Compute pace update

[jmgoff]

Summary

With the growing number of atomic cluster expansion (ACE) users, a LAMMPS compute is needed at this time that can evaluate the descriptors for various methods and workflows such as FitSNAP and ML-IAP. This update features a new LAMMPS compute, compute pace, that evaluates ACE descriptors. The style mirrors the compute for SNAP descriptors, compute snap, and provides users a familiar method to evaluate ACE descriptors in LAMMPS for data analysis, for model training, etc.

Related Issue(s)

N/A

Author(s)
(Alphabetical)

Aidan Thompson (Sandia National Laboratories)
Andrew Rohskopf (Sandia National Laboratories)
James Goff (Sandia National Laboratories) ( goffjimmy1@gmail.com )
Yury Lysogorskiy (Ruhr University for lammps-user-pace backend)

Licensing

By submitting this pull request, I agree, that my contribution will be included in LAMMPS and redistributed under either the GNU General Public License version 2 (GPL v2) or the GNU Lesser General Public License version 2.1 (LGPL v2.1).

Backward Compatibility
No problems with old inputs.

Implementation Notes

The calculation of ACE descriptors and descriptor gradients were validated numerically against those from the existing pair_style pace in LAMMPS. They have also been validated against other ACE software that is not in LAMMPS.

This initial implementation of compute pace requires minor changes to the lammps-user-pace library that is downloaded whenever the ML-PACE D flag is turned on for compilation. The required changes to the lammps-user-pace library are essentially just allocating arrays to store ACE descriptors and gradients. Until such changes are made in the default lamps-user-pace repo at ICAMS, some errors may be thrown unless a separate fork of the lammps-user-pace library is used. Future updates and integration with the (external ACE compute)[https://github.com/ICAMS/python-ace/issues/14] developed by @yury-lysogorskiy at ICAMS would resolve this as well.

This first compute pace is here to provide users with the ACE descriptor compute capabilities immediately and serves as the base framework to integrate more standardized descriptor + gradient calculators in the future.

Post Submission Checklist

• The feature or features in this pull request is complete
• Licensing information is complete
• Corresponding author information is complete
• The source code follows the LAMMPS formatting guidelines
• Suitable new documentation files and/or updates to the existing docs are included
• The added/updated documentation is integrated and tested with the documentation build system
• The feature has been verified to work with the conventional build system
• The feature has been verified to work with the CMake based build system
• Suitable tests have been added to the unittest tree.
• A package specific README file has been included or updated
• One or more example input decks are included

Further Information, Files, and Links

Link to the original ICAMS PACE libraries and those including an updated descriptor + gradient calculator: https://github.com/ICAMS/lammps-user-pace and ICAMS/python-ace#14
Link to modified lammps-user-pace: https://github.com/jmgoff/lammps-user-pace-1

[akohlmey]

I've added a comment for a required change to the compute_pace.h file.
This code also seems to be dependent on a different version of the ACE library since our CI testing fails to compile with errors indicating missing definitions. See, e.g. https://ci.lammps.org/job/dev/job/pull_requests/job/fedora/job/cmake_kokkos_mpi_openmp_clang_shared/7591/console
Thus the corresponding build files need to be updated to download a suitable version.

[stanmoore1]

Overall this looks good, but needs a little clean up as I mentioned in some of the comments.

[akohlmey]

@jmgoff I've merged with upstream and implemented requested changes (and a couple of mostly cosmetic updates, too).

Please check it out and let us know if this is now a final version or still waiting for additional modifications.

[jmgoff]

@akohlmey Thank you for making those changes and updates requested by @stanmoore1 (and for your patience with my slow updates on this)! It looks good to me and is the final version for this PR.