* elementwise updated

* add support for constructing any order arg list

* add comment for using node_names

* use array instead of vector for args/output

* fix per review comment

* remove dead code

* Remove "view"
Simplify layout

* Fix merge error

* fix build error

* PR1602. IntelGPU backend. Compilation fixed.

* [AIBT-446] Run nGraph-ONNX tests on nGraph PRs

* Cloning nGraph from NervanaSystems

* Updated CI path to "ngraph/.ci/onnx/jenkins"

* Updated permissions

* Fixed path in find command

* Changes to notification message

* Editing CMakeCache.txt instead of removing

* Removing old ngraph wheels; cloning pybind11 if necessary

* Fixed path in removing ngraph wheel

* [ONNX] Non-linear operators

* Review fix pt. 1

* Review fix pt. 2

* Non-linear tests

* style check

* Exception fix

* Test fix

* add unsupported_op exception

* unsupported_op test

* add printout of unsupported op in model

* fix GPU dispatcher check

* fix test designation

* catch exceptions on single file runs too

* add unsupported_op exception where needed

* remove unsupported_op class

* add unassigned op exception

* add unit test

* catch unsupported op in nbench

* add cpu test back

* update all latest merges

* mode change

Clang chooses to use a __vectorcall optimization in which
address pointers are vector loaded in the gpu::invoke_primitive.
This results in a segfault when stack alignment is absent.
Since the GPU transformer does not rely on CPU for compute,
we disable the optimizations of the emitted function.

* Modify DEX OneHot op: use generator.

* Cast index to int.

* topological sort with cdeps

* add control deps API, fix unit tests

* rollback adjoints changes

* fix test failures,add more tests

* remove dead code

* address scott's feedback

* Add in_place suport for ReplaceSlice

* Add emit_replace_slice_inplace kernel

* changed file permissions to original

* Formatted code using maint/apply-code-format.sh

* Removed data type check and removed dead code

* Removed setting mkldnn_op(true). ReplaceSlice is not mkldnn op

* Add missing header.

* Test for ReduceSum

* Simple tests for reductions

- L1/L2/LogSum/LogSumExp/Max/Mean/Min/Prod/SumSquare.

* Add floating point literal suffix

* Fix typo

* Update fusion doc and add ONNX build flag to buildlb doc

* Fix PR comments

* Final PR review comments addreswsed

* Fix link on reformmatted doc README

* Delete index.rst.save

* add option to copy intput/output data for each iteration

* add support for stale buffers

* Add support for Quantized Pooling(Max + Avg) op via mkldnn for IA backend (codegen + DEX)

* Add checks for min and max

* Extracting out the common code from codegen and DEX

* Use call_with_validate

* [ONNX] Shape operator

* Review fix pt. 1

* Style check

* wip

* interperter use switch instead of if/else

* more cleanup

* make nop elimination run on all backends

* revert

* use single include file to define all ops so there is only one instance

* move op.tbl to ngraph/op dir as it is useful. Added useage example.

* add some comments where needed

* revert some changes to reduce delta

* add const

* add more const

* simplify using NodeWrapper

* update per review comments

* update per review comments

* update per review comments

* remove switch warning as it is not supported in older gcc

* Add conv add fusion

* Updated file permissions and cpu_fusion order

* Formatted code using maint/apply-code-format.sh

* Fixed minor review comments

* Use NODE_VALIDATION_ASSERT instead of throw ngraph_error;\nupgrade baseline and fix issues

* Some more fixes

* IntelGPU backend: BatchNorm operation optimization

* PR1579. Function moved by request

* ReduceSum and ReduceSumSquare ONNX operations.

* Add new reduction ops.

- ReduceLogSum,
- ReduceLogSumExp,
- ReduceMax,
- ReduceMin,
- ReduceMean,
- ReduceProd.

* Add ReduceL1 and ReduceL2

* Utility generic functions generating monotonic sequences of values.

* Review comments: return AxisSet not std::vector

* Use common functions for generating monotonic sequence.

* Review comments.

* Add support for Dequantize op via mkldnn for IA backend (codegen + DEX)

* Remove unused variable

* Static cast target range

* constant + pad

* adding broadcast test back

* Implement TopK.

* Update python wrappers for TopK, ArgMin and ArgMax.

* Address some reviewer comments.

* Add type property check tests for TopK.
Set correct TopK behavior for K==0.

* TopK: Add 1d and 3d unit tests.

* Address more reviewer comments.

* Apply code style.

* Update doc build v and fix doc on captioning

* Clarify to build the library

* update link on README

Double curly-brace initialization (required by clang for non-templated functions) causes a compiler error in centos. (#1561)

Since the warning is not enforced in clang for templated functions, we can get around the centos compiler error with only a single set of curly braces here.

* onnx: add missing header files
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnxifi: implementation of onnxGetBackendIDs
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnxifi: add unit tests for onnxGetBackendIDs
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnxifi: change std::out_of_range to std::length_error
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnxifi: after review changes
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* simplify result copy elimination

* gpu fix

* remove include header

* circumvent gpu issue

* add a whitepace

* add cuda reduce

* clang format

* fix bugs

* fix bug

* add 1d reduce

* clang format

* fix bugs

* unroll loop

* remove debug info

* revert tests

* unroll 1D reduce op

* add comments

* using cudnn for nd to scalar reduction

* remove cuda 1d reduction since cudnn version is faster

* remove 1D kernel

* fix bugs

* 1d multi block size

* remove debug

* change kernel name

* add reduce to scalar optimization, add test

* fix bugs and tune parameters

* clang format

* update comments

* update comments

* update comments

* clang format

* update comments

* remove wrong comments, apply clang format

* resolve Bob's comment

* clang format

* pass shared mem size from cuLaunchKernel, set unroll loop size through host code

* remove unused code.clang format

* change reduce to thread with shfl for each warp first

* add seed

* unroll size