Post #3892071
2026-07-15 18:32 UTC
Has anyone here tried out one of the small language models, tinyLLM for example? Are they truly self-contained? Don't need an internet connection to run?
To run on against the contents of /home/tomj.
Does this idea piss people off? Be blunt. Assuming it would be actually self-contained and not a tool tethered to one of the corps.
Replies (1)
-
@petrillic@hachyderm.io 2026-07-15 19:29
@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.