1. Download the NVIDIA driver
Recently learning artificial intelligence Paddle and Tensorflow, you need to install NVIDIA series graphics card drivers. The download and installation methods of this set of drivers still have some characteristics, so record them for future use (note the download of the following software, you must register before downloading, my The package version installed on the machine is as follows:)
419.17-desktop-win10-64bit-international-whql.exe
cuda_10.1.168_425.25_win10.exe
cudnn-10.1-windows10-x64-v7.6.5.32.zip
1. Baidu search input: NVIDIA official website, click: NVIDIA Leads Artificial Intelligence Computing - NVIDIA;
official website: https://www.nvidia.cn/
Drivers official website: https://www.nvidia.cn/geforce/drivers/
2. In the opened NVIDIA window, we click: Driver;
3. The following page opens the NVIDIA driver download page (NVIDIA series graphics card), here we need to enter the product type, product series, product, operating system and language , Download type selection.
4. In the first step, we choose the product type, and take GTX 1060 as an example to illustrate the operation steps, such as
NVIDIA GeForce GTX 1060 graphics card belongs to the GeForce series, so choose GeForce for the product type;
5. The second step is to select the product series. GTX 1060 belongs to the 10-series graphics card. We choose GeForce10 Series (note: for desktop computers, choose GeForce10 Series; if it is a notebook computer, choose the series with Notebooks in the suffix, that is, the series with M).
6. The third step is to select the product family, that is, the graphics card model: GTX 1060;
7. The fourth step is to select the operating system. This machine is installed with the Windows 10-64-bit professional version system, so choose: Windows 10 64-bit;
(If you don’t know what type your computer is, you can right-click the property column on the desktop icon, and then open the system information to see the system type)
8. The fifth step is to choose the language, the simplified Chinese I choose here: Chinese (Simplified);
9. So far, the product type, product series, product family, operating system and language of the graphics card have been selected, click: start searching;
10. After clicking Search, the latest version of the driver that needs the graphics card will appear, we click: Download;
11. Then ❶ click: agree and start downloading; ❷ click: download;
12. After the download is complete, we find the downloaded driver file in the local disk and double-click it;
13. We click: OK to start decompressing the compressed file, and wait for the decompression to complete;
14. After the decompression is completed, the NVIDIA installer window will open automatically. If we have selected the correct "product type, product series, product family, operating system and language" above, the installation of the GTX 1060 graphics card driver will be completed automatically.
15. If the selected "product type, product series, product family, operating system and language" is different from the local graphics card, the graphics card driver cannot be installed, and an error message will appear:
NVIDIA installer cannot continue
This graphics driver was unable to find compatible graphics hardware.
The above is: the download and installation method of the NVIDIA series graphics card driver, for the reference of friends who use the NVIDIA series graphics card.
2. Differences in the use of GEFORCE® driver download versions
GAME READY DRIVER AND STUDIO DRIVER
If you're a gamer who wants to be supported the first time new games, patches, and game add-on downloadable content (DLC) are released on launch day, choose Game Ready Drivers.
If you're a content creator who prioritizes stability and quality in creative workflows such as video editing, animation, photography, graphic design, and live streaming, choose the Studio driver. Still confused? Don't worry, both drivers support great games and creative applications.
3. Download the nvidia driver version suitable for your computer configuration
- Download the nvidia driver version suitable for your computer configuration. remember! ! ! The download URLs of the new and old versions are as follows:
4. The old NVIDIA GeForce 419.17 WHQL graphics card driver: attached driver download
NVIDIA GeForce 419.71 WHQL official download: Windows 10 64bit:
Note: Download CUDA and CUDNN according to the driver, otherwise it cannot be used after installation.
https://cn.download.nvidia.com/Windows/419.17/419.17-desktop-win10-64bit-international-whql.exe
5. Check the CUDA version number in Windows system (from the control panel)
5. Check the CUDA version number in Windows system (from the control panel)
Information to be confirmed2. You need to confirm that your Windows 7/8/10 is a 64-bit operating system*
3. You need to have an nVidia graphics card that supports CUDA and install CUDA 10.1 correctly
4. Note: Windows only supports CUDA 8.0/9.0/ 10.1 single-card mode; does not support CUDA 9.1/9.2/10.1
5. Need to use cuDNN 7.6.5 6.
Windows does not support NCCL for now
7. Confirm that the Anaconda/miniconda environment is installed correctly
8. Enter the command line terminal of Anaconda and add Paddle conda Tsinghua source
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/Paddle/
conda config --set show_channel_urls yes
Installation process:
First, create the environment:
first check the python information you installed, if you need to create a new conda virtual environment for Paddle (you can choose the environment name after –name): see the figure below
conda create --name paddle python=3.7
In the middle there is a query Procced([y]/n)?
Second, activate the created paddle environment: activate paddle
Confirm that your conda virtual environment and the Python that needs to be installed with PaddlePaddle are where you expected, because your computer may have multiple Pythons
Enter the command line terminal of Anaconda, and enter the following command to confirm the location of Python: where python
Third, check whether your machine is suitable for installing the propeller
and confirm that Python and pip are 64 bits, and the processor architecture is x86_64 (or called x64, Intel 64 , AMD64) architecture, currently PaddlePaddle does not support arm64 architecture.
The first line below outputs "64bit", and the second line outputs "x86_64", "x64" or "AMD64":
Python -c "import platform;print(platform.architecture()[0]);print(platform.machine())"
Fourth, refer to the article Paddlepaddle Quick Installation for installation. The website is as follows: https://www.paddlepaddle.org.cn/install/quick?
My machine uses CUDA 10.1, cuDNN 7.6+, and the command to install the GPU version is :
conda install paddlepaddle-gpu cudatoolkit=10.1 -c paddle
Fifth, check whether the installation is successful
Check the version of Paddle
Enter in Anaconda Prompt: python -c "import paddle; print(paddle. version )" to view the version number of Paddle:
Check whether Paddle is installed successfully
Use python3 to enter the python interpreter, enter import paddle.fluid, and then enter paddle.fluid.install_check.run_check().
If Your Paddle Fluid is installed succesfully! appears, it means that you have successfully installed it (as shown in the figure below).
So far, the artificial intelligence paddle (Paddle) has been successfully installed.