ZvikaZ/evo_attack

Python

Prepare the environment

conda create -n attackar numpy=1.19 pytorch tensorflow python=3.9
conda activate attackar
pip install piqa adversarial-robustness-toolbox
pip install git+https://github.com/RobustBench/robustbench@v0.2.1

Download the trained models

Download model from Google Drive link at https://github.com/huyvnphan/PyTorch_CIFAR10
Unzip it, to get state_dicts/*.pt
TODO: get models/*.pth

Run

python main.py --model=inception_v3 --dataset=cifar10 --eps=0.025 --pop=70 --gen=600 --images=1 --tournament=25