hello i am trying to convert mxnet model
i guess some topology are not supported
script used to convert model
python mo_mxnet.py --input_model /media/prateek/prateek_space/model_files/helmet_n_vest/external/models/openvino/mobilenet1.0-0000.params -o ~/prateek_space/model_files/openvino_model/ --input_symbol /media/prateek/prateek_space/model_files/helmet_n_vest/external/models/openvino/mobilenet1.0-symbol.json
Operation '_arange' not supported. Please register it as custom op.
For more information please refer to Model Optimizer FAQ (https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_Model_Optimizer_FAQ.html), question #86.
Unexpected exception happened during extracting attributes for node yolov30_yolooutputv30__mulscalar2.
Original exception message: Operation '_arange' not supported. Please register it as custom op.
Can you help me..how do I go about converting this model?
This is weird because '_arange' layer/operation is mentioned on the supported operations for MXNet (https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_Supported_Frameworks_Layers.html). What version of OpenVINO are you running? (R1, R2, R3?)
The latest version of OpenVINO(2019_R3.1) is openvino_2019.3.379. Please confirm that you are using this version because the '_arange' MXNet* symbol is not supported on any previous versions.
If the issue still persist on the latest version, can you share your model and necessary files for me to replicate this? If required, I can sent you a PM to share the model privately.
Are you working on windows based system?
that could be one of the reason for different version number
its yolo3 with depth .25 on mobilenet backbone its external model..
i only have param file.
To convert an MXNet* model using Model Optimizer, there should be both model-file-symbol.json and model-file-0000.params in the specified file format.
On the command you mentioned to convert the model, the required files are used in expected file format. Can you provide the same?
Also, you are right! For Linux, openvino_2019.3.376 is the latest version.