English

FALQON: Accelerating LoRA Fine-tuning with Low-Bit Floating-Point Arithmetic

Machine Learning 2025-10-29 v1 Artificial Intelligence

Abstract

Low-bit floating-point (FP) formats, such as FP8, provide significant acceleration and memory savings in model training thanks to native hardware support on modern GPUs and NPUs. However, we analyze that FP8 quantization offers speedup primarily for large-dimensional matrix multiplications, while inherent quantization overheads diminish speedup when applied to low-rank adaptation (LoRA), which uses small-dimensional matrices for efficient fine-tuning of large language models (LLMs). To address this limitation, we propose FALQON, a novel framework that eliminates the quantization overhead from separate LoRA computational paths by directly merging LoRA adapters into an FP8-quantized backbone during fine-tuning. Furthermore, we reformulate the forward and backward computations for merged adapters to significantly reduce quantization overhead, and introduce a row-wise proxy update mechanism that efficiently integrates substantial updates into the quantized backbone. Experimental evaluations demonstrate that FALQON achieves approximately a 3×\times training speedup over existing quantized LoRA methods with a similar level of accuracy, providing a practical solution for efficient large-scale model fine-tuning. Moreover, FALQON's end-to-end FP8 workflow removes the need for post-training quantization, facilitating efficient deployment. Code is available at https://github.com/iamkanghyunchoi/falqon.

Keywords

Cite

@article{arxiv.2510.24061,
  title  = {FALQON: Accelerating LoRA Fine-tuning with Low-Bit Floating-Point Arithmetic},
  author = {Kanghyun Choi and Hyeyoon Lee and SunJong Park and Dain Kwon and Jinho Lee},
  journal= {arXiv preprint arXiv:2510.24061},
  year   = {2025}
}

Comments

NeurIPS 2025

R2 v1 2026-07-01T07:08:58.017Z