:orphan:

Core Gallery
============

Examples about the core ONNX builder, evaluators, graph utilities, and shape expressions.


.. raw:: html

  <div id='sg-tag-list' class='sphx-glr-tag-list'></div>


.. raw:: html

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

.. thumbnail-parent-div-open

.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example builds a small ONNX model, saves it to a temporary file, and then demonstrates the different output formats produced by the python -m yobx print command.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_print_command_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_print_command`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Command Line: python -m yobx print</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="decompose_einsum, decompose_einsum_2inputs, and the native ONNX Einsum operator all represent the same computation but as very different ONNX graphs:">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_einsum_cost_comparison_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_einsum_cost_comparison`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Comparing Computational Cost of Three Einsum→ONNX Strategies</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.translate.translate converts an onnx.ModelProto into Python source code that, when executed, recreates the same model.  Five output APIs are available:">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_translate_comparison_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_translate_comparison`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Comparing the five ONNX translation APIs</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx ships three evaluators that share the same run(outputs, feeds) interface but differ in their backend:">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_evaluator_comparison_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_evaluator_comparison`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Comparing the three evaluators</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example explains how FLOPs (floating-point operations) cost is estimated for ONNX models in yobx, and programmatically lists the formula used for every supported operator.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_cost_formulas_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_cost_formulas`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Computation Cost: How It Works and Supported Operator Formulas</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example shows how yobx.xshape.shape_builder_impl.BasicShapeBuilder tracks symbolic dimension expressions through a sequence of Add, Concat, and Reshape nodes, and compares the result with the standard onnx.shape_inference.infer_shapes and the onnx-shape-inference package from PyPI.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_computed_shapes_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_computed_shapes`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Computed Shapes: Add + Concat + Reshape</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="The ONNX Einsum operator is very expressive but not all runtimes support it.  yobx.helpers.einsum_helper.decompose_einsum converts an einsum equation into a sub-graph built from simpler operators (``Transpose``, Reshape, MatMul, Mul, ReduceSum, …) that every compliant ONNX runtime understands.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_einsum_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_einsum`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Decompose Einsum into Regular ONNX Operators</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="When an ONNX model contains dynamic (unknown) input dimensions, yobx.xshape.shape_builder_impl.BasicShapeBuilder represents every output dimension as either a plain integer or a symbolic string expression built from the names of the input dimensions.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_shape_expressions_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_shape_expressions`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Expressions in Shape Computation</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.container.ExtendedModelContainer extends the standard onnx.model_container.ModelContainer to handle large weight tensors — numpy arrays stored separately from the main .onnx file instead of being serialised inside the protobuf.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_extended_model_container_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_extended_model_container`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">ExtendedModelContainer: large-initializer ONNX models</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.reference.ExtendedReferenceEvaluator extends onnx.reference.ReferenceEvaluator with additional operator kernels for non-standard domains such as com.microsoft.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_extended_reference_evaluator_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_extended_reference_evaluator`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">ExtendedReferenceEvaluator: running models with contrib operators</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.helpers.mini_onnx_builder.MiniOnnxBuilder creates minimal ONNX models whose only purpose is to store tensors as initializers and return them when the model is executed.  The model has no inputs — running it simply replays the stored values.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_mini_onnx_builder_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_mini_onnx_builder`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">MiniOnnxBuilder: serialize tensors to an ONNX model</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.helpers.dot_helper.to_dot converts an onnx.ModelProto into a DOT string that can be rendered by Graphviz.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_dot_graph_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_dot_graph`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">ONNX Graph Visualization with to_dot</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="yobx.helpers.mermaid_helper.to_mermaid converts an onnx.ModelProto into a Mermaid flowchart TD string that can be rendered by any Mermaid-compatible viewer (e.g. GitHub Markdown, the Mermaid live editor, or Sphinx with the sphinxcontrib-mermaid extension).">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_mermaid_graph_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_mermaid_graph`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">ONNX Graph Visualization with to_mermaid</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="This example shows how to compute the symbolic FLOPs cost of an ONNX model using yobx.xshape.shape_builder_impl.BasicShapeBuilder with inference=InferenceMode.COST.">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_symbolic_cost_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_symbolic_cost`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Symbolic Cost of a Model: Attention Block</div>
    </div>


.. raw:: html

    <div class="sphx-glr-thumbcontainer" tooltip="stats_tree_ensemble computes per-tree and per-feature statistics for TreeEnsembleClassifier and TreeEnsembleRegressor ONNX nodes (from the ai.onnx.ml domain).">

.. only:: html

  .. image:: /auto_examples_core/images/thumb/sphx_glr_plot_tree_statistics_thumb.png
    :alt:

  :doc:`/auto_examples_core/plot_tree_statistics`

.. raw:: html

      <div class="sphx-glr-thumbnail-title">Tree-Ensemble Statistics</div>
    </div>


.. thumbnail-parent-div-close

.. raw:: html

    </div>


.. toctree::
   :hidden:

   /auto_examples_core/plot_print_command
   /auto_examples_core/plot_einsum_cost_comparison
   /auto_examples_core/plot_translate_comparison
   /auto_examples_core/plot_evaluator_comparison
   /auto_examples_core/plot_cost_formulas
   /auto_examples_core/plot_computed_shapes
   /auto_examples_core/plot_einsum
   /auto_examples_core/plot_shape_expressions
   /auto_examples_core/plot_extended_model_container
   /auto_examples_core/plot_extended_reference_evaluator
   /auto_examples_core/plot_mini_onnx_builder
   /auto_examples_core/plot_dot_graph
   /auto_examples_core/plot_mermaid_graph
   /auto_examples_core/plot_symbolic_cost
   /auto_examples_core/plot_tree_statistics


.. 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_core_python.zip </auto_examples_core/auto_examples_core_python.zip>`

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

      :download:`Download all examples in Jupyter notebooks: auto_examples_core_jupyter.zip </auto_examples_core/auto_examples_core_jupyter.zip>`


.. only:: html

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

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