First, I generated training data by solving the Heat equation with a spectral method in heat1d_data.ipynb. The dataset is stored in heat1d_dataset.npy.
Then, I trained a Fourier Neural Operator on the data in train_heat1d.ipynb and saved the model in heat1d.eqx.
Finally, I generate a random scenario in eval_heat1d.ipynb and evaluate it using the model. I demonstrate zero-shot superresolution by evaluating the model on a 1000-point mesh; it was trained on a 100-point mesh.