tf.train.write_graph( graph_or_graph_def, logdir, name, as_text=True )
See the guide: Training > Training Utilities
Writes a graph proto to a file.
The graph is written as a text proto unless
v = tf.Variable(0, name='my_variable') sess = tf.Session() tf.train.write_graph(sess.graph_def, '/tmp/my-model', 'train.pbtxt')
v = tf.Variable(0, name='my_variable') sess = tf.Session() tf.train.write_graph(sess.graph, '/tmp/my-model', 'train.pbtxt')
logdir: Directory where to write the graph. This can refer to remote filesystems, such as Google Cloud Storage (GCS).
name: Filename for the graph.
True, writes the graph as an ASCII proto.
The path of the output proto file.