pyCARPool provides a custom class and functions to implement the Convergence Acceleration by Regression and Pooling (CARPool) method; this method exploits the correlation between simulations and surrogates to compute fast, reduced-variance statistics of large-scale structure observables without model error at the cost of only a few simulations. The method's estimates are unbiased, and achieve unbiased variance reduction factors of up to ∼10 without any further tuning. CARPool can also remove model error from ensembles of fast surrogates by combining them with a few high-accuracy simulations.

LtU-ILI (Learning the Universe Implicit Likelihood Inference) performs machine learning parameter inference. Given labeled training data or a stochastic simulator, the LtU-ILI piepline automatically trains state-of-the-art neural networks to learn the data-parameter relationship and produces robust, well-calibrated posterior inference. The package comes with a wide range of customizable complexity, including posterior-, likelihood-, and ratio-estimation methods for ILI, including sequential learning analogs, and various neural density estimators, including mixture density networks, conditional normalizing flows, and ResNet-like ratio classifiers. It offers fully-customizable, exotic embedding networks, including CNNs and Graph Neural Networks, and a unified interface for multiple ILI backends such as sbi, pydelfi, and lampe. LtU-ILI also handles multiple marginal and multivariate posterior coverage metrics, and offers Jupyter and command-line interfaces and a parallelizable configuration framework for efficient hyperparameter tuning and production runs.