DMRA
DMRA copied to clipboard
About learning rate
Hi Jiwei!
Why you try a little bit more bigger learning rate in your training phase?
Or
You try some bigger lr like lr = 3e-4
or lr = 1e-6
, can you suggest some useful experiential value?
Hello,
Actually, I haven't done some experiments about learning rate. But I suggest you can set the lower lr at the beginning, because the higher lr may lead to less convergence.
However, choosing a suitable lr requires a lot of experiments, which is very time-consuming. So I think you can refer to some great baseline, such as ResNet, VGGNet.
I'm sorry I don't help you much.