Error "dlquantizer workflow failed" Deep Network Quantizer App

I have tried to follow the example "Import a dlquantizer Object into the Deep Network Quantizer App" from the link https://it.mathworks.com/help/deeplearning/ref/deepnetworkquantizer-app.html, but when I used the calibrate function I got this error:

Does anyone know how can I fix the problem?

[Asvin Kumar]

That example works well for me. I used squeeznet for my convenience.

Looking at your screenshot, it doesn't look like you're importing a dlquantizer object. It also doesn't seem like we're both using the same network. Could you share steps to reproduce this error that you're seeing?

[Will Kaczmarek]

@Asvin Kumar I am having the same error, following the steps listed here. Once I click "Quantize and Validate", it thinks for a second then gives me this error. Calibration seems to work fine for me. Below is the error I receive when running from the Command Window:

Error using dlquantizer/validate
dlquantizer workflow failed.

Error in quantizeTest (line 66)
valResults = validate(quantObj,augimdsValidation,quantOpts);

--and when I try to break on errors--
dlquantizer.validate is not currently in a debuggable state. Skipping the frame.
In base workspace.

[Asvin Kumar]

Hi @Will Kaczmarek

Would you mind sharing the following:

1. Your trained network that you used to create the dlquantizer
2. The calibration and validation data sets
3. Any script that you're using (quantizeTest?)
4. MATLAB Release details (save output of ver command)

Put it all in a MAT file. Save them all even if they're the same data from the example(s).

I would recommend creating a new question thread along with your attachments instead of tagging it to this one. You can mention me in the question comments.

[Asvin Kumar]

@Giusy Giulia Tuccari I notice that you're using resnet50. Can you try the following command and let me know the errors, if any?

calResults = calibrate(quantObj, aug_calData)

If you're seeing errors, it would be nice if you can attach the details I requested for in the comment before this one.