Zero-Shot Knowledge Distillation in Deep Networks Pytorch

Model	Method	Dataset	top1-accuracy
LeNet5-LeNet5Half	Paper	Mnist	98.77
LeNet5-LeNet5Half	Ours	Mnist	96.98

Usage

python main.py --dataset=mnist --lr=3.0 --t_train=False --num_sample=24000 --batch_size=100

TO-DO

Hard to repruduce the results in paper. The paper author also does not reply.