This reverts commit 8ba38e07.

* update 1.8 apis, test=develop

* udpate contents, test=develop

* add NLLLoss cn doc test=document_preview

* add NLLLoss cn doc test=document_preview

* add NLLLoss cn doc test=document_preview

* add cn doc of trace API & diag_embed API (#2039)

* add cn doc of trace API

* add cn doc of diag_embed API

* add cn doc of diag_embed & trace API

* add bmm arange api_doc test=develop

* Add randperm cn doc, test=develop, test=document_preview

* add doc for kron op, test=document_preview

* add support for integer types, test=document_preview

* add cross, roll, index_select, nonzero api_cn

* test=document, add cross,index_select,nonzero,roll cn doc

* test=document, bug fix for syntax error

* add paddle.sin, sqrt, tanh, atan cn doc to fluid docs

test=develop

* Add Chinese document of API: randn

paddle.add doc for elementwise_add operator 
paddle.div doc for elementwise_div operator 

fix chinese doc of gradient_clip 

add the the mul and pow api to the fluid tensor doc 

* add cn doc for dot and logsumexp

add the chinese doc of paddle.norm and the preview doc 

As the title

add Chinese doc in ops(sum, zeros, zeros_like, ones, ones_like, mm, elemenwise_equal, elememtwise_sum, sort, argmax)

add Chinese doc in ops(sum, zeros, zeros_like, ones, ones_like, mm, elemenwise_equal, elememtwise_sum, sort, argmax)

* add eye,t,min,max,log1p API doc_cn; test=develop

* add new apis
quote> test=develop

* refine doc
test=develop

* use p
test=develop

* rm no_grad_guard
test=develop

* fix sample code diff for OP(retinanet_detection_output, retinanet_target_assign), fix  default value for OP(sigmoid_focal_loss). test=develop

* fix float value. test=develop

* add int type for gamma and alpha of sigmoid_focal_loss.test=develop

* change cn.rst

* change cn.rst

* change fluid site

* change site paddle.*

* add register_forward_hook api for Layer, test=develop

* fix sample code, test=develop

fix Chinese doc bug in argsort, add the dtype support of dtypes