[Anonymous]

Hi everyone,

I installed the SOTA Deep Learning Toolkit and had access to the beta version. Yesterday, I tried converting the exact model featured in the demo video. However, after starting the conversion process, Notepad opens up and nothing further happens on the LabVIEW side (2024 Q3) — the conversion doesn’t complete or display any output.

Has anyone encountered a similar issue? Any insights or troubleshooting tips would be greatly appreciated.

Thanks in advance!

Topic

[Anonymous]

Up !

[Youssef MENJOUR]

Hi Akram,

Sorry for the late response. We no longer support LabVIEW 2024 Q3. Please switch to version 2025 Q1.

Could you specify which model you are referring to? Please be more precise.

[Michal Kozák]

Hi All,

I encoutered a very similar issue in version LabVIEW 2025 Q3.

In my case, I loaded several models in h5 or onnx format in FIG and tried to export them.

• If I choose Complete LabVIEW exportation mode, only txt summary of a graph similiar to the one above emerges and no scripting is performed and FIG disappears. (Then I must restart LabVIEW to be FIG operational again).
• If I choose Normal, a Untitled VI emerges with Graph object (with some data inside) but without any other code. Txt summary of a “empty” graph popped out.

I tried to load some my models, even some yours examples (Accelerator\sobel filter\Model\sobel_filter.onnx or Dueling DDQN\LunarLander\training\trained_model.onnx).

Is there anyone else with the same issue in LV 2025? Or am I doing anything wrong?

Thank for any advice,

Michal

[Youssef MENJOUR]

Sorry for the late reply.

Your screenshot actually shows a “Normal” export. This mode generates the LabVIEW class associated with the ONNX file, along with a small summary (.txt file) listing the model’s inputs and outputs to help you run it later. You can then modify the class through Get/Set methods, but if you want more precise and modular control, the “Complete” export option is the better choice.

That said, based on your description, even with the “Complete” export, the model isn’t being scripted. I tried to reproduce this in LabVIEW 2025 Q3, and scripting worked fine on my end. Here’s what I did:

Opened LabVIEW
Launched FIG
Clicked Load File and selected sobel_filter.onnx
Clicked Export, then chose Complete at the bottom of the window
→ At this point, a new VI opened and scripting automatically started, recreating the full graph.
Can you double-check that you selected the correct export type? The first four buttons in the Export window only define the type of summary that gets generated. It’s the two buttons at the bottom that determine whether the graph is fully scripted or if only the class (containing the graph) is created.

 

To help narrow down the issue, could you share a bit more about your setup? For example:

Which version of Windows are you running?
Do you see any error messages or warnings during the export?

• This reply was modified 1 month ago by Youssef MENJOUR.

[Corentin Maravat]

Hi Michal,

Sorry for the inconvenience. We found that the issue was caused by a Windows 11–related bug.
This will be fixed in the upcoming release of the DeepLearning toolkit, which should be available tomorrow.

[Corentin Maravat]

Hi again Michal,

The new version has just been released, and on our side it fixed the Windows 11 issue with FIG. We suggest trying with this version, and if it still doesn’t work, we’ll be glad to help investigate further and find a solution.

Best,
Corentin

[Michal Kozák]

Hi Youssef and Corentin,

thank you for your replies and such a fast response, it is as you both say – normal export was ok and after update FIG is perfectly functional. Thanks a lot.

Best, Michal.

• This reply was modified 1 month ago by Michal Kozák.

[Author]

Replies