Summary of A Framework to Implement 1+n Multi-task Fine-tuning Pattern in Llms Using the Cgc-lora Algorithm, by Chao Song and Zhihao Ye and Qiqiang Lin and Qiuying Peng and Jun Wang
A Framework to Implement 1+N Multi-task Fine-tuning Pattern in LLMs Using the CGC-LORA Algorithm
by Chao Song, Zhihao Ye, Qiqiang Lin, Qiuying Peng, Jun Wang
First submitted to arxiv on: 22 Jan 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 The proposed unified framework combines multiple-task fine-tuning and low-rank adaptation to efficiently adapt pre-trained language models to various tasks. By implementing a 1 + N multi-task fine-tuning pattern, the framework overcomes high computing costs and seesawing issues. The Customized Gate Control (CGC) Low-rank Adaptation (LoRA) algorithm is designed to leverage both Multi-Task Learning (MTL) and Plug-and-Play Language Models (PEFT) schemes. The framework’s innovative layer includes two types of experts as additional trainable parameters, making it compatible with MTL. Experimental results on two public datasets show that the proposed framework outperforms benchmarks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers has developed a new way to train language models for many tasks at once. They wanted to find a solution that is both efficient and effective. To do this, they created a special algorithm called Customized Gate Control (CGC) Low-rank Adaptation (LoRA). This algorithm combines two different approaches: multi-task learning and plug-and-play language models. The team tested their method on two large datasets and found that it worked better than other methods. |
Keywords
* Artificial intelligence * Fine tuning * Lora * Low rank adaptation * Multi task
