ForestFlow

Lyman-alpha Cosmology Emulator. This code is a normalising flow emulator for the 3D flux power spectrum of the Lyman-alpha forest.

Documentation

The documentation includes the user guide and complete API reference generated from NumPy-style docstrings.

• Online documentation
• Documentation source

To build the documentation locally:

python -m pip install -e ".[docs]"
make docs

Open docs/_build/html/index.html after the build completes. Read the Docs uses the repository's .readthedocs.yaml file to perform the same build online.

Emulator parameters:

These are the parameters that describe each individual P3D(k, mu) power spectrum. We have detached these from redshift and traditional cosmology parameters.

Cosmological parameters:

Delta2_p is the amplitude of the (dimensionless) linear spectrum at k_p = 0.7 1/Mpc

n_p is the slope of the linear power spectrum at k_p

IGM parameters:

mF is the mean transmitted flux fraction in the box (mean flux)

sigT_Mpc is the thermal broadening scale in comoving units, computed from T_0 in the temperature-density relation

gamma is the slope of the temperature-density relation

kF_Mpc is the filtering length (or pressure smoothing scale) in inverse comoving units

Tutorials:

In the Notebooks folder, there are several tutorials one can run to learn how to use the emulators and archives.

• Archive tutorial: notebooks/Tutorial_archive.ipynb
• Emulator tutorial: notebooks/Tutorial_emulator.ipynb

Installation

ForestFlow requires Python 3.12 or newer. We recommend installing it in a dedicated environment:

conda create -n forestflow python=3.12
conda activate forestflow

ForestFlow uses LaCE for cosmology and simulation archive support. Install LaCE by following its installation instructions, then clone and install ForestFlow:

git clone https://github.com/igmhub/ForestFlow.git
cd ForestFlow
python -m pip install -e .

The editable installation is recommended for development. To install the documentation and testing tools as well, use:

python -m pip install -e ".[docs,test]"

Optional features

Install the dependencies required by the cross-power routines with:

python -m pip install -e ".[px]"

To generate notebooks from the Jupytext sources:

python -m pip install jupytext
jupytext --to ipynb notebooks/*/*.py

To expose the environment as a Jupyter kernel:

python -m pip install ipykernel
python -m ipykernel install --user --name forestflow --display-name forestflow