.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "auto_examples/core/plot_onnxruntime_inference.py"
.. LINE NUMBERS ARE GIVEN BELOW.

.. only:: html

    .. note::
        :class: sphx-glr-download-link-note

        :ref:`Go to the end <sphx_glr_download_auto_examples_core_plot_onnxruntime_inference.py>`
        to download the full example code.

.. rst-class:: sphx-glr-example-title

.. _sphx_glr_auto_examples_core_plot_onnxruntime_inference.py:


.. _l-plot-onnxruntime-inference:

Running inference with onnxruntime
===================================

This example shows how to build a small ONNX model and run it with
:mod:`onnxruntime`.  The model is a two-layer MLP created with the
:func:`~yaourt.doc.demo_mlp_model` helper.

.. GENERATED FROM PYTHON SOURCE LINES 11-16

.. code-block:: Python


    import numpy as np
    import onnxruntime
    from yaourt.doc import demo_mlp_model








.. GENERATED FROM PYTHON SOURCE LINES 17-24

1. Build the model
------------------

:func:`demo_mlp_model <yaourt.doc.demo_mlp_model>` returns an
``onnx.ModelProto`` representing a small fully-connected network:
``x (3×10) → MatMul → Add → Relu → MatMul → Add → output (3×1)``.
The input shape is fixed to a batch size of 3.

.. GENERATED FROM PYTHON SOURCE LINES 24-30

.. code-block:: Python


    model = demo_mlp_model("")
    print("Opset:", model.opset_import[0].version)
    print("Inputs :", [i.name for i in model.graph.input])
    print("Outputs:", [o.name for o in model.graph.output])





.. rst-class:: sphx-glr-script-out

 .. code-block:: none

    Opset: 18
    Inputs : ['x']
    Outputs: ['output_0']




.. GENERATED FROM PYTHON SOURCE LINES 31-36

2. Run inference
----------------

We create a random input tensor matching the model's fixed batch size
and call :meth:`~onnxruntime.InferenceSession.run`.

.. GENERATED FROM PYTHON SOURCE LINES 36-46

.. code-block:: Python


    x = np.random.randn(3, 10).astype(np.float32)

    sess = onnxruntime.InferenceSession(model.SerializeToString(), providers=["CPUExecutionProvider"])
    (output,) = sess.run(None, {"x": x})

    print("Input  shape:", x.shape)
    print("Output shape:", output.shape)
    assert output.shape == (3, 1), f"Unexpected output shape: {output.shape}"





.. rst-class:: sphx-glr-script-out

 .. code-block:: none

    Input  shape: (3, 10)
    Output shape: (3, 1)




.. GENERATED FROM PYTHON SOURCE LINES 47-51

3. Run multiple times
---------------------

The same session can be reused for multiple calls with different data.

.. GENERATED FROM PYTHON SOURCE LINES 51-59

.. code-block:: Python


    for i in range(3):
        xi = np.random.randn(3, 10).astype(np.float32)
        (yi,) = sess.run(None, {"x": xi})
        print(f"  run={i}  output shape: {yi.shape}")
        assert yi.shape == (3, 1)

    print("All runs passed.")




.. rst-class:: sphx-glr-script-out

 .. code-block:: none

      run=0  output shape: (3, 1)
      run=1  output shape: (3, 1)
      run=2  output shape: (3, 1)
    All runs passed.





.. rst-class:: sphx-glr-timing

   **Total running time of the script:** (0 minutes 0.248 seconds)


.. _sphx_glr_download_auto_examples_core_plot_onnxruntime_inference.py:

.. only:: html

  .. container:: sphx-glr-footer sphx-glr-footer-example

    .. container:: sphx-glr-download sphx-glr-download-jupyter

      :download:`Download Jupyter notebook: plot_onnxruntime_inference.ipynb <plot_onnxruntime_inference.ipynb>`

    .. container:: sphx-glr-download sphx-glr-download-python

      :download:`Download Python source code: plot_onnxruntime_inference.py <plot_onnxruntime_inference.py>`

    .. container:: sphx-glr-download sphx-glr-download-zip

      :download:`Download zipped: plot_onnxruntime_inference.zip <plot_onnxruntime_inference.zip>`


.. only:: html

 .. rst-class:: sphx-glr-signature

    `Gallery generated by Sphinx-Gallery <https://sphinx-gallery.github.io>`_