Improving Prompt Tuning with Learned Prompting Layers