The catalogue

Everything in Syndex is a dataset: a grid, an instrument, or some other named piece of data. Each dataset has one or more releases, the versions of its file that have been published.

Kinds of data

Most of the catalogue is one of three kinds, each with a set file layout that Syndex knows how to read and check:

Grids

Spectra, lines and ionising luminosities from SPS and AGN models, over a grid of parameters.

Grids

Dust grids

Dust attenuation curves or dust emission spectra, over a grid of parameters.

Dust grids

Instruments

Saved Synthesizer instruments: filters or wavelengths, with resolution, PSFs and noise where they have them.

Instruments

The rest supports Synthesizer’s tests and tools. It comes in whatever format its producer wrote, so a reviewer sorts it rather than an automatic check:

simulation_data

Cut-down simulation output for Synthesizer’s tests and examples.

generation_data

The raw model inputs grids were built from, kept so they can be rebuilt.

synference_data

Inputs for Synference, the simulation-based inference toolkit built on Synthesizer.

reference_data

Reference outputs that regression tests compare against.

cache

Archived copies of external services, such as SVO filter curves, so tests do not depend on them being up.

Names

Every dataset has two names. The catalogue name is the one you download it by, such as bpass-2p2p1-bin-chabrier03-0p1-300p0. It uses lowercase letters, digits and hyphens, with p in place of a decimal point, and runs from the most general part to the most specific. The display name is the readable title the portal shows.

Important

A catalogue name never changes once it is published, because scripts and papers refer to it. Each kind’s page shows how names are built for it.

Releases

A new version of a dataset is published as a new release under the same name. Nothing is overwritten: older releases keep their files and stay downloadable with --release.

Markers

Some datasets carry a marker in the portal:

known bug

A problem has been found in the current release since it was published. The dataset’s page explains it.

recommended

The maintainers’ default choice for this kind of data.

reduced

A small, cut-down copy for tests and examples. Not for science.

CI

Downloaded by Synthesizer’s own test suite.