continue

• null (default) / weights, checkpoint_bestmodel.pth.tar

---

If the continue tag is set to weights, the training process restarts from the LAMMPS potential file(potential_saved). If the tag is set to checkpoint_bestmodel.pth.tar, the training process restarts from the checkpoint file. The usage of continue is as below.

#input.yaml
neural_network:
    continue: weights / checkpoint_bestmodel.pth.tar

Note

Users need to copy pca and scale_factor and potential files if you use LAMMPS potential(change the name of potential file into potential_saved).

Users need to copy checkpoint_bestmodel.pth.tar into your running directory if you use checkpoint file.