* [ONNX] Add documentation

* Update documentation contributor's instructions

* Doc theme to match ngraph.ai

* Minor formatting fixes and PR feedback

* ToC fixes

* ToC fixes

* Add changes

* Be consistent with BUILDDIR

* Be consistent with substitution

* Update Makefile

* Update README.md

Lots of updates to the broken links, typos, and other improvements to be noted for the latest release.

* Update README.md

fix typo

* Try asciidoc version of README

* Update README.md

* Tox is stupid

* Add both versions for kicks

* Add PR review feedback

* Remove text version of readme

* [ONNX] Add documentation

* Update documentation contributor's instructions

* Added extra case for conv bias affine folding

* Added unit test

* Check for fused ops in unit test

* [ONNX] Change default broadcast dim value, gemm improvement

* update comments: ones -> zeros

* Add erf op, interpreter kernel and unit tests

* Add review feedback and fix unit tests

* Add unsupported op to GPU and IntelGPU

* Add one more test

* Do not set num_threads_per_pool to 1 in cpu executor.
Ignore cacheability in memory assignment when not reusing memory.

* Fix debug output.

* [WIP] First commit

* Remove some commented code

* Further changes

* Further changes

* Add method to remove patterns with just one concat

* Add tests

* Add more tests

* Fix fan out case

* refactor code

* refactor code

* Added NGRAPH_DEBUG statements

* Use INTERPRETER as backend instead of CPU...travis build failure

* clang

* minor edit

* add more checks in the tests

* Incorporate Bob's comment

* Removed some NGRAPH_DEBUG statements and incorporated Pruthvi's comment

* Incorporate Xiaoyu's comments

* some refactoring

min_signal is intended to optionally skip checking float
distances when numbers are close enough to 0.
Only skips when both numbers are < min_signal.
Intention is to allow tighter float testing in certain
cases where most values are not near 0, but values near
zero are differing by more bits than values farther from 0.
Should be used with caution in limited cases.

* i) added support to handle default convolution kernels

* - added test case for rotated convolution

* added debug info for conv exception

* disbale handling conv exception

* WIP onnx unit test reproducer

* - Fallback to reference kernel if its negative padding for convolution

* - corrected the conditional check for -ve padding in conv

* added dyn_reshape and dyn_slice.

* style fix.

* some fixes.

* added dyn reshape type prop.

* fixed gpu build.

* added headers to gpu emitter.

* Inital DynPad implementation

* Inital DynPad implementation

* Fixed DynPad validation. Added Unit-test

* Nits and white-space fixes

* - PR feedback.
- Added padding rank check tests

* Minor comment fix

* Fix merge error

* Fix accuracy issue

* Style fix

Cleanup runtime context in CPU_CallFrame Destructor

Moving the channel part of the sum into an inner loop lets padding be ignored
once per position rather than once per channel/position. This order of non-precise
summation also seems to be closer to other convolution implementations.

Add options to test generator for controlling precision of tests for debugging purposes,
and fix another python3 problem.

* Delete old stack diagram

* Update first paragraph

* Update introduction.rst

* Update introduction.rst

* Update introduction.rst

* Update introduction.rst

* Introduction graph optimization diagram

* Update introduction.rst

* Introduction kernel to deep learning frameworks

* Update introduction.rst

* Update introduction.rst

* Delete intro_kernel_to_fw.png

* Introduction kernel to frameworks with accent

* Update introduction.rst

* Update introduction.rst

* Update introduction.rst

* Introduction multiplying kernels

* Update introduction.rst

* Update introduction.rst

* Update introduction.rst

* Update introduction.rst

* Update introduction.rst

* Delete intro_kernel_to_fw_accent.png

* Introduction kernel explosion image

* Update introduction.rst

* Update introduction.rst

* Update introduction.rst

* Edit introduction to be relevant to a wider audience

* Refine intro

* writing thru another draft

* Another batch of updates

* Tie the conclusion to the intro

* Clean up the doc

* First reviewer feedback done!

* fix typo

* fix typos

* Minor formatting fixes

INTELGPU: Fix names duplication of generated supporting primitives for Convolution* operations (#2642)

* Adds new core op DynBroadcast

* Adds new core op DynBroadcast

* Fixes build error caused by recent changes in node validation API

* Addresses code review comments.

* Moves new op under experimental.

* Fixes style errors.

* Silee2/external project rpath (#2525)

* Set rpath for mkl-dnn.

* Set library rpath to  on Linux.

* Use patchelf to set rpath of prebuilt mklml library.

* Add patchelf to Linux Dockerfiles.

* Revert "Add patchelf to Linux Dockerfiles."

This reverts commit 1769505a866061552942e19467ddcc0dad0922e8.

* Revert "Use patchelf to set rpath of prebuilt mklml library."

This reverts commit 726f6553a0450520328607177d64baf48fa93dd2.

* Copy cldnn runtime.

* Copy mlsl libraries.

* add unit tests for the two versions of Backend create_tensor (#2607)

* add unit tests for the two versions of Backend create_tensor

* disable new unit test on GPU until we have time to address it

* Resolves merge conflicts

* Addresses code review comments.

* Fixes merge issues

* Fixes style errors

* Fixes type check to use compatible()

* Reverts unintenional change

* Reverts unintenional change

* Fixes typo in comment

* Addresses code review comments.

* Add support for negative padding

* Use std::bind in pad builder check

* Add support for negative padding in CPU backend

* Updated kernel to do pad+slice

* Remove type conversion warnings

* Fix review comments

* Remove interior padding from core op and interpreter stuff

* Update backends other than GPU for retirement of padding_interior

* Skeleton of support for edge/reflect padding

* Post-merge cleanup

* Attempt reference implementation for EDGE.

* Fix the edge-padding reference, and add some unit tests

* Implement REFLECT padding ref; add tests

* Fixes to the CPU stuff so it compiles now

* Fix test

* Add support for different pad modes

* Restore a stub get_padding_interior function, and tweak some stale comments

* Update ONNX importer to not supply interior padding value; add checks for padding-too-small for EDGE and REFLECT

* Typo

* Bop a warning

* Attempt fix to INTELGPU backend

* Attempt another fix to INTELGPU backend

* Fix pyapi

* Style apply

* Add support for padding modes

* Remove unnecesary node validation checks

* Remove tests for minimal reflect and edge pad

* Remove commented tests

* Remove unnecesary Asserts

* Little update of pad documentation

* Monospace for pad_mode options

* Revert "Remove tests for minimal reflect and edge pad"

This reverts commit 81e4787ea47195b832cab1452dde698bc05776fe.

* Revert "Remove unnecesary node validation checks"

This reverts commit 7e68db7564f3c9b1fd40e7db1d1bda4e0677cad9.

* Test only spatial dims

* axis -> spatial axis

* Fix typo

* Style check

* Update test

* Add CoordinateDiff include

* Remove pad_mode from tree visualization

* Convert padding into NVShape

* Skip failing tests on GPU

* Revert mode change

* Remove merge artifact

* Rename pad kernel into pad_ref

* - add method to conditional check for mkl version
- use conv_auto if feature is available

* fix conditional check for mkl version

* fix typo

* - added funx to query mkldnn_version

* WIP debug

* -   added test case for conv winograd algorithm selection

* i) style fix

* - fix clang errors

* - fix clang error ( clang has issue around conditional operator)
- changes to mkldnn::utils conv_heuristics helper to return mkldnn:algorithm instead of bool value

* - add right checks based on the choosen conv algorithm in the unit test

* - utililty function to determine isa_type in unit test

* - address PR comments
- add unit test for conv_winograd and executes with CPU backend

* i) fix style ii) remove unit test

* Localize mkldnn_version() based decisions to mkldnn_utils

* Handle mixed length sequences.

* UT for mixed sequence length LSTM.

* Style apply.

* Fix typos.

* Add std:: prefix to preserve unified style within file.

* Fix debugger tests crash for CODEGEN.

* Address PR feedback.

* Expect exception for CODEGEN.

* Address PR feedback.

* Do not allow in place slice if the arg is in CONSTANT buffer set.

* Add a unit test.

* Address PR feedback.

Pre-build tbb wasn't getting its headers into the master header
Reference convolution changes needed to be added to codegen

* Add NodeInput and NodeOutput classes

* Review comments

Also fix formatting

* Update sitemap.rst

* Update sitemap.rst

using the structure suggested by Kevin P instead

* add method to get underlying executable

* Copy the function because copy cleans up the graph linkage/users

* style

* change float comparisons from == to all_close_f

* style

* address a few more direct float comparisons

* add missing include

* specify tightest tolerance for Broadcast and Reshape tests

* Increased tightness of float testing

Increased tightness of float testing via MIN_FLOAT_TOLERANCE_BITS parameter

* style

* Added sqrt to constant folding

* Added negative sqrt checking

* add copy_from on GPUTensor

* add test

* remove unused code in test

* fix warning

* fix type

* fix comment texg

* using all_cloase for test

* Enable onnx, protobuf and onnx import on Windows.

* Apply style.