stable_datasets.features package

Submodules

stable_datasets.features.array module

Array-based feature codecs.

class Array3D(shape: tuple, dtype: str = 'uint8')

Bases: FeatureType

Fixed-shape 3D array stored as flat bytes.

encode(value, *, cache_dir: Path | None = None) → bytes | None

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type() → DataType

stable_datasets.features.base module

Core feature descriptors shared across modalities.

class ClassLabel(names: list[str] | None = None, num_classes: int | None = None)

Bases: FeatureType

Categorical label with name-to-int mapping.

encode(value, *, cache_dir: Path | None = None)

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

int2str(idx: int) → str

str2int(name: str) → int

to_arrow_type() → DataType

class FeatureType

Bases: object

Base class for feature type descriptors.

arrow_metadata() → dict[bytes, bytes]

encode(value, *, cache_dir: Path | None = None)

fingerprint_data() → str

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type() → DataType

class Sequence(feature: FeatureType)

Bases: FeatureType

Variable-length list of a sub-feature.

encode(value, *, cache_dir: Path | None = None)

to_arrow_type() → DataType

class Value(dtype: str)

Bases: FeatureType

Scalar value type. Maps dtype strings to PyArrow types.

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type() → DataType

stable_datasets.features.image module

Image feature codec.

class Image(encode_format: str = 'PNG')

Bases: FeatureType

Image feature stored as raw bytes in Arrow.

encode(value, *, cache_dir: Path | None = None) → bytes | None

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type()

stable_datasets.features.video module

Video feature codec and lazy reference objects.

class Video(storage: str = 'path', allowed_extensions: tuple[str, ...] = ('.mp4', '.avi', '.mov', '.webm', '.mkv'))

Bases: FeatureType

Video feature with validated path, bytes, or specialized frame storage.

arrow_metadata() → dict[bytes, bytes]

encode(value, *, cache_dir: Path | None = None)

fingerprint_data() → str

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type() → DataType

class VideoRef(cell: Mapping[str, Any], cache_dir: Path | None = None)

Bases: object

Lazy reference to a cached video asset.

property bytes: bytes

cache_dir: Path | None = None

cell: Mapping[str, Any]

property checksum: str | None

property extension: str

property media_type: str

property mode: str

property path: Path | None

property size: int

Module contents

Feature codec modules.

class Array3D(shape: tuple, dtype: str = 'uint8')

Bases: FeatureType

Fixed-shape 3D array stored as flat bytes.

encode(value, *, cache_dir: Path | None = None) → bytes | None

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type() → DataType

class ClassLabel(names: list[str] | None = None, num_classes: int | None = None)

Bases: FeatureType

Categorical label with name-to-int mapping.

encode(value, *, cache_dir: Path | None = None)

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

int2str(idx: int) → str

str2int(name: str) → int

to_arrow_type() → DataType

class FeatureType

Bases: object

Base class for feature type descriptors.

arrow_metadata() → dict[bytes, bytes]

encode(value, *, cache_dir: Path | None = None)

fingerprint_data() → str

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type() → DataType

class Image(encode_format: str = 'PNG')

Bases: FeatureType

Image feature stored as raw bytes in Arrow.

encode(value, *, cache_dir: Path | None = None) → bytes | None

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type()

class Sequence(feature: FeatureType)

Bases: FeatureType

Variable-length list of a sub-feature.

encode(value, *, cache_dir: Path | None = None)

to_arrow_type() → DataType

class Value(dtype: str)

Bases: FeatureType

Scalar value type. Maps dtype strings to PyArrow types.

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type() → DataType

class Video(storage: str = 'path', allowed_extensions: tuple[str, ...] = ('.mp4', '.avi', '.mov', '.webm', '.mkv'))

Bases: FeatureType

Video feature with validated path, bytes, or specialized frame storage.

arrow_metadata() → dict[bytes, bytes]

encode(value, *, cache_dir: Path | None = None)

fingerprint_data() → str

format(value, *, format_type: str, decode_images: bool = True, cache_dir: Path | None = None)

to_arrow_type() → DataType

class VideoRef(cell: Mapping[str, Any], cache_dir: Path | None = None)

Bases: object

Lazy reference to a cached video asset.

property bytes: bytes

cache_dir: Path | None = None

cell: Mapping[str, Any]

property checksum: str | None

property extension: str

property media_type: str

property mode: str

property path: Path | None

property size: int