def calculate_accuracy()

in use-cases/model-fine-tuning-pipeline/model-eval/src/validate_fine_tuned_model.py [0:0]

• 25 lines of code
• 2 McCabe index (conditional complexity)

    def calculate_accuracy(self):
        ground_truth = pd.DataFrame(self.training_df["Answer"])
        total_test_size = len(self.df)
        logger.info(f"Test dataset size: {total_test_size}")

        product_names = self.extract_product_names(self.output_file)

        true_positives_count, false_positives_count = self.count_tp_fp(
            product_names, ground_truth
        )
        none_predictions = self.count_no_products_prediction(product_names)
        logger.info(f"True Positives Count: {true_positives_count}")
        logger.info(f"False Positives Count: {false_positives_count}")
        logger.info(
            f"Number of predictions with no product details: {none_predictions}"
        )

        accuracy = round((true_positives_count / total_test_size) * 100, 2)
        logger.info(f"Accuracy of Gemma2 9B IT model on test dataset is {accuracy}%")

        if true_positives_count | false_positives_count:
            precision = round(
                (true_positives_count / (true_positives_count + false_positives_count))
                * 100,
                2,
            )
            logger.info(
                f"Precision of Gemma2 9B IT model on test dataset is {precision}%"
            )