I am trying to do a CNN project on digit recognization. But in the model_fit part, I got an unsupported operand types error.Here, x_train.shape is (336000,28,28,1) and batch