Do you thinking re-writing this using the C++ tensorflow api will lead to a dramatic speed up?
Do you thinking re-writing this using the C++ tensorflow api will lead to a dramatic speed up?