Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

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

About

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.

Shihan Wu, Ji Zhang, Pengpeng Zeng, Lianli Gao, Jingkuan Song, Heng Tao Shen• 2024

Related benchmarks

TaskDatasetResultRank
Image ClassificationEuroSAT
Accuracy91.57
497
Image ClassificationFood-101
Accuracy87.13
494
Image ClassificationDTD
Accuracy74.07
487
Image ClassificationFlowers102--
478
Image ClassificationStanford Cars
Accuracy87.43
477
Image ClassificationSUN397
Accuracy76.8
425
Image ClassificationUCF101
Top-1 Acc87.7
404
Image ClassificationImageNet
Top-1 Accuracy73.83
324
Image ClassificationFood101--
309
Image ClassificationStanfordCars--
266
Showing 10 of 39 rows

Other info

Code

Follow for update