Recently years have witnessed a rapid development of large language models (LLMs). Despite the strong ability in many language-understanding tasks, the heavy computational burden largely restricts the application of LLMs especially when one needs to deploy them onto edge devices. In this paper, we propose a quantization-aware low-rank adaptation (QA-LoRA) algorithm. The motivation lies in the imbalanced degrees of freedom of quantization and adaptation, and the solution is to use group-wise operators which increase the degree of freedom of quantization meanwhile decreasing that of adaptation. QA-LoRA is easily implemented with a few lines of code, and it equips the original LoRA with two-fold abilities: (i) during fine-tuning, the LLM’s weights are quantized (e.g., into INT4) to reduce time and memory usage; (ii) after fine-tuning, the LLM and auxiliary weights are naturally integrated into a quantized model without loss of accuracy. We apply QA-LoRA to the LLaMA and LLaMA2 model families and validate its effectiveness in different fine-tuning datasets and downstream scenarios. Code will be made available at this https URL.
deleted by creator
Thank you very much for your explanation. I can understand that one. This is exactly the important difference. In my words it’d be: They figured out a way to improve on the maths, making the calculations faster. (by reducing an important matrix multiplication in dimensionality)
But there is another important aspect to it. They keep the quanzized property after the fine-tuning which QLoRA doesn’t. Which makes it a bit more precise than doing another (lossy) quantization after the fact.
Your explanation got me on track to figure it out. Thanks. I wrote another longer reply to noneabove1182. I’m not going to repeat everything, but I think I’m satisfied now.