How to read an image created with the legacy torchtext?

I build plenty of models before I forced myself to upgrade the pytorch/torchtext. All the model images include also vocabs. I am saving them within Problem is that the old Vocab is now in torchtext.legacy. When I read an image with pytorch.load() then the vocab type is: <class ‘torchtext.vocab.Vocab’>. But the image is actually legacy Vocab and of course it does not work.
How can I get the legacy Vocab from the image so I can convert it to the new one?

Thanks for any hint