chuhong-wang Goto Github PK

** MLIP Registration: https://mlip.skoltech.ru/download

Get started with learning on-the-fly MD (LOTF-MD)

1. Compile MLIP
2. Installation (VASP is already available)
3. LOTF_MD_test.sh

First, MLIP-2 and LAMMPS with mlip interface need to be installed.
1. MLIP (-> make test)
2. Lammps interface with MLIP
3. Lammps
4. Pymatgen (pre- and post- data processing)

User-defined values

1. VASP path, mlp path, LAMMPS path
   
2. training weight energy-force-stress
   

convert AIMD OUTCAR to MLIP input file format .cfg

mlp convert-cfg OUTCAR train.cfg --input-format=vasp-outcar

initiate MTP training

mpirun mlp train pot.mtp train.cfg --max-iter=100 --energy-weight=100 --force-weight=0.1 --stress-weight=1 --curr-pot-name=pot.mtp --trained-pot-name=pot.mtp

compute the matrix later used for extrapolation grade calculation

mlp calc-grade pot.mtp train.cfg train.cfg temp.cfg 

run MTP-MD with active learning on LAMMPS Before running MTP, you need the input files

1. input_lmp (lammps MD setting)
2. mlip.ini (MTP interface, call the trained MTP)
3. start.inp (structure data file, generated by obtain_lmp_inputData.py from POSCAR or CONTCAR)

python obtain_lmp_inputData.py
lmp_g++_serial < input_lmp