my own security policy
#1
by
Forrest20231206
- opened
If I want to use my own security policy, do I have to reconstruct the data set and train it from scratch.
Our models are very focused on our policy. Yet, it is possible to change the policy prompt for inference and even introduce new categories. As safety usually does not allow for much error, relabeling and retraining is a sensible choice, too.
thanks for your reply.
Forrest20231206
changed discussion status to
closed