ML Deformer provided sample retrained does not provide the same data?

Hi, I am currently testing and trying to understand provided ML deformer sample went through documentation for it multiple times and I can not find the solution. Retraining the ML with the same settings as provided and with adjusting the setting before training always gives me results where:

“Num Active Morphs” is 100%
“Nearest Neighbor Ids” is -1

I have tested many different setting for training from Compression, Delta Threshold, PCA (ON/OFF), different num basis and more.

I also tried with my own mesh from Houdini and after training it I get the same result as with sample data.

Feels like changing any settings always gives the same result.

It almost makes me wonder if I maybe missing some python or nvidia plugin/extension.

How can I fix this and get correct result?

Thank you in advance!

Solved

I was missing NN layer in sections