TF ssdv2 to IRv10 error: "Preprocessor/mul" has 2 producers

[Tr__Ra]

Hello, this is the first time I get this error. Last week it was working.

I tried both 20.01 and 20.03.

See below for command and error. Thanks!

    python /opt/intel/openvino_2020.3.194/deployment_tools/model_optimizer/mo.py \

    --input_model frozen_inference_graph.pb \

    --tensorflow_use_custom_operations_config /opt/intel/openvino_2020.3.194/deployment_tools/model_optimizer/extensions/front/tf/ssd_v2_support.json \

    --tensorflow_object_detection_api_pipeline_config pipeline.config \

    --reverse_input_channels \

    --output_dir ./frozen_model \

    --data_type FP16\

Output:

[ ERROR ]  Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.front.ChangePlaceholderTypes.ChangePlaceholderTypes'>): Something bad has happened with graph! Data node "Preprocessor/mul" has 2 producers

[Max_L_Intel]

Hi @Tr__Ra 

Have you ever successfully converted this exact frozen_inference_graph.pb model or are you doing this for the first time?

If you already did this, then please ensure that your TF frozen model did not get corrupted after that and that you correctly provide Model Optimizer conversion command, as per the example below:

<INSTALL_DIR>/deployment_tools/model_optimizer/mo_tf.py --input_model=/tmp/ssd_inception_v2_coco_2018_01_28/frozen_inference_graph.pb --transformations_config <INSTALL_DIR>/deployment_tools/model_optimizer/extensions/front/tf/ssd_v2_support.json --tensorflow_object_detection_api_pipeline_config /tmp/ssd_inception_v2_coco_2018_01_28/pipeline.config --reverse_input_channels

If you do this for the first time, then please specify if this is publicly available model (e.g. from Open Model Zoo) or custom trained one (topology used, TF Object Detection API version, etc.)

Thanks.
Best regards, Max.

[Tr__Ra]

yes i have until last week.

Seems to be a problem with obj det model coming from training using obj det api.

Models that i train in colab now produce the error above.

Models from a week or so or older convert successfully.

[Max_L_Intel]

Hi @Tr__Ra 

Okay, so then it was not exactly the same model's instance, but different instance of the same topology.
In that case please check for changes that you did to model training process (configuration, TF API version, etc.) in order to isolate the rootcause.

Thank you.

[Kamma__Vijayakumar]

Even i am facing the same issue. I have not changed any training pipeline and all. only i have increased the amount of data. Not sure exactly.

Please help !!

[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.front.ChangePlaceholderTypes.ChangePlaceholderTypes'>): Something bad has happened with graph! Data node "Preprocessor/mul" has 2 producers 

[Max_L_Intel]

Hi @Kamma__Vijayakumar 

Please create new topic in this forum section and provide all the available details including information that I requested here earlier (if this is publicly available model or custom trained one, what is topology used, TensorFlow Object Detection API version).

Thanks.
Best regards, Max.

[Tr__Ra · Accepted solution]

The error comes from TF Obj Det API. Use models/research github version prior to july 1st. Their latest commit l.eads to this error.