LvGuangzu
LvGuangzu
> 谢谢你的工作。我导入模型，并将其推送到GPU，但错误发生了：运行时错误：参数#2的张量“权重”在CPU上，但期望它在GPU上（同时检查参数cudnn_batch_norm）期待您的回复！ Have you solved this problem?
I tried to run the model from the GPU, but he told me there was weight on the CPU in the middle
> I tried to run the model from the GPU, but he told me there was weight on the CPU in the middle have solved
Have you solved this problem? I'm running without changing any code, my loss has been around 2, can you communicate if it's convenient for you
Hello, my loss rate has been around 2 and cannot be reduced, can you communicate if it is convenient for you
Have you solved this problem? I also have this problem, looking forward to your reply, if it is not solved we can communicate a bit.
I used the data set BIPED in the project in the training process, and the training set contains 200 pictures. My loss dropped to 0.9 in the 34th round, but...
 Then I trained it again using the 0.9 pth file as my pretrained model and found that he started to oscillate around 1.7.
I read that your experiment in the paper was performed 150K times. I don't know if it is because I trained too few times. I trained for 100 rounds and...
> Yes, I changed some hyperparameters. I didn't change any hyperparameters on the first run, but found that the loss did not converge, I think it was a problem that...