aligned.exposed_model.mlflow

module documentation
(source)

Undocumented

Class InMemMLFlowAlias Undocumented
Class MlflowConfig Undocumented
Class MLFlowServer Describes a model exposed through a mlflow server.
Function in_memory_mlflow A model that is loaded from MLFlow using the given model name and alias.
Function mlflow_server Calls an MLFlow server to get the model predictions.
Function mlflow_spec Undocumented
Function reference_metadata_for_features Undocumented
Function reference_metadata_for_input Undocumented
Function references_from_metadata Decodes the feature references in stored in a metadata key.
Function signature_for_contract Undocumented
Function signature_for_features Undocumented
Variable logger Undocumented
def in_memory_mlflow(model_name: str | ConfigValue, model_alias: str | ConfigValue = 'champion', reference_tag: str = 'feature_refs', mlflow_config: MlflowConfig | None = None) -> ExposedModel: (source)

A model that is loaded from MLFlow using the given model name and alias.

This will also run in memory, and not require that mlflow is installed.

def mlflow_server(host: str, model_alias: str = 'champion', model_name: str | None = None, timeout: int = 30, mlflow_config: MlflowConfig | None = None) -> ExposedModel: (source)

Calls an MLFlow server to get the model predictions.

This will load the features described in the model signature, and pass them to the server.

def mlflow_spec(feature: Feature, is_multiple_columns: bool) -> ColSpec | TensorSpec | list[ColSpec] | str: (source)

Undocumented

def reference_metadata_for_features(features: list[FeatureReference]) -> list[str]: (source)

Undocumented

def reference_metadata_for_input(requests: list[RetrievalRequest] | RetrievalRequest) -> list[str]: (source)

Undocumented

def references_from_metadata(metadata: dict[str, Any], reference_key: str = 'feature_refs') -> list[FeatureReference] | None: (source)

Decodes the feature references in stored in a metadata key.

def signature_for_contract(model_contract: ModelFeatureStore) -> ModelSignature: (source)

Undocumented

def signature_for_features(inputs: Iterable[Feature], outputs: Iterable[Feature]) -> ModelSignature: (source)

Undocumented

logger = (source)

Undocumented