[ONNX] Update images and APIs to onnx_dynamo.rst (#144358)

Update the result image of exporting, and delete the functions/class that belongs to `torch.onnx.dynamo_export`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/144358
Approved by: https://github.com/justinchuby, https://github.com/malfet

docs/Makefile

@@ -19,7 +19,6 @@ figures:
 
 onnx:
 	@$(PYCMD) source/scripts/onnx/build_onnx_torchscript_supported_aten_op_csv_table.py
-	@$(PYCMD) source/scripts/onnx/build_onnx_dynamo_diagnostics_rules_md.py $(SOURCEDIR)/generated/onnx_dynamo_diagnostics_rules
 
 opset:
 	@$(PYCMD) source/scripts/build_opsets.py

docs/source/onnx_dynamo.rst

@@ -23,15 +23,6 @@ bytecode analysis that preserves the dynamic nature of the model instead of usin
 In addition, during the export process, memory usage is significantly reduced compared to the TorchScript-enabled exporter.
 See the :doc:`memory usage documentation <onnx_dynamo_memory_usage>` for more information.
 
-The exporter is designed to be modular and extensible. It is composed of the following components:
-
-  - **ONNX Exporter**: :class:`Exporter` main class that orchestrates the export process.
-  - **ONNX Export Options**: :class:`ExportOptions` has a set of options that control the export process.
-  - **ONNX Registry**: :class:`OnnxRegistry` is the registry of ONNX operators and functions.
-  - **FX Graph Extractor**: :class:`FXGraphExtractor` extracts the FX graph from the PyTorch model.
-  - **Fake Mode**: :class:`ONNXFakeContext` is a context manager that enables fake mode for large scale models.
-  - **ONNX Program**: :class:`ONNXProgram` is the output of the exporter that contains the exported ONNX graph and diagnostics.
-  - **ONNX Diagnostic Options**: :class:`DiagnosticOptions` has a set of options that control the diagnostics emitted by the exporter.
 
 Dependencies
 ------------
@@ -85,6 +76,12 @@ See below a demonstration of exporter API in action with a simple Multilayer Per
 As the code above shows, all you need is to provide :func:`torch.onnx.export` with an instance of the model and its input.
 The exporter will then return an instance of :class:`torch.onnx.ONNXProgram` that contains the exported ONNX graph along with extra information.
 
+``onnx_program.optimize()`` can be called to optimize the ONNX graph with constant folding and elimination of redundant operators. The optimization is done in-place.
+
+.. code-block:: python
+
+  onnx_program.optimize()
+
 The in-memory model available through ``onnx_program.model_proto`` is an ``onnx.ModelProto`` object in compliance with the `ONNX IR spec <https://github.com/onnx/onnx/blob/main/docs/IR.md>`_.
 The ONNX model may then be serialized into a `Protobuf file <https://protobuf.dev/>`_ using the :meth:`torch.onnx.ONNXProgram.save` API.
 
@@ -93,12 +90,15 @@ The ONNX model may then be serialized into a `Protobuf file <https://protobuf.de
   onnx_program.save("mlp.onnx")
 
 Two functions exist to export the model to ONNX based on TorchDynamo engine.
-They slightly differ in the way they produce the :class:`ExportedProgram`.
+They slightly differ in the way they produce the :class:`torch.export.ExportedProgram`.
 :func:`torch.onnx.dynamo_export` was introduced with PyTorch 2.1 and
 :func:`torch.onnx.export` was extended with PyTorch 2.5 to easily switch
 from TorchScript to TorchDynamo. To call the former function,
 the last line of the previous example can be replaced by the following one.
 
+.. note::
+    :func:`torch.onnx.dynamo_export` will be deprecated in the future. Please use :func:`torch.onnx.export` with the parameter ``dynamo=True`` instead.
+
 .. code-block:: python
 
   onnx_program = torch.onnx.dynamo_export(model, tensor_x)
@@ -112,20 +112,6 @@ You can view the exported model using `Netron <https://netron.app/>`__.
     :width: 40%
     :alt: MLP model as viewed using Netron
 
-Note that each layer is represented in a rectangular box with a *f* icon in the top right corner.
-
-.. image:: _static/img/onnx/onnx_dynamo_mlp_model_function_highlight.png
-    :width: 40%
-    :alt: ONNX function highlighted on MLP model
-
-By expanding it, the function body is shown.
-
-.. image:: _static/img/onnx/onnx_dynamo_mlp_model_function_body.png
-    :width: 50%
-    :alt: ONNX function body
-
-The function body is a sequence of ONNX operators or other functions.
-
 When the conversion fails
 -------------------------
 
@@ -133,25 +119,6 @@ Function :func:`torch.onnx.export` should called a second time with
 parameter ``report=True``. A markdown report is generated to help the user
 to resolve the issue.
 
-Function :func:`torch.onnx.dynamo_export` generates a report using 'SARIF' format.
-ONNX diagnostics goes beyond regular logs through the adoption of
-`Static Analysis Results Interchange Format (aka SARIF) <https://docs.oasis-open.org/sarif/sarif/v2.1.0/sarif-v2.1.0.html>`__
-to help users debug and improve their model using a GUI, such as
-Visual Studio Code's `SARIF Viewer <https://marketplace.visualstudio.com/items?itemName=MS-SarifVSCode.sarif-viewer>`_.
-
-The main advantages are:
-
-  - The diagnostics are emitted in machine parseable `Static Analysis Results Interchange Format (SARIF) <https://docs.oasis-open.org/sarif/sarif/v2.1.0/sarif-v2.1.0.html>`__.
-  - A new clearer, structured way to add new and keep track of diagnostic rules.
-  - Serve as foundation for more future improvements consuming the diagnostics.
-
-.. toctree::
-   :maxdepth: 1
-   :caption: ONNX Diagnostic SARIF Rules
-   :glob:
-
-   generated/onnx_dynamo_diagnostics_rules/*
-
 .. toctree::
     :hidden:
 
@@ -162,14 +129,14 @@ API Reference
 
 .. autofunction:: torch.onnx.dynamo_export
 
+.. autoclass:: torch.onnx.ONNXProgram
+    :members:
+
 .. autoclass:: torch.onnx.ExportOptions
     :members:
 
 .. autofunction:: torch.onnx.enable_fake_mode
 
-.. autoclass:: torch.onnx.ONNXProgram
-    :members:
-
 .. autoclass:: torch.onnx.ONNXRuntimeOptions
     :members:
 

docs/source/onnx_torchscript.rst

@@ -2,7 +2,7 @@ TorchScript-based ONNX Exporter
 ===============================
 
 .. note::
-    To export an ONNX model using TorchDynamo instead of TorchScript, see :func:`torch.onnx.dynamo_export`.
+    To export an ONNX model using TorchDynamo instead of TorchScript, please see :doc:`Learn more about the TorchDynamo-based ONNX Exporter <onnx_dynamo>`
 
 .. contents:: :local:
 

docs/source/scripts/onnx/build_onnx_dynamo_diagnostics_rules_md.py

@@ -1,41 +0,0 @@
-import argparse
-import os
-from dataclasses import fields
-
-from torch.onnx._internal import diagnostics
-from torch.onnx._internal.diagnostics import infra
-
-
-def gen_docs(out_dir: str):
-    os.makedirs(out_dir, exist_ok=True)
-    for field in fields(diagnostics.rules):
-        rule = getattr(diagnostics.rules, field.name)
-        if not isinstance(rule, infra.Rule):
-            continue
-        if not rule.id.startswith("FXE"):
-            # Only generate docs for `dynamo_export` rules. Excluding rules for TorchScript
-            # ONNX exporter.
-            continue
-        title = f"{rule.id}:{rule.name}"
-        full_description_markdown = rule.full_description_markdown
-        assert (
-            full_description_markdown is not None
-        ), f"Expected {title} to have a full description in markdown"
-        with open(f"{out_dir}/{title}.md", "w") as f:
-            f.write(f"# {title}\n")
-            f.write(full_description_markdown)
-
-
-def main() -> None:
-    parser = argparse.ArgumentParser(
-        description="Generate ONNX diagnostics rules doc in markdown."
-    )
-    parser.add_argument(
-        "out_dir", metavar="OUT_DIR", help="path to output directory for docs"
-    )
-    args = parser.parse_args()
-    gen_docs(args.out_dir)
-
-
-if __name__ == "__main__":
-    main()