Summary of Break the Sequential Dependency Of Llm Inference Using Lookahead Decoding, by Yichao Fu et al.
Break the Sequential Dependency of LLM Inference Using Lookahead Decoding
by Yichao Fu, Peter Bailis, Ion Stoica, Hao Zhang
First submitted to arxiv on: 3 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computation and Language (cs.CL)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper tackles the problem of accelerating autoregressive decoding in large language models (LLMs), which is currently memory bandwidth bounded and results in high latency. The authors propose Lookahead decoding, an exact, parallel algorithm that can be used without auxiliary models or data stores. This method allows for trading off per-step log(FLOPs) to reduce the number of total decoding steps, making it more parallelizable on single or multiple modern accelerators. In comparison tests, Lookahead decoding was able to speed up autoregressive decoding by up to 1.8x on MT-bench and 4x with strong scaling on multiple GPUs in code completion tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models are very good at understanding human language, but they can be slow when trying to generate text. This is because they have to look at a lot of information in their memory to figure out what words come next. The authors of this paper came up with a new way called Lookahead decoding that makes it faster for these models to generate text. It’s like having a superpower that lets them process lots of information all at once, instead of one piece of information at a time. This can make it take as much as 4 times less time for the model to come up with text. |
Keywords
* Artificial intelligence * Autoregressive
