joshuacwnewton/onnx-to-tflite

Converting the ONNX model representation to the TensorFlow Lite representation.

Hi! I'm an open source maintainer working for the NeuroPoly research lab at Polytechnique Montreal. 👋

For the past 3 years, I've maintained the Spinal Cord Toolbox, a highly-regarded Python-based software package in a niche, data-driven research field. Our user base is small but meaningful; we serve the needs of medical researchers by providing them with a wealth of spinal cord image processing tools, with our work serving as the backbone for countless vital clinical studies published in many high-impact journals.

onnx-to-tflite's Issues

ValueError: Invalid tensors 'input' were found.

All works fine till using

converter = tf.compat.v1.lite.TFLiteConverter.from_saved_model(pb_path,
                                                      input_arrays=input_nodes,
                                                      output_arrays=output_nodes,
                                                               )
tflite_rep = converter.convert()

log

[/usr/local/lib/python3.10/dist-packages/tensorflow/lite/python/util.py](https://localhost:8080/#) in get_tensors_from_tensor_names(graph, tensor_names)
    164   # Throw ValueError if any user input names are not valid tensors.
    165   if invalid_tensors:
--> 166     raise ValueError("Invalid tensors '{}' were found.".format(
    167         ",".join(invalid_tensors)))
    168   return tensors

ValueError: Invalid tensors 'input' were found.

Tensorflow addons not supported for 1.x

Hi this is the same issue as #1, reopening it as a suitable solution isn't mentioned before closing it.

Thank you.

Stuck in step 2.2 while running on Colab

Step 2.2 is currently not only spitting warnings, it seems to not work at all unfortunately:

import onnx
from onnx_tf.backend import prepare

onnx_model = onnx.load(onnx_path)
tf_rep = prepare(onnx_model)

results in:

/usr/local/lib/python3.7/dist-packages/tensorflow_addons/utils/ensure_tf_install.py:67: UserWarning: Tensorflow Addons supports using Python ops for all Tensorflow versions above or equal to 2.3.0 and strictly below 2.5.0 (nightly versions are not supported). 
 The versions of TensorFlow you are currently using is 1.15.2 and is not supported. 
Some things might work, some things might not.
If you were to encounter a bug, do not file an issue.
If you want to make sure you're using a tested and supported configuration, either change the TensorFlow version or the TensorFlow Addons's version. 
You can find the compatibility matrix in TensorFlow Addon's readme:
https://github.com/tensorflow/addons
  UserWarning,
---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-4-f6cf1dfe42ba> in <module>()
      1 import onnx
----> 2 from onnx_tf.backend import prepare
      3 
      4 onnx_model = onnx.load(onnx_path)
      5 tf_rep = prepare(onnx_model)

7 frames
/usr/local/lib/python3.7/dist-packages/tensorflow_addons/utils/types.py in <module>()
     21 
     22 # TODO: Remove once https://github.com/tensorflow/tensorflow/issues/44613 is resolved
---> 23 from tensorflow.python.keras.engine import keras_tensor
     24 
     25 

ImportError: cannot import name 'keras_tensor' from 'tensorflow.python.keras.engine' (/tensorflow-1.15.2/python3.7/tensorflow_core/python/keras/engine/__init__.py)

---------------------------------------------------------------------------
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

To view examples of installing some common dependencies, click the
"Open Examples" button below.
---------------------------------------------------------------------------