About¶

torch-fem is a GPU-accelerated Finite Element Method (FEM) library built on PyTorch. By leveraging PyTorch’s automatic differentiation, it enables seamless sensitivity analysis and gradient-based optimization for structural mechanics, allowing you to treat FEM solvers as differentiable layers in your machine learning or optimization workflows.

Highlights

• ⚙️ Support for many element types with linear and quadratic interpolation
• 🧱 Built-in material models from linear elasticity to finite-strain hyperelasticity
• 📈 GPU acceleration via PyTorch (and optional CuPy support)
• 🔁 Differentiable solvers enabling end-to-end gradient-based optimization
• 📓 Rich examples - from basic cantilevers to topology optimization notebooks