* registers nnp tester

* check output size

* rename to nnp tester

* revert changes in graph_util

* update CentOS requs

* doc link imports title of section so remove redundant word

* CentOS getting its own section for install

* don't need both http and ssh clone for centos

* don't need to be root to install ngraph

* Integrate PR comments

* Update section on CentOS install for nGraph with PR feedback

* CMake: Generate code for the host CPU

* CMake: Separate machine-specific flag

* Increase tolerance for convolution tests

* test graph.

* rnn matrix fusion wip.

* WIP.

* matrix fused.

* clean up.

* additional refactor.

* fixed merge errors.

* testing.

* added test case.

* more testing.

* more testing.

* bias wip.

* debuggging.

* fusing bias too.

* disabled debug outputs.

* removed debugging.

* removed testing models.

* removed debug models.

* reset nbench.

* clean up and testing.

* removed debug code.

* updated comments and var names.

* rename var.

* removed debug code.

* removed debug code.

* fixed compiler warnings.

* refactored test.

* Added const to parameter.

* address some of the review comments.

* added comments and addressed reviews.

updating ngraph theme with IntelNeoSans font on headings, better OFL font for doc readability, better rendering of certain text, etc (#727)

* Normalize op implementations

* update custom backend ops

* Add test for ng.absolute

- remove unnecesary repetitive call to as_node

* Add Acos operation with UT.

- update docsrings
- refactor and parameterize unit tests

* Review refactoring fix.

* Review refactoring fix part 2.

* Add Asin and Atan nGraph operation wrappers.

* [Py] Add UT for avg_pooling_3d(Cover uncontigous data input)

Issue #725

* make sure we don't get insidious perf bugs due to doule op::Results
* fix graph_partition tests

* make matmulbias callback aware that addition is commutative

* make sure deserializer doesn't add op::Result twice

* Added new ctor for bn which supports Inference
- added mkldnn emitter code for bn inference
* Added test case for bn inference
- added support for layout propogation for bn inference
* added sanity checks for gamma, beta, mean, variance shape in bn
* added serializer support for bn inference

* general dot for gpu

* Current cudnn implementations use only
a single dimension for the ngraph tensor data (width).
In this case the tensor format should be set to

CUDNN_TENSOR_NCHW

so that adjacent memory accesses are coalesced (stride=1 for width).

* * Added some kernel emitter helpers that are reused often.
* Renamed EmitElementwise -> emit_elementwise to match emit<T>.
* op::Sum now handles trivial case of dim(input_tensor) = dim(output_tensor)
  by performing a memcpy as no axes are reduced.

*   Added general case for Nd descriptors which is used when the tensor
  has more than 4 dimensions. Currently a naive reduce is performed,
  in the future a coordinate transformation could be performed to
  improve the memory layout for the reduction.

* Switched to codegen::CodeWriter::block_begin/end.
It appears that CodeWriter::block_begin/end is not frequently used for emitters (in cpu and gpu transformers)
because a block comment is often desired. To this end I added prefix/suffix default parameters to CodeWriter::block_begin/end
so that this functionality is captured.

* Added backprop op for relu and enabled tests.

* Remove examples from pybind wrapper
* Adding back example (basic.py)

* update nnp transformer path

Adjust CallFrame argument order to match Function

* rename directories to be consistent
* rename reference namespace to match directory

* Remove unnecessary Optional type annotations.

* Implement PR review changes

* cleanup neon integration documentation