Turning Trash into Treasure: Accelerating Inference of Large Language Models with Token Recycling
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866916755830472704 |
|---|---|
| author | Luo, Xianzhen Wang, Yixuan Zhu, Qingfu Zhang, Zhiming Zhang, Xuanyu Yang, Qing Xu, Dongliang |
| author_facet | Luo, Xianzhen Wang, Yixuan Zhu, Qingfu Zhang, Zhiming Zhang, Xuanyu Yang, Qing Xu, Dongliang |
| contents | Massive parameters of LLMs have made inference latency a fundamental bottleneck. Speculative decoding represents a lossless approach to accelerate inference through a guess-and-verify paradigm. Some methods rely on additional architectures to guess draft tokens, which need extra training before use. Alternatively, retrieval-based training-free techniques build libraries from pre-existing corpora or by n-gram generation. However, they face challenges like large storage requirements, time-consuming retrieval, and limited adaptability. Observing that candidate tokens generated during the decoding process are likely to reoccur in future sequences, we propose Token Recycling. It stores candidate tokens in an adjacency matrix and employs a breadth-first-search (BFS)-like algorithm to construct a draft tree, which is then validated through tree attention. New candidate tokens from the decoding process are then used to update the matrix. Token Recycling requires \textless2MB of additional storage and achieves approximately 2x speedup across all sizes of LLMs. It significantly outperforms existing train-free methods by 30\% and even a widely recognized training method by 25\%. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_08696 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Turning Trash into Treasure: Accelerating Inference of Large Language Models with Token Recycling Luo, Xianzhen Wang, Yixuan Zhu, Qingfu Zhang, Zhiming Zhang, Xuanyu Yang, Qing Xu, Dongliang Computation and Language Machine Learning Massive parameters of LLMs have made inference latency a fundamental bottleneck. Speculative decoding represents a lossless approach to accelerate inference through a guess-and-verify paradigm. Some methods rely on additional architectures to guess draft tokens, which need extra training before use. Alternatively, retrieval-based training-free techniques build libraries from pre-existing corpora or by n-gram generation. However, they face challenges like large storage requirements, time-consuming retrieval, and limited adaptability. Observing that candidate tokens generated during the decoding process are likely to reoccur in future sequences, we propose Token Recycling. It stores candidate tokens in an adjacency matrix and employs a breadth-first-search (BFS)-like algorithm to construct a draft tree, which is then validated through tree attention. New candidate tokens from the decoding process are then used to update the matrix. Token Recycling requires \textless2MB of additional storage and achieves approximately 2x speedup across all sizes of LLMs. It significantly outperforms existing train-free methods by 30\% and even a widely recognized training method by 25\%. |
| title | Turning Trash into Treasure: Accelerating Inference of Large Language Models with Token Recycling |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2408.08696 |