.int64,
        },
    )
    ds = instruction_ds.interleave(
        lambda f: tf.data.TFRecordDataset(
            f['filename']).skip(f['skip']).take(f['take'])
    )
    ```

    When `skip=0` and `take=-1`, the full shard will be read, so the `ds.skip`
    and `ds.take` could be skipped.

    Returns:
      A `dict(filename, take, skip)`
    """
    return _make_file_instructions(
        split_infos=[self],
        instruction=str(self.name),
    )

  @property
  def filenames(self) -> list[str]: