freePruner: A Training-free Approach for Large Multimodal Model Acceleration

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Xu, Bingxin, Shang, Yuzhang, Ge, Yunhao, Lou, Qian, Yan, Yan
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917846318055424
author Xu, Bingxin
Shang, Yuzhang
Ge, Yunhao
Lou, Qian
Yan, Yan
author_facet Xu, Bingxin
Shang, Yuzhang
Ge, Yunhao
Lou, Qian
Yan, Yan
contents Large Multimodal Models (LMMs) have demonstrated impressive capabilities in visual-language tasks but face significant deployment challenges due to their high computational demands. While recent token reduction methods show promise for accelerating LMMs, they typically require extensive retraining or fine-tuning, making them impractical for many state-of-the-art models, especially those with proprietary training data. We propose freePruner, a training-free token reduction approach that can be directly applied to any open-source LMM without additional training. Unlike existing methods that rely heavily on token merging operations, freePruner employs a two-stage token selection strategy: (1) identifying pivotal tokens that capture high-level semantic information using our designed contribution degree metric, and (2) selecting complementary tokens that preserve essential low-level visual details through attention pattern analysis. Extensive experiments demonstrate that freePruner achieves 2x acceleration while maintaining comparable performance across mainstream visual question-answering benchmarks in the training-free setting. Moreover, freePruner is orthogonal to and can be combined with other post-training acceleration techniques, such as post-training quantization, providing a practical solution for efficient LMM deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2411_15446
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle freePruner: A Training-free Approach for Large Multimodal Model Acceleration
Xu, Bingxin
Shang, Yuzhang
Ge, Yunhao
Lou, Qian
Yan, Yan
Computer Vision and Pattern Recognition
Artificial Intelligence
Large Multimodal Models (LMMs) have demonstrated impressive capabilities in visual-language tasks but face significant deployment challenges due to their high computational demands. While recent token reduction methods show promise for accelerating LMMs, they typically require extensive retraining or fine-tuning, making them impractical for many state-of-the-art models, especially those with proprietary training data. We propose freePruner, a training-free token reduction approach that can be directly applied to any open-source LMM without additional training. Unlike existing methods that rely heavily on token merging operations, freePruner employs a two-stage token selection strategy: (1) identifying pivotal tokens that capture high-level semantic information using our designed contribution degree metric, and (2) selecting complementary tokens that preserve essential low-level visual details through attention pattern analysis. Extensive experiments demonstrate that freePruner achieves 2x acceleration while maintaining comparable performance across mainstream visual question-answering benchmarks in the training-free setting. Moreover, freePruner is orthogonal to and can be combined with other post-training acceleration techniques, such as post-training quantization, providing a practical solution for efficient LMM deployment.
title freePruner: A Training-free Approach for Large Multimodal Model Acceleration
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2411.15446