Function main

Contest/ExampleSimAnalysis/TestSetEval.py:194–226

Main function of program for predicting similarity testset samples Parameters: - args -- Parsed command line arguments as object returned by ArgumentParser

(args)

            _i += 1
    
def main(args):
    """
    Main function of program for predicting similarity testset samples

    Parameters:
    - args  -- Parsed command line arguments
               as object returned by ArgumentParser
    """
    if not os.path.exists(args.source_code):
        sys.exit(f"Directory {args.source_code} with source code is not found")
    if not os.path.exists(args.test):
        sys.exit(f"File {args.test} with test pairs is not found")
    if not os.path.exists(args.tokenizer):
        sys.exit(f"Tokenizer {args.tokenizer} is not found")
    if not os.path.exists(args.dnn):
        sys.exit(f"Check point with dnn model {args.dnn} is not found")
    ds = makeDataset(args.source_code, args.test, args.tokenizer)
    labels = loadLabels(args.labels)
    #Load trained DNN from TF checkpoint
    dnn = tf.keras.models.load_model(args.dnn)
    if labels is not None:
        if ds[0].shape[0] == labels.shape[0]:
            #Evaluate DNN accuracy on the testset
            loss, acc = dnn.evaluate(ds, labels, verbose = args.progress)
            print("\nEvaluation accuracy is {:5.2f}%".format(acc * 100))
            print("Evaluation loss is {:5.2f}".format(loss))
        else:
            print(f"Numers of labels {labels.shape[0]} " +
                f"and samples {ds[0].shape[0]} is different ")
            print("Accuracy of DNN on this test cannot be evaluated")
    #Compute probabilities of similarity predicted by DNN
    prob = dnn.predict(ds, verbose = args.progress)
    writePredictions(args.test, prob, args.predictions)
##############################################################################
# Program arguments are described below
##############################################################################