LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models