FitSNAP - v3.1.0

This release includes a refactor of the library API to be more Pythonic, allowing:

1. Multiple instances of FitSnap since ParallelTools is no longer a singleton.
2. No reliance on input scripts; each FitSnap instance may have its own dictionary of settings.
3. Separate and custom MPI communicators for each FitSnap instance.
4. Only need to import FitSnap, no more separate importing and object creation for Config and ParallelTools (this all happens under the hood for each instance).

Overall these changes allow seamless integration with other Python libraries like pyiron, JARVIS, and more.

Scientific Software - Peer-reviewed - Python
Published by rohskopf over 2 years ago

FitSNAP - v3.0.1

This release is the first new build version since the upgrade to FitSNAP 3. Main changes include:

• Refactor of the code to include a library interface and API.
• Nonlinear solvers for fitting neural network (NN) potentials, with different options for multi-element support.
• NNs with SNAP and ACE descriptors.
• Custom descriptors and pairwise NNs.
• Deploying NN models to LAMMPS.
• Streamlined documentation using the library API and Sphinx, deployed at https://fitsnap.github.io/

Scientific Software - Peer-reviewed - Python
Published by rohskopf almost 3 years ago