创建MindSpore虚拟环境
- 创建虚拟环境并安装依赖库
conda create -n mindspore python=3.7.5 cudatoolkit=10.1 cudnn=7.6.5 gmp=6.1.2 nccl openmpi
或者分步安装:
conda create -n mindspore python=3.7.5
conda activate mindspore
conda install cudatoolkit=10.1 cudnn=7.6.5
conda install gmp=6.1.2
conda install nccl
conda install openmpi
打印环境所有安装的库:
conda list
# packages in environment at /home/devil/anaconda3/envs/mindspore: # # Name Version Build Channel _libgcc_mutex 0.1 main _openmp_mutex 4.5 1_gnu asttokens 2.0.5 pypi_0 pypi astunparse 1.6.3 pypi_0 pypi ca-certificates 2021.5.25 h06a4308_1 certifi 2021.5.30 py37h06a4308_0 cffi 1.14.5 pypi_0 pypi cudatoolkit 10.1.243 h6bb024c_0 cudnn 7.6.5 cuda10.1_0 decorator 5.0.9 pypi_0 pypi easydict 1.9 pypi_0 pypi gmp 6.1.2 h6c8ec71_1 libedit 3.1.20210216 h27cfd23_1 libffi 3.2.1 hf484d3e_1007 libgcc-ng 9.3.0 h5101ec6_17 libgfortran-ng 7.5.0 ha8ba4b0_17 libgfortran4 7.5.0 ha8ba4b0_17 libgomp 9.3.0 h5101ec6_17 libstdcxx-ng 9.3.0 hd4cf53a_17 mindspore-gpu 1.2.1 pypi_0 pypi mpi 1.0 openmpi mpmath 1.2.1 pypi_0 pypi nccl 2.8.3.1 hcaf9a05_0 ncurses 6.2 he6710b0_1 numpy 1.21.0 pypi_0 pypi openmpi 4.0.2 hb1b8bf9_1 openssl 1.1.1k h27cfd23_0 packaging 21.0 pypi_0 pypi pillow 8.3.0 pypi_0 pypi pip 21.1.3 py37h06a4308_0 protobuf 3.17.3 pypi_0 pypi psutil 5.8.0 pypi_0 pypi pycparser 2.20 pypi_0 pypi pyparsing 2.4.7 pypi_0 pypi python 3.7.5 h0371630_0 readline 7.0 h7b6447c_5 scipy 1.7.0 pypi_0 pypi setuptools 52.0.0 py37h06a4308_0 six 1.16.0 pypi_0 pypi sqlite 3.33.0 h62c20be_0 sympy 1.8 pypi_0 pypi tk 8.6.10 hbc83047_0 wheel 0.36.2 pyhd3eb1b0_0 xz 5.2.5 h7b6447c_0 zlib 1.2.11 h7b6447c_3
所安装的依赖软件库和官方给出的有一定差别,但是后面验证发现可以正常使用,因此这样安装是完全可以的。
具体说明,参考:https://zhuanlan.zhihu.com/p/364284533
为 cuda 和 cudnn 配置环境路径:
本人使用anaconda3创建的Python环境地址为:
/home/devil/anaconda3/envs/mindspore/
在 anaconda3中配置环境:
创建文件夹 etc/conda/activate.d :
mkdir -p etc/conda/activate.d
配置进入虚拟环境后加入的环境变量:
vim /home/devil/anaconda3/envs/mindspore/etc/conda/activate.d/env_vars.sh
配置内容:
# add library path export LD_LIBRARY_PATH=/home/devil/anaconda3/envs/mindspore/lib:$LD_LIBRARY_PATH # then, add system path export PATH=/home/devil/anaconda3/envs/mindspore/bin:$PATH # you should modify the code as: # export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/{your_path_to_install_conda}/envs/{your_virtual_env_name}/lib # export PATH=$PATH:/{your_path_to_install_conda}/envs/{your_virtual_env_name}/bin
退出环境,重新进入:
conda deactivate mindspore
conda activate mindspore
测试是否安装配置成功:
测试文件:
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))
成功运行,证明虽然安装的软件版本与官方的有略微差别但是其兼容性还是不影响code的运行的。
==============================================================
参考: