Vega

https://readthedocs.org/projects/lyafit/badge/?version=latest

Vega is a tool for computing 3D correlation function models for tracers used by the Ly-α forest group (such as Ly-α flux, Quasar positions or different metal lines) and for fitting data produced by picca primarily to measure Baryon Acoustic Oscillations (BAO).

Free software: GPL-3.0 License
Documentation: https://vega.readthedocs.io.
Referencing: If you use Vega in a publication please give the link to this repository (https://github.com/andreicuceu/vega). Right now there is no Vega paper. The best descriptions of what the code does are found in Cuceu et al. 2022 (https://arxiv.org/abs/2209.12931) and du Mas des Bourboux et al. 2020 (https://arxiv.org/abs/2007.08995).

Installation

We recommend to start by creating a fresh conda environment. The following code will create this and also install all the dependencies:

conda create --name vega pip ipython jupyter jupyterlab ipykernel numpy scipy astropy numba h5py setuptools "iminuit>=2.0.0" cachetools matplotlib
conda activate vega
pip install mcfit

You can either clone the public repository:

git clone https://github.com/andreicuceu/vega.git

Or download the tarball:

curl -OJL https://github.com/andreicuceu/Vega/tarball/master

Once you have a copy of the source, you can install it with:

cd vega
pip install -e .

If you are at NERSC and want your vega environment to show up as Jupyter kernel, you can run the following command:

python -m ipykernel install --user --name vega --display-name Vega

Both of the samplers and a few other modules in Vega need mpi4py. If you are at NERSC, you should install this using the NERSC-specific command:

MPICC="cc -shared" pip install --force-reinstall --no-cache-dir --no-binary=mpi4py mpi4py

Vega currently has interfaces for two samplers: Polychord and PocoMC. You do not need to install either of them to run the iminuit minimizer. Alternatively, if you only want to use one of the samplers, you only need to install that one (see instructions below).

Installing Polychord

Here are instructions for installing Polychord at NERSC. Note that this requires the default Perlmutter environment with no changes (except module load python). Start by following the steps above to install vega and its dendencies. After that clone Polychord:

git clone https://github.com/PolyChord/PolyChordLite.git
cd PolyChordLite

In the PolyChordLite folder, you will find a make file named "Makefile_gnu". You need to open and edit this file by changing lines 2-4 from:

FC = mpifort
CC = mpicc
CXX = mpicxx

FC = ftn
CC = CC
CXX = CC

After that, you can install PolyChord:

make veryclean
make COMPILER_TYPE=gnu
pip install -e .

You can test if PolyChord works by running the test script on an interactive node:

srun -n 2 python run_pypolychord.py

Finally, you should add this line to your .bashrc file, or at the beginning of your scripts (make sure to replace it with the correct path to your version of PolyChord):

export LD_LIBRARY_PATH=/path/to/PolyChordLite/lib:${LD_LIBRARY_PATH}

Installing PocoMC

Here are instructions for installing PocoMC at NERSC. First, install Pytorch in CPU mode (see this for more details):

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

Finally, install PocoMC:

pip install pocomc

Usage

Vega needs one "main.ini" file with the configuration, and at least one correlation config file. These correlation config files are generally of the form "lyaxlya.ini" for the Lyman alpha forest auto-correlation, or "qsoxlya.ini" for its cross-corelation with quasars. More complex cases also appear if we use the part of the Lyman alpha forest that appears left of the Lyman beta peak (i.e. in the Lyman beta part of the forest). These are generally called lyalyaxlyalyb.ini, which means we correlate Lya absorption in the Lya forest, denoted Lya(Lya), with Lya absorption in the Lyb part of the forest, denoted Lya(Lyb).

In the examples folder you can find examples of these config files with a lot of comments explaining what each option does. If you don't understand something, or we missed something, please open an issue.

Vega now has a Config Builder that is designed to create full Vega config files with minimal input. This is now the preffered way of interacting with Vega, as it automates fits and reduces the chance of mistakes. You can use the BuildConfig class interactively (e.g. in a notebook) as shown in this tutorial.

Using the terminal

You can call Vega from a terminal using the scripts in the bin folder, and pointing them to a "main.ini" file like this:

python run_vega.py path_to/main.ini

The "run_vega.py" script can be used for computing model correlations and for running the fitter. However, these can also be run interactively (see next section).

On the other hand the sampler (PolyChord) cannot be run interactively and needs to be called using the second script like this:

python run_vega_mpi.py path_to/main.ini

We strongly suggest you run the sampler in parallel on many cores, as normal run-times are of the order 10^2 - 10^4 core hours.

Interactive use

You can run Vega interactively using Ipython or a Jupyter notebook. This example notebook takes you through the steps of intializing Vega, computing a model and performing a fit.

This process is much more powerful compared to running in terminal as you directly have access to all the output, model components and fit results. Additionally, Vega was built in a modular structure with the aim of the user being able to call each module independently. Therefore, you have access to much more functionality this way. The documentation is the best source on how to run these modules independently, but if you can't find something there, please open an issue and we will try to help you and also improve the documentation.

Vega also has a FitResults module for analysing the results of a fit. You can find example usage of it in this notebook.

Credits

This package is based on picca fitter2 found here: https://github.com/igmhub/picca/tree/master/py/picca/fitter2, and was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.