Efficient Deep Neural Network Model Training By Reducing Memory And Compute Demands