Retain Python async run resources

[Akshay Sonawane (apsonawane)]

This pull request updates the ONNX Runtime Python bindings to improve memory management and API flexibility for asynchronous inference sessions. The main changes ensure that Python objects used as inputs to run_async are properly referenced and not prematurely garbage collected, and that the API is more Pythonic by accepting py::object for session and run options. It also adds a test to verify that input arrays are kept alive during asynchronous execution.

Improvements to memory management and API flexibility:

• The AsyncResource struct now stores references to Python feed objects, the session, and run options to ensure they are kept alive for the duration of async execution (onnxruntime/python/onnxruntime_pybind_state.cc).
• The run_async binding now accepts py::object for the session and run options, and internally casts them as needed. This makes the API more Pythonic and flexible (onnxruntime/python/onnxruntime_pybind_state.cc).
• When preparing feeds for run_async, the code now stores each input object in feed_objects before creating the corresponding OrtValue, ensuring the Python objects are not garbage collected too early (onnxruntime/python/onnxruntime_pybind_state.cc).

Testing and validation:

• A test using weakref was added to verify that input arrays passed to run_async remain alive until the callback is invoked, preventing premature garbage collection (onnxruntime/test/python/onnxruntime_test_python.py) [1] [2] [3].
• The test also exercises explicit deletion and garbage collection of inputs, session, and run options to ensure robustness (onnxruntime/test/python/onnxruntime_test_python.py).

[Copilot]

Pull request overview

This PR updates the ONNX Runtime Python bindings’ InferenceSession.run_async path to retain the Python-owned resources (session, run options, and feed objects) for the full duration of asynchronous execution, preventing premature garbage collection of inputs used by run_async. It also adds a Python test that explicitly deletes references and forces GC to validate the lifetime behavior.

Changes:

• Extend AsyncResource to hold strong references to feed Python objects, the session, and run options for the duration of RunAsync.
• Update the run_async binding to accept py::object for session and run options, casting internally and preserving the objects.
• Add a weakref-based regression test that deletes input/session/run options, triggers GC, and asserts the input is still alive in the callback.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
onnxruntime/python/onnxruntime_pybind_state.cc	Retains Python objects (feeds/session/run_options) inside AsyncResource to prevent GC during async execution and updates run_async binding signature accordingly.
onnxruntime/test/python/onnxruntime_test_python.py	Adds a weakref + forced GC test to verify run_async keeps input arrays alive until callback completion.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.