How to set the training parameters on mnasnet

rwightman / gen-efficientnet-pytorch

Pretrained EfficientNet, EfficientNet-Lite, MixNet, MobileNetV3 / V2, MNASNet A1 and B1, FBNet, Single-Path NAS

Apache License 2.0

1.57k stars 214 forks source link

How to set the training parameters on mnasnet #6

Closed stigma0617 closed 5 years ago

stigma0617 commented 5 years ago

Hi,

I have tried to reproduce mnasnet.

Could you let me know how to set the training parameters?

or Can I get the same results with your reported ones by using your default setting in your training code?

Tx in advance.

rwightman commented 5 years ago

Please refer to https://github.com/rwightman/pytorch-image-models/issues/11

You could try something similar to my suggested Mnasnet/MobileNet-V3 settings at the top of that thread... or perhaps something similar to the reported EfficientNet-B0 settings near the bottom .... you'll want to scale the learning rate/batch size to match your gpu type and number of them and you'll likely need to do some experimenting