Failing to install TensorRT 10 on Jetson Orin NX

[talcs]

Description

I tried to install the new TensorRT version, which seemed to be successful. However, when running /usr/src/tensorrt/bin/trtexec, I still see the old version: [TensorRT v8502].

• I downloaded the new TensorRT version from here. Since it is Jetson (with JetPack) that uses Ubuntu, I tried to install one of the DEB packages from the JetPack list:

• The behavior was consistent for the 2 DEB packages:

  • When trying to install using apt install, I got an explicit error:

$ sudo apt install nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb 
Reading package lists... Done
Building dependency tree       
Reading state information... Done
E: Unable to locate package nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb

• On the other hand, when trying to install using dpkg -i, it seemed to be successful. However, as stated at the beginning, nothing is actually installed and the trtexec version remains the old one:

$ sudo dpkg -i nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb 
(Reading database ... 162568 files and directories currently installed.)
Preparing to unpack nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb ...
Unpacking nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4 (1.0-1) over (1.0-1) ...
Setting up nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4 (1.0-1) ...

I later followed the instructions from here, but still have errors:

$ sudo cp /var/nv-tensorrt-local-repo-l4t-10.0.1-cuda-12.4_1.0-1_arm64/*-keyring.gpg /usr/share/keyrings/
$ sudo apt-get update
$ 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: libnvinfer-samples (= 10.0.1.6-1+cuda12.4) but it is not going to be installed
            Depends: python3-libnvinfer-dev (= 10.0.1.6-1+cuda12.4) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.

How can I make the new version work?

[lix19937]

Download nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb at local, then sudo dpkg -i nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb

[zerollzeng]

Download nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb at local, then sudo dpkg -i install nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb

No, cross-aarch64 is for cross compilation, TRT is bundled with Jetpack on embedded platform for user. Please wait for the new JP release.

[zerollzeng]

Other usage is not verified by QA.

[lix19937]

Download nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb at local, then sudo dpkg -i install nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb

No, cross-aarch64 is for cross compilation, TRT is bundled with Jetpack on embedded platform for user. Please wait for the new JP release.

Yes, cross-aarch64 is for cross compilation, I just means loacl/host machine to compile.

[ttyio]

closing since there is already answer in the thread, thanks all!

[krishnarajk]

Does tensorrt 10 will work with jetson orin nano jetpack 6?

[lix19937]

No, @krishnarajk more to see https://forums.developer.nvidia.com/t/upgrade-tensorrt-and-cuda-on-agx-orin-without-jetpack-6-0/280066

[quanfeifan]

Download nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb at local, then sudo dpkg -i install nv-tensorrt-local-repo-cross-aarch64-l4t-10.0.1-cuda-12.4_1.0-1_all.deb

No, cross-aarch64 is for cross compilation, TRT is bundled with Jetpack on embedded platform for user. Please wait for the new JP release.

Yes, cross-aarch64 is for cross compilation, I just means loacl/host machine to compile.

not work, the same as https://github.com/NVIDIA/TensorRT/issues/3912#issue-2322734647. Any more solutions?