When using TensorRT to convert the onnx model to engine, the [TRT] Cuda initialization failure with error
. Detailed error information is as follows
[TensorRT] ERROR: CUDA initialization failure with error 222. Please check your CUDA installation: http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html
The reason for this error is due to the NVIDIA driver version. I used the official TensorRT mirror nvcr.io/nvidia/tensorrt:21.03-py3
of NVIDA. I used Driver Version: 460.84
and Driver Version: 460.80
on the other two servers. Can run.
And the NVIDIA
driver version of this server is 460.56
.
Upgrade the NVIDIA driver version to Driver Version: 460.84
to solve the problem