How to change RAM limits on X-Cube AI for a dual core MCU.

LMain.1 · ‎2023-05-13

I am using a STM32WL55JC2 Nucleo Board that mounts a dual core M4/M0+, I changed the linker script in order to have more RAM and FLASH on the M4, but this change is not reflected on the X-CUBE AI interface when adding a new network.

Is it possible to do so? Since I am trying to convert a larger network.

fauvarque.daniel · ‎2023-05-15

The max memory given in X-CUBE-AI is the one reported by the STM32CubeProgrammer database, it is not linked with the linker script.

The only workaround to your problem is to generate the project with a model that will fit in the memory, then call the stm32ai command line by hand with your model and change the generated project with the generated files for your project.

In order to give better visibility on the answered topics, please click on 'Accept as Solution' on the reply which solved your issue or answered your question.

View solution in original post

fauvarque.daniel · ‎2023-05-15

The max memory given in X-CUBE-AI is the one reported by the STM32CubeProgrammer database, it is not linked with the linker script.

The only workaround to your problem is to generate the project with a model that will fit in the memory, then call the stm32ai command line by hand with your model and change the generated project with the generated files for your project.

In order to give better visibility on the answered topics, please click on 'Accept as Solution' on the reply which solved your issue or answered your question.

LMain.1 · ‎2023-05-15

Thank you! That's what I thought.

I think an option to generate the network code would be to use the STM32Cube.AI Developer Cloud as well.

fauvarque.daniel · ‎2023-05-15

Yes of course you can do that.

You can also take another M4 with more ram/flash that meets your network and then copy the generated files into your initial project

In order to give better visibility on the answered topics, please click on 'Accept as Solution' on the reply which solved your issue or answered your question.

RMand · ‎2023-08-20

This 'workaround' makes no sense whatsoever. If the OP were able to generate a model that 'fits in the memory' why would he even bother asking this question in the forum?