M2_SETI/IA/seti_master-master
2023-01-29 16:56:40 +01:00
..
code IA 2023-01-29 16:56:40 +01:00
doc IA 2023-01-29 16:56:40 +01:00
README.md IA 2023-01-29 16:56:40 +01:00

This repository provide all the necessary material for the practical session of SETI Master - AI section.

The main file is a jupyter notebook code/tutorial.ipynb, slides for the first part of the tutorial are under doc/slides.pdf.

There are several dependancies to install.

Getting started

  • We suppose python3.9 is installed and pip

  • First we can move into code folder

    cd code
    
  • We recommend creating a dedicated environment for the tutorial, using for instance virtualenv. If virtualenv is not installed run:

    python3.9 -m pip install virtualenv
    

    Afterwards you can create the virtualenv using:

    virtualenv ~/.virtualenvs/tuto_seti
    

    consider adding -p /usr/bin/python3.9 pointing to your python version if you have multiple python installed. Then activate the environment using

    source ~/.virtualenvs/tuto_seti/bin/activate
    
  • Once in this venv, we can install python dependencies needed for this tutorial with :

    jupyter notebook tutorial.ipynb
    

Tools installation

  • Z3: your linux distribution should have a package (named z3):

    sudo apt install z3
    

    Otherwise check here for a detailed installation instruction

  • ISAIEH can be found compiled in the bin folder.

  • Marabou can be found compiled in the bin folder.

  • PyRAT comes as a precompiled binary in this repository that requires Python 3.9

When everything is installed, jupyter notebook tutorial.ipynb should open the tutorial notebook in your web browser.

In case compiled version does not work

  • ISAIEH: The next is to install dependencies for ISAIEH, which we will use:

    • install the ocaml opam build system with:
    sudo apt install opam
    
    • Additional system dependencies with m4 and gmp are needed at minimum. Install them using
    sudo apt install m4 libgmp3-dev
    

    on a Debian-like.

    • Finally we can install ISAIEH with
    git clone https://git.frama-c.com/pub/isaieh.git && cd isaieh && opam init --disable-sandboxing && opam install -y . && eval $(opam env) && dune build
    
  • Marabou can be compiled from source using:

    sudo apt install cmake libboost-all-dev
    
    git clone https://github.com/NeuralNetworkVerification/Marabou
    cd Marabou
    mkdir build
    cd build
    cmake ..
    cmake --build .
    

    If this does not work an older version binary can be downloaded with

    wget https://aisafety.stanford.edu/marabou/marabou-1.0-x86_64-linux.zip
    

Authors

  • Augustin Lemesle
  • Julien Girard-Satabin
  • Guillaume Charpiat
  • Zakaria Chihani
  • Marc Schoenauer