Skip running onnx tests in python mac os pipeline (microsoft#5416)

orttraining_rc3.1 release

orttraining_rc3 release

fix quantization of EmbeddingLayerNorm (microsoft#5321)

fix quantization of EmbeddingLayerNorm (microsoft#5321)

For ORTTraining RC2 release to include hotfix microsoft#4639.

It will be used to create docker image for orttraining examples.

Onnxruntime 1.4.0 Release Tag

Cherry pick more commits to Rel-1.3.1 (microsoft#4231)

* Fix deprecated CentOS link for Linux CI pipeline (microsoft#4000)

* Fix Linux_CI_GPU_Dev

* centos6

* Fix crash reported in microsoft#4070. (microsoft#4091)

* Fix crash reported in microsoft#4070.

* Add newline to warning message

* Add comment for using cout instead of the logger

* fix optional input/outputs (microsoft#4229)

Co-authored-by: Ryan Lai <[email protected]>
Co-authored-by: Pranav Sharma <[email protected]>
Co-authored-by: Tracy Sharpe <[email protected]>

Cherry pick fixes to release branch rel-1.3.0 (microsoft#3936)

* Fix DirectML nuget creation in Nuget pipeline (microsoft#3929)

* Added onnxruntime aarch64 wheel to pypi publishing pipeline (microsoft#3903)

* Added onnxruntime aarch64 wheel to pypi publishing pipeline

* Support nightly build flag

* Add support for nightly build

* Fix error handling in LearningModelSession.cpp (microsoft#3920)

* Update DML Nuget version and DML EP Doc (microsoft#3945)

Update DML Nuget version and DML EP Doc

* Fix ordering of APIs. (microsoft#3951)

Co-authored-by: Ryan Lai <[email protected]>
Co-authored-by: Prabhat <[email protected]>
Co-authored-by: Jeff Bloomfield <[email protected]>
Co-authored-by: Pranav Sharma <[email protected]>

Cherry pick 3 fixes to rel-1.2.0 (microsoft#3158)

* Publish release symbols (microsoft#3152)

* Publish release symbols

* Publish symbols if IsReleaseBuild

* Disable delayload for cuda dlls (microsoft#3147)

This change fixes microsoft#3129. When running onnxruntime as dll on Windows, CUDA does some internal cleanups when process exits. After this, any call to CUDA would cause crash. Delayload makes thread_local destructor to happen after CUDA cleanup, thus the crash.

* Update Gelu Fusion to support new graph pattern from PyTorch 1.4 (microsoft#3148)

* update GeluFusion to support pattern from PyTorch 1.4; 
* Fix a bug that missing the check of an edge between mul2 and root.
* update script to fuse gelu from PyTorch 1.4
* Add test for python optimizer

Co-authored-by: Tiago Koji Castro Shibata <[email protected]>
Co-authored-by: KeDengMS <[email protected]>
Co-authored-by: Tianlei Wu <[email protected]>