Single-DGOD
Single-DGOD copied to clipboard
Loss is
Loss is nan
The learning rate is set to 0.001. And the batchsize is set to 4. Welcome to communicate with me. Thanks.
The learning rate is set to 0.001. And the batchsize is set to 4. Welcome to communicate with me. Thanks.
My learning rate is set to 0.001 too.The difference is that I used two gpus with a batch size of 8.The only way to get it to work properly is to reduce the learning rate to 1e-5, but then it won't converge, why is that?Sincerely look forward to your guidance, although this may be a naive question for you.
Thank you for your reply. The problem has been successfully resolved.