table of Contents
One, install Anaconda
1. Download Anaconda
Download URL, click to download
2. Install Anaconda
Choose the installation directory by yourself. Later, as the installed modules will take up more space,
you can check the automatic configuration environment or uncheck the manual configuration environment.
Choose to install VScode according to your personal needs.
3. Add anaconda environment
After the addition is successful,
Win+R starts cmd, enter the following command to view your installed version
conda -V
Two, install TensorFlow-GPU, Keras
1. Create a TensorFlow environment
Win+R start cmd, enter the following command in the command prompt:
1 | conda create –n tensorflow-gpu python=3.7
2 | activate tensorflow-gpu
3 | pip install tensorflow-gpu==1.13.2
4 | pip install keras==2.1.5
If the above command download is too slow, you can use Douban source to speed up the download:
3 | pip install -i https://pypi.doubanio.com/simple/ tensorflow-gpu==1.13.2
4 | pip install -i https://pypi.doubanio.com/simple/ keras==2.1.5
Three, install CUDA and cuDnn
1. Check your computer GPU model
This computer -> right click management
Click the following URL to view your GPU computing power, you can download the corresponding Cuda and cudnn according to the graphics card computing power to
view your own graphics card model
2. Download CUDA10.0 and cuDnn7.4.1.5
CUDA10.0 download official website
cuDnn need to manually find 7.4.1.5 install
cuDnn download official website The
above method download is too slow, you can use the following link to download
Baidu network disk link: https://pan.baidu.com/s/1Prck69Mei9DyJxjS0nvTCg
extraction code: p281
3. Install CUDA10.0
4. Install cuDnn7.4.1.5
After installing CUDA, open the following location
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0
and unzip all the contents of Cudnn and copy them to the above directory.
Four, test
import tensorflow as tf
tf.test.gpu_device_name()
Prompt the following information to install successfully