YinxiaCao
YinxiaCao
> Hi XU， Thank you for your quick reply! I am just a little confused why the teacher can yield better result than student. In the tranditional mean-teacher mechenism, this...
Hi, @YonghaoXu. Thank you very much. Your work is great. I understand your explanation. I still have two issues and hope to know your thoughts. 1. The target images with...
Hi, @YonghaoXu. Thank you very much. I want to know why the `teacher_net.train()` was not added? If this code was not added, then the batch normalization and dropout layers in...
Hi, @YonghaoXu. Oh, you are right. I feel that this step is easily ignored when we use Resnet or other networks with bn and dropout layers. Anyway, thank you very...
@Junjue-Wang, Thank you very much. I still have one question about the `clip_grad.clip_grad_norm_(filter(lambda p: p.requires_grad, model.parameters()), max_norm=35, norm_type=2)`. I guess the `max_norm` should be changed according to different datasets. Whether...
Oh, thank you anyway. I just want to know why `max_norm` is set to `35` . Do you set it by experiments?
Oh, I see. Thank you a lot.
> Have same issue. Changing user agent to actual(like my web browser) doesn't work for me. Me, too. Have you solved it?
I see. Thank you very much.
Hi Ilker GURCAN, I want to know whether you have successfully built the ossim in Windows.