* cherry-pick:add softmax_switch for softmax_with_cross_entropy_op, test=develop

* add softmax_switch for softmax_with_cross_entropy_op, test=develop

* delete using EigenMatrix in softmax_with_cross_entropy_op.h, test=develop

* add REGISTER_OP_VERSION for softmax_switch attr of softmax_with_cross_entropy_op, test=develop

* cherry-pick:add softmax_switch for softmax_with_cross_entropy_op,test=develop

* change softmax_switch to use_softmax, test=develop

* fix code format for softmax_with_cross_entropy_op.cc, test=develop

 if all input grads are zero, the output of clip_by_norm will be inf or nan. This pr is used to fix this bug.

* Remove PE special profiler (#30886)

* remove pe special profiler

* add profiler info

* add truncated gaussian random (#30922)

add truncated gaussian random

* 【Paddle.Fleet】fix dataset zip py3 bug (#31441)

* fix zip py3 bug

* 【Paddle.Fleet】Fix one ps gradient clip  (#31664)

* fix one ps gradient clip

* fix en doc for emb (#31980)

* fix en doc for emb, test=document_fix;
Change-Id: I4757e67caacd7189f068493ed45a7445f87ffb40

* LOG CLEAN (#31819)

* upgrade vlog

* train from dataset fetch optimize

* OneDNN hardswish integration (#30211)

* keep only conv + hardswish in this PR
Co-authored-by: Njakpiase <62569058+jakpiase@users.noreply.github.com>

* [Paddle-TRT] TRT inference support for BERT/Transformer in paddle 2.0 api (#31744)

* support multihead_matmul_fuse_pass_v3

* fix compile problems

* embedding_eltwise_ln pass support lookup_table_v2

* suppoort matmul and matmul_v2 in qkv matmul

* map_matmul_to_mul_pass support 3dim

* fix cache key in concat oneDNN kernel

* key simplified

* update, test=develop (#30692)

* align the default value of some configuration for fleet to that of single cards (#30740)

* update, test=develop

This reverts commit 20e68a22.

* solve build gpu task core (#30626)

* build gpu task core

* format

* dump to cpu (#30750)

* dump to cpu

* format

* format

* format

* support multi node in heterps (#31102)

* push multi node

* multi node

* MultiThread

* remove log

* solve bug in 30829

* optimizer

[Cherry-pick] Fix dtype unmatched in custom op API

cherry-pick of #31305

* modify custom op dependent from paddle_framework to paddle_custom_op (#31195)

* [Custom Op] Remove unsupport dtypes (#31232)

* remove remove_unsupport_dtype

* remove remove_unsupport_dtype

* remove test dtype

* add more include

* change dtype.h's enum as enum class to avoid conflict with inference lib

* make enum as enum class

* remove additional test

* merge develop

* polish code

* [Custom OP] Support stream set on Custom Op (#31257)

* [Custom OP] change the user header file format, test=develop (#31274)

* [Custom OP]add PD_THROW and PD_CHECK for User Error message (#31253)

* [Custom OP]add PD_THROW and PD_CHECK for User error message

* PD_THROW and PD_CHECK, fix comment

* fix Windows error message

* fix Windows error message

* fix CI

* [Custom OP]add MSVC compile check on Windows (#31265)

* fix test_check_abi
Co-authored-by: NZhou Wei <52485244+zhouwei25@users.noreply.github.com>
Co-authored-by: NJiabin Yang <marsyang199376@gmail.com>
Co-authored-by: N石晓伟 <39303645+Shixiaowei02@users.noreply.github.com>
Co-authored-by: Nzhouwei25 <zhouwei25@baidu.com>

* [CustomOp] Add Modeling with Custom op unittest (#31218)

* add unittest for static/dygraph/dy2stat

* add PE unittet

* remove usless code

* add unittest in CMakeList.txt

* [CustomOp] Split build op marco & polish details (#31229)

* split build op marco & polish details

* revert register api del

* fix other unittest

* [CustomOP]Support Incremental compilation and Add Version management (#31228)

* Support Incremental compilation and Add Version management

* replace hash with hashlib

* fix test_op_num unittest

* Revert "fix test_op_num unittest"

This reverts commit 2f78de976e1d7ca60915b2310717b38a32ae204a.
Co-authored-by: NChen Weihang <chenweihang@baidu.com>

* fix error message & label check in softmax_with_cross_entropy

* fix error message & label check in softmax_with_cross_entropy

* fix print comment

* fix ignore_index check in softmax_with_cross_entropy

ATT，cherry pick PR #29260

[Cherry-pick] The Second part of new custom op extension in 2.0.1

Change-Id: I6210ce9c60bed48f3323c47b16500302b66cedf2

cherry-pick #31068

* add get_cublas_handle() api

* update format

* add unittests

* alter function name

Cherry-pick double grad for clip

* fix entry

* fix distributed lookup table fuse case

* fix entry bug at first time

* move entry from paddle.fluid -> paddle.distributed

* fix ut with paddle.enable_static()
Co-authored-by: Nmalin10 <malin10@baidu.com>
Co-authored-by: Nmalin10 <malin10@baidu.com>

* added support for fake_quantize_dequantize_abs_max op in quantization inference pass

* remove const_cast to pass ci

* remove compare operator to pass ci-coverage

* added detailed error message for unregistered tensorrt_subgrah_pass

[CustomOp] New custom operator extension mechanism in 2.0.1

Cherry-pick New custom operator basic implementation related PRs

[BUG FIX] Fix softmax cross entropy overflow problem.

* [Kunlun] Add condition_variable and notify() in BindThreadedSSAGraphExecutor (#30586)

* [Kunlun] fix dead lock for exec_op_count_ (#30718)

* Fix the problem that the number of ops executed by xpu is wrong (#30961)
Co-authored-by: Nliuyuhui <liuyuhui@baidu.com>

ATT, cherry pick of #31132

* A fix for oneDNN matmul kernel. Fixes issue #30309 (#30723)

* A fix for #30309 with oneDNN 1.6

* test=develop, save/load, shrink
Co-authored-by: NseiriosPlus <tangwei12@baidu.com>
Co-authored-by: N123malin <malin10@baidu.com>