Community
cancel
Showing results for 
Search instead for 
Did you mean: 
bibas__koby
Beginner
66 Views

Model optimizer fails- MobileNet-SSD mxnet

Hi,

I'm trying to convert the following model with intel optimizer:

    https://github.com/zhreshold/mxnet-ssd/releases/download/v0.7-alpha/mobilenet-ssd-512.zip

I used the following command:

sudo python3 mo_mxnet_converter.py --model-name mobilenet-ssd-512 --models-dir $INPUT_DIR --output-dir $OUTPUT_DIR --img-format BGR --img-width 224 --img-height 224 --img-channels 3

 

and got the error:


[07:44:26] src/nnvm/legacy_json_util.cc:209: Loading symbol saved by previous version v0.12.0. Attempting to upgrade...
[07:44:26] src/nnvm/legacy_json_util.cc:217: Symbol successfully upgraded!
/usr/local/lib/python3.5/dist-packages/mxnet/module/base_module.py:53: UserWarning: You created Module with Module(..., label_names=['softmax_label']) but input with name 'softmax_label' is not found in symbol.list_arguments(). Did you mean one of:
    data
    label
  warnings.warn(msg)
Traceback (most recent call last):
  File "mo_mxnet_converter.py", line 105, in <module>
    convert_model(args)
  File "mo_mxnet_converter.py", line 82, in convert_model
    args.img_channels, enable_fuse, visualize_path)
  File "mo_mxnet_converter.py", line 42, in get_model_config
    model = load_model(n_name, iteration_number, input_shape, data_path)
  File "mo_mxnet_converter.py", line 31, in load_model
    return read_mxnet_json(json_path, input_shape, mx_loaded_model=model_loaded)
  File "/opt/intel/deeplearning_deploymenttoolkit_2017.1.0.5675/deployment_tools/model_optimizer/mxnet_converter/mxnet_convertor/mxnet_utils.py", line 26, in read_mxnet_json
    return MxNetModel(data, input_shape, mx_loaded_model)
  File "/opt/intel/deeplearning_deploymenttoolkit_2017.1.0.5675/deployment_tools/model_optimizer/mxnet_converter/mxnet_convertor/mxnet_primitives.py", line 53, in __init__
    json_layers = MxNetJsonModelPatternMatcher.remove_and_change_layers(json_layers)
  File "/opt/intel/deeplearning_deploymenttoolkit_2017.1.0.5675/deployment_tools/model_optimizer/mxnet_converter/mxnet_convertor/mxnet_pattern_matcher.py", line 148, in remove_and_change_layers
    MxNetJsonModelPatternMatcher.ssd_pattern_remove_reshape(json_layers)
  File "/opt/intel/deeplearning_deploymenttoolkit_2017.1.0.5675/deployment_tools/model_optimizer/mxnet_converter/mxnet_convertor/mxnet_pattern_matcher.py", line 109, in ssd_pattern_remove_reshape
    if l2['inputs'][0][0] != i or l2['op'] != 'Reshape':
IndexError: list index out of range

 

Did I missed something?

Regards,

Koby

0 Kudos
2 Replies
66 Views

Hi Koby,

We have new version of CVSDK here, (new brand name OpenVINO)

      https://software.intel.com/en-us/openvino-toolkit

And could not reproduce your issue.

Please try again with this new version.

Regards,

Peter.

66 Views

Also,

Our dev team found your model (https://github.com/zhreshold/mxnet-ssd/releases/download/v0.7-alpha/mobilenet-ssd-512.zip) contain topology for training. You should convert it to deploy version first. (Use deploy.py script from https://github.com/zhreshold/mxnet-ssd repo)

Thank you.

Regards,

Peter.