Ability to load in a distribution

Hi, I am fairly new to Pyro and am looking for a feature to do the following.

Say I have a NN.
I create a model, and a guide, and begin training my distribution using SVI.step().

After 50 iterations, I receive some new information that gives me a distribution over a particular weight in my NN. How can I load in this distribution?, and what do I have to reinitialize when doing so?