-
Notifications
You must be signed in to change notification settings - Fork 7
The quality of feedback
Marcus Green edited this page May 27, 2024
·
8 revisions
The quality of feedback is closely related to the quality of the prompt. However it is sometimes not obvious how to refine a prompt to get the response you want. Most development was done against ChatGPT (gpt-4) model. However I am committed to ensuring it works with other systems such as Ollama and the models available to that system. However a prompt that works as you want with ChatGPT will often give quite different answers with models under Ollama.
I have had very encouraging results from the LLama3 model and found the hosting from Groq.com gives excellent performance and a very reasonable price. So far it has cost me US$0.
The responses from the smallest languages I have found not to be useful.