* [nBench] Add support to dump result tensors

A new flag '--dump_results' is added to dump result tensors to std::out.

* address feedback

* fix enum unhandled
Co-authored-by: Ashok Emani <ashok.emani@intel.com>
Co-authored-by: Robert Kimball <robert.kimball@intel.com>

Co-authored-by: Evgenya Stepyreva <evgenya.stepyreva@intel.com>

* There is the files "non_zero.cpp & non_zero.hpp" for which operator non_zero is defined"
The two files were added: non_zero.cpp & non_zero.hpp

* The pull request present a study of extension of the NonZero operatos used for ML models:
1) The various type of files was modified: cpp & hpp, onnx files & ng files
2) Non_zero operator supports the onnx standard (there are added the files in which the operator is enabled: non_zero.cpp & non_zero.hpp)
i

* Operator description: refactor

* GPU emitter - is not implemented

* GPU emitter: v0 version

* NoN zero: the input and output is adjusted to output shape & input_shape

* Move the NonZero op to v3 namespace

* Move NonZero to opset3

* Correct shape inference for NonZero

* NonZero op constant folding for 0D and 1D inputs

* Constant folding for NonZero op

* Correct output shape for NonZero & scalars

* Helper function to test NonZero

* NonZero constant folding UT for floats

* Enable more data types in NonZero CF

* NonZero type prop tests

* NonZero constant folding tests (directly)

* Use is_constant instead of casting in UT

* NonZero op doxygen docs

* onnx_importer docs adjustment

* Correct version of the NonZero core op

* Disable NonZero in GPU backend

* Short circuit if all elems in data are identical

* find_indices() optimization

* Assert on the input shape in NonZeroElements

* CF of NonZero with all non-zero values

* NonZero CF test for scalars

* bool support in NonZero

* Missing include in NonZero CF

* Dont throw if NonZero CF fails

* Update src/ngraph/pass/constant_folding_non_zero.cpp
Co-Authored-By: Robert Kimball <robert.kimball@intel.com>

* Removing warning
Co-authored-by: tomdol <tomasz.dolbniak@intel.com>
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>
Co-authored-by: Robert Kimball <robert.kimball@intel.com>

* Experiment with removing Threads

* Use private Threads

* Removed Threads

* Added Threads to benchmark

* Revert Threads

* Remove threads
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* Better performance for graph.huge test. More robust Function validation

* Fix parameter replace in unit test
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* Update includes in reference runtime

Add additional headers to reference runtime files to make them self-sufficient

* style
Co-authored-by: Robert Kimball <robert.kimball@intel.com>
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* [ VariadicSplit ] Dynamic shape inference

* Small code fixes

* Add tests for partial shape inference

* Style-apply

* Style-apply
Co-authored-by: Michal Karzynski <michal.karzynski@intel.com>
Co-authored-by: raramer01 <rebecca.ramer@intel.com>
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* Add value_type to Dimension, deprecate int64_t cast

* Need a conversion

* Initial PR to enable embedding images in Doxygen

* Address review comments

Signed-off-by: Alexander Peskov <alexander.peskov@intel.com>
Co-authored-by: Alexander Peskov <alexander.peskov@intel.com>

* remove dynreshape

* remove header

* use v1::Reshape for shape relevance

* remove return of bool due to function refactor

* remove gpu emit v0 DynReshape

* Merge cleanups

* throw exception for reshape node with dyn shape

* fix style
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>
Co-authored-by: Wenzhe Xue <wenzhe.xue@intel.com>
Co-authored-by: Scott Cyphers <scott.cyphers@intel.com>
Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>

* Let doxygen handle NGRAPH_API macro on objects

* More fixes to doxygen config fdile

* Cleanup

* Fix ngraph as a class

* Add missing API

* Fix build error

* [MLIR] Use mkldnn callback for ConvBias.

* Add try catch.

Fix opAttrsVec.

Add rank check for Gemm and MatMul.

* Fix merge error.

* Fix a bug.

* Fix lit test.

* Modify unit test.

* Fix merge error.

* Address PR feedback.

* Address PR feedback.

* Insert callback_init function to module.

* Fix lit tests.

* Fix a bug.

* Use a set of GlobalOps for attributes.

* Address PR feedback.

* Address PR feedback.

* Fix merge error.

* Fix style error.

* Fix style error.
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* [ONNX] Add Tile op

* Fix test model output shape

* Add static test for Tile

* Test Case: use expected shape, not computed output shape

* Exclude dynamic Tile op tests

* Add tests using BackendMode::DYNAMIC

* Code review comments
Co-Authored-By: Katarzyna Mitrus <katarzyna.mitrus@intel.com>

* Apply suggestions from code review
Co-Authored-By: Katarzyna Mitrus <katarzyna.mitrus@intel.com>

* Code review
Co-authored-by: Katarzyna Mitrus <katarzyna.mitrus@intel.com>
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>
Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>

* Enable all of -Wall

* Revert 32 bit change
Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>

* [ONNX] Enable cross-compiling and ninja build for onnx_importer

- Change minimum cmake version to 3.11
- Enable cross-compilation of protobuf
- Rewrite ONNX external project into FetchContent
- Fix some issues with ninja build on windows

* Remove -Werror from ONNX CMAKE_CXX_FLAGS

* Unit-test: Link gcpu_backend if NGRAPH_GENERIC_CPU_ENABLE is ON

* Remove overwriting of main CMAKE_CXX_FLAGS from the onnx cmake file

* Fix onnx usage of CMAKE_CXX_FLAGS

* Ignore warning for onnx on windows, restore Windows CXX replacement

* Resolve some problems on OV ci

* Fix missing quotes

* Ignore range-loop-construct warning in onnx

* Just stop be pedantic for onnx

* Remove -Werror as well

* Remove unnecesary code

* style

* Add NGRAPH_API macro to builder::opset1::reorder_axes for OV transformations

* Move NGRAPH_API to header
Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>
Co-authored-by: Robert Kimball <robert.kimball@intel.com>
Co-authored-by: Scott Cyphers <scott.cyphers@intel.com>
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* Stricter checks on padding attributes for v1::Convolution

* Move pad argument checks to op constructor
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>
Co-authored-by: Michał Karzyński <postrational@users.noreply.github.com>
Co-authored-by: Robert Kimball <robert.kimball@intel.com>
Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>

* Added check if getting rank is needed.

* Removed normalization from importer.

* Added normalisation to Concat op.

* Added test for Concat.

* Added test for Unsqueeze

* Added missed test.
Co-authored-by: Michał Karzyński <postrational@users.noreply.github.com>

* more fixes

* Fix test

* Cleanup comment

* style

* Update per review comments

* Change get_arguments to input_values
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* GetOutputElement removal preparation

Not all outputs are used so don't force them to be connected in replace
Add pattern that matches on any output
Remove GOEs by default, allow to disable
Fix failing core passes/tests with GOE dependency

* Fix MLIR call

* Fix value handle assignment

* Cleanup

* style

* review comments

* Fix onnx tests

* Allow simplifcations to work on multi-values nodes

* Disable goe removal for MLIR test

* null init of Output
Co-authored-by: nmostafa <nagy.h.mostafa@intel.com>

* Trying to compile with hidden visility

* Export more functions needed for Inference Engine

* Update CMakeLists.txt
Co-authored-by: Robert Kimball <robert.kimball@intel.com>

* float16 round halves to even

* NaN details are platform-dependent

* typo

* typos
Co-authored-by: Ilya Churaev <ilya.churaev@intel.com>

* Use v1::ReduceMean in ONNX ReduceMean

* Add support for keep_dims

* Replace binds with lambdas
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* test

* Don't indent reserialized graph

* Code cleanup.

* Update LLVM to 9.0.1
Co-authored-by: Robert Kimball <robert.kimball@intel.com>
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

* Split api to onnx_utils, headers refactor

* Improved ONNX importer API doc

* styles applied

* Code review remarks introduced

* remarks introduced, style applied

* Code review remarks introduced

* namespaces description added

* Code review remarks introduced

* styles applied

* Fixed onnx_utils doc and margin alignment

* code review remarks introduced
Co-authored-by: Michał Karzyński <postrational@users.noreply.github.com>

* Cleanup get_output_shape

* More cleanup

* Update more

* output tensor

* fix input

* get input source ptr

* cleanup

* more changes

* wip

* wip

* more input helpers

* style
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>

* remove placement index

* style

* Fix overwriting of value

* style

* Style

* remove operator size_t in dimension

* use get_length()

* transform uint64_t to size_t for func to_shape()

* more replacement of get_length

* more replacement of get_length in test

* make style-apply

* change in onnx

* more change onnx test

* fix build

* add size_t back to dldt test
will do another PR after dldt update
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>
Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>
Co-authored-by: Robert Kimball <robert.kimball@intel.com>

Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>
Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>

Co-authored-by: Scott Cyphers <diyessi@users.noreply.github.com>
Co-authored-by: Sang Ik Lee <sang.ik.lee@intel.com>