Awesome pre-trained models toolkit based on PaddlePaddle. (400+ models including Image, Text, Audio, Video and Cross-Modal with Easy Inference & Serving)【安全加固,暂停交互,请耐心等待】
hi, I noticed that the U2Net module.py does not have the meta=SegmentationModule in the @moduleinfo decorator. I believe as a result, we cannot access the training_step method to perform fine tuning of the pre-trained model with our customize images set.
Is this exclusion deliberate or is it a missed-out? Is there any quick way to get around this to continue with the fine-tuning? I had tried changing the downloaded copy at ~/.paddlehub/modules/U2Net/module.py, but it seem that hub.Module(name='U2Net') will detect that the code had been changed and re-download the model https://bj.bcebos.com/paddlehub/paddlehub_dev/U2Net.tar.gz and overwrite the changes..
hi, I noticed that the U2Net module.py does not have the meta=SegmentationModule in the @moduleinfo decorator. I believe as a result, we cannot access the training_step method to perform fine tuning of the pre-trained model with our customize images set.
Is this exclusion deliberate or is it a missed-out? Is there any quick way to get around this to continue with the fine-tuning? I had tried changing the downloaded copy at ~/.paddlehub/modules/U2Net/module.py, but it seem that hub.Module(name='U2Net') will detect that the code had been changed and re-download the model https://bj.bcebos.com/paddlehub/paddlehub_dev/U2Net.tar.gz and overwrite the changes..