Summary of Learning Harmonized Representations For Speculative Sampling, by Lefan Zhang et al.
Learning Harmonized Representations for Speculative Sampling
by Lefan Zhang, Xiaodan Wang, Yanhua Huang, Ruiwen Xu
First submitted to arxiv on: 28 Aug 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 Speculative sampling is a promising approach to accelerate the decoding stage for Large Language Models (LLMs). Recent advancements that leverage target LLM’s contextual information, such as hidden states and KV cache, have shown significant practical improvements. However, these approaches suffer from inconsistent context between training and decoding. We also observe another discrepancy between the training and decoding objectives in existing speculative sampling methods. To address these issues, we propose HArmonized Speculative Sampling (HASS) that learns harmonized representations to accelerate the decoding stage without adding inference overhead through harmonized objective distillation and harmonized context alignment. Our experiments on four LLaMA models demonstrate that HASS achieves a significant speedup ratio of 2.81x-4.05x averaging across three datasets, surpassing EAGLE-2 by 8%-20%. The code is available at this URL. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making computers faster at understanding human language. Right now, it takes a lot of time for computers to translate what we say into text. Scientists are trying to find ways to make this process go faster without sacrificing accuracy. One way they’re doing this is by using something called “speculative sampling”. This approach looks at the context around what we’re saying to help the computer understand better. The problem is that sometimes this approach doesn’t work as well as it should because of inconsistencies between how the computer learns and how it applies what it’s learned. The scientists in this paper propose a new way called “HArmonized Speculative Sampling” (HASS) that helps fix these issues. They tested HASS on four different language models and found that it made them work about 2-4 times faster. |
Keywords
» Artificial intelligence » Alignment » Distillation » Inference » Llama
