Why does the model of first stage for finetune train all parameters instead of using LoRA?

[ohheysherry66]

Thank you for your great job! The question is described in title.

[fcjian]

@ohheysherry66 You can also fine-tune SD using LoRA. We tried it at the time and it achieves similar performance.

[ysysys666]

Excuse me, have you successfully reproduced the experimental results of the detector in the paper? I am looking forward to your reply very much! @ohheysherry66