DogFit: Domain-guided Fine-tuning for Efficient Transfer Learning of Diffusion Models

TL;DR

DogFit introduces a domain-guided fine-tuning method for diffusion models that improves transfer learning efficiency and controllability without extra computational costs, outperforming prior guidance techniques.

Contribution

The paper proposes DogFit, a novel guidance mechanism that internalizes domain-aware guidance during fine-tuning, enabling efficient and controllable transfer learning of diffusion models.

Findings

DogFit outperforms prior guidance methods in FID and FDDINOV2 metrics.

It requires up to 2x fewer sampling TFLOPS.

It improves generation quality and training stability across diverse domains.

Abstract

Transfer learning of diffusion models to smaller target domains is challenging, as naively fine-tuning the model often results in poor generalization. Test-time guidance methods help mitigate this by offering controllable improvements in image fidelity through a trade-off with sample diversity. However, this benefit comes at a high computational cost, typically requiring dual forward passes during sampling. We propose the Domain-guided Fine-tuning (DogFit) method, an effective guidance mechanism for diffusion transfer learning that maintains controllability without incurring additional computational overhead. DogFit injects a domain-aware guidance offset into the training loss, effectively internalizing the guided behavior during the fine-tuning process. The domain-aware design is motivated by our observation that during fine-tuning, the unconditional source model offers a stronger marginal estimate than the target model. To support efficient controllable fidelity-diversity trade-offs at inference, we encode the guidance strength value as an additional model input through a lightweight conditioning mechanism. We further investigate the optimal placement and timing of the guidance offset during training and propose two simple scheduling strategies, i.e., late-start and cut-off, which improve generation quality and training stability. Experiments on DiT and SiT backbones across six diverse target domains show that DogFit can outperform prior guidance methods in transfer learning in terms of FID and FDDINOV2 while requiring up to 2x fewer sampling TFLOPS.