Docker-mmdetection3d

Docker-mmdetection3d

ubuntu18.04 3090Ti

docker安装

https://blog.csdn.net/b1134977524/article/details/120442417
常用命令：

• https://blog.csdn.net/u013378306/article/details/86668313
• https://blog.csdn.net/awodwde/article/details/122047721（wsl）

部署cuda

https://blog.csdn.net/weixin_50008473/article/details/119464898

报错

• 用docker命令时报错，提示：Cannot connect to the Docker daemon at unix:///var/run/docker.sock.
• 出现问题原因：
  可能是上一次没有正常退出docker，所以docker没有正常启动，在相应的/var/run/路径下找不到docker进程。

  [root@node02 ~]# systemctl stop docker.socket
  [root@node02 ~]# sudo systemctl restart docker
  [root@node02 ~]# docker ps
  CONTAINER ID   IMAGE     COMMAND   CREATED   STATUS    PORTS     NAMES
  

• 参考：https://blog.csdn.net/weixin_42449832/article/details/114994451

安装mmdetection3d

参考：https://blog.csdn.net/weixin_41010198/article/details/116133545
可以按需修改docker/下的DockerFile

docker build -t mmdetection3d docker/

生成or启动mmdetection3容器

sudo docker run --gpus all --shm-size=8g -it -v /home/wtw/workspace/:/mmdetection3d/data mmdetection3d

/home/wtw/workspace/映射一个本地的路径到容器中，用于存放数据，这样不会导致你容器删除的时候出现数据丢失！！！

暂时安装镜像没有报错，但是测试demo时，报错mmcv1.3.8冲突

修改mmcv,从https://download.openmmlab.com/mmcv/dist/cu101/torch1.6.0/index.html查询版本

pip uninstall mmcv-full
pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.6.0/mmcv_full-1.3.14-cp37-cp37m-manylinux1_x86_64.whl

如果有问题，参考https://blog.csdn.net/weixin_36670529/article/details/116664344 不同方法

安装报错

ERROR: After October 2020 you may experience errors when installing or updating packages. This is because pip will change the way that it resolves dependency conflicts.We recommend you use --use-feature=2020-resolver to test your packages with the new resolver before it becomes the default.imageio 2.10.3 requires pillow>=8.3.2, but you'll have pillow 8.1.1 which is incompatible.

没有A4B469963BF863CC key，添加

RUN apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv A4B469963BF863CC

Certificate verification failed
将https改成http

Dockerfile中修改pip镜像

RUN pip3 config set global.index-url http://mirrors.aliyun.com/pypi/simple
RUN pip3 config set install.trusted-host mirrors.aliyun.com

还是存在pillow问题，不过容器生成了
手动将pillow升级后，保存新镜像
sudo docker ps
sudo docker commit relaxed_gauss mmdet3d_modify
移除多余镜像

测试

python demo/pcd_demo.py demo/data/kitti/kitti_000008.bin configs/second/hv_second_secfpn_6x8_80e_kitti-3d-car.py data/testmm3d/hv_second_secfpn_6x8_80e_kitti-3d-car_20200620_230238-393f000c.pth --out-dir data/output_result/

• 显示测试结果
  https://mmdetection3d.readthedocs.io/en/latest/useful_tools.html#visualization

修改的Dockerfile

ARG PYTORCH="1.8.0"
ARG CUDA="11.1"
ARG CUDNN="8"FROM pytorch/pytorch:${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-develENV TORCH_CUDA_ARCH_LIST="5.0 7.0 8.0 8.6+PTX"
ENV TORCH_NVCC_FLAGS="-Xfatbin -compress-all"
ENV CMAKE_PREFIX_PATH="$(dirname $(which conda))/../"RUN cp /etc/apt/sources.list /etc/apt/sources.list.bak  \&& echo  "# 默认注释了源码镜像以提高 apt update 速度，如有需要可自行取消注释\ndeb http://mirrors.tuna.tsinghua.edu.cn/ubuntu/ bionic main restricted universe multiverse\n# deb-src http://mirrors.tuna.tsinghua.edu.cn/ubuntu/ bionic main restricted universe multiverse\ndeb http://mirrors.tuna.tsinghua.edu.cn/ubuntu/ bionic-updates main restricted universe multiverse\n# deb-src http://mirrors.tuna.tsinghua.edu.cn/ubuntu/ bionic-updates main restricted universe multiverse\ndeb http://mirrors.tuna.tsinghua.edu.cn/ubuntu/ bionic-backports main restricted universe multiverse\n# deb-src http://mirrors.tuna.tsinghua.edu.cn/ubuntu/ bionic-backports main restricted universe multiverse\ndeb http://mirrors.tuna.tsinghua.edu.cn/ubuntu/ bionic-security main restricted universe multiverse\n# deb-src http://mirrors.tuna.tsinghua.edu.cn/ubuntu/ bionic-security main restricted universe multiverse" >  /etc/apt/sources.list && cat  /etc/apt/sources.list\&& apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv A4B469963BF863CCRUN apt-get update && apt-get install -y ffmpeg libsm6 libxext6 git ninja-build libglib2.0-0 libsm6 libxrender-dev libxext6 \&& apt-get clean \&& rm -rf /var/lib/apt/lists/*
# set pip channel
RUN pip3 config set global.index-url http://mirrors.aliyun.com/pypi/simple 
RUN pip3 config set install.trusted-host mirrors.aliyun.com 
# Install MMCV, MMDetection and MMSegmentation
RUN pip install openmim
RUN mim install mmdet
RUN pip install mmsegmentation
#RUN pip install mmcv-full==1.3.16 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.8.0/index.html
#RUN pip install mmdet==2.17.0
#RUN pip install mmsegmentation==0.18.0# Install MMDetection3D
RUN conda clean --all
RUN git clone https://github.com/open-mmlab/mmdetection3d.git /mmdetection3d
WORKDIR /mmdetection3d
ENV FORCE_CUDA="1"
RUN pip install -r requirements/build.txt
#RUN pip uninstall mmcv-ful -y && pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.6.0/mmcv_full-1.6.0-cp38-cp38-manylinux1_x86_64.whl
RUN pip uninstall mmcv-ful -y && pip install mmcv-full==1.6.0   pillow==8.3.2   ipython==7.23.1   traitlets==5.2.2
RUN pip install --no-cache-dir -e .

cuda支持的算力要高于实际的显卡算力——软件支持算力高于硬件
修改setup.py文件里的TORCH_CUDA_ARCH_LIST，将这个值改成你当前使用的GPU对应算力，可以使cuda（软件算力）降到硬件范围

使用现成的镜像

 docker search mmdetection

挑选了这个朋友的镜像，pulling…

docker pull  hquan7395/mmdetection3d-lab

run起来，/home/{共享路径}/（本地）:/mmdetection3d/data（docker容器）

sudo docker run --gpus all --shm-size=8g -it -v /home/{共享路径}/:/mmdetection3d/data hquan7395/mmdetection3d-lab

测试mm3d（需要下载模型权重文件）

python demo/pcd_demo.py demo/data/kitti_000008.bin configs/second/hv_second_secfpn_6x8_80e_kitti-3d-car.py checkpoints/hv_second_secfpn_6x8_80e_kitti-3d-car_20200620_230238-393f000c.pth

通过共享路径将结果文件夹传出来
show

ps:这是在自己的1050ti笔记本上测试的