Target-Driven Distillation: Consistency Distillation with Target Timestep Selection and Decoupled Guidance
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910588252192768 |
|---|---|
| author | Wang, Cunzheng Guo, Ziyuan Duan, Yuxuan Li, Huaxia Chen, Nemo Tang, Xu Hu, Yao |
| author_facet | Wang, Cunzheng Guo, Ziyuan Duan, Yuxuan Li, Huaxia Chen, Nemo Tang, Xu Hu, Yao |
| contents | Consistency distillation methods have demonstrated significant success in accelerating generative tasks of diffusion models. However, since previous consistency distillation methods use simple and straightforward strategies in selecting target timesteps, they usually struggle with blurs and detail losses in generated images. To address these limitations, we introduce Target-Driven Distillation (TDD), which (1) adopts a delicate selection strategy of target timesteps, increasing the training efficiency; (2) utilizes decoupled guidances during training, making TDD open to post-tuning on guidance scale during inference periods; (3) can be optionally equipped with non-equidistant sampling and x0 clipping, enabling a more flexible and accurate way for image sampling. Experiments verify that TDD achieves state-of-the-art performance in few-step generation, offering a better choice among consistency distillation models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_01347 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Target-Driven Distillation: Consistency Distillation with Target Timestep Selection and Decoupled Guidance Wang, Cunzheng Guo, Ziyuan Duan, Yuxuan Li, Huaxia Chen, Nemo Tang, Xu Hu, Yao Computer Vision and Pattern Recognition Consistency distillation methods have demonstrated significant success in accelerating generative tasks of diffusion models. However, since previous consistency distillation methods use simple and straightforward strategies in selecting target timesteps, they usually struggle with blurs and detail losses in generated images. To address these limitations, we introduce Target-Driven Distillation (TDD), which (1) adopts a delicate selection strategy of target timesteps, increasing the training efficiency; (2) utilizes decoupled guidances during training, making TDD open to post-tuning on guidance scale during inference periods; (3) can be optionally equipped with non-equidistant sampling and x0 clipping, enabling a more flexible and accurate way for image sampling. Experiments verify that TDD achieves state-of-the-art performance in few-step generation, offering a better choice among consistency distillation models. |
| title | Target-Driven Distillation: Consistency Distillation with Target Timestep Selection and Decoupled Guidance |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2409.01347 |