I
Research workflow
From material model to inversion loop.
Start here to build an acquisition, choose a boundary and storage policy, inspect receiver data, and optimize ε or σ.
Differentiable electromagnetics, expressed directly
Maxwell modeling as a computable operator.
TIDE connects FDTD propagation, model-space derivatives, and inversion workflows through one explicit PyTorch API for 2D TM and full 3D fields.
One physics model, four derivative views
Read the library as an operator.
The public API follows the mathematics. Move across the operators to see what changes, what stays fixed, and where each path is documented.
InputEM model + experiment
Propagate one electromagnetic model into receiver data.
Returnsd
Two reading layers
Enter through the question you have.
II
API and internals
From public contract to native kernel.
Use this layer to inspect named results, backend capabilities, derivative sessions, storage ownership, and C/CUDA implementation notes.
Before scale