BasePredict

Overview¶

The predict pipeline is a pipeline that performs common actions concerning predictions, such as making predictions from new data, versioning predictions, and analzing predictions.

Attributes¶

BasePredict contains no default attributes.

Configuration¶

BasePredict contains no default or required configuration.

Interface¶

The following methods are part of BasePredict and should be implemented in any class that inherits from this base class:

compare_data¶

Compares the prediction dataset to another dataset, such as the training dataset. The intestion here is to capture data skew from training time to prediction time.

def compare_data(self, model_version, dataset_version, data, *args, **kwargs)

Arguments:

model_version (object): The model version object that is being promoted.
dataset_version (object): The dataset version object containing to old data to compare the prediction data to.
data (object): The prediction data.

Returns:

Nothing

get_predictions¶

Runs a prediciton job to get predictions from a new dataset

def get_predictions(self, model, model_version, data, *args, **kwargs) -> tuple[Any, Any]

Arguments:

model (object): The model object to use to make predictions.
model_version (object): The model version containing the model
data (object): The prediction data

Returns:

predictions (object): The predictions
prediction_job (object): The prediction job that created the predictions, created from a metadata_tracker

track_predictions¶

Versions the results of a prediction job.

def track_predictions(self, prediction_job, data, *args, **kwargs)

Arguments:

prediction_job (object): The prediction job from the metadata_tracker
data (object): The predictions.

Returns:

Nothing

analyze_prediction_drift¶

Analyzes drift between the current predictions and the previously produced predictions.

def analyze_prediciton_drift(self, dataset_version, prediction_job, data, *args, **kwargs)

Arguments:

dataset_version (object): The dataset_version corresponding to the current predictions.
predition_job (object): The prediction job that created the predictions.
data: (object): The prediction data.

Returns:

Nothing

check_predictions¶

Runs prediction checks.

def check_predictions(self, data, prediction_job,  *args, **kwargs)

Arguments:

data (object): The predictions
predition_job (object): The prediction job that created the predictions.

Returns:

Nothing

save_predictions¶

Save predictions.

def save_predictions(self, data, target, *args, **kwargs)

Arguments:

data (object): The predictions
target (str): The place to save the predictions. Could be a local file, path in an object store, or a table in a data warehouse, etc.

Returns:

Nothing