提出块验证机制，让扩散模型推理提速6.3%且无需额外训练。
Accelerating Speculative Diffusions via Block Verification
- 引入块验证技术，高效实现扩散模型的推测采样。
- 自由起草者（Free Drafter）在无训练下提升草案接受率。
- 相比现有方法，提速最高达6.3%，开销几乎为零。
推测解码通过使用草稿模型生成令牌，并采用接受-拒绝机制确保输出符合目标分布，从而加速大语言模型推理。将其适配到连续扩散模型面临挑战：推测采样需从残差分布中采样，而在连续空间中高效采样该残差非易事。因此，现有扩散模型方法要么计算效率低，要么依赖替代方案。本文提出一种新机制，可高效实现扩散模型的原始推测采样。该方法的关键优势在于：首次将块验证从大语言模型成功迁移至扩散模型——其能严格提升草案接受率。此外，我们形式化并分析了无需训练的自由起草者（Free Drafter），一种启发式自推测草稿机制。得益于块验证，该自由起草者在不增加训练成本的前提下，相较现有推测方法实现最高6.3%的推理速度提升，额外开销仅来自现有的并行验证步骤。
原文摘要 · Abstract (English)
Speculative decoding speeds up LLM inference by using a draft model to generate tokens, with an acceptance-rejection scheme that ensures that the output matches the target distribution. Adapting this to continuous diffusions is difficult because speculative sampling requires drawing from a residual distribution. While straightforward in discrete spaces, efficiently sampling this residual in continuous space is non-trivial. Consequently, existing diffusion adaptations either use computationally inefficient sampling techniques or rely on an alternative scheme. In this work, we introduce a novel scheme that efficiently implements the original speculative sampling mechanism for diffusion models. Our approach offers a critical advantage over current methods: it enables us to adapt block verification from LLMs to diffusions -- which provably improves the acceptance rate of drafts. Furthermore, we formalize and analyze the Free Drafter, a heuristic self-speculative drafter for diffusions that requires no training. By enabling block verification, our Free Drafter yields up to a 6.3% speedup over existing speculative methods with no additional training and negligible overhead beyond the existing parallel verification pass.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。