The content of activation is reverse

Bisonai / mobilenetv3-tensorflow

Unofficial implementation of MobileNetV3 architecture described in paper Searching for MobileNetV3.

Apache License 2.0

225 stars 81 forks source link

The content of activation is reverse #11

Closed namikimeiko closed 5 years ago

namikimeiko commented 5 years ago

In layers.py line 41~56:

HardSigmoid() contains h-swish, and HardSwish() contains h-sigmoid. Maybe simple mistake.

Regards.

martinkersner commented 5 years ago

Hi @namikimeiko The difference between HardSigmoid and HardSwish is just minor. HardSwish just additionally multiplies with its input, therefore you maybe got confused.

Cheers, Martin

namikimeiko commented 5 years ago

Oh, I understand. I'm very sorry, I mistook, you are true.

Thank you for your response.