Support for LayerNormalization in STM32Cube.AI

[M_Benz]

Hi everyone,

I’m trying to deploy a wav2vec model using STM32Cube.AI and ran into an issue during onnx conversion.

NOT IMPLEMENTED: Unsupported layer types: LayerNormalization

I’ve read the documentation at supported onnx operators and it seems that LayerNormalization is not currently supported.

I have a couple of questions:

• Are there plans to support LayerNormalization in future releases?

• If not, does anyone know a recommended workaround other than just modifying the model to use another type of normalization?

[Julian E. · Accepted solution]

Hello @M_Benz,

Here is what the dev team says:

"Support of LayerNormalization is under development. It will be supported in the next ST Edge AI Core release, but only for software targets"

Software target means that it will not be supported using the NPU.

You can expect to see the next version of the Core in November, but it is still under discussion.

Have a good day,

Julian