I’m using the chat completion API (using the python module, currently, but planning to switch to base HTTP soon.)
I ask for a small number of completions – typically 3 – so I have some to choose from, using some admission predicate I apply after getting them back.
Currently, I pass in a certain temperature and it is applied to all three completions.
However, I’d rather make a request saying something like “please give me 3 completions, with temperature 0.0, 0.4, and 0.8”
Is there some way of doing this, other than making three API requests in parallel?
1 Like
Welcome to our dev community. Hope you stick around.
This sounds useful, but as far as I know, it’s not doable at the moment. You would need to send three requests, each with a different temp. Again, though, that sounds like a really useful feature for those who do ask for more than one response.
1 Like
Thanks for the answer, and confirming what I suspected.