windows下，完全离线安装anaconda与tensorflow-gpu，并用其进行训练

最新推荐文章于 2024-07-26 14:32:27 发布

dna葡萄糖

最新推荐文章于 2024-07-26 14:32:27 发布

阅读量2.7k

点赞数 3

文章标签： tensorflow 深度学习离线安装

本文链接：https://blog.csdn.net/qq_35942419/article/details/89962188

版权

因项目需求，需要出差并在离线的情况下进行深度学习目标检测，由于电脑没有网络连接功能，无法通过直接迁移环境来实现配置，同时在尝试离线安装的时候，安装低版本的anaconda报错，故采用此种方式安装软件环境。（如果利用prompt下载慢的话也可以采用这种方式）

假定本机环境：win10 GTX1050ti （目标主机环境目前未知）

参考教程：https://blog.csdn.net/qq_40434140/article/details/84800901

https://blog.csdn.net/qq_40434140/article/details/88083142

https://user.qzone.qq.com/892054726/blog/1467193413?_t_=0.12324156286134635&

安装步骤：（你需要一个迅雷，最好冲个会员，方便一口气复制所有链接，所有下载链接在https://repo.anaconda.com/pkgs 以及 https://repo.continuum.io/pkgs 里面，如果下载链接有问题可以自行在其他网站找相同名称的文件）

1、 安装框架

安装显卡驱动（不要太旧，不要太新，试过cuda ga2 对最低配置376.xx，正好是此版本的时候会报错）

安装Anaconda https://www.anaconda.com/download/

安装时选择加入path

根据显卡安装CUDA https://developer.nvidia.com/cuda-toolkit-archive

安装对应版本cudNN https://developer.nvidia.com/rdp/cudnn-download

--将解压后的文件拷贝到CUDA development文件夹下

注：a、CUDA提示警告的时候，可以通过回退显卡驱动版本来取消警告，回退显卡驱动在相关品牌电脑技术支持可找到；

b、cudNN版本需要与CUDA版本对应

本机所安装Anaconda 为 Anaconda3-2019.03-Windows-x86_64.exe

CUDN为8.0 GA2 cudnn 7.1.4

2、 配置与安装tensorflow

1）创建python3.5编程环境（由于anaconda的python本身是3.7版本的，首先需要创建一个3.5的环境，考虑到conda install <路径+文件名>太慢了，采用创建本机channel的方式进行安装，原理就是在本地建立一个镜像源，把所有后续安装可能需要的文件从网络镜像源挪到本地）

a、关闭anaconda 网络环境源（或者每条命令加--offline）

conda config --set offline true

b、建立一个channel文件夹用于存放各种库文件，切换到这个目录下

cd D:\deep_learning\channel

c、在channel目录下新建一个文件夹命名为win-64

将已经下载好的以下文件放入win-64中（可以一口气把下面所有文件都放进去）

https://repo.anaconda.com/pkgs/main/win-64/python-3.5.6-he025d50_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/certifi-2018.8.24-py35_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/pip-10.0.1-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/setuptools-40.2.0-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/vc-14.1-h0510ff6_4.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/vs2015_runtime-14.15.26706-h3a45250_4.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/wheel-0.31.1-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/wincertstore-0.2-py35hfebbdb8_0.tar.bz2

d、创建路径索引文件，执行命令，在文件下产生repodata.json与repodata.json.bz2等文件

conda index win-64

conda config –add channels D:\deep_learning\channel

记事本查看 c:/用户/<用户名>/.condarc 文件，在channels下应有D:\deep_learning\channel路径，没有就自己加上

e、创建环境

conda create –n tensorflow python=3.5

2）安装tensorflow-gpu的前置库

将已经下载好的以下文件放入win-64中

https://repo.anaconda.com/pkgs/main/win-64/_tflow_190_select-0.0.1-gpu.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/absl-py-0.4.1-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/astor-0.7.1-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/blas-1.0-mkl.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/cudatoolkit-9.0-1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/cudnn-7.3.1-cuda9.0_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/gast-0.2.0-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/grpcio-1.12.1-py35h1a1b453_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/icc_rt-2019.0.0-h0cc432a_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/intel-openmp-2019.3-203.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/libprotobuf-3.6.0-h1a1b453_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/markdown-2.6.11-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/mkl-2018.0.3-1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/mkl_fft-1.0.6-py35hdbbee80_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/mkl_random-1.0.1-py35h77b88f5_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/numpy-1.15.2-py35ha559c80_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/numpy-base-1.15.2-py35h8128ebf_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/protobuf-3.6.0-py35he025d50_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/six-1.11.0-py35_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/tensorboard-1.9.0-py35he025d50_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/tensorflow-1.9.0-gpu_py35h0075c17_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/tensorflow-base-1.9.0-gpu_py35h6e53903_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/tensorflow-gpu-1.9.0-hf154084_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/termcolor-1.1.0-py35_1.tar.bz2

https://repo.anaconda.com/pkgs/main/noarch/werkzeug-0.15.2-py_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/zlib-1.2.11-h62dcd97_3.tar.bz2

conda index win-64

打开在win-64文件下生成的repodata.json文件，查找tensorflow-base-1.9.0-gpu_py35h6e53903_0这一项，将cudnn 7.1.* 改为 cudnn >=7.1.4,<8.0a0 （每次conda index 后都需要这样操作一遍）

注：如果出现任何配置错误，可以查看repodata.json文件确定其前置包，或者通过conda search <包名> -i 来查看详细配置信息，通过官网来寻找对应包

3）安装tensorflow-gpu

conda install tensorflow-gpu

4）安装spyder （如果安装的是python对应版本的anaconda这步省略，这步装的我恶心。。可以找一台装的是对应版本anaconda 把它目录/pkgs下面的统统拷贝过来也可（会缺少部分文件，缺少的去下面https://repo.anaconda.com/***下载））

将已经下载好的以下文件放入win-64中

https://repo.continuum.io/pkgs/free/win-64/alabaster-0.7.9-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/astroid-1.4.7-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/babel-2.3.4-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/chardet-3.0.4-py35_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/cloudpickle-0.5.5-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/colorama-0.3.7-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/decorator-4.0.10-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/docutils-0.12-py35_2.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/entrypoints-0.2.2-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/icu-57.1-vc14_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/imagesize-0.7.1-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/ipykernel-4.5.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/ipython-5.1.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/ipython_genutils-0.1.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/jedi-0.9.0-py35_1.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/jinja2-2.8-py35_1.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/jpeg-8d-vc14_2.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/jsonschema-2.5.1-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/jupyter_client-4.4.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/jupyter_core-4.2.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/lazy-object-proxy-1.2.1-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/libpng-1.6.22-vc14_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/markupsafe-0.23-py35_2.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/mistune-0.7.3-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/nbconvert-4.2.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/nbformat-4.1.0-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/numpydoc-0.8.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/openssl-1.0.2j-vc14_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/path.py-8.2.1-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/pep8-1.7.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/pickleshare-0.7.4-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/prompt_toolkit-1.0.3-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/psutil-5.4.7-py35hfa6e2cd_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/pycodestyle-2.4.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/pyflakes-1.3.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/pygments-2.1.3-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/pylint-1.5.4-py35_1.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/pyqt-5.6.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/pytz-2016.6.1-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/pyzmq-15.4.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/qt-5.6.0-vc14_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/qtawesome-0.3.3-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/qtconsole-4.2.1-py35_2.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/qtpy-1.5.0-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/rope-0.11.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/simplegeneric-0.8.1-py35_1.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/sip-4.18-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/snowballstemmer-1.2.1-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/sphinx-1.4.6-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/spyder-3.0.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/tornado-4.4.1-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/traitlets-4.3.0-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/wcwidth-0.1.7-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/win_unicode_console-0.5-py35_0.tar.bz2

https://repo.continuum.io/pkgs/free/win-64/wrapt-1.10.6-py35_0.tar.bz2

conda index win-64

conda install spyder=3.0.0 –offline

5）测试tensorflow-gpu

运行spyder或在prompt输入python启动编辑器

输入

import tensorflow as tf

a = tf.constant([1.0, 2.0,3.0,4.0,5.0,6.0],shape=[2,3], name =“a”)

b = tf.constant([1.0, 2.0,3.0,4.0,5.0,6.0],shape=[3,2], name =“b”)

c=tf.matmul(a,b)

sess = tf.Session(config = tf.ConfigProto(log_device_placement=True))

print(sess.run(c))

得到结果即证明安装成功

如果报错“Failed to create session”，需要更新显卡驱动版本，拷贝较新的（2017年的）（不要最新的）（也不要卡要求底线）驱动版本安装即可。

3、配置tensorflow object detection API环境

1）配置前置库

将已经下载好的以下文件放入win-64中

https://repo.anaconda.com/pkgs/main/win-64/freetype-2.9.1-ha9979f8_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/libtiff-4.0.9-h36446d0_2.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/olefile-0.46-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/pillow-5.2.0-py35h08bbbbd_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/tk-8.6.8-hfa6e2cd_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/jpeg-9b-hb83a4c4_2.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/libpng-1.6.34-h79bbb47_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/pyqt-5.6.0-py35ha878b3d_6.tar.bz2

https://repo.anaconda.com/pkgs/free/win-64/qt-5.6.2-vc14_6.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/vc-14-h21ff451_10.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/zlib-1.2.11-h8395fce_2.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/cython-0.28.5-py35h6538335_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/libiconv-1.15-h1df5818_7.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/libxml2-2.9.9-h464c3ec_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/libxslt-1.1.33-h579f668_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/lxml-4.2.5-py35hef2cd61_0.tar.bz2

https://repo.anaconda.com/pkgs/free/win-64/ipywidgets-5.2.2-py35_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/jupyter-1.0.0-py35_7.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/jupyter_console-5.2.0-py35_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/notebook-5.3.1-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/send2trash-1.5.0-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/widgetsnbextension-3.4.1-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/cycler-0.10.0-py35hcc71164_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/kiwisolver-1.0.1-py35h6538335_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/matplotlib-2.2.2-py35had4c4a9_2.tar.bz2

https://repo.anaconda.com/pkgs/main/noarch/pyparsing-2.4.0-py_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/python-dateutil-2.7.3-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/libpng-1.6.37-h2a8f88b_0.tar.bz2

conda index win-64

conda install Cython

conda install pillow

conda install lxml --offline

conda install jupyter --offline

conda install matplotlib

（protobuf在之前步骤已被默认安装）

2）下载tensorflow/models

下载地址：https://github.com/tensorflow/models，下载zip格式

将下载好的文件名改为models，将文件解压到某文件夹下，如“D:\TensorflowDET\models”

3）安装protobuf

a）下载链接：https://github.com/protocolbuffers/protobuf/releases?after=v3.4.1

在Javascript下面有一个Assets 23 打开找protoc-3.4.0-win32.zip 下载

或者直接通过连接地址下载（如下）

https://github.com/protocolbuffers/protobuf/releases/download/v3.4.0/protoc-3.4.0-win32.zip

b）将其解压后的文件放入“D:\TensorflowDET”文件夹下

c）将D:\TensorflowDET\bin加入到系统变量path中

d）利用win+r调出cmd窗口，

cd D:\TensorflowDET\models\research

protoc object_detection/protos/*.proto --python_out=.

会在“D:\TensorflowDET\models\research\object_detection\protos”下出现文件

4） tf slim库安装

a）新建系统变量变量名为PYTHONPATH 变量值为D:\TensorflowDET\models\research; D:\TensorflowDET\models\research\slim

b）利用prompt，在tensorflow环境下，切换路径“D:\TensorflowDET\models\research” 执行 python setup.py install

打开“D:\TensorflowDET\models\research\slim”文件，删掉build文件

prompt 切换路径“D:\TensorflowDET\models\research\slim”执行 python setup.py install

5）测试安装效果

cd .. 将路径切换回“D:\TensorflowDET\models\research”

object_detection/builders/model_builder_test.py

不报错就成功

6）测试demo

cd D:\TensorflowDET\models\research\object_detection

jupyter notebook

注：如果出现cannot impport name ‘ensure_dir_exists’ 的错误将下列文件移至channel（有些是不必要的，但我不知道哪些是哪些不是）

https://repo.anaconda.com/pkgs/free/win-64/libtiff-4.0.2-1.tar.bz2

https://repo.anaconda.com/pkgs/free/win-64/pillow-2.9.0-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/numpy-1.13.3-py35h5c71026_4.tar.bz2

https://repo.anaconda.com/pkgs/free/win-64/matplotlib-2.0.2-np113py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/ca-certificates-2019.1.23-0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/sqlite-3.28.0-he774522_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/xz-5.2.4-h2fa13f4_4.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/zstd-1.3.7-h508b16e_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/icu-58.2-ha66f8fd_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/jpeg-9b-hb83a4c4_2.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/jupyter_client-5.2.3-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/jupyter_core-4.4.0-py35_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/libtiff-4.0.10-hb898794_2.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/openssl-1.1.1b-he774522_1.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/pyqt-5.9.2-py35h6538335_2.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/qt-5.9.7-vc14h73c81de_0.tar.bz2

https://repo.anaconda.com/pkgs/main/win-64/sip-4.19.8-py35h6538335_0.tar.bz2

cd D:\deep_learning\channel

conda index win-64

conda update jupyter_core jupyter_client –offline

可解决此问题

打开此目录下的object_detection_tutorial.ipynb

注释掉imports下的

if StrictVersion(tf.__version__) < StrictVersion('1.12.0'):

raise ImportError('Please upgrade your TensorFlow installation to v1.12.*.')

上方cell→run all

在最后显示出检测图片，就代表整套东西已顺利安装

dna葡萄糖

关注

3
点赞
踩
5

收藏

觉得还不错? 一键收藏
0
评论
复制链接

分享到 QQ

分享到新浪微博

扫一扫