Right now the optim module takes model.paramters() during construction. Is there a way to use optim when the model grows during training?
For example, if I am doing online learning and I would like to grow and simultaneously optimize my model, how would I go about doing so?