* add data-parallel model doc

* Polish DyGraph doc

* skip PipeReader in IO

* skip PipeReader in IO
test=develop

test=document_preview fix the defination of nms_eta in api_cn of op retinanet_detection_output (#1289)

* add data loader doc, test=develop

* follow comments, test=develop

* follow comments again, test=develop

* test=document_preview
Fix the op API of diag, greater_equal, greater_than, less_equal, linspace, not_equal, rank, reduce_all, reduce_any, rsqrt, sign, unique_with_counts, where, zeros_like.

* test=document_preview
Fix some problems in fisrt commit, main format problem.

* test=document_preview
Fix document, the return type of greater_equal, greater_than, less_equal, not_equal.

* test=document_preview
Fix the format in doc of op rsqrt.

update cn doc of "logsigmoid, softsign, softplus, softshrink, tanh, tanh_shrink, NormalInitializer, TruncatedNormalInitializer, UniformInitializer" (#1265)

* test=develop update cn fluid doc

* test=develop update cn doc

* refine deformable roi pooling doc, test=document_preview

* test=document_preview refine api_cn for retinanet_target_assign_cn, sigmoid_focal_loss_cn, retinanet_detection_output

* test=document_preview refine the api_cn of op retinanet_target_assign, retinanet_detection_output, sigmoid_focal_loss

* test=document_preview refine api_cn of op retinanet_target_assign_cn

* improve fc api cn doc test=document_preview

* refine param num_flatten_dims doc test=document_preview

* refine output dataType test=document_preview

* refine return type

* add LoDTensor

* refine paraAttr, bias_attr

更新FluidDoc/external中的book repo 至9月5日。

* Update chinese doc of layers.sums.
test=develop
test=document_preview

* Add more examples and more detail imformation of output.
test=develop
test=document_preview

* Refine the format.
test=develop
test=document_preview

* Refine the comment in example codes.
test=develop
test=document_preview

* Refine the introduce of out.
test=develop
test=document_preview

* Refine the introduction of out.
test=develop
test=document_preview

Since Program_cn.rst is a Binary file we delete it first then we will add a regular rst file back

* rewrite the content of load_inference_model_cn.rst.

* add AdadeltaOptimizer cn doc, test=develop

* refine adadelta optimizer doc again, test=develop

* fix APIS,stanh,accuracy,auc,smooth_l1,linear_chain_crf,sigmoid,softmax, test=document_preview

* fix notification, test=document_preview

* fix docs, test=develop

* fix LoDTensor ref bug, test=develop

* follow lanxianghit's comments, test=develop

* add range of output random int tensor, test=develop

* remove unused repo to increase CI

* remove three repos

* add Name and ParamAttr index

* add default value

* deprecate paddle.fluid.memory_optimize

* fix comment

* fix comment

modified cn doc for image_resize and resize_nearest/bilinear/trilinear, test=develop, test=document_preview (#1161)

* refine flags doc,test=develop

* follow comments, test=develop

add cn doc for crop_tensor
modified cn doc for crop

* Update WIndows whl link

* Update WIndows whl link for 1.5.2, test=develop

* Update WIndows whl link for 1.5.2, test=develop

* update slice_cn.rst test=document_preview

*  update reshape_cn.rst test=document_preview

*  update pow_cn.rst test=document_preview

* update expand_cn.rst test=document_preview

* add result of output. test=document_preview

* modify deformable conv chinese doc, test=document_preview

* MKL-DNN Data transform doc design

- Cosmetic

- Changes

- CPU -> Paddle conversion

- Finished data transform draft

- Cosmetic fixes

* - Fixes after internal review

* Exemption mkldnn

* update

* test=develop

* update install doc for release 1.5.2, test=develop

* update Tables.md for 1.5.2, test=develop

* update Tables.md to 1.5.1 for windows, test=develop

With the release of 1.5.2, CUDA and CUDnn support will also need to be updated for Windows installation.

Now，release of 1.5.2 support CUDA10, and CUDA8 should match up cuDNN v7.1+.

* fix the installation document of Windows, test=document_preview

*  CUDA and CUDnn support will need to be updated for Windows installation, test=develop