Conversion of mxnet model (.params and .json) to a openvino IR (.xml and .bin) using model optimizer

[Cronje__Abrie]

I trained a custom object detection model (framework: MXNet) (base network: resnet-50) (training image: 811284229777.dkr.ecr.us-east-1.amazonaws.com/object-detection:latest) on AWS Sagemaker, resulting in a .params and .json file.

When I try to optimize it using the openvino model optimizer with the code below:

python3 mo.py
--input_model deploy_model_algo_1-0000.params
--output_dir openvino_models/ir/
--input_shape "[1,300,300,3]"

I get the following error

[ ERROR ]  Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.front.user_data_repack.UserDataRepack'>): Original placeholders: 'relu4_3_scale, data, conv1_1_weight, conv1_1_bias, conv1_2_weight, conv1_2_bias, conv2_1_weight, conv2_1_bias, conv2_2_weight, conv2_2_bias, conv3_1_weight, conv3_1_bias, conv3_2_weight, conv3_2_bias, conv3_3_weight, conv3_3_bias, conv4_1_weight, conv4_1_bias, conv4_2_weight, conv4_2_bias, conv4_3_weight, conv4_3_bias, relu4_3_cls_pred_conv_weight, relu4_3_cls_pred_conv_bias, conv5_1_weight, conv5_1_bias, conv5_2_weight, conv5_2_bias, conv5_3_weight, conv5_3_bias, fc6_weight, fc6_bias, fc7_weight, fc7_bias, relu7_cls_pred_conv_weight, relu7_cls_pred_conv_bias, relu4_3_loc_pred_conv_weight, relu4_3_loc_pred_conv_bias, relu7_loc_pred_conv_weight, relu7_loc_pred_conv_bias'. Freezing was requested for ''. --input_shape was provided without --input. Can not deduce which node shape to override

These 'placeholders' all refer to sections in the .json file where the 'op' field is null, for instance:
 

{
      "op": "null", 
      "name": "conv5_1_weight", 
      "attrs": {
        "kernel": "(3, 3)", 
        "num_filter": "512", 
        "pad": "(1, 1)"
      }, 
      "inputs": []
    },

When I do not specify the input shape, using the call below,

python3 mo.py
--input_model deploy_model_algo_1-0000.params
--output_dir openvino_models/ir/

I get the following error,

[ ERROR ]  Cannot infer shapes or values for node "relu7_loc_pred_conv_bias".
[ ERROR ]  Parameter node relu7_loc_pred_conv_bias should have `shape` attribute. Please use cli options to set model input shape
[ ERROR ]  
[ ERROR ]  It can happen due to bug in custom shape infer function <function Parameter.infer at 0x7fc65aa2ec20>.
[ 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 "relu7_loc_pred_conv_bias" node.
 For more information please refer to Model Optimizer FAQ (https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_Model_Optimizer_FAQ.html), question #38.

The suggested question 38 was not very helpful to me.

38. What does the message "Stopped shape/value propagation at node" mean?

Model Optimizer cannot infer shapes or values for the specified node. It can happen because of a bug in the custom shape infer function, because the node inputs have incorrect values/shapes, or because the input shapes are incorrect.

I am quite new to openvino and any help would be greatly appreciated.

[Munesh_Intel]

Hi Abrie,

Would you please provide us the trained model files, together with .params and .json files, so that we can reproduce your issue?

Thank you.

Regards,

Munesh

[Martino__Evan]

I'm having the same error converting RetinaFace, with the .json and .params files available at https://www.dropbox.com/s/53ftnlarhyrpkg2/retinaface-R50.zip?dl=0

Any help would be appreciated!

[Munesh_Intel]

Hi Evan,

Greetings to you.

Please create a new thread to help us serve you better, since this thread is not monitored anymore..

Thank you.

Regards,

Munesh