SnapML sample code throws error

Follow

Marco Vanossi

Snap Lens Network Member Lens Studio Team

• June 12, 2020 11:25

Hi,

 I'm following the instructions on https://lensstudio.snapchat.com/guides/machine-learning/ml-frameworks/export-from-tensorflow/ in order to freeze a TensorFlow PB file .

 I am using the recommended version 1.15 for TensorFlow.

 This is the error I get.

python3 freeze.py

Traceback (most recent call last):

  File "freeze.py", line 17, in <module>

    saved_model_tags=tag_constants.SERVING)

TypeError: 'module' object is not callable

 

thanks

thanks

0

• 

  Marco Vanossi

  Snap Lens Network Member Lens Studio Team

  • June 12, 2020 11:38

  Correct definition is 

   

  def freeze_graph(input_graph,
  	input_saver,
  	input_binary,
  	input_checkpoint,
  	output_node_names,
  	restore_op_name,
  	filename_tensor_name,
  	output_graph,
  	clear_devices,
  	initializer_nodes,
  	variable_names_whitelist="",
  	variable_names_blacklist="",
  	input_meta_graph=None,
  	input_saved_model_dir=None,
  	saved_model_tags=tag_constants.SERVING,
  	checkpoint_version=saver_pb2.SaverDef.V2):

  Comment actions Permalink
• 

  Marco Vanossi

  Snap Lens Network Member Lens Studio Team

  • June 12, 2020 11:42

  Here's code that worked for me

   

  from tensorflow.python.tools import freeze_graph

  from tensorflow.python.saved_model import tag_constants

  freeze_graph.freeze_graph(input_graph=None,

               input_saver=None,

               input_binary=False,

               input_checkpoint=None,

               output_node_names="xxx",

               restore_op_name=None,

               filename_tensor_name=None,

               output_graph='frozen_graph.pb',

               clear_devices=False,

               initializer_nodes="",

               variable_names_whitelist="",

               variable_names_blacklist="",

               input_meta_graph=None,

               input_saved_model_dir=".",

               saved_model_tags=tag_constants.SERVING)

   

  Comment actions Permalink
• 

  Marco Vanossi

  Snap Lens Network Member Lens Studio Team

  • June 12, 2020 11:49

  This code snippet provided also has errors:

   

  def export_from_frozen_graph(frozen_graph_filename,

                               input_node_names=['placeholder'],

                               output_node_names=['output'],

                               output_filename='frozen_graph.pb',

                               optimize=True):

      tf.reset_default_graph()

      graph_def = tf.GraphDef()

   

      with tf.gfile.GFile(frozen_graph_dir, "rb") as f:

   

   

  frozen_graph_dir should be replaced with frozen_graph_filename

  Comment actions Permalink
• 

  Marco Vanossi

  Snap Lens Network Member Lens Studio Team

  • June 12, 2020 12:00

  The following lines should be removed as well from the sample snippet as nowhere in the page it says the code is running on google collab

  # download frozen graph
  from google.colab import files
  files.download('mobilenet_v1_optimized.pb') 

  Comment actions Permalink
• 

  Pavel Antonenko

  Snap Lens Network Member Lens Studio Team

  • June 12, 2020 14:10

  Thanks a lot for feedback, Marco!

  Comment actions Permalink
• 

  Eric Hu

  Snap Lens Network Member Lens Studio Team

  • June 12, 2020 18:42

  Hi Marco, thanks for the feedback! We've updated the Tensorflow guide and changes should be reflected shortly.

  Comment actions Permalink

Please sign in to leave a comment.