artifact_serializer

serialize_artifact(value_node_id, execution_id, reference, name, storage_backend=None, **kwargs)

Serialize artifact using various backend.

Currently, most objects are using lineapy as the backend for serialization. The only exception is MLflow-supported model flavors. In order to use MLflow for ml model serialization, following conditions need to be satisfied:

1. The artifact(ML model) should be a MLflow-supported flavor
2. MLflow is installed
3. storage_backend should be mlflow or storage_backend is None and options.get("default_ARTIFACT_STORAGE_BACKEND")=='mlflow'

Parameters:

Name	Type	Description	Default
value_node_id	LineaID	Value node id in Linea Graph	required
execution_id	LineaID	Execution id	required
reference	Any	Same as reference in lineapy.save()	required
name	str	Same as reference in lineapy.save()	required
storage_backend	Optional[ARTIFACT_STORAGE_BACKEND]	Same as reference in lineapy.save()	None
**kwargs		Same as reference in lineapy.save()	{}

Returns:

Type	Description
Dict	Dictionary with following key-value pair: backend: storage backend used to save the artifact metadata: metadata of the storage backed

Source code in lineapy/api/artifact_serializer.py

def serialize_artifact(
    value_node_id: LineaID,
    execution_id: LineaID,
    reference: Any,
    name: str,
    storage_backend: Optional[ARTIFACT_STORAGE_BACKEND] = None,
    **kwargs,
) -> Dict[str, Any]:
    """
    Serialize artifact using various backend.

    Currently, most objects are using lineapy as the backend for serialization.
    The only exception is MLflow-supported model flavors. In order to use
    MLflow for ml model serialization, following conditions need to be
    satisfied:

    1. The artifact(ML model) should be a MLflow-supported flavor
    2. MLflow is installed
    3. `storage_backend` should be `mlflow` or `storage_backend` is `None` and
    `options.get("default_ARTIFACT_STORAGE_BACKEND")=='mlflow'`

    Parameters
    ----------
    value_node_id: LineaID
        Value node id in Linea Graph
    execution_id: LineaID
        Execution id
    reference: Union[object, ExternalState]
        Same as reference in [`lineapy.save()`][lineapy.api.api.save]
    name: str
        Same as reference in [`lineapy.save()`][lineapy.api.api.save]
    storage_backend: Optional[ARTIFACT_STORAGE_BACKEND]
        Same as reference in [`lineapy.save()`][lineapy.api.api.save]
    **kwargs:
        Same as reference in [`lineapy.save()`][lineapy.api.api.save]

    Returns
    -------
    Dict
        Dictionary with following key-value pair:

        - backend: storage backend used to save the artifact
        - metadata: metadata of the storage backed
    """
    if _able_to_use_mlflow(storage_backend):
        model_info = try_write_to_mlflow(reference, name, **kwargs)
        if model_info is not None:
            return {
                "backend": "mlflow",
                "metadata": model_info,
            }

    pickle_name = _pickle_name(value_node_id, execution_id)
    _try_write_to_pickle(reference, pickle_name)
    return {
        "backend": "lineapy",
        "metadata": {"pickle_name": str(pickle_name)},
    }

---

Was this helpful?

Help us improve docs with your feedback!

Send