Connect input and output tensors of two different graphs. Flow tensor

Question

I have 2 ProtoBuf Files that I currently upload and forward each one separately, calling -

out1=session.run(graph1out, feed_dict={graph1inp:inp1})

followed by

final=session.run(graph2out, feed_dict={graph2inp:out1})

where graph1inp and graph1out are introduced by node and the node of graph 1 and similar terminology for graph 2 are displayed

Now I want to connect graph1out with graph2inp , so I need to run graph2out while you feed graph1inp > with inp1 . In other words, connecting the input and output tensors from the two involved graphs in such a way that one run is enough to perform the output in both ProtoBuf trained files.

Answer 1

Assuming your Protobuf files contain serialized tf.GraphDefprotos, you can use the argument input_map tf.import_graph_def()to connect two graphs:

# Import graph1.
graph1_def = ...  # tf.GraphDef object
out1_name = "..."  # name of the graph1out tensor in graph1_def.
graph1out, = tf.import_graph_def(graph1_def, return_elements=[out_name])

# Import graph2 and connect it to graph1.
graph2_def = ...  # tf.GraphDef object
inp2_name = "..."  # name of the graph2inp tensor in graph2_def.
out2_name = "..."  # name of the graph2out tensor in graph2_def.
graph2out, = tf.import_graph_def(graph2_def, input_map={inp2_name: graph1out},
                                 return_elements=[out2_name])

Answer 2

, , . :

import tensorflow as tf
from tensorflow.python.framework import meta_graph

with tf.Graph().as_default() as graph1:
    input = tf.placeholder(tf.float32, (None, 20), name='input')
    ...
    output = tf.identity(input, name='output')

with tf.Graph().as_default() as graph2:
    input = tf.placeholder(tf.float32, (None, 20), name='input')
    ...
    output = tf.identity(input, name='output')

graph = tf.get_default_graph()
x = tf.placeholder(tf.float32, (None, 20), name='input')

tf.train.export_meta_graph, CollectionDef meta_graph.import_scoped_meta_graph . , , input_map.

meta_graph1 = tf.train.export_meta_graph(graph=graph1)
meta_graph.import_scoped_meta_graph(meta_graph1, input_map={'input': x}), import_scope='graph1',
out1 = graph.get_tensor_by_name('graph1/output:0')

meta_graph2 = tf.train.export_meta_graph(graph=graph2)
meta_graph.import_scoped_meta_graph(meta_graph2, input_map={'input': out1}, import_scope='graph2')

, .

print(tf.global_variables())

.