Thanks! So I studied the output of solve(....) (for the second case where I don’t define a RHS function I studied the output of NeuralODE(....)). The only difference that I found was that for interp, the cache is different as I have shown in the figure below. The variable in the top, pred_neuralnew is the output from NeuralODE and pred_ode is the output from solve for the ODEProblem. Is there a way to change it so that the cache is the same for both cases?
