Issue with data mapping on new fine tuning job

We are not getting proper response from our prompt using fine tuning model with our own dataset. When we are asking for information the fine tuned model is not pointing to our own dataset instead it is pointing to other dataset and giving irrelevant information.

  • Format of jsonl file is message role and content type format.

Where is your dataset?

Just so you are aware, Fine-Tuning will not encode new data, it only shows the model new patterns and ways of matching those patterns.

If you wish to store large datasets then you should investigate a RAG pipeline where your data is stored in embeddings and then searched for matching sections of that data using a vector database.