How does one store and restore models for AutoML?

I saw this warning:

warnings.warn("pickle support for Storage will be removed in 1.5. Use `torch.save` instead", FutureWarning)

that came from pytorch.

I am generating models on the fly so I do not have a class for them as the recommended suggest (https://pytorch.org/docs/master/notes/serialization.html#recommend-saving-models).

If my pickled code won’t work later, how am I suppose to save the model that is agnostic to model class code? (i.e. I do NOT have a way to do:

the_model = TheModelClass(*args, **kwargs)

)

Pickling is really a fantastic solution for my case since I just do:

child_model = db['child_model']

Can we keep the pickling code avaible for us please?


gitissue: https://github.com/pytorch/pytorch/issues/41341