Fail on STM32N6 with ONNX model: “index 0 is out of bounds for axis 0 with size 0”

[giuug]

Hi,

I am trying to analyze a custom Speech Enhancement ONNX model targeting STM32N6, but the tool fails immediately with an internal error.

Command executed (cmd)

stedgeai analyze -m onnx\best_trained_model.onnx --target stm32n6 --compression none --verbosityST Edge AI Core v3.0.0-20426

Tool version

ST Edge AI Core v3.0.0-20426

Error

TOOL ERROR: index 0 is out of bounds for axis 0 with size 0

 The model is an ONNX export of a neural network used for speech enhancement. I used the code on stm32-model-zoo-services to train this model. Training and evaluation were fine, indeed i've used the .onnx that this phase generates. Then the quantization phase generate another error:

Error (symbolic_shape_infer.py)

Exception: Incomplete symbolic shape inference

What should i do?
@Julian E. 

[Julian E.]

Hi @giuug,

 

It seems to be a compiler bug coming form the node:

_dpgrnn_dpgrnn_0_intra_ln_LayerNormalization_output_0

 

In the ST Edge AI Core 4.0, the error is a bit different:

ValueError: Shape and shape map lengths must be the same: [33, 16] vs. (CH,)

 

The model might be malformed as the infershape fails (onnx).

 

This is transmitted to the dev team but will be worked on depending on other priorities.

 

Have a good day,

Julian