Download no longer available on ST Edge AI Developer Cloud

[francovaro]

Hi

 

After the latest update, I can't download C Code from Edge Ai Developer Cloud, there are no boards in the list.

 

 

 

fv

[Julian E.]

Hi @francovaro,

 

This is indeed the case. it seems to be a bug for now.

 

So how to work in the meantime? it depends on your target:

 

Non NPU target:

We advise that until March, for "normal" stm32, you use the version 2.2 of the core (so X cube AI 10.2 and you should also be able to use it in the dev cloud). This is because the new 3.0 version is mainly a NPU improvement.

 

For NPU target (your case):

CubeMX do not really support the N6 correctly, so everything is available directly in the ST Edge AI Core install folder. Which means that you can use the 3.0

 

The validation application is in C:\ST\STEdgeAI\3.0\Projects\STM32N6570-DK\Applications\NPU_Validation

 

Here is a guide on how to use it, for more details please visit: https://stedgeai-dc.st.com/assets/embedded-docs/stneuralart_getting_started.html 

 

• Go ST\STEdgeAI\2.2\N6_Scripts
• Copy your model in \models
• Run the generate command, in your case, something like:
• stedgeai.exe generate --model ./models/model.onnx --target stm32n6 --st-neural-art profile@user_neuralart.json

 

Then edit the config.json file, here is mine for reference:

 

You need STM32CubeIDE and most paths can be found by going to the pluging folder of the local install of STM32CubeIDE, searching for each element and copying the path.

• Save the config file
• Plug your N6 dk board on dev mode
• Then, still in the N6_scripts folder, run the command: Python n6_loader.py

 

With the configuration done, you should not have an error. If you get an error for the make path, add it to the configuration file. This is what you should get:

Finally, run the following validate command:

stedgeai validate --model ./models/model.onnx --target stm32n6 --mode target -d serial:921600

 

Note: if you unplug the N6, you will need to reflash the validation firmware with the n6_loader.py before running the validate command.

 

by default, it uses random data as input. It would be better if tested on a real input dataset using the option:  -–vi dataset.npz

 

It can take quite a while, but you should get something like this:

 

Have a good day,

Julian

 

[Julian E.]

This should be fixed.

[francovaro]

Yes it's fixed!

 

thank you