Summary of Besa: Pruning Large Language Models with Blockwise Parameter-efficient Sparsity Allocation, by Peng Xu et al.
BESA: Pruning Large Language Models with Blockwise Parameter-Efficient Sparsity Allocation
by Peng Xu, Wenqi Shao, Mengzhao Chen, Shitao Tang, Kaipeng Zhang, Peng Gao, Fengwei An, Yu Qiao, Ping Luo
First submitted to arxiv on: 18 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 This paper introduces a novel technique for pruning large language models (LLMs) called blockwise parameter-efficient sparsity allocation (BESA). BESA aims to reduce the computational footprint of LLMs while maintaining their impressive performance. The approach targets individual transformer blocks and allocates layer-specific sparsity, ensuring reduced performance degradation after pruning. Experiments show that BESA achieves state-of-the-art performance on LLaMA1, LLaMA2 with 7B-70B parameters, pruned efficiently on a single A100 GPU in five hours. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make big language models more efficient and useful. It’s like taking a giant puzzle apart into smaller pieces to make it easier to solve. The new method, called BESA, is better than other ways of making the models smaller because it keeps them working just as well. Scientists used this method on some really large models and found that it worked amazingly well, even on powerful computers. |
Keywords
* Artificial intelligence * Parameter efficient * Pruning * Transformer
