Hello,
I am currently using X-CUBE-AI library to run neural networks on a stm32h747. I successfully generated and validated on board single input and output networks, but when I execute the validation through CLI with a multi output network, I get an internal ERROR (text below).
I can't find any information about this error on the web or in the documentation. Where could I find any information on the error ?
$ stm32ai validate --mode stm32 -m Data/models_stm/343/model.tflite -d 7500000 --verbosity 2
Neural Network Tools for STM32 v1.3.0 (AI tools v5.1.1)
Using TensorFlow backend.
-- Importing model
-- Importing model - done (elapsed time 0.201s)
-- Building X86 C-model
-- Building X86 C-model - done (elapsed time 0.517s)
-- Setting inputs (and outputs) data
Using random inputs, shapes=[(10, 6400)] dtype=[float32]
-- Running STM32 C-model
ON-DEVICE STM32 execution ("network", None, 7500000)..
<Stm32com id=0x7f0423450eb8 - CONNECTED(/dev/ttyACM0/7500000) devid=0x450/STM32H743/53/50xx and STM32H745/55/47/57xx msg=2.1>
0x450/STM32H743/53/50xx and STM32H745/55/47/57xx @480MHz/240MHz (FPU is present) lat=4 Core:I$/D$
found network(s): ['network']
description : 'network' 1->[43]->2 macc=49850547 rom=259.68KiB ram=208.88KiB
tools versions : rt=(5, 1, 0) tool=(5, 1, 1)/(1, 3, 0) api=(1, 1, 0) "Tue Aug 11 09:40:31 2020"
Running with inputs (10, 80, 80, 1)..
.INTERNAL ERROR: E100(Stm32comError): Stm32com exception: Invalid state: 4 instead 2
