Question about DPO learning rate - comparison to neural-chat-7b-v3 training #25
Description
The learning rate default in the dpo recipe config is set to 5e-7 and https://huggingface.co/Intel/neural-chat-7b-v3 was trained with a learning rate of 1e-4 (using of course a different data set https://huggingface.co/datasets/Open-Orca/SlimOrca).
However, I am wondering about the significant difference in lr and yet both models seem to perform well. Any insights about that, that you can share?
Thank you