serialization - Example to save and load a sub graph in TensorFlow? -
i created model within variable/name scope , store graph definition , variables disk later load without defining graph again. how can save , load operations , variables within given variable/name scope?
conveniently, use tf.saver
. save()
has option store meta graph restore()
not seem import it. moreover, in saver constructor, can specify list of parameters, not name scope control operations saved.
there tf.train.write_graph()
couldn't find explanation of , how relates saver class , meta graph.
Comments
Post a Comment