Cuda Error in loadKernel: 3

Same issue as the link. The NVIDIA moderator says that it’s a Jetson issue and call me to raise the issue here.

Anyone can help?

Thanks a lot.

Hi,

Do you try to deserialize a TensorRT engine file created on other platforms?

Please noted that TensorRT engines are not portable.
You will need to serialize it directly on the Xavier.

Thanks.

Hi @AastaLLL
Yes, I created the engine on other platform, but I can load engine by trtexec successful on the platform that have problem. Isn’t it means that I cant use the engine on the platform?

Thanks.

Hi,

You need to serialize it again.

For example, if you create the engine from an ONNX model.
Please copy the ONNX file into Xavier and apply the conversion and serialization again.

Thanks.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.