paltas is a package for conducting simulation-based inference on strong gravitational lensing images. The package builds on lenstronomy to create large datasets of strong lensing images with realistic low-mass halos, Hubble Space Telescope (HST) observational effects, and galaxy light from HST's COSMOS field. paltas also includes the capability to easily train neural posterior estimators of the parameters of the lensing system and to run hierarchical inference on test populations.
paltas is installable via pip:
$ pip install paltasThe default paltas requirements do not include tensorflow, but if you are interested in using the modules contained in the Analysis folder, you will have to install tensorflow:
$ pip install tensorflowThe main functionality of paltas is to generate realistic datasets of strong gravitational lenses in a way that's modular, scalable, and user-friendly. To make a dataset with paltas all you need is a configuration file which you can then pass to the generate.py script:
$ python generate.py path/to/config/file path/to/output/folder --n 100Running the line of code above would generate 100 lenses and output them in the specified folder. paltas comes preloaded with a number of configuration files which are described in Configs/Examples/README.rst. For example, to create a dataset with HST observational effects, subhalos, and line-of-sight halos run:
$ python generate.py Configs/Examples/config_all.py example --n 100We provide a tutorial notebook that describes how to generate your own config file.
paltas comes with a tutorial notebook for users interested in modifying the simulation classes.
- Implement your own source, line-of-sight, subhalo, or main deflector model.
- Training a neural posterior estimator of simulation parameters.
- Running hierarchical inference on a population of strong lenses.
Code for generating the plots included in some of the publications using paltas can be found under the corresponding arxiv number in the notebooks/papers/ folder.
We welcome improvements via GitHub pull request! The workflow for contributing to paltas is the same as it is for lenstronomy, please see the lenstronomy CONTRIBUTING guidelines. for the rundown.
In brief, to contribute to paltas you will need to fork the repo on GitHub, clone it, and pip install with editable option, like this:
$ git clone git@github.com:<your-account>/paltas.git
$ cd paltas
$ pip install -e .You should then work in a local development branch, while keeping your main branch up to date with the base repo on GitHub. (That means you'll need to git remote add base https://github.com/swagnercarena/paltas.git so that you can git pull base main from your main branch).
See the lenstronomy CONTRIBUTING guidelines. for help with python style, docstring formats, creating pull requests, and so on. We recommend that you start your pull request (PR) when you _begin work on a new feature or bug fix, so that the rest of the paltas community can see what you are doing and coordinate with you - when your improvement is ready you can request code review from others via the PR interface. Try and restrict PRs to one feature per branch, to help keep them small and manageable. (Note that paltas uses Google docstring style.)
If you use paltas or its datasets for your own research, please cite the paltas package (Wagner-Carena et al. 2022) as well as the lenstronomy package (Birrer & Amara 2018, Birrer et al. 2021).
The following is a list of the zenodo uploads associated to papers using paltas. These uploads will include additional chains, test sets, and model weights required to reproduce the paper results.
