Does it exist ‘lr_mult' and 'decay_mult' in optimizer?

1

I saw a code like this,

{'params': bn, 'lr_mult': 1, 'decay_mult': 0,
             'name': "BN scale/shift"}

so, dose ‘lr_mult’ and ‘decay_mult’ avaliable in pytorch optimizer ?

2

It depends on what optimizer you are using, some have it and some don’t. See details in the doc http://pytorch.org/docs/master/optim.html.

3

The optimizer is SGD. Both the documents about SGD and the method step of SGD never mentioned about lr_mult and lr_decay.

4

Sorry, you are right. Those args are indeed not used by pytorch. Maybe the author forgot to convert them from caffe :slight_smile:

5

Hi there, does it actually used in pytorch? I meet the same question.

6

As @SimonW mentioned above, those args are not used in PyTorch.