This is the main directory of the new generation fitting code for the NNPDF collaboration. For a more technical description of the code, please consult the NNPDF documentation. The first release of NNPDF fits using this code corresponds to the NNPDF40 release.
It is recommended to utilize the officially released conda package which can be installed with:
conda install nnpdf -c https://packages.nnpdf.science/condaIf you are a developer please consult the guidelines before starting.
Runcards examples are provided in the runcards directory in this folder.
Specific runcard options are discussed here.
The basic command is:
n3fit runcard.yml <replica_number>For a full tutorial on how to run a fit using this code please see here Other relevant tutorials are the ones dedicated to fit analysis using validphys 2.
Please note that this repository is in active development and only releases tagged as such are guaranteed to produce correct results.
In order to ensure reproducibility all official NNPDF releases are run with a fixed version of not only the nnpdf fitting code but also
of all external packages that could potentially influence the results.
A list of conda environments with the exact versions of the nnpdf fitting code as well as all external packages used for official releases
of the code can be found in the environments directory.
The papers describing the implementation, most significant developments and release of the code are:
- Towards a new generation of parton densities with deep learning models S. Carrazza, J. Cruz-Martinez, Eur.Phys.J.C 79 (2019) 8, 676
- NNPDF40 paper NNPDF Collaboration Eur.Phys.J.C 82 (2022) 5, 428