Fixed cpack information, removed some links (#976)
authorIlya Churaev <ilya.churaev@intel.com>
Wed, 17 Jun 2020 14:17:32 +0000 (17:17 +0300)
committerGitHub <noreply@github.com>
Wed, 17 Jun 2020 14:17:32 +0000 (17:17 +0300)
ngraph/python/README.md [deleted file]
ngraph/python/setup.py
ngraph/src/ngraph/CMakeLists.txt
ngraph/src/ngraph/frontend/onnx_import/op/quant_conv.cpp
ngraph/src/ngraph/ngraph.hpp
ngraph/src/ngraph/op/group_conv.cpp

diff --git a/ngraph/python/README.md b/ngraph/python/README.md
deleted file mode 100644 (file)
index 21181a8..0000000
+++ /dev/null
@@ -1,89 +0,0 @@
-## nGraph Compiler stack
-
-[nGraph][ngraph_github] is an open-source graph compiler for Artificial 
-Neural Networks (ANNs). The nGraph Compiler stack provides an inherently 
-efficient graph-based compilation infrastructure designed to be compatible 
-with the many upcoming processors, like the Intel Nervana&trade; Neural Network 
-Processor (Intel&reg; Nervana&trade; NNP), while also unlocking a massive performance 
-boost on any existing hardware targets in your neural network: both GPUs 
-and CPUs. Using its flexible infrastructure, you will find it becomes 
-much easier to create Deep Learning (DL) models that can adhere to the 
-"write once, run anywhere" mantra that enables your AI solutions to easily
-go from concept to production to scale.
-
-Frameworks using nGraph to execute workloads have shown [up to 45X] performance 
-boost compared to native implementations.
-
-### Using the Python API 
-
-nGraph can be used directly with the [Python API][api_python] described here, or 
-with the [C++ API][api_cpp] described in the [core documentation]. Alternatively, 
-its performance benefits can be realized through frontends such as 
-[TensorFlow][frontend_tf], [PaddlePaddle][paddle_paddle] and [ONNX][frontend_onnx].
-You can also create your own custom framework to integrate directly with the 
-[nGraph Ops] for highly-targeted graph execution.
-
-## Installation
-
-nGraph is available as binary wheels you can install from PyPI. nGraph binary 
-wheels are currently tested on Ubuntu 16.04. To build and test on other 
-systems, you may want to try [building][ngraph_building] from sources.
-
-Installing nGraph Python API from PyPI is easy:
-
-    pip install ngraph-core
-
-## Usage example
-
-Using nGraph's Python API to construct a computation graph and execute a 
-computation is simple. The following example shows how to create a minimal 
-`(A + B) * C` computation graph and calculate a result using 3 numpy arrays 
-as input.
-
-
-```python
-import numpy as np
-import ngraph as ng
-
-A = ng.parameter(shape=[2, 2], name='A', dtype=np.float32)
-B = ng.parameter(shape=[2, 2], name='B', dtype=np.float32)
-C = ng.parameter(shape=[2, 2], name='C', dtype=np.float32)
-# >>> print(A)
-# <Parameter: 'A' ([2, 2], float)>
-
-model = (A + B) * C
-# >>> print(model)
-# <Multiply: 'Multiply_14' ([2, 2])>
-
-runtime = ng.runtime(backend_name='CPU')
-# >>> print(runtime)
-# <Runtime: Backend='CPU'>
-
-computation = runtime.computation(model, A, B, C)
-# >>> print(computation)
-# <Computation: Multiply_14(A, B, C)>
-
-value_a = np.array([[1, 2], [3, 4]], dtype=np.float32)
-value_b = np.array([[5, 6], [7, 8]], dtype=np.float32)
-value_c = np.array([[9, 10], [11, 12]], dtype=np.float32)
-
-result = computation(value_a, value_b, value_c)
-# >>> print(result)
-# [[ 54.  80.]
-#  [110. 144.]]
-
-print('Result = ', result)
-```
-
-[up to 45X]: https://ai.intel.com/ngraph-compiler-stack-beta-release/
-[frontend_onnx]: https://pypi.org/project/ngraph-onnx/
-[paddle_paddle]: https://ngraph.nervanasys.com/docs/latest/frameworks/paddle_integ.html 
-[frontend_tf]: https://pypi.org/project/ngraph-tensorflow-bridge/
-[ngraph_github]: https://github.com/NervanaSystems/ngraph "nGraph on GitHub"
-[ngraph_building]: https://github.com/NervanaSystems/ngraph/blob/master/python/BUILDING.md "Building nGraph"
-[api_python]: https://ngraph.nervanasys.com/docs/latest/python_api/ "nGraph's Python API documentation"
-[api_cpp]: https://ngraph.nervanasys.com/docs/latest/backend-support/cpp-api.html
-[core documentation]: https://ngraph.nervanasys.com/docs/latest/core/overview.html
-[nGraph Ops]: http://ngraph.nervanasys.com/docs/latest/ops/index.html
-
-
index f3bf421..07f4e7b 100644 (file)
@@ -375,8 +375,7 @@ setup(
     description="nGraph - Intel's graph compiler and runtime for Neural Networks",
     version=__version__,
     author="Intel Corporation",
-    author_email="intelnervana@intel.com",
-    url="https://github.com/NervanaSystems/ngraph/",
+    url="https://github.com/openvinotoolkit/openvino",
     license="License :: OSI Approved :: Apache Software License",
     long_description=open(os.path.join(PYNGRAPH_ROOT_DIR, "README.md")).read(),
     long_description_content_type="text/markdown",
index 1c2bea3..03c34ef 100644 (file)
@@ -788,12 +788,9 @@ install(FILES ${CMAKE_CURRENT_BINARY_DIR}/version.hpp
 
 set(CPACK_GENERATOR "DEB")
 set(CPACK_DEBIAN_PACKAGE_SHLIBDEPS ON)
-set(CPACK_PACKAGE_DESCRIPTION_SUMMARY "A most excellent graph library")
+set(CPACK_PACKAGE_DESCRIPTION_SUMMARY "nGraph library")
 set(CPACK_PACKAGE_NAME "nGraph")
-set(CPACK_PACKAGE_CONTACT "Robert Kimball")
-# set(CPACK_PACKAGE_VERSION
-set(CPACK_PACKAGE_VENDOR "Intel Nervana")
-# set(CPACK_PACKAGE_DESCRIPTION_FILE "")
+set(CPACK_PACKAGE_VENDOR "Intel")
 
 set(CPACK_PACKAGE_VERSION_MAJOR ${NGRAPH_VERSION_MAJOR})
 set(CPACK_PACKAGE_VERSION_MINOR ${NGRAPH_VERSION_MINOR})
index 23765eb..8e960f7 100644 (file)
@@ -80,8 +80,6 @@ namespace ngraph
                         {
                             // Split one convolution op to N ops where N is the number of groups
                             // and concat results after computation.
-                            // reference:
-                            // https://github.com/NervanaSystems/ngraph-mxnet/blob/fdd692/src/ngraph/ngraph_emitter.cc#L822-L856
                             std::size_t n_data_channels{data->get_shape().at(1)};
                             std::size_t n_filters_channels{filters->get_shape().at(0)};
 
index 021ef63..f2ebc68 100644 (file)
@@ -47,7 +47,7 @@ namespace ngraph
 }
 
 /// \namespace ngraph
-/// \brief The Intel Nervana Graph C++ API.
+/// \brief The Intel nGraph C++ API.
 
 /// \namespace ngraph::descriptor
 /// \brief Descriptors are compile-time representations of objects that will appear at run-time.
index 7aff64f..97bfbfe 100644 (file)
@@ -727,8 +727,6 @@ NodeVector op::v0::GroupConvolution::decompose_op() const
     auto filters_shape = get_input_shape(1);
     // Split one convolution op to N ops where N is the number of groups
     // and concat results after computation.
-    // reference:
-    // https://github.com/NervanaSystems/ngraph-mxnet/blob/fdd692/src/ngraph/ngraph_emitter.cc#L822-L856
     NodeVector convolution_nodes;
 
     // slice data