I have been lying in a pit all night. Now I am writing an article to sum up my successful experience

I hope you can give a thumbs-up to this article

First take a look at your configuration:

1) System Win10 64bit

                             

2) Video card 1050TI

                             

First of all go to the www.tensorflow.org/install/gpu Tensorflow website above to view their own version requirements

It’s clearly written up here

Cuda will be released in version 10.0 and above

The drive is above 410.x

Cudnn is 7.4.1 and above



So the first thing is to look at your graphics card if I could use CUDAhttps://developer.nvidia.com/cuda-gpus




Here to update your driver www.nvidia.com/Download/in…

It says that the drive we need is above 410.x




Pick your device and look it up and update it

(I’m going to draw a circle here for you to circle. This is very important. It’s very important to do it.

Above must look carefully carefully otherwise easy to step on the pit behind

Above must look carefully carefully otherwise easy to step on the pit behind

Above must look carefully carefully otherwise easy to step on the pit behind

After the preparations are complete, the environment installation begins

Install Anaconda & Tensorflow-GPU

Anaconda download https://www.anaconda.com/download/

According to their own needs



Once you’ve installed it




Install CUDA and cuDNN

Here to find the corresponding CUDA version download remember to the local version developer.nvidia.com/cuda-downlo…


Find Conda 10.0 here



download




Be careful not to choose the lite version when installing



Download it and install it





Configure environment variables. If not, configure them yourself


Check whether the installation was successful. Use this to see if the following screen appears

nvcc -V



Start installing cuDNN7.3

Here you can download cuDNN, to apply for account developer.nvidia.com/rdp/cudnn-d…




Download the files you can see after the escort, and copy these folders to

(C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0)



Put these files in CUDA and rename them CUDNN




Configure variable environment



Install Tensorflow in CMD

pip install  --upgrade tensorflow-gpu



Test success



The GPU test succeeds



Here I want to thank Teacher Longlong

A full set of netease cloud classroom video link: study.163.com/course/intr…