78 create prodigy app #79
Conversation
…o press the accept button)
…reate-prodigy-app
|
Hey @RFOxbury, first off, thanks a lot for getting started with this - it's been super helpful! I've now pushed the changes to the recipe and an expanded readme with instructions. I've also moved a bunch of code in a "scrap" folder, as I ended up not using it for this PR (which is about setting up the platform). I've gitignored the scrap folder for now as it was throwing up formatting errors from git commit hooks. I can perhaps still reuse some of it when I'll do some quick analysis of the evaluation results. That's just to say that we still have the code you wrote for eg fetching stuff from s3 or viewing annotation results, but it's at the moment not visible for simplicity purposes. It'd be great if you can take a look and see if everything makes sense - no need to spend too much time with the review or re-running all the scripts. Simply checking the readme and looking at the changes in the code will also be fine. Thank you! |
| prompt = MessageTemplate(role="user", content=question) | ||
| response = TextGenerator.generate( | ||
| model="gpt-4", | ||
| temperature=0.6, |
There was a problem hiding this comment.
Just curious, why is this the temperature setting?
There was a problem hiding this comment.
Good question - not a really good answer, but Kostas said this value worked well for him in other tasks, which is why I've left it at 0.6.
In principle, as I understand this value would suggest a balance between stable (temperature=0) and more random (max temperature=2) modes of operation, with a preference towards the more stable end.
No description provided.