GitLab

In reference to #729 added comments to clarify the naming and action of the layers performing downsampling in resnets.

* Zero-init the residual branch in resnet

* Add zero_init_residual as an option

* use predefined function

* cleanup resnet

* fix stride

* revert inplanes since it isnt constant

* lint

* AdaptiveAvgpool2d integrated

The Average Pool Layer on SqueezeNet and ResNet was hardcoded, this was changed, now any the input size is accepted.

* error fixed

* linting fixed

nit: replace 4 by expansion

* fix for PyTorch v0.4 (remove .data and add _ for the initializations in nn.init)

* fix m.**.**() style to nn.init.**(**) style

* remove .idea

* fix lines and indents

* fix lines and indents

* change to use `kaming_normal_`

* add `.data` for safety

* add nonlinearity='relu' for sure

* fix indents

    git ls-files | grep '\.py$' | xargs -n1 -P`nproc` autopep8 -i

Also add pre-trained ResNet-152 model.

  ResNet-152: Prec@1 78.312 Prec@5 94.046