Summary of Recurformer: Not All Transformer Heads Need Self-attention, by Ruiqing Yan et al.
RecurFormer: Not All Transformer Heads Need Self-Attention
by Ruiqing Yan, Linghan Zheng, Xingbo Du, Han Zou, Yufeng Guo, Jianfei Yang
First submitted to arxiv on: 10 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); 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 RecurFormer, an innovative architecture that reduces the computational costs associated with transformer-based large language models (LLMs) during inference. By replacing certain attention heads with linear recurrent neural networks (RNNs), specifically Mamba, RecurFormer minimizes memory overhead and retains the ability to model long-range dependencies. The approach leverages the recency-aware distribution of attention weights, which focuses on local and short-range dependencies. This modification allows for reusing pre-trained Transformer-based LLMs’ weights with continual training while maintaining generation quality. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary RecurFormer is a new architecture that helps large language models be faster and more efficient when processing long texts. It replaces some of the attention heads in traditional transformer models with simpler, linear networks. This change reduces the amount of memory needed during inference without sacrificing the model’s ability to understand long-range relationships in text. |
Keywords
» Artificial intelligence » Attention » Inference » Transformer
