Changes in version 2.18.0.9000                     

                 Changes in version 2.18.0 (2025-08-20)                 

  - New dataset_rebatch().
  - dataset_batch() gains args num_parallel_calls, name.
  - dataset_interleave() gains args deterministic, num_parallel_calls,
    name.
  - text_line_dataset() gains args num_parallel_calls, buffer_size,
    name.
  - Updated documentation to fix cross links (#96).

                 Changes in version 2.17.0 (2024-07-16)                 

  - Updates for TensorFlow v2.17.0, Keras 3.

                 Changes in version 2.9.0 (2022-06-29)                  

  - New dataset_unbatch()
  - New dataset_group_by_window()
  - New dataset_take_while()
  - New as_tensor() and as.array() methods which can be used on TF
    Datasets with a single element.

                 Changes in version 2.7.0 (2021-11-09)                  

  - Added compatability with Tensorflow version 2.7
  - as_iterator(), iter_next() and iterate() are is now reexported from
    {reticualte}.
  - New as_array_iterator(), for converting a dataset into an iterable
    that yields R arrays. (as_iterator() yields tensorflow tensors)
  - New dataset_bucket_by_sequence_length()
  - New dataset_rejection_resample()
  - New dataset_unique()
  - New choose_from_datasets()
  - sample_from_datasets() gains argument stop_on_empty_dataset.
  - dataset_batch() gains arguments num_parallel_calls and
    deterministic.
  - dataset_padded_batch(): Fixed error raised when drop_remainder=TRUE
    with recent TF versions. Added examples, docs, and tests.
  - dataset_concatenate() gains ... and the ability to combine multiple
    datasets in one call.

                 Changes in version 2.6.0 (2021-09-07)                  

  - New dataset_options() for setting and getting dataset options.
  - New length() method for tensorflow datasets.
  - New dataset_enumerate().
  - New random_integer_dataset().
  - New dataset_scan(), a stateful variant of dataset_map().
  - New dataset_snapshot() for persisting the output of a dataset to
    disk.
  - range_dataset() gains a dtype argument.
  - dataset_prefetch() argument buffer_size is now optional, defaults to
    tf$data$AUTOTUNE

                 Changes in version 2.4.0 (2021-04-03)                  

  - Fixed problem when saving models with feature specs (#82).

                 Changes in version 1.13.1 (2019-04-09)                 

  - Add datatset_window method.
  - Allow purrr style lambda functions in dataset_map.
  - Added a NEWS.md file to track changes to the package.
  - Added a new feature spec interface that can be used to easily create
    feature_columns. (#42)