def do_solver_configuration()

in amzn-smt-prediction/scripts/generate_predictions.py [0:0]

• 14 lines of code
• 9 McCabe index (conditional complexity)

def do_solver_configuration(theory, endpoint_fmf, on_features):

    # Check that the config file includes a reference to a SageMaker endpoint for the Solver Configuration Model
    if endpoint_fmf == None:
        missing_endpoint_error(theory, "endpoint_fmf")
        exit(1)

    # Already generated the features for this model (the online features from the Runtime Prediction Model above)
    # So we can skip right to calling the predictor

    ##### GENERATE INFERENCE #####

    # Create a Predictor object which references the endpoint in AWS SageMaker
    predictor = Predictor(endpoint_fmf, serializer=CSVSerializer())
    # Call the endpoint to get a prediction for our example
    prediction = get_prediction(predictor, on_features) # This model only uses the online features

    confidence = round(prediction * 100, 2)

    if v == 'Full' or v == 'Pretty':
        if confidence <= 50:
            print(start_green + "But, the Solver Configuration Model predicted -- with " + str(100 - confidence) + "% confidence -- that CVC4 will find a solution faster with --strings-fmf DISABLED." + end_green)
        else:
            print(start_green + "Additionally, the Solver Configuration Model predicted -- with " + str(confidence) + "% confidence -- that CVC4 will find a solution fastest with --strings-fmf ENABLED." + end_green)
    elif v == 'Vector':
        print(prediction)