0

I am trying to do the Tensorflow for Poets tutorial and I am on step 6 which is using the retrained model. I am trying to run this command:

python -m scripts.label_image --graph=tf_files/retrained_graph.pb -- image=tf_files/flower_photos/daisy/21652746_cc379e0eea_m.jpg

and this error is returned:

2019-03-06 16:13:27.270248: I Tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
Traceback (most recent call last):
File "C:\Users\rzara\AppData\Local\Programs\Python\Python37\lib\runpy.py", line 193, in _run_module_as_main
"__main__", mod_spec)
 File "C:\Users\rzara\AppData\Local\Programs\Python\Python37\lib\runpy.py", line 85, in _run_code
exec(code, run_globals)
File "C:\Users\rzara\Documents\GitHub\tensorflow-for-poets- 2\scripts\label_image.py", line 121, in <module>
input_operation = graph.get_operation_by_name(input_name);
File "C:\Users\rzara\AppData\Local\Programs\Python\Python37\lib\site- packages\tensorflow\python\framework\ops.py", line 3606, in get_operation_by_name
return self.as_graph_element(name, allow_tensor=False, allow_operation=True)
File "C:\Users\rzara\AppData\Local\Programs\Python\Python37\lib\site- packages\tensorflow\python\framework\ops.py", line 3478, in as_graph_element
return self._as_graph_element_locked(obj, allow_tensor, allow_operation)
File "C:\Users\rzara\AppData\Local\Programs\Python\Python37\lib\site-packages\tensorflow\python\framework\ops.py", line 3538, in 
_as_graph_element_locked
"graph." % repr(name))
**KeyError: "The name 'import/Mul' refers to an Operation not in the graph."**

I have already tried to open label_image.py and change:

input_height = 299
input_width = 299
input_layer = 'Mul'

but it still does not run. Please help. I have already checked another post on here but it did not help: tensorflow for poets: "The name 'import/input' refers to an Operation not in the graph."

rzaratx
  • 756
  • 3
  • 9
  • 29

1 Answers1

0

Well apparently the layer is wrong, looked at the official github and there they use the name 'input', maybe that works. If that does not work, you need to load in your model and check the tensorboard graph, if you followed the tutorial you've most likely already done that. It should be clear from the graph which one your input tensor is, the first one. If it doesn't work, let me know

T. Kelher
  • 1,188
  • 10
  • 9