Skip to content

Latest commit

 

History

History
28 lines (18 loc) · 946 Bytes

README.md

File metadata and controls

28 lines (18 loc) · 946 Bytes
Raw

EdgePro

In this repository, code is for our paper EdgePro: Edge Deep Learning Model Protection via Neuron Authorization

Installation

Install Python>=3.6.0 and Pytorch>=1.4.0

Usage

Prepare the dataset: MNIST and CIFAR-10 dataset: MNIST and CIFAR will be automatically download

Code structures

normal_training.py: normal training on datasets
EdgePro_training.py: lock training on datasets
ranking_neurons.py: neuron importance ranking code
select_az_neurons.py: select authorization neurons
options.py: a parser of the configs, also assigns EdgePro given the configs

You can execute the task using the following commands:

python src/EdgePro_training.py --dataset mnist --model LeNet_AZ --epoch 30 --lam 0 --gpu 0

Check out parser.py for the use of the arguments, most of them are self-explanatory. Prior to training with EdgePro, it is necessary to run ranking_neurons.py and select_az_neurons.py to select authorized neurons.