why RPNLogLoss=nan and RPN1Loss=nan?

Open lt573152873 opened this issue 8 years ago • 3 comments

I use one GPU; after bash /scripts/train_alternate.sh ; Just started showing the correct information； when Batch[680], showing RPNLogLoss=nan and RPN1Loss=nan; why?

Dec 12 '17 02:12 lt573152873

You can try to lowering the learning rate.

Dec 12 '17 05:12 tkuanlun350

I solved the problem .Thanks. but But every training time，Iteration for some time，My computer will restart。why?

Dec 12 '17 06:12 lt573152873

Why do you restart the computer while training? @tkuanlun350

Dec 12 '17 06:12 lt573152873