Any plans for being able to fine tune models in rustformers llm?
moonstripe opened this issue · comments
Just curious if there are any plans to implement a fine tuning feature that loads a training set and fine tunes a model for specific tasks.
Closing here, since I got a response on Discord.
For those interested, here's the response I got:
I think I recall a brief discussion about this earlier and the answer was to fine tune with Python/torch etc and then quantize for ggml etc
^^^ codesoda
that's correct, yeah - we don't have any current plans as ggml doesn't support it + getting it right is quite complicated
I'm not a huge fan of the Python ML ecosystem, but limited resources and time mean we have to pick our battles
^^^ philpax