Installation

Requirements

Julia 1.10 or newer.

Norma's full-order solver, Schwarz coupling, and kernel/Operator-Inference reduced-order models are pure Julia and require no Python. Only the optional neural-network Operator-Inference model needs Python (see below).

Clone the repository

cd /path/to
git clone git@github.com:sandialabs/Norma.jl.git
cd Norma.jl
julia

Set up the environment

In the Julia package manager (press ] in the REPL):

pkg> activate .
pkg> registry update
pkg> update
pkg> instantiate

Press Backspace or Delete to leave the package manager.

Optional: neural-network reduced-order models

Only the neural-network Operator-Inference model needs Python (PyTorch via PyCall); it is an optional package extension. To enable it, add PyCall to the environment and make sure its Python has torch installed:

pkg> add PyCall

Julia then loads the backend automatically. Without it, requesting a neural-network model aborts with a message telling you to install PyCall. The reduced-order capabilities as a whole are documented with the norma-opinf companion package and are outside the scope of this guide.

Sandia network

If you are on Sandia's network and hit SSL/TLS certificate errors when installing or fetching packages, see README-sandia.md in the repository for a complete setup guide, and Troubleshooting.