Torch Geopooling is an innovative extension for PyTorch that introduces geospatial pooling modules, enhancing the capability to build geospatial neural networks. By providing both adaptive and regular pooling modules, Torch Geopooling supports decomposition of 2D spatial data, crucial for geospatial applications. Adaptive modules dynamically adjust during training, while regular modules use precomputed decompositions. Installation is straightforward via PyPI, and the library integrates seamlessly with PyTorch's automatic gradient computation.