[Feature Request] Jetson Thor Cuda 13.0 Onnxruntime-gpu

[SnophiTheDeveloper]

Can be added aarch64 gpu version for jetson cuda 13 devices?

Hi,
I am currently using Jetson Thor for my project (CUDA 13). Some of NGC containers are back compatible with Thor but TensorRT-25.06. It's too hard to combine different packages.

I checked https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ort-cuda-13-nightly/PyPI/onnxruntime-gpu/overview/1.24.0.dev20260120001 and i couldn't see aarch64.

[TimotejFasiang]

There is supposed to be one specifically made for jetson thor in https://pypi.jetson-ai-lab.io/sbsa/cu130, but as seen in these issues (https://forums.developer.nvidia.com/t/method-to-compile-onnxruntime-gpu-on-thor/347563/3) (https://forums.developer.nvidia.com/t/still-no-onnxruntime-gpu-in-sbsa-cu130/358085), it's not there.
You can build it manually though: https://github.com/dusty-nv/jetson-containers/tree/master/packages/ml/onnxruntime