Tensorrt 8.0 installation

The following packages have unmet dependencies:
tensorrt : Depends: libnvinfer8 (= 8.0.0-1+cuda10.2) but 8.0.0-1+cuda11.3 is to be installed
Depends: libnvinfer-plugin8 (= 8.0.0-1+cuda10.2) but 8.0.0-1+cuda11.3 is to be installed
Depends: libnvparsers8 (= 8.0.0-1+cuda10.2) but 8.0.0-1+cuda11.3 is to be installed
Depends: libnvonnxparsers8 (= 8.0.0-1+cuda10.2) but 8.0.0-1+cuda11.3 is to be installed
Depends: libnvinfer-bin (= 8.0.0-1+cuda10.2) but it is not going to be installed
Depends: libnvinfer-dev (= 8.0.0-1+cuda10.2) but 8.0.0-1+cuda11.3 is to be installed
Depends: libnvinfer-plugin-dev (= 8.0.0-1+cuda10.2) but 8.0.0-1+cuda11.3 is to be installed
Depends: libnvparsers-dev (= 8.0.0-1+cuda10.2) but 8.0.0-1+cuda11.3 is to be installed
Depends: libnvonnxparsers-dev (= 8.0.0-1+cuda10.2) but 8.0.0-1+cuda11.3 is to be installed
Depends: libnvinfer-samples (= 8.0.0-1+cuda10.2) but it is not going to be installed
Depends: libnvinfer-doc (= 8.0.0-1+cuda10.2) but it is not going to be installed

Using, Ubuntu 18.04 and cuda 10.2

Hi,
Please refer to the installation steps from the below link if in case you are missing on anything
https://docs.nvidia.com/deeplearning/tensorrt/install-guide/index.html
Also, we suggest you to use TRT NGC containers to avoid any system dependency related issues.
https://ngc.nvidia.com/catalog/containers/nvidia:tensorrt

Thanks!

Hi @fjoseph,

Please make sure you downloaded correct deb package suitable for your CUDA version. or you can upgrade CUDA version to 11.3 and install 11.3 supported TensorRT package.
https://developer.nvidia.com/nvidia-tensorrt-8x-download

Thank you.

Thanks for the suggestion. I did download the 10.2 version which makes me wonder why I am getting the above error message.
I will try upgrading to 11.3.

@fjoseph,

Are you still facing this issue?

I found that tensorrt 8 didn’t offer the solution I needed and hence didn’t pursue upgrading to tensorrt 8.

Hi @fjoseph , I hit the same problem with ubuntu16.04 and cuda10.2. I found a possible solution that we can install tensorrt and its dependencies one by one manually. The following commands work well on my machine and hope they are helpful to you.

dpkg -i nv-tensorrt-repo-ubuntu1604-cuda10.2-trt8.0.1.6-ga-20210626_1-1_amd64.deb
cd /var/nv-tensorrt-repo-ubuntu1604-cuda10.2-trt8.0.1.6-ga-20210626
dpkg -i libcudnn8_8.2.1.32-1+cuda10.2_amd64.deb
dpkg -i libcudnn8-dev_8.2.1.32-1+cuda10.2_amd64.deb
dpkg -i libnvinfer8_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvinfer-dev_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvinfer-plugin8_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvinfer-plugin-dev_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvonnxparsers8_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvonnxparsers-dev_8.0.1-1+cuda10.2_amd64.deb
dpkg -i onnx-graphsurgeon_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvparsers8_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvparsers-dev_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvinfer-bin_8.0.1-1+cuda10.2_amd64.deb
dpkg -i python3-libnvinfer_8.0.1-1+cuda10.2_amd64.deb
dpkg -i python3-libnvinfer-dev_8.0.1-1+cuda10.2_amd64.deb
dpkg -i libnvinfer-samples_8.0.1-1+cuda10.2_all.deb
dpkg -i libnvinfer-doc_8.0.1-1+cuda10.2_all.deb
dpkg -i tensorrt_8.0.1.6-1+cuda10.2_amd64.deb
8 Likes

Thanks I shall try this if I am going down this path again.

I am facing a similar issue while tring to build TensorRT OSS latest release (8.0.1). An error pops when running make -j.
These are my systems variables:

TensorRT Version : 8.0.1.6
GPU Type : Nvidia RTX 2060
Nvidia Driver Version : 470.42.01
CUDA Version : 11.3
CUDNN Version : 8.2.1
Operating System + Version : Ubuntu 20.04 LTS
Python Version (if applicable) : 3.8.10
TensorFlow Version (if applicable) : 2.5
Cmake version: 3.17

Any ideas about what might be happening?

2 Likes

where did you download all of deb files ?

run command dpkg -i nv-tensorrt-repo-ubuntu1604-cuda10.2-trt8.0.1.6-ga-20210626_1-1_amd64.deb and you will find all deb files in dir /var/nv-tensorrt-repo-ubuntu1604-cuda10.2-trt8.0.1.6-ga-20210626

1 Like

thanks @shuo-ouyang

it work for me with ubuntu18.04 and cuda10.2
thanks!