This repository contains the official implementation of our CVPR 2025 paper, "LoRA Subtraction for Drift-Resistant Space in Exemplar-Free Continual Learning."
This code is implemented in PyTorch, and we perform the experiments under the following environment settings:
- python = 3.11.4
- torch = 2.0.1
- torchvision = 0.15.2
- timm = 0.6.7
The code has been tested on Linux Platform with a GPU (RTX3080 Ti).
If you see the following error, you may need to install a PyTorch package compatible with your infrastructure.
RuntimeError: No HIP GPUs are available or ImportError: libtinfo.so.5: cannot open shared object file: No such file or directory
For example if your infrastructure only supports CUDA == 11.1, you may need to install the PyTorch package using CUDA11.1.
pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html
If you see the following error, you can resolve it by installing a lower version of timm, such as pip install timm==0.6.7.
TypeError: 'PretrainedCfg' object is not subscriptable
- Create a folder
data/ - CIFAR 100: should automatically be downloaded
- ImageNet-R: retrieve from link. After unzipping, place it into
data/folder
The JSON configuration files in configs/ are preconfigured for 20-task scenarios. You can modify init_cls, increment and total_sessions parameters in configs/[dataset].json to configure different CIL settings.
-
CIFAR100:
python main.py --config configs/cifar100.json -
ImageNet-R:
python main.py --config configs/imagenetr.json
@inproceedings{liu2025lora,
title={LoRA Subtraction for Drift-Resistant Space in Exemplar-Free Continual Learning},
author={Liu, Xuan and Chang, Xiaobin},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year={2025}
}We appreciate the following repositories for their contributions of useful components and functions to our work.