My understanding was that fine tuned models take a base model and train it on specific data.
I am having trouble understanding llamaindex, why does it use an openAI API key to work? Shouldn’t it be able to make inferences offline without the use of any third party apps.
llamaindex is a framework to connect LLM’s together the API is required if you wish to use any of OpenAI’s models and services, ada-002 for embedding, davinchi-003 for base model fine tuning, etc, etc.
I should also note that fine tuning does not teach the model new data directly, it teaches it new ways to think, new patterns to detect. Embeddings usually do a better job of encoding new information and fine tunes encode new ways to process data.
The base models like davinci are only available through the OpenAI API. That being said, I also recommend embeddings over fine-tuning unless you have a really specific use-case