Using vision.models with the CIFAR dataset?

[stsievert (Scott)]

I’ve looked around but haven’t found much, and have found that

• “they [all retrained models] are all pre-trained on Imagenet-12” Is there pretrained CNN (e.g. ResNet) for CIFAR-10 or CIFAR-100?
• Removing avgpool worked with resnet + cifar10: [Help] Debugging ResNet on CIFAR10
• Wideresnet works (but it’s under a different repo): Is there pretrained CNN (e.g. ResNet) for CIFAR-10 or CIFAR-100?

This makes me think think vision#140 and vision#155 are relevant.

Is there any way to use the model implemented in vision.models with CIFAR?

[smth]

The pre-trained models are all based on Imagenet-12. Most of the model definitions also need 224x224 as the minimum input size. For CIFAR, you are better off with other repos.

[stsievert (Scott)]

Thanks. I’ve got another repo working now:

xternalz/WideResNet-pytorch