Summary of Less Memory Means Smaller Gpus: Backpropagation with Compressed Activations, by Daniel Barley et al.
Less Memory Means smaller GPUs: Backpropagation with Compressed Activations
by Daniel Barley, Holger Fröning
First submitted to arxiv on: 18 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 method reduces the memory footprint and data movement during deep neural network (DNN) training by compressing activation maps for the backward pass using pooling. This approach maintains prediction accuracy while decreasing peak memory consumption by 29%. The technique is demonstrated on ResNet, a common vision architecture, showing convergence and examining effects on feature detection. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research helps solve a big problem in machine learning called “ever-growing scale”. It means that as deep neural networks get bigger, they require more and more computing power. Some networks are so large that they need supercomputers to train them! This new method makes it possible to reduce the amount of memory needed for training, which is important because computer chips don’t have enough space for all those calculations. |
Keywords
* Artificial intelligence * Machine learning * Neural network * Resnet
