Summary of Eora: Training-free Compensation For Compressed Llm with Eigenspace Low-rank Approximation, by Shih-yang Liu et al.
EoRA: Training-free Compensation for Compressed LLM with Eigenspace Low-Rank Approximation
by Shih-Yang Liu, Maksim Khadkevich, Nai Chit Fung, Charbel Sakr, Chao-Han Huck Yang, Chien-Yi Wang, Saurav Muralidharan, Hongxu Yin, Kwang-Ting Cheng, Jan Kautz, Yu-Chiang Frank Wang, Pavlo Molchanov, Min-Hung Chen
First submitted to arxiv on: 28 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
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 redefines the model compression problem as a customized compensation issue, aiming to introduce residual low-rank paths in compressed models to balance accuracy and overhead. The authors propose Training-free Eigenspace Low-Rank Approximation (EoRA), a method that directly minimizes compression-induced errors without training, achieving fast optimization using minimal calibration data. EoRA projects errors into the eigenspace of input activations, prioritizing high-importance error components based on eigenvalues. It can be seamlessly integrated with fine-tuning and quantization to improve effectiveness and efficiency. The authors demonstrate EoRA’s superiority in compensating errors for compressed LLaMA2/3 models on various tasks, such as language generation and math reasoning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making machines more efficient without sacrificing their ability to do tasks correctly. It introduces a new way of fixing mistakes made when compressing machine learning models. The authors created a method called EoRA that can quickly fix errors in compressed models using just a little bit of extra information. EoRA works by looking at the patterns in how the model is used and focusing on the most important parts to make it more accurate. This new approach makes machines more flexible and able to do different tasks better. |
Keywords
» Artificial intelligence » Fine tuning » Machine learning » Model compression » Optimization » Quantization
