Fix ORTModule python doc generation (#7704)

* Fix ORTModule python doc generation

* Address comment

docs/python/training/conf.py

@@ -28,4 +28,8 @@ html_static_path = ['_static']
 
 # -- Options for intersphinx extension ---------------------------------------
 
-intersphinx_mapping = {'https://docs.python.org/': None}
+intersphinx_mapping = {
+    'python': ('https://docs.python.org/3', None),
+    'numpy': ('https://numpy.org/doc/stable', None),
+    'torch': ('https://pytorch.org/docs/stable/', None),
+}

docs/python/training/content.rst

@@ -47,8 +47,8 @@ in the simple case where the entire model can be offloaded to ONNX Runtime:
 API
 ===
 
-.. automodule:: onnxruntime.training.ortmodule
+.. automodule:: onnxruntime.training.ortmodule.ortmodule
     :members:
     :show-inheritance:
     :member-order: bysource
-..    :inherited-members:
+    :inherited-members:

orttraining/orttraining/python/training/ortmodule/ortmodule.py

@@ -16,28 +16,10 @@ from typing import Iterator, Optional, Tuple, TypeVar
 T = TypeVar('T', bound='Module')
 
 class ORTModule(torch.nn.Module):
-    """Specializes a user torch.nn.Module to leverage ONNX Runtime graph execution.
+    """Extends user's :class:`torch.nn.Module` model to leverage ONNX Runtime super fast training engine.
 
-    ORTModule specializes the user's torch.nn.Module and provides forward, backward
-    implementations be leveraging ONNX Runtime.
-
-    ORTModule interacts with:
-    - GraphExecutionManagerFactory: Which returns a GraphExecutionManager based on
-    whether or not the user's torch module is in training mode or eval mode.
-    - GraphExecutionManager: Responsible for building and executing the forward and backward graphs.
-        - InferenceManager(GraphExecutionManager): Responsible for building, optimizing
-        and executing the inference onnx graph.
-        - TrainingManager(GraphExecutionManager): Responsible for building, optimizing
-        and executing the training onnx graph.
-
-        The GraphExecutionManager first exports the user model into an onnx model.
-        Following that, GraphExecutionManager interacts with OrtModuleGraphBuilder to optimize the onnx graph.
-        Once the onnx graph has been optimized, an ExecutionAgent is instantiated that
-        facilitates in executing the forward and backward subgraphs of the onnx model.
-
-    - _ortmodule_io: Provides utilities to transform the user inputs and outputs of the model.
-        - It facilitates in flattening the output from the user's PyTorch model (since exporting
-        of nested structures is not supported at the moment)
+    ORTModule specializes the user's :class:`torch.nn.Module` model, providing :meth:`~torch.nn.Module.forward`,
+    :meth:`~torch.nn.Module.backward` along with all others :class:`torch.nn.Module`'s APIs.
     """
 
     def __init__(self, module):
@@ -76,6 +58,13 @@ class ORTModule(torch.nn.Module):
 
         self._execution_manager = GraphExecutionManagerFactory(self._flattened_module)
 
+    # IMPORTANT: DO NOT add code here
+    # This declaration is for automatic document generation purposes only
+    # The actual forward implementation is bound during ORTModule initialization
+    def forward(self, *inputs, **kwargs):
+        '''Dummy documentation for forward method'''
+        ...
+
     def _is_training(self):
         return self._flattened_module.training and torch.is_grad_enabled()