This repository contains the code for Section 3.7 Application: Continual Learning of the paper "Squeeze, Recover and Relabel: Dataset Condensation at ImageNet Scale From A New Perspective". [SRe2L Project Page]
- Python 3.8
- PyTorch 1.13.1
- torchvision 0.14.1
- numpy
- scipy
- tqdm
Firstly, follow SRe2L to get the squeezed model and distilled Tiny-ImageNet dataset (100 IPC). Then, run the following script to run continual learning on the distilled dataset.
python main.py \
--steps 5 --lr_net 0.5 \
-T 20 --num_eval 3 --ipc 100 \
--train_dir /path/to/distilled_tiny \
--teacher_path /path/to/tiny-imagenet/resnet18_E50/checkpoint.pth \
| tee cl_sre2l_T20_step5.txt
You can find the example output at cl_sre2l_T20_step5.txt.
If you find our code useful for your research, please cite our paper.
@inproceedings{yin2023squeeze,
title={Squeeze, Recover and Relabel: Dataset Condensation at ImageNet Scale From A New Perspective},
author={Yin, Zeyuan and Xing, Eric and Shen, Zhiqiang},
booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
year={2023}
}
This repository is built upon the codebase of https://github.com/VICO-UoE/DatasetCondensation. We thank the authors for their great work.