Matching Pairs: Attributing Fine-Tuned Models to their Pre-Trained Large Language Models