Module: tfds.core
API to define datasets.
Classes
class BeamBasedBuilder: Beam based Builder.
class BeamMetadataDict: A tfds.core.Metadata object supporting Beam-generated datasets.
class BuilderConfig: Base class for DatasetBuilder data configuration.
class DatasetBuilder: Abstract base class for all datasets.
class DatasetCollectionLoader: Loader class for dataset collections.
class DatasetIdentity: Identity of a dataset that completely identifies a dataset.
class DatasetInfo: Information about a dataset.
class DatasetNotFoundError: Exception raised when the dataset cannot be found.
class Experiment: Experiments which can be enabled/disabled on a per version basis.
class FileFormat: Format of the record files.
class GeneratorBasedBuilder: Base class for datasets with data generation based on file adapter.
class Metadata: Abstract base class for DatasetInfo metadata container.
class MetadataDict: A tfds.core.Metadata object that acts as a dict.
class Path: Abstract base class for pathlib.Path-like API.
class ReadInstruction: Reading instruction for a dataset.
class SequentialWriter: Class to write a TFDS dataset sequentially.
class ShardedFileTemplate: Template to produce filenames for sharded datasets.
class SplitDict: Split info object.
class SplitGenerator: Defines the split information for the generator.
class SplitInfo: Wraps proto.SplitInfo with an additional property.
class Version: Dataset version MAJOR.MINOR.PATCH.
class lazy_imports: Lazy importer for heavy dependencies.
Functions
add_data_dir(...): Registers a new default data_dir to search for datasets.
as_path(...): DEPRECATED. Please use from etils import epath with epath.Path().
gcs_path(...): Returns the GCS URI path.
tfds_path(...): Path to tensorflow_datasets/ root dir.