alibaba / ChatLearn

A flexible and efficient training framework for large-scale alignment tasks
Apache License 2.0
216 stars 17 forks source link

[Feature] dpo场景 为什么 onlinedpo效果不如dpo,rlhf会有下降? #159

Closed yiyepiaoling0715 closed 1 day ago

yiyepiaoling0715 commented 4 days ago

Is your feature request related to a problem? Please describe. A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like A clear and concise description of what you want to happen.

Describe alternatives you've considered A clear and concise description of any alternative solutions or features you've considered.

Additional context Add any other context or screenshots about the feature request here.

image

adoda commented 2 days ago

开源HH数据集中没有coding相关数据,这块效果会比较随机。这里关注avg指标即可。生产训练中,可以按需准备训练数据。