This PR enhance CPU kernel and add new CUDA kernel to make grid_sample support:

- align_corners: with bool type.
- padding mode: which can be in ['zeros', 'reflect', 'border']
- Interpolation mode: which ca be in ['bilinear', 'nearest']

The old CPU and CUDNN version only support align_corners=true, padding_mode='zeros' and interpolation_mode='bilinear'.

The behavior of the new version op in default mode is compatible with the old version.

change to use bce_loss op, add numel check for bce_loss.

* add div, floor_div, remainder

* add alpha_dropout in nn.functional and nn.layer, test=develop

* refine Interface and assertion, test=develop

* fix ci import error, test=develop

* fix alias and use layers.scale, test=develop

* fix doc and training params, test=develop

* refine details in doc, test=develop

* fix doc details, test=develop

* update Conv2d Conv3d conv2d conv3d api

* update conv_transpose api

* add binary cross entropy with logit loss

* [API2.0] add dropout, dropout2d and dropout3d, test=develop

* refine Interface and assertion after review, test=develop

* fix alias p=1 and use scale, test=develop

* fix doc and training, test=develop

* fix doc in Dropout2D, test=develop

* add functional ctc_loss and CTCLoss class.

* modified docstring of ctc_loss and CTCLoss

Update the error message for the margin_ranking_loss

* add cross_entropy to nn/layer and nn/functional, test=develop

* use functional/cross_entropy in layer/CrossEntropy

* use functional/cross_entropy in layer/CrossEntropy, test=develop

* Update `paddle.clamp`

rename to `paddle.clip`
add fast path for dygraph mode
remove `out`
rename `input` -> `x`
update doc sample

* Fix leftover `Variable` wording

* Indent doc with spaces

* Remove `:alias` in docs

* Update `enable_imperative` -> `disable_static`

* Remove `imperative`

also trigger CI

* Update tests for better coverage

* Rebase to fix `cosine_similarity`

* Fix `cosine_similarity` some more

* add adaptive_avg_pool2d

* add adaptive_avg_pool3d

Fix the API of bce loss and add functional API binary_cross_entropy 

* add input.py  file

* write input.py

* fix init file

* add unit tests

* fix dygraph and input shape

* Revert "fix dygraph and input shape"

This reverts commit 89ad8664.

* fixed pylint

* fix deprecated

* fix old op

* fix old op

* set check_dygraph=True

* Revert "set check_dygraph=True"

This reverts commit a8e93e33.

* test commit

* fix doc and change test file name

* add pad func

* add pad

* test=develop, add pad op and apis

* restore pad2d

* test=develop, fix paddl declare

* fix pad interface

* test=develop, fix pad

* test=develop, add all pad api and cos_sim

* test=develop, remove padding default value

* test=develop, rename var to tensor

* test=develop, add more tests

* test=develop, rename tovar to totensor

* test=develop, fix init

* test=develop, add more test

* test=develop, add more tests

* logsigmoid and LogSigmoid, test=develop

* add elu gelu relu, test=develop

* update to_variable to to_tensor, test=develop

* address review comments, test=develop

* address review comments, test=develop

* change to_variable to to_tensor in test, test=develop

add nn.functional.normalize API for API2.0, fix L1Loss API typo, test=develop

* add nll loss API, update demo code of the comment

* upate the target to label, test=develop

* Update the code for the margin_ranking_loss, test=develop

Update the sigmoid, Sigmoid layer for the api2.0

* expose and unify the Tensor concepts to the user

* expose tensor to user

* add copy place for Tensor

* add copy place for Tensor

* add note

* add macro PADDLE_WITH_CUDA

* remove RUN_TYPE=DIST

* fix some error

* add Class KLDivLoss and function kl_div

* add paddle.nn.functional.mse_loss

* add name

* fix conflict

add the api and doc for the margin_ranking_loss and MarginRankingLoss

* add nll loss API, update demo code of the comment