Tutorial on numpy API for ONNX

_doc/sphinxdoc/source/conf.py

@@ -58,6 +58,7 @@
 epkg_dictionary.update({
     '_PredictScorer': 'https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/metrics/scorer.py#L168',
     'airspeed-velocity': 'https://github.com/airspeed-velocity/asv',
+    'ast': 'https://docs.python.org/3/library/ast.html',
     'asv': 'https://github.com/airspeed-velocity/asv',
     'bench1': 'http://www.xavierdupre.fr/app/mlprodict_bench/helpsphinx/index.html',
     'bench2': 'http://www.xavierdupre.fr/app/mlprodict_bench2/helpsphinx/index.html',
@@ -73,6 +74,7 @@
     'DOT': 'https://en.wikipedia.org/wiki/DOT_(graph_description_language)',
     'einsum': 'https://numpy.org/doc/stable/reference/generated/numpy.einsum.html',
     'exec': 'https://docs.python.org/3/library/functions.html#exec',
+    'FunctionTransformer': 'https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.FunctionTransformer.html',
     'GaussianProcessRegressor': 'https://scikit-learn.org/stable/modules/generated/sklearn.gaussian_process.GaussianProcessRegressor.html',
     'Iris': 'https://en.wikipedia.org/wiki/Iris_flower_data_set',
     'IR_VERSION': 'https://github.com/onnx/onnx/blob/master/docs/IR.md#onnx-versioning',
@@ -105,6 +107,7 @@
     'run_asv.sh': 'https://github.com/sdpython/mlprodict/blob/master/bin/run_asv.sh',
     'Rust': 'https://www.rust-lang.org/',
     'sklearn-onnx': 'https://github.com/onnx/sklearn-onnx',
+    'sklearn-onnx tutorial': 'http://www.xavierdupre.fr/app/onnxcustom/helpsphinx/tutorial.html',
     'tensorflow': 'https://www.tensorflow.org/',
     'Tokenizer': 'https://github.com/microsoft/onnxruntime/blob/master/docs/ContribOperators.md#com.microsoft.Tokenizer',
     'tqdm': 'https://github.com/tqdm/tqdm',

_doc/sphinxdoc/source/index.rst

@@ -95,6 +95,7 @@ does not have any dependency on :epkg:`scikit-learn`.
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.linear_model import LinearRegression
@@ -157,6 +158,7 @@ predictions.
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     from sklearn.linear_model import LogisticRegression
     from sklearn.datasets import load_iris

_doc/sphinxdoc/source/onnx.rst

@@ -31,6 +31,7 @@ of modules below:
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
     :rst:
 
     from mlprodict.onnxrt.validate.validate_helper import modules_list

_doc/sphinxdoc/source/onnx_bench.rst

@@ -46,6 +46,7 @@ of modules below:
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
     :rst:
 
     from mlprodict.onnxrt.validate.validate_helper import modules_list
@@ -80,6 +81,7 @@ describes the list of considered problems.
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
     :rst:
 
     from mlprodict.onnxrt.validate.validate import sklearn_operators, find_suitable_problem

_doc/sphinxdoc/source/onnx_conv.rst

@@ -13,6 +13,7 @@ can be fully converted.
 
 .. runpython::
     :rst:
+    :warningout: DeprecationWarning
     :showcode:
 
     from mlprodict.onnx_conv.register import register_converters

_doc/sphinxdoc/source/onnx_runtime.rst

@@ -21,6 +21,7 @@ Main class is :class:`OnnxInference
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.linear_model import LinearRegression
@@ -94,7 +95,7 @@ the cause of the error if it does not work.
 .. runpython::
     :showcode:
     :rst:
-    :warningout: PendingDeprecationWarning UserWarning RuntimeWarning FutureWarning
+    :warningout: DeprecationWarning PendingDeprecationWarning UserWarning RuntimeWarning FutureWarning
 
     from logging import getLogger
     from pyquickhelper.loghelper import noLOG
@@ -142,7 +143,7 @@ intermediate node anymore.
 
 .. runpython::
     :showcode:
-    :warningout: FutureWarning
+    :warningout: FutureWarning DeprecationWarning
 
     import numpy
     from sklearn.ensemble import AdaBoostRegressor
@@ -171,7 +172,7 @@ We create a table similar to :ref:`l-onnx-pyrun-tbl`.
 .. runpython::
     :showcode:
     :rst:
-    :warningout: PendingDeprecationWarning UserWarning RuntimeWarning
+    :warningout: DeprecationWarning PendingDeprecationWarning UserWarning RuntimeWarning
 
     from logging import getLogger
     from pyquickhelper.loghelper import noLOG
@@ -240,7 +241,7 @@ object to :epkg:`ONNX`. We create a table similar to
 .. runpython::
     :showcode:
     :rst:
-    :warningout: PendingDeprecationWarning UserWarning RuntimeWarning
+    :warningout: DeprecationWarning PendingDeprecationWarning UserWarning RuntimeWarning
 
     from logging import getLogger
     from pyquickhelper.loghelper import noLOG

_doc/sphinxdoc/source/tutorial/index.rst

@@ -9,5 +9,6 @@ one piece this module can do. More should follow.
     :maxdepth: 1
 
     onnx
+    onnx_numpy
     optim
     benchmark

_doc/sphinxdoc/source/tutorial/onnx.rst

@@ -22,6 +22,7 @@ to convert many :epkg:`scikit-learn` models.
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.datasets import load_iris
@@ -48,6 +49,7 @@ begin.
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.datasets import load_iris
@@ -74,6 +76,7 @@ The verbosity can be increased.
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.datasets import load_iris
@@ -103,6 +106,7 @@ can also call :epkg:`onnxruntime` to compute the predictions by using
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.datasets import load_iris
@@ -129,6 +133,7 @@ to look into intermediate results.
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.datasets import load_iris
@@ -157,6 +162,7 @@ As a consequence, interdiate results cannot be seen anymore.
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.datasets import load_iris
@@ -190,6 +196,7 @@ It works exactly the same:
 
 .. runpython::
     :showcode:
+    :warningout: DeprecationWarning
 
     import numpy
     from sklearn.datasets import load_iris
@@ -209,121 +216,3 @@ It works exactly the same:
 
     oinf = OnnxInference(model_def, runtime='python')
     print(oinf.run({'X': X_test[:5]}))
-
-From ONNX to Python
-+++++++++++++++++++
-
-The Python Runtime can be optimized by generating
-custom python code and dynamically compile it.
-:class:`OnnxInference <mlprodict.onnxrt.OnnxInference>`
-computes predictions based on an ONNX graph with a
-python runtime or :epkg:`onnxruntime`.
-Method :meth:`to_python
-<mlprodict.onnxrt.onnx_inference_exports.OnnxInferenceExport.to_python>`
-goes further by converting the ONNX graph into a standalone
-python code.
-
-.. _l-numpy2onnx-tutorial:
-
-From numpy to ONNX
-++++++++++++++++++
-
-.. index:: algebric function
-
-*mlprodict* implements function
-:func:`translate_fct2onnx
-<mlprodict.onnx_grammar.onnx_translation.translate_fct2onnx>`
-which converts the code
-of a function written with :epkg:`numpy` and :epkg:`scipy`
-into an :epkg:`ONNX` graph.
-
-The kernel *ExpSineSquared*
-is used by :epkg:`sklearn:gaussian_process:GaussianProcessRegressor`
-and its conversion is required to convert the model.
-The first step is to write a standalone function which
-relies on :epkg:`scipy` or :epkg:`numpy` and which produces
-the same results. The second step calls this function to
-produces the :epkg:`ONNX` graph.
-
-.. runpython::
-    :showcode:
-    :process:
-    :store_in_file: fct2onnx_expsine.py
-
-    import numpy
-    from scipy.spatial.distance import squareform, pdist
-    from sklearn.gaussian_process.kernels import ExpSineSquared
-    from mlprodict.onnx_grammar import translate_fct2onnx
-    from mlprodict.onnx_grammar.onnx_translation import squareform_pdist, py_make_float_array
-    from mlprodict.onnxrt import OnnxInference
-
-    # The function to convert into ONNX.
-    def kernel_call_ynone(X, length_scale=1.2, periodicity=1.1, pi=3.141592653589793):
-
-        # squareform(pdist(X, ...)) in one function.
-        dists = squareform_pdist(X, metric='euclidean')
-
-        # Function starting with 'py_' --> must not be converted into ONNX.
-        t_pi = py_make_float_array(pi)
-        t_periodicity = py_make_float_array(periodicity)
-
-        # This operator must be converted into ONNX.
-        arg = dists / t_periodicity * t_pi
-        sin_of_arg = numpy.sin(arg)
-
-        t_2 = py_make_float_array(2)
-        t__2 = py_make_float_array(-2)
-
-        t_length_scale = py_make_float_array(length_scale)
-
-        K = numpy.exp((sin_of_arg / t_length_scale) ** t_2 * t__2)
-        return K
-
-    # This function is equivalent to the following kernel.
-    kernel = ExpSineSquared(length_scale=1.2, periodicity=1.1)
-
-    x = numpy.array([[1, 2], [3, 4]], dtype=float)
-
-    # Checks that the new function and the kernel are the same.
-    exp = kernel(x, None)
-    got = kernel_call_ynone(x)
-
-    print("ExpSineSquared:")
-    print(exp)
-    print("numpy function:")
-    print(got)
-
-    # Converts the numpy function into an ONNX function.
-    fct_onnx = translate_fct2onnx(kernel_call_ynone, cpl=True,
-                                  output_names=['Z'])
-
-    # Calls the ONNX function to produce the ONNX algebric function.
-    # See below.
-    onnx_model = fct_onnx('X')
-
-    # Calls the ONNX algebric function to produce the ONNX graph.
-    inputs = {'X': x.astype(numpy.float32)}
-    onnx_g = onnx_model.to_onnx(inputs, target_opset=12)
-
-    # Creates a python runtime associated to the ONNX function.
-    oinf = OnnxInference(onnx_g)
-
-    # Compute the prediction with the python runtime.
-    res = oinf.run(inputs)
-    print("ONNX output:")
-    print(res['Z'])
-
-    # Displays the code of the algebric function.
-    print('-------------')
-    print("Function code:")
-    print('-------------')
-    print(translate_fct2onnx(kernel_call_ynone, output_names=['Z']))
-
-The output of function
-:func:`translate_fct2onnx
-<mlprodict.onnx_grammar.onnx_translation.translate_fct2onnx>`
-is not an :epkg:`ONNX` graph but the code of a function which
-produces an :epkg:`ONNX` graph. That's why the function is called
-twice. The first call compiles the code and a returns a new
-:epkg:`python` function. The second call starts all over but
-returns the code instead of its compiled version.