Add possibility to share model with remote state_dict

[gmuraru]

Description

Currently, to share a model we need to have it locally and then call share - in the back we share the state_dict for the weight and the bias. We want to also be able to share the state_dict if it is on another machine without having it locally.

Are you interested in working on this improvement yourself?

• Yes, I am.

Additional Context

Take a look in module/__init__.py to check how the share is doing for the moment (the state_dict is local, but for this issue we need to check if we can share the state_dict if it is on another machine)

[gmuraru]

This is an stretch goal