Can not get the desire answer after train the model

I have a huge descriptive document. I want to fine tune my model such a way that if some one ask some question the system will generate the answer by going through the predefined document.

What I have Done.
I have convert my entire document as question and answer in which the questions are in prompt and the answer is in completion.I have done some separation also to segregate the prompt and completion.
Here are one example of prompt and completion.

{“prompt”: “Question:What are the objectives of an Underwriter?\n\nAnswer:\n\n”, “completion”:" The objectives of an Underwriter are to identify and verify income available to meet the mortgage payment, other shelter expenses, debts and obligations, and family living expenses; and to evaluate whether the verified income is stable and reliable, anticipated to continue during the foreseeable future, and sufficient in amount. END" }

Following is my data

But I cant get the proper answer after train the model.

{“prompt”:“Can I use a personal loan for any purpose?”,“completion”:" Yes, personal loans are versatile and can be used for a variety of purposes. Whether you need funds for consolidating debt, home improvements, medical expenses, wedding costs, or even a vacation, you can typically use a personal loan for these purposes. However, it’s important to borrow responsibly and use the loan for legitimate needs. Be sure to check with the lender if they have any specific restrictions on the loan’s usage###"}

use thiis format
and if you are using fullstop at the end of completion so plaese make sure you set the value of stop sequence .END