Error "Name:'bn_data' Status Message: Invalid input scale: NumDimensions() != 3"

[luan1412167]

Describe the bug
Hi all,
I'm using onnx runtime from source (latest version support spatial BN = 0 ). my model here

Non-zero status code returned while running BatchNormalization node. Name:'bn_data' Status Message: Invalid input scale: NumDimensions() != 3

Thanks so much for your help

Urgency
I must to implement retinaface and arcface on C++ for my thesis at my university.

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 18.04
• ONNX Runtime installed from (source or binary): source
• ONNX Runtime version: latest
• Python version:
• Visual Studio version (if applicable):
• GCC/Compiler version (if compiling from source): 3.15
• CUDA/cuDNN version:
• GPU model and memory:

[hariharans29]

Hi Luan,

Did you convert this model from MxNet by any chance ? It has the same issue as the ArcFace model where the BN nodes are stamped with spatial == 0 and the inputs (scale, B, mean, var) do not conform to ONNX spec -

The input shape [3] is acceptable only when spatial == 1. As it is, this is an invalid ONNX model. Please try using the script I shared in the ArcFace model issue to correct this model and try again.

Closing this, because it is an invalid ONNX opset-8 model.