Release of SFT tuned model
#8
by
yakazimir
- opened
As I understand it, these instruction tuned models are released with SFT + DPO training, presumeably trained in different stages. Is there a plan to release just the SFT tuned models? This would be quite helpful for those of us working on DPO tuning.
zkz
Would it have much of an effect whether the model is aligned or not? Meaning - you can still align the model further to your preference.
