TensorRT Jetson: no matching manifest for linux/arm64/v8 in the manifest list entries

Description

Environment

Jetson Nano CUDA 10.2 Ubuntu 18.04

Steps To Reproduce

git clone -b master https://github.com/nvidia/TensorRT TensorRT cd TensorRT git submodule update --init --recursive sudo docker build -f docker/ubuntu-18.04.Dockerfile --build-arg UBUNTU_VERSION=18.04 --build-arg CUDA_VERSION=10.2 --tag=tensorrt-ubuntu .

Output

Sending build context to Docker daemon  63.45MB
Step 1/31 : ARG CUDA_VERSION=11.3.1
Step 2/31 : ARG OS_VERSION=18.04
Step 3/31 : FROM nvidia/cuda:${CUDA_VERSION}-cudnn8-devel-ubuntu${OS_VERSION}
10.2-cudnn8-devel-ubuntu18.04: Pulling from nvidia/cuda
no matching manifest for linux/arm64/v8 in the manifest list entries```

Sep 07 '21 12:09 hoonkai

I got the same issue, and my specs are the same!

Sep 17 '21 13:09 nrpr161293

I get the same issue with the same specs, is there any update on this?

Oct 05 '21 08:10 PeterHorvath10

same with me. please help me!

Platform

Jetson NX CUDA_VERSION=10.2 OS_VERSION=18.04

Error Info

docker build -f docker/ubuntu-cross-aarch64.Dockerfile --build-arg CUDA_VERSION=10.2 --build-arg uid=0 --build-arg gid=0 --tag=tensorrt-jetpack-cuda10.2 . Sending build context to Docker daemon 61.76MB Step 1/33 : ARG CUDA_VERSION=10.2 Step 2/33 : ARG OS_VERSION=18.04 Step 3/33 : FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${OS_VERSION} 10.2-devel-ubuntu18.04: Pulling from nvidia/cuda no matching manifest for linux/arm64/v8 in the manifest list entries

Nov 05 '21 03:11 yajinwuzl

same

Nov 26 '21 13:11 willosonico

same...

Dec 20 '21 13:12 jarmi-rm

same

Jan 13 '22 20:01 jumpercc

same

Feb 04 '22 13:02 jpacifico

I've also met same error. Does anyone has solution?

Apr 06 '22 10:04 dFuong

I managed to find partial solution https://github.com/NVIDIA/TensorRT/issues/1730#issuecomment-1047843945

Apr 06 '22 11:04 jumpercc

@jumpercc , I've tried running as following your guide. But, I still meet this error. I need to deploy on jetson xavier-nx. Can u help me again?

Apr 08 '22 01:04 dFuong

same here, would love to know how we can resolve this. Thanks!

Jun 09 '22 22:06 Cococatty

The issue is that NVIDIA CUDA docker image does not have aarch64 version: https://hub.docker.com/r/nvidia/cuda/tags?page=1&name=10.2

There isn't much what we can do. Some options include:

Cross-build from x86 using https://github.com/NVIDIA/TensorRT/issues/1730#issuecomment-1047843945
Run TRT on bare-metal without using docker (i.e. directly install TRT on the board and build TRT OSS outside of the docker container)

Jun 15 '22 09:06 nvpohanh

Closing due to >14 days without activity. Please feel free to reopen if the issue still exists. Thanks

Jul 01 '22 06:07 nvpohanh

This is a real problem and is not solved. I recommend this issue should be re-opened.

Jul 01 '22 11:07 johnwalicki

I see. reopen.

Jul 01 '22 14:07 nvpohanh

Still no solution?

Nov 01 '22 07:11 jamesfaure

is this still an open issue?

Jul 17 '24 13:07 krishnarajk