NVIDIA / TensorRT

NVIDIA® TensorRT™ is an SDK for high-performance deep learning inference on NVIDIA GPUs. This repository contains the open source components of TensorRT.
https://developer.nvidia.com/tensorrt
Apache License 2.0
10.81k stars 2.13k forks source link

ERROR [13/25] COPY docker/jetpack_files /pdk_files #4140

Open krishnarajk opened 1 month ago

krishnarajk commented 1 month ago

Description

I am trying to setup tensort rt oss on the jetson orin nano. While executing the https://github.com/NVIDIA/TensorRT/blob/release/8.6/docker/ubuntu-cross-aarch64.Dockerfile file it is throwing the error with `ERROR [13/25] COPY docker/jetpack_files /pdk_files 0.0s

[13/25] COPY docker/jetpack_files /pdk_files:

ubuntu-cross-aarch64.Dockerfile:69

67 | # Skip installing PyPI packages and NGC client on cross-build container 68 |
69 | >>> COPY docker/jetpack_files /pdk_files 70 | COPY scripts/stubify.sh /pdk_files 71 |

ERROR: failed to solve: failed to compute cache key: failed to calculate checksum of ref 156867a5-6c53-4894-8bca-0c027714172f::f1lz936xp5y2zkhe4z4gs28dg: "/docker/jetpack_files": not found ` I have modified the file to accept the cuda version 12.2.0

Environment

TensorRT Version: 8.6.1

NVIDIA GPU: jetson orin nano

NVIDIA Driver Version:

CUDA Version: 12.2

CUDNN Version:

lix19937 commented 1 month ago

You can find the scripts/stubify.sh file in you env or in github repo.

I upload my stubify.sh as follow, you can save it as stubify.sh

#!/bin/bash
#
# This short shell script will extract all the strong "text" symbols from the
# shared library and create a new "stub" shared library with the same symbols.
# The body of these functions will be empty and therefore have no dependencies.
# This scripts uses whatever CC is defined in the user's environment.
#

set -o pipefail

# check arguments
if [ $# -ne 2 ] ; then
    echo "Usage: $(basename $0) IN_LIBFILE OUT_LIBFILE"
    exit 1
fi

IN_LIBFILE="$1"
OUT_LIBFILE="$2"

# check compiler
if [ -z "${CC}" ] ; then
    echo "Error: Environment variable 'CC' has not been defined"
    exit 1
fi

SONAME=$(readelf -d "${IN_LIBFILE}" | grep '(SONAME)' | cut -d [ -f 2 | cut -d ] -f 1)

# make stub library
nm -D "${IN_LIBFILE}" | \
    awk '{if ($2 == "T") { print "void",$3,"() {}" }}' | \
    "${CC}" -x c -O0 -fPIC -shared -Wl,-soname=${SONAME} -Wl,--strip-all -o "${OUT_LIBFILE}" -

exit $?

@krishnarajk

krishnarajk commented 1 month ago

@lix19937 thank you for your reply. It also throws error for

COPY docker/jetpack_files /pdk_files

i couldnt find the jetpack_files in the repo. How can I get those files?

lix19937 commented 1 month ago

Do you has install JetPack SDK ?

Download and launch the JetPack SDK manager. Login with your NVIDIA developer account. Select the platform and target OS, and click Continue. Under Download & Install Options change the download folder and select Download now, Install later. Agree to the license terms and click Continue. Move the extracted files into the /docker/jetpack_files folder. @krishnarajk

krishnarajk commented 3 weeks ago

I was trying with tensort 10.3 release. with jetpack 6.1 still files are missing after copying files from the jetpack manager, it has files named l4t-cuda-terga-repo*.deb etc

``ERROR [16/24] RUN dpkg -i /pdk_files/cuda-repo-cross-aarch64.deb /pdk_files/cuda-repo-ubuntu_amd64.deb && sudo cp /var/cuda-repo-cross-aarch64/cuda-keyring.gpg /usr/share/keyrings/ && sudo 0.6s


[16/24] RUN dpkg -i /pdk_files/cuda-repo-cross-aarch64.deb /pdk_files/cuda-repo-ubuntu_amd64.deb && sudo cp /var/cuda-repo-cross-aarch64/cuda-keyring.gpg /usr/share/keyrings/ && sudo cp /var/cuda-repo-ubuntu2204/cuda-keyring.gpg /usr/share/keyrings/ && apt-get update && apt-get install -y cuda-cross-aarch64 && rm -rf /var/lib/apt/lists/:
0.469 dpkg: error: cannot access archive '/pdk_files/cuda-repo-cross-aarch64
.deb': No such file or directory