Code for the generation of the benchmark used in the paper "Verifying Neural Networks with SMT Solvers: a Comparison."
The benchmarks used in the paper can be found in the folder "paper_benchmarks/". The code to replicate the generation of the benchmark is in the script benchmark_generation.py The code to replicate the testing of the smt solvers can be found in exp_launcher.py. It should be noted that the solver will need to be installed autonomously and minor modification to the file will be needed to replicate the experiment. To replicate the benchmark generation the package pynever (and its dependencies) is needed.