Tuning text models with RLHF - custom model vs base model offerings?

[rosh]

1

2

3

Looking through several instances of the information provided here, it seems that only the 6 base models offered here can be RLHF tuned. What if we would like to bring our own SFT model and use it, is that a possibility? Or using a custom SFT task specific model for RLHF tuning? Where can I find these resources?