How to split train and eval after reward_model_annotate?

[zjuruizhechen]

As titled.

Thanks.

[yumeng5]

Hi,

In our experiments, we did train/test set split in accordance with the original train/test prompt split of the UltraFeedback dataset. However, you may also perform your customized train/test split based on your specific needs.

Best,
Yu