v0.1.1 Synference Release

Small bump to version to ensure unique filename for PyPI. Otherwise identitcal to 0.1.0.

v0.1.0 Synference Release

First public Synference release, alongside paper (Harvey et al. 2025) submitted to MNRAS.

Synference is a Python package aimed to make Simulation-Based Inference (SBI) methods for galaxy SED fitting easier. SBI encompasses a family of related techniques that train a flexible neural network to learn the relationship between model parameters and observables, enabling e.g. Bayesian posterior estimation orders of magnitude more rapidly than traditional techniques.

Brief Overview of Features

1. Generation of synthethic library data using Synthesizer for arbitary forward models and observables. Stored in compressed HDF5 files. Also supports user-provided forward models e.g. from forward modelled hydro sims, CLOUDY outputs, radiative transfer codes etc.
2. Conversion of training library into survey-specific format with feature transformations to maximize model performance. Realistic noise modelling to match observations.
3. SBI model trainining using the LtU-ILI package, which implements best-practices for SBI. Model optimization is offered through Optuna.
4. Validation and diagnostic measurements for trained model, and checks for e.g. model misspecification.
5. Methods to apply trained model to observed catalogues, and recover predicted SEDs, SFHs, photometry etc.

Please raise issues in the repository if things don't work. There will still be some lingering issues given that this is the first release.

V0.1.0alpha Test Release

Inital Synference release to claim PyPI name. Full release notes will follow with full v0.1.0 release.