Model Tuning or Prompt Tuning? A Study of Large Language Models for Clinical Concept and Relation Extraction