NEP-GPUMD-tutorial

An example of training a NEP potential on Pt@zeolite database and running a GPUMD simulation.
For installation instructions, please watch the second half of this video.
Please follows these links to learn more about NEP/GPUMD and PLUMED.

NEP training

1. Use the command python3 1_train_test_split.py to create the training data (train.xyz) and testing data (test.xyz).
2. After installing the GPUMD software, Edit the command in run_NEP.sh to 'Path to your GPUMD folder'/src/nep. This calls the nep excecutable that you installed.
3. Run sbatch run_NEP.sh to train your NEP potential.

GPUMD Simulation

1. Transfer the nep.txt file from your NEP training folder to the GPUMD folder (This is your forcefield).
2. Take an initial trajectory and save it as model.xyz (This can be easily done through ASE GUI).
3. Run the following commands to create an initial_traj.pdb for PLUMED. You can do it easily on the terminal by typing python.

from ase.io import read,write 
read=read('model.xyz',index=':')
write=write('initial_traj.pdb',read,format="proteindatabank") 

4. Make sure that you have your own plumed.dat (for PLUMED) and run.in (for GPUMD) files. Example scripts already provided for reference in this repo.
5. Edit the command in run_GPUMD.sh to 'Path to your GPUMD folder'/src/gpumd. This calls the gpumd excecutable that you installed.
6. Run sbatch run_GPUMD.sh to run your very own GPUMD based metadynamics simulation.