search

tonylins / pytorch-mobilenet-v2

A PyTorch implementation of MobileNet V2 architecture and pretrained model.
Apache License 2.0
1.37k stars 328 forks source link

Training Details #33

Closed YUE-FAN closed 5 years ago

YUE-FAN commented 5 years ago

Hi,

Thanks for sharing this awesome work. Could you please also provide the batch size, the hyper-parameters for the optimizer and the decay steps of the cosine learning rate?

tonylins commented 5 years ago

Hi,

The batch size is 256. I used the default hyper-parameter of SGD in PyTorch (momentum=0.9). As for the learning rate, I just used the half cosine decayed learning rate, which does not have decay steps.