Scale-wise Distillation of Diffusion Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914363582971904 |
|---|---|
| author | Starodubcev, Nikita Drobyshevskiy, Ilya Kuznedelev, Denis Babenko, Artem Baranchuk, Dmitry |
| author_facet | Starodubcev, Nikita Drobyshevskiy, Ilya Kuznedelev, Denis Babenko, Artem Baranchuk, Dmitry |
| contents | Recent diffusion distillation methods have achieved remarkable progress, enabling high-quality ${\sim}4$-step sampling for large-scale text-conditional image and video diffusion models. However, further reducing the number of sampling steps becomes more and more challenging, suggesting that efficiency gains may be better mined along other model axes. Motivated by this perspective, we introduce SwD, a scale-wise diffusion distillation framework that equips few-step models with progressive generation, avoiding redundant computations at intermediate diffusion timesteps. Beyond efficiency, SwD enriches the family of distribution matching distillation approaches by introducing a simple patch-level distillation objective based on Maximum Mean Discrepancy (MMD). This objective significantly improves the convergence of existing distillation methods and performs surprisingly well in isolation, offering a competitive baseline for diffusion distillation. Applied to state-of-the-art text-to-image/video diffusion models, SwD approaches the sampling speed of two full-resolution steps and largely outperforms alternatives under the same compute budget, as evidenced by automatic metrics and human preference studies. Project page: https://yandex-research.github.io/swd |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_16397 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Scale-wise Distillation of Diffusion Models Starodubcev, Nikita Drobyshevskiy, Ilya Kuznedelev, Denis Babenko, Artem Baranchuk, Dmitry Computer Vision and Pattern Recognition Recent diffusion distillation methods have achieved remarkable progress, enabling high-quality ${\sim}4$-step sampling for large-scale text-conditional image and video diffusion models. However, further reducing the number of sampling steps becomes more and more challenging, suggesting that efficiency gains may be better mined along other model axes. Motivated by this perspective, we introduce SwD, a scale-wise diffusion distillation framework that equips few-step models with progressive generation, avoiding redundant computations at intermediate diffusion timesteps. Beyond efficiency, SwD enriches the family of distribution matching distillation approaches by introducing a simple patch-level distillation objective based on Maximum Mean Discrepancy (MMD). This objective significantly improves the convergence of existing distillation methods and performs surprisingly well in isolation, offering a competitive baseline for diffusion distillation. Applied to state-of-the-art text-to-image/video diffusion models, SwD approaches the sampling speed of two full-resolution steps and largely outperforms alternatives under the same compute budget, as evidenced by automatic metrics and human preference studies. Project page: https://yandex-research.github.io/swd |
| title | Scale-wise Distillation of Diffusion Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2503.16397 |