comb_graph.py

        with tf.gfile.GFile(modelfilename, 'rb') as f:
            graph_def = tf.GraphDef()
            graph_def.ParseFromString(f.read())
            self.graph_def = graph_def

        with tf.Graph().as_default() as g1:
            #gi_name = model_name+'/'+gen_input
            gi = tf.placeholder(tf.float32,
                              [None, self.z_dim],
                              name='g1_gi')
            go, gl = tf.import_graph_def(self.graph_def,
                                input_map={
                                    gen_input: gi
                                },
                                return_elements=[gen_output, gen_loss],
                                op_dict=None,
                                producer_op_list=None,
                                name=model_name+'g1')
            self.image_shape = go.shape[1:].as_list()

        with tf.Graph().as_default() as g2:
            #di_name = model_name+'/'+disc_input
            di = tf.placeholder(tf.float32,
                              [None] + self.image_shape,
                              name='g2_di')
            do, = tf.import_graph_def(self.graph_def,
                                input_map={
                                    disc_input: di
                                },
                                return_elements=[disc_output],
                                op_dict=None,
                                producer_op_list=None,
                                name=model_name+'g2')

        gdef_1 = g1.as_graph_def()
        gdef_2 = g2.as_graph_def()


        with tf.Graph().as_default() as g_combined: #merge together
            self.gi = tf.placeholder(tf.float32,
                      [None, self.z_dim],
                      name='comb_di')
            self.di = tf.placeholder(tf.float32,
                  [None] + self.image_shape,
                  name='comb_di')
            
            tf.import_graph_def(gdef_1, input_map={'g1_gi': self.gi}, name='g1_comb')
            print([n.name for n in tf.get_default_graph().as_graph_def().node if 'Tanh' in n.name])
            self.go = g_combined.get_tensor_by_name('g1_comb/' + model_name + 'g1' +'/'+ gen_output)


            tf.import_graph_def(gdef_2, input_map={'g2_di': self.di}, name='g2_comb')
            self.do = g_combined.get_tensor_by_name('g2_comb/' + model_name + 'g2' +'/'+ disc_output)
            
            self.graph = g_combined