A parameter-efficient fine-tuning method that combines quantization with low-rank adaptation to reduce memory usage while training large models.