Scimba
Scimba is a library that implements sciML methods and allows hybridization with more traditional numerical methods. The first block implements methods such as PINNs, neural Galerkin methods or neural operators for parametric elliptic, hyperbolic and kinetic PDEs on general geometries. It is coded in Pytorch. The numerical blocks, currently being written, (DG methods, particle methods) will be coded in a taichi differentiable framework compatible with Pytorch.
The public version is available here. A documation and tutorial will be available soon.
