Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Randomized Asymmetric Chain of LoRA: The First Meaningful Theoretical Framework for Low-Rank Adaptation

About

Fine-tuning has become a popular approach to adapting large foundational models to specific tasks. As the size of models and datasets grows, parameter-efficient fine-tuning techniques are increasingly important. One of the most widely used methods is Low-Rank Adaptation (LoRA), with adaptation update expressed as the product of two low-rank matrices. While LoRA was shown to possess strong performance in fine-tuning, it often under-performs when compared to full-parameter fine-tuning (FPFT). Although many variants of LoRA have been extensively studied empirically, their theoretical optimization analysis is heavily under-explored. The starting point of our work is a demonstration that LoRA and its two extensions, Asymmetric LoRA and Chain of LoRA, indeed encounter convergence issues. To address these issues, we propose Randomized Asymmetric Chain of LoRA (RAC-LoRA) -- a general optimization framework that rigorously analyzes the convergence rates of LoRA-based methods. Our approach inherits the empirical benefits of LoRA-style heuristics, but introduces several small but important algorithmic modifications which turn it into a provably convergent method. Our framework serves as a bridge between FPFT and low-rank adaptation. We provide provable guarantees of convergence to the same solution as FPFT, along with the rate of convergence. Additionally, we present a convergence analysis for smooth, non-convex loss functions, covering gradient descent, stochastic gradient descent, and federated learning settings. Our theoretical findings are supported by experimental results.

Grigory Malinovsky, Umberto Michieli, Hasan Abed Al Kader Hammoud, Taha Ceritli, Hayder Elesedy, Mete Ozay, Peter Richt\'arik• 2024

Related benchmarks

TaskDatasetResultRank
Image ClassificationOffice-Home (test)
Mean Accuracy89.33
402
Image ClassificationOfficeHome--
161
Paraphrase DetectionMRPC
Empirical Error G(W)0.37
90
Relational ReasoningCLUTRR
Empirical Generalization Error G(W)1.01
60
Linguistic AcceptabilityCoLA (test)
Avg Accuracy68.1
56
Linguistic AcceptabilityCOLA
Generalization Error Approximation (G(W))0.25
40
Image ClassificationOfficeHome
Generalization Error (G(W))0.1
40
Paraphrase IdentificationPAWS
G(W) Error Approximation1.27
40
Question ClassificationTREC50
Generalization Error G(W)2.24
30
Image ClassificationCIFAR-10 (test)
Test Accuracy99
24
Showing 10 of 25 rows

Other info

Follow for update