Summary of Everybody Prune Now: Structured Pruning Of Llms with Only Forward Passes, by Lucio Dery et al.
Everybody Prune Now: Structured Pruning of LLMs with only Forward Passes
by Lucio Dery, Steven Kolawole, Jean-François Kagy, Virginia Smith, Graham Neubig, Ameet Talwalkar
First submitted to arxiv on: 8 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computation and Language (cs.CL)
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 As machine learning educators writing for technical audiences, we can summarize the abstract by saying: This paper addresses the challenge of making Large Language Models (LLMs) more accessible to practitioners with limited resources. Currently, LLMs are becoming increasingly inaccessible due to their growing size and computational requirements. The authors propose a novel approach to structured pruning of LLMs using only forward passes, which enables practitioners to prune models that are too large for their available hardware. This method, called Bonsai, is gradient-free and perturbative, allowing it to deliver small, fast, and accurate pruned models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make big language models smaller so people with limited computers can use them. Right now, these models are getting too big for most people’s computers, which makes them hard to work with. The authors found a way to make the models smaller by using only one kind of calculation (forward passes). This means that anyone can take a large model and make it smaller so they can run calculations on their own computer. |
Keywords
* Artificial intelligence * Machine learning * Pruning
