* remove tensor offset from tensor read/write calls because it was never used

* fix build errors

* fix build errors

* fix python test errors

* more python fixes

* revert change

* Make old version of read/write deprecated

* fix python read overload

* one more try to fix python binding

* fix python

* yet another try

* why is this so hard

* fix?

* add text to changes.md

* Add create_dynamic_tensor function to Backend signature

* Add skeleton of dynamic wrapper class

* Simple test for construction

* Some progress on dynamic tensor creation

* Compile dynamic wrapper into core; change config string to a factory option

* Add wrapped executable, working on simple example

* Add wrapped dynamic tensor

* Update to match new tensor API

* Update wrapper to clone the graph; basic dyn test now passes

* Update call_with_validate to grok dynamic

* Make the unit tests a little prettier

* Make get_element_type virtual

* Change dynamic tests to be backend-parametric; fix get_element_type check for dyn tensors

* Add dynamic.in.cpp

* CODEOWNERS entry for dynamic_wrapper

* Remove overly-accommodating behavior for create_dynamic_tensor on non-dynamic BEs

* Pass enable_performance_collection to wrapped compile() calls

* More comments in the 'abc' test

* Fix 'create' wrapper in pybind11

* Remove unnecessary WrappedStaticTensor class

* Better names for the classes, and add some docstrings

* Fix pybind for 'create' to return shared_ptr now

* Interpreter working

* cleanup

* cleanup

* add CPU backend

* Update the rest of the backends

* update python API

* update test_case to use new shared_ptr

* Remove parameter, result, and node vector files and class, use vector

* Fix gcc issues

* style

* Remove python references

* update python files

* fix error

* fix build error

* Don't create a separate .so for python onnx importer.

* Pass outer make's -j values to sub make used in python_wheel.

* Add MPI_Bcast to nGraph core

* Add mpi_broadcast op and pass the unit test on CPU backend

* Modify API for MLSL

* Remove unused variable

* Fix the format

* Add the op call in python/pyngraph/ops/regmodule_pyngraph_op.cpp.

* Fix emitter function for DistBroadcast op

* Fix format

* Fix the emitter function for distbroadcast op

* Fix format

* Delete the reference code for distbroadcast op

* Include mlsl.h from ngraph build for genernal external usage

* Change back for build

* Fix alphabetical order issue and line space

* Rename disbroadcast to broadcastdistributed

* Change back continue

* Update the changes

* Fix style

* Delete unused variables

* Change some text format

* Add support for interpreter backend for broadcastdistributed

* Delete some redundant code

* Fix format

* rename

* remove unused file

* More renames

* renaming

* Remove one unused file

* Fix a bug

* add back redundant

* One more commit

* More commits

* Unify distributed_send and distributed_recv functions

* Fix format

* 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

* Extend compile API to take in PassConfig object

* fix override warning

* remove extra semicolon

* cmake fixes to support cases where include_path has multiple directories

* Help pybind find the overloaded compile methods

* Limit compile-api exposed through PyBind (#2530)

* clang-format

* Remove setter for compilation mode to prevent post-init changes. Add compile-mode warning.

* Removed pass_config include

* fix merge

* rename and document the node name methods

* address more references to renamed methods

* fix compile error

* fix build error

* Add a direct-to-Tile op

* Disable dequantize_dynamic_offset

* Add missing Py op defn

* Generic passthrough op; serialization

* Appease Linux builds

* Add gpu handlers

* Disable floor_int32 for now

* 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

* update licenses for 2019

* style

* make GOE extend from util::Op

* fix build breaks

* 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

This reverts commit c5b082c6.

* onnx: return signle nGraph function with multiple outputs
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: remove excessive code
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* Update ngraph python unit tests

* 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

* remove forward op

* fix bbrks

* fix pybind c-tor for max_pool_bprop

* Add new c-tor to MaxPoolBackprop that takes MaxPool as an explicit argument. Add serializer support.

* Add nvgpu support for new backward pooling c-tor, and calculate fprop when it isn't available.

* Add extra layout for 3 arg maxpool backprop.

* Formatting.

* cpu_workspace_insertion to expect 3-arg maxpool bprop

* GPU: add bprop_needs_pooling flag to primitive hash

* Update INTELGPU arguments_check for MaxPoolBackprop and GPU invocation for avg pool.

* 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

* onnx: enable external weights to enable Caffe2 support
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: update ONNX importer interface documentation
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

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

* create unique_ptr backend

* unit test cleanup

* address more code that was recently added

* change from reference to pointer when passing backend to reduce the number of lines changed.

* fix build error

* fix python wrapper

* style

* more specific treatment for unique_ptr

* 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)

Add support for an environment variable `NGRAPH_ONNX_IMPORT_ENABLE` to setup.py
This variable controls whether the Python API is build with or without support for ONNX import functionality built into core nGraph.

* rename files

* rename runtime TensorView to Tensor

* rename HostTensorView to HostTensor

* remove get_tensor method

* remove get_element_count as it is redundant to get_size

* fix a few issues with examples and python wrapper

* rename get_size to get_element_count and add get_size_in_bytes method

* 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.

* Validate/infer types as a virtual function
BroadcastLike, ScalarConstantLike
Set output shape separately from output types
Remove some obsolete tests
Replace *Like ops with fixed type versions

* Review comments

* Remove TensorViewType

* Remove PrimaryTensorView

* Reviewer comments

* Style fix.

* Another try at GPU

* GPU

* Simplify class hierarchy

* More of previous.

* Better errors

* Remove unneeded classes

* Merge

* Update Python bindings

* clang-format

* Remove redundant files

* Remove commented out lines

* clang-format

* Group includes

* Missing tests

* Merge error

* Expose onnx_import C++ interface in Python API.

* Pretty printing.

* Add Computation interface using Ngraph Function objects plus UT.

* Apply code format.

* Remove unnecessary stream open mode.

- Code formatting.

* Fix onnx_import submodule visibility.

- Folder restructurization.

* Fix some small errors.

- Wrong function type annotations.
- Class doc.
- Code formatting.
- Class inheritance from object.

* Use modified Runtime class interface.

* Add model for test_onnx_import.

* Revert back to old API.

- Use of Function object in Computation class.

* Use of previous verions API.

* Small refactoring

* Code cleanup

* use line comments instead of multiline comments for license header

* update more

* update new files

* more header updates

* style

* normalize comments

* address review comments

* [Py] Wrapper for LRN

* Add missing header

* Add default param values, docs and some unit tests

* Fixes

* clang-format

* [Py] Add __repr__ to Strides and CoordDiff

* Apply clang-format

* Repr fix

* Apply clang-format