alaa-shubbak
commented
2 years ago I am trying to train a yolov3 model on my custom dataset in a form of soft-teacher. unfortunately , i got such an error. .
TypeError: grad_clip={'max_norm': 35, 'norm_type': 2} in child config cannot inherit from base because grad_clip is a dict in the child config but is of type <class 'NoneType'> in base config. You may setdelete=Trueto ignore the base config
i did not understand what i should do to solve it ,and run effectively the training
Hello, Thanks for such a nice work that have been done . I would like to ask about using such code and config at which the teacher model will be different that student model. for example; teacher model could be faster RCNN( resent 101 ) and student model will be Faster RCNN (resnet 50)
is that possible ? what shall i do for config the system like that? I could not understand where shall i change the teacher and where shall i change the student in the model. any help?
what about using it in some different way, like yolo v3 as teacher and faster RCNN as student? do you think that could be worked ? did you try it before?
looking forward to your answer. thanks.