Request for PaddlePaddle Support on JetPack 5.1.2 (CUDA 11.4)

[Zahrachab]

I'm trying to run PaddlePaddle on a Jetson device with JetPack 5.1.2, which includes CUDA 11.4, cuDNN 8.6, and TensorRT 8.5. I’ve encountered difficulties finding a compatible prebuilt PaddlePaddle wheel or clear instructions to build it from source for this setup.
Could you please advise on any of the following:

1. Are there any official support for JetPack 5.1.2 ?
2. Is there a compatible prebuilt wheel or Docker image available?
3. If not, could you provide guidance or an updated build guide for compiling PaddlePaddle from source under this configuration?

[magic-alt]

1.Official prebuilt GPU wheel for JetPack 5.1.2 (aarch64)? No—build from source.

2.Official Docker you can run on Jetson? No—NGC Paddle images are amd64 (x86 servers).

3.What to do today? Follow a source build as above (CUDA 11.4 / cuDNN 8.6 / TRT 8.5), or use Paddle Inference/Paddle Lite if you only need deployment libraries.