Problems in Active learning
Hello, I encountered several issues while using mlip-3 for active learning.
1.Are the steps for active learning in mlip-3 still similar to mlip-2? Can I still use case learning from mlip-2?
2.The documentation for mlip-3 mentions the LAMMPS input file "pair_style mlip load_from=curr.almtp extrapolation_control=true extrapolation_control:threshold_break=10 extrapolation_control:threshold_save=2 extrapolation_control:save_extrapolative_to=preselected.cfg", indicating that mlip.ini and state.als files are no longer required. Is that correct? "mlp calculate_grade curr.almtp train.cfg diff.cfg" is already a complete command, and there is no "out.cfg --als-filename=state.als" afterwards.
3.I encountered the issue "Rank 3, Breaking threshold exceeded (MV-grade: 26315.789474) on 3 processor" during LAMMPS training, but I'm still unsure how to handle it after reading your previous response. Could you provide some suggestions for resolving it?