This fitter is designed to analyze correlation functions generated from lattice QFT calculations. This fitting package is currently undergoing rapid development, and there is no promise of backwards compatibility yet. Version numbers will be used to support reproducibility.
This package requires the lsqfit and gvar libraries developed by Peter Lepage.
This package is now locally pip-installable in editable mode (which means, you can actively develop the code and re-install quickly). Tested locally with a clean anaconda installation, the following works for installing on a mac OS at least.
[skip these steps unless you want to test a "clean" install]
conda create -n bare3.8 python=3.8
conda activate bare3.8
[rest of the install]
git clone https://github.com/callat-qcd/c51_corr_fitter
cd c51_corr_fitter
pip install -e .
This will result in an installation that claims there are pip install errors, but in practice, the installation works. To test success, type
which fit_twopt
and if this returns a binary, the installation has worked. If you do not have an anaconda install, or virtual env, you can add "--user" at the end of the pip install to install the binary in the $HOME/.local location to keep it isolated from the rest of your python site-packages, if desired.
When the code is updated, to re-install the new features, one can simple go to the root source dir and re-install
cd <path_to_repo>/c51_corr_fitter
pip install -e .
To build up a fit, one usually looks at effective mass plots etc., and starts to guess the input ground state energy and overlap factors. With a working installation, you should be able (from the source directory) do
cd tests
fit_twopt input_file/callat_a12m220XL_test_params.py --eff --fit
which will generate effective mass plots and perform the fit of the states specified in the input file.