* add leaky_relu forward and backward in activation_op.cu

* support control flow

* supoort sync_parameters_buffers

* fix the bug of sparse embedding

* add custom init grad for backward function

* add custom init grad for backward function

* handle when the grad_tensor is none

* handle when the grad_tensor is none

* fix the args type error on windows platform

* modify the args order and doc

* format code

* add grad_tensor to xpu

* modify the grad_tensor type check

* add paddle.backward api to support multi tensors gradient compute

* add paddle.backward api to support multi tensors gradient compute

* add paddle.atuograd module and backward api

* change tensor.backward func args

* modify tensor backward api

* remove create_graph intputs args

* add doc and examplex code for backward api

* when have the same tensor, throw error

* modify test Init func args

* modify the execute.Init func args in test files

* add paddle.autograd package in setup.py.in

* modify error msg, remove _run_backward method in class Tensor

* add test cases for backward api

* upgrade vlog

* train from dataset fetch optimize

* add anchor generator op plugin

* add anchor generator unit_test

* remove dbg info

* remove redundant line

* replace assertion with paddle enforce

* dynamic plugin replaces assertion with paddle enforce

* anchor generator support dynamic shape on spatial axis

* anchor generator test with fp16, dynamic shape

* add anchor generator test all

* add back main

* reduce test input size to not exceed the timelimit of ci

* change super to InferencePassTest for python2 compatibility

* reuse paddle operator anchor generator

* move creator construct to header with default

* add cuda ifdef

* reduce line

* change super to InferencePassTest for python2 compatibility

* fix anchor generator fp16 serialize setting

* split unittest from test_all

* restrict anchor generator input format before version 7234

* anchor generator only support greater than trt7.1

* change min_graph_size to 2

* min_graph size to 3 if dynamic shape

* reduce dynamic shape size to avoid trt search tactic too long to exceed time limit

* remove anchor from fetch list

* anchor generator support all trt version

* fix memory not allocated but if serialized

* new group

* ci compatible fix

* assert nccl

* refactor and simplify hook design

* fix reducer add hook error

* add Tensor.register_hook basic impl

* refine prepare data impl

* revert prepare data change

* support register_hook for Tensor

* add hook test in model

* polish tests and doc example

* fix double grad test failed

* remove reduce hook func

* fix set empty error

* polish code by comments

* change reduce_hook to mutable_hook

* remove useless tmp_ins

* fix shape code format error

* fix shape code format error

* delete include framework.pb.h

* fix error

* delete fluid_train

* [Parallel UT]improve Parallel UT level on Windows/Linux

* [Parallel UT]improve Parallel UT level on Windows/Linux

* [Parallel UT]Improve Parallel UT level on Windows/Linux

* [Parallel UT]Improve Parallel UT level on Windows/Linux

* fix CI

* update, test=develop

* fix one error massage

* fix a error message

* new fix three error messages

* new fix three error messages

* new fix some error

* new fix one error message

* polish tensor pipeline. test=develop

* update eigen version to f612df27, test=develop

* fix compilation error, test=develop

* remove patch command in eigen, test=develop

* fix compilation error caused by call Eigen function with float16 and bfloat16, test=develop

* fix unittest error, test=develop

* fix unittest error caused by precision, test=develop

* remove patch files used by old version eigen, test=develop

* update compilation with C++14, test=develop

* fix compilation error in eigen, test=develop

* fix split core

* format

* bugfix for warpctc

* fix warpctc commit id

* fix warpctc commit id

* fix warpctc commit id

* fix warpctc commit id

* fix warpctc commit id

* fix WARPCTC_WITH_HIP invalid

* Add logs to find out why can not dlopen libwarpctc.so

* fix warpctc commit id

* fix unit test test_warpctc_op

* Optime failed log for dlopen

* Optime failed log for dlopen

* Delete extra changes

* fix warpctc commit id

* fix warpctc commit id

* Add is_compiled_with_rocm for test_warpctc_op

* fix warpctc commit id

* Cancel optimize dlopen failed reason, move to next pr, due to it makes windows ci failed

* Cancel optimize dlopen failed reason, move to next pr, due to it makes windows ci failed

* Cancel optimize dlopen failed reason, move to next pr, due to it makes windows ci failed

* fix code style problems

* 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

* Remove old custom OP to reduce whl package volume

* [Custom OP]Remove old custom OP to reduce whl package volume

* fix batchnorm when inpu dims < 3

* add unittest for batchnorm dims = 2

* yolobox converter and plugin

* yolobox unittest

* add dynamic shape restriction

* fix git merge log

* fix cmake model path

* update cmake

* fix unittest

* fix unittest

* give shape related contructor and reshape warning

* change line num to fit ut

* change ut to fit

* remove useless code

* call resize directly in constructor

* add roi_align_plugin

* add roi align unit_test

* add roi align serialization

* remove roi align static plugin because of batch dim issue

* refine roi align unittest and add fp16/serialization

* add trt roi align condition to op_teller

* refine error message

* remove unnecessary reshape layer

* trt affine channel converter

* add trt affine channel base test

* add trt affine channel NHWC

* remove asterisk for python2 compatibility

* trt affine channel converter

* add trt affine channel base test

* add trt affine channel NHWC

* remove asterisk for python2 compatibility

* fix rebase

* move LodTensor to Tensor

* add dbg info

* affine channel converter only support NCHW

* scale,bias are parameters, use create_parameters api

* reduce test input size to not exceed the timelimit of ci

* refine affine channel unittest and add serialization/dynamic test

* change super to InferencePassTest for python2 compatibility

* change super to InferencePassTest for python2 compatibility

* fix affine channel fp16 serialize setting

* Use layer to calculate output scale
* add backward for moving_average_abs_max_scale and save output scales to op's attr