Added

-   Support for community datasets on GCS.
-   [API] `tfds.builder_from_directory` and `tfds.builder_from_directories`, see
 https://www.tensorflow.org/datasets/external_tfrecord#directly_from_folder.
-   [API] Dash ("-") support in split names.
-   [API] `file_format` argument to `download_and_prepare` method, allowing user
 to specify an alternative file format to store prepared data (e.g.
 "riegeli").
-   [API] `file_format` to `DatasetInfo` string representation.
-   [API] Expose the return value of Beam pipelines. This allows for users to
 read the Beam metrics.
-   [API] Expose Feature `tf_example_spec` to public.
-   [API] `doc` kwarg on `Feature`s, to describe a feature.
-   [Documentation] Features description is shown on
 [TFDS Catalog](https://www.tensorflow.org/datasets/catalog/overview).
-   [Documentation] More metadata about HuggingFace datasets in TFDS catalog.
-   [Performance] Parallel load of metadata files.
-   [Testing] TFDS tests are now run using GitHub actions - misc improvements
 such as caching and sharding.
-   [Testing] Improvements to MockFs.
-   New datasets.

Changed

-   [API] `num_shards` is now optional in the shard name.

Removed

-   TFDS pathlib API, migrated to a self-contained `etils.epath` (see
 https://github.com/google/etils).

Fixed

-   Various datasets.
-   Dataset builders that are defined adhoc (e.g. in Colab).
-   Better `DatasetNotFoundError` messages.
-   Don't set `deterministic` on a global level but locally in interleave, so it
 only apply to interleave and not all transformations.
-   Google drive downloader.

Added

-   [API] `split=tfds.split_for_jax_process('train')` (alias of
 `tfds.even_splits('train', n=jax.process_count())[jax.process_index()]`).
-   [Documentation] update.

Fixed

-   Import bug on Windows (3709).

Added

-   [API] Better split API:
 -   Splits can be selected using shards: `split='train[3shard]'`.
 -   Underscore supported in numbers for better readability:
     `split='train[:500_000]'`.
 -   Select the union of all splits with `split='all'`.
 -   [`tfds.even_splits`](https://www.tensorflow.org/datasets/splits#tfdseven_splits_multi-host_training)
     is more precise and flexible:
 -   Return splits exactly of the same size when passed
     `tfds.even_splits('train', n=3, drop_remainder=True)`.
 -   Works on subsplits `tfds.even_splits('train[:75%]', n=3)` or even
     nested.
 -   Can be composed with other splits: `tfds.even_splits('train', n=3)[0] +
     'test'`.
-   [API] `serialize_example` / `deserialize_example` methods on features to
 encode/decode example to proto: `example_bytes =
 features.serialize_example(example_data)`.
-   [API] `Audio` feature now supports `encoding='zlib'` for better compression.
-   [API] Features specs are exposed in proto for better compatibility with
 other languages.
-   [API] Create beam pipeline using TFDS as input with
 [tfds.beam.ReadFromTFDS](https://www.tensorflow.org/datasets/api_docs/python/tfds/beam/ReadFromTFDS).
-   [API] Support setting the file formats in `tfds build
 --file_format=tfrecord`.
-   [API] Typing annotations exposed in `tfds.typing`.
-   [API] `tfds.ReadConfig` has a new `assert_cardinality=False` argument to
 disable cardinality.
-   [API] `tfds.display_progress_bar(True)` for functional control.
-   [API] DatasetInfo exposes `.release_notes`.
-   Support for huge number of shards (>99999).
-   [Performance] Faster dataset generation (using tfrecords).
-   [Testing] Mock dataset now supports nested datasets
-   [Testing] Customize the number of sub examples
-   [Documentation] Community datasets:
 https://www.tensorflow.org/datasets/community_catalog/overview.
-   [Documentation]
 [Guide on TFDS and determinism](https://www.tensorflow.org/datasets/determinism).
-   [[RLDS](https://github.com/google-research/rlds)] Support for nested
 datasets features.
-   [[RLDS](https://github.com/google-research/rlds)] New datasets: Robomimic,
 D4RL Ant Maze, RLU Real World RL, and RLU Atari with ordered episodes.
-   New datasets.

Deprecated

-   Python 3.6 support: this is the last version of TFDS supporting Python 3.6.
 Future versions will use Python 3.7.

Fixed

-   Misc bugs.

Added

-   [API]
 [`PartialDecoding` support](https://www.tensorflow.org/datasets/decode#only_decode_a_sub-set_of_the_features),
 to decode only a subset of the features (for performances).
-   [API] `tfds.features.LabeledImage` for semantic segmentation (like image but
 with additional `info.features['image_label'].name` label metadata).
-   [API] float32 support for `tfds.features.Image` (e.g. for depth map).
-   [API] Loading datasets from files now supports custom
 `tfds.features.FeatureConnector`.
-   [API] All FeatureConnector can now have a `None` dimension anywhere
 (previously restricted to the first position).
-   [API] `tfds.features.Tensor()` can have arbitrary number of dynamic
 dimension (`Tensor(..., shape=(None, None, 3, None)`)).
-   [API] `tfds.features.Tensor` can now be serialised as bytes, instead of
 float/int values (to allow better compression): `Tensor(...,
 encoding='zlib')`.
-   [API] Support for datasets with `None` in `tfds.as_numpy`.
-   Script to add TFDS metadata files to existing TF-record (see
 [doc](https://www.tensorflow.org/datasets/external_tfrecord)).
-   [TESTING] `tfds.testing.mock_data` now supports:
 -   non-scalar tensors with dtype `tf.string`;
 -   `builder_from_files` and path-based community datasets.
-   [Documentation] Catalog now exposes links to
 [KnowYourData visualisations](https://knowyourdata-tfds.withgoogle.com/).
-   [Documentation] Guide on
 [common implementation gotchas](https://www.tensorflow.org/datasets/common_gotchas).
-   Many new reinforcement learning datasets.  Changed
-   [API] Dataset generated with `disable_shuffling=True` are now read in
 generation order.

Fixed

-   File format automatically restored (for datasets generated with
 `tfds.builder(..., file_format=)`).
-   Dynamically set number of worker threads during extraction.
-   Update progress bar during download even if downloads are cached.
-   Misc bug fixes.

Added

-   [API] `dataset.info.splits['train'].num_shards` to expose the number of
 shards to the user.
-   [API] `tfds.features.Dataset` to have a field containing sub-datasets (e.g.
 used in RL datasets).
-   [API] dtype and `tf.uint16` support in `tfds.features.Video`.
-   [API] `DatasetInfo.license` field to add redistributing information.
-   [API] `.copy`, `.format` methods to GPath objects.
-   [Performances] `tfds.benchmark(ds)` (compatible with any iterator, not just
 `tf.data`, better colab representation).
-   [Performances] Faster `tfds.as_numpy()` (avoid extra `tf.Tensor` <>
 `np.array` copy).
-   [Testing] Support for custom `BuilderConfig` in `DatasetBuilderTest`.
-   [Testing] `DatasetBuilderTest` now has a `dummy_data` class property which
 can be used in `setUpClass`.
-   [Testing] `add_tfds_id` and cardinality support to `tfds.testing.mock_data`.
-   [Documentation] Better `tfds.as_dataframe` visualisation (Sequence, ragged
 tensor, semantic masks with `use_colormap`).
-   [Experimental] Community datasets support. To allow dynamically import
 datasets defined outside the TFDS repository.
-   [Experimental] Hugging-face compatibility wrapper to use Hugging-face
 datasets directly in TFDS.
-   [Experimental] Riegeli format support.
-   [Experimental] `DatasetInfo.disable_shuffling` to force examples to be read
 in generation order.
-   New datasets.

Fixed

-   Many bugs.

Added

-   [CLI] `tfds build` to the CLI. See
 [documentation](https://www.tensorflow.org/datasets/cli#tfds_build_download_and_prepare_a_dataset).
-   [API] `tfds.features.Dataset` to represent nested datasets.
-   [API] `tfds.ReadConfig(add_tfds_id=True)` to add a unique id to the example
 `ex['tfds_id']` (e.g. `b'train.tfrecord-00012-of-01024__123'`).
-   [API] `num_parallel_calls` option to `tfds.ReadConfig` to overwrite to
 default `AUTOTUNE` option.
-   [API] `tfds.ImageFolder` support for `tfds.decode.SkipDecoder`.
-   [API] Multichannel audio support to `tfds.features.Audio`.
-   [API] `try_gcs` to `tfds.builder(..., try_gcs=True)`
-   Better `tfds.as_dataframe` visualization (ffmpeg video if installed,
 bounding boxes,...).
-   [TESTING] Allow `max_examples_per_splits=0` in `tfds build
 --max_examples_per_splits=0` to test `_split_generators` only (without
 `_generate_examples`).
-   New datasets.

Changed

-   [API] DownloadManager now returns
 [Pathlib-like](https://docs.python.org/3/library/pathlib.html#basic-use)
 objects.
-   [API] Simpler `BuilderConfig` definition: class `VERSION` and
 `RELEASE_NOTES` are applied to all `BuilderConfig`. Config description is
 now optional.
-   [API] To guarantee better deterministic, new validations are performed on
 the keys when creating a dataset (to avoid filenames as keys
 (non-deterministic) and restrict key to `str`, `bytes` and `int`). New
 errors likely indicates an issue in the dataset implementation.
-   [API] `tfds.core.benchmark` now returns a `pd.DataFrame` (instead of a
 `dict`).
-   [API] `tfds.units` is not visible anymore from the public API.
-   Datasets updates.

Deprecated

Removed

-   Configs for all text datasets. Only plain text version is kept. For example:
 `multi_nli/plain_text` -> `multi_nli`.

Fixed

-   [API] Datasets returned by `tfds.as_numpy` are compatible with `len(ds)`.
-   Support 0-len sequence with images of dynamic shape (Fix 2616).
-   Progression bar correctly updated when copying files.
-   Better debugging and error message (e.g. human readable size,...).
-   Many bug fixes (GPath consistency with pathlib, s3 compatibility, TQDM
 visual artifacts, GCS crash on windows, re-download when checksums updated,
 ...).

Added

-   It is now easier to create datasets outside TFDS repository (see our updated
 [dataset creation guide](https://www.tensorflow.org/datasets/add_dataset)).
-   When generating a dataset, if download fails for any reason, it is now
 possible to manually download the data. See
 [doc](https://www.tensorflow.org/datasets/overview#manual_download_if_download_fails).
-   `tfds.core.as_path` to create pathlib.Path-like objects compatible with GCS
 (e.g. `tfds.core.as_path('gs://my-bucket/labels.csv').read_text()`).
-   `verify_ssl=` option to `tfds.download.DownloadConfig` to disable SSH
 certificate during download.
-   New datasets.  Changed
-   All dataset inherit from `tfds.core.GeneratorBasedBuilder`. Converting a
 dataset to beam now only require changing `_generate_examples` (see
 [example and doc](https://www.tensorflow.org/datasets/beam_datasets#instructions)).
-   `_split_generators` should now returns `{'split_name':
 self._generate_examples(), ...}` (but current datasets are backward
 compatible).
-   Better `pathlib.Path`, `os.PathLike` compatibility: `dl_manager.manual_dir`
 now returns a pathlib-Like object. Example: `python text =
 (dl_manager.manual_dir / 'downloaded-text.txt').read_text()` Note: Other
 `dl_manager.download`, `.extract`,... will return pathlib-like objects in
 future versions. `FeatureConnector`,... and most functions should accept
 `PathLike` objects. Let us know if some functions you need are missing.
-   `--record_checksums` now assume the new dataset-as-folder model.

Deprecated

-   `tfds.core.SplitGenerator`, `tfds.core.BeamBasedBuilder` are deprecated and
 will be removed in a future version.

Fixed

-   `BuilderConfig` are now compatible with Beam datasets 2348
-   `tfds.features.Images` can accept encoded `bytes` images directly (useful
 when used with `img_name, img_bytes =
 dl_manager.iter_archive('images.zip')`).
-   Doc API now show deprecated methods, abstract methods to overwrite are now
 documented.
-   You can generate `imagenet2012` with only a single split (e.g. only the
 validation data). Other split will be skipped if not present.

Fixed

-   `tfds.load` when generation code isn't present.
-   GCS compatibility.

Added

-   Dataset-as-folder: Dataset can now be self-contained module in a folder with
 checksums, dummy data,... This simplify implementing datasets outside the
 TFDS repository.
-   `tfds.load` can now load dataset without using the generation class. So
 `tfds.load('my_dataset:1.0.0')` can work even if `MyDataset.VERSION ==
 '2.0.0'` (See 2493).
-   TFDS CLI (see https://www.tensorflow.org/datasets/cli for detail).
-   `tfds.testing.mock_data` does not require metadata files anymore!
-   `tfds.as_dataframe(ds, ds_info)` with custom visualisation
 ([example](https://www.tensorflow.org/datasets/overview#tfdsas_dataframe)).
-   `tfds.even_splits` to generate subsplits (e.g. `tfds.even_splits('train',
 n=3) == ['train[0%:33%]', 'train[33%:67%]', ...]`.
-   `DatasetBuilder.RELEASE_NOTES` property.
-   `tfds.features.Image` now supports PNG with 4-channels.
-   `tfds.ImageFolder` now supports custom shape, dtype.
-   Downloaded URLs are available through `MyDataset.url_infos`.
-   `skip_prefetch` option to `tfds.ReadConfig`.
-   `as_supervised=True` support for `tfds.show_examples`, `tfds.as_dataframe`.
-   tfds.features can now be saved/loaded, you may have to overwrite
 [FeatureConnector.from_json_content](https://www.tensorflow.org/datasets/api_docs/python/tfds/features/FeatureConnector?version=nightly#from_json_content)
 and `FeatureConnector.to_json_content` to support this feature.
-   Script to detect dead-urls.
-   New datasets.

Changed

-   `tfds.as_numpy()` now returns an iterable which can be iterated multiple
 times. To migrate: `next(ds)` -> `next(iter(ds))`.
-   Rename `tfds.features.text.Xyz` -> `tfds.deprecated.text.Xyz`.

Removed

-   `DatasetBuilder.IN_DEVELOPMENT` property.
-   `tfds.core.disallow_positional_args` (should use Py3 `*,` instead).
-   Testing against TF 1.15. Requires Python 3.6.8+.

Fixed

-   Better archive extension detection for `dl_manager.download_and_extract`.
-   Fix `tfds.__version__` in TFDS nightly to be PEP440 compliant
-   Fix crash when GCS not available.
-   Improved open-source workflow, contributor guide, documentation.
-   Many other internal cleanups, bugs, dead code removal, py2->py3 cleanup,
 pytype annotations,...
-   Datasets updates.

Fixed

-   Issue with GCS on Windows.

Added

-   [API] `tfds.ImageFolder` and `tfds.TranslateFolder` to easily create custom
 datasets with your custom data.
-   [API] `tfds.ReadConfig(input_context=)` to shard dataset, for better
 multi-worker compatibility (1426).
-   [API] The default `data_dir` can be controlled by the `TFDS_DATA_DIR`
 environment variable.
-   [API] Better usability when developing datasets outside TFDS: downloads are
 always cached, checksums are optional.
-   Scripts to help deployment/documentation (Generate catalog documentation,
 export all metadata files, ...).
-   [Documentation] Catalog display images
 ([example](https://www.tensorflow.org/datasets/catalog/sun397#sun397standard-part2-120k)).
-   [Documentation] Catalog shows which dataset have been recently added and are
 only available in `tfds-nightly`
 <span class="material-icons">nights_stay</span>.
-   [API] `tfds.show_statistics(ds_info)` to display
 [FACETS OVERVIEW](https://pair-code.github.io/facets/). Note: This require
 the dataset to have been generated with the statistics.

Deprecated

-   `tfds.features.text` encoding API. Please use
 [tensorflow_text](https://www.tensorflow.org/tutorials/tensorflow_text/intro)
 instead.

Removed

-   `tfds.load('image_label_folder')` in favor of the more user-friendly
 `tfds.ImageFolder`.

Fixed

-   Fix deterministic example order on Windows when path was used as key (this
 only impacts a few datasets). Now example order should be the same on all
 platforms.
-   Misc performances improvements for both generation and reading (e.g. use
 `__slot__`, fix parallelisation bug in `tf.data.TFRecordReader`, ...).
-   Misc fixes (typo, types annotations, better error messages, fixing dead
 links, better windows compatibility, ...).

Added

-   [API] `tfds.builder_cls(name)` to access a DatasetBuilder class by name
-   [API] `info.split['train'].filenames` for access to the tf-record files.
-   [API] `tfds.core.add_data_dir` to register an additional data dir.
-   [Testing] Support for custom decoders in `tfds.testing.mock_data`.
-   [Documentation] Shows which datasets are only present in `tfds-nightly`.
-   [Documentation] Display images for supported datasets.

Changed

-   Rename `tfds.core.NamedSplit`, `tfds.core.SplitBase` -> `tfds.Split`. Now
 `tfds.Split.TRAIN`,... are instance of `tfds.Split`.
-   Rename `interleave_parallel_reads` -> `interleave_cycle_length` for
 `tfds.ReadConfig`.
-   Invert ds, ds_info argument orders for `tfds.show_examples`.

Deprecated

-   `tfds.features.text` encoding API. Please use `tensorflow_text` instead.

Removed

-   `num_shards` argument from `tfds.core.SplitGenerator`. This argument was
 ignored as shards are automatically computed.
-   Most `ds.with_options` which where applied by TFDS. Now use `tf.data`
 default.

Fixed

-   Better error messages.
-   Windows compatibility.

Added

-   `DownloadManager` is now pickable (can be used inside Beam pipelines).
-   `tfds.features.Audio`:
 -   Support float as returned value.
 -   Expose sample_rate through `info.features['audio'].sample_rate`.
 -   Support for encoding audio features from file objects.
-   More datasets.

Changed

-   New `image_classification` section. Some datasets have been move there from
 `images`.
-   `DownloadConfig` does not append the dataset name anymore (manual data
 should be in `<manual_dir>/` instead of `<manual_dir>/<dataset_name>/`).
-   Tests now check that all `dl_manager.download` urls has registered
 checksums. To opt-out, add `SKIP_CHECKSUMS = True` to your
 `DatasetBuilderTestCase`.
-   `tfds.load` now always returns `tf.compat.v2.Dataset`. If you're using still
 using `tf.compat.v1`:
 -   Use `tf.compat.v1.data.make_one_shot_iterator(ds)` rather than
     `ds.make_one_shot_iterator()`.
 -   Use `isinstance(ds, tf.compat.v2.Dataset)` instead of `isinstance(ds,
     tf.data.Dataset)`.

Deprecated

-   The `tfds.features.text` encoding API is deprecated. Please use
 [tensorflow_text](https://www.tensorflow.org/tutorials/tensorflow_text/intro)
 instead.
-   `num_shards` argument of `tfds.core.SplitGenerator` is currently ignored and
 will be removed in the next version.

Removed

-   Legacy mode `tfds.experiment.S3` has been removed
-   `in_memory` argument has been removed from `as_dataset`/`tfds.load` (small
 datasets are now auto-cached).
-   `tfds.Split.ALL`.

Fixed

-   Various bugs, better error messages, documentation improvements.