Fine tune specific models #441
Description
Description & Motivation
I was thinking, with all the ways in which there is optimisation being done, certain ones that pass a mark could be collected for fine tuning the model? Is there something on the roadmap for auto fine tuning the models given a task? There is only so much room for few shot prompting ultimately.
Pitch
What if while optimizing, a dataset for fine tuning is collected and then on command/ trigger it would fine tune the models and then swap. This would under go testing etc.. and again train and swap? Or revert and retrain and swap. You get the point.
Alternatives
No response
Additional context
No response