mp.tasks.retrieval.UniversalEmbedderOptions

Options for the universal embedder task.

mp.tasks.retrieval.UniversalEmbedderOptions(
    base_options: mp.tasks.BaseOptions,
    l2_normalize: bool = True,
    max_input_length: Optional[int] = None,
    vision_tokens_per_image: Optional[int] = None,
    activation_data_type: Optional[mp.tasks.retrieval.ActivationDataType] = None,
    cache_dir: Optional[str] = None,
    text_delegate: Optional[mp.tasks.BaseOptions.Delegate] = None,
    vision_delegate: Optional[mp.tasks.BaseOptions.Delegate] = None,
    audio_delegate: Optional[mp.tasks.BaseOptions.Delegate] = None
)

base_options Dataclass field
l2_normalize Dataclass field
max_input_length Dataclass field
vision_tokens_per_image Dataclass field
activation_data_type Dataclass field
cache_dir Dataclass field
text_delegate Dataclass field
vision_delegate Dataclass field
audio_delegate Dataclass field

Methods

to_ctypes

View source

to_ctypes() -> _MpUniversalEmbedderOptionsC

Generates a ctypes MpUniversalEmbedderOptionsC object.

__eq__

__eq__(
    other
)

Return self==value.

activation_data_type None
audio_delegate None
cache_dir None
l2_normalize True
max_input_length None
text_delegate None
vision_delegate None
vision_tokens_per_image None