在BML平台使用容器镜像服务CCR
更新时间：2023-07-26
在BML平台使用容器镜像服务CCR
平台支持用户在用户资源池上关联容器镜像服务CCR作为资源池的镜像仓库，在使用用户资源池提交任务时，可以使用镜像仓库中的镜像。
当前支持的容器镜像服务CCR类型：
- 容器镜像服务CCR-企业版
- 容器镜像服务CCR-个人版
当前支持使用容器镜像服务CCR提交的任务：
- 自定义作业-训练作业任务、自动搜索作业任务
前提条件
- 用户在平台上已经挂载了容器引擎CCE资源作为用户资源池，点击了解容器引擎CCE；
- 用户已经创建了容器镜像服务CCR，点击了解容器镜像服务CCR。
- 容器镜像服务CCR能够被容器引擎CCE资源访问到，也即能被对应的VPC访问到。
创建镜像仓库
- Step1:进入平台管理-资源池管理，已挂载并运行正常的用户资源池支持“镜像仓库”的操作项，点击即可开始查看镜像仓库。
- Step2:点击镜像仓库，即可进入镜像仓库列表。
- Step3：点击添加镜像仓库，即可进入添加流程。
企业版：支持选择资源池对应区域和VPC下的，归属于主账号的容器镜像服务CCR-企业版的实例，并填写账号密码进行添加。
个人版：支持选择归属于主账号的容器镜像服务CCR-个人版的实例，并填写账号密码进行添加。
使用镜像提交自定义作业任务
在算法配置阶段，如果用户选择了用户资源池，即支持选择该资源池所关联的CCR镜像环境提交任务。（训练作业和自动搜索作业任务的提交过程一致）
- 企业版：支持依次选择镜像仓库-命名空间-镜像-版本，从而选中一个唯一确定的镜像用于提交任务。
- 个人版：支持依次选择镜像仓库-镜像-版本，从而选中一个唯一确定的镜像用于提交任务。
在选择完镜像后，需要根据镜像中的深度学习框架来选择分布式框架，其中：
- Paddlefleet：PaddleFleet是PaddlePaddle推出的分布式图引擎以及大规模参数服务器，用于支撑⻜桨框架大规模分布式训练能力。
- Horovod：分布式训练框架，支持tensorflow、pytorch、ray、mxnet等业界著名的开源的机器学习框架。通过对底层tensorflow等框架进行上层分布式调度、分布式通信、梯度计算等封装，完成大规模集群下模型的训练。horovod主要支持在GPU资源集群上的大规模分布式的同步训练。
后续填写代码文件、启动命令、输出路径等信息后，即可提交自定义作业任务。
附录：自定义镜像规范
Paddle镜像
- python版本要求：python3.7及以上
- 安装bos：dockerfile实现
RUN /bin/bash -c 'mkdir /home/bos && \
cd /home/bos && \
wget --no-check-certificate https://sdk.bce.baidu.com/console-sdk/linux-bcecmd-0.3.0.zip && \
unzip linux-bcecmd-0.3.0.zip && \
echo "export LANG=en_US.UTF-8" >> ~/.bashrc && \
echo "export PATH="/home/bos/linux-bcecmd-0.3.0:${PATH}"" >> ~/.bashrc && \
source ~/.bashrc'或者自行在镜像中安装到/home/bos/linux-bcecmd-0.3.0目录下，确保 “/home/bos/linux-bcecmd-0.3.0/bcecmd” 这句命令在命令行能弹出相关bcecmd帮助信息，即命令能被系统识别
- 安装搜索需要的 sdk
- 用户自行安装wheel包：rudder_autosearch-1.0.0-py3-none-any.whl，点击下载。
- 安装 jq 解析sts需要：apt-get install jq
- 安装 curl 解析sts需要：apt-get install curl
- 安装protobuf 搜索sdk需要：建议安装3.20.1版本 pip install protobuf==3.20.1
- 设置虚拟化环境变量初始值：dockerfile实现
ENV NVIDIA_VISIBLE_DEVICES ""Sklearn镜像
FROM ubuntu16.04-python3
# Configure time zone
RUN apt-get update && \
apt-get install -y tzdata && \
ln -snf /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && \
dpkg-reconfigure -f noninteractive tzdata && \
apt-get clean
RUN apt-get install -y --no-install-recommends\
build-essential \
libopencv-dev \
libssl-dev \
dnsutils \
unzip \
vim \
jq \
curl \
wget && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
#py3
RUN bash -c 'cd /tmp && \
wget --no-check-certificate https://repo.anaconda.com/miniconda/Miniconda3-py37_4.8.3-Linux-x86_64.sh && \
bash Miniconda3-py37_4.8.3-Linux-x86_64.sh -b -p ~/miniconda3 && \
echo "source ~/miniconda3/bin/activate" >> ~/.bashrc && \
echo "export PATH="~/miniconda3/bin:${PATH}"" >> ~/.bashrc && \
source ~/.bashrc && \
rm -rf Miniconda3-py37_4.8.3-Linux-x86_64.sh' && \
/root/miniconda3/bin/python -m pip config set global.index-url https://pypi.douban.com/simple/ && \
/root/miniconda3/bin/python -m pip install --upgrade pip && \
/root/miniconda3/bin/python -m pip install --upgrade setuptools && \
/root/miniconda3/bin/python -m pip install numpy==1.17.4 && \
/root/miniconda3/bin/python -m pip install albumentations==0.4.3 && \
/root/miniconda3/bin/python -m pip install Cython==0.29.16 && \
/root/miniconda3/bin/python -m pip install pycocotools==2.0.0 && \
/root/miniconda3/bin/python -m pip install ruamel.yaml && \
/root/miniconda3/bin/python -m pip install ujson && \
/root/miniconda3/bin/python -m pip install scipy==1.5.3 && \
/root/miniconda3/bin/python -m pip install scikit-learn==0.23.2 && \
/root/miniconda3/bin/python -m pip install pandas
RUN /root/miniconda3/condabin/conda clean -p && \
/root/miniconda3/condabin/conda clean -t
RUN rm -rf ~/.cache/pip
RUN rm -rf /usr/bin/python3 && ln -s /root/miniconda3/bin/python /usr/bin/python3
#sklearn
RUN /root/miniconda3/bin/python -m pip install xgboost==1.3.1
#安装bos
RUN /bin/bash -c 'mkdir /home/bos && \
cd /home/bos && \
wget --no-check-certificate https://sdk.bce.baidu.com/console-sdk/linux-bcecmd-0.3.0.zip && \
unzip linux-bcecmd-0.3.0.zip && \
echo "export LANG=en_US.UTF-8" >> ~/.bashrc && \
echo "export PATH="/home/bos/linux-bcecmd-0.3.0:${PATH}"" >> ~/.bashrc && \
source ~/.bashrc'
#添加搜索作业SDK install 3.20.1 protobuf for searchjob
COPY rudder-autosearch-1.0.0-py3-none-any.whl /home/rudder-autosearch-1.0.0-py3-none-any.whl
RUN pip install /home/rudder-autosearch-1.0.0-py3-none-any.whl && \
pip install protobuf==3.20.1
ENV NVIDIA_VISIBLE_DEVICES ""
ENTRYPOINT ["/bin/bash"]如果是pytorch/tensorflow的多机分布式作业，则需要安装额外的依赖包
- 安装openmpi
#安装openmpi
RUN mkdir /tmp/openmpi && \
cd /tmp/openmpi && \
wget https://download.open-mpi.org/release/open-mpi/v4.0/openmpi-4.0.0.tar.gz && \
tar zxf openmpi-4.0.0.tar.gz && \
cd openmpi-4.0.0 && \
./configure --enable-orterun-prefix-by-default && \
make -j $(nproc) all && \
make install && \
ldconfig && \
rm -rf /tmp/openmpi
# Create a wrapper for OpenMPI to allow running as root by default
RUN mv /usr/local/bin/mpirun /usr/local/bin/mpirun.real && \
echo '#!/bin/bash' > /usr/local/bin/mpirun && \
echo 'mpirun.real --allow-run-as-root "$@"' >> /usr/local/bin/mpirun && \
chmod a+x /usr/local/bin/mpirun
# Configure OpenMPI to run good defaults:
# --bind-to none --map-by slot --mca btl_tcp_if_exclude lo,docker0
RUN echo "hwloc_base_binding_policy = none" >> /usr/local/etc/openmpi-mca-params.conf && \
echo "rmaps_base_mapping_policy = slot" >> /usr/local/etc/openmpi-mca-params.conf && \
echo "btl_tcp_if_exclude = lo,docker0" >> /usr/local/etc/openmpi-mca-params.conf- 安装horvod（MPI形式的作业，需要安装horovod，并进行相关配置。horovod安装需要cmake3.13+，可以apt-get install 指定版本安装）
RUN ldconfig /usr/local/cuda/targets/x86_64-linux/lib/stubs && \
HOROVOD_GPU_ALLREDUCE=NCCL HOROVOD_WITH_PYTORCH=1 pip install --no-cache-dir horovod && \
ldconfig- 安装并配置nccl
# Set default NCCL parameters
RUN echo NCCL_DEBUG=INFO >> /etc/nccl.conf- 安装openssh并配置
# Install OpenSSH for MPI to communicate between containers
RUN apt-get install -y --no-install-recommends openssh-client openssh-server && \
mkdir -p /var/run/sshd
# Allow OpenSSH to talk to containers without asking for confirmation
RUN cat /etc/ssh/ssh_config | grep -v StrictHostKeyChecking > /etc/ssh/ssh_config.new && \
echo " StrictHostKeyChecking no" >> /etc/ssh/ssh_config.new && \
mv /etc/ssh/ssh_config.new /etc/ssh/ssh_configpytorch/tf 镜像示例
FROM cuda11.0.3-cudnn8-devel-ubuntu18.04
# Configure time zone
RUN rm /etc/apt/sources.list.d/cuda.list && rm /etc/apt/sources.list.d/nvidia-ml.list
RUN apt-get update && \
DEBIAN_FRONTEND="noninteractive" TZ="Asia/Shanghai" \
apt-get install -y tzdata && \
ln -snf /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && \
dpkg-reconfigure -f noninteractive tzdata && \
apt-get clean
#cmake需要大于3.13版本
RUN apt-get install -y \
build-essential \
libopencv-dev \
libssl-dev \
dnsutils \
unzip \
vim \
git \
jq \
curl \
cmake \
wget \
ca-certificates \
libjpeg-dev \
libpng-dev \
wget && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
#py3
RUN bash -c 'cd /tmp && \
wget --no-check-certificate https://repo.anaconda.com/miniconda/Miniconda3-py37_4.8.3-Linux-x86_64.sh && \
bash Miniconda3-py37_4.8.3-Linux-x86_64.sh -b -p ~/miniconda3 && \
echo "source ~/miniconda3/bin/activate" >> ~/.bashrc && \
echo "export PATH="~/miniconda3/bin:${PATH}"" >> ~/.bashrc && \
source ~/.bashrc && \
rm -rf Miniconda3-py37_4.8.3-Linux-x86_64.sh' && \
/root/miniconda3/bin/python -m pip config set global.index-url https://pypi.douban.com/simple/ && \
/root/miniconda3/bin/python -m pip install --upgrade pip && \
/root/miniconda3/bin/python -m pip install --upgrade setuptools && \
/root/miniconda3/bin/python -m pip install numpy==1.17.4 && \
/root/miniconda3/bin/python -m pip install albumentations==0.4.3 && \
/root/miniconda3/bin/python -m pip install Cython==0.29.16 && \
/root/miniconda3/bin/python -m pip install pycocotools==2.0.0 && \
/root/miniconda3/bin/python -m pip install ruamel.yaml && \
/root/miniconda3/bin/python -m pip install ujson && \
/root/miniconda3/bin/python -m pip install scikit-learn==0.23.2 && \
/root/miniconda3/bin/python -m pip install pandas
RUN /root/miniconda3/condabin/conda clean -p && \
/root/miniconda3/condabin/conda clean -t
RUN rm -rf ~/.cache/pip
RUN rm -rf /usr/bin/python3 && ln -s /root/miniconda3/bin/python /usr/bin/python3
#torch
RUN /root/miniconda3/bin/python -m pip install torch==1.7.1+cu110 torchvision==0.8.2+cu110 torchaudio===0.7.2 \
-f https://download.pytorch.org/whl/torch_stable.html
# Install Open MPI 4.0.0
RUN mkdir /tmp/openmpi && \
cd /tmp/openmpi && \
wget https://download.open-mpi.org/release/open-mpi/v4.0/openmpi-4.0.0.tar.gz && \
tar zxf openmpi-4.0.0.tar.gz && \
cd openmpi-4.0.0 && \
./configure --enable-orterun-prefix-by-default && \
make -j $(nproc) all && \
make install && \
ldconfig && \
rm -rf /tmp/openmpi
# Install Horovod, temporarily using CUDA stubs
# /usr/local/cuda links to /usr/local/cuda-10.1
RUN ldconfig /usr/local/cuda/targets/x86_64-linux/lib/stubs && \
HOROVOD_GPU_ALLREDUCE=NCCL HOROVOD_WITH_PYTORCH=1 pip install --no-cache-dir horovod && \
ldconfig
# Create a wrapper for OpenMPI to allow running as root by default
RUN mv /usr/local/bin/mpirun /usr/local/bin/mpirun.real && \
echo '#!/bin/bash' > /usr/local/bin/mpirun && \
echo 'mpirun.real --allow-run-as-root "$@"' >> /usr/local/bin/mpirun && \
chmod a+x /usr/local/bin/mpirun
# Configure OpenMPI to run good defaults:
# --bind-to none --map-by slot --mca btl_tcp_if_exclude lo,docker0
RUN echo "hwloc_base_binding_policy = none" >> /usr/local/etc/openmpi-mca-params.conf && \
echo "rmaps_base_mapping_policy = slot" >> /usr/local/etc/openmpi-mca-params.conf && \
echo "btl_tcp_if_exclude = lo,docker0" >> /usr/local/etc/openmpi-mca-params.conf
# Set default NCCL parameters
RUN echo NCCL_DEBUG=INFO >> /etc/nccl.conf
# Install OpenSSH for MPI to communicate between containers
RUN apt-get install -y --no-install-recommends openssh-client openssh-server && \
mkdir -p /var/run/sshd
# Allow OpenSSH to talk to containers without asking for confirmation
RUN cat /etc/ssh/ssh_config | grep -v StrictHostKeyChecking > /etc/ssh/ssh_config.new && \
echo " StrictHostKeyChecking no" >> /etc/ssh/ssh_config.new && \
mv /etc/ssh/ssh_config.new /etc/ssh/ssh_config
#安装bos
RUN /bin/bash -c 'mkdir /home/bos && \
cd /home/bos && \
wget --no-check-certificate https://sdk.bce.baidu.com/console-sdk/linux-bcecmd-0.3.0.zip && \
unzip linux-bcecmd-0.3.0.zip && \
echo "export LANG=en_US.UTF-8" >> ~/.bashrc && \
echo "export PATH="/home/bos/linux-bcecmd-0.3.0:${PATH}"" >> ~/.bashrc && \
source ~/.bashrc'
#添加搜索作业SDK install 3.20.1 protobuf for searchjob
COPY rudder-autosearch-1.0.0-py3-none-any.whl /home/rudder-autosearch-1.0.0-py3-none-any.whl
RUN pip install /home/rudder-autosearch-1.0.0-py3-none-any.whl && \
pip install protobuf==3.20.1
ENV NVIDIA_VISIBLE_DEVICES ""
ENTRYPOINT ["/bin/bash"]