Evaluating Fine-Tuned LLMs with WeightWatcher Part II: PEFT / LoRa Models