Summary of Autosculpt: a Pattern-based Model Auto-pruning Framework Using Reinforcement Learning and Graph Learning, by Lixian Jing et al.
AutoSculpt: A Pattern-based Model Auto-pruning Framework Using Reinforcement Learning and Graph Learning
by Lixian Jing, Jianpeng Qi, Junyu Dong, Yanwei Yu
First submitted to arxiv on: 24 Dec 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- 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 This paper proposes AutoSculpt, a novel automated pruning framework that leverages graph learning and deep reinforcement learning (DRL) to optimize deep neural networks (DNNs) for constrained computational resources on edge devices. Existing auto-pruning methods face challenges due to the diversity of DNN models, various operators, and the difficulty in balancing pruning granularity with model accuracy. AutoSculpt aims to address these limitations by identifying and pruning regular patterns within DNN architectures that can be recognized by existing inference engines, enabling runtime acceleration. The framework consists of three key steps: constructing DNNs as graphs to encode their topology and parameter dependencies, embedding computationally efficient pruning patterns, and utilizing DRL to iteratively refine auto-pruning strategies until the optimal balance between compression and accuracy is achieved. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary AutoSculpt is a new way to make deep neural networks work better on devices with limited power. It uses special learning techniques to find parts of the network that can be removed without affecting how well it works. This makes the network use less energy and run faster, which is important for things like self-driving cars or smartphones. The paper shows that AutoSculpt is good at making networks work better, even when they are very complex. |
Keywords
» Artificial intelligence » Embedding » Inference » Pruning » Reinforcement learning
