[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.load.tf.l

[resh]

Basically, I am trying to get IR for TF2 supported saved_model.pb and this actually throws this error, an anyone make me understand the cause for it?

[Wan_Intel]

Hi Reshu Singh,

Thank you for reaching out to us.

I noticed that you are using OpenVINO™ 2020.3.

We regret to inform you that OpenVINO™ 2020.3 does not support the conversion of TensorFlow 2 models to IR. However, OpenVINO™ 2021.3 does support the conversion of TensorFlow 2 models to IR.

OpenVINO™ 2021.3 is available for download at the following page:

https://software.intel.com/content/www/us/en/develop/tools/openvino-toolkit/download.html

The steps to convert TensorFlow 2 models to IR is available at the following page:

https://docs.openvinotoolkit.org/2021.3/openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_TensorFlow.html#Convert_From_TF2X

Regards,

Wan

[resh]

Hi @Wan_Intel !

I tried downloading 2021.3 version but in that not able to see model_optimizer in <INSTALL_DIR>/deployment_tools/model_optimizer

Here is the screenshot-

[Wan_Intel]

Hi Reshu Singh,

Thanks for your information.

The issue you encountered is not a common issue when installing OpenVINO™ 2021.3. However, we found that this issue happens when the user installs OpenVINO™ 2021.3 runtime package.

Please download OpenVINO™ 2021.3 at the following page:

https://software.intel.com/content/www/us/en/develop/tools/openvino-toolkit/download.html

Select the options of Operating System (OS), Distribution, Version Type, and Installer Type, a “Download” button will appear. Click on the Download button to download the OpenVINO™ Toolkit installer on your local machine.

Next, install OpenVINO™ 2021.3 by following the steps given in the installation guides:

https://docs.openvinotoolkit.org/2021.3/openvino_docs_install_guides_installing_openvino_linux.html

To verify OpenVINO™ 2021.3 is installed correctly on your local machine, please run demo scripts located in Get Started Guides:

https://docs.openvinotoolkit.org/2021.3/get_started_guides.html

For example, if you are using Linux OS, you can run the Image Classification Demo Scripts:

https://docs.openvinotoolkit.org/2021.3/openvino_docs_get_started_get_started_linux.html#image_classification_demo_script

Please get back to us should you face any difficulty in downloading and installing OpenVINO™ 2021.3.

Regards,

Wan

[resh]

Hi @Wan_Intel !

I actually now am able to successfully download 2021.3 version but the error with my very basic model, the error is attached as

Can you help in identifying the issue?

C

My simple sequential model is here as -

def create_model():

model = tf.keras.models.Sequential([

keras.layers.Dense(512, activation='relu', input_shape=(784,)),

keras.layers.Dropout(0.2),

keras.layers.Dense(10)

])

model.compile(optimizer='adam',

loss=tf.losses.SparseCategoricalCrossentropy(from_logits=True),

metrics=[tf.metrics.SparseCategoricalAccuracy()])

return model

[resh]

[ ERROR ] Cannot infer shapes or values for node "StatefulPartitionedCall/sequential/dense/MatMul".
[ ERROR ] MatMul input shapes are incorrect. COL_INDEX_DIMs are not equal. Node: StatefulPartitionedCall/sequential/dense/MatMul. Shapes: [array([ 1, 512, 784, 512]), array([ 1, 512, 784, 512])]
[ ERROR ]
[ ERROR ] It can happen due to bug in custom shape infer function <function MatMul.infer at 0x7f0862f54840>.
[ ERROR ] Or because the node inputs have incorrect values/shapes.
[ ERROR ] Or because input shapes are incorrect (embedded to the model or passed via --input_shape).
[ ERROR ] Run Model Optimizer with --log_level=DEBUG for more information.
[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.middle.PartialInfer.PartialInfer'>): Stopped shape/value propagation at "StatefulPartitionedCall/sequential/dense/MatMul" node.
For more information please refer to Model Optimizer FAQ, question #38. (https://docs.openvinotoolkit.org/latest/openvino_docs_MO_DG_prepare_model_Model_Optimizer_FAQ.html?question=38#question-38)

[Wan_Intel]

Hi Reshu Singh,

I have compiled your sequential model and successfully converted the model to IR using the following model optimizer command:

python mo_tf.py --saved_model_dir <path_to_saved_model_dir> --input_shape [1,512,512,784]

General guidelines on specifying input shape is available at the following links:

https://docs.openvinotoolkit.org/2021.3/openvino_docs_MO_DG_prepare_model_convert_model_Converting_Model_General.html

https://docs.openvinotoolkit.org/2021.3/openvino_docs_MO_DG_prepare_model_convert_model_Converting_Model_General.html#when_to_specify_input_shapes

Regards,

Wan

[Wan_Intel]

Hi Reshu Singh,

This thread will no longer be monitored since we have provided a solution.

If you need any additional information from Intel, please submit a new question.

Regards,

Wan

[resh]

Resolved! Thanks for helping out