* Re-organize files. Create MLIR backend classes

* WIP

* Refactored. Code compiles

* Moved context to Runtime class to outlive compilation and execution

* style-apply

* Base Runtime class. Few other modifications

* Minor fixes

* Fixed Runtime::run() to take type-erased pointer

* renamed core compiler

* rename backend compiler

* rename runtime compiler

* PR feedback

* Fix build fails

* - refactor LowerNGDialect
- addgraph pass to lower standardDialect to llvm module

* Addressed PR Comments

* i) Style fix ii) restore deleted file

* Introduced reshape:v1 downgrade

* Enable downgrade pass in dynamic backend

* Add unit test for dynamic backend downgrade pass

* Clang styles applied

* Apply unit tests name refactor based on code review
Co-Authored-By: Michał Karzyński <postrational@users.noreply.github.com>

* Removed redundant pass

* Changed order of downgrade pass

* Changed order of passes in dynamic backend

* Rename input arg -> data

* Update node validation and shape propagation to support data rank and number of repeats to be different.

* Add type_prop tests

* Add UT for new tile cases

* style

* Modify CPU backend to support the new case.

* Fix backend selection in test

* Add new line back.

* Re-organize files. Create MLIR backend classes

* WIP

* Refactored. Code compiles

* Moved context to Runtime class to outlive compilation and execution

* style-apply

* Base Runtime class. Few other modifications

* Minor fixes

* Fixed Runtime::run() to take type-erased pointer

* renamed core compiler

* rename backend compiler

* rename runtime compiler

* PR feedback

* Fix build fails

* DynSlice was adjusted to specyfication

* Fixed documentation

* Fixed styles

* Fixed clang warning

* Introduced StridedSlice as separate op

* revert dyn_slice

* Fixed StrideSlice implementation

* Enable downgrade pass in dynamic backend

* Implemented upgrade/downgrade pass

* Added serialization

* Styles applied

* Styles applied. Part.2

* Updated dynamic backends list

* Move StridedSlice outside experimental

* Make Slice as v0 of StridedSlice

* Styles applied

* Styles applied

* Fixed clang error

* Code review remarks introduced

* Move GenerateMask up in Opset0Downgrade to keep alphabetical sort

* Style apply

* Initial draft start

* Add provenance doc

* Another effort at resolving doc build warning

* Revert file

* Revise provenance intro

* Ensure Clang dependencies update with new minimum version of clang needed

* PR review feedback, cleanup other change added to collab_ngai

* - WIP fusion pattern for softmax + cross entropy

* fix compiler error

*  make summation axis integer for the fusion pattern

* - Fusion pattern for sigmoid cross entropy bprop

* WIP callback implementation for fused sigmod+crossentropy fprop

* - implemented fprop softmax+crossentropy as single layer for numerical
stabilization
- added broadcasting nodes to fix elementwise assertions

* Added unit test case for functionality test

* Move the softmax + crossentropy fusion pass to core

* i) style fix ii)added missing header

* - Added new Fused Op for Softmax + CrossEntropy
- moved the decomposition to the Softmax + CrossEntropy FusedOp

* - Add SoftmaxCrossEntropy for fused tablegen
- Add serializer support for SoftmaxCrossEntropy
- fix documentation

* Added missing json file for unit test case

* Addressed PR comment

* Addressed PR comments

* - Fix fusion string

* - Style fix

* - Added Bprop for Softmax + crossEntropy

* - added SoftmaxCrossEntropy support when soft_lable is provided
- serailizer and deserializer support for SoftmaxCrossEntropyBprop

* - Added support in decompose_op for SM+CE bprop when ignore_mask is specified

* Updated Doc strinng

* - unit test case for SoftmaxCrossEntropy backprop with soft lables
- fixed decompose_op bug in bprop

* - if soft_label=true, capture pattern only if the labels dont have one
hot encoding

* - SoftmaxCrossEntropyBprop Support if ignore_index is specified

* add serialized files for unit test

* - fix softmax + CE pattern bug
- fix softmax + CE decompose_op() bug

* - change reduction_axes to int64_t type in fprop and bprop ctor

* - add soft_labels and ignore_index attribute to SM+CE fprop ctor

* - addition asserts in unit test to ensure SM + CE fprop and bprop fusion is successful

* - move reduction_axis computation to decompose_op from ctor to relax on
dynamic shapes

* Addressd PR Comments

* - suppprt for SM+CE for ignore_index and softmax=false

* - test case for SM+CE fprop with ignore_mask, soft_labels=false
- fix bug in decompose_op

* - refactor unit test case

* - fix PDPD unit test

* broadcast delta if shape mismatches

* -fix bdcast issue in decompose_op

* Concat operator is updated to support dynamic_shape
1) Added new concat_negative_indexing test
2) Replaced size_t -> int64_t
3) Support for a negative indexing, calculate axis = axis + int64(input_rank)
   if (axis < 0) { axis = axis + int64_t(this_input_rank); }

* Remove unwanted #include "ngraph/op/constant.hpp" header

* Refactoring:
1) The name of variable is replaced: m_concatenation_axis -> m_axis

* Concat negative indexing test is adjusted to support dynamic_shape tensor
auto pshape_a = PartialShape::dynamic(); for each tensor {a,b,c} result tensor has dynamic_shape

* The backend supports dynamic shapes
auto backend = runtime::Backend::create("${BACKEND_NAME}", true);

* Other shape is supported by concat:
set_output_type(0, inputs_et, PartialShape::dynamic(concatenation_axis_output_dim));

* The NODE_VALIDATION_CHECK was moved up to be for a dynamic_shape

* [Test] The shape of output tensor was changed
[Concat CPU] Added support for a negative indexing on a cpu

* Review changes:
1) Added axis re-calculate for a reference version of concat
2) axis is not replied

* Review changes: support for a negative axis

* Comment about variable is added to concat.hpp file
Removed unused variable

* Added auto generated configs for nGraph

* Fixed absolute paths

* Fixed comments

* Add constant folding for v1 reduce ops.

* Add reference/mean.hpp.

* Remove extra semicolon.

* Address PR feedback.

* Add constant folding for Squeeze and Unsqueeze.

* Address PR feedback.

* add partial slice op placeholder

* fprop implemented

* add a dynamic ut

* implement bprop

* support decrease axis

* add bprop ut

* address comment

* use descriptive var name

* Updated nGraph backend to be compatible with OpenVINO plugins

* Resolve issue with gtest

* Resolve comments in PR

* Resolve comments in PR

* Code style fix

* Rename runtime2 to ov_runtime

* Style check fix

* Dynshape support for GenerateMask

* fix clang error

* Remove comments

* Test case correction

* Disable plaidml

* Merge

* Added relative paths for python build

* Fixed Ninja generator

* Implemented downgrade pass

* Added v1 support in onnx importer

* Clang style applied

* Pooling factory refactor

* Removed unused variables

* clang styles apllied

* Fixed avg pool bprop UT

* [MLIR] Modify mlir subgraph extraction pass.

* Address PR feedback.

* Modify sub-graph construction algorithm.

* Address PR feedback.

* Address PR feedback.

* Change macro to function.

* Remove debug output

* More cleanup

* unfold attributes

* Remove unnecesary if

* Rename run() -> lstm_pass()

* Unify usage of LSTMForward for one and bi directional LSTM

* Unify LSTMForward return values for one and bi directional LSTM

* Dirty moving LSTMForward into fused directory

* Accept lstm direction as string instead of enum

* Fused op which uses decompose_op in onnx_importer

* Rename LSTMForward -> LSTMSequence

* Split LSTMSequence to cpp and hpp

* Remove LSTMDirection enum

* Add getters for class fields

* Adjust constructors

* Add direction validation.

* Add support of LSTMSequence op in serializer

* Reorder fused op input order

* Style fix

* Fix for reorder of inputs

* Use NodeTypeInfo instead of static string

* Node -> value in doc

* Add doc for prepare_input method

* Fix shape inference

* Use enum instead of string for direction

* Add Type prop unit test

* Fix style

* Check size requirement before creating scratchpad.

* Check max scratchpad size before allocating scratchpad_buffer.

* Add the same checks for CODEGEN.

* Fix unused-parameter warning.

* Fix a typo.

* Address PR feedback.

* Fix a bug.

* Fix style error.

* AvgPoolBprop dynamic shape support

* Add opset_transform test

* Call the correct shape relevance function

* Unused variable

* Support auto broadcast in FakeQuantize op

* style fix

* add pdpd unit test to plaidml manifest

* Make axis for TopK dynamic

* set_input_is_relevant_to_shape

* Proper handling of output shape with dynamic top_k_axis

* fix ut

* remove unused line

* check is_constant

* add v0 version

* update provenance

* use set_argument

* fix clang

* Minor cleanup

* Add support for ONNX 1.5 version of TopK

* Add unit tests

* Style apply

* Exclude failing tests

* Exclude failing tests

* Add support for ONNX 1.6 TopK attribures: larges and sorted

* Support for ONNX 1.6 TopK

* If k_node is a Constant, recreate as constant with Shape{}

* Extend `interpret_as_scalar` function

* Extend `interpret_as_scalar` function

* Remove merge artifact

* Add doc string

* Exclude failing tests

* Exclude failing tests

* Refactor function

* Remove unnecessary template param

* Use get_k function in OpSet 10 TopK

* Style apply

* Remove merge artifact

* Add tests for `interpret_as_scalar`

* Revert "Add tests for `interpret_as_scalar`"

This reverts commit 8b85965acb39c75ff9e66b06ad8f64df16e1a9da.

* - Added support for v1 Broadcast op specification
- Added upgrade/downgrade conversions between v0 and v1

* Added unit test for pdpd broadcast

* Make numpy default autobroadcast type and some style fixes

* Added support in Dynamic wrapper for dyn elimination and copied over unit tests from DynBroadcast

* Addressed PR feedback

* Addressed PR feedback on documentation