* add dist_attr for dist op

* add unitest

* update inputname

* update function name

* add unitest

* update CMakeLists.txt for CI

* fix dis_matmul

* fix compile error

* update matmul to matmul_v2

Put Nvidia's cusparse library into paddle.

* add beam_search npu op

* fix CMakeList and add unittest

* fix bug of beam search npu op

* fix unittest

* let input ids become int64

* set output ids to int64_t

* delete check_dygraph

* fix beam_width=1

* support numpy.ndarray index.

* polish code.

* [NPU] fix depthwise_conv2d_grad, test=develop

* remove debug files, test=develop

Co-authored-by: Nxiaoxiaohehe001 <hiteezsf@163.com>

* 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

Add paddle.cuda.device.stream_guard API 

* add_split_op

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* Enhance Check mechanism of fetch_list in Executor

* support single tuple

* fix typo

* fix typo

* add layerwise learning rate for adamw

* fix format

* add unitest

* add NotImplementedError

* add gpu unitest

* update gpuplace

* add anchor_generator test

* add anchor_generator test

* add clip convert test

* add swish convert test

* modify
Co-authored-by: Nxiaoxiaohehe001 <hiteezsf@163.com>

* new function: share third party cache among servers to fasten build speed

* modified code according to zhouwei25's comment

* add wget install step, move cd build to the last of if condition

* block note and error of third_party share; change bce upload method

* change third_party sub_dir in bos, since third party in different cuda version cant share

* set sub_dir by get nvcc version

* change third_party local path to be same with bos path

* add anchor_generator test

* add anchor_generator test

* add clip convert test

* delete wrong file
Co-authored-by: Nxiaoxiaohehe001 <hiteezsf@163.com>

Implement FunctionTraits to support two kinds of elementwise functor and remove some old codes for broadcast. (#35688)

delete blank lines at the beginning of the file

delete main function

1. optimize the error message of softmax_with_cross_entropy_op;2. add input value check for cross_entropy, if the dimention of input is zero, raise error. test = develop

* add dropout convert test

* modify dropout convert test
Co-authored-by: Nxiaoxiaohehe001 <hiteezsf@163.com>

* add affine_channel test

* modify affine_channel convert test
Co-authored-by: Nxiaoxiaohehe001 <hiteezsf@163.com>