REVIEW 8 cited by
Efficient Diffusion Training via Min-SNR Weighting Strategy
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
abstract
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-$\gamma$. 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.
Forward citations
Cited by 8 Pith papers
-
Entropy as a Structural Prior: How a Log-Barrier on DiT Belief Space Drives Musical Diversity and Development
An entropy-based log-barrier on DiT outputs acts as an online curriculum in supervised diffusion fine-tuning, producing higher thematic development and textural diversity than standard training on MusicCaps.
-
ELT: Elastic Looped Transformers for Visual Generation
Elastic Looped Transformers share weights across recurrent blocks and apply intra-loop self-distillation to deliver 4x parameter reduction while matching competitive FID and FVD scores on ImageNet and UCF-101.
-
ELT: Elastic Looped Transformers for Visual Generation
Weight-shared looped transformers trained with intra-loop self-distillation match MaskGIT-class FID/FVD at roughly 4x fewer parameters and support any-time inference across loop counts.
-
Forecasting implied volatility surface with generative diffusion models
A conditioned diffusion model with SNR-weighted arbitrage penalty generates one-day-ahead arbitrage-free implied volatility surfaces and outperforms baselines on market data.
-
Gaussian Variation Field Diffusion for High-fidelity Video-to-4D Synthesis
A video-to-4D model that encodes mesh animations into compact Gaussian variation latents and diffuses them conditioned on the video and a canonical Gaussian splat.
-
Spatio-Temporal Conditional Diffusion Models for Forecasting Future Multiple Sclerosis Lesion Masks Conditioned on Treatments
A treatment-conditioned diffusion model generates future multiple sclerosis lesion masks from baseline MRI and better predicts lesion activity than population-level baselines.
-
Zero123++: a Single Image to Consistent Multi-view Diffusion Base Model
Zero123++ produces high-quality 3D-consistent multi-view images from a single input by fine-tuning Stable Diffusion with targeted conditioning and training methods.
-
Multimodal Image Colorization: Quantifying the Impact of Text-Conditioned Guidance on Grayscale-to-Color Translation
Text conditioning improves PSNR by ~5.7%, SSIM by ~1.4%, colorfulness by up to 36.6%, and reduces LPIPS by ~9.5% across U-Net and Stable Diffusion colorization models.
Discussion (0). Sign in to comment.