Does predictor.mobile_exporter (caffe2) actually do anything special for running on mobile?

Looking at the code of the Export function, I don’t see anything being done that distinguishes the model as something for mobile (as in android/ios). Is there any advantage to using this over Caffe2Backend.onnx_graph_to_caffe2_net to get the init_net and predict_net if one wants to run it on mobile?