【Bug已解决】DPOTrainer: ref_model is None during eval when gradient_checkpointing=True 解决方案
【Bug已解决】DPOTrainer: ref_model is None during eval when gradient_checkpointingTrue 解决方案原始报错:DPOTrainer: ref_model is None during eval when gradient_checkpointingTrue
场景:用 DPOTrainer 做偏好训练,开了 gradient_c…
2026/7/22 5:30:14