Fine tuning doesn't quite work that way. You have to format the training data set as request/response. The idea of fine tuning is to get the model to output things in a specific format, style or structure.
Your use case is better suited to RAG. This is where you retrieve data from a large dataset and inject it into the user's request so the AI model has the context it needs to answer accurately.
But that's not a silver bullet and you would need to spend significant time on chunking strategy and ranking of results to hopefully get a decent response accuracy.
Comments
Fine tuning doesn't quite work that way. You have to format the training data set as request/response. The idea of fine tuning is to get the model to output things in a specific format, style or structure.
Your use case is better suited to RAG. This is where you retrieve data from a large dataset and inject it into the user's request so the AI model has the context it needs to answer accurately.
But that's not a silver bullet and you would need to spend significant time on chunking strategy and ranking of results to hopefully get a decent response accuracy.