Model train with Cifar10

This topic is made to ask for help. I am running some experiments under the directory /mxnet/example/image-classification, but I can not find a network which can lead to high train accuracy and high validation accuracy. Even though some models under the /mxnet/example/gluon can achieve this goal, but the dcasgd optimizer cannot be applied (I do not know why). Does anyone can provide me some models that run under mxnet/example/image-classification using cifar10 and achieve both high training accuracy and validation accuracy. Thank you very much.