In the PyTorch 1.6 release note I saw that the old pickle-based serialization format was replaced by a new zipfile-based format.
I want to save the entire PyTorch model and NOT the state_dict for deployment reasons. “The disadvantage of this approach is that the serialized data is bound to the specific classes and the exact directory structure used when the model is saved. The reason for this is because pickle does not save the model class itself.” (Link). My question: does the new format eliminate this disadvantage? Is it saving the model class or not?