bardisafa / eoal Goto Github PK

I found that in the argument parser there is the option of choosing the tinyimagenet dataset, however in the dataset.py there isn't any link of that.

Moreover the dictionary __factory do not have the tinyimagenet key:

__factory = {
    'cifar100': CIFAR100,
    'cifar10': CIFAR10
}

Maybe because the tinyimagenet is not implemented on torchvision.dataset, so how do you get and use it?

Thanks for the feedback!

How long does the training run take on Cifar100 dataset, with a mismatch rate of 20% and other options as default?
I took 20 hours up to 5 cycles with RTX3090. Is it right for it to take this long?

Hi,

Thank you for sharing your research.
I'm wondering if you used the official code for MQNet and LfOSA or reproduced it for your experiments.

If you reproduced it, can you share it?

best regards,