[Question] about fine-tune Qualitative question and Quantitative question

[phamkhactu]

I have fine-tuned LLama2 with LoRa, I found that:

• Model can generate Qualitative question very excellent. For example:

  
  [Question]: Who is Donal Trump?

[Answer]: Donald John Trump is an American politician, media personality, and businessman who served president of the United States from 2017 to 2021. Trump graduated from the University of Pennsylvania with a bachelor's degree in economics in 1968.

However, with quantitative question, model not good. Model can not extractly response what I want. 

[Question]: Which presidency number was Donald Trump? [Answer]: Donald John Trump is an American politician, media personality, and businessman. He graduated from the University of Pennsylvania with a bachelor's degree in economics in 1968.

The answer should be: ``` Donald Trump was the 45th President of the United States.```

How can I enhance performance model for quantitative question?
Thank you.

[hunxuewangzi]

Hello, may I ask if you changed the model directly when fine-tuning llama2? Would it have any impact? I'm not sure if you could use this code to fine-tune Llama-2-chat. Thank you

[phamkhactu]

Hello, may I ask if you changed the model directly when fine-tuning llama2? Would it have any impact? I'm not sure if you could use this code to fine-tune Llama-2-chat. Thank you

Hi @hunxuewangzi, I am so sorry, I don't provide correctly. I use Llama-2-chat-hf