1.下载anaconda
点开下面的链接，下载版本Anaconda3-4.1.1-Windows-x86_64.exe
https://repo.anaconda.com/archive/
2.安装anaconda
一路点击安装
3.安装tensorflow
两种方式安装TF，第一种是通过命令来安装，在anaconda prompt下PIP安装：
下载非常的慢，一个变通的方式是， 我们看到通过PIP安装时会将包文件资源的整个路径显示出来，我们可以直接拷贝这个路径，然后用迅雷等下载工具进行下载：
下载完tensorflow二进制包后，我们可以看到，它的名字是有明显特征的，比如CP35表示对应的Python版本为3.5版.
安装二进制包的方法是，anaconda prompt进入到包所的目录，执行
pip install tensorflow-2.3.1-cp35-cp35m-win_amd64.whl
验证的时候，如果遇到 "no module named tensorflow"
这个时候，在命令行下执行conda list查看，如果没有tensorflow等包，需要在conda下再次安装一遍
C:\Users\DELL>conda list
# packages in environment at C:\Users\DELL\Anaconda3:
#
_nb_ext_conf 0.2.0 py35_0
absl-py 0.13.0 <pip>
alabaster 0.7.8 py35_0
anaconda 4.1.1 np111py35_0
anaconda-client 1.4.0 py35_0
anaconda-navigator 1.2.1 py35_0
argcomplete 1.0.0 py35_1
astropy 1.2.1 np111py35_0
astunparse 1.6.3 <pip>
babel 2.3.3 py35_0
backports 1.0 py35_0
beautifulsoup4 4.4.1 py35_0
bitarray 0.8.1 py35_1
blaze 0.10.1 py35_0
bokeh 0.12.0 py35_0
boto 2.40.0 py35_0
bottleneck 1.1.0 np111py35_0
bzip2 1.0.6 vc14_3 [vc14]
cachetools 4.2.2 <pip>
cffi 1.6.0 py35_0
chest 0.2.3 py35_0
click 6.6 py35_0
cloudpickle 0.2.1 py35_0
clyent 1.2.2 py35_0
colorama 0.3.7 py35_0
comtypes 1.1.2 py35_0
conda 4.1.6 py35_0
conda-build 1.21.3 py35_0
conda-env 2.5.1 py35_0
configobj 5.0.6 py35_0
console_shortcut 0.1.1 py35_1
contextlib2 0.5.3 py35_0
cryptography 1.4 py35_0
curl 7.49.0 vc14_0 [vc14]
cycler 0.10.0 py35_0
cython 0.24 py35_0
cytoolz 0.8.0 py35_0
dask 0.10.0 py35_0
datashape 0.5.2 py35_0
decorator 4.0.10 py35_0
dill 0.2.5 py35_0
docutils 0.12 py35_2
dynd-python 0.7.2 py35_0
entrypoints 0.2.2 py35_0
et_xmlfile 1.0.1 py35_0
fastcache 1.0.2 py35_1
flask 0.11.1 py35_0
flask-cors 2.1.2 py35_0
freetype 2.5.5 vc14_1 [vc14]
gast 0.3.3 <pip>
get_terminal_size 1.0.0 py35_0
gevent 1.1.1 py35_0
google-pasta 0.2.0 <pip>
greenlet 0.4.10 py35_0
grpcio 1.40.0 <pip>
h5py 2.10.0 <pip>
h5py 2.6.0 np111py35_0
hdf5 1.8.15.1 vc14_4 [vc14]
heapdict 1.0.0 py35_1
idna 2.1 py35_0
imagesize 0.7.1 py35_0
importlib-metadata 4.8.1 <pip>
ipykernel 4.3.1 py35_0
ipython 4.2.0 py35_0
ipython_genutils 0.1.0 py35_0
ipywidgets 4.1.1 py35_0
itsdangerous 0.24 py35_0
jdcal 1.2 py35_1
jedi 0.9.0 py35_1
jinja2 2.8 py35_1
jpeg 8d vc14_0 [vc14]
jsonschema 2.5.1 py35_0
jupyter 1.0.0 py35_3
jupyter_client 4.3.0 py35_0
jupyter_console 4.1.1 py35_0
jupyter_core 4.1.0 py35_0
Keras-Preprocessing 1.1.2 <pip>
libdynd 0.7.2 0
libpng 1.6.22 vc14_0 [vc14]
libtiff 4.0.6 vc14_2 [vc14]
llvmlite 0.11.0 py35_0
locket 0.2.0 py35_1
lxml 3.6.0 py35_0
Markdown 3.3.4 <pip>
markupsafe 0.23 py35_2
matplotlib 1.5.1 np111py35_0
menuinst 1.4.1 py35_0
mistune 0.7.2 py35_0
mkl 11.3.3 1
mkl-service 1.1.2 py35_2
mpmath 0.19 py35_1
multipledispatch 0.4.8 py35_0
nb_anacondacloud 1.1.0 py35_0
nb_conda 1.1.0 py35_0
nb_conda_kernels 1.0.3 py35_0
nbconvert 4.2.0 py35_0
nbformat 4.0.1 py35_0
nbpresent 3.0.2 py35_0
networkx 1.11 py35_0
nltk 3.2.1 py35_0
nose 1.3.7 py35_1
notebook 4.2.1 py35_0
numba 0.26.0 np111py35_0
numexpr 2.6.0 np111py35_0
numpy 1.11.1 py35_0
numpy 1.18.5 <pip>
odo 0.5.0 py35_1
openpyxl 2.3.2 py35_0
openssl 1.0.2h vc14_0 [vc14]
opt-einsum 3.3.0 <pip>
pandas 0.18.1 np111py35_0
partd 0.3.4 py35_0
path.py 8.2.1 py35_0
pathlib2 2.1.0 py35_0
patsy 0.4.1 py35_0
pep8 1.7.0 py35_0
pickleshare 0.7.2 py35_0
pillow 3.2.0 py35_1
pip 8.1.2 py35_0
ply 3.8 py35_0
protobuf 3.18.0 <pip>
psutil 4.3.0 py35_0
py 1.4.31 py35_0
pyasn1 0.1.9 py35_0
pyasn1 0.4.8 <pip>
pyasn1-modules 0.2.8 <pip>
pycosat 0.6.1 py35_1
pycparser 2.14 py35_1
pycrypto 2.6.1 py35_4
pycurl 7.43.0 py35_0
pyflakes 1.2.3 py35_0
pygments 2.1.3 py35_0
pyopenssl 0.16.0 py35_0
pyparsing 2.1.4 py35_0
pyqt 4.11.4 py35_6
pyreadline 2.1 py35_0
pytables 3.2.2 np111py35_4
pytest 2.9.2 py35_0
python 3.5.2 0
python-dateutil 2.5.3 py35_0
pytz 2016.4 py35_0
pywin32 220 py35_1
pyyaml 3.11 py35_4
pyzmq 15.2.0 py35_0
qt 4.8.7 vc14_8 [vc14]
qtconsole 4.2.1 py35_0
qtpy 1.0.2 py35_0
requests 2.10.0 py35_0
rope 0.9.4 py35_1
rsa 4.5 <pip>
ruamel_yaml 0.11.7 py35_0
scikit-image 0.12.3 np111py35_1
scikit-learn 0.17.1 np111py35_1
scipy 0.17.1 np111py35_1
setuptools 23.0.0 py35_0
simplegeneric 0.8.1 py35_1
singledispatch 3.4.0.3 py35_0
sip 4.16.9 py35_2
six 1.10.0 py35_0
six 1.16.0 <pip>
snowballstemmer 1.2.1 py35_0
sockjs-tornado 1.0.3 py35_0
sphinx 1.4.1 py35_0
sphinx_rtd_theme 0.1.9 py35_0
spyder 2.3.9 py35_0
sqlalchemy 1.0.13 py35_0
statsmodels 0.6.1 np111py35_1
sympy 1.0 py35_0
tensorboard-data-server 0.6.1 <pip>
tensorboard-plugin-wit 1.8.0 <pip>
tensorflow-estimator 2.3.0 <pip>
termcolor 1.1.0 <pip>
tk 8.5.18 vc14_0 [vc14]
toolz 0.8.0 py35_0
tornado 4.3 py35_1
traitlets 4.2.1 py35_0
typing-extensions 3.10.0.2 <pip>
unicodecsv 0.14.1 py35_0
vs2015_runtime 14.0.25123 0
werkzeug 0.11.10 py35_0
wheel 0.29.0 py35_0
wrapt 1.12.1 <pip>
xlrd 1.0.0 py35_0
xlsxwriter 0.9.2 py35_0
xlwings 0.7.2 py35_0
xlwt 1.1.2 py35_0
zipp 3.5.0 <pip>
zlib 1.2.8 vc14_3 [vc14]
执行conda命令： conda install tensorflow
C:\Users\DELL>conda install tensorflow
Fetching package metadata .......
Solving package specifications: ..........
Package plan for installation in environment C:\Users\DELL\Anaconda3:
The following packages will be downloaded:
package | build
---------------------------|-----------------
blas-1.0 | mkl 6 KB
conda-env-2.6.0 | 0 498 B
vs2015_runtime-14.0.25420 | 0 2.0 MB
vc-14 | 0 703 B
libprotobuf-3.2.0 | vc14_0 9.1 MB
markdown-2.6.9 | py35_0 101 KB
requests-2.14.2 | py35_0 705 KB
ruamel_yaml-0.11.14 | py35_1 221 KB
backports.weakref-1.0rc1 | py35_0 8 KB
html5lib-0.9999999 | py35_0 185 KB
protobuf-3.2.0 | py35_0 463 KB
bleach-1.5.0 | py35_0 22 KB
pyopenssl-16.2.0 | py35_0 70 KB
tensorflow-1.2.1 | py35_0 21.0 MB
conda-4.3.30 | py35hec795fb_0 541 KB
------------------------------------------------------------
Total: 34.4 MB
The following NEW packages will be INSTALLED:
backports.weakref: 1.0rc1-py35_0
blas: 1.0-mkl
bleach: 1.5.0-py35_0
html5lib: 0.9999999-py35_0
libprotobuf: 3.2.0-vc14_0
markdown: 2.6.9-py35_0
protobuf: 3.2.0-py35_0
tensorflow: 1.2.1-py35_0
vc: 14-0
The following packages will be UPDATED:
conda: 4.1.6-py35_0 --> 4.3.30-py35hec795fb_0
conda-env: 2.5.1-py35_0 --> 2.6.0-0
pyopenssl: 0.16.0-py35_0 --> 16.2.0-py35_0
requests: 2.10.0-py35_0 --> 2.14.2-py35_0
ruamel_yaml: 0.11.7-py35_0 --> 0.11.14-py35_1
vs2015_runtime: 14.0.25123-0 --> 14.0.25420-0
Proceed ([y]/n)? y
Fetching packages ...
blas-1.0-mkl.t 100% |###############################| Time: 0:00:00 2.10 MB/s
conda-env-2.6. 100% |###############################| Time: 0:00:00 251.87 kB/s
vs2015_runtime 100% |###############################| Time: 0:00:02 748.93 kB/s
vc-14-0.tar.bz 100% |###############################| Time: 0:00:00 408.96 kB/s
libprotobuf-3. 100% |###############################| Time: 0:00:02 4.00 MB/s
markdown-2.6.9 100% |###############################| Time: 0:00:00 447.80 kB/s
requests-2.14. 100% |###############################| Time: 0:00:00 865.03 kB/s
ruamel_yaml-0. 100% |###############################| Time: 0:00:00 528.02 kB/s
backports.weak 100% |###############################| Time: 0:00:00 2.12 MB/s
html5lib-0.999 100% |###############################| Time: 0:00:00 207.23 kB/s
protobuf-3.2.0 100% |###############################| Time: 0:00:00 645.29 kB/s
bleach-1.5.0-p 100% |###############################| Time: 0:00:00 1.02 MB/s
pyopenssl-16.2 100% |###############################| Time: 0:00:00 364.58 kB/s
tensorflow-1.2 100% |###############################| Time: 0:00:07 2.84 MB/s
conda-4.3.30-p 100% |###############################| Time: 0:00:00 723.63 kB/s
Extracting packages ...
[ COMPLETE ]|##################################################| 100%
Unlinking packages ...
[ COMPLETE ]|##################################################| 100%
Linking packages ...
[ COMPLETE ]|##################################################| 100%
C:\Users\DELL>
此时，再次执行conda list,发现tensorflow已经存在:
这时，再次执行import tensorflow as tf，不再报错：
还有两个命令或许对上面的问题有效:
conda create -n tensorflow-cpu tensorflow
activate tensorflow-cpu
(C:\Users\DELL\Anaconda3) C:\Users\DELL>conda create -n tensorflow-cpu
Fetching package metadata .............
Solving package specifications:
Package plan for installation in environment C:\Users\DELL\Anaconda3\envs\tensorflow-cpu:
Proceed ([y]/n)? y
#
# To activate this environment, use:
# > activate tensorflow-cpu
#
# To deactivate an active environment, use:
# > deactivate
#
# * for power-users using bash, you must source
#
(C:\Users\DELL\Anaconda3) C:\Users\DELL>conda activate tensorflow-cpu
CommandNotFoundError: 'activate'
(C:\Users\DELL\Anaconda3) C:\Users\DELL>activate tensorflow-cpu
(tensorflow-cpu) C:\Users\DELL>
(tensorflow-cpu) C:\Users\DELL>
这里 激活 开发环境是指，在 Anaconda 下我们可以有多个开发环境，比如如果你想对比一下 CPU 和 GPU 计算速度的差距，可以同时安装 2 个开发环境，然后根据需要切换到 CPU 开发环境，或者 GPU 开发环境，非常方便。如果不用 Anaconda 而是一个 Python 裸奔的话，要么使用 VirtualEnv，要么就只能反复安装卸载不同的开发环境了。
跑minist网络用例:
执行一段绘图程序：
安装ONNX
安装onnx需要依赖cmake，所以首先安装cmake
pip install cmake
pip install onnx==1.8.1
C:\Users\DELL>pip install cmake
Collecting cmake
Downloading https://files.pythonhosted.org/packages/d3/7e/8fc8632ef7de5c3443af88d58147a6ae32f3386a5b5ee5f473847e59e700/cmake-3.21.2-py2.py3-none-win_amd64.whl (37.3MB)
100% |████████████████████████████████| 37.3MB 34kB/s
Installing collected packages: cmake
Successfully installed cmake-3.21.2
You are using pip version 8.1.2, however version 21.2.4 is available.
You should consider upgrading via the 'python -m pip install --upgrade pip' command.
C:\Users\DELL>pip install onnx==1.8.1
Collecting onnx==1.8.1
Downloading https://files.pythonhosted.org/packages/ec/c7/422de621ed9a5a56c166b0456c585f463e04802a7d281067ee97334fc387/onnx-1.8.1-cp35-cp35m-win_amd64.whl (6.9MB)
100% |████████████████████████████████| 6.9MB 213kB/s
Collecting typing>=3.6.4 (from onnx==1.8.1)
Downloading https://files.pythonhosted.org/packages/f2/5d/865e17349564eb1772688d8afc5e3081a5964c640d64d1d2880ebaed002d/typing-3.10.0.0-py3-none-any.whl
Requirement already satisfied (use --upgrade to upgrade): typing-extensions>=3.6.2.1 in c:\users\dell\anaconda3\lib\site-packages (from onnx==1.8.1)
Requirement already satisfied (use --upgrade to upgrade): numpy>=1.16.6 in c:\users\dell\anaconda3\lib\site-packages (from onnx==1.8.1)
Requirement already satisfied (use --upgrade to upgrade): six in c:\users\dell\anaconda3\lib\site-packages (from onnx==1.8.1)
Requirement already satisfied (use --upgrade to upgrade): protobuf in c:\users\dell\anaconda3\lib\site-packages (from onnx==1.8.1)
Installing collected packages: typing, onnx
Successfully installed onnx-1.8.1 typing-3.10.0.0
You are using pip version 8.1.2, however version 21.2.4 is available.
You should consider upgrading via the 'python -m pip install --upgrade pip' command.
C:\Users\DELL>
测试发现，安装onnx 1.8.1版本才能成功，其它的版本都会失败.
安装剩余的软件包
C:\Users\DELL>pip install numpy scipy sklearn pandas pillow matplotlib keras -i https://pypi.tuna.tsinghua.edu.cn/simple
Requirement already satisfied (use --upgrade to upgrade): numpy in c:\users\dell\anaconda3\lib\site-packages
Requirement already satisfied (use --upgrade to upgrade): scipy in c:\users\dell\anaconda3\lib\site-packages
Collecting sklearn
Using cached https://pypi.tuna.tsinghua.edu.cn/packages/1e/7a/dbb3be0ce9bd5c8b7e3d87328e79063f8b263b2b1bfa4774cb1147bfcd3f/sklearn-0.0.tar.gz
Requirement already satisfied (use --upgrade to upgrade): pandas in c:\users\dell\anaconda3\lib\site-packages
Requirement already satisfied (use --upgrade to upgrade): pillow in c:\users\dell\anaconda3\lib\site-packages
Requirement already satisfied (use --upgrade to upgrade): matplotlib in c:\users\dell\anaconda3\lib\site-packages
Collecting keras
Using cached https://pypi.tuna.tsinghua.edu.cn/packages/5a/38/04d9b7fb53acdf861df2c4505fa96b06c779817a511e94b8882d284ba360/keras-2.6.0-py2.py3-none-any.whl
Requirement already satisfied (use --upgrade to upgrade): scikit-learn in c:\users\dell\anaconda3\lib\site-packages (from sklearn)
Requirement already satisfied (use --upgrade to upgrade): python-dateutil>=2 in c:\users\dell\anaconda3\lib\site-packages (from pandas)
Requirement already satisfied (use --upgrade to upgrade): pytz>=2011k in c:\users\dell\anaconda3\lib\site-packages (from pandas)
Requirement already satisfied (use --upgrade to upgrade): cycler in c:\users\dell\anaconda3\lib\site-packages (from matplotlib)
Requirement already satisfied (use --upgrade to upgrade): pyparsing!=2.0.4,>=1.5.6 in c:\users\dell\anaconda3\lib\site-packages (from matplotlib)
Requirement already satisfied (use --upgrade to upgrade): six>=1.5 in c:\users\dell\anaconda3\lib\site-packages (from python-dateutil>=2->pandas)
Building wheels for collected packages: sklearn
Running setup.py bdist_wheel for sklearn ... done
Stored in directory: C:\Users\DELL\AppData\Local\pip\Cache\wheels\6f\cd\11\f6acd1062135d70bc0a7066808561580d256b3149055cb33ad
Successfully built sklearn
Installing collected packages: sklearn, keras
Successfully installed keras-2.6.0 sklearn-0.0
You are using pip version 8.1.2, however version 21.2.4 is available.
You should consider upgrading via the 'python -m pip install --upgrade pip' command.
C:\Users\DELL>
抓取MINST数据集并输出
#coding:utf-8
from tensorflow.examples.tutorials.mnist import input_data
import numpy as np
import matplotlib.pyplot as plt
#%matplotlib inline
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True) #MNIST数据输入
X_train = mnist.train.images
y_train = mnist.train.labels
X_test = mnist.test.images
y_test = mnist.test.labels
# 输入图像大小是 28x28 大小
X_train = X_train.reshape([-1, 28, 28, 1])
X_test = X_test.reshape([-1, 28, 28, 1])
plt.imshow(X_train[0].reshape((28, 28)), cmap='gray')
plt.imshow(X_train[1].reshape((28, 28)), cmap='gray')
plt.imshow(X_train[2].reshape((28, 28)), cmap='gray')
print ('输入数据打shape:',mnist.train.images.shape)
print ('输入数据打shape:',mnist.test.images.shape)
print ('输入数据打shape:',mnist.validation.images.shape)
安装tf2onnx
C:\Users\DELL>pip install tf2onnx
Collecting tf2onnx
Downloading https://files.pythonhosted.org/packages/46/52/fa6a3af9f8ea0560d460d727096edc1453b11daeb226d1093aa93e27ebcc/tf2onnx-1.9.2-py3-none-any.whl (430kB)
100% |████████████████████████████████| 440kB 585kB/s
Collecting flatbuffers~=1.12 (from tf2onnx)
Downloading https://files.pythonhosted.org/packages/eb/26/712e578c5f14e26ae3314c39a1bdc4eb2ec2f4ddc89b708cf8e0a0d20423/flatbuffers-1.12-py2.py3-none-any.whl
Requirement already satisfied (use --upgrade to upgrade): requests in c:\users\dell\anaconda3\lib\site-packages (from tf2onnx)
Requirement already satisfied (use --upgrade to upgrade): numpy>=1.14.1 in c:\users\dell\anaconda3\lib\site-packages (from tf2onnx)
Requirement already satisfied (use --upgrade to upgrade): onnx>=1.4.1 in c:\users\dell\anaconda3\lib\site-packages (from tf2onnx)
Requirement already satisfied (use --upgrade to upgrade): six in c:\users\dell\anaconda3\lib\site-packages (from tf2onnx)
Requirement already satisfied (use --upgrade to upgrade): typing>=3.6.4 in c:\users\dell\anaconda3\lib\site-packages (from onnx>=1.4.1->tf2onnx)
Requirement already satisfied (use --upgrade to upgrade): protobuf in c:\users\dell\anaconda3\lib\site-packages (from onnx>=1.4.1->tf2onnx)
Requirement already satisfied (use --upgrade to upgrade): typing-extensions>=3.6.2.1 in c:\users\dell\anaconda3\lib\site-packages (from onnx>=1.4.1->tf2onnx)
Installing collected packages: flatbuffers, tf2onnx
Successfully installed flatbuffers-1.12 tf2onnx-1.9.2
You are using pip version 8.1.2, however version 21.2.4 is available.
You should consider upgrading via the 'python -m pip install --upgrade pip' command.
C:\Users\DELL>pip install asserts
Collecting asserts
Downloading https://files.pythonhosted.org/packages/15/ef/a02b2af8228be2a08ffd7e7630084e441030fbd3e6426483ddcdf905ac34/asserts-0.11.1-py2.py3-none-any.whl
Installing collected packages: asserts
Successfully installed asserts-0.11.1
You are using pip version 8.1.2, however version 21.2.4 is available.
You should consider upgrading via the 'python -m pip install --upgrade pip' command.
C:\Users\DELL>
安装keras
C:\Users\DELL>conda install keras
Fetching package metadata .............
Solving package specifications: .
Package plan for installation in environment C:\Users\DELL\Anaconda3:
The following NEW packages will be INSTALLED:
keras: 2.2.2-0
keras-applications: 1.0.4-py35_1
keras-base: 2.2.2-py35_0
keras-preprocessing: 1.0.2-py35_1
patch: 2.5.9-1
The following packages will be UPDATED:
anaconda: 4.1.1-np111py35_0 --> custom-py35_0
conda: 4.3.30-py35hec795fb_0 --> 4.5.11-py35_0
conda-env: 2.6.0-0 --> 2.6.0-1
pycosat: 0.6.1-py35_1 --> 0.6.3-py35hfa6e2cd_0
Proceed ([y]/n)?
keras-applicat 100% |###############################| Time: 0:00:00 292.58 kB/s
keras-preproce 100% |###############################| Time: 0:00:00 959.92 kB/s
keras-base-2.2 100% |###############################| Time: 0:00:00 1.03 MB/s
keras-2.2.2-0. 100% |###############################| Time: 0:00:00 3.03 MB/s
C:\Users\DELL>conda install keras
结束！
