Intalling TensorRT: Depends: libnvinfer8 (= 8.2.2-1+cuda11.4) but it is not going to be installed

[niclastrelle]

I do have a docker container with the following environment:

• OS: Linux Ubuntu 18.04
• Base Image: pytorch/pytorch:1.9.1-cuda11.1-cudnn8-runtime
• GPU: Tesla K80
• Cuda Version: 11.1
• PyTorch Version: 1.9.1
• downloaded .deb: nv-tensorrt-repo-ubuntu1804-cuda11.4-trt8.2.2.1-ga-20211214_1-1_amd64.deb

information from the downloaded .deb:
TensorRT 8.2 GA Update 1 for Ubuntu 18.04 and CUDA 11.0, 11.1, 11.2, 11.3, 11.4 and 11.5 DEB local repo Package

As I followed the installing instructions on the Nvidia Webside an error ocurred:

Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
 python3-libnvinfer : Depends: libnvinfer8 (= 8.2.2-1+cuda11.4) but it is not going to be installed
                      Depends: libnvinfer-plugin8 (= 8.2.2-1+cuda11.4) but it is not going to be installed
                      Depends: libnvparsers8 (= 8.2.2-1+cuda11.4) but it is not going to be installed
                      Depends: libnvonnxparsers8 (= 8.2.2-1+cuda11.4) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.

[kct22aws]

I am running into same issue as well. Any fix or suggestions?

[mcdonasd1212]

Same issue

[fevz10]

Same issue...

[laszloturanyi]

Install python3-libnvinfer-dev first, then it should work.

sudo apt-get install python3-libnvinfer-dev

Order in Installation guide is mixed up

[niclastrelle]

@laszloturanyi I tried that, but it is not able to locate the package for some odd reason.

apt-get install python3-libnvinfer-dev

into:

E: Unable to locate package python3-libnvinfer-dev

This problem occurs for any other libnv-package too.

[niclastrelle]

I now used the given container from nvidia as workaround:

https://github.com/NVIDIA/TensorRT/tree/main/docker

[montmejat]

I am also having the same issues... I have tried this (after downloading and installing the .deb package):

$ sudo apt-get install tensorrt
Reading package lists... Done
Building dependency tree       
Reading state information... Done
Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
 tensorrt : Depends: libnvinfer8 (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvinfer-plugin8 (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvparsers8 (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvonnxparsers8 (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvinfer-bin (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvinfer-dev (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvinfer-plugin-dev (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvparsers-dev (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvonnxparsers-dev (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvinfer-samples (= 8.2.3-1+cuda11.4) but it is not going to be installed
            Depends: libnvinfer-doc (= 8.2.3-1+cuda11.4) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.

A quick search gives me this result:

$ apt-cache search tensorrt
libnvinfer-plugin-dev - TensorRT plugin libraries
libnvinfer-bin - TensorRT binaries
tensorrt - Meta package of TensorRT
libnvinfer-doc - TensorRT documentation
graphsurgeon-tf - GraphSurgeon for TensorRT package
libnvinfer-dev - TensorRT development libraries and headers
onnx-graphsurgeon - ONNX GraphSurgeon for TensorRT package
python3-libnvinfer - Python 3 bindings for TensorRT
libnvparsers-dev - TensorRT parsers libraries
libnvinfer-plugin8 - TensorRT plugin libraries
libnvinfer8 - TensorRT runtime libraries
libnvonnxparsers8 - TensorRT ONNX libraries
uff-converter-tf - UFF converter for TensorRT package
libnvonnxparsers-dev - TensorRT ONNX libraries
libnvinfer-samples - TensorRT samples
python3-libnvinfer-dev - Python 3 development package for TensorRT
libnvparsers8 - TensorRT parsers libraries
nv-tensorrt-repo-ubuntu2004-cuda11.4-trt8.2.1.8-ga-20211117 - nv-tensorrt repository configuration files

I have also tried:

$ sudo apt-get install python3-libnvinfer-dev
Reading package lists... Done
Building dependency tree       
Reading state information... Done
Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:

 python3-libnvinfer-dev : Depends: python3-libnvinfer (= 8.2.3-1+cuda11.4) but it is not going to be installed
                          Depends: libnvinfer-dev (= 8.2.3-1+cuda11.4) but it is not going to be installed
                          Depends: libnvinfer-plugin-dev (= 8.2.3-1+cuda11.4) but it is not going to be installed
                          Depends: libnvparsers-dev (= 8.2.3-1+cuda11.4) but it is not going to be installed
                          Depends: libnvonnxparsers-dev (= 8.2.3-1+cuda11.4) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.

[AsianZeus]

any updates or fixes for this issue??

[dmitrii-khizbullin]

I face the same issue and still do not have any solution for it

[dmitrii-khizbullin]

Found solution here https://forums.developer.nvidia.com/t/tensorrt-8-0-installation/180498/6

[sergeyleyko]

the solution is to install cudnn packages through the deb installation, not tar,
as described here: https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html

[leekiaxnn]

same issue too, anyone know how to fix it?

[iuskov-nablet]

I have bumped to same issue when I installed the cuda + driver from the .run package. It turned out that result of .run installation is NOT same as after .deb repository unpacking: install from .run script does not register cuda packages at all. In result the tensorrt installation does not see necessary cuda packs and fails.

NOTE: It is important to uninstallfiles instelled by .run script correctly else cuda installation from .deb will fail.
There are two commands are necessary:
sudo /usr/local/cuda-xx.x/bin/./cuda-uninstaller
then
sudo nvidia-uninstall
then reboot
then install cuda+driver from a .deb repo

[Heppa5]

I have bumped to same issue when I installed the cuda + driver from the .run package. It turned out that result of .run installation is NOT same as after .deb repository unpacking: install from .run script does not register cuda packages at all. In result the tensorrt installation does not see necessary cuda packs and fails.

NOTE: It is important to uninstallfiles instelled by .run script correctly else cuda installation from .deb will fail. There are two commands are necessary: sudo /usr/local/cuda-xx.x/bin/./cuda-uninstaller then sudo nvidia-uninstall then reboot then install cuda+driver from a .deb repo

Thanks @iuskov-nablet.
I previously installed Cuda with a run file from NVidia's archive and could not get it to work.

Just a quick note to anybody using their archive.
The older versions of cuda are shipped with old nvidia drivers in the archive, which caused an issue for me because I use the 40-series graphics cards.
If you want older cuda versions but with newer drivers, then take a look here.
Once it's setup, then you can install cuda like this:

sudo apt-get install cuda-11-3

For the above command I got nvidia-driver-535 shipped along with cuda11.3 :)