Do you have any plan to add fine tuning guide in HuggingFace?
#9
by
kjh76363
- opened
I've tried your algorithm and overall it seems to be very good, so I'd like to apply it to my application, but I don't know how to do it in the usual way (preprocessing with ImageProcessor.from_pretrained and fine-tuning with Trainer or Torch). Do you have any resources to help me?
Hi, Joohyeok, thanks for your interest! You can use the full codes on the GitHub repo: https://github.com/ZhengPeng7/BiRefNet. You can set the resume in train.sh to load some well-trained weights, and turn on/off the backbone freezing for fine-tuning.
Many people have been fine-tuning their custom data. You can find some of their problems in the issues there.
ZhengPeng7
changed discussion status to
closed