Summary of Mixed Non-linear Quantization For Vision Transformers, by Gihwan Kim et al.
Mixed Non-linear Quantization for Vision Transformers
by Gihwan Kim, Jemin Lee, Sihyeong Park, Yongin Kwon, Hyungshin Kim
First submitted to arxiv on: 26 Jul 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 The proposed mixed non-linear quantization method, which assigns the most error-minimizing quantization method from known methods to each non-linear layer based on layer-wise quantization sensitivity measured by SQNR difference metric, outperforms state-of-the-art models I-BERT, FQ-ViT, and I-ViT in both 8-bit and 6-bit settings for ViT, DeiT, and Swin models. This improvement is achieved through the consideration of non-linear operation-specific quantization methods, showcasing the effectiveness of this approach. The proposed method also outperforms I-BERT and I-ViT when training time is limited. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A group of researchers developed a new way to make computer models smaller without losing their ability to learn. They did this by giving different parts of the model special treatment depending on how much they change the information being processed. This approach worked better than previous methods for reducing the size of certain types of models used in image recognition. The new method can also work faster and be more accurate when training time is limited. |
Keywords
» Artificial intelligence » Bert » Quantization » Vit
