Adaptive Layer-skipping in Pre-trained LLMs