Han Liang

New research accelerates diffusion model inference and improves sampling techniques

Researchers are developing new methods to accelerate the inference process for diffusion models, which are computationally intensive for image generation. ChebBooster, a training-free framework, uses Chebyshev polynomia…