optimize_for_inference.py

"""Removes parts of a graph that are only needed for training.
There are several common transformations that can be applied to GraphDefs
created to train a model, that help reduce the amount of computation needed when
the network is used only for inference. These include:
 - Removing training-only operations like checkpoint saving.
 - Stripping out parts of the graph that are never reached.
 - Removing debug operations like CheckNumerics.
 - Folding batch normalization ops into the pre-calculated weights.
 - Fusing common operations into unified versions.
This script takes either a frozen binary GraphDef file (where the weight
variables have been converted into constants by the freeze_graph script), or a
text GraphDef proto file (the weight variables are stored in a separate
checkpoint file), and outputs a new GraphDef with the optimizations applied.
If the input graph is a text graph file, make sure to include the node that
restores the variable weights in output_names. That node is usually named
"restore_all".
An example of command-line usage is:
bazel build tensorflow/python/tools:optimize_for_inference && \
bazel-bin/tensorflow/python/tools/optimize_for_inference \
--input=frozen_inception_graph.pb \
--output=optimized_inception_graph.pb \
--frozen_graph=True \
--input_names=Mul \
--output_names=softmax
"""

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import argparse
import os
import sys

from google.protobuf import text_format

from tensorflow.core.framework import graph_pb2
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import graph_io
from tensorflow.python.platform import app
from tensorflow.python.platform import gfile
from tensorflow.python.tools import optimize_for_inference_lib

FLAGS = None


def main(unused_args):
  if not gfile.Exists(FLAGS.input):
    print("Input graph file '" + FLAGS.input + "' does not exist!")
    return -1

  input_graph_def = graph_pb2.GraphDef()
  with gfile.Open(FLAGS.input, "rb") as f:
    data = f.read()
    if FLAGS.frozen_graph:
      input_graph_def.ParseFromString(data)
    else:
      text_format.Merge(data.decode("utf-8"), input_graph_def)

  output_graph_def = optimize_for_inference_lib.optimize_for_inference(
      input_graph_def,
      FLAGS.input_names.split(","),
      FLAGS.output_names.split(","),
      _parse_placeholder_types(FLAGS.placeholder_type_enum),
      FLAGS.toco_compatible)

  if FLAGS.frozen_graph:
    f = gfile.FastGFile(FLAGS.output, "w")
    f.write(output_graph_def.SerializeToString())
  else:
    graph_io.write_graph(output_graph_def,
                         os.path.dirname(FLAGS.output),
                         os.path.basename(FLAGS.output))
  return 0


def _parse_placeholder_types(values):
  """Extracts placeholder types from a comma separate list."""
  values = [int(value) for value in values.split(",")]
  return values if len(values) > 1 else values[0]


def parse_args():
  """Parses command line arguments."""
  parser = argparse.ArgumentParser()
  parser.register("type", "bool", lambda v: v.lower() == "true")
  parser.add_argument(
      "--input",
      type=str,
      default="",
      help="TensorFlow \'GraphDef\' file to load.")
  parser.add_argument(
      "--output",
      type=str,
      default="",
      help="File to save the output graph to.")
  parser.add_argument(
      "--input_names",
      type=str,
      default="",
      help="Input node names, comma separated.")
  parser.add_argument(
      "--output_names",
      type=str,
      default="",
      help="Output node names, comma separated.")
  parser.add_argument(
      "--frozen_graph",
      nargs="?",
      const=True,
      type="bool",
      default=True,
      help="""\
      If true, the input graph is a binary frozen GraphDef
      file; if false, it is a text GraphDef proto file.\
      """)
  parser.add_argument(
      "--placeholder_type_enum",
      type=str,
      default=str(dtypes.float32.as_datatype_enum),
      help="""\
      The AttrValue enum to use for placeholders.
      Or a comma separated list, one value for each placeholder.\
      """)
  parser.add_argument(
      "--toco_compatible",
      type=bool,
      default=False,
      help="""\
      If true, only use ops compatible with Tensorflow
      Lite Optimizing Converter.\
      """)
  return parser.parse_known_args()


if __name__ == "__main__":
  FLAGS, unparsed = parse_args()
  app.run(main=main, argv=[sys.argv[0]] + unparsed)