TensorFlow environment configuration and installation (win10+GeForce GTX1060+CUDA 9.0+cuDNN7.3+tensorflow-gpu 1.12.0+python3.5.5)

Record the installation of win10+GeForce GTX1060+CUDA 9.0+cuDNN7.3+tensorflow-gpu 1.12.0+python3.5.5

Pycharm, Anaconda and VS2013 have been installed before, so the installation record will start from then on

The overall steps are roughly as follows:

1. Confirm whether your computer graphics card model supports CUDA (there is a pit here)

 

 

 

There is a pit here! Don't worry about CUDA9.2.148 displayed in the NVIDIA control panel component.

The CUDA you downloaded does not necessarily need to match, especially CUDA9.2. It is best to use CUDA9.0. I just fell into this pit.

2. Download CUDA and cuDNN, pay attention to the corresponding version

①Check the version matching:

https://www.tensorflow.org/install/source_windows

Guess you like

Origin blog.csdn.net/ch206265/article/details/106897338