Summary of Towards Low-bit Communication For Tensor Parallel Llm Inference, by Harry Dong et al.
Towards Low-bit Communication for Tensor Parallel LLM Inference
by Harry Dong, Tyler Johnson, Minsik Cho, Emad Soroush
First submitted to arxiv on: 12 Nov 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Machine Learning (cs.LG)
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 proposes a novel approach to reducing the communication cost in large language model (LLM) inference by introducing a quantization method that leverages consistent outliers in communicated features. The authors demonstrate the effectiveness of this approach, achieving an average reduction of 4.2 bits in communicated values while preserving nearly all original performance across various tasks and models, such as Gemma 2 27B and Llama 2 13B. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about finding a way to make big language models work more efficiently on many devices. Right now, it takes a lot of communication between these devices to get the job done, which can slow things down. The authors came up with a clever trick to reduce this communication cost by focusing on parts that don’t change much. They show that this approach works well for big language models like Gemma 2 and Llama 2, keeping most of their original performance. |
Keywords
» Artificial intelligence » Inference » Large language model » Llama » Quantization
