MULTIFLOW: Shifting Towards Task-Agnostic Vision-Language Pruning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Farina, Matteo, Mancini, Massimiliano, Cunegatti, Elia, Liu, Gaowen, Iacca, Giovanni, Ricci, Elisa
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911830768615424
author Farina, Matteo
Mancini, Massimiliano
Cunegatti, Elia
Liu, Gaowen
Iacca, Giovanni
Ricci, Elisa
author_facet Farina, Matteo
Mancini, Massimiliano
Cunegatti, Elia
Liu, Gaowen
Iacca, Giovanni
Ricci, Elisa
contents While excellent in transfer learning, Vision-Language models (VLMs) come with high computational costs due to their large number of parameters. To address this issue, removing parameters via model pruning is a viable solution. However, existing techniques for VLMs are task-specific, and thus require pruning the network from scratch for each new task of interest. In this work, we explore a new direction: Task-Agnostic Vision-Language Pruning (TA-VLP). Given a pretrained VLM, the goal is to find a unique pruned counterpart transferable to multiple unknown downstream tasks. In this challenging setting, the transferable representations already encoded in the pretrained model are a key aspect to preserve. Thus, we propose Multimodal Flow Pruning (MULTIFLOW), a first, gradient-free, pruning framework for TA-VLP where: (i) the importance of a parameter is expressed in terms of its magnitude and its information flow, by incorporating the saliency of the neurons it connects; and (ii) pruning is driven by the emergent (multimodal) distribution of the VLM parameters after pretraining. We benchmark eight state-of-the-art pruning algorithms in the context of TA-VLP, experimenting with two VLMs, three vision-language tasks, and three pruning ratios. Our experimental results show that MULTIFLOW outperforms recent sophisticated, combinatorial competitors in the vast majority of the cases, paving the way towards addressing TA-VLP. The code is publicly available at https://github.com/FarinaMatteo/multiflow.
format Preprint
id arxiv_https___arxiv_org_abs_2404_05621
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle MULTIFLOW: Shifting Towards Task-Agnostic Vision-Language Pruning
Farina, Matteo
Mancini, Massimiliano
Cunegatti, Elia
Liu, Gaowen
Iacca, Giovanni
Ricci, Elisa
Computer Vision and Pattern Recognition
While excellent in transfer learning, Vision-Language models (VLMs) come with high computational costs due to their large number of parameters. To address this issue, removing parameters via model pruning is a viable solution. However, existing techniques for VLMs are task-specific, and thus require pruning the network from scratch for each new task of interest. In this work, we explore a new direction: Task-Agnostic Vision-Language Pruning (TA-VLP). Given a pretrained VLM, the goal is to find a unique pruned counterpart transferable to multiple unknown downstream tasks. In this challenging setting, the transferable representations already encoded in the pretrained model are a key aspect to preserve. Thus, we propose Multimodal Flow Pruning (MULTIFLOW), a first, gradient-free, pruning framework for TA-VLP where: (i) the importance of a parameter is expressed in terms of its magnitude and its information flow, by incorporating the saliency of the neurons it connects; and (ii) pruning is driven by the emergent (multimodal) distribution of the VLM parameters after pretraining. We benchmark eight state-of-the-art pruning algorithms in the context of TA-VLP, experimenting with two VLMs, three vision-language tasks, and three pruning ratios. Our experimental results show that MULTIFLOW outperforms recent sophisticated, combinatorial competitors in the vast majority of the cases, paving the way towards addressing TA-VLP. The code is publicly available at https://github.com/FarinaMatteo/multiflow.
title MULTIFLOW: Shifting Towards Task-Agnostic Vision-Language Pruning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2404.05621