Aoti CUDA runtime libraries

[larryliu0820]

This pull request adds a CUDA backend to the ExecutorTorch project, enabling Ahead-Of-Time Inductor (AOTI) model execution on CUDA devices. The changes include new build rules, backend implementation, and a test runner for CUDA, as well as minor related improvements. The most important changes are summarized below.

CUDA Backend Implementation:

• Added a new C++ implementation for the CUDA backend, including dynamic loading of AOTI CUDA model containers, tensor management between CPU and GPU, and execution logic. (backends/cuda/runtime/cuda_backend.cpp)
• Introduced a CMake build file for the CUDA backend, which sets up the necessary libraries, dependencies, and installation rules. (backends/cuda/CMakeLists.txt)
• Added a test runner executable for the CUDA backend, allowing end-to-end testing of model loading and execution. (backends/cuda/tests/voxtral_runner.cpp)

Build System Integration:

• Updated the top-level CMakeLists.txt to conditionally build the CUDA backend and its dependencies when EXECUTORCH_BUILD_CUDA is enabled, and to register the backend for use. (CMakeLists.txt)
• Modified the executor runner build logic to include the flat tensor extension library if built, improving modularity. (CMakeLists.txt)

Supporting Changes:

• Added a type alias for AOTITensorHandle in the AOTI model container header to improve code clarity and type safety. (backends/aoti/aoti_model_container.h)
• Minor Python code fixes and deduplication in the CUDA backend Python module, including resolving a merge conflict and removing redundant enum definitions. (backends/cuda/cuda_backend.py) [1] [2]

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/14731

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.