華爲人工智能atlasA800-9000物理服務器離線安裝及CANN安裝和MindSpore安裝和Tensorflow安裝

目錄

華爲人工智能atlas A800-9000 物理服務器全程離線安裝驅動以及CANN安裝部署和MindSpore安裝部署和Tensorflow安裝部署

A800-9000 物理服務器安裝驅動

使用鏡像配置本地apt源

建立普通用戶並設置密碼

安裝驅動以及固件

驗證是否安裝成功

CANN開發環境部署安裝

安裝環境以及依賴

安裝完成後查看版本

安裝Python3.7.5

使用Python3.7.5環境安裝pip依賴包

安裝開發套件包

CANN訓練環境部署安裝

說明

安裝訓練軟件包

安裝MindSpore

安裝whl包

配置環境變量

測試是否可行

安裝mindinsight

安裝whl包

配置環境變量

啓動及使用

安裝Tensorflow

編譯hdf5

配置環境變量及軟鏈接

安裝whl包

安裝Pytorch

華爲人工智能atlas A800-9000 物理服務器全程離線安裝驅動以及CANN安裝部署和MindSpore安裝部署和Tensorflow安裝部署

背景

Atlas 800 訓練服務器（型號：9000）是基於華爲鯤鵬920+昇騰910處理器的AI訓練服務器，具備最強算力密度、超高能效與高速網絡帶寬等特色。該服務器普遍應用於深度學習模型開發和訓練，適用於智慧城市、智慧醫療、天文探索、石油勘探等須要大算力的行業領域。

連接：

https://e.huawei.com/cn/products/cloud-computing-dc/atlas/atlas-800-training-9000

CANN (Compute Architecture for Neural Networks)

是華爲公司針對AI場景推出的異構計算架構，經過提供多層次的編程接口，支持用戶快速構建基於昇騰平臺的AI應用和業務。

連接：

https://e.huawei.com/cn/products/cloud-computing-dc/atlas/cann

MindSpore，新一代AI開源計算框架。

創新編程範式，AI科學家和工程師更易使用，便於開放式創新；該計算框架可知足終端、邊緣計算、雲全場景需求，能更好保護數據隱私；可開源，造成廣闊應用生態。

連接：

https://www.mindspore.cn/

TensorFlow最初由谷歌大腦團隊開發，用於Google的研究和生產，於2015年11月9日在Apache 2.0開源許可證下發布。

連接：

https://www.tensorflow.org/

A800-9000物理服務器安裝驅動

使用鏡像配置本地apt源

root@ubuntu:/etc/apt# mkdir/media/cdrom
root@ubuntu:/etc/apt# mount/home/cby/ubuntu-18.04.5-server-arm64.iso /media/cdrom
mount: /media/chrom: WARNING:device write-protected, mounted read-only.
root@ubuntu:/etc/apt# apt-cdrom-m -d=/media/cdrom/ add
root@ubuntu:/etc/apt# cat/etc/apt/sources.list

建立普通用戶並設置密碼

root@ubuntu:~#groupadd HwHiAiUser
root@ubuntu:~#useradd -g HwHiAiUser -d /home/HwHiAiUser -m HwHiAiUser
root@ubuntu:~#passwd HwHiAiUser
Enter newUNIX password:
Retype newUNIX password:
passwd:password updated successfully

安裝驅動以及固件

root@ubuntu:~# cd /home/cby/
root@ubuntu:/home/cby# ll
total 98324
drwxr-xr-x 4 cby  cby     4096 Apr 21 21:41 ./
drwxr-xr-x 4 root root     4096 Apr 21 21:44 ../
-rw-r--r-- 1 cby  cby 99728721 Apr 21 21:41A800-9000-npu-driver\_20.2.0\_ubuntu18.04-aarch64.run
-rw-r--r-- 1 cby  cby   912335 Apr 21 21:41 A800-9000-npu-firmware\_1.76.22.3.220.run
root@ubuntu:/home/cby# chmod +x\*.run
root@ubuntu:/home/cby# ll
total 98324
drwxr-xr-x 4 cby  cby     4096 Apr 21 21:41 ./
drwxr-xr-x 4 root root     4096 Apr 21 21:44 ../
-rwxr-xr-x 1 cby  cby 99728721 Apr 21 21:41 A800-9000-npu-driver\_20.2.0\_ubuntu18.04-aarch64.run\*
-rwxr-xr-x 1 cby  cby   912335 Apr 21 21:41 A800-9000-npu-firmware\_1.76.22.3.220.run\*
root@ubuntu:/home/cby# aptinstall gcc
root@ubuntu:/home/cby# aptinstall make
root@ubuntu:/home/cby#./A800-9000-npu-driver\_20.2.0\_ubuntu18.04-aarch64.run –run
root@ubuntu:/home/cby#./A800-9000-npu-firmware\_1.76.22.3.220.run –run

*注意：安裝完成後須要重啓服務器

驗證是否安裝成功

root@ubuntu:/home/cby#npu-smi info

CANN開發環境部署安裝

安裝環境以及依賴

root@ubuntu:/home/cby# apt install g++
root@ubuntu:/home/cby# cd cmake/
root@ubuntu:/home/cby/cmake# ll
total 4356
drwxr-xr-x 2 root root    4096 Apr 21 23:48 ./
drwxr-xr-x 7 cby  cby    4096 Apr 21 23:48 ../
-rw-r--r-- 1 cby  cby 2971248 Apr 21 23:45 cmake\_3.10.2-1ubuntu2.18.04.1\_arm64.deb
-rw-r--r-- 1 cby  cby 1331524 Apr 21 23:45 cmake-data\_3.10.2-1ubuntu2.18.04.1\_all.deb
-rw-r--r-- 1 cby  cby   69166 Apr 21 23:47 libjsoncpp1\_1.7.4-3\_arm64.deb
-rw-r--r-- 1 cby  cby   71788 Apr 21 23:48 librhash0\_1.3.6-2\_arm64.deb
root@ubuntu:/home/cby/cmake# apt install./\*
root@ubuntu:/home/cby/cmake# make–version

root@ubuntu:/home/cby# apt install./zlib1g-dev\_1%3a1.2.11.dfsg-0ubuntu2\_arm64.deb
root@ubuntu:/home/cby# apt install./libbz2-dev\_1.0.6-8.1ubuntu0.2\_arm64.deb
root@ubuntu:/home/cby# apt install ./libsqlite3-dev\_3.22.0-1ubuntu0.4\_arm64.deb

root@ubuntu:/home/cby# cd libssl-dev/
root@ubuntu:/home/cby/libssl-dev#apt install ./\*

root@ubuntu:/home/cby# cd libxslt1-dev/
root@ubuntu:/home/cby/libxslt1-dev#ll
total 13596
drwxr-xr-x  2 root root   4096 Apr 22 00:37 ./
drwxr-xr-x 10 cby  cby    4096 Apr 22 00:37 ../
-rw-r--r--  1 cby cby    18528 Apr 22 00:30gir1.2-harfbuzz-0.0\_1.7.2-1ubuntu1\_arm64.deb
-rw-r--r--  1 cby cby   170204 Apr 22 00:27icu-devtools\_60.2-3ubuntu3.1\_arm64.deb
-rw-r--r--  1 cby cby   983364 Apr 22 00:37libglib2.0-0\_2.56.4-0ubuntu0.18.04.8\_arm64.deb
-rw-r--r--  1 cby cby    61832 Apr 22 00:33libglib2.0-bin\_2.56.4-0ubuntu0.18.04.8\_arm64.deb
-rw-r--r--  1 cby cby  1297600 Apr 22 00:31libglib2.0-dev\_2.56.4-0ubuntu0.18.04.8\_arm64.deb
-rw-r--r--  1 cby cby    99676 Apr 22 00:31libglib2.0-dev-bin\_2.56.4-0ubuntu0.18.04.8\_arm64.deb
-rw-r--r--  1 cby cby    14528 Apr 22 00:32libgraphite2-dev\_1.3.11-2\_arm64.deb
-rw-r--r--  1 cby cby   280584 Apr 22 00:28libharfbuzz-dev\_1.7.2-1ubuntu1\_arm64.deb
-rw-r--r--  1 cby cby    12556 Apr 22 00:30libharfbuzz-gobject0\_1.7.2-1ubuntu1\_arm64.deb
-rw-r--r--  1 cby cby     5348 Apr 22 00:29libharfbuzz-icu0\_1.7.2-1ubuntu1\_arm64.deb
-rw-r--r--  1 cby cby  8890124 Apr 22 00:26libicu-dev\_60.2-3ubuntu3.1\_arm64.deb
-rw-r--r--  1 cby cby    14412 Apr 22 00:28libicu-le-hb0\_1.0.3+git161113-4\_arm64.deb
-rw-r--r--  1 cby cby    29760 Apr 22 00:27libicu-le-hb-dev\_1.0.3+git161113-4\_arm64.deb
-rw-r--r--  1 cby cby    18756 Apr 22 00:26libiculx60\_60.2-3ubuntu3.1\_arm64.deb
-rw-r--r--  1 cby cby   120696 Apr 22 00:35libpcre16-3\_2%3a8.39-9\_arm64.deb
-rw-r--r--  1 cby cby   113240 Apr 22 00:35libpcre32-3\_2%3a8.39-9\_arm64.deb
-rw-r--r--  1 cby cby   459316 Apr 22 00:33libpcre3-dev\_2%3a8.39-9\_arm64.deb
-rw-r--r--  1 cby cby    15124 Apr 22 00:35libpcrecpp0v5\_2%3a8.39-9\_arm64.deb
-rw-r--r--  1 cby cby   673384 Apr 22 00:25libxml2-dev\_2.9.4+dfsg1-6.1ubuntu1.3\_arm64.deb
-rw-r--r--  1 cby cby   395564 Apr 22 00:24libxslt1-dev\_1.1.29-5ubuntu0.2\_arm64.deb
-rw-r--r--  1 cby cby    42802 Apr 22 00:33pkg-config\_0.29.1-0ubuntu2\_arm64.deb
-rw-r--r--  1 cby cby   144176 Apr 22 00:37python3-distutils\_3.6.9-1~18.04\_all.deb
root@ubuntu:/home/cby/libxslt1-dev#apt install ./\*

root@ubuntu:/home/cby# cd libffi-dev/
root@ubuntu:/home/cby/libffi-dev#ls
libffi-dev\_3.2.1-8\_arm64.deb
root@ubuntu:/home/cby/libffi-dev#apt install ./\*

root@ubuntu:/home/cby#apt install unzip
root@ubuntu:/home/cby# apt install./libblas-dev\_3.7.1-4ubuntu1\_arm64.deb

root@ubuntu:/home/cby# cd gfortran/
root@ubuntu:/home/cby/gfortran#ll
total 7844
drwxr-xr-x  2 root root   4096 Apr 22 00:50 ./
drwxr-xr-x 12cby  cby     4096 Apr 22 00:50 ../
-rw-r--r--  1 cby cby     1344 Apr 22 00:48gfortran\_4%3a7.4.0-1ubuntu2.3\_arm64.deb
-rw-r--r--  1 cby cby  7464740 Apr 22 00:48gfortran-7\_7.5.0-3ubuntu1~18.04\_arm64.deb
-rw-r--r--  1 cby cby   248176 Apr 22 00:50libgfortran4\_7.5.0-3ubuntu1~18.04\_arm64.deb
-rw-r--r--  1 cby cby   300500 Apr 22 00:49libgfortran-7-dev\_7.5.0-3ubuntu1~18.04\_arm64.deb
root@ubuntu:/home/cby/gfortran#apt install ./\*

root@ubuntu:/home/cby# cd libblas3/
root@ubuntu:/home/cby/libblas3#apt install ./libblas3\_3.7.1-4ubuntu1\_arm64.deb

root@ubuntu:/home/cby# cdlibopenblas-dev/
root@ubuntu:/home/cby/libopenblas-dev#ll
total 3412
drwxr-xr-x  2 root root   4096 Apr 22 00:56 ./
drwxr-xr-x 14cby  cby     4096 Apr 22 00:56 ../
-rw-r--r--  1 cby cby  1813748 Apr 22 00:55libopenblas-base\_0.2.20+ds-4\_arm64.deb
-rw-r--r--  1 cby cby  1668126 Apr 22 00:54libopenblas-dev\_0.2.20+ds-4\_arm64.deb
root@ubuntu:/home/cby/libopenblas-dev#apt install ./\*

安裝完成後查看版本

gcc --version
g++ --version
make --version
cmake --version
dpkg -l zlib1g| grepzlib1g| grep ii
dpkg -l zlib1g-dev|grep zlib1g-dev| grep ii
dpkg -l libbz2-dev|grep libbz2-dev| grep ii
dpkg -llibsqlite3-dev| grep libsqlite3-dev| grep ii
dpkg -l openssl| grepopenssl| grep ii
dpkg -l libssl-dev|grep libssl-dev| grep ii
dpkg -l libxslt1-dev|grep libxslt1-dev| grep ii
dpkg -l libffi-dev|grep libffi-dev| grep ii
dpkg -l unzip| grepunzip| grep ii
dpkg -l pciutils|grep pciutils| grep ii
dpkg -l net-tools|grep net-tools| grep ii
dpkg -l libblas-dev|grep libblas-dev| grep ii
dpkg -l gfortran|grep gfortran| grep ii
dpkg -l libblas3|grep libblas3| grep ii
dpkg -llibopenblas-dev| grep libopenblas-dev| grep ii

安裝Python3.7.5

root@ubuntu:/home/cby/python#tar xvf Python3.7.5.tar
root@ubuntu:/home/cby/python# cdPython-3.7.5/
root@ubuntu:/home/cby/python/Python-3.7.5#./configure --prefix=/usr/local/python3.7.5 --enable-loadable-sqlite-extensions--enable-shared
root@ubuntu:/home/cby/python/Python-3.7.5#make
root@ubuntu:/home/cby/python/Python-3.7.5#make install
 
root@ubuntu:/home/cby# sudo ln -s/usr/local/python3.7.5/bin/pip3 /usr/local/bin/pip3.7.5
root@ubuntu:/home/cby# sudo ln-s /usr/local/python3.7.5/bin/python3 /usr/local/bin/python3.7.5
root@ubuntu:/home/cby/cann\_xunlian#sudo ln -s /usr/local/python3.7.5/bin/python3 /usr/local/bin/python3.7
root@ubuntu:/home/cby/cann\_xunlian#sudo ln -s /usr/local/python3.7.5/bin/pip3 /usr/local/bin/pip3.7
 
root@ubuntu:/home/cby# vim ~/.bashrc
exportLD\_LIBRARY\_PATH=/usr/local/python3.7.5/lib:$LD\_LIBRARY\_PATH
      
root@ubuntu:/home/cby#python3.7.5 --version
Python 3.7.5
root@ubuntu:/home/cby# pip3.7.5--version
pip 19.2.3 from /usr/local/python3.7.5/lib/python3.7/site-packages/pip(python 3.7)

使用Python3.7.5環境安裝pip依賴包

root@ubuntu:/home/cby/pip-pack# tar xvfpip\_pack.tar
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./attrs-20.3.0-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./numpy-1.17.2-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./decorator-5.0.6-py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./mpmath-1.2.1-py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./sympy-1.4-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./pycparser-2.20-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./cffi-1.12.3.tar.gz
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./PyYAML-5.3.1.tar.gz
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./six-1.15.0-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./pathlib2-2.3.5-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./psutil-5.8.0.tar.gz
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./protobuf-3.15.8-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./scipy-1.6.0-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./chardet-3.0.4-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./idna-2.10-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./urllib3-1.25.10-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./certifi-2020.6.20-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./certifi-2020.6.20-py2.py3-none-any.whl
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./requests-2.24.0-py2.py3-none-any.wh
root@ubuntu:/home/cby/pip-pack/pip\_pack#pip3.7.5 install ./xlrd-1.2.0-py2.py3-none-any.whl

 *注意：以上pip包的安裝必須以該順序依次進行安裝

安裝開發套件包

root@ubuntu:/home/cby/cann#./Ascend-cann-tfplugin\_20.2.rc1\_linux-aarch64.run –install
root@ubuntu:/home/cby/cann#./Ascend-cann-toolkit\_20.2.rc1\_linux-aarch64.run –install

    出現install success後表示安裝成功。

CANN訓練環境部署安裝

說明

       訓練環境的Python3.7.5和環境以及依賴，和開發環境下的安裝方式同樣，可參考《CANN開發環境部署安裝》文檔進行安裝。在已經搭建好的開發環境中，進行安裝訓練環境僅需安裝一下訓練軟件包和實用工具包便可。

安裝訓練軟件包

root@ubuntu:/home/cby/cann\_xunlian# chmod+x ./\*.run
root@ubuntu:/home/cby/cann\_xunlian# ./Ascend-cann-nnae\_20.2.rc1\_linux-aarch64.run–install
root@ubuntu:/home/cby/cann\_xunlian#./Ascend-cann-toolbox\_20.2.rc1\_linux-aarch64.run –install

       出現install success後表示安裝成功。

安裝MindSpore

安裝whl包

        安裝Ascend 910 AI處理器配套軟件包提供的whl包，whl包隨配套軟件包發佈，升級配套軟件包以後須要從新安裝。

root@ubuntu:/home/cby/mindspore\_ascend#pip3.7.5 install /usr/local/Ascend/ascend-toolkit/latest/fwkacllib/lib64/hccl-0.1.0-py3-none-any.whl
root@ubuntu:/home/cby/mindspore\_ascend#pip3.7.5 install /usr/local/Ascend/ascend-toolkit/latest/fwkacllib/lib64/te-0.4.0-py3-none-any.whl
root@ubuntu:/home/cby/mindspore\_ascend#pip3.7.5 install /usr/local/Ascend/ascend-toolkit/latest/fwkacllib/lib64/topi-0.4.0-py3-none-any.whl
 
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install easydict-1.9.tar.gz
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install ./wheel-0.36.2-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install ./astunparse-1.6.3-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install ./Pillow-8.2.0-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install ./asttokens-2.0.4-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install ./cffi-1.14.5-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install ./pyparsing-2.4.7-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install ./packaging-20.9-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindspore\_ascend/pip#pip3.7.5 install ../mindspore\_ascend-1.1.1-cp37-cp37m-linux\_aarch64.whl

*注意：安裝時必須以此順序進行安裝

配置環境變量

\# control log level.0-DEBUG, 1-INFO, 2-WARNING, 3-ERROR, default level is WARNING.
 
export GLOG\_v=2
 
# Conda environmentaloptions
 
LOCAL\_ASCEND=/usr/local/Ascend
 
# the root directoryof run package
 
# lib libraries thatthe run package depends on
 
exportLD\_LIBRARY\_PATH=${LOCAL\_ASCEND}/add-ons/:${LOCAL\_ASCEND}/ascend-toolkit/latest/fwkacllib/lib64:${LOCAL\_ASCEND}/driver/lib64:${LOCAL\_ASCEND}/opp/op\_impl/built-in/ai\_core/tbe/op\_tiling:${LD\_LIBRARY\_PATH}
 
 
# Environmentvariables that must be configured
 
exportTBE\_IMPL\_PATH=${LOCAL\_ASCEND}/ascend-toolkit/latest/opp/op\_impl/built-in/ai\_core/tbe
 
# TBE operatorimplementation tool path
 
exportASCEND\_OPP\_PATH=${LOCAL\_ASCEND}/ascend-toolkit/latest/opp
 
# OPP path
 
exportPATH=${LOCAL\_ASCEND}/ascend-toolkit/latest/fwkacllib/ccec\_compiler/bin/:${PATH}
 
# TBE operatorcompilation tool path
 
exportPYTHONPATH=${TBE\_IMPL\_PATH}:${PYTHONPATH}
 
# Python library thatTBE implementation depends on

測試是否可行

Python代碼內容：

import numpy as np
from mindspore importTensor
import mindspore.opsas ops
importmindspore.context as context
 
context.set\_context(device\_target="Ascend")
 
x =Tensor(np.ones(\[1,3,3,4\]).astype(np.float32))
y =Tensor(np.ones(\[1,3,3,4\]).astype(np.float32))
print(ops.tensor\_add(x,y))

出現此結果便是安裝部署完成

\[\[\[\[2. 2. 2. 2.\]
   \[2. 2. 2. 2.\]
   \[2. 2. 2. 2.\]\]
 
  \[\[2. 2. 2. 2.\]
   \[2. 2. 2. 2.\]
   \[2. 2. 2. 2.\]\]
 
  \[\[2. 2. 2. 2.\]
   \[2. 2. 2. 2.\]
   \[2. 2. 2. 2.\]\]\]\]

安裝mindinsight

安裝whl包

root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./itsdangerous-1.1.0-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./Werkzeug-1.0.1-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./MarkupSafe-1.1.1-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./Jinja2-2.11.3-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./click-7.1.2-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./Flask-1.1.2-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./Flask\_Cors-3.0.10-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./yapf-0.31.0-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./future-0.18.2.tar.gz
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./treelib-1.6.1.tar.gz
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./grpcio-1.37.0-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./google\_pasta-0.2.0-py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./pytz-2021.1-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./python\_dateutil-2.8.1-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./pandas-1.2.3-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./gunicorn-20.1.0.tar.gz
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./marshmallow-3.11.1-py2.py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./threadpoolctl-2.1.0-py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./joblib-1.0.1-py3-none-any.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./scikit\_learn-0.24.1-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/mindinsight/Mindinsight#pip3.7.5 install ./mindinsight-1.1.1-cp37-cp37m-linux\_aarch64.whl

*注意：安裝必須以此順序進安裝

配置環境變量

在配置文件中配置以下變量

PATH=$PATH:/usr/local/python3.7.5/bin/

root@ubuntu:/home/cby#source /etc/profile

啓動及使用

root@ubuntu:/home/cby#mindinsight start
Workspace:/root/mindinsight
Webaddress: http://127.0.0.1:8080
servicestart state: success

    出現該消息後，說明可視化已經啓動成功，若須要外機訪問的話，須要進行反向代理到0.0.0.0上面便可，好比frp工具便可實現該操做

    在訓練完成的Python代碼目錄下，使如下命令便可啓動並展現該目錄下的訓練數據，debugger的參數可以使用false或者true

mindinsightstart --summary-base-dir . --port 8080 --enable-debugger True --debugger-port50051

使用以下命令便可啓動訓練

root@ubuntu:/home/cby/lenet/lenet#python3.7.5 lenet.py --device_target=Ascend

安裝Tensorflow

編譯hdf5

root@ubuntu:/home/cby/Tensorflow/Tensorflow#cd hdf5-1.10.5/
root@ubuntu:/home/cby/Tensorflow/Tensorflow/hdf5-1.10.5#./configure --prefix=/usr/include/hdf5
root@ubuntu:/home/cby/Tensorflow/Tensorflow/hdf5-1.10.5#make
root@ubuntu:/home/cby/Tensorflow/Tensorflow/hdf5-1.10.5#make install

配置環境變量及軟鏈接

exportCPATH="/usr/include/hdf5/include/:/usr/include/hdf5/lib/"
root@ubuntu:/home/cby/Tensorflow/Tensorflow/hdf5-1.10.5#ln -s /usr/include/hdf5/lib/libhdf5.so /usr/lib/libhdf5.so
root@ubuntu:/home/cby/Tensorflow/Tensorflow/hdf5-1.10.5#ln -s /usr/include/hdf5/lib/libhdf5\_hl.so /usr/lib/libhdf5\_hl.so

安裝whl包

root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./Cython-0.29.21-py2.py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./h5py-2.10.0-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./grpcio-1.30.0.tar.gz
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./gast-0.2.2.tar.gz
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./opt\_einsum-3.3.0-py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./Keras\_Applications-1.0.8-py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./Keras\_Preprocessing-1.1.2-py2.py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./astor-0.8.1-py2.py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./typing\_extensions-3.7.4.3-py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./zipp-3.4.1-py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./importlib\_metadata-3.10.1-py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./Markdown-3.2.2-py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./tensorboard-1.15.0-py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./wrapt-1.12.1.tar.gz
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./tensorflow\_estimator-1.15.1-py2.py3-none-any.whl
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./termcolor-1.1.0.tar.gz
root@ubuntu:/home/cby/Tensorflow/Tensorflow#pip3.7.5 install ./tensorflow-1.15.0-cp37-cp37m-linux\_aarch64.whl

注意：必須依次安裝

安裝Pytorch

root@ubuntu:/home/cby/pytorch/Pytorch#pip3.7.5 install ./apex-0.1+ascend-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/pytorch/Pytorch#pip3.7.5 install ./torch-1.5.0+ascend.post2-cp37-cp37m-linux\_aarch64.whl
root@ubuntu:/home/cby/pytorch/Pytorch#pip3.7.5 install ./future-0.18.2.tar.gz

該文章所配套的軟件包關注微信公衆號回覆 ai 便可獲取所須要的全部軟件包

Linux運維交流社區

Linux運維交流社區，互聯網新聞以及技術交流。

20篇原創內容

公衆號