Hi all, I am training a model with Fine-tune and gpt3.5.
When testing it in the playground tab, I notice that it does not respond with the example questions I have generated. This makes me understand that the gpt3.5 model is not the right one for what I need?
If I enter the global theme, it answers 90% of the questions well, but I have to give it the introduction.
Does anyone have any ideas, examples or documentation?
Hi and welcome to the developer forum!
Fine tuning does not teach the model new data, it teaches it new ways to think and new patterns to respond to and with.
Think of fine tuning the model with the written works of an author, the model will now be able to write in the style of that author, but it will not remember any of their books.
You may need to look into “Embeddings” for information retrieval, or a combination of them both for specific Q/A styles and responses.