[FEATURE] Instantiate PyTorch models from .nam files

[sdatkinson]

In order to upgrade old .nam models with loudness metadata, we need to be able to instantiate them as PyTorch models. so that we can run the method on them.

This Issue is for instantiating models from .nam files

[dimentorium]

sounds good, lets keep the discussion here ;-)
So do you think this will be handled by the pytorch api or is there a special wavenet api?
Might this article here help? https://pytorch.org/tutorials/beginner/saving_loading_models.html#what-is-a-state-dict