After two months of training, my poor old Vaio has finally handled 30k iterations. LOL
So I decided to try Google colab, and I got it to work! Thanks to the great advice found on these forums, of course.
Well, now that I'm able to get training on Google colab, I was wondering if it is possible to perform the extraction process as well, since there I don't have (my laptop, I mean) enough free VRAM for parallel processing and that means painful hours to barely get some dozens of faces. I find it odd that this hasn't been discussed - or at least I couldn't find such discussion.