TensorFlow C++ Session API reference documentation

TensorFlow's public C++ API includes only the API for executing graphs, as of version 0.5. To control the execution of a graph from C++:

Build the computation graph using the Python API.
Use tf.train.write_graph() to write the graph to a file.

Load the graph using the C++ Session API. For example:

// Reads a model graph definition from disk, and creates a session object you
// can use to run it.
Status LoadGraph(string graph_file_name, Session** session) {
 GraphDef graph_def;
 TF_RETURN_IF_ERROR(
     ReadBinaryProto(Env::Default(), graph_file_name, &graph_def));
 TF_RETURN_IF_ERROR(NewSession(SessionOptions(), session));
 TF_RETURN_IF_ERROR((*session)->Create(graph_def));
 return Status::OK();
}