Efficient Diffusion Training via Min-SNR Weighting Strategy
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2023
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910359618584576 |
|---|---|
| author | Hang, Tiankai Gu, Shuyang Li, Chen Bao, Jianmin Chen, Dong Hu, Han Geng, Xin Guo, Baining |
| author_facet | Hang, Tiankai Gu, Shuyang Li, Chen Bao, Jianmin Chen, Dong Hu, Han Geng, Xin Guo, Baining |
| contents | Denoising diffusion models have been a mainstream approach for image generation, however, training these models often suffers from slow convergence. In this paper, we discovered that the slow convergence is partly due to conflicting optimization directions between timesteps. To address this issue, we treat the diffusion training as a multi-task learning problem, and introduce a simple yet effective approach referred to as Min-SNR-$γ$. This method adapts loss weights of timesteps based on clamped signal-to-noise ratios, which effectively balances the conflicts among timesteps. Our results demonstrate a significant improvement in converging speed, 3.4$\times$ faster than previous weighting strategies. It is also more effective, achieving a new record FID score of 2.06 on the ImageNet $256\times256$ benchmark using smaller architectures than that employed in previous state-of-the-art. The code is available at https://github.com/TiankaiHang/Min-SNR-Diffusion-Training. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2303_09556 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Efficient Diffusion Training via Min-SNR Weighting Strategy Hang, Tiankai Gu, Shuyang Li, Chen Bao, Jianmin Chen, Dong Hu, Han Geng, Xin Guo, Baining Computer Vision and Pattern Recognition Denoising diffusion models have been a mainstream approach for image generation, however, training these models often suffers from slow convergence. In this paper, we discovered that the slow convergence is partly due to conflicting optimization directions between timesteps. To address this issue, we treat the diffusion training as a multi-task learning problem, and introduce a simple yet effective approach referred to as Min-SNR-$γ$. This method adapts loss weights of timesteps based on clamped signal-to-noise ratios, which effectively balances the conflicts among timesteps. Our results demonstrate a significant improvement in converging speed, 3.4$\times$ faster than previous weighting strategies. It is also more effective, achieving a new record FID score of 2.06 on the ImageNet $256\times256$ benchmark using smaller architectures than that employed in previous state-of-the-art. The code is available at https://github.com/TiankaiHang/Min-SNR-Diffusion-Training. |
| title | Efficient Diffusion Training via Min-SNR Weighting Strategy |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2303.09556 |