very excellent job and I add something

[ju1987yetchung]

this programm file: "train_surrogates.py" is designed only recieve one channel data. for example, the shape should be (h,w,d).
and the model is designed to use 5D data, whose shape should be like (b,c,h,w,d), so, in the file".\neurop_invdes-main\fno_field_prediction\models_base.py", will add a dimension in position one, to make the shape be like (b,1,h,w,d).

so, if somebody want to train the data which already contains 2 or more channels, you should go to _base.py, and replace line 15 “y_pred = self(x.unsqueeze(1))” to be "y_pred = self(x)". And replace line 25 "y_pred = self(x.unsqueeze(1))" to "y_pred = self(x)" also.