lsstdesc-diffsky 1.1.1

Library for differentiable generation of synthetic skies for LSST DESC

lsstdesc-diffsky is a python library based on JAX for producing mock catalogs tailored to LSST DESC science requirements.

Installation

The latest release of lsstdesc-diffsky is available for installation with pip. To do so, we recommend starting from a base environment of jax managed by conda-forge:

$ conda create -c conda-forge -n diffsky_env python=3.9 numpy jax pytest ipython jupyter matplotlib scipy h5py
$ conda activate diffsky_env

To install the latest release:

$ pip install lsstdesc-diffsky

To install the current tip of the main branch, you will need to clone and install the source code:

$ cd /path/to/root/lsstdesc-diffsky
$ pip install .

Documentation

Online documentation for lsstdesc-diffsky is available at lsstdesc-diffsky.readthedocs.io

Testing

To run the suite of unit tests:

$ cd /path/to/root/lsstdesc-diffsky
$ pytest

To build html of test coverage:

$ pytest -v --cov --cov-report html
$ open htmlcov/index.html

Available Catalogs

A mock catalalog is available at NERSC and is called roman_rubin_2023_v1.1.0.

Working at NERSC

If you want to use the desc-python environment at NERSC, you will need to add lsstdesc-diffsky and its dependencies yourself. The directions for cloning and adding packages to the environment can be found here. Once you have built the basic environment, you will need to activate it and install lsstdesc-diffsky and its dependencies. Note that the appropriate version of the diffsky codes can vary with the mock being used.

If you wish to use the code in a jupyter notebook, you will also need to install a python kernel as follows:

$ python -m ipykernel install --user --name=my_kernel_name

demo_load_catalog is a demo notebook showing how to load the catalog from a notebook running at NERSC and perform some basic operations.