Summary of Flora: Low-rank Adapters Are Secretly Gradient Compressors, by Yongchang Hao et al.
Flora: Low-Rank Adapters Are Secretly Gradient Compressors
by Yongchang Hao, Yanshuai Cao, Lili Mou
First submitted to arxiv on: 5 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
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 solution to reduce memory usage in large neural networks, which is essential for training these models efficiently. The authors introduce Flora, an algorithm that approximates low-rank adaptation (LoRA) by randomly projecting optimization states while maintaining high-rank updates. This approach achieves sublinear space complexity and improves model performance across various tasks and architectures. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Flora is a new way to train big neural networks without using too much memory. It helps reduce the memory needed for training by approximating an old technique called LoRA, which wasn’t very effective because it limited how much the network could learn. Flora gets around this problem by randomly changing the way optimization states are stored. This makes it possible to use even larger neural networks without running out of memory. |
Keywords
* Artificial intelligence * Lora * Low rank adaptation * Optimization
