Fast Inference from Transformers via Speculative Decoding

Authors: Yaniv Leviathan, Matan Kalman, Yossi Matias (2023)

arXiv: 2211.17192

Domains

Inference

TLDR (English)

Uses small draft model to predict multiple tokens, large model verifies in one pass, achieving nearly lossless 2-3x speedup. Standard technique in all inference engines (vLLM, TensorRT-LLM) today.

TLDR（中文）

用一个小 draft model 预测多个 token，再让大模型一次校验，几乎无损地获得 2-3x 加速。是当下所有推理引擎（vLLM、TensorRT-LLM）的标配技术之一。

Appears in These Articles

• Sampling 与 Decoding：从概率到文字
• Sampling and Decoding: From Probabilities to Text

Co-cited Papers

These papers appear in the same articles as this one

• Language Models are Few-Shot Learners Tom Brown et al. (2020)
• Accelerating Large Language Model Decoding with Speculative Sampling Charlie Chen et al. (2023)
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters Charlie Snell et al. (2024)
• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models Jason Wei et al. (2022)

Related Papers

Other papers in the same domain

• DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model DeepSeek-AI (2024)
  ArchitectureInference
• FlashAttention-3: Fast and Accurate Attention with Asympotic IO Complexity Jay Shah et al. (2024)
  Inference
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters Charlie Snell et al. (2024)
  ReasoningInference
• GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints Joshua Ainslie et al. (2023)
  InferenceArchitecture
• Accelerating Large Language Model Decoding with Speculative Sampling Charlie Chen et al. (2023)
  Inference
• FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning Tri Dao (2023)
  Inference