Add onnxruntime 1.18.1 for Linux aarch64 GPU

[csukuangfj]

See also #1630

The pre-built onnxruntime libs are provided by the community using the following command:

./build.sh --build_shared_lib --config Release --update \
  --build --parallel --use_cuda \
  --cuda_home /usr/local/cuda \
  --cudnn_home /usr/lib/aarch64-linux-gnu 2>&1 | tee my-log.txt

See also microsoft/onnxruntime#11226

info about the board:

Software part of jetson-stats 4.3.1 - (c) 2024, Raffaello Bonghi
Model: NVIDIA Jetson Orin Nano Engineering Reference Developer Kit Super - Jetpack 6.2 [L4T 36.4.3]
NV Power Mode[1]: 25W
Serial Number: [XXX Show with: jetson_release -s XXX]
Hardware:
 - P-Number: p3767-0003
 - Module: NVIDIA Jetson Orin Nano (8GB ram)
Platform:
 - Distribution: Ubuntu 22.04 Jammy Jellyfish
 - Release: 5.15.148-tegra
jtop:
 - Version: 4.3.1
 - Service: Active
Libraries:
 - CUDA: 12.6.68
 - cuDNN: 9.3.0.75
 - TensorRT: 10.3.0.30
 - VPI: 3.2.4
 - Vulkan: 1.3.204
 - OpenCV: 4.8.0 - with CUDA: NO

Build logs are attached below
build_log.txt

---

You can try the pre-built binaries and libs for v1.10.45 at
https://huggingface.co/csukuangfj/sherpa-onnx-libs/blob/main/aarch64/sherpa-onnx-v1.10.45-linux-aarch64-shared-gpu-onnxruntime-1.18.1.tar.bz2

[drlima]

@csukuangfj, it's my understanding that we only have compatibility with onnxruntime 1.17:

sherpa-onnx/build-android-x86-64.sh

Line 72 in 7740dbf

onnxruntime_version=1.17.1

Do you see any blockers in trying to bump it to the newest versions, like 1.20?

My particular use case is that I wanna use sherpa-onnx in an application that strongly depends on the onnxruntime-genai, and it depends on the ORT version 18, so the current 1.17 version is a blocker for me.

[csukuangfj]

please see #1719 for how to.use a different version of onnxruntime.

not all models in sherpa-onnx.work with onnxruntime> 1.17.1

please test it by yourself.

[drlima]

Thanks! I'll give it a try.