DDIM
An efficient sampling algorithm for diffusion models that accelerates the image generation process by skipping steps in the reverse diffusion trajectory path.
Think of It Like This
Like taking an express train that skips several local stops to reach the final destination much faster without significantly changing the route.
Denoising Diffusion Implicit Models (DDIM) modify the Markov chain process used in standard diffusion by relying on a non-Markovian forward process. This allows the model to generate high-quality samples in 10 to 50 steps rather than the hundreds or thousands required by earlier methods. As a result, inference costs are drastically reduced.