Hi!

I am following your tutorial ( https://omid.al/posts/2017-02-20-Tutorial-Build-Your-First-Tensorflow-Android-App.html ).

When I try to generate the .pb file with freeze_graph I get an error.

I have all the required files in the same directory.
Any idea why I am getting this error?

Thanks in advance.

`---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
in ()
23 filename_tensor_name,
24 output_frozen_graph_name,
---> 25 clear_devices, "")

/home/atig/GitKraken/hallgato-grofattila/Training/freeze_graph.py in freeze_graph(failed resolving arguments)
80
81 # 'input_checkpoint' may be a prefix if we're using Saver V2 format
---> 82 if not saver_lib.checkpoint_exists(input_checkpoint):
83 print("Input checkpoint '" + input_checkpoint + "' doesn't exist!")
84 return -1

AttributeError: module 'tensorflow.python.training.saver' has no attribute 'checkpoint_exists'`

Hi Omimo,
I ran your scripts, but when I execute this statement,
inferenceInterface.fillNodeFloat(INPUT_NODE, INPUT_SIZE, inputFloats);
it is throwing me the following error.
java.lang.IllegalArgumentException: No Operation named [I] in the Graph

omid,

do you think it would be easy to implement another model with your sample?

I surfed the net a lot and found out that there is a variety of ways for using tensorflow model in android,
as I runned your code successfully, I am wondering if using a bit complicated model may work with that also?
It would be best if you could provide this step by step teaching to a more complicated case like this
https://github.com/laviavigdor/twitter-sentiment-analysis

sepas

hey omid,
why when I am running your simple NN code it raises error:
return bytes_or_text.decode(encoding)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 85: invalid start byte

It has problem with last part I mean optimization,
persian girl

I really appreciate your work here. I've been having trouble putting together the TF app through Bazel, and seeing this app plugged in and dropped into Android Studio was really refreshing! cheers ^____^

Why cant we use .tflite file instead of .pb file in Tensorflow Inference model.
And if I want to use .tflite model than how to implement it with Tensorflow Inference model.