DeCoOp: Robust Prompt Tuning with Out-of-Distribution Detection

🏠 Homepage • 📃 Paper

OPT Problem Setting

In this paper, we examine a problem setting known as Open-world Prompt Tuning (OPT), which focuses on tuning prompts for base classes and evaluating their performance on a combination of base and new classes. This setting allows for a comprehensive evaluation of the discriminabilities among the base-class, base-to-new, and new-class categories. We demonstrate that the accuracy in OPT does not align consistently with the previous H metric, as shown in the figure in our paper, indicating the need for a new evaluation metric for OPT.

DeCoOp Approach

We propose the Decomposed Context Optimization (DeCoOp) approach to solve the OPT problem setting. DeCoOp integrates out-of-distribution (OOD) detection into prompt tuning, introducing new-class detectors to enhance the discriminability between the base and new classes. Additionally, DeCoOp employs sub-classifiers to further enhance the discriminability within the base class, thereby improving the performance of the base-class data. The original prompts are retained to ensure the robust performance of the new-class data. To address the issue of not having knowledge of the new-class data during training, we introduce an ensemble strategy to train the DeCoOp approach. The experimental results demonstrate that our DeCoOp approach surpasses state-of-the-art methods, effectively solving the OPT problem setting.

Quick Start

1. Prepare datasets

Prepare the datasets according to the instructions in DATASETS and put them in the DATA directory.

2. Prepare Python Environment

Clone DeCoOp repository, create conda environment, and then install the required packages.

git clone https://github.com/WNJXYK/DeCoOp.git
cd DeCoOp
conda create -n decoop python==3.8
conda activate decoop
pip install -r requirements.txt

3. Run DeCoOp Approach

Run the DeCoOp approach using the command script bash eval_decoop.sh {GpuID} {Architecture} {Logdir}, where GpuID is an integer indicating the GPU you want to use, Architecture is the backbone model of CLIP with options Vit-B16 or Vit-B32, and Logdir is the path to save the experimental results.

For example, to run DeCoOp using GPU 0 and the Vit-B16 backbone model, with the corresponding experimental logs saved to ./results, use the following command:

bash eval_decoop.sh 0 Vit-B16 ./results

TODO List

• Launch project homepage
• Release official code
• Release code based on Dassl.pytorch toolbox

Citation

Please cite the paper if you refer to our code or paper from DeCoOp.

@inproceedings{zhou24decoop,
    author       = {Zhi Zhou and Ming Yang and Jiang-Xin Shi and Lan-Zhe Guo and Yu-Feng Li},
    title        = {DeCoOp: Robust Prompt Tuning with Out-of-Distribution Detection},
    booktitle    = {Proceedings of the 41st International Conference on Machine Learning},
    year         = {2024}
}