Error on installing TensorRT with SDK Manager 1.3.1

Autonomous Vehicles DRIVE AGX Xavier DRIVE AGX Xavier General
1

Please provide the following info:
Hardware Platform: DRIVE AGX Xavier Developer Kit
Software Version: DRIVE Software 10
Host Machine Version: native Ubuntu 18.04
SDK Manager Version: 1.3.1.7110

Hi, I updated my SDK Manager today and wanted to reflash the Drive AGX Pegasus System. I get an error when it tries to install TensorRT on the host system.

I have tried to uninstall and redownload everything but no success. Any Idea on how to resolve this?

Here is relevant part of the log:

14:49:37 INFO: TensorRT (with cuDNN 7.5) x86: exec_command: sudo apt-get update; sleep 0.5; sudo apt-get -y --allow-downgrades install libcudnn7=7.5* libcudnn7-dev=7.5* graphsurgeon-tf=5.1* uff-converter-tf=5.1* libnvinfer5=5.1* libnvinfer-dev=5.1* libnvinfer-samples=5.1* python3-libnvinfer=5.1* python3-libnvinfer-dev=5.1* python-libnvinfer=5.1* python-libnvinfer-dev=5.1* tensorrt=5.1* libgie-dev=5.1*

[
]

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: Some packages could not be installed. This may mean that you have

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: requested an impossible situation or if you are using the unstable

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: distribution that some required packages have not yet been created

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: or been moved out of Incoming.

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: The following information may help to resolve the situation:

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: The following packages have unmet dependencies:

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: libnvinfer-dev : Breaks: libgie-dev (< 5.1.5-1+cuda10.1) but 5.1.4-1+cuda10.2 is to be installed

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: tensorrt : Depends: libnvinfer5 (= 5.1.4-1+cuda10.2) but 5.1.5-1+cuda10.1 is to be installed

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: Depends: libnvinfer-dev (= 5.1.4-1+cuda10.2) but 5.1.5-1+cuda10.1 is to be installed

14:11:06 ERROR: TensorRT (with cuDNN 7.5) x86: E: Unable to correct problems, you have held broken packages.

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: [ Package Install Finished with Error ]

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: [host] [ 117.50 MB used. Disk Avail: 143.42 GB ]

14:11:06 INFO: TensorRT (with cuDNN 7.5) x86: [ NV_TENSORRT_HOST_COMP Install took 22s ]

14:11:06 ERROR: TensorRT (with cuDNN 7.5) x86: [error]: Error when apt install failed; [exec_command]: sudo apt-get update; sleep 0.5; sudo apt-get -y --allow-downgrades install libcudnn7=7.5* libcudnn7-dev=7.5* graphsurgeon-tf=5.1* uff-converter-tf=5.1* libnvinfer5=5.1* libnvinfer-dev=5.1* libnvinfer-samples=5.1* python3-libnvinfer=5.1* python3-libnvinfer-dev=5.1* python-libnvinfer=5.1* python-libnvinfer-dev=5.1* tensorrt=5.1* libgie-dev=5.1* ; [error]: exit status 100; [deb_path]: /home/fabian/Downloads/nvidia/sdkm_downloads/nv-tensorrt-repo-ubuntu1804-cuda10.2-trt5.1.4.2-ga-20190506_1-1_amd64.deb

14:11:06 ERROR: TensorRT (with cuDNN 7.5) x86: command terminated with error

14:11:06 SUMMARY: TensorRT (with cuDNN 7.5) x86: APT system is broken and requires manual fix

2

Hi @fabian.h,

Please share ~/.nvsdkm/logs/DRIVE_Software_10.0_Linux_OS/NV_TENSORRT_HOST_COMP.log for our investigation. At the same time, could you try to narrow down what’s the problem in below command by running it manually?

sudo apt-get update; sleep 0.5; sudo apt-get -y --allow-downgrades install libcudnn7=7.5* libcudnn7-dev=7.5* graphsurgeon-tf=5.1* uff-converter-tf=5.1* libnvinfer5=5.1* libnvinfer-dev=5.1* libnvinfer-samples=5.1* python3-libnvinfer=5.1* python3-libnvinfer-dev=5.1* python-libnvinfer=5.1* python-libnvinfer-dev=5.1* tensorrt=5.1* libgie-dev=5.1*
3

Hi @VickNV,

there seems to be an dependency issue. I get the same error as in the logs when running it manually, but there is some more info:

Building dependency tree
Reading state information
 Done
Selected version ‘7.5.1.14-1+cuda10.2’ (NVIDIA CUDA:localhost [amd64]) for ‘libcudnn7’
Selected version ‘7.5.1.14-1+cuda10.2’ (NVIDIA CUDA:localhost [amd64]) for ‘libcudnn7-dev’
Selected version ‘5.1.4-1+cuda10.2’ (NVIDIA CUDA:localhost [amd64]) for ‘graphsurgeon-tf’
Selected version ‘5.1.4-1+cuda10.2’ (NVIDIA CUDA:localhost [amd64]) for ‘uff-converter-tf’
Selected version ‘5.1.5-1+cuda10.1’ (NVIDIA CUDA:developer.download.nvidia.com [amd64]) for ‘libnvinfer5’
Selected version ‘5.1.5-1+cuda10.1’ (NVIDIA CUDA:developer.download.nvidia.com [amd64]) for ‘libnvinfer-dev’
Selected version ‘5.1.4-1+cuda10.2’ (NVIDIA CUDA:localhost [all]) for ‘libnvinfer-samples’
Selected version ‘5.1.5-1+cuda10.1’ (NVIDIA CUDA:developer.download.nvidia.com [amd64]) for ‘python3-libnvinfer’
Selected version ‘5.1.5-1+cuda10.1’ (NVIDIA CUDA:developer.download.nvidia.com [amd64]) for ‘python3-libnvinfer-dev’
Selected version ‘5.1.5-1+cuda10.1’ (NVIDIA CUDA:developer.download.nvidia.com [amd64]) for ‘python-libnvinfer’
Selected version ‘5.1.5-1+cuda10.1’ (NVIDIA CUDA:developer.download.nvidia.com [amd64]) for ‘python-libnvinfer-dev’
Selected version ‘5.1.4.2-1+cuda10.2’ (NVIDIA CUDA:localhost [amd64]) for ‘tensorrt’
Selected version ‘5.1.4-1+cuda10.2’ (NVIDIA CUDA:localhost [all]) for ‘libgie-dev’
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:
libnvinfer-dev : Breaks: libgie-dev (< 5.1.5-1+cuda10.1) but 5.1.4-1+cuda10.2 is to be installed
tensorrt : Depends: libnvinfer5 (= 5.1.4-1+cuda10.2) but 5.1.5-1+cuda10.1 is to be installed
Depends: libnvinfer-dev (= 5.1.4-1+cuda10.2) but 5.1.5-1+cuda10.1 is to be installed
E: Unable to correct problems, you have held broken packages.

Seems like there is a problem with the selected cuda-versions. My version is 10.2

/usr/local/cuda/bin/nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Oct_23_19:24:38_PDT_2019
Cuda compilation tools, release 10.2, V10.2.89
fabian@LB-FABIANH:~$ cat /usr/local/cuda/version.txt
CUDA Version 10.2.89

Running apt clean and apt auto-clean did not help.

Here is the complete log:

Info: [host] [ Disk Avail:124.35 GB ]
Info: NV_TENSORRT_HOST_COMP include 1 package
Info: [ Package Install Started ]
Info: deb installer start to install
Info: exec_command: dpkg -f /home/fabian/Downloads/nvidia/sdkm_downloads/nv-tensorrt-repo-ubuntu1804-cuda10.2-trt5.1.4.2-ga-20190506_1-1_amd64.deb Package | tr -d ‘\n’
Info: exec_command: dpkg -f /home/fabian/Downloads/nvidia/sdkm_downloads/nv-tensorrt-repo-ubuntu1804-cuda10.2-trt5.1.4.2-ga-20190506_1-1_amd64.deb Version | tr -d ‘\n’
Info: exec_command: sudo -E apt-get install -y --allow-downgrades ‘/home/fabian/Downloads/nvidia/sdkm_downloads/nv-tensorrt-repo-ubuntu1804-cuda10.2-trt5.1.4.2-ga-20190506_1-1_amd64.deb’
Reading package lists

Building dependency tree

Reading state information

nv-tensorrt-repo-ubuntu1804-cuda10.2-trt5.1.4.2-ga-20190506 is already the newest version (1-1).
0 upgraded, 0 newly installed, 0 to remove and 248 not upgraded.
Info: exec_command: dpkg -L nv-tensorrt-repo-ubuntu1804-cuda10.2-trt5.1.4.2-ga-20190506 | grep ‘.pub$’ | tr -d ‘\n’
Info: exec_command: sudo apt-key add /var/nv-tensorrt-repo-cuda10.2-trt5.1.4.2-ga-20190506/7fa2af80.pub
Warning: apt-key output should not be parsed (stdout is not a terminal)
OK
Info: exec_command: sudo apt-get check
Reading package lists

Building dependency tree

Reading state information

Info: exec_command: sudo apt-get update; sleep 0.5; sudo apt-get -y --allow-downgrades install libcudnn7=7.5* libcudnn7-dev=7.5* graphsurgeon-tf=5.1* uff-converter-tf=5.1* libnvinfer5=5.1* libnvinfer-dev=5.1* libnvinfer-samples=5.1* python3-libnvinfer=5.1* python3-libnvinfer-dev=5.1* python-libnvinfer=5.1* python-libnvinfer-dev=5.1* tensorrt=5.1* libgie-dev=5.1*
Get:1 file:/var/cuda-repo-10-2-local-10.2.19-430.17 InRelease
Ign:1 file:/var/cuda-repo-10-2-local-10.2.19-430.17 InRelease
Get:2 file:/var/cuda-repo-10-2-local-10.2.19-cross-aarch64 InRelease
Ign:2 file:/var/cuda-repo-10-2-local-10.2.19-cross-aarch64 InRelease
Get:3 file:/var/nv-tensorrt-repo-cuda10.2-trt5.1.4.2-ga-20190506 InRelease
Ign:3 file:/var/nv-tensorrt-repo-cuda10.2-trt5.1.4.2-ga-20190506 InRelease
Get:4 file:/var/cuda-repo-10-2-local-10.2.19-430.17 Release [574 B]
Get:5 file:/var/cuda-repo-10-2-local-10.2.19-cross-aarch64 Release [574 B]
Get:6 file:/var/nv-tensorrt-repo-cuda10.2-trt5.1.4.2-ga-20190506 Release [574 B]
Get:4 file:/var/cuda-repo-10-2-local-10.2.19-430.17 Release [574 B]
Get:5 file:/var/cuda-repo-10-2-local-10.2.19-cross-aarch64 Release [574 B]
Get:6 file:/var/nv-tensorrt-repo-cuda10.2-trt5.1.4.2-ga-20190506 Release [574 B]
Hit:7 Index of /graphics-drivers/ppa/ubuntu bionic InRelease
Hit:8 Index of /ubuntu bionic InRelease
Hit:9 http://dl.google.com/linux/chrome/deb stable InRelease
Hit:10 Index of /ubuntu bionic-updates InRelease
Hit:11 Index of linux/ubuntu/ bionic InRelease
Hit:12 Index of /ubuntu bionic-backports InRelease
Hit:13 Index of /repos/ms-teams/ stable InRelease
Ign:15 Index of /compute/cuda/repos/ubuntu1804/x86_64 InRelease
Hit:16 Index of /ubuntu bionic-security InRelease
Ign:18 Index of /compute/machine-learning/repos/ubuntu1804/x86_64 InRelease
Hit:19 Index of /compute/cuda/repos/ubuntu1804/x86_64 Release
Hit:20 Index of /compute/machine-learning/repos/ubuntu1804/x86_64 Release
Reading package lists

Reading package lists

Building dependency tree

Reading state information

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:
libnvinfer-dev : Breaks: libgie-dev (< 5.1.5-1+cuda10.1) but 5.1.4-1+cuda10.2 is to be installed
tensorrt : Depends: libnvinfer5 (= 5.1.4-1+cuda10.2) but 5.1.5-1+cuda10.1 is to be installed
Depends: libnvinfer-dev (= 5.1.4-1+cuda10.2) but 5.1.5-1+cuda10.1 is to be installed
E: Unable to correct problems, you have held broken packages.
Info: [ Package Install Finished with Error ]
Info: [host] [ 72.00 KB used. Disk Avail: 124.35 GB ]
Info: [ NV_TENSORRT_HOST_COMP Install took 14s ]
Error: [error]: Error when apt install failed; [exec_command]: sudo apt-get update; sleep 0.5; sudo apt-get -y --allow-downgrades install libcudnn7=7.5* libcudnn7-dev=7.5* graphsurgeon-tf=5.1* uff-converter-tf=5.1* libnvinfer5=5.1* libnvinfer-dev=5.1* libnvinfer-samples=5.1* python3-libnvinfer=5.1* python3-libnvinfer-dev=5.1* python-libnvinfer=5.1* python-libnvinfer-dev=5.1* tensorrt=5.1* libgie-dev=5.1* ; [error]: exit status 100; [deb_path]: /home/fabian/Downloads/nvidia/sdkm_downloads/nv-tensorrt-repo-ubuntu1804-cuda10.2-trt5.1.4.2-ga-20190506_1-1_amd64.deb

4

a URI in a source.list file including 5.1.5 packages may cause the problem.
Please back up all /etc/apt/sources.list.d/* files, remove them and then try again.

5

Hi @VickNV

these where the files I removed (also the list.save files):

ls ~/sources_backup/*.list
/home/fabian/sources_backup/cuda-10-2-local-10.2.19-430.17.list
/home/fabian/sources_backup/cuda-10-2-local-10.2.19-cross-aarch64.list
/home/fabian/sources_backup/cuda.list
/home/fabian/sources_backup/google-chrome.list
/home/fabian/sources_backup/graphics-drivers-ubuntu-ppa-bionic.list
/home/fabian/sources_backup/nvidia-machine-learning.list
/home/fabian/sources_backup/nv-tensorrt-cuda10.2-trt5.1.4.2-ga-20190506.list
/home/fabian/sources_backup/teams.list

now it can’t find the packages:

Info: exec_command: sudo apt-get update; sleep 0.5; sudo apt-get -y --allow-downgrades install libcudnn7=7.5* libcudnn7-dev=7.5* graphsurgeon-tf=5.1* uff-converter-tf=5.1* libnvinfer5=5.1* libnvinfer-dev=5.1* libnvinfer-samples=5.1* python3-libnvinfer=5.1* python3-libnvinfer-dev=5.1* python-libnvinfer=5.1* python-libnvinfer-dev=5.1* tensorrt=5.1* libgie-dev=5.1*
Hit:1 Index of linux/ubuntu/ bionic InRelease
Hit:2 Index of /ubuntu bionic InRelease
Hit:3 Index of /ubuntu bionic-updates InRelease
Hit:4 Index of /ubuntu bionic-backports InRelease
Hit:5 Index of /ubuntu bionic-security InRelease
Reading package lists

Reading package lists

Building dependency tree

Reading state information

E: Version ‘7.5*’ for ‘libcudnn7’ was not found
E: Version ‘7.5*’ for ‘libcudnn7-dev’ was not found
E: Unable to locate package graphsurgeon-tf
E: Unable to locate package uff-converter-tf
E: Unable to locate package libnvinfer5
E: Unable to locate package libnvinfer-dev
E: Unable to locate package libnvinfer-samples
E: Unable to locate package python3-libnvinfer
E: Unable to locate package python3-libnvinfer-dev
E: Unable to locate package python-libnvinfer
E: Unable to locate package python-libnvinfer-dev
E: Unable to locate package tensorrt
E: Unable to locate package libgie-dev
Info: [ Package Install Finished with Error ]
Info: [host] [ 24.00 KB used. Disk Avail: 81.40 GB ]
Info: [ NV_TENSORRT_HOST_COMP Install took 12s ]
Error: [error]: Error when apt install failed; [exec_command]: sudo apt-get update; sleep 0.5; sudo apt-get -y --allow-downgrades install libcudnn7=7.5* libcudnn7-dev=7.5* graphsurgeon-tf=5.1* uff-converter-tf=5.1* libnvinfer5=5.1* libnvinfer-dev=5.1* libnvinfer-samples=5.1* python3-libnvinfer=5.1* python3-libnvinfer-dev=5.1* python-libnvinfer=5.1* python-libnvinfer-dev=5.1* tensorrt=5.1* libgie-dev=5.1* ; [error]: exit status 100; [deb_path]: /home/fabian/Downloads/nvidia/sdkm_downloads/nv-tensorrt-repo-ubuntu1804-cuda10.2-trt5.1.4.2-ga-20190506_1-1_amd64.deb

6

I suspect the issue is from one of below lists (propably nvidia-machine-learning.list).

/home/fabian/sources_backup/cuda.list
/home/fabian/sources_backup/google-chrome.list
/home/fabian/sources_backup/graphics-drivers-ubuntu-ppa-bionic.list
/home/fabian/sources_backup/nvidia-machine-learning.list
/home/fabian/sources_backup/teams.list

So please copy “only” below three source lists (should be from DRIVE Software 10 installation) to /etc/apt/sources.list.d directory and try again.

cuda-10-2-local-10.2.19-430.17.list
cuda-10-2-local-10.2.19-cross-aarch64.list
nv-tensorrt-cuda10.2-trt5.1.4.2-ga-20190506.list

If it still doesn’t help, please try clean reinstallation by install after running below commands.

$ rm /etc/apt/sources.list.d/*
$ rm -rf ~/.nvsdkm* ~/Downloads/nvidia/sdkm_downloads* ~/nvidia/nvidia_sdk*
8

Thanks, that worked!

1 Like