search

opendatalab / HA-DPO

Beyond Hallucinations: Enhancing LVLMs through Hallucination-Aware Direct Preference Optimization
https://opendatalab.github.io/HA-DPO
Apache License 2.0
66 stars 6 forks source link

Dose the ref_model really need? #8

Closed liuxingbin closed 4 months ago

liuxingbin commented 4 months ago

Hi, Thanks for the great work, I am curious about the red model, Can I set the ref_model to None, as I see here: https://github.com/opendatalab/HA-DPO/blame/main/ha_dpo/trainer/base_dpo_trainer.py#L298. When you set reference_free=False, you do not need a ref-model to save computation resources.