abICS

Software framework for training a machine learning model to reproduce first-principles energies and then using the model to perform configurational sampling in disordered systems. It has been developed with an emphasis on multi-component solid-state systems such as metal and oxide alloys. At present, Quantum Espresso, VASP and OpenMX can be used as first-principles energy calculators, and aenet can be used to construct neural network potentials.

Official Page: http://www.pasums.issp.u-tokyo.ac.jp/abics/en

GitHub: https://github.com/issp-center-dev/abICS

• The directory where sample scripts and input files are stored.
  

  /home/issp/materiapps/intel/abics/

• How to run abICS

1. 1.  Preparing the computing environments 
      

      $ source /home/issp/materiapps/intel/abics/abicsvars.sh

      This will pass a PATH to python3 with the latest version of abICS installed.
   2. Preparing  input files
      

      $ cp -rf $ABICS_ROOT/examples . 
      $ cd examples/active_learning_qe/baseinput_ref 
      $ sh ./download_pp.sh 
      $ cd .. 
      $ cp -rf /home/issp/materiapps/intel/abics/sample_jobscript/* .

   3. Submitting job

      • For system B

        To learn aenet first, active learning is done by typing the following command

        $ sbatch AL.sh

        After the calculation is completed, the following command is entered to perform the Monte Carlo calculation.

        $ sbatch MC.sh

        For more information, see the tutorial in the manual.
      • For system C

        To learn aenet first, active learning is done by typing the following command

        $ qsub AL.sh

        After the calculation is completed, the following command is entered to perform the Monte Carlo calculation.

        $ qsub MC.sh

        For more information, see the tutorial in the manual.