Ubuntu下自动安装MindSpore1.7GPU版

https://www.mindspore.cn/install

(1)前置安装

  • 确认是安装 Ubuntu 是x86架构64位操作系统。
  • 安装Minicanda或者Anaconda。
  • 安装Python 环境 3.7.5 或3.9.0(如何使用Conda安装可以直接创建命令即可,如果手动安装则需要配置Python环境变量)
  • 安装 Cuda10.1/11.1 驱动、GCC、glibc、OpenSSL
  • 查看GPU驱动
nvidia-smi

image-20220619222854958.png

(2)自动安装

  • 下载脚本
wget https://gitee.com/mindspore/mindspore/raw/r1.7/scripts/install/ubuntu-gpu-pip.sh

image-20220619220842049.png

  • 编辑脚本
vim ubuntu-gpu-pip.sh

该脚本会执行以下操作:

  • 更改软件源配置为华为云源。
  • 安装MindSpore所需的依赖,如GCC,gmp。
  • 通过APT安装Python3和pip3,并设为默认。
  • 下载CUDA和cuDNN并安装。
  • 通过pip安装MindSpore GPU版本。
  • 如果OPENMPI设置为on,则安装Open MPI。

自动安装脚本执行完成后,需要重新打开终端窗口以使环境变量生效。

  • 执行脚本,完成安装
# 安装MindSpore 1.7.0,Python 3.7和CUDA 10.1
MINDSPORE_VERSION=1.7.0 CUDA_VERSION=10.1 bash -i ./ubuntu-gpu-pip.sh

image-20220619221631399.png

  • 安装完成

image-20220619224009745.png

  • 安装其他版本,修改参数
# 如需指定安装Python 3.9,CUDA 10.1以及MindSpore 1.6.0,使用以下方式
# PYTHON_VERSION=3.9 CUDA_VERSION=10.1 MINDSPORE_VERSION=1.6.0 bash -i ./ubuntu-gpu-pip.sh
  • 完整脚本学习
set -e

PYTHON_VERSION=${PYTHON_VERSION:-3.7}
MINDSPORE_VERSION=${MINDSPORE_VERSION:EMPTY}
CUDA_VERSION=${CUDA_VERSION:-11.1}
OPENMPI=${OPENMPI:-off}

version_less() {
    test "$(echo "$@" | tr ' ' '\n' | sort -rV | head -n 1)" != "$1";
}

if [ $MINDSPORE_VERSION == "EMPTY" ] || version_less "${MINDSPORE_VERSION}" "1.6.0"; then
    echo "MINDSPORE_VERSION should be >=1.6.0, please check available versions at https://www.mindspore.cn/versions."
    exit 1
fi

available_py_version=(3.7 3.8 3.9)
if [[ " ${available_py_version<li>} " != *" $PYTHON_VERSION "* ]]; then
    echo "PYTHON_VERSION is '$PYTHON_VERSION', but available versions are [${available_py_version<li>}]."
    exit 1
fi

if [[ "$PYTHON_VERSION" == "3.8" && ${MINDSPORE_VERSION:0:3} == "1.6" ]]; then
    echo "PYTHON_VERSION==3.8 is not compatible with MINDSPORE_VERSION==1.6.x, please use PYTHON_VERSION==3.7 or 3.9 for MINDSPORE_VERSION==1.6.x."
    exit 1
fi

available_cuda_version=(10.1 11.1)
if [[ " ${available_cuda_version<li>} " != *" $CUDA_VERSION "* ]]; then
    echo "CUDA_VERSION is '$CUDA_VERSION', but available versions are [${available_cuda_version<li>}]."
    exit 1
fi
declare -A minimum_driver_version_map=()
minimum_driver_version_map["10.1"]="418.39"
minimum_driver_version_map["11.1"]="450.80.02"
driver_version=$(modinfo nvidia | grep ^version | awk '{printf $2}')
if [[ $driver_version < ${minimum_driver_version_map[$CUDA_VERSION]} ]]; then
    echo "CUDA $CUDA_VERSION minimum required driver version is ${minimum_driver_version_map[$CUDA_VERSION]}, \
        but current nvidia driver version is $driver_version, please upgrade your driver manually."
    exit 1
fi
cuda_name="cuda-$CUDA_VERSION"

declare -A version_map=()
version_map["3.7"]="${MINDSPORE_VERSION/-/}-cp37-cp37m"
version_map["3.8"]="${MINDSPORE_VERSION/-/}-cp38-cp38"
version_map["3.9"]="${MINDSPORE_VERSION/-/}-cp39-cp39"

# add value to environment variable if value is not in it
add_env() {
    local name=$1
    if [[ ":${!name}:" != *":$2:"* ]]; then
        echo -e "export $1=$2:\$$1" >> ~/.bashrc
    fi
}

# use huaweicloud mirror in China
sudo sed -i "s@http://.*archive.ubuntu.com@http://repo.huaweicloud.com@g" /etc/apt/sources.list
sudo sed -i "s@http://.*security.ubuntu.com@http://repo.huaweicloud.com@g" /etc/apt/sources.list
sudo apt-get update

sudo apt-get install curl make gcc-7 libgmp-dev linux-headers-"$(uname -r)" -y

# python
sudo add-apt-repository -y ppa:deadsnakes/ppa
sudo apt-get install python$PYTHON_VERSION python$PYTHON_VERSION-distutils python3-pip -y
sudo update-alternatives --install /usr/bin/python python /usr/bin/python$PYTHON_VERSION 100
# pip
python -m pip install -U pip -i https://pypi.tuna.tsinghua.edu.cn/simple
echo -e "alias pip='python -m pip'" >> ~/.bashrc
python -m pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple

# install cuda/cudnn
echo "installing CUDA and cuDNN"
cd /tmp
declare -A cuda_url_map=()
cuda_url_map["10.1"]=https://developer.download.nvidia.cn/compute/cuda/10.1/Prod/local_installers/cuda_10.1.243_418.87.00_linux.run
cuda_url_map["11.1"]=https://developer.download.nvidia.cn/compute/cuda/11.1.1/local_installers/cuda_11.1.1_455.32.00_linux.run
cuda_url=${cuda_url_map[$CUDA_VERSION]}
wget $cuda_url
sudo sh ${cuda_url##*/} --silent --toolkit
cd -
sudo apt-key adv --fetch-keys https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
sudo add-apt-repository "deb https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu1804/x86_64/ /"
sudo add-apt-repository "deb https://developer.download.nvidia.cn/compute/machine-learning/repos/ubuntu1804/x86_64/ /"
sudo apt-get update
declare -A cudnn_name_map=()
cudnn_name_map["10.1"]="libcudnn7=7.6.5.32-1+cuda10.1 libcudnn7-dev=7.6.5.32-1+cuda10.1"
cudnn_name_map["11.1"]="libcudnn8=8.0.4.30-1+cuda11.1 libcudnn8-dev=8.0.4.30-1+cuda11.1"
sudo apt-get install --no-install-recommends ${cudnn_name_map[$CUDA_VERSION]} -y

# add cuda to path
set +e && source ~/.bashrc
set -e
add_env PATH /usr/local/cuda/bin
add_env LD_LIBRARY_PATH /usr/local/cuda/lib64
add_env LD_LIBRARY_PATH /usr/lib/x86_64-linux-gnu
set +e && source ~/.bashrc
set -e

# optional openmpi for distributed training
if [[ X"$OPENMPI" == "Xon" ]]; then
    echo "installing openmpi"
    cd /tmp
    curl -O https://download.open-mpi.org/release/open-mpi/v4.0/openmpi-4.0.3.tar.gz
    tar xzf openmpi-4.0.3.tar.gz
    cd openmpi-4.0.3
    ./configure --prefix=/usr/local/openmpi-4.0.3
    make
    sudo make install
    add_env PATH /usr/local/openmpi-4.0.3/bin
    add_env LD_LIBRARY_PATH /usr/local/openmpi-4.0.3/lib
fi

arch=`uname -m`
python -m pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/${MINDSPORE_VERSION}/MindSpore/gpu/${arch}/${cuda_name}/mindspore_gpu-${version_map["$PYTHON_VERSION"]}-linux_${arch}.whl --trusted-host ms-release.obs.cn-north-4.myhuaweicloud.com -i https://pypi.tuna.tsinghua.edu.cn/simple

# check mindspore installation
python -c "import mindspore;mindspore.run_check()"

# check if it can be run with GPU
cd /tmp
cat > example.py <<END
import numpy as np
from mindspore import Tensor
import mindspore.ops as ops
import mindspore.context as context

context.set_context(device_target="GPU")
x = Tensor(np.ones([1,3,3,4]).astype(np.float32))
y = Tensor(np.ones([1,3,3,4]).astype(np.float32))
print(ops.add(x, y))
END
python example.py
cd -
posted @ 2022-07-17 17:43  Skytier  阅读(77)  评论(0编辑  收藏  举报