ALLoRA: Adaptive Learning Rate Mitigates LoRA Fatal Flaws
Keywords: Large Language Models, low rank adaption, finetuning, dropout
Abstract: Low-Rank Adaptation (LoRA) is the bread and butter of Large Language Model (LLM) finetuning. LoRA learns an additive low-rank perturbation of a pretrained matrix parameter to align the model to a new task or dataset. We identify three core limitations to LoRA for finetuning--with only a limited amount of training steps. First, it employs Dropout as a means to prevent overfitting. We prove that Dropout is only suitable for long training episodes but fails to reliably regularize training for short training episodes, e.g., finetuning. Second, LoRA’s parameters initialization is at $0$ makes the optimization landscape poorly conditioned during the first steps of training. That poor conditioning combined with the need to move away from $0$ lead to slow training dynamics. Third, the scaling factor that multiply each LoRA additive perturbation create ``short-sighted'' interactions between the LoRA modules of different layers. Motivated by principled analysis of those limitations, we find an elegant solution: a Dropout-free, scaling-free, LoRA with Adaptive Learning rate--coined ALLoRA. By scaling the per sample and per parameter gradients with a coefficient inversely proportional to parameters’ $\ell_2$ norm, ALLoRA alleviates those three limitations. As a by-product, ALLoRA removes two hyper-parameters from LoRA: the scaling factor and the dropout rate. Empirical results show that ALLoRA admits better accuracy than LoRA on various settings, including against recent LoRA variants such as Weight-Decomposed Low-Rank Adaptation (DoRA). Ablation studies show our solution is the optimal in a family of weight-dependent / output-dependent approaches.
Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning
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.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
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: 13143
Loading