the libcudnn8 version is not correct, when use the version who is not the latest one.

[l1351868270]

Description

The libcudnn8 should libcudnn8_8.9.1.23-1+cuda11.8_arm64, but, it is libcudnn8_8.9.1.23-1+cuda12.1_arm64.
The libcudnn8-dev should libcudnn8-dev_8.9.1.23-1+cuda11.8_arm64, but, it is libcudnn8-dev_8.9.1.23-1+cuda12.1_arm64.
So, when use the command trtexec --onnx=resnet50_onnx_model.onnx --saveEngine=resnet_engine_onnx.trt --explicitBatch, it will core dump.
I download the correct version: https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu2004/sbsa/libcudnn8_8.9.1.23-1+cuda11.8_arm64.deb and https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu2004/sbsa/libcudnn8-dev_8.9.1.23-1+cuda11.8_arm64.deb.
And down the version dpkg -i libcudnn8_8.9.1.23-1+cuda11.8_arm64.deb and dpkg -i libcudnn8-dev_8.9.1.23-1+cuda11.8_arm64.deb .
Then, it is ok for the command.

Environment

TensorRT Version:
8.5
I think other version who is not the latest one also has this problem.

NVIDIA GPU:A40

NVIDIA Driver Version:520.61.05

CUDA Version:11.8.0

CUDNN Version:
cudnn8
But, it is wrong

Operating System:ubuntu20.04, aarch64

Python Version (if applicable):3.8.10

Tensorflow Version (if applicable):None

PyTorch Version (if applicable):None

Baremetal or Container (if so, version):Container, 24.0.1

Relevant Files

Model link:https://github.com/NVIDIA/TensorRT/blob/release/8.4/quickstart/IntroNotebooks/4.%20Using%20PyTorch%20through%20ONNX.ipynb

Steps To Reproduce

Commands or scripts:
./docker/build.sh --file docker/ubuntu-20.04-aarch64.Dockerfile --tag tensorrt-aarch64-ubuntu20.04-cuda11.8 --cuda 11.8.0

trtexec --onnx=resnet50_onnx_model.onnx --saveEngine=resnet_engine_onnx.trt --explicitBatch
Have you tried the latest release?:

Can this model run on other frameworks? For example run ONNX model with ONNXRuntime (polygraphy run <model.onnx> --onnxrt):

[zerollzeng]

@rajeevsrao ^^

[ttyio]

We removed the cudnn dependency in the latest 10.0 release, I will close this, thanks all!