2023-06-06 04:02 AM - edited 2023-11-20 03:50 AM
I am currently trying to run an ML model on the STM32H7A3. The model is uploaded correctly and successfully analysed.
The first issue is that when generating the code, the middleware folder which contains ST>ai (which contains ai_platform) is deleted by CubeIDE. I've had to overcome this by copying the middleware folder from an older project, as there is no way to generate it. Hopefully, ST knows about this.
Anyway, when I run the model, the correct input and outputs are present. This means I can fill in the input buffer. I do not enable the use activation buffer for input buffer as shown in the screen show:
When I run the model batch is returned as 1, and the error type is 17 and code 16, which means:
AI_ERROR_INVALID_VALUE = 0x17,
AI_ERROR_CODE_ARRAY = 0x0016
I can't find any documentation on the error codes. I also can't seem to fix the problem as everything else works OK.
Is this a known issue with CubeAI at the moment? or is it because the middleware is not being generated anymore? I've seen other posts where ST has answered, saying that there will be an update in mid-June to Fix CubeAI with update 8.1.
Thanks
Solved! Go to Solution.
2023-06-07 02:37 AM
For the middleware folder beeing removed, the fix will be in 8.1. in the mean time just use the project generation option not to remove the files not generated.
For the input/output issue, this is probably due to the way you pass the data to the model, you can look at this thread for an example https://community.st.com/s/question/0D73W000000bAqBSAU/detail
Regards
Daniel
2023-06-07 02:37 AM
For the middleware folder beeing removed, the fix will be in 8.1. in the mean time just use the project generation option not to remove the files not generated.
For the input/output issue, this is probably due to the way you pass the data to the model, you can look at this thread for an example https://community.st.com/s/question/0D73W000000bAqBSAU/detail
Regards
Daniel