Iterative procedure prerequisites

When training a neural network potential (NNP) for a chemical system (or several systems that you want to describe with the same NNP), you will often want to explore the chemical space as diversely as possible. In ArcaNN, this is made possible by the use of systems. A system corresponds to a particular way of exploring the chemical space that interests you and will be represented by specific datasets within the total training set of the NNP. A dataset corresponds to an ensemble of structures (e.g., atomic positions, types of atoms, box size, etc.) and corresponding labels (e.g., energy, forces, virial).

You get the idea: you need a subsystem for every kind of chemical composition, physical state (temperature, density, pressure, cell size, etc.), biased reactive pathway, and more that you wish to include in your final training dataset.

Attention, systems are defined once and for all in the Initialization of the procedure. Because of this, every time you want to include a new subsystem (such as transition state structures, see SN2 example), you will need to initialize the procedure again. This is very simple—you only need to create a new $WORK_DIR and include the necessary files in user_files/ for each extra system you want to add.

To initiate the iterative training procedure, you should create in your $WORK_DIR two folders: user_files/ and data/.

In user_files/ you will store all the files needed for each step. You can find some templates to start with in the GitHub Repository, now available in your machine at your ArcaNN installation location arcann_training/examples/user_files/.

  • For the exploration step, you must adapt the template files found in exploration_lammps/ or exploration_sander_emle/ depending of your choice. You will need the following files:

    • The input files: SYSTEM.in for LAMMPS and SYSTEM.xml for i-PI.
    • The plumed files: plumed_SYSTEM.dat where SYSNAME refers to the system name (additional PLUMED files can be used as plumed_*_SYSNAME.dat, which will also be taken into account for explorations).

  • For the labeling step, use the template files found in labeling_cp2k/ or labeling_orca/. You will need the following files:

    • The input files : [1-2]_SYSNAME_labeling_XXXXX_[cluster].inp, where [cluster] refers to the short string selected for the labeling cluster in the machine.json; see Labeling.

  • For the training step, from training_deepmd you will need:

    • A DeePMD-kit JSON file named dptrain_VERSION.json, where VERSION is the DeePMD-kit version that you will use (e.g., 2.1; currently supported versions are 2.0, 2.1, and 2.2).

  • The SLURM scripts for individual jobs and for job arrays are organised by step in several fordels. Prepare your files according to your software choice for each step. You can fin them in:

    • job_exploration_lammps_slurm and job_exploration_sander_emle_slurm for exploration.
    • job_labeling_orca_slurm and job_labeling_cp2k_slurm for the labeling.
    • job_training_deepmd_slurm for training.
    • job_test_deepmd_slurm for testing.

We strongly advise you to create the previous files starting from the templates, as they contain replaceable strings for the key parameters that will be updated by the procedure.

  • A representative file in the LAMMPS Data Format format for each system, named SYSTEM.lmp, where SYSTEM refers to the system name (we will refer to them as LMP files). This file represent the configuration of the system, with the number of atoms and the number of types of atoms, the simulation cell dimension, the atomic masses for each type and the atomic geometry of your system. They will be used as starting point for the first exploration.

The order of the atoms in the LMP files must be identical for every system and must match the order indicated in the "type_map" keyword of the DeePMD-kit dptrain_VERSION.json training file.

  • A properties file must be provided and named properties.txt. This is file will be used by atomsk. In the case you have several systems with different chemical composition, the properties file must be the same for all systems to ensure consistant type mapping.

Then, gather all these files and store them inside the user_files/ directory. Don't create subdirectories inside user_files.

Finally, you also need to prepare at least one initial dataset which will be used for your first neural networks training, that you will store in the data/ directory. It must follow DeePMD-kit standards and should contain a type.raw file and set.000/ folder with box.npy, coord.npy, energy.npy and force.npy (see DeePMD-kit documentation). You can prepare as many initial datasets as you wish and they should all be stored in the $WORK_DIR/data/ folder with a folder name starting with init_.