Summary of Singular Value Scaling: Efficient Generative Model Compression Via Pruned Weights Refinement, by Hyeonjin Kim et al.
Singular Value Scaling: Efficient Generative Model Compression via Pruned Weights Refinement
by Hyeonjin Kim, Jaejun Yoo
First submitted to arxiv on: 23 Dec 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
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 A novel pruning technique, Singular Value Scaling (SVS), is introduced to refine pruned weights in generative models like GANs and Diffusion models. The approach targets the dominant singular vectors in pruned weights that hinder fine-tuning efficiency and performance. By minimizing these disparities, SVS enhances weight initialization, improving fine-tuning and compression performance without additional training costs. Experimental results demonstrate improved compression performance on StyleGAN2, StyleGAN3, and DDPM. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Pruning is a technique used to make computer models smaller and more efficient. Researchers have been trying to figure out how to make this process work better for certain types of models called generative models. These models are used for tasks like generating new images or videos that look realistic. One problem with pruning these models is that it can make them harder to fine-tune, which means making small adjustments to get the best results. A team of researchers has come up with a new way to prune generative models called Singular Value Scaling (SVS). This method helps to make the model’s weights more suitable for fine-tuning and improves its performance without needing additional training. The team tested their approach on different types of generative models and found that it worked well. |
Keywords
» Artificial intelligence » Fine tuning » Pruning
