Ok thanks for telling me your reasoning. But… when i use a model which was built ob pirated data and i am a pirate myself - i think its fine.
Comment on Self-Hosting Survey 2026: What are your favorite self-hosted apps?
gwl@lemmy.blahaj.zone 6 days agoWhere’s the training data from?
loppwn@sh.itjust.works 6 days ago
brucethemoose@lemmy.world 6 days ago
If you run Nemotron, open datasets.
gwl@lemmy.blahaj.zone 6 days ago
Okay, but we know most don’t
brucethemoose@lemmy.world 5 days ago
Admittedly, no, but this is slowly improving. Stepfun published their SFT dataset, smaller labs are publishing their datasets for task specific tunes. I believe there was another Chinese lab that published a lot of pretraining data, but I can’t find it in my history at the moment.