English

DiffFit: Unlocking Transferability of Large Diffusion Models via Simple Parameter-Efficient Fine-Tuning

Computer Vision and Pattern Recognition 2023-07-28 v6

Abstract

Diffusion models have proven to be highly effective in generating high-quality images. However, adapting large pre-trained diffusion models to new domains remains an open challenge, which is critical for real-world applications. This paper proposes DiffFit, a parameter-efficient strategy to fine-tune large pre-trained diffusion models that enable fast adaptation to new domains. DiffFit is embarrassingly simple that only fine-tunes the bias term and newly-added scaling factors in specific layers, yet resulting in significant training speed-up and reduced model storage costs. Compared with full fine-tuning, DiffFit achieves 2×\times training speed-up and only needs to store approximately 0.12\% of the total model parameters. Intuitive theoretical analysis has been provided to justify the efficacy of scaling factors on fast adaptation. On 8 downstream datasets, DiffFit achieves superior or competitive performances compared to the full fine-tuning while being more efficient. Remarkably, we show that DiffFit can adapt a pre-trained low-resolution generative model to a high-resolution one by adding minimal cost. Among diffusion-based methods, DiffFit sets a new state-of-the-art FID of 3.02 on ImageNet 512×\times512 benchmark by fine-tuning only 25 epochs from a public pre-trained ImageNet 256×\times256 checkpoint while being 30×\times more training efficient than the closest competitor.

Keywords

Cite

@article{arxiv.2304.06648,
  title  = {DiffFit: Unlocking Transferability of Large Diffusion Models via Simple Parameter-Efficient Fine-Tuning},
  author = {Enze Xie and Lewei Yao and Han Shi and Zhili Liu and Daquan Zhou and Zhaoqiang Liu and Jiawei Li and Zhenguo Li},
  journal= {arXiv preprint arXiv:2304.06648},
  year   = {2023}
}

Comments

Tech Report

R2 v1 2026-06-28T10:05:02.014Z