Skip to content

HowToBuild.md

Latest commit

 

History

History
61 lines (44 loc) · 2.2 KB

HowToBuild.md

File metadata and controls

61 lines (44 loc) · 2.2 KB

Build Instructions

Follow these steps to build n2p2:

  1. Navigate to the interface directory:

    cd <n2p2_directory>/src/interface/

  2. Modify the makefile at line 58 to enable the necessary LAMMPS packages:

    cd lammps-hdnnp/src/ && $(MAKE) yes-ml-hdnnp && $(MAKE) yes-kspace && $(MAKE) yes-molecule && $(MAKE) yes-manybody && $(MAKE) yes-misc && $(MAKE) yes-rigid && $(MAKE) yes-qeq && $(MAKE) yes-spin && $(MAKE) yes-misc && $(MAKE) yes-tally && $(MAKE) yes-phonon && $(MAKE) mpi

  3. Navigate to the source directory and build the LAMMPS interface:

    cd <n2p2_directory>/src/
export LD_LIBRARY_PATH=~/anaconda3/lib:$LD_LIBRARY_PATH
make clean && make -j8 lammps-hdnnp

    Note: If fails, fix the issue and run make clean-lammps-hdnnp && make -j8 lammps-hdnnp

  4. Install required libraries via conda:

    conda install eigen gsl

  5. Update the makefile.gnu configuration:

    • Set line 8: PROJECT_GSL=<conda_environment_path>/include/
    • Set line 9: PROJECT_EIGEN=<conda_environment_path>/include/eigen3/
    • Set line 24: PROJECT_LDFLAGS_BLAS=-L<conda_environment_path>/lib -lopenblas -lgsl -lgslcblas

  6. Build all targets:

    make -j8 all

Optional: Control charge feature for Q-HDNNP and 4G-HDNNP

You can control whether the predicted charge is included as a feature in the energy neural network for Q-HDNNP and 4G-HDNNP types using the following flags:

  • INCLUDE_CHARGE_IN_ENERGY_Q=1 (default: enabled)
  • INCLUDE_CHARGE_IN_ENERGY_4G=1 (default: enabled)

To disable charge as a feature for either type, set the flag to 0 when building, e.g.:

make clean && make -j8 lammps-hdnnp INCLUDE_CHARGE_IN_ENERGY_Q=0 INCLUDE_CHARGE_IN_ENERGY_4G=1

This allows you to independently control charge inclusion for Q-HDNNP and 4G-HDNNP builds.

Alternative method with system packages

If you have sudo permissions, you can install system packages instead of using conda:

sudo apt install -y libopenmpi-dev libgsl-dev libeigen3-dev libblas-dev liblapack-dev

Replace <n2p2_directory> with the actual path to your n2p2 installation and <conda_environment_path> with the path to your conda environment (e.g., ~/anaconda3).