Summary of Ftuner: a Fast Dynamic Shape Tensors Program Auto-tuner For Deep Learning Compilers, by Pengyu Mu et al.
FTuner: A Fast Dynamic Shape Tensors Program Auto-Tuner for Deep Learning Compilers
by Pengyu Mu, Linquan Wei, Yi Liu, Rui Wang
First submitted to arxiv on: 31 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Distributed, Parallel, and Cluster Computing (cs.DC)
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 This paper addresses the challenge of optimizing dynamic tensors in artificial intelligence models, where input data has varying lengths and resolutions. The shape of these tensors affects model performance, making it difficult to optimize them beforehand. Two common solutions are proposed: adding useless data to match a pre-optimized tensor library or creating small basic tensors and tuning them to minimize padding. However, the latter approach can be time-consuming. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Artificial intelligence models deal with input data of different sizes and shapes, which is tricky because it affects how well they work. The paper looks at ways to fix this problem. There are two main ideas: adding extra information to make the input match a pre-prepared library or creating small building blocks to create the right shape for the input. While the second idea can take time, both approaches try to solve the same issue. |
