* support matmulv2 in multihead

* Polish codes and memory usage for fused_gate_attention.

* Fix wrong reduce_dims in fused_gate_attention when computing gradient of nonbatched_bias.

* optimize dump_to_cpu for dymf

* code clean. test=develop

* fix func. test=develop

* fix code style. test=develop

* fix. test=develop

* add method for post ops

* format code

* change post-ops pattern

* code style

* add use_master_acc_grad

* add ut

* Update Profiler

* make HostEventRecorder templated

* fix bugs of reducer

* update

* update

* [MLU]support cast double type

* [MLU]fix cast test

* Update creation.py

* Update search.py

* Update search.py

* Update xavier.py

* Update xavier.py

* Update pooling.py

* Update pooling.py

* Update pooling.py

* Update search.py

Transpose optimization with assitant of  Chengdu Supercomputing Center and auto_tune operation (#42704)

fix the unittest bug of none grad of margin_cross_entropy when FLAGS_retain_grad_for_all_tensor change default setting (#43241)

* modify paddle2onnx cmake

* modify paddle2onnx cmake

* modify export interface

* modify paddle2onnx export

* paddle2onnx add mac windows

* modify paddle2onnx mac windows cmake

* modify paddle2onnx mac windows cmake

* modify paddle2onnx cmake support windows

* modify paddle2onnx cmake support windows

* modify paddle2onnx cmake support windows
Co-authored-by: Nxiegegege <xiege01@baidu.com>

* fix gradient merge

* bug fix

* update annotation

* test=document_fix

* cpplint set version 1.6.0;test=document_fix

* use yapf to format all python file

* yapf exclude two unittests file for they rely on writing and reading file, and format will break them

* disable diff_py_file because too many diff files cause command following failed