Customising models · Question 12 of 12
A team appends 30 examples to every prompt so that the model keeps to a special format. With millions of calls per month the token costs become high. Which options make sense?
2 answers are correct. Hard
Show answer
Answer
- C Fine-tune a model on the format
- E Cache the unchanging block of examples with prompt caching
Explanation
At high volume fine-tuning can pay off, because the examples no longer need to be in every prompt afterwards. Prompt caching lowers the cost of the recurring block. More examples raise costs further, and own pre-training is out of all proportion.