RandLoRA: Full rank parameter-efficient fine-tuning of large models

Published: 22 Jan 2025, Last Modified: 11 Feb 2025ICLR 2025 PosterEveryoneRevisionsBibTeXCC BY 4.0
Keywords: parameter-efficient; finetuning; CLIP; vision-language; full rank;
TL;DR: We propose a full rank alternative to LoRA to finetune large models
Abstract: Low-Rank Adaptation (LoRA) and its variants have shown impressive results in reducing the number of trainable parameters and memory requirements of large transformer networks while maintaining fine-tuning performance. However, the low-rank nature of the weight update inherently limits the representation power of the fine-tuned model, potentially compromising performance on complex tasks. This raises a critical question: when a performance gap between LoRA and standard fine-tuning is observed, is it due to the reduced number of trainable parameters or the rank deficiency? This paper aims to answer this question by introducing RandLoRA, a parameter-efficient method that performs full-rank updates using a learned linear combinations of low-rank, non-trainable random matrices. Our method limits the number of trainable parameters by restricting optimization to diagonal scaling matrices applied to the fixed random matrices. This allows us to effectively overcome low-rank limitations while maintaining low parameter count and memory usage during training. Through extensive experimentation across vision, language, and vision-language benchmarks, we systematically evaluate the limitations of LoRA and existing random basis methods. Our findings reveal that full-rank updates are beneficial across vision and language tasks separately, but especially so for vision-language tasks, where RandLoRA significantly reduces---and sometimes eliminates---the performance gap between standard fine-tuning and LoRA, demonstrating its efficacy.
Supplementary Material: zip
Primary Area: applications to computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5665
Loading