PyTorch implementation of our TPAMI paper "Variational Data-Free Knowledge Distillation for Continual Learning"
Title: Variational Data-Free Knowledge Distillation for Continual Learning
Authors: Xiaorong Li, Shipeng Wang, Jian Sun, Zongben Xu
Email: lixiaorong@stu.xjtu.edu.cn, lixiaorongxjtu@gmail.com
- VDFD (w/o SSL)
cd classification/task_incremental
sh scripts/vdfd.sh
- VDFD
cd classification/task_incremental
sh scripts_ssl/vdfd.sh
For 5-Dataset, you can download this dataset at here.
- VDFD (w/o SSL)
cd classification/class_incremental
sh scripts/vdfd.sh
- VDFD
cd classification/class_incremental
sh scripts_ssl/vdfd.sh
cd segmentation
sh scripts/vdfd.sh
Python (3.6)
PyTorch (1.8.0)
timm
tensorboardX
tqdm
Python (3.6)
Pytorch (1.8.1+cu102)
torchvision (0.9.1+cu102)
tensorboardX (1.8)
apex (0.1)
inplace-abn (1.0.7)
@ARTICLE{Li_2023_tpami,
author={Li, Xiaorong and Wang, Shipeng and Sun, Jian and Xu, Zongben},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
title={Variational Data-Free Knowledge Distillation for Continual Learning},
year={2023},
volume={},
number={},
pages={1-17},
doi={10.1109/TPAMI.2023.3271626}}
The code of classification is based on Adam-NSCL and Continual-Learning-Benchmark.