scala - What is the Tensorflow Java Api `toGraphDef` equivalent in Python? -

i using tensorflow java api load created tensorflow model jvm. using example: tensorflow/examples/labelimage.java

here simple scala code:

import java.nio.file.{files, path, paths} import org.tensorflow.{graph, session, tensor}  def readallbytesorexit(path: path): array[byte] = files.readallbytes(path) val graphdef = readallbytesorexit(paths.get("path_to_a_single_file_describing_tf_model.pb")) val g = new graph() g.importgraphdef(graphdef) val session = new session(g) val result: tensor = session.runner().feed("input", image).fetch("output").run().get(0)) 

how save model both session , graph stored in same file. described in "path_to_a_single_file_describing_tf_model.pb" above.

described here mentions:

the serialized representation of graph, referred graphdef, can generated tographdef() , equivalents in other language apis.

what equivalents in other language apis? dont find obvious

note: looked @ mnist_saved_model.py under tensorflow_serving saving through procedure gives me .pb file , variables folder. when trying load .pb file get: java.lang.illegalargumentexception: invalid graphdef

currently java api of tensorflow, found how save graph graphdef (i.e. without variables , meta-data). can done writing array[byte] file:

files.write(paths.get(modeldir, modelname), mygraph.tographdef) 

here mygraph java object graph class.

i suggest save model python api, using savedmodel api defined here. save model in folder both serialized graph in .pb file , variables in folder. note tag_constants use you'll need in scala/java code load model variables. graph , session variables loaded savedmodelbundle java class java api. returns wrapper both graph , session containing variables values:

val model = savedmodelbundle.load(modeldir, modeltag) 

if tried this, maybe can share code see why returned invalid graphdef.

another option freeze graph, i.e. turned variable nodes constant nodes self-contained in .pb file. mores infos here freezing part