FAQs
cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Search instead for 
Did you mean: 

I use cubemx.ai convert a simple conv-pool model bulit by keras, but when it was mounted on the board of STM32L452RE, all the outputs were -NaN

ZShi.1
Associate

‎2020-09-24 10:14 PM

the model I have bulit:

0693W000004H4WLQA0.png​the config of my project:

only use cubemx.ai for Application Template，without any other config in my project

0693W000004H4X4QAK.png

0693W000004H4XEQA0.png0693W000004H4XJQA0.png

in keil project, run the model and see the output in debug window, I find it's all -NaN:

0693W000004H4Y2QAK.png

in earlier experiment, when I use only ONE conv-pool layer, it can work smoothly, however, if I add another conv-pool, the outputs get NaN immediately. WHY?? is something wrong with my model? or any other problem?:loudly_crying_face: :loudly_crying_face:

Labels:
0 Kudos
0 REPLIES 0
Top