A genetic algorithm to find the stable configuration of HfNiTi
To use this code, you'll first need to get access to the mlip-2 repository by filling out the form at https://mlip.skoltech.ru/register/ Once you have access, you can chock out the MLIPPY branch. Follow the instructions on the readme file.
You will need the pymatgen package. The most recent version for me didn't work with my environment, so if you run into that problem, you can get an older version that works with this command: python -m pip install pymatgen==v2021.3.3 --force-reinstall
If you're running this on MaryLou, than I found I needed to load these modules:
module load intel-compilers libfabric intel-mkl intel-mpi
and add these directories to the python path:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/apps/intel_parallel_studio_xe/2017_update8/compilers_and_libraries_2017.8.262/linux/mpi/intel64/lib:/apps/libfabric/1.8.0/lib:/apps/intel_parallel_studio_xe/2017_update8/compilers_and_libraries_2017.8.262/linux/compiler/lib/intel64:/apps/intel_parallel_studio_xe/2017_update8/compilers_and_libraries_2017.8.262/linux/mkl/lib/intel64/
when I compiled MLIPPY and each time I wanted to use it.
I've included some tools that can help analyze the data from the iterations of the genetic algorithm (analyse_blacklist.py, analyse_mlippy.py, view_mlippy.py, etc.) The files that are vital for the genetic algorithm are start_children.py (generate the first generation of samples), generate_unit_cell.py, and gen_alg.py.
My lowest-energy results are included under the "special" directories in luke_data