Skip Tuning: Pre-trained Vision-Language Models are Effective and Efficient Adapters Themselves

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Shihan, Zhang, Ji, Zeng, Pengpeng, Gao, Lianli, Song, Jingkuan, Shen, Heng Tao
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913735644282880
author Wu, Shihan
Zhang, Ji
Zeng, Pengpeng
Gao, Lianli
Song, Jingkuan
Shen, Heng Tao
author_facet Wu, Shihan
Zhang, Ji
Zeng, Pengpeng
Gao, Lianli
Song, Jingkuan
Shen, Heng Tao
contents Prompt tuning (PT) has long been recognized as an effective and efficient paradigm for transferring large pre-trained vision-language models (VLMs) to downstream tasks by learning a tiny set of context vectors. Nevertheless, in this work, we reveal that freezing the parameters of VLMs during learning the context vectors neither facilitates the transferability of pre-trained knowledge nor improves the memory and time efficiency significantly. Upon further investigation, we find that reducing both the length and width of the feature-gradient propagation flows of the full fine-tuning (FT) baseline is key to achieving effective and efficient knowledge transfer. Motivated by this, we propose Skip Tuning, a novel paradigm for adapting VLMs to downstream tasks. Unlike existing PT or adapter-based methods, Skip Tuning applies Layer-wise Skipping (LSkip) and Class-wise Skipping (CSkip) upon the FT baseline without introducing extra context vectors or adapter modules. Extensive experiments across a wide spectrum of benchmarks demonstrate the superior effectiveness and efficiency of our Skip Tuning over both PT and adapter-based methods. Code: https://github.com/Koorye/SkipTuning.
format Preprint
id arxiv_https___arxiv_org_abs_2412_11509
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Skip Tuning: Pre-trained Vision-Language Models are Effective and Efficient Adapters Themselves
Wu, Shihan
Zhang, Ji
Zeng, Pengpeng
Gao, Lianli
Song, Jingkuan
Shen, Heng Tao
Computer Vision and Pattern Recognition
Prompt tuning (PT) has long been recognized as an effective and efficient paradigm for transferring large pre-trained vision-language models (VLMs) to downstream tasks by learning a tiny set of context vectors. Nevertheless, in this work, we reveal that freezing the parameters of VLMs during learning the context vectors neither facilitates the transferability of pre-trained knowledge nor improves the memory and time efficiency significantly. Upon further investigation, we find that reducing both the length and width of the feature-gradient propagation flows of the full fine-tuning (FT) baseline is key to achieving effective and efficient knowledge transfer. Motivated by this, we propose Skip Tuning, a novel paradigm for adapting VLMs to downstream tasks. Unlike existing PT or adapter-based methods, Skip Tuning applies Layer-wise Skipping (LSkip) and Class-wise Skipping (CSkip) upon the FT baseline without introducing extra context vectors or adapter modules. Extensive experiments across a wide spectrum of benchmarks demonstrate the superior effectiveness and efficiency of our Skip Tuning over both PT and adapter-based methods. Code: https://github.com/Koorye/SkipTuning.
title Skip Tuning: Pre-trained Vision-Language Models are Effective and Efficient Adapters Themselves
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.11509