MetaflowOfflinePredict¶

The MetaflowOfflinePredict class is a subclass of BasePredict and provides methods for running a Metaflow flow object and obtaining artifacts from the MetaflowOfflinePredict pipeline.

Attributes¶

METAFLOW_CLASS - The name of the class that inherits from metaflow's FlowSpec

Configuration¶

Required Configuration¶

MetaflowOfflinePredict requires the following component types:

data_connector
metadata_tracker
resource_version_control
model_explainer
data_checker
data_profiler

Methods¶

run¶

This method loads and runs a provided Metaflow flow with the provided inputs.

def run(self, model, model_version, data, dataset_version, **kwargs)

Arguments:

model (object): The model object to use for predictions.
model_version (object): The model version of the model object
data (object): The prediction data
dataset_version (object): The dataset version of the prediction data

Returns:

None

get_artifacts¶

def get_artifacts(self, artifact_keys)

This method gets the artifacts generated by the Metaflow flow.

Arguments:

artifact_keys (list): A list of artifact keys.

Returns:

artifacts (dict): A dictionary containing the requested artifacts.

MetaflowOfflinePredictSpec Methods¶

MetaflowOfflinePredictSpec contains the following methods. These are mirrored from the OfflinePredict class, and you should see that documentation for more information (Note: instead of these method explicitly using arguments, they instead access saved artifacts during the Metaflow run).

start
compate_data
get_predictions
track_predictions
analyze_prediction_drift
check_predicitons
save_predictions
end