* 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

* split build op marco & polish details

* revert register api del

* fix other unittest

Change-Id: I6210ce9c60bed48f3323c47b16500302b66cedf2

* add cache for VariableWrapper

* modify args names and vlog level

* format code style

* add log when set cache to variable_wrapper

* add log when set cache to variable_wrapper

* add comment to variableWrapper cache

* format code style

* add simple attr support and test

* add int, float attr support

* support other attribute

* add custom attrs test in cmake

* polish details

* fix test failed

* add backward test

* update test flags

* revert the modification of set_expected_place

* set device before op run

* add ut

* update, test=develop

* update, test=develop

* push multi node

* multi node

* MultiThread

* remove log

* solve bug in 30829

* add get_cublas_handle() api

* update format

* add unittests

* alter function name

* add group norm plugin

* fix compile problems

* move concat axis check to trt op teller

* add nbDims for scale and bias nv dims

* add group norm unit test

* fix unittest

* add trt version restriction for group norm op teller

* fix unittest

* add new custom op so

* fix use new method error

* fix test failed

* 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>

* fix some problem of Windows custom op

* fix some problem of Windows custom op

* fix some problem of Windows custom op

* 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

* fix windows for optimization of elementwise_add Op

* Unification of bfloat16 enablement process and refactor

* Remove unnecessary function

* Standardize the output name search

[BUG FIX] Fix softmax cross entropy overflow problem.

* save multi table one path

* format