Finetuning an LLM-Based Spam Classifier with LoRA from Scratch (github.com/rasbt)