Start / Install

Install¶

binference needs Python 3.13 or later. Its core depends on NumPy (2.1 or later), SciPy (1.14.1 or later) and threadpoolctl. threadpoolctl stops parallel chains from starting more threads than there are cores.

binference is in alpha. Install it from PyPI, with the plotting extra:

pip install "binference[plot]"

For the development version, install from GitHub instead:

pip install "binference[plot] @ git+https://github.com/joshspeagle/binference"

Leave out [plot] for the core alone, or name several extras, such as [plot,xarray].

Extras¶

Extra	Adds	For
`plot`	matplotlib	`result.plot()` and the functions of `binference.plotting`
`xarray`	xarray	`result.to_datatree()`, an ArviZ-compatible export
`all`	both of the above

To work on binference itself, follow the set-up in the contributing guide.

Check the installation¶

This prints the installed version, then fits a simulated catalogue of 500 objects on 10 bins and prints the fitted population fractions.

import binference

print(binference.__version__)
mock = binference.simulate.mock_1d(500, binference.Bins.linear(0, 3, 10), rng=0)
print(binference.infer(mock.likelihoods, method="map").rho.round(3))

Next, the Quick start takes you from your own arrays to a population.