Hi there!
It looks like you are intending to use fine-tuning for Q&A. Unfortunately, this is not what it is intended for - fine-tuning within this specific context - is primarily designed to get a model to behave in a certain way, e.g. adopt a certain style in output or approach a task in a specific way. It is not a recommended way to get the model to absorb specific knowledge.
As a similar topic came up just a few days ago, I’ll refer you to another thread where I have discussed this at greater length along with references to other resources.
Bottom line is that you should be looking at retrieval augmented generation (RAG) approach for your use case.
Let us know if you have any follow-up questions once you’ve had a chance to review the other thread.
Good luck in any case.