LoRA: Low-Rank Adaptation of Large Language Models

TL;DR

LoRA introduces a low-rank adaptation method that significantly reduces the number of trainable parameters in large language models, enabling efficient fine-tuning without sacrificing performance.

Contribution

The paper proposes LoRA, a novel low-rank adaptation technique that drastically decreases trainable parameters and memory usage during fine-tuning of large language models.

Findings

LoRA reduces trainable parameters by 10,000 times compared to full fine-tuning.

LoRA achieves comparable or better performance than traditional fine-tuning.

LoRA has no additional inference latency and improves training throughput.

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.