Is there any hope for OpenAI to allow us to actually train our Assistants? I have been able to take advantage of the function calling feature to great affect, however what I really want to do is train my model so that it doesn’t have to call out to fetch a ton of data every time. The function calling has limitations too as to how many tokens can be processed for a given prompt. If my Assistant had a lot of background information it was trained on, it could create a lot better experience.
The new Assistants v2 now allows you to use a fine-tuned version of 3.5. Perhaps that’s what you’re looking for?
Release notes here: https://platform.openai.com/docs/assistants/whats-new
1 Like