LoRA-SP: streamlined partial parameter adaptation for resource efficient fine-tuning of large language models | Synapse