Summary of Apollo: Sgd-like Memory, Adamw-level Performance, by Hanqing Zhu et al.
APOLLO: SGD-like Memory, AdamW-level Performance
by Hanqing Zhu, Zhenyu Zhang, Wenyan Cong, Xi Liu, Sem Park, Vikas Chandra, Bo Long, David Z. Pan, Zhangyang Wang, Jinwon Lee
First submitted to arxiv on: 6 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Performance (cs.PF)
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 paper addresses the issue of large language models requiring high amounts of memory during training, particularly when using the AdamW optimizer. This limitation can be a barrier to scaling up model training on available hardware or reducing batch sizes. To overcome this challenge, researchers have proposed various memory-efficient optimizers that reduce memory usage. However, these alternatives face significant challenges in terms of computational cost, performance trade-offs compared to AdamW, and remaining memory overhead. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about making language models work better on computers by using less memory during training. This is important because big models take up a lot of space and can be hard to train without using too many powerful computers or reducing the amount of data used for training. The authors want to find ways to make model training more efficient, but they face some big challenges to do so. |
