Jetson Orin NX Engineering Performance Developer Kit Super使用Docker安装部署TensorRT-YOLO

huomama

已于 2025-03-27 09:48:51 修改

阅读量195

点赞数 10

分类专栏： Jetson 文章标签： docker YOLO 容器 tensorrt Jetson

于 2025-03-26 17:59:42 首次发布

本文链接：https://blog.csdn.net/huomama/article/details/146525018

版权

Jetson 专栏收录该内容

2 篇文章

订阅专栏

参考

教程参考

https://blog.csdn.net/lida2003/article/details/145322174
https://blog.csdn.net/fengshengwei3/article/details/146228414

下载

https://developer.nvidia.com/rdp/cudnn-archive
https://forums.developer.nvidia.com/t/pytorch-for-jetson/72048
https://pypi.jetson-ai-lab.dev/jp6/cu126
https://elinux.org/Jetson/L4T/TRT_Customized_Example#YoloV11_using_Pytorch
https://pytorch.org/TensorRT/getting_started/jetpack.html
https://developer.nvidia.com/cudnn-9-3-0-download-archive?target_os=Linux&target_arch=aarch64-jetson&Compilation=Native&Distribution=Ubuntu&target_version=22.04&target_type=deb_local
https://developer.download.nvidia.com/compute/redist/jp/v61/pytorch/
https://forums.developer.nvidia.com/t/jetpack-6-1-cuda-12-6-support-cusparselt-pytorch-installation/310222
https://developer.nvidia.com/cusparselt-downloads?target_os=Linux&target_arch=aarch64-jetson&Compilation=Native&Distribution=Ubuntu&target_version=22.04&target_type=deb_local

开发环境

基础环境：
jetson orin nx engineering performance developer kit super 【Mem 16G】
jetpack 6.2 【https://developer.nvidia.com/embedded/jetpack-sdk-62】
cuda: 12.6.68
cudnn: 9.3.0.75
tensorrt: 10.3.0.30
python: 3.10

docker：
docker image: nvcr.io/nvidia/l4t-tensorrt:r10.3.0-devel
cmake version: 3.14
考虑到适配性问题，我下载10.3.0的tensorrt镜像，cuda版本和cudnn版本与基础环境保持一致。
查看cuda版本命令nvcc -V, 当然也可以用dpkg -l|grep cuda查看安装的包
可以用dpkg -l|grep cudnn查看安装的cudnn

Dockerfile

FROM nvcr.io/nvidia/l4t-tensorrt:r10.3.0-devel

# Required to build Ubuntu 22.04 without user prompts with DLFW container
ARG DEBIAN_FRONTEND=noninteractive

# Set environment and working directory
ENV TZ=Asia/Shanghai \
    LANG=C.UTF-8

RUN sed -i 's|http://ports.ubuntu.com/ubuntu-ports/|https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/|g' /etc/apt/sources.list


# Install requried libraries
RUN apt-get update && apt-get install -y --no-install-recommends \
    software-properties-common \
    build-essential \
    gdb \
    libopencv-dev \
    wget \
    git \
    pkg-config \
    ssh \
    pbzip2 \
    bzip2 \
    unzip \
    axel \
    vim \
    curl

# 设置 pip 使用清华源
RUN pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple \
    && pip config set install.trusted-host pypi.tuna.tsinghua.edu.cn \
    && pip install --upgrade pip \
    && pip install --upgrade setuptools>=41.0.0 \
    && pip install --upgrade "pybind11[global]"

WORKDIR /workspace

# 准备好依赖，包括torch等
COPY dependencies/ /workspace/dependencies/
    
# 安装torch, torchvision, 参考https://blog.csdn.net/lida2003/article/details/145322174下载torch,下载torchvision参考https://elinux.org/Jetson/L4T/TRT_Customized_Example#YoloV11_using_Pytorch
RUN cd /workspace/dependencies/ \
	&& pip install torch-2.5.1+l4t36.4-cp310-cp310-linux_aarch64.whl \
	&& pip install torchvision-0.20.0-cp310-cp310-linux_aarch64.whl

# 安装最新cmake
RUN cd /workspace/dependencies/ \ 
	&& tar -zxvf cmake-3.31.6-linux-aarch64.tar.gz \
	&& mv cmake-3.31.6-linux-aarch64 / 
ENV PATH=/cmake-3.31.6-linux-aarch64/bin:$PATH

# 安装(cudnn)[https://developer.nvidia.com/cudnn-9-3-0-download-archive?target_os=Linux&target_arch=aarch64-jetson&Compilation=Native&Distribution=Ubuntu&target_version=22.04&target_type=deb_local]

RUN wget https://developer.download.nvidia.com/compute/cudnn/9.3.0/local_installers/cudnn-local-tegra-repo-ubuntu2204-9.3.0_1.0-1_arm64.deb \
	&& dpkg -i cudnn-local-tegra-repo-ubuntu2204-9.3.0_1.0-1_arm64.deb \
	&& cp /var/cudnn-local-tegra-repo-ubuntu2204-9.3.0/cudnn-*-keyring.gpg /usr/share/keyrings/ \
	&& apt-get update \
	&& apt-get -y install cudnn \
	&& rm cudnn-local-tegra-repo-ubuntu2204-9.3.0_1.0-1_arm64.deb

# 安装缺失库
RUN wget https://developer.download.nvidia.com/compute/cusparselt/0.7.1/local_installers/cusparselt-local-tegra-repo-ubuntu2204-0.7.1_1.0-1_arm64.deb \
    && dpkg -i cusparselt-local-tegra-repo-ubuntu2204-0.7.1_1.0-1_arm64.deb \
	&& cp /var/cusparselt-local-tegra-repo-ubuntu2204-0.7.1/cusparselt-*-keyring.gpg /usr/share/keyrings/ \
	&& apt-get update \
	&& apt-get -y install libcusparselt0 libcusparselt-dev  \
	&& rm cusparselt-local-tegra-repo-ubuntu2204-0.7.1_1.0-1_arm64.deb

RUN apt update \
	&& apt install -y --no-install-recommends libopenblas-dev \
    && apt-get install libnuma-dev
    
# 下载不下来可以用https://gitee.com/menglongyue-love/TensorRT-YOLO.git
RUN git clone https://github.com/laugh12321/TensorRT-YOLO.git /workspace/TensorRT-YOLO \
	&& cd /workspace/TensorRT-YOLO \
	&& cmake -S . -B build -DTENSORRT_PATH=/usr/local/tensorrt -DBUILD_PYTHON=ON \
	&& cmake --build build -j$(nproc) --config Release

RUN cd  /workspace/TensorRT-YOLO/python \
    && pip install --upgrade build \
    && python -m build --wheel \
    && pip install dist/tensorrt_yolo-6.0.0-py3-none-any.whl[export]

RUN pip install dill \
    && pip install onnxexplorer \
    && pip install numpy==1.26.3

RUN rm -rf /workspace/dependencies && rm -rf /var/lib/apt/lists/*

VOLUME /workspace

RUN ["/bin/bash"]

Build

build.sh

docker build -t  jetson-trtyolo-base:jetpack62 .

运行构建镜像
bash build.sh

测试

docker run -dit --gpus all --shm-size=8g  --restart=always  --net=host --name test --runtime nvidia -e NVIDIA_DRIVER_CAPABILITIES=video,compute,utility -v /etc/localtime:/etc/localtime jetson-trtyolo-base:jetpack62 /bin/bash

过程中遇到的问题

python -c "import torch; print(torch.version.cuda, torch.backends.cudnn.version())"

Traceback (most recent call last):

  File "<string>", line 1, in <module>

  File "/usr/local/lib/python3.10/dist-packages/torch/__init__.py", line 367, in <module>

    from torch._C import *  # noqa: F403

ImportError: libcudnn.so.9: cannot open shared object file: No such file or directory

使用以下命令找到主机上的 cuDNN 库路径：

ls /usr/lib/aarch64-linux-gnu/libcudnn*

找不到直接装，安装[cudnn](https://developer.nvidia.com/cudnn-9-3-0-download-archive?target_os=Linux&target_arch=aarch64-jetson&Compilation=Native&Distribution=Ubuntu&target_version=22.04&target_type=deb_local)


ImportError: libopenblas.so.0: cannot open shared object file: No such file or directory
这个错误表示你的程序（通常是 Python 的 NumPy 或 SciPy 等库）尝试加载 OpenBLAS 库，但系统找不到它。OpenBLAS 是一个优化的 BLAS（Basic Linear Algebra Subprograms）库，用于执行基本的线性代数运算，它对数值计算性能至关重要。

find / -name "libopenblas.so.0"
找不到装
apt update
apt install libopenblas-dev


ImportError: libcusparseLt.so.0: cannot open shared object file: No such file or directory
这个错误表明你的程序（通常是依赖于 CUDA 的深度学习框架，如 PyTorch 或 TensorFlow）无法找到 `libcusparseLt.so.0` 共享库。`libcusparseLt.so.0` 是 NVIDIA cuSPARSELt 库的一部分，它提供了用于稀疏矩阵运算的优化函数，特别是在深度学习中用于加速 Transformer 模型等。
find / -name libcusparseLt.so.0
https://forums.developer.nvidia.com/t/jetpack-6-1-cuda-12-6-support-cusparselt-pytorch-installation/310222
https://developer.nvidia.com/cusparselt-downloads?target_os=Linux&target_arch=aarch64-jetson&Compilation=Native&Distribution=Ubuntu&target_version=22.04&target_type=deb_local


wget https://developer.download.nvidia.com/compute/cusparselt/0.7.1/local_installers/cusparselt-local-tegra-repo-ubuntu2204-0.7.1_1.0-1_arm64.deb
dpkg -i cusparselt-local-tegra-repo-ubuntu2204-0.7.1_1.0-1_arm64.deb
cp /var/cusparselt-local-tegra-repo-ubuntu2204-0.7.1/cusparselt-*-keyring.gpg /usr/share/keyrings/
apt-get update
apt-get -y install libcusparselt0 libcusparselt-dev


ImportError: libnuma.so.1: cannot open shared object file: No such file or directory
apt-get install libnuma-dev