About group number error

[ChenHongYen]

Thanks for your job！
But I found that if (output channels / groups) != 1 and equals to other integer, the net couldn't work.
e.g. If input channels = 32, groups = 32 and output channels = 64, the loss of net will not decrease. In Imagenet1000 training, the loss is still at 6.9.
Do you know how to solve this problem?
Thx!

[刘灏@megvii.com]

The operation which changes the output channel nums as you descript is not the classic depthwise convolution in mobilenet paper.
Sorry, my implemence does not suppport it.

[ChenHongYen]

THx!
I do this because in the official caffe conv_layer.hpp says that output channels / groups can be bigger than 1.
like this
group (\b optional, default 1). The number of filter groups. Group

• convolution is a method for reducing parameterization by selectively
• connecting input and output channels. The input and output channel dimensions must be divisible
• by the number of groups. For group @f$ \geq 1 @f$, the
• convolutional filters' input and output channels are separated s.t. each
• group takes 1 / group of the input channels and makes 1 / group of the
• output channels. Concretely 4 input channels, 8 output channels, and
• 2 groups separate input channels 1-2 and output channels 1-4 into the
• first group and input channels 3-4 and output channels 5-8 into the second
• group.
  So I guess your depthwise convolution layer can do the same thing like official caffe conv layer.

[刘灏@megvii.com]

Sorry, it is simplethat work like pooling.

[刘灏@megvii.com]

Sorry, It is not functional as group is, it is just a simple implement which operate on same output and input.

[ChenHongYen]

Thx you for answering my question.
I will try to implement it with group funtion.