test=develop, test=release/1.6, test=document_fix

* refine eng doc for hard_sigmoid op
test=develop
test=document_fix

* refine the description of hard_sigmoid
 test=develop
test=document_fix

* update API.spec
test=document_fix

* Refine the decription of parameters of HardSigmoid op
test=develop, test=document_fix

* Update API.spec for hard_sigmoid op
test=develop, test=document_fix

* fix APIs: relu, relu6, hash

test=develop
test=document_fix

* fix relu6 doc

test=develop
test=document_fix

* fix API.spec

test=develop
test=document_fix

* add description link for hash

test=develop
test=document_fix

* fix API:cos, exp, ceil, elu, brelu English doc
test=develop
test=document_fix

modify en api doc

- leaky_relu
- less_than 
- log
- logical_and
- logical_or
- logical_xor
- logical_not

* fix APIs,test=develop,test=document_fix

* fix conflict, test=develop, test=document_fix

* fix confict, test=develop, test=document_fix

* fix confict, test=develop, test=document_fix

* fix API.spec, test=develop, test=document_fix

* change fluid.layers.data to fluid.data,test=develop, test=document_fix

* fix bug on example code, test=develop, test=document_fix

* fix API.spec,  test=develop, test=document_fix

* update elementwise double grad to save gpu memory, test=develop

* update elementwise_mul/div_grad_grad to save memory, test=develop

* remove eval function in eigen statement to save memory, test=develop

* add unittest for elementwise_div_grad_grad without dout, test=develop

* add unittest for elementwise_add_grad_grad without ddx, test=develop

* add float16 cuda kernel for elementwise double grad op, test=develop

improve pow op according to reviews:
1. Delete unnecessary judgement statements in PowGradOpDescMaker;
2. Improve test of test_api;

overload GetKernelTypeForVar

add stop_gradient=True when attr(factor) is tensor Variable, change examples in API pow.
test=develop,test=document_preview

* make leaky relu inplacable, test=develop

* force add unittests to pass coverage, test=develop

* add hard_swish activation op (new op)
test=develop

* remove redundancy files

* modify document content of HardSwish OP

* add API test in test_layers.py

* add dynamic_graph for test_hard_swish

test=develop

test=develop

* feature/buffer_shared_inplace, test=develop

* refine code, test=develop

* fix elementwise_add op cpu inplace and sum inplace bug, test=develop

* add unittest and debug log, test=develop

* fix parallel_executor scope bug, polish code, test=develop

* fix sum op, activation op, single_in_place_inference bug, test=develop

* remove kLocalExecScopeName, test=develop

* fix unittest,test=develop

* fix out_var first version bug, test=develop

* follow comments,test=develop

* Refine Infershape in activation_op for double_grad.

* fix softrelu doc

test=develop

* update API doc

test=develop

* Update backward.py:
     - If there is no input grad var in all outputs of previous ops, do not append this op into graph.
     - Only apply this stragety when double backward.
* Update some double backward op.
* Update sum_op to judge whether a tensor is empty by numel or IsInitialized().

* double backward sqrt

* refine unittest. test=develop

* refine test. test=develop

* remove alpha in unittest. test=develop

* add double grad for square. test=develop

* formax code. test=develop

* fix for grad sum. test=develop

* refine shape. test=develop

* refine extract. test=develop

* Add conv2d_grad_grad_op
* Extracte the cuDNN conv algo searching code in conv_cudnn_helper.h.
    - Now use it in conv2d_grad_grad.
    - Will simply the searching code in conv2d and conv2d_grad in next PR.
* Enhance and fix bug in unit testing of gradient_checker.
* Support to fetch empty variables，return None in Python.

backward of backward: leaky_relu

Support backward of backward for Relu and add a new gradient checker by comparing theoretical and numerical Jacobian. (#16862)

* Support backward of backward and a new gradient checker
* Rename decorators.py to decorator_helper.py, since Python on Windows CI has decorators package.

1. Add ReluDoubleGradMaker when register relu_grad.
2. Add a new gradient checker by comparing theoretical and numerical Jacobian.  Check double gradients by double_grad_check.

test=develop

add rsqrt op

* temoprarily disable the code of use kCUDNN, test=develop

* add TODO comment, test=develop

test=develop

test=develop

test=develop

test=develop

* staged.

* polish code

* polish code. test=develop

* polish code. test=develop

* api change. test=develop

* fix default value. test=develop

* fix default value. test=develop

* staged.

* polish code

* polish code. test=develop

* polish code. test=develop

* api change. test=develop

* fix default value. test=develop

* fix default value. test=develop

test=develop

test=develop

* expose square
test=develop

* fix activation
test=develop

* Add square API
test=develop

* add necessary op

* code refine

* fix API.spec
test=develop

* fix unit test
test=develop

* add unit test sparse_grad_clip
test=develop

* fix API.spec
test=develop

* remove mac test for test_gradient_clip
test=develop

* remove selectedrows_mul_tensor
test=develop