I'd like to use a fine-tuned model in the Chat API (as opposed to the completion API). I asked ChatGPT about this, but it thinks I can, but then when shown copy pastes from the docs, it expresses uncertainty. Which is cool, but yeah, I'd still like to know... so I'm here : )
#Can I use my fine-tuned model in the Chat API?
next sequoia
wanton whale
I'd like to use a fine-tuned model in the Chat API (as opposed to the completion...
The fine-tuned model is the model you use... Also the new Chat models cannot be fine-tuned. What purpose do you have that requires the use of a fine-tuned model?
rotund flax
You could (in theory) have a [USER]: and [ASSISTANT]: field with the stop token being [USER]:, but I’m not sure how effective this would be
next sequoia
I want to try to "teach" ChatGPT to be more expert at particular subject, namely my subjects: Creativity and Innovation. For example we do really high level brainstorming with groups of scientists. I'd love to teach it to partake. I could give it giant piles of data on how to generate science ideas for a science question, or build on one idea to generate many more.