* optimize cache-key by replace GraphToProgram to Dot string

* fix compile failure bug

* fix build error of gtest dependency, test=develop

* update when WITH_TESTING is OFF, test=develop

* modify logic

* test=allcase

* test=document_fix

* add trt runtime version check

* use different wrap, and change to major version check

* Add dlnne engine runtime

* Fix log

* Remove <const_cast> and remove unrelated modify with dlnne, +clang-format

* Fix CMakeList format error

* Add copyright message

* Fix dlnne CMakeList.txt

* Add some paddlepaddle_pass to support more networks

* Fix some format bug

* delete include framework.pb.h

* fix error

* fix tensorrt output varible reshape

* move padding shape x 1 x 1 in ernie to qkv and fc

* update layer name

* fix softmax when input is dynamic, fc not padding any more

* fix varlen

* move fc x_dim assert to op_teller

* fix engine key in trt int8 calibration

* fix unit test

* support trt serialize when load model from memory

* delete conv_bn_fuse_pass before tensorrt, with which trt serialize engine id is not stable

* Revert "delete conv_bn_fuse_pass before tensorrt, with which trt serialize engine id is not stable"

performance degradation, fix in the future

This reverts commit fa6cd17e60b15df351efda379ddd00e9e9c1fea9.

* add delete conv_bn

* delete path when delete_cache_files

* update paddle_fluid.so to paddle_inference.so

* Add conv transpose BF16

* Share function GetWeightsTz

* Adjust to review and fix op compatibility

* Add bias to unique handler name

* Remove errors related to paddle enforce

* Add conv2d_transpose to bf16 list and kernel refator

* use iwyu clean include second time, test=develop

* add dla

* add dla done

* add python api
Co-authored-by: Nshangzhizhou <root@szth-rp-fanyi-opera49.szth.baidu.com>

* snap

* add inference api: DisableTensorRtOPs

* fix code style

* update api to experimental

* update variable name

* register instance norm, test=op_version

* add the support the op version check for matmul, test=op_version

* add the op version check for the elementwise ops, test=op_version

* add the support check for elementwise_ops, test=op_version

* fix tensorrt output shape error

* fix unittest tensorrt_engine_op_test

* fix code style for unitest

* register op_version for addto

* upgrade pass capability

* change eq to le

* change eq to le

* fix merge

* skip_layernorm_op done

* add unittest

* slice op convertor support trt < 6

* skip_layernorm only work in ernie

* Add bfloat16 softmax and gelu

* Add pass attr bfloat16_enabled_op_types

* Changes from review

* fp16 result ok

* change -DWITH_NVINFER_PLUGIN toconfig.EnableTensorRtOSS

* auto detect special slice op converter for ernie with trt oss

* ernie oss only support fp16

* fix special_slice_plugin serialize bug

* matmul in tensorrt ok

* ernie unittest ok

* add matmul tensorrt unittest

* remove demo code