embeddings.ObjectEmbedding

vector_search.embeddings.ObjectEmbedding()

Abstract class that can be used to create embeddings for Objects of a specific format.

Methods

Name Description
dimensions Returns the number of dimensions of the embedding vectors.
embed Creates embedding vectors for objects. Returns a numpy array of embedding vectors.
init_kwargs Returns a dictionary containing kwargs that can be used to re-initialize the ObjectEmbedding.
load Loads the model in order to be ready for embedding objects.
vector_type Returns the datatype of the embedding vectors.

dimensions

vector_search.embeddings.ObjectEmbedding.dimensions()

Returns the number of dimensions of the embedding vectors.

embed

vector_search.embeddings.ObjectEmbedding.embed(objects, metadata)

Creates embedding vectors for objects. Returns a numpy array of embedding vectors. There is no enforced restriction on the object format. ObjectReaders and ObjectEmbeddings should use compatible object and metadata formats.

Parameters

Name Type Description Default
objects OrderedDict An OrderedDict, containing the object data, having structure similar to TileDB-Py read results. required
metadata OrderedDict An OrderedDict, containing the object metadata, having structure similar to TileDB-Py read results. required

init_kwargs

vector_search.embeddings.ObjectEmbedding.init_kwargs()

Returns a dictionary containing kwargs that can be used to re-initialize the ObjectEmbedding.

This is used to serialize the ObjectEmbedding and pass it as argument to UDF tasks.

load

vector_search.embeddings.ObjectEmbedding.load()

Loads the model in order to be ready for embedding objects.

This method will be called once per worker to avoid loading the model multiple times.

vector_type

vector_search.embeddings.ObjectEmbedding.vector_type()

Returns the datatype of the embedding vectors.