Error Code 1: CuTensor (Internal cuTensor permutate execute failed) Error Code 1: Cuda Runtime (invalid resource handle)

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin NX
1

Hi,
I am trying to make inference on the images of a ros topic but I have the problem:

[05/07/2025-19:05:58] [TRT] [E] 1: [reformat.cpp::executeCutensor::387] Error Code 1: CuTensor (Internal cuTensor permutate execute failed)
[05/07/2025-19:05:58] [TRT] [E] 1: [checkMacros.cpp::catchCudaError::202] Error Code 1: Cuda Runtime (invalid resource handle)

Environment:

  • TensorRT Version: 8.5.2.1
  • NVIDIA Driver Version: 35.3.1
  • CUDA Version: V11.4.315
  • CUDNN Version: 8.6.0.166
  • L4T Version: 35.3.1
  • Jetpack Version: 5.1.1
  • Operating System: L4T 35.3.1 (Based on linux 20.04)

THE CODE IS
code.txt (5.1 KB)

3

Hi,

As we have several newer software for Orin now, are you able to test it with the TensorRT 10.3?
If the issue continues, please share the ONNX file with us so we can reproduce this in our environment.

Thanks.

4

Hi,

I think I can’t install tensorrt 10.3 since I work with jetpack 5.1 and ubuntu 20.04, maybe I’m wrong.

I attach the onnx file.
model_onnx.zip (38.1 MB)

My problem is when I work with ROS1 Noetic, since if I run as a python file (attached example) I don’t have the problem.
code_for_python.txt (16.0 KB)
code.txt (5.1 KB)

5

Hi,

Do you have dependencies on JetPack 5.1?
For OrinNX, you can upgrade the JetPack version to 6.2 and TensorRT 10.3.

Thanks.