Fast Sampling of Score-Based Models With Cyclical Diffusion Sampling

Diffusion models have recently exhibited significant potential in generative modeling, surpassing generative adversarial networks concerning perceptual quality and autoregressive models in density estimation. However, a notable drawback of these models is their slow sampling time, requiring numerous model evaluations to generate high-quality samples. This research proposes a technique termed cyclical diffusion sampling that incorporates cyclical stochastic gradient Langevin dynamics (SGLD) with elucidated diffusion models (EDM) sampler to enhance stability when utilizing a limited number of sampling steps. Cyclical step-size scheduling has been demonstrated to enhance the effectiveness of SGLD in learning complex multimodal distributions. We have shown, via empirical assessments, that cyclical diffusion sampling significantly enhances image quality and markedly decreases inference time. Notably, the method preserves simplicity and requires no alterations to the network architecture, which promotes straightforward reproducibility and seamless integration with current methodologies.