DeePMD-kit
Trains and deploys Deep Potential interatomic models from reference data, connecting machine-learning energy and force predictions to atomistic simulation engines.
Where it fits
Longer or larger simulations after a suitable potential has been trained and validated.
Limits and practical considerations
Training-domain coverage and backend compatibility limit transferability; extrapolation needs validation.
Licensing and access
LGPL-3.0-or-later code; pretrained model and training-data terms need separate review.