spectracles 0.6.4

Unified spectrospatial models: glasses for your spectra.

Unified spectrospatial models for integral field spectroscopy in JAX

Glasses for your spectra

Spectracles is a Python library for inferring properties of IFU/IFS spectra as continuous functions of sky position.

It can also be used as a general-purpose statistical model library that extends equinox to allow for composable models that may have coupled parameters. It also implements some other nice features that are a bit awkward in equinox out of the box, like easily updating model parameters between fixed and varying.

Installation

From PyPI with pip:

pip install spectracles

Or with uv (recommended):

uv add spectracles

From source:

git clone [e-mail]:TomHilder/spectracles.git
cd spectracles
pip install -e .

Note: fftw must be installed or the dependency jax-finufft will fail to build.

Features

• Parameter sharing - Couple parameters across model components
• Declarative optimization schedules - Specify which parameters are free/fixed per phase
• Glob patterns - Use wildcards like "gp.kernel.*" to match parameters
• JAX integration - Built on equinox, fully compatible with JAX transformations
• Rich output - Pretty-printed model trees and gradient diagnostics

Documentation

Full documentation: tomhilder.github.io/spectracles

Citation

Coming soon.

License

MIT