- CUDA 11.8
- Python >=3.10
-
Install CUDA 11.8 by following the official documentation provided by NVIDIA.
-
Create a virtual environment (optional but recommended):
python3 -m venv myenv source myenv/bin/activate
-
Install the necessary PyTorch package using pip:
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
-
Run the main script /experiment.py using the following commands:
python experiment.py --run_all
from the repo source to run all experiments. Note however that the images used are not published and would, therefore have to be added for the script to run. An example of the Attack Model training is shown in /attack_model.ipynb.
If the experiments ran, they can be visualized with
python experiment.py --plot --clip