Fourier Neural Operator is a family of resolution invariant network architectures that use spectral convolutions to learn mappings between function spaces. The model is described in the paper here by Li et.al.
The pretrained model checkpoint comes from the Darcy flow example which trains an Adaptive Fourier neural operator (AFNO) model as described in the Modulus documentation.
The model was trained on Darcy flow dataset. For more information please refer the Modulus documentation.
You can use the inference script to run the model and vary the parameters of the design. You can use the training script as a starting point to train for a different geometry. Refer Modulus Documentation for more details.
The model accepts pearmeability field on a cartesian grid. Since the model is resolution invariant, the cartesian grid can have a resolution different that the training resolution.
The model outputs pressure field on the same resolution as the input permeability.