SM-Net: A Learned Continuous Spectral Manifold from Multiple Stellar Libraries

This record contains the trained SM-Net “lookup” model weights used by the SM-Net local Dash/Plotly web interface for rapid synthetic stellar spectrum generation from (Teff,log⁡g,log⁡Z). The weights correspond to models trained on a combined grid spanning the PHOENIX–Husser, C3K–Conroy, TMAP–Werner, and OB–PoWR libraries. These files are intended to be used with the open-source SM-Net repository (installation instructions and the local web interface are provided there):https://github.com/ICRAR/SM_Net The repository ships lightweight metadata caches (*.meta.npz) for parameter bounds and wavelength grids; this Zenodo deposit provides the large binary weight archives required to run the models.DOI: 10.5281/zenodo.18883385 Recommended citation: cite this Zenodo record for the weights and the GitHub repository for the code and interface.