【CANN安装】昇腾驱动固件+CANN昇腾社区版+mindspore安装

[jianyunchao]

910逻辑，没有固件驱动，从头开始操作

已经安装好固件驱动，从这里开始： https://github.com/jianyunchao/mindlearning/issues/5#issuecomment-1670808750

[jianyunchao]

官网安装教程：

https://www.mindspore.cn/install

安装昇腾AI处理器配套软件包章节中点击对应超链接跳转昇腾下载固件、驱动与CANN包

mindspore 2.0.0的cann社区版为例：

• 固件和驱动

点击“固件和驱动”超链接，搜索6.3.RC1.alpha003版本下载，我这边执行arch是aarch64操作系统，对应是服务器的9000，下载对应driver和firmware的run包即可 https://www.hiascend.com/hardware/firmware-drivers?tag=community

• CANN社区版

点击“CANN社区版”超链接，搜索6.3.RC1.alpha003版本下载，我的需求是910训练，下载对应toolkit的run包 https://www.hiascend.com/software/cann/community-history

[jianyunchao]

在root用户下，安装driver

首次安装，新增HwHiAiUser用户组和用户 groupadd HwHiAiUser useradd -g HwHiAiUser -d /home/HwHiAiUser -m HwHiAiUser -s /bin/bash

当出现 Reboot now or after driver installation for the installation/upgrade to take effect. 需要reboot重启

[jianyunchao]

在root用户下，安装firmware

当出现 Reboot now or after driver installation for the installation/upgrade to take effect. 需要reboot重启

安装完成driver和firmware，执行命令npu-smi info验证是否成功

[jianyunchao]

root登录，安装toolkit包

[jianyunchao]

常见安装问题

• 登录其他普通用户，再su切换root会报错

[Toolkit] [2023-08-09 07:07:28] [ERROR]: usergroup=root not right! Please check the relatianship of group1 and root [Toolkit] [2023-08-09 07:07:28] [INFO]: End Time: 2023-08-09 07:07:28 [Toolkit] [20230809-07:07:29] [WARNING] delete operation, the /usr/local/Ascend/ascend-toolkit is not exist or not directory. [Toolkit] [20230809-07:07:29] [ERROR] CANN-toolkit-6.4.12.1.241-linux.aarch64.run install failed [Toolkit] [20230809-07:07:29] [ERROR] package install failed, Please refer to the log for more details: /var/log/ascend_seclog/ascend_toolkit_install.log [Toolkit] [20230809-07:07:29] [ERROR] process exit

解决方法： 直接用root登录操作

• 缺少HwHiAiUser 用户组

解决方法： groupadd HwHiAiUser useradd -g HwHiAiUser -d /home/HwHiAiUser -m HwHiAiUser -s /bin/bash

• 缺少ifconfig工具报错：[ERROR]The list of missing tools: ifconfig,

解决方法： apt install net-tools

• 系统内核版本不匹配，需要重建编译内核

解决方法： apt install build-essential -y apt install cmake -y apt install dkms -y

[jianyunchao]

安装minicoda

下载官网对应系统的miniconda 直接执行 bash Miniconda3-latest-Linux-aarch64.sh 一直enter下一步，yes，使用默认安装路径

激活环境变量 source ~/.bashrc

新建conda环境 conda create -n ms21_py38 python=3.8 -y

[jianyunchao]

安装配置mindspore

https://mindspore.cn/install

• 配置pip镜像源为清华源

执行命令： mkdir ~/.pip vim ~/.pip/pip.conf 写入： [global] index-url = https://pypi.tuna.tsinghua.edu.cn/simple [install] trusted-host = https://pypi.tuna.tsinghua.edu.cn/ timeout = 120

• 升级pip

python -m pip install -U pip

• 安装依赖whl包：

pip install sympy pip install /usr/local/Ascend/ascend-toolkit/latest/lib64/te--py3-none-any.whl pip install /usr/local/Ascend/ascend-toolkit/latest/lib64/hccl--py3-none-any.whl

• 安装mindspore

export MS_VERSION=2.1.0 pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/${MS_VERSION}/MindSpore/unified/aarch64/mindspore-${MS_VERSION/-/}-cp37-cp37m-linux_aarch64.whl --trusted-host ms-release.obs.cn-north-4.myhuaweicloud.com -i https://pypi.tuna.tsinghua.edu.cn/simple

• 验证是否安装成功：

python -c "import mindspore;mindspore.set_context(device_target='Ascend');mindspore.run_check()"