git clone https://github.com/cutz-j/ExRep
We use torchvision built-in CIFAR-10.
The pretrained weights can be downloaded by running the file in dataset.
# Download the pre-trained weights
cd weights
bash download_weights.sh
cd ..
pip install -r requirements.txt
We utilize timm repository. https://github.com/rwightman/pytorch-image-models/tree/master/timm
git clone https://github.com/rwightman/pytorch-image-models
pip install timm
cd pytorch-image-models
cp -r timm/ your-work-dir
We utilize the repository of CRD paper author [1]. https://github.com/HobbitLong/RepDistiller
# Dataset and model weights need to be download.
# If you want to change config, check /config/test.yaml
python demo.py --dataset dataset dir \
--pt_weight pt_weight_dir \
[1] Tian, Yonglong, Dilip Krishnan, and Phillip Isola. "Contrastive representation distillation." arXiv preprint arXiv:1910.10699 (2019).