scverse/scvi-tools

Built on PyTorch and AnnData, scvi-tools provides modular probabilistic models for tasks like dimensionality reduction, data integration, automated annotation, doublet detection, and spatial deconvolution across single-cell and spatial omics datasets. The framework leverages PyTorch Lightning and Pyro as foundational building blocks, enabling researchers to rapidly develop and deploy custom probabilistic models with GPU acceleration and Scanpy interoperability. High-level APIs include standard save/load functions and integrate seamlessly into the broader scverse ecosystem for omics analysis.

1,582 stars and 147,559 monthly downloads. Used by 5 other packages. Actively maintained with 11 commits in the last 30 days. Available on PyPI.