wuyuanhang94

shouldn't group in_channels inside 1x1 conv

2

`nn.Conv2d(in_channels, C * D, kernel_size=1, groups=C, bias=False),` doesn't make sense.

建议site加上苏州

1

苏州不是有Momenta和微软之类的大厂吗？感觉可以新开一个site

it should be 4, taking RexNext-50 or 101 into account.

There should be Aux classifier in GoogLenet, but there is not now.