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