Python API for HOST_ACCESSIBLE OrtValue allocation

[ericcraw]

Description

Adds memory_info= parameter to OrtValue.ortvalue_from_shape_and_type(), backed by two new C-level factory methods that look up the registered shared allocator via the full OrtMemoryInfo (including mem_type).

This is required because the current shared allocator query doesn't include the memory type making HOST_ACCESSIBLE invisible to python. UsesCpuMemory() is used in GetPyObjFromTensor so that tensors in HOST_ACCESSIBLE memory are returned as zero-copy numpy views.

Motivation and Context

Enable zero copy interop between numpy and ortvalue.

This is a follow up for #28037

[yuslepukhin]

No Python test exercising the new memory_info= parameter or verifying that HOST_ACCESSIBLE OrtValues produce zero-copy numpy views.

[Copilot]

Pull request overview

Adds Python-level support for allocating OrtValue tensors using an explicit OrtMemoryInfo (including mem_type) so plugin EP HOST_ACCESSIBLE shared allocators can be selected, enabling zero-copy numpy interop for those tensors.

Changes:

• Update tensor-to-numpy conversion to treat HOST_ACCESSIBLE tensors as CPU-memory-compatible via OrtDevice::UsesCpuMemory().
• Add new pybind factory methods to allocate OrtValue from shape/type using a full OrtMemoryInfo lookup.
• Extend OrtValue.ortvalue_from_shape_and_type() Python API with an optional memory_info= parameter to route allocations through those new factories.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
onnxruntime/python/onnxruntime_pybind_state.cc	Enables zero-copy numpy views for HOST_ACCESSIBLE tensors via UsesCpuMemory().
onnxruntime/python/onnxruntime_pybind_ortvalue.cc	Adds OrtMemoryInfo-based OrtValue allocation factories using shared allocator lookup.
onnxruntime/python/onnxruntime_inference_collection.py	Exposes memory_info= on OrtValue.ortvalue_from_shape_and_type() and dispatches to new C++ factories.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[yuslepukhin]

Should we wait for #28037 to merge?

[ericcraw]

I think this can go in parallel to #28037. We need a way to allocate an ort value with a host accessible device allocator + create np views into them. We just may not avoid all copies until #28037 merges.

[ericcraw]

@yuslepukhin now that #28037 is merged, are we ok to merge this?