pith. sign in

arxiv: 2501.14713 · v2 · pith:QEAEAKVN · submitted 2025-01-24 · cs.CL · cs.LG

FlexiGPT: Pruning and Extending Large Language Models with Low-Rank Weight Sharing

pith:QEAEAKVNopen to challenge →

classification cs.CL cs.LG
keywords performancebenchmarkslanguagelow-rankmodelsblockscompressiondemonstrate
0
0 comments X
read the original abstract

The rapid proliferation of large language models (LLMs) in natural language processing (NLP) has created a critical need for techniques that enable efficient deployment on memory-constrained devices without compromising performance. We present a method to prune LLMs that selectively prunes model blocks based on an importance score and replaces them with a low-parameter replacement strategy. Specifically, we propose a principled metric to replace each pruned block using a weight-sharing mechanism that leverages unpruned counterparts from the model and block-specific low-rank adapters. Furthermore, we facilitate the learning of these replacement blocks with output feature normalization and an adapter initialization scheme built on low-rank SVD reconstructions. Empirical evaluations demonstrate substantial performance gains over existing methods, achieving state-of-the-art performance on 5/6 benchmarks for a compression rate of 30% and 6/6 benchmarks for a compression rate of 40%. We also demonstrate that our approach can extend smaller models, boosting performance on 6/6 benchmarks using only ~0.3% tokens of extended training with minimal additional parameter costs.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.