由于工作平台原因，需要在内网下安装深度学习框架，不联网的话，需要提前下载好第三方库的安装包，再离线安装。这里以Anaconda和Tensorflow为例，看过几个博客，坑不少，所以这里写点心得。这里以python3.8，conda4.10，tensorflow-2.6为例，先下载好Anaconda3-2021.05-Windows-x86_64与tensorflow-2.6.0-cp38-cp38-win_amd64。
一、下载情况
1.Anaconda的下载：https://www.anaconda.com/download/
2.Tensorflow的下载，可以去清华大学开源软件镜像站，好多博主给的网站都是之前的，不可用，新地址为https://pypi.tuna.tsinghua.edu.cn/simple/
3.Tensorflow的支撑包需要以下这些：
Requirement already satisfied: keras~=2.6
Requirement already satisfied: keras-preprocessing~=1.1.2
Requirement already satisfied: gast==0.4.0
Requirement already satisfied: wrapt~=1.12.1
Requirement already satisfied: h5py~=3.1.0
Requirement already satisfied: termcolor~=1.1.0
Requirement already satisfied: grpcio<2.0,>=1.37.0
Requirement already satisfied: astunparse~=1.6.3
Requirement already satisfied: tensorboard~=2.6
Requirement already satisfied: wheel~=0.35
Requirement already satisfied: opt-einsum~=3.3.0
Requirement already satisfied: absl-py~=0.10
Requirement already satisfied: google-pasta~=0.2
Requirement already satisfied: typing-extensions~=3.7.4
Requirement already satisfied: numpy~=1.19.2
Requirement already satisfied: protobuf>=3.9.2
Requirement already satisfied: flatbuffers~=1.12.0
Requirement already satisfied: six~=1.15.0
Requirement already satisfied: tensorflow-estimator~=2.6
Requirement already satisfied: clang~=5.0
Requirement already satisfied: setuptools>=41.0.0
Requirement already satisfied: requests<3,>=2.21.0
Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0
Requirement already satisfied: tensorboard-plugin-wit>=1.6.0
Requirement already satisfied: werkzeug>=0.11.15
Requirement already satisfied: google-auth<2,>=1.6.3
Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1
Requirement already satisfied: markdown>=2.6.8
Requirement already satisfied: rsa<5,>=3.1.4
Requirement already satisfied: pyasn1-modules>=0.2.1
Requirement already satisfied: cachetools<5.0,>=2.0.0
Requirement already satisfied: requests-oauthlib>=0.7.0
Requirement already satisfied: pyasn1<0.5.0,>=0.4.6
Requirement already satisfied: urllib3<1.27,>=1.21.1
Requirement already satisfied: certifi>=2017.4.17
Requirement already satisfied: idna<3,>=2.5
Requirement already satisfied: chardet<5,>=3.0.2
Requirement already satisfied: oauthlib>=3.0.0
二、安装过程
1.安装Anaconda
2.进入Anaconda安装路径，到\Lib\site-packages目录下，创建文件夹tensorflow
3.在创建的tensorflow文件夹中，解压上文下载的支撑包.rar，将支撑包文件夹中的所有文件都放到tensorflow文件夹中
4.将tensorflow-2.6.0-cp38-cp38-win_amd64.whl文件放到tensorflow文件夹中
5.解压protoc-3.17.3-win64.zip，将bin中的protoc.exe复制到C:\windows\system32中
6.直接解压protobuf-python-3.17.3.zip，得到protobuf-3.17.3.文件夹，将第5步中的protoc.exe文件也复制到protobuf-3.17.3\src文件夹中
7.在cmd命令中更改路径，进入到\Lib\site-packages\tensorflow\protobuf-3.17.3\python路径中，执行python setup.py install;
8.切换安装路径到\Lib\site-packages\tensorflow，接下来先安装tensorflow的各个支撑包，顺序不对，按提示进行安装
whl文件直接就利用pip install 文件名；rar压缩文件形式的安装，与安装protobuf一致，先解压，然后切换路径，再执行python setup.py install。
三、本教程各支撑包如图版本在这里插入图片描述
，其他版本的自行从网站下载：https://pypi.org/search/
记得要安装和python对应各个版本，否则会出错
如numpy-1.19.2-cp38-cp38-win_amd64.whl 表示numpy的版本是python3.8,64位操作系统的。
