Summary of Dota: Weight-decomposed Tensor Adaptation For Large Language Models, by Xiaolin Hu et al.
DoTA: Weight-Decomposed Tensor Adaptation for Large Language Models
by Xiaolin Hu, Xiang Cheng, Peiyu Liu, Wei Liu, Jian Luan, Bin Wang, Yong Liu
First submitted to arxiv on: 30 Dec 2024
Categories
- Main: Computation and Language (cs.CL)
- 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 addresses the computational and memory demands of fine-tuning large language models (LLMs) by proposing Weight-Decomposed Tensor Adaptation (DoTA). LoRA, a low-rank adaptation method, approximates updates with low-rank matrices but fails to capture high-dimensional structures. DoTA leverages Matrix Product Operator (MPO) decomposition for effective initialization in fine-tuning LLMs. The authors also introduce QDoTA, a quantized version of DoTA designed for 4-bit quantization. Experiments on commonsense and arithmetic reasoning tasks show that DoTA outperforms random initialization methods with fewer parameters, while QDoTA achieves comparable performance to DoTA on commonsense reasoning tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make big language models smaller and more efficient by finding a better way to start learning new skills. They tried different approaches and found one that works really well. This new method is called Weight-Decomposed Tensor Adaptation (DoTA). It’s like taking apart the model’s existing knowledge, reorganizing it, and then adding new skills on top. The authors also made a version of DoTA that uses less memory, so it can be used even more easily. |
Keywords
» Artificial intelligence » Fine tuning » Lora » Low rank adaptation » Quantization
