Quick start

Prepare VASP calculations

Run VASP calculations at this step.

Collect paths of VASP calculations

We provided examples of VASP outputs at VASP_calculations_example.
Find all directories containing OUTCAR file:
```
find . -name OUTCAR > paths.log
```
Remove the string ‘OUTCAR’ in paths.log.
```
sed -i 's/OUTCAR$//g' paths.log
```
Specify the absolute paths in paths.log.
```
sed -i "s#^.#${PWD}#g" paths.log
```

Build database

from agat.data import BuildDatabase
if __name__ == '__main__':
    database = BuildDatabase(mode_of_NN='ase_dist', num_of_cores=16)
    dataset = database.build()

Train AGAT model

from agat.model import Fit
f = Fit()
f.fit()

Application (geometry optimization)

from ase.optimize import BFGS
from ase.io import read
from agat.app import AgatCalculator

model_save_dir = 'agat_model'
graph_build_scheme_dir = 'dataset'

atoms = read('POSCAR')
calculator=AgatCalculator(model_save_dir,
                          graph_build_scheme_dir)
atoms = Atoms(atoms, calculator=calculator)
dyn = BFGS(atoms, trajectory='test.traj')
dyn.run(fmax=0.05)

Application (high-throughput prediction)

from agat.app.cata import HtAds

model_save_dir = 'agat_model'
graph_build_scheme_dir = 'dataset'
formula='NiCoFePdPt'

ha = HtAds(model_save_dir=model_save_dir, graph_build_scheme_dir=graph_build_scheme_dir)
ha.run(formula=formula)

Tips:

See API doc for more details. For example:

Manipulating agat.dataset:
AGAT molecular dynamics simulations:
More options for controlling the AGAT training process.