RoSTE: An Efficient Quantization-Aware Supervised Fine-Tuning Approach for Large Language Models
This work addresses the efficiency-performance trade-off in deploying fine-tuned LLMs, offering a practical solution for resource-constrained applications, though it is incremental as it builds on existing quantization and fine-tuning techniques.
The paper tackles the problem of performance degradation when quantizing fine-tuned large language models by proposing RoSTE, a quantization-aware supervised fine-tuning approach that combines adaptive rotation with straight-through estimation, achieving superior performance across various models and tasks compared to conventional post-training quantization methods.
Supervised fine-tuning is a standard method for adapting pre-trained large language models (LLMs) to downstream tasks. Quantization has been recently studied as a post-training technique for efficient LLM deployment. To obtain quantized fine-tuned LLMs, conventional pipelines would first fine-tune the pre-trained models, followed by post-training quantization. This often yields suboptimal performance as it fails to leverage the synergy between fine-tuning and quantization. To effectively realize low-bit quantization of weights, activations and KV caches in LLMs, we propose an algorithm named Rotated Straight-Through-Estimator (RoSTE), which combines quantization-aware supervised fine-tuning (QA-SFT) with an adaptive rotation strategy that identifies an effective rotation configuration to reduce activation outliers. We provide theoretical insights on RoSTE by analyzing its prediction error when applied to an overparameterized least square quantized training problem. Our findings reveal that the prediction error is directly proportional to the quantization error of the converged weights, which can be effectively managed through an optimized rotation configuration. Experiments on Pythia, Qwen and Llama models of different sizes demonstrate the effectiveness of RoSTE. Compared to existing post-SFT quantization baselines, our method consistently achieves superior performances across various tasks and different LLM architectures. Our code is available at https://github.com/OptimAI-Lab/RoSTE.