Upgrading torchvision module makes old model useless?

Once upon a time, I downloaded the vgg models by jonhson and loaded with torchvision.models.vgg16/19 without a problem.

Lately, I pip upgraded torchvision modules, and the same old vgg models fail to load, error goes like:
KeyError: 'unexpected key "classifier.1.weight" in state_dict'

Then I check stat_dict of current upgraded torchvision.models.vgg19, goes like this:
In [5]: v19.state_dict().keys()
Out[5]: odict_keys([‘features.0.weight’, ‘features.0.bias’, ‘features.2.weight’, ‘features.2.bias’, ‘features.5.weight’, ‘features.5.bias’, ‘features.7.weight’, ‘features.7.bias’, ‘features.10.weight’, ‘features.10.bias’, ‘features.12.weight’, ‘features.12.bias’, ‘features.14.weight’, ‘features.14.bias’, ‘features.16.weight’, ‘features.16.bias’, ‘features.19.weight’, ‘features.19.bias’, ‘features.21.weight’, ‘features.21.bias’, ‘features.23.weight’, ‘features.23.bias’, ‘features.25.weight’, ‘features.25.bias’, ‘features.28.weight’, ‘features.28.bias’, ‘features.30.weight’, ‘features.30.bias’, ‘features.32.weight’, ‘features.32.bias’, ‘features.34.weight’, ‘features.34.bias’, ‘classifier.0.weight’, ‘classifier.0.bias’, ‘classifier.3.weight’, ‘classifier.3.bias’, ‘classifier.6.weight’, ‘classifier.6.bias’])

Indeed, no such a key classifier.1.weight, so the new torchvision models made some changes to the stat_dict?

I’ve alerted @jcjohnson about this issue: https://github.com/jcjohnson/pytorch-vgg/issues/3

1 Like

Thanks for the clarification, smth!

Hi, @smth , could you please take a look on my issue? Any suggestions would be great!