DeepXDE

DeepXDE is a mature Physics-Informed Neural Network (PINN) framework known for usability; its official docs and examples are very friendly to Chinese-speaking users, and it supports TensorFlow, PyTorch and other backends. Its declarative API describes PDEs and boundary conditions, ideal for quickly validating physics-constrained learning.

Use cases: PINN teaching; small/medium inverse and forward PDE problems; research prototyping.