paltax is a package for conducting simulation-based inference on strong gravitational lensing images.

paltax is installable via pip:

$ pip install paltax

For the most up-to-date version of paltax install directly from the git repository.

$ git clone https://github.com/swagnercarena/paltax.git
$ cd path/to/paltax/
$ pip install -e .

The main functionality of paltax is to train (sequential) neural posterior estimators with on-the-fly data generation. To train a model with paltax you need a training configuration file that is passed to main.py:

$ python main.py --workdir=path/to/model/output/folder --config=path/to/training/configuration

paltax comes preloaded with a number of training configuration files which are described in paltax/TrainConfigs/README.rst. These training configuration files require input configuration files, examples of which can be found in paltax/InputConfigs/.

paltax comes with a tutorial notebook for users interested in using the package.

• Using an input configuration file to generate a batch of images.

Code for generating the plots included in some of the publications using paltax can be found under the corresponding arxiv number in the paltax/notebooks/papers/ folder.

If you use paltax for your own research, please cite the paltax package (Wagner-Carena et al. 2024)

paltax builds off of the publically released Google DeepMind codebase jaxstronomy.