* Add count_include_pad support to AvgPool

* Fix python_cpp/test_topi_pooling.py

* Change auto to explicitly type, and fix format.

* add conv2d_NCHWc compute template

* add conv_NCHWc compute decl and schedules

* allow default avx schedule

* fix lint

* remove unused schedule

* remove import nnvm.reg

* pass schedule object to compute and schedule

* support dilation in conv2d and depthwise_conv2d

* handle dilated conv in extern libs (cudnn, miopen)

* add layout_transform. add schedule support for ndim>=5 for x86

* fix lint

* generalize pool2d to arbitrary layout

* explain more the layout support for pool

* allow missing factor size for pooling

* explain what factor size is used for

* fix typo

* name idx -> axis

Signed-off-by: Janboe Ye <yeyuanbo@xiaomi.com>

* explicit import testing

* Enable init api for extension modules

* Better error message handling for contrib

* fix lint

* fix testcase

* fix test

This reverts commit 12d15704d7f5d30cff7540f1fd16be64c6baca68.

* [RUNTIME] Refactor extension type handling, now it is header only

* update c++ pool and softmax

* clean up reduce axis

* Add basic dense scheduler

* Revert to put back cpp dense registration

* Fix lint

* add winograd for mali

* fix lint

* add padding

* fix comment

* [Documentation Changes] Parameter description change for reduction api

The parameter description is updated for max, min, argmax and argmin

* Lint changes on patch-3 of reduction.py

* add 4 dim softmax

* update for NHWC layout

* remove layout param from softmax

* fix typo

* minor fix to pool

support axis=1 ndims=5 softmax.

add softmax axis

* few fix for softmax

* fix typo

* add more doc

* minor doc fix

* fix upsampling output shape

* fix lint

* cleanup softmax

* minor fix

* raise exception instead of assert, handles negative axis

* check axis after axis transformation

* conv2d schedules for Intel CPU (AVX2 & AVX512)

* fix lint

* remove override register