The time required for training the model reflect the computation cost the model, which is also one consideration when applying deep learning models in practice. Therefore, training time is one indicator evaluating the model performance and influential factors have also been investigated in the experiments.