While most numerical solvers focus on moving forward, sometimes you need to take a step back. Have you ever faced challenges like these?
- Solving Lagrangian inverse problems without an adjoint solver?
- Performing shape optimization, but your boundaries are non-differentiable?
- Building closure models when your SPH solver lacks gradients?
If you answered yes to any of these or similar questions, we’re excited to announce the first public release of Rene’s DiffSPH, our differentiable Smoothed Particle Hydrodynamics solver. DiffSPH can tackle all these challenges and more. It supports compressible, weakly-compressible, and incompressible fluids, all while being fully differentiable and written in PyTorch.
- Explore our repository: https://diffSPH.fluids.dev
- Check out a short live demo of a 1D Sod Shock Tube simulation in Colab
