:orphan:

Scikit-learn Gallery
====================

Examples about converting scikit-learn estimators and pipelines to ONNX.



.. raw:: html

    <div class="sphx-glr-thumbnails">

.. thumbnail-parent-div-open

.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.sklearn.to_onnx converts a fitted sklearn.cluster.KMeans into an onnx.ModelProto that can be executed with any ONNX-compatible runtime.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_kmeans_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_kmeans`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Converting a scikit-learn KMeans to ONNX</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.sklearn.to_onnx converts fitted scikit-learn estimators and pipelines into onnx.ModelProto objects that can be executed with any ONNX-compatible runtime.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_pipeline_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_pipeline`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Converting a scikit-learn Pipeline to ONNX</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="spox is a Python library for constructing ONNX graphs that exposes strongly-typed, opset-versioned operator functions rather than the raw onnx.helper API.  Each operator is a Python function (e.g. op.MatMul(A, B)) so the construction code is type-safe, IDE-friendly, and always produces a graph that is valid for the chosen opset.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_with_spox_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_with_spox`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Converting a scikit-learn model to ONNX with spox</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example converts a sksurv.linear_model.IPCRidge survival regression model into ONNX using yobx.sklearn.to_onnx.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_sksurv_ipc_ridge_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_sksurv_ipc_ridge`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Converting sksurv IPCRidge to ONNX</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.sklearn.to_onnx can convert a FunctionTransformer that wraps any numpy function — without writing a custom converter.  The conversion works by tracing: the function is re-executed with lightweight NumpyArray proxies instead of real arrays.  Every numpy operation performed on those proxies is recorded as an ONNX node, so the resulting graph exactly mirrors the Python code.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_function_transformer_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_function_transformer`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Exporting FunctionTransformer with numpy tracing</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="By default yobx.sklearn.to_onnx produces a flat ONNX graph where every operator from every estimator is inlined directly in the graph proto.  This is fine for most use cases, but sometimes you want to keep the high-level structure visible in the model — for example to make the graph easier to inspect, to share weights between identical sub-models, or to target a runtime that supports ONNX local functions natively.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_function_options_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_function_options`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Exporting sklearn estimators as ONNX local functions</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.sklearn.to_onnx accepts a convert_options argument that enables extra outputs on tree and ensemble estimators without changing the core converter interface.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_convert_options_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_convert_options`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Exporting sklearn tree models with convert options</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="A common source of surprise when deploying scikit-learn models with ONNX is numerical discrepancy between scikit-learn and the ONNX runtime output. The culprit is almost always a dtype mismatch: the model was trained with float64 (the NumPy default), but the ONNX graph was exported with float32 inputs and weights.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_pls_float32_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_pls_float32`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Float32 vs Float64: precision loss with PLSRegression</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.sklearn.to_onnx converts a fitted sklearn.neighbors.KNeighborsClassifier or sklearn.neighbors.KNeighborsRegressor into an onnx.ModelProto.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_knn_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_knn`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">KNeighbors: choosing between CDist and standard ONNX</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.sklearn.to_onnx ships with built-in converters for a curated set of scikit-learn estimators (``StandardScaler``, LogisticRegression, DecisionTree, MLP, and Pipeline). For any estimator that is not covered by the built-in registry, you can supply your own converter through the extra_converters keyword argument.">

.. only:: html

  .. image:: /auto_examples_sklearn/images/thumb/sphx_glr_plot_sklearn_with_sklearn_onnx_thumb.png
    :alt:

  :doc:`/auto_examples_sklearn/plot_sklearn_with_sklearn_onnx`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Using sklearn-onnx to convert any scikit-learn estimator</div>
    </div>


.. thumbnail-parent-div-close

.. raw:: html

    </div>


.. toctree::
   :hidden:

   /auto_examples_sklearn/plot_sklearn_kmeans
   /auto_examples_sklearn/plot_sklearn_pipeline
   /auto_examples_sklearn/plot_sklearn_with_spox
   /auto_examples_sklearn/plot_sklearn_sksurv_ipc_ridge
   /auto_examples_sklearn/plot_sklearn_function_transformer
   /auto_examples_sklearn/plot_sklearn_function_options
   /auto_examples_sklearn/plot_sklearn_convert_options
   /auto_examples_sklearn/plot_sklearn_pls_float32
   /auto_examples_sklearn/plot_sklearn_knn
   /auto_examples_sklearn/plot_sklearn_with_sklearn_onnx


.. only:: html

  .. container:: sphx-glr-footer sphx-glr-footer-gallery

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

      :download:`Download all examples in Python source code: auto_examples_sklearn_python.zip </auto_examples_sklearn/auto_examples_sklearn_python.zip>`

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

      :download:`Download all examples in Jupyter notebooks: auto_examples_sklearn_jupyter.zip </auto_examples_sklearn/auto_examples_sklearn_jupyter.zip>`


.. only:: html

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

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