Post #3892070

@tomjennings@tldr.nettime.org so, we have internally used small language models trained on our own data for classification purposes and had really good results. But it was trained on our own data on our own hardware. And honestly, I think the training time was a couple hours on a single GPU.