3. Run FitSNAP

If you want to get started immediately with interactive examples, please see our Colab Python notebook tutorial

There are two ways to run FitSNAP: (1) the python executable or (2) as a python library. The executable version is reserved for the most basic operation of fitting on a dataset, where data is scraped from a directory, configurations are input to LAMMPS to calculate descriptors, and the machine learning problem is solved. The python library provides more flexibility and allows one to modify the steps in that fitting process.

Before using the executable, you must set the path to your FitSNAP directoy in your PYTHONPATH environment variable, e.g.

export PYTHONPATH="path/to/FitSNAP:$PYTHONPATH"

which is conveniently placed in a ~/.bashrc or ~/.bash_profile file. Then fits can be performed with the executable by doing

python -m fitsnap3 input.in --option

where input.in is a FitSNAP input file and option is an acceptable command line option. Options and input files are explained below.

• 3.1. Command-line options
• 3.2. Input files
• 3.3. FitSNAP Outputs
• 3.4. Running as library