embedding

galactic.embedding.get_embeddings(self, input_field, embedding_field='__embedding', model='bge-micro', backend='cpu', normalize=False, pad=False, split_strategy='even')

Get embeddings for a field in the dataset. :type input_field: str :param input_field: The field to get embeddings for. :type input_field: str :type embedding_field: str :param embedding_field: The field to store the embeddings in. :type embedding_field: str, default=’__embedding’ :type model: Literal['gte-small', 'gte-tiny', 'bge-micro', 'bge-micro-v2', None] :param model: The model to use for embeddings, if computing locally. Options: ‘gte-small’, ‘gte-tiny’, ‘bge-micro’.

# Replace "field_name1"
ds.get_embeddings(field="field_name1")

Parameters:

• input_field (str) –

• embedding_field (str) –

• model (Literal['gte-small', 'gte-tiny', 'bge-micro', 'bge-micro-v2', None]) –

• backend (Literal['cpu', 'gpu', 'openai', 'replicate', 'modal', 'onnx', 'ctranslate2']) –

• normalize (bool) –

• pad (bool) –

• split_strategy (Literal['truncate', 'greedy', 'even']) –

galactic.embedding.get_nearest_neighbors(self, query, k=5, embedding_field='__embedding')

Get the nearest neighbors to a query.

Parameters:

• query (str | numpy.ndarray | list[float]) –

• k (int) –

• embedding_field (str) –

galactic.embedding.initialize_embedding_model(self, backend='cpu', model='bge-micro', max_requests_per_minute=None, max_tokens_per_minute=None)

Initializes the embedding model based on the backend.

Parameters:

• backend (str, default='auto') – The backend used for embedding. Options: ‘cpu’, ‘gpu’, ‘openai’.

• model (Literal['gte-small', 'gte-tiny', 'bge-micro', 'bge-micro-v2', None]) –

• max_requests_per_minute (int | None) –

• max_tokens_per_minute (int | None) –

Raises:

ValueError – Raises an exception if the backend is ‘openai’ but openai_api_key is not set, or if the backend is unknown.

Example:

initialize_embedding_model('cpu')