- What is the recommended hardware for training using PhysicsNeMo framework?
- What model architectures are in PhysicsNeMo?
- How do I use physics-informed training with PhysicsNeMo?
- What can I do if I dont see a PDE in PhysicsNeMo?
- What is the difference between the pip install and the container?
Please refer to the recommended hardware section: System Requirements
Nvidia PhysicsNeMo is built on top of PyTorch and you can build and train any model architecture you want in PhysicsNeMo. PhysicsNeMo however has a catalog of models that have been packaged in a configurable form to make it easy to retrain with new data or certain config parameters. Examples include GNNs like MeshGraphNet or Neural Operators like FNO. PhysicsNeMo samples have more models that illustrate how a specific approach with a specific model architecture can be applied to a specific problem. These are reference starting points for users to get started.
You can find the list of built in model architectures here.
PhysicsNeMo includes a physicsnemo.sym module (install with
pip install "nvidia-physicsnemo[sym]") that provides symbolic PDE definition,
automatic spatial derivative computation, and physics-informed residual evaluation.
Define your equations using SymPy, then use PhysicsInformer to compute PDE
residuals automatically.
See the LDC PINNs example and the Darcy physics-informed example for complete training scripts.
Note: The separate PhysicsNeMo-Sym repository is being archived. Its core functionality has been upstreamed into PhysicsNeMo. See the migration guide for details.
Define your PDE using SymPy and the physicsnemo.sym.eq.pde.PDE base class.
See the LDC PINNs example for an inline
Navier-Stokes definition, or the
MHD PINO example for a custom MHD PDE.
There is no functional difference between the two. This is to simplify the ease of installing and setting up the PhysicsNeMo environment. Please refer to the getting started guide on how to install using Pip or using the container.