Developer Cloud network.c Version Test - STM32N6570-DK

[Duckpaddle]

Yesterday I built a model from a 256x256x3 .tflite. The Cloud produced a zip file that included a  network_atonbuf.xSPI2.raw file. This file is about 27KB in size.  

What triggered the change from the network_data files?

The network.c knew about this .raw file and pointed at it 0x71000000.  That's all good. I just rename it to .bin and loadi it with CubeProgrammer.  But what I don't understand is the test for version: 

#if LL_ATON_VERSION_MAJOR != 1 || LL_ATON_VERSION_MINOR != 1 || LL_ATON_VERSION_MICRO != 3 || LL_ATON_VERSION_DEV != 262
#  error "Possible mismatch in ll_aton library used"
#endif

Working with the STM32N6-GettingStarted-ObjectDetection as my base this version test failed in Make.  I just commented it out. 

Am I going to have a problem?

[hamitiya · Accepted solution]

Hello @Duckpaddle 

By default, when using ST Neural-ART, weights are produced as "raw" (binary) files instead of .c files. They need to be loaded via STM32CubeProgrammer at the specified address (here, 0x7100_0000).

In order to change this behavior, you can do it by editing the memory pools description:

- in C:\ST\STEdgeAI\4.0\Utilities\windows\targets\stm32\resources\NPU\STM32N6xx

- For XSPI2, replace FORMAT_RAW with FORMAT_C

Code you are refering to is in order to prevent mismatch between Neural-ART Library in your project and the one used to generate the code.

If you have mismatch, it means you used a different version of ST Edge AI Core.

LL_ATON 1.1.3.262 refers to ST Edge AI Core 4.0.0-20500 (latest available)

Best regards,

Yanis

[Duckpaddle]

For the other people that might have this issue and me after I forget,

I removed the comment block around the version test in network.c

I deleted the ./Middlewares/stedgeai-lib/Npu, then I copied my new ST Cloud subdirectory  into the Npu subdirectory under ./Middlewares/stedgeai-lib/Npu 

After a clean and make it worked great.