Can you share codes of training and data augmentation?

[c747903646]

Hi,

I hope to reproduce the same training result by runing program on my machine. Can you share codes of training and data augmentation that were used to get the result reported in this paper?

[gasvn]

Please refer to the offical pytorch training code: https://github.com/pytorch/examples/tree/master/imagenet

[c747903646]

These 3 py files didn't include all model reported in paper, such as Res2NeXt-29, 8c×25w×4s. Can I implement this model by adding codes as below to the end of res2next.py? Is there anything else I should do?

def res2next29_8cx25wx4scale(pretrained=False, **kwargs): model = CifarRes2NeXt(Bottle2neckX, layers = [3, 3, 3], baseWidth = 25, cardinality=8, scale = 4, num_classes=100) return model

[gasvn]

The cifar-res2next is based on cifar-resnext. The implemention of cifar-resnext is different from ImageNet-resnext, as described in the resnext paper. And you can replace the bottleneck block in cifar-resnext with the res2next block.

[c747903646]

OK, thanks！