Installation Instructions
This package works on both Linux and MacOS (intel and arm64) platforms, and has so-far been tested using Python 3.9 - 3.11. There are two (plus one) methods to install the code.
Preliminaries
To enable GPU functionality for network training, make sure you have CUDA installed (or python 3.8+ if using apple silicon).
You will need some way to generate galaxy power spectrum multipoles to generate training sets. One option is to download and install both ps_1loop and ps_theory_calculator. You might need to request access to those repositories, in which case you can contact Yosuke Kobayashi (yosukekobayashi@arizona.edu). We have also included a version of FAST-PT to satisfy this requirnment.
From pip (recommended)
In a clean enviornment, simply run,
pip install mentat-lss
Alternatively, if you would like to install from source (for example, you want to add to the package, or would like easier access to the provided config files), you can do so in two different ways.
From source (automatic)
Download this repository to your location of choice.
In the base directory, simply run
install.shin the terminal. This script will create a new anaconda enviornment, fetch the corresponding version of PyTorch, and install the code, all automatically.
From source (manual)
1. Download this repository to your location of choice.
1. install the corresponding PyTorch version. If your machine doesn’t have a GPU, you can skip this step.
2. In the base directory, run python -m pip install ., which should install this repository as a package and all required dependencies.
To run the provided unit-tests, you can run the following command in the base repo directory,
python -m pytest tests