flâneur

[2402.12374] Sequoia: Scalable, Robust, and Hardware-aware Speculative Decoding

arxiv.org · 7,986 words · saved by 2 readers

Abstract:As the usage of large language models (LLMs) grows, performing efficient inference with these models becomes increasingly important. While speculative decoding has recently emerged as a promising direction for speeding up inference, existing methods are limited in their ability to scale to larger speculation budgets, and adapt to different hyperparameters and hardware. This paper introduces Sequoia, a scalable, robust, and hardware-aware algorithm for speculative decoding. To attain better scalability, Sequoia introduces a dynamic programming algorithm to find the optimal tree structure for the speculated tokens. To achieve robust speculative performance, Sequoia uses a novel sampling and verification method that outperforms prior work across different decoding temperatures. Finally, Sequoia introduces a hardware-aware tree optimizer that maximizes speculative performance by automatically selecting the token tree size and depth for a given hardware platform. Evaluation shows that Sequoia improves the decoding speed of Llama2-7B, Llama2-13B, and Vicuna-33B on an A100 by up to $4.04\times$, $3.73\times$, and $2.27\times$. For offloading setting on L40, Sequoia achieves as low as 0.56 s/token for exact Llama2-70B inference latency, which is $9.96\times$ on our optimized offloading system (5.6 s/token), $9.7\times$ than DeepSpeed-Zero-Inference, $19.5\times$ than Huggingface Accelerate.

S EQUOIA: Scalable and Robust Speculative Decoding Zhuoming Chen1∗ Avner May2∗ Ruslan Svirschevski3,4∗ Yuhsun Huang1 Max Ryabinin2 Zhihao Jia1 Beidi Chen1,5 1 Carnegie Mellon University 2 Together AI 3 Yandex 4 National Research University Higher…

saved by

related reading