TensorFlow There are two versions of cpu and gpu. The gpu version requires support from NVIDIA CUDA and cuDNN, and the cpu version is not required. Here I installed the gpu version,
ready to work (four steps in total):
Step1: Confirm that your graphics card supports CUDA, and the computing power is in line with the amount of computation of the code you want to run (generally in the experiment of the paper) In the description, please check it yourself.
Display adapter (graphics card) view: Enter "start devmgmt.msc" in cmd.
我的卡卡GPU: NVIDA GeForce GTX 750
@GPU Computing Capability Query: Query page
My graphics card computing power: 5.0
Step2: Confirm that you have installed VS2015 or 2013 or 2010 I used 2013.
Step3: python 3.5 (confirm that your Python version is 3.5 64-bit, it is not recommended to use a higher Python version. Python recommends installing Anaconda, because this integrates a lot of libraries necessary for scientific computing, can avoid many dependencies The problem is also very convenient to install.)
Anaconda Official website Recent version: Download page
Anaconda All versions of the official website: Download page
Version selection: Anaconda3-4.2.0-Windows-x86_64.exe (The The version of Anaconda comes with Python version 3.5)
Step4: pip 9.0.1 (confirm that the pip version is at least 8.1, use pip -V to view the current pip version, upgrade with "python –m3 pip install -U pip" Pip .)
After finishing the above preparations, you can take a break and then check out the second blog for follow-up work.