pacemaker
pacemaker is a tool for fitting of interatomic potentials in a general nonlinear Atomic Cluster Expansion (ACE) form.
Documentation
- Installation
- Quick start
pacemaker
workflow- Input file
pacemaker
command line interface- Utilities
- Frequently asked questions (FAQ)
- What is a good value for batch_size?
- My fit on GPU crushes with OOM error, what to do ?
- Can I toggle between CPU and GPU when starting pacemaker?
- I dont have a GPU. Should I use backend, evaluator=
tensorpotential
orpyace
? - How to continue fitting ?
- I want to preserve the “shape” of potential, but refit it from scratch
- My potential behaves unphysical at short distances, how to fix it?
- How to split train/test data for the fitting?
- I want to change the cutoff, what should I do ?
- How better to organize my dataset files ?
- What are good values for regularization parameters ?
- How to fit only certain part of the potential, i.e. binary interaction only ?
- I see different metrics text files during the fit, what is it ?
- Optimization stops too early due to too small updates, but I want to run it longer…
- How to create a custom weights dataframe for ExternalWeightingPolicy?
- How to add more weights to certain structures ?
- How to run on-the-fly validation of the potential ?