Parameter-Efficient Fine-Tuning of LLaMA for the Clinical Domain