VRL is a toolbox for verification in reinforcement learning systems.
Require Python2.7 and usual python packages such as numpy, scipy, matplotlib, ...
- pip install tensorflow
- pip install tflearn
- pip install numba .
- From https://github.com/Z3Prover/z3/releases, download Z3 binary.
export PYTHONPATH=PATH_UNPACK_Z3/z3/bin/:
.
- Download Julia-0.6 from https://julialang.org/downloads/oldreleases.html.
export PATH="$PATH:PATH_UNPACK_JULIA/julia/bin"
Download and licsence the Mosek solver
-
Download the MAC OS 64bit x86 MOSEK Optimization Suite distribution from https://mosek.com/downloads/ and unpack it into a chosen directory.
-
For MAC, run the command:
python <MSKHOME>/mosek/8/tools/platform/osx64x86/bin/install.py
where<MSKHOME>
is the directory where MOSEK was installed. This will set up the appropriate shared objects required when using MOSEK.
-
Add the path
<MSKHOME>/mosek/8/tools/platform/osx64x86/bin
to the OS variable PATH. -
License the tool.
-
Enter Julia command line.
-
Enter following commands:
julia> Pkg.status()
julia> Pkg.add("SumOfSquares")
julia> Pkg.status("SumOfSquares")
SumOfSquares 0.2.0
julia> Pkg.status("JuMP")
JuMP 0.18.2
julia> Pkg.add("MathOptInterface")
julia> Pkg.add("DynamicPolynomials")
julia> Pkg.add("SCS")
julia> Pkg.add("MathOptInterfaceMosek")
julia> Pkg.add("Clp")