Summary of A General and Efficient Training For Transformer Via Token Expansion, by Wenxuan Huang et al.
A General and Efficient Training for Transformer via Token Expansion
by Wenxuan Huang, Yunhang Shen, Jiao Xie, Baochang Zhang, Gaoqi He, Ke Li, Xing Sun, Shaohui Lin
First submitted to arxiv on: 31 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
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 Token Expansion (ToE) scheme aims to accelerate the training of Vision Transformers (ViTs) while maintaining the consistency and universality of original transformer networks. By introducing an “initialization-expansion-merging” pipeline, ToE preserves the intermediate feature distribution, preventing the loss of crucial learnable information during training. This approach can be seamlessly integrated into various transformer architectures, including DeiT and LV-ViT, as well as efficient training frameworks like EfficientTrain. Extensive experiments demonstrate that ToE achieves 1.3x faster training with no performance loss or even gains over full-token training baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new way to train special kinds of artificial intelligence (AI) called Vision Transformers is introduced. These AI models are very good at recognizing images, but they take a long time to learn. To speed up this learning process, researchers developed a new method that works consistently across different AI models and training settings. This new method, called Token Expansion, helps the AI models learn faster without losing their ability to recognize images accurately. The results show that Token Expansion can train Vision Transformers 1.3 times faster than before, with some cases even performing better than traditional methods. |
Keywords
* Artificial intelligence * Token * Transformer * Vit
