* Divert dynamic offset to reference

* update gpu manifest

* Avoid dynamic cast

* Retire Reduce, ReduceWindow, SelectAndScatter

* Remove lingering AnyAllReplacement code

* Remove apparently-now-unused macro

* Remove lingering op/reduce.hpp includes

* Remove FunctionCall

* Update Python stuff to remove FunctionCall, Reduce

* Add Any/All tests to GPU manifest

* Remove deleted .hpp #include from gpu_compiled_function.cpp

* Restore reduce_function.cpp since that is where the All/Any implementations ended up residing

* Add reduce_function.cpp back into CMakeLists.txt

* Remove #include of deleted reduce.hpp

* Re-remove AnyAllReplacement from GPU passes

* Remove deleted #includes from gpu_internal_function.cpp

* Remove git conflict stuff (whoops)

* Add newline at EOF, per review comment

* Address flake8 complaint about unused import

* Simplified & tightened all_close_f parameters

Removed specification of mantissa bits for all_close_f in favor
of just specifying tolerance bits. Tightened up all_close_f default.
Fixed LRN unit test which had insufficient result precision to pass
tighter all_close_f tolerance.

* Addressed PR comments.

Reworked mantissa bit and tolerance constants.
Clarified and improved graph comparison tolerance calculation flexibility.
Clarified unit test tolerance testing.

* update licenses for 2019

* style

* make validate public

* move compile call outside of call for unit tests

* fix compile error

* one more error

* Enable in place slice when slice's arg is function input.

* Add a corner case.

* Add codegen support.

* Set the correct offset for in-place-slice when there is a chain of in place ops starting from parameter.

* files pulled from bob/winbuild

* fix compile problems

* fix a few windows build errors

* add windows file to exclude from git

* add comment why change was made

* revert obsolete change

* more cleanup

* building interpreter and unit test on windows with DLLs

* Add flag for windows to export all symbols. Short term fix.

* enable MD build

* address warnings

* dump all windows build results to a single directory

* fix windows backend dll open issue

* remove debug

* fix file iterator for windows

* fix merge error

* fix test failure

* change header from h to hpp in hopes of making python happy

* address more linux build issues

* fix visibility enable

* change compile call to return Handle

* make CPU require compile() before call()

* fix unit tests to call compile() before call()

* fix failing ops

* update unit test

* revert some changes

* more fixups

* more diff cleanup

* a few more issues addressed

* more fixes

* update API

* more updates

* fix test_ops.py

* fix

* another attempt to fix

* fix unit test

* fix test error

* - modified cpu_assignment pass to support bn with input 5D
- added test cases for 5D bn and 5D bn+relu

* - Address PR comments
- used mkldnn_utils to validate bn for mkldnn

* fix compilation error

* Addressed PR comments
- added helpers in mkldnn_utils for assigning ngraph Op as MKLDNN op
- helper funnction for bn mkldnn assignment

* fix clang error

* Fix batchnorm argument order, cleanup some comments, fix backprop

* Merge error

* Clean up training function, organize inference test

* BatchNormInference tests

* Training case

* Training test

* Fix autodiff BatchNorm test

* Cleanup

* Move file to doc checkout

* Update disabled test name in igpu manifest
Fix unnused variable

* Unit tests disables

* Review comments

* add cuda reduce for product and max and added tests

* a quick fix for empty reduce axis and 0 axis

* adding min cuda reduce

* add for min

* fix bug and format

* add another min test

* adding sum to the mix and adding tests to intelGPU manifest

* Incorporate Chris's first comment + clang

* Some mods to the last commit

* Addressed Bob's comments + added more tests int8

* Added more int8 tests + added tests to IntelGPU manifest

* CI test failure debug attempt

* clang

* edit

* Adding the CPU failing test to manifest

* pulled changes from master to address travis ci build failure

* Move ParameterVector and ResultVector to the ngraph namespace where they belong

* update python wrapper

* more python fixes

* style

* Update setup.py

* fix some new code

* adding i32 as quantized type

* code format

* mask gpu unit tests

* unused variable

* intel gpu unit test manifest

* fix typo in unit test manifest

* add softmax cuda support

* optimize block size

* remove debug info

* remove debug

* style

* remove unused

* remove cudnn softmax

* format

* using nullptr

* move helper, add test

* fix style

* using all_close_f

* using kahansum

* style

* remove commentted out code

* Add ShapeOf op

* Helps to check in the source files

* Add shape_of_scalar to unit test manifests

* Add missing include to gpu_emitter.cpp

* Change 'this op is experimental' wording per @indie's suggestion

* New idea: let's try not mallocing 300 terabytes

* Update interpreter implementation

* Add in-place-slice optimization for CPU backend.

* Modify slice emitter function for in place slice.

* Allow arg node to have multiple outputs for in place slice.

* Remove unused variable.

* Add CPUExecutionContext argument to slice builder.

* Address PR feedback: move computation out of the functor.

* Move size computation out of the functor for in place concat.

* When CUDNN_DATA_TYPE == CUDNN_DATA_DOUBLE, it appears that the cudnn calculated workspace size is incorrect.
Adding a temporary fix here until the underlying issue is found.

* Add softmax test illustrating bug in cudnn impl.

* disable new unit test in intel GPU

* add nop backend

* nop backend

* fix flag name

* add new switch to cmake output of switch settings

* add new unit test to igpu manifest

* remove redundant test

* generate mask

* add codegen+dex

* states in context

* CPU dropout

* remove dead code

* remove dead code

* remove state.cpp

* change perms, add state.cpp

* address bobs feedback

* restore igpu unit-test manifest after a bad merge

* better error msgs

* throw on GPUs to keep a compiler happy

* address more feedback

* fix tests

* fresh commit for the changes

* Working topk on ndims for nvGPU

* fix

* clang

* Added unit test, improved kernel hash and Bob's comment

* int64 test+clang

* Moved argReduce and topk tests to a separate file

* TopK unsupported for IntelGPU

* addressed Fenglei and Chris's comments

* addressed Fenglei and Chris's comments

* add split concat

* fix bug

* fix bug

* fix bug

* add test

* fix test bug

* add comments

* format

* return intead of check processed

* remove .back() since it's not vector anymore.

* format

* change to paramter tests based on Geoff's comments

* types-> type

* change split size to 256

* Add builder for {de}quantize

* Add declaration in header

* Add mkldnn support for {de}quantize

* Add support for {de}quantize with mkldnn

* Add Dex support

* Generalizing some api's and adding a test case for DQ in backend_test.in.cpp

* Unify scale between ngraph and mkldnn

* Check for nullptrs

* PR feedback

* fix unit test failure

* Adding tests for builder and deleting the backend tests

* curly braces

* test rename

* Update ArgReduce to handle i64 indices.

* Formatting.

* Add throw for output types other than int32/64.

* Add output type to hash.

* Add type to throw.

* Interpreter doesn't currently support 64bit output indices for argmin/max and so disabling this test [JIRA:NGRAPH-3183].

* Temp bn update commit.

* Add CUDNNBatchNorm which adds two additional outputs to batchnorm, the batch mean and batch inv variance.
The batch mean is the same as the output mean if the cummulative average factor is 1.0. Add BatchNormCache pass which replaces all BatchNorm ops that are inputs to BatchNormBackprop
with CUDNNBatchNorm which outputs the saved batch statistics directly to the backprop step.

* Updated bn cache pass, removed extra tests, added test checking that provided stats are used in bprop instead of batch stats.
This test was disabled for interpreter as the reference kernel needs to be updated to use provided statistics.

* Formatting.

* Update to new batch norm API.

* CUDNNBatchNorm -> BatchNormTrainingWithStats

* new line

* Preprocess input variance into BN denominator for cudnn (#1885)

* Add explicit cuda kernel to calculate what cuDNN describes as the inverse
variance. In reality, the backward cudnn kernel for BN requires 1.0f / sqrt(variance + eps),
which is the batchnorm denominator for each channel (a numerically stable inverse stddev).

This introduces op annotations for batch norm backprop and updates the cudnn_emitter to support the insertion of this cuda kernel when required.

* Disable second test on INTERPRETER.

* added half_toward_zero; all previous tests passing

* all rounding modes added with unit tests

* fix cpu emitter

* round mode doc

* round out round modes

* doc typo

* using  names for round modes

* use ceil/floor for rounding functions instead of round/nearbyint

* clean up doc

* equidistant

* split bn into bn_inference bn_training

* fix warnings

* Add GPU support for the new BN ops (#1569)

* Add GPU support and change batchnorm_globalstats test to use BNInference.

* Changed test back to using BNTraining for global stats and updated cudnn backend to account for it.

* Fix issues in merge with master.

* Formatting.

* CPU fixes

* remove 5-arg training BN for now

* more fixes

* python batchnorm changes

* fix onnx_import

* fix a call BatchNormInference c-tor

* yet another fix to BatchNormInference c-tor

* AND yet another fix to batchnorm_inference c-tor

* ops.py

* address adam's feedback

* Remove unnecessary parameter/argument.

* remove batch_norm_training_relu_with_global_stats

* remove bn_relu (training)

* comparisons

* move more unit test out of backend_test.in.cpp

* move more tests

* move more tests

* Improved AvgPool unit test coverage. Fixed small bug that was revealed.

* Renamed disabled unit tests to reflect new names.

* Ran clang-format on backend_test.in.cpp to fix format.

* Renamed cpu_results->backend_results in two unit tests.

* Project initialization commit

* Added unit tests for 3D tensors for argmax

* Refactored reduce to be used by argmax argmin. argmax argmin still has some issues. WIP

* [WIP]First working version of ArgMax ArgMin

* added reduce buffer for the cudnn api calls

* added reduce buffer for the cudnn api calls

* Further modifications. Using rvalues to pass enums to build reduce method

* more unit tests added

* Incorporate Fenglei's comments

* Incorporating Chris's first set of comments

* small change to test file

* Resolving clang issue that was causing argmin test to fail

* Incorporate Chris's  comments

* clang format issue

* Add pad with fill operator using the outward-in index pattern.

* Remove static pad and rename build_pad_dynamic -> build_pad. Update maxpool 1d padding.

* Formatting.

* Split build_pad_dynamic into build_pad and build_pad_fill.

* Add test coverage for fixed bug in op::Pad for gpu.

* add a test failed on gpu, pass on cpu

* fixed bug

* get datatype size

* add descript for test

* update comment

* update comments and name

* rename files

* rename runtime TensorView to Tensor

* rename HostTensorView to HostTensor