I am working on notebook files on Ubuntu. When running the following command, I got the following error message:
!python3 /opt/intel/openvino/deployment_tools/model_optimizer/mo_tf.py --input_model tf_model/top_layers.iv3.pb --input_shape=[1,299,299,3]
[ ERROR ] List of operations that cannot be converted to Inference Engine IR:
[ ERROR ] FusedBatchNormV3 (94)
[ ERROR ] batch_normalization_1_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_2_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_3_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_4_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_5_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_6_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_7_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_8_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_9_1/FusedBatchNormV3_1
[ ERROR ] batch_normalization_10_1/FusedBatchNormV3_1
I got top_layers.iv3.bin file, but failed to create top_layers.iv3.xml file.
Hi Jung Soon,
- Which version of OpenVINO are you using?
- What type of model are you using? Is it a public model or custom-trained model?
- Can you share the model and other necessary files with us, to replicate the issue? If required, I can send a PM to share the files privately.
The version is openvino-toolkit-p-2019.3.376. I've downloaded the notebook files from DC to Edge Course.
I was not able to upload files on this forum. So, I put the project folder on google drive and the following
is the link:
I could not follow the installation instruction, as it needs to be done with conda (I did not install conda).
This might cause the error.
The support for FusedBatchNorm operation is added in the latest version of OpenVINO(2020.1). I have generated the IR files from the attached model (Notebooks/tf_model/top_layers.iv3.pb), using OpenVINO(2020.1).
Hope this helps.