Accelerated Diffusion Models via Speculative Sampling
This work accelerates diffusion model generation, which is beneficial for applications requiring high-quality image or data synthesis, though it is incremental as it adapts an existing technique to a new domain.
The paper tackled the problem of slow inference in diffusion models by extending speculative sampling from discrete sequences to continuous vector-valued Markov chains, achieving a halving of function evaluations while generating exact samples from the target model.
Speculative sampling is a popular technique for accelerating inference in Large Language Models by generating candidate tokens using a fast draft model and accepting or rejecting them based on the target model's distribution. While speculative sampling was previously limited to discrete sequences, we extend it to diffusion models, which generate samples via continuous, vector-valued Markov chains. In this context, the target model is a high-quality but computationally expensive diffusion model. We propose various drafting strategies, including a simple and effective approach that does not require training a draft model and is applicable out of the box to any diffusion model. Our experiments demonstrate significant generation speedup on various diffusion models, halving the number of function evaluations, while generating exact samples from the target model.