Skip to main content

EmbeddingList

Addedv2.6.x

An EmbeddingList instance represents a list of vector embeddings. You can use an EmbeddingList instance to build the query vectors in a search against a vector field in an Array of Structs field.

class pymilvus.EmbeddingList

Constructor​

Constructs an empty embedding list or a list of given vector embeddings.

EmbeddingList(
    embeddings: Optional[Union[np.ndarray, List[np.ndarray]],
    dim: Optional[int],
    dtype: Optional[Union[np.dtype, str, DataType]]
)

PARAMETERS:

  • embeddings (np.ndarray, List[np.ndarray) -

    A list of vector embeddings, which can be either of the following types:

    • np.ndarray with shape (n, dim), indicating a list of multiple vector embeddings

    • np.ndarray with shape (dim,), indicating a single vector embedding

    • List[np.ndarray], indicating a list of vector embedding arrays

  • dim (int) -

    The dimensionality of the vector embeddings that are specified in the embedding parameter, for validation purposes.

    If provided, all specified vector embeddings must adhere to the dimensionality restriction.

  • dtype (np.dtype, str, DataType) -

    • np.dtype, such as np.float32, np.float16, or np.unit8

    • string, such as 'float32', 'float16', or 'uint8'

    • DataType, such as DataType.FLOAT_VECTOR, DataType.FLOAT16_VECTOR, DataType.BFLOAT16_VECTOR, DataType.INT8_VECTOR, or DataType.BINARY_VECTOR

RETURN TYPE:

EmbeddingList

RETURNS:

An EmbeddingList instance.

Examples​

from pymilvus import EmbeddingList

# create an empty embedding list
embeddingList1 = EmbeddingList()

# create an embedding list with a single vector embedding of 5 dimensions
embeddingList2 = EmbeddingList(
    embeddings=[0.1, 0.2, 0.3, 0.4, 0.5],
    dim=5
)

# create an embedding list with two vector embeddings, each having five dimensions
embeddingList3 = EmbeddingList(
    embeddings= [[0.1, 0.2, 0.3, 0.4, 0.5], [0.5, 0.4, 0.3, 0.2, 0.1]],
    dim=5
)