Cuda Runtime (no kernel image is available for execution on the device)

Autonomous Machines Jetson & Embedded Systems Jetson Xavier NX
1

Problem：
I need to deploy the pytorch model on the board, the pth was converted to onnx successfully. But when onnx was converted to trt some errors about cuda appeared.
NX has JetPack5.1.3/cuda11.4 installed. During the work，model conversion requires tensorrtversion 8.6.1, and this version of tensorrt requires cuda12.0. So I installed cuda12.0/12.2 and tensorrt8.6.1 manually, and i set the environment variables for cuda12.x and trt.

SF Environment：(JetPack5.1.3/cuda11.4/tensorrt8.5.2)
Jetson Xvaier NX ()
cuda：11.4/12.0/12.2
tensorrt：8.6.1

Onnx model is here:
pcdSeg.zip (1.8 MB)

Error:

error2
error21920×1440 369 KB

4

Dear @846443037,
Does this issue noticed with TRT installed with Jetpack?

5

Other error when converting models using TRT installed with Jetpack.

The error:

error3
error31702×1276 241 KB

6

There is no update from you for a period, assuming this is not an issue any more.
Hence we are closing this topic. If need further support, please open a new one.
Thanks

Dear @846443037,
Did you try the polygraphy tool as suggested in error?