Over the past decade, Deep Learning (DL) has become an integral part of our daily lives. This surge in DL usage has heightened the need for developing reliable DL software systems. Given that fault localization is a critical task in reliability assessment, researchers have proposed several fault localization techniques for DL-based software, primarily focusing on faults within the DL model. While the DL model is central to DL components, there are other elements that significantly impact the performance of DL comp…