- My training data has 16 categories with multiple subjects with multiple Q&A.
- I asked ChatGPT and it says I can also add meta data like this {“prompt” : …, “completion” : …, “subject” : …, “summary” : …, “category” : … } but when I ran fine_tunes.prepare_data again it removes all my additional keys:
Based on the analysis we will perform the following actions:
- [Necessary] Remove additional columns/keys: [‘subject’, ‘summary’, ‘category’].
Keeping only {“prompt” : …, “completion” : …}
How can I fine-tune with meta data? What is the best approach to fine-tune a model with multiple categories, summaries and questions and answers?