Improved Methods for Model Pruning
Keywords: model pruning, optimization, alignment, attention
TL;DR: We proposed an improved model pruning performance optimization technique for large language and vision models.
Abstract: Model pruning is a performance optimization technique for large language and vision models. However, existing pruning methods often lead to significant performance degradation or require extensive retraining and fine-tuning. This technique aims to identify and remove neurons, connections unlikely leading to the contribution during the machine generation phase. Our goal is to obtain a much smaller and faster foundational model that can quickly generate content almost as good as those of the unpruned models. We propose MAMA (short for Movement and Magnitude Analysis), an improved pruning method that effectively reduces model size and network computational complexity while maintaining performance comparable to the original unpruned model even at extreme pruned levels. The improved method is based on weights, bias, activations and proposed novel pruning indicators. Empirical results show that our method outperforms and be comparable to state-of-the-art methods across various pruning levels. All our code, models, dataset, and demo are publicly available.
Primary Area: optimization
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 3614
Loading