[Py] Update README for PyPI (#2151)

* Update README for PyPI

* Update README for PyPI

* Remove redundant newlines

* Fix links

python/BUILDING.md

+# Building the Python API for nGraph
+
+## Building nGraph Python Wheels
+
+[nGraph's build instructions][ngraph_build] give detailed instructions on building nGraph on different operating systems. Please make sure you specify the options `-DNGRAPH_PYTHON_BUILD_ENABLE=ON` and `-DNGRAPH_ONNX_IMPORT_ENABLE=ON` when building nGraph. Use the `make python_wheel` command to build nGraph and create Python packages. 
+
+Basic build procedure on an Ubuntu system: 
+    
+    # apt-get install build-essential cmake clang-3.9 clang-format-3.9 git curl zlib1g zlib1g-dev libtinfo-dev unzip autoconf automake libtool
+    # apt-get install python3 python3-dev python python-dev python-virtualenv
+    
+    $ git clone https://github.com/NervanaSystems/ngraph.git
+    $ cd ngraph/
+    $ mkdir build
+    $ cd build/
+    $ cmake ../ -DNGRAPH_PYTHON_BUILD_ENABLE=ON -DNGRAPH_ONNX_IMPORT_ENABLE=ON -DNGRAPH_USE_PREBUILT_LLVM=ON
+    $ make python_wheel
+
+After this procedure completes, the `ngraph/build/python/dist` directory should contain Python packages. 
+
+    $ ls python/dist/
+    ngraph-core-0.10.0.tar.gz  
+    ngraph_core-0.10.0-cp27-cp27mu-linux_x86_64.whl  
+    ngraph_core-0.10.0-cp35-cp35m-linux_x86_64.whl
+
+### Using a virtualenv (optional)
+
+You may wish to use a virutualenv for your installation.
+
+    $ virtualenv -p $(which python3) venv
+    $ source venv/bin/activate
+    (venv) $
+
+### Installing the wheel
+
+You may wish to use a virutualenv for your installation.
+
+    (venv) $ pip install ngraph/build/python/dist/ngraph_core-0.10.0-cp35-cp35m-linux_x86_64.whl
+
+## Running tests
+
+Unit tests require additional packages be installed:
+
+    (venv) $ cd ngraph/python
+    (venv) $ pip install -r test_requirements.txt
+
+Then run tests:
+
+    (venv) $ pytest test/ngraph/
+
+[ngraph_build]: http://ngraph.nervanasys.com/docs/latest/buildlb.html

python/README.md

-# Python binding for the nGraph Library
+## nGraph Neural Network compiler
 
-## Build the nGraph Library (required)
+[nGraph][ngraph_github] is Intel's open-source graph compiler and runtime for Neural Network models. Frameworks using nGraph to execute workloads have shown up to [45X](https://ai.intel.com/ngraph-compiler-stack-beta-release/) performance boost compared to native implementations.
 
-Follow the [build instructions] to build nGraph. When you get to the `cmake` 
-command, be sure to specify the option that enables ONNX support in the Library: 
+nGraph can be used directly thought it's [Python API][api_python] or [C++ API][api_cpp]. Alternatively it can be used through one of its frontends, e.g. [TensorFlow][frontend_tf], [MXNet][frontend_mxnet] and [ONNX][frontend_onnx].
 
-    $ cmake ../ -DNGRAPH_ONNX_IMPORT_ENABLE=ON 
+## Installation
 
+nGraph is available as binary wheels you can install from PyPI. nGraph binary wheels are currently tested on Ubuntu 16.04 and require a CPU with AVX-512 instructions, if you're using a different system, you may want to [build][ngraph_building] nGraph from sources.
 
-Next, clone the `pybind11` repository:
+Installing nGraph Python API from PyPI is easy:
 
-    $ cd ngraph/python
-    $ git clone --recursive https://github.com/pybind/pybind11.git
+    pip install ngraph-core
 
+## Usage example
 
-Set the environment variables:
+Using nGraph's Python API to construct a computation graph and execute a computation is simple. The following example shows how to create a simple `(A + B) * C` computation graph and calculate a result using 3 numpy arrays as input.
 
-    export NGRAPH_CPP_BUILD_PATH=$HOME/ngraph_dist
-    export LD_LIBRARY_PATH=$HOME/ngraph_dist/lib
-    export DYLD_LIBRARY_PATH=$HOME/ngraph_dist/lib # (Only needed on MacOS)
-    export PYBIND_HEADERS_PATH=pybind11
+```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)>
 
-Install the wrapper (Python binding):
+model = (A + B) * C
+# >>> print(model)
+# <Multiply: 'Multiply_14' ([2, 2])>
 
-    $ python setup.py install
+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)>
 
-Unit tests require additional packages be installed:
+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)
 
-    $ pip install -r test_requirements.txt
+result = computation(value_a, value_b, value_c)
+# >>> print(result)
+# [[ 54.  80.]
+#  [110. 144.]]
 
+print('Result = ', result)
+```
 
-Then run a test:
-
-    $ pytest test/test_ops.py
-    $ pytest test/ngraph/
-
-
-## Running tests with tox
-
-[Tox] is a Python [virtualenv] management and test command line tool. In our 
-project it automates:
-
-* running of unit tests with [pytest]
-* checking that code style is compliant with [PEP8] using [Flake8]
-* static type checking using [MyPy]
-* testing across Python 2 and 3
-
-
-Installing and running test with Tox:
-
-    $ pip install tox
-    $ tox
-
-
-You can run tests using only Python 3 or 2 using the `-e` (environment) switch:
-
-    $ tox -e py36
-    $ tox -e py27
-
-
-You can check styles in a particular code directory by specifying the path:
-
-    $ tox ngraph/
-
-
-If you run into any problems, try recreating the virtual environments by 
-deleting the `.tox` directory:
-
-    $ rm -rf .tox
-    $ tox
-
-[build instructions]:http://ngraph.nervanasys.com/docs/latest/buildlb.html
-[Tox]:https://tox.readthedocs.io/
-[virtualenv]:https://virtualenv.pypa.io/
-[pytest]:https://docs.pytest.org/
-[PEP8]:https://www.python.org/dev/peps/pep-0008
-[Flake8]:http://flake8.pycqa.org
-[MyPy]:http://mypy.readthedocs.io
\ No newline at end of file
+[frontend_onnx]: https://pypi.org/project/ngraph-onnx/
+[frontend_mxnet]: https://pypi.org/project/ngraph-mxnet/ 
+[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/howto/ 

python/build_wheel.sh.in

@@ -49,5 +49,6 @@ if [ "${PYTHON3_DETECTED}" == 1  ]; then
     source build/venv3/bin/activate
     pip install --upgrade setuptools pip wheel
     python ${SOURCE_DIR}/setup.py bdist_wheel
+    python ${SOURCE_DIR}/setup.py sdist
     deactivate
 fi

python/setup.py

@@ -293,7 +293,7 @@ ext_modules = [
     ),
 ]
 
-if NGRAPH_ONNX_IMPORT_ENABLE == 'TRUE':
+if NGRAPH_ONNX_IMPORT_ENABLE in ['TRUE', 'ON', True]:
     onnx_sources = [
         'pyngraph/pyngraph_onnx_import.cpp',
         'pyngraph/onnx_import/onnx_import.cpp',
@@ -370,13 +370,14 @@ with open(os.path.join(PYNGRAPH_ROOT_DIR, 'requirements.txt')) as req:
 
 setup(
     name='ngraph-core',
-    description=open(os.path.join(PYNGRAPH_ROOT_DIR, 'README.md')).read(),
+    description='nGraph - Intel\'s graph compiler and runtime for Neural Networks',
     version=__version__,
     author='Intel',
     author_email='intelnervana@intel.com',
-    url='https://ai.intel.com/',
+    url='https://github.com/NervanaSystems/ngraph/',
     license='License :: OSI Approved :: Apache Software License',
-    long_description='',
+    long_description=open(os.path.join(PYNGRAPH_ROOT_DIR, 'README.md')).read(),
+    long_description_content_type='text/markdown',
     ext_modules=ext_modules,
     package_dir=package_dir,
     packages=packages,

python/test/ngraph/test_onnx_import.py

@@ -17,20 +17,25 @@
 import os
 import numpy as np
 
-from ngraph.impl.onnx_import import load_onnx_model_file
 from test.ngraph.util import get_runtime
 
+try:
+    from ngraph.impl.onnx_import import load_onnx_model_file
 
-def test_import_onnx_function():
-    model_path = os.path.join(os.path.dirname(__file__), 'models/add_abc.onnx')
-    ng_function = load_onnx_model_file(model_path)[0]
+    def test_import_onnx_function():
+        model_path = os.path.join(os.path.dirname(__file__), 'models/add_abc.onnx')
+        ng_function = load_onnx_model_file(model_path)[0]
 
-    dtype = np.float32
-    value_a = np.array([1.0], dtype=dtype)
-    value_b = np.array([2.0], dtype=dtype)
-    value_c = np.array([3.0], dtype=dtype)
+        dtype = np.float32
+        value_a = np.array([1.0], dtype=dtype)
+        value_b = np.array([2.0], dtype=dtype)
+        value_c = np.array([3.0], dtype=dtype)
 
-    runtime = get_runtime()
-    computation = runtime.computation(ng_function)
-    result = computation(value_a, value_b, value_c)
-    assert np.allclose(result, np.array([6], dtype=dtype))
+        runtime = get_runtime()
+        computation = runtime.computation(ng_function)
+        result = computation(value_a, value_b, value_c)
+        assert np.allclose(result, np.array([6], dtype=dtype))
+
+except ImportError:
+    # Do not test load_onnx_model_file if nGraph was build without ONNX support
+    pass