How can I contribute pretrained models?

[forjiuzhou]

I have trained ResNet18 with 73.01 accuracy without tricks like label smoothing, randaugmentation or random erasing. The models is trained only with spherical knowledge distillation (arixiv: 2010.07485) and basic data augmentation, and with 100 epochs.
I have confidence to achive more accuracy models with more epochs and more training technologys. So is there a way to share the pretrained weights in this repo?
Thanks.

PS: I notice this repo trained very effiecient models compared to pytorch official release. But I'm not able to find a document elaborating the training details. I'm curious how these models are trained.