Great Work! Are there any pre-trained models available? If so, could you please share some of them, especially on CIFAR and ImageNet?