/Prompt-based-CL-Toolbox

Primary LanguagePython

Prompt-tuning based Incremental Learning Toolbox

This repository contains awesome prompt-based incremental learning methods:

L2P: Wang, Zifeng, et al. "Learning to prompt for continual learning." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2022.
DualPrompt: Wang, Zifeng, et al. "Dualprompt: Complementary prompting for rehearsal-free continual learning." Computer Vision–ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XXVI. Cham: Springer Nature Switzerland, 2022.
S-Prompts: Wang, Yabin, Zhiwu Huang, and Xiaopeng Hong. "S-Prompts Learning with Pre-trained Transformers: An Occam’s Razor for Domain Incremental Learning." Advances in Neural Information Processing Systems.

Environment

Ubuntu 20.04.1 LTS
NVIDIA GeForce RTX 3090
Python 3.8 CUDA 116

install using the requirements.txt

pip install -r requirements.txt

Run

bash run.sh

You can also start with:

python -m torch.distributed.launch \
    --nproc_per_node <customize_by_yourself> \
    --master_port <customize_by_yourself> \
    --use_env main.py 
        <sprompt_cddb_slip, l2p_cifar100, dualp_cifar100,
        l2p_core50, dualp_core50>

Acknowledgement

Official Jax Implementation of L2P and DualP
Reimplemented in PyTorch of L2p
Reimplemented in PyTorch of DualP
Official implementation of S-prompts
PyCIL: A Python Toolbox for Class-Incremental Learning