Data batch channel count (1) does not match filter input channel count (256)

[EnricoBeltramo]

System information (version)

• OpenVINO => 2020.4
• Operating System / Platform => ubuntu 64 Bit
• Compiler => python
• Problem classification: Model Conversion
• Framework: pytorch (if applicable)
• Model name: custom (if applicable)

Detailed description

I successfully converted a model from pytorch to onnx to openvino, anyway when I load the model for inference, I have follow error: Check 'Dimension::merge(merged_channel_count, data_channel_count, filter_input_channel_count)' failed at ngraph/core/src/validation_util.cpp:341: While validating node 'v1::ConvolutionIE ConvolutionIE_11300 (188[0]:f32{1,256,29,29}, 185[0]:f32{1,256,5,5}) -> (dynamic?)' with friendly_name 'ConvolutionIE_11300': Data batch channel count (1) does not match filter input channel count (256). I don't understand why the issue, because both layers have same batch dimension (1)

I attach the xml of openvino model

Steps to reproduce

converted model from pytorch to onnx converted model from onnx to openvino loaded model for inference

Issue submission checklist

• [x] I report the issue, it's not a question

• [x] I checked the problem with documentation, FAQ, open issues, Stack Overflow, etc and have not found solution

• [x] There is reproducer code and related data files: images, videos, models, etc.

mobilenethead.zip

[jgespino]

Hi @EnricoBeltramo

Do you see the same issue with the latest OpenVINO 2021.1 release? Please provide the full Model Optimizer command used to convert the model, from the XML I don't see --input_shape or --batch was passed. Could you try specifying one of those two parameters?

Regards, Jesus

[EnricoBeltramo]

Hi @EnricoBeltramo

Do you see the same issue with the latest OpenVINO 2021.1 release? Please provide the full Model Optimizer command used to convert the model, from the XML I don't see --input_shape or --batch was passed. Could you try specifying one of those two parameters?

Regards, Jesus

Yes, the openvino version is version is 2021.1.

I tested different commands, anyway I fall always in similar error. I tried configuring the batch size: python mo.py --input_model '/home/ulix/Progetti/pysot/siamrpnmobilenet.onnx' --output_dir /home/ulix/Progetti/pysot/openvinosiamrpnmobilenet/ --data_type FP16 --batch 1 Model Optimizer arguments: Common parameters:

• Path to the Input Model: /home/ulix/Progetti/pysot/siamrpnmobilenet.onnx
• Path for generated IR: /home/ulix/Progetti/pysot/openvinosiamrpnmobilenet/
• IR output name: siamrpnmobilenet
• Log level: ERROR
• Batch: 1
• Input layers: Not specified, inherited from the model
• Output layers: Not specified, inherited from the model
• Input shapes: Not specified, inherited from the model
• Mean values: Not specified
• Scale values: Not specified
• Scale factor: Not specified
• Precision of IR: FP16
• Enable fusing: True
• Enable grouped convolutions fusing: True
• Move mean values to preprocess section: None
• Reverse input channels: False ONNX specific parameters: Model Optimizer version: 2021.1.0-1237-bece22ac675-releases/2021/1

[ SUCCESS ] Generated IR version 10 model. [ SUCCESS ] XML file: /home/ulix/Progetti/pysot/openvinosiamrpnmobilenet/siamrpnmobilenet.xml [ SUCCESS ] BIN file: /home/ulix/Progetti/pysot/openvinosiamrpnmobilenet/siamrpnmobilenet.bin [ SUCCESS ] Total execution time: 26.08 seconds. [ SUCCESS ] Memory consumed: 370 MB.

and configuring all inputs shapes (the model has 2 inputs): python mo.py --input_model '/home/ulix/Progetti/pysot/siamrpnmobilenet.onnx' --output_dir /home/ulix/Progetti/pysot/openvinosiamrpnmobilenet/ --data_type FP16 --input_shape [1,3,256,7,7],[1,3,224,224] --input z,x Model Optimizer arguments: Common parameters:

• Path to the Input Model: /home/ulix/Progetti/pysot/siamrpnmobilenet.onnx
• Path for generated IR: /home/ulix/Progetti/pysot/openvinosiamrpnmobilenet/
• IR output name: siamrpnmobilenet
• Log level: ERROR
• Batch: Not specified, inherited from the model
• Input layers: z,x
• Output layers: Not specified, inherited from the model
• Input shapes: [1,3,256,7,7],[1,3,224,224]
• Mean values: Not specified
• Scale values: Not specified
• Scale factor: Not specified
• Precision of IR: FP16
• Enable fusing: True
• Enable grouped convolutions fusing: True
• Move mean values to preprocess section: None
• Reverse input channels: False ONNX specific parameters: Model Optimizer version: 2021.1.0-1237-bece22ac675-releases/2021/1

[ SUCCESS ] Generated IR version 10 model. [ SUCCESS ] XML file: /home/ulix/Progetti/pysot/openvinosiamrpnmobilenet/siamrpnmobilenet.xml [ SUCCESS ] BIN file: /home/ulix/Progetti/pysot/openvinosiamrpnmobilenet/siamrpnmobilenet.bin [ SUCCESS ] Total execution time: 28.44 seconds. [ SUCCESS ] Memory consumed: 368 MB.

I attach the xml file of latest version generated siamrpnmobilenet.zip

here a link to onnx model and converted version: https://drive.google.com/file/d/1HlRnqdXd9Ziq5y8bTLTHsyziQQbCczRe/view?usp=sharing

And the error (I changed sightly the model in order to reduce the number of inputs) is: Check 'Dimension::merge(merged_channel_count, data_channel_count, filter_input_channel_count)' failed at ngraph/core/src/validation_util.cpp:341: While validating node 'v1::ConvolutionIE ConvolutionIE_16773 (1125[0]:f32{1,256,25,25}, 1122[0]:f32{1,256,5,5}) -> (dynamic?)' with friendly_name 'ConvolutionIE_16773': Data batch channel count (1) does not match filter input channel count (256).

If I convert only the backbone part (with only second input [1,3,224,224]), it works fine

[EnricoBeltramo]

There are some news about that? I was able to run successfully the network using ONNX runtime with Openvino provider, so I suppose the network is ok and able to run in Openvino. May be the problem is in the network conversion or how to load it in Openvino inference. I would to use native Openvino in order to manage better the Openvino optimizations.

[luoyiroy]

Hi @EnricoBeltramo This issue has been fixed by #3056 You can now inference the model with the master benchmark to verify it.