LoRA: Low-Rank Adaptation of Large Language Models

TLDR

Large‑scale pre‑training followed by task‑specific adaptation is common, but as models grow, full fine‑tuning becomes infeasible and deploying separate 175B‑parameter fine‑tuned copies is prohibitively expensive. LoRA aims to enable efficient adaptation of large language models by freezing pre‑trained weights and inserting trainable low‑rank matrices into each Transformer layer. The method injects rank‑decomposition matrices per layer, drastically reducing trainable parameters, and the authors provide a PyTorch package with implementations and checkpoints for RoBERTa, DeBERTa, and GPT‑2. LoRA cuts trainable parameters by 10,000× and GPU memory by 3×, while matching or surpassing fine‑tuning performance on RoBERTa, DeBERTa, GPT‑2, and GPT‑3 with higher throughput and no extra inference latency.

Abstract

An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA.