Fail to untar the model for tiredImagenet

[ucasligang]

Thanks for your promising work. There is an issue, I fail to untar the model for tiredImagenet because of the file is broken. Could you check it out, please?

[nupurkmr9]

For the tieredImagenet model untar is not required. You should be able to directly do torch.load(). Can you check that.

[ucasligang]

Thank you for you response. I have another little problem, the pretrained tieredImagenet model is first train a model with rotation self-supervision and finetune the model using both manifold mixup and rotation self-supervision, isn't it ?

[ucasligang]

If I only use the pretrained tieredImagenet model to extract the feature of tieredImagenet, I didn't need to fine-tune, is right?

[nupurkmr9]

yes, the model is trained by S2M2 training technique.

[ucasligang]

Thank you very much, we'll cite your article if our work gets published.