Exploring LoRA – Part 1: The Idea Behind Parameter Efficient Fine-Tuning