Installed version:
python version: 3.6.4
release or time: 2017-12-19
download address: https://www.python.org/ftp/python/3.6.4/python-3.6.4-amd64.exe
Anaconda version 3.5.1
Release version or time: latest
Download address: https://repo.continuum.io/archive/Anaconda3-5.1.0-Windows-x86_64.exe
TensorFlow 1.7.0
1. Prepare the environment
TensorFlow supports the installation of python 3.5.1 and above.
Win7 first upgrade to the service pack version:
System upgrade method: https://jingyan.baidu.com/article/ff411625b7fd1812e48237f0.html
Install: api-ms-win-runtime-1-1-0.dll
download link:
Just run the install
2. Install python
Run the installation package (win10 system):
Page 1: Select Add Environment Variables and Optional Installation Configuration to specify the installation path.
The second page: the default can be Next
Page 3: The installation options are default, or you can add them by yourself to modify the installation path. Install to install
Page 4: After the installation is complete, click close.
To verify whether python is installed successfully, enter cmd in win+r, and enter python -V in the console to print the version, and the installation is successful.
Step 3: Install Anaconda
Run the installation package, the first page directly next.
Page 2, I Agree
The third page, default or all users, Next
The fourth page, configure the installation path
On the fifth page, check to add the replacement variable and use the installed python, and a prompt window appears to confirm. Install installation, the installation time is long.
On the sixth page, directly Next after completion.
On the seventh page, you are prompted to install the vscode editor, click skip.
On the eighth page, cancel the learning introduction and finish directly
Verify that the installation is successful and that there is no shortcut added to the desktop. Find it directly from all software, run Anaconda Navigator, and the startup is successful.
Step 4: Install TensorFlow
Start cmd as win10 administrator. Run, pip install TensorFlow or (python -m pip install TensorFlow), the cpu version is installed by default, if a step in the installation process fails, re-run the installation command.
cpu version: pip install TensorFlow
gpu version: pip install TensorFlow-gpu
Installation is complete, verify.
First python, in import tensorflow as tf, and finally tf.__version__, note that there are two _ on the left and right. Print version installed successfully
Run the tensorflow upgrade method: pip install --upgrade tensorflow
Specify the version: pip install tensorflow==1.4.0
Bug fixes:
If the following prompt appears when importing tensorflow as tf, this is not an error, but a reminder.
The solution is to reinstall and lower the numpy package version, pip install numpy==1.13.3, and the installation is complete.