STM32N6: Many compile errors after changing NN model with CubeMX + X-CUBE-AI

[实在太懒于是不想取名]

Hi,

I'm working on STM32N6 with CubeMX + X‑CUBE‑AI to deploy a neural network model.

• The first AI model works fine: no compile errors, runs normally.
• When I replace the neural network model in CubeMX and regenerate code, I get dozens of compile errors (undefined reference, multiple definition, missing files, etc.).
• • Clean project in IDE
  • Delete build / Objects / Listings folders
  • Reimport the new model in CubeMX
  • Regenerate code completelyI have tried:

But the issue still exists. It seems old AI files are not cleaned properly when changing models.

Has anyone faced similar issues on STM32N6 when switching AI models with CubeMX + X‑CUBE‑AI?
How did you solve this?

Thanks for help.

[chanchal94]

I am facing the similar issue while working with STM32 cube AI studio + cube IDE. 

As soon as I initialize the peripherals in MX, it starts giving many errors. Before that it was building fine.

Changing in cube MX and regenerating code breaks the AI middlewares.

I am still unable to solve it!

Thanks & Regards

Chanchal

[实在太懒于是不想取名]

I totally feel you — this issue is so frustrating and such a huge time-waster.

Every time you just tweak a small peripheral setting in CubeMX and regenerate code, it immediately breaks the entire Cube.AI middleware, throws tons of weird errors, and makes the whole project unbuildable. It’s incredibly annoying how unstable this workflow is.

I’ve ended up having to copy a completely clean, non-AI project every single time just to avoid this mess — it’s not a real solution, just a  workaround we’re forced to use.

Really hope ST fixes this  workflow soon. It’s such a pain dealing with these unnecessary bugs.