* fix QuantizeLinear kernel and pass in QAT (#44784)

* Add Reduce Max in Quant (#44825)
Co-authored-by: NChang Xu <molixu7@gmail.com>

* slice_to_2.3

* add cast trt convert

* skip cast trt convert when input dtype is bool

* code format

* fix bug

* update unittest

* fix bug

* Adjust the relative error of QR's grad (#42221)

* Fix the format

* [Inference]ort backend optimizer (#44136)

* add ort clone interface

* paddle2onnx update to 1.0.0rc

* ort input_tensor use mutable data of scope

* clone ort_predictor reuse session (#44703)

* ort backend support output mutable data (#44724)

* 2.3 interface is different from the Develop interface

* 2.3 interface is different from the Develop interface

* 2.3 interface is different from the Develop interface

Co-authored-by: Ngongweibao <gongweibao@baidu.com>

* fix record event for operator type in new dygraph (#44582)

* fix new dygraph record event for op

* update unit test

* fix file mode

修复poisson op单测随机挂

原因：由于随机OP的无法直接验证数值正确性，该单测随机采样100万个样本，统计落到直方图各区间的数量，计算出粗略的概率密度函数，与标准概率密度函数对比，这种测试方式会有一定误差。
当采样数量越小，误差越大，因此该PR增大采样样本数量（100万->200万），误差进一步减小在rtol范围内。

cherry-pick #43626

* add serialization for new field in event node (#43405)

* add serialization for new field in event node

* fix a bug

* add more field to memory record (#43578)

* Add infer shape in dygraph (#43822)

* record memory and op supplement info

* update

* update

* fix a bug

* fix memory recording

* fix a bug

* update

* update

* fix a bug

* update

* fix a bug

* fix a bug

* fix a bug

* update dygraph record

* add infer shape record

* fix

* fix

* fix

* add comments

* fix a bug

* fix

* fix

* add record op info

* fix file mode

* add op input shape info

* fix dependency

* add new field for event node

* fix

* fix bug

* fix bug

* fix clang

* fix clang format

* fix code format

* [Cherry-pick] Apply IOU to test_parallel_executor_seresnext_base_gpu (#43812)
1. Fix the conflict between #43812 and current release/2.3 branch
2. test_parallel_executor_seresnext_base_gpu failed on 2 P100 GPUs with `470.82` driver.

* fix emb pass for ernie3.0

* fix emb pass for ernie3.0

* fix emb pass for ernie3.0

Cherry-pick of #43536

Backgroud in #43262

In elementwise_div UT, the numeric gradient (validation) has large relative error in comparison to analytic gradient (Paddle OP).

    The default rtol for UTs is 0.005
    The rtol for float32 and float64 elementwise_div OP is set to be 0.05
    The rtol for float16 and bfloat16 elementwise_div OP is set to be 1.0

The relative error is too large, so this PR provides user defined gradients to test elementwise_div followed by the analytic method.

* cherry pick 43890

* Update api docs (#42725)

* Fix max_pool3d doc, test=document_fix (#42715)

* fix pooling doc

* fix typo test=document_fix

* fix doc typo, test=document_fix

* fix adaptive_avg_pool1d doc bug (#42721)

* fix adaptive_avg_pool1d doc bug

* fix adaptive_avg_pool1d doc bug

* fix spectral_norm en doc (#42728)

* Fix example code bugs (#42739)

* update readme test=document_fix

* fix api docs bugs test=document_fix

* fix code example bugs;test=document_fix
Co-authored-by: NLinjie Chen <40840292+linjieccc@users.noreply.github.com>
Co-authored-by: NWei Shengyu <weisy11@163.com>
Co-authored-by: NWalter <dongshl1226@hotmail.com>
Co-authored-by: Nwangna11BD <79366697+wangna11BD@users.noreply.github.com>

* Update kaiming.py

* Update initializer.py

* fix doc bug;test=document_fix

* fix doc;test=document_fix

* Update initializer.py

* Update kaiming.py

* for ci;test=document_fix
Co-authored-by: NLigoml <39876205+Ligoml@users.noreply.github.com>
Co-authored-by: NLigoml <39876205+Ligoml@users.noreply.github.com>

* elementwise support

* commit

* update quantization clip and round

* fix quantization clip and round Attribute

* fix typo

* Create Tensor by paddle::empty  in custom operator (#41840)

* create tensor by empty in custom op

* fix some bug

* update relu custom op demo (#43173)

* Fix incompatible error for custom op Placetype (#43749)

* fix incompatible error

* rmeove default constructor

* add macro

* fix cpu make error

* add DefaultGPUPlace api
Co-authored-by: Nzyfncg <zhangyunfei07@baidu.com>

* lazy creating work queue

* fix dry_run

* Use all sitepackages path as the library/include path (#42940)

* Fix several unit tests and increase the unit tests stability (#43670)

* Reduce gather op unit tests size and increase the timeout

* Add NVIDIA_TF32_OVERRIDE for multi-processes environment

* Remove record test for device event ut

* Fix 3 unittest errors (#43532)

* Fix test_fuse_resnet_unit failure

* Fix test_imperative_auto_mixed_precision failure

* Fix sparse_attention_op error

* Fix sparse_attention_op error

* Use fixed random seed (#43659)

* for CI test_collective_sendrecv_api
Co-authored-by: Nzlsh80826 <rewang@nvidia.com>
Co-authored-by: NShijie <505749828@qq.com>

* fix structure infos conflict in static return_list mode. test=develop

* fix format. test=develop

* fix format. test=develop

* cherry pick form develop 43621

* code format

* paddle2onnx update to 0.9.8

* Correct elementwise quantization (#43693)

* [Bug fix] Do not quantize weights Y when matmul X and Y both other ops outputs (#43297)

* fix some matmul that X and Y both other ops outputs, do not dequantize the Y.

* fix CI format

* fix according to review
Co-authored-by: Njoanna.wozna.intel <joanna.wozna@intel.com>

* add bilinear_interp_v2 converter

* update op_teller.cc

* add unittest for bilinear_interp_v2 converter

* code format

* bug fix

* code format and add unittest

* remove merged modify in op_teller.cc

* code format

* code format

* fix scale init error

* fix conflict

* improve the doc

cherry-pick #42750。

QA反馈，#42750 优化后，solov2模型性能可提升6%，故cherry-pick到2.3。因#41096 将linspace python实现从fluid.layers.tensor挪到了paddle.tensor.creation下，该pr不在release/2.3分支中，故将#42750 中python修改同步到fluid.layers.tensor.linspace中。