jgespino
commented
2 years ago Hi @KodeWorker
I'm still looking into the issue, I tested with the latest pre-release dev package and I am seeing a different error message. Did you just retrained the model with --data_format "channels_first" or did you make other changes to the model? Could you provide the model in checkpoint format?
mo --input_model custom-efficientnet-b0.pb
Model Optimizer arguments:
Common parameters:
- Path to the Input Model: C:\Users\user\gh10633\custom-efficientnet-b0.pb
- Path for generated IR: C:\Users\user\gh10633\.
- IR output name: custom-efficientnet-b0
- Log level: ERROR
- Batch: Not specified, inherited from the model
- Input layers: Not specified, inherited from the model
- Output layers: Not specified, inherited from the model
- Input shapes: Not specified, inherited from the model
- Source layout: Not specified
- Target layout: Not specified
- Layout: Not specified
- Mean values: Not specified
- Scale values: Not specified
- Scale factor: Not specified
- Precision of IR: FP32
- Enable fusing: True
- Enable grouped convolutions fusing: True
- Move mean values to preprocess section: None
- Reverse input channels: False
- Use legacy API for model processing: False
- Use the transformations config file: None
TensorFlow specific parameters:
- Input model in text protobuf format: False
- Path to model dump for TensorBoard: None
- List of shared libraries with TensorFlow custom layers implementation: None
- Update the configuration file with input/output node names: None
- Use configuration file used to generate the model with Object Detection API: None
- Use the config file: None
- OpenVINO runtime found in: c:\users\user\openvino_dev215\lib\site-packages\openvino
OpenVINO runtime version: 2022.1.0-6682-121d59aa80a
Model Optimizer version: 2022.1.0-6682-121d59aa80a
[ ERROR ] Cannot infer shapes or values for node "efficientnet-b0/model/head/dense/MatMul".
[ ERROR ] MatMul input shapes are incorrect. COL_INDEX_DIMs are not equal. Node: efficientnet-b0/model/head/dense/MatMul. Shapes: [masked_array(data=[ 1, 1280],
mask=False,
fill_value=-1000000007,
dtype=int64), masked_array(data=[320, 2],
mask=False,
fill_value=-1000000007,
dtype=int64)]
[ ERROR ]
[ ERROR ] It can happen due to bug in custom shape infer function <function MatMul.infer at 0x00000288710DAD38>.
[ 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 'openvino.tools.mo.middle.PartialInfer.PartialInfer'>): Stopped shape/value propagation at "efficientnet-b0/model/head/dense/MatMul" node.
For more information please refer to Model Optimizer FAQ, question #38. (https://docs.openvino.ai/latest/openvino_docs_MO_DG_prepare_model_Model_Optimizer_FAQ.html?question=38#question-38)Regards, Jesus
KodeWorker
System information
Detailed description
I trained a custom efficientnet-b0 model using --data_format "channels_first": https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet
And I convert the checkpoint to frozen graph model: https://gist.github.com/KodeWorker/6bb37a7864d98836356b85bdea2d39eb
Then, I tried to convert the frozen graph model to OpenVINO IR. I was a failure and the error messages is shown below:
However, I used the pretrained model which is "channels_last" and the conversion is a success. Why the "channels_first" model leads to a failed conversion?
Model Download Links:
Steps to reproduce