LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention

Renrui Zhang, Jiaming Han, Aojun Zhou, Xiangfei Hu, Shilin Yan, Pan Lu, Hongsheng Li, Peng Gao, Yu Qiao

Abstract

We present LLaMA-Adapter, a lightweight adaption method to efficiently fine-tune LLaMA into an instruction-following model. Using 52K self-instruct demonstrations, LLaMA-Adapter only introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs. Specifically, we adopt a set of learnable adaption prompts, and prepend them to the word tokens at higher transformer layers. Then, a zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively preserves its pre-trained knowledge. With our efficient training, LLaMA-Adapter can generate high-quality responses, comparable to Alpaca with fully fine-tuned 7B parameters. Besides language commands, our approach can be simply extended to multi-modal instructions for learning image-conditioned LLaMA model, which achieves superior reasoning performance on ScienceQA and COCO Caption benchmarks. Furthermore, we also evaluate the zero-initialized attention mechanism for fine-tuning other pre-trained models (ViT, RoBERTa) on traditional vision and language tasks, demonstrating the superior generalization capacity of our approach. Code is released at https://github.com/OpenGVLab/LLaMA-Adapter.
BibTeX
@misc{zhang2023llama,
  author = {Renrui Zhang and Jiaming Han and Aojun Zhou and Xiangfei Hu and Shilin Yan and Pan Lu and Hongsheng Li and Peng Gao and Yu Qiao},
  title = {{LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention}},
  year = {2023},
  eprint = {2303.16199},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2303.16199},
}
Figure