* Fix bug of reduce_sum op. When input.numel() > INT32_MAX, its result
is wrong.

* support pure bfloat16

* support bf16 linear

* update PR to pass CI

* tiny fix where_grad_kernel.cu

* Support bfloat16 type for reducer and sharding.

* Fix some bug.

* Polish code.

* Polise code.

* Add bfloat16 datatype in fill_grad kernels.
Co-authored-by: Nsneaxiy <sneaxiy@126.com>

* namespace modify

* update by comment

* Revert "【Hackathon No.56&38】deformable_conv_v1 算子实现 float16 数据类型支持&前向运行加速 (#46111)"

support fp16 for deformable conv

* fix linspace error in amp

* fix log

* fix amp error

* Revert "Simplify size op impl (#45808)"

This reverts commit c252b1de.

* fix_seg

* fix detection
Co-authored-by: NChen Weihang <sunny_cwh@163.com>

* fix linspace error in amp

* fix log

* fix amp error

* fix ocr error which caused by amp

* add more check

* rename dtype ns

* polish code comments

* polish data_device_transform.cc

* move sum

* fix ci bugs

* fix ci bugs

* fix set_lod bugs

* fix infershape bugs

* fix ci bugs

* fix ci unittest bug

* fix ci bugs

* perfect code

* update code according comment

* add unittest

* fix ci bugs

This reverts commit c252b1de.

* fix return order error and duplicate results with specific inputs

* perfect softmax functor

* fix compile bugs

* fix ci bugs

* add softmax infer kernel

* simplify size op

* trans to cuda manuly

* fix copy error

Adapt tensor output_size for conv2d_transpose and depthwise_conv2d_transpose

* copy kernel file to phi

* delete some code

* migrate uniform_random, test=kunlun

* fix input error, test=kunlun

* fix gpu register error, test=kunlun

* add include file, test=kunlun

* try fix error from CI, test=kunlun

* polish other PR

* fix CI-coverage error, test=kunlun

* enhance grid_sampler cpu kernel to 5D input

* fix bug when 5D input tensor running on the cudnn kernel

* move elementwise test=kunlun

* move add/sub/mul/div kernel to elementwise_kernel, test=kunlun

* fix ci bugs,test=kunlun

* fix ci bugs

* test=kunlun

* migrate truncated_gaussian_random kernel to phi, test=kunlun

* reuse CPU kernel, test=kunlun

* debug kernel, test=kunlun

* migrate truncated_gaussian_random kernel to phi, test=kunlun

* split truncated_normal, test=kunlun

* try fix error from CI, test=kunlun

* [OpAttr]output_size of unpool support Tensor type

* fix coverage

* fix contain_var

* fix coverage

* fix split bug

* solve function redefine

* fix fluid.layers.split and add unit test

* delete splitInferMeta register in unary.cc

* modify test_split_op GPU unit test

* modify test_split_op GPU unit test place param

* refactor split op and fix infershape bugs

* add () in && and ||

* fix split C++ unit test

* fix split infershape

* Adapt tensor axis for argmin/max

* Add UT

* Polish UT

* [OpAttr]Adapt tensor axis for reduce_min/max/mean/sum/prod

* move incubate to geometric

* add paddle.geometric

* fix unittest bug

* add float16 support for segment op

* change reindex and sample neighbors flag name

* add heter graph reindex

* move sample_neighbors.py to neighbors.py

* delete khop_sampler in geometric

* delete unused code

* change sample_neighbors api input order

* fix en doc

* fix unittest

* fix unittest

* change reindex

* fix division by 0

* delete unnecessary input argument

* delete final_state

* Add omp parallel for directives

* Revert "Add omp parallel for directives"

This reverts commit f4e4f8ddb12454018d9c1e49c074af2543659de6.

* Add #pragma omp parallel for to correct file

* Add check for _OPENMP definition

* Disable omp on gpu

* Trigger CI

* Readd check for _OPENMP definition

* Change macro disabling changes on GPU

* Improve macro readability

* [OpAttr]min/max of Uniform_rand support Tensor type

* fix typo

* make full_like support double_max in dygraph

* fix bug

* make tensor_util contains no cuda code

* refine isfinite

* revert ut

* move isfinite function to its op

* fix test

* fix compile

* std::isnan is not defined for int type on windows

* fix windows compile

* fix fp16

* fix rocm compile

* revert gradient node