第一:Comparity of tensorflow GPU version and CPU version:
CPU The versatility of
In order to handle a variety of different data types, at the same time, logical decisions will introduce a large number of branch jumps and interrupts. These make the internal structure of the CPU extremely complicated.
While GPUThe face is a highly unified, non-dependent large-scale data and a pure computing environment that does not need to be interrupted
So: CPU and GPU Presenting a very different architecture.
第二: How to install the related package
1: Install the python environment. Here is just a glimpse of how you can go online to Baidu. It is recommended to use Anaconda installation. Then there is a default operation.
2: Find yourself and the computer and the matching version. I am the God of War i7+Z7-kp7sc+window10+CUDA9.0+CUNN9.0+tensorflow-GPU. You must choose the right version, so you can avoid it. The error behind.
3: Installation of CUDA9.0
CUDNN library download:
After downloading, do not install, directly copy the files in the corresponding folder to the corresponding folder of cuda.
After the installation is complete, test whether the installation is successful. Enter nvcc -V
in cmd and open cmd and then enter the following code:
import tensorflow as tf
hello = tf.constant("Hello!TensorFlow")
sess = tf.Session()
Select your own installed python The path is OK.
Finally, you can run it. You can find your own test code online.