Lid-Driven Cavity (LDC) PINN

The Lid-Driven Cavity PINN is the canonical introductory example for physics-informed training in PhysicsNeMo v2.0. It solves the steady, incompressible 2D Navier–Stokes equations on a unit-square cavity with a moving top wall, using physicsnemo.sym.eq.phy_informer.PhysicsInformer to enforce the PDE residuals as soft constraints inside an explicit PyTorch training loop.

The complete, runnable example — including the model definition, the NavierStokes PDE class, the PhysicsInformer setup, the boundary condition treatment, the Hydra configuration, and a step-by-step walkthrough of the training loop — lives in the PhysicsNeMo repository and is the source of truth for this tutorial:

• examples/cfd/ldc_pinns/ on GitHub.

Run it with:

git clone https://github.com/NVIDIA/physicsnemo.git
cd physicsnemo/examples/cfd/ldc_pinns
pip install "nvidia-physicsnemo[sym]"
python train.py

For an inverse-problem counterpart that recovers unknown PDE coefficients from data (with the same PhysicsInformer machinery), see examples/cfd/inverse_pinns/.

If you are migrating an existing physicsnemo-sym LDC script, see the PhysicsNeMo Sym → physicsnemo.sym section of the PhysicsNeMo v2.0 Migration Guide.