petals icon indicating copy to clipboard operation
petals copied to clipboard

Published 20 hours ago • verified publisher icon bigscience-workshop

content of 'labels' when doing prompt tuning of llama-2 on QA

Open caochuxueeee opened this issue 3 months ago • 0 comments

I saw that on classification tasks, the 'labels' are target values. When using CausalLM model to tune it on QA dataset, the format should be

  1. input_ids: Q; labels:A or
  2. input_ids: Q+A, labels: Q+A or
  3. input_ids: Q+A, labels: IGNORE_TOKENs+A ?

Thank you!

caochuxueeee avatar Mar 12 '24 06:03 caochuxueeee