torchflows and pytorch-flows
These are competitors offering overlapping implementations of normalizing flow algorithms in PyTorch, where the choice between them depends on whether you prioritize modern design and active maintenance (torchflows) versus established algorithm coverage (pytorch-flows).
About torchflows
davidnabergoj/torchflows
Modern normalizing flows in Python. Simple to use and easily extensible.
This library helps machine learning researchers and practitioners train generative models and estimate data density using modern normalizing flows. You provide your dataset, and it outputs a model that can generate new, similar data points or calculate the likelihood of existing ones. It's designed for those working with advanced machine learning models who need flexible tools for generative tasks.
About pytorch-flows
ikostrikov/pytorch-flows
PyTorch implementations of algorithms for density estimation
This project helps researchers and data scientists analyze complex, high-dimensional data to understand the underlying probability distribution. It takes raw numerical datasets, like those found in physics or genetics, and produces a model that can estimate the likelihood of specific data points or generate new, realistic data samples. It's for anyone needing to model complex data distributions without making strong assumptions about their shape.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work