Saving a SavedModel after loading a graph

[Luca Mcarthur]

Hi, is there a way to save the SavedModel after training having loaded the graph from a SavedModel (as opposed to initialising the layers & vars like in examples/xor.rs)?

It seems like the SavedModelBuilder requires a collection of variables & scope which doesn't seem straightforward to get just from an already existing graph (nor the meta graph needed in SavedModelSaver).

Btw repost from https://github.com/tensorflow/rust/issues/279#issue-722415545 (just found this community).

Cheers,

Luca