* Add New Op: gumbel_softmax

* Add New Op: gumbel_softmax

* Add New Op: gumbel_softmax (amend)

* add __main__ function in unit test

* fix bugs when test in windows ci

* update en docs

* delete reletive error in unit test

* delete relative error in unit test

* set hard=True in unit test

* Support fix seed in Python for test

* add gradient kernel of det op and slogdet op

* fix CI APPROVAL problem

* Add elementwise_sub_mkldnn_op without grad

* Add test to static_mode_white_list

* Refactor code, change license years

* Remove invalid grad implementation

* Fix element_wise_sub_op test

* Fix CI Approval error

* Remove unnecessary EltwiseSubMKLDNNGradKernel class

* Fix CI Approval 2

* Fix CI Approval 3

* Fix CI Approval Attempt #4

* Fix CI Approve Attempt #5

* Fix CI Approval Attempt #6

* Fix CI Approval Attemt #7

* Change test names containing add to sub

* Fix old tests testing add instead of sub

* Copy grad implementation from elementwise_add_mkldnn

* CI test fix attempt

* Revert "CI test fix attempt"

This reverts commit c647cacf41e6a87c715385a185de5cbf65fc8900.

* Fix CI attempt 2

* Fix elementwise_sub tests, temporary mkldnn broadcast test disable

* Add working implementation of elementwise_sub grad

* Fix build errors caused by pull

* Fix format error

* Fix format error 2

* Disable elementwise_sub_mkldnn test on GPU

* Apply fix for paddle.fluid import

* Revert changes of test_elementwise_sub and Fix mkldnn test

* Revert "Apply fix for paddle.fluid import"

This reverts commit fc3b122fec8e12f2bcb32928a2685ba4d20fd742.

* fix bug of module 'paddle' has no attribute 'fluid' for python3.6 (#35862)

* Add changes suggested by reviewers

* Change @unittest.skipIf... to @OpTestTool.skip_if_not_cpu_bf16() to satisfy Approval CI

* Remove check_dygraph=False to satisify CI Approval
Co-authored-by: Nzhangbo9674 <82555433+zhangbo9674@users.noreply.github.com>

* add pool2d convert test

* modify error

* modify error

* modify error

* modify error

* modify error

* modify error

* - candidate fix

* - More fixes to #34554

* - another incosnstent fix to key

* - Remvoed unneeded line

* - matching the cache behaviour to other ops

* Add linalg.solve op, test=develop

* Fix a bug caused by accidental deletion

* updated description and fix a bug: missing a comma

* Add linalg.solve op, test=develop

* updated solve op backward logic

* updated solve op backward logic again

* Add linalg.solve Op, test=develop

* Updated and modified to fit CI requirements

* Fix a bug

* 1)Add more test cases; 2)Fix a wrong usage in reduces operation; 3)Remove redundant code

* Remove redundant comments

* 1)Removed redundant code; 2)Updated to enhance code robustness

* Removed redundant code

* Updated API documents

* add multihead_matmul trt converter test case

* move attribute check to op_teller

* add the shape check for the matmul

* remove the test case for the linear

* add align for WorkQueue

* WorkQueue update

* Revert "WorkQueue update"

This reverts commit 14ce793dbb204f8ddec63c34b3b72a73c7cdb93a.

* optimize WorkQueue

* add argmax and iou_similarity for kunlun

* add argmax and iou_similarity for kunlun

* add argmax and iou_similarity for kunlun

* Fix copy elision warning

* Remove redundand code

* add randperm_op_npu

* fix test_set_value_op_npu

* Pass compat of conv_transpose_bias_mkldnn_fuse_pass

* Fix a bug of strided_slice op, about the axes parameter access memory out of bounds

* Fix a bug of transpose op, about accessing memory out of bounds of the perm param

* op:transpose_op supports bool type

Add new API : paddle.linalg.det & paddle.linalg.slogdet

API Alias：paddle.det& paddle.slogdet

* support ernie-int8 test and prune op attribute test

* remove using and use namespace

* remove macro and use shell instead

* Revert "remove macro and use shell instead"

This reverts commit 615964b149d7de7825b341936b42be22a4bc0091.

* fix grammar error

* fix shell error

* support nnadapter and ascend310

* modify code

* add anchor_generator convert test

* add gelu convert test

* add conv2d convert test

* modify anchor_operator convert test

* modify conv2d test

* modify con2d convert test

* modify conv2d convert test

* modify conv2d convert test

* modify conv2d test

* fix WITH_PYTHON compile error

* modify test file

* modify test file

* modify test file

* modify test file

* modify test file

* modify test file

* modify test file

* modify test file
Co-authored-by: Nxiaoxiaohehe001 <hiteezsf@163.com>
Co-authored-by: Njiweibo <jiweibo@baidu.com>

* refine gc for new_executor, test=develop

* refine, test=develop

* refine, test=develop

* merge, test=develop

* Modify H2D and D2H as kQueue::Sync

* fix interface error

* support extern third_party lapack on Linux/Windows/Mac

* fix ci

* Create stateful OneDNNAXPYHandler object.

This makes it possible to call it multiple times without recreating the
oneDNN primitives every time.

* Prepare SGDOpKernel to reuse its implementation from OneDNN kernel.

* OneDNN SGD kernel.

* Update call to use new OneDNNAXPYHandler object api.

* Setup seed in proper place.

* Enable OneDNN kernel only for single case.

* For dense param and sparse grad.

* Small refactor.

* Enable oneDNN by op attr or by cmd line flag.

* Use int64_t type for number of elements.

* Support dense param and grad from OneDNN kernel.

* Enable SGD OneDNN kernel when use MP BF16 optimizer.

* Force non-copyable/movable OneDNNAXPYHandler.

* Reuse OneDNNAXPYHandler for spare tensors in SUM op.

* Fix SFINAE rules.

* Remove recording event inside AXPY.

* Get rid of internal primitive caching.

* Stop use PP cache mechanims to store mem and primitive obj.
* Handler obj store and reuse needed desc & prim

* Do not derive from MKLDNNHandlerT

* Optimization of pool2d grad, first commit.

* remove useless print codes

* refine codes

* refine codes

* seal more operation into template specialization

* fix template struct error in MaxPool2dGrad.

* Fix header including error

* refine code with comment

* Seal the param-preparation codes into function for common use.

* Seal the param-preparation codes into function for common use.

* Seal the param-preparation into funciton and make it common for other kernels

* polish code and erase useless template speicalization

* Rerun triger

* rerun trigger

Add basic Cost Model, it uses executor to run program and profile it to get op time.

This is an early basic version, we will add more functions in the future.