DiffFit: Unlocking Transferability of Large Diffusion Models via Simple Parameter-Efficient Fine-Tuning
About
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$\times$512 benchmark by fine-tuning only 25 epochs from a public pre-trained ImageNet 256$\times$256 checkpoint while being 30$\times$ more training efficient than the closest competitor.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Image Generation | Oxford Flowers | FID14.02 | 15 | |
| Image Generation | CUB200 | FID5.48 | 10 | |
| Dataset Distillation | ImageWoof surrogate (train) | MMD5.3 | 10 | |
| Image Generation | Food | FID6.96 | 8 | |
| Image Generation | Caltech | FID33.84 | 8 | |
| Image Generation | ArtBench | FID20.87 | 8 | |
| Image Generation | SUN | FID8.55 | 8 | |
| Image Generation | DF-20M | FID17.35 | 8 | |
| Image Generation | Standard Cars | FID9.9 | 8 | |
| Class-conditioned generation | Food101 | FID7.8 | 7 |