How to load the saved_model pb graph as concrete functions

[lnshi]

With TF2 python tf.saved_model.load(export_dir, tags=None, options=None) API, i can load the saved_model as a collection of concrete functions, then i call them like conc_func(**input_tensors) according to different signatures, how can i achieve the same with rust APIs?

Or currently with rust APIs only the model_bundle.session.run(&mut run_args) option is available?

[Corallus-Caninus]

I do something similar, this may help

https://github.com/Corallus-Caninus/Brains/blob/main/src/lib.rs#L444

the example code also has loading and saving (where I originally got this).