Problems with running ruimtehol on Windows R

[trinitrobenzol]

Hi! Using ruimtehol on a Mac and happy with the result! Thank you. But then I tried to run the same script and the same training-label datasets on Windows. Something very strange, the result is completely different on PC. Looks like completely untrained. R version is same (4.0.3) , all packages like UDPipe are updated to last versions. What can be the problem ? (have tried dif 2 Mac and 2 PC).

[jwijffels]

Initial randomisation of the embediings?

[trinitrobenzol]

All the same, I load the same one excel file

[jwijffels]

Can you please provide reproducible examples.

[jwijffels]

were you training wit ngrams > 1?