Speculative Diffusion Decoding: Accelerating Language Generation through Diffusion
This accelerates inference for large language models, addressing efficiency bottlenecks in real-time applications, though it is an incremental improvement over existing speculative decoding methods.
The paper tackles the limitation of speculative decoding in language generation by proposing Speculative Diffusion Decoding, which uses discrete diffusion models to parallelize drafting and verification, achieving up to 7.2x speedups over standard generation and 1.75x over existing speculative decoding.
Speculative decoding has emerged as a widely adopted method to accelerate large language model inference without sacrificing the quality of the model outputs. While this technique has facilitated notable speed improvements by enabling parallel sequence verification, its efficiency remains inherently limited by the reliance on incremental token generation in existing draft models. To overcome this limitation, this paper proposes an adaptation of speculative decoding which uses discrete diffusion models to generate draft sequences. This allows parallelization of both the drafting and verification steps, providing significant speedups to the inference process. Our proposed approach, $\textit{Speculative Diffusion Decoding (SpecDiff)}$, is validated on standard language generation benchmarks and empirically demonstrated to provide up to 7.2x speedups over standard generation processes and up to 1.75x speedups over existing speculative decoding approaches.