Can you add GPU SM110 (Jetson Thor) support to the tensorrt:25.06-py3 container?

AI & Data Science NGC Catalog Feedback
, ,
1

Hi,
I am trying to use onnxruntime-gpu on Jetson Thor but it’s not compatible with Cuda 13 yet. So i pulled this container to use Cuda 12.9 with back compability to use onnxruntime-gpu 1.23.9. But realized ;
[TensorRT EP] TensorRT version is 101100
[TensorRT EP] CUDA version is 12090
…
2026-01-22 07:24:52.273885646 [I:onnxruntime:Default, tensorrt_execution_provider.cc:2789 GetCapability] [TensorRT EP] Whole graph will run on TensorRT execution provider

2026-01-22 07:24:52.274201999 [V:onnxruntime:Default, tensorrt_execution_provider.cc:3165 CreateNodeComputeInfoFromGraph] [TensorRT EP] FP16 mode is enabled

2026-01-22 07:24:52.287030771 [E:onnxruntime:Default, tensorrt_execution_provider.h:90 log] [2026-01-22 07:24:52 ERROR] IBuilder::buildSerializedNetwork: Error Code 9: API Usage Error (Target GPU SM 110 is not supported by this TensorRT release.)

It hasn’t compute capabilty to SM-110. Can you please add back compabilty to Jetson Thor.
I also opened an issue to onnxruntime-gpu for Cuda13.

Thank you!

2

You can build onnxruntime-gpu with the script in the below repo:

(Still no onnxruntime_gpu in sbsa/cu130)